pmid stringlengths 8 8 | pmcid stringlengths 8 11 ⌀ | source stringclasses 2
values | rank int64 1 9.78k | sections unknown | tokens int64 3 46.7k |
|---|---|---|---|---|---|
35492994 | PMC9051554 | pmc | 360 | {
"abstract": "Fascinating and challenging, the development of repairable materials with long-lasting, sustainable and high-performance properties is a key-parameter to provide new advanced materials. To date, the concept of self-healing includes capsule-based healing systems, vascular healing systems, and intrinsic healing systems. Polyurethanes have emerged as a promising class of polymeric materials in this context due to their ease of synthesis and their outstanding properties. This review thereby focuses on the current research and developments in intrinsic self-healing polyurethanes and related composites. The chronological development of such advanced materials as well as the different strategies employed to confer living-like healing properties are discussed. Particular attention will be paid on chemical reactions utilized for self-healing purposes. Potential applications, challenges and future prospects in self-healing polyurethane fields are also provided.",
"conclusion": "VIII. Conclusions and challenges PU materials are highly used in our daily life while finding approaches to extend their lifetime which dictate their use in future applications, requiring highly advanced materials. Herein, self-healing PUs provide opportunities for designing novel materials with high performance, lifetime and reliability. Although the number of publications reporting self-healing PUs has grown exponentially the past decade, so far, their presence on the market is still limited. Non-exhaustive list of drawbacks limiting their commercialization are the following: (i) no practical up-scaling synthesis: multi-step synthesis, low yield, production price, use of solvent; (ii) limited healing recovery: in terms of reactive functions ( e.g. damage-induced self-healing PUs), cycles ( e.g. hindered urea) or mobility ( e.g. self-healing PU composites); (iii) high energy input required for self-healing: thermal or UV irradiations tend to be replaced by softer stimuli as humidity or visible light; (iv) healing limited to microscopic damages: limited to micrometer sized scratch or requiring physical reattachment of cut material; (v) non-functional: only applicable to commodity applications (unless addition of functional nanoparticles is considered) and (vi) no detection of damage in PUs (as a prevention): up to now, only mechanochromic PUs provide damage probing feature. PUs addressing these issues will be considered as the perfect self-healing PUs. To achieve such non-utopist challenges, combination of the different concepts described in the present review should open future directions in this challenging but fascinating quest. Developing chemical or physical approaches separately will always be less efficient than their combination in a single material. Finally, and equally important, the development of spectroscopic techniques and knowledges allowing to highlight healing mechanism in PUs still remain a key-factor for designing efficient systems."
} | 738 |
31903216 | PMC6936270 | pmc | 361 | {
"abstract": "To effectively forage in natural environments, organisms must adapt to changes in the quality and yield of food sources across multiple timescales. Individuals foraging in groups act based on both their private observations and the opinions of their neighbours. How do these information sources interact in changing environments? We address this problem in the context of honeybee colonies whose inhibitory social interactions promote adaptivity and consensus needed for effective foraging. Individual and social interactions within a mathematical model of collective decisions shape the nutrition yield of a group foraging from feeders with temporally switching quality. Social interactions improve foraging from a single feeder if temporal switching is fast or feeder quality is low. When the colony chooses from multiple feeders, the most beneficial form of social interaction is direct switching, whereby bees flip the opinion of nest-mates foraging at lower-yielding feeders. Model linearization shows that effective social interactions increase the fraction of the colony at the correct feeder (consensus) and the rate at which bees reach that feeder (adaptivity). Our mathematical framework allows us to compare a suite of social inhibition mechanisms, suggesting experimental protocols for revealing effective colony foraging strategies in dynamic environments.",
"introduction": "1. Introduction Social insects forage in groups, scouting food sources and sharing information with their neighbours [ 1 – 3 ]. The emergent global perspective of animal collectives helps them adapt to dynamic and competitive environments in which food sources’ quality and location can vary [ 4 ]. Importantly, decisions made by groups involve nonlinear interactions between individuals, temporally integrating information received from neighbours [ 5 ]. For example, honeybees waggle dance 1 to inform nest-mates of profitable nectar sources [ 6 , 7 ], and use stop signalling 2 to dissuade them from perilous food sources [ 8 ] or curb recruitment to overexploited sources [ 9 ]. While waggle dancing rouses bees from indecision, stop signalling prevents decision deadlock and builds consensus when two choices are of similar quality [ 10 ]. Thus, both positive and negative feedback interactions within the group are important for regulating collective decisions and foraging [ 11 , 12 ]. Honeybee colonies live in dynamic environments, in which the best adjacent nest or foraging sites can vary across time [ 13 , 14 ]. Bees adapt to change by abandoning less-profitable nectar sources for those with higher yields [ 15 , 16 ], and by modifying the number of foragers [ 17 , 18 ]. Prior studies focused on how waggle dance recruitment or the division of individual bee roles shape colony adaptivity [ 19 , 20 ]. Inhibitory social interactions, whereby bees stop each other from foraging, have been mostly overlooked as a communication mechanism for facilitating collective adaptation to change [ 21 , 22 ]. We propose that inhibitory social interactions are important for foraging groups to adapt to change in a fluid world. To study how social inhibition shapes foraging yields, we focus on a task in which the nectar quality of feeders is switched periodically. Related situations probably occur in nature due to the dynamics of competitor and predator prevalence, crowding by nest-mates, and weather fluctuations [ 23 – 25 ]. Precisely periodic dynamics do not occur naturally but can be generated in controlled experiments [ 16 , 20 ]. There are important distinctions between the goals of colonies in foraging as opposed to those searching for a new home site. Once a colony establishes a permanent nest site, this is the starting and ending point for each food foraging excursion. The colony does not need to reach consensus to obtain nutrition from foraging, since food is brought to the nest regardless of how many foraging sites the group is split between [ 25 ]. By contrast, when a honeybee swarm looks for a nest, it must reach consensus for all bees and the queen to fly to the selected site. If not, their transition to a permanent nest site will be delayed, or the swarm might split. Bees use stop signals to obtain this needed consensus when house-hunting, especially when two potential sites are of similar quality [ 26 ]. Consensus is not essential when foraging for food, but, as we will show, increasing the fraction of the colony at the best foraging site increases foraging yields. Foraging colonies appear to be able to adapt to change. In prior studies [ 15 , 16 , 20 ], colony foraging targets shifted in response to food quality switches, suggesting bee collectives can detect such changes. Uncommitted inspector bees can lead bees away from feeders whose nectar quality has dropped [ 20 ], and recruitment via waggle dancing appears to be unimportant for effective foraging in changing environments (see also [ 27 ]). Here, we also find recruitment can be detrimental, but social inhibition can rapidly pull bees from low- to high-yielding feeders. This, paired with ‘abandonment’ whereby bees spontaneously stop foraging, facilitates temporal discounting of prior evidence. By contrast, strong positive feedback via recruitment causes bees to congregate at feeders even after food quality has dropped, biasing a colony’s behaviour based on past states of the world. We quantify the contribution of these positive and negative feedback interactions within a mathematical model of a foraging colony. Our study focuses on four potential inhibitory social interactions—discriminate and indiscriminate stop signalling [ 8 , 26 ], direct switching [ 28 , 29 ] and self-inhibition—by which foraging bees alter the behaviour of other foraging bees. Self-inhibition has not been reported in honeybee foraging experiments, but we consider its effects as a potential social inhibitory mechanism, claiming it could be observable in behavioural assays for which it is advantageous (e.g. single switching feeder). Strategies are compared by measuring the rate of foraging yield over the timescale of feeder quality switches. When bees have a single feeder, social interactions are less important unless temporal switching is fast and food quality is low, but in the case of two feeders the performance of different forms of social interactions is clearly delineated. Direct switching, by which a bee converts another forager to their own preference, is the most effective means for a colony to adapt to feeder quality changes. Also, foraging yields are most sensitive to changes in group-wide interactions in rapidly changing environments with lower food quality. Model linearizations allow us to calculate a correspondence between social interaction parameters and the consensus (steady-state fraction of bees at the high-yielding feeder) and adaptivity (the rate of switching from low- to high-yielding feeders). This provides a clear means of determining the impact of social interactions on a colony’s foraging efficacy.",
"discussion": "3. Discussion Foraging animals constantly encounter temporal and spatial changes in their food supply [ 38 ]. The success of foraging animal groups thus depends on how efficiently they communicate and act upon environmental changes [ 39 ]. Our bee colony model analysis pinpoints specific social inhibition mechanisms that facilitate adaptation to changes in food availability and consolidate consensus at better feeders. If bees interact by direct switching, they can immediately update their foraging preference without requiring recruitment, keeping foragers active following environmental changes. Recruitment is less important to the foraging success of a colony in dynamic conditions; bees can initiate commitment via their own scouting behaviour. Individuals should balance their social and private information in an environment-dependent way to decide and forage most efficiently [ 40 , 41 ]. Efficient group decision-making combines individual private evidence accumulation and information sharing across the cohort [ 42 ]. However, in groups where social influence is strong, opinions generated from weak and potentially misleading private evidence can cascade through the collective, resulting in rapid but lower value decisions [ 10 , 43 , 44 ]. Our analysis makes these notions quantitatively concrete by associating the accuracy of the colony’s foraging decisions with the consensus fraction at the better feeder and the speed of decisions with the adaptivity or rate the colony approaches steady-state consensus ( figure 6 ). The best foraging strategies balance these colony-level measures of decision efficiency. Social insects do appear to balance the speed and accuracy of decision to increase their rate of food intake [ 45 , 46 ], and collective tuning is probably influenced by individuals varying their response to social information. We find that social recruitment can speed along initial foraging decisions, but it can limit adaptation to change. This is consistent with experimental studies that show a reduction in positive feedback can help collectives steer away from lower value decisions. For example, challenging environmental conditions (e.g. volatile and low food quality) are best managed by honeybee colonies whose individuals do not wait for recruitment but rely on their own individual scouting [ 27 ]. Ants encountering crowded environments tend to deposit less pheromone to keep their nest-mates from less-efficient foraging paths [ 47 ]. These experimental findings suggest social insects adapt to changing environmental conditions by limiting communication that promotes positive feedback [ 48 ]. Foragers must then be proactive in dynamic environments, since they cannot afford to wait for new social information [ 49 ]. Thus, the advantages of social learning depend strongly on environmental conditions [ 50 ]. In concert with a reduction in recruitment, we predict that honeybee colonies foraging in volatile environments will benefit from strengthening inhibitory mechanisms at the individual and group level. Bees enacting social inhibition dissuade their nest-mates from foraging at opposing feeders. We found the most efficient form of social inhibition is direct switching whereby bees flip the opinion of committed bees to their own opinion. So do honeybees use this mechanism in dynamic environments? Observations of bee colonies making nest site decisions show scouts directly switching their dance allegiance [ 28 , 51 ], but these events seem to be relatively rare in the static environments of typical nest site selection experiments [ 31 ]. Other forms of social inhibitory signals, especially stop signalling, appear to be used to promote consensus in nest decisions [ 10 , 26 ] and escaping predators while foraging [ 8 , 24 ]. Thus, the role and prevalence of social inhibition as a means for foraging adaptively in dynamic environments warrants further investigation. Most work studying the effects of social inhibition on honeybee colony decisions has focused on swarms choosing a place to build a nest from sites whose qualities are fixed in time [ 26 , 28 ]. Social inhibition is needed in this context to promote consensus, generating a consistent opinion across the swarm and preventing the deadlock and group splitting. On the other hand, it is not immediately obvious that social inhibition would improve foraging if it primarily increases consensus, since colonies can obtain and store food even when foragers are split between multiple feeders, though stop signals can reduce crowding [ 25 ]. Nonetheless, we found that when the colony can rapidly switch opinion so nearly all bees agree to forage from the most profitable feeder, this does increase the nutrition yield of the colony overall. However, consensus is only advantageous in dynamic environments if it does not come at a cost to adaptivity: the opinion around which consensus is built should change with the environment. Our simple parametrized model, developed from previously validated house-hunting models [ 26 , 28 , 29 , 32 ], is amenable to analysis and could be validated with time-series measurements from dynamic foraging experiments. Past experimental work focused on shorter time windows in which only a few switches in feeder quality occurred [ 16 , 20 ], which may account for the relatively slow adaptation of the colonies to environmental changes. We predict bees will slowly tune their social learning strategies to suit the volatility of the environment, but this could require several switch observations. Foraging tasks conducted within a laboratory could be controlled to track bee interactions over long time periods using newly developed automated monitoring techniques [ 52 ]. Our study also identifies key regions in parameter space in which different foraging strategies diverge in their performance, suggesting that placing colonies in rapid environments with relatively low food supplies will help distinguish which social communication mechanisms are being used. Previous computational modelling studies of honeybee collective decisions primarily focused on groups solving house-hunting problems in static environments [ 26 , 29 ], emphasizing how social interactions shape the speed at which consensus is obtained within a collective. However, less attention has been paid to how such collectives must adapt to change, and how social communication determines group adaptivity. Some previous work has discussed the importance of uncommitted inspector bees in affording group adaptivity [ 20 ], but our work is the first to systematically compare how different forms of social communication [ 8 , 26 , 28 , 29 ] shape group adaptivity. Social communication by which one bee can switch the foraging preference of another appear to be most effective in providing groups with the ability to both build consensus and adaptive to change. Our findings are fairly robust to considerations of interaction heterogeneity within the colony (see appendix C(e) and figure 13 ). A colony whose bees have individualized rates of recruitment and abandonment exhibits slight decreases in consensus and adaptivity, but qualitatively the group still remained responsive to change.\n There are a number of possible extensions of our work here. For instance, one could consider separate populations of scouts and foraging recruits as in some previous modelling studies [ 1 , 53 ]. Our analysis assumes bees can fluidly transition between scouting and foraging behaviour, as documented in several previous studies [ 54 , 55 ]. Overall, a strict and unchanging division of labour within the hive provides an incomplete description of colony organization. For instance, bees may switch to foraging when the environment demands it [ 56 ] or when socially signalled to do so [ 57 ], and thus a strict caste divide between scouts and recruits may be unrealistic [ 55 ]. Honeybees' roles appear to be strongly determined by the changing requirements of the colony, such as the influx or availability of nectar, rather than strictly due to some genetic predisposition [ 54 , 58 ]. Bees that scout and forage tend to be in the same life cycle phase, and as such are more amenable to temporal caste switching [ 59 ]. Such flexibility may even be a rule rather than exception to colony labour organization [ 60 ]. We could also have considered the effects of crowding at feeders [ 25 ], so nutrition yields would scale sublinearly with the fraction of bees at the feeder, possibly reordering the efficacy of social inhibition strategies. Collective decision strategies and outcomes can depend on group size [ 61 , 62 ], though decision accuracy does not necessarily increase with group size [ 63 ]. We approximated bee colony dynamics using a population level model, which is the deterministic mean-field limit of a stochastic agent-based model [ 26 ]. Finite group size considerations would result in stochastic models, in which the same conditions can generate different colony dynamics [ 10 ]. The qualitative predictions of our mean-field model did not change dramatically when considering stochastic finite-size effects (see appendix C(f) and figure 14 ). However discriminate stop-signalling colonies exhibit bistable decision dynamics ( figure 3 d , f ), so the stochasticity in the finite-sized model could allow colonies to break free from less-profitable feeders, similar to noise-driven escapes of particles in double potential well models [ 64 ]. Fluctuation-induced switching may thus provide an additional mechanism for flexible foraging [ 65 , 66 ], and would be an interesting extension of our present modelling work. Moreover, besides their importance to understanding decisions of biological collectives, our mathematical modelling results could inform efficient strategies for organizing distributed decision-making in inanimate groups, like swarm robotics and artificial communication networks [ 67 , 68 ]."
} | 4,242 |
39908388 | PMC11797559 | pmc | 362 | {
"abstract": "Biological neurons use diverse temporal expressions of spikes to achieve efficient communication and modulation of neural activities. Nonetheless, existing neuromorphic computing systems mainly use simplified neuron models with limited spiking behaviors due to high cost of emulating these biological spike patterns. Here, we propose a compact reconfigurable neuron design using the intrinsic dynamics of a NbO 2 -based spiking unit and excellent tunability in an electrochemical memory (ECRAM) to emulate the fast-slow dynamics in a bio-plausible neuron. The resistance of the ECRAM was effective in tuning the temporal dynamics of the membrane potential, contributing to flexible reconfiguration of various bio-plausible firing modes, such as phasic and burst spiking, and exhibiting adaptive spiking behaviors in changing environment. We used the bio-plausible neuron model to build spiking neural networks with bursting neurons and demonstrated improved classification accuracies over simplified models, showing great promises for use in more bio-plausible neuromorphic computing systems.",
"introduction": "INTRODUCTION Neuromorphic computing aims at emulating neuronal and synaptic computations of a biological brain to improve the cognitive capabilities and processing efficiencies of an intelligent system ( 1 – 3 ). Nonetheless, comprehensive and faithful simulation of brain functions is challenging because of large diversity of brain organizations differentiated by neuron types ( 4 ), connections ( 5 ), and functionalities ( 6 ). Several bio-plausible neuron models, such as Hodgkin-Huxley (H-H) ( 7 ), Fitzhugh-Nagumo ( 8 ), Chay ( 9 ), Hindmarsh-Rose ( 10 ), and Izhikevich ( 11 ) were proposed to characterize the electrophysical spiking process inside a biological neuron. By tuning the parameters within these models, various spiking patterns such as fast spiking, phasic spiking, and bursting ( 12 ) can be obtained and applied to spatial-temporal processing ( 13 ), rhythmic recurrent activity ( 14 ), and sensory information encoding ( 15 ). However, a bio-plausible neuron could have more than three orders of computing complexity ( 16 , 17 ), making it too costly to be used in a practical spiking neural network (SNN). On the other hand, simplified neuron models such as Leaky-Integrate-and-Fire (LIF) ( 18 ) were proposed and widely adopted in both neuromorphic hardware designs ( 19 , 20 ) and algorithms ( 21 , 22 ). Nonetheless, these models can only emulate the timing of the spike signals, whereas the rich information expressed by a biological spike pattern is lost in model simplification, therefore limiting neuromorphic systems to reach the desired cognitive performance ( 23 ). Recently, emerging nanoelectronic devices such as memristors ( 24 – 26 ) have been used to implement neuronal and synaptic functions for efficient neuromorphic computing ( 27 – 31 ). The complex computations needed to emulate biological behaviors are offloaded to hardware circuits in which ionic motions inside these emerging devices naturally match the intrinsic dynamics of the biological cells. As a result, bio-plausible neuron models such as H-H neuron ( 32 – 34 ) could be efficiently implemented using emerging devices with the addition of a few circuit elements, producing complex firing patterns with much smaller footprints compared to a complementary metal-oxide semiconductor circuit ( 35 ). The compact neuron designs have been used in human motion recognition, robot obstacle avoidance, and human-machine interface applications and demonstrated low latency and energy efficient computing capabilities ( 36 – 38 ). In light of their rich expressions of spike firing patterns, changing firing modes in these neuron circuits can only be achieved through change of input amplitudes ( 30 , 39 , 40 ) or change of circuit elements ( 32 , 33 , 37 ). However, the association of a firing mode to a particular input amplitude would limit its application in SNNs since neuron inputs depend on diverse upstream spike signals and cannot be fixed for neuron configurations. Meanwhile, modulation of firing modes may be achieved by tweaking multiple circuit elements and control signals, which may be possible by element multiplexing but would add notable cost to implementation, thus offsetting the scalability provided by using emerging devices. As a result, although these neuron designs have been used in different applications, using and actively changing the neuron behaviors in hardware is still limited. Developing a bio-plausible neuron circuit that supports reconfigurable functions is still highly desired for a neuromorphic chip design ( Fig. 1A ). Fig. 1. Reconfigurable spiking neuron for bio-plausible neuromorphic computing. ( A ) Schematic diagram of a reconfigurable chip supporting various spiking modes. ( B ) Design of the bio-plausible neuron circuit emulating fast-slow neuronal dynamics. ( C ) Core components of the neuron circuit including a nonvolatile ECRAM memory device and a volatile NbO 2 device. Scale bar, (optical micrographs) 40 μm. ( D ) Reconfiguration between four spiking modes modulated by different ECRAM resistances. Constant input voltage biases of 4 V were used in all cases. In this work, we reported a bio-plausible reconfigurable spiking neuron composed of a compact two-stage circuit that emulated the fast-slow dynamics in biological neurons ( 41 ). Reconfiguration of the neuron functions was achieved by heterogenous integration of NbO 2 -based neuronal spiking units ( 42 ) and an electrochemical memory (ECRAM) unit ( 43 ). Changes in ECRAM resistance were used as a slow variable of the neuron circuit, which demonstrated highly effective modulation of the spiking dynamics, contributing to reconfiguration between four spiking patterns including fast spiking, adaptive spiking, phasic spiking, and bursting. Our reconfigurable neuron offered a viable path to build a neuromorphic system with rich computing functions. Under bursting mode, the SNNs with bio-plausible neurons have shown better performance in both artificial neural network (ANN) converted SNNs and directly trained SNNs. Under adaptive mode, the neuron can efficiently encode input light signals with adaptive adjustment of firing rate, just like how our retina responds to external visual stimulations. Our robust reconfigurable neuron design could be used in a neuromorphic chip to expand its functionality beyond current homogeneous SNNs and pave the way for the development of more bio-plausible neuromorphic systems ( 44 ).",
"discussion": "DISCUSSION This work demonstrated a compact reconfigurable neuron capable of expressing various spike patterns for bio-plausible neuromorphic computing. The reconfiguration of the neuronal functions was conveniently achieved by modulating the resistance of the ECRAM device without the need to modify other circuit components or change input amplitudes. This simple reconfiguring mechanism can greatly improve the scalability of the circuits, which enabled us to achieve comparably small footprints with rich firing behaviors, as shown in table S2. The performance of the bio-plausible neuron in SNNs was evaluated through simulations, which demonstrated evident improvement in classification tasks compared to a simplified neuron model. The use of our neuron circuit in a future integrated chip could establish a previously unidentified platform for neuromorphic computing that closes the gap between the dynamical computing paradigms in biology and more digitized implementations in existing systems to enhance the cognitive capability of neuromorphic computing toward more advanced artificial intelligence."
} | 1,915 |
35287716 | PMC8922893 | pmc | 363 | {
"abstract": "Background O-Acetyl-L-homoserine (OAH) is an important potential platform chemical. However, low levels of production of OAH are greatly limiting its industrial application. Furthermore, as a common and safe amino acid-producing strain, Corynebacterium glutamicum has not yet achieved efficient production of OAH. Results First, exogenous L-homoserine acetyltransferase was introduced into an L-homoserine-producing strain, resulting in the accumulation of 0.98 g/L of OAH. Second, by comparing different acetyl-CoA biosynthesis pathways and adding several feedstocks (acetate, citrate, and pantothenate), the OAH titer increased 2.3-fold to 3.2 g/L. Then, the OAH titer further increased by 62.5% when the expression of L-homoserine dehydrogenase and L-homoserine acetyltransferase was strengthened via strong promoters. Finally, the engineered strain produced 17.4 g/L of OAH in 96 h with acetate as the supplementary feedstock in a 5-L bioreactor. Conclusions This is the first report on the efficient production of OAH with C. glutamicum as the chassis, which would provide a good foundation for industrial production of OAH. Supplementary Information The online version contains supplementary material available at 10.1186/s13068-022-02114-0.",
"conclusion": "Conclusion In this study, exogenous L-homoserine acetyltransferase was introduced into an L-homoserine-producing strain. Then, the effects of the introduction of the acetyl-CoA biosynthesis pathway and the addition of various feedstocks on the OAH biosynthesis were compared, resulting in improving OAH production to 3.2 g/L. Through the strong promoters to control the expression of L-homoserine acetyltransferase, the titer of OAH increased to 5.2 g/L. Finally, the OAH titer reached 17.4 g/L at 96 h in a 5-L bioreactor. This is the first time to achieve efficient production of OAH in C. glutamicum .",
"introduction": "Introduction O-Acetyl-L-homoserine (OAH) is a potential platform chemical for the production of high-value compounds, such as L-methionine [ 1 ] and γ-butyrolactone [ 2 ]. In biological systems, neither L-homoserine nor OAH is directly involved in protein biosynthesis, but they are precursors in the biosynthesis of L-methionine and S-adenosylmethionine. L-Methionine biosynthesis is strictly regulated, and its industrial production by microbial fermentation has not been realized. The industrial production is usually carried out by enzyme conversion and chemical synthesis with L-homoserine or OAH as the precursor [ 3 , 4 ]. When L-homoserine is used as the precursor, it needs to be activated by HCl before reacting with methanethiol to produce L-methionine [ 5 ]. Whereas, when OAH is selected as the precursor, it can directly react with methanethiol or 3-methylthiopropionaldehyde to form L-methionine [ 6 ]. Therefore, the production of OAH is very important for the industrial production of L-methionine. Escherichia coli and Corynebacterium glutamicum are the most popular strains used for the production of amino acids and their derivatives, such as L-glutamate, L-lysine, L-threonine, L-serine, L-histidine [ 7 – 9 ], 5-aminolevulinic acid and L-ornithine [ 10 , 11 ]. Compared with E. coli , C. glutamicum is a safe industrial microorganism, which is more reliable for the production of food and drug-related compounds. Reports have shown that L-homoserine and OAH have been produced efficiently in E. coli [ 12 – 16 ]. However, thus far, C. glutamicum has only achieved efficient production of L-homoserine [ 17 , 18 ]. L-Homoserine and acetyl-CoA are the substrates for OAH biosynthesis, whereas the production of OAH in C. glutamicum has not been reported. L-Homoserine can be efficiently accumulated in C. glutamicum , indicating that OAH should also be efficiently accumulated through efficient acetyl transfer [ 19 ]. Unfortunately, the engineered C. glutamicum strain only efficiently accumulated L-homoserine but not OAH without knock-out of the metX gene in our previous studies [ 17 , 18 ], suggesting that some problems need to be solved to achieve OAH accumulation. These problems may include the total enzyme activity, specific enzyme activity, heat resistance of L-homoserine acetyltransferase (MetX), and even the supply of acetyl-CoA [ 20 – 22 ]. Acetyl-CoA is not only a key intermediate metabolite that plays an irreplaceable role in cell growth and metabolic regulation, but also is the precursor of acetyl-CoA derivatives, whose accumulation needs to strengthen metabolic flow of acetyl-CoA biosynthesis [ 23 , 24 ]. There are many biosynthetic pathways of acetyl-CoA based on different substrates, such as pyruvic acid, acetic acid, and fatty acids [ 25 , 26 ]. Pyruvate forms acetyl-CoA through decarboxylation using the pyruvate dehydrogenase complex (PDH) [ 27 ]; acetate forms acetyl-CoA through the reversible Pta–Ack pathway or the irreversible ACS pathway [ 28 – 30 ]; fatty acids form acetyl-CoA through β-oxidation [ 31 ]. In contrast to glucose, acetate can be converted to acetyl-CoA without carbon loss. Moreover, the carbon content of acetic acid and glucose is equal, and acetate is cheaper than glucose. Therefore, at the same price, the mass of acetate is more than that of glucose [ 32 ]. In addition, strengthening the biosynthesis of CoA is another way to improve the biosynthesis of acetyl-CoA [ 13 ]. By engineering these pathways, the biosynthesis of acetyl-CoA in many microorganisms has been strengthened, resulting in the efficient accumulation of high-value acetyl-CoA derivatives [ 33 ]. In this study, an efficient OAH-producing strain was constructed via metabolic engineering based on an efficient L-homoserine-producing C. glutamicum strain reported in our previous study [ 18 ]. First, various L-homoserine acetyltransferase genes were individually introduced into the efficient L-homoserine-producing C. glutamicum strain. The best performer was chosen for further study. Then, different acetyl-CoA biosynthesis pathways were introduced to strengthen acetyl-CoA biosynthesis and explore the effects of acetyl-CoA on OAH accumulation. More importantly, different feedstocks (including acetate, citrate, and pantothenate) were added to the medium, resulting in significant increases in OAH accumulation. The production of OAH was further increased through the expression of L-homoserine dehydrogenase and L-homoserine acetyltransferase via strong promoters. These results showed that C. glutamicum efficiently accumulated not only L-homoserine, but also OAH. This system has great potential for the industrial production of OAH.\n\nIntroduction of different acetyl-CoA biosynthesis pathways OAH did not accumulate efficiently after enhanced MetX expression though the pEC-XK99E with high copy number and strong promoter. Therefore, we turned to the supply of acetyl-CoA, which was a precursor of OAH biosynthesis in addition to L-homoserine. In order to enhance the biosynthesis of acetyl-CoA, different acetyl-CoA biosynthesis pathways were introduced. Before introducing the acetyl-CoA biosynthesis pathways, the metX r from L. meyeri ( metX r _Lm ) gene was integrated into the genome of strain Cg-Hser with three strong promoters (P NCgl1676 , P sod , P tuf ) [ 37 ] (Fig. 2 C), generating Cg-1, Cg-2, and Cg-3, respectively. As shown in Fig. 2 B, the titers of OAH in these strains were 0.97 g/L, 0.76 g/L, 0.71 g/L, respectively. Then, we chose to upregulate the endogenous or introduce exogenous acetyl-CoA biosynthesis pathways, whose substrates were acetic acid or pyruvate, into strain Cg-1 to generate Cg-4, Cg-5, Cg-6, Cg-7, Cg-8, Cg-9, Cg-10, Cg-11, Cg-12, and Cg-13, respectively. However, the results showed that the enhancement of acetyl-CoA biosynthesis pathways did not improve the accumulation of OAH, and some of these strains even exhibited reduced accumulation (Fig. 3 ). Fig. 3 Effects of different acetyl-CoA biosynthetic pathways on OAH production Acetyl-CoA is a direct precursor of OAH biosynthesis, and a very key intermediate metabolite and regulator in organisms [ 38 ]. Therefore, the effective supply of acetyl-CoA should be an important factor for the efficient production of OAH. Attempts were made to strengthen the acetyl-CoA biosynthesis by introducing different acetyl-CoA biosynthesis pathways, but none of them had a positive effect on OAH accumulation, and some even had negative effects. At the same time, L-homoserine production was diminished, indicating that the introduction of acetyl-CoA biosynthesis pathways led to the reduction of metabolic flow in the direction of biosynthesis of L-homoserine and OAH. The acetyl-CoA biosynthesis pathway (derived from S. enterica and P. putida ) with acetate as its substrate had no positive or negative effects on OAH accumulation, probably because this pathway did not compete with L-aspartate family amino acids for pyruvate [ 39 ]. These results were very different from those in E. coli , in which acetyl-CoA biosynthesis was directly improved to promote the efficient accumulation of OAH on the basis of the efficient production of L-homoserine [ 13 ]. This suggested that, as a branch substance, the rational distribution of pyruvate was very important when it formed two direct substrates of the target product in C. glutamicum . Therefore, the factors limiting the further accumulation of OAH in C. glutamicum needed to be further explored.",
"discussion": "Results and discussion Construction of the OAH-producing strain Usually, the wild-type C. glutamicum ATCC 13032 has no capacity to accumulate OAH, even if there is a relevant biosynthetic pathway. Recently, engineered C. glutamicum strains have exhibited the ability to biosynthesize many amino acids including L-homoserine [ 18 ]. In biological systems, L-homoserine is the precursor of the biosynthesis of OAH [ 34 ]. Therefore, on the basis of high production of L-homoserine, a strain should be able to accumulate OAH via L-homoserine acetyltransferase. However, in our previous study, a high L-homoserine-producing strain without knock-out of the metX gene did not accumulate detectable OAH. This may be because the native expression of the metX gene was too low, and the enzyme activity was strictly regulated, resulting in the failure of acetyl transfer to L-homoserine [ 35 ]. In order to achieve the accumulation of OAH in C. glutamicum , an L-homoserine-producing strain (Cg13) was used as the starting strain, which was renamed Cg-Hser [ 18 ]. Strain Cg-Hser was derived from C. glutamicum ATCC 13032. In detail, some genes were successively knocked out, including mcbR (encoding a regulatory protein), metD (encoding amino acid import protein), thrB , pck , metB , and metY . The native genes including lysC T311I , asd , hom , pyc P458S , brnFE , and the heterologous aspC (from E. coli K12-MG1655) were upregulated though strong promoter P sod in the genome. The native genes including dapA and icd were downregulated though weak start codon replacement in the genome. However, the engineered strain Cg-Hser without knock-out of the metX gene failed to accumulate detectable OAH. Therefore, we should strengthen the expression of L-homoserine acetyltransferase. The MetX from Leptospira meyeri and C. glutamicum ATCC 13032, whose properties have been tested in vitro in previous study [ 13 ], were chosen. Same as previous studies [ 17 , 36 ], we directly expressed the metX genes from Leptospira meyeri and C. glutamicum ATCC 13032 by plasmid pEC-XK99E in vivo for faster screening of better performing enzymes, which generated strains Hser-1, Hser-2, respectively (Fig. 1 ). These engineered strains could accumulate about 0.9 g/L of OAH, and strain Hser-1 with expression of the metX variant gene ( metX r ) from L. meyeri could accumulate the highest titer of OAH (0.98 g/L) (Fig. 2 A), as in E. coli [ 13 ]. Fig. 1 Construction of an OAH-producing strain Fig. 2 Effects of introducing different sources of metX on OAH accumulation. A Effects of different sources of metX on OAH accumulation; B the OAH production of strains with metX r _Lm gene expression under the control of different promoters; C the expression intensity of three promoters (P NCgl1676 , P sod , P tuf ) Introduction of different acetyl-CoA biosynthesis pathways OAH did not accumulate efficiently after enhanced MetX expression though the pEC-XK99E with high copy number and strong promoter. Therefore, we turned to the supply of acetyl-CoA, which was a precursor of OAH biosynthesis in addition to L-homoserine. In order to enhance the biosynthesis of acetyl-CoA, different acetyl-CoA biosynthesis pathways were introduced. Before introducing the acetyl-CoA biosynthesis pathways, the metX r from L. meyeri ( metX r _Lm ) gene was integrated into the genome of strain Cg-Hser with three strong promoters (P NCgl1676 , P sod , P tuf ) [ 37 ] (Fig. 2 C), generating Cg-1, Cg-2, and Cg-3, respectively. As shown in Fig. 2 B, the titers of OAH in these strains were 0.97 g/L, 0.76 g/L, 0.71 g/L, respectively. Then, we chose to upregulate the endogenous or introduce exogenous acetyl-CoA biosynthesis pathways, whose substrates were acetic acid or pyruvate, into strain Cg-1 to generate Cg-4, Cg-5, Cg-6, Cg-7, Cg-8, Cg-9, Cg-10, Cg-11, Cg-12, and Cg-13, respectively. However, the results showed that the enhancement of acetyl-CoA biosynthesis pathways did not improve the accumulation of OAH, and some of these strains even exhibited reduced accumulation (Fig. 3 ). Fig. 3 Effects of different acetyl-CoA biosynthetic pathways on OAH production Acetyl-CoA is a direct precursor of OAH biosynthesis, and a very key intermediate metabolite and regulator in organisms [ 38 ]. Therefore, the effective supply of acetyl-CoA should be an important factor for the efficient production of OAH. Attempts were made to strengthen the acetyl-CoA biosynthesis by introducing different acetyl-CoA biosynthesis pathways, but none of them had a positive effect on OAH accumulation, and some even had negative effects. At the same time, L-homoserine production was diminished, indicating that the introduction of acetyl-CoA biosynthesis pathways led to the reduction of metabolic flow in the direction of biosynthesis of L-homoserine and OAH. The acetyl-CoA biosynthesis pathway (derived from S. enterica and P. putida ) with acetate as its substrate had no positive or negative effects on OAH accumulation, probably because this pathway did not compete with L-aspartate family amino acids for pyruvate [ 39 ]. These results were very different from those in E. coli , in which acetyl-CoA biosynthesis was directly improved to promote the efficient accumulation of OAH on the basis of the efficient production of L-homoserine [ 13 ]. This suggested that, as a branch substance, the rational distribution of pyruvate was very important when it formed two direct substrates of the target product in C. glutamicum . Therefore, the factors limiting the further accumulation of OAH in C. glutamicum needed to be further explored. Effects of several feedstocks on OAH accumulation The introduction of the exogenous acetate derived acetyl-CoA biosynthesis pathway failed to improve the accumulation of OAH. We speculated that this might be because there was no acetate available as a substrate for the biosynthesis of acetyl-CoA. Although some C. glutamicum strains can accumulate acetate [ 40 ], an analysis of the fermentation broth components found that all engineered strains in this study could not accumulate acetate under the culture conditions of this study. Therefore, to improve the biosynthesis efficiency of acetyl-CoA from acetate, it was necessary to add acetate to the culture medium. To avoid an adverse effect on cell growth caused by a sudden drop in pH, ammonium acetate was selected as the additive instead of acetic acid. Previous studies showed that L-homoserine could accumulate a high titer only after fermentation for 24 h [ 17 , 18 ]. To convert the added acetate into acetyl-CoA that could be used for acetylation of L-homoserine, 2.5 g/L of acetate was added at 24 h and 36 h. The results showed that the OAH titer of the engineered strains (Cg-4, Cg-7) did not increase when the metX r _Lm gene was only integrated into the genome. Whereas, the OAH titer increased significantly when the metX r _Lm gene was overexpressed via the plasmid. As shown in Fig. 4 A, the OAH titers of strains Cg-15 and Cg-16 were 2.1 g/L and 1.5 g/L, respectively. At this time, acetate was fully utilized, and the consumption of glucose did not change much, but the OD 600 of the strains increased. Interestingly, the OAH titer of strain Cg-14, which only expressed the metX r _Lm gene without introducing acetyl-CoA biosynthase, was higher after addition of acetate, up to 2.5 g/L. Fig. 4 Effects of several feedstocks on OAH accumulation. A Effects of acetate on OAH production. B Effects of citrate and pantothenate on the OAH production of strain Cg-14. C The addition methods of the several feedstocks Acetate was completely consumed, indicating that its addition may be the limiting factor in the OAH accumulation. Therefore, five feeding methods of acetate were chosen to study the effects on the OAH titers. The OAH titer of Cg-14 was the highest (3.2 g/L) after 5.0 g/L of acetate was added at 24 h and 36 h, and this was 28% higher than when 2.5 g/L of acetate was added at 24 h and 36 h (Fig. 4 A). However, when the addition of acetate was increased by 100%, the OAH titer increased by only 28%. Pantothenate is the precursor of CoA, which is the precursor of Acetyl-CoA. Acetyl-CoA is the competitive precursor for the biosynthesis of citric acid and OAH. To enhance the supply of acetyl-CoA for OAH biosynthesis and reduce the consumption of acetyl-CoA for citric acid biosynthesis in TCA cycle, pantothenate and citrate were also added to the culture medium. Strain Cg-14 was again employed and five feeding methods were chosen. The results showed that the citrate feeding significantly increased the biomass of the strain, but the OAH titer decreased sharply to about 0.5 g/L, indicating that citrate was not conducive to OAH accumulation. Different from citrate, the pantothenate feeding did not affect the OAH accumulation (Fig. 4 B). Corynebacterium glutamicum has an acetic acid biosynthesis pathway and the ability to accumulate acetic acid, but this ability is different under different culture conditions [ 41 ]. Under the culture conditions of this study, the strains could not accumulate acetic acid. Therefore, acetate needed to be added to make the introduced acetyl-CoA synthase function [ 42 ]. After acetate feeding, the titers of L-homoserine and OAH were both increased. Unexpectedly, the titer of OAH decreased after the introduction of acetyl-CoA biosynthase, indicating that the strains had a sufficient native capacity of acetate acetylation [ 43 ], and overexpression could cause a metabolic burden. Acetyl-CoA condenses with oxaloacetic acid by citrate synthase to form citric acid and then enters the TCA cycle. The citric acid feeding can improve the efficiency of the TCA cycle, but it leads to a sharp decrease in OAH accumulation, which may be because the ability to biosynthesize acetyl-CoA was strongly inhibited by citric acid, resulting in insufficient supply for L-homoserine acetylation even though acetate was added [ 44 ]. Enhanced expression of L-homoserine acetyltransferase In addition to acetate feeding, the introduction of different exogenous acetyl-CoA biosynthesis pathways and the citrate and pantothenate feeding failed to improve OAH accumulation. We speculated that the expression level of pathway enzymes would become the main limiting factor for the OAH accumulation after acetate feeding. Therefore, the thrA S345F gene (encoding bifunctional L-aspartokinase and L-homoserine dehydrogenase) from E. coli and the metX r _Lm was overexpressed by pEC-XK99E in strain Cg-1, resulting in strains Cg-17 and Cg-18, respectively. However, the OAH titer only reached 1.1 g/L and 3.2 g/L when 5.0 g/L of acetate was added at 24 h and 36 h, respectively (Fig. 5 A). In order to further enhance OAH accumulation, we used three strong promoters (P trc , P tac , P NCgl1676 ) to control the expression of the metX r gene from L. meyeri after overexpression of the thrA S345F gene in plasmid pEC-XK99E [ 37 ], resulting in strains Cg-19, Cg-20, and Cg-21, respectively. As shown in Fig. 5 A, the L-homoserine titers of the strains were 8.5 g/L, 8.2 g/L, and 8.0 g/L, respectively; and the OAH titers were 3.5 g/L, 4.8 g/L, and 5.2 g/L, which increased by 9.4%, 50.0%, and 62.5% compared with the control strain Cg-14, when 5.0 g/L of acetate was added at 24 h and 36 h, respectively. These results showed that the supply of acetyl-CoA was improved after the addition of acetate, and the OAH titer could be increased through the enhanced expression of L-homoserine acetyltransferase. Fig. 5 Effects of expression of thrA S345F and metX r genes with strong promoters on OAH production. A Effects of plasmid expression of thrA S345F and metX r genes with different strong promoters on OAH production; B OAH production in a 5-L bioreactor The 5-L bioreactor for OAH production A high concentration of acetate has a strong inhibitory effect on strain growth, and the pH of the fermentation process cannot be controlled in the shaking flask. Therefore, in order to explore the potential of acetate as a feedstock and improve OAH production, a 5-L bioreactor was used to carry out further experiments. As L-homoserine is the precursor of OAH biosynthesis, we chose the same conditions as in the previous L-homoserine production for OAH production. Before that, the cas9 and recET genes in the genome of strain Cg-21 were deleted, generating strain Cg-22. According to the above experiments, acetate (20% v/v) was added at 24 h to reach the concentration of 5 g/L, then it was added every 12 h. As shown in Fig. 5 B, the OAH titer of strain Cg-22 reached 17.4 g/L after 96 h, which was the highest titer, with 14.1 g/L of L-homoserine. These results suggested that acetate could improve the conversion of L-homoserine and the titer of OAH. This will provide a good basis for the industrial production of OAH."
} | 5,586 |
29921923 | PMC6008303 | pmc | 364 | {
"abstract": "Memristors with tunable resistance states are emerging building blocks of artificial neural networks. However, in situ learning on a large-scale multiple-layer memristor network has yet to be demonstrated because of challenges in device property engineering and circuit integration. Here we monolithically integrate hafnium oxide-based memristors with a foundry-made transistor array into a multiple-layer neural network. We experimentally demonstrate in situ learning capability and achieve competitive classification accuracy on a standard machine learning dataset, which further confirms that the training algorithm allows the network to adapt to hardware imperfections. Our simulation using the experimental parameters suggests that a larger network would further increase the classification accuracy. The memristor neural network is a promising hardware platform for artificial intelligence with high speed-energy efficiency.",
"introduction": "Introduction With the introduction of hardware accelerators 1 – 4 for inference in deep neural networks (DNNs) 5 – 9 , the focus on improving overall energy and time performance for artificial intelligence applications is now on training. One promising approach is in-memory analog computation based on memristor crossbars 10 – 18 , for which simulations have indicated potentially significant speed and power advantages over digital complementary metal-oxide-semiconductor (CMOS) 19 – 23 . However, experimental demonstrations to date have been limited to discrete devices 24 , 25 or small arrays and simplified problems 26 – 31 . Here we report an experimental demonstration of highly efficient in situ learning in a multilayer neural network implemented in a 128 × 64 memristor array. The network is trained on 80 000 samples from the Modified National Institute of Standards and Technology (MNIST) 32 handwritten digit database with an online algorithm, after which it correctly classifies 91.71% of 10 000 separate test images. This level of performance is obtained with 11% devices in the crossbar unresponsive to programming pulses and the training algorithm blind to the defectivity, demonstrating the self-adapting capability of the in situ learning to hardware imperfections. Our simulation based on the memristor parameters suggested that the accuracy could be higher than 97% with a larger (e.g., 1024 × 512) memristor array. Our results indicate that analog memristor neural networks can achieve accuracy approaching that of state-of-the-art digital CMOS systems with potentially significant improvements in speed-energy efficiency. Memristors offer excellent size scalability (down to 2 nm) 33 , fast switching (faster than 100 ps) 34 , and low energy per conductance update (lower than 3 fJ) 34 . Their tunable resistance states can be used both to store information and to perform computation, allowing computing and memory to be integrated in a highly parallel architecture. However, given the level of technology maturity, attempts to implement memristive neural networks have struggled with device non-uniformity, resistance level instability, sneak path currents, and wire resistance, which have limited array sizes and system performance. In particular, learning in memristor neural networks has been hampered by significant statistical variations and fluctuations in programmed conductance states and the lack of linear and symmetric responses to electric pulses 35 . Here we develop a reliable two-pulse conductance programming scheme utilizing on-chip series transistors to address the challenges in memristor conductance programming. This in situ training scheme enables the network to continuously adapt and update its knowledge as more training data become available, which significantly improves accuracy and defect tolerance.",
"discussion": "Discussion A further potential benefit of utilizing analog computation in a memristor-based neural network is a substantial improvement in speed-energy efficiency. The advantages mainly come from the fact that the computation is performed in the same location used to store the network data, which minimizes the time and energy cost of accessing the network parameters required by the conventional von-Neumann architecture. The analog memristor network is also capable of handling analog data acquired directly from sensors, which further reduces the energy overhead from analog-to-digital conversion. The memristors we used maintain a highly linear IV relationship, allowing for the use of voltage-amplitude as the analog input for each layer. This also minimizes circuit complexity and hence energy consumption for future hardware hidden neurons and output current readout. While the external control electronics we use in this work is not optimized for fast speed and low power consumption yet, previous literature on circuit design 45 , 51 and architecture 21 , 53 suggest an on-chip integrated system would yield significant advantages in speed-energy efficiency. In summary, we have demonstrated the in situ and self-adaptive learning capability of a multilayer neural network built by monolithically integrating memristor arrays onto a foundry-made CMOS substrate. The transistors enable reliable, linear, and symmetric synaptic weight updates, allowing the network to be trained with standard machine learning algorithms. After training with a SGD algorithm on 80 000 images drawn from the MNIST training set, we achieved 91.71% accuracy on the complete 10,000-image test set. This accuracy is 2.4% lower than an idealized simulation despite an 11% defect rate for the memristors used. The demonstrated performance with in situ online training and inference suggests that memristor crossbars are a promising high speed and energy efficiency technology for artificial intelligence applications. The software neurons used in this demonstration indicate that a hybrid digital processor and neuromorphic analogue approach for DNNs can be effective, but all the software functions used in the present demonstration can be integrated as hardware onto a full-function chip in the near future."
} | 1,519 |
19246237 | null | s2 | 365 | {
"abstract": "Bacterial populations utilize a variety of signaling strategies to exchange information, including the secretion of quorum-sensing molecules and contact-dependent signaling cascades. Although quorum sensing has received the bulk of attention for many years, contact-dependent signaling is forging a niche in the research world with the identification of novel systems and the emergence of more mechanistic data. Contact-dependent signaling is probably a common strategy by which bacteria in close contact, such as within biofilms, can modulate the growth and behavior of both siblings and competitors. Ongoing work with diverse bacterial systems, including Myxococcus xanthus, pathogenic Escherichia coli strains, Bacillus subtilis, and dissimilatory metal-reducing soil bacteria, is providing increasingly detailed insight into the dynamic mechanisms and potential of contact-dependent signaling processes."
} | 226 |
21613980 | PMC3130561 | pmc | 366 | {
"abstract": "Single-cell quantification of the input–output relation of the quorum-sensing circuit reveals how Vibrio harveyi employs multiple feedback loops to simultaneously control quorum-sensing signal integration and to ensure signal transmission fidelity.",
"conclusion": "Conclusion The integration of multiple signals is a common challenge faced by all living organisms. Here, we have exploited single-cell fluorescence microscopy to explore the integration of quorum-sensing signals by the model social bacterium V. harveyi . Multiple feedback loops in the V. harveyi quorum-sensing circuit actively regulate receptor ratios to control signal integration, expand the input and compress the output dynamic range, and regulate noise. The sophistication of the circuitry appears to reflect the complex requirements of responding appropriately to multiple signals in a dynamically changing environment.",
"introduction": "Introduction Cell-to-cell communication is fundamental to both unicellular and multicellular life. Cells often detect multiple chemical communication cues simultaneously and integration of the information encoded in these cues guides their behavior. Correctly integrating signals generally requires complex signal transduction pathways ( Pawson and Scott, 2010 ). In insulin signaling, pancreatic β islets regulate cardiac contractility and insulin secretion by the synergistic action of multiple second messenger molecules such as cAMP and calcium-responsive effectors ( Saltiel and Kahn, 2001 ). Integration of two signals enhances precision of some biological events ( Pawson and Scott, 2010 ) and errors in neuronal signal integration underlie many human diseases. For example, proper neurodevelopment requires that the protein disrupted-in-schizophrenia-1 (DISC1) integrate signals from two parallel pathways ( Mao et al, 2009 ). Mutations in DISC1 are associated with schizophrenia, a psychiatric disorder of social interaction ( Millar et al, 2000 ). In prokaryotes, bacterial chemotaxis provides a paradigm for cellular response to multiple environmental stimuli ( Armitage, 1992 ). The well-studied chemotactic signaling circuit in Escherichia coli receives both positive and negative signals, and generates an integrated response ( Khan et al, 1995 ). In addition to integrating signals, regulatory circuits must ensure signal transmission fidelity. Information can be lost or corrupted by internally generated noise (e.g. fluctuations in protein numbers) or by external perturbations (e.g. changes in temperature) and circuits must be designed to compensate for such factors. For example, the chemotaxis network of E. coli is designed to function robustly in the presence of gene-expression noise ( Kollmann et al, 2005 ) and circadian circuits accurately compensate for temperature variation ( Virshup and Forger, 2009 ). While signal integration and high-fidelity signal transmission have been addressed separately, little is known about the mechanisms cells use to solve both tasks simultaneously. Here, we report how the model bacterium Vibrio harveyi simultaneously integrates and faithfully transmits chemical signals. In a process called quorum sensing, bacteria communicate by synthesizing, releasing, and detecting signal molecules called autoinducers (AIs). The bioluminescent marine bacterium V. harveyi integrates three AI signals into its quorum-sensing circuit: AI-1, an intra-species signal, CAI-1 an intra-genera signal, and AI-2 a ‘universal’ signal. Each signal is detected by a cognate receptor AI-1/LuxN, CAI-1/CqsS, and AI-2/LuxPQ ( Figure 1A ). The information contained in the three AIs is transduced through a shared signaling pathway. At low cell density, in the absence of AIs, the receptors autophosphorylate and pass phosphate to the phosphorelay protein LuxU, which in turn shuttles phosphate to the response regulator LuxO. Phosphorylated LuxO (LuxO∼P) activates transcription of genes encoding five small regulatory RNAs, Qrr1-5, that repress translation of the mRNA encoding the master quorum-sensing regulator LuxR. At high cell densities, the AIs accumulate, bind their receptors, and convert the receptors to phosphatases, thereby draining phosphate from LuxU and LuxO. Consequently, the Qrr sRNAs are not produced and luxR mRNA is translated. LuxR protein controls >100 genes that underpin collective behaviors including bioluminescence and biofilm formation. There are five feedback loops in the V. harveyi quorum-sensing circuit ( Figure 1A ). First, LuxO autorepresses its own transcription. Second, the Qrr sRNAs repress luxO translation. Third, LuxR autorepresses its own transcription. Fourth, LuxR activates expression of the qrr 2-4 genes, which in turn, repress luxR translation. Fifth, as we show below, the luxMN operon, encoding the AI-1 synthase and receptor, is repressed by the Qrr sRNAs. In a previous study, Long et al (2009) showed that information from the V. harveyi AIs is integrated strictly additively, with a close balance between the strengths of the different signals. That study did not, however, address how the circuit uses shared components to distinguish between multiple AI inputs or what role each feedback loop has in signal integration and transmission. To explore these features, here we examine the input–output relation between AIs and LuxR, using a suite of strains with specific feedback loops either present or disrupted. We found, first, that feedback onto LuxN allows V. harveyi to actively adjust its relative sensitivity to AI signals as cells transition from low to high densities, and, second, that the other feedback loops control the input and output dynamic ranges and the noise in the circuit. Remarkably, by functioning together, these feedback loops compress a 3 order of magnitude input range into a six-fold output range. Our results reveal that the V. harveyi quorum-sensing circuit employs multiple feedback loops to actively regulate signal integration while maintaining signaling fidelity.",
"discussion": "Discussion The V. harveyi quorum-sensing circuit uses multiple feedback loops to control AI signal integration and signal transmission. Intriguingly, we found that a feedback loop regulates the level of the AI-1 receptor LuxN, to adjust the relative sensitivity to AI signals. In principle, this active regulation allows cells to ‘pay attention’ to particular signals at different stages during the transition from low to high cell density, and potentially helps explain how the quorum-sensing circuit can distinguish multiple inputs using a single output. Our study also provides quantitative evidence that multiple feedback loops control the input and output dynamic ranges and noise level of the circuit. Why does the circuit actively regulate receptors? Our results suggest that, in addition to the input signals, the receptor ratios ( Supplementary Figure S9 ) also have a role in determining the circuit output. One advantage of this regulation is to assure that cells only respond to steady, reliable signals. Cells control protein levels on the time scale of the cell cycle, so the regulation of receptor levels may aid in filtering out rapid signal fluctuations from one particular AI channel in the environment. In this scenario, at low cell density, a transient increase in AI-1 will not cause cells to commit to the quorum-sensing transition because the LuxN receptor level is low. Rather, cells will only respond to persistently high levels of AIs, allowing time for LuxN to build up, before committing to the quorum-sensing transition. As a corollary, however, because of the low copy number of LuxN at low cell density, the circuit is more sensitive to an impulse of AI-2. Such a changing balance in the sensitivity to AIs with increasing cell density could constitute a useful property of the quorum-sensing circuit. Mehta et al (2009) previously suggested that feedback on receptor number allows bacteria to focus attention on one input in order to monitor different stages of development. Our results suggest the cells pay more attention to AI-2 than AI-1 at low cell density and then pay more attention to AI-1 than to AI-2 at high cell density. Nadell et al (2008) previously suggested that the ‘universal’ signal AI-2 is more informative for a mixed-species community at the early stage of biofilm formation and AI-1 is more informative for a single-species community at a later stage of biofilm formation. Thus, it is possible that the directed sensitivity of the quorum-sensing circuit evolved in response to a canonical progression from a mixed-species to a single-species community during biofilm development. Quantitatively, the differences in WT Loop strain receptor ratios inferred from the slopes of the output contours in Figure 7A are modest, ∼40% over the full AI range. Why are receptor ratios held within such a narrow range? Our data in Figure 6B suggest that cells lose all sensitivity to one channel if the expression level of the receptor for another channel becomes too high. Receptor ratios therefore cannot undergo extreme variations if cells are to maintain sensitivity to multiple AI channels. Our direct measurements of LuxN expression ( Supplementary Figure S5 ) indicate a large increase in LuxN levels with increasing AI concentration, but we have not observed a parallel increase in LuxPQ expression. Thus, how the effective receptor ratio between LuxN and LuxPQ is maintained within the narrow range inferred from the signaling data ( Figure 7A ) remains unknown. In the case of LuxN, we observed a large increase in receptor number at high AI concentrations. One consequence of higher receptor levels might be to establish dominance of AI signaling over any other inputs that impinge on the quorum-sensing pathway. For example, in Vibrio cholerae the VarS/VarA two-component system affects the activity of LuxO ( Lenz et al, 2005 ). By acting as phosphatases at high cell density, receptors prevent other kinases from phosphorylating and thereby activating LuxO, and so higher receptor levels might insulate the pathway from other inputs once high cell density is achieved. A second consequence of higher receptor levels might be to accelerate the transition from high cell density to low cell density. The accumulation of receptors at high cell density maximizes the total kinase rate upon a transition from high cell density to low cell density. The resulting strong kinase activity will rapidly phosphorylate LuxO, leading in turn to rapid production of Qrr sRNAs and an accelerated switch to the low cell density program of gene expression. Since transitions from high cell density to low cell density in the natural environment may be very fast, for example during shedding from a biofilm of expulsion from a eukaryotic host, cells may put a premium on rapid induction of the low cell density program of expression. The circuit tightly regulates both the input and output dynamic range Previous work has shown that negative feedback in signaling can contribute to the linearization of the input–output relationship. Nevozhay et al (2009) demonstrated that negative autoregulation coupled with pairing between a repressor and an inducer can robustly convert a sigmoidal dose response to a linear dose response. Yu et al (2008) identified a related negative feedback effect in the pheromone response of yeast—negative feedback aligns the dose response of consecutive steps in a pathway, which improves the linearity of the overall relation between the input pheromone concentration and the pathway output. In our study, the input dynamic range is ∼100-fold larger for WT Loop strain ( Figure 4B ) than for the Five Loop mutant ( Figure 4R ) and, likewise, the dose response for the WT Loop strain is much more linear (up to ∼2-fold difference in Hill coefficients) ( Figure 5C ). These results demonstrate that the multiple feedback loops act together to significantly broaden the input dynamic range. Specifically, the Five Loop mutant has a sharp transition with regard to both AIs, so that this feedback-lacking strain acts as a ‘coincidence detector’ for AIs, only responding to simultaneously high levels of both AIs. As a result, the circuit without negative feedback loops loses information about the full range of AI inputs. In contrast to such an ‘on–off’ switch, the circuit in the WT Loop strain gradually responds over a broad range of AI signals. We speculate that in the natural environment it may be important for V. harveyi to correctly respond to different AI blends. As mentioned, it is possible that different combinations of AIs encode information about the developmental stage of the community (e.g. the development of a biofilm). If so, it could be advantageous for cells to coordinate their behaviors at multiple developmental stages by expanding and actively controlling the AI dose–response range via negative feedback loops. The mean LuxR contour in the WT Loop strain ( Figure 4B ) clearly demonstrates that the circuit tightly controls the output dynamic range to a modest six-fold difference. How do the feedback loops regulate this range? We observed that the output range is 20-fold for both the LuxOLuxR Loop mutant ( Figure 4J ) and the Five Loop mutant ( Figure 4R ). At high cell density, the maximum level of LuxR is much larger for the LuxOLuxR Loop mutant and the Five Loop mutant (∼1200 copies) than for the WT Loop strain (∼650 copies). Apparently, LuxR autorepression places an upper bound on LuxR levels. As shown in Figure 5A , we found that at low cell density, the minimum level of LuxR is the same (∼60 copies) for the LuxOLuxR Loop mutant and the Five Loop mutant, and lower than the level (∼110 copies) for WT and the LuxN Loop mutant. We reason that LuxR can never be fully repressed by the Qrr sRNAs because the high level of Qrr sRNAs needed to do this would also fully repress LuxO, which is required for qrr expression. Indeed, disruption of Qrr-mediated repression of LuxO did lead to lower minimum levels of LuxR, but only by ∼45%, suggesting that other mechanisms also prevent LuxR from being fully repressed. Why does the circuit limit the output dynamic range to only six-fold? In a previous study ( Svenningsen et al, 2008 ), it was shown that in V. cholerae , the transition from high cell density to low cell density is accelerated by HapR activation of qrr gene expression (recall that HapR is the V. cholerae homolog of V. harveyi LuxR). In particular, the accumulation of HapR at high cell density activates rapid production of Qrr sRNAs upon the switch to low cell density. While the Qrr sRNAs lead to rapid destruction of hapR mRNA, the levels of HapR protein decrease only slowly, due to dilution by growth. V. harveyi LuxR protein exhibits the same behavior ( Tu and Bassler, 2007 ). It is possible that the circuit tightly limits maximal LuxR levels to prevent LuxR from being present for an overextended time following the transition from high cell density to low cell density. Thus, both the tight control of LuxR levels and the accumulation of receptors at high cell density suggest that the quorum-sensing circuit is at least in part designed to accelerate the response of cells to sudden transitions from high to low cell density. Feedback noise in two-component systems The WT Loop strain ( Figure 4B ) displays higher LuxR noise in an AI-1-only environment than in an AI-2-only environment. By contrast, in the LuxN Loop mutant ( Figure 4N ), the noise is roughly identical under these two conditions. We concluded that the LuxN feedback loop contributes to LuxR noise, and we speculated that the loop promotes positive feedback when LuxN acts as a phosphatase, which increases the noise in the circuit. Previously, Christian et al (2010) showed that a two-component system can change the sign of its feedback depending on the signal level and negative feedback often reduces noise while positive feedback leads to phenotypic heterogeneity. In our case, we expect the LuxN feedback to be positive at high AI-1 concentrations, and thus increase the noise. However, the overall change in noise level is modest, and it is not clear whether the increased noise at high AI-1 concentrations has any beneficial role or is simply an unavoidable consequence of the regulated increase in LuxN levels at high cell density."
} | 4,093 |
24320083 | PMC3896932 | pmc | 367 | {
"abstract": "Biologically produced methane (CH 4 ) from anaerobic digesters is a renewable alternative to fossil fuels, but digester failure can be a serious problem. Monitoring the microbial community within the digester could provide valuable information about process stability because this technology is dependent upon the metabolic processes of microorganisms. A healthy methanogenic community is critical for digester function and CH 4 production. Methanogens can be surveyed and monitored using genes and transcripts of mcrA , which encodes the α subunit of methyl coenzyme M reductase – the enzyme that catalyses the final step in methanogenesis. Using clone libraries and quantitative polymerase chain reaction, we compared the diversity and abundance of mcrA genes and transcripts in four different methanogenic hydrogen/CO 2 enrichment cultures to function, as measured by specific methanogenic activity (SMA) assays using H 2 /CO 2 . The mcrA gene copy number significantly correlated with CH 4 production rates using H 2 /CO 2 , while correlations between mcrA transcript number and SMA were not significant. The DNA and cDNA clone libraries from all enrichments were distinctive but community diversity also did not correlate with SMA. Although hydrogenotrophic methanogens dominated these enrichments, the results indicate that this methodology should be applicable to monitoring other methanogenic communities in anaerobic digesters. Ultimately, this could lead to the engineering of digester microbial communities to produce more CH 4 for use as renewable fuel.",
"introduction": "Introduction Anaerobic wastewater treatment is an environmentally and economically beneficial process in which the biological degradation of organic wastes results in the production of CH 4 as a carbon-neutral energy source (Zitomer et al ., 2008 ). Unfortunately, the widespread application of this ‘green’ technology has been hampered by concerns about process stability. Although treatment failure can be a serious problem, monitoring of anaerobic biomass can be used to measure the efficacy of bioaugmentation or system control used to prevent digester failure or encourage faster recovery of stressed digesters (Castellano et al ., 2007 ; Schauer-Gimenez et al ., 2010 ). Digester failure occurs when the microbial community is stressed by organic overload or toxicants or other abrupt environmental changes (Castellano et al ., 2007 ). Therefore, the results of assays that rapidly monitor the microbial community in anaerobic biomass could provide useful information to operators seeking to manage digester function. In practice, however, the microorganisms in the anaerobic microbial community are rarely monitored directly. In fact, the microbial community in anaerobic digesters has been a black box throughout most of the history of this technology (Rivière et al ., 2009 ). Specific methanogenic activity (SMA) assays, methane production rates, biogas composition, chemical oxygen demand (COD) removal, pH, granule morphology, acetate utilization rates, methanethiol concentration and quantification of volatile fatty acids have all been suggested or used to evaluate digester function (Coates et al ., 1996 ; Zitomer et al ., 2000 ; Castellano et al ., 2007 ; Conklin et al ., 2008 ; Molina et al ., 2009 ). While these parameters are closely related to the metabolic functions of the microbial community, they do not directly quantify microorganisms. Successful removal of organic waste from the influent wastewater and methane production depend upon the collaborative efforts of the members of an interdependent microbial community, so knowledge of the structure and function of the community in anaerobic wastewater digesters can be very useful when attempting to stabilize or increase the efficiency of waste removal and biogas production. For example, SMA assays have been used to evaluate changes in biomass activity by quantifying the production of methane per the amount of volatile suspended solids (VSS) per unit time (Coates et al ., 1996 ; 2005 ). The methanogens are the source of the methane, and they can be directly targeted using molecular microbiological methods. Methanogen genomes encode a unique operon for the enzyme methyl coenzyme M reductase (MCR). Previous studies have established that the presence and transcription of the gene for the alpha subunit of MCR ( mcrA ) can be used to detect the presence, abundance and/or activity of methanogens in natural and engineered environments (Springer et al ., 1995 ; Luton et al ., 2002 ; Juottonen et al ., 2008 ; Gagnon et al ., 2011 ; Kampmann et al ., 2012 ; Munk et al ., 2012 ). Other studies have demonstrated that the methane flux from peat and biogas production from anaerobic biomass correlated with the abundance of mcrA (Freitag and Prosser, 2009 ; Freitag et al ., 2010 ; Traversi et al ., 2012 ). Based in part on these reports, it was hypothesized that quantification of mcrA genes and/or transcripts by quantitative polymerase chain reaction (qPCR) would correlate with SMA results and could thus be used in their stead. The substitution could provide a substantial benefit because qPCR assays can be completed within 24 h from biomass collection – whereas SMA assays can take up to 7–10 days to complete, giving digester operators information about the activity of the biomass much more rapidly than SMA assays. Herein, we report an evaluation of the use of qPCR of mcrA genes and transcripts in comparison with traditional SMA assays on the biomass from four different H 2 /CO 2 enriched bioreactors.",
"discussion": "Discussion In anaerobic wastewater treatment, methanogens are critically important, serving as both the final step in organic degradation and the source of CH 4 . Therefore, studies of methanogen dynamics can provide valuable information for the development and monitoring of this form of biotechnology. The enrichment cultures used in these analyses were fed with H 2 /CO 2 as the primary substrates. Thus, it was expected that hydrogenotrophic methanogens would dominate and analysis of the mcrA clone libraries (Fig. 4 ) generally supported this expectation. The exception was that mcrA sequences from Methanosaeta , acetoclastic methanogens adapated to low acetate concentrations (Jetten et al ., 1990 ), represented about 60% of the clones in R4 in one set of analyses, but these contributed to only ∼ 5% of the mcrA transcripts. The low abundance and/or mcrA transcript activity of acetoclastic methanogens were probably the major reasons why the SMA analysis with acetate (although limited) was below detection. Using H 2 /CO 2 also limited the necessity of methanogens obtaining H 2 from syntrophic acetogens which contributed to the SMA using propionate also being below detection. However, the SMA values measured herein for H 2 /CO 2 (Fig. 1 ) were within the previously reported range (30–1500 ml CH 4 g −1 VSS-h −1 for pure cultures) for strictly anaerobic cultures (Pavlostathis and Giraldo-Gomez, 1991 ). The qPCR results indicated a significant correlation between the abundance of primarily hydrogenotrophic methanogens ( mcrA copy number) and H 2 /CO 2 SMA values (Fig. 3 ). This finding complements previous studies which linked mcrA gene copy number to methane flux (Traversi et al ., 2012 ). However, the qRT-PCR results did not demonstrate a correlation between mcrA transcripts and SMA values. This is in contrast to results found by Munk and colleagues ( 2012 ) who found a correlation between methane productivity and the concentration of mcrA transcripts. These authors did not use a SMA approach to estimate methane production rates. The results in the present study indicated that the number of methanogens present was more important for the rates of methane production in these H 2 /CO 2 enrichment cultures. This finding is encouraging in that it indicates that qPCR of mcrA , which can be performed within a 24-h time frame, provides information which correlates with SMA, a 2-day to 1-week standard method for determining the activity of anaerobic biomass. While our values for mcrA genes and transcripts were higher than those in peat reported in a study by Freitag and Prosser ( 2009 ), this result was expected because of the H 2 /CO 2 enrichment of the bioreactors for hydrogenotrophic methanogens. However, the previous study did not detect the same strong correlation between mcrA gene abundance and measurements of methane flux (Freitag and Prosser, 2009 ). Instead, transcript to gene copy ratios showed the best correlation with methane production (r 2 = 0.79, Freitag and Prosser, 2009 ), but transcript to gene copy ratios for mcrA did not correlate with SMA in this study. This difference may be due to several factors including sample type, the diversity of the methanogens in each environment and methods of measurement for methane production. The variation in the qPCR and qRT-PCR results from each enrichment culture across the three sampling dates (Fig. 2 A and B) may have been due to the fact that the biomass samples were not collected at any specific time of day, especially in reference to the daily pulse feeding and biomass wasting. Still, the trend (abundance of mcrA genes in R1 and R3 > R2 and R4) across all three dates was clearly the same, and the correlation of the mean values was significant. qPCR and qRT-PCR assays temporally represent a ‘snapshot’ of methanogen abundance, while SMA assays take much longer to complete. Therefore, it made sense to use multiple extractions over time to generate qPCR results for comparison. Results for qPCR and qRT-PCR were normalized to ng of DNA or RNA respectively. In a similar study with peat, Freitag and Prosser ( 2009 ), used both nanogram of nucleic acids and gram of soil. Although it would have been possible to normalize to VSS, a measure of the organic content, or millilitre of culture, the respective nucleic acids were chosen for several reasons. VSS measures all organic content including recalcitrant organic substrates and dead organisms, and thus, the actual active biomass component could have varied widely among the samples. The VSS for each culture ranged from 162 mg VSS l −1 in R1 to 515 mg VSS l −1 in R2, with VSS values in R3 and R4 falling in between. Therefore, 1 ml of biomass from each reactor could have contained vastly different amounts of organic compounds, and although VSS is not an ideal measure of active biomass, using equal amounts of samples with wide disparity in VSS could represent variation in the abundance of methanogens as well. Using nucleic acids for normalization allowed us to calculate mcrA genes or transcripts as a proportion of the DNA or RNA extracted from the sample, making comparison among bioreactors as straightforward as possible. Other possible explanations for the similar SMA results from cultures R1 and R3 or R2 and R4 were that the methanogens in these cultures were alike or dominated by a particular species, or that the active methanogen populations in cultures with similar SMAs were comparable. We performed community analysis on methanogens in order to rule out these possibilities. After analysing the communities from each culture two ways (Fig. 4 A and B), we found no correlation between the community structure of the methanogens in the cultures and SMA values. Therefore, based on our analyses, community structure was not related to methane production rates in these cultures. Variation among methanogen transcription and translation rates for mcrA , as well as the half-life and stability of the mRNA and the resulting protein, may all have affected the outcomes of this study; however, very little of these data is available for methanogen genera. Furthermore, while mcrA has been demonstrated to be a valuable gene for use in the investigation of methanogens in the environment, the data obtained from PCR-based methods using primers for mcrA are subject to biases inherent in the process (von Wintzingerode et al ., 1997 ). However, the primer set designed by Luton and colleagues ( 2002 ) has previously been shown to consistently amplify mcrA from a wide range of methanogen genera, making the set a sound choice for the examination of methanogens in environmental samples (Luton et al ., 2002 ; Banning et al ., 2005 ; Juottonen et al ., 2006 ). Further physiological information about methanogens and mcrA would be useful for interpreting these data as the link between genetic differences in mcrA and MCR activity has not been explored. In summary, the data from this study may be used to better understand methanogenic community structure in anaerobic digesters even though the enrichment process favoured hydrogenotrophic methanogens. Recent papers have indicated the importance of hydrogenotrophic methanogens in anaerobic digesters under some conditions (Kampmann et al ., 2012 ; Sundberg et al ., 2013 ). Quantification of mcrA genes was correlated with SMA values, and therefore, qPCR assays could be a valuable, time saving method for monitoring and assessing anaerobic biomass. Future studies that include lab-scale and industrial-scale digester biomass containing a both hydrogenotrophic and acetoclastic methanogens will be performed to assess this method for wider application. We report a significant correlation between the abundance of mcrA gene copies and SMA results. We include analysis of mcrA DNA and cDNA clone libraries from each of the bioreactors in order to rule out the influence of similarities among methanogen community structure on these results. These results suggest that SMA assays of biomass activity may be replaced by a faster method, qPCR of mcrA ."
} | 3,446 |
35733974 | PMC9207759 | pmc | 368 | {
"conclusion": "Conclusions In conclusion, many studies have provided evidence that G. sulfurreducens expresses e-pili comprised of the pilin monomer PilA. It remains a mystery as to why Gu et al. ( 2021 ) did not recover filaments comprised of PilA from their strain of ‘wild-type' G. sulfurreducens when so many other studies, including several by the senior author of Gu et al., had previously found PilA in filament preparations. Furthermore, e-pili comprised of PilA can clearly be seen emanating from cells of G. sulfurreducens . Other microbes can heterologously express the G. sulfurreducens PilA and assemble it into the same type of e-pili found in G. sulfurreducens . Consistent with these observations, G. sulfurreducens nanowire conductivity is readily tuned simply by changing the abundance of aromatic amino acids in the pilin expressed. Expression of poorly conductive pili has demonstrated the importance of e-pili in Fe(III) oxide reduction, electron transfer to other microbial species, and for generating high current densities in bioelectrochemical systems. Therefore, at present the preponderance of evidence is that e-pili, comprised of PilA, not only exist, but are an important feature in Geobacter extracellular electron exchange. The pilins and archaellins of phylogenetically distinct bacteria and archaea are assembled into conductive filaments and it seems likely that e-pili and e-archaella are spread throughout the microbial world (Walker et al., 2018 , 2019 , 2020 ; Bray et al., 2020 ; Lovley and Holmes, 2020 ).",
"introduction": "Introduction There is a debate whether Geobacter sulfurreducens produces electrically conductive pili (e-pili) from its pilin monomer, PilA, a protein encoded by gene GSU 1496. G. sulfurreducens assembly of the PilA into e-pili was proposed over a decade ago (Reguera et al., 2005 ). As detailed below, many subsequent studies have provided additional data consistent with this concept ( Figure 1 ). However, Gu et al. have recently concluded that G. sulfurreducens does not express e-pili from PilA (Gu et al., 2021 ). Figure 1 Evidence consistent with the hypothesis that Geobacter sulfurreducens expresses e-pili comprised of the pilin monomer, PilA and that PilA can be assembled into conductive filaments. This is not a controversy over small details of the physiology of one microbe. Geobacter species play an important role in natural environments and biotechnologies. For example, Geobacter species are typically abundant in soils and sediments in which Fe(III) oxide reduction has a significant impact on the biogeochemical cycling of carbon, nutrients, and trace metals as well as in bioremediation (Lovley et al., 2011 ; Reguera and Kashefi, 2019 ; Lovley and Holmes, 2022 ). Geobacter species are also often abundant in soils and anaerobic digesters in which direct interspecies electron transfer (DIET) appears to be an important mechanism for methane production (Zhao et al., 2020 ; Lovley and Holmes, 2022 ). Geobacter and closely related species are often enriched on the anodes of electrodes harvesting electricity from organic matter and G. sulfurreducens generates the highest current densities of all known electroactive isolates (Lovley et al., 2011 ; Logan et al., 2019 ). Although other microbes, most notably Shewanella species, have been helpful for developing an understanding of key extracellular electron transfer mechanisms (Shi et al., 2016 ; Lovley and Holmes, 2022 ), there are no pure cultures that are as effective in Fe(III) oxide reduction, DIET, and current production as G. sulfurreducens and its close relative G. metallireducens . Furthermore, if it were true that PilA cannot be assembled into conductive filaments, this would mean that attempts to develop new protein-based electronic materials based on concepts for electron transport along e-pili (Creasey et al., 2018 ; Dorval Courchesne et al., 2018 ; Gutermann and Gazit, 2018 ; Cosert et al., 2019 ; Roy et al., 2020 ) may be misguided. The reported heterologous expression of e-pili from PilA in Pseudomonas aeruginosa (Liu et al., 2019 ) or Escherichia coli (Ueki et al., 2020 ) for mass production of e-pili for the fabrication of electronics would require new, non-obvious explanations to describe how introducing G. sulfurreducens PilA confers the capacity for conductive filament expression. Other apparent accomplishments for electronics applications, also achieved simply by modifying the structure of PilA, such as tuning of the conductivity of G. sulfurreducens filaments or the introduction of novel binding sites on filaments to enhance sensor selectivity (Lovley and Yao, 2021 ), would also need reevaluation. The function of electronic devices for electricity generation (Liu et al., 2020b ), sensing (Liu et al., 2020a ; Smith et al., 2020 ), and neuromorphic memory (Fu et al., 2020 , 2021 ) would need to be reconsidered."
} | 1,234 |
35358187 | PMC8970472 | pmc | 369 | {
"abstract": "With this work we introduce a novel memristor in a lateral geometry whose resistive switching behaviour unifies the capabilities of bipolar switching with decelerated diffusive switching showing a biologically plausible short-term memory. A new fabrication route is presented for achieving lateral nano-scaled distances by depositing a sparse network of carbon nanotubes (CNTs) via spin-coating of a custom-made CNT dispersion. Electrochemical metallization-type (ECM) resistive switching is obtained by implanting AgAu nanoparticles with a Haberland-type gas aggregation cluster source into the nanogaps between the CNTs and shows a hybrid behaviour of both diffusive and bipolar switching. The resistance state resets to a high resistive state (HRS) either if the voltage is removed with a retention time in the second- to sub-minute scale (diffusive) or by applying a reverse voltage (bipolar). Furthermore, the retention time is positively correlated to the duration of the Set voltage pulse. The potential for low-voltage operation makes this approach a promising candidate for short-term memory applications in neuromorphic circuits. In addition, the lateral fabrication approach opens the pathway towards integrating sensor-functionality and offers a general starting point for the scalable fabrication of nanoscaled devices.",
"conclusion": "Conclusion In this work sparse CNT networks as a new approach for obtaining nanoscaled distances in a lateral geometry as well as those networks with implanted AgAu nanoparticles as a novel lateral memristor with short-term memory capabilities and a hybrid switching behaviour between diffusive and bipolar switching have been presented. The sparse CNT networks have shown to offer a general starting point for introducing nanoscaled gaps into laterally oriented systems. The CNTs exhibit the function of bridging substantial distances between the electrodes while gaps between individual CNTs have shown to be in the nanometer range. AgAu nanoparticles implanted into this CNT network showed ECM-type resistive switching. The switching behaviour is based on providing a substantial but still limited reservoir of silver by deposited AgAu nanoparticles, yielding a diffusive switching behaviour with a second-scale retention as well as the ability of bipolar devices to reset to the HRS by reverse voltages. It has been shown that the “memory span”, i.e. the retention, is positively correlated to the width of the Set voltage pulse since the retention is prescribed by the filament’s lifetime and thus its thickness. Also, it has been shown that the CNT/AgAu networks are able to reach switching voltages in the range providable by chips fabricated with complementary metal-oxide semiconductor (CMOS) technology, while the network provides an integrated serial resistance limiting the current to the nA to μA range. This makes CNT/AgAu networks a promising approach for low-power short-term memory components in neuromorphic circuits, though for a deeper understanding further investigations with respect to the detailed correlation of Set pulse width and retention time as well as the impact of the network topology are necessary.",
"introduction": "Introduction After the memristor’s postulation by Chua et al . [ 1 ] and the reported link between memristor theory and resistive switching in TiO 2 thin films by Strukov et al . in 2008 [ 2 ], the potential of resistive switching phenomena has led to a broad variety of research directions. The application potential ranges from non-volatile memory [ 3 ] over bio-inspired neural networks as a promising approach to overcome the von-Neumann bottleneck [ 4 ] to the concept of a “memsensor” joining memristive with sensitive functionality, allowing for unique features such as habituation to a permanent background signal [ 5 ]. Different types of memristive devices based on their switching mechanisms have been reported such as valence change mechanism (VCM), phase change materials (PCM) or electrochemical metallization (ECM), with different characteristics e.g. bipolar, unipolar or diffusive switching [ 6 – 12 ]. ECM cells commonly consist of a dielectric layer of a few nm in thickness between two metal electrodes, where one is electrochemically active e.g. Cu or Ag [ 9 , 13 ]. Field-driven oxidation and motion of metal ions as well as their subsequent reduction at the cathode lead to the formation of a metal filament switching the device resistance from matrix determined to metal determined. Due to their bipolar switching behaviour ECM cells have been commonly discussed as candidates for non-volatile memory applications [ 14 ]. However, recent advances in ECM systems included introducing metal nanoparticles into the dielectric matrix between inert electrodes to act as an ion reservoir under exploitation of the inherent local field enhancement of nanoparticles [ 15 ], with stable diffusive switching properties that have been reported for AgAu and AgPt nanoparticles (NP) embedded in a SiO 2 matrix in [ 16 ]. These and most other memristive devices are based on vertical stacks of thin films to achieve the nanometer scaled distances necessary for resistive switching phenomena to occur [ 6 , 17 – 28 ]. Whereas designing memristive components in a lateral geometry makes the active interfaces on one hand accessible for investigation by surface sensitive or imaging methods such as electron microscopy and on the other hand allows them to be reached by external stimuli e.g. for surface plasmon resonance excitation [ 29 – 31 ] or for opening the path towards integrating sensor-features into memristive systems. However, while the layer thickness in a vertical sandwich structure can be precisely controlled by well-established deposition methods, obtaining nanoscaled distances laterally requires sophisticated and time-consuming techniques like electron beam lithography or focused ion beam deposition [ 10 , 32 ]. In the scope of this work, we present a facile and scalable fabrication route for sparse CNT networks implanted with AgAu nanoparticles (in the following termed CNT/AgAu networks) in a lateral geometry, reaching the nanometer scale required for resistive switching by a combination of three length scales, as indicated in Fig 1a : Electrodes are fabricated with standard ultraviolet (UV) lithography to provide a spacing in the micrometer range (6–8 μm). The sparse CNT network provides gaps between the CNTs of up to a few hundreds of nanometers. The fabrication process uses a custom-made CNT dispersion circumventing the detrimental effects of the additives of commercially available CNT dispersions as well as allowing for a quick spin coating deposition method by using a volatile solvent. Finally, the AgAu nanoparticles, sputter deposited with a gas aggregation source (GAS) [ 33 ], yield spacings that reach the lower nanometer range. The nanoparticles implanted into gaps provided by the CNTs act as silver ion reservoirs for ECM-type switching, as illustrated in Fig 1b . The switching behaviour is a hybrid of diffusive switching with a retention time in the second- to sub-minute-scale and bipolar switching as it is possible to reset to the high resistive state (HRS) by applying reverse voltages. The retention time of a memristive system describes the time it is able to retain its resistive state, most commonly the low resistive state (LRS) [ 34 ]. For non-volatile memory applications the retention time is required to be as high as possible to prevent data loss. However, the CNT/AgAu networks with their short retention time are useful for implementing a “short-term-memory” in neuromorphic circuits. Short-term memories are efficient for storing information that becomes deprecated quickly, as the information does not have to be removed explicitly, and save power by automatically returning to a HRS. Additionally, the capability of the CNT/AgAu networks for explicit reset retains the flexibility of a traditional bipolar memory cell. 10.1371/journal.pone.0264846.g001 Fig 1 Schematic illustration of the key features and switching mechanism of the CNT/AgAu. a) Vital components of a CNT/AgAu network from left to right: The inert electrodes, the sparse CNT network and the AgAu nanoparticles inside a nanogap between two individual CNTs. b) The switching mechanism between two NPs when exposed to a potential U. LRS = Low resistive state, HRS = High resistive state. In the following sections of this work, first the fabrication route for the CNT/AgAu networks is presented. The challenges and applied methods at each fabrication step are discussed starting with the custom-made CNT dispersion. The discussion of the nanoparticle deposition includes the in-operando percolation measurement showing the sequential usage of three length scales to obtain the nanoscaled distances necessary for resistive switching. Afterwards, the results of the morphological characterization by means of scanning electron microscopy (SEM) are presented, revealing, that the fabrication route yielded an underpercolated network of CNTs and nanoparticles. In the following, the electrical characterizations are presented consisting of three different measurement modes: Current-voltage cycles showing the distinct high resistive and low resistive states. Current-voltage cycles into reverse voltage regimes showing the capability for voltage induced reset. Time-resolved current measurements showing the short-term memory effect of the decelerated diffusive switching behaviour and a positive correlation of the retention time to the duration of the Set voltage pulse. Additionally, the data indicates that the CNT network acts as an integrated serial resistance limiting the current flow to the nA to μA regime without additional external circuitry [ 16 ] while also potentially being operable at low-voltages, resulting in a low power consumption. Finally, the proposed switching mechanism and retention are discussed with respect to filament formation and lifetime.",
"discussion": "Results and discussion Most resistive switching phenomena require nanometer scaled distances due to the resulting strong electrical fields acting as a driving force for the respective switching mechanism [ 17 , 27 , 28 ]. While there are reliable and scalable methods to achieve these distances in vertical orientation by the deposition of thin films, the available methods to obtain this in a lateral orientation are time-consuming and inscalable [ 10 , 32 ]. We developed a new method for obtaining nanometer scaled gaps with a network of CNTs deposited on a substrate with patterned gold electrodes. This CNT network must meet certain requirements: The network must fill the space between the electrodes. The CNTs must be finely dispersed, so that there are no dense agglomerations of CNTs. The network must be just below the percolation point, so that the distance between individual tubes is in the nanometer range. The CNTs must not be heavily coated by surfactants or other additives, as that would inhibit to remove short-circuiting paths by Joule heating. For the deposition of CNT networks a custom made CNT dispersion has been prepared by mixing the following ingredients: Pristine CNTs as dry powder, so that they are not coated with additives initially Ethanol as a fast evaporating solvent to facilitate a quick sequential application of dispersion droplets onto the substrate during spin coating Poly(3,4-ethylenedioxythiopene) polystyrene sulfonate (PEDOT:PSS) as an additive, keeping the CNTs finely dispersed [ 35 ] The mixture has been sonicated with an ultrasonicator to break up the CNT particles and disperse them. The PEDOT:PSS prevents re-agglomeration without impeding the resistive heating step. A thin film of dispersion has been deposited onto the substrates via spin coating. The dispersion was dropped onto the substrate sequentially dropwise during spinning, where each drop has been left to evaporate before applying the next one, allowing for precise control over the particle density. After deposition of a CNT network a voltage ramp has been applied to it to remove any continuous CNT paths short-circuiting the electrodes by resistive heating, which has been indicated by a sudden drop in the current readout (see S1 Fig ). The AgAu nanoparticles have been deposited by direct current (DC) magnetron sputtering using a Haberland-type gas aggregation source (GAS) identical to the one reported in [ 36 ], with a AgAu target as in [ 33 ] attached to the DC planar magnetron source. This deposition method enables precise control over particle composition and density without affecting the CNT network on the substrate [ 33 ]. The deposition time for the nanoparticles has been set to stay below the percolation point [ 37 ]. The deposition time for the percolation point has been determined by performing electrical measurements in-operando during deposition. A schematic of the setup and the percolation measurement of a substrate with deposited CNT network is shown in Fig 2 . After deposition of AgAu nanoparticles, a thin layer of SiN has been deposited on top as a protective layer without breaking vacuum. Experimental details about any step of the procedure can be found in the section “Materials & Methods”. 10.1371/journal.pone.0264846.g002 Fig 2 Percolation measurement for AgAu nanoparticle deposition. a) Schematic of the in-operando percolation measurement setup. b) Time-resolved current measurement across adjacent electrodes at a voltage of 3V. The time where the flowing current shows a significant increase is taken as the percolation time (337 s). The deposition has been stopped at the percolation point. The red dashed line indicates the progression of the current, if the deposition had continued, leading to an overpercolated layer of nanoparticles. Fig 3 shows a sequence of SEM images of a CNT/AgAu network revealing the homogeneous distribution and sparseness of the CNTs. A substantial fraction of CNTs have been broken into smaller pieces during the ultrasonication step, which we assume to provide two advantages for the fabrication process: Firstly, an alleviation of entanglement and agglomeration of CNTs and secondly an increased sparseness of the network providing more gaps between CNTs and preventing individual tubes from bridging the whole network. 10.1371/journal.pone.0264846.g003 Fig 3 SEM micrographs of a finished [CNT/AgAu network] without SiN layer. a+b) Homogeneous sparse CNT network between the electrodes. c) A nanogap between two CNTs with deposited AgAu NP. The samples shown in the images have not been coated with SiN. Fig 3c shows a CNT network with deposited underpercolated AgAu nanoparticles. The particle distribution shows spacing in the lower nanometer range enabling ECM-type memristive switching between nanoparticles [ 16 ]. Samples solely prepared with AgAu nanoparticles with an equal filling factor showed no switching behaviour in the considered voltage regime (see S2 Fig ) verifying that the sparse network of CNTs fulfils its expected functionality of providing suitable nanogaps to enable the resistive switching of the nanoparticles. For CNT/AgAu networks, that did not show resistive switching below 10 V, an electric preforming step has been performed by cycling to a voltage of ± 20 V (0 V -> 20 V -> -20 V -> 0 V) over several cycles (see S3 Fig ). A stable HRS corresponding to Fig 1b is reached during the second cycle, from where on the resistive switching occurs. This indicates that nanoparticle gaps in the conduction path become persistent conductive elements by forming stable filaments that are not collapsing, when the electrical field is removed. This yields lower switching voltages by decreasing the number of gaps over which the overall voltage drops. After the preforming procedure, when cycled in low voltage regimes, operation becomes stable as shown in Fig 4 . 10.1371/journal.pone.0264846.g004 Fig 4 Resistive switching behaviour. a) Cyclic voltage pattern, showing resistive switching behaviour and an ON/OFF ratio of around 81 at a Read voltage of 1.5 V. b) HRS and LRS currents at Read voltage of 1.5 V for each cycle. Fig 4a shows a cyclic IV-measurement of a CNT/AgAu network with an 8 μm spacing between electrodes. The voltage has been cycled from 0 V to 2.5 V and back to 0 V at a rate of 250 mV/s over ten cycles, where after each cycle the voltage has been held at 0 V for 5 s. The current response shows a clear distinction of two resistive states (see Fig 4b ) and stable operation across the ten cycles with the HRS current being 1.9 ± 0.8 nA and the LRS current being 155 ± 55 nA yielding a current ratio between HRS and LRS of ≈ 81 at a Read voltage of 1.5 V. The data indicates the potential for low-voltage operation and shows that the network itself acts as a series resistance limiting the current flow and thus ensuring low power consumption without additional circuitry (cf. [ 16 ]). Fig 5 shows two subsequent cycles of a CNT/AgAu network, recorded at a ramp speed of 25 mV/s. The cycle shown in Fig 5a starts in an HRS, switching into the LRS and retaining it, until a negative voltage of -1.5 V is applied. The CNT/AgAu network then switches into its LRS again upon reaching its Set voltage of 2.5 V. The subsequent cycle seen in Fig 5b starts in the LRS and shows that this effect is symmetrical. Reducing the ramp speed of 10 mV/s did not change the reset voltage, indicating that the effect is not accountable to the diffusive reset behaviour. These measurements indicate that, while the CNT/AgAu network also returns to its HRS over time without the application of a voltage like a diffusive device, it shows the ability of bipolar switching i.e. to be reset from LRS to HRS by reverse voltages. 10.1371/journal.pone.0264846.g005 Fig 5 Reset behaviour with reverse voltage. a) Cycle starting in HRS. The LRS is retained until -1.5 V is applied. b) Subsequent cycle starting in LRS. The reset behaviour is symmetrical in the positive and negative voltage range. The numbers indicate in which order the resistive switching occurred. The time-dependent retention has been investigated by time resolved current measurements with the voltage pattern shown in the upper graph of Fig 6 . The measurement starts at 0 V for several seconds to serve as a reference. The current readout at the Read voltage of 0.5 V before the first Set pulse shows that the CNT/AgAu network is in its HRS initially. After each Set pulse, each with different pulse durations, the current readouts at the subsequent Read voltage indicate a switch of the CNT/AgAu network into the LRS. The Read voltage has been held until it returned into its HRS, where the time from returning to the Read voltage until the current returned to the HRS regime has been taken as the retention time (see Fig 6 ). The series of pulse durations for the Set pulses indicates a positive correlation between the time in the Set state and the resulting retention time of the CNT/AgAu network i.e. a longer Set pulse results in a longer retention time (see also S4 Fig ). 10.1371/journal.pone.0264846.g006 Fig 6 Time-resolved current measurements showcasing the retention. The upper graph shows the applied voltage pattern: Read voltage = 0.5 V, Set voltage = 5.5 V. The retention time is taken as the time from returning from Set to Read voltage to when the current reaches the HRS current. The numeric values indicate the duration of the Set voltage pulse. The Set time and retention time show a positive correlation, which can be found in the (see S4 Fig ). In ECM devices the resistive states originate from metal filaments formed by movement and reduction of metal ions in the electrical field [ 12 , 16 , 38 , 39 ]. With a formed filament the device is in its LRS. When the filament ruptures, it returns to its HRS, which is due to surface tension of the filament as a restoring force, making it energetically favourable to form spherical particles depending on the thickness of the filament and the surrounding matrix material [ 7 , 40 ]. In bipolar memristive ECM devices the electrodes provide a metal reservoir large enough to form stable filaments with a retention of several years [ 40 ]. The CNT/AgAu networks use the AgAu nanoparticles as metal ion reservoirs limiting the amount of silver atoms available for filament formation. Singular or few nanoparticles have been reported to show no retention due to the formed filament being thin enough to immediately collapse as soon as the electrical field as a driving force is removed [ 16 ]. In this work, however, the density of deposited AgAu nanoparticles in the CNT networks’ nanogaps yield a suitable silver reservoir to provide a substantial amount of silver ions for filament formation. The resulting filament in the gaps, where resistive switching occurs, is thick enough to be stable, so that the rupture does not happen right away without electrical field, but instead is delayed until diffusion thinned down the filament enough to collapse by surface tension. At the same time though the amount of silver atoms is still limited such that the filament formed by an electrical field is not outright long-term stable, like in a typical bipolar device. Instead the filament formed initially is sufficiently thin to show diffusive switching. However, with prolonged application of the electrical field, the filament grows by material diffusion from other nearby nanoparticles. The longer the Set voltage is applied, the thicker the filament becomes. Thus, the retention time increases i.e. the time it takes for the filament to collapse after removing the electrical field. As long as the filament holds, a reverse voltage excitation is able to break the filament. In typical bipolar ECM type systems the electrode materials are asymmetrical, so that only one voltage polarity yields a filament while the reverse polarity leads to its dissolution [ 12 , 14 , 38 ]. As seen in Fig 5b however, it is evident, that the behaviour of the CNT/AgAu network is symmetrical. The reset mechanism is presumed to be based on drift of the filament’s silver towards the cathode until the filament is sufficiently thinned down on the anode side to collapse, switching the CNT/AgAu network into its HRS."
} | 5,575 |
30842458 | PMC6403400 | pmc | 370 | {
"abstract": "Depth from defocus is an important mechanism that enables vision systems to perceive depth. While machine vision has developed several algorithms to estimate depth from the amount of defocus present at the focal plane, existing techniques are slow, energy demanding and mainly relying on numerous acquisitions and massive amounts of filtering operations on the pixels’ absolute luminance value. Recent advances in neuromorphic engineering allow an alternative to this problem, with the use of event-based silicon retinas and neural processing devices inspired by the organizing principles of the brain. In this paper, we present a low power, compact and computationally inexpensive setup to estimate depth in a 3D scene in real time at high rates that can be directly implemented with massively parallel, compact, low-latency and low-power neuromorphic engineering devices. Exploiting the high temporal resolution of the event-based silicon retina, we are able to extract depth at 100 Hz for a power budget lower than a 200 mW (10 mW for the camera, 90 mW for the liquid lens and ~100 mW for the computation). We validate the model with experimental results, highlighting features that are consistent with both computational neuroscience and recent findings in the retina physiology. We demonstrate its efficiency with a prototype of a neuromorphic hardware system and provide testable predictions on the role of spike-based representations and temporal dynamics in biological depth from defocus experiments reported in the literature.",
"introduction": "Introduction The complexity of eyes’ inner structure implies that any visual stimuli from natural scenes contains a wide range of visual information, including defocus. Several studies have shown that defocus is essential in completing some tasks and more specifically for depth estimation 1 , 2 . Although a large body of research on Depth From Defocus (DFD) exists since the early 60’s, there is currently a gap between the information output from biological retinas and the existing literature both in the vision science and computer vision that uses images as the sole source of their studies. Although images are perfect to display static information, their use in acquiring dynamic contents of scenes is far from being optimal. The use of images implies a stroboscopic acquisition of visual information (unknown to biological systems) at a low sampling frequency. They are thus unable to describe the full dynamics of observed scenes. On the other hand, retinal outputs are massively parallel and data-driven: ganglion cells of biological retinas fire asynchronously according to the information measured in the scene 3 , 4 at millisecond precision. Recent neuroscience findings show that this temporal precision can also be found in other subcortical areas, like the lateral geniculate nucleus (LGN) 5 , 6 and the visual cortex 7 . The last decade has seen a paradigm shift in neural coding. It is now widely accepted that precise timing of spikes open new profound implications on the nature of neural computation 8 , 9 . The information encoded in the precise timing of spikes allows neurons to perform computation with a single spike per neuron 10 . Initially supported by theoretical studies 11 , this hypothesis has been later confirmed by experimental investigations 12 , 13 . Here, we present a novel approach to the depth from defocus, inspired by biological retina ouput, which is compatible with ultra low latency and low power neuromorphic hardware technologies 14 . In particular, we exploit advances made in both mixed signal Analog/Digital VLSI technology and computational neuroscience which enabled us to combine a new class of retina-like artificial vision sensors with brain-inspired spiking neural processing devices to build sophisticated real-time event-based visual processing systems 15 – 17 . We show how precise timing of spiking retinas allows the introduction of a novel, fast and reliable biologically plausible solution to the problem of estimating depth from defocus directly from the high temporal properties of spikes. Silicon retinas located at the core of the hereby presented system are a novel piece of hardware which do not sense scenes as a serie of frames. Conventional cameras wastefully record entire images at fixed frame rates(30–60 Hz) that are too slow to match the temporal sub-millisecond resolution of human senses. Silicon retinas are asynchronous and clock-less, every pixel is independent from its neighbors and only reacts to changes caused by movements in a scene. Data are transmitted immediately and are scene driven, resulting in a stream of events with a microsecond time precision equivalent to conventional high-speed vision sensors, with the addition of being low power and sparse 18 . This type of acquisition increases the sensor dynamic range and reduces power computation. Spiking Neural Networks (SNNs 19 ) are computational models using neural stimulation. It has been shown that such networks are able to solve constraint satisfaction problems 20 , 21 , depth extraction from stereovision 22 , 23 or flow computation 24 , 25 . As they are mimicking real neurons behavior, they allow a massively parallel, low power calculation, which is highly suitable for embedded computation. The use of a SNN in this work is a natural choice to build a complete neuromorphic event-based system, from the signal acquisition to the final output of the depth information. This is advantageous because of the resulting low-power system promised by the spiking/neuromorphic technology. The developed architecture is particularly adapted on a variety of existing neuromorphic spiking chips such as the SpiNNaker 26 , TrueNorth 27 or LOIHI 28 neural chips. More specific neuromorphic hardware, such as the 256 neurons ROLLS chip 29 , can also be used. When combined with an event-based camera, power as low as 100 mW is proven to be sufficient to achieve a realtime optical flow computation 25 . We are showing with this work that a low-power (≤100 mW), computationally inexpensive and realtime DFD system can be similarly achieved. Among the multitude of techniques developed by vision scientists to estimate depth, those called depth from focus (DFF) or depth from defocus (DFD) have the great advantage of requiring only a monocular camera 30 . The DFF method uses many images, and depth clues are obtained from the sharpness at each pixel. This method is computationally expensive and the amount of data to process is substantial. On the other hand, DFD estimates the variance of spatially varying blur spots based on a physical model. This technique requires less images but at the cost of a greater error in positioning. Current methods that use DFD or DFF generate depth maps for static scenes only 31 as they are limited by the frame rate of the camera driven at maximum of 25 fps. The computer vision and engineering community have described a number of algorithms for defocus computation 32 – 34 . However, they typically require multiple concurrent images 35 – 37 , lightfield systems 38 , specific lens apertures 35 , 39 , correlations 40 , specific hardware 41 or light with known patterns projected onto the environment 37 . The use of images and luminance implies high computational costs of around 17 ms to process a single frame 40 . These approaches cannot serve as conceivable models of defocus estimation in natural visual systems, as mammalian usually operate on a complete different data format and acquisition principles. Early studies 42 , 43 show that the border between blurred and sharp regions can be used to establish the depth-order of objects. For example, an out-of-focus target with a blurry textured region and a blurry border was perceived to be located proximal to the plane of focus, while an out-of-focus target with a blurry region and a sharp border was perceived to be located distant to the plane of focus. Recent findings in neuroscience show that blur perception in human is a dynamic process that allows depth assessment. In particular, the retinal defocus blur provides information regarding the relative and/or absolute distance of objects in the visual field 44 . Recently 45 , it has been demonstrated that subjects were able to detect the relative distance of two vertical edges, justifying that the retinal blur allowed the subjects to judge target distance deferentially without any other depth cues. Other studies demonstrated that motor efference and/or sensory feedback related to the blur-driven accommodative response contain sufficient information to estimate the absolute distance of visual targets 46 . In addition, information derived from image blur can be integrated by the visual system with other visual cues (e.g., retinal disparity, size, interposition, etc.), which would assist in enabling one to judge the depth order of objects over a range of distances 43 , 47 – 50 . The addition of blur information can improve the speed and accuracy in such a depth-ordering task 51 .",
"discussion": "Conclusions and Discussions In this paper we proposed a spiking neural network model that solves the depth from focus efficiently by exploiting an event-based representation amenable to neuromorphic hardware implementations. The network operates on visual data in the form of asynchronous events produced by a neuromorphic silicon retina. It processes these address-events in a data-driven manner using artificial spiking neurons computation units. This work introduces a valid explanation and a robust solution to depth estimation from defocus that has not been reported in the literature. The overall system matches recent existing literature of neuroscience, biological retinas and psychophysics studies on the role of defocus in the visual system. This network is nonetheless an abstract simplification of the depth estimation problem that must surely combine more complex information in biological systems. More importantly, this study should be coined depth from focus rather than from defocus as the neural structure developed aims at detecting the exact time of focus during a sweep. During the last five decades of research, DFD has remained an unsolved issue. The fundamental difference and novelty of this work is that the proposed network operates using exclusively precisely-timed contrast events. These events are measured directly from the neuromorphic silicon retina, which models only the transient responses of retinal cells (i.e., of the Y-ganglion cells), without including the sustained ones, yet present in the system. While the sustained information is present in the silicon retina used, we show that this information is not necessary to provide depth estimation from defocus. Silicon retina transient responses produce single events. Their precise timing plays a crucial role in the estimation of blur and more importantly in determining when the observed object is in focus. In contrast, the vast majority of computational models of depth from defocus are based on images that are known to be absent from the visual system and only rely on luminance information. Additionally, none of them use the precise timing of spikes. In these models, convolutions techniques are used to determine the level of blur. These methods are computationally expensive and meaningfully slower as several acquisitions are often needed to provide an accurate result. By contrast, the model we presented does not incorporate any notion of filtering or convolutions. These choices are based on the perception of spatial contrast, whereas the presented model solely responds to temporal contrast. Whether the brain is using such a technique to estimate depth from defocus is an open question. However due to the nature of precisely timed information output by biological retinas 64 convolutions algorithms cannot provide a viable explanation as the stroboscopic nature of image acquisition and luminance use is incompatible with neural systems. Instead, we show that the change of polarity at the pixel level contains sufficient information to estimate depth from defocus. Recent findings in physiology show that several mechanisms used by our methodology exist in Nature. Biological retinas contain several types of ganglion cells, each informing the brain about a particular content of the visual scene, such as motion, edges or chromatic composition. In a recent paper, a newly discovered ganglion cell type ‘On-delayed’ is described 65 . This cell has been shown to respond vigorously to increasing blur. Its degree of firing directly encodes the amount of high spatial frequencies contained in its receptive field. More importantly, this cell gets input from both ON and OFF polarities. While it is currently unknown how this defocus information is used by the brain, it is most likely that this information projects to the visual thalamus and cortex and also to midbrain structures where accommodation is controlled 66 . We expect the most significant impact of our model to be in the field of artificial vision. Today’s machine vision processing systems face severe limitations imposed both by the conventional sensors front-ends (which produce very large amounts of data with fixed sampled frame-rates), and the classical Von Neumann computing architectures (which are affected by the memory bottleneck and require high power and high bandwidths to process continuous streams of images). The emerging field of neuromorphic engineering has produced efficient event-based sensors, that produce low-bandwidth data in continuous time, and powerful parallel computing architectures, that have co-localized memory and computation and can carry out low-latency event-based processing. This technology promises to solve many of the problems associated with conventional computer vision systems. However, the progress so far has been chiefly technological, whereas related development of event-based models and signal processing algorithms has been comparatively lacking (with a few notable exceptions). This work elaborates on an innovative model that can fully exploit the features of event-based visual sensors. In addition, the model can be directly mapped onto existing neuromorphic processing architectures. Results show that the full potential is leveraged when single neurons from the neural network are individually emulated in parallel. In order to emulate the full-scale network, however, efficient neuromorphic hardware device capable of emulating large-scale neural networks are required. The developed architecture requires few neurons per pixel and is implementable on a variety of existing neuromorphic spiking chips such as the SpiNNaker 26 , TrueNorth 27 or LOIHI 28 neural chips."
} | 3,685 |
27774375 | null | s2 | 372 | {
"abstract": "Hydration is central to mitigating surface fouling by oil and microorganisms. Immobilization of hydrophilic polymers on surfaces promotes retention of water and a reduction of direct interactions with potential foulants. While conventional surface modification techniques are surface-specific, mussel-inspired adhesives based on dopamine effectively coat many types of surfaces and thus hold potential as a universal solution to surface modification. Here, we describe a facile, one-step surface modification strategy that affords hydrophilic, and underwater superoleophobic, coatings by the simultaneous deposition of polydopamine (PDA) with poly(methacryloyloxyethyl phosphorylcholine) (polyMPC). The resultant composite coating features enhanced hydrophilicity (i.e., water contact angle of ~10° in air) and antifouling performance relative to PDA coatings. PolyMPC affords control over coating thickness and surface roughness, and results in a nearly 10 fold reduction in "
} | 244 |
36844511 | PMC9948195 | pmc | 373 | {
"abstract": "In this study, we used a simple and cost-effective method\nto fabricate\ntriboelectric nanogenerators (TENGs) based on biowaste eggshell membranes\n(EMs). We prepared stretchable electrodes with various types of EMs\n(hen, duck, goose, and ostrich) and employed them as positive friction\nmaterials for bio-TENGs. A comparison of the electrical properties\nof the hen, duck, goose, and ostrich EMs revealed that the output\nvoltage of the ostrich EM could reach up to 300 V, due to its abundant\nfunctional groups, natural fiber structure, high surface roughness,\nhigh surface charge, and high dielectric constant. The output power\nof the resulting device reached 0.18 mW, sufficient to power 250 red\nlight-emitting diodes simultaneously, as well as a digital watch.\nThis device also displayed good durability when subjected to 9000\ncycles at 30 N at a frequency of 3 Hz. Furthermore, we designed an\nostrich EM-TENG as a smart sensor for the detection of body motion,\nincluding leg movement and the pressing of different numbers of fingers.",
"conclusion": "Conclusions We have investigated four types of EMs\nas friction materials with\nstretchable electrodes. Among them, the ostrich EM displayed the highest\nelectrical output due to its abundance of functional groups, natural\nfiber structure, high surface roughness, high surface charge, and\nhigh dielectric constant. The optimized and fabricated ostrich EM-TENG\nexhibited output powers of up to 0.18 mW, sufficient to operate 250\nred LEDs instantaneously. Furthermore, this device could generate\npower effectively under various external resistance loads. The output\nvoltage of the EM-TENG was stable during long-term cycling. Furthermore,\nwe found that the ostrich EM-TENG could be used as a smart sensor,\nwith applications in a wide range of wearable electronics: for powering\nvarious small-scale electronic devices, in biomedical monitoring systems,\nor for sensing mechanical motions. The electrospinning method is a\nnanofiber fabrication technology. Other varieties of triboelectric\nnanofiber materials could be fabricated using the electrospinning\nmethod.",
"introduction": "Introduction The drive to replace traditional fossil\nfuels with sustainable\nand renewable energy sources has grown rapidly in the past decade,\ndue to possible energy crises and higher standards for environmental\nprotection. As a consequence, natural and green energy sources, including\nsolar, wind, and water, have attracted much research interest in the\npursuit of sustainable and renewable energy. 1 − 4 At the same time, with rapid progress\nin the Internet of Things (IoT), smart and portable electronic products\nhave become essential parts of our daily lives, leading to a steadily\ngrowing demand for independent power sources detached from the power\ngrid. 5 Nevertheless, accommodating conventional\nrigidly structured portable power sources without the self-charging\nability (particularly batteries and supercapacitors) limits the potential\ndesign space for electronic devices. Hence, there is a great desire\nto develop sustainable and flexible energy harvesting technologies\nthat can convert ambient energy into valuable electricity for use\nas green power supplies for next-generation electronics. A popular\ncandidate for a portable energy source that harvests\nenergy from the environment is the triboelectric nanogenerator (TENG),\nfirst introduced by Wang et al. TENGs have many attractive features,\nincluding low fabrication costs, lightweight, high energy efficiency,\nand a wide range of materials and structures to choose from when designing\na device, 7 − 9 leading to a large variety of viable device setups. 10 − 13 Among other factors, the contact area between the triboelectric\nlayers plays a major role in determining the properties of a TENG. 14 − 19 Previous studies have focused on investigating the effects of material\nselection, 20 − 22 surface modification, 23 − 26 and surface morphology, 27 − 30 on the mechanism of operation and performance of TENGs. Because\nincreasing the dielectric constant can enhance the capacitance of\nthe dielectric layer, thereby increasing the surface charge density,\nthe dielectric constant of a triboelectric material is an important\nfactor affecting triboelectric performance. 31 One particular drawback of the early TENG designs is the lack\nof\nflexibility of the involved materials that would be required for applications\nin next-generation electronics, including electronic skin, flexible\nand touch screen displays, electronic watches, and biomechanical monitoring\nsensors. 50 − 54 Consequently, extensive research has been undertaken to make the\ntriboelectric layer—responsible for the generation of triboelectric\ncharges in the TENG assembly—fully stretchable and reliable\nfor wearable electronics. 54 Another challenge\nis fabricating flexible electrodes while still facilitating the transfer\nof charges to the external circuit with minimum loss. The influence\nof the mechanical flexibility of a wearable TENG on its output performance\nhas been reported. 55 , 56 Apart from inherent material\nproperties, it is also important to achieve high degrees of contact\nat the contact interface. Because of the generally insulating nature\nof dielectric tribomaterials, charge transfer from the triboelectrification\nlayer to the electrode (through electrostatic induction) in a TENG\nis of great concern. Therefore, appropriate engineering at the electrode\nsurface and friction layer–electrode interface, along with\noptimization of the combination of electrodes with appropriate work\nfunctions and triboelectric materials, is required to ensure an increased\nsurface charge density and power output. 57 When selecting materials for TENGs with applications in wearable\nconsumer electronics, there are other important aspects to consider.\nTraditional choices involved polymers, such as poly(tetrafluoroethylene)\n(PTFE), poly(vinylidene difluoride) (PVDF), polydimethylsiloxane (PDMS),\nand nylon, as friction layers, which are neither cost-effective nor\nbiodegradable. 32 − 34 Furthermore, those commercial polymers can produce\npotentially harmful chemicals, restricting their applications in biomedicine.\nHence, the development of bio-inspired, non-toxic, ultra-sensitive,\nand flexible TENGs has become a great challenge for next-generation\nbiomedical applications. 35 − 38 Although several biowaste materials,\nincluding rice husks, 22 seagrasses, 39 leaves, 40 , 41 sunflower husks, 42 peanut shell powders, 43 and fish bladders, 44 have been used to\nform the friction layers, the processes for preparing biowaste material\nfriction layers have been complicated and required the usage of harmful\nsolvents. Moreover, those biowaste materials do not have a natural\nfiber structure, thereby minimizing the electrical output of the nanogenerator.\nA promising candidate for a biodegradable material with a natural\nfiber structure would be eggshell membranes (EMs). 45 EMs have been used as cost-efficient, eco-friendly, and\nflexible materials for printed electronic applications because they\npossess large numbers of COOH and NH 2 functional groups.\nFor example, a highly efficient EM-based electrode material has been\nfabricated for supercapacitor applications. 46 , 47 In addition, EMs are capable of being piezoelectric, triboelectric,\ncapacitive, and humidity sensing due to their nanofibrous structures\nand a variety of proteins within their structures. 48 , 49 In this study, we used various types of EMs (hen, duck, goose,\nand ostrich) as positive friction materials to prepare stretchable\nelectrodes for biodegradable TENGs. Among them, the output voltage\nof the ostrich EM could reach up to 300 V, due to its abundance of\nfunctional groups, natural fiber structure, high surface roughness,\nhigh surface charge, and high dielectric constant. The output power\nof the device reached 0.18 mW, sufficient to power 250 red light-emitting\ndiodes simultaneously, as well as a digital watch. This ostrich EM-TENG\nnot only exhibited excellent electrical properties, but we also used\nit as a smart sensor to detect body motion.",
"discussion": "Results and Discussion Figure 1 a illustrates\nthe procedure for preparing the plasma-treated PDMS substrate. First,\na PDMS solution was fabricated by mixing the monomer and curing agent\n(mass ratio, 10:1) and degassing the mixture in a vacuum oven. The\nPDMS solution was then poured into a Petri dish and cured in a vacuum\noven at 60 °C for 1–2 h to obtain the PDMS substrate.\nThe surface-modified PDMS substrate was subjected to N 2 atmospheric plasma treatment to induce functional groups on the\nPDMS surface to form better connections. The N 2 plasma\npower is 18 W, and the N 2 pressure is 1.5 kgf/cm 2 . PDMS surface roughness (rms) before and after N 2 plasma\ntreatment is 0.834 and 0.289 nm, respectively. A layer of Ag was formed\non the PDMS surface through vapor deposition. A Ag layer was also\ndeposited on the surface of the untreated PDMS substrate ( i.e. , without N 2 plasma treatment) to obtain\ncorresponding Ag/PDMS substrates. The surfaces of the PDMS substrates\nwere subjected to contact angle analysis ( Figure 1 b). The surface of the untreated PDMS substrate\nexhibited a water contact angle of 106.6°. After N 2 atmospheric plasma treatment, the contact angle decreased to 7.9°,\nindicating that the surface of this PDMS substrate was hydrophilic.\nIts contact angle increased from 7.9 to 55.9° after exposure\nto an atmospheric environment for 12 h, suggesting that the content\nof oxygen-containing functional groups on the plasma-treated surface\nof the PDMS substrate decreased upon increasing the exposure time. 58 Therefore, the plasma treatment of the PDMS\nsubstrate subjected to thermal evaporation must be performed within\n1 h to avoid such a decrease in the availability of oxygen-containing\nfunctional groups. XPS revealed that the C 1s and O 1s peaks shifted\nafter N 2 plasma treatment, indicating a strong connection\nbetween the Ag layer and PDMS ( Figure 1 c,d). We used SEM to investigate the morphologies of\nthe Ag/PDMS substrates with and without plasma treatment ( Figure 1 e,f, respectively).\nThe images revealed that relatively continuous 50-nm-thick Ag nanoparticle\nlayers had been deposited on the PDMS substrates. For the unmodified\nPDMS substrate ( Figure 1 e), the surface morphology of the Ag nanoparticle layer featured\nisolated nanoparticles, potentially resulting in poor conductivity.\nAfter N 2 atmospheric plasma treatment, however, the morphology\nof the Ag nanoparticle layer changed to a relatively continuous layer\n( Figure 1 f). Figure 1S (Supporting Information) presents SEM\nimages of the Ag films deposited on PDMS substrates that had been\nsubjected to various durations of N 2 plasma treatment. Figure 2 a–d\npresents the surface morphologies of the various types of EMs. Among\nthem, the hen EM exhibited the highest roughness. The three-dimensional\n(3D) surface topographical image revealed that the surface of the\nhen EM was not flat, potentially resulting in lower electrical output\ndue to poor draping with the friction layer. In contrast, the ostrich\nEM had greater roughness (RMS: 0.328 μm) but was a flat film\nthat provided a large interspace and a high specific surface area,\nleading to efficient contact under pressure; moreover, it features\nunique and uniform particles on the fiber. Figure S2 displays SEM images of the various types of EMs, revealing\ntheir natural fiber structures with porosity on the contact surface,\npotentially helping to store more electrons in the pores, thereby\nprolonging the electrical output. We also used Fourier transform infrared\n(FTIR) spectroscopy to analyze the structures of the various types\nof EMs ( Figure S3 ). The collagen fibrils\nin EMs are structurally stable because of their regular chemical bonding\nand crystallinity. Figure 2 e provides the schematic configuration of the EM-TENG device.\nFor its fabrication, an EM was taken with a desired square shape (2.0\ncm × 2.0 cm). One side of the selected EM portion was used as\nthe positive tribomaterial, and the other was attached to a stretchable\nAg electrode, which was used as the conducting electrode. We used\na polyimide (PI) film as the negative tribomaterial, with conductive\naluminum (Al) tape covering the PI surface as the electrode. Figure 2 f presents an overview\nof the working principle of an EM-TENG in vertical contact–separation\nmode, involving a combination of triboelectrification and electrostatic\ninduction. Prior to contact, no electron flow occurs in the separated\nmodel. Once an external force is applied to bring the EM and Kapton\ninto contact, surface charge transfer occurs at the interface due\nto a triboelectric effect. The direction of the charge transfer is\ndetermined by the relative triboelectric polarity of the two layers.\nIn this study, we used Kapton as the negative triboelectric material\ndue to its triboelectric polarity. When the external force is released,\nthe EM and Kapton surfaces become separated. At this stage, the separation\nof the surface charges leads to an increasingly strong dipole moment,\ncreating an electric potential difference between the electrodes.\nAs a consequence, electrons begin to flow from a negative to positive\npotential, with charges accumulating on the electrodes, resulting\nin a positive electrical signal. Several physical properties of triboelectric\nmaterials—in particular, their surface roughness, electron\naffinity, friction, and dielectric constant—affect the performance\nof TENGs. Among them, a high dielectric constant is the most important\nproperty for improving the output performance of the devices, measured\nin terms of their output voltages and currents. 6 , 27 Figure 2 (a–d)\nSEM images of various types of EMs: (a) hen, (b) duck,\n(c) goose, and (d) ostrich. Insets: 3D views of the surface topographies.\n(e, f) Schematic representations of (e) the EM-TENG device and (f)\nthe working mechanism of the EM contact-mode TENG. We measured the electrical properties of the ostrich\nEM-TENG while\nvarying the operating frequency, applied force, and external load. Figure 3 a,b displays the\nelectrical output voltages of the ostrich EM at various frequencies\nand mechanical forces, respectively. When the frequency was varied\nfrom 1 to 3 Hz under an applied force of 30 N, the peak output voltage\nfor the ostrich EM mats increased from approximately 280 V to approximately\n300 V ( Figure 3 a).\nIncreasing the operating frequency resulted in more intense friction,\nwhich generated more charges due to faster contact between the ostrich\nEM and the Kapton layer. Nevertheless, the output voltage became unstable\nat frequencies greater than 5–10 Hz. This instability arose\nbecause, at these high frequencies, the contact and separation processes\nof the triboelectric layers were incomplete, preventing the surface\ncharge from reaching its maximum value. With a fixed operating frequency\nof 3 Hz, the output voltage of the ostrich EM increased from approximately\n150 V to approximately 300 V upon increasing the applied force from\n10 to 40 N. This behavior presumably resulted from an increased compressive\nforce, leading to significantly improved contact between the triboelectric\nlayers, thereby resulting in the generation of more electric charges.\nThe output voltage was almost saturated, however, when the compressive\nforce was greater than 30 N. Thus, the optimized conditions for subsequent\nstudies involved a cycled compressive force of 30 N at an applied\nfrequency of 3 Hz. Figure S4 provides the\nelectrical output currents of the ostrich EM measured at various frequencies\nand mechanical forces. Figure 3 c displays the maximum output performance data of the various\nEMs under a cycled compressive force of 30 N at an applied frequency\nof 3 Hz. The output voltages of the hen, duck, goose, and ostrich\nEMs were approximately 250, 150, 200, and 300 V, respectively, under\nthe same mechanical force. The output current density of the ostrich\nEM reached up to approximately 0.6 μA/cm 2 , higher\nthan those of the duck and goose EMs ( Figure 3 d). The combination of the open-circuit voltage\nand short-circuit current led to the maximal power of the ostrich\nEM (18 mW) being greater than those achievable for the hen, duck,\nand goose EMs alone (17.5, 4.5, and 8.0 mW, respectively). Furthermore, Figure 3 e,f presents the\nmeasured voltage outputs, current outputs, and power densities generated\nby the ostrich EM-TENG under various external load resistances when\noperated at 3 Hz and 40 N. The output voltage increased upon increasing\nthe external load resistance from 470 Ω to 1 MΩ, with\nthe output current decreasing until the external load resistance reached\n10 MΩ. Consequently, the ostrich EM-TENG exhibited a maximum\noutput power density of 270 μW/cm 2 at a resistance\nof 10 MΩ. Figure 5S provides the\nmeasured voltage outputs, current outputs, and power densities generated\nby the hen, duck, and goose EM-TENG under various external load resistances.\nCompared with the hen EM-TENG, the power density of the ostrich EM-TENG\nwas 1.1 times higher, due to its higher surface roughness, surface\ncharge, surface potential, and dielectric constant. Figure 3 (a, b) Electrical output\nvoltages of the various EMs measured at\nvarious (a) frequencies and (b) mechanical forces. (c) Maximum output\nvoltages and (d) currents of the ostrich EM-TENG, measured at 30 N\nand 5 Hz. (e) Output voltage and current density, and (f) power density\nof the ostrich EM-TENG, plotted with respect to resistance. Several studies have revealed that the surface\ncharge density can\nbe enhanced by incorporating microscopic structures or engineering\nfunctional groups onto the surfaces of triboelectric materials. 27 , 57 , 59 , 60 Figure 4 a reveals\nthe amounts of charge transferred for the various types of EMs. Among\nthem, the ostrich EM possessed the highest amount of transferred charge\ndue to its good contact area and surface roughness (as determined\nusing SEM and AFM), thereby generating more electrons through friction\nwhen compared with the other various EMs. Moreover, the maximum transferred\ncharge in the ostrich EM suggests improvements in the capture and\nstorage of the triboelectric electrons, thereby enhancing the electrical\noutput. The initial surface potentials of the hen, duck, goose, and\nostrich EMs were approximately 4.5, 4.7, 3.0, and 9.5 kV, respectively.\nAmong them, the ostrich EM had the highest initial surface potential,\nindicating that its highest surface charge would result in the highest\nelectrical output. According to the relationship between the transferred\ncharge and the dynamic capacitance of a TENG, the first derivative\nof the charge with respect to time can be expressed as 1 Equation 1 indicates that the rate at which the capacitance changes\nwith respect to time is a critical parameter. Hence, the output performance\ncan be optimized by changing the dielectric film’s compressibility,\nfor example, by incorporating porous sponge structures into triboelectric\nmaterials. Figure 4 c plots the dielectric constants for the various eggshell membranes\nwith respect to frequency. All EMs exhibited a dielectric constant\nof approximately 3 at a frequency of 10 kHz. Nevertheless, the overall\ndielectric constant of the EM was small because of its porous nature.\nThe dielectric constant of each of the four EMs decreased upon increasing\nthe frequency, possibly due to a polarization effect. When the frequency\nbecomes too high, reorientation of the dipoles becomes impossible,\ncausing the dielectric constant to decrease toward 1.0, close to the\ndielectric constant of air. 61 As displayed\nin Figure 4 c, the dielectric\nconstant of the ostrich EM was higher than those of the others, due\nto a lower volume of pores ( Table 1 ) and a higher amount of transferred charge. The surface\ncharge densities and dielectric constants of triboelectric materials\nare known to impact the output performance of TENGs. 59 , 60 In this study of four different kinds of EM-TENGs, we used the finite-element\nsimulation tool (COMSOL multiphysics) to calculate the electric potentials\nof the EM-TENGs ( Figure 4 d). The distance between the two triboelectric materials was set\nat 3 mm. For the hen, duck, goose, and ostrich EM-TENGs, the measured\ntriboelectric charges were 0.80, 0.65, 0.65, and 1.17 mC/m 2 , respectively, with dielectric constants of 2.81, 2.15, 2.57, and\n3.05, respectively, used for simulation of the electric potential.\nBecause of the influence of the relative permittivity, the material\nwith the lower relative permittivity would be negatively charged,\nwhile the other triboelectric material would be positively charged.\nWe found that the ostrich EM possessed the highest surface charge\nand dielectric constant, resulting in the highest electrical output. Figure 4 (a) Transferred\ncharges and (b) retention times of the surface\npotentials for various types of EMs after contact friction with a\nKapton film. (c) Dielectric constants for the various types of EMs.\n(d) Finite-element simulation of the electric potential variation\nfor the various types of EMs. Table 1 (a) SEM-derived data for pore sizes\nand fiber sizes; and (b) Brunauer–Emmett–Teller (BET)-derived\ndata for surface areas and pore volumes of the hen, duck, goose, and\nostrich EMs analysis item a. Hen b. Duck c. Goose d. Ostrich (a) SEM pore size (μm) 3.8 3.9 8.0 3.9 fiber size (μm) 1.5 1.6 1.9 2.7 (b) BET BET surface area (m 2 /g) 1.51 1.84 1.29 1.17 pore volume (cm 3 /g) 0.000573 0.00121 0.000546 0.00133 A device’s robustness is also critical to its\noutput performance.\nThe wear that occurs to an EM-TENG device after long hours of operation\nwill cause its output to be unstable, gradually degrading. To examine\nthe mechanical stability of the optimal EM-TENG device, we monitored\nthe electrical output of the ostrich EM-TENG over a duration of 9000\ncycles at 40 N with a frequency of 3 Hz ( Figure 5 a), revealing the high durability and stability\nof the ostrich EM-TENG. Figure 5 b displays the voltage curves obtained when charging capacitors\nof varying capacitance (0.1, 1.0, 2.2, 4.7, and 10 μF) with\nan ostrich EM-TENG, with the 0.1 μF capacitor undergoing charging\nto 6.5 V within 37 s. Furthermore, an LED bulb array and a digital\nwatch could be powered by the EM-TENG, as displayed in Movie S1 . Moreover, the ostrich EM-TENG device\ncould be used not only as a self-powering device but also as a sensor.\nWe employed this EM-TENG to detect leg movement by attaching it to\na leg ( Figure 5 d, Movie S2 ). During leg movement, the ostrich EM-TENG\ngenerated an output signal that varied based on the speed of motion.\nIn another instance, we assessed the sensitivity of the ostrich EM-TENG\nby varying the number of fingers pressed onto the device ( Figure 5 e). Interestingly,\nthe shape of the output signal, namely the number of peaks, changed\ndepending on the number of fingers pressing on the device. We also\nsuccessfully implemented the fabricated EM-TENG sensor to detect air\nblowing from a mouth ( Figure 5 f), suggesting that this EM-TENG could be used to harvest\nwind energy. Accordingly, this ostrich EM-TENG device has potential\nfor several applications in monitoring human actions and as a power\nsource. Figure 5 (a) Electrical outputs during long-term cycling of the ostrich\nEM-TENG. (b) Using the ostrich EM-TENG to charge capacitors of various\ncapacitances. (b) (I) Schematic representation of the operating circuit\nfor LED bulbs with a full-wave bridge rectifier and (II) photograph\nof 250 serially connected LEDs powered by a TENG formed from ostrich\nEMs (device size: 2.5 cm × 2.5 cm). (d–f) Sensitivity\nof the ostrich EM-TENG, measured by (d) leg movement, (e) varying\nthe number of fingers pressing on the device, and (f) air blowing."
} | 5,951 |
24466132 | PMC3899268 | pmc | 374 | {
"abstract": "Generation of renewable energy is one of the grand challenges facing our society. We present a new bio-electric technology driven by chemical gradients generated by photosynthesis and respiration. The system does not require pure cultures nor particular species as it works with the core metabolic principles that define phototrophs and heterotrophs. The biology is interfaced with electrochemistry with an alkaline aluminum oxide cell design. In field trials we show the system is robust and can work with an undefined natural microbial community. Power generated is light and photosynthesis dependent. It achieved a peak power output of 33 watts/m 2 electrode. The design is simple, low cost and works with the biological processes driving the system by removing waste products that can impede growth. This system is a new class of bio-electric device and may have practical implications for algal biofuel production and powering remote sensing devices.",
"introduction": "Introduction The search for renewable energy sources has renewed interest in finding ways to use biological systems to generate electrical energy. Specifically there is an interest in systems that use biology to convert light into electrical energy as way to use the advantages of biology to harvest a sustainable energy source. These devices are collectively known as photo-bioelectric systems. In this study we aim to develop a new photo-bioelectric system through the combination of several well-known technologies and widely conserved biological phenomena. The system is distinct from a microbial fuel cell (MFC), as it uses an aluminum oxide cell design to interface phototrophic and heterotrophic metabolisms with power production. Many photo-bioelectric systems are modeled on microbial fuel cells and have been described [1] , [2] . A MFC has microbes associated with the anode oxidizing organic compounds under normally anaerobic conditions, using the anode as the terminal electron acceptor [3] . Electrons are shuttled to a platinum cathode where they combine with O 2 and H + , yielding water [3] . In the closest related class of photo-MFCs, the MFC is fed organic carbon from algae [4] – [6] or plant root exudates [7] – [12] . Energy stored by photosynthesis is liberated when the organic matter is oxidized by bacteria. In another design, algae are added to the cathodic side of the MFC, supplying O 2 in the electron consuming reaction [5] . In both cases electrons are donated from electron transfer chains to the anode. Cyanobacteria can also donate electrons to the anode during respiration of cellular carbon reserves in dark phases with redox shuttles such as HNQ [13] – [15] . The need for a redox shuttle to move electrons from algae to anode limits this type of cell to closed systems. Recently it was shown pure cultures of cyanobacteria could directly donate electrons to the anode [16] . The authors postulate these organisms donate electrons via nanowires when CO 2 is limiting [16] . It is also possible to extract electrons from photosynthesis by using hydrogen as an intermediary [2] . Hydrogen is produced by hydrogenases or nitrogenases and then oxidized at a platinum electrode, recovering the electrons [2] . Ryu et al. take a different approach by inserting nano-electrodes directly into photosynthetic membranes of alga, extracting electrons using an overvoltage [17] . This eliminates light to chemical energy conversion losses, theoretically increasing efficiency, but consumes energy needed by the organism for growth and sustained survival. In this study we present a system that is distinct from established photo-bioelectric systems. It is designed around the normal processes that occur when phototrophs and heterotrophs grow and replicate. Phototrophs and heterotrophs pump carbon through ecosystems, shifting inorganic carbon equilibrium reactions and in the process affecting pH. Algae alter pH by removing CO 2 and HCO 3 \n − , which shift the equilibrium and produces hydroxide ions, resulting in pH values as high as 11 [18] . Respiration oxidizes organic compounds to CO 2 and organic acids, reducing pH. The system extracts energy from acids and bases which are waste products of heterotrophic and phototrophic metabolism respectively. These products of metabolism are interfaced with an electrical system by using an aluminum fuel cell under the alkaline conditions. In this type of fuel cell, water and base react with aluminum to form aluminum oxides and electrons [19] . In an isolated system the electrons typically reduce hydrogen ions resulting in hydrogen gas, however they can also be drawn into an external circuit with an applied voltage. The goal of this study is to demonstrate a proof of principle of exploiting the chemical gradient that occurs when photosynthesis and respiration are separated. The mechanistic data presented focus on the algal side of the cell. We show power production in the cell is pH dependent in the absence of algae. Power production is light and photosynthesis dependent when the algae are present. Finally a cell was deployed in the field to test the robustness of the design outside of the laboratory.",
"discussion": "Results and Discussion The system is shown in figure 1A with the proposed mechanism of action. By separating heterotrophic and phototrophic metabolism a pH gradient can be generated. This pH gradient can be used to generate power. Importantly, the presented system is effectively a photosynthesis dependent battery in the current stage of development due to the use of aluminum (Al) as the anode and does not yet constitute a standalone energy generation system. Substitution of the Al electrode with a catalyst would make it a stand-alone biologically driven power source. The cell has an algae-filled high pH side with an Al electrode and low pH side with a platinum electrode, separated by a paper membrane ( Fig. 1B ). The Al reacts with basic water produced by algal carbon fixation. Al is ionized, leaving electrons on the anode which pass through an external circuit, combining with hydrogen ions at the platinum electrode producing H 2 (shown below). \n 10.1371/journal.pone.0086518.g001 Figure 1 Principle and design of the algal electric cell. ( A ) Respiration decreases pH through the oxidation of organic compounds, generating CO 2 which shifts the inorganic carbon equilibrium and produces hydrogen ions. Photosynthesis increases pH by removing CO 2 and HCO 3 \n − , which shifts the inorganic carbon equilibrium and produces hydroxide ions, resulting in pH values as high as 11 23 . Separation of photosynthesis and respiration results in a pH gradient which can be used to generate electricity. ( B ) Design of test cell. Initially we show the cell is responsive to pH by increasing pH stepwise on the basic side of the cell in the absence of algae while voltage was monitored. The data confirm increased pH results in increased voltage ( Fig. 2A ). Because voltage is dependent on having high pH on one side of the cell and low pH on the other, the greater the difference the higher the voltage as seen in Fig. 2A . 10.1371/journal.pone.0086518.g002 Figure 2 pH dependence of voltage and light dependence of pH and voltage. ( A ) Voltage is dependent on pH of the Al side of the cell. pH was adjusted stepwise in the absence of algal cells. pH was measured every minute and voltage was measured every ten seconds ( B ) Voltage and pH over light/dark cycles (white/black bar). Algal photosynthesis increases pH, increasing voltage. pH and voltage were each measured every 1.5 minutes for 24 hours. At 16 hours fresh f/2 medium was added to compensate for evaporative losses, resulting in slightly higher voltage at a lower pH due to more of the electrode being submerged. The light/dark data presented are from a single cell and are a representative dataset from four independent trials. Results were obtained in cell-free media, showing living cells are not necessary for pH-dependent power production. We observed a steady-state voltage at lower pH values but at pHs higher than ca. 10, voltage continued to increase. One possibility is that the reaction has not yet achieved steady state at the surface of the aluminum at these higher pHs. Aluminum oxidation chemistry at the surface of the metal is complex and not yet completely understood, however we hypothesize that the hydroxide is responsible for the formation of an initial oxide which then reacts with water to form the metal hydroxide. One or more of these reactions may not yet have reached equilibrium. Support for the proposed Al reaction was observed by comparing electrodes before and after ten days of operation. Electrode ionization was observed near the air-liquid interface ( Fig. S1 ). These data support hydroxide dependent production of current through the oxidation of Al. The exact mechanism of Al oxidation in our system is unknown and may proceed by several parallel reactions [19] . While the intermediate steps in Al oxidation are important, here we focus on developing a process linking the biology of algae to an electron producing reaction. Optimal operating load was determined with a power curve, where current was measured at increasing external resistances ( Fig. S2 ). Optimal resistance was 500 Ω, where 285 mW/m 2 Pt electrode was produced. Often in photo-bioelectric systems either algae or a bacterial intermediary must first colonize the electrode [2] . In our system voltage is produced immediately upon submersion of the electrodes, showing colonization is not required, which is consistent with power generation being driven by the pH. In addition, different alga in different media with similar pH values produce similar voltages including Microcystis aeruginosa LE-3 (freshwater cyanobacterium), Scenedesmus sp. HTB1 (freshwater green alga) and Nannochloropsis oceanica IMET1 (marine alga). We hypothesized production of hydroxide ions through the removal of inorganic carbon via photosynthesis drives power production. In our next set of experiments we tested if voltage is light dependent. Algae in the cell were passed through a series of light/dark cycles while voltage and pH were monitored. During light phases, pH and voltage increase ( Fig. 2B ). Photosynthesis likely causes the pH increase and leads to higher cell voltage. During dark phases, photosynthesis stops and voltage decreases as hydroxide is consumed in the reaction with Al. In a separate trial the light/dark dependence was tested and found to be stable over the two week course of the experiment ( Fig. 3 ). These data show increases in voltage are light dependent and likely due to pH increases from photosynthesis. The results also show that the same magnitude of voltage is produced when the same magnitude of pH gradient is present. To confirm power output is dependent on photosynthesis we used DCMU (3-(3,4-dichlorophenyl)-1,1-dimethylure), a specific inhibitor of photosystem II. DCMU was added to cultures under constant illumination while voltage and pH were monitored. Upon addition of the inhibitor pH increase stopped, whereas the pH of the solvent control lacking DCMU continued to climb ( Fig. 4A ). Cell voltage resembles the pH curve, voltage increase stopped after addition of the inhibitor, while voltage continued to climb in the control ( Fig. 4B ). Voltage plateaued before pH in the control, potentially indicating inhibition by constraints on the cathodic reaction imposed by the size of the Pt electrode, or the internal resistance of the cell. The data show specific inhibition of photosynthesis results in lower pH and lower voltage, confirming voltage is light and photosynthesis dependent. 10.1371/journal.pone.0086518.g003 Figure 3 Change in voltage is light dependent. Algal electric cell with algae exposed to light (white) and dark (black) cycles for 13 days. Voltage increases when light falls on the algal electric cell, plateaus and then decreases in the dark. 10.1371/journal.pone.0086518.g004 Figure 4 Voltage and pH are dependent on photosynthesis. Addition of the photosynthesis inhibitor DCMU (orange arrow) ( A ) inhibits pH increase, and ( B ) voltage increase. Control and inhibition experiments were carried out on separate algal stocks on separate days showing biological reproducibility of pH and voltage curves prior to addition of inhibitor. Data presented are from a single cell and are a representative dataset from three independent trials. Next we were interested in testing the limits of this type of cell. We hypothesized power was limited by internal resistance; this was tested with a cell cast into 1.5% agar with 5 M NaCl. This setup produced a peak power output of 33 W/m 2 Pt electrode with a stable power output of 0.7 W/m 2 Pt. This peak power output is over 33 fold higher than the next highest photo-bioelectric cell [6] and nearly 6 fold higher than the highest bioelectric system [22] . High power is possible due to liberation of stored energy in the Al electrode through reaction with algae-derived hydroxide. The Al electrode could be substituted with a catalyst to extract electrons from hydroxide ions such as a perovskite oxide [23] . It should also be possible to increase power by decreasing the pH of the acidic side of the cell, though the focus of the current paper is on the algal side of the cell. The current device is useful for powering remote sensors in aquatic systems and for algal biofuel production to control pH and provide power for mixing ponds. The mechanism of action suggests the cell is species neutral. The species neutral nature of the cell was tested with a larger version deployed in a Chesapeake Bay tributary. This cell incorporated the entire proposed design principle and is shown in figure 5A . A power curve showed peak power output occurred at 50 Ω resistance ( Fig. S2 ). Power was produced immediately upon submerging the cell at the site, where pH was 8.3. The cell interior maintained a lower pH than the surrounding water throughout the experiment. A maximum power output of 192 mW/m 2 Pt electrode was recorded shortly after start up with an average power output of 11 mW/m 2 Pt ( Fig. 5B ). Unlike the laboratory-scale test, the cell was affected by environmental factors including storms and tides. Power produced roughly follows the day/night cycle, deviations from this cycle correlate well with changes in tides. Changes in tides affect power output by altering the area of submerged Pt electrode. After the fourth day a storm moved over the site and stirred up sediment, correlating with a drop in power to 0.3–1 mW/m 2 Pt electrode ( Fig. 5B ). We speculate this was due to a pH decrease from inhibition of photosynthesis. After this period the cell recovered, and resumed power production. From these data we conclude the cell can successfully produces power with an undefined freshwater microbial community. The system we presented represents a novel class of photo-bioelectric cell, where photosynthesis and respiration are linked to electrical power production via known electrochemical reactions. This type of cell can generate high current densities and operate outside of the laboratory in a species neutral manner. Power production from different species is likely dependent on the rate at which algae removed inorganic carbon from water. This rate will differ based on the algal species present and the densities on those algae. To maintain a constant voltage from the system under differing conditions of algal carbon fixation the external resistance can be adjusted to obtain a desired voltage. 10.1371/journal.pone.0086518.g005 Figure 5 Design and field-testing of the outdoor photo-bioelectric cell. ( A ) Cutaway of the cell. Inside of the cell is water with the local biota which is kept in the dark, inhibiting photosynthesis while allowing respiration. Pt cloth electrode is wrapped around the center. Paper membrane separates the inside of the cell from the surrounding water. Al electrode is wrapped around the cell. ( B ) Power produced by the cell. Power is correlated with day/night cycle (white/black bar) and tides (blue). After day four a storm suspended sediment, likely inhibiting photosynthesis and power production. Voltage was measured every 30 seconds for two weeks over a 50 Ω resistance. Data presented are from a single cell deployed on-site. It is important to remember that the current cell design is dependent on oxidation of aluminum and therefore does not yet constitute a stand-alone sustainable energy source. It may however be used as a way to liberate the energy stored in the reduced metal in a portable manner. This would allow aluminum to be used effectively as an energy storage medium for transferring energy from areas that have access to large reserves of renewable energy, such as hydroelectric power, to areas that do not. Further research may allow power to be generated directly from the pH gradient that occurs when photosynthetic carbon fixation is separated from respiration."
} | 4,258 |
26868597 | PMC4824173 | pmc | 375 | {
"abstract": "Reef-building corals depend on symbiotic mutualisms with photosynthetic dinoflagellates in the genus Symbiodinium . This large microalgal group comprises many highly divergent lineages (“Clades A–I”) and hundreds of undescribed species. Given their ecological importance, efforts have turned to genomic approaches to characterize the functional ecology of Symbiodinium . To date, investigators have only compared gene expression between representatives from separate clades—the equivalent of contrasting genera or families in other dinoflagellate groups—making it impossible to distinguish between clade-level and species-level functional differences. Here, we examined the transcriptomes of four species within one Symbiodinium clade (Clade B) at ∼20,000 orthologous genes, as well as multiple isoclonal cell lines within species (i.e., cultured strains). These species span two major adaptive radiations within Clade B, each encompassing both host-specialized and ecologically cryptic taxa. Species-specific expression differences were consistently enriched for photosynthesis-related genes, likely reflecting selection pressures driving niche diversification. Transcriptional variation among strains involved fatty acid metabolism and biosynthesis pathways. Such differences among individuals are potentially a major source of physiological variation, contributing to the functional diversity of coral holobionts composed of unique host–symbiont genotype pairings. Our findings expand the genomic resources available for this important symbiont group and emphasize the power of comparative transcriptomics as a method for studying speciation processes and interindividual variation in nonmodel organisms.",
"conclusion": "Conclusions Comparisons among deeply sequenced transcriptomes can reveal the extent and function of molecular variation that is critical to speciation in nonmodel organisms. Such work provides important baselines against which experimentally manipulated samples might be compared and more accurately interpreted. Our data reveal the extent of expression variation that occurs among strains of Symbiodinium and emphasizes how natural selection on existing populations may play a critical role in the response of coral–dinoflagellate symbioses to climate change. The genomic resources described here should improve functional investigations into marine symbiosis biology, particularly as model systems continue to be developed ( Baumgarten et al. 2015 ). Future studies should examine the same strains exposed to different stressors (thermal, osmotic, and/or light) in order to characterize the relationship between physiological and gene expression phenotypes. Each strain should also be brought into an experimental host (e.g., the model Aiptasia [= Exaiptasia ]) and observed in symbiosis, which would provide insight into how changes in gene expression work to maintain stable cnidarian–dinoflagellate mutualisms. Our findings underscore that important transcriptional differences exist at different taxonomic ranks among dinoflagellates, from clades to species to strains. Future Symbiodinium genomics experiments should be designed such that clade-level questions incorporate different species to serve as a representative sampling of the clade under study, while species-level questions should incorporate distinct strains to serve as a representative sampling of the species under study. Such designs will improve our understanding of Symbiodinium genetic, functional, and phylogenetic diversity.",
"introduction": "Introduction The concept that adaptation and speciation are driven largely by natural selection on variant individuals of a population is central to evolutionary biology. Much like other types of genetic diversity, gene expression variation is extensive, highly heritable, and subject to selection ( Ranz and Machado 2006 ; Voolstra et al. 2007 ; Wittkopp et al. 2008 ). The role of differential gene expression in ecological speciation has received renewed interest in the genomics era because the molecular biology of nonmodel organisms with unique evolutionary histories can now be studied in great detail at relatively low cost ( Wolf et al. 2010 ). For example, among two recently diverged species of cordgrass, only one is successful at invading environments perturbed by climate change, and it exhibits unique expression patterns at growth- and stress-related genes ( Chelaifa et al. 2010 ). A similar study in daisies illustrated that a comparative transcriptomic framework can be used to identify selective processes affecting ecological speciation ( Chapman et al. 2013 ). Additionally, transcription-based assays of microbial metagenomes have revealed unique niche diversification (e.g., specialization for certain substrates, metabolic pathways, or environments) that is otherwise hidden due to functional redundancy in the genomes of many bacteria ( Gifford et al. 2013 ). Thus, comparative genomics can also provide a means to recognize important functional variation in organisms that are difficult to probe phenotypically, such as corals and their symbionts. Coral reef ecosystems support tremendous marine biodiversity and ecological goods and services ( Moberg and Folke 1999 ). Coral productivity and growth depend on a mutualism with endosymbiotic dinoflagellates known as Symbiodinium ( Muscatine and Porter 1977 ; Muscatine 1990 ; Yellowlees et al. 2008 ). This microalgal “genus” is incredibly diverse, encompassing at least nine major lineages that show ribosomal divergence equivalent to that found among different genera, families, or even orders of other dinoflagellates ( Rowan and Powers 1992 ). Likewise, Symbiodinium exhibit many unique ecologies, ranging from “host-specialized” taxa commonly found as symbiotic partners of corals ( Parkinson, Coffroth, et al. 2015 ), to “ecologically cryptic” taxa with alternate nonsymbiotic lifestyles ( LaJeunesse et al. 2015 ), to completely “free-living” taxa that thrive independently in the water column ( Jeong et al. 2014 ). Unlike their mostly obligate coral hosts, many Symbiodinium can survive ex hospite and are maintained in culture. In the natural environment, stressful conditions can cause the association between corals and host-specialized symbionts to break down in a process called coral bleaching, which can lead to colony mortality ( Fitt et al. 2001 ). Climate change is predicted to drive more frequent and intense bleaching events ( Hoegh-Guldberg 1999 ), prompting a major research focus on how climate-related stressors might affect coral–dinoflagellate symbioses in the future. Accordingly, the last decade has generated many studies describing coral host transcription in various contexts ( Meyer and Weis 2012 ), but comparable studies in Symbiodinium are still in their early stages ( Leggat et al. 2007 ; Leggat, Yellowlees, et al. 2011 ; Lin 2011 ). With the incorporation of next-generation sequencing technology, genomic resources for Symbiodinium have expanded greatly despite their status as a nonmodel organism. The first draft genome was released in 2013 ( Shoguchi et al. 2013 ), with the complete chloroplast genome following shortly thereafter ( Barbrook et al. 2014 ). Multiple mRNA transcriptomes are available ( Bayer et al. 2012 ; Ladner et al. 2012 ; Baumgarten et al. 2013 ; Rosic et al. 2014 ; Xiang et al. 2015 ), representing the four major clades known to associate with scleractinian corals (Clades A, B, C, and D). Recent efforts have expanded in important new directions, such as the description of Symbiodinium microRNAs ( Baumgarten et al. 2013 ), the comparison of orthologous genes among clades ( Voolstra et al. 2009 ; Ladner et al. 2012 ; Barshis et al. 2014 ; Rosic et al. 2014 ), the completion of another draft genome ( Lin et al. 2015 ), and the development of the Aiptasia – Symbiodinium system for in-depth cellular and physiological research ( Weis et al. 2008 ; Sunagawa et al. 2009 ; Lehnert et al. 2012 , 2014 ; Xiang et al. 2013 ; Baumgarten et al. 2015 ). Dinoflagellate genomes are unique among eukaryotes for multiple reasons ( Leggat, Yellowlees, et al. 2011 ). Of particular note, dinoflagellates including Symbiodinium modulate nuclear-encoded protein levels predominantly by posttranscriptional processes ( Morse et al. 1989 ; Leggat, Seneca, et al. 2011 ). It is now understood that dinoflagellates also exhibit some measure of transcriptional regulation, albeit changes in expression profiles are minimal when exposed to different environmental conditions ( Erdner and Anderson 2006 ; Moustafa et al. 2010 ). For example, the number and magnitude of expression changes among Symbiodinium exposed to thermal stress are relatively small compared with their animal hosts ( Leggat, Seneca, et al. 2011 ). Barshis et al. (2014) found that two Symbiodinium spp. in Clades C and D did not alter gene expression when exposed to temperature stress in hospite, even though the host response involved the modulation of hundreds of genes ( Barshis et al. 2013 , 2014 ). Interestingly, a large number of transcriptional differences were maintained, or fixed, for the two species from different clades regardless of temperature treatment ( Barshis et al. 2014 ). This suggests that fixed expression differences are likely to be evident in strains cultured ex hospite under identical controlled environmental conditions. Differences in these “stable-state” expression profiles among lineages may strongly reflect evolutionary divergence, some of which may be adaptive. These expression patterns may also correspond to functional differences among distantly related species. If lineage-specific expression extends to the subcladal level—that is, between species within clades or among individual strains within species—it will be critical to recognize this source of variation when interpreting Symbiodinium genomic data and account for it in future experimental designs. By comparing different isoclonal cell lines (strains), it is possible to reveal intraspecific variation in genomic features that underlie ecological and physiological phenotypes. For example, unique genes distinguish strains of nitrogen-fixing rhizobial bacteria with different symbiotic efficiencies and host specificities ( Galardini et al. 2011 ; Österman et al. 2015 ). At the level of transcription, toxic and nontoxic strains of the dinoflagellate Alexandrium minutum maintain fixed expression differences at shared genes ( Yang et al. 2010 ). We may expect similar patterns among Symbiodinium strains, but this idea has never been tested. Symbiodinium belonging to Clade B are ideal candidates for further genomic characterization because several ecologically distinct species within this group were recently described ( LaJeunesse et al. 2012 ; Parkinson, Coffroth, et al. 2015 ), a draft genome exists for the member species S ymbiodinium minutum ( Shoguchi et al. 2013 ), and multiple genetically distinct cultures are available for several species. Currently, the extent of variation among species within a single Symbiodinium clade and among individual strains within a single species is mostly unknown ( Parkinson and Baums 2014 ). To address this knowledge gap, we analyzed stable-state gene expression among four species representing the two major evolutionary radiations within Clade B: The Pleistocene (B1) radiation and the Pliocene (B19) radiation (sensu LaJeunesse 2005 ). For each radiation, two species with different ecologies were studied: Either host-specialized taxa or ecologically cryptic taxa. Although these latter species were originally cultured from coral tissues, they have never been detected as the numerically dominant symbionts in cnidarian mutualisms, and therefore were probably commensals or free-living contaminants isolated from the mucus or gastrovascular cavity ( Parkinson, Coffroth, et al. 2015 ). Where available, we incorporated biological replication in the form of distinct isoclonal cell cultures. The genomic resources developed here should assist in the design and interpretation of future comparative transcriptional analyses among Symbiodinium strains, species, and clades, as well as broaden our understanding of speciation among microeukaryotes.",
"discussion": "Discussion Fixed differences in gene expression ultimately influence the phenotypic variation available for selection to act upon. We anticipated that a comparative analysis of Symbiodinium spp. transcription would improve our understanding of adaptation and speciation among microeukaryotes. Indeed, we found that despite an overall similarity in gene content and expression among Clade B species with distinct ecologies, all cultures exhibited lineage-specific expression differences diagnostic for each species. Overrepresentation of photosynthesis-related gene expression variation among species likely reflects adaptation to unique light regimes over evolutionary time. Extensive disparity in the expression of fatty acid metabolism genes among strains within species may translate into differences in membrane composition, thermal tolerance, energy reserves, and growth rates. These differences may play a crucial role in coral–dinoflagellate symbiosis ecology and evolution. By examining the stable-state transcriptomes of cultures reared independently of their hosts under identical environmental conditions, we infer that these differences stem from genotypic rather than environmental factors. Our efforts reinforce the utility of comparative transcriptomics for studying speciation and functional variation in dinoflagellates and other nonmodel organisms ( Chelaifa et al. 2010 ; Chapman et al. 2013 ; Gifford et al. 2013 ). Partitioning the Variation in Gene Expression When comparing multiple species, expression differences can be attributed to 1) technical variation, 2) within-species variation, and 3) among-species variation, with the proportion of variable genes expected to increase from 1) to 2) to 3) ( Whitehead and Crawford 2006 ). Our results matched this general trend. Technical variation was inferred to be low based on the agreement between our data and transcriptome statistics from other studies that included the same S. minutum Mf1.05b strain, the high mapping success achieved between our Mf1.05b reads and the draft genome derived from the same strain (73%), and the nonrandom distribution of expression differences among species ( fig. 4 c ). The percentage of orthologous genes differentially expressed within species (1.54% for S. minutum and 1.18% for S. psygmophilum ; fig. 4 ) was roughly half that found between species (2.33%; fig. 2 b ). Overall, DEGs make up a small proportion of the entire transcriptome, as has been found before for Symbiodinium and other dinoflagellates ( Baumgarten et al. 2013 ; Barshis et al. 2014 ; Xiang et al. 2015 ). Between-Species Variation By comparing closely related species within a clade, we greatly expanded our comparative power to determine what genetic changes underlie speciation among Symbiodinium. We were able to identify at least four times as many orthologs shared between Clade B species as has been possible using similar methods to compare species across separate clades ( Ladner et al. 2012 ; Barshis et al. 2014 ; Rosic et al. 2014 ). Genetic divergence between clades is massive ( Rowan and Powers 1992 ), and thus comparisons among species within clades reveal finer-scale differences likely to be important in physiological and ecological processes. Overall, stable-state gene expression was similar among Clade B Symbiodinium . Of the nearly 20,000 orthologs shared by S. aenigmaticum , S. minutum , S. pseudominutum , and S. psygmophilum , only 452 (2.3%) were differentially expressed between species. Thus a substantial portion of the transcriptome maintains relatively constant expression levels across members of Clade B. This result mirrors similar studies in other systems such as flowering plants where only a small proportion of interspecific orthologs were differentially expressed ( Chapman et al. 2013 ). The species comparison with the greatest number of DEGs was S. minutum versus S. psygmophilum ( fig. 1 b ), which fit expectations for several reasons. First, our replication scheme (four strains per species) may have enhanced our ability to detect fixed differences between these species’ transcriptomes (though this is unlikely; see Materials and Methods). Second, both species associate with different hosts and likely diverged in part due to coevolutionary constraints imposed by those hosts, whereas ecologically cryptic species may not have faced the same constraints. Third, they are from phylogenetically divergent lineages. Finally, S. minutum is warm water adapted, while S. psygmophilum is cold water adapted ( Thornhill et al. 2008 ), likely contributing to expression differences. Interestingly, the contrasts with the second- and third-most abundant DEG counts both involved S. aenigmaticum ( fig. 1 b ), a very distinct species from the Pliocene radiation and one that appears to have undergone rapid evolution ( fig. 1 a ; LaJeunesse 2005 ; Parkinson, Coffroth, et al. 2015 ). The three species pair comparisons with the least number of DEGs all involved S. pseudominutum ( fig. 1 b ). In fact, this species was roughly equidistant from all other species based on DEG number and MDS position ( fig. 3 c ). Its position might be explained on the one hand by its close evolutionary history with S. minutum , and on the other by its cryptic ecology shared with S. aenigmaticum . Based on these results, fixed differences in gene expression may not always correspond to phylogenetic similarity. Multidimensional scaling offered a complementary analysis for visualizing the similarities in expression among all strains without a priori knowledge of species membership ( fig. 3 ). By restricting the data set to only non-DEGs, almost all replicates from all species (8 of 10) clustered together ( fig. 3 a ), matching the expectation that at stable-state these Clade B Symbiodinium generally maintain similar expression profiles. When both non-DEGs and DEGs were included in the analysis, each species was mostly resolved, showing that non-DEGs contributed little to either species-level signal or noise ( fig. 3 b ). As expected, when only the DEGs were considered, all species resolved well ( fig. 3 c ). Note however that the distant positioning of S. aenigmaticum in all three MDS plots indicates that a large proportion of expression variation for this species is unique. In addition to pairwise comparisons, we also contrasted groups of replicate species by lineage (2 species from the Pleistocene radiation vs. 2 Pliocene radiation species) or by ecology (2 host-specialized species vs. 2 ecologically cryptic species). The Pleistocene—Pliocene contrast was equivalent to the S. minutum – S. psygmophilum comparison in terms of identity of DEGs, meaning that the species contrast either captured all the differences between major lineages, or that adding just one more strain to each group did not affect expression variation sufficiently to alter our detection of DEGs, even though the strain belonged to a different species. Similarly, the ‘host-specialized’—‘ecologically cryptic’ contrast only recovered four unique genes that had not been identified in any of the species contrasts. The probable identity of only one of these DEGs was determined (a general mRNA splicing factor). These results indicate that differential expression of a particular set of genes does not necessarily explain shared ecological attributes of phylogenetically distinct species. Photosynthesis Gene Expression Differences between Species Expression differences among closely related species were consistently enriched for photosynthesis genes ( supplementary table S2 , Supplementary Material online ). Here, overrepresentation of plastid genes cannot be attributed to light intensity differences because all cultures were reared under identical light conditions. In fact, although we might expect these genes to be regulated by light intensity in Symbiodinium as they are in other photosynthetic organisms ( Escoubas et al. 1995 ; Pfannschmidt 2003 ), only minor (or no) changes in photosynthesis-related gene expression have been detected in cultures exposed to varying light levels ( McGinley et al. 2013 ; Xiang et al. 2015 ). Thus, we conclude that different species evolved unique expression levels among photosynthesis-related genes. These differences may relate to inherent variation in the circadian rhythm among species ( Van Dolah et al. 2007 ; Sorek and Levy 2012 ) or, more likely, to functional variation in photosynthesis biochemistry. For example, during heat stress, thermally sensitive Symbiodinium taxa suffer physiological disruption of PSII photochemistry ( Warner et al. 1999 ; Robison and Warner 2006 ) and associated downregulation of core photosynthesis genes ( McGinley et al. 2012 ), whereas thermally tolerant species do not. The maintenance of distinct expression patterns at key genes may underlie the capacity for certain Symbiodinium species to occupy distinct niches, as has been demonstrated for three diatom species in the genus Pseudonitzschia ( Di Dato et al. 2015 ). Evolutionary Significance of Gene Expression Variation In biogeographic surveys of marine mutualisms, depth and latitude (correlates of light availability) are often primary factors explaining the distribution of Symbiodinium diversity ( Rowan and Knowlton 1995 ; LaJeunesse et al. 2004 , 2014 ; Frade et al. 2008 ; Finney et al. 2010 ; Sanders and Palumbi 2011 ). Thus, light availability represents a main axis of niche differentiation for this group. Symbiodinium possess a diverse array of light-harvesting proteins ( Boldt et al. 2012 ), which may be both the cause and consequence of ecological specialization. Many such genes have been transferred to the nuclear genome ( Bachvaroff et al. 2004 ), while others are encoded on plastid minicircles ( Zhang et al. 1999 ; Moore et al. 2003 ; Barbrook et al. 2014 ). Minicircles are subject to different transcriptional mechanisms than nuclear encoded genes ( Dang and Green 2010 ), which may also facilitate specialization to different light regimes. Given that a majority of expression variation between divergent species is expected to accumulate neutrally over time ( Khaitovich et al. 2005 ), it is intriguing that expression differences between Symbiodinium species are consistently enriched for photosynthesis genes ( Baumgarten et al. 2013 ; Barshis et al. 2014 ; Rosic et al. 2014 ; this study). This evidence suggests that species-specific differences in gene expression are functionally important and influenced by natural selection tied to niche diversification. Within-Species Variation Within each of the two species with four isoclonal cultures, we detected hundreds of DEGs: 659 unique genes among S. minutum strains ( fig. 4 a ) and 506 unique genes among S. psygmophilum strains ( fig. 4 b ). Interestingly, only four annotated genes differentially expressed among S. minutum overlapped with those among S. psygmophilum , and enriched categories only overlapped for housekeeping genes which regulate biochemical processes like nucleic acid synthesis and microtubule organization ( supplementary table S2 , Supplementary Material online ). Thus, transcriptional variation among strains differs from species to species ( fig. 5 ). Furthermore, nonrandom gene expression differences among strains of a given species exist even under identical rearing conditions ( fig. 4 c ), emphasizing that a degree of expression variation among Symbiodinium strains is genetically determined and potentially subject to natural selection. Thus, the extent of variation among isoclonal strains may be much greater than previously assumed. Although inter-individual differences are known to play a significant role in symbiosis ecology and evolution in terrestrial systems ( Shuster et al. 2006 ; Whitham et al. 2006 ; Hughes et al. 2008 ), such evidence has been lacking for coral–dinoflagellate associations ( Parkinson and Baums 2014 ). Although ∼500 of the ∼40,000 genes represents a small fraction of the transcriptome, such differences may be important, especially because overall differential expression of genes within a Symbiodinium species responding to stress seems low ( Barshis et al. 2014 ; but see Baumgarten et al. 2013 ). For example, putative “symbiosis genes” have been identified by comparing symbiotic versus aposymbiotic cnidarian hosts ( Meyer and Weis 2012 ). The expression levels of similar genes in the symbiont may also play a role in maintaining functional associations. Two such genes varied among S. minutum strains: An ABC transporter (up to 4.2-fold) and a glutathione reductase (up to 9.5-fold). There were also clear differences in the expression of long chain fatty acid CoA ligase (up to 12.2-fold), long chain acyl-CoA synthetase (up to 8.8-fold), and six acetyl-CoA carboxylases (up to 12.5-fold), indicating that certain strains regulate fatty acid metabolism differently. These genes may be related to cell membrane composition, which in turn can affect thermal sensitivity ( Tchernov et al. 2004 ; Diaz-Almeyda et al. 2011 ). They may also relate to energy storage and nutrient availability, perhaps contributing to different growth rates observed among some of these strains ex hospite ( Parkinson and Baums 2014 ). Under environmental change, these functional differences may impact stress tolerance among genotypic host–symbiont combinations in a population ( Parkinson and Baums 2014 ; Parkinson, Banaszak, et al. 2015 ), partly explaining why some coral colonies of a given species bleach while others do not, even when sharing the same symbiont species ( Goulet et al. 2008 ; LaJeunesse et al. 2010 ). Similar fine-scale variation has been observed among maize strains with distinct flavonoid content ( Casati and Walbot 2003 ) and among dinoflagellate strains with distinct toxicities ( Yang et al. 2010 )."
} | 6,552 |
39423261 | PMC11488576 | pmc | 376 | {
"abstract": "Renewable alternatives for nonelectrifiable fossil-derived chemicals are needed and plant matter, the most abundant biomass on Earth, provide an ideal feedstock. However, the heterogeneous polymeric composition of lignocellulose makes conversion difficult. Lignin presents a formidable barrier to fermentation of nonpretreated biomass. Extensive chemical and enzymatic treatments can liberate fermentable carbohydrates from plant biomass, but microbial routes offer many advantages, including concomitant conversion to industrial chemicals. Here, testing of lignin content of nonpretreated biomass using the cellulolytic thermophilic bacterium, Anaerocellum bescii , revealed that the primary microbial degradation barrier relates to methoxy substitutions in lignin. This contrasts with optimal lignin composition for chemical pretreatment that favors high S/G ratio and low H lignin. Genetically modified poplar trees with diverse lignin compositions confirm these findings. In addition, poplar trees with low methoxy content achieve industrially relevant levels of microbial solubilization without any pretreatments and with no impact on tree fitness in greenhouse.",
"introduction": "INTRODUCTION The environmental impact of climate change motivates reduced dependence on nonrenewable fossil feedstocks for producing carbon-based fuels and chemicals. Plant biomass, as the most abundant renewable material on Earth, is the primary candidate for replacing fossil feedstocks, but this requires efficient deconstruction and subsequent conversion of the polymers contained within lignocellulose to be economically viable ( 1 , 2 ). Lignocellulose is composed of polysaccharides cross-linked with the phenolic polymer lignin such that processes using lignocellulose need to liberate carbohydrates from lignin ( 3 ). Mechanical, chemical, and enzymatic pretreatments, alone or in combination, have been used to achieve this. However, microbial routes, which can solubilize the carbohydrate content of plant biomass and concomitantly produce fuels and chemicals (i.e., consolidated bioprocessing), are particularly attractive ( 4 , 5 ). Metabolic engineering has been used to create microbial strains that convert simple fermentable sugars to industrial products, but there are clear advantages if the microbes can also catabolize carbohydrate polymers from plant biomass, especially microcrystalline cellulose ( 6 ). Certain thermophilic bacteria (i.e., genera Caldicellulosiruptor, Anaerocellum , and Acetivibrio ) natively solubilize and use a wide range of plant polysaccharides using large sets of extracellular enzymes and have been metabolically engineered to make fuels and chemicals ( 4 ). Among these, Anaerocellum (f. Caldicellulosiruptor ) bescii , belonging to the extremely thermophilic Caldicellulosiruptorales ( T opt > 70°C), has demonstrated the capacity to degrade a wide range of plant biomasses, resist contamination, and produce industrially relevant products ( 2 , 7 , 8 ). However, it is clear that lignin remains a barrier to highly efficient biomass degradation by A. bescii , which is reflected in the disparate levels of carbohydrate solubilization of low and high lignin plant biomasses (such as soybean hulls as compared to poplar wood) ( 2 , 8 ). The lignin barrier extends to other thermophilic microbes, such as Acetivibrio thermocellus ( 8 ). The key to expanding the use of plant feedstocks is to make them less recalcitrant and easier to degrade without compromising fitness. An ideal plant biomass feedstock should be genetically tractable, fast growing, require minimal use of pesticides, and grow on marginal lands to not compete with food crops; Populus trichocarpa (black cottonwood) fits these requirements ( 4 , 9 , 10 ). A. bescii fermentation of genetically modified P. trichocarpa lines has been used to screen tree lines for improved feedstock qualities (i.e., more efficient deconstruction) ( 11 ); A. bescii reached nearly 90% carbohydrate solubilization of certain low-lignin poplar lines ( 12 ). Unfortunately, poplar lines that were the most amenable to degradation had substantial fitness issues ( 11 ). However, as recently reported, multiplexed CRISPR-edited poplar trees can have superior wood properties for fiber pulping, producing trees with lower lignin, higher carbohydrate-to-lignin ratio, and increased S/G ratio all while preserving the overall fitness of the trees ( 9 ). The goal of this work was to determine whether plant biomasses, including genetically modified poplar, that are superior for fiber pulping are also amenable to microbial conversion and if features can be identified that are indicative of their ability to be solubilized into fermentable sugars.",
"discussion": "DISCUSSION The results here provide important insights into the relationship between lignin content and composition as this relates to recalcitrance to carbohydrate solubilization and conversion by A. bescii and Ac. thermocellus , not just for poplar but also for a wide variety of plant biomasses. The extension of these findings to other (hemi)cellulolytic, anaerobic, thermophilic bacteria (and less thermophilic fermentative microorganisms) needs to be determined but likely follows the findings here, based on other studies on biomass solubilization and conversion ( 4 , 8 ). In particular, consolidated bioprocessing systems (such as those using Ac. thermocellus and Thermoanaerobacterium species) and lignocellulolytic anaerobic digester systems likely follow these trends and should be evaluated ( 16 – 18 ). These results may only apply to microbes that produce carbohydrate active enzymes and not to lignin-degrading aerobic microorganisms. The challenge of developing modified biomasses with good growth and fitness characteristics and low recalcitrance has been met for genetically modified poplar lines (at least as 6-month-old seedlings) ( 9 ), but targeting low methoxy content provides a clear-cut objective. Correlation between methoxy content and microbial access to carbohydrates explains the variability reported previously for lignin impact on solubilization; data either poorly correlate with total lignin [as seen in Figs. 1 (A and B) and 3C ] or are inconsistent on the impact of S/G ratio ( 11 , 19 – 21 ). Here, incorporating the large number of poplar lines with markedly different lignin contents and compositions, the unifying variable, methoxy content, was demonstrated as the primary barrier for microbial plant biomass solubilization. For lignocellulose with the same lignin compositions, the methoxy content number ( Eq. 1 ) reduces to total lignin content alone; this reflects the weak correlations seen between solubilization and total lignin. Historically, correlations with total lignin content and lignin composition (i.e., S/G ratio) have been conducted independently resulting in variable reports on S/G ratio influence on microbial solubilization. S/G ratio does not correlate with microbial solubilization and conversion because the ratio fails to account for the absolute quantities of S or G lignin in relation to total plant biomass. For microbial solubilization, monolignol preference appears to be H > G > S. This allows lines like H-4-1 (high G) and line 54 (high H) to have low methoxy content numbers at higher total lignin levels than if lignin composition was identical to WT poplar. In other words, for poplar with WT lignin composition to have the same methoxy content as H-4-1 or line 54, ~10.1 or ~6.2% total lignin would be required, respectively (undoubtedly affecting tree fitness). This observation also extends to other biomasses. Grasses (such as wheat straw, corn stover, and switchgrass) are higher in H lignin and G lignin compared to hardwoods like WT poplar. These lignin compositions contribute less methoxy content per total lignin ( Tables 1 and 2 ) when compared to the S-unit–rich WT poplar. Softwoods, like Fraser fir examined here, resist anaerobic microbe’s access to carbohydrates; despite being high in G lignin and low in S lignin, their methoxy content remains high due to high total lignin (37.8% for Fraser fir). Conversely, spent coffee beans, which are high in H lignin (60%) and total lignin (32.8%), can still have low methoxy content, enabling efficient microbial access to carbohydrates. H-4-1 poplar would be economically and technologically relevant in an industrial biorefinery setting [≥65% substrate utilization ( 2 )]. However, opportunities remain for further optimization. As H-4-1 is predominantly G lignin, a shift to higher H and even lower S lignin could generate poplar with even lower methoxy content. The relationship between low lignin and methoxy content as they relate to plant biomass recalcitrance is not completely clear. Lower lignin improves both chemical and microbial (and by association, enzymatic) deconstruction, but the influence of methoxy content and lignin composition diverge. Increased methoxy content may result in a higher degree of carbohydrate-lignin cross-links that are less enzymatically available. This may be reflected in the wide-field and confocal images ( Fig. 5 ), where the carbohydrate (blue) and lignin (red) appear to be intertwined for WT and E-3-1 poplar, while H-4-1 poplar appears to have more distinct phases of lignin and carbohydrate. Higher S/G ratio has been associated with less condensed lignin polymer harboring lower levels of carbon-carbon linkages between subunits of lignin, which may reduce interference during chemical deconstruction ( 22 ). Phenol ring methoxy substitutions affect resonance structures of monolignol radicals during lignin polymerization ( 13 ). Methoxy groups prevent formation of radicals in the 3′ and 5′ phenol ring positions, instead favoring the 4′ hydroxyl or the 1′3-hydroxypropenyl group. This may explain why H > G > S lignin leads to a more condensed structure with a higher amount of lignin-lignin (phenol-phenol) linkages. As lignin-carbohydrate complexes are not associated directly with the phenol ring carbons ( 23 ), this likely reduces cross-linking with carbohydrates. It is clear from Fig. 5 that the residual cellulose (blue) is not localized, suggesting a chemical rather than physical reason for the recalcitrant residual carbohydrate; this is consistent with the lignin-carbohydrate cross-linking hypothesis. High S lignin is favorable for alkaline and chemical treatments (i.e., pretreatments or pulp and paper), in contrast to conditions that favor microbial solubilization and conversion. This explains why alkaline pretreatments work well with A. bescii ( 24 ), as it likely breaks these recalcitrant lignin-carbohydrate linkages. Furthermore, methoxy content could be used to guide the intensity of chemical pretreatments required. Where low methoxy biomasses (≤~0.17) do not require pretreatment with A. bescii , higher methoxy substrates may benefit from pretreatments. The exact methoxy content cutoff is likely dependent on processes and microbes selected. This also suggests that postfermentation chemical treatments may liberate the residual, recalcitrant carbohydrates. As both microbial and chemical treatments favor lower total lignin, but diverge in context of lignin composition, they are not complete inverses of each other. S/G ratio historically has been used to explain variation unaccounted for in total lignin content for chemical treatments. Genetically modified plants, such as the poplar examined in this study, now span a wider range of total lignin content and lignin compositions. Chemical treatment correlations with S/G ratio may not continue to hold for these plants. Beyond methoxy content, other factors likely influence microbial solubilization of lignocellulose. For example, when A. bescii was compared with Ac. thermocellus for solubilization of certain agricultural wastes (corn fiber and soybean hulls), important differences were noted ( 8 ). The carbohydrate composition of these biomasses is substantially different than the cellulose- and xylan-rich lignocellulose examined here. Most likely, A. bescii and Ac. thermocellus that lack carbohydrate active enzyme(s) needed to deconstruct specific carbohydrates within corn fiber and soybean hulls, respectively. In addition, deviation of line 80 poplar from the fit in Fig. 6 (A and B) is likely due to the incorporation of specific monolignol intermediates (most likely aldehydes) into the final lignin polymers. How hydroxycinnamaldehyde units in lignin contribute to the overall methoxy content correlation remains unclear from these data, and further evaluation of these aldehyde units in lignin is needed. Despite this, most lignocellulosic substrates are primarily cellulose, xylan, and lignols. For these substrates, correlations with methoxy content number are highly predictive. As methoxy content represents the primary barrier to microbial solubilization and conversion of cellulose- and xylan-rich lignocellulose, the recalcitrance of plant biomasses can now be predicted. This establishes specific lignin compositional goals for feedstock engineering for microbial biorefineries. While low methoxy content substrates do not require chemical pre- or post-treatments, opportunity exists to reduce recalcitrance in higher methoxy substrates through chemical means that leverage the divergent lignin composition preferences for chemical and microbial solubilization."
} | 3,356 |
37662333 | PMC10473687 | pmc | 377 | {
"abstract": "Achieving sustainable chemical synthesis and a circular economy will require process innovation to minimize or recover existing waste streams. Valorization of lignin biomass has the ability to advance this goal. While lignin has proved a recalcitrant feedstock for upgrading, biological approaches can leverage native microbial metabolism to simplify complex and heterogeneous feedstocks to tractable starting points for biochemical upgrading. Recently, we demonstrated that one microbe with lignin relevant metabolism, Acinetobacter baylyi ADP1, is both highly engineerable and capable of undergoing rapid design-build-test-learn cycles, making it an ideal candidate for these applications. Here, we utilize these genetic traits and ADP1’s native β-ketoadipate metabolism to convert mock alkali pretreated liquor lignin (APL) to two valuable natural products, vanillin-glucoside and resveratrol. En route, we create strains with up to 22 genetic modifications, including up to 8 heterologously expressed enzymes. Our approach takes advantage of preexisting aromatic species in APL (vanillate, ferulate, and p -coumarate) to create shortened biochemical routes to end products. Together, this work demonstrates ADP1’s potential as a platform for upgrading lignin waste streams and highlights the potential for biosynthetic methods to maximize the existing chemical potential of lignin aromatic monomers.",
"conclusion": "Conclusion This work demonstrates the unique potential for Acinetobacter baylyi ADP1 within sustainable chemistry, specifically with respect to the biotechnological upgrading of lignin feedstocks. ADP1 combines the metabolic utility of the β-ketoadipate pathway with facile genetic engineering. ADP1’s engineerability is clearly demonstrated in the >20 successive, scarless genetic knockouts, including expression of eight separate heterologous enzymes (COMT, UGT, mtn, luxS, metK, CysE*, metA*, and metB) simultaneously. Of note, all of the necessary cloning to generate these strains was carried out by a single graduate student over the course of one year. This scale of cloning was enabled by a combination of ADP1’s native homologous recombination, natural competency, and our recently developed markerless and scarless Cas9-based counterselection method 23 . Such a scope of genetic modification would be challenging for even well-established hosts such as E. coli and S. cerevisiae , and this work provides only a glimpse of the potential for synthetic biology applications. In addition to the facile cloning, ADP1 grows quickly and doesn’t require specialize media or reagents, and thus could be readily adopted by any lab that already conducts E. coli cloning. Although these strengths are promising, it is worth mentioning that this work suggests a heterologous expression capacity restraint in ADP1, which is worth further exploration. In addition, it remains to be seen how ADP1, a strict aerobe, will perform at scale. ADP1 is one among several microbial hosts being explored for lignin valorization, ( Pseudomonas putida 48 – 50 and Rhodoccocus jostii 51 representing two others), and each of these microbes have advantages to offer. It remains to be seen if a single host will emerge as ideal. Continued development of each in the near-term will allow for more flexibility as these determinations are made, especially if synthetic microbial communities are to be explored 9 . The upgrading of aromatic lignin monomers to aromatic products via the metabolic funneling approach is also an important demonstration, specifically our use of ADP1 as a “molecular sieve” where only selected lignin aromatic species are retained. Others have recently suggested taking a more atom economic approach to lignin upgrading 52 . However, previous examples have not used complex feedstock mixtures and included heterologous biosynthetic steps, like shown here. This work clearly demonstrates this potential. Lastly, the natural products made in this study have value in their own right. Vanilla is the most utilized flavor compound globally, and its current supply is primarily provided by chemical synthesis from petroleum derived compounds 53 . Consumer preferences disfavor this approach, yet the alternative of natural sourcing via Vanilla plantifolia is not feasibly scalable. Biotechnological production, as others have pursued 54 – 56 , provides a possible middle ground for obtaining classifiably natural vanilla via a sustainable and renewable process. Utilizing lignin for this application, which already contains vanillin and other chemically proximal species, offers a synthetically advantaged (fewer biochemical steps) route. Even traditional chemical approaches are looking to leverage vanillin synthesis directly from lignin 57 . Likewise, resveratrol is a desirable nutraceutical, and utilizing the p -coumarate in mock APL provides a biochemically advantaged route for its production. Both approaches circumvent the need to take a feedstock down through glycolysis (or another catabolic pathway) and then back up through shikimate or chorismate biosynthesis (and beyond). This study provides a key demonstration for taking greater advantage of lignin’s chemical potential, which will be essential for achieving a biomass-based bioeconomy and will help move towards closed loop carbon cycles via waste valorization.",
"introduction": "Introduction Progress towards sustainable chemical synthesis will require a transition away from the current linear model of manufacturing and consumption, where resources are extracted, fashioned, utilized, and discarded, to one that is instead circular 1 . Non-food biomass has significant potential as a feedstock for sustainable chemical synthesis because of its inherent linkage to circularity through photosynthetic CO 2 fixation and because biomass utilization, in exchange of non-renewable feedstocks, can alleviate ecosystem burden 2 , 3 . To be successful, though, biomass-based processes must be efficient and be capable of generating a broad array of products, including high value products, especially in view of competing traditional processes 4 , 5 . A present and acute need within biomass-based processes is for more complete biomass utilization 6 , 7 . Specifically, lignin utilization has lagged behind that of the other two primary components (cellulose and hemicellulose) 8 . Lignin, which provides the structure for plants, is a heterogenous aromatic polymer, and its complex nature has proved challenging for traditional means of upgrading 9 . Instead, lignin is often treated as a waste product of biomass processes and is burned for heat 10 . Creating new processes that generate value-added products from lignin would represent a significant advance for carbon circularity broadly and biomass circularity in particular. Microbial systems have been proposed as a solution for lignin valorization, as part of a “biological funneling” concept 11 ( Figure 1 ). In this approach, lignin is pretreated to generate a mixture of aromatic monomers, such as p -coumarate, ferulate, vanillate, and p -hydroxybenzoate. Following, this mixture is fed to microbial systems that leverage the β-ketoadipate pathway to “funnel” the various lignin aromatic monomers to tractable starting points for biochemical upgrading. Previous work within the lignin biological funneling paradigm has primarily focused on selectively retaining a single metabolite (i.e. vanillin) 12 , converting all metabolites to commodity chemicals like muconic acid 13 , or building a pathway from the TCA cycle 14 , the entry point of the β-ketoadipate pathway to central carbon metabolism. Few, if any, approaches leverage the existing chemical potential of the available aromatic species. Utilizing this chemical potential to synthesize high value aromatic products from lignin could significantly impact the economic feasibility of lignin-based biomanufacturing and thus biomass utilization as a whole. Here, we use the bacterium Acinetobacter baylyi ADP1 to synthesize the aromatic natural products vanillin-glucoside and resveratrol from the existing aromatic monomer species of a mock alkali pretreated liquor lignin (APL) 15 – 17 ( Figure 1 ). Conscientious of long-term economic considerations, all processes in this work utilized this feedstock and in the context of a minimal medium (M9) with only trace metals supplemented. Vanillin-glucoside and resveratrol were chosen because, as with many valuable natural products, their demand outpaces natural sourcing, where low native yields and arable land requirements constrain scalability 18 – 20 . Motivated by this constraint and consumer preferences away from petroleum-derived products, metabolic engineering approaches have previously been developed as a sustainable solution for scaled production of each of these products, primarily using glucose as a feedstock 21 , 22 . By instead synthesizing these products from lignin aromatics, we take advantage of significantly shorter biosynthetic pathways, simultaneously leveraging ADP1’s native metabolism of ferulate (vanillin-glucoside) and p -coumarate (ferulate). ADP1 was chosen as a host organism because of its possession of the β-ketoadipate pathway for aromatic monomer catabolism and, more importantly, its engineerability that we have previously demonstrated 23 . Biomanufacturing success is largely a function of the ability to modify the host organism toward engineering goals 24 . ADP1 excels in this area, and its facile genetics enabled the creation of strains with up to 22 distinct scarless genetic knock outs and 8 heterologously expressed enzymes. Taken together, our works is a key proof-of-concept demonstration for high-value aromatic product synthesis from waste lignin biomass.\n\nIntroduction of vanillin synthesis pathway to ADP1 Having obtained a strain capable of improved vanillin retention, individual enzymes in the vanillin-glucoside synthesis pathway were tested for activity in ADP1 with two studies used as the basis for the enzymes chosen for testing 22 , 27 . First, catechol O-methyltransferase (COMT) from Homo sapiens was tested and found capable of converting PCA to vanillin in ADP1 ( Figure S3 ). Next, the UDP-glucose dependent glycosyl transferase (UGT) step that converts vanillin to vanillin-glucoside was tested. The glycosylation was tested second as ADP1 maintained some vanillin degradation capacity but is unable to degrade vanillin-glucoside, so any successful conversion of vanillin to vanillin-glucoside should be observable. For this UGT step, UGT72E2 of Arabidopsis thaliana was chosen and found capable of converting vanillin to vanillin-glucoside in ADP1 ( Figure S4 ). Finally, carboxylic acid reductase (Car) from Nocardia iowensis was tested for its ability to convert vanillate to vanillin. Car was paired both with its native phosphopantetheinyl transfer partner ppt and an alternative partner from Bacillus subtilis ( sfp ), which had been found successful in a previous work 27 . Car enzymatic tests were carried out in the context of a strain also expressing UGT to trap any vanillate converted to vanillin at vanillin-glucoside. While both pairings were successful, Car/sfp was the superior partnership, as it showed both greater amounts of vanillin-glucoside and vanillyl alcohol (a re-metabolization of vanillin) and as the full amount of vanillate was utilized ( Figure S5 ). All three enzymes were expressed in the context of Δ16 with COMT integrated at vanAB under Trc expression, UGT integrated at pcaHG under Trc-BCD9 expression, and with Car/Sfp expressed via plasmid. In a 24-hour cultivation in M9 with simplified mock APL, vanillin-glucoside production was observed ( Figure S6 – 7 ).",
"discussion": "Results and Discussion We begin with our biological funneling approach for vanillin-glucoside production from mock APL ( Figure 2 ). Like the latter steps of glucose-based approaches to vanillin or vanillin-glucoside synthesis 22 , 25 – 28 , our synthesis pathway proceeds from protocatechuate (PCA) to vanillate by catechol O-methyltransferase (COMT), from vanillate to vanillin by carboxylic acid reductase (Car), and from vanillin to vanillin-glucoside by a UDP-glucose dependent glycosyltransferase (UGT) ( Figure 2 ). Vanillin-glucoside was prioritized as a final product over vanillin because it has previously been shown less toxic to production hosts, because of its superior aqueous solubility, and as it naturally exists in Vanilla planifolia 22 , 26 , 29 . Our biosynthetic pathway begins with PCA owing to the composition of our feedstock (APL) and ADP1’s metabolism of it. Mock APL, which based on an alkali pretreated corn stover 15 – 17 , contains glucose, acetate, and the aromatic monomers p -coumarate, p -hydroxybenzoate, p -hydroxybenzaldehyde, vanillate, and ferulate. As mentioned, ADP1 metabolizes the aromatic monomers of mock APL via the β-ketoadipate pathway, which funnels these aromatics to either catechol or PCA 30 . Importantly, all of the aromatic monomers in APL belong to the PCA branch of the β-ketoadipate pathway, and are thus funneled to PCA to form the basis for vanillin-glucoside synthesis. The remaining glucose and acetate of APL is used for cell growth and maintenance. It is also worth noting that one APL species (vanillate) is already part of the vanillin-glucoside synthesis pathway and that the first step of ferulate’s native degradation is to another a pathway metabolite (vanillin), allowing maximum utilization of the feedstock. While fundamentally enabling for this study, ADP1’s aromatic funneling also presents a challenge. The capability to degrade various aromatic species to tractable starting points for metabolic engineering necessitates that vanillin-glucoside intermediate pathway metabolites can be metabolized unproductively for growth. To ensure that synthesis the pathway proceeds primarily the forward direction, and that all aromatic species can be utilized productively, we engineered ADP1 to inhibit or prevent PCA, vanillate, and vanillin consumption. First, the enzymes responsible for PCA and vanillate degradation in ADP1 were removed, pcaHG and vanAB respectively. A strain with both pairs of enzymes removed was fed 1 mM p-coumarate and 1 mM vanillin, along with glucose and acetate for growth. As expected, this strain retained PCA (derived from p -coumarate) and vanillate (derived from vanillin) ( Figure 3 ). While removing the genes responsible for vanillate and PCA degradation was straightforward, removing the dehydrogenase activity responsible for vanillin degradation activity was not. While other microbes that possess the β-ketoadipate pathway have annotated vanillin dehydrogenase ( vdh ) enzymes, ADP1 lacks an enzyme with this specific annotation. Therefore, we utilized Psi-BLAST 31 searches of ADP1’s genome with known vanillin dehydrogenases from both Pseudomonas and Sphingomonas species and ald from Corynebacterium glutamicum to identify candidates for deletion ( Tables S1 ). The top most likely candidates based on homology from this search were hcaB (annotated as hydroxybenzaldehyde dehydrogenase) and areC (annotated as benzaldehyde dehydrogenase II). Both hcaB and areC were removed from the strain with vanAB and pcaHG already removed to create the strain Δ2, with Δ2 a reference to the number of putative vanillin dehydrogenase genes removed. When Δ2 was grown on the previously described medium with p -coumarate and vanillin, the retained product profile changed measurably ( Figure 3 ). While some PCA and vanillate (the previously observed oxidation products) still appeared, new species emerged including vanillyl alcohol and p -hydroxybenzyl alcohol (reduction products). However, no vanillin was observed after the 24-hour cultivation, suggesting that alternative degradation pathways were being utilized by ADP1. Others have observed a similar need to remove multiple aldehyde dehydrogenase enzymes to prevent vanillin degradation 27 , 32 . Accordingly, we systematically deleted putative vanillin dehydrogenase enzymes from our candidate list prioritizing those with high homology to known vanillin dehydrogenases and with increased expression from growth on quinate (an aromatic molecule) compared to succinate (non-aromatic molecule) as determined by a prior study 33 ( Table S2 for list of genes removed). At approximately strain Δ16, we identified a strain that retained vanillin along with p -hydroxybenzaldehyde, which is likely degraded by the same promiscuous dehydrogenases. Additional knocks outs neither appeared to improve vanillin retention nor change the retained product profile ( Figure 3 , comparing Δ16 and Δ20, Figures S1 for full set of knockouts up to Δ16). In addition, strains with more knockouts generally showed reduced growth, but only in the minimal medium condition with APL and not in LB ( Figures S1 – 2 ). Therefore, Δ16 was used for subsequent work, as it enabled vanillin-glucoside pathway intermediate retention and thus provided a foundation for vanillin-glucoside pathway engineering in ADP1. Introduction of vanillin synthesis pathway to ADP1 Having obtained a strain capable of improved vanillin retention, individual enzymes in the vanillin-glucoside synthesis pathway were tested for activity in ADP1 with two studies used as the basis for the enzymes chosen for testing 22 , 27 . First, catechol O-methyltransferase (COMT) from Homo sapiens was tested and found capable of converting PCA to vanillin in ADP1 ( Figure S3 ). Next, the UDP-glucose dependent glycosyl transferase (UGT) step that converts vanillin to vanillin-glucoside was tested. The glycosylation was tested second as ADP1 maintained some vanillin degradation capacity but is unable to degrade vanillin-glucoside, so any successful conversion of vanillin to vanillin-glucoside should be observable. For this UGT step, UGT72E2 of Arabidopsis thaliana was chosen and found capable of converting vanillin to vanillin-glucoside in ADP1 ( Figure S4 ). Finally, carboxylic acid reductase (Car) from Nocardia iowensis was tested for its ability to convert vanillate to vanillin. Car was paired both with its native phosphopantetheinyl transfer partner ppt and an alternative partner from Bacillus subtilis ( sfp ), which had been found successful in a previous work 27 . Car enzymatic tests were carried out in the context of a strain also expressing UGT to trap any vanillate converted to vanillin at vanillin-glucoside. While both pairings were successful, Car/sfp was the superior partnership, as it showed both greater amounts of vanillin-glucoside and vanillyl alcohol (a re-metabolization of vanillin) and as the full amount of vanillate was utilized ( Figure S5 ). All three enzymes were expressed in the context of Δ16 with COMT integrated at vanAB under Trc expression, UGT integrated at pcaHG under Trc-BCD9 expression, and with Car/Sfp expressed via plasmid. In a 24-hour cultivation in M9 with simplified mock APL, vanillin-glucoside production was observed ( Figure S6 – 7 ). Enhancing o-methyl transferase activity for PCA Both our own testing ( Figure S7 ) and prior work 25 , 28 , 34 indicate an apparent bottleneck at the first step of the pathway with the conversion of PCA to vanillate. As all the aromatic metabolites present in mock APL are on the PCA side of the β-ketoadipate pathway, this could present an issue for fed-batch experiments where PCA could accumulate and become toxic. Not only is metabolite accumulation generally detrimental in metabolic engineering, but PCA is known to behave as an iron chelator 35 , which could cause further issues. Accordingly, we prioritized improving the conversion of PCA to vanillate. Prior studies in Escherichia coli showed that conversion of PCA to vanillate can be limited by enzyme expression and cofactor availability 27 . Therefore, for an initial experiment, we tested increasing COMT expression using a plasmid, providing the precursor for the methyl donating cofactor s-adenosylmethionine (SAM) that COMT utilizes to convert PCA to vanillate (L-methionine), along with testing the addition of trace metals that have been shown to improve strain performance. Each intervention showed slight improvement for conversion of PCA to vanillate ( Figure S8 ), providing direction for subsequent optimization. To optimize COMT expression, we first modified the expression of the integrated COMT by testing different ribosomal binding site (RBS) variants to replace the initial “agga” RBS. For this we used bicistronic design (BCD) constructs 36 that we had previously validated in ADP1 23 . Increased expression provided modest benefit up to a point, with BCD20 providing a 24% increase in turnover compared to the “agga” RBS. However, the constructs with higher expression strength (BCD14 and BCD9) showed reduced COMT turnover ( Figure S9 ). Interestingly, the high expression strength constructs also showed a sharp reduction in growth ( Figure S9 ). As SAM is required for essential processes in the cell (phospholipid biosynthesis, protein post-translational modification, DNA methylation) 37 , this reduction could be the result of an exhaustion of SAM availability in the cell, which would make improvement of SAM availability critical for pathway performance. Motivated by this possibility, we next sought to test the impact of modifying SAM availability. Leveraging previous work 28 and the metabolic mapping of ADP1 38 , we identified six genes from E. coli that could be incorporated to potentially improve the SAM pool in ADP1, mtn , luxS , metK , CysE , metA , and metB . Figure S10 provides a “map” of SAM related enzymes, considering both ADP1 and E. coli pathways, that could help generate (or regenerate) the cofactor. Worth noting, ADP1 and E. coli have different homoserine to homocysteine pathways, with ADP1 preferring acetylation and E. coli preferring succinylation. In addition, by utilizing heterologous E. coli enzymes in ADP1, we hypothesized that we might avoid endogenous regulation. While we observed a benefit with L-methionine addition ( Figure S8 ), suggesting that the activity of metK may not be limiting, as was the finding of an E. coli study 28 . However, for thoroughness, we still tested the addition of metK and two enzymes involved in the recycling of SAH (the metabolic product of SAM after methyl donation), mtn and luxS. The six E. coli SAM genes were integrated in different groupings and altogether in the context of the Δ16 strain. Each SAM cofactor enzyme underwent a scarless and markerless integration and was integrated with a Trc promoter and lacO operator. In terms of integration location, metK was integrated at ACIAD2929, the feedback resistant mutant metA* at acoD, metB at ACIAD1577-ACIAD1578, the feedback resistant cysE* at dhbA, mtn at quiA, and luxS at calB. Though several combinations showed benefit over the reference strain with no additional SAM enzyme expression, metK improved turnover the most (>4-fold improvement in COMT activity, Figure S11 ). The flux through this enzymatic step was increased to the point that a potential isomer “isovanillate” was observed ( Figure S12 ), where a “shoulder” in the HPLC trace on the vanillate peak potentially represents a promiscuous para O-methylation in place of the desired meta O-methylation 34 . The best overall combination of SAM enzymes was that of luxS, mtn, and metK (given the shorthand name “lmmK”). Though on their own they provided benefit over the reference strain, the inclusion of CysE* and metA* with “lmmK” hindered turnover ( Figure S11 ). Combining the COMT expression and SAM pool optimization, a new strain was constructed from the Δ16 background with COMT under Trc-BCD20 chromosomal expression, the set of “lmmK” SAM enzymes chromosomally integrated, and with UGT chromosomally expressed as before (Trc-BCD9). Curiously, though this new strain showed a 48% improvement over the increased expression of COMT with BCD20 alone, and a 2.58-fold improvement over the reference strain of Δ16, it showed a reduction in activity compared to “lmmk” alone ( Figure S13 ). This result, including the lower OD 600 of this strain compared to “lmmk,” suggests that either a potential total capacity for heterologous expression has been reached or that the SAM pool is still being exhausted. In the context of the full vanillin-glucoside pathway, this strain showed a slight decrease in vanillin-glucoside titer from mock APL compared to the Δ16 reference strain giving 10.9 mg/L compared to 12.7 mg/L in a 24-hour cultivation in M9 with trace metals at the 3 mL scale ( Figure S14 ). However, the strain did show a modest (8%) increase in vanillate compared to the reference strain (p-value 0.074) ( Figure S15 ). Vanillyl alcohol recovery with vanillyl alcohol oxidase During cultivations with the vanillin dehydrogenase knockout strains, vanillin degradation was still observed, even though the primary degradation pathway (from vanillin to vanillate) had been removed. One of the alternative vanillin degradation pathways involves the reduction of vanillin to vanillyl alcohol, with p -hydroxybenzaldehyde being similarly degraded to benzoyl alcohol instead of p -hydroxybenzoate ( Figure S16 ). Both of these secondary degradation products were observed by HPLC in strains with greater numbers of putative vanillin dehydrogenases removed ( Figure 3 ). As it would be desirable to recapture this carbon for vanillin-glucoside production, we tested an enzyme known as vanillyl alcohol oxidase (VAOX) 39 , 40 , which had been previously used in an enzyme cascade to synthesize vanillin 41 , for activity in ADP1. Feeding vanillyl alcohol to ADP1, we observe VAOX activity ( Figure S17 ), indicating that this enzyme could be useful to include in our engineered strain. Based on the results of the COMT optimization in the context of the full pathway, we integrated VAOX into a strain with only metK expressed among SAM cofactor enzymes and COMT expressed under a weak RBS. This strain showed 17.1 mg/L vanillin-glucoside, along with 19.5 mg/L vanillin, for a total of 36.6 mg/L of the two primarily flavor compounds ( Figure 4 ). Resveratrol production In an effort to demonstrate that this approach with ADP1 has applicability as platform for lignin upgrading beyond vanillin-glucoside synthesis, we sought to identify an additional natural product that could be made from aromatic lignin monomers. Because the highest concentration lignin monomer in APL is p -coumarate, we wanted to target a product that could be readily made from this aromatic species. To this end, p -coumarate is only two biosynthetic steps away from the valuable nutraceutical flavonoid resveratrol ( Figure 5 , top), which has been a frequent target of metabolic engineers due to limits for plant extraction (grape) and chemical synthesis 21 . To date, approaches for resveratrol biosynthesis have been primarily carried out in common metabolic chassis strains ( E. coli , S. cerevisiae ) using traditional feedstocks such as glycerol and sucrose, often supplementing L-tyrosine or p -coumarate to the medium 42 , 43 . These approaches leverage amino acid biosynthesis and subsequently convert tyrosine to p -coumarate by tyrosine ammonia lyase (TAL) 44 . Following, p -coumarate is converted to resveratrol by two additional enzymatic steps 45 . First, 4-coumarate-CoA ligase (4CL) is used for CoA addition 44 . Second, resveratrol synthase (stilbene synthase, STS) converts p -coumaroyl-CoA to resveratrol using three units of malonyl-CoA 45 . Though a simple pathway, resveratrol synthesis is intriguing because of its inherent value, as it can function as a proof of concept for the broader flavonoid family of molecules 46 , and because resveratrol can form the basis for further decoration to create derivatives such as the methylated pinostilbene or pterostilbene 47 . Our approach to resveratrol synthesis is shown in Figure S18 . Fortuitously, the first step of resveratrol biosynthesis from p -coumarate and ADP1’s native metabolism of p -coumarate and ferulate are the same CoA ligation ( Figure S19 ). To determine if the activity of the enzyme often used for resveratrol synthesis (4CL of Arabidopsis thaliana ) was interchangeable with ADP1’s native enzyme for this activity (hcaC), we first deleted hcaC . As expected, a Δ hcaC strain of ADP1 was not able to degrade p -coumarate or ferulate ( Figure S20 ). Next, to test if 4CL was interchangeable for hcaC, we added 4CL under plasmid expression (pBAV1k-lacI-Trc-4CL) into Δ hcaC . Complementing the Δ hcaC strain with 4CL restored degradation of p -coumarate and ferulate ( Figure S20 ), suggesting that ADP1 could perform the first step of resveratrol synthesis from p -coumarate without heterologous expression of 4CL. The interchangeability of hcaC and 4CL for CoA-ligation of p -coumarate and ferulate provides both an advantage and a disadvantage. Helpfully, it reduces the number of heterologous enzymes necessary to synthesize resveratrol in ADP1. However, it prevents the metabolite pooling approach used for vanillin-glucoside, where native degradation genes were removed to prevent carbon loss of APL aromatic metabolites to cell growth. Here, the activity of hcaC/4CL cannot be removed, as the pathway to resveratrol would then be incomplete. Moreover, removing the subsequent step in the pathway, hcaA ( Supplemental Figure S19 ), would trap ferulate in its degradation at 4-hydroxy-3-methoxyphenyl-β-hydroxypropyl-CoA (feruloyl-CoA) and would cause a potentially toxic metabolite buildup. We identified vanillin synthase of Vanilla planifolia 29 as a potential solution to this issue. Vanillin synthase directly converts ferulate to vanillin, and previous studies have shown it is selective for ferulate, and does not act on p -coumarate or another related aromatic metabolite caffeic acid 29 . We incorporated vanillin synthase by plasmid expression into Δ hcaC . Though this strain showed a statistically significant decrease in ferulate concentration (p-value 0.0042), and an unexpected statistically significant increase in p -coumarate concentration (p-value 0.00245), the activity was weak resulting in only a 4% decrease in ferulate concentration ( Figure S20 ). We hypothesize the increase in p -coumarate may be due to promiscuous activity of downstream catabolic enzymes for ferulate degradation that can cleave the O-methyl group at the meta position ( vanAB ) and that may be able to cleave the subsequent meta hydroxy group, thus creating p -coumarate from ferulate. Though promising, this strategy will likely only be effective if vanillin synthase can be engineered for superior turnover, and thus was not further pursued in this study. Having established that ADP1’s native metabolism carries out the first necessary step to convert p -coumarate to resveratrol, we added the remaining enzyme, resveratrol synthase (STS) from Vitis vinifera . We compared chromosomal and plasmid-based expression in the context of supplying only p -coumarate or in full mock APL. These experiments showed resveratrol production in both media contexts, but only with plasmid-based STS expression ( Figure 5 ). Chromosomal expression presumably provided insufficient STS expression. While a greater titer was obtained from p -coumarate feeding alone (13.1 mg/L), full APL did show resveratrol production (5.4 mg/L), thus demonstrating ADP1 can successfully synthesize another natural product class (flavonoid/polyketide) from mock APL."
} | 7,906 |
39814725 | PMC11735814 | pmc | 378 | {
"abstract": "Neural reuse can drive organisms to generalize knowledge across various tasks during learning. However, existing devices mostly focus on architectures rather than network functions, lacking the mimic capabilities of neural reuse. Here, we demonstrate a rational device designed based on ferroionic CuInP 2 S 6 , to accomplish the neural reuse function, enabled by dynamic allocation of the ferro-ionic phase. It allows for dynamic refresh and collaborative work between volatile and non-volatile modes to support the entire neural reuse process. Notably, ferroelectric polarization can remain consistent even after undergoing the refresh process, providing a foundation for the shared functionality across multiple tasks. By implementing neural reuse, the classification accuracy of neuromorphic hardware can improve by 17%, while the consumption is reduced by 40%; in multi-task scenarios, its training speed is accelerated by 2200%, while its generalization ability is enhanced by 21%. Our results are promising towards building refreshable hardware platforms based on ferroelectric-ionic combination capable of accommodating more efficient algorithms and architectures.",
"introduction": "Introduction Brain-inspired computing, as an emerging computing paradigm, aims to develop an efficient computing system by learning from the structure and functions of biological nervous systems. Within this field, neuromorphic devices based on memristors, are regarded as important components for constructing brain-like computing systems. They can utilize distinct memristive mechanisms to mimic various forms of biological plasticity. However, these prototype devices primarily focus on implementing the functionalities of individual neurons or synapses 1 . Thus, they lack the bionic capabilities of some important network-level properties, particularly the reusability of neural circuits in biology. According to neural reuse theory 2 , 3 , neural circuits established can be redeployed to different uses, without losing their original functions (Fig. 1a ). This intrinsic mechanism drives organisms to generalize knowledge across various tasks during learning, rather than starting anew each time. Predictably, achieving this property in neuromorphic hardware will enable adaptable repurposing of the existing components across various tasks, thereby allowing for performing more tasks with greater resilience and lower resource consumption (Fig. 1b ). Fig. 1 The hardware implementation schematic of neural reuse, alongside the illustrative diagram of the reconfiguration and refresh strategies. a Schematic of neural reuse, where shared original neural circuits are specialized across multiple tasks through redeployment (blue arrow) and restoration (red arrow) processes. b Schematic hardware implementation of neural reuse, encompassing the refresh and restore process, along with the cyclic configuration between source model and task model. c Schematic of the reconfiguration strategy and the refresh strategy. It encompasses characteristics of switching between volatility and non-volatility, as well as the allocation of them across diverse tasks. d Schematic of the refreshable memristor proposed in this work. The left and right sides of the diagram represent the energy barriers that need to be crossed for ferroelectric polarization reversal and ion migration, respectively. The dynamic allocation of order-disorder phases in CuInP 2 S 6 allows for the circular configuration of two distinct memristive mechanisms within a single material. However, it poses a significant challenge for memristor-based neural networks. As shown in Fig. 1b , it requires the network to accommodate two distinct yet related functions (refresh from source model to task model) while ensuring their respective stability and mutability. This seemingly contradictory yet interrelated property bears a resemblance to the volatile and nonvolatile characteristics exhibited by memristors. This led us to contemplate integrating the complete memristive dynamics process of the memristor into a unified framework. Specifically, the stable source model and the tunable task model can be constructed based on the non-volatile and volatile states of a single device, respectively. Then, through cyclical configuration between the different working modes, refresh and restoration operations are accomplished (Fig. 1b ). Given that, utilization of memristor reconfiguration strategy appears to be a possible solution. However, as illustrated in Fig. 1c , current research on reconfiguration solutions has primarily focused on controlling the conversion between volatility and non-volatility, which is leveraged to handle different types of tasks separately 1 , 4 – 8 . Thus, despite possessing switchable working modes, a single device typically can perform only one mode within one task. This restriction, which divides application scenarios based on memristive properties, has hindered the sharing of general knowledge (represented by non-volatility) across different tasks. Therefore, there is a pressing need to explore a targeted refresh strategy, wherein the volatility and non-volatility of a single device can collaborate on the same task, and its non-volatility can be utilized across multiple tasks. As shown in Fig. 1c , this requires the memristor to possess both two distinctive characteristics: firstly, non-volatility always appears earlier and serves as the foundation of volatility; secondly, non-volatility remains consistent both before and after mode switching. The former ensures that the non-volatile function can be refreshed with seamless conductance changes, while the latter ensures its continued utility across multiple tasks. Albeit a significant challenge, the growing demand for searching innovative memristor materials and mechanisms, the blooming of two-dimensional (2D) layered materials with an atomically thin nature and intriguing physical properties provide a promising material family 9 . Among them, the exceptional current manipulation capabilities of van der Waals ferroionic CuInP 2 S 6 (CIPS) position it as a highly potential candidate (Fig. 1d ). Within this typical room-temperature order-disorder ferroelectric semiconductor, the displacive instability of copper (Cu) leads to two intriguing properties of CuInP 2 S 6 : ferroelectricity, characterized by off-center ordering 10 , 11 , and ionic conduction, characterized by long-range migration disordering 12 – 14 . For ferroelectricity, its electric dipoles primarily arise from the displacement of Cu + ions within the crystal lattice. Below the Curie temperature, spontaneous ordering of electric dipoles produces macroscopic polarization that can be switched by an external electric field 10 . This polarization exhibits excellent retention and has found widespread applications in devices such as ferroelectric diodes 10 and ferroelectric tunnel junctions 15 . On the other hand, Cu + ions also exhibit ionic conduction through incoherent, long-range motion 16 , 17 . This behavior is facilitated by, and can itself generate, atomic disorder 18 . And it typically requires an electric field for sustainment, forms the basis of the volatile memory. Intriguingly, the ordered site serves as both the initial site for this reversible ion motion and its stable site after relaxation, which bears a great resemblance to the mechanism of neural reuse. Given that ferroelectric polarization and ion migration in CIPS share a common origin, as they are essentially mere reflections of the order-disorder dynamics of Cu + ions 18 , this dual memristive mechanism is expected to be highly integrated into space and perform refresh without reconfiguration constraints. Here, we proposed the concept of the refreshable memristor and constructed the device based on the vertical structure with a MoS 2 /CIPS/MoS 2 configuration. Electrical characterization validates its capacity to execute multi-state operations in both ferroelectric polarization and ion migration working modes. Moreover, the ferroelectric state can remain consistent before and after multiple redeployments, which is completely different from existing reconfiguration solutions and aligns seamlessly with the crucial requirement for neural reuse. Furthermore, transfer learning, by leveraging parameter sharing to apply general knowledge across various downstream tasks 19 , 20 , can be regarded as a concrete application of the concept of neural reuse in algorithms. This renders transfer learning well-suited to evaluating the device’s capability to mimic neural reuse. Simulation results based on realistic refreshable device properties show that our device characteristics are sufficient to fulfill the requirements of neural reuse. Compared to the randomly initialized model, repurposing the well-trained source model can yield an accuracy improvement of over 17% while requiring only 60% of the cost for the same duration of training. In multi-task scenarios, neural reuse can accelerate the training speed for new tasks by up to 22 times, effectively enhancing its generalization ability by over 21%. These results indicate that leveraging the distinctive refresh capability of our devices can mimic neural reuse mechanism, thereby narrowing the gap in cognitive ability between memristor-based neural networks and biological nervous networks. Crucially, leveraging this network-level property as the inner driving, this refreshable device empowers memristor-based neural networks to accommodate more efficient algorithms and architectures, thereby substantially expanding the upper limit of their applications.",
"discussion": "Discussion In this work, we have demonstrated the integration of neural reuse with memristor-based neural networks through a variable configuration between the ferroelectric polarization and ion migration. This implies that the division of application scenarios based on memristor properties is no longer a necessity. Instead, they can be integrated into a progressive and more powerful framework, which is lacking in most reported reconfiguration strategies. In addition, through the elaborate modulation of the order-disorder phase in CIPS, the proposed device not only allows for unlocking the dual working modes driven by distinct memristor mechanisms but also ensures ferroelectric polarization remains consistent after undergoing repeated cycles of ion migration and relaxation. Leveraging the refresh ability of the realistic device, further simulations indicate that integrating neural reuse can lead to improvements in the training efficiency and generalization ability of memristor-based neural networks, whether compared to the randomly initialized model or in specific multi-task scenarios. This work opens up the possibility for implementation of neural reuse in memristor-based neural network platforms, thereby loading a more potent inner engine into the hardware."
} | 2,733 |
36178156 | PMC10092204 | pmc | 380 | {
"abstract": "Abstract Cable bacteria are long, filamentous, multicellular bacteria that grow in marine sediments and couple sulfide oxidation to oxygen reduction over centimetre‐scale distances via long‐distance electron transport. Cable bacteria can strongly modify biogeochemical cycling and may affect microbial community networks. Here we examine interspecific interactions with marine cable bacteria (Ca. Electrothrix) by monitoring the succession of 16S rRNA amplicons (DNA and RNA) and cell abundance across depth and time, contrasting sediments with and without cable bacteria growth. In the oxic zone, cable bacteria activity was positively associated with abundant predatory bacteria (Bdellovibrionota, Myxococcota, Bradymonadales), indicating putative predation on cathodic cells. At suboxic depths, cable bacteria activity was positively associated with sulfate‐reducing and magnetotactic bacteria, consistent with cable bacteria functioning as ecosystem engineers that modify their local biogeochemical environment, benefitting certain microbes. Cable bacteria activity was negatively associated with chemoautotrophic sulfur‐oxidizing Gammaproteobacteria ( Thiogranum , Sedimenticola ) at oxic depths, suggesting competition, and positively correlated with these taxa at suboxic depths, suggesting syntrophy and/or facilitation. These observations are consistent with chemoautotrophic sulfur oxidizers benefitting from an oxidizing potential imparted by cable bacteria at suboxic depths, possibly by using cable bacteria as acceptors for electrons or electron equivalents, but by an as yet enigmatic mechanism.",
"introduction": "INTRODUCTION Cable bacteria grow as long multicellular filaments and act as centimetre‐scale electrical conductors in aquatic sediments, providing a conduit for rapid electron transport from deeper sulfidogenic horizons up to an oxic sediment surface (Nielsen et al., 2010 ). Since their discovery about a decade ago (Pfeffer et al., 2012 ), cable bacteria have been observed in a wide range of depositional sedimentary environments (Aller et al., 2019 ; Burdorf et al., 2016 ; Hermans et al., 2020 ; Scholz et al., 2021 ), and frequently at high cell densities (Malkin et al., 2017 ). Especially where these bacteria achieve high biomass, their metabolic activity can drive intense localized changes in pH and strongly influence the cycling of oxygen, sulfur, iron, manganese, phosphorus, carbonate, organic carbon, and trace metals (Huo et al., 2022 ; Rao et al., 2016 ; Risgaard‐Petersen et al., 2012 ; Sulu‐Gambari et al., 2016a , 2016b ; van de Velde et al., 2016 ), with impacts on the ecosystem level (Seitaj et al., 2015 ). The biogeochemical consequences of cable bacteria growth in sediments are not limited to their direct metabolic activities, but additionally involve their interspecific interactions with a myriad of other microorganisms in their complex sedimentary environment (Daghio et al., 2016 ; Liu et al., 2022 ; Scholz et al., 2021 ). Therefore, gaining a better understanding of how cable bacteria interact with co‐occurring microbes would provide a better picture of how these bacteria affect local and system‐scale biogeochemical cycling. In addition to the natural environment, cable bacteria also frequently grow well under laboratory conditions, which readily enable the opportunity to observe their growth dynamics (Schauer et al., 2014 ; Yin et al., 2021 ). In organic‐rich sulfate reducing marine sediments incubated in aerated aquaria, cable bacteria have been observed to grow rapidly to great densities (e.g., up to 25% of microbial cells; Schauer et al., 2014 ). Such dominance suggests an adaptive strategy for competitive exclusion, potentially by an ability to out‐compete other sulfide oxidizers by creating a suboxic space that limits their ability to simultaneously access sulfide together with oxygen or nitrate. By contrast, evidence from recent studies has pointed to the growth of cable bacteria as a stimulating factor for other microbes, promoting the activity of other bacterial taxa and/or biogeochemical processes (Kessler et al., 2019 ; Vasquez‐Cardenas et al., 2015 ), and driving an increase in microbial network complexity (Liu et al., 2022 ). These findings point to a different ecological role for cable bacteria—as providers of resources for other microbes, for example by solubilizing ferrous iron (Kessler et al., 2019 ), or by acting as an electron acceptor for chemoautotrophy, as hypothesized by Vasquez‐Cardenas et al. ( 2015 ) and Otte et al. ( 2018 ), and in freshwater or low salinity systems, by recycling sulfate at suboxic depths (Huo et al., 2022 ; Liu et al., 2022 ; Sandfeld et al., 2020 ). Cable bacteria may perform different ecological functions in different sediment horizons. In the oxic zone, they may be effective competitors in minimizing microbial diversity, while in the deeper suboxic zone, cable bacteria may act as facilitators for the growth of other microbes, by modifying the environment in ways favourable to more microbes and/or potentially by serving as an electron acceptor. Cable bacteria may particularly benefit other microbes at a depth where oxidants and labile organic carbon are depleted. The objective of this study was to identify key interspecific associations between cable bacteria and members of their associated indigenous community. We examined the successional changes in the microbial community through depth over a time course of 46 days and, in suboxic sediments, contrasted communities with and without cable bacteria across time. In one set of sediment cores, marine cable bacteria (Ca. Electrothrix) were allowed to grow unimpeded. In a separate subset of sediment cores, 0.2 μm polycarbonate barrier filters were embedded at 0.5 cm depth, which allowed for porewater diffusion, but blocked the downward growth of cable bacteria below the filters, as previously demonstrated by Pfeffer et al. ( 2012 ). Sediment cores were destructively sampled and sectioned at six time points, from which cell counts (single cells and cable bacteria filaments) were performed in tandem with DNA‐ and RNA‐based amplicon sequencing of 16S rRNA genes. We computed the product of the fractional read abundance of the RNA‐based amplicons and the direct cell counts (‘FRAxC activity’), and examined correlations of this activity metric for each amplicon sequence variant (ASV) against the FRAxC activity of cable bacteria (Ca. Electrothrix). Based on previously observed relationships with cable bacteria and their geochemical influence on sediments, we hypothesized that associations with cable bacteria activity would be depth‐dependent. We hypothesized negative correlations between cable bacteria and sulfur‐oxidizing bacteria in the oxic zone (i.e., exhibiting competition) and positive correlations with sulfur‐oxidizing chemoautotrophs in the suboxic zone (i.e., exhibiting facilitation or syntrophy). We furthermore hypothesized positive correlations with bacteria that have high iron requirements, due to the anticipated porewater ferrous iron production associated with cable bacteria activity.",
"discussion": "DISCUSSION In this study, we aimed to identify which microbes' growth and activity were correlated with the growth and activity of Ca. Electrothrix (marine cable bacteria). Accurately identifying such relationships remains a major challenge in molecular ecology, and this may be particularly true in anaerobic sediments where biases known to exist (e.g., nucleic acid extraction efficiency) are particularly challenging to constrain. Examining RNA amplicons of the 16S rRNA gene is frequently used as a sensitive marker to examine the active microbial community because they exist in greater copy numbers than DNA and the RNA pool broadly scales with cellular growth rate (Molin & Givskov, 1999 ). RNA amplicons were examined here to obtain greater sensitivity in identifying relationships between actively growing microbes, though we acknowledge that inherent trade‐offs exist between sensitivity to detect relationships where they exist and robustness against spurious correlations. The compositional sequencing data were normalized to cell counts, to mitigate the extreme bias that would have been imposed by the compositionality of the data (Gloor et al., 2016 ), particularly given the dominance of Ca. Electrothrix reads. To minimize spurious correlations, we considered only taxa that exhibited a trend with time and a change on the last date, when cable bacteria abundance and activity had crashed. However, given that the RNA‐based FRAxC metric is affected by the growth rates of the microbes, which vary with environmental conditions and was not in steady, it would not be appropriate to contrast the FRAxC values across studies. An underlying assumption in this experiment with cable bacteria was that the activity and growth of most microbial taxa in the sediments would tend to slow down during the incubations, given that we were not adding new organic matter to maintain steady‐state conditions. This is broadly supported by the decreases over time in DOU in the sediments without cable bacteria (Figure S1 ). Cable bacteria are unusual in such incubation conditions, in that they are stimulated by the dissolution of FeS, which in turn is enhanced by their activity (i.e., their growth stimulates positive feedback). Thus, positive correlations of a given ASV in the FRAxC metric with cable bacteria are likely to reflect a real stimulation of the activity of the given taxon. Conceptual model of cable bacteria bioenergetics We report here that cell‐specific transcriptional activity and inferred rates of cable bacteria growth demonstrated depth‐dependency. We estimated doubling times of 42 h among the most surficial sediment section (i.e., in the upper 0.5 cm of sediment, which included cells in the oxic zone and cells immediately below), with doubling times decreasing with depth, down to 10 h among the deepest growing cells. These rates bound estimates from a previous study which estimated marine cable bacteria doubling times of 20 h during exponential growth, based on an assumption of uniform cell division rates (Schauer et al., 2014 ). We additionally observed that cable bacteria cells exhibited higher transcriptional activity (as a ratio of RNA‐ to DNA‐based amplicons) with increasing depth, down to a maximum observed at 1.5–2.0 cm, possibly reflecting the depth distribution of sulfide flux to cable bacteria cells in our particular experimental sediment. Increasing rates of bacterial cell proliferation are strongly correlated with higher ribosomal content and transcriptional activity (Scott et al., 2010 ). Our experimental observations of cable bacteria are consistent with this paradigm, though further confirmation by single filament experimentation is warranted. At the community level, we found up to 65% of the RNA‐based amplicon reads were attributed to the activity of Ca. Electrothrix, which was extraordinary. The dominance of the RNA amplicon pool by Ca. Electrothrix is consistent with these bacteria having a fast growth rate and an attendant large cellular ribosomal content. This finding aligns with the concept of cable bacteria exhibiting an opportunistic life history and an ability to dominate sediment microbial metabolism (Nielsen et al., 2010 ; Risgaard‐Petersen et al., 2014 ). Bringing these observations together, we can describe our results in the context of an emerging conceptual model of cable bacteria physiology and bioenergetics. This conceptual model proposes that cable bacteria exhibit a division of labour between cells, based not on cell differentiation, but instead based on the different environmental conditions experienced by individual cells along the filament (Geerlings et al., 2020 ; Kjeldsen et al., 2019 ). Cells inhabiting the suboxic zone putatively perform anodic sulfide oxidation by a reversed canonical sulfate reduction pathway similar to Desulfurivibrio alkaliphilus (Thorup et al., 2017 ), which is coupled to energy conservation (ATP formation) and generation of reductants required for carbon‐fixation via the Wood‐Ljungdahl pathway (Kjeldsen et al., 2019 ). These suboxic zone (anodically‐operating) cells demonstrate high levels of biosynthesis (Geerlings et al., 2020 ; Kjeldsen et al., 2019 ). Here, we add supporting evidence that these cells exhibit higher transcriptional activity and have higher inferred growth rates. By an incompletely understood process, electrons generated from this sulfide oxidation are transferred onto the filament's conductive network, comprising nickel‐protein fibres housed in a periplasmic space that run the length of the cable bacteria filament, which rapidly ushers electrons from the suboxic zone to the oxic zone (Bjerg et al., 2018 ; Boschker et al., 2021 ; Cornelissen et al., 2018 ; Jiang et al., 2018 ). Cells located in the oxic zone function cathodically, transferring electrons from the conductive network to oxygen (or nitrate/nitrite) via periplasmic reductases, producing water and consuming protons (Kjeldsen et al., 2019 ; Marzocchi et al., 2022 ). Cells in the oxic zone do not appear to have membrane‐associated reductases required to conserve energy (Kjeldsen et al., 2019 ). Instead, the function of cable bacteria cells in the oxic zone appears to be primarily to dissipate electrons arising from the sulfide oxidation. Single filament experiments have demonstrated that cable bacteria cells are adapted for exceptionally high rates of cell‐specific oxygen reduction (i.e., electron ‘flaring’) in the oxic zone (Scilipoti et al., 2021 ). Cable bacteria filaments are motile (Bjerg et al., 2016 ) and can adjust their position to optimize their depth distribution based on changing environmental conditions (Malkin et al., 2014 ). Our observations here affirm that a small proportion (13% or less) of the total length of cable bacteria cells are occupying the oxic zone, similar to previous estimates of 8%–10% of the filament cells (Schauer et al., 2014 ; Scilipoti et al., 2021 ). In addition to affirming this conceptual model of cable bacteria bioenergetics based on cell positioning across redox zones, we also found that the interactions between cable bacteria and their microbial community differed by redox zone. In the oxic zone, we detected more negative (33 ASVs) than positive (10 ASVs) associations, while in the suboxic zone, we observed the converse, with more positive (9 ASVs) than negative (2 ASVs) associations. In the remaining discussion, we interpret the putative nature of these interactions, which are summarized in Figure 6 . In short, we find correlative evidence to support predator–prey relationships with cable bacteria, primarily in the oxic zone (Figure 6A ); microbial taxa that are likely benefiting from the geochemical impacts attributable to cable bacteria (Figure 6B ); and putative sulfur‐oxidizing taxa that may be inhibited by cable bacteria in the oxic zone by competition, but apparently benefitting from cable bacteria in the suboxic zone, potentially by an energy facilitation or syntrophy (Figure 6C,D ). FIGURE 6 Conceptual diagrams of potential microbial interactions with marine cable bacteria. These include (A) predator–prey interactions, for example by obligate intracellular predators affiliated to Bdellovibrionota and/or swarming facultative predators affiliated to Myxococcota and Bradymonadales; (B) cable bacteria acting as ecosystem engineers by altering local geochemistry, such as stimulating bacteria with high iron requirements (e.g., Magnetovibrio ) by acidic dissolution of FeS, or altering distribution of acidity, promoting alkaliphilic bacteria at the sediment surface (e.g., Thioalkalispira ) and impeding sulfur disproportionators (e.g., Desulfocapsa ) by acidification of the suboxic zone; (C) competitive interactions such as with single celled sulfur oxidizers (e.g., Thiogranum , Sedimenticola ), and (D) syntrophic or facilitative interactions, for example as hypothesized for putative chemoautotrophic sulfur oxidizers potentially benefitting energetically from the activity of cable bacteria (e.g., Thiogranum , Sedimenticola ). Cable bacteria are illustrated as green multicellular filaments. The overlying oxygenated water is illustrated as blue, and the sediment, with increasingly reducing conditions at depth, is illustrated by brown to black gradient. Panel (B) includes a typical depth profile of pH and sulfide removal attributable to cable bacteria long distance electron transport activity. Putative predator–prey interactions Cable bacteria activity was positively associated with several bacterial taxa that have predatory lifestyles. In the oxic zone, these included members of the genus‐level OM27 clade (Bdellovibrionota), the genus Haliangium (Myxobacteria), and unidentified members of the order Bradymonadales (Desulfobacterota), and in the suboxic zone, this included an ASV affiliated with the Class Polyangia (Myxobacterium). Myxobacteria in general, and Haliangium in particular, are facultative predators, described as capable of degrading live or dead cells by secreting hydrolytic enzymes, and specialize in congregating as swarms with ‘wolf‐pack’ attack strategies (Fudou et al., 2002 ; Muñoz‐Dorado et al., 2016 ; Pérez et al., 2016 ). Members of Bradymonadales too, unlike other members of the phylum Desulfobacterota, exhibit a predatory lifestyle similar to Myxococcales, with facultative predatory swarming behaviour (Mu et al., 2020 ; Wang et al., 2015 ). Bdellovibrionota is described as motile and aerobic predators, which typically replicate by infiltrating the periplasmic space of Gram negative bacteria, where they grow and divide, and which they subsequently lyse (Pérez et al., 2016 ; Sockett, 2009 ). They are generally considered obligate predators of Gram negative bacteria, with only rare exceptional observations of host‐independent replication (Sockett, 2009 ). Different species or strains of Bdellovibrionota can exhibit specialist, generalist, or versatile associations with their prey (Chen et al., 2011 ). As Gram negative bacteria, cable bacteria are potential prey of Bdellovibrionota, at least where oxygen is present. We considered whether the associations between the potential predators and cable bacteria suggest a potential control on cable bacteria population abundance or activity separately in the oxic and suboxic zones. It has been observed repeatedly that cable bacteria biomass and activity can decline rapidly after a period of prolific growth, in both laboratory studies (Rao et al., 2016 ; Schauer et al., 2014 ) and field studies of seasonally hypoxic coastal basins (Malkin et al., 2022 ; Seitaj et al., 2015 ). The decline of cable bacteria has been primarily attributed to a bottom‐up control: diminishing stocks of FeS, which are dissolved by the activity of cable bacteria, but are not replenished under static lab conditions or during seasons of low bottom water oxygen (Larsen et al., 2015 ; Schauer et al., 2014 ; Seitaj et al., 2015 ). Yet, other hypotheses have also been posed to explain the rapid decline, including viral attack (Kjeldsen et al., 2019 ). Cable bacteria genomes exhibit adaptations to viral predation, including the presence of restriction enzyme and CRISPR sequences (Kjeldsen et al., 2019 ). Additionally, the conductive network of the cable bacteria exhibits properties of a ‘fail safe’ network, ensuring continuity of electrical conduction in the event of individual cell injury (Eachambadi et al., 2020 ), which may be an adaptation to predation pressure. Nevertheless, in the suboxic zone, it seems unlikely that bacterial predators could be principally responsible for controlling cable bacteria population sizes in sediments. The only predator correlated with cable bacteria activity was a Myxobacterium (Class Polyangia ), a group that is not obligate predators, and which had low RNA‐based fractional amplicon read abundance (maximum 0.03%). By contrast, in the surface sediments, we observed positive associations between cable bacteria activity and multiple putative bacterial predators, including abundant Bdellovibrionota ( OM27 clade). The putatively obligate predators of the OM27 clade in particular were observed at up to 2.8% of the RNA‐based amplicon reads in surface sediment on Day 20, during peak cable bacteria biomass and activity, suggesting a potentially quantitatively important impact on cable bacteria (Figure 6A ). During a bloom of cable bacteria observed in the Chesapeake Bay (following the sedimentation of the spring phytoplankton bloom), there was an uptick in amplicons affiliated with the OM27 clade (up to 0.6% of DNA‐based amplicon reads; increasing from below detection prior to the sedimentation event; Malkin pers. obs .; NCBI Bioproject PRJNA613483), further suggesting this association may be ecologically relevant. Cable bacteria filaments appear to minimize their exposure to the oxic zone, beyond meeting their requirement for electron discharge (Scilipoti et al., 2021 ). Our results suggest that the oxic environment may be additionally hostile to cable bacteria cells due to these motile bacterial predators. The extent to which these aerobic predators may inhibit cable bacteria activity, or inflict a mortal threat to the entire filament by infecting and lysing cathodic cells, is worthy of further exploration, and invites the hypothesis that cathodic cells may be sacrificial and require continual replacement. Along with testing the relative importance of bottom‐up resource control on cable bacteria (e.g., by sulfide and FeS availability), and top‐down control by viral predation, the impacts of bacterial predation on cable bacteria activity and population density in the oxic zone are worthy of further research. Cable bacteria as ecosystem engineers Several of the correlations between Ca. Electrothrix activity and other microbes were consistent with cable bacteria acting as ecosystem engineers, modifying the local biogeochemical environment in ways that were favourable to some taxa. At the sediment surface, there was a strong positive association between Ca. Electrothrix and the genus Magnetovibrio (Class Alphaproteobacteria). This genus is described as magnetotactic, microaerophilic or anaerobic, and chemolithoautotrophic, capable of using thiosulfate and sulfide as electron donors coupled with oxygen or nitrous oxide as terminal electron acceptors (Bazylinski et al., 2013 ). This genus is also described as having a high growth requirement for iron (Bazylinski & Frankel, 2004 ), due to the presence of magnetosomes, composed of magnetite (Fe 3 O 4 ) or greigite (Fe 3 S 4 ) (Simmons & Edwards, 2007 ; Yan et al., 2012 ). We infer that the ASVs affiliated with this genus in our dataset were bacteria with magnetosomes and that this association was potentially driven by cable bacteria activity increasing the availability of dissolved ferrous iron (Larsen et al., 2015 ; Rao et al., 2016 ; Risgaard‐Petersen et al., 2012 ). Dissolved ferrous iron production attributed to cable bacteria anodic activity stimulates DNRA activity in sediments (Kessler et al., 2019 ), though whether this involves shifts in the microbial community composition has not been previously reported. In a separate experiment, Magnetovibrio were notably enriched and formed biofilms on graphite rods embedded in marine sediments (11% 16S rRNA gene reads), which the authors attributed to alleviation of iron limitation at the rod surfaces (Matturro et al., 2017 ). Otte et al. ( 2018 ) additionally found positive correlations between cable bacteria and ASVs of putative ferrous iron oxidizers affiliated to Pedomicrobium , Hoeflea , Chlorobium , and Rhodopseudomonas , based on field surveys of surface sediments. Taken together, these observations are consistent with cable bacteria enabling greater availability of ferrous iron, and in turn stimulating the activity of bacteria with either high iron requirements for biosynthesis or as an electron donor (Figure 6B ). At suboxic depths, positive associations were additionally detected between Ca. Electrothrix activity and several sulfate‐reducing Desulfobacterota, including ASVs affiliated to the family Desulfobulbaceae , the order Desulfobacterales , and the genus Desulfatiglans . Desulfobacterota is the dominant contributors to dissimilatory sulfate reduction in the environment, with many isolates of this phylum capable of performing chemoorganotrophic sulfate reduction and/or dissimilatory reduction of sulfur compounds of intermediate redox state (Waite et al., 2020 ). Desulfosarcina , Desulfatiglans , and SEEP‐SRB1 are canonical sulfate‐reducing bacteria capable of utilizing diverse carbon substrates, with some isolates capable of degrading methane and other hydrocarbons (Jochum et al., 2018 ; Petro et al., 2019 ; Watanabe et al., 2017 ). Experiments have demonstrated that sulfide oxidation carried out by cable bacteria, combined with enhanced anion diffusion generated by their electric field, enriches sulfate at suboxic depths (Huo et al., 2022 ; Rao et al., 2016 ; Risgaard‐Petersen et al., 2012 ), which in turn can stimulate sulfate reduction, at least in sediments where sulfate reduction is limited by sulfate supply at depths occupied by cable bacteria (Marzocchi et al., 2020 ; Sandfeld et al., 2020 ). The association between sulfate reducers and cable bacteria activity has implications for biogeochemistry and bioremediation, for example by suppressing methanogenesis (Scholz et al., 2020 ), and stimulating the anaerobic degradation of organic compounds, including hydrocarbons (Liu et al., 2022 ), at least in low salinity systems. Studies examining the influence of cable bacteria on sediment biogeochemical cycling and microbial composition have found contrasting results vis à vis whether there are shifts in the composition of the community associated with increased sulfate supply (Sandfeld et al., 2020 ; c.f. Liu et al., 2022 ). Whether the putative influence of the cable bacteria activity on sulfate reducer activity in our experiment was due to an increase in sulfate, decrease in sulfide, or some other factor, is not certain. Multiple environmental factors shape the spatial distribution and niche partitioning among sulfate reducers (Marshall et al., 2019 ), but there is no evidence at this time for a specialized taxon‐specific relationship between cable bacteria and sulfate reducers. Finally, the effects of cable bacteria on the distribution of porewater acidity may have strong impacts on their associated microbial communities. There was a negative association observed between the FRAxC activity of Ca. Electrothrix and Desulfocapsa , which was notable as the only negative association among the Desulfobacterota at suboxic sediment depths. Rather than performing sulfate reduction, this genus is specialized for sulfur disproportionation (Finster et al., 2013 ). We infer this negative relationship is attributable to the increased porewater acidity generated by the anodic activity of cable bacteria (i.e., from pH 7.4 to pH 5.4; Figure 2 ). In similar sediments, but at circumneutral pH, sulfur disproportionation provides a low energy yield, estimated to be near the minimum energy required for ATP synthesis (Müller & Hess, 2017 ). Greater acidity diminishes the thermodynamic energy yield of sulfur disproportionation, which we estimate makes this environment unfavourable for their growth (Table S2 ). Additionally, Thioalkalispira was positively associated with cable bacteria activity at the sediment surface. The type species of the genus, T. microaerophila , is alkaliphilic (Sorokin et al., 2002 ). The increase of this ASV in the surface sediments is consistent with a preference for alkaline conditions, which are imparted by cable bacteria cathodic activity. Interactions with sulfur‐oxidizing bacteria The Gammaproteobacterial genera Thiogranum and Sedimenticola were negatively correlated with Ca. Electrothrix in surface sediments, while ASVs affiliated to these genera were positively associated with Ca. Electrothrix in suboxic sediments. The genus Thiogranum and the type species Thiogranum longum are described as chemolithoautotrophs capable of oxidizing reduced sulfur compounds (thiosulfate, sulfite, elemental sulfur, sulfide, and tetrathionate), with an obligate requirement for oxygen (Mori et al., 2015 ). Sedimenticola is described as chemolithoautotrophs, capable of growing on thiosulfate, sulfide, tetrathionate, and elemental sulfur using oxygen, nitrate, nitrite, or selenate as electron acceptors (Flood et al., 2015 ; Narasingarao & Häggblom, 2006 ). We infer the negative association between these single cell sulfur‐oxidizing bacteria and cable bacteria in the surface sediments is a consequence of substrate competition (Figure 6C ). These single cell sulfur‐oxidizing bacteria are traditionally thought to require simultaneous access to reduced sulfur compounds and oxygen or nitrate (Jørgensen & Nelson, 2004 ). Cable bacteria, by efficiently removing porewater sulfide in proximity to the sediment surface and thereby creating a suboxic space, appear to be highly effective at outcompeting these single cell microbes near the sediment surface. More perplexing, however, is the positive association observed between Ca. Electrothrix and members of these genera ( Sedimenticola and Thiogranum ) at suboxic depths. Oxygen and nitrate have a high redox potential and are consumed rapidly in sulfate reducing sediments (Llobet‐Brossa et al., 2002 ; Marzocchi et al., 2014 ). Selenate additionally has a low crustal abundance and is not likely a quantitatively significant electron acceptor in these sediments (Stolz & Oremland, 1999 ). These compounds were therefore unlikely to be available below 0.5 cm depth, at the suboxic depths sampled. One hypothesis to explain their positive association with cable bacteria was first proposed by Vasquez‐Cardenas et al. ( 2015 ) who postulated that cable bacteria may be used as an electron acceptor by other chemoautotrophic bacteria affiliated to Gammaproteobacteria or Campylobacterota (formerly Epsilonproteobacteria) at depth. The concept proposes that the conductive fibre network of cable bacteria could be an environmental electron sink, similar to a poised anode. In microbial fuel cells, anodes select for microbes with a specialized ability to perform extracellular electron transport (‘exoelectricigens’), some of which have the specialized ability to exchange electrons with other phylogenetically distant microbes as a form of syntrophy (i.e., ‘direct interspecies electron transfer’; DIET; Summers et al., 2010 ; Rotaru et al., 2014 ; McGlynn et al., 2015 ; Wegener et al., 2015 ; Li et al., 2017 ). In recent years, diverse pathways for extracellular electron transport have been uncovered among phylogenetically distant Gram negative and Gram positive bacteria, Archaea, and Eukarya (e.g., Yang et al., 2021 , and reviewed in Lovley & Holmes, 2022 ), broadening our understanding of the adaptive significance and biogeochemical consequences of this process. Additionally, some fermenters may be electro‐active, benefitting energetically from interactions with electrodes in bioreactors, by various mechanisms including by secreting redox‐active compounds that may mediate interactions with anodes (reviewed in Schievano et al., 2016 ). Taken together, there has been a transformational change in our understanding of the plausibility that some bacteria may be capable of utilizing other bacteria as electron acceptors under natural conditions. In the particular case of whether anodic cells of cable bacteria may accept electrons from other sediment associated microbes remains speculative. Although cable bacteria have been observed in close contact with sediment anodes, suggesting the potential for an electro‐active association (Reimers et al., 2017 ), the nature of the relationship remains to be confirmed. Likewise, none of the correlations with cable bacteria was found with known exoelectricigens (Lovley & Holmes, 2022 ; Verma et al., 2021 ), and there is currently no evidence to confirm Sedimenticola or Thiogranum as having electroactive associations. Nevertheless, Sedimenticola was previously observed as a quantitatively significant component of biofilms on anodic zones of graphite rods embedded in marine sediment (9.6% 16S rRNA gene reads), which the authors hypothesized was an electroactive association (Matturro et al., 2017 ). It is possible these Gammaproteobacteria have a previously undocumented ability to perform DIET and are using cable bacteria directly as electron acceptors. Alternatively, these bacteria may have other previously undocumented metabolic capacities for anaerobic growth, and benefitted from cable bacteria growth for other reasons, such as sulfide removal. These bacteria could also conceivably benefit energetically from cable bacteria growth without invoking DIET, if cable bacteria were acting as a sink for electron equivalents. Sachs et al. ( 2022 ) recently observed that in freshwater and aquifer sediments, cable bacteria proliferation was linked with an enrichment of putative fermenters affiliated to Firmicutes , Bacteroidetes , and Spirochaetes at suboxic depths, consistent with cable bacteria acting as a sink for electron equivalents via unidentified redox‐active mediators. Given that microbes in complex sedimentary environments commonly exhibit metabolic capacities that extend beyond their observed capacities in culture (Dyksma et al., 2016 ; Lenk et al., 2012 ; Mußmann et al., 2017 ), the interactions between cable bacteria and these putative chemoautotrophic Gammaproteobacteria at suboxic depths is worthy of further exploration. Disentangling the mechanisms of these associations may be fundamental for understanding the biogeochemical role of cable bacteria in marine sediments."
} | 8,537 |
36686915 | PMC9811985 | pmc | 382 | {
"abstract": "Oil–water separation using porous superhydrophilic materials is a promising method to circumvent the issue of oil-polluted water by separating water from oil–water mixtures. However, fabricating metal-based porous superhydrophilic materials with stable superhydrophilicity that can recover their strong hydrophilicity and have acceptable oil–water separation efficiency without complex external stimuli is still a challenge. Inspired by the anti-wetting behavior of broccoli buds, this study successfully fabricated metal-based superhydrophilic and underwater superoleophobic porous materials by hydrothermal treatment of stainless steel meshes (SSMs) combined with magnetron sputtering of metallic Ti and W. The process was then followed with annealing at 300 °C for 4 hours. The effects of coating materials, annealing temperature, and surface structure on the wetting behavior of the prepared meshes were studied and analyzed. The modified meshes exhibited unique broccoli-like microstructures coated with thin TiO 2− x N x /WO 3 films and showed superhydrophilicity with a 0° water contact angle (WCA) and underwater superoleophobicity with underwater oil contact angles (UOCAs) higher than 155°. They also maintained strong hydrophilicity for more than three weeks with WCAs of less than 13°. Besides, they could recover their initial superhydrophilicity with a 0° WCA after post-annealing at 80 °C for 30 minutes. Notably, the broccoli-like structures and the strong hydrophilic coatings contributed to a significant water flow rate ( Q ) of 3650 L m −2 h −1 and satisfactory oil–water separation efficiency of 98% for more than 15 separation cycles toward various oil–water mixtures. We believe that the presented method and fabricated material are promising and can be applied to induce hydrophilicity of various metallic materials for practical applications of oil–water separation, anti-fouling, microfluidic transport, and water harvesting.",
"conclusion": "4. Conclusions In summary, this article presented a facial method to construct bio-inspired metal-based superhydrophilic meshes by the combined actions of the hydrothermal treatment, magnetron sputtering, and annealing. The experimental results showed that annealing at 300 °C oxidized both W and TiN films and resulted in the formation of the WO 3 and N-doped TiO 2 . The annealed films contained more hydroxyl groups which enhanced their strong hydrophilicity. Besides, when the prepared films were subjected to post-annealing at 80 °C after three weeks, they recovered their initial high hydrophilicity. These results have shown that the combined action of magnetron sputtering and annealing is a promising technique for constructing strong metal-based hydrophilic materials that can recover their strong hydrophilicity without the need for complex external stimuli. In addition, the broccoli-like structure was crucial to enhance the hydrophilicity of the TiN300 film and endowed the SSMs with stable and strong hydrophilicity, as well as satisfactory oil–water separation.",
"introduction": "1. Introduction Oil is an essential source of energy and it is used daily in most industrial facilities. However, the unsafe practices of waste oil disposal, machinery operation, oil exploration, and oil transportation can result in oil-contaminated water, which causes severe damage to the ecosystem and affects the quality of human life. 1–3 Recently, porous superhydrophilic materials, including membranes, meshes, foams, and sponges have shown effective results in reducing the negative effects of oil-contaminated water by separating the water from oil–water mixtures. 4–6 These materials can separate and adsorb the water from the oil–water mixture due to the high affinity to water and high repellence to oil simultaneously. 7 Through the study of some biological surfaces with excellent superhydrophilicity and underwater superoleophobicity in nature, it is found that the synergy of micron/nano scale structure and hydrophilic surface chemical components is the key for achieving outstanding porous superhydrophilic materials for oil-in-water separation applications. 8–11 On this basis, feasible approaches and materials have been designed and developed to construct superhydrophilic and underwater superoleophobic materials. 12–14 For example, Xin et al. fabricated superhydrophilic and underwater superoleophobic stainless steel meshes by depositing TiO 2 using liquid phase deposition (LPD). 15 Their results showed that the fabricated meshes could separate various types of oil-in-water mixtures with a separation efficiency of 99%. Dong et al. also coated a stainless steel mesh with TiO 2 using the sol–gel method. 6 The original stainless steel mesh exhibited poor wettability to the water with a 124° WCA. In contrast, the WCA of the treated SSM decreased to almost 0°, and extremely low oil adhesion was demonstrated. In the oil–water separation tests, the treated mesh was found to separate oil from the oil–water mixture with a separation efficiency of 99% even in corrosive and harsh environments. Ye et al. fabricated superhydrophilic surfaces using a different approach based on the femtosecond laser ablation of micro-holes drilling of the titanium foil. 10 The WCA on the original titanium foil was 66.3 ± 2.1°. However, when the water droplet came into contact with the treated titanium foil, it wetted the surface with a WCA of 0°. In addition, the prepared titanium foil separated the oil from the oil–water mixture with an efficiency of over 98%. This improvement in wettability and separation efficiency after laser ablation was attributed to the formation of TiO 2 and the increase in surface roughness. Despite the feasibility of the recently utilized methods in fabricating superhydrophilic and underwater superoleophobic materials for oil–water separation, there are still challenges in preparing porous metal-based superhydrophilic surfaces with desired surface structure, stable wetting behavior, low cost, and minimal secondary pollution. 15–18 As compared to other methods of fabricating superhydrophilic materials, hydrothermal treatment is inexpensive, with no secondary pollution, and can create superhydrophilic surfaces with special nano/microstructures. 19,20 However, the application of the hydrothermal treatment in previous studies have been limited to inducing the hydrophilicity of strong hydrophilic materials such as TiO 2 and SiO 2 , which limits its employment for practical applications. 14,16 Therefore, using the hydrothermal treatment to fabricate the desired surface structures, followed by the deposition of strong hydrophilic materials is a promising technique to control the hydrophilicity of various porous metallic surfaces for oil–water separation applications. Among various strong hydrophilic coating materials, metal-based photosensitive materials such as TiO 2 and WO 3 have been widely utilized to fabricate superhydrophilic and underwater superoleophobic materials due to their ability to recover their initial strong hydrophilicity under UV or visible light illuminations. 21,22 For instance, Gao et al. prepared an ultrathin composite film that exhibited superhydrophilicity and underwater superoleophobicity after UV light illumination. 23 The film was based on a single-walled carbon nanotube and TiO 2 nanocomposite network (SWCNTs)/TiO 2 and prepared by the sol–gel process. The as-prepared (SWCNTs)/TiO 2 showed poor hydrophilicity with a WCA of 82°. After one hour of irradiation by UV light, however, the WCA decreased to nearly 0° and the surface exhibited underwater superoleophobicity with UOCAs higher than 150° toward different types of oil droplets. However, it should be noted that the hydrophilicity and photo-induced superhydrophilicity of materials of this kind depend on their chemical compositions. Therefore, constructing metal-based superhydrophilic materials capable of maintaining strong hydrophilicity and recovering satisfactory separation efficiency without complex external stimuli is crucial. Inspired by the anti-wetting behavior of broccoli buds, this work presented a simple method to construct metal-based superhydrophilic and underwater superoleophobic porous materials by the hydrothermal treatment of stainless steel meshes combined with the magnetron sputtering of metallic Ti and W materials. The coated meshes were finally post-annealed at 300 °C for 4 hours. As a result, the modified SSMs exhibited a unique broccoli-like microstructure with excellent superhydrophilicity and robust superoleophobicity underwater. Notably, they could maintain strong hydrophilicity for more than three weeks with WCAs of less than 13°, and after annealing at 80 °C for 30 minutes, the WCAs decreased to 0° again. Meanwhile, they demonstrated satisfactory oil–water separation efficiency of 98% toward various oil–water mixtures, even in corrosive HCl and NaOH solutions. Since the proposed technique is neither complex nor expensive, we believe that this method can be used to induce the hydrophilicity of metal-based surfaces for the practical application of oil–water separation, antifouling, microfluidic transportation, and water harvesting.",
"discussion": "3. Results and discussions 3.1 Wettability and characteristics of broccoli buds The wetting behavior of some creatures in nature, including fish scales and the latus leaves, has always helped researchers to design and construct materials with special wetting behavior for practical applications. 24,25 For instance, this study observed that broccoli buds exhibit strong hydrophobicity in the air with a WCA of nearly 134° as shown in Fig. 2(a) . According to the Wenzel and Cassie models, this unique wetting behavior of broccoli buds might result from their capability of locking the air (air has low surface free energy) in their nano/microscale structures as shown in Fig. 2(b) . 26 Fig. 2 Wetting mechanism of (a) broccoli flower head, (b) hydrophobic surfaces, (c) underwater oleophobic surfaces. Comparable to the wetting mechanism of hydrophobic surfaces, the underwater Cassie model states that the wetting mechanism of superhydrophilic and underwater superoleophobic surfaces depends on locking water (water surface tension is higher than oil surface tension) rather than the air in the nano/microscale structures as shown in Fig. 2(c) . 27 In this respect, this study attempted to follow the anti-wetting properties of broccoli buds to fabricate underwater superoleophobic surfaces by first constructing broccoli bud-like structures on stainless steel meshes via the hydrothermal treatment. The process was then followed by fabricating thin TiO 2− x N x /WO 3 (TiN300) layers on the hydrothermally treated stainless steel meshes to enhance their ability to lock water (rather than air) so that they can exhibit superoleophobicity in the oil–water–solid interfaces system. 3.2 Wettability of the prepared coatings on flat surfaces We began to demonstrate the optimal strong hydrophilic coating by depositing thin TiN and TiN300 layers on flat stainless steel samples. As shown in Fig. 3(a and b) , Videos S1 and S2, † the flat stainless steel sample with a TiN layer exhibited excellent hydrophilicity in the air with a water contact angle of about 8°. After annealing at 300 °C, the surfaces demonstrated strong hydrophilicity and the WCA decreased to 2°. Interestingly, Fig. 3(c) shows that the thin TiN300 layer could maintain strong hydrophilicity for more than two weeks with a WCA of about 17° without external stimuli and less than 10° after UV light illumination for 30 minutes provided by a pressurized mercury lamp with a wavelength of ( λ = 365 nm). It also could recover its initial strong hydrophilicity after annealing at 80 °C for 20 minutes as shown in Fig. 3(c) , Videos S3 and S4. † Fig. 3 WCA measurements on TiN and TiN300. (a) WCA on TiN on day one, (b) WCA on TiN300 on day one, and (c) change of WCAs on TiN and TiN300 from day one to day 60 under the effect of dirt contamination, post-annealing, and UV light illumination. According to the data from (XPS) in Fig. 4(a–e) , SEM in Fig. 4(f and g) , and XRD pattern in Fig. S3, † the TiN and TiN300 exhibited different chemicals composition and similar surface structures. Therefore, the difference in the WCAs could be attributed to the different chemical compositions. The XPs survey data shown in Fig. 4(a) reveals the presence of tungsten, titanium, oxygen, nitrogen, and carbon on both layers. However, the atomic percentage of these elements changed after annealing at 300 °C. As illustrated in Table S1, † the atomic percentage of Ti and O increased while the atomic percentage of C, N, and W decreased after annealing at 300 °C. The corresponding Ti (2p) high-regulation spectra in Fig. 4(b) show that the TiN layer was composed of Ti and TiN components. After annealing at 300 °C, the Ti (2p) region showed three peaks at 458.34, 463.93, and 456.3 eV binding energies. These peaks were assigned respectively to Ti 4+ 2p 1/2 , Ti 4+ 2p 3/2 , and Ti–N bonding, which indicated that the TiN thin film was oxidized by oxygen atoms in air. 28,29 We further investigated the N (1s) states in TiN and TiN300 films. As shown in the corresponding high-resolution N (1s) spectra in Fig. 4(c) , the TiN300 film showed a main peak at 395.58 eV, which originated from the N–Ti–O bond in TiN300. 30–32 These results indicate that nitrogen successfully doped the TiO 2 atoms in TiN after annealing at 300 °C. 32,33 The core level spectra of O (1s) of TiN and TiN300 films were also investigated and shown in Fig. 4(d) . We can see that the O 2 spectra of TiN and TiN300 films exhibited main peaks at 529.90 and 529.79 eV corresponding to W \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 and Ti–W–O bonds, respectively. 34 Moreover, the shoulder peaks at 531.91 and 531.73 eV in TiN and TiN300 films were assigned to the hydroxyl group of the surface. 5,28 Finally, the elemental state of the tungsten in TiN and TiN300 was investigated. As shown in Fig. 4(e) , the peaks at 35.58 and 37.68 eV (BE) were respectively assigned to W4f 7/2 and W4f 5/2 , which indicate that the W atoms were in the W 6+ state. 35 Fig. 4 XPS and SEM data for TiN and TiN300 films. (a) XPS survey, (b–e) the corresponding high-resolution of (b) Ti (2p), (c) O (1s), (d) N (1s), and (e) W (4f), and (f and g) SEM data for TiN and TiN300 films. Following the XPS and SEM data analysis, the strong hydrophilicity of the TiN300, therefore, might result from its high hydroxyl group content, low carbon content, the presence of N-doping in TiO 2 , and the presence of WO 3 under N-doped TiO 2 . As reported by previous studies, the presence of hydroxyl groups increases hydrogen bonding with water which is a key rule for enhancing hydrophilicity. In addition, N-doped TiO 2 and WO 3 under TiO 2 were found to exhibit better photo-induced superhydrophilicity compared to pure TiO 2 , even under typical light illumination, due to the photosensitivity of N-doped TiO 2 and WO 3. 36–39 In the case of the restoration of strong hydrophilicity after annealing at 80 °C and ultraviolet light illumination, it may be due to the removal of the low surface energy dirt contamination. 9,23 The photo-induced hydrophilicity of the TiO 2 /WO 3 can be explained by Fig. 5 . During the UV illumination, the electrons generated in TiO 2 transfer to WO 3 , while the holes generated in WO 3 transfer to TiO 2 due to the large positive potential of the balance and conduction bands of WO 3 . 40,41 Meanwhile, the photo-induced holes in TiO 2 react with water and generate – OH radicals, which oxidize organic compounds on the surface of TiO 2 and increase the hydrogen bonds with water, which determines the hydrophilicity. 41,42 Fig. 5 Schematic illustration of the photo-induced superhydrophilicity of the thin TiN300 film. 3.3 Surface characterization and wetting behavior of stainless steel meshes The thin TiN300 layer was then fabricated on the hydrothermally treated SSMs. As shown in Fig. 6(a) , the hydrothermally treated for 12 hours SSM with TiN300 film was covered with flake nanoparticles and irregular broccoli-like hierarchal structures. By reducing the hydrothermal treatment time to 8 hours, the SSM with a thin TiN300 film was successfully covered with dense and uniform broccoli-like structures as shown in Fig. 6(b) . In addition, each broccoli-like microstructure was totally covered with bud-like nanoparticles. Fig. 6 SEM and WCA measurements of the SSMs (a) SEM of the hydrothermally treated for 12 hours SSM with TiN300 film, (b) SEM of the hydrothermally treated for 8 hours SSM with TiN300 film, (c) a water droplet on the unprocessed SSM, (d and e) time sequence of a water droplet spreading on the modified SSM. The corresponding wetting investigations of the unprocessed SSM and the hydrothermally treated for 8 hours SSM coated with a thin TiN300 film are presented in Fig. 6(c–e) . We can see that the unprocessed stainless steel mesh in Fig. 6(c) exhibited poor wettability with a WCA of about 78°. In contrast, the modified SSM in Fig. 6(d and e) showed remarkable superhydrophilicity with a 0° WCA, which was less than the WCA on the TiN300 film on flat stainless steel samples. Besides, it absorbed the water droplet in less than one second, which confirms their superhydrophilicity (Video S5 † ). Subsequently, we carried out oil contact angle measurements and oil adhesion tests for the SSMs before and after modification. Before the tests, all the stainless steel meshes were pre-wetted with deionized water. As shown in Fig. 7(a and c) , the unprocessed SSM exhibited poor self-cleaning and underwater oleophobicity with UOCAs less than 120°. The oil contaminated the mesh and strongly adhered to its surface, even after immersion in a water bath. In contrast, the modified SSM experienced satisfactory anti-fouling performance and underwater oleophobicity. The oil droplet on the modified SSM slid off immediately after immersion in water as shown in Fig. 7(b) . This anti-fouling was confirmed by the underwater oil contact angle (UOCA) measurements and oil adhesion tests shown in Fig. 7(d and e) . We can see that the modified SSM expressed UOCAs higher than 155° toward paraffin liquid and 1,2-dichloroethane droplets, and also no trace of oil was observed after the oil adhesion test as shown in Fig. 7(e) . Fig. 7 Oil contamination tests (a) paraffin liquid on the as-received SSM, (b) paraffin liquid on the modified SSMs (c) UOCA on the unprocessed SSM, (d) UOCA on the modified SSMs (e) paraffin liquid adhesion test on the modified SSMs. The fascinating superhydrophilicity and underwater superoleophobicity of the freshly modified SSMs resulted from the abundant hydroxyl group in TiN300 and the unique broccoli bud-like rough structure. The rule of surface roughness in inducing hydrophilicity can be explained by the Wenzel model shown in Fig. 8(b) . Simply, the increase of surface roughness after the hydrothermal treatment increased the solid–water contact area, which enhanced the hydrophilicity. 43 The WCA on the rough surface can be calculated using eqn (2) as follows: 2 cos θ r = r cos θ where θ is the water contact angle on flat surfaces, and r is the roughness factor, the ratio of the unfolded area to the projected surface area, r = 1 for the flat surface and >1 for the rough surface. Fig. 8 Schematic illustration of the wetting states (a) flat surface with water in air, (b) rough surface with water in air, (c) rough surface with oil in water. Underwater, the anti-oil adhesion behavior of the modified SSMs can be explained based on the modified underwater Cassie–Baxter model. 44 As shown in Fig. 8(c) , the strong hydrophilic TiN300 film and the presented broccoli bud-like structures absorbed the water to form a thin water-film, which prevented the oil adhesion due to the repulsive behavior of water toward oil. The underwater oil contact angle on a rough-hydrophilic surface ( θ OW r ) can be evaluated using the Cassie–Baxter equation as follows: 45 3 cos θ OW r = f cos θ OW + f − 1 where f is the fraction of the surface which is in contact with the oil droplet, and θ OW is the underwater oil contact angle on the flat surface, which can be described by the modified underwater Young's equation as follows: 46 4 where γ OW , γ WA , and γ OA are the oil–water interface tension, water–air interface tension, and oil–air interface tension, respectively. θ O and θ W are the contact angles of oil and water in the air, respectively. 3.4 Oil-in-water separation test The oil–water separation mechanism using superhydrophilic materials can be explained using the Young–Laplace theory and Fig. 9(a) as follows: 45,47 5 where Δ P W is the water break-through pressure, the critical pressure applied on the porous materials to allow the water to permeate. γ LV is the water interface tension and d is the pore radius. Finally, B is the geometric pore coefficient, B = 1 for cylindrical pores and 0 < B < 1 for non-cylindrical pores. Fig. 9 Oil–water separation mechanism and efficiency (a and b) oil–water separation mechanism using a superhydrophilic mesh, (c) oil–water separation process using the modified SSMs, and (d) oil–water separation efficiency. According to eqn (5) and Fig. 9(a) , the water can penetrate the porous material without external pressure as the intrinsic WCA on the porous material is less than 90°. On the other hand, if the WCA is much higher than 90°, external pressure is required to force the water to penetrate the porous material. Similarly, the ability of a hydrophilic porous material in blocking oil in the oil–water–sold interfaces system can be explained by the following equation: 6 where ΔP O is the oil break-through pressure in the oil–water–solid interfaces system, γ OW is the water/oil interfacial tension, and θ OW is the UOCA on the flat surface. Based on eqn (6) , when θ OW is higher than 90°, external pressure is required to allow the oil to penetrate the porous material. In addition, eqn (5) and (6) show that the pore size is an essential factor that affects the oil and water penetration. When the pore size ( d ) is much smaller than the water capillary length (WCL), a continuous water-film (colored blue in Fig. 9(b) ) can be formed between the adjacent wires. 48 This thin water film can drive the water through the porous material due to the capillary effect. 49 Meanwhile, it can block oils due to the repellent property of water toward oil. 7 However, if the pore size is much larger than the capillary length of the water, the water film can break off so that both water and oil may penetrate. 50,51 Besides, the continuous water-film can only block a certain amount of oil before it is disrupted by the pressure of the blocked/accumulated oil. The actual break-through pressure (Δ P o-act ) related to the accumulated oil on porous materials can be calculated using the following equation: 5 7 Δ P o-act = ρgh max where ρ is the density of the oil, g is the acceleration of the gravity, and h max is the maximum height of the oil that the porous material can block before the oil infiltrate through it. The flux permission ( J ) of porous materials is an important characteristic to determine their penetration capacity. The relation among the permission flux ( J ), porosity ( ε ), pores radius ( r P ), viscosity of the liquid ( μ ), and pressure drop Δ P can be elucidated using Hagen–Poiseuille's equation as follows. 47 8 According to eqn (8) , the flux permission will increase as the pore size, porosity, and applied pressure increase. However, as mentioned earlier, if the pore size is larger than the water capillary length, the continuous water-film between the adjacent wires may break off and affect the oil–water separation efficiency. Therefore, for achieving high oil–water separation performance, the pressure drop (Δ P ) in eqn (8) must be greater than the breakthrough pressure of water (Δ P W ) and smaller than the breakthrough pressure of oil (Δ P O ). Finally, the actual flow rate ( Q ) of the water passing porous materials can be estimated using eqn (9) as follows: 52 9 Q = V / At where V is the volume of the water passing the porous materials, A is the flow area, and t is the time for the specific volume of water to completely pass the porous materials. To test the oil–water separation efficiency of the SSMs before and after modification, three oil–water mixtures were prepared by mixing 30 mL of deionized water with 10 mL of different oils, including paraffin liquid, 1,2-dichloroethane, and engine oil. The separation process is shown in Fig. 9(c) . Before the separation tests, we pre-wetted all the SSMs with deionized water to prevent oil adhesion on their surfaces. The pre-wetting process revealed that the modified SSM exhibited high water flow rate ( Q ) of about 3650 L m −2 h −1 , compare with 1633 L m −2 h −1 for the unprocessed SSM (Video S6 † ). During the oil–water separation tests, the waited time between each separation cycle was maintained at one minute. Finally, the blocked oil was collected to estimate the amount of oil in the penetrated water using eqn (1) . The results showed that the unprocessed SSM experienced poor oil–water separation performance of less than 40% toward the oil–water mixtures as shown in Fig. S5(a). † In contrast, the modified SSM exhibited considerable separation efficiency of 98% for the first 10 separation cycles and decreased to less than 85% after 20 separation cycles as shown in Fig. 9(c) , Video S7, and Fig. S5(b). † This separation efficiency was recovered to 95% for another 10 separation cycles after ultrasonic cleaning in petroleum ether, ethanol, and deionized water for 10 minutes each, and finally, UV illumination under a high-pressure mercury lamp for 30 minutes as shown in Fig. 9(c) . It was also observed that the modified mesh could block a certain amount of oil to a certain height ( h max ) before it allowed the oil to penetrate as shown in Fig. S5(c). † The maximum amount of the accumulated oil was measured for the engine oil with a maximum high ( h max ) of about 17 cm and equivalent to actual break-through pressure ( ΔP o-act ) of 1.451 kPa. Table 2 shows the h max of each blocked oil and their Δ P o-act . Finally, the separation efficiency toward oil-in-corrosive aqueous solutions, mainly 1 M HCl and 1 M NaOH, was also investigated and shown in Fig. 9(c) and S5(d). † We can see that the modified mesh could keep high separation efficiency of about 97% for the first five separation cycles. The worst separation efficiency was observed for the soybeans oil-in-1 M NaOH. The actual oil break-through pressure and corresponding h max of the modified SSMs Type of oil Engine oil Paraffin liquid Soybean oil Parameter Δ P o-act (kPa) \n h \n max (cm) Δ P (kPa) \n h \n max (cm) Δ P (kPa) \n h \n max (cm) Value 1.451 17 1.319 16 1.125 12.5"
} | 6,887 |
33869078 | PMC8044998 | pmc | 383 | {
"abstract": "Bacterial biofilms are important medically, environmentally and industrially and there is a need to understand the processes that govern functional synergy and dynamics of species within biofilm communities. Here, we have used a model, mixed-species biofilm community comprised of Pseudomonas aeruginosa PAO1, Pseudomonas protegens Pf-5 and Klebsiella pneumoniae KP1. This biofilm community displays higher biomass and increased resilience to antimicrobial stress conditions such as sodium dodecyl sulfate and tobramycin, compared to monospecies biofilm populations. P. aeruginosa is present at low proportions in the community and yet, it plays a critical role in community function, suggesting it acts as a keystone species in this community. To determine the factors that regulate community composition, we focused on P. aeruginosa because of its pronounced impact on community structure and function. Specifically, we evaluated the role of the N-acyl homoserine lactone (AHL) dependent quorum sensing (QS) system of P. aeruginosa PAO1, which regulates group behaviors including biofilm formation and the production of effector molecules. We found that mixed species biofilms containing P. aeruginosa QS mutants had significantly altered proportions of K. pneumoniae and P. protegens populations compared to mixed species biofilms with the wild type P. aeruginosa . Similarly, inactivation of QS effector genes, e.g. rhlA and pvdR , also governed the relative species proportions. While the absence of QS did not alter the proportions of the two species in dual species biofilms of P. aeruginosa and K. pneumoniae , it resulted in significantly lower proportions of P. aeruginosa in dual species biofilms with P. protegens . These observations suggest that QS plays an important role in modulating community biofilm structure and physiology and affects interspecific interactions.",
"introduction": "Introduction Bacterial biofilms are made up of surface-attached or suspended aggregates encased in self-produced extracellular polymeric substances (EPS). Biofilms in natural systems normally contain a rich mixture of species. In addition to the species level diversity, biofilms can also contain cells that are physiologically diverse due to various chemical gradients (e.g. oxygen, carbon source etc.) in different parts of the biofilm (exposed surface vs. interior) ( Davey and O’Toole G, 2000 ). Biofilm communities have been shown to display a number of emergent properties, including enhanced stress and antimicrobial tolerance, increased biomass and reduced genetic variation relative to comparable monospecies biofilms ( Flemming et al., 2016 ). Such effects are not observed for planktonic communities suggesting that these behaviors are biofilm specific. Biofilm formation may enable such responses due to the ability of species to optimize their spatial organization, e.g. to colocalize with collaborating species or to avoid competitors. The biofilm matrix plays a key role in these processes, enabling and stabilizing localization. The matrix further facilitates the establishment of gradients that result in for the formation of microniches that further allow for specialized growth, e.g. growth of anaerobes under otherwise oxygenic conditions ( Flemming and Wingender, 2010 ). In this context, it is clear that microorganisms can modify their environment for optimal growth. Interactions between individual species that contribute to the assembly and stability of a multi-species biofilm are still not well characterized. It is also clear that, just as observed in macroscale ecosystems, successionary processes are important for microbial communities. For example, in an oral biofilm, the initial colonizer tends to be Streptococcus sp., followed by the bridging species Fusobacterium nucleatum and finally by Porphyromonas gingivalis ( Kommerein et al., 2018 ). Another example is the association of Aggregatibacter actinomycetemcomitans utilizing lactate produced by other oral bacteria such as Streptococcus gordonii to form biofilms of higher biomass ( Brown and Whiteley, 2007 ). Thus, it is clear that bacteria undergo ecological processes that are similar to those described for higher organisms. In this study, we investigated a 3-species biofilm community consisting of Pseudomonas aeruginosa PAO1, Pseudomonas protegens Pf-5 and Klebsiella pneumoniae KP1. These three species are reported to occur together as a community in the gut of Bombyx mori and in other environments along with various bacterial species ( Chazal, 1995 ; Anand et al., 2010 ). Moreover, this mixed-species biofilm is highly reproducible in structure and viable cell counts under well-defined conditions, and therefore suitable for molecular studies into community dynamics ( Jackson et al., 2001 ; Stoodley et al., 2001 ; Lee et al., 2014 ). Previously, we showed that this mixed-species community generated higher biomass and resistance to stress conditions, such as antibiotics (tobramycin), detergent sodium dodecyl sulfate (SDS) and predation (unpublished), than individual biofilms ( Lee et al., 2014 ). Further, these effects were only observed when the community grew in a spatially organized biofilm and were not apparent for planktonic communities. Interestingly, P. aeruginosa , arguably one of the best studied model biofilm forming organisms, was the least abundant, between 1-10% of the entire community. When the percentage of P. aeruginosa was artificially increased by substituting the wild type strain with a more competitive, natural genetic variant, many of the emergent properties, e.g. stress resistance of the community, were lost. This would argue that, despite its low overall abundance, P. aeruginosa exerts a disproportionate effect on the rest of the community. Such species are frequently described as keystone organisms. We therefore wanted to explore the mechanisms by which this keystone organism controls community biofilm development. One of the mechanisms used by P. aeruginosa PA01 to regulate biofilm formation, maturation and interspecies interactions is quorum sensing (QS). QS is a cell-cell communication system where bacteria produce and respond to signals to coordinate gene expression at the population level. There are four primary QS gene circuits reported for P. aeruginosa PAO1, the AHL-dependent Las and Rhl systems and the quinolone dependent PQS and IQS signaling systems ( Lee et al., 2013 ; Lee and Zhang, 2015 ). The Las QS system, which produces and responds to 3-Oxododecanoyl)-L-homoserine lactone (3-oxo-C12-HSL), is the dominant system that can regulate both the Rhl and PQS systems ( Gambello and Iglewski, 1991 ). The Rhl system produces and responds to N-butanoyl-L-homoserine lactone (C4-HSL) while the PQS system mainly produces and responds to 2-heptyl-3-hydroxy-4-quinolone (PQS) ( Ochsner and Reiser, 1995 ; Pearson et al., 1995 ; Lee and Zhang, 2015 ). QS controls the expression of multiple effector molecules that have been shown to play important roles in surface colonization, nutrient acquisition and virulence for monospecies biofilms. Most studies of the role of QS in species interactions have focused on cross-species signaling. For example, it has been shown that signal production by Pantoea agglomerans or Erwinia toletana could rescue virulence in a signal deficient mutant of Pseudomonas savastanoi ( Hosni et al., 2011 ). Similarly, it has been shown that P. aeruginosa and Burkholderia cepacia display signal cross talk in a mouse lung infection model ( Riedel et al., 2001 ). While there are several reports of the role of QS in controlling expression of compounds with antimicrobial activity ( Moons et al., 2006 ; Lesic et al., 2009 ; Abdel-Mawgoud et al., 2010 ), the overall role of QS as a regulator of interspecies interactions with mixed species biofilms is relatively poorly understood, especially in the context of QS regulated effectors. Here, we report a role for P. aeruginosa PAO1 AHL QS in determining the composition of the three species biofilm community consisting also of P. protegens Pf-5 and K. pneumoniae KP1. We have generated isogenic, double-mutants of P. aeruginosa PAO1 in Las and Rhl systems and determined its effect on composition and structural properties of the 3-species biofilm.",
"discussion": "Discussion In this work, we investigated the role of QS in interspecies interactions in a defined 3-species bacterial community during biofilm formation. We observed that the proportions of K. pneumoniae and P. protegens were dictated by the presence or absence of a functional P. aeruginosa QS system. We also show that there is no effect of deletion of the various QS genes on the relative fitness of the strains, further suggesting that it is the loss of QS function that drives the changes in community composition and structure. This effect was surprising as P. aeruginosa represents only 1-10% of the mature biofilm community, suggesting that even at a relatively low abundance, this species can significantly modulate community interactions. QS was shown to play a significant role in specifically modulating community composition between P. aeruginosa and P. protegens , with little to no effect on K. pneumoniae . We also showed that several downstream QS targets cumulatively mediate this competitive effect. The complemented lasRrhlR mutant completely restored known QS regulated phenotypes, e.g. swarming, elastase activity and protease activity in in vitro assays. In addition, complementation also restored competitive advantage over P. protegens in growth assays on agar surface ( \n Figure 4 \n ). While the relative abundances of the complemented strains were not significantly different from the wild type when grown in the mixed species biofilm, they were none the less slightly lower than for the wild type ( \n Figure S2 \n ). The lower abundances achieved for the complemented strains might be due to higher and earlier expression of the QS genes from the multi-copy plasmids used for complementation. In addition to the amount of signal production, the level of signal receptor, LasR, production is known to be a critical factor in QS mediated gene expression linked to biofilm formation ( Schuster et al., 2003 ; Rutherford and Bassler, 2012 ). This may in part be due to the expression of LasR and RhlR at late logarithmic and early stationary phase in planktonic cultures. Thus, given that the receptor genes were expressed on a multicopy plasmid, it is likely that the receptors were produced earlier in the complemented strain than normal. This could lead to the expression of QS regulated functions at a lower cell density by the complemented strain, which might consequently lead to its earlier detection by the competing bacterial species, P. protegens , likely leading to its removal and lower abundance. Therefore, optimal timing of expression of QS regulators LasR and RhlR might play an important role in the competition with P. protegens during early biofilm formation. \n K. pneumoniae KP1 and P. protegens Pf-5 carry one and two luxR homologs, respectively, but they do not produce AHLs. It is possible that both K. pneumoniae and P. protegens in multispecies communities might sense and respond to foreign AHLs to regulate biofilm formation. This has previously been reported for multispecies biofilms formed by P. aeruginosa and Burkholderia cepacia in the lungs of CF patients ( Van Delden and Iglewski, 1998 ; Huber et al., 2001 ; Riedel et al., 2001 ). Indeed, cross-species or cross-kingdom signal perception is the most commonly reported role of QS for communities. However, both of the P. aeruginosa lasRrhlR and lasIrhlI mutants show similar proportions in the 3-species community as do the individual QS effector mutants. This suggests that it is the P. aeruginosa QS regulon that mediates interaction with the other two species and not simply signal perception by those species through their LuxR homologs. However, it will be interesting to study whether AHLs directly influence biofilm formation by P. protegens and K. pneumoniae in future studies. In dual species biofilms, there was no effect of QS on the relative abundance of either K. pneumoniae or P. aeruginosa ( \n Figure 2 \n ) under the conditions used in this study. Also, the K. pneumoniae - P. protegens interaction appeared to be neutral, where both species were present in equal abundance (data not shown). However, QS provided a competitive advantage to P. aeruginosa when cultivated with P. protegens ( \n Figure 3 \n ). Although there are several reports on the role of QS in laboratory dual species co-culture systems and it is predicted to influence more complex community dynamics, there is not much direct evidence for the latter, which has been shown here ( Mazzola et al., 1992 ; Flemming and Wingender, 2010 ; Tan et al., 2014 ). Specifically, P. aeruginosa interaction with B. cepacia sp., A. tumefaciens , S. aureus and B. cenocepacia has been shown to be due to QS regulated antimicrobial production and AHL eavesdropping in planktonic cultures ( Lewenza et al., 2002 ; An et al., 2006 ; Costello et al., 2014 ; Smalley et al., 2015 ). In Agrobacterium tumefaciens-P. aeruginosa dual species biofilms, the wild type P. aeruginosa completely blanketed A. tumefaciens and progressively excluded A. tumefaciens from the biofilm while a P. aeruginosa QS mutant did not. This was controlled by antimicrobial production and swarming ( An et al., 2006 ). We propose that P. aeruginosa - P. protegens interactions in dual species biofilms is a two-way interaction involving functions from both bacteria which provide them competitive advantage and results in mutual exclusion. A number of QS regulated P. aeruginosa products such as rhamnolipids, hydrogen cyanide, pyocyanin and pyoverdine are known to be important for competition with other bacteria; these effects are likely to be multifactorial with some antimicrobials having a greater effect ( Smalley et al., 2015 ). Similarly, P. protegens is also known to produce a cocktail of both antifungal and antibacterial secondary metabolites including 2,4-diacetylphloroglucinol, pyoluteorin, pyrrolnitrin, extracellular protease, hydrogen cyanide and siderophores pyochelin and pyoverdine ( Paulsen et al., 2005 ). It has been proposed that the natural function of multi-drug resistance efflux pumps in P. aeruginosa could be to excrete a particular class of secondary metabolite ( Poole, 1994 ). It is thus possible that QS regulated efflux systems also protects P. aeruginosa against one or more antimicrobials produced by P. protegens and future work will address these mechanisms. Our results with QS and target mutants revealed that most of the mutants except aprD were proficient for biofilm formation in batch cultivation assays ( \n Figure S4 \n ). Previous reports indicate that P. aeruginosa QS is required for biofilm formation. Specifically, the Las system has been shown to control progression from reversible to irreversible attachment whereas the Rhl system is responsible for biofilm maturation ( Jaffar-Bandjee et al., 1995 ; Davies et al., 1998 ). However, in our experiments, monospecies biofilms, formed by lasIrhlI and lasRrhlR mutants under flow-cell conditions did not show any difference compared to P. aeruginosa wild type in terms of biovolume and thickness ( \n Figure S8 \n ). Moreover, among the QS target mutants, pqsA , pqsE and rhlA have also shown to be defective for biofilm maturation ( Yang et al., 2009 ; Darveau et al., 2012 ; Hajishengallis et al., 2012 ). The effects of Pqs system on biofilm structure was shown to be iron dependent. Although rhlA mutant has been shown to form flat biofilms lacking heterogeneity of wild type biofilms in the flow-cell system, in the same study it has been reported that in microtiter dish assays the rhlA mutant is more proficient at early colonization (up to 24 h) than is the wild type ( Darveau et al., 2012 ). Active iron uptake has been shown to be important for P. aeruginosa biofilm architecture as pvdS and pvdA mutants form thin biofilms ( Banin et al., 2005 ; Yang et al., 2009 ). However, pvdA and pvdE mutants form nearly normal biofilms in microtiter dish assays ( Kang and Kirienko, 2017 ). Our experiments to test biofilm formation by QS target mutants was carried out under batch culture conditions. As described above, there are differences in the medium and carbon source/s, cultivation temperature and cultivation system between the previous studies with these mutants and the work we have presented here; this might account for the observed biofilm formation phenotypes. For example, Davies et al. ( Davies et al., 1998 ) showed that the QS defect in biofilm formation by P. aeruginosa could be overcome by growing the mutant in complex (LB) rather than defined media. In order to narrow down the processes involved in competition, we investigated interactions between P. aeruginosa and P. protegens in dual and 3-species biofilms where the wild type P. aeruginosa was substituted by mutants of several QS regulated genes, including rhlA and pvdR . rhlA mutant was significantly less competitive than wild type in dual species community with P. protegens ( \n Figure 5 \n ). Interestingly, when wild type P. aeruginosa was substituted by the rhlA mutant in the 3-species community, there was a significant increase in the biomass of the rhlA mutant ( \n Figure 6 \n ). Mutants in rhlA are defective for the production of the biosurfactant rhamnolipid which promotes motility and affects the architecture of monospecies P. aeruginosa biofilms ( Davey et al., 2003 ). Rhamnolipids might affect both cell-cell and cell-surface interactions ( Neu, 1996 ). In fact, in the wild type, when rhamnolipids are produced, they might prevent self-colonization as well as colonization by other species. This inhibition is lacking in the rhlA mutant community. Only pvdR showed trends of alteration of both K. pneumoniae and P. protegens similar to the QS mutants in 3-species biofilms, although in dual species pvdR community this difference was not significant. PvdR is involved pyoverdine transport and recycling in P. aeruginosa and its mutants accumulate pyoverdine in the periplasmic space and are unable to acquire ferric iron efficiently via pyoverdine pathway ( Imperi et al., 2009 ). This suggests that optimal iron uptake and possibly use of heterologous siderophores might be important for maintaining the 3-species community composition as fluorescent Pseudomonads are known to produce multiple pyoverdines ( Cornelis and Matthijs, 2002 ). Together, these results suggest that QS regulated effectors affect the community composition in different ways. Specifically, we found that reduction of K. pneumoniae biomass was not always linked to increase in P. protegens biomass in the 3-species biofilm. It is also interesting that QS did not alter the relationship between P. aeruginosa and K. pneumoniae despite the clear reduction in the latter when the QS mutants of P. aeruginosa are present in the three species community. Further work will be needed to elucidate how QS effectors operate in the three species community to control composition. Similarly, the mechanisms of competition mediated by secondary metabolites of P. protegens and their role in biofilms require further investigation and are beyond the scope of the present study. The least abundant bacterium in this 3-species biofilm community, P. aeruginosa , determined the proportions of the other two species. This finding shows an effect of QS on non-clonal cells even when present in low abundance in a polymicrobial community. It has been similarly reported that Microbacterium oxydans , the least abundant species, determined the spatial organization of others in a four-species community ( Liu et al., 2017 ). Additionally, P. gingivalis , which is present in relatively low abundance, is responsible for triggering changes in both amount and composition of oral commensal microbiota ( Hajishengallis et al., 2011 ). These bacteria are examples of keystone species in a community ( Hajishengallis and Lamont, 2016 ). Low abundant species in communities are often overlooked due to the limitation of omics approaches in delineating their role in these biofilms. Despite some of those limitations, it is none the less possible to identify potential keystone species in complex communities, such as wastewater systems, through network analysis. Indeed, such network analyses have been extended to functional genes within sludge communities to identify those genes, termed keystone genes, that play essential, linking roles, between community members and presumably sludge function ( Roume et al., 2015 ). The role of low-abundance keystone pathogens that promote the formation and stabilization of dysbiotic and disease-inducing communities in the human gut have been reported ( Hajishengallis et al., 2012 ). There are also keystone organisms that act as stabilizers that kill opportunistic pathogens but not the indigenous gut bacteria ( Stecher et al., 2013 ). For example, it has been shown that P. aeruginosa senses peptidoglycan and diffusible signals produced by other bacteria to modify community composition and persistence in P. aeruginosa -dominated disease conditions ( Twomey et al., 2012 ; Korgaonkar et al., 2013 ). The role played by low abundance bacteria producing these signals in shaping community-level changes would be consistent with their position as keystone bacteria. Our findings clearly demonstrate that a bacterium does not need to be the most abundant organism in the biofilm to be important for controlling the community structure and composition of bacterial species. In this way, P. aeruginosa serves as a keystone species in this biofilm community. The results presented here also demonstrate how a genetically amenable system can be well suited to test hypotheses about the roles and mechanisms of keystone species in mixed species communities. Of the three species used in this study, P. aeruginosa and K. pneumoniae are opportunistic human pathogens known to occur together in human-associated infections such as respiratory tract, chronic wound and urinary tract infections ( Harmsen et al., 2010 ; Rezaei et al., 2011 ; Magill et al., 2014 ). Both bacteria adapt to airways and other biotic surfaces through biofilm formation which contributes to the success of these pathogens ( Riquelme et al., 2018 ). These bacteria often occur with other bacteria in polymicrobial interactions associated with human disease and some of these interactions have been studied ( Peters et al., 2012 ). P. protegens , predominantly studied as a plant-protecting bacterium, is not generally considered a human pathogen although bacteria of the P. fluorescens complex have been isolated in clinical samples from mouth, stomach and lungs ( Paulsen et al., 2005 ; Scales et al., 2014 ). Colonization studies of roots and biofilm formation by P. protegens have been reported but interaction with other members of rhizosphere communities is less studied. Most of the community studies use a top-down approach to interrogate the properties of community dynamics and function. However, these studies have yielded less information regarding specific mechanisms, interactions and community assembly. Understanding the different 2-species interactions in our mixed-species community is an advantage for future studies with increasing community complexity made up of bacterial species having either clinical or environmental relevance. In conclusion, P. aeruginosa QS is an important modulator of the composition and structure of the 3-species biofilm community. In this system, P. aeruginosa functions as a keystone species that modulates the relative abundances of the species within its local environment. Future studies need to focus on the mechanisms behind changes in the proportions of both K. pneumoniae and P. protegens in the 3-species biofilm community. Competitive interactions between two members of this community, P. aeruginosa and P. protegens , appear to be responsible for this phenotype. Therefore, further studies on P. aeruginosa QS regulated functions and P. protegens secondary metabolites might help us to understand principles of antagonistic and competitive interactions within a developing biofilm."
} | 6,156 |
39527666 | PMC11714237 | pmc | 384 | {
"abstract": "Abstract Inspired by the extensive signal processing capabilities of the human nervous system, neuromorphic artificial sensory systems have emerged as a pivotal technology in advancing brain‐like computing for applications in humanoid robotics, prosthetics, and wearable technologies. These systems mimic the functionalities of the central and peripheral nervous systems through the integration of sensory synaptic devices and neural network algorithms, enabling external stimuli to be converted into actionable electrical signals. This review delves into the intricate relationship between synaptic device technologies and neural network processing algorithms, highlighting their mutual influence on artificial intelligence capabilities. This study explores the latest advancements in artificial synaptic properties triggered by various stimuli, including optical, auditory, mechanical, and chemical inputs, and their subsequent processing through artificial neural networks for applications in image recognition and multimodal pattern recognition. The discussion extends to the emulation of biological perception via artificial synapses and concludes with future perspectives and challenges in neuromorphic system development, emphasizing the need for a deeper understanding of neural network processing to innovate and refine these complex systems.",
"conclusion": "7 Conclusion and Perspective The advancement of robotics and AI has significantly accelerated the design and development of NAS systems. These systems emulate the functionality and architecture of biological synapses and nervous systems. NAS systems with the ability to perceive and process sensory information like the neural pathways, not only detect various external stimuli but also convert these signals into transient channel currents via synaptic devices, thereby mimicking the somatosensory mechanisms found in biological sensory nerves. [ \n \n 144 \n , \n 145 \n , \n 146 \n , \n 147 \n , \n 148 \n \n ] The integration of these synaptic signals with neural networks, governed by supervised or unsupervised machine learning algorithms, facilitates the direct processing of signals emanating from the diverse sensory inputs such as light, [ \n \n 149 \n , \n 150 \n , \n 151 \n \n ] touch, [ \n \n 152 \n , \n 153 \n \n ] smell, [ \n \n 28 \n , \n 154 \n \n ] taste, and sound, enabling AI and robotics to perform complex recognition and decision‐making tasks. As highlighted in this review, significant advancements have been achieved in the development of NAS systems with various device structures and operational principles that effectively simulate several synaptic functions. Specifically, the integration of artificial sensory nerves with biological efferent nerves also offers a novel approach for replacing damaged peripheral nerves and generating voluntary movements, providing a new standard for neuromorphic prosthetic devices. Consequently, artificial synapses can be engineered to produce synaptic responses indistinguishable from the electrophysiological signals of biological neurons. Incorporating a closed‐loop feedback mechanism, NAS systems could establish a real‐time adaptive feedback loop, autonomously adjusting neural interfacing activities to restore complex motor functions. These artificial systems with reliable synaptic outputs can serve as human‐machines interfaces, narrowing the gap between prosthetic devices and the biological nervous system. Ongoing research on nervetronics, encompassing neural‐inspired electronics such as neural prostheses, exoskeletons, and soft robotics that connect NAS systems with various living organs, are expected to accelerate due to rapid advancements in materials science, computer science, AI, healthcare, and synaptic devices. Despite the progress discussed above, challenges in the practical application of NAS systems still exist, and further research investigations are needed. Here, we suggest several areas for future research. 1. Scalability: The fabrication of NAS systems often entails the use of separate sensors and processing units, which presents challenges in enhancing device density and achieving system‐level integration. In pursuit of energy‐efficient and compact NAS systems, strategies for monolithic integration, which minimize the distance between sensors and processing units, have been reported. [ \n \n 66 \n , \n 68 \n , \n 108 \n \n ] Nevertheless, further research investigation on device‐device uniformity, stability, and other performance metrics are required to implement large‐scale integration. 2. Algorithm improvement: The processing of sensory information in NAS systems integrated with neural networks typically entails tasks such as recognition and classification, facilitated by the input of pre‐processed sensory information. [ \n \n 155 \n , \n 156 \n \n ] This functionality enables NAS systems to execute decisions or take actions based on the interpretation of sensory inputs. However, this intelligent capability is highly dependent on large amounts of sensory data, rendering the training process time‐consuming and susceptible to overfitting. To address these challenges, the technique of neural network pruning granularities can be leveraged to enhance computational efficiency and accelerate training time and help mitigate overfitting. [ \n \n 157 \n \n ] This technique involves reducing the size and complexity of neural network models by eliminating less critical components, such as neurons, weights, or layers, without significantly degrading the network performance. [ \n \n 157 \n \n ] \n 3. Power consumption: Innovative strategies, such as iontronics, have been employed to realize low power consumption in synaptic devices. Nonetheless, the incorporation of machine learning algorithms and the separation of system components lead to substantial power consumption due to the extensive amount of energy required for data processing. Given that biological neurons demonstrate exceptionally low power consumption while maintaining high efficiency, [ \n \n 4 \n , \n 5 \n \n ] future development of NAS systems should integrate technologies that allow operation with minimal energy sources. 4. Memory retention: The biological neural system possesses the capability for both STM and LTM characteristics. Information in STM that is not actively rehearsed tends to diminish gradually, whereas repeated rehearsing transforms it into LTM. STM facilitates the removal of unnecessary information over time, whereas LTM enables the retrieval of information that has been stored for extended periods. [ \n \n 6 \n , \n 158 \n , \n 159 \n \n ] The integration of STM and LTM functionalities is desirable in NAS systems, yet the memory retention abilities of existing NAS systems, particularly in terms of LTM, are inferior to their biological counterparts, with information loss occurring within a few hours. The development of advanced materials and design strategies is necessary to realize LTM retention capabilities that are on par with or exceed those of the biological neural system. Research efforts to overcome these challenges will accelerate the advancement of next‐generation neuroprosthetics, nervetronics, and neurorobotics that are energy‐efficient and suitable for implantation. We believe that continued research into the development of high‐density iontronic neural device arrays [ \n \n 160 \n \n ] could significantly diminish the physical dimensions of devices, enhance scalability, and reduce power consumption.",
"introduction": "1 Introduction The human nervous system plays a crucial role in enabling signal transmission between different body parts and the brain, facilitating the reception and transmission of sensory data for coordinated actions. [ \n \n 1 \n , \n 2 \n , \n 3 \n \n ] This system is categorized into the central nervous system (CNS, encompassing the brain and spinal cord) and the peripheral nervous system (PNS, including receptors, sensory, and motor nerves). [ \n \n 4 \n , \n 5 \n \n ] The PNS contains diverse sensory receptors (such as mechanoreceptors, photoreceptors, and nociceptors) that detect and react to environmental stimuli like light, sound, touch, and chemicals. [ \n \n 6 \n \n ] These stimuli are then conveyed to the CNS through sensory nerves for neuronal computations, learning, cognition, and memory. [ \n \n 7 \n \n ] The CNS interprets this sensory information and sends instructions to specific organs and tissues via motor nerve fibers to generate appropriate responses. [ \n \n 6 \n \n ] This complex information processing involves an interconnected network of neurons and synapses, with presynaptic neurons generating nerve impulses (action potentials) in response to stimuli. These impulses are transmitted to postsynaptic neurons through synapses. [ \n \n 8 \n \n ] The efficiency of impulse transmission between two neurons (known as synaptic weight) changes during neuronal activities. These changes in synaptic weight result in the modification of the strength of connections between the neurons (referred to as synaptic plasticity), facilitating short‐term plasticity (STP) and long‐term plasticity (LTP) depending on the retention time. [ \n \n 4 \n , \n 8 \n , \n 9 \n \n ] STP enables synapses to execute essential computational tasks including signal encoding, filtering out extraneous signals, and facilitating decision‐making within the neural networks. [ \n \n 10 \n , \n 11 \n , \n 12 \n \n ] Through adequate training, STP can transition into LTP by regulating the frequency and intensity of spikes (impulses) generated by presynaptic neurons. [ \n \n 13 \n , \n 14 \n , \n 15 \n \n ] This process underpins the foundational mechanisms of learning and memory. [ \n \n 5 \n , \n 9 \n \n ] This complex system allows humans to perceive and interpret sensory inputs ( Figure \n 1 \n , left side), through perceptual learning and adaptation. Figure 1 An illustration depicting a visual representation of both human and artificial signal processing pathways from various sensory inputs. The human nervous system consists of receptors, sensory nerves, and neurons for sensing, transmitting, and processing of sensory information (left part). Similarly, neuromorphic artificial sensory system mirroring the human nervous system, comprises sensing, transmission, and neural network algorithms for detecting and processing of sensory data (right part). Drawing inspiration from the sensory nervous system of humans (PNS and CNS), the development of neuromorphic artificial sensory (NAS) systems has garnered significant research interest due to their potential to create brain‐like computing systems that offer faster data processing and storage with low power consumption. [ \n \n 8 \n , \n 16 \n , \n 17 \n \n ] Such advancement promises significant progress in next‐generation humanoid robotics, prosthetics, and wearable technologies. [ \n \n 18 \n , \n 19 \n \n ] NAS systems are primarily operated with three functionalities; sensory recognition (visual, [ \n \n 20 \n , \n 21 \n , \n 22 \n \n ] tactile, [ \n \n 23 \n , \n 24 \n , \n 25 \n \n ] auditory, [ \n \n 26 \n , \n 27 \n \n ] olfactory, [ \n \n 28 \n , \n 29 \n , \n 30 \n \n ] and gustatory [ \n \n 31 \n , \n 32 \n \n ] ), neural signal processing, [ \n \n 33 \n , \n 34 \n , \n 35 \n \n ] and interpretation of the sensory information based on neural network processing algorithms. [ \n \n 21 \n , \n 36 \n , \n 37 \n \n ] NAS system transforms external sensory stimuli into electrical signals via artificial neurons and neural‐like devices, facilitating neural signal processing of external sensory information. Artificial neurons play a crucial role in bio‐inspired neuromorphic systems, emulating biological neural networks for bio‐sensing and bio‐interfacing applications. These neurons, particularly organic electrochemical neurons [ \n \n 38 \n \n ] and organic artificial neurons (OANs), [ \n \n 39 \n , \n 40 \n \n ] mimic neuronal excitability and firing using organic mixed ionic‐electronic conductors. By interacting with biological carriers like ions and neurotransmitters, they closely replicate natural neuronal processes. OANs, for instance, exhibit spiking behavior modulated by the concentration of Na + and K + ions, allowing real‐time operation with biological tissues. [ \n \n 39 \n , \n 40 \n , \n 41 \n \n ] In the context of bio‐sensing and bio‐interfacing, these neurons integrate sensors, oscillators, and synaptic transistors to convert physical stimuli into electrical impulses, mimicking action potentials in biological neurons. This capability is critical for applications like neuroprosthetics, biohybrid systems, and neuromorphic sensing. Their operation in biological environments makes them highly suitable for direct interaction with living tissues, paving the way for advancements in medical diagnostics, robotics, and neural interfaces, effectively linking artificial electronics with biological systems. [ \n \n 39 \n \n ] It is noteworthy that the characteristics of input spike signals–their width, duration, amplitude, and frequency–vary depending on the specific type of sensory stimuli. Exploiting these sensory pulse signals, NAS system enables neural signal processing of sensory information, thereby facilitating the generation of synaptic memory signals, encompassing short‐term and long‐term memory (STM and LTM). In recent research, NAS technologies have steadily progressed to augment artificial sensory neural network computation capabilities, providing brain‐like sensory information processing across various disciplines. [ \n \n 42 \n \n ] Nevertheless, the predominant focus of the research on NAS systems has revolved around diverse materials for the synaptic device architecture, their sensing mechanisms across the five senses, and synaptic characteristics. [ \n \n 4 \n , \n 5 \n , \n 43 \n , \n 44 \n , \n 45 \n \n ] Therefore, there exists an urgent necessity for a comprehensive exploration of neural network processing mechanisms for sensory information and its practical application, aiming to deepen our understanding of NAS systems and drive further advancements in the field. In this review, we highlight the representative NAS systems utilizing neural network algorithms to process synaptic sensory information in various applications. We begin with an analysis on the latest developments in artificial synaptic properties stimulated by optical, mechanical, sound, chemical, and hybrid inputs, focusing on elucidating their neural network processing capabilities for the sensory synaptic signals. Moreover, we delve into the cutting‐edge applications of NAS systems that replicate biological neuromorphic perceptions and neural interface behaviors, and finally, concluding with a discussion on the challenges and future directions for development of NAS technologies."
} | 3,669 |
37996910 | PMC10668361 | pmc | 385 | {
"abstract": "Background The positive effects of exposing corals to microorganisms have been reported though how the benefits are conferred are poorly understood. Here, we isolated an actinobacterial strain (SCSIO 13291) from Pocillopora damicornis with capabilities to synthesize antioxidants, vitamins, and antibacterial and antiviral compounds supported with phenotypic and/or genomic evidence. Strain SCSIO 13291 was labeled with 5 (and − 6)-carboxytetramethylrhodamine, succinimidyl ester and the labeled cell suspension directly inoculated onto the coral polyp tissues when nubbins were under thermal stress in a mesocosm experiment. We then visualized the labelled bacterial cells and analyzed the coral physiological, transcriptome and microbiome to elucidate the effect this strain conferred on the coral holobiont under thermal stress. Results Subsequent microscopic observations confirmed the presence of the bacterium attached to the coral polyps. Addition of the SCSIO 13291 strain reduced signs of bleaching in the corals subjected to heat stress. At the same time, alterations in gene expression, which were involved in reactive oxygen species and light damage mitigation, attenuated apoptosis and exocytosis in addition to metabolite utilization, were observed in the coral host and Symbiodiniaceae populations. In addition, the coral associated bacterial community altered with a more stable ecological network for samples inoculated with the bacterial strain. Conclusions Our results provide insights into the benefits of a putative actinobacterial probiotic strain that mitigate coral bleaching signs. This study suggests that the inoculation of bacteria can potentially directly benefit the coral holobiont through conferring metabolic activities or through indirect mechanisms of suppling additional nutrient sources. Supplementary Information The online version contains supplementary material available at 10.1186/s40793-023-00540-7.",
"introduction": "Introduction Coral reefs are the largest structures created by living organisms and are home to highly diverse marine organisms [ 1 ]. Reef-building corals are associated with a diverse microbiota embracing photosynthetic Symbiodiniaceae, as well as bacteria, fungi, archaea, and viruses, which together with their coral host comprise the holobiont [ 2 ]. Symbiodiniaceae fix carbon and provide photosynthates to the coral host. Other microbial symbionts also play crucial roles in maintaining coral host fitness and survival by participating in carbon, nitrogen, and sulfur cycling, providing nutrients, and defending against pathogens [ 2 ]. The coral-associated microbiome is sensitive to environmental perturbation [ 2 ]. The coral-Symbiodiniaceae symbiosis may be disrupted by thermal stress, resulting in the elimination of algal endosymbionts and coral bleaching, leading to a shortage of energy for coral hosts [ 3 ]. It is known that elevated temperature damages Symbiodiniaceae’s photosynthetic apparatus and leads to the generation of reactive oxygen species [ 3 ]. Oxidative stress due to increased temperatures can activate innate immunity in corals, and host cells undergoing apoptosis seek to remove damaged symbionts by host cell suicide [ 3 , 4 ]. Exocytosis of algal symbionts has also been observed in coral bleaching [ 3 ]. In addition to oxidative stress induced by elevated temperature, nutrient reduction due to damaged photosystems in Symbiodiniaceae may also lead to coral bleaching through the photooxidative pathway [ 5 ]. At the molecular level, heat shock proteins and antioxidant enzymes are commonly found to be upregulated in corals in response to heat stress [ 3 , 4 , 6 – 8 ]. The overexpression of p 53, a pro-apoptotic transcription factor, after exposure to thermal stress activates caspases, which are involved in apoptosis and increase under thermal stress in corals [ 7 – 10 ]. Symbiodiniaceae also exhibited significant expression alterations in the face of elevated temperatures, such as genes associated with photosynthesis, metabolism, antioxidant activity, and the immune response [ 7 , 8 , 11 – 13 ]. Globally, coral reefs are in decline largely as a result of increased scale, frequency, and intensity of coral bleaching events driven by anthropogenic global warming [ 14 – 16 ]. Given the significant functional roles of the bacterial symbionts in the coral holobiont [ 2 , 17 ], the application of beneficial microorganisms for corals (BMCs) to enhance coral tolerance and resilience to environmental stress is promising and has received increasing attention in recent years [ 18 , 19 ]. By selecting bacterial isolates with potential beneficial traits to corals, such as facilitating enhanced nitrogen (e.g. fixation and denitrification) or sulphur cycling, catalase activity, and inhibition activity to putative pathogens (e.g. Vibrio coralliilyticus ), assembled BMC consortiums applied to corals have been shown to mitigate coral bleaching and mortality when corals were challenged with higher seawater temperatures or putative coral pathogens [ 18 – 20 ]. Currently however, it is not clear if the added microbial strains successfully establish a symbiotic relationship with the corals and confer microbial mediated traits that benefit the host or indirectly buffer coral stress through provision of additional nutrients [ 18 – 21 ]. Actinobacteria are ubiquitously associated with stony corals [ 22 , 23 ], with several culturable coral-associated actinobacteria demonstrating high antimicrobial activity and the potential to synthesize a range of putative beneficial biosynthetic compounds [ 24 ]. However, the probiotic potential of actinobacterial strains has not been investigated extensively [ 19 ]. During the exploration of culturable coral-associated bacteria, we isolated the facultative anaerobic Actinobacteria strain SCSIO 13291 from Pocillopora damicornis tissues. Here, we explored the genetic and phenotypic properties of this strain. Given its capabilities, we further tested if this strain was capable of ameliorating environmental stress when inoculated onto the coral P. damicornis exposed to controlled thermal stress. Bacterial cells were labelled with fluorescent dye and inoculated onto the tissues of coral polyps with visualization, coral physiological, transcriptome and microbiome analyses conducted to elucidate the interactions and effect this strain conferred on the coral holobiont.",
"discussion": "Discussion Inoculation of the actinobacterial strain SCSIO 13291 to thermally stressed coral nubbins partially mitigated the visual signs of bleaching and maintained higher Symbiodiniaceae cell densities and photosynthetic Fv / Fm average rates relative to untreated corals. These effects were similar to previous reported studies that inoculated bacterial consortia onto heat stressed or pathogen challenged coral, thereby conferring benefits to the coral host [ 18 , 19 ]. The mechanism by which the inoculated bacteria increase the resilience of the coral host to stress or challenge is unknown though the associated bacterial communities restructuring and coral transcriptional reprogramming were observed [ 18 , 19 ]. Inoculated bacteria can potentially directly benefit the coral through conferring metabolic traits that boost holobiont performance or through indirect mechanisms of suppling additional nutrient sources that promote symbiont growth and associated coral health indicators. The SCSIO 13291 strain inoculated in this study was labeled with the vital fluorescent stain TAMRA/SE, which has no adverse effects on cell viability and has been used in monitoring bacterial transport in subsurface environments [ 33 ] and hemocytic phagocytosis of bacteria [ 34 ]. Microscopy images of coral polyps detected fluorescent signals consistent with TAMRA/SE stained bacteria near the oral disk of the polyp and attached to the column of inoculated coral samples collected at both T2 and T3 timepoints, indicating that the inoculated bacterial strain is associated with the corals throughout the experimental period. However, the Propionibacteriaceae affiliated 16S rRNA gene sequences recovered from the inoculated corals represented less than 1% of the total sequences retrieved and representative ASVs displayed only 90–94.1% similarity to the 16S rRNA gene sequence of the inoculated SCSIO 13291 strain. Previous studies that have inoculated bacteria to coral (BMC cocktails) similarly reported low recovery of 16S rRNA gene sequences affiliated to the inoculated strains [ 18 , 19 , 35 ]. Hence results based on 16S rRNA gene sequencing suggest that the inoculated strains may not establish an association with the corals nubbins, though amplicon sequencing also has limitations including biases associated with DNA extraction and amplification [ 36 ]. Elucidating if the inoculated bacteria establish an association with the coral host, localizing where these association are, how long they are maintained and subsequently characterizing if metabolic activity of the added bacteria directly benefit the coral host is essential in future studies [ 21 ]. Alternatively, the inoculation of bacteria may provide a short-term stimulatory benefit to the coral holobiont through nutrient acquisition, modulating the associated bacterial community structure plus Symbiodiniaceae and coral host metabolic activity. Inoculation of the SCSIO 13291 strain resulted in significant changes in the coral associated microbial structure, consistent with previous studies profiling the bacterial communities of BMC-treated corals [ 18 – 20 ]. Network analyses identified that the modularity of the interaction between bacterial communities associated with FASW-treated corals at 31 °C was lower (less stable [ 37 , 38 ]) than those associated with corals maintained at 27 °C and SCSIO 13291-treated corals at 31 °C, although the average degree and link were higher. These alterations might be induced by additional heterotrophic nutrients supplied by dosing the SCSIO 13291 strain to the corals or alternatively through specific active metabolites produced by the SCSIO 13291 strain. Although cooccurring networks of the coral microbiome have been elucidated across different treatment scenarios [ 39 – 41 ], the relationship between the tissue specific bacterial interactions and the resilience and tolerance of coral holobionts subjected thermal stress is complicated, though crucial for understanding the resilience of coral holobionts to ocean warming. At the ASVs level, several potential pathogens were significantly decreased in relative abundance in the SCSIO 13291 inoculated coral nubbins under thermal stress, while ASVs affiliated with Endozoicomonas and Candidatus Amoebophilus (T2 & T3) were enriched. Endozoicomonas affiliated taxa often dominant coral microbiomes, accumulated in aggregates (Cell Associated Microbial Aggregates) within the coral tissues and proposed to be a beneficial symbiont [ 42 – 46 ]. However, under environmental stress the relative abundance of Endozoicomonas taxa decrease likely through dysbiosis processes [ 7 , 47 , 48 ]. The maintenance of Endozoicomonas population in the tissues of the SCSIO 13291 inoculated corals likely benefited the coral holobiont during the heat stress treatment. The Bacteroidetes genus “Candidatus Amoebophilus ” is of interest due to characteristic endosymbiotic traits, including smaller genomes and associated reduced metabolic capabilities while also harboring a high count of host cell interaction genes [ 49 – 52 ]. Its role in the coral holobiont, especially enriched in the SCSIO 13291 inoculated corals which displayed better physiological status under temperature stress, is unclear. Additionally, several ASVs enriched in SCSIO 13291-treated samples under thermal stress, including an unknown Gammaproteobacteria, Denitrovibrio , Fulvivirga , Coxiella , and Halodesulfovibrio marinisediminis , were also determined to be keystone species in the co-occurrence networks. Enrichment of Denitrovibrio might enhance nitrate reduction to ammonium [ 53 ], and thus supply the preferred nitrogen source to Symbiodiniaceae [ 54 ], which may be linked to the downregulation of ammonium transporters in the coral host. Fulvivirga affiliated taxa, like Endozoicomonas have been found to be more abundant in healthy corals [ 55 ], while Coxiella has previously been reported as a common coral associated symbiont [ 56 ]. A novel member of the sulfate-reducing bacterium Halodesulfovibrio has been proposed to syntrophically interact with coral-associated Prosthecochloris , which is capable of carbon and nitrogen fixation in coral skeletons [ 57 ]. However, little information is known about the potential function of this genus in coral tissues in addition to many of the other affiliated taxa recovered in this study. Given their potential important roles identified within the network links of the coral microbiome, further investigations are required to establish their roles in structuring the coral-associated bacterial community and influence on thermal tolerance. Transcriptional profiling demonstrated significant differences in gene expression patterns between SCSIO 13291-treated corals maintained at 31 °C and FASW-treated corals subjected to 27 and 31 °C. Photosystem and photosynthesis-associated genes were upregulated in Symbiodiniaceae of coral nubbins inoculated with SCSIO 13291, especially at T3, and the Fv / Fm rates were higher than those of the FASW-treated corals (albeit lower than corals at 27 °C) suggesting mitigation of damage to the photosystems for Symbiodiniaceae in inoculated corals [ 19 , 58 ]. Several genes putatively involved in apoptosis and exocytosis were also downregulated in the SCSIO 13291 inoculated coral hosts, correlating with these higher Symbiodiniaceae densities. Apoptosis and exocytosis are the processes through which corals expel algal symbionts under thermal stress and ultimately manifest a bleaching phenotype [ 3 , 4 ]. Therefore, inoculation of the bacterial strain potentially resulted in less damaged Symbiodiniaceae photosystems and downregulated putative apoptosis- and exocytosis-related proteins of the coral host which together mitigated Symbiodiniaceae loss. The underlying mechanisms for these responses are unknown however, and maybe the result of release of the Symbiodiniaceae from nutrient limitation within the coral gastrodermal cell layer [ 5 , 59 ]. This is somewhat supported through analysis of the metabolite transportation expression patterns. We found the downregulation of ammonium and peptide transporters and the folliculin complex involved in sensing amino acid availability and the upregulation of the sugar transporter SWEET1-like in SCSIO 13291-treated coral hosts. The peptide transporter and folliculin complex were previously found to be overexpressed in alga-hosting cells in single–cell transcriptional analysis [ 31 ]. The SCSIO 13291-treated coral may transfer available sugars supplied by the dosed bacteria rather than peptides or amino acids derived from algal symbionts. Expression of peroxiredoxin, thioredoxin, glutaredoxin, superoxide dismutases and catalase associated genes were upregulated in SCSIO 13291-treated coral hosts, while ascorbate peroxidase and manganese superoxide dismutase were upregulated in Symbiodiniaceae. Hence actinobacterial inoculation appeared to enhance activity of these antioxidants, which could protect corals and Symbiodiniaceae from tissue damage caused by reactive oxygen species (ROS) under thermal stress. Strain SCSIO 13291 possessed catalase activity and the capability to synthesize heme, both of which potentially functioned in ROS elimination. In addition, strain SCSIO 13291 has the ability synthesize nisin and labyrinthopeptin compounds with demonstrated antibacterial and antivirus activities [ 60 ] and which may contribute to the defense against pathogens and viruses in the coral holobiont. The presence of these antioxidative and antimicrobial capabilities supports the beneficial effects of SCSIO 13291, though further investigations are required to determine if these bioactive compounds are their associated specific activities are conferred directly to the coral holobiont and maintained over time to mitigate stress. In the face of global climate change, the tolerance of coral holobionts to heat stress is one key factor to their long-term survival [ 61 ]. The results of this study showed that inoculation of an actinobacterial strain can mitigate coral bleaching signs and this was accompanied by alterations in the coral-associated bacterial community and their ecological networks. A suite of coral host and Symbiodiniaceae responses were observed in SCSIO 13291-treated samples, including protection from ROS and light damage, attenuated apoptosis and exocytosis, and metabolites utilization, potentially underlying coral thermal tolerance. These responses across the coral host, Symbiodiniaceae, and associated bacterial community could be result of nutrients supplied through inoculation of the bacteria or through direct metabolic activity of the bacteria associating with the coral polyp (Fig. 5 ). It is worthy to include the killed bacterium in the future experiment to acquire more evidence [ 62 ]. Irrespective of these direct or indirect mechanisms, our results provide insights into the beneficial effects of a putative actinobacteria coral probiotic and supports the feasibility of enhancing coral resistance to adverse environmental conditions through modulation of the associated microbiome. \n Fig. 5 Summary of the alterations in bacterial community and expression profile in strain SCSIO 13291-treated coral samples that showed mitigated bleaching signs. We proposed that these responses could be results of nutrients supplied through inoculation of the bacterium (indirect mechanism) or metabolic activity of the added bacterium associating with the coral polyp (direct mechanism)"
} | 4,498 |
36151070 | PMC9508249 | pmc | 386 | {
"abstract": "Get in-depth understanding of each part of visual pathway yields insights to conquer the challenges that classic computer vision is facing. Here, we first report the bioinspired striate cortex with binocular and orientation selective receptive field based on the crossbar array of self-powered memristors which is solution-processed monolithic all-perovskite system with each cross-point containing one CsFAPbI 3 solar cell directly stacking on the CsPbBr 2 I memristor. The plasticity of self-powered memristor can be modulated by optical stimuli following triplet-STDP rules. Furthermore, plasticity of 3 × 3 flexible crossbar array of self-powered memristors has been successfully modulated based on generalized BCM learning rule for optical-encoded pattern recognition. Finally, we implemented artificial striate cortex with binocularity and orientation selectivity based on two simulated 9 × 9 self-powered memristors networks. The emulation of striate cortex with binocular and orientation selectivity will facilitate the brisk edge and corner detection for machine vision in the future applications.",
"introduction": "Introduction The history of machine vision spans more than several decades 1 . Nevertheless, the robust and general solutions to the major issues such as motion detection, object recognition, vision-based navigation and activity recognition are still beyond reach of present computer vision system 2 . Biologically visual system is hierarchical organization including retina, optic nerve, lateral geniculate nucleus (LGN) and striate cortex 3 – 5 . It is evident that the different level of visual system process different types of visual information with receptive fields covering different region of visual field 6 . Get understanding of each part of visual pathway playing in visual perception yields insights of the challenges that classic computer vision is facing 7 . The receptive field is a restricted retinal area where the light shining on it could influence the firing rate of corresponding units 4 , 8 . The ganglion cells possess concentric receptive field with an “off” center and an “on” border or an “on” center and “off” border. The “on” and “off” region in receptive field are mutually antagonistic 9 . So that the light spots with circular form and restricted to the “on” area are more effective stimuli for activating retinal ganglion cells 3 , 10 – 12 . Its next part, LGN, possesses similar concentric receptive field 13 , 14 . However, the receptive field of cells in the striate cortex is narrow, long, vertically oriented region which differs strikingly from that in retinal ganglion cells and LGN 15 . Therefore, the vertical slit-shaped spot of light superimposing on the center of receptive field of striate cortical cells often evoke brisk response 16 – 18 . The key role of receptive field of striate cortex is supporting the edge and corner detection so the motion detection is usually processed here 19 . Along the visual pathway, the receptive field of different level of visual system shows convergence process 20 . The receptive field of one ganglion can be viewed as the collective of the many photosensory cells which synapses with it. In turn, the group of ganglion cells form the receptive field of striate cortical cells (Fig. 1a ). Therefore, the receptive field of one striate cortical neuron can be viewed as the collective of the many retinal photosensory cells (rods and cones) which indirectly synapse with it 21 . The modulation of synaptic connections between the photosensory cells and striate cortical cells is crucial to develop the narrow, slit-shaped receptive field, ensuring the orientation selective in striate cortex 22 , 23 . Fig. 1 Artificial striate cortex based on self-powered memristor. a The hierarchy of human visual system for light perception and processing. b The convergence process of receptive field in different level of visual system. c The concept of artificial striate cortex based on self-powered perovskite memristor. d The development of narrow, slit-shaped receptive field of striate cortex with binocular orientation selectivity based on BCM learning rule. Compared with Purely Hebbian, that is the modification of synapse based on the multiplication of the pre- and post-stimuli and the synaptic weight stabilized by controlling cortical responses below the maximum, the rate-based Bienenstock-Cooper-Munro (BCM) learning rule is more biorealistic for the synaptic modification and neuronal response selectivity in the experience-dependent modification that observed in striate cortex 24 . It describes that the sign of synapse weight modification is determined by whether the postsynaptic response exceeds a threshold. The postsynaptic firing rate higher than a sliding threshold induces the strengthening of synapse while the postsynaptic responses below the threshold weaken the synapse 25 , 26 . The sliding threshold is dependent on the average activity of the postsynaptic neuron, ensuring a history- or experience-dependent characteristic, which is the figure of merit of BCM 27 – 29 . In order to realize BCM, a triplet-STDP which introduces a third presynaptic or postsynaptic spike to pair-STDP has been employed to reproduce frequency effect of the pair protocol 30 , 31 . The frequency effect stems from the pair spikes-induced paired term and the previous spike of triplet-STDP induced triplet term 32 . Furthermore, the triplet-STDP can be employed to realize rate-based BCM learning rule through an All-to-All framework 33 . Except for a strong preference for a particular orientation, the visual response of striate cortical neuron is also binocular 34 . In the biological visual system, signals from the left and the right eyes first converge in the striate cortex, V1. Neurons in adult striate cortex are binocular with a strong preference for contours of a particular orientation. In biology visual system, the newly born interocular neurons have different orientation preference which means the binocular response is inconsistent for the same field of view (FoV). A matching process between two eyes is required to form normal binocular perception for depth and stereopsis 35 . Binocular neurons in the striate cortex must match their orientation tuning through the two eyes in order for the animal to perceive coherently. Currently, the vision sensors are emerging to mimic receptive field of ganglion cell in retina for simultaneously sensing and processing. The emulation of experience-dependent modifications of synaptic strength to form binocular, orientation selective receptive field that observed in the striate cortex lags considerably behind that of the retina 36 – 38 . In this work, we first report the bioinspired striate cortex with binocularity and orientation selectivity based on the crossbar array of self-powered memristors where each cross-point contains one CsFAPbI 3 perovskite solar cell and one CsPbBr 2 I perovskite memristor. The second-order CsPbBr 2 I memristor with mobile halogenic vacancy similar to Ca 2+ dynamics in the striate cortical synapse, which allows the emulation of rate-based plasticity. While the CsFAPbI 3 perovskite solar cell can be viewed as photosensory retinal cells to synapse with striate cortical cell for converting external optical signals into electrical signals (For biological visual system, the cortical cell synapses with LGN, LGN synapses with ganglion cells, and ganglion cell synapses with retinal sensory cells. In our hardware implementation, we directly synapse artificial retinal photosensory cells with cortical cell for simplicity). Furthermore, plasticity of 3 × 3 crossbar array of self-powered memristor has been successfully modulated based on rate-based BCM learning rules for pattern recognition by light illumination. Finally, we realized artificial striate cortex with binocularity and orientation selectivity based on two 9 × 9 self-powered memristor networks, following the generalized BCM learning rule. By varying the type of input for (1) normal binocular contour vision, (2) monocular deprivation, (3) binocular deprivation, we highly reproduced the experience-dependent modifications that have been observed experimentally in kitten striate cortex. To the best of our knowledge, this is the first time to realize the hardware implemented striate cortex with binocular and orientation selectivity. The bio-inspired striate cortex is highly compatible with high-density and low power consumption machine vision owing to its crossbar paradigm and homotypic materials system.",
"discussion": "Discussion In summary, based the crossbar array of self-powered memristors, we first emulated artificial striate cortex with binocularity, orientation selectivity based on the BCM learning rule. The crossbar array of self-powered memristors is monolithic all perovskite system where each-cross point contains one CsFAPbI 3 perovskite solar cell (photosensory retinal cell) to convert external optical signals into electrical signals and one CsPbBr 2 I perovskite memristor (cortical synapse) to implement plasticity modulating. Based on this artificial striate cortex, we investigated the triplet-STDP rules under optical stimuli. The asymmetry characteristic of triplet-STDP is beneficial to the following realization of BCM-rate learning rule compared to pair-STDP. By constructing the 3 × 3 crossbar array of self-powered memristor, the critical characteristics of BCM, synapse depression/potentiation takes place at low/high postsynaptic firing rate region, and the history-dependent sliding threshold were realized which has been further applied in the optical-encoded pattern recognition. Finally, artificial striate cortex with binocularity and orientation selectivity based on two simulated 9 × 9 self-powered memristor networks, following the generalized BCM learning rule. By varying the type of input for (1) normal binocular contour vision, (2) monocular deprivation, (3) binocular deprivation, we highly reproduced the experience-dependent modifications that have been observed experimentally in kitten striate cortex. Two-terminal structure of self-powered memristor based on monolithic all-perovskite system ensures the bio-inspired striate cortex to be extendable to crossbar array structure for high-density and low power consumption machine vision, which has not been realized yet."
} | 2,593 |
31796086 | PMC6889567 | pmc | 387 | {
"abstract": "Nitrogen is an essential element of life, and nitrogen availability often limits crop yields. Since the Green Revolution, massive amounts of synthetic nitrogen fertilizers have been produced from atmospheric nitrogen and natural gas, threatening the sustainability of global food production and degrading the environment. There is a need for alternative means of bringing nitrogen to crops, and taking greater advantage of biological nitrogen fixation seems a logical option. Legumes are used in most cropping systems around the world because of the nitrogen-fixing symbiosis with rhizobia. However, the world's three major cereal crops—rice, wheat, and maize—do not associate with rhizobia. In this review, we will survey how genetic approaches in rhizobia and their legume hosts allowed tremendous progress in understanding the molecular mechanisms controlling root nodule symbioses, and how this knowledge paves the way for engineering such associations in non-legume crops. We will also discuss challenges in bringing these systems into the field and how they can be surmounted by interdisciplinary collaborations between synthetic biologists, microbiologists, plant biologists, breeders, agronomists, and policymakers.",
"conclusion": "Conclusions and perspectives The Food and Agriculture Organization of the United Nations estimates that the Earth will have two billion more people to feed in 2050 [ 217 ]. Given that half of the world population is currently sustained through synthetic fertilizers, it would not be reasonable to claim that biological nitrogen fixation will replace the Haber-Bosch process entirely. But, as indicated earlier, the extreme dependence of the global food supply on synthetic fertilizers is not sustainable. Now is the time for a “Symbiotic Revolution” to combine food production and sustainable soil health. So, are we there yet? Many avenues to improve biological nitrogen fixation in non-leguminous crops have been described in this review (Fig. 2 ). Some of them could bring solutions in the next decade, and some will probably bear fruit in the longer term [ 83 , 128 , 208 , 211 , 212 ]. Some of the natural systems we presented can provide significant amounts of biologically fixed nitrogen. Microbial and plant natural diversity is a resource and a source of knowledge that should be explored more, and that could deliver practical solutions in a relatively short time. We take as an example the Sierra Mixe maize, where an unexplored system is capable of sustaining most of the nitrogen requirement for the crop over several months and at a critical period of the growing season [ 83 ]. Such unexpected discoveries reinforce the need to preserve natural diversity in our crops and their wild relatives. These Sierra Mixe landraces cannot be used directly in most cropping systems or environments due to their size and their long growing season. Breeding the trait in more conventional accessions of maize is necessary, but this process will take time. Once this trait is introduced into more conventional varieties, agronomic questions such as the amount of fertilizer saved by the trait, the effect of soil nitrogen, or the yield cost of the trait will need to be addressed. This will require efforts and funding, but it seems achievable to use such natural traits in the next decades. Approaches aiming at engineering root nodules in cereals are more complex and will likely take more time. Procedures to engineer crops able to fix their nitrogen without the bacteria seem even longer term. Nevertheless, as discussed previously, these long-term approaches are promising and likely to succeed.\n Fig. 2 Main approaches to engineer or improve biological nitrogen fixation in cereals. 1 Improving nitrogen-fixing bacteria: ( a ) [ 153 ], ( b ) [ 111 , 113 ], ( c ) [ 154 ], ( d ) [ 152 ], ( e ) [ 110 , 144 , 171 ]. 2 Making crops better hosts for nitrogen-fixing bacteria: ( a ) [ 127 ], ( b ) [ 83 ], ( c ) [ 56 ]. 3 Allowing crops to fix their nitrogen without microbes [ 209 ] A critical justification for pursuing the range of approaches described in this review is the significant effect that the environment has on many of these biological systems. Taking the example of the Sierra Mixe maize again, mucilage production by the maize aerial roots is dependent on rain [ 83 ]. While this trait seems to be directly usable in many regions of the world, it will be more challenging to adapt it to arid environments. If we look at the worldwide distribution of nodulating legumes, engineered root nodules may be efficient in a broader range of situations, but legume nodulation itself is affected by environmental factors such as soil nitrogen or flooding [ 218 – 220 ]. The environmental dependence of plants fixing their own nitrogen is, at this point, entirely speculative. Lastly, as indicated earlier, the process of nitrogen fixation, whether biological or industrial, requires significant amounts of energy. In all the approaches to improve nitrogen fixation discussed in this review, the energy for nitrogen fixation comes from plant photosynthesis and will have a cost on plant carbon. Despite the tight autoregulation of nodulation, legumes dedicate 10–20% of their carbon to nodules [ 117 ]. This does not necessarily decrease yield, as carbon cost is offset by increased photosynthetic capacity due to the nitrogen gained from biological nitrogen fixation. However, symbiotic nitrogen fixation will not be energetically competitive compared to nitrogen fertilization, and growers in developed countries are not ready to suffer any yield loss while fertilization remains cheap. Estimating the potential yield penalty for different strategies will be necessary. Environmental policies may provide more incentives for a reduction of synthetic fertilizers in the future [ 221 ]. In developing countries, any nitrogen input will be valuable for poor farmers where nitrogen is the most important factor limiting their production [ 222 ]. International projects such as Realizing Increased Photosynthetic Efficiency (RIPE) are currently working on improving photosynthetic efficiency, which could offset the yield penalty of relying on biological nitrogen fixation [ 223 , 224 ]. Improving nitrogen fixation in non-leguminous crops has been a dream of the agronomic community for more than a century. The global challenges that our world is facing make the realization of this dream urgent. Fortunately, natural diversity holds solutions that the scientific community overlooked possibly because of the intense focus on legume nodules. Technological developments such as the advent of next-generation sequencing, gene editing, and synthetic biology allow the dissection and manipulation of plants and microbes at an unprecedented scale. We are confident that combining the prospecting of plant and bacterial natural diversity with genetic engineering will deliver solutions in the short and long terms and will help to feed the world in a more sustainable manner.",
"introduction": "Introduction Nitrogen is an essential component of life, required for building proteins and DNA, and despite being abundant in the atmosphere, only limited reserves of soil inorganic nitrogen are accessible to plants, primarily in the form of nitrate and ammonium. Thus, agricultural yields are often limited by nitrogen availability [ 1 ]. This limitation was battled for centuries by crop rotation or co-culture with legumes and the use of fertilizers in the form of animal waste, wood ash, or seaweed. At the beginning of the 20 th century, two German chemists, Fritz Haber and Carl Bosch, invented a process allowing nitrogen fixation, the conversion of dinitrogen into ammonium, on an industrial scale [ 1 , 2 ]. The use of synthetic fertilizers was the main factor for drastically increase crop production during the Green Revolution, especially in developing countries, and the subsistence of nearly half of the world population is currently dependent on the use of such fertilizers [ 3 ]. Breaking the triple bonds of dinitrogen requires vast amounts of energy (1–2% of the global energy supply) and about one ton of natural gas is dedicated to the production of one ton of synthetic nitrogen fertilizers [ 4 , 5 ]. Not surprisingly, the cost of fertilizers is highly dependent on the price of natural gas, which is currently low due to the practice of hydraulic fracturing or fracking [ 6 ]. However, the dependence of so much food production on natural gas, a finite resource, is concerning. Ironically, even biofuel production (e.g., corn ethanol) depends on the use of synthetic fertilizers and therefore fossil fuel, which defeats the very purpose of biofuels. All these examples reveal that nitrogen availability for crops is a threat to the sustainability of our agricultural systems, economy, and food supply. Besides these global sustainability considerations, the intensive use of fertilizers also creates specific issues in developed and developing countries. Addition of Haber-Bosch derived nitrogen, sometimes more than 200 kg N ha −1 yr −1 , has increased yields but also led to the contamination of groundwater and eutrophication of rivers, causing massive community shifts for inland and coastal aquatic microbiota and impacting human health [ 7 – 9 ]. In contrast, subsistence farmers are unable to access fertilizers at an affordable price. Lack of local production and poor transportation infrastructure also contributes to low yields and, thus, cycles of food insecurity and poverty [ 10 ]. Bacteria and Archaea have been fixing atmospheric nitrogen for hundreds of millions of years [ 11 ]. This biological nitrogen fixation accounts for much of the nitrogen input of natural systems, considerably more so than rock weathering or lightning [ 12 ]. Biological fixation in prokaryotes is performed by the nitrogenase complex, a metalloenzyme complex composed of the catalytic protein dinitrogenase, and an ATP-dependent electron-donating iron protein, the dinitrogenase reductase. The catalytic domain of dinitrogenases commonly contains a molybdenum-iron cofactor, but some species use two other classes of dinitrogenases, defined by the presence of vanadium-iron or iron-only cofactors [ 13 ]. The nitrogen fixation genes (commonly referred to as nif genes) encode the components of nitrogenase and other regulatory proteins. The nifHDK operon encodes the dinitrogenase and the dinitrogenase reductase, but additional proteins are required to produce a fully functional holoenzyme. About 20 nif genes have been found in nature across the three classes of nitrogenases [ 13 – 15 ]. Nitrogen-fixing prokaryotes, also called diazotrophs, can be free-living or exist in symbiotic associations with Eukaryotes, with examples including fungi ( Geosiphon ), sponges ( Dysidea ), termites, and plants [ 16 ]. A successful symbiosis requires an appropriate host and diazotrophic partner, combined with environmental conditions to allow nitrogen fixation. Diazotrophic bacteria are highly diverse and are found in various ecological niches (free-living or in association with different organisms; Fig. 1 ) and have a wide range of metabolic characteristics [ 27 – 29 ]. In plant–bacteria interactions, the energy-intensive nitrogen fixation is powered by photosynthates from the plant, in exchange for a portion of the fixed nitrogen. Most of the time, \"symbiotic nitrogen fixation\" has referred only to symbioses leading to the development of root nodules. By definition, however, symbiosis is a long-term association between two different organisms that is beneficial for at least one of them [ 30 ]. Associative nitrogen fixation obviously meets this definition, as the plant benefits from growth promotion (both via increased nitrogen nutrition and several other benefits) and the bacteria gains carbon from plant photosynthesis. Thus, in this review, we will refer to both root nodule symbioses and associative nitrogen fixation as \"symbiotic nitrogen fixation\" (Fig. 1 ). Some publications already employ these terms similarly [ 31 ], but we believe that the community should also adopt this terminology more widely.\n Fig. 1 Different types of nitrogen-fixing associations with plants. The three challenges of biological nitrogen fixation are solved with different efficiency by these types of interactions—energy source, oxygen protection, and transfer of fixed nitrogen to the plant. The efficiency of each bacterial partner is indicated by + (low), ++ (moderate), or +++ (high). The nitrogen fixation rates depend on the efficiency of the interaction. a Root nodule symbiosis, 50–465 kg N ha −1 y −1 [ 17 , 18 ]; b associative nitrogen fixation, 2–170 kg N ha −1 y −1 [ 19 – 23 ]; and c, d free-living nitrogen fixation, 1–80 kg N ha −1 y −1 [ 24 – 26 ] The root nodule symbiosis and the unexploited diversity of nitrogen-fixing microorganisms in nature Root nodule symbioses are only found in plants of a monophyletic clade often referred to as “FaFaCuRo” ( Fabales , Fagales , Curcubitales , and Rosales ) but are incredibly diverse in modes of infection by rhizobia or Frankia , nodule anatomy, and metabolism [ 32 – 35 ]. Associations between legumes and rhizobia are so efficient that legumes are found in a wide range of environments across the globe and used in nearly all cropping systems [ 36 ]. Genetic approaches have been essential to the dissection of the molecular mechanisms that control the establishment of these associations [ 37 – 39 ]. Genetic tools were first developed in rhizobia, with rhizobial mutants unable to trigger the development of root nodules, allowing the identification of nod , nol , and noe genes [ 40 – 42 ]. Some nod genes encode regulatory NodD proteins that bind to diffusible signals present in legume root exudates (flavonoids, isoflavonoids, and betaines) and regulate the expression of other nod genes that control the production of Nod factors [ 15 ]. Nod factors are lipo-chitooligosaccharides (LCOs) with a short chitin backbone of three to five residues of N-acetylglucosamine, with an acyl chain at the reducing end [ 43 ]. Nod factors are decorated with various substitutions (methyl, acetyl, fucose, arabinose, and others) that are the primary determinant of the often high levels of host specificity observed in the rhizobia–legume symbiosis [ 44 ]. Symbiotic interaction between the actinobacteria Frankia and actinorhizal plants may use different recognition factors, in which as yet unknown diffusible signals drive the pre-infection responses, instead of chitin-based signals used by rhizobia [ 45 ]. Bacterial exopolysaccharides are also often required and recognized by specific receptors for successful colonization [ 39 , 42 ]. The “common nod genes” are found in most rhizobia and control the production of the lipo-chitooligosaccharide backbone. In contrast, \"specificity nod genes\" are present in some but not all rhizobial strains and control the addition of substitutions on this chitin backbone and therefore host specificity [ 46 ]. For instance, a Sinorhizobium meliloti nodH mutant is no longer able to nodulate its natural host alfalfa but nodulates vetch [ 47 ]. A few bradyrhizobia have been shown to nodulate some legumes in the absence of nod genes, but the vast majority of rhizobia require nod genes and Nod factors to associate with their legume hosts [ 48 – 50 ]. The genetic mechanisms that control root nodulation have been deeply dissected in two model legumes, Medicago truncatula and Lotus japonicus [ 51 – 55 ]. The host mechanisms include three distinct processes that can be uncoupled genetically: mutual recognition, colonization (often called infection), and nodule development (organogenesis) [ 56 , 57 ]. Mutual recognition begins with the perception of Nod factors by lysin motif receptor-like kinases [ 58 , 59 ]. A mechano-stimulation from the microbes may modulate symbiotic signaling [ 60 ]. The activation of these receptors leads to the activation of the “common symbiosis pathway” (CSP), a signaling pathway controlling intracellular colonization and presumably adapted from the more ancient symbiosis between land plants and arbuscular mycorrhizal fungi [ 61 – 63 ]. In legumes, the CSP induces expression of transcription factors including NODULE INCEPTION (NIN) and members of the NF-Y family, which control nodule organogenesis in concert with the cytokinin signaling pathway [ 53 , 56 , 64 , 65 ]. Strikingly, NIN-like and NF-Y proteins are also involved in lateral root organogenesis in many plants, although cytokinin often acts to repress lateral root initiation [ 66 – 68 ]. Root nodules and lateral roots are both lateral root organs. Their similarities and differences have been the subject of debates over decades. The nodules of some legume species do not have a persistent meristem (determinate), but the nodules of many legumes have a persistent meristem (indeterminate) like lateral roots [ 69 ]. Classically legume root nodules have been differentiated from lateral roots by the presence of peripheral vasculature, whereas lateral roots have central vasculature. However, some actinorhizal plants have nodules with a central vasculature [ 34 ]. Genetic evidence supports the idea that the mechanisms used by plants for developing nodules have been co-opted and slightly modified from those used to form lateral roots in most plants [ 70 ]. For example, legume mutants in the transcription factor NOOT form nodules with a meristem that reverts to a lateral root identity [ 71 ]. Altogether, genetic and evolutionary studies indicate that root nodulation evolved from recruiting pre-existing mechanisms of arbuscular mycorrhizal associations and lateral root development, connecting NIN and possibly other proteins into the CSP, and bringing auxin and cytokinin together to drive nodule development [ 72 – 75 ]. The root nodules of the legume plants provide an excellent environment for nitrogen fixation, with rates of 50–465 kg N ha −1 yr −1 in agricultural settings, and has been a significant focus of the agronomic community over the last decades [ 17 , 18 , 76 ]. Intracellular nitrogen-fixing symbioses outside the legume and actinorhizal lineages are rare. Gunnera species host the cyanobacteria Nostoc in stem glands, and this symbiosis can also provide substantial amounts of fixed nitrogen (15 kg N ha −1 yr −1 ) [ 77 ]. Associations between plants and epiphytic or free-living diazotrophs can also provide significant amounts of nitrogen to the host plant. These associative symbioses are quite diverse. For instance, the interactions between lichens and mosses and cyanobacteria can contribute up to 3 kg N ha −1 yr −1 to subarctic and boreal forest communities [ 19 , 78 ]. Rice paddies are naturally fertilized by \"green manure\", comprised of aquatic ferns ( Azolla ) extracellularly associated with Anabaena azollae [ 1 , 79 ]. Many other plants accommodate Nostoc extracellularly, including cycads on modified (collaroid) roots and in slime-filled cavities in many bryophytes [ 80 ]. The nitrogen amounts fixed by these symbioses are poorly evaluated [ 19 ]. Endophytic diazotrophs, such as Gluconacetobacter diazotrophicus , Herbaspirillum seropedicae , Herbaspirillum rubrisubalbicans , and Burkholderia silvatlantica , can fix nitrogen in the vasculature and intercellular spaces of sugarcane stems [ 81 , 82 ]. Diazotrophs, including Herbaspirillum species, living in mucilage released from the aerial roots of maize landraces from Sierra Mixe, Mexico, can provide up to 82% of the host nitrogen [ 83 ]. Plants also benefit from nitrogen fixed by bacteria in the soil, which obtain their energy either from degrading organic matter in the ground (heterotrophs) or from photosynthesis (autotrophs), but the contribution of this fixed nitrogen to crops is lower than from symbioses [ 24 , 84 ]. In 2016, Ladha et al. estimated that biological nitrogen fixation in the rhizosphere of rice, wheat, and maize contributed up to 25% (13–22 kg N ha −1 yr −1 ) of the total nitrogen in harvested grain, but it was not possible to quantify the respective contributions of associative and free-living fixation [ 20 ]. Symbiotic nitrogen fixation contributes to the growth-promotion effect seen in plant growth-promoting rhizobacteria, although it is generally not the only benefit that these bacterial symbionts provide to the plant host. Rhizobacteria can increase plant access to other nutrients, enhance defense against pathogens or abiotic stresses, and secrete plant hormones [ 85 – 88 ]. It is often challenging to differentiate the contribution of biological nitrogen fixation from plant growth promotion by other factors [ 89 – 91 ]. The techniques used to evaluate how much nitrogen is fixed and transferred to the plants have strengths and pitfalls (described in Table 1 ). These issues have led to many conflicting reports and confusion in the literature. We believe that proper estimation of nitrogen fixation can only come from using several if not all of the techniques mentioned in Table 1 [ 83 ].\n Table 1 Estimating the contribution of biological nitrogen fixation Determining the rate of nitrogen fixation is a difficult task, especially in field conditions. Five categories of techniques have been used, and all of them have their pitfalls. (1) The acetylene reduction assay (ARA) is a sensitive and accurate method of assessing nitrogenase activity, via the indirect measure of reduction from acetylene to ethylene by nitrogenase. However, different types of nitrogenases reduce acetylene differently, leading to discrepancies with other methods, and this method is challenging in field conditions due to the flammable acetylene gas and difficulties in tightly enclosing the plant. Most importantly, this technique cannot evaluate how much of the fixed nitrogen is assimilated by the plant. (2) The 15 N natural abundance technique relies on the higher abundance of this naturally occurring and stable nitrogen isotope in most soils [ 92 ]. A diazotroph acquiring its nitrogen from the air and its host will, therefore, have a lower 15 N abundance than plants only obtaining their nitrogen from the soil. Variations in isotope ratios are reported as ∂-values, commonly expressed in parts per mil (‰). These variations are measured using isotope-ratio mass spectrometry. This technique is high throughput and can be performed in fields. The stable nature of 15 N isotopes allows storing and shipping samples efficiently. Unfortunately, variations in 15 N abundance across the experimental field or from a geographical location to another and soil horizons can lead to artifacts, and the use of abundant controls, including soil samples, is required. (3) 15 N isotope dilution is a variant of the previous technique where the soil is enriched with a 15 N-enriched nitrogen source to increase the differential between the ground and the air and limits the natural variations in 15 N abundance. However, the cost of 15 N-enriched nitrogen restricts the scale of these experiments. 15 N-enriched sources can also move vertically or horizontally during the growing season, which mandates frequent soil sampling for controls [ 92 , 93 ]. (4) Another 15 N-based technique, called 15 N gas enrichment, is conceptually the reverse of the previous ones. In this case, dinitrogen from the air is labeled with 15 N and the incorporation of 15 N in bacteria and its host plant indicates that they acquired some of their nitrogen from the air. This technique is one of the best pieces of evidence to prove that plants obtained nitrogen through nitrogen fixation. However, bacterial contaminations must always be considered as another source of N reaching the host. Sensitivity can be enhanced using radioactive nitrogen isotopes, such as 13 N, but these are challenging to use given their short half-life [ 94 ]. Determining if 15 N was incorporated in the host tissues is best achieved by mass spectrometry imaging or by extracting plant-specific metabolites such as chlorophyll [ 95 , 96 ]. (5) Nitrogen-balance experiments evaluate the amount of nitrogen acquired by the plant from the soil and the total amount of nitrogen in the plant. The difference between the two measurements gives the amount of nitrogen from the air. However, evaluating soil nitrogen is difficult, introducing a significant level of uncertainty in these evaluations. Biological challenges for efficient nitrogen-fixing symbioses Extending efficient symbiotic nitrogen fixation from legumes to cereals has been a dream of agronomists since the understanding of the benefits behind the legume crop rotation system [ 97 ]. As early as 1917, scientists attempted to cultivate the rhizobia from legumes and inoculate these into other species [ 98 ]. To date, however, none of these attempts to transfer the complex root nodule to non-legume plants has succeeded. Symbiotic nitrogen fixation can take many forms in nature, but the main challenges solved by these different biological systems are quite similar: energy source, oxygen protection, and efficiency of nutrient exchange. The same problems also face any new approach that aims at improving or creating a nitrogen-fixing symbiosis.\n Nitrogen fixation is energy expensive, with the reduction of dinitrogen into ammonia requiring at least 16 ATP per dinitrogen fixed (Table 2 ). However, the real cost is estimated to be 20–30 ATP, accounting for the production of the nitrogenase complex, the reductive power, and recycling the toxic dihydrogen waste resulting from the process [ 99 , 101 ]. The catalytic [4Fe-4S] cluster of dinitrogenase, which is exposed between the subunits, is permanently oxidized in minutes, while the dinitrogenase reductase—the ATP-dependent iron protein—is inactivated in seconds [ 102 – 104 ]. Thus, the entire complex is highly vulnerable to destruction by molecular oxygen. This oxygen sensitivity leads to the oxygen paradox of biological nitrogen fixation, as the most efficient source to produce ATP is aerobic respiration, which requires the presence of oxygen [ 105 ]. One solution to this paradox is to avoid oxygen entirely, respiring on sulfate, hydrogen, or metal ions. These systems are not possible in conditions in which plants can grow, so active diazotrophs must tightly regulate internal oxygen tension to supply aerobic respiration while limiting harm to nitrogenase. In legume nodules, the physical structure, including the suberin in the endodermis, acts as a physical barrier to oxygen diffusion and the leghemoglobin acts as an oxygen buffer to maintain a low oxygen tension [ 106 ]. To preserve respiratory capacity and energy production, the terminal oxidase of the electron transport chain of rhizobia binds oxygen more strongly than in most microbes, even pulling oxygen out of the leghemoglobin [ 107 ]. In non-legume symbioses, the viscous mucus excreted by maize aerial roots limits oxygen diffusion while the root and microbes in it consume oxygen, leading to low internal oxygen tension [ 83 , 108 ]. Bacteria produce exopolysaccharides and biofilms on root surfaces to similar effect [ 109 , 110 ]. Autotrophic cyanobacteria must produce oxygen from photosynthesis to power fixation, protecting their nitrogenase either by separating the nitrogenase physically in dedicated heterocyst cells or temporally by fixing nitrogen only at night. Soil diazotrophs like Azotobacter contain an additional respiratory chain dedicated to consuming oxygen to maintain an anoxic cytoplasm [ 111 ]. This is complemented by conformational protection, where iron-sulfur Shethna proteins form part of the nitrogenase complex and cover the active site in the presence of oxygen, temporally inactivating the enzyme but preventing permanent oxidative damage [ 111 – 113 ]. The efficiency of nutrient exchange between the two partners is also critical. Fixed carbon must be fed to the symbiont for energy and nitrogen exported to the host while limiting losses to other organisms or the environment. In root nodules the bacteria fix nitrogen within plant cells (endosymbiosis), which provides a large surface of contact to exchange nutrients between host and symbionts with striking structural and molecular similarities to mycorrhizal arbuscules [ 114 – 116 ]. In aerial roots of Sierra Mixe maize, nitrogen released by the bacteria in the gel is actively taken up by aerial roots. \n Table 2 Idealized nitrogen fixation equation N 2 + 8 H + + 8 e - + 16 ATP ➔ 2 NH 3 + H 2 + 16 ADP + 16 Pi [ 99 , 100 ] Energy must be expended to support bacterial growth even in the most efficient legume systems, increasing the cost of nitrogen fixation for the plant. Estimating this cost is complicated, given that additional nitrogen leads to more photosynthesis, but a loss of 5.6–8.0 g of carbon per gram of reduced nitrogen obtained by legumes appears a reasonable estimate. This represents around 30–40% efficiency relative to the theoretical cost of 2.5 g of carbon per gram of reduced nitrogen [ 117 ]. One solution to this inefficiency loss would be to express the nitrogenase complex directly in the plant. This would also prevent losses during the nutrient exchange but is a much more complex technical challenge. Manipulating the bacterial partner to increase biological nitrogen fixation in non-leguminous plants The search for microbes to improve both monocot crop nitrogen nutrition and development is a long-standing aspiration [ 118 – 120 ]. After the ‘70s, with the efforts of Dr. Johanna Döbereiner, the association between diazotrophs and cereal crops received more attention. Azotobacter and Beijerinckia were first isolated from sugarcane and cereal grasses in 1961 [ 121 ]. Enterobacter cloacae was found in corn roots in 1972, and in rice, wheat, and tropical grasses in 1973 [ 122 ]. Spirillum sp. strains were first isolated in 1975 from surface sterilized maize roots, and their nitrogenase activity demonstrated [ 123 ]. In the ‘80s, the endophyte Herbaspirillum seropedicae was isolated from maize, sorghum, and rice, and Gluconacetobacter diazotrophicus from sugarcane [ 124 , 125 ]. After the advent of the acetylene reduction assay (ARA), it was possible to test bacteria for nitrogen-fixation ability directly [ 126 ]. Diazotrophs isolated from sugarcane and cereals, including but not limited to G. diazotrophicus , Herbaspirillum frisingense , H. seropedicae , and Azospirillum brasilense , were shown to contribute at various levels to the plant nitrogen requirements via nitrogen fixation under laboratory and field conditions [ 21 , 22 , 127 – 131 ]. Next-generation sequencing made possible the identification of free-living, endophytic, and epiphytic diazotrophs on a massive scale, using genes encoding core proteins of the nitrogenase complex as markers for screening metagenomes [ 132 – 134 ]. However, the presence of these genes merely reflects the potential of the microbiota for nitrogen fixation [ 135 – 137 ]. We believe that these DNA-based surveys should be more systematically complemented with transcriptomic and possibly proteomic approaches to determine if these nif genes are actually expressed. Global methods are also not sufficient to evaluate the benefits provided to the host, which requires isolation. The right nitrogen-fixing symbiont for crops must both be an efficient colonizer of the root system and release a significant portion of its fixed nitrogen to the plant. Ideally, it would keep fixing nitrogen even in fertilized fields. Attempts to isolate better ammonium releasers have used ethylenediamine to deregulate glutamine synthase. One example, Azospirillum brasilense HM053, allowed the model C4 monocot Setaria viridis to grow in nitrogen-free media and promoted wheat growth in laboratory conditions [ 129 , 131 , 138 ] and maize growth in field conditions [ 139 ]. This effect appears common, as ethylenediamine-treated Pseudomonas sp. also increased the biomass of plants grown under nitrogen-limiting conditions [ 140 ]. Significant progress has been made in understanding the biochemical, physiological, and ecological aspects of diazotroph associations with cereals. Many diazotrophs also promote plant growth through other mechanisms, such as the production of plant hormones, phosphate solubilization, and the acquisition of other nutrients like calcium, potassium, iron, copper, magnesium, and zinc (reviewed in [ 141 , 142 ]). These mechanisms can further increase plant nitrogen access by increasing root growth and relieving nutrient deficiencies. However, the genetic mechanisms that drive the establishment of cereal–microbe interaction are still poorly understood, and this must be corrected if we are to exploit these associations more effectively. Genetic tools have been developed to study the endophytic diazotroph Azoarcus sp. BH72, and allowed the characterization of the molecular mechanisms controlling its interaction with plants [ 143 ]. Interestingly, Azoarcus sp. BH72 induced to fix nitrogen cannot be returned to culture, suggesting that it undergoes terminal differentiation in a way perhaps similar to the differentiation of rhizobia into bacteroids in the rhizobium –legume symbiosis [ 144 ]. Recently, Faoro et al. [ 145 ] isolated a new strain, Azoarcus olearius DQS-4T, in oil-polluted soils. This DQS-4T strain demonstrated significant plant growth promotion activity and an active nitrogenase [ 145 ]. This finding highlights the importance of continuing to prospect, in a wide range of environments, for better nitrogen fixers, better colonizers, and plant growth promoters. Genetic engineering strategies towards better nitrogen-fixing microsymbionts Engineering microsymbionts may make it possible to confer nitrogen-fixing ability on non-diazotrophs or to improve the benefits of natural associations between diazotrophs and crops significantly [ 146 ]. The transfer of fixation capacity to a non-diazotroph was first achieved in 1971, with the transfer of a nif cluster from Klebsiella pneumonia e into Escherichia coli [ 147 ]. Subsequently, many researchers have produced transgenic bacteria capable of fixing nitrogen, discovering the minimum set of nif genes required for the production of a functional nitrogenase [ 148 – 150 ]. An exciting goal of engineering increased nitrogen fixation is to remove the inhibition of nitrogenase by nitrogen and oxygen and to alter metabolism so that more ammonium is released to the plant rather than incorporated into bacterial metabolism. A nifL mutant in Azotobacter vinelandii was isolated in the 90s that could fix and release nitrogen even in the presence of 15 mM ammonium [ 151 ]. Deletions of nifL in Azotobacter and Pseudomonas also improved the excretion of ammonium and increased expression of the nif genes in the presence of oxygen [ 152 , 153 ]. Manipulating the bacterial ammonium assimilation pathway is also a straightforward strategy to increase the amount of ammonium released by diazotrophs. Mus et al. achieved that by mutation of glnE in A. vinelandii , preventing the posttranslational repression of glutamine synthetase by ammonium [ 154 ]. This improved diazotrophic growth but impaired growth and reduced fitness on ammonium-containing medium. Similarly, deleting the ammonium transporter amtB led to increased ammonium excretion [ 153 ]. Also, decreased glutamine synthetase activity resulted in ammonium release in A. vinelandii glnA mutants and A. caulinodans glnB or glnK mutants [ 155 , 156 ]. For a more significant review of nitrogen-fixation regulation see [ 13 , 146 ]. As mentioned previously, a significant challenge of biological nitrogen fixation is that the nitrogenase is irreversibly inactivated by oxygen [ 113 , 157 ]. It had been reported that Streptomyces thermoautotrophicus UBT1 possesses a novel class of nitrogenase that was supposedly insensitive to oxygen. This would have been a significant finding. Unfortunately, further studies demonstrated that the described nitrogenase is not present in the S. thermoautotrophicus genome, and the diazotrophic phenotype could not even be recapitulated [ 158 , 159 ]. It is thus unclear if an oxygen-insensitive nitrogenase is even possible. However, efforts are in progress to transfer oxygen protection systems, like the Shethna protein of A. vinelandii , to other diazotrophs [ 160 ]. Efficient biological nitrogen fixation requires close interactions between bacteria and the plant host Plant growth promotion is the result of interactions between soil type, microbiota, and the host plant. Benefits to the plant can come from a wide array of mechanisms [ 161 , 162 ]. Unfortunately, much of the work on these benefits has been limited to describing phenotypes rather than the underlying genetics. The host genotype is also an essential player in defining the microbial communities and their benefits to the interaction partners [ 163 ]. Azoarcus is known to be a very efficient colonizer and was the first non-rhizobial diazotroph with a sequenced genome [ 143 ]. Using mutagenesis studies and labeled bacteria, the mechanisms involved in the Azoarcus –rice interaction have been well described, but not yet translated into practical applications to the field. Early field studies for Azospirillum seem more promising [ 164 – 167 ]. Azospirillum is part of a broad group of plant growth-promoting bacteria, together with endophytic diazotrophs from the genera Herbaspirillum , Gluconacetobacter , Klebsiella , and Burkholderia [ 168 – 171 ]. Infection and colonization of grasses by these endophytes have been well described at the microscopic and physiological levels. In the genus Pseudomonas , several species can colonize plants and promote plant growth efficiently. The transfer of the nitrogenase from Pseudomonas stutzeri to the non-nitrogen fixing root-associated Pseudomonas protegens Pf-5 was suggested to supply nitrogen to several crops [ 128 , 130 ], but, to our knowledge, these results have yet to be replicated by other teams. The same authors showed that heterologous polyhydroxybutyrate production might regulate nitrogenase activity. Polyhydroxybutyrate is a carbon storage polymer that can be mobilized under stressful physiological conditions, increasing the survival of bacteria in the soil. Indeed, recently, H. seropedicae strains overproducing polyhydroxybutyrate were shown to have better colonization fitness when compared to wild-type strains [ 172 ]. This highlights the importance of studies that integrate nitrogen fixation into bacteria with better plant colonization ability. Finding genes that improve this colonization ability will open new avenues to increase inoculant efficiency and survival between crops. Another approach towards searching for better colonizers are microbiome studies and, in particular, those going beyond 16S-based classification to look at functional genes. Such efforts include the Earth Microbiome Project, which collected information for more than 30,000 microbiota samples across the globe [ 173 ]. Recent work compared 3837 bacterial genomes, aiming to identify plant-associated gene clusters, and found that plant-associated bacteria genomes encoded more carbohydrate metabolism genes than related non-plant-associated genomes and determined 64 plant-associated protein domains that possibly mimic plant domains [ 174 ]. This can potentially lead us to a comprehensive set of genes that directly affect the symbiotic interaction between bacteria and non-legume hosts. The search for better plant hosts for nitrogen-fixing bacteria In the quest for nitrogen-fixing crops, a lot of the community efforts have been focused on legumes, which, as reviewed previously, led to a wealth of knowledge on root nodule symbiosis, but the practical applications of this knowledge are probably a long-term goal. Previous attempts involved the transfer of seven core CSP genes from M. truncatula to a variety of non-fixing eudicots and were unsuccessful at inducing nodulation [ 175 ]. We now know that these species already contained functional orthologs of these genes, as they are conserved for signaling in the ancestral arbuscular mycorrhizal symbiosis. A complete rebuild of the nodulation pathway in a non-host would probably require a large number of genes and may as yet be impossible with our current understanding of the symbiosis. But, efforts to ‘brute force’ a new host by transfer of all or a substantial core set of nodulin genes to monocotyledonous crops is likely unnecessary. The very concept of ‘nodulin’ genes is questionable as large-scale transcriptomic approaches demonstrate that many of these genes are expressed in other tissues or conditions [ 176 ]. Most if not all genes involved in nodulation have been repurposed from existing conserved families, including roles in homeotic flower development (NOOT) [ 177 ], root architecture in response to nitrogen (NIN family transcription factors) [ 66 ], and the autoregulation of nodulation pathway [ 178 ] and defense (Nodule cysteine-rich peptides) [ 35 , 179 – 181 ]. A more efficient approach is probably to exploit the conservation of most ‘nodulin’ genes outside of the FaFaCuRo clade, taking an evolution-guided ‘minimal change’ approach to engineering. Griesmann et al. suggested that the change that enabled nodulation was the coordination of expression of ‘nodulin’ genes, rather than the appearance of new genes not seen outside the FaFaCuRo clade [ 35 ]. The same idea, that the evolution of nodulation was a gain of regulatory elements rather than protein coding sequences, was also proposed by Doyle [ 182 ]. Taking this evolution-guided approach step by step through the stages of the nodule symbiosis, we first observe that all plants release at least the basal flavonoid naringenin, which is known to activate nod gene expression in several rhizobial species [ 183 ]. Thus, it is likely easier to move the NodD gene from these species to other rhizobia than alter flavonoid metabolism in the plant. All mycorrhizal plants contain LCO receptors capable to some extent of binding rhizobial Nod factors, although these ‘mycorrhizal’ LysM receptors seem to have lower sensitivity than their legume homologs [ 184 ]. Adding legume receptors, co-evolved for millions of years for specificity with their symbiont, to non-nodulating plants may help improve a new symbiosis but is unlikely to be necessary to trigger the CSP in response to the Nod factors of an engineered symbiont. Most legumes undergo “root hair infection” where a microcolony of rhizobia is enclosed by a curling root hair, and invagination of the host membrane forms an intracellular “infection thread” through which the bacteria move into the root cortex. However, root hair infection is dispensable for symbiotic nitrogen fixation, as demonstrated by nodulators with “crack entry” mechanisms such as peanut, and the L. japonicus mutants root hairless and slippery root , where the rhizobia enter the root via the crack formed by an emerged lateral root and form infection structures directly in the cortex [ 185 ]. The enclosure of bacteria in a host membrane (called the symbiosome) is likely an essential step for the efficiency of the symbiosis. However, this basal ‘infection module’, the group of genes that permit intracellular infection by microsymbionts, is conserved in all plants able to associate with arbuscular mycorrhizal fungi (Table 3 ). The genes that form this module are not yet fully characterized, but their conservation in nodulation and arbuscular mycorrhization is demonstrated by the example of VAPYRIN and VAMP721d /e, which are essential to both symbioses, as they establish the secretory pathway used to build the symbiosome during nodulation, and peri-arbuscular membranes during mycorrhization [ 190 ]. Symbiosomes in legumes are endocytosed from the plasma membrane, but in other nodulating plants, such as Parasponia andersonii , the infection thread remains contiguous with the plasma membrane, as it does in the arbuscular mycorrhizal symbiosis. Rhizobia still differentiate into bacteroids and fix nitrogen at high efficiency in these fixation threads, supporting the concept of intermediate stages of evolution that an engineering project could take advantage of [ 191 ].\n Table 3 The common symbiosis pathway (CSP) controls the establishment of rhizobia–legume associations and the arbuscular mycorrhizal symbiosis The common symbiosis pathway (CSP) controls the establishment of rhizobia–legume associations and the arbuscular mycorrhizal symbiosis. Arbuscular mycorrhizal fungi (Mucoromycotina) produce diffusible Myc factors composed of short chitin oligomers as well as lipo-chitooligosaccharides similar to rhizobial Nod factors. These fungal signals are perceived by LysM RLKs similar to the Nod factor receptors [ 62 , 186 , 187 ]. The arbuscular mycorrhizal association appeared with the first land plants about 450 million years ago and is still found in more than 70% land plants, including most legumes and cereals [ 188 ]. In contrast, root nodule symbioses appeared much more recently, around 100 million years ago, and are restricted to plants of the “FaFaCuRo clade” [ 182 , 189 ]. It seems likely that the nitrogen-fixing bacteria mimicked fungal signals and co-opted the ancient and widespread mycorrhizal pathway. A critical missing link for nodulation outside the FaFaCuRo clade is likely the activation of ‘nodulins’ by the calcium oscillations of the CSP [ 72 , 192 ]. Thus, the main challenge of the ‘minimum change' approach would be to add or alter promoter elements in an essential set of conserved ‘nodulin’ genes to coordinate their expression in response to nuclear calcium spiking. Some of the genes that make up this essential set are currently known (for example, NFRs, LYK3, CCaMK, IPD3/CYCLOPS, CASTOR/POLLUX, NIN, NSP1, NSP2, LHK1), but others would have to be elucidated through further research. One potential issue with this strategy is that it is contingent on how plants in the FaFaCuRo clade differentiate between arbuscular mycorrhizal and rhizobial signaling (as the same calcium spiking appears to elicit different gene expression, suggesting the existence of unknown secondary pathways [ 189 , 193 , 194 ]), but this is still a genetic black box beyond the scope of current knowledge). In contrast to infection, shaping the organogenesis module of nodulation also may need only a few significant changes in non-nodulating plants. The basal actinorhizal plants produce nodules with central vasculature that arise from the pericycle, differentiated from lateral roots only by the cessation of growth and hosting of symbiosomes [ 34 ]. Nodules within the legumes are more elaborate, probably recruiting further genes to aid the symbiosis (for example, leghemoglobin for oxygen protection) [ 106 , 195 ]. However, while these changes likely improve efficiency, they are probably dispensable and may be replaceable by bacterial functions [ 117 , 196 , 197 ]. So, what is necessary to trigger organogenesis of a nodule rather than a lateral root? NIN, NF-Y proteins, and other components that regulate lateral root initiation in response to nitrogen starvation must be repurposed, activating their expression in the tissue layer destined to give rise to the nodule, in response to bacterial signaling. In legumes, this is characterized by a coordinated buildup of cytokinin and auxin to drive cell dedifferentiation and activation of the cell cycle, so links between these transcription factors and hormone synthesis will need to be confirmed or added in non-nodulating plants [ 67 , 75 ]. A key difference between legume nodules with peripheral vasculature and lateral roots with a central vasculature appears to be controlled by homeotic transcription factors of the NOOT family. In the nodules of legume noot mutants the vasculature alternates between a peripheral or central location along the length of the nodule, an apparent reversion to a lateral root or to an actinorhizal nodule identity [ 198 , 199 ]. NOOT orthologs are present in non-legumes, but their function is unknown. Attempts to demonstrate the feasibility of this evolution-guided approach would be best undertaken in a close relative of the FaFaCuRo clade, to maximize the protein similarity of conserved ‘nodulins’. Of these relatives, poplar ( Populus sp.) is an attractive model for the expansion of nodulation, given the ease of transformation and the phylogenetic proximity to the FaFaCuRo clade. The long-term objective of such approaches is, of course, to engineer root nodulation in cereals crops. Nitrogen-fixing associations outside the FaFaCuRo clade open new horizons Engineering associative nitrogen fixation should, in theory, be more straightforward than engineering root nodules and intracellular infection or expressing the nitrogenase in plants. However, the expansion of symbiotic associative fixation faces a significant challenge due to the poor understanding of the genetic requirements that allow a host to associate with and benefit from diazotrophs. The benefit obtained by the host is likely governed by three factors: nitrogen uptake efficiency at low concentrations, defense responses, and the amount of carbon available to the diazotrophs. Blind manipulation of the latter two is likely to lead to problems with pathogens or competition from non-fixing rhizospheric microorganisms. Nitrogen uptake efficiency has been a breeding target, though it is often in a trade-off with the efficient uptake at high concentrations that intensity agriculture breeds for. Many crops benefit from some level of soil fixation (usually > 20 kg N ha −1 yr −1 , but decreasing on nitrogen fertilization [ 123 , 200 , 201 ]), likely powered by photosynthates in root exudates. However, many more elaborate, and more efficient, nitrogen-fixing symbioses have been discovered in nature [ 12 ]. Of particular interest is the fixation on aerial roots of maize landraces from the Sierra Mixe [ 83 ]. These maize accessions produce aerial roots on many more nodes than conventional maize accessions. Upon rain, these roots secrete a sugar-rich mucilage, which houses diazotrophs that contribute 29–82% of the plant’s nitrogen [ 83 ]. Preliminary evidence suggests that tropical accessions of other cereals like sorghum may possess the same trait of abundant mucilage production by aerial roots [ 202 ]. Another example is Brazilian sugarcane, which obtains nitrogen from bacteria (most notably Gluconacetobacter diazotrophicus ) housed within the stem, contributing up to 30% of the plant’s nitrogen [ 127 ]. The rate of biological nitrogen fixation is known to depend on the plant cultivar, and the phenotype seems dependent on the environment, but we are not aware of any exploration of the genetic basis of this trait [ 127 , 203 ]. This type of associative nitrogen fixation provides an enormous well of untapped potential, and more efforts should be devoted to their study. Advantages and environmental concerns with nitrogen-fixing crops and microbes engineered for biological nitrogen fixation Intensive agriculture leads to environmental degradation on a global scale. Microbial inoculants promise an alternative eco-friendly practice, reducing the amount of fertilizer usage. However, it is worth remembering that legumes themselves can lead to significant nitrogen leaching when crop residues are mineralized; thus, agronomic practices such as reduction of tillage and using cover crops must also be considered to solve these environmental issues [ 204 ]. Commercially available bioinoculants for non-legumes uses plant growth-promoting rhizobacteria, but the efficiency of these products in incorporating fixed nitrogen is still limited and variable depending on the environment (for an extensive review see [ 205 ]). Azospirillum is a versatile inoculant because it not only fixes nitrogen but also mineralizes nutrients from the soil and sequesters iron [ 206 ] (for a more comprehensive review see [ 207 ]). On the other hand, endophytic bacteria, such as Azoarcus sp., Herbaspirillum sp., and G. diazotrophicus , appear promising candidates as they colonize the intercellular spaces, so fixed nitrogen is likely released directly to the plant without competition from the rhizosphere community [ 22 , 120 , 129 , 145 ]. However, these endophytic bacteria display only a mild plant growth promotion effect. Thus, it is essential to improve the efficiency of the ammonium release from live microbes as opposed to relying on the release after cell death. It will also be necessary to understand better the microbial traits required for plant colonization, persistence, and competitiveness in the plant microbiota. Similarly, the impacts of plant growth-promoting rhizobacteria on endogenous microbial communities are understudied. The effect of these newcomers on preexisting microbial populations, and the useful ecosystem services which they provide, is unknown. One of the first studies to demonstrate the influence of the environment in the establishment of beneficial microbe–host interactions was conducted by Dr. Johanna Döbereiner, who showed in 1961 that the growth promoter of sugarcane, Beijerinckia , was dependent on rainfall [ 121 ]. Similarly, the rain on aerial roots of maize is required for secretion of mucilage [ 83 ]. More generally, the concept of a disease triangle in which host, microbe, and environment interact can be applied to beneficial microbes too. We believe that prospecting for better diazotrophs and better host plant genotypes combined with engineering approaches has the potential to deliver transformative agricultural tools and that different host–microbe combinations may be necessary for different environments. Can we shortcut the bacteria and develop plants that fix nitrogen directly? Engineering plants is generally more challenging than manipulating bacteria, primarily due to generation time and the bottleneck of plant transformation. However, a crop that fixes nitrogen without the need for microbes would have an agronomic impact without precedent. Current attempts to generate a nitrogen-fixing eukaryote have favored assembling the active nitrogenase inside chloroplasts or mitochondria. These organelles are the main sites of ATP synthesis, and so are most able to meet the high energetic requirements of the nitrogenase. López-Torrejón et al. showed that yeast mitochondria were anoxic enough to allow for the accumulation of active NifU and NifH and that, in the presence of NifM, NifH could incorporate endogenous mitochondrial Fe-S clusters [ 208 ]. Attempts to engineer transgenic yeast expressing nitrogenase have led to the identification of a minimal nif cassette of nine genes sufficient for nitrogen fixation. The stoichiometric ratios of these nine nitrogenase components are critical for the assembly of a functional holoenzyme [ 209 ]. Burén et al. showed that refactoring approaches could be used to recapitulate that in eukaryotes [ 210 ]. Assembly of large hetero-tetrameric complexes has proved challenging. The use of ‘giant gene’ constructs separated by peptides cleaved by the ribosome or proteases have been attempted, but the cleavage overhangs have been shown to impair both targeting and folding. Re-assemblies using giant genes have been demonstrated to fix nitrogen in E. coli , but functionality in a eukaryotic system is yet to be shown [ 211 ]. Allen et al. demonstrated that these lessons could be applied to plants, expressing 16 nif genes in the tobacco mitochondrial matrix [ 212 ]. For extensive reviews about strategies to transfer nif genes to eukaryotes refer to [ 209 , 213 ]. If expressed in the chloroplast, the nitrogenase would be exposed to an ATP-rich environment and should not be exposed to high oxygen levels during the night [ 214 ]. Some cyanobacteria like Synechococcus perform photosynthesis during the day and fix nitrogen during the night, thus uncoupling photosynthesis and nitrogen fixation temporally [ 25 ]. The evolutionary relatedness of plant chloroplasts to cyanobacteria suggests that it may be possible to engineer such a “night shift” in plant chloroplasts. Ivleva et al. produced transplastomic tobacco plants expressing NifH/NifM, which was active in vitro under low oxygen conditions (10% O 2 ) in the presence of the molybdenum-iron protein from A. vinelandii [ 215 ]. The current lack of evidence of nitrogenase function in eukaryotes, combined with the lack of a high-throughput plastid transformation procedure for monocots, means that the development of nitrogen-fixing cereals is still a long-term prospect. Advantages of and concerns with nitrogen-fixing crops Developing plants that could fix and assimilate nitrogen without the help of microbial partners would alleviate the adverse effects of nitrogen fertilizers on the environment and benefit developing countries by facilitating higher yield in low input systems. Despite the genetic challenge, a plant capable of directly fixing nitrogen will be more robust than symbiotic nitrogen fixation, as it would decrease nitrogen loss to other organisms. Ammonium produced by nitrogenase could likely be coupled to plant metabolism in the plastid or mitochondria, further increasing efficiency [ 16 , 209 ]. However, this approach will probably need much careful refinement, because if nitrogenase activity is not coupled to substrate delivery, the nitrogenase could divert large proportions of cellular resources to the futile evolution of hydrogen, imposing a significant yield drag on the plant [ 216 ]. Another considerable advantage of self-fixing plants would be the freedom from the partner requirement of symbiotic nitrogen fixation, as germline transmission would provide for more straightforward distribution and require less infrastructure from the farmer, compared to a symbiotic nitrogen fixation approach which would require inoculation. Concerning food security, transgenic plants are regulated and cultivated in many countries, and so far no soil, environment, or health issues have been correlated to it. It is necessary that the current regulation is revisited to avoid unnecessary fears preventing society from benefiting from this technology, which has the potential to make food production more environmentally sustainable and help feed the increasing world population. Conclusions and perspectives The Food and Agriculture Organization of the United Nations estimates that the Earth will have two billion more people to feed in 2050 [ 217 ]. Given that half of the world population is currently sustained through synthetic fertilizers, it would not be reasonable to claim that biological nitrogen fixation will replace the Haber-Bosch process entirely. But, as indicated earlier, the extreme dependence of the global food supply on synthetic fertilizers is not sustainable. Now is the time for a “Symbiotic Revolution” to combine food production and sustainable soil health. So, are we there yet? Many avenues to improve biological nitrogen fixation in non-leguminous crops have been described in this review (Fig. 2 ). Some of them could bring solutions in the next decade, and some will probably bear fruit in the longer term [ 83 , 128 , 208 , 211 , 212 ]. Some of the natural systems we presented can provide significant amounts of biologically fixed nitrogen. Microbial and plant natural diversity is a resource and a source of knowledge that should be explored more, and that could deliver practical solutions in a relatively short time. We take as an example the Sierra Mixe maize, where an unexplored system is capable of sustaining most of the nitrogen requirement for the crop over several months and at a critical period of the growing season [ 83 ]. Such unexpected discoveries reinforce the need to preserve natural diversity in our crops and their wild relatives. These Sierra Mixe landraces cannot be used directly in most cropping systems or environments due to their size and their long growing season. Breeding the trait in more conventional accessions of maize is necessary, but this process will take time. Once this trait is introduced into more conventional varieties, agronomic questions such as the amount of fertilizer saved by the trait, the effect of soil nitrogen, or the yield cost of the trait will need to be addressed. This will require efforts and funding, but it seems achievable to use such natural traits in the next decades. Approaches aiming at engineering root nodules in cereals are more complex and will likely take more time. Procedures to engineer crops able to fix their nitrogen without the bacteria seem even longer term. Nevertheless, as discussed previously, these long-term approaches are promising and likely to succeed.\n Fig. 2 Main approaches to engineer or improve biological nitrogen fixation in cereals. 1 Improving nitrogen-fixing bacteria: ( a ) [ 153 ], ( b ) [ 111 , 113 ], ( c ) [ 154 ], ( d ) [ 152 ], ( e ) [ 110 , 144 , 171 ]. 2 Making crops better hosts for nitrogen-fixing bacteria: ( a ) [ 127 ], ( b ) [ 83 ], ( c ) [ 56 ]. 3 Allowing crops to fix their nitrogen without microbes [ 209 ] A critical justification for pursuing the range of approaches described in this review is the significant effect that the environment has on many of these biological systems. Taking the example of the Sierra Mixe maize again, mucilage production by the maize aerial roots is dependent on rain [ 83 ]. While this trait seems to be directly usable in many regions of the world, it will be more challenging to adapt it to arid environments. If we look at the worldwide distribution of nodulating legumes, engineered root nodules may be efficient in a broader range of situations, but legume nodulation itself is affected by environmental factors such as soil nitrogen or flooding [ 218 – 220 ]. The environmental dependence of plants fixing their own nitrogen is, at this point, entirely speculative. Lastly, as indicated earlier, the process of nitrogen fixation, whether biological or industrial, requires significant amounts of energy. In all the approaches to improve nitrogen fixation discussed in this review, the energy for nitrogen fixation comes from plant photosynthesis and will have a cost on plant carbon. Despite the tight autoregulation of nodulation, legumes dedicate 10–20% of their carbon to nodules [ 117 ]. This does not necessarily decrease yield, as carbon cost is offset by increased photosynthetic capacity due to the nitrogen gained from biological nitrogen fixation. However, symbiotic nitrogen fixation will not be energetically competitive compared to nitrogen fertilization, and growers in developed countries are not ready to suffer any yield loss while fertilization remains cheap. Estimating the potential yield penalty for different strategies will be necessary. Environmental policies may provide more incentives for a reduction of synthetic fertilizers in the future [ 221 ]. In developing countries, any nitrogen input will be valuable for poor farmers where nitrogen is the most important factor limiting their production [ 222 ]. International projects such as Realizing Increased Photosynthetic Efficiency (RIPE) are currently working on improving photosynthetic efficiency, which could offset the yield penalty of relying on biological nitrogen fixation [ 223 , 224 ]. Improving nitrogen fixation in non-leguminous crops has been a dream of the agronomic community for more than a century. The global challenges that our world is facing make the realization of this dream urgent. Fortunately, natural diversity holds solutions that the scientific community overlooked possibly because of the intense focus on legume nodules. Technological developments such as the advent of next-generation sequencing, gene editing, and synthetic biology allow the dissection and manipulation of plants and microbes at an unprecedented scale. We are confident that combining the prospecting of plant and bacterial natural diversity with genetic engineering will deliver solutions in the short and long terms and will help to feed the world in a more sustainable manner."
} | 16,122 |
39003244 | PMC11287213 | pmc | 388 | {
"abstract": "Abstract Growing environmental concerns and the need to adopt a circular economy have highlighted the importance of waste valorization for resource recovery. Microbial consortia-enabled biotechnologies have made significant developments in the biomanufacturing of valuable resources from waste biomass that serve as suitable alternatives to petrochemical-derived products. These microbial consortia-based processes are designed following a top-down or bottom-up engineering approach. The top-down approach is a classical method that uses environmental variables to selectively steer an existing microbial consortium to achieve a target function. While high-throughput sequencing has enabled microbial community characterization, the major challenge is to disentangle complex microbial interactions and manipulate the structure and function accordingly. The bottom-up approach uses prior knowledge of the metabolic pathway and possible interactions among consortium partners to design and engineer synthetic microbial consortia. This strategy offers some control over the composition and function of the consortium for targeted bioprocesses, but challenges remain in optimal assembly methods and long-term stability. In this review, we present the recent advancements, challenges, and opportunities for further improvement using top-down and bottom-up approaches for microbiome engineering. As the bottom-up approach is relatively a new concept for waste valorization, this review explores the assembly and design of synthetic microbial consortia, ecological engineering principles to optimize microbial consortia, and metabolic engineering approaches for efficient conversion. Integration of top-down and bottom-up approaches along with developments in metabolic modeling to predict and optimize consortia function are also highlighted. One-Sentence Summary This review highlights the microbial consortia-driven waste valorization for biomanufacturing through top-down and bottom-up design approaches and describes strategies, tools, and unexplored opportunities to optimize the design and stability of such consortia.",
"conclusion": "Conclusions There have been unprecedented efforts to harness microbial consortia to develop several biotechnologies for biomanufacturing. This review specifically highlights the potential of microbial consortia and discusses the top-down and bottom-up engineering design approaches including the challenges and future recommendations. The top-down strategy still remains the most commonly used approach for waste valorization due to its relative ease of implementation. The bottom-up approach holds promise, but successful implementation will require developing techniques for stable consortia assembly, process optimization using ecological principles, and metabolic engineering to develop stable synthetic microbial consortia. There has been significant progress in engineering individual strains for biomanufacturing, however, less focus has been placed on consortium engineering. While models exist for single microbial metabolism, there is a need for models that can predict and identify the metabolic features governing interactions and long-term stability within microbial consortia. Besides focusing on the technical advancements, the economics should also be considered as more resources are needed for developing synthetic microbial consortia using the bottom-up approach compared to the traditional top-down approach. For instance, it might be economically beneficial to invest in developing and improving synthetic consortia for bioprocesses that produce high-value specialty chemicals. Future research should also explore integrating undefined natural microbial consortium (top-down) with defined synthetic consortium (bottom-up) for accomplishing efficient conversion of complex waste biomass. An interdisciplinary approach of combining bioprocess engineering, multi-omics analysis, metabolic engineering, system biology, and metabolic modeling is needed for a deeper understanding and optimization of microbial consortia-driven biomanufacturing for a sustainable future.",
"introduction": "Introduction A staggering 85% of the world's energy consumption is derived from nonrenewable fossil fuels (Cleveland & Morris, 2014 ; World Energy Use, 2022 ). This reliance has led to greenhouse gas (GHG) emissions and severely impacted natural ecosystem and biodiversity, necessitating sustainable alternatives to petrochemical-derived products. Biomanufacturing harnesses the power of microorganisms or enzymes to produce biofuels and bioproducts thus reducing our dependence on fossil fuels. Biomass-based products could replace up to 16% of crude oil consumption in the U.S., generating an additional $812 billion profit (Bioproducts to Enable Biofuels Workshop Summary Report, 2015 ). Biomass provided 5% of the total energy consumption in the U.S. in 2022, reducing the reliance on traditional energy sources (Biomass Explained, 2023 ). To achieve the ultimate goal of achieving a net-zero emission by 2050 and a carbon-neutral economy, biobased products will be further incentivized the U.S. It is estimated that 2.01 billion tons of municipal solid waste is generated annually with more than 33% of total waste managed in an environmentally unsafe manner (Kaza et al., 2018 ). According to the U.S. Environmental Protection Agency, 66.2 million tons of food waste was generated from food retail and an additional 40.1 million tons from food processing and manufacturing, causing significant economic loss and environmental issues in 2019 (2019 Wasted Food Report, 2023 ). Traditional waste management methods, such as composting, landfill disposal, and incineration have limitations such as low process efficiency and negative environmental impacts, including space requirements, generation of odor, contaminated leachate, and toxic pollutants, and ash emissions (Phua et al., 2019 ; Andraskar et al., 2021 ; Parvin & Tareq, 2021 ). Inexpensive and readily available wet waste (Tayou et al., 2022 ), lignocellulosic waste (Li et al., 2020 ; Shahab et al., 2020a ; Wongfaed et al., 2023 ), and C1 waste (CO 2 , CO) (Diender et al., 2016 ; Jiang et al., 2021 ), are promising renewable feedstocks for the production of value-added products to enable biomanufacturing. Implementing innovative technologies for converting such untapped waste to useful products can address negative impacts due to waste accumulation and management problems, substitute fossil fuel-based products, and strengthen the bioeconomy. Microbial communities greatly impact various aspects of life ranging from biogeochemical cycles, medicine, bioremediation, and public health, to biomanufacturing and resource recovery. Microbial consortia are suited to convert complex waste biomass due to their high enzyme diversity and the concerted and syntrophic activity of microorganisms belonging to different functional groups. Biotechnologies employing microbial consortia for sustainable waste valorization have emerged as promising alternatives to the petrochemical refinery processes to produce valuable biofuels, bioplastics, biochemicals, enzymes, and single-cell proteins (Venkateswar Reddy & Venkata Mohan, 2012 ; Zhou et al., 2017 ; Chi et al., 2018 ; Reddy et al., 2018 ; Valentino et al., 2018 ; Li et al., 2020 ; Pagliano et al., 2020 ; Tayou et al., 2022 ). Several of these biotechnologies have moved beyond lab and pilot scales and seen commercialization and expansion in recent years, contributing to the bioeconomy. For instance, anaerobic digestion (AD) has been widely adopted as a biological waste treatment and resource recovery technology to produce biogas which is further converted into electricity and heat. According to the World Biogas Association, an estimated 132 000 small, medium, and large-scale ADs are operational globally (World Biogas Association Global Report, 2019 ). The advancement in AD has been possible by understanding the role of microbial consortia and engineering it for efficient degradation of a wider range of feedstocks, tolerance to inhibitory compounds, and resilience to environmental perturbations (Werner et al., 2011 ; Blair et al., 2021 ). Microbial consortia-based biotechnologies harness the metabolic capacity of microorganisms and their synergistic interactions by employing either top-down or bottom-up approaches for microbiome engineering. The top-down approach involves providing selective pressure by manipulating environmental or operating conditions to steer the structure and activity of the natural microbial consortia toward a desired function. On the contrary, the bottom-up approach starts by understanding individual microbial characteristics to rationally assemble native or engineered microorganisms into a new synthetic consortium. However, such microbial consortia-based processes still suffer from undesirable side reactions, low process efficiency, and inability to control and maintain stability for a long term, which is partly due to exposure to external perturbations and unpredictable intercellular interactions. Previous reviews on microbial consortia have focused on either the top-down or bottom-up microbiome engineering approaches for environmental, public health, medical, or biotechnology applications (Duncker et al., 2021 ; Hu et al., 2022 ; Sauer & Marx, 2023 ; Zhou et al., 2024 ). However, there is a lack of comprehensive reviews that provide critical perspectives on both top-down and bottom-up approaches, and their integration to enable biomanufacturing from waste streams. This perspective paper first explores the promising potential of microbial consortia, highlighting their industrial and environmental applications utilizing diverse waste streams. Next, the paper elucidates recent advancements and knowledge gaps in top-down and bottom-up approaches with a focus on the design and assembly of synthetic microbial consortia, ecological engineering for process optimization, and metabolic engineering of microbial consortia. Finally, we also discuss the combination of the top-down and bottom-up approaches to maximize the potential of microbial consortia in different scenarios as well as metabolic modeling to predict and guide the microbial consortia design."
} | 2,569 |
39003244 | PMC11287213 | pmc | 388 | {
"abstract": "Abstract Growing environmental concerns and the need to adopt a circular economy have highlighted the importance of waste valorization for resource recovery. Microbial consortia-enabled biotechnologies have made significant developments in the biomanufacturing of valuable resources from waste biomass that serve as suitable alternatives to petrochemical-derived products. These microbial consortia-based processes are designed following a top-down or bottom-up engineering approach. The top-down approach is a classical method that uses environmental variables to selectively steer an existing microbial consortium to achieve a target function. While high-throughput sequencing has enabled microbial community characterization, the major challenge is to disentangle complex microbial interactions and manipulate the structure and function accordingly. The bottom-up approach uses prior knowledge of the metabolic pathway and possible interactions among consortium partners to design and engineer synthetic microbial consortia. This strategy offers some control over the composition and function of the consortium for targeted bioprocesses, but challenges remain in optimal assembly methods and long-term stability. In this review, we present the recent advancements, challenges, and opportunities for further improvement using top-down and bottom-up approaches for microbiome engineering. As the bottom-up approach is relatively a new concept for waste valorization, this review explores the assembly and design of synthetic microbial consortia, ecological engineering principles to optimize microbial consortia, and metabolic engineering approaches for efficient conversion. Integration of top-down and bottom-up approaches along with developments in metabolic modeling to predict and optimize consortia function are also highlighted. One-Sentence Summary This review highlights the microbial consortia-driven waste valorization for biomanufacturing through top-down and bottom-up design approaches and describes strategies, tools, and unexplored opportunities to optimize the design and stability of such consortia.",
"conclusion": "Conclusions There have been unprecedented efforts to harness microbial consortia to develop several biotechnologies for biomanufacturing. This review specifically highlights the potential of microbial consortia and discusses the top-down and bottom-up engineering design approaches including the challenges and future recommendations. The top-down strategy still remains the most commonly used approach for waste valorization due to its relative ease of implementation. The bottom-up approach holds promise, but successful implementation will require developing techniques for stable consortia assembly, process optimization using ecological principles, and metabolic engineering to develop stable synthetic microbial consortia. There has been significant progress in engineering individual strains for biomanufacturing, however, less focus has been placed on consortium engineering. While models exist for single microbial metabolism, there is a need for models that can predict and identify the metabolic features governing interactions and long-term stability within microbial consortia. Besides focusing on the technical advancements, the economics should also be considered as more resources are needed for developing synthetic microbial consortia using the bottom-up approach compared to the traditional top-down approach. For instance, it might be economically beneficial to invest in developing and improving synthetic consortia for bioprocesses that produce high-value specialty chemicals. Future research should also explore integrating undefined natural microbial consortium (top-down) with defined synthetic consortium (bottom-up) for accomplishing efficient conversion of complex waste biomass. An interdisciplinary approach of combining bioprocess engineering, multi-omics analysis, metabolic engineering, system biology, and metabolic modeling is needed for a deeper understanding and optimization of microbial consortia-driven biomanufacturing for a sustainable future.",
"introduction": "Introduction A staggering 85% of the world's energy consumption is derived from nonrenewable fossil fuels (Cleveland & Morris, 2014 ; World Energy Use, 2022 ). This reliance has led to greenhouse gas (GHG) emissions and severely impacted natural ecosystem and biodiversity, necessitating sustainable alternatives to petrochemical-derived products. Biomanufacturing harnesses the power of microorganisms or enzymes to produce biofuels and bioproducts thus reducing our dependence on fossil fuels. Biomass-based products could replace up to 16% of crude oil consumption in the U.S., generating an additional $812 billion profit (Bioproducts to Enable Biofuels Workshop Summary Report, 2015 ). Biomass provided 5% of the total energy consumption in the U.S. in 2022, reducing the reliance on traditional energy sources (Biomass Explained, 2023 ). To achieve the ultimate goal of achieving a net-zero emission by 2050 and a carbon-neutral economy, biobased products will be further incentivized the U.S. It is estimated that 2.01 billion tons of municipal solid waste is generated annually with more than 33% of total waste managed in an environmentally unsafe manner (Kaza et al., 2018 ). According to the U.S. Environmental Protection Agency, 66.2 million tons of food waste was generated from food retail and an additional 40.1 million tons from food processing and manufacturing, causing significant economic loss and environmental issues in 2019 (2019 Wasted Food Report, 2023 ). Traditional waste management methods, such as composting, landfill disposal, and incineration have limitations such as low process efficiency and negative environmental impacts, including space requirements, generation of odor, contaminated leachate, and toxic pollutants, and ash emissions (Phua et al., 2019 ; Andraskar et al., 2021 ; Parvin & Tareq, 2021 ). Inexpensive and readily available wet waste (Tayou et al., 2022 ), lignocellulosic waste (Li et al., 2020 ; Shahab et al., 2020a ; Wongfaed et al., 2023 ), and C1 waste (CO 2 , CO) (Diender et al., 2016 ; Jiang et al., 2021 ), are promising renewable feedstocks for the production of value-added products to enable biomanufacturing. Implementing innovative technologies for converting such untapped waste to useful products can address negative impacts due to waste accumulation and management problems, substitute fossil fuel-based products, and strengthen the bioeconomy. Microbial communities greatly impact various aspects of life ranging from biogeochemical cycles, medicine, bioremediation, and public health, to biomanufacturing and resource recovery. Microbial consortia are suited to convert complex waste biomass due to their high enzyme diversity and the concerted and syntrophic activity of microorganisms belonging to different functional groups. Biotechnologies employing microbial consortia for sustainable waste valorization have emerged as promising alternatives to the petrochemical refinery processes to produce valuable biofuels, bioplastics, biochemicals, enzymes, and single-cell proteins (Venkateswar Reddy & Venkata Mohan, 2012 ; Zhou et al., 2017 ; Chi et al., 2018 ; Reddy et al., 2018 ; Valentino et al., 2018 ; Li et al., 2020 ; Pagliano et al., 2020 ; Tayou et al., 2022 ). Several of these biotechnologies have moved beyond lab and pilot scales and seen commercialization and expansion in recent years, contributing to the bioeconomy. For instance, anaerobic digestion (AD) has been widely adopted as a biological waste treatment and resource recovery technology to produce biogas which is further converted into electricity and heat. According to the World Biogas Association, an estimated 132 000 small, medium, and large-scale ADs are operational globally (World Biogas Association Global Report, 2019 ). The advancement in AD has been possible by understanding the role of microbial consortia and engineering it for efficient degradation of a wider range of feedstocks, tolerance to inhibitory compounds, and resilience to environmental perturbations (Werner et al., 2011 ; Blair et al., 2021 ). Microbial consortia-based biotechnologies harness the metabolic capacity of microorganisms and their synergistic interactions by employing either top-down or bottom-up approaches for microbiome engineering. The top-down approach involves providing selective pressure by manipulating environmental or operating conditions to steer the structure and activity of the natural microbial consortia toward a desired function. On the contrary, the bottom-up approach starts by understanding individual microbial characteristics to rationally assemble native or engineered microorganisms into a new synthetic consortium. However, such microbial consortia-based processes still suffer from undesirable side reactions, low process efficiency, and inability to control and maintain stability for a long term, which is partly due to exposure to external perturbations and unpredictable intercellular interactions. Previous reviews on microbial consortia have focused on either the top-down or bottom-up microbiome engineering approaches for environmental, public health, medical, or biotechnology applications (Duncker et al., 2021 ; Hu et al., 2022 ; Sauer & Marx, 2023 ; Zhou et al., 2024 ). However, there is a lack of comprehensive reviews that provide critical perspectives on both top-down and bottom-up approaches, and their integration to enable biomanufacturing from waste streams. This perspective paper first explores the promising potential of microbial consortia, highlighting their industrial and environmental applications utilizing diverse waste streams. Next, the paper elucidates recent advancements and knowledge gaps in top-down and bottom-up approaches with a focus on the design and assembly of synthetic microbial consortia, ecological engineering for process optimization, and metabolic engineering of microbial consortia. Finally, we also discuss the combination of the top-down and bottom-up approaches to maximize the potential of microbial consortia in different scenarios as well as metabolic modeling to predict and guide the microbial consortia design."
} | 2,569 |
39003244 | PMC11287213 | pmc | 389 | {
"abstract": "Abstract Growing environmental concerns and the need to adopt a circular economy have highlighted the importance of waste valorization for resource recovery. Microbial consortia-enabled biotechnologies have made significant developments in the biomanufacturing of valuable resources from waste biomass that serve as suitable alternatives to petrochemical-derived products. These microbial consortia-based processes are designed following a top-down or bottom-up engineering approach. The top-down approach is a classical method that uses environmental variables to selectively steer an existing microbial consortium to achieve a target function. While high-throughput sequencing has enabled microbial community characterization, the major challenge is to disentangle complex microbial interactions and manipulate the structure and function accordingly. The bottom-up approach uses prior knowledge of the metabolic pathway and possible interactions among consortium partners to design and engineer synthetic microbial consortia. This strategy offers some control over the composition and function of the consortium for targeted bioprocesses, but challenges remain in optimal assembly methods and long-term stability. In this review, we present the recent advancements, challenges, and opportunities for further improvement using top-down and bottom-up approaches for microbiome engineering. As the bottom-up approach is relatively a new concept for waste valorization, this review explores the assembly and design of synthetic microbial consortia, ecological engineering principles to optimize microbial consortia, and metabolic engineering approaches for efficient conversion. Integration of top-down and bottom-up approaches along with developments in metabolic modeling to predict and optimize consortia function are also highlighted. One-Sentence Summary This review highlights the microbial consortia-driven waste valorization for biomanufacturing through top-down and bottom-up design approaches and describes strategies, tools, and unexplored opportunities to optimize the design and stability of such consortia.",
"conclusion": "Conclusions There have been unprecedented efforts to harness microbial consortia to develop several biotechnologies for biomanufacturing. This review specifically highlights the potential of microbial consortia and discusses the top-down and bottom-up engineering design approaches including the challenges and future recommendations. The top-down strategy still remains the most commonly used approach for waste valorization due to its relative ease of implementation. The bottom-up approach holds promise, but successful implementation will require developing techniques for stable consortia assembly, process optimization using ecological principles, and metabolic engineering to develop stable synthetic microbial consortia. There has been significant progress in engineering individual strains for biomanufacturing, however, less focus has been placed on consortium engineering. While models exist for single microbial metabolism, there is a need for models that can predict and identify the metabolic features governing interactions and long-term stability within microbial consortia. Besides focusing on the technical advancements, the economics should also be considered as more resources are needed for developing synthetic microbial consortia using the bottom-up approach compared to the traditional top-down approach. For instance, it might be economically beneficial to invest in developing and improving synthetic consortia for bioprocesses that produce high-value specialty chemicals. Future research should also explore integrating undefined natural microbial consortium (top-down) with defined synthetic consortium (bottom-up) for accomplishing efficient conversion of complex waste biomass. An interdisciplinary approach of combining bioprocess engineering, multi-omics analysis, metabolic engineering, system biology, and metabolic modeling is needed for a deeper understanding and optimization of microbial consortia-driven biomanufacturing for a sustainable future.",
"introduction": "Introduction A staggering 85% of the world's energy consumption is derived from nonrenewable fossil fuels (Cleveland & Morris, 2014 ; World Energy Use, 2022 ). This reliance has led to greenhouse gas (GHG) emissions and severely impacted natural ecosystem and biodiversity, necessitating sustainable alternatives to petrochemical-derived products. Biomanufacturing harnesses the power of microorganisms or enzymes to produce biofuels and bioproducts thus reducing our dependence on fossil fuels. Biomass-based products could replace up to 16% of crude oil consumption in the U.S., generating an additional $812 billion profit (Bioproducts to Enable Biofuels Workshop Summary Report, 2015 ). Biomass provided 5% of the total energy consumption in the U.S. in 2022, reducing the reliance on traditional energy sources (Biomass Explained, 2023 ). To achieve the ultimate goal of achieving a net-zero emission by 2050 and a carbon-neutral economy, biobased products will be further incentivized the U.S. It is estimated that 2.01 billion tons of municipal solid waste is generated annually with more than 33% of total waste managed in an environmentally unsafe manner (Kaza et al., 2018 ). According to the U.S. Environmental Protection Agency, 66.2 million tons of food waste was generated from food retail and an additional 40.1 million tons from food processing and manufacturing, causing significant economic loss and environmental issues in 2019 (2019 Wasted Food Report, 2023 ). Traditional waste management methods, such as composting, landfill disposal, and incineration have limitations such as low process efficiency and negative environmental impacts, including space requirements, generation of odor, contaminated leachate, and toxic pollutants, and ash emissions (Phua et al., 2019 ; Andraskar et al., 2021 ; Parvin & Tareq, 2021 ). Inexpensive and readily available wet waste (Tayou et al., 2022 ), lignocellulosic waste (Li et al., 2020 ; Shahab et al., 2020a ; Wongfaed et al., 2023 ), and C1 waste (CO 2 , CO) (Diender et al., 2016 ; Jiang et al., 2021 ), are promising renewable feedstocks for the production of value-added products to enable biomanufacturing. Implementing innovative technologies for converting such untapped waste to useful products can address negative impacts due to waste accumulation and management problems, substitute fossil fuel-based products, and strengthen the bioeconomy. Microbial communities greatly impact various aspects of life ranging from biogeochemical cycles, medicine, bioremediation, and public health, to biomanufacturing and resource recovery. Microbial consortia are suited to convert complex waste biomass due to their high enzyme diversity and the concerted and syntrophic activity of microorganisms belonging to different functional groups. Biotechnologies employing microbial consortia for sustainable waste valorization have emerged as promising alternatives to the petrochemical refinery processes to produce valuable biofuels, bioplastics, biochemicals, enzymes, and single-cell proteins (Venkateswar Reddy & Venkata Mohan, 2012 ; Zhou et al., 2017 ; Chi et al., 2018 ; Reddy et al., 2018 ; Valentino et al., 2018 ; Li et al., 2020 ; Pagliano et al., 2020 ; Tayou et al., 2022 ). Several of these biotechnologies have moved beyond lab and pilot scales and seen commercialization and expansion in recent years, contributing to the bioeconomy. For instance, anaerobic digestion (AD) has been widely adopted as a biological waste treatment and resource recovery technology to produce biogas which is further converted into electricity and heat. According to the World Biogas Association, an estimated 132 000 small, medium, and large-scale ADs are operational globally (World Biogas Association Global Report, 2019 ). The advancement in AD has been possible by understanding the role of microbial consortia and engineering it for efficient degradation of a wider range of feedstocks, tolerance to inhibitory compounds, and resilience to environmental perturbations (Werner et al., 2011 ; Blair et al., 2021 ). Microbial consortia-based biotechnologies harness the metabolic capacity of microorganisms and their synergistic interactions by employing either top-down or bottom-up approaches for microbiome engineering. The top-down approach involves providing selective pressure by manipulating environmental or operating conditions to steer the structure and activity of the natural microbial consortia toward a desired function. On the contrary, the bottom-up approach starts by understanding individual microbial characteristics to rationally assemble native or engineered microorganisms into a new synthetic consortium. However, such microbial consortia-based processes still suffer from undesirable side reactions, low process efficiency, and inability to control and maintain stability for a long term, which is partly due to exposure to external perturbations and unpredictable intercellular interactions. Previous reviews on microbial consortia have focused on either the top-down or bottom-up microbiome engineering approaches for environmental, public health, medical, or biotechnology applications (Duncker et al., 2021 ; Hu et al., 2022 ; Sauer & Marx, 2023 ; Zhou et al., 2024 ). However, there is a lack of comprehensive reviews that provide critical perspectives on both top-down and bottom-up approaches, and their integration to enable biomanufacturing from waste streams. This perspective paper first explores the promising potential of microbial consortia, highlighting their industrial and environmental applications utilizing diverse waste streams. Next, the paper elucidates recent advancements and knowledge gaps in top-down and bottom-up approaches with a focus on the design and assembly of synthetic microbial consortia, ecological engineering for process optimization, and metabolic engineering of microbial consortia. Finally, we also discuss the combination of the top-down and bottom-up approaches to maximize the potential of microbial consortia in different scenarios as well as metabolic modeling to predict and guide the microbial consortia design."
} | 2,569 |
39003244 | PMC11287213 | pmc | 389 | {
"abstract": "Abstract Growing environmental concerns and the need to adopt a circular economy have highlighted the importance of waste valorization for resource recovery. Microbial consortia-enabled biotechnologies have made significant developments in the biomanufacturing of valuable resources from waste biomass that serve as suitable alternatives to petrochemical-derived products. These microbial consortia-based processes are designed following a top-down or bottom-up engineering approach. The top-down approach is a classical method that uses environmental variables to selectively steer an existing microbial consortium to achieve a target function. While high-throughput sequencing has enabled microbial community characterization, the major challenge is to disentangle complex microbial interactions and manipulate the structure and function accordingly. The bottom-up approach uses prior knowledge of the metabolic pathway and possible interactions among consortium partners to design and engineer synthetic microbial consortia. This strategy offers some control over the composition and function of the consortium for targeted bioprocesses, but challenges remain in optimal assembly methods and long-term stability. In this review, we present the recent advancements, challenges, and opportunities for further improvement using top-down and bottom-up approaches for microbiome engineering. As the bottom-up approach is relatively a new concept for waste valorization, this review explores the assembly and design of synthetic microbial consortia, ecological engineering principles to optimize microbial consortia, and metabolic engineering approaches for efficient conversion. Integration of top-down and bottom-up approaches along with developments in metabolic modeling to predict and optimize consortia function are also highlighted. One-Sentence Summary This review highlights the microbial consortia-driven waste valorization for biomanufacturing through top-down and bottom-up design approaches and describes strategies, tools, and unexplored opportunities to optimize the design and stability of such consortia.",
"conclusion": "Conclusions There have been unprecedented efforts to harness microbial consortia to develop several biotechnologies for biomanufacturing. This review specifically highlights the potential of microbial consortia and discusses the top-down and bottom-up engineering design approaches including the challenges and future recommendations. The top-down strategy still remains the most commonly used approach for waste valorization due to its relative ease of implementation. The bottom-up approach holds promise, but successful implementation will require developing techniques for stable consortia assembly, process optimization using ecological principles, and metabolic engineering to develop stable synthetic microbial consortia. There has been significant progress in engineering individual strains for biomanufacturing, however, less focus has been placed on consortium engineering. While models exist for single microbial metabolism, there is a need for models that can predict and identify the metabolic features governing interactions and long-term stability within microbial consortia. Besides focusing on the technical advancements, the economics should also be considered as more resources are needed for developing synthetic microbial consortia using the bottom-up approach compared to the traditional top-down approach. For instance, it might be economically beneficial to invest in developing and improving synthetic consortia for bioprocesses that produce high-value specialty chemicals. Future research should also explore integrating undefined natural microbial consortium (top-down) with defined synthetic consortium (bottom-up) for accomplishing efficient conversion of complex waste biomass. An interdisciplinary approach of combining bioprocess engineering, multi-omics analysis, metabolic engineering, system biology, and metabolic modeling is needed for a deeper understanding and optimization of microbial consortia-driven biomanufacturing for a sustainable future.",
"introduction": "Introduction A staggering 85% of the world's energy consumption is derived from nonrenewable fossil fuels (Cleveland & Morris, 2014 ; World Energy Use, 2022 ). This reliance has led to greenhouse gas (GHG) emissions and severely impacted natural ecosystem and biodiversity, necessitating sustainable alternatives to petrochemical-derived products. Biomanufacturing harnesses the power of microorganisms or enzymes to produce biofuels and bioproducts thus reducing our dependence on fossil fuels. Biomass-based products could replace up to 16% of crude oil consumption in the U.S., generating an additional $812 billion profit (Bioproducts to Enable Biofuels Workshop Summary Report, 2015 ). Biomass provided 5% of the total energy consumption in the U.S. in 2022, reducing the reliance on traditional energy sources (Biomass Explained, 2023 ). To achieve the ultimate goal of achieving a net-zero emission by 2050 and a carbon-neutral economy, biobased products will be further incentivized the U.S. It is estimated that 2.01 billion tons of municipal solid waste is generated annually with more than 33% of total waste managed in an environmentally unsafe manner (Kaza et al., 2018 ). According to the U.S. Environmental Protection Agency, 66.2 million tons of food waste was generated from food retail and an additional 40.1 million tons from food processing and manufacturing, causing significant economic loss and environmental issues in 2019 (2019 Wasted Food Report, 2023 ). Traditional waste management methods, such as composting, landfill disposal, and incineration have limitations such as low process efficiency and negative environmental impacts, including space requirements, generation of odor, contaminated leachate, and toxic pollutants, and ash emissions (Phua et al., 2019 ; Andraskar et al., 2021 ; Parvin & Tareq, 2021 ). Inexpensive and readily available wet waste (Tayou et al., 2022 ), lignocellulosic waste (Li et al., 2020 ; Shahab et al., 2020a ; Wongfaed et al., 2023 ), and C1 waste (CO 2 , CO) (Diender et al., 2016 ; Jiang et al., 2021 ), are promising renewable feedstocks for the production of value-added products to enable biomanufacturing. Implementing innovative technologies for converting such untapped waste to useful products can address negative impacts due to waste accumulation and management problems, substitute fossil fuel-based products, and strengthen the bioeconomy. Microbial communities greatly impact various aspects of life ranging from biogeochemical cycles, medicine, bioremediation, and public health, to biomanufacturing and resource recovery. Microbial consortia are suited to convert complex waste biomass due to their high enzyme diversity and the concerted and syntrophic activity of microorganisms belonging to different functional groups. Biotechnologies employing microbial consortia for sustainable waste valorization have emerged as promising alternatives to the petrochemical refinery processes to produce valuable biofuels, bioplastics, biochemicals, enzymes, and single-cell proteins (Venkateswar Reddy & Venkata Mohan, 2012 ; Zhou et al., 2017 ; Chi et al., 2018 ; Reddy et al., 2018 ; Valentino et al., 2018 ; Li et al., 2020 ; Pagliano et al., 2020 ; Tayou et al., 2022 ). Several of these biotechnologies have moved beyond lab and pilot scales and seen commercialization and expansion in recent years, contributing to the bioeconomy. For instance, anaerobic digestion (AD) has been widely adopted as a biological waste treatment and resource recovery technology to produce biogas which is further converted into electricity and heat. According to the World Biogas Association, an estimated 132 000 small, medium, and large-scale ADs are operational globally (World Biogas Association Global Report, 2019 ). The advancement in AD has been possible by understanding the role of microbial consortia and engineering it for efficient degradation of a wider range of feedstocks, tolerance to inhibitory compounds, and resilience to environmental perturbations (Werner et al., 2011 ; Blair et al., 2021 ). Microbial consortia-based biotechnologies harness the metabolic capacity of microorganisms and their synergistic interactions by employing either top-down or bottom-up approaches for microbiome engineering. The top-down approach involves providing selective pressure by manipulating environmental or operating conditions to steer the structure and activity of the natural microbial consortia toward a desired function. On the contrary, the bottom-up approach starts by understanding individual microbial characteristics to rationally assemble native or engineered microorganisms into a new synthetic consortium. However, such microbial consortia-based processes still suffer from undesirable side reactions, low process efficiency, and inability to control and maintain stability for a long term, which is partly due to exposure to external perturbations and unpredictable intercellular interactions. Previous reviews on microbial consortia have focused on either the top-down or bottom-up microbiome engineering approaches for environmental, public health, medical, or biotechnology applications (Duncker et al., 2021 ; Hu et al., 2022 ; Sauer & Marx, 2023 ; Zhou et al., 2024 ). However, there is a lack of comprehensive reviews that provide critical perspectives on both top-down and bottom-up approaches, and their integration to enable biomanufacturing from waste streams. This perspective paper first explores the promising potential of microbial consortia, highlighting their industrial and environmental applications utilizing diverse waste streams. Next, the paper elucidates recent advancements and knowledge gaps in top-down and bottom-up approaches with a focus on the design and assembly of synthetic microbial consortia, ecological engineering for process optimization, and metabolic engineering of microbial consortia. Finally, we also discuss the combination of the top-down and bottom-up approaches to maximize the potential of microbial consortia in different scenarios as well as metabolic modeling to predict and guide the microbial consortia design."
} | 2,569 |
26234179 | PMC5014196 | pmc | 390 | {
"abstract": "Summary Bacterial macrocolony biofilms grow into intricate three‐dimensional structures that depend on self‐produced extracellular polymers conferring protection, cohesion and elasticity to the biofilm. In E \n scherichia coli , synthesis of this matrix – consisting of amyloid curli fibres and cellulose – requires CsgD , a transcription factor regulated by the stationary phase sigma factor RpoS , and occurs in the nutrient‐deprived cells of the upper layer of macrocolonies. Is this asymmetric matrix distribution functionally important or is it just a fortuitous by‐product of an unavoidable nutrient gradient? In order to address this question, the RpoS ‐dependent csgD promoter was replaced by a vegetative promoter. This re‐wiring of csgD led to CsgD and matrix production in both strata of macrocolonies, with the lower layer transforming into a rigid ‘base plate’ of growing yet curli‐connected cells. As a result, the two strata broke apart followed by desiccation and exfoliation of the top layer. By contrast, matrix‐free cells at the bottom of wild‐type macrocolonies maintain colony contact with the humid agar support by flexibly filling the space that opens up under buckling areas of the macrocolony. Precisely regulated stratification in matrix‐free and matrix‐producing cell layers is thus essential for the physical integrity and architecture of E \n . coli macrocolony biofilms.",
"introduction": "Introduction Bacterial biofilms are multicellular aggregates of cells surrounded by a self‐produced matrix that provides cohesion and protection. Matrix components can include cell appendages such as adhesive pili and flagella, amyloid fibres, secreted proteins, exopolysaccharides (EPSs) and extracellular DNA (eDNA) (Flemming and Wingender, 2010 ). Although often depicted as rather unstructured heaps of matrix‐surrounded cells, biofilms in fact show an intricate and highly regulated supracellular architecture (Parsek and Tolker‐Nielsen, 2008 ; Serra and Hengge, 2014 ). Macrocolony biofilms that grow on agar‐solidified complex media for several days – thus mimicking biofilms on decaying organic materials in nature – even fold and buckle up to produce an intricate macroscopic morphology with intertwined wrinkles, elongated folds and ridges and/or concentric ring patterns (Serra et al ., 2013a , 2013b ; Okegbe et al ., 2014 ). This structural organization of macrocolony biofilms depends on the ability to produce an extracellular matrix (Römling, 2005 ; Romero et al ., 2010 ; Serra and Hengge, 2014 ). Moreover, it is intimately associated with physiological stratification within biofilms, where zones of actively growing cells and highly stress‐resistant stationary phase cells are generated along gradients of nutrients and oxygen, which emerge as a function of consumption and diffusion (Rani et al ., 2007 ; Lenz et al ., 2008 ; Stewart and Franklin, 2008 ; Williamson et al ., 2012 ; Serra et al ., 2013a , 2013b ). Whether zones of active growth are found in the top or bottom layers of macrocolonies, exquisitely depends on the specific metabolism of a given species, in particular whether oxygen (in the absence of other respiratory electron acceptors) is absolutely required for growth or not (see discussion in Serra and Hengge, 2014 ). As an enteric bacterium, Escherichia coli can grow by fermentation when oxygen becomes limiting, which makes the nutrient supply the major determinant for entry into stationary phase. Therefore, the top layer of E. coli macrocolonies, which is most remote from the nutrient‐providing agar support, features small stationary phase cells that are tightly embedded in an extracellular matrix network, whereas the bottom layer contains elongated and dividing cells that produce flagella (Serra et al ., 2013b ). The matrix in the top layer consists of amyloid curli fibres and cellulose that together form a composite material that confers tissue‐like properties, i.e. strong cohesiveness and elasticity, which is a prerequisite for the very flat and large macrocolonies to buckle up into a pattern of ridges and wrinkles (Serra et al ., 2013a ). A similar stratification of matrix production has also been observed for a uropathogenic E. coli grown in pellicles on static liquid (Hung et al ., 2013 ). Notably, the classical laboratory E. coli K‐12 strains are cellulose‐negative (due to a nonsense mutation in bcsQ in the cellulose biosynthesis operon) and therefore produce a curli‐only matrix in the top layer, which is brittle and during growth of the much thicker macrocolonies breaks into a pattern of concentric rings. However, a ‘de‐domesticated’ derivative of the E. coli K‐12 strain W3110, with bcsQ ‘repaired’ in the chromosome, has been generated and is used in the present study (strain AR3110) (Serra et al ., 2013a ). Activation of curli and cellulose production depends on a hierarchical control network with RpoS (σ S ), the stationary phase sigma subunit of RNA polymerase (RNAP) acting as a master regulator. In addition, several diguanylate cyclases (DGCs) and phosphodiesterases (PDEs), which control the second messenger c‐di‐GMP, provide for decisive input into this network (Pesavento et al ., 2008 ; Lindenberg et al ., 2013 ). Moreover, various small regulatory RNAs play a modulatory role (summarized in Mika and Hengge, 2014 ). The output of this multiple signal‐integrating network is the control of expression of CsgD, a transcription factor, which directly drives the expression of the curli operon csgBAC and indirectly activates cellulose biosynthesis by driving the expression of YaiC, a DGC which is essential to activate cellulose synthase (summarized in Hengge, 2009 ; 2010 ). The overall result is a strong accumulation of CsgD and matrix components in the upper macrocolony layer, with a transition zone of slow growth between bottom and top layers, in which CsgD and matrix are produced heterogeneously (Serra and Hengge, 2014 ). This asymmetric distribution of matrix components in macrocolony biofilms raises some questions. Is matrix production in the upper layer only just a fortuitous by‐product of RpoS dependency and the nutrient gradient that unavoidably builds up in these biofilms? But why then is the obviously energy‐intensive production of large amounts of extracellular protein and polysaccharide confined to the energy‐limited cells in the upper layer, with an underlying regulation that could hardly be more complex? This suggests that this kind of matrix stratification is functionally important – but for what? Here, we demonstrate that genetically re‐wiring CsgD expression in a way that results in matrix production in both strata has striking consequences for macrocolony integrity, supracellular architecture and macroscopic morphology. These point to an essential spatial division of labour between matrix‐producing and matrix‐free cells in the building of an environmentally robust biofilm.",
"discussion": "Discussion Macrocolony biofilms of E. coli exhibit a clear physiological stratification with vegetatively growing cells in the bottom layer and the outer edges of the expanding colony and slowly growing and finally stationary phase cells in the upper layer where an extracellular scaffold of curli fibres and cellulose is produced that is arranged in a highly distinct architecture (Serra and Hengge, 2014 ). This matrix confers strong cohesiveness and high elasticity, i.e. tissue‐like properties to the macrocolonies which therefore fold and buckle up vertically as a result of tension building up by cellular proliferation and crowding (Serra et al ., 2013a ). The result can be rather spectacular macrocolony morphologies – in fact a trait that is ubiquitous among microbes, with even yeasts growing into similar colony patterns (see Okegbe et al ., 2014 for a multispecies collection of macrocolony images). In E. coli , the asymmetric matrix production in the upper macrocolony layer only is a consequence of (i) the nutrient gradient building up in this biofilm which leads to physiological stratification and (ii) CsgD, the master regulator of matrix biosynthesis, being under control of the stationary phase sigma factor RpoS (Serra et al ., 2013b ). Re‐wiring CsgD expression in a manner which resulted in matrix production in the upper and lower strata of the macrocolonies allowed us to address the question whether the vertical stratification of matrix production is physiologically important. Thus, artificially engineered matrix production in both layers resulted in perturbation of the supracellular architecture and macroscopic morphology and, in the longer run, in detachment, desiccation, breakage and exfoliation of the upper macrocolony layer. Thus, matrix production in the upper layer only is essential for physical integrity of E. coli macrocolony biofilms. Moreover, these phenotypes also directly indicate why this matrix stratification is essential: trapping the growing cells of the lower stratum in a scaffold of dense matrix – in fact a rigid base plate‐like structure – no longer allows these cells to flexibly invade the space that opens up when the upper layer buckles into ridges and wrinkles or – if curli only is produced – into dome‐like ring patterns. As a consequence, the upper layer detaches and can no longer maintain its water content via capillary action, therefore drying out and exfoliating like paper (after all, paper is the technical equivalent of desiccated cellulose). Overall, the necessity for matrix production in the upper layer only is thus a consequence of the physical properties conferred by the hydrated curli‐cellulose composite matrix, which make the macrocolony react like a soft elastic sheet under mechanical tension (Cerda and Mahadevan, 2003 ). Similar properties have recently also been described for pellicle biofilms (Trejo et al ., 2013 ). However, what is the benefit of such tissue‐like behaviour, especially since it entails the difficult task of energy‐intensive production of large amounts of extracellular polymers in the energy‐deprived cells of the upper macrocolony layer? An extensive matrix that essentially encloses cells at the surface of the macrocolony biofilm (Serra et al ., 2013a ) clearly protects against abiotic or biotic stresses, such as toxic chemicals or predation (Branda et al ., 2005 ; Flemming and Wingender, 2010 ; DePas et al ., 2014 ). Moreover, cellulose, which is the matrix component that can bind large amounts of water and confers elasticity to the macrocolony (Zogaj et al ., 2001 ), is arranged in vertical filaments, sheets and sheaths that give rise to the vertical ‘pillars’ of cells in the lower region of the matrix‐producing layer (Figs 6 and 7 ) (Serra et al ., 2013a ). These long‐range connections seem to connect the lower and upper colony layers, which forces growing cells to spread out laterally and thus produces very flat macrocolonies of a very even height of approximately 60–65 μm (by contrast, the ring‐forming macrocolonies of curli‐only producing E. coli grow 200 μm and higher; compare Figs 3 and 6 ). This pronounced flatness as well as the buckling up of the flat elastic macrocolonies increases the surface‐to‐volume ratio of the biofilm and thus clearly optimizes access to oxygen (Dietrich et al ., 2013 ; Serra et al ., 2013a ; Kempes et al ., 2014 ). These advantages are obviously significant enough to outweigh the price of having to deal with the topological consequences, i.e. the necessity of a complex regulation that ensures (i) the absence of matrix in the lower layer of cells that have to be flexible enough to spread into openings below the folds and thus maintain contact and capillary action between the vertically folding layer of matrix‐immobilized cells and the humid support and (ii) the (still unknown) logistics of providing the already carbon‐ and energy‐limited cells in the upper layer with sufficient resources to produce the massive amount of extracellular matrix. This interplay between metabolic and matrix stratification and biophysical properties that shape E. coli macrocolonies invites a comparison with other bacterial species. While wrinkles and ridges are rather ubiquitous features of macrocolony morphology, metabolic requirements – in particular the dependence on oxygen for growth – are very different and represent an adaptation to the primary niches of different species. This has consequences for the physiological stratification of macrocolony biofilms, which in turn should affect the regulation of extracellular matrix production. Thus, Pseudomonas aeruginosa – another well‐studied Gram‐negative biofilm model bacterium – depends on oxygen for metabolic activity and growth (if not provided with an electron acceptor for anaerobic respiration). As a consequence, metabolically active cells, which are in a state of post‐exponential growth, populate the upper layer of a macrocolony, where they find sufficient oxygen as well as nutrients, which can diffuse upwards through the lower layer of metabolically inert cells (Lenz et al ., 2008 ; Williamson et al ., 2012 ; Zhang et al ., 2013 ). What does this mean for the regulation of synthesis of the biofilm matrix, which in the case of P. aeruginosa contains the EPSs Psl and Pel as well as eDNA? Unfortunately, the regulation of these genes is not fully understood and, to our knowledge, the localization of their expression or of the matrix itself within macrocolony biofilms has not been studied. However, in order to benefit from protection by a matrix and to generate the intricate macrocolony morphology with wrinkles and high ridges (Dietrich et al ., 2013 ; Kempes et al ., 2014 ; Okegbe et al ., 2014 ), which is very similar to that of E. coli , one has to assume that matrix is produced by the post‐exponentially growing cells in the upper layer of P. aeruginosa macrocolonies. This would not only facilitate the energy‐intensive matrix production (in comparison with E. coli ), but is consistent with the expression of rpoS and the quorum sensing regulator gene rhlR in this layer (Pérez‐Osorio et al ., 2010 ), since psl genes are activated by RpoS (Irie et al ., 2010 ) and three interconnected quorum sensing systems (Las, Rhl, Pqs) play multiple regulatory roles in the production of Psl, Pel and eDNA (Davies et al ., 1998 ; Davey et al ., 2003 ; Allesen‐Holm et al ., 2006 ; Gupta and Schuster, 2012 ). The inverse physiological stratification of macrocolonies of P. aeruginosa and E. coli – with the latter not even possessing acyl‐homoserine lactone‐based quorum sensing – could thus provide an explanation for the striking differences in biofilm regulation in these two gammaproteobacteria. In addition, the question arises, how much matrix is present in the lower layer and how P. aeruginosa macrocolonies remain in tight contact with the moist agar surface in areas where they buckle up? Future studies should therefore investigate amounts and spatial distribution of the various matrix components in P. aeruginosa macrocolonies. \n Bacillus subtilis , a model organism for biofilm formation by Gram‐positive bacteria, demonstrates an intriguingly different solution to the problem of buckling macrocolony desiccation. Bacillus subtilis macrocolonies also form complex patterns with wrinkles and elongated elevated structures that depend on the amyloid fibre protein TasA, an EPS and the BslA protein as matrix components (Aguilar et al ., 2007 ; Romero et al ., 2010 ; Kobayashi and Iwano, 2012 ; Hobley et al ., 2013 ; Vlamakis et al ., 2013 ). Whereas flagella genes are expressed at the outer edges, matrix genes are active in principle everywhere else in macrocolonies, with sporulation genes finally turning on in cells close to the air‐exposed surface, i.e. farthest away from the nutrient‐providing agar (Vlamakis et al ., 2008 ). In these macrocolonies, which apparently do not show matrix stratification, localized cell death precedes, facilitates and possibly determines the site of buckling (Asally et al ., 2012 ). In contrast to E. coli or P. aeruginosa macrocolonies, which fold into ridges with a thin bilayer structure that are pushed higher and higher (Serra et al ., 2013a ), the entire B. subtilis macrocolony (which is about 200 μm thick) buckles up into broader openings that can close at the bottom (either by growth or by the flat parts of the growing biofilm pushing together). This results in the formation of tube‐like channels, which are filled with liquid due to water flow that arises from evaporation from the colony surface (Wilking et al ., 2013 ). While it has been pointed out that this network of horizontal channels facilitates nutrient transport (Wilking et al ., 2013 ), it is obviously also a system that keeps the macrocolony hydrated. Moreover, with the channels closing at the bottom, the lower part of the macrocolony is flat again and remains in tight contact with the moist agar surface. In conclusion, the extracellular matrix which tightly surrounds the cells in bacterial biofilms is not only protective and cohesive but it confers elastic tissue behaviour to macrocolony biofilms. As a consequence, cellular proliferation and crowding require flat macrocolonies to buckle up into wrinkles and ridges. As pointed out in our study this behaviour comes at the risk of losing contact to the humid support followed by desiccation and potential exfoliation of large parts of the air‐exposed biofilm. Different species of bacteria have found different solutions to this problem that are intimately connected to their specific metabolic properties, which dictates physiological stratification of these biofilms, which in turn impacts on regulatory mechanisms of matrix biosynthesis. Insight into these mechanisms also has implications for any attempts by synthetic biology to produce and ‘shape’ bacterial communities with specific properties."
} | 4,513 |
34987488 | PMC8721230 | pmc | 393 | {
"abstract": "Understanding how microorganism-microorganism interactions shape microbial assemblages is a key to deciphering the evolution of dependencies and co-existence in complex microbiomes. Metabolic dependencies in cross-feeding exist in microbial communities and can at least partially determine microbial community composition. To parry the complexity and experimental limitations caused by the large number of possible interactions, new concepts from systems biology aim to decipher how the components of a system interact with each other. The idea that cross-feeding does impact microbiome assemblages has developed both theoretically and empirically, following a systems biology framework applied to microbial communities, formalized as microbial systems ecology (MSE) and relying on integrated-omics data. This framework merges cellular and community scales and offers new avenues to untangle microbial coexistence primarily by metabolic modeling, one of the main approaches used for mechanistic studies. In this mini-review, we first give a concise explanation of microbial cross-feeding. We then discuss how MSE can enable progress in microbial research. Finally, we provide an overview of a MSE framework mostly based on genome-scale metabolic-network reconstruction that combines top-down and bottom-up approaches to assess the molecular mechanisms of deterministic processes of microbial community assembly that is particularly suitable for use in synthetic biology and microbiome engineering.",
"conclusion": "Conclusion Deciphering ecological processes taking place within a microbial community is the only way to obtain a mechanistic view of its functioning. Ecological interactions, particularly cross-feeding, must thus be taken into account in any microbial ecology project, notably in synthetic biology and microbiome engineering, with many applications including human health and sustainable agriculture ( Toju et al., 2018 ; Henriques et al., 2020 ). With this goal in view, MSE frameworks are being developed to unify top-down and bottom-up approaches in an iterative design-build-test-learn cycle ( Lawson et al., 2019 ). Still, MSE should be used cautiously to avoid being drowned under hundreds of irrelevant models. Whenever possible, predictions of an MSE framework should be tested experimentally ( Röling and Van Bodegom, 2014 ; Muller et al., 2018 ; Vázquez-Castellanos et al., 2019 ), and in return, experimental observations should improve models. To build reliable and in-depth knowledge, efforts should focus on a few aspects, such as GEM quality (in order to go beyond research on conserved, well-known metabolic pathways), the integration of -omics data ( Franzosa et al., 2015 ), notably the microbial secretome with exometabolomics, and cross-talk with other approaches such as niche modeling or dynamics modeling ( Jacoby and Kopriva, 2019 ).",
"introduction": "Introduction Deciphering the assembly rules of microbial communities is vital for a mechanistic understanding of the general principles driving microbiome activity and functions ( Vellend et al., 2014 ; Morrison-Whittle and Goddard, 2015 ). Microbial communities are governed by both stochastic and deterministic factors ( Vellend, 2010 ; Stegen et al., 2012 ), and recent advances show that deterministic processes largely contribute to shaping microbial community assembly. Their relative contribution varies however according to the ecology of microorganisms (e.g., specialists or generalists) and the stability of the environment ( Figure 1E , Stegen et al., 2012 ; Ning et al., 2020 ; Xu et al., 2020 ). Ecological interactions including commensalism, competition, and mutualism contribute to the self-organizational properties of microbiomes ( Stegen et al., 2013 ). However, how these different interactions act in concert to shape microbial assemblages remain poorly understood ( Nemergut et al., 2013 ). Microbial communities are likely not only driven by antagonistic interactions but also by cooperative symbioses, defined in 1879 by De Bary (2019) as the “living together of unlike organisms.” Symbioses (thus cooperation) are now recognized as central drivers of (co-)evolution, and are often associated with obligate mutualism but are actually a continuum of interactions between mutualism and parasitism ( Ewald, 1987 ; Drew et al., 2021 ), implying dependency of one organism on another ( Figure 1A ; Raina et al., 2018 ). Among these interactions, metabolic dependencies by cross-feeding likely explain patterns in microbial communities ( Mas et al., 2016 ; Zomorrodi and Segrè, 2017 ; Amor and Bello, 2019 ; Coyte and Rakoff-Nahoum, 2019 ; Pacheco and Segrè, 2019 ; Seif et al., 2020 ; Zhu et al., 2020 ). In community ecology, competition and related competitive exclusion were previously considered to be the main drivers of community assembly. The competitive exclusion principle (also often referred to as Gause’s law) states that two species with the same ecological niche cannot coexist because of competition, which leads either to the extinction of species or to the differentiation of their ecological niche ( Gause, 1960 ; Hardin, 1960 ; Pocheville, 2015 ). This role of competition was questioned by the observation of unexpectedly complex microbial communities according to general ecology theories ( Pacheco and Segrè, 2019 ). Hence, cross-feeding is increasingly believed to play an important role in the complexity of microbial communities ( Zengler and Zaramela, 2018 ). In this mini-review, we summarize the definitions of cross-feeding and its underlying mechanisms, as well as its importance in structuring microbial communities. Then, we describe microbial systems ecology (MSE), a discipline at the crossroads of systems biology and microbial community ecology aiming to explain coexistence. Figure 1 Cross-feeding among co-occurring microorganisms and its integration to microbial systems ecology (MSE). (A) Symbiosis is the interaction between living entities along a gradient from mutualism to parasitism, depending on the effect of the receiver (also referred to as “beneficiary,” blue bacteria symbols) on the fitness of the provider (green bacteria symbols). (B) There are several subcategories of cross-feeding ( Smith et al., 2019 ). The type of secreted compounds, i.e., wastes (garbage icons) or other metabolites (red triangles) and on the directionality of the exchange (mutual or not, blue triangles) particularly matter in the classification of cross-feeding (see glossary for associated definitions). Enzymes (orange circle) can also be secreted to degrade complex molecules, making them available both for the producer and the receiver(s). (C) The existence of cross-feeding depends on the secretion, transport, and assimilation capacity of the public good ( D’Souza et al., 2018 ). (D) Metabolic interactions are environment-dependent, notably regarding available nutrients. If a required nutrient (red triangle) is freely available in the growth medium, then cross-feeding is not indispensable for the receiver organism. Otherwise, when a particular nutrient is not available, but is synthesized by the producer from another substrate (brown square), cross-feeding becomes obligatory for the receiver. (E) Graphical abstract summarizing the study of ecological interactions, notably cross-feeding, in microbial communities with MSE."
} | 1,843 |
28054770 | null | s2 | 394 | {
"abstract": "It is known that smooth, hydrophobic solid surfaces exhibit low ice adhesion values, which have been shown to approach a lower ice adhesion strength limit (∼150 kPa) defined by the water receding contact angle. To overcome this limit, we have designed self-lubricating icephobic coatings by blending polydimethylsiloxane (PDMS)-poly(ethylene glycol) (PEG) amphiphilic copolymers into a polymer matrix. Such coatings provide low ice adhesion strength values (∼50 kPa) that can substantially reduce the lower bound of the ice adhesion strength achieved previously on smooth, hydrophobic solid surfaces. Different molecular mechanisms are responsible for the low ice adhesion strength attained by these two approaches. For the smooth hydrophobic surfaces, an increased water depletion layer thickness at the interface weakens the van der Waals' interactions between the ice and the polymeric substrate. For the self-lubricating icephobic coatings, the PEG component of the amphiphilic copolymer is capable of strongly hydrogen bonding with water molecules. The surface hydrogen-bonded water molecules do not freeze, even at substantial levels of subcooling, and therefore serve as a self-lubricating interfacial liquid-like layer that helps to reduce the adhesion strength of ice to the surface. The existence of nonfrozen water molecules at the ice-solid interface is confirmed by solid-state nuclear magnetic resonance (NMR) spectroscopy."
} | 359 |
24011134 | null | s2 | 395 | {
"abstract": "Quorum sensing (QS) enables bacteria to sense and respond to changes in their population density. It plays a critical role in controlling different biological functions, including bioluminescence and bacterial virulence. It has also been widely adapted to program robust dynamics in one or multiple cellular populations. While QS systems across bacteria all appear to function similarly-as density-dependent control systems-there is tremendous diversity among these systems in terms of signaling components and network architectures. This diversity hampers efforts to quantify the general control properties of QS. For a specific QS module, it remains unclear how to most effectively characterize its regulatory properties in a manner that allows quantitative predictions of the activation dynamics of the target gene. Using simple kinetic models, here we show that the dominant temporal dynamics of QS-controlled target activation can be captured by a generic metric, 'sensing potential', defined at a single time point. We validate these predictions using synthetic QS circuits in Escherichia coli. Our work provides a computational framework and experimental methodology to characterize diverse natural QS systems and provides a concise yet quantitative criterion for selecting or optimizing a QS system for synthetic biology applications."
} | 335 |
23563810 | PMC3619133 | pmc | 396 | {
"abstract": "Compact and power-efficient plastic electronic synapses are of fundamental importance to overcoming the bottlenecks of developing a neuromorphic chip. Memristor is a strong contender among the various electronic synapses in existence today. However, the speeds of synaptic events are relatively slow in most attempts at emulating synapses due to the material-related mechanism. Here we revealed the intrinsic memristance of stoichiometric crystalline Ge 2 Sb 2 Te 5 that originates from the charge trapping and releasing by the defects. The device resistance states, representing synaptic weights, were precisely modulated by 30 ns potentiating/depressing electrical pulses. We demonstrated four spike-timing-dependent plasticity (STDP) forms by applying programmed pre- and postsynaptic spiking pulse pairs in different time windows ranging from 50 ms down to 500 ns, the latter of which is 10 5 times faster than the speed of STDP in human brain. This study provides new opportunities for building ultrafast neuromorphic computing systems and surpassing Von Neumann architecture.",
"discussion": "Discussion Compared with other emerging electronic synapses 5 6 11 13 14 16 27 , the analog chalcogenide synapse displayed the advantages of ultralow operation voltage, ultrafast synaptic events and feasibility of time window tuning ( Fig. 4d ). While chalcogenide devices have already demonstrated a drastic reduction in power and time requirements for operation 51 52 , the mature CMOS-compatible fabrication processes of chalcogenide materials enable us to intergrate the electronic synaptic matrix with traditional CMOS neuron chips with the goal of performing more complex cognitive functions in one chip, such as visual pattern cognition and associative learning 28 . And based on the network arthictecture (crossbar and 3D stacking), the important superiority of neural system, parrallelism, could be further investigated. In addition, the energy accumulation phenomenon associated with chalcogenide's amorphous state has been proposed to emulate neuronal threshold spiking behavior, which was verified in our memristor device ( Supplementary Fig. S8 and Fig. S9 ). We envisage a complete neuron-synapse system being implemented in a chalcogenide-based matrix, which could be an attractive option for the design of future neuromorphic system. This study demonstrated that ultrafast synaptic events can be implemented in chalcogenide synapses. We revealed that the intrinsic memristive characteristics of crystalline GST originated from the charge trapping and releasing by defects in the material. The memristance offered us a foundation for emulating synaptic weight modulation, and we used it to implement four different STDP forms. Furthermore, the feasibility of STDP time window tuning in this material was also demonstrated from 50 ms to 500 ns, the latter of which is 10 5 times faster than the speed of STDP in human brain. In principle, the scaling up of the number of analog/plastic synapses with low power consumption and ultrafast response times is in urgent demand for further evolution of neuromorphic chips. We believe that the completely passive, non-volatile chalcogenide synapse is a strong contender for replacing the existing transistor- and capacitor-based electronic synapse and provides a promising method of surpassing Von Neumann architecture."
} | 839 |
26839589 | PMC4736482 | pmc | 397 | {
"abstract": "Background Biogas production is an economically attractive technology that has gained momentum worldwide over the past years. Biogas is produced by a biologically mediated process, widely known as “anaerobic digestion.” This process is performed by a specialized and complex microbial community, in which different members have distinct roles in the establishment of a collective organization. Deciphering the complex microbial community engaged in this process is interesting both for unraveling the network of bacterial interactions and for applicability potential to the derived knowledge. Results In this study, we dissect the bioma involved in anaerobic digestion by means of high throughput Illumina sequencing (~51 gigabases of sequence data), disclosing nearly one million genes and extracting 106 microbial genomes by a novel strategy combining two binning processes. Microbial phylogeny and putative taxonomy performed using >400 proteins revealed that the biogas community is a trove of new species. A new approach based on functional properties as per network representation was developed to assign roles to the microbial species. The organization of the anaerobic digestion microbiome is resembled by a funnel concept, in which the microbial consortium presents a progressive functional specialization while reaching the final step of the process (i.e., methanogenesis). Key microbial genomes encoding enzymes involved in specific metabolic pathways, such as carbohydrates utilization, fatty acids degradation, amino acids fermentation, and syntrophic acetate oxidation, were identified. Additionally, the analysis identified a new uncultured archaeon that was putatively related to Methanomassiliicoccales but surprisingly having a methylotrophic methanogenic pathway. Conclusion This study is a pioneer research on the phylogenetic and functional characterization of the microbial community populating biogas reactors. By applying for the first time high-throughput sequencing and a novel binning strategy, the identified genes were anchored to single genomes providing a clear understanding of their metabolic pathways and highlighting their involvement in anaerobic digestion. The overall research established a reference catalog of biogas microbial genomes that will greatly simplify future genomic studies. Electronic supplementary material The online version of this article (doi:10.1186/s13068-016-0441-1) contains supplementary material, which is available to authorized users.",
"conclusion": "Conclusions This study demonstrated that the metagenomic assembly and binning of the shotgun sequences obtained from biogas reactors allowed the identification of 106 GBs that can be assigned into the context of the biogas degradation food chain by means of bioinformatic analysis. This is a major step forward in the characterization of the biogas microbial community especially when compared to previous studies, where the functional roles have been inferred from those assigned to the more similar species identified considering 16S rRNA similarity. In the case of the biogas microbial community, the identified GBs are distantly related to species for which the genomes are available in the databases and, as previously discussed, a predictive metagenomics approach is not accurate. This is clearly demonstrated by the high fraction of new GBs identified and assigned only at high taxonomic level, as for example the newly identified methanogenic archaea (Eu03). Another concluding remark drawn by the binning process revealed that approximately 70 % of the assembly cannot be assigned to a specific GB. This suggests the presence of more than 450 GBs in the biogas microbial community. As this is the first attempt to deeply characterize the AD microbiome, it is expected that further studies performed under different operational conditions (e.g., different temperatures and substrate) will allow in the next future to enrich the genome database. Finally, this study opens new avenues in deciphering the functional interactions between microbial species involved in the AD process and provides a solid reference that will greatly simplify further metatranscriptomics and metaproteomics analyses.",
"discussion": "Results and discussion Approximately, 340 millions high-quality paired-end reads (~51 gigabases of metagenomic sequence) were obtained from high throughput sequencing of 15 samples collected from 8 anaerobic digesters, representing conventional biogas reactors. The assembly of the reads resulted in 409,831 scaffolds (~686 Mbp) ranging in size from 500 to 313,754 bp (N50 2338). The percentage of reads aligned to the assembly varied from 57 to 73 % (with a mean average of 67 %) as shown in Additional file 1 : Table S1. No differences were found between the samples included in the assembly and those used only for the binning process, suggesting that the assembly was fairly representative for all the reactors. It should be mentioned that ~242 Mbp are in scaffolds larger than 5 kbp (“ Methods ” and Additional file 1 ). After the assembly process, the gene finding and annotation are more reliable and led to the identification of nearly one million protein encoding genes, 23.6 % of which could be assigned to GBs (Additional file 2 ). The protein encoding genes were annotated using COG [ 22 ], KEGG [ 23 ], and Pfam [ 24 ] (Additional file 2 ). The results showed that 569,645 genes (60.8 %) had a match in the COG database, 418,103 (44.6 %) in KEGG and 579,337 (61.8 %) had a protein domain annotated in Pfam. Finally, 277,604 genes (29.6 %) were completely unknown. The number of predicted proteins is approximately 70 times and 3.7 times more than those obtained in the two best previous assemblies of a biogas microbial community reported in the literature [ 6 , 8 ]. The number of genes belonging to each KEGG category in the assembly was compared with the scaffold coverage, which is directly related to species abundance (Additional file 1 and Additional file 1 : Figure S2). This means that the categories with the higher ratio between “coverage” and “number of genes in the category” are those associated with most abundant GBs. This analysis allowed an evaluation of the relevance of the KEGG classes considering both the number of genes in the pathway and the abundance of the species in the microbial community. From these data, it was evident that some metabolic pathways included genes with a high average coverage because they were encoded (also) in the genomes of the more abundant species of the microbial community as shown in Additional file 1 : Figure S2. The metabolic pathway of methanogenesis is the most straightforward example indicating that some methanogenic archaea (i.e., Eu01) dominate the microbial community in terms of abundance. We can assume that, for the same reason, the riboflavin KEGG pathway, which led to the biosynthesis of the proteins’ cofactor F430 involved in methanogenesis [ 25 ], is one of the highly ranked in the list. On the contrary, the KEGG pathway modules related to the degradation of xenobiotic compounds like “styrene,” “naphthalene,” “fluorobenzoate,” and “aromatic compounds” were mainly encoded in low abundant species and frequently belong to scaffolds that could not be assigned through the binning process. For example, only 22 % of the genes involved in “xylene degradation” were binned vs. 36 % of the “RNA-transport” and 33 % of the “riboflavin metabolism” (Additional file 2 ). This suggests that the degradation of xenobiotic compounds is specific to the rare biosphere in the biogas reactors. The only notable exception is “nitrotoluene degradation” but this is expected as the degradation of this compound and incorporation into the bacterial biomass in anaerobic conditions has been previously demonstrated [ 26 – 28 ]. Carbohydrate phosphotransferase (PTS) system, despite being represented by 1261 genes, is the second least abundant category considering the ratio “coverage/number of genes.” This suggests that mainly low abundant community members utilize this system to transport sugars. This result is totally unexpected, as PTS is widely spread among bacteria [ 29 ]. However, our data evidence that, in the AD microbiome, ABC transporters are more frequent in the high abundant species. Moreover, it was found that “nitrogen metabolism” includes genes mostly represented in low abundance species. This could be due to the average low nitrogen concentration contained in cattle manure (in comparison for example to pig or poultry manure) [ 30 ]. It can be expected that, due to the high dynamicity of the biogas community [ 3 , 14 , 31 ], modification of the manure composition (for example a higher quantity of ammonia) can lead to an increase in the abundance of some species that in our experiment are associated to the rare biosphere. Binning process and taxonomic classification Mapping reads from each sampling point to the assembled scaffolds indicated that the microbial species were differentially represented due to heterogeneous manure feedstock composition. The differences in the microbial abundance allowed the clustering of the scaffolds and resulted in the extraction of 106 GBs from the total assembly. A detailed explanation of the binning assembly procedure is reported in Additional file 1 together with a schematic representation of the binning strategy in Additional file 1 : Figure S3. In the first part of our procedure, high-quality GBs were manually extracted using the procedure of Albertsen et al. [ 16 ]. These GBs served as internal controls and were used to drive the second part of the binning. By an automatic extraction process based on clustering of scaffolds having similar coverage profiles, 61 additional GBs were identified. The estimated completeness of the GBs, based on the presence of 107 conserved marker genes [ 32 ], ranged from 15 % to more than 99 % (with a mean of 83 %) (Additional file 3 ). In order to validate this finding, an additional analysis was performed using CheckM [ 33 ], which gave as output very similar values (85 % completeness on average). The level of genome contamination was estimated both considering the number of duplicated essential genes and also with CheckM; the contamination was found to be extremely low and ranging on average between 3 and 5 % depending on the method used (Additional file 3 ). With our procedure, we have successfully identified 60 genomes with estimated completeness higher than 90 % considering the 107 essential genes, or 51 genomes according to CheckM. The result obtained is of a very high quality if compared with previous studies obtained from single-cell genome sequencing, where genome completeness averages around 40 % [ 34 ]. An additional analysis was performed using MetaBAT software [ 21 ] in order to evaluate the performance of our binning strategy. Considering as thresholds a) genome completeness higher than 90 % and b) contamination level lower than 20 %, MetaBAT managed to extract 42 GBs, while our binning strategy led to the identification of 51 GBs. Even by lowering the completeness threshold (e.g., to 70 %), our binning strategy was able to extract more GBs. The outcome of this comparison validated the high accuracy and efficiency of the binning strategy presented in the current manuscript. Taxonomic assignment showed that none of the GBs could be assigned to species level, only 10 GBs were assigned to the genus level, while the vast majority were assigned to phylum level (Table 1 ; Additional file 3 ). This confirms that most of the species in the biogas microbial community were not previously characterized at a genomic level [ 35 ]. The more affordable taxonomic assignments were obtained for Euryarchaeota Eu01, Eu02, Eu04, Eu05 suggesting that archaea are better characterized than bacteria in the biogas community (with the exception of Eu03). On the contrary, bacteria are completely unknown at genomic level. The results revealed that the biogas microbial community is dominated by the phylum Firmicutes (69 GBs) followed by the phyla Proteobacteria (10 GBs) and Bacteroidetes (6 GBs), which is in accordance with other studies [ 36 – 40 ] (Fig. 1 ). Sixty-nine of the GBs belong to Firmicutes (Fig. 5 ; Additional file 3 ). The species included in this division are extremely relevant from a functional point of view since they are involved in many metabolic processes including the degradation of carbohydrates, fatty acids utilization, Wood–Ljungdahl pathway (WLP) (homoacetogenesis) or syntrophic acetate oxidation (SAO). The comprehensive high-resolution microbial tree (Fig. 1 ) evidenced that these GBs can be subdivided into six main sub-groups (Additional file 3 ). Three GBs belong to Eubacteriaceae , 17 to the family Clostridiaceae , seven to the family Syntrophomonadaceae and 22 can be assigned only to the class Clostridia and are distantly related to the other Firmicutes . It is worth mentioning that the GBs assigned to the class Clostridia are the most cryptic inside the community, as they are distantly related to other Firmicutes , showing deeply branched GBs (Fig. 1 ). Moreover, four of the GBs initially assigned to the Firmicutes using the 107 essential genes were then re-assigned to the family Acholeplasmataceae , of the phylum Tenericutes (Te01, Te02, Te03, Te04) using Phylophlan. Table 1 Taxonomic assignment and basic genome characteristics of the 106 GBs extracted from biogas reactors Genome bin ID Genome bin “species name” GB size (Mbp) Estimated completeness (%) Genome bin ID Genome bin “species name” GB size (Mbp) Estimated completeness (%) Pr02 \n Gammaproteobacteria sp. DTU038 \n 4.2 84 Fi16 \n Clostridia sp. DTU025 \n 2.0 95 Fi48 \n Clostridiaceae sp. DTU079 \n 3.1 99 Fi13 \n Clostridia sp. DTU022 \n 2.0 89 Fi49 \n Clostridia sp. DTU080 \n 3.1 86 Fi32 \n Clostridiales sp. DTU060 \n 2.0 88 Pr05 \n Alcaligenaceae sp. DTU041 \n 2.9 96 Fi21 \n Halothermothrix sp. DTU029 \n 2.0 94 Fi40 \n Clostridiales sp. DTU070 \n 2.9 97 Ac01 \n Actinomycetales sp. DTU046 \n 1.9 67 Fi30 \n Clostridiales sp. DTU058 \n 2.9 99 Ba02 \n Rikenellaceae sp. DTU002 \n 1.9 88 Pr01 \n Gammaproteobacteria sp. DTU037 \n 2.8 96 Ba01 \n Rikenellaceae sp. DTU001 \n 1.9 95 Eu04 \n Methanosarcina sp. DTU009 \n 2.8 95 Fi17 \n Clostridia sp. DTU026 \n 1.9 82 Ba06 \n Porphyromonadaceae sp. DTU048 \n 2.7 84 Fi19 \n Clostridiales sp. DTU053 \n 1.9 96 Fi65 \n Pelotomaculum sp. DTU098 \n 2.6 97 Fi52 \n Clostridiales sp. DTU083 \n 1.9 93 Fi67 \n Clostridiales sp. DTU100 \n 2.6 80 Fi35 \n Clostridiales sp. DTU064 \n 1.9 86 Fi09 \n Syntrophomonas sp. DTU018 \n 2.6 97 Sy04 \n Synergistales sp. DTU085 \n 1.9 93 Fi43 \n Clostridiales sp. DTU074 \n 2.6 92 Fi53 \n Clostridia sp. DTU084 \n 1.9 79 Fi28 \n Clostridiales sp. DTU055 \n 2.6 91 Fi22 \n Clostridia sp. DTU030 \n 1.8 94 Fi39 \n Clostridiales sp. DTU069 \n 2.6 92 Eu03 \n Euryarchaeota sp. DTU008 \n 1.8 98 Fi62 \n Clostridia sp. DTU095 \n 2.5 88 Fi69 \n Clostridiales sp. DTU071 \n 1.8 52 Fi08 \n Syntrophomonas sp. DTU017 \n 2.5 88 Fi06 \n Clostridia sp. DTU015 \n 1.7 90 Fi15 \n Clostridiales sp. DTU024 \n 2.5 94 Ba05 \n Porphyromonadaceae sp. DTU047 \n 1.7 88 Fi12 \n Clostridia sp. DTU021 \n 2.5 87 Pr07 \n Campylobacterales sp. DTU103 \n 1.7 86 Fi51 \n Clostridiales sp. DTU082 \n 2.5 75 Fi33 \n Clostridia sp. DTU062 \n 1.7 79 Fi57 \n Clostridiales sp. DTU089 \n 2.5 92 Fi29 \n Bacilli sp. DTU057 \n 1.7 98 Pr10 \n Alcaligenaceae sp. DTU106 \n 2.4 87 Sp02 \n Treponemaceae sp. DTU108 \n 1.7 71 Fi34 \n Tepidanaerobacter sp. DTU063 \n 2.3 95 Fi02 \n Clostridia sp. DTU011 \n 1.7 83 Ba03 \n Porphyromonadaceae sp. DTU003 \n 2.3 84 Fi11 \n Clostridiales sp. DTU020 \n 1.7 71 Pr11 \n Desulfobulbaceae sp. DTU107 \n 2.3 86 Fi42 \n Clostridiales sp. DTU073 \n 1.7 93 Pr06 \n Alcaligenaceae sp. DTU102 \n 2.3 76 Fi23 \n Clostridiales sp. DTU031 \n 1.6 82 Fi07 \n Syntrophothermus sp. DTU052 \n 2.3 97 Fi24 \n Clostridiales sp. DTU032 \n 1.6 89 Fi05 \n Clostridia sp. DTU014 \n 2.3 94 Fi41 \n Clostridiales sp. DTU072 \n 1.6 96 Fi68 \n Clostridiales sp. DTU101 \n 2.2 75 Sy02 \n Synergistaceae sp. DTU044 \n 1.6 85 Fi20 \n Clostridiaceae sp. DTU054 \n 2.2 91 Sy03 \n Synergistaceae sp. DTU045 \n 1.5 92 Fi66 \n Clostridiales sp. DTU099 \n 2.2 88 Ba07 \n Rikenellaceae sp. DTU049 \n 1.5 68 Fi36 \n Clostridia sp. DTU065 \n 2.2 94 Fi46 \n Clostridia sp. DTU077 \n 1.5 68 Pr04 \n Gammaproteobacteria sp. DTU040 \n 2.2 91 Fi26 \n Clostridiales sp. DTU035 \n 1.5 90 Fi38 \n Clostridia sp. DTU068 \n 2.2 93 Te02 \n Acholeplasmatales sp. DTU061 \n 1.5 87 Fi55 \n Clostridiales sp. DTU087 \n 2.2 94 Fi25 \n Clostridiales sp. DTU033 \n 1.5 93 Fi47 \n Clostridiales sp. DTU078 \n 2.2 91 Te03 \n Acholeplasmatales sp. DTU067 \n 1.5 94 Fi10 \n Syntrophomonas sp. DTU019 \n 2.2 91 Th01 \n Thermotogaceae sp. DTU111 \n 1.4 82 Fi31 \n Clostridiaceae sp. DTU059 \n 2.2 94 Fi58 \n Clostridiales sp. DTU090 \n 1.4 75 Eu01 \n Methanoculleus sp. DTU006 \n 2.2 93 Fi50 \n Clostridiales sp. DTU081 \n 1.4 71 Fi18 \n Peptococcaceae sp. DTU027 \n 2.1 93 Fi27 \n Clostridiales sp. DTU036 \n 1.4 77 Fi37 \n Clostridiales sp. DTU066 \n 2.1 90 Fi14 \n Clostridiale sp. DTU023 \n 1.4 82 Fi04 \n Clostridiales sp. DTU013 \n 2.1 89 Sy06 \n Synergistales sp. DTU110 \n 1.4 55 Fi03 \n Clostridiales sp. DTU012 \n 2.1 94 Sy01 \n Anaerobaculum sp. DTU043 \n 1.4 59 Fi45 \n Clostridiales sp. DTU076 \n 2.1 96 Eu05 \n Methanothermobacter sp. DTU051 \n 1.2 78 Fi54 \n Clostridiales sp. DTU086 \n 2.1 90 Te01 \n Acholeplasmatales sp. DTU056 \n 1.2 95 Fi60 \n Clostridiales sp. DTU092 \n 2.1 90 Tm01 \n TM7 DTU050 \n 1.2 65 Sp01 \n Spirochaeta sp. DTU042 \n 2.1 90 Fi56 \n Clostridia sp. DTU088 \n 1.2 48 Fi64 \n Clostridia sp. DTU097 \n 2.1 72 Fi59 \n Erysipelothrix sp. DTU091 \n 1.1 96 Fi01 \n Clostridiales sp. DTU010 \n 2.1 92 Te04 \n Acholeplasmatales sp. DTU094 \n 0.8 85 Eu02 \n Methanoculleus sp. DTU007 \n 2.0 97 Sy05 \n Synergistaceae sp. DTU109 \n 0.8 57 Fi44 \n Clostridiales sp. DTU075 \n 2.0 92 Fi63 \n Eubacteriaceae sp. DTU096 \n 0.7 34 Fi61 \n Clostridiales sp. DTU093 \n 2.0 89 Pr09 \n Desulfomicrobium sp. DTU105 \n 0.7 30 Pr08 \n Rhodocyclaceae sp. DTU104 \n 2.0 74 Th02 \n Thermotogales sp. DTU112 \n 0.6 15 Fig. 1 Phylogenetic assignment of the 106 GBs. High-resolution microbial tree of life with taxonomic annotations, microbial phylogeny, and putative taxonomy, obtained with PhyloPhlAn using 400 broadly conserved proteins used to extract phylogenetic signal [ 66 ]. The tree was built using FigTree and contains a total of 3737 microbial genomes plus the 106 GBs identified (represented by small colored dots ). Organisms are colored based on phyla, those in light grey color text , were absent Proteobacteria is the second most abundant group (10 GBs), including Alcaligenaceae (Pr05, Pr06, Pr10), a group of three GBs that can only be assigned to the Gammaproteobacteria group (Pr01, Pr02, Pr04), two GBs belonging to the Deltaproteobacteria (Pr09, Pr11) and one belonging to the Campylobacteraceae (Pr07). Alcaligenaceae are not frequently reported in analysis of the biogas reactors [ 41 , 42 ] and the analysis of their genomic composition can provide a first glimpse into their possible role. GB Pr09 has been tentatively assigned to the Desulfomicrobium group and it is one of the most interesting members of Proteobacteria because it is competing with methanogens in anaerobic enrichment cultures degrading oleate and palmitate [ 43 , 44 ]. Finally, Pr11 is relevant because members of the Desulfobacterales are involved in acetate oxidation by parallel reduction of sulfur, a key process in the biogas microbial community [ 45 ]. Bacteroidetes is the third most abundant group (6 GBs) which is composed of two subgroups: Porphyromonadaceae (Ba03, Ba05, Ba06) and Rikenellaceae (Ba01, Ba02, Ba07). Both subgroups are dominant microorganisms in biogas plants [ 46 , 47 ]. Synergistetes , which was a recently introduced phylum having only Synergistaceae family, was represented by 6 GBs in our study. The remaining GBs are included in the phyla Actinobacteria (Ac01), Thermotogae (Th01, Th02), and Spirochaete (Sp01, Sp02). The abundance of all these GBs was low in the samples examined. Species of the phylum Thermotogae were identified also in thermophilic biogas-production plants utilizing renewable primary products for biomethanation, even at low abundances [ 48 – 50 ]. Their role in utilization of complex carbohydrates has been recently suggested on the basis of the gene content of Defluviitoga tunisiensis [ 51 ]. The low frequency of Spirochaetes is in agreement with relevant works in anaerobic digesters and their abundance seems to be highly variable depending on the operational conditions of the reactor [ 52 ]. Also, Actinobacteria have been previously reported at low abundance in biogas reactors [ 53 , 54 ] but in the cited research their functional role was difficult to be predicted due to the lack of genomic sequences and their highly variable physiological and metabolic properties. In the tree of life obtained using PhyloPhlAn, Tm01 was one of the most difficult taxonomic assignments as this GB was deeply branched from the candidate phylum TM7 composed only by the Candidatus Saccharimonas aalborgensis [ 16 ]. Despite its genome that is not completely closed (72 % completeness), it is one of the most complete TM7 reported in database and its small genome size (~1.2 Mbp) confirms data reported in the literature indicating that it is one of the smallest in the AD microbial community. Functional characterization of the biogas microbial community In the cited literature, the role of the majority of microbial groups involved in biogas production has been hypothesized considering the functional characteristics of distantly related species. Nevertheless, the lack of genome sequences prevents a clear understanding of their physiology and behavior. Therefore, our analyses targeted to give answers to two questions; (a) how much specialized are the microorganisms, and (b) which are their roles in the metabolic pathways of AD process. In order to elucidate the microorganisms’ specialization, we converted the results from SEED analysis (“ Methods ” and Additional file 1 ) into a Network Representation of the Biogas Functional Organization (NRBFO). For the construction of the NRBFO, we selected only the GBs that were ranked among the top 1/8 of each SEED functional category based on the number of corresponding genes. If one GB belonged in two categories, these were connected with an edge (Fig. 2 ; Additional file 4 ). This revealed that the GBs in the AD microbiome could be classified into two distinct groups according to their functional properties. Fig. 2 Network Representation of the Biogas Functional Organization (NRBFO). Nodes represent SEED functional categories. The size of each node is correlated to the number of GBs ranked among the top one-eighth of each functional category. Edges thickness is proportional to the number of GBs shared by two nodes; edge colors were used to simplify the visual observation of the connections. Thick edges connect nodes including GBs with high number of SEED feature counts in the two categories. Categories having thin edges are those comprising GBs that tend to have specialized functions The first group consists of GBs specialized on a single metabolic process, as the ones enriched in genes belonging to the general functional category of “carbohydrate utilization and metabolism” (blue nodes in Fig. 2 ). More specifically, some GBs were only involved in “central carbohydrate metabolism,” others in “aminosugars utilization,” or “di- and oligosaccharides utilization,” and so on, generating a very complex and faceted organization inside the microbial community. On the contrary, the second group includes GBs possessing multifunctional roles (i.e., they have high number of genes in more SEED categories). These GBs are inside the nodes connected by thick edges in the network (Fig. 2 ). It was found that 10 GBs have high number of SEED feature counts both in “sugar fermentation” and “fatty acids oxidation,” Three of these GBs belong to Syntrophomonadaceae family (Fi07, Fi08, Fi09), two belong to Alcaligenaceae family (Pr05, Pr10), two to Gammaproteobacteria (Pr01, Pr02) and three to Clostridia class (Fi12, Fi62, Fi68) (Additional file 5 ). It is known that common functionalities can be shared by species of the same taxonomic group [ 11 ]. However, our analysis proved that in some cases, species of completely different taxonomic groups can share the same functional role and therefore compete for the same niche. As an additional step, the species were functionally classified considering the proposed organization of the AD process, which is divided in four layers (i.e., hydrolysis, acidogenesis, acetogenesis, and methanogenesis) (Fig. 3 ; Additional file 1 : Figures S4–S7; Additional files 5 , 6 , 7 , 8 ). In order to do this, a putative functional role for the GBs was assigned taking into account their annotation obtained by COG, KEGG, SEED, and Pfam (Additional file 1 ). The assignment showed that the AD microbiome bears resemblance to a funnel concept; during the initial step of organic substrate degradation (i.e., carbohydrates, proteins, and lipids), a wide variety of GBs (even belonging to different phyla) are involved. In contrast, while proceeding to the next steps of the AD process (i.e., acetogenesis, acidogenesis, and methanogenesis), the involved GBs become gradually more specialized. Fig. 3 Functional roles of the GBs in the biogas production “food chain.” The main steps of the anaerobic degradation process are highlighted, together with the more relevant GBs involved. Functional roles were defined considering nearly complete KEGG pathways (Wood–Ljungdahl pathway, methanogenesis, propionate and butyrate metabolism), SEED categories (fatty acid degradation, carbohydrates utilization, denitrification, sulfate reduction), COG (amino acids fermentation) and Pfam (polysaccharides). Ovals refer to the compounds used by the microbial community (carbohydrates, fatty acids, proteins), intermediates (volatile fatty acids (VFA)-propionate, butyrate), and final products (carbon dioxide and methane) Particular attention was drawn to key functional steps of the AD process in order to elucidate the role of GBs. For example, proteins involved in polysaccharide degradation are important as it is well known that the raw manure contains a high fraction of fibers due to animal nutrition. These proteins were identified using the SEED annotation (polysaccharides category) and also by selecting those with significant matches to at least one of the carbohydrate-binding modules proposed by Hess et al. [ 55 ]. Analysis of the Pfam domains was performed in order to minimize the dependence on the overall sequence similarity of candidate genes to known carbohydrate-active enzymes. Out of 7161 carbohydrate-binding proteins found in the global assembly, 1896 of them (~26 %) were assigned to specific GBs. Most of the GBs with high number of carbohydrate-binding modules belong to Clostridiales and considering the similarity of the 107 essential genes obtained using BLAST, can be related to Ruminoclostridium or Clostridium . These genera are well known for their involvement in polysaccharides degradation, and some species have been previously isolated in biogas plants [ 56 ]. The carbohydrates utilization process involves numerous species which are specialized in degradation of different carbohydrates groups (Fig. 3 ). These microorganisms cooperate with species involved in lipids or proteins degradation to generate the byproducts for the subsequent steps of methanogenesis. In the fermentation of sugars to organic acids, an important role is played by the Wood-Ljungdahl pathway (WLP), which is characteristic for some acetogenic bacteria and archaea [ 57 ]. In this process, carbon dioxide is reduced to carbon monoxide and then converted to acetyl-CoA, with hydrogen serving as electron donor. From KEGG analysis performed on selected genes of the WLP, it was found that a specific subset of 8 bacterial species (Fig. 3 ; Additional file 1 : Figure S6) features a complete or nearly-complete pathway. All these bacteria were assigned to Firmicutes and more specifically to Clostridia sp. (Fi12, Fi13, Fi38, Fi46, and Fi62), Clostridiales sp. (Fi61), Peptococcaceae sp. (Fi18), and Tepidianaerobacter sp. (Fi34). It is known that specific microbes are capable to perform also the reverse WLP (i.e., the so called Syntrophic Acetate Oxidation, SAO), which includes the same genes of the WLP. By this pathway, they oxidize acetate to hydrogen and carbon dioxide when growing syntrophically with hydrogenotrophic methanogens that utilize the hydrogen and carbon dioxide produced to generate methane [ 58 ]. The overall process can be viewed as an additional mechanism of methane formation from acetate, and was originally proposed by Barker [ 59 ] and later confirmed by Zinder and Koch [ 58 ]. The mechanism was initially described in thermophilic anaerobic processes [ 58 , 60 , 61 ] and was later on identified to occur also in reactors operating at mesophilic temperatures [ 62 – 64 ]. Another finding is related to the synergistic behavior between Synergistetes with other microorganisms. SEED subsystem revealed that the most similar sequenced species to Sy02, Sy03, Sy05, and Sy06 is Thermanaerovibrio acidaminovorans DSM 6589 and for Sy01 is Anaerobaculum hydrogeniformans ATCC BAA-1850. Therefore, the presence of numerous ABC transporters for branched-chain amino acids (AA) (Additional file 5 ) together with the large number of genes involved in AA metabolism (Additional file 1 ) indicates that Synergistetes , similarly to T. acidaminivorans , could operate synergistically with other species, to ferment AAs to acetate and propionate [ 65 ]. Archaeal community characterization As previously discussed, the archaeal species are the best characterized in the biogas community. Archaea are dominated by the hydrogenotrophic methanogen Eu01 belonging to the Methanoculleus genus. Eu01, together with Eu02 (another Methanoculleus sp.), features all the central enzymes and complexes of methanogenesis: Mcr, Mtr, Fpo, and Hdr (Fig. 4 ). In addition, they feature all the complementary genes necessary for the reduction of CO 2 to methane: fmd / fwd , ftr , mch , mtd , and mer . On the contrary, they both lack the gene phosphate acetyltransferase ( pta ), involved in the conversion of acetate to methane (aceticlastic pathway) and also all the genes coding for the methylamine and methanol corrinoid proteins, essential for the conversion of methyl groups from methanol and methylamines to methane (methylotrophic pathway). Fig. 4 Comparison of the KEGG methane pathways of the 5 archaeal GBs (Eu01–05). In the upper part of the figure the reference KEGG methane metabolism pathway is represented, in the lower part archaeal GBs’ genes present and absent in the pathway are highlighted . Genes identified in the archaeal GBs were labeled with a small colored dot . Genes absent in the GBs and present in the reference genomes are marked with a “X” (Eu01–Eu02— Methanoculleus marisnigri ; Eu03— Candidatus Methanoplasma termitum ; Eu04— Methanosarcina acetivorans ; Eu05— Methanothermobacter thermoautotrophicus ). Genes identified in the GBs and absent in the reference are labeled with a circled dot \n Eu04, belonging to the Methanosarcinales genus, has a very low abundance (Fig. 5 ). It features all the genes belonging to the methylotrophic and aceticlastic pathway. Interestingly, it lacks the gene coding for Mtd, which catalyzes the fourth reaction in the reduction of CO 2 to methane. As stated before, Eu01 and Eu02 are instead able to perform all the reactions of this pathway, and the fact that these two GBs are approximately 4000 and 270 folds more abundant than Eu04 (Fig. 5 ) is an indication that the hydrogenothrophic pathway is the most favorable at the tested conditions. This finding is in accordance with several studies performed in similar conditions [ 5 , 6 , 10 ]. Fig. 5 Graphic representation of the GBs abundance in the biogas microbial community. The GBs coverages are represented as circles where the area is proportional to the coverage. GBs are grouped considering the taxonomic assignment at phylum level (Sp Spirochetes , Sy Synergistetes , Th Thermotogae , Pr Proteobacteria , Fi Firmicutes , Te Tenericutes , Ac Actinobacteria , Ba Bacteroidetes , Tm TM7 phylum, Eu Euryarchaeota ). Outlines colors correspond to those reported in Fig. 1 \n Interestingly, the second most abundant archaeon is represented by a completely new Euryarchaeota (Eu03). It is remarkable that Eu03 was the second most abundant archaeal species in the microbial community. It has very small genome size (~1.76 Mbp) similarly to Candidatus Methanoplasma termitum and Ca . Methanomassiliicoccus intestinalis (1.48 and 1.93 Mbp, respectively). Comparative analysis of the methane pathway (performed on all the archaeal GBs) using KEGG (Fig. 4 ) revealed that, differently from the other archaeal GBs, Eu03 completely lacks the coenzyme F420 biosynthesis pathway. This feature is also evident in the recently sequenced Ca. M. termitum [ 66 ]. Interestingly, it was possible to identify pivotal methanogenic genes in Eu03, including some belonging to the methylotrophic pathway, which are instead absent in the Methanomassiliicoccales species previously sequenced. Due to the small number of archaeal genomes in public databases, all the taxonomic analyses performed on the essential genes (Additional file 3 ) failed to assign Eu03 to any previously defined lineage. Only by using BLASTP similarity search and analysis of the 16S rRNA, it was possible to identify a distant correlation with the recently discovered seventh order of methanogens, the Methanomassiliicoccales (previously referred to as “ Methanoplasmatales ”). It is important to highlight that this new uncultured archaeon (Eu03) was identified, quantified, and assigned to a putative functional role only thanks to the binning strategy, which is the fundament of the present work."
} | 8,555 |
34659393 | PMC8519689 | pmc | 398 | {
"abstract": "The deep Q-network (DQN) is one of the most successful reinforcement learning algorithms, but it has some drawbacks such as slow convergence and instability. In contrast, the traditional reinforcement learning algorithms with linear function approximation usually have faster convergence and better stability, although they easily suffer from the curse of dimensionality. In recent years, many improvements to DQN have been made, but they seldom make use of the advantage of traditional algorithms to improve DQN. In this paper, we propose a novel Q-learning algorithm with linear function approximation, called the minibatch recursive least squares Q-learning (MRLS-Q). Different from the traditional Q-learning algorithm with linear function approximation, the learning mechanism and model structure of MRLS-Q are more similar to those of DQNs with only one input layer and one linear output layer. It uses the experience replay and the minibatch training mode and uses the agent's states rather than the agent's state-action pairs as the inputs. As a result, it can be used alone for low-dimensional problems and can be seamlessly integrated into DQN as the last layer for high-dimensional problems as well. In addition, MRLS-Q uses our proposed average RLS optimization technique, so that it can achieve better convergence performance whether it is used alone or integrated with DQN. At the end of this paper, we demonstrate the effectiveness of MRLS-Q on the CartPole problem and four Atari games and investigate the influences of its hyperparameters experimentally.",
"conclusion": "5. Conclusion How to improve convergence and stability of the DQN algorithm is one of the key issues in deep RL. In this paper, we propose MRLS-Q, a linear RLS function approximation algorithm with the similar learning mechanism to DQN. MRLS-Q can be used not only alone but also as the last layer of DQN. Similar to LS-DQN, the Hybrid-DQN with MRLS-Q can enjoy rich representations from deep RL networks as well as stability and data efficiency of the RLS method, but it can seamlessly integrate MRLS-Q and thus is easier to use. In MRLS-Q, we use the experience replay to break the correlation between training samples, present an average RLS optimization method to improve the convergence performance and reduce the computational complexity, employ an L 1 regularization technique to prevent overfitting, and propose a new method to define the feature function for alleviating the feature change of the same state and integrating MRLS-Q into DQN. Experiment results on the CartPole problem demonstrate that MRLS-Q has better convergence than Adam-Q and reveal the hyperparameter influences on MRLS-Q. In addition, experiment results on four Atari games demonstrate that DQN can improve convergence and stability by integrating with MRLS-Q.",
"introduction": "1. Introduction Reinforcement learning (RL) is an important machine learning methodology for solving sequential decision-making problems. In theory, by interacting with an initially unknown environment, the RL agent can learn the optimal action policies at different states to maximize the cumulative expected return [ 1 ]. Unfortunately, in the past several decades, due to the so-called “curse of dimensionality,” RL can only be used to solve some real-world problems with the small-scale discrete or low-dimensional continuous state space. It is not until 2013 that this dilemma was partially solved by Mnih et al. [ 2 ]. By combining the Q-learning algorithm with deep learning, they proposed the preliminary version of the deep Q-network (DQN) algorithm. Two years later, Mnih et al. [ 3 ] presented the normal version of DQN, which achieves the human-level performance on 49 classical Atari games. Since then, DQN has attracted more and more research attention, and many other novel deep RL algorithms [ 4 , 5 ] and new applications [ 6 , 7 ] have been proposed, and thus deep RL has become a thriving research branch in artificial intelligence. However, although DQN has succeeded in some more complicated problems [ 8 – 10 ], it still has many drawbacks, such as slow convergence, instability, and low sample efficiency. Therefore, we will focus on how to improve the DQN's performance in this paper. Currently, there are three main categories of research work on improving DQN. The first category mainly focuses on how to estimate action values accurately. For example, Hasselt et al. [ 11 ] proposed the double DQN, which can reduce the observed overestimation by exploiting the idea of double Q-learning. Wang et al. [ 12 ] introduced a dueling network architecture, which separately estimates state values and advantage values to improve the policy evaluation. Hausknecht and Stone [ 13 ] presented the deep recurrent Q-network, which is more suitable for solving partial observation problems, by adding recurrent LSTM layers to convolutional networks. Kim et al. [ 14 ] combined the mellowmax method with DQN to calculate the target action values, preventing overestimation effectively. Anschel et al. [ 15 ] proposed the averaged DQN, which uses some previously learned action-value estimates to produce the current action value. This algorithm can reduce the approximation error variance in the target values. The second category mainly focuses on how to explore or exploit samples efficiently. Schaul et al. [ 16 ] presented a prioritized experience replay, which can make the effective use of historical samples to improve the DQN's convergence performance. Fortunato et al. [ 17 ] proposed the noisynet DQN, which adds noise to the deep network parameters for aiding efficient exploration. Lee et al. [ 18 ] introduced an episodic backward update to improve the sample efficiency. The third category mainly focuses on how to reduce memory and computation. Mnih et al. [ 19 ] proposed asynchronous variants of four standard reinforcement learning algorithms, such as the asynchronous one-step Q-learning algorithm and the asynchronous n-step Q-learning algorithm. Interestingly, this work also opens the door to research the asynchronous advantage actor-critic (A3C) algorithm. In traditional RL, Q-learning algorithms often use linear functions to approximate action values, which have better stability and fewer hyperparameters to be trained than DQNs [ 20 ]. In particular, the least squares (LS) type RL algorithms, such as the least squares policy iteration (LSPI) algorithm [ 21 ], the fitted-Q iteration (FQI) algorithm [ 22 ], and the recursive least squares temporal difference with forgetting factor (RLS-TD-f) algorithm [ 23 ], not only have better stability but also have faster convergence. In the research community of adaptive filtering, the LS and the recursive least squares (RLS) algorithms are famous for their fast convergence rate. Obviously, the success of LS-type RL algorithms mainly benefits from this merit. In recent years, many new machine learning algorithms, such as the extreme learning machine (ELM) [ 24 ] and the broad learning system [ 25 , 26 ], have been proposed by combining LS or RLS algorithms. In practice, the last layer of the neural network used for DQN is usually a linear layer, which means that we probably can improve the DQN's performance by integrating DQN with the LS-type RL algorithms. In fact, Levine et al. [ 20 ] proposed a hybrid approach—the least squares deep Q-network (LS-DQN), which combines DQN with LSPI or FQI. By retraining the last layer of the policy network with a batch least squares update periodically, LS-DQN can obtain better convergence performance than DQN, whereas LS-DQN is not easy to use. At each update by using LSPI or FQI, LS-DQN needs to use the current network parameters to generate a training dataset, which requires running a forward pass of the deep network for each sample in the experience replay buffer. In addition, LS-DQN needs to generate new state-action features and compute the matrix inverse. From the DQN's learning mechanism, a perfect integrated LS-type algorithm should be able to use the inputs of the DQN's last layer for approximating action values and should have the same learning mode as DQN. In our previous work [ 27 ], we propose two policy control algorithms called ESNRLS-Q and ESNRLS-Sarsa. They seem to meet the above requirements to some extent, although they are also difficult to integrate with DQNs. They use the same experience replay and minibatch learning mode as DQN. In addition, they can avoid computing the matrix inverse and are more suitable for online learning by using recursive least squares (RLS). Based on this work and inspired by the work of Levine et al., we propose a novel minibatch RLS Q-learning algorithm with linear function approximation, called the MRLS-Q. Our main contributions are as follows. (1) By borrowing the experience replay to remove the temporal correlation between the observed transitions, we first combine the traditional Q-learning algorithm with the RLS optimization technique. (2) By using state features rather than state-action features for linear function approximation, we make MRLS-Q able to be used alone and also be integrated into DQN seamlessly. (3) In order to reduce the computational complexity and make the RLS method suitable for training parameters in the minibatch mode, we present an average approximation method for updating the RLS autocorrelation matrix. (4) In order to alleviate the feature change of the same state and integrate MRLS-Q into DQN, we present a new method to define the feature function of MRLS-Q. (5) We demonstrate the effectiveness of MRLS-Q, alone and as the last layer of DQN, by using the CartPole problem and four Atari games, respectively. We also test the influences of its hyperparameters experimentally. The remainder of this paper is organized as follows. Section 2 describes the related theories and algorithms of MRLS-Q. Section 3 represents the detailed derivation and the practical implementation of MRLS-Q. Then, in Section 4 , comparison experiments on the CartPole problem and four Atari games are conducted to separately verify the effectiveness of MRLS-Q used alone and as the last layer of DQN. Finally, Section 5 summarizes the whole paper."
} | 2,551 |
28509908 | PMC5607359 | pmc | 401 | {
"abstract": "On contemplating the adaptive capacity of reef organisms to a rapidly changing environment, the microbiome offers significant and greatly unrecognised potential. Microbial symbionts contribute to the physiology, development, immunity and behaviour of their hosts, and can respond very rapidly to changing environmental conditions, providing a powerful mechanism for acclimatisation and also possibly rapid evolution of coral reef holobionts. Environmentally acquired fluctuations in the microbiome can have significant functional consequences for the holobiont phenotype upon which selection can act. Environmentally induced changes in microbial abundance may be analogous to host gene duplication, symbiont switching / shuffling as a result of environmental change can either remove or introduce raw genetic material into the holobiont; and horizontal gene transfer can facilitate rapid evolution within microbial strains. Vertical transmission of symbionts is a key feature of many reef holobionts and this would enable environmentally acquired microbial traits to be faithfully passed to future generations, ultimately facilitating microbiome-mediated transgenerational acclimatisation (MMTA) and potentially even adaptation of reef species in a rapidly changing climate. In this commentary, we highlight the capacity and mechanisms for MMTA in reef species, propose a modified Price equation as a framework for assessing MMTA and recommend future areas of research to better understand how microorganisms contribute to the transgenerational acclimatisation of reef organisms, which is essential if we are to reliably predict the consequences of global change for reef ecosystems.",
"conclusion": "Conclusion We propose here an inclusive concept of adaptation sensu lato in coral reef organisms, that encompasses in addition to (epi)genetic adaptation of the host, a pervasive role of MMTA via symbiont shuffling and switching, genetic mutation, and horizontal transfer of beneficial genes that are inherited by offspring. Successful modelling of future reefs therefore requires an accurate assessment of several as yet understudied holobiont processes ( Figure 1 ). To date, most experimental research has focussed on determining the phenotypic responses of reef organisms in single-factor, short-term, single generation experiments, severely limiting our ability to assess the potential for evolutionary adaptation. Recommendations for future research on reef species therefore include (i) determining if variation in host fitness due to assembly of different microbiomes with unique microbial functions ultimately drives a multigenerational response to selection, (ii) experimentally assessing the mechanisms and rates of transgenerational acclimatisation and adaptation, (iii) manipulative experiments to alter microbial composition and assess phenotypic responses, (iv) determining the extent of functional equivalence in microbiomes to assess whether hidden diversity is critical for holobiont functioning under ocean change, and (v) using a modified Price equation ( Price, 1970 ; Govaert et al. , 2016 ) to integrate the relative contributions of (epi)genetic adaptations in the host via (a) germ line and mitotic mutations, (b) microbiome changes established by both shuffling and switching and (c) evolution within microbial strains including HGT."
} | 835 |
26601279 | PMC4646793 | pmc | 402 | {
"abstract": "DNA sequencing uncovers the entire architecture of below-ground plant–fungus networks.",
"introduction": "INTRODUCTION Ever since its introduction to ecology, network theory has repeatedly reorganized our understanding of the laws and processes that drive ecological community dynamics ( 1 – 3 ). The architecture of ecological networks and its consequences for community stability have been intensively investigated in mutualistic interactions involving plants and their pollinating or seed-dispersing animal partners ( 4 – 7 ). These plant–animal interactions commonly exhibit a “nested” network architecture ( 3 , 8 ), in which specialists (that is, species with narrow partner ranges) mainly interact with subsets of the partners of generalists ( 4 , 8 ) ( Fig. 1A ). Because nestedness is so prevalent in plant–animal mutualistic networks ( 3 , 8 ), understanding its impact on plant community processes is the key to understanding how biotic environmental changes (for example, extinction of indigenous partners or introduction of alien partners) can alter plant community structure ( 9 ). Fig. 1 Plant–fungus network architecture. ( A ) Schematic example of nested and antinested plant–fungus associations. In a nested network, specialists (that is, species with narrow partner ranges) interact with subsets of the partners of generalists (that is, species with broad partner ranges). Networks whose nestedness estimates are higher/lower than that expected by chance are regarded as nested/antinested. Antinestedness can result from compartmentalized and checkerboard network patterns. ( B to K ) Observed network structure. In each network of (B) cool-temperate (CL) (36 plants and 278 fungi), (C) warm-temperate (WM) (33 plants and 343 fungi), and (D) subtropical (ST) (36 plants and 580 fungi) forests, ectomycorrhizal fungi (red), arbuscular mycorrhizal fungi (blue), and fungi with unknown functions (gray) are linked with their host plants (white). The size of circles represents the relative abundances of fungi or plants in each network. The network-level interaction specialization (E), modularity (F), and nestedness (G) of each network (red bar) are shown with those calculated for randomized networks (blue bar; ±SD). Asterisks represent significant deviations from randomized index values. In addition, checkerboard scores representing how the overlap of host/symbiont plants is avoided in fungal (H) or plant (I) communities are shown. With the use of those network indices, a principal component analysis was also performed (J). The examined ecological networks are plotted on the surface defined by principal components axes (PC1, principal component 1; PC2, principal component 2). For the networks of plants and their root-associated fungi, each index was calculated also for the partial networks representing associations between particular functional or taxonomic groups of fungi and their host plants. The “rank abundance” curve representing the compositional evenness/unevenness of each plant community (K) is also shown. Theory predicts that a nested network architecture can promote plant species coexistence by offsetting among-plant competition, increasing persistence against random extinctions, and promoting facilitation ( 3 , 4 , 9 ). Moreover, a recent study ( 2 ) argued that a nested architecture can enhance species coexistence in mutualistic networks by increasing structural stability, which is mathematically defined as the range of parameter values that realize both feasible and dynamically stable equilibria. Thus, knowledge of such a potential link between network nestedness and species coexistence is crucial to preventing further plant biodiversity loss and consequential degrading of terrestrial ecosystem services worldwide. Despite the potential importance of network architecture to plant species coexistence and the advances promoted by the study of plant–animal mutualism, we remain ignorant of the network structural features of most major forms of plant–partner interactions. In terrestrial ecosystems, plant species mutualistically interact not only with pollinators and seed dispersers but also with functionally and taxonomically diverse root-associated fungi ( 10 , 11 ). Mycorrhizal symbiosis ecologically differs from interactions between plants and most of their pollinating and seed-dispersing agents because mycorrhizal fungi form physiologically intimate associations with host plants and some can simultaneously interact with more than one host individual ( 11 , 12 ). Furthermore, because the symbiosis of plants and root-associated fungi originated early in the history of land plants, it is a ubiquitous and major component of all terrestrial ecosystems ( 10 , 13 , 14 ) involving broader taxonomic ranges of plant species than pollination and seed dispersal interactions ( 15 – 17 ). In below-ground plant–fungus symbiosis, not only mycorrhizal fungi, but also various clades of endophytic fungi supply host plants with soil nutrients and receive photosynthetic carbohydrates in return ( 10 , 11 ). Among the major groups of root-associated fungi, ectomycorrhizal fungi display a relatively high host specificity and are considered to promote the dominance of plants in specific families such as Fagaceae, Pinaceae, and Dipterocarpaceae ( 11 , 18 ). In contrast, arbuscular mycorrhizal and root-endophytic fungi generally have broad host ranges ( 11 , 19 ), potentially working as interaction network hubs ( 5 , 20 ) in plant–fungus networks and thereby connecting otherwise isolated groups of species ( 5 ). In addition to these mutualistic fungi, potentially commensalistic or antagonistic fungi may also be involved in interaction webs with plants ( 15 – 17 ). Consequently, analyses of how these symbiotic networks embrace multiple functional groups of fungi and their host plants provide pivotal opportunities for examining relationships between network architecture and species coexistence in ecologically complex interactions ( 1 , 3 , 4 ). In a previous study, we found that a large network of 33 plant species and their root-associated fungi in a warm-temperate forest (35°02′N, 135°47′E) had unique properties, including the absence of nestedness ( 21 ). In this study, we examine the generality of those architectural features in plant–fungus networks by compiling additional large next-generation sequencing (NGS) data sets of the symbiosis of plants and their root-associated fungi in two contrasting forests in Japan [cool-temperate (42°40′N, 141°36′E) ( 17 ) and subtropical (30°26′N, 130°30′E) ( 15 ) sites] (fig. S1 and data file S1). Incorporation of these additional communities, together with the use of NGS, allows not only a comparative analysis of variation in the architectural properties of these networks of whole plant and fungal communities along a latitudinal gradient but also a deeper analysis of the phylogenetic and functional complexity of ecological networks than has been previously possible. We evaluate how these results differ in some important respects from patterns found in other kinds of ecological networks.",
"discussion": "DISCUSSION This study is the first to provide a picture of whole kingdom-to-kingdom ecological networks at three sites along a latitudinal gradient by exploring how phylogenetically and functionally diverse fungi constitute structured (that is, nonrandom or nested or antinested) interaction networks with plant communities. Our results build on studies that have provided insights into plant–mycorrhizal fungus networks based on DNA sequencing data sets ( 16 , 29 – 31 ). Previous studies of plant–fungus networks, however, have focused on either arbuscular mycorrhizal fungi ( 29 – 32 ) or ectomycorrhizal fungi ( 25 ) partly because these two fungal functional groups had been thought to considerably differ in their host specificity ( 11 ). Recent high-throughput sequencing analyses have altered the conventional view by showing that both arbuscular mycorrhizal and ectomycorrhizal fungal clades include host-specific and generalist fungi ( 16 , 17 , 33 ). Moreover, a plant individual can be simultaneously colonized not only by arbuscular mycorrhizal and ectomycorrhizal fungi ( 34 ) but also by diverse endophytic fungi ( 15 , 28 ). By targeting both arbuscular mycorrhizal and ectomycorrhizal fungi as well as endophytic fungi, we have shown how functionally and phylogenetically diverse fungi constituted networks with the whole plant communities. This study therefore provides a basis for contemplating how plant community structure is organized by the entire root-associated fungal community in light of findings in above-ground plant–partner networks. Although relatively comprehensive, our study is nevertheless limited in its ability to test the mechanisms that we broadly propose or extrapolate to other systems at larger scales. Even so, it provides several testable hypotheses that future research in plant–microbe networks will be more able to address with more resources and rapidly advancing methods. Our study lacks replicate habitats along the latitudinal gradient. We can therefore observe patterns that suggest mechanism but not readily confirm them. Furthermore, there are continuing challenges in applying operational taxonomic units (OTUs) in biologically realistic ways in molecular microbial community research ( 21 , 35 ). Thus, although we have more fully characterized the relationship between network architecture and species interactions than previous studies and have uncovered a diversity of intriguing network patterns, our findings primarily provide a more solid basis for future experimental studies and theoretical models. Our strongest finding is that, although most plant–pollinator and plant–seed disperser networks have nested architectures, the species-rich below-ground networks between plants and fungi are consistently antinested. There are several potential biological explanations for why the below-ground networks differ in structure from the above-ground plant–partner networks. These plant–fungus networks are composed of symbiotic interactions that could ecologically range from mutualism to antagonism, depending on environmental conditions ( 36 ). Generally, networks of symbiotic interactions are often more compartmentalized and less nested than networks of interaction among free-living species ( 26 ), and antagonistic networks are sometimes less nested than mutualistic networks ( 3 , 8 ). In addition, below-ground plant–fungus symbioses are unique in that some fungi are potentially able to establish symbiotic interactions with two or more host individuals ( 37 , 38 ). Finally, the potentially competitive aspects of the interactions identified here may themselves contribute to antinestedness. Collectively, these ecological aspects of plant root–associated fungus interactions may partially or fully explain the observed patterns, but studies partitioning these effects are needed. The observed antinestedness in plant–fungus networks suggests that the relationship between network architecture and ecological community processes [for example, relationship between nestedness and species coexistence ( 2 – 4 )] is more complex and variable than has previously been anticipated based on the studies of above-ground plant–pollinator and plant–seed disperser networks. Network architectural or compositional properties other than nestedness might be major determinants of community persistence in below-ground plant–fungus symbiosis. Recent theoretical studies have shown, for example, that the existence of antagonists in predominantly mutualistic interaction networks ( 39 ) could greatly affect community stability. The observed antinested architecture also highlights the hypothesis that below-ground plant–fungus symbiosis may restrict, rather than support, the species coexistence of plants in terrestrial ecosystems [sensu ( 40 )]. Ectomycorrhizal fungi, for instance, can promote the monodominance of their specific host-plant species in otherwise highly species-rich tropical forests ( 18 ). Thus, plants are simultaneously involved in several above-ground and below-ground networks, of which some (for example, plant–pollinator and plant–seed disperser networks) may promote the coexistence of their species, whereas others may counteract the facilitative effects. Accordingly, a comprehensive understanding of plant community structure and dynamics can be achieved only when we consider both above-ground and below-ground processes of plant–partner interactions. Although our results provide insights into how some plant–fungus networks are organized, our present data are limited in their ability to extrapolate the potential mechanisms by which plant community structure is organized by below-ground fungal communities worldwide. This study included no replicate sites at each climatic region, and the present observational approach precluded explicit tests of the relationship between plant–fungus network architecture and community persistence. In addition, more rare fungi will be detected by deeper sequencing in future analyses, and use of OTUs in microbial diversity analyses in itself deserves continuing methodological attention ( 21 , 35 ). The influence of spatial autocorrelations in sampling plots on the estimated plant–fungus network architecture also deserves further theoretical analysis ( 21 , 25 ). Consequently, our present study primarily provides a basis for important working hypotheses on the linkage between above-ground and below-ground community dynamics, instead of giving an opportunity for strict hypothesis testing. Future theoretical and experimental studies will deepen our understanding of how plant communities are organized by both above-ground and below-ground partners."
} | 3,453 |
36903681 | PMC10005145 | pmc | 403 | {
"abstract": "Memristors have been considered to be more efficient than traditional Complementary Metal Oxide Semiconductor (CMOS) devices in implementing artificial synapses, which are fundamental yet very critical components of neurons as well as neural networks. Compared with inorganic counterparts, organic memristors have many advantages, including low-cost, easy manufacture, high mechanical flexibility, and biocompatibility, making them applicable in more scenarios. Here, we present an organic memristor based on an ethyl viologen diperchlorate [EV(ClO 4 )] 2 /triphenylamine-containing polymer (BTPA-F) redox system. The device with bilayer structure organic materials as the resistive switching layer (RSL) exhibits memristive behaviors and excellent long-term synaptic plasticity. Additionally, the device’s conductance states can be precisely modulated by consecutively applying voltage pulses between the top and bottom electrodes. A three-layer perception neural network with in situ computing enabled was then constructed utilizing the proposed memristor and trained on the basis of the device’s synaptic plasticity characteristics and conductance modulation rules. Recognition accuracies of 97.3% and 90% were achieved, respectively, for the raw and 20% noisy handwritten digits images from the Modified National Institute of Standards and Technology (MNIST) dataset, demonstrating the feasibility and applicability of implementing neuromorphic computing applications utilizing the proposed organic memristor.",
"conclusion": "4. Conclusions In summary, a redox system consisting of ethyl viologen diperchlorate ([EV(ClO 4 )] 2 ) and triphenylamine-containing polymer (BTPA-F) was fabricated and used as the RSL of the organic memristor. When sandwiched between two metal electrodes, the bilayer-structured RSL exhibits memristive behaviors and excellent long-term synaptic plasticity. We evaluated the electrical characteristics and memristive behaviors of the memristor by applying consecutive positive and negative voltage sweeps for conductance setting and resetting, and small voltages of ± 0.2 V for current reading. The long-term potentiation and long-term depression properties together with the weight modulation rules were then investigated to verify the long-term plasticity of the device. At last, a three-layer artificial neural network was designed and implemented by customizing a synaptic crossbar array employing the presented memristor as the synapses. The experimental results demonstrate the feasibility and applicable of our device in implementing neuromorphic computing systems.",
"introduction": "1. Introduction In the big data era, huge volumes of data with much irregularity are rapidly generated every day, which is becoming a big challenge for information processing systems [ 1 , 2 , 3 ]. Recently, emerging neuromorphic computing methodologies, including artificial neural networks (ANNs), have attracted much attention due to their abilities to imitate the functions of synapses and neurons of the human brain and to deal with tricky tasks involving big data [ 4 , 5 , 6 ]. Numerous studies have also demonstrated their successful application in a wide range of fields in relationship with big data, such as computer vision (CV), pattern recognition, and natural language processing (NLP) [ 7 , 8 , 9 ]. However, the existing neuromorphic computing hardware platforms, including the General-Purpose Graphics Processing Unit (GPGPU) and application-specific accelerators such as Google’s Tensor Processing Unit (TPU) [ 10 ], are mainly designed on the basis of the traditional von Neumann architecture, where the computation and memory units are physically separated from each other, inevitably incurring frequent data movement back and forth between them. Consequently, this leads to the memory wall problem and brings about a performance bottleneck and low-energy efficiency to the system [ 11 , 12 ]. The shortcomings are exposed more obviously especially when the von-Neumann-architecture-based systems are used to deal with data-intensive applications such as neuromorphic computing. As in the human brain, synapses are the fundamental and key elements of a neuromorphic system. The synapses in existing neuromorphic systems are usually realized using CMOS devices, hindering the design of synapses and neurons with a higher density in neuromorphic systems. Therefore, it is quite essential to explore novel devices beyond von Neumann computing paradigms to design more performance- and energy-efficient synapses, neurons, as well as neuromorphic computing systems. Recently, memristors have been widely used to design artificial synapses and neurons, owing to their numerous advantages such as such as nonvolatility, high density, low-power consumption, and CMOS compatibility [ 13 , 14 , 15 , 16 , 17 , 18 ]. The resistive switching and multiconductance state properties allow the memristor-based synapses and neurons to do in situ computing in an analogue fashion with the help of the Ohm’s Law and Kirchhoff’s Current Law (KCL), inherently fusing the functions of memory and computation into the identical devices [ 19 ]. As a result, the frequent data movement problem can be addressed, improving both the performance and energy efficiency of the system. Additionally, the nonvolatility of memristors enables the stored synaptic weights to be kept unchanged even when the system is powered off, which can further lower the overhead during a system initiation. Furthermore, when stimuli of appropriate amplitudes and widths are applied, biological synaptic behaviors such as spike-timing-dependent plasticity (STDP), potentiation, and depression can be observed on memristors [ 20 ]. Afterwards, a neuromorphic computing system beyond the von Neumann paradigm could be established where memristors are utilized to construct artificial electronic synapses and neurons. Therefore, memristors are considered to be one of the most promising next-generation neuromorphic devices. Numerous reports [ 20 , 21 , 22 , 23 , 24 , 25 , 26 , 27 , 28 , 29 , 30 ] have broadly investigated and demonstrated the applicability of memristors in such fields. However, most of them employed memristors based on inorganic materials. For instance, memristors based on metal oxide materials such as HfO x or ZnO were used to realize artificial synapses in [ 20 , 21 , 22 ], while sulfide materials such as Ag 2 S or Cu 2 S were used in [ 23 , 24 ]. Additionally, other works [ 27 , 28 , 29 ] also employed memristors based on 2D materials to realize artificial synapses and neuromorphic computing systems. Compared with inorganic counterparts, memristors based on organic materials have many more strengths, including low cost, easy manufacture, high-mechanical flexibility, biocompatibility, and more importantly, tunable electronic properties [ 19 , 31 , 32 ]. This allows them to be applied in scenarios such as wearable devices or even skin-implantable systems. In this work, a two-terminal organic memristor using ethyl viologen diperchlorate [EV(ClO 4 )] 2 /triphenylamine-containing polymer (BTPA-F) as the resistive switching layer (RSL) is presented. The bilayer-structured RSL between two metal electrodes exhibits memristive behavior and excellent long-term synaptic plasticity, which is of great importance for artificial synapses and neurons. Additionally, the conductance states can be precisely modulated by applying appropriate voltage pulses, making it possible to design multilevel-weight synapses using the device. A multilayer perception (MLP) neural network [ 33 ] was designed and implemented using the EV(ClO 4 ) 2 /BTPA-F-based memristor to examine the feasibility of implementing neuromorphic computing systems employing the proposed device. By taking advantage of the nonvolatile and tunable conductance of the device and the KCL, the simulated hardware network is capable of storing synaptic weights and doing neuromorphic computing on identical devices, which can significantly improve the performance and lower the power of a neuromorphic computing system. The synaptic weights of the network were trained and modulated on the basis of the memristor’s synaptic plasticity characteristics and conductance modulation rules. Recognition accuracies of 97.3% and 90% were achieved, respectively, for the raw and 20% noisy handwritten digits images from the Modified National Institute of Standards and Technology (MNIST) dataset [ 34 ]. Thus, the EV(ClO 4 ) 2 /BTPA-F-based organic memristor is a promising candidate for neuromorphic computing applications.",
"discussion": "3. Results and Discussion 3.1. Electrical Characteristics and Long-Term Synaptic Plasticity The EV(ClO4) 2 /BTPA-F RSL exhibits memristive behavior when sandwiched between the top electrode tantalum and the bottom electrode platinum. Compared with the related literature [ 35 , 36 , 37 , 38 ], the devices prepared with the dielectric materials in this manuscript have more stable linear conductance states, as well as better endurance and biocompatibility. The current–voltage curve of the device is shown in Figure 2 a. During the set process, the window displayed by the high- and low-resistance-state transitions indicates that there is a large switching ratio (>10). During the reset process, there are many slowly changing stable and controllable resistance states rather than instantaneous flipping. Taking the ternary conductance as a typical value, the electrical parameters of the system test the device at room temperature. Experimental results show that the conductance state can be repeatedly programmed and accessed within 500 cycles ( Figure 2 b) and remains for at least 10 4 s ( Figure 2 c). It was found that, by means of ion transport and compensatory doping, electrons can be removed from the main chain of triphenylamine polymers through redox to generate holes which can not only increase the concentration of mobile carriers but also generate a new polaron energy level in the original energy gap. Thus, the carrier mobility can be further adjusted by taking advantage of the change of the energy level between the adjacent groups. Consequently, the conductance of the bilayer-structured RSL can be precisely modulated. Experiments were conducted to observe the tuning process of the memristive behavior. As indicated in Figure 3 a, the conductance of the device will gradually increase (from 0.04 mS to 0.1 mS) when consecutive positive voltage sweeps of 0 V → 1 V → 0 V are applied to the top and bottom electrodes of the device. Correspondingly, the conductance of the device will decrease (from 0.05 mS to 0.1 mS) when consecutive negative voltage sweeps of 0 V → −1 V → 0 V are applied to the device, as shown in Figure 3 b. Utilizing the nonvolatility of the device, applying the same positive or negative scanning voltage to it sequentially seven times, the conductance value will show seven continuously increasing or decreasing conductance values. The curves of different colors in the figure represent different memristive states, and the variation range of the resistive state shows excellent symmetry. Different from those bistable memristors with abrupt changing conductance [ 39 , 40 ], our device shows a slower and smoother conductance tuning trend, which is more useful in artificial electronic synaptic applications [ 31 ]. The conductance value is positive and the weight has positive and negative values. Here, we add the maximum value (Gmax) of the device to the minimum value (Gmin) and then divide the sum by two. Use this value as the critical point. If the conductance value of the device is greater than this value, it will be a positive weight, otherwise it will be negative. When the activity between the presynaptic neuron and postsynaptic neuron increases or decreases, the synaptic connection will be strengthened or weakened. The change in the strength of the synaptic connection is defined as synaptic plasticity [ 41 ]. As one of the basic elements of synaptic plasticity, long-term plasticity indicates long-lasting changes in synaptic weight and is believed to be related to the learning and memory mechanisms in the human brain [ 42 ]. The phenomenon of the long-lasting or permanent increase in synaptic weight is referred to as long-term potentiation (LTP). By contrast, the phenomenon of long-lasting or permanent decrease in synaptic weight is referred to as long-term depression (LTD) [ 43 ]. LTP and LTD can be used as the basic rules of synaptic weight renewing and modulating in neuromorphic computing systems. The Ta/EV(ClO 4 ) 2 /BFPA-F/Pt memristor can be used as electrical synapse with synaptic plasticity, where the top electrode tantalum acts as the presynaptic neuron while the bottom electrode platinum acts as the postsynaptic neuron. Figure 4 a depicts the conductance response of our device on applying consecutive positive or negative voltage pulses, which demonstrates the LTP and LTD properties of the memristor. To begin with, 50 consecutive positive voltage pulses with the amplitude of 1 V, duration of 10 ms, and period of 2 s are applied to the Ta/EV(ClO 4 ) 2 /BFPA-F/Pt memristor. Subsequently, 50 consecutive negative voltage pulses with the identical amplitude, duration, and period were immediately applied to the device. The positive and negative voltage pulse stimuli caused the occurrence LTP and LTD, as indicated by the blue and red curves, respectively. Additionally, the LTP and LTD can also be charactered by the spike-timing-dependent plasticity (STDP) [ 44 , 45 ]. STDP is a temporally asymmetric Hebbian learning rule induced by tight temporal correlations between presynaptic and postsynaptic neuronal spikes through which the connection strength between neurons can be modulated [ 46 , 47 ]. Figure 4 b demonstrates the STPD properties of the Ta/EV(ClO 4 ) 2 /BTPA-F/Pt memristor through the schematic illustration of the anti-STDP window [ 48 , 49 ]. Δ W denotes the synaptic weight change in the device and can be calculated by the following equation [ 50 ]\n (1) Δ W = I p o s t − I p r e I p r e \nwhere I p o s t and I p r e denote the current of presynaptic and postsynaptic spikes and Δ t ( t p o s t − t p r e ) denotes the time interval between the post- and presynaptic spikes. When the postsynaptic spike arrives before the presynaptic ( Δ t < 0 ), the synaptic weight change ( Δ W ) will be positive, wherein the value of synaptic weight will increase gradually, indicating an LTP process. On the contrary ( Δ t > 0 ), the synaptic weight change will be negative wherein the value of synaptic weight will decrease gradually, indicating an LTD process. The synaptic weight retention performance of our device in response to temperature change is also examined in this work. The result shown in Figure 4 c demonstrates that our device can tolerate a wide range of temperature without obvious synaptic weight loss, which allows the device to be applied in a wide range of temperature environment. 3.2. Neuromorphic Network Implementation The biological presynapse and postsynapse can be respectively mapped to the top and bottom electrodes of the memristor, and the conductance value corresponds to the synaptic weight. Applying pulse voltage on the memristive device can be used to replace the nerve stimulation signal of neurons. The characteristics of nerve stimulation signals corresponding to different synaptic functions can be simulated by changing the shape, frequency, duration, and other parameters of the pulse voltage. As shown in Figure 5 a, a three-layer MLP neural network was designed and implemented utilizing the Ta/EV(ClO 4 ) 2 /BTPA-F/Pt memristor for the purpose of demonstration of the feasibility of our device in implementing neuromorphic computing systems. The supervised learning based on the backpropagation (BP) algorithm was employed to train the network using 60,000 images from the MNIST database, a standard benchmark widely used to gauge machine learning algorithms. The grayscale of the image is represented by the conductance value. According to the corresponding relationship between the pulse and the memristive state, the grayscale of the image in the database is mapped to the number of spike pulses that need to be applied. Each input image was scaled to 8 pixels by 8 pixels to match up the size of our custom network. Through cropping and bicubic interpolation downsampling methods, the effective information of the image is preserved under the condition of adapting the input quantity of the network. It is worth mentioning that the more integrated memristive network has more input features, thus achieving higher resolution image recognition. A total of 64 input neurons of the network corresponded to the total amount of pixels of one image, while 10 output neurons corresponded to 10 handwritten Roman numerals. The weights were updated during the learning process based on the experimental data sampled by testing on the Ta/EV(ClO 4 ) 2 /BTPA-F/Pt synaptic memristor according to the LTP and LTD modulation rules. A dedicated crossbar array based on the presented memristor was then designed to simulate the custom neural network, as illustrated in Figure 5 b. The top electrodes of the devices on an identical row were connected to a word-line (WL) while the bottom electrodes of the devices on an identical column were connected to a bit-line (BL). The top and bottom electrodes of each individual memristor device mimicked the pre- and postsynaptic neuron, respectively, while the bilayer-structured RSL of the device acted as the synapse. The custom MLP neural network consist of 5920 (80 rows × 74 rows) artificial synapses, each of which was initialized to the minimum conductance of the presented. The training dataset from MNIST was used in the training duration, with a mini-batch size of 60. As illustrated in Figure 5 c, the network training was composed of two stages: feedforward inference and feedback weight update. The synaptic weight of each synapse was kept unchanged and used in each feedforward inference iteration and updated in each backward iteration by applying voltage pulses according to the LTP and LTD modulation rules of the device. The feedforward inference was performed layer by layer sequentially, as was the backward weight update. The input voltage vector for the first layer was a feature vector from the dataset, while the input vector for the subsequent layer was the output vector of the previous layer. The analogue weighted sum can be performed along bit-lines according to the Ohm’s law and Kirchhoff’s law [ 51 , 52 ], demonstrating that the in situ computing is enabled in the memristor array. The total current of each bit-line was the summation of the currents through each device in the same column, while each current was the product of the conductance and the corresponding voltage across the memristor. The input signal of hidden neurons can be derived from the Equation (2): (2) I j l = ∑ i = 1 64 W i j l V i l \nwhere V i l denotes the input voltage vector applied to the top electrodes of the synaptic devices, I j l denotes the readout current vector from the bottom electrodes of the synaptic devices, while W i j l denotes the weight matrix of layer l . Then, the current results were activated by a nonlinear sigmoid transfer function. The activated result of a hidden layer was transferred to the output neurons. The inference result was calculated by Equation (3): (3) V i l + 1 = σ ( I i l ) = { c I i l , I i l > 0 0 , I i l ≤ 0 \nwhere c is 800 V/A, which is a scaling factor matching the voltage range of the device. The resulting voltage elements exceeding 0.8 V were clipped to avoid changing the memristor state. V 2 and V 3 are the output signals of the second and third layers of the network, respectively. The forward propagation process ends at this point. During the BP process, delta weights are calculated and transferred to modify the synaptic weights with the driving circuit, as the Equations (4) and (5) show: (4) δ j l ( n ) = { ∂ f ∂ v j l + 1 ( n ) [ t j ( n ) − y j ( n ) ] l = 2 ∂ f ∂ v j l + 1 ( n ) ∑ i W i j l + 1 δ i l + 1 l = 1 \nwhere y j ( n ) and t j ( n ) represent the input feature vector and the target output vector (label), respectively; v j l + 1 and f denotes the output voltage vector of the postsynaptic electrode and the activation function, respectively.\n (5) Δ W i j l = η ∑ n = 1 60 δ j l ( n ) V i l ( n ) \nwhere η is the learning rate and δ j l is the calculated error between the real output and the corresponding target value during the training process. When the feedback was transferred to the weights of the first layer, an epoch finished. The MNIST database contains a total of 70,000 handwritten digital pictures, of which 60,000 are used to train the neural network and the remaining 10,000 are used to test and validation the result of network accuracy. There are two training methods: online and offline. Online training is real-time training and testing in the hardware circuit. Offline training is training to obtain weights first and then adjusting the memristor conductance value. Online training requires more calculations and time and is prone to overfitting; therefore, we adopted offline training. The training database was used to train the custom MLP neural network. After being trained for 40 epochs, the recognition performance of the network implemented using the presented memristor was examined using the testing database ( Figure 6 a) from the same database. After a feedforward inference process, the maximum value of the output neuron was taken as the inference result. As a result, a recognition accuracy of 97.3% was achieved. Figure 6 b shows the inference results for the ten digits from “0” to “9”, indicating that the neural network based on the presented memristor exhibits excellent recognition performance. Figure 7 a shows the recognition accuracy in response to neural network structure with different layer numbers. With the increase of training epochs, the recognition accuracy of the neural network with two hidden layers is significantly higher than that with one layer. The performance of the custom network under different Signal-to-Noise Ratio (SNR) was then tested in this work ( Figure 7 b). We used MATLAB (2020a, Natick, MA, USA) to randomly generate noisy matrices with different signal-to-noise ratios and superimposed them on the original image matrix, where the noise amplitude obeyed Gaussian distribution. The recognition accuracies of 93.7%, 90%, and 81.3% were achieved under an SNR of 10, 5, and 1, respectively, indicating that the custom neural network implemented using the presented memristor can tolerant a relatively high noise contamination. Figure 7 c–f show two cases indicating that noise contamination may bring about accuracy drop for the recognition task."
} | 5,725 |
29062483 | null | s2 | 404 | {
"abstract": "Superhydrophobic surface simultaneously possessing exceptional stretchability, robustness, and non-fluorination is highly desirable in applications ranging from wearable devices to artificial skins. While conventional superhydrophobic surfaces typically feature stretchability, robustness, or non-fluorination individually, co-existence of all these features still remains a great challenge. Here we report a multi-performance superhydrophobic surface achieved through incorporating hydrophilic micro-sized particles with pre-stretched silicone elastomer. The commercial silicone elastomer (Ecoflex) endowed the resulting surface with high stretchability; the densely packed micro-sized particles in multi-layers contributed to the preservation of the large surface roughness even under large strains; and the physical encapsulation of the microparticles by silicone elastomer due to the capillary dragging effect and the chemical interaction between the hydrophilic silica and the elastomer gave rise to the robust and non-fluorinated superhydrophobicity. It was demonstrated that the as-prepared fluorine-free surface could preserve the superhydrophobicity under repeated stretching-relaxing cycles. Most importantly, the surface's superhydrophobicity can be well maintained after severe rubbing process, indicating wear-resistance. Our novel superhydrophobic surface integrating multiple key properties, i.e. stretchability, robustness, and non-fluorination, is expected to provide unique advantages for a wide range of applications in biomedicine, energy, and electronics."
} | 394 |
37895680 | PMC10608025 | pmc | 405 | {
"abstract": "The von Neumann architecture has faced challenges requiring high-fulfillment levels due to the performance gap between its processor and memory. Among the numerous resistive-switching random-access memories, the properties of hexagonal boron nitride (BN) have been extensively reported, but those of amorphous BN have been insufficiently explored for memory applications. Herein, we fabricated a Pt/BN/TiN device utilizing the resistive switching mechanism to achieve synaptic characteristics in a neuromorphic system. The switching mechanism is investigated based on the I–V curves. Utilizing these characteristics, we optimize the potentiation and depression to mimic the biological synapse. In artificial neural networks, high-recognition rates are achieved using linear conductance updates in a memristor device. The short-term memory characteristics are investigated in depression by controlling the conductance level and time interval.",
"conclusion": "4. Conclusions In this study, we aimed to operate BN-based RRAM as a synaptic device in an artificial neural network. With a Pt/BN/TiN structure, XPS analysis revealed the involvement of boron, nitrogen, and oxygen bonds. To understand the device’s switching mechanism, we used cells with various sizes and adopted the 100 × 100 μm size which yielded minimal variations. We obtained identical I–V curves from 10 randomly selected cells during 100 cycles. Based on the DC switching results, we proceeded with potentiation and depression. For potentiation, we compared linearity by applying multiple pulses within the same period. As the pulse frequency increased, the conductance and nonlinearity increased. For depression, six states were set from 100 to 350 μS, and readings were performed over time for all of them. When XX was operated as hardware for MNIST using the optimal values, we achieved a high-recognition rate equal to 94.96%. This study demonstrated that BN-based RRAM can achieve high performance in neuromorphic systems.",
"introduction": "1. Introduction The recent advances in the Internet of Things (IoT) and artificial intelligence (AI) have led to increased information influx, thus necessitating higher-performing processors, minimal power consumption, and increased computation speeds [ 1 , 2 ]. However, conventional computing systems, based on the von Neumann architectural design have bottlenecks attributed to their serial data processing structures [ 3 , 4 , 5 ]. In response to this challenge, a recent study proposed an alternative architecture to that used in von Neumann systems [ 6 , 7 ]. A neuromorphic system features parallel networks comprised of neurons and synapses. It performs complex tasks, such as pattern recognition, very efficiently because the processor and memory are closer within a single semiconductor chip compared with their arrangement in the traditional computing system [ 8 ]. Next-generation memories, such as phase-change memory (PRAM) [ 9 , 10 ], ferroelectric random-access memory (FRAM) [ 11 , 12 , 13 ], and resistive-switching random-access memory (RRAM) are commonly employed in neuromorphic systems [ 14 , 15 , 16 ]. The use of numerous materials in the construction of the resistor imparts different characteristics to RRAM, with each material having different electrical and chemical properties as well as ion-migration behavior. Consequently, RRAM chooses resistive switching layers and metal electrodes that suit its requirements. RRAM stores data in the low-resistance state (LRS) (or ON state) and high-resistance state (HRS) (or OFF state). The mechanism of RRAM is classified as either a filamentous or nonfilamentary type. A filament-type RRAM has nonuniform set/reset voltages. The nonuniformity of the switching is caused by the random formation and rupture of the conduction filament. Owing to the rapid change in the set process from HRS to LRS, the filament type of the RRAM device was also used with the selector as a synaptic device in neuromorphic systems [ 17 ]. Conversely, the nonfilamentary RRAM type possesses bidirectional (gradually increasing) conductance. Therefore, the set/reset voltages can be uniformly obtained [ 18 ]. Transition metal oxides (TMOs) are deposited using a variety of processes, including sputtering [ 19 , 20 , 21 , 22 , 23 , 24 , 25 , 26 ]. Recently, nitride-based devices have been studied to create efficient synaptic and memory devices to control accurately conductive paths and metal/semiconductor barriers in these devices [ 27 , 28 , 29 ]. The amorphous BN thin films have attracted significant attention for memory applications owing to wide-bandgap semiconductors with high-thermal conductivity and chemical stability [ 30 , 31 ]. The integration of BN with RRAM enables the construction of neural network models with a smaller memory footprint and capability for faster inferences [ 32 , 33 , 34 ]. In this study, the potentiation and depression of the Pt/BN/TiN device in neuromorphic computing applications are mimicked using changes in conductance values, especially with multi-level cell (MLC) configurations. The performance as a synaptic device is tested by analyzing Modified National Institute of Standards and Technology database (MNIST) data, thus demonstrating the high potential and promise of the Pt/BN/TiN device as a viable candidate for neuromorphic computing implementation. Potentiation is a synaptic plasticity effect induced by the continuous accumulation of conductance, thus promoting the enhancement of connection strength. Conversely, depression involves constant conductance reduction to weakened synaptic connections [ 35 ]. The MNIST, well-known for its handwritten digit dataset, is essential for enabling thorough evaluations of deep-learning and machine-learning models. The collection of MNIST images with handwritten numbers plays an essential role in enabling a comprehensive examination and appraisal of various deep- and machine-learning models [ 36 ]. The impact of BN films on the resistive and synaptic characteristics is discussed in this study. Pt serves as the top electrode of the device and TiN as the bottom electrode. The materials used for both electrodes affect the resistive switching window and stability. The study focuses on the remarkable nonfilamentary resistive switching characteristics observed in the Pt/BN/TiN memory device.",
"discussion": "3. Results and Discussion X-ray photoelectron spectroscopy (XPS) is a valuable tool used for the analysis of the chemical compositions and properties of target materials. We utilized the XPS in surface mode to investigate the amorphous BN (a-BN) film. In the case of BN, like AlN, it oxidizes easily. Therefore, when deposited using sputter techniques, it is unavoidable to encounter oxygen in XPS analysis [ 37 , 38 ]. Figure 2 a displays the B 1s spectra of a-BN that reveal two peaks related to the B-N (190.5 eV) and B-O (192 eV) bonds [ 39 , 40 ]. This suggests that oxygen participates in the a-BN bonding during the deposition of the metal electrode as well as the dielectric. Figure 2 b illustrates the N 1s spectra, which have two peaks at 399.15 and 399.85 eV. They are assigned to N-B and N-H, respectively. The O 1s spectra, as shown in Figure 2 c, show peaks at 529.7 eV (related to the OH groups) and 531.4 eV (related to the O-B bonds). The oxygen element did not originate from natural oxidation, but from residual oxygen gas inside the equipment during BN sputtering, ultimately indicating the formation of a BNO film. To characterize the electrical properties of the Pt/BN/TiN device, the I–V curves were constructed. The I–V curves with different sizes of BN cells are shown for 100 × 100 μm 2 ( Figure 3 a), 50 × 50 μm 2 ( Figure S1a ), 30 × 30 μm 2 ( Figure S1b ), and 10 × 10 μm 2 ( Figure S1c ). The entire process was conducted in the minimum compliance current range to prevent the breakdown of the device. Moreover, in the positive voltage range, a set switching occurred from the HRS to the LRS, whereas in the negative voltage range, a reset switching occurred from the LRS to HRS, thus demonstrating typical bipolar switching behavior. The cell in Figure 3 c switches at 4 V with a compliance current (CC) of 0.1 mA and an applied voltage to the cells of −8 V to return to the HRS. For 100 cycles, there is virtually no shift when the condition values are fixed. The cell (with an area of 50 × 50 μm 2 ) changes to LRS at approximately 3.2 V at a lower CC of 0.01 mA and to the HRS at −5 V. Given its smaller size, this cell works at a lower CC level as there are relatively fewer nitrogen and oxygen ions (vacancies) involved in the electrical switching [ 41 ]. This trend is also observed in the cells with areas equal to 30 × 30 μm 2 and 10 × 10 μm 2 . The cell with an area of 30 × 30 μm 2 switches using a current of 0.01 mA at 2.7 V in the set and at −2 V in the reset processes. The reduced ions cause a voltage shift, thus leading to variations during cycling [ 42 ]. The cell with an area of 10 × 10 μm 2 operates at an even lower CC of 0.001 mA, but the reset operation occurs only in the first cycle. Moreover, the endurance deteriorates further compared with the cell with an area of 30 × 30 μm 2 . Figure 3 b shows the variation of the resistance range (difference between the HRS and LRS) for 100 cycles. The on/off ratio varied from 19.7 to 8.5. Figure 3 c shows the I–V curves for 20 cell−to−cell switching cycles. Randomly selected 20 cells were measured under the same measurement conditions (set: ~4 V, reset: −8 V, and current range: ~0.1 mA). There are no major malfunctions during operation (over 10 cycles and in each cell). Figure 3 d shows the change in resistance over time that demonstrates the retention of the device. Over a period of 10,000 s, both HRS and LRS read at 1.5 V increase from 1.86 × 10 7 to 3.92 × 10 9 and 4.26 × 10 5 to 2.53 × 10 9 , respectively. This indicates that the cell with an area of 100 × 100 μm 2 also exhibits short-term memory characteristics. The BN-based memristor used as a synaptic device emulates the connection between neurons, as shown in Figure 4 a. When a stimulus is transmitted from a presynaptic neuron to a postsynaptic neuron, the synapse is involved wherein the neurotransmitter is released [ 43 ]. Similarly, the memristor device transfers the signal from the top electrode (Pt) to the bottom electrode (TiN) via the BN layer. Nitrogen ions and nitrogen vacancies that coexist in BN layer due to the presence of anti-Frenkel pairs could affect resistive switching [ 44 , 45 ]. Figure S2 illustrates the multilevel characteristics achieved by varying the compliance current and reset voltage. Applying a fixed set voltage of 6 V while varying the compliance current from 50 μA to 1 mA results in distinct LRS. This outcome is attributed to an increased movement of ions within the BN layer’s interface as the compliance current increases, leading to a higher current flow and the appearance of multilevel characteristics. Additionally, the achievement of multilevel states is demonstrated by increasing the reset voltage. Varying the reset voltage from −7 V to −8.5 V is associated with recombination processes. With an increase in the reset voltage, more recombination occurs with the vacancies, consequently resulting in different HRS. Furthermore, the action of potentiation is achieved, which enhances the weight and weakens depression in the artificial synapse device. Linear conductance update is needed in RRAM devices to implement the neuromorphic system. In other words, the nonlinear update of conductance is a challenge as it severely degrades the performance of neuromorphic systems. Unpredictable values during the conductance updating process are not desirable. Thus, the pulse design is tailored to allow potentiation and depression predictions when identical pulses are used. Figure 5 a–c illustrates the potentiation performed by setting differently only the number of pulses and interval time within a period of 10 s (as detailed in Figure S3 ). All data are reset to conductance based on an initializing process before potentiation that involves 10 inputs at −5 V and 50 ms each. The nonlinearity parameter (α) is calculated using the following formula and is depicted in Figure 5 d [ 46 ].\n (1) G = { ( ( G L R S α − G H R S α ) × w + G H R S α ) 1 α i f α ≠ 0 , G H R S × ( G L R S / G H R S ) α i f α = 0 . \nwhere G LRS and G HRS are the minimum and maximum conductance, respectively, α is a parameter that attains linearity and symmetry, and w is a synaptic weight increased or decreased according to the pulse (from zero to one). Inputting more pulses within the same period is equivalent to a faster pulse input. These rapid pulses result in a larger conductance and induce abrupt potentiation. This indicates that the pulse oscillation has an impact on nonlinearity. For depression, the pulse shown in Figure S4 is transmitted to the device. Like potentiation, after initialization, the conductance is intentionally varied to take values in the range of 100–350 μS, as shown in Figure 6 a. Herein, it is only read to suppress dramatic changes, utilizing short-term memory characteristics that are gradually forgotten over time. All the conductance noticeably decreases from 100 ms and become negligibly small after 1 ms. During the period from 10 s to 14 s, these conductance values nearly return to the initial state. Based on this, the decreased conductance can be defined by modulating the interval time between read pulses. The cells are randomly selected and averaged for both potentiation ( Figure 7 a) and depression ( Figure 7 b). Unlike the previous approach, the impact of pulse width was investigated. As observed in Figure 7 a, an expansion in pulse width corresponds to an increase in both nonlinearity and conductance. This suggests that the pulse width amplifies the change required to reach the next level. A larger conductance demands a longer depression time, as explicitly indicated in Figure 7 b. Finally, Figure 7 c shows the endurance response of Pt/BN/TiN RRAM devices at the HRS and LRS states after the repetitive application of pulse trains comprising 8 V, 1.5 V, −3 V, and 1.5 V with pulse widths of 10 μs. Throughout the period spanning 10,000 cycles, the HRS and LRS are maintained at values equal to or greater than 3.73. The summary of different types of BN and their applications in inorganic materials is provided in Table 1 [ 47 , 48 , 49 , 50 , 51 ]. Furthermore, to assess the response speed of the devices, we employed SET pulses with a width of 5 ms and amplitudes of 5 V to initiate the resistance state changes, as depicted in Figure S5 . Read pulses of 5 ms/3 V were applied both before and after the SET pulses to monitor the resistance switching. Using the pulse pair of 5 ms/5 V, the Pt/BN/TiN device could switch to the LRS within 52.34 µs, while consuming 11.34 nJ of energy during the SET process. This can be calculated by using the formula: (2) W = V × I × T \nwhere V is the applied voltage of the pulse, I is the response current, and T is the response time, respectively [ 52 ]. Potentiation and depression are key elements of the software MNIST in artificial neural networks [ 53 , 54 ]. Handwritten numbers, which are sometimes difficult to distinguish, even for humans, cannot be easily recognized by computers. The weight update implemented by the software MNIST and the RRAM hardware make this task even more challenging. When a binary image is input, the system outputs a prediction based on training using the stored dataset for numbers ranging from zero to nine. Each image is represented as a 28 × 28 matrix, where colors closer to white correspond to values closer to 255 in Figure 8 a while the color black part converges to zero. These results are used to update the conductance of the RRAM. Finally, the expected value is output following calculations. The potentiation and depression from random cells are depicted in Figure 8 b with α being 0.88 and 0.36, respectively. An epoch refers to the number of times the 60,000 images are repeated. When implemented 10 times, the Pt/BN/TiN device achieves a high accuracy equal to 94.96%."
} | 4,049 |
38385710 | PMC11326117 | pmc | 406 | {
"abstract": "ABSTRACT Global warming is a key issue that causes coral bleaching mainly because of the thermosensitivity of zooxanthellae. Compared with the well-studied zooxanthellae Symbiodiniaceae in coral holobionts, we rarely know about other coral symbiotic algae, let alone their thermal tolerance. In this study, a zoochlorellae, Symbiochlorum hainanensis, isolated from the coral Porites lutea, was proven to have a threshold temperature of 38°C. Meanwhile, unique high-temperature tolerance mechanisms were suggested by integrated transcriptomics and real-time quantitative PCR, physiological and biochemical analyses, and electron microscopy observation. Under heat stress, S. hainanensis shared some similar response strategies with zooxanthellae Effrenium sp., such as increased ascorbate peroxidase, glutathione peroxidase, superoxide dismutase activities and chlorophyll a, thiamine, and thiamine phosphate contents. In particular, more chloroplast internal layered structure, increased CAT activity, enhanced selenate reduction, and thylakoid assembly pathways were highlighted for S. hainanensis ’s high-temperature tolerance. Notably, it is the first time to reveal a whole selenate reduction pathway from SeO 4 2− to Se 2− and its contribution to the high-temperature tolerance of S. hainanensis . These unique mechanisms, including antioxidation and maintaining photosynthesis homeostasis, efficiently ensure the high-temperature tolerance of S. hainanensis than Effrenium sp. Compared with the thermosensitivity of coral symbiotic zooxanthellae Symbiodiniaceae, this study provides novel insights into the high-temperature tolerance mechanisms of coral symbiotic zoochlorellae S. hainanensis , which will contribute to corals’ survival in the warming oceans caused by global climate change. IMPORTANCE The increasing ocean temperature above 31°C–32°C might trigger a breakdown of the coral-Symbiodiniaceae symbioses or coral bleaching because of the thermosensitivity of Symbiodiniaceae; therefore, the exploration of alternative coral symbiotic algae with high-temperature tolerance is important for the corals’ protection under warming oceans. This study proves that zoochlorellae Symbiochlorum hainanensis can tolerate 38°C, which is the highest temperature tolerance known for coral symbiotic algae to date, with unique high-temperature tolerance mechanisms. Particularly, for the first time, an internal selenium antioxidant mechanism of coral symbiotic S. hainanensis to high temperature was suggested.",
"conclusion": "Conclusions Compared with the thermosensitive zooxanthellae Effrenium sp. (threshold temperature: 32°C), zoochlorellae S. hainanensis has a heat-resistant temperature of 38°C, which is the highest thermal tolerance of coral symbiotic algae. Besides the similar heat response strategies as Effrenium sp., e.g., increased APX, GPX, and SOD activities and chlorophyll a, thiamine, and thiamine phosphates’ contents, S. hainanensis has unique high-temperature tolerance mechanisms, including more chloroplast internal layered structure, increased CAT activity, and enhanced selenate reduction and thylakoid assembly pathways. Particularly, for the first time, an internal selenium antioxidant mechanism of coral symbiotic S. hainanensis to high temperature was suggested. The revealed unique high-temperature tolerance mechanisms of zoochlorellae S. hainanensis efficiently remove ROS to maintain the low-level inner cellular superoxide (O 2 •− ) content and ensure photosynthesis homeostasis. The revealed 38°C high-temperature tolerance and the related molecular mechanisms of S. hainanensis greatly expand our understanding of the heat resistance of coral symbiotic algae.",
"introduction": "INTRODUCTION Coral reef ecosystems, known as the rainforest of the oceans, are suffering a dramatic worldwide decline because of global climate change, e.g., ocean warming ( 1 , 2 ). The well-studied coral symbiotic photosynthetic algae are dinoflagellates (Symbiodiniaceae, also known as zooxanthellae), which are generally sensitive to thermal stress with a lower threshold temperature ca. 31°C–32°C ( 3 ). It is for this reason the thermal response of coral symbiotic Symbiodiniaceae has been investigated ( 3 – 9 ). For instance, under elevated temperatures, chlorophyll synthesis in Breviolum sp. B1, Cladocopium goreaui C1, and Durusdinium trenchii D1a is upregulated ( 4 ), and the heat shock proteins (HSPs) display different expression changes among Cladocopium (clade C1) ( 5 , 6 ), Cladocopium (clade C3K) ( 7 ), Durusdinium (clade D1) ( 5 ), Durusdinium (clade D2) ( 6 ), and Fugacium (clade F) ( 3 ). In addition, increased ascorbate peroxidase (APX) activity was detected in Fugacium (clade F1) at 33°C ( 8 ). Upregulated expression of genes encoding glutathione peroxidase (GPX), peroxiredoxin (Prdx), superoxide dismutase (SOD), and its different metalloforms are upregulated under heat stress ( 6 , 9 ). These reports provide a preliminary understanding of the thermal response of coral symbiotic zooxanthellae Symbiodiniaceae and suggest that hosting high-temperature tolerant symbiotic algae is a valid strategy for corals to survive under higher temperatures ( 10 ). The thermal tolerance of coral symbiotic algae is very important for the health and survival of coral holobionts in the warming oceans. Studies have found that some Symbiodiniaceae types have relatively higher thermal tolerance, e.g., Symbiodinium thermophilum ( 11 ) and Durusdinium trenchii ( 10 , 12 ), but totally, coral symbiotic zooxanthellae Symbiodiniaceae is sensitive to thermal stress with lower threshold temperature, i.e., 31°C–32°C. The increasing ocean temperature above 32°C might trigger a breakdown of the coral-Symbiodiniaceae symbioses ( 13 ) or coral bleaching caused by thermosensitive Symbiodiniaceae’s escape or hypopigmentation ( 1 , 2 , 10 , 14 ); therefore, the exploration of other kinds of coral symbiotic algae with high-temperature tolerance is important for the corals’ protection under warming oceans. Besides zooxanthellae Symbiodiniaceae, corals host other kinds of symbiotic photosynthetic algal symbionts, e.g ., Ostreobium ( 15 ). In 2018, a zoochlorellae, Symbiochlorum hainanensis (Chlorophyta, Ulvophyceae), was first isolated from the bleached scleractinian coral Porites lutea in the South China Sea and named by us ( 16 ). Meanwhile, we found that S. hainanensis was wildly distributed in scleractinian corals Platygyra lamellina , Porites lutea, and Favia speciosa . Particularly, the abundance of S. hainanensis became higher when these corals were bleached under thermal stress, accompanied by an abundant decrease in coral symbiotic Symbiodiniaceae ( 17 ). This phenomenon indicates the possible roles of S. hainanensis in maintaining the coral holobionts’ health under warming oceans by replacing zooxanthellae Symbiodiniaceae. Zoochlorellae has been found to be able to enhance the acclimation capacity of green hydra under heat stress ( 18 ), but we rarely know about the response of coral zoochlorellae to thermal stress compared to coral zooxanthellae. In 2020, we proved that S. hainanensis could maintain growth at 32°C, which is generally a lethal temperature to most Symbiodiniaceae ( 19 ). Thus, the high-temperature tolerance of zoochlorellae S. hainanensis arouses our great interest, and it is hypothesized that it probably has unique high-temperature tolerant mechanisms that are different from zooxanthellae Symbiodiniaceae. In this study, unique high-temperature tolerance mechanisms of coral symbiotic Symbiochlorum hainanensis to a high temperature of 38°C were predicted by transcriptomics first and then verified by real-time quantitative PCR (RT-qPCR) and physiological and biochemical analysis along with electron microscopy observation using thermosensitive zooxanthellae Effrenium sp. as a control.",
"discussion": "DISCUSSION Based on the thermal response of coral Symbiodiniaceae ( 3 , 5 – 9 ) and Effrenium sp., in this study, S. hainanensis has the highest temperature tolerance known for coral symbiotic algae, i.e., 38°C. The expression of HSPs is commonly considered to be associated with stress ( 20 ), HSP-related DEGs in zoochlorellae S. hainanensis and zooxanthellae Effrenium sp. under thermal stress indicate that algal cells were indeed in a stress response state ( Fig. 5 ). HSPs are known to have function in protein processing, such as protein folding, protein translocation, and maintaining the conformation of unstable/wrong-folded proteins, as well as their signaling functions ( 20 ). In S. hainanensis , the upregulation of genes encoding HSPs, especially small HSPs, HSP70 and HSP90, indicated that the heat response system was triggered, and the alga was conserving its protein homeostasis. In contrast to S. hainanensis , a few HSPs were upregulated in Effrenium sp., indicating that some of its protein homeostasis maintaining mechanism might be damaged or dysfunctional under thermal stress, which would result in the destruction of protein homeostasis inside algal cells. However, the unchanged O 2 •− content and decreased H 2 O 2 content in S. hainanensis indicated its higher ability to remove reactive oxygen species (ROS) than Effrenium sp. ( Fig. 3A through F ); this is probably one of the reasons why S. hainanensis can survive under the extreme high temperature of 38°C. Compared with Effrenium sp., the presence of CAT, the remaining activity of Prdx, and the reducing product of the selenate reduction pathway (Se 2− ) could be the mechanisms that contribute to the enhanced ROS removal capacity in S. hainanensis . The upregulated DEGs related to antioxidases, selenate reduction, thiamine biosynthesis, chlorophyll a synthesis, and thylakoid assembly in S. hainanensis X1 are summarized in Fig. 7 , showing the unique thermal resistance mechanisms of S. hainanensis . Considering the wide distribution and increased abundance of S. hainanensis in bleaching corals ( 16 , 17 , 19 , 21 ), S. hainanensis might play important roles in corals’ resistance to thermal stress, particularly when thermosensitive zooxanthellae escape in the warming oceans caused by global climate change. Meanwhile, the transplanting of S. hainanensis might be a strategy to help corals survive in warming oceans caused by global climate change because of its high thermal tolerance. Fig 7 Schematic summary of the upregulated DEGs in S. hainanensis X1 response to thermal stress. DEGs were selected as adjusted P value < 0.01 and log2 (fold change) ≥ ±1 in thermal stressed groups (X1W1 group, 32°C and X1W2 group, 38°C) compared to the control (X1C1 group, 26°C) on the third day. Genes related to antioxidases, selenate reduction, thiamine biosynthesis, chlorophyll a synthesis, and thylakoid formation are shown. Solid and dotted arrows represent one-step or multi-step reactions, respectively. Italics represent genes. The semicircle indicates the change of gene relative expression or substance content in algae cells: left semicircle represents S. hainanensis X1, and right semicircle represents Effrenium sp. S1; color in the semicircle indicates the change trend: red represents upregulated or increased, blue represents downregulated or decreased, and white represents no significant chang. The dotted black border represents that this gene or substance was not detected. The specific contribution of antioxidant enzymes, particularly CAT, to the high-temperature tolerance of S. hainanensis Elevated temperature not only affects temperature-dependent biochemical reactions but also increases intracellular oxidative pressure ( 22 ). In this study, the MDA content change may reflect the oxidative pressure particularly under the ultimate temperatures, i.e., 38°C and 32°C, for S. hainanensis and Effrenium sp., respectively ( Fig. 3C and F ). The increase in ROS such as O 2 •− and H 2 O 2 under heat stress will cause algal oxidative damage ( 22 ). Hence, the scavenging of ROS will contribute to heat resistance. The O 2 •− content in S. hainanensis under different experiment temperatures remained low, and the H 2 O 2 content under the extreme temperature significantly decreased ( Fig. 3A and B ). On the contrary, O 2 •− and H 2 O 2 contents in Effrenium sp. increased ( Fig. 3D and E ), indicating that its ROS removal capacity was weakened under thermal stress. The lower O 2 •− and H 2 O 2 contents in S. hainanensis indicate its higher ability to relieve oxidative pressure than Effrenium sp. The expression of genes encoding antioxidant enzymes SOD, APX, CAT, and GPX in S. hainanensis was upregulated under thermal stress ( Fig. 5A ); consequently, the activities of SOD, APX, CAT, and GPX increased ( Fig. 3G through K ). It is known that antioxidases are able to transfer O 2 •− to H 2 O 2 and finally to H 2 O ( 22 ); therefore, ROS inside the algal cells is maintained at a low level. However, the different activity changes of these four enzymes indicated their different contributions to the heat resistance of S. hainanensis . Specifically, the GPX and SOD activities were higher when the algal cells were heated (32°C and 38°C), but the APX and CAT activities were only higher under an extremely high temperature (38°C). It can be speculated that GPX and SOD are the core antioxidant enzymes in the heat resistance of S. hainanensis , and APX and CAT are reserves, which can only be called under the extremely high temperature. The increased intercellular SOD activity in Chlorella ellopsoidea ( 23 ), Breviolum (clade B1), and Cladocopium (clade C1) ( 8 ) indicates SOD’s role in the algal response to heat stress. In Chlamydomonas reinhardtii , APX activity was found to be increased under elevated temperatures ( 24 ). Although transcripts encoding antioxidant enzymes like APX and SOD displayed a decreasing trend in Effrenium sp. under thermal stress, their expression changes did not reach a significant level. Although catalase peroxidase (KatG) has been found in Breviolum (clade B1) ( 25 ), the capacity of H 2 O 2 degradation in Breviolum (clade B1) and Effrenium (clade E1) displayed no significant change under thermal stress ( 25 , 26 ). Accordingly, it is speculated that different coral symbiotic algae have different response patterns of antioxidant enzymes to thermal stress. In particular, our results suggested the importance of antioxidase CAT in the high-temperature tolerance of S. hainanensis because no CAT activity was detected in Effrenium sp. Based on the result from Bayer et al. ( 27 ), Symbiodinium sp. CassKB8 and Breviolum sp. Mf1.05b appear to lack CAT ( 23 ), and no CAT activity change was detected in dinoflagellate Cladocopium goreaui during the thermal exposure period ( 28 ). Thus, CAT might lead to a much more effective antioxidant system in S. hainanensis and aid its higher tolerance than Effrenium sp. Selenate reduction and thiamine biosynthesis related to the high-temperature tolerance of S. hainanensis In S. hainanensis , besides the roles of multiple antioxidant enzymes, the enhancement of selenate reduction and thiamine biosynthesis pathways probably contributes to the tolerance of S. hainanensis to high temperatures ( Fig. 7 ). It is worth mentioning that a whole pathway of selenate reduction was detected in S. hainanensis ( Fig. 6E ), which was correlated with this algal high-temperature tolerance. In this pathway, SeO 4 2− is successively catalyzed to Se 2− by sulfate adenylyltransferase (EC 2.7.7.4, encoded by MET3 ) and thioredoxin reductase (EC 1.8.1.9, encoded by TRR1 ). The upregulated expression of two genes MET3 and TRR1 ( Fig. 5A and 6A ) and the increased content of Se 2− in S. hainanensis cells ( Fig. 6B ) were detected under heat stress in this study. As a result of this upregulation, theoretically, the content of substrate (SeO 4 2− ) and intermediate (SeO 3 2− ) should be reduced, which was supported by the decreased contents of SeO 4 2− and SeO 3 2− and the increase of Se 2− in the algal cells. In contrast to S. hainanensis , there is no significant content change in SeO 3 2− or Se 2− in Effrenium sp. cells ( Fig. 6H ), as well as no significant change in the related genes’ expression ( Fig. 5B ), indicating this selenate reduction pathway does not contribute to the heat resistance of Effrenium sp. Selenium has been reported to play a key role in the cellular antioxidant defense mechanism ( 29 ). For example, Se at low concentration positively promoted the antioxidative effect of Chlorella pyrenoidosa by increasing the levels of glutathione peroxidase, catalase, linolenic acid, and photosynthetic pigments ( 30 ) and increased the activity of antioxidant enzymes (SOD and CAT) and the amount of antioxidant metabolites (phenols, flavonoids, and carotenoids) in Ulva sp. ( 31 ). Maronić et al. ( 32 ) also highlighted the importance of the algal Se detoxification strategy, especially the role of selenoenzymes and other selenoproteins with antioxidant function. Similarly, based on the upregulation of specific genes and the increased Se 2− yield concentration under heat stress, this study suggests an internal selenium antioxidant mechanism of S. hainanensis to high temperature. Taken together the present knowledge of the thermal response mechanisms of well-studied Chlamydomonas ( 33 ), coral symbiotic Symbiodiniaceae ( 3 – 9 ), and the thermal response of Effrenium sp. in this study, it is the first time to find a correlation between the upregulated selenate reduction pathway and high-temperature tolerance of coral symbiotic algae, which could be one of the reasons why its antioxidant system is more effective than Effrenium sp. Tunc-Ozdemir et al. ( 34 ) found the role of thiamine in the protection of cells against oxidative damage in Arabidopsis thaliana and found that thiamine-induced tolerance to oxidative stress was accompanied by decreased production of reactive oxygen species, as evidenced from decreased protein carbonylation and hydrogen peroxide accumulation. In this study, the expression of six genes ( MET17 , iscS , dxs , TH2 , thiN, and adk ) responsible for synthesizing thiamine and its three phosphates was upregulated in S. hainanensis under heat stress ( Fig. 5A , 6C, D, and F ), which was proved by the increased contents of thiamine, TMP, TDP, and TTP under heat stress. Prior studies have noted that the antioxidant/anti-heat function of thiamine and TDP is common in algae and plants, such as the cyanobacterium Nodularia spumigena , dinoflagellate Prorocentrum minimum ( 35 ), Zea mays, and Arabidopsis thaliana ( 36 ). Thus, thiamine biosynthesis could contribute to thermal tolerance in S. hainanensis by increasing the content of its antioxidant products thiamine and TDP. Combined with the similar increase of thiamine and TDP in thermally stressed Effrenium sp., it could be proposed that this is a universal mechanism in the thermal stress response of coral symbiotic algae. The increased TTP of algae under thermal stress ( Fig. 6D ) suggests TTP’s possible correlation with thermal tolerance. However, to date, we rarely know about the function of TTP in stress response, except that it was suggested to serve as “alarmones” when cells are under starvation ( 37 ). Thus, TTP’s roles in the antioxidation of S. hainanensis need further study. Maintenance of photosynthesis homeostasis by enhancing thylakoid assembly for the high-temperature tolerance of S. hainanensis Based on the KEGG and GO enrichment analyses, in addition to antioxidation, photosynthesis homeostasis maintenance might be another contributor to the high-temperature tolerance of S. hainanensis . Under heat stress, the expression of seven genes ( ChlD , ChlH , ChlI , CHLM , CTH1 , HEMC, and PPOX ) involved in chlorophyll a biosynthesis was upregulated ( Fig. 5A ). Consistent with this result, the chlorophyll a content in S. hainanensis cells increased ( Fig. 1C ). The enhanced chlorophyll a biosynthesis indicated that S. hainanensis was compensating for heat-induced chlorophyll loss or increasing the energy inflow under heat stress. In addition, the increased thylakoid formation was found to be involved in the response of S. hainanensis to heat stress ( Fig. 7 ). Five genes ( ALB3.2 , SQD1 , TatA , Thf1, and VIPP1 ) associated with thylakoid formation were upregulated in S. hainanensis under heat stress ( Fig. 5A ). These genes are involved in thylakoid membrane lipid synthesis ( SQD1 ), thylakoid membrane protein synthesis ( TatA ), integration of light-harvesting complex into thylakoid membrane ( ALB3.2 ), as well as thylakoid assembly and stacking ( Thf1 and VIPP1 ). In contrast, the increased thylakoid assembly was not observed in Effrenium sp. ( Fig. 2G through L , Fig. 5B ), indicating the possible damage to photosynthesis homeostasis of this alga under heat stress. The internal layers of chloroplasts in S. hainanensis cells under higher temperatures ( Fig. 2D and F ) became more abundant compared with the control ( Fig. 2B ), whereas there was no significant change in internal layers in Effrenium sp. chloroplasts under heat stress. Thus, it can be speculated that S. hainanensis probably increase the assembly or the formation of thylakoids under thermal stress ( Fig. 7 ). Presumably, it might be compensating for the thylakoid losses caused by heat or forming new thylakoid de novo to maintain photosynthesis energy inflow in S. hainanensis . Similar to our results, the formation of aberrant prolamellar body-like structures was observed in the chloroplast of heat-tolerating Chlamydomonas reinhardtii under elevated temperature, which is considered to be associated with photosynthesis maintenance ( 38 ). Similarly, the enlargement of chloroplasts along with the increase in chlorophyll fluorescence and pigment content of S. hainanensis were detected in our previous study ( 19 ). Coupled with the morphologic change of chloroplasts in both the 32°C and 38°C groups ( Fig. 2 ), it is presumably suggested that S. hainanensis probably try to maintain the photosynthesis homeostasis by increasing the assembly of thylakoid and more chloroplast internal layered structure. Conclusions Compared with the thermosensitive zooxanthellae Effrenium sp. (threshold temperature: 32°C), zoochlorellae S. hainanensis has a heat-resistant temperature of 38°C, which is the highest thermal tolerance of coral symbiotic algae. Besides the similar heat response strategies as Effrenium sp., e.g., increased APX, GPX, and SOD activities and chlorophyll a, thiamine, and thiamine phosphates’ contents, S. hainanensis has unique high-temperature tolerance mechanisms, including more chloroplast internal layered structure, increased CAT activity, and enhanced selenate reduction and thylakoid assembly pathways. Particularly, for the first time, an internal selenium antioxidant mechanism of coral symbiotic S. hainanensis to high temperature was suggested. The revealed unique high-temperature tolerance mechanisms of zoochlorellae S. hainanensis efficiently remove ROS to maintain the low-level inner cellular superoxide (O 2 •− ) content and ensure photosynthesis homeostasis. The revealed 38°C high-temperature tolerance and the related molecular mechanisms of S. hainanensis greatly expand our understanding of the heat resistance of coral symbiotic algae."
} | 5,896 |
33841761 | PMC8019053 | pmc | 407 | {
"abstract": "Abstract \n Plants typically interact with multiple above‐ and below‐ground organisms simultaneously, with their symbiotic relationships spanning a continuum ranging from mutualism, such as with arbuscular mycorrhizal fungi (AMF), to parasitism, including symbioses with plant‐parasitic nematodes (PPN). Although research is revealing the patterns of plant resource allocation to mutualistic AMF partners under different host and environmental constraints, the root ecosystem, with multiple competing symbionts, is often ignored. Such competition is likely to heavily influence resource allocation to symbionts. Here, we outline and discuss the competition between AMF and PPN for the finite supply of host plant resources, highlighting the need for a more holistic understanding of the influence of below‐ground interactions on plant resource allocation. Based on recent developments in our understanding of other symbiotic systems such as legume–rhizobia and AMF‐aphid‐plant, we propose hypotheses for the distribution of plant resources between contrasting below‐ground symbionts and how such competition may affect the host. We identify relevant knowledge gaps at the physiological and molecular scales which, if resolved, will improve our understanding of the true ecological significance and potential future exploitation of AMF‐PPN‐plant interactions in order to optimize plant growth. To resolve these outstanding knowledge gaps, we propose the application of well‐established methods in isotope tracing and nutrient budgeting to monitor the movement of nutrients between symbionts. By combining these approaches with novel time of arrival experiments and experimental systems involving multiple plant hosts interlinked by common mycelial networks, it may be possible to reveal the impact of multiple, simultaneous colonizations by competing symbionts on carbon and nutrient flows across ecologically important scales.",
"conclusion": "4 CONCLUSION Plant roots are integral to the concept of an ecosystem, existing as critical components of below‐ground systems that underpin those above‐ground, rather than as single entities interacting with individual symbionts. Within this system, the complex interactions between multiple below‐ground symbionts in competition for plant resources remain critically under‐explored, resulting in a number of important, outstanding research questions such as: how is allocation of plant resources regulated in the AMF‐PPN‐host system? Does regulation occur at the site of competition, systemically within the single host, or is it modulated by the common mycelial network across multiple plants? By addressing these, a more holistic understanding of contrasting tripartite below‐ground interactions may be acquired that may allow us to evaluate the roles of AMF in the field and the influence on plant community, and subsequently ecosystem, structure, and function. Tracking the movement of resources in plants with different combinations of root symbionts will help to determine the consequences of PPN infection for plant–AMF associations, potentially also helping to explain the disparity between the species‐specific effects of different AMF and nematode species on AMF biocontrol efficacy and the role(s) of common mycelial networks in agro‐ecosystems. The mutualistic AMF and parasitic PPN symbioses discussed here represent the granularity of the below‐ground ecosystem. There are a huge number of other organisms that interact at the root–soil interface, with each component of the soil community impacting each other either directly or indirectly. As there are finite plant resources available that the below‐ground community is in competition for, particularly in agri‐ecosystems, the influence of such interactions on plant resource allocation must be understood in order that it may be managed to promote beneficial interactions and optimum plant growth conditions. By investigating such granularities of the below‐ground ecosystem, future research will provide a platform from which to begin to unpick and understand other complex, multitrophic interactions that exist in nature and thereby promote a fuller, more holistic understanding of the rhizosphere.",
"introduction": "1 INTRODUCTION In nature, plants engage in a variety of complex below‐ground symbioses spanning the mutualistic relationships formed with the near‐ubiquitous arbuscular mycorrhizal fungi (AMF), through to parasitic interactions (Johnson et al., 1997 ) with pathogenic organisms such as plant‐parasitic nematodes (PPN). These interactions seldom occur in isolation, for instance more than half (51%) of the plant species that play host to PPN are also colonized by AMF, highlighting the prevalence of this tripartite interaction in natural and agro‐ecosystems (FungalRoot [Soudzilovskaia et al., 2020 ] and Nemabase [Ferris, 2020 ]). A wealth of other organisms interacts concurrently at the plant:soil interface, such as fungi, oomycetes, bacteria, and nematodes, and exert deleterious or beneficial effects on the plant. These interactions can include beneficial nitrogen fixation (Jacoby et al., 2017 ) and mineralization of nutrients from the soil for plant uptake (Richardson et al., 2009 ), whilst others can result in plant disease (Raaijmakers et al., 2009 ). This greatly affects the structure and function of the soil community, with important, economical implications for crops. In agro‐ecosystems, AMF and PPN both form complex root‐symbiont interfaces that facilitate the transfer of nutrients and will invariably compete for the finite supply of host plant resources in the form of photosynthetically fixed carbon‐based molecules. The same compounds (such as glucose, fructose, and galactose) are often directly acquired by both symbionts (Helber et al., 2011 ; Rodiuc et al., 2014 ). However, the mechanisms underpinning resource allocation between competing root symbionts in AMF‐PPN‐plant associations remain unresolved. Plant parasitism has evolved independently at least four times within the phylum Nematoda (Kikuchi et al., 2017 ), and there are >4,100 species of PPN from ectoparasitic, semi‐endoparasitic, and endoparasitic groups (Jones et al., 2013 ). The most damaging are the sedentary endoparasites, which have developed similar, intricate lifestyles to maximize assimilation of plant resources (Bird et al., 2015 ). The phytophagous ability of PPN has made them one of the four most economically important plant pathogens (Jones et al., 2013 ), estimated to result in crop yield losses of >US$80 billion per annum (Nicol et al., 2011 ). Due to their economic importance on key crops worldwide, the majority of research to date has focused on the sedentary root‐knot and cyst nematode species, which we discuss here in relation to their impact on agro‐ecosystems. AMF, on the other hand, typically confer a variety of potential growth‐enhancing benefits on their plant host (Smith & Smith, 2011 ). Nutrients that are either beyond the root depletion zone or in an inaccessible form to plant roots may be acquired by AMF and translocated from the soil to host plant roots via an extensive mycelial network, in some cases providing the plant with up to 90% of their phosphorus requirements whilst also contributing toward plant nitrogen and micronutrient needs (Ezawa & Saito, 2018 ; Field & Pressel, 2018 ; van der Heijden et al., 2015 ). The extent of the benefits conferred on plants by AMF is known to vary considerably according to the identities of plant and fungus, as well as various abiotic factors (Field & Pressel, 2018 ; Johnson et al., 1997 ). Therefore, it is critical that the impact(s) of environmental factors, including biotic and abiotic, are taken into account when symbiotic function and degree of plant benefits obtained through symbiosis with AMF are considered. As obligate biotrophs, both PPN and AMF rely on plant‐fixed photoassimilates for nutrition. However, unlike AMF, PPN offer no benefits to their host plant. Instead, PPN typically confer severe impediments on plant growth and productivity (Jones et al., 2013 ). Despite simultaneous associations with AMF and PPN being common in both agro‐ and natural ecosystems, relatively little is known about how plants cope with these competing root symbionts, especially regarding the allocation of plant resources in crop plants. Here, we do not attempt to review AMF‐PPN‐host interactions in their entirety, rather we highlight the currently ill‐defined key factors that may regulate the distribution of host plant resources to the competing symbionts. To instigate new ideas and research into this area, we hypothesize alternate scenarios for nutrient allocation within tripartite interactions between plants, AMF, and PPN and the potential outcomes for each. By defining this area, a more holistic understanding of complex nutrient allocation at the plant–soil interface may be reached, with considerations of impacts on plant defenses and the potential exploitation of AMF as biocontrol agents to enhance sustainable agro‐ecosystems."
} | 2,262 |
23981909 | PMC3755279 | pmc | 408 | {
"abstract": "Despite extensive progress, current icephobic materials are limited by the breakdown of their icephobicity in the condensation frosting environment. In particular, the frost formation over the entire surface is inevitable as a result of undesired inter-droplet freezing wave propagation initiated by the sample edges. Moreover, the frost formation directly results in an increased frost adhesion, posing severe challenges for the subsequent defrosting process. Here, we report a hierarchical surface which allows for interdroplet freezing wave propagation suppression and efficient frost removal. The enhanced performances are mainly owing to the activation of the microscale edge effect in the hierarchical surface, which increases the energy barrier for ice bridging as well as engendering the liquid lubrication during the defrosting process. We believe the concept of harnessing the surface morphology to achieve superior performances in two opposite phase transition processes might shed new light on the development of novel materials for various applications.",
"discussion": "Discussion Previously we discussed the influence of micro/nanoscale roughness on the delay of frost formation and inter-droplet freezing wave propagation. In particular, we demonstrated that the controlled-introduction of microscale structures with inclined edges not only reinforces spontaneous condensate droplet departure behavior for delaying frost growth, but also increases additional structural barrier for the ice bridging, both of which contribute to the suppression of inter-droplet freezing wave propagation. Note that these spatially controlled micro-edges are distinctively different from those large sample edges defining the geometry boundary of our samples which are associated with low heterogeneous ice nucleation energy barrier owing to their geometric singularity. In contrast, the patterned micro-edges dramatically impede individual frost formation and inter-droplet freezing wave propagation through a continuous process of droplet nucleation, coalescence, departure, and/or evaporation. To explain the efficient frost removal on the hierarchical surface, we quantified the variation of fracture density over time. The fracture density is defined as the number of visible fractures per unit area during the defrosting process. Among three samples studied, the flat hydrophobic surface displays the maximum fracture density during the defrosting process ( Fig. 5a ). Owing to the small contact angle (~110°), the flat hydrophobic surface yields a strong affinity to the frozen ice and melting liquid, and accordingly the frozen ice and melting liquid are tightly anchored at the bottom surface with a strong adhesion (step 1a–3a in Fig. 5b ). In contrast, the frost at the upper layers has a relatively weak interaction. Thus, it is expected that the large difference in interaction at the bottom and upper layers leads to the emergence of pronounced fractures upon defrosting. In contrast, the incorporation of nanosacale roughness in the nanograssed or hierarchical surfaces contributes to the formation of “Cassie-droplet” during the defrosting process (step 1b–3b, 1c–3c in Fig. 5b ), resulting in the formation of interconnected network without encountering the severe fracture as opposed to that on the hydrophobic surface. Notably, the incorporation of 3-D microscale edge structures aids the mobility of the melting underneath bulk frost layer in both lateral and vertical directions (step 1c–2c in Fig. 5b ). Moreover, coupled with the minimal frost adhesion to the substrate enabled by the trapped air pockets in the nanoscale roughness, the enhanced mobilization allows for the preservation of the integrity of the whole frost layer during the defrosting process. As a result, the melting droplets at the micro/nanostructured interface serve as liquid lubricant, accelerating the efficient removal of frost layer as described above (step 3c in Fig. 5b ). Whereas without the 3-D structure and two-tier roughness, the melting water droplets on the nanograssed surface retain with spherical shapes, and don't slip away from the surface (step 3b in Fig. 5b ). We expect the efficient frost removal on the hierarchical surface endowed by the synergic cooperation of micro/nano-scale roughness and the microscale edge effect could dramatically lower the energy cost associated with the defrosting. In summary, we developed a robust icephobic surface that allows for enhanced frost formation retardation through the suppression of inter-droplet freezing propagation over the entire surface, as well as efficient frost removal by self-lubrication. Our work investigates both frosting and defrosting processes on engineered icephobic surfaces in an integrated approach. We demonstrate that these improved performances are ascribed to the microscale edge effect and the synergistic cooperation of micro/nano-scale roughness. In particular, we find that the spatial control of microscale edges in the hierarchical surface not only increases the energy and structural barriers for the interdroplet freezing wave propagation through a continuous process of droplet nucleation, coalescence, departure and/or evaporation, but also enhances the bulk frost mobility during the defrosting stage. We envision that the concept of harnessing surface morphology to achieve superior performances in two opposite phase transition processes (frosting/defrosting) might open up a new avenue for the development of efficient materials for various applications ranging from anti-icing, dropwise condensation and water harvesting."
} | 1,401 |
36332017 | PMC9635822 | pmc | 409 | {
"abstract": "Engineered living materials (ELMs) are gaining traction among synthetic biologists, as their emergent properties and nonequilibrium thermodynamics make them markedly different from traditional materials. However, the aspiration to directly use living cells as building blocks to create higher-order structures or materials, with no need for chemical modification, remains elusive to synthetic biologists. Here, we report a strategy that enables the assembly of engineered Saccharomyces cerevisiae into self-propagating ELMs via ultrahigh-affinity protein/protein interactions. These yeast cells have been genetically engineered to display the protein pairs SpyTag/SpyCatcher or CL7/Im7 on their surfaces, which enable their assembly into multicellular structures capable of further growth and proliferation. The assembly process can be controlled precisely via optical tweezers or microfluidics. Moreover, incorporation of functional motifs such as super uranyl-binding protein and mussel foot proteins via genetic programming rendered these materials suitable for uranium extraction from seawater and bioadhesion, respectively, pointing to their potential in chemical separation and biomedical applications.",
"introduction": "INTRODUCTION Biological tissues or organs are remarkable multicellular materials that can self-grow, regenerate, evolve, and adapt. De novo design of these “living” materials has been unthinkable within traditional materials science. It is until recently that the convergence of synthetic biology and materials science has made it possible to create some engineered living materials (ELMs) that can recapitulate the natural ones to some extent ( 1 , 2 ). One typical example is to genetically or chemically reengineer some naturally occurring soft biomaterials such as bacterial biofilms ( 1 , 2 ). However, this approach has largely focused on the self-assembling biofilm proteins at the molecular level while exerting little control over the organization of cells. Although several other studies have demonstrated the controlled assembly of eukaryotic cells (e.g., yeasts or mammalian cells), often with the help of exogenous polymers, into tissue-like structures via DNA hybridization or click chemistry ( 3 – 8 ), these processes involved extensive chemical modifications, which might pose uncertainties for cellular fitness and biofunctionality. As the long-range organization of cells has been the hallmark of multicellular tissues and organs in higher organisms, the use of engineered cells as building blocks to construct high-order structures, while circumventing chemical modification, would add a new dimension to the design of ELMs ( 9 , 10 ). “Ultrahigh-affinity” protein/protein interactions (PPIs) have often been used to assemble recombinant proteins into functional materials ( 11 – 17 ). For instance, SpyTag/SpyCatcher, a peptide/protein pair that is originally derived from a split bacterial adhesin domain, is capable of covalently stitching together protein molecules under mild physiological conditions, which has led to the creation of a number of uncommon protein architectures and materials ( Fig. 1A ) ( 18 , 19 ). Those noncovalent pairs, such as cohesin, X module/dockerin, the WW domain/its cognate proline-rich peptide, and the inactivated colicin E7 deoxyribonuclease/immunity protein 7 (CL7/Im7), can form strong physical interactions with dissociation constants ( K d ) at low picomolar or even femtomolar, which have contributed to the formation of injectable hydrogels with shear-thinning behavior ( 20 , 21 ). The recently developed CL7/Im7, with an ultrahigh affinity ( K d ~10 −14 to 10 −17 M), has also provided a means to purify delicate protein complexes from biological systems ( Fig. 1B ). For materials design and synthesis, these PPI motifs can be readily introduced because of their genetic encodability and modularity. We envisioned that not only can biomolecules be assembled into macroscopic materials but also cells can serve as building blocks to construct high-order living materials. To accomplish this, the key is to endow the cells with the ability to form stable and specific intercellular interactions, the latter of which can be realized through the combined use of suitable protein chemistries and robust cell surface display systems. Fig. 1. Schematic showing the assembly of multicellular living material enabled by ultrahigh-affinity PPIs. ( A ) Structure of the SpyTag/SpyCatcher complex and the formation of isopeptide bond [Protein Data Bank identifier (PDB ID): 4MLI]. ( B ) Structure of the CL7/Im7 complex (PDB ID: 7CEI). ( C ) Assembly of Saccharomyces cerevisiae cells. AGA1 and AGA2 proteins enable the display of the target proteins, such as SpyTag, SpyCatcher, CL7, and Im7, on the cell surface. Intercellular PPIs assemble individual cells into networks. Saccharomyces cerevisiae (baker’s yeast) cells have been noted for their asexual cellular aggregation at a high cell density, a natural phenomenon also known as flocculation, when cells adhere reversibly to each other via the physical interactions between surface glycoproteins, leading to the formation of macroscopic aggregates during culturing ( 22 ). Inspired by the yeast cell aggregation, in this study, we accomplished controlled assembly of single-celled S. cerevisiae into highly stable, multicellular ELMs via genetically programmed intercellular interactions. The cells were engineered to produce and display ultrastrong PPI motifs such as SpyCatcher/SpyTag or CL7/Im7. Further introduction of functional motifs such as metalloproteins and underwater adhesive domains into these single-celled components gave rise to the corresponding emergent properties at the macroscopic material level. This study illustrates a simple strategy for creating functional living materials, which holds great promise for real-world challenges ranging from chemical separation to wound closure.",
"discussion": "RESULTS AND DISCUSSION Yeast surface display of ultrahigh-affinity binding partners To achieve stable cell-to-cell conjugation, we chose two representative PPI partners—the covalent SpyTag/SpyCatcher that spontaneously forms a Lys-Asp isopeptide bond and the noncovalent, ultrahigh-affinity CL7/Im7 ( Fig. 1, A and B ). Yeast surface display was accomplished via the widely used AGA1/AGA2 system ( 23 ). The genes encoding SpyTag, SpyCatcher, Im7, and CL7 were cloned into the pCTcon2 vector and then transformed into the S. cerevisiae strain EBY100 ( 24 ). The resulting yeast cells, denominated as ScA , ScB , Sc Φ, and Sc Ψ, were expected to be able to produce and display the fusion proteins, AGA2-SpyTag, AGA2-SpyCatcher, AGA2-CL7, and AGA2-Im7, respectively, on their surfaces ( Fig. 1C and table S1). To confirm the expression and display of these proteins, we treated these cells, ScA , ScB , Sc Φ, and Sc Ψ, correspondingly with the fluorescent proteins, SpyCatcher-elastin-like-protein (ELP)–enhanced green fluorescent protein (EGFP)–ELP-SpyCatcher, SpyTag-ELP-EGFP-ELP-SpyTag, Im7-ELP-EGFP-ELP-Im7, and CL7-ELP-EGFP-ELP-CL7 (table S2). These engineered cells, but not the wild-type ones, were labeled with EGFP efficiently, showing successful protein expression and surface display ( Fig. 2 ). Fig. 2. Yeast surface display of interacting motifs confirmed by EGFP labeling. S. cerevisiae cells displaying SpyTag ( ScA ) ( A ), SpyCatcher ( ScB ) ( B ), CL7 ( Sc Φ) ( C ), and Im7 ( Sc Ψ) ( D ) were labeled with SpyCatcher-ELP-EGFP-ELP-SpyCatcher (B-GFP-B), SpyTag-ELP-EGFP-ELP-SpyTag (A-GFP-A), Im7-ELP-EGFP-ELP-Im7 (Ψ-EGFP-Ψ), and CL7-ELP-EGFP-ELP-CL7 (Φ-EGFP-Φ), respectively. The yeast cells that did not display the proteins served as control. Scale bars, 50 μm. Multicellular assembly at a high cell density Being widely used to assemble biomolecules into high-order structures ( 14 , 16 ), genetically programmable PPIs may well be used to assemble larger mesoscopic objects like living cells into ELMs. Because of their natural tendency to adhere to each other at a high cell density, S. cerevisiae cells provide an ideal platform for achieving PPI-mediated ultrastable multicellular assembly while circumventing diffusion barriers. To test this hypothesis, we simply mixed the matched cell types, ScA + ScB and Sc Φ + Sc Ψ, at an approximately 1:1 ratio and a cell density of ~2.0 × 10 8 per milliliter. To assess the cell-to-cell conjugation and also distinguish the two types of cells, the mixtures, ScA + ScB and Sc Φ + Sc Ψ, were treated sequentially with the fluorescent proteins, SpyCatcher-ELP-Dronpa/SpyTag-ELP-mCherry-ELP-SpyTag and Im7-ELP-mCherry/CL7-ELP-EGFP-ELP-CL7, respectively. To avoid the undesirable reactions between the fluorescent proteins that may confound the cell labeling, we first treated the cell mixtures with the monovalent fluorescent proteins, SpyCatcher-ELP-Dronpa and Im7-ELP-mCherrry, to label ScA and Sc Φ, respectively, followed by the other divalent ones, SpyTag-ELP-mCherry-ELP-SpyTag and CL7-ELP-EGFP-ELP-CL7, to label ScB and Sc Ψ; the other way around may doubly label ScB and Sc Ψ. The subsequent confocal microscopic imaging revealed that the matched pairs, ScA + ScB and Sc Φ + Sc Ψ, led to multicellular clusters in numbers significantly larger than the mismatched pairs, ScA + Sc Φ, ScA + Sc Ψ, ScB + ScΦ, and ScB + Sc Ψ, which suggest that the observed clustering is mainly driven by the specific interactions mediated by SpyTag/SpyCatcher and CL7/Im7 on the cell surfaces ( Fig. 3 ). Fig. 3. Cell-to-cell conjugation mediated by surface-displayed SpyTag/SpyCatcher and CL7/Im7 at a high cell density. ( A ) Schematic showing the mixing and assembly process of the matched cell pairs at high density. ( B ) Representative micrographs of the matched pairs, ScA + ScB and Sc Φ + Sc Ψ, and the mismatched pairs, including ScA + Sc Ψ, ScB + Sc Ψ, Sc Φ + ScA , and Sc Φ + ScB . Scale bars, 25 μm. ( C ) Number of multicellular clusters that consist of ≥3 cells. The two-sample t test was used. Data are presented as means ± SD ( n = 5). * P < 0.01. Intercellular assembly controlled by optical tweezers Optical tweezers are powerful, yet noninvasive tools for manipulating living cells ( 25 ). Their exquisite capabilities in force measurement also make it possible to quantitatively analyze intercellular interactions ( 26 , 27 ), so as to distinguish the engineered interactions (i.e., SpyTag/SpyCatcher or Im7/CL7) from the innate ones mediated by the natural glycoproteins such as flocculins. Figure 4A illustrates the way that we trapped a yeast cell in the microfluidic laminar flow chamber using a 1064-nm laser beam. Optically trapping two cells simultaneously with two laser beams allowed us not only to control the intercellular conjugation but also to quantify its strength by pulling the conjugate apart ( Fig. 4B ). When the matched cells were trapped and paired, strong intercellular interactions were observed ( Fig. 4C and movies S1 to S10) ( 28 , 29 ), as the cells remained attached until the pulling forces reached 334.6 ± 20.1 and 335.0 ± 48.7 pN for ScA + ScB and Sc Φ + Sc Ψ, respectively. Moreover, the conjugations between the matched cells, ScA + ScB and Sc Φ + Sc Ψ, controlled by optical trapping proved to be robust by the consistent outcomes observed from at least 10 repeats ( Fig. 4, D and E ). By contrast, negligible interaction forces (<10 pN) were detected between the mismatched cells, ScA + ScA or ScB + Sc Φ, confirming the essential roles of specific PPIs such as SpyTag/SpyCatcher and Im7/CL7 in the stable cell-to-cell conjugation ( Fig. 4, D and E , and fig. S1). Together, these results demonstrated the feasibility of using optical trapping to delicately control and quantitatively assess the assembly of living cells, the latter of which further enabled us to differentiate the designed ultrahigh-affinity intercellular interactions from the innate weak ones. Fig. 4. Cell-to-cell conjugation controlled by optical tweezers. ( A ) Schematic illustration of the trapping of a single cell by an optical tweezer. ( B ) Schematic illustration of the intercellular conjugation and separation by optical tweezers. ( C ) Intercellular conjugation and separation controlled by optical tweezers. Scale bars, 5 μm. ( D and E ) Measurement of interaction forces between cells as a function of optical tweezer movement, where Trap 2 was pulled away at a constant velocity (200 nm/s). n = 10. Scale bars, 5 μm. Growth and proliferation of stable multicellular assemblies Self-propagation is an important characteristic of living matter. To assess whether the multicellular assemblies were able to self-propagate while retaining the assembled structures, the bicellular conjugates that formed via optical trapping served as seeds to be further cultured in the SGCAA medium ( Fig. 5A ). It turned out that, after 130 h of growth and proliferation, the new cells remained bound to the initial bicellular seeds ( ScA + ScB or Sc Φ + Sc Ψ), leading to the formation of three-dimensional (3D) multicellular clusters ( Fig. 5B ). The force measurement with the optical tweezers showed that these multicellular clusters that originated from the matched pairs were remarkably stable; to pull apart the new cells that originated from ScA + ScB and Sc Φ + Sc Ψ, considerable forces, up to 248.0 ± 19.2 and 351.4 ± 28.7 pN ( n = 3), were needed, respectively, substantially larger than those observed among the newly divided cells lacking the complexed SpyTag/SpyCatcher or Im7/CL7 (~36.7 to 133.3 pN), the latter of which might arise from the budding and nonspecific intercellular interactions ( Fig. 5, C and D ). The marked mechanical stability of these clusters grown out of the paired seeds, ScA + ScB and Sc Φ + Sc Ψ, strongly suggests that the designed intercellular PPIs have been not only preserved among the old cells but also inherited by the new ones throughout the growth and proliferation. Moreover, the multicellular structures from the matched pairs exhibited a 3D stacking, relatively compact, while those derived from the mismatched pairs were flat and spread on the bottom of the chamber. These results confirmed that the materials arising from the controlled conjugation of living cells were able to self-propagate and self-assemble while maintaining the compact 3D structures overall as well as the stable intercellular connections. These multicellular ELMs are truly capable of self-propagation, with no need for additional chemical modification, which is therefore superior to traditional polymeric materials or even some other ELM prototypes, the latter of which are often the hybrids of living cells and exogenous polymers ( 3 – 8 ). These findings also point to a cost-effective approach to future biomaterial manufacturing. Fig. 5. Growth and proliferation of bicellular conjugates into stable multicellular assemblies. ( A ) Schematic illustration of the creation of a bicellular seed by optical trapping and its subsequent growth into a multicellular structure. ( B ) Growth of initial bicellular seeds into multicellular assemblies via cell proliferation and intercellular conjugation mediated by Spy chemistry or IM7/CL7. Scale bars, 5 μm. ( C and D ) Measurement of interaction forces between the new cells, originated from the seeds, ScA + ScB (C) and Sc Φ + Sc Ψ (D), as a function of optical tweezer movement, where Trap 2 was pulled away at a constant velocity (200 nm/s). The forces between the mismatched cells were also measured as control. Scale bars, 5 μm. Directed assembly of S. cerevisiae via dielectrophoresis Programmable, long-range assembly of cells, though typical for higher-order organisms, has been rare among the existing ELMs. To control the assembly of S. cerevisiae cells, we leveraged dielectrophoresis (DEP), which has been often used in microfluidics to manipulate and assemble dielectric particles and cells ( 30 , 31 ). A microdevice was created as described in Methods, which can induce positive DEP (pDEP) and negative DEP (nDEP) via high-frequency (500 kHz to 3 MHz) and low-frequency (3 to 100 kHz) alternating electric fields, respectively ( Fig. 6, A and B ). To direct the assembly of the cells into ELMs with tailored structures, the doped conductive silicon electrode was patterned with the nonconductive polymer, poly(ethylene glycol) diacrylate (PEGDA), which also had antifouling properties and prevented cell adhesion (fig. S2). pDEP and nDEP were expected to drive the cells to the sites of the maximum and minimum electric field norms, respectively ( Fig. 6, A and B ). It was confirmed by the respective distributions of the cells on these electrodes at field frequencies of 1 MHz and 100 kHz ( Fig. 6C ); the cells clustered on the exposed silicon electrodes through pDEP within 10 s but moved away when nDEP was the dominant electrokinetic phenomenon ( Fig. 6C ). This behavior created the possibility for the formation of stable monodisperse multicellular aggregates or that of a patterned multicellular porous material using PPIs. Fig. 6. Directed assembly of yeast cells through DEP. ( A and B ) Schematic showing of directed assembly of S. cerevisiae by positive DEP (pDEP) (A) and negative DEP (nDEP) (B). UV, ultraviolet; ITO, indium tin oxide; PEGDA, poly(ethylene glycol) diacrylate. ( C ) Micrographs of wild-type yeast cells driven by pDEP and nDEP at 0 and 10 s after applying electric field. Darker shades correspond to the regions with more cells. Scale bars, 100 μm. ( D ) Multicellular assemblies consisting of ScA + ScB and Sc Φ + Sc Ψ, induced by nDEP, are stable and resistant toward disruptive dielectrophoretic forces, generated by switching from nDEP to pDEP (1 min). The mismatched pair, ScA + Sc Ψ, served as a control. Scale bars, 200 μm. To create stable multicellular assemblies from DEP, we mixed the matched cell types, ScA + ScB and Sc Φ + Sc Ψ, at a ratio of ~1:1, in the microfluidic device, followed by the introduction of an electric field to generate nDEP. Lattice-like multicellular structures emerged, which remained intact for 1 min for both the ScA + ScB and Sc Φ + Sc Ψ pairs after switching from nDEP to pDEP, which generated an opposite dielectrophoretic force ( Fig. 6D ). By contrast, the mismatched pair, ScA + Sc Ψ, assembled via nDEP was transient and fully dissipated 1 min after pDEP was applied ( Fig. 6D ). These results demonstrated the feasibility of using DEP, alongside the specific intercellular PPIs, to control the assembly of cells into stable, patterned structures. Multicellular ELMs for uranium extraction Directly assembling engineered microorganisms into macroscopic living materials, along with their capability of self-propagation and ease of harvesting (compared with single cells), may provide some cost-effective solutions to the challenges facing chemical separation and energy industries, such as uranium extraction from seawater ( 32 ). There is an enormous uranium reserve (4 billion tons in total) in the ocean—1000 times that on land ( 33 ). However, large-scale extraction of uranyl (UO 2 2+ ), a major form of oceanic uranium, is extremely challenging because of its very low concentration (~14 nM) and the abundance of competing ions such as calcium, magnesium, and carbonate. On the other hand, unlike other types of heavy metals, which tend to accumulate in the bottom or certain regions of the ocean, uranyl ions are found to be quite evenly distributed across the ocean, regardless of its depth and location ( 33 ). A variety of synthetic polymeric materials have proved to be feasible for extracting uranium from seawater but at a very high cost ( 34 ). Developing the ability to tap oceanic uranium in a cost-effective manner will be crucial to our goal of achieving clean energy and sustainable development. To create functional ELMs capable of uranium extraction, we introduced the gene encoding the super uranyl-binding protein (SUP) ( 35 ) into the yeast surface display systems, generating two strains, ScA-SUP and ScB-SUP , which can produce and display AGA2-SUP-SpyTag and AGA2-SUP-SpyCatcher, respectively. The introduction of SUP exhibited little effect on the efficiency of the protein surface display, as evidenced by efficient labeling of these cells by the corresponding fluorescent proteins, SpyTag-ELP-EGFP-ELP-SpyTag and SpyCatcher-ELP-EGFP-ELP-SpyCatcher ( Fig. 7A and fig. S3). These cells formed clusters upon mixing, which was indicative of the intercellular interactions mediated by SpyTag/SpyCatcher chemistry ( Fig. 7A ). We further assessed the abilities of these ELMs to extract uranyl, the predominant form of uranium in seawater. It turned out that the yeast cells, even in the absence of SUP, were already able to sequester uranyl, with ~60 to 70% removal from seawater, which might be attributed to the negatively charged glycopeptides on the cell surface ( 36 ). Introducing SUP further enhanced their efficiency in uranyl sequestration; ~90% uranyl was removed from seawater by the assembled ELM comprising ScA-SUP + ScB-SUP , and ~89 and ~82% uranyl was removed by the free ScA-SUP and ScB-SUP cells, respectively ( Fig. 7B ). Together, these results demonstrated as proof of principle the synthesis of functional ELMs by directly assembling engineered single-celled microorganisms, which might offer an economical approach to energy and environmental issues. Fig. 7. Multicellular ELMs for uranium extraction. ( A ) Intercellular conjugation of the yeast cells that display SpyTag–super uranyl-binding protein (SUP) ( ScA-SUP , green) and SpyCatcher-SUP ( ScB-SUP , red) enabled by SpyTag/SpyCatcher chemistry. The cells, ScA-SUP and ScB-SUP , were labeled with the SpyCatcher-ELP-Dronpa and SpyTag-ELP-mCherry-ELP-SpyTag proteins, respectively. Scale bars, 25 μm. ( B ) Uranyl extraction from seawater by free and assembled cells. Data are presented as means ± SD ( n = 3). The two-sample t test was used. * P < 0.01. Multicellular ELMs for strengthened mechanics and underwater adhesion Self-proliferating underwater adhesives can benefit emergency care amid extreme situations such as in a battlefield or a desolate area without basic health care infrastructures and resources. To explore the possibility of developing multicellular assemblies into living adhesives, we introduced into the yeast cells a pair of interacting motifs, mussel foot protein 3/5 (MFP3/5), both of which originated from the marine organism, Mytilus edulis , and crucial for its underwater adhesion ( 37 , 38 ). MFP3 and MFP5 can bind to a variety of substrates in a water-resistant manner ( 39 ), which is largely attributed to their unique amino acid compositions. They are rich in lysine, a positively charged amino acid, and 3,4-dihydroxyphenylalanine, a posttranslationally modified residue derived from tyrosine oxidation by tyrosinase. We first examined the cell-to-cell conjugation between the yeast cells, ScA-MFP3 (red) and ScA-MFP5 (green), which produce and display AGA2-MFP3-SpyTagand AGA2-MFP5-SpyCatcher, respectively ( Fig. 8A and fig. S5). It turned out that the intercellular interactions remained after the introduction of the MFP motifs, as the SpyCatcher- and SpyTag-displaying cells were still able to cluster with each other ( Fig. 8A ). MFP3 and MFP5 are mechanically strong PPI partners and are responsible for the direct interaction between native mussel foot and underwater substrates. To assess the influence of these domains on the mechanics of ELMs, we performed dynamic rheological tests in frequency and strain sweep modes ( Fig. 8, B and C ). The ELMs comprising ScA-MFP5 + ScB-MFP3 proved to be elastic solids; both, in the presence and absence of tyrosinase, exhibited a steady storage modulus G ′ of ~10 5 Pa and a loss modulus G ″ of ~10 4 Pa over the frequency range of 0.01 to 100 rad/s, substantially higher than those free of MFP3/MFP5, i.e., ScA + ScB (fig. S6A), suggesting that MFP3 and MFP5 are essential for the observed strong mechanics. Using shear strength adhesive testing, we further assessed the adhesiveness of these multicellular assemblies. The ELM comprising ScA-MFP5 + ScB-MFP3 in the presence of tyrosinase proved to be an effective glue on glass and porcine skin ( Fig. 8D ), generating adhesion strengths of 8.1 ± 2.1 and 8.2 ± 1.5 kPa, respectively, substantially higher than that in the absence of tyrosinase (2.9 ± 1.2 and 2.1 ± 0.7 kPa). This living glue was also stronger than the one free of MFP3/5, i.e., ScA + ScB , with an adhesive strength of 3.1 ± 0.7 kPa toward porcine skin, suggesting that the presence of the MFP3/5 domains is crucial for strong adhesiveness. On the other hand, either ScA-MFP5 or ScB-MFP3 alone turned out to be weaker, even after tyrosinase treatment, in adhesion (~3.0 to 5.0 kPa) than their assembled counterpart, ScA-MFP5 + ScB-MFP3 , which suggests that the latter’s strong adhesion is likely to be an emergent property out of the combined contributions from the MFP3/5 domains and the multicellular structure ( Fig. 8E ). Fig. 8. Strengthened mechanics and underwater adhesion of ELMs by MFP3/5. ( A ) Multicellular assembly of the yeast cells displaying SpyTag-MFP5 ( ScA-MFP5 , green) and SpyCatcher-MFP3 ( ScB-MFP3 , red). ScA-MFP5 and ScB-MFP3 cells were labeled by the SpyCatcher-ELP-Dronpa and SpyTag-ELP-mCherry-ELP-SpyTag proteins, respectively. Scale bars, 25 μm. ( B ) Dynamic frequency sweep tests of ScA-MFP5 + ScB-MFP3 before and after tyrosinase oxidation. ( C ) Dynamic strain sweep tests of ScA-MFP5 + ScB-MFP3 before and after tyrosinase oxidation. ( D ) Adhesion tests of ScA-MFP5 + ScB-MFP3 on glass or porcine skin with and without tyrosinase oxidation. ( E ) Adhesion tests of different S. cerevisiae cells on porcine skin with and without tyrosinase oxidation. Data are presented as means ± SD ( n = 3). The two-sample t test was used. * P < 0.01. In summary, we have reported a new approach to self-propagating ELMs that involved the use of strong intercellular PPIs to assemble single-celled microorganisms. This assembly process can be precisely controlled in shape and stability via optical tweezers or microfluidics. Depending on the functional motifs introduced by genetic programing, the resulting multicellular ELMs have given rise to emergent properties that hold great promise for applications ranging from uranium extraction to bioadhesion. This study has thus exemplified the use of engineered single cells as versatile building blocks, while free from chemical modification, for synthesizing higher-order living structures or materials. In this sense, this study stands for a new way of thinking for synthetic biology; these single cells are to synthetic biology what atomic/molecular building blocks are to synthetic chemistry."
} | 6,747 |
37737271 | PMC10516961 | pmc | 411 | {
"abstract": "Deep convolutional neural networks (CNNs) have achieved promising performance in the field of deep learning, but the manual design turns out to be very difficult due to the increasingly complex topologies of CNNs. Recently, neural architecture search (NAS) methods have been proposed to automatically design network architectures, which are superior to handcrafted counterparts. Unfortunately, most current NAS methods suffer from either highly computational complexity of generated architectures or limitations in the flexibility of architecture design. To address above issues, this article proposes an evolutionary neural architecture search (ENAS) method based on improved Transformer and multi-branch ConvNet. The multi-branch block enriches the feature space and enhances the representational capacity of a network by combining paths with different complexities. Since convolution is inherently a local operation, a simple yet powerful “batch-free normalization Transformer Block” (BFNTBlock) is proposed to leverage both local information and long-range feature dependencies. In particular, the design of batch-free normalization (BFN) and batch normalization (BN) mixed in the BFNTBlock blocks the accumulation of estimation shift ascribe to the stack of BN, which has favorable effects for performance improvement. The proposed method achieves remarkable accuracies, 97.24 \\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}$$\\%$$\\end{document} % and 80.06 \\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}$$\\%$$\\end{document} % on CIFAR10 and CIFAR100, respectively, with high computational efficiency, i.e. only 1.46 and 1.53 GPU days. To validate the universality of our method in application scenarios, the proposed algorithm is verified on two real-world applications, including the GTSRB and NEU-CLS dataset, and achieves a better performance than common methods.",
"conclusion": "Conclusions and future work This article proposed an efficient ENAS algorithm combing multi-branch ConvNet with batch-free normalization Transformer backbone to evolve compact CNN architectures for image classification task. To effectively search network architectures with promising performance, a search space was designed based on multi-branch block and the proposed BFNTBlock. Furthermore, a flexible encoding strategy was developed to adaptively evolve the configurations of CNN in a variable-length manner. The designed crossover and mutation operators provide effective local search and global search ability for our proposed algorithm. The computational complexity of the algorithm is significantly reduced by a parallel computational component. The proposed algorithm is examined on two challenging classification benchmark datasets, CIFAR10 and CIFAR100, achieving accuracies 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}$$97.24\\%$$\\end{document} 97.24 % , \\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}$$80.06\\%$$\\end{document} 80.06 % , respectively, with low search cost. The proposed algorithm is superior to almost all the hand-crafted architectures and ENAS algorithms, and achieves a competitive performance among the non-EAs-based NAS competitors. Furthermore, the proposed algorithm obtains competitive performance on two real-world applications, including the GTSRB and NEU-CLS dataset. A limitation of our approach is the search space based on richly experienced building blocks limit the diversity of architectures. More flexible search spaces deserve to be explored. Extending NAS methods to other complicated downstream tasks like semantic segmentation and object detection is another meaningful future work. In addition, we plan to focus on developing effective evolutionary methods to notably accelerate the process of fitness evaluation.",
"introduction": "Introduction Convolutional neural networks (CNNs), as the foundation of deep learning 1 , have made prominent achievements in various specialized fields, such as image classification 2 , natural language processing 3 and object detection 4 . The performance of CNNs counts on its architecture to a great extent. However, discovering optimal architectures for different tasks is extremely challenging, because architectures have a large number of parameters and relatively complex structure. At an early stage, the vast majority of the most advanced CNN architectures are hand-crafted by specialists in both neural networks and images processing, such as GoogleNet 5 , ResNet 6 and DenseNet 7 . Unfortunately, in practice, most users are limited by scarce domain knowledge. Moreover, CNNs are often used to solve specific problems, and once assignments changes, the architecture must be redesigned correspondingly. As a result, neural architecture search (NAS) has attracted unprecedented attention. NAS aims to automatically design network architectures that achieves the best possible performance with minimal human intervention. Existing NAS algorithms are divided into three main categories: Reinforcement Learning (RL)-based NAS 8 , gradient (GD)-based NAS and Evolutionary Algorithms (EAs)-based NAS 9 . The RL requires large computational resources even on the median-scale dataset like CIFAR10 dataset. Compared with RL, the gradient-based algorithms decrease the consumption of computational resources, but requires to construct a supernet in advance, which have high requirements for expertise. The ENAS searches for superior performance architectures by using EAs. Specifically, EAs are a cluster of algorithms based on biological evolutionary mechanisms, such as natural selection, simulating the evolution of species, to solve optimization problems. Real et al. 10 conducted a comparison between RL-based NAS and EAs-based NAS, the results demonstrated that EAs-based NAS can converge faster with the same hardware, especially in the initial stage of search. The methods to design neural network using EAs can be main divided in two branches: Neuroevolution (NE) 11 and Evolutionary neural architecture search (ENAS). The former exploits EAs to optimize neural networks, and also enables important capabilities, including learning neural network building blocks, the design of topologies, and hyper-parameters 12 . The latter designs neural architectures through Evolutionary Computation (EC) methods. Owing to the evolution of population, the performance of the architecture can constantly improve to a relatively high level on the research task. Recently, ENAS has attracted great attention due to its superior performance. However, since EAs are search methods based on population, plenty of fitness require to be evaluated, leading to ENAS consume enormous computational resources. For instance, CNN-GA 13 needs to operate on several GPUs for 35 days on CIFAR10 datasets. Therefore, ENAS focus on decreasing the computational resources and accelerating the search process of the neural architecture. Techniques to reduce computational overheads mainly contain two aspects. One common approach to reduce computational overheads is converting the global search into modular search space 14 . In contrast to the global search space that searches entire neural structure, the modular search space only requires searching a few small cell structures, after which cells are stacked to form the final architecture. Additionally, it is convenient for modular search space to migrate on other tasks, which is usually impossible for global search space. Therefore, compact and flexible is the superiority of the modular search space. The block-based NAS methods generally perform well since they restrict the search space and are inclined to design compact CNN architectures 15 . AE-CNN 16 constructed a search space based on ResNet block 6 and DenseNet block 7 to automatically design CNN architectures. CNN-GA 13 employed skip connections, enabling deep CNN architectures to be efficiently trained. In the two aforementioned algorithms, an efficient variable-length coding strategy was designed to adaptively search for the unpredictable optimal depth of CNNs. The other method is fitness evaluation. Early versions of NAS 17 usually find the best neural architecture according to the performance, which is extremely time-consuming since many candidate neural architectures need to be compared. For example, AE-CNN 16 , a block-based search method, takes 27 GPU days on CIFAR10 dataset. Hence, a majority of NAS algorithms focus on how to avoid resource consumption. FP-DARTS 18 constructs a super network with two-parallel-path to accelerate the training process. EBNAS 19 proposes an efficient binary neural architecture search algorithm that designs a search space simplification strategy to improve search efficiency. To search for the network architecture with exceptional performance and further improve search efficiency, this paper proposes an ENAS algorithm combining multi-branch CNNs with improved Transformer. The main contributions of this paper are threefold as follows: From the perspective of architecture search space, the multi-branch block (MBB) is employed to generate architectures with exceptional performance. MBB extracts abundant features and enhance the representational capacity of a single convolution by combining diverse operations having various receptive fields. Unlike a majority of traditional novel ConvNet architectures, CNN architecture constructed by MBB has a larger model capacity and can be trained to achieve remarkable performance. To search for CNN architecture with the optimal depth, we designed a novel variable-length encoding strategy, which can accommodate various basic units including attention module. Meanwhile, the genetic operators tailored for the novel encoding scheme are designed to avoid the decline of population diversity and improve the optimization efficiency. A simple yet powerful backbone “batch-free normalization Transformer Block” (BFNTBlock) is proposed to capture long-range dependencies. The proposed BFNTBlock leverages the merits of long range dependence and spatial adaptability derived from MHSA, as well as properties of CNNs, such as inductive bias and local receptive field. In particular, the batch-free normalization (BFN) introduced in BFNTBlock independently normalizes each sample without across batch dimension, which impedes the accumulative estimation shift of BN. This relieves the degeneration of performance if a distribution shift occurs. In addition, an asynchronous computational component is applied to accelerate the design of CNN architectures and discover the optimal network architecture within an acceptable time. From the results compared with state-of-the-art peer competitors on widely used image classification datasets in NAS: CIFAR10 and CIFAR100, the proposed method is not only computationally efficient but also highly competitive in terms of performance among all compared methods. To further validate the universality in application scenarios, the proposed algorithm is valuated on the GTSRB and the NEU-CLS datasets, and outperforms the commonly used methods.",
"discussion": "Results and discussion In this section, we firstly present the overall comparison results between the proposed algorithm and peer competitors, while the evolutionary trajectories are displayed to demonstrate the convergence of the proposed algorithm within the parameter settings. Then, the discovered optimal CNN architectures on CIFAR10 and CIFAR100 datasets are shown. Again, we conduct experiments on various attention mechanisms to certify the superior performance and adaptability of proposed BFNTBlock. To validate the effectiveness of our algorithm in real applications, we perform experiments on two application scenarios. Finally, ablation studies are conducted to verify the contribution of each component in our algorithm for performance improvement. Overall results We investigate our method and the compared algorithms in terms of test classification accuracy, the parameters and FLOPs (if available) of output architecture and consumed computational cost. In particular, referring to existing studies, the “GPU Days” 86 is selected as the quantitative metric for computational cost. The comparison results grouped into three different categories are displayed in Table 1 . The symbol “-” implies that the corresponding algorithm does not publicly report results. “SMBO” stands for the “sequential model based global optimization”. “BO” represents the “Bayesian optimization”. Note that the results of competitors are extracted from their seminal papers, since the results in papers are usually the best. As observed in Table 1 , the proposed algorithm outperforms all state-of-the-art handcrafted CNNs on CIFAR10. These methods use well-designed building blocks to improve the compactness and the performance of the architecture, such as Dense block in DenseNet and MBConv block in MobileNetV2. It is shown that the proposed algorithm obtains \\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}$$3.67\\%$$\\end{document} 3.67 % 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}$$0.7\\%$$\\end{document} 0.7 % accuracy improvements over ResNet-101 and DenseNet on CIFAR10 dataset, respectively, while having far fewer parameters than DenseNet (4.73M < 25.6M). Moreover, the FLOPs of the proposed algorithm is much smaller than that of the Wide-ResNet and DenseNet. Compared with ResNext-29, our method shows better performance, and has a smaller parameters. In addition, our proposed algorithm achieves the highest classification accuracy among Wide-ResNet, Pre-ResNet, SENet, MobileNetV2, ShuffleNet, IGCV3-D on CIFAR10. On CIFAR100, the proposed algorithm employs \\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}$$84\\%$$\\end{document} 84 % , \\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}$$49\\%$$\\end{document} 49 % 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}$$45\\%$$\\end{document} 45 % fewer parameters compared to Wide-ResNet, SENet and Pre-ResNet, respectively, and improves performance by \\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}$$0.56\\%$$\\end{document} 0.56 % , \\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}$$2.77\\%$$\\end{document} 2.77 % 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}$$2.77\\%$$\\end{document} 2.77 % , respectively. Though the performance of our algorithm is inferior to those of DenseNet and ResNext-29, the number of parameters is dramatically decreased. The proposed algorithm also outperforms ResNet-101, MobileNetV2, ShuffleNet, IGCV3-D in terms of the classification accuracy. Compared to the lightweight networks, MobileNetV2 and ShuffleNet, the accuracy of our algorithm achieve significant improvements on both dataset. Compared with non-evolutionary NAS methods in the second category, the proposed algorithm consumes fewer GPU days than all RL-based NAS algorithms except ENAS and FPNAS + Cutout, which uses a sharing parameters approach to accelerate convergence. Moreover, it is observed that, our proposed algorithm outperforms NAS V3 ( \\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}$$1.71\\%$$\\end{document} 1.71 % ), ENAS ( \\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}$$0.18\\%$$\\end{document} 0.18 % ), MetaQNN ( \\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}$$4.16\\%$$\\end{document} 4.16 % ), NASNet-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}$$0.97\\%$$\\end{document} 0.97 % ), Block-QNN-S ( \\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}$$1.62\\%$$\\end{document} 1.62 % ), RENAS ( \\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}$$0.12\\%$$\\end{document} 0.12 % ), Path-level EAS ( \\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}$$0.23\\%$$\\end{document} 0.23 % ), FPNAS + Cutout ( \\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}$$0.25\\%$$\\end{document} 0.25 % ) and PNAS ( \\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}$$0.65\\%$$\\end{document} 0.65 % ) on CIFAR10 datasets, while merely consumes about one GPU days. Compared with GD-based algorithms, i.e. SNAS, DARTS, RC-DARTS-C42, Firefly and AdaptNAS-S, our method achieves competitive performance on CIFAR10. However, GD-based methods propose a method to enable the search space continuous in order to optimize the architecture using SGD, which requires a huge hand-crafted CNN as the building block. Once hand-crafted CNN is smaller than the optimal one, GD-based methods will never discover the best CNN architecture. There are no such limitations for the proposed algorithm, which designs promising CNN architectures without domain knowledge. Though our proposed algorithm obtained a slightly worse classification than Proxyless ( \\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}$$-0.68\\,\\%$$\\end{document} - 0.68 % ), the proposed algorithm has fewer parameters (4.7M < 5.7M) and much less GPU days. For BO-based NAS algorithm, the proposed algorithm surpasses NAGO, GP-NAS in accuracy, and obtains competitive performance with BANANAS in much fewer computational cost. On CIFAR100, our method achieves better performance compared to MetaQNN ( \\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}$$7.2\\%$$\\end{document} 7.2 % ), Block-QNN-S ( \\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}$$0.71\\%$$\\end{document} 0.71 % ), SNAS ( \\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}$$0.15\\%$$\\end{document} 0.15 % ) and NAGO ( \\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}$$0.76\\%$$\\end{document} 0.76 % ). Though the classification accuracy of the proposed algorithm is bit lower than DARTS, but the search only takes 1.53 GPU days. In the second category, the proposed algorithm have the lowest number of FLOPs. Compared to the proposed algorithm, the major limitation of the non-evolutionary NAS algorithms is the extensively requirement of domain expertise for users. For instance, the CNNs generated by Block-QNN-S cannot be directly used, which must be inserted into a larger hand-crafted CNN in advance, and the final performance of Block-QNN-S depends on whether the larger network is properly designed. The reasons for the proposed algorithm outperforming NAS V3 and Meta-QNN can be summarized as follows. First, there are not apply the crossover operator in Large-scale Evolution, losing the local search ability. In addition, because NAS and Meta-QNN are designed based on RL, lacking the process of fitness evaluation, which often consume more computational resources. Table 1 Comparison of the proposed algorithm with peer competitors in terms of accuracy ( \\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}$$\\%$$\\end{document} % ), the number of parameters(M), and the consumed GPU days on CIFAR10 and CIFAR100 datasets. Architecture CIFAR10 CIFAR100 Parameters GPU days Search method ACC FLOPs ACC FLOPs ResNet-101 93.57 – 74.84 – 1.7 – Manual Wide-ResNet 95.83 – 79.50 5248 36.5 – Manual Pre-ResNet 95.36 – 77.29 – 10.3 – Manual DenseNet 96.54 9388 82.82 – 25.6 – Manual ResNext-29 96.42 – 82.69 – 68.1 – Manual SENet 95.95 – 77.29 – 11.2 – Manual MobileNetV2 94.56 – 77.09 – 2.1 – Manual ShuffleNet 90.87 – 77.14 – 1.06 – Manual IGCV3-D 94.96 – 77.95 – 2.2 – Manual NAS V3 95.53 – – – 7.1 22,400 RL ENAS 97.06 – – – 4.2 0.5 RL MetaQNN 93.08 – 72.86 – – 100 RL Block-QNN-S 95.62 – 79.35 – 6.1 90 RL Proxyless NAS 97.92 – – – 5.7 1,500 RL NASNet-B 96.27 – – – 2.6 2,000 RL FPNAS + Cutout 96.99 – – – 5.8 0.83 RL Path-level EAS 97.01 – – – 5.7 200 RL RENAS 97.12 – – – 3.5 6 RL+EA SNAS 97.15 – – – 2.8 1.5 GD SNAS – – 79.91 422 2.8 1.5 GD DARTS 97.00 547 – – 3.3 1.5 GD DARTS – – 82.46 528 3.4 4 GD RC-DARTS-C42 97.19 – – – 3.3 1 GD Firefly 97.27 – – – 3.3 1.5 GD AdaptNAS-S 97.50 – – – 5.3 2 GD PNAS 96.59 – – – 3.2 150 SMBO BANANAS 97.36 – – – – 11.8 BO NAGO 96.60 – 79.30 – – – BO GP-NAS 96.21 – – – – 0.9 BO Large-scale evolution 94.60 – – – 5.4 2,750 EA Large-scale evolution – – 77.00 – 40.4 2,750 EA LEMONADE 97.42 – – – 13.1 90 EA AmoebaNet-A 96.66 – 81.07 – 3.3 3,150 EA AE-CNN+E2EPP 94.70 – – – 4.3 7 EA AE-CNN+E2EPP – – 77.98 – 20.9 10 EA NSGANet 97.25 1290 – – 3.3 4 EA NSGANet – – 79.26 1290 3.3 8 EA Hierarchical evolution 96.37 – – – 15.7 300 EA CNN-GA 96.78 - - - 2.9 35 EA CNN-GA – – 79.47 – 4.1 40 EA AE-CNN 95.30 – – – 2.0 27 EA AE-CNN – – 77.60 – 5.4 36 EA SI-EvoNet 96.02 – – – 0.51 0.46 EA SI-EvoNet – – 79.16 – 0.99 0.81 EA FPSO 95.16 – – – 0.7 1.25 EA EffPNet 96.51 – 81.51 – 2.54 \\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}$$<3$$\\end{document} < 3 EA Proposed 97.24 167 – – 4.73 1.46 EA Proposed – – 80.06 260 5.71 1.53 EA Bold highlights the performance of our algorithm. For ENAS algorithms in the third category, the proposed algorithm shows superior performance over Large-scale Evolution, AmoebaNet-A, AECNN+E2EPP, Hierarchical Evolution, CNN-GA, AE-CNN and FPSO on CIFAR10, but slightly worse than LEMONADE ( \\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}$$-0.18\\%$$\\end{document} - 0.18 % ). Note that Large-scale Evolution and Hierarchical Evolution accelerate search phrase by hardware distributed computation, while CNN-GA and AE-CNN design search space based on efficient building blocks to decrease computational cost. However, they still consumes enormous computational resources. Compared to these ENAS algorithms with high computational costs, the proposed algorithm only takes 1.46 and 1.53 GPU days on CIFAR10 and CIFAR100, respectively. In addition, AE-CNN+E2EPP uses an random forest-based performance predictor to accelerate the process of fitness evaluation, but the proposed algorithm still consumes considerably less GPU days than AE-CNN+E2EPP without using any training tricks, which indicates CNN architectures searched by proposed algorithm are well compact. Compared with the lightweight architecture SI-EvoNet, the performance of proposed algorithm is significantly improved ( \\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}$$1.22\\%$$\\end{document} 1.22 % ). In particular, the proposed algorithm with fewer GPU days achieves competitive classification accuracy compared to the classical NSGANet. For EffPNet, the proposed algorithm is substantially lower in search cost, and achieves better performance on CIFAR10. On CIFAR100, our method perform much better than Large-scale Evolution and AE-CNN+E2EPP, while only has 5.71M parameters, saving \\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}$$86\\%$$\\end{document} 86 % parameters compared to Large-scale evolution. Moreover, the proposed algorithm outperforms NSGANet, CNN-GA, AE-CNN and SI-EvoNet. Compared with AmoebaNet-A and EffPNet, the performance on CIFAR100 is weaker against it, but our method consumes much less GPU Days. In summary, our proposed algorithm outperforms the advanced manually designed CNNs and most of ENAS algorithms in terms of classification accuracy, the number of parameters and the computational overheads. Meanwhile, the discovered architecture is more lightweight and compact than most of the state-of-the-art CNNs. Moreover, the proposed algorithm has much fewer FLOPs than that of the other methods which have been reported. Although RL-based and GD-based architecture algorithms display similar (or slightly better) classification accuracy to that of the proposed method, our algorithm achieves preferable speed-accuracy trade-offs. Designed CNN architectures As seen from Fig. 7 a, the best architecture on CIFAR10 dataset designed by proposed algorithm consists of nine multi-branch blocks, three pooling blocks and one BFNTBlock, namely containing 18 blended layers. The best architecture on CIFAR100 displayed in Fig. 7 b is similar to that on CIFAR10, which also includes 18 hybrid layers. Compared with the most state-of-the-art CNNs that are solely built on basic blocks like residual block and DenseNet block, the architectures based on multi-branch blocks and BFNTBlocks automatically discovered by proposed algorithm have highly competitive performance with the less number of parameters. This means ensemble blocks may be more effective. CIFAR100 is a more complicated benchmark dataset than CIFAR10, and CNN architectures on CIFAR100 dataset typically contain more layers than that of CIFAR10. However, according to the optimal architectures displayed in Fig. 7 , the complexity of best architecture on CIFAR100 is the same as that of on CIFAR10, which indicates the proposed algorithm can design architectures with the appropriate depth according to diverse tasks. Figure 7 The discovered optimal CNN architectures on CIFAR10 ( a ) and CIFAR100 ( b ). Evolutionary trajectory Aiming at intuitively illustrating the evolutionary process of discovered CNN architectures, the evolutionary trajectories of our algorithm on benchmark datasets are displayed in Fig. 8 . The horizontal and vertical axis indicate generation number and classification accuracy respectively. The red line refers to the average classification accuracy of each generation, while the green area is depicted by the best and worst individual in each generation. As depicted in Fig. 8 a, the mean classification accuracy remains almost unchanged from the first generation to third generation. Due to the random initialization of population, which lead to generate inferior individuals, the mean accuracy slightly downturn at the fourth generation. Then, the individuals with uncompetitive fitness are eliminated, the mean accuracy improves and steadily moves forward until the 7th generation. After that, mean classification accuracy increases to nearly \\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}$$82\\%$$\\end{document} 82 % as the evolution process proceeds. Finally, the mean classification keep a steady state which implies the population converges. An analogous situation can be seen from Fig. 8 b, the mean classification accuracy remains constant when the evolution terminates. Meanwhile, the upper and lower boundaries of the light-green areas (i.e. the best and worst classification accuracy) gradually rise and converges to a stead state. This indicates the parameter setup is reasonable since the proposed algorithm well converges. In the case of more computational resources, the evolutional generation and population size can be set to larger numbers. Application to traffic sign recognition task \n Figure 8 Evolutionary trajectories of the proposed algorithm on CIFAR10 ( a ) and CIFAR100 ( b ). \n To further demonstrate the superior performance and generalization of the proposed algorithm in real applications, we conducted experiments on GTSRB. The German Traffic Sign Recognition Benchmark (GTSRB) dataset 87 is the standard benchmark dataset for traffic sign classification task, which contains 43 classes of traffic signs and has a total of 51,839 images, including 39,209 training images and 12,630 test images. The champion on GTSRB is MDCNN proposed by IDSIA team 88 . The MDCNN is comprised of 25 deep neural networks (DNNs) with same architecture, achieving an accuracy 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}$$99.46\\%$$\\end{document} 99.46 % on recognition accuracy. The optimal CNN architecture on GTSRB mainly consists of nine multi-branch blocks, three pooling blocks and one BFNTBlock (see Fig. 9 ). Compared with the state-of-the-art CNNs 6 , the optimal architecture comprises fewer layers and a simpler structure. In particular, the merits of discovered CNN architectures are in three folds: (1) multi-branch block possesses different paths with the ability to sufficiently extract abundant features while maintain highly competitiveness in terms of performance. (2) The proposed BFNTBlock concentrates on the aggregation of global information, which is complementary to CNNs that focus on local information. The effective aggregation between global and local information is of great significance to generate excellent network architecture. (3) The design of BN and GN mixed in the BFNTBlock obstruct the accumulated estimation shift due to the stack of BN. Table 2 refers to the classification accuracy on GTSRB obtained by state-of-the-art approaches. The proposed algorithm achieves an accuracy of 99.61 \\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}$$\\%$$\\end{document} % with a small number of parameters, which outperforms the human performance and other methods. This domenstrates the proposed algorithm can design network architecture efficiently and automatically for specific tasks. Table 2 Comparison of the proposed algorithm with peer competitors on GTSRB dataset. Team Methods Acc ( \\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}$$\\%$$\\end{document} % ) Proposed ENAS 99.61 IDSIA Committee of CNNs 99.46 INI-RTCV Human performance 98.84 Sermanet Multi-scale CNNs 98.31 CAOR Random forests on HOG 96.14 INI-RTCV LDA on HOG 95.68 The best records are marked in bold. \n Figure 9 The best architecture found by proposed algorithm on GTSRB. \n Application to defect classification NEU-CLS is the surface defect classification dataset, containing six types of defects in the hot-rolled steel strip, each of which includes 300 images with resolution of 200 × 200. The training, validate and testing dataset are divided with the proportion of 6.4:1.6:2 89 . The comparison results on NEU-CLS are presented in Table 3 . AECLBP 90 and BYEC 91 are manual feature extraction methods. The Decaf 92 , ResNet34-MFN 93 and SDC-SN-ELF+MRF 94 , are the CNN-based methods. The NAS-SDC-B 95 is a NAS-based method. As shown in Table 3 , the proposed method achieved 99.44 \\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}$$\\%$$\\end{document} % prediction accuracy. It outperforms other classical methods and obtains the competitive performance against SDC-SN-ELF+MRF and NAS-SDC-B. Table 3 Comparison of the proposed algorithm with peer competitors on NEU-CLS dataset. Methods Acc ( \\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}$$\\%$$\\end{document} % ) AECLBP 98.93 BYEC 96.30 Decaf 99.27 ResNet34-MFN 99.17 SDC-SN-ELF+MRF 100 NAS-SDC-B 99.63 Proposed 99.44 The best records are marked in bold. Comparisons with state-of-the-art attention modules To demonstrate the superiority of the BFNTBlock, we introduced different competitive attention modules in the individuals searched on three datasets to compare performance, including SE 44 , CBAM 45 , Shuffle attention 96 , Triplet attention 97 , MHSA 48 , BoTNet 53 . Evaluation criteria include the parameters of the network, accuracy and GPU days. For a fair comparison, all experiments are performed on training settings described in Sect. \" Parameters settings \". The baseline is MBB. Concretely, CBAM, a lightweight and general module, infers attention maps along channel and spatial dimensions, after which are aggregated to form the feature map. Similar to CBAM, SE is also a light attention module, proposing a method to enhance the ability of the representation for the network by modelling channel-wise relationships. MHSA allows a model to jointly process information from diverse subspaces. Shuffle Attention proposes a lightweight attention module, which constructs channel attention and spatial attention simultaneously for each sub-feature. Triplet attention captures cross-dimensional interactions through a three-branch structure to calculate attention weights. Table 4 Comparisons of diverse attention modules on CIFAR10, CIFAR100 and GTSRB in terms of accuracy ( \\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}$$\\%$$\\end{document} % ) and parameters(M). Attention module CIFAR10 CIFAR100 GTSRB Baseline 95.14 – – Baseline – 75.12 – Baseline – – 98.42 SE 95.14 – – SE – 77.52 – SE – – 99.43 CBAM 94.92 – – CBAM – 75.56 – CBAM – – 98.92 MHSA 94.12 – – MHSA – 75.38 – MHSA – – 98.49 SA 94.94 – – SA – 75.56 – SA – – 99.01 TA 95.12 – – TA – 75.17 – TA – – 98.74 BoTNet 95.82 – – BoTNet – 77.37 – BoTNet – – 98.86 BFNTBlock \\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}$$_{GN}$$\\end{document} GN -P1 96.06 – – BFNTBlock \\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}$$_{GN}$$\\end{document} GN -P3 – 77.82 – BFNTBlock \\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}$$_{GN}$$\\end{document} GN -P1 – – 99.29 SA and TA denote Shuffle attention and Triplet attention respectively. Table 4 represents the comparisons of the proposed BFNTBlock with diverse attention modules. It can be observed that the BFNTBlock enhances classification performance by \\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}$$0.92\\%$$\\end{document} 0.92 % , \\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}$$2.7\\%$$\\end{document} 2.7 % , \\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}$$0.87\\%$$\\end{document} 0.87 % on CIFAR10, CIFAR100 and GTSRB respectively compared with the baseline. Note that the classification accuracy on CIFAR10 degrades after incorporating CBAM, SA and MHSA due to the lack of adaptability. The experimental results verify that the proposed BFNTBlock can be better integrated with architectures designed by the proposed algorithm, and achieves the optimal performance improvement. The improvements mainly benefit from the followings: (1) the proposed BFNTBlock leverages the merits of both MHSA and convolutions, including global receptive field and inductive bias (2) GN employed in the BFNTBlock obstruct the accumulative estimation shift arising from the stack of BN. Advanced training strategies Besides the standard training strategies described in Sect. \" Parameters settings \", we adopt advanced training strategies to augment the training data: (1) cutout 98 with a patch length of 16, 8, 2 on CIFAR10, CIFAR100 and GTSRB respectively (2) mixup 99 training with a mix factor of 0.2. ‘ST’ indicates standard training. The baseline model comprises MBB and BoTNet. As shown in Table 5 , the networks with BFNTBlock generated by proposed algorithm also consistently outperforms the baseline by a remarkable margin. This reconfirms the remarkable performance of the proposed BFNTBlock. Table 5 The accuracy ( \\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}$$\\%$$\\end{document} % ) on CIFAR10, CIFAR100 and GTSRB using advanced training strategies. Training strategies CIFAR10 CIFAR100 GTSRB Baseline \\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}$$\\hbox {BFNTBlock}_{GN}$$\\end{document} BFNTBlock GN -P1 Baseline \\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}$$\\hbox {BFNTBlock}_{GN}$$\\end{document} BFNTBlock GN -P3 Baseline \\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}$$\\hbox {BFNTBlock}_{GN}$$\\end{document} BFNTBlock GN -P2 ST 95.82 96.06 77.37 77.82 98.86 98.95 MixUp 94.62 96.50 76.67 78.62 99.33 99.42 CutOut 95.68 96.80 75.84 79.17 99.51 99.43 MixUp+CutOut 96.04 97.24 77.92 80.86 99.53 99.61 Ablation studies Positions of BFN in a BFNTBlock We investigate the position of BFN applied in the BFNTBlock. For fair comparison, all experiments are conducted on the settings according to Sect. \" Parameters settings \". We use GN (group number=64) as BFN in this paper. We design three BFNTBlock variants, which replace the first, second and third BN in the BoTNet (referred as ’BFNTBlock \\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}$$_{GN}$$\\end{document} GN -P1’, ’BFNTBlock \\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}$$_{GN}$$\\end{document} GN -P2’ and ’BFNTBlock \\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}$$_{GN}$$\\end{document} GN -P3’). We substitute these BFNTBlocks for BoTNet and report the results in Table 6 . Our baseline is the model containing MBB and BoTNet. Table 6 shows the comparison results of applying a GN at different positions in the BFNTBlock. It can be discovered that in addition to the slight performance decline of BFNTBlock \\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}$$_{GN}$$\\end{document} GN -P3 on CIFAR10 and that of BFNTBlock \\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}$$_{GN}$$\\end{document} GN -P1 on CIFAR100, the performances of other networks with BFNTBlock outperform the baseline to an extent. This further verify the comprehensive availability of our BFNTBlock. Table 6 Results of GN applied at the different positions in a BFNTBlock. Methods CIFAR10 CIFAR100 GTSRB Baseline 95.82 77.37 98.86 \\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}$$\\hbox {MBB+BFNTBlock}_{GN}$$\\end{document} MBB+BFNTBlock GN -P1 96.06 \\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}$${\\textbf {(+0.24)}}$$\\end{document} ( + 0.24 ) 77.31 \\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}$${\\textbf {(-0.06)}}$$\\end{document} ( - 0.06 ) 99.29 \\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}$${\\textbf {(+0.43)}}$$\\end{document} ( + 0.43 ) \\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}$$\\hbox {MBB+BFNTBlock}_{GN}$$\\end{document} MBB+BFNTBlock GN -P2 95.90 \\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}$${\\textbf {(+0.08)}}$$\\end{document} ( + 0.08 ) 77.43 \\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}$${\\textbf {(+0.06)}}$$\\end{document} ( + 0.06 ) 98.95 \\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}$${\\textbf {(+0.09)}}$$\\end{document} ( + 0.09 ) \\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}$$\\hbox {MBB+BFNTBlock}_{GN}$$\\end{document} MBB+BFNTBlock GN -P3 95.76 \\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}$${\\textbf {(-0.06)}}$$\\end{document} ( - 0.06 ) 77.82 \\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}$${\\textbf {(+0.45)}}$$\\end{document} ( + 0.45 ) 98.94 \\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}$${\\textbf {(+0.08)}}$$\\end{document} ( + 0.08 ) We evaluate the test classification accuracy ( \\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}$$\\%$$\\end{document} % ) on CIFAR10, CIFAR100 and GTSRB datasets. Bold indicates the corresponding improvement ratio. Comprehensive analysis for components of proposed algorithm We thoroughly analyse the components of the proposed algorithm and validate the contribution of the designed BFNTBlock to improve the expression ability of CNN architecture. First, we conduct experiments on the baseline model merely containing multi-branch block. After that, considering the MHSA is one of the most essential components of BoTNet, on which the proposed BFNTBlock is designed, the MHSA, BoTNet and BFNTBlock are sequentially incorporated into neural architectures to test performance. All parameters are consistent with configurations set in Sect. \" Parameters settings \". It can be obviously observed from Table 7 , the performance on CIFAR10 remarkably descends when only incorporate MHSA due to the lack of normalization, which has a great impact on learning robust significant coefficients and may lead to degradation of performance. However, the performance on CIFAR10, CIFAR100 and GTSRB remarkably improve when the proposed BFNTBlock is introduced in the discovered architecture. BFNTBlock sufficiently leverages the advantages of convolutions and MHSA. Convolutional layer possesses better generalization ability with faster convergence due to its inherent prior of inductive bias, while MHSA layer has larger model capacity that can benefit from larger datasets. In particular, BFNTBlock obstruct accumulation of estimation shift, which ensures a significant performance improvement. Furthermore, BFNTBlock is a lightweight and powerful module, it can be seamlessly integrated into CNN architectures with negligible overheads. The experimental results demonstrate all components in our algorithm are indispensable for the improvement of performance. Table 7 Performance comparisons of the proposed algorithm with different components. Methods CIFAR10 CIFAR100 GTSRB Parameters MBB 95.14 - - 3.81 MBB – 75.12 – 4.08 MBB – – 98.42 1.62 MBB + MHSA 94.12 – – 3.9 MBB + MHSA – 75.38 – 2.92 MBB + MHSA – – 98.49 3.73 MBB + BoTNet 95.82 – – 4.73 MBB + BoTNet – 77.37 – 5.71 MBB + BoTNet – – 98.86 3.52 MBB + BFNTBlock \\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}$$_{GN}$$\\end{document} GN -P1 96.06 – – 4.73 MBB + BFNTBlock \\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}$$_{GN}$$\\end{document} GN -P3 – 77.82 – 5.71 MBB + BFNTBlock \\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}$$_{GN}$$\\end{document} GN -P1 – – 99.29 3.52"
} | 13,722 |
34577769 | PMC8466037 | pmc | 414 | {
"abstract": "With advances in internet of things technology and fossil fuel depletion, energy harvesting has emerged rapidly as a means of supplying small electronics with electricity. As a method of enhancing the electrical output of the triboelectric nanogenerator, specialized for harvesting mechanical energy, structural modification to amplify the input force is receiving attention due to the limited input energy level. In this research, a lever structure was employed for delivering the amplified input force to a triboelectric nanogenerator. With structural optimization of a 2.5 cm : 5 cm distance ratio of the first and second parts using two lever structures, the highest electrical outputs were achieved: a V OC of 51.03 V, current density of 3.34 mA m −2 , and power density of 73.5 mW m −2 at 12 MΩ in the second part. As applications of this triboelectric generator, a vertical vibration sensor and a wearable reloading trigger in a gun shooting game were demonstrated. The possibility for a wearable finger bending sensor with low-level input was checked using a minimized device. Enhanced low-detection limit with amplified input force from the structural advantage of this lever-based triboelectric nanogenerator device can expand its applicability to the mechanical trigger for wearable electronics.",
"introduction": "1. Introduction As the demand for electrical energy is continuously growing with the increasing number of wearable electronics and internet of things (IoT) devices, developing and using novel energy sources are recently attracting attention [ 1 , 2 ]. Additionally, the depletion of fossil fuels is accelerating the replacement of the fossil fuel-based conventional energy sources [ 3 , 4 , 5 ]. Therefore, multitudinous researchers have been focusing on how to harvest the ambient energy. A lot of energy is wasted in the form of heat, light, and sound [ 6 ]. In operating small electronics, the scavenged vibrational energy from humans, machines, and phenomena of nature can highly increase the operating time with the use of conventional batteries [ 7 ]. Accordingly, the energy harvesting technology can be an effective way to reuse the wasted energy. In energy harvesting technology which captures and stores energy from external sources, solar cells (SCs), electromagnetic generators (EMGs), thermoelectric generators (ThEGs), piezoelectric nanogenerators (PENGs), and triboelectric nanogenerators (TENGs) are considered to be mainstream technologies. The SC harvests the light energy from the sun or other light sources with the principle of the photoelectric effect [ 8 ]. Direct current (DC) can be generated with this solar cell device mainly on sunny days or on cloudy days, even if its output would not be the best performance. The ThEG scavenges the thermal energy using the principle of the Seebeck effect [ 9 , 10 ]. Based on the temperature difference maintained between the top and bottom sides of the device, the ThEG can also generate stable DC output. EMGs, which consist of coils and magnets, have been widely used for generating electricity ever since Michael Faraday discovered the operating principle of the EMG in 1831 [ 11 , 12 , 13 ]. EMGs generate an alternate current by changing the magnetic flux in the coil, showing the high energy conversion efficiency in a large-scale device [ 14 , 15 ]. As mechanical energy harvesters, PENGs with using a piezoelectric material [ 16 , 17 , 18 ] and TENGs operating with a simple contact-separation motion [ 19 ] are widely used these days. Both of these mechanical energy harvesters have been used for generating electricity or sensing an external stimulus. These harvesters also present the advantages of light weight and possibility of miniaturization. In particular, the TENGs, which operate on the principle of the conjunction of contact electrification and electrostatic induction, show distinctive advantages in the easy material selection, simple structure, and high output power compared to other small energy harvesters [ 20 , 21 ]. From the easy material selection property, low-cost and eco-friendly materials can be adopted in parts of the TENG device. Moreover, the device can be easily fabricated without using complicated equipment. To enhance the electrical output of the TENG without hybridizing, several methods were reported: material selection [ 22 , 23 , 24 , 25 , 26 ], contact surface modification [ 27 , 28 , 29 , 30 , 31 ], using an external circuit [ 32 ], and structural modification [ 33 , 34 , 35 ]. For example, a simple method to enhance the electrical output of the TENG is adopting two materials far from each other in the triboelectric series. Increasing the contact surface area of the dielectric or that of the counter material is another method for increasing the output in the first stage. Recently, some researchers have succeeded in generating high electrical output by using a power management integrated circuit (PMIC) as an auxiliary circuit [ 36 , 37 , 38 ]. Moreover, structural modification techniques were also adopted to effectively convert the limited input source into electrical energy. In this paper, using the structural modification method, the electrical output of the TENG is increased with a simple machine structure of a lever. The lever consists of a bar to inject the force into the object and a fulcrum to transfer the injected force. When using the lever structure, higher output force can be applied to the object with smaller input force when the fulcrum is located closer to the object and farther from the injecting point at once. When this structure is applied to the TENG, the double electrical output from the first (hereafter ‘1st’) part (directly applied) and the second (hereafter ‘2nd’) part (beyond the fulcrum) of lever-based TENGs (L-TENGs) can be generated. With a commercial force sensor, the optimized point that can generate more output force can be found. The output from the 2nd part of the TENG can be increased with the optimized distance ratio of the fabricated device. Moreover, the fabricated L-TENG device can detect the small input force from the electrical output signals with the boosted output force at the 2nd part. Inductively coupled plasma (ICP) etching is applied to the contact-dielectric layers, further enhancing the electrical output of the L-TENG [ 39 ]. A finite element method (FEM) can be utilized to check the electrical potential distribution with variation of the distance ratio. Several electrical output characteristics of the frequency response, durability, and charging capacitors are analyzed to use this L-TENG as an energy harvester. LEDs are also illuminated by connecting and operating the L-TENG device to represent the ability for driving electronic devices. With additional structural advantages, this device can be applied to a vibration sensor and a wrist bending sensor with the enhanced accuracy from the output boosting characteristic. This L-TENG will open up a new path for enhancing energy harvesting efficiency and additional use of TENG devices as wearable sensors.",
"discussion": "4. Discussion In this study, a lever-based TENG was fabricated by simple 3D printing technology and analyzed with force and electrical outputs. The two double-electrode TENGs were adopted in the 1st part- and 2nd part-TENGs and ICP etching was conducted on the dielectric layer of the PTFE film. The FEM simulation presented higher electric potential in a shorter distance between the axis and the TENG due to the stronger contact force. The compressing force of the lever structure shows a higher value inversely proportional to the distance from the fulcrum using the ‘U’ shaped L-TENG device. With the use of the force sensor, the input force to each part of the TENG was measured with a 2 mm-gap of the 2nd part-TENG. The 2 mm gap case showed the highest input force in the 2nd part due to the effective separation and transmitted force. Moreover, the flat L-TENG device with the distance ratio of 2.5 cm of the 1st part and 5 cm of the 2nd part displayed optimal electrical outputs of the V OC of 51.03 V and a current density of 3.34 mA m −2 in the 2nd part-TENG and a V OC of 9.33 V and a current density of 0.199 mA m −2 in the 1st part-TENG. Each part of the TENG presented power density of 0.92 mW m −2 at 50 MΩ and 73.5 mW m −2 at 12 MΩ in the 1st part- and 2nd part-TENGs, respectively. With this condition, the 1st part- and 2nd part-TENGs showed the frequency response with gradients of 12.2 nA Hz −1 and 42.1 nA Hz −1 , respectively. The durability of the 2nd part-TENG was verified with 2.5% degradation after being measured for 25,200 cycles. Combining the electrical outputs from the 1st part- and 2nd part-TENGs, the output voltage through the charging capacitor increased compared with the separated case and it can turn on 24 green LEDs. As practical applications, a vertical vibration sensor with a sensitivity of −0.4 Hz −1 and a wrist bending sensor trigger in a gun shooting game were designed with the use of the L-TENG. As a finger bending sensor, the minimized L-TENG showed the considerably low-detection limit of input force and the ability to distinguish between the weak and strong bending conditions. This L-TENG shows potential for enhancing the output from the lever structure and can be applied to a wearable sensor with the enlarged detection range from the structural advantage of the lever."
} | 2,352 |
28542245 | PMC5443480 | pmc | 415 | {
"abstract": "Abstract Given the numerous threats against Earth’s coral reefs, there is an urgent need to develop means of assessing reef coral health on a proactive timescale. Molecular biomarkers may prove useful in this endeavor because their expression should theoretically undergo changes prior to visible signs of health decline, such as the breakdown of the coral-dinoflagellate (genus Symbiodinium ) endosymbiosis. Herein 13 molecular- and physiological-scale biomarkers spanning both eukaryotic compartments of the anthozoan- Symbiodinium mutualism were assessed across 70 pocilloporid coral colonies sampled from reefs of Fiji’s easternmost province, Lau. Eleven colonies were identified as outliers upon employment of a detection method based partially on the Mahalanobis distance; these corals were hypothesized to have been displaying aberrant sub-cellular behavior with respect to their gene expression signatures, as they were characterized not only by lower Symbiodinium densities, but also by higher levels of expression of several stress-targeted genes. Although these findings could suggest that the sampled colonies were physiologically compromised at the time of sampling, further studies are warranted to state conclusively whether these 11 scleractinian coral colonies are more stress-prone than nearby conspecifics that demonstrated statistically normal phenotypes.",
"introduction": "Introduction Earth’s coral reefs are currently threatened by a number of anthropogenic insults [ 1 – 2 ], most notably global climate change [ 3 – 4 ]. There is consequently an urgent need to develop means of assessing coral health on a proactive timescale [ 5 ]. Unfortunately, traditional coral reef surveys (e.g., [ 6 ]) involve the documentation of dead or dying corals; although the ensuing data are indeed of interest to managers, they come too late to benefit the resident corals. Ideally, an assessment of coral health could be made prior to visible manifestations of stress, such as bleaching, whereby the coral-dinoflagellate (genus Symbiodinium ) endosymbiosis that serves as the foundation of all coral reefs deteriorates [ 7 ]. Molecular biology-based approaches have shed light on numerous aspects of the fundamental and stress biology of anthozoan-dinoflagellate endosymbioses [ 9 – 12 ], and molecular biomarkers [ 8 ], in particular, may hold promise for coral health diagnostics since they require only a single sampling event; therefore, they could theoretically be used to make inferences about coral physiology prior to visible signs of stress, such as bleaching. A particular stress protein, for instance, may demonstrate up-regulation in response to an environmental shift well before any loss of Symbiodinium from the coral gastrodermal tissues. If a significant proportion of a reef’s corals are expressing highly abnormal levels of well-validated biomarkers, then it is conceivable that a manager could be alerted to attempt to ameliorate the impact of the local-scale stressors (e.g., water pollution) in order to promote coral resilience. To validate a potential biomarker, such as an mRNA, one would ideally collect data from control specimens to establish a “normal” concentration level. However, even if traditional tank studies are employed (e.g., [ 13 ]), what is considered a control expression level of a biomarker in one region may be aberrant in another. It should be noted here that “aberrant” does not refer to health, but only to divergence from a norm/average. Unfortunately, the fact that no reefs on Earth are devoid of any human impact precludes the ability to simply present typical ranges for each marker (above or below-which signifies stress) in the absence of data from corals sampled prior to the Industrial Revolution. As potential evidence for this, high expression levels of stress genes have been measured in corals from some of the most remote, least populated regions of the Pacific Ocean, such as the Austral and Cook Islands [ 14 ]. This phenomenon was hypothesized previously [ 15 – 16 ] to represent mRNA “front-loading,” whereby high expression levels of mRNAs encoding stress proteins (e.g., heat shock proteins [HSPs][ 17 ]) occur at all times in order for the corals to have the capacity to rapidly translate such stress proteins when temperatures change abruptly due to, for instance, upwelling [ 18 – 19 ]. However, corals of the Austral and Cook Islands experience relatively low and stable temperatures [ 20 ], suggesting that this may not only be a strategy employed by corals residing within thermally extreme and dynamic environments (such as those of Southern Taiwan [ 15 ]). It is worth noting that corals are amongst the only organisms currently known to exhibit such an “always stressed” phenotype given the significant cellular energy expenditure required to do so [ 21 ]. Despite issues with using absolute expression levels of individual genes or proteins to predict whether or not a coral is stressed, it is possible that multivariate statistical approaches (MSA) could nevertheless be used to identify corals behaving significantly differently from what is normal in a particular region. Colonies displaying statistically unusual phenotypes may ultimately be found to be those either experiencing stress or, in contrast, those of enhanced resilience (assuming the front-loading hypothesis to be true). To test the notion that molecular biomarkers could be used to identify aberrantly behaving coral colonies, the model coral Pocillopora damicornis [ 22 – 25 ] was targeted across reefs of Fiji’s frontier province, Lau (Figs 1 and 2 ), and 13 molecular-physiological response variables were measured in each colony. MSA were used to analyze the dataset and identify outliers, and it was hypothesized that certain environmental parameters might significantly influence outlier frequency; for instance, it was predicted that corals displaying statistically aberrant behavior would be more likely to be found on reefs with higher temperatures and light levels. Collectively, it was hoped that this MSA-based approach for assessing the environmental physiology of this model reef coral could serve as a conceptual platform for others looking to use, in particular, molecular biology-driven approaches for not only identifying outliers, but also for simply establishing baseline functional data for invertebrate-dinoflagellate endosymbioses in understudied regions of the Indo-Pacific. 10.1371/journal.pone.0177267.g001 Fig 1 Map of Fiji’s Lau Province and pie graphs depicting the proportional genetic breakdown of the pocilloporids sampled. PA = P . acuta (red; Sebastian Schmidt-Roach [SSR; taxonomic authority] genotype β). PB = P . brevicornis (blue). PD = P . damicornis (black; SSR genotype α). PM = P . meandrina (orange). PV = P . verrucosa (green). The color codes for the five species are used throughout all of the manuscript’s figures. All images of the sampled colonies (including “macro” images of the polyps) can be found on the following website: http://coralreefdiagnostics.com under the “Fiji” sub-heading. 10.1371/journal.pone.0177267.g002 Fig 2 Maps of nine of the islands/atolls within Fiji’s Lau Archipelago whose reefs were surveyed. Two islands, Vanua Vatu (June 15, 2013) and Nayau (June 16, 2013), have not been depicted, as the corals sampled were not processed in full for all response variables. For a detailed description of the reefs of the former, including high-resolution maps, please see Saul and Purkis [ 26 ]. Site 53 (Vanua Balavu) has not been labeled due to its proximity to site 56 (see GPS coordinates in Table 1 .).",
"discussion": "Discussion The relationship between Symbiodinium density and gene expression A combination of univariate and multivariate statistical approaches were used herein to uncover corals displaying statistically aberrant behavior. Interestingly, two exploratory approaches used to depict variation in the dataset and similarity amongst samples (PCA and MDS, respectively) were able to uncover several colonies positioned away from the normal “core” physiological response region; all such samples were ultimately found to be outliers based on a more quantitative approach featuring the Mahalanobis distance and the heat map score. These 11 outliers had ~30% lower Symbiodinium densities and 3- to 4-fold higher stress gene expression levels than non-outliers, and the negative relationship between Symbiodinium density and expression of both hsp90 and ubiq-lig was much more pronounced for outliers than non-outliers. Reef-building corals require high densities (~10 6 cells cm -2 ) of Symbiodinium to maintain their metabolic needs [ 36 ]. As environments change, particularly with respect to temperature, low levels of bleaching can take place [ 37 ], resulting in lower densities of Symbiodinium in hospite . Such a hypothetical, bleaching-inducing environmental change would likely also affect cell physiology, specifically the expression of genes encoding stress proteins (such as hsp90 and ubiq-lig ). Therefore, it is unsurprising that stress-sensitive genes were expressed at higher levels in corals exhibiting lower Symbiodinium densities. Outlier frequency and environment In contrast to what was hypothesized, there was no effect of any environmental parameter on outlier frequency; corals displaying statistically aberrant behavior were just as likely, for instance, to be found in the lagoon as on the fore reef. In fact, there were numerous instances in which outliers were sampled from a reef in which normally behaving corals were also sampled. For instance, of the six corals sampled at site FJMT13 (Matuku), only one was considered an outlier (sample 25). Another member of the same species ( P . verrucosa ) of nearly identical size was collected within ~100 m at nearly the same depth and light level, and this sample was deemed normal with respect to the 11 molecular physiological response variables. Therefore, it is possible that intra-site environmental variation, or, alternatively, differing life histories, led to the aberrant behavior of colony 25, which appeared normally pigmented (albeit only 7-cm long and therefore presumably young). There was only one site in which multiple outliers were sampled: FJFL29 (a lagoonal patch reef at Fulaga); although only one such outlier is listed for Fulaga (colony 56) in Table 3 , when looking at the heat map scores, it is clear that colony 54 was likely an outlier. It could not be labeled as such because of to the inability to calculate its respective Mahalanobis distance due to poor DNA extraction efficiency. Therefore, JMP’s “multivariate normal” imputation algorithm featuring a shrinkage estimate was used to impute missing data (with off-diagonals scaled by a factor of 0.75), and the corresponding Mahalanobis distance for sample 54 was 8.6, the highest value in the entire dataset. Samples 54 and 56 were collected within 10 minutes of each other (~16:00) in <3 m of water in an area characterized by such high sediment loads (not quantified) that visibility was <1 m; such sedimentation may have contributed to the aberrant sub-cellular behavior of these samples either directly (e.g., by smothering the tissues and therefore necessitating a stress response [thereby affecting ubiq-lig ]) or indirectly (e.g., via modification of the corals’ light environment [thereby affecting rbcL ]). Biomarker profiling in the absence of pristine control reefs Comparisons with biomarker expression signatures of samples from controlled tank studies conducted elsewhere (e.g., [ 38 – 41 ]) are risky, as what is considered a control level of expression for a certain target molecule in a region like Taiwan, whose reefs abut some of the world’s highest human population densities, may actually be “stress-indicative” in a place like Lau Province; although the reefs of the Lau Archipelago are far from pristine due in part to a virtual absence of sea cucumbers from over-harvesting by Chinese fleets (unpublished data), the region is only sparsely populated (~11,000 people across the 60 islands, only about half of which are populated). However, as documented in Southern Taiwan [ 15 ] and even uninhabited South Pacific atolls (e.g., Maria Atoll, French Polynesia; [ 14 ]), all 70 of the Lau samples expressed high levels of stress marker genes, including the Symbiodinium stress genes ubiq-lig and hsp90 (but not apx1 ) and the host coral stress genes cu-zn-sod and gfp-cp . Although the latter is not a classical stress gene, per se, in corals it is known to be up-regulated at high PAR levels [ 42 ]; the respective chromoproteins absorb excess light that might otherwise lead to photoinhibition [ 43 ] and consequently bleaching. Whether or not these generally high levels of expression of genes encoding stress proteins indicates that these corals were indeed stressed at the time of sampling or were, alternatively, better prepared for future environmental changes (as discussed in the Introduction), remains to be determined. Furthermore, little congruency between mRNA and protein expression was documented for another reef-building pocilloporid [ 40 ]; therefore, it is possible that the respective proteins may show entirely different expression patterns. If nine of ten corals on a reef display very similar molecular phenotypes, whereas the latter (i.e., the “outlier”) is characterized by a completely different one, this does not necessarily mean that the outlier is stressed and the former nine colonies are healthy; such a guess could only be made if there existed a detailed knowledge of the environmental history of the samples, or, alternatively, if the corals’ growth and reproductive output were monitored over a multi-week timescale. Therefore, it is not currently possible to state whether the 11 outliers identified herein were stressed, despite their being characterized by lower Symbiodinium densities and higher stress gene expression; it can only be stated with the data in hand that they were behaving significantly differently from the other 59 colonies analyzed. It would be fruitful to return to the same sites at which both outliers and non-outliers were found, such as FJFL29, during a period of anomalously high temperatures to see if the outliers are actually of diminished resilience to environmental change than the conspecific non-outliers. If such were found to be the case, the validity of this approach for use as a coral stress test would be substantiated. Future directions in coral health assessment Although at current average costs (~US$150/sample, excluding bioinformatics costs), next generation sequencing technology was prohibitively expensive for analyzing these 70 samples (in comparison to ~US$30/sample spent to assess the 11 molecular-scale response variables herein), sequencing prices will continue to drop, and so it may ultimately be possible to profile entire transcriptomes of several dozen, or even hundred, coral samples in the coming years at reasonable costs. Then, it may be found that there are mRNAs that are only expressed by truly stressed colonies (as determined by tank experiments in which growth, reproductive output, and Symbiodinium densities/pigmentation are also tracked over a long-term timescale); if such a result could be corroborated in conspecifics sampled across numerous study sites around the globe, then it is conceivable that a molecular biomarker assay for conclusive determination of coral health could ultimately be developed. In that case, the conceptual and statistical framework reported herein could be used in conjunction with such biomarkers in order to not only label a coral as being an outlier or not, but, more specifically, to assign each coral sample of interest a health index score (e.g., from 1 to 10, with 1 being nearly dead and 10 being healthy)."
} | 3,975 |
36249713 | null | s2 | 416 | {
"abstract": "Discovered in the 1920's, polyhydroxyalkanoates (PHA) are a naturally occurring class of biopolyesters that have long been touted as a renewable, biodegradable plastic alternative. Demand for sustainable products and over a half century of research have led to moderate commercial success of PHA. Yet, these materials are not pervasive. Therefore, an important question to address is, \"what is the barrier that prevents widespread application of these materials?\" PHA can be made from an incredibly diverse class of monomers that incorporate both simple and complex organic acids. Herein, we provide an updated list of unique PHA monomers that are substrates for a PHA polymerase. Unfortunately, most unique monomers are incorporated only after feeding a structurally related feedstock to a PHA accumulating bacterium. Therefore, we put forward an argument that research must now turn to developing feedstock-independent, synthetic pathways to produce an increased diversity of PHAs capable of competing with petroleum-derived plastics."
} | 259 |
24453885 | PMC3881690 | pmc | 420 | {
"abstract": "Application of microbial fuel cells (MFCs) to wastewater treatment for direct recovery of electric energy appears to provide a potentially attractive alternative to traditional treatment processes, in an optic of costs reduction, and tapping of sustainable energy sources that characterizes current trends in technology. This work focuses on a laboratory-scale, air-cathode, and single-chamber MFC, with internal volume of 6.9 L, operating in batch mode. The MFC was fed with different types of substrates. This study evaluates the MFC behaviour, in terms of organic matter removal efficiency, which reached 86% (on average) with a hydraulic retention time of 150 hours. The MFC produced an average power density of 13.2 mW/m 3 , with a Coulombic efficiency ranging from 0.8 to 1.9%. The amount of data collected allowed an accurate analysis of the repeatability of MFC electrochemical behaviour, with regards to both COD removal kinetics and electric energy production.",
"conclusion": "5. Conclusions A laboratory-scale SC-MFC, with air-cathode, operating in batch mode with internal recirculation, was built and studied. The MFC was fed with two different synthetic wastewaters, in order to test exoelectrogenic bacteria behaviour under various conditions. An average COD removal efficiency of 86% was achieved, with mean waste retention time (HRT) of 150 hours. The MFC produced an average power density of 13.2 mW/m 3 , with peaks of 20–30 mW/m 3 . Electrical energy recovered amounted to 7.9 kJ per m 3 of treated waste. Coulombic efficiency was in the range 1-2%, with mean value of 1.2%. These values are lower than those achievable by chemical Fuel Cells; however, this is of relative importance since the intended fuel (urban wastewater) in this case is actually a waste that must be disposed of at a nonnegligible cost. The synthetic wastewater adopted (SW2), very similar to actual urban wastewater, showed good results in terms of experimental repeatability and will be useful for future investigations on MFC process and for benchmarking of the process when shifting to actual urban wastewater. Synthetic waste allowed also to test continuous flow operation showing an improvement of electric behaviour over time. This suggested that a well-designed continuous flow plant could ensure better bioelectrochemical performances than abatch one, once in steady-state conditions. The results of this study will be applied to improve the design of the tested MFC and to switch to actual urban wastewater substrate operation. Although electric power generation was modest, this study shows that MFCs are feasible, although in need of improvement, process for urban wastewater treatment allowing direct energy recovery from a waste source.",
"introduction": "1. Introduction Fuel cells are devices that use a combustion reaction without resorting to a thermal process, thus achieving direct conversion of chemical energy (of a generic “fuel” or “substrate”) into electrical energy. In particular, microbial fuel cells (MFCs) directly convert the chemical energy, contained in an organic bioconvertible substrate, into electrical energy, through the mediation of exoelectrogenic bacteria that act as catalyser of the half-reaction of substrate oxidation [ 1 , 2 ]. The first evidence of this phenomenon was discovered in 1911 by Potter [ 1 ], but very few practical advances were achieved in the field until the first patent of mediatorless MFCs, dated 1999 [ 3 ]. MFCs-like technology has already been used as an energy source in field applications such as environmental sensors or process biomonitoring [ 4 – 8 ]. Some applications to brewery wastewaters treatment are also reported [ 9 , 10 ]. Application of MFCs to municipal wastewater treatment appears to provide a potentially attractive alternative to traditional treatment processes that may include indirect energy recovery from wastes (e.g., anaerobic digestion with methane fermentation), as these devices are suitable to operate with low concentration substrates and temperatures below 20°C [ 11 ]. Although MFCs were tested for the first time in 2004 by Liu et al. [ 12 ], as of today their main applications remain confined to laboratory-scale plants. The limiting factors for MFC application to natural scale plants are, in fact, high initial capital costs (especially for electrode construction and membranes) and the limited power density that can be achieved [ 13 , 14 ]. MFC's working principle relies on splitting the semireactions of oxidation and reduction that make up a typical redox reaction, allowing them to occur in two different compartments. In the anodic compartment, exoelectrogen bacteria catalyse substrate oxidation and transfer the electrons, released from cellular respiratory chain, to a metal electrode (i.e., anode) [ 14 ]. Electrons then flow through an external electric circuit towards the cathodic compartment, where they reduce the terminal electron acceptor (TEA, usually oxygen) [ 15 ]. For each electron released at the anode, an H + ion must reach the cathode through the electrolytic solution saturating the cell, in order to internally close the circuit and reestablish electric neutrality. Hence electrons and protons react with oxygen on the cathode, generating water vapour H 2 O [ 15 ]. The maximum current that can be produced by a MFC depends on the actual rate of substrate biodegradation, whereas maximum theoretical cell voltage (also called electromotive force or emf ) depends on Gibbs free energy of the overall reaction and can be calculated as the difference between the standard reduction potentials of the cathodic oxidant (oxygen) and the chosen anodic substrate (e.g., as per reference [ 16 ]) [ 2 , 14 , 17 ]. However, the cell's emf is a thermodynamic value that does not take into account any internal losses [ 17 ]. Measured experimental values are always substantially lower than theoretical ones. The typical MFC configuration is that of a dual-chamber cell, where the anodic and cathodic volumes (chambers) are separated by a protonic or cationic exchange membrane (PEM or CEM), which allows internal ionic fluxes but prevents mixing of anodic reducing solution and cathodic oxidant [ 17 ]. This membrane, however, is one of the principal cost factors in a MFC plant and increases the cell's internal resistance (i.e., the measurement of all internal voltage losses, occurring when current flows throughout the system) [ 13 , 14 , 17 ]. For this reason, current research on MFCs has shifted towards the use of single-chamber, membraneless cells (so called SC-ML-MFCs), where the cathode is directly exposed to the atmosphere (so called air-cathode) [ 13 ]. Dual-Chamber MFCs are usually still investigated when the specific aim is to exploit the cathodic reduction semireaction, for the removal of nutrients from wastewater or groundwater lacking organics (e.g., [ 18 – 20 ]). In the case of SC-MFCs, the cathode has proved to be the most critical component of the process. The cathode must in fact provide the interface between three separate phases: the oxidant gas (atmospheric oxygen), the liquid electrolyte (containing the moving H + ions), and the solid conductor (external circuit), through which electrons flow. It is therefore likely to be the cathode limiting electrode for power generation [ 13 ]; several studies looked at ways to improve its electrical performances, avoiding at the same time the adoption of expensive chemical catalysts and/or ionic exchange membranes/resins [ 21 – 25 ]. Determination of the optimal material for electrode construction and definition of the most suitable dimensional ratios between electrode surfaces and cell volume are still object of investigation [ 13 ]. 1.1. MFC Reactor Parameters In order to compare the performance of different reactor configurations and electrodes, a series of parameters and experimental methods have been proposed, to determine the bioelectrochemical performances of MFCs. One of the more encompassing reviews on this subject has been written by Logan et al. [ 17 , 26 ]. Firstly, the mean power (and current) produced by the cell must be normalized by a relevant geometric characteristic of the reactor that could be one of the electrodes' surface area or the volume of anodic chamber (when dealing with SC-MFCs). In this study we have chosen to express current density with respect to cathode surface (the limiting electrode in our process) and power density with respect to the total reactor volume. The polarization curve is a synthetic method to analyse the behaviour of a MFC [ 17 ]: the curve represents the dependence of cell's voltage on the electrical current flowing in the circuit and allows to estimate the values of electrode overpotentials and the internal resistance of the cell, representing an overall measurement of cell's internal voltage losses and defined, geometrically, by the slope of the linear region of the polarization curve [ 26 ]. The power curve is calculated from the polarization curve and describes the power output of the cell as a function of the current. Usually it has a parabolic shape with a single point of maximum (called Maximum Power Point or MPP), which occurs when the external resistance of the circuit equals internal resistance of the cell. From the wastewater treatment engineer's point of view, it is possible to evaluate the substrate conversion rate of a MFC, in terms of chemical oxygen demand (COD), through the determination of the COD removal efficiency or, better, of its removal rate (thus taking into account the retention time of the substrate in the cell). Finally, an important parameter for the evaluation of MFCs performance is its Coulombic efficiency, defined as the ratio of actual transferred electric charge and its maximum value obtainable, if all of the substrate's removal were to produce a current [ 17 ]. In this study, a laboratory-scale SC-MFC, with an air-cathode of novel design, operating in batch mode with internal recirculation was built and operated. The cell was sequentially fed with different wastewaters, both synthetic and natural, in order to test exoelectrogenic bacteria behaviour under various substrate load conditions. The aims of the study conducted were primarily to characterize the cell under both an electrical and a substrate removal point of view, through the construction of polarization and power curves, determination of COD removal rates, mean electrical power, and Coulombic efficiencies.",
"discussion": "4. Results and Discussion 4.1. Testing with SW1 Wastewater After the initial biomass growing period, the MFC was fully monitored for two weeks while feeding the anodic chamber with simple synthetic wastewater containing CH 3 COONa as the only oxidable compound (SW1). Theoretical emf , calculated assuming COD concentration of 300 mg/L, neutral pH, and temperature of 23°C (maintained constant by means of a thermostat), was equal to 1.1 V [ 26 ]. Experimental values of our MFC voltage at the external resistance of 1000 Ω (hereafter E \n 1000 ) were always substantially lower ( Table 2 ). Reasons for this probably consisted of high electrodes overpotentials (i.e., activation losses) and low ionic strength of the wastewater, indeed the linear region of polarization curve exhibited only a moderate gradient, which resulted in values of R \n int comprised between 49 and 630 Ω ( Figure 3 ). From Figure 3 it is clear that R \n int increased as the residence time of wastewater in the cell. That was mainly due to reduction of maximum extractable current, rather than Open Circuit Voltage variation. E \n 1000 values remained quite constant around 220–270 mV, at least for the first 40 hours. Table 3 shows final results of the continuous MFC monitoring, throughout the two batch treatment cycles performed with SW1. Both T.C.1 and -2 exhibited a rapid power drop after the first 72 hours of operation ( Figure 4 ), when the energy recover had already reached 80% of its final value, though wastewater COD was still higher than 100 mg/L. T.C.2 showed a mean power output of 8.7 mW/m 3 , 58% higher than T.C.1, but this can be ascribed to a higher COD concentration in the influent, as both the kinetic constant and Coulomb efficiency determined for these cycles assumed almost the same values. 4.2. Testing with SW2 Wastewater After the first two weeks, three treatment cycles with a more complex synthetic waste (SW2) were carried out. SW2 was designed to mimic behaviour of a natural substrate [ 32 ], but without the presence of toxic compounds and with limited COD oscillations. The cell was not completely emptied and cleaned at the end of each cycle but simply filled with anodic biomass and a concentrated dose of wastewater. This is in order to simulate a continuous operation of the cell, attempting to achieve maximum biomass growth at the anode and, eventually, maximum concentration of endogenous electron mediators in wastewater. Results are summarized in Table 4 (polarization curves) and Table 5 (continuous monitoring). Polarization curves showed how MFC electric behaviour improved by not emptying and cleaning the cell ( Figure 5 ) but finally resulted in a P \n max of 25.9 mW/m 3 , almost the same value reached with SW1 under previous operating conditions. Although from an analysis of the polarization curves it seems that the cell behaviour was not overly affected by the adopted feed waste composition, continuous monitoring revealed that both biological and electrochemical behaviour improved considerably utilizing SW2. The first order kinetic rate constant, b , and the COD removal efficiency reached very stable values ( Table 5 ), while the average power production and Coulombic efficiency increased from one cycle to the next, proving how the MFC's biomass progressively improved its exoelectrogen characteristics."
} | 3,460 |
26406279 | PMC4598564 | pmc | 421 | {
"abstract": "Triboelectric nanogenerators have been invented as a highly efficient, cost-effective and easy scalable energy-harvesting technology for converting ambient mechanical energy into electricity. Four basic working modes have been demonstrated, each of which has different designs to accommodate the corresponding mechanical triggering conditions. A common standard is thus required to quantify the performance of the triboelectric nanogenerators so that their outputs can be compared and evaluated. Here we report figure-of-merits for defining the performance of a triboelectric nanogenerator, which is composed of a structural figure-of-merit related to the structure and a material figure of merit that is the square of the surface charge density. The structural figure-of-merit is derived and simulated to compare the triboelectric nanogenerators with different configurations. A standard method is introduced to quantify the material figure-of-merit for a general surface. This study is likely to establish the standards for developing TENGs towards practical applications and industrialization.",
"discussion": "Discussion We have developed methods for standardized evaluations on the performance of TENGs. Starting from the built-up voltage V -transferred charge Q plot, the CMEO with infinite load resistance was derived to have the maximized output energy per cycle, which represents the maximum energy production of TENG, similar to the Carnot cycle in heat engines. On the basis of the maximum output energy per cycle, and considering both the maximized energy-conversion efficiency and the maximized average output power, the FOM P was derived to evaluate each TENG design, composed by a structural FOM and a FOM M . The structural FOMs for different structures of TENGs were simulated by analytical formulae and FEM, respectively, showing the maximum value of structural FOM for each TENG structure. The standard evaluation of the FOM M was also demonstrated by measuring triboelectric surface charge density via contacting the materials with liquid metals, and then the normalized triboelectric charge density and FOM DM were defined and derived for various materials. The standards and evaluation methods provide here set the foundation for the further applications and industrialization of TENG technology."
} | 576 |
27420605 | PMC5241762 | pmc | 422 | {
"abstract": "ABSTRACT The synthesis of renewable bioproducts using photosynthetic microorganisms holds great promise. Sustainable industrial applications, however, are still scarce and the true limits of phototrophic production remain unknown. One of the limitations of further progress is our insufficient understanding of the quantitative changes in photoautotrophic metabolism that occur during growth in dynamic environments. We argue that a proper evaluation of the intra- and extracellular factors that limit phototrophic production requires the use of highly-controlled cultivation in photobioreactors, coupled to real-time analysis of production parameters and their evaluation by predictive computational models. In this addendum, we discuss the importance and challenges of systems biology approaches for the optimization of renewable biofuels production. As a case study, we present the utilization of a state-of-the-art experimental setup together with a stoichiometric computational model of cyanobacterial metabolism for quantitative evaluation of ethylene production by a recombinant cyanobacterium Synechocystis sp. PCC 6803.",
"conclusion": "Conclusions: Toward systematic strain improvements The challenges of the 21 st century, including mitigation of climate change, food and fresh water supply security or sustainable energy development make the domestication of cyanobacteria as a human resource of high importance. Large-scale biotechnological applications using photosynthetic prokaryotes, however, are still in their infancy. We argue that further integration of metabolic modeling with high-quality measurements under controlled conditions will result in significant advances in our understanding of phototrophic production limits. Beyond phenomenological analysis, the combination of experimental data with theoretical modeling allows to identify key principles of rational metabolic engineering strategies - as shown in the recent study aiming at identification and characterization of suitable strain design strategies for cyanobacterial bioproducts synthesis. 24 In case of ethylene bioproduction, significant improvement of the strains productivity has been reported. 20 However, the ethylene yields are still significantly lower when compared to heterotrophic cultivations 2 as well as when compared to yields of other bioproducts such as ethanol or lactic acid. 16 It also remains unknown to what extent industrial production is feasible, since reliable data for large-scale cultivation are not available. The only preliminary study evaluating large-scale ethylene bioproduction only pointed out the need of usage of sea water and coastal areas for operation of production plants to achieve positive environmental impact. 27 While many challenges remain, we conjecture that theoretical modeling, in conjunction with experimental data, will ultimately result in rational strain design strategies that consider the host cell and its environment as interacting system - and thereby enable green biotechnology for the 21 st century.",
"introduction": "Introduction In our recent work, we characterized the impact of different light intensities on ethylene production by a recombinant cyanobacterium Synechocystis sp. PCC 6803. Ethylene is among the most widely used compounds in chemical industry and is currently mainly derived from fossil resources. A technology for its renewable production is therefore highly desirable. In addition to the formation in plants, ethylene is produced naturally by various microorganisms either by oxidation of 2-keto-4-methylthiobutyric acid or by utilization of α-ketoglutarate and arginine as substrates in a reaction catalyzed by the ethylene-forming enzyme (EFE). 1 The EFE was previously expressed in Escherichia coli, Saccharomyces cerevisiae, Trichoderma viride and Trichoderma reesei . 2 A promising alternative to heterotrophic microbial synthesis is the light-driven conversion of atmospheric CO 2 into ethylene by photosynthetic microorganisms. The EFE has been introduced in the cyanobacterial strains Synechococcus elongatus sp. PCC 7942 and in Synechocystis sp. PCC 6803, 2 resulting in stable production rates of approximately 200 nL (C 2 H 4 ) mL culture −1 h −1 OD 730/750 −1 with the highest reported productivity of 171 mg (C 2 H 4 ) L culture −1 d −1 achieved using dense cultures. 3 To facilitate sustainable and economically viable production, however, further improvements and novel strategies with respect to yield, genetic stability, and robustness of production strains are urgently needed. In this Addendum, we argue that overcoming the current challenges with respect to cyanobacteria as a resource for bioproducts requires a better integration of quantitative measurements (using highly-controlled bioreactor setups) together with computational analysis and quantitative models at both single-cell and culture levels. As yet, systematic attempts to identify and quantify possible intra- and extracellular limitations of cellular production are still in their infancy and computational frameworks allowing identification and ranking of suitable modifications for phototrophic production improvements are still very rare. 4,5"
} | 1,297 |
37854702 | PMC10579436 | pmc | 423 | {
"abstract": "Summary Methanogenesis allows methanogenic archaea to generate cellular energy for their growth while producing methane. Thermophilic hydrogenotrophic species of the genus Methanothermobacter have been recognized as robust biocatalysts for a circular carbon economy and are already applied in power-to-gas technology with biomethanation, which is a platform to store renewable energy and utilize captured carbon dioxide. Here, we generated curated genome-scale metabolic reconstructions for three Methanothermobacter strains and investigated differences in the growth performance of these same strains in chemostat bioreactor experiments with hydrogen and carbon dioxide or formate as substrates. Using an integrated systems biology approach, we identified differences in formate anabolism between the strains and revealed that formate anabolism influences the diversion of carbon between biomass and methane. This finding, together with the omics datasets and the metabolic models we generated, can be implemented for biotechnological applications of Methanothermobacter in power-to-gas technology, and as a perspective, for value-added chemical production.",
"introduction": "Introduction Solutions are needed to mitigate the devastating effects of greenhouse gas emissions, primarily carbon dioxide (CO 2 ), and defossilize the energy and industrial sectors. Societies must efficiently implement: (1) renewable electric power to replace fossil sources; and (2) the use of the emitted CO 2 as a feedstock for the production of commodities within a circular carbon economy. Power-to-gas technologies can convert excess renewable electric power into dioxygen (O 2 ) and molecular hydrogen (H 2 ) via water electrolysis. Thus, power-to-gas systems can provide H 2 to store renewable electric energy but the hydrogen storage and transportation infrastructure is not well-established. 1 , 2 Legal regulatory frameworks typically limit the injection of H 2 into the established natural gas grid to under 10% v/v due to the different physical properties of H 2 and natural gas, which result in limited infrastructure and material compatibility. 1 , 3 , 4 Instead, methane (CH 4 ), the main constituent of natural gas, can be injected into the existing natural gas grid infrastructure to replace fossil natural gas without limitations for storage, distribution, and consumption. 5 , 6 In power-to-gas technology, CH 4 can be derived from an additional methanation step in which the H 2 is combined with CO 2 to produce CH 4 and water ( Equation 1 ). (Equation 1) C O 2 ( g ) + 4 H 2 ( g ) → C H 4 ( g ) + 2 H 2 O ( l ) The methanation step in power-to-gas can be performed thermochemically via the hydrocarbon-forming Sabatier process at high temperatures (>200°C) and pressures (>1 MPa) with metal catalysts (e.g., iron, nickel, cobalt, ruthenium), which are sensitive to gas impurities. 7 , 8 , 9 , 10 Alternatively, the methanation step can be performed biologically with microbes as biocatalysts. Different groups of hydrogenotrophic methanogenic archaea (methanogens) natively metabolize CO 2 and H 2 , producing CH 4 ( Equation 1 ), 6 , 9 , 11 and are therefore being investigated as biocatalysts for the methanation step in power-to-gas technology. 12 , 13 Many methanogens exhibit low growth rates and biomass yields, which limits their potential for large-scale industrial processes. An exception to this are thermophilic methanogenic species of the genus Methanothermobacter , which readily grow in minimal salt media without any complex compounds. Their thermophilic growth temperature coupled with the exothermic methanogenic process results in lower temperature-control demands at industrial scales, 14 explaining the success of Methanothermobacter as production strains in large-scale industrial processes. 13 , 15 , 16 , 17 As an example, the company Electrochaea GmbH is creating an industrial-scale biotechnology platform with Methanothermobacter . A systematic understanding of the biocatalyst’s metabolic capacities is required to advance this biotechnology platform further. Previous studies expanded knowledge on the core metabolism of the genus Methanothermobacter , which includes methanogenesis. 18 Exemplary is an extensive comparative genome study between Methanothermobacter thermautotrophicus ΔH (formerly Methanobacterium thermoautotrophicum ΔH) and Methanothermobacter marburgensis Marburg (formerly Methanobacterium thermoautotrophicum Marburg). 19 However, only limited studies have looked at Methanothermobacter strains systematically in chemostat bioreactors and at the transcriptomic or proteomic level. 20 , 21 , 22 , 23 , 24 , 25 , 26 In fact, more than 500 hypothetical proteins and many pathways have not yet been resolved. 27 Some strains, such as Methanothermobacter thermautotrophicus Z-245 (formerly Methanobacterium thermoformicicum Z-245 18 ), can utilize formic acid (or the deprotonated form, formate) as a sole growth substrate ( Equation 2 ) in addition to CO 2 and H 2 due to the presence of a catabolic formate dehydrogenase. 23 (Equation 2) 4 HCOOH ( aq ) → C H 4 ( g ) + 3 CO 2 ( g ) + 2 H 2 O ( l ) Formic acid is derived from the (electro)chemical reduction of CO 2 and has the advantage of being in the liquid phase. Therefore, in the context of a circular carbon economy, formic acid can serve as a potential intermediate storage molecule in converting CO 2 into valuable chemicals, with formatotrophic (i.e., formic acid-utilizing) microbes as vital biocatalysts. 28 A promising approach to systematically predict phenotypes and identify potential bottlenecks in microbial metabolisms is through mathematical analyses of their potential metabolic networks. 29 Genome-scale network reconstructions represent the theoretical metabolic capacities of a microbe. A genome-scale metabolic model (GEM) is a mathematically constrained reconstruction. These mathematical models enable in-silico investigations, including investigating the effects of gene deletions and insertions for metabolic engineering purposes. Thousands of reconstructions and GEMs have been assembled across all three domains of life; however, archaea remain underrepresented. Only 127 of 6239 (2%) are archaeal GEMs. 30 Of those, only ten models have been manually curated, which is a procedure that leads to higher-quality models. 30 The applications of GEMs and related models have been extensively reviewed, covering methods that are based on GEMs from flux balance analysis to kinetic modeling and machine learning. 31 , 32 , 33 , 34 , 35 , 36 To further optimize methanogens for large-scale applications, metabolic engineering of these biocatalysts and an in-depth understanding of their metabolism is required. However, well-developed genetic tools exist for only a couple of mesophilic methanogens, namely different species of the genus Methanosarcina and Methanococcus maripaludis . 37 , 38 In previous work, we generated a genetic system for M. thermautotrophicus ΔH, 39 , 40 which allows genetically modifying this strain. To use Methanothermobacter strains in more versatile biotechnological applications, such as chemical production, it is necessary to further expand the available systems biology datasets and methodologies. Here, we conduct a multi-disciplinary systematic approach to assess differences in three species of the genus Methanothermobacter : M . thermautotrophicus ΔH, M. marburgensis Marburg, and M. thermautotrophicus Z-245. With this approach, we reveal differences in the formate metabolism of closely related strains, including a genetically engineered strain. 40 We observed that these differences alter the diversion of carbon between biomass and CH 4 production in the three strains. Our observations could, in future work, be exploited to implement metabolic engineering strategies for chemical production with species of the genus Methanothermobacter .",
"discussion": "Discussion Although methanogenesis is well studied, there are still knowledge gaps surrounding methanogenic metabolism. In particular, understanding the metabolic strengths and limitations of different strains is necessary to select the most appropriate strain for a given biotechnological application, such as optimization of CH 4 production or engineering strains to produce value-added products. In this study, we systematically compared M. thermautotrophicus ΔH, M. thermautotrophicus Z-245, and M. marburgensis Marburg to identify differences in their metabolism. Under the conditions of our bioreactor experiments, M. thermautotrophicus ΔH had a higher specific CH 4 production rate than the other strains, while M. marburgensis Marburg reached higher biomass production rates. From our modeling results (with the maximization of the ATPM reaction), we deduced that M. thermautotrophicus ΔH has the highest and M. marburgensis Marburg has the lowest non-growth-associated maintenance energy ( Data S1 - Table S7 ). The non-growth-associated maintenance energy represents the dissipation of ATP (as no storage compounds are known). 50 The higher the non-growth-associated maintenance energy, the more CH 4 (and ATP) must be produced per biomass unit. Based on our multi-omics analysis we propose that differences in anabolism could explain this observation, while further investigations are required to confirm this hypothesis. Our data indicate that the three Methanothermobacter strains might use three different enzymes to produce formate for biomass growth, such as for purine biosynthesis. 43 , 44 , 45 From an ecological perspective, the observed differences in formate anabolism might have an impact on the adaptation to different ecological niches in which, for example, a higher CH 4 -production rate over biomass yield would provide a selective benefit to M. thermautotrophicus ΔH. Furthermore, as ATP is the primary energy currency in the cell, these findings are important to optimize the use of Methanothermobacter strains as cell factories by metabolic engineering in the future. 51 Intriguingly, while M. marburgensis Marburg reached higher biomass production rates ( Figure 1 C), none of the putative enzymes for formate anabolism were abundant in our proteomics analyses. If the low-abundant Fdh is responsible for formate production in M. marburgensis Marburg, questions on the kinetic properties of this enzyme remain. Alternatively, formate could be produced via other, yet unconsidered pathways, for example, through a hydrogen-dependent CO 2 reductase (HDCR)-like activity. 52 While formate is not an intermediate in CO 2 reduction during methanogenesis, a weak formate dehydrogenase activity has been ascribed to the formylmethanofuran dehydrogenases (Fmd/Fwd). 53 Recently, a large enzyme complex was shown to be involved in the electron-bifurcating step of CO 2 reduction without the release of ferredoxin. 54 This enzyme complex may provide the formate that is required for anabolism by an apparent HDCR-like activity. Understanding the metabolism of a microbe is especially important to guide its rewiring for biotechnological purposes. Here, we demonstrated the long-term ability of the strain M. thermautotrophicus ΔH pMVS1111A:P hmtB - fdh Z-245 , encoding the catabolic Fdh cassette from M. thermautotrophicus Z-245 on a plasmid, to grow on formate in chemostat bioreactors. Introducing the Fdh cassette into M. thermautotrophicus ΔH pMVS1111A:P hmtB - fdh Z-245 resulted in a growth behavior that is more similar to wildtype M. thermautotrophicus Z-245 than to M. thermautotrophicus ΔH. This finding supports our hypothesis that, indeed, the different solutions for the three methanogens to produce formate in anabolism have an impact on the CH 4 -to-biomass ratios. The CH 4 -to-biomass ratio is an essential parameter in bioprocessing (it should be high for CH 4 production but low for chemical production). These observations have implications to design strategies for the redirection of carbon from CH 4 to other high-value products with metabolic engineering, although further investigation is needed (e.g., with biochemical investigations of the enzyme systems). The ability to combine the GEM with the genetic system will be a considerable step toward chemical production with thermophilic methanogens because the use of validated GEMs will enable phenotypes to be predicted. In summary, our combined methods lay the foundation for archaeal biotechnology to optimize existing power-to-gas processes or enable the production of value-added chemicals by metabolic engineering of microbes of the genus Methanothermobacter . Limitations of the study Based on our findings, M. thermautotrophicus ΔH appears the most suitable among the three investigated strains for industrially relevant power-to-gas systems due to its superior kinetics and CH 4 specificity. However, it is essential to consider the fermentation conditions of our study. We applied a volume gas per volume bioreactor per minute (vvm) of 0.08, while others used a much higher vvm of up to 2.01. 13 It must be further noted that the composition of the growth medium was not optimized for any of the three strains, to avoid a bias toward one of the strains. Whether the superior behavior of M. thermautotrophicus ΔH would hold at the higher vvm and with strain-optimized specific media was not evaluated here and should be considered. Our modeling approach is based on assumptions of steady state and considers static fluxes in the metabolism. Many additional factors need to be considered, such as enzyme kinetics and reaction capacities, which may result in bottlenecks in certain pathways. Those considerations were not included in the modeling approach that we applied here, in part due to limited data on enzyme kinetics for the Methanothermobacter strains. However, our manually curated genome-scale reconstruction is the prerequisite for follow-up modeling approaches including but not limited to: (1) integrating transcriptomics and proteomics with alternative algorithms and methods, such as iMAT 55 , 56 or GIMMEp; 57 (2) model constraining with additional types of data, such as enzyme kinetic data (e.g., kinetome 58 ) and 13 C fluxomics; 59 and (3) generating superimposed models, such as metabolic and expression models (ME-models). 60 Our multi-disciplinary systematic approach led us to hypothesize that the three investigated Methanothermobacter strains produce formate for biomass growth in distinct ways. By including the genetically engineered M. thermautotrophicus ΔH pMVS1111A:P hmtB - fdh Z-245 strain, we further support this hypothesis. However, future experiments will need to confirm our observations and hypotheses, for example, by measuring the involved intracellular metabolites or biochemically assaying the enzymatic reactions."
} | 3,723 |
30323959 | PMC6186167 | pmc | 424 | {
"abstract": "The widely observed positive relationship between plant diversity and ecosystem functioning is thought to be substantially driven by complementary resource use of plant species. Recent work suggests that biotic interactions among plants and between plants and soil organisms drive key aspects of resource use complementarity. Here, we provide a conceptual framework for integrating positive biotic interactions across guilds of organisms, more specifically between plants and mycorrhizal types, to explain resource use complementarity in plants and its consequences for plant competition. Our overarching hypothesis is that ecosystem functioning increases when more plant species associate with functionally dissimilar mycorrhizal fungi because differing mycorrhizal types will increase coverage of habitat space for and reduce competition among plants. We introduce a recently established field experiment (MyDiv) that uses different pools of tree species that associate with either arbuscular or ectomycorrhizal fungi to create orthogonal experimental gradients in tree species richness and mycorrhizal associations and present initial results. Finally, we discuss options for future mechanistic studies on resource use complementarity within MyDiv. We show how mycorrhizal types and biotic interactions in MyDiv can be used in the future to test novel questions regarding the mechanisms underlying biodiversity–ecosystem function relationships.",
"conclusion": "Conclusions and Outlook During the first two years of the MyDiv experiment, tree community productivity increased marginally significantly with tree species richness. However, the most productive tree communities were not, as hypothesized, the ones with both mycorrhizal types, but rather those that associate with AMF only. This result may be due to the fast growth of most of the AMF species. The strongest influence of tree diversity on tree productivity was observed in the EMF-communities and slightly weaker effects in mixed communities. The observed increases in productivity with tree diversity were due to the inclusion of highly productive species, primarily Betula pendula Roth (EMF), Tilia platyphyllos Scop. (EMF), and Prunus avium (L.) L. (AMF). Consequently, we found that selection effects drove early biodiversity effects in the EMF-communities. This is consistent with experiments in grasslands (e.g., Marquard et al. 2009 ) and forests (e.g., Tobner et al. 2014 ) showing that selection effects are more important than complementarity effects in early stages of experiments. Complementarity effects may become more important with time ( Fargione et al. 2007 , Reich et al. 2012 ). We speculate that these strong selection effects in EMF-communities may be explained by differences in life histories of the EMF-species (e.g., growth rates; Brzeziecki and Kienast 1994 ). Thus, the presented results have to be interpreted with caution, as the magnitude and mechanisms driving biodiversity effects on ecosystem functioning in MyDiv will likely change over time ( Guerrero-Ramirez et al. 2017 ) as mycorrhiza colonize and continue to develop their hyphal networks. The long-term perspective of the experiment may shed light on processes underlying those temporal dynamics of biotic interactions. MyDiv addresses the need to integrate biotic interactions, realized in the form of mycorrhizal types, into the experimental design of BEF experiments. MyDiv is one of the first experiments that focuses on the effects of identity and diversity of mycorrhizal types that typically co-occur in forest ecosystems and mediate resource use complementarity. To better understand the processes behind resource uptake strategies in the two mycorrhizal types and to test the predicted conceptual framework presented in Fig. 1 , the future use of resource tracer experiments may be particularly promising ( Gockele et al. 2014 ). MyDiv offers the opportunity for further subplot treatments to explore the basis of resource use complementarity. For instance, addition of different nutrients (e.g., combinations with nitrogen and phosphorus additions) and stable isotope tracers could further illuminate the role of specific groups of mycorrhizal fungi in nutrient uptake. Furthermore, the long-term perspective of this tree diversity experiment allows for studying temporal dynamics in the contribution of mycorrhizal type identity and diversity to resource use among plants. In addition, it allows scaling up of species interactions and physiological processes from individuals to neighborhoods to plot-level ecosystem functions. For instance, the use of PhytOakmeters helps in conducting such measurements in a highly standardized way, but also all other tree species are well replicated along the diversity gradient. The PhytOakmeters combine the advantages of a laboratory study system, which to date is commonly used in studies with mycorrhiza and controlled multitrophic interactions ( Herrmann et al. 2016 ), with the necessity to study biotic interactions with plants in more natural ecosystems, such as tree plantations where they interact with their biotic and abiotic environment. In future studies, a second clone may be introduced, for instance, a tree species associating with AMF. Thus, MyDiv implemented functional characteristics of plants in an experimental design to provide a conceptual framework for predicting resource use complementarity by considering biotic interactions with mycorrhizal fungi.",
"introduction": "Introduction Concern that unprecedented rates of biodiversity change will alter ecosystem functioning and the provisioning of ecosystem services has prompted more than two decades of biodiversity–ecosystem function (BEF) research ( Schulze and Mooney 1993 , Loreau et al. 2001 , Cardinale et al. 2012 ). This field of research has provided compelling evidence for a largely positive relationship between biodiversity and ecosystem functioning ( Balvanera et al. 2006 , Cardinale et al. 2012 ) in controlled experiments as well as in nature ( Hautier et al. 2014 , Duffy et al. 2017 ). Despite this emerging consensus regarding the significant role of biodiversity for ecosystem functioning, the underlying mechanisms driving this relationship are still not well understood. Theory predicts that positive plant diversity effects on ecosystem functioning should arise if intraspecific competition in communities is higher than interspecific competition ( Loreau and Hector 2001 ). As a consequence, plant traits related to resource use may be particularly influential drivers of competitive interactions in plant communities. If species are dissimilar in their resource use strategies, they avoid competition for limiting resources (hereafter resource use complementarity). This reduction in interspecific competition should provide higher levels of ecosystem functioning than a community of species with more similar resource use strategies ( Heemsbergen et al. 2004 , Jousset et al. 2011 ). For example, species asynchrony (which may indicate resource use complementarity over time; de Mazancourt et al. 2013 , Hautier et al. 2014 ) and spatial dissimilarity in light use in tree crowns (resource use complementarity in space; Williams et al. 2017 ) have been suggested as significant biological mechanisms that underlie positive BEF relationships. Consequently, much effort has been placed to identify species traits that are essential drivers of BEF relationships ( Ebeling et al. 2014 , Tobner et al. 2014 ). Resource use complementarity among species also relies on biotic interactions across guilds of organisms ( Eisenhauer 2012 , Hines et al. 2015 ). In plant communities, the acquisition of soil nutrients is not only a function of rooting depth ( Mueller et al. 2013 , Oram et al. 2017 ) and root traits, but requires interaction partners like mycorrhizal fungi ( de Kroon et al. 2012 ). Plant mycorrhization occurs in most of the terrestrial plant species and is commonly known to be beneficial to plants by enhancing their growth (e.g., Smith and Read 2010 , van der Heijden et al. 2015 ). Mycorrhizal fungi supply plants with water and nutrients in exchange for photosynthates and therefore co-determine the outcome of plant competition ( Fitter 1977 , Zobel and Moora 1995 , van der Heijden et al. 1998 , Scheublin et al. 2007 , Wagg et al. 2011 a , b , Merrild et al. 2013 ). In fact, mycorrhizal taxa themselves have evolved ways to reduce competition in space and time and possess traits as various as the plants with which they associate ( Koide 2000 , Smith et al. 2000 , Jansa et al. 2005 , van der Heijden and Scheublin 2007 , Thonar et al. 2010 ). As a consequence, mycorrhizal fungi are thought to play a critical role in the maintenance of plant diversity ( Francis and Read 1994 , 1995 , van der Heijden et al. 1998 ) and positive BEF relationships ( Klironomos et al. 2000 , Schnitzer et al. 2011 , Eisenhauer 2012 ). However, mycorrhizal associations are not beneficial in all cases. They form a continuum from being beneficial to being detrimental that depends on factors like environmental conditions and the developmental state of the associations ( Johnson et al. 1997 ). For mycorrhizal associations to maintain plant diversity and improve ecosystem function, the presence and diversity of fungal associations should increase resource partitioning among the different plant species with which they associate ( Klironomos et al. 2000 , Bever et al. 2010 , Wagg et al. 2015 ). This may be true for different plant species associating with different mycorrhizal fungal species and also with different mycorrhizal types. Mycorrhizal types considerably differ in their morphological and physiological traits that facilitate dissimilar soil nutrient uptake processes. Several studies have shown the significance of arbuscular mycorrhizal fungal (AMF) species diversity for plant performance ( Vogelsang et al. 2006 , Maherali and Klironomos 2007 , Wagg et al. 2011 a , b , 2015 , Reinhart and Anacker 2014 ). However, including both mycorrhizal types as potential biotic interactions driving resource use complementarity in studies is crucial as they typically co-occur in natural ecosystems. Our paper provides a conceptual framework for including positive biotic interactions across guilds of organisms—more specifically between plants and mycorrhizal types—to study potential mechanisms behind resource use complementarity of plants as well as the consequences for plant competition and BEF relationships. First, we provide an overview of the current understanding of the effects of biotic interactions on resource use complementarity and how this might enhance ecosystem function. Second, we highlight the urgent need for including plant–mycorrhiza interactions in studies to deepen the mechanistic understanding of resource use complementarity. Third, we introduce a recently established field experiment that utilizes this conceptual framework. The study uses different pools of tree species that associate with dissimilar types of mycorrhizal fungi to create experimental gradients in tree species richness. The different experimental combinations between tree species and mycorrhizal fungi span a hypothesized gradient in coverage of resource niche space and thus provide a predictive framework for resource use complementarity. Fourth, we provide an outlook of potential future studies in this experimental setup that may advance the mechanistic understanding of BEF relationships."
} | 2,880 |
31723597 | PMC6839939 | pmc | 426 | {
"abstract": "Skin-inspired semiconductor that is intrinsically stretchable, self-healable, and strain-sensitive for advanced sensory devices.",
"conclusion": "CONCLUSION We present here an approach to enable strain-sensitive, stretchable, and self-healable semiconductor film for fabrication of skin-like active-matrix strain sensor array. We observed that supramolecular dynamic cross-linked network of semiconducting polymer and a self-healing elastomer provides strain sensitivity to our blended film. In addition, the dynamical cross-linked network by metal-ligand coordination enabled the semiconductor to be not only highly stretchable (>1500%) but also autonomously self-healable at room temperature. The measured GF, an indicator for strain sensitivity of the film, was 5.75 × 10 5 with up to 100% strain, which is the highest reported value for strain-sensitive semiconducting materials. With these properties, a stretchable active-matrix strain sensor transistor array was designed and fabricated with our semiconducting film. Highly stretchable interconnects were developed to enable reliable data acquisition from the active-matrix sensor. The stretchable active-matrix sensor–based e-skin is able to detect pressure-induced deformation of the e-skin, with simultaneous visualization of the applied strain. Last, for skin-like sensory devices, which are fully self-healable and can be operated within range of medically safe voltage, the integration with self-healable conductor and especially high- k dielectric material still need to be developed.",
"introduction": "INTRODUCTION Recent progress in stretchable electronic materials ( 1 , 2 ) and devices ( 3 , 4 ) that emulate the sensing and self-healing properties of human skin has accelerated the development of skin-inspired devices, soft robots, and biomedical devices ( 5 – 14 ). Various rigid sensing modules have been integrated into an ultrathin platform using strain-engineered designs for interconnects and fabrication by transfer printing ( 15 , 16 ). Bioinspired structures/materials are created to further improve sensitivity and compatibility with the human body ( 2 , 17 , 18 ). A modulation of mechanical stimuli to electrical signals has been a representative function of the electronic skin (e-skin), which mimics the human skin sensory function ( 3 , 19 ). The active-matrix transistor array–based sensors provide high-quality sensing signals with reduced cross-talk between the individual pixels ( 20 – 26 ). In this case, each pixel consists of a sensor connected with a transistor. Previous works used strain engineering to integrate rigid sensors and transistors into stretchable biomimetic systems ( 4 , 7 ). To overcome mechanical mismatch between rigid and soft components, both sensors and transistors need to be intrinsically stretchable. Stretchable semiconductor that shows strain-dependent electrical behavior is a potential candidate that combines sensing and transistor switching. Such a strain-sensing transistor can potentially simplify fabrication processes and improve mechanical robustness and conformability. In addition, a self-healing ability would be an added benefit to e-skin to warrant a longer lifetime of the e-skin ( 27 – 38 ). Here, we present an intrinsically stretchable and self-healable semiconducting film that has strain-sensitive electrical behavior when incorporated into a stretchable transistor. The semiconducting film was prepared by a blend of polymer semiconductor and an insulating elastomer. Both materials, which contain metal ligand dynamic bonding sites, were recently reported by our group for stretchable and self-healable electronic materials ( 6 , 30 ). Here, we intend to demonstrate a new property by fusing the two materials together for strain-sensitive semiconducting film through dynamic cross-linking of polymer semiconductor and insulating elastomer. The metal coordination bond, once broken, can spontaneously reconstruct, rendering the brittle semiconducting film stretchable, tough, and self-healing. Moreover, the elastomer in the blended film is highly elastic with low modulus, effectively absorbing the external mechanical strain conferred. Thus, this approach can be a new way for developing multifunctional electronic materials. The strain sensitivity that is defined as its field-effect mobility modulation of the semiconducting film to external stimuli can be optimized by controlling the weight ratio of the semiconducting polymer and insulating elastomer. The nanoparticle-like phase separation of the semiconducting film enabled it to be further strain sensitive, since the local distance between zero-dimensional (0D) structures is highly correlated to the effective charge transportation during stretching and releasing ( 39 – 42 ). The optimized semiconducting film showed high strain sensitivity [gauge factor (GF), 5.75 × 10 5 at 100% strain] and intrinsic stretchability (fracture strain, >1300% strain; fig. S2). The GF was calculated as a general equation for piezo-resistive semiconductor and conductor, GF = (Δ R / R )/ε, and the method is further explained in the Materials and Methods section. In addition, the broken dynamic bonding can be spontaneously reformed, enabling the recovery of the damaged film. The cut semiconducting film was observed to be autonomously self-healed after 1 day at room temperature, and its field-effect mobility was almost completely recovered. Next, we fabricated a crack-based stretchable gold (Au) nanomembrane interconnector that is highly conductive and durable up to 100% strain according to literature report ( 43 ). Last, we proceeded to fabricate a stretchable active-matrix sensory transistor array. The semiconducting film, dielectric, electrode, and interconnect are all effectively integrated into an active-matrix array platform using a transfer-printing process. Our stretchable active-matrix skin-like sensor array is successfully capable of monitoring strain distribution of the external force. In addition, a transfer-printed passivation layer enabled the semiconductor/dielectric interface of the sensor array to be waterproof even after we drop-casted artificial sweat on it for 15 hours. Our demonstrated strain-sensitive, stretchable, and self-healable semiconducting film would change the paradigm of the e-skin to further expand its application.",
"discussion": "RESULTS AND DISCUSSION Figure 1 shows the overall material design strategy for the strain-sensitive, stretchable, and self-healable semiconducting film and its mechanical and electrical properties. We choose poly(3,6-di(thiophen-2-yl)diketopyrrolo[3,4-c]pyrrole- 1,4-dione-alt-1,2-dithienylethene) with 10 mol% 2,6-pyridinedicarboxamine moieties (DPP-TVT-PDCA) as the semiconducting material due to its good charge carrier mobility, as we reported previously ( 6 ), and the PDCA units that can be used to bind to the insulating and stretchable poly(dimethylsiloxane-alt-2,6-pyridinedicarbozamine) (PDMS-PDCA) polymer ( Fig. 1A ) ( 30 ). We previously reported the use of PDCA to form metal-ligand coordination complex with a molar ratio of Fe(III) ion to PDCA ligand of 1:2 ( 30 ). It was shown that Fe(III)-PDCA coordination has multiple dynamic bonds with three different bonding strengths (Fe-N pyridyl , strong; Fe-N amido , medium; and Fe-O amido , weak) facilitating the dynamic cross-linking for intrinsic stretchability and self-healing ability ( 30 ). The metal-ligand coordination bond was formed first during the preparation of PDMS-PDCA-Fe elastomer [PDCA:Fe(III) = 2:1] to prevent chemical damage of the semiconducting polymer from a strong acidic by-product (hydrochloric acid). The semiconducting film was prepared by blending semiconducting DPP-TVT-PDCA and PDMS-PDCA-Fe elastomer. The cross-linked PDMS-PDCA-Fe chains in organic solvent can exchange metal-ligand bonding with PDCA segment in DPP-TVT chains. Figure 1B shows a schematic illustration of Fe(III)-PDCA ligand bonding of PDMS-PDCA and DPP-TVT-PDCA in the blend film. Fig. 1 Design and characterizations of strain-sensitive, stretchable, and self-healable semiconducting film. ( A ) Chemical structure of DPP semiconducting polymer, PDMS, and PDCA moiety introduced in both polymer backbones as dynamic bonding sites through metal-ligand interaction. Structure of the [Fe(HPDCA) 2 ] + moiety that is reversible dynamic bonds by force. ( B ) Schematic illustration of DPP and PDMS dynamically cross-linked through Fe(III)-PDCA complexation. ( C ) STEM dark-field and STEM-EDS elemental mapping of the DPP-TVT-PDCA (1):PDMS-PDCA-Fe (5) blend film. ( D ) Field-effect mobilities of the blend film organic thin-film transistors (OTFTs) (source and drain electrode: Au, 40 mn; dielectric layer: SiO 2 , 300 nm; gate electrode: highly doped silicon substrate) as a function of blending-weight ratio (semiconductor:elastomer). ( E ) Strain cyclic testing of the blend film (1:5). ( F ) Plot of dichroic ratio (α ⫽ /α ⊥ ) of 1:5 blend film as a function of strain. ( G ) Relative degree of crystallinity (rDoC) calculated from (200) peak for both “parallel” and “perpendicular” directions to x-ray beam line. ( H ) Proposed mechanism for reinforcement of stretchability in blend film via metal-ligand dynamic bonding based on analyzed information. We evaluated the dependency of field-effect mobility on semiconductor (DPP-TVT-PDCA), with an elastomer (PDMS-PDCA-Fe) weight ratio ranging from 1:1 to 1:20, as shown in Fig. 1D and fig. S1. The optimized weight ratio of the blend film was observed to be 1:5, which is the minimum weight ratio of the semiconducting polymer to give reasonable charge carrier mobilities, which suggest that sufficient electrical percolation paths (highlighted in blue color in Fig. 1D ) are still preserved in the blend film. The observed volume faction was 0.166, as determined by sulfur element, indicating semiconducting polymer in Fig. 1C using ImageJ software. This value is very close to the theoretical percolation threshold volume fraction of 0.16 for spherical particles ( 44 ). The blend film was observed to have a high stretchability (fracture strain, >1300%; fig. S2) with a Poisson’s ratio of 0.462 (fig. S3), and its Young’s modulus (~300 kPa) is similar to that of human skin, while typical semiconducting polymers are in the range of hundreds of megapascals to gigapascals ( 45 ). Next, the rheology analysis of the blend film at room temperature showed that the storage modulus ( G ′) is higher than the loss modulus ( G ″) in a typical frequency range from 10 −3 to 10 3 Hz and a temperature range from 10° to 60°C, respectively, which indicates that the blend film behaves like a solid material due to the metal-ion coordination cross-linking with the more rigid DPP-TVT (figs. S4 and S5). In addition, the low glass transition temperature of the blend film (below −90°C; fig. S6) is similar to that of typical PDMS rubber ( 46 ). To further analyze the stretchability of the blend film, we conducted a repeated strain cyclic test, as shown in Fig. 1E . We observed that the blend film began to display stress-strain hysteresis when it was stretched to above 30% strain. We attribute this hysteresis to energy dissipation, resulting from the breakage of Fe(III)-PDCA coordination bonds, leading to stress relaxation (fig. S7). However, even when the blend film was elongated by stretching it to over 100% strain, we observed that it was able to revert to its initial length after 1 hour of rest (fig. S8). This recovery is attributed to the reorganization of the polymer chains to, approximately, their initial configurations driven by the energetic gain of the configurational entropy of this initial state. To characterize the electrical percolation path in the blend film, the morphology of the blend film was characterized by transmission electron microscopy (TEM). The morphology of the blend film depended on the blend ratio from 1:1 to 1:20 (semiconductor:elastomer; fig. S9). We observed nanoparticles uniformly distributed in the 1:5 blend film on the order of 100 nm. These nanoparticles were formed during thermal annealing through phase separation driven by difference of surface energy of the two different materials. Element mapping of the scanning TEM (STEM) image was subsequently performed by energy-dispersive x-ray spectroscopy (EDS) to identify elements in the film ( Fig. 1C ). Sulfur (S) and silicon (Si) peaks are used to determine the presence of semiconducting and insulating polymers, respectively (see chemical structures in Fig. 1A ). In the nanoparticle regions, strong S and Fe signals can be seen, while other regions showed strong Si signals. These results indicate that the nanoparticles are primarily composed of the semiconducting polymers with Fe-PDCA, potentially at the interface between the semiconducting domains and the PDMS regions. Although the connection between the nanoparticle semiconducting domains cannot be seen clearly, the charge carrier mobility still maintained a value as high as 0.1 cm 2 /Vs. This suggests that the nanoparticles are connected by a small amount of DPP-TVT. This also suggests that the blend film maybe very sensitive to strain. To characterize the molecular level changes of the semiconducting polymer during stretching, the chain alignment was measured using polarized ultraviolet-visible (UV-vis) spectroscopy (fig. S10), and the degree of chain alignment was quantified in terms of its dichroic ratio (α ⫽ /α ⊥ ), as shown in Fig. 1F . The dichroic ratio of the blend film was observed to linearly increase up to 50% strain (presumably due to strain-induced chain alignment) and plateaued at 100% strain, while the neat semiconducting polymer’s dichroic ratio linearly increased up to 100% with a higher slope. The observed difference in response to strain is may be because the elastomer is easier to stretch by strain due to its lower modulus than the semiconducting polymer in the blend film. The relative degree of crystallinity of the blend film was measured using grazing-incidence wide-angle x-ray diffraction to understand the change of film morphology upon stretching cycles ( Fig. 1G and figs. S11 and S12). The initial crystallinity of blend film was maintained at ~80%, although it was stretched up to 100% strain and fully recovered regardless of stretching direction with preserving full-width at half maximum value (parallel and perpendicular stretching directions to x-ray beam). This observation, together with the dichroic results, indicates that the applied strain is mainly absorbed into the elastomer while preserving the crystalline regions of the semiconducting polymer. Together, this constitutes the proposed stretching mechanism of the blend film ( Fig. 1H ). To evaluate the strain-sensitive charge transport of the semiconducting film, organic thin-film transistors (OTFTs) were fabricated using transfer printing of the semiconducting film (200 nm), as shown in Fig. 2A . The semiconducting film was stretched from 0 to 100% strain on PDMS elastomer stamp and transferred onto the surface of SiO 2 (dielectric) on a heavily doped Si (gate) substrate. Atomic force microscopy (AFM) images did not find any nanocracks in the transferred film, indicating no mechanical damage due to strain ( Fig. 2B ). Gold, as an electrode material, was then thermally evaporated on the blend film. The OTFT showed typical transistor output and transfer curves (fig. S13). They were observed to be highly sensitive on-current as a function of the applied strain. Specifically, the on-current of the transistor decreased from 2.79 × 10 −5 A at 0% strain to 4.85 × 10 −10 A at 100% strain ( Fig. 2C ). The GF was 5.75 × 10 5 at 100% strain ( Fig. 2D ), which is among the highest values previously reported for semiconducting strain gauges and even comparable with the state-of-art conductor-based strain gauges. As a comparison, the highest value previously reported were for graphene-polymer nanocomposite-based strain sensors and mechanical crack-based strain sensors, which showed GFs of more than 500 and 2000 (0 to 2% strain), respectively ( 47 – 50 ). The morphology of the blend film, as observed by optical microscopic images and AFM, during stretching did not show any visible cracks up to 100% strain (fig. S14). The devices showed fully reversible current-voltage characteristics ( Fig. 2E ) and repeatable cycling behavior within a strain range of 30% ( Fig. 2F and fig S15), a value similar to the stretchability of human skin. Fig. 2 Strain-sensitive property of self-healable semiconducting film. ( A ) Schematic illustration for sequential fabrication procedures of the OTFT with stretchable self-healable semiconducting film (200 nm) using transfer-printing assembly. ( B ) AFM height images for pristine and stretched (100%) semiconducting films. Scale bars, 1 μm. ( C ) Transfer curves of OTFTs as a function of strain applied to semiconducting film along the tensile stretching direction and ( D ) GFs extracted from on-current of OTFTs. ( E ) Field-effect mobilities on strain and after releasing strain measured for the same device. ( F ) Field-effect motility as a function of stretching cycle at different strains. ( G ) Schematics for fabrication methods of the self-healed semiconducting film that was cut by bending a partially cracked PDMS stamp and its OTFT. ( H ) Optical microscope (OM) images of damaged semiconducting film through self-healing process and ( I ) self-healed film. Inset: Corresponding dark-field OM images. ( J ) Transfer curves and ( K ) field-effect mobility of pristine and autonomously healed OTFTs. R.T., room temperature. Self-healing is a unique function for next-generation e-skin. Our semiconducting film can be self-healed through the dynamic metal-ligand coordination bonding. To evaluate the film’s self-healing ability, the blend film (200 nm in thickness) was cut and left at room temperature. After 24 hours, we observed the scar of the cut film autonomously disappeared, i.e., self-healed. Furthermore, the healed film could be stretched again to more than 200% strain before fracturing (fig. S16). Since the thickness of the semiconducting layer in a thin film transistor is only on the order of tens of nanometers, it needed to be supported by a substrate for cutting and self-healing tests. Figure 2G shows a process to perform these tests. Briefly, the self-healable semiconducting film was transferred to a precut PDMS stamp from an octadecyl trichlorosilane (OTS)–treated SiO 2 /Si substrate and then was broken by bending the semiconducting film/PDMS stamp quickly. The two broken pieces of semiconducting films were thus self-aligned to contact with each other when the PDMS substrate was returned to a flat structure ( Fig. 2G ). Figure 2H depicts an optical microscope image of the semiconducting film (200 nm) after breaking on the stamp. For autonomous self-healing, the damaged film was left without any post-treatment at room temperature. The scar in the damaged film was observed to be completely gone after self-healed for 1 day ( Fig. 2I ). Next, the electrical property of the healed semiconducting film was evaluated in a bottom-contact bottom-gate–structured OTFT shown in Fig. 2G using a soft-contact method. Figure 2J shows the transfer curves before and after self-healing, in which the field-effect mobility of the healed semiconducting film was recovered from 0.047 ± 0.013 to 0.028 ± 0.047 cm 2 /Vs ( Fig. 2K ). In comparison, the device with cut semiconducting film without allowing to self-heal did not show any transistor-like current-voltage behavior. To demonstrate the potential of our newly developed semiconducting film for e-skin applications, we proceeded to fabricate a 5 by 5 fully stretchable strain-sensitive active-matrix transistor array ( Fig. 3 ). Typically, resistive sensors require an active-matrix backplane for multiplexing without cross-talk between pixels. The use of the above strain-sensitive transistors combines a strain-resistive sensor and a transistor into one single device and reduces the complexity in device fabrication and potentially gives higher sensitivity. To achieve high-speed scanning of multiple lines without signal delay or loss in the active-matrix architecture, a highly stretchable and conducting interconnect is required. Although many stretchable electrodes have been reported using nanowires or nanotubes, or nano/microparticles, these approaches generally involve a trade-off between electrical conductivity and mechanical stretchability. Recently, a wrinkled and cracked metal nanomembrane supported on elastomer substrates was reported for stretchable electrodes ( 4 , 43 , 51 , 52 ). Building on this concept, we further developed highly conductive stretchable electrode using Au and polystyrene-block-poly(ethylene-ran-butylene)-block-polystyrene (SEBS) elastomer for the interconnect of active-matrix transistor array. Figure 3A shows a plot of resistance of the Au (thickness of 80 nm) nanomembrane/SEBS free-standing electrode as a function of time with different tensile strains (black, 50%; gray, 70%; teal, 100%) while stretching the electrode for a total of 10 cycles. Inset indicates photographs of the electrode before and after stretching up to 100% strain. Compared to previous electrodes, absolute resistance values are relatively low and stable even at the tensile strain of 100%. To confirm the mechanical reliability of our electrode, we performed repeated cyclic testing up to 100 cycles under 50% strain ( Fig. 3B ). Superior performance was obtained from the reversible wrinkled and cracked nanostructures supported on a SEBS (elastic modulus, 3.5 MPa) with high elasticity Fig. 3C ). The good mechanical and electrical stability of the stretchable interconnect is important to allow good contact to our semiconducting layer via a conventional evaporation process and transfer-printing to give a 5 × 5 strain-sensitive transistor array ( Fig. 3, D and E ). The active area of each pixel has a channel width of 1 mm and channel length of 150 μm. We observed that all devices showed good uniformity with a maximum mobility of 0.11 cm 2 /V s and an average mobility of 0.076 ± 0.019 cm 2 /V s ( Fig. 3, F and G , and fig. S17). To verify its reversible strain-sensing operations, we stretched the device up to 30% strain and then released it while applying a gate voltage of −60 V at a read voltage of −60 V ( Fig. 3H ). As expected, the fully stretchable transistor is sensitive to the applied strain, while the saturation drain current was linearly reduced when applied strain was increased. It was also able to fully recover to the original level after the strain was released. Normalized on-current of the device depending on strain is presented in Fig. 3I . Fig. 3 Characterizations of stretchable active-matrix transistor sensor array. ( A ) In situ measurement of resistance of Au/SEBS stretchable interconnect during 10 stretching cycles at different strains (50, 70, and 100%). Inset: Photographs of Au/SEBS interconnect at 0% (left) and 100% (right) strain. ( B ) Resistance change of Au/SEBS stretchable interconnect as a function of stretching cycle at 0 and 50% strain. ( C ) OM images of pristine (0% strain, upper left), stretched (100% strain, upper right), released (0% strain, lower right), and stretched (100% strain; 100 cycles, lower left) Au/SEBS stretchable interconnect. ( D ) Architecture and ( E ) photograph of a fully stretchable 5 × 5 active-matrix transistor strain sensor array fabricated via our developed strain-sensitive, stretchable, and self-healable semiconducting film. Scale bar, 5 mm. ( F ) Mapping and ( G ) statistical distribution of the field-effect mobility in our stretchable active-matrix transistor array. ( H ) Transfer curves and ( I ) normalized on-current of fully stretchable transistor in active-matrix array as a function of strain. Photo credits: Jin Young Oh, Department of Chemical Engineering, Kyung Hee University and Donghee Son, Biomedical Research Institute, Korea Institute of Science and Technology. SQRT, square root. For practical application of the stretchable strain sensor array as e-skin, a lower device-operating voltage is desirable for long-term sustainability and safety. Hence, the drain voltage of device was reduced from −60 to −5 V. Despite the thick dielectric layer, due to the low-threshold voltage, the device still showed ideal transfer curves in accordance to applied drain voltages and was sensitive on applied strain under the reduced drain voltage (−5 V) ( Fig. 4A and fig. S18). Next, waterproof performance is highly desirable since the ions from sweat as generated from human skin may result in malfunction of the device. Thus, our fabricated 5 × 5 sensory transistor array was passivated using SEBS elastomer to protect it against sweat ( Fig. 4B ). The resulting encapsulated device was observed to maintain its electrical performance from undesired leakage sources for a 15-hour operating duration when submersed in artificial sweat ( Fig. 4C ). The distinctive property of our fabricated active-matrix skin-like sensor array is that this monolithic sensing system enables 3D mapping of e-skin surface deformation with a simplified fabrication process as it combines one sensor and one transistor architecture into one transistor device. To demonstrate its functionality as the stretchable strain sensor, we used a plastic tip to poke on the sensor array ( Fig. 4D ) while simultaneously recording the on-currents from multiple pixels of the array ( Fig. 4E and table S1). The obtained 3D hemispheric shape of the normalized on-current mapping corresponded to that of the “poked” e-skin. To quantify the on-current changes of the active-matrix sensor array upon poking it, we simulated the applied strain using finite-element method. Figure 4F shows a mapping of calculated maximum principal strain on our active-matrix sensor array, which is able to subsequently calibrate the current changes in our device through the calculated principal strains, as shown in fig S19. Fig. 4 Strain-sensitive stretchable active-matrix transistor array as skin-like stretchable strain sensor. ( A ) Transfer curves of the stretchable active-matrix transistor array as a function of drain voltage with four different drain/source voltages. ( B ) Photograph of the stretchable active-matrix transistor array under artificial sweat and ( C ) on- and off-currents of the stretchable active-matrix transistor array as a function of time. ( D ) Photograph of stretched active-matrix transistor array by poking with a plastic bar and ( E ) normalized on-current of the poked active-matrix transitory array. ( F ) Simulation result of strain applied by poking to the stretchable active-matrix array. Photo credits: Jin Young Oh, Department of Chemical Engineering, Kyung Hee University."
} | 6,724 |
33888697 | PMC8062479 | pmc | 428 | {
"abstract": "Environmental composition is a major, though poorly understood, determinant of microbiome dynamics. Here we ask whether general principles govern how microbial community growth yield and diversity scale with an increasing number of environmental molecules. By assembling hundreds of synthetic consortia in vitro, we find that growth yield can remain constant or increase in a non-additive manner with environmental complexity. Conversely, taxonomic diversity is often much lower than expected. To better understand these deviations, we formulate metrics for epistatic interactions between environments and use them to compare our results to communities simulated with experimentally-parametrized consumer resource models. We find that key metabolic and ecological factors, including species similarity, degree of specialization, and metabolic interactions, modulate the observed non-additivity and govern the response of communities to combinations of resource pools. Our results demonstrate that environmental complexity alone is not sufficient for maintaining community diversity, and provide practical guidance for designing and controlling microbial ecosystems.",
"introduction": "Introduction Microbial communities form the basis for an enormous range of biological processes, from cycling of nutrients in the ocean to regulation of human health 1 – 5 . Despite our growing knowledge of community compositions in various biomes 6 – 8 and of the role of individual nutrients in modulating community properties 9 , 10 , relatively little is known about how the nutrient complexity of an environment (i.e., the number of different available nutrients) affects community ecology. Understanding this relationship is crucial to disentangling the effects of the environment on natural microbial ecosystems, which are exposed to a multitude of different nutrients 9 , 10 , as well as the effects of diet on microbiome structure and function. In the gut microbiome, for example, recent work has highlighted how community composition depends strongly on the diversity of available nutrients 11 – 13 . However, reports in natural ecosystems 14 – 16 and in synthetic microcosms 17 – 19 conflict on how this environmental complexity modulates growth yields and taxonomic diversity, raising questions as to how these relationships vary within and across communities, and hindering nascent efforts to engineer microbiomes with defined functions 20 – 22 . An additional unknown in the effect of environmental complexity on community assembly is the extent to which different nutrient compositions drive ecosystems towards predictable states, as opposed to stochastically driven outcomes. While previous work has shown a combination of determinism and stochasticity in community assembly 23 , 24 , studies have also shown how particular environments can be associated with long-lasting stable communities 25 . It is therefore important to understand to what degree these patterns will impact synthetic consortia cultured on increasingly complex combinations of defined nutrients. A number of quantitative frameworks have previously been used to address similar questions, and have provided possible clues as to how a microbial community could depend on the complexity of its environment. In classical ecology, for example, theories based on competitive exclusion and niche partitioning suggest that there would be greater opportunities for biodiversity in environments with a greater breadth of nutrient types 26 , 27 . Although this is an intuitive hypothesis, factors such as organism-specific resource use capabilities 28 , 29 , ecological niche overlap 30 , 31 , and interspecies interactions 24 , 32 , 33 can lead to significant deviations from this expectation. From a very different perspective, the question of how different perturbations in biological systems would be expected to jointly affect a given phenotype is captured by the classical genetic concept of epistasis 34 , 35 . Epistasis between two genetic mutations, for example, quantifies how much the phenotypic effect of one mutation is affected by the presence of the other. This concept constitutes a broader systems biology framework for quantifying the nonlinear behavior of biological systems 36 – 39 , and can be used to estimate the non-additivity of microbial community phenotypes 40 – 42 (note that we will use the terms ‘non-additive’ and ‘nonlinear’ interchangeably throughout this paper). Specifically, one may extend this notion to define epistasis between environments, by comparing community properties observed on combinations of nutrient sets against those on individual nutrient sets. Here, we determine how increasingly complex environmental compositions affect the growth yield and taxonomic structure of synthetic microbial communities. In addition to mapping the phenotypes of these communities along the axis of environmental complexity (the number of different carbon sources present in the medium), the design of our experiments allows us to quantify how communities are shaped by the combination of sets of carbon sources relative to their properties under each constituent set. By testing the effects of increasing numbers of up to 32 different carbon sources on over 280 synthetic microcosms, we examine how yield and diversity differ from expectations based on those in simpler environments. We further contextualize our results through the use of mathematical models, which reveal how these environment-phenotype relationships can be explained by a set of ecological rules for combining environments, with implications for the ecology of natural and engineered microbiomes.",
"discussion": "Discussion Deciphering how multispecies microbial communities grow on mixtures of resources remains highly challenging. Here, we showed how a hierarchical experimental design paired with extensive consumer resource modeling can be applied to address this question, revealing the role of resource specialization and niche overlap in determining the scaling of community properties with environmental complexity. Although our simplified experimental system is still far from the complexity of natural microbiomes 48 , 53 , it captures properties that go beyond those observable in small artificial consortia. In particular, we identified a simple additive principle that explains how average growth yields can remain invariant with increasing environmental complexity—a consequence of all available resources being efficiently utilized given enough organisms with varied metabolic capabilities. Although one could expect this behavior to arise in communities well adapted to a specific environment, it is surprising that it also emerged in our synthetic consortia composed of organisms from different biomes grown on artificial combinations of carbon sources. Despite this additive relationship in some communities, our experiments and modeling showed how decreasing the degrees of community niche overlap could lead to non-additively increasing growth yields, reminiscent of observations of overyielding in various ecological studies 54 – 56 . However, while overyielding generally pertains to species mixtures displaying higher yields relative to monoculture, we describe a pattern by which these increases in yield are brought about by increasingly complex environments. To contextualize these observations, we drew from descriptions of nonlinearities in genetics and devised an ‘epistatic’ metric that allowed us to quantify our observed non-additive scaling of growth yield. The versatility of the concept of epistasis allowed us to define similar metrics to quantitatively describe changes in taxonomic diversity. In contrast to our growth yield epistasis distributions that were either centered at zero or positively skewed, our distributions of diversity epistasis were centered on negative values. While the magnitude of negative epistasis was also highly dependent on organism resource specialization and niche overlap, our results raise the question of whether different distributions could be observed given an alternative formulation of our epistatic metrics. Indeed, the question of which mathematical definition best establishes a baseline for biological nonlinearities is a longstanding one in genetics 57 , 58 , raising the prospect of new definitions as the basis for expanded quantitative evaluations of ecological nonlinearities. Irrespective of our formal definitions, however, our results showed how increased environmental complexity does not guarantee greater taxonomic diversity beyond that already possible on individual carbon sources 24 . This result underscores the dependence of biodiversity on an interplay of features, such an appropriate balance of generalists and specialists and the existence of evolved interdependencies 59 – 61 . Furthermore, it raises the prospect for systematic exploration of additional mechanisms that can impact the relationship between environmental complexity and community ecology. Of particular interest are experimental concerns such as the timescale and regime of medium dilutions 62 , or metabolic ones such as the impacts of different molecular currencies (e.g., nitrogen or phosphorus 63 – 65 ) and the ability of organisms to either sequentially or simultaneously utilize multiple resources 66 , 67 . Such extensions would further clarify how communities respond to combinations of resources, facilitating the design of synthetic microbial ecosystems and improving multiscale models of communities adapted to complex environments, such as host-associated microbiomes and communities involved in biogeochemical cycles 68 – 72 ."
} | 2,409 |
37188092 | PMC10175680 | pmc | 429 | {
"abstract": "Flexible electronic devices play a key role in the fields of flexible batteries, electronic skins, and flexible displays, which have attracted more and more attention in the past few years. Among them, the application areas of electronic skin in new energy, artificial intelligence, and other high-tech applications are increasing. Semiconductors are an indispensable part of electronic skin components. The design of semiconductor structure not only needs to maintain good carrier mobility, but also considers extensibility and self-healing capability, which is always a challenging work. Though flexible electronic devices are important for our daily life, the research on this topic is quite rare in the past few years. In this work, the recently published work regarding to stretchable semiconductors as well as self-healing conductors are reviewed. In addition, the current shortcomings, future challenges as well as an outlook of this technology are discussed. The final goal is to outline a theoretical framework for the design of high-performance flexible electronic devices that can at the same time address their commercialization challenges.",
"conclusion": "4 Conclusion In summary, this review focuses on the development and design strategies of stretchable and self-healing conductors in electronic skin, while the development of intrinsically stretchable conductors is relatively blank. Based on the current progress in artificial intelligence technology and the concept of sustainable and environmentally friendly materials, developing biodegradable flexible devices is a good idea. In addition, the demand for flexible electronic devices such as electronic skin in the pharmaceutical, aerospace, and new energy industries will increase day by day, making it a promising field in the future.",
"introduction": "1 Introduction After years of research and development, electronic products based on organic materials have made significant progress in performance, stability, and production costs compared to traditional materials, such as organic field-effect transistors (OFET) ( Vissenberg et al., 1998 ; Meijer et al., 2003 ; Payne et al., 2005 ; Muccini, 2006 ; Wang et al., 2012 ; He et al., 2014 ), organic photovoltaic cells (OPVs), and organic solar cells (OSCs) ( Siebbeles, 2010 ; Gruverman et al., 2011 ; Zhang et al., 2012 ; Bredas, 2014 ; Cnops et al., 2014 ; Tumbleston et al., 2014 ). With the efforts of scientific researchers, the above products have achieved good practical use efficiency and have great commercial prospects. Most organic electronic devices are assembled from several components, which ensure that the device exhibits good functionality and performance. For example, OFET is typically composed of gate electrodes, drain electrodes, organic semiconductor layers, insulating layers, and grid electrode. Although these components provide performance, these materials typically have great rigidity, which resulting in the inflexibility of electronic devices, and limits the development of stretchable and skin like electronic products ( Shcherbina et al., 2017 ; Etiwy et al., 2019 ; Bent et al., 2020 ). With the progress of the organic electronics industry, the research of flexible electronic materials is also advancing. The emergence of electronic skin (flexible electronic devices) represents the phased significance of stretchable, flexible, and dexterous electronic products. Electronic skin, also known as a new wearable flexible bionic tactile sensor, is a new type of electronic material that can simulate human skin and provide biocompatibility. The simple structure of electronic skin can be processed into various shapes, which attracting increasing attention. To simulate the good performance of human skin during actual use, corresponding materials need to have a certain degree of flexibility and self-healing ability ( Kim et al., 2011 ; Jeong et al., 2012 ). So far, some work has been done on stretchable and self-healing organic electronic devices, and significant progress has been made ( Liang and Stephen, 2010 ; Kuribara et al., 2012 ; Hammock et al., 2013 ; Bauer et al., 2014 ; Kim et al., 2014 ; Son et al., 2014 ; Xie and Wei, 2014 ; Kim and Lee, 2015 ; Minev et al., 2015 ; Larson et al., 2016 ). In this review, we summarize the methods used to develop stretchable and self-healing materials and their synthetic devices."
} | 1,094 |
30880142 | null | s2 | 431 | {
"abstract": "From biosynthesis to bioremediation, microbes have been engineered to address a variety of biotechnological applications. A promising direction in these endeavors is harnessing the power of designer microbial consortia that consist of multiple populations with well-defined interactions. Consortia can accomplish tasks that are difficult or potentially impossible to achieve using monocultures. Despite their potential, the rules underlying microbial community maintenance and function (i.e. the task the consortium is engineered to carry out) are not well defined, though rapid progress is being made. This limited understanding is in part due to the greater challenges associated with increased complexity when dealing with multi-population interactions. Here, we review key features and design strategies that emerge from the analysis of both natural and engineered microbial communities. These strategies can provide new insights into natural consortia and expand the toolbox available to engineers working to develop novel synthetic consortia."
} | 262 |
25141306 | PMC4139316 | pmc | 432 | {
"abstract": "Waggle dancing bees provide nestmates with spatial information about high quality resources. Surprisingly, attempts to quantify the benefits of this encoded spatial information have failed to find positive effects on colony foraging success under many ecological circumstances. Experimental designs have often involved measuring the foraging success of colonies that were repeatedly switched between oriented dances versus disoriented dances (i.e. communicating vectors versus not communicating vectors). However, if recruited bees continue to visit profitable food sources for more than one day, this procedure would lead to confounded results because of the long-term effects of successful recruitment events. Using agent-based simulations, we found that spatial information was beneficial in almost all ecological situations. Contrary to common belief, the benefits of recruitment increased with environmental stability because benefits can accumulate over time to outweigh the short-term costs of recruitment. Furthermore, we found that in simulations mimicking previous experiments, the benefits of communication were considerably underestimated (at low food density) or not detected at all (at medium and high densities). Our results suggest that the benefits of waggle dance communication are currently underestimated and that different experimental designs, which account for potential long-term benefits, are needed to measure empirically how spatial information affects colony foraging success.",
"introduction": "Introduction Colony success in social insects often depends on the colony's ability to mobilize workers and allocate them to where work is needed [1] - [4] . Accordingly, insects have evolved different ways to communicate in these situations, such as when nestmates must be recruited to valuable resources [1] , [2] . One of the most remarkable means of recruitment is the honeybee waggle dance. Foragers perform this dance-like behavior inside the nest after finding a profitable food source or on a swarm during nest-hunting to advertise suitable nest-sites [3] , [5] - [8] . Other foragers following a waggle dance learn the location and are subsequently able to fly to the area of the food source [8] – [12] , where they use additional visual and olfactory information to localize the food [6] , [8] , [13] , [14] . Only recently have attempts been made to quantify empirically the colony level benefits of this spatial information [15] – [19] . Surprisingly, these studies found that colonies would often not benefit from spatial communication [16] – [19] . For example, Donaldson-Matasci and Dornhaus [17] tested the effect of spatial communication in five different habitats, but found a positive effect of communication only in one. Dornhaus & Chittka [18] found benefits only in a tropical habitat, but not in temperate European habitats. These findings suggest that the benefits of spatial information strongly depend on the spatiotemporal distribution of food sources, a conclusion that is in agreement with theoretical studies [20] – [22] . Recruitment by waggle dances is costly in terms of time and energy [21] , [23] , [24] and it is thus conceivable that these costs outweigh the benefits under certain conditions, for example when food sources are easy to find by independent search (scouting). However, the apparent absence of benefits resulting from spatial information use in many habitats could also be the result of confounding effects caused by the experimental designs used to quantify these benefits. In previous studies, researchers took advantage of the fact that honeybees are unable to perform oriented dances on a horizontal comb with no or only diffuse light [16] – [19] . It is thus possible to create colonies with oriented (with spatial information; SI) or disoriented (no direction information; NI) dances and compare the foraging success of colonies in these two conditions. Importantly, in these studies, colonies were kept in one condition for relatively short time periods (2 or 3 days in [16] – [18] ; a variable number of days with a mean of ∼12 days in [19] ) and repeatedly switched between the two experimental conditions (SI vs. NI), presumably to use paired statistical tests. If a bee that was recruited on the last day of the SI treatment returns to that food source on the following day, the food she collects on that day will be added to the NI treatment, even though her success was a consequence of acquiring spatial information during the SI treatment. Thus, the continued availability of food sources combined with “site fidelity” (persistency) leads to confounded results and makes the interpretation of such data challenging. This seems particularly important since flower patches can remain in bloom for weeks or even months [25] , and honeybees often return to the same foraging sites for days and up to 3 weeks [6] , [8] , [26] – [34] . Here we explored, using an agent-based simulation model, whether there is a benefit to the use of spatial information over longer periods of time than previously explored. We hypothesized that forager persistency causes long-term benefits of successful recruitment events, and therefore the benefits of spatial information might have been underestimated in experimental designs such as those used in empirical studies. Furthermore, we hypothesized that the degree to which results are confounded depends on various factors, including the duration of the experimental period, the longevity of food patches, the density of food sources, and the size of the colony.",
"discussion": "Discussion Spatial location information of waggle dancing bees increased colony foraging success under almost all simulated circumstances. Spatial information did not improve colony success when patches lasted just 1 day, which is unrealistic for most natural habitats [25] , [54] – [56] . The most surprising finding is that the relative benefits of spatial information increased as environmental stability increased ( Figure 2 ). This contradicts the common assumption that dance information is most beneficial in an environment with ephemeral food sources [18] , [19] , [29] , [57] . A closer examination of our model and the existing literature suggests that our results of long term benefits of the dance are plausible. Recruitment in honeybees is costly [21] , [23] , [24] : recruits need to wait for a dancing bee [21] , and they usually require several field excursions before locating the advertised food source after following a dance [23] , [58] – [60] . Thus, potential recruits incurred both energy and opportunity costs in our model. But after the advertised food source is located, the energetic returns are higher than for scouts. This result is consistent with empirical studies, which found that recruits tend to discover food sources of higher quality than scouts [23] , [24] , [60] . By dancing only for high quality food sources (see Figure S1 and [3] , [8] ) foraging bees effectively filter information, which allows recruits to exploit selectively the best food sources known to the colony [3] , [61] , [62] . After successful recruitment, costs of continuing to visit the food patch are low as bees quickly locate the previously visited patches using route-memory [63] , [64] , while benefits potentially remain high. The more stable the environment, the longer these benefits can accumulate. This is true even if there are short term fluctuations in a given patch's quality, such that it switches back and forth between profitable and non-profitable, because colonies with persistent individuals using private information can quickly decide which is the most profitable patch at a given time [65] . In contrast, if the environment completely changes (e.g., profitable patches become permanently unprofitable and vice versa), as in the case where patches only lasted for a day, recruits repeatedly have to pay the recruitment costs, while benefits remain short-lived. One major difference between our model and previous models is our assumption that patches could last several days (and up to 4 weeks). Three lines of evidence suggest that this is a realistic assumption. First, it is known that flower patches in both tropical and temperate habitats often bloom for several days or even months [25] , [54] – [56] , [66] . Second, observations on naturally foraging honeybees show that foragers often return to the same patches for days and or even weeks [6] , [8] , [26] , [28] – [30] , [32] – [34] . Bees recruited by waggle dances are particularly likely to show site fidelity because food sources located by recruits are more profitable than those located by individually exploring bees [23] , [24] and profitability positively affects site fidelity [33] , [37] . In our model, foragers visited the same food patch for an average of 1 to 2.8 days, depending on the longevity of food patches ( Figure S3 ). Third, honeybees readily learn the time of day when food sources are most rewarding and return to feeding sites on subsequent days at the appropriate time of day [27] , [31] , [67] , [68] . This suggests that foragers are adapted to an environment where the spatiotemporal availability of the currently exploited food source is predictive for the next day. On the other hand, studies analyzing foraging locations by decoding waggle dances [3] show considerable daily changes in the locations that are advertised by dances. However, while such daily changes in dance maps inform us about the number of patches that receive increased exploitation (pp. 48 in [3] ), they provide no information about how long individual foragers exploit patches. Clearly, the long-term foraging behavior of honeybee foragers in natural flower patches deserves further study. Our findings help to explain the puzzling observation that experienced foragers following dances frequently ignore spatial waggle dance information [29] , [69] – [74] . Our results show that foragers should continue to visit familiar food sources if these remain profitable in order to avoid recruitment costs and the lower benefits of individual exploration [23] , [24] , [60] . Decoding the spatial information of dances that advertise alternative food patches would become more beneficial if the currently visited patch deteriorates and using memory no longer provides rewards [29] , [61] , [75] . Accordingly, a recent simulation study found that individuals do best if they rely on learned behaviors most of the time and tailor social information-use to circumstances when the environment changes [76] . It is not yet entirely clear why experienced foragers follow dances at all if they subsequently ignore the spatial information. It is possible that these foragers follow dances mainly to acquire up-to-date information about the time period a particular flower species produces rewards, rather than its location, and is therefore worth inspecting by other foragers at other locations [6] , [13] , [60] , [69] , [77] . We hypothesized that treating colonies by repeatedly switching between SI and NI states leads to confounded results and obscures the benefits of spatial information. Our simulations corroborate this hypothesis. When we used short treatment periods (2-day or 3-day cycles), as in previous empirical studies [16] – [18] , we often found no improved collection performance during periods with spatial information ( Figure 1 and 2 ), even in environments where colonies with continuous access to SI (no switch) outperformed those without SI. The best estimates of dance benefits were obtained with 12-day treatment periods (similar to [19] ). However, even this experimental design lead to a considerable underestimation of the relative benefits of spatial information in some environmental situations ( Figure 1 and 2 ). Our simulations suggest that the problems of repeatedly switching between SI and NI are caused by carry-over effects ( Figure 1 ): the foraging success of colonies during the first days of a new treatment is affected by the foraging success during the last days of the previous treatment ( Figure 1 ). It takes a few days before these carry-over effects are no longer apparent. Allowing for site-fidelity, foragers collecting at a good food source have a high probability to continue visiting this food sources irrespective of treatment changes. Hence, colonies newly switched to the NI state will initially perform well because foragers continue to visit the high quality food sources to which they were recruited by waggle dances during SI periods. If we prevented the agents from being true to a site, these carry-over effects disappeared ( Figure S7 ). An additional problem for the interpretation of data from switch-experiments is that the degree to which data is confounded depends on the spatiotemporal distribution of food sources ( Figure 1 ). This makes it especially challenging to meaningfully compare the foraging success of colonies in switch-experiments in different environments or seasons. To solve the methodological problems, we propose the following changes to experimental designs: First, switch experiments should allow colonies to recover for several days between SI and NI periods. Second, SI and NI periods should not be shorter than the average patch-age to make sure the environment changes substantially. With such a design, differences in the gained weight (or energy) should be noticeable between treatments. Alternatively, if only a qualitative result needs to be obtained (whether SI or NI is better), researchers could look at the change in weight gain over time instead of the weight gain. In other words, instead of recording the weight of colonies and calculating the day to day weight gain or loss ((weight on day t + 1) – (weight on day t)), one would calculate the day to day change in the weight gain or loss ((weight change on day t + 1) – (weight change on day t)). This latter solution does not involve a different experimental protocol and should therefore be straightforward to implement, but because of carry-over effects it would still be impossible to gain quantitative data on the difference between pure SI and NI treatments. So far, we have discussed our results in the context of honeybee recruitment by waggle dances. However, the main findings of the model – the beneficial effects of social information, particularly in more stable environments – are probably not restricted to honeybees alone. If acquiring social information is (i) more costly and (ii) subsequently associated with higher rewards than asocial information (e.g. individual exploration), then we would expect the relative payoff of social information to increase with increasing environmental stability. In support of the first assumption (i), Dechaume-Moncharmont and colleagues [21] show with a model that social information is often costlier than asocial information due to time costs (waiting for a demonstrator). The recent finding that social information is usually of high quality because demonstrators perform the most successful behavior they know of [62] supports the second assumption (ii). However, empirical research is needed to estimate costs and benefits of different types of information and test the role of environmental stability. Communication about food source locations is common in the Apini and many Meliponini (as pheromone trails), but not in bumble bees. The specific circumstances that lowered the costs or increased the benefits to dancing in Apis bees so it could evolve in the first place are not known, and our model is not concerned with the initial evolution of the dance. Recent phylogenies suggest that dancing evolved only once (reviewed in [5] ), suggesting some constraints to dance evolution even if simpler forms of recruitment seem to evolve readily [78] . Temperate bumble bees might be exposed to different resource distribution [78] . Additionally, even though the range of tested colony sizes did not affect our main result, a critical colony size for the dance to be beneficial is likely: if the waiting costs to recruits are too long, for example because only a small work force is collecting food, the benefit of finding high quality food might not outweigh these waiting costs [79] – [81] . In summary, our study and previous simulation studies suggest that dancing is most beneficial in environments where food sources vary greatly in quality [20] , are hard to find [22] , [38] and persist for several days or weeks. In such an environment, spatial information helps a colony to allocate its foragers to highly profitable, but hard-to-find food sources and exploit those for extended time periods. We argue that dancing is beneficial in almost every natural environment, but new empirical studies using different experimental designs are needed to test this prediction."
} | 4,237 |
37596630 | PMC10439622 | pmc | 433 | {
"abstract": "Background Reef-building corals are acutely threatened by ocean warming, calling for active interventions to reduce coral bleaching and mortality. Corals associate with a wide diversity of bacteria which can influence coral health, but knowledge of specific functions that may be beneficial for corals under thermal stress is scant. Under the oxidative stress theory of coral bleaching, bacteria that scavenge reactive oxygen (ROS) or nitrogen species (RNS) are expected to enhance coral thermal resilience. Further, bacterial carbon export might substitute the carbon supply from algal photosymbionts, enhance thermal resilience and facilitate bleaching recovery. To identify probiotic bacterial candidates, we sequenced the genomes of 82 pure-cultured bacteria that were isolated from the emerging coral model Galaxea fascicularis . Results Genomic analyses showed bacterial isolates were affiliated with 37 genera. Isolates such as Ruegeria , Muricauda and Roseovarius were found to encode genes for the synthesis of the antioxidants mannitol, glutathione, dimethylsulfide, dimethylsulfoniopropionate, zeaxanthin and/or β-carotene. Genes involved in RNS-scavenging were found in many G. fascicularis -associated bacteria, which represents a novel finding for several genera (including Pseudophaeobacter ). Transporters that are suggested to export carbon (semiSWEET) were detected in seven isolates, including Pseudovibrio and Roseibium . Further, a range of bacterial strains, including strains of Roseibium and Roseovarius , revealed genomic features that may enhance colonisation and association of bacteria with the coral host, such as secretion systems and eukaryote-like repeat proteins. Conclusions Our work provides an in-depth genomic analysis of the functional potential of G. fascicularis -associated bacteria and identifies novel combinations of traits that may enhance the coral’s ability to withstand coral bleaching. Identifying and characterising bacteria that are beneficial for corals is critical for the development of effective probiotics that boost coral climate resilience. \n Video Abstract Supplementary Information The online version contains supplementary material available at 10.1186/s40168-023-01622-x.",
"conclusion": "Conclusions The current study adds to the culture collection and publicly available genomes of coral-associated bacterial strains. Pure cultures are crucial for probiotic inoculation experiments [ 121 ], and bacterial genome sequences provide insights into bacterial functional potential and the relevance of bacteria to the coral holobiont. Since G. fascicularis has gained increased attention as an emerging coral model in recent years, this collection will support research aimed at establishing this model. We focused on bacteria with coral bleaching mitigation features via ROS and RNS-scavenging and provide an in-depth list of putative beneficial functions of bacteria isolated from G. fascicularis , some of which are novel for certain bacterial genera. We provide novel insights into the potential of coral-associated bacteria to export carbon. The functionality of each trait, as well as the impacts of proposed probiotic strains on coral holobiont performance during thermal stress remains to be assessed in controlled inoculation experiments. Temporal stability and localisation of the probiotic candidates within the coral holobiont also remains to be investigated. While the field of coral probiotics is still in its infancy and functioning of bacteria within the coral holobiont is not well understood, this study provides an important step for identifying suitable probiotic bacterial strains aimed at building coral climate resilience.",
"introduction": "Introduction Reef-building corals and coral reefs are under imminent threat from climate change. Increasing sea surface temperatures in combination with high irradiance levels, which often occur during summer heat waves driven by climate change, are the major cause of mass coral bleaching [ 1 ]. Coral bleaching is the breakdown of the obligate symbiosis between the coral host and its algal symbionts of the family Symbiodiniaceae, which results in the separation of the symbionts from the coral host tissues. This leaves the host in a carbon-deprived state [ 2 ], which is often followed by coral death and the degradation of coral reefs. There is a growing concern that ocean warming is progressing too rapidly for natural adaptation of corals to keep pace due to their relatively long generation times. This notion has led to a new field of research aimed at accelerating evolutionary processes to enhance coral bleaching resilience [ 3 ]. The concept of assisted evolution of corals includes, among other approaches, the manipulation of coral-associated microbial symbionts, such as bacteria. Coral-associated bacteria are important players in coral health and fitness as they defend the coral host from pathogens through the synthesis of antimicrobial compounds [ 4 ], produce antioxidants and cycle nutrients such as nitrogen, sulphur, carbon and phosphorus [ 5 ]. Correlations between the composition of coral-associated bacterial communities and coral heat tolerance suggest bacteria play beneficial roles in coral heat tolerance [ 6 ], but the underlying mechanisms are currently unknown. Microbiome manipulation has been successfully applied in fields like agriculture and medicine [ 7 ], but it is still in its infancy in cnidarians. Nevertheless, the feasibility of coral microbiome manipulation has recently been demonstrated [ 8 , 9 ]. Further, recent success has been achieved in treating coral white pox disease [ 10 ] and stony coral tissue loss disease [ 11 ]. Studies aimed at increasing coral bleaching resilience via microbiome manipulation have also shown positive results, even though it remains to be explored if and how the added bacteria were driving the improved tolerance to heat stress. These studies inoculated corals with bacteria isolated from corals and seawater and that were tested for potentially beneficial functions (antimicrobial activities against pathogens, activity of the antioxidant enzyme catalase and the presence of genes responsible for sulphur and nitrogen cycling) and demonstrated reduced thermal bleaching and reduced phenotypic responses to pathogen infection in Pocillopora damicornis [ 12 ] and enhanced bleaching recovery in Mussismilia hispida [ 13 ]. Another study showed increased bleaching tolerance in heat sensitive Pocillopora sp. and Porites sp. that were inoculated with whole microbiomes from heat-tolerant conspecifics [ 14 ]. The main theory of bleaching is the oxidative stress hypothesis which poses that increased temperature and light damage photosystem II, reductive pentose phosphate cycle reactions and thylakoid membranes [ 15 , 16 ] in the photosymbionts, which leads to an overproduction of toxic reactive oxygen species (ROS) [ 17 ]. This can overwhelm antioxidant responses and excess ROS diffuse into the coral host cells where they cause damage to macromolecules (e.g. DNA) and trigger a cellular cascade that leads to bleaching [ 17 , 18 ]. Various ROS such as singlet oxygen ( 1 O 2 ), superoxide (O 2 − ), hydrogen peroxide (H 2 O 2 ) and hydroxyl radicals (OH − ) are continuously produced during photosynthesis in Symbiodiniaceae even under non-stress conditions [ 19 ] and are promptly scavenged by the antioxidant defence system consisting of various enzymatic and non-enzymatic mechanisms in the photosymbiont and host cells. Scavenging enzymes include catalase and superoxide dismutase, and non-enzymatic antioxidants comprise mannitol, glutathione and carotenoids [ 20 ]. In addition to ROS, reactive nitrogen species (RNS), such as nitric oxide, may be involved in coral bleaching [ 18 , 21 ]. Increased levels of nitric oxide in Symbiodiniaceae cultures and the sea anemone Exaiptasia diaphana , and increased activities of nitric oxide-producing enzymes in Symbiodiniaceae have been correlated with cnidarian bleaching and thermal stress [ 22 – 25 ]. Several studies suggest that nitric oxide plays a role in inducing host apoptotic pathways in response to symbiont dysfunction during bleaching [ 25 – 28 ]. Signalling pathways of ROS and RNS may also interact [ 18 , 22 , 26 ]. One interaction is the generation of peroxynitrite (ONOO − ) from O 2 − and nitric oxide, which disrupts electron transport within mitochondria [ 21 ] and has been linked to thermal stress in Symbiodiniaceae [ 26 ]. Aside from the oxidative stress theory which poses that the overproduction of ROS and RNS caused by light and temperature stress triggers a cellular cascade resulting in bleaching, some studies postulate that the bleaching cascade is triggered by the host’s inability to provide enough CO 2 to the faster growing Symbiodiniaceae under increased temperatures [ 29 , 30 ]. This is believed to disrupt the Calvin-Benson cycle and result in an overproduction of ROS by Symbiodiniaceae, which may leak into the host cells and initiate the bleaching cascade. A third theory is that elevated temperatures affect nutrient cycling between the coral and its algal symbiont [ 2 ]. Heat stress increases host respiration and catabolic processes, resulting in ammonium becoming available to the Symbiodiniaceae. Consequently, Symbiodiniaceae are freed from their normally nitrogen-limited state permitting their growth to increase and resulting in them using most of their photosynthate for their own growth rather than translocating it to the coral host. The Symbiodiniaceae quickly run out of phosphorus and the ensuing N:P imbalance is believed to cause a change in the composition of their thylakoid membranes and impair the photosystem [ 31 ], creating again an overproduction of ROS which may trigger the loss of the Symbiodiniaceae from the coral host. Based on the roles of ROS and RNS in coral bleaching, mechanisms that neutralise these molecules may boost coral heat tolerance. Indeed, corals and coral model animals (sea anemones) bleached less when ROS levels were decreased through the addition of exogenous antioxidant compounds [ 32 , 33 ], and a nitric oxide scavenging compound mediated decreases in photosynthetic efficiency of Symbiodiniaceae under heat stress [ 25 ]. We hypothesise that enhancing ROS and RNS-degradation within the coral holobiont by microbiome manipulation, such as inoculating corals with bacteria that have a high ROS and/or RNS-scavenging ability, may be a useful conservation strategy. While this may reduce the expulsion of algal symbionts, the coral host is likely still carbon-limited due to higher respiration rates and lower carbon translocation from the Symbiodiniaceae. Thus, ROS and RNS-scavenging bacteria that have the additional potential to translocate carbon to the host may provide an added benefit by minimising host starvation [ 34 ]. The presence of sugar transporters that can translocate carbon from the bacterial cell is therefore a likely beneficial trait. The “sugars will eventually be exported transporter” (SWEET) found in plants and other eukaryotes [ 35 , 36 ] can bidirectionally transport small sugar molecules, in particular glucose [ 37 ]. SemiSWEET proteins are the bacterial homologues of SWEET proteins. Here, we identify bacterial probiotic candidates to mitigate thermal stress in the scleractinian coral Galaxea fascicularis , which is an emerging coral model [ 38 ]. We analysed the genome sequences of 82 pure-cultured bacterial strains isolated from G. fascicularis with a focus on traits involved in ROS and RNS-scavenging, and sugar export mechanisms. Further, we surveyed for a range of other potentially beneficial metabolic pathways and genomic features that may indicate a stable host association, such as secretion systems that have been found in mutualistic endosymbionts and are known to facilitate evasion of eukaryotic host immune systems. Further, eukaryote-like repeat proteins (ELPs), including microbial ankyrin-repeat proteins (ARPs) and WD40-repeats, are suggested to facilitate host infection and generally promote stable symbiosis by assisting protein–protein interactions [ 39 , 40 ].",
"discussion": "Discussion To identify probiotic bacterial candidates that may assist in building coral bleaching resilience, we examined the functional potential of Galaxea fascicularis -associated bacteria via genomic screening. We identified G. fascicularis -associated bacteria with various novel combinations of putative beneficial functions, such as ROS and/or RNS-scavenging that may mitigate thermal stress in the coral holobiont, carbon translocation which might aid bleaching resistance and recovery, and genomic features suggested to enhance bacterial colonisation of and association with the coral holobiont. In G. fascicularis , ROS is believed to be a major driver of thermal bleaching [ 74 ], and the selected probiotic candidates may neutralise ROS and reduce or prevent bleaching in this coral species. Our selection of bacterial probiotic candidates may also be relevant for a broad range of other coral species as they are taxa commonly found in scleractinian corals. Further, G. fascicularis is known to be widespread across a broad spectrum of reef environments worldwide and even is the dominant species on some inshore fringing reefs [ 75 ]. Whole-genome sequenced bacterial isolates represent G. fascicularis -associated bacterial diversity The 82 bacterial isolates for which genomes were obtained comprise 16 families and 37 genera. These families make up the majority of the bacterial microbiome associated with the three GBR-sourced G . fascicularis colonies. The most abundant genera that were identified by 16S rRNA gene metabarcoding, Ruegeria (11.68% average relative abundance) and Alteromonas (7.73%), were also obtained in pure culture. Bacteria that were found in the G. fascicularis bacterial microbiome but which we were not able to culture included genera such as Endozoicomonas , which has gained increased attention as a potential indicator for coral health [ 76 ] and which has been shown to play a role in the coral sulphur cycle by metabolising DMSP to DMS [ 77 ]. In general, the microbiome of the GBR colonies is similar to that of G. fascicularis from the South China Sea [ 78 , 79 ]. When expanding the focus to bacterial microbiomes of scleractinian corals in general, all bacterial classes in our culture collection (i.e. Gammaproteobacteria, Alphaproteobacteria, Bacilli and Flavobacteriia) have been found to be coral-associated [ 80 , 81 ]. Taken together, our collection of bacterial genomes was a comprehensive representation of the bacterial diversity associated with G. fascicularis and scleractinian corals in general. Potential of G. fascicularis -sourced probiotic candidates to scavenge ROS and RNS Three of the selected probiotic candidates, i.e. Ruegeria sp. DMG2200320, Roseibium sp. DMG3300306 and Roseovarius sp. DMG2200361, exhibit the potential of producing the two antioxidants DMS and DMSP [ 82 ]; Pseudophaeobacter sp. DMG2200305 contains genes for the production of DMS only. DMS synthesis via demethylation or cleavage of DMSP has previously been reported for strains of Pseudophaeobacter sp. [ 83 ], Ruegeria sp. [ 84 , 85 ] and Roseovarius sp. [ 86 ], whereas DMSP biosynthesis via dsyB is novel for the latter two genera. Both DMS and DMSP synthesis are known for Roseibium sp. [ 87 , 88 ]. Generally, DMSP and DMS are both considered effective antioxidants that scavenge OH − , with DMS being the more reactive compound [ 82 ]. DMSP and DMS can also act as a carbon and sulphur source for microbes [ 89 ] or can shape coral microbial communities via antimicrobial properties [ 90 ]. DMSP can also act as a chemo-attractant for the coral pathogen Vibrio coralliilyticus to colonise the coral host [ 91 ]. We detected the potential to synthesise the antioxidant mannitol in three of the selected probiotic candidates ( Ruegeria sp. DMG2200320, Roseibium sp. DMG3300306, Pseudovibrio denitrificans DMG2200345), which is a novel observation for Ruegeria sp. and P. denitrificans . Mannitol scavenges OH − [ 92 ] and could mitigate thermal stress in corals. For instance, exogenous addition of mannitol to corals reduced Symbiodiniaceae loss in Agaricia tenuifolia during heat stress [ 32 ], and reduced DNA damage in Pavona divaricate host tissues during thermal stress [ 93 ]. It also mitigated bleaching in E. diaphana [ 94 ] . The synthesis of other antioxidants, zeaxanthin and β-carotene, was identified in the probiotic candidate Muricauda sp. DMG2200308, supporting previous findings about this genus [ 95 , 96 ]. Carotenoids belong to the most potent antioxidants by quenching the highly reactive singlet oxygen [ 97 , 98 ]. Zeaxanthin produced by Muricauda sp. strain GF1 mitigated light and thermal stress via the reduction of ROS in cultured Symbiodiniaceae [ 95 ]. Zeaxanthin produced by Muricauda sp. isolated from coastal marine sands was demonstrated to scavenge nitric oxide [ 96 ]. Thus, zeaxanthin is an antioxidant that might mitigate both ROS and RNS overproduction in the coral holobiont. All selected probiotic candidates display nitric oxide reduction potential via norBC , whereas Roseibium sp. DMG3300306 and Pseudophaeobacter sp. DMG2200305 also contains hmp for this property. Furthermore, all candidates, except for Roseovarius sp. DMG2200361, show the potential to convert nitrous oxide to nitrogen. Nitric oxide ( norBC ) and nitrous oxide reductases ( nosZ ) were previously documented in Ruegeria [ 84 ], Muricauda [ 99 ], Roseibium [ 100 ], Pseudovibrio [ 101 ] and Roseovarius ( norBC only) [ 102 ], while the discovery of hmp in Roseibium sp. DMG3300306 and Pseudophaeobacter sp. DMG2200305 is novel for these genera. The reduction of nitric oxide via probiotic bacteria might be advantageous for the coral under thermal stress, especially if targeting its algal symbionts, since the addition of a nitric oxide scavenging compound alleviated a decrease in photosynthetic performance in Symbiodiniaceae cultures under heat stress [ 25 ]. Three of the selected probiotic candidates also show the potential to reduce peroxynitrite to nitrate via ahpC ( Muricauda sp. DMG2200308, P. denitrificans DMG2200345 and Pseudophaeobacter DMG2200305), which is a novel pathway for Muricauda and Pseudophaeobacter . A relevant observation from an earlier study is that ahpC from coral-associated Bacillus aquimaris protected Escherichia coli from oxidative stress [ 103 ]. The presence of all four RNS-scavenging features is novel for Pseudophaeobacter , calling for further studies testing their functionality. Probiotic candidates with carbon translocation potential The ability to export carbon, especially glucose [ 104 ], to the coral host may be an advantageous trait for coral probiotic candidates, and may be used to provide energy to enhance bleaching tolerance and facilitate bleaching recovery of the carbon-starved host [ 34 , 105 ]. This study provides the first report of any coral-associated bacteria possessing semiSWEET protein genes, potentially giving them the ability to export small sugar molecules just like the eukaryotic homologue SWEET, although sugar export has not yet been confirmed in bacteria [ 72 ]. SemiSWEET transporters were discovered, among others, in the selected probiotic candidates Roseovarius sp. DMG2200361, Pseudophaeobacter sp. DMG2200305 and P. denitrificans DMG2200345, novel findings for these genera, suggesting that these strains might export carbon to the coral host. Probiotic candidates possess putative traits for a stable host association For bacterial probiotics to be a viable intervention to enhance coral climate resilience, long-term beneficial effects on the coral holobiont must be achieved [ 106 ]. It is therefore important that probiotic bacteria form a stable association with the holobiont. In this study, we identified proteins with ARPs and/or WD40-repeat proteins in four of the selected probiotic candidates ( Ruegeria sp. DMG2200320, Muricauda sp. DMG2200308, Roseibium sp. DMG3300306 and Roseovarius sp. DMG2200361). These ELPs have been hypothesised to promote stable symbiotic associations through bacterial protein-eukaryotic host protein interactions, as indicated for a range of coral-associated bacteria from Porites lutea [ 107 ] and the suggested coral bacterial symbiont Endozoicomonas sp. [ 77 ]. Muricauda sp. DMG2200308 and Roseibium sp. DMG3300306 exhibited the highest numbers of ARPs and WD40-repeats among G. fasciularis -associated bacterial isolates, and we hypothesise that those may facilitate symbiotic interactions with the coral host and Symbiodiniaceae. T4SS, detected in Roseovarius sp. DMG2200361 and Pseudophaeobacter sp. DMG2200305, might also assist in their association with the coral holobiont by translocating ankyrin-repeat-containing effectors, a mechanism that has been reported for a range of bacteria [ 40 ]. One study suggested that T4SS found in Roseovarius mucosis aids colonisation of its dinoflagellate host Alexandrium ostenfeldii [ 108 ]. T6SS, found in Roseibium sp. DMG3300306 and P. denitrificans DMG2200345, is a commonly found secretion system in bacteria that plays a role in virulence and antibacterial activity [ 109 ] and might promote bacterial communication [ 110 ]. T6SS in Vibrio fischeri has been described to play a role in the establishment of the symbiosis with the bobtail squid via eliminating bacterial competitors [ 111 ]. This secretion system has also been detected in the coral bacterial symbiont Endozoicomonas sp. [ 112 ]. Whether both T6SS and T4SS, which are the only secretion systems (including T3SS) that can transport proteins across an extra host cell membrane [ 113 ], may play beneficial roles in the establishment of the discussed probiotic candidates with the coral host and/or Symbiodiniaceae requires further studies. Some of the selected probiotic candidate genera isolated in this study are known to form stable associations with corals, such as Ruegeria spp. [ 114 ]. For example, members of Ruegeria have been observed in both early and adult life stages of Pocillopora damicornis [ 115 ]. Moreover, some of the selected probiotic candidate genera are closely associated with Symbiodiniaceae or other algae. For example Ruegeria pomeroyi forms a symbiosis with the diatom Thalassosira pseudonana , providing it with the essential vitamin B 12 [ 116 ]. Symbiodiniaceae and corals cannot generate vitamin B 12 [ 117 ], but require it as a cofactor for enzyme functioning in central metabolism [ 118 ]. For instance, cultured Symbiodiniaceae can grow without vitamin B 12 addition to the culture medium so long as bacteria are present [ 119 ]. Thus, the proposed probiotic candidate Ruegeria sp. DMG2200320 might also contribute to coral holobiont functioning by providing the essential vitamin B 12 to Symbiodiniaceae and the coral. The genera Muricauda, Roseibium and Roseovarius sp. are associated with different Symbiodiniaceae species in culture [ 120 ]. Using probiotic candidates which are co-localised with Symbiodiniaceae is particularly appealing as excess production of ROS and RNS mostly occurs there. In a recent study, Roseovarius sp. with high ROS-scavenging ability isolated from Symbiodiniaceae contributed to Symbiodiniaceae growth under elevated temperatures after inoculation (Heric K, Maire J, Deore P, Perez-Gonzalez A, van Oppen MJH: Inoculation with Roseovarius increases thermal tolerance of the coral photosymbiont, Breviolum minutum, under review), further suggesting that strains of this genus could be beneficial for coral probiotics."
} | 5,964 |
29038467 | PMC5739288 | pmc | 435 | {
"abstract": "Bacterial cell-cell signaling, or quorum sensing, is characterized by the secretion and group-wide detection of small diffusible signal molecules called autoinducers. This mechanism allows cells to coordinate their behavior in a density-dependent manner. A quorum-sensing cell may directly respond to the autoinducers it produces in a cell-autonomous and quorum-independent manner, but the strength of such self-sensing effect and its impact on bacterial physiology are unclear. Here, we explored the existence and impact of self-sensing in the Bacillus subtilis ComQXP and Rap-Phr quorum-sensing systems. By comparing the quorum-sensing response of autoinducer-secreting and non-secreting cells in co-culture, we found that secreting cells consistently showed a stronger response than non-secreting cells. Combining genetic and quantitative analyses, we demonstrated this effect to be a direct result of self-sensing and ruled out an indirect regulatory effect of the autoinducer production genes on response sensitivity. In addition, self-sensing in the ComQXP system affected persistence to antibiotic treatment. Together, these findings indicate the existence of self-sensing in the two most common designs of quorum-sensing systems of Gram-positive bacteria.",
"introduction": "Introduction In bacterial quorum-sensing systems, a secreted, diffusible signal molecule called an autoinducer, activates a cognate receptor to control a wide array of quorum-sensing responses in a density-dependent manner [ 1 ]. The design of quorum-sensing systems may allow the receptor to interact cell autonomously with autoinducer produced by the same cell. The overall concentration sensed by the receptor will be the sum of the local effect and the average concentration in the environment ( Supplementary Discussion ) [ 2 ]. We term the autocrine component, self-sensing, to distinguish it from the non-autonomous, quorum-sensing sources [ 3 ]. Mathematical analysis suggests that the self-sensing may depend on the rate of autoinducer secretion, diffusion and degradation, and on possible compartmentalization of the autoinducer and its receptor ( Supplementary Discussion ). Recently, strong self-sensing was observed in a synthetic yeast quorum-sensing system based on the alpha mating factor [ 4 ], but the impact of self-sensing in endogenous quorum-sensing systems has not been significantly explored. This work aimed to explore the existence and impact of self-sensing in the quorum-sensing, Gram-positive bacterium B. subtilis. This species codes for two types of quorum-sensing systems; ComQXP and Rap-Phr, which represent the two main types of quorum-sensing families found in Gram-positive bacteria – a membranal receptor sensing a long or modified peptide and a cytoplasmic receptor sensing a short unmodified peptide ( Supplementary Fig. 1 ) [ 5 , 6 ]. Both the ComQXP system and many Rap-Phr systems control the response regulator ComA, which controls the production of surfactin and the induction of the K-state through its regulation of the srfA operon [ 7 ]. In the K-state, bacteria become competent for DNA transformation and persist in the presence of antibiotics [ 8 , 9 ]. The ComQXP system is encoded by the comQXP operon. In this system, the ComX autoinducer is encoded by the comX gene and post-translationally cleaved and prenylated by ComQ [ 10 – 12 ]. ComX binds to and activates ComP, a membranal histidine kinase receptor, which then phosphorylates the response regulator ComA [ 13 ]. The Rap-Phr systems on the other hand, code for cytoplasmic Rap receptors and short, unmodified Phr autoinducers. The Phr autoinducer is expressed as a pre-peptide, which is secreted through the general secretory pathway and undergoes further extracellular cleavage events to form the mature, unmodified autoinducer peptide. The Phr peptide is then imported through the oligopeptide permease system into the cytoplasm, where it prevents its cognate Rap receptor from repressing its target response regulators and is degraded by peptidases. Here, we found that in both systems, the autoinducer-secreting cells had a stronger quorum-sensing response than non-secreting cells when co-cultured. A combination of genetic and quantitative analyses ascribed this difference to a self-sensing mechanism, as opposed to a regulatory one.",
"discussion": "Discussion Formation of concentration gradients around autoinducer-secreting cells is a natural consequence of the diffusion process, even in a well-mixed environment. However, our theoretical estimation suggests that a self-sensing mechanism which is exclusively diffusion-based cannot explain the level of self-sensing observed in our analysis of both the ComQXP and Rap-Phr systems ( Supplementary Discussion ). Enhanced self-sensing may occur if the autoinducer and the receptor interact within a subcellular compartment ( Supplementary Discussion ), but the design of both B. subtilis quorum-sensing systems seems to prevent compartmentalization of signal and receptor within the cytoplasm. For the Rap-Phr system, the presented data suggest that self-sensing may arise from pre-peptide secretion failures, which will lead to cytoplasmic co-occurrence of Rap and the cell-autonomously produced mature Phr peptide ( Supplementary Fig. 9, Supplementary Fig. 10 ). The impact of self-sensing in the Rap-Phr system is mitigated in the wild-type, by co-regulation of ComA by the ComQXP system ( Supplementary Fig. 9b,c ) and by the transcriptional regulation of many Rap-Phr systems by ComA [ 17 , 22 ]. The underlying mechanism for self-sensing in the ComQXP system is unknown. However, the membranal localization of the ComP receptor and the hydrophobicity of the ComX prenyl chain suggest that self-sensing in the ComQXP system may occur through membranal compartmentalization of the receptor with the autoinducer prior to its secretion ( Supplementary Discussion ) [ 11 – 13 ]. The specific mechanisms underlying quorum sensing can have a significant impact on the quantitative aspects of quorum-sensing response and on its evolutionary fate [ 23 , 24 ], raising the question whether self-sensing is an adaptive feature of these systems, and results from a direct selective pressure. Quorum sensing is known to control various types of activities with a different impact on individual and group fitness [ 25 ]. Self-sensing is disadvantageous when controlling a public benefit, but provides an advantage when controlling private traits, such as antibiotic persistence ( Fig. 3 ). This explanation for the existence of self-sensing is problematic, as a similar private benefit would arise from constitutive activation of the quorum-sensing regulated factors [ 26 ]. Finally, some activities are intermediate between individual and public, leading to a selective advantage for the trait in aggregated, but not in planktonic form [ 27 – 29 ]. A hydrophobic autoinducer, such as ComX, may better inform cells on their aggregate status than a hydrophilic one [ 30 ]. In this case, self-sensing would be a tolerable side effect of aggregation-sensing. Altogether, our results demonstrate that self-sensing is observed in the two most common designs of Gram-positive quorum-sensing systems – a membranal extracellular receptor, with a modified or long peptide autoinducer (ComQXP) and a cytoplasmic receptor, with an unmodified peptide autoinducer (Rap-Phr). Theoretically, these designs better compartmentalize signal production and sensing than the design of Acyl homoserine lactone based systems, where both signal production and reception are intracellular, yet they still show a self-sensing behavior. Further work will be required to identify the mechanisms underlying self-sensing, its impact on the design and evolution of quorum-sensing systems and its prevalence in other types of quorum-sensing systems."
} | 1,959 |
39636109 | PMC11705803 | pmc | 436 | {
"abstract": "ABSTRACT Understanding the extracellular electron transfer mechanisms of electroactive bacteria could help determine their potential in microbial fuel cells (MFCs) and their microbial syntrophy with redox-active minerals in natural environments. However, the mechanisms of extracellular electron transfer to electrodes by sulfate-reducing bacteria (SRB) remain underexplored. Here, we utilized double-chamber MFCs with carbon cloth electrodes to investigate the extracellular electron transfer mechanisms of Desulfovibrio vulgaris Hildenborough ( Dv H), a model SRB, under varying lactate and sulfate concentrations using different Dv H mutants. Our MFC setup indicated that Dv H can harvest electrons from lactate at the anode and transfer them to cathode, where Dv H could further utilize these electrons. Patterns in current production compared with variations of electron donor/acceptor ratios in the anode and cathode suggested that attachment of Dv H to the electrode and biofilm density were critical for effective electricity generation. Electron microscopy analysis of Dv H biofilms indicated Dv H utilized filaments that resemble pili to attach to electrodes and facilitate extracellular electron transfer from cell to cell and to the electrode. Proteomics profiling indicated that Dv H adapted to electroactive respiration by presenting more pili- and flagellar-related proteins. The mutant with a deletion of the major pilus-producing gene yielded less voltage and far less attachment to both anodic and catholic electrodes, suggesting the importance of pili in extracellular electron transfer. The mutant with a deficiency in biofilm formation, however, did not eliminate current production indicating the existence of indirect extracellular electron transfer. Untargeted metabolomics profiling showed flavin-based metabolites, potential electron shuttles. IMPORTANCE We explored the application of Desulfovibrio vulgaris Hildenborough in microbial fuel cells (MFCs) and investigated its potential extracellular electron transfer (EET) mechanism. We also conducted untargeted proteomics and metabolomics profiling, offering insights into how DvH adapts metabolically to different electron donors and acceptors. An understanding of the EET mechanism and metabolic flexibility of Dv H holds promise for future uses including bioremediation or enhancing efficacy in MFCs for wastewater treatment applications.",
"conclusion": "Conclusions In this study, we found that Dv H was able to attach to the electrodes of MFC and produced different types of filaments connecting the bacterium to the electrode surface; however, direct tests of the conductivity of the pili were inconclusive. During electroactive respiration with DvH on both electrodes in a dual chamber MFC, there was electricity production with a maximum power density of ~0.074 W/m 2 . The ratio of electron donor to acceptor in both chambers and the presence of pili and biofilm were found to be key variables for electricity generation. Untargeted metabolomics profiling showed flavin-based metabolites, which are potential electron shuttles. Taken together, these results indicated that Dv H likely has multiple extracellular electron transfer pathways facilitating growth on solid surfaces that can be utilized either as the electron acceptor or donor ( Fig. 6 ). Future work is needed to confirm the activity of this variety of possible mechanisms in contributing to energy production in this metabolically versatile microorganism.",
"introduction": "INTRODUCTION Microbial fuel cells (MFCs) have been used as a promising technology for electrical energy generation, which uses microbes to transfer the chemical energy of organic compounds into electricity ( 1 ). Novel insights have been incorporated into MFCs for energy generation as well as the microbial transformation of wastes. For instance, MFCs have been investigated for the conversion of wastewater containing organic compounds and sulfate to electricity using sulfate-reducing bacteria (SRB) and sulfide-oxidizing bacteria ( 2 ). These sulfate-containing wastewaters are produced by many processes including mining, food processing, pulp and paper wastewater, animal husbandry, and more ( 3 ). By culturing SRB in MFCs, previous studies have achieved the removal of sulfate and organic compounds with electricity production ( 4 , 5 ), ranging from 0.013 W/m 2 to 0.68 W/m 2 ( 6 – 8 ). The overall electricity generation of an MFC mainly relies on the efficiency of extracellular electron transfer (EET) from electrogenic bacteria to the electrode ( 9 ). Electrogenic bacteria can route their electron transport chain to the exterior of the cell through various EET mechanisms ( 10 ). Two major EET mechanisms are direct electron transfer and indirect electron transfer. Direct electron transfer mainly relies on outer surface redox molecules, proteins, and conductive nanowires. For instance, in many species, such as Geobacter sulfurreducens , Shewanella oneidensis , and Acidithiobacillus ferrooxidans , EET can be mediated by outer membrane c -type cytochromes (e.g., OmcA-MtrCAB protein complexes) ( 11 ). G. sulfurreducens also exchange electrons through nanowires, which are pili formed by protein filaments ( 12 ). S. oneidensis MR-1 forms nanowires through extensions of the outer membrane and periplasm that include the multiheme cytochromes that are responsible for EET ( 13 ). Indirect electron transfer involves the transfer of electrons through small redox-active organic molecules, electron shuttles, excreted by cells or added exogenously. Different species secrete various extracellular electron carriers such as flavins and phenazine derivatives ( 14 ). Previous studies demonstrated the coexistence of direct electron transfer and indirect electron transfer in S. oneidensis ( 15 ). S. oneidensis could simultaneously transfer electrons through direct contact with the electron acceptor and also produce flavins ( 15 ). Large numbers of studies have demonstrated that Geobacter and Shewanella species can power MFCs with high power density ( 11 , 12 , 16 , 17 ), due to their unique EET mechanism. SRB can utilize organic compounds and gases (e.g., hydrogen) as electron donors ( 18 ). Recent studies also found that some SRB can use electrodes as electron donors for energy production ( 19 ). Despite this, the mechanism for SRB extracellular electron uptake is not clear due to the difficulty in distinguishing the electron uptake reaction (e.g., EET mechanism) and hydrogen evolution on the electrode surface. Knowledge of EET mechanisms has major implications for being able to understand, control, or intervene in several environmental problems caused by SRB, such as corrosion of steel, concrete, and electrode ( 20 ); souring of oil ( 21 ); altering mobility of toxic heavy metals (e.g., Cr and U) ( 22 ); and providing for syntrophic growth with other microorganisms (e.g., methanogens) ( 23 ). Desulfovibrio vulgaris Hildenborough ( Dv H), a model SRB strain, was reported to cause the corrosion of carbon steels due to their ability to harvest extracellular electrons from elemental iron oxidation ( 24 ). The intracellular electron transport of Dv H from the electron donor (i.e., lactate) to the electron acceptor (i.e., sulfate) was proposed via two ways: (i) hydrogen cycling pathway that uses hydrogen as an intermediate electron carrier between the periplasm and the cytoplasm, and (ii) a pathway that bypasses hydrogen cycling and transfers electrons directly to the membrane-bound menaquinone pool ( 25 ). However, there is little in-depth knowledge regarding the EET including direct electron transfer and indirect electron transfer from Dv H to electrodes. It was found that D. ferrophilus IS5 was able to adopt the multi-heme cytochromes containing at least four heme-binding motifs in acquiring energy from solid electron donors ( 26 ). Kang et al. found that D. desulfuricans was able to conduct direct electron transfer through cytochrome c proteins ( 27 ). However, so far, no outer membrane c -type cytochromes of Dv H have been identified ( 28 ). Deng et al. ( 29 ) found that Dv H biosynthesized iron sulfide (FeS) nanoparticles on the cell membrane, which could enhance extracellular electron uptake significantly. Zhou et al. ( 30 ) also demonstrated the accumulation of iron sulfide crystallite on the surface of the cell by obtaining electrons intracellularly. Thus, one of the direct electron transfer mechanisms of Dv H could be via iron sulfide nanoparticles. Moreover, D. desulfuricans utilized electrically conductive nanoscale filaments to transfer electrons to insoluble electron acceptors (i.e., iron(III) oxide) ( 31 ). However, the major characteristics of these filaments of D. desulfuricans have not been identified. Dv H could also form a biofilm, which is dependent on protein filaments such as flagella and pili ( 32 ). However, the role of pili, flagella, and relevant biofilm in Dv H EET has not been investigated. Flavins, such as riboflavin and flavin adenine dinucleotide (FAD), are well-known electron shuttles ( 33 ), which carry electrons among multiple redox reactions and play an important role in indirect electron transfer. An investigation of the EET mechanisms both to and from Dv H to the electrode will not only fill the knowledge gap but also encourage the extensive utilization of MFCs in the treatment of sulfate-containing wastewater by SRBs. In this study, we investigated the different EET mechanisms employed by SRB when growing with the anode as the electron acceptor and the cathode as the electron donor in an anaerobic MFC system. First, this study aimed to determine the effects of the electron donor (i.e., lactate)/electron acceptor (i.e., sulfate) ratio in the anodic chamber and cathodic chamber, respectively, on the electricity generation ( Fig. 1c ). This investigation can confirm the capacity of Dv H in the electricity generation in MFC using lactate and sulfate. To further unveil the role of pili and biofilm in EET of Dv H, the electricity generation of MFCs inoculated with Dv H JWT700, Dv H JW3422 (a mutant with a deletion of the gene coding for the pilin protein), and Dv H JWT716 (a mutant has a deficiency in biofilm formation) ( 34 ) was compared separately. Subsequently, extracellular metabolites of Dv H JWT700, Dv H JW3422, and Dv H JWT716 under electroactive respiration were analyzed to screen for potential electron shuttles. Finally, the conductivity of the surface structures of Dv H and their role in electricity generation were evaluated. Although having Dv H in both chambers of the MFC is neither typical nor what would be utilized in an industrial setting, we confirmed that Dv H could employ different EET mechanisms depending on its growth mode. These findings will allow for future improved design and implementation of SRB on MFCs or to treat sulfate-containing wastewaters. Fig 1 Schematic ( a ) and laboratory-scale prototype ( b ) of the MFC. A schematic representation of experimental flow (c). A schematic and a photograph show an experimental MFC setup with an anode and cathode separated by a PEM, displaying lactate and sulfite conversion and EET by bacteria DvH, as well as a list of main research objectives for studying EET mechanisms.",
"discussion": "DISCUSSION In this study, we observed that Dv H can harvest and send electrons to and from an electrode by varying the concentrations of the electron donor/electron acceptor ratio in the anodic chamber and cathodic chamber, respectively. By testing different Dv H mutants (i.e., Dv H JW3422 [pilin-] and Dv H JWT716 [biofilm-]), we determined that pili and biofilm contributed to the EET processes of Dv H. We also identified the extracellular metabolites excreted during growth under electroactive respiration and found molecules that could act as electron shuttles. Our results in conjunction with previous literature for DvH and other members of the Desulfovibrio genus led us to draw the schematic in Fig. 6 , which shows all the potential mechanisms of EET in DvH that we will now discuss. Fig 6 All potential EET mechanisms of Dv H. Indirect electron transfer based on electron shuttles (1) and direct electron transfer based on: EPS (2), unknown outer membrane c -type cytochromes (3), conductive pili/filaments (4), and FeS clusters (5) are presented. A diagram shows electron transfer and chemical reactions within a bacterial cell interacting with an electrode, highlighting the electron movement and oxidation pathways of lactate, acetate, pyruvate, and formate. It is important to note that these EET mechanisms are distinct from the biocorrosion processes, which are well described for Dv H, as corrosion occurs primarily during growth under dissimilatory sulfate-reducing conditions ( 45 ). Previous studies focused mainly on the biocorrosive capacity of Dv H ( 46 , 47 ) and overlooked the electricity production capacity of Dv H through MFCs. The reported coulombic efficiency in the current study is much lower ( Tables 2 and 3 ) than the ones reported in other relevant studies, which varied from 6.7% to 98.9% when different feed compositions and SRB were applied ( 8 , 48 , 49 ). The highest current densities come from mixed cultures that are usually dominated by the genus Geobacter ( 50 ). The differences in coulombic efficiency were caused by differences in MFC configurations and SRB species, which have different electron transfer mechanisms or use different electron donors ( 16 ). Interestingly, Dv H grew on both the anode and the cathode of the MFC while producing electricity, which means that Dv H is capable of bidirectional EET. Likewise, the proteomics results showed a complete proteome rearrangement for electroactive respiration in a biofilm compared with DSR biofilm and indicated that different sets of proteins were involved in the use of the anode as the electron acceptor and use of the cathode as the electron donor. From this observation, we then asked how EET was being carried out by Dv H and if it involved different EET mechanisms when the anode was used as the electron acceptor and the cathode as the electron donor. Possible indirect electron transfer mechanisms: electron shuttle molecules The first possible mechanism depicted in Fig. 6.1 is indirect electron transfer. In our study of Dv H, there were several experimental results that suggested the possibility of indirect electron transfer occurring in this anaerobic bacterium. In the MFC runs with Dv H wild-type and mutant strains, it was seen that neither the lack of biofilm nor the lack of pili completely diminished EET. This suggested the involvement of IET. For other bacteria with known IET mechanisms, small molecules called electron shuttles ferry electrons between the cell and the electron source or sink have been characterized. Shewanella species were found to secrete flavins (i.e., FMN and riboflavin) as electron shuttles, which mediate EET by binding to outer membrane cytochromes ( 51 ). S. oneidensis MR-1 can use the interaction of flavin/outer membrane c -type cytochrome complexes to regulate extracellular electron transport ( 52 ). Although Geobacter species have abundant c -type cytochromes and are thought to transfer electrons by direct contact, flavin synthesis and excretion genes are widely distributed in Geobacter species ( 53 ). Studies indicated that Geobacter sulfurreducens can uptake self-secreted riboflavin as bound cofactors for EET ( 54 ). In the metabolite analysis of DvH, it was not surprising that we detected riboflavin since we provided riboflavin as a component of the media ( Table 5 ). However, under electroactive respiration, Dv H strains did have different patterns of riboflavin utilization. Given the experimental setup for these metabolomics experiments, we cannot quantify or provide relative quantitation for these changes. However, in the proteomics experiments which were done quantitatively, we saw increases in the abundance of riboflavin synthesis genes (e.g., DVU1199, DVU1200, and DVU1201) in electroactive conditions compared with Dv H under dissimilatory sulfate reduction respiration samples that matched the observations of riboflavin usage in the metabolomics. It was previously found both riboflavin and FAD accelerated pitting corrosion and weight loss on the stainless steel caused by Desulfovibrio vulgaris biofilm ( 53 ). Flavin-like molecules were also found to increase in electron donor-limiting conditions by mass spectrometry in Desulfovibrio alaskensis G20 biofilms ( 33 ). Thus, flavins may act as electron shuttles of Dv H. In Dv H, flavin adenine dinucleotide (FAD) in quinone form which is the electron shuttle could accept two electrons and two protons to become hydroquinone form (FADH 2 ) ( 55 ). Further evidence for the possibility of IET in Dv H was seen in the cyclic voltammogram, which gave indications of both the effect of some corrosion from hydrogen sulfide as well as the possibility of electron shuttling. CVs indicated the larger EDL capacitance of plain carbon cloths than the electrodes with Dv H biofilms ( Fig. 3a and b ), which can be due to hydrogen sulfide produced by Dv H mutants poisoning the Pt wires on the surface of carbon cloths as shown in the schematic in Fig. 1 ( 56 ). Those poisonings largely reduced the pseudocapacitive contributions from Pt wires. Furthermore, the lower R ct of the anode with Dv H JWT700 (wild-type) demonstrated the formation of more effective biofilm on the anode that enhanced the electron transfer process ( Fig. 3d and c ) as reported for Klebsiella variicola ( 57 ). However, non-pili-forming Dv H JW3422 and non-biofilm-forming Dv H JWT716 did not enhance the electron transfer process of anodes and non-biofilm-forming Dv H JWT716 presented less conductive biofilm which may be due to the semiconductive EPS and the poisoning of Pt wires via hydrogen sulfide ( 56 , 58 ). Irreversible oxidation processes appear in both the anodic and cathodic suspensions of Dv H JWT700. In both chambers, the suspension of Dv H JWT700 (wild-type) had a larger oxidation current density than Dv H JW3422 and Dv H JWT716, suggesting that Dv H JWT700 secreted more stable and oxidative compounds than the other two. The obtained CV curves are very similar to those in a previous study that identified these compounds as quinone or benzene derivatives, which suggests similar molecules may participate in DvH EET ( 59 ). However, the cell-surface redox-active proteins as well as further characterization of small molecule contributions need to be investigated to determine the free-flavin-mediated electron-shuttling mechanism of Dv H. Possible direct electron transfer mechanisms: Dv H utilized filaments that facilitated but not necessarily conducted electron transfer from cell-to-cell and to the electrode Throughout our various running conditions in the MFC, we consistently observed that greater attachment and larger biofilm led to better electricity generation in parameters such as power density and columbic efficiency, which suggests the possibility for direct electron transfer. Furthermore, in these high-performing runs, microscopy images revealed the presence of biogenic extracellular structures appearing to connect the cell to the electrode surface and frequently other cells in the biofilm. To evaluate the possibilities for direct electron transfer mechanisms in DvH EET, we considered known extracellular structures and enzymes on the cell surface or integral outer membrane proteins. Biofilms of bacteria are often composed of EPS or extracellular polymeric substances ( Fig. 6.2 ). In our study, a thin biofilm of DvH attached to the electrode surface facilitated by a network of extracellular structures was beneficial for the growth of DvH in electroactive conditions. Dv H did not produce an extensive exopolysaccharide matrix and its biofilm formation is reported to be dependent upon protein filaments ( 32 ). EPS is not known to be conductive, but it was previously demonstrated that extracellular polymeric substances include polysaccharides, proteins, glycoproteins, glycolipids, and humic substances, which these molecules can possess some semiconductive properties ( 60 ). Thus, EPS or biofilm proteins could trap conductive molecules which may be contributing to or mediating a direct electron transfer mechanism. It is more likely however that EPS or the biofilm matrix of DvH provides the structural support or decreased distance that enables EET from the cell to an electrode or conductive surface. The set of extracellular structures we considered are conductive outer membrane proteins such as outer membrane cytochrome c ’s (OMCC) ( Fig. 6.3 ). In other bacteria, OMCCs are a well-described direct electron transfer mechanism. Different outer membrane c -type cytochromes such as MtrC, OmcA, OmcE, and OmcS offer various routes of electron transfer extracellularly for S. oneidensis and G. sulfurreducens ; however, DvH lacks homologs to any of these characterized direct electron transfer proteins ( 61 , 62 ). Previous studies demonstrated that Dv H contains several membrane-bound redox complexes such as Qrc (quinone reducing complex) and Hmc (high molecular weight cytochrome) complex that can accept electrons ( 63 , 64 ). Qrc complex can accept electrons from the low-redox potential hemes of a type one soluble cytochrome c 3 protein, TpI c 3, whereas Hmc transferred the electrons by transporting H + from cytoplasmic lactate oxidation to the periplasmic cytochrome c 3 network ( 65 ). However, none of these proteins were located at the outer membrane. Additional evidence in the CV had a small redox peak in the case of Dv H JW3422 (lacking pilin) revealing that it might contribute to the oxidation reaction through unknown outer membrane c -type cytochromes. Despite this, some uncharacterized proteins such as DVU1174, DVU0401, DVU1359, DVU0842, and DVU2997 that we observed on electrodes should be further investigated to determine their functions in EET of Dv H but are not predicted to be cytochrome containing and computational predictions of their subcellular localization was inconclusive. Based on findings in a previous study that extracellular enzymes such as hydrogenase and formate dehydrogenase could mediate a direct electron uptake for Methanococcus maripaludis, we considered if such enzymes could have a role in Dv H EET ( 66 ). However, Desulfovibrio species mainly have cytoplasm-located and periplasmic hydrogenases that contribute to intracellular EET and hydrogen formation ( 67 , 68 ). More outer membrane redox complexes need to be investigated to reveal the possibility of direct electron transfer mediated by OM proteins in Dv H. Finally, for direct electron transfer mechanisms, we considered the possibility of conductive nanowires made out of protein ( Fig. 6.4 ) or iron sulfide minerals ( Fig. 6.5 ). A previous study reported that the presence of nanowires of Candidatus Desulfofervidus HotSeep-1 cell depended on substrates hydrogen, which contributes to the interspecies electron transfer ( 69 ). Deltaproteobacteria have many different but not always fully described extracellular electron transfer mechanisms including groups like cable bacteria ( 70 , 71 ), indicating that Dv H may have similar electron transfer mechanisms with other bacteria in Deltaproteobacteria , such as Geobacter sp., as indicated by protein sequence similarity of their pilins (Fig. S7). Our results from several experiments indicated that the pili appear to play an important role in mediating the extracellular electron transfer processes in DvH, whether by facilitating attachment or possibly through conduction. Although the SEM-preparatory dehydration process may collapse or separate outer-membrane materials and break filaments, it was observed that although Dv H exhibited then biofilm formation, it produced numerous filaments or pili on the surface of carbon cloths ( Fig. 4 ). We noticed that pili and/or filaments were effective for the attachment of cells of Dv H JWT700 on the surface electrode or functional as the networks between cells ( Fig. 4 ). However, no such attachment using filaments was observed for mutant strain Dv H JW3422 (pilin-). In addition to attachment, pili-related proteins (DVU2118, DVU2227, and DVU1262) were present in the anodic chamber in electroactive conditions but were not found in Dv H under dissimilatory sulfate reduction respiration ( Fig. 5 ; Table S1 ). Crucially, in this study, MFC cultivating non-pili-forming Dv H JW3422 exhibited a lower power density than the wild-type strain ( Fig. 2 ). However, no definitive answer as to the conductivity of pili from DvH can be obtained based on the current cAFM findings. Given the inconclusive results of our experiment, we compared our results and DvH proteins with reports in the literature on conductive pili. Type IV pili can be categorized into two subclasses, type IVa pili and type IVb pili, based on the sequence and length of the pilin subunit ( 72 ). Geobacter sulfurreducens, Geobacter bremensis, Desulfuromonas thiophila , and so on that were reported to have conductive pilin (e-pili) have type IVa structure ( 73 ). Although Holmes et al. ( 73 ) reported pili in Desulfovibrio vulgaris were the long type IVa pilin, according to the findings in the protein sequence alignment (Figs. S6 and S7), the Dv H pilin system is closer to the type IVb system. Until now, the pili structure of Dv H has not been well studied. Normally, the conductivity depends on the composition of the amino acid chain of the major pili ( 74 ). A high density of aromatic amino acids and a lack of substantial aromatic-free gaps along the length of long pilins may be important features of e-pilin ( 74 ). Two pilin proteins were found on the genome of Dv H: one is major pili, which belong to Flp family type IVb pili (DVU2116) (Fig. S6), and the other, prepilin-type N-terminal cleavage/methylation domain-containing protein, which may be the minor pili (putative PilE) of Dv H (Fig. S7). We further compared the major pilin and minor pili with the e-pili reported previously. No similar trend between sequences of the amino acid chain of major pili and e-pili was found. However, minor pili of Dv H had a lower E value and higher query cover percentage, indicating the minor pili shared a certain similarity with e-pili (Table S5). E-pili normally has phenylalanine (F), which aromatic amino acid, at the N terminus, and the majority have leader peptides with less than 12 amino acids ( 74 ). Instead of phenylalanine, the minor pili have Tyrosine (Y), which is also an aromatic amino acid. Thus, it is possible that the minor pili that are conductive may contribute to EET of Dv H. Being similar to Dv H, D. desulfuricans produces nanoscale filaments ( 31 ). These unidentified filaments were confirmed to be electrically conductive for extracellular electron transfer. D. desulfuricans also have both Flp family type IVb pilin and prepilin-type N-terminal cleavage/methylation domain-containing protein according to the identical protein group database in NCBI. Through the above analysis, pili of Dv H did not account for 100% of EET, suggesting additional co-occurring direct or indirect mechanisms in this study. Dv H also has flagella, which are composed of flagellin proteins (i.e., DVU1441, DVU2444, and DVU2082) ( 75 ), and these proteins do not have sequence similarity with pilin of Dv H nor e-pilin. Flagellar and histidine kinase-related proteins were dominant unique peptides in the anode and cathode when compared with dissimilatory sulfate reduction respiration ( Fig. 5 ). It was reported that Dv H forms motility halos on solid media that are mediated by flagella-related mechanisms via the CheA3 histidine kinase ( 35 ). This indicated that Dv H had increased motility or surface attachment in the anodic chamber, which may promote but not conduct EET. Another type of extracellular structure was reported in previous studies that proposed biosynthesized FeS mediates the electron transport from Dv H to the electrode surface ( 30 , 76 ). For instance, Deng et al. ( 76 ) found Dv H biosynthesized FeS nanoparticles on the cell membrane in the presence of sulfate and iron as an electron conduit enabling Dv H to utilize solid‐state electron donors via direct electron uptake. However, no biosynthesized FeS was observed on the surface of the cathode based on the results of EDS elements analysis, but they were observed as precipitates at the bottom of MFCs. The FeS nanocrystallites may be washed off during the fixation preparation process of SEM. Besides the direct electron transfer conducted by FeS nanoparticles ( 77 ), our results indicated that direct contact through pili/filaments may be one of the major routes for direct electron transfer of Dv H. Our results showed that indirect electron transfer mechanisms potentially via electron shuttle small molecules, which are flavin-related ones are also happening in Dv H. This indicates that Dv H can use different possible electron transfer mechanisms to survive in less-than-ideal conditions with solid surfaces as both an electron acceptor and a donor. A comprehensive understanding of Dv H’s electron transfer mechanisms can expand its application beyond MFCs to address sulfate-containing water and wastewater in various contexts and applications. Conclusions In this study, we found that Dv H was able to attach to the electrodes of MFC and produced different types of filaments connecting the bacterium to the electrode surface; however, direct tests of the conductivity of the pili were inconclusive. During electroactive respiration with DvH on both electrodes in a dual chamber MFC, there was electricity production with a maximum power density of ~0.074 W/m 2 . The ratio of electron donor to acceptor in both chambers and the presence of pili and biofilm were found to be key variables for electricity generation. Untargeted metabolomics profiling showed flavin-based metabolites, which are potential electron shuttles. Taken together, these results indicated that Dv H likely has multiple extracellular electron transfer pathways facilitating growth on solid surfaces that can be utilized either as the electron acceptor or donor ( Fig. 6 ). Future work is needed to confirm the activity of this variety of possible mechanisms in contributing to energy production in this metabolically versatile microorganism."
} | 7,668 |
32859083 | PMC7558274 | pmc | 437 | {
"abstract": "Memristive systems can provide a novel strategy to conquer the von Neumann bottleneck by evaluating information where data are located in situ. To meet the rising of artificial neural network (ANN) demand, the implementation of memristor arrays capable of performing matrix multiplication requires highly reproducible devices with low variability and high reliability. Hence, we present an Ag/CuO/SiO 2 /p-Si heterostructure device that exhibits both resistive switching (RS) and negative differential resistance (NDR). The memristor device was fabricated on p-Si and Indium Tin Oxide (ITO) substrates via cost-effective ultra-spray pyrolysis (USP) method. The quality of CuO nanoparticles was recognized by studying Raman spectroscopy. The topology information was obtained by scanning electron microscopy. The resistive switching and negative differential resistance were measured from current–voltage characteristics. The results were then compared with the Ag/CuO/ITO device to understand the role of native oxide. The interface barrier and traps associated with the defects in the native silicon oxide limited the current in the negative cycle. The barrier confined the filament rupture and reduced the reset variability. Reset was primarily influenced by the filament rupture and detrapping in the native oxide that facilitated smooth reset and NDR in the device. The resistive switching originated from traps in the localized states of amorphous CuO. The set process was mainly dominated by the trap-controlled space-charge-limited; this led to a transition into a Poole–Frenkel conduction. This research opens up new possibilities to improve the switching parameters and promote the application of RS along with NDR.",
"conclusion": "5. Conclusions The Ag/CuO/SiO 2 /p-Si memristive device improves the switching characteristics by reducing the variability. It was successfully demonstrated by comparing the switching response without the native oxide. The traps generated by localized states in CuO nanoparticles control charge transfer and enable abrupt switching from HRS to LRS leading to a high switching ratio. The NDR phenomena were explained by the Schottky barrier and trapping/detrapping in interfacial traps in SiO 2 . The I-V characteristics on a device without oxide were investigated to determine the effect on NDR. The temperature-dependent I-V characteristics prove that the conducting mechanism consists of space-charge-limited conduction (SSLC). The combined mechanism includes SCLC, conductive filament (CF), and trapping/de-trapping in the native oxide layer: these features result in RS and NDR. In short, a heterostructure device with Ag/CuO/SiO 2 /p-Si gave highly stable non-volatile RS along with NDR that facilitates many applications.",
"introduction": "1. Introduction A new era of non-volatile resistive memory devices has emerged due to the physical limitations of existing memory devices. Due to their scope, memristors have drawn substantial attention in the previous decades. Several specific properties make memristors a favorable and promising candidate for non-Boolean neuromorphic computing [ 1 ]. Some of its most common attributes include low power consumption, high scalability, multiple switching states and non-damaging readout. Long retention characteristics allow memristors to take over the complementary metal-oxide-semiconductor (CMOS) and to integrate it with CMOS technology [ 2 , 3 ]. Negative differential resistance (NDR) along with resistive switching can have additional applications such as resonant tunneling transistors [ 4 ], high-frequency oscillators, [ 5 ] memory devices, and multi-level logic devices [ 6 ]. The NDR effect in negative bias region reduces variability in reset, which results in a controlled reset. Memristors are two-terminal memory cells with a sandwiched active layer. This simple two-terminal structure has shown a promising output in many metal oxides such as TiO 2 [ 1 ], NiO [ 2 ], ZnO [ 7 ], perovskites [ 8 ], and transition metal di-chalcogenide monolayers [ 9 , 10 ]. Among all oxides, CuO is the most widely investigated and reported material because of its good availability, easy synthesis, good reliability, non-toxic behavior, and low cost. CuO has recently been explored for memristive devices with a superior retention time, low power consumption, good compatibility with CMOS technology [ 11 ], and excellent endurance [ 12 , 13 , 14 , 15 , 16 ]. This gives CuO substantial advantages over other materials. The NDR effect found several applications in high-frequency oscillators and multi-level switching. Hence, the industry is focused on understanding, analyzing, and discovering the NDR effect along with resistive switching. Several reports have been published on NDR with ZnO [ 17 ], polymers, graphene oxide, nanocomposites [ 1 , 17 , 18 , 19 , 20 , 21 ], other transition metal oxides (TiO x [ 22 ], and FeO x [ 23 ]). However, the NDR effect is rarely reported in CuO. Recently Kadima et al. demonstrated an NDR device with ZnO nano-rod arrays [ 24 ]. Among all switching mechanisms, the conductive filament (CF) model is widely analyzed to explain switching. However, the CF model has some shortcomings. The reset in the CF model involves the rupture of filaments making a scattering voltage distribution. This deteriorates the device endurance and leads to variability in switching. In this study, we investigate the effect of native oxide on resistive switching, which reduced the variability and demonstrated the hysteresis current–voltage (I-V) characteristics along with NDR. We used an ultra-spray pyrolysis method to deposit amorphous CuO nanoparticles on p -Si to obtain a multifunctional device with resistive switching and NDR properties. The topology of the device was well defined by scanning electron microscopy (SEM). The phase purity of CuO nanoparticles was confirmed by Raman spectroscopy. The resistive switching and negative differential resistance were obtained by I-V characteristics. To understand the conduction mechanism, temperature-dependent I-V characteristics were carried to substantiate the space-charge-limited current conduction (SCLC). The log-log plot further verifies the SCLC. The prime goal of this study was to reduce the variability in switching and to reduce power consumption by introducing the native oxide layer. Here, we fabricated Ag/CuO/ITO to substantiate the effect of native oxide. I-V characteristics of Ag/CuO/ITO were compared with that of Ag/CuO/SiO 2 /p-Si. The heterostructure device improved the resistive switching (RS) by reducing the switching variability. This multifunctional device has great potential for advanced multifunctional non-volatile memories.",
"discussion": "4. Results and Discussion Figure 2 a shows the SEM image of CuO nanoparticles deposited by ultra-spray pyrolysis (USP). The CuO nanoparticles are uniformly deposited on the SiO 2 /p-Si substrate. The size of CuO nanoparticles is from 20 nm to 200 nm with an average size of 110 nm. Typical Raman peaks were observed at 268 cm −1 , 305 cm −1 , and 605 cm −1 as shown in Figure 2 b [ 25 ]. These Raman peaks are assigned to the symmetry of A g (1), B g (1), and B g (2) modes of CuO, which confirms the CuO phase formation. The I-V characteristics in Figure 3 a clearly show the hysteresis curve, which demonstrates the non-volatile resistive switching along with negative differential resistance. The voltage was swept from 0 » 3 V » 0 » −3 V » 0. A stable resistive switching (RS) is observed when the bias voltage sweeps from 0 V to 3 V; the current switches from 10 −6 A to 10 −3 A at a set voltage of 1.7 V. The device maintains a low resistive state (LRS) when the bias voltage sweeps back from positive (3 V) to negative (−0.7 V), NDR is observed at −0.8 V, and the device switches off. Figure 3 b demonstrates repeatability in switching up to 50 cycles, emphasizing the reproducibility and stability of the device. The inset in Figure 3 b shows a clear cut of the NDR. The current decreases sharply with an increase in potential. Figure 3 c shows the endurance performance of the device. Resistance was taken as a function of the number of cycles, which show the consistency and stability in low resistive state (LRS) and high resistive state (HRS). The semi-log I-V curves in Figure 3 d give a more accurate picture of NDR and RS with a switching ratio of 10 3 . Figure 4 presents the response of the device at different sweep rates. The device shows a stable response at higher sweep rates and performs well. The presence of trap charges in native oxide provides triggering of the next switching cycle at the same location, which improves the response. To understand the mechanism, we implemented and proved the amorphous model. The traps generated by localized states in amorphous CuO nanoparticles film had a strong influence on the injected current in response to applied voltage [ 27 ]. The interaction of the injected carriers in defect states affects the magnitude of current, which also affects the current–voltage characteristics. To prove the RS and NDR mechanism and investigate the role of traps, Log V vs. Log I and temperature-dependent I-V characteristics were plotted in Figure 5 a,b, respectively. The current was proportional to voltage ( IαV ) at a lower voltage. Hence, the space-charge-limited current (SCLC) was close to negligible, and Ohm’s law dominated the I-V characteristics [ 2 , 28 ]. The transport is explained by the presence of thermal equilibrium free carriers given by Equation (1): n = N [exp [( E − E c / kT )]] (1) Here, n is a free carrier, N are the effective density of states in the conduction band, E C is the energy at the bottom of the conduction band, k is Boltzmann’s constant, and T is the temperature. As the voltage increases further, the SCL current becomes obvious as seen in the slope via IαV (Equations (2) and (3)) [ 28 ]. Hence, we inferred that the conduction was dominated by the shallow trap SCL current. The shallow trap square law is given by:\n (2) J = 9 8 µ ε V 2 L 3 \nwhere V , L , μ and ε are the voltage, the distance between Top Electrode and Bottom Electrode, the free-charge mobility, and the dielectric constant, respectively. As the applied field further increases, the strong potential caused the effective depth of a coulombic-attractive trap to be reduced due to the Poole–Frenkel (PF) effect [ 14 , 28 ]. The reduction in depth reduces the barrier height resulting in an increase in the current level. This steep jump in current is higher than that predicted by the standard SCL current theory. The steep increase in the current is obvious in Figure 5 a. The shallow trap SCL current density equation incorporating the PF effect can be given as: (3) J = μ Θ ε ξ exp ( β √ ε k T ) d ε d x \nwhere Θ is the ratio of free to trapped charge concentration, ξ is an electric field and β is the Poole–Frenkel constant. Trap-induced space-charge-limited current was further proved by plotting temperature-dependent I-V in Figure 4 . The data suggested that the change in temperature led to a variation in the trap distribution. The voltage dependence of space-charge-limited current is given by\n (4) I α V ( T C / T ) + I \nwhere T c is a characteristic temperature describing the distribution of traps in energy. An increase in temperature does not alter the total amount of space charge but does increase the fraction of space charge in the conduction band. Equation (4) indicates that the temperature should be shifted in the current–voltage curve along with the voltage axis toward lower voltages [ 29 ]. Beyond room temperature, the switching and NDR effects were not observed over 25 °C. The tunneling current had a negative temperature coefficient, which proved the tunneling current is dominated by the conduction process in the negative region. Based on the mechanism, the RS can be described by the formation of an Ag conductive filament due to a trap control charge transfer enabling the abrupt switching from HRS to LRS [ 24 ]. Ag is an active metal, and the Ag atoms are possibly ionized for the Ag/CuO/SiO 2 /p-Si device. The Ag atoms can be ionized into Ag ions with the electric field. The electric field drives the Ag ions into defects level of CuO nanoparticles. When the Ag ions accumulate to a certain extent, the conductivity of the material will increase substantially because the Ag ions play the role of conductive filaments to finish the set process leading to the formation of conductive paths [ 21 ]. To further investigate the role of native oxide, p-Si, and Ag, we fabricated a device consisting of ITO as the bottom electrode. Figure 6 shows the I-V characteristics of the device with Ag as a top electrode and ITO as a bottom electrode with CuO nanoparticles acting as an active layer. In the absence of SiO 2 , the device showed resistive switching without NDR. The switching was described by the formation of Ag filament based on the linear I-V characteristics. The device did not show NDR in the absence of oxide. The abrupt uncontrollable switching also characterizes the absence of native oxide. The results showed no NDR in the case of Ag/CuO/ITO, and we inferred that NDR in Ag/CuO/SiO 2 /p-Si was mainly due to electric field-induced charge transfer between CuO and SiO 2 ; the high on/off ratio was due to the oxide barrier that contributes to an increase in the off-state resistance. Considering all the above results, we propose a model to describe the NDR and RS behaviors shown in Figure 7 a–e. The absence of NDR in Ag/CuO/ITO provided a piece of evidence that NDR was due to oxide-trapped charges associated with defects in SiO 2 . The device is initially HRS due to traps in the CuO nanoparticles and the Schottky barrier at the Ag/CuO/SiO 2 /Si interface. Initially, the Schottky barrier at the SiO 2 /p-Si interface reduces when a forward voltage sweep is applied, resulting in an injection of charges. These injected charges are then trapped in oxide interfacial traps [ 30 ]. Simultaneously, the electric field drives the Ag ions into defect levels of CuO nanoparticles. This initiates the forward conduction as shown in Figure 7 a. The filled interfacial traps reduce oxide resistance [ 31 ]. Further increases in the electric field accumulate Ag ions, which then play the role of conductive filaments. Therefore, the current increases sharply with further increases in the electric field due to CF; the filled traps are shown in Figure 7 b. The device remains in an LRS until a negative voltage is applied to rupture the CF. The trapped electrons are then released under bias ( Figure 7 c). With a further increase of the negative voltage, the charge injection from the Ag electrode is prevented by the Schottky barrier at the interface. The current decreases gradually upon increasing the sweeping voltage. This explains the NDR behavior ( Figure 7 d). The current decreases until the trapped electrons are released completely ( Figure 7 e)."
} | 3,755 |
31227022 | PMC6588946 | pmc | 438 | {
"abstract": "Background Coral reefs are facing unprecedented pressure on local and global scales. Sensitive and rapid markers for ecosystem stress are urgently needed to underpin effective management and restoration strategies. Although the fundamental contribution of microbes to the stability and functioning of coral reefs is widely recognised, it remains unclear how different reef microbiomes respond to environmental perturbations and whether microbiomes are sensitive enough to predict environmental anomalies that can lead to ecosystem stress. However, the lack of coral reef microbial baselines hinders our ability to study the link between shifts in microbiomes and ecosystem stress. In this study, we established a comprehensive microbial reference database for selected Great Barrier Reef sites to assess the diagnostic value of multiple free-living and host-associated reef microbiomes to infer the environmental state of coral reef ecosystems. Results A comprehensive microbial reference database, originating from multiple coral reef microbiomes (i.e. seawater, sediment, corals, sponges and macroalgae), was generated by 16S rRNA gene sequencing for 381 samples collected over the course of 16 months. By coupling this database to environmental parameters, we showed that the seawater microbiome has the greatest diagnostic value to infer shifts in the surrounding reef environment. In fact, 56% of the observed compositional variation in the microbiome was explained by environmental parameters, and temporal successions in the seawater microbiome were characterised by uniform community assembly patterns. Host-associated microbiomes, in contrast, were five-times less responsive to the environment and their community assembly patterns were generally less uniform. By applying a suite of indicator value and machine learning approaches, we further showed that seawater microbial community data provide an accurate prediction of temperature and eutrophication state (i.e. chlorophyll concentration and turbidity). Conclusion Our results reveal that free-living microbial communities have a high potential to infer environmental parameters due to their environmental sensitivity and predictability. This highlights the diagnostic value of microorganisms and illustrates how long-term coral reef monitoring initiatives could be enhanced by incorporating assessments of microbial communities in seawater. We therefore recommend timely integration of microbial sampling into current coral reef monitoring initiatives. Electronic supplementary material The online version of this article (10.1186/s40168-019-0705-7) contains supplementary material, which is available to authorized users.",
"conclusion": "Conclusion Our study provides the first holistic microbial baseline spanning multiple free-living and host-associated microbiomes for selected GBR sites. Results suggest that there is realistic scope to enhance long-term reef monitoring initiatives by incorporating seawater microbiome observations for assessments of environmental change over space and time, especially for rapid and sensitive identification of early signs of declining ecosystem health. The establishment of microbial observatories [ 65 ] and DNA biobanks for long-term biomonitoring [ 66 ] will be paramount to successfully inferring ecosystem state and/or perturbations from microbial communities. We therefore recommend timely integration of microbial sampling into current coral reef monitoring initiatives. Further refinement of the sampling and data analysis techniques should focus on selection and validation of additional indicator taxa as well as assessment of ecologically important microbial functions. A further consideration is to explore which monitoring objectives would benefit most from assessments of microbial communities. For example, it is likely that the rapid response time of microbial indicators makes them better suited to early-warning, impact or compliance monitoring programs than to monitoring of slower, long-term changes.",
"discussion": "Discussion Sensitive and rapidly responding markers of coral ecosystem stress are needed to underpin effective management and restoration strategies. In this study, we used a range of statistical tests and machine learning approaches across multiple free-living and host-associated reef microbiomes to assess their diagnostic value as sensitive indicators of environmental state. Our results show that the microbial community in reef seawater has the highest diagnostic value when compared to other free-living (e.g. sediment) and host-associated microbiomes (e.g. coral, sponge and macroalgae). Our conclusion is based on the microbiome’s (1) habitat-specificity, (2) uniformity of its community assembly, (3) sensitivity towards environmental fluctuations and (4) accuracy to predict environmental parameters. This assessment of the diagnostic capacity of various free-living and host-associated coral reef microbiomes to extrapolate environmental variations provides crucial information for ecosystem management initiatives aimed at incorporating microbial monitoring. In general, high habitat-specificity was observed across free-living and host-associated microbiomes, confirming previous reports on the compositional variability of microbial communities between coral reef habitats [ 40 ], host species [ 15 , 41 – 43 ] and even between host compartments [ 44 ]. High compositional divergence of microbial communities across different reef habitats can be due to the variation of available resources and/or biotic interactions [ 21 ]. High habitat-specificity contributes to the overall high diversity and complexity across different microbial communities on coral reefs, highlighting the importance of holistic studies that focus on microbial interactions across the benthic-pelagic realm. Bacterial community structure associated with water and sediment is thought to be primarily governed by deterministic processes [ 45 ]. Our results are consistent with this, showing uniform community assembly patterns within time point replicates. In contrast, host-associated microbiomes displayed little compositional similarity within a sampling time point, suggesting a non-uniform temporal response. Host-associated microbiomes were also only marginally affected by environmental parameters, indicating that their community assembly pattern is variable between conspecific individuals [ 45 ]. A higher variability in community assembly can lead to increased community heterogeneity, also referred to as dispersion, which has been described as a common characteristic of host-associated microbiomes [ 18 , 46 – 48 ]. Furthermore, lower microbial compositional similarities among replicates may be driven by increased niche space (e.g. host compartments) [ 44 ] and host genotype effects (e.g. host genetics) [ 42 ]. Collectively, our results show that free-living microbial communities have a higher potential to infer environmental parameters (such as standard measures in environmental monitoring programs) than host-associated microbial communities due to their higher uniformity and environmental sensitivity. Importantly however, previous metaproteomic research on reef sponges has shown that while microbial community composition can appear stable when seawater temperatures increase, disruption to nutritional interdependence and molecular interactions (such as reduced expression of transporters involved in the uptake of sugars, peptides and other substrates) actually occurs prior to detectable changes in community structure [ 49 ]. Hence, considering the importance of microbes to reef invertebrate health, more sensitive transcriptomic/proteomic approaches may still be warranted for sensitive detection of microbial responses to environmental perturbations. The diagnostic potential of microbial communities, especially in combination with machine learning approaches, has gained momentum across multiple research fields, including disease identification by characterisation of the human gut-microbiome [ 50 ], evaluation of the environment and host genetics on the human microbiome [ 51 ], prediction of hydrological functions in riverine ecosystems [ 52 ] and assessment of macroecological patterns in soil samples [ 53 ]. This development of microbial-based diagnostics is largely due to availability of high-throughput sequencing of the 16S rRNA gene and streamlined analytical pipelines that facilitate rapid assessment of microbial community composition [ 54 , 55 ]. In addition to its utility for inferring environmental fluctuations, the seawater microbiome possesses numerous characteristics desirable for environmental monitoring programs: (i) non-destructive collection and simple processing methods facilitate large-scale collections alongside existing programs that sample water quality measurements, (ii) high fractional contribution of abundant microbes minimises the impacts of sequencing biases (Additional file 1 : Figure S9) and (iii) sampling is conducive to future automated, high throughput analyses such as in-line flow cytometry on vessels and real-time DNA/RNA sequencing for community characterisation. Incorporation of seawater microbial community data into coral reef monitoring approaches should enhance our ability to describe environmental conditions and changes more holistically. For example, temperature fluctuations drive structural variations in seawater microbial communities [ 56 , 57 ], and elevated seawater temperatures on coral reefs are highly correlated with coral bleaching [ 1 , 58 ]. The inclusion of microbial community data alongside water quality parameters could therefore improve our ability to predict the likelihood of ecosystem stress. For instance, our sample sites, located in the central sector of the GBR, were not affected by the 2016 bleaching that primarily affected the northern sector [ 59 ]; however, they were impacted by the 2017 bleaching event [ 60 ]. In the months prior to bleaching (late December 2016 till March 2017), we observed two to four times higher relative abundances of high temperature indicator assemblages than when compared to the equivalent period at the beginning of 2016 (Fig. 4 a), where no bleaching was observed. Interestingly, high temperature indicator assemblages included putative coral pathogens (e.g. Vibrio ) and opportunistic bacteria (e.g. Rhodobacteraceae , Verrucomicrobia and Flavobacterium ). Coral pathogens, such as Vibrio corallilyticus , increase their efficiency and motility behaviours with rising seawater temperatures [ 61 – 63 ], and the higher abundance of these microbes may explain the increased prevalence of coral disease post bleaching [ 64 ]. Hence, microbial monitoring could help inform managers about impending disease outbreaks."
} | 2,697 |
21151704 | null | s2 | 439 | {
"abstract": "We report an approach to the fabrication of superhydrophobic thin films that is based on the 'reactive' layer-by-layer assembly of azlactone-containing polymer multilayers. We demonstrate that films fabricated from alternating layers of the azlactone functionalized polymer poly(2-vinyl-4,4-dimethylazlactone) (PVDMA) and poly(ethyleneimine) (PEI) exhibit micro- and nanoscale surface features that result in water contact angles in excess of 150º. Our results reveal that the formation of these surface features is (i) dependent upon film thickness (i.e., the number of layers of PEI and PVDMA deposited) and (ii) that it is influenced strongly by the presence (or absence) of cyclic azlactone-functionalized oligomers that can form upon storage of the 2-vinyl-4,4-dimethylazlactone (VDMA) used to synthesize PVDMA. For example, films fabricated using polymers synthesized in the presence of these oligomers exhibited rough, textured surfaces and superhydrophobic behavior (i.e., advancing contact angles in excess of 150º). In contrast, films fabricated from PVDMA polymerized in the absence of this oligomer (e.g., using freshly distilled monomer) were smooth and only moderately hydrophobic (i.e., advancing contact angles of ~75º). The addition of authentic, independently synthesized oligomer to samples of distilled VDMA at specified and controlled concentrations permitted reproducible fabrication of superhydrophobic thin films on the surfaces of a variety of different substrates. The surfaces of these films were demonstrated to be superhydrophobic immediately after fabrication, but they became hydrophilic after exposure to water for six days. Additional experiments demonstrated that it was possible to stabilize and prolong the superhydrophobic properties of these films (e.g., advancing contact angles in excess of 150° even after complete submersion in water for at least six weeks) by exploiting the reactivity of residual azlactones to functionalize the surfaces of the films using hydrophobic amines (e.g., aliphatic or semi-fluorinated aliphatic amines). Our results demonstrate a straightforward and substrate-independent approach to the design of superhydrophobic and reactive polymer-based coatings of potential use in a broad range of fundamental and applied contexts."
} | 573 |
36838420 | PMC9964548 | pmc | 440 | {
"abstract": "Cyanobacteria are photosynthetic microorganisms capable of using solar energy to convert CO 2 and H 2 O into O 2 and energy-rich organic compounds, thus enabling sustainable production of a wide range of bio-products. More and more strains of cyanobacteria are identified that show great promise as cell platforms for the generation of bioproducts. However, strain development is still required to optimize their biosynthesis and increase titers for industrial applications. This review describes the most well-known, newest and most promising strains available to the community and gives an overview of current cyanobacterial biotechnology and the latest innovative strategies used for engineering cyanobacteria. We summarize advanced synthetic biology tools for modulating gene expression and their use in metabolic pathway engineering to increase the production of value-added compounds, such as terpenoids, fatty acids and sugars, to provide a go-to source for scientists starting research in cyanobacterial metabolic engineering.",
"conclusion": "6. Conclusions This review covered the current state of cyanobacterial biotechnology, including cyanobacteria with industrially relevant traits, the synthetic biology tools available to control gene expression and a description of several metabolic engineering approaches applied for photosynthetic bioproduction. We anticipate that more synthetic biology tools for the efficient development of cyanobacterial cell factories will be rapidly developed. Combined with large-scale efforts to generate genetic libraries (overexpression and deletion mutants), such as those available for E. coli , to improve understanding of cyanobacterial metabolism and physiology and to develop advanced photobioreactors, these capabilities will enable the full potential of cyanobacteria and sustainable and photosynthetic bioproduction to be realized.",
"introduction": "1. Introduction Cyanobacteria are photosynthetic unicellular microorganisms with powerful biotechnological features. They are Gram-negative prokaryotes that belong to the bacterial domain and are considered one of the oldest and largest groups of bacteria on Earth. The oldest fossil dates back to the Archean era. Cyanobacteria were essential for forming the biosphere, creating oxygenating conditions by releasing O 2 into the atmosphere [ 1 ]. They are also able to fix nitrogen, occupying a prominent role in the nitrogen cycle [ 2 ]. Given their long history, their ability to adapt to environmental changes on Earth is one of their principal characteristics. For instance, they have differentiated specialized nitrogen-fixing cell types, facilitating the dispersion of species [ 3 ] and are found in marine, freshwater and terrestrial environments. Cyanobacteria possess a typical prokaryotic cellular organization; however, they lack the cell wall usually found in bacteria. Plant chloroplasts are thought to be derived from endosymbiotic cyanobacteria [ 4 ], explaining the similarity of their photosynthetic apparatus embedded in the thylakoid membranes [ 5 ]. Despite being a relatively ‘young’ organism from a biotechnological point of view (compared to the well-characterized industrial chassis Escherichia coli and Saccharomyces cerevisiae ), cyanobacteria’s capability to use solar energy for generating reducing power and energy along with their prokaryotic cellular organization make them attractive biotechnological agents to produce valuable compounds. Cyanobacteria convert inorganic carbon dioxide (CO 2 ) and H 2 O into biomass and valuable products and some species can also fix molecular nitrogen. They have a clear advantage compared to the microorganisms, such as E. coli and S. cerevisiae , which are currently the preferred cell platforms in industrial biotechnology, but rely on reduced carbon and nitrogen sources, typically sugars and ammonia, increasing the production costs of the target compounds [ 6 , 7 ]. Their photosynthetic biomass production rate is also higher than that of plants [ 8 , 9 , 10 ]. Moreover, genetic modifications of cyanobacteria are faster and more efficient than in plants or algae [ 8 , 9 ]. Many molecules of commercial interest are natively produced by cyanobacteria, for instance: terpenoids, chlorophylls, fatty acids, sugar and amino acids [ 11 , 12 ]. Together, these traits make cyanobacteria ideal candidates for sustainable, low-cost biological production of high-value chemicals. Various strategies have been implemented to optimize cyanobacteria for bioproduction, including flux enhancement through a determined pathway, removal of competitive pathways, or augmenting cell fitness against high product concentrations [ 13 , 14 , 15 , 16 ]. Most of the metabolic engineering approaches have led to scarce results in terms of productivity and titer in comparison to those achieved in heterotrophic microbes [ 10 ]. The lack of knowledge about cyanobacterial regulatory mechanisms, a gap in available genetic tools and the inconsistency in performance of characterized genomic parts across different cyanobacterial strains can explain this discrepancy. The increased number of cyanobacterial genome sequencing data (3858 cyanobacterial genome assemblies available in GenBank [ 17 ]) has greatly facilitated the integration of transcriptomics, proteomics and metabolomics studies and helped raise awareness beyond model organisms and identify new industrial relevant species [ 18 ]. The evolution of the synthetic biology paradigm as the systematic reconstitution of new standardized biological parts, modules and devices to produce a particular cellular output has further spurred the availability of well-characterized genetic parts (e.g., promoters, ribosome binding sites (RBS) and coding sequences) and improved methods for cyanobacteria metabolic engineering. This review will introduce the principal cyanobacterial model organisms and lesser-known species with relevant biotechnological traits and applications. Secondly, we will focus on available genetic engineering and synthetic biology tools. Finally, we present the main achievements in terms of metabolic engineering of cyanobacteria, taking a closer look at the production of terpenoids, fatty acids and carbohydrates. We will also introduce several new engineering approaches still majorly unexplored in cyanobacteria and highlight some challenges and prospects of cyanobacterial cell factory development."
} | 1,604 |
37566650 | PMC10421036 | pmc | 441 | {
"abstract": "Climate change–amplified marine heatwaves can drive extensive mortality in foundation species. However, a paucity of longitudinal genomic datasets has impeded understanding of how these rapid selection events alter cryptic genetic structure. Heatwave impacts may be exacerbated in species that engage in obligate symbioses, where the genetics of multiple coevolving taxa may be affected. Here, we tracked the symbiotic associations of reef-building corals for 6 years through a prolonged heatwave, including known survivorship for 79 of 315 colonies. Coral genetics strongly predicted survival of the ubiquitous coral, Porites (massive growth form), with variable survival (15 to 61%) across three morphologically indistinguishable—but genetically distinct—lineages. The heatwave also disrupted strong associations between these coral lineages and their algal symbionts (family Symbiodiniaceae), with symbiotic turnover in some colonies, resulting in reduced specificity across lineages. These results highlight how heatwaves can threaten cryptic genotypes and decouple otherwise tightly coevolved relationships between hosts and symbionts.",
"introduction": "INTRODUCTION Extreme climatic events, such as heatwaves, wildfires, floods, and droughts, have become substantial catalysts for ecological change, jeopardizing biodiversity and natural ecosystems worldwide ( 1 – 3 ). Such events are driving factors behind many species range contractions, extirpations, and invasions ( 4 – 8 ) and may also influence the genetic makeup of taxa through selection imposed by rapid environmental change ( 9 – 11 ). Recent studies have demonstrated that selection through extreme events can be directional, favoring individual genotypes or lineages that carry adaptive traits ( 10 , 12 – 14 ). While this selection may drive adaptation and help taxa persist through similar types of future events, reductions in the diversity of genotypes or lineages may limit the capacity for future adaptation to disturbances of a different nature ( 15 , 16 ), such as unexpected stressors and pathogens ( 17 ). In the ocean, some of the most profound impacts of climate change are experienced during marine heatwaves ( 5 , 18 )—pulse heat stress events in which water temperatures are abnormally high for unusual lengths of time ( 19 , 20 ). While it is established that heatwaves can threaten marine biodiversity by driving conspicuous species losses ( 4 , 5 , 21 ), selection imposed during heatwaves may also drive cryptic losses of diversity within taxa, such as the loss of morphologically indistinguishable genotypes or cryptic species that remain challenging to characterize ( 9 , 10 , 12 ). This phenomenon is predicted to be an outcome of extreme climatic events but, due to the lack of baseline genetic data, has only rarely been demonstrated in marine systems ( 10 , 12 , 22 , 23 ). Losses of cryptic diversity could have long-term evolutionary consequences for taxa by limiting their scope for future adaptation ( 10 , 12 , 24 ). For example, introgression between differentiated lineages or cryptic species can be a critical mechanism behind rapid adaptation ( 25 , 26 ), a process that would be hindered by the loss of cryptic diversity. Our ability to anticipate the true consequences of marine heatwaves will therefore depend on understanding the extent to which marine heatwaves drive differential mortality among cryptic genotypes, lineages, or species, thereby altering the genetic makeup of taxa ( 10 ). To date, the few studies that have investigated this phenomenon have demonstrated that marine heatwaves can alter the relative abundance of genotypes ( 12 , 22 ), but linking these patterns to fitness components, which requires tracking survival and/or reproductive output of individuals, remains a challenge [but see ( 23 )]. Under climate change, tropical reef-building corals may be particularly susceptible to losses of cryptic diversity, both because cryptic species complexes are exceptionally common in this group [e.g., ( 27 – 30 )] and because corals are highly sensitive to thermal stress ( 18 ). Heatwaves disrupt the critical relationship between reef-building corals and their obligate endosymbionts (family Symbiodiniaceae), causing them to bleach ( 18 , 31 , 32 ) and making them vulnerable to starvation and disease ( 33 ). Intense or prolonged marine heatwaves can cause mass bleaching and widespread coral mortality, with profound ecological and socioeconomic impacts ( 18 ). This is especially true given that these ecosystems are the most biologically diverse and among the most economically valuable in the ocean ( 34 ). As in a wide range of taxa, molecular investigations of reef-building corals over the past two or more decades have drastically reshaped our understanding of their evolution and diversity [e.g., ( 27 , 32 , 35 )]. Similar to macroalgae and other invertebrates [e.g., ( 36 , 37 )], many morphologically defined coral species actually represent cryptic species complexes consisting of multiple morphologically similar, or even indistinguishable, lineages that are partially or completely reproductively isolated from one another [e.g., ( 27 – 30 )]. Moreover, constituent cryptic species or lineages often show evidence of introgression, suggesting that they may have the potential to exchange adaptive genomic diversity ( 27 , 38 ). Although coral cryptic lineage complexes are common, the number of studies testing for differences in heat tolerance between cryptic lineages is limited ( 39 , 40 ). Moreover, heatwave studies of cryptic coral lineages tend to assess their bleaching tolerance [e.g., ( 39 )], rather than their survival [but see ( 23 )], despite the fact that these two processes can be decoupled ( 41 ). Consequently, climate change may already be threatening the persistence and coexistence of cryptic lineages, but the lack of genotype-specific time series data makes this problem challenging to detect. Quantifying differential mortality across cryptic coral genotypes, lineages, or species will therefore be essential to understanding and predicting the influence of future climate change on the diversity and adaptive potential of threatened coral reefs. Although a handful of coral studies have tracked the stability of symbioses through marine heatwaves and shown that differential bleaching can strongly depend on algal symbionts ( 11 , 39 , 41 ), to our knowledge, only one has directly assessed the impacts of these events on the population genetics of coral taxa in natural systems ( 23 ). Moreover, despite growing awareness that cryptic coral lineages can harbor unique assemblages of symbionts, which could be the primary determinants of their climate change vulnerability (or resilience) [e.g., ( 39 , 42 )], no study to date has simultaneously tested for shifts in both host population genetics and associated symbiont assemblages through an extreme heatwave event. Thus, while experimental work points to important functional differences between cryptic coral lineages ( 40 , 43 , 44 ), the extent to which heatwaves may be threatening particular cryptic lineages in nature remains unclear. Furthermore, given tight associations between some cryptic lineages and their symbiotic partners ( 29 , 42 ), it remains an open question how marine heatwaves alter symbiont specificity across co-occurring cryptic coral lineages, and if they potentially threaten rare or heat-sensitive lineages of either symbiotic partner ( 11 , 45 ). Between 2014 and 2017, a series of heatwaves unfolded across much of the world’s tropical reefs ( 18 , 46 ). This period, considered chronologically as the third global coral bleaching event on record, was unprecedented in terms of the severity, duration, and geographic spread ( 46 ). This event led to mass coral bleaching and mortality across many coral reefs in the Pacific and Indian Oceans, including extensive damage to the Great Barrier Reef ( 18 , 41 , 47 , 48 ). Species-level assessments of coral mortality have demonstrated that there were winners and losers in the face of this widespread bleaching, with survival varying substantially across coral taxa ( 47 , 49 ). However, it is not known whether differential survival during this global bleaching event affected the genetic composition of coral species or species complexes, nor how underlying local anthropogenic stressors—a feature of virtually all coral reefs—might modulate impacts on this critical facet of diversity. Here, we directly assessed the extent to which marine heatwaves drive differential mortality across cryptic coral lineages and alter the specificity of host-symbiont pairings. We focused on one of the most widespread, ecologically important, and well-studied coral genera, Porites, and tracked the fate and algal symbiont composition of individual Porites colonies (massive growth form; identified in the field as Porites lobata ) in the central equatorial Pacific Ocean, through the third global coral bleaching event. Within this region, the coral atoll Kiritimati experienced some of the highest levels of accumulated heat stress ever documented on a coral reef, rivaled only by nearby Jarvis Island during this same time period ( 49 ). This heatwave lasted 10 months (June 2015 to April 2016), imposing up to ~31.6 degree heating weeks (°C-weeks) on Kiritimati’s coral reefs ( 41 ). Despite this, massive Porites had relatively high survivorship (~80% at some sites), with highly variable bleaching severity and survival among colonies and sites ( 49 ). We leveraged this extreme climatic event as a natural experiment to directly test whether coral bleaching susceptibility or survivorship could be predicted by the genetic identity or lineage of the affected colonies and/or their associated algal symbionts. To accomplish this, we analyzed genetic data on colonies at two levels: tracking individual colonies ( n = 79) from before to after the heatwave to directly compare survival, and analyzing a broader population-level sampling of massive Porites colonies before, during, and after the heatwave ( n = 315 colonies total). Recent molecular studies have determined that the genus Porites comprises at least eight clades, some characterized by complex genetic structure ( 28 , 50 ), possibly reflecting cryptic or pseudo-cryptic lineages within each clade. Our study focused on one of these clades [clade V from ( 50 ); also known as the P. lobata / lutea clade] facilitating a deeper look into the functional differences between finer-scale cryptic lineages than has previously been achieved in this group. Our objectives were to examine (i) if cryptic coral lineages were present across Kiritimati, (ii) if survivorship of coral colonies varied across cryptic lineages or by their underlying exposure to chronic local human disturbance, (iii) if cryptic coral lineages were associated with specific symbionts and if there is evidence of host-symbiont coevolution (i.e., cophylogeny), and (iv) whether the specificity of these symbiotic partnerships was affected by mass bleaching and mortality during the heatwave.",
"discussion": "RESULTS AND DISCUSSION Sympatric cryptic lineages of Porites We identified three genetic lineages of massive Porites (hereafter referred to as PKir-1, PKir-2, and PKir-3) that were found sympatrically across the reefs of Kiritimati before the 2015–2016 El Niño–driven heatwave ( Fig. 1 ). Ordination [based on >12,000 single-nucleotide polymorphisms (SNPs) from 2b-RAD] revealed three distinct genomic clusters with no intermediate genotypes, which was further supported by ADMIXTURE analyses showing the lowest cross-validation (CV) error for k = 3, where every sample was assigned to a lineage with >85% probability ( Fig. 1, A and B ). Global F ST values between lineages were also high, suggesting relatively high levels of differentiation across cryptic lineages (table S1). PKir-1 and PKir-2 (global F ST = 0.263) were found to be more genetically similar to each other than either was to PKir-3 (global F ST = 0.361 and 0.326, respectively; fig. S1). However, evidence for historical gene flow was detected between all lineages (fig. S2), suggesting that, although these lineages appear reproductively isolated in the present day, they have likely experienced introgression in the past. Fig. 1. Cryptic lineages of massive Porites across forereef sites on Kiritimati. ( A ) Principal coordinate analysis (PCoA) of 2b-RAD data (using 1-Pearson correlation matrices through ANGSD) showing three population clusters. ( B ) Results of ADMIXTURE analysis showing the assignment of colonies to one of three lineages, arranged by collection site. ( C ) Map with pie charts showing the relative abundance of each lineage at each site before the heatwave. Numbers indicate the number of colonies sampled and sequenced with either 2b-RAD or ITS2 metabarcoding. Circles labelled with site names represent sites colored by level of human disturbance. Semi-transparent red circles indicate the location of villages, scaled by human population size. Demographic analyses using Moments ( 51 ) based on allele frequency spectra (AFS) infer some limited gene flow between the three lineages with asymmetrical introgression across lineages and regions of the genome. The best-fit model supported the hypothesis of heterogeneous gene flow across the genome, with a small proportion of the genome experiencing particularly high gene flow (fig. S2). Moreover, we inferred higher gene flow from PKir-3 to PKir-1 and PKir-2 compared to the reverse direction (fig. S2). Analysis of one-dimensional AFS with StairwayPlot ( 52 ) revealed that effective population sizes ( N e ) were similar across all three lineages, and all showed contractions in recent millennia (fig. S3). Leveraging host sequences in the ITS2 metabarcoding dataset, we were able to expand lineage assignment beyond the n = 67 colonies that were sequenced with 2b-RAD to n = 300 colonies, including all n = 161 colonies that were sampled before the heatwave (table S2). As expected, all Porites ITS2 sequences belonged to the same previously described Porites clade, clade V—the P. lobata / lutea clade ( 50 ) (fig. S4). However, examining colonies that were sequenced using both ITS2 and 2b-RAD ( n = 67), we found that host ITS2 sequences were consistently dissimilar across the cryptic Porites lineages such that particular ITS2 barcode sequences could be used to assign host lineages to those colonies not sequenced with 2b-RAD (see Materials and Methods). Using all samples collected before the heatwave for which lineage assignment was possible ( n = 157 of 161), we found that, although the relative abundance of each Porites lineage varied across the atoll before the heatwave ( Fig. 1C ) (Fisher’s exact test: P < 0.001), with significant differences between the southeast side of the atoll and other regions (fig. S5), lineage distribution was not driven by local human disturbance (multinomial regression: χ 2 = 1.5373, P = 0.4636). We note that 1 colony out of the 79 with known survivorship (see below) could not be assigned to a lineage due to ambiguous ITS2 sequences, indicating that it was from either PKir-1 or PKir-2. Survivorship, cryptic host lineage, and human disturbance Tracking individual colonies through nine time points that span the 2015–2016 heatwave, we found strong evidence of differential survival among lineages—an effect that was more pronounced at minimally disturbed sites ( Fig. 2 ). Mortality of tagged colonies began during the heatwave (first observed in March 2016) and continued for at least a year following the heatwave, as colonies that had mostly died during the event finally experienced complete mortality in the months following the heatwave. We measured mortality (up to and including 2017) in two ways. First, we quantified colony mortality as a binary variable (died or survived, regardless of how much colony tissue remained) to determine whether colonies were eliminated from the population or remained to potentially grow and reproduce. Second, we examined variation in tissue loss across colonies, comparing the percentage of each colony that died from the heatwave (see Materials and Methods). We found that for both metrics, mortality varied significantly among lineages and levels of anthropogenic disturbance. Total mortality was greatest for PKir-3 (17 of 20; 85%) compared to the other two lineages (PKir-1: 16 of 30, 53%; PKir-2: 11 of 28, 39%) ( Table 1 ; binomial regression—lineage: χ 2 = 12.2437, P = 0.0022) and varied significantly with human disturbance (binomial regression—disturbance: χ 2 = 16.1409, P < 0.001). This effect was driven by PKir-1 and PKir-2; while mortality was high across all values of human disturbance for PKir-3, total mortality sharply increased from 17 to 33% at minimally disturbed sites to 88 to 100% in PKir-1 and PKir-2 at sites exposed to very high human disturbance. Assessing tissue loss, we found that cryptic lineage and human disturbance were also significant predictors of this metric (quasibinomial logistic regression—lineage: χ 2 = 12.7531, P = 0.0017, Disturbance: χ 2 = 8.7798, P = 0.0030; Fig. 2B ). For both metrics of survival, there were no significant interactions between lineage and human disturbance (binomial regression—lineage*Disturbance: χ 2 = 1.8726, P = 0.3921; quasibinomial regression—lineage*Disturbance: χ 2 = 0.8086, P = 0.6674), likely due to uncertainty in the effect of human disturbance on PKir-3. Fig. 2. Survivorship by coral cryptic lineage and chronic human disturbance. The probability that coral colonies experienced complete mortality between 2015 and 2017 ( A ) and the average percentage of partial mortality (i.e., tissue loss) that each colony experienced ( B ) both in relation to human disturbance and Porites lineage are shown. In (A), each point represents a colony that either survived (0) or died (1). The proportion of colonies that died at each value is estimated by the logistic regression line (shown with 95% confidence intervals). In (B), proportion mortality indicates the area of tissue loss experienced by each colony from 2015 to 2017, while each line shows the model fit with human disturbance for each lineage. Note that human disturbance is a relative metric based on fishing pressure and distance to Kiritimati’s villages [see ( 87 )]. Note that data points are jittered for visualization. Table 1. Survival and relative abundance of cryptic coral lineages before and after the 2015–2016 marine heatwave. Numbers indicate the number (and percentage) of colonies in each category. Lineage Tagged colonies Population-level sampling Survived Died Relative abundance before Relative abundance after Change in relative abundance PKir-1 14 (47%) 16 (53%) 61/161 (38%) 84/173 (49%) +29% PKir-2 17 (61%) 11 (39%) 50/161 (31%) 64/173 (37%) +19% PKir-3 3 (15%) 17 (85%) 44/161 (27%) 16/173 (9%) −77% Unassigned 1 0 6/161 (4%) 9/173 (5%) \n We also found a significant effect of cryptic lineage identity on bleaching score using two different methods of categorizing bleaching (see Materials and Methods for description) such that PKir-3 tended to have the highest level of bleaching at both time points during the heatwave (2015—method 1: deviance = 7.8441, P = 0.0198; 2015—method 2: deviance = 11.303, P = 0.0035; 2016—method 1: deviance = 11.409, P = 0.0033; 2016—method 2: deviance = 13.155, P = 0.0014; fig. S6). There was no significant effect of colony size (area; cm 2 ) on survival through the heatwave, and lineages themselves did not differ in the size of colonies tracked through the heatwave (fig. S7). However, there was a trend toward a possible interaction between lineage and size [quasibinomial generalized linear model (GLM)—size*lineage: χ 2 = 5.8273, P = 0.0543]. This was driven by increased survival (but not decreased tissue loss) in larger colonies of PKir-1 and PKir-2, while PKir-3 had similarly high mortality across all colony sizes (fig. S7). To explore the hypothesis that genomic differences may explain differences in heat tolerance, we tested for local genomic differentiation across the three Porites lineages and identified genes near outlier loci. While we found numerous genes near outlier loci when comparing lineage pairs (PKir-1 versus PKir-2: n = 47; PKir-1 versus PKir-3: n = 63; PKir-2 versus PKir-3: n = 42; file S1), the only gene near an outlier locus when comparing both PKir-1 and PKir-2 to PKir-3 matched the ETS-related transcription factor Elf-2 (~57% similarity) with possible links to coral immunity ( 53 ), suggesting that genome-level functional differences may exist between cryptic Porites lineages. Disruption of lineage-specific symbioses We found strong associations between coral lineage and symbiont assemblage composition before the marine heatwave. Across all colonies sampled before the heatwave (i.e., population-level sampling), there was a strong relationship between coral lineage and recovered C15-type Cladocopium sequence variants—i.e., variants of the Cladocopium C15 lineage [permutational multivariate analysis of variance (PERMANOVA): F = 175.41, R 2 = 0.73, P < 0.001]. Specifically, Cladocopium sequences formed two distinct clusters ( Fig. 3, A and C ), with variants in one cluster associating almost exclusively with PKir-3 colonies (all but one case, ~2.5%, although two PKir-3 colonies, 5%, also had symbiont sequences from the other cluster). This pattern was consistent across both maximum entropy decomposition (MED; Fig. 3A ) and amplicon sequence variant (ASV; Fig. 3C ) methods (see Materials and Methods), highlighting the robustness of these patterns to multiple analytical approaches. According to ITS2 profile data output from SymPortal ( 54 ), which attempts to characterize putative Symbiodiniaceae taxa (by accounting for the multicopy nature of the ITS2 locus), most of the coral samples taken at any time point (554 of 674 samples; 82%) were each characterized by a single C15-type Cladocopium profile (i.e., a profile from the C15 lineage), although profile identity was variable across colonies. Most of the remaining colonies (82 of 674 samples; ~12%) had mixed assemblages that included one C15-type profile and one or two additional profiles from other Symbiodiniaceae lineages (e.g., Cladocopium C116, C1, C3, Durusdinium D1, D4). A single sample had four associated profiles (including one C15-type). In total, we identified 47 C15-type profiles (110 profiles from all Symbiodiniaceae lineages) across all colonies successfully sequenced. According to SymPortal, no corals were associated with more than one C15-type profile at a given time point and only ~3% (19 of 674) of samples lacked a C15-type profile altogether (mostly bleached colonies from March 2016; see below). Unlike PKir-1 and PKir-2, which initially associated with 9 and 13 profiles, respectively, PKir-3 had highly specific symbiotic associations before the heatwave, with 95% of colonies (38 of 40) associated with one of just three C15-type symbiont profiles that were absent or rare in other lineages (PKir-1: 0%, PKir-2: 3%). Moreover, all three of these ITS2 profiles had the same dominant sequence [as identified using MED methods; referred to by SymPortal as “defining intragenomic variants (DIVs)], suggesting that they reflect closely related symbiont populations. Specifically, PKir-3 tended to be associated with profiles that are dominated by the “ C15cu ” DIV, while the other two lineages tended to be associated with the “ C15 ” DIV. Fig. 3. Impact of the marine heatwave on lineage-specific symbioses. The results of PCoA based on unifrac dissimilarity of Cladocopium C15 maximum entropy decomposition (MED) sequences for all colonies sampled ( A ) before and ( B ) after the heatwave are shown. PCoAs based on Bray-Curtis dissimilarity of amplicon sequence variants (ASVs) from Cladocopium for each colony sampled ( C ) before and ( D ) after the heatwave are also shown. Ellipses at the 95% level are shown for each assigned coral lineage. Only samples with >500 sequence reads were included. Note that the ellipse for PKir-3 in (A) largely overlaps the points and is therefore challenging to visualize. Leveraging the colonies that were sequenced using both 2b-RAD (host genomics) and ITS2 metabarcoding (symbiont characterization), we tested for evidence of cophylogeny between symbiont and host lineages. There was a significant phylogenetic signal on the abundance of several Cladocopium DIV sequences (Moran’s I : 22 of 31, 71%; Blomberg’s K : 10 of 31, 32%; table S3) and ITS2 profiles (Moran’s I : 3 of 3, 100%; Blomberg’s K : 1 of 3, 33% out of three profiles found in >10% of these samples; table S4), highlighting the specificity of symbiont sequences to particular host lineages. We also performed procrustes analysis of cophylogeny (PACo) on data from these same colonies to test for cophylogeny between Porites and C15-type ITS2 profiles. Given the multicopy nature of the ITS2 locus, we used unifrac dissimilarities as a proxy for phylogenetic distance to produce a tree of C15-type profiles. Using a goodness-of-fit test, we found strong support for cophylogeny between massive Porites and their C15-type symbionts ( m 2 XY = 2.106617, P < 0.001, n = 10,000; fig. S8). Together, these analyses support the hypothesis that cryptic lineages had coevolved, to some extent, with their symbiotic partners before the heatwave. These tight relationships between Porites lineages and their Cladocopium symbionts were disrupted during the heatwave. Population-level sampling of colonies from after the heatwave indicated that the potentially coevolutionary associations between symbiont and host lineages were absent after the heatwave, with both MED and ASV sequences from all lineages forming a single cluster ( Fig. 3, B and D ). Following the heatwave, only one colony (whose lineage was unknown due to a lack of host sequence reads) was found to still have sequence variants common in PKir-3 before the heatwave. These patterns were also captured by analysis of ITS2 profiles (figs. S9 to S13), where C15cu ITS2 profiles (which were tightly associated with PKir-3 before the heatwave) were completely absent from PKir-3 colonies sampled after the heatwave ( n = 12; Fig. 4 ). Symbiont assemblages across PKir-2 colonies also became more homogeneous, in general, with decreases in the relative abundance of colonies associated with profiles dominated by “ C15m ” and “ C15 / C15cs ” DIVs in favor of colonies associated with profiles dominated by the C15 DIV ( Fig. 5 and figs. S9 to S13). This pattern is similar to a recent study on Acropora spp. showing losses in host-associated Symbiodiniaceae diversity at the reef scale during the same global bleaching event ( 55 ). Fig. 4. The relative abundances of ITS2 profiles across all colonies of each lineage before and after the heatwave. Sample sizes indicate the number of colonies. Only samples with >500 sequence reads were included. Fig. 5. Temporal stability of Symbiodiniaceae associated with tracked colonies of each cryptic Porites lineage. Each row represents an individual colony, with color at each time point indicating the most abundant symbiont profile. Colonies within each lineage are arranged in order of human disturbance (lowest on top, highest on bottom). Note that colonies that survived to 2017 were considered alive for survivorship analyses. Colonies that were considered to have switched between C15-type profiles (i.e., between ITS2 profiles with different dominant DIVs; n = 4) are indicated with an asterisk. Expeditions during the heatwave are shown in red text. Approximately half of the colonies that were tracked for two or more time points (~51%; 80 of 158) had the same C15-type profile in every case; the remaining colonies hosted variable symbiont profiles over time (see Fig. 5 and fig. S9). Most notably, a few colonies sampled both before and after the heatwave appeared to recover from bleaching (observed in November 2016 or later) with a different C15-type profile. For example, one of the three surviving colonies of PKir-3 was associated with a C15cu profile (i.e., dominated by the C15cu sequence variant) before the heatwave but was later associated with a C15 profile (dominated by the C15 sequence variant) after the heatwave; the other two PKir-3 colonies that survived were not successfully sequenced with ITS2 . Similarly, three PKir-2 colonies were associated with C15m profiles before bleaching (which were previously only found in that Porites lineage, PKir-2) and recovered from bleaching with C15 profiles. Notably, the C15m profiles were not found to dominate any colonies after the heatwave, despite their prevalence in PKir-2 before the heatwave. In all four colonies in which there were notable changes in C15-type profile identity, the initial symbiont profiles were completely absent in the colonies following their symbiotic shift. Moreover, all four colonies experienced turnover in the composition of DIVs, with some of the most common DIVs before the heatwave found to be absent following (fig. S14). Thus, shifts between ITS2 profiles with different dominant DIVs (C15-type) appeared consistent with symbiont switching [i.e., symbiotic replacement by environmentally acquired symbionts ( 31 , 56 , 57 )]. However, given that symbiont populations can be present in coral tissues at levels undetectable by metabarcoding methods ( 58 ) or may be present in different parts of the colony than were sampled ( 59 ), we cannot definitely differentiate between symbiont switching and shuffling [i.e., changes in relative abundance of two co-occurring symbionts ( 56 )]. Some additional colonies (62 of 158) also appeared to shift between similar profiles (often sharing their most common DIVs but having additional minor sequence variants; e.g., C15-C15ct-C15ch-C15cp-C15hj versus C15-C15ct-C15ch-C15cp-C15cq ). The overall similarity of these latter profiles (fig. S15) suggests that they may represent closely related members of the same symbiont population that may have even been assigned as different profiles due to sequencing artifacts (e.g., missing a rare sequence variant) rather than true biological differences. Alternatively, they may represent multiple co-occurring C15-type symbionts where the most common population is shared across time points. Because of this uncertainty in interpreting differences between similar symbiont ITS2 profiles, we only consider symbiont shifts between C15-type profiles when they differ in their most common DIV. Late in the heatwave, several of the bleached colonies of all lineages were associated with Cladocopium C1- and C3-type ITS2 profiles that were only ever present during that time point (March/April 2016; ~10 months into the heat stress; Fig. 5 ). In all cases, these colonies were severely bleached when sampled and all three surviving colonies that were sampled at later dates had recovered with symbionts characterized by C15-type sequence variants (see Fig. 5 ). Thus, due to the temporary nature of these three C1 and C3 profiles (1— C1 / C3-C1c-C1b-C42.2-C1bh-C1br ; 2— C1-C1c-C1al ; 3 —C3-C1bp-C3dg-C3-df-C3-dh ) on Kiritimati, we interpret these associations as transient Symbiodiniaceae infections, although we note that C1 and C3 symbionts have been found to stably associate with Porites spp. in other parts of the world [e.g., ( 60 )]. So, while it remains unclear whether these C1- and C3-profile symbionts offered any benefit to bleached hosts, it is possible that they helped colonies maintain some basic nutritional requirements during the period between initial bleaching and subsequent recovery of C15-type symbionts. A similar pattern was observed during bleaching of Pocillopora spp. in the eastern Pacific, where bleached colonies were temporarily colonized by a presumed opportunistic Breviolum population ( 11 ). The functional and ecological importance of these short-lived symbioses remains unclear but offers an interesting avenue for future research. In our study, other symbiont types (e.g., C116 or Durusdinium ) were generally rare and found inconsistently across samples, suggesting that these symbionts represent minor or opportunistic constituents of the Porites holobiont; they were not further examined here. However, we note that these rare profiles, although appearing only transient in massive Porites on Kiritimati reefs, may still be of functional importance to the Porites holobiont. Early studies assumed that massive Porites only acquired their symbionts vertically and therefore showed fixed symbiont dominance and high specificity ( 61 ). However, multiple studies have now shown that a single massive Porites colony can harbor mixed Cladocopium and Durusdinium communities ( 62 , 63 ) as well as different profiles from the C15 lineage ( 59 ). Thus, although vertical transmission occurs in Porites (and may be the dominant mode of transmission), these corals likely also have the ability to acquire new Symbiodiniaceae via horizontal transmission and/or “shuffle” dominant symbionts ( 64 , 65 ). Our data suggest that Porites can potentially associate with symbionts that have different ITS2 profiles following extreme bleaching and can temporarily associate with C1- and C3 -type symbionts. The ability to switch or shuffle symbionts is presumably adaptive by allowing corals to avoid evolutionary “dead-ends,” whereby a vertically transmitted symbiont is fixed across the host population, but may be maladaptive under future warming ( 31 , 57 ). Laboratory experiments on Montipora, which also transmits its symbionts vertically, have shown that changes in symbiont assemblages acquired in one generation can be transferred to the next ( 66 ), providing an avenue for intergenerational plasticity in coral holobiont function ( 67 ). This suggests that the loss of variation in symbiont identity across colonies may persist for generations. High degrees of vertical transmission and heritability in symbiont genotypes can lead to strong patterns of phylosymbiosis and/or cophylogeny ( 68 – 70 ), where closely related corals share similar algal symbiont communities, a pattern clearly reflected across host lineages in our dataset (fig. S16 and tables S3 and S4). Differences in symbiotic assemblages between cryptic lineages have now been documented across multiple coral genera ( 29 , 39 , 42 ). Under strong selection from the heatwave, however, this pattern of co-occurrence between coral host lineage and algal symbiont sequence variants was disrupted in our study. The erosion of phylosymbiosis across coral lineages is likely driven largely by changing symbiont associations that occurred during recovery from bleaching. Among the tracked colonies, four (three from PKir-2 and one from PKir-3) that bleached and recovered post-heatwave did so with symbionts characterized by profiles that had different dominant DIVs than those they hosted before bleaching (all C15-type; Fig. 5 ). However, given that profiles dominated by the C15 DIV were associated with ~5% of PKir-3 colonies before the heatwave (based on population-level sampling), differential mortality across colonies that had differing profiles initially (i.e., before the heatwave) may have also played a role in driving observed shifts in the population-level symbiont assemblages of each lineage. Consequently, although symbiont acquisition appears the most parsimonious explanation, the specific mechanism accounting for these symbiont changes cannot be unambiguously identified. Although we lack the statistical power to definitively tease apart the role of symbiont from that of host lineage (due to high overlap among tracked colonies), the breakdown of host-symbiont associations and the observed switching between symbionts in 4 of 79 tracked colonies (5%) are most parsimoniously explained by fine-scale functional differences between some C15-type symbionts. In particular, we hypothesize that corals associated with symbiont profiles dominated by the C15cu DIV (and possibly the C15m DIV) are less thermally tolerant than those dominated by C15 profiles. This hypothesis is further supported by the fact that all sampled colonies with less than 50% bleaching late in the heatwave (i.e., March/April 2016) were associated with C15 profiles. This included the only two colonies assigned to PKir-3 that were not severely bleached when sampled during population-level sampling on that time point (fig. S6) and that were associated with C15 symbionts despite the lineage being more typically associated with C15cu profiles. All colonies known to be associated with C15cu profiles before the heatwave bleached and experienced severe partial, or complete, mortality. Although functional differences between Symbiodiniaceae genera are well documented [i.e., Cladocopium versus Durusdinium ( 41 , 57 , 64 , 71 )], it has remained unclear until recently whether more closely related symbionts (e.g., different C15-type variants) can confer functional variation on their hosts that is meaningful in the face of heat stress. However, recent work showed that Porites cylindrica and Porites rus —two clearly defined (i.e., both morphologically and genetically) species—differ in thermal tolerance and also associate with different C15-type symbionts ( 72 ). Moreover, variants of C3 in the Persian Gulf have rapidly evolved increased thermal tolerance relative to close relatives from nearby areas ( 73 ). Our study offers evidence in support of the hypothesis that some closely related algal symbionts can vary meaningfully in function (or confer differences in function on their hosts) by demonstrating how intense warming can result in the near-complete loss of a previously prominent symbiont type while increasing the relative abundance of its close relatives. Whether C15cu profiles are more thermally sensitive could be confirmed by using laboratory experiments or by continuing to track colonies of PKir-3 through the next major heatwave event, to test whether these colonies, having lost their association with C15cu , are now equally tolerant of heat stress as PKir-1 and PKir-2. In this way, remaining colonies of PKir-3 may now be better suited to future heatwave events ( 57 ). However, if this is the case, then the beneficial response was confined to a minority of the PKir-3 population, with a mortality rate exceeding 80% and reductions in the overall relative abundance of that lineage (fig. S17). Implications for cryptic complexes facing extreme events Cryptic lineages are rapidly being uncovered across a broad range of taxa [e.g., ( 36 , 74 )], but their functional importance remains unclear. Theory would predict functional differences between cryptic lineages if they have diverged as a result of ecological speciation ( 75 – 77 ), but close relatives tend to be ecologically similar when comparing to a broader pool of taxa ( 78 ). Thus, it remains unclear whether climate change has the potential to differentially affect cryptic lineages in nature, threatening their persistence in the face of environmental change. Here, we demonstrated that cryptic lineages can be disproportionately affected (and some threatened) by marine heatwaves. This finding has important implications for our understanding of how climate change is affecting the diversity of marine species and populations. Moreover, the decoupling of cryptic lineages from their distinct symbiont assemblages demonstrates that highly specific species interactions may also be threatened in the face of climate change. Our inferred results would have differed substantially had we not characterized cryptic lineages in our study. While we would have still captured the major change in symbiont compositions that occurred during the heatwave, we would have missed the fact that the heatwave threatened a critical facet of diversity by disproportionately affecting one of the three cryptic lineages. While it is unclear whether Porites lineages represent early or incipient species or just highly differentiated populations, high (~80%) mortality across PKir-3 has the potential to threaten the long-term persistence of this lineage on Kiritimati. Given the widespread nature of cryptic diversity in corals and other habitat-forming species, we suggest that assessing the differential susceptibility of cryptic species or lineages to extreme events should be a major goal in the field of coral ecology and conservation. For example, differential success of distinct Pocillopora lineages that vary in their ability to associate with heat-tolerant symbionts ( Durusdinium glynni ) has recently been implicated in the long-term persistence of coral reefs in the eastern tropical Pacific ( 79 ). Some of the paradigms that underlie our understanding of population and species-level ecology (e.g., variation in traits, performance, and survival) could be reconsidered in light of cryptic diversity ( 74 ). We also found that human disturbance increased the susceptibility of tagged colonies to heatwave-driven mortality. While the interaction between human disturbance and coral lineage was not significant using either metric of mortality, it is noteworthy that mortality appeared equally high across all lineages on highly disturbed reefs, despite clear differences in mortality at lower disturbance sites ( Fig. 2 ). Past work has found that human disturbance can have either negative ( 41 , 49 ) or positive effects ( 80 , 81 ) on coral thermotolerance or survivorship through heatwaves and this has been linked to disturbance-driven changes in the symbiont composition ( 41 , 81 ) or density ( 80 ). Here, we found negative impacts of human disturbance irrespective of symbiont community composition (figs. S10 to S12). Overall, our results suggest that, for Porites , human disturbance negatively affects survivorship through heatwaves, regardless of cryptic host lineage or symbiotic associations. Moreover, given differences in survivorship between high and low disturbance sites for PKir-1 and PKir-2, our results demonstrate that the combined impacts of human disturbance and prolonged thermal stress may exceed the tolerance of even putatively stress-tolerant lineages. Thus, while the preservation or future-proofing of undisturbed reefs may benefit from the use of stress-tolerant genotypes or lineages [see, e.g., ( 67 , 82 )], these types of efforts may have less impact in highly disturbed areas. Limitations of study Although our research provides important insights into the impacts of heatwaves on cryptic diversity of reef-building corals and their symbiotic partners, its limitations may serve as interesting avenues for future research. In particular, it is challenging to directly tease apart the impacts of host genetics from those of symbiont identity in our system given that these nearly completely overlapped before the heatwave (see Figs. 3 to 5 ). While various patterns support the hypothesis that fine-scale differences between some C15-type symbiont populations explain some of the observed patterns in differential mortality and abundance, host genetics may also play a role. For example, we identified one gene that was an outlier between PKir-3 and both PKir-2 and PKir-1: ETS-related transcription factor Elf-2 , which may have possible links to coral immunity ( 53 ). Thus, it is possible that genetic differences between these cryptic lineages influenced the probability of survival, for example, by increasing bleaching propensity or susceptibility to disease following bleaching. Further experimental work or colony tracking of massive Porites lineages is required to definitively tease apart the role of host versus symbiont genetics in bleaching susceptibility or survival. Given the widespread nature and ecological role of Porites spp. on coral reefs globally, this offers a potentially critical direction for future investigation. Other limitations involve our ability to track all colonies and assign them to their respective cryptic lineages due to challenges (e.g., weather and safety) associated with tagging corals on remote reefs. Moreover, even when colonies were tracked, sequencing sometimes failed or samples were not taken because the surviving tissue area was deemed too small, leaving gaps in our knowledge of host lineage or symbiont identity. For example, two of the PKir-3 colonies that survived the heatwave could not be sequenced following the event. These colonies both bleached and experienced substantial partial mortality during the event; thus, it would be informative to know which symbionts they associated with after the event. Nonetheless, consistent patterns between tracked colonies and population-level sampling before and after the heatwave suggest that symbionts closely associated with PKir-3 before the heatwave were virtually absent afterward. Another limitation lies in our use of both 2b-RAD and coral ITS2 sequences to assign colonies to their respective lineages. While this multi-method approach allowed us to substantially increase replication, there are associated limitations. First, a number of colonies could not be assigned to a coral lineage because they had ITS2 barcode sequences that were ambiguous or did not match a reference sequence from the colonies also sequenced with 2b-RAD ( n = 15 colonies; see Materials and Methods). It is likely that these latter colonies are members of the three lineages but have ITS2 variants that were not represented in the colonies sequenced with 2b-RAD ( n = 67) by chance. Second, there is a higher risk of false lineage assignment when using barcoding rather than genomic data to assign colonies to cryptic lineages, although we note that barcoding sequences are commonly used to assign cryptic lineages or species [e.g., ( 36 , 42 )]. Introgression or incomplete lineage sorting, for example, could lead to a misleading pattern when looking at a single locus [e.g., ( 83 , 84 )]. Misassignment is probably more likely between PKir-1 and PKir-2 than between either one and PKir-3, based on the distribution of ITS2 barcodes across colonies sequenced with both methods (figs. S18 and S19), and due to the close affiliation of PKir-1 and PKir-2 in the genomic data. Thus, our conclusions about PKir-3 relative to the other two lineages should be robust to errors introduced by misleading barcode sequences. Nonetheless, as a sensitivity test, we tested whether cryptic lineage predicted survivorship using only the subsample of colonies sampled with 2b-RAD and with known survival ( n = 61) and found similar overall statistical results (fig. S20). Finally, our study focused on only a single depth range; thus, the full distributions of these cryptic lineages remain to be characterized. Although all three Porites lineages were sympatric across Kiritimati, the extent to which they fully overlap across the seascape remains unclear. Past work on cryptic lineages has demonstrated that they often inhabit discrete microenvironments even if they do overlap over broader spatial scales [e.g., ( 30 , 39 )]. While depth was standardized across sites and tagged colonies (see Materials and Methods), it is possible that asymmetrical gene flow inferred across lineages could be the result of differences in habitat. For example, if PKir-3 is found across a larger range of habitats than the other two lineages, this could help to explain why gene flow was reduced into PKir-3. Given the functional differences in thermal tolerance through a major heat stress event, it is possible that these lineages generally occupy different depth ranges but co-occur in the moderate forereef environment where we sampled [as observed in Pocillopora spp. in Mo’orea ( 42 )]. Understanding the distribution of cryptic coral lineages across different environments will be important for elucidating the processes driving and reinforcing differentiation across these lineages and for better predicting the outcomes of future bleaching events ( 27 ). Summary By coupling host genomic sequencing and Symbiodiniaceae metabarcoding with longitudinal coral colony tracking, we showed that differential mortality during a marine heatwave resulted in a substantial change in the relative abundances of cryptic lineages of massive Porites , with a relative decrease of ~80% in one lineage (PKir-3) across the atoll following the heatwave ( Table 1 ). This provides direct evidence that heatwaves have the potential to threaten cryptic genetic diversity, even among one of the most common and stress-tolerant coral genera. These cryptic Porites lineages had specific symbiont associations that recombined during the heatwave, highlighting a likely mechanism behind differential survival of lineages. Moreover, mortality was strongly predicted by human disturbance but only in two of the three cryptic lineages, illustrating that anthropogenic drivers can mediate the strength of selection during extreme events. High mortality in PKir-3 decreased its overall population size, increasing the probability that this lineage may be extirpated in the future. Although changes in symbiont associations in this lineage may facilitate adaptive change ( 31 ) in the face of future heatwaves, with unknown functional trade-offs ( 57 ), the loss of this specific host-symbiont pairing demonstrates how heatwaves may be eroding specific biotic interactions in addition to threatening diversity. Our study demonstrates that strong marine heatwaves may threaten biodiversity at finer scales than have generally been appreciated to date. Overall, these findings underscore the need to better understand cryptic diversity within our current taxonomic framework. They also illustrate how climate change may threaten the persistence of undiscovered diversity, causing Centinelan extinctions—losses of taxa that are never described by science and are therefore unrecorded ( 85 ). Moreover, this undescribed diversity may help explain enigmatic variation in coral bleaching and mortality, and improve future predictions of bleaching severity and impact."
} | 12,614 |
30519408 | PMC6262932 | pmc | 442 | {
"abstract": "Abstract For animals that harbor photosynthetic symbionts within their tissues, such as corals, the different relative contributions of autotrophy versus heterotrophy to organismal energetic requirements have direct impacts on fitness. This is especially true for facultatively symbiotic corals, where the balance between host‐caught and symbiont‐produced energy can be altered substantially to meet the variable demands of a shifting environment. In this study, we utilized a temperate coral–algal system (the northern star coral, Astrangia poculata, and its photosynthetic endosymbiont, Symbiodinium psygmophilum ) to explore the impacts of nutritional sourcing on the host's health and ability to regenerate experimentally excised polyps. For fed and starved colonies, wound healing and total colony tissue cover were differentially impacted by heterotrophy versus autotrophy. There was an additive impact of positive nutritional and symbiotic states on a coral's ability to initiate healing, but a greater influence of symbiont state on the recovery of lost tissue at the lesion site and complete polyp regeneration. On the other hand, regardless of symbiont state, fed corals maintained a higher overall colony tissue cover, which also enabled more active host behavior (polyp extension) and endosymbiont behavior (photosynthetic ability of Symbiondinium ). Overall, we determined that the impact of nutritional state and symbiotic state varied between biological functions, suggesting a diversity in energetic sourcing for each of these processes.",
"introduction": "1 INTRODUCTION The differential utilization of alternative energy sources can directly influence an organism's growth, reproduction, behavior, and survival (Heino & Kaitala, 2001 ). In organisms that can obtain carbon flexibly from multiple pathways, energetic dynamics can be particularly complex. For example, corals harboring photosynthetic algal symbionts ( Symbiodinium ) can obtain energy through transfer of photosynthate from the endosymbiont or by predation on plankton (Grottoli, Rodrigues, & Palardy, 2006 ; Palardy et al., 2008 ). When obtaining energy via photosynthesis, the coral holobiont (host animal plus symbionts) is functioning as an autotroph, and when obtaining its energy via predation, it is functioning as a heterotroph. However, corals often obtain energy through multiple sources simultaneously, and there may be interactions between autotrophy and heterotrophy. Additionally, as colonial organisms, energy must be translocated across individual polyps to maintain colony function (Fine, Oren, & Loya, 2002 ; Oren, Rinkevich, & Loya, 1997 ). As such, energy budgeting within coral colonies is complex, dynamic, and not yet well understood. For tropical corals in well‐lit, shallow environments, host colonies can meet or exceed their metabolic needs through transfer of photosynthate from Symbiodinium spp. (Muscatine, 1990 ). It has been hypothesized that these corals prey on zooplankton mainly to supplement the energy they receive from the endosymbiont and to supply essential nutrients (such as phosphorus and nitrogen; Johannes, Cole, & Kuenzel, 1970 ; Tanaka, Miyajima, Koike, Hayashibara, & Ogawa, 2006 ) and that prolonged heterotrophic compensation may be a stress response that increases resilience under conditions unfavorable to autotrophy (Hughes & Grottoli, 2013 ; Levas et al., 2015 ). Additionally, heterotrophic feeding can enhance growth rate, protein, and chlorophyll concentrations, as well as calcification rates in daylight and in darkness (Ferrier‐Pagès, Witting, Tambutté, & Sebens, 2003 ; Houlbrèque, Tambutté, & Ferrier‐Pagès, 2003 ). However, the degree to which a colony can supplement lost photosynthetic resources appears to vary by species (Anthony & Fabricius, 2000 ; Grottoli et al., 2006 ), and studies have suggested that the balance between energy sources might not be fixed (Piniak, 2002 ). In the temperate realm, a highly variable environment can lead to a wide variety of flexible feeding strategies, such as those employed by facultatively symbiotic corals like Astrangia poculata (= A. danae ; Peters, Cairns, Pilson, & Wells, 1988 , Figure 1 ), Oculina patagonica (Fine, Zibrowius, & Loya, 2001 ), and Oculina arbuscula (Leal et al., 2014 ). Heterotrophy has many effects on the metabolism and physiology of these facultatively symbiotic temperate corals: (a) It can mitigate thermally induced “bleaching” (a sharp reduction in symbiont density caused by exposure to elevated temperatures; Aichelman et al., 2016 ); (b) it increases nitrogen uptake and excretion (Szmant‐Froelich & Pilson, 1984 ); (c) it increases calcification and growth (Jacques & Pilson, 1980 ; Jacques, Marshall, & Pilson, 1983 ; Miller, 1995 ); and (d) it reduces damage from sedimentation (Peters & Pilson, 1985 ). Symbiotic state can impact the effects of heterotrophy, although the presence of photosynthetic symbionts does not preclude heterotrophy. For example, symbiotic colonies of A. poculata can retain more carbon ( 14 C) from heterotrophic sources than aposymbiotic colonies (Szmant‐Froelich, 1981 ), and there is evidence for transfer of photosynthetic carbon to coral host tissue (Schiller, 1993 ). Additionally, the endosymbiont in fed A. poculata colonies fix carbon more efficiently (but translocate less 14 C) than their starved counterparts (Szmant‐Froelich, 1981 ). This suggests a potentially high degree of interconnectivity between energy strategies (Piniak, 2002 ) as well as a complex dynamic between simultaneous autotrophy and heterotrophy. Figure 1 (a) Symbiont states in Astrangia poculata : With polyps contracted, fully aposymbiotic colonies appear white in color; fully symbiotic colonies appear brown in color. (b) [a] Aposymbiotic, [m] mottled, and [s] symbiotic colonies of A. poculata occur concurrently in the field. (c) Adjacent polyps in a mottled colony demonstrate [a] aposymbiotic and [s] symbiotic densities. Photographs by E.M. Burmester The northern star coral A. poculata has an expansive range along the east coast of North America, from Florida and the Gulf of Mexico to southern Massachusetts (Dimond & Carrington, 2007 ; Dimond et al., 2013 ). In nature, these corals can exist in one of three basic symbiotic states with the endosymbiont, Symbiodinium psygmophilum (Lajeunesse, Parkinson, & Reimer, 2012 ): Fully symbiotic corals appear brown; aposymbiotic corals harbor far fewer symbionts, and they appear white; symbiont density can also vary from polyp to polyp, producing a mottled, mixed colony comprising both white and brown polyps (Cummings, 1983 ). Unlike in tropical corals, in A. poculata , the aposymbiotic state is not the result of stress (i.e., bleaching); white colonies of A. poculata are as “healthy” as brown colonies and can persist indefinitely in nature (Grace, 2004 ). The relatively low density of S. psygmophilum is actively maintained by the regular expulsion of the symbiont (Dimond & Carrington, 2008 ). Regardless of symbiont state, temperate colonies rely heavily on heterotrophy (Farrant, Borowitzka, Hinde, & King, 1987 ; Szmant‐Froelich & Pilson, 1984 ), with symbiont density only explaining an estimated 23% of growth in the field (Dimond & Carrington, 2007 ). This study investigates the interaction of feeding and symbiotic state on wound healing in Astrangia poculata . There is ample evidence to suggest that both lesion recovery and colony maintenance (i.e., maintaining a healthy layer of tissue cover) are energetically costly activities that are often in conflict with each other and in conflict with other critical physiological functions such as reproduction, calcification, and growth (Anthony, Connolly, & Willis, 2002 ; Richmond, 1987 ; Rinkevich, 1996 ; Rotjan and Lewis, 2009 ; Ward, 1995 ). In addition, the process of lesion repair can require a high degree of colonial energy integration, which can vary by wound and colony characteristics (Oren, Benayahu, Lubinevsky, & Loya, 2001 ; Szmant‐Froelich, Yevich, & Pilson, 1980 ). Because of its flexibility and tractability, A. poculata makes an ideal study organism for investigating the dynamics between energy sourcing and organismal health. This study uses (a) small‐scale wound lesion and total colony tissue recovery, (b) foraging behavior, and (c) symbiont density and photosynthetic efficiency metrics to assess colony health and stress response in the presence and absence of both autotrophic and heterotrophic nutritional strategies in naturally occurring symbiotic and aposymbiotic A. poculata colonies.",
"discussion": "4 DISCUSSION Our findings highlight some of the dynamic pathways through which coral colonies might obtain, distribute, and utilize energetic resources during the process of recovering from physical abrasion. This study suggests that autotrophy plays an important role in wound recovery and that there may be an important interplay and feedback (both positive and negative) between autotrophy and heterotrophy. As previously found in A. poculata , symbiotic state had a significant role on healing initiation and success as well as proportional surface area recovery to wounds (Burmester et al., 2017 ; DeFilippo et al., 2016 ). However, symbiont state alone was not enough to maximize healing potential. Starved‐symbiotic and fed aposymbiotic healed comparably, while there was an additive negative feedback between starved aposymbiotic corals (no nutrition from either source; little/no healing), and an additive positive feedback between fed symbiotic corals (nutrition from both sources; highest healing ability) (Figures 3 and 4 ). While nutritional state impacted healing initiation, it had no statistical effect on healing success or surface area recovery, presumably because there was some autotrophic compensation. On the other hand, only nutritional treatment (and not symbiont state) appeared to play a role in total colony tissue maintenance. These findings suggest that energy might not be regulated or distributed uniformly across levels of body organization, which is to be expected in a colonial organism that can translocate resources. This is consistent with other studies, where branching growth tips of Stylophora pistillata had significantly less 14 C products than fragments from below branch tips (Rinkevich & Loya, 1983 ). Additionally, both symbiotic state and lesion induction can alter the quantity and directionality of carbon translocation across a coral colony (Fine et al., 2002 ; Oren et al., 1997 ). In O. patagonica , preferential translocation to recovering tissue proceeded from a distance of 4–5 cm, but this phenomenon does not occur in colonies that were fully or partially (30%–80%) bleached (Fine et al., 2002 ). The pace and completion of wound recovery are subject to the impacts of several intrinsic and extrinsic factors (such as colony size, wound size, wound location, temperature, disease state, sedimentation [as reviewed by Henry & Hart, 2005 and for example: Van Veghel & Bak, 1994 , Meesters, Noordeloos, & Bak, 1994 , Meesters, Wesseling, & Bak 1996 , Meesters, Pauchli, & Bak, 1997 , Nagelkerken & Bak, 1998 , Nagelkerken, Meesters, & Bak, 1999 , Kramarsky‐Winter & Loya, 2000 , Rotjan & Lewis, 2005 , Edmunds, 2009 , Denis et al., 2011 , Cameron & Edmunds, 2014 ]), which also have the potential to interact with energy sourcing and nutritional state. The type of damage inflicted may also play a role in how energy is regulated or redirected to recovery and other biological processes (DeFilippo et al., 2016 ). Dislodged colonies of Pocillopora damicornis with edge damage experience a decrease in overall energy allocation, resulting in higher mortality rates and decreased growth and reproduction (Ward, 1995 ). Meanwhile, fragmentation bears no significant impact on growth and mortality, but results in higher overall energy allocation and increased reproduction (Ward, 1995 ). Therefore, it is likely that tissue maintenance and damage are regulated differently for small‐scale local wounds (e.g., the single polyp removal demonstrated in this study) and across a coral's total colony tissue cover (e.g., DeFilippo et al., 2016 ), potentially due to underlying compartmentalized nutritional gradients across an energetically integrated colony (Conlan, Humphrey, Severati, & Francis, 2018 ). Interestingly, our results indicate that symbiont state is more important to the regulation of tissue surface area at the wound level while overall maintenance of total colony tissue cover is more greatly impacted by the presence or absence of prey items. Therefore, there could be an added cost to lesion recovery during and after bleaching events that may not be fully supplemented via heterotrophy. This is consistent with tropical corals that rely more heavily on autotrophy; for example, wounded Orbicella colonies recovered more slowly from bleaching compared to intact colonies (Rotjan et al., 2006 ). Additionally, this added cost may be compounded by the influence of other external disturbances to wound recovery, such as ocean acidification (Edmunds & Yarid, 2017 ) or elevated sea surface temperatures (Bonesso, Leggat, & Ainsworth, 2017 ). Since recent evidence also suggests that physiological integration (i.e., high integration) may increase risk of bleaching stress (Swain et al., 2018 ), understanding how corals utilize, store, and distribute energy from multiple nutritional sources may prove invaluable to conservation efforts. Both the availability (stimulus) of prey items and the history of heterotrophic opportunity significantly influenced polyp foraging behavior. Fed colonies maintained a higher degree of polyp expansion than unfed colonies at all time points, and the introduction of food particles induced even greater expansion. In tropical, obligate symbiotic scleractinians, symbiotic photosynthetic energy resources have been shown to influence heterotrophic activity. Colonies of Pocillopora damicornis maintained under dark conditions for 2 weeks ingested less Artemia nauplii than those in lighted conditions, suggesting a dependence on energy from photosynthesis to meet the metabolic needs required for sustainable foraging behavior (Clayton & Lasker, 1982 ). In the present study, there was no observed statistical difference in foraging activity between symbiotic and aposymbiotic colonies. These results are similar to those found for other facultatively symbiotic corals. Piniak ( 2002 ) found that prey capture efficiency varied by prey type and flow rate, but observed no difference between (fed) symbiotic and aposymbiotic colonies of Oculina arbuscula . Coral colonies may also forage advantageously regardless of photosynthetic activity, as even obligate, tropical corals have been shown to seek heterotrophic nutrition even if metabolic carbon requirements are met via autotrophy (Ferrier‐Pagès, Allemand, Gattuso, & Jaubert, 1998 ; Grottoli et al, 2006 ). Additionally, the regular availability of heterotrophic food sources increased foraging activity in fed colonies both with and without a food stimulus. Therefore, colonies with stable nutritional inputs are better able to maintain a fuller, long‐term foraging effort, allowing them to not only respond to a food stimulus, but to also survey their environment. This suggests a heterotrophic, rather than autotrophic, mechanism for inducing appropriate behavior to meet metabolic demands in temperate, facultatively symbiotic corals. In this study, while the photosynthetic efficiency (maximum quantum yield) of fed symbiotic colonies was significantly higher than that of all aposymbiotic (fed and starved) colonies, there was no difference between aposymbiotic colonies and starved‐symbiotic colonies. This phenomenon does not appear to derive from a loss of chlorophyll, which suggests an energetic cost to symbiont photosynthesis that must be fulfilled via host heterotrophic means. In fact, zooxanthellae have been documented to exhibit heterotrophic behavior inducing a parasitic metabolic burden on the facultatively symbiotic anemone Aiptasia pulchella (Steen 1986; Baker, Freeman, Wong, Fogel, & Knowlton, 2018 ). Previous studies have documented an enhancement to photosynthesis in temperate corals after feeding (Jacques & Pilson, 1980 ). Similarly, rates of photosynthesis increased (2–10×) after the introduction of heterotrophic food sources to Stylopora pistilla (Houlbrèque et al., 2003 ). The decline in photosynthetic efficiency for starved, symbiotic colonies could also potentially be attributed to their higher rates of polyp contraction. A. poculata go through a winter quiescence phase, when polyps enter a state of metabolic dormancy (Jacques et al., 1983 ) and tentacles no longer elicit a tactile feeding response (Grace, 2017 ). Quiescence corresponds with wintertime food scarcity in New England due to relatively oligotrophic waters compared to summer nutrient conditions and corresponding plankton blooms (Grace, 2017 ). During quiescence, A. poculata colonies in New England experience a decline in photosynthetic efficiency and in ACD (Dimond & Carrington, 2007 ). Although the cold wintertime temperatures have assumed to be a driver of quiescence behavior, the polyp behavior and photosynthetic efficiency of starved corals in our experiment suggest that quiescence may instead have a nutritional cue, since ambient temperatures (18°C) were maintained throughout the experiment. Photoperiod and/or angle of incidence may also play a role, as Fabricius and Klumpp ( 1995 ) found reduced photosynthetic productivity and increased required levels of irradiance to achieve photosynthetic compensation and saturation in contracted large‐polyped soft corals. Though again, PAR was maintained throughout the 60 days of this experiment. The dynamic relationship between Astrangia poculata and Symbiodinium psygmophilum is well‐documented, with both symbiotic states characterized across its range (Dimond et al., 2013 ), and the potential for state‐switching under experimental conditions (Dimond & Carrington, 2007 ). The aposymbiotic state is common in nature (Grace, 2004 ) despite relevant losses in recovery ability (Burmester et al., 2017 ; DeFilippo et al., 2016 ) as well as resilience to stress (Holcomb, Cohen, & McCorkle, 2012 ; Holcomb, McCorkle, & Cohen, 2010 ). It has been hypothesized that the persistence of the aposymbiotic life history may be due, in part, to the relative reduction in polyp loss under cold temperatures during winter quiescence (Dimond et al, 2013 ). Despite its thermal tolerance and resilience to chronic cold exposure (Thornhill, Kemp, Bruns, Fitt, & Schmidt, 2008 ), S. psygmophilum experiences a rapid decline and cessation in maximum quantum yield at winter temperatures. Combined with metabolic dormancy and a lack of feeding response, the demonstrated decline in photosynthetic efficiency in the absence of heterotrophy in implied energetic cost of these symbionts to the host coral could explain the reduced polyp loss (and correspondingly higher biomass compared to symbiotic corals) of aposymbiotic colonies under overwinter conditions. While feeding behavior was ensured among all polyps for each colony, this study did not specifically determine the quantity of food consumed nor the amount of carbon incorporated. Likewise, while we recorded light availability (PAR) and photosynthetic efficiency ( F \n v / F \n m ), neither of these measurements provide accurate insight into photosynthetic carbon production for this coral. As such, it would be difficult to infer how specific pathways might be impacted by differences in symbiont state and experimental feeding treatments on the cellular level. Additionally, these results represent the nutritional dynamics of a single, facultatively symbiotic species that may not be directly applicable to tropical coral species with higher dependencies on autotrophic pathways. However, the results of this study demonstrate significant and predictable morphological and stress‐tolerant responses that influence key life history strategies in temperate corals and broadly highlight the importance in understanding the complexity of energy sourcing when establishing energy budgets for maintaining organismal health."
} | 5,101 |
37299637 | PMC10254461 | pmc | 443 | {
"abstract": "In this paper, we have fabricated non-volatile memory resistive switching (RS) devices and analyzed analog memristive characteristics using lateral electrodes with mesoporous silica–titania (meso-ST) and mesoporous titania (meso-T) layers. For the planar-type device having two parallel electrodes, current–voltage (I–V) curves and pulse-driven current changes could reveal successful long-term potentiation (LTP) along with long-term depression (LTD), respectively, by the RS active mesoporous two layers for 20~100 μm length. Through mechanism characterization using chemical analysis, non-filamental memristive behavior unlike the conventional metal electroforming was identified. Additionally, high performance of the synaptic operations could be also accomplished such that a high current of 10 −6 Amp level could occur despite a wide electrode spacing and short pulse spike biases under ambient condition with moderate humidity (RH 30~50%). Moreover, it was confirmed that rectifying characteristics were observed during the I–V measurement, which was a representative feature of dual functionality of selection diode and the analog RS device for both meso-ST and meso-T devices. The memristive and synaptic functions along with the rectification property could facilitate a chance of potential implementation of the meso-ST and meso-T devices to neuromorphic electronics platform.",
"conclusion": "4. Conclusions In this study, we have introduced synaptic device functions with meso-ST and meso-T active layers with long channel lateral electrodes. Asymmetric pinched hysteresis approves the self-rectification with different threshold voltages, which was significant to RS-based memory devices as it evaded the cross talk. Both devices showed potentiation and depression synaptic properties. In the near future, the meso-ST and meso-T devices are expected to be successfully integrated in planar structured neuromorphic electronics.",
"introduction": "1. Introduction Since intensive introductions of resistive switching (RS) devices as novel memory element after Chua’s perception, many kinds of RS devices have developed and demonstrated [ 1 , 2 , 3 , 4 ]. The adoption of RS devices has been mainly resided on digital switching element due to first introduction of RS phenomena on ReRAM devices. Therefore, restrictions of utilization of RS devices should include a lack of a lateral electrode device with long channel structure to show digital RS switching or analog RS switching [ 1 , 2 ]. For the RS device, transition metal oxides (TMOs) such as TiO 2 , Ta 2 O 5 , and NiO, have been extensively adopted and successfully implemented into the RS active layers including memristors and synaptic devices. Typically, the TMO active devices have been fabricated by sputtering as well as solutional processes such as the sol-gel process [ 3 , 4 ]. For the RS property mechanisms of the TMO devices, filament formation from metal electrodes such as aluminum (Al) or silver (Ag) has been typically proposed for capacitor-type devices or crossbar-type devices. Usually, the electroforming process has been required for the RS behavior to show reversible formation and rupture of metallic filament, which could be observed even in lateral electrode devices upon 100~1000 nm long TiO 2 channel [ 3 ]. In this case, physically migrated or protruded Al mass could be one or multiple electrical paths to touch the other side electrode to accomplish the RS property [ 4 , 5 ]. In addition, for the RS property of the planar electrode device, field-assisted activated hopping of metal ions could be a resource for the filament formation and diminution. However, its structural instability with poor retention had to block the further development for a neuromorphic device application [ 3 ]. Meanwhile, non-filamental RS behaviors have been continuously demonstrated with TMO active layers [ 4 , 5 , 6 ]. Typically, the operation mechanism of the non-filamental RS devices have been laid with formation of conduction filaments (CFs) inside of TMO materials, which were involved with evolutions of oxygen ions or oxygen gas [ 4 ]. The RS phenomena have been based on movement, accumulation, and resultant agglomeration of oxygen vacancies transited into CFs [ 6 ]. For the mechanism of CFs formation, oxygen deficient and non-stoichiometric phase has been believed to be generated in the CFs, which could be reversibly disrupted or resurrected under direct current (DC) voltage or pulsed spike voltage under electrode-stack-type or crossbar-type device structures. Meanwhile, there have been very limited studies about lateral-electrode-formatted RS devices, which has long RS active layer between two planar electrodes. For the lateral electrode study, 0.05~0.1 μm long RS active layer between planar electrodes have been examined to elucidate and analyze the possible existence of metal filament or oxygen vacancy-based CFs through a direct tracking tool [ 6 ]. For example, with a high resolution transmission electron microscopy (HR-TEM) observation, the filament formation between two lateral electrodes could be explicitly observed, which were formed by d-orbital overlap covering electrode metal (Pt) and reduced metal ions of the TMO layer [ 6 ]. Particularly, through the real-time TEM analysis, O 2 gas bubble formation could be identified during the oxygen vacancy build-up process inside of the TMO layer. However, the length of CFs was as short as 0.05~0.1 μm between two electrodes, which was commensurate with the thickness of stack-type or crossbar-type devices. Separately, with the TMO-based RS devices, nanoparticles (NPs)-shaped TMO materials can be applied to fabricate the RS devices with or without a polymeric matrix [ 7 , 8 ]. However, the NPs-based devices were strongly dependent on the charge trapping mechanisms, which could confine the dimension of the RS phenomena, especially in length scale [ 7 , 8 , 9 ]. For the TMO RS devices, electroforming of oxygen vacancy through conduction filaments (CFs) has been widely accepted for RS phenomena mechanism of transited current level at high resistance state (HRS) up to low resistance state (LRS) [ 10 , 11 , 12 ]. Specifically, for TiO 2 -based RS devices, mesoporous silica (SiO 2 ) titania (TiO 2 ) (meso-ST) has been introduced recently. Mainly, the meso-ST composites have been characterized with high surface area, degree of hydrophilicity, excellent ion exchange ability, and stable framework structure, which have drawn considerable attention due to their catalytic properties [ 11 ]. However, the meso-ST could have shown memristive characteristics of long-term potentiation (LTP) associated with spike time dependent plasticity (STDP) under cap-type or crossbar-type device layouts [ 11 ]. Here, the mesoporous nature could provide higher chances of forming CFs bundles with empty spaces or rooms to facilitate formation of O 2 gas or its diminution, which would be critical to oxygen vacancy generation or metallic cation’s reduction event. Very recently, there have been continuous demonstrations of synaptic transistors, which should have relatively long channels (~10 μm) between the lateral source and drain electrodes. However, no sufficient discussions have been dedicated to the mechanism of memristive behaviors [ 12 , 13 ]. In fact, for the conduction mechanism of SiO 2 -based RS active material, ionic transport should be significantly considered as a major mechanism of current path with charge carrier of proton (H + ) migration [ 12 , 13 , 14 , 15 ]. Even in these cases, the meso-ST and meso-T layers can be considered as an electrolyte-based active channel, which may facilitate fast switching of migrative ions [ 13 , 14 ]. For example, in mesoporous silica, ions or charge carriers could be mainly protons, which have the smallest size apart from typical charge carriers of electron or hole [ 13 ]. Therefore, ionic current would be responsible for current–voltage (I–V) curves having hysteresis behaviors. Along with TMOs, a broad area of materials has been shown the applications in RS switching such as 2D materials including perovskite and chalcogenide with long channels [ 16 , 17 ]. For the perovskite, there have been many reports about dual functionality of digital and analog RS properties, which could be successfully demonstrated synaptic device operations [ 16 ]. However, there have not been many lateral device formats so far with the perovskites, instead of the vertical device format. For the perovskite case, metal ions such as Ag + could migrate through the electrolyte-like perovskite layer for the RS phenomena [ 16 ]. For the case of chalcogenides such as MoS 2 , a possibility of lateral-type format for the synaptic device purpose was strongly identified [ 17 ]. The mechanism of the metal dichalcogenide device could be investigated with scanning tunneling microscope (STM), which concretely detected defects throughout longitude of the 2D layer. The origin of defects was identified as metal substitutions into sulfur vacancies, which can open up a precisely tuned defect engineering for the planar-type memristor device [ 17 ]. In this study, meso-ST and meso-T have been examined and demonstrated as memristive active layer with long channel gap (20~100 μm) between two Al lateral electrodes. Using the mesopores, more accessible O 2 formation and diminution were available through the pores and, as a result, CFs formation through the long gap between planar electrodes could be achieved to show neuromorphic memristive characteristics. Through this study, unlike the common expectation of the catalytic application about meso-ST and meso-T matrix, a novel application to exploit synaptic device function can be drawn in integrated circuits (ICs) of brain-inspired electronics. For comparisons of the meso-ST and meso-T with other TiO 2 and SiO 2 -based RS materials, Table 1 is presented in terms of deposition method, unipolar or bipolar switching behavior, digital or analog characteristics, and conduction mechanisms [ 18 , 19 , 20 ]. Above all, most of the TiO 2 -based RS layers have been prepared by sputtering method so far [ 3 , 4 , 6 , 10 ]. Even an electron beam evaporation could be particularly utilized for the TiO 2 -based RS layer as shown in Table 1 [ 18 ]. However, both evaporation and sputtering methods require ultra-high vacuum (UHV) system to prepare the films. Meanwhile, the meso-ST and meso-T films could be prepared by a simple spin-coating method. For another comparison of the meso-ST with the SiO 2 -based RS material, SiO 2 could be usually prepared by plasma-enhanced chemical vapor deposition (PECVD), which also requires massive film deposition equipment unlike the spin-coating method [ 19 ]. In addition, there have been reports to prepare the SiO 2 -based films by the spin coatings based on sol-gel formation [ 12 , 14 , 20 ]. However, the SiO 2 -based films were not RS-active [ 12 ], non-synaptic without an additional semiconductor [ 14 ], and solely digital functioned [ 20 ].",
"discussion": "3. Results and Discussions Figure 1 shows device structures with long channel (22.64 μm) between the lateral electrode for the meso-ST device. As shown in Figure 1 , mesopores could be identified by TEM image with the meso-ST layer. Both meso-ST and meso-T films were prepared by evaporation-induced self-assembly (EISA) method, which constructed well-ordered mesopores of ~10 nm as shown in the TEM image of Figure 1 . As shown in Figure 1 , all the lateral electrodes were Al, which could be easily oxidized in contact of the meso-ST or meso-T layers. Even though there could be some oxidations on the Al surface, it was assumed that current injection was still available. In addition, since the channel length was very long in this study, the risk of oxygen or oxygen ion (O 2− ) transport could be minimized, which could be critical under the existence of aluminum oxide (AlO x ) layer. Moreover, as illustrated in Figure 1 , the meso-ST and meso-T layers were formed on an insulating layer of 300 nm thermally grown SiO 2 surface. Therefore, there were no leakage current between electrodes and p-doped Si in these devices. Figure 2 a,b show analog non-volatile memory I–V curves as well as rectifying behavior shown for both meso-ST and meso-T devices. Even under high possibility of humidity effect on the meso-ST and meso-T devices, all the experimental results have been taken in ambient conditions. Furthermore, during each test, the relative humidity (30–90%) was recorded. The devices were stable during the measurements in ambient conditions. As shown in Figure 2 a, small hysteresis could be found for I–V curves of the meso-ST device with 22.64 μm channel. The length was so long that electrodes could not trigger simple metallic migration, which eventually would form metallic filaments. Therefore, the conduction mechanism is believed to be from CFs out of oxygen vacancy and ionic or proton transport [ 4 , 6 , 11 , 14 , 15 ]. Actually, low temperature (300 °C) processed silica layer was assumed as an electrolyte film having proton-based electrodynamics [ 14 ]. Since the meso-ST layer has the low temperature annealed silica portion, the long channel could provide the memristive I–V curves based on the proton migration. The inlet graph of Figure 2 a shows EDS elemental analysis, which proved that there were no Al elements in channel area after repeated memristive I–V sweeps [ 3 ]. Therefore, it is suggested that no metallic filaments were involved for the memristive I–V behaviors as identified in Figure 2 a. For Figure 2 b, active channel length of the meso-T device was up to 100 μm, which could not be originated from metallic migration as well [ 3 ]. Here, due to the huge difference channel length (20 μm vs. 100 μm), the orders of current levels were varied from the meso-ST device (~10 −7 Amp) to the meso-T device (~10 −8 Amp) significantly. After I–V characterizations, the devices were very stable to show the memristor behaviors for one year in ambient conditions. With an operating voltage over ± 5 V, a high current value of 10 −7 Amp was observed for the meso-ST device’s excursive I–V curves. Here, the eight-wise excursions in Figure 2 a could be another representation of oxygen-based CFs evolution mechanism as reported in previous literature [ 11 ]. Furthermore, the meso-T device in Figure 2 b shows a decrease in the current level with each successive I–V cycle which implicitly plays a role in depression property in a typical neuromorphic characteristic. In addition, as shown in I–V curves, the direction of I–V hysteresis was counter-clockwise (CCW) for both meso-ST and meso-T. In addition, the I–V curves shows analog RS memory having concurrent threshold RS, which can have rectifying characteristics. The threshold RS property can act as a selection diode, which can reduce crosstalk among RS or memristive devices under a high integration state. This rectification property could help the device to show unidirectional current flow which can be useful to mimic unidirectional synaptic property which helps the device to improve learning and unlearning phenomenon for electrical synapse [ 11 ]. Figure 3 shows pulsed spike effect of LTP and LTD for the ( a ) meso-ST device and ( b ) meso-T device, respectively. The y-axis represents currents under read voltage of +0.1 V with repeated pulse or spike operations. In the implementation of the pulse measurement, through repeated measurements of 50 times for meso-ST and 30 times for meso-T in each potentiation and depression condition, a certain tendency of synaptic characteristics was identified. In Figure 3 a, the spike time width was 0.3 s with ±7 V spike voltage for the meso-ST device. However, for the meso-T device as shown in Figure 3 b, long spike time width such as 0.3 s could now demonstrate discrete potentiation or depression. Instead, a very short spike time width of 0.05 s could generate noticeable synaptic potentiation and depression. Compared with a cap-type device, the extent of LTP in Figure 3 a was much smaller than the stacked meso-ST active device due to the long active length between lateral electrodes [ 11 ]. Overall, synaptic responses of the meso-ST device were more apparent than those of the meso-T device, apart from channel length difference (20 μm vs. 100 μm). The higher synaptic efficacy with the meso-ST is believed to be from ionic or proton transport through mesoporous matrix of silica [ 12 ]. At first, since the Si atoms are smaller than Ti atoms in meso-ST film, the meso-ST having silica would be more advantageous for H + transport than meso-T film without silica [ 12 ]. Figure 4 shows XRD spectra for the meso-ST layer. The meso-ST revealed peaks positioned at 39.9, 46, 67.5, 82.2, and 86.1, corresponding to the planes of anatase (111), (200), (220), and rutile (311), (222) in accordance with JCPDS card no. (21–1276), additionally having hematite plane (440), respectively. Here, the anatase phase of titania is dominant probably by the fact that a small fraction of Ti atoms interacted or entered into the lattice structure and induced the lattice expansion. Here, the Si ion radius is known to be smaller than the Ti 4+ (r = 0.079 nm). The average of crystallite size of ~9 nm is calculated for the meso-ST as-synthesized sample. It shrinks down to 8 nm after 1 h heating and again expands up to 10 nm after 2 h heating [ 11 ]. However, since the annealing temperature was still as low as 300 °C, it is believed that there is a significant portion of amorphous phase in this film. Figure 5 shows the Raman spectra of meso-ST as-synthesized, annealed at 200 °C and 300 °C for 1 h and 2 h, respectively. The intense peaks at 785 cm −1 in meso-ST as-synthesized and annealed samples are ascribed to vibrational absorbance of TiO 2 or Ti-O-Ti bonds stretching. In addition, different phases of TiO 2 such as rutile and anatase showed strong absorption bands in range of 650~850 cm −1 [ 11 ]. However, in Figure 5 , there are no corresponding peaks for rutile or anatase state, which could suggest an amorphous state of the m-ST. Sharp bands at 1080~1083 cm −1 in separate samples are attributed to asymmetric Si-O-Si vibrational signals, which are very weak for the as-synthesized sample. Meanwhile, edge peak at 1197 cm −1 on three meso-ST samples correspond to asymmetric stretching vibrations of Si-O-Si [ 11 ]. Additionally, there are no peaks detected for bulk Si or crystalline Si in Figure 5 [ 21 ]. In the meso-ST layer, broad peaks around 1203 cm −1 and 1371 cm −1 are attributed to the O-C=O symmetric and asymmetric stretch modes. They represent organic functional groups or volatile compounds of surfactant F127. The bands could be diminished with further annealing as shown in Figure 5 . A noticeable peak at 1741 cm −1 in meso-ST as-synthesized sample is ascribed to C=O bond’s stretching or vibration, which decreased after 2 h annealing. The bands at 2876 cm −1 , 2946 cm −1 , and 2279 cm −1 of the meso-ST can be attributed to the C-H stretching bonds, respectively, which are constituents of organic surfactant. The bands around 1655 cm −1 , which shifted to 1631 cm −1 in three meso-ST layers, are assigned to bending or vibration mode of H-O-H. The peak for the H-O-H can be from a protonated silanol group, which can greatly influence the proton transport under an electrical field [ 12 , 14 , 15 ]. The peculiar band at 3283 cm −1 (O-H stretching) in the meso-ST became a gradually broader form as-synthesized to annealed at 3013 cm −1 , which was ascribed to the presence of Si-OH and Ti-OH bonds. Particularly, the silanol group (Si-OH) is believed to be critical to ionic current by surface-mediated transport depending upon its degree of protonation (-O + H 2 ) within mesopores [ 12 ]. The annealed meso-ST is known to be evolved into solid semi-spheres, which showed the presence of mesopores (~10 nm diameter) among the particles. The generated mesopores would allow O 2 gas to penetrate within the meso-ST’s interior by diffusion, which can influence the conduction mechanism of the meso-ST layer."
} | 5,037 |
38400487 | PMC10892219 | pmc | 444 | {
"abstract": "Organizations managing high-performance computing systems face a multitude of challenges, including overarching concerns such as overall energy consumption, microprocessor clock frequency limitations, and the escalating costs associated with chip production. Evidently, processor speeds have plateaued over the last decade, persisting within the range of 2 GHz to 5 GHz. Scholars assert that brain-inspired computing holds substantial promise for mitigating these challenges. The spiking neural network (SNN) particularly stands out for its commendable power efficiency when juxtaposed with conventional design paradigms. Nevertheless, our scrutiny has brought to light several pivotal challenges impeding the seamless implementation of large-scale neural networks (NNs) on silicon. These challenges encompass the absence of automated tools, the need for multifaceted domain expertise, and the inadequacy of existing algorithms to efficiently partition and place extensive SNN computations onto hardware infrastructure. In this paper, we posit the development of an automated tool flow capable of transmuting any NN into an SNN. This undertaking involves the creation of a novel graph-partitioning algorithm designed to strategically place SNNs on a network-on-chip (NoC), thereby paving the way for future energy-efficient and high-performance computing paradigms. The presented methodology showcases its effectiveness by successfully transforming ANN architectures into SNNs with a marginal average error penalty of merely 2.65%. The proposed graph-partitioning algorithm enables a 14.22% decrease in inter-synaptic communication and an 87.58% reduction in intra-synaptic communication, on average, underscoring the effectiveness of the proposed algorithm in optimizing NN communication pathways. Compared to a baseline graph-partitioning algorithm, the proposed approach exhibits an average decrease of 79.74% in latency and a 14.67% reduction in energy consumption. Using existing NoC tools, the energy-latency product of SNN architectures is, on average, 82.71% lower than that of the baseline architectures.",
"conclusion": "5. Conclusions This paper introduces a comprehensive tool flow designed for the exploration and implementation of high-level NN architectures, mainly focusing on SNN models. This tool flow integrates the use of Python Keras libraries and the SNN-TB, along with our innovatively developed SNN-GPA. The SNN-GPA is instrumental in partitioning and positioning the SNN within an NoC architecture. The methodology proposed herein demonstrates remarkable efficiency in converting ANN architectures into SNNs, incurring an average error penalty of only 2.65%. Moreover, in comparison to a baseline model, the SNN-GPA significantly reduces synaptic communication weights, with an average reduction of 14.22% in inter-communication weights and 87.58% in intra-communication weights. This underscores the efficacy of the proposed algorithm in optimizing neural network communication pathways and emphasizes the effectiveness of the proposed approach in enhancing the operational efficiency of SNN models. In contrast to a baseline graph-partitioning algorithm, the suggested approach demonstrates an average latency reduction of 79.74% and a decrease in energy consumption of 14.67%. Using the proposed methodology, the synthetic synthetic_4k network exhibits a 97% reduction and the realistic Zambrano_mnist network exhibits a 94.17% reduction in the energy-latency product compared to the baseline model.",
"introduction": "1. Introduction Spiking neural networks (SNNs) [ 1 ] represent the vanguard in the evolution of artificial neural networks (ANNs), drawing inspiration from the intricate workings of biological organisms. SNNs offer several advantages and unique features compared to traditional ANNs, including biological plausibility, making them more biologically plausible than ANNs. They operate using spikes, similar to the firing of neurons in the human brain. In addition, SNNs are inherently event-driven, meaning they process information only when there is a change (spike). This event-driven nature can lead to energy-efficient computations, especially in applications where continuous processing is not necessary. SNNs also naturally capture temporal information through the timing of spikes. This is essential for tasks where the sequence and timing of events matter, such as in sensory processing or dynamic pattern recognition. Most importantly, the sparsity and binary nature of spikes in SNNs can lead to energy-efficient hardware implementations. This is particularly advantageous for applications in edge computing and IoT devices, where power consumption is critical to enable real-time processing. In contrast to ANNs, SNNs can exhibit robustness to input noise, as their spike-based processing can filter out irrelevant information. This can be beneficial in applications where input data may have inherent noise. However, the advanced neural network paradigm finds efficient implementation in neuromorphic platforms, which are characterized by manycore systems, wherein a predetermined quantity of neuronal computation is meticulously mapped to individual cores. Communication between these neurons, facilitated through synapses, is orchestrated using the network-on-chip (NoC) fabric—a quintessential design choice for engendering seamless communication within a multicore system. In the traditional SNN architecture, non-biological spiking neurons and grids, akin to the architecture of cache memory [ 1 ], are employed. The neuron undergoes a firing event, or spike production, immediately upon surpassing its action potential threshold, with the crossbars serving as repositories for synaptic weights [ 2 ]. The computational efficacy, gauged by execution latency and energy consumption, of an SNN-based computing system is contingent upon the judicious allocation of neurons to computing units (i.e., cores) with minimal communication latency. Nevertheless, the electrical constraints of the input load and output load impose limitations on the number of input–output connections per neuron, necessitating the incorporation of multiple crossbars through NoC architectures. In this context, extant algorithmic methods for mapping SNN unitary computational components to cores in a manycore system need more consideration for the underlying NoC models to ensure the attainment of optimal communication delay. Furthermore, our investigation has identified multiple pivotal challenges in designing large-scale SNNs on actual hardware systems. These challenges include (i) a dearth of comprehensive guidelines for constructing a software-level model translating to hardware deployment, (ii) the absence of design-automation devices and the imperative need for a breadth of domain expertise, and (iii) limitations in existing neuron clustering approaches, which are incapable of handling a large number of neurons in an SNN. This research addresses the aforementioned challenges by offering an existing graph-partitioning algorithm [ 3 ] and effectuating the placement of SNN architectures onto an NoC model, employing a methodology of a generic nature. In this manuscript, we address a significant limitation present in current graph-partitioning algorithms [ 1 , 4 , 5 ], specifically the constraint on the number of vertices, which typically remains below 10,000. We introduce our novel greedy graph-partitioning algorithm, which has the capacity to effectively manage graphs comprising over 100,000 vertices, thereby mitigating a substantial amount of communication overhead when integrated into crossbar hardware configurations [ 3 ]. In particular, the key contributions of this work are as follows: We introduce our novel design and automation methodology that systematically transforms any neural network architecture into an SNN for the purpose of optimizing energy efficiency in neuromorphic computations. We introduce our novel graph-partitioning algorithm devised for implementing extensive SNNs. We map partitioned SNN architectures to a state-of-the-art NoC tool flow to show the efficiency of the proposed methodology. We conduct benchmark assessments on diverse deep neural network (DNN) and convolutional neural network (CNN) architectures and seamlessly integrate multiple applications to demonstrate the efficacy of our tool flow. Compared to a baseline graph-partitioning algorithm, the proposed method showcases an average decrease of 79.74% in latency and a 14.67% reduction in energy consumption. Collectively, the proposed approach exhibits, on average, an 82.71% reduction in the energy-latency product compared to the baseline approaches.",
"discussion": "4. Experimental Results and Discussion 4.1. Experimental Setup In the preliminary phase of this study, an evaluation was conducted using a computing system equipped with a 32-core Intel Xeon Gold processor, complemented by 64 GB of RAM, and an NVIDIA Quadro P4000 GPU, operating under Ubuntu 18.04. The Python programming language was employed for the development of the tool flow. Our analysis incorporated both synthetic and realistic network models. Synthetic SNNs often involve simplified and abstract models of networks. These models capture the essential elements of spiking neurons but aim to avoid replicating the full complexity of complex NN architectures. In contrast, realistic SNNs aim to closely mimic actual NN architectures and feature extractors. This includes modeling detailed aspects of filter characteristics and modeling SNNs considering neuronal ion channels, neurotransmitter dynamics, and complex synaptic interactions. These networks are particularly useful in computational neuroscience for studying how real neural systems work. They help in forming hypotheses about neural computations and brain function. Regarding synthetic networks, we utilized a three-layer (fully connected) synthetic network, which encompassed 4K neurons and approximately 3.75 million synapses, referred to herein as the “synthetic_4k” network. For the examination of realistic networks, a variety of models were employed, including CNN_mnist [ 1 ], LeNet_mnist [ 45 ], Zambrano_mnist [ 46 ], Rueckauer_Cifar10 [ 30 ], LeNet_cifar10 [ 47 ], AlexNet_mnist [ 48 ], multilayer perceptron for mnist (MLP_mnist), a CNN for DogsVsDogs [ 49 ], a CNN for Fruits360 [ 50 ], and AlexNet_CatsVsDogs [ 48 ] for benchmarking. The benchmarks for this evaluation were conducted using several datasets, notably the mnist handwritten digit dataset [ 45 ], the Cifar10 dataset [ 51 ], and the CatsVsDogs dataset [ 49 ]. These datasets were instrumental in evaluating the performance and efficacy of the networks under study. For the final SNN implementation, we used an existing NoC simulator (i.e., Noxim [ 52 ]). 4.2. Implementation Results In the course of this research, each network was meticulously implemented using the TensorFlow framework and subsequently trained to utilize the Keras library. For the purpose of converting a CNN or ANN into a spiking neural network (SNN), we employed a modified iteration of the SNN Tool Box (SNN-TB) [ 30 ]. It is imperative to underscore the significance of weight normalization in CNNs to create accurate SNN models. A notable challenge encountered with the existing SNN-TB implementation is requiring all layer activations corresponding to the normalization dataset to be concurrently loaded into the GPU memory. This stipulation presents a considerable limitation, as large-scale models coupled with extensive datasets often exceed the memory capacity of most GPUs. To circumvent this limitation, we developed an innovative normalization workaround. While producing equivalent results, this solution reallocates the memory burden from the GPU to the system memory, thereby overcoming the previously mentioned constraint. The firing rate of neurons in an SNN determines the amount of information being transmitted through the network. It can affect the network’s ability to learn and process information. In many SNN learning algorithms, such as spike-timing-dependent plasticity (STDP), the timing and frequency of spikes are crucial for the adjustment of synaptic weights. The firing rate thus directly influences how learning occurs in the network. At the same time, the simulation time refers to the duration for which the SNN is simulated during the training process. It is crucial for allowing the network to learn from the temporal patterns in the input data. Figure 2 shows the SNN’s million operations (MOps) versus simulation time curve considering the CNN_mnist network. We can observe increased activity in neurons as the learning time increases. We employed Pearson correlation coefficients to ascertain the efficacy of the conversion process from ANNs to SNNs. This statistical measure was utilized to compare the activations in ANNs with the spike rates in SNNs. Figure 3 provides a graphical representation of the correlation coefficients for each layer of the CNN_mnist network [ 3 ]. This network was specifically trained using the mnist dataset, and the coefficients presented are the averages calculated over all the data batches [ 3 ]. Furthermore, this study meticulously tracked the progression of classification errors throughout the training phase across various communication periods. These data are visually represented in Figure 4 [ 3 ]. The simulation duration was measured in discrete steps of 1 millisecond. In the graphical depiction, the green scatter points denote the top-1 classification errors over time, whereas the blue ones denote the top-5 classification errors observed during the same period. Additionally, the shaded regions in the figure indicate the standard deviation in the classification errors observed in both SNNs and ANNs. Table 1 [ 3 ] presents a comparative analysis of the accuracy metrics for both the ANN and SNN models when applied to realistic networks. These networks utilized approximately 173.8 million synapses and 0.39 million neurons, on average. Furthermore, the mean spike count and the average simulation time, denoted as S T , were recorded as 8476.14 million and 320 s, respectively. The methodology proposed herein has demonstrated its efficacy by converting ANN architectures to SNNs with a minimal average error penalty of only 2.65%. To assess the effectiveness of our graph-partitioning methodology, we examined the synaptic weights, specifically I n t r a W and I n t e r W . This analysis encompassed evaluating synthetic and realistic networks utilizing the mnist dataset. Furthermore, an SNN architecture, specifically graph_edgedet, was employed for standard edge detection, as shown in Table 2 [ 3 ]. When applied to the Zambrano_mnist network, the proposed graph-partitioning algorithm demonstrated a reduction of 6.65% in inter-communication weights and an impressive reduction of 99.86% in intra-communication weights compared to a baseline model. In an overarching evaluation, the proposed SNN-GPA achieved a reduction of 14.22% in inter-communication weights and 87.58% in intra-communication weights compared to a baseline model. Upon completion of partition creation, the proposed methodology facilitated the placement of these neurons onto a designated NoC grid. We formulated a 2D mesh NoC architecture, adopting a grid length of 2 nm and employing a Cartesian coordinate system. Notably, this grid length is a design decision, and it can range from 10 μ m to several hundred micrometers [ 53 ]. For illustrative purposes, Figure 5 presents a representative diagram depicting the placement of the Zambrano_mnist network on a 120 × 120 mm chip using the proposed tool flow [ 3 ]. 4.3. Noxim NOC-Based Implementation Results In order to understand the effectiveness of the proposed methodologies, we used the Noxim [ 52 ] simulator. The Noxim simulator was developed using SystemC, which is a library written in C++. The Noxim runtime engine (NRE) is a cycle-accurate simulator that can execute various NoC architectural features and models. Noxim supports different topologies, buffer and packet sizes, traffic distributions, routing algorithms, packet injection rates, etc. The Noxim simulator uses Tile as its primary component, which comprises a processing element (PE), local computational memory, and a router. The PE is workload-dependent and primarily responsible for consuming and generating data packets. In our analysis, we used a mesh-based NOC architecture, which has better scalability and energy efficiency compared to shared bus-based architectures—the data packet travels through the router using the existing XY algorithm. In addition, Noxim permits a wormhole mechanism rather than the traditional store-and-forward mechanism for transferring flits from one router to another. In the wormhole mechanism, data packets are broken into smaller flits, which are then sent over the network in a wormhole fashion, whereas the conventional approach involves copying the entire data packet into the router before moving it to the next node. As a result, the wormhole approach enables resource sharing across multiple users. To demonstrate the efficacy of the proposed methodology, synthetic and realistic networks were employed, and network computations were mapped using the Noxim simulator. The results of this analysis are shown in Table 3 . The synthetic_4k network exhibits the highest latency improvement of 96.88% and an energy improvement of 3.8% compared to the baseline architectures. Among the realistic networks, the latency improvement ranges from 6.27% (for LeNet_mnist_padded) to 93.83% (for Zambrano_mnist), and the energy improvement ranges from 5.66% (for Zambrano_mnist) to 56.12% (for MLP_mnist) compared to the baseline architectures, as shown in Table 3 . Compared with a baseline graph-partitioning algorithm, the proposed approach demonstrates an average latency reduction of 79.74%. At the same time, the state-of-the-art SNN mapping algorithm [ 1 ] reported an average latency improvement of 45% compared to a baseline model. Using the proposed SNN-GPA algorithm and the Noxim tool led to a significant improvement in the energy-latency product. Figure 6 depicts the energy-latency product efficiency of the proposed algorithm. Using the proposed methodology, the synthetic synthetic_4k network exhibits a 97% reduction in the energy-latency product compared to the baseline model. Similarly, the Zambrano_mnist network exhibits a 94.17% reduction in the energy-latency product compared to the baseline model, with the highest energy-latency efficiency among the realistic networks. On average, the networks listed in Table 3 exhibit an 82.71% reduction in the energy-latency product compared to the baseline architectures."
} | 4,695 |
35505991 | PMC9043302 | pmc | 445 | {
"abstract": "Common mycorrhizal networks (CMNs) that connect individual plants of the same or different species together play important roles in nutrient and signal transportation, and plant community organization. However, about 10% of land plants are non-mycorrhizal species with roots that do not form any well-recognized types of mycorrhizas; and each mycorrhizal fungus can only colonize a limited number of plant species, resulting in numerous non-host plants that could not establish typical mycorrhizal symbiosis with a specific mycorrhizal fungus. If and how non-mycorrhizal or non-host plants are able to involve in CMNs remains unclear. Here we summarize studies focusing on mycorrhizal-mediated host and non-host plant interaction. Evidence has showed that some host-supported both arbuscular mycorrhizal (AM) and ectomycorrhizal (EM) hyphae can access to non-host plant roots without forming typical mycorrhizal structures, while such non-typical mycorrhizal colonization often inhibits the growth but enhances the induced system resistance of non-host plants. Meanwhile, the host growth is also differentially affected, depending on plant and fungi species. Molecular analyses suggested that the AMF colonization to non-hosts is different from pathogenic and endophytic fungi colonization, and the hyphae in non-host roots may be alive and have some unknown functions. Thus we propose that non-host plants are also important CMNs players. Using non-mycorrhizal model species Arabidopsis, tripartite culture system and new technologies such as nanoscale secondary ion mass spectrometry and multi-omics, to study nutrient and signal transportation between host and non-host plants via CMNs may provide new insights into the mechanisms underlying benefits of intercropping and agro-forestry systems, as well as plant community establishment and stability.",
"conclusion": "7 Concluding remarks and future perspectives Experimental data have directly proved that host-supported mycorrhizal fungi could penetrate or colonize non-host plant roots, without forming typical mycorrhizal structures. The colonization of non-host plant roots by mycorrhizal fungi differs to that with pathogenic or endophytic fungi, since similar early AM fungi-host recognition can be initiated normally, but plant defense responses could be triggered in a late stage. Moreover, the non-host plants are most likely to be adversely affected by host-supported mycorrhizal fungi and the growth of hosts are also variously impaired; the growth-defense tradeoffs alone could not explain these phenomenon sufficiently. AM and EM fungi show some different impacts on plant growth and nutrient acquisition in the tripartite system. These findings result in to a promising conclusion that non-mycorrhizal or non-host plants are also involved in CMNs ( Fig. 4 ), which would offer excite future studies to answer the following questions: 1. Can mycorrhizal fungi get and/or translocate carbon, nitrogen and other mineral nutrients from non-host plants through hyphal colonization? 2. Do host-supported mycorrhizal hyphae always negatively affect the growth of non-host plants? And for host plants? What's the mechanisms underlying the effects of the tripartite on host and non-host plant growth and nutrient acquisition? 3. How rhizosphere traits such as root exudates and rhizosphere microbiome changes in response to the host, mycorrhizal fungi and non-host plant tripartite? 4. Is there any molecular signal communication within the tripartite, such as microRNAs or peptides/proteins? 5. Can pest or pathogenic organisms induced warning signals transmit from host to non-host plants? And vice versa ? 6. Will intercropping of AMF crops and non-AMF crops or EM trees and crops lead to resistance to diseases but reduction of yield? 7. In natural ecosystems, are all plants connected by mycorrhizal fungi belowground? And what is the role of mycorrhizal fungi in shaping plant community? 8. Are there any new techniques, suitable plant species or plant cultural system that can be used to promote the research in this area? Fig. 4 A conceptual diagram showing potential common mycorrhizal linkages or networks among arbuscualr mycorrhizal (AM), ectomycorrhizal (EM) and/or non-mycorrhizal (NM) plants (Picture drew by Wenjun Xu, Kunming Institute of Botany, Kunming, China), as well as the potential functions of this kind of networks. Ectomycorrhizal (EM) trees or arbuscular mycorrhizal (AM) plants are connected by EM hyphae (red lines) or AM hyphae (blue lines), respectively (note: some trees can have dual EM and AM symbioses); and host supported both AM and EM hyphae may penetrate roots of non-mycorrhizal (NM) plants and/or their non-host plant species. Thus, all the plants could be connected underground via AM and/or EM fungi. Fig. 4 Currently, intercropping is attracting attention due to its ability to produce high yields at lower inputs and to suppress pests and diseases ( Li et al., 2020 ; Tang et al., 2021 ). Given the fact that a cereal and a legume is by far the most common intercrop combination worldwide ( Martin-Guay et al., 2018 ; Li et al., 2020 ), mycorrhizal effects on intercropping of a mycorrhizal (e.g. cereal or legume) and a non-mycorrhizal (e.g. canola) plant species with mycorrhizal fungus inoculation warrant attention. Understand the interaction between mycorrhizal and non-host plants will also benefit agro-forestry mixed system and underpin important drivers for plant community establishment under natural conditions. In future, growing non-mycorrhizal Brassicaceae species with mycorrhizal cereal species as their neighbors would be an excellent system to study the interactive impacts and mechanisms between host and non-host plants. Application of new techniques such as SIMS, stable isotope of C and N, and multi-omics (e.g. transcriptome, metabolome and proteome) will facilitate the research progress in this direction.",
"introduction": "1 Introduction About 80–90% of ≥10,000 plant species in terrestrial ecosystems have symbiotic relationships with mycorrhizal fungi to form mycorrhizas, which are beneficial to both the plant and fungi, thereby increasing their chances of surviving ( Wang and Qiu, 2006 ; Smith and Read, 2008 ; Brundrett, 2009 ), although only < 1.0% of all classified plant species have been evaluated for their mycorrhizal status ( Albornoz et al., 2021 ). At present six mycorrhizal types are categorized as arbuscular mycorrhiza, arbutoid mycorrhiza, ectomycorrhiza, ericoid mycorrhiza, monotropoid mycorrhiza and orchid mycorrhiza ( Smith and Read, 2008 ). Among them, the first and second common mycorrhizas, also the most economically important mycorrhizas in agricultural and natural ecosystems, are arbuscular mycorrhiza (AM) and ectomycorrhiza, which colonize ~80% and 2% of all tested plant species, respectively ( He et al., 2003 ; Wang and Qiu, 2006 ; Smith and Read, 2008 ; Brundrett, 2009 ). The mycorrhizas improve plant nutrient uptake, growth and yield, while the fungus receives photosynthetically assimilated carbon (C) from the associated mycorrhizal host plant (hereafter host) ( Smith and Read, 2008 ; Bonfante and Genre, 2010 ; Wagg et al., 2015 ). In nature, a plant may associate with multiple fungi and each fungal individual may associate with more than one plant, when mycorrhizal fungi connect individual plants of the same or different species together belowground, (common) mycorrhizal networks (CMNs) formed ( Pringle, 2009 ; van der Heijden et al., 2015 ). Through a CMN, C, nitrogen (N), and other nutrients can be transferred among individual plants, and plants can also use these networks to communicate the presence of pests and diseases, and to release chemicals that provide plants with a competitive advantage over other non-mycorrhizal plants ( He et al., 2003 ; Hoeksema, 2015 ; Wagg et al., 2015 ; Gilbert and Johnson, 2017 ). However, there are ~10% of land plants are non-mycorrhizal species ( Wang and Qiu, 2006 ; Smith and Read, 2008 ; Brundrett, 2009 ). Moreover, each mycorrhizal fungus has a limited number of host plant species, resulting in a large number of plant species that are non-mycorrhizal host plants for a specific mycorrhizal fungus. Currently if and how non-mycorrhizal or non-host plants are able to involve in CMNs remains unknown, while relevant studies can help to understand plant community establishment and stability in terrestrial ecosystems. Additionally, intercropping and agro-forestry systems become globally popular and thus AM and/or EM mediating host and non-host plant interactions warrant attention. Great progresses on host plants-mycorrhizal fungi interactions and their ecological functions have been made in the past two decades (e.g. He et al., 2003 ; Simard and Durall, 2004 ; Smith and Read, 2008 ; Wagg et al., 2015 ; Wipf et al., 2019 ). Regarding non-mycorrhizal or non-host plants, studies are mainly focused on their contribution to understand the mechanisms and evolution of mycorrhizal symbiosis (see reviews Giovannetti and Sbrana, 1998 ; Cosme et al., 2018 ). To date, studies on the mycorrhizal host, mycorrhizal fungus and non-mycorrhizal host tripartite relation (hereafter tripartite) have not attracted enough attention, leading to the effects of the tripartite interaction on plant growth and nutrient acquisition and their relevant underlying mechanisms remain largely unknown ( Fernandez et al., 2019 ). Currently, to our knowledge, from the point of view of the interaction between mycorrhizal and non-mycorrhizal plants, only two reviews have discussed the performance and coexistence of mycorrhizal and non-mycorrhizal species under extremes of nutrient availability, particularly under phosphorus deficiency ( Lambers and Teste, 2013 ; Lambers et al., 2018 ). In this review, we first define the concepts and scopes of non-mycorrhizal and non-host plants. Next, we summarize experimental data from published studies to discuss (1) if AM or EM hyphae could penetrate or colonize roots of non-host plants; (2) how plant growth and nutrient acquisition has been affected by host-mycorrhizal fungi-non-host tripartite; (3) molecular dialogue between AM hyphae and non-host model plant Arabidopsis thaliana ; (4) the induced system resistance (ISR) or induced system susceptibility (ISS) responses when non-host plant roots were colonized by mycorrhizal hyphae. Finally, we discuss future perspectives in this research area. The overarching goal is to review recent progress in mycorrhizas mediated host and non-host plant interaction, and discuss if non-host plants can be involved in CMNs."
} | 2,654 |
26388936 | PMC4573693 | pmc | 446 | {
"abstract": "Background Currently, hydrogen fuel is derived mainly from fossil fuels, but there is an increasing interest in clean and sustainable technologies for hydrogen production. In this context, the ability of some photosynthetic microorganisms, particularly cyanobacteria and microalgae, to produce hydrogen is a promising alternative for renewable, clean-energy production. Among a diverse array of photosynthetic microorganisms able to produce hydrogen, the green algae Chlamydomonas reinhardtii is the model organism widely used to study hydrogen production. Despite the well-known fact that acetate-containing medium enhances hydrogen production in this algae, little is known about the precise role of acetate during this process. Results We have examined several physiological aspects related to acetate assimilation in the context of hydrogen production metabolism. Measurements of oxygen and CO 2 levels, acetate uptake, and cell growth were performed under different light conditions, and oxygenic regimes. We show that oxygen and light intensity levels control acetate assimilation and modulate hydrogen production. We also demonstrate that the determination of the contribution of the PSII-dependent hydrogen production pathway in mixotrophic cultures, using the photosynthetic inhibitor DCMU, can lead to dissimilar results when used under various oxygenic regimes. The level of inhibition of DCMU in hydrogen production under low light seems to be linked to the acetate uptake rates. Moreover, we highlight the importance of releasing the hydrogen partial pressure to avoid an inherent inhibitory factor on the hydrogen production. Conclusion Low levels of oxygen allow for low acetate uptake rates, and paradoxically, lead to efficient and sustained production of hydrogen. Our data suggest that acetate plays an important role in the hydrogen production process, during non-stressed conditions, other than establishing anaerobiosis, and independent of starch accumulation. Potential metabolic pathways involved in hydrogen production in mixotrophic cultures are discussed. Mixotrophic nutrient-replete cultures under low light are shown to be an alternative for the simultaneous production of hydrogen and biomass. Electronic supplementary material The online version of this article (doi:10.1186/s13068-015-0341-9) contains supplementary material, which is available to authorized users.",
"conclusion": "Conclusion H 2 production by Chlamydomonas cultures is still far for large-scale implementations due to the low yields obtained so far. Among other limitations, the H 2 partial pressure is an important limiting factor, as shown in this work. Moreover, most of the studies about H 2 production on Chlamydomonas rely in the use of nutrient starved cultures, which reduces the viability of large-scale productions. The approach presented here, although also has very low H 2 production yields, opens up a new possibility to the straightforward attainment of both H 2 and biomass in non-stressed cultures. This could have a great biotechnological interest if current H 2 production yields are sufficiently improved. Another factor limiting H 2 production in algae is the O 2 sensitivity of the hydrogenases. However, an intriguing conclusion of this work is that low levels of O 2 can actually benefit H 2 production by the means of facilitating acetate uptake in mixotrophic cultures. However, the role of acetate during the H 2 photoproduction in Chlamydomonas cultures remains uncertain. We have shown that H 2 production in mixotrophic cultures is mostly associated with the PSII-independent pathway (80 %), indicating that a source of reductants is needed to feed the PQ pool. However, although acetate assimilation can favor starch accumulation, we clearly demonstrate that mobilization of the starch reserves does not provide with reductants for H 2 production under our experimental conditions. We discuss potential metabolic pathways that may be involved in H 2 production and linked to the dissimilation/assimilation of acetate, but certainly, more studies are needed to unravel the precise mechanisms that trigger H 2 production in the presence of acetate in Chlamydomonas cultures. Understanding such mechanisms could help us improving H 2 production efficiency through new physiological manipulations and genetic engineering.",
"discussion": "Discussion H 2 partial pressure critically inhibits H 2 production Hydrogenases have a reversible nature and are capable of both biosynthesizing H 2 and dissociating it into H + and electrons. This latter activity is commonly known as H 2 uptake. As any chemical reaction, the equilibrium between the biosynthesis and uptake of H 2 depends on the concentrations of substrates and products. This would imply that H 2 accumulation within bioreactors increases H 2 partial pressure in the headspace, while reducing H 2 biosynthetic rates to eventually cease any net hydrogenase bioproduction activity. Inhibition of the Chlamydomonas H 2 production by injection of different amounts of pure H 2 in the headspaces has previously been observed by others [ 21 , 27 ]. Moreover, Mignolet et al. [ 14 ] demonstrated how H 2 elimination from hermetic cultures by flushing with N 2 can enhance H 2 production. Finally, Kosourov et al. [ 27 ] also studied the effect of different volumes of purged headspaces on H 2 production in Chlamydomonas cultures under S- and phosphorous-deprivation. Here, we certainly confirm that the release of the H 2 accumulated in the headspace greatly enhances H 2 production, even when this release is accompanied by O 2 supplementation to the cultures (Fig. 2 a). Daily release of the H 2 accumulated in the headspaces of the bioreactors vessels, through either aeration, or purging with N 2 , resulted in H 2 production 2.4 and 3.4 times higher than in non-aerated cultures, respectively (Fig. 3 a). In our experimental designs, release of the H 2 partial pressure was executed every 24 h. Since non-aerated cultures did not evolve substantial H 2 after 24 h (Fig. 1 a), we conclude that under our nutrient-replete LL conditions, the equilibrium between H 2 production and H 2 uptake was reached before the 24-h aeration events. In fact, H 2 uptake can be observed after 24 h in LL, non-aerated cultures (Fig. 1 a). For dark non-aerated cultures, the equilibrium is reached after 7–8 days (Fig. 1 a). However, for both non-aerated LL and dark cultures, the maximal accumulation of H 2 in the headspaces reached about 1.9 % (0.76 ml H 2 out of 40 ml headspace). Hence, we assume that 1.9 % is near the maximal percentage of H 2 that can be accumulated under our specific experimental nutrient-replete conditions. Interestingly, in S-depleted, non-aerated cultures (Additional file 3 : Fig. S3A), the maximal H 2 accumulation was higher than in non-aerated S-replete cultures and reached 3.2 % of the headspace. This could indicate that different physiological conditions can probably alter the substrates availability of the hydrogenases (H + and electrons) exerting different equilibrium pressures towards H 2 formation. Overcoming the inhibition of H 2 accumulation within bioreactors must be dealt with to optimize H 2 production. Utilization of a larger gas to liquid phase ratios or ventilation systems that liberate H 2 partial pressure are straightforward ways to overcome the hydrogenase reversibility. However, this may lead the production of highly diluted H 2 , which may in turn reduce the potential for commercial applications. Utilization of bioreactors covered with materials, such as metallacarboranes, with high reversible capacity for binding H 2 at low temperatures [ 30 ] might be a way to avoid H 2 uptake. Production of both biomass and H 2 is possible in non-stressed Chlamydomonas cultures Hydrogenase sensitivity to O 2 is probably the main drawback for sustained H 2 photoproduction in photosynthetic organisms. Reduction of net O 2 levels can be accomplished by reducing PSII activity and maintaining high rates of mitochondrial respiration. At physiological level, nutrient stresses (in the presence of acetate) are widely employed strategies to establish this condition in Chlamydomonas. Sulfur limitation is the most widely employed [ 6 ], although phosphorous [ 31 ], nitrogen [ 32 ] and magnesium [ 33 ] deficiencies have been used successfully to attain H 2 production in Chlamydomonas. These approaches imply a two-stage process: first biomass is grown under aerobic conditions in nutrient-replete media, and then a H 2 production phase can be induced by removing nutrients and O 2 from the media. This procedure normally requires a solid–liquid separation step of biomass and culture medium after the growth phase, followed by several washing steps and O 2 purging processes [ 6 , 34 , 35 ]. Continuous or semi-continuous regimes of cultivation [ 36 – 38 ] or re-addition of nutrients to the medium [ 39 ] are alternatives to the solid–liquid separation steps, which extend the sustainability of H 2 production. However, all these protocols normally require high-energy inputs, do not support algae biomass and H 2 production simultaneously, are time-consuming and reduce culture viability. In this work, we show an alternative straightforward methodology for H 2 production in nutrient-replete media under LL. This methodology neither requires initial high cell density cultures nor the removal of nutrients from the media. Moreover, under LL and in the presence of acetate, the atmospheric O 2 is totally depleted after 24 h, which makes the purging process, required in other strategies, unnecessary. Finally, H 2 production can start within 24 h avoiding the typical lag phase (2–8 days) observed under nutrient-depleted cultures [ 15 , 29 , 31 – 33 ]. Overall, this strategy opens up new possibilities for production of both H 2 and biomass, the latter being important for further downstream biotechnological applications. Here, we demonstrated that Chlamydomonas, nutrient-replete, mixotrophic, cultures grown under LL (<50 PAR) can reach anoxia after 24 h and produce H 2 , highlighting the inverse relationship between H 2 production rates and light intensities. Similar results have been previously reported by Degrenne et al. [ 26 ] who studied the induction of anoxia and H 2 production in mixotrophic Chlamydomonas cultures using batch and continuous mode bioreactors under different light regimes. Here, we have extended these previous studies using cultures under different light intensities and oxygenic regimes. We have described in detail how light intensity, O 2 availability, acetate uptake, and H 2 production are processes closely linked to each other. Here, we have shown that O 2 allows acetate uptake and cell growth. Providing O 2 to batch mode cultures under LL allowed us to achieve simultaneously H 2 production and biomass (Fig. 2 a; Additional file 1 : Fig. S1). Moreover, in aerated fed-batch mode bioreactors, non-continuous supplementation with acetic acid and O 2 greatly enhanced H 2 production (Fig. 4 a). Likely continuous mode bioreactors, with continuous supply of acetic acid and low levels of O 2 (plus other basic nutrients), would allow for sustained H 2 and biomass production. In our study, maximal H 2 production was obtained when cultures under 12PAR were not aerated but N 2 -purged daily. The purging procedure allowed prolonged H 2 production (10 days in purged cultures vs 3 days in aerated cultures) (Fig. 3 ) with a substantial reduction of the acetate consumption, which is of interest for potential large-scale productions. However, N 2 -purged cultures did not grow and biomass did not increase, likely because O 2 levels were too low. Optimization of these procedures must be achieved to obtain both maximal H 2 and biomass production. For example, purging gases that include low concentration of O 2 can be finely tuned to allow for high H 2 production rates and cell growth. Roles of acetate and oxygen in the production of H 2 under LL in nutrient-replete cultures Most of the studies related to H 2 production in Chlamydomonas have been performed with S-depleted acetate-containing media [ 6 , 10 , 34 , 40 ]. The goal of these studies, however, did not focus on identifying the role of acetate during H 2 production. When autotrophic Chlamydomonas cultures are used to produce H 2 in either nutrient-replete [ 41 ] or S-depleted media [ 29 , 35 , 42 ], H 2 production is low compared with media containing acetate, which highlight the importance of acetate in the context of H 2 production under these conditions. From studies using nutrient-replete and S-depleted media, it is widely assumed that acetate facilitates H 2 production by helping to establish anoxia in the cultures through its oxidation via the Tricarboxylic Acid Cycle (TCA) and oxidative phosphorylation [ 6 , 17 , 21 , 26 , 34 , 43 ]. The addition of acetate to Chlamydomonas cultures decreases photosynthesis efficiency, net O 2 evolution, and CO 2 fixation, as well as promotes the transition from state I to state II and mitochondrial respiration [ 44 – 47 ]. All these factors help to establish anoxia in sealed cultures. Repression of CO 2 fixation may also contribute to H 2 production by reducing the competition for electrons, for HYDA1, at the level of FDX1 [ 10 ]. In any case, the role of acetate in stimulation of H 2 production in the light has not yet been carefully examined. We have shown that acetate uptake rates are not directly proportional to H 2 production rates (e.g., purged cultures vs aerated cultures, Fig. 3 a, b). Indeed, the faster the acetate is consumed, the faster O 2 accumulates (once acetate is depleted from the media) and the faster H 2 -production phase is inhibited (e.g., Fig. 2 a–c). On the other hand, we have shown that acetate uptake is highly dependent on O 2 availability (e.g., comparing aerated and non-aerated cultures with DCMU, Figs. 1 g, 2 g). Under severe anoxic conditions, such as in non-aerated cultures with DCMU or daily purged cultures with DCMU, there is no acetate uptake and H 2 production is very limited (Figs. 1 g, e, 3 a, b). This suggests that a minimal threshold of acetate uptake is needed to support H 2 production and that the role of acetate must rely on something else than just the establishment of anoxic condition. Moreover, paradoxically, O 2 is a key substrate that under limiting concentrations allows acetate uptake and thereby H 2 production. Acetate uptake does not directly depend on the PSII activity per se (Fig. 2 g); however, PSII activity is crucial to provide O 2 to hermetically sealed cultures and allow acetate uptake (e.g., Fig. 1 c). Overall, our results suggest that an important aspect to be considered when producing H 2 in mixotrophic Chlamydomonas cultures is that optimal H 2 production does not rely on reaching very high respiration rates or establishing severe anoxic conditions but rather on sustaining relatively low acetate uptake rates under hypoxic conditions. LL is an optimal physiological condition where O 2 production is high enough to allow low acetate uptake rates but low enough to prevent hydrogenase inhibition. The H 2 production observed in acetate-containing media is clearly enhanced in the light, although, under our experimental conditions, dark fermentation may contribute to about 10–20 % of the total H 2 production detected at 12PAR. The stimulatory effect of light on H 2 production necessarily requires an active photosynthetic electron flow that can feed electrons to HYDA1 via either the PSII-dependent or PSII-independent pathway. The inhibitory effect of DCMU on algal H 2 photoproduction has frequently been used to differentiate between PSII-dependent and -independent pathways [ 7 , 10 , 48 ]. Previously, it was observed that the effect of DCMU on cells grown in nutrient-replete media was unclear. Some studies reveal no impact of DCMU on H 2 photoproduction [ 49 ]. However, others observed substantial inhibition of both H 2 production (90 %) and acetate uptake (66 %) [ 21 , 24 ]. Here, we demonstrate that the addition of DCMU to nutrient-replete, acetate-containing cultures can have very different effects on both H 2 production (e.g., at 12 PAR, inhibition of 72.4 % in non-aerated cultures vs 19.2 % in aerated cultures) and acetate uptake (e.g., at any light intensity, severe inhibition in non-aerated cultures vs slight inhibition in aerated cultures). The level of inhibition of H 2 production by DCMU is subjected to the specific culture conditions, more specifically to O 2 availability and acetate uptake. This may be problematic for determining the contribution of the PSII-dependent pathway on cells grown in media containing acetate. Experimental conditions (e.g. O 2 availability, pre-culture conditions, purging processes, etc.) may affect the effect of DCMU addition on H 2 production and in turn overestimate the contribution of the PSII-dependent pathway. This may partly explain some discrepancies found in the literature concerning the relative importance of this pathway. Nonetheless, we conclude that H 2 production in LL mixotrophic cultures is mainly via PSII-independent pathway, since up to 80 % of the maximal H 2 production can be obtained when PSII is inhibited (Figs. 2 e, 4 e). It has been proposed that acetate may enhance the PSII-independent H 2 production pathway under nutrient-replete media by favoring the accumulation of starch during oxygenic conditions, which is degraded later during anaerobiosis [ 17 , 24 , 49 , 50 ]. However, our data do not seem to support this possibility since we observe no correlation between starch accumulation/degradation patterns and H 2 production under our experimental conditions (Figs. 3 a, 5 ). Our data are in agreement with previous observations of starch-deficient mutants that can produce H 2 under both S-replete and S-depleted conditions [ 10 ]. Alternatively, starch degradation may not be the source of reductive equivalents but rather acetate assimilation may provide the chloroplast with reductive equivalents for PSII-independent H 2 photoproduction. The localization of two enzymes involved in acetate assimilation, succinate dehydrogenase (SDH) and malate dehydrogenase (MDH), in the chloroplast of Chlamydomonas, may provide this organelle with the reductive equivalents (FADH 2 and NADPH, respectively) needed to feed the PQ pool [ 51 , 52 ] (Fig. 6 ). Additionally, the oxaloacetate produced in the chloroplast from MDH can also potentially contribute to H 2 production. Recently, it was shown that the pyruvate:ferredoxin oxidoreductase (PFR) enzyme of Chlamydomonas possesses an important affinity for oxaloacetate [ 53 ], which opens up the possibility that acetate, via the glyoxylate pathway, might also be coupled to fermentative H 2 production (Fig. 6 ). Fig. 6 Tentative model for H 2 production pathways in acetate-containing cultures under LL. Black dashed arrows multiple enzymatic steps. Red dashed arrow electrons flow. DCT dicarboxylate transport system, Cytb6f cytochrome b6f, HYDA1 hydrogenase 1, FDX1 ferredoxin 1, MDH malate dehydrogenase, NDA2 NAD(P)H:plastoquinone oxidoreductase 2, PFR pyruvate:ferredoxin oxido reductase, PSI photosystem I, PQ plastoquinone, SDH succinate dehydrogenase, TCA tricarboxylic acid cycle In any case, under our experimental conditions, the potential contribution of these pathways (either PSII-independent via SDH/MDH or fermentative via PFR) needs to be necessarily linked to light-dependent reactions since dark H 2 production is clearly lower than in the light. For the SDH/MDH pathway, it is easy to explain the light dependence since electron flow from the PQ pool to HYDA1 would require an active PSI. At the same time, an active PSI would promote PQ re-oxidation and would allow the turnover of FAD + and NADP + needed to maintain SDH and MDH activities, respectively, which in turn would provide oxaloacetate to the chloroplast, leading to high PFR rates under anoxia. Still, the relative importance, if any, of these potential pathways for H 2 production in nutrient-replete acetate-containing cultures, remains to be determined appropriately. Enzyme localization and proteomics studies have helped in identifying the acetate assimilation and dissimilation pathways, which require the participation of enzymes localized in the cytosol, the mitochondria, and the chloroplast [ 51 , 52 , 54 – 56 ]. Based on these studies and on the physiological data we have provided in this study, we propose a tentative model that aims at explaining the physiology and metabolic pathways involved in H 2 production in nutrient-replete, acetate-containing medium (Fig. 6 ). Acetate can be simultaneously routed to different pathways. Acetate can be dissimilated in the TCA cycle or assimilated by the glyoxylate cycle. The acetate entering the TCA cycle would provide energy to the cell and would contribute in maintaining low O 2 levels. The glyoxylate cycle would provide succinate to the chloroplast for carbon skeletons. Chloroplastic SDH and MDH would participate in the conversion of chloroplast succinate into oxaloacetate and would provide reductive equivalents to the PQ pool. Chloroplastic oxaloacetate would either be redirected towards different biosynthetic pathways (e.g., lipids or starch biosynthesis) or used by chloroplast PFR if anoxic/hypoxic conditions are established. All these pathways occur simultaneously and their relative activities can be affected by O 2 availability. Light intensity greatly affects O 2 availability through PSII activity, and this O 2 could be used by the TCA cycle to dissimilate acetate. If the light intensity is not too high, then the TCA cycle would be able to maintain the cells under hypoxic conditions. However, even though hypoxic conditions are established, O 2 photoproduction rates would regulate the TCA cycle activity, acetate uptake rate, and cell growth. The higher these parameters, the lower would be the H 2 production. Under LL, however, O 2 availability is very low and so would be the TCA activity, acetate uptake rates, and ATP generation. Under this scenario, the relative importance of the glyoxylate cycle is enhanced, as well as chloroplastic SDH and MDH activities, which could provide reductive equivalents to the PQ pool. Also, due to low ATP levels, the use of oxaloacetate for biosynthetic pathways and cell growth is impaired and PFR activity is favored. All these processes would favor the H 2 production under LL in acetate-containing medium. Under severe anoxic conditions, no acetate would be taken up, and H 2 production would be low. Finally, in the dark, and independent of O 2 availability, PSII-independent H 2 production is not active, and the turnover of FADH 2 and NADPH would limit oxaloacetate availability and its use by PFR."
} | 5,760 |
34653368 | null | s2 | 447 | {
"abstract": "For microbiome biology to become a more predictive science, we must identify which descriptive features of microbial communities are reproducible and predictable, which are not, and why. We address this question by experimentally studying parallelism and convergence in microbial community assembly in replicate glucose-limited habitats. Here, we show that the previously observed family-level convergence in these habitats reflects a reproducible metabolic organization, where the ratio of the dominant metabolic groups can be explained from a simple resource-partitioning model. In turn, taxonomic divergence among replicate communities arises from multistability in population dynamics. Multistability can also lead to alternative functional states in closed ecosystems but not in metacommunities. Our findings empirically illustrate how the evolutionary conservation of quantitative metabolic traits, multistability, and the inherent stochasticity of population dynamics, may all conspire to generate the patterns of reproducibility and variability at different levels of organization that are commonplace in microbial community assembly."
} | 285 |
39702455 | PMC11659276 | pmc | 450 | {
"abstract": "Coral thermotolerance research has focused on the ability of coral holobionts to maximize withstanding thermal stress exposure. Yet, it’s unclear whether thermal thresholds adjust across seasons or remain constant for a given species and location. Here, we assessed the thermal tolerance thresholds over time spanning the annual temperature variation in the Red Sea for Pocillopora verrucosa and Acropora spp. colonies. Utilizing the Coral Bleaching Automated Stress System (CBASS), we conducted standardized acute thermal assays by exposing corals to a range of temperatures (30 to 39 °C) and measuring their photosynthetic efficiency ( F v /F m ). Our results reveal species-specific thermal tolerance patterns. P. verrucosa exhibited significant seasonal changes in their thermal thresholds of around 3 °C, while Acropora spp. remained rather stable, showing changes of around 1 °C between seasons. Our work shows that thermal thresholds can vary with seasonal temperature fluctuations, suggesting that coral species may acclimate to these natural temperature hanges over short periods in a species-specific manner.",
"introduction": "Introduction Coral reefs are one of the most diverse and valuable ecosystems on Earth, providing significant ecological, economic, and societal benefits 1 , 2 . Coral reefs have rapidly declined worldwide due to climate change and human activities, resulting in reduced ecosystem services 3 . Recent predictions highlighted that continued warming could lead to significant coral loss 4 . Corals rely on the symbiotic relationship with the dinoflagellate photosynthetic microalgae of the family Symbiodiniaceae 5 , 6 . The microalgae cells live inside the coral host and provide the required energy to build the calcium carbonate skeletons, exerting a crucial role in the holobiont 5 . Prolonged exposure to elevated temperatures can break this relationship, leading to bleaching (i.e., the loss of the coral-associated Symbiodiniaceae 7 , which is also followed by the loss of the algae pigments and is indicative of coral health decline) and, ultimately, coral death 8 . Marine heatwaves, characterized by periods of extremely high sea surface temperature that can span thousands of kilometres, have become more frequent, contributing to four global coral bleaching events 9 . These conditions challenge corals’ ability to maintain their symbiotic relationship, often resulting in widespread coral mortality 10 . Hence, efforts are underway to identify coral species and populations with enhanced heat stress resistance for restoration and conservation efforts 5 , 11 . Understanding how corals respond to temperature changes, as well as their capacity for long-term acclimation and adaptation, is essential for developing effective strategies to mitigate the impacts of these increasingly common marine heatwaves. The thermal threshold of corals (the maximum temperature that a coral can tolerate) has been studied for a variety of species across many locations and appears to commonly align with local long-term thermal summer maxima 11 , 12 . Thus, while absolute thermal thresholds differ, corals typically exhibit bleaching under prolonged heat stress exceeding only 1–2 °C above their maximum summer mean temperature (i.e., MMM) 13 , 14 . The exposure to repeated warming events (e.g. heatwaves) may enhance thermal tolerance, improving projections of coral survival in successive warming events 15 , 16 . In addition, the variability of thermal regimes experienced by corals was shown to affect their thermal tolerance so that higher daily or seasonal thermal fluctuations positively affect thermal resilience and thus may provide a higher adaptive capacity of tolerance to heat stress 17 , 18 . In recent years, the development of standardized short-term acute heat stress assays 18 has allowed the experimental assessment of coral thermal tolerance (and resilience), as well as their connection to the prevailing local and/or regional thermal regime 19 . Coral physiology (e.g. photosynthetic efficiency, calcification, symbiont density) has been demonstrated to vary seasonally 20 – 23 . However, the extent of seasonal variation has been shown to be species-specific 20 . More importantly, whether physiological acclimation is affecting seasonal thermal thresholds, in particular in the framework of standardized stress testing (in the form of acute thermal assays, e.g. CBASS) is currently unresolved. Such information may also aid to better predict responses to changes in temperature due to climate change 24 , 25 . Thus, it is key to assess whether coral thermal tolerance thresholds change over time, based on seasonal thermal fluctuations, and how these changes potentially affect the capacity of corals to respond to changes in temperature. The Red Sea is considered an underwater laboratory because of its particular features and extreme thermal regimes experienced by the organisms inhabiting it. Corals in the central Red Sea are exposed to temperatures exceeding 30 °C in summer, while the coldest temperatures recorded are around 24 °C 18 , 26 , 27 . The natural thermal variation experienced by corals in this region suggests mechanisms of acclimation and/or adaptation for the different species to cope with this fluctuation year-round. Here, we experimentally assess the thermal tolerance threshold of two coral species using standardized acute thermal stress assays 18 , complemented by in situ monitoring of the photochemical efficiency of the coral-associated Symbiodiniaceae community (F v /F m ) as a response to seasonal changes in temperature in the central Red Sea.",
"discussion": "Results and discussion Colonies of P. verrucosa and Acropora spp. were examined during different periods of the year, encompassing all four seasons in Al Fahal Reef (central Red Sea) to determine whether coral thermal threshold varies as a function of the in situ temperature. The in situ photosynthetic efficiency (F v /F m ) of both species varied significantly with seawater temperature (Fig. 1 , Supplementary Table 1 ), exhibiting an unimodal thermal response curve. The maximum F v /F m performance for both species was recorded between November and December 2021 when in situ daily mean seawater temperatures were 28-29 °C, in agreement with previous results from the same region 26 , 28 , 29 . Minimum F v /F m values were recorded during the temperature peak in the summer of 2021 and 2022, when seawater temperatures were above 31 °C. The thermal optima ( T opt ) was 28.62 °C (95% CI [28.49, 28.73]) for Acropora spp. and 28.60 °C (95% CI [28.25, 28.79]) for P. verrucosa , with no statistically significant differences in the thermal performance curve between both species (Supplementary Table 1 ). The overall T opt estimated considering the combined thermal performance curves for P. verrucosa and Acropora spp. was 28.36 °C (95% CI [28.18, 28.58]). The trend of high F v /F m values in winter and low F v /F m in summer has also been reported previously for colonies of Stylophora pistillata and Favia favus from the Gulf of Aqaba 22 . In this case, the temperature at which the maximum performance was measured was around 23-25 °C, lower than that reported in our study. This difference is not surprising due to the lower ambient seawater temperatures in the northern Red Sea compared to the central Red Sea 11 . Fig. 1 Relationship between in situ photosynthetic efficiency ( F v /F m ) and in situ mean seawater temperature for Acropora spp. and P. verrucosa . The transparent points represent coral colonies. Bright colored points represent the mean and error bars show the standard error of the mean (s.e.m). Lines denote the thermal performance curve estimated by the Sharpe-Schoofield model and the shadow area around the fitting line is the 95% confidence interval estimated using bootstrapping method. The number of experiments included were n P.verrucosa = 12 and n Acropora spp . = 7. The number of biological replicates included in each experiment is given in Supplementary Table 6 . To assess species-specific differences in thermal tolerance over seasons, we experimentally determined the thermal thresholds for each species at different sampling periods. Thermal thresholds were estimated by calculating the temperature at which the photosynthetic efficiency rate ( F v /F m ) is reduced to 50% (based on the ecotoxicological term ED50) 18 , 30 . ED50 estimated using photosynthetic efficiency rates ( F v /F m ) has been demonstrated as a standardized proxy for coral thermotolerance and it captures differences in the susceptibility of corals to bleach 18 . We determined the thermal threshold of two different species at different times of the year including in situ temperatures ranging from 24.5 to 32 °C. We observed fluctuations in the estimated ED50 values during the various sampling periods. The shape of the performance curves based on experimental F v /F m rates over assay temperatures differed markedly between the two coral species, indicating a species-specific response to increasing temperature (Fig. 2 ). P. verrucosa exhibited different thermal response curves at different sampling times: corals sampled during the coldest period exhibited a gradual and continuous decrease in F v /F m at increasing experimental temperatures, whereas corals sampled in the warmer period maintained their F v /F m values up to a threshold and then dropped drastically (Fig. 2 ). Differences in the shape of the thermal performance curves, similar to those observed here for P. verrucosa colonies, have been previously reported 11 . Although our study cannot pinpoint the exact mechanisms behind the changes in the shape of the performance curves, it appears to be related to the colonies’ exposure to a gradual increase in temperature over time. The ED50 values estimated for P. verrucosa increased from 35.42 ± 0.53 °C in colder months to 38.10 ± 0.54 °C in warmer periods (August 2021), corresponding to the peak in seawater temperature. Fig. 2 Thermal performance curves for each CBASS experiment. Changes in F v / F m with assay temperature for each CBASS experiment performed using P. verrucosa ( A ) and Acropora spp. ( B ). Dots represent the relationship between F v /F m and experimental temperature for each CBASS run. Lines reflect the log-logistic model fitted to each experiment ( n P. verrucosa = 10 & n Acropora spp = 5). Color code represents in situ temperature at the time of sample collection for the CBASS experiment. The dots’ shape and lines represent the different seasons: winter (January to March), Spring (April to June), Summer (July to September) and Autumn (October to December). The details on the biological replicates sampled in each experiment are given in Supplementary Table 6 . On the other hand, Acropora spp. had a more stable pattern, sustaining F v /F m to a threshold before exhibiting a sharp decline (Fig. 2 ). The fact that different corals exhibit different shapes in their thermal performance curves after being exposed to thermal stress might suggest differential susceptibility to warming 11 . The ED50 range for Acropora spp. (36.76 ± 0.68 °C to 37.73 ± 1.02 °C) was narrower (~1 °C) than that of P. verrucosa (~3 °C), despite both being exposed to similar seasonal temperature ranges ( P. verrucosa : 24.5-32 °C, Acropora spp.: 24.5-31.4) and experimental conditions (30–39 °C). A quantitative comparison of thermal thresholds indicated that ED50 of P. verrucosa significantly differed between warmer and colder periods, while Acropora spp. showed no significant changes (Table 1 ). Differences in the thermal threshold of around 1 °C between summer and winter seasons for Pocillopora damicornis have been previously reported using a different experimental approach, consisting on 5-day tank experiments using corals located on the Great Barrier Reef 31 . However, here we observe differences up to 3 °C using a standardized comparison. These findings suggest that different acclimation mechanisms to environmental conditions may occur in the same area, depending on the species and even different genotypes, which is evidenced by the high variability observed between biological replicates (Supplementary Figs. 1 and 2 ). Table 1 Experimentally estimated thermal threshold for each species at each sampling time Exp_ID EnvT N ED50 sd se ci Letters Species CRG0(10) 31.42 21 37.38 1.54 0.34 0.70 bc P.verrucosa MEE1(13) 29.09 5 36.01 1.08 0.48 1.33 abc P.verrucosa MEE2(14) 24.53 5 35.55 0.75 0.33 0.93 ab P.verrucosa PvN1(1) 31.97 15 38.10 0.54 0.14 0.30 c P.verrucosa PvN2(2) 29.84 15 36.62 1.10 0.28 0.61 abc P.verrucosa PvN3(3) 28.72 15 36.27 0.66 0.17 0.37 ab P.verrucosa PvN4(4) 28.01 15 35.42 0.53 0.14 0.30 a P.verrucosa ReI1(5) 30.29 10 37.11 1.70 0.54 1.21 abc P.verrucosa ReI2(6) 30.15 10 36.97 1.50 0.47 1.07 abc P.verrucosa ReI3(7) 28.69 10 35.83 1.05 0.33 0.75 ab P.verrucosa CRG0(10) 31.42 21 37.45 0.65 0.14 0.30 Acropora spp MEE1(13) 29.09 5 37.23 0.99 0.44 1.22 Acropora spp MEE2(14) 24.53 5 37.25 0.64 0.29 0.80 Acropora spp ReE0(8) 29.99 8 37.73 1.02 0.36 0.86 Acropora spp ReE1(9) 28.69 8 36.76 0.68 0.24 0.57 Acropora spp This table shows the mean thermal threshold (ED50, °C) of each species and experiment including all biological replicates, in situ temperatures (EnvT, °C), N of biological replicates ( N ), standard deviation (sd), standard error of the mean (se) and 95% confidence interval (ci). Letters represent differences between sampling experiments for P. verrucosa (different letters show significant differences, Dunn Test: p .adj < 0.05). The metadata associated with each experiment is available in Supplementary Table 6 . To investigate whether prevailing seawater temperature had an effect on the observed variability of the coral thermal thresholds, we studied the relationship between the changes in ED50 and the average seawater temperature measured in situ at the sampling time. P. verrucosa thermal threshold values (ED50) showed a flat response until T opt was reached (28.36 °C), followed by a steep linear correlation with the increase of the seawater temperature (Fig. 3A , Supplementary Table 2 , linear mixed-effect model: p < 0.001). Although the pattern was similar to P. verrucosa , no significant increase was observed in the ED50 for Acropora spp. (Fig. 3B , Supplementary Table 3 , linear mixed-effect model: p = 0.08), suggesting a more stable thermal threshold across different seasons and a different species strategy to respond to temperature change. Some variation in thermal tolerance has been previously reported for P. damicornis 31 ; however, establishing a clear relationship with the in situ temperature has not been previously possible due to the lack of year-round coral monitoring data. Although the identification of the mechanism driving the differential response between species is out of the scope of this work, the seasonal changes observed here suggest that some coral species (e.g. genus Pocillopora ) might be capable of acclimatizing to gradual changes in temperature in the short-term. Fig. 3 Linking the thermal tolerance threshold of the corals with in situ performance and seawater temperature. Changes in the thermal tolerance threshold (ED50) with seawater temperature at the time of sampling for P. verrucosa ( A ) and Acropora spp. ( B ). The gray dots correspond to the individual colonies, and the fitting line is the result of the linear regression considering in situ temperature above T opt . The gray area represents the 95% confidence interval (CI). Full dots correspond to the mean ED50 of each experiment and the error bars to the standard error of the mean (s.e.m.). We found that ED50 was positively correlated with in situ temperature (Linear mixed-effect model, p < 0.001) for P. verrucosa , while this relationship was positive but not significant for Acropora spp (Linear mixed-effect model, p = 0.08) (Supplementary Tables 2 and 3 ). Relationship between in situ photosynthetic efficiency, F v /F m , and ED50 for P. Verrucosa ( C ) and Acropora spp. ( D ). The gray dots correspond to the individual replicates corresponding to the experiments performed at in situ temperatures above the T opt of the coral. The solid line corresponds to the linear model fitting, and the gray area represents the 95% CI. The full dots represent the average ED50 and F v /F m and the error bars correspond to the s.e.m of each variable. We found that ED50 was negatively correlated with in situ F v /F m in P . v errucosa (Linear mixed-effect model, p = 0.01), but not in Acropora spp (Linear mixed-effect model, p = 0.94) (Supplementary Table 4 and 5 ). The number of experiments included were n P.verrucosa = 8 and n Acropora spp . = 4. The number of biological replicates included in each experiment is given in Supplementary Table 6 . The species included in this study exhibited different experimentally derived standardized thermal tolerance thresholds (ED50s), surpassing (~7 °C) the maximum monthly mean (MMM) of the region (MMM AL Fahal reef = 30.75 °C) in summer. This offset between ED50 and local MMM is typically observed and aligns with what has been previously shown in the area 11 , 30 , 32 . Importantly, the fact that the thermal threshold is not a fixed value (for a given species and location), underlines the significance of characterizing seasonal variations in coral thermal tolerance. This understanding aids in predicting responses to temperature changes throughout the year, which is crucial when abrupt temperature shifts occur (e.g. heatwaves events). Some coral species, like P. verrucosa , show a gradual thermotolerance increase, aiding to cope with summer heat (~32–34 °C) and adapting to seasonal temperature variations, including winter cold stress (~20–24 °C) 27 , 29 . The decline in P. verrucosa thermal threshold during winter may be an acclimation strategy against cold stress. While cold stress data are limited compared to warm stress, corals have shown paling during winter months when temperatures drop below thermal optima 27 , 29 . Our findings suggest that the short-term thermal threshold flexibility of P. verrucosa could offer an advantage for coping with temperature changes. We then explored the link between thermal thresholds (ED50) and in situ F v /F m . Despite a similar performance of the studied coral species (Fig. 1 ), we observed a significant inverse correlation between P. verrucosa in situ F v /F m and estimated ED50 (Fig. 3C , Supplementary Table 4 , linear mixed-effect model: p = 0.01), while Acropora spp. exhibited no correlation (Fig. 3D , Supplementary Table 5 , linear mixed-effect model: p = 0.94). These results, in agreement with the previous results reported here, suggest distinct strategies for coping with changing environmental conditions. The flexibility of P. verrucosa might stem from modifying thermal thresholds in response to rising temperatures, potentially due to metabolic compensation processes 33 , indicating short-term acclimation to temperature. Differential responses to disparate environmental conditions have been shown in Pocillopora and Acropora genera 34 . Additionally, Acropora spp. has been shown to exhibit higher susceptibility after surpassing their thermal threshold 12 , as was also evident in our study from the marked drop in the dose-response curve (Fig. 2 ). The recovery after bleaching has been observed to be faster in P. verrucosa compared to Acropora spp. 12 . The species-specific responses reported here could also be aligned with the previously observed differences in microbiome flexibility between Pocillopora and Acropora species 25 , 34 . Pocillopora spp. maintain a stable microbiome, even under stress, whereas Acropora spp. harbor a more flexible bacterial community in response to various impacts, sites, and environmental conditions. Differences in microbiome flexibility could potentially readjust species-specific performance traits (growth, respiration, photosynthetic efficiency) to better cope with seasonal temperature changes. The reduction in coral performance (photosynthetic efficiency) during acute stress assessments has been correlated with in situ warming events and local temperature oscilation 18 , 19 . In addition, F v /F m derived ED50 has been demonstrated to reveal differences in the thermal threshold of the coral colonies even at a genotypic level using short-term acute stress experimental levels and be comparable with long-term heat stress experiment 11 , 18 , 30 . However, other physiological parameters may also be suitable for consistently comparing coral performance and thermal tolerance, which is still under study. Similarly to the rest of the performance rates, F v /F m rates are likely to be affected by temperature changes in an unimodal way, as we observe here and in other studies 31 . The performance increases up to an optimum and then exhibits a rapid decrease once the optimum has been surpassed. Therefore, it is not surprising that precisely above the T opt we observe a stronger relationship between thermal threshold and in situ temperature. The decrease in the performance observed in this part of the performance curve associated with the increase in the thermal tolerance of P. verrucosa species is indicative of a possible re-adjustment in the metabolic fluxes to compensate for the energy needed to acclimatize to the in situ environmental variation increasing the thermal threshold accordingly. These compensation mechanisms have been previously reported in microbes 33 , 35 and are likely playing a role here. However, more specific analyses are required to confirm the mechanisms. Changes in abiotic conditions other than seawater temperature, such as light intensity and water quality, have also been reported to affect the performance of the symbionts 21 , 22 . While the effects of light cannot be entirely dismissed in our observations, the strong correlation with temperature suggests it is the primary driver of the patterns we have observed. Although our study did not explore the underlying mechanisms, we hypothesize that coral species may have evolved to cope with environmental stress. These strategies are likely linked to microbial flexibility, and short-term acclimation patterns, which are significant for their ability to adapt to climate change. This study highlights the utility and the need of using standardized methods, such as CBASS, to identify which coral species or genotypes exhibit acclimation to seasonal variations 12 , 18 , 30 . The ability of a coral species or genotype to acclimate seasonally may differ depending on their physiological and genetic underpinnings, in addition to Symbiodiniaceae and bacterial assemblage 25 . These variables have also been demonstrated to differ between heat tolerant and heat susceptible colonies and their capacity to recover 36 . Therefore, the identification of coral species or genotypes that show short-term (seasonal) acclimatization of thermal tolerance thresholds could potentially be used as an indicator of which coral species are more sensitive to ocean warming. To conclude, our study highlights distinct seasonal fluctuations in thermal tolerance among two coral species in the Red Sea. P. verrucosa displayed a response reflective of prevailing seawater variations, possibly resulting from metabolic adjustments in response to gradual temperature shifts. Conversely, Acropora spp . maintained a more consistent thermal threshold, potentially supported by higher flexibility and, therefore, adaptation level of the associated microbiome 25 , 34 , likely selected from the seasonally different microbial communities available in the surrounding seawater. These findings imply that coral species may respond differently to a changing climate because it may be easier for some coral species, likely the most sensitive, to adapt and respond to gradual temperature changes, slowly enhancing their resilience. However, these species might become more vulnerable to stress and bleaching under acute thermal stress conditions. Future investigations should explore the mechanisms underlying these short-term responses to environmental changes within the holobiont, such as the host gene expression and microbiome shifts."
} | 6,120 |
24658114 | PMC4063449 | pmc | 451 | {
"abstract": "Many natural biological systems - such as biofilms, shells and skeletal\ntissues - are able to assemble multifunctional and environmentally responsive\nmultiscale assemblies of living and non-living components. Here, by using\ninducible genetic circuits and cellular communication circuits to regulate\n Escherichia coli curli amyloid\nproduction, we show that E. coli cells\ncan organize self-assembling amyloid fibrils across multiple length scales,\nproducing amyloid-based materials that are either externally controllable or\nundergo autonomous patterning. We also interfaced curli fibrils with inorganic\nmaterials, such as gold nanoparticles (AuNPs) and quantum dots (QDs), and used\nthese capabilities to create an environmentally responsive biofilm-based\nelectrical switch, produce gold nanowires and nanorods, co-localize AuNPs with\nCdTe/CdS QDs to modulate QD fluorescence lifetimes, and nucleate the formation\nof fluorescent ZnS QDs. This work lays a foundation for synthesizing,\npatterning, and controlling functional composite materials with engineered\ncells."
} | 266 |
36157340 | PMC9500351 | pmc | 452 | {
"abstract": "Biotreatment of acidic rare earth mining wastewater via acidophilic living organisms is a promising approach owing to their high tolerance to high concentrations of rare earth elements (REEs); however, simultaneous removal of both REEs and ammonium is generally hindered since most acidophilic organisms are positively charged. Accordingly, immobilization of acidophilic Galdieria sulphuraria ( G. sulphuraria ) by calcium alginate to improve its affinity to positively charged REEs has been used for simultaneous bioremoval of REEs and ammonium. The results indicate that 97.19%, 96.19%, and 98.87% of La, Y, and Sm, respectively, are removed by G. sulphuraria beads (GS-BDs). The adsorption of REEs by calcium alginate beads (BDs) and GS-BDs is well fitted by both pseudo first-order (PFO) and pseudo second-order (PSO) kinetic models, implying that adsorption of REEs involves both physical adsorption caused by affinity of functional groups such as –COO– and –OH and chemical adsorption based on ion exchange of Ca 2+ with REEs. Notably, GS-BDs exhibit high tolerance to La, Y, and Sm with maximum removal efficiencies of 97.9%, 96.6%, and 99.1%, respectively. Furthermore, the ammonium removal efficiency of GS-BDs is higher than that of free G. sulphuraria cells at an initial ammonium concentration of 100 mg L −1 , while the efficiency decreases when initial concentration of ammonium is higher than 150 mg L −1 . Last, small size of GS-BDs favors ammonium removal because of their lower mass transfer resistance. This study achieves simultaneous removal of REEs and ammonium from acidic mining drainage, providing a potential strategy for biotreatment of REE tailing wastewater.",
"conclusion": "4 Conclusions In this work, simultaneous removal of both REEs and ammonium from acidic rare earth mining wastewater is achieved by calcium alginate-immobilized Galdieria sulphuraria beads (GS-BDs). Four major conclusions are reached as follows: First, 95.47%, 94.15%, and 96.72% of La, Y, and Sm, respectively, are removed from REE wastewater by BDs, while slightly higher amounts of La, Y, and Sm can be removed by GS-BDs owing to the REE bioaccumulation capacity of microalgae. Second, the adsorption of REEs by BDs and GS-BDs is well fitted by PFO and PSO kinetic models because they are a combination of both physical adsorption due to the affinity of functional groups and ion exchange of Ca 2+ with REEs. The removal efficiencies of REEs at 80–160 mg L −1 are ranked as Sm > La > Y, with maximum removal efficiencies of 99.1%, 97.9%, and 96.6%, respectively. Third, the recovery efficiencies of La, Y, and Sm with initial concentrations ranging from 80 to 160 mg L −1 by GS-BDs are contrary to their removal efficiencies since the affinity of BDs for REEs is positively related to their REE adsorption capacity. Fourth, the ammonium removal efficiencies of both GS-BDs-15 and GS-BDs-5 are higher than the ammonium removal efficiency of free G. sulphuraria at an initial concentration of 100 mg L −1 but are lower when the initial ammonium concentration is higher than 150 mg L −1 because electrostatic repulsion hinders the further adsorption of ammonium by algae cells. Additionally, reducing the size of GS-BDs from 6.0 mm to 1.5 mm favors ammonium removal owing to both the increased surface area and lower mass transfer resistance at smaller diameters.",
"introduction": "1 Introduction Rare earth elements (REEs), composed of 15 lanthanide elements and two pseudolanthanide elements (Sc and Y), are present in small quantities with vast applications in manufacturing, the nuclear industry, electronic devices, and medicine [ 1 ]. REEs primarily exist in ore from specific regions of the world, which means that their extraction requires a series of specific processes, including physical meshing, floating, magnetic separation, and subsequent chemical processes, such as liquid‒liquid or solid‒liquid hydrometallurgical methods [ 2 , 3 ]. Accordingly, in situ chemical leaching with ammonium salts without extensive physical pretreatment is generally applied in REE mining operations, as a result of which local surface and groundwater are unavoidably contaminated by ammonium sulfate-rich effluent [ 4 ]. REE wastewater has been reported to be an acidic pollutant (pH value of 3.5–5.0) composed of NH 4 + (50–200 mg L −1 ), NO 3 − (10–80 mg L −1 ), SO 4 2− (200–700 mg L −1 ), trace organics (<10 mg chemical oxygen demand (COD) per L), and residual REEs, leading to serious environmental pollution problems [ 5 ]. Therefore, REE wastewater treatment is significant to local environmental protection. Considering the strategic properties and potential metal pollution of REEs, many studies regarding the recovery of REEs from REE wastewater through sorption to active carbon [ 6 , 7 ], biosorbents [ 8 , 9 ], nanoparticles [ 10 , 11 ], and clay minerals [ 12 ] and through precipitation with Fe oxyhydroxides, Al–Fe hydroxides, and even carbonates at high pH values have been conducted [ [13] , [14] , [15] ]. However, these strategies are extremely limited in practical applications since the adsorption capacity and precipitation efficiency are pH dependent and low recovery efficiencies are obtained under acidic conditions [ 16 ]. Additionally, alkali and alkaline metals such as Na + , K + , Ca 2+ , and Mg 2+ may compete with REEs for adsorption sites, leading to a limited recovery efficiency of REEs [ 17 ]. Accordingly, nanofiltration (NF), a pressure-driven process that can discriminate monovalent and multivalent ions, has alternatively been investigated for the recovery of REEs. However, membranes generally exhibit poor chemical stability under low pH conditions, leading to low REE rejection rates after filtration [ 18 ]. Therefore, the fabrication of acid-tolerant membranes has been extensively investigated to reduce the cost of membrane replacement [ 19 ]. Although NF can recover multivalent REEs under appropriate conditions, the high transmembrane pressure and loss of selectivity after long-term use in acid wastewater hinder their practical applications [ 20 ]. Biosorption of REEs via acidophilic living organisms has been investigated for the recovery of REEs owing to the high tolerance of such organisms to REEs under acidic conditions [ 21 ]. Initially, bacteria such as B. subtilis , L. methylohalidivorans , and P. inhibens were employed for the biosorption of REEs under acidic conditions [ 22 ]; however, extensive organic carbon is consumed by the metabolism of heterotrophic bacteria, making bacteria-based biosorption an energy-intensive methodology [ 23 ]. Therefore, autotrophic algae with CO 2 as a carbon source have attracted the attention of researchers. Furthermore, algal cell walls are generally composed of functional groups that act as active sites for binding REEs, such as –OH, –COOH, –NH 2 , oxygen-containing groups, and other sulfated groups [ 24 ]. Nevertheless, both macroalgae and microalgae exhibit limited REE tolerance below 25 mg L −1 and a relatively low recovery efficiency of 70% [ 25 ]. The biosorption of positively charged ciprofloxacin via electrostatic attraction by Chlamydomonas sp. Tai-03 with 100% removal efficiency strongly inspired our interest in REE adsorption [ 22 ]. Chemical modifications to change the electrostatic interactions of biosorbents and REEs, including functional group grafting [ 26 ], acid or base treatment [ 27 ], and alginate base immobilization [ 28 , 29 ], have been investigated for enhancing metal removal from wastewater, but ammonium nitrogen, another primary pollutant in REE acidic mining effluent, is rarely reduced or recycled simultaneously. However, some research on the biotreatment of ammonium/nitrate nitrogen-rich REE mining drainage by living microalgae has been reported owing to the remarkable nitrogen-fixation abilities of these organisms [ 8 ]. In addition to inorganic nitrogen, organic nitrogen, such as sulfamethoxazole, can be effectively biotransformed by microalgae [ 30 ]. Acidophilic microalgae, such as Chlorococcum sp., Galdieria sulphuraria ( G. sulphuraria ) [ 31 ], Scenedesmus sp. and Parachlorella sp. [ 5 ], are generally employed for deammonification owing to their high tolerance of REEs and low pH value. Although an ultrahigh ammonium recovery efficiency is reported in these studies, simultaneous recovery of REEs has rarely been reported. According to previous research, these acidophilic microalgae generally have a low removal efficiency (≤25 mg L −1 ), although they exhibit high REE tolerance. Based on our previous research, acidic microalgae are generally positively charged with a zeta potential of approximately 2–3 mV, which may hinder the adsorption of positively charged REEs [ 32 ]. Simultaneous recovery of high concentrations of REEs and ammonium from acidic mining drainage is still a great challenge for REE wastewater treatment. Calcium alginate-based immobilization is a promising strategy to functionalize the surface of microorganisms with specific functional groups or surface charges, potentially providing a path for the simultaneous recovery of REEs and ammonium [ 33 ]. On this basis, G. sulphuraria , a thermoacidophilic red algae, was immobilized by sodium alginate to simultaneously remove both REEs and ammonium nitrogen from acidic mining drainage. The biosorption efficiencies and adsorption kinetics of La, Y, and Sm on free G. sulphuraria , blank beads (BDs), and beads of immobilized G. sulphuraria (GS-BDs) were evaluated and simulated by a pseudo-first-order (PFO) kinetic model and pseudo-second-order (PSO) kinetic model. To explore the mechanism of adsorption of REEs, Fourier transform infrared (FTIR) spectroscopy and X-ray photoelectron spectroscopy (XPS) analysis of the surface of preadsorption and postadsorption GS-BDs were investigated in detail. The bioaccumulation of ammonium in free G. sulphuraria , BDs, and GS-BDs with variable sizes was also studied under different initial concentrations of ammonium to achieve the simultaneous removal of REEs and high concentrations of ammonium nitrogen.",
"discussion": "3 Results and discussion 3.1 REE removal efficiencies of algae, BDs, and GS-BDs Free G. sulphuraria , BDs, and GS-BDs were used to recover REEs from synthetic rare earth tailwater containing 40 mg L −1 La, Y, and Sm. Although G. sulphuraria exhibited remarkable biological activity in the synthetic rare earth tailwater, the concentrations of La, Y, and Sm remained at approximately 35.0 mg L −1 without a significant reduction during the cultivation of free G. sulphuraria algal cells in synthetic rare earth tailwater ( Fig. 1 a). Conversely, 95.47%, 94.15%, and 96.72% of La, Y, and Sm, respectively, were removed from REE wastewater containing 100 mg L −1 ammonium by blank beads (BDs). There are slight decreases in the removal efficiencies of La, Y, and Sm with increasing ammonium concentration from 100 to 250 mg L −1 . Analogously, 97.19%, 96.19%, and 98.87% of La, Y, and Sm, respectively, were removed by GS-BDs. The REE removal efficiencies of GS-BDs also decrease with increasing ammonium concentration from 100 to 250 mg L −1 . Notably, the REE removal efficiencies of GS-BDs are all higher than those of BDs, except at 200 mg L −1 ammonium ( Fig. 1 b–d). Unlike negatively charged microalgae such as Chlorella sp [ 36 ]. and Chlamydomonas sp . Tai-03 [ 22 ], G. sulphuraria are spheres with an average diameter of 5.84 μm and zeta potential of 2.69 mV. The electrostatic repulsion between the surface of algal cells and REE ions hinders the physical adsorption of REEs, similar to the electrostatic repulsion between Dunaliella acidophila and heavy metal ions [ 37 ], subsequently leading to constant REE concentrations during the cultivation of free G. sulphuraria . In contrast, GS-BDs composed of sodium alginate are rich in functional groups such as –OH and –COOH ( Fig. S1 ). These functional groups provide abundant active sites for the adsorption of REEs, resulting in the high REE removal efficiency of BDs. However, a blueshift in the absorption peak of O–H and an obvious increase in the absorption peak of asymmetric vibrations of –COO are observed owing to monodentate complexation, bidentate complexation or bridging of functional groups with REEs after biotreatment of REE wastewater ( Fig. S1 ). To confirm the existence of REEs in GS-BDs, XPS analysis of GS-BDs-15 after adsorption of REEs was conducted, and the results are presented in Fig. 2 . Peaks are observed at 852.3 and 835.4 eV, attributed to La 3d 3/2 and 3d 5/2 , respectively ( Fig. 2 b); 160.95 and 158.05 eV, attributed to Y 3d 3/2 and 3d 5/2 ( Fig. 2 c); and 1110.1 and 1083.1 eV, attributed to Sm 3d 3/2 and 3d 5/2 ( Fig. 2 d), implying the adsorption of REEs by GS-BDs. However, the REE removal efficiencies of both BDs and GS-BDs all decrease with increasing ammonium concentration, possibly due to competition of NH 4 + with REEs for active sites on the BDs [ 38 ]. The REE removal efficiency of GS-BDs is slightly higher than the REE removal efficiency of BDs because the extracellular polymeric substances of algal cells are rich in functional groups, facilitating bioaccumulation in G. sulphuraria cells [ 39 ]. Even though the increase in the REE removal efficiency of GS-BDs is limited, their advantages in removing ammonium reduce the cost of ammonium removal from REE wastewater. Fig. 1 a , Removal of La, Y, and Sm by free G. sulphuraria . b–d, Removal percentages of La ( b ), Y ( c ), and Sm ( d ) by BDs and GS-BDs-15. Fig. 1 Fig. 2 XPS spectra of GS-BDs-15 after adsorption of REEs. Fig. 2 3.2 Kinetics of adsorption of REEs To explore the mechanism of adsorption of REEs by BDs and GS-BDs, the adsorption kinetics of REEs were fitted by a PFO kinetic model and a PSO kinetic model. As demonstrated in Fig. 3 , La, Y, and Sm are all quickly adsorbed by both BDs and GS-BDs within 3 min, while the adsorption rates slow after 3 min and finally reach equilibrium in 8 min. In comparison, a slight decrease in the adsorption rates of REEs by BDs and GS-BDs with increasing ammonium concentration is observed. REEs have been reported to be able to coordinate with carboxylic groups and hydroxylic groups through monodentate complexation, bidentate complexation or bridging on the surface of BDs and GS-BDs, leading to the fast accumulation of REEs within 3 min [ 40 ]. The subsequent decrease in the adsorption efficiency of REEs is possibly due to the following two reasons. On the one hand, most of the active sites for the adsorption of REEs are occupied by La, Y, or Sm after 3 min, leading to decreases in the adsorption rates of REEs. On the other hand, the surface charges of BDs and GS-BDs become more positive with increasing concentrations of REEs, subsequently inhibiting further adsorption of REEs by enhanced electrostatic repulsion between the beads and REE ions. The decreases in the REE removal rates of BDs and GS-BDs are consistent with the variation in their relative removal efficiencies, possibly owing to the competition of NH 4 + with REEs for active adsorption sites [ 41 ]. The parameters of the PFO model for the adsorption of La, Y and Sm by BDs and GS-BDs are listed in Table 1 , Table 2 , respectively. As presented, the adsorption of REEs by both BDs and GS-BDs is well fitted by the PFO model, with R 2 values above 0.98 ( Table 1 , Table 2 ), implying that the initial adsorption of REEs by both BDs and GS-BDs is a physical process. Fig. 3 Simulated PFO model of REE adsorption by BDs under different ammonium concentrations: a , La; b , Y; c , Sm. Simulated PFO model of REE adsorption by GS-BDs under different ammonium concentrations: d , La; e , Y; f , Sm. Fig. 3 Table 1 Parameters of the PFO model for the adsorption of La, Y, and Sm by BDs. Table 1 NH 3 –N (mg L −1 ) La Y Sm q e ’ (μg g −1 ) k (min −1 ) R 2 q e ’ (μg g −1 ) k (min −1 ) R 2 q e ’ (μg g −1 ) k (min −1 ) R 2 100 113.5 0.8876 0.9921 112.2 0.8412 0.9907 113.2 1.085 0.9918 150 113.4 0.8927 0.9997 111.7 0.8364 0.9995 114.9 1.035 1.0 200 114.2 0.5549 0.9970 111.5 0.5695 0.9990 119.8 0.6367 0.9774 250 108.4 0.7427 0.9992 105.6 0.7194 0.9989 111.6 0.8421 0.9991 Table 2 Parameters of the PFO model for the adsorption of La, Y, and Sm by GS-BDs. Table 2 NH 3 –N (mg L −1 ) La Y Sm q e ’ (μg g −1 ) k (min −1 ) R 2 q e ’ (μg g −1 ) k (min −1 ) R 2 q e ’ (μg g −1 ) k (min −1 ) R 2 100 115.6 0.8014 0.9978 117.1 0.7454 0.9942 115.5 0.9751 0.9985 150 116.1 0.7243 0.9931 117.2 0.6967 0.9887 117.7 0.8003 0.9915 200 111.1 0.7186 0.9904 112.1 0.6902 0.9850 113.9 0.7801 0.9897 250 114.6 0.5481 0.9896 115.6 0.5312 0.9897 117.0 0.5992 0.9942 Considering the possible ion exchange during the adsorption of REEs by BDs and GS-BDs, the kinetics of the adsorption of REEs were also simulated via the PSO model. According to Fig. 4 , the adsorption of REEs by both BDs ( Fig. 4 a–c) and GS-BDs ( Fig. 4 d–f) is well fitted by the PSO model with average R 2 values of approximately 0.9906 ( Table 3 ) and 0.9768 ( Table 4 ), respectively, implying that chemical reactions also take place during the adsorption of REEs. Accordingly, the variations in the concentrations of REEs and Ca 2+ with time were determined and are presented in Fig. 5 . Remarkably, the concentration of Ca 2+ increases from 24.1 to 260.9 mg L −1 with the adsorption of REEs from wastewater, while the concentrations of La, Y, and Sm decrease from 39.4, 37.1, and 37.5 mg L −1 to 3.09, 3.86, and 1.89 mg L −1 , respectively. According to stoichiometric analysis, approximately 0.872 mmol L −1 REEs are adsorbed by BDs, which implies that 1.31 mmol L −1 Ca 2+ is released due to ion exchange during adsorption; this relationship indicates that an extra stoichiometric amount of Ca 2+ is released in this process because of the release of Ca 2+ from BDs or GS-BDs after the initial enrichment of Ca 2+ from the culture medium. Overall, we propose that the adsorption of REEs from acidic mining drainage follows two processes: first, initial physical adsorption due to the affinity between REEs and functional groups of calcium alginate; and second, ion exchange between REEs and Ca 2+ . Ion exchange of Ca 2+ with REEs after physical adsorption is a limiting step for further bioaccumulation of REEs. REEs are stored in BDs or GS-BDs at an equilibrium concentration following these two processes. Fig. 4 Simulated PSO model of REE adsorption by BDs under different ammonium concentrations: a , La; b , Y; c , Sm. Simulated PSO model of REE adsorption by GS-BDs under different ammonium concentrations: d , La; e , Y; f , Sm. Fig. 4 Table 3 Parameters of the PSO model for the adsorption of La, Y, and Sm by BDs. Table 3 NH 3 –N (mg L −1 ) La Y Sm q e ’ (μg g −1 ) k 2 × 10 3 (g μg −1 min −1 ) R 2 q e ’ (μg g −1 ) k 2 × 10 3 (g μg −1 min −1 ) R 2 q e ’ (μg g −1 ) k 2 × 10 3 (g μg −1 min −1 ) R 2 100 128.2 9.7 0.9930 127.4 9.09 0.9927 125.6 13.51 0.9923 150 128.6 9.61 0.9958 127.7 8.87 0.9966 128.1 12.18 0.9933 200 139.5 4.21 0.9880 135.2 4.56 0.9929 143.5 4.93 0.9538 250 125.8 7.41 0.9966 123.1 7.22 0.9970 127.6 8.79 0.9948 Table 4 Parameters of the PSO model for the adsorption of La, Y, and Sm by GS-BDs. Table 4 NH 3 –N (mg L −1 ) La Y Sm q e ’ (μg g −1 ) k 2 × 10 3 (g μg −1 min −1 ) R 2 q e ’ (μg g −1 ) k 2 × 10 3 (g μg −1 min −1 ) R 2 q e ’ (μg g −1 ) k 2 × 10 3 (g μg −1 min −1 ) R 2 100 133.2 7.75 0.9892 136.5 6.71 0.9814 130.1 10.77 0.9922 150 136.0 6.43 0.9775 138.2 5.91 0.9710 135.9 7.50 0.9675 200 130.3 6.60 0.9718 132.5 6.05 0.9643 131.9 7.41 0.9709 250 140.0 4.14 0.9783 142.5 3.83 0.9746 140.6 4.73 0.9829 Fig. 5 Variations in the concentrations of REEs and Ca 2+ with time. Fig. 5 The adsorption capacities and removal percentages of La, Y, and Sm by GS-BDs-15 with initial concentrations of 80, 120, and 160 mg L −1 were also tested. As demonstrated, La, Y, and Sm still have a high removal efficiency when a high initial concentration is used ( Fig. 6 ), exhibiting losses of 11.3, 14.6, and 7.16 mg L −1 , respectively, after 40 min with an initial REE concentration of 160 mg L −1 in wastewater ( Fig. 6 c). The removal efficiencies of REEs at 80–160 mg L −1 are ranked as Sm > La > Y, with maximum removal efficiencies of 99.1%, 97.9%, and 96.6%, respectively ( Fig. 6 d). The removal rate of metal ions generally depends on the properties of the adsorbent, electronegativity of the metal ion, ionic radius, coordination number, and affinity constant between the metal and adsorbent [ 42 , 43 ]. Among these parameters, the affinity constants of La, Y, and Sm are 10 6.8 , 10 6.8 , and 10 7.0 , respectively, indicating that Sm has the highest affinity [ 44 , 45 ]. Accordingly, the removal percentages of Sm are higher than those of La and Y under the same conditions. Fig. 6 Variations in the concentrations of REEs with time under different initial concentrations of REEs: a , 80 mg L −1 ; b , 120 mg L −1 ; c , 160 mg L −1 d , Removal percentages of REEs under different concentrations. Fig. 6 3.3 Recovery of REEs The recovery of REEs from GS-BDs after adsorption was also tested by dissolving GS-BDs in sodium citrate. According to Fig. 7 , the recovery percentages of La, Y, and Sm decrease from 72.7%, 76.1%, and 63.5% with an initial concentration of 80 mg L −1 to 65.1%, 66.4%, and 56.5%, with an initial concentration of 120 mg L −1 , respectively. However, they notably increase to 72.3%, 74.2%, and 64.3% with an initial concentration of 160 mg L −1 , respectively. We suppose that most REE ions are physically adsorbed by BDs and GS-BDs and may therefore be more easily detached by sodium citrate, while REEs might chemically bind with alginate via ion exchange with calcium ions, leading to decreases in the recovery efficiencies of REEs. However, we suppose that the amount of physically adsorbed REEs increases since extra REEs are physically adsorbed at an initial concentration of approximately 160 mg L −1 , subsequently enhancing the recovery of REEs from GS-BDs. Another point worth mentioning is that the recovery efficiencies of REEs at initial concentrations of 80–160 mg L −1 are ranked as Y > La > Sm, which is totally different from the relative removal efficiencies of GS-BDs from wastewater. The high affinity of REEs such as Sm for BDs possibly leads to low recovery efficiency, and vice versa. In addition to demonstrating potential for the recovery of REEs, immobilized G. sulphuraria can be harvested via filtration and resuspended for the next sequence of immobilization, possibly reducing the cost for the treatment of REE wastewater. Fig. 7 Recovery percentages of REEs from GS-BDs-15 after treatment of REE wastewater under 80, 120, and 160 mg L −1 La, Y, and Sm with 200 mg L −1 NH 4 + . Fig. 7 3.4 Removal of ammonium by GS-BDs In addition to the adsorption of REEs from acidic mining wastewater, the bioremoval of ammonium, another important pollutant, was extensively investigated. According to Fig. 8 a–d, ammonium at initial concentrations of 100, 150, 200, and 250 mg L −1 is initially adsorbed by BDs within 1–3 days, after which it is slowly released to reach equilibrium concentrations of 95.4, 128.0, 160.6, and 210.0 mg L −1 , respectively. Conversely, significant decreases in ammonium concentrations are observed with the addition of free G. sulphuraria and GS-BDs with initial biomass concentrations of 15 and 5 g L −1 , named GS-BDs-15 and GS-BDs-5, respectively. Notably, both GS-BDs-15 and GS-BDs-5 have higher ammonium removal efficiencies than free algae cells of G. sulphuraria at an initial ammonium concentration of 100 mg L −1 , while the ammonium removal rates of GS-BDs decrease with increasing initial concentration of ammonium. In addition, average equilibrium concentrations of 73.6, 88.5, 115.2, and 128.3 mg L −1 ammonium are obtained after REE treatment with an initial ammonium concentration of 250 mg L −1 by BDs, free G. sulphuraria , GS-BDs-15, and GS-BDs-5, as demonstrated in Fig. 8 a–d. As illustrated, NH 4 + might be adsorbed by functional groups, such as –COO–, and –OH, during the adsorption of REEs, leading to a dramatic decrease in NH 4 + concentration. Nevertheless, the adsorbed NH 4 + would be released to reach an equilibrium NH 4 + concentration owing to the competition of REEs for active sites. Ammonium is an important nitrogen source for the metabolism of G. sulphuraria . Accordingly, a continuous decrease in ammonium concentration is observed due to bioaccumulation by G. sulphuraria . However, a lower ammonium removal rate is observed for free algae cells than immobilized algae cells at an initial ammonium concentration of 100 mg L −1 , possibly owing to both the reduced charges of algae cells and the increased functional groups resulting from immobilization. Moreover, the ammonium removal rate decreases with increasing initial ammonium concentration since extra ammonium adsorbed by BDs would electrostatically repel further ammonium adsorption to algae cells. Furthermore, there is a slight increase in ammonium concentration at the 3rd or 4th day of cultivation of GS-BDs with initial ammonium concentrations of 150, 200, and 250 mg L −1 , proving that adsorption equilibrium of ammonium is reached with alginate-based immobilization. Although the ammonium removal efficiencies after 5 days with initial concentrations of 150, 200, and 250 mg L −1 are lower for GS-BDs than free G. sulphuraria , the adsorption of REEs by GS-BDs offsets the extra cost for immobilization of microalgae and reduced ammonium removal efficiency. Fig. 8 Removal of ammonium by BDs, free G. sulphuraria cells (GS), GS-BDs-15, and GS-BDs-5 with a mixture of coexisting REEs (40 mg L −1 La, 40 mg L −1 Y, and 40 mg L −1 Sm) under the following initial concentrations: a , 100 mg L −1 ; b , 150 mg L −1 ; c , 200 mg L −1 ; d , 250 mg L −1 . Fig. 8 Unlike the adsorption of REEs, the removal of ammonium is primarily attributed to the metabolism of algal cells. Thus, the effect of the size of GS-BDs, which determines the mass transfer resistance of NH 4 + from wastewater to algae cells, on the ammonium removal rate was also investigated. GS-BDs with diameters of 1.5, 3 and 6 mm, named GS-BDs-1.5, GS-BDs-3.0, and GS-BDs-6.0, were used to remove ammonium from acidic mining wastewater. According to Fig. 9 , 58.6, 74.6, and 112.0 mg L −1 ammonium is observed after the treatment of REE wastewater by GS-BDs-1.5, GS-BDs-3.0, and GS-BDs-6.0, respectively. Obviously, reducing the size of GS-BDs from 6.0 mm to 1.5 mm is more favorable to ammonium removal owing to both increased surface area and lower mass transfer resistance at smaller diameters. However, a continuous decrease might lead to larger mass transfer resistance because of both the formation of tightly bound BDs caused by enhanced ion exchange and self-shading of GS-BDs, subsequently leading to a decrease in ammonium removal efficiency [ 46 ]. Fig. 9 Ammonia nitrogen removal by GS-BDs (5 g L −1 ) with particle sizes of 1.5, 3, and 6 mm. Fig. 9"
} | 6,784 |
28777381 | PMC5649168 | pmc | 453 | {
"abstract": "Aerobic methanotrophic bacteria have evolved a specialist lifestyle dependent on consumption of methane and other short-chain carbon compounds. However, their apparent substrate specialism runs contrary to the high relative abundance of these microorganisms in dynamic environments, where the availability of methane and oxygen fluctuates. In this work, we provide in situ and ex situ evidence that verrucomicrobial methanotrophs are mixotrophs. Verrucomicrobia-dominated soil communities from an acidic geothermal field in Rotokawa, New Zealand rapidly oxidised methane and hydrogen simultaneously. We isolated and characterised a verrucomicrobial strain from these soils, Methylacidiphilum sp. RTK17.1, and showed that it constitutively oxidises molecular hydrogen. Genomic analysis confirmed that this strain encoded two [NiFe]-hydrogenases (group 1d and 3b), and biochemical assays revealed that it used hydrogen as an electron donor for aerobic respiration and carbon fixation. While the strain could grow heterotrophically on methane or autotrophically on hydrogen, it grew optimally by combining these metabolic strategies. Hydrogen oxidation was particularly important for adaptation to methane and oxygen limitation. Complementary to recent findings of hydrogenotrophic growth by Methylacidiphilum fumariolicum SolV, our findings illustrate that verrucomicrobial methanotrophs have evolved to simultaneously utilise hydrogen and methane from geothermal sources to meet energy and carbon demands where nutrient flux is dynamic. This mixotrophic lifestyle is likely to have facilitated expansion of the niche space occupied by these microorganisms, allowing them to become dominant in geothermally influenced surface soils. Genes encoding putative oxygen-tolerant uptake [NiFe]-hydrogenases were identified in all publicly available methanotroph genomes, suggesting hydrogen oxidation is a general metabolic strategy in this guild.",
"introduction": "Introduction Aerobic methane-oxidising bacteria (methanotrophs) consume the potent greenhouse gas methane (CH 4 ) ( Kirschke et al. , 2013 ). They serve as the primary biological sink of atmospheric methane (~30 Tg annum −1 ) ( Hanson and Hanson, 1996 ) and, together with anaerobic methane-oxidising archaea, also capture the majority of biologically and geologically produced CH 4 before it enters the atmosphere ( Oremland and Culbertson, 1992 ). Relative to their global impact as greenhouse gas mitigators, aerobic methanotrophs exhibit low phylogenetic diversity and are presently limited to 26 genera in the Alphaproteobacteria and Gammaproteobacteria ( Euzéby, 1997 ), two candidate genera in the phylum Verrucomicrobia ( Op den Camp et al. , 2009 ; van Teeseling et al. , 2014 ), and two representatives of candidate phylum NC10 ( Ettwig et al. , 2010 ; Haroon et al. , 2013 ). Reflecting their aerobic methylotrophic lifestyle, methanotrophs thrive in oxic–anoxic interfaces where CH 4 fluxes are high, including peat bogs, wetlands, rice paddies, forest soils and geothermal habitats ( Singh et al. , 2010 ; Knief, 2015 ). However, they also exist within soil and marine ecosystems where CH 4 and oxygen (O 2 ) are more variable ( Knief et al. , 2003 ; Tavormina et al. , 2010 ; Knief, 2015 ). Based on current paradigms, aerobic methanotrophs are thought to primarily grow on one-carbon (C1) compounds in the environment ( Dedysh et al. , 2005 ). All species can grow by oxidising CH 4 to methanol via particulate or soluble methane monooxygenase. They subsequently oxidise methanol to carbon dioxide (CO 2 ), yielding reducing equivalents (e.g. NADH) for respiration and biosynthesis. Proteobacterial methanotrophs generate biomass by assimilating the intermediate formaldehyde via the ribulose monophosphate or serine pathways ( Hanson and Hanson, 1996 ). In contrast, verrucomicrobial methanotrophs oxidise methanol directly to formate ( Keltjens et al. , 2014 ) and generate biomass by fixing CO 2 \n via the Calvin–Benson–Bassham cycle ( Khadem et al. , 2011 ). While these specialist C1-based metabolisms are thought to be the primary growth strategy under optimal conditions (i.e. CH 4 and O 2 replete conditions), they would presumably be less effective in dynamic environments where CH 4 and oxidant availability are likely to fluctuate. To add to this complexity, the methane monooxygenase reaction (CH 4 +O 2 +[NAD(P)H+H + ]/QH 2 →CH 3 OH+NAD(P) + /Q+H 2 O) ( Hakemian and Rosenzweig, 2007 ) is metabolically demanding, given it requires simultaneous sources of CH 4 , endogenous reductant (NAD(P)H or quinol) and exogenous O 2 to proceed. Methanotrophs therefore must carefully allocate resources to meet carbon, energy and reductant demands ( Hanson and Hanson, 1996 ). This complex balancing act provokes that, in order to be viable in environments limited for CH 4 and O 2 gases ( Knief et al. , 2003 ; Tavormina et al. , 2010 ), methanotrophs should be able to supplement C1 usage with other energy-yielding strategies. Recent pure culture studies have provided evidence that CH 4 -oxidising bacteria are indeed more metabolically versatile than previously thought. A minority of conventional methanotrophs can meet energy demands by oxidising the trace concentrations of CH 4 (1.8 ppmv) found in the atmosphere ( Kolb et al. , 2005 ; Ho et al. , 2013 ; Cai et al. , 2016 ). Contrary to the long-held paradigm that methanotrophs are obligate methylotrophs, species from three alphaproteobacterial genera have been shown to grow on simple organic acids, alcohols and short-chain alkane gases ( Dedysh et al. , 2005 ; Crombie and Murrell, 2014 ). Most recently, it has been shown that some methanotrophs are not exclusive heterotrophs: the verrucomicrobium Methylacidiphilum fumariolicum SolV can sustain chemolithoautotrophic growth on molecular hydrogen (H 2 ) through the activity of two [NiFe]-hydrogenases ( Mohammadi et al. , 2016 ). Proteobacterial methanotrophs can also consume H 2 , though to date this process has only been reported as providing reductant to supplement methanotrophic growth ( Chen and Yoch, 1987 ; Shah et al. , 1995 ; Hanczár et al. , 2002 ). Our recent findings demonstrating a widespread distribution and diversity of hydrogenases in aerobic bacteria, in specific methanotrophs ( Greening et al. , 2014a , 2014b , 2015 , 2016 ), led us to surmise that H 2 metabolism could serve a multifaceted role in adaptation of methanotrophic bacteria to their environment. Specifically, H 2 may serve as an important electron donor for the organism to meet carbon, energy and reductant demands in response to fluctuations in CH 4 and oxidant availability. In this work, we addressed this hypothesis by conducting an interdisciplinary investigation of the role of H 2 in defining the physiology and ecology of verrucomicrobial methanotrophs. Evidence obtained from in situ field studies indicate that Verrucomicrobia simultaneously oxidised CH 4 and H 2 in geothermally heated soils in Rotokawa, New Zealand, suggesting they are mixotrophic with respect to energy metabolism. Pure culture studies on a verrucomicrobium representative isolated from this site confirmed that the microorganism grew most efficiently through a mixotrophic lifestyle and depended on H 2 consumption to acclimate to fluctuations in CH 4 and O 2 availability. Integrating these findings with genome surveys, we propose that H 2 oxidation expands the ecological niche of methanotrophs, enabling them to meet energy and biomass demands in dynamic environments where O 2 and CH 4 concentrations are variable. We provide evidence that, while methanotrophic bacteria are often pervasively viewed as C1 specialists, their niche space is likely broader than previously recognised. Combining heterotrophic and lithotrophic electron donors allows for a more flexible growth/survival strategy with clear ecological benefits ( Semrau et al. , 2011 ).",
"discussion": "Results and discussion Verrucomicrobia-dominated surface soils serve as a sink of geothermally derived H 2 and CH 4 in Rotokawa geothermal field We performed a geochemical, molecular and biochemical survey of CH 4 and H 2 metabolism in an acidic geothermal soil in Rotokawa, New Zealand. The Rotokawa geothermal field is a predominately steam-driven system dominated by acidic and sulphurous springs and heated soil features. We selected a geothermally heated and acidic soil where previous studies have indicated methanotrophic activity (Sharp et al. , 2014). Substantial vertical gradients in temperature, pH and mixing ratios of CH 4 , H 2 and O 2 were observed in the soil profile ( Figure 1a ). Consistent with the geothermal activity at the site, high soil mixing ratios of CH 4 (47 000 ppmv) and H 2 (280 ppmv) were detectable at the deepest soil depths sampled. The levels of both gases decreased in the upper 30 cm of soil and, in the case of H 2 , dropped towards atmospheric levels by 10 cm depth ( Figure 1b ). These sharp decreases suggested that there were active methanotrophs and hydrogenotrophs in the oxic zone of the soil that consume most geothermally derived gas before it is emitted into the atmosphere. Indeed, microcosm incubations containing surface soils and associated communities rapidly consumed H 2 and CH 4 introduced into ambient air headspaces. Rates of H 2 oxidation exceeded that of CH 4 , suggesting that H 2 serves as a major energy source for this geothermal soil community ( Figures 1c and Figures 1d ). To infer the microorganisms responsible for CH 4 and H 2 uptake, we determined the microbial community structure of the soil profile. Consistent with findings in other acidic soil ecosystems ( Golyshina, 2011 ; Sharp et al. , 2014 ; Lee et al. , 2016 ), the euryarchaeotal order Thermoplasmatales was dominant at all depths. Methanotrophic verrucomicrobial genera, specifically Methylacidiphilum spp. and Methylacidimicrobium spp., were the dominant bacterial OTUs in surface soils and accounted for 47% of all bacterial 16S rRNA gene sequences in the soil profile ( Figure 1e ). Bacteria from these genera have been previously isolated in acidic geothermal soils in New Zealand ( Dunfield et al. , 2007 ; Sharp et al. , 2014 ), Kamchatka ( Islam et al. , 2008 ) and Italy ( Pol et al. , 2007 ; van Teeseling et al. , 2014 ). As the only known acidophilic methanotrophs ( Op den Camp et al. , 2009 ), it is probable that the verrucomicrobial phylotypes were solely responsible for CH 4 consumption in this ecosystem. Moreover, given putative uptake [NiFe]-hydrogenases have been detected in the genomes of Methylacidiphilales but not in Thermoplasmatales ( Hou et al. , 2008 ; Khadem et al. , 2012b ; Greening et al. , 2016 ), it was likely that the Verrucomicrobia detected in these soils serve as major sinks of H 2 in this ecosystem. To test this possibility, we designed PCR primers to detect the presence of genes encoding the large subunits of the three particulate methane monooxygenases ( pmoA ) and a single oxygen-tolerant uptake hydrogenase ( hyaB ) encoded in the genome of Methylacidiphilum infernorum V4 ( Hou et al. , 2008 ) ( Supplementary Table S1 ). These primers were applied in qPCRs on DNA extracts from the Rotokawa soil profile. Both the hydrogenase and methane monoxygenase genes were detected at all depths, with the most abundant templates detected in the top 10 cm of soil ( Figure 1f ), corresponding to the zones with the highest relative abundance of Verrucomicrobia-affiliated sequences and where the lowest CH 4 and H 2 soil gas concentrations were detected ( Figure 1b ). A verrucomicrobial strain isolated from Rotokawa constitutively oxidises CH 4 and H 2 gas To gain insight into the metabolic strategies that Verrucomicrobia use to dominate bacterial assemblages in geothermally heated acidic soil ecosystems, we isolated a thermotolerant methanotroph from surface soils. The strain, Methylacidiphilum sp. RTK17.1 ( Supplementary Table S2 ), grew optimally at pH 2.5, 50 °C ( T max 60 °C) and shared 99% 16S rRNA gene sequence identity with Methylacidiphilum infernorum V4 ( Dunfield et al. , 2007 ). Bacteriological characterisation confirmed that the strain, in common with other verrucomicrobial methanotrophs ( Khadem et al. , 2012a , 2011 ), oxidised CH 4 , fixed CO 2 and accumulated glycogen. In addition, cultures rapidly consumed exogenous H 2 ( Supplementary Figures S2 and S3 ). Real-time amperometric measurements confirmed that the strain oxidised H 2 under oxic conditions at rates proportional to increases in cell density ( Figure 2a ). H 2 oxidation occurred in all batch culture conditions tested, including when CH 4 was absent ( Supplementary Figure S1A ), when CH 4 was in excess ( Supplementary Figure S1C ), and following inhibition of CH 4 oxidation with acetylene ( Supplementary Figure S2 ). This suggests that H 2 and CH 4 are oxidised independently serving to energise the respiratory chain through the reduction of the quinone pool. Moreover, this expands the role of hydrogenases in aerobic methanotrophs beyond their previously suggested role of providing reductant for pMMO ( Shah et al. , 1995 ; Hanczár et al. , 2002 ). The observation that RTK17.1 can constitutively oxidise H 2 and CH 4 parallels results from the soil study showing that both H 2 and CH 4 are simultaneously oxidised ( Figure 1 ). Considering that Verrucomicrobia are dominant among taxa putatively capable of oxidising H 2 or CH 4 , this provides further evidence that verrucomicrobial methanotrophs adopt a mixotrophic lifestyle with respect to their energy metabolism. We sequenced the genome of Methylacidiphilum sp. RTK17.1 to obtain further insights into the potential functionality of this taxon ( Supplementary Table S2 ). Genes encoding key enzymes and pathways for CH 4 oxidation to CO 2 , CO 2 fixation through the Calvin–Benson–Bassham pathway, and aerobic respiration ( Figure 3 ) were highly conserved with those identified in other Methylacidiphilum strains ( Hou et al. , 2008 ; Khadem et al. , 2012b ; Erikstad and Birkeland, 2015 ). We also detected two [NiFe]-hydrogenase-encoding gene clusters in the genome ( Supplementary Figure S3 ) and confirmed their expression during aerobic growth with CH 4 and H 2 by RT-PCR ( Supplementary Figure S4 ). The gene clusters were classified as groups 1d ( hyaABC ) and 3b ( hyhBGSL ) [NiFe]-hydrogenases based on phylogenetic affiliation with biochemically characterised enzymes ( Supplementary Figure S3 ; Greening et al. , 2016 ; Søndergaard et al. , 2016 ). Biochemically characterised group 1d [NiFe]-hydrogenase are H 2 -uptake multimeric proteins that are membrane-bound via their cytochrome b subunit ( hyaC ) and function by transferring electrons into the respiratory chain via the quinone pool ( Fritsch et al. , 2011 ). Consistent with the observed activity of the [NiFe]-hydrogenase in the presence of O 2 ( Figure 2a ), enzymes of this class are predicted to be O 2 -tolerant due to the presence of a novel [4Fe3S] cluster that protects the O 2 -sensitive active site from oxidative damage ( Fritsch et al. , 2011 ; Shomura et al. , 2011 ). Indeed, the six cysteine residues required to ligate such a cluster were conserved in the deduced RTK17.1 HyaB protein sequence. In comparison, biochemically characterised group 3b hydrogenases are reversible cytosolic enzymes that are relatively O 2 -sensitive ( Kwan et al. , 2015 ); they directly couple NAD(P)H oxidation to H 2 formation during fermentation ( Berney et al. , 2014a ) and, in some cases, H 2 oxidation by these enzymes supports CO 2 fixation through the production of reduced electron carriers ( Yoon et al. , 1996 ). The RTK17.1 [NiFe]-hydrogenase combination differs from Methylacidiphilum fumariolicum SolV, where group 1 h/5 ( hhyLH ; a putative high-affinity H 2 uptake hydrogenase) and group 1d [NiFe]-hydrogenases (annotated as hupSLZ ) were reported ( Mohammadi et al. , 2016 ), with our previous survey ( Greening et al. , 2016 ) also showing SolV encodes a group 3b enzyme ( Supplementary Figure S3 ). H 2 oxidation supports aerobic respiration and CO 2 fixation in the verrucomicrobial isolate The observation that RTK17.1 encodes and utilises [NiFe]-hydrogenases prompted biochemical studies to investigate the role of H 2 in the metabolism of this bacterium. Biochemical assays targeting the group 1d [NiFe]-hydrogenase demonstrated that it is a membrane-bound uptake hydrogenase linked to the aerobic respiratory chain, consistent with our genome-based predictions. Fractionation experiments confirmed the activity was membrane-localised, as shown by the 31-fold increase in activity in membranes when compared to the cytosolic fraction ( Figure 2b ). We next tested the effect of the ionophores nigericin and valinomycin on rates of H 2 oxidation on whole cells (in the absence of CH 4 ). These compounds (nigericin and valinomycin) dissipate components of the electrochemical gradient used for ATP synthesis (pH and charge gradient, respectively) and the cellular response is to increase respiration to replenish the electrochemical gradient ( Cook et al. , 2014 ) in a phenomenon known as uncoupling. H 2 oxidation increased upon treatment with these ionophores ( Figures 2c and d ), showing hydrogenase activity behaves as expected from a component of the energy-conserving respiratory chain. This uncoupled activity rapidly ceased, due to O 2 consumption by the cells suspended in the sealed chamber, but could be restored by further supplementation with O 2 . These results show hydrogenase is a bona fide component of this microorganism’s respiratory chain and is coupled to the activity of terminal cytochrome oxidases. Under these conditions, the onset of O 2 -limitation was likely exacerbated by the catabolism of endogenous glycogen reserves ( Supplementary Figure S5 ). Collectively, these findings demonstrate that this group 1d [NiFe]-hydrogenase is a membrane-bound, respiratory-linked, O 2 -tolerant/dependent enzyme that drives ATP synthesis as has been observed in other aerobic hydrogenotrophs. To test whether H 2 oxidation coupled to O 2 reduction could support CO 2 fixation in RTK17.1, we transferred mixotrophically grown, log phase cultures into a new microoxic headspace (O 2 1% v/v) in which H 2 (8% v/v of the headspace) was present as the sole exogenous electron donor and CO 2 (8% v/v) as the sole carbon source. Trace 14 CO 2 (0.1% of total CO 2 supplied) was added and the amount fixed into biomass sampled over time was determined by measuring disintegrations per minute (DPM) via liquid scintillation counting. We observed systematic increases in DPMs associated with cells sampled over a 20 h period relative to controls, indicating that CO 2 was rapidly incorporated into biomass in a time-dependent manner. The number of DPMs associated with cells after 20 h of incubation were 200-fold greater in live than heat-killed cells ( Figure 2e ), showing that biological CO 2 fixation occurs in cultures supplied with H 2 as the sole reductant and O 2 as the sole oxidant. This activity was not observed in the absence of exogenous H 2 , indicating that H 2 serves as the source of reductant for CO 2 fixation under these conditions. H 2 oxidation supports adaptation of verrucomicrobial methanotrophs to CH 4 and O 2 limitation The finding that Methylacidiphilum sp. RTK17.1 couples H 2 oxidation to aerobic respiration and carbon fixation suggests that it can grow chemolithoautotrophically. Consistent with this hypothesis, we observed a small but significant increase in biomass in cultures grown under microoxic conditions when H 2 , CO 2 and O 2 (1% headspace concentration) were supplied as the sole electron donor, carbon source and electron acceptor, respectively ( Supplementary Figure S2A ). This biomass increase was concomitant with increased amounts of CO 2 fixed ( Figure 2e ), and the amount of carbon fixed per cell (40 to 80 fmol cell −1 ) produced over this time period was consistent with biomass yields from other studies ( Maestrini et al. , 2000 ). However, the observed growth rates (0.005 h −1 ) were substantially lower than observed when RTK17.1 was supplied with CH 4 , CO 2 and H 2 (0.037 h −1 ) and when Methylacidiphilum fumariolicum SolV was grown autotrophically in similar conditions (0.047 h −1 ) ( Mohammadi et al. , 2016 ). We also observed that autotrophic growth was only sustained when RTK17.1 was incubated under microoxic conditions (1% O 2 ) rather than in an oxic (20% O 2 ) headspace. We speculate that, under such microoxic conditions, sufficient O 2 is available to drive hydrogenotrophic aerobic respiration through activity of the group 1d [NiFe]-hydrogenase. Simultaneously, it is likely that the O 2 -sensitive group 3b [NiFe]-hydrogenase can remain active and is able to supply reducing equivalents required for CO 2 fixation. In support of this notion, H 2 oxidation sustained energy-conservation via the group 1d hydrogenase of non-growing CH 4 -limited cultures in the presence of ambient O 2 ( Supplementary Figure S2C ) and enhanced growth yields in CH 4 -replete cultures ( Supplementary Figure S2D ). To better understand the role of H 2 and CH 4 oxidation during mixotrophic growth, we compared growth and gas consumption kinetics of the cells cultivated in a chemostat under six different conditions ( Table 1 ). We observed that H 2 addition into the feedgas of Methylacidiphilum sp. RTK17.1 increased growth yields under O 2 -replete and O 2 -limiting conditions. Whereas CH 4 oxidation predominated under O 2 -replete conditions, the specific consumption rate of H 2 increased 80-fold and exceeded rates of CH 4 oxidation under O 2 -limiting conditions. In combination, these results show that Methylacidiphilum sp. RTK17.1 grows mixotrophically and modulates rates of H 2 and CH 4 consumption in response to the availability of O 2 in order to balance energy-generation and carbon fixation. H 2 oxidation may be a general ecological strategy for verrucomicrobial and proteobacterial methanotrophs In this work, we demonstrated that a verrucomicrobial methanotroph adopts a mixotrophic lifestyle both in situ and ex situ . The environmental isolate Methylacidiphilum sp. RTK17.1 sustains aerobic respiration and carbon fixation by using organic (CH 4 ) and inorganic (H 2 ) electron donors either in concert or separately depending on substrate availability. Through the dual use of both electron donors, the bacterium is able to more flexibly adjust its metabolism to meet energy and carbon demands in response to simulated environmental change. A model of how CH 4 and H 2 metabolism is integrated into the physiology of this microorganism, based in part on genomic information ( Supplementary Table S2 ), is shown in Figure 3 . Integrating our genomic, physiological and biochemical findings, we conclude that the group 1d [NiFe]-hydrogenase is a membrane-bound uptake hydrogenase that is directly linked to the aerobic respiratory chain and supplements the quinone pool to power the methane monoxygenase reaction or feed directly into Complex III. The organism is also capable of using H 2 as a reductant to support CO 2 fixation through the Calvin–Benson-Bassham pathway, likely through the cytosolic NAD(P)-coupled group 3b [NiFe]-hydrogenase. The [NiFe]-hydrogenase combination (group 1d and group 3b) in this RTK17.1 only supports weak autotrophic growth, but provides multiple layers of support for a mixotrophic lifestyle. In addition to supporting growth and survival during periods of CH 4 limitation, our data show that H 2 is the preferred electron donor during O 2 -limiting conditions. Under these conditions, rates of H 2 consumption increased by 77-fold and exceeded observed rates of CH 4 oxidation ( Table 1 ). This is likely to be a consequence of two factors. Firstly, some hydrogenases such as the group 3b [NiFe]-hydrogenase are inhibited at high O 2 concentrations ( Kwan et al. , 2015 ). Secondly, methanotrophy is more resource-intensive than canonical aerobic hydrogenotrophy, given it requires O 2 both as a substrate for methane monooxygenase and as the terminal electron acceptor for respiration ( Hakemian and Rosenzweig, 2007 ). Comparison of our independent findings with those made by Mohammadi et al. (2016) suggests that verrucomicrobial methanotrophs have evolved a range of strategies to integrate H 2 metabolism into their physiology. Both Methylacidiphilum sp. RTK17.1 and Methylacidiphilum fumariolicum SolV are capable of sustaining chemolithoautotrophic growth on H 2 /CO 2 under microoxic conditions. However, the strains grow at drastically different rates of 0.005 and 0.047 h −1 ( Mohammadi et al. , 2016 ), respectively, under optimal conditions. These differences may reflect that, while both strains possess group 1d ( hyaABC/hupLSZ ) and group 3b ( hyhBGSL ) hydrogenases, SolV has also acquired a group 1 h enzyme ( hhyLH ) ( Greening et al. , 2016 ) with surprisingly fast whole-cell kinetics ( Mohammadi et al. , 2016 ). It is possible that, with this enhanced hydrogenase suite, SolV may be able to more efficiently partition electrons derived from H 2 oxidation between respiration and carbon fixation. In addition to supporting hydrogenotrophic growth, both organisms also modulate hydrogenase expression and H 2 oxidation rates in response to simulated environmental change, such as CH 4 and O 2 availability ( Mohammadi et al. , 2016 ). While the physiological significance of this regulation was not explored in SolV, our studies inferred that H 2 co-oxidation with CH 4 enhanced yields during CH 4 surplus and sustained survival during CH 4 limitation in RTK17.1. Further differences between the strains are reflected in the regulatory profile, with the group 1d enzyme constitutively expressed in RTK17.1, but repressed in favour of the group 1 h enzyme under oxic conditions in SolV ( Mohammadi et al. , 2016 ). Overall, our findings suggest that SolV may fulfil a similar ecological niche to classical Knallgas bacteria (e.g. Ralstonia eutropha ), switching between efficient heterotrophic and autotrophic growth dependent on energy availability. In contrast, H 2 metabolism appears to be more important for optimising growth and survival of RTK17.1 in response to energy and O 2 availability. In this regard, this organism’s metabolism more closely resembles the mixotrophic strategy employed by Mycobacterium smegmatis ( Berney et al. , 2014a ; Greening et al. , 2014b ). Future studies would benefit from side-by-side comparisons of these strains under equivalent conditions and further exploration of the physiological role and biochemical features of the group 3b and group 1 h [NiFe]-hydrogenase enzymes. More generally, we predict that H 2 oxidation is likely to support the majority of aerobic methanotrophic bacteria. Whereas only a few methanotrophic genera appear to be capable of heterotrophic generalism ( Crombie and Murrell, 2014 ), genomic surveys ( Peters et al. , 2015 ; Greening et al. , 2016 ) show that all 31 publicly available aerobic methanotrophic genomes harbour the capacity to metabolise H 2 ( Supplementary Figure S3 ). As with SolV ( Mohammadi et al. , 2016 ) and RTK17.1, most of the surveyed methanotroph genomes that were found to encode for [NiFe]-hydrogenases have been shown to support aerobic respiration (groups 1d, 1h, 2a) and carbon fixation (groups 3b, 3d, 1 h) ( Greening et al. , 2016 ). Reports showing H 2 oxidation by several proteobacterial strains further support the classification of these enzymes as uptake hydrogenases ( Chen and Yoch, 1987 ; Shah et al. , 1995 ; Hanczár et al. , 2002 ). H 2 is likely to be a particularly attractive energy source for methanotrophs because of its relative ubiquity when compared to C1 compounds. H 2 is biologically produced by diverse organisms across the three domains of life as a result of fermentation, photobiological processes and nitrogen fixation ( Peters et al. , 2013 ; Schwartz et al. , 2013 ; Poudel et al. , 2016 ). Moreover, verrucomicrobial and proteobacterial methanotrophs harbouring the recently described group 1 h [NiFe]-hydrogenases ( Greening et al. , 2014a , 2015 ) may be capable of scavenging atmospheric H 2 to survive CH 4 starvation. Considering these observations in concert, it seems likely that hydrogenases in aerobic methanotrophs function to supplement the energetic and reductant requirements in environments where CH 4 and O 2 gases are limiting or variable. Thus, while methanotrophic bacteria are often pervasively viewed as C1 specialists, we propose that, via the utilisation of hydrogenases as part of a mixotrophic strategy, the niche space of methanotrophs is much broader than previously recognised. Combining heterotrophic and lithotrophic electron donors allows for a more flexible growth/survival strategy, with clear ecological benefits ( Semrau et al. , 2011 ). We therefore predict that most methanotrophs will be able to use H 2 to support either autotrophic growth, mixotrophic growth or long-term persistence/maintainence. Finally, our geochemical and microbial community diversity investigation of the Rotokawa geothermal field provides ecological support to our assertion that the metabolic flexibility of methanotrophs enhances niche expansion in situ . We provide genetic and biochemical evidence that methanotrophic Verrucomicrobia inhabiting the near-surface soils co-metabolised CH 4 and H 2 gas ( Figure 1 ) and in doing so adopt a clear mixotrophic strategy. In acidic geothermal soils, we demonstrated that verrucomicrobial methanotrophs have grown to be the dominant bacterial taxon by simultaneously consuming gases primarily of geothermal and atmospheric origin, that is, CH 4 and H 2 as energy sources, respectively, CO 2 as a carbon source and O 2 as oxidant. Their metabolic flexibility also ensures resilience to temporal and spatial variations in the availability of key substrates allowing for CH 4 oxidation via the monooxygenase reaction. More generally, the prevalent narrative that methanotrophic bacteria are methylotrophic specialists is based on studies under optimal growth conditions and ignores the requirement of these organisms to adapt to environmental variations requiring a certain level of metabolic versatility. Intimate evolutionary and ecological interactions are likely to have selected for a spectrum of different lifestyles across methanotrophic lineages, ranging from strict C1 specialism to broad substrate generalism, depending on the environment. However, based on the presence of [NiFe]-hydrogenase in numerous methanotroph genomes ( Supplementary Figure S3 ) and the data presented here, we contend it is likely that most methanotrophs depend on H 2 oxidation to some extent to support either growth and/or survival. This finding has broad implications for future investigations on the ecology of methanotrophs as well as the biogeochemical cycles of H 2 and CH 4 ."
} | 7,817 |
34542731 | PMC8452823 | pmc | 454 | {
"abstract": "Highlights \n An up-to-date review of hybrid triboelectric-electromagnetic nanogenerators is provided. Rotational, pendulum, linear, sliding, cantilever, flexible blade, multidimensional, and magnetoelectric hybrid technologies are thoroughly analyzed. Promising results highlight the potential of these hybrid technologies for both small-scale and large-scale powering.",
"conclusion": "Concluding Remarks and Future Prospects This review deeply and systematically explores the most relevant breakthroughs already carried out in the scope of hybrid triboelectric-electromagnetic nanogenerators driven by mechanical energy harvesting. The theoretical transduction mechanism models developed for both TENGs and EMGs were exposed, as well as reported structural design prototypes and corresponding electromechanical characteristics and applications. Hybridized E-TENGs can offer a more efficient vibration energy conversion by taking advantage of both of their desirable complementary high voltage and high current characteristics and wider operating bandwidths. The TENGs are effectively able to harvest electric energy from low-frequency (< 1 Hz) and low amplitude (< 1 mm) kinetic energy, providing large output voltages. The EMGs are effective technologies to harvest at high frequencies and amplitudes of excitation, providing high output currents. E-TENGs can also be connected in series or in parallel, or used independently, to output high voltages or currents to fulfill customized requirements and suitability for particular applications. Thus, by yielding peak output powers higher than 100 mW, they might find use as micro/nanopower sources or in self-powered sensors by scavenging general forms of vibration energy, wheel rotation energy, biomechanical energy, blue energy, wind energy, thermal energy and more. Therefore, E-TENG generators hold potential to power captivating technologies, such as those related to the internet of things, wireless sensor networks, portable electronics, implantable biomedical devices, etc. Despite the observed impressive progress, some important issues remain to be addressed: (i) Energy conversion efficiency—The energy scavenging capabilities of TENGs are still not sufficient to continuously power most conventional electronic devices. Since its output power increases with the square of the triboelectric surface charge density, enhancing this value is mandatory. At least five different strategies have been explored in this regard: modification in material composition, improvement in effective contact area, surface charge pumping, ionized-air injection and control of environmental conditions. The composition of the triboelectric materials can be modified through chemical surface functionalization [ 182 ], by changing the functional groups on the surface or bulk composition manipulation [ 183 ], by using high dielectric constant materials as fillers, such that the charge-attracting or charge trapping capability of the materials can be enhanced. The effective contact area can be improved by introducing nano/microstructures onto the triboelectric materials, such as nanowires and pyramid/cube-like arrays [ 137 , 184 ] or by using fluid–solid interfaces [ 185 ]. The charge density of a triboelectric nanogenerator can be pumped up by a charge-shuttling technique [ 186 ] or enhanced by ionized-air injection [ 187 ]. Charge accumulation strategy was successfully employed to boost the output voltage providing sustainable OC voltages over 20 kV in a TENG with rotating tribo-electrodes and transporting electrodes [ 188 ]. Controlling environmental conditions such as temperature and air pressure to increase surface charge density and dielectric breakdown voltage is another valid strategy [ 189 ]. (ii) Power management—Coupling the output powers from the TENG and EMG units is a tremendous issue due to the huge impedance mismatch between them and between TENGs and storage devices. This is also complicated by the dependence of the equivalent internal impedance of TENGs and resonant EMGs with the excitation frequency. A better understanding of the combined energy harvesting efficiency in hybrid systems is required. It is also relevant the conversion of irregular AC electrical outputs into stable DC input power to supply electronic systems. The use of conventional rectifiers and transformers introduces large energy losses so that advanced power management methods to achieve high energy harvesting efficiency are critical [ 190 ]. Impedance matching has already been achieved with TENGs by designed complex power converters, employing multiple temporary storage capacitors, coupled inductors and switches controlled by logic units [ 103 , 191 ]. In such systems, a first stage with a small temporary capacitor (with large equivalent impedance) is charged through a bridge rectifier from 0 V. Once the voltage reaches an optimal level after the impedance match condition is achieved, the energy stored in the temporary capacitor is transferred to another stage with a storage unit (e.g., a large capacitor) through coupled inductors. Thus, such an optimized charging cycle can lead to a maximization of the charging efficiency up to ca. 75% [ 190 ]. Another strategy makes use of a group of capacitors that are periodically charged in a series connection and discharged to a load in a parallel connection [ 192 ]. Energy transfer maximization has also been explored through a rationally modulated charging cycle controlled by a motion-triggered switch placed in parallel to the energy storage unit [ 193 ]. The undesirable small output current from TENGs can also be enhanced, e.g., using sliding micro-grating [ 141 ], radial-arrayed [ 142 ] or multi-layered stacked structures [ 194 ]. (iii) Durability and stability—Stable and reliable materials and devices are required for real life applications, which could constitute a problem mainly in the case of high-friction lateral-sliding TENGs. More durable composite structures must be investigated for TENGs or using a conjunction of working modes. Strategies already tested have included mechanically switchable structures with automatic transition between contact (triboelectric charge build-up) and non-contact working states [ 195 ] and low-friction free-standing structures with intermediate rolling rods [ 149 ]. Structures with flexible hair brushes have been used to greatly reduce the operation friction resistance, enhance the device durability and solve the charge dissipation problem [ 76 ]. The friction forces of TENGs fabricated by PTFE were shown to be lower than those with PVC and FEP under the same pressure, namely in wind energy harvesting excited by the movement of high-speed rail vehicles [ 196 ]. Thus, based on the reasonable selection of dielectric materials with minimal friction coefficient, the results highlight that the friction force of a double-layer PTFE/Kapton elastic rotation TENG was greatly reduced, the energy harvesting efficiency was doubled and its durability increased by 4 times [ 196 ]. Another effective strategy to simultaneously enhance the power density and durability of sliding-mode TENGs is via liquid lubrication interface [ 197 ]. Operation in a squalene solution has been shown to reduce the friction force, suppress the interfacial electrostatic breakdown and decrease the charge loss by more than 50% [ 197 ]. (iv) Packaging—The performance of the nanogenerators, especially TENGs, sensibly depends on the moisture and temperature. Thus, good packaging methods for devices with moving parts are required to guarantee a thorough water sealing [ 65 ], while raising the working temperature of the devices could also prove useful [ 198 ]. Humidity-resistive TENGs were also fabricated making use of metal organic framework composites to increase the electron-trapping capacity and dielectric constant [ 123 ]. (v) Energy storage—Energy storage of irregular electrical outputs is required. Improvements must be carried out regarding the leakage problem in supercapacitors and new designs of Li-ion batteries should be tested [ 199 ]. (vi) Operation input—Wider bandwidths of input frequencies and amplitudes are valuable. This could be achieved, e.g., by periodic tuning of the resonant frequency, mechanical stoppers, nonlinear springs or bi-stable structures [ 200 ]. Low frequencies and amplitudes of excitation are associated with the low input powers and predominance of velocity independent Coulomb’s mechanical dry friction. Some solutions to overcome this problem might include the use of flexible brushes as contact materials [ 76 ], rational selection of low-friction dielectric materials [ 196 ] and interface liquid lubrication [ 197 ]. (vii) Understanding of the conversion mechanism—Obtaining a better understanding of the triboelectric effect and its relation to the environmental conditions could be useful to optimize the performance of the harvesters. Experimental studies using, e.g., atomic force microscopy and Kelvin probe microscopy [ 112 , 113 ] and theoretical first principle quantum physical studies are still lacking. New techniques to accurately measure the surface charge density and its relation to the dielectric properties and surface morphology of the materials would be welcome. An updated quantitative standard triboelectric series could be important for material choice guidance. New computational models to predict the electromechanical output for realistic irregular tri-dimensional motions of combined hybrid harvesters should also be developed. (viii) Quantitative standardization—Quantitative experimental standards should be employed for comparing and calibrating the performance of developed E-TENG prototypes [ 118 ], since testing conditions vary widely in the literature. Reporting of average power or conversion efficiency values should also be preferred to maximum instantaneous peak power due its high sensitivity to the time form of the input mechanical excitation. Values of the input displacement, acceleration or force amplitudes should be provided. (ix) System integration—Rational designs should be developed for system integration with different kinds of vibration environments. Biomechanical and hydrodynamic fluid flow studies could be important [ 78 , 107 ]. Economic studies should be carried out to estimate the maintenance and production costs of hybrid nanogenerators to evaluate their competitiveness in the energy market. The possible impact of relatively large stray magnetic and electric fields associated with EMGs and TENGs, respectively, should also be addressed. (x) Miniaturization—Miniaturization of E-TENGs for the integration with wearable equipment or incorporation within innovative implantable bioelectronic medical devices should be explored [ 201 ]. Significant output powers from biomechanical sources and the ability to work in biological environment are relevant requirements for future advanced applications of miniaturized E-TENGs. Small-scale individual generators are expected to provide small output powers, although multiple arrays of such devices can theoretically yield improved results. Shape-adaptive TENGs made of elastomeric materials and mounted under shoes or weaved into clothing for powering wearable electronics have already been developed [ 202 , 203 ]. Fiber-based TENGs for wearable power sources that could be woven into fabric have also been tested [ 204 , 205 ]. EMGs are expected to suffer from poor volume downscaling with rapidly dropping output powers and increasing resonance frequencies. Their characteristic low output voltages can also result in no output power because of the certain voltage level required to polarize diodes in the rectification circuit. Assembling small-scale coils and magnets might also prove challenging. TENGs should have a better volume scaling behavior although problems related to charge leakage due to imperfect insulation and dielectric breakdown due to thinner dielectrics are anticipated. The predominance of surface effects, including stray electric fields at the edges and corresponding parasitic capacitances, is also expected. (xi) Large-scale manufacturing—While large-scale EMGs are easier to manufacture, as only conventional machining technology is required, large-scale triboelectric generators require special attention, because upscaling of the laboratory production techniques such as pulsed laser deposition (PLD) or atomic layer deposition (ALD) to a mass production level is difficult. Moreover, standard techniques reported in the literature to boost the performance of TENGs involve nano/micro-patterning of the electrodes, which is complex, tedious and expensive. The solutions for the large-scale patterning could be soft lithography as well as surface treatment processes, including different kinds of plasma treatment processes and chemical synthesis methods suitable for large-scale manufacturing [ 206 ]. For the layer deposition, several techniques have been used such as s flexible printed circuit manufacture, inkjet printing, screen printing and roll-to-roll patterning [ 206 ]. The choice of the method is determined by the materials used and type of the device to be manufactured. (xii) Multi-energy hybrid cells—Hybrid cells to simultaneously harvest energy from multiple available types of environmental energy, including solar, thermal and mechanical, will most likely be very useful in future [ 71 ]. These questions raise many opportunities in the development of hybrid triboelectric-electromagnetic nanogenerators for researchers and entrepreneurs across all fields. We thus expect this field to experience a rapid growth in the next decade.",
"introduction": "Introduction Energy is one of the most important resources with a huge demand from the part of modern society due to its central standing to the general quality of life [ 1 ]. The energy transition may be one of the greatest challenges facing humanity today driven by environmental and limited fossil fuel concerns. Thus, the demand for new large-scale sustainable forms of energy, such as solar, wind and hydroelectric, is enormous. The near future is further expected to see a worldwide distribution of a huge number of mobile electronics driven by the recent progress observed in the development of the internet of things (IoT), wireless sensor networks (WSN), microelectromechanical systems (MEMS), wearable electronics, implantable medical devices, robotics and artificial intelligence (AI), all of them requiring new compact and efficient sources of energy [ 2 – 5 ]. Foreseen problems related to the limited lifetime and periodic replacement of conventional external power sources, such as batteries, in a large number of devices must be addressed. Fortunately, a continuous reduction in size and power requirements of integrated circuits to values lower than a few µW has been seen last decade [ 6 , 7 ]. Thus, the concept of nanoenergy was created, meaning the use of nanomaterials and nanotechnology for the creation of integrated power sources able to harvest various forms of energy, such as light, thermal and mechanical [ 8 – 12 ]. Several types of nanogenerators (NG) have thus already been explored, including those for self-powered industrial and health monitoring systems, smart housing and clothing, ambient intelligence, self-powered portable electronic devices, intelligent traffic systems, vibration dampers, wireless power transmitters, etc. [ 13 – 15 ]. Among the various forms of ambient energy, low-level ambient vibrations are ubiquitous, originating from sources such as human motion (walking, talking, breading, typing, etc.), working machinery, vehicles, ocean and lake waves, wind flows, etc., and typically range in acceleration amplitudes of 0.01‒1 g and frequencies of 1‒200 Hz [ 7 , 12 , 16 , 17 ]. The four main transduction mechanisms between mechanical and electrical energy are: piezoelectric [ 8 – 13 , 18 – 20 ], electromagnetic [ 21 , 22 ], electrostatic [ 23 , 24 ] and triboelectric [ 25 , 26 ]. In particular, triboelectric nanogenerators (TENGs) have been developed and very widely investigated during the last decade [ 27 – 31 ]. The operation mechanism of TENGs is based on the combined phenomena of triboelectric and electrostatic induction effects [ 32 – 35 ]. The first consists of a contact-induced electrification, where two different material surfaces become electrically charged after physically contacting each other. The second consists of a flow of free charges in an external circuit induced by the electric fields produced by the mechanical separation between the aforementioned surfaces with opposite electrostatic charges. Alternatively, electromagnetic generators (EMGs) have been known since the nineteenth century, being currently widely used in conjunction with various turbines in hydraulic, natural gas, nuclear and coal-based large-scale power generation. The fundamental operation of EMGs is based on the phenomenon of electromagnetic induction, or the generation of an electromotive force in a wire loop by a changing magnetic flux as described by Faraday’s law of induction. Several types of EMGs can be broadly categorized as mechanically into rotatory [ 36 ], linear [ 37 ] or multidimensional [ 38 ] kinetic input and electrically into direct, alternate or variable current output [ 13 , 39 – 45 ]. TENGs use a wide range of conventional materials and are currently known to be mechanically flexible, lightweight, cost-effective and easily scalable with low operation frequencies and large bandwidths [ 46 ]. EMGs are well-established, efficient, versatile, reliable, effective at large scales, have an easily controllable internal impedance and high frequency of operation. Electrically, TENGs behave as low current sources with high parallel internal impedance because of the electrostatic induction mechanism and the nature of the insulator-to-insulator or insulator-to-metal interface. Mesoscale devices typically have high output open-circuit (OC) voltages (~ 1–1000 V) and low short-circuit (SC) currents (~ 1–1000 µA) and capacitive internal impedance characteristics. Their characteristics of high output voltage and low current and susceptibility to wear, ambient humidity and temperature, as well as low and unstable charge density on tribo-layers, still limit the practical applications of TENGs [ 46 ]. EMGs, on the other hand, behave as low voltage sources with low series internal impedance due to the electromagnetic induction mechanism and the high conductivity of the coils. They typically have low output OC voltages (~ 1–1000 mV) and high SC currents (~ 1–1000 mA) and resistive and inductive internal impedance characteristics. As such, both technologies are complementary as explained in the following. Depending on the environment of operation of self-powered electronic devices, multiple types of energy might be available and the final goal is to power such devices using all available resources. TENG hybrid cells have been developed to simultaneously harvest energy from various sources, such as mechanical [ 47 , 48 ], solar [ 49 – 51 ], thermal [ 51 ] or chemical [ 49 , 52 ]. Each type of mechanical energy harvesters provides its own benefits and unique advantages/drawbacks [ 53 ]. TENG/EMG hybrid cells (E-TENGs) can be used to simultaneously scavenge vibrational energy by taking advantage of their complementarity: high voltage (TENG) and high current (EMG) or, alternatively, to use either of these to meet the requirements of particular applications [ 54 , 55 ]. Furthermore, they can be used to broaden the operating bandwidth of the nanogenerator due to TENG’s high efficiency at low frequencies and amplitudes of excitation and increased performance of EMGs at high frequencies and amplitudes. In fact, early comparative studies between transduction mechanisms have suggested that piezoelectric and triboelectric energy harvesters provide superior performance in relation to electromagnetic generators at low frequencies and low dimensions [ 56 – 60 ], as well as for small displacement amplitudes of excitation [ 61 ]. By taking into account constitutive equations for their respective conversion mechanisms, scaling analysis of the output power of different transducer types as a function of effective material volume ( V ) has shown that it should be roughly proportional to V 2 and V 2/3 for the electromagnetic and electrostatic generators, respectively [ 8 , 56 , 58 ]. Thus, below a critical volume of ~ 0.5 cm 3 the triboelectric mechanism can become more attractive [ 58 ]. Nevertheless, technological difficulties may still be encountered at smaller size scales, e.g., high magnetic flux gradients, assembling components in EMGs and surface potential decay due to imperfect dielectric insulation and stray electric field at the edges in TENGs. Recently, linear TENGs and EMGs, with similar geometry and size, were fabricated and their electrical output characteristics were systematically studied as a function of the amplitude and frequency of excitation, yielding significantly larger output powers for the TENG in the low-frequency (⪅ 1 Hz) and small-amplitude regime (⪅ 1 mm), as illustrated in Fig. 1 a, b [ 61 ]. The most important electromechanical characteristics of TENGs and EMGs are summarized in Table 1 and Fig. 1 c. Another study compared the applied torque and energy conversion efficiencies between rotational TENGs and EMGs [ 62 ]. The input mechanical torque of the EMG was shown to be balanced by the friction and electromagnetic resisting torques, which increased with increasing rotation rate due to Ampère’s force. The input torque of the TENG was balanced by the friction and electrostatic resisting torques, which were nearly constant with the rotation rate. The energy conversion efficiency of the EMG was observed to increase with increasing mechanical power inputs, while the one of the TENG remains nearly constant. These results suggest that the TENG can be superior to the EMG for harvesting mechanical energy with low input powers (⪅ 11.4 mW). Fig. 1 a Domain of excitation amplitude and frequency values where the energy harvesting performance of a test EMG or TENG is superior. The light red area denotes the dominant scope of the TENG in low-frequency and small-amplitude while the light green area denotes that of the EMG. b Maximum average output power ratio of the EMG and TENG versus amplitude and frequency. c Summarized overall comparison table between the complementary characteristics of EMGs and TENGs. Reproduced with permission from Ref. [ 61 ]. Copyright 2019, Nano Energy. (Color figure online) Table 1 Summary of the electromechanical parameters of the E-TENGs reported in the literature, generally expressed in terms of maximum peak values: open-circuit voltage ( V OC ), short-circuit current ( I SC ) and maximum power at a matching load resistance E-TENG Triboelectric materials TENG mode Energy source Dimensions External excitation Electric characterization TENG Electric characterization EMG Electric characterization Hybrid Rotating segmentally structured disk (Zhang et al.) [ 63 ] Kapton nanorods/Al Lateral-sliding General rotation TENG: 45.2 cm 3 EMG: 288 cm 3 20 π rad/s (600 rpm) DC \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC ≈ 80 V DC \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 6.96 µA DC Power = 140.4 µW @ 13.8 MΩ DC \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 67.6 mV DC \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC ≈ 6 mA DC Power = 102.6 µW @ 12.3 Ω DC \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC ≈ 110 mV DC \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC ≈ 7 µA DC Power = 209.7 nW @ 18.1 kΩ Rotating disk (Zhong et al.) [ 64 ] Polyamine/radial-arrayed Cu Lateral-sliding General rotation Φ140 mm × 5 mm 200 rpm \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 75 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 330 µA Power = 8.6 mW @ 0.2 MΩ \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 0.62 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 57.8 mA Power = 8.4 mW @ 12 Ω ND Waterproof rotating disk (Guo et al.) [ 65 ] FEP nanowires/Cu Sliding free-standing Wind energy/water flow Φ100 mm 1600 rpm \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 500 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 100 µA Power = 7 mW @ 10 MΩ \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 2.9 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 15 mA Power = 4.5 mW @ 60 Ω \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC ≈ 5 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC ≈ 2.3 mA Power ≈ 2.9 mW @ ≈ 1.1 kΩ Rotating-disk-based (Chen et al.) [ 66 ] PTFE/Radial-arrayed Au Sliding free-standing General rotation Φ280 mm 900 rpm \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 127 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 4.2 mA Power = 217.8 mW @ 20 kΩ \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 33.5 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 39 mA Power = 137.39 mW @ 300 Ω ND Rotating-sleeve-based (Cao et al.) [ 67 ] FEP nanowires/Cu Sliding free-standing Wind energy ND 400 rpm \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 600 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 150 µA Power = 12.7 mW @ 9 MΩ \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 0.85 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 11.5 mA Power ≈ 2.1 mW @ 50 Ω \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 48 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 2.2 mA Power ≈ 13 mW @ ≈ 8 kΩ Rotating-disk-based (Zhang et al.) [ 68 ] PTFE nanostructured/Al Single-electrode Wind energy/Road traffic ≈ 3.14 × 10 –4 m 3 1000 rpm \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 240 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 10 µA Power = 3.4 mW @ 50 MΩ \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 7.5 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 9 mA Power = 16.2 mW @ 400 Ω \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 3.5 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 5 mA Power = 17.5 mW @ 700 Ω Ultra-low friction (Wang et al.) [ 69 ] FEP/Cu Sliding-mode free-standing Wind energy ≈ Φ65 mm × 40 mm 1000 rpm \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 65 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 8.2 µA Power = 438.9 mW/kg @ 40 MΩ \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 7 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 100 mA Power = 181 mW/kg @ 80 Ω ND Blue energy harvester (Shao et al.) [ 70 ] PVDF nanowires/Al Contact-separation Blue energy ND 100 rpm \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 315.8 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 44.6 µA Power avg = 90.7 µW @ ~ 100 MΩ \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 0.59 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 1.78 mA Power avg = 79.6 µW @ 318 Ω ND Broad-band blue energy nanogenerator (Wen et al.) [ 71 ] FEP nanowires/foam Sliding-mode free-standing Blue energy ≈ Φ70 mm × 80 mm 300 rpm, 20 mm \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 375 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 14.12 µA Power avg = 15.67 µW/cm 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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 360 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 8.05 µA Power avg = 4.07 µW/cm 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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 1.79 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 11.57 mA Power avg = 27.12 µW/cm 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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 0.51 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 2.91 mA Power avg = 6.33 µW/cm 2 ND Easily assembled hybrid (Zhong et al.) [ 72 ] FEP/Cu Sliding-mode free-standing Wind energy/water flow Φ65 mm × 55 mm 500 rpm \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 200 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC ≈ 10 µA Power = 1.05 mW @ 4 MΩ \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 5 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC ≈ 60 mA Power = 58.3 mW @ 20 Ω ND Thermomagnetic generator (Ahmed et al.) [ 73 ] FEP/Al Sliding-mode free-standing Thermal energy ≈ Φ95 mm × 100 mm 46.5 °C (263 rpm) \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 15.6 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 9.82 µA Power = 14.4 µW @ 10 MΩ \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 4.56 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 20.61 mA Power = 15.62 mW @ 100 Ω \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 15.34 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC ≈ 20.61 mA Power = 12.1 mW @ 120 Ω Chaotic pendulum (Chen et al.) [ 74 ] PTFE/Au Sliding-mode free-standing Blue energy Φ100 mm × 167 mm 2.5 Hz \\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}$$V_{{{\\text{OCpp}}}}$$\\end{document} V OCpp = 197.03 V \\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}$$I_{{{\\text{SCpp}}}}$$\\end{document} I SCpp = 3 µA Power = 15.21 µW @ 400 MΩ \\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}$$V_{{{\\text{OCpp}}}}$$\\end{document} V OCpp = 1.08 V \\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}$$I_{{{\\text{SCpp}}}}$$\\end{document} I SCpp = 4 mA Power = 1.23 mW @ 400 Ω ND Rotational pendulum (Hou et al.) [ 75 ] FEP/Cu Contact-separation Human motion/blue energy Φ71 mm × 40 mm 10 cm \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 230 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 7 µA Power = 0.65 mW @ 10 MΩ \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 17.5 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 50 mA Power = 265 mW @ 75 Ω ND Soft-contact swing generator (Feng et al.) [ 76 ] FEP/Rabbit hair Sliding-mode free-standing Blue energy ≈ Φ100 mm × 50 mm 10 cm, 0.1 Hz \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 640 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 4.57 µA Power = 1.29 mW @ 150 MΩ \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 2.9 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 11.9 mA Power = 3.5 mW @ 300 Ω Power = 4.8 mW Rotating hexagonal prism (Fan et al.) [ 77 ] PDMS pyramids /ITO Contact-separation Rotating tire Side ≈ 50 mm ND ND ND \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 40 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 5 mA Linear magnetic levitation (Hu et al.) [ 47 ] Kapton nanowires/Cu Lateral-sliding/Contact-separation General vibration ND ND \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC ≈ 500 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC ≈ 325 µA Power = 22.5 mW @ 10 MΩ \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC ≈ 1.4 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC ≈ 16 mA Power = 5.8 mW @ 90 Ω \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 4.6 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 2.2 mA Power ≈ 10 mW @ R = 5 kΩ Elastic-impact-based hybrid (Rahman et al.) [ 78 ] PTFE nanowires/nylon Contact-separation Biomechanical energy ND 1 g \\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}$$V_{{{\\text{OCpp}}}}$$\\end{document} V OCpp = 230 V \\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}$$I_{{{\\text{SCpp}}}}$$\\end{document} I SCpp = 77 µA Power = 1.21 mW @ 9 MΩ \\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}$$V_{{{\\text{OCpp}}}}$$\\end{document} V OCpp = 15.10 V \\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}$$I_{{{\\text{SCpp}}}}$$\\end{document} I SCpp = 115.88 mA Power = 142.42 mW @ 76 Ω \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 13.15 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC ≈ 60 mA Power = 144.1 mW @ 1.5 kΩ Hybrid energy cell (Wu et al.) [ 79 ] PDMS pyramids/PA Contact-mode single-electrode General vibration 45 mm × 45 mm × 200 mm 22 Hz \\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}$$V_{{{\\text{OCpp}}}}$$\\end{document} V OCpp = 600 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 3.5 µA Power = 0.25 mW @ 100 MΩ \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 3 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 1 mA Power = 0.58 mW @ 10 kΩ ND Heaving point absorber harvester (Saadatnia et al.) [ 80 , 81 ] PTFE/Nylon Sliding-mode free-standing Blue energy Φ19.05 mm × 203.2 mm ND \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC ≈ 100 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC ≈ 20 µA Power ≈ 115 W/m 3 @ ≈ 10 MΩ \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC ≈ 2 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC ≈ 60 mA Power ≈ 210 W/m 3 @ ≈ 40 Ω ND Traffic generator (Askari et al.) [ 82 ] PTFE/Al Sliding-mode free-standing Road traffic ND 1500 N \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC ≈ 450 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC ≈ 150 µA Power ~ 20.92 W/m 3 @ ≈ 10 GΩ \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC ≈ 7.5 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC ≈ 90 mA Power ≈ 50.81 W/m 3 @ ≈ 100 Ω ND Dual Halbach array-based (Salauddin et al.) [ 83 , 84 ] PDMS pyramids/Al Contact-separation Human motion 65 mm × 26 mm × 18 mm 0.5 g V = 60 V Power = 88 µW @ 10 MΩ V = 1.22 V Power = 11.5 mW @ 32.5 Ω Power = 3.1 mW @ 700 Ω E-TENG–PENG (He et al.) [ 85 ] Silicon with carbon nanotubes/ NdFeB Contact-mode free-standing General vibration Φ48 mm × 27 mm 2.5 mm \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 15 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 2 µA Power = 78.4 µW @ ≈ 10 MΩ \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 7 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 7 mA Power = 38.4 mW @ ≈ 1000 Ω ND Integrated E-TENG–PENG (Ma et al.) [ 86 ] PTFE/Cu Contact-separation General rotation 115 mm × 100 mm × 30 mm 45 rpm \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 250 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 4.1 µA Power = 712.3 µW @ 72 MΩ \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 9.8 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 21 mA Power = 30.9 mW @ 200 Ω ND E-TENG–PENG (He et al.) [ 87 ] PDMS patterned/Cu Contact-separation General vibration Φ65 mm × 45 mm 0.5 g \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 13.3 V Power = 4.6 µW @ ≈ 1.4 MΩ \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 362.1 mV Power = 66.5 µW @ ≈ 343.1 Ω ND Stacked E-TENG–PENG (Rodrigues et al.) [ 88 ] PTFE/Nylon Contact-separation Human walking 50 mm × 30 mm × 23 mm 322.5 N \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC ≈ 75 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 38 µA Power = 2.9 mW @ ≈ 1 MΩ \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 2.9 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 46 mA Power = 33 mW @ ≈ 70 Ω \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 75 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 45 mA Power = 32 mW @ ≈ 100 Ω Linear spring generator (Zhang et al.) [ 89 ] PDMS pyramids/Al Contact-separation Human walking 50 mm × 50 mm × 25 mm ND \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 268 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 61 µA Power = 4.9 mW @ 6 MΩ \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 4.9 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 3.6 mA Power = 3.5 mW @ 2 kΩ ND Linear spring generator (Liu et al.) [ 90 ] FEP/Al Contact-separation General vibration Φ48 mm × 14.8 mm 1 g \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 250 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC ≈ 14 µA Power = 1.09 mW @ 15 MΩ \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 0.52 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC ≈ 0.8 mA Power = 0.28 mW @ 200 Ω ND Broadband nonlinear generator (Gupta et al.) [ 91 ] PTFE/ITO Contact-separation General vibration 40 mm × 40 mm × 25 mm 2 g RMS \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 9.5 V RMS \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 70 nA RMS Power = 0.166 µW @ 100 MΩ RMS \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 55 mV RMS \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC ≈ 4 mA RMS Power = 50 µW @ 15 Ω ND Linear spring generator (Quan et al.) [ 92 ] PDMS/Al Contact-separation General vibration 4625 + 348 mm 3 ND \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 84 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 43 µA Power = 1.2 mW @ 2 MΩ \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 9.9 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 7 mA Power = 17.4 mW @ 2 kΩ ND Arc-shaped brace structure (Huang et al.) [ 93 ] PTFE/Cu Contact-separation General vibration 25 mm × 25 mm × 8 mm 8 m/s 2 (~ 0.8 g) \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 28 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 90 µA Power = 675 µW/cm 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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 0.1 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 0.8 mA Power = 80 µW/cm 2 ND Non-contact nanogenerator (Ren et al.) [ 94 ] Fe 3 O 4 nanoparticles PVDF/Al Contact-separation Biomechanical 55 mm × 30 mm × 20 mm 2 Hz \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 88 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 6 µA Power = 0.23 mW @ 25 MΩ \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 4 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 10.5 mA Power = 3.4 mW @ 200 Ω ND Multifunctional hybrid solar (Shao et al.) [ 95 ] PTFE nanowires/Al Contact-separation Blue energy 250 mm × 100 mm × 30 mm 2 Hz \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 142 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 23.3 µA Power avg = 31.5 µW @ ~ 100 MΩ \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 0.66 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 2.14 mA Power avg = 66.9 µW @ 261 Ω ND Honeycomb hybrid (Feng et al.) [ 96 ] PTFE nanowires /Al Contact-separation/Sliding-mode free-standing Blue energy Φ140 mm × 4 mm 3.5 cm, 4 Hz \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 550 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 1.25 µA Energy = 21.7 µJ @ 50 MΩ \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 3 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 4.65 mA Power avg = 8.23 µW @ 350 Ω ND Ship-shaped generator (Wang et al.) [ 97 ] Silicone/Cu Contact-separation/Sliding-mode free-standing Blue energy ND 2 Hz \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 290 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 2.8 µA Power = 165 µW @ 20 MΩ \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 4.3 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 15 mA Power = 9 mW @ 100 Ω ND Sliding linear-grating (Zhang et al.) [ 98 ] Cu/Nylon Sliding-mode free-standing Sliding vibration 120 mm × 40 mm × 16 mm 20 m/s 2 (~ 2 g) \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 118.4 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 0.9 mA Power = 102.8 mW @ 0.4 MΩ \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 55.7 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 7.7 mA Power = 103.3 mW @ 6 kΩ ND Shared-electrode-based (Quan et al.) [ 99 ] FEP/Nylon Sliding-mode free-standing Sliding vibration 10 mm × 5 mm 5 m/s 2 (~ 0.5 g) \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 245 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 2.8 µA Power = 0.22 mW @ 200 MΩ \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 0.13 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 3.8 mA Power = 0.08 mW @ 40 Ω ND Resonant wideband cantilever (Zhu et al.) [ 100 ] PTFE/Cu stopper Free-standing General vibration ND 2 g \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC ≈ 4.5 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC ≈ 42 nA Power avg = 30 µW/m 2 @ ≈ 100 MΩ Power avg = 38.1 µW/m 2 @ 40 Ω ND Fully enclosed resonant cantilever (Quan et al.) [ 101 ] FEP/Cu Contact-separation General vibration 80 mm × 55 mm × 50 mm ND \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 24 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 20 µA Power = 130 µW @ 0.8 MΩ \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 0.8 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 0.5 mA Power = 80 µW @ 3 kΩ ND Triboelectric-piezoelectric-electromagnetic cantilever (Du et al.) [ 102 ] PTFE nanoparticles/PET Contact-separation Rotating tire 35 mm × 40 mm × 57 mm ND \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC ≈ 0.1 mA Power = 1.2 mW @ 0.4 MΩ \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC ≈ 0.8 mA Power = 7.4 mW @ 40 kΩ \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC ≈ 0.6 mA Power = 8 mW @ 20 kΩ Air-flow harvester (Wang et al.) [ 48 ] PTFE/Kapton Contact-separation Air-flow 67 mm × 45 mm × 20 mm 18 m/s \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 55.7 µA Power = 3.5 mW @ 3 MΩ \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 3.3 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 2.3 mA Power = 1.8 mW @ 2 kΩ \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC ≈ 4 mA Air-flow harvester (Wang et al.) [ 103 ] FEP/Cu Contact-separation Air-flow 130 mm × 15 mm × 22 mm 18 m/s \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 51 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 40 µA Power = 1.7 mW @ 10 MΩ \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 3 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 3.7 mA Power = 2.5 mW @ 1 kΩ ND Two-dimensional wave harvester (Hao et al.) [ 104 ] Silicone pyramids/Al Sliding-mode free-standing Wave energy 100 mm × 88 mm × 43 mm ND \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 75 V \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC = 1.2 µA Power = 80 µW @ 100 MΩ \\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}$$V_{{{\\text{OC}}}}$$\\end{document} V OC = 9 V = 1 mA Power = 14.9 mW @ 1 kΩ ND Wrist generator (Quan et al.) [ 105 ] PVB nanowires-PDMS/Nylon Contact-separation Human motion 36 mm × 36 mm × 30 mm ND \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC ≈ 7 µA Power = 100 µW @ 6 MΩ \\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}$$I_{{{\\text{SC}}}}$$\\end{document} I SC ≈ 4 mA Power = 2.8 mW @ 700 Ω ND Water wave energy harvester (Wu et al.) [ 106 ] PTFE/Cu Sliding-mode free-standing Wave energy Φ100 mm 30 rpm V OC ≈ 60 V I SC ≈ 1.2 µA V OC ≈ 1.7 V I SC ≈ 1 mA ND 3D full-space nanogenerator (He et al.) [ 38 ] Silicone pyramids/ Polystyrene Sliding-mode free-standing General 3D vibration Φ120 mm 100 mm V OC = 77 V I SC = 0.7 µA Power = 18 µW @ 200 MΩ V OC = 2.4 V I SC = 1 mA Power = 640 µW @ 1 kΩ ND Wrist-wearable generator (Maharjan et al.) [ 107 ] PTFE nanowires/NdFeB Sliding-mode free-standing Human walking Φ13 mm × 120 mm ND V OC = 19 V I SC = 3 µA Power = 0.223 µW/cm 3 @ 13 MΩ V OC = 8 V I SC = 18 mA Power = 5.15 mW/cm 3 @ 49 Ω ND Magneto-mechano-triboelectric generator (Lim et al.) [ 108 ] PFA nanostructured/Al Contact-separation AC current/Magnetic field 60 mm × 20 mm 7 Oe V OCpp = 708 V I SC = 277 µA Power = 21.8 mW @ 2 MΩ ND ND ND Not described Many designs of hybridized E-TENGs have already been proposed and tested in applications such as general vibration energy harvesting, wheel rotation energy, biomechanical energy, blue energy (wave energy and fluid flow), wind energy, thermal energy, among others. These generators achieved maximum output peak powers up to ca. 100 mW [ 66 , 75 , 78 ], average powers around 1 mW [ 63 , 70 , 71 ] and peak power densities up to 1 mW cm −3 [ 72 , 75 , 85 , 88 , 93 , 98 , 107 ]. However, no literature reviews are currently available on this topic except of more general ones [ 109 , 110 ]. Then, this study systematically an deeply analyzes, for the first time, major breakthroughs in emerging hybrid E-TENG technologies, with the emphasis on theoretical transduction mechanisms, structural designs and applications, experimental electromechanical output parameters and future outlook."
} | 22,335 |
38562581 | PMC10983830 | pmc | 455 | {
"abstract": "Abstract Genetically engineered silkworms have been widely used to obtain silk with modified characteristics especially by introducing spider silk genes. However, these attempts are still challenging due to limitations in transformation strategies and difficulties in integration of the large DNA fragments. Here, we describe three different transformation strategies in genetically engineered silkworms, including transcription-activator-like effector nuclease (TALEN)-mediated fibroin light chain (FibL) fusion (BmFibL-F), TALEN-mediated FibH replacement (BmFibH-R), and transposon-mediated genetic transformation with the silk gland-specific fibroin heavy chain (FibH) promoter (BmFibH-T). As the result, the yields of exogenous silk proteins, a 160 kDa major ampullate spidroin 2 (MaSp2) from the orb-weaving spider Nephila clavipes and a 226 kDa fibroin heavy chain protein (EvFibH) from the bagworm Eumeta variegate , reach 51.02 and 64.13% in BmFibH-R transformed cocoon shells, respectively. Moreover, the presence of MaSp2 or EvFibH significantly enhances the toughness of genetically engineered silk fibers by ∼86% in BmFibH-T and ∼80% in BmFibH-R silkworms, respectively. Structural analysis reveals a substantial ∼40% increase in fiber crystallinity, primarily attributed to the presence of unique polyalanines in the repetitive sequences of MaSp2 or EvFibH. In addition, RNA-seq analysis reveals that BmFibH-R system only causes minor impact on the expression of endogenous genes. Our study thus provides insights into developing custom-designed silk production using the genetically engineered silkworm as the bioreactor.",
"introduction": "Introduction Silks are natural protein fibers synthesized by arthropod, most notably silkworms and spiders ( 1–3 ). Silk fibers have attracted both scientific and economic interest in the fields of biology, medical, chemistry, and materials science, owing to their extraordinary mechanical properties, favorable biocompatibility and biodegradability ( 4–7 ). The mechanical properties of silk are predominantly determined by the primary structure and condensed state structure of silk protein ( 7–10 ). As one of the most excellent silk fibers in nature, spider dragline silk combines high strength and superior extensibility ( 11–14 ). The strength of dragline silk from the golden orb-weaver Nephila clavipes is comparable with technological materials, such as high-tensile steel and Kevlar ( 15 , 16 ). Moreover, N . clavipes dragline silk shows highly overall toughness, which depends strongly on the extensibility ( 13 , 16–18 ). The major ampullate spidroin 2 (MaSp2), one of the major compositions of the dragline silk, is proposed to be responsible for the extensibility. MaSp2 consists of regular repetitive motifs and nonrepetitive regulatory N-terminal and C-terminal domains (NTD and CTD, respectively) ( 13 , 19–21 ). The repetitive motifs of MaSp2 can be classified into two categories: polyalanine motifs (polyA/GA) and GPGGX (X = typically Y, P, or Q) motifs. The polyA/GA motifs form β-sheet and assemble ordered crystalline region of MaSp2, while the GPGGX motifs are proposed to be involved in β-turn spiral formation, giving rise to the disordered noncrystalline regions of MaSp2 ( 13 , 22–24 ). Compared with major ampullate spidroin 1 (MaSp1), the spidroin supposed to be implicated in high-tensile strength of silk fiber, MaSp2 is poorly studied ( 25 , 26 ). The main difference between MaSp1 and MaSp2 lines in the proline content in sequence and polymeric patterns ( 13 , 27 ). Based on the characteristics of amino acid sequence and block structure, proline-related motif GPGGX is hypothesized to be responsible for the extensibility of the fiber ( 13 , 23 , 28 ). The bagworm, Eumeta variegate , produces silk that combines with plant materials, forming portable bag against natural enemies ( 29 ). In recent years, the bagworm silk has been paid great attention due to its superior mechanical properties including extraordinarily high modulus, strength, and excellent tensile deformation behavior ( 30–32 ). Previous study reported that the silk from the Darwin's bark spider Caerostris darwini is the toughest known biomaterial compared with any previously described silk ( 33 , 34 ). Surprisingly, the toughness of bagworm silk is even higher than that of C . darwini silk. The mechanical properties of bagworm silk are predominantly attributable to the contribution of silk fibroin heavy chain (EvFibH) ( 31 , 32 ). There are three characteristic motifs, polyA, poly (GA), and GGX in EvFibH amino acid sequence, combining characteristics of that in Bombyx mori, spider dragline silk and Saturniidae moths, despite several decisive distinctions from their motifs ( 32–36 ). The silk from the domesticated silkworm, B. mori , is the only secretory natural protein fiber that be produced and utilized in large scales. Silk from B. mori is composed of fibroin proteins coated with adhesive sericin proteins. Fibroins are core fibrous proteins primarily determining mechanical properties of silk fibers. The fibroin is a 2,300 kDa molecular complex consists of the fibroin heavy chain (BmFibH, 350 kDa), the fibroin light chain (BmFibL, 25.8 kDa), and the fibrohexamerin protein (BmP25, 25.7 kDa) with a molar ratio of 6:6:1 ( 37–39 ). More importantly, B. mori is an ideal bioreactor for expressing exogenous proteins attributing to high capacity for protein synthesis in the silk gland (SG) and the efficient genetic manipulation technology has been established ( 40–42 ). Numerous efforts have attempted to improve mechanical properties of silk fibers via expressing spider silk proteins ( 20 , 23 , 26 ). In the previous work, we achieved massive spider silk protein production through performing transcription-activator-like effector nuclease (TALEN)-mediated homology-directed repair to replace the BmFibH with partial MaSp1 gene ( 39 ). It is also reported that production of spider silk proteins in the silkworm through employing clustered regularly interspaced short-palindromic repeats (CRISPR) and CRISPR-associated protein 9 (CRISPR/Cas9) technology can significantly improve mechanical properties of silk fibers ( 43 , 44 ). However, such study is still challenging due to limitations in transformation strategies and difficulties in integration of the large DNA fragments. In the current study, we established comprehensive transformation strategies including TALEN-mediated BmFibL fusion system (BmFibL-F), TALEN-mediated BmFibH replacement system (BmFibH-R) and piggyBac -mediated transgenic system (BmFibH-T) in B. mori . The presence of the 160 kDa MaSp2 protein significantly enhanced the extensibility of silk fiber in transformed silkworm lines. In comparison, the BmFibH-R system achieved the highest yield of MaSp2 production with minor influence on the silkworm physiology. The RNA-seq analysis also revealed that expression of endogenous genes only slightly affected in BmFibH-R silkworms. We further established two silkworm lines to express the chimeric 226 kDa EvFibH protein and the transformed EvFibH silk fiber showed significantly improved mechanical properties, especially the strength. Notably, the protein expression amount of MaSp2 and EvFibH by using the BmFibH-R system reach unprecedented yield with 51.02 and 64.13%, respectively, in transformed cocoon shells. Our data indicated that MaSp2 and EvFibH were responsible for extensibility and strength of transformed silk fibers, respectively, paving the way for future custom-designed silk production using B. mori as a promising bioreactor.",
"discussion": "Discussion In the current study, we achieved ectopic expression of MaSp2 and EvFibH proteins in genetically engineered silkworms by using various expression strategies including BmFibL-F, BmFibH-R, and BmFibH-T systems. In silkworm lines constructed by BmFibH-T system, negligible difference was detected in physiological property compared with that of WT animals. The comprehensive mechanical properties of transgenic silkworm fibers were significantly promoted. Nevertheless, the yields of exogenous protein were not so desirable, consistent with previous studies, limiting further improvement of silk mechanical properties ( 26 , 46 ). We presumed the transgenic system was fit for verifying properties of sequence with particular motifs. By comparison, BmFibL-F system can achieve higher yield of exogenous protein; however, the transformed silkworms displayed evident defects in PSG development and endogenous gene expression. In addition, the comprehensive mechanical properties of fibers from transformed silkworms were not satisfactory compared with other lines, indicating BmFibL-F system is not suitable for improving mechanical properties of silk fibers. In contrast, the silkworm lines constructed by BmFibH-R system can produce extraordinary amount of exogenous protein with minimal abnormality in physiological properties. More importantly, the mechanical properties of silk fibers from transformed silkworms were drastically improved, suggesting that BmFibH-R system can serve as the ideal strategy. In addition, the BmFibH-R system can achieve up to 12 kb fragment integration precisely, revealing its tremendous potential for expressing protein with high molecular weight. In our study, the BmFibH-R system achieved expression of a 226 kDa protein, which has higher molecular weight than the 150 kDa minor ampullate spidroin (MiSp) reported in the recent study ( 43 ). Compared with the previous study inserting spider silk genes into intron of BmFibH and BmFibL thorough CRISPR/Cas9 mediated nonhomologous end joining, the BmFibH-R strategy can achieve precise seamless insertion ( 44 ). In our study, the efficiency of TALEN-mediated transformation strategy was lower than that of piggyBac-based transgenic system. We speculated this was attributed from multiple insertion sites in piggyBac -mediated transformation strategy. The physical properties of silk fibers are primarily determined by the primary structure and secondary structure of silk protein ( 7–10 ). Based on the sequence characteristics, MaSp2 and EvFibH were supposed to be responsible for extensibility and strength of silk fibers, respectively. Nonetheless, mechanical properties of silk fibers cannot be easily explained by the condensed state structures made from diverse motifs. Our study provided experimental evidence that MaSp2 and EvFibH are implicated in extensibility and strength of silk fibers, providing sequence materials for exploiting custom-designed silks. In our previous study, we achieved 67 kDa chimeric MaSp1 protein expression with a yield of 35.2% in a single cocoon shell using the BmFibH-R system. However, it was difficult to obtain single fibers from the cocoon shells of homozygous transformed silkworms, which also showed apparent defects in PSG development, likely being responsible from low molecular weight of MaSp1 compared with that of original 350 kDa BmFibH ( 39 ). In the current study, we obtained transformed silkworm lines with the expression of 160 kDa MaSp2 and chimeric 226 kDa EvFibH, two proteins with much higher molecular weight compared with that of MaSp1, by using the BmFibH-R system. Remarkably, the amounts of exogenous proteins can reach 51.02 and 64.13% in cocoon shells of H-Sp2-R and H-EvFH-R lines, respectively, verifying our previous hypothesis that integrating proteins of higher molecular weights can further increase protein yield. Moreover, mechanical properties of silk fibers from the homozygous transformed silkworms, especially H-EvFH-R line, were vastly enhanced. In addition, the higher molecular weight of integrated protein appeared to alleviate the unfavorable consequence caused by BmFibH deletion on SG development. It is worth mentioning that the comprehensive mechanical performance of H-EvFH-R fiber is much better than that of H-EvFH-T line. By comparison, fiber of H-Sp2-T line displayed better mechanical properties than that of H-Sp2-R line, indicating that higher content of exogenous protein with higher molecular weight can further promote mechanical properties of silk fibers. In this study, chimeric EvFibH protein consists of 21 repetitive motifs, we speculated the mechanical properties of silk fibers and yield of protein can be further improved when increasing repetition times of the motifs. Numerous studies have been attempted to promote mechanical properties of silk fibers through expression exogenous proteins, notably spider silk proteins. Nonetheless, the majority of studies were restricted to the expression of one kind of protein with certain physical properties, such as extensibility, strength, or stickiness ( 26 , 43 , 44 , 46–49 ). Future clarifying relationship between the sequence structure and physical properties of silk proteins will be useful for integrating chimeric proteins combining with multiple characteristic motifs. Overall, our study provides the ideal strategy for exploiting custom-designed, mass silk production in genetically modified silkworms."
} | 3,278 |
23741341 | PMC3669319 | pmc | 457 | {
"abstract": "A central focus in studies of microbial communities is the elucidation of the relationships between genotype, phenotype, and dynamic community structure. Here, we present a new computational method called community flux balance analysis (cFBA) to study the metabolic behavior of microbial communities. cFBA integrates the comprehensive metabolic capacities of individual microorganisms in terms of (genome-scale) stoichiometric models of metabolism, and the metabolic interactions between species in the community and abiotic processes. In addition, cFBA considers constraints deriving from reaction stoichiometry, reaction thermodynamics, and the ecosystem. cFBA predicts for communities at balanced growth the maximal community growth rate, the required rates of metabolic reactions within and between microbes and the relative species abundances. In order to predict species abundances and metabolic activities at the optimal community growth rate, a nonlinear optimization problem needs to be solved. We outline the methodology of cFBA and illustrate the approach with two examples of microbial communities. These examples illustrate two useful applications of cFBA. Firstly, cFBA can be used to study how specific biochemical limitations in reaction capacities cause different types of metabolic limitations that microbial consortia can encounter. In silico variations of those maximal capacities allow for a global view of the consortium responses to various metabolic and environmental constraints. Secondly, cFBA is very useful for comparing the performance of different metabolic cross-feeding strategies to either find one that agrees with experimental data or one that is most efficient for the community of microorganisms.",
"introduction": "Introduction In nature, microbes generally occur in communities. These microbial communities play important roles: they are essential for global nitrogen, carbon and energy cycling [1] and contribute to a healthy human physiology as part of our oral and gut flora [2] . In such complex systems, the physiology, behavior, and fitness of the species are interdependent. It is a major challenge to understand how the interplay between microbes determines community dynamics and robustness, and how the genotype of each of the microorganisms ultimately influences ecosystem properties. Today, advanced molecular methods (meta-omics) facilitate the detailed characterization of microbial communities, providing information at an unprecedented level of molecular detail. These methods catalogue the active molecular processes, the ecotypes present, and report the identity and abundances of specific microbial species [3] . While such approaches are generally high-throughput, comprehensive and broadly applicable, they give little insight into the rationales behind the metabolic behaviors of individual microbial species. Why do microbes choose a particular physiological state out of their full range of metabolic capacities? How do these decisions depend on the metabolic coupling between species? Which metabolic interactions determine community structure and how do selective pressures influence this? Answering these questions will require integrative computational approaches that link genes to species metabolisms and community-level structure and offer a consistent framework for describing community level interactions [4] , [5] . The promise of these methods, combined with in depth molecular characterization, is the rational design, manipulation and control of microbial communities in biotechnology and medicine. Constraint-based stoichiometric modeling of genome-scale metabolic networks is a set of computational methods developed in systems biology for studying the comprehensive metabolic capacities of organisms [6] , [7] . This collection of computational methods considers the entire metabolic network of an organism as reconstructed from genomic and physiological information [8] . Flux distributions in metabolic networks for optimal biomass or product formation can be predicted from the resulting genome-scale stoichiometric models with flux balance analysis (FBA), for instance as function of the nutrient conditions and as a response to enzyme knock-outs [6] . These models generally compute steady states of metabolic networks and consider only reaction stoichiometry and omit enzyme kinetic information [9] . Constraint-based stoichiometric modeling of genome-scale metabolic networks is widely used in biotechnology and medicine [7] . In microbial communities, a new level of complexity is added on top of microbial metabolism that complicates the application of constraint-based stoichiometric modeling to microbial communities. Besides the presence of all metabolic reactions in each of the microorganisms, the exchange of metabolites between species and biomass abundances of each of the microbial species has to be considered. In addition, each of these microorganisms has specific nutrient requirements for growth, which it can meet through metabolic cross-feeding, nutrient-competition or by uptake from the environment. On top of that, selective pressures at the level of single species change the metabolic interactions between species through mutations, which leads to accumulation of genetic variants and co-evolution of metabolic partnerships. These forces together shape the structure of microbial communities. In such systems, the actions of individual species are constrained by their own biochemical processes and by their interactions with other species. Computational methods are essential to address those complex aspects of biological systems [10] . Considerable effort has been invested in recent years to develop suitable computational approaches that, in principle, can consider all the metabolic reactions occurring in a microbial community [11] , [12] , [13] , [14] , [15] , [16] , [17] , [18] , [19] , [20] . These studies differ in the computational and mathematical methods employed, some are for instance limited to mutualistic metabolic interactions or compartmentalized approaches. Compartmentalized approaches [12] , [16] , [21] , [11] , [22] to microbial communities consider only metabolite exchanges without the explicit consideration of biomass abundances of individual species, even though this aspect of microbial community composition is a major research subject in microbial ecology. More advanced methods can consider competition for resources and variable biomass abundance. They typically make use of dynamic flux balance analysis [23] , [24] . The recently introduced method ‘OptCom’ [20] is arguably the most advanced method and uses sophisticated multi-objective optimization techniques to predict the biomass composition of the community along with the growth rates of each species. OptCom takes a multi-objective optimization (Pareto optimization) approach to interrelate the objectives of individual organisms. The reason why constraint-based stoichiometric modeling of microbial communities is much more complicated than for single organisms originates from the interdependencies between the metabolic objectives of microorganisms in the community. Generally, in constraint-based stoichiometric modeling, a single metabolic objective is postulated for the organism, such as optimization of biomass yield, to give rise to a manageable solution space of flux distributions [25] . In a microbial community, the metabolic performance (fitness) of each organism is dependent on all others, either directly or indirectly. As a consequence, the community metabolic state in a constraint-based stoichiometric model of the microbial community would have to emerge from the multi-objective optimization of each of those performances, taking into account trade-offs and possibly allowing for suboptimal strategies. Hence, a nonlinear multi-objective optimization perspective appears most appropriate and this is indeed the approach taken by OptCom. In this paper, we show that the multi-objective optimization task greatly simplifies when microbial communities are considered that engage in balanced growth. This leads to the formulation of a new approach for constraint-based modeling of microbial communities, which we shall refer to as community FBA (cFBA). It requires fewer assumptions than full-blown multi-objective optimization approaches and is much easier to interpret. Balanced growth occurs when internal metabolism is at steady state while the cells grow exponentially at a fixed growth rate. Here, we extend this definition to microbial ecosystems and, hence, require all metabolites (intra- and extracellular) to attain a steady state level. Under those growth conditions, our computational method, community flux balance analysis (cFBA), predicts the fractional biomass abundances of all the participating microorganisms in the community as well as the intra- and extracellular flux distributions and metabolic exchanges. cFBA predicts the complete state of the microbial community engaging in balanced growth and postulates only a single objective. cFBA applies to microbial ecosystems that function in a fairly constant environment, such as specific microbial communities involved in bioremediation, waste water treatment or in laboratory settings, e.g. chemostats. The community flux balance analysis (cFBA) that we will present in this paper is a direct translation of FBA for single organisms to microbial communities and requires only a few community-specific constraints. It is based on the concept of balanced growth of microorganisms and the corresponding metabolic network states. It is therefore a fundamental description of the microbial community structure resulting from basic concepts of microbiology. It directly applies to the study of stable microbial communities under laboratory conditions in controlled bioreactors, either resulting from laboratory evolutionary experiments or from direct samples from the environment, or for microbial communities in nature that are exposed to prolonged stable environmental conditions. cFBA predicts the optimal flux distribution, growth rate, and abundance of all species in the consortium as well as the exchange fluxes between species and the community environment. As a proof of concept, we first investigate a simple microbial community in which species are mutually dependent for their growth. Next, we study an evolved syntrophic E. coli consortium with genome-scale models.",
"discussion": "Discussion In this paper, we presented an approach to interrelate genotype, phenotype, and community structure for microbial communities at steady state. Our cFBA method allows for the prediction of metabolic fluxes, community growth rate, and the fractional biomass abundance given (genome-scale) stoichiometric models of the participating species and constraints derived from biochemistry, thermodynamics, microbial physiology, and ecology. We derived the cFBA method from the microbiological principles of balanced growth and mass flow in microbial communities. We thus extended the concept of balanced growth of a single organism in microbiology to microbial ecology. At present, cFBA is limited to microbial communities at balanced growth, such that all microorganisms grow equally fast and have an intracellular metabolism operating at steady state. The resulting condition of equal specific growth rate can be directly used as the objective function that is to be maximized computationally. The optimization leads to the identification of an optimal community structure. This structure encompasses the rates of all the metabolic fluxes in the community and fractional biomass abundances. cFBA applies to microbial communities where the environmental changes are slow enough for the entire community to settle to a steady state. Communities involved in wastewater treatment or bioremediation can attain steady state levels when done in specific bioreactors [32] , [33] , [34] . But also communities in natural environments can be exposed to fairly constant environmental conditions, such as communities found in the (deep) subsurface or on inert surfaces [23] . Another application of cFBA is the study of mixed cultures in controlled bioreactors for new environmental biotechnological applications, such as the production of bioelectricity [35] or bio-plastics [36] . Constraint-based stoichiometric modeling approaches for the metabolic networks of microbial communities, such as cFBA or other methods [11] , [12] , [13] , [14] , [15] , [16] , [18] , [19] , [20] , can be a great tool to supplement experimental microbial community analysis. These computational methods can address specific questions unanswered by molecular characterization of communities and the mathematical models are natural ways to integrate heterogeneous data. For instance, after identification of the microbial species (or ecotypes) making up the community and the (partial) functional annotation of their genome, the metabolic network can be reconstructed from this genome information [8] . Then, depending on the level of genome annotation, the majority of the metabolic capacities of the microorganisms are known. How those metabolic capacities together give rise to ecosystem level properties regarding biomass abundances, growth rate, and metabolic activities can subsequently be addressed with constrained-based stoichiometric modeling approaches. The output of those computational methods can be directly compared to available experimental data about fluxes and biomass abundances. cFBA extends the growing arsenal of such computational methods and has major advantage compared to previous methods that it is straightforward and predicts the entire state of a microbial community, including biomass abundances of individual species. It can address an unlimited number of species and any type of species interaction. Future extensions of cFBA will consider dynamic scenario’s where one or several nutrients are limiting consortium growth and depleting slowly. The constraint-based stoichiometric modeling of microbial communities is still largely in its infancy. Much is still to be learnt from studies where the predictions of these modeling approaches are critically compared to experimental data. How can we study microbial communities in a sensible fashion by using an optimization approach? Do we need to consider multi-objective nonlinear optimization approaches? If so, the computational approaches will quickly run into problems related to computational speed and uniqueness of solutions. However, even if not all the assumptions hold, as is also often the case for single-species FBA, these models will be useful to explore the metabolic potential of microbial communities. We thus expect approaches such as cFBA to be already sufficiently informative to become a vital component of the workflow of studying communities with modern “omics” approaches."
} | 3,738 |
34026457 | PMC8132064 | pmc | 459 | {
"abstract": "Abstract Spin‐torque memristors are proposed in 2009, and can provide fast, low‐power, and infinite memristive behavior for neuromorphic computing and large‐density non‐volatile memory. However, the strict requirements of combining high magnetoresistance, stable domain wall pinning and current‐induced switching in a single device pose difficulties in physical implementation. Here, a nanoscale spin‐torque memristor based on a perpendicular‐anisotropy magnetic tunnel junction with a CoFeB/W/CoFeB composite free layer structure is experimentally demonstrated. Its tunneling magnetoresistance is higher than 200%, and memristive behavior can be realized by spin‐transfer torque switching. Memristive states are retained by strong domain wall pinning effects in the free layer. Experiments and simulations suggest that nanoscale vertical chiral spin textures can form around clusters of W atoms under the combined effect of opposite Dzyaloshinskii–Moriya interactions and the Ruderman–Kittel–Kasuya–Yosida interaction between the two CoFeB free layers. Energy fluctuation caused by these textures may be the main reason for the strong pinning effect. With the experimentally demonstrated memristive behavior and spike‐timing‐dependent plasticity, a spiking neural network to perform handwritten pattern recognition in an unsupervised manner is simulated. Due to advantages such as long endurance and high speed, the spin‐torque memristors are competitive in the future applications for neuromorphic computing.",
"conclusion": "3 Conclusion In conclusion, a spin‐torque memristor with a high TMR ratio, a low RA, a low working current, and a nanoscale size is obtained by engineering an atomic‐thickness W spacer in the free layer of a MTJ. A memristive TMR is achieved during STT‐induced switching in both directions. By comparing the intrinsic pinning field for domain wall motion in the free layer films and the filed required to break the intermediate TMR at low temperature, it is proved that the memristive behavior of the device originates from strong domain wall pinning effects. NV center measurements prove that the magnetization of the free layer is quite homogenous in a saturated state. Experiments and micromagnetic simulations show that a chiral vortex domain wall could form at a cluster of W atoms because of the opposing interfacial DMIs and the RKKY interaction. The energy fluctuation induced by the domain wall structural transition between trivial coupled Bloch configuration and vertical chiral vortex configuration may be one reason for the strong pinning effect. The STDP functionality of the device has been experimentally demonstrated based on its synaptic property. A compact model is developed based on experimental data and system‐level simulation is performed, showing that our spin‐torque memristor can be performant and competitive for neuromorphic computing, such as unsupervised learning.",
"introduction": "1 Introduction Memristors are considered to be essential elements for realizing neuromorphic computing. [ \n \n 1 \n , \n 2 \n , \n 3 \n \n ] Traditional memristors rely on ion motion and ionic valence changes in materials. [ \n \n 1 \n , \n 4 \n , \n 5 \n \n ] However, most of them suffer from certain limitations, such as finite endurance [ \n \n 6 \n \n ] or relatively low switching speed, [ \n \n 1 \n \n ] which hinder their real applications in systems requiring long endurance such as neural networks with “on‐chip” learning. Spintronic devices, in which the state is modulated by magnetic variation and thus promise a much longer endurance, provide an alternative solution. [ \n \n 7 \n , \n 8 \n \n ] The concept of a spin‐torque memristor based on the current‐induced magnetic domain wall motion in the free layer of a magnetic tunnel junction (MTJ) was first proposed in 2009. [ \n \n 9 \n \n ] Nevertheless, a real device with nanoscale dimension and all‐spin‐torque operation, for example, without the assistance of an external magnetic field, is still missing. The intermediate tunneling magnetoresistance (TMR) is difficult to stabilize against thermal activation or stimulating currents, especially in devices with nanoscale dimensions. [ \n \n 10 \n \n ] A free layer in the partially switched state with domain walls usually has higher energy than monodomain states because both Heisenberg exchanges and magnetic anisotropy favor a collinear spin texture. Several possible solutions have been proposed, such as creating an intermediate state with the assistance of shape anisotropy, [ \n \n 11 \n \n ] manipulating memristive switching through domain wall pinning in some complex geometries, [ \n \n 12 \n , \n 13 \n \n ] or by engineering the reference layer. [ \n \n 14 \n \n ] However, these solutions require a large device size or an external magnetic field to realize memristive behaviors. The interfacial Dzyaloshinskii–Moriya interaction (DMI), [ \n \n 15 \n , \n 16 \n \n ] a form of antisymmetric exchange that favors a chiral spin texture, makes it possible to obtain intrinsically stable noncollinear magnetic structures in nanometer‐scale magnet and provides new possibilities to realize memristive MTJs with nanoscale dimensions. [ \n \n 17 \n , \n 18 \n \n ] \n In this work, we experimentally demonstrate a nanoscale spin‐torque memristor based on a perpendicular‐anisotropy MTJ with W‐inserted free layer. A high TMR ratio, a low resistance‐area product (RA), and spin‐polarized‐current‐induced switching are achieved. The memristive behavior is proved to originate from strong domain wall pinning effects in the free layer. Measurements with nitrogen‐vacancy (NV) color center in diamond indicate that the saturated magnetization is quite homogeneous in the free layer of MTJ stacks. Whereas, experiments and simulations suggest that a type of vertical chiral spin vortex could form around clusters of W atoms under the combined effect of Ruderman–Kittel–Kasuya–Yosida (RKKY) interactions and interfacial DMIs. [ \n \n 15 \n , \n 16 \n \n ] This spin texture leads to the fluctuations of the domain wall surface energy, which may be one reason for the strong domain wall pinning effects. The spike‐timing‐dependent plasticity (STDP) functionality is experimentally validated. A compact model of the spin‐torque memristors is created according to experimental data and is integrated into a spiking neural network (SNN), which was then employed for the Mixed National Institute of Standards and Technology database (MINST) handwritten pattern recognition. [ \n \n 19 \n \n ] Arabic handwritten numeral can be recognized with our devices. Thanks to advantages such as long endurance, [ \n \n 7 \n , \n 20 \n , \n 21 \n \n ] the spin‐torque memristor studied in this work provides a new opportunity for the application of spintronic devices in neuromorphic computing with “on‐chip” learning.",
"discussion": "2 Results and Discussion 2.1 Memristive MTJs \n Figure \n 1 a introduces the layer structure used to fabricate the memristive MTJ device: synthetic antiferromagnet/W (0.25)/CoFeB (1.0)/MgO (0.8)/CoFeB (1.3)/W (0.2)/CoFeB (0.5)/MgO (0.75)/Ta (3.0) (thickness in nanometer). The stack is prepared with a Singulus magnetron sputtering machine and is annealed at 390 °C for 1 h after deposition. In contrast to traditional MTJs with a single‐free layer structure, in this study, an atomic‐thickness W layer is inserted between two free layers during sputtering deposition to engineer the free layer properties. [ \n \n 22 \n , \n 23 \n \n ] A transmission electron microscopy (TEM) image of the multilayer stack is presented in Figure 1b . Figure 1 Layer structure and electric test of the device. a) The stack structure of the MTJ. b) Cross‐sectional TEM image showing the layer quality of the stack. c) Optical microscopy image of an MTJ with four electrodes. d) STT‐induced switching using stepwise increasing voltage, with each step lasting 100 ms. A 20‐mT perpendicular field is applied during the test to compensate the bias of stray fields from the reference layer, the same is also true throughout the text. The MTJ film with composite free layer is patterned into circular nanopillars with a 200‐nm radius (R) using electron beam (e‐beam) lithography and Ar ion milling and is instrumented with gold electrodes, as shown in Figure 1c . Then, continuous and stepwise increasing voltage is applied and resistance of the device is measured simultaneously. As shown in Figure 1d , the spin‐transfer torque (STT)‐induced magnetization switching is achieved, with multiple intermediate states on both switching directions. A TMR ratio as large as 200% and an RA of 7 Ω µm 2 are obtained at room temperature. The switching current is on the order of MA cm −2 . To investigate the stability of the intermediate states, voltage pulses with a duration T \n P = 100 ms are applied and after each pulse, the resistance is immediately measured with a 0.01‐V reading voltage, as shown in Figure \n 2 a . The voltage‐dependent decrease of the antiparallel resistance is avoided using this low reading voltage. [ \n \n 24 \n \n ] Intermediate states are reproduced, which correspond well with those in Figure 1d , confirming their good stability. Furthermore, sequences of shorter voltage pulses with T \n P = 200 ns are applied to observe the STT‐switching and more intermediate states appeared, as shown in Figure 2b . As can be observed from the minor loops, the intermediate states are stable even when the polarity of the applied voltage is reversed. It is worth noting that the pulse width in our tests cannot be further compressed because of the oscillating noise. Much more intensive intermediate states can be expected if the stimulating pulse width scales down to nanoseconds, which is the typical spin‐flip time in STT‐induced switching. Figure 2c gives the current–voltage loop obtained by a train of T \n P = 200 ns switching voltage pulses. One can find that once the applied voltage increases over a threshold value, the resistance of the device varies quasi‐continuously, which is typical memristive behaviors. Figure 2 Spin‐torque‐induced and magnetic‐field‐induced switching of an MTJ with R = 200 nm. a) Resistance–voltage loop. In this measurement, a train of voltage pulses with a duration T \n P of 100 ms and an increase of 0.02 V per step is applied. The resistance is measured using a voltage of 0.01 V after each stimulus pulse. b) Resistance–voltage minor loops. In this measurement, short voltage pulses with T \n P = 200 ns are applied. c) Current–voltage loop obtained using a train of voltage pulses with T \n P = 200 ns. The resistance is measured when switching voltage is applied in both B and C. d) Black: full resistance‐perpendicular field hysteresis loop of the device; red and blue: stability of intermediate state against external magnetic fields at 35.5 K. The stability of the intermediate states against external magnetic fields is checked. First, the resistance‐magnetic field hysteresis loop in a device with R = 200 nm is measured at a low temperature, as shown in Figure 2d . As expected, the coercive field is very large in such a small device and no intermediate state is observed. Whereas, after creating an intermediate state with a STT current and then applying an external field, we find that the intermediate state remained stable under certain external magnetic fields. The field needed to destroy the intermediate state is ≈15.5 mT after offsetting the bias caused by the stray field from the reference layer. 2.2 Strong Domain Wall Pinning Effect in the Composite Free Layer The magnetoresistance of an MTJ depends on the relative magnetic state of the free layer with respect to the pinned layer. [ \n \n 24 \n \n ] A stable intermediate state is rare in traditional single‐free layer MTJs with submicron dimensions since once magnetic switching begins, the nucleated domain wall moves forward immediately and leads to the complete switching. [ \n \n 10 \n \n ] Here, we get a different process in the composite free layer MTJs. To obtain a deeper insight into the properties of the composite free layer, we deposited a MgO/CoFeB/W/CoFeB/MgO film with the same structure as that in the MTJ stack (called FL‐film in the following). The deposition and annealing conditions remained the same. As seen from the hysteresis loop of this film (see Figure \n 3 a ) and the MTJ stack (see the Supporting Information 1 (online)), the two CoFeB layers are globally ferromagnetically coupled and show strong perpendicular magnetic anisotropy. An intermediate state of the MTJ should not be caused by inconsistent magnetization of the upper free layer (Up‐layer) and the lower free layer (Lw‐layer). The field‐induced magnetization reversal of the free layers appears to be gradual, indicating that the threshold field for domain wall propagation is much larger than the domain wall nucleation field in this film. Next, the field‐induced domain wall motion in the FL‐film is imaged using Kerr microscopy. Figure 3b shows a dendritic trace after domain wall motion induced by a magnetic field of 3.6 mT, which appears much rougher than an ordinary single‐layer CoFeB film with low pinning effect, [ \n \n 25 \n \n ] indicating a strong pinning effect in the composite free layer film. The velocity of the field‐induced domain wall motion is measured and it leaves the thermally activated creep regime [ \n \n 26 \n \n ] until the driving field reaches ≈ μ \n 0 H C = 16 mT, as shown in Figure 3c . This critical field (conventionally called intrinsic pinning field) can be seen as an indicator of the domain wall pinning strength in a magnetic material, below which the wall motion is not possible without the assistance of thermal activation. [ \n \n 26 \n \n ] This value is consistent with the field (15.5 mT, see Figure 2d ) required to destroy an intermediate state in the MTJ device at low temperature, confirming that the intermediate states are stabilized by the strong domain wall pinning effect in the free layer. Figure 3 Strong domain wall pinning in a MgO/CoFeB/W/CoFeB/MgO film (FL‐film). a) The perpendicular hysteresis loop of the FL‐film. b) Kerr image showing the dendritic trace of the domains after domain wall motion driven by a perpendicular field of 3.6 mT in the FL‐film. c) Velocities of domain wall motion driven by a perpendicular field in FL‐film (in blue circle) and a W/CoFeB (1.0 nm)/MgO film (in red diamond), and the linear fit using the creep law. d) Scanning image of the stray fields distribution above a saturated FL‐film. The NV center is 20.7±6.7 nm above the sample and the quantization axis tilts 35° from the sample plane. According to Leon Chua's definition, a memristor can be described as, v = R ( w , i ) i and d w d t = f ( w , i ) , where v and i are the voltage and current. w can be a set of state variables and R and f can in general be explicit functions of time. [ \n \n 27 \n , \n 28 \n \n ] Obviously, the behavior of the present device consists of this definition, where w can be defined as the domain wall position. Theoretically, the number of intermediate states depends on the pinning sites distribution. Increasing the density and strength of pinning sites allows to obtain more intermediate states in a device with finite size. 2.3 Hybrid Chiral Domain Wall Structure The domain wall pinning effect in a heavy metal/CoFeB/MgO stack is usually very weak. [ \n \n 25 \n \n ] For example, we measured the domain wall motion velocity in a W/CoFeB (1 nm)/MgO film for comparison and found the intrinsic pinning field is 4.1 mT, as shown in Figure 3c . The domain wall pinning effects in double layer film with W insertion is four times larger than CoFeB single‐layer films. Scanning microscopy with NV center in diamond is used to check the magnetic quality of the FL‐film. [ \n \n 29 \n , \n 30 \n \n ] The NV center is placed 20.7±6.7 nm above a magnetically saturated FL‐film and the distribution of stray fields along the NV center quantization axis is obtained, as shown in Figure 3d . Supposing a fluctuation of the magnetization by 10% occurs in a 10 nm × 10 nm local area, which is a minimum dimension for an eventual defect to pin a domain wall, the fluctuation of the stray field caused by this defect should be in the order of 300 μT according to simulations (see Supporting Information 1 (online)). This fluctuation may be caused by a local change of the magnetic layers thickness or the saturation magnetization. Obviously, the measurement results via NV center rule out the major role of this kind of pinning effect. According to the energy‐dispersive X‐ray spectroscopy (EDS) mapping ( Figure \n 4 a ), the spatial distribution of W atoms in the composite free layer of MTJ stack is inhomogeneous, with some stochastic overlap and some breakage, which may be caused by atom diffusion during the annealing process. According to the RKKY theory, two ferromagnets separated by a thin metal layer exhibit periodic ferro‐/antiferromagnetic exchange, where the period depends on the type and lattice structure of the metal. [ \n \n 31 \n \n ] Based on the experimental data reported by S. Parkin [ \n \n 32 \n \n ] and our fitting results according to the RKKY law, two ferromagnets neighboring a W spacer begin to exhibit an antiferromagnetic coupling when the W‐thickness reaches ≈0.44 nm, reaching a peak at a W‐thickness of 0.55 nm, corresponding to approximately two atomic layers. [ \n \n 33 \n \n ] The antiferromagnetic coupling between CoFeB layers separated by 0.6‐nm thick W is also observed in a recent study. [ \n \n 34 \n \n ] Therefore, although the two free layers of our MTJ stack exhibit globally ferromagnetic exchange, the stochastic overlapping of the W atoms (W clusters) may result in antiferromagnetic exchange in some local regions. Figure 4 Domain wall profile in CoFeB/W/CoFeB multilayers dominated by the competition between interlayer coupling and opposing DMIs. a) EDS mapping showing the inhomogeneous distribution of W atoms in the multilayer stack. b) Side‐view of domain wall profiles under different interlayer coupling strengths obtained via micromagnetic simulations. c) Variations in the azimuth angle of the domain wall magnetization (0 means chiral vortex wall and π /2 means coupled Bloch wall) and the surface energy as functions of the interlayer coupling strength. d) Schematic showing the chiral vortex domain wall structure around a W cluster. On the other hand, considerable DMIs can arise at a W/CoFeB interface because of the spin‐orbit coupling, [ \n \n 35 \n , \n 36 \n \n ] and also at a CoFeB/MgO interface. [ \n \n 37 \n , \n 38 \n , \n 39 \n \n ] This interaction is particularly large for a Fe‐rich CoFeB composition, as in our case, and promotes a chiral magnetic texture in the magnetic layer. [ \n \n 40 \n , \n 41 \n \n ] We have measured the DMI in films with different symmetry breaking via asymmetric domain wall motion when both the perpendicular field and in‐plane field are applied. The DMI is about 0.45 mJ m −2 and favors a left‐handed chirality in a MgO/CoFeB(1.7 nm)/W film; while DMI reaches 0.65 mJ m −2 and favors a right‐handed chirality in a W/CoFeB(1 nm)/MgO film (see Supporting Information 1 (online)). These results suggest that in the composite free layer of MTJs, opposite chiralities are favored for the domain walls in the Up‐layer and Lw‐layer. On the one hand, because of the global ferromagnetic coupling of the two free layers, the total DMI cancels out (only 0.02 mJ m −2 is detected in the FL‐film) and the domain wall configuration shows no chirality (see the Supporting Information 1 (online)). In contrast, when the exchange coupling between the Up‐layer and Lw‐layer is weak or even antiferromagnetic in a local region where W atoms overlap, the domain wall chiralities in the two free layers should be determined by the corresponding DMIs. The magnetization of the domain walls center in the two free layers are opposite, and a chiral vortex could form around the W cluster. The magnetic state of the two 1‐nm‐thick ferromagnetic layers separated by a 0.2‐nm spacer is simulated using the OOMMF code, [ \n \n 42 \n \n ] as shown in Figure 4b . In the first (second) case, the ferromagnetic coupling strength J \n ex is set to be 1 mJ m −2 (0.01 mJ m −2 ), where a positive (negative) sign of J \n ex means ferromagnetic (antiferromagnetic) coupling. The interfacial DMI constant D is set to 0.5 mJ m −2 in both cases, with a negative (positive) sign in the Up‐layer (Lw‐layer). The results confirm the above analysis: a coupled Bloch domain wall forms in the two free layers with strong coupling, while a vertical vortex forms under weak coupling, the chirality of which is dominated by the DMIs in each layer. Furthermore, the magnetization direction in the center of a domain wall and the domain wall surface energy as a function of the interlayer coupling strength are calculated and given in Figure 4c (see detailed calculations in Supporting Information 1(online)). As the interlayer coupling strength decreases below 0.35 mJ m −2 , the domain wall profile transforms to a chiral vortex, and the domain wall surface energy drops rapidly. In fact, the formation of the chiral vortex domain wall minifies both the energy associated with DMIs and with dipole–dipole interactions. [ \n \n 43 \n \n ] According to the fitting result based on the data reported by S. Parkin, [ \n \n 32 \n \n ] the interlayer coupling decreases from 0.6 mJ m −2 to −0.03 mJ m −2 as the thickness of the W increases from 0.2 to 0.55 nm in ferromagnet/W/ferromagnet structure. When a moving domain wall encounters a W cluster, the domain wall will locally transform into a chiral vortex, as illustrated in Figure 4d . The energy well due to the transition of domain wall configuration could be one important reason for the strong pinning effects in the free layer of our MTJ stack (see the video showing this effect in Supporting Information 2 (online)). Indeed, the transition of magnetic chirality from a vortex‐Néel domain wall to a degenerate Bloch–Néel wall mediated by increasing RKKY interactions has been experimentally observed in similar structures. [ \n \n 44 \n \n ] It should be noted that, according to our model, even a partial rotation of the domain wall angle could cause a large fluctuation of wall energy (Figure 4c ). Besides, the local variation of magnetic anisotropy due to the inhomogeneous distribution of W or local change of exchange coupling due to W—Co or W—Fe alloying may also contribute to the domain wall pinning effect. Moreover, the dendritic domains morphology after field‐driven domain wall motion (Figure 3b ) also suggest that the strong demagnetizing field due to the thick composite free layers in this sample (1.8 nm in total) plays a non‐negligible role in maintaining the uncompleted switching state in the free layers (see the Supporting Information 1 (online)). 2.4 Spike‐Timing‐Dependent Plasticity Due to the memristive and non‐volatile magnetoresistance, this two‐terminal MTJ can be used as synapse for neuromorphic computing. [ \n \n 45 \n \n ] The plasticity, an essential property of an electronic synapse, [ \n \n 46 \n \n ] is investigated by applying two types of spike stimulus. First, a train of spike voltage pulses with stepwise increasing magnitude is applied and the resistance is measured immediately after each stimulating pulse, as shown in the top panel and inset of Figure \n 5 a . When the spike voltage reaches a threshold value (0.39 V/−0.37 V), the resistance of the device begins to increase/decrease gradually, corresponding to the potentiation/depression function of the synapse. The non‐volatile quasi‐continuous variation of resistance with stimulus shown here is a characteristic of long‐term functional synaptic plasticity. [ \n \n 2 \n \n ] In the second case, the plasticity is explored using sequences of voltage pulses with constant amplitude (0.54 V/−0.44 V) and duration ( T \n P = 200 ns), as shown in Figure 5b , and plastic behavior is also observed. It is worth noting that with different types of stimulating signals, two different memristive profiles are obtained on the same device and both of them could be employed in neuromorphic computing. [ \n \n 2 \n , \n 47 \n \n ] \n Figure 5 a) Plasticity explored by pulse sequence with ramped amplitude (from 0 to 6.2 V/−5.5 V with 0.01‐V increase per step) and constant duration ( T \n P = 200 ns per pulse). The upper plot shows the stimulating and detecting pulses (blue and red in the inset, respectively). A pulse with 0.05‐V amplitude and 1‐µs duration is applied after each stimulating pulse as the detecting signal. The lower plot shows the corresponding detected resistance. b) Plasticity explored by constant amplitude pulses sequence (0.54 V/−0.44 V) with T \n P = 200 ns. The resistance is measured via low‐voltage pulse (0.05 V, T \n P = 1 µs). c) The pre‐spike and post‐spike waveforms and the across voltage obtained as the superposition of the pre‐ and post‐spikes. d) An example of the resistance change induced by an across voltage with Δ t = 80 µs. e) STDP learning curve. Then, STDP, a fundamental learning function for artificial neural networks, [ \n \n 47 \n , \n 48 \n \n ] is investigated based on these devices. Two sequences of pulses ( T \n P = 200 ns) with opposite polarity, the amplitude of which ramped below the threshold switching voltage, are used as pre‐spike. A couple of opposite‐voltage pulses ( T \n P = 200 ns and ±0.2 V), with a delay time Δ t from the center of pre‐spike, are used as post‐spike, as shown in Figure 5c . Here, Δ t is defined as positive (negative) when the pre‐spike stimulus is applied before (after) the post‐spike. Since the amplitude of both the pre‐ and post‐spike stimulus are below the switching voltage of the device, the resistance variation is determined only by the polarity and amplitude of the pick of the across voltage, which depends uniquely on the delay time Δ t . Figure 5d gives an example showing the resistance variation corresponding to an across voltage with Δ t = 80 µs. We define the synaptic weight as the ratio of the initial resistance to the resistance after spiking. Then, we perform a set of tests to observe the weight changes under varying delay time Δ t , ranging from −110 to 140 µs, and the results are given in Figure 5e . Statistically, when the pre‐spike occurs before (Δ t >0) the post‐spike, the synaptic weight increases, mimicking the potentiation of synapse in a neural system, vice versa; The smaller |Δ t |, meaning the higher correlation between the two spikes, the deeper potentiation/depression of synapse. 2.5 Spin Memristor‐Based SNNs In order to investigate the availability and performance of the device in spiking neuromorphic networks, a device‐system simulation with a hierarchical framework is performed. A behavioral model of the device that matches the experimental data is developed and then integrated as synapse in the network, interfacing with the Leaky‐Integrate‐Fire neurons. For both the AP‐P and P‐AP switching, the variation of the resistance along with the stimulating time can be phenomenologically fitted with an exponential function, with a characteristic time constant τ for each given voltage, as shown in Figure \n 6 a,b . ln τ is observed to be linear to the voltage, consistent with the domain wall pinning‐dominated switching model. [ \n \n 49 \n \n ] A noise factor is added into the device model considering the device's variations and stochastic nature of the domain wall depinning effect. Some examples of the switching behaviors of the device model are given in Figure 6a , b . Then, unsupervised learning is conducted on a subset of the MNIST database [ \n \n 19 \n \n ] as shown in Figure 6c . Three stylized Arabic numerals (0, 1, and 2) are used for both training and testing. To classify the 20 × 20‐pixel black‐and‐white images into 3 classes, a two‐layer network with 20 × 20 neurons as input layer and 3 neurons as the output layer is built. Each input neuron is connected with one pixel of the image, which is then fed to the output layer through the synaptic connection built by spin‐torque memristors. Thus, the firing rates of output neurons were mainly determined by the plasticity of the spin‐torque memristor. The specified unsupervised learning protocol can be referred to. [ \n \n 47 \n \n ] Figure 6d plots the synaptic conductance of spin‐torque memristors (or weights) learned by the system during the complete training session for all the letters. Initially, all the weights are randomly distributed between 1/ R \n ap to1/ R \n p . Increasing the number of epochs from 0 to 5, 20th, and 50, the weight of the depressed synapses (purple lines) gradually converge to the lowest conductance and the potentiated synapses (yellow lines) converge to the highest conductance. After 50th epochs of unsupervised training, the neural network is trained and the letters are recognizable in the synaptic layer. The specified system simulation method and intermediate data could be found in the Supplementary Materials. Figure 6 Spin‐torque memristor‐based SNN implementation and simulations. Resistance change in spin‐torque memristor corresponding to the a) P‐AP (V = 0.54 V) and b) AP‐P (V = −0.44 V) switching process. Red lines represent the averaged result based on 10 experimental tests and the dashed black lines represent the fitting result with the exponential function. Thin blue lines show the switching behavior given by the built model (with noise) used in simulations for 5 tests under identic stimulating voltage. c) The topological structure of proposed SNN with 20 × 20 pre‐neurons connected to 3 post‐neurons through a 400 × 3 artificial synapses array. Each pixel of the images shown to the network is associated with 3 handwritten numerals (0, 1, and 2). d) Synaptic weight evolution of the MTJs matrix in the SNN during unsupervised learning session for 50 epochs, trained with the three numerals. The color bar on the right indicates the conductance range from 1/ R \n ap to1/ R \n p . The resistance of the presented spin memristor is around hundreds of Ohms and the ON/OFF ratio is lower than conventional metal‐oxide memristors. [ \n \n 28 \n \n ] These disadvantages could be overcome by using multiple devices in series as one synapse or implementing sense amplifiers in the synaptic array in real applications. [ \n \n 50 \n \n ] Since the response speed of the device increases in an exponential manner with the applied voltage, a spiking time reduced to nanoseconds can be expected if the voltage is slightly increased. In fact, tens nanometers of STT‐induced domain wall motion are possible within this time scale (10). Considering a spiking voltage of about 0.6 V and the average device resistance of 130 Ω, the power consumption per spike is estimated to be in the order of 10 pJ for a single device. It is worth noting that the memristive behavior of our devices is achieved through magnetic switching, without ion motion or ionic valence changes. Their endurance is expected to be more than 10 15 cycles, [ \n \n 7 \n , \n 20 \n , \n 21 \n \n ] which is promising for neuromorphic computing that requires high reliability on devices, for example, neural networks with “on‐chip” learnings. [ \n \n 51 \n \n ]"
} | 7,808 |
28683118 | PMC5500281 | pmc | 460 | {
"abstract": "Shallow marine ecosystems naturally experience fluctuating physicochemical conditions across spatial and temporal scales. Widespread coral-bleaching events, induced by prolonged heat stress, highlight the importance of how the duration and frequency of thermal stress influence the adaptive physiology of photosymbiotic calcifiers. Large benthic foraminifera harboring algal endosymbionts are major tropical carbonate producers and bioindicators of ecosystem health. Like corals, they are sensitive to thermal stress and bleach at temperatures temporarily occurring in their natural habitat and projected to happen more frequently. However, their thermal tolerance has been studied so far only by chronic exposure, so how they respond under more realistic episodic heat-event scenarios remains unknown. Here, we determined the physiological responses of Amphistegina gibbosa , an abundant western Atlantic foraminifera, to four different treatments––control, single, episodic, and chronic exposure to the same thermal stress (32°C)––in controlled laboratory cultures. Exposure to chronic thermal stress reduced motility and growth, while antioxidant capacity was elevated, and photosymbiont variables (coloration, oxygen-production rates, chlorophyll a concentration) indicated extensive bleaching. In contrast, single- and episodic-stress treatments were associated with higher motility and growth, while photosymbiont variables remained stable. The effects of single and episodic heat events were similar, except for the presumable occurrence of reproduction, which seemed to be suppressed by both episodic and chronic stress. The otherwise different responses between treatments with thermal fluctuations and chronic stress indicate adaptation to thermal peaks, but not to chronic exposure expected to ensue when baseline temperatures are elevated by climate change. This firstly implies that marine habitats with a history of fluctuating thermal stress potentially support resilient physiological mechanisms among photosymbiotic organisms. Secondly, there seem to be temporal constraints related to heat events among coral reef environments and reinforces the importance of temporal fluctuations in stress exposure in global-change studies and projections.",
"conclusion": "Conclusions Our laboratory experiment represents the first study focusing on the physiological responses of LBF to temperature fluctuations. Although some physiological responses showed high variability, this study illustrates how thermal variation has different effects on the foraminifera and their photosymbionts compared to chronic exposure despite the same peak temperature. We also showed how reoccurring stress did not induce acclimatization, likely because A . gibbosa populations from the Florida Keys are already adapted to the applied pattern and amount of temperature variability. This study, together with coral research, conveys how temperature fluctuations affect reef ecosystems differently than chronic exposure, provided that the intensity and duration of transient thermal stress events do not exceed naturally occurring extremes [ 28 , 54 ]. This study further demonstrates that experimental studies and projections of global change effects on reef calcifiers must consider temporal fluctuations in stress exposure. In a warming ocean, fluctuations in stress level can be an important factor to facilitate recovery from chronic heat stress [ 30 ], which either allow for short-term acclimatization [ 27 ], or induce physiological acclimation [ 8 ] by enhancing metabolic efficiency [ 26 ]. The energetic costs of acclimatization through high physiological plasticity [ 18 ], such as possible suppression of reproduction, are important aspects that need to be addressed in future research. Overall, marine habitats with fluctuating temperature regimes may bear highly resilient reef calcifiers with a high potential to seed or serve as potential reef refugia [ 28 – 30 ], and therefore need to be primary focal points of coral reef research to guide global conservation efforts.",
"introduction": "Introduction The health and the geographical distribution of coral reefs are rapidly declining with ever increasing local and global pressures [ 1 ]. Among the most prominent causes for this decline is long-term ocean warming, often manifested as transient heat events, which induce the loss of photosynthetic microalgae and/or photopigments from reef organisms, known as bleaching [ 2 ]. The bleaching phenomenon was first observed among corals [ 2 , 3 ] and has since been documented among other photo-symbiotic tropical organisms including large benthic foraminifera (LBF) [ 4 ]. In recent years shallow-water tropical reef regions (e.g., the Great Barrier Reef) have undergone massive bleaching events [ 5 ], which are expected to become regular occurrences in the coming decade [ 6 ]. The ongoing decline of coral populations and degradation of coral reefs has kindled interest in the thermal tolerance, adaptive value and stability of algal-invertebrate symbioses in these environments under higher temperature regimes [ 7 , 8 ]. The LBF Amphistegina spp. is a circumglobal, warm-water, calcifying eukaryote inhabiting oligotrophic coral-reef and shallow-shelf environments and hosting diatom photosymbionts [ 9 ]. Facilitated by their algal symbionts, LBF are vital constituents of coral-reef ecosystems [ 10 ] and important marine calcifiers, responsible for the global production of approximately 0.1 Gt/year of carbonate sediments [ 11 ]. Due to their physiological sensitivity, LBF are commonly used as bioindicators for past and present environmental conditions such as water quality and coral reef health [ 12 , 13 ]. The LBF are exceptionally useful models for studying the effects of global change on marine photosymbiotic calcifiers, primarily due to their abundance, fast growth, and easy handling in culture. Previous studies have shown that extreme and chronic thermal stresses have direct detrimental effects on calcification and overall host and photosymbiont (i.e., holobiont) fitness [ 14 – 17 ]. These studies have characterized either the immediate response to elevated temperatures or the effects of chronic exposure. Yet, how LBF react to episodic stress events, followed by intervals of thermal respite, is currently unknown. This is a vital aspect of adaptive physiology, because episodic stress followed by a phase of recovery, represents a realistic scenario for predicting the consequences of present and future global warming [ 18 ]. Thermal stress appears to affect LBF primarily by impairing the function of the photosynthetic apparatus of the algal symbionts [ 15 , 17 , 19 ]. Such impairment can include reduced expression of the rate-limiting carbon-fixation enzyme RuBisCO (ribulose 1-5-biphosphate carboxylase/-oxygenase) [ 20 ], reduced photopigment concentrations and photosynthetic performance [ 14 – 17 , 21 , 22 ] and reduced oxygen-production rates [ 16 ]. Collectively, thermal stress can cause reduced growth, calcification, survivorship and fecundity [ 14 – 16 , 22 – 24 ], as well as host inactivity [ 17 ]. The exact kinds of molecular damage and cellular stress-related mechanisms that mediate these effects remain unknown. Similarly, the processes of recovery of LBF after stress exposure have not been previously reported. Recovery potential, however, is important in the context of episodic stress exposure, as such potential may facilitate survival despite peak temperatures reaching the bleaching threshold, and could even increase thermal tolerance [ 7 , 8 ]. Recovery responses could explain how LBF thrive in habitats where local temperatures can exceed temperatures that induce mortality when persistent over several days [ 13 ]. The local effects of global warming include fluctuating physicochemical conditions across spatial and temporal scales [ 18 , 25 ]. In response to dynamic atmospheric and hydrographic processes, including cloud formation, wind-driven advection, diurnal heating and cooling, tides and internal waves, many abiotic parameters (e.g., intensity of solar irradiance, pH, temperature, and nutrient availability) can be altered on scales from hours to weeks [ 7 , 26 – 28 ]. Such fluctuations can be experienced from the surface of the ocean to mesophotic depths within coral-reef habitats [ 29 , 30 ]. For example, the Florida Keys already experience high levels of thermal stress on a near-annual basis [ 6 ]. Common daily subsurface temperature fluctuations here are on the order of 2 to 5°C [ 31 ], but peak within-day ranges during summer can reach as much as 7 to 9°C at 20 to 30 meters depth, respectively [ 29 ]. Environmental heterogeneities influence the sensitivity of organisms to changing ocean conditions [ 18 ] and should be considered when assessing their thresholds and tolerances. For instance, when temperature fluctuations are incorporated into model projections of global warming scenarios, the effects on species performance are stronger [ 25 ], highlighting the necessity to understand resilience to episodic stress events. In this study, we investigated how the effects of episodic exposure to thermal stress, followed by recovery phases of thermal respite differ from the effects of chronic exposure to heat stress in LBF. Along the lines of earlier studies conducted on corals [ 8 , 26 , 27 , 32 ], our hypothesis emphasizes the role of thermal variations on the physiological performance of LBF. Specifically, we carried out a laboratory-based culturing experiment, exposing the common western Atlantic LBF species Amphistegina gibbosa to one of four treatments, (a) control, (b) a single thermal-stress event, (c) episodic thermal-stress events or (d) chronic thermal stress. Our goals were to determine (i) whether single or episodic event exposure to thermal stress causes similar physiological response as chronic exposure, (ii) if the initial physiological response recovers after the stress is released and (iii) if acclimatization occurs to repeated short-term stress events.",
"discussion": "Discussion The results demonstrate that the physiological effects of single and episodic stress events on photosymbiotic calcifiers are markedly different compared to chronic stress. Single and episodic thermal peaks did not impair the function of A . gibbosa , while chronic stress damaged the algal photosymbionts, induced an antioxidant defense response, and compromised the overall holobiont health and activity. The divergence in physiological responses between the chronic and episodic thermal stress seems to have developed between day 3 and day 12 of the experiment (Figs 3 and 4 ). This divergence emphasizes not only the temporal tipping point and damage associated with chronic stress but also the importance of respite phases during thermal stress. The temperature conditions in this experiment emulate water temperature variability and duration (hours to days) shifts >5°C recorded in tropical reefs [ 18 , 29 ] and therefore present real-life scenarios of temperature stress. Control treatment Since the photosymbionts in the control treatment flourished, the host grew, reproduction likely occurred (seen as mortality), and ACAP values did not rise above natural population averages, we can use the experimental conditions and observed response patterns to predict how field populations respond. For instance, Chl a concentration increased over time, indicated by lower L* and increasing b* values, as well as greater net oxygen-production rates by the end of the experiment ( S2 Fig ). This response can be explained by natural increase in numbers of photosymbionts during the ontogeny of the foraminifera and may also reflect an increase symbiont density in response to low light levels in culture conditions [ 19 , 49 ] or a possible feeding-related rise in the availability of fixed nitrogen, which could increase the amount of nutrients supplied from the host to the symbionts. The former process of photo-acclimation is known from corals, which can increase the amount of chloroplasts in their photosymbionts to meet their energy demands despite low-light conditions in their environment [ 50 ] and might act similarly in LBF by increasing the amount of symbionts or their chloroplasts. The ACAP values of the control population are comparable to values reported for a population of A . lobifera in the Great Barrier Reef, which shows elevated resilience towards temperature and nutrient stress, probably due to preconditioning based on environmental fluctuations [ 23 ]. Specifically, both the absolute ACAP values and the temporal trend in the control resemble those measured by Prazeres et al. [ 23 ], indicating that the population of A . gibbosa in our study is possibly acclimatized or adapted to comparable conditions. Chronic thermal stress Chronic thermal stress induced gradual bleaching, which is reflected by reduced photopigment concentrations and ultimately decreasing photophysiological performance. This observation is in line with previous studies on LBF [ 14 , 16 , 17 , 19 , 23 ]. Although oxygen-production rates were negative after 30 days of chronic stress exposure, holobiont respiration rates indicated that the remaining photosymbionts were still photosynthetically active ( S2 Fig ). Those specimens that exhibited intense bleaching showed accumulation of brown material at the periphery of the shell and close to the aperture ( S1D Fig ) resulting from the deterioration of chloroplasts, typically followed by degradation or expulsion of the photosymbiont residues [ 17 , 19 ]. Despite survival of some photosymbionts, their decreased concentration and activity likely impaired the fitness of the holobiont, by reduced translocation of metabolites causing lower growth rates [ 15 – 17 , 21 , 22 ], reduced motility [ 17 ] and probably also less reproductive activity [ 24 ] (here seen as mortality). Although growth rates across all treatments gradually slowed this is most likely due to the same natural aging trends known for benthic foraminifera [ 51 ]. It is remarkable that that the chronic-stress treatment seems to have reduced growth by ~50% in comparison to the other treatments after the first measurements in the treatment. This early reduction in growth indicates that the primary response to chronic thermal stress is likely due to the holobionts using their energy to maintain homeostasis. Respiration rates could indicate that bacteria, which were feeding on the remains of dead foraminifers were respiring very actively. Alternatively, the respiration rates could indicate that A . gibbosa specimens from the chronic-stress treatment were still alive at the end of the experiment, although they did not reproduce, ceased to move, and did not grow after 21 days of chronic exposure. Together with previous studies, our results support the hypothesis that foraminiferal hosts are more resistant to thermal stress than their endosymbionts [ 19 , 40 ]. Cytological analyses revealed that prolonged temperature stress under low light conditions (6–8 μmol photons m -2 s -1 ) induced significant declines in photosymbiont densities and lipid bodies, while some host endoplasm remained intact [ 19 ]. In our experiment, similar exposure temperature and duration (32°C for one month) led to bleaching but was sub-lethal to the host, which reconfirms that LBF can survive bleaching, however with the overall reallocation of metabolic activity. The lack of mortality in the chronic-stress treatment in our experiment seems to be at odds with other long-term chronic exposure studies, which showed increased mortality at elevated temperatures [ 16 , 22 , 23 ]. This could be related to variations between LBF and photosymbiont species, or durations and intensities of stress exposure in the different studies. However, the functionality of the holobiont at the end of our experiment appears to have been so severely impaired that more profound effects will likely ensue if stress continues or other interacting pressures occur [ 22 ]. Hallock et al. [ 33 ] reported a variety of issues associated with bleaching in A . gibbosa , including reproductive failure, epibiont infestations and calcification anomalies. Here, we show for the first time that ACAP in A . gibbosa is greatly enhanced by chronic thermal stress (Figs 4E and 5 ). The only other study to measure ACAP in LBF in response to thermal stress showed that after 30 days at 29°C, the ACAP of A . lobifera had not increased significantly [ 23 ]. The lack of ACAP response from A . lobifera compared to A . gibbosa from our study may have resulted from the 3°C higher exposure temperature in our experiment, species-specific temperature tolerances, different local adaptations or symbiont communities. The function of elevated ACAP is associated with defense mechanisms against amplified oxygen radicals produced by photosynthesis under higher temperature, as seen amongst cnidarians [ 52 ]. Although the density of symbionts, which are expected to produce radical oxygen species, decreased over time (e.g. Fig 4A ) the ACAP increased continuously. This implies that either the remaining but more and more damaged symbionts were still producing sufficient oxygen radicals for the host’s defense system to require higher ACAP to compensate for these, or that the antioxidant capacity was responding to the oxidative stress with a time lag. Since we did not measure gene or protein expression as in other studies [ 20 , 27 ], but on the level of enzyme and non enzymatic low-molecular-weight scavenger (e.g. glutathione, ascorbic acid, uric acid, vitamin E and carotenoids) capacity [ 42 , 52 ], these might be produced more slowly and, more importantly, might remain functional over considerably longer time periods. Overall, our chronic stress scenario suggests a reallocation of host energy towards defense and repair mechanisms, thereby reducing calcification, motility and reproductive activity but preventing mortality. Single and episodic stress events The A . gibbosa coped well with fluctuating temperatures simulated by single and episodic thermal stress events. Most photosymbiont and holobiont response variables did not change significantly over the term of the experiment. This seems contradictory to former studies that analyzed the responses of LBF to short-term thermal stress, which found lower Chl a concentrations, reduced photosynthetic efficiency [ 16 , 17 ], and lower quantities of RuBisCO [ 20 ] after hours to days of exposure. These studies, however, focused on the immediate response to stress, while our results represent their physiological response after they were released from the thermal stress. It is therefore possible that A . gibbosa and most of the photosymbiont variables (e.g., Chl a and color values) had already recovered within 24 hours after the peak thermal stress, demonstrating the capability of this species to quickly recover from short-term stress. Similarly, oxygen-consumption rates required only a few hours to recover from extreme temperatures, in contrast to photosynthesis rates that needed several days to recover [ 40 ]. After the single stress event, net photosynthesis varied strongly over time. Because these variations were ongoing throughout the experiment, we interpret them as most likely related to the presumably high incidence of reproduction in this treatment (data lacking). Reproductive activity even exceeded the control specimens and represented the only variable in which single stress and episodic stress responses differed. Because half of the shells in the single-stress treatment were empty by the end of the experiment, the single thermal peak followed by stable conditions might have stimulated reproduction. In contrast, episodic stress appeared to suppress or delay reproduction in the same way as in the chronic-stress treatment. Correspondingly, suppression of asexual reproduction in adults and failure to normally calcify were reported from A . gibbosa specimens collected during summer, which also exhibited photosymbiont deterioration [ 33 ]. Previous studies [ 24 , 51 ] related reduced reproduction and fecundity to low light intensities. Since in our study the light level was the same in all aquaria and reproduction presumably occurred in other treatments, this does not seem to be the driving factor here. In the case that recurring stress induces malfunction or impairment of reproductive activities, this would imply important long-term consequences for foraminiferal life cycles, population densities and community structures with severe impacts on carbonate budgets and overall health of coral-reef environments [ 10 , 24 , 33 , 41 ] and should be addressed in future studies. Analogy to other coral reef calcifiers While there are no comparable studies on the effects of fluctuating temperatures on LBF, other photosymbiotic reef organisms have been subjected to temperature variations and showed that even short temperature reductions can reduce immediate thermal damage within coral reefs. Such examples include large-amplitude internal waves, which cause pH and temperature to drop within minutes, allowing short-term relief, and have been shown to reduce the physiological effect of heat stress on corals [ 30 ]. Daily temperature fluctuations can be beneficial to the photosynthetic efficiency of coral larvae [ 26 ], but led to strong declines in photosymbiont densities, while maintaining or even increasing calcification in studies on adult coral colonies [ 53 ]. Corals that are exposed to extreme natural temperature fluctuations during spring-tide upwelling events increase most physiological and molecular parameters, suggesting that the holobiont may acclimate to fluctuating temperatures by the symbionts capacity to increase photosynthesis and carbon fixation [ 27 ]. These results and our study support the hypothesis that temperature fluctuations, in contrast to chronic thermal stress, have substantially different effects on photosymbiotic reef calcifiers. The impact of thermal stress appears to not only depend on exposure level and duration, but also on whether the stress is constant or discontinuous because intermittent stress provides respite periods permitting repair mechanisms to alleviate or entirely prevent the detrimental effects of thermal stress. Interactive effects of multiple contemporaneous or consecutive stressors could produce different outcomes and should be targeted by future research. Besides the immediate effects of temperature variations, thermal history is an important factor among photosymbiotic reef organisms, because local acclimatization or adaptation to thermal stress may enhance thermal resistance through higher phenotypic and metabolic plasticity. This is evident by elevated thermal tolerance in corals from habitats where they naturally experience temperature fluctuations, such as large-amplitude internal waves [ 30 ] or lagoon pools [ 7 ]. Furthermore, coral colonies that were experimentally pre-stressed before exposure to severe prolonged thermal stress revealed more effective photoprotective mechanisms [ 8 ]. Similar to coral studies, A . lobifera populations from stable offshore environments are more sensitive to stress than those from inshore habitats that experience stronger fluctuating conditions [ 23 ]. Comparably, our results indicate that local conditions increased the tolerance of A . gibbosa to environmental changes, considering long-term subsurface temperature variability in the Florida Keys [ 29 ]. Specifically at the sampling location, Tennessee Reef situated in the Middle Keys, reefs were historically exposed to intermediate levels of sea-surface temperature variability [ 54 ]. These intermediate thermal fluctuations seem to be beneficial to biodiversity, survival, and recovery of the local stony-coral assemblages [ 28 ]. It is therefore highly probable that the population of A . gibbosa sampled for our experiment is adapted or acclimatized to thermal variability such that single- and episodic-stress treatments did not exceed its tolerance range. Indeed, time-series studies of A . gibbosa populations from the Florida Keys through the 1990s revealed that bleaching followed the solar cycle of irradiance, such that peak bleaching consistently occurred well before the late summer temperature maximum and the populations were typically already showing recovery when temperature peaked [ 4 , 33 ]. No acclimatization to repeated stress events occurred in our study, but the LBF under chronic stress arrived close to the thermal tipping point. In some corals, elevated thermal tolerance can be independent of local variation in ocean temperature, such that their acclimatization capacity to future warming is limited [ 32 ]. Whether A . gibbosa is generally characterized by high thermal tolerance or if the high physiological plasticity found in this study is specific to the local population assessed, which would suggest a high acclimatization capacity, has to be targeted in future research. This raises the discussion on whether the resilience of these foraminifers is a product of short-term acclimatization due to recent thermal history, or if long-term adaptation has increased the tolerance of these photosymbiotic calcifiers. Such questions could be disentangled with the use of ‘omics’ approaches, which can determine the influence of environmental stressors on the gene or protein level and therefore reveal meaningful insights into underlying molecular processes governing acclimatization/adaptation pathways. Furthermore, research on the flexibility and physiological plasticity of the photosymbiont community would further improve our understanding of LBF adaptive potential."
} | 6,464 |
34188937 | PMC8226191 | pmc | 462 | {
"abstract": "Lay summary High flow speeds can mediate damaging impacts of sub-lethal thermal stress on a branching coral species, showing increased levels of endosymbiont photosystem health compared to corals under low flow conditions. However, eventual declines in physiology and no difference in bleaching under high and low flow conditions indicate that eventually thermal stress may overwhelm positive impacts of flow.",
"conclusion": "Conclusions and implications Investigating flow patterns at the scale of metres within the context of putative beneficial physiological impacts reveals some interesting avenues for coral reef management. For example, current metres deployed on the reef slope showed that reef topography can influence local flow patterns within niche habitats on coral reefs. Variability in flow could be an ecologically relevant but overlooked phenomenon and requires further investigation. Acclimatization due to variability in historic SST trajectories has been shown to significantly impact the molecular mechanisms that underpin thermal tolerance in corals ( Oliver and Palumbi, 2011 ; Schoepf et al. , 2015 ; Ainsworth et al. , 2016 ; Thomas et al. , 2018 ). A recent study investigating the transcriptional responses of corals during tidal fluxes suggested that variability in tide (and therefore flow) is related to acclimatory mechanisms in corals, such as ‘front loading’( Ruiz-Jones and Palumbi, 2017 ). This highlights a limitation of our experiment, in that corals were exposed to constant high or low flow conditions. These conditions were applied for the purpose of understanding whether high or low flow may affect coral functioning under thermal stress, but future studies should look to replicate more closely the diurnal variations in flow speed that corals experience in different reef environments. The predictable timing of neap and spring tide times could also be used to inform management decisions proceeding predicted bleaching events. At larger reef scales, if lower flow caused by neap tides was to occur alongside cloudless days and low wind speeds, shallow reefs could be at risk of high bleaching and mortality ( DeCarlo et al. , 2017 ; Burt et al. , 2019 ). Knowledge of the predominant flow direction at a site may also be able to explain local patterns of bleaching severity. At Coral Gardens (reef slope), the prevailing flow direction at the two exposed metres (1 and 2) was north-westerly, with the highest flow speeds also occurring in this direction. This same pattern has been highlighted in other studies in this region, where a north-westward flow pattern was also measured along the Capricorn Bunker group ( Griffin et al. , 1987 ). These patterns also changed at each metre during neap and spring tidal phases. The morphology of coral colonies provides resistance to the natural flow of water, slowing down and causing turbulence in the water column ( Reidenbach et al. , 2006a ; Reidenbach et al. , 2006b ; Reidenbach et al. , 2007 ). This means that under some circumstances, flow can be greatly reduced at downstream faces of a colony compared to upstream. This effect is dependent on coral morphology but has been hypothesized as causing mortality from the inner sections of branching coral colonies, which further progressed towards colony fringes during an anomalous thermal event in Iriomote, Japan ( Baird et al. , 2018 ). In conclusion, the results presented here indicate that high flow has a mediating effect on the health of endosymbiont photosystems in A. aspera compared to low flow conditions under both sub-bleaching and bleaching levels of thermal stress. We uncover a synergistic interaction between high flow and the protective effect of a pre-stress pulse in temperature. Measurements of physiology indicate the beneficial action of high flow on the efficiency, damage and recovery rate of endosymbiont photosystems under both direct thermal stress, and acclimatory thermal treatments. Endosymbiont photosystems retained photosynthetic efficiency in high flow heat treated corals, despite occurring in similar densities to low flow heat treated corals at the end of both experimental periods. We hypothesize that this effect may be due to increased rates of mass transfer under higher flow. However, despite a delayed onset, under sustained exposure to high levels of thermal stress corals under both high and low flow exhibit declines in photo-physiology and bleaching severity. Further work is needed to understand the effects of flow on coral host responses, in addition to coral recovery after temperature stress has subsided. Through taking a holistic approach to understanding how an environmental driver impacts coral responses to thermal stress, we have investigated the impacts an environmental factor like flow can have on the health and persistence of reefs into the future.",
"introduction": "Introduction Tropical coral reefs are invaluable in their ability to support marine biodiversity ( Reaka-Kudla, 1997 ; Hughes et al. , 2002 ), provide resources to coastal communities ( Costanza et al. , 1997 ; Woodhead et al. , 2019 ) and absorb energy protecting large areas of coastline ( Harris et al. , 2018 ; Osorio-Cano et al. , 2019 ). Underlying this remarkable capacity for ecosystem function is a symbiotic partnership between scleractinian coral species and an algal dinoflagellate symbiont (of the family Symbiodinaceae; LaJeunesse et al. , 2018 ). The endosymbiont photosynthesizes and translocates fixed carbon in the form of sugars to the host, providing the coral with a majority of the carbon needed to grow and accrete calcium carbonate from the surrounding water column ( Muscatine and Porter, 1977 ). Yet, coral symbiosis is increasingly threatened by a myriad of disturbances from global ( Hughes et al. , 2017 ; Hughes et al. , 2018 ) to local scales ( McLean et al. , 2016 ; Wolff et al. , 2018 ). Exposure to environmental conditions to which a coral may not be locally acclimated can cause coral bleaching—a stress response resulting in the loss or reduction of endosymbionts and/or damage or loss of their associated pigments from the coral host cells ( Gates et al. , 1992 ; Hoegh-Guldberg, 1999 ). For example, coral bleaching events are often linked with exposure to above average temperatures (i.e. thermal stress), exacerbated by high light conditions (and low wind speeds; Fordyce et al. , 2019 ). The impacts of climate change over reefs are varied: marine heatwaves may be severe causing widespread bleaching and mortality ( Fordyce et al. , 2019 ; Leggat et al. , 2019 ), while more moderate events may cause bleaching with little to no mortality ( Heron et al. , 2016 ). Warmer temperatures have also been linked to outbreaks of coral disease ( Heron et al. , 2010 ; Randall and van Woesik, 2015 ) and the emergence of novel species interactions ( Miranda et al. , 2018 ; Vergés et al. , 2019 ). As such the preservation of reefs requires effective management plans that can mitigate the cumulative negative impacts of both local and global stressors to conserve ecosystem function. As a result of the rapid changes now being documented on coral reefs worldwide, there are a number of novel intervention actions that have been proposed ( Bay et al. , 2019 ; Board, Ocean Studies, National Academies of Science Engineering and Medicine, 2019 ; Ainsworth et al. , 2020 ; Morrison et al. , 2020 ), including the development of coral cryobiology ( Hagedorn and Carter, 2016 ), coral probiotics ( Rosado et al. , 2019 ) and the potential engineering of ‘super corals’ ( Camp et al. , 2018a ; Camp et al. , 2018b ; Buerger et al. , 2020 ). There has also been a push to use new technologies to more effectively monitor reefs ( Bajjouk et al. , 2019 ; Calders et al. , 2019 ) and various restoration methods are now becoming more refined and scalable ( Suggett et al. , 2019 ). These approaches are not without pitfalls but can provide innovative ways to preserve such critical ecosystems into the future. Similarly, there now is also the development of reef conservation approaches that aim to harness existing beneficial environmental and ecological interactions and utilize these to support coral health and survival ( Halpern et al. , 2007 ; Timpane-Padgham et al. , 2017 ; Ladd et al. , 2018 ; Ainsworth et al. , 2020 ). Importantly, reef environments are often highly heterogeneous in both abiotic and biotic interactions ( Lenihan et al. , 2008 ; Guadayol et al. , 2014 ; Rogers et al. , 2016 ; Hoogenboom et al. , 2017 ). The interplay between these, referred to as ‘bio-physical’ interactions, ultimately has the capacity to affect the responses of colonies to sources of physiological stress ( West and Salm, 2003 ; van Woesik et al. , 2005 ; Smith and Birkeland, 2007 ; Hoogenboom and Connolly, 2009 ; Schutter et al. , 2010 ; Davis et al. , 2011 ; Ainsworth et al. , 2016 ; DeCarlo et al. , 2017 ; Darling and Côté, 2018 ; Page et al. , 2019 ). Interestingly, the temperature regime that an individual coral experiences prior to surpassing its thermal bleaching threshold can also influence physiological and therefore ecological outcomes ( Ainsworth et al. , 2008 ; Ainsworth et al. , 2016 ). Even week-long increases in mild stress with sub-lethal effects during a conditioning period can play a protective role, directly causing a reduction in magnitude of subsequent stress responses such as apoptosis ( Bellantuono et al. , 2012 ; Ainsworth et al. , 2016 ). Water flow conditions over reefs have been shown to mediate the physiological outcome of thermal stress on coral health ( Nakamura and van Woesik, 2001 ; McClanahan et al. , 2005 ; Skirving et al. , 2006 ; Smith and Birkeland, 2007 ; DeCarlo et al. , 2017 ; Wolanski et al. , 2017 ; Page et al. , 2019 ). The relationship between bleaching patterns and flow was recognized after the 1998 bleaching event in Japan where coral survival was positively associated with areas of higher flow ( Nakamura and van Woesik, 2001 ). Similarly, the survival of corals on offshore islands during the 2002 bleaching event in the Arabian Gulf was linked to higher levels of water motion compared to conditions further inshore ( Riegl, 2003 ). High water flow can enhance primary production, dark respiration and particle capture in corals, thereby positively affecting growth ( Patterson et al. , 1991 ; Sebens and Johnson, 1991 ; Finelli et al. , 2006 ; Finelli et al. , 2007 ; Mass et al. , 2010 ; Osinga et al. , 2017 ). In contrast, low flow speeds (<0.03 ms −1 ) have been recorded as contributing to ‘extreme’ bleaching conditions, leading to rapid coral mortality over reefs ( DeCarlo et al. , 2017 ; Baird et al. , 2018 ). The relationship between water flow and coral function at the colony level is hypothesized to be linked to increased flow speeds creating thinner boundary layers, which increase mass transfer of gases and metabolites, increasing the rates of physiological processes ( Patterson, 1992 ; Falter et al. , 2005 ; Carpenter and Williams, 2007 ; Falter et al. , 2007 ; van Woesik et al. , 2012 ; Page et al. , 2019 ). Specifically, under thermal stress it has been suggested that higher flow speeds may positively impact coral function through reducing heat-induced oxidative stress ( Nakamura and van Woesik, 2001 ; Nakamura, 2010 ). Although experimental studies to date offer support for the beneficial impacts of high flow on coral responses to general environmental stress ( Nakamura and van Woesik, 2001 ; Nakamura et al. , 2003 ; Nakamura et al. , 2005 ), limited experimental comparisons and detail of the responses measured, coupled with exposure to extreme treatments (e.g. high flow conditions of 0.5 to 0.7 m s −1 and 95% irradiance; Nakamura and van Woesik, 2001 ), result in insufficient evidence to conclude the extent to which flow may be causing differential responses to thermal stress. In this paper, we look to assess the impacts of flow on coral function and in doing so provide experimental evidence to characterize how flow may modulate thermal stress responses in corals. Resolving the biological consequences of flow at reef and within-reef scales is important to comprehensively understand how flow affects resistance of corals to thermal stress. Here we investigate whether water flow speeds (high and low) can mediate the impacts of multiple climate change-related thermal regimes on a thermally susceptible coral species, Acropora aspera ( Loya et al. , 2001 ; van Woesik et al. , 2011 ; Nitschke et al. , 2018 ). Using Heron Island on the Southern Great Barrier Reef (GBR) as a case study, we contextualized experimental flow simulations through characterizing the flow conditions to which corals are exposed within two reef habitats: the flat intertidal region of reef lagoon (flat) and the sloping reef towards a deep water channel (slope). To investigate the impacts of water flow on an initial stress response through to bleaching, we exposed fragments of A. aspera collected from the reef flat to a range of thermal stress regimes over two experimental periods. Specifically, we designed three temperature treatments based on the thermal threshold for Heron Island Reef Flat: a sub-bleaching trajectory, a sub-bleaching trajectory with a pre-stress pulse in temperature ( Ainsworth et al. , 2016 ) and a bleaching trajectory involving a direct increase in temperature to the thermal threshold for Heron Island Reef Flat (34°C). This relatively high bleaching threshold is owed to the extremely variable environment to which corals are acclimatized to ( Kline et al. , 2012 ; Ruiz-Jones and Palumbi, 2017 ). The three temperature treatments were achieved through manipulation of the rate of temperature increase and maximum daily temperatures, and the stress responses of the coral host and their endosymbiotic algae were measured using photophysiology and quantification of endosymbiont densities.",
"discussion": "Discussion Flow conditions over reefs have the potential to mediate the physiological damage and function of hard coral species under thermal stress ( Nakamura and van Woesik, 2001 ; Nakamura et al. , 2003 ; Nakamura et al. , 2005 ; Nakamura, 2010 ; van Woesik et al. , 2012 ; Page et al. , 2019 ). Further understanding of the extent to which an environmental factor like flow may contribute to the potential resistance and resilience of reef areas at relevant ecological scales has wide implications for both adaptive management, as well as more novel intervention techniques. Through conducting physiological experiments and contextualizing flow treatments to conditions measured in both reef flat and slope habitats, the present study examines the impacts of high (~0.15 m s −1 ) and low (~0.03 m s −1 ) flow speeds on the physiological responses of an important reef building coral under different levels of thermal stress. Flow conditions over Heron Island reef Hydrodynamic models provide information on the temporal and spatial patterns of flow conditions over a reef, but these are often limited by coarse spatial resolution (e.g. eReefs smallest resolution is 1 km grids; Steven et al. , 2019 ). Understanding the range of flow conditions at scales within which corals live, i.e. metres (individual colonies) to 10s of metres (coral beds), allows experimental conditions to be guided by an ecological context and allows for abiotic heterogeneity to be acknowledged in local reef management plans. In the current study, flow speeds measured over the reef flat at Heron Island were generally consistent across all transects with few outliers. The high and low flow treatments used in both experiments were found to occur within the upper and lower confidence intervals across all transects on the reef flat at the time the study was undertaken (0.16 m s −1 and 0.093 m s −1 ). Recorded reef flat flow speeds are also representative of speeds measured over other reef flats. For example, time-average flow speeds across Kaneohe Bay Barrier Reef flat in Hawaii ranged between 0.08 and 0.22 m s −1 ( Falter et al. , 2004 ). In the present study, measurements were taken between high and low tide to capture the full range of flow speeds experienced on the flat ( Roberts and Suhayda, 1983 ) and in doing so we provide observations of the flow speeds to which corals may be exposed. This daily variability in tidal flow was visible in time series recorded on the reef slope in the current study. Flow conditions on the reef slope are inherently different to those on the flat due to differences in average depth, where flow typically decreases with increasing depth ( Lentz et al. , 2016 ). However, at the shallow depths where metres were deployed (in 3 m of water), it is clear that flow speeds over time are still driven by tidal patterns, wind and wave stress (Figs S3 – S6 ). Indeed hydrodynamic circulation within coral reefs is primarily thought to be driven by wave and tidal forcing and to a lesser extent wind and buoyancy effects Monismith (2006 ). The role of water flow in a simulated sub-bleaching thermal stress event Exposure to elevated sea-surface temperatures below the bleaching threshold has been shown to result in a number of sub-cellular and cellular responses in corals ( Ainsworth et al. , 2008 ; Bonesso et al. , 2017 ), including continued declines in condition until severe bleaching is reached (PSII yield < 0.3, cell density reduction > 50%) ( Ainsworth et al. , 2008 ). In this experiment, the lowest quantum yields recorded were in the SB high and low flow corals (PSII yields of ~0.4). At these values, corals showed signs of physiological damage to photosystems by the end of the experimental period but notably do not show significant differences in endosymbiont density (a proxy for bleaching severity) when compared to their respective control treatments ( Fig. 3b ). In contrast, endosymbiont densities recorded in corals exposed to PS SB heat treatment at the end of the experimental period were significantly lower than controls: 72% reduction in endosymbiont density was recorded in low flow corals, compared with a 47% reduction in high flow corals. Notably, these reductions in densities were not reflected in measures of photophysiology ( Fv/Fm ). Similar mismatches in declines of endosymbiont densities and photophysiology have been recorded in other coral bleaching studies ( Middlebrook et al. , 2010 ; Krueger et al. , 2015 ) and could be indicative of either remaining populations of endosymbionts retaining photosynthetic efficiency or the presence of non-photophysiological impacts of thermal stress initiating expulsion ( Baird et al. , 2009 ). Corals exposed to low flow and PS SB trajectory showed significantly reduced yields on the final day, compared to PS SB high flow corals that retained comparatively high yields (PSII yield of ~0.55, Fig. 3a ). These results clearly indicate that exposure to high and low flow speeds does interact with potential acclamatory capacity and impacts of sub-lethal thermal stress on coral health. \n Ainsworth et al. (2016) investigated the effect of the PS SB temperature trajectory on the stress responses of A. aspera . After exposure to a pre-stress pulse of 4 days with peak temperatures of 32°C, the study recorded significantly lower stress responses in pre-stressed corals compared to those exposed to only a single increase in temperature to 34°C. In the current study, significant declines were measured in endosymbiont densities in PS SB treated corals irrespective of the maintenance of higher yields. The pulse duration used in this experiment was shorter, instead temperature gradually increased to 32°C over 6 days, with a total of 2 days that reached a maximum of 32°C. Although there is evidence that this pre-stress pulse in temperature did have a beneficial effect on photosystem efficiency, the physiological acclimatory ‘signal’ may not have been strong enough to result in less severe bleaching. Equally, values of quantum yield are independent of endosymbiont density, indicating that although corals exposed to the PS SB treatment may have lower densities of endosymbionts, the populations that remain maintained high quantum yields compared to populations present in fragments exposed to a single increase in temperature (SB). The results of this experiment suggest that corals exposed to a pre-stress pulse in temperature and high flow show an increased physiological performance compared to those under low flow conditions and/or a single bleaching trajectory. This suggests a positive interactive relationship between the protective impacts of pre-stress heating and higher flow conditions. However, the damaging effects that sub-lethal thermal stress has on the physiology of coral fragments prior to bleaching may have masked any initial beneficial effect that higher flow has on the resistance of corals to thermal stress. The role of water flow in a simulated bleaching thermal stress event The temperature treatment applied in Experiment 2 reached the thermal threshold for Heron Island Reef flat (34°C) over an extended time period. A maximum daily temperature of 34°C was reached after 16 days of heat accumulation, reaching a total of 4.95°C weeks by the end of the experimental period (a total of 20 days). This meant that even though both SB (Experiment 1) and B treatments exposed corals to 4 days at 34°C, the corals exposed to B trajectory had been exposed to a greater accumulation of light and temperature stress, than those exposed to the SB trajectory. This slower ramping rate successfully allowed us to further evaluate the differential responses of corals to high and low flow under temperature stress. Corals exposed to high and low flow speeds showed differential responses when exposed to temperatures at their thermal threshold of 34°C. Buffering effects of high flow were apparent, where high flow corals maintained a higher level of photosynthetic function when exposed to thermal stress than low flow corals. Specifically, photophysiological measurements showed differences in the efficiency, damage and recovery potential of endosymbiont PSII. There is a time offset in responses between high and low flow treated corals, where the onset of quantum yield decline was two days later in high flow heat treated corals (Day 19, eDHW = 4.56°C weeks) than low flow heat treated corals (Day 17, eDHW = 3.70°C weeks). This response is also reflected in the maximum quantum yield at the end of the recovery phase ( Fig. 4b ), where significant declines are recorded in heat-treated high flow corals on Day 17 and heat-treated low flow corals on Day 15. There were also 3 days (Days 16 through to 18) when differences were seen between high and low flow, heat treated corals; high flow corals retained higher yields, but low flow heat treated corals had already declined significantly. However, similarly to the sub-bleaching experiment, by the end of the experimental period both high and low flow heat treated corals showed similar declines in endosymbiont densities and therefore coral bleaching. This result indicates that any beneficial effect of high flow was short term under sustained thermal stress. Why might flow have this effect? There are a number of putative mechanisms through which higher flow speeds may be able to reduce the vulnerability of a coral to thermal stress. The primary determinant of coral bleaching is generally described as the accumulation of oxidative damage caused by the production and accumulation of reactive oxygen species (ROS) produced during light and thermal stress to the endosymbiont and thermal stress in the host ( Lesser et al. , 1990 ; Gates et al. , 1992 ; Lesser, 1996 ; Nii and Muscatine, 1997 ; Hoegh-Guldberg, 1999 ; Davy et al. , 2012 ). Eventually, the rate of damage overwhelms capacity for the host and/or endosymbiont to repair, leading to a breakdown of the symbiotic relationship. This occurs through the expulsion of the endosymbiont and/or apoptosis of the host gastrodermal cells or the endosymbionts themselves ( Weis, 2008 ; Davy et al. , 2012 ). High flow has been hypothesized as augmenting passive diffusion of ROS away from coral tissue, limiting the amount of cellular damage that occurs ( Nakamura et al. , 2003 ; Nakamura et al. , 2005 ). Increased ROS removal has previously been described as the mechanism by which high flow lowers rates of light induced photoinhibition ( Nakamura et al. , 2005 ) and enhances recovery ( Nakamura et al. , 2003 ). However, ROS are volatile molecules that need to cross cell walls, membranes and tissue layers (symbiosome, gastrodermal cell wall, mesoglea and epithelial cells) before the diffusion boundary layer is reached. This represents an opportunity for damage to occur before removal from the coral tissue has taken place. The extent to which diffusion of ROS across the boundary layer is impacted by flow is yet to be explored. Furthermore, an experiment by Mass et al. (2010 ) directly looking at the effects of flow on photosynthesis under no temperature stress was unable to detect the role of a reactive species in impacting photosynthesis. Alternatively, the effect of flow could be related to an increase in flux of carbon dioxide and oxygen from the coral tissue to the surrounding water column and follow-on effects that this has on photosynthesis and dark respiration of the coral tissue. In this study, the photosystems of endosymbionts retained higher photosynthetic efficiency in high flow, heat treated corals, despite algal cells occurring in similar reduced densities to low flow heat treated corals at the end of the experimental period. This result could be related to increased rates of mass transfer induced by thinner boundary layers under high flow conditions. During the day, the effects of flow on photosynthesis have been suggested to operate at the level of the Rubisco enzyme, a crucial protein used in the dark reaction of photosynthesis for carbon fixation ( Mass et al. , 2010 ). Rubisco is capable of using both carbon dioxide and oxygen as a substrate, the former representing photosynthesis and the latter photorespiration. Photorespiration is a wasteful process when compared to photosynthesis ( Ort and Baker, 2002 ). During the day, higher rates of photosynthetic efficiency and higher effluxes of oxygen from coral tissue have been measured in corals under high flow conditions ( Finelli et al. , 2006 ; Mass et al. , 2010 ), potentially building energy reserves and maintaining a symbiotic relationship for longer. In the same way, measurements of the diffusion boundary layer near the surface of corals have revealed oxygen depletion in low flow conditions during the night ( Shashar et al. , 1993 ), which in turn restricts rates of dark respiration. There is also the potential for high flow conditions to raise coral and endosymbiont respiration rates at night ( Patterson et al. , 1991 ) through the increased flux of oxygen into coral tissues. Higher rates of photosynthetic efficiency could be related to increased rates of mass transfer induced by thinner boundary layers under high flow conditions. Increased rates of mass transfer may reduce any sink limitation (i.e. photorespiration and/or electron flow) causing higher levels of photosynthetic efficiency and increased levels of respiration at night ( Jones et al. , 1998 ) compared to corals under low flow. Under heat stress, this putatively allows endosymbionts of corals under high flow to maintain photosynthetic efficiency for longer and potentially increase levels of respiration and general energetic capacity for repair. This may in turn lead to greater recovery from thermal and light induced damage at night compared to corals under low flow conditions. Equally, the results of this study show that there were no differences in endosymbiont densities between high and low flow corals at the end of both experimental periods. This indicates that thermal stress accumulation throughout both experiments caused enough damage to induce a bleaching response in high and low flow corals. Endosymbiont densities were only recorded at the end of the experimental period, which means we were not able to capture timing of initial declines in density in addition to any deviation in response between high and low flow corals. An interactive effect between flow and temperature treatment on endosymbiont densities was recorded in Experiment 2, indicating that there is some effect of flow on bleaching severity. Further work is needed to uncover the relationship between flow speeds and bleaching responses under thermal stress accumulation. There are a number of other mechanisms through which flow can impact coral function during thermal stress. The thermal boundary layer is analogous to the diffusion boundary layer, where the transfer of heat instead of the diffusion of molecules takes place. It therefore directly affects the temperature to which a coral may be exposed under lower flow conditions, where a thicker boundary layer can limit the exchange of heat transfer ( Jimenez et al. , 2011 ). Under high irradiance and low flow, corals can be 0.2–0.6°C warmer than surrounding water column ( Jimenez et al. , 2011 ). High flow conditions can also increase the probability of particle food capture ( Sebens and Johnson, 1991 ). In this experiment, corals were not artificially fed but because water was taken unfiltered from the reef flat we cannot rule out the possibility of increased heterotrophic feeding of high flow corals. Our study concentrated methodologically on impacts to endosymbiont physiology and the process of endosymbiosis breakdown in coral bleaching. Although generalizable to the coral meta-organism, further investigation is needed into the effects of flow on the coral host, its physiology and heat stress responses. Future studies should also look to quantify whether the beneficial effects of high flow are held into recovery after thermal stress has ceased and survival if thermal stress continued within the system. By the end of both experimental periods, populations of endosymbionts that remained may have been less damaged under high flow, compared to low flow, which would indicate that these fragments have the potential to recover their photosynthetic yields sooner when temperatures return to ambient conditions. Equally, if temperature stress was to continue, at some point (indicated by eventual declines in yields and endosymbiont populations) it seems that the beneficial effect of flow is no longer apparent. Conclusions and implications Investigating flow patterns at the scale of metres within the context of putative beneficial physiological impacts reveals some interesting avenues for coral reef management. For example, current metres deployed on the reef slope showed that reef topography can influence local flow patterns within niche habitats on coral reefs. Variability in flow could be an ecologically relevant but overlooked phenomenon and requires further investigation. Acclimatization due to variability in historic SST trajectories has been shown to significantly impact the molecular mechanisms that underpin thermal tolerance in corals ( Oliver and Palumbi, 2011 ; Schoepf et al. , 2015 ; Ainsworth et al. , 2016 ; Thomas et al. , 2018 ). A recent study investigating the transcriptional responses of corals during tidal fluxes suggested that variability in tide (and therefore flow) is related to acclimatory mechanisms in corals, such as ‘front loading’( Ruiz-Jones and Palumbi, 2017 ). This highlights a limitation of our experiment, in that corals were exposed to constant high or low flow conditions. These conditions were applied for the purpose of understanding whether high or low flow may affect coral functioning under thermal stress, but future studies should look to replicate more closely the diurnal variations in flow speed that corals experience in different reef environments. The predictable timing of neap and spring tide times could also be used to inform management decisions proceeding predicted bleaching events. At larger reef scales, if lower flow caused by neap tides was to occur alongside cloudless days and low wind speeds, shallow reefs could be at risk of high bleaching and mortality ( DeCarlo et al. , 2017 ; Burt et al. , 2019 ). Knowledge of the predominant flow direction at a site may also be able to explain local patterns of bleaching severity. At Coral Gardens (reef slope), the prevailing flow direction at the two exposed metres (1 and 2) was north-westerly, with the highest flow speeds also occurring in this direction. This same pattern has been highlighted in other studies in this region, where a north-westward flow pattern was also measured along the Capricorn Bunker group ( Griffin et al. , 1987 ). These patterns also changed at each metre during neap and spring tidal phases. The morphology of coral colonies provides resistance to the natural flow of water, slowing down and causing turbulence in the water column ( Reidenbach et al. , 2006a ; Reidenbach et al. , 2006b ; Reidenbach et al. , 2007 ). This means that under some circumstances, flow can be greatly reduced at downstream faces of a colony compared to upstream. This effect is dependent on coral morphology but has been hypothesized as causing mortality from the inner sections of branching coral colonies, which further progressed towards colony fringes during an anomalous thermal event in Iriomote, Japan ( Baird et al. , 2018 ). In conclusion, the results presented here indicate that high flow has a mediating effect on the health of endosymbiont photosystems in A. aspera compared to low flow conditions under both sub-bleaching and bleaching levels of thermal stress. We uncover a synergistic interaction between high flow and the protective effect of a pre-stress pulse in temperature. Measurements of physiology indicate the beneficial action of high flow on the efficiency, damage and recovery rate of endosymbiont photosystems under both direct thermal stress, and acclimatory thermal treatments. Endosymbiont photosystems retained photosynthetic efficiency in high flow heat treated corals, despite occurring in similar densities to low flow heat treated corals at the end of both experimental periods. We hypothesize that this effect may be due to increased rates of mass transfer under higher flow. However, despite a delayed onset, under sustained exposure to high levels of thermal stress corals under both high and low flow exhibit declines in photo-physiology and bleaching severity. Further work is needed to understand the effects of flow on coral host responses, in addition to coral recovery after temperature stress has subsided. Through taking a holistic approach to understanding how an environmental driver impacts coral responses to thermal stress, we have investigated the impacts an environmental factor like flow can have on the health and persistence of reefs into the future."
} | 8,789 |
27781084 | PMC5054820 | pmc | 463 | {
"abstract": "Modern society is fueled by fossil energy produced millions of years ago by photosynthetic organisms. Cultivating contemporary photosynthetic producers to generate energy and capture carbon from the atmosphere is one potential approach to sustaining society without disrupting the climate. Algae, photosynthetic aquatic microorganisms, are the fastest growing primary producers in the world and can therefore produce more energy with less land, water, and nutrients than terrestrial plant crops. We review recent progress and challenges in developing bioenergy technology based on algae. A variety of high-value products in addition to biofuels can be harvested from algal biomass, and these may be key to developing algal biotechnology and realizing the commercial potential of these organisms. Aspects of algal biology that differentiate them from plants demand an integrative approach based on genetics, cell biology, ecology, and evolution. We call for a systems approach to research on algal biotechnology rooted in understanding their biology, from the level of genes to ecosystems, and integrating perspectives from physical, chemical, and social sciences to solve one of the most critical outstanding technological problems.",
"introduction": "Introduction Our present day petroleum reserves are the legacy of phytoplankton growing over hundreds of millions of years. The modern day descendants of these sources of fossil energy have usefully retained the ability to produce the same energy-rich compounds that made their ancestors essential to the development of modern society. The tantalizing possibility that biotechnology may harness the capacity of photosynthetic microorganisms to generate energy by fixing carbon from the atmosphere has stimulated a burst of research activity. It remains to be seen when and how photosynthetic microbial biofuel production will help solve the conundrum of how we maintain and extend our modern standards of living without further disrupting the environment. Will microbial biofuels prove to be a silver bullet, one element of a broader solution to the energy economy, or perhaps just an expensive lesson in the limits of biotechnology? Microscopic algae offer clear advantages over terrestrial crops in that they grow at far faster rates and can be cultivated on non-arable land and with non-potable water, lessening the pressure placed on existing food production systems. Here we describe recent progress in understanding the cultivation of microscopic algae for the production of energy. We outline key technical and economic gaps in the pathway toward large-scale commercialization and discuss opportunities for further progress. We argue that realization of the full potential of bioenergy from algae demands a perspective rooted in systems biology that integrates understanding of the genetics, cell biology, physiology, evolution, and ecology of photosynthetic microorganisms. The genetic and physiological origin of traits that determine biochemical composition and growth under variable conditions must be understood in order to optimize strains through classical genetics, breeding, or targeted molecular manipulation of the genome. The interactions between cultivated strains and the diverse assemblages of microbial life that invariably colonize outdoor production ponds or enclosed systems are equally important to commercial success."
} | 844 |
24352945 | PMC3871314 | pmc | 464 | {
"abstract": "Ecosystems have a limited buffering capacity of multiple ecosystem functions against biodiversity loss (i.e. low multifunctional redundancy). We developed a novel theoretical approach to evaluate multifunctional redundancy in a microbial community using the microbial genome database (MBGD) for comparative analysis. In order to fully implement functional information, we defined orthologue richness in a community, each of which is a functionally conservative evolutionary unit in genomes, as an index of community multifunctionality (MF). We constructed a graph of expected orthologue richness in a community (MF) as a function of species richness (SR), fit the power function to SR (i.e. MF = c SR a ), and interpreted the higher exponent a as the lower multifunctional redundancy. Through a microcosm experiment, we confirmed that MF defined by orthologue richness could predict the actual multiple functions. We simulated random and non-random community assemblages using full genomic data of 478 prokaryotic species in the MBGD, and determined that the exponent in microbial communities ranged from 0.55 to 0.75. This exponent range provided a quantitative estimate that a 6.6–8.9% loss limit in SR occurred in a microbial community for an MF reduction no greater than 5%, suggesting a non-negligible initial loss effect of microbial diversity on MF.",
"introduction": "1. Introduction The rapid and continued development of molecular biology and genomic techniques has unveiled immense soil, freshwater and ocean microbial diversity [ 1 – 8 ]. However, the quantitative relationship between microbial diversity and ecosystem function remains unclear. Recent advances in our understanding of plant diversity and ecosystem function in terrestrial ecosystems, which have been achieved by large-scale field experiments [ 9 – 13 ] clearly demonstrate the paucity of studies in microbial communities. Instead of field experiments that manipulate natural microbial diversity levels, ecologists have applied microbial communities as a ‘model’ to test general theory in microcosm settings (see [ 14 , 15 ] for a review). Even in experiments considering relatively high species richness (SR), results indicated the levels of manipulated microbial richness are less than 100, much lower than levels in natural communities [ 16 ]. Furthermore, it remains unclear whether variation in ecosystem functions was attributable to differences in community composition at the species level, highly frequent horizontal functional gene exchange among species [ 17 , 18 ], or rapid functional trait evolution of individual species [ 19 ]. Many studies have demonstrated that the relationships between microbial diversity and ecosystem functions are weak [ 20 and references therein]. Microbial decomposer communities often exhibit high redundancy for a single function, such as microbial respiration and biomass production, which has been shown in more extensive plant biodiversity and ecosystem function studies in terrestrial systems [ 13 , 21 ]. Nielsen et al. [ 20 ] reviewed 57 studies, and concluded that the saturating relationship between microbial richness and a single ecosystem function was dominant in soil ecosystems, suggesting high functional redundancy in soil microbes. In aquatic ecosystems, linear and saturating patterns are a challenge to distinguish owing to a limited range in microbial richness. The relationship demonstrated by experiments (SR was directly manipulated or indirectly manipulated by a dilution–extinction method) was positive [ 22 – 25 ], negative [ 24 ] or non-significant [ 23 – 27 ]. A correlation between microbial richness and a function observed in environmental gradients was positive, fit by an exponential function [ 28 ], linearly negative [ 29 ] and nonlinearly negative for bacterial production and bacterial respiration [ 30 ]. It is notable that the relationship differed when focused on alternative functions [ 23 – 25 , 30 ]. Some studies indicated it was not SR, but heterogeneity in community composition that explained variability in some functions [ 24 , 25 , 31 – 36 ], however, a significant relationship between composition and function was not detected in other studies [ 32 , 35 , 37 , 38 ]. Previous studies found that functional redundancy in microbial communities was high [ 16 , 20 ], suggesting an initial loss in microbial diversity was unlikely to substantially affect ecosystem functions. However, this view (i.e. low effects from initial biodiversity loss on ecosystem functions) was highly sensitive to quantitative measures of microbial function. Peter et al. [ 24 , 27 ] and Langenheder et al. [ 32 ] demonstrated that if the focal function was more specific (e.g. ability to decompose recalcitrant carbon substrates) than general functions (e.g. respiration and biomass production), the link between SR or community composition and function was greater. More importantly, multifunctional redundancy was generally lower (the degree of multiple functional dependence on diversity was higher) than single-functional redundancy [ 27 , 39 ]. The functional composition associated with multiple carbon substrate utilization patterns (revealed by EcoPlate) was often linked to species composition [ 25 , 34 ]. Gilbert et al. [ 40 ] reported a positive correlation between transcript richness (a type of functional richness), and phylogenetic richness by the metatranscriptome approach. These lines of evidence strongly indicated the need to quantify the multifunctionality (MF) of microbial communities. In this way, the role of microbial diversity in ecosystem functioning could be thoroughly evaluated. A growing social demand exists to better project the future magnitudes of change in decreased microbial diversity, and its consequences on ecosystem functions. Reductions in diversity and shifts in microbial species composition may occur in various ecosystems owing to anthropogenic impacts, e.g. increased nitrogen deposition [ 41 ], invasive species introduction and establishment [ 42 ], and toxic substance contamination [ 43 ]. A quantitative assessment/projection of anthropogenic impacts should be undertaken owing to potential trade-offs between ecosystem functions and services provided from natural microbial assemblages (e.g. regulating services) and ecosystem functions provided from artificial ecosystem modifications (e.g. increased crop production and enhanced bioremediation; [ 12 ]). Quantitative MF assessment of natural microbial communities prevents the underestimation of potentially important ecosystem services provided by natural microbial species. In this study, we provided a new theoretical approach for quantitative evaluation of a microbial community by assessing the following two multifunctional indices: (i) MF and (ii) multifunctional redundancy. As much functional information as possible was incorporated into the evaluation of potential microbial functions by analysing the richness of an evolutionary unit of genetic material, i.e. an orthologue; a gene in different species derived from a common ancestor from speciational processes is an orthologue. Orthologues are generally expected to be functionally conservative; therefore, orthologous genes tend to exhibit a similar function [ 44 ]. Because common orthologues are shared by multiple species, we proposed to evaluate orthologue richness in a microbial community, which represents the potential range of functions in the community, and the MF index at the community level. For defining multifunctional redundancy, we used an orthologue accumulation curve , which is a graph of the observed orthologue number (i.e. MF) as a function of SR observed in a community. We subsequently hypothesized that the orthologue accumulation curve can be approximated by the power-law relationship, MF = c SR a . The exponent a serves as a multifunctional redundancy index, whereas c represents the average MF of single species in a community. A smaller a -value can be interpreted as larger multifunctional redundancy, indicating that a loss in SR exhibits fewer impacts on MF. These settings are a natural extension defined for redundancy of a single function [ 21 ]. This approach provides a new method to quantitatively evaluate the impact of change in microbial diversity on ecosystem functions. In order to test the above hypotheses, we conducted community simulations by integrating genomic and ecological information from the database of microbial metagenomics (microbial genome database (MBGD) for comparative analysis, [ 45 ]). We also tested the linkage between MF index defined by the orthologue richness and MF observed in microcosm bacterial communities. This was the initial step to quantify the relationship between microbial diversity and MF of a microbial community. Genomic and ecological information enabled us to conduct extensive simulations, which demonstrated multifunctional redundancy was generally low (0.55 < a < 0.75), and therefore quantitatively supported the importance of maintaining microbial diversity.",
"discussion": "4. Discussion (a) The relationship between biodiversity and multifunctionality The importance of SR for maintaining ecosystem MF has been demonstrated in plant [ 39 , 55 ] and microbial communities [ 27 , 39 , 56 ]. In this study, we proposed a new method for quantifying MF in a microbial community, which uses accumulating metagenomic information. We defined orthologue richness in the community as MF. We also defined multifunctional redundancy by the exponent of power function estimated by fitting the power function to the orthologue accumulation curve; the larger the exponent, the lower the multifunctional redundancy. Our community simulations demonstrated that multifunctional redundancy is generally low. Reich et al. [ 21 ] showed that monofunctional redundancy became lower through time using two datasets derived from grassland biodiversity experiments; the largest exponent values were 0.42 and 0.51 in each dataset. Our simulations generated exponents ranging from 0.57 to 0.71 ( figure 2 a ), which suggests lower multifunctional redundancy rather than monofunctional redundancy, as reported in Gamfeldt et al. [ 39 ]. Deep sea benthic metazoan communities also show low multifunctional redundancy; the functional diversity (trait diversity) is the power function of SR with the exponent 0.59 [ 57 ]. These results imply the consistency of limited multifunctional redundancy in terrestrial plant, metazoan and prokaryotic communities. In our study, the estimated exponent (0.832), which was less than unity, derived from the relationship between species genome size and orthologue richness ( figure 1 a ) suggested a certain degree of multifunctional redundancy within an individual genome. It might be argued that orthologue richness in a community can be predicted (extrapolated) by community genome size using the power-law relationship fit within a genome ( figure 1 a ). However, community simulations demonstrated that orthologue richness in a community was much smaller than predicted by the relationship within a genome ( figure 1 b ). In addition, estimated multifunctional redundancy was higher (i.e. estimated exponent 0.689 was lower) than within an individual genome (exponent 0.832). These results indicated the community cannot be regarded as a single super species with a very large genome. Instead, sharing common orthologues among species is a key mechanism, responsible for multifunctional redundancy (figures 1 b , 2 and 3 ). Randomization simulations indicated that multifunctional redundancy (exponent a ) was influenced by the degree of data coverage: SR ( figure 2 a ) and orthologue richness ( figure 3 ). However, concurrently multifunctional redundancy differed among domains, habitats and oxygen requirements ( figure 2 a ), suggesting multifunctional redundancy was dependent on habitats and environmental conditions in natural systems. Furthermore, when the estimated exponent in a specific community was outside the 95% CI from randomly assembled communities, the exponent was mostly less than the lower 95% CI limit ( figure 2 a ). These results indicated that genome orthologue composition similarities among species in a specific community can be significantly higher than in randomly assembled communities, suggesting environmental selection (environmental filtering [ 11 ]) on orthologue composition. Between domains, multifunctional redundancy was higher in Archaea (A, SR = 58) than Bacteria (B, SR = 420). When the redundancy was separately calculated for each domain in extreme environment community (Ex_t), the exponents were 0.589 and 0.625 for Archaea (SR = 30) and Bacteria (SR = 54), respectively, indicating the higher redundancy in Archaea. It might imply the stronger environmental selection on Archaea community and/or phylogenetically more aggregated choice of species from MBGD for Archaea. Between habitats, an interesting pattern from pseudo-communities is that the marine environment (Mr) showed lower redundancy than the freshwater environment (Fw). Between oxygen requirements, it is worth noting that the facultative groups (Facl) had higher redundancy than anaerobic (An) and aerobic (Aer) groups. Although there are only five examples, results from natural assemblages (TrH, pond, lake, ocean1, ocean2) may imply that the synthesized communities of isolated strains from the identical media (pond and TrH) show functionally more redundant than communities described by non-cultured methods (lake and ocean1, ocean2; the exponent from TrH and pond are much smaller than lake and oceans). (b) Theoretical implications of community simulations and future directions Many gene functions remain elusive despite sequence analyses. For example, in the MBGD database, more than 180 000 orthologues are classified simply as hypothetical proteins. Characterizing the physiological traits of prokaryotes is challenging, even for isolated species. Orthologues (genes that presumably share functions) are an effective unit of distinct function, and a cautious approach to avoid underestimating microbial functions, assuming higher orthologue richness implies higher functional diversity or MF. Our simulations found that removing orthologous genome segments from the analysis overestimated multifunctional redundancy ( figure 3 ). In general, species functions are linked to ecosystem functions, which are characterized by functional traits. These traits determine how organisms respond to environmental changes or effect the environment [ 11 ]. Ecological functions are characterized by ecological traits, which are responsible for ecological interactions between the focal species and other species. Therefore, it is reasonable to assume that all functional genes are potentially related to ecosystem and/or ecological functions, which, in turn, participate in the diverse array of ecosystem services [ 12 ]. Orthologous genes in different species have diverged from a single gene in a common ancestor. Therefore, increased richness of orthologue groups in a community suggests the maintenance of diversification processes during prokaryotic evolution. In other words, MF, which is defined as community orthologue richness, can be characterized as functional diversity, but also evolutionary diversity. Therefore, our approach provides a means to incorporate an evolutionary perspective into biodiversity science (cf. [ 58 ] a phylogenetic diversity measure [ 59 ]). Owing to the limited availability of ecological information in genome databases (e.g. only 16 habitat types) [ 46 , 47 ], MBGD pseudo-communities may include species that do not co-occur in natural environments. Therefore, additional examples of a natural microbial community are valuable to evaluate the robustness of the patterns discovered from MBGD pseudo-communities. These data will provide the concrete SR–MF relationship in a specific environment as is shown in our five examples ( figure 2 ). Multifunctional redundancy comparisons between/among different environments generate a more robust depiction of regional variation in the vulnerability of a microbial community. In addition, species versus orthologue richness can be plotted using data from many regions, which provides biodiversity–MF relationships at larger spatial scales. From methodological point of view, it is also notable that the similarity of naturally occurring species found by a culture-independent method to the most genetically related strain in a genome database can be lower (e.g. lower than 90%; see the electronic supplementary material, table S4) than species found by a culture-dependent method (greater than 97%; see the electronic supplementary material, table S6), owing to the dominance of unculturable species in natural assemblages. Under the limited availability of information of uncultured species in a genome database, the combination of culture-independent and culture-dependent approaches would lead to better understanding of the link between microbial SR and genetic (orthologue) richness in natural environments. Our microcosm experiment was the first step to ascertain the relationship between orthologue richness and microbial processes. Although the degree of manipulation of SR and orthologue richness was small (about 5% reduction from the control), we found that the reduction of the MF directly measured by carbon substrates usage ability could be predicted by the reduction of orthologue richness ( figure 4 ). This result strongly implied the link from potential MF to the actual expression of multiple functions. At the same time, the predictability was generally low (e.g. adjusted r 2 = 0.14 when T = 0.3 in figure 4 ) probably, because we count all of the orthologues equally independently of whether or not their functions are already predicted in MBGD. More detailed analysis on genes with ecological functions will improve the reliability of MF index predicted from genomic data. More interestingly, the low predictability would imply that the community-level functions cannot be fully predicted just by the sum of genetic functions of each species; inter-specific ecological interactions might also matter. Larger scale experiments with larger variations of microbial SR and orthologue richness than our settings will also elucidate the effectiveness of community orthologue richness as the index of ecosystem processes and MF. (c) Importance of quantitative information for conservation and sustainable biodiversity use Former studies that focus on a single ecosystem function or service (e.g. productivity, nutrient retention and resistance to species invasion) indicated that direct supporting evidence of the importance of species number was limited by the small number of species present (10 or fewer) [ 60 ]. Therefore, Diaz et al. [ 60 ] reported the possibility that ‘a reduction in the number of species may initially have small effects’ was difficult to exclude. However, applying a new community MF measure (using a prokaryotic community as a model), our theoretical result demonstrated that even an initial small loss of SR had proportional effects on community (multi)functionality (equation (3.1) and table 1 ) as well as our experimental result from the microcosms ( figure 4 ). In other words, the levels of reduction in SR required to avoid substantial declines in ecosystem services must be very low (e.g. less than 10%). It is also worth noting that our simulations confirmed a general hypothesis generated in other studies that multifunctional redundancy increases as functional diversity decreases ( figure 3 ) [ 27 , 39 , 55 , 56 ]. A contemporary concern for biodiversity conservation is that decision makers require quantitative biodiversity evaluations as part of science-based negotiations and communications. The Convention on Biological Diversity (CBD), Article 14 addresses the importance of appropriate assessment to minimize adverse effects of anthropogenic impacts on biological diversity [ 61 ]. For example, the introduction of living modified organisms (LMOs), including crops and microorganisms to natural environments is suspected to result in unfavourable changes to microbial communities supported by soils and watersheds, leading to a loss in microbial diversity and ecosystem function. The Cartagena Protocol on biosafety (a supplement to CBD) requests decision-making based on scientifically sound risk assessments to identify and evaluate the potential adverse effects of LMOs on the conservation and sustainable use of biological diversity [ 62 ]. However, the quantitative methods to assess the relationship between biological diversity and ecosystem MF remain underexplored, which prevents any quantitative assessment of LMOs adverse effects, leading to biological diversity and ecosystem services decline. This is the case not only for LMOs. In general, enhancement of one ecosystem service in agro- (such as crop production) and natural ecosystems is accompanied by changes in community structure and SR, which in turn degrades other ecosystem services (e.g. water purification) [ 12 ]. Under such trade-offs among different ecosystem service components, and increasing public demands for sustainable biodiversity use that balances costs and benefits, quantitative evaluation of the impacts of biodiversity loss on MF will have greater future importance. The approach proposed in this study serves as a foundation for additional risk assessment developmental procedures to facilitate scientifically sound international decision-making."
} | 5,407 |
36186561 | PMC9516702 | pmc | 469 | {
"abstract": "Smart surfaces with superhydrophobic/superhydrophilic\ncharacteristics\ncan be controlled by external stimuli, such as temperature. These\ntransitions are attributed to the molecular-level conformation of\nthe grafted polymer chains due to the varied interactions at the interface.\nHere, tunable surfaces were prepared by grafting two well-known thermo-responsive\npolymers, poly( N -isopropylacrylamide) (PNIPAM) and\npoly(oligoethylene glycol)methyl ether acrylate (POEGMA 188 ) onto micro-pollen particles of uniform morphology and roughness.\nDirect Raman spectra and thermodynamic analyses revealed that above\nthe lower critical solution temperature, the bonded and free water\nat the interface partially transformed to intermediate water that\ndisrupted the “water cage” surrounding the hydrophobic\ngroups. The increased amounts of intermediate water produced hydrogen\nbonding networks that were less ordered around the polymer grafted\nmicroparticles, inducing a weaker binding interaction at the interface\nand a lower tendency to wet the surface. Combining the roughness factor,\nthe bulk surface assembled by distinct polymer-grafted-pollen microparticles\n(PNIPAM or POEGMA 188 ) could undergo a different wettability\ntransition for liquid under air, water, and oil. This work identifies\nnew perspectives on the interfacial water structure variation at a\nmultiple length scale, which contributed to the temperature-dependent\nsurface wettability transition. It offers inspiration for the application\nof thermo-responsive surface to liquid-gated multiphase separation,\nwater purification and harvesting, biomedical devices, and printing.",
"conclusion": "Conclusions In summary, we studied the temperature-dependent\ninterfacial properties\non thermo-responsive surfaces. Specifically, we investigated the assemblies\nof microparticles grafted with two types of LCST polymers, PNIPAM\nand POEGMA 188 , and correlated the interfacial water structure\nvariation at a multiple length scale with the wettability transition\nof the integrated surface (formed by the self-assembly of microparticles),\nas revealed by Raman measurements, supplementary rheology experiments,\nand confocal microscopy. From the analysis at the nanoscale and macroscale\nlength scales, we concluded that the increased intermediate water\n(decreased bonded water structure) with combined surface roughness\nresulted in the enhanced hydrophobicity. This surface design strategy\nprovides information that correlates the molecular-level conformational\ntransition with the macroscopic surface wettability. Moreover, the\nknowledge and fundamental understanding derived from this study demonstrate\nthe potential application of PNIPAM and POEGMA 188 by controlling\nthe interfacial water structure at the solid–liquid interface\nin other systems, such as biobased responsive surfaces. 43",
"introduction": "Introduction Thermo-responsive polymer composites and\ncoatings are a class of\nsmart materials that find broad applications in wearable devices and\ndrug delivery due to their switchable and programmable properties. 1 Particularly, the thermo-responsive behavior\nof polymers provides an effective strategy to design systems with\ntunable properties. These systems possess interesting physics associated\nwith the conformational transition triggered by temperature that is\nstill not completely understood. Thus far, PNIPAM and POEGMA 188 are the two most widely\nstudied thermo-responsive polymers due to their sharp thermal transition\nand their LCSTs being close to the body temperature. These transitions\nhave been studied using many different techniques, 2 , 3 such\nas laser light scattering, fluorescence spectroscopy, turbidimetry,\ndifferential scanning calorimetry (DSC), infrared spectroscopy, nuclear\nmagnetic resonance, and Raman spectroscopy, to understand the phase\ntransition of thermo-responsive polymers in aqueous solutions. Two\nmain conclusions were derived from these experiments, the coil-to-globule\nstate of polymer and associated functional group transition. 4 For thermal-responsive polymers grafted to bulk\nsurfaces, the current understanding is that the wettability transition\nnear the LCST depended on the exposed functional groups. 5 Nevertheless, the studies consistently claimed\nthat the functionalized surface groups could only affect the water\nmolecule directly in contact with or extremely close to (normally\n1–2 nm) the interface. 6 Thus, fundamental\nunderstanding on how such a short-range interaction that influences\nthe macroscopic wettability transition is necessary for the manipulation\nof the bulk surface wettability. Lycopodium sporopollenin extine\nshell (L.SEC) microparticles have\ngained increasing attention for applications, such as drug carriers, 7 sensors, and soft robotics. 8 Owing to the unique morphology and versatile physical and\nchemical characteristics, L.SEC particles could be a good substrate\nfor grafting thermo-responsive polymer brushes that offer us a flexible\nplatform to investigate the polymer–water interactions. From\nthe microscopic perspective, the grafting of thermo-responsive polymer\nbrushes on rigid particles can enhance the structural stability that\nminimizes the aggregation of the polymer chains. In addition, the\nsurface functionality of the L.SEC offers sufficient amounts of chemical\nreactive sites (hydroxyl groups) for the grafting of the polymer chains.\nMost importantly, these microparticles are naturally produced in plants,\nand hence they are renewable and abundant and are a good source of\nmaterials for a variety of applications. Herein, we prepared\nthermally induced surfaces by grafting two\nrepresentative thermo-responsive polymers, poly( N -isopropylacrylamide) (PNIPAM) and poly(oligoethylene glycol)methyl\nether acrylate (POEGMA 188 ) onto pollen microparticles with\na uniform morphology and roughness to investigate the water–surface\ninteraction at different length scales. We demonstrated that the temperature-dependent\ninterfacial properties of the polymer grafted L.SEC particles are\nassociated with the interaction between interfacial water film and\napolar/polar groups of the polymer brushes, as revealed by in-situ Raman spectra and thermodynamic analysis. Additional\nrheological measurements suggested that the transformation of the\ninterfacial water structure near the polymer brushes can be amplified\non each polymer grafted L.SEC particle due to the rearrangement of\nthe hydrogen bonding network during the LCST transitions. The formation\nof the distinct hydrogen bonding network at the microscale on rough\nsurfaces can effectively induce the bulk surface wettability transition,\nwhich can be elucidated by surface free energy calculations, 3D confocal\nmicroscopy imaging, and liquid contact angles under different environmental\nconditions. This study provides fundamental insight and understanding\ninto the relationship between the interfacial water structure and\nsurface wettability transition. Furthermore, these findings offer\na new route to design thermo-responsive colloids and surfaces (or\nother stimuli-responsive systems) for a wide range of applications,\nsuch as liquid-gated multiphase separation, 9 water purification and harvesting, 10 − 12 biomedical devices, 13 and printing. 14",
"discussion": "Results and Discussion Polymer-L.SEC Microparticles The thermo-responsive\nL.SEC microparticles were synthesized by performing cerium nitrate\n(CAN) free radical polymerization of N -isopropylacrylamide\n(NIPAM) and (oligoethylene glycol) methyl ether acrylate (OEGMA 188 ) in water ( Figure 1 A). First, L.SEC microparticles with a tripartite structure\ndecorated with honeycomb-like microridges (1–2 μm height\nand 200 nm width) on the external surface with a uniform shape of\n29.02 μm were prepared via the KOH extraction process ( Figure S1 ). 8 Notably,\nthe hollow L.SEC with a large surface area offered a facile method\nto control the external polymer layer architecture, consisting of\npolymer grafting density and chain length. 15 We manipulated these two factors by changing the monomer/initiator\nmass ratio and polymerization time. L.SECs were designated as PNm-g-L.SEC\nand POm-g-L.SEC, where m corresponds to the molar\nratio of the monomer to 0.1 equiv of the initiator varying in 5, 10,\n20, and 40 mmol. A brown L.SEC powder was obtained and characterized\nby scanning electron microscopy (SEM), showing the surface morphology\nof PNm-g-L.SEC and POm-g-L.SEC with a higher roughness nanostructure\non micro-ridges owing to the polymer grafted canopy compared to pristine\nL.SECs ( Figure 1 B).\nSpecifically, the architecture of the polymer grafted canopy of PNm-g-L.SEC\nparticles transformed from a sparse to dense structure with a rough\nnanostructure that increased with the increasing grafting ratio ( Figure S2 ). Moreover, changes in the particle\nsize were strongly associated with the polymer layer thickness as\nsummarized in Table S2 . For example, the\nsize of PN-g-L.SEC increased from 30.51 to 31.62 μm as m increased from 5 to 40 mmol, corresponding with the epicuticular\npolymers covering the microridges on the outer surface becoming dense\nand increasing the thickness from 260.9 to 762.3 nm. POm-g-L.SEC showed\na similar structural change as PNm-g-L.SEC. The induced structural\nand hydrophobicity change of modified L.SECs (PNm-g-L.SEC and POm-g-L.SEC)\nhad a profound impact on the L.SEC-based thermo-responsive surfaces.\nFourier transform infrared spectroscopy (FT-IR) further confirmed\nthe successful grafting of PNIPAM and POEGMA 188 on the\nextine of L.SEC. 16 The IR spectrum of PN10-g-L.SEC\ndisplayed two characteristic peaks that confirmed the presence of\namine groups, where the first band at 1650 cm –1 is\nassociated with the N-C=O bond, while the absorption peak at 1550\ncm –1 corresponds to the N–H bonds. The isopropyl\ngroups (IP) were confirmed by the IR spectra over the range of 2500\nto 4000 cm –1 . The peak at 2970 cm –1 was assigned to the antisymmetric and symmetric CH stretch of the\nmethyl groups, while the peaks at around 2850 and 2871 cm –1 were derived from the symmetric stretches of CH 2 and\nCH 3 , respectively. 17 As shown\nin Figure 1 C, PO10-g-L.SEC\ndisplayed a unique peak on 1740 cm –1 , which corresponds\nto the ester linkage between the methacrylate and oligoethylene glycol\nside chains of the POEGMA 188 graft brushes. 4 Further evidence on the successful polymerization of NIPAM\nand OEGMA from the surface is provided by X-ray photoelectron spectroscopic\n(XPS) elemental analyses, confirming that this layer contained organic\nmolecules with the expected changes in the ratio of the C–N\nbond on the PNm-g-L.SEC surface and C–O bond on the POm-g-L.SEC\nsurface. ( Figure S3 and Table S1 ). 18 Figure 1 (A) Schematic of preparation of thermo-responsive L.SEC\nparticles\nand surfaces by grafting PNIPAM and POEGMA 188 . (B) SEM\nimages of pristine L.SEC. (C) FT-IR spectra of PN10-g-L.SEC, PO10-g-L.SEC,\nand L.SEC. (D) DSC measurement on PNm-g-L.SEC ranging the temperature\nfrom 15 to 50 °C with a scanning rate of 1 °C/min. (E) Turbidity\nchange of PN10-g-L.SEC aqueous solution measured by UV–vis\nspectrophotometry. (F) Surface tension of PNm-g-L.SEC detected by\na tensiometer. With regards to thermal responsive characteristics,\nthe polymer\nconformational transition endowed the L.SEC with different interactions\nwith water molecules that are dependent on the thermal characteristics,\nsuch as the lower critical solution temperature (LCST). The LCST driven\nby temperature was typically associated with the turbidity change,\nwhich could be measured by UV–vis spectrophotometry. Figure 1 E reveals the LCST\nof PN10-g-L.SEC solutions at ∼35.2 °C caused by the hydrophobic\neffect of the isopropyl groups (IP) and bound water around the amine\n(−NH) and carbonyl groups (-C=O). In addition, the phase transition\nof PNm-g-L.SEC was further investigated by differential scanning calorimetry\n(DSC) over consecutive heating/cooling cycles. 19 The LCST transition of the PNm-g-L.SEC revealed two different\nthermal processes as shown in Figure 1 D, a typical endothermic peak at ∼35.8 °C\nand an exothermic peak at ∼36.2 °C of PN10-g-L.SEC. These\nthermal transitions are related to the dehydration of apolar groups\n(IP) and the interactions between polar groups and water molecules.\nBelow the LCST, the polymer chains are in a random coil conformation,\nwith “water cages” surrounding the IP and water molecules\nor bonding with the -NH and -C=O groups. 20 At temperatures above the LCST, the entropy of the polymer–water\nsystem dominated, which was unfavorable for the exothermic formation\nof hydrogen bonds. Thus, the water cages surrounding the IP groups\nwere disrupted together with the bound water molecules that were released\nto increase their entropy and the polymers collapsed into a globular\nstate. 21 The phase transition of PN10-g-L.SEC\nwas evident from the inset of Figure 1 E determined from the changes in the solution turbidity.\nPN10-g-L.SEC was well-dispersed in an aqueous solution with an average\nradius of 31.07 μm at 20 °C, and the solution transformed\ninto a turbid dispersion at 40 °C that became insoluble and the\nsuspension transformed into a brown opaque color, confirming that\nthe grafted PNIPAM chains collapsed and transformed into globules,\nwhere the PN10-g-L.SEC possessed an average radius of 30.09 μm\n( Figure S4 ). The polymer chain conformational\ntransition can be traced by the\ninterfacial arrangement of these hydrophobic/hydrophilic segments\non PN10-g-L.SEC, which was indicated by changes in the dynamic temperature-dependent\nsurface tension and interfacial behavior between polar/apolar solvents.\nThe dependence of the surface tension with changes in temperatures\nare displayed in Figure S8 D, where γ LV of PN10-g-L.SEC was 52.1 mN m –1 at 20\n°C decreasing to 38.7 mN m –1 at 40 °C.\nThe results demonstrated the predominately strong hydrogen bonding\ninteracting between water molecules with hydrophilic −NH and\n-C=O groups at low temperature while the exposure of large amounts\nof hydrophobic IP groups above the LCST reduced the surface tension.\nThe surface tension of PN10-g-L.SEC was thermally responsive and reversible\nduring the heating-and-cooling cycles, while the rearrangement of\nhydrophobic/hydrophilic groups induced by temperature could also be\ndetected via the time-dependent interfacial tension data ( Figure S5 ). 22 With\nincreasing temperature, the hydrophilic groups formed intramolecular\ninteractions instead of hydrogen bonds with water, where the well-dispersed\nPN10-g-L.SEC particles in the water phase became hydrophobic and rearranged\nat the polar/apolar interface. This interfacial behavior of PN10-g-L.SEC\nresulted in the change of the droplet shape, and the simulated value\nof the interfacial tension decreased around their LCST, which was\nconsistent with the previous phase transition results. Moreover,\ndifferent polymer chain structural arrangements were\ntriggered by temperature, which could be deduced from surface tension.\nAs a result, the equilibrium surface tension (γ LV ) of pure L.SEC aqueous solution was about 71.9 mN m –1 , which was close to pure water (72.8 mN m –1 ).\nWith increasing grafting polymer ratio, the amounts of grafted polymer\nchains on L.SEC increased and γ LV decreased sharply\nfrom 68.3 for PN5-g-L.SEC to 44.1 mN m –1 for PN40-g-L.SEC( Figure 1 F). The results demonstrated\nthe amounts of grafted polymers as deduced from the grafting density\nand chain length at a low grafting ratio, where the IP groups on the\nside chain displayed a flat conformation on the pollen surface. However,\nat a high grafting ratio, the rearrangement of IP groups resulted\nin an extended and ordered chain conformation, that reduced the water\naffinity. In addition, the polymer-L.SEC particle morphology could\naffect the adsorption at the air/water interface. To demonstrate this,\nwe compared the interfacial behaviors between PN10-g-L.SEC microparticles\nand PNIPAM microgel, where the surface tension variation was recorded.\nAs shown in Figure S7, S8 , the surface\ntension of PN10-g-L.SEC cycled over a smaller range (∼12 °C),\nand the process was reversible over several cycles, while surface\ntension of the PNIPAM soft microgel possessed a larger temperature\nrange (∼17–18 °C), and the temperature decreased\nwith each cycle. This is caused by the soft particles deforming and\nspreading over a larger interfacial area since they possessed a higher\nadsorption energy compared to rigid particles. 23 When two deformed microgels are forced into close proximity,\ntheir size and shape changed irreversibly resulting in a reduced surface\ntension after several cyclings of between 20 and 40 °C. However,\nthe polymer-L.SEC particles adsorbed at the interface according to\nthe Young–Dupré relationship, 24 endowing them with reversible chemical structural and morphological\ntransition that further ensure a more flexible interaction between\nwater and the stable interparticle interaction compared to the microgel.\nTherefore, the polymer-L.SEC particles are good candidates for the\nconstruction of thermo-responsive surfaces to achieve a tunable and\nreversible wettability transition. To compare the influence\nof molecular structuring on the hydrophobic\ntransition, POm-g-L.SECs were used as reference samples since their\npolymer chain transitions were different from PNIPAM. The LCST of\nPO10-g-L.SEC occurred in a reversible phase transition at ∼26.9\n°C ( Figure S6 ). Given the mechanism\ngoverning the phase transition, the polyethylene glycol (PEG) side\nchains were solubilized at low temperatures due to the extensive hydrogen\nbonding between the ether oxygen and water hydrogen atoms. When the\ntemperature was increased beyond its LCST of 26.9 °C, the polymer–polymer\ninteraction became more thermodynamically favorable compared to polymer–water\ninteraction, causing the PEG chains to collapse onto the methacrylate\nbackbone forming an insoluble globule resulting in a turbid solution. 25 The hydrophobic methyl methacrylates (MMA) counterbalanced\nthis hydrophilic character of oligo(ethylene glycol) (OEG) groups\nleading to a competitive hydrophobic effect. In addition, the LCST\ntemperature of PNm-g-L.SEC was found, similarly to POm-g-L.SEC, to\nbe relatively independent of polymer stereoregularity ( Figure S6 ). Temperature-Dependent Macroscopic Surface Wettability Transformation To gain a deep insight into the interactions between water and\nPNm-g-L.SEC and POm-g-L.SEC, we prepared surfaces with the L.SEC microparticles\nvia the self-assembly of L.SEC microparticles onto a substrate. The\ncombined hierarchical structure reinforced the surface hydrophobic\ntransformation that highlighted the interaction between polymer chain\nand water molecules, providing a clear and visual picture to demonstrate\nthe process. 26 The surface topography and\nroughness factor of the L.SEC-based surface were investigated by SEM\nand laser confocal microscopy. The PN10-g-L.SEC surface possessed\na porous network structure with a roughness factor ( R q ) of 2.32 μm, where the green–yellow–red\nregions corresponded to the surface protrusions consisting of assembled\nL.SEC microparticles and blue regions represented the “valleys”\nbetween the protrusions ( Figure 2 B,C). The convex protrusions comprised self-assembled\nL.SEC microparticles as indicated by the SEM image ( Figure S9 ). Notably, this surface design strategy could amplify\nthe molecular-level conformational transition for tuning the macroscopic\nsurface characteristics. Figure 2 (A) Illustration of surface wettability transition\ncorresponding\nto the coil-to-globule transition driven by temperature. Surface topography\nof PN10-g-L.SEC surface measured by laser confocal microscopy: (B)\nmagnification 10×, and(C) magnification 50×. (D) Variation\nof contact angle on the PN10-g-L.SEC surface under different environmental\nconditions. The surface wettability behavior is a key parameter\nto determine\nthe relationship between water and the substrate at the macroscale\nas characterized by contact angles. Since the surface conformation\nand structure influenced the water contact angle under air (θ w ), we separated the surfaces into three types for comparison.\nAt a low grafting ratio of polymer (m ∼ 5 mmol), the grafted\nchain was randomly distributed on the L.SEC surface with low nanoscale\nroughness, resulting in insufficient functional groups on the outer\nsurface yielding a mild response to temperature changes. As shown\nin Figure 3 A, the PN5-g-L.SEC\nsurface was hydrophilic (θ w ∼ 22°) regardless\nof the temperature, and the molecular conformation transition at the\nnanoscale could not alter the surface wettability at the macroscale. 27 A similar surface wettability characteristic\nwas observed for PO5-g-L.SEC surfaces, ( Figure 3 D) showing hydrophilicity with a θ w of 22.4 o (20 °C) and 22.1° (40 °C)\nat low grafting density. Notably, the grafting ratio on L.SEC was\na key factor in determining the nano/microstructure of polymer-L.SEC,\nwhich further influenced the overall surface topography. Figure 3 Variation of\ncontact angles including water contact angle under\nair and oil contact angle under water: (A) PN5-g-L.SEC, (B) PN10-g-L.SEC,\n(C) PN40-g-L.SEC, (D) PO5-g-L.SEC, (E) PO10-g-L.SEC, and (F) PO40-g-L.SEC. As for PN40-g-L.SEC surface ( Figure S9 ), the surface roughness ( R q ∼\n1.28 μm) was reduced due to the high grafting ratio of polymer\nbrushes that covered the L.SEC walls and microbridges. The result\nshowed that the high grafting ratio of the polymer altered the hierarchical\nnano/microstructure, where the increased nano roughness dramatically\nreduced the micro-roughness. In the case of PN40-g-L.SEC, the surface\ndisplayed a hydrophilic/hydrophobic characteristic with a θ w of 54 ° at 20 °C and 103° at 40 °C ( Figure 3 C). The reduced microstructure\nled to the hydrophilic/hydrophobic transition of PN40-g-L.SEC surface,\nwhich exhibited a similar trend as the flat polymer surface. 28 Here, with the enhanced nano/microstructure,\nthe surface wettability transformation phenomenon was induced by the\nmolecular structure, orientation, and restructuring of the polymer\nchain occurring at the water/PNIPAM interface driven by temperature.\nAt low temperatures, the -C=O and -NH groups displayed strong hydrogen\nbonding with water molecules around the PNIPAM chains that impacted\nthe air/PNIPAM/water interface. When the temperature exceeded the\nLCST, the polymer chains collapsed and the hydrophilic -C=O and -NH\ngroups interacted via hydrogen bonds while the exposed hydrophobic\nIP moieties near the surface of the collapsed chains contributed to\nthe hydrophobicity of the PN-g-L.SEC outer surface. In addition, based\non the theory of similarity-inter miscibility, hydrophobic groups 3 ) extended to the oil phase and the hydrophilic segments (−NH 2 , C=O groups) extended toward the water phase. The\norientation of the hydrophobic/hydrophilic moieties of the functional\ngroups on the polymer brushes was confirmed by the oil CA under water\n(θ o/w ) and the water CA under oil (θ w/o ). PN40-g-L.SEC(20 °C) surface was in a metastable state, possessing\nunder-oil superhydrophobic (θ wo ∼ 150.5 ° ) and under-water oleophobic (θ ow ∼151°) characteristics\n( Figure S10 ). Figure 3 F shows the thermal switching between hydrophilic\nand hydrophobic states for the PO40-g-L.SEC surface, where the static\nθ w oscillated between 22° at 20 °C and\n64° at 40 °C. Although this surface showed a similar thermal\nwettability transition from hydrophilic to hydrophobic, the molecular\nstructural transformation of POm-g-L.SEC surfaces were different from\nPNm-g-L.SEC. To elucidate the surface wettability generated\nby the hydrophobic/hydrophilic\ngroups, the conformational transition of the polymer brushes on the\nL.SEC surfaces were analyzed by comparing the total interfacial energy\nof different wetting states ( Figure S13 ). 29 We compared the total interfacial\nenergies of L.SEC-based surfaces at different temperatures that were\ncompletely wetted by either water ( E w )\nor an arbitrary immiscible oil (hexane) ( E o ) or without a fully wetted immiscible water floating on top ( E wo ). 30 The surface\nwas wetted preferentially by water or oil, and the conformational\ntransition of hydrophobic and hydrophilic moieties on the PNIPAM and\nPOEGMA 188 triggered by temperature could be estimated.\nFor the PO40-g-L.SEC (20 °C) surface, the comparison of interfacial\nenergy ( E w < E o but E w > E wo ) indicated that the surface would be preferentially\nwetted\nby water, forming a stable water–solid interface and displaying\nhydrophilicity in oil and oleophobicity in water ( Table S3 ). Under this wetting state, the ether oxygens of\nPEG on the exterior of the collapsed globules bonded with the water\nmolecules. However, the interfacial energy indicated that the water-(PO40-g-L.SEC\n(40 °C)) surface was thermodynamically unstable, whereas the\nexperimental results showed that the PO40-g-L.SEC(40 °C) surface\nexhibited an under-oil superhydrophilic (θ wo ∼\n155.5°) and under-water superoleophobic (θ ow ∼ 152.0°) characteristics. The under-oil superhydrophilic\ncharacteristics suggested that the hydrophobic moieties consisting\nof the methyl groups on the main chain of MMA favored the exterior\nenvironment and repelled water, resulting from the enhanced polymer–polymer\ninteractions on the PO40-g-L.SEC (40 °C) surface. In addition,\nthe under-water superoleophobic surface suggested that the hydrogen\nbonding between the side chain of POEGMA 188 and water persisted\nin forming a layer of water film that repelled oil, preventing its\ninfusion to the surface. Thus, the molecular conformational transition\nof the POm-g-L.SEC system associated with the ether oxygens of PEG\non the outer surface of the polymer chains bonded with water molecules\nbelow the LCST (26.9 °C). Above the LCST, this balance was disrupted\nand the interaction between the side chain of POEGMA 188 and water was reduced, resulting in the enhanced polymer–polymer\ninteractions over polymer–water interactions. Mechanism Investigation of Interfacial Water at Multiple Length\nScale and Surface Wettability Transition Although previous\nstudies suggested that the surface wettability change was induced\nby the collapsed and extended state of the polymer chains, the associated\nwater structure at the molecular interfaces needed to be determined.\nTo better explore the macroscopic wettability phenomenon at the air/polymer-g-L.SEC/water\ninterface, in situ Raman spectroscopy was used to investigate the\nchanges in the structure and dynamics of water induced by the functional\ngroups on the polymer chains. 31 Here, we\npresented experimental evidence that revealed a similarity between\nthe structure of water around the hydrophilic/hydrophobic groups and\nat macroscopic air/PNm-g-L.SEC/water interfaces. Figure 4 shows a typical Raman spectrum\nof a fully prewetted and hydrated PN40-g-L.SEC surface over the temperature\nrange from 20 to 40 °C. A broad Raman band extending from 3000\nto 4000 cm –1 is related to the vibration of hydrogen\nbonds in water, where the 3250, 3410, 3520, and 3630 cm –1 were assigned to different types of water structures. Gaussian fittings\nof the spectra showed that the O–H stretching band could be\nresolved into three distinct components, corresponding to three types\nof O–H stretching vibrations. The low wavenumber component\nof the 3250 cm –1 peak is associated with the vibration\nof 4-coordinate hydrogen-bonded water (4-HBW), which is attributed\nto free water with four hydrogen bonds (①) and IP groups surrounded\nby polyhedral cages composed of tetrahedrally hydrogen-bonded water\nmolecules (②). 20 Whereas, the 3410\ncm –1 is associated with the in-phase vibrations\nof water molecules captured by the -C=O or -NH groups (③) of\nPNIPAM, which is regarded as bound water (BW). The high wavenumber\ncomponents at 3520 and 3630 cm –1 correspond to the\nstretching of the weak or non-hydrogen bonded water molecules (④⑤),\nwhich is regarded as intermediate water (IW), reflecting the hydrophobic\ndisordered water in the hydration shell. 32 Note that the hydration-shell OH band possessed a different shape\nwith changes in temperature ranging from 20 to 40 °C. At 20 °C,\nthe resulting spectra revealed two small dangling (non-hydrogen-bonded)\nwater OH peaks near 3520 and 3630 cm –1 as well as\ntwo broad overlapping hydrogen-bonded OH features near 3250 and 3410\ncm –1 . Moreover, at 40 °C, the relative intensity\nof the Raman band near 3250 cm –1 decreased significantly,\nwhereas the high wavenumber components at 3520 and 3630 cm –1 showed a dramatic increase. The shift in the spectrum for water\nstructure transformation was attributed to two stages in the polymer\nchain transition. On one hand, the -C=O or -NH groups formed a strong\nhydrogen bond around water molecules at low temperatures, which was\nreplaced by intramolecular interaction with each other that weakened\nthe interaction with water molecules above the LCST. On the other\nhand, IP groups are surrounded by the hydrophobic hydration shells\n(“water cage”) that resemble solid clathrate hydrates\nbelow the LCST, which were tetrahedral with fewer weak hydrogen bonds\ncompared to bulk water. When the temperature was increased to 40 °C,\nthe hydration shell transformed dramatically into a less-ordered and\nweaker H-bonded structure, accompanied by the peaks with a lower intensity\nof 3250 cm –1 and enhanced intensity near 3520 and\n3630 cm –1 . These results were observed for the PN40-g-L.SEC\nsurface, where the phase transition of PNIPAM chains was accompanied\nby a water structure transformation. Figure 4 Water structure along the PNIPAM polymer\nchain: (A) below LCST\nand (B) above LCST. Proportion of the water ratio measured by Raman\nspectroscopy: (C) PN40-g-L.SEC surface below LCST and (D) PN40-g-L.SEC\nsurface above LCST. A better description of the three types of water\nstructures on\nthe PN40-g-L.SEC surface can be described by comparing the proportion\nof the O–H stretching vibrational band. Analysis of the results\nin Figure 4 C,D revealed\nthat as the temperature was increased from 20 to 40 °C, the proportion\nof 4-HBW decreased from 35.7 ± 1.5 to 21.1 ± 1.3%, BW varied\nbetween 36.8 ± 1.4 and 33.0 ± 1.3%, and IW increased from\n27.5 ± 1.0 to 45.9 ± 1.2%. Because of the proportion of\n4-HBW and BW due to the disruption of the “water cage”\nsurrounding IP groups and dehydration of -C=O and −NH, respectively,\nwe could compare the ratio of 4-HBW/BW to analyze these hydrophilic/hydrophobic\norientation and transition around the LCST. Raman mappings on an area\nof 40 × 40 μm 2 clearly showed the changes of\nthe hydrophilic/hydrophobic orientation and transition on the PN40-g-L.SEC\nsurface at different temperatures ( Figure 6 ). The signal contrast between 20 and 40\n°C indicated that the ratio of 4-HBW/BW was higher at 20 °C,\nrevealing that the surface possessed more bound water and the hydrophilic\n-C=O and −NH groups assembled mainly at the air/PNm-g-L.SEC/water\ninterface. However, the ratio of 4-HBW/BW showed a dramatic drop over\nthe whole area at 40 °C, indicating that the water transformed\ninto a less ordered and weaker hydrogen-bonded structure and the preferred\nexposure of hydrophobic IP groups rearranged at the interface. Furthermore, the O–H stretching spectra of the PO40-g-L.SEC\nsurface was selected to investigate the relationship between the water\nstructure and polymer conformation. Theoretical predictions of the\nconformation of POEGMA 188 brushes possessed a hydrophobic\nmain chain and a hydrophilic side chain. Figure 5 shows the Raman spectra of PO40-g-L.SEC\nat 20 and 40 °C, and the proportion of 4-HBW decreased from 35.9\n± 1.3 to 28.7 ± 1.1%, BW varied between 37.8 ± 1.3\nand 34.1 ± 1.3%, and IW increased from 26.3 to 37.2 ± 1.0%.\nMoreover, the ratio of 4-HBW/BW on PO40-g-L.SEC also showed lower\nfluctuations. These results confirmed that the POEGMA 188 comprised a strong interaction between C–O and water and\na weak hydrophobic hydration shell around the −CH 3 of the main chain. Additionally, the signal contrast on Raman mapping\nof PO40-g-L.SEC revealed that the change in the ratio of 4-HBW/BW\nwas less than the PN40-g-L.SEC surface. Owing to the strong water\naffinity of C–O groups and the chain structure, the IW layer\noccurred on the main chain with increasing temperature surrounded\nby a BW layer ( Figure 6 ). Thus, the hydrophobic hydration shell\nwas different from the “water cage” surrounding IP groups\nof PNIPAM that did not disappear with increasing temperature, resulting\nin a lower hydrophobic character of the −CH 3 groups. Figure 5 Water\nstructure along the POEGMA 188 polymer chain: (A)\nbelow LCST and (B) above LCST. Proportion of water ratio measured\nby Raman spectroscopy: (C) PO40-g-L.SEC surface below LCST and (D)\nPO40-g-L.SEC surface above LCST. Figure 6 (A) Illustration of temperature-dependent Raman spectroscopy.\nWater\nratio of 4-HBW/BW changing with the temperature detected by Raman\nmapping (the ratio was measured at each pixel ranging from 0.6 to\n1.0 that recorded as red to blue): (B) PN40-g-L.SEC surface at 20\n°C, (C) PN40-g-L.SEC surface at 40 °C, (D) PO40-g-L.SEC\nsurface at 20 °C, and (E) PO40-g-L.SEC surface at 40 °C. The water structural variation occurred around\nthe thermo-responsive\npolymer chains on a molecular level. However, as exhibited by the\nmacroscopic evidence (wettability transformation), such a short-ranged\ninterfacial effect determined the macroscale surface wettability transition\nwhen combined with the surface roughness. This could be attributed\nto the grafting of polymer brushes on the rigid and rough lycopodium\npollen, where the water structure variation near the thermo-responsive\npolymer chain was optimum for the whole pollen particle due to the\ntendency of the water to maintain the integrity of its hydrogen bond\nnetwork. 33 , 34 This phenomenon could be demonstrated by\nrheological analysis of concentrated pollen suspensions. The\ndense suspensions with high volume fraction of particles were\nprepared near the maximum packing volume fraction, and shear rheological\nmeasurements were performed. 35 In general,\nthe rheological properties are extremely sensitive to the interparticle\ninteractions and hydration shell around the particles at the nanoscale.\nDue to the higher content of 4-coordinate hydrogen-bonded water (4-HBW)\nand bound water (BW) around the exposed and swollen polymer chain\non the PN40-g-L.SEC surface, their ordered hydrogen bond network 36 induced a stable and thick lubricating hydration\nshell, which kept the particle surfaces separated until a critical\nload was exceeded resulting in the interpenetration of the brushes\nbelow the LCST. This stable lubrication layer would be disrupted at\na high shear rate (10 2 s –1 ), and the\nhydrodynamic rearrangement of the particles generated larger clusters\nof aggregated particles, leading to a smooth and reversible viscosity\nincrease (continuous shear thickening in Figure S11 ). 37 Above the LCST, the interfacial\nwater structure would be substantially altered, where more intermediate\nwater structures were formed around the collapsed polymer chain disrupting\nthe original tetrahedral hydrogen bond structure, yielding distorted\nand heterogeneous network brushes. 38 Thus,\nthe thinner hydration shell and unstable hydrogen bond system would\nform at the interface, where the lubrication layer (hydration shell)\ncould be readily disrupted under the hydrodynamic force that produced\na higher viscosity signified by the shear thickening behavior. The\nchanges in the rheological properties with temperature could also\nbe observed for the dense PNIPAM microgel suspension. The shear stress\nwould be suppressed, and discontinuous shear thickening occurred at\ntemperatures exceeding the LCST (red open triangles of Figure S11 ). The rheological profiles agreed\nwith the prediction of recent molecular dynamics calculation and experiments. 35 The PO40-g-L.SEC displayed similar trend in\nthe rheological behavior, where the shear thickening behavior was\nenhanced at high temperature. However, the conformational transition\nof POEGMA 188 brushes led to lower amounts of intermediate\nwater contents as determined by the Raman measurements. Therefore,\nit could still form a stable hydrogen network between the PO40-g-L.SEC\nparticles when compared with PN40-g-L.SEC at temperatures beyond the\nLCST, showing a smaller shear thickening enhancement ( Figure S11B ). In addition, additional rheological\nexperiments were conducted on another type of polymer grafted pollen\nmicroparticles (Lotus), where the Lotus pollen possessed a similar\nsize but with a different surface roughness ( Figure S12 ). Both the PN40-g-L.SEC and PN40-g-Lotus particles displayed\ntemperature-dependent viscosity variations. However, above the LCST,\nPN40-g-L.SEC possessed a higher shear thickening effect compared to\nPN40-g-Lotus. 39 This is attributed to the\nrougher PN40-g-L.SEC particle with an enhanced proportion of polymer–water\ninterface, leading to a larger heterogeneity of the water structures\ndue to the increased density of intermediate water. 40 Moreover, surface free energy (γ s ) was affected\nby the rearrangement of hydrophilic/hydrophobic moieties, which is\na key parameter to determine the wettability at the integrated air–liquid–solid\ninterface. According to the Owens, Wendt, Rabel, and Kaelble (OWRK)\ntheory, 41 the surface free energy consisted\nof both polar (γ s p ) and dispersive components (γ s d ), where the two polar -C=O, −NH\ngroups, and apolar IP groups of PNIPAM contributed to γ s p and γ s d . Tables S3 and S4 provide a summary of the changes of surface\nfree energy on PN-g-L.SEC and PO-g-L.SEC surfaces for the temperature\nat 20 and 40 °C. For the PN40-g-L.SEC surface, when the temperature\nwas increased from 20 to 40 °C, γ s p decreased from 15.8 to 0.04 mJ m –2 and γ s d decreased from 39.5 to 32.6 mJ m –2 , suggesting that the surface switched from hydrophilic to hydrophobic\nand γ s of 55.2 mJ m –2 decreased\nto 32.6 mJ m –2 . It can be concluded that the surface\nfree energy contributed to the increased hydrophobicity of the surface\nowing to the polar/apolar component’s conformational transition. The temperature-dependent polymer conformational transition followed\nby the surrounding water structural transformation and surface free\nenergy fluctuations further impacted the water structure at a macroscopic\nair–water-solid interface. When the grafting ratio increased\nto a critical value, the surface could display a switchable wettability\nbehavior driven by temperature that amplified the interaction modes\nbetween the water molecules and the polymer chains. We selected two\nrepresentative states of PN10-g-L.SEC and PO10-g-L.SEC surfaces to\nillustrate the surface superhydrophobicity transformation. Interestingly,\nfor the PN10-L.SEC surface with an R q of\n2.32 μm, the surface displayed a superhydrophobic/superhydrophilic\nchange with a CA of 3.1 ° at 20 °C and 154.3° at 40\n°C ( Figure 3 B, Movie S1 and S2 ).\nThe experimental results showed that the PN10-g-L.SEC (20 °C)\nsurfaces with under-water superoleophobicity (θ ow ∼ 151.5°) changed to θ ow ∼ 1.4°\nat 40 °C ( Figure 2 D). This demonstrated that the hydrophilic moieties of PNIPAM brushes\non the PN10-g-L.SEC surface extended toward the water phase at 20\n°C. Above the LCST (at 40 °C), the dehydration of hydrophobic\ngroups induced the oil phase to displace the water phase on the surface.\nWhen the surface was prewetted by hexane, the IP groups interacted\nwith hexane and repelled water driven by solvation ( Figure 2 D). Thus, the PN10-g-L.SEC\nsurface exposed more hydrophilic groups at low temperatures, which\ntransformed into the hydrophobic groups at higher temperature. A large\namount of hydrophobic groups (2 methyl/per unit) endowed the surface\nwith a low surface free energy of 29.3 mJ m –2 , and\nin combination with the hierarchical structure, it produced the superhydrophobic\ncharacteristic as indicated by the CA above the LCST. A similar thermal\nswitching between superhydrophobicity and superhydrophilicity was\nobserved for the PN20-g-L.SEC surface ( Figure S14 ). Additionally, the PN10-g-L.SEC surface possessed a rapid\ntransformation between superhydrophilicity and superhydrophobicity\nsince a single cycle took only several minutes, changing between 2\nand 150° on the PN10-g-L.SEC surface when the temperature cycled\nbetween 20 and 40 °C. This reversibility of the surface hydrophobicity\nremained after the sample was stored without special protection for\nmore than three months, confirming that the polymer-L.SEC was robust\nand stable. Additionally, we observed that the PO10-g-L.SEC\nsurface displayed\nsuperhydrophilic characteristics in contrast to PN10-g-L.SEC. The\ninterfacial behavior at air/ POm-g-L.SEC/water and the corresponding\nwater affinity behavior were determined by the surface free energy\nof the hydrophobic MMA and hydrophilic OEG groups. Interestingly,\nthe R q was 2.25 μm for the PO10-g-L.SEC\nsurfaces, where the CAs remained constant at 1 o (20 °C)\nand 2° (40 °C), displaying superhydrophilicity without wettability\ntransition ( Figure 3 E). Below the LCST (26.9 °C), the ether groups on PEG segments\nformed hydrogen bonds with water molecules. However, above the LCST,\nthis balance was disrupted and the interaction between the side chain\nof POEGMA 188 and water decreased, resulting in the enhanced\npolymer–polymer interactions over polymer–water interactions.\nThe POEGMA 188 chains collapsed into a globule conformation\nwith the OEG chains/groups surrounding the hydrophobic MMA backbone\nyielding a less hydrophobic state (θ i <90°), where the water repellent characteristic was less\nsevere compared to PN10-g-L.SEC. 42 This\nphase transition behavior led to a more hydrophobic of POEGMA 188 globules above the LCST compared to the solvated chains\nat low temperature, with the overall characteristics being somewhat\nhydrophilic. Since the OEG segments resided on the outer surface of\nthe collapsed chains with a higher surface free energy, hence, the\nwater droplets deposited on the PO10-g-L.SEC surface would spread\nwith a low CA of 2°. These results confirmed the intrinsic hydrophilic\ncharacteristic of the polymer chain conformation together with the\nhierarchical structure that controlled the surface wettability. To\nconclude, the thermally responsive switching between superhydrophobic\nand superhydrophilic states of PNm-g-L.SEC surfaces (m ∼ 10,\n20) was observed due to the reduced low surface free energy caused\nby the hydrophobic IP moieties orientation at air/solid/water interface.\nIn contrast, POm-g-L.SEC surfaces (m ∼ 10, 20) possessed a\nhigher surface free energy due to the hydrophilic PEG segments being\nexposed to the interface resulting in a non-switchable wettability\nphenomenon. To further confirm the mechanism of hydrophobic\nenhancement induced\nby the surface roughness and the interfacial water structure transformation,\nwe imaged the 3D contact interface between the liquid droplet and\nthe PN10-g-L.SEC surface at 20 and 40 °C via confocal laser microscopy.\nBelow the LCST, the stable and ordered hydrogen bond network promoted\nthe wetting of the microparticle surface by water molecules that also\noccupied the gap between the microparticles generating superhydrophilic\ndomains as indicated in Figure S15A . The\ngreen dots (fluorescence-stained) persisted from the base substrate\nto the outer surface, demonstrating the fully wetted state. Above\nthe LCST, the droplet contacts with the PN10-g-L.SEC surface revealed\nthat the liquid baseline was suspended between particles, indicating\na non-wetted state ( Figure S15B ). This\nobservation further demonstrates that the more disordered water structure\n(intermediate water) caused by the polymer chain transition induced\na weaker binding interaction, resulting in a lower tendency to wet\nthe surface. Thus, the liquid would not penetrate the air-pockets\nto fill the surface resulting in the observed superhydrophobic character\nof the substrate. The molecular structure and chemical composition\nof PNIPAM and\nPOEGMA 188 impacted the interaction between water molecules\nand polymer brushes, which controlled the hydrophobicity transition\ncharacteristics of the surface. Preferential exposure of the hydrophobic\nor hydrophilic moieties of the polymer-L.SEC altered the interfacial\ncharacteristics of the surrounding solvents (water or oil) and the\nsurface, which could be used to manipulate the macroscopic wettability.\nThus, these pollen-based thermo-responsive surfaces offer a novel\ndesign strategy to control the surface wettability transformation\nand exploit for various on-demand applications, such as emulsion separation.\nThe switchable oil and water repellency driven by temperature can\nbe conducted by alternately prewetting with water and oil, which gives\nthe separation membrane the versatility to handle oil–water\nmixtures ( Figure S16 )."
} | 11,249 |
35654796 | PMC9163127 | pmc | 471 | {
"abstract": "Biophotovoltaics (BPV) generates electricity from reducing equivalent(s) produced by photosynthetic organisms by exploiting a phenomenon called extracellular electron transfer (EET), where reducing equivalent(s) is transferred to external electron acceptors. Although cyanobacteria have been extensively studied for BPV because of their high photosynthetic activity and ease of handling, their low EET activity poses a limitation. Here, we show an order-of-magnitude enhancement in photocurrent generation of the cyanobacterium Synechocystis sp. PCC 6803 by deprivation of the outer membrane, where electrons are suggested to stem from pathway(s) downstream of photosystem I. A marked enhancement of EET activity itself is verified by rapid reduction of exogenous electron acceptor, ferricyanide. The extracellular organic substances, including reducing equivalent(s), produced by this cyanobacterium serve as respiratory substrates for other heterotrophic bacteria. These findings demonstrate that the outer membrane is a barrier that limits EET. Therefore, depriving this membrane is an effective approach to exploit the cyanobacterial reducing equivalent(s).",
"introduction": "Introduction Recently, various technologies have been developed to utilize reducing equivalents produced by oxygenic photosynthesis, which absorbs light and oxidizes water to produce high energy electrons. Emerging among them is biophotovoltaics (BPV), in which the reducing equivalents are transferred extracellularly to electrodes or to other bacteria, a phenomenon called extracellular electron transfer (EET), eventually generating electrical power 1 – 4 . Cyanobacteria, which are gram-negative bacteria capable of performing oxygenic photosynthesis, have been extensively studied in the field of BPV research 1 – 3 . This is because cyanobacteria have higher energy conversion efficiency of photosynthesis than terrestrial plants, much like eukaryotic algae 5 – 7 . Moreover, they are easy to culture, amenable to genetic manipulation 8 , grow fast, and possess a simple cell structure compared with eukaryotes. Many studies on the mechanism of EET have been conducted using cyanobacteria 9 – 15 , and so far, direct EET via conductive nanowires 16 and indirect, mediated EET by endogenous mediators 14 , 15 , 17 have been put forward as possible EET pathways; some suggest that the latter is more likely than the former to occur in the case of cyanobacteria 2 , 18 . Although much progress has been made in the field of BPV in the past decade, the low EET activity of cyanobacteria remains a limitation. The EET activity of cyanobacteria, both in the dark and under illumination, is markedly lower compared with that of mineral-reducing, electricity-generating bacteria, e.g. the genera Shewanella and Geobacter 2 , which are capable of utilizing diverse electron acceptors including an anode 19 . As previously pointed out, the main mechanism for the low EET activity of cyanobacteria is their autotrophy, in which EET could be totally useless and wasteful because electrons originating from phosynthesis should be fully utilized to provide enough reducing equivalents and energy to fix carbon 1 . Here, we hypothesize an additional mechanism for the low EET activity of cyanobacteria: the low permeability of the outer membrane. The outer membrane of cyanobacteria exhibits more than 20-fold lower permeability to organic substrates than that of Escherichia coli , the model gram-negative bacteria 20 . This low permeability is thought to reflect its autotrophic life style 20 , 21 , which does not necessarily rely on uptake of extracellular nutrients, although various transport systems do exist and function in cyanobacteria 22 – 24 . Here, we show that, using an outer membrane-deprived Synechocystis sp. PCC 6803 (hereafter Synechocystis ) mutant, slr0688i, in which the interaction between the outer membrane and the peptidoglycan is weakened so that the outer membrane is detached from the cell, a significant enhancement in cyanobacterial EET activity is achievable. EET activity is evaluated as extensively as possible in terms of photocurrent generation, ferricyanide reduction, and electron donation capacity as respiratory substrates (Fig. 1a ). This study verifies our hypothesis that the low permeability of the outer membrane contributes to the low EET activity of cyanobacteria. Fig. 1 EET and photocurrent generation from Synechocystis cells. a Schematic summary of EET from outer membrane-deprived Synechocystis cells, slr0688i. From slr0688i, reducing equivalents could be transferred to an exterior electrode via secreted compounds (mediated EET) or to an artificial electron acceptor, ferricyanide (ferricyanide-mediated EET). In addition, the extracellularly derived reducing equivalents could serve as respiratory substrates for other bacteria (here, Bacillus cereus ), i.e., electron donors in BPV systems. b Slr0688i (OD 730 = 1.5, 4 mL; red points) and dCas9 (OD 730 = 1.5, 4 mL; black points) were injected by gravity onto plane ITO electrodes, and +0.25 V vs Ag/AgCl was applied to obtain chronoamperograms. Averages ± 2 SE from 33 and 21 biological replicates for slr0688i and dCas9, respectively, are presented. c Representative chronoamperograms of slr0688i (OD 730 = 1.5, 4 mL; red line) and dCas9 (OD 730 = 1.5, 4 mL; black line). d The supernatant of slr0688i was substituted either with that of dCas9 (pink line) or with fresh BG11 medium (green line). Also shown as controls are slr0688i and dCas9 suspended in their respective supernatants (red and black lines, respectively), and dCas9 in slr0688i supernatant (blue line). All cell suspensions were adjusted to OD 730 = 1.5 (4 mL) and injected onto flat ITO, and +0.25 V vs Ag/AgCl was applied. e Slr0688i (OD 730 = 14.6, 4 mL) after 6 days of culture were injected by gravity onto a piece of carbon paper placed upon flat ITO, and +0.25 V vs Ag/AgCl was applied. The inset shows the enlarged view from 0 to 1200 s of the measurement. The upward and downward arrowheads indicate the beginning and end of illumination ( b – d , 100 µmol photons m −2 s −1 ; e, 420 µmol photons m −2 s −1 ), respectively. Source data are provided as a Source Data file.",
"discussion": "Discussion In this study, using an outer membrane-deprived Synechocystis mutant, slr0688i 25 , we achieved an order-of-magnitude enhancement in photocurrent generation and a significantly higher rate of exogenous ferricyanide reduction. These results verified our hypothesis that the low permeability of the outer membrane contributes to low cyanobacterial EET activity. This study demonstrated that an order-of-magnitude increase in current generation from Synechocystis cells is achievable by genetic engineering 25 . Moreover, based on the data shown in Fig. 3 and the chlorophyll content of slr0688i (5.6 µg Chl (mL OD 730 ) −1 ), the rate of EET from slr0688i to ferricyanide is calculated to be 14.6 nmol electrons nmol Chl −1 h −1 under illumination; this is higher than the fastest rate reported with terminal oxidase mutants, i.e., 2.7 nmol electrons nmol Chl −1 h −1 in the Cyd/ARTO double mutant, which we calculated using the data reported in Table S2 in Bradley et al. 12 . The focus of this study was to reveal and characterize the effects of outer membrane deprivation on EET activity of Synechocystis cells with regard to current generation, ferricyanide reduction, and electron donation capacity to other bacteria. Therefore, we simply measured the photocurrent generation by the cell suspension that was placed by gravity on the electrode, and did not optimize the electrode or cell–electrode interaction. This study clearly showed that outer membrane deprivation is a promising technology to improve the efficiency of BPV systems. The photocurrent is expected to be further enhanced by combining various other approaches: optimization 54 and modification of the electrodes 14 , 33 , 34 , strengthening the cell–electrode interaction 13 , 14 , 30 , 32 , 55 , optimizing the cell culture conditions 56 , and genetic engineering 12 , 31 in addition to outer membrane deprivation. Indeed, as demonstrated in Fig. 1e , photocurrent generated from slr0688i could be increased more than 100-fold by using larger amount of the cells and placing carbon paper onto ITO as an additional anode; the observed photocurrent is the highest among those reported for mediator-less setups utilizing untreated whole cells of Synechocystis 12 – 14 , 31 – 34 . When using a carbon paper anode, we noticed that the current level tended to remain high after switching off the light (Fig. 1e ), and it took approx. 1 h to decline back to the basal level. We speculate that this is probably because pores of carbon paper (about 100 µm) 57 prevent mediators reduced by cells from diffusive dispersion into the bulk electrolyte and/or trap cells so densely that sufficient CO 2 is not available, leading to the overaccumulation of intracellular reducing equivalents which could possibly reduce mediators even in the dark. This study provides the useful means for EET enhancement by structural alteration of a genetically manipulated cell surface. The outer membrane of the slr0688i mutant is detached from the peptidoglycan layer by suppression of pyruvylation of peptidoglycan-linked polysaccharides (Supplementary Fig. 1a ), which weakens the interaction between peptidoglycans and the S-layer homologous (SLH) domain of outer membrane proteins 25 . This leads to liberation of periplasmic and thylakoid lumenal components 25 , which were proven in this study to have an electron donation capacity equal to 8 µM glucose (Fig. 4b ). Although slr0688i exhibits growth retardation compared to the wild type (Supplementary Fig. 1b ), photosynthetic and respiratory activity were comparable to those of the wild type 25 as judged by the oxygen-evolving activity (Supplementary Fig. 2 ). The importance of cyanobacterial cell membranes on the photocurrent generation was previously demonstrated by Saper et al., where Synechocystis cells treated either through a low-pressure microfluidizer at 10–15 psi under the presence of 400 mM NaCl or by a simple osmotic shock using 400 mM NaCl produced markedly enhanced photocurrents compared to the untreated cells 15 . These observations show that a gentle physical treatment, which presumably causes a modest damage(s) to the cell membranes, leads to a higher electrogenic activity and thus their study provided a notable approach to improve the bio-photoelectrochemical system. However, although both the ‘physically-treated’ cells and the outer membrane-deprived cells presented by our study show the enhanced electrogenic activity, their underlying mechanisms are considered different from each other for the following reasons. First, aforementioned physical treatments might partially perturb the structural integrity of the cell surface but is not expected to deprive the outer membrane from the cells because the cyanobacterial outer membrane is well known to be tightly anchored to the peptidoglycan even after the mechanical disruption of the cells, such as passing through a French pressure cell press at 14,000 psi 20 or mechanical cracking 58 , 59 . Detaching the outer membrane thus requires the enzymatic degradation of the peptidoglycan and/or solubilization of the outer membrane by detergent treatment 20 , 58 – 60 . Second, suspending the cells in 400 mM NaCl unequivocally generates the osmotic pressure to the cytoplasmic membrane rather than to the outer membrane because the outer membrane contains abundant non-specific channels those allow the rapid passage of ions across the outer membrane 20 . Indeed, Reed et al. 61 showed that treating the cells with 490 mM NaCl affected the cytoplasmic membrane and resulted in the leakage of various cytosolic metabolites, which may include the possible mediator(s) for EET. Third, genetically engineered outer membrane deprivation avoids interfering the cytoplasmic membrane function and cytosolic metabolites release, so that even abundant cytosolic molecules, such as NADP(H), was not detected in the culture supernatant of the outer membrane-deprived cells 25 (Supplementary Fig. 21 , which is discussed in detail later in Discussion section). Consequently, it appears reasonable to assume that the electron transfer path from the physically-treated cells and that from the outer membrane-deprived cells, to the electrodes, are different; as a matter of fact, this assumption was verified by a striking difference in their photocurrent properties that DCMU enhances the photocurrent of the former cells 15 but abolishes the photocurrent from the latter cells (Fig. 2a ). Photocurrent measurements under the presence of a number of photosynthesis inhibitors demonstrated that the electrons stem from the pathway(s) downstream of PSI (Fig. 2 , Supplementary Figs. 9 – 13 ). Among the effects of the inhibitors, the difference in effects of GA and pCMB appeared to be important; both compounds are suggested to inhibit the Calvin cycle (Supplementary Figs. 10 – 13 ) 39 , 42 , but the former inhibitor abolished the photocurrent, whereas the latter markedly enhanced it (Fig. 2a ). One obvious difference in effects between GA and pCMB was that only GA-treated cells exhibited a high P700 oxidation level under FR light illumination (Supplementary Fig. 12 ), suggesting that GA inhibits respiration- and/or CET-related pathway(s). It is not unreasonable to assume the inhibitors affect pathways other than the Calvin cycle because GA is known to be inherently pleiotropic owing to its electrophilicity; indeed, GA is reported to inhibit yeast growth and fermentation 62 . An additional difference between GA and pCMB was evident from NADPH fluorescence measurement (Supplementary Fig. 13 ); the pCMB-treated cells showed a slowed decay of NADPH fluorescence after switching off the light and NADPH formation in the subsequent dark period was never observed. The slowed NADPH fluorescence decay may be attributable to Calvin cycle inhibition, but the lack of subsequent NADPH formation in the dark suggests the existence of other inhibitory effects of pCMB on the metabolic pathway(s) that reduces NADP, such as the oxidative pentose phosphate pathway 63 , 64 . We speculate that inhibition of dark NADPH formation changes cellular metabolism to accumulate alternative reducing equivalents instead of NADPH and this may be linked to the enhanced EET activity. Supporting this speculation, the pCMB-treated cells showed greatly enhanced EET current even in the dark (see the amperometric i-t curve in Fig. 2a ). Taken together, the electrons involved in photocurrent generation seem to be derived from intermediates of metabolic pathway(s) that is linked to PQ reduction but not to NADP reduction. Elucidation of the action of GA and pCMB will provide further insights into the EET mechanism. The notion that the photocurrent stems from the pathway(s) downstream of PSI is largely consistent with previous studies 2 , 11 , 12 , 17 . In addition, the possible involvement of the pathway(s) connected to PQ reduction agreed with the previous observation that a mutant deficient in the functional NDH complex, which supplies electrons to PQ, exhibits significant enhancement of ferricyanide reduction 12 . However, contradictory observations hinder a straightforward understanding of the EET mechanism of this cyanobacterium. (I) Strikingly, some studies reported that DCMU enhances photocurrent generation 15 , 17 , whereas others, including this study, reported the opposite effect 11 , 13 , 14 . (II) Recently, Shlosberg et al. reported that NADPH serves as the electron mediator in this bacterium 17 . However, this mechanism is not necessarily applicable to this study because of the following two reasons. (I) We confirmed that the supernatant of slr0688i cells, regardless of before and after the electrochemical measurement, contained no detectable level of NADPH that was analyzed by means of the two-dimensional fluorescence mapping (2D-FM), enzymatic analysis, and LC-MS/MS analysis (Supplementary Fig. 21 ). The 2D-FL revealed the presence of strong fluorescent molecules and one of which exhibited the fluorescence pattern of excitation wavelength (Ex) of 360 nm and emission wavelength (Em) of 450 nm, which was similar to those of NADPH (Ex 340 nm/ Em 460 nm). However, the small discrepancy of Ex/Em led us to further analyze whether this fluorescence was actually derived from NADPH; neither an enzymatic assay whose detection limit was 0.0625 µM nor a LC-MS/MS assay using the 50-fold concentrated supernatant detected NADPH. On the basis of these results, we concluded that the fluorescent molecules found in the supernatant was not NADPH. The absence of NADPH in the supernatant is reasonable because the bacterial periplasm is well known to be an oxidizing environment, and bacteria possess various systems to convey reducing power to periplasmic space rather than directly exporting the cytosolic reducing equivalents such as NAD(P)H (one of the examples can be found in Depuydt et al. 65 ). Thus, just depriving the outer membrane is not considered to result in releasing NADPH into the external environment. (II) Furthermore, intracellular NADPH fluorescence intensity was not significantly changed even in pCMB-treated cells despite their showing greatly enhanced photocurrent generation (Fig. 2a , Supplementary Fig. 13a ). One possible explanation for these contradictory observations is the assumption that the EET pathway may depend on the cellular physiology and intracellular electron flux, which are sensitive to numerous growth condition parameters, such as light intensity or CO 2 availability. Thus, further investigation is necessary to elucidate the EET mechanism of this organism 66 – 68 . Deprivation of the outer membrane is undoubtedly beneficial not only for the improvement of BPV systems but also for enhancing production of various chemicals ranging from biofuels to high-value compounds 69 – 73 . Supporting this is a previous study showing that slr0688i secretes enough nutrients to support the growth of heterotrophs 25 ; moreover, this study also showed that the supernatant of slr0688i provides respiratory substrates to Bacillus cells as efficiently as 8 µM glucose. Many beneficial aspects of the outer membrane-deprived mutant remain to be explored, and full utilization of its photosynthetic reaction is a promising way to achieve a clean and sustainable future."
} | 4,647 |
30368244 | PMC6204281 | pmc | 473 | {
"abstract": "Background The expansion of renewable energy produced by windmills and photovoltaic panels has generated a considerable electricity surplus, which can be utilized in water electrolysis systems for hydrogen production. The resulting hydrogen can then be funneled to anaerobic digesters for biogas upgrading (biomethanation) purposes (power-to-methane) or to produce high value-added compounds such as short-chain fatty acids (power-to-chemicals). Genome-centric metagenomics and metatranscriptomic analyses were performed to better understand the metabolic dynamics associated with H 2 injection in two different configurations of anaerobic digesters treating acidic wastes, specifically cheese manufacturing byproducts. These approaches revealed the key-genes involved in methanation and carbon fixation pathways at species level. Results The biogas upgrading process in the single-stage configuration increased the CH 4 content by 7%. The dominant methanogenic species responsible for the upregulation of the hydrogenotrophic pathway in this reactor was Methanothermobacter wolfeii UC0008. In the two-stage configuration, H 2 injection induced an upregulation of CO 2 fixation pathways producing short-chain fatty acids, mainly acetate and butyrate. In this configuration, the abundant species Anaerobaculum hydrogeniformans UC0046 and Defluviitoga tunisiensis UC0050 primarily upregulated genes related to electron transport chains, suggesting putative syntrophisms with hydrogen scavenger microbes. Interestingly, Tepidanaerobacter acetatoxydans UC0018 did not act as an acetate-oxidizer in either reactor configurations, and instead regulated pathways involved in acetate production and uptake. A putative syntrophic association between Coprothermobacter proteolyticus UC0011 and M . wolfeii UC0008 was proposed in the two-stage reactor. In order to support the transcriptomic findings regarding the hydrogen utilization routes, an advanced bioconversion model was adapted for the simulation of the single- and two-stage reactor setups. Conclusions This is the first study investigating biogas reactor metatranscriptome dynamics following hydrogen injection for biomethanation and carbon fixation to short-chain fatty acids purposes. The same microbes showed different patterns of metabolic regulation in the two reactor configurations. It was observed an effect of the specialized acidogenic reactor on the overall microbial consortium composition and activity in the two-stage digester. There were also suggested the main species responsible for methanation, short-chain fatty acids production, and electron transport chain mechanisms, in both reactor configurations. Electronic supplementary material The online version of this article (10.1186/s40168-018-0583-4) contains supplementary material, which is available to authorized users.",
"conclusion": "Conclusions H 2 injection induced different transcriptional regulation responses in the same MAGs inhabiting the two reactors. Specifically, they favored methanation in the single-stage reactor (power-to-methane), and SCFAs production in the two-stage configuration (power-to-chemicals). The above finding was also confirmed by model simulations. Gene expression results revealed that C . proteolyticus UC0011 and A . hydrogeniformans UC0046 mainly upregulated pathways involved in acetate and butyrate production. However, a 7% increase in CH 4 content in the biogas of the single-stage reactor was observed, mainly due to the dominant hydrogenotrophic M . wolfeii UC0008. In contrast, a doubling of total SCFAs by CO 2 fixation was evidenced in the two-stage configuration, with A . hydrogeniformans UC0046 and D . tunisiensis UC0050 upregulating genes involved in electron transport chains. Interestingly, the SAOB T . acetatoxydans UC0018 did not act as acetate-oxidizer in either reactor configuration, but primarily inhibited sugar metabolism in the single stage and boosted acetate uptake via the reductive TCA cycle in the two-stage configuration. A putative syntrophism between C . proteolyticus UC0011 and M . wolfeii UC0008 was proposed in the serial reactor configuration.",
"discussion": "Results and discussion Metagenomic and metatranscriptomic investigations were performed at two time points; the first point referred to the reactors’ steady-state performance before H 2 injection (phase I) and the second occurred 1 week after H 2 injection (phase II). To verify the stability of the microbial community during the reactors’ stable operation, an additional set of metagenomic samples was collected from R1 and R3 at multiple time points and was analyzed using 16S rRNA gene amplicon sequencing. The overview of the sequencing depth obtained with the different NGS data type showed that the microbial community under consideration was well captured (Additional file 1 : Table S1). OTUs taxonomy showed that the biological process was adequately captured in terms of microbial composition; results from beta diversity demonstrated that the microbial community was stable during time, and thus, the selected point chosen for in depth analysis was representative of the steady-state period (Additional file 1 : Table S2 and Figure S1). In particular, the overall OTUs’ taxonomic distribution in both R1 and R3 was in agreement with the profile of the reconstructed MAGs (Additional file 1 : Table S2). PCoA results and OTUs relative abundances (i.e., 1.5 average fold change) revealed negligible variations among the different time points in R1 (Additional file 1 : Table S2 and Figure S2). Regarding the reactor R3, the dominant OTUs abundances were coherent with MAGs coverages (Fig. 2a and Additional file 1 : Table S2). The observed differences in the PCoA results were mainly attributed to the microbial diversity of a minor subset of OTUs in the middle sampling point. Considering the results from the biochemical parameters, the reactor operation was stable, indicating that this OTUs subset was not primarily involved in the methanation process. Fig. 2 CH 4 yield of the two configurations ( a ), pH and VFAs trends in R1 ( b ), R2 ( c ) and R3 ( d ), before (phase I) and after (phase II) H 2 injection. The orange and green arrows highlight the DNA/RNA sampling points for the single and the two-stage configuration, respectively The reconstructed MAGs identified in the microbial community represented more than 60% of the entire microbiome. Therefore, the results from the current work covered successfully the majority of the transcriptional changes occurring in the reactors excluding only a minor fraction of the information present in the shotgun reads. Moreover, the total number of protein-encoding genes identified in the assembly was slightly higher than 196,000, out of which 80% had at least 1 read in 1 of the samples examined and 27% had 10 or more reads. Consequently, the outcomes of the identified genes confirm that the transcriptional study was representing the expression of a considerable fraction of the total genes in the microbiome. In order to acquire a global overview, analysis of the total expressed genes (not assigned to MAGs) was carried out considering COG classification (Additional file 1 : Figure S3, Additional file 2 : Dataset S3). In both reactor configurations, the most differentially expressed categories (excluding R and S categories, representing the general and unknown functions, respectively) belonged to the carbohydrate and amino acid transports and metabolisms. However, a high fraction of genes within the C category (energy production and conversion) was also differentially expressed in both single and two-stage reactors. Analysis of the expressed genes was subsequently performed in a genome-centric perspective to decipher the roles of the individual MAGs. The investigation was focused on the most abundant and active species, having more than 1000 expressed genes after H 2 injection. However, the analysis was exceptionally expanded to two MAGs ( Methanothermobacter wolfeii UC0008 and Tepidanaerobacter acetatoxydans UC0018) that were considered of particular interest, despite the fact that they showed less than 1000 expressed genes. Single-stage reactor: power-to-methane The single-stage reactor (R1) exhibited a pH trend ranging between 6.3 and 7.3 during phase I (Fig. 2b ). Total VFAs were highly concentrated (9.7 ± 1.1 g/L) and composed mainly of acetate (6.1 ± 1.0 g/L) (Fig. 2b ). These conditions inhibited the activity of methanogenic archaea, resulting in a CH 4 yield equal to 31% of the theoretical value, which is 350 mL CH 4 /g COD (Fig. 2a and Table 1 ). High VFAs concentrations lower the pH of the reactor, and thus, concomitantly lead to alteration of the microbial activities [ 37 ]. This effect is especially evident during the anaerobic digestion of acidic substrates characterized by poor buffering capacity [ 10 ]. In particular, methanogens are the most sensitive species to over acidification events, since their optimal growth rate ranges between the pH values of 6.5 and 8.5 [ 38 ]. Total VFAs increased by ~ 1 g/L in phase II, mainly due to higher butyrate concentration (Fig. 2b ). This increase could be caused by the high acetate levels present in this reactor (6.7 ± 0.8 g/L), which may have hampered syntrophic butyrate oxidation [ 39 ]. Despite the further over acidification during phase II, the CH 4 yield in R1 increased by 10% compared to the previous experimental phase (Fig. 2a and Table 1 ), indicating a positive effect of H 2 injection on the methanogenic consortia. Table 1 Reactors’ performance at phase I (steady state, before H 2 injection) and phase II (1 week after H 2 injection) Reactor configuration Phase I (pre-H 2 ) Phase II (post-H 2 ) CH 4 yield (mL CH 4 /g COD added ) CH 4 (%) CO 2 (%) CH 4 yield (mL CH 4 /g COD added ) CH 4 (%) CO 2 (%) H 2 (%) H 2 consumption rate (mL/L day) CO 2 conversion rate (mL/L day) Single stage 110 ± 21 44.6 ± 0.1 55.4 ± 0.1 142 ± 16 51.6 ± 0.1 23.0 ± 0.1 25.4 ± 0.1 648 ± 24 182 ± 56 Two-stage 276 ± 34 57.3 ± 0.1 42.7 ± 0.1 152 ± 16 39.7 ± 0.1 30.0 ± 0.1 30.3 ± 0.1 303 ± 64 173 ± 35 Three MAGs were identified as dominant (77% of the microbial community) in R1 during phase I, specifically Coprothermobacter proteolyticus UC0011, Anaerobaculum hydrogeniformans UC0046, and Defluviitoga tunisiensis UC0050 (Fig. 3 , Additional file 1 : Figure S4, Additional file 2 : Dataset S1). This microbial core reached 85% of relative abundance after H 2 injection, with C . proteolyticus UC0011 as the dominant species (61% relative abundance) (Fig. 3 , Additional file 1 : Figure S4, Additional file 2 : Dataset S1). These results highlight the strong microbial selection operated by both the feed characteristics such as the acidic pH and low buffer capacity, and the increased H 2 partial pressure inside the reactor. A significative correlation between C . proteolyticus UC0011 in phase II and H 2 content in the reactor was also highlighted by statistical analysis (Additional file 1 : Figure S5). Transcriptional data showed that C . proteolyticus UC0011 responded to H 2 addition by differentially expressing genes related to carbon metabolism, specifically the pyruvate metabolic pathway (Table 2 and Fig. 4 ). Genes associated with the pyruvate dehydrogenase complex and pyruvate-formate lyase (Aco, Ace, and Pfl), both involved in acetyl-CoA production, increased their expression by ~ 3-fold in C . proteolyticus UC0011 (Fig. 5 and Additional file 2 : Dataset S2). This upregulation suggests that C . proteolyticus UC0011 is involved in the acetate accumulation observed in R1 (Fig. 4 and Fig. 5 ). In contrast, expression of the ATP-dependent protease Clp was inhibited by ~ 4-fold in C . proteolyticus UC0011 (Fig. 4 and Additional file 2 : Dataset S2), indicating a specific repression of the proteolytic activity of this enzyme, which causes H 2 release [ 40 , 41 ]. Fig. 3 Heat map of relative abundance of the 50 MAGs (R1 and R1H: single stage pre- and post-H 2 , respectively; R2 and R2H: acidogenic reactor of the two-stage pre- and post-H 2 , respectively; R3 and R3H: methanogenic reactor of the two-stage pre- and post-H 2 , respectively; a – c : replicates). Up and down arrows indicate the statistically significant shifts in abundance of the MAGs (increase and decrease, respectively) between the two conditions (pre-/post-H 2 ) Table 2 Number of differentially expressed (DE) genes per KEGG category of selected MAGs in R1 (single-stage configuration) and R3 (methanogenic reactor of the two-stage configuration) MAG M . thermophila UC0006 M . wolfeii UC0008 C . proteolyticus UC0011 T . acetatoxydans UC0018 A . hydrogeniformans UC0046 D . tunisiensis UC0050 Reactor KEGG category R3 R1 R3 R1 R3 R1 R3 R1 R1 R3 ABC transporters 3 5 4 2 0 7 10 4 0 11 Amino acids metabolism 19 5 0 0 5 5 35 8 0 21 Bacterial chemotaxis 0 0 0 0 0 3 8 1 0 5 Biosynthesis of amino acids 43 4 3 0 4 3 37 2 0 39 Biosynthesis of antibiotics 18 8 4 3 10 4 28 1 0 6 Biosynthesis of secondary metabolites 13 8 3 3 5 4 28 0 0 20 Butanoate metabolism 3 1 1 1 6 0 3 0 0 7 Carbon fixation pathways in prokaryotes 3 4 2 0 7 0 5 1 0 5 Carbon metabolism 18 21 12 4 11 4 19 1 0 11 Citrate cycle (TCA cycle) 1 4 1 2 9 0 6 0 0 5 Fatty acid metabolism 0 0 0 0 0 0 0 2 0 1 Flagellar assembly 0 0 0 0 0 0 7 0 0 0 Galactose metabolism 1 0 0 2 0 5 6 0 0 1 Glycolysis/Gluconeogenesis 1 1 1 2 7 4 11 0 0 5 Metal transport system 6 0 5 0 1 0 6 0 0 8 Methane metabolism 21 64 14 0 0 0 7 0 0 3 Nitrogen metabolism 5 0 0 0 1 1 3 1 0 1 Peptidoglycan biosynthesis 0 0 0 0 1 0 3 0 0 2 Propanoate metabolism 1 0 1 1 1 1 2 2 0 1 Purine metabolism 0 4 0 1 0 1 0 0 0 0 Pyrimidine metabolism 6 5 1 1 3 0 0 0 0 1 Pyruvate metabolism 0 3 0 3 0 1 0 0 0 0 Quorum sensing 4 0 0 1 2 0 13 0 3 7 Reductive acetyl-CoA pathway (Wood-Ljungdahl) 1 1 0 0 0 2 0 0 0 0 Ribosome 28 25 5 3 1 0 16 0 0 5 Sugar, amino acid and oligo-peptide transport system 0 0 0 0 1 0 25 0 0 1 Triacylglycerol biosynthesis 1 0 0 0 0 0 1 0 0 2 Two-component system 2 0 0 0 0 0 9 0 0 5 Fig. 4 Schematic representation describing the main degradation pathways of the substrates and the responsible MAGs. Green and red arrows indicate pathways enriched with up- and downregulated genes, respectively. Orange arrow indicates the connection between MAGs which mostly upregulated electron transport chain mechanisms, and hydrogenotrophic archaea. Orange dashed arrow specifically highlights the proposed syntrophic association between Coprothermobacter proteolyticus UC0011 and Methanothermobacter wolfeii UC0008. Metabolic representation of R2 was based on the change in abundance of the indicated MAGs, while for R1 and R3 it was based on gene expression data. R1 single-stage reactor, R2 acidogenic reactor of the two-stage configuration, R3 methanogenic reactor of the two-stage configuration, extH 2 external hydrogen, intH 2 internal hydrogen Fig. 5 Main upregulated genes after H 2 injection by the selected MAGs (indicated with colored dots and squares) in R1 (single-stage reactor) and R3 (methanogenic reactor of the two-stage configuration). rTCA reductive tricarboxylic acid cycle, ETF electron transfer flavoprotein Analysis of A . hydrogeniformans UC0046 revealed the differential expression of genes encoding ABC transporters related to amino acid translocation across the plasma membrane (Table 2 and Fig. 4 ). Expression of acetyl-CoA acetyltransferase (AtoB), acetyl-CoA:acetoacetyl-CoA transferase (AtoD), and butyrate kinase (Buk) increased ~ 3-fold, all involved in butyrate metabolism (Fig. 5 and Additional file 2 : Dataset S2). Indeed, Ato enzymes participate in the degradation of acetoacetate intermediate, which can be subsequently funneled to the central energy-gaining step, where crotonyl-CoA is converted to butyryl-CoA [ 42 ]. Moreover, the gene coding for Buk enzyme, which catalyzes the final step for butyrate formation, is frequently used as biomarker for the identification of butyrate-producing communities [ 43 ]. Therefore, these results indicate that A . hydrogeniformans UC0046 contributes to the increased butyrate concentration found in R1 after H 2 injection. Only 11 genes of D . tunisiensis UC0050 were differentially expressed after H 2 addition, and 3 were involved in quorum sensing activities (Table 2 ). RpoD (sigma 70) (Additional file 2 : Dataset S2) is the main bacterial sigma factor responsible for housekeeping gene transcription [ 44 ], and showed decreased expression. This regulation pattern suggests an inhibition of basal gene expression in this species during phase II. The known syntrophic acetate-oxidizing bacterium (SAOB) T . acetatoxydans UC0018 [ 41 ] slightly decreased in abundance after H 2 addition (Fig. 3 , Additional file 1 : Figure S4, Additional file 2 : Dataset S1). Transcriptomic data showed that T . acetatoxydans UC0018 differentially expressed genes encoding ABC transporters and enzymes involved in amino acid and sugar metabolism (Table 2 and Fig. 4 ). Sugar intake was decreased by the downregulation of specific ABC transporters (Mgl permease) as well as genes related to glucose and galactose metabolism ( nag sugar kinase, fruK , fba ) by 4- and 8-fold, respectively (Additional file 2 : Dataset S2). This regulation suggests the existence of a “feedback mechanism” to limit excessive acetate production via sugar catabolism. Regarding the methanogenic consortia, there was a clear dominance of one archaeal species, the hydrogenotrophic M . wolfeii UC0008, which was reduced in abundance by half after H 2 injection (Fig. 3 , Additional file 1 : Figure S4, Additional file 2 : Dataset S1). A significative reduction was also observed for the less abundant hydrogenotrophic Methanothermobacter thermautotrophicus UC0010 (Fig. 3 , Additional file 1 : Figure S4, Additional file 2 : Dataset S1). The growth inhibition of Archaea may be partly due to the low alkalinity intrinsic to cheese whey permeate, along with the increased acidification of the system following H 2 addition (Fig. 2 b) [ 38 ]. Despite this inhibition, H 2 injection induced a significant upregulation of the hydrogenotrophic methanogenesis pathway in M . wolfeii UC0008 (Additional file 1 : Figure S6). Such transcriptional behavior led to increased CH 4 content in the biogas, with a ~ 54% CO 2 conversion efficiency (Table 1 ). The high accumulation of acetate measured in this reactor could also be related to the lack of aceticlastic methanogens, whose growth was probably not favored by the conditions established in the single-stage configuration. In addition to the difficulty in maintaining the pH in a proper range for methanogenic growth, toxicity related to the accumulation of cations (i.e., potassium) or lipids has been previously hypothesized [ 10 ]. Two-stage configuration, acidogenic reactor: power-to-chemicals The hypothesis for applying the H 2 only into the acidogenic reactor (R2) of the two-stage configuration was that it could better withstand a potential pH increment that could be caused by the transformation of CO 2 into methane. This reactor indeed maintained a stable pH (~ 4) throughout the process. The main VFA produced in R2 was butyrate (3.9 ± 0.7 g/L), which increased by ~ 1 g/L after H 2 injection (Fig. 2c ). Bifidobacterium crudilactis UC0001 was the dominant species inhabiting this reactor, and it showed a change in abundance after H 2 addition, decreasing from 82 to 52% of the total microbiome (Fig. 3 , Additional file 1 : Figure S4, Additional file 2 : Dataset S1). In contrast, the heterofermentative lactic acid bacteria Leuconostoc pseudomesenteroides UC0016 strongly increased in abundance (~ 3-fold). This variation could be related to the higher butyrate concentration present in R2 during phase II (Fig. 2 c), since the lactose fermentation to lactate by L . pseudomesenteroides UC0016 can enhance the cross-feeding of Clostridiales species involved in the conversion of lactic acid to butyrate [ 45 ]. It was indeed observed a ~ 4-fold increase of Clostridiaceae sp. UC0025, Clostridiaceae sp. UC0028, and Clostridium sp. UC0030 during phase II (Fig. 3 , Additional file 1 : Figure S4, Additional file 2 : Dataset S1). Moreover, butyrate production by clostridial-type fermentation is also known to be favored under high H 2 partial pressures [ 46 – 50 ], as can occur during exogenous H 2 injection in R2. The significative effect of butyrate increase in phase II on microbial distribution was also evidenced by statistical analysis (Additional file 1 : Figure S5). Two-stage configuration, methanogenic reactor: power-to-chemicals The methanogenic reactor of the two-stage configuration (R3) maintained the pH between 6.7 and 7.5 during phase I, exhibiting lower accumulation of total VFAs (primarily acetate) than the single-stage configuration (3.4 ± 1.3 g/L) (Fig. 2d ). These operating conditions resulted in a CH 4 yield equal to 80% of the theoretical value (Fig. 2 a, d). However, H 2 addition induced an increase in total VFAs (primarily butyrate), which doubled in concentration to 6.1 ± 0.3 g/L (Fig. 2d ). The CH 4 yield was highly reduced under these conditions (Fig. 2a and Table 1 ), and the increased butyrate and acetate levels along with the decreased CH 4 content seen in phase II indicate that CO 2 fixation toward SCFAs overtook the methanation pathways. The most abundant MAGs were the same as those found in the single stage R1, specifically C . proteolyticus UC0011, A . hydrogeniformans UC0046, and D . tunisiensis UC0050, which accounted for 47% of the microbiome in the reactor (Fig. 3 , Additional file 1 : Figure S4, Additional file 2 : Dataset S1). This microbial core reached 81% of relative abundance after H 2 injection, and D . tunisiensis UC0050 was the dominant species (54% of the total community) (Fig. 3 , Additional file 1 : Figure S4, Additional file 2 : Dataset S1). Transcriptomic data indicated that D . tunisiensis UC0050 differentially expressed genes involved in carbon metabolism and fixation pathways for energy production (Table 2 and Fig. 4 ). It was observed a ~ 4-fold increase in NADH:ubiquinone oxidoreductase expression (the NuoE subunit, forming the NADH dehydrogenase module), which may function as electron acceptor for the also consistently highly expressed flavodoxin FldA (Fig. 5 and Additional file 2 : Dataset S2). The NADH:ubiquinone oxidoreductase enzyme is indeed a proton pump (also known as complex I), which couples electron transfer with the translocation of four protons through the membrane [ 51 ]. This electron flow can be mediated via a reduced flavodoxin, such as FldA, which acts as intermediate between central carbon metabolism (e.g., TCA cycle) and complex I [ 52 ]. Thus, the upregulation of these genes suggests an increased activity of the electron transfer chain via H 2 oxidation [ 51 ], and may be involved in syntrophic relationships with hydrogenotrophic species throughout the increased proton extrusion from the cell. Similarly to D . tunisiensis UC0050, C . proteolyticus UC0011 also differentially expressed genes related to carbon fixation pathways (Table 2 and Fig. 4 ). Specifically, C . proteolyticus UC0011 boosted the reductive tricarboxylic acid cycle (rTCA) by a ~ 6-fold increase in the expression of pyruvate:ferredoxin oxidoreductase (PFOR) and phosphoenolpyruvate carboxykinase (Pck) (Fig. 5 and Additional file 2 : Dataset S2). Such regulation indicates an uptake of the excess acetate for pyruvate production, from which other central metabolic intermediates can be formed. C . proteolyticus UC0011 also regulated genes involved in amino acids metabolism, including a ~ 4-fold upregulation of the ATP-dependent protease Clp, along with enzymes metabolizing various amino acids (arginine, alanine, glutamate, tryptophan, aspartate) (Fig. 4 , Additional file 2 : Dataset S2). Since H 2 is one of the main products derived from proteins and amino acids degradation by C . proteolyticus [ 40 , 41 ], it cannot be excluded that this microbial species can form syntrophic association with hydrogen-scavenger microorganisms, such as the hydrogenotrophic methanogen M . wolfeii UC0008. Previous studies pointed out a synergistic effect operated by the co-existence of proteolytic anaerobes and hydrogen-consuming methanogens, revealing an augmented cell growth and protein degradation efficiency [ 53 ]. A partnership between C . proteolyticus and archaeal species belonging to the Methanothermobacter genus has also been recently proposed [ 54 , 55 ]. A . hydrogeniformans UC0046 responded to H 2 injection by upregulating a H + -ATPase (NtpB) (Fig. 5 and Additional file 2 : Dataset S2) that extrudes protons through ATP hydrolysis, and by downregulating the expression of the Na + /proline symporter (PutP) and Na + /H + antiporters (MnhC, NhaC) (Additional file 2 : Dataset S2). These mechanisms are used by various anaerobic bacteria to regulate internal pH and to control the transmembrane electrochemical gradient [ 56 ]; however, a syntrophic mechanism also cannot be excluded for this species. Additionally, A . hydrogeniformans UC0046 upregulated the coenzyme F420-reducing hydrogenase (FrhA), the electron transfer flavoprotein (ETF: FixA), and the NADH:ubiquinone oxidoreductase (NuoE), all known to be involved in mechanisms of electron flow and energy production [ 51 , 52 ] (Fig. 5 and Additional file 2 : Dataset S2). As for D . tunisiensis UC0050, the upregulation of these genes suggests an involvement of A . hydrogeniformans UC0046 in syntrophic relationships with hydrogen-scavenger microorganisms. The less abundant SAOB T . acetatoxydans UC0018 increased by almost 8-fold after H 2 injection (Fig. 3 , Additional file 1 : Figure S4, Additional file 2 : Dataset S1). There was an upregulation of glucose metabolism (FruK, Fba), as well as sugar and branched-chain amino acid ABC transporters (Rbs, Mgl, and LivK) in this species (Fig. 4 , Fig. 5 , Additional file 2 : Dataset S2). T . acetatoxydans UC0018 also upregulated the rTCA key enzymes pyruvate:ferredoxin oxidoreductase and phosphoenolpyruvate carboxykinase (PorA and PckA) by 8- and 4-fold, respectively (Fig. 5 and Additional file 2 : Dataset S2), indicating an acetate uptake probably aimed to increase the energy store capacity. It is indeed known that the utilization of the TCA cycle in the reductive direction by many autotrophic anaerobes is aimed at producing metabolic intermediates via acetyl-CoA incorporation [ 57 ]. The rTCA upregulation seen in T . acetatoxydans UC0018 indicates the different metabolic direction taken by T . acetatoxydans , which did not act as a SAO by upregulating enzymes for acetate oxidation. A significative correlation between T . acetatoxydans UC0018 in phase II and acetate concentration in the reactor was also indicated by statistical analysis (Additional file 1 : Figure S5). The archaeal consortium was composed of three equally abundant methanogens: M . wolfeii UC0008, the generalist Methanosarcina thermophila UC0006 [ 58 ], and the methylotrophic Methanomassiliicoccus sp. UC0009 [ 59 ]. Only the latter decreased in abundance after H 2 injection (Fig. 3 , Additional file 1 : Figure S4, Additional file 2 : Dataset S1), showing a ~ 4-fold reduction and indicating that this species may be more sensitive to the new condition. The remarkable decrease of Methanomassiliicoccus sp. UC0009 and therefore the CH 4 produced by the methylotrophic pathway could be also one determinant of the lower methanation seen in R3 after H 2 addition. Additionally, although M . thermophila UC0006 and M . wolfeii UC0008 remained quantitatively stable and upregulated the aceticlastic ( M . thermophila UC0006) and hydrogenotrophic pathways in phase II (Additional file 1 : Figures S7 and S8), the drop in CH 4 content seen in R3 was unchanged (Fig. 2a and Table 1 ). However, it is worth to highlight that the presence of M . thermophila UC0006 in R3 may have allowed the lower accumulation of acetate compared to the single-stage configuration. Simulation of hydrogen utilization routes in R1 and R3 The same bacterial species had different regulatory responses in the two reactors. This diverse regulation could be due to the reactor configurations, resulting in different physicochemical conditions, and consequently different H2 utilization capability. To support the experimental findings based on gene expression results (Fig. 4 and Fig. 5 ), which showed the main metabolic pathways undertaken by the MAGs, a computational model of the two reactor configurations was also developed. Moreover, mass balance calculations contributed to clarify the processes occurring in the reactors after H 2 injection (Additional file 1 ). It was found that approximately 40% of the H 2 moles injected in the single-stage configuration (R1) per day were effectively utilized to produce CH 4 . Software simulation results for R1 showed trends similar to those obtained experimentally. In particular, both simulated methane production and total VFA concentration curves agreed with the measured values, although being slightly lower (Additional file 1 : Figures S9 and S10). However, the remaining 60% of the added H 2 was not enough to account for the butyrate increase seen in the same reactor (Additional file 1 ). Therefore, the additional utilization of internal H 2 produced by lactose fermentation to acetate and butyrate should be considered. The most reasonable hypothesis in terms of demand for indigenous H 2 moles suggests that propionate reduction increased butyrate (Additional file 1 ). Concerning the acidogenic reactor of the two-stage configuration (R2), the fractions of exogenous H 2 moles utilized for the butyrate augmentation were ~ 97%, 60%, and 30%, based on the substrate reduced (CO 2 , acetate, and propionate, respectively). Since the amount of propionate in R2 was negligible and butyrate increase cannot be primarily based on CO 2 reduction (considering the 30% residual hydrogen in the effluent gas from R3), the most probable explanation for butyrate production is via the acetate reduction. This was further confirmed by the slight decrease of acetate concentration in R2 (Fig. 2c ). Additionally, the acetate rise seen in the methanogenic reactor of the same configuration (R3) during phase II mostly relied on butyrate oxidation (~ 60%), also considering the augmented butyrate feeding from R2. Overall, model simulations of the methane production and changes in total VFA concentration were in agreement with the above assumptions (Additional file 1 : Figures S15 and S16). Finally, since only ~ 10% of the acetate rise in R3 could be explained via the acetogenic pathway, it is reasonable that the remaining 30% of the acetate mole increase was likely due to an accumulation effect, which may have inhibited the acetogenic pathway. Computer-aided simulations, combined with mass balance calculations, indicate the most probable H 2 availabilities in the two reactor configurations and therefore the different metabolic routes for H 2 utilization used by the anaerobic digestion microbiome."
} | 7,872 |
36132897 | PMC9417421 | pmc | 474 | {
"abstract": "Harvesting energy from the surrounding environment, particularly from human body motions, is an effective way to provide sustainable electricity for low-power mobile and portable electronics. To get adapted to the human body and its motions, we report a new fiber-based triboelectric nanogenerator (FTNG) with a coaxial double helix structure, which is appropriate for collecting mechanical energy in different forms. With a small displacement (10 mm at 1.8 Hz), this FTNG could output 850.20 mV voltage and 0.66 mA m −2 current density in the lateral sliding mode, or 2.15 V voltage and 1.42 mA m −2 current density in the vertical separating mode. Applications onto the human body are also demonstrated: the output of 6 V and 600 nA (3 V and 300 nA) could be achieved when the FTNG was attached to a cloth (wore on a wrist). The output of FTNG was maintained after washing or long-time working. This FTNG is highly adaptable to the human body and has the potential to be a promising mobile and portable power supply for wearable electronic devices.",
"conclusion": "Conclusions In summary, we developed a FTNG with a coaxial double helix structure that utilizes the general Nylon/Cu fiber and PTFE/Cu fiber for effectively harvesting the mechanical energy from the human body. Under a small displacement (10 mm, 1.8 Hz), this FTNG could output voltage of 850.20 mV and current density of 0.66 mA m −2 in the lateral sliding mode, and 2.15 V and 1.42 mA m −2 in the vertical separating mode. Even after a washing process or a 3 hours continuous working (∼10 000 cycles), no observable output performance's degradation was found. The large output of FTNG and its properties of being flexible, light-weighted, and robust structure make it suitable for continuous power harvesting. When the FTNG was worn on a human wrist, it delivered an output of 3 V voltage and 200 nA current via shaking hands. It can also harvest the energy in swing arms during walking and generates an output of 6 V voltage and 600 nA current when attached to the cloth. Furthermore, this FTNG is highly adaptable for harvesting the rotating mechanical energy. These features make this FTNG a promising mobile and portable power supply for wearable electronic devices.",
"introduction": "Introduction Nowadays, mobile and portable electronics have been widely applied in communication, monitoring personal health care, monitoring environmental safety, and so on. 1–9 For powering mobile and portable electronics, numerous batteries or supercapacitors have been utilized, but their lifetime is usually limited, resulting in some problems, such as the problem of frequent charging, which greatly hinders their applications. 10–12 An effective way to solve such issues is the harvesting and utilization of the widely existing energy from the surrounding environment. In our ambient environment, numerous energy sources could be harvested and utilized, such as solar energy, mechanical energy, thermal energy, and chemical energy. As a kind of mechanical energy, human body motion energy is closely related to human activities, which make it ubiquitously available in the applicable environment for mobile and portable devices. Thus, it is important to collect and convert the human body motion energy into electricity as a mobile energy source for mobile and portable electronics. Aimed at harvesting the widely existing mechanical energy, triboelectric nanogenerators (TENGs) were invented based on triboelectrification and electrostatic induction. 13–19 Using this technology, many groups have developed lots of wearable TENGs to harvest and convert human body motion energy into electricity, 20–35 which makes the body motion energy a feasible and available power source for mobile and portable electronics. Because of fiber's merits of being small, lightweight, bendable, and washable property, the fiber-based wearable TENGs have been widely studied. 36–38 Zhong et al. fabricated a fiber-based TENG to convert the biomechanical motions/vibration energy into electricity using one CNT-coated cotton fiber, and one PTFE and CNT-coated cotton fiber in an entangled structure. 39 Kim et al. fabricated a fabric-based TENG for powering wearable electronics by weaving fibers consisting of Al fibers and PDMS tubes. 40 Zhao et al. developed a wearable TENG by directly weaving Cu-coated PE fibers and polyimide-coated Cu-PET fibers in two vertical directions. 41 Dong et al. developed a 3-dimensional TENG for harvesting biomechanical energy using three types of fibers: blended fiber consisting of stainless steel fiber and polyester fiber, PDMS-coated energy-harvesting fiber, and binding fibers in three directions. 42 Wen et al. demonstrated a TENG built using a Cu-coated-EVA tube along with PDMS and Cu-coated EVA tube to collect random body motion energies. 43 After that, He et al. fabricated a fiber-based TENG with a silicone rubber fiber, in which the CNT layer and the copper fiber function as two electrodes. 44 Chen et al. reported a wearable TENG from commercial PTFE, carbon, and cotton fibers with the traditional shuttle weaving technology. 45 By choosing and processing materials together with designing a flexible structure, the fiber-based TENGs in all the above-mentioned studies can effectively scavenge mechanical energy in different forms. However, the complexity and fragility of the hybrid composites composed of cotton thread, carbon nanotube, PDMS, etc. could reduce their robustness and lead to disconnection or short-circuits, and further affect the output performance. Being an important factor in affecting the TENG's capacity in practical applications, it is attractive to implement a highly robust fiber-based TENG with cost-effective materials and flexible structure. Herein, we report a fiber-based triboelectric nanogenerator (FTNG) with high robustness that can effectively scavenge biomechanical energy from both weak physiological motions and vigorous behavior. In FTNG, the positive and negative triboelectric layers (Nylon and PTFE fibers) were first wrapped with their electrodes to form the two core–shell parts of FTNG, and then these two core–shell structure fibers were twined into a coaxial double helix structure FTNG. The FTNG can be worn on the human wrist or attached to the cloth to harvest the gentle energy of body motion. Besides, it could also be used to harvest the spinning energy from a rapid rotation.",
"discussion": "Results and discussion As shown in Fig. 1a , FTNG consists of one Nylon fiber-wrapped Cu fiber, and one PTFE fiber-wrapped Cu fiber, and they are twined into a double helix structure. Nylon and PTFE act as the frictional surfaces and the Cu fibers present inside work as the positive and negative electrodes, respectively. The fabrication process of FTNG is illustrated in Fig. S1a † and the experimental section in detail. The optical photo of the as-fabricated FTNG shown in Fig. 1b indicates the high flexibility of this structure. An enlarged view of the part in the red square is shown in Fig. 1c , in which the Nylon/Cu fiber and PTFE/Cu fiber were interwoven at regular intervals along the axial direction. The lateral image and the cross-sectional image of FTNG are shown in Fig. S1b and c, † respectively. FTNG weighs only 0.58 g, as shown in Fig. S1d, † which demonstrates that it is light enough to be used as the wearable power source with little discomfort. Since the tensile operation is a common motion in daily activities, a tensile-loading test was essential to examine the ability of FTNG to endure mechanical operations. As shown in Fig. 1d , the composite fiber exhibits a strength of 200 MPa with a tension strain of 150%, which is a little higher than that of the Nylon-wrapped Cu fiber (∼150 MPa) and PTFE-wrapped Cu fiber (∼95 MPa). Moreover, a weight of 500 g can be steadily hung on the composite fiber, as shown in the inset of Fig. 1d . It can be found that the double helix structure improves the tensile property, which is beneficial for the robustness of FTNG, thereby enabling its characteristics feature of harvesting mechanical energy from violent activities. Fig. 1 (a) The structure of the FTNG. (b) Digital photography of the FTNG. (c) An enlarged view of the part marked in red square of (b). (d) Stress–strain curves of the Nylon-wrapped Cu fiber, PTFE-wrapped Cu fiber, and the composite FTNG fiber. The inset is a digital photography of the FTNG hanging a 500 g weight. To test the output performance of FTNG, it was tensioned by fixing its two ends, and a sewing polyester thread of 0.20 mm in diameter as a contact object was rubbed against FTNG, as shown in Fig. S1e. † When sliding and contacting occurred, FTNG starts to work and generates electricity. Fig. 2a shows a full working cycle of the FTNG's operation in the sliding mode. When the polyester fiber moves towards the Nylon fiber, the surfaces of the two fibers come in contact and rub against each other. On the basis of the frictional series, the gaining electron's ability of polyester was relatively stronger than that of Nylon due to which electrons migrated from the Nylon surface to the polyester surface, resulting in polyester and Nylon with negative and positive charges, respectively. As displayed in Stage I, when the polyester fiber moves to the right, the contact surface between polyester and PTFE rubs against each other, and then the electrons migrate from polyester to PTFE, making the PTFE fiber a negatively charged surface. Meanwhile, the net reduced electric field drives electrons from the PTFE's electrode to the Nylon's electrode until the net electric field gets shielded by the induced charges moving from two electrodes. As shown in Stage II, when the polyester wire continues sliding to the right, the contact stage comes to the aligned position, where the positive and negative triboelectric charges were completely balanced. In the case of polyester wire sliding towards left, the contact position goes back to the misaligned condition, with the free electrons being driven from the Nylon's electrode to the PTFE's electrode, as presented in Stage III, leading to the backflow of induced free electrons. This process continues until the polyester wire keeps sliding towards the left in an aligned position (Stage IV). However, when the leftward sliding continues in a misaligned position, a reversed flow of induced electrons was observed (Stage V). Consequently, the power generation process of FTNG in one cycle was completed. By sliding the polyester fiber back and forth along the FTNG, charges got alternately transferred between the two electrodes of Nylon and PTFE. Under a 10 mm displacement at 1.8 Hz, FTNG generates an output voltage and output current of 850.20 mV and 19.52 nA, as shown in Fig. S1f and g, † respectively. An enlarged view of the output voltage peak in one cycle is shown in Fig. 2b , from which the four working stages could be observed clearly. As demonstrated in Fig. 2c , the current density could reach 0.66 mA m −2 , which is the ratio of the output current value and the sliding contact area. The actual contact area, here in the sliding process, was about 29.41 mm 2 , which is demonstrated in detail in Table 1 in ESI. † Fig. 2d is the integral curve of the output current curve from which the accumulative charge quantity reaches 16.73 nC, and the charging rate reaches 1.67 nC s −1 . Here, the charging rate is an average electric quantity in one second calculated from the integral value of the current curves. Fig. 2 (a) Working mechanism of the FTNG under a lateral sliding mode. (b) Enlarged output voltage of the FTNG a under frequency of 1.8 Hz with the displacement of 10 mm. (c) The current density of FTNG via dividing the output current in Fig. S1e † by the contact area. (d) The accumulative charge quantity via integrating the output current in Fig. S1e. † The moving speed and frictional area can largely affect the TENG's output performance. As shown in Fig. 3a and b , with a constant 10 mm sliding displacement, both the FTNG's output voltage and its output current increase with an increase in the frequency. With the increase in the sliding frequency, the sliding speed becomes larger, which shortens one working cycle time and further increases the working cycle number in a fixed time. Consequently, the peak value of the open-circuit voltage increased from 140.53 mV at 0.3 Hz to 688.27 mV at 1.5 Hz ( Fig. 3a ). As shown in Fig. 3b , the current peak value increases from 3.13 nA at 0.3 Hz to 14.43 nA at 1.5 Hz, which means that the peak current density increases from 0.11 mA m −2 at 0.3 Hz to 0.48 mA m −2 at 1.5 Hz. During the relative sliding process, the actual frictional area between the FTNG and polyester fiber will increase with the increase in the sliding displacement. To investigate this aspect, FTNG was made to work at different sliding displacements at a fixed sliding frequency of 1 Hz. When the sliding displacement varied between 3 mm and 7 mm, the voltage peak value increased from 70.56 mV to 441.02 mV ( Fig. 3c ). Moreover, the current peak value increased from 0.77 nA to 6.45 nA ( Fig. 3d ). When divided by the contact area of 29.41 mm 2 , the peak current density increased from 0.03 mA m −2 at 3 mm to 0.22 mA m −2 at 7 mm. It can be explained by stating that the larger contact area leads to increased displacement. This measurement exhibits the FTNG's ability to scavenge mechanical energy from low frequency and small amplitude motion widely existing in daily activities of humans. Fig. 3 (a) Output voltage and (b) output current of FTNG under different lateral sliding frequencies. (c) Output voltage and (d) output current of FTNG under different lateral sliding displacements. (e) Output current before and after washing operation. (f) Output current when continually working for 3 hours. As shown in Fig. 3e and ESI Video S1, † FTNG was immersed in water and stirred by a glass rod to test if FTNG can be washed like the clothes without a reduction in the performance. The comparison of the output currents before and after the washing process ( Fig. 3e ) shows no obvious reduction, which indicates that FTNG possesses good washing durability. Further, the robustness of FTNG was tested by continuously working for 3 h (∼10 000 times) at 1 Hz sliding frequency and 10 mm sliding displacement. As shown in Fig. 3f , after the charge accumulation process, the output current remains stable, which indicates the high robustness and long-term stability of FTNG. Aside from harvesting the sliding mechanical energy along the fiber's length direction in the above sliding mode, FTNG could also harvest the mechanical energy when it contacts and separates with other fabrics such as polyester fabric, as shown in Fig. S2a † in a contact-separating TENG working mode. As FTNG contacts and separates with the fabric, the electrons flow back-and-forth between the two electrodes in an external circuit, generating an alternating current output. Here, the polyester fabric, having fibers of 0.20 mm in diameter, was used to contact and separate with FTNG. At 1.8 Hz frequency and 20 mm displacement, the output voltage of 2.15 V and output current of 69.3 nA were achieved (Fig. S2b and c † ). The corresponding peak output current density was 1.42 mA m −2 . The actual contact area, here in the vertical separating process, was about 48.80 mm 2 , which is demonstrated in detail in Table 2 in the ESI. † The effect of the separation frequency on the FTNG's output performance was investigated with the separating distance being set at 10 mm. As shown in Fig. S2d and e, † when the frequency increases from 0.3 to 1.5 Hz, the output voltage and current increase from 436.78 mV, 4.97 nA to 1291.83 mV, 22.05 nA, respectively. Also, the corresponding current density, which can be obtained by dividing with the contact area of 48.80 mm 2 , increases from 0.10 mA m −2 at 0.3 Hz to 0.47 mA m −2 at 1.5 Hz. To test the ability of harvesting energy from the human activity, 3 FTNGs were woven into a bracelet in a parallel manner, and placed on one experimenter's wrist, as presented in Fig. 4a and ESI Video S2. † In this case, the friction occurs between FTNG and the wrist from shaking hands. Fig. 4b and c show the corresponding output voltage and output current generated by the experimenter's hand shaking motion. The output signals easily reach to 3 V and 200 nA (the enlarged view of output signals is shown in Fig. S3 † ). Then, 7 FTNGs were attached to the waist of one experimenter in parallel, as shown in Fig. 4d and ESI Video S3. † When the experimenter swings his arm alternately along the lateral direction and vertical direction, the friction occurs between FTNG and clothes. Fig. 4e and f show the output voltage and output current generated by the swing-arm movements, respectively. The output signals easily reach to 6 V and 600 nA, and every wave packet in them corresponds to one arm swing. The enlarged view of the output signals can be found in Fig. S4. † Therefore, this FTNG has the potential to act as a mobile and portable power supply for wearable devices. Fig. 4 (a) Demonstration of the application of FTNG to harvest the wrist motion energy. (b) Output voltage and (c) output current of FTNG driven by the wrist's motion. (d) Photograph showing the application of FTNG attached on the cloth to harvest the body motion energy in walking or jogging. (e) Output voltage and (f) output current of FTNG attached on the cloth. To further test the adaptability of FTNG for versatile mechanical energy harvesting, a whirligig is used to drive FTNG. The whirligig is a circular disc that spins when pulling on strings passes through its center (radius of the central disc was ∼35 mm). As shown in Fig. 5a , two FTNGs are inserted into the two center holes, respectively. The one cycle movement of the circular disc consists of two processes of forward and backward windings (ESI Video S4 † ). During the forward winding process, the input stretching force on FTNG (exerted by the human hands) accelerates the circular disc to its maximum rotating speed. Simultaneously, the two FTNGs begin to come in contact with each other and reach a tightly coiled state. Then, in the back winding process, no input force was on the FTNG due to which the disc rotates in the reverse direction. Consequently, the two FTNGs separate from each other and get into a parallel state. After this position, the inward force was applied again, and the two FTNGs begin to come in contact with each other again. This cycle of winding and unwinding of the FTNGs repeats itself, in which electricity is generated by the two FTNGs. Fig. 5b and c show the output voltage and current driven by the spinning of the circular disc. The corresponding voltage and current were 1.2 V and 40 nA, respectively. The enlarged view of one wave packet of the output voltage and current is shown in Fig. 5d and e , respectively, which corresponds to a winding or unwinding process. This demonstration implies that such a tough FTNG can also be extended to harvest the motion energy from the high-speed and vigorous movements. Fig. 5 (a) Schematic of the application of FTNG on harvesting the spinning energy. (b) Output voltage and (c) output current of FTNG driven by the spinning movement. The corresponding enlarged view for one wave packet of the output voltage (d) and output current (e)."
} | 4,867 |
23766295 | null | s2 | 477 | {
"abstract": "Cathodic reactions in biofilms employed in sediment microbial fuel cells is generally studied in the bulk phase. However, the cathodic biofilms affected by these reactions exist in microscale conditions in the biofilm and near the electrode surface that differ from the bulk phase. Understanding these microscale conditions and relating them to cathodic biofilm performance is critical for better-performing cathodes. The goal of this research was to quantify the variation in oxygen, hydrogen peroxide, and the pH value near polarized surfaces in river water to simulate cathodic biofilms. We used laboratory river-water biofilms and pure culture biofilms of Leptothrix discophora SP-6 as two types of cathodic biofilms. Microelectrodes were used to quantify oxygen concentration, hydrogen peroxide concentration, and the pH value near the cathodes. We observed the correlation between cathodic current generation, oxygen consumption, and hydrogen peroxide accumulation. We found that the 2 e(-) pathway for oxygen reduction is the dominant pathway as opposed to the previously accepted 4 e(-) pathway quantified from bulk-phase data. Biofouling of initially non-polarized cathodes by oxygen scavengers reduced cathode performance. Continuously polarized cathodes could sustain a higher cathodic current longer despite contamination. The surface pH reached a value of 8.8 when a current of only -30 μA was passed through a polarized cathode, demonstrating that the pH value could also contribute to preventing biofouling. Over time, oxygen-producing cathodic biofilms (Leptothrix discophora SP-6) colonized on polarized cathodes, which decreased the overpotential for oxygen reduction and resulted in a large cathodic current attributed to manganese reduction. However, the cathodic current was not sustainable."
} | 453 |
28799236 | null | s2 | 478 | {
"abstract": "Self-healing polymers crosslinked by solely reversible bonds are intrinsically weaker than common covalently crosslinked networks. Introducing covalent crosslinks into a reversible network would improve mechanical strength. It is challenging, however, to apply this concept to \"dry\" elastomers, largely because reversible crosslinks such as hydrogen bonds are often polar motifs, whereas covalent crosslinks are nonpolar motifs. These two types of bonds are intrinsically immiscible without cosolvents. Here, we design and fabricate a hybrid polymer network by crosslinking randomly branched polymers carrying motifs that can form both reversible hydrogen bonds and permanent covalent crosslinks. The randomly branched polymer links such two types of bonds and forces them to mix on the molecular level without cosolvents. This enables a hybrid \"dry\" elastomer that is very tough with fracture energy 13500 Jm"
} | 227 |
38205027 | PMC10776380 | pmc | 479 | {
"abstract": "Excessive mining and utilization fossil fuels has led to drastic environmental consequences, which will contribute to global warming and cause further climate change with severe consequences for the human population. The magnitude of these challenges requires several approaches to develop sustainable alternatives for chemicals and fuels production. In this context, biological processes, mainly microbial fermentation, have gained particular interest. For example, autotrophic gas-fermenting acetogenic bacteria are capable of converting CO, CO 2 and H 2 into biomass and multiple metabolites through Wood-Ljungdahl pathway, which can be exploited for large-scale fermentation processes to sustainably produce bulk biochemicals and biofuels (e.g. acetate and ethanol) from syngas. Clostridium autoethanogenum is one representative of these chemoautotrophic bacteria and considered as the model for the gas fermentation. Recently, the development of synthetic biology toolbox for this strain has enabled us to study and genetically improve their metabolic capability in gas fermentation. In this review, we will summarize the recent progress involved in the understanding of physiological mechanism and strain engineering for C . autoethanogenum , and provide our perspectives on the future development about the basic biology and engineering biology of this strain.",
"conclusion": "2 Conclusions and outlook Microbial gas fermentation is a promising technology to address drastic environmental consequences caused by modern industry. C. autoethanogenum represents one promising platform for sustainable production of bulk chemicals from syngas. To further understand the basic biology for this strain, the high-throughput functional genomics tools remains to be developed for investigating the gene interaction network to provide a new understanding of the relationship between genotype and phenotype. The uncovering of genome-wide gene regulatory networks is also a key tool allowing us to understand the complex mechanisms of transcriptional gene regulation and to better control the expressional behavior of native and foreign genes. For the engineering biology, metabolic engineering efforts in C. autoethanogenum are constrained by their energy requirements, and computational approaches will be of great advantage in the metabolic pathway design and enzyme engineering. Moreover, the recent development of synthetic biology tools, such as combinatorial libraries of DNA parts and regulatory circuits, also allows for optimizing the expression of synthetic pathway and balancing metabolic fluxes. In general, metabolic engineering in the C. autoethanogenum is still in its infancy. Alternatively, constructing the synthetic microbial communities in which sygnas-untilizing C. autoethanogenum produce the organic carbon resource for the other heterotrophic member, can take advantage of syngas fermentation and expand the spectrum of products.",
"introduction": "1 Introduction The development of modern industry is still mostly depending on the non-renewable fossil fuel resources for the production of biochemicals and biofuels, which caused many environmental concerns about global warming. When addressing these challenges, microbial fermentation process has shown its great promising as they allow efficient conversion of carbonaceous substrates into target products. The first generation of biofuel such as ethanol production by bacteria and yeasts or acetone, butanol, and ethanol fermentation by Clostridia are mainly relying on the fermentation of starch and sugar materials, which are making the development of economic processes extremely challenging [ 1 , 2 ]. One promising route to overcome these challenges is leveraging acetogenic bacteria to sustainably produce fuels and chemicals from single carbon (C1) gases CO and CO 2 in a process called gas fermentation. Optimization of this technical route followed by industrial promotion is expected to build a new and sustainable biomanufacturing model. Chemoautotrophic gas-fermenting bacterium, such as Clostridium autoethanogenum [ 3 ], Clostridium ljungdahlii [ 4 ], Clostridium carboxidivorans [ 5 ], Clostridium ragsdalei , Clostridium coskatii [ 6 ] and Acetobacterium woodii [ 7 ], possess the Wood-Ljungdahl pathway (WLP) of carbon fixation [ 2 ], which allows the conversion of C1-gases into the biomass precursor acetyl-CoA, acetate and other specific products, such as ethanol or butanol, while generating ATP for growth [ 8 , 9 ]. The WLP is known as the most efficient pathway among the six native biological carbon fixation pathways as it requires the least reaction steps and energy consumption [ 10 ]. C. autoethanogenum is one representative of chemoautotrophic bacteria belonging to Firmicutes, can fix gaseous carbons via WLP. Therefore, the wild-type C. autoethanogenum is able to utilize Cl gases derived from the gasification of domestic and agricultural wastes or different industrial off-gases and produce acetate and ethanol as main product [ 11 ]. And it is noteworthy, C. autoethanogenum has been applied in carbon-negative production of bulk chemicals by gas fermentation at industrial scale [ 12 ]. The genetic manipulation toolkit has already been available for the C . autoethanogenum [ 13 ]. Clustered Regularly Interspaced Short Palindromic Repeats (CRISPR)-Cas tools have also been developed to genetically modify this organism [ 14 , 15 ]. These tools not only help us understand the molecular mechanism for C1 fixation and energy metabolism involved, but also allow us to improve strain performance through metabolic engineering as well as to diversify and enhance their metabolic capabilities. In this review, we will focus on the recent development of synthetic biology toolbox, biogenetics and metabolic engineering for the C . autoethanogenum . 1.1 Carbon fixation and native product synthesis in C. autoethanogenum C. autoethanogenum is a strictly anaerobic, Gram-positive, spore forming, rod-like, motile bacterium. It was first isolated from rabbit faeces in 1994 under an atmosphere of carbon monoxide, nitrogen and carbon dioxide, with carbon monoxide as the sole energy source and was identified as a facultative chemolithotroph [ 16 , 17 ]. The whole genome of C. autoethanogenum DSM10061 has been published, according to this sequence, the bacterium has a chromosome length of 4,352,205 base pairs with a G + C content of 31.1 %, with 4161 predicted genes, 4042 of which are potentially protein-coding genes with 18 pseudogenes present, and 18 RNA genes [ 18 ]. Compared with its close species C . ljungdahlii, another model acetogenic bacteria, C. autoethanogenum displayed important phenotypic, genomic and metabolic differences although the two are indistinguishable at the 16S rRNA gene level [ 16 , [18] , [19] , [20] , [21] , [22] ]. For example, the two strains behave very differently at the transcriptional level [ 23 ], and C. autoethanogenum produced ethanol and 2,3-butanediol (2,3-BDO) production more efficiently [ 19 , 20 ]. Similar to other syngas-utilizing acetogenic bacteria, C. autoethanogenum is able to convert CO 2 and CO into low-carbon fuels and chemicals through WLP. Many excellent review articles have described the biochemistry related to this pathway in great detail [ 24 , 25 ]. The WLP consists of a methyl and carbonyl branch ( Fig. 1 ). In the methyl branch, CO or CO 2 is reduced to methyl group through a sequence of tetrahydrofolate- and cobalamin-dependent reactions, which will be then combined with CO (used either directly or after enzymatic reduction of CO 2 ) to form acetyl coenzyme A (acetyl-CoA). In the WLP, carbon monoxide dehydrogenase (CODH) catalyzes the reversible oxidation of CO to CO 2 , and acetyl-CoA synthase (ACS) combines with CODH to form a CODH/ACS complex for acetyl-CoA fixation. The overall structure of this complex adopts the classic α2β2 architecture, the α- and β-subunits catalyzed the formation of acetyl-CoA and the reduction of CO 2 to CO, respectively [ 26 ]. Both of these two enzymes are predicted to be essential to the function of WLP. Whole genome sequencing revealed the presence of three putative genes encoding CODHs: acsA (CAETHG_1620-162), cooS1 (CAETHG_3005) and cooS2 (CAETHG_3899). Mutagenesis study shows that, of the three genes, acsA is essential for autotrophy, by contrast, cooS1 and cooS2 are dispensable for autotrophy [ 13 ]. In fact, the WL pathway is energy-consuming and requires reducing equivalent supply which comes from CO or hydrogen under autotrophic growth conditions. Especially, the oxidation of hydrogen requires the participation of hydrogenases, and the Fdh/Hydrogenase complex directly utilizes the reducing force from hydrogen to reduce CO 2 . Alternatively, Hydrogenase catalyzes H 2 oxidation to produce Fd 2− and NADPH by electron bifurcation, storing the reducing power in cofactors [ 27 ]. Besides, many genes for gluconeogenesis also plays important roles in the carbon fixation during syngas fermentation. For example, transcriptomics revealed that the largest transcriptional fold change in the glycolytic enzymes observed was for one glyceraldehyde-3-phosphate dehydrogenase (GAPDH, CAETHG_3424) during autotrophic growth. The CAETHG_3424 deficient strain could not grow autotrophically on gas as the sole carbon and energy source. It indicated that CAETHG_3424 is one critical gene for the syngas utilization [ 23 ]. Additionally, the phosphoenolpyruvate carboxykinase (PCK) is responsible for the conversion of oxaloacetate to phosphoenolpyruvate using a molecule of ATP and controls the rate-limiting step of gluconeogenesis [ 28 ]. In the genome of C. autoethanogenum , gene CAETHG_2721 is annotated as an ATP-dependent PCK and significantly upregulated during autotrophic growth. Disruption of CAETHG_2721 impaired growth when cultured on gas only, the growth of ΔCAETHG_2721 was restored in the presence of both gas and fructose. This result showed that CAETHG_2721 also control the rate-limiting step of gluconeogenesis in the C. autoethanogenum [ 23 ]. Recently, Craig Woods et al. has developed the transposon insertion sequencing techniques for high-through functional genomics in C. autoethanogenum and identified 758 genes (19 % of the genome) essential genes for the autotrophic growth [ 29 ]. Fig. 1 Detailed mechanism of WLP and energy conservation mechanism in C. autoethanogenum. NAD + : nicotinamide adenine dinucleotide; NADP + : nicotinamide adenine dinucleotide phosphate. Fdh, formate dehydrogenase; THF, tetrahydrofolate; CoFeSP, corrinoid iron-sulfur protein; Acetyl-CoA, acetyl coenzyme A; CODH: carbon monoxide dehydrogenase; ADP, adenosine diphosphate; Pi, phosphate; ATP, adenosine triphosphate; Etf: electron transfer flavoprotein; The electron-bifurcating enzyme Nfn is responsible for the interconvesion of Fd, NADH and NADPH. C. autoethanogenum oxidizes hydrogen an enzyme complex of hydrogenase (HytA-E) and formate dehydrogenase (Fdh), achieving the reduction of ferredoxin (Fd) and NAD + or NADP + . The membrane-bound Rnf complex and ATPase are proposed to couple the electron transfer from reduced ferredoxin (Fd 2− ) to NAD + with the generation of ATP. Fig. 1 Like C. ljungdahlii, a variety of compounds can be produced from acetyl-CoA, acetate and ethanol are main products in the C . autoethanogenum ( Fig. 3 ). The formation of acetate is directly derived from acetyl-CoA through the pathway with phosphotransacetylase and acetate kinase. For the synthesis of ethanol, there are two pathways proposed: (1) The classic pathway via a bi-functional aldehyde/alcohol dehydrogenase (AdhE). (2) Acetate reduction to acetaldehyde and further to ethanol via an aldehyde: ferredoxin oxidoreductase (AOR) and alcohol dehydrogenase [ 30 ]. Fungmin Liew et al. demonstrated that AOR-based pathway is critical to ethanol formation in C. autoethanogenum [ 31 ]. During gas fermentation C . autoethanogenum is also known to synthesize other two valuable chemical compound, lactate and 2,3-BDO ( Fig. 3 ). The lactate is produced through direct reduction of pyruvate catalyzed by the lactate dehydrogenase (CAETHG_1147). The synthetic pathway for 2,3-BDO includes three key enzyme, acetolactate synthase (CAETHG_1740), acetolactate decarboxylase (CAETHG_2932) and 2,3-BDO dehydrogenase (CAETHG_0385), all of the genes for those enzymes have been identified in the genome ( Fig. 3 ). When reconstructing the complete pathway through expressing those enzymes in the E.coli , under anaerobic conditions, the resulting E. coli strain is able to produce 2,3-BDO comparable to the level produced by C. autoethanogenum during growth on CO containing waste gases [ 11 ]. 1.2 Energy metabolism in C. autoethanogenum The WLP is a net energy-consuming process and takes in one ATP molecule and eight reducing equivalents. In acetogenic bacteria, energy conservation is mainly relying on the membrane-associated reduced ferredoxin: NAD + oxidoreductase (Rnf), which can convert reduced ferredoxin (Fd 2− ）to Fd by coupling the translocation of protons across the cell membrane. The proton gradient generated will drive the ATPase to synthesize ATP for the carbon fixation. Besides, electron bifurcation is another mechanism of energetic coupling that saves cellular ATP for the bacteria, in which electron bifurcating enzymes deliver the electrons from the electron donor to two different electron acceptors coupling endergonic redox reactions to exergonic redox reactions [ 32 ] to overcome thermodynamic barriers and minimize free energy waste [ 33 ]. When oxidizing hydrogen, the electron-bifurcating and ferredoxin-dependent transhydrogenase (Nfn) or hydrogenase will produce Fd 2− and NADH/NADPH. Although there are six various types of hydrogenase systems encoded in the genome, the cells appear to contain only one active hydrogenase HytA-E1/E2 (CAETHG_2794-99) [ 27 ]. When CO is used as the sole carbon source, CO dehydrogenase will oxidize partial CO to CO 2 by coupling the formation of Fd 2− and NADH/NADPH [ 34 ] ( Fig. 1 ). Most acetogens can reduce CO 2 with H 2 to acetic acid via the WLP. By contrast, a few species such as C. autoethanogenum , C. ljundahlii and C. ragsdalei grow optimally between pH 5 and pH 5.5, and is able to form ethanol and acetate when fermenting H 2 and CO 2 [ 35 ]. With C. autoethanogenum as model, Johanna Mock et al. investigated the underlying special energy conservation mechanisms for its growth during ethanol formation from H 2 and CO 2 . It is shown that the presence of Rnf, Nfn and AOR with very high specific activities in H 2 /CO 2 -grown cells is key point to understand the energy metabolism of C. autoethanogenum [ 27 ]. 1.3 Advances in genetic tools for C. autoethanogenum An efficient DNA transfer method to introduce and express foreign DNA molecules is prerequisite to genetically engineer acetogens. Conjugation is widely used for DNA transfer in C. autoethanogenum, which relies on cell-to-cell contact between the donor strain, usually Escherichia coli and the receiving host [ 36 , 37 ]. ClosTron, a group Il-intron-based retrohoming gene inactivation method, has been widely used for gene inactivation in C. autoethanogenum [ 13 , 27 ]. ClosTron can be well used for the single gene mutation, but generating double genes mutation is proved to be impossible for this technique [ 31 ] ( Fig. 2 ). Fungmin Liew et al. developed an allelic exchange method for C. autoethanogenum, which utilizes a pseudo-suicide vector reliant on the pCD6 replicon [ 36 ] and counter selection marker composed of an orotate phosphoribosyltransferase gene from C. acetobutylicum [ 38 ] ( Fig. 2 ). This system was successfully used for making double mutation such as adhE1 + adhE2 and aor1 + aor2 [ 31 ]. Fig. 2 The genetic tools available for the C. autoethanogenum. ClosTron method is most widely used tools to disrupt gene expression in C. autoethanogenum. The CRISPR-Cas genome editing system has been efficiently used in C. autoethanogenum . The homologous recombination system for gene deletion is based on double crossover. CRISPR: clusted regularly interspaced short palindromic repeats; Cas: CRISPR-associated protein; Chr: chromosome; DSB: double-stranded break; HR: homologous recombination; LHA: left homology arm; RHA: right homology arm. Fig. 2 Fig. 3 The main products of wild-type or engineered C. autoethanogenum. NAD + : nicotinamide adenine dinucleotide; NADP + : nicotinamide adenine dinucleotide phosphate. ThlA, thiolase; CoA-SH: Coenzyme A; Acetyl-CoA: Acetyl Coenzyme A; Hbd: 3-hydroxybutyryl-CoA dehydrogenase; Crt: crotonase; Bcd: butyryl-CoA dehydrogenase; Fd 2− : reduced ferredoxin; Fd: ferredoxin; Ptb: phosphotransbutyrylase; Pi:phosphate; BuK: butyrate kinase; ATP: adenosine triphosphate; ADP: adenosine diphosphate; AdhE: aldehyde/alcohol dehydrogenase; CtfA/B: electron-transferring flavoprotein A and electron-transferring flavoprotein B; 2,3-Bdh: 2,3-butanediol dehydrogenase; Adc:acetoacetate decarboxylase; sAdh: secondary alcohol dehydrogenase; Pfor: pyruvate:ferredoxin oxidoreductase; Als: acetolactate synthase; BudA: acetoin decarboxylase; Ldh: lactate dehydrogenase; Pta: phosphotransacetylase; AOR: aldehyde:ferredoxin oxidoreductase; Ack: acetate kinase; Cit/D/E/F: citrate lyase; AcnB: aconitase; AceA: isocitrate lyase; GhrA: glyoxylate reductase; AldA: Glycolaldehyde dehydrogenase; fucO: lactaldehyde reductase. r-Box: reverse ꞵ-oxidation. Fig. 3 CRISPRs are adaptive immune systems evolved by bacteria to protect against exogenous genetic elements, such as phages and plasmids [ 39 ], and CRISPR-Cas-based genetic tools have been adapted for many organisms including some Clostridium [ [40] , [41] , [42] , [43] ]. Due to toxicity caused by uncontrolled Cas9 protein expression, CRISPR-Cas9-mediated desired gene deletion is unsuccessful in C. autoethanogenum . To address this, Shilpa Nagaraju et al. constructed and screened a small library of tetracycline-inducible promoters that can be used to finely tune gene expression [ 15 ]. With a new inducible promoter, the efficiency of target gene deletion in C. autoethanogenum was improved to over 50 %, making it a viable tool for engineering C. autoethanogenum [ 15 ] ( Fig. 2 ). To increase the pace of engineering progress, Nicholas Fackler et al. developed an inducible CRISPR interference (CRISPRi) system for C. autoethanogenum , which is used to repress the expression of crucial genes for the native production of 2,3-BDO [ 14 ]. More recently, a novel riboswitch-based editing tool, RiboCas, has been engineered to overcome excessive Cas9 toxicity by tightly repressing cas9 expression using a theophylline-inducible riboswitch [ 44 ]. Originally demonstrated in four non-acetogenic clostridial species, it has now been shown to function effectively for the generation of mutants in C. autoethanogenum [ 45 , 46 ]. Compared to Cas9, the utilization of Cas12a proteins is still not implemented in C. autoethanogenum , although they were more adapted to Clostridium genomes for the T-rich recognition and lower toxicity [ 47 ]. Based on their successful application in redirecting carbon flux of the closely related acetogenic bacterium ( C. ljungdahlii ) [ 48 ], Cas12a-based genome editing of C. autoethanogenum may be more applicable and promising. 1.4 Strain engineering A variety of strategies for manipulating the culture conditions were explored to optimize the yields of the native chemicals and fuels in C. autoethanogenum , which includes tuning pH value, adding the amino acids, optimizing the basic composition and modulating the composition of gas [ [49] , [50] , [51] , [52] ]. To improve our understanding of this organism's metabolism, Marcellin et al. has constructed genome metabolic model of C. autoethanogenum by combining multiomics and experimental data [ 23 ]. Moreover, Kaspar Valgepea et al. performed a systems-level steady-state gas fermentation study to build quantitative links between carbon, energy, and redox metabolism of this microorganism [ 53 ]. These studies have significantly accelerated the metabolic engineering of C. autoethanogenum as cell factories, which has been engineered to produce a wide range of biofuels and industrially relevant chemicals including primary and secondary alcohols, acids and terpenes ( Table 1 ). Notably, Lanzatech has genetically manipulated C. autoethanogenum through multiple metabolic engineering strategies and constructed an engineered strain capable of producing 10 g/L/h ethanol and 95 % selectivity under the optimal syngas fermentation condition, which resulted in one commercialized process for the production of bioethanol. Table 1 Strain engineering of C. autoethanogenum . Table 1 Year Genetic manipulations Results Scale References 2011 Expressing C. acetobutylicum butanol synthesis pathway genes 1.54 g/L Butanol 0.31 g/L Butyrate Lab [ 21 , 60 ] 2013 Expressing diol dehydratase genes from Klebsiella opylova 1.4 g/L 1-propanol 0.012 g/L 2-butanol Lab [ 61 ] 2013 Expressing exogenous mevalonate pathway enzymes and/or DXS pathway enzymes not quantified Isoprene Lab [ 62 ] 2014 Expressing native PFOR, alsS and alsD 9 g/L 2,3-BDO, 33 % selectivity Pilot [ 63 ] 2017 Inactivating aldehyde/alcohol dehydrogenase (AdhE) 2.46 g/L ethanol Lab [ 31 ] 2019 Expressing the PHB synthetic genes from Cupriavidus necator 10 % (w/w) of CDW Lab [ 64 ] 2019 Expressing operon aceA-ghrA- aldA and fucO from E. coli 0.029 g ethylene glycol/g fructose Lab [ 46 ] 2021 Multiple strategies in metabolic engineering 10 g/L/h ethanol, 95 % selectivity Commercial [ 14 ] 2020 Expressing the 3-hydroxybutyrate and butanol synthetic pathway optimized by the CFS 14.63 g/L 3-hydroxybutyrate 1. 63 g/L butanol 0.5 g/L 1,3-butanediol Lab [ 57 ] 2021 Expressing heterologous phenypyruvate decarboxylase, phenyacetaldehyde reductase and native key enzyme for the phenypyruvate synthesis pathway 0.28 g/L 2-phenylethanol Lab [ 58 ] 2022 Expressing the 1-hexanol synthetic pathway optimized by the CFS 0.26 g/L 1-hexanol Lab [ 59 ] 2022 Expressing the acetone or isopropanol synthetic pathway optimized by the CFS 3 g/L/h acetone, 90 % selectivity 3 g/L/h isopropanol, 90 % selectivity Pilot [ 12 ] 2022 Expressing the ꞵ-alanine pyruvate aminotransferase/γ-aminobutyrate transaminases, malonic semialdehyde reductase, aspartate decarboxylase/2,3-alanine aminomutase from multiple microorganisms 0.358 g/L/d 3-hydroxypropionate 0.0822 mg/L/d 1,3-propanediol Lab [ 65 ] Recently, the cell-free systems (CFS) are emerging as powerful platforms for synthetic biology applications, which allows for testing of hundreds to thousands of different designs within days. Particularly, this innovative way is used to inform strain design in absence of high-throughput capabilities [ 54 ]. Antje Krüger et al. developed and optimized one batch CFS platform for C . autoethanogenum , with which protein synthesis can reach up to 260 μg/mL. A key feature of the platform is that both circular and linear DNA templates can be applied directly to the CFS reaction to program protein synthesis [ 55 , 56 ]. The CFS has been successfully used to engineer this microorganism for the production of multiple chemicals ( Table 1 ). For example, Ashty S. Karim et al. screened 54 different cell-free pathways for 3-hydroxybutyrate production with CFS, which improved in vivo 3-HB production in Clostridium by 20-fold up to 15 g/L in a 1.5 L continuous system. The same group also optimized a six-step butanol pathway across 205 permutations and increased the production of butanol up to 22 mM in the batch ferment, which is the 10 folds than the highest yield previously reported for the engineered acetogenic bacteria [ 57 ]. Bastian Vögeli et al. have optimized and implemented reverse β-oxidation pathway in the C . autoethanogenum by the CFS platform, which generated C . autoethanogenum strains able to produce 1-hexanol from syngas, achieving a titer of 0.26 g/L in a 1.5 L continuous fermentation [ 58 ]. Most recently, combining genome mining and kinetic modeling prototyping cell-free system to optimize metabolic flux, Fungmin Eric Liews et al. generated one engineered strain capable of producing either acetone or isopropanol using syngas as a feedstock with productivity of over 3 g/L/h and 90 % selectivity in continuous pilot scale production, and life cycle analysis confirmed a negative carbon footprint for the products [ 12 ]."
} | 6,158 |
23141473 | null | s2 | 480 | {
"abstract": "Demand for sustainable materials motivates the development of microorganisms capable of synthesizing products from renewable substrates. A challenge to commercial production of polyhydroxyalkanoates (PHA), microbially derived polyesters, is engineering metabolic pathways to produce a polymer with the desired monomer composition from an unrelated and renewable source. Here, we demonstrate a metabolic pathway for converting glucose into medium-chain-length (mcl)-PHA composed primarily of 3-hydroxydodecanoate monomers. This pathway combines fatty acid biosynthesis, an acyl-ACP thioesterase to generate desired C₁₂ and C₁₄ fatty acids, β-oxidation for conversion of fatty acids to (R)-3-hydroxyacyl-CoAs, and a PHA polymerase. A key finding is that Escherichia coli expresses multiple copies of enzymes involved in β-oxidation under aerobic conditions. To produce polyhydroxydodecanoate, an acyl-ACP thioesterase (BTE), an enoyl-CoA hydratase (phaJ3), and mcl-PHA polymerase (phaC2) were overexpressed in E. coli ΔfadRABIJ. Yields were improved through expression of an acyl-CoA synthetase resulting in production over 15% CDW--the highest reported production of mcl-PHA of a defined composition from an unrelated carbon source."
} | 307 |
23141473 | null | s2 | 481 | {
"abstract": "Demand for sustainable materials motivates the development of microorganisms capable of synthesizing products from renewable substrates. A challenge to commercial production of polyhydroxyalkanoates (PHA), microbially derived polyesters, is engineering metabolic pathways to produce a polymer with the desired monomer composition from an unrelated and renewable source. Here, we demonstrate a metabolic pathway for converting glucose into medium-chain-length (mcl)-PHA composed primarily of 3-hydroxydodecanoate monomers. This pathway combines fatty acid biosynthesis, an acyl-ACP thioesterase to generate desired C₁₂ and C₁₄ fatty acids, β-oxidation for conversion of fatty acids to (R)-3-hydroxyacyl-CoAs, and a PHA polymerase. A key finding is that Escherichia coli expresses multiple copies of enzymes involved in β-oxidation under aerobic conditions. To produce polyhydroxydodecanoate, an acyl-ACP thioesterase (BTE), an enoyl-CoA hydratase (phaJ3), and mcl-PHA polymerase (phaC2) were overexpressed in E. coli ΔfadRABIJ. Yields were improved through expression of an acyl-CoA synthetase resulting in production over 15% CDW--the highest reported production of mcl-PHA of a defined composition from an unrelated carbon source."
} | 307 |
28152002 | PMC5289496 | pmc | 482 | {
"abstract": "Global climate change not only leads to elevated seawater temperatures but also to episodic anomalously high or low temperatures lasting for several hours to days. Scleractinian corals are detrimentally affected by thermal fluctuations, which often lead to an uncoupling of their mutualism with Symbiodinium spp. (coral bleaching) and potentially coral death. Consequently, on many Caribbean reefs scleractinian coral cover has plummeted. Conversely, gorgonian corals persist, with their abundance even increasing. How gorgonians react to thermal anomalies has been investigated utilizing limited parameters of either the gorgonian, Symbiodinium or the combined symbiosis (holobiont). We employed a holistic approach to examine the effect of an experimental five-day elevated temperature episode on parameters of the host, symbiont, and the holobiont in Eunicea tourneforti , E . flexuosa and Pseudoplexaura porosa . These gorgonian corals reacted and coped with 32°C seawater temperatures. Neither Symbiodinium genotypes nor densities differed between the ambient 29.5°C and 32°C. Chlorophyll a and c 2 per Symbiodinium cell, however, were lower at 32°C leading to a reduction in chlorophyll content in the branches and an associated reduction in estimated absorbance and increase in the chlorophyll a specific absorption coefficient. The adjustments in the photochemical parameters led to changes in photochemical efficiencies, although these too showed that the gorgonians were coping. For example, the maximum excitation pressure, Q m , was significantly lower at 32°C than at 29.5°C. In addition, although per dry weight the amount of protein and lipids were lower at 32°C, the overall energy content in the tissues did not differ between the temperatures. Antioxidant activity either remained the same or increased following exposure to 32°C further reiterating a response that dealt with the stressor. Taken together, the capability of Caribbean gorgonian corals to modify symbiont, host and consequently holobiont parameters may partially explain their persistence on reefs faced with climate change.",
"introduction": "Introduction Global climate change affects many ecosystems, including coral reefs [ 1 ]. One aspect of climate change is the rise of seawater temperatures that is anticipated to continue into the future [ 1 , 2 ]. In addition, short-term fluctuations in prevailing temperatures over several hours or days are also projected to occur more frequently [ 3 – 5 ]. Exposure to seawater temperatures even 2°C above the mean summer maximum can adversely affect corals and their mutualistic endosymbiotic dinoflagellate algae, Symbiodinium spp. [ 3 ]. Numerous studies have investigated the predominantly detrimental effects of elevated seawater temperatures on scleractinian coral— Symbiodinium symbioses (reviewed in [ 6 , 7 , 8 ]), but such data on other abundant coral reef cnidarians, such as octocorals, lag behind. In the Caribbean, for example, over the past few decades, scleractinian coral cover has dramatically declined [ 9 , 10 ] concurrent with a rise in seawater temperatures by 0.2–0.4°C/decade between 1985 and 2006 [ 11 ]. On the other hand, the abundance of Caribbean octocorals, predominantly gorgonian corals, has remained the same or even increased [ 12 – 15 ]. In fact, gorgonian corals constitute the dominant benthic fauna on many Caribbean reefs [ 13 , 14 , 16 , 17 ], where they provide food and shelter to a variety of invertebrates and fish [ 18 – 21 ]. Therefore, in order to understand the future of Caribbean reefs, it is imperative to determine the effects of potential stressors, such as elevated seawater temperatures, on gorgonian corals. In corals, thermal stress often leads to a reduction in Symbiodinium numbers and/or the amount of chlorophyll within the remaining Symbiodinium , which is commonly referred to as coral bleaching [ 22 ]. The elevated temperatures can disrupt Symbiodinium photosynthesis by hindering the repair of damaged photosystems [ 23 ], increasing the production of reactive oxygen species (ROS) that impair the thylakoid membranes [ 24 ], and inhibiting enzymes responsible for carbon fixation [ 25 ]. In addition, the production of high levels of nitric oxide (NO) in thermally stressed Symbiodinium can result in apoptosis [ 26 ]. Sensitivity to thermal stress can vary between different Symbiodinium clades and sub-cladal types [ 27 , 28 ]. Detrimental effects on the Symbiodinium may alter the nutrient exchange between the partners. Symbiodinium supply their host with carbohydrates, lipids, and essential and mycosporine-like amino acids [ 29 – 31 ], while the host provides Symbiodinium with carbon, nitrogen, nutrients, and an environment for photosynthesis [ 32 – 35 ]. Disruption of the symbiosis may alter nutrient exchange between the partners, the amount of energy required to maintain homeostasis, and drive the coral host and its symbionts to utilize their energy reserves [ 36 ]. For example, thermally stressed scleractinian corals and octocorals in the Indo-Pacific exhibit a drop in tissue reserves like lipids, proteins and carbohydrates [ 37 – 40 ]. In scleractinian corals, tolerance to, and the capacity to recover from, thermal stress is linked to the amounts of tissue reserves available [ 41 , 42 ]. Faced with stressors, Symbiodinium and corals can utilize several mechanisms to mitigate dysfunction in their cells. By increasing the activities of antioxidant enzymes like superoxide dismutase (SOD) they can convert superoxide to H 2 O 2 , and then further break H 2 O 2 down with peroxidase (POX) and catalase (CAT) to water and O 2 [ 28 , 43 , 44 ]. Corals can also reduce damage to proteins by increasing the production of heat shock proteins (Hsp) [ 43 – 45 ]. As in Symbiodinium , the ability of corals to cope with thermal stress can vary between different host taxa [ 43 , 46 ]. For example, Porites cylindrica , which possessed higher levels of SOD and Hsp than Stylophora pistillata , was better able to cope with, and recover from, thermal stress [ 43 ]. In contrast to the plethora of studies on the effects of elevated temperatures on scleractinian corals, only a handful of studies investigated the potential consequences of elevated seawater temperatures on Caribbean gorgonians. These studies focused only on a few parameters such as on the production of ROS, NO and Hsp90 [ 47 , 48 ], the effects of pathogens on gorgonian corals at ambient and elevated temperatures [ 49 – 51 ] and the effects of ultraviolet radiation in conjunction with elevated temperatures [ 52 ]. We decided to employ a holistic approach to determine the effects of elevated temperature on multiple parameters of the gorgonian host, the Symbiodinium and the subsequent holobiont in representative species of these important Caribbean reef taxa.",
"discussion": "Discussion Thermal fluctuations that expose coral reefs to anomalously high or low seawater temperatures for several hours or days are projected to occur more frequently in the future [ 3 – 5 ]. In scleractinian corals subjected to experimental conditions simulating such events, a 50–80% reduction in Symbiodinium density often occurs [ 39 , 43 , 91 – 96 ]. Subsequently, scleractinian corals may recover from the bleaching event. Conversely, the loss of Symbiodinium , compounded with the other stress responses, may lead to the demise of the host [ 39 , 43 , 94 – 96 ]. While we mimicked a short term thermal event by exposing branches of three gorgonian species, Eunicea tourneforti , E . flexuosa , and Pseudoplexaura porosa , to an elevated 32°C seawater temperature, the Symbiodinium densities in these branches did not significantly differ from Symbiodinium densities in branches from the same colonies maintained at the ambient temperature of 29.5°C. Furthermore, in P . porosa , Symbiodinium densities at the elevated temperature were actually higher, not lower, compared to those at the ambient temperature, although not significantly so ( Fig 2 ). The ability to continue hosting the same Symbiodinium density at elevated temperatures may be one reason why Caribbean gorgonians are maintaining or increasing their abundance on Caribbean coral reefs while scleractinian coral cover is declining [ 9 , 10 , 12 – 15 ]. Although the elevated temperature did not significantly alter the Symbiodinium density in the three gorgonian species, the Symbiodinium in branches of the gorgonian corals did react to the change in environmental conditions by modifying other Symbiodinium parameters. For example, at 32°C there was less Chl a and Chl c 2 per Symbiodinium cell which, in turn, affected the amount of chlorophyll per surface area ( Fig 2 ). Absorbance was less and, concomitantly, a* was more at the elevated temperature. Chlorophyll content can be altered over short timescales in response to changes in the environment (reviewed in [ 97 ]), and is a quicker response than Symbiodinium re-population, following symbiont loss, which can take from six weeks [ 58 ] to over two years [ 98 ]. The adjustments in pigments and subsequent light capture could in turn affect photochemical efficiency. In scleractinian corals, thermal stress can hamper symbiont photochemistry and photosynthesis [ 57 , 91 , 92 ], resulting in reductions in both Fv/Fm and ΔF/Fm`[ 43 , 99 , 100 ], leading to Q m values of 0.8 or above [ 99 , 101 , 102 ]. In the three gorgonian species, Fv/Fm at the elevated temperature was reduced, but Fv/Fm can also be lower due to activation of photoprotective processes [ 103 ]. In the Eunicea species, ΔF/Fm`did not mirror the reduction in Fv/Fm and was not affected by the elevated temperatures ( Fig 3 ). In P . porosa a 10% significant reduction in ΔF/Fm`occurred, but this reduction was much smaller than the >50% reduction in ΔF/Fm`recorded in thermally stressed scleractinian corals [ 43 , 99 , 100 ]. Furthermore, looking at daily changes in ΔF/Fm`in all three gorgonian species demonstrated that by day 5, ΔF/Fm`was actually higher at the elevated than at the ambient temperature. Lastly, in all three gorgonian species, the Q m in branches exposed to 32°C was either similar to or lower than the Q m in branches held at the 29.5°C ambient temperature, with Q m values being lower than 0.5 ( Fig 3 ). Given the numerous parameters related to photosynthesis, in addition to the maintenance of symbiont densities, the Symbiodinium appeared to not be photosynthetically compromised. Concomitantly, the elevated temperature did not lead to a change in the Symbiodinium genotypes in any of the three gorgonian symbioses. Lack of symbiont turnover either over time [ 104 , 105 ] or following environmental perturbation or disease in gorgonian corals ([ 49 , 78 ], Ramsby et al. unpubl., McCauley et al. unpubl.), other octocorals [ 106 ] and in numerous studies on scleractinian coral species [ 107 – 111 ] and sea anemones [ 27 ] has been demonstrated repeatedly. Although the gorgonians did not change their symbiont complement, the three gorgonian species did host different Symbiodinium types ( Fig 1 ), and analysis of the microsatellite Sym15 flanker region indicated that these Symbiodinium belonged to three distinct lineages of the “B1” radiation [ 72 , 112 ]. Symbiodinium type B41 in E . tourneforti in our study (previously referred to as B1l in [ 64 ]) fell within the same Symbiodinium lineage as the Symbiodinium hosted by E . flexuosa at >20m depth [ 73 ] and Symbiodinium endomacracis that associate with the scleractinian coral Madracis sp. [ 74 ] in the Caribbean. Symbiodinium types B41a and B41b (previously referred to as B1b in [ 64 ]) which we found in E . flexuosa , belong to a separate lineage that includes the Symbiodinium inhabiting E . flexuosa found at <5m depth [ 73 ]. Symbiodinium types B1i and B42 found in P . porosa belong to a novel lineage. Given the lack of a change in Symbiodinium , any response and potential acclimation to the stressor was accomplished by modifying parameters within the existing host/symbiont genotypic combination. The gorgonian species hosting different Symbiodinium , with these symbionts exhibiting different physiologies [ 62 ], may have contributed to the differences between the response of the gorgonian species to the elevated temperature. In addition to elevated temperature potentially affecting Symbiodinium , the entire symbiosis, including the host may be detrimentally affected [ 113 ]. Activation of cellular mechanisms to deal with stressors could increase the amount of energy required to maintain homeostasis [ 45 , 114 ], and thereby alter metabolism, and consequently the biochemical composition of tissues. In our study, compared to branches maintained at the ambient temperature, gorgonian branches exposed to elevated temperature exhibited higher sclerite content driven by lower protein and lipid contents per dry weight ( Fig 4 ), and also lower protein content when protein was assessed per organic matter. Thus, as seen in scleractinian corals [ 39 , 115 ], exposure to elevated temperature led to a reduction in the amount of tissue biomass present within gorgonian branches. Compared to scleractinian corals, however, this reduction was relatively small. For example, in bleached scleractinian corals, total biomass, mean protein, lipid and carbohydrate contents can be 40–70% lower [ 98 , 116 ], and mean energy content can be 22–37% lower [ 115 ] compared to tissues of unbleached corals. In our study, in the branches of the Eunicea species and P . porosa , protein per organic matter at the elevated temperature was only 6.5–14.3% lower compared to branches at ambient temperature. Furthermore, lipid, carbohydrate, and total energy content in tissues did not significantly differ between the gorgonian branches at ambient and elevated temperatures. Thus, despite some changes in biomass and protein content, exposure to elevated temperature did not affect the amount of energy available to these gorgonian species. An integral part of maintaining the symbiosis under thermal stress involves managing the levels of ROS produced in the chloroplasts of Symbiodinium and the mitochondria of the host [ 26 , 44 , 117 ]. Oxidative outbursts after exposure to elevated temperature have been recorded in Caribbean gorgonian corals, and their magnitude can vary between species [ 47 ]. Both Symbiodinium and their host cnidarian possess antioxidant enzymes that neutralize ROS [ 26 , 28 , 43 , 44 ]. In this study, SOD activity did not significantly vary between branches of the Eunicea species at ambient and elevated temperatures. Therefore, despite the nearly two-fold difference in SOD activity between the Eunicea species, basal levels of SOD in both species were sufficient to convert any excess O 2 - to H 2 O 2 [ 44 , 45 ]. H 2 O 2 itself is damaging because it can readily diffuse across membranes from one partner to the other, affect distant cell organelles, and trigger apoptosis [ 26 , 44 ]. The enzymes POX and CAT neutralize H 2 O 2 . POX activity in E . tourneforti was two times higher in branches exposed to elevated temperature than in those maintained at ambient temperature while in E . flexuosa branches, POX and CAT activity did not differ between the two temperatures. Therefore, the Eunicea species maintained or increased the activities of antioxidant enzymes when exposed to elevated temperature indicating that they managed oxidative stress. Looking at Symbiodinium , holobiont and enzymatic parameters, the three gorgonian- Symbiodinium symbioses examined dealt with the potential stress of elevated temperature, although the way in which they did so differed. In P . porosa many Symbiodinium parameters were modified in response to the elevated temperature. Not only did the largest reduction in Fv/Fm occur in P . porosa but Fv/Fm also progressively declined with the duration of exposure to elevated temperature. A reduction in chlorophyll content of symbiont cells also only occurred in P . porosa . Among the three species P . porosa has the highest symbiont density and pigment content in tissues, and the lowest a * Chl a (this study, [ 62 , 64 ]). These parameters along with attributes of the photosynthesis-irradiance curves, led Ramsby et al. [ 62 ] to hypothesize that the symbionts in P . porosa were comparatively less efficient at absorbing and utilizing light than those in E . tourneforti . Since high light levels can exacerbate thermal stress, the inefficient utilization of light may alleviate the negative effects of elevated temperature, and promote photoacclimation through adjusting photochemistry over losing symbiont cells from tissues. Furthermore, net photosynthesis in P . porosa is two to three times higher than in E . tourneforti [ 62 ], and P . porosa possess significantly greater amounts of tissue reserves than the Eunicea species (this study, [ 64 ]). Thus, the low efficiency of photosynthesis per symbiont cell coupled with high net photosynthesis and tissue resources may enable P . porosa to tolerate disruptions in symbiont photosynthesis that may occur when exposed to elevated temperature. In the Eunicea species, the modifications that occurred in the symbioses were predominantly at the holobiont rather than the symbiont level. Exposure to elevated temperature led to greater reductions in mean protein content per organic matter in tissues of E . tourneforti (14.30%) and E . flexuosa (12.11%) than in those of P . porosa (6.51%) and to a doubling of POX activity in E . tourneforti . Even under ambient conditions, the Eunicea species had lower Symbiodinium density, pigment content, and energy reserves than P . porosa (this study, [ 64 ]). Due to the lower tissue resources at their disposal, the Eunicea species may attenuate changes in symbiont parameters by maintaining or increasing antioxidant activity to survive unfavorable conditions. In the literature, bleaching of Caribbean gorgonian corals is seldom reported [ 118 – 121 ] and the three gorgonian species in our study did not exhibit a decline in Symbiodinium density although a reduction in the amount of chlorophyll per gorgonian surface area did occur, with P . porosa having a larger drop than within the Eunicea species. The varied responses of the gorgonian corals in our study match the inter-species differences in a visual assessment of bleaching on the reef [ 121 ]. In a 2005 bleaching event in Puerto Rico, 22% of Pseudoplexaura spp. colonies bleached [ 121 ]. In contrast, bleaching was observed in only a few E . flexuosa colonies and none of the other Eunicea species bleached [ 121 ]. Our study, however, suggests that even with a reduction in chlorophyll at 32°C, Symbiodinium photosynthesis in P . porosa was not compromised, and therefore the changes in pigment content were potentially part of an acclimatory response. This may explain why, with the exception of Muricea sp., the gorgonian species that were observed bleached in 2005 survived the thermal event [ 121 ]. Furthermore, in the 2005 bleaching event, bleaching in the Caribbean gorgonian species occurred much after most scleractinian corals, hydrocorals and zoanthids had bleached [ 121 ]. Taken together, the response of the gorgonian symbioses to elevated temperature in this study, and the few reports on bleaching in gorgonian corals [ 118 – 121 ], suggest that in the Caribbean, gorgonian corals may be comparatively more tolerant to thermal stress than many scleractinian coral species."
} | 4,933 |
31608130 | PMC6774398 | pmc | 483 | {
"abstract": "ABSTRACT Since 2012, a triboelectric nanogenerator (TENG) has provided new possibilities to\nconvert tiny and effective mechanical energies into electrical energies by the physical\ncontact of two objects. Over the past few years, with the advancement of materials’\nsynthesis and device technologies, the TENGs generated a high instantaneous output power\nof several mW/cm 2 , required to drive various self-powered systems. However,\nTENGs may suffer from intrinsic and practical limitations such as air breakdown that\naffect the further increase of the outputs. This article provides a comprehensive review\nof high-output TENGs from fundamental issues through materials to devices. Finally, we\nshow some strategies for fabricating high-output TENGs.",
"conclusion": "4. Conclusions We have summarized the recent progress in the development of high-output TENGs from\nfundamentals to devices. Since the discovery of the TENG on 2012, various TENG devices\nincluding the materials’ synthesis technologies generated quite high instantaneous power\ndensities up to several tens of mW/cm 2 under practical input force conditions\nassociated with various mechanical energy sources. Because of the rapid development during a\nfew years, various potential applications such as portable power source and self-powered\nsensors successful demonstrated, giving the possibility of the commercialization,\nespecially, related with self-powered wireless transmission technologies. The charge density\nof the TENGs was continuously increased, by introducing ions injection and poling process\nunder high electric field to multi-layered film or composites-type film. Further increase of\nthe charge density may be limited by air breakdown, meaning that designing concepts for new\ndevices are needed, such as metal–metal contacts and self-charging technologies. We believe\nthat this review can be useful and helpful for designing energy harvesters which are\npossible to provide sufficient energy with portable electronic devices.",
"introduction": "1. Introduction Since the first report of triboelectric nanogenerator (TENG) on 2012, the rapid development\nof a variety of functional devices and high-performance materials dramatically enhanced the\noutput power, proven to be one of the highly efficient, simple, robust and cost-effective\ntechniques for converting mechanical energies around us to electricity [ 1 – 5 ]. The mechanical energy sources involve a wide\nvariety of classifications for types of energy such as human motion, wind flow, flowing\nwater, vibration and any other mechanical motion. Under surrounding environments of ambient\ntemperature and relative humidity, and under practical input force conditions associated\nwith each energy sources, quite high instantaneous power densities up to several tens of\nmW/cm 2 were routinely reported in several papers so far [ 1 , 6 – 11 ]. The technological\nadvances made the TENGs to ensure continuous and reliable supply of power for many devices\nsuch as wearable devices, sensors, smartphones and medical devices, giving the realization\nof various self-powered systems [ 8 , 12 – 17 ]. Among many potential applications in the near future, the self-powered wireless\ncommunication technologies may be quite attractive because of the broad applicability such\nas sensor network system, security systems, intelligent transportation systems and patient\nmonitoring and telemedicine [ 18 , 19 ].\nSo far, the technologies have transformed the means of information exchange or sharing,\ncommunication and transactions, leading to the new digital economy. They are also deeply\nrelated with the realization of The Fourth Industrial Revolution, stimulated by numerous\ninternet of things (IOT) sensors and wireless transmission of data or information. Thus, the\nglobal market in the wireless data communication market was reported to be valued at 794.6\nmillion USD in 2018, expected to reach 1867.8 million USD by 2023 [ 20 ]. Introducing a new concept of self-powered technologies will allow many\ndevices to be free from any power cables or battery-changing task, which will further\nincrease the market, more than as expected. Figure 1 shows the\npower consumption of the Bluetooth, compared with the instantaneous output power densities\nof various TENGs developed so far. It is obviously seen that the Bluetooth energy has\ncontinuously decreased, while the output power of the TENGs has increased up to several\nmW/cm 2 [ 1 , 7 – 9 , 11 ]. This means that the output power\nmay become comparable to the Bluetooth energy in the near future. However, for the\nsuccessful implementation of TENGs in practical applications, it is essential to store the\nsufficient output power in a capacitor or a battery, to operate the devices. Although there\nwas some progress in increasing the charging efficiency via the optimized circuit design,\nthe efficiency was still so low [ 21 , 22 ]. 10.1080/14686996.2019.1655663-F0001 Figure\n1. Power consumption for Bluetooth and instantaneous output powers\nof various TENGs. In general, many strategies to increase the output powers of TENGs are based on the\nenhancement of the charge transfer occurring between two contacted surfaces to increase the\ntransferred charge density because it determines the electric potential between two\nelectrodes. According to the triboelectric series, many combinations associated with various\npolymers have been applied to various TENGs to show quite high charge densities up to\nseveral hundred μC/m 2 [ 6 , 8 , 10 , 23 , 24 ]. Some advances associated with the contacted materials involved electronic and\nstructural modifications, such as a large work-function difference, high porosity, large\ndielectric constant and large surface roughness [ 17 , 25 – 32 ]. The maximum charge density of\nthese TENGs in practical environments, however, requires additional processes, such as\nartificial charge injection and poling process under high electric field [ 6 , 10 , 33 ].\nRecently, several modifications to amplify the current flowing through the external circuit\nhave been successfully demonstrated [ 34 , 35 ]. Metal-dielectric-metal structures or metal-metal contacts have been\nintroduced and have been proven to increase the charge density of the TENG by several times\n[ 8 , 36 ]. Here, we review various\nstrategies for ultrahigh outputs generation in triboelectric energy harvesting technologies\nfrom the charge transfer mechanism to the fabrication of the devices."
} | 1,601 |
33291306 | PMC7731204 | pmc | 484 | {
"abstract": "Superhydrophobic surfaces have attracted intensive attention in the antifouling field because of their excellent anti-bioadhesive performance and environmental friendliness. However, promising surfaces have met great challenges of poor mechanical robustness under harsh serving conditions. Herein, an organic-inorganic composite strategy, that the silane-modified TiO 2 nanoparticles are compounded into the porous framework provided by the stable and indurative aluminum oxide film, is proposed to address the common serious problem in superhydrophobic surfaces. Different from the traditional superhydrophobic surfaces, this composite film possesses a ~18 μm thick layer which can provide strong support to silane-modified TiO 2 nanoparticles. The resulting film can reserve superhydrophobicity to the surface even after a thickness loss of ~15 μm under continuous abrasion. At the same time, the results of the bacterial adhesive tests also verify that the film has the same long-term anti-bioadhesive performance. The film with superhydrophobicity, excellent anti-bioadhesive property, and stable robustness will make it a promising candidate for serving in a harsh environment, and the design concept of this film could be applied to various substrates.",
"conclusion": "4. Conclusions In summary, we design an organic–inorganic superhydrophobic surface with robustness against abrasion by filling the high-hardness nanoporous aluminum oxide film with the low-surface-energy PFDS/TiO 2 . The surface exhibits superhydrophobicity with a large WCA of 159°, and the WCA can keep above 140° even after a ~15 μm thickness loss. Besides, the coating performs durable efficient inhibition of the bacterial adhesion under continuous abrasion. For engineering applications, this superhydrophobic surface has the intrinsic high hardness and strong adhesive force, and it can be achieved by a high-output fabricating method at a low-cost. Furthermore, the design strategy of the robust surface with the special microstructure can be widely adapted to the other materials that need stable and robust armor for service in harsh environments.",
"introduction": "1. Introduction Anti-bioadhesive surfaces have become research hotpots because of the urgent need for seeking the alternative to replace the traditional antibacterial coatings that always contain fungicides [ 1 , 2 ]. Numerous researchers [ 3 , 4 ] have done much investigation on the strategies to develop environmentally friendly anti-bioadhesive surfaces, among which, superhydrophobic surfaces with high water contact angle (WCA, ≥150°) have shown super anti-bioadhesive performance and nontoxicity to environments. To achieve the superhydrophobic surfaces, a common approach is modifying the surface with the low-surface-energy chemistry as well as decreasing the contact area between liquid and solid via creating the micro-/nanoscale textures or fabricating the nanoporous structures [ 5 , 6 , 7 ]. To increase the surface hydrophobic property, many methods have been reported, including etching [ 8 ], electrochemical deposition [ 9 ], magnetron sputtering [ 10 ], lithographing [ 11 ], a thermochemical synthetic method [ 12 ], and sol–gel processing [ 13 ]. For example, Maharana et al. fabricate a nano-hierarchical structured Cu-ZrO 2 nano-cone arrays to acquire hydrophobicity [ 14 ]. These researches indicate that the effects of the microstructure geometries, the high aspect ratio guard ring structure, and the hierarchical surface roughness are the key factors that can improve the superhydrophobicity. Whereas, a small fraction of the overall area in contact with liquid results in high local pressure on the contact area, which weakens the surface robustness against abrasion [ 5 , 6 , 15 ]. To improve the mechanical robustness of the superhydrophobic coatings, various approaches have been explored, such as enhancing the bonding force between the coating and the substrate [ 16 ], strengthening the hardness of the coating [ 17 ], fabricating a biomimetic self-healing surface [ 18 , 19 , 20 ], and creating a self-similar structure by the low-output methods, such as lithographing and magnetron sputtering [ 10 , 11 , 21 , 22 ]. Among these approaches, coatings with high hardness and strong bonding force have shown improved robustness comparing with the general coatings. For instance, Du et al. [ 23 ] have reported a super-robust hydrophobic coating fabricated by multi-arc ion plating, and the coating exhibits enhanced hardness, good adhesive force, and excellent wear resistance, whereas this strategy has resulted in only modest advancements in mechanical robustness and it will cause rapid failure once the coating suffers damage [ 6 ]. For the bionic strategy, it is often too hard and expensive to mimic the intrinsically self-repairing ability [ 5 ]. Therefore, it is urgently needed to develop a robust superhydrophobic coating that not only possesses high hardness, strong bonding force, and large self-similar layer but also can be mass-produced at a low cost. Aluminum oxide film [ 24 , 25 ] with high hardness and good anti-corrosion property, has widely used as the surface protective layer of the structural material applied in ships, aircraft, and vehicles, and so on. The nanoporous microstructures of the aluminum oxide film can be controlled by adjusting the fabricating parameters [ 26 ], and the selected nanoporous surfaces can act as the container to load the functional nanofiller. Many powders—such as TiO 2 [ 27 ], Cu 2 O [ 28 ], zirconia [ 29 , 30 , 31 ], silica [ 32 ], ZTA [ 33 ], and SiC [ 34 ]—have reported as the functional powders to modify the metallic or ceramic surfaces, and these powders are obtained by various methods including sol–gel [ 35 ], coprecipitation [ 36 ], subcritical drying [ 37 ], and so on. For instance, M.Taha et al. have fabricated a SiC-reinforced 6061 Al to achieve the physical, mechanical, and electrical properties [ 34 ]. Among these powders, TiO 2 nanoparticles, with good dispersibility, will provide promising applications in surface engineering since they can effectively increase the surface roughness and improve the specific surface area [ 27 ]. Moreover, the nanoparticles enable the formation of the nanocomposite induced by the low-surface-energy organic modifiers [ 38 ]. Perfluorodecyltrimethoxysilane (PFDS), with the lowest surface energy and the best wetting-repellency, is a commercial surface-active agent widely used in the coatings of electronic screens [ 39 ]. However, the adhesive force between the substrate and the PFDS-contained coating is just through the van der Waals force in the traditional applications. Herein, we suggest combining the TiO 2 nanoparticles with PFDS to form a nanoparticle-induced sol–gel, and then filling the sol–gel into the nanopores of the aluminum oxide film. Thus, the active hydrophobic groups will be transplanted to the nanoporous surface through the strong chemical bonds between pore-walls and active groups [ 40 ]. In this work, we design a superhydrophobic surface with excellent mechanical robustness against abrasion that is achieved via a simple process. The microstructure of the surface contains a thick nanoporous framework filled with functional nanoparticles to provide water repellency and durability. The choice of the nanoporous frameworks with different microstructure geometries has been discussed based on the Cassie–Baxter theory. As an application, the anti-bioadhesive performance of the surface before and after abrasion has been investigated. The research sense of this work is to provide an alternative strategy of fabricating an excellent robust superhydrophobic surface in an affordable manufacturing method and to launch a design concept that can be conveniently applied to the other metal surfaces.",
"discussion": "3. Results and Discussion 3.1. Design Strategy to Obtain the Robust Superhydrophobic Coating The strategy for achieving a robust superhydrophobic surface is to fabricate an organic–inorganic composite layer on 6061 Al through filling the aluminum oxide film with PFDS/TiO 2 , as shown in the schematic illustration in Figure 1 a. aluminum oxide film, as the foundation of the framework, intrinsically has high hardness, strong bonding force with the substrate, corrosion resistance, and good processing property. Furthermore, the nanopores of the aluminum oxide film can be used as the storage of the hydrophobic functional groups. After being filled with the nanoparticle-induced low-surface-energy nanofiller, the surface with the nanoporous framework will become to repel water. The wetting state of the surface can be expressed by the Cassie–Baxter model ( Figure 1 c) that the droplet keeps above the Cassie interface [ 6 , 23 ] due to the existence of the air cushion in the nanoporous structures. According to Cassie–Baxter theory, the WCA of the superhydrophobic surface, θ, can be expressed by the Cassie–Baxter relation [ 6 ]\n (2) cos θ γ = f ls × ( cos θ 0 + 1 ) − 1 \nwhere, θ γ is the apparent contact angle, θ 0 is the intrinsic contact angle, f ls is the percentage of the liquid-solid interface area. From the relation, it can get that θ γ has a negative correlation with f ls ( f ls < 1 ). Therefore, minimizing the liquid-solid interface can improve the WCA, but it will cause a decrease in mechanical robustness. To optimize the robustness and superhydrophobicity, we prepare these aluminum oxide films in different electrolyte systems under various anodizing voltage. SEM images in Figure 2 show the morphology of aluminum oxide film formed in the three electrolytes under various anodizing voltage, and the corresponding average pore diameter, porosity (see the porosity calculation method in Figure S2 ), and hardness (see the testing details in Figure S3 ) are listed in Table 2 . Results show that aluminum oxide film prepared in the H 3 PO 4 electrolyte system ( Figure 2 a) has large nanopores (~130 nm), and the diameter of the nanopores increases with the anodizing voltage. However, the hardness of aluminum oxide film formed in the phosphoric acid system is below 140 HV, which is not very conducive to engineering application. Aluminum oxide film fabricated in the H 2 SO 4 electrolyte system tends to form small nanopores (~85 nm) and thin pore-wall ( Figure 2 b), and the hardness is above 270 HV. Whereas, the small nanopores are not fit as the modifier storage which will be filled with nanoparticle-induced nanofiller. Figure 2 c displays the morphology of the aluminum oxide film formed in the OA electrolyte that possesses large pores, and the hardness is relatively high. Therefore, the pore size of the fabricated aluminum oxide film in different electrolytes increases in the following order: H 2 SO 4 < OA < H 3 PO 4 , and the hardness heightens in the following order: H 3 PO 4 < H 2 SO 4 < OA. Therefore, the optimized aluminum oxide film that simultaneously has a large pore size, high porosity, and high hardness can be achieved in the OA electrolyte system ( Figure 2 c). As shown in Figure 2 c, the nanoporous aluminum oxide film formed in OA electrolyte exhibits fine roughness and a unique nanoporous structure with high uniformity. According to the above comprehensive analysis of the requirement in the filling process, the mechanical robustness, and the wetting model, we select the aluminum oxide film with a pore size of ~100 nm, the porosity of 45.9% (see the calculating details in Figure S2 ), and the hardness more than 350 HV ( Table 2 ) as the metal framework of the coating. Three reasons are listed to explain the choice: first, the nanopore sized ~100 nm is large enough for the nanofiller with TiO 2 nanoparticles sized ~15 nm; besides, based on the Cassie–Baxter relation, the theoretical WCA of the filled aluminum oxide film with the porosity of 45.9% can reach ~160° (see the calculating details in Supporting Information ); moreover, the high hardness (above 350 HV) is sufficient to meet the requirements of the engineering application. Thus, the optimized aluminum oxide film has been picked out as the metal framework of the coating. Previous studies [ 30 , 41 ] have pointed to that a large diameter of dispersed TiO 2 nanoparticles (sized ~15 nm) is recommended as the structural modifier to increase the surface roughness and improve the specific surface area. For both of the above reasons, an aluminum oxide film with ~100 nm diameter is selected as the nanoporous framework of the surface, and TiO 2 nanoparticles with ~15 nm diameter are chosen as the nanofiller. 3.2. Analysis of the Morphologies and Components Morphologies of the surface and cross-section are characterized by SEM ( Figure 3 ). Figure 3 a displays the aluminum oxide film structure that is composed of nanopores (sized ~100 nm) and pore walls (sized ~35 nm), and the nanopores and walls are evenly spaced with no presence of collapses and cracks. To obtain the superhydrophobic surface, the aluminum oxide film is filled with low-surface-energy PFDS/TiO 2 . As shown in Figure 3 b, the filled surface involves the random and evenly honeycomb-like structures, and the nanoparticles inside of the nanopores have an average diameter of 15 nm. To achieve a durable superhydrophobicity, the organic–inorganic framework is designed to form a honeycomb-like structure with a large thickness in the depth direction, that can endow the coating with robustness against abrasion by sacrificing the upper layers in a self-similar manner. Figure 3 c displays the cross-section of the coating, and the length from the upper surface to the bottom is shown to be ~18 μm. In the magnified SEM image of the cross-section ( Figure 3 d), it can be seen that the nanoparticles are deep filled into the coating bottom. The fully filled coating at ~18 μm thick can provide the sustaining superhydrophobicity and excellent robustness under continuous abrasion. TOF-SIMS represents the elemental peaks on the coating, and the quantificational elemental contents of Al, Ti, Si, C, O, and H are listed in the insert table in Figure 4 . Strictly speaking, a large amount of Al element attributes to the aluminum oxide film and the 6061 Al substrate, the existence of Ti is due to TiO 2 nanoparticles, Si and C are the main elements of PFDS, and O element, with a high percentage, derives from TiO 2 , aluminum oxide film, and PFDS. To distinguish the elemental forms, XRD and FTIR are employed to characterize the metal composition and organic matter, respectively. Figure 5 shows the XRD pattern of the TiO 2 nanoparticles, the AAO, and the fabricated PFDS/TiO 2 @AAO surface. The XRD of TiO 2 nanoparticles reveals the presence of the anatase polymorph with the crystallographic planes in orientations of (101), (103), (004), (112), (200), (105), (211), (204), (116), (220), and (215) [ 30 , 42 ]. These crystallographic planes are consistent with those reported in the previous paper [ 22 ]. As the XRD result, AAO has a peak curve bread that represents amorphous and three crystallographic planes in orientations of (200), (220), and (311). These crystallographic planes of TiO 2 nanoparticles have been found in the PFDS/TiO 2 @AAO surface, which indicates that it cannot cause the crystallographic-plane transformation of TiO 2 nanoparticles in the modifying process. FTIR ( Figure 6 ) records the organic groups in the coating. The absorption peaks appear at 1033 cm −1 and 1150 cm −1 (consistent with the Si-O-Si stretching), 1420 cm −1 (indicating the C-O stretching), 2853 cm −1 , and 2926 cm −1 (referring to the C-H of methylene) [ 43 ]. This indicates that the functional organic groups are attached to the surface and endow the surface with the good water-repellent property. However, these results cannot illustrate the bonding form between the organic groups and the aluminum oxide film. Here, XPS is performed to investigate the chemical conditions of the superhydrophobic surface. Figure 7 a shows that the elements of C, F, and Si are detected in the surface, implying that the surface has been covered with silane. Based on the previous research, it can be deduced that the chemical bonds of Al-O-Al in aluminum oxide film and the Si-O in PFDS are opened in the hydroxylation process, and the -OH groups (existing in ethyl alcohol) are grafted to these open chemical bonds [ 44 , 45 ]. Then, the hydroxylated aluminum oxide film and PFDS are chemically combined in the dehydration process [ 46 , 47 ] as the schematic shown in Figure 7 b. Thus, the low-surface-energy organic groups are grafted on the framework of the coating through the two processes of hydroxylation and dehydration. 3.3. Superhydrophilicity, Anti-Bioadhesive Performance, and Robustness WCA of the samples before and after filled with PFDS/TiO 2 are tested to investigate the water repellent performance. Contrastively, the WCA changes from 68° ( Figure 8 a) to 159° ( Figure 8 e) after the surface is modified with PFDS/TiO 2 , indicating that the water repellent performance transforms from hydrophilic state to superhydrophobic state. Due to the good water repellent property, the superhydrophobic PFDS/TiO 2 @AAO is hardly adhered to by the bacterial colonies of E. coli and S. aureus ( Figure 8 g,h). The superhydrophobic PFDS/TiO 2 @AAO performs a 100% inhibition ratio (see calculating details in Experimental Section) against the two bacteria, displaying excellent anti-bioadhesive performance. However, the unmodified aluminum oxide film is seriously adhered to by the bacteria ( Figure 8 c,d), exhibiting no inhibition effect of bacterial adhesion. These results suggest that the water-repellent property and the anti-bioadhesive performance can be significantly improved by creating thus a superhydrophobic coating on the surface. Standard sandpaper abrasion is performed to test the mechanical robustness of the superhydrophobic PFDS/TiO 2 @AAO coating (see the experimental details in the Experimental Section). The WCA changes as functions of the abrasive thickness loss are shown in Figure 9 . The WCA remains above 145° after 5 μm thickness loss ( Figure 9 ), revealing that slight abrasion damage has little impact on the coating hydrophobicity. With the increasing thickness loss, the WCA decreases slightly, but still maintains above 140° before a thickness loss of 15 μm. The steep decline in WCA occurs after ~15 μm thickness loss, and the WCA falls to 101° after a total thickness loss of ~20 μm, indicating the failure of the superhydrophobic coating. Here, to keep a consistent surface roughness of the testing samples, all the testing surfaces are kept at the same scale in the microstructures that are characterized by AFM ( Figure 10 ). To verify the durability of the anti-bioadhesive performance, bacterial adhesion testing is performed on the superhydrophobic surface before/after abrasion. In the initial state, the surface before abrasion shows good inhibition of bacterial adhesion against E. coli and S. aureus with no bacterial colony adhesion ( Figure 10 (a3,a4)). After 5 μm thickness loss, it still displays excellent anti-bioadhesive performance as shown in Figure 10 (b3,b4). However, the adhesive bacterial colonies are sporadically distributed on the testing plate after the coating is abraded off 15 μm ( Figure 10 (d3,d4)). Results in Figure 10 (e3,e4) indicate that the coating is entirely ineffective with a mass of bacterial colonies adhered to after ~20 μm thickness loss. These results suggest that the coating can provide effective inhibition of bacterial adhesion within ~15 μm thickness. The robust and durable superhydrophobic surface can be applied in the harsh environment, such as the guild rails of the automobile skylights that serves under the abrasive condition (see the details in Figure S5 ). The mechanism schematic in Figure 11 reveals why the PFDS/TiO 2 @AAO is robust against abrasion. As the discussion in Figure 7 and Figure 3 c,d, the functional organic groups have been grafted onto the wall of the nanopores, and the nanoparticles-induced modifier is filled deep into the bottom of the nanopores. During the abrasion ( Figure 11 b,c), the upper layer of the PFDS/TiO 2 @AAO is abraded while the remaining layer can still provide the hydrophobicity benefiting from the self-similar structure of the nanoporous framework. Until the thickness loss achieves to ~20 μm ( Figure 11 d), the PFDS/TiO 2 @AAO is completely expended during the continuous abrasion, and the bulk substrate material exposes, which fails the superhydrophobicity. In practice, adhesion parameters are important to a superhydrophobic surface. To evaluate the adhesion of the coating, scratch test, and bending test are employed in this work. Scratch test can provide the anti-scratch ability of the surface. Figure 12 a shows the scratch that broadens and deepens with the increasing load. The variations of friction and sound signal with the increasing loading are recorded by the tester. Figure 12 b shows that the friction and sound signal concurrently occur saltation under 40 N load, which indicates the coating fracture. The bending test can directly give the bending strength of the coating. Figure 13 exhibits the variation of the bending strength with the displacement. It can be seen that the curve has saltation at the bending strength of 234 MPa that implies the coating cracks. It can be seen that the cracking strength of the coating is near to the yield strength of the sample (244 MPa), which suggests that the coating has an excellent anti-bending performance. SEM image of the bending sample ( Figure 14 a) displays that the surface coating has cracked after the bending test. EDS mapping ( Figure 14 b–g) verify the cracks happen on the total coating because the substrate has been exposed between the cracks."
} | 5,461 |
39851362 | PMC11761944 | pmc | 485 | {
"abstract": "This study evaluates the potential of biorefinery and dairy wastewater as substrates for electricity generation in double chamber Microbial Fuel Cells (DCMFC), focusing on their microbial taxonomy and electrochemical viability. Taxonomic analysis using 16S/18S rDNA-targeted DGGE and high-throughput sequencing identified Proteobacteria as dominant in biorefinery biomass, followed by Firmicutes and Bacteriodota. In dairy biomass, Lactobacillus (77.36%) and Clostridium (15.70%) were most prevalent. Biorefinery wastewater exhibited the highest bioelectrochemical viability due to its superior electrical conductivity and salinity, achieving a voltage yield of 65 mV, compared to 75.2 mV from mixed substrates and 1.7 mV from dairy wastewater. Elevated phosphate levels in dairy wastewater inhibited bioelectrochemical processes. This study recommends Biorefinery wastewater as the most suitable purely organic substrate for efficient bioelectricity generation and scaling up of MFCs, emphasising the importance of substrate selection for optimal energy output for practical and commercial viability.",
"conclusion": "4. Conclusions and Recommendations 4.1. Conclusions In the quest to ensure compliance with stringent industrial wastewater quality standards, the adoption of effective treatment and advanced dewatering technologies is imperative. Among these technologies, the Bioelectrochemical System (BES), specifically the Microbial Fuel Cells (MFCs) system, has exhibited remarkable reliability, sustainability, and renewability in addressing the challenges posed by high-strength complex industrial substrates and contaminants, ultimately yielding bioelectricity [ 24 , 56 , 57 , 58 ]. The domain of Microbial Fuel Cells (MFCs) is rapidly evolving and demands the development of reliable and scalable commercial voltage generation units [ 58 , 59 , 60 , 61 , 62 ]. Achieving optimal MFC operation necessitates a multifaceted approach encompassing scientific knowledge, wet chemistry, microbial studies, and electro-engineering aspects, all tailored toward enhancing voltage yield generation [ 23 , 42 , 61 , 63 , 64 , 65 ]. This study underscores the critical importance of identifying the most dependable and readily available raw wastewater samples for use as anolytes within the MFC system. Moreover, the necessity for well-cultured and active microbial consortia to facilitate the biodegradation of complex organic compounds into electrical energy has been underscored [ 61 , 66 , 67 , 68 , 69 ]. Based on the taxonomical classification, along with the concurrent analysis of small RNA (sRNA) and ribosomal DNA (rDNA) identification from the three complex wastewater sources, the following conclusions have emerged: Biorefinery wastewater exhibits substantial conductivity capacity, elevating conductivity levels to approximately 12,000 µS·cm, making it the most saline raw water source. It yielded an overall voltage of approximately 230 mV when processed from pure and untreated biorefinery wastewater. The mixed stream wastewater source ranks second in terms of conductivity and salinity reliability. Remarkably, it recorded the highest voltage yield of 76 mV when processed from pure substrates. Clover wastewater, on the other hand, is situated at the lower end of the taxonomical classification, yielding a mere 1–5 mV when subjected to raw anolyte feeding. This result is correlated with its notably high phosphate content, which appears to inhibit the bioelectricity generation process, [ 18 , 40 , 74 , 75 , 76 ]. These findings highlight the potential of different wastewater sources for bioelectricity generation in MFCs, with biorefinery and mixed stream wastewater sources showing promise for further exploration and optimisation in scaling up voltage yield generation. 4.2. Recommendations Based on the findings presented in Figure 11 a,b, sourced from Shabangu et al. [ 70 , 71 , 72 , 73 ] several important recommendations have been drawn: i. Biorefinery and mixed wastewater streams emerge as highly viable options for serving as reliable anolyte and inoculum sources in operating this benchtop DCMFC unit utilising purely raw industrial wastewater as an electrogenic bacterial community and efficient anolyte. ii. The heterotrophs classified in the morphology section, specifically Proteobacteria and Bacteroidetes, are considered as viable sources of electrons for operating this benchtop DCMFC unit in subsequent experiments of this study. Biorefinery wastewater stands out for its reliability, renewability, and sustainability in terms of being the primary DCMFC bioelectricity source. iii. Organic removal or biodegradable contaminants as shown in Figure 11 a, b, demonstrated that in all three different wastewater sources, the overall percentage removal was achieved within a short span of time, specifically, within 72 h of treatment incubation in the DCMFC technology. Complete 100% removal was observed with mixed wastewater substrates within the same 72 h treatment period. From a scientific perspective, precisely in the field of bioelectrochemistry, wet chemistry, and biodegradation principles, it is evident that the duration of high-percentage organic removal is influenced by the length of incubation periods, in the DCMFC unit. This finding of this study primarily relates and validates previous work on conventional wastewater treatment strategies reported in studies conducted by Shabangu et al. [ 70 , 71 , 72 , 73 ], Logan et al., [ 14 , 15 ] and Kim et al. [ 20 ] using traditional H-shaped Microbial Fuel Cells. In summary, the choice of biorefinery and mixed wastewater streams as anolyte sources in MFC technology and the identification of Proteobacteria and Bacteroidetes as viable sources of electrons are key takeaways from this study, emphasising the efficiency and rapidity of organic removal within MFC units. These findings contribute to innovating the field of industrial wastewater treatment and bioelectrochemical processes for practicality and applicability at commercial scale platforms. 4.3. Future Perspectives Based on the above underpinned findings and conclusions of this study, the following future directions can be highlighted: i. Expansion of genetically engineered microorganisms for tailored applications. ii. Real-time signal processing improvements with MATLAB/SIMSCAPE-simulated empirical algorithms. iii. Focus on optimising the bioenergy capacity of the MFC unit via integration with power boosting electrical components towards meeting the national grid connection IEEE standards for the practicality and applicability of this technology in fighting the current energy and freshwater scarcity in South Africa. This paper establishes DCMFC-organic pollutant-based unit as a promising technology for environmental pollution monitoring and treatment while generating bioelectricity as a multifaceted approach. It highlights the need for ongoing optimisation to address current challenges commercial scale bioenergy production and practical application.\n\n4.1. Conclusions In the quest to ensure compliance with stringent industrial wastewater quality standards, the adoption of effective treatment and advanced dewatering technologies is imperative. Among these technologies, the Bioelectrochemical System (BES), specifically the Microbial Fuel Cells (MFCs) system, has exhibited remarkable reliability, sustainability, and renewability in addressing the challenges posed by high-strength complex industrial substrates and contaminants, ultimately yielding bioelectricity [ 24 , 56 , 57 , 58 ]. The domain of Microbial Fuel Cells (MFCs) is rapidly evolving and demands the development of reliable and scalable commercial voltage generation units [ 58 , 59 , 60 , 61 , 62 ]. Achieving optimal MFC operation necessitates a multifaceted approach encompassing scientific knowledge, wet chemistry, microbial studies, and electro-engineering aspects, all tailored toward enhancing voltage yield generation [ 23 , 42 , 61 , 63 , 64 , 65 ]. This study underscores the critical importance of identifying the most dependable and readily available raw wastewater samples for use as anolytes within the MFC system. Moreover, the necessity for well-cultured and active microbial consortia to facilitate the biodegradation of complex organic compounds into electrical energy has been underscored [ 61 , 66 , 67 , 68 , 69 ]. Based on the taxonomical classification, along with the concurrent analysis of small RNA (sRNA) and ribosomal DNA (rDNA) identification from the three complex wastewater sources, the following conclusions have emerged: Biorefinery wastewater exhibits substantial conductivity capacity, elevating conductivity levels to approximately 12,000 µS·cm, making it the most saline raw water source. It yielded an overall voltage of approximately 230 mV when processed from pure and untreated biorefinery wastewater. The mixed stream wastewater source ranks second in terms of conductivity and salinity reliability. Remarkably, it recorded the highest voltage yield of 76 mV when processed from pure substrates. Clover wastewater, on the other hand, is situated at the lower end of the taxonomical classification, yielding a mere 1–5 mV when subjected to raw anolyte feeding. This result is correlated with its notably high phosphate content, which appears to inhibit the bioelectricity generation process, [ 18 , 40 , 74 , 75 , 76 ]. These findings highlight the potential of different wastewater sources for bioelectricity generation in MFCs, with biorefinery and mixed stream wastewater sources showing promise for further exploration and optimisation in scaling up voltage yield generation.",
"introduction": "1. Introduction Industrial wastewater, a byproduct of virtually all manufacturing processes, often contains chemical contaminants and reagents, leading to significant environmental pollution and adverse effects on human health when discharged untreated into surface water bodies [ 1 , 2 ]. Dairy wastewater, characterised by discontinuous production processes and significant variability in chemical composition, includes high concentrations of COD, BOD, FOG, nitrogen, and phosphorus, alongside inhibiting cleaning agents [ 1 , 3 , 4 , 5 , 6 , 7 , 8 , 9 ]. While the heterogeneity of dairy wastewater complicates its characterisation, the high-water usage in milk processing generates waste streams with elevated temperatures, variable pH, and sharp fluctuations in contaminant concentrations [ 4 , 5 , 6 , 7 , 8 , 9 ]. Research on wastewater from other dairy sectors remains limited, necessitating further investigation [ 4 , 5 , 6 , 7 , 8 , 9 ]. The sugar industry also produces substantial wastewater during seasonal operations, requiring 1500–2000 L of water per ton of cane crushed, generating approximately 1000 L of wastewater [ 10 , 11 , 12 , 13 ]. Processing stages, including juice extraction, clarification, evaporation, and crystallisation, contribute to significant wastewater volumes, posing environmental challenges [ 10 , 11 , 12 , 13 ]. Addressing these challenges, Microbial Fuel Cell (MFC) technology offers a bioelectrochemical approach for treating complex wastewater substrates while simultaneously generating electricity. MFCs efficiently biodegrade organic pollutants (e.g., COD, BOD, TOC, and TSS) using diverse microbial consortia with unique metabolic capabilities [ 14 ]. Certain organic-rich substrates, such as food processing wastewater and swine manure, enhance microbial growth and bioelectrochemical activity [ 14 ]. MFC technology not only reduces energy demands and sludge production compared to conventional anaerobic digestion but also produces valuable byproducts [ 1 ]. This dual functionality has garnered increasing attention for sustainable wastewater treatment and renewable energy production [ 1 , 14 , 15 , 16 ]. However, to commercialise MFC technology, further improvements in performance, efficiency, and scalability are necessary, along with identifying viable inoculants and substrates to enhance bioelectrochemical processes [ 1 , 14 , 15 , 16 ]. This study focuses on taxonomically classifying three distinct wastewater sources based on their organic substrate complexity and pollutant strength to assess their viability as electron donors for MFCs. Key wastewater parameters, including electrical conductivity, salinity, total dissolved solids, resistivity, oxidation-reduction potential, and organic pollutant concentrations, were analysed to evaluate their suitability for MFC applications. By characterising pollutant capacities and correlating them with energy generation potential, the study aims to optimise MFC performance and maximise bioenergy production. The research also explores the thermodynamic relationship between pollutant composition and electricity generation efficiency, using the Gibbs free energy principles to elucidate energy conversion mechanisms in MFCs. Wastewater streams will be inoculated into double-chamber H-type MFCs to assess substrate removal and bioelectricity production, identifying the most reliable anolyte among the studied organic wastewater sources. This novel approach leverages pure industrial and organic wastewater substrates as electrogenic sources, advancing MFC technology for wastewater treatment and renewable energy applications. This study provides new insights into the taxonomic composition and electrochemical properties of distinct wastewater sources, linking their pollutant profiles to energy generation potential in a double chamber Microbial Fuel Cell (DCMFCs). It introduces a novel application of the Gibbs free energy principles to analyse the thermodynamic relationship between pollutant composition and bioelectricity production efficiency. By leveraging pure industrial and organic wastewater substrates as electrogenic sources, the research highlights the potential of these streams for enhancing DCMFC performance and scalability. This dual focus on wastewater treatment and renewable energy production establishes a foundation for optimising bioelectrochemical processes in real-world applications.",
"discussion": "3. Results and Discussions 3.1. Statistical Classification of the Three Wastewater Streams Used in This by Welch Student t-Test and ANOVA Mean Average Test: Biorefinery Wastewater—TH; Dairy Wastewater—(CL) and MIXED-STREAM—(MX) Figure 4 A,B presents the correlation charts that were performed in R Statistical software as part of the student’s t -test method with 95% confidence levels in comparison of both the physico and organic industrial wastewater complex substrates. Figure 4 A,B, presents a statistical correlation analysis of key physicochemical parameters related to wastewater quality and their potential implications for bioelectricity production in Microbial Fuel Cells (MFCs). Panel A focuses on the relationships among total organic carbon (TOC), chemical oxygen demand (COD), and phosphates ( P O 4 3 − ), while Panel B highlights correlations among salinity (Sal), dissolved oxygen (DO), electrical conductivity (EC), total dissolved solids (TDS), and resistivity (Res). Significant correlations are indicated by asterisks (***), and the red regression lines demonstrate linear relationships. In R statistical analysis, the number of asterisks next to a p -value in the output of a statistical test or regression summary indicates the level of statistical significance. These asterisks correspond to specific thresholds for the p -value: Asterix ( * **) Indicates a very high level of significance ( p < 0.001). Figure 4 A shows TOC, COD, and Phosphates: TOC and COD (r = 0.81, p < 0.001): A strong positive correlation suggests that TOC levels are proportional to COD, indicating that wastewater with higher organic carbon content tends to have higher chemical oxygen demand. This relationship highlights the potential of TOC-rich wastewater as a substrate for MFCs, as both TOC and COD serve as indicators of available organic matter for microbial metabolism and subsequent electricity generation. TOC and PO (r = 0.35, p < 0.001): A weaker positive correlation between TOC and phosphate concentrations suggests that while phosphates are present in wastewater, they do not scale proportionally with organic carbon content. Phosphates, as a nutrient source, could enhance microbial growth but are less directly linked to bioelectricity production compared to organic carbon. COD and ( P O 4 3 − ), (r = 0.31, p < 0.001): Like TOC and ( P O 4 3 − ), the weak correlation suggests that COD is only moderately related to phosphate levels. This implies that phosphate concentrations may not directly limit or drive COD biodegradation during MFC operations. Figure 4 B shows Salinity, DO, EC, TDS, and Resistivity: Salinity and EC (r = 0.96, p < 0.001) and Salinity and TDS (r = 0.95, p < 0.001): These strong positive correlations indicate that higher salinity corresponds to increased electrical conductivity and dissolved solids, which are critical parameters for ion transport within the MFC. Elevated salinity and conductivity enhance the ionic strength of the electrolyte, improving electron flow and energy recovery efficiency. Salinity and Resistivity (r = −0.84, p < 0.001): The strong negative correlation between salinity and resistivity underscores the inverse relationship, as higher salinity reduces resistivity. This is advantageous for MFC performance, as lower resistivity minimises internal resistance and improves power output. EC and TDS (r = 0.97, p < 0.001): The nearly perfect correlation suggests that TDS predominantly governs the electrical conductivity of wastewater. This reinforces the importance of TDS as a key parameter for assessing the suitability of wastewater as an MFC substrate. DO and other parameters (weak correlations): The relatively weak relationships between dissolved oxygen and other parameters suggest limited dependence of DO on salinity, conductivity, or TDS. Since MFCs operate in anoxic conditions, the influence of DO on bioelectricity production may be minimal compared to other physicochemical factors. Implications for bioelectricity production in DCMFCs: Strong correlations between TOC and COD confirm the potential of these parameters as predictors of organic substrate availability for microbial metabolism. Higher TOC and COD levels are desirable for enhanced electricity generation. Strong correlations among salinity, EC, and TDS highlight the importance of ionic strength in optimising electron transfer and reducing internal resistance in the MFC [ 19 , 20 , 21 , 26 , 27 , 28 , 29 , 30 , 31 ]. Wastewater with higher salinity and conductivity is better suited for MFC applications. Phosphate levels, though weakly correlated with TOC and COD, remain essential for microbial growth and activity, indirectly supporting bioelectricity production [ 32 , 33 , 34 , 35 , 36 , 37 ]. The inverse relationship between salinity and resistivity emphasises the need for low-resistivity substrates to maximise MFC efficiency. Overall, this statistical analysis underscores the critical role of organic matter (TOC, COD) and electrolyte conductivity (salinity, EC, TDS) in determining the feasibility of wastewater as an MFC substrate for bioelectricity production. Tables S1 and S2 in the Supplementary section present the Welch Student’s t -test methods used to statistically validate significant differences between the mean average values sourced from the three wastewater streams studied: biorefinery, dairy, and mixed 50% ( v / v ) (biorefinery-dairy) wastewater streams. The Welch Student’s t-test analysis was conducted to compare the basic physicochemical parameters commonly referred to as pollutant strengths or complex substrate compositions in industrial wastewater [ 1 , 14 , 15 , 16 , 24 , 25 , 38 ]. The objective is to discern significant differences among these wastewater streams and identify the stream with higher and more statistically significant pollutants as a viable source of electrons and proper electron donor or anolyte for a bioelectrochemical technology process to generate bioelectricity in a lab-scale Microbial Fuel Cell (MFC). Precisely, the mean average statistical significances presented in Tables S1 and S2 are derived from thorough statistical analysis using the Welch two-sample Student’s t-test method and 95% confidence level comparison between all wastewater streams for physicochemical parameters [ 1 , 14 , 24 , 25 , 38 ]. This analysis presents mean average values, p -values, t-values, 95% confidence interval levels, and sample mean differences when pairing biorefinery-mixed wastewater, dairy-biorefinery, and dairy-mixed wastewater streams. A concluding statement or overview of the statistical regression under the empirical hypothesis that the true difference in mean average values between sample groups listed above for both organic and physicochemical parameters is not equal to zero [ 25 ]. For most parameters, the p -values, discussed based on a set threshold value of 1.02 × 10 −13 p ≤ 0.05, showed strong significance when observed between all groups, as presented in Tables S1 and S2 . Additionally, the coefficient of variation (CV), a statistical measure of the dispersion of data points in a data series around the mean, was analysed. The coefficient of variation represents the ratio of the standard deviation to the mean, making it a useful statistic for comparing the degree of variation from one data series to another, even if the means differ drastically from one another [ 16 , 24 , 25 ]. Based on the findings in Tables S1 and S2 , a general hypothesis is proposed that the magnitude of variability of the paired groups’ data was on average low, ranging from 0.43 ≤ CV ≤ 1.473. These low values suggest a very solid observation that the magnitude of the standard deviation to the mean value shown by the CV is minimal. This implies that the average sample data are very close to the mean average value, essentially presenting very small deviations or variability from the mean average data value. Empirically, the data demonstrate a strong confidence that the sample populations for all the groups are not dispersed far from the mean values. Consequently, these values can be reliably used to simulate the chemical or organic matter contained in these complex raw water substrates to predict the amount of bioenergy that can be produced in the form of bioelectricity. Table S3 in the Supplementary section , presents the Analysis of Variation (ANOVA) between sample groups: biorefinery-dairy; biorefinery-mixed wastewater and dairy- mixed wastewater. This test was performed precisely for the physicochemical parameters: Salinity, Dissolved Oxygen, Electrical Conductivity, Total Dissolved Solids, and Resistivity. A Tukey multiple comparisons of mean average values approach was adopted according to Mutombo et al. [ 24 ]. The findings in Table S3 support the statistical hypothesis, indicating a 95% confidence interval for the validity and dispersion of the data based on the sample mean for each population group, be as it may for dairy wastewater, biorefinery wastewater, and the mixed raw wastewater organic substrates. The values presented in Table S3 can be attained at any point of the wastewater treatment plant and any time of the week or process operation. These physicochemical parameter values confirm the conductivity and suitability of all three wastewater organic substrates as anolytes and active electron donors, highlighting their potential as reliable and sustainable sources of bioelectricity in bioelectrochemical systems (BES) precisely the DCMFC system for this study [ 8 , 9 ]. The overall observation of the p -value across the comparison groups was 0.000 ≤ p ≤ 0.166. This p -value presents strong significance of the organic strength of these raw organic industrial wastewater substrates [ 1 , 14 , 16 , 24 , 25 , 38 ]. 3.2. Taxonomy and Characterisation of the 3-Complex Substrates Based on Organic Constituents as Viable Electron Donor Before MFC Treatment Stage 3.2.1. Total Organic Carbon (TOC) vs. Chemical Oxygen Demand (COD) Profile for All Wastewater Streams Harvested from the Wastewater Treatment Plant Figure 5 a–c scientifically proves the strong correlation between total organic carbon and chemical oxygen demand as per the principle of biochemistry and wet chemistry for raw organic substrates principally carbon enriched sources. The literature has reported that a correlation between TOC and COD is normally depicted in the form of linear empirical model [ 19 , 20 , 21 , 39 ]: (8) T o t a l O r g a n i c C a r b o n T O C m g T O C L = C O D m g C O D L + 49.2 3 Total organic carbon (TOC) and chemical oxygen demand (COD) monitoring serve as established standards for assessing water quality, both at the point of water injection and treatment [ 39 ]. While TOC measures the level of organic molecules or pollutants in water analytically, COD provides a measurement relating to virtually all degradable carbon present in wastewater, [ 40 ]. Ongoing advances in the precision and sensitivity of monitoring technologies play a pivotal role in understanding this emerging challenge [ 9 , 19 , 20 , 21 , 26 , 39 ]. The findings of this study clearly demonstrate a linear relationship between TOC and COD, with a strong correlation indicated by the root mean factor for the three different sources of organic substrates. For instance, Figure 5 a illustrates the dairy wastewater effluent’s R 2 mean factor of 0.989, signifying a strong correlation between TOC and COD. An empirical polynomial correlation model of y = 0.0003x 2 + 2.6426x − 50.633 was derived to depict the significant relationship between TOC and COD. This model aligns with the theoretical principle that the TOC magnitude in organic industrial wastewater is typically approximately half the composition of COD within the system [ 19 , 39 ], a principle validated by the study. Additionally, dairy wastewater presents a significant amount of chemical energy that could be biodegradable in the form of cATP and chemically converted into bioenergy, provided there are adequate active microorganisms to carry out the biodegradation and electron donation process in bioelectrochemistry within a suitable bioelectrochemical technology such as the MFC system. Figure 5 b for biorefinery wastewater effluent attained a R 2 mean factor of 0.9588 strong correlation significance between TOC and COD. An empirical polynomial correlation model of y = −3 × 10 −5 x 2 + 2.9965x − 37.649 was attained in relation to the strong significance between TOC and COD organic substrates. This model is in line with the theoretical principle that the magnitude of the TOC contained in the complex organic industrial wastewater is always almost half the of the composition of COD within the system [ 19 , 20 , 21 , 39 ]. The biorefinery stream articulated a good source of chemical energy and potential for being a reliable source of electron donor in the DCMFC system for sustainable energy production. The complex waste matter will be simply harnessed and converted into bioelectricity through the bioelectrochemical principle reported by Logan et al. [ 1 , 14 , 15 , 16 , 20 , 25 , 38 , 41 ]. Figure 5 c presented the mixed stream 50% ( v / v ) (dairy and biorefinery wastewater effluent systems). The results attained relates to a R 2 mean factor of 0.693 good correlation significance between TOC and COD. This aspect is scientifically validated by the nature of this wastewater stream. An empirical linear correlation model of y = 2.4768x + 196.74 was attained in relation to the strong significance between TOC and COD organic substrates. This aspect proves that a fraction of COD content in the raw wastewater source correlates to a certain magnitude to the organic composition of TOC in the effluent stream [ 19 , 20 , 21 , 26 , 39 ]. This model is in line with the principle that the magnitude of the TOC contained in the organic substrate is always almost half the of the composition of COD [ 19 ]. 3.2.2. Salinity Taxonomical Classification and Profile for All Wastewater Streams, Harvested from the Wastewater Treatment Plant The resistivity of wastewater reflects the ability of wastewater to efficiently resist an electrical current [ 19 , 20 , 21 ] in a typical MFC system as a viable anolyte. This section scrutinises resistivity of the three wastewater sources and how it relates to viable MFC applications. There is a close link between conductivity and resistivity [ 19 ]. While conductivity is a measurement of how well electrical current can flow through wastewater, resistivity is a measurement of how well wastewater can resist electricity flow [ 19 ]. An increase in salinity results in an increase in conductivity due to dissolved salts that tend to exude an electrical current [ 19 , 21 ]. The findings presented in the Supplementary section by Figure S1a shows a strong opposing mechanism between salinity and resistivity, as stated above. A clear characteristic potential is observed between the above streams that dairy wastewater exudes high resistivities and moderate salinity concentrations, referenced in Figure S1a . The highest salinity of 8 ppt and resistivity of 0.0006 mΩ.cm was captured in the dairy wastewater stream referenced by Figure S1a . Contrary, Figure S1b presented high salinity assays of 10 ppt and low resistivities of 0.0001 mΩ.cm. The biorefinery stream exudes the characteristics of a viable anolyte or electron donor in the MFC operation for the production electrical energy, according to Eddie and Metcalf [ 19 ]. The concept is validated further by the comparison curves in Figure 6 a,b, respectively. Figure 6 b presents the high salinity content of a biorefinery wastewater stream as compared to the rest; likewise, Figure 6 a clearly shows that biorefinery wastewater poses the least resistance in terms of electron flow sequence with the circuit, hence less resistivity capacity. Precisely, biorefinery seems as a recommendable MFC anolyte for optimised and scaled up MFC electricity production and practical applications with non-exogenous organic complex substrates source. 3.2.3. ORP Taxonomical Profiles for All Streams Harvested from Wastewater Treatment Plant Khumalo et al. [ 42 ] emphasises the crucial role of oxidation-reduction potential (ORP) in determining the electrical conductivity of wastewater, which is pivotal for identifying suitable electron donors or anolyte streams in bioelectrochemical systems (BES) for bioelectric energy generation. Additionally, Khumalo et al. [ 27 ] elaborate on the correlation between pH and ORP, indicating the significance of pH levels within the context of conductivity and electron donation [ 29 , 30 , 31 , 43 , 44 ]. The findings presented in Figure 7 a–c confirm that all the investigated wastewater streams exhibit favourable electrical conductivity properties, aligning with established thresholds for viable electron donors in MFC technology [ 19 , 27 , 29 , 31 , 43 , 44 , 45 ]. Notably, the optimal pH range for these wastewater streams was observed to be between 6 and 10, accompanied by corresponding ORP values ranging from −300 to −50. This pH-ORP combination signifies an ideal electro-potential conducive to effective electron donation, which is essential for electrical generation in BES and MFC units [ 27 , 28 ]. 3.2.4. Taxonomic Classification of Electrical Conductivity (EC) Profiles for All Wastewater Streams The conductivity of water is influenced by temperature, as indicated by thermodynamic principles [ 21 ]. Temperature has a direct impact on the solubility constant of a solution, affecting the overall solubility of minerals present [ 44 , 45 ]. This relationship is particularly relevant in the context of the three distinct wastewater sources examined in this study. An increase in temperature generally enhances solubility, hence positively influencing the conductivity of the liquid [ 27 , 32 , 45 , 46 ]. It is important to note that the conductivity of wastewater, which is closely linked to its total dissolved solids, varies across different temperature ranges, such as psychrophilic, mesophilic, or thermophilic conditions [ 27 , 33 ]. Future investigations will explore the specific effects of psychrophilic and mesophilic temperature regimes on electricity production in Double Chamber Microbial Fuel Cells (DCMFCs), utilising Proteobacteria and Bacteriodota as biocatalysts. These studies aim to provide comprehensive insights into this phenomenon, building upon the existing literature [ 19 , 20 , 21 ]. Elevated conductivity levels are commonly observed at higher temperatures [ 19 ]. Based on the findings presented in the Figure 8 comparisons graph for electrical conductivity (EC) for all organic wastewater streams, biorefinery exudes the most viable capacities of all streams considering these vital electrochemical parameters imperative as a perfect bioelectrochemical generation source in the DCMFC unit. The precise range, respectively, for EC for dairy, biorefinery, and mixed wastewater stream chronologically were: 1300 ≤ EC ≤ 8000 mS·cm 2 ; 4000 ≤ EC ≤ 12,200 mS·cm 2 . The salinity levels, as presented in the above sections, also favoured high electrochemical potentials of the biorefinery wastewater stream hence validating its feasibility as a recommended anolyte for reliable bioelectricity generation in a DCMFC in this study, [ 1 , 2 , 10 , 12 , 13 , 19 , 21 ]. 3.2.5. Comparison of the Present Study Wastewater Classification Profiles of Other Studies Dairy wastewater composition is commonly milk processing effluents that have an increased temperature and large variations in pH, TSS, biological oxygen demand (BOD), chemical oxygen demand (COD), total nitrogen (TN), total phosphorus (TP), and fat, oil, and grease (FOG) [ 4 , 5 , 6 , 7 , 8 , 9 ]. There is little information on industrial-scale dairy effluent composition, [ 9 ]. The information on the general characteristics of dairy wastewater, biorefinery wastewater and mixed wastewater stream is shown in Table S4 in the Supplementary section . Typically, dairy wastewater is white in colour (whey is yellowish green) and has an unpleasant odour and turbid character [ 2 , 10 , 27 ]. With annual temperatures of 17–25 °C, dairy waste streams are warmer than municipal wastewater (10–20 °C) [ 7 , 8 ], which results in faster biological degradation compared to sewage treatment plants [ 28 ]. The average temperatures of industrial dairy effluents range from 17 to 18 °C in winter and 22–25 °C in summer [ 6 ]. Using the Arrhenius model, as in the literature, the biodegradation rates and oxygen consumption can be predicted to be 1.5 times higher in summer than in winter [ 6 , 7 , 8 , 9 ]. The design winter temperature of 15 °C is adopted for this type of wastewater due to the utilisation of hot water for washing and cleaning of equipment [ 6 , 7 , 8 , 9 , 10 , 11 , 12 , 13 , 17 , 18 , 22 , 40 ]. In the sugar industry, water is used for cleaning purposes in the different sections of the factory generates wastewater [ 12 , 33 ]. Practically, there are no single units that generate wastewater, but the wastewater is produced mainly by washing on the milling house floor, boiling house like evaporators, clarifiers, vacuum pans, centrifugation [ 33 , 34 ]. Periodic cleaning of lime water and SO₂ production facilities also significantly contributes to the large volume of wastewater, as well as periodic descaling of heat exchangers and evaporators using NaOH, Na 2 CO 3 , and HCl for descaling of heater and neutralisation [ 33 ]. Precisely, mill houses, and process houses are the two main sections of wastewater generated in sugar factories [ 33 ]. The mill house wastewater is polluted mainly with oil, grease, and suspended solids; whereas the wastewater generated from the process house is contaminated with high organic matter such as COD, BOD 5 , and pH [ 12 , 33 ]. Studies of physicochemical properties of the sugar industrial effluent, dairy industrial effluent, and the mixed stream industrial effluent has been sampled and collected from local wastewater treatment plants and analysed as presented in the supplementary summative Table S4 . The detailed values of both physicochemical and organic parameters indicated that the effluent pollutant qualities and quantities are quite different [ 12 , 33 ]. Precisely, looking at the common organic real time process plant data monitoring parameters that were investigated to characterise the wastewater pollutant strength in this study, listing the few critical ones: COD, BOD, TOC, EC, and ORP for the current studied streams, comparatively to previous authors work, the same range of results has been attained. The results align with expectations based on reputable literature [ 12 , 33 ], with mixed wastewater streams presenting high COD and BOD levels due to their organic content [ 33 ]. TOC analyses were unique to this study, since most researchers rely on COD for real-time monitoring of wastewater treatment plant performance. The biodegradability ratios of COD: BOD, respectively, closely related for the current study streams: 1.66 ± 1 and previous authors with 1.66 and 1.85, respectively. This current study also thoroughly covered the characteristic physico parameters to present the bioelectrochemical capabilities of these streams. That has been thoroughly digressed in the previous sections and clearly recommended for BES processes application especially in the generation of electricity in MFC processes. 3.3. Taxonomy on Biomass rDNA Sequencing and Analysis of Phylum’s Class as Viable Bioelectrochemical Inoculate, After Wastewater Treament Palnt Harvest 3.3.1. Biorefinery (Tongaat Hullet) Biomass–Phylum Classification Blueprint for Sugar Biorefinery Biomass from Wastewater Treatment Plant This section of the manuscript contains the metagenomics analysis of full-length 16s gene amplicons. As aforementioned, samples were sequenced on the Sequel system by PacBio ( www.pacb.com ). Raw sub-reads were processed through the SMRTlink (v11.0) Circular Consensus Sequences (CCS) algorithm to produce highly accurate reads (>QV40). These highly accurate reads were processed through DADA2 ( https://benjjneb.github.io/dada2/index.html ) and qiime2 ( https://docs.qiime2.org/2021.11/ ) for quality control assessment and taxonomic classification, respectively. Create_vsearch_single_sample_pdf_report_pacbio.py 0732022_M13_bc1004_F-M13_bc1068_R.hifi_reads-filtered-feature-table-asv.tsv-M13_bc1004_FM13_bc1068_R0732022 220705_Cell1 16S-QIMME2. Table S1 presents the full layout of the Top phylum family classification of the Biorefinery biomass sample. 3.3.2. Biorefinery (Sugar Mill) Biomass Taxonomical Graphical Classifications The detailed genomic analysis was carried out through a series of taxonomical sequences as prelisted: Top-Phylum; Top-Class Classification; Top-Order Classification; Top-Family Classification then finally the Top-Genus Classification on Inqaba Biotec full-length 16s metagenomics report—Sample 3_CloverBiomass_0732022 as presented in Figure 9 . The top phylum graphical layout of the bacterial colony’s proportions in the biorefinery sample which correlates with Table S4 (Supplementary Material) was considered for bacterial colonies taxonomical classification. Organisms are grouped into taxa (singular: taxon), and these groups are given a taxonomic rank; groups of a given rank can be aggregated to form a more inclusive group of higher rank, thus creating a taxonomic hierarchy. Precisely, the taxonomy of this bacterial consortium is based on the top phylum graphical classification of the bacterial population, which correlates with the Table S1 family phylum classification, as attached in the Supplementary Materials , which shows a wide margin dominance of Proteobacteria by a read count index ratio of 46.66%. Firmicutes and Bacteriodota follow up sequentially at 22.98% and 7.36%, respectively. A precise observation is that Proteobacteria the dominant species has been applicable in some bio-electrochemical technologies reported by Hossain et al. [ 35 ]. Firmicutes phylum is historically related to Geobacter [ 11 , 13 ] and Bacteroidetes biologically and is based on its genomic and rDNA analysis, as reported by [ 1 , 34 , 35 , 36 , 37 , 47 , 48 , 49 ]. This species has been widely reported as a viable electrogene in the generation of bioelectricity in MFCs and other METs, as reported by Logan et al. [ 36 , 37 , 50 , 51 , 52 ]. Firmicutes and Bacteriodota have also been reported by Logan et al. [ 34 , 48 , 52 , 53 ] as a good source of electrons or viable anolytes and are efficient for high-strength organic biodegradation and good biological catalyst for the production bioelectricity in an MFC technology. Automatic identification of the characteristic peaks in the Energy Dispersive X-ray Spectrum (EDS) is a valuable software tool that has been progressively developed with the rise in computing power and speed [ 35 , 36 , 48 , 49 ]. The EDS is modern sophistication tool, commonly used for commercial automatic peak identification procedures that are frequently used in the labelling of high intensity peaks that correspond to major constituents [ 36 , 48 , 49 ]. Based on the Supplementary Material images presented by Figures S2–S4 ; the EDS pick analysis precisely corresponds to the SEM scans and images that relate a smooth biomass sample with visible patches of white and milky way phases in both scan images, Figure S2 . However, the scale bar can be used to possibly identify microstructures and microorganisms, based on the overall size. These SEM images precisely on the red circled areas relates to the EDS peaks analysis that conveys a low peak energy dispersive analysis on the overall biomass sample content, as presented in Figure S2 . The lack of calcium on this biorefinery biomass sample will not inhibit the overall process of being a carbon enriched source to facilitate the organic pollutants biodegradation in a typical BES, in the MFC system, referenced by Figure S3 . The energy content peaks of carbon in both samples 1 and 2 images shot up to more than 25 ≤ cps/ev ≤ 40 and an average of 51 wt. % of the mass composition out of 100% total sample population, which makes it a mass fraction of 0.51 (kg·C/Kg·Biomass·Sample), as presented in Figure S4 . A basic scientific hypothesis further relating to Figure S4 articulates that the elemental mapping of a biomass sample using Energy Dispersive X-ray Spectroscopy (EDS) for various key elements: Ca, O, C, S, Mg, K, and Cu. These elements are crucial indicators of the bioelectrochemical efficiency of biomass samples, particularly for Microbial Fuel Cell (MFC) applications. Calcium and magnesium are essential for microbial growth and biofilm formation, which enhance microbial activity in MFCs. Their presence suggests that the biomass supports microbial colonisation and stability. The oxygen mapping indicates potential electron acceptor sites or oxidative environments. However, in anaerobic regions (typical for MFC anodic chambers), oxygen levels would need to be limited for effective electron transfer [ 36 , 37 , 47 , 48 , 49 ]. According to Logan et al. [ 54 ] and Sarkar et al. [ 4 , 5 , 6 , 7 , 8 , 9 ], the abundant presence of carbon shows the availability of organic matter, which is vital as an electron donor. Microorganisms can metabolise this carbon to produce electrons for bioelectricity generation in MFCs. Sulphur compounds are often involved in microbial metabolism, particularly for bacteria capable of sulphur reduction. This enhances the variety of metabolic pathways and can contribute to electron donation. Potassium is a vital element for maintaining cellular functions, while copper can act as a catalyst in electron transport processes. The distribution of these elements suggests that the biomass may also provide favourable conditions for electron transfer reactions. In terms of the biorefinery sample viability, the elemental distribution, especially carbon, sulphur, and other nutrients, indicates that the biomass is rich in organic matter and supportive of microbial metabolic activity. These properties enhance the biomass’s capacity as an electron donor in an MFC, thus improving its bioelectrochemical performance. This analysis implies that the biomass sample could serve as a viable feedstock for sustainable energy generation via microbial-driven processes in MFC systems. 3.3.3. Dairy (Clover) Biomass—Phylum Classification and Taxonomical Blueprint, After Harvest from Wastewater Treatment Plant Furthermore, this section of the article summarises the metagenomic analysis of full-length 16s gene amplicons of the biomass samples that have been clinically disseminated as per the previous section—biorefinery genomic analysis. Samples were sequenced on the Sequel system by PacBio. Raw sub-reads were processed through the SMRTlink (v11.0) Circular Consensus Sequences (CCS) algorithm to produce highly accurate reads (>QV40). These highly accurate reads were processed through DADA2 ( https://benjjneb.github.io/dada2/index.html ) and qiime2 ( https://docs.qiime2.org/2021.11/ ) for quality control assessment and taxonomic classification, respectively. create_vsearch_single_sample_pdf_report_pacbio.py-3_Clover_Biomass_1432022_M13_bc1004_F--M13_bc1070_R.hifi_reads-filtered-feature-table-asv.tsv-M13_bc1004_F--M13_bc1070_R 3_Clover_Biomass_1432022 220705_Cell1 16S-QIMME2. Table S4 shows the thorough phylum classification of the dairy biomass. This genomic analysis precisely presents the phylum taxonomical array of the heterotrophic bacterial species, a precise read count, and its ratio analysis. Sample 1432022, full-length 16s metagonomical report by Inqaba Biotec laboratories: The taxo-classification of the dairy wastewater bacterial colonies was carried out through a various classification hierarchy as prelisted: The Top-phylum classification was based on Inqaba Biotec full-length 16s metagenomics report—Sample 3_Clover_Biomass_1432022; the Top-Class Classification was based on Inqaba Biotec full-length 16s metagenomics report—Sample 3_Clover_Biomass_1432022; the Top-Order Classification was based on Inqaba Biotec full-length 16s metagenomics report—Sample 3_Clover_Biomass_1432022; the Top-Family Classification, then finally the Top-Genus Classification on Inqaba Biotec full-length 16s metagenomics report—Sample 3_Clover Biomass_1432022. For the simplicity and reliability of this study, the top-phylum taxonomy was considered to analyse the bacterial dominance of the dairy wastewater harvested biomass sample, as presented above in Figure 10 . Also, Table S5 in the supplementary section presents the top phylum classification of the bacterial population, which correlates with the graphical family phylum classification, as seen in the figure below. A precise observation that the Lactobacillus bacterial species comprising classification counts and 77.36% of the read counts index was the most dominant bacterial species for this family population [ 17 ]. Sequentially and second to the hierarchy of bacterial dominance is Clostridium, at 15.70%. The chloroflexi phylum is vastly related to anaerolineceae, as reported by Liang et al. [ 54 ]. Anaerolineceae phylum comprise obligate anaerobes, as a majority in alkali-degrading bacterial colonies, as stated by Logan et al. [ 54 ]. Anearolineacea may be associated with the anaerobic degradation of oil-related compounds—this bacterial lineage was also reported as the most frequently encountered bacteria taxon in anaerobic degradation [ 54 , 55 ]. Lactobacillus has been reported by Sarkar et al. [ 4 , 5 , 6 , 7 , 8 , 9 ] as a common species in the dairy wastewater solid waste and effluent system. A concise understanding is based on the background and origin of the dairy sample [ 7 , 9 ]. Dairy waste organic matter and suspended solid masses essential disintegrates to forms these actively classified biomass genus [ 3 , 4 , 6 , 17 ]. Since it is a dairy based sample, it will have to be lactose based and, therefore, the Lactobacillus proved dominant in the metagenomic genus dairy based sample [ 8 , 9 ]. 3.3.4. Dairy Biomass Samples 1 and 2 FEG SEM-EDX Analysis via Zeiss Ultra A separate significant issue is the reliability of elemental identification in the EDS [ 35 , 36 , 48 , 49 ]. Automatic identification of the characteristic peaks in the Energy Dispersive X-ray Spectrum (EDS) is a valuable software tool that has been progressively developed with the rise in computing power and speed [ 36 ]. The EDS is a modern sophistication tool, commonly used for commercial automatic peak identification procedures that are frequently used in the labelling of high intensity peaks that correspond to major constituents [ 36 , 48 , 49 ]. Based on the Supplementary Material images, as presented by Figures S5–S7 (Supplementary Section) ; precisely the EDS pick analysis corresponds to the SEM scans and images that relate a smooth biomass sample with visible patches of white and milky way phases in both scan images, Figure S7 . However, the scale bar can be used to possibly identify microstructures and microorganisms, based on the overall size. SEM images highlight circled areas corresponding to EDS peak analysis, indicating low peak energy dispersive values for the overall biomass sample, as shown in Figure S5 . Figure S6 intricately presents SEM images of two dairy based samples (Clover 1 and Clover 2) with highlighted areas representing notable surface structures. These structures can be associated with the biofilm formation and microbial colonisation critical for bioelectrochemical activity in Microbial Fuel Cells (MFCs). The observed surface roughness and specific textures likely indicate regions where microorganisms may anchor, enhancing electron transfer processes between the microbial community and the anode. Such biofilm formations are integral to efficient energy generation in MFCs. The variations in surface characteristics between dairy biomass sample 1 and 2 could imply differences in microbial adhesion or electron transfer efficiency, potentially affecting their performance as MFC inoculants [ 36 ]. Further investigation of these samples could confirm their suitability for bioelectrochemical applications, particularly in wastewater treatment and renewable energy generation. The energy content peaks of carbon in both sample 1 and 2 images shot up to more than 40 cps/ev and more than 77% mass composition out of 100% total sample population which makes it a mass fraction of 0.77 (kg·C/Kg·Biomass·Sample), as presented in Figure S7 . 3.3.5. Comparison of the Three-Wastewater Substrates and Their Dominant Phylum for DCMFC Treatment and Production of Bioelectricity Ensuring discharged industrial wastewater meets the required quality standards relies on effective treatment and advanced dewatering technologies, e.g., the BES technology; the MFC system has proven reliability, sustainability, and renewability towards the efficiency of high strength complex industrial substrates and contaminants subsequently generating bioelectricity [ 24 , 56 , 57 , 58 ]. Microbial Fuel Cells (MFC) research is a rapidly evolving niche area of research that needs reliable and scalable commercial voltage yield generation from this unit. It is still tricky to compare devices on an equivalent basis [ 58 , 59 , 60 , 61 , 62 ]. Proper operation of this unit requires knowledge of various scientific, wet chemistry, microbial studies, and essentially electro-engineering aspects to be fine-tuned towards scaling up voltage yield generation [ 23 , 42 , 61 , 63 , 64 , 65 ]. This study has clearly outlined the need to identify the most reliable and readily available raw wastewater sample as a perfect anolyte for the system. More-so, a need for perfectly cultured and active microbes to facilitate the biodegradation aspects of disintegrating chemical organic energy into electrical energy has been raised [ 61 , 66 , 67 , 68 , 69 ]. In view of the taxonomical classification sourced from the comparative analysis of organic substrate in the start-up sequence for double chamber Microbial Fuel Cell (DCMFC) sourced from Shabangu et al. [ 70 , 71 , 72 , 73 ]. Figure 11 a–c demonstrates intricately the three wastewater streams concurrent to the dissemination of the sRNA and rDNA identification, the subsequent conclusions have been reached; the biorefinery wastewater exudes a massive conductivity capacity, as it shoots up conductivity to around 12,000 uS·cm 2 and is the most saline raw water source. An overall voltage yield of about 230 mV was reached from pure and raw biorefinery wastewater. The mixed stream source comes second in reliability of conductivity and salinity but exuded the highest recorded voltage yield of 76 mV from pure substrates. Clover wastewater was at the bottom of the taxonomical classification with as low as 1–5 mV generation from a purely raw anolyte feed coupled with the highest phosphates contained compared to all sources. The high phosphates contained explains the inhibition of the bioelectricity generation process [ 18 , 40 , 74 , 75 , 76 ]. Based on the findings of Figure 11 a–c, sourced from Shabangu et al. [ 73 , 74 ], biorefinery and mixed wastewater streams are highly recommendable for consideration as reliable anolyte and inoculum in the operation of a typical lab scale unit using purely raw industrial wastewater as a source of electrogeneses and anolyte. The heterotrophs identified and classified in the morphology section states that Protobacteria and Bacteroidetes are recommended as a viable source of electrons when operating the benchtop MFC unit in the subsequent experiments of this study. Biorefinery presented reliability, renewability, and sustainability in terms of being the MFC voltage source supply. Moreso, an organic removal of biodegradable contaminants profile was also shown in Figure 11 a–c. A clear depiction is that in all three of the different wastewater sources. The overall percentage removal was achieved within 72 hours of treatment incubation in the MDC unit. Complete 100% removal was observed with mixed wastewater substrates within 72 h of the treatment time. From a scientific point of view, in particular, a bioelectrochemical and wet chemistry or biological perspective, the longer high percentage organic removal is a function of longer incubation periods even in the MFC unit, which relates positively with convectional traditional wastewater treatment strategies. Logan et al. [ 14 , 15 ] and Kim et al. [ 20 ] have reported corresponding findings to Shabangu et al. [ 66 , 67 ] in numerous studies they completed in the wastewater treatment capacities using traditional H-shaped Microbial Fuel Cells. 3.3.6. Comparative Summary of Findings of Current Paper vs. Previous Studies on MFCs Huang et al. [ 77 ] report advancements in Microbial Fuel Cell (MFC) technology for detecting organic matter, focusing on biochemical oxygen demand (BOD) and chemical oxygen demand (COD). It was gathered that dual-chamber and single-chamber MFCs are prominent, with single-chamber designs offering simplicity and cost-effectiveness but reduced stability. Emerging designs, such as miniaturised, submersible, and coupled MFCs (e.g., wetland-integrated), have expanded application scenarios and improved detection capabilities. However, limitations include the need for stable microbial communities, resistance to fouling, and interference from toxicants or competing electron acceptors. The study highlights the transition of MFCs from prototypes to practical environmental monitoring tools, emphasising design and operational innovations for scalable, reliable, and self-powered systems. Integration with artificial intelligence is proposed to enhance data interpretation and adaptability. Loveccho et al. [ 78 ] present a customised multichannel measurement system for Microbial Fuel Cell (MFC) characterisation, featuring an expandable design capable of simultaneously measuring up to 12 MFCs. The system includes multi-step discharge protocols, long-term measurement capabilities with variable time steps, and calibration procedures to ensure accurate low-current signal detection. Key contributions include bridging the gap between laboratory-grade equipment and cost-effective, scalable tools for MFC research, with robust calibration and multi-channel capabilities suitable for complex systems. This system has the potential to accelerate advancements in renewable energy and wastewater treatment technologies. Limitations include dependence on rigorous calibration, reliance on simulated testing, and scalability challenges for applications beyond 12 MFCs, requiring specialised hardware and software modifications. The current research study explores the potential of a Double Chamber Microbial Fuel Cell (DCMFC) for electricity generation using three distinct wastewater sources—biorefinery, dairy, and mixed streams. It emphasises the importance of substrate selection and microbial community composition for optimising bioelectrochemical processes. Key findings highlight that biorefinery wastewater offers the highest bioelectrochemical potential due to superior electrical conductivity and salinity, achieving a voltage yield of 65 mV, compared to 75.2 mV for mixed streams and 1.7 mV for dairy wastewater. Proteobacteria, Firmicutes, and Bacteriodota were identified as dominant phyla in biorefinery samples, while Lactobacillus and Clostridium were prevalent in dairy samples. Limitations include the inhibitory effects of high phosphate levels in dairy streams and challenges in scalability and microbial stability. As new knowledge contributions, the study bridges the gap between industrial wastewater characterisation and bioelectricity generation, leveraging metagenomics to identify suitable microbial consortia. It establishes correlations between wastewater physicochemical properties (e.g., salinity, conductivity) and energy output, advancing MFC scalability and reliability. Lastly, this study proposes biorefinery wastewater as a viable source for sustainable electricity production and efficient substrate removal. Furthermore, the findings support the use of MFCs in wastewater treatment and renewable, sustainable and reliable bioenergy generation. Challenges such as optimising microbial stability, scalability for viable commercial application, and overcoming substrate-specific limitations like high phosphate interference, specifically in the dairy wastewater organic substrate as an anolyte, are clearly highlighted. Moreover, this work underpins the foundation for future advancements in commercial scale bioelectrochemical systems. The following comparative Table 1 lays the key concepts from the current study comparatively to previous work."
} | 14,750 |
32749905 | null | s2 | 486 | {
"abstract": "Numerous bacterial behaviors are regulated by a cell-density dependent mechanism known as Quorum Sensing (QS). QS relies on communication between bacterial cells using diffusible signaling molecules known as autoinducers. QS regulates physiological processes such as metabolism, virulence, and biofilm formation. Quorum Quenching (QQ) is the inhibition of QS using chemical or enzymatic means to counteract behaviors regulated by QS. We examine the main, diverse QS mechanisms present in bacterial species, with a special emphasis on AHL-mediated QS. We also discuss key Enabled "
} | 145 |
32749905 | null | s2 | 487 | {
"abstract": "Numerous bacterial behaviors are regulated by a cell-density dependent mechanism known as Quorum Sensing (QS). QS relies on communication between bacterial cells using diffusible signaling molecules known as autoinducers. QS regulates physiological processes such as metabolism, virulence, and biofilm formation. Quorum Quenching (QQ) is the inhibition of QS using chemical or enzymatic means to counteract behaviors regulated by QS. We examine the main, diverse QS mechanisms present in bacterial species, with a special emphasis on AHL-mediated QS. We also discuss key Enabled "
} | 145 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.