pmid stringlengths 8 8 | pmcid stringlengths 8 11 ⌀ | source stringclasses 2
values | rank int64 1 9.78k | sections unknown | tokens int64 3 46.7k |
|---|---|---|---|---|---|
38786238 | PMC11121321 | pmc | 4,288 | {
"abstract": "Traditional anti-corrosion and anti-fouling coatings struggle against the harsh marine environment. Our study tackled this by introducing a novel dual-layer hydrogel (A-H DL) coating system. This system combined a Cu 2 O–SiO 2 –acrylic resin primer for anchoring and controlled copper ion release with a dissipative double-network double-anchored hydrogel (DNDAH) boasting superior mechanical strength and anti-biofouling performance. An acrylamide monomer was copolymerized and cross-linked with a coupling agent to form the first irreversible network and first anchoring, providing the DNDAH coating with mechanical strength and structural stability. Alginate gel microspheres (AGMs) grafted with the same coupling agent formed the second reversible network and second anchoring, while coordinating with Cu 2+ released from the primer to form a system buffering Cu 2+ release, enabling long-term antibacterial protection and self-healing capabilities. FTIR, SEM, TEM, and elemental analyses confirmed the composition, morphology, and copper distribution within the A-H DL coating. A marine simulation experiment demonstrated exceptional stability and anti-fouling efficacy. This unique combination of features makes A-H DL a promising solution for diverse marine applications, from ship hulls to aquaculture equipment.",
"conclusion": "3. Conclusions The present study successfully addressed the limitations of traditional marine anti-fouling coatings by developing the innovative A-H DL coating system. This dual-layer design, comprising an AR primer with controlled Cu 2+ release and a DNDAH coating with robust anchoring and self-lubricating properties, demonstrated exceptional performance in marine environments. Extensive testing revealed the A-H DL coating’s superior anti-biofouling capabilities. After 30 days of immersion, the coating exhibited remarkable stability, with no corrosion or microbial attachment. The uniform copper distribution, the impressive protein adsorption resistance, and the significantly reduced microbial attachment confirmed the effectiveness of the system’s anti-fouling mechanisms. In conclusion, the A-H DL coating presents a promising solution for overcoming the challenges of marine anti-fouling. This coating focuses on the problems of environmental friendliness and a short service life in traditional ship coatings, and, by combining these with the characteristics of various new anti-biofouling coatings, it achieves a long-term underwater anti-biofouling performance while reducing pollution of the oceans. Its unique design, superior performance, and potential for environmental benefits make it a valuable addition to the arsenal of marine technologies aimed at protecting our oceans and ensuring the sustainable operation of marine infrastructure. Despite the promising laboratory results, the potential environmental impacts of the copper released by the coating require comprehensive evaluation. Further research and development are also essential to enhancing the A-H DL system and exploring its wider applications across various marine scenarios. Developing a new coating entails various challenges that warrant consideration. For instance, scaling-up experiments may pose construction issues, longer-term durability testing is required, and a comprehensive environmental assessment involving systematic scientific approaches is essential. To address these limitations and refine the A-H DL coating system, future research could focus on optimizing the application methods to overcome scalability issues, conducting extended durability tests under varying environmental conditions, and performing comprehensive environmental impact assessments. Additionally, exploring the underlying mechanisms driving this coating’s anti-biofouling properties through theoretical frameworks or computational modeling could provide valuable insights. Moreover, investigating the integration of novel materials or additives to enhance efficacy and environmental sustainability would further refine the A-H DL coating system for broader applicability in marine environments.",
"introduction": "1. Introduction The maritime water environment presents significant challenges, including chemical corrosion and biological fouling, prompting intensified efforts to develop effective anti-corrosion and anti-fouling solutions for marine vessels [ 1 , 2 , 3 ]. The traditional coatings used in shipbuilding exhibit inherent limitations, such as a susceptibility to microbial adhesion, a rapid depletion of anti-fouling agents [ 4 ], a restricted lifespan [ 5 ], and environmental concerns [ 6 ], underscoring the need for innovative approaches. Nowadays, the research on emerging anti-fouling technologies or coatings is forced in two directions. The first one is replacing the traditional coatings with more environmentally friendly fungicides [ 7 ], and the second one involves biocide-free coatings that exploit surface chemistry to reduce adhesion or dislodge adhered organisms [ 8 ]. Both of those strategies can solve one aspect of the challenges mentioned above. However, they all have some limitations [ 1 ]. For instance, these more environmentally friendly fungicides, such as degradable synthetic organic or natural product-based biocides, have a lower efficiency than the traditional ones, such as copper-based coatings, and the biocide-free coatings with a special surface cannot completely restrain bio-fouling and ensure the mechanical performance. In response to these challenges, hydrogel coatings have emerged as a promising solution [ 9 , 10 , 11 ]. Characterized by an intricate network of interwoven water molecules and polymer chains, hydrogel coatings demonstrate an exceptional durability [ 12 , 13 ]. Specifically, within the realm of marine hydrogel coatings, the surface microstructure of gel coatings proves less conducive to microorganism adhesion, resulting in an extended service life. Moreover, the smooth surface of hydrogels reduces resistance during a ship’s forward movement [ 14 ]. This combination of attributes positions hydrogel coatings as a promising avenue for tackling the challenges associated with marine environments. In the realm of hydrogel coating research, pivotal attention is directed toward two key areas: the mechanical properties of the hydrogel itself and its interfacial toughness with the substrate. For the mechanical properties of the hydrogel, its performance is a critical benchmark for engineering applications [ 12 ]. Among the existing studies, double-network hydrogels have emerged as a common choice for enhancing these mechanical properties. These hydrogels boast exceptional tensile [ 15 ] and impact resistance owing to their unique dual-network structure. In current research, dual-network hydrogels are typically fashioned by combining a dissipative network and a solid network [ 16 ]. The solid network, primarily cross-linked irreversibly, provides the hydrogel with significant mechanical strength [ 9 , 16 , 17 , 18 ]. Conversely, the dissipative network, typically cross-linked reversibly, dissipates external stress by undergoing the fracture and regeneration of its cross-linking nodes, thereby fortifying the hydrogel’s toughness and refining its mechanical characteristics. Considering the application scenarios of hydrogel marine coatings, dual-network hydrogels prove particularly well-suited [ 16 , 19 , 20 , 21 , 22 , 23 ]. They ensure the stability of the gel coatings over extended periods, even under diverse underwater conditions, thereby offering prolonged protection to the substrate [ 24 ]. Turning to interfacial toughness, hydrogel materials have remained a focal point across various application domains since their inception [ 9 ]. The work of Yuk et al., researchers who achieved groundbreaking progress by covalently anchoring hydrogels to non-porous surfaces and elastomers, introduced a paradigm shift in hydrogel coating research and development [ 24 , 25 ]. On the basis of this research, hydrogel coating technology in shipbuilding has seen rapid advancement in recent years, with various structures demonstrating excellent underwater anti-fouling performance. However, challenges persist, particularly regarding the covalent anchoring strategy. While this method has shown promise, its applicability is limited to substrates with active hydroxyl groups on their surfaces, such as glass and ceramics [ 26 , 27 ]. For materials commonly used in ships, such as aluminum and steel, surface plasma treatment is required to impart anchoring activities [ 28 , 29 ], which is costly and not feasible for industrial applications. Moreover, the soft nature of hydrogel materials renders them susceptible to damage in the unpredictable ocean environment, exposing substrates to corrosion [ 9 , 12 ]. In addition, efforts to incorporate antibacterial agents, such as Cu 2+ , into hydrogels for functional antibacterial effects face challenges, but the antibacterial agents tend to leach into seawater at an accelerated rate, diminishing the functional lifespan of the hydrogel coating. Building upon the aforementioned research on hydrogel marine coatings, this study introduces an innovative acrylic resin–hydrogel double-layer (A-H DL) marine anti-corrosion and anti-fouling coating structure. The A-H DL coating comprises an acrylic resin (AR) primer, which includes polyvinylpyrrolidone-cuprous oxide (PVP-Cu 2 O) microcapsules and 3-(trimethoxymethylsilyl) propyl methacrylate-modified nano-silica (TMAPMS-SiO 2 ), and a dissipative double-network double-anchored hydrogel (DNDAH) coating, showcasing commendable mechanical properties, robust adhesion, and effective anti-biosorption functionality. This double-layer hydrogel coating structure addresses the challenges associated with the low-cost anchoring and long-term underwater anti-biosorption of hydrogel coatings by incorporating a primer coating between the gel layer and the substrate. The primer is designed to provide anchoring active sites and facilitate copper ion release, enhancing antibacterial performance. Additionally, in the event of gel coating damage, the primer serves as a final protective barrier. The A-H DL coating, which boldly uses a hydrogel surface to achieve chemical structure anti-fouling, adheres to the design concept of environmental protection of modern marine coatings while retaining the efficient anti-biofouling performance of traditional copper-based coatings, and it also improves its mechanical performance through innovation, thus achieving dual-anchoring, enhanced adhesion, and superior anti-biofouling properties for marine applications.",
"discussion": "2. Result and Discussion 2.1. Construction of the A-H DL Coating System The coating system of A-H DL combined a Cu 2 O-SiO 2 -AR primer for anchoring and controlled copper release with a mechanically strong and anti-fouling DNDAH. The detailed internal structure of the coating is depicted in Figure 1 . For the primer system, an acrylic resin coating served as the foundation, with the addition of Cu 2 O particles coated with PVP to impart an antibacterial functionality. Incorporating TMSPMA-modified nano-silica particles enhanced the anchoring ability of the primer and gel system. For the hydrogel component, we adopted the construction strategy of a double-network hydrogel. The AAm monomer was copolymerized and cross-linked with a coupling agent to form the first network, with the coupling agent condensing to create irreversible cross-links. The second network incorporated AGM grafted with a coupling agent. Alginate coordinated with the Cu 2+ released from the Cu 2 O in the primer, forming reversible cross-links. The Cu 2+ was retained in the gel for an extended period, ensuring long-term antibacterial protection for the hull. The coating’s outstanding mechanical properties and prolonged anti-biosorption function were experimentally verified. The two networks interconnected through the condensation of the coupling agents, and the anchoring groups on both networks’ coupling agents underwent condensation with the silicon hydroxyl groups of the SiO 2 in the primer, forming covalent anchoring with verified super-anchoring performance to form the A-H DL coating. The chemical composition, excellent mechanical properties, and marine anti-fouling functionality of the A-H DL coating were systematically demonstrated through subsequent characterization and performance experimental test results. 2.2. Analysis of the Internal Chemical Structure of the Coatings 2.2.1. Internal Structure of DNDAH The hydrogel coating in this study consisted of a robust first network, formed by the copolymerization and cross-linking of a coupling agent (TMSPMA) and acrylamide, and a dissipative second network created by the physical cross-linking of copper and a long chain of alginate (Cu 2+ gradually released from the primer), modified by a coupling agent (APES). Firstly, we conducted an FTIR analysis to identify the functional groups in the hydrogel samples. The results for TMSPMA-PAAm, AGM, and DNDAH are presented in Figure 2 a. In the TMSPMA-PAAm sample spectrum, the broad absorption peak at 3420 cm −1 corresponded to the silanol (Si–OH) stretching vibration, while 3197 cm −1 was the characteristic absorption peak of the associated –NH 2 . Additionally, the peak at 1120 cm −1 corresponded to the Si–O–Si bond’s antisymmetric stretching vibration, and the peak at 1670 cm −1 corresponded to the stretching vibration of the C=O bond in amides. Notably, there was no absorption peak for the terminal ene groups. This infrared spectrum confirmed the successful copolymerization of the AAm monomer with TMSPMA, with some coupling agents undergoing condensation reactions to cross-link into a hydrogel, while others retained the silicon hydroxyl groups as the anchor points. In the AGM sample spectrum, a similar backbone with TMSPMA-AAm was observed. A broad absorption peak at 3400 cm −1 was attributed to the silanol (Si–OH) stretching vibration, and the adsorption peak at 1670 cm −1 corresponded to the stretching vibration of the C=O bond in amides. Additionally, 1290 cm −1 represented the ether bond absorption peak in the long chain of alginate, while the peak at 1092 cm −1 corresponded to the Si–O–Si bond’s antisymmetric stretching vibration. Importantly, there was no absorption peak for the carboxyl (–COOH) and primary amine (–NH 2 ) groups. The infrared spectrum results indicated the successful grafting of the coupling agent onto the main chain of alginate, with some coupling agents cross-linked into hydrogels through condensation reactions, while others retained the silyl hydroxyl group as an anchor point and cross-linking site with the first network. On the other hand, the spectrum of the DNDAH sample combined the infrared spectrum characteristics of the TMSPMA-AAm and AGM samples. It exhibited an absorption peak at 3400 cm −1 corresponding to alcohol hydroxyl, originating from the two types of networks. The peak at 3197 cm −1 represented the characteristic absorption peak of the associated –NH 2 , derived from the primary amine in the TMSPMA-AAm network. Additionally, the peak at 1290 cm −1 corresponded to the ether bond’s absorption peak, consistent with the spectrum of AGM. Therefore, it was preliminarily inferred that the DNDAH sample comprised both TMSPMA-AAm and AGM networks, demonstrating an anchoring performance. Furthermore, the SEM images of the hydrogel sample ( Figure 2 d,e) showed the 3D network structure of the hydrogel to a certain extent. The water that initially occupied most of the space in the hydrogel was removed by freeze-drying, leaving only the polymer skeleton, which retained the original hollow shape to a certain extent. Firstly, Figure 2 d illustrates that the internal structure of the polymer is porous, aligning with the internal structural characteristics of the hydrogel [ 30 ]. Additionally, the polymer exhibits a high degree of cross-linking, and research indicates that the higher the cross-linking degree of a material, the better its anti-swelling performance. Therefore, the low swelling rate of the hydrogel coating can be explained in terms of the 3D structure, which is conducive to the long-term adhesion of the hydrogel coating onto the primer surface. Secondly, in Figure 2 e, the presence of smaller spherical particles in the network structure is apparent. Upon zooming in, Figure 2 e shows microspheres with a diameter of approximately 20 μm. According to the theoretical gel structure, it can be inferred that these are coupling agent-cross-linked alginate gel microspheres dispersed in the hydrogel system, confirming the theoretical structure of the hydrogel’s double network. 2.2.2. Internal Structure of AR Primer The AR primer contained acrylic resin and some additives that helped the coating solidify, as well as nano-PVP-Cu 2 O particles and modified nano-SiO 2 dispersed in the coating. The following mainly displayed the chemical structure and morphological characteristics of the two types of nano-particles. On the one hand, for the PVP-Cu 2 O sample, we characterized its chemical composition and morphological characteristics. Firstly, we conducted an FTIR analysis to verify the composition and embedding integrity of the PVP-Cu 2 O microcapsules. The results of Cu 2 O, PVP, and PVP-Cu 2 O are shown in Figure 2 b. Characteristic absorption peaks at 2953 cm −1 and 1431 cm −1 (–CH), 1667 cm −1 (C–O), and 1283 cm −1 (C–N) were found. The absorption peak at 3470 cm −1 was the characteristic peak of –OH, which was supposed to be derived from the water adsorbed by the PVP. For the spectrum of Cu 2 O, the absorption peak at 628 cm −1 corresponded to Cu 2 O, and the absorption peak at 3310 cm −1 was supposed to be caused by the moisture in the tested sample. For the spectrum of PVP-Cu 2 O, the characteristic absorption peak of Cu 2 O was found at 607 cm −1 . Moreover, characteristic absorption peaks at 2971 cm −1 and 1416 cm −1 (–CH) and 1653 cm −1 (C–O) were also found. These results show that the components of the produced microcapsules were PVP and Cu 2 O [ 31 ]. In addition, we further validated the chemical composition of the PVP-Cu 2 O microcapsules by XRD. Figure 2 i shows the XRD profiles of Cu 2 O, PVP, and PVP-Cu 2 O, respectively. It can be observed from the XRD pattern of Cu 2 O that six major characteristic X-ray diffraction peaks appeared at positions 2θ of 31.3°, 38.1°, 44.1°, 63.2°, 75.5°, and 79.4°, whereas PVP, as an organic noncrystallographic material, had no obvious crystal characteristic peaks in the XRD pattern. For the PVP-Cu 2 O XRD pattern, the characteristic diffraction peaks of 38.1° and 44.1° still existed, but the intensity was greatly reduced, and the remaining weaker characteristic peaks were intermixed with the hybrid peaks of PVP, indicating that the PVP phase interfered with the X-ray diffraction intensity of Cu 2 O, further proving the presence of Cu 2 O and PVP in PVP-Cu 2 O. We could observe the specific morphological characteristics and particle size of the microcapsules through TEM. Figure 2 f,g show the Cu 2 O particles without PVP and the microcapsules with PVP, respectively. Firstly, these microcapsules have a spherical shape in appearance and a distinct core–shell structure, with a spherical radius of approximately 200 nm and a shell thickness of approximately 40 nm, proving the inclusion structure of PVP-Cu 2 O. Then, we also characterized the chemical composition and morphological characteristics of the TMSPMA-SiO 2 sample. First of all, we performed an FTIR analysis to explore the influence of coupling agent modification on the chemical structure of nano-SiO 2 . The results of SiO 2 and TMSPMA-SiO 2 are shown in Figure 2 c. For the characteristic absorption peak of Si–OH at 1056 cm −1 , the absorption of TMSPMA-SiO 2 was significantly weaker than that of the unmodified SiO 2 , indicating that the modification of the coupling agent reduced the hydroxyl’s density on the silica surface, thus weakening the agglomeration of silica during the high-temperature curing of the paint film. Moreover, we observed the morphology and particle size of TMSPMA-SiO 2 through TEM. Figure 2 h shows that the particle size of TMSPMA-SiO 2 was around 50 nm. 2.3. Anchoring Performance and Mechanical Properties of A-H DL Coating 2.3.1. Internal Tensile Strength of DNDAH The tensile strength is one of the crucial indicators of mechanical properties [ 32 ]. We conducted tensile strength tests on SNSAH and DNDAH at room temperature and obtained Figure 3 . The tensile properties of SNSAH were significantly weaker than those of DNDAH. According to the calculation results, the calculated tensile strength of SNSAH was 0.020 MPa, and the tensile strength of DNDAH without Cu 2+ was 0.29 MPa, which was the same as that of SNSAH, whereas that of DNDAH was 0.073 MPa ( Figure 3 a). From Figure 3 b, we can observe that the modulus of DNDAH was obviously higher than those of SNSAH and DNDAH without Cu 2+ . This verified the dissipation effect of copper AGM on external stress. This tensile strength was similar to the tensile strength of interpenetrating double-network hydrogels reported in previous studies [ 33 , 34 ], indicating that the mechanical properties and dissipative properties of the dissipative gel microspheres and those of an interpenetrating dissipative gel are the same, but AGM can provide a second anchoring effect on the primer. Figure 3 c demonstrates the mechanism of how DNDAH dissipates external mechanical energy. 2.3.2. The Adhesion Properties of A-H DL The adhesion properties of the product included the anchoring properties between the DNDAH and the AR primer and the adhesion between the AR primer and the substrate. All the results of the experiments are displayed in Figure 4 . Firstly, we proved the excellent adhesion properties between the AR primer and the substrate via a cross-cut test. The results of the experiment are demonstrated in Figure 4 a. We found that almost no paint film was peeled off from the substrate by the tape (the tape marked by the red box is very clean). Then, the AR primer that had been immersed in water for 30 days showed a similar experiment phenomenon as the primer mentioned above. Hence, the AR primer had great adhesion properties to the substrate. Furthermore, we verified the excellent anchoring performance between the DNDAH and the AR primer coating through an underwater swelling and detachment experiment and peeling experiments. The anchoring structure was formed by the condensation of the hydroxyl (–OH) provided by TMSPMA-SiO 2 on the AR primer surface with two silane coupling agents in the gel system. The interface adhesion and dissipation mechanisms of the SA hydrogel are shown in Figure 4 j. In the underwater swelling and detachment experiment, shown in Figure 4 b, there was a change in the area percentage of the hydrogel anchored to the primer over time. It was clear that the percentage of the hydrogel with double-anchoring (DA) remained constant over 30 days, while the percentage of the single-anchoring (SA) hydrogel suffered a moderate decrease. However, the hydrogel without anchoring detached completely. The anchoring of the hydrogel coating and the primer coating after immersion in water for 30 days is shown in Figure 4 d–g. Figure 4 e shows the paint coat composed of the primer without TMSPMA-SiO 2 combined with DNDAH; Figure 4 f shows the paint coat composed of the AR primer with TMSPMA-SiO 2 combined with SNSAH; Figure 4 g illustrates the AR primer composed of the primer with TMSPMA-SiO 2 combined with DNDAH. Upon comparing the results shown in Figure 4 e,g, it could be observed that the DNDAH coating on the primer without TMSPMA-SiO 2 had almost completely detached. However, the DNDAH coating on the primer with added TMSPMA-SiO 2 exhibited no edge warping or bulging. This phenomenon proved that TMSPMA-SiO 2 provided the Si-OH capable of undergoing a condensation reaction with the coupling agent in the DNDAH coating, forming a chemically anchored interface between the two coatings and the primer. Additionally, upon comparing the results in Figure 4 f,g, it was evident that SNSAH displayed partial edge warping, while the double-anchored hydrogel did not show any edge warping. This phenomenon indicated that the DNDAH had a more robust and enduring anchoring effect on the AR primer when TMSPMA-SiO 2 was added. In the peeling experiment, the unanchored sample could peel off the hydrogel coating from the AR primer with only 0.2 N, while the SA hydrogel and DA hydrogel failed to peel off the hydrogel coating from the AR primer, until it broke, indicating that the interfacial toughness between the hydrogel coating and the AR primer was greater than the internal toughness of the hydrogel ( Figure 4 c). The interface adhesion performances of the peeling DA and SA hydrogels are illustrated separately in Figure 4 h,i. We can observe that some part of the DA hydrogel remained on the AR primer, while the SA hydrogel could be peeled off successfully. To sum up, the two layers of the A-H DL coating could be connected extremely firmly via the covalent anchoring structure, and the entire coating had great adhesion properties with respect to the substrate. Hence, the brilliant adhesion of the A-H DL coating could be proven. 2.4. Functional Evaluation of the Coatings 2.4.1. Cu 2+ Release Performance The results of the Cu 2+ release properties’ experiment are demonstrated in Figure 5 a–c. Regarding the results of the determination of the Cu 2+ release rate ( Figure 5 a), the initial copper ion release rate of the sample AR primer reached up to 20 μg·cm −2 ·d −1 . Then, there was a significant decline, during which the rate dropped to 9.8 μg·cm −2 ·d −1 on the fifth day. Subsequently, after rising to 13.6 μg·cm −2 ·d −1 on the fifteenth day, the rate dropped again and tended to stabilize gradually. It could be deduced that, due to the presence of a large amount of Cu 2 O in the coating, the release rate of the copper ions was very fast in the initial stage. Then, the Cu 2 O value in the surface layer of the coating rapidly decreased, resulting in a significant decrease in the release rate. As the soaking time progressed, the internal structure of the coating became loose, intensifying the release of copper oxide and increasing the release rate of the copper ions to the maximum value. Finally, with the consumption of Cu 2 O, the release rate gradually decreased and eventually stabilized [ 35 ]. The behavior of the AR primer releasing Cu 2+ into the artificial seawater was similar to that of the AR primer releasing Cu 2+ into DNDAH because of the highwater content of the hydrogel. This release pattern ensured that the Cu 2+ could release into the hydrogel immediately to form the reversible cross-linking structure. Meanwhile, the release rates of A-H DL were always zero before 20 days and were still very small after that time. This phenomenon indicated that DNDAH provided a good buffering effect on the release process of Cu 2+ , thus decreasing the pollution of Cu 2+ in the marine environment. Figure 5 b shows the mapped scan area. Figure 5 c displays the elemental content of the samples. Figure 5 d,e indicate that there are many C and O elements in DNDAH and further prove that the skeleton of this hydrogel is connected by carbon hydrogen oxide compounds [ 36 ]. In Figure 5 f, the presence of copper in the DNDAH is evident, and it can be observed that the distribution of copper elements is relatively uniform. This indicates that the Cu 2 O in the primer could successfully release Cu 2+ into the DNDAH coating through an electrochemical reaction. The working mechanism of Cu 2 O releasing Cu 2+ into DNDAH to coordinate with AGM and Cu 2+ releasing from DNDAH into seawater to play an anti-fouling role is shown Figure 5 g. 2.4.2. Anti-Protein Adsorption Ability The anti-fouling characteristics of the DNDAH coating were analyzed via anti-protein-fouling using BSA as a protein model. After 1 day of immersion in BSA, more protein agglomeration was observed on the primer coating. The DNDAH coating exhibited a much better resistance, with an 88% reduction in protein attachment, compared to the coating without DNDAH, which exhibited a 74% resistance. To comprehend the reasons behind the enhanced protein contamination resistance of the hydrogel coating, we conducted measurements of the DNDAH, the AR primer, and the iron wettability and surface free energy [ 37 ]. Figure 6 a illustrates the dynamic water contact angle on the DNDAH, the AR primer, and iron over 30 min, and the pictures of the contact angle are shown in Figure 6 c. Initially, the DNDAH coating exhibited hydrophobic characteristics, with a high contact angle value of 100°. However, as time progressed, the contact angle dropped to 13°, indicating a transition from hydrophobic to super-hydrophilic. However, the contact angle of water on the AR primer and iron surfaces did not change significantly, and the small change in the angle was speculated to be due to the influence of gravity. We counted the surface free energy (mJ/m 2 ) of the three surfaces. Figure 6 b demonstrates that the values of the iron and the AR primer both maintained high levels over 30 min, even if the rates of the iron went through a more obvious decrease; however, the value of the DNDAH dropped to nearly zero. This transition was attributed to the migration of hydrophilic polymer chains from the interior to the surface of the hydrogel coating, forming a hydration layer which acted as an effective barrier against contact with proteins and other substances [ 19 ]. Previous findings have indicated that surfaces with anti-protein adsorption abilities generally exhibit characteristics of hydrophilicity and electrical neutrality. The surface of the hydrogel coating in this study met these requirements, demonstrating an excellent anti-protein adsorption performance. In contrast, the primer sample without the hydrogel coating was hydrophobic, with a significantly higher surface free energy compared to the DNDAH coating. The former was prone to surface-charging underwater, leading to an inferior anti-protein adsorption performance. 2.4.3. Marine Simulation Experiment In order to test the long-term anti-biological adsorption performance of the A-H DL coatings in real marine environments, we simulated a seawater environment in a fish tank and immersed four types of coatings into it for 30 days. The microbial attachment on the samples from the fish tank after a 30-day-long immersion in seawater is illustrated in Figure 7 a, while Figure 7 b presents the ratio of untouched coating areas on four different coatings. The AR primer samples, which lacked the DNDAH coating, indicated a substantial microbial attachment, whereas the DNDAH coating revealed a minimal adhesion of microorganisms. This evidence validated the greater anti-microbial adsorption efficacy of the DNDAH coating compared to traditional marine coatings, hereby represented by the AR primer. Despite the difficulty in discerning the differences in the anti-microbial adsorption performance of the two A-H DL coatings using the naked eye, thin slices of the hydrogel surface of each coating were observed under an optical microscope. As highlighted in Figure 7 a, the DNDAH coating not anchored with the PVP-Cu 2 O AR primer showed a minimal microbial attachment of about 700 cells cm −2 , as indicated by the red circle in the image. In stark contrast, the hydrogel coating anchored with the PVP-Cu 2 O AR primer demonstrated no microbial attachment. This observation verified that the Cu 2+ released by the AR primer into the hydrogel fortified the DNDAH coating’s anti-biosorption ability. Previous studies have established the antibacterial tendencies of Cu 2+ . Hence, in an environment housing Cu 2+ , organisms’ survival rates dip significantly. Consequently, the DNDAH coating was said to employ its inherent hydrophilicity to resist biological attachment after the Cu 2 O in the AR primer released Cu 2+ into the hydrogel coating by way of an electrochemical reaction. The experimental results demonstrated that A-H DL had a good anti-fouling performance over 30 days in seawater [ 19 ], which was sufficient to prove that it could withstand harsh marine environments over an extended period without losing its anti-fouling properties. The mechanism of resisting biological adherence is illustrated in Figure 7 c. Firstly, the surface of the hydrogel was not fit for microorganisms to live on, and, once a microorganism adhered to it accidentally, the Cu 2+ in the gel would kill it immediately. For the algae in the sea, the hydrogel had a low modulus, thus the roots of algae could not grow on the surface easily. As for the proteins which generally come from mussels, the hydration layer formed on the surface of the hydrogel in the sea became a barrier to prevent these proteins from adhering to the ship. Such a process not only resisted biological adherence but also disrupted the physiological behavior of organisms that accidentally adhered to the coating through the Cu 2+ present in the DNDAH coating. This feature allowed for further resistance to biological adsorption. Briefly, the A-H DL coating has long-term anti-biofouling properties in real marine environments based on the surface characteristics of the hydrogel and the existence of Cu 2+ , and it has significant advantages in terms of its anti-biofouling properties compared to traditional ship coatings. 2.5. Considerations about Monitoring and Maintenance Strategies The conditions in a real sea environment are very complex, so the functional lifespan of the A-H DL coating is unsure. Therefore, it is necessary to construct a mechanism to monitor the protective performance of the A-H DL coating and invent a convenient method to maintain it [ 38 , 39 ]. We propose two monitoring strategies here. The first is morphological observation, which is the traditional method of checking for mechanical damage to a coating. If there is some damage on the DNDAH coating, based on the reversible structure of the hydrogel system, workers could maintain the damage by filling the hydrogel precursor directly in the damaged part. The second method is sample testing. There are two aspects to this test, which overall involves sampling from the surface of the ship, including the coatings and any attached organisms. During this test, researchers first detect the Cu 2+ content in the DNDAH to assess its current anti-biofouling performance. Second, the samples are sent to the laboratory for a microbiological analysis to determine which types of microorganisms have begun to exceed the prevention and control standards. If the lifespan of A-H DL is over, the whole coating, including the AR primer and the DNDAH, should be updated. In conclusion, in practical application scenarios, workers should consider two problems of this coating—appearance integrity and functional effectiveness—and should apply local or global maintenance according to the specific situations."
} | 8,814 |
36648151 | PMC9796772 | pmc | 4,289 | {
"abstract": "Abstract While horticulture tools and methods have been extensively developed to improve the management of crops, systems to harness the rhizosphere microbiome to benefit plant crops are still in development. Plants and microbes have been coevolving for several millennia, conferring fitness advantages that expand the plant’s own genetic potential. These beneficial associations allow the plants to cope with abiotic stresses such as nutrient deficiency across a wide range of soils and growing conditions. Plants achieve these benefits by selectively recruiting microbes using root exudates, positively impacting their nutrition, health and overall productivity. Advanced knowledge of the interplay between root exudates and microbiome alteration in response to plant nutrient status, and the underlying mechanisms there of, will allow the development of technologies to increase crop yield. This review summarizes current knowledge and perspectives on plant–microbial interactions for resource acquisition and discusses promising advances for manipulating rhizosphere microbiomes and root exudation.",
"introduction": "INTRODUCTION Diverse microbes (archaea, bacteria, fungi and protists) cohabiting with plants are collectively known as the plant microbiota (Bulgarelli et al., 2013 ; Compant et al., 2019 ). The plant microbiota, its inhabitants, habitats, genomes and surrounding environmental conditions are termed collectively the plant microbiome (Berg et al., 2020 ; Marchesi & Ravel, 2015 ), which is presently considered an extended plant trait with functional capabilities that contribute to plant host nutrition, development and immunity (Lemanceau et al., 2017 ; Teixeira et al., 2019 ; Vandenkoornhuyse et al., 2015 ). The highly diverse plant‐associated microbial communities are shaped by biotic and abiotic constraints varying on time, and space (Hassani et al., 2018 ; Xiong et al., 2021 ). Plants gradually enrich microbes in specific plant compartments creating microbial habitats that typically start from the bulk soil and can move into above‐ground internal plant tissues. Thus, the composition of the plant microbiome is compartment specific and is divided into rhizosphere (soil surrounding the plant roots; Zhang et al., 2017 ), the endosphere (interior of the above and below plant organs; Compant et al., 2021 ), and the phyllosphere (above‐ground portion of the plant; Koskella, 2020 ). The deeper these soil bacteria, fungi and other micro‐organisms move into these different plant compartments, the more they are filtered out or selectively recruited by the plant's signaling molecules and immune system (Xiong et al., 2021 ). The rhizosphere is dominated by prokaryotic phyla including Proteobacteria, Actinobacteria, Bacteroidetes, Firmicutes and Acidobacteria. In terms of the fungal phyla, the rhizosphere is dominated by Ascomycota and Basidisomycota, which are also the most common taxonomical phyla in soils (Mohanram & Kumar, 2019 ). Deciphering plant and microbial interaction is a multi‐disciplinary research endeavor that integrates different branches of biology including ecology, microbial, plant and molecular biology applying informatics, statistics and modelling as well as biotechnology (Berg et al., 2020 ). Efforts to link specific microbial processes to specific microbial taxa have been accelerated with genomic data (i.e. marker gene, genomic and metagenomic) by grouping taxa according to similarity in strategies and functional attributes (Carrión et al., 2019 ; Song et al., 2020 ). These methodological and conceptual advances have accelerated our understanding of the plant microbiome (Fierer, 2017 ). By describing and understanding the plant‐associated microbial communities and their functional features, we can manipulate the plant rhizosphere microbiome to enhance plant health and productivity (Xun et al., 2021 )."
} | 969 |
27136705 | PMC4895246 | pmc | 4,290 | {
"abstract": "Colony\nbiofilms of Bacillus subtilis are a widely\nused model for studying cellular differentiation. Here, we applied\nmatrix-assisted laser desorption/ionization (MALDI) mass spectrometry\nimaging (MSI) to examine cellular and molecular heterogeneity in B. subtilis colony biofilms. From B. subtilis cells cultivated on a biofilm-promoting medium, we detected two\ncannibalistic factors not found in previous MALDI MSI studies of the\nsame strain under different culturing conditions. Given the importance\nof cannibalism in matrix formation of B. subtilis biofilms, we employed a transcriptional reporter to monitor matrix-producing\ncell subpopulations using fluorescence imaging. These two complementary\nimaging approaches were used to characterize three B. subtilis strains, the wild type isolate NCIB3610, and two mutants, Δ spo0A and Δ abrB , with defective and\nenhanced biofilm phenotypes, respectively. Upon deletion of key transcriptional\nfactors, correlated changes were observed in biofilm morphology, signaling,\ncannibalistic factor distribution, and matrix-related gene expression,\nproviding new insights on cannibalism in biofilm development. This\nwork underscores the advantages of using multimodal imaging to compare\nspatial patterns of selected molecules with the associated protein\nexpression patterns, obtaining information on cellular heterogeneity\nand function not obtainable when using a single method to characterize\nbiofilm formation.",
"conclusion": "Conclusions We developed a multimodal imaging method to compare metabolite\ndistribution and gene expression patterns on the surfaces of a widely\nused B. subtilis biofilm model. Using fluorescence\nimaging coupled with a transcriptional reporter, we were able to analyze\nmatrix production in biofilm development. This key phenotype had not\nbeen examined in previous MALDI MSI studies of B. subtilis , perhaps due to the incompatibility of sample preparation methods\nbetween metabolites and protein amyloids. Combining MALDI and fluorescence\nimaging enabled the detection of distinct populations of cells in\na biofilm previously assumed to be comprised of an identical population.\nThe combination of genetic tools, fluorescence imaging, and MALDI\nMSI is applicable to other bacterial biofilms for examining spatial\ndistributions in a range of complex biofilm communities.",
"introduction": "Introduction Biofilms are microbial\ncommunities of surface-associated microorganisms\nembedded in a self-secreted extracellular matrix. 1 Research on microbial biofilms improves our understanding\nof fundamental biological processes, such as cell differentiation 2 , 3 and intercellular communication. 4 These\nefforts have also delivered immediate applications in biomedical and\nbiotechnological settings, including prevention of biofilm growth\non indwelling clinical devices 5 and wastewater\ntreatment systems. 6 Although most naturally\noccurring biofilms consist of multiple microbial species, single-species\nbiofilms are widely employed in laboratory research as experimental\nmodels. 7 Bacillus subtilis , a nonpathogenic Gram-positive\nbacterium, is among the most popular models for studying biofilms. 8 B. subtilis can develop\ndiverse types of biofilms under different culture conditions, including\ncolony biofilms at the air–solid interface, floating pellicles\nat the air–liquid interface, and submerged biofilms at the\nliquid–solid interface. 8 In particular,\nwhen grown on agar media with a biofilm-inducing composition (minimal\nsalts glycerol glutamate medium, or MSgg medium), B. subtilis forms wrinkled colony biofilms with complex structures. 9 During biofilm development, functionally distinct\ncell subpopulations arise from genetically identical ancestors, following\na seemingly ordered differentiation sequence: motile cells–matrix-producing\ncells–sporulating cells. 10 The matrix-producing\ncells have also been suggested to be the same subpopulation of cannibals,\nwhich secrete toxins that can lyse a fraction of their sensitive siblings. 11 As the cannibal/matrix-producing cells exhibit\nresistance to these toxins, they utilize the released nutrients from\nthe cannibalized cells and increase in number, which leads to enhanced\nmatrix production and promotes biofilm formation. 11 Phenotypic heterogeneity in B. subtilis biofilms is considered to be a result of spatiotemporal crosstalk\nbetween chemical signals and gene expression, 10 and yet the ability to characterize such complex interactions requires\nmajor advances in currently available analytical approaches. Imaging techniques have been instrumental in understanding spatial\nheterogeneity in biofilms. Mass spectrometry imaging (MSI) is a label-free\nmolecular imaging technique that can provide two- or even three-dimensional\nvisualization of metabolite distribution in biological samples, and\nhas been increasingly used in microbiological research. 12 , 13 The most broadly used mass spectrometry (MS)-based imaging techniques\nin microbiology are matrix-assisted laser desorption/ionization (MALDI)\nimaging, secondary ionization mass spectrometry (SIMS) imaging, and\ndesorption electrospray ionization imaging. 12 Several applications using the aforementioned MSI techniques have\nbeen performed successfully to unravel dynamic spatial or temporal\nchemical information for metabolites in various microbial systems,\nsuch as single B. subtilis colonies or coculture\nsystems, 14 − 16 Pseudomonas aeruginosa biofilms, 17 − 20 and plant–microbe cocultures. 21 With the aid of MSI, in situ visualization of the spatial distribution\nof individual molecules is enabled without the need for chemical derivatization\nor immunostaining. The ion images (or m / z images) obtained provide insight into microbe–microbe or\nmicrobe–plant interactions and metabolic exchange, and enable\ndiscovery of novel natural products. 14 , 15 , 21 , 22 Given the complex\nnature of biofilms, the combination of multiple\nimaging techniques with complementary figures of merit has potential\nfor unraveling biofilm biology. 23 In our\nprevious work, confocal Raman microscopy (CRM) was combined with MALDI 18 and SIMS imaging 19 , 20 to study biofilms\nof the opportunistic pathogen Pseudomonas aeruginosa . CRM can be used to visualize molecular distributions based on characteristic\nvibrational modes of chemical functional groups, and correlated CRM\nand MSI has been successfully used to expand chemical coverage, resolve\nsubtly differing compounds, and cross-validate molecular distributions. 18 − 20 Besides measuring small-molecule toxins, we also wanted to\nexamine\ngene expression patterns to study cellular heterogeneity, and then\ncorrelate these results with the presence of specific gene products.\nWhile in principle CRM or MSI could provide this information, there\nare several issues. For example, CRM can detect typical bands of DNA\nand RNA molecules in bacterial biofilms, 18 , 20 but it does not differentiate the identities of different genetic\nspecies. Moreover, MALDI MS of intracellular oligonucleotide species\nrequires complicated sample preparation processes, such as cell lysis\nand purification, 24 which are difficult\nto adapt for an MSI experiment. On the other hand, fluorescence imaging\nhas long been a primary tool in assisting biologists in monitoring\nspecific gene expression via the use of transcriptional reporters. 25 These genetic reporters are constructed by fusing\na fluorescence reporter gene with the upstream DNA sequence (promoter)\nfrom a gene of interest; expression of the targeted gene can be correlated\nto the abundance of fluorescence proteins. 10 , 26 For example, using fluorescence reporters of the hag , yqxM , and sspB genes, which are\nexclusively expressed in motile, matrix-producing and sporulation\ncells, respectively, revealed that these different cell types exist\nin distinct spatial locations within Bacillus biofilms. 10 For this work we combined MALDI and fluorescence\nimaging to compare\nmetabolite distributions to spatial patterns of differentiated cell\ntypes, revealing molecular mechanisms that are impossible to study\nwhen single imaging methods are used alone. Using MALDI MSI to study B. subtilis colony biofilms grown on biofilm-promoting\nMSgg agar media, we observed two cannibalistic toxins that were absent\nin previous studies in which the rich medium ISP2 was used to cultivate\nthe same strain. We then combined MALDI MSI and fluorescence stereoscopy\nto examine one wild type biofilm and two mutant strains with distinct\nbiofilm phenotypes, Δ spo0A and Δ abrB . We report correlated changes in metabolite abundance\nand matrix-related gene expression upon deletion of key transcriptional\nfactors, and discuss these observations in the context of the genetic\nregulation of B. subtilis biofilm formation.",
"discussion": "Results and Discussion MALDI\nMSI of Bacillus Colony Biofilms on MSgg\nMedia In this work, we focused on the wild type isolate NCIB3610\nstrain, which is a model for studying B. subtilis colony biofilms. 8 − 10 Elegant previous MALDI MS studies of B. subtilis culture produced exciting data on several molecular distributions, 14 , 15 but only reported images from ISP2 medium, on which relatively unstructured\ncolonies form ( Figure S1 ). To examine whether\ncell differentiation is associated with the spatial heterogeneity\nof metabolite distribution in B. subtilis biofilms,\nwe employed MSgg medium, a widely used condition to induce biofilm\ndevelopment. 9 When grown on MSgg, NCIB3610\nformed characteristic wrinkled colony biofilms ( Figure S1 ), consistent with previous reports. 9 As different medium compositions often require customized\nsample preparation for MALDI MSI of microbial cultures on thin-layer\nagar, 28 we screened appropriate MALDI matrices\nand matrix application approaches for MSgg. Three commonly used MALDI\nmatrices—DHB, CHCA, and a 1:1 mixture of CHCA and DHB—were\ntested and sieved to saturate agar samples. In comparison to CHCA\nand the mixture of CHCA and DHB, the DHB matrix powder provided more\nhomogeneous matrix layers and a broader coverage of target analytes,\nincluding surfactins, plipastatins, and subtilosin, as well as two\ncannibalistic factors, sporulation killing factor (SKF) and sporulation\ndelaying protein (SDP) ( Figure 1 ). Our assignment of these aforementioned known compounds\nis based on a comparison of our tandem MS results with fragmentation\ndata reported in the literature 14 , 15 , 29 ( Table S2 and Figure S2 ). Figure 1 MALDI imaging analysis\nof colony biofilms of B. subtilis NCIB3610 and\nits mutants. (A,B) Single-pixel MALDI-MS spectra of\nsurfactins, plipastatins, subtilosin, SKF and SDP from the inner (A)\nand outside (B) regions of B. subtilis NCIB3610\nbiofilms. (C) Selected MALDI images for B. subtilis biofilms. Each column represents (left to right): m / z 1030 (surfactin-C13, [M + Na] + ); m / z 1044 (surfactin-C14, [M + Na] + ); m / z 1058 (surfactin-C15, [M\n+ Na] + ); m / z 1485 (plipastatin-C16-Ala,\n[M + Na] + ); m / z 1499\n(plipastatin-C17-Ala, [M + Na] + ); m / z 1513 (plipastatin-C16-Val, [M + Na] + ); m / z 1527 (plipastatin-C17-Val, [M + Na] + ); m / z 2782 (SKF, [M + H] + ); m / z 3422 (subtilosin,\n[M + Na] + ); m / z 4334\n(SDP, [M + Na] + ); and an overlay of ion images (green:\nSKF; bright pink: SDP). The ion intensity is reflected by the intensity\nof colors. Each column of ions is displayed using the same intensity\nscale, optimized per each metabolite and normalized to the TIC. Scale\nbar = 5 mm. To confirm these identifications,\nstrains with gene deletions in\nsurfactin and plipastatin biosynthesis pathways were analyzed. 15 The absence of surfactin and plipastatin ions\nwas observed in the Δ srfAA and Δ ppsB strains, respectively ( Figure 1 ). Furthermore, MALDI imaging revealed region-specific\ndistributions of various metabolites ( Figure 1 ). Surfactins, including surfactin-C13, surfactin-C14,\nand surfactin-C15, were mainly detected as Na + and K + adduct ions. Surfactins were visualized in both agar media\nand bacteria cells, with ion intensities being higher in the cells\nthan in the agar. In contrast, plipastatins (plipastatin-C16-Ala,\nplipastatin-C17-Ala, plipastatin-C16-Val, and plipastatin-C17-Val)\nand subtilosin were preferably located in the agar media outside of\nthe biofilms ( Figure 1 ), and were detected as H + , Na + , and K + adduct ions. For the cannibalistic factors, SKF was merely\ndetected as H + ions, but SDP was detected as H + , Na + , and K + adduct ions. While both SKF and\nSDP were primarily associated with the biofilms, SKF was preferably\nlocated at the center and SDP was detected across the entire biofilm\nsurface ( Figure 1 ). Notably, SKF and SDP were not detected in previous MALDI imaging\nstudies when the same strain was cultivated on ISP2 medium. 14 Considering the importance of cannibalism in\nmatrix formation and biofilm development, 11 it is not surprising that the absence of SKF and SDP correlates\nwith the lack of biofilm formation in NCIB3610 cultures on ISP2 ( Figure S1 ). 14 , 15 Moreover,\nprevious studies on SKF and SDP primarily employed the laboratory B. subtilis strain PY79, which is incapable of forming\nbiofilms, 9 and MALDI imaging revealed that\nboth SKF and SDP were located evenly on the surfaces of PY79 colonies. 14 In contrast, SKF and SDP exhibited distinct\nspatial patterns on NCIB3610 biofilms ( Figure 1 ), suggesting an association between heterogeneity\nin metabolite distribution and cell differentiation. Furthermore,\na previous study suggested that surfactins were required to activate\ncannibalism, based on the appearances of biofilms formed by a few\nmutant B. subtilis strains. 11 While our results ( Figure S1 ) agree with the prior reports on the aberrant morphology of Δ srfAA biofilms, 9 , 11 we still observed SKF\nand SDP production in the absence of surfactins ( Figure 1 ). Together, the capability\nto detect SKF and SDP from B. subtilis biofilms\nusing MSgg suggests that in certain systems, it is necessary to use\nspecific cultivation media to replace ISP2, the most commonly used\nmedium in MALDI MSI analysis of microbial agar cultures. Monitoring\nMatrix-producing Cells Using a Fluorescence Reporter The\nmetabolites we characterized via MALDI MSI perform different\nroles in B. subtilis biofilm development: surfactins\ncan promote surface swarming of B. subtilis cells, 30 and they have been proposed as a class of signaling\nmolecules that trigger biofilm initiation. 26 , 31 SKF and SDP reduce the proportion of nonmatrix-producing cells to\nenhance biofilm formation; 11 , 14 however, the biofilm-related\nfunctions of plipastatins and subtilosin remain elusive. Although\nextracellular matrix production is a key step in biofilm development\nto encase constituent cells into a structurally integrated community, 32 , 33 matrix components were not detected in this and previous MALDI imaging\nstudies. 14 , 15 The B. subtilis biofilm\nmatrix mainly consists of exopolysaccharide and proteins. 8 , 34 In particular, amyloid fibers formed by the TasA protein are major\nmatrix components, 35 providing structural\nintegrity to B. subtilis biofilms. 36 The absence of TasA in previous MALDI MSI studies 14 , 15 may be due to the low desorption/ionization efficiency of this amyloid\nprotein from thick biofilms, or its formation into high molecular\nweight insoluble fibers. In addition, the typical sample preparation\nsteps for analyzing amyloids from thin sections of mammalian tissues 37 may not be suitable for studying biofilms, as\nrinsing and tryptic digestion may cause biofilm flaking and dislocation\nof small metabolites. Therefore, instead of targeting matrix molecules,\nwe focused on cell-type-specific gene expression to visualize spatial\ndistributions of matrix-producing cells. We chose to use a fluorescence\ntranscriptional reporter, P yqxM - CFP, whereby the CFP gene is fused to the promoter of the yqxM-sipW-tasA operon. 26 As this\nTasA-encoding operon is highly expressed in matrix-producing cells,\nthis subpopulation becomes CFP positive during biofilm development\nin strains harboring the P yqxM - CFP reporter. 26 We cultivated\ncolony biofilms of a B. subtilis NCIB3610 strain\nintegrated with P yqxM - CFP, 26 and used flow cytometry to measure\nCFP expression at the single cell level. Compared to the negative\ncontrol strain without P yqxM - CFP, the reporter strain (herein referred to as WT) exhibited a 62\n± 25% (mean ± SD) increase in CFP fluorescence ( Figure 2 ), confirming the\nfunctional reconstitution of the genetic reporter. Figure 2 Flow cytometric analysis\nof B. subtilis colony\nbiofilms with the P yqxM -CFP reporter. B. subtilis strains were cultured on MSgg agar media\nfor 24 h before harvest. Colony biofilms were dispersed into single\ncells using sonication for analysis via flow cytometry. Control denotes\nthe NCIB3610 strain without the reporter. WT denotes the NCIB3610\nstrain integrated with a P yqxM - CFP cassette. Two gene deletion strains based on WT are\ndenoted as spo0A- and abrB-. (A) Histogram of the flow cytometry.\n(B) Mean fluorescence intensities of 10 000 cells. CFP intensities\nwere normalized to the control. Error bars indicate the SDs of three\nbiological replicates. P values were calculated using\nthe independent two-tailed, two-sample t -test for\nequal sample sizes and equal variance: 0.042 (WT/Control), 0.944 (spo0A-/Control),\nand 0.013 (abrB-/Control). We then explored whether mutations in genetic regulators\nmay affect\nmatrix-related gene expression. Formation of B. subtilis colony biofilms is controlled by a complex regulatory network in\nresponse to environmental stimuli. 8 , 38 In particular,\nSpo0A and AbrB are two important transcriptional factors modulating\nexpression of many biofilm-related genes. It has been reported that\nmutations of the spo0A and abrB genes\ncan result in defective and enhanced biofilm phenotypes, respectively. 9 , 26 , 39 We disrupted the spo0A or the abrB gene in the B. subtilis reporter strain, and found that these gene deletions exerted opposite\neffects on CFP intensity. Whereas Δ spo0A deletion\ncompletely abolished CFP signals, Δ abrB increased\nCFP signals substantially ( Figure 2 ). These observations correlate well with the biological\nfunctions of Spo0A and AbrB. As Spo0A and its phosphorylated forms\nare required for initiating matrix production, 40 the P yqxM promoter remains\nsilent in the absence of Spo0A. AbrB inhibits gene expression from\nP yqxM , and hence, the P yqxM promoter can be activated upon removal of its genetic repressor\nAbrB. 41 Accordingly, a genetic fluorescence\nreporter was successfully constructed to monitor gene expression in B. subtilis . Analysis of Colony Biofilms Using MALDI MSI\nand Fluorescence\nMicroscopy With an optimized MALDI MSI protocol and a functional\nfluorescence reporter, we sought to combine MS and fluorescence imaging\nto characterize both chemical and genetic heterogeneity in B. subtilis biofilms. Three B. subtilis strains harboring the P yqxM - CFP reporter were examined: the WT, Δ spo0A , and Δ abrB strains. When cultivated on the\nbiofilm-promoting MSgg medium, the WT formed colony biofilms with\ncharacteristic wrinkles ( Figure S1 ), and\nCFP intensity was higher at the edge of biofilms than in the central\nregion ( Figure 3 ).\nCompared with the wild type NCIB3610 strain, the recombinant NCIB3610\nstrain integrated with the P yqxM - CFP reporter exhibited no observable changes in biofilm\nmorphology ( Figure S1 ) or metabolite distribution\n( Figure 1 and Figure 3 ). On the other hand,\nthe mutant lacking spo0A formed unstructured colonies,\nand the mutant lacking abrB produced biofilms with\nhyper-wrinkly structures ( Figure S1 ). Figure 3 Fluorescence\nimaging and MALDI MSI of B. subtilis biofilms\nof NCIB3610 and its mutants integrated with the P yqxM -CFP reporter. CFP images were acquired using\na fluorescence stereoscope before MALDI analysis. Each individual\ncolumn represents (left to right): m / z 1030 (surfactin-C13, [M + Na] + ); m / z 1044 (surfactin-C14, [M + Na] + ); m / z 1058 (surfactin-C15, [M + Na] + ); m / z 1485 (plipastatin-C16-Ala, [M + Na] + ); m / z 1499 (plipastatin-C17-Ala,\n[M + Na] + ); m / z 1513\n(plipastatin-C16-Val, [M + Na] + ); m / z 1527 (plipastatin-C17-Val, [M + Na] + ); m / z 2782 (SKF, [M + H] + ), m / z 3422 (subtilosin, [M + Na] + ), m / z 4334 (SDP, [M + Na] + ), and an overlay of ion images (green: SKF; bright pink:\nSDP). The ion intensity is reflected by the intensity of colors. Each\ncolumn of ions is displayed using the same intensity scale, optimized\nper each metabolite and normalized to the TIC. Scale bar = 2 mm for\nfluorescence images and 5 mm for optical and ion images. Consistent with the flow cytometry results ( Figure 2 ), biofilms of the\nΔ spo0A and Δ abrB strains\nshowed substantially reduced\nand enhanced CFP fluorescence compared with the WT, respectively ( Figure 3 ). Furthermore, Δ spo0A and Δ abrB also exerted different\neffects on metabolite abundance and distribution. For surfactins,\nplipastatins, subtilosin, and SDP, the Δ abrB mutant exhibited no observable changes relative to the WT, and the\nabundance of SKF was enhanced in Δ abrB compared\nto the WT ( Figure 3 ). On the other hand, while surfactins of the WT were observed both\nin agar and cells, surfactins of the Δ spo0A mutant were mostly confined within the biofilm region ( Figure 3 ). These results\nare in agreement with previous findings that surfactin secretion is\nabsent upon spo0A deletion. 42 Also, the abundances of plipastatins, subtilosin, SKF, and SDP were\ngreatly reduced by Δ spo0A deletion ( Figure 3 ), consistent with\nobservations in earlier reports. 14 Notably,\nwhile the Δ spo0A and Δ abrB strains exhibited moderately decreased growth rates compared with\nWT in liquid LB media ( Figure S3A ), cell\ngrowth was significantly ( p < 0.001) impaired\non MSgg agar media for both mutants relative to WT ( Figure S3B ). Therefore, it may be reasonable to attribute\nenhanced SKF signals in the Δ abrB biofilms\nto increased biosynthesis, but it was unclear whether reduced metabolite\namounts in the Δ spo0A biofilms resulted from\ndecreased cell numbers or suppressed biosynthesis ( Figure S3B ). Our results confirm the current thought\non the regulatory functions\nof Spo0A and AbrB in biofilm development. Spo0A is a master transcriptional\nregulator that modulates expression of many genes involved in biofilm\nformation and sporulation. 39 , 40 A single aspartate\nresidue of Spo0A is subjected to phosphorylation modification, and\nthe cellular concentration of Spo0A∼P (phosphorylated Spo0A\nprotein) determines a specific transcriptomic state. 43 For example, intermediate Spo0A∼P concentrations\nactivate matrix production, but high concentrations trigger sporulation.\nHence, diverse Spo0A∼P levels in a cell population over the\ntime course of biofilm formation enable temporospatial cellular differentiation. 43 Consistent with this model, unstructured colonies\nwere formed by the Δ spo0A strain, suggesting\ncellular differentiation was absent. Also, expression from the P yqxM promoter ( Figure 2 ) was blocked upon the Δ spo0A deletion, indicating that the pathway for matrix synthesis is subjected\nto genetic regulation by Spo0A∼P levels. For AbrB, it\nrepresses a number of genes, including those involved\nin biofilm formation via direct association with promoter sequences. 44 AbrB is negatively regulated by Spo0A∼P\nvia both transcriptional repression 45 and\nallosteric inhibition. 46 When environmental\nsignals trigger phosphorylation of Spo0A∼P, cellular AbrB abundancy\nis reduced, which in turn activates matrix gene expression and biofilm\nformation. 41 In the Δ abrB strain, hyper-wrinkly biofilms were formed ( Figure S1 ), consistent with the role of AbrB in the negative\nregulation of biofilm development. For the P yqxM reporter, fluorescence intensity was greatly enhanced by Δ abrB ( Figure 2 and Figure 3 ), in\nagreement with previous reports that the yqxM-sipW-tasA operon is a repression target of AbrB. 41 Compared with the WT, SKF abundance was greatly increased in the\nΔ abrB strain ( Figure 3 ), suggesting that repression of the SKF\npathway by AbrB was not fully relieved in the WT strain when cultivated\nusing MSgg. In contrast, production of the other metabolites observed\ndid not exhibit obvious changes between the WT and the Δ abrB mutant, indicating that the activation of corresponding\npathways was no longer limited by AbrB in the WT when cultivated on\nMSgg. However, caution needs to be taken when interpreting quantitative\nchanges among different strains in the fluorescence and MALDI imaging\nresults, which can be affected by both total cell numbers and protein/metabolite\nabundance in individual cells. In particular, as cell growth was substantially\nimpaired in the Δ spo0A mutant relative to WT\non MSgg agar ( Figure S3 ), it was difficult\nto conclude whether decreased CFP and MS signals ( Figure 3 ) resulted from reduced cell\nnumbers or suppressed biosynthesis, or both, based solely on the imaging\nresults. Several approaches can be used to alleviate this concern.\nFirst, if changes in fluorescence/MS imaging signals and cell amounts\nare in opposite directions, it may be reasonable to speculate modified\nbiosynthetic productivity. In this study, it is highly likely that\nenhanced CFP signal and SKF abundance in the Δ abrB strain ( Figure 3 )\nwere due to increased biosynthesis, as there were fewer cells in the\nmutant colony biofilms compared with WT ( Figure S3B ). Second, complementary approaches can be used to provide\nsingle cell-level information. For example, flow cytometry measures\nfluorescence intensities of individual cells, and therefore, one may\nconclude that the Δ spo0A and Δ abrB mutations changed CFP transcription from the P ypxM promoter ( Figure 2 ). Third, other analytical techniques more\namenable to quantitative analysis than MALDI imaging (such as liquid\nchromatography) can be performed to estimate aggregated averages of\nmetabolite abundance in individual cells. Together, as the TIC normalization\nmethod for processing MALDI imaging data does not account for cell\nnumber variations, it requires careful assessment when comparing metabolite\nabundance results between different microbial strains. It is\nalso important to note that the combined use of MALDI and\nfluorescence imaging supports insights on cannibalism and matrix production\nin B. subtilis biofilm development. Previously,\nmatrix-producing and cannibal cells were considered the same population. 11 The main evidence for this assertion was obtained\nusing transcriptional reporters, whereby biosynthesis pathways responsible\nfor TasA and SKF production were coordinately activated in the same\nsubpopulation. 11 In contrast, by using\nMALDI and fluorescence imaging we observed that in WT biofilms, SDP-,\nSKF- and matrix-producing cells exhibited distinct spatial patterns\n( Figure 3 ), suggesting\nfurther differentiation in the cannibal/matrix-producing subpopulation.\nSuch differentiation may be mediated by AbrB, as the distributions\nof SDP-, SKF-, and CFP-positive cells were almost overlapping in Δ abrB biofilms ( Figure 3 ). This discrepancy may be a result of the analytical\nmethods used in the other study, 11 which\ndispersed biofilms into single cells before flow cytometric analysis,\nand so did not yield spatial information. We believe that even if\nfluorescence microscopy were to be used to monitor the spatial activities\nof multiple transcriptional reporters, it still cannot match the information-rich\ndata from our combined method. Given the broad bandwidths of emission\nspectra, only two transcriptional reporters are commonly used in a\nsingle B. subtilis strain. 10 , 26 Therefore, visualization of three different subpopulations in B. subtilis biofilms using fluorescence reporters required\nconstruction of at least two different strains. 10 In contrast, the mass resolution of MALDI allows detection\nof a myriad of preselected and unanticipated compounds from a biofilm,\nminimizing the effort in strain creation and reducing sample-to-sample\nvariations as compared to using fluorescence microscopy alone."
} | 7,088 |
27562316 | PMC5072199 | pmc | 4,291 | {
"abstract": "Summary Considering wide utilization and high methane fluxes from anaerobic biological stabilization ponds ( ABSP s), understanding the methanogenesis in ABSP s is of fundamental importance. Here we investigated the variation and impact factors of methanogenesis in seven ABSP s that spanned from the north to the south of China. Results showed that methanogen abundance (7.7 × 10 9 –8.7 × 10 10 copies g −1 dry sediment) and methanogenic activities (2.2–21.2 μmol CH \n 4 g −1 dry sediment h −1 ) were considerable for all sediments. Statistical analysis demonstrated that compared with other factors (ammonium, pH , COD and TOC ), mean annual temperature ( MAT ) showed the lowest P value and thus was the most important influencing factor for the methanogenic process. Besides, with the increasing MAT , methanogenic activity was enhanced mainly due to the shift of the dominant methanogenic pathway from acetoclastic (49.8–70.7%) in low MAT areas to hydrogenotrophic (42.0–54.6%) in high MAT areas. This shift of methanogenic pathway was also paralleled with changes in composition of bacterial communities. These results suggested that future global warming may reshape the composition of methanogen communities and lead to an increasing methane emission from ABSP s. Therefore, further research is urgently needed to globally estimate methane emissions from ABSP s and re‐examine the role of ABSP s in wastewater treatment.",
"conclusion": "Conclusion To our knowledge, this is the first study to investigate the variation of methanogenesis with a MAT gradient in ABSPs sediments. A high methanogenic activity and abundance of methanogens were observed in the ABSPs sediments. MAT was verified to be the most important factor on methanogenesis. Higher methanogenic activities were obtained in HM compared with LM areas. The acetoclastic methanogenic pathway dominated in LM areas while the hydrogenotrophic pathway prevailed in HM areas. All of these results suggested that methane emission happened in these uncovered ponds and more methane would be emitted in response to the future global warming. Further studies are urgently needed to estimate methane emission from ABSPs all over the world and adjust the application of ABSPs in wastewater treatment to prevent more methane emission.",
"introduction": "Introduction Anaerobic biological stabilization ponds (ABSPs) are widely used for wastewater treatment around the world due to their simple operation, effective cost and low maintenance requirements. Particularly, in regions where land is readily available (Kivaisi, 2001 ), such as non‐metropolitan and regional USA, Australia, central Europe and China, ABSPs are the preferred wastewater treatment process. For swine farms located far from the city with large areas of available land, ABSPs have been recommended as the most effective wastewater treatment process for the removal of chemical oxygen demand (COD) (Liu et al ., 2014 ). However, a considerable amount of greenhouse gas could emit from these uncovered ponds. It has been reported that methane flux of ABSPs in North America (Adler, 1994 ), New Zealand (Park and Craggs, 2006 ), French Mediterranean coast (Picot et al ., 2003 ) and Mexico (Hernandez‐Paniagua et al ., 2014 ) ranged from 7.7 to 64 g m −2 d −1 , which were much higher than that of paddy soils (0.092–1.1 g m −2 d −1 ) (Seiler et al ., 1983 ) and wetlands (0.048–0.71 g m −2 d −1 ) (Baker‐Blocker et al ., 1977 ). Moreover, a long‐term input of high concentration of COD and an anaerobic condition of ABSPs sediments could be expected to promote the degradation of organic matters (Toprak, 1995 ), which would lead to an increase of methane emissions and thereby accelerated global warming. Despite the environmental and climate importance of this process, the composition of functional microbial community structure, especially methanogens in ABSPs is still unclear. What is more, little is known about the influence of biotic (syntrophic‐related bacteria) (Fotidis et al ., 2014 ) and abiotic drivers on the methanogenesis in ABSPs. Temperature, in particular, has been implicated as a major factor in methane emission and methanogenesis. Based on a database of 1553 measurements of methane emission and temperature across wetlands, rice paddies and aquatic ecosystems, meta‐analysis revealed that both methane emission and ratio of methane to CO 2 emission markedly increased with increasing seasonal temperature (Yvon‐Durocher et al ., 2014 ). In addition, this temperature dependence was consistent with the catabolism of methanogens and composition of methanogen communities. First, the precursors of methanogenesis were accumulated in the peatland during colder periods of the year and then metabolized via acetoclastic methanogenesis during warmer seasons (Shannon and White, 1996 ). The turnover rate of acetate increased from 3 to 26 nmol l −1 h −1 with temperature increasing from 4 to 25°C (Kotsyurbenko et al ., 2004 ). Second, the composition of methanogenic community was affected by temperature variation: relative abundance of Methanocella increased at 15°C while that of Methanoregula and Methanosaeta increased at 5°C (Schmidt et al ., 2015 ). However, previous studies mainly focused on paddy soil, wetland and uplands. Systematic research into the variation of methane emission and methanogenesis under different temperature conditions in ABSPs does not appear to have been carried out. To date, only three studies mentioned the temperature conditions in their research about methane emission from ABSPs (Picot et al ., 2003 ). About 2.3‐ to 5.2‐fold higher methane flux was observed in the ABSPs with an average temperature of 37°C than 15°C (Park and Craggs, 2006 ). It could be postulated that elevated temperature would also increase the methane emission and impact on the methanogenesis in ABSPs. However, further research is needed to confirm this influence and explore the underlying mechanisms – variability of methanogenic community diversity and structure in response to temperature change. China spans a latitude gradient of 50° and covers a wide range of climate zones with a large mean annual temperature (MAT) gradient from north to south. This MAT gradient provides an ideal environment to investigate the distribution of microbial community structure in ABSPs under different MAT conditions. In this study, sediment and water samples were collected from ABSPs for piggery wastewater treatment across six provinces with different latitudes in China in 2014 (Fig. 1 A). From north to south, sampling sites were located in Beijing (BJ, N40°12′0.14″ E116°33′45.18″), Shandong (SD1, N36°04′25.1″ E116°43′08.2″ and SD2, N36°03′24.7″ E116°38′26.1″), Zhejiang (ZJ, N30°24′40.22″ E119°54′2.90″), Sichuan (SC, N30°14′31.54″ E103°34′38.14″), Fujian (FJ, N24°24′44″ E118°04′07″) and Guangdong (GD, N21°52′4.94″ E111°57′55.76″), with significantly different MATs (Fig. 1 B) (China Meteorological Administration, 2004‐2013 ). Based on the differences in MAT (more than 5°C), these sites were divided into a low‐MAT (LM) area (SC, BJ, SD1 and SD2) and a high‐MAT (HM) area (ZJ, FJ and GD). In addition, the relationship between environmental variables (i.e. pH, ammonium, nitrite, nitrate, COD, TOC, water content and MAT) and microbial community distribution was investigated in order to elucidate whether MAT was an influential environmental factor. These results are important to our understanding of methanogenesis in ABSPs along a MAT latitude gradient. Moreover, even though some research concerning the influence of temperature on methane emission has been reported (Toprak, 1995 ), the research on the underlying mechanism, especially microbial community composition related to the methanogenesis is still lacking. Further study about microbial community related to methanogenesis will help us better estimate the influences of climate change on the methane emission from ABSPs in the future. Figure 1 Sampling sites. (A) The geographical location (this figure was generated using Microsoft Powerpoint 2013 and Adobe Photoshop 7.0). (B) MAT (12.9–20.6°C).",
"discussion": "Results and discussion Methanogenic activity in ABSPs Serum bottle incubation experiments showed that the mean methanogenic activity of all sediment samples was 7.33 μmol CH 4 g −1 dry sediment h −1 , ranging from 2.2 to 21.25 μmol CH 4 g −1 dry sediment h −1 . Compared with the mean methane oxidation activity (0.7 μmol CH 4 g −1 dry sediment h −1 ), methanogenic activity was 10.1‐fold higher (Table 1 ). Compared with paddy soil (< 0.02 μmol CH 4 g −1 dry sediment h −1 ) (Dong et al ., 2013 ), ABSPs sediments in our study also showed a much higher methanogenic activity. Table 1 Physicochemical characteristics of the sediment samples Sediment sample a \n Ammonium (g kg −1 ) Nitrite (mg kg −1 ) Nitrate (g kg −1 ) TOC (g kg −1 ) pH Water content Methanogenic activity b \n Methane oxidation activity b \n MAT (2004–2013) LM SC 3.38 ± 0.24 0.41 ± 0.06 0.08 ± 0.01 154.17 ± 9.44 8.30 ± 0.11 0.86 ± 0.01 2.83 ± 0.12 0.46 ± 0.05 12.99 ± 0.67 BJ 8.77 ± 0.02 0.22 ± 0.01 0.17 ± 0.04 202.45 ± 0.28 8.28 ± 0.11 0.88 ± 0.01 2.20 ± 0.07 0.49 ± 0.02 13.20 ± 0.57 SD1 3.58 ± 0.09 0.55 ± 0.02 0.09 ± 0.01 170.65 ± 9.19 8.04 ± 0.09 0.87 ± 0.03 2.91 ± 0.09 0.44 ± 0.03 13.50 ± 0.39 SD2 2.42 ± 0.04 0.28 ± 0.01 0.05 ± 0.00 173.22 ± 3.85 7.95 ± 0.06 0.85 ± 0.03 2.82 ± 0.19 0.62 ± 0.02 13.50 ± 0.39 HM ZJ 10.44 ± 2.06 1.84 ± 0.05 0.06 ± 0.00 186.70 ± 1.41 7.79 ± 0.07 0.91 ± 0.02 21.25 ± 3.97 0.77 ± 0.23 18.48 ± 0.59 FJ 7.36 ± 0.13 3.80 ± 0.06 0.05 ± 0.01 108.70 ± 6.78 8.34 ± 0.11 0.92 ± 0.01 8.92 ± 0.06 1.17 ± 0.07 20.50 ± 0.82 GD 12.65 ± 0.21 3.62 ± 0.64 0.11 ± 0.00 254.32 ± 5.16 8.46 ± 0.11 0.92 ± 0.02 10.38 ± 2.74 1.11 ± 0.16 22.60 ± 0.71 a Means ± standard deviation. b The measurement unit is μmol CH 4 g −1 dry sediment h −1 . John Wiley & Sons, Ltd The high methanogenic activities in ABSPs sediments can be attributed to the following reasons. First, the high carbon and nitrogen concentration of water and sediments could provide enough substrate for methane production (Zitomer and Shrout, 2000 ). Physicochemical analyses revealed that almost all of the ABSPs had a high content of COD (382.3–973.3 mg l −1 ) (Table S1) and high ammonium (166.7–715.1 mg l −1 ) in water samples and high total organic carbon (108.7–254.3 g kg −1 ) and ammonium (2.42–12.65 g kg −1 ) (Table 1 ) in sediments. Additionally, the low dissolved oxygen concentrations (0.15–0.23 mg l −1 ) (Table S1) of the ABSPs water suggested that an anaerobic environment existed in the ABSPs, which provides a suitable habitat for methanogens (Großkopf et al ., 1998 ). As expected, Illumina Miseq sequencing of archaea targeting V5‐V6 region revealed that methanogens accounted for more than 82.7% of the total archaea. Meanwhile, a small fraction of methane‐oxidizing archaea (< 0.07%) and bacteria (NA) was observed. qPCR results (Fig. 2 , Table S2 & Fig. S2) showed that the abundances of 16S rRNA gene of methanogens in sediments and water were 1.89 × 10 4 –2.10 × 10 5 ‐fold and 2.28–10.55‐fold higher than pmoA gene, respectively, which further demonstrated that the methanogenic process was predominant. In addition, the abundances of 16S rRNA gene of methanogens in the pond sediments were higher than observed in UASB sludge granules when treating swine wastewater (1.9 × 10 7 –5.7 × 10 7 copies ml −1 ) (Song et al ., 2010 ), anaerobic batch digesters (2.8 × 10 7 –4.5 × 10 8 copies ml −1 ) (Lee et al ., 2009 ) and fixed‐bed anaerobic digester (2.8 × 10 7 –7.6 × 10 9 copies ml −1 ) (Sawayama et al ., 2006 ). Totally, the high methanogenic activity and abundance of methanogens in our study suggested that considerable amount of methane would be emitted from the ABSPs for swine wastewater treatment. In addition, the activities in our research were not measured in situ but in the laboratory. Therefore, the measured activities might be different from the real methane emissions from ABSPs, which is recommended to be measured in the future. Figure 2 Abundance of 16S rRNA gene of methanogens and pmoA gene (copies g −1 dry sediment) in the sediment samples. Error bars indicate standard deviation of the mean of triplicate qPCR reactions. Microbial community diversity A total of 172 993 and 205 918 sequences were obtained using the archaeal V5‐V6 primers and bacterial V4 primers respectively. The approach of an asymptote in the rarefaction curve indicated that the archaeal and bacterial community were well captured at the sequencing depth (Fig. S1). Besides, the non‐parametric statistics analyses (Table 2 ) showed that the observed sequences covered 46–52% and 62–79% of the total archaeal and bacterial sequences. Table 2 Richness and diversity of archaea and bacteria in the sediment samples collected from ponds located from northern to southern parts of China (estimated by 97% OTU clusters) Sediment sample Archaea Bacteria Richness estimators Diversity estimators Richness estimators Diversity estimators S obs \n a \n Chao ACE Chao/ACE Simpson Shannon % Inv. Comp. b \n S obs \n a \n Chao ACE Chao/ACE Simpson Shannon % Inv. Comp. b \n LM SC 185 401.0 433 0.92 0.44 1.38 46 835 1322 1353 0.98 0.06 3.85 63 BJ 263 554.6 583 0.95 0.46 1.67 47 628 1018 963 1.06 0.13 3.17 62 SD1 266 542.3 561 0.97 0.30 1.92 49 1109 1549 1564 0.99 0.04 4.43 72 SD2 275 530.0 545 0.97 0.29 2.01 52 1055 1405 1446 0.97 0.04 4.30 75 HM ZJ 287 621.4 836 0.74 0.16 2.59 46 1457 1927 1931 1.00 0.02 5.21 76 FJ 287 622.6 739 0.84 0.14 2.84 46 1665 2104 2144 0.98 0.01 5.38 79 GD 267 582.4 700 0.83 0.19 2.29 46 1041 1427 1427 1.00 0.05 4.41 73 a Observed richness. b Total observed richness/Chao estimate ×100. John Wiley & Sons, Ltd The microbial richness and diversity of the sediment samples from different ponds were compared using the Shannon, Simpson, S obs , Chao and ACE index (Table 2 ). For archaea and bacteria, the highest richness and diversity were both found in sediments collected from FJ and ZJ. The microbial richness and diversity increased with the increasing MAT. Furthermore, based on the explanation of 89% and 48% of total variation, the distribution of principal coordinate analyses (PCoA) plots (Fig. 3 ) revealed that the distributions of archaeal and bacterial communities in the sediment samples area were similar and could be clustered into LM and HM areas respectively. Figure 3 Principal coordinate analyses results of archaeal and bacterial communities in the sediment samples. Bacterial community structure A total of 20 major bacterial phyla (> 0.2%) were identified in all sediment samples, including Firmicutes , Bacteroidetes , Proteobacteria , Chloroflexi , Planctomycetes , Verrucomicrobia , Actinobacteria , Spirochaetes , OP8 , Synergistetes , OD1 , Caldithrix , OP9 , Hyd24‐12 , NKB19 , H‐178 , Tenericutes , Lentisphaerae , Acidobacteria and Armatimonadetes . The three predominant phyla in all sediment samples were assigned to Firmicutes (21.20–80.45%), Proteobacteria (5.43–18.46%) and Bacteroidetes (3.34–40.92%) (Fig. 4 & Table S3). The percentage of Firmicutes (mainly the families of Clostridiaceae and Peptostreptococcaceae) decreased while Proteobacteria (mainly the class of δ‐ Proteobacteria in the sediment samples from HM area and mainly the classes of β‐ Proteobacteria , γ‐ Proteobacteria and ε‐ Proteobacteria in the sediment samples from LM area) and Bacteroidetes (mainly the classes of Bacteroidia , Flavobacteriia and Sphingobacterii ) increased with increasing MAT. In contrast, Actinobacteria and NKB19 only existed in the sediment samples from LM area, while OD1 , Spirochaetes and Caldithrix were only found in the sediment samples from HM area. Synergistetes showed a higher percentage in the sediment samples from LM area, whereas Verrucomicrobia was assigned a larger percentage in the sediment samples from HM area. Figure 4 Comparison of bacterial community among sediment samples from the different sampling sites. The area of the circle represents the percentage of the species. Different colours show the sediment samples collected from different mean annual temperature areas. Archaeal community structure In accordance with the results of PCoA, the sediment samples shared similar archaeal community structure for HM area and LM area. Four different archaeal phyla including Euryarchaeota , Parvarchaeota , Crenarchaeota and a newly discovered Thaumarchaeota were detected in all sediment samples. Euryarchaeota (> 55%) was the most abundant phylum of archaea among all those sediment samples and this phylum showed an especially high abundance in the sediment samples from LM area (> 85.5%). Parvarchaeota showed a higher percentage in the sediment samples from HM area (> 14.34%) than LM area, while Crenarchaeota (4.5%) and Thaumarchaeota (0.04%) were evenly distributed in all sediment samples. More differences of the four achaeal phyla distribution were observed among sediment samples from different MAT area at deeper levels (Fig. 5 & Table S4). At the class level, all the observed archaea were classified into 12 classes. Methanomicrobia was the predominant class in all sediment samples, which showed a decreasing tendency with increasing MAT (> 84.5% and > 44.7% of archaea in the sediment samples from LM and HM areas respectively). In the sediment samples from HM area, Parvarchaea (> 14.2%), Thermoplasmata (> 7.9%) and Methanobacteria (> 1.8%) were widely observed while they showed much lower percentages in the sediment samples from LM area (< 5.74%, 1.55% and 0.69% respectively) than HM area. MCG was present in all sediment samples, which accounted for more than 2% of all the archaeal members. Many other little fraction classes (< 0.34%), including Micrarchaea , DSEG , Nitrososphaerales and Thermoprotei , were mostly found in the sediment samples from HM area. Aigarchaeota , Halobacteria and Methanococci were identified only in the sediment samples from HM area. At the order level, a total of 17 archaeal orders were identified in all sediment samples. On average, seven dominant orders (> 0.2%) belonged to Methanosarcinales (28–70%), Methanomicrobiales (13.5–34.5%), WCHD3‐30 (1.14–27.1%), Thermoplasmatales (0.68–11.4%), pGrfC2 (2.16–6.65%), Methanobacteriales (0.37–5.34%) and Micrarchaeles (0–0.34%). The percentages of Methanosarcinales , Methanomicrobiales and pGrfC2 decreased with the increasing MAT. The opposite tendency was observed in the orders of WCHD3‐30, Thermoplasmatales and Methanobacteriales . The orders of ArA07 and F99a103 only existed in the sediment samples from HM area while YC_E6 only existed in LM area. Nitrosophaeraceae (0.01–0.08%), unclassified Thermoprotei (0–0.07%), Methanocellales (0–0.04%) and Cenarchaeales (0.01–0.03%) were evenly distributed in all sediment samples. At the family level, 16 identified families and 10 unclassified families were present in all sediment samples. Methanosaetaceae (49.5–70.46%) and unclassified Methanomicrobiales (12.15–28.32%) were the predominant families and showed a higher percentage in the sediment samples from LM area than HM area, while unclassified WCHD3‐30 (13.99–27.14%) and Methanomassiliicoccaceae (7.84–11.39%) showed a higher percentage in the sediment samples from HM area. ANME‐2D and ANME‐2c were also detected in all sediment samples at this level. Figure 5 Comparison of archaeal community structure in the sediment samples collected from LM (left) and HM (right) area. For each figure, the archaeal composition at phylum, class, order, family and genus levels are shown from outside to inside of the circle. Each colour represents one phylum except the red one, which represents the typical species or obviously more abundant species than another. Controls on methanogenesis in ABSPs sediments All the results confirmed our hypothesis that MAT played the most important factor in impacting on methanogenesis in ABSPs sediments. First, serum bottle incubation experiments and qPCR analysis suggested that the methane emission might be increased with an elevated temperature. A 5‐fold higher methanogenic activity was observed in the sediment samples from HM area compared with those from LM area (Table 1 ). Nevertheless, only 2.01‐fold higher methane oxidation activity was observed in the sediment samples from HM area than those from LM area. From the perspective of functional microbes abundance, the abundance of 16S rRNA gene of methanogens in the sediment from HM area were 4.74‐fold larger than LM area, while only 1.88‐fold larger was observed for the abundance of pmoA gene (Fig. 2 ). In the water, the abundance of 16S rRNA gene of methanogens from HM area were 21.11‐fold larger than LM area, while only 13.41‐fold larger was observed for the abundance of pmoA gene (Fig. S2). In general, a much stronger increasing tendency was observed for the methane production of ABSPs, which implied that an increase in MAT more strongly favoured methanogenesis than methanotrophy. Elevated temperature might increase the methane emission from ABSPs through the influence on activities and abundance of functional microbes. Second, statistical analysis including Pearson's correlation analysis and archaeal redundancy analysis (RDA) demonstrated that MAT played the most important factor in impacting on methanogenesis in ABSPs sediments for the lowest P value compared with other factors. Pearson's correlation analysis showed that ammonium of pond water and COD of influent were significantly correlated with the abundance of 16S rRNA gene of methanogens ( P < 0.05) (Table S5). However, compared with ammonium and COD, MAT presents a stronger significant correlation with the abundance of 16S rRNA gene of methanogens ( P < 0.01) (Table S5). Archaeal RDA analyses indicated that archaeal community structures of sediment samples collected from LM and HM area were significantly different (Fig. 6 ). Among all the environmental factors, the lowest P value of MAT ( P = 0.012) indicated that MAT was the most important determinant of the archaeal community structure. Figure 6 Results of redundancy analysis ( RDA ) at the family level of archaeal community structure. The blue solid dots represent the dominant families in the sediment samples from LM area. The red ones represent the dominant families in the sediment samples from HM area. The analyses indicated that archaeal community structures of sediment samples collected from LM and HM area were significantly different. The environmental factors in the first two RDA axes explained 84.2% and 10.3% of the total variation of archaeal community structure. Among all the environmental factors, MAT ( P = 0.012), water content ( P = 0.018) and nitrite concentration ( P = 0.028) of sediment samples significantly impacted on the archaeal community structure. Besides, methanogenic community distribution analyses revealed that methanogenic pathways shifted in the sediment samples from different MAT areas (Fig. 7 ). Acetoclastic methanogens (49.8–70.7%) were predominant in the sediment samples from LM area while hydrogenotrophic methanogens (42–54.6%) were mainly dominant in the sediment samples from HM area. Strikingly, MAT also influenced the distribution of bacteria in the sediment samples, which could support the distinct methanogenic pathways. For example, Clostridiaceae , Flavobacteriaaceae , Turicibacteraceae , Anaerolinaceae , Actinomycetales and Synergistales (Fernandez et al ., 2008 ; Overmann, 2008 ; Jumas‐Bilak et al ., 2009 ; Ito et al ., 2012 ; Wang et al ., 2014 ), which were well known for their roles in the fermentation of macromolecular organic matter (lipid, cellulose, propionate, glucose, fatty acid and amino acid) into acetate, showed a higher abundance in the bacterial community of the sediment samples from LM area than HM area. Comamonadaceae and Pseudomonadaceae , which were known to be homo‐acetogens and could provide acetate for acetoclastic methanogens by consuming H 2 and CO 2 (Morrill et al ., 2014 ), were enriched in the sediment samples from LM area. In contrast, in the sediment samples from HM area, more than 31.9% of bacteria mainly belonged to the phylum of Bacteroidetes (Qiu et al ., 2008 ) and Proteobacteria such as Syntrophaceae (Briones et al ., 2009 ) and Syntrophorhabdaceae (Gray et al ., 2011 ), and are known to have a syntrophic relationship with hydrogenotrophic methanogens. Rikenellaceae , Porphyromonadaceae , Bacteroidaceae , Ruminococcaceae , Spirochaetes , OP8 , OD1 , Peptococcaceae , Lentisphaerae and Thermotogacaea (Jehmlich et al ., 2010 ; Cheng et al ., 2013 ; Zhang et al ., 2013 ; Veeravalli et al ., 2014 ; Wang et al ., 2014 ), which accounted for about 14.2% of the bacteria in the sediments from HM area, were able to produce hydrogen. It has also been reported in reactor experiments that hydrogenotrophic methanogens become more important under higher temperature (55°C) with a syntrophic relationship with acetate‐oxidizing bacteria (Guo et al ., 2014 ). Our results further verified this tendency in environmental sediments with a relatively low temperature. Figure 7 Methanogens distribution in each sediment sample. Different colours represent methanogens with different methanogenic pathways. The red ones are acetoclastic methanogens, the green ones are hydrogenotrophic methanogens, green ones are other types of methanogens and other archaea. Overall, it could be concluded that MAT is an important factor to the methanogenesis process in ABSPs sediments. With the increasing MAT, methane emission was enhanced mainly due to the shift of methanogenic pathways being constituted by different fractions of methanogens with different methanogenic activities. In addition, the distribution of bacteria in the sediments, which was impacted by MAT, supported our conclusion that the methanogenic pathways shifted from acetoclastic to hydrogenotrophic with increasing MAT. Ecological significance The high methanogenic activity and the dominant role of methanogenesis in the ABSPs sediments in our study suggested that high biodegradable COD concentration of ABSPs could provide sufficient substrate for methanogens and thus might result in large quantities of methane emission from these uncovered ponds. Therefore, considering the important impact of actual loading rates to methanogenesis (Toprak, 1995 ), it would be better to conduct further research to assess the underlying mechanisms of actual loading rates’ influence on methanogenesis in ABSPs. From the perspective of methane emission control, ABSPs for swine wastewater treatment are not suitable for application in treating anaerobic digestion effluent but more likely to be utilized as the tailwater treatment process. Moreover, ecological regulation analysis (Röling, 2007 ) and specific methanogenic activity (SMA) tests (Enright et al ., 2005 ) both suggested that a much higher methanogenic activity might be obtained when hydrogenotrophic methanogens were dominant in the environment compared with the acetoclastic methanogens. This is in accordance with our study that the sediment samples from HM area were dominated by hydrogenotrophic methanogenic pathway and showed a higher methanogenic activity than LM area. Therefore, methane emission from ABSPs in HM area will be largely different from LM area. In the future, methane emission from the ABSPs for swine wastewater treatment should be estimated according to the differences of temperature in different areas. Finally, we confirmed that elevated temperature could increase the methane emission and impact the methanogenesis in ABSPs sediments. According to IPCC, 2013, the global surface temperature showed a warming of 0.85 (0.65–1.06) °C over the period 1880–2012 and was projected to increase by 2.6–4.8°C under ‘Representative Concentration Pathways (RCP) 8.5’ by the end of the 21st century (Pachauri, 2013 ). Therefore, more methane is expected to be emitted from ABSPs for swine wastewater treatment in response to the increasing temperature, possibly because of a variation of methanogenic pathway in the future."
} | 7,099 |
37072126 | PMC10173695 | pmc | 4,292 | {
"abstract": "Biomolecular nanotechnology has helped emulate basic\nrobotic capabilities\nsuch as defined motion, sensing, and actuation in synthetic nanoscale\nsystems. DNA origami is an attractive approach for nanorobotics, as\nit enables creation of devices with complex geometry, programmed motion,\nrapid actuation, force application, and various kinds of sensing modalities.\nAdvanced robotic functions like feedback control, autonomy, or programmed\nroutines also require the ability to transmit signals among subcomponents.\nPrior work in DNA nanotechnology has established approaches for signal\ntransmission, for example through diffusing strands or structurally\ncoupled motions. However, soluble communication is often slow and\nstructural coupling of motions can limit the function of individual\ncomponents, for example to respond to the environment. Here, we introduce\nan approach inspired by protein allostery to transmit signals between\ntwo distal dynamic components through steric interactions. These components\nundergo separate thermal fluctuations where certain conformations\nof one arm will sterically occlude conformations of the distal arm.\nWe implement this approach in a DNA origami device consisting of two\nstiff arms each connected to a base platform via a flexible hinge\njoint. We demonstrate the ability for one arm to sterically regulate\nboth the range of motion and the conformational state (latched or\nfreely fluctuating) of the distal arm, results that are quantitatively\ncaptured by mesoscopic simulations using experimentally informed energy\nlandscapes for hinge-angle fluctuations. We further demonstrate the\nability to modulate signal transmission by mechanically tuning the\nrange of thermal fluctuations and controlling the conformational states\nof the arms. Our results establish a communication mechanism well-suited\nto transmit signals between thermally fluctuating dynamic components\nand provide a path to transmitting signals where the input is a dynamic\nresponse to parameters like force or solution conditions.",
"conclusion": "Conclusions We established a mechanism to transmit\ninformation between distal\ndynamic components of DNA nanodevices via steric interactions, demonstrated\nwith a dynamic device comprising a base platform and two dynamic arms\nthat can sterically interact. We demonstrated the utility of steric\ninteractions for the modulation of conformational landscapes of distal\ncomponents both by influencing range of motion and distributions of\nbinding states. Our mesoscopic simulations revealed that steric modulation\ncan be accurately predicted from individual arm properties. The experimentally\nvalidated model allowed us to explore the full 2D conformational (and\nfree energy) landscapes, which revealed interesting features, such\nas an accumulation of conformations near the sterically occluded region\nin some designs ( Figure 3 C,D), which would have been challenging to identify with experiments\nalone due to limited sampling. The model developed here can also serve\nas a useful tool for the design of future dynamic devices that leverage\nsteric interactions. We found interactions between the arms depended\non structural, mechanical, and dynamic design parameters. For example,\nthe longer length of the right arm led to an asymmetry in excluded\nangles, and the balance of competing latching interactions depended\non the number and sequence of latching overhangs. Our simulations\nfurther revealed that the steric interactions lead to kinetic modulation\nof binding states through occlusion of binding sites. These\nresults establish the engineering of steric interactions\nas a useful approach to transmit information (e.g., motion or binding)\nbetween distal dynamic components. The fact that these components\nremain dynamic suggests they can still be sensitive to local environmental\ninputs such as ion concentrations, 50 temperature, 51 or forces. 52 As a\nproof of concept, we measured the dynamic latching of several versions\nof the SteriDyn at increased MgCl 2 concentrations, which\ngenerally led to increased overall latching ( Supporting Figures S13 and S14 ). For the LC 3 |RC 3 design, we observed that increasing ion conditions led to the right\narm (i.e., RC 3 latching increased) more effectively outcompeting\nthe left arm (i.e., LC 3 latching decreased), illustrating\nthat the steric communication can be sensitive to the local environment.\nHence, similar designs could convert environmental inputs into steric\nsignals that regulate device functions. The possibility of designing\nsystems with more than two dynamic arms, or even arrays of dynamic\narms, could provide a foundation for leveraging steric interactions\nto perform logic steps 53 or transmit signals\nover longer distances. While such tasks have previously been accomplished\nby means of strand displacement reactions, steric interactions promise\nrobust and faster dynamics over larger length scales and environment\nsensitivity.",
"introduction": "Introduction DNA nanotechnology 1 , 2 has\ngarnered increasing interest\nfor the development of nanoscale robotic systems due to the precise\ncontrol over geometry afforded by this approach and the ability to\ndesign devices with complex motion 3 and\ntunable mechanical properties. 4 In addition\nto these design features, advanced robotic devices and materials typically\nintegrate functions like actuation, sensing, and communication, which\ncombined can enable advanced capabilities like feedback control or\nautonomous behavior. While actuation 5 − 11 and sensing 12 − 16 have been widely demonstrated within the context of DNA nanotechnology, 17 communication—which entails transmitting\ninformation from one part of the system to another—has been\nmuch less studied. Many natural biomolecules have evolved mechanisms\nfor signal transmission. One example of such mechanisms is allosteric\ncommunication within proteins, where local binding events regulate\nfunctions at distal locations through conformational changes or modulation\nof dynamic properties such as amplitudes of motion or conformational\ndistributions. 18 − 21 Inspired by protein allostery, here we develop a mechanism for transmitting\ninformation via steric interactions between dynamic components within\na DNA nanostructure. Prior efforts have implemented approaches\nto transmit signals within\nDNA nanostructures either via transport of components, such as single\nstrands 22 − 24 or larger constructs, 25 − 29 along predefined tracks, or via conformational changes\nof the nanostructure resulting from structurally coupled motions. 30 , 31 The first of these approaches introduces hybridization and dissociation\nof nucleotide sequences as the driving mechanisms for communication. 32 These mechanisms enable very precise communication,\ndue to the sequence specificity of DNA, at the expense of speed and\ndistance. Other studies have leveraged the precise geometric design\nof DNA origami 33 , 34 to demonstrate larger motions\nin nanostructures exhibiting conformational changes involving stiff\nlinks coupled by flexible joints 7 , 8 to achieve kinematically\nconstrained motion or in devices where conformational changes propagate\namong small repetitive structural units. 30 , 35 Such deployable structures allow for the transfer of forces and\nmotions among large DNA components, which have been used to control\nmolecular interactions, 31 , 36 and can be combined\nwith triggering events, such as local enzymatic modifications or binding\nof an input molecule, to regulate device properties at distal locations. 36 , 37 Introduction of such conformational changes enables allosteric communication.\nHowever, structurally coupling the motion of components removes degrees\nof freedom and limits the overall design flexibility. In this\nwork, rather than transferring motion through kinematic\nconstraints or coupling local conformational changes, we establish\nan approach to transmit mechanical signals through the steric interaction\namong thermally fluctuating components. These interactions enable\nfluctuating components to regulate the dynamic properties of distal\nones, such as amplitudes of motion or thermodynamics of binding interactions.\nSteric interactions have recently been demonstrated to constrain rotational\nmotions of tight-fitting components in DNA origami rotary mechanisms, 38 including facilitating processive rotation in\nnanoscale motors. 39 Here we establish the\nability to convey steric interactions between two distal thermally\nfluctuating components. We present a nanodevice comprised of two fluctuating\narms connected to a base platform, where the arms are long enough\nto come into contact. We modulate the steric interactions by engineering\nthe conformational distributions of individual arms and show that\nthe conformational fluctuations of each arm can influence the dynamic\nmotion of the other arm. In addition, we introduce binding interactions\nbetween arms and base platform to further control the conformational\ndistribution of the individual arms and demonstrate that each fluctuating\narm can also modulate the equilibrium binding distributions of the\ndistal one. These results establish steric interactions as a mechanism\nfor transmitting mechanical signals between thermally fluctuating\ncomponents and lay a foundation for allosteric devices where local\nevents regulate dynamic properties at distal locations.",
"discussion": "Results and Discussion Design and Fabrication of a DNA Origami Device with Interacting\nArms A DNA origami nanodevice composed of two fluctuating\narms connected to a base platform via flexible hinges was designed\nin MagicDNA. 40 To allow for steric interactions,\nthe arms were designed to be longer than half the length of the base\nplatform, with lengths of the left arm l L = 41 nm, right arm l R = 46 nm, and base\nplatform arm l C = 57 nm. The arms were\ndesigned with different lengths to study how length influences steric\ninteractions. Each of the three components was made up of an 18-helix\nbundle organized into three layers of six helices. Two additional\nbundle layers were added to the bottom of the left-hand side of the\nbase platform to allow for easy identification of the left and right\narms in images ( Figure 1 A). The flexible hinge connection between both arms and the base\nplatform comprises six ssDNA scaffold linkers. Three of these ssDNA\nconnections were 2-nt-long pieces of scaffold (black strands in Figure 1 A) arranged along\na line defining the hinge axis of rotation, while the other three\nssDNA connections were 30-nt-long DNA strands (blue strands in Figure 1 A) introduced to\nfacilitate tuning of the angular distribution of the individual arms. 41 At the ends of the bundles, poly thymines (poly-T,\neither 5 or 10 T long) strands (orange strands in Figure 1 A) were added to prevent base\nstacking of the arms with the base platform. We refer to this device\nas the “SteriDyn”, since it was designed to exhibit\ndynamic behavior regulated by steric interactions. In addition, the\ndevice contains sites for mutually complementary ssDNA overhangs on\nthe arms and base platform to latch the left or right arms down to\nthe platform (green and red strands). The design can harbor up to\nthree latching overhangs for either arm to achieve a high yield of\nlatching (see Figure 1 B). Structures were folded in 20 mM MgCl 2 (gel electrophoresis\nshown in Figure S3 ) and purified by PEG\ncentrifugation, 42 and conformational distributions\nof thermally fluctuating SteriDyn structures were analyzed via TEM\n( Figure 1 B,C). Figure 1 Design and\nfabrication of the DNA origami device. (A) Isometric\nview of our device with a 41-nm-long left arm, 46-nm right arm, and\n57-nm central arm. The central arm included an additional 6 ×\n2 square lattice structure at the bottom of one of its sides to distinguish\nthe structure orientation. Green and red strands represent the extended\noverhangs introduced to latch the arms to the central arm. The hinge\nconnection is made of six scaffold connections (three black 2-nt-long\nssDNAs and three blue 30-nt-long ssDNAs). Orange strands represent\n5-nt-long poly-T overhangs. (B) TEM images of the devices show three\ntypical conformations: left arm closed (red arrows), right arm closed\n(green arrows), and unlatched arms (black arrow). (C) Schematics and\nTEM images depict an example of the left arm latched by a green strut\n(green circle); an example of both arms open (black circle), with\nour convention for the left-arm angle, θ L , and the\nright-arm angle, θ R , shown; and an example of the\nright arm latched by a red strut (red circle). Scale bar: 50 nm. Hinge and Overhang Properties Control the Free Energy Landscape\nof the Arms Prior studies have shown that the properties\nof the ssDNA linkers in DNA origami hinge joints modulate the flexibility\nof the hinge. 7 , 41 , 43 We first investigated SteriDyn devices with fluctuating arms where\nthe hinge joint scaffold linkers were left single stranded and 5T\noverhangs were included on the ends of the arms and base platform.\nWe observed that this baseline condition favors open conformations\nof the arms (see Figure 2 A and F). Under this “free hinge” condition, the arms\nbehave independently since they both sample large angles where no\nsteric interactions occur. We then changed the properties of each\narm individually, while leaving the other arm free to measure individual\nhinge properties. We first used a hinge design approach inspired by\nprior work, 41 where scaffold linkers were\nmade double-stranded in two sections by extending six staples (three\nfrom the arm and three from the base-platform) by 15 nt to base-pair\nthe scaffold linkers. We also extended the poly-T overhangs from 5T\nto 10T to inhibit base-stacking interactions more strongly. This “constricted\nhinge” condition allowed us to shift the angular distributions\nof each arm toward smaller angles, and thereby more closed conformations,\nwith a mean angle of 74° ( Figure 2 B and F). In addition to these dynamic conformations\nwhere arms can sample a range of angles, we designed sites in both\nthe arms and platforms, which allow for the placement of up to three\noverhangs capable of latching the arms onto the base platform. A single\n8 bp latching overhang leads to a mixture of latched and unlatched\nstates ( Figure 2 C),\nwhile two and three latching overhangs lead to nearly all arms exhibiting\na latched conformation ( Figure 2 D and E). We ran similar experiments to demonstrate control\nover conformations of the right arm with the left arm being in the\nfree configuration ( Figure S6 ). Figure 2 F shows a summary\nof the left and right arm conformations in each condition tested in\nthe absence of steric interactions (i.e., other arm at open angles).\nThe two arms exhibit similar mechanical properties for the designs\ntested. Figure 2 Engineering single arm angular distributions with scaffold connection\nand overhangs. (A) Baseline device with both left and right free arms\n(LF|RF, 160°|153° mean angles) results in open conformations.\n(B) Increasing the poly-T length of the left arm connections from\n5 nt to 10 nt and base pairing these connections with complementary\nstaples shift the arm’s angular distribution toward more closed\nconformations (LC|RF, 74°|150°). (C) Modifying LC|RF by\nintroducing the duplex DNA overhangs that latch the left arm shifts\nthe distribution toward even smaller angles. LC 1 |RF (left),\n(D) LC 2 |RF (middle), and (E) LC 3 |RF (right)\nrepresent arms latched by one, two, and three pair(s) of overhangs,\nrespectively. Angular histograms were normalized by setting a maximum\nbin height of 1. (F) Summary of the hinge angles, θ, for the\nleft (cyan) and right (dashed black line) arms compared for all five\ncases where the opposing arm is in the free configuration (individual\ndistributions for right arm shown in Figure S6 ). Outliers where p < 0.02, where p is the normalized probability, were removed from the distribution\nfor clear comparison (full distributions are shown in Figure S7 ). Sample sizes N for\nLF|RF, LC|RF, LC 1 |RF, LC 2 |RF, and LC 3 |RF are 303, 521, 240, 193, and 180, respectively. Steric Interactions Modify Energy Landscapes of Distal Dynamic\nComponents Leveraging these approaches to engineer individual\narm conformations, we next attempted to engineer steric interactions\nbetween the arms. Geometrically, steric interactions should yield\na region of the 2D angular conformational space where both arms cannot\ncoexist due to steric clash (i.e., a “dead zone” in\nthe conformational space). When both arms can independently adopt angles inside this conformational space, each arm will exclude\nthe other. This steric competition should then modify the angular\ndistributions sampled by the individual arms. We hypothesized that\nthe behavior of SteriDyn devices with interacting arms could be predicted\nfrom individual arm properties ( Figure 1 ) and steric interactions. While coarse-grained models\nsuch as oxDNA can reproduce the mechanical properties and predominant\nmotions of DNA origami structures, 43 − 46 they cannot access the time scales\nrequired to obtain large data sets of possible conformations of complex\nDNA origami nanostructures, such as the SteriDyn. To this end, we\nconstructed a minimalistic mesoscopic model of our structures that\naccurately captures the geometric features and motions involved in\nour system ( Figure 3 A). Figure 3 Angular conformational space of the interacting arms. (A) Schematic\nof our coarse-grained model and angle convention. The central arm\n(pink beads) of our device is functionalized to the two mobile arms\n(cyan beads) by means of physical hinges (involving blue and green\nbeads as well as the closest arm beads) as shown in Figure S8 . (B–F) Experimental counts (crosses) and\nsimulated energy landscapes (color) of left ( x axis)\nand right ( y axis) angles for different systems of\ninteracting arms. The area inside the dashed line represents the boundary\nof the “dead zone”. Experimental counts were clustered\nby means of a distance cutoff (7.5°), and the sizes of the crosses\nare logarithmically associated with the number of points in each cluster.\nClusters were built from largest to smallest sizes to maximize the\nnumber of points in each cluster. Energy landscapes were constructed\nby Boltzmann inversion from 15 simulations, where measurements were\ntaken every 100,000 steps on a total of 100 million steps, and smoothed\nby means of a mean filter over a rectangle of size 5° by 5°.\nA single visit was associated with angular conformations not sampled\nin our simulations. Experimental results are also shown as histograms\nin marginal plots. Dashed lines are the reference for hinges experiencing\nno steric effects (open opposite arms). Sample sizes in experiment\nfrom B to F are 521, 383, 264, 356, and 232, respectively. Briefly, our mesoscopic model treats the interacting\narms as rigid\nbodies tethered to a rigid platform and capable of rotating around\nthe hinge attachment points. The arms and the base platforms consist\nof spherical beads with a diameter of 6 nm (∼3 DNA helices)\nplaced to resemble the geometry of the SteriDyn (a base platform with\na length of 9 beads, left arm of length 7 beads, and right arm of\nlength 8 beads, with all three elements having a thickness of 1 bead\nand a width of 3 beads). The two arms are capable of rotating around\ntheir axes by means of experimentally informed energetic penalties U (θ) applied to the corresponding hinges. These penalties,\nwhich implicitly model the effects of hinge designs and latching overhangs,\nwere chosen to follow the Boltzmann distribution U (θ) = − k B T ln P (θ) + U 0 ,\nwhere P (θ) is the experimentally measured probability\ndensity associated with finding a chosen arm at a particular angle\nθ (in the absence of interactions with the other arm), and U 0 is a shift factor that makes U (θ) = 0 at the angle of minimal energy (or maximum probability\ndensity). Steric interactions between the interacting arms and between\nthese arms and the rigid platform were introduced by means of short-range\nrepulsive potentials, which allowed us to reproduce the steric exclusion\nbetween the mobile components of our devices. We first tested\nthis model for a design where the left arm hinge\nis constricted (LC) and the right arm hinge is free (RF); note that\nthis case is the same as that presented in Figure 2 A and represents a structure with negligible\nsteric interactions. Langevin dynamics simulations of this structure\nwere performed, and the free energy landscapes were reconstructed\nfrom angular distributions sampled by the two arms using Boltzmann\ninversion from 15 independent simulations, where measurements were\ntaken every 100 000 time steps for a total of 100 million steps.\nThe color contour map plotted in Figure 3 B depicts the 2D angular free energy landscape\nconstructed from simulations. To compare these predictions against\nexperiments, we overlaid onto this map the set of angles sampled by\nour SteriDyn device as measured from the TEM images. These angles\nare indicated by black crosses (X), where the size of the crosses\nindicates the number of events at a particular location. Experimental\ncounts were clustered by means of a distance cutoff (7.5°), and\nthe sizes of the crosses are logarithmically associated with the number\nof points in each cluster. The experimentally observed conformations\nof the device arms are consistent with the predicted energy landscape\n(i.e., most events fall in the low-energy regions). 1D histograms\ndepicting the experimental angular probability distributions of each\narm ( Figure 3 B, marginal\nplots) also agree well with the corresponding 1D angular distributions\nobtained from simulations in the absence of steric interactions ( Figure S10 ). To engineer steric interactions\nbetween the two dynamic arms, we\nconstructed SteriDyn devices with both left and right hinges in the\nconstricted condition (LC and RC, respectively), which cause the arms\nto adopt relatively small angles. We again observe that the experimentally\nmeasured conformations are in good agreement with the simulation predictions\n( Figure 3 C and Figure S10 ). Both arms exhibit a wide range of\nangles including small angles, but the conformational distribution\nreveals the clear presence of a sterically clashed “dead zone”,\nwhich can be defined by the geometry of the two arms (red dashed line,\nsee Methods section for detailed calculation).\nThe dead zone is asymmetric due to the difference in length between\nthe arms, showing that the longer right arm indeed excludes more conformational\nspace of the shorter left arm than vice versa . Focusing\non the left arm, the depletion of small angle configurations due to\nthe sterically clashed dead zone results in an increase in large-angle\nconformations, leading to a bimodal distribution with a small-angle\npopulation and a large-angle population observed in the 1D θ L distribution. The 2D angular distribution reveals that the\nlarge left-arm angles (θ L ) correspond primarily to\nsmall right-arm angles (θ R ) less than ∼60°\n(angles where the right arm can interact with the left arm), whereas\nfor θ R larger than ∼60°, we observe the\nmajority of θ L at small angles. The 1D angle distribution\nof the left arm ( Figure 3 C, top) similarly reveals a depletion of small-angle configurations\nand a corresponding increase in large-angle configurations. Based on these results, we reasoned that forcing the right arm\nto be at small angles would yield stronger steric interactions to\nalter the left-arm angle distribution. Hence, we constructed various\nSteriDyn devices where the left arm was constricted (LC) and the right\narm was constricted (RC) and could further latch to the base platform\nvia one, two, or three overhang attachments, termed RC 1 ( Figure 3 D), RC 2 ( Figure 3 E),\nand RC 3 ( Figure 3 F), respectively. We again observe a clear sterically clashed\ndead zone and a stronger shift toward larger angles of the left arm,\nθ L (with corresponding depletion of small angles).\nInterestingly, LC|RC 1 , where the right arm exhibits both\nlatched and unlatched states, results in the largest increase in the\npopulation of devices with θ L larger than 120°.\nLC|RC 2 and LC|RC 3 , which are nearly all latched,\ncause a clear depletion of angles below ∼40°, but when\nthe right arm is latched (i.e., θ R ≤ 15°),\nthe left arm can rotate down on top of the right arm, leading to several\ndevices that accumulate near the lower envelope of the steric dead\nzone. These results demonstrate the capacity to transmit a signal\n(i.e., influence conformational distributions) between distal dynamic\ncomponents via steric interactions. Steric Interactions Modulate Binding States of Distal Dynamic\nComponents Besides concerted conformational shifts, allosteric\nmechanisms can also control binding/unbinding reactions in distal\ncomponents. 21 , 47 − 49 Here we studied\nhow a range of SteriDyn arm constraints could regulate the binding\nstate (i.e., latched vs unlatched) of the opposing arm. Focusing on\nthe left arm, we constructed SteriDyn designs with a wide range of\nlatching efficiencies ( Figure 4 ). We defined an arm as “latched” if it adopts\na configuration of θ ≤ 15°. Hence, even with no\noverhangs, it is possible for an arm to appear latched, although with\nlow probability (<3%), and SteriDyn conformations where neither\narm has θ ≤ 15° are considered to have both arms\nunlatched. We tested designs where the left arm is constricted without\nlatching connections (LC), with one or two latching connections (LC 1 and LC 2 ), and two versions of the LC with three\nconnections: LC 3 introduced earlier and one with a higher\nfraction of GC base-pairs (LC 3(HGC) ) in the latching connections.\nWe used the constricted right arm (RC) as the baseline condition.\nThese designs ranged from 2% to 97% in the fraction of latched left\narms ( Figure 4 ). Figure 4 (A) Simulated\n(dashed, squares) and experimental (solid, triangles)\npercentages of left (top plot) and right (bottom plot) latched arms\nfor different numbers of overhangs. Simulation and experimental data\nwere slightly offset from each other in the horizontal direction for\nvisibility. A threshold θ = 15° was considered for the\nlatched state. The x axis is associated with right\narm designs RC, RC 1 , RC 2 , and RC 3 , from lower to increasing propensity for latching. The error bars\nfor experimental data were calculated by bootstrapping (see Methods for details). The error bars for simulation\ndata were calculated from the standard error associated with 15 simulations\nof each device (50 for strongly interacting arms). Error bars that\nare smaller than the symbols are not shown. (B) Normalized dwell times\nin the unlatched state for the left arm (top plot) and right arm (bottom\nplot) obtained from simulations, where the average dwell time for\neach design was normalized by the case with no latching of the right\narm (i.e., all cases were divided by LC X paired with RC),\nshowing the increased latching affinity of the right arm led to an\nincrease in unlatched dwell times of the left arm, but does not affect\nunlatched dwell times of the right arm. The error bars were calculated\nfrom the standard error associated with 15 simulations of each device\n(50 for strongly interacting arms). To regulate the binding state of the left arm through\nsteric interactions,\nwe constructed designs where the right arm could also latch to the\nbottom platform via one, two, or three interactions (RC 1 , RC 2 , and RC 3 ). We find that due to steric\ninteractions, latching of the right arm prohibits latching of the\nleft arm. For each left-arm design (LC, LC 1 , LC 2 , LC 3 , and LC 3(HGC) ), the propensity of left\narm latching decreases with increasing latching affinity of the right\narm. Low-affinity latching of the right arm (RC 1 ) causes\na minor decrease in latching across all left-arm designs. However,\nadding two latching overhangs to the right arm (RC 2 ) leads\nto sharper drops in the left-arm latching propensities for all designs\nexcept LC 3(HGC) . Finally, adding three latching overhangs\non the right arm (RC 3 ) leads to continued decreases in\nthe left-arm latching of LC, LC 1 , and LC 2 and\na significant drop in the left-arm latching of LC 3(HGC) . These results show that dynamic arm components can regulate the\nbinding interactions of distal components through a geometric design\nthat mutually excludes binding states. Interestingly, we observed\ncases where despite both arms latching\nless than 50% of the time, we still observed steric communication\nbetween the latching arms. For example, RC 1 decreased the\nlatching of LC 1 from 30% (when paired with RC) to 18% (when\npaired with RC 1 ) even though RC 1 individually\n(i.e., when paired with LC) only exhibited 32% latching. We hypothesized\nthat this steric communication was due to modulation of binding kinetics;\nthat is, left-arm latching sites being occluded part of the time due\nto binding of the right arm would decrease the latching rate of the\nleft arm. To explore this kinetic modulation, we modified our Langevin\ndynamics simulations by introducing experimentally informed binding\nand unbinding probabilities p on and p off between the arms and the platform at an\nangle θ 0 , so that the dynamics for θ > θ 0 (“open state”) follow the energy landscape U (θ) of the corresponding arm in the absence of overhangs,\nwhile the dynamics for θ ≤ θ 0 (“latched\nstate”) are consolidated into a partially absorbing boundary.\nWe then quantified dwell times in the unlatched and latched states\nfrom these simulations. As we do not have experimental data on absolute\ndwell times, we report relative dwell times, normalizing to the baseline\ncase where there is no latching of the other arm. For example, we\ncalculated the average dwell times from simulations in the unlatched\nstate for each LC 1 design and divided by the average dwell\ntime for the LC 1 paired with RC (i.e., no latching of the\nright arm). These results illustrate that increasing latching affinity\nof the right arm increases dwell times of the left arm in the unlatched\nstate ( Figure 4 B) and\ndoes not influence the dwell times of the left arm in the latched\nstate ( Figure S12 ). Thus, dynamic binding\nof structural components can sterically modulate the binding kinetics\nof distal dynamic components. Our results also show that for\ncases with similar latching affinity\nof the left and right arms (e.g., LC 1 |RC 1 , LC 2 |RC 2 , and LC 3 |RC 3 ), the right\narm outcompeted the left arm to achieve a higher latching efficiency.\nFor example, for the SteriDyn device combining LC 1 |RC 1 , the right arm latched with 32% efficiency and the left arm\nlatched with 18% efficiency. This asymmetry results from minor differences\nin the flexibility of the left and right hinges, as observed in the\nfree-arm angle distributions ( Figure 2 C) and in the binding kinetics of individual arms,\nas the right arm requires a lower binding constant to be found in\nits closed conformation as often as its left counterpart ( Table S2 )."
} | 7,766 |
29067002 | PMC5641323 | pmc | 4,293 | {
"abstract": "Lignocellulosic biomass (LCB) is an attractive source of carbon for the production of sugars and other chemicals. Due to its inherent complexity and heterogeneity, efficient biodegradation requires the actions of different types of hydrolytic enzymes. In nature, complex microbial communities that work efficiently and often synergistically accomplish degradation. Studying such synergisms in LCB degradation is fundamental for the establishment of an optimal biological degradation process. Here, we examine the wheat straw degradation potential of synthetic microbial consortia composed of bacteria and fungi. Growth of, and enzyme secretion by, monocultures of degrader strains were studied in aerobic cultures using wheat straw as the sole carbon and energy source. To investigate synergism, co-cultures were constructed from selected strains and their performance was tested in comparison with the respective monocultures. In monoculture, each organism – with a typical enzymatic profile – was found to mainly consume the cellulose part of the substrate. One strain, Flavobacterium ginsengisoli so9, displayed an extremely high degradation capacity, as measured by its secreted enzymes. Among 13 different co-cultures, five presented synergisms. These included four bacterial bicultures and one bacterial–fungal triculture. The highest level of synergism was found in a Citrobacter freundii / Sphingobacterium multivorum biculture, which revealed an 18.2-fold increase of the produced biomass. As compared to both monocultures, this bacterial pair showed significantly increased enzymatic activities, in particular of cellobiohydrolases, mannosidases, and xylosidases. Moreover, the synergism was unique to growth on wheat straw, as it was completely absent in glucose-grown bicultures. Spent supernatants of either of the two partners were found to stimulate the growth on wheat straw of the counterpart organism, in a directional manner. Thus, the basis of the LCB-specific synergism might lie in the specific release of compounds or agents by S. multivorum w15 that promote the activity of C. freundii so4 and vice versa.",
"introduction": "Introduction Millions of tons of agricultural waste are generated globally every year ( Väisänen et al., 2016 ). Examples are wheat and maize straws, sugarcane bagasse and corn stover. Such lignocellulosic biomass (LCB) is useful as raw material for the production of value-added materials as well as fuels. LCB is composed of lignin, cellulose and hemicellulose, whereas pectin, proteins, small molecules, and minerals can also be present ( Guerriero et al., 2016 ). The exact composition of LCB depends on factors such as plant cultivar type, plant age, local growth conditions, harvesting season and the quality of the soil used for cultivation. For instance, depending on cultivar, age and local conditions, wheat straw can contain 30–44% cellulose, 23–50% hemicellulose, and 7.7–15% lignin ( Van Dyk and Pletschke, 2012 ). A clear impediment to the widespread use of wheat straw as raw material for value-added compounds is its relatively recalcitrant nature, which means it does not easily break down into its monomers. This recalcitrance is clearly caused by its complex chemical composition, and it relates to a major extent to the tight linkages between the lignin, hemicellulose, and cellulose parts. Moreover, the LCB physical structure, i.e., the degree of crystallinity and polymerization of cellulose and polysaccharide, is an important parameter that influences its degradability ( Van Dyk and Pletschke, 2012 ; Bhattacharya et al., 2015 ). As a reflectance of its inherent complexity, a large variety of organisms (producing diverse enzymes) is commonly needed to efficiently degrade LCB like into its monomer compounds. In nature, microbial communities commonly degrade it in a dynamic and time-dependent manner. The degraders are thus presumed to show dynamic responses to the substrate, reaching higher biomass when working together when than acting alone. This process is known as synergistic growth. Moreover, the degrading organisms may use enzymes with complementary activities (enzymatic synergism). Synergism in growth and that in enzymatic activity therefore reflect two processes that are often closely linked in microbial communities ( Van Dyk and Pletschke, 2012 ; Cragg et al., 2015 ). We took these two definitions into our own work on microbial consortia, as proposed in the recent literature ( Mitri and Foster, 2013 ; Deng and Wang, 2016 ). Given the fact that in natural systems synergism in LCB degradation processes is the rule rather than the exception, we surmised it is exacerbated in soil-derived microbial consortia selected on LCB. What mechanisms are behind synergistic behavior in LCB degradation? According to classical knowledge and theory, microorganisms growing together on one substrate, when coexisting, most often divide labor, in a process called niche partitioning. Metabolic complementarity is the main process behind such niche partitioning, as revealed by the classical example of biofuel and hydrogen production through co-cultures of Bacillus and Clostridium on rice straw compost ( Chang et al., 2008 ). So far, it has been relatively unknown to what extent complex substrates like LCB foster processes leading to coexistence. However, recently a co-culture of Trichoderma reesei and Escherichia coli growing on (pretreated) corn stover was found to be optimal in isobutanol production ( Minty et al., 2013 ). The strategy was based on division of function between the two organisms. T. reesei secreted cellulolytic enzymes that transformed the LCB into soluble saccharides, whereas E. coli fermented these into isobutanol. Another recent study reported that, along the same lines, co-cultures of Clostridium cellulovorans (743B) and C. beijerinckii (NCIMB 8152) also successfully produced butanol, under mesophilic conditions ( Wang et al., 2015 ). These studies thus show the key importance of metabolic complementarity in LCB degradation, in which the cooperation between synergistic pairs is driven by exchanges of key metabolites, or by niche partitioning. However, we still do not understand the plethora of mechanisms, as well as the dynamism, that play roles in the microbial attack on the LCB wheat straw ( Pandhal and Noirel, 2014 ; Dolinšek et al., 2016 ; Ghosh et al., 2016 ; Jia et al., 2016 ; Jiang et al., 2017 ). For instance, it remains unclear to what extent the composition/structure of the substrate affects the interactions between collaborating degraders. Moreover, the dynamism in the interactions and activities of collaborative organisms remains understudied. In our previous work, a suite of microbial strains was isolated from three lignocellulolytic microbial consortia that had been selected by repeated growth on raw wheat straw as the single carbon and energy source. Most of the strains had shown promising lignocellulolytic capabilities ( Cortes-Tolalpa et al., 2016 ). We here hypothesized that the wheat straw substrate, being complex and spatially structured, will promote ‘division of labor,’ and so cooperation, between some of the degrader strains. The aim of this study was, therefore, to uncover such synergisms and determine their potential. In this endeavor, we also addressed the potential mechanism behind the synergisms. The data showed that cooperative behavior was relatively ‘common’ in microbial consortia growing on wheat straw, but broke down when strain combinations were grown on simple substrates like glucose.",
"discussion": "Discussion The interest in using co-cultures or consortia in the LCB bioprocess industry has increased recently. For instance, microbial consortia have been proposed as key agents in the degradation of wheat straw ( Jiménez et al., 2013 ; Ghosh et al., 2016 ; Jia et al., 2016 ). The underlying assumption was that they provide a perfect mix of diverse lignocellulolytic enzymes required to degrade the recalcitrant compounds in wheat straw. In particular, metabolic cooperation between microorganisms and synergistic action of secreted enzymes may allow for an efficient degradation process ( Taha et al., 2015 ; Jiménez et al., 2017 ). In this study, we aimed at characterizing to what extent cooperation between individual populations from the microbial consortia affects lignocellulose degradation, by characterizing co-cultures (in comparison to monocultures) of lignocellulose-degrading bacteria and fungi. The cultures were monitored through time, thus providing a dynamic view of both growth and enzyme activities. Our results clearly indicate that bacterial synergism does play a substantial role in subsets of organisms in such consortia and that the relationship between strains inhabiting the same system is dependent on the complexity of the carbon source. Metabolic Complementarity Overall, a positive relationship was found between the abundance of particular degrading bacteria (in raw wheat straw derived consortia) and their capacity to grow on the substrate ( Figure 1 and Supplementary Figure S1 ). This finding corroborated the conclusion that the enrichment process used indeed allowed the selection of strains with high LCB degradative capacity. We further addressed the ability of selected lignocellulose degrading strains to establish a [positive] relationship with each other, as suggested in an earlier study ( Cortes-Tolalpa et al., 2016 ). Synergistic interactions were indeed observed in five of 13 co-cultures, and metabolic complementarity of the component strains was invoked as the most likely mechanism involved. For instance, the most promising synergistic pair, composed of C. freundii so4 and S. multivorum w15 (biculture A) displayed superior growth in co-culture as compared to the respective monocultures, with synergistic activities of several hydrolytic enzymes ( Figure 2A ). C. freundii and S. multivorum differ widely in their metabolic properties. C. freundii is a member of the Enterobacteriaceae , a facultatively anaerobic family, with motility by flagella. It is able to grow on glycerol as well as citrate as sole carbon sources ( Rosenberg et al., 2014a ). S. multivorum belongs to the Sphingobacteriaceae . It is a strict aerobe, which does not possess flagellar motility. It is able to produce acid from a large variety of carbohydrates (including α- D -glucopyranoside and α- D -mannopyranoside) by oxidative processes. In fact, the organism is able to grow on p -hydroxy-butyrate as a single carbon source, but not on glycerol, like C. freundii . Moreover, S. multivorum is well known as a producer of extracellular enzymes, mainly xylosidases, proteases, and lipases ( Rosenberg et al., 2014b ). Both strains are capable of transforming cellobiose. Division of Labor In our study, S. multivorum w15 probably contributes to cultures growing on wheat straw with efficient extracellular enzymes. In particular the release of different types of xylosidases seems to be a common feature among S. multivorum strains ( Malfliet et al., 2013 ; Lian et al., 2016 ). Here, growing on raw wheat straw, S. multivorum w15 produced powerful cellobiohydrolases and β-xylosidases; such enzymes were not found with C. freundii so4 when grown under the same conditions ( Figure 2A ). We also found highly active β-xylosidases from S. multivorum strains w15 and so22, grown on wheat straw singly and in co-culture ( Figure 2 and Supplementary Figure S1 ). Moreover, it has been indicated that S. multivorum has lignin-degradation potential, which suggests the organism may also play a role in the degradation of the lignin present in wheat straw ( Taylor et al., 2012 ). Such key metabolic activities allow S. multivorum to establish positive interactions with C. freundii so4. On the other hand, C. freundii so4 showed excellent growth on glucose, as opposed to S. multivorum w15. However, strain w15 did grow well in the glucose bicultures, which indicates that C. freundii so4 exerted a positive metabolic effect on its counterpart strain ( Figure 3 ). We hypothesized that it probably provides redox power and contributes to the degradation of oligosaccharides to simpler sugars. This might be stimulated by its high motility, allowing it to explore the substrate. Furthermore, given the strict aerobic metabolism of S. multivorum w15, it is very likely that C. freundii so4 produces metabolic intermediates that S. multivorum w15 can consume, allowing it to reach higher cell densities in co-culture than in monoculture. Furthermore, the observed growth stimulation of the S. multivorum w15 as well as the C. freundii so4 monocultures following treatment with the supernatant of the counterpart wheat-straw-grown strain further corroborates the contention that synergistic interactions take place when growing on wheat straw. We speculate that, in both cases, the recipient strain was capable of reaching increased cellular density after receiving, from the donor, a considerable number of secreted enzymes, next to (potentially) other compounds. With respect to the latter, signaling could be involved. This is corroborated by the fact that a quorum sensing system has been found in C. freundii ( Rosenberg et al., 2014a ; Wang and Zhou, 2015 ). Although we cannot precisely pinpoint the mechanisms that drive the interactions in our co-cultures, as well as the large increase of enzymatic activities observed in them ( Figures 1 , 2 ), the supernatant-induced growth stimuli ( Figure 4 ) provide clear evidence for synergistic interactions. Moreover, the metabolic differences between the two strains suggest that they divide ‘labor’ in the transformation of the heterogeneous wheat straw, allowing their co-cultures to build up an enhanced biomass. Importantly, the synergism was only observed with supernatants harvested from cells growing on raw wheat straw, but not with those from glucose-grown cells, indicating the relevance of the chemical complexity of the substrate (see below). Influence of the Carbon Source The complexity of carbon sources can have a substantial influence on the metabolism of heterotrophic organisms ( Deng and Wang, 2016 ). Klitgord and Segrè (2010) , using flux balance analysis, found that different medium formulations (based on carbon, nitrogen, sulfur, and phosphorus) affect the interactions between microorganisms ( Klitgord and Segrè, 2010 ). In our study, the more complex the substrate was, the more synergistic the relationship between C. freundii so4 and S. multivorum w15 became. Thus, the emergence of synergism in subsets of the original wheat-straw-grown microbial consortia can be linked to the inherent heterogeneity of the substrate, suggesting that the complexity of the carbon source can strongly modify the relationship between degrader strains. Specifically, we hypothesized that the level of synergism between bacteria involved in LCB degradation processes is related to the differential presence of bonds in substrates of different complexity. In the SLB, the three main components (cellulose, xylan, and lignin) were not tightly bound together in a matrix, such as was the case for the raw wheat straw. Thus, the finding that the collaborative bacterial pair showed synergism only at the end of the experiment is in line with this lower number of bonds ( Figure 3B ). Specifically, the presence of bonds between lignin and the complex carbohydrates cellulose and hemicellulose, or between them, may have been at the basis of the observed synergism. Such bonds determine to some extent the recalcitrance of the LCB ( Du et al., 2014 ; Arnling Bååth et al., 2016 ). Notwithstanding our enhanced understanding of the bias of synergism and the link to recalcitrant bond numbers, further studies are necessary to understand this phenomenon in greater detail. Overall, the data indicate that, when grown on raw wheat straw as the sole C and energy source, degradative strains first consume the labile parts of the substrate, after which they are in need to collaborate to access the remaining recalcitrant sources of carbon. We here posit that ‘multipolymer’ or ‘peeling’ synergism could be a model description of the mechanism involved in the synergism between S. multivorum w15 and C. freundii so4 on raw wheat straw. In this type of synergism, proposed by Selig et al. (2008) and Várnai et al. (2011) , cellulose and hemicellulose are, at the same time, “peeled off” by enzymatic action, exposing new structures of the substrate to the hydrolytic enzymes that are or become available. For the complete hydrolysis of the raw wheat straw, different types of lignocellulolytic enzymes are probably required, in a temporally and spatially dynamic manner ( Selig et al., 2008 ; Várnai et al., 2011 ). Final Remarks Overall, this study reveals that, in LCB degradation processes, co-cultures of particular nature are superior to monocultures, as they allow division of labor in the metabolic processes that are required by the substrate. Clearly, microorganisms often lack some key metabolic pathways, which may be supplemented by others ( Mikesková et al., 2012 ; Abreu and Taga, 2016 ; Ghosh et al., 2016 ). Thus, LCB degradation, in the end, may impose ‘group selection’ pressure on the process participants, in which ‘group’ is not defined by ‘kin’ but is rather determined by complementarity in a spatially- and temporally-explicit process. Our findings are consistent with recent data that show that co-cultures often present improved performance over corresponding monocultures. The mechanisms involved may include enhanced substrate utilization, overcoming of nutritional limitations, reduction of the levels of cheaters/scavengers and achieving superior overall activity, conversion and enzymatic action ( Feng et al., 2011 ; Okeke and Lu, 2011 ; Zuroff et al., 2013 ; Liao et al., 2015 ; Valdez-Vazquez et al., 2015 )."
} | 4,499 |
37823039 | PMC10562921 | pmc | 4,296 | {
"abstract": "Utilizing anaerobic metabolisms for the production of biotechnologically relevant products presents potential advantages, such as increased yields and reduced energy dissipation. However, lower energy dissipation may indicate that certain reactions are operating closer to their thermodynamic equilibrium. While stoichiometric analyses and genetic modifications are frequently employed in metabolic engineering, the use of thermodynamic tools to evaluate the feasibility of planned interventions is less documented. In this study, we propose a novel metabolic engineering strategy to achieve an efficient anaerobic production of poly-(R)-3-hydroxybutyrate (PHB) in the model organism Escherichia coli . Our approach involves re-routing of two-thirds of the glycolytic flux through non-oxidative glycolysis and coupling PHB synthesis with NADH re-oxidation. We complemented our stoichiometric analysis with various thermodynamic approaches to assess the feasibility and the bottlenecks in the proposed engineered pathway. According to our calculations, the main thermodynamic bottleneck are the reactions catalyzed by the acetoacetyl-CoA β-ketothiolase (EC 2.3.1.9) and the acetoacetyl-CoA reductase (EC 1.1.1.36). Furthermore, we calculated thermodynamically consistent sets of kinetic parameters to determine the enzyme amounts required for sustaining the conversion fluxes. In the case of the engineered conversion route, the protein pool necessary to sustain the desired fluxes could account for 20% of the whole cell dry weight.",
"conclusion": "4 Conclusions In this study, different approaches were combined to evaluate the thermodynamic and physiological feasibility of an efficient anaerobic conversion of glucose/sucrose in PHB. It was shown how a handful of reactions and metabolites involved in a novel conversion strategy can have a drastic effect in its feasibility, highlighting key targets for eventual genetic modifications. Special interest deserve the reactions catalyzed by the acetoacetyl-CoA β-ketothiolase and the acetoacetyl-CoA reductase, where the presence of a substrate channeling mechanism and/or the availability/discover/engineering of enzymes with higher turnover numbers could be game changers.",
"introduction": "1 Introduction Stoichiometric analyses have become invaluable tools for a deeper understanding of complex metabolic networks and the rational design of metabolic engineering strategies [ 1 , 2 ]. However, the metabolic network analyses restricted to the stoichiometric relationships are blind to important factors such as the thermodynamic feasibility of the involved reactions or the kinetic properties of the participating enzymes. Consideration of kinetic and thermodynamic constrains is frequently crucial for the success of metabolic engineering approaches [ [3] , [4] , [5] ]. Therefore, we think that the inclusion of thermodynamic and kinetic analyses is an important step during the assessment of the feasibility of the metabolic engineering strategies. Two methods used for the identification of possible thermodynamic and kinetic bottlenecks in metabolic pathways are the Max-Min Driving Force (MDF) [ 6 ] and the Enzyme Cost Minimization (ECM) [ 7 , 8 ]. MDF calculates the maximum possible thermodynamic driving force of the conversion pathway, given a set of metabolite concentration ranges [ 6 ]. It aids in identifying the reactions and metabolite concentrations that are key for the thermodynamic feasibility and favorability of the conversion. On the other hand, the ECM method calculates the minimum amount of enzyme required to sustain a given metabolic flux through a conversion pathway. Thus, the comparison between the experimentally observed amount of a given enzyme (which could be determined by enzymatic assays, proteomics, immunoassays, etcetera) and the minimal amount calculated by ECM could guide the genetic engineering approaches towards a cellular resource allocation more suitable for the biotechnological purpose. MDF and ECM methods are particularly valuable for analyzing anaerobic conversions, which dissipate less free energy and thus have narrower ranges of thermodynamically feasible intracellular metabolite concentrations. Despite the thermodynamic constrains, anaerobic conversions offer two key advantages compared to aerobic processes: (i) reduced material and/or energy input (no air supply, low stirring and less cooling) and (ii) increased product yields. Remarkably, previous thermodynamic analyses have shown that many biotechnologically relevant products could be produced using anaerobic processes [ 9 ]. The implementation of metabolic engineering designs that enable the substitution of conventional aerobic conversions with anaerobic conversions has the potential to enhance the competitiveness and reduce the environmental footprint of certain bioprocesses. One such bioprocess is the production of poly-(R)-3-hydroxybutyrate (PHB). PHB is a polymer that can be synthetized by various microorganisms, and it has properties similar to some petrol-based plastics. In 1988, two research groups independently showed that the expression of the phaCAB1 gene from Cupriavidus necator is sufficient to induce PHB accumulation in Escherichia coli [ 10 , 11 ]. Since these groundbreaking studies, the study of PHB accumulation in E. coli has attracted considerable attention given the large availability of tools to genetically modify this bacterium and its natural inability to degrade this polymer. In addition, E. coli can assimilate different carbon sources and it can grow both aerobically and anaerobically. Carlson and co-workers studied anaerobic PHB production in E. coli . They combined experimental observations with analysis of the stoichiometric possibilities, specifically focusing on elementary modes enabling the highest possible product yields, given the glycolytic pathways naturally present in this bacterium [ 12 ]. The study concluded that the co-supply of glucose and acetate, in a 2:1 ratio, should result in the production of (R)-3-hydroxybutyryl monomers and formic acid, in a 1:1 ratio, allowing for the recovery of 80% of the carbon in the PHB. Experimentally, the anaerobic co-generation of PHB and ethanol [ [13] , [14] , [15] ] or PHB and hydrogen [ 16 ] have been achieved. However, the obtained PHB yield are very low (6 mg PHB /g glucose [ 16 ]; 8 mg PHB /g xylose [ 14 ]; 14 mg PHB /g xylose [ 15 ]). Moreover, the continuous generation of PHB as the sole fermentation product during anaerobic metabolism has not been reported yet. We are proposing here a novel metabolic engineering strategy for E. coli to achieve an anaerobic conversion of glucose or sucrose into PHB, while maximizing product yield and ATP output. In addition to the stoichiometric analysis, MDF and ECM calculations were applied to identify key bottlenecks in the proposed engineered route, assessing the thermodynamic and biological feasibility of this strategy. Moreover, we employed these computational tools to compare the proposed engineered conversion pathway with the natural glycolytic pathways from E. coli .",
"discussion": "3 Results and discussion 3.1 A novel strategy towards an efficient anaerobic production of PHB In wild type E. coli cells, glucose can be converted to acetyl-CoA with a positive ATP yield, through two glycolytic routes: the Embden-Meyerhof-Parnas (EMP) and the Entner-Doudoroff (ED) pathways, complemented with the reactions catalyzed by the pyruvate formate lyase (EC 2.3.1.54) ( Fig. 1 A). A stoichiometric analysis of these catabolic routes shows a mismatch between the amount of acetyl-CoA and reducing equivalents produced by these pathways and the amount of acetyl-CoA and reducing equivalents required by the PHB biosynthetic route ( Fig. 1 B). Consequently, to achieve the redox balance for PHB production from sugars under anaerobic conditions, it is necessary that cells co-produce another fermentation product [ 14 ], take-up volatile fatty acids [ 24 ], or use an external electron acceptor [ 25 ]. In this study, we are proposing an alternative solution to overcome this mismatch by generating a fraction of the acetyl-CoA through the non-oxidative glycolysis. The non-oxidative glycolysis (NOG) is an artificial pathway, firstly proposed by Bogorad and co-workers, enabling the conversion of glucose into acetate without oxidation steps [ 26 ]. Specifically, if 75% of the acetyl-CoA is generated by-passing the oxidation steps catalyzed by the glyceraldehyde-3-phosphate dehydrogenase and the pyruvate dehydrogenase complex, the resulting proportion of acetyl-CoA and reducing equivalents matches the redox requirements for PHB synthesis ( Fig. 1 C). Therefore, it should be feasible to design a metabolic engineering strategy that combines reactions from the NOG and EMP pathways. Thus, we will refer to this new strategy as NOGEMP. Fig. 1 Efficient PHB accumulation requires a match between acetyl-CoA (AcCoA) and electrons. A : In E. coli wild type, glucose is oxidized through the Embden-Meyerhof-Parnas (EMP) and/or the Entner-Doudoroff (ED) pathways. The numbers represent the relative metabolic fluxes through the EMP (red) and the ED (blue) pathways, respect to one molecule of oxidized glucose. The harvested ATP is “consumed” in an ATP hydrolysis reaction, mimicking the ATP consumed for maintenance or other metabolic processes. Names of the metabolites and enzymes are detailed in the Supplementary Material. B : It is possible to see that, either using the EMP (red arrow) or the ED pathway (blue arrows), there is a mismatch between the catabolic supply of AcCoA and reducing equivalents and the demand of AcCoA and reducing equivalents of the PHB synthesis pathway: for every two molecules of AcCoA generated, four pairs of electrons must be harvested by NAD, NADP or released as formate/H 2 . However, only one NAD(P)H is re-oxidized during the synthesis of 3-(R)-hydroxybutyrate (the monomer of the PHB). The other molecules of NAD(P)H have to be re-oxidized in processes delivering electrons outside of the cells (respiration or fermentation). Circles represent enzymatic steps catalyzed by oxido-reductases. C : Proposed engineered pathway enabling an efficient anaerobic conversion of glucose to PHB. Due to the combination of reactions from the Embden-Meyerhof-Parnas and the Non-Oxidative Glycolysis pathways, we nicknamed this engineered pathway as NOGEMP. Note that all the reducing equivalents coming from the carbon source can be harvested in the product, approaching the maximum theoretical product yield. Names of the metabolites and enzymes are detailed in the Supplementary Material. Fig. 1 The implementation of NOGEMP can be divided in five modules. First , glucose uptake must occur by diffusion [ 27 ], and glucose-6-phosphate must be produced in a reaction catalyzed by a glucokinase (EC 2.7.1.2) to uncouple the phosphorylation of glucose from the flux through the lower EMP or ED pathways. Second , the upper glycolytic flux should go through the NOG. For this redirection of the glycolytic flux, phosphoketolase (EC 4.1.2.9) and fructose-1,6-biphosphatase (EC 3.1.3.11) activities are required [ 26 ]. The 6-phosphofructokinase (EC 2.7.1.11) activity must be suppressed to avoid an ATP futile cycle caused by the simultaneous activities of phosphofructokinase and fructose-1,6-biphosphatase [ 28 ]. Third , the reduction of acetoacetyl-CoA to 3-(R)-hydroxybutyryl-CoA must be catalyzed by an NADH-preferring acetoacetyl-CoA reductase (EC 1.1.1.36) [ 25 ]. This will transform PHB in a sink for the electrons carried by NADH ( i.e. , a fermentation product). Fourth , the formation of ethanol, acetate and lactate must be suppressed [ 29 ]. The suppression of the mentioned products enforces PHB synthesis as an obligate mechanism for NADH re-oxidation and avoids the conversion of acetyl-CoA to acetate. The formation of succinate and formate/H 2 cannot be fully suppressed because it is required for biomass formation under anaerobic conditions. Fifth , modifications in the promoter region controlling the expression of the genes encoding for the pyruvate dehydrogenase complex are required to become active under anaerobic conditions [ 30 ]. Overall, the implementation of the five previously described modules will imply that, under anaerobic conditions, PHB accumulation will be coupled to biomass formation. If we consider the dimer of two monomers of (R)-3-hydroxybutyrate as the minimal unit of PHB, the global reactions of the anaerobic conversion of glucose to PHB using the EMP, the ED or the NOGEMP are: EMP: 3 Glucose + 6 ADP + 6 Pi = PHB + 2 Ethanol + 6 Formate + 5 H 2 O + 6 ATP ED: 3 Glucose + 3 ADP + 3 Pi = PHB + 2 Ethanol + 6 Formate + 2 H 2 O + 3 ATP NOGEMP: 3 Glucose + ADP + Pi = 2 PHB + 2 CO 2 + 5 H 2 O + ATP. The comparison of the yields of these conversion pathways highlights the advantage of implementing the NOGEMP pathway: 90% of the carbon and 100% the electrons contained in the source should be conserved in the product ( Table 1 ). The improvements in the production yields are due to the correction of the mismatch between the acetyl-CoA and the electrons produced in the oxidation of the glucose and the acetyl-CoA and electrons consumed in the PHB formation, eliminating the necessity of generating fermentation products. Table 1 Specific consumption/production (q-) rates and yields expected for the operation of the pathways under analysis. Table 1 Pathway q glucose q Ethanol q Formate q ATP q PHB Y carbon ∗ (Cmol in PHB/Cmol in glucose) Y ∗∗ (g PHB /g glucose ) fraction of Y max # (%) EMP −15 10 30 30 5 0.444 0.32 50 ED −15 10 30 15 5 0.444 0.32 50 NOGEMP −15 0 0 5 10 0.889 0.64 100 ∗ The PHB molecule was considered as the union of two monomers of 3-R-hydroxybutyrate. ∗∗ The molecular weight of the considered PHB molecule is 172 g/mol. # Maximum theoretical yield. However, our stoichiometric analyses do not consider the thermodynamic characteristics of the intermediary reactions and the kinetic properties of the involved enzymes. To overcome these limitations, we firstly focused on the identification of the thermodynamic bottlenecks in NOGEMP and the assessment of the metabolite concentration ranges enabling the operation of this pathway. 3.2 Max-Min Driving Force analyses show that the reaction catalyzed by the acetoacetyl-CoA β-ketothiolase is the main thermodynamic bottleneck The thermodynamic driving forces of the ED, EMP and NOGEMP pathways were compared using a wide concentration range (between 1 μM and 10 mM). The ED pathway exhibited the highest thermodynamic driving force, while the NOGEMP pathway had the lowest thermodynamic driving force (MDF EMP = 8.53 kJ/mol, MDF ED = 11.38 kJ/mol, MDF NOGEMP = 7.17 kJ/mol). Previous studies have shown that, in the absence of PHB synthesis reactions, the MDF of ED is higher than that of EMP [ 7 ]. According to Flamholz and co-workers, the higher ATP conservation for EMP comes with a price: a lower MDF and a higher enzyme cost. Our results show that, even with the inclusion of PHB synthesis reactions and considering a wide metabolite concentration range, ED still has a higher MDF than EMP. It must be noticed that the products generated in the EMP and ED in one hand, and NOGEMP in the other hand, are not the same. According to the calculations with the group contribution method [ 21 ], the free energy variation of the global reactions of these pathways differs: 3 Glucose + H 2 O = (R)-3-hydroxybutanoyloxybutanoate + 2 Ethanol + 6 Formate (Δ r G′° = −560.0 ± 16.8 kJ/mol) 3 Glucose = 2 (R)-3-hydroxybutanoyloxybutanoate + 2 CO 2 + 4 H 2 O (Δ r G′° = −639.4 ± 15.6 kJ/mol]) This difference impacts the potential for energy conservation. However, the stoichiometric analysis shows that despite its higher total free energy variation, NOGEMP showed the lowest ATP conservation and the lowest MDF value. This indicates the co-existence in the same pathway of steps with a large energy dissipation (the reactions catalyzed by phosphoketolase and fructose-1,6-biphosphatase) and steps with a low metabolic driving force (the linked and highly reversible reactions catalyzed by phosphoglucose isomerase (EC 5.3.1.9), transaldolase (EC 2.2.1.2), transketolase (EC 2.2.1.1), ribose-5-phosphate isomerase (EC 5.3.1.6) and ribose-5-phosphate epimerase (EC 5.1.3.1)) ( Fig. S2 ). MDF calculations were repeated, this time considering experimentally validated narrower ranges of metabolite concentrations to estimate the thermodynamic driving forces at physiological conditions ( Table S5 ). The results (MDF EMP = 6.17 kJ/mol, MDF ED = 6.62 kJ/mol, MDF NOGEMP = 3.77 kJ/mol) show that the thermodynamic driving forces of all three pathways decreased due to the more constrained metabolite concentration ranges ( Fig. S2 ). However, it is important to note that all three pathways remained feasible under these physiological conditions. To gain a deeper understanding of the factors limiting the achievable thermodynamic force in the pathways under analysis, a concentration variability analysis (CVA) was implemented. This analysis aimed to determine the metabolite concentrations ranges that enable the operation of a given pathway at a specific thermodynamic driving force. At the thermodynamic driving force calculated by the MDF analysis, at least one of the metabolites reached a critical concentration threshold. This means that the concentration of that particular metabolite fell outside the range that enables thermodynamic feasibility of at least one reaction. Therefore, CVA calculations must be performed by setting the thermodynamic driving force at a value lower or equal to the thermodynamic driving force calculated by MDF. As expected, the higher the allowed thermodynamic driving force, the more constrained were the metabolite concentration ranges that enable the achievement of such thermodynamic driving force ( Fig. 2 , Fig. S3 ). CVA also enabled us to calculate which are the metabolites with extreme concentrations during the operation of the pathway, even at a low thermodynamic driving force. For this purpose, we performed CVA analyses setting the thermodynamic driving force at 1 kJ/mol. In the three analyzed pathways, minimal acetyl-CoA concentrations of 4.8 mM and maximal acetoacetyl-CoA concentrations of 21 μM are required to enable the operation of any of the three pathways under analysis, at a thermodynamic driving force of 1 kJ/mol ( Table S9 ). These metabolites with extreme concentrations indicate that the reaction catalyzed by the acetoacetyl-CoA β-ketothiolase is the primary thermodynamic bottleneck, as it involves both acetyl-CoA and acetoacetyl-CoA. Fig. 2 Metabolite concentration ranges enabling the operation of the EMP and NOGEMP pathways at two different driving forces (1 kJ/mol and 3 kJ/mol). The limits of the ordinate axes were fixed at convenient values, not necessarily corresponding to the thermodynamically feasible metabolite concentrations ranges. These metabolite concentration ranges were obtained with the Concentration Variability Analysis approach. It is possible to see how some metabolite concentration ranges are constrained with the increase in the thermodynamic driving force; for example, the range for the acetoacetyl-CoA (AcAcCoA). Fig. 2 The reaction catalyzed by the acetoacetyl-CoA β-ketothiolase is a Claissen condensation between two molecules of acetyl-CoA with a standard free energy of +25.2 kJ/mol. Substrate channeling has been observed in other enzymes catalyzing reactions with a similar functional chemistry, such as the 3-ketoacyl-CoA thiolase participating in the β-oxidation in Pseudomonas fragi [ 31 ], the citrate synthase in the Krebs cycle in pigs [ 32 ] and the acetoacetyl-CoA reductase participating in the formation of 3-hydroxy-3-methylglutaryl-CoA in the mevalonate pathway from Archaea [ 33 ]. Therefore, it is likely that substrate channeling plays a role during the formation of PHB, as previously suggested [ 25 , 34 ]. The finding of experimental evidence of this phenomenon requires dedicated biochemical procedures beyond the scope of this research. Yet, we explored the effects of substrate channeling between the acetoacetyl-CoA β-ketothiolase and the acetoacetyl-CoA reductase on the thermodynamic profile of the anaerobic conversion of glucose in PHB, using the EMP, ED and NOGEMP pathways. For the analyses of the effect of substrate channeling, the reactions catalyzed by the acetoacetyl-CoA β-ketothiolase and the acetoacetyl-CoA reductase were replaced by the lumped reaction 2 AcCoA + NADH → (R)-3-hydroxybutyryl-CoA + CoA + NAD + . Consistent with the hypothesis of considering the reaction catalyzed by the acetoacetyl-CoA β-ketothiolase as the thermodynamic bottleneck in the three pathways, higher driving forces were obtained using the lumped reaction (MDF EMP = 7.05 kJ/mol, MDF ED = 8.16 kJ/mol, MDF NOGEMP = 4.67 kJ/mol). Moreover, the concentration variability analyses showed a decrease in the minimal concentration of acetyl-CoA required to have feasible conversions ( Tables S10–S12 ). Bearing in mind a cytoplasmic volume of 1.9 mL/gCDW [ 35 ] and the molecular weight of the acetyl-CoA (809.57 g/mol), the change in the minimal concentration of acetyl-CoA from 4.8 mM to 0.12 mM due to the substrate channeling, implies a decrease from 0.7% to 0.02% in the contribution of acetyl-CoA to the cellular weight. Given the large molecular weight of the coenzyme A and the thermodynamic constraints associated with the Claissen condensation, it seems likely that some mechanism enabling a substrate channeling between the acetoacetyl-CoA β-ketothiolase and the acetoacetyl-CoA reductase had been developed by evolution. However, given the lack of experimental evidence and to avoid bias, the rest of the calculations here reported were performed without considering this substrate channeling mechanism. Our MDF calculations show that the reaction catalyzed by the acetoacetyl-CoA β-ketothiolase is the main thermodynamic bottleneck for the anaerobic production of PHB using the EMP, the ED or the engineered NOGEMP pathway. Previous experimental observations of PHB accumulation in E. coli cells under anaerobic conditions showed that this thermodynamic bottleneck can be surpassed [ 12 , 36 ]. However, to achieve the flux distribution envisioned in the proposal of the NOGEMP, an extensive re-wiring of the metabolic fluxes is required. Provided that (i) the enzymes catalyzing the formation of fermentation products are eliminated by genetic engineering and that (ii) the enzymes participating in NOGEMP are expressed in the required amounts, the only way to re-oxidize the produced NADH under anaerobic conditions is with the flux distribution proposed for NOGEMP. The elimination of the enzymes involved in the generation of fermentation products can be achieved by genetic engineering, using techniques described elsewhere [ 25 , 37 ]. On the other hand, to calculate the minimal amount of the enzymes required to sustain the metabolic fluxes, we applied the ECM approach. 3.3 NOGEMP pathway could be sustained by E. coli cells after further modifications aiming an increase in ATP conservation For the ECM calculations, we defined the absolute fluxes as the metabolic fluxes required to sustain the ATP cost for maintenance (see Supplementary Material for a more detailed explanation). The respective ATP yields ( Table 2 ) and fluxes distributions ( Fig. S4 ) were calculated by Flux Balance Analysis. Kinetic parameters found in literature come from diverse experiments, executed under different conditions; therefore, taken as a whole they could be inconsistent with the laws of thermodynamics. To obtain more accurate results, the turnover rates (in the forward and the backward direction) and the Michaelis constants of a given enzyme should be thermodynamically consistent, satisfying the corresponding Haldane relationship. To accomplish this requirement, thermodynamically consistent sets of kinetic parameters were obtained using the Parameter Balancing Tool [ 38 ] ( Fig. S1 , Table S8 ). To run the Parameter Balancing Tool, we constructed (i) SBML models representing each pathway and (ii) the corresponding *.tsv files containing the estimated means and standard deviations of the equilibrium constants and the kinetic parameters. The files containing the SBML models were constructed using the freely available tool COPASI (v4.34) [ 39 ]. For the construction of the *.tsv files, experimentally determined parameters were gathered from different sources (see values and references in Table S6 ). The uncertainties associated to the equilibrium constants were taken directly from the estimations of eQuilibrator. The uncertainties associated to the known kinetic parameters were taken from the literature (when available) or declared as unknown values (when information about the dispersion of the kinetic estimates was not available in the reviewed literature). Overall, for all the pathways under study, more than 70% of the parameters employed as input for the Parameter Balancing Tool were known ( Table S7 ). Certainly, the number of unknown values affects the accuracy of the obtained results: the larger the number of unknown values, the larger the uncertainties associated with the calculated consistent parameters. The resulting sets of thermodynamically consistent kinetic parameters, with their associated uncertainties, are available in Table S8 . Table 2 Physiological impact of the enzyme cost. Table 2 ATP yield (mol ATP /mol hexose ) enzyme cost (g Prot /L cyt ) contribution Thio + AAR to whole enzyme cost (% of enzyme cost) contribution of the pathway to whole proteome (%) ED 1 10.9 43 3.0 EMP 2 11.9 28 3.2 NOGEMP 0.333 126.4 25 34.3 NOGEMP sucrose symporter 0.667 72.4 23 20 NOGEMP sucrose uniporter 1.667 60.5 22 16 After including the molecular weight of the participating enzymes, the ECM analysis yielded their results in units of grams of protein per liter of cytoplasm (g Prot /L cyt ). This conversion enabled to evaluate the physiological impact of the protein cost, calculated as the fraction of the total cellular protein pool occupied by the enzymes of a given pathway. To calculate these fractions, we considered that every liter of cytoplasm corresponds to 526 g of cell dry weight [ 35 ], and that proteins account for 70% of the cell dry weight [ 18 ]. The resulting equation is: e n z y m e c o s t ( g p a t h w a y P r o t L c y t ) 526 g C D W L c y t * 0.7 g t o t a l P r o t g C D W * 100 % The protein fractions obtained for the different pathways are shown in Table 2 . Our results show that the combined contributions of the glycolytic enzymes, without considering acetoacetyl-CoA β-ketothiolase and acetoacetyl-CoA reductase, is 6.21 g Prot /L cyt for the ED pathway and 8.55 g Prot /L cyt for the EMP pathway. This indicates that, despite its higher ATP yield, the amount of protein required for the operation of the EMP pathway is higher than the amount required for the ED pathway, which is consistent with previous observations [ 7 ]. Considering the costs of acetoacetyl-CoA β-ketothiolase plus acetoacetyl-CoA reductase in the EMP and the ED pathways, these enzymes contribute with 3.33 g Prot /L cyt and 4.68 g Prot /L cyt , respectively. Due to the fact that the ATP yield of the ED pathway is half of the ATP yield of the EMP pathway, to cover the same ATP cost for maintenance the flux through the enzymes acetoacetyl-CoA β-ketothiolase and acetoacetyl-CoA reductase in the ED pathway doubles the flux through the same enzymes in the EMP pathway ( Fig. S4 ). Nevertheless, the protein cost of these enzymes in the ED pathway does not double the cost of these enzymes in the EMP pathway because the thermodynamic conditions are different for each pathway (Tables S10 and S11). This case exemplifies how the contribution to the whole proteome of a subset of enzymes, with identical kinetic properties, could be different, depending on the thermodynamic conditions found in each network. This calculation is an example of a (biotechnologically) relevant information that it is possible to get with the thermodynamic and kinetic analyses and that it is not possible to get with simple stoichiometric analyses. In the case of the NOGEMP pathway, the enzyme cost raises to 126 g Prot /L cyt . One of the reasons explaining this higher cost is the higher glycolytic flux required to cover the same ATP expenses for maintenance but using a glycolytic network with a lower ATP yield ( Table 1 ). On the other hand, as previously addressed, the NOGEMP has more reactions operating closer to their thermodynamic equilibrium; therefore, at a given time, more active sites are engaged in catalyzing backward reactions, increasing the amount of enzyme required to sustain a given net flux in the forward direction [ 8 ]. The enzyme cost calculated for NOGEMP would represent 34% of the whole cellular proteome. This number exceeds the estimated fractions of the whole proteome dedicated to the catabolic functions (between 10% [ 40 , 41 ] and 20% [ 42 ]). To overcome this problem, we are proposing to increase the ATP conservation by fueling the cells with sucrose instead of glucose, followed by sucrose phosphorolysis [ 43 ]. Two mechanisms were considered for the transport of sucrose across the cytoplasmic membrane: (i) facilitated diffusion through a uniporter and (ii) co-transport of sucrose with protons ( Fig. S5 ). In the latter case, one molecule of sucrose is co-transported with one proton. To maintain the cytoplasmic pH, the excess of protons is pumped out the cytoplasm. The amount of ATP required to pump-out one mol of protons depends on the difference in pH between the cytoplasm and the periplasmic compartment, the transmembrane voltage and the concentrations of ATP, ADP and Pi. According to our calculations, three mol of proton can be pumped-out with the energy released by the hydrolysis of one mol of ATP (see calculations in the Supplementary material and Table S13 ). Therefore, depending on the sucrose uptake mechanism, the ATP yield should rise to 0.667 ATP/hexose (sucrose symporter) or to 1.667 ATP/hexose (uniporter). After implementing the strategies to increase the ATP conservation, the enzyme cost of NOGEMP should decrease, representing 20% (symporter) or 16% (uniporter) of the whole proteome ( Table 2 ). Although these values are in the upper half of the previously reported range of 10–20% of the proteome dedicated to catabolism, it is important to highlight that these previously reported values are based on aerobic growth. It has been shown that during the shift from aerobic to anaerobic conditions most of the newly formed proteins are involved in glycolysis and pathways aimed at preventing glycolysis grinding to a halt by a cellular redox imbalance [ 44 ]. Moreover, it has been observed that, during anaerobic conditions, the abundance of the proteins associated with glycolysis and fermentation is duplicated and the abundance of the enzymes of the Krebs cycle are reduced up to five times [ 45 ]. In this scenario, alcohol dehydrogenase, pyruvate formate lyase, enolase and glyceraldehyde-3-phosphate dehydrogenase represented each one 2–3% of the proteome mass. Regarding the choices to increase the ATP conservation, it is key to discuss that although the passive diffusion of sucrose had been described in different plants [ [46] , [47] , [48] ] and some data support its successful heterologous expression in Saccharomyces cerevisiae [ 43 ], other data indicate that the actual operative mechanism is a proton:sucrose symporter [ 49 ]. Even if sucrose diffusion is possible, the feasibility of this mechanism will depend on the existence of a concentration gradient between the extracellular medium and the cytoplasm. Considering that the Michaelis constants for sucrose of the best studied sucrose phosphorylases are around 10–20 mM [ 50 ], the semi-saturation of this enzyme would require intracellular concentrations of sucrose above 3.5 g/L. Therefore, even considering its higher cost in comparison with the sucrose uniporter, we consider that the implementation of the strategy based on the use of a proton:sucrose symporter is more safe. Complementary approaches such as directed evolution and protein/genomic engineering could help to decrease the enzyme cost. We would also want to highlight that the ECM analysis yielded insights into the contributions of the thermodynamic, saturation, and capacity terms to the costs of individual enzymes ( Fig. 3 , Fig. S6 ). This information is valuable because it allows to assess the potential for reducing enzyme costs through improvements in the catalytic efficiency of the involved enzymes. While protein engineering approaches cannot change the thermodynamic properties of a given reaction, the availability of faster and/or more saturated enzyme can lead to a decrease in enzyme costs. Fig. 3 Relative contributions of the thermodynamic driving force (reversibility), catalytic power of the enzyme (turnover) and metabolite concentrations (saturation) to the enzyme costs, according to the results of the ECM analysis, for the EMP, ED and NOGEMP pathways (stacked bars graphics). In the Supplementary Material it is shown how these contributions were calculated. Using the molecular weights of the polypeptides, the corresponding masses were calculated and plotted (pie charts). Proteins whose contributions were below 1% were not represented in the pie charts. Fig. 3 3.4 Oxygen-limited cultures could increase PHB yield In the previous section, we estimated the fractions of the proteome dedicated to sustain the glycolytic flux required to cover the ATP costs for maintenance. However, the actual glycolytic flux (and its associated enzyme cost) will be higher after considering the resources required to make biomass. One possibility to decrease the glycolytic flux required to cover the ATP demands is to “burn” a fraction of the taken-up carbon source by oxidative phosphorylation, a process that yields a much larger amount of ATP per hexose. With the aim to quantify the effect on the PHB accumulation of implementing a limited supply of oxygen, we expanded the stoichiometric analysis including the reactions required for the tricarboxylic acid cycle, oxidative phosphorylation and biomass synthesis. Sucrose uptake was implemented as a proton:sucrose symporter. Two different cultivation modes were simulated: (i) a sucrose-limited continuous culture with different levels of oxygen supply, and (ii) a batch culture with different levels of oxygen supply. For the simulation of the continuous culture, the growth rate was fixed to 0.05 h −1 (corresponding to the biomass composition of the employed in silico model [ 18 ]) and the optimization goal was the minimization of the sucrose consumption rate. On the other hand, for the simulation of the batch culture, sucrose uptake rate was fixed to 10 mmol/gCDW/h (according to experimentally observed sucrose uptake rates in E. coli W [ 51 ]) and maximization of the biomass formation was set as the optimization goal. To evaluate the effect of the maintenance ATP costs, the simulations were executed for two experimentally obtained values of this bioenergetics parameter: 3.2 mmol ATP /gCDW/h [ 18 ] and 16.4 mmol ATP /gCDW/h [ 52 ]. According to our simulations, both for the case of a continuous culture or a culture in batch, the increase in the oxygen supply, until a certain critical value, increases the PHB yield ( Fig. 4 and Fig. S7 ). However, an increase in the oxygen supply beyond this critical value decreases the PHB yield. The oxygen consumption rates corresponding to the critical value change depending on the culture setup and the maintenance ATP cost. The reason behind this behavior is that the availability of oxygen enables the re-oxidation to menaquinone of the menaquinol generated in the anabolic reaction catalyzed by the dihydroorotate dehydrogenase (EC 1.3.5.2) without requiring the formation of succinate in the reaction catalyzed by the succinate dehydrogenase (EC 1.3.5.1). Therefore, with the increase in the availability of oxygen until the critical value, a lower flux through the succinate dehydrogenase is required for the re-oxidation of the menaquinol generated in the anabolism. This relationship between the oxygen availability, the production of succinate and the PHB yield can be appreciated both in the continuous culture and the culture in batch ( Fig. 4 and Fig. S7 ). Beyond the critical value (zero succinate production), the supply of oxygen exceed the amount required for the re-oxidation of menaquinol and it drains electrons that otherwise can be sink in the PHB. Using the previously described approach, the PHB content could reach values of 80–90% of the total cell weight, which is comparable with the highest values reported for E. coli [ 53 , 54 ], but with a yield of 0.74 g PHB /g hexose (batch) or 0.92 g PHB /g hexose (continuous) which would be two times higher than the values of 0.36–0.4 g PHB /g hexose previously reported [ 53 , 54 ]. Fig. 4 Different production parameters as a function of the oxygen consumption rate (q O2 ) for a continuous culture (D = 0.05 h −1 ) of engineered E. coli cells with the NOGEMP pathway, using sucrose as the sole carbon source. Different production parameters as a function of the oxygen consumption rate (q O2 ). Two different maintenance ATP costs were considered during the calculations: 3.2 mmol ATP /g CDW /h (upper row) and 16.4 mmol ATP /g CDW /h (lower row). Shown values correspond to the simulation of a continuous culture with a dilution rate of 0.05 h −1 . On the left side, sucrose consumption rate ( * ) and PHB content ( o ); on the right side, succinate production rate ( x ) and PHB yield ( + ). Fig. 4 Using a limited supply of oxygen, the amount of proteins required to sustain glycolysis will decrease but other proteins will be required. However, many of the enzymes required for the oxidation of glucose/sucrose have to be expressed anyways to generate the building blocks of the biomass. Nevertheless, the use of oxygen implies more operational costs. Still, another possibility to implement the proposed metabolic engineering approach is to divide the whole biotechnological process into a phase of aerobic biomass formation followed by a phase of anaerobic PHB accumulation. Anyways, the actual costs will depend on many factors, and a techno-economical evaluation of what is the more profitable option is beyond the scope of this manuscript. 3.5 On the experimental implementation of NOGEMP in E. coli According to our stoichiometric analysis, the NOGEMP pathway enables the coupling of PHB accumulation and NADH re-oxidation, with net ATP generation. In the Supplementary Material we outlined a possible genetic engineering strategy to materialize this metabolic engineering proposal. Currently, there is a handful of techniques to efficiently perform gene knock-in/outs and modulate the gene expression. A specific discussion about which is the most suitable genetic tool to implement the NOGEMP is beyond the scope of this paper. However, we acknowledge the necessity of elaborate more about some key parts or modules required for the functioning of the engineered pathway. First, it should be noticed that the pyruvate dehydrogenase activity is required to implement the NOGEMP. The most effective way to achieve pyruvate dehydrogenase activity under anaerobic conditions in E. coli is through the modification of the regulatory elements in the promoter, as already published [ [55] , [56] , [57] ]. Another alternative would be to maintain the usual anaerobic conversion of pyruvate to AcCoA catalyzed by the pyruvate formate lyase (E.C. 2.3.1.54), but substituting the formate dehydrogenase (E.C. 1.17.1.9) from E. coli (which use menaquinone as cofactor) by the NAD-dependent formate dehydrogenase from Candida boidinii [ 58 ]. Second, our choice for the acetoacetyl-CoA reductase was the engineered enzyme obtained by a combination of structural elements from the protein encoded by the phaB1 from C. necator and an acetoacetyl-CoA dehydrogenase isolated from Candidatus Accumulibacter phosphatis [ 59 ]. The resultant enzyme (Chimera 5) has a high preference for NADH under physiological conditions and one of the highest turnover recorded among the kinetically characterized homologues. Third, the functionality of the NOGEMP depends on the operation of the synthetic non-oxidative glycolysis [ 26 ]. The key step in this pathway is the reaction catalyzed by a phosphoketolase. The phosphoketolase from Bifidobacterium adolescentis seems a good choice, as it has both xylulose-5-phosphate-dependent (E.C. 4.1.2.9) and fructose-6-phosphate-dependent (E.C. 4.1.2.22) activities, and it is functional in E. coli [ 26 , 28 ]. Moreover, it was already successfully expressed in a previous effort aiming an increase in PHB production [ 60 ]. The introduction of the non-oxidative glycolysis implies that two thirds of the glycolytic flux goes to acetyl-CoA, skipping oxidation steps, and that a smaller flux goes through the lower EMP. This modification decreases the NADH yield per hexose and matches the catabolic production and the PHB-linked consumption of NADH. Remarkably, this smaller flux through the lower EMP rules out the use of the PEP phosphotransferase system to fuel the glucose uptake because the flux through the reaction catalyzed by a PEP synthase (E.C. 2.7.9.2) required to replenish the PEP pool would imply zero net ATP production. On the other hand, the decreased flux through the lower EMP implies a decrease in the ATP yield per hexose, increasing the absolute flux required to cover the ATP expenses, stretching the enzyme cost. To overcome this situation, we propose (i) to extend the metabolic engineering intervention to include the passive diffusion and phosphorolysis of sucrose and/or (ii) the use of oxygen-limiting cultures. An important observation arising from our calculations is the potential role of substrate channeling in PHB accumulation. The molecular mechanisms of substrate channeling and the positive effect of multi-enzyme assemblies and scaffolds for reaction kinetics are current subjects of debate [ 61 , 62 ]. Engineering spatial proximity between multiple enzymes of a metabolic cascade has been shown to increase product titers in other cases [ 63 ]. Indeed, the three enzymes enabling the formation of PHB from acetyl-CoA have been physically approximated in E. coli with the use of molecular scaffolds, and an increase of 2.5-fold in PHB production was observed after the introduction of such modification [ 34 ]. To represent the substrate channeling in our calculations, we used the approach proposed by Bar-Even and co-workers, where the reactions involved in a substrate channeling event can be considered as a single lumped reaction [ 6 , 64 ]. Although it has not been extensively elaborated, that model assumes the transfer of intermediates between two enzymes as a change between different transition states of the net reaction. The existence of this kind of transition states is not considered in the typical calculations of Δ r G’. This model remains to be experimentally validated, but if it is indeed validated, MDF, CVA and ECM analyses would be powerful tools for identifying reactions likely engaged in substrate channeling. In any case, for the final calculations of the enzyme costs, substrate channeling was not considered. There are two important potential limitations of our approach that require some discussion. First, for the estimations of the enzyme costs, we based our calculations in the fluxes required to cover 3.2 mmol ATP /g CDW /h for maintenance. Values as high as 16.4 mmol ATP /g CDW /h had been previously reported for E. coli growing under anaerobic conditions [ 52 ]. However, this latter value was obtained assuming a different value of P/O ratio and in the presence of acetate, which could cross back into the cytoplasm as acetic acid and dissociate inside the cells. The dissociation of the acetic acid generates protons, and the pumping-out of these protons cost energy. In our proposed approach, the generation of acetate should be decreased or suppressed by genetic engineering. The second important limitation is that the protein costs determined by ECM correspond to the most favorable thermodynamic driving force (MDF) calculated for a simplified network. It is possible that the actual metabolite concentrations ranges are more constrained because other reactions (anabolism) need to be thermodynamically feasible as well. The smaller the MDF, the higher the protein cost because the enzymes will be operating in a less favorable thermodynamic situation. Nevertheless, it is possible that some of the design choices defined in our strategy may themselves enhance the thermodynamic driving forces. Non-oxidative glycolysis has been shown to significantly increase acetyl-phosphate concentrations (likely increasing acetyl-CoA concentrations as well) [ 26 ]. Additionally, the deletion of genes involved in the formation of ethanol and lactate could increase the NADH/NAD ratio as it eliminates potential electron sinks other than PHB. Moreover, the suppression of the formation of acetate should eliminate a process draining acetyl-CoA. Finally, it is important to address that, as steady state based optimization algorithms, both MDF and ECM may not accurately predict in vivo dynamic conditions, but these methods can be effective tools for identifying key factors hindering the effectiveness of a metabolic engineering strategy [ 65 ]. They are, definitively, a very useful complement to stoichiometry-based tools such as Flux Balance Analysis or the Elementary Modes Analysis, as they consider thermodynamic and kinetic properties of the reactions conforming the conversion network."
} | 11,510 |
28303150 | PMC5332412 | pmc | 4,298 | {
"abstract": "Cyanobacteria are oxygenic photosynthetic prokaryotes that are able to assimilate CO 2 using solar energy and water. Metabolic engineering of cyanobacteria has suggested the possibility of direct CO 2 conversion to value-added chemicals. However, engineering of cyanobacteria has been limited due to the lack of various genetic tools for expression and control of multiple genes to reconstruct metabolic pathways for biochemicals from CO 2 . Thus, we developed SyneBrick vectors as a synthetic biology platform for gene expression in Synechococcus elongatus PCC 7942 as a model cyanobacterium. The SyneBrick chromosomal integration vectors provide three inducible expression systems to control gene expression and three neutral sites for chromosomal integrations. Using a SyneBrick vector, LacI-regulated gene expression led to 24-fold induction of the eYFP reporter gene with 1 mM isopropyl β-D-1-thiogalactopyranoside (IPTG) inducer in S. elongatus PCC 7942 under 5% (v/v) CO 2 . TetR-regulated gene expression led to 19-fold induction of the GFP gene when 100 nM anhydrotetracycline (aTc) inducer was used. Gene expression decreased after 48 h due to degradation of aTc under light. T7 RNA polymerase-based gene expression resulted in efficient expression with a lower IPTG concentration than a previously developed pTrc promoter. A library of T7 promoters can be used for tunable gene expression. In summary, SyneBrick vectors were developed as a synthetic biology platform for gene expression in S. elongatus PCC 7942. These results will accelerate metabolic engineering of biosolar cell factories through expressing and controlling multiple genes of interest.",
"conclusion": "Conclusion A synthetic biology platform of SyneBrick vectors for gene expression in S. elongatus PCC 7942 was developed with different gene inducible systems and chromosomal integration sites. LacI-pTrc, TetR-pTetA, T7RAP-T7-based SyneBrick vectors controlled expression of target genes and showed strong gene expression in S. elongatus PCC 7942 under 5% CO 2 . A synthetic T7 promoter library was used for a broad range of gene expression in S. elongatus PCC 7942. Thus, the SyneBrick vectors could be useful for expressing target genes in synthetic biology to engineer S. elongatus PCC 7942 as a biosolar cell factory for directly converting CO 2 to value-added chemicals.",
"introduction": "Introduction Increasing concerns about limited fossil fuels and the advent of global warming have drawn attention to direct conversion of CO 2 to high-value products through engineering cyanobacteria and microalgae (Ducat et al., 2011 ; Angermayr et al., 2015 ). These methods have high productivity per acre compared to terrestrial crops (Parmar et al., 2011 ). Metabolic engineering of cyanobacteria has been applied to development of biosolar factories from CO 2 with the aim of enhancing production of bio-products through modification of metabolism and introduction of heterologous metabolic pathways (Atsumi et al., 2009 ; Liu et al., 2011 ; Choi et al., 2016 ; Chwa et al., 2016 ; Lee et al., 2017 ). To reconstruct heterologous metabolic pathways in a cyanobacterium, multiple genes must be inserted into either a plasmid or the genome. Controllable expression of those genes is required to balance metabolic flux toward desired products. Tools for controlling gene expression in cyanobacteria have been developed by engineering TetR-regulated promoters in Synechocystis sp. PCC 6803 (Huang and Lindblad, 2013 ). Other tools include isopropyl β-D-1-thiogalactopyranoside (IPTG) and anhydrotetracycline (aTc)-based inducible promoters and a transacting small RNA system for Synechococcus sp. PCC 7002 (Markley et al., 2015 ; Zess et al., 2016 ), designed oxygen-responsive genetic circuits in Synechocystis sp. PCC 6803 (Immethun et al., 2016 ), and theophylline riboswitches in Synechococcus elongatus PCC 7942 (Nakahira et al., 2013 ). In addition, synthetic biology-based plasmid vectors (e.g., pPMQAK1) have been developed for Synechocystis sp. PCC 6803 (Huang et al., 2010 ). A synthetic platform for gene expression in S. elongatus PCC 7942, a model cyanobacterium for metabolic engineering and circadian rhythms, has not been reported. This study aimed to develop SyneBrick™ vectors as synthetic platform for controllable gene expression in S. elongatus PCC 7942 under 5% (v/v) CO 2 . The standardization of biological parts and their assembly is a core idea behind synthetic biology. Standardized assembly approaches such as BglBrick cloning do not require PCR amplification and use sequence-based homologous recombination. Thus, these cloning methods are not limited by automation or in the number of DNA fragment copies used (Hillson, 2011 ). BglBrick standard vectors and the derivative CoryneBrick vectors have been successfully applied for metabolic engineering of Escherichia coli and Corynebacterium glutamicum to express multiple target genes for synthetic biology (Lee et al., 2011 , 2014 ; Kang et al., 2014 ). Based on these synthetic biology platforms, we developed SyneBrick vectors with regulatory and promoter sections for controlling gene expression. We expect that SyneBrick vectors will facilitate metabolic engineering applications in S. elongatus PCC 7942.",
"discussion": "Results and discussion SyneBrick vectors: features To engineer S. elongatus PCC 7942 for gene expression, we developed SyneBrick vectors capable of expressing multiple genes controlled by different promoters and transcriptional regulators. An chromosomal integration plasmid with BglBrick cloning features, termed SyneBrick vector, was constructed based on BglBrick (Lee et al., 2011 )/CoryneBrick vectors (Kang et al., 2014 ), using the CPEC cloning method (Quan and Tian, 2011 ; Figures 1A,B ). The set of SyneBrick vectors used three different chromosomal integration sites (Neutral site I, II, and III) where no genetic interference occurred. Although many combinations of neutral sites with antibiotic selection markers (Spec R , Km R , and Cm R ) were possible, we chose three pairs of SyneBrick vectors, NSI-Spec R , NSII-Km R , NSIII-Cm R , to generate pSe1Bb1s-GFP, pSe2Bb1k-GFP, and pSe3Bb1c-GFP, respectively. These three vectors can be used for the seven combinatorial construction of the cyanobacterial strains: single integrations, Se1Bb1s, Se2Bb1k, Se3Bb1c; double integrations, Se12Bb11sk, Se13Bb11sc, Se23Bb11kc; and a triple integration, Se123Bb111skc. Additional neutral sites could be identified using transcriptome sequencing analysis (Ng et al., 2015 ). The initial SyneBrick vectors have been successfully applied to metabolic engineering (Choi et al., 2016 ; Chwa et al., 2016 ; Lee et al., 2017 ). To expand a selection of the SyneBrick vectors, LacI-pTrc in the standard SyneBrick vectors pSe1Bb1s-GFP and pSe2Bb1k-eYFP were replaced with transcriptional regulator TetR-pTetA and pT7 promoter, respectively. This constructed strains Se1Bb2s-GFP and Se2Bb7k-eYFP. For interpreting the features of SyneBrick vectors, the nomenclature is Se# for S. e longatus and neutral sites (#) and Bb for BglBrick vector-originated (Figure 1B ). Annotations of BglBrick vectors are one for the trc promoter, two for the tetA promoter, seven for the T7 promoter, s for spectinomycin, k for kanamycin, and c for the chloramphenicol resistant gene. Name of genes of interest are followed by a dash. SyneBrick vectors share the features of BglBrick/CoryneBrick vectors. A PCR step is not necessary for gene cloning and multiple gene assembly is possible with repeated enzyme digestions and ligations without choosing unique restriction enzyme sites. The details of the cloning steps have been described (Lee et al., 2011 ; Kang et al., 2014 ). In summary: A target gene of interest digested at Eco RI/ Bam HI sites is ligated to Eco RI/ Bgl II sites of SyneBrick vectors (Table 1 ) by replacing GFP or eYFP. Each target gene with a BglBrick compatible sequence is sequentially inserted into the SyneBrick vectors in multiple gene clonings. Alternatively, multiple genes built into other BglBrick (Lee et al., 2011 ) or CoryneBrick vectors (Kang et al., 2014 ), or BglBrick-formatted genes used in Synechocystis sp. PCC 6803 (Huang et al., 2010 ) or S. elongatus PCC 7002 (Markley et al., 2015 ) can be cloned into SyneBrick vectors by one-step digestion and ligation. Target genes must be checked for the presence of Eco RI, Bgl II, Bam HI, and Xho I sites prior to multiple-gene assembly in the vector. Development of controllable gene expression for S. elongatus PCC 7942 To test for controllable gene expression, GFP-expressing, or eYFP-expressing S. elongatus PCC 7942 strains (Se1Bb1s-eYFP, Se1Bb1s-GFP, and Se1Bb2s-GFP) were constructed after natural transformation with SyneBrick vectors. We tested the wild-type strain with IPTG up to 10 mM and found no growth inhibition (Figure 2 ). Subsequently, engineered strains were cultivated in BG-11 medium with 5% (v/v) CO 2 bubbling in the presence of IPTG or aTc, depending on the strain. GFP or eYFP fluorescent levels were measured. Figure 2 Gene expression analysis of engineered S. elongatus PCC 7942 using SyneBrick vectors. (A) Growth of S. elongatus PCC 7942 wild type under 5% (v/v) CO 2 bubbling in BG-11 supplemented with indicated concentrations of IPTG measured at OD 730 (black square, 0 mM; red circle, 0.01 mM; blue triangle, 0.1 mM; inverted pink triangle, 1 mM; green diamond, 10 mM). (B) Growth of engineered S. elongatus PCC 7942 (Se1Bb1s-eYFP and Se1Bb1s-GFP) under 5% (v/v) CO 2 bubbling in BG-11 supplemented with indicated concentrations of IPTG, measured at OD 730 (upper panel). Specific fluorescence (intensity per OD 730 ) for cyanobacterial cultures (lower panel). Se1Bb1s-None was the control strain (open black square). Symbols for IPTG used are as in (A) . (C) Strain Se1Bb2s-GFP was used to measure growth and specific fluorescence under aTC induction (black square, 0 mM; red circle, 10 nM; blue triangle, 100 nM; inverted pink triangle, 200 nM; green diamond, 1000 nM). Experiments were performed in triplicate cultures. All data are mean ± standard deviation ( SD ) from triplicate cultures. IPTG induction up to 10 mM and aTc up to 1 mM did not cause growth inhibition of the engineered strains although slight growth inhibition with 10 mM IPTG or 1 mM aTC was observed in the engineered strains compared to the wild type. Strain Se1Bb1s-None was constructed as a control for calibrating values for gene expression of engineered strains because the wild type shows auto-fluorescence in the range of measurements (McEwen et al., 2013 ). Compared with a control, strains Se1Bb1s-eYFP, Se1Bb1s-GFP, and Se1Bb2s-GFP showed low levels of leaky GFP/eYFP expression in the absence of inducers (Figure 2B ). When 0.01, 0.1, 1, and 10 mM of IPTG was used, higher IPTG concentrations resulted in higher induction of the reporter gene. Increased induction levels over cultivation days were also observed. Compared with 0 mM IPTG, induction of Se1Bb1s-GFP with 1 mM resulted in a 10-fold change in gene expression and 10 mM resulted in an 11-fold change. This result was consistent with the 16-fold induction of β-glucuronidase activity in a reporter assay using an IPTG-inducible pTrc gene expression system in S. elongatus PCC 7942 (Geerts et al., 1995 ). Compared with Se1Bb1s-GFP, Se1Bb1s-eYFP had a 24-fold higher induction with 1 mM IPTG and 26-fold with 10 mM. This discrepancy could be due to differences in signal intensities between intracellular GFP and eYFP in cyanobacteria. At least 10-fold induction was achieved with either 1 or 10 mM IPTG in S. elongatus PCC 7942 using SyneBrick vectors. For controllable gene expression with SyneBrick vectors with the 1 promoter, 1 mM IPTG induction is recommended. Se1Bb2s-GFP with the TetR-pTetA transcriptional system was tested with four concentrations of aTc: 10, 100, 200, and 1000 nM in 5% (v/v) CO 2 . The highest induction of 19-fold occurred when 100 nM aTc was used, compared to without aTc (Figure 2C ). However, induction was decreased dramatically when more than 100 nM aTC was used. Thus, 100-nM aTC induction was recommended with SyneBrick vectors with the 2 promoter and 24-h culture. In previous studies, an aTc-dependent induction system for Synechococcus sp. PCC 7002 had induction ranges of 6- and 32-fold with 1000 ng/mL (≅ 2200 nM) aTc (Zess et al., 2016 ) until 48 h. Dynamic ranges of inductions up to 290-fold have been reported for Synechocystis sp. PCC 6803 under light-activated heterotrophic growth with 10 μg/mL (≅ 22,000 nM) aTC for 48 h (Huang and Lindblad, 2013 ). Consistent with these results, our TetR system showed decreased gene expression after 48 h due to aTc degradation in light. However, our TetR-pTetA systems required much less aTC to induce gene expression, compared with previous studies (Huang and Lindblad, 2013 ; Zess et al., 2016 ). Thus, time-dependent gene expression in Se1Bb2s-GFP under light conditions could be possible by adding aTc in a pulse-type control mode. Engineering of our TetR-pTetA transcriptional system in S. elongatus PCC 7942 with light wavelength (i.e., red light or white light) and intensities carefully considered for SyneBrick applications with aTC as a light-sensitive inducer. Sequence modification of the pTetA promoter with tetO operators could alter intrinsic expression and induction of target genes. This could be easily incorporated into SyneBrick platform vectors. Expanding synthetic biology platforms using T7 gene expression systems To narrow the dynamic range of IPTG induction with tunable gene expression in S. elongatus PCC 7942, T7 RNA polymerase (RNAP) and its cognate T7 promoter were developed in SyneBrick vectors (pSe2Bb7k-eYFP and pSe3Bb1c-T7RNP; Figure 3 ). We tested the possibility of T7 promoter-driven transcription in the absence of T7 RNAP. No gene expression was observed in Se2Bb7k-eYFP under control of the T7 promoter, indicating that the T7 promoter was not recognized by cyanobacterial RNA polymerases (data not shown). Thus, the T7 RNAP gene 1 encoding for T7 RNAP from E. coli BL21(DE3) was integrated into Se2Bb7k-eYFP using pSe3Bb1c-T7RNP, yielding strain Se23Bb7kc-eYFP/P. Figure 3 Design features of SyneBrick vectors 2.0 for gene expression using the T7 promoter in S. elongatus PCC 7942. (A) Scheme for genetic control of target gene (e.g. eYFP). With IPTG, T7 RNA polymerase (T7 TNAP) was expressed and its cognate T7 promoter was used to express target genes. Sequences of original T7 promoter and sequence variants (T7.3 and T7.4) with binding and strength regions (Temme et al., 2012 ). (B) Construction of cyanobacterial strains using pSe2Bb7k-eYFP, pSe2Bb7.3k-eYFP, pSe2Bb7.4k-eYFP, and pSe3Bb1c-T7RNAP vectors for expression of eYFP protein under control of T7 variant promoters. Sequencing primers are in Table S1 . (C) Gel images of recombinant cyanobacterial strains. Strains were verified by PCR using oligonucleotides (arrows). Sequences are in Table S1 . DNA fragments from wild-type and strains 1, 2, and 3 are shown in gel images. DNA sequences were verified. Compared with the LacI-pTrc gene expression system of Se1Bb1s-eYFP, the T7RNAP-T7 gene expression system of Se23Bb7kc-eYFP/P showed efficient induction of target gene with IPTG. Specific fluorescence intensities of Se23Bb7kc-eYFP/P increased 6-fold with 0.01 mM IPTG and 1.5-fold with 0.1 mM compared to Se1Bb1s-eYFP (Figure 4 ). Thus, induction of the T7RNAP-T7 gene expression system with 0.1 mM IPTG resulted in the same levels of induction as 1 mM IPTG in the LacI-pTrc gene expression system. The efficient induction with the low IPTG concentration could be due to an inherently fast mRNA elongation rate of the T7 RNA polymerase. Decreased gene expression was detected when more than 1 mM IPTG was used. The reason that expression was reduced with high concentrations of IPTG was because the RNA polyadenylation and degradation mechanisms in cyanobacteria are different from E. coli (Rott et al., 2003 ). Thus, orthogonal expression of the T7 polymerase gene 1 can be alternatively applied by replacing the pTrc promoter with the pTetA promoter to prevent interference with expression from the T7 promoter. The T7RNAP-T7 gene expression system induced well with 0.01 or 0.1 mM IPTG. Figure 4 Gene expression analysis of engineered S. elongatus PCC 7942 using SyneBrick 2.0 vectors. (A) Growth of engineered S. elongatus PCC (pSe23Bb7kc-eYFP/P, pSe23Bb7.3kc-eYFP/P, and pSe23Bb7.4kc-eYFP/P) under 5% (v/v) CO 2 bubbling in BG-11 supplemented with indicated concentrations of IPTG, measured at OD 730 (black square, 0 mM; red circle, 0.01 mM; blue triangle, 0.1 mM; inverted pink triangle, 1 mM; green diamond, 10 mM). (B) Specific fluorescence (intensities per OD 730 ) were calculated for strains. Experiments were performed in triplicate cultures. All data are presented as mean ± standard deviation ( SD ) from triplicate cultures. In addition, variants of T7 promoters (T7.3 and T7.4) in which a 5-bp strength-determining region was modified (Temme et al., 2012 ) were replaced with the original T7 promoter to alter promoter strength without RNA specificity (Figure 3 ). This generated strain Se23Bb7.3kc-eYFP/P with 82% eYFP expression compared to Se23Bb7kc-eYFP/P and Se23Bb7.4kc-eYFP/P, with 90% expression. These reductions in gene expression in cyanobacteria are different from gene expression levels in E. coli , which are 28% for T7.3 and 17% for T7.4 from T7 (Temme et al., 2012 ). The order of promoter strength using a T7 library in cyanobacteria was consistent with E. coli . Using a library of T7 promoter SyneBrick vectors, the process of refactoring of a gene cluster encoding for complex proteins such as α-carboxysomes would be possible as rewriting the genomes to decipher their functions in cyanobacteria. T7 lysozyme can also be used as an inhibitor for fine-tuning gene expression of the T7 promoter by controlling T7 RNA polymerase activity. The development of clustered regulatory interspaced short palindromic repeats (CRISPR) interference technology has allowed control of gene expression by inducing dCas9 encoding genes with aTc. This method has successfully been applied to lactate production from 1% CO 2 (Gordon et al., 2016 ) and to repress multiple genes in the cyanobacterial genome (Yao et al., 2016 ). Thus, SyneBrick vectors could be further expanded to precisely control gene expression by integrating CRISPR technology for metabolic engineering of S. elongatus PCC 7942."
} | 4,662 |
35990433 | PMC9386839 | pmc | 4,300 | {
"abstract": "Slippery liquid-infused surfaces (SLISs) are developed\nas a potential\nalternative to superhydrophobic surfaces (SHSs) to resolve the issues\nof poor durability in corrosion protection and wear resistance. In\nthis work, we used a simple laser processing technology to prepare\na SLIS on the aluminum alloy (7075) surface. The superhydrophobicities\nof the modified surface and the oil film formed by liquid injection\nmake the corrosive medium difficult to directly contact the surface\nand thus have a significant effect on corrosion resistance. The water\nand oil repellent SLIS exhibits durable corrosion resistance and excellent\ntribological properties compared with the SHS. The anticorrosion and\nwear resistance performances provided by the composite film have been\nassessed by multiple methods including the electrochemical test, immersion\ntest, and friction wear test. The results indicate that compared to\nthe bare surface, laser-ablated surface (LAS), and fluoroalkyl silane-modified\nSHS, the SLIS composite coating has better corrosion resistance and\nwear resistance, which is of great significance to expand the potential\napplications of 7075 aluminum alloys. The work provides a research\nbasis for expanding the practical application of SLISs in complex\nenvironments.",
"conclusion": "4 Conclusions In this work, a SLIS with\ncorrosion-resistant and wear-resistant\nproperties has been obtained on the Al alloy surface by simple laser\nprocessing technology. The electrochemical test, immersion test, and\nfriction wear test were performed to investigate and evaluate the\nperformance of the SLIS. The results showed that the SLIS exhibited\nexcellent corrosion-resistant and tribological properties compared\nto the Al alloy substrate, LAS, and fluoroalkyl silane-modified SHS.\nSpecifically, the results of the electrochemical experiment indicated\nthat the I corr of the SLIS was 1.88 ×\n10 –7 A·cm –2 , decreased by\n3 orders of magnitude compared to that of the Al alloy substrate.\nBesides, the | Z | value of the SLIS increased around\n3 orders of magnitude, and its efficiency in corrosion resistance\nwas significantly improved ( P EF = 99.75%).\nEven being immersed in the 3.5 wt % NaCl solution for 21 days, it\nstill exhibited an excellent corrosion-resistant behavior. In the\nfriction wear test, the friction coefficient of the SLIS surface was\nonly 0.12, reduced by 84.4% compared to that of the Al alloy substrate,\nand the fewest scratches were found, having an excellent tribological\nproperty. The lubricant on the surface of the SLIS provided an opportunity\nto effectively prevent Cl – in the NaCl solution\nfrom corroding the Al alloy substrate, with a great potential for\ndelaying the process of corrosion. Moreover, the array pit structures\non the Al alloy surface were not only conducive to reducing the actual\ncontact area between the SUS440C stainless steel ball and the surface\nbut also capable of trapping wear debris and reducing the damage of\nthe wear debris to the surface during scratching, leading to a reduced\nfriction coefficient. During the relative movement, the lubricant\non the surface adheres to the ball surface to provide lubrication.\nThis research provides a novel and effective strategy for corrosion\nprotection and wear resistance of Al alloy materials, and this method\nis simple in operation, low in cost, and easy to industrialize for\nhigh-volume manufacturing.",
"introduction": "1 Introduction Aluminum (Al) alloy materials\nare indispensable in the industrial\nfield due to their high strength, low density, and good processing\nperformance, which are widely used in various fields such as electronic\nengineering, petrochemical engineering, and aerospace. 1 − 4 However, corrosion occurs when all alloy materials are exposed to\nthe atmosphere, seawater, soil, and other media, resulting in equipment\ndamage and huge economic losses. 5 − 9 Al has high electrochemical activity, and Al alloys are easy to\nbe corroded in many application environments, especially in chloride\nsolutions, which seriously restricts their promotion and applications\nin various fields. 10 , 11 Therefore, how to effectively\nreduce the surface corrosion of materials has become an important\nissue in the surface treatment of Al alloy materials. 12 − 14 In recent years, with the in-depth study of superhydrophobic\nsurfaces\n(SHSs), scholars have found that metal materials with special wettability\nare of great significance in the field of corrosion protection. 15 − 17 The SHS prevents the corrosive liquid from coming into contact with\nthe material surface due to its special liquid repellent property,\nwhich is a method for effectively protecting the Al alloy substrate\nfrom corrosion. 18 − 20 In practical applications, it is found that the superhydrophobic\nproperty of coatings is not stable, which will lead to the failure\nof superhydrophobic coatings under high temperature, high pressure,\nor surface damage. 21 Compared with the\nSHS, the air layer in the micro/nanostructures is replaced by the\nlubricating liquid, forming a more stable solid–liquid composite\nlayer. The slippery liquid-infused surface (SLIS) is super slippery,\nliquid repellent, and stable even under high-temperature and high-pressure\nconditions, and the surface has self-healing and repairability after\nphysical damage. 22 − 25 On this basis, the researchers have found that the SLIS is more\nconducive for improving the surface corrosion performance due to its\nsuperior liquid repellent properties. 26 − 28 In addition, the lubricant\non the SLIS can form a lubricating film and reduce the friction and\nextrusion damage on the material surface during use. 29 Therefore, constructing a SLIS is an important method to\nfurther solve the problems of corrosion and wear of Al alloy materials. 30 , 31 To date, researchers have proposed many methods for preparing\nSLISs,\nsuch as electrochemical etching/anodizing and chemical etching methods,\nsol–gel method, spraying method, and layer-by-layer self-assembly\nmethod. 32 − 35 However, most of the methods have various disadvantages such as\nthe lack of safety, environmental pollution, and wear of coatings.\nSpecifically, electrochemical etching/anodizing and chemical etching\nmethods involve a large amount of strong acid and alkali, and an easy-to-scatter\norganic solvent needs to be used when applying the spraying method. 36 − 38 In this case, the safety of operators and the pollution of the environment\ncannot be ignored. For the sol–gel method, spraying method,\nand layer-by-layer self-assembly method, the obtained coating and\nmaterial substrate are mainly mechanically bonded, so the bonding\nstrength to the surface is low, which limits the actual use of SLISs. 39 , 40 Recently, lasers have proven to be one of the most powerful\ntools\nin the field of advanced micro/nanomanufacturing, which has been successfully\napplied in the field of surface science to regulate the wettability\nof material surfaces. 41 − 43 To date, the femtosecond laser processing technology\nhas been used as one of the main preparation methods of SLISs. Compared\nwith femtosecond laser technology, nanosecond laser technology has\nthe advantages of low cost, high efficiency, and few requirements\nfor environmental conditions, which is more suitable for the industrial\nproduction of SLISs. 44 − 46 In this work, a simple method has been used to produce\na corrosion resistance SLIS on the Al alloy substrate. Besides, the\narray pit structures have also been constructed by nanosecond laser\ntechnology. The obtained microstructures can be modified by low-surface-energy\nmaterials to fabricate an SHS, and the SLIS can be obtained by infusing\nlubricant into the SHS. The corrosion resistance and wear resistance\nof the SHS and SLIS were compared and evaluated.",
"discussion": "3 Results and Discussion 3.1 Surface Morphology, Chemical State, and Wetting\nBehavior Figure 1 shows the SEM images of the Al alloy surface after laser\nprocessing. After laser processing, it can be seen that there is an\narray pit structure on the surface, and a convex structure is formed\naround the pit due to the accumulation of molten deposits on each\nother ( Figure 1 a).\nBy further magnifying the SEM images ( Figure 1 b,c), it is observed that there are nanoscales\non the surface of the microscale pits and convex structures, which\nconstitute micro/nanoscale composite structures of the laser-machined\nAl alloy surface. The F element was observed on the fluoroalkyl silane-modified\nSHS based on the EDS results and distributed evenly on the entire\ncoating, as shown in Figure 1 d. Figure 1 SEM images of the Al alloy surface after laser processing: (a–c)\nSEM images of the laser-ablated Al alloy surface and (d) element distribution\nmaps of the fluoroalkyl silane-modified SHS. XPS was used to further analyze the chemical states\nof the surface\nelements of the samples before and after fluoroalkyl silane modifications. Figure 2 a shows the high-resolution\nXPS spectra of the LAS and the fluoroalkyl silane-modified SHS. Compared\nwith the LAS, the fluoroalkyl silane-modified SHS presents a strong\npeak of F 1s at 688.4 eV ( Figure 2 b), and the peak corresponding to −CF2 is observed\nin Figure 2 c. The results\nindicated that the fluoroalkyl silane-modified SHS was coated with\nfluoroalkyl silane. 47 , 48 Figure 2 High-resolution XPS spectra of the LAS\nand fluoroalkyl silane-modified\nSHS. Figure 3 shows the\nwetting states of water and hexadecane droplets to different Al alloy\nsurfaces. From the images ( Figure 3 a,b), the bare Al alloy surface is hydrophilic (67.8\n± 4.1°), while the laser-ablated Al alloy surface is superhydrophilic\n(0°). The Wenzel model can explain the change in wettability. 49 The actual solid–liquid contact area\nis larger than the apparent contact area, so the CA becomes smaller\nas the surface roughness increases. 50 After\nmodification with fluoroalkyl silane, a water droplet on the base\nAl alloy surface exhibits hydrophobicity with a CA of 102.5 ±\n2.1° ( Figure 3 c). Fluorination treatment can reduce the surface free energy and\nincrease the hydrophobicity of the surface, but the CA of the water\ndroplet is not greater than 120°. 51 , 52 The perfluorinated\nlubricant is injected into the hydrophobic surface and the CA is 93.9\n± 1.5°. When the surface is turned 90°, the water droplets\nstill adhere to the surface ( Figure 3 d). A water droplet on the fluoroalkyl silane-modified\nLAS exhibits superhydrophobicity, with a CA of 157.7 ± 1.5°\nand an SA of less than 2° ( Figure 3 e,f). The superhydrophobic properties of the surface\nare the result of the interaction between the micro/nanostructure\nand the chemical composition. 53 , 54 Figure 3 g–j shows the CAs and sliding states\nof the droplets to the SLIS. The CAs of the water and hexadecane droplets\nare 112.1 ± 2.3 and 53.9 ± 1.9°, respectively, and\nthe sliding speeds of the two droplets to the surface with an inclination\nangle of 5° are 0.14 ± 0.01 and 0.31 ± 0.02 mm·s –1 , respectively. In addition, the results of different\ntypes of oil droplets (glycerin, ethylene glycol, olive oil, dichloroethane,\nchloroform, n -dodecane, and hexadecane) moving on\nthe SLIS are shown in Figure 4 , and it can be seen that various droplets easily slid over\nthe surface with an inclination angle of 10°. According to the\nrelevant criteria of SLIS, the substrate of micro/nanostructure is\nconducive to the complete wetting and adhesion of the lubricating\nliquid. 55 , 56 After the perfluorinated lubricant penetrates\nand wets the micro/nanostructure, the formed lubricating film prevents\nthe droplets from directly contacting the surface of the substrate,\nresulting in super slippery properties. Therefore, when the surface\nis tilted at a certain angle, the droplets that land on the SLIS can\neasily slide off. Figure 3 Wetting states of water and hexadecane droplets to different\nAl\nalloy surfaces: (a–f) water droplets on the bare surface, LAS,\nhydrophobic surface, bare Al alloy surface injected with the perfluorinated\nlubricant, and fluoroalkyl silane-modified SHS. (g, h) CAs and (i,\nj) sliding states of (g, i) water and (h, j) hexadecane droplets to\nthe SLIS. Figure 4 Sliding states of different types of droplets on SLIS\n(inclination\nangle of 10°): (a) water, (b) glycerol, (c) ethylene glycol,\n(d) olive oil, (e) dichloroethane, (f) chloroform, (g) n -dodecane, and (h) hexadecane. The shape of the water droplet on the fluoroalkyl\nsilane-modified\nLAS was close to a circular shape (157.7 ± 1.5°), and the\nCA of the hexadecane droplet to the surface was 46.1 ± 3.4°\n( Scheme 2 a,b). The\nperfluorinated lubricant droplet diffused rapidly on the surface,\nand the CA was 0°, which was manifested as superoleophilicity\nfor the perfluorinated lubricant ( Scheme 2 c). The results show that the chemical affinities\nof the perfluorinated lubricant and the surface are higher than those\nof water or other oil droplets (glycerol, ethylene glycol, olive oil,\ndichloroethane, chloroform, n -dodecane, and hexadecane)\nand the surface. According to the theory of SLIS, when water or other\noil droplets fall onto the SLIS injected by the perfluorinated lubricant,\nthe formed perfluoro-oil film prevents the droplets from passing through,\navoiding direct contact with the surface and forming a stable oil–liquid–solid–gas\ncontact zone. Scheme 2 CAs of Droplets on a Fluoroalkyl Silane-Modified Laser-Machined\nSurface (a) Water, (b) hexadecane,\nand\n(c) perfluorinated lubricant. To evaluate\nthe air stability of the SLIS and verify whether the\nsurface lubricant evaporates over time and the performance of SLIS\nis invalid, the air stability of the SLIS is tested. The sample is\nplaced in the air for 21 days, and the CAs and SAs are measured every\n3 days ( Figure 5 ).\nThe CAs are maintained at 112 ± 3.5°, and the SAs are less\nthan 5°, indicating that the prepared SLIS has excellent air\nstability. Figure 5 Wettability of the SLIS placed in the air for different days. 3.2 Corrosion Resistance Figure 6 shows the open circuit potential\n(OCP) curves of the bare surface, LAS, fluoroalkyl silane-modified\nSHS, and SLIS in a 3.5 wt % NaCl solution. It can be seen that after\n1800 s, the OCP values of different sample surfaces have reached a\nrelatively stable state, and the OCP values are in the following order:\nbare surface > fluoroalkyl silane-modified SHS > SLIS > LAS.\nThe electrochemical\npotential of the bare surface is the highest, and the LAS is the lowest.\nIn addition, the OCP curve of the fluoroalkyl silane-modified SHS\nfluctuates greatly in the early stage, which may be caused by the\nsuperhydrophobicity of the surface. Figure 6 OCP curves of Al alloy surfaces treated\nunder different processing\nconditions. Figure 7 shows the\nPP curves of the bare surface, LAS, fluoroalkyl silane-modified SHS,\nand SLIS. Table 2 shows\nthe E corr and I corr , and the anodic/cathodic Tafel polarization slopes β a /β c of the different surfaces. The Stern–Geary\nequation is used to calculate the polarization resistance R p of different sample surfaces 57 1 where A is the exposed area\nof different samples, A = 1 cm 2 . Figure 7 Polarization\ncurves of Al alloy surfaces treated under different\nprocessing conditions. Table 2 E corr and I corr of Al Alloy Surfaces Treated under Different\nProcessing Conditions in the NaCl Solution sample E corr (V) I corr (A·cm –2 ) P EF (%) β a (mV·dec –1 ) β c (mV·dec –1 ) R p (Ω) bare Al alloys –1.27 1.01 × 10 –4 314 86.8 2.92 × 10 2 LAS –1.27 2.62 × 10 –4 326 137 1.60 × 10 2 fluoroalkyl silane-modified SHS –1.28 7.55 × 10 –6 90.73 159 83.7 3.15 × 10 3 SLIS (40 μm) –1.34 1.88 × 10 –7 99.75 144 77.4 1.16 × 10 5 Table 2 shows the\npolarization resistance R p values of different\nsample surfaces in the following order: R p-SLIS (1.16 × 10 5 Ω) > R p-SHS (3.15 × 10 3 Ω) > R p-Al (2.92 × 10 2 Ω)\n> R p-LAS (1.60 × 10 2 Ω). The corrosion current densities\nof LAS (2.62 × 10 –4 A·cm –2 ) were higher than that of the bare surface (1.01 × 10 –4 A·cm –2 ), indicating that the laser processing\nreduced the corrosion resistance of the Al alloy surface. According\nto the analysis, the LAS was superhydrophilic, which caused the chloride\nion in the corrosion solution to directly contact the surface and\npromoted the occurrence of the corrosion reaction, leading to a decrease\nin surface corrosion resistance. After modification with fluoroalkyl\nsilane, the I corr of the LAS was significantly\nreduced to 7.55 × 10 –6 A·cm –2 , suggesting that the fluoroalkyl silane modification increased the\ncorrosion resistance of the surface in addition to improving the hydrophobicity.\nFor the SLIS, I corr was 1.88 × 10 –7 A·cm –2 , and it was reduced\nby 3 orders of magnitude compared with that of the bare surface. This\nexcellent corrosion resistance property is due to the fact that the\nliquid lubricant completely covered the fluoroalkyl silane-modified\nSHS, inhibiting electron transfer between the substrate and the etching\nsolution. The corresponding corrosion protection efficiency ( P EF ) is calculated as follows 58 2 where R pp refers\nto the polarization resistance of the Al alloy substrate, and R cp is the polarization resistance of the Al\nalloy surfaces treated with different conditions. According to eq 2 , the corrosion protection\nefficiencies of the fluoroalkyl silane-modified SHS and the SLIS are\n90.73 and 99.75%, respectively. Figure 8 shows the\nNyquist plots of the bare surface, LAS, fluoroalkyl silane-modified\nSHS, and SLIS. It is found that the SLIS has the largest diameter\nin the capacitive impedance arc, followed by SHS and Al alloy substrate\nand LAS in the NaCl solution. As a matter of fact, the diameter of\nthe capacitive impedance arc is proportional to the corrosion resistance\nof the sample. In a specific way, an increasing diameter of the capacitive\nimpedance arc means an increase in the resistance of the electrochemical\nreaction. The Al alloy substrate’s Nyquist plot is composed\nof dual capacitive impedance arcs in the high- and low-frequency regions.\nThe Nyquist plot of LAS shows a capacitive impedance arc in the high-frequency\nregion and an inductive impedance arc in the low-frequency region.\nIn the pitting corrosion model of Al alloy, inductive impedance means\nthe occurrence of the pitting corrosion, and therefore, the low-frequency\ninductive impedance arc represents that the pitting corrosion occurs\non the LAS. It is concluded that both SLIS and SHS are capable of\nimproving the corrosion resistance of the Al alloy substrate. For\nthe SLIS surface, it can be explained that the lubricant present on\nthe SLIS surface is helpful to isolate the corrosion media from the\nAl alloy substrate, especially in water because the lubricant is incompatible\nwith water and can be caught for a long period of time, slowing down\nthe process of corrosion. The sample with SHS coating is immersed\nin the NaCl solution, and hence, a large amount of air is trapped\nin the micro/nanostructures of the coating. The air layer makes it\npossible to reduce the contact area between the NaCl solution and\nthe coating, delaying the occurrence of corrosion. Figure 8 Nyquist plots of Al alloy\nsurfaces treated with different conditions. As shown in Figure 9 a, the low-frequency impedance modulus (| Z |) value\nis 10 6.5 Ω·cm 2 and is significantly\nhigher than 10 3.9 Ω·cm 2 of the bare\nsurface, which is an increase of about 3 orders of magnitude. It is\nwell known that the increased | Z | value means an\nincrease in corrosion resistance. The | Z | value of\nthe LAS is the smallest (10 2.7 Ω·cm 2 ), indicating that the LAS is unacceptable in corrosion resistance.\nAccording to the analysis, the LAS is superhydrophilic, which causes\nthe chloride ions in the corrosion solution to directly contact the\nsurface and promotes the occurrence of the corrosion reaction. For\nthe fluoroalkyl silane-modified SHS, the | Z | is significantly\nincreased, and the | Z | value is 10 5.7 Ω·cm 2 . Figure 9 Bode plots of Al alloy surfaces treated with different conditions. As shown in Figure 9 b, all samples have two time constants. The time constant\nin the\nlow-frequency region represents local corrosion that occurred on the\nsurface, and the time constant in the medium-frequency region means\nthe charge transfer resistance ( R ct ) and\ndouble electric layer capacitance (CPE dl ) during the process\nof corrosion. The inhibiting effect generated by the uniform and dense\ncoating results in the appearance of a high-frequency capacitance\nloop. 59 On the other hand, two time constants\nare available to the Al alloy substrate, LAS, and fluoroalkyl silane-modified\nSHS. One is in the low-frequency region and the other is in the medium-frequency\nregion. The two time constants of SLIS are in the medium- and high-frequency\nregions. Zahner analysis software is used to fit the electrochemical\ndata\nso as to explore the process of coating corrosion. The equivalent\ncircuit is obtained, as shown in Figure 10 , where R s means\nthe solution resistance between the sample and the reference electrode. R ct and CEP dl represent the charge\ntransfer resistance and double electric layer capacitance between\nthe coating and the substrate, respectively. R ′ o is the additional resistance of the solution within the pit. 60 CEP o is the double electric layer\ncapacitance of the oxide layer. R SLIS and\nCEP SLIS are the resistance and capacitance of the oil film. R SHS and CEP SHS refer to the resistance\nand capacitance of the SHS. R Laser and\nCEP Laser are the resistance and capacitance of the LAS,\nand the inductance element L is connected in series\nin the R L circuit, which represents the\ninductance behavior due to the pitting corrosion. CPE dl is represented by the constant phase element (CPE) Q . The value of CPE dl can be calculated by Brug’s\nformula 61 , 62 3 where Q is the value of CPE\nand n is the CPE’s dimensionless exponent. 63 When n = 1, Q is equivalent to an ideal capacitor. It is observed from the fitting\nresults listed in Table 3 that R ct is increased from 2.85 kΩ·cm 2 of the Al alloy substrate to 122 kΩ·cm 2 of the SLIS, and the CPE dl value is decreased 4 orders\nof magnitude accordingly. Generally, the charge transfer resistance\nis associated with corrosion. The higher the charge transfer resistance,\nthe lower the corrosion rate. A change in the CPE dl value\nillustrates the permeation behavior of the electrolyte on the coating,\nand the increased corrosion solution permeated into the coating causes\nan increase in the CPE dl value. 63 It can be concluded that the SLIS plays a significant role in protecting\nthe Al alloy substrate from the corrosion of chloride ions. Figure 10 Equivalent\ncircuit models for the EIS fitting of different samples:\n(a) bare surface, (b) LAS, (c) fluoroalkyl silane-modified SHS, and\n(d) SLIS. Table 3 Fitting Results of the EIS Data of\nDifferent Samples in the 3.5 wt % NaCl Solution sample R ct (kΩ·cm 2 ) Q dl (Ω –1 ·s – n ·cm –2 ) n dl CEP dl (μF·cm –2 ) R coat (kΩ·cm 2 ) Q coat (Ω –1 ·s – n ·cm –2 ) bare Al alloys 2.85 7.64 × 10 –4 1 7.64 × 10 –4 5.20 1.50 × 10 –5 LAS 2.09 5.68 × 10 –6 0.815 7.07 × 10 –7 0.41 7.10 × 10 –6 fluoroalkyl\nsilane-modified SHS 44.6 7.82 × 10 –7 0.958 4.91 × 10 –7 1.02 ×10 3 8.16 × 10 –6 SLIS (40 μm) 122 6.17 × 10 –8 1 6.17 × 10 –8 3.88 × 10 3 1.14 × 10 –7 The corrosion resistance of the surface was further\ninvestigated\nby immersing the sample in the 3.5 wt % NaCl solution. Figure 11 shows the optical photographs\nof the bare surface, LAS, fluoroalkyl silane-modified SHS, and SLIS\nbefore and after immersion. Figure 11 a shows that the surfaces of the four samples that\nare not immersed are smooth and clean. As shown in Figure 11 b, after immersion for 21\ndays, it can be seen that there are many salt deposits on the bare\nand LASs. In contrast, the contaminants on the fluoroalkyl silane-modified\nSHS and the SLIS are reduced, and no significant signs of corrosion\nare observed. Also, the presence of an oil film is observed on the\nSLIS, demonstrating superior corrosion resistance. Figure 11 Optical photographs\nof Al alloy surfaces treated with different\nconditions before (a) and after (b) immersing in the 3.5 wt % NaCl\nsolution for 21 days. Among them, sample 1 is the bare surface, sample\n2 is the LAS, sample 3 is the fluoroalkyl silane-modified SHS, and\nsample 4 is the SLIS. Figure 12 shows\nthe SEM images of the bare surface, LAS, fluoroalkyl silane-modified\nSHS, and SLIS after being immersed in the 3.5 wt % NaCl solution.\nAs shown in Figure 12 a, corrosion holes with a diameter of about 10 μm appeared\non the bare Al alloy surface. For the LAS, after being immersed for\n21 days into the corrosive solution, it is found that the number of\ncorrosion holes increased, and the diameter increased to about 20\nμm ( Figure 12 b), indicating that the corrosion of LAS is significant. For the\nfluoroalkyl silane-modified SHS and the SLIS, the morphology did not\nchange ( Figure 12 c,d),\nindicating that the two surfaces have good stability in the 3.5 wt\n% NaCl solution. According to the analysis, the superhydrophobicity\nof the modified surface and the oil film formed by liquid injection\nmake the corrosive medium difficult to directly contact the surface\nand thus produce a significant anticorrosion effect. Figure 12 SEM images of Al alloy\nsurfaces treated with different conditions\nafter immersing in the 3.5 wt % NaCl solution for 21 days: (a) bare\nsurface, (b) LAS, (c) fluoroalkyl silane-modified SHS, and (d) SLIS. Table 4 shows the\nchanges of the elements Al, O, F, Na, and Cl on the bare surface,\nLAS, fluoroalkyl silane-modified SHS, and SLIS before and after the\nsamples were immersed in the 3.5 wt % NaCl solution for 21 days. For\nthe bare Al alloy surface, the weight percentage of element Al decreased\nfrom 98.3 to 92.1%, while the increase of element O (from 1.7 to 5.2%)\nand new elements Na and Cl (weight percentages of 0.9 and 1.8%, respectively)\nis observed on the surface, indicating that the bare surface is corroded\nby the NaCl solution. For the LAS, fluoroalkyl silane-modified SHS,\nand SLIS, the elements Na and Cl are found on all three surfaces,\nand the weight percentage of element O shows an increase after 21\ndays. The difference is that the weight percentages of elements O,\nCl, and Na on the LAS are 19.4, 1.7, and 3.2%, respectively, which\nare more than those of the bare surface after being immersed in the\nNaCl solution, indicating that the LAS has more serious corrosion.\nHowever, the fluoroalkyl silane-modified SHS and the SLIS have little\nchange in element O, and elements Cl and Na are also less, caused\nby residual chloride on the surface after immersion in the NaCl solution\nfor a long period of time. Simultaneously, element F is also detected\non the fluoroalkyl silane-modified SHS and SLIS, and their contents\nare 2.9 and 3.0%, respectively, which are unchanged compared to the\nunsoaked fluoroalkyl silane-modified SHS (3.0%). The results show\nthat the low-surface-energy materials are not damaged, indicating\nthat the fluoroalkyl silane-modified SHS and SLIS are substantially\nnot corroded after being immersed in the 3.5 wt % NaCl solution for\n21 days. Table 4 Changes of Elements Al, O, F, Na,\nand Cl Before and After Corrosion of Al Alloy Surfaces element\ncomposition and content (wt %) sample time (days) Al O F Na Cl bare Al alloys 0 98.3 1.7 bare Al alloys 21 92.1 5.2 0.9 1.8 LAS 0 93.4 6.6 LAS 21 76.1 19.4 1.7 3.2 fluoroalkyl silane-modified SHS 0 90.2 6.9 3.0 fluoroalkyl silane-modified SHS 21 89.0 7.1 2.9 0.2 0.8 SLIS (40 μm) 21 89.8 6.9 3.0 0.1 0.2 3.3 Corrosion Mechanism It can be seen\nfrom the above corrosion test that in the NaCl solution, the chloride\nion erodes the oxide film on the surface of the Al alloy substrate,\nirregular pits appear on the surface of the Al alloy, and pitting\ncorrosion occurs ( Figure 12 ). 60 The anodic reaction is mainly\nas follows 64 4 5 According to eq 5 , the acidity of the anode position is enhanced, and\nthe chloride ion promotes the anodic dissolution of Al to form aluminum\nchloride. The following reactions may occur at the cathode 6 7 8 In the corrosion solution, the bare surface\nand the LAS are hydrophilic, and the corrosive solution is in direct\ncontact with the surface of the sample, resulting in serious corrosion\nbehavior. In addition, the micro/nanoscale composite structures of\nthe LAS increase the real contact area between the corrosion solution\nand the surface of the sample, resulting in more serious surface corrosion.\nThe fluoroalkyl silane-modified SHSs are superhydrophobic. When the\nfluoroalkyl silane-modified SHS is immersed in the corrosion solution,\nan air layer is formed on the surface, which can reduce the real contact\narea between the corrosion solution and the fluoroalkyl silane-modified\nSHS and effectively improve the corrosion resistance of the fluoroalkyl\nsilane-modified SHS ( Figure 13 a). 65 However, in the corrosive\nsolution, the superhydrophobic property of SHS gradually disappeared\nunder external pressure for a long time. 66 Compared with the fluoroalkyl silane-modified SHS, the air layer\nin the micro/nanoscale composite structures is replaced by the lubricant\nto form a more stable solid–liquid composite layer. Especially\nin water, the lubricant stored in the micro/nanoscale composite structures\nis incompatible with water, effectively isolating the contact between\nthe corrosion solution and the Al alloy substrate ( Figure 13 b). In addition, when the\nSLIS is damaged, the lubricant on the surface can quickly repair the\ndamaged area under the surface energy-driven capillary action, providing\nmore long-term and stable protection for the Al alloy substrate. 67 Figure 13 Schematic images of the corrosion-resistant mechanism\nmodel of\nthe solid–liquid interface between the Al alloy surface and\nthe 3.5 wt % NaCl solution under different treatment conditions: (a)\nfluoroalkyl silane-modified SHS and (b) SLIS. 3.4 Tribological Behaviors Figure 14 shows the friction\ncoefficient curves of different sample surfaces. It can be seen that\nthe friction coefficient of the bare surface is the largest, reaching\n0.77, and the friction coefficient of the SLIS is the smallest, about\n0.12. The friction coefficient is reduced by 84.4% compared to the\nbare surface. At the beginning of the friction test (0–700\ns), the friction coefficient of the hydrophobic surface is relatively\nsmall. The fluoroalkyl silane modification reduces the free energy\nof the bare surface. Larger surface free energy corresponds to stronger\nadhesion, which leads to higher friction between the two surfaces. 68 But with the increase of the friction test time,\nthe hydrophobic surface wears, and the friction coefficient gradually\nincreases close to the friction coefficient of the bare surface. Compared\nwith the bare surface and hydrophobic surface, the friction coefficient\nof the bare surface injected with the perfluorinated lubricant is\nsmaller. This is because when the relative movement between the friction\npairs occurs, the lubricant on the bare Al alloy surface adheres to\nthe surface of the SUS440C stainless steel ball with a lubricating\neffect. Compared with the bare Al alloy surface, the friction coefficients\nof the LAS and fluoroalkyl silane-modified SHS are small, and the\nmicropit structures on the LAS can store the wear debris generated\nduring the friction process and reduce the contact between the wear\ndebris and the substrate. However, the friction coefficient of the\nSHS is not significantly lower than that of the LAS. This is because\nthe SHS wears and the superhydrophobic performance becomes invalid\nwhen the relative movement between the friction pairs occurs. Most\nsurprisingly, the friction coefficient of the SLIS is significantly\nlower than that of the bare surface injected with the perfluorinated\nlubricant. When the friction pairs move relatively, the lubricant\non the surface of the substrate spontaneously adheres to the surface\nof the stainless steel ball, providing a lubricating medium for the\ncontact interface. However, with the continuous friction test, the\nsubstrate surface is not conducive to the storage of lubricant, and\nthe lubricant on the surface is lost, which cannot continuously provide\na lubricating medium for the contact interface, and the friction coefficient\nincreases. In contrast, the micropit structures of the SLIS are more\nconducive to the storage of the lubricant, preventing the loss of\nthe lubricant on the surface during frictional movement and providing\ncontinuous lubrication for the friction pair. The results show that\nthe lubricating effect of the oil film and the storage function of\nthe microstructures play an important role, and the phenomenon that\nthe grinding debris adheres to the surface to increase the frictional\nforce is avoided. Figure 14 Average values of the friction coefficients of different\nsample\nsurfaces. Figure 15 shows\nthe SEM images of the surface wear of the bare surface, hydrophobic\nsurface, bare Al alloy surface injected with the perfluorinated lubricant,\nLAS, fluoroalkyl silane-modified SHS, and SLIS. It can be seen in Figure 15 a,b that the bare\nand hydrophobic surfaces have clear furrow structures in the sliding\narea along the sliding direction, and there is wear debris on the\nwear scar. The wear is mainly adhesive wear, with some abrasive wear.\nMoreover, there are cracks and depressions on the bare surface, which\nare consistent with the sliding direction ( Figure 15 a2). 69 The width\nof the wear scar to the bare surface is 189 μm, which has the\nlargest value in the width of wear scars. The wear scar on the hydrophobic\nsurface is relatively light, about 177 μm. The wear scar on\nthe surface of the Al alloy substrate injected with the perfluorinated\nlubricant is 89 μm ( Figure 15 c), and there is no obvious furrow structure on the\nsurface, and the adhesion wear and the wear scar are significantly\nreduced. As shown in Figure 15 d,e, the LAS and fluoroalkyl silane-modified SHS have larger\nvalues in the widths of wear scars, while the depths of wear scars\nto the two surfaces are significantly reduced relative to the bare\nsurface described above, indicating that the microstructures of the\nlaser processing improve the wear resistance of the surface to a certain\nextent. 70 The wear scar on the SLIS is\n123 μm, indicating that the wear resistance of the Al alloy\nsurface is significantly improved. Figure 15 SEM images of the surfaces of different\nsamples in the friction\nand wear tests: (a) bare surface, (b) hydrophobic surface, (c) bare\nAl alloy surface injected with the perfluorinated lubricant, (d) LAS,\n(e) fluoroalkyl silane-modified SHS, and (f) SLIS. Based on the above results, the micropit structures\nconstructed\non the Al alloy surface by laser processing technology have proved\nto be capable of improving the tribological performance. The micropit\nstructures on the LAS can store the wear debris, reducing the contact\nbetween the wear debris and the substrate during the friction process,\nthereby reducing the friction coefficient and the probability of the\noccurrence of abrasive wear. In addition, when the relative movement\nbetween the friction pairs occurs, the perfluorinated lubricant injected\non the surface spontaneously adheres to the surface of the small ball\nand hence provides a lubricating medium for the contact surface. The\ncombination of nanosecond laser processing and surface lubricant injection\nimparts excellent tribological properties to the Al alloy surface."
} | 8,906 |
28323280 | PMC5480597 | pmc | 4,301 | {
"abstract": "Microbial communities are essential to a wide range of ecologically and industrially important processes. To control or predict how these communities function, we require a better understanding of the factors which influence microbial community productivity. Here, we combine functional resource use assays with a biodiversity–ecosystem functioning (BEF) experiment to determine whether the functional traits of constituent species can be used to predict community productivity. We quantified the abilities of 12 bacterial species to metabolise components of lignocellulose and then assembled these species into communities of varying diversity and composition to measure their productivity growing on lignocellulose, a complex natural substrate. A positive relationship between diversity and community productivity was caused by a selection effect whereby more diverse communities were more likely to contain two species that significantly improved community productivity. Analysis of functional traits revealed that the observed selection effect was primarily driven by the abilities of these species to degrade β-glucan. Our results indicate that by identifying the key functional traits underlying microbial community productivity we could improve industrial bioprocessing of complex natural substrates.",
"introduction": "Introduction Microbial communities underpin the functioning of natural ecosystems ( Soliveres et al. , 2016 ) and the efficiency of a wide range of industrial bioprocesses (for example, waste bioreactors) ( Cydzik-Kwiatkowska and Zielińska, 2016 ; Widder et al. , 2016 ). The form of the biodiversity–ecosystem functioning (BEF) relationship is therefore an important property of microbial communities both in nature and the simpler communities used in a range of industrial bioprocesses. Several studies have identified positive BEF relationships for microbial community productivity ( Bell et al. , 2005 ; Gravel et al. , 2011 ), stability ( Awasthi et al. , 2014 ), micropollutant degradation ( Johnson et al. , 2015 ) and resistance to invasion ( Elsas et al. , 2012 ), suggesting that for a range of functions microbial community performance improves with increasing species richness. Positive BEF relationships can arise via the complementarity effect, whereby diverse communities use more of the available resource space through niche differentiation or facilitation ( Salles et al. , 2009 ; Singh et al. , 2015 ), or the selection effect (also termed the sampling effect), whereby diverse communities are more likely to contain species which have a large impact on community functioning ( Hooper et al. , 2005 ; Langenheder et al. , 2012 , 2010 ; Awasthi et al. , 2014 ). Both complementarity and selection effects depend on the functional traits of constituent species and several studies have now shown functional diversity to be a better predictor of community function than phylogenetic diversity ( Mokany et al. , 2008 ; Salles et al. , 2009 ; Krause et al. , 2014 ). However, for many ecologically and biotechnologically important microbial communities it is still unclear how the functional traits of individual species scale-up to determine the performance of a diverse community. One of the most important ecosystem functions microbial communities perform is the decomposition of plant material and subsequent nutrient cycling ( Van Der Heijden et al. , 2008 ; McGuire and Treseder, 2010 ). Understanding the decomposition of plant material also has important industrial relevance. Plant biomass (collectively referred to as lignocellulose) is the most abundant raw material on Earth ( Pauly and Keegstra, 2008 ). It is typically composed of approximately 40–50% cellulose, 20–40% hemicellulose and 20–35% lignin which together form a complex, recalcitrant structure ( Himmel et al. , 2007 ; Liao et al. , 2016 ). The high sugar content and abundance of lignocellulose make it a promising substrate for biofuel production ( Naik et al. , 2010 ). However, lignin is highly recalcitrant to enzymatic attack causing a bottleneck in the efficient conversion of lignocellulose to biofuels reducing cost-effectiveness ( Jorgensen et al. , 2007 ; Naik et al. , 2010 ). Understanding how natural microbial communities (for example, in soils ( Lynd et al. , 2002 ), compost ( Lopez-Gonzalez et al. , 2014 ) or termite guts ( Brune, 2014 )) achieve efficient lignocellulose degradation could inform both the prediction of nutrient cycling in natural systems and the design of efficient microbial communities for industrial processes ( Wei et al. , 2012 ). Both biodiversity and the presence of certain species have been shown to influence the rate of decomposition by bacterial communities ( Bell et al. , 2005 ; Bonkowski and Roy, 2005 ; Langenheder et al. , 2012 ) but the mechanisms which determine community decomposition performance remain poorly understood ( McGuire and Treseder, 2010 ). A key question therefore is to what extent community functioning is predictable from the combined functional traits of constituent species? Using culturable bacterial strains isolated from compost we performed a random partition design BEF experiment ( Bell et al. , 2009 ) to test the contributions of species richness and composition to productivity of communities when grown on wheat straw. Although using only the culturable fraction of the community is likely to overlook some functionally important species in the natural community, culturability is a key feature of microbes that could feasibly be used in industrial bioprocessing. Next we tested how the functional traits of individual species shaped the productivity of these communities to determine the extent to which community productivity was predictable from the functional traits of the constituent species and to determine the contribution of each functional trait to overall productivity. We quantified the functional resource use traits of each species by their ability to utilise a range of known components of lignocellulose (that is, cellulose, hemicellulose, pectin and lignin).",
"discussion": "Discussion Understanding the factors that influence microbial community productivity has potentially important ecological and industrial applications ( Widder et al. , 2016 ). The ability of community niche to predict functioning in well-defined media has been demonstrated previously ( Salles et al. , 2009 ). Here, we define for communities growing in complex undefined media, the key functional resource use traits that predict decomposer community productivity. Crucially, functional resource use traits explained more variation in productivity than either species richness or measures of community niche. Indeed, a single function, the ability to degrade β-glucan, explained a larger proportion of variation than community niche. This key functional trait was shared by two dominant strains which were shown to significantly increase the productivity of communities. As with several previous BEF studies ( Bell et al. , 2009 ; Gravel et al. , 2011 ; Awasthi et al. , 2014 ), we identified a positive relationship between species richness and community productivity. By analysing the effect of community composition we found that the presence of two highly functioning species, Paenibacillus sp. A8 and C. flavigena D13, significantly increased community productivity suggesting this positive BEF relationship is driven by the selection effect. To determine if the dominance of these two species could be explained by their functional traits, we compared the ability of these species to utilise the various carbon sources used in functional trait assays to the other species. With the exception of Rhodococcus sp. E31, Paenibacillus sp. A8 and C. flavigena D13 were the highest performing species on β-glucan ( Figure 2 ). The ability to utilise β-glucan may suggest these species are able to metabolise the cellulose portion of wheat straw in addition to the more labile hemicellulose and pectin fractions. Interestingly, when the productivity of communities containing either one, both or neither of these species is compared across each day of the experiment ( Supplementary Figure 3 ), it is noticeable that communities containing neither of these species have very low productivity during the later days of the experiment. A possible explanation is that easily accessible labile substrates are being used within the first two days of growth after which only recalcitrant and inaccessible substrates remain. The ability to degrade cellulose would allow Paenibacillus sp. A8 and C. flavigena D13 to maintain higher levels of growth when labile substrates become depleted. Interestingly, Paenibacillus sp. A8 and C. flavigena D13 have similar functional traits which would indicate they occupy overlapping niche space and may be in direct competition with each other. However, communities containing both these species were significantly more productive than communities containing only one or neither suggesting complementarity or facilitation effect between these species, that is, they are able to exploit a wider niche space when grown together potentially because they each produce enzymes or by-products that improve the overall community productivity. Wohl et al. (2004) found a similar result whereby functionally redundant cellulose degrading bacteria were more productive in communities than in monoculture. The ability of species within communities to utilise β-glucan was a better predictor of community productivity than measures of community niche or species richness. The significance of this activity is consistent with the composition of wheat straw lignocellulose, which is made up of 40–50% cellulose. Interestingly, functional trait assays revealed that Rhodococcus sp. E31 achieved the second highest growth on β-glucan but this species did not significantly increase community productivity compared to an average species ( Supplementary Figure 2 ). In addition, Rhodococcus sp. E31 was able to utilise lignin as well as the more labile hemicellulose substrates ( Figure 2 ). It might have reasonably been expected that as lignin is the major contributing factor to recalcitrance, species able to degrade it would increase community productivity by increasing accessibility of saccharification enzymes to cellulose. The limited contribution of Rhodococcus sp. E31 to community productivity may be explained in part by structural differences between Kraft lignin used in functional trait assays and native lignin present in lignocellulose ( Vishtal and Kraslawski, 2011 ). Alternatively, although able to achieve efficient degradation of all substrates in monoculture growth assays, Rhodococcus sp. E31 may be outcompeted in communities and unable to achieve the functional potentials revealed by trait assays. Recalcitrant substrates may require more energy expensive breakdown pathways than labile substrates ( Lynd et al. , 2002 ) which may put species that are specialised to degrade such substrates, for example, Rhodococcus sp. E31, at a competitive disadvantage in communities. Measuring the abundance of species in each community would allow us to better determine the functional traits present in communities assuming that enzyme expression does not differ between monoculture and communities. Alternatively, it may be possible to match functional traits to community productivity by comparing the transcriptome and proteome of focal communities, although any such approach is necessarily limited by the correct annotation of functional genes and/or proteins. Rivett et al. (2016) found that the ability of species to degrade labile resources could be explained by metabolic plasticity whereas the ability to degrade more recalcitrant substrates required evolutionary adaptation. Species best adapted to utilise the accessible labile substrates may be able to dominate communities during initial growth stages but as labile substrates become depleted, species able to adapt to utilise the remaining recalcitrant substrates will become more dominant in communities. When comparing the contribution of species across each day of the BEF experiment, we found that the contribution of species did not noticeably differ throughout the seven days of growth. Paenibacillus sp. A8 significantly improved community productivity relative to the average species on each day while C. flavigena D13 made a significantly higher contribution than the average species from day 2 onwards ( Supplementary Figure 3 ). The presence of Rheinheimera sp. D14A made a significantly above average contribution to community productivity on day one of the experiment, though for the remaining 6 days the contribution of this species did not significantly differ from that of an average species. Of the remaining 9 species, contributions remained lower than or did not significantly differ from the average species throughout the 7 days. The ability of C. flavigena D13 and Paenibacillus sp. A8 to efficiently degrade both recalcitrant and labile substrates may allow them to outcompete other species before they are able to adapt to utilise recalcitrant substrates. Allowing the species used here a period of evolutionary adaptation to the wheat straw substrate may increase their ability to degrade recalcitrant substrates and alter the dominance hierarchy within these communities and is an interesting topic for future study. In conclusion, we have identified key functional traits that define the productivity of communities degrading lignocellulose. We found that the degradative abilities of communities against β-glucan, arabinoxylan and xylan were able to predict community productivity more effectively than either measures of community niche or species richness. Furthermore, we found that two species, Paenibacillus sp. A8 and C. flavigena D13, made greater than average contributions to community productivity suggesting a key role for the selection effect in driving the observed positive BEF relationship. Our results suggest that, using simple experiments, it is possible to identify the important functional traits and species that drive microbial community productivity on complex natural substrates like wheat straw, potentially simplifying efforts to predict the functioning of natural communities and the assembly of highly performing communities for biotechnological industrial applications."
} | 3,618 |
35633108 | PMC9543596 | pmc | 4,303 | {
"abstract": "Summary \n Arbuscular mycorrhizal fungi (AMF) can help mitigate plant responses to water stress, but it is unclear whether AMF do so by indirect mechanisms, direct water transport to roots, or a combination of the two. Here, we investigated if and how the AMF Rhizophagus intraradices transported water to the host plant Avena barbata , wild oat. We used two‐compartment microcosms, isotopically labeled water, and a fluorescent dye to directly track and quantify water transport by AMF across an air gap to host plants. Plants grown with AMF that had access to a physically separated compartment containing 18 O‐labeled water transpired almost twice as much as plants with AMF excluded from that compartment. Using an isotopic mixing model, we estimated that water transported by AMF across the air gap accounted for 34.6% of the water transpired by host plants. In addition, a fluorescent dye indicated that hyphae were able to transport some water via an extracytoplasmic pathway. Our study provides direct evidence that AMF can act as extensions of the root system along the soil–plant–air continuum of water movement, with plant transpiration driving water flow along hyphae outside of the hyphal cell membrane.",
"introduction": "Introduction Arbuscular mycorrhizal fungi (AMF) form symbiotic associations with 80% of surveyed land plant species and are well‐recognized for accessing and transferring nutrients to plants (Smith & Read, 2008 ). Yet, AMF also perform other essential functions, notably improving plant–water relations (Augé, 2001 ). Plants with AMF symbionts can have different rates of water movement into and out of roots, which affect tissue hydration and leaf physiology, and often lead to higher drought tolerance (Augé, 2001 ). Indeed, mycorrhizal plants typically have higher water contents than non‐mycorrhizal plants in the same environment (Faber et al ., 1991 ) and have been shown to access soil water below the permanent wilting point of non‐mycorrhizal plants (Dakessian et al ., 1986 ; Bethlenfalvay et al ., 1988 ; Franson et al ., 1991 ). The role of AMF in plant–water relations is most commonly attributed to indirect mechanisms such as enhancement of plant nutrition and osmoregulation in the host plants (Ruiz‐Lozano, 2003 ; Porcel & Ruiz‐Lozano, 2004 ; Augé et al ., 2015 ; Mo et al ., 2016 ). However, some evidence suggests AMF may directly transport water to plants (Faber et al. , 1991 ; Ruiz‐Lozano & Azcón, 1995 ; Khalvati et al ., 2005 ; Püschel et al ., 2020 ). Overall, the relative contribution of direct and indirect AMF mechanisms to the amelioration of plant–water relations remains unclear. Arbuscular mycorrhizal fung can improve plant–water relations via several indirect mechanisms (as reviewed by Augé, 2001 ). By enhancing plant nutrition, AMF not only improve plant health, which boosts plants’ resilience to environmental stresses, but also increase plants’ ability to osmoregulate via the production of nontoxic compatible solutes (Ruiz‐Lozano, 2003 ; Wu & Xia, 2006 ; Zulfiqar et al ., 2020 ). In addition, AMF can reduce drought‐induced oxidative stress in their host plants (Porcel & Ruiz‐Lozano, 2004 ; Talbi et al ., 2015 ), and help roots absorb more water by improving soil water retention properties and modulating root hydraulic conductivity (Aroca et al ., 2007 , 2009 ; Maurel et al ., 2008 ; Querejeta, 2017 ; Bitterlich et al ., 2018 ; Chen et al ., 2018 ; Quiroga et al ., 2019a , b ). The mycelia of AMF improve soil structure and soil moisture characteristics, so plants with AMF can more efficiently deplete soil water (Augé et al ., 2001 ; Querejeta, 2017 ; Bitterlich et al ., 2018 ; Chen et al ., 2018 ). Root hydraulic conductivity and symplastic flow also tend to increase in plants colonized by AMF, possibly through increased expression of root aquaporins, which allow plants to uptake more water (Aroca et al ., 2007 , 2009 ; Maurel et al ., 2008 ; Ruiz‐Lozano et al ., 2009 ; Li et al ., 2013 ; Quiroga et al ., 2019a , b ). Indeed, non‐mycorrhizal roots have decreased levels of water permeability and cell hydraulic conductivity when water‐stressed, whereas mycorrhizal roots maintain the same levels as non‐water‐stressed counterparts (Quiroga et al ., 2019a , b ). AMF may also help regulate stomatal conductance in host plants, leading to 50% higher conductance rates during moderate drought and > 100% higher rates during severe drought compared to non‐mycorrhizal plant hosts (Kaschuk et al ., 2009 ; Augé et al ., 2015 ). By consuming plant‐fixed carbon (C), which amplifies the translocation of C out of leaves and reduces its concentration in the mesophyll, AMF also stimulate stomatal opening (Jarvis & Davies, 1998 ; Kaschuk et al ., 2009 ). Relatively little is known about direct mechanisms of water transport via AMF to plants. While investigating nutrient transport, Faber et al . ( 1991 ) discovered that mycorrhizal plants with intact hyphae transpired about 20% more than mycorrhizal plants with severed hyphae. In two experiments where AMF were allowed to access a separate compartment where roots had been excluded, the water content of the soil in the no‐plant compartment declined, and it was estimated that AMF contributed 4–20% of the water transpired by their host plants (Khalvati et al ., 2005 ; Ruth et al ., 2011 ). However, a more recent study using deuterated water found that although AMF transported some water to plants, the volume carried was low compared to the transpiration demand of the plants (Püschel et al ., 2020 ). Thus, the ability of AMF to transport a significant volume of water to host plants remains ill‐defined. The physical pathways of direct water movement from AMF hyphae to roots are also unknown. Water could be transferred via hyphae by travelling along the outside of the fungal cell wall or through the cell wall matrix itself (Allen, 2007 ). The composition of cell walls can differ between fungal genera, developmental stages and conditions, and includes different proportions of polysaccharides and glycoproteins, primarily chitin and glucan (as reviewed by Bowman & Free, 2006 ; Feofilova, 2010 ). The cross‐linking of polysaccharides and glycoproteins forms a complex network (Bago et al ., 1996 , 1998 ) that creates a space and surface outside the plasma membrane where water can travel (Allen, 2007 ). We refer to this external pathway as ‘extracytoplasmic’, in contrast to an internal ‘cytoplasmic’ pathway where transport occurs inside the fungal cell membrane. Arbuscular mycorrhizal fungi hyphae are coenocytic so the cytoplasm can stream long distances in the space within the plasma membrane of the hyphae without being slowed or stopped by septa (Jany & Pawlowska, 2010 ; Purin & Morton, 2011 ). In order to resolve these knowledge gaps, we investigated if and how the AMF Rhizophagus intraradices transported water to the host plant Avena barbata , wild oat. In a glasshouse experiment, we used isotopically labeled water and a fluorescent dye to directly track and quantify water transport by AMF to plants. We specifically assessed whether AMF could access water in soil unavailable to roots and transport it across an air gap to their host plant. Finally, we estimated the relative contribution of direct and indirect AMF mechanisms to the improvement of plant–water relations.",
"discussion": "Discussion Water availability limits plant growth and is a pressing issue in the context of climate change as drought conditions are becoming more prevalent in many regions around the world (Kirkham, 2005 ). Plants have evolved multiple strategies to increase their tolerance of soil water deficit and alleviate its detrimental effects (Augé, 2001 ; Mo et al ., 2016 ), including associations with AMF. Plants with AMF symbionts tend to cope with soil water limitations more effectively, due in part to indirect mechanisms including enhanced plant nutrition, osmoregulation and root hydraulic conductivity (Ruiz‐Lozano, 2003 ; Porcel & Ruiz‐Lozano, 2004 ; Augé et al ., 2015 ; Quiroga et al ., 2019a , b ). However, it has remained unclear whether AMF also are able to directly transport water to their host plants. In our experiment, using 18 O‐labeled water, we found that direct water transport by AMF accounted for 34.6 ± 10.5% (95% CI) of the water transpired by +AMF plants. Our results indicate that AMF have the ability to bridge air gaps in soil, penetrate small pores and access water that is inaccessible to roots, attributes that could be especially important in dry soils where water films are discontinuous. We found that +AMF plants transpired an average of 1.22 ml more than −AMF plants each day, and 0.885 ml of this amount was derived from direct hyphal transport. Thus, in our experiment, direct water transport by AMF to roots accounted for over 2/3 of the extra 1.22 ml transpired by +AMF plants. We presume that indirect benefits of the AMF symbiosis accounted for the remaining 1/3 of the extra water transpired by +AMF plants each day. Notably, +AMF plants had a significantly higher P content than −AMF plants despite −AMF plants receiving more nutrient solution. Taken together, these results indicate that AMF improve plant–water relations both directly by helping plants access more water and indirectly by providing other benefits to plant health. We hypothesize that the relative contribution of direct and indirect AMF mechanisms to the improvement of plant–water relations may change depending on environmental conditions and the identity of the plant and fungal species involved in the symbiosis. In order to investigate direct AMF water transport, Püschel et al . ( 2020 ) used an experimental design similar to ours, but employed deuterated water injected into the no‐plant compartment. Puschel and coworkers then measured deuterium incorporation in plant biomass, but did not assess isotope content in transpired water. They concluded that the amount of water AMF transported to plants was low compared to the volume of water that they estimated was transpired. Our direct measurement of isotope ( 18 O) in transpired water indicated a much larger amount of water carried by AMF to their host plant. The differences between our findings could be due to the different plants and AMF taxa used, and/or to our direct measurement of isotope label in transpiration water. It seems likely that the specific amount of water AMF transport will depend on the plant–AMF species pairing as well as environmental conditions, in the same way that AMF can behave mutualistically or parasitically in their trade of nutrients for photosynthetic C (Johnson et al ., 1997 ; Klironomos, 2000 , 2003 ). We recognize that our glasshouse experiment took place in an artificial environment where conditions were crafted to encourage AMF root colonization and increase the likelihood of water transport by AMF to host plant. We designed the experiment so that if water transport by hyphae is a phenomenon that occurs, it would reveal itself in a context in which we could detect and measure it. Yet, the low nutrient and water content profiles plants and AMF experienced in our study can occur in the field. Plants and soil microbes experience intermittent or permanent conditions of low nutrient and/or water availability in numerous climatic zones and ecosystems, including in Mediterranean‐type climates and arid and semi‐arid regions around the world (Kirkham, 2005 ). In soil, fine roots and root hair can access water in macropores (> 80‐μm diameter) and mesopores (> 30‐μm diameter), but hyphae also can enter micropores down to 2‐μm diameter (Kirkham, 2005 ; Smith & Read, 2008 ; Brady & Well, 2009 ). In unsaturated soils and as previously saturated soils dry out, water is taken from larger pores first and the water films become discontinuous (Kirkham, 2005 ). Hyphae extend beyond zones that roots and root hairs inhabit, and enter smaller pores that remain water‐filled; they bridge the gap between soil particles by serving as a surface on which water can travel (Allen, 2007 ). These are the conditions we attempted to simulate with our microcosms with an air gap and a soil compartment that only hyphae could enter. Quantification of 18 O in transpired water indicated that 34.6% of water transpired by +AMF plants was transported by AMF; this proportion may be on the higher end of what occurs in the field. Roots are more efficient than hyphae at water uptake when soil water is readily available, so water transport by hyphae to host plants will probably be most critical under drought conditions when hyphae can access water unavailable to roots. Nevertheless, we observed AMF transporting a significant volume of water to their host plants, with some of that water traveling via an extracytoplasmic pathway along hyphae. Numerous studies have observed improved productivity in plants with AMF associations and have reported benefits of the AMF symbiosis in terms of increased above‐ and belowground biomass (as reviewed in Smith & Smith, 1996 ; Smith & Read, 2008 ; Diagne et al ., 2020 ). Although AMF symbionts often have a positive effect on plant growth, in our study we found no significant difference between above‐ or belowground biomass in our +AMF and −AMF treatments. +AMF plants had significantly higher P content and transpired significantly more than −AMF plants, but this did not translate to increased above‐ or belowground growth. One possibility is that conditions other than P and water availability were limiting growth in all treatments. Another possibility is that +AMF plants with higher transpiration fluxes had higher fitness in metrics other than biomass. Indeed, the cost–benefit analysis of the plant–AMF symbiosis can be complicated, and biomass alone might not be a good indicator of reproductive fitness (Johnson et al ., 1997 ). It also is worth noting that plants in both +AMF and −AMF treatments had AMF symbionts, but in the −AMF treatment, hyphae were not allowed to access the no‐plant compartment and the additional water and nutrients it contained. As described in the Materials and Methods section, −AMF plants received additional water and nutrients (to make up for what +AMF plants could obtain from the no‐plant compartment) because our goal was to grow plants of all treatments in the same conditions, except for the last few days of the experiment when the labeled water was injected into the no‐plant compartment. This also may explain why differences in above or below ground biomass between treatments were not observed. A strength of our study is the complementary combination of isotopically labeled water and a fluorescent dye. Although the isotopically labeled water allows quantification, it can move across the airgap from the no‐plant to the plant compartment by non‐AMF means (e.g. in the gas phase), resulting in a background level of plant enrichment. By contrast, the fluorescent dye provides a complementary determination because it cannot cross the airgap in the gas phase, and as shown in our experiments, was not apparent at even a background level in the −AMF treatment. These orthogonal data provides solid evidence that the AMF in our study acted as a direct means of water transport from the no‐plant compartment to the plant. In plants, roots can transport water via both apoplastic and symplastic pathways, and plants can regulate the relative contribution of each route based on environmental conditions (Bárzana et al ., 2012 ). The symplastic pathway, which tends to be favored when water availability is limited (Steudle & Peterson, 1998 ; Steudle, 2000 ), is slower because water has to flow from cell to cell via the cytoplasm, crossing plasma membranes or plasmodesmata, following an osmotic gradient (Steudle & Peterson, 1998 ; Bárzana et al ., 2012 ). The apoplastic pathway, which is favored when plants are not water‐stressed, is faster because water travels extracellularly through the cell wall and matrix and moves directly and continuously via the transpiration stream, facing little resistance (Steudle & Peterson, 1998 ; Steudle, 2000 ). Interestingly, Bárzana et al . ( 2012 ) found that plants with AMF associations have an increased apoplastic water flow in both drought and nondrought conditions, and have a greater ability to switch between water transport pathways, compared to plants with no AMF associations. They further suggest that AMF hyphae could contribute water to the apoplastic flow in roots, which is consistent with our observations. We found that the LYCH dye travelled from the no‐plant compartment to the plant compartment via hyphae. As this dye cannot cross cell membranes, it must have travelled extracellularly within the hyphal cell wall matrix and/or outside the cell wall. Indeed, we observed the dye on hyphae and even within the wall of fungal spores (Fig. 2g–i ). This finding suggests that AMF can act as extensions of the root system along the soil–plant–air continuum of water movement, with plant transpiration driving water flow along hyphae outside of the hyphal cell membrane. The soil solution generally contains nutrient ions that move to roots by diffusion and mass flow. When soil water content is very low, these nutrient supply paths are disrupted by the discontinuity of soil water films and plant nutrient deficiencies are common. Extracytoplasmic hyphal transport of water by AMF would not only supply water to roots during dry conditions, but also could enable the movement of nutrient ions to roots in dry soils. In addition, extracytoplasmic water flow along hyphae has been shown to be a way for AMF to carry phosphate‐solubilizing bacteria to areas with organic P and thereby enhance P mobilization (Jiang et al ., 2021 ). Our study provides strong evidence supporting the existence of extracytoplasmic water transport in hyphae. It is possible, however, that cytoplasmic transport also occurs in hyphae at the same time. Plants and AMF both have aquaporins at the soil interface and at the arbuscule–plant cell interface (Aroca et al ., 2007 , 2009 ; Maurel et al ., 2008 ; Li et al ., 2013 ) and, under drought conditions, the gene expression of AMF and plants aquaporins and the hydraulic conductivity and symplastic flow in roots with AMF have been shown to increase (Aroca et al ., 2007 , 2009 ; Maurel et al ., 2008 ; Ruiz‐Lozano et al ., 2009 ; Li et al ., 2013 ; Sánchez‐Romera et al ., 2017 ; Quiroga et al ., 2019a , b ). However, it has been argued that based on physical principles, notably the Hagen–Poiseuille equation, AMF hyphal diameters (2–20 μm) are too small and so their flow rates are too slow to transport significant volumes of water inside the cytoplasm (Allen, 2007 ). Evidence from the membrane‐impermeable tracer LYCH and physical principles of fluid dynamics strongly suggest that hyphae are able to transport water extracellularly to roots. That said, we cannot exclude cytoplasmic flow in hyphae, and it is possible that cytoplasmic water transport also plays an important complementary role in water transport to host plants; therefore, both pathways are illustrated in our conceptual diagram, Fig. 4 . Fig. 4 Simplified representation of water transport from soil through an arbuscular mycorrhizal fungus (AMF) hypha to a plant root. Extracytoplasmic water transport in a hypha, represented by a light blue arrow, joins apoplastic transport in a plant root, represented by a yellow arrow. Cytoplasmic transport in a hypha, represented by a dark blue arrow, joins symplastic transport in a plant root, represented by a purple arrow. 1, AMF hypha; 2, root; 3, soil water; 4, soil particles; 5, arbuscule; 6, appressorium; 7, Casparian strip. Our experimental method pairing of H 2 \n 18 O and a fluorescent tracer provides strong evidence that AMF are able to bridge air gaps in soil and bring water to plants that is inaccessible to roots. In addition, our results indicate that water can be transported to plants via an extracytoplasmic pathway in hyphae. Our findings have implications for the management of water and plant drought tolerance in the context of climate change. Plant‐AMF symbioses are key players in the maintenance of soil and plant productivity when water is limited, making them essential not only in arid and semi‐arid regions around the world, but also in areas experiencing short‐term droughts, especially as changing climatic conditions increase the occurrence of water‐limiting conditions."
} | 5,149 |
22931250 | null | s2 | 4,305 | {
"abstract": "A hallmark of the biofilm architecture is the presence of microcolonies. However, little is known about the underlying mechanisms governing microcolony formation. In the pathogen Pseudomonas aeruginosa, microcolony formation is dependent on the two-component regulator MifR, with mifR mutant biofilms exhibiting an overall thin structure lacking microcolonies, and overexpression of mifR resulting in hyper-microcolony formation. Using global transcriptomic and proteomic approaches, we demonstrate that microcolony formation is associated with stressful, oxygen-limiting but electron-rich conditions, as indicated by the activation of stress response mechanisms and anaerobic and fermentative processes, in particular pyruvate fermentation. Inactivation of genes involved in pyruvate utilization including uspK, acnA and ldhA abrogated microcolony formation in a manner similar to mifR inactivation. Moreover, depletion of pyruvate from the growth medium impaired biofilm and microcolony formation, while addition of pyruvate significantly increased microcolony formation. Addition of pyruvate to or expression of mifR in lactate dehydrogenase (ldhA) mutant biofilms did not restore microcolony formation, while addition of pyruvate partly restored microcolony formation in mifR mutant biofilms. In contrast, expression of ldhA in mifR::Mar fully restored microcolony formation by this mutant strain. Our findings indicate the fermentative utilization of pyruvate to be a microcolony-specific adaptation of the P. aeruginosa biofilm environment."
} | 386 |
35498583 | PMC9050365 | pmc | 4,306 | {
"abstract": "Electricity generation in microbial fuel cells can be restricted by a few factors, such as the effective area of the anode for biofilm attachment, diffusion limitation of substrates and internal resistance. In this paper, a suspended anode (carbon-based felt granule)-type microbial fuel cell was developed to make full use of the volume of the anode chamber and provide a larger surface area of the anode for the growth of exoelectrogenic bacteria. The current collector was rotated in the anodic chamber to contact with the suspended granules intermittently and achieve better mixing. The open-circuit voltage reached steady state at around 0.83 V. The maximum power density obtained from each scenario increased steadily with the increase in mixing rate. The internal resistance decreased when the rotational rate and the content of the carbon granules were increased. The maximum power density reached 951 ± 14 mW m −3 with a corresponding minimum internal resistance of 162.9 ± 3.5 Ω when the mass of carbon granules was 50 g and the rotational rate was 300 rpm. The suspended microbes made negligible contribution to the power density. The microbial fuel cell with a higher content of carbon granules had lower coulombic efficiency and lower relative abundance of exoelectrogenic bacteria.",
"conclusion": "4. Conclusions A novel configuration of microbial fuel cell was constructed for enhanced electricity generation from wastewater. The main conclusions are: • A suspended carbon felt granule anode in MFCs was used to overcome the diffusion limitations of the substrates and increased the effective area of the biofilms. • The performance of the MFCs with the carbon granules was better than that of the control without carbon granules. The suspended bacteria had a negligible effect on the electricity generation. • The maximum power density reached 951 ± 14 mW m −3 with a corresponding minimum internal resistance of 162.9 ± 3.5 Ω when the mass of carbon granules was 50 g and the rotational rate was 300 rpm. • The CE of MFC was lower with the higher content of carbon granules because the relative abundance of exoelectrogenic bacteria decreased and contact between the anode and current collector occurred less frequently.",
"introduction": "1. Introduction Typical domestic wastewater contains about 1.23 kW h m −3 of energy, which is mainly in the form of biodegradable organics. Microbial fuel cells (MFCs) have received much attention because of their potential applications in wastewater treatment and energy generation. 1–5 An exoelectrogen is an essential part of MFCs and has many special properties, including the metabolism of organics and the generation of electrons. 6 The electrons generated by exoelectrogenic bacteria spontaneously transfer to the anode, which generates a low current flow in an external circuit in a clean and mild condition. However, the practical applications of MFCs are restricted by many issues including the low voltage output and the difficulty of scaling up, which are discussed in more detail below. 7 Anode performance is one of the key factors affecting the MFC performance. It has been investigated extensively in many previous studies. 8 The internal resistance of MFCs, effective area of biofilms attached to the anode surface, and mass transfer rate of substrates play important roles in anode performance. However, high internal resistance increases the loss of electrical energy during transfer, the effective area of biofilms can directly affect the electricity generation, and the low mass transfer rate causes concentration polarization, especially in large-scale reactors. 9–11 Special configurations of MFCs have been studied to solve these issues. For example, membrane electrode assembly-type MFCs were applied to reduce the internal resistance, 12 and fluidized electrodes or packed-bed electrodes acting as three-dimensional electrodes were used to overcome the diffusion limitations of substrates and enhance the effective area of biofilms in large reactors. 13–18 A fluidized electrode-type MFC was analyzed to assess the capacitive characteristics of granular activated carbon (GAC) particles. 19 Electrons can be stored in the GAC particles in the form of an electric double layer, and a continuous current can be generated with intermittent contact between the GAC particles and the current collector. 20 In these MFCs, the electrode acts as a filter and can be easily clogged by suspended solids or biofilms, which can block the mass-transfer channel frequently during the operation. These issues need to be solved before building a large reactor using these electrodes. A carbon-based flow-through composite anode configuration was constructed as carbon-based particles have been proved to enhance the power generation in microbial fuel cells. 21 In our study, a novel configuration of MFC reactor was developed, which indicated the untapped potential of larger MFCs. Specifically, charges were stored in carbon-based felt granules suspended in the anodic chamber, while the current flowed with the intermittent contact between the carbon-based felt granules and a rotational current collector. In order to solve the problems of flow blockage in previous fluidized electrode-type MFCs, the anode was suspended in a large reactor rather than flowing in a channel. Compared to the traditional 3D porous anode material with magnetic stirring, the current collector could contact with the anode better but did not assemble and block. The suspended anode-typed MFC was designed to make full use of the reactor volume, to overcome diffusion limitations, and to reduce the internal resistance. Furthermore, we investigated the bio-capacitor characterization of the electrodes and the performance of the MFCs, including the maximum power density, internal resistance, and coulombic efficiency (CE). In addition, the community structure of the microbes was investigated.",
"discussion": "3. Results and discussion 3.1 MFC start-up The voltage output during the acclimatization is shown in Fig. 2 . The initial voltage was around 0.10 V and fluctuated in the first 30 h, then steadily increased to 0.25 V within 20 h and maintained this level for 100 h. After 150 h, the voltage output was increased rapidly, which means the exoelectrogens were in a logarithmic growth phase. This was accompanied by a rapid increase in the voltage to around 0.63 V after 200 h. Simultaneously, the open-circuit voltage reached 0.83 V. The start-up time was longer than that of the traditional two-chamber MFCs because of the scouring from stirring by the current collector, which meant it was hard for the biofilm to attach to the anode. Fig. 2 Temporal profiles of the voltage output during the acclimatization. 3.2 Biocapacitor characterization of the electrodes The anodic granules of the MFCs were characterized by SEM-EDS, and the results are shown in Fig. 3 and Table 1 . As shown in Fig. 3a and b , the biofilm was attached to the carbon felt. From the morphology, it was indicated that microorganisms, like bacillus and coccus, were attached to the granules. The EDS results show the element ratio difference from the carbon felt before and after acclimatization. The percentages of N, O, and other trace element in the microbes were significantly higher than in the control, which indicated that the biofilm was attached. Fig. 3 Images obtained by SEM of the anodic granules after acclimatization ((A) at 3000×, (B) at 10000×) and the granules without microorganisms ( i.e. , the control, (C) at 10000×). Relative abundance of the elements of the anodic granules after acclimatization and the control without microorganisms (atomic percent) Element Anodic granules (%) Control group (%) C 69.88 91.33 N 8.83 03.84 O 15.66 04.82 Na 0.29 — Mg 0.32 — P 2.22 — S 0.33 — K 0.16 — Ca 0.75 — Fe 1.57 — 3.3 Voltage output from the on–off experiment The voltage output from the on–off experiment is shown in Fig. 4 . When the agitator was off, the voltage output immediately decreased from 0.45 V to about 0.15 V. This could be because without mixing only a few bioanodes were contacted with the current collector and the internal resistance increased. As a consequence, electrons generated by the biofilms could not transfer to the current collector, and were stored in the granules of the carbon felt. Borsje et al. found that the charge could be stored in the electric double layer of single carbon granules. 19 When the agitator was restarted, a peak value with around 50 mV variations appeared in the voltage output curve, which verified the capacitance characteristic of the carbon felt granules and that the current was not only produced at the moment of contact. A similar finding was reported by Liang et al. , who reported that the power density of MFCs instantly increased by applying transient-state regulation ( i.e. , alternating open-circuit and closed-circuit) because of the anode capacitance. 24 Carbon-based capacitive anodes were used in the MFCs in some pervious studies to enhance the performance of the MFCs, as the capacitance of carbon-based granules mainly consists of electrode double-layer capacitance. 18 Fig. 4 Temporal profiles of the voltage output from the on–off experiment, and the changing portions of 0.2 min intervals after switching on are enlarged (top). 3.4 Electricity generation of the microbial fuel cells 3.4.1 Power density curves The effects of the rotational rate and the content of carbon granules on the power density were investigated, as shown in Fig. 5 . For a typical power density curve at a given rotational rate ( e.g. , Fig. 5a ), the power density initially increased with the growth of the current. After reaching a maximum value, the power density gradually decreased with higher current. The maximum power densities obtained at different rotational rates were compared ( Fig. 5g ) and the results showed that the maximum power density increased with the increase in the rotational rate and the current at maximum power density also increased. This was consistent with the results in the tests with 20, 30, 40, and 50 g of suspended anodes ( Fig. 5b–e ). On the one hand, when the rotational rate increased, the anolyte was mixed more homogeneously. On the other hand, the higher rotational rate allowed for a better contact between the carbon granules and the current collector, and thus electrons could transfer more quickly. However, the further increases in the mixing rate to 350 and 400 rpm did not result in an obvious increase in the maximum power density ( Fig. 5f ). This could be because the higher rotational rate was accompanied by higher hydraulic shear force, which made it difficult for the biofilms to attach to the anode. This was also supported by a previous study that reported that the performance of rotating disk MFCs became poor when the rotational rate was higher than 400 rpm. 25 Fig. 5 Power density curves obtained at different rotational rates with 10 g (a), 20 g (b), 30 g (c), 40 g (d), and 50 g (e) of suspended anode and at a high rotational rate with 50 g of suspended anode (f). The maximum power density of each condition is shown (g). There was a positive correlation between the power density and the amount of carbon granules. The power density reached 951 ± 14 mW m −3 with 50 g of the suspended anode and 300 rpm, while the current density was 2.18 A m −3 . The current density of 0 rpm with 50 g of anode was 0.58A m −3 . When the rotating rate was 0 rpm, the anode granules were deposited at the bottom of anodic chamber and were in continuous contact with the current collector. It could be found that the suspension plays an important role in the performance. The I max values from similar studies are compared in Table 2 , but it is to be noted that these researchers used different methods to make the capacitive anode contact with current collector intermittent. It was found that the gas lift, liquid pump, and stirring rate all increased the performance of MFCs. A larger amount of carbon granules could provide more sites for biofilm attachment and hence could increase the exoelectrogen concentration, which is beneficial for electron transfer. In addition, having more carbon granules provides a larger capacitance, which allows the electrons generated by exoelectrogens to be stored more effectively. However, when higher than 50 g, agitation in the MFCs became hard and some granules could not have continuous contact with the current. Performance overview of various fluidized capacitive anode MFCs Reactor type \n V \n total (mL) \n I \n max (A m −3 at P max ) Reference Suspended carbon felt granules anode MFC with stirring 1000 2.18 This work MFC-fluidized bed membrane bioelectrochemical reactor with liquid pump 1700 (700 in MFC) 20 (13 in MFC) \n 26 \n Membrane-free fluidized-bed MFC with pump 1000 0.8 \n 27 \n Fluidized capacitive bioanode (GAC) MFC with gas lift 2102 0.7 \n 14 \n Fluidized GAC anode with stirring 7 260 \n 20 \n To make a comparison between power generation and stirring, the stirring power was calculated in theory by the following equation: 28 P = N p ρn 3 d 5 where N p is the power number, ρ is the density of solution, n is the rotating rate, and d is the diameter of the agitator. After calculating, the power was 4.19 W with 50 g of the suspended anode and 300 rpm when the power density reached the maximum. In fact, the power generated by the MFC could not offset the power of stirring, while the maximum total power of this reactor was 0.93 W. This study focused on the effect of stirring on the performance of this anode. Furthermore, in further applications, low-grade energy, like wind energy, could be used to offset the power consumption. 3.4.2 Internal resistance The effect of the rotational rate and the amount of carbon granules on the polarization curves are investigated in this chapter, as shown in Fig. 6 . The cathode potentials in this reactor did not significantly change when the condition of the anode changed ( Fig. 6f ). The polarization was mainly influenced by the anode since the polarization was affected by activation polarization, ohmic polarization, and concentration polarization at low, middle, and high levels of current density, respectively. Fig. 6 Polarization curves obtained at different rotational rates and contents of suspended anode ((a) 10 g, (b): 20 g, (c) 30 g, (d) 40 g, (e) 50 g). (f) The cathode potential (solid line) and anode potential (dotted line) ( vs. Ag/AgCl) in a 50 g batch. (g) The internal resistance of each condition is shown. In the low current density range, the activation polarization affected the polarization predominantly. We can find that it had no discernible effects on the polarization in this experiment. That was attributed to the high temperature of 28 °C and high roughness and specific surface of the electrodes, which were proven to decrease the activation polarization in a previous study. 7 In the middle current density range, the ohmic polarization played a dominant role. The internal resistance between the suspended anode and cathode was used to describe the resistance between the suspended anode and current collector since the resistance between the current collector and cathode was stable. The internal resistance decreased when the rotational rate and the content of carbon granules increased ( Fig. 6f ). In the MFCs ( Fig. 6a–e ), the polarization was intense when the rotational rate was zero, which could be related to the high internal resistance. Without mixing, the current collector could not contact with the suspended anode efficiently, and thus the internal resistance was large. The minimum internal resistance was 162.9 ± 3.5 Ω, which was achieved with 300 rpm and 50 g of suspended anode. The decrease in internal resistance was because the higher rotational rate enabled a better contact between the carbon granules and the current collector, and thus quicker electron transfer. The higher the content of carbon granules was, the higher the electrical conductivity. However, the current became higher not only because of the low internal resistance, but also because of the greater contact and higher electron flow. It is more accurate to describe this parameter as the apparent internal resistance. There is little discussion of the internal resistance in fluidized anode-type MFCs because the actual internal resistance is hard to identify due to the complex contact. The minimum apparent internal resistance of a membrane-free fluidized-bed MFC that used a pump to maintain intermittent contact was 902 Ω, which had a similar volume to our reactor. 27 In the high current density range, some batches, e.g. , 0 rpm in both batches and 50 rpm in the 30 g (c), 40 g (d), and 50 g (e) batches of carbon granule, were affected by concentration polarization, especially at a low rotational rate and high content of carbon granules. When the rotational rate was low, the diffusion rate in the anolyte was lower. High shear forces led to faster diffusion in biofilms, and thus no pronounced concentration polarization occurred. 29 A high content of carbon granules caused concentration polarization because faster agitation was needed to achieve a homogeneous distribution of the granules. 3.4.3 Effect of the suspended bacteria on the electricity generation To investigate the contribution of the suspended bacteria (not attached to the anode) on the electricity generation, the suspended granules were taken out from the MFC under a nitrogen environment. The maximum power density in the batch tests without the suspended granules at 200 rpm was 43.24 ± 11.94 mW m −3 , which was far below the maximum power density in the batch tests with the suspended anode and when the internal resistance was 15 725 ± 10 943 Ω. This indicated that the suspended bacteria had a negligible effect on the electricity generation. The following reasons can account for this phenomenon: (1) the bacteria in suspension had no place to attach and were scoured by agitation, making it difficult to form stable biofilms; (2) the carbon granules acted as capacitors in the MFCs. Electrons generated by microbes had no place to be stored without the suspended granules and thus the electrons were hard to be transferred to the current collector. 3.5 Coulombic efficiency The coulombic efficiency was calculated to investigate the performance of the suspended anode-type MFCs. As shown in Fig. 7 , the CE of the MFCs was 20.6 ± 2.2% in the test with 10 g of suspended carbon granules and decreased gradually with the increase in the amount of carbon granules. The CE was relatively low in this experiment and could be affected by many operational parameters. 30 The intermittent contact between the anode and the current collector could be one of the key reasons for the low CE given the higher biodegradation rate of non-exoelectrogens. 31 The electron transfer could only happen when the anode is in contact with the current collector, otherwise the electrons generated by the MFCs are temporarily stored in the carbon granule anode, but there are some energy losses too during the charge period and storage. Volume also plays an important role in CE since a large volume involves a complex microbial community. Considering the large concentration of microbes, especially non-exoelectrogens, in these experiments, the electron-transfer resistance was low under the suspended set-up for the anode. Suspending the anode causes another issue in MFCs: the direct contact between the suspended anode and the current collector is reduced, which could affect the community structure of the microbes in MFCs. In general, the CE of MFCs with mixed strains was lower than that with pure cultures because the energy was partially consumed by bacteria that could not generate electrons. 32,33 Fig. 7 Coulombic efficiency tests with different amounts of suspended anode. The coulombic efficiency was determined in a full batch experiment with a rotational rate of 200 rpm. 3.6 Microbial communities In order to investigate the decrease of CE, the community structure of the microbes was determined by 16S rDNA analysis. The taxonomic tree of 100 kinds of microbes with the highest relative abundance in MFCs is shown in Fig. 8 , where A1 is the sample of inoculated sludge, B1 is from sampling when the granules content was 20 g L −1 in MFC-1, while C1 was 40 g L −1 . Fig. 8 Taxonomic tree of 100 kinds of microbes with the highest relative abundances and distribution bar plot of the inoculated sludge (A1), 20 g L −1 granules in MFC-1 (B1) and 40 g L −1 granules in MFC-1 (C1). Top 20 microbes with the highest relative abundances are marked by asterisk and different phyla are indicated by different colors. The outer ring is a thermodynamic chart in which the samples are marked by different colors and the samples with a higher relative abundance are shown in a darker color. The abundances of the microbiological species or strains are shown in the outer ring of the taxonomic tree. There were more species in the inoculated sludge ( i.e. , A1) than in the granules of the MFCs ( i.e. , B1 and C1). This indicated that the microbes in samples B1 and C1 were screened by acclimatization and hence had lower diversity. There were many species of electrogenesis bacteria, such as Ottowia , Malikia , Tistrella , Pseudomonas , and Fontibacter , which was similar to in previous studies. 34–37 Tistrella in Rhodospirillales was widely found in the anodic chambers of the MFCs and was in high abundance. In addition, the abundance of Tistrella increased when the content of granules increased. As the distribution bar plot shows, the kinds of microbial species decreased substantially when the content of granules increased. The dominant strains were screened by the reaction. However, there was an increase in many non-electricity-production species with the increase in granules, e.g. , Proteiniphilum , Longilinea , Acholeplasma , and so on. An entire circuit contributed to the accumulation of exoelectrogens in MFCs. The total number of granules contacted with the current collector, which is equal to an entire circuit, increased with increasing the number of granules, and thus the performance and exoelectrogens increased. The intermittent contact between the anode and the current collector was unfavorable for the growth of the electrogenesis bacteria and hence its relative abundance. Thus, the exoelectrogens in the biofilms attached to the anode found it difficult to be the dominant bacteria. This also caused the lower CE with the higher content of carbon granules."
} | 5,662 |
25376371 | PMC4223642 | pmc | 4,307 | {
"abstract": "Although microbes directly accepting electrons from a cathode have been applied for CO 2 reduction to produce multicarbon-compounds, a high electron demand and low product concentration are critical limitations. Alternatively, the utilization of electrons as a co-reducing power during fermentation has been attempted, but there must be exogenous mediators due to the lack of an electroactive heterotroph. Here, we show that Clostridium pasteurianum DSM 525 simultaneously utilizes both cathode and substrate as electron donors through direct electron transfer. In a cathode compartment poised at +0.045 V vs. SHE, a metabolic shift in C. pasteurianum occurs toward NADH-consuming metabolite production such as butanol from glucose (20% shift in terms of NADH consumption) and 1,3-propandiol from glycerol (21% shift in terms of NADH consumption). Notably, a small amount of electron uptake significantly induces NADH-consuming pathways over the stoichiometric contribution of the electrons as reducing equivalents. Our results demonstrate a previously unknown electroactivity and metabolic shift in the biochemical-producing heterotroph, opening up the possibility of efficient and enhanced production of electron-dense metabolites using electricity.",
"discussion": "Discussion Here, we have shown a direct electron transfer from a cathode to C. pasteurianum and an electricity-derived reducing power induced NADH-consuming metabolic pathway in glucose and glycerol fermentation. This is the first report to present a wild type of pure culture ( C. pasteurianum ) that not only accepts electrons directly from the cathode even in the presence of electron-rich glucose and glycerol (heterotrophic growth), but that also produces a higher amount of net NADH-consuming compounds from substrates, beyond that which supplemented electricity can theoretically achieve. The immediate current recovery in the 2 nd batch of BES ( Supplementary Fig. 2 ) supports direct electron transfer. In addition, considering that cell adhesion to insoluble electron acceptors (ferric iron oxide) is not required for Shewanella in the presence of soluble electron shuttles 24 , the biofilm of C. pasteurianum on the cathode also supports direct electron transfer from the cathode to the cells. The mechanism of direct electron transfer in C. pasteurianum is as yet unknown. Gram-positive bacteria lack an outer membrane and it is structurally hard for them to carry electrons, whereas Gram-negative bacteria like Geobacteraceae have complicated electron transfer chains 27 . Nonetheless, it was also reported that Gram-positive Clostridium strains like C. ljungdahlii and C. aceticum were able to consume electrons from the cathode to produce small amounts of multicarbon-compounds using CO 2 as the sole carbon source. Our studies suggest that the surface interaction between electrode and microorganism could affect extracellular electron transfer and it may be future study to elucidate the mechanism of the direct electron transfer of C. pasteurianum . Electrochemically introduced reducing power appeared to significantly affect intracellular redox conditions ( Fig. 5 ) and consequently stimulate the production of net NADH-consuming metabolites like butanol in glucose fermentation and, more notably, 1,3-PD in glycerol fermentation. In early studies, C. acetobutylicum increased butanol production up to 26% using electrically reduced methyl viologen as an electron mediator 17 . However, C. pasteurianum was able to directly accept electrons from the cathode without any mediators and butanol production increased more than 2-fold compared to the case of glucose fermentation without electricity. The metabolic shift to the net NADH-consuming reaction in glycerol fermentation was clearly current consumption-dependent ( Fig. 4c ) and more significant than it was in glucose fermentation. This result is likely due to the presence of a simple and separate reduction pathway in the glycerol fermentation process (i.e., 1,3-PD production); this pathway likely responds to BES conditions actively by consuming reducing power. Lactate production as an intermediate from glucose electrofermentation also supports the idea of the importance of a relative simple pathway to efficiently consume NADH in BES. Recently, the utilization of electrons from a cathode as a reducing equivalent in biofuel production has been studied with carbon dioxide as a carbon source; this process mimics natural photosynthesis in plants 28 29 30 . However, a much higher electron uptake is required for CO 2 reduction compared to that necessary for microbial biochemical production from general carbon sources such as sugars and glycerol. This high demand for electrons and, as mentioned in the case of glucose electrofermentation, the lack of a relatively simple and active reduction pathway, might limit the production of electrofuel from CO 2 . Alternatively, co-utilizing organic substrates and electrons from a cathode as a reducing power through electrofermentation has been suggested as a way to overcome the limitations of CO 2 reduction using electrons 2 3 ; however, there have been no reports on single pure microorganism that can perform electrofermentation. It is of interest that C. pasteurianum actively attached to the cathode to uptake electrons even though there was abundant amount of electron-rich glucose. In the case of microbial fuel cell (which is an electron-producing process), G. metallireducens only donated electrons to an anode in the absence of soluble electron acceptors 31 32 . On the contrary, C. pasteurianum used dual electron donors, i.e., glucose or glycerol (soluble), and the cathode (insoluble). This characteristic of C. pasteurianum of simultaneously utilizing a reducing equivalent derived from the cathode and substrates will be very advantageous to facilitate electrofermentation for biofuel and biochemical production. It is generally known that electroactive microorganisms do not possess biochemical production pathways and that biochemical-producing heterotrophic microorganisms are not electroactive 33 . Therefore, this unique characteristic of C. pasteurianum (an electroactive biochemical-producing heterotrophic microorganism) is extraordinary. It should be noted that the electroactivity of C. pasteurianum had not been discovered until we demonstrated it here, although C. pasteurianum is a well-known Gram-positive bacterium that has been studied for a long time by many researchers 14 15 16 26 34 . This indicates that, in addition to isolating novel electroactive microorganisms, exploring known microorganisms might yield unexpected opportunities to discover previously unknown electrochemical activity. The results presented here could be the next step toward unveiling a natural bio-electrochemical versatility, including such a versatility in heterotrophs. Further study on the mechanism of direct electron transfer and elucidation of the metabolic regulation in BES will provide valuable information to open the next door toward efficient biofuel and biochemical production using sustainable electricity."
} | 1,789 |
35713370 | PMC9542400 | pmc | 4,308 | {
"abstract": "Abstract The assembly of microbial communities through sequential invasions of microbial species is challenging to study experimentally. Here, I used genome‐scale metabolic models of multiple species to model community assembly. Each such model represents all known biochemical reactions that a species uses to build biomass from nutrients in the environment. Species interactions in such models emerge from first biochemical principles, either through competition for environmental nutrients, or through cross‐feeding on metabolic by‐products excreted by resident species. I used these models to study 250 community assembly sequences. In each such sequence, a community changes through successive species invasions. During the 250 assembly sequences, communities become more species‐rich and invasion‐resistant. Resistance against both constructive and destructive invasions – those that entail species extinction – is associated with high community productivity, high biomass, and low concentrations of unused carbon. Competition for nutrients outweighs the influence of cross‐feeding on the growth rate of individual species. In a community assembly network of all communities that arise during the 250 assembly sequences, some communities occur more often than expected by chance. These include invasion resistant “attractor” communities with high biomass that arise late in community assembly and persist preferentially because of their invasion resistance. Genome‐scale metabolic models can reveal generic properties of microbial communities that are independent of the resident species and the environment.",
"introduction": "1 INTRODUCTION In his 1958 popular science book “The ecology of invasions by animals and plants” British ecologist Charles S. Elton set the foundation for invasion biology as an independent field of science (Elton, 1958 ). In this short book, Elton proposed a number of hypotheses that became the subject of intensive research in subsequent decades. Perhaps the most prominent of them is the “biotic resistance” or “diversity‐stability” hypothesis. It posits that more species‐rich communities are less susceptible to invasion by non‐native species. Pertinent empirical evidence accumulated in the decades since Elton's book was published. Some such evidence, including from grasslands, subtropical wetlands, riparian and agricultural ecosystems, supports the hypothesis (Boughton et al., 2011 ; Brown & Peet, 2003 ; Dukes, 2002 ; Kennedy et al., 2002 ; Naeem et al., 2000 ; Peltzer & MacLeod, 2014 ; Peng et al., 2019 ; Tilman, 1997 ). Other evidence argues that species‐rich communities are not less but more susceptible to invasion (Altieri et al., 2010 ; Brown & Peet, 2003 ; Fridley et al., 2004 ; Levine, 2000 ; Lonsdale, 1999 ; Stohlgren et al., 2003 ; Stohlgren, Barnett, et al., 2006 ; Stohlgren, Jarnevich, et al., 2006 ). To resolve this so‐called “invasion paradox”, some studies suggest that diversity helps reduce invasibility at small spatial scales, but increases it at large spatial scales (Brown & Peet, 2003 ; Fridley et al., 2004 , 2007 ; Hui & Richardson, 2017 ; Richardson & Pysek, 2006 ). Here, I studied how species invasions help assemble microbial communities, where species invasions have been less well‐studied than in macroscopic organisms. Several reasons are responsible for the relative neglect of invasion dynamics in microbial communities. First, such dynamics are difficult to observe experimentally through standard microbial ecology techniques, such as 16S rDNA sequencing, or bulk measurements of metabolite concentrations (Aguirre de Carcer, 2020 ; Nemergut et al., 2013 ). In addition, microbes can disperse globally and show complex mosaic phylogenies shaped by horizontal gene transfer (Nemergut et al., 2013 ; Soucy et al., 2015 ), which can blur the distinction between native and invading species. Moreover, even though microbial community ecology is a very active field, it has developed independently from invasion biology (Latombe et al., 2021 ). It asks different questions, such as about principles of species coexistence, convergent assembly trajectories, historical contingency (priority effects), and multistable equilibria (Bittleston et al., 2020 ; Blasche et al., 2021 ; Datta et al., 2016 ; Estrela et al., 2021 ; Friedman et al., 2017 ; Furman et al., 2020 ; Goldford et al., 2018 ; Gralka et al., 2020 ; Hoek et al., 2016 ; Jami et al., 2013 ; Kehe et al., 2021 ; Kelsic et al., 2015 ; Ratzke et al., 2020 ; Sanchez‐Gorostiaga et al., 2019 ; Wolfe et al., 2014 ). Because it is challenging to observe species invasions and community assembly in microbes, mathematical and computational models remain important to identify general assembly principles. Such models have a long tradition in studies of community assembly (Drake, 1990 ; Drossel et al., 2001 ; Hui et al., 2021 ; Law & Morton, 1996 ; McKane, 2004 ; Minoarivelo & Hui, 2016 , 2018 ; Post & Pimm, 1983 ). Pertinent models fall into broad classes, such as (generalized) Lotka‐Volterra models (Baiser et al., 2010 ; Gonze et al., 2018 ; Hofbauer & Sigmund, 1988 ; Kuntal et al., 2019 ; Minoarivelo & Hui, 2018 ; Mittelbach & McGill, 2019 ; Servan et al., 2018 ), consumer resource models (Cuddington & Hastings, 2016 ; Goldford et al., 2018 ; Marsland et al., 2019 , 2020 ; Mittelbach & McGill, 2019 ; Valdovinos et al., 2018 ), and adaptive dynamics models (Hui et al., 2021 ). Existing models have several limitations when applied to microbial community assembly. First, with notable exceptions (Brunner & Chia, 2019 ; Goldford et al., 2018 ; Marsland et al., 2019 ), they are designed to understand predator–prey or plant‐pollinator communities, but not microbial communities, which are governed by trophic interactions mediated by the abiotic environment (Bittleston et al., 2020 ). For example, microbial species frequently feed on small molecules that other microbial species excrete as by‐products of their metabolism. Second, existing models usually assume a constant environment that is not changed by community members. In contrast, microbial species can secrete multiple metabolic by‐products that change the chemical environment and thus construct new ecological niches for other microbes (Basan et al., 2015 ; Bittleston et al., 2020 ; Pacheco et al., 2019 ; San Roman & Wagner, 2018 ). Third, existing models generally allow only for pairwise interactions among species, whereas higher order interactions can be important (Grilli et al., 2017 ; Mickalide & Kuehn, 2019 ; Sanchez‐Gorostiaga et al., 2019 ). Finally and perhaps most importantly, existing models make specific and often ad hoc assumptions about which species interact. They are not suited to allow species interactions to emerge from first biological principles. These limitations can be alleviated by the modelling framework used here (Khandelwal et al., 2013 ; Klitgord & Segre, 2010 ; Levy & Borenstein, 2013 ; Libby et al., 2019 ; Pacheco et al., 2019 ; San Roman & Wagner, 2021 ; Stolyar et al., 2007 ). It is built on genome‐scale metabolic models of microbial species (Amalric & Dehaene, 2019 ). Each such model represents all known metabolic biochemical reactions that take place in a given species. Its establishment requires a combination of genome sequence data and extensive biochemical information. Manual curation, together with increasingly advanced automatic model construction techniques have made such models available for hundreds of species (Devoid et al., 2013 ; Gu et al., 2019 ; Magnúsdóttir et al., 2017 ). With the computational method of flux balance analysis (FBA, see Methods ), a genome‐scale model can be used to predict the instantaneous biomass growth rate of a species in a given chemical environment (Orth et al., 2010 ), as well as the amounts of metabolic by‐products that a microbe excretes into the environment. The predictions of this method have been experimentally validated in multiple experiments (Edwards et al., 2001 ; Feist et al., 2007 ; Segre et al., 2002 ; Varma & Palsson, 1994 ). Genome‐scale metabolic modelling and FBA allow species interactions to emerge from basic biochemical principles. First, they allow species to compete by consuming the same nutrients available in the environment. Second, they can predict the identity and amount of metabolic by‐products that a species excretes and that can help other species grow. In doing so, they also allow a community's species to change their environment and create new ecological niches. Finally, they also allow higher order trophic species interactions to emerge naturally, because all species in a community have access to the same chemical environment. Here, I use these methods to study the assembly of two different kinds of communities in silico. The first harbour a set of 100 human gut microbial species, and are assembled in an anaerobic chemical environment similar to a Western diet. The second harbour a subset of 100 random viable metabolisms (“species”), each of which contains a random complement of chemical reactions sampled from a known “universe” of such reactions realized in the biosphere, under the sole constraint that each such metabolism must be at least viable on glucose as the sole carbon source. I assemble the latter communities in an aerobic environment that contains only glucose as a carbon source. The two classes of metabolic systems are dramatically different, both in terms of their origins (biological/synthetic), and their environments (complex/simple, anaerobic/aerobic). Thus, any properties they share are probably generic properties of microbial communities assembled through trophic interactions, and not just peculiarities of a specific set of microbial species. For each type of community, I model 250 community assembly sequences. Each sequence starts out with a single species and proceeds through the sequential invasion of randomly chosen species from the species pool. I show that over time, both types of communities become more species‐rich, show higher biomass, and become more invasion resistant, largely because they become occupied by superior competitors. Some communities recur more often than expected by chance alone. They represent either recurrent “attractors” of community assembly or transient sources of new communities.",
"discussion": "4 DISCUSSION For both gut microbes and random viable “species”, I find that communities become more invasion resistant, especially during community assembly. These observations are consistent with earlier modelling work (Drake, 1990 ; Law & Morton, 1996 ; Post & Pimm, 1983 ). However, in contrast to the Lotka‐Volterra models of earlier work, species interactions in my genome‐scale metabolic modelling framework emerge from first biochemical principles. My observations thus suggest that increased invasion resistance is a generic feature of microbial communities in which trophic interactions mediated by environmental nutrients are prevalent. The models I use also help explain why invasion resistance increases. At the center of this explanation is a community's biomass productivity. It helps explain resistance against both constructive and destructive invasions, but for different reasons. Productive communities harbour few opportunities for constructive invasions, because they harbour few resources that are underused by resident species, which an invader can exploit without interfering with the persistence of a resident. Conversely, productive communities also harbour few opportunities for destructive invasions, because one or more of their resident species convert nutrients efficiently into biomass, such that few invaders can outcompete them. In the communities I study, most pairwise species interactions are competitive (>80%) rather than facilitative (<17%), that is, a species' growth rate is reduced rather than increased by the presence of another species. This is not surprising, because all species can survive on the externally provided nutrients, and are thus competing for these nutrients. It holds even though 70% or more of species pairs cross‐feed. The prevalence of competition together with extensive cross‐feeding imply that any one species grows its biomass to a greater extent from externally supplied nutrients than from the excretions of other species. It does not mean, however, that cross‐feeding is unimportant for the assembly of species‐rich communities. This becomes clear from random viable metabolisms, whose environments provide only a single carbon‐source niche in the form of glucose. Without cross‐feeding, competitive exclusion dictates that only single species communities are possible in this environment, because a superior competitor would always replace the resident species (MacArthur & Levins, 1964 ; Stewart & Levin, 1973 ). In contrast, I find that communities with more than four species are formed even in this simple environment. My predictions are consistent with recent experiments that cultured microbial communities from various natural habitats in a synthetic environment with only a single carbon source. Due to extensive cross‐feeding on metabolic by‐products, this simple environment could support communities of up to 22 species (Estrela et al., 2021 ; Goldford et al., 2018 ). I emphasize that this prevalence of competition over facilitation may depend on the environment and on the invading species. For example, in nutrient‐poor environments, or in environments where many species can grow only on metabolic excretions rather than externally supplied nutrients, facilitation may be more important than competition (Marsland et al., 2019 ). Destructive invasions become more frequent during community assembly, but they never account for more than 50% of all invasions (Figure S2 ). This observation underlines that the external environment and metabolic excretions create ample free niche space to accommodate invaders without leading to the extinction of resident species. Constructive invasions, however, generally lead to smaller increases in community productivity than destructive invasions. Genome‐scale models can provide a new perspective on long‐standing debates in invasion biology. One of these debates revolves around Elton's biotic resistance hypothesis. Evidence for this hypothesis is mixed, possibly because diversity may prevent invasions at small spatial scales, while facilitating invasions at large spatial scales (Altieri et al., 2010 ; Fridley et al., 2007 ; Jeschke et al., 2018 ). The present study system can help validate the hypothesis without space as a complicating factor, because it models a well‐mixed environment. And it shows that more species‐rich communities tend to be invasion resistant, because their efficient resource use leaves fewer opportunities for a new invader. Elton argued that high species‐richness entails high invasion resistance. Part of the reason is that species‐rich communities may be associated with few ecological niches that are free for an invader to exploit. However, this association need not be universal. Many organisms create new ecological niches through “niche construction” or “ecosystem engineering” (Jones et al., 1994 ; Odling‐Smee et al., 2003 ). Bacteria do so by excreting by‐product metabolites that other bacteria can exploit through cross‐feeding. Such niche construction can help sustain large microbial communities even in chemically minimal environments (Estrela et al., 2021 ; Goldford et al., 2018 ; San Roman & Wagner, 2021 ). Thus, species‐rich communities might harbour more free niches, and these niches might also facilitate species invasions, contradicting Elton's reasoning. Indeed, my simulations show that species‐rich communities excrete a greater number of metabolites into the environment (Figure 3 ). The reason why they are nonetheless more invasion resistant is that competition for externally supplied nutrients prevails over facilitation via niche construction. In the community assembly network I study, nodes are equilibrium communities. Two communities A and B are connected by a directed edge if a successful invasion of community A creates community B. Subsets of 100 species can form many more communities (≈10 30 ) than one can explore. Even if only a small fraction of them form equilibrium communities in which all species coexist, the network of Figure 4a is only a sparse sample of the complete community assembly network. In addition, it is biased towards small communities, because I assemble successively larger communities from single species. Despite these limitations, the network representation can provide a new perspective on community assembly dynamics. For example, it leads to a natural distinction between “transient” communities, which have many outgoing edges and often give rise to other communities, and to “attractor” communities, which have many incoming edges and are frequent products of community assembly. Transient communities occur early during an assembly sequence, have low productivity and little invasion resistance. In contrast, attractor communities are late arising, with high productivity and high invasion resistance (Figure S9 ). The most extreme attractor communities are those that have no outgoing edges – they are end points of assembly, because they cannot be invaded by any species. However, because the network of Figure 4a is a sparse sample of a complete assembly network, I cannot conclude that its 244 communities with outdegree zero are not invasible. This illustrates the limitations of sampling an assembly network. It could be overcome by studying assembly sequences for a much smaller species pool with a manageable number of possible communities. Complete assembly networks also could help answer a number of other questions that sparse sampling cannot. For example, they could help identify stationary distributions of community assembly, quantify non‐transitive interactions between communities and invading species through cycles in the network, and determine whether subsets of communities preferentially give rise to each other, thus creating assembly networks with a “modular” structure. Questions like these remain exciting tasks for future studies. A main technical limitation of my work is its computational cost. For 250 assembly sequences of 5000 h (50,000 time steps of 0.1 h), maximal biomass‐growth needs to be computed 250 × 50,000 = 1.25 × 10 7 times, for each of a community's species, which is represented by a model of approximately 10 3 chemical reactions. Despite the numerical efficiency of FBA, this cost prevented me from asking a number of additional questions. For example, how does invasion resistance relate to other measures of community resilience, such as the ability to survive perturbations in species biomass or environmental nutrients. Do communities show multiple equilibria and priority effects? These questions too remain tasks for future studies. While genome‐scale models are an important step towards biologically realistic models of community assembly, they also have biological limitations. First, they do not represent interference competition, in which bacteria release toxins or antibiotics to slow the growth of competitors. Second, they are not suited to represent viral parasitism or predator–prey interactions, which do occur in bacterial communities (Pérez et al., 2016 ), Third, they neglect that biomass growth is not only influenced by nutrient uptake. For example, different microbial species achieve maximal growth at different pH values, and metabolic by‐products often change the pH of the environment (Ratzke et al., 2020 ; Ratzke & Gore, 2018 ). Fourth, they neglect intercellular communication mechanisms such as quorum sensing, which can help a community coordinate the exploitation of nutrients (Goo et al., 2015 ). The ability to persist over time is advantageous for a community, because it allows a community to “outlive” other communities. On the level of individual organisms, advantageous properties usually arise through Darwinian evolution. However, Darwinian evolution requires selection among competing individuals in a population. Because no analogue of a population exists during the assembly of a single community, an alternative mechanism is needed to explain increased resistance to invasions or other perturbations (Wagner, 2005 ). This mechanism has also been called “sequential selection”, “selection through survival alone” or “systemic selection” (Borrelli et al., 2015 ; Doolittle, 2014 ; Lenton et al., 2018 ). In this form of selection, ecological systems which experience changes that increase their ability to persist become more prevalent over time, because they outlive other, more ephemeral systems (Wagner, 2005 ). In my study system, these events are invasions of highly productive species, which increase the community's ability to resist future invaders. The same principle has been invoked to explain mechanisms that reduce oscillations in predator–prey systems, which make them prone to species extinctions (Doolittle, 2014 ), as well as planetary‐scale homeostatic mechanisms that are required for the Gaia hypothesis (Doolittle, 2014 ; Lenton et al., 2018 ). Darwinian evolution is not the only process that can create living systems with advantageous traits. The same can be achieved by ecological processes as fundamental as resource competition and repeated invasion."
} | 5,384 |
22018234 | null | s2 | 4,310 | {
"abstract": "In bacterial communities, \"tight economic times\" are the norm. Of the many challenges bacteria face in making a living, perhaps none are more important than generating energy, maintaining redox balance, and acquiring carbon and nitrogen to synthesize primary metabolites. The ability of bacteria to meet these challenges depends heavily on the rest of their community. Indeed, the most fundamental way in which bacteria communicate is by importing the substrates for metabolism and exporting metabolic end products. As an illustration of this principle, we will travel down a carbohydrate catabolic pathway common to many species of Bacteroides, highlighting the interspecies interactions established (often inevitably) at its key steps. We also discuss the metabolic considerations in maintaining the stability of host-associated microbial communities."
} | 213 |
22018234 | null | s2 | 4,311 | {
"abstract": "In bacterial communities, \"tight economic times\" are the norm. Of the many challenges bacteria face in making a living, perhaps none are more important than generating energy, maintaining redox balance, and acquiring carbon and nitrogen to synthesize primary metabolites. The ability of bacteria to meet these challenges depends heavily on the rest of their community. Indeed, the most fundamental way in which bacteria communicate is by importing the substrates for metabolism and exporting metabolic end products. As an illustration of this principle, we will travel down a carbohydrate catabolic pathway common to many species of Bacteroides, highlighting the interspecies interactions established (often inevitably) at its key steps. We also discuss the metabolic considerations in maintaining the stability of host-associated microbial communities."
} | 213 |
28776038 | PMC5517106 | pmc | 4,313 | {
"abstract": "Nanocoatings with exceptional mechanical, barrier, and flame-retardant properties were fabricated via one-step coassembly.",
"introduction": "INTRODUCTION Through millions of years of evolution, many biological systems have developed to realize virtually perfect unification of their structures and thus optimized properties. They are usually made of organic and inorganic components arranged in a complicated but amazingly hierarchical structure, enabling them to have a unique combination of remarkable stiffness, strength, toughness, low density, and possibly extra functionality ( 1 , 2 ). One of the most outstanding and representative examples is nacre. Nacre is an organic/inorganic composite with outstanding strength, stiffness, and toughness ( 2 – 6 ). Nacre is composed of ca. 95 volume percent (volume %) of inorganic calcium carbonate (in the form of aragonite) and ca. 5 volume % of organic biopolymers (β-chitin and silk fibroin proteins), both having ordinary mechanical properties ( 7 , 8 ). The striking contrast between the exceptional mechanical properties of nacre and their ordinary components has inspired materials scientists to synthesize organic/inorganic hybrids with a similar structure for practical applications. The key structural features of nacre are a high concentration of well-aligned nanosheets (fig. S1) and a strong interface. Nature has adopted an elaborate strategy to create nacre ( 9 – 13 ), involving a multistep biomineralization process ( 14 ). Although this process has been mimicked in vitro ( 11 , 15 ), it is very difficult to scale up this highly delicate biological process. In addition to mineralization ( 16 , 17 ), a number of approaches, including ice-templated synthesis ( 6 , 18 ), layer-by-layer (LbL) self-assembly ( 19 – 22 ), and electrophoretic deposition ( 23 ), have been explored to form a nacre-like microstructure. Although each of the above approaches has its own advantages, it remains a huge challenge to achieve large-scale continuous mass production of large-sized samples. It is well known and intuitively understandable that flow can help induce orientation ( 24 – 26 ). However, concentrated suspensions of fillers can pose difficulties in achieving filler alignment ( 24 ). Here, we design to create a low-viscosity liquid flow containing both inorganic nanosheets and polymer binders, to help align nanosheets on a substrate surface along the flow direction. During the flow-induced orientation, the nanosheets and polymer chains are expected to coassemble to form a highly ordered layered structure with dozens of layers within a single step, whereas the ratio of the nanosheets and polymer can be easily adjusted, both of which make this process distinctively different from and easier than the LbL assembly ( 21 ).",
"discussion": "DISCUSSION The above systematic characterizations have proven that we managed to assemble and align dozens of layers of MMT nanosheets per cycle of coating with a high efficiency, which can probably be attributed to three causes: (i) The flow helps to induce initial and rough orientation; (ii) the highly crowded nanosheets force themselves to remain highly oriented with each other to accommodate neighboring ones before the coating is dried; and (iii) the final drying process further helps align the nanosheets (fig. S6A). A control experiment was conducted to prepare a PVA/MMT-50-C nanocoating by horizontally casting the same dispersion on the same film substrate. The XRD pattern of the resultant nanocoating is shown in fig. S6B, which showed a broader peak but a similar interlayer distance in comparison to the one prepared vertically with flow. This horizontally coated film also exhibited a much higher O 2 permeability of 8.3 × 10 −19 cm 3 (STP)·cm/cm 2 ·s·Pa, more than five times higher than that of the vertically coated one. These results further suggest that the initial flow-induced orientation is very critical for the alignment of nanosheets in the entire nanocoating formation. Another disadvantage of the conventional horizontal casting method is that it is more difficult to achieve continuous large-scale fabrication, whereas the vertical casting process can be easily scaled up for mass production. Although LbL can continuously achieve uniform thin films, it is much more complicated. In summary, we have developed a facile and effective self-assembly process to form nanocoatings with a nacre-like structure. This new coating process has no limit on lateral dimensions but can only coat to form thin films. Dozens of nanosheets could be well assembled and aligned during each coating cycle, in high contrast to the LbL technique that can assemble only one layer during each cycle ( 21 ). This approach can be easily adopted to be a continuous process (for example, a moving plastic film through a liquid container) for mass production. The composition of the final nanocoating can be easily adjusted, leading to a high composition flexibility and property tunability. The nanocoatings exhibited outstanding mechanical, barrier, and flame-retardant properties, promising for widespread application."
} | 1,276 |
23346080 | PMC3548243 | pmc | 4,315 | {
"abstract": "Symbioses in marine sponges involve diverse consortia of microorganisms that contribute to the health and ecology of their hosts. The microbial communities of 13 taxonomically diverse Great Barrier Reef (GBR) sponge species were assessed by DGGE and 16S rRNA gene sequencing to determine intra and inter species variation in bacterial symbiont composition. Microbial profiling revealed communities that were largely conserved within different individuals of each species with intra species similarity ranging from 65–100%. 16S rRNA gene sequencing revealed that the communities were dominated by Proteobacteria , Chloroflexi, Acidobacteria, Actinobacteria, Nitrospira , and Cyanobacteria . Sponge-associated microbes were also highly host-specific with no operational taxonomic units (OTUs) common to all species and the most ubiquitous OTU found in only 5 of the 13 sponge species. In total, 91% of the OTUs were restricted to a single sponge species. However, GBR sponge microbes were more closely related to other sponge-derived bacteria than they were to environmental communities with sequences falling within 50 of the 173 previously defined sponge-(or sponge-coral) specific sequence clusters (SC). These SC spanned the Acidobacteria, Actinobacteria, Proteobacteria, Bacteroidetes, Chloroflexi, Cyanobacteria, Gemmatimonadetes, Nitrospira , and the Planctomycetes-Verrucomicrobia-Chlamydiae superphylum. The number of sequences assigned to these sponge-specific clusters across all species ranged from 0 to 92%. No relationship between host phylogeny and symbiont communities were observed across the different sponge orders, although the highest level of similarity was detected in two closely related Xestospongia species. This study identifies the core microbial inhabitants in a range of GBR sponges thereby providing the basis for future studies on sponge symbiotic function and research aiming to predict how sponge holobionts will respond to environmental perturbation.",
"introduction": "Introduction Associations between sponges and bacteria have existed for 600 million years making them one of the most ancient of all symbioses between microbes and metazoa (Wilkinson, 1984 ). Most sponges host diverse and abundant communities of microorganisms (Hentschel et al., 2006 ; Taylor et al., 2007 ; Webster et al., 2010 ), which contribute to host health, ecology and evolution [reviewed in (Taylor et al., 2007 ) and (Webster and Taylor, 2012 )]. The importance of the relationship between sponges and their associated microbial communities is supported by the fact that microorganisms can contribute to more than 35% of the sponge biomass (Vacelet, 1975 ) and undertake diverse functional roles including nutrition, cycling of metabolites and host defense [reviewed in (Hentschel et al., 2012 ; Webster and Taylor, 2012 )]. At least 32 bacterial phyla and candidate phyla have so far been found in sponges, either via cultivation or molecular characterization (Taylor et al., 2007 ; Schmitt et al., 2012b ; Webster and Taylor, 2012 ). Some of these may be transient members of the sponge microbiota, potentially being filtered from the seawater as food. However, the “core” taxa, thought to represent the stable sponge inhabitants or true symbionts, include the Acidobacteria, Actinobacteria, Chloroflexi, Cyanobacteria, Gemmatimonadetes, Nitrospira, Proteobacteria , (especially Alpha , Delta , Gamma classes) and the candidate phylum “ Poribacteria ” (Taylor et al., 2007 ; Schmitt et al., 2012b ). The existence of sponge-specific microorganisms was first reported over a decade ago based on the finding that distantly related sponges from geographically separated regions shared microbes that had not been recovered from any other source, including the surrounding seawater (Hentschel et al., 2002 ; Taylor et al., 2007 ). Clusters of sponge-specific sequences were defined if groups of at least three rRNA gene sequences derived from more than two sponge species shared higher similarity to each other compared to sequences from other environments (Hentschel et al., 2002 ). Recently, the concept of sponge-specific microbes was comprehensively explored by performing phylogenetic analyses of all publicly available 16S and 18S rRNA gene sequences that originated from sponges. In total, 27% of all sponge-derived sequences fell into monophyletic, sponge-specific sequence clusters (SC) within the Bacteria, Archaea, and Fungi (Simister et al., 2012a ) and additional sequences fell within clusters containing both sponge and coral derived sequences (SCC). Within the Bacteria, Chloroflexi, Cyanobacteria , “ Poribacteria ”, Betaproteobacteria , and Acidobacteria contained the highest abundance of these SCs. However, deep sequencing of diverse marine environments including seawater, sediments, hydrothermal vents, salt marshes, microbial mats, and corals has recently demonstrated that the rare biosphere may be a reservoir for some previously designated “sponge-specific” microbes (Webster et al., 2010 ; Taylor et al., 2012 ). It is well established that particular sponge species can host stable microbial populations that are different to the communities in other species (Wilkinson et al., 1981 ; Taylor et al., 2004 , 2005 ; Webster et al., 2004 , 2010 ; Holmes and Blanch, 2007 ; Erwin and Thacker, 2008b ; Turque et al., 2008 ; Lee et al., 2009 ; Erwin et al., 2011 , 2012a , b ; Schmitt et al., 2012b ) and some molecular evidence also supports the potential for host-symbiont coevolution (Erpenbeck et al., 2002 ; Thacker and Starnes, 2003 ; Fan et al., 2012 ). Throughout the sponge literature and within this manuscript, the term “symbiosis” is used in its loosest possible definition, referring to the stable host-microbe association rather than implying any symbiotic function to the relationship. To test whether geographic or host-specific subpopulations of sponge microbes exist, Schmitt and colleagues performed 454 amplicon sequencing on 32 sponge species collected from eight locations around the world (Schmitt et al., 2012b ). Whilst tropical sponge species shared more similarity in their microbial communities than they did with sub-tropical species, no other geographic or host phylogeny patterns were detected (Schmitt et al., 2012b ). Interestingly, only a small “core” bacterial community was present in all 32 sponge species with the majority of bacterial operational taxonomic units (OTUs) occurring in only a single sponge species. Whilst the different sponge species were found to contain unique microbial populations, most of these sponge-associated bacteria were more closely related to other sponge-derived bacteria than they were to microbes from other environmental origins. These findings suggest that exploration of further sponge species may reveal additional novel sequences which would enhance our understanding of these sponge-specific microbial communities. Over 8500 sponge species have been described globally (van Soest et al., 2012 ) with an estimated 2500 species occurring in Australian waters, although many of these remain to be formally described (Hooper, 2009 ). Baseline data on the composition and stability of symbiotic microbial communities is lacking for most sponge species and this knowledge gap makes it difficult to determine the role of microorganisms in sponge morbidity and mortality events. Sponge disease and mass mortalities have increased over recent years (Webster, 2007 ; Maldonado et al., 2010 ; Angermeier et al., 2011 , 2012 ), including numerous reports of diseases that affect Great Barrier Reef (GBR) sponge species (Webster et al., 2002 ; Webster, 2007 ; Luter et al., 2010a , b ). The study of sponge disease (including the identification of causative agents) is hampered by the high bacterial diversity within sponges combined with inadequate knowledge of the inherent microbial communities for most species. Understanding how sponges are likely to respond to a rapidly changing environment has also been a recent research focus (Webster et al., 2008 ; López-Legentil et al., 2010 ; Simister et al., 2012b ), but scientists require a much better understanding of the diversity and specificity of microbes before assessments of environmental change can be validated. Enhanced understanding of the ecological and evolutionary implications of sponge-bacterial symbioses gained over the past decade has prompted considerable new research in this field (Webster and Taylor, 2012 ), however, some regions such as the GBR remain largely understudied. The GBR is home to over 1500 sponge species although the microbial associates of the vast majority of species are yet to be explored. Here we surveyed the taxonomic composition of microbial associates in 13 of the most abundant and ecologically important GBR sponge species spanning five orders within the Porifera. By profiling replicate individuals per sponge species we are also able to address questions related to intra and inter species specificity.",
"discussion": "Discussion The microbial communities of taxonomically diverse GBR sponges were dominated by phyla previously reported to associate with marine sponges. The microbial populations were highly conserved within different individuals of each species yet varied significantly between species. This is consistent with patterns for other GBR sponges (Luter et al., 2010b , 2012 ; Webster et al., 2011 ; Fan et al., 2012 ) and in species from other broad geographic locations (Schmitt et al., 2012b ; Webster and Taylor, 2012 ). Whilst BLAST analysis did not reveal common indicator species between the DGGE and the clone library analyses, both datasets were dominated by sequences that had sponge and coral derived bacteria as their closest relatives. Further highlighting the specific nature of GBR sponge-bacterial populations was the finding that all species (with the exception of Hamigera sp.) contained clone sequences that fell within previously defined clusters of sponge-specific bacterial sequences (Simister et al., 2012a ). The dominance of Proteobacteria, Chloroflexi, Acidobacteria, Actinobacteria, Nitrospira , and Cyanobacteria across the GBR sponge species accords with the dominant microbial groups described from other studies of sponge-associated microbial assemblages (Taylor et al., 2007 ; Webster and Taylor, 2012 ). Two of the GBR sponge species from this study ( C. coralliophila and Stylissa sp.) were recently assessed using metagenomic shotgun sequencing (Fan et al., 2012 ). Whilst the different sequencing approaches were generally consistent, some methodological bias was evident with metagenomic sequencing detecting an abundance of Chloroflexi in C. coralliophila and a dominance of Archaea in Stylissa sp. which were not detected by DGGE or clone sequencing. The community composition in X. testudinaria from the present study showed high similarity with the microbial community described for this species in Indonesia (Montalvo and Hill, 2011 ). However, Cyanobacteria , which are well studied associates of Caribbean Xestospongia spp., were not detected in either of the GBR Xestospongia species (Steindler et al., 2005 ; Erwin and Thacker, 2007 , 2008a ). Episodes of both cyclic and fatal bleaching have dramatically effected populations of Xestospongia muta in the Caribbean (López-Legentil et al., 2008 ; McMurray et al., 2011 ), whereas there is currently no evidence of bleaching within GBR Xestospongia species. The absence of photosynthetic cyanobacteria in GBR species not only distinguishes these geographically separated congeners, but raises questions of GBR populations being less vulnerable to bleaching impacts associated with a changing climate. Results from the present study indicate that C. singaporensis , C. foliascens , and C. coralliophila are likely to be phototrophic species due to the high representation of Cyanobacteria . This is an important finding for studies that may want to distinguish differential responses of heterotrophs or phototrophs to environmental perturbation. The Chloroflexi were present in over 50% of the surveyed GBR sponge species and were particularly abundant in Cinachyra sp., X. exigua , and C. matthewsi . Another recent investigation compared the presence of Chloroflexi in sponge species containing either high (HMA) or low (LMA) microbial abundance (Schmitt et al., 2011 ). Clear differences were observed between HMA and LMA species including a greater abundance, diversity and sponge-specificity of Chloroflexi in HMA sponges. In contrast, LMA species tended to be highly variable and contain Chloroflexi sequences related to seawater-derived microorganisms. These findings indicate a functional importance for Chloroflexi in HMA sponges and whilst we would predict that Cinachyra sp., X. exigua , and C. matthewsi from the GBR are HMA species based on sequence analysis of Chloroflexi , further microscopy analysis would be required to determine the microbial abundance within each of our GBR sponges. No relationship between host phylogeny and bacterial communities were observed within the Dictyoceratida, Haplosclerida, Poescilosclerida, Halichondrida, and Pirophorida from the GBR. This finding is consistent with a 454 amplicon pyrosequencing study that surveyed sponge microbes across 32 different host species encompassing nine sponge orders (Schmitt et al., 2012b ). Species within the same orders did not contain more similar microbial communities than species in different orders nor was there any clear correlation to host phylogeny when three species each within the genera Aplysina, Hyrtios , and Ircinia were compared (Schmitt et al., 2012b ). Cospeciation of sponges and their symbionts would be most likely to occur if the microbial associates were transmitted strictly vertically (from adult to gametes/larvae). Research over the past decade has revealed that many sponges use both vertical (Usher et al., 2001 ; Oren et al., 2005 ; Enticknap et al., 2006 ; Schmitt et al., 2007 ; Sharp et al., 2007 ) and horizontal (Taylor et al., 2007 ; Schmitt et al., 2008 ; Webster et al., 2010 ) transmission strategies to maintain their complex and diverse microbial communities. The lack of correlation between host phylogeny and bacterial composition in the GBR sponges also suggests a strategy of symbiont acquisition that incorporates both vertical and horizontal transmission. It has been hypothesized that abundant microbes common to all samples within a given habitat must be essential for the function of that community (Shade and Handelsman, 2012 ). Hence, the identification of core microbiomes (microorganisms that are shared between two or more samples) in complex microbial habitats can enhance our understanding of systems ecology. Across the 13 GBR sponge species investigated here, the core microbiome within each species was high yet the microbiome shared between species was low. However, it should be noted that additional sequencing for each species may reveal further sequences common to multiple samples. No OTUs were present in more than five different sponge species and 91% of the total OTUs were species-specific. Schmitt and colleagues ( 2012b ) also identified a minimal core bacterial community in 32 different sponge species. In another study of five Mediterranean sponge species, 72% of the detected OTUs were species-specific, 26% were common to two or more species and only 2% were shared amongst all five species (Schmitt et al., 2012a ). The study of three sympatric Mediterranean Ircinia sp. also revealed host species-specific assemblages and identified one proportion of the OTUs that were shared between the two most phylogenetically related species and a second component that was shared between the two species sharing the same cryptic habitat (Erwin et al., 2012a ). These findings indicate that factors relevant to each host species can contribute to structuring the distinct microbial communities. Whilst next generation sequencing would further elucidate the host-specific nature of GBR sponge microbial communities, our results contribute to this growing body of evidence for species-specific microbial associations in sponges. The vast majority of GBR sponges surveyed in the current study had high proportions of sequences that were more similar to other sponge-derived sequences than they were to sequences from the environment or other sources. These findings are consistent with our understanding of sponge-specific SC (Hentschel et al., 2002 ; Taylor et al., 2007 ; Simister et al., 2012a ) and further our knowledge of their distribution and geographic range. In the global sponge microbial survey, each of the 32 species was also found to host bacteria that were more closely related to each other than they were to non-sponge derived bacterial species (Schmitt et al., 2012b ). However, a recent survey of the rare microbial biosphere from ~650 marine samples collected from diverse habitats detected 77 of the 173 known sponge-specific SC in environmental (non-sponge) samples, although these were generally found at extremely low abundances (Taylor et al., 2012 ). To date, little is known about the functional roles of these “sponge-specific” bacteria although single celled genomics of a sponge-specific Poribacteria indicated potential symbiotic functions including autotrophic carbon fixation and vitamin B12 production (Siegl et al., 2011 ). Interestingly, sponge species from the present study which contained no sequences within sponge-specific clusters still maintained highly conserved bacterial populations, indicating that species without these “sponge-specific” sequences are still capable of structuring their communities and are not just reflecting the microbial composition of the surrounding seawater. For example, Hamigera sp. had no sequences within SC/SCCs and only 2% of Paramyxilla sp. fell within a SC yet the replicate individuals within each species shared 90 and 80% similarity respectively. Similarly, the highest number of sequences falling within SC/SCCs occurred in L. variabilis (92% of total sequences), yet this species had the lowest intra-species similarity based on microbial profiling (65%). Whilst the number of microbes shared between different GBR sponge species was small, the core microbiome within each species was large, with the vast majority of microbes conserved in all replicate individuals per species. This indicates that these microbes are likely to be functionally important to the ecology of each species. Metagenomic sequence analysis recently highlighted how core symbiont functions can be provided in different sponge species by functionally equivalent microbes and analogous enzymes or biosynthetic pathways (Fan et al., 2012 ). Therefore, a “core microbiome” at the phylogenetic level does not need to exist across species. As indicated by Shade and Handelsman ( 2012 ), identifying the “core” microorganisms in any habitat is essential for defining the healthy community and subsequently predicting how the community will respond to perturbation. Given the increasing incidence of sponge disease and health declines associated with climate change, an enhanced understanding of the species-specific symbionts in GBR sponges will be a valuable resource. Conflict of interest statement The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest."
} | 4,903 |
24860240 | null | s2 | 4,316 | {
"abstract": "Darwin proposed two seemingly contradictory hypotheses for a better understanding of biological invasions. Strong relatedness of invaders to native communities as an indication of niche overlap could promote naturalization because of appropriate niche adaptation, but could also hamper naturalization because of negative interactions with native species ('Darwin's naturalization hypothesis'). Although these hypotheses provide clear and opposing predictions for expected patterns of species relatedness in invaded communities, so far no study has been able to clearly disentangle the underlying mechanisms. We hypothesize that conflicting past results are mainly due to the neglected role of spatial resolution of the community sampling. In this study, we corroborate both of Darwin's expectations by using phylogenetic relatedness as a measure of niche overlap and by testing the effects of sampling resolution in highly invaded coastal plant communities. At spatial resolutions fine enough to detect signatures of biotic interactions, we find that most invaders are less related to their nearest relative in invaded plant communities than expected by chance (phylogenetic overdispersion). Yet at coarser spatial resolutions, native assemblages become more invasible for closely-related species as a consequence of habitat filtering (phylogenetic clustering). Recognition of the importance of the spatial resolution at which communities are studied allows apparently contrasting theoretical and empirical results to be reconciled. Our study opens new perspectives on how to better detect, differentiate and understand the impact of negative biotic interactions and habitat filtering on the ability of invaders to establish in native communities."
} | 436 |
23986752 | PMC3750489 | pmc | 4,317 | {
"abstract": "Hyperthermophilic bacteria from the Thermotogales lineage can produce hydrogen by fermenting a wide range of carbohydrates. Previous experimental studies identified a large fraction of genes committed to carbohydrate degradation and utilization in the model bacterium Thermotoga maritima . Knowledge of these genes enabled comprehensive reconstruction of biochemical pathways comprising the carbohydrate utilization network. However, transcriptional factors (TFs) and regulatory mechanisms driving this network remained largely unknown. Here, we used an integrated approach based on comparative analysis of genomic and transcriptomic data for the reconstruction of the carbohydrate utilization regulatory networks in 11 Thermotogales genomes. We identified DNA-binding motifs and regulons for 19 orthologous TFs in the Thermotogales . The inferred regulatory network in T. maritima contains 181 genes encoding TFs, sugar catabolic enzymes and ABC-family transporters. In contrast to many previously described bacteria, a transcriptional regulation strategy of Thermotoga does not employ global regulatory factors. The reconstructed regulatory network in T. maritima was validated by gene expression profiling on a panel of mono- and disaccharides and by in vitro DNA-binding assays. The observed upregulation of genes involved in catabolism of pectin, trehalose, cellobiose, arabinose, rhamnose, xylose, glucose, galactose, and ribose showed a strong correlation with the UxaR, TreR, BglR, CelR, AraR, RhaR, XylR, GluR, GalR, and RbsR regulons. Ultimately, this study elucidated the transcriptional regulatory network and mechanisms controlling expression of carbohydrate utilization genes in T. maritima . In addition to improving the functional annotations of associated transporters and catabolic enzymes, this research provides novel insights into the evolution of regulatory networks in Thermotogales .",
"introduction": "Introduction Carbohydrates constitute the most abundant single class of organic substances found in nature. Plant cell walls constitute ~70% of the worldwide biomass production by land plants, but only ~2% of this biomass is currently utilized by humans (Pauly and Keegstra, 2010 ). Carbohydrate composition of plant biomass is highly diverse and differs significantly between plants. A limited number of mono—or disaccharides compose a majority of plant biopolymers. Cellulose, hemicelluloses, and pectins are major polysaccharides of the plant cell wall. Hemicelluloses are characterized by a large diversity of building blocks including pentoses (xylose, arabinose) and hexoses (glucose, mannose, galactose, and rhamnose, and uronic acids). Pectins are composed of galacturonate and rhamnose residues with various branching side chains. Microbial degradation of plant cell wall polysaccharides and their conversion to biofuels is a vital objective for society and requires novel efficient technologies. Sugar utilization pathways are major feed lines of carbon and energy for central metabolism in a large variety of heterotrophic bacteria. Although these pathways and their transcriptional regulation were extensively studied in model bacteria, projection of this knowledge to diverse bacteria is a major challenge due to chemical diversity of carbohydrates and a matching variability of microbial sugar utilization genes, pathways, and regulons. This variability includes alternative biochemical routes, non-orthologous gene replacements, and functionally heterogeneous families of paralogs. Due to this complexity, genomic annotations of sugar utilization genes derived solely from sequence similarity analysis are often imprecise and incomplete, especially for distantly related bacteria. Metabolic reconstruction based on a combination of various types of genomic context analysis (Yang et al., 2006 ; Rodionov et al., 2007 ; Leyn et al., 2012 ), primarily operons and regulons, allows us to substantially improve the quality of functional annotations and predictions, enabling more accurate metabolic modeling. In this study, we applied an integrative approach to simultaneous genomics-based reconstruction of both metabolic and associated regulatory networks to study carbohydrate utilization networks in a hyperthermophilic marine bacterium from the phylogenetically deep-branching Thermotogales group. Thermotoga spp. are anaerobic fermentative bacteria that are able to grow on various simple and complex carbohydrates including glucose, starch, cellobiose, xylan, and pectin while producing hydrogen, carbon dioxide, and acetate (Chhabra et al., 2002 ; Kluskens et al., 2003 ; Conners et al., 2006 ). T. maritima MSB8, a model bacterium in the Thermotoga group, was isolated from geothermally heated marine sediments of Volcano Island in Italy (Huber et al., 1986 ). A closely related bacterium, T. neapolitana , was isolated from a submarine thermal vent near Lucrino, Bay of Naples, Italy (Jannasch et al., 1988 ). Other Thermotogales have a broad geographic distribution (see Figure 1 ) including hydrothermal vents in the Azores [ Themotoga sp. RQ-2, (Swithers et al., 2011 )], a sulfate-reducing bioreactor in Europe [ T. lettingae , (Balk et al., 2002 )], a hot spring in New Zealand [ Fervidobacterium nodosum , (Patel et al., 1985 )], and an offshore oil reservoir in Japan [ T. naphthophila, T. petrophila , (Takahata et al., 2000 )]. Figure 1 Properties of 11 Thermotogales species analyzed in this study. (A) The geographic isolation sites for the Thermotogales strains. (B) Distribution of sugar utilization pathways and cognate transcriptional regulators. The presence of orthologous genes encoding regulators and associated sugar catabolic pathways is shown by “+” and colored circles, respectively. The phylogenetic species tree was constructed using the concatenated alignment of 78 universal bacterial proteins in the MicrobesOnline database ( http://www.microbesonline.org/cgi-bin/speciesTree.cgi ). The complete genome sequence of T. maritima revealed a large proportion of genes (10–15%) involved in carbohydrate metabolism (Nelson et al., 1999 ). A number of T. maritima and T. neapolitana enzymes that can degrade β-glucan, hemicelluloses, and pectin have been studied by functional genomics or biochemistry (Conners et al., 2006 ). A three-dimensional reconstruction of the central metabolic network of T. maritima includes 478 enzymes, 120 of which were determined experimentally (Zhang et al., 2009 ). Still, many metabolic gaps remained in sugar catabolic networks of T. maritima . We recently investigated some of those gaps using a combination of bioinformatics and experimental techniques (Rodionova et al., 2012a , b , 2013 ). The comprehensive genome-scale metabolic modeling of T. maritima also requires an understanding of mechanisms, components, and behavior of the transcriptional regulatory machinery. The integration of genome-scale metabolic and regulatory models significantly improves growth phenotype predictions and allows for the interpretation of systems biology datasets (Faria et al., 2013 ). Functional genomics approaches were previously used to study global gene expression in response to growth of T. maritima on sugars. The transcriptional responses of T. maritima to various mono- and polysaccharides identified subsets of genes that are coordinately regulated in response to specific carbohydrates (Chhabra et al., 2003 ; Nguyen et al., 2004 ; Conners et al., 2005 ; Frock et al., 2012 ). However, the molecular mechanisms underlying these transcriptional responses remain unclear. To infer sugar-responsive transcriptional regulons in Thermotoga , we started from the initial metabolic reconstruction, applying a combination of comparative genomics-based methods of regulatory reconstruction (Rodionov, 2007 ). Previously, we applied these methods to reconstruct a large number of metabolic regulons in diverse taxonomic groups of bacteria (Rodionov et al., 2011 ; Leyn et al., 2013 ; Ravcheev et al., 2013 ). Integration of metabolic and regulatory reconstructions is particularly efficient for elucidation of carbohydrate utilization networks in a group of taxonomically related bacteria, as recently demonstrated by the comparative genomic study of Shewanella spp. (Rodionov et al., 2010 ). In previous studies, we also combined genomics-based inferences with experimental elucidation to identify and characterize several novel regulons in Thermotoga , including Rex controlling the central carbon metabolism and hydrogen production (Ravcheev et al., 2012 ) and seven ROK-family transcription factors controlling the sugar utilization pathways (Kazanov et al., 2013 ). Here, we extended this analysis toward the entire transcriptional network for sugar catabolism in T. maritima compared to 10 additional Thermotogales species with completely sequenced genomes. The reconstructed networks allowed us to improve gene annotations, refine associated pathways, and identify novel, yet uncharacterized, enzymes and transporters tentatively implicated in the sugar utilization machinery. Some of the critical inferences about newly identified regulons were experimentally verified. A comparison with global transcriptomic data obtained for T. maritima grown on different carbohydrates provided additional validation and enrichment of the genomic reconstruction.",
"discussion": "Discussion Transcriptional regulation is a highly variable component of carbohydrate utilization networks. Although our current knowledge of sugar catabolic regulons in model bacteria such as Escherichia coli and Bacillus subtilis is nearly comprehensive, the projection of this knowledge to the sequenced genomes of bacteria from distant taxonomic groups is still a challenge. The major difficulties for bioinformatics-based propagation of TF regulons include duplications and losses of TFs and their binding sites, rapid diversification of specificities of TFs toward DNA sites and sugar effectors, non-orthologous replacements of entire regulatory systems, and frequent horizontal transfers. Comparative genomic analysis of specific sugar utilization pathways has led to substantial progress in the identification and reconstruction of their cognate TF regulons in diverse bacterial lineages (Yang et al., 2006 ; Rodionov et al., 2010 ; Leyn et al., 2012 ). In this study, we used comparative genomics to reconstruct novel TF regulons for sugar catabolism in hyperthermophilic bacteria from the deep-branching lineage Thermotogales (Figure 1 ). In T. maritima , a model bacterium in this lineage, the identified sugar catabolic regulatory network includes 18 TFs, 40 TF binding sites, and 163 target genes comprising 15 known and three hypothetical metabolic pathways (Table 1 ; Figures 2 , 3 ). All studied sugar-responsive TFs in T. maritima are encoded within their cognate regulated gene loci, and thus are subject to autoregulation. All sugar utilization regulons reconstructed in Thermotogales are controlled by TFs that are homologous to known sugar-related regulators from six families: DeoR, GntR, IclR, LacI, ROK, and RpiR. For most of these TFs, phylogenetic and genome context analysis of distant homologs did not reveal their potential functional orthologs outside the Thermotogales lineage. These genomic observations suggest that the analyzed sugar metabolic regulators have evolved and functionally specialized after the separation of the Thermotogales . Our previous phylogenetic analysis suggests that the expansion of the ROK family represented by seven paralogs in T. maritima was likely due to massive duplications and subsequent functional diversification of regulators during the evolution of Thermotogales (Kazanov et al., 2013 ). The DeoR-family regulator FruR controlling the fructose utilization operon in two closely related Thermotoga spp. represents an exceptional case of relatively recent horizontal gene transfer from thermophilic Clostridia. All genes from the fructose utilization operon in Thermotoga spp., including fruR , are highly similar to the orthologous genes in Caldicellulosiruptor spp. (70–80% identity). We propose that the entire fructose utilization operon was laterally transferred between these two lineages that likely share the same ecological niche. The observed variations in the distribution of sugar utilization regulons among the Thermotogales are also remarkable (Figure 1 ). Only six regulons are present in at least two different genera (CelR, ChiR, GalR, GloR, RbsR, and TreR), whereas the remaining regulons are present only in the Thermotoga genus. Moreover, most of the latter regulons are restricted to the group of five closely related Thermotoga spp, and only three regulons have orthologs in a more distant genome of T. lettingae . Conservation of all Thermotoga -specific regulons except FruR suggests their likely emergence in the common ancestor of this genus. We observed several cases of species-specific regulon loss in which an entire regulon (including all operons from a regulated pathway) is missing only in a single Thermotoga spp., including the inositol and ribose utilization regulons in the RQ-2 strain and the glucose utilization regulon GluR in T. petrophila . In another case, the RhaR regulator was lost in a single strain of Thermotoga sp. RQ-2, however, the rhamnose utilization operons are still retained in the genome. Many other cases of the absence of orthologous regulators for sugar catabolic pathways in the genomes of more distant Thermotogales can be explained by their control by non-orthologous TFs that have yet to be characterized (Figure 1 ). Overall, the reconstructed carbohydrate utilization regulatory network in T. maritima contains 18 local TF regulons (FruR is absent from T. maritima ), each controlling between one and seven operons. In contrast to other model bacteria (e.g., E. coli and B. subtilis ) that employ various transcriptional mechanisms for global catabolite repression of sugar metabolism, our genomic-based analysis did not reveal any global regulons for sugar utilization genes in Thermotoga . This observation correlates with the genome-wide transciptome data for T. maritima obtained in this and previous studies (Chhabra et al., 2003 ; Frock et al., 2012 ). With the exception of a fructose-specific system in two Thermotoga strains, the Thermotogales genomes lack orthologs of sugar phosphotranferase (PTS) systems that are essential components of global carbon catabolic repression mechanisms in both Gram-negative and Gram-positive bacteria (Deutscher, 2008 ). Thus, bacteria from the Thermotogae phylum appear to use different regulatory strategies for sugar utilization. Another remarkable feature of sugar catabolic networks in Thermotoga is the existence of multiple interconnections between distinct sugar regulons. Two xylan catabolic gene loci, xtp and xlo , were found to be controlled by the xylose-responsive regulator XylR and the predicted glucuronate utilization regulator KdgR in T. maritima (Figure 2 ) and in other Thermotoga genomes (Table S2 in Supplementary Material). The first gene locus encodes a secreted endoxylanase, a putative xylan oligosaccharide ABC transporter, and a cytoplasmic glucuronidase, which might be involved in the cleavage of glucuronate residues from a xylan-derived oligosaccharide. The second gene locus encodes another secreted endoxylanase, a xylose oligosaccharide ABC transporter, and a cytoplasmic xylosidase and is presumably involved in the utilization of xylose-containing oligosaccharides. From transcriptomic data in T. maritima , both XylR/KdgR-regulated gene loci are upregulated during growth on xylose (Tables S3 in Supplementary Material) and xylan (Conners et al., 2005 ). Another gene locus, glo , encoding hypothetical glucose oligosaccharide utilization genes, is predicted to be controlled by both the locally encoded regulator GloR and the distal regulator CelR, which also controls five other operons involved in cellobiose and glucan utilization (Figure 2 ). Lastly, the trehalose transporter treEFG in five Thermotoga spp. is controlled by the trehalose- and glucose-responsive regulators TreR and GluR, respectively (Figure 4 ). We confirmed by RT-PCR that the treE gene in T. maritima is upregulated on glucose and trehalose (Figure 7 ). The observed partial overlaps between distinct local regulons in Thermotoga point to the existence of coordinated regulatory responses to particular types of complex polysaccharides and disaccharides. We assessed the reconstructed sugar regulatory network in T. maritima by microarray gene expression profiling of cells grown on a variety of mono- and disaccharides and pectin as a single carbon source. The obtained gene induction patterns demonstrate overall consistency with individual regulons reconstructed by genomic analysis. In addition, the previously determined effectors of the ROK-family regulators XylR, TreR, GluR, and BglR (Kazanov et al., 2013 ) were found to be in good agreement with the in vivo gene expression results for xylose, trehalose, glucose, and cellobiose, respectively (Figure 2 ). In addition, the reconstructed regulons correlated with previous microarray expression studies in T. maritima (Chhabra et al., 2003 ; Nguyen et al., 2004 ; Conners et al., 2005 ). For instance, the predicted galactoside utilization regulon GalR was induced by galactose in our study and was previously shown to be upregulated on lactose, a galactose-glucose disaccharide (Nguyen et al., 2004 ). The predicted cellobiose and glucan utilization regulon CelR was induced by cellobiose in our study and was previously shown to be upregulated by barley and glucomannan (Conners et al., 2005 ). The reconstructed regulatory network allowed us to suggest and/or refine specific functional assignments for sugar-specific ABC transporters that constitute the most abundant class of uptake transporters in sugar utilization pathways of Thermotogales . In T. maritima , the previously uncharacterized transporter ChiEFG was tentatively assigned chitobiose specificity based on the chitobiose-responsive regulon ChiR (Figure 3 ). Similarly, the characterized GluR and TreR regulons and their respective effectors allowed us to propose glucose and trehalose specificities for the GluEFK and TreEFG transporters, respectively, and these functional assignments were recently experimentally validated (Boucher and Noll, 2011 ). Finally, the reconstructed regulons of other Thermotoga species allowed us to predict sugar specificities for multiple novel ABC transporters that are non-orthologous to transport systems from T. maritima (Table S1 in Supplementary Material). Interestingly, despite of the absence of FruR regulon, T. maritima demonstrates some growth on fructose as a sole carbon source (Figure 5 ). This observation correlates with the presence of functional fructokinase TM0296 (Rodionova et al., 2012b ), however, the fructose uptake transporter remains unknown in T. maritima . In summary, the comparative genomics-based regulon and pathway reconstruction combined with some experimental data allowed us to identify the integrated metabolic and regulatory network of sugar utilization in T. maritima and revealed its substantial diversity between different Thermotogales species. Our results revealed a high level of consistency between the in silico –predicted TF regulons, the in vitro –determined TF effector specificities, and the in vivo –measured gene expression changes. The described integrative genomics–based approach for regulon analysis may be applied to other yet uncharacterized taxonomic groups of microorganisms for which multiple closely related genomes are available."
} | 4,974 |
36990451 | PMC10103138 | pmc | 4,318 | {
"abstract": "DNA nanotechnology\nis a unique field, where physics, chemistry,\nbiology, mathematics, engineering, and materials science can elegantly\nconverge. Since the original proposal of Nadrian Seeman, significant\nadvances have been achieved in the past four decades. During this\nglory time, the DNA origami technique developed by Paul Rothemund\nfurther pushed the field forward with a vigorous momentum, fostering\na plethora of concepts, models, methodologies, and applications that\nwere not thought of before. This review focuses on the recent progress\nin DNA origami-engineered nanomaterials in the past five years, outlining\nthe exciting achievements as well as the unexplored research avenues.\nWe believe that the spirit and assets that Seeman left for scientists\nwill continue to bring interdisciplinary innovations and useful applications\nto this field in the next decade.",
"conclusion": "9 Conclusions and Outlook In recent years,\nDNA nanotechnology has made a great leap in developing\nnew self-assembly methodologies as well as generating DNA-based nanomaterials\nwith diverse functions. Crucially, effective solutions have been found\nto remove or circumvent bottlenecks in the field. For instance, the\nassembly of DNA nanostructures usually requires a large number of\nssDNA species of considerable quantity, and thus the high cost of\nDNA synthesis has been an issue for quite some time. To this end,\nseveral methods were developed to reduce the cost of DNA synthesis,\nincluding the chip-synthesized DNA followed by parallel enzymatic\namplification 509 and the bacteriophage-based\nproduction of single-stranded precursor DNA followed by cleavage using\nDNAzymes. 509a − 511 Consequently, the cost of DNA-based\nnanomaterials is largely reduced, especially for large and complex\nDNA structures. This would substantially help to expand the applications\nof DNA nanotechnology in many research fields. However, when compared\nto that of conventional nanomaterials, such as lipid NPs, polymers,\ninorganic metallic/nonmetallic materials, etc., the cost of DNA production\nand structural fabrication is still higher in general. Novel approaches\nare needed to achieve large-scale, reliable, and good manufacturing\npractice-compliant production of DNA origami-based nanomaterials. There remain many outstanding challenges and open questions in\nthe next decade that await innovative solutions and further investigations.\nFrom the perspective of self-assembly methodologies, the development\nof new computational tools would be invaluable to the optimization\nof the assembly process, as well as the accurate prediction of the\nassembly pathways and outcomes. Current DNA origami design schemes\nstill largely rely on researchers’ intuition and experiences,\nwhich may differ considerably from person to person. In particular,\nfor complicated 3D origami structures, minor design differences often\nlead to large variations in the assembly yield. The design optimization\ncan thus be both time- and cost-consuming. Advanced computational\ntools will help to standardize the design process and reduce the trial\nand error involved to generate desired assembly products. Furthermore,\nthe perspective of interfacing protein engineering\nwith DNA nanotechnology is extremely exciting, owing to the proteins’\nmultifaceted cellular functionalities and excellent biocompatibility.\nRecent work by Praetorius and Dietz showed DNA–protein hybrid\nstructures that may one day be produced in cells, opening exciting\nopportunities to encode, assemble, and operate nanodevices made of\nDNA–protein complexes in vivo . 327 Nevertheless, such DNA–protein hybrid\nstructures mainly take advantage of the nucleic acid-binding proteins,\nwhile functional proteins are excluded from the final structures.\nHence, the development of a general strategy to design genetically\nencoded protein assemblies that can form inside cells would be of\ngreat interest. Cell-compatible functional protein assemblies would\nopen up a new research area, where genetically encoded proteins can\nautonomously assemble into target geometries and carry out designated\ntasks in vivo . One of the possible solutions would\nbe the assembly of protein–RNA origami from co-transcription.\nConsidering the high structural diversity of proteins, however, this\nambitious goal may involve substantial challenges. The main issue\nis that protein folding is far more complicated than DNA self-assembly,\nand the correct assembly of protein structures often relies on additional\nmolecular machineries for protein transport, modification, and quality\ncontrol. Nevertheless, we remain optimistic in this direction considering\nthe rapid development of tools for protein structure design and prediction. 511a , 511b While many DNA nanomachines were demonstrated, fewer RNA nanomachines\nwere reported. The reasons for this might lie in the chemical instability\nof RNA structures in vitro and the limited available\ntools for designing complicated RNA nanostructures. However, RNA assemblies\nmay possess advantages over their DNA counterparts because of the\nhigher structural and functional diversity of RNA. Moreover, RNA could\nfold co-transcriptionally in vitro and in\nvivo . It would thus be appealing to develop RNA nanomachines\nthat can form and function in vivo for applications\nsuch as gene regulation and signal transduction. When considering\nthe performance and efficacy of DNA-based drug\ndelivery systems, the behaviors of DNA origami structures in complex\nbiochemical environments are still not fully understood. In recent\nyears, several groups reported that PEGylated oligolysine could enhance\nthe stability of DNA origami in physiological conditions. 512 − 514 For instance, polymer-covered DNA nanostructures exhibited remarkable\nnuclease resistance and prolonged circulating half-life in\nvivo . 515 In addition, photo-cross-linking\ncould improve the stability of DNA nanostructures that contain acrydite\nmodification 516 or thymines in close proximity. 517 Despite a series of careful studies, 518 − 523 the pharmacokinetics, biodistribution, and clearance mechanisms\nof DNA origami structures in vivo as well as their\ncellular uptake and trafficking, all of which are crucial for biomedical\napplications, need to be better characterized. In particular, the\ninteractions between DNA origami with macromolecules in biological\nenvironments remain a relatively underexplored area. As an\nexciting new frontier in DNA nanotechnology, DNA-engineered\nmembrane materials have the potential to transform basic research\nin membrane biology, synthetic biology, and nanomedicine. To push\nthis frontier, researchers could devote endeavors to the following\nareas. First, the success of current membrane manipulation methods\nhinges on selecting suitable membrane anchors for DNA nanodevices.\nAlthough DNA oligonucleotides with certain hydrophobic modifications\n(e.g., cholesterols 410 ) are readily available\nfrom commercial sources, many other useful DNA modifications (e.g.,\nalkyl chains 415 ) require in-house DNA synthesis\nor conjugation reactions. As recent studies pointed out, 400 , 411 subtle differences in chemical structure could profoundly impact\nthe membrane anchor’s ability to recruit and insert into membranes.\nTherefore, the precision and versatility of the DNA-based membrane-engineering\ndevices could be vastly enhanced with a library of chemically modified\nDNA oligonucleotides with a spectrum of size and hydrophobicity. Second,\nwhile the existing DNA-based methods focus on controlling the geometry\nand chemical modification of membranes (including the membrane-integrated\nstructures, e.g., nanopores), very few DNA nanodevices have been developed\nto directly control or respond to the mechanical and electrical properties\nof the membranes. 524 Because membrane tension\nand potential play vital roles in cell signaling and metabolism, the\nability to modulate and measure these properties in biological and\nsynthetic systems is highly desirable to fully understand the molecular\ndeterminants of cellular processes. Third, the complexity of cell\nmembranes, such as the leaflet asymmetry, underlying cytoskeleton,\nas well as various peripheral and integral membrane proteins, presents\nboth challenges and opportunities to the next generation of DNA nanodevices\nthat act on living cells. 525 So far, DNA\nnanostructures have been built to deliver molecular payloads to cells, 341 , 526 detect cellular contraction forces, 527 activate cell surface receptors, 4 , 329 , 528 puncture cell membranes, 529 and cluster cells. 530 In the future,\nit is conceivable that DNA nanodevices could extract patches of cell\nmembranes, dynamically remodel the cell surface landscape, as well\nas communicate intracellular signals among cells. Fourth, it would\nbe possible to scale up the complexity of the membrane engineering\ntools by building DNA devices that work collaboratively to accomplish\nmultistep tasks on the membranes, such as sorting, tagging, packaging,\nand sending membrane-associated molecular cargos, or by organizing\nmultiple DNA-engineered membrane compartments into interconnected\nchemical reactors, 531 similar to a network\nof neurons passing biochemical information to one another. Li et al.\nshowed that transmembrane nanopores (∼7 nm wide) could be coupled\nto micrometer-long DNA channels, which facilitated the translocation\nof molecules as small as 0.8 nm in radius with minimal leakage, providing\na promising method for direct, long-range intercell communication. 532 Fifth, the DNA-enabled membrane-engineering\ntechniques may find applications in biomedicine, such as formulating\ntherapeutic exosomes or lipid NPs, as well as in DNA computing, like\ncompartmentalizing data storage and computing units for more stable\nand controllable operations. From the perspective of nanophotonics,\nDNA origami has solidified\nits role in this field in the past years, but challenges and exciting\nopportunities still lie ahead. First, 3D nanofabrication with arbitrary\ngeometries is one of the extraordinary capabilities of DNA origami.\nHowever, to achieve considerable scattering or absorption cross-sections\nfor generating pronounced plasmonic effects, large metallic NPs are\nneeded. To date, spherical AuNPs up to 150 nm in diameter 533 and anisotropic NPs, such as AuNRs up to 50\nnm in length, 484 have been successfully\napplied in DNA origami-templated nanophotonic structures. Functionalizing\nlarger metallic NPs and stabilizing them in salty environments that\nDNA origami structures prefer are tricky. New chemical functionalization\nstrategies for stable conjugation of large NPs with DNA origami are\nthus desirable. Second, to build complex nanophotonic architectures,\nsufficiently large origami templates are required, for instance, to\nhost big NPs or even to form lattices with long-range order. The dimensions\nof origami structures folded from single scaffolds are restricted\nby the lengths of the scaffolds. Even though longer scaffolds can\nbe used, this does not essentially solve the problem of scaling up\nDNA origami. Large DNA origami can be created by the so-called superorigami\napproach, 2 , 534 which utilizes a long scaffolding strand\nto link multiple origami components together. Nevertheless, the superorigami\nstill has the issues of restricted geometry, symmetry, and size. Alternatively,\nlarge origami superstructures can be created by polymerizing origami\nmonomers through blunt-end, 74 , 76 , 535 , 536 sticky-end, 368 , 537 , 538 or shape-complementary interactions. 59 Third, in many optical applications, the optical\ncomponents in the supercell of a lattice should not be regularly reiterated.\nFor instance, optical metasurfaces, a rapidly developing research\nsubject in nanophotonics, comprise arrays of antennas that can modulate\nthe amplitude, phase, and polarization of the incident light. In a\nsupercell of a metasurface based on the Pancharatnam–Berry\n(PB) phase, the optical antennas (e.g., AuNRs) are arranged with different\norientations to generate a complete phase coverage from 0 to 2π\nfor light modulation. Therefore, the individual origami components\nin the supercell should have high specificity to position the AuNRs\nwith designated orientations. In addition, periodic patterning of\nthese origami supercells in a lattice with sufficient size on the\norder of tens of micrometers is required for experimental characterization\nand good optical performance. Inspiring design strategies are highly\nencouraged to develop for such technical challenges. 539 The efforts along this line will be very rewarding, because\nDNA origami-templated optical metasurfaces could possess unprecedented\nproperties, including addressability, programmability, and reconfigurability\non the single-antenna level, which cannot yet be offered by other\nnanotechniques at visible frequencies. Fourth, dynamic reconfiguration\nis another distinct advantage of\nDNA origami. A multitude of external inputs to dynamically reconfigure\nthe assembled nanostructures have been demonstrated, including temperature\ncontrol, 481 pH changes, 540 ion concentration, 59 light, 541 DNA, 206 RNA, 479 aptamers, 480 enzymes,\nproteins, 542 and small molecules, 543 , 544 among others. 545 − 547 A common limitation of the dynamic DNA structures\nwas the reconfiguration rate, typically below 1 nm/min. A remarkable\nDNA-based AuNP motor powered by chemical fuels was demonstrated by\nthe Salaita group. 548 It could processively\ntranslocate on a functionalized flat surface\nat an average speed of 50 nm/s. In addition, the Dietz and Simmel\ngroups reported an electrically driven DNA origami rotary ratchet\nmotor with a rotational speed of up to 250 rpm, 231 which approached already the speed of natural molecular\nrotors, ATP synthase. 549 These exciting\nexamples open the pathway to realizing dynamic nanophotonic devices\nwith high speed and excellent performance. Fifth, although optical\nelements of different materials have been successfully functionalized\non DNA origami, AuNPs and AgNPs remain the major choices for optical\nantennas. 550 Other metals, such as palladium,\nplatinum, magnesium (Mg), aluminum (Al), and their alloys with noble\nmetals, exhibit interesting catalytic properties, yet their optical\nresponses remained vastly unexplored, when templated on DNA origami.\nParticularly, Mg and Al are also excellent candidates for ultraviolet\n(UV) plasmonics, as their permittivity becomes negative in the UV\nrange. Another important aspect is that metals are very lossy due\nto their high absorption in the visible range. All-dielectric NPs,\nsuch as Si (silicon) and TiO 2 (titanium dioxide), that\nexhibit Mie resonances with low losses have so far been overlooked\nfor DNA-guided assembly. Protocols for functionalizing all-dielectric\nNPs on DNA origami will be a promising solution to creating low-loss\nnanophotonic devices at visible frequencies. Sixth, long-term\nhigh performance without degradation is the fundamental\nprerequisite for practical optical devices. This property is still\ncritical for DNA-assembled nanomaterials, which generally degrade\nover time. Nguyen et al. suggested coating DNA origami with an ultrathin\nlayer of silica to prevent the origami from degradation and aggregation. 264 However, the resulting DNA origami post silica-coating\nwould lose most of the modification and functionalization possibilities.\nStrategies to enhance the rigidity and stability of DNA origami and\nmaintain all the DNA functions are worthy of further investigation.\nOne further aspect is the stability of the optical elements, such\nas metallic NPs and quantum emitters. This investigation is vital\nnot only for DNA nanotechnology but also for self-assembly in general.\nSeventh, the combination of DNA self-assembly and top-down techniques\nis a powerful route to constructing a new generation of nanophotonic\narchitectures with advanced optical properties. One of the significant\nchallenges is to accurately control the positions and orientations\nof the DNA origami-assembled nanostructures during their immobilization\non a substrate. The Gopinath, Cha, and Rothemund groups are the pioneers\nwho brought a remarkable leap forward in exploring DNA origami patterning\non solid supports. 493 , 551 − 553 To achieve nanophotonic devices with designated long-range order,\nfor instance, the aforementioned dynamic optical metasurfaces based\non the PB phase, innovative protocols have to be developed to enable\nthe correct positioning of each antenna within a lattice and meanwhile\nachieve the dynamic control of each antenna orientation in a fully\nindependent fashion. Taken together, since the original proposal\nof Nadrian Seeman,\nthe field of DNA nanotechnology has flourished on a scale that was\nunimaginable 40 years ago. The invention of DNA origami by Paul Rothemund\nhas further pushed this field to a new horizon and fostered a plethora\nof concepts, models, methodologies, and applications that were not\nthought of before. In particular, DNA origami-engineered nanomaterials\nhave brought exciting and vastly unexplored research avenues in materials\nscience, greatly enriching the portfolio of DNA-based applications.\nUndoubtedly, the synergetic efforts and collaborations among scientists\nwith different research backgrounds will continue to bring innovations\nto this field in the next decade. There shall always be plenty of\nroom at the bottom to explore, and we will certainly have much to\ncelebrate at the 50th anniversary of DNA nanotechnology.",
"introduction": "1 Introduction Molecular self-assembly\nplays a fundamental role in the structural\ncomplexity and functionality of biological systems. Nature evolves\nsophisticated ways to self-assemble information-carrying materials\ninto well-organized cellular architectures from the nanoscale to the\nmicroscale. In particular, the cell can be viewed as a biological\nfactory containing various molecular machines that work in concert.\nThe ultimate goal of synthetic biology is to build biological mimics\nthat can complete individual tasks as well as support artificial signaling\nand communication in a fully controllable manner. In comparison to\nproteins or other biomolecules, DNA is ideally suited for constructing\nbiological mimics or, more broadly speaking, functional structures\nand materials by molecular self-assembly due to many reasons. First,\nDNA has a well-defined B-form structure: a right-handed double helix\nformed by two complementary single strands with 10.5 bases per helical\nturn and about 2 nm in diameter. Second, DNA strands undergo predictable\ninteractions. Single-stranded DNA (ssDNA) hybridizes into a double\nhelix following the Watson–Crick base pairing, where A pairs\nwith T and G pairs with C. Third, facile synthesis and chemical stability\nof DNA render many practical applications possible. Fourth, DNA can\nbe readily modified and functionalized with a variety of nanoscale\nentities that possess interesting biological, chemical, magnetic,\nelectrical, or optical properties. Fifth, DNA self-assembly is a highly\nparallel bottom-up fabrication method with spatial accuracy and resolution\nat the nanoscale. Sixth, dynamic DNA structures exhibit excellent\nspatiotemporal responses to a multitude of external stimuli with the\ninherent sequence specificity, programmability, and addressability\nof DNA. The idea of using nucleic acids, the primary genetic\nmaterial in\ncells, as building blocks for the construction of functional structures\nand materials was conceived by Nadrian Seeman in 1982. 1 In the past four decades, the field of DNA nanotechnology\nhas made significant advances, from assemblies of small DNA motifs\nto giant DNA superstructures of gigadalton molecular weight, from\nstatic to complex dynamic structures in response to environmental\nfactors, and from unmodified DNA/RNA nanostructures to functional\nconstructs for a wide range of applications in biomedical engineering,\ndrug delivery and therapy, nanophotonics and electronics, energy harvesting\nand transfer, biochemical sensing, super-resolution imaging, nanomachinery,\nbiomimetics, and synthetic cells. A plethora of nucleic acid self-assembly\napproaches have been developed, enabling the creation of complex DNA\nassemblies and novel hybrid nanomaterials from one dimension (1D)\nto two and three dimensions (2D and 3D). Meanwhile, the nucleic acid\nself-assembly mechanisms have also been better elucidated, facilitating\nthe design, prediction, simulation, and optimization of diverse DNA-based\nsystems. This review focuses on the most recent advances in\nDNA-origami-based\nnanomaterials, including DNA assemblies, superstructures, nanodevices,\nfunctional hybrid systems, and many others in the recent five years.\nPrevious important achievements have been discussed in the reviews\nby Hong et al. 2 and others. 3 , 4 Our review is structured as follows (see Figure 1 ). We start with a brief overview of the\nmilestones in DNA self-assembly, followed by a comprehensive presentation\nof the recent breakthroughs in self-assembly methodologies and advanced\nDNA architectures. Subsequently, we recapitulate the working mechanisms\nand experimental realizations of DNA origami-templated functional\nnanomaterials, ranging from inorganic nanoclusters to biologically\nrelevant molecules. Furthermore, we summarize the recent progress\nin bridging DNA nanotechnology with other research fields, such as\ndrug delivery and nanomedicine, membrane biology, and nanophotonics.\nFinally, we finish this review with conclusions and an outlook to\nhighlight the future avenues and promising directions in this vigorous,\nmultidisciplinary field."
} | 5,401 |
36568361 | PMC9779932 | pmc | 4,319 | {
"abstract": "Horizontal gene transfer mediated by conjugation is considered an important evolutionary mechanism of bacteria. It allows organisms to quickly evolve new phenotypic properties including antimicrobial resistance (AMR) and virulence. The frequency of conjugation-mediated cargo gene exchange has not yet been comprehensively studied within and between bacterial taxa. We developed a frequency-based network of genus-genus conjugation features and candidate cargo genes from whole-genome sequence data of over 180,000 bacterial genomes, representing 1,345 genera. Using our method, which we refer to as ggMOB, we revealed that over half of the bacterial genomes contained one or more known conjugation features that matched exactly to at least one other genome. Moreover, the proportion of genomes containing these conjugation features varied substantially by genus and conjugation feature. These results and the genus-level network structure can be viewed interactively in the ggMOB interface, which allows for user-defined filtering of conjugation features and candidate cargo genes. Using the network data, we observed that the ratio of AMR gene representation in conjugative versus non-conjugative genomes exceeded 5:1, confirming that conjugation is a critical force for AMR spread across genera. Finally, we demonstrated that clustering genomes by conjugation profile sometimes correlated well with classical phylogenetic structuring; but that in some cases the clustering was highly discordant, suggesting that the importance of the accessory genome in driving bacterial evolution may be highly variable across both time and taxonomy. These results can advance scientific understanding of bacterial evolution, and can be used as a starting point for probing genus-genus gene exchange within complex microbial communities that include unculturable bacteria. ggMOB is publicly available under the GNU licence at https://ruiz-hci-lab.github.io/ggMOB/",
"introduction": "1 Introduction Several mechanisms of horizontal gene transfer (HGT) allow bacteria to exchange genetic material. One of these mechanisms, termed conjugation, occurs when bacterial cells form direct physical contacts that allow for passage of genetic material from one bacterium to another. The machinery required to form these contacts and initiate genetic exchange is often contained within integrative and conjugative elements (ICE), plasmids, and other mobile genetic elements (MGEs) ( Frost et al., 2005 ; Wozniak and Waldor, 2010 ; Roberts and Mullany, 2011 ; Wiedenbeck and Cohan, 2011 ; Perry and Wright, 2013 ; Johnson and Grossman, 2015 ; Singer et al., 2016 ). The conditions that induce excision and conjugation are not fully elucidated, but DNA damage and subsequent SOS response seem to be an important trigger ( Waldor et al., 2004 ; Koraimann and Wagner, 2014 ). The cost of acquiring and maintaining the new genetic material also influences the success of transfer events ( Uhlemann et al., 2021 ). Genes exchanged between bacteria during conjugation include functional domains associated with conjugative machinery (e.g., excisionases, integrases, conjugative transport proteins) as well as intervening “accessory” genes that are not necessary for conjugation, often termed “cargo genes” ( Johnson and Grossman, 2015 ). By pairing conjugative machinery with an array of diverse cargo genes, bacterial communities can significantly expand their genetic repertoire, including between bacteria of diverse taxonomy ( Guglielmini et al., 2011 ; Bellanger et al., 2014 ; Neil and Allard, 2021 ). Functions commonly associated with conjugative cargo include antimicrobial resistance (AMR) and virulence ( Roberts and Mullany, 2011 ; Perry and Wright, 2013 ; Johnson and Grossman, 2015 ; Cury et al., 2017 ), which can pose a risk to human and animal health if transferred into pathogens ( Partridge et al., 2018 ). Therefore, understanding the microbial ecology of conjugative elements and cargo genes (i.e., their distribution and behavior across bacterial taxa) is important in assessing the risk posed by various bacterial communities ( Gaston et al., 2021 ). For example, how often do different bacterial taxa carry conjugative machinery and AMR genes; what resistance phenotypes are commonly associated with the presence of conjugative machinery within the genome; how often do different commensal bacterial taxa carry out conjugation to exchange cargo genes with pathogens; and what conditions foster conjugative exchange of specific cargo genes between pathogens and non-pathogens? These questions are fundamental to understanding how bacterial communities respond to external stimuli, and how these responses increase the overall risk posed by microbial communities of varying composition ( Martínez et al., 2015 ; Oh et al., 2018 ). However, the process of conjugative HGT is highly stochastic and therefore, difficult to predict ( Lopatkin and Collin, 2020 ). One reason for this stochasticity is variability in the conjugation competency of donor and recipient bacterial cells for a given conjugative MGE; as well as variability in the capacity of a given type of conjugative MGE to also transfer unrelated cargo genes. Recent meta-analyses of conjugation rates for specific bacterial species and/or MGEs have highlighted these complexities ( Alderliesten et al., 2020 ; Sheppard et al., 2020 ). Historically, the scientific process for estimating conjugative likelihood has stemmed from highly controlled in vitro experiments between pairs of bacterial isolates and specific MGEs. Results from such studies have been crucial for uncovering the behavior of MGEs and their importance for functions such as AMR. However, reductive experiments typically do not generalize well to the complex microbial communities found in situ , including host and environmental microbiomes. Furthermore, these experiments are necessarily restricted in their ability to characterize the full bacterial host range of a given MGE, as they typically involve only several distinct bacterial taxa. One major challenge that remains is to generate a conjugation likelihood for every host-donor-MGE combination observed across all bacterial taxa and MGE. Insight into this challenge can be gained through the plethora of whole genome sequence (WGS) data which is now publicly available. As an example, the analysis of HGT-associated genes from just 336 genomes across 16 phyla was sufficient to significantly improve bacterial phylogenies as compared to those obtained from conserved marker genes ( Abby et al., 2012 ). An analysis of 1,000 genomes demonstrated that ICE machinery is ubiquitous across diverse prokaryotes, and likely one of the most common mechanisms of bacterial evolution ( Guglielmini et al., 2011 ). Currently, public datasets contain orders of magnitude more WGS data, which can be used to improve our understanding of the mechanisms by which critically important genes and pathogens emerge and persist ( Botelho et al., 2020 ). However, despite the importance of HGT in bacterial evolution and pathogenicity, there has not yet been a comprehensive, systematic survey of the frequency of conjugation and cargo genes within or between bacterial genera. The objective of this work was to describe intra- and inter-genus conjugation-cargo dynamics by leveraging the comprehensive set of WGS data and conjugation sequences currently available within the Reference Sequence (RefSeq) and Short Read Archive (SRA) databases at the National Center for Biotechnology Information (NCBI). In particular, we analyzed 186,887 WGS datasets to identify putative conjugation events and corresponding candidate cargo genes, as well as to characterize the frequency of AMR genes with respect to the frequency of their occurrence with conjugative proteins. We were able to identify over 95,000 genomes containing conjugative proteins, and more than 4 billion putative cargo genes between genomes. We summarize and disseminate this analysis through an open-source network that describes the genus-genus sharing of conjugation features and cargo genes, representing genomes from over 1,300 different genera. Our network, which we refer to as ggMOB, allows users to filter for both conjugation features and putative cargo genes. Using ggMOB we analyzed the ratio of AMR gene representation in conjugative versus non-conjugative genomes and found it to be greater than 5:1, confirming that conjugation is a critical force for AMR spread across genera. Finally, we demonstrated that clustering genomes by conjugation profile sometimes correlated well with taxonomic structuring, but in some cases was highly discordant, suggesting that the importance of the accessory genome in driving bacterial evolution may be highly variable across genera. These results demonstrates that ggMOB can be used to further probe potential genus-genus mobilization dynamics, and thus, provide insight into conjugative mobilization between unculturable bacteria and complex interactions involving multiple genera.",
"discussion": "3 Discussion The public availability of large scale genomic data makes it possible to apply cloud computing technology and big data techniques to study important phenomena in molecular and microbiology. Curating these data in a relational database with biologically structured entity relations (i.e., linking genomes, genes, proteins, domains, and metadata) provides a powerful method with which to ask biological questions about the data. We leveraged this approach in the current study of cargo and conjugation, which is an essential mechanism by which bacteria acquire new phenotypes, transmit molecular functions, and adapt to stress. Furthermore, these events are critical for understanding bacterial evolution and phylogeny ( Guglielmini et al., 2011 ; Abby et al., 2012 ; Bellanger et al., 2014 ). Our work not only sheds light on conjugation-mediated cargo transfers between and within genera, but also demonstrates the ability of mining and analyzing large datasets in improving our understanding of bacterial evolutionary dynamics. The network of putative genus-genus conjugation features and candidate cargo genes can be dynamically visualized using the ggMOB tool, which supports hypothesis generation and testing related to intra- and inter-genus conjugation dynamics. In our analysis we identified sets of proteins with the strongest evidence as conjugation and cargo proteins. This was accomplished by selecting only those proteins that exhibited both 100% sequence identity and co-occurrence in pairs of genomes containing identical conjugation-associated sequences. With this strict selection process, the putative cargo proteins exhibited a high degree of spatial correlation within assembled contigs (i.e., they were highly adjacent to each other, as well as to the conjugation protein itself). Other proteins in these genomes may also have been transferred (or are transferable) by bacterial conjugation, but they did not meet our strict selection criteria. Considering only strictly-selected candidate cargo proteins, we were able to profile the frequency of conjugation-mediated protein exchange within and between genera. Our results suggest that conjugation-mediated exchange is not uncommon, affirming prior studies ( Guglielmini et al., 2011 ; Bellanger et al., 2014 ). Conjugation-related proteins were observable in 51% of bacterial genomes and in 631 of 1,345 genera (approximately 47%). Frequency of intra- and inter-genus conjugation-mediated exchange varied significantly depending on the taxa involved, suggesting that taxonomy greatly influences genetic exchange of, e.g., AMR or pathogenicity proteins ( Delavat et al., 2017 ). By quantifying this across a large database of high-quality WGS data, we measured the “exchange likelihood” between different genera. These likelihoods can be visualized dynamically in the ggMOB tool, which reveals distinct clusters of genera that share conjugative features with exact sequence match. This suggests that the likelihood of protein transfer varies substantially by genus pair, and that the bacterial composition within a given environment is an important consideration when attempting to evaluate mobilization potential within a microbial community ( Lopatkin and Collin, 2020 ; Neil and Allard, 2021 ). While we have conducted this analysis for a specific set of conjugation features ( Table 1 ), the analytic approach can be applied to any MGE(s) and cargo protein(s) of interest. As such, our overall approach represents a method for obtaining a long-range evolutionary view of transfer likelihood between diverse bacterial taxa, including pathogens and commensal bacteria ( Guglielmini et al., 2011 ). These baseline exchange likelihoods are critical parameters for risk analysis at the microbial community level, including for applications such as personalized microbiome medicine, and microbiome-centric surveillance. Bacterial taxon is not the only significant driver of exchange likelihood; we also observed that putative, successful transfer events were more likely to involve cargo proteins that infer fitness advantage to the involved bacterial populations, such as AMR. While any gene can, in principle, be transmitted as a cargo gene in conjugative exchange, only a subset of transferred proteins will increase the fitness of the receiving organism. The likelihood of observing successful transfer depends on a large number of factors including the environment, the existing proteins in the recipient chromosome, the cargo proteins themselves, and the survival probability of the organism ( Cohen et al., 2010 ). Conjugation-mediated protein transfer that improves fitness may increase survival probability. Therefore, chromosomal arrangements that group fitness-conferring cargo proteins near the conjugation machinery will be observed more frequently than those arrangements that involve neutral or disadvantageous proteins. Conversely, very common proteins that aid in stress response may be less likely to be transferred as cargo, since the relative fitness advantage is diminished for proteins that are already likely to be present within a bacterium (i.e., proteins that confer redundant function). The particular stressor—as well as the specific advantageous proteins of interest—depend on phenotype of interest. This view is exemplified by the data in Table 4 which shows that rare AMR proteins are more likely to be found as cargo in genomes that also contain conjugative proteins, as compared to genomes that do not. Conversely, common AMR proteins are less likely to be found in genomes that contain conjugative machinery. One might hypothesize that, with chromosomal rearrangement, nature effects a real world experiment to dynamically optimize cargo protein collections—thereby spreading rare (but useful) proteins and gene combinations over time. The particular cargo proteins shared between chromosomes varied by conjugation feature, as demonstrated by the AMR proteins analyzed in Figures 1 , 4 . Considering all conjugation features used in this study, our results suggest that conjugation dynamics are important in structuring genomic content, and thus driving phylogenetic evolution. Based on Figure 5 , it seems that sometimes these evolutionary conjugation dynamics can sometimes overpower other taxonomic drivers, such that genus-level genomes do not always cluster together. To demonstrate this conjugation-driven phylogeny, we used the data in Figure 5 to generate Figure 2 , which represents the distance between all pairs of genomes based on Euclidean distance between their representations as normalized conjugation feature vectors. The resulting hierarchical clustering shows that the dominant conjugation features are represented in genomes across different genera and, conversely, that individual genera include genomes with differing conjugation profiles. This abrogation of genus-level taxonomy due to conjugation-related genomic content is an inevitable consequence of the inter-genus transfers visualized in ggMOB. Given the reality of conjugative exchange, there is no reason to expect that taxonomic classification by organism name will always predict the composition of conjugation-associated cargo proteins. However, by selecting genomes based on a particular phenotype of interest, it is possible to classify organisms and genome-genome distances based on a feature space defined by conjugation (or other mobilization) proteins, as in Figure 2 . Given the ubiquity and diversity of conjugation and other types of HGT ( Guglielmini et al., 2011 ), these types of genome clustering techniques may provide crucial information about bacterial evolution that is not contained within traditional phylogenies. In this regard, the ability to filter the ggMOB data based on conjugation features of interest may particularly useful. FIGURE 1 Individual genomes typically contained more than one conjugative feature ( Table 1 ), and often contained more than one protein per feature. (A) Histogram (log frequency) of the number of conjugative features per genome, and (B) Histogram (log frequency) of the number of conjugative proteins per genome for all 106,433 genomes containing at least one conjugative feature. FIGURE 2 Heatmap showing the relative genomic positions of conjugative features and putative cargo proteins for the 2,000 genomes with the greatest number of cargo proteins. Conjugative features are represented as color pixels based on the colors shown in Table 1 , with yellow representing proteins assigned specific IPR codes and red representing protein families from the literature. Cargo proteins are shown in grey and other chromosomal DNA in white. Each genome is bit shifted to the left until the first conjugative feature is centered in the figure. Most genomes contained more than one conjugative protein (all contain at least one). The inset highlights the genomes at indices 1450–1550 in order to expand a subset of the data."
} | 4,512 |
36952444 | PMC10035835 | pmc | 4,322 | {
"abstract": "Giant kelp and bull kelp forests are increasingly at risk from marine heatwave events, herbivore outbreaks, and the loss or alterations in the behavior of key herbivore predators. The dynamic floating canopy of these kelps is well-suited to study via satellite imagery, which provides high temporal and spatial resolution data of floating kelp canopy across the western United States and Mexico. However, the size and complexity of the satellite image dataset has made ecological analysis difficult for scientists and managers. To increase accessibility of this rich dataset, we created Kelpwatch, a web-based visualization and analysis tool. This tool allows researchers and managers to quantify kelp forest change in response to disturbances, assess historical trends, and allow for effective and actionable kelp forest management. Here, we demonstrate how Kelpwatch can be used to analyze long-term trends in kelp canopy across regions, quantify spatial variability in the response to and recovery from the 2014 to 2016 marine heatwave events, and provide a local analysis of kelp canopy status around the Monterey Peninsula, California. We found that 18.6% of regional sites displayed a significant trend in kelp canopy area over the past 38 years and that there was a latitudinal response to heatwave events for each kelp species. The recovery from heatwave events was more variable across space, with some local areas like Bahía Tortugas in Baja California Sur showing high recovery while kelp canopies around the Monterey Peninsula continued a slow decline and patchy recovery compared to the rest of the Central California region. Kelpwatch provides near real time spatial data and analysis support and makes complex earth observation data actionable for scientists and managers, which can help identify areas for research, monitoring, and management efforts.",
"conclusion": "Conclusions Over the past decade, there has been an increased focus on the long-term declines of kelp forests both regionally and globally, usually in the context of warming ocean conditions, competition with other reef space holders, and increases in herbivore abundance [ 67 – 70 ]. While kelp forests in many regions have undoubtedly experienced severe and unprecedented declines in recent years [ 11 , 71 ], time series of kelp dynamics are often limited by short durations or punctuated field campaigns. These time series limitations can obscure the true nature of kelp forest change especially given the rapid dynamics of kelp. For example, a recently observed decline may be related to a decadal marine climate oscillation and similar periods of low kelp abundance may have occurred before the time series was initiated or were missed due to logistical or funding constraints. Therefore, it is essential to produce a long, continuous, and calibrated time series in order to put recent declines in the context of long-term dynamics. While Landsat observations can be limited by cloud cover and can only detect fluctuations in surface canopy, the uninterrupted satellite continuity (1984 to present), rapid repeat frequency (16 days from 1984 to 1998; 8 days from 1999 to present), and large spatial domain (global) offer an unparalleled opportunity to track kelp forest dynamics [ 18 , 72 ]. The assessment of continuous kelp dynamics allows for the observation of decadal cycles in canopy cover that often result from changes in regional nutrient regimes (e.g., the North Pacific Gyre Oscillation; [ 6 , 40 , 53 ]) or sudden regional-scale crashes and recoveries in canopy cover resulting from El Niño and La Niña events, respectively [ 21 ]. Recently, an ensemble of climate models was used to determine the appropriate time series length needed to distinguish a climate change precipitated trend from natural variability for several biogeochemically relevant marine variables and found that time series of at least 40 years in length are necessary to define the natural variability of biotic variables (phytoplankton chlorophyll concentration and production dynamics; [ 56 ]). With close to 40 years of observations as of the time of this analysis, the Landsat-derived data available on Kelpwatch are beginning to approach the length necessary to observe changes in kelp canopy across decadal cycles and detect long-term trends. Tools like Kelpwatch make earth observation data actionable and will help scientists and managers identify areas to focus research and monitoring efforts to understand how kelp forests respond to marine heatwaves and other pressures, and to place these dynamics in historical context to inform strategic management interventions. The analyses in this study are descriptive in nature and further work by researchers and kelp forest managers is required to identify the drivers of kelp canopy response to and recovery from disturbance events such as the 2014 to 2016 heatwave period. The near coincident occurrence of high ocean temperatures, reduced seawater nutrient concentrations, and increased density of herbivorous sea urchins during the heatwave period necessitate additional spatial data streams to identify driver timing and strength [ 11 ]. Additionally, the recovery of kelp populations is subject to additional demographic (e.g., spore supply; [ 73 ]) and ecological processes such as hysteresis [ 74 ] that may cloud the recovery of population dynamics if only abiotic drivers are considered [ 10 ]. All the analyses in this study were completed using data downloaded directly from the Kelpwatch platform and analyzed in commonly used geospatial or statistical programs for transparency and reproducibility. Having a calibrated, open-access, and continuous time series of kelp canopy dynamics puts the ability to examine near real-time observations of kelp canopy and spot problem areas for kelp loss in the hands of scientists and managers.",
"introduction": "Introduction Along the west coast of North America, underwater forests of kelp provide the foundation for a productive and diverse nearshore ecosystem [ 1 ]. The dominant and iconic species of kelp in this region are giant kelp ( Macrocystis pyrifera ) and bull kelp ( Nereocystis luetkeana ), both of which create large, floating canopies. Both species have high rates of primary production [ 2 ] and create complex structure [ 3 ], thereby providing food and habitat for many ecologically and economically important species. However, the abundance of these kelp species fluctuates rapidly and is sensitive to environmental changes [ 4 ]. Stressors such as climate change, overgrazing, and coastal development have been linked to declines in kelp abundance [ 5 ] and there is high spatial variability in the response of kelp forests to changing environmental conditions [ 6 ]. From 2014 to 2016 the west coast of North America experienced a series of extreme marine heatwaves that had significant impacts to coastal marine ecosystems [ 7 , 8 ] and was the warmest three-year period on record for the California Current [ 9 ]. This heatwave period initially led to widespread declines in the abundance of giant and bull kelp [ 10 , 11 ], but the magnitude and duration of these impacts varied widely. In northern California, the combined effects of the heatwaves, the loss of an important sea urchin predator (sunflower sea stars) due to disease [ 12 ], and a subsequent explosion in sea urchin populations led to a collapse in bull kelp abundance, with devastating ecological and economic impacts [ 11 , 13 ]. However, despite the regional loss of sunflower sea stars [ 12 ], bull kelp populations in southern Oregon were relatively insensitive to the heatwave events [ 14 ]. Around the Monterey Peninsula in Central California, increased sea urchin abundance has reduced the once expansive giant kelp forests to a patchwork of urchin barrens and kelp stands that are maintained by sea otters (an important sea urchin predator) selectively feeding on healthy urchins within the remaining kelp areas [ 15 ]. In southern California and across the Baja California Peninsula there were widespread declines in giant kelp abundance immediately following the heatwave events, but recovery in subsequent years was spatially variable [ 10 ]. Frequent and widespread monitoring of kelp forests is crucial for understanding patterns and drivers of kelp forest trends and their response to disturbances, which is a key component of effective kelp forest ecosystem-based management [ 16 ]. Many species of kelp (including bull kelp and giant kelp) have populations that are highly variable through time [ 17 , 18 ]. Boom and bust cycles are common, and collapse of kelp forests can be sudden [ 13 , 19 ]. Kelp forest dynamics are also highly variable on small spatial scales (e.g., kilometers, [ 20 ]), which leads to high amounts of variability in patterns of recovery, even following widespread disturbance events such as continental-scale marine heatwaves [ 10 , 21 ]. Remote sensing is a powerful tool for monitoring canopy forming kelps such as bull kelp and giant kelp, and recent increases in the availability of airborne and spaceborne imagery is enabling regular monitoring across multiple space and time scales [ 5 , 14 , 22 ]. For example, inexpensive small unoccupied aerial systems (UAS) can provide very high-resolution monitoring of canopy extent at local scales [ 23 , 24 ], constellations of CubeSats can provide high-resolution data on regional scales [ 25 ], while moderate resolution satellites can be used to map kelp canopy dynamics at global scales [ 26 , 27 ]. The Landsat satellite program is particularly valuable for kelp monitoring, as it provides imagery with continuous global coverage at a 30 m resolution since 1984, and can be used to detect long-term trends in kelp canopy area, biomass, and abundance and put recent changes in a broader historical context [ 18 , 20 ]. Landsat imagery has been used to map both giant kelp and bull kelp canopy density and extent [ 14 , 18 , 28 – 30 ] and kelp abundance [ 18 , 20 ] on seasonal time scales from 1984 to present for the west coast of the United States and Baja California, Mexico [ 14 , 18 , 28 – 30 ] and other regions of the world [ 27 , 31 , 32 ]. One of the most valuable aspects of this dataset is its extensive spatial and temporal coverage, especially for distinguishing the impacts of climate change on kelp populations from other sources of variability [ 5 ]. However, the size of the dataset also makes it difficult to use, especially for those without extensive experience working with large geospatial datasets and more complicated file formats. This accessibility barrier has limited the use of the Landsat dataset for mapping and monitoring canopy forming kelps. To increase accessibility of Landsat imagery among researchers, management agencies, and the public, we created Kelpwatch.org, a visualization and analysis web tool that allows users to select a region, time frame, and season(s) of interest to interactively display changes in kelp canopy over time and freely download data. The primary objective of Kelpwatch.org is to make published kelp canopy data from Landsat imagery actionable for restoration practitioners and researchers, and promote data-driven resource management (e.g., targeted restoration efforts, adaptively managing kelp harvest leases, changing fisheries seasons or catch limits). Analogous web tools have demonstrated success in facilitating data-driven management of other foundational ecosystems by making earth observation data actionable (e.g., Global Forest Watch, Allen Coral Atlas, Global Mangrove Watch; [ 33 – 36 ]). Kelpwatch.org (hereafter referred to as Kelpwatch) provides a user-friendly interface to analyze and download seasonal kelp canopy observations at 30 m resolution for the west coast of North America from central Baja California, Mexico to the Washington-Oregon border since 1984. To demonstrate the types of analyses that can be completed using Kelpwatch, we used data downloaded directly from Kelpwatch to ask the following questions: (1) What were the regional trends in kelp canopy area over the past 38 years? (2) What were the spatial patterns of kelp canopy area in response to and recovery from the 2014 to 2016 marine heatwave events? and (3) Given the recent spatial alterations to kelp forests by sea urchins around the Monterey Peninsula, California [ 15 ], how do local-scale patterns in recent kelp canopy area in this subregion compare to historical data?",
"discussion": "Discussion Long-term trends in kelp canopy across regions Significant long-term trends in kelp canopy area were observed at the 10 x 10 km scale and across entire regions. Kelp canopy in Oregon was characterized by three short periods of high canopy area early and in the middle of the time series, followed by a longer period of moderate kelp canopy starting the mid-2000’s, resulting in a significant, overall regional decline from 1984 to 2021 ( Fig 2B ). Total kelp canopy area in Oregon was dominated by the dynamics on Orford, Blanco, and MacKenzie reefs ( Fig 2A ), a large rocky reef complex southwest of Cape Blanco totaling ~50 km 2 with depths ranging from 10 to 25 m [ 46 ]. Five of the nine 10 x 10 km cells in the Oregon region declined over the study period, the only region where a majority of cells showed significant long-term trends ( Table 1 ). While there are a paucity of studies examining subtidal kelp dynamics and environmental drivers in Oregon, Hamilton and others [ 14 ] used Landsat imagery to assess five of the region’s largest kelp forests from 1984 to 2018. Two of the five forests examined displayed long-term declines, including Orford Reef [ 14 ]. Kelp forests in Oregon have also been periodically assessed via aerial imagery by the Oregon Department of Fish and Wildlife with imagery collected in 1990, 1996 to 1999, and 2010 [ 46 – 50 ]. A change in the image collection methodology from color-infrared photography (1990’s) to digital multispectral imagery (2010) made intercomparison difficult as photographs from the 1990’s were delineated by hand while individual 1-meter pixels were classified as kelp canopy from the multispectral imagery [ 50 ]. Without robust calibration between sensors through time, as done with the Landsat sensors [ 18 ], long-term trend analysis is impossible. Kelp forests off the coast of Northern California have experienced historic lows from 2014 to 2021 [ 11 , 13 , 24 ], however, while 20% of the 10 x 10 km cells showed a negative trend, a significant regional decline was not observed in the full time series. Kelp canopy was highly dynamic across the region, with three periods of high multiyear kelp canopy area during the late 1980’s to early 1990’s, late 1990’s to early 2000’s, and late 2000’s to early 2010’s, with the two highest years occurring in 2008 and 2012 ( Fig 2C ). Despite the recent period of low or absent kelp canopy across the region from 2014 to 2021, high levels of canopy present in the years immediately preceding the marine heatwave (2008, 2012 to 2013) demonstrates the interannual oscillatory nature of the kelp canopy in this region. The Central California region did display a significant regional decline in canopy area across the time series ( Fig 2D ), driven by a large decrease in canopy area around the Monterey Peninsula since 2014 ( Fig 3B ; see Local Declines Around Monterey Peninsula ). High kelp canopy area early in the time series resulting from both high canopy density and large canopy extent likely contributed to this negative trend. The Central California region represents a transition zone between the two major canopy forming kelp species but is mostly dominated by giant kelp [ 1 ]. While over 35% of 10 x 10 km cells displayed negative trends ( Table 1 ), Central California also possessed two cells with positive trends, including the northernmost cell near Point Año Nuevo, which showed increasing post-heatwave canopy area. Southern California and Baja California Norte both showed a low percentage of cells with significant trends (~10%), with every Southern California cell with significant trends located on the offshore Channel Islands ( Fig 2A ). Both regions suffered major reductions in plant and stipe density and canopy area during the 1997/1998 El Niño event, reaching regional minimums in 1998 ( Fig 2E and 2F ; [ 51 , 52 ]). However, both regions responded positively to the 1999/2000 La Niña event and canopy area continued to increase regionally with positive North Pacific Gyre Oscillation index values and the associated elevated seawater nutrients [ 6 , 53 ] with regional maximums in 2005 for Southern California and 2009 for Baja California Norte ( Fig 2F and 2G ). Baja California Sur had the shortest regional time series due to a lack of available imagery at the beginning of the time series ( S1 Table ). While no regional trend was detected, the two positive long-term trends from individual 10 x 10 km cells should be treated with skepticism since missing data occurred during a period when canopy area was relatively high across regions. The relationship between regional kelp canopy dynamics and decadal marine climate oscillations [ 6 , 40 , 54 ] produce multiyear periods of high (or low) kelp canopy that make the identification of long-term trends difficult [ 18 ]. This interannual oscillatory nature of regional kelp canopy dynamics is apparent in the regional time series and may have resulted in greater then 80% of 10 x 10 km cells showing no significant long term trend ( Fig 2 ). A recent analysis has shown that the synchrony of giant kelp canopy is highly coherent with the North Pacific Gyre Oscillation on long time scales (4 to 10 years; [ 55 ]), meaning that sites within regions tend to increase and decrease similarly according to the fluctuations of the large-scale ocean climate. Since regular oscillatory patterns make the detection of long-term trends difficult [ 18 , 56 ], perhaps the most beneficial use for these data is to investigate the spatial heterogeneity of the response in kelp canopy to major climate events, such as the 2014 to 2016 marine heatwaves. Here, this type of analysis uncovered sub-regional (and potentially local-scale) variability in kelp canopy and could allow researchers to hone in on areas showing disparate patterns and elucidate underlying drivers. Kelp canopy response to and recovery from the 2014 to 2016 marine heatwave events During the summer of 2014, an unprecedented warm water temperature event spread across the Northeastern Pacific leading to negative impacts across both nearshore and pelagic ecosystems [ 43 , 57 , 58 ]. This marine heatwave, known as ‘The Blob’, was closely followed by a strong El Niño event in 2015 to 2016, contributing to an extended period of anonymously high ocean temperatures, low seawater nutrients, and low productivity throughout the region [ 59 ]. Notably, there were historic and widespread declines in kelp forest ecosystems associated with these events, both in Northern California [ 13 ] and Baja California [ 60 ], although ecosystem response was not constant across regions [ 61 ]. We found that kelp canopies across all regions declined in response to the marine heatwave period, but that these declines were significantly related to latitude. Cavanaugh and others [ 10 ] found that the resistance/response of giant kelp canopy across Southern and Baja California was associated with an absolute temperature threshold (23°C) and not a relative temperature anomaly. Since ocean temperature generally decreases in the California Current with increasing latitude [ 62 ], it is perhaps not surprising that the response of giant kelp canopy to the marine heatwave events was more negative in southern regions given that giant kelp responds strongly to temperatures over an absolute threshold. Interestingly, regions that are primarily composed of bull kelp also exhibited a similar latitudinal response separate from the one displayed by regions dominated by giant kelp ( Fig 4A ). This implies that each species may possess specific temperature thresholds for growth and mortality and in fact, recent laboratory experiments show that bull kelp blades maximize elongation rate at 11.9°C with precipitous declines at temperatures above 16°C [ 63 ]. While previous marine heatwave events have resulted in short-term declines in kelp abundance across regions, the recovery of the kelp canopy following the heatwave can be spatially variable and often occur at smaller spatial scales (meters to kilometers; [ 51 ]). While a significant positive relationship between canopy recovery and latitude was found for giant kelp, the relationship was more variable when compared to its response to the heatwave period, with examples of high recovery in all four giant kelp dominated regions ( Fig 4B ). This is a similar result to previous studies that found no clear relationship between large-scale high ocean temperatures and heatwave variables to kelp recovery [ 10 ]. One striking example of kelp canopy recovery is the kelp forest that surrounds Bahía Tortugas in Baja California Sur near the southern range limit for giant kelp in the Northern Hemisphere ( Fig 5B ). This kelp forest displayed high canopy area before the heatwave, a complete loss of canopy during the heatwave, and a complete recovery to greater canopy area in the years following. However, this kelp forest is surrounded by 10 x 10 km cells that displayed little recovery during the five years post-heatwave events ( Fig 3 ), implying a driver acting over a smaller spatial scale than a latitudinal temperature gradient. The coast of Baja California has a varied geometry leading to distinct sub-regional upwelling zones that are oriented parallel to the dominant wind direction [ 64 ]. The three upwelling zones located within the Baja California study domain exist at 31.5°N, 29°N, and 27°N and all correspond to cells with high recovery. The sub-regional nature of coastal upwelling, delivering cool, nutrient-rich seawater to the nearshore, may be vital to kelp forest recovery after heatwave events and future studies comparing kelp dynamics to localized upwelling are needed. Regions dominated by bull kelp showed a significant relationship between recovery and latitude, driven by little recovery in Northern California and varied recovery in Oregon ( Fig 4B ). While signs of kelp canopy recovery in Northern California did not begin until 2021, some sites in Oregon displayed increases in canopy area throughout and after the heatwave events. An example of this is Rogue Reef, where little canopy was present prior to the heatwave, small increases occurred during the heatwave, and a large canopy formed post-heatwave ( Fig 5A ). This incredible level of recovery versus the historical mean canopy (~480%) represents one of the few areas with increasing kelp canopy during the marine heatwave events. As there are fewer subtidal monitoring programs in Oregon compared to California [ 65 ], more work is needed to understand the spatial drivers of kelp forest dynamics across Oregon. Local declines around Monterey Peninsula While the Central California region exhibited relatively high levels of recovery to the marine heatwave events, the five 10 x 10 km cells surrounding the Monterey Peninsula showed less than 20% recovery compared to the historical mean ( Fig 3 ). Prior to the heatwave events, kelp canopy area around the Monterey Peninsula was seasonally dynamic, with large winter waves removing whole plants and/or canopy each year leading to a reduction in kelp abundance [ 6 , 66 ]. However, kelp canopies in this subregion were persistently high on annual time scales, with decreases during the heatwave events of 2014 to 2016 and with further reductions post-2016 ( Fig 6A ). This cluster of cells with low sustained recovery warranted a local-scale analysis made possible by altering the domain of the spatial input polygons uploaded to Kelpwatch. During the post-heatwave years (2014 to 2021) the vast majority of 1 x 1 km cells showed less than 50% of their mean historical canopy area (1984 to 2013) with only a few local-scale examples of high recovery ( Fig 6B ). An examination of the Landsat imagery used to generate the kelp canopy data for Kelpwatch further illustrates these declines. The kelp forests near Pescadero Point ( Fig 5c1 ), Carmel Point ( Fig 5c2 ), and Point Lobos ( Fig 5c3 ) can be clearly seen as the offshore red patches in the false color imagery during September 2011. By September 2016, the Pescadero Point kelp forest canopy (1) had nearly disappeared, Carmel Point (2) was similar in area to 2011, and Point Lobos (3) had been reduced to a few patches. By September 2021, the Pescadero Point kelp forest (1) was showing some patchy recovery, Carmel Point (2) had been reduced to patches and Point Lobos (3) had nearly disappeared. These spatial patterns exposed by Kelpwatch support a recent field-based analysis examining the role of sea otters, an important predator of herbivorous sea urchins. Smith and others [ 15 ] found that the spatial pattern of sea otter foraging was associated with the distribution of energetically profitable urchins, that is, restricted to areas that maintained high kelp densities and well-fed sea urchins. This resulted in a patchy mosaic of kelp forest stands interspersed with sea urchin barrens, possibly enhancing the resistance of existing stands but not directly contributing to the resilience of areas without kelp [ 15 ]. While this explains the spatial patchiness and lack of recovery of kelp canopy around the Monterey Peninsula, it does not explain why this subregion showed less recovery than other areas in Central California. The sustained decline in kelp canopy around the Monterey Peninsula detected using the Kelpwatch tool represents the longest period of low canopy cover for this area over the length of the Landsat time series, suggesting that more research and monitoring attention should be directed at this location. Understanding differences in environmental conditions and trophic interactions around the Monterey Peninsula and nearby locations that have exhibited high kelp canopy recovery may shed light on important drivers that are best assessed by instrumented moorings and diver-based survey methods. This case study demonstrates how a decision-support tool like Kelpwatch can be used to make complex data actionable for managers, restoration practitioners, and researchers, and promote data-driven resource management."
} | 6,640 |
21523460 | null | s2 | 4,323 | {
"abstract": "Thanks to the confluence of genome sequencing and bioinformatics, the number of metabolic databases has expanded from a handful in the mid-1990s to several thousand today. These databases lie within distinct families that have common ancestry and common attributes. The main families are the MetaCyc, KEGG, Reactome, Model SEED, and BiGG families. We survey these database families, as well as important individual metabolic databases, including multiple human metabolic databases. The MetaCyc family is described in particular detail. It contains well over 1,000 databases, including highly curated databases for Escherichia coli, Saccharomyces cerevisiae, Mus musculus, and Arabidopsis thaliana. These databases are available through a number of web sites that offer a range of software tools for querying and visualizing metabolic networks. These web sites also provide multiple tools for analysis of gene expression and metabolomics data, including visualization of those datasets on metabolic network diagrams and over-representation analysis of gene sets and metabolite sets."
} | 270 |
21523460 | null | s2 | 4,324 | {
"abstract": "Thanks to the confluence of genome sequencing and bioinformatics, the number of metabolic databases has expanded from a handful in the mid-1990s to several thousand today. These databases lie within distinct families that have common ancestry and common attributes. The main families are the MetaCyc, KEGG, Reactome, Model SEED, and BiGG families. We survey these database families, as well as important individual metabolic databases, including multiple human metabolic databases. The MetaCyc family is described in particular detail. It contains well over 1,000 databases, including highly curated databases for Escherichia coli, Saccharomyces cerevisiae, Mus musculus, and Arabidopsis thaliana. These databases are available through a number of web sites that offer a range of software tools for querying and visualizing metabolic networks. These web sites also provide multiple tools for analysis of gene expression and metabolomics data, including visualization of those datasets on metabolic network diagrams and over-representation analysis of gene sets and metabolite sets."
} | 270 |
21980383 | PMC3184130 | pmc | 4,325 | {
"abstract": "Springtails, arthropods who live in soil, in decaying material, and on plants, have adapted to demanding conditions by evolving extremely effective and robust anti-adhesive skin patterns. However, details of these unique properties and their structural basis are still unknown. Here we demonstrate that collembolan skin can resist wetting by many organic liquids and at elevated pressures. We show that the combination of bristles and a comb-like hexagonal or rhombic mesh of interconnected nanoscopic granules distinguish the skin of springtails from anti-adhesive plant surfaces. Furthermore, the negative overhang in the profile of the ridges and granules were revealed to be a highly effective, but as yet neglected, design principle of collembolan skin. We suggest an explanation for the non-wetting characteristics of surfaces consisting of such profiles irrespective of the chemical composition. Many valuable opportunities arise from the translation of the described comb-like patterns and overhanging profiles of collembolan skin into man-made surfaces that combine stability against wear and friction with superior non-wetting and anti-adhesive characteristics.",
"introduction": "Introduction Water-repellent and self-cleaning surfaces that protect plants in humid environments under high pathogen pressures have recently gained much interest. Those superhydrophobic plant surfaces result from hierarchically aligned structural elements, always including nanoscale wax crystals forming rather fragile and continuously regenerated structures which are often needle-like [1] – [7] . As a consequence, the contact area between the plant surface and liquids or particles is minimized by surface roughness and heterogeneous wetting [8] – [11] . Many recent efforts have tried to mimic key features of superhydrophobic plant surfaces in artificial materials and coatings, but the inherently low mechanical stability of the structures results in rather limited durability [5] , [7] . Anti-wetting phenomena are also known from some arthropods and their eggs [12] – [14] . Springtails (Collembola, Entognatha), a wingless arthropod group of more than 7000 species which live in soil, in decaying material, and on plants, have adapted to demanding environmental conditions by evolving extremely effective and robust anti-adhesive skin patterns. They are among the most abundant of all macroscopic animals and considered a separate evolutionary lineage that branched much earlier than the separation of crustaceans and insects [15] . Springtails often live in habitats where water is heavily contaminated by surface-active substances originating from decaying organic matter, and where potentially harmful microorganisms are present [12] . In consequence, they exhibit a very unusual skin structure that reflects an even more pronounced adaptation than that observed in plants [16] – [19] . While it has been previously recognized that the prevention of cuticle wetting is critically important for survival because springtails depend on epidermal respiration previous research efforts have mainly examined the mechanisms Collembola use for surviving drought, freezing and dispersal ability and only few investigations considered the repellent properties of collembolan skin [20] – [24] . Here we explore the structural elements ( \n Figure 1 \n \n and Figure S1 ) of Collembola skin which control interfacial phenomena. We demonstrate that the skin can resist wetting by many organic liquids and at elevated pressures and we suggest a general explanation for the non-wetting characteristics of the related structures. 10.1371/journal.pone.0025105.g001 Figure 1 Springtail skin combines bristles and a unique nanoscopic comb pattern. The rhombic or hexagonal comb pattern is formed by small primary granules connected by ridges. Additionally, some - but not all - species possess papillous secondary granules (SG), which can significantly differ in shape, depending on the specific habitat and body size of the respective species [for details see Information S1 ].",
"discussion": "Results and Discussion Scanning electron microscopy studies showed that the skin of springtails exhibits a hierarchical structure of nanoscopic interconnected granules (primary granules) combined with bristles or feathered hairs ( \n Figure 1 \n , for O. stachianus and T. bielanensis , for additional information on 35 different species from 16 families and 4 orders, comprising animals with quite different shape, size and habitat see Table S1 and Information S1 ). 18 out of 35 investigated species were found to possess microscopic, papillous granules (secondary granules). Bristles are tens of microns in length and their hinge-like base allows them to bend in all directions in response to mechanical forces. Distal bristle diameters are very small, ranging from 90 to 150 nm. Comparing the occurrence of secondary granules with the habitat of the related species suggests that these granules mechanically protect the integrity of the nanostructures (for details on the mechanical stability see Figures S2 , S3 and Information S1 ). The triangular and quadrangular primary granules of the skin have side lengths of about 200–300 nm and are connected by thinner bars. These connections produce a hexagonal or rhombic comb-like pattern of nanocavities, which covers the whole body of the springtails. The structure size can vary between animals of the same species in response to different environmental influences indicating the capability of the skin to undergo ecomorphological adaptation [25] , [26] . In particular, the occurrence of secondary granules was observed on animals living in the soil. Rhombic and hexagonal comb patterns can occur both on the skin of the same animal, whereby the rhombic pattern is present on segments requiring higher elasticity, e.g. for bending. To characterize the anti-wetting performance of collembolan skin in some detail, we applied different liquids and condensation experiments. Very stable plastrons (air cushions) were observed around Collembola upon forced immersion in bulk liquids and resist elevated pressures up to values higher than 3.5 atmospheres ( \n Figures 2 \n \n and Figure S4 ). This is in contrast, to the plastron preservation of most other arthropods which was reported to be clearly below two atmospheres [12] . Plastrons of springtails were found to persist for many days and occurred not only in water but also in many polar and non-polar liquids with much lower surface tensions ( \n Table 1 \n ). Importantly, the collembolan skin not only exhibits superhydrophobicity but similarly superoleophobic characteristics as demonstrated here by the resistance against wetting with a variety of organic liquids, even including tridecane. 10.1371/journal.pone.0025105.g002 Figure 2 Immersion and water condensation experiments. (A) (left) T. bielanensis in water, (right) Orthonychiurus stachianus immersed in ethanol resist wetting through the formation of a shiny air cushion. Results of immersion experiments with various liquids ( \n Table 1 \n ) revealed a resistance of the collembolan skin against wetting by non-polar liquids with surface tensions down to approximately 25 mJ/m 2 . No immersion occurred with any polar liquid. When exposed to increasing pressure, the plastron shrank and the shiny cover disappeared at pressures exceeding 3.5–4.0 bar. After the disappearance of the plastrons, the animals lost their buoyancy and sank. However, different from previously described superhydrophobic surfaces, the shiny plastron reappeared after pressure normalization if the time at reduced pressure did not exceed one minute. This suggests a reversible, pressure-dependent transition between the visible macroplastron and non-visible nanoplastrons enabled by the unique skin topography. (B) At elevated humidity condensation started on the skins of T. bielanensis , as a patchy droplet pattern with sizes of around 1 µm (ESEM image). Growing droplets fused or were absorbed by larger drops, leaving behind a completely non-wetted surface on which the described condensation process repeatedly occurred. Repeated droplet fusion finally led to the upward movement of larger drops to the structure tops, which is designated as anti-fogging . 10.1371/journal.pone.0025105.t001 Table 1 Results of immersion tests of three different species in polar and non-polar liquids. polar liquids skin wetting α [mJ/m 2 ] nonpolar liquids skin wetting α [mJ/m 2 ] ethanol no 22.1 hexane yes 18.0 methanol no 22.2 decane yes 23.5 acetone no 23.4 cyclohexane yes 24.7 butanone no 23.9 dodecane yes 24.9 1-pentanol no 25.3 tridecane no 25.6 2-heptanone no 26.1 chloroform no 26.9 Water no 72.3 hexadecane no 27.1 Condensation was explored in situ under environmental scanning electron microscopy (ESEM) conditions and confirmed the remarkable resistance of collembolan skin against wetting ( \n Figure 2 \n ). Droplet formation was observed on the secondary granule tops, indicating a stable heterogeneous wetting regime. Even tiny drops with diameters of only a few microns exhibited a spherical shape. Larger droplets occurred in a heterogeneous pattern typically observed with superhydrophobic surfaces and showed contact angles higher than 160° ( \n Figure 2B \n ). The fusion of droplets during growth was often accompanied by a lateral displacement of the drops, confirming a very low hysteresis in the wetting behaviour, in line with earlier reports on superhydrophobic surfaces at condensation [27] . The drops formed by this fusion process attained a more spherical shape than the initial drops and a smaller net liquid-solid contact area resulting in durable anti-fogging characteristics of the skin. Together, these features prevent the formation of a continuous water film and thus suffocation of the springtails in high-humidity environments. To further explore the principles behind these unique non-wetting characteristics we analysed the nanoscale features of collembolan skin. Transmission electron microscopy (TEM) revealed overhanging cross-sections with negative curvature of the smallest structural elements; the primary granules and the connecting bars in the mesh structure ( Figure S1 ). This peculiar profile enlarges the skin-air interface to facilitate respiration and creates a remarkably strong resistance against wetting according to a previously unknown but surprisingly simple principle: As depicted in \n Figure 3 \n , due to the negative curvature of the overhanging profile, an energy barrier must be overcome by the advancing liquid phase before wetting becomes irreversible even for liquids with very low surface tension. Interestingly, engineered surfaces with isolated microelements containing overhanging profiles were recently reported to exhibit amazing super-oleophobic characteristics and provide -through the variability of structural elements- very valuable insights into design criteria for non-wetting surfaces [28] – [32] . As of now, however, those engineered structures did not include any negative curvatures in the overhanging microelements (as illustrated in \n Figure 3 \n ) nor connections of the structural elements, two major components of Collembola skin that explain its uniquely effective and durable anti-adhesive properties. The dimensions of the structural elements of Collembola skin are furthermore at least one magnitude smaller as compared to any of the artificial surfaces considered so far. We suggest that the anti-wetting barrier resulting from the array of nanocavities and the curvature in the shape of the smallest elements was evolutionarily optimized to protect the springtails if acoustic vibrations, pressure jumps or mechanical forces temporarily impose additional energy on the system. While the chemical composition of the springtail skin remains to be analysed in detail, the described design principle can protect surfaces irrespective of their actual chemistry. 10.1371/journal.pone.0025105.g003 Figure 3 Three levels of protection - the anti-wetting skin morphology of springtails. Multiple design principles are combined to protect collembolan skin against wetting: (A) The hairy cover is the first wetting barrier; liquids can be pinned on the bristle tips. If external forces or very low surface tensions enable liquids to conquer this first barrier, a second principle comes into play: (B) Nanoscopic comb structures of interconnected primary granules can still pin liquids by effective retention of entrapped gas nanobubbles within the surface nanocavities. (C) Gas retention is enforced by the previously unknown fact that the overhanging topographies of the structural elements exhibit a negative curvature (with respect to an orthogonal axis to the surface). The result is a forced Cassie state, through which a dramatically reduced solid–liquid contact area leads to increased macroscopic contact angles of drops on the skin surface. As schematically shown in (D), the design principle protects the surface against wetting independent of the surface chemistry and even at very low surface tensions of the liquid and at elevated pressures. Water-repellent, self-cleaning plant surfaces and springtail skin share a hierarchical surface structure with papillous microelements. The collembolan skin, however, is substantially more mechanically stable due to incorporated, flexible bristles and the comb-like alignment of granules (The higher mechanical stability is obvious from the comparison of the surface structures per se and confirmed in a sand abrasion experiment described in the supplement, see Figures S2 , S3 and Table S2 .). Due to embedded nanocavities, springtail skin resists wetting more effectively. In line with this, the skin was also found to exhibit outstanding repellence to particles and bacterial or fungal contamination. None of the microscopically investigated samples in our study ever showed a trace of any adhering material. Furthermore, we massively exposed springtails to Escherichia coli , Staphylococcus aureus and Candida albicans (representing Gram-positive bacteria, Gram-negative bacteria and fungi), respectively, for periods of four days under standard culture conditions without observing any significant deposition. Many valuable opportunities arise from the translation of the described surface structure of collembolan skin into man-made materials and coatings that combine stability against wear and friction with superior non-wetting and anti-adhesive characteristics, addressing critical limitations of the currently employed concepts of superhydrophobic surfaces."
} | 3,655 |
22372639 | null | s2 | 4,326 | {
"abstract": "Highly resilient synthetic hydrogels were synthesized by using the efficient thiol-norbornene chemistry to cross-link hydrophilic poly(ethylene glycol) (PEG) and hydrophobic polydimethylsiloxane (PDMS) polymer chains. The swelling and mechanical properties of the hydrogels were controlled by the relative amounts of PEG and PDMS. The fracture toughness (G(c)) was increased to 80 J/m(2) as the water content of the hydrogel decreased from 95% to 82%. In addition, the mechanical energy storage efficiency (resilience) was more than 97% at strains up to 300%. This is comparable with one of the most resilient materials known: natural resilin, an elastic protein found in many insects, such as in the tendons of fleas and the wings of dragonflies. The high resilience of these hydrogels can be attributed to the well-defined network structure provided by the versatile chemistry, low cross-link density, and lack of secondary structure in the polymer chains."
} | 239 |
27635330 | PMC5012276 | pmc | 4,327 | {
"abstract": "Anemones of genus Exaiptasia are used as model organisms for the study of cnidarian-dinoflagellate (genus Symbiodinium ) endosymbiosis. However, while most reef-building corals harbor Symbiodinium of clade C, Exaiptasia spp. anemones mainly harbor clade B Symbiodinium (ITS2 type B1) populations. In this study, we reveal for the first time that bleached Exaiptasia pallida anemones can establish a symbiotic relationship with a clade C Symbiodinium (ITS2 type C1). We further found that anemones can transmit the exogenously supplied clade C Symbiodinium cells to their offspring by asexual reproduction (pedal laceration). In order to corroborate the establishment of stable symbiosis, we used microscopic techniques and genetic analyses to examine several generations of anemones, and the results of these endeavors confirmed the sustainability of the system. These findings provide a framework for understanding the differences in infection dynamics between homologous and heterologous dinoflagellate types using a model anemone infection system.",
"introduction": "Introduction The sea anemone Exaiptasia pallida is a widespread species that has been well-adopted as a model animal for the study of cnidarian endosymbiology, particularly with associations featuring the dinoflagellate algae Symbiodinium sp. ( Weis et al., 2008 ; Grajales & Rodriguez, 2016 ). In the laboratory, bleached anemones can be prepared by cold shock treatment ( Muscatine, Grossman & Doino, 1991 ) and then maintained for several years in laboratory culture. Recently, genetic examinations of field-collected specimens and laboratory infection demonstrate that E. pallida anemones primarily harbor Symbiodinium spp. of Symbiodinium minutum (ITS2 type B1) and Symbiodinium A4 (ITS2 type A4), and in rare cases, a mixed population of Symbiodinium B1 and C1 ( Thornhill et al., 2013 ; Grajales, Rodriguez & Thornhill, 2016 ), which can be readily isolated from these anemones and cultured in vitro ( Kinzie III et al., 2001 ; Wang et al., 2008 ; Peng et al., 2012 ; Xiang et al., 2013 ). By infecting bleached anemones with free-living Symbiodinium , the endosymbiotic relationship can then be re-established and tracked in order to understand the recognition processes that occur at the molecular level and culminate in successful mutualisms ( Weis et al., 2008 ; Kinzie & Chee, 1979 ; Lin, Wang & Fang, 2000 ; Chen et al., 2004 ; Hong et al., 2009 ; Wang et al., 2013 ; Hambleton et al., 2014 ; Xiang et al., 2013 ). In our previous research involving the infection of E. pallida anemones with various Symbiodinium sp., we discovered the uptake and consequent cellular proliferation of a cultured Symbiodinium of clade C, a lineage of dinoflagellates known to predominantly infect reef corals ( Chen et al., 2005 ; LaJeunesse et al., 2003 ; LaJeunesse et al., 2004a ; LaJeunesse et al., 2008 ; LaJeunesse et al., 2010 ; Lien, Fukami & Yamashita, 2012 ). This is a particularly interesting finding, and it needs to be confirmed if this Exaiptasia -clade C Symbiodinium association is a sustainable endosymbiotic relationship. Furthermore, if the nature of this association is proven to be similar to that of corals, this Exaiptasia -clade C Symbiodinium association deserves even more merit as a model system for understanding reef corals, which cannot be successfully bleached and re-infected due to the stress it imposes on the corals (i.e., they are obligately endosymbiotic). Metabolic relationships between corals and Symbiodinium are functionally diverse, depending largely on the genetic identity of the latter ( Baker et al., 2004 ; Abrego et al., 2008 ; Stat, Morris & Gates, 2008 ; Yuyama, Harii & Hidaka, 2012 ; Yuyama & Higuchi, 2014 ). For instance, corals associated with Symbiodinium of clade D have been shown to possess an enhanced degree of thermal tolerance ( Baker et al., 2004 ). Understanding the physiological consequences of engaging in an endosymbiotic relationship with dinoflagellates of differing identity would then be useful in formulating predictions as to how anemones, or even reef corals, may respond to global climate change. To further corroborate our previously unpublished findings and gain greater insight into the ability to develop a heterologous anemone- Symbiodinium infection system, we co-cultured exogenously supplied Symbiodinium C1 with bleached anemones (infection trial). The Symbiodinium C1-infected anemones were then maintained in the laboratory for more than one year, and asexual reproduction (pedal laceration) of Symbiodinium C1-infected anemones was observed.",
"discussion": "Results and Discussion To determine whether the clade C Symbiodinium (CCMP2466; ITS2 type C1) could be taken up by Exaiptasia anemones and then be transmitted to offspring after proliferation through the adult tissues, anemones infected with the clade C Symbiodinium were cultured for more than one year. During this period, juvenile anemones infected with Symbiodinium C1 grew from 2–3 mm in height to 2–3 cm in height, and Symbiodinium cells proliferated throughout the bodies of the specimens ( Fig. 1B ). Asexual reproduction (pedal laceration) was also observed and recorded ( Fig. 1B ). An aboral view of a representative anemone clearly illustrates the process of asexual reproduction, pedal laceration, in which the newly budded lacerates surround the pedal disk of the anemone ( Fig. 1B ). Furthermore, Symbiodinium cells were found to aggregate within the lacerates ( Fig. 1B ). 10.7717/peerj.2358/fig-1 Figure 1 A representative bleached and a representative clade C Symbiodinium -infected anemone and its lacerates. (A) Representative image of a bleached anemone that had lost its brownish coloration following the expulsion of Symbiodinium cells during cold shock-induced bleaching. (B) Aboral view of a representative clade C Symbiodinium -infected anemone showing brownish Symbiodinium cells distributed throughout the body, with a notable degree of dinoflagellate aggregation in the margins of the pedal disk (arrow), as well as within the newly budded lacerates (triangles). Scale bars: 0.5 cm. The lacerates of Symbiodinium C1-infected anemones developed into juvenile anemones within 9 days ( Fig. 2 ). On day 1, red autofluorescence of the Symbiodinium cells was already readily observed with a fluorescence stereomicroscope ( Fig. 2A ). On days 3 and 4, tentacle tissue began forming and extended out from the top of the lacerates, as shown in Figs. 2C and 2D . Between days 5 and 9, the tentacle tissue extended significantly to form the shape of a juvenile anemone ( Figs. 2E – 2I ). During lacerate development, Symbiodinium cells were re-distributed to the tentacles, where they are predominantly localized in healthy, adult anemones. Such localization suggests that these clade C Symbiodinium cells had successfully established a symbiotic relationship with the juvenile anemones. 10.7717/peerj.2358/fig-2 Figure 2 Development of a representative lacerate collected from a clade C Symbiodinium -infected anemone. First, the lacerate was transferred to a new dish immediately following laceration from the pedal disk of a clade C Symbiodinium -infected anemone. The development of the lacerate and spread of Symbiodinium was recorded daily using a fluorescent stereomicroscope (A–I). The red spots in the images indicate chlorophyll autofluorescence of the Symbiodinium cells. Scale bars: 100 µm. To confirm the sustainability and genetic identity of the clade C Symbiodinium within the infected anemones and their offspring, specimens of three generations of anemones ( Fig. 3 ) were subjected to RFLP analysis of partially digested Symbiodinium 18S rDNA. Our results reveal that all three generations of the clade C Symbiodinium -infected anemones had the same RFLP pattern ( Fig. 4 ). The major fragments were ∼900 and ∼750 bp when digested with Taq I and ∼900 and ∼500 bp when digested with Sau 3AI ( Fig. 4 lanes 2–10); such patterns are diagnostic of Symbiodinium C1 ( Rowan & Powers, 1991a ; Rowan & Powers, 1991b ; Rowan & Knowlton, 1995 ). Furthermore, RFLP patterns of the clade C Symbiodinium -infected anemones were well-differentiated from the representative clade B Symbiodinium -infected anemone ( Fig. 4 , Lane 1), which came from the same clonal line as the clade C-infected anemones but was not subject to bleaching and re-infection. The RFLP patterns were ∼900 and ∼500 bp upon digestion with Taq I and ∼800 and 500 bp upon digestion with Sau 3AI ( Fig. 4 , lane 1). These results clearly demonstrate that the anemones which originally harbored clade B Symbiodinium now harbor clade C and transmitted the symbiont for many generations. 10.7717/peerj.2358/fig-3 Figure 3 Images of the clade C Symbiodinium -infected anemones and their offspring. (A) Generation 1 (G1) anemones maintained in the laboratory for more than one year; (B) Generation 2 (G2) anemones cultured for more than three months; (C) Generation 3 (G3) of anemone cultured for 15 days following laceration. Scale bar: 1 cm. 10.7717/peerj.2358/fig-4 Figure 4 Restriction fragment length polymorphism (RFLP) analysis of three generations of the clade C Symbiodinium -infected anemones. Amplified genomic fragments of small subunit ribosomal RNA genes (18S rDNA) from Symbiodinium were digested using the restriction enzymes Taq I and Sau 3A I (A and B, respectively). Lane 1 (control): Normal anemone harboring clade B Symbiodinium . Lanes 2–4: Generation 1 (G1) anemones. Lanes 5–7: Generation 2 (G2) anemones. Lanes 8–10: Generation 3 (G3) anemones. Ep, Exaiptasia pallida ; MW, molecular weight. In addition to the RFLP analysis, the bright field images ( Fig. 5 ) and diameter measurement of Symbiodinium further support the presence of clade C Symbiodinium cells in the anemones in contrast with the presence of clade B Symbiodinium . As shown in Fig. 5 , the clade C Symbiodinium cells appear more darkly brownish than clade B cells. Moreover, the clade C Symbiodinium cells averaged 8.30 ± 0.09 µm ( n = 855) in diameter in hospite , significantly larger ( p < 0.01) than in hospite clade B cells (7.72 ± 0.13 µm, n = 992). Since the cell size of the clade C Symbiodinium (6.70 ± 0.84 µm, n = 1,000) is significantly larger ( p < 0.01) than clade B Symbiodinium (6.00 ± 0.68 µm, n = 1,000) in free-living cultures, the larger Symbiodinium cells in the anemones demonstrate it harbors the clade C Symbiodinium instead of the clade B. It also demonstrates that like the clade B Symbiodinium ( Pasaribu et al., 2015 ), the cell size of the clade C Symbiodinium increase in endosymbiotic condition. This experimental evidence reveals that the exogenously supplied clade C Symbiodinium can be transmitted to multiple generations of progeny via pedal laceration under laboratory conditions. We thus conclude that the Exaiptasia anemone species used in this study can establish a symbiotic relationship with the clade C Symbiodinium (CCMP2466; ITS2 type C1). In addition, our study and previous work that has infected Exaiptasia with heterologous Symbiodinium types originally from other host species ( Schoenberg & Trench, 1980 ; Hawkins et al., 2016 ) also demonstrate that Exaiptasia pallida could potentially establish symbiotic relationships with other clades of Symbiodinium . 10.7717/peerj.2358/fig-5 Figure 5 The morphology of clade B and clade C Symbiodinium cells within their host anemone tissue. (A) Tentacular tissues harboring clade B Symbiodinium ; (B) Tentacular tissues harboring clade C Symbiodinium . Scale bar: 20 µm. Recently, the scientific name of several anemones used widely as a model system for the study of cnidarian-dinoflagellate endosymbiosis has been revised to Exaiptasia pallida ( Grajales & Rodriguez, 2014 ; Grajales & Rodriguez, 2016 ; Grajales, Rodriguez & Thornhill, 2016 ). This change was based on the updated morphological and genetic population study of the newly collected specimen from around the world. According to these global comparative investigations, the anemones used in our study, which originate from the same wild population as the studies cited above (N22 03 00.08 E120 41 42.88), belong to a widespread lineage that hosts Symbiodinium minutum (ITS2 type B1). These updated studies also found that, in addition to the previously known Florida lineage of Exaiptasia pallida which hosts Symbiodinium A4, B1, and a mixed population of B1 and C1 ( Thornhill et al., 2013 ), Exaiptasia hosts a mixed population of Symbiodinium B1 and C1 in Bermuda and Symbiodinium type A4 in Mexico and the Bahamas ( Grajales & Rodriguez, 2016 ; Grajales, Rodriguez & Thornhill, 2016 ). Therefore, although the present study is not the first to document Exaiptasia spp. hosting clade C Symbiodinium , it does show for the first time that the Exaiptasia pallida anemones predominantly known to host the clade B Symbiodinium ( LaJeunesse, Parkinson & Reimer, 2012 ; Thornhill et al., 2013 ; Kinzie et al., 2001 ) can be bleached and re-infected with exogenously supplied dinoflagellate algae of the clade C Symbiodinium . The clade C Symbiodinium (CCMP2466) used in this study was originally isolated from a corallimorph ( Discosoma sanctithomae) , not a reef-building coral, but the genetic data shows that it belongs to a line (ITS2 type C1; Krueger et al., 2015 ) of diverse clade C Symbiodinium that are mainly harbored by reef-building corals ( Van Oppen et al., 2001 ; LaJeunesse et al., 2003 ; LaJeunesse et al., 2004b ; Chen et al., 2005 ; LaJeunesse et al., 2010 ) and are sensitive to thermal stress ( Baker et al., 2004 ; Litmann, Bourne & Willis, 2010 ). Since CCMP2466 could infect aposymbiotic larvae of a reef-building coral, Acropora tenuis , and then establish a monoclonal Symbiodinium C1-infected coral association ( Yuyama & Higuchi, 2014 ), CCMP2466 has been successfully applied to the study of thermal tolerance of corals when it harbors the thermal sensitive clade C (CCMP2466) or thermal tolerant clade D Symbiodinium ( Yuyama et al., 2016 ). These updated reports and this present study imply that the Exaiptasia-Symbiodinium C1 association is an interesting and useful system for studying the functional diversity between cnidarian hosts and their symbionts. In conclusion, the present study opens up a window for future studies to determine whether the molecular pathways and the type of symbiosis underlying the establishment of the Exaiptasia – Symbiodinium B1 association are similar to the Exaiptasia – Symbiodinium C1 endosymbiotic association. If this relationship is later shown to be mutualistic, as is Exaiptasia- clade B Symbiodinium association, then this could serve as a potentially valuable model for the study of cnidarian- Symbiodinium endosymbiosis."
} | 3,738 |
37445045 | PMC10342640 | pmc | 4,328 | {
"abstract": "In this paper, we present a sustainable approach for the creation of superhydrophobic (SP) coating on a stainless-steel substrate based on a biological metal–organic framework (MOF). The MOF was synthesized using aspartic acid as a linker and copper ions as a core metal. Two SP coatings were well constructed on stainless steel utilizing electrodeposition of nickel (Ni) and nickel altered by MOF (Ni@Bio-MOF) coatings followed by soaking in a solution of stearic acid in ethanol. The results of Fourier transform infrared spectroscopy demonstrate that the stearic acid-grafted nickel coating (Ni@SA) and the stearic acid-grafted Ni@Bio-MOF composite (Ni@Bio-MOF@SA), were effectively deposited on the stainless steel. The wettability findings displayed that the water contact angle of Ni@SA and Ni@Cu-As MOF@SA are 160° ± 1.1°, and 168° ± 1.2°, respectively. The prepared SP coating was also found to be chemically and mechanically stable. The results show that the Ni@SA coating maintains SP characteristics in a pH range of 3–11 while the Ni@Cu-As MOF@SA coating retained SP characteristics in a pH range of 1–13. Additionally, the superhydrophobic Ni@SA coating demonstrated SP characteristics up to a length of abrasion equal to 1300 mm, while the Ni@Cu-As MOF@SA coating exhibited SP characteristics up to a length of abrasion equal to 2700 mm. Furthermore, the Ni@SA and Ni@Cu-As MOF@SA coatings exhibited significantly improved corrosion protection in a 0.5 M NaCl solution compared with bare stainless steel, with protection efficiencies of approximately 94% and 99%, respectively. The results of this study demonstrate that the proposed approach is a promising method for the fabrication of eco-friendly and corrosion-resistant SP coatings on stainless steel substrate.",
"conclusion": "4. Conclusions In our study, an eco-friendly approach to the construction of SP coatings on SS metal based on biological metal–organic frameworks (MOFs) was developed. The MOF was synthesized using aspartic acid as the organic linker and copper ions as the metal center. The water contact angles of SS coated with Ni@SA and Ni@Cu-As MOF@SA are 160° ± 1.1°, and 168° ± 1.2°, respectively. The chemical stability results show that Ni @SA coating maintains SP characteristics in the pH range of 3–11, whereas the Ni@Cu-As MOF@SA coating retains SP characteristics in the pH range of 1–13. The mechanical stability results show that the created SP Ni@SA coating demonstrates SP characteristics up to an abrasion length of 1300 mm, while the Ni@Cu-As MOF@SA coating exhibits SP characteristics up to an abrasion length of 2700 mm. The corrosion resistance of the coated SS was also significantly improved. The corrosion current density of the coated SS with Ni@SA is 0.0041915 µA/cm 2 , and Ni@Cu-As MOF@SA is 0.0041915 µA/cm 2 , which was much lower than that of the bare SS (0.0710569 µA/cm 2 ). The results of this study demonstrate that the proposed approach is a promising method for the construction of eco-friendly SP coatings with excellent corrosion resistance.",
"introduction": "1. Introduction Extremely non-wettable surfaces, also known as superhydrophobic (SP) surfaces are surfaces that have a contact angle greater than 150 degrees and are highly resistant to water [ 1 ]. These surfaces have attracted interest due to their potential applications in various fields such as antifouling technologies, microfluidic devices, biomedical, solar cells, sensors, drag reduction, oil–water separation, and corrosion resistance [ 2 , 3 , 4 , 5 ]. However, creating SP surfaces can be difficult and most methods require extreme conditions, especially when environmental issues are present. There are a variety of techniques used to create SP surfaces, including electrodeposition, spraying, anodization, electrospinning, sol-gel, and chemical vapor deposition [ 6 , 7 ]. Electrodeposition is a technique that is relatively simple, low-cost, and flexible, making it an excellent option for creating artificial SP surfaces [ 8 ]. The scalability of the electrodeposition method depends on a number of factors such as the size and shape of the object being coated. While it may be feasible to use electrodeposition to coat large objects, such as wings, aircrafts, or the rotor blades of wind turbines, doing so would require significant modifications to the electrodeposition system to ensure uniform coating and adhesion over such large surfaces. By raising surface roughness, a vital prerequisite for superhydrophobicity and lowering surface energy, another crucial prerequisite for superhydrophobicity, SP films with significant water repellency can be produced [ 9 ]. As roughness increases, the surface can trap air pockets within the gaps between surface features, preventing the liquid from fully wetting the surface and resulting in a Cassie–Baxter state and superhydrophobicity. The degree of superhydrophobicity also depends on the shape of the protrusions and their spacing [ 10 ]. Historically, perfluorinated compounds have been used to lower surface energy because of their own very low surface energy, but these compounds have been shown to have toxic and harmful environmental effects [ 11 ]. As a result, eco-friendly techniques, and materials for producing SP surfaces are required. To enhance roughness, researchers have utilized a range of nanomaterials such as metal–organic frameworks (MOFs), carbon nanotubes, SiO 2 , TiO 2 , ZnO, and CuO [ 12 , 13 , 14 , 15 , 16 , 17 ]. MOFs are combination materials with porous structures and periodic network arrangements produced via organic linkers and core metal ions. These have gained popularity due to their unique properties and potential applications [ 1 ]. However, traditional methods for creating MOFs have their own drawbacks, while electrochemical techniques offer various benefits, such as clean-up, convenience of use, mild reaction conditions and metal ions formed in situ via anodic oxidation, which avoids the need to utilize problematic anions [ 18 ]. Recently, biological metal–organic frameworks, Bio-MOFs, have been gaining reputation as renewable framework materials that are recyclable, and green [ 19 ]. Bio-MOFs use biomolecules, such as amino acids, peptides, nucleobases, sugars, and other bio-based materials, as building blocks and water as a solvent and have been used to produce SP surfaces [ 19 ]. Stainless steel (SS) is widely used in the marine engineering, aerospace, automotive, petrochemical, nuclear engineering and biomedical sectors [ 20 , 21 ]. Though SS has the advantages of an improved corrosion resistance, a good toughness and a high strength, it is subjected to severe corrosion in media containing chloride ions [ 22 , 23 ]. Corrosion is a significant problem for society as it affects both public safety and the economy [ 24 , 25 , 26 ]. The impact of corrosion on public safety can be significant as it can weaken the structural integrity of various infrastructure, such as buildings, bridges, pipelines, power plants, transportation vehicles, etc. Corrosion can potentially lead to failures and accidents that may cause injuries or fatalities and can also have serious consequences for the environment. Nuclear power plants are particularly susceptible to corrosion-related issues that can compromise their safety and reliability. Therefore, it is crucial to develop effective corrosion protection and mitigation strategies to ensure the safety and reliability of our infrastructure and systems. One of the most effective ways to reduce SS corrosion is by creating SP coats that drastically increase the corrosion resistance of SS [ 27 ]. Nickel is a valuable metal in the industrial sector, known for features such as corrosion resistance, hardness, and magnetism. The corrosion of SS is slowed when nickel is applied to it. When combined with superhydrophobicity, the nickel coat is able to deliver additional advantages, such as better corrosion resistance and self-cleaning properties [ 28 ]. The goal of this research is to create an SP coat on an SS substrate using Bio-MOF. The electrochemical process was utilized to prepare a Cu-As MOF coating which contains aspartic acid as the linker and copper as the core metal. Two rough coats of Ni, and Ni modified with Bio-MOF (Ni@Bio-MOF), were grafted on an SS substrate through electrostatic deposition. The coats were soaked in an ethanol solution containing stearic acid (SA) to create SP surfaces. To increase the mechanical and chemical stability of the SP coats on stainless steel, we innovatively employed a composite of Cu-As bio-MOF, synthesized by electrodeposition, and nickel. Stearic acid was utilized as a low-surface energy compound due to it being an eco-friendly substance that is more cost-effective in comparison with other options, such as toxic fluorinated polymers and silanes. The prepared SP coatings’ wettability, mechanical and chemical stability, and corrosion resistance in a solution of 0.5 M NaCl were evaluated.",
"discussion": "3. Results and Discussion 3.1. Thermogravimetric Results of the Prepared Cu-As MOF The thermogravimetric results of Cu-aspartic acid MOF, shown in Figure 1 , would likely center around the three distinct regions observed in the graph, and the changes in weight that occur in each region. The first region, between 32 and 101 °C, may be characterized by a relatively low rate of weight loss as the MOF loses adsorbed water or other weakly bound species. The second region, between 101 and 218 °C, may be characterized by a more rapid weight loss as the MOF loses more strongly bound species or undergoes structural changes. The third region, between 218 to 288 °C, may be characterized by a slower rate of weight loss, as the MOF reaches its maximum decomposition temperature. This region may indicate that the MOF is losing its structural integrity. 3.2. FTIR Results The FTIR spectra of coated SS with Ni@Cu-As MOF, Ni@Cu-As MOF@SA, and Ni@SA are presented in Figure 2 . The FTIR results for the coated SS with Ni@Cu-As MOF likely indicate the presence of several functional groups in the material. The band at 3463 cm −1 , and 3111 cm −1 may be due to the N-H 2 stretch of aspartic acid [ 27 ]. The bands at 2980 cm −1 , and 2899 cm −1 may be due to the presence of C-H symmetry and a symmetry vibration of -CH 2 - groups [ 1 ]. The band at 1724 cm −1 is due to the stretching vibration of C=O and the band at 1439 cm −1 may be due to the presence of C-N stretching vibrations, indicating the presence of amine groups [ 1 ]. The band at 864 cm −1 is due to the stretching of C-H bonds in an aspartic acid compound [ 1 ]. The band at 728 cm −1 is due to the presence of Ni(OH) 2 bending in the coating while the band at 506 cm −1 is due to the metal–oxygen stretching vibrations and the band at 429 cm −1 is due to the metal–oxygen bending vibrations, indicating the coordination of the copper ions with the oxygen atoms of the carboxylate and amine groups [ 1 ]. The spectrum of the SS coated with Ni@Cu-As MOF@SA displays similar bands to that of the Cu-As MOF, but with slight changes in the position of the band of the C=O stretch and the N-H 2 stretch band, which appear at 1733 cm −1 and 3294 cm −1 , respectively. This suggests that the Cu-As MOF has been doped with SA [ 27 ]. The spectrum of the coated SS with Ni@SA coat displays a band at 3530 cm −1 which is likely due to the presence of the hydroxyl groups of stearic acid [ 27 ]. The bands at 2932 cm −1 , and 2894 cm −1 may be due to the presence of C-H symmetry and a symmetry vibration of -CH 2 - groups [ 27 ]. The band at 1698 cm −1 is associated with the stretching vibrations of C=O in the stearic acid and the band at 1454 cm −1 is likely due to the bending vibrations of the CH 2 groups in the stearic acid [ 29 ]. The bands at 1255 and 971 cm −1 are due to CH stretch. The band at 689 cm −1 is likely attributed to the presence of Ni(OH) 2 bending in the coating [ 30 ]. 3.3. SEM and Wettability The SEM of the SS coated by Ni@Cu-As MOF@SA, and Ni@SA are presented in Figure 3 . The discussion of SEM results of SP-coated SS with Ni@Cu-As MOF@SA, and Ni@SA likely centers around the differences in microstructure and surface roughness between the two coatings. The SEM micrographs likely show that the coating made with Ni@Cu-As MOF@SA has smaller circular microstructures compared with the coating made with Ni@SA. The usage of an MOF in the Ni@Cu-As MOF@SA coating may operate as a nucleation center for the electrodeposition process, speeding up the nucleation process rather than crystal growth and producing smaller structures, increasing the surface’s roughness. The smaller size of the microstructures may also contribute to the SP properties of the coating, as smaller structures can lead to a more roughness, and stable water-repellent coating. The wettability of the Ni@SA, and Ni@Cu-As MOF@SA was examined by the measurement of CA. The Ni@SA has CA of 160° ± 1.1°, and an SA of 4° ± 0.1°, while Ni@Cu-As MOF@SA has a CA of 168° ± 1.2°, and an SA of 1° ± 0.1°, so the two coats showed excellent SP properties. The micrograph of the water droplet on the SP prepared with Ni@SA and Ni@Cu-As MOF@SA is shown as an inset in Figure 3 . The rolling/bouncing of the water droplet on the SS coated by Ni@Cu-As MOF@SA is illustrated in Video S1 . 3.4. Chemical Stability Chemical stability is considered an essential requirement for SP coatings to work well over time in harsh solution conditions. The correlations between the CAs and SAs of water droplets on the SP coatings and the solution pH are depicted in Figure 4 . The shape of the water droplet on the SS coated with Ni@Cu-As MOF@SA after being immersed in a solution of pH 7 for 5 h is illustrated in Figure 5 . Video S2 demonstrates the superhydrophobicity and rolling of a water droplet on the SS coated with a Ni@Cu-As MOF@SA surface after immersion in the pH 7 solution for 5 h. According to the findings, Ni@SA films are SP in the pH range of 3–11, while Ni@Cu-As MOF@SA films are SP in the pH range of 1–13, where the CAs are frequently higher than 150° and the SAs are less than 10°. The chemical stability of SP-coated SS with Ni@Cu-As MOF@SA is higher than that of SS coated with Ni@SA because the Cu-As MOF enhances the coating superhydrophobicity and provides an additional layer of protection. The SP-coated SS with Ni@Cu-As MOF@SA has a superior chemical stability to numerous values that have been reported previously [ 31 , 32 , 33 , 34 ]. 3.5. Mechanical Stability SP surfaces often have limited practical applications due to their mechanical fragility. When touched with a finger, some surfaces with SP characteristics can crash [ 4 ]. The produced SP films’ resistances to mechanical abrasion were assessed utilizing abrasion and sand impact tests. Figure 6 depicts the variations in CAs and SAs of the manufactured SP films with respect to the abrasion length. The Ni@SA SP film maintains its SP characteristics up to a 1300 mm abrasion length. In comparison, the SP Ni@Cu-As MOF@SA film preserves its SP characteristics up to a 2700 mm abrasion length. The SP-coated SS with Ni@Cu-As MOF@SA exhibits larger abrasion resistance than numerous stated values [ 27 , 35 ]. The mechanical abrasion test for SS coated with Ni@Cu-As MOF@SA film for abrasion length of 20 cm is shown in Video S3 . The mechanical abrasion resistance of SP-coated SS via Ni@Cu-As MOF@SA is higher than that of SP SS coated with Ni@SA only because the MOF layer enhances the superhydrophobicity and provides an additional layer of protection, resulting in a more durable and abrasion-resistant coating [ 36 , 37 ]. As seen in Figure 7 , the sand abrasion assessments were undertaken to evaluate the mechanical performance of the SP coatings. The Ni@SA film maintains SP characteristics up to 10 sand impact cycles, while the Ni@Cu-As MOF@SA film exhibits superhydrophobicity up to 20 sand impact cycles. Ni@Cu-As MOF@SA exhibit a sand impact resistance larger than several previously stated values [ 33 , 38 ]. 3.6. Corrosion Measurements 3.6.1. Potentiodynamic Polarization Results The corrosion behaviors of uncoated and SP-coated SS by Ni@SA, Ni@Cu-As MOF@SA were studied using the potentiodynamic polarization technique. The potentiodynamic polarization plots of bare and SP-coated SS in 0.5 M NaCl are shown in Figure 8 . The observation of limited diffusion currents during cathodic polarization suggests that the cathodic process is governed by the transfer of oxygen gas from the bulk to the electrode surface. Pitting corrosion for bare SS or the development of a passive layer for SS cured with an SP coating prevent the formation of an ideal anodic Tafel area [ 39 , 40 ]. Table 2 shows the potentiodynamic polarization variables for bare and SP-coated SS, containing corrosion potential (E corr ), protection efficiency (%P), and corrosion current density (i corr ). The %P was determined utilizing Equation (2) [ 41 ].\n %P = [(i o − i )/ i o ] × 100 (1) \nwhere, i o and i are the corrosion current densities of the bare SS and the SP-coated SS, respectively. The i corr value for coated SS with Ni@SA is lower than that for bare SS due to the superhydrophobicity of the coated SS. Air trapped in the SP coating microstructures can diminish the surface area between the solution and SS, which causes the i corr value to fall more quickly [ 42 ]. The doping of the superhydrophobic Ni@SA coat with MOF enhances the SP property, leading to a greater decline in the contact area between the medium and SS. Therefore, SS coated with Ni@Cu-As MOF@SA has a higher protection efficiency than SS coated with Ni@SA. 3.6.2. Electrochemical Impedance Spectroscopy Results Nyquist, Bode and Theta plots of the bare and SP-coated SS in a 0.5 M NaCl solution are presented in Figure 9 . The Nyquist plots, shown in Figure 9 a, exhibit a diffusion tail at low frequency and a depressed capacitive semicircle at high frequency. The depressed capacitive semicircle observed at high frequencies in the Nyquist plots is due to the interfacial charge transfer reaction [ 43 ]. The diffusion tails observed at low frequencies are attributed to mass transfer. Based on these observations, it can be inferred that the improved charge transfer resistance of SS coated with Ni@SA compared with bare SS is attributed to the presence of a protective SP layer. The SS coated with Ni@Cu-Asp MOF@SA exhibits the largest capacitive semicircle, suggesting that it provides the highest level of protection. The incorporation of MOF onto Ni@SA enhances the superhydrophobicity of the surface, making the superhydrophobic Ni@Cu-Asp MOF@SA coating more effective in restricting the diffusion of corrosive species such as Cl − and H 2 O into the SS substrate. When the superhydrophobic-coated SS was immersed in a 0.5 M NaCl solution, it demonstrated significantly higher impedance magnitudes at lower frequencies on the Bode plots, shown in Figure 9 b, compared with bare steel. This clearly indicates that the created SP coatings have successfully protected the SS substrate. The phase angle plot, shown in Figure 9 c, shows two time constants at low and intermediate frequencies. The time constant observed at the low-frequency region is attributed to the unprotective corrosion products of bare SS or the protective SP coating. On the other hand, the time constant observed at the moderate frequency is attributed to the electrical double layer. The theta angle, which is approximately 45 degrees at a moderate frequency, indicates that the corrosion process is under diffusion control. The equivalent circuit displayed in Figure 10 was utilized to fit the results of the electrochemical impedance spectroscopy experiment and the Zsimpwin program was utilized to calculate the impedance parameters. The components of the equivalent circuit include the solution resistance (R s ), the charge transfer resistance (R ct ), the double-layer constant phase element (CPE dl ), and the Warburg element (W). Table 3 demonstrates the electrochemical impedance spectroscopy parameters for both bare SS and SP-coated SS. Equation (2) was utilized to estimate the %P [ 27 ]: %P = [(R ct − R cto )/R ct ] × 100 (2) \nwhere the charge transfer resistances for uncoated and SP-coated SS are R cto and R ct , respectively. Table 3 presents the obtained impedance parameters. It is clear that, each of R ct and %P of the bare SS < SS + Ni @SA < SS + Ni@Cu-As MOF@SA. The corrosion performance of the SP-coated SS by Ni@Cu-As MOF@SA is greater than many previously documented values [ 44 , 45 , 46 , 47 ] and lower than other previously reported values [ 48 , 49 ]. The corrosion resistance of SP-coated SS with a Ni@Cu-As MOF@SA coating is higher than that of SS coated with Ni@SA. This may be due to the way in which the Cu-As MOF increases the superhydrophobicity and forms a protective barrier against corrosive agents such as water and oxygen, thus slowing down the corrosion process and enhancing the corrosion resistance of the coating."
} | 5,262 |
23882213 | PMC3714542 | pmc | 4,332 | {
"abstract": "Efficient path planning and navigation is critical for animals, robotics, logistics and transportation. We study a model in which spatial navigation problems can rapidly be solved in the brain by parallel mental exploration of alternative routes using propagating waves of neural activity. A wave of spiking activity propagates through a hippocampus-like network, altering the synaptic connectivity. The resulting vector field of synaptic change then guides a simulated animal to the appropriate selected target locations. We demonstrate that the navigation problem can be solved using realistic, local synaptic plasticity rules during a single passage of a wavefront. Our model can find optimal solutions for competing possible targets or learn and navigate in multiple environments. The model provides a hypothesis on the possible computational mechanisms for optimal path planning in the brain, at the same time it is useful for neuromorphic implementations, where the parallelism of information processing proposed here can fully be harnessed in hardware.",
"introduction": "Introduction One of the central problems for neurobiology is to understand the computational effectiveness of the brains of higher animals. Brains rapidly carry out extraordinary feats of visual scene analysis or problem solving through thinking on “wetware” that is tens of millions times slower than modern digital hardware. Part of the explanation is brute-force anatomical parallelism. In this paper we develop a model of parallel computational processing in the context of path planning and spatial navigation. We propose that spatial navigation can be solved through simultaneous mental exploration of multiple possible routes. A typical mental exploration task for an animal might involve knowing an extensive terrain containing a few water sources, being motivated (being thirsty) to seek the nearest water source. Hopfield ( 2010 ) recently described a way that serial mental search for a useful route could be done by a moving clump of activity and synapse modification in a hippocampus-like neural network 1 . We show here that a best path can rapidly be found by parallel search in the same kind of network, but by a propagating wave of spiking activity. The process of path planning and navigation, as proposed in our model, consists of the following steps: (1) expanding waves of neural activity are initiated from the place cells corresponding to selected target location(s); (2) the propagating waves alter synaptic connectivity within the network through spike-timing-dependent plasticity and create a directed synaptic vector field (SVF) converging on the goal locations; (3) this vector field is used by an animal to navigate toward targets; (4) whenever a new planning process is necessary, all synapses are reset to the baseline state and waves of activity can be initiated from the new target locations. Can animals employ such parallel mental exploration to solve novel problems? Indeed can humans do so? Recent electrophysiology experiments demonstrated existence of expanding, traveling waves of neural activity in the hippocampus, associated with theta-oscillations (Lubenov and Siapas, 2009 ; Patel et al., 2012 ), as well as with much faster sharp wave ripples (Ellender et al., 2010 ), yet, no link between such waves and spatial planning has been shown so far. One of the major roles of theory is to elucidate interesting consequences and possibilities inherent in our incomplete experimental knowledge of a system. The fact that hippocampus-like neural substrate can support parallel mental exploration, as explored here, is such a possibility. New experimental paradigms could easily test for parallel mental exploration in rats. These ideas also form the basis for novel neuromorphic circuits in engineering, which could be used to implement effectively certain Artificial Intelligence algorithms such as those based on the idea of a wave-front propagation (Dorst and Trovato, 1988 ; Dorst et al., 1991 ; LaValle, 2006 ) by taking advantage of the true parallelism of the neuromorphic hardware systems (Boahen, 2005 ; Misra and Saha, 2010 ).",
"discussion": "Discussion The problem of planning and executing a complex motion over a protracted time-period which will optimally take an autonomous agent from its present location and configuration to a desired target location and configuration is common to both animal behavior and robotics. In its simplest manifestation there is only a single target, a single known environment, and a short or fast path is preferred over a longer or slower one. The trajectory planning must accommodate the physical constraints posed by the environment. Additional complexities might include the simultaneous presence of multiple targets, possibly of different intrinsic values, terrain which affects the value of trajectories in a non-trivial fashion, and multiple environments. The neurally-inspired network presented in our work has been shown to solve the planning problem in several steps. First, in an exploratory phase it learns an environment by developing a set of “place” cells whose locations reflect all possible trajectory boundaries due to kinematic constraints or constraints in the behavior arena. It develops in this exploration process interconnections between all pairs of places that can be visited in temporal contiguity, and thus can be possible candidates for a section of a trajectory. Second, given the expected set of synaptic connections, the excitation of a target location (or locations) initiates a wavefront of single spike- or single burst activity that propagates outward from the initiation site(s). The wave propagation process is terminated when a wavefront reaches the present location of the agent. The passage of such a wavefront produces synapse modification pattern that can be described as a vector field. The desired trajectory is simply the path along the SVF from the present location to the (or a) target. Since the SVF lines are produced by an expanding circular wavefront, they converge when followed backward toward a source, and thus provide stable guidance for going to a target location. The full extent of the parallelism available in our concept is perhaps best illustrated in Figure 4 . The system simultaneously selects the closest target and the best route to that target from a single propagation of the exploration wave. Conventionally, a best path would be found for each target sequentially, using a serial algorithm to rate possible paths, and a choice of target then made between these optimal single-target paths. The conceptualization of the parallel search method and the demonstration by simulation that best trajectories can be followed in neuromorphic simulation are the major accomplishments of this paper. Network analysis As mentioned before, the goal of our paper was to present a concept of parallel exploration through propagating waves of neural activity and STDP-altered SVFs. We have illustrated our concept in a set of simulations, but we have not attempted to quantify our results. An interesting extension of our work thus would be to perform an analysis of the properties of our system. Interestingly, such an analysis has recently been offered for the network proposed in (Hopfield, 2010 ), which is of the same type and topology as the one considered in our work. Indeed, Monasson and Rosay ( 2013 ) provided an indepth theoretical analysis of the dynamics and storage capacity of that network as a function of such parameters as: network size, level of neural activity, level of noise, or size of place cells. Specifically, using the statistical mechanics tools, the authors analysed conditions necessary for the network to learn multiple maps (environments). The storage of a map manifests itself through the fact that the neural activity is localized, and acquires a clump-like shape in the corresponding environment. Remarkably, according to the analysis performed by the authors, a moderate level of noise can slightly increase the capacity storage with respect to the noiseless case. However, when the number of environments or the noise are too high the neural activity cannot be localized any longer in any one of the environments. For high noise, the activity, averaged over time, becomes uniform over space. For high loads the activity is not uniform, but is delocalized with spatial heterogeneities controlled by the cross-talks between the maps. The paper provides quantitative results for the transition between these states. The authors also analyse storage capacity of the network, that is a maximum number of environments for which a stable representation of a given environment can still be retrieved, as a function of network size and topology. For the network of the type considered in (Hopfield, 2010 ), and so also in our work, the storage capacity is proportional to the network size and is estimated to be of the order of 10 −3 bits per synapse (for the 2 dimensional space representation and under the optimal conditions). Interestingly, these results are consistent with an earlier analysis for a network with a similar topology but with a different neuron type given in Battaglia and Treves ( 1998 ). Related models The wave-propagation concept has first been introduced by Dorst and Trovato as an efficient parallel method for path planning (Dorst and Trovato, 1988 ; Dorst et al., 1991 ) and since then has widely been used in robotics and computer science (LaValle, 2006 ). The wave-front methods are essentially the same as exhaustive or heuristic versions of a classical A * search algorithm (Dijkstra, 1959 ; Hart et al., 1968 ) of whose optimality is proven. Several neural models for spatial navigation using the concept of propagating waves have been proposed so far (for reviews see, e.g., Lebedev et al., 2005 ; Qu et al., 2009 ). However, only a few models addressed a question on how the propagating neural activity can be transformed into an appropriate configuration of synaptic connectivity able to later guide an agent to a target location (Roth et al., 1997 ; Gorchetchnikov and Hasselmo, 2005 ; Qu et al., 2009 ; Ivey et al., 2011 ). To the best of our knowledge, our model is the first one to demonstrate that biologically plausible, temporally asymmetric synaptic plasticity rules can achieve this goal. Also, most of the previous models assumed multiple trials for learning a complete set of optimal paths for every new selected target location. In contrast, in our model, once an agent becomes familiar with an environment, a single passage of an activity wavefront through the network is sufficient to create a SVF guiding an animal from any possible location in the experienced environment to a target location. Interesting enough, such an ability of animals to rapidly replan routes if the starting and goal points are changed to new, random locations within a known environment has recently also been observed experimentally (Pfeiffer and Foster, 2013 ). Biological relevance Parallel exploration as proposed in our model requires mechanisms that support stable propagation of expanding waves of neural activity throughout the network. Conditions for such stable propagation of spiking activity in biological neural circuits have been examined both theoretically (Diesmann et al., 1999 ; Kumar et al., 2008 , 2010 ) and experimentally (Reyes, 2003 ; Wu et al., 2008 ; Nauhaus et al., 2012 ). Recent electrophysiological results suggest existence of expanding waves of neural activity in the hippocampus during, so called, sharp wave ripple (SWR) episodes (Ellender et al., 2010 ). Sharp wave ripples are brief high-frequency bursts of neural activity observed during sleep or at awake rest (Buzsaki, 1986 ). Hippocampal SWRs are frequently accompanied by sequential reactivation of place cells occuring in the same- or reverse temporal order as previously experienced during behavior, but replayed at a compressed time scale (Pavlides and Winson, 1989 ; Wilson and McNaughton, 1994 ; Foster and Wilson, 2006 ). Interestingly, reactivation patterns observed in the awake animals are not always just a simple function of experience (Gupta et al., 2010 ), and have also been reported to represent trajectories never directly or fully experienced by an animal, suggesting a possible role of the awake SWRs in planning, navigation or decision making (Pastalkova et al., 2008 ; Buhry et al., 2011 ; Foster and Knierim, 2012 ; Singer et al., 2013 ). These results point to the awake-state SWRs as a possible biological candidate process for parallel mental exploration as required in our model. Moreover, it has been suggested that the SWRs provide optimal conditions for the activation of synaptic plasticity processes, such as STDP (Sadowski et al., 2011 )—which, again, is consistent with our assumption that a propagating wave of neural activity should be able to modify connectivity within the network in order to create structured SVFs. The SVFs are in turn used in our model to guide behavior. Indeed we assume that the movement of an agent (an animal) is guided by the activity of places cells surrounding the present agent location. Therefore, the problem is to generate motor forces which will bring into better alignment two “bumps” of neural activity, one coming from the sensory system representing the actual location of the agent, and the other clump of neural activity having a location biased by the modified synapses. In our paper, this problem is solved by a mathematical algorithm (cf. Methods). However, neurophysiological experiments suggest that the same problem can also be solved by a biological neural network, for it is isomorphic to the problem of moving the two eyes so that the image of one bright spot is centered on both fovea (Ohzawa et al., 1990 , 1997 ). A relatively inefficient but fully neural solution to this two-bump problem was given in (Hopfield, 2010 ). As mentioned already, generation of directed connections for SVFs requires asymmetric STDP rules. Such asymmetry in the STDP learning windows has been found in the synaptic connections between hippocampal cells, first in cultured cells (Bi and Poo, 1998 ) and more recently also in slice preparations (Aihara et al., 2007 ; Campanac and Debanne, 2008 ). “Anti-” or “reverse-” STDP, in which a pairing of a pre-synaptic spike that precedes a post-synaptic spike decreases the strength of a synapse (Bell et al., 1997 ; Kampa et al., 2007 ), was used in our model to produce the SVF. There are two important reasons for why “normal” (or “pro”) STDP cannot be used in the model. If parameters are set in the fashion of (Hopfield, 2010 ) so that a clump of activity, once initiated by sensory input, is stable when sensory input is removed, that clump of activity will move, following the vector field. Thus, when the “anti” sign is used, the agent can rehearse mentally the chosen trajectory from its present location to the chosen goal. It could even, with slight elaboration, communicate a sequential list of way points. Such a natural behavior of mental rehearsal in sequential order from the starting point is not available with “pro” STDP, for the clump of activity in this case moves away from the target. Initiating a clump of activity at the target location does not create an equivalent in reverse order because the vector field diverges from that point. Another advantage of using anti-STDP over STDP is apparent for navigation in the presence of neural noise or external perturbation (physical forces pushing the agent away from the original path). When using anti-STDP, flow field lines converge when looking toward the source of the expanding circular wavefront that generated the field. When following in this direction, nearby vector field lines all converge toward the same destination, so noise is attenuated by the following process and has little effect. When following away from a source, as would be the case for normal STDP, vector field lines diverge, the effect of a noise error is amplified, and effects of noise accumulate. Our model assumes that whenever a new planning process is necessary, all synapses are reset to the baseline state and waves of activity can be initiated from the present target locations to create new SVFs. There are several candidate phenomena observed in the nervous system that could potentially realize the necessary resetting mechanism. One hypothesis, that seems to have both theoretical and experimental support, is that the population bursts during sharp wave ripples could serve this task by desynchronizing neurons through STDP (Mehta, 2007 ; Lubenov and Siapas, 2008 ). If this is the case indeed, the SWR episodes in our model would need to serve both tasks: memory erasing (hypothetically during the synchronious activation of populations of neurons) and formation of new memories (during the reactivation). To the best of our knowledge though, no such double-function of the SWR has been reported in the experimental literature so far. Another hypothetic mechanism for resetting synaptic connectivity in the hippocampus is through the neuromodulators. For example Bouret and Sara ( 2005 ) point to the role of noradrenaline in reorganizing the network structure in a way necessary for memory erasing. We recognize that not all mechanisms proposed in our work have experimental support from the studies on hippocampus. Hence, biological relevance of our model remains hypothetical. Nevertheless, we believe our approach is useful as a conceptual model, laying grounds for efficient parallel neural computation for navigation and path planning. Outlook Our model can be usefully expanded in many ways. As mentioned before, different costs can be associated with the particular pathways or spatial locations through the uneven distribution of place cells and/or uneven distribution of strength of synaptic connections. This will affect the speed and the shape of the particular wavefronts, and consequently will determine the boundaries of the basins of attraction and best path within each basin. Giving an animal the ability to actively control the speed of the wavefront propagation through the different regions of the network would provide a way to encode certain features of the environment in the path planning algorithm. Imagine that there is a cost associated with a certain path, e.g., an animal has to go through a “hazardous” area. This cost can be represented in the network through relatively weaker or “shorter” connections between neurons along this path. As a consequence, a wavefront will have a lower velocity when propagating through the place cells associated with this path, making the choice of this pathway less likely. Another possible way to dynamically control the local speed of the wavefront propagation as a function of environmental features, is by enabling interactions of the mental map considered in our present model with other mental maps, each one encoding for different features of the same environment. In this case, mental selection of particular path planning criteria (for example, “find the shortest/the fastest/the safest path”) would activate interactions between the “path planning map” and the appropriate feature maps. These interactions could be implemented through the local excitatory or inhibitory feedback loops between the “path planning” map and the selected “feature maps,” triggered by the propagating wavefront and resulting in the local changes of neuronal excitability, and so of the wavefront propagation speed in the “path planning map.” In our model we use place cells distributed uniformly, having a single spatial scale, and a simple place field in each of several separate environments. None of these are literally true in the hippocampus. However, by being an oversimplified idealization, it has allowed an exploration of rapid computational possibilities in a network that perhaps over-represents space, and seems a profligate use of neurons. An interesting extension of our work could be a hierarchical model, where space (or more generally memories) would be represented by different groups of neurons at different levels of abstraction. Several recent studies suggest that the hippocampus can encode memories at multiple levels of “resolution,” from a detailed rendition of specific places or events within a single experience, to a broad generalization across multiple environments or experiences (Steinmetz et al., 2011 ; Komorowski et al., 2013 ). Indeed, when we think about our own experience, we seem to be using a context-dependent switching between different representations of space. For example, when we plan to drive from our present location to another place in a town, we typically only focus on specific points in space when decisions about further route need to be taken (e.g., “turn left or turn right”)—at this point we typically don't think about the details of a highway we drive on, but rather on “when and where to turn or what exit to take.” To the contrary, when we need to change a lane on a highway, we quickly switch to the “high-resolution” local map and we use a spatial map of our surround to navigate between other cars and objects. A similar mechanism could be used in an extension of our model to increase efficiency of the implementation and to reduce the demand on resources (number of neurons), without compromising performance and robustness of computation. From the application point of view our neural model can be extended to the path planning problems in systems with more than two dimensions or in tasks with extra constraints, such as, e.g., non-holonomic navigation, arm movement planning. Our model, as a particular implementation of the wavefront expansion algorithm, can also be used for solving variety of optimality problems from other domains than motor control (Dorst et al., 1991 ; LaValle, 2006 )."
} | 5,493 |
25370498 | PMC4222110 | pmc | 4,333 | {
"abstract": "ABSTRACT The yeast Saccharomyces cerevisiae is a widely used cell factory for the production of fuels, chemicals, and pharmaceuticals. The use of this cell factory for cost-efficient production of novel fuels and chemicals requires high yields and low by-product production. Many industrially interesting chemicals are biosynthesized from acetyl coenzyme A (acetyl-CoA), which serves as a central precursor metabolite in yeast. To ensure high yields in production of these chemicals, it is necessary to engineer the central carbon metabolism so that ethanol production is minimized (or eliminated) and acetyl-CoA can be formed from glucose in high yield. Here the perspective of generating yeast platform strains that have such properties is discussed in the context of a major breakthrough with expression of a functional pyruvate dehydrogenase complex in the cytosol."
} | 217 |
37768063 | PMC10654089 | pmc | 4,336 | {
"abstract": "ABSTRACT Laboratory experimental evolution provides a powerful tool for studying microbial adaptation to different environments. To understand the differences and similarities of the dynamic evolutionary landscapes of two model species from the Bacillus genus as they adapt to abiotic and biotic surfaces, we revived the archived population samples from our four previous experimental evolution studies and performed longitudinal whole-population genome sequencing. Surprisingly, higher number of mutations, higher genotypic diversity, and higher evolvability were detected in the biotic conditions with smaller population size. Different adaptation strategies were observed in different environments within each species, with more diversified mutational spectrum detected in biotic conditions. The insertion sequences of Bacillus thuringiensis are critical for its adaptation to the plastic bead-attached biofilm environment, but insertion sequence mobility was a general phenomenon in this species independent of the selection condition. Additionally, certain parallel evolution has been observed across species and environments, particularly when two species adapt to the same environment at the same time. Furthermore, our results suggest that the population size might be an important driver of evolution. Together, these results provide the first comprehensive mutational landscape of two bacterial species’ biofilms that is adapted to an abiotic and biotic surface. IMPORTANCE Biofilm formation is a vital factor for the survival and adaptation of bacteria in diverse environmental niches. Experimental evolution combined with the advancement of whole-population genome sequencing provides us a powerful tool to understand the genomic dynamic of evolutionary adaptation to different environments, such as during biofilm development. Previous studies described the genetic and phenotypic changes of selected clones from experimentally evolved Bacillus thuringiensis and Bacillus subtilis that were adapted under abiotic and biotic biofilm conditions. However, the full understanding of the dynamic evolutionary landscapes was lacking. Furthermore, the differences and similarities of adaptive mechanisms in B. thuringiensis and B. subtilis were not identified. To overcome these limitations, we performed longitudinal whole-population genome sequencing to study the underlying genetic dynamics at high resolution. Our study provides the first comprehensive mutational landscape of two bacterial species’ biofilms that is adapted to an abiotic and biotic surface.",
"introduction": "INTRODUCTION Biofilms are matrix-enclosed microbial communities that adhere to biotic or abiotic surfaces. These complex assemblages confer emergent properties to their inhabitants and represent a much higher level of organization than free-living bacterial cells do ( 1 , 2 ). The biofilm lifestyle not only protects bacteria to survive in harsh environments but also facilitates the colonization of new niches during dispersal from these microbial clusters. Bacteria form biofilms on almost all natural and artificial surfaces ( 2 , 3 ), utilizing diverse mechanisms that depend on environmental conditions and specific species ( 4 ). Biofilms can be categorized into three models depending on the type of occupied niche, including pellicle biofilms at the air-liquid interface, colony biofilms at the air-solid interface, and submerged biofilms at the solid-liquid interface ( 5 , 6 ). Regardless of the type, these biofilm types can be formed under either abiotic conditions or in connection with a host. Members of the Gram-positive Bacillus genus form various types of biofilms exhibiting either beneficial or pathogenic impact. Here, we focused on two model species from the Bacillus genus and experimental evolution of their biofilms. Bacillus thuringiensis is commonly used as a biological pesticide belonging to the Bacillus cereus sensu lato group, and its spores can be isolated from diverse environments ( 7 ). Bacillus subtilis is a soil-dwelling, non-pathogenic bacterium that is commonly found in association with plants and their rhizosphere ( 6 ). Both species can form biofilms in diverse environments. They share a large number of transcriptional factors, including Spo0A, σ B , SinI, and SinR ( 8 ), which play crucial roles in the intertwined regulation of sporulation and biofilm formation ( 9 – 13 ). Between these two species, however, there are also some important differences in the regulatory pathways and transposable elements. For example, SigD, the motility-specific sigma factor, and the DegU/DegS two-component system, are absent from the species of the B. cereus sensu lato group; while the virulence regulator PlcR, which plays an important role in B. cereus and B. thuringiensis physiology ( 14 , 15 ), is absent in B. subtilis . Besides, diverse insertion sequences (ISs) have been described in different B. thuringiensis isolates ( 16 – 19 ), while only limited number of ISs have been reported in B. subtilis ( 20 ). ISs are the simplest transposable elements and play important role in shaping their host genomes ( 21 ). Insertions of IS elements can result in both gene inactivation and activation, or alter the expression of neighboring genes ( 22 ). In addition, IS-mediated changes have been described to both promote and constrain evolvability of Escherichia coli in a long-term evolution experiment (LTEE) ( 23 ). Experimental evolution has been widely used to study the evolutionary processes that occur in experimental populations in response to the surrounding environment ( 24 – 26 ). Experimental evolution provides a powerful tool to study microbial adaptation to different environments in real time. Combined with the advancement of whole-population genome sequencing, it is possible to understand the genomic dynamic of evolutionary adaptation ( 27 – 30 ). Plenty of evolutionary studies have been performed on planktonic forms of bacteria and yeasts, under diverse environmental conditions ( 27 , 30 – 35 ). In contrast, only a relatively limited number of evolution experiments have been performed with biofilm populations ( 28 , 36 – 40 ). Most of the biofilm evolution studies have focused on the quick emergence of morphotypic, phenotypic, and genotypic variation within biofilms. In biofilms, astonishing parallelism has been observed at the different biological hierarchy levels, from fitness level to gene level and even nucleotide level, both between replicate lineages within the same evolution experiment and across different evolution experiments ( 24 ). Previously, we have experimentally evolved B. thuringiensis and B. subtilis using generally comparable adaptation concepts. For B. thuringiensis, experimental evolution was performed using two approaches, a plastic bead-attached biofilm model (Bth_bead) ( 41 ) and Arabidopsis thaliana root colonization model (Bth_root) ( 42 ). Similarly, B. subtilis, a static air-medium floating biofilm transfer mode (Bs_pellicle) ( 43 ), and an A. thaliana root colonization model (Bs_root) ( 44 ) were used. Thus, for each bacterial species, biofilm formation proceeded under abiotic conditions (plastic bead or air-liquid interface) and on biotic surfaces ( A. thaliana root). In these laboratory evolution setups, 5 to 7 parallel lineages (i.e., subsequent populations with an individual evolutionary trajectory) were followed using 30 to 40 transfers. The detailed descriptions of each experiment were previously published ( 41 – 44 ). In the Bth_bead study, all evolved lineages displayed significantly enhanced biofilm production accompanied by the appearance of a B. thuringiensis fuzzy spreader (FS) colony morphotype variant. This FS variant showed higher competitive ability in most multicellular traits compared to the ancestral strain, suggesting an important role for diversification during adaptation of B. thuringiensis to the abiotic surface biofilm lifestyle. Furthermore, genetic characterization showed that the guanylyltransferase gene was disrupted by an IS element in the FS, which altered the aggregation and hydrophobicity of this variant ( 41 ). In the Bth_root experiments, bacterial lineages displayed enhanced root colonization ability compared with the ancestral strain. Single isolates from two of the evolved lineages showed higher recolonization efficiency of new roots compared with the other lineages, in addition to exhibiting altered bacterial differentiation and pathogenicity. Investigation of a key mutation in the gene encoding the Rho transcription termination factor in these lineages demonstrated how transcriptional rewiring alters cell fate decisions in B. thuringiensis ( 42 ). In the case of Bs_pellicle approach, B. subtilis diversified into four distinct colony variants that dramatically differed in biofilm formation abilities and expression of biofilm-related genes. Genome comparison suggested that major phenotypic transformations between the morphotypes can be triggered by subtle genetic differences ( 43 ). Finally, in the Bs_root study, B. subtilis was shown to rapidly adapt to the A. thaliana root environment, several evolved isolates displayed altered colony morphologies. Two selected evolved isolates from independent populations from the final transfer outcompeted the ancestor during root colonization. Re-sequencing of single evolved isolates from independent populations and different time points revealed mutations in genes related to different bacterial traits. The examined evolved isolates also displayed robust biofilm formation in response to plant polysaccharides, impaired motility, and altered growth on plant-derived compounds ( 44 ). In each experiment of these four studies, the evolved clones diversified into distinct pheno- and/or genotypic colony variants in time, including dramatically increased biofilm development of the isolated clones at the end of the experiments (Fig. S1). These findings provided novel insights into how B. thuringiensis and B. subtilis rapidly adapt to abiotic and biotic surface environment and revealed the evolutionary consequences. However, these four studies only studied the genomic changes of certain selected isolates at selected time points or only focused on certain lineages, while the full understanding of the dynamic evolutionary landscapes was lacking over the full experimental evolution setup. Furthermore, the differences and similarities of adaptive mechanisms in B. thuringiensis and B. subtilis were not identified when these two species were adapted to the same environment and when one species is adapted to two distinct environments. To overcome these limitations, we revived the archived population samples from these four studies and performed longitudinal whole-population genome sequencing to study the underlying genetic dynamics at high resolution. Generally, parallel evolution was verified in B. thuringiensis and B. subtilis when adapted to respective selective environments (i.e., mutations were detected in overlapping genes), except in the Bth_root experimental evolution system. Interestingly, we found that transposable elements in B. thuringiensis possibly play a critical role in adaptive evolution. Our study provides the first comprehensive mutational landscape of two bacterial species’ biofilms that is adapted to an abiotic and biotic surface.",
"discussion": "DISCUSSION We used longitudinal whole-population genome sequencing ( 27 , 28 , 30 ) to study the underlying genetic dynamics of the two species adapt to abiotic and biotic environment at high resolution; this method allowed us to capture the mutations with a frequency of 5%. We observed hundreds of mutations in each evolved condition: there were higher number of mutations, relatively few intergenic mutations, higher fixed mutation rate, and higher genetic diversity in biotic conditions compared with abiotic conditions in both species. In addition, there were more fixed mutations in biotic conditions in both species, and extremely more fixed cases were found in Bth_root which possibly related with the small population size. The dN/dS ratio is slightly higher in B. thuringiensis, and we only detected IS-mediated mutations in B. thuringiensis . IS elements are small and autonomous transposable elements with variable numbers and copies found in most bacterial genomes and have been reported to play important roles in shaping the genomes of their hosts ( 23 , 71 ). These simplest transposable elements usually contain only the genes responsible for their transposition ( 17 ). The IS-mediated transpositions could lead to gene inactivation or to the activation or alteration of the expression of neighboring genes ( 22 ), and consequently, the effect could be either beneficial, deleterious, or neutral. Both the number and activity of the IS elements of a given bacterium will influence the genome structure and gene expression, which will further impact the fitness of the bacterium in certain environments ( 22 , 71 ). IS-mediated changes have been described to both promote and constrain evolvability of Escherichia coli in a long-term evolution experiment (LTEE) ( 23 ), and an experimental evolution using cyanobacterium Acaryochloris marina revealed that the vast majority of beneficial mutations during laboratory evolution are due to transposition of a single IS element ( 21 ). ISs play a critical role in allowing B. thuringiensis to adapt to the plastic bead-attached biofilm environment ( 41 ). In fact, we identified a total of 13 IS transposition cases in B. thuringiensis. Notably, the rfbM gene was disrupted by IS 231 A in three out of six lineages in Bth_bead. The FS variant that was identified previously to harbor this insertion showed altered aggregation and hydrophobicity. Interestingly, this rearrangement of ISs in the genomes of the evolved lineages and final populations’ isolates showed high parallelism of adaptation. We also observed four IS-mediated mutations in Bth_root (three mediated by IS 231 A and one mediated IS 110 ), while insert positions were distributed in a different area of the genome. This implies that these ISs may only contribute little to adapt to the A. thaliana . Our study has important implications for understanding how IS elements in the B. thuringiensis genome affect the fitness, biofilm formation, and adaptation. We frequently observed clonal interference in all experimentally evolving lineages. The frequency of mutations fixation rate was different among the four conditions, and mutations seemed to be more frequent in biotic conditions while also negatively correlated with the population size. When we reviewed the population size of four experiments, we found that the biotic conditions harbored much lower population size than abiotic conditions at the end of each timepoints and each transfer. Our results suggest that a stronger population bottleneck will lead to a more frequent fixation of specific genotypes due to genetic drift. Furthermore, the difference in the frequency of fixed mutations might be explained by the different population sizes, the strong effect of competing beneficial mutations in the abiotic populations, and the strong impact of genetic drift in the biotic populations ( 72 ). We also observed nested fixations in some of the evolving lineages, with some even fixed simultaneously as cohort ( 33 , 68 ), especially in Bth_root (four of five lineages), in which the size of the fixation cohorts was the largest among all conditions (up to seven mutations fixed simultaneously). However, it is very unlikely that these mutations occurred simultaneously when considering the typical mutation rates in bacteria. One possible explanation is that the beneficial mutation(s) occurs in a background that already has certain sets of neutral or deleterious mutation, and therefore, the hitchhiker and beneficial driver mutations fix at the same time ( 68 ). We found this possible phenomenon frequently in Bth_root, in lineage D, the ywdH : T186T mutation is possibly a neutral mutation that serves as a background. Subsequently, beneficial co-drivers occurred sequentially, leading the genotype containing the cohort to fix. Although dN/dS ratios vary among the four conditions, they all approximate a value of 1, which is relatively low compared with the high dN/dS ratio observed in previous studies ( 27 , 40 ). This implies that neutral or weakly deleterious hitchhikers occurred frequently in our experimental evolution models. Thus, the higher fixation rates of mutations (or cohorts) were possibly caused by the smaller population size and strong population bottleneck (genetic drift) under the biotic conditions. We observed obvious parallelism at genotype level in Bth_bead, Bs_pellicle, and Bs_root conditions, but not in Bth_root, which is consistent with the parallelism at phenotype level. In Bth_root, isolate from lineages E and F displayed improved iterative ecesis on roots and increased virulence against insect larvae. The motility ability, sporulation kinetics, and cell morphologies were also different from other lineages ( 42 ). When we reviewed the population size of four experiments, we found that the Bth_root contains the smallest population size. It only contains from 1.13E+03 to 1.05E+06 cells at the end of each timepoints and each transfer. Theory and empirical studies suggest that strong selection and large population sizes increase the probability for parallel evolution at the phenotypic and genotypic levels ( 33 , 70 , 73 – 75 ). Differences in population sizes between populations exposed to similar conditions can affect the degree of parallelism ( 73 ). The use of different-sized experimental populations of the unicellular alga Chlamydomonas reinhardtii adapting to a high salt environment demonstrated that adaptation to salt was repeatable at the fitness level in medium and large populations, but not in small populations ( 74 ). Similarly, evolution experiments with large and small yeast populations revealed that beneficial mutations occur more consistently in larger populations ( 33 ). Furthermore, smaller population size generally led to a greater among-population variation than large population sizes do in a viral adaptation model ( 76 ). As shown above, the abiotic conditions contain much larger population size than that in biotic conditions, which is consistent with the Jaccard index ( J ) and the muller plots; thus, higher degree of parallelisms is observed in the large population size conditions in each species. Generally, as larger population size permits higher number of mutations, we would expect to detect more mutations, higher diversity, and more intense clonal interference in larger populations for both B. thuringiensis and B. subtilis in the same environment. In contrast, higher number of mutations and alpha genetic diversity could be detected in total and at the end of the experiment of each lineage in biotic condition in each species in spite of the lower population size under these conditions compared to the abiotic environment. The four experimental evolution setups used here differed in several parameters, e.g., cultivation temperature and volume, from which the biofilm surface, growth temperature, and cultural volume (which would affect nutrition supply) together may play a pivotal role in the adaptation and also possibly affect the population size. Furthermore, we speculate that the population bottleneck size is correlated with the population size, although it was not possible to calculate the exact number of cells transferred from one round to another in these experimental setups. While the B. subtilis pellicle population was consecutively diluted 100-fold each time, the generation number could be calculated. We could not determine the exact number of cells colonizing the new root or beads in the other three experimental evolution setups. The strong bottleneck imposed during the transfer process likely influenced the effectiveness of selection, making genetic drift more robust. This was especially evident in Bth_root, which showed the lowest parallelism among the four conditions, both phenotypically and genotypically ( 72 ). In addition, a sharper fluctuation of mutation number was also observed in biotic conditions, which suggests that a lot of mutations were lost during transfer process due to the small population bottleneck size in these conditions. The higher number of mutations, higher genotypic diversity, and higher evolvability (accumulated mutations in each genotype at the last timepoint) are puzzling in the populations evolved in the biotic conditions. The averaged mutation number per transfer in each lineage was calculated to be 0.60, 1.20, 0.90, and 1.67 for Bth_bead, Bth_root, Bs_pellicle, and Bs_root, respectively, nearly twice as much in biotic conditions than in abiotic. This is also possibly caused by the stronger bottleneck in biotic conditions, where deleterious and neutral mutations were easier to fix or increase to detectable frequency than that in abiotic conditions with large population size. Furthermore, the spatial niche heterogeneity in biotic conditions may also contribute to the higher genetic diversity within each lineage, while plays a less important role compared to the bottleneck size. In conclusion, our results indicate that both evolutionary (clonal interference) and ecological (population size) factors could potentially influence the genomic landscape of the evolving population of Bacilli in distinct environment. Furthermore, the different conditions could affect the degree of adaptation parallelism, while population size is an important driver of evolution."
} | 5,457 |
35469253 | PMC9034392 | pmc | 4,339 | {
"abstract": "In nature, barnacles and bacterial biofilms utilize self-assembly amyloid to achieve strong and robust interface adhesion. However, there is still a lack of sufficient research on the construction of macroscopic adhesives based on amyloid-like nanostructures through reasonable molecular design. Here, we report a genetically programmed self-assembly living-cell bioadhesive inspired by barnacle and curli system. Firstly, the encoding genes of two natural adhesion proteins (CsgA and cp19k) derived from E. coli curli and barnacle cement were fused and expressed as a fundamental building block of the bioadhesive. Utilizing the natural curli system of E. coli , fusion protein can be delivered to cell surface and self-assemble into an amyloid nanofibrous network. Then, the E. coli cells were incorporated into the molecular chain network of xanthan gum (XG) through covalent conjugation to produce a living-cell bioadhesive. The shear adhesive strength of the bioadhesive to the surface of the aluminum sheet reaches 278 kPa. Benefiting from living cells encapsulated inside, the bioadhesive can self-regenerate with adequate nutrients. This adhesive has low toxicity to organisms, strong resistance to the liquid environment in vivo, easy to pump, exhibiting potential application prospects in biomedical fields such as intestinal soft tissue repair.",
"conclusion": "4 Conclusions The development of biomimetic adhesives based on the natural adhesion systems of marine organisms is rapidly progressing. Although some recombinant or native mfps based on mussel systems can achieve adhesion strengths of 300–500 kPa or higher, but these proteins can only be produced in milligram quantities, which is difficult to meet practical applications [ 34 ]. More importantly, these in vitro expressed proteins lack the unique self-assembly and self-regeneration properties of biological systems. The shear adhesion strength of the adhesives in this study was between 100 and 300 kPa, which was similar to the reported living cell adhesive based on mussel system [ 6 ]. However, the currently reported mussel-inspired living glue system relies on the action of tyrosinase, which requires the use of high concentrations of Cu (II) to ensure enzymatic activity, and thus may cause toxicity to living organisms in practical applications. In this study, we have developed a novel living-cell bioadhesive by simulating the natural adhesion mechanism of barnacle cement and E. coli curli, using three self-assembly bioadhesive materials, cp19k, CsgA and XG. We have demonstrated that CsgA and cp19k fusion proteins have the ability to rapidly self-assemble into amyloid fibrils in a seawater environment. Amyloid fibrils can provide beneficial material properties such as resistance to degradation and mechanical strength, which allow the adhesive to remain stable in harsh environments. XG is a non-toxic natural material with good safety and biocompatibility, and its network structure gives the adhesive good thixotropy for easy pumping by injection and easy filling defects of any shape. In addition, the living cells in the bioadhesive endow it with the ability to regenerate itself. The adhesive we designed has low toxicity to organisms, strong resistance to the in vivo liquid environment, and easy pumpability, so it has potential applications in biomedical fields such as intestinal soft tissue repair. Based on flexible genetic engineering designs, we envision the future of making smart adhesives capable of environmentally-triggered repairs that could be used in a wider range of biomedical and industrial settings. For example, tissue repair by triggering adhesive regeneration in response to in vivo biochemical signals.",
"introduction": "1 Introduction Nowadays, biocompatible, processable and strong adhesives are highly demanded both in industrial and biomedical fields, such as tissue adhesives and hemostatic materials [ 1 ]. Interestingly, several natural organisms such as mussels, barnacles, bacterial biofilms and marine flatworms have provided a wealth of inspiration for the manufacture of biomimetic adhesives [ 2 ]. To date, a large number of biomimetic adhesives based on marine bioadhesion systems have been developed. Hybrid materials that combine two or more independent natural adhesion systems have the potential to combine the advantages of different systems and are therefore attractive. Recently, some researchers have used genetic engineering methods to produce recombinant proteins expressed by fusion of mussel foot protein (Mfp) and bacterial curli, which have an underwater adhesion energy approaching 20.9 mJ m −2 [ 3 ]. However, although these materials are designed in cells, they do not fully exploit the characteristics of living biological systems, and their purification process is complex and the yield is low, which cannot meet the needs of practical applications. Through synthetic biology techniques, living cells can be engineered into tunable multiscale materials with programmable properties of living systems such as self-regeneration and environmental responsiveness, which provides new ideas for the development of biomimetic adhesives [ 4 ]. As the major proteinaceous component of enteric biofilms of E. coli and Salmonella , curli is a typical example of functional adhesive amyloids, which can be used as a protein scaffold to synthesize various hybrid materials [ 5 ]. Some researchers have used the Mfp domains to functionalize curli on cell surface, thus making bacterial biofilm into living glues [ 6 ]. However, the currently reported mussel-inspired living glue system relies on the action of tyrosinase, which requires the use of high concentrations of Cu (II) to ensure the enzyme activity, which limits its practical application [ 7 ]. Barnacle cement is another attractive natural adhesion system. Barnacles can tenaciously adhere to a wide range of underwater substrates by secreting and curing multiple protein components which are integrated into amyloid-like nanofibrous network and eventually form a highly insoluble cement material [ 8 ]. In amyloid nanofibers, the β chain is oriented perpendicular to the fibril axis and connected by a dense hydrogen bond network to form a supramolecular β sheet, which increases the contact area of peptide chains to substrates and improve the adhesive ability [ 9 ]. A 19 kDa protein (cp19k) found at the barnacle cement-substrate boundary is considered to play the key surface coupling role [ 10 ]. The primary structure of cp19k has been revealed, which contains two alternating blocks, one of which is dominated by Gly, Ser, Thr, Val and Ala residues, and the other is rich in hydrophobic and charged amino acids [ 11 ]. The unique molecular design of amino acids has been proven to endow cp19k with high interfacial activity and enable it to self-assemble into amyloid fibrils in seawater analogs, exhibiting higher adhesion than before self-assembly [ 12 ]. Recently, some researchers have used engineered E. coli curli as a host system to display barnacle adhesion component AACP43, which can be exported through the periplasmic space and aggregated into fibrous structures in the biofilm [ 13 ]. Therefore, the use of barnacle cement protein-functionalized curli system is expected to make a new type of living glue. The curli on surface of engineered cells can also be used to synthesize functional composites with other inorganic or organic materials, such as minerals or metals, to enrich its applications [ 14 , 15 ]. Xanthan gum (XG) is an extracellular anionic polysaccharide purified from Xanthomonas [ 16 ]. XG has an ordered secondary structure under certain conditions, and its conformation has been identified as a network formed by self-assembly of interlaced molecular chains connected through dense hydrogen bonds, in which each single chain is formed with an intrinsic double-stranded helix [ 17 ]. Benefiting from this network structure, XG gels in aqueous solution, and its viscosity can reach more than 10 5 mPa s at low shear rate, exhibiting good thickening properties [ 18 ]. Moreover, the rich hydroxyl groups of XG provide possible sites for chemical modification and crosslinking [ 19 ]. Therefore, XG has the potential as one of the raw materials to improve the adhesion and viscoelasticity of hybrid adhesives. In view of the similar self-assembly capabilities of the three bioadhesive materials, it is theoretically possible to use barnacle cp19k, E. coli curli and XG to develop stronger hybrid adhesives than single component. In this research, we have designed a bioinspired self-assembly living-cell adhesive ( Fig. 1 ). Specifically, cp19k and CsgA recombinant proteins are fused and expressed in E. coli to work as a fundamental building block. Here, CsgA not only plays a role in accelerating amyloid self-assembly, but also acts as a platform to display recombinant protein on the E. coli cell surface and form a hierarchically assembled network of amyloid nanofibers. Then, we incorporated the E. coli cells into the network of XG via an amidation reaction, and used XG as the matrix forming material of the bioadhesive. The engineered E. coli cells fixed inside the living-cell bioadhesive can serve as a living factory which give it extraordinary properties such as genetic programmability and self-regeneration. Fig. 1 Schematics of the modular design strategy for engineering the bioinspired living-cell adhesive. The development of artificial bioadhesive is inspired by two independent natural adhesion systems- E. coli curli and barnacle cement. The genes of two kinds of natural adhesion elements (CsgA and cp19k) were fused and cloned into the pET-28a (+) and pET-22b (+) vectors for protein purification and curli formation, respectively. When incubated in vitro, the purified protein CsgA-cp19k can self-assemble into amyloid nanofibers and shows a strong surface binding strength. When delivered to the extracellular area, CsgA-cp19k can form a recombinant curli network on cell surface. By conjugation with XG, curli-producing E. coli cells can be processed into a living-cell bioadhesive, which can self-regenerate under suitable conditions. Fig. 1",
"discussion": "3 Results and discussion 3.1 Expression of recombinant proteins and characterization of in vitro self-assembly CsgA, cp19k and their fusion protein CsgA-cp19k were expressed in E. coli BL21 (DE3) ΔCsgA cells. The CsgA and CsgA-cp19k proteins were obtained at a yield of about 50 mg/L, and cp19k was obtained at a higher yield of about 160 mg/L. The SDS-PAGE result shows single bands, and the molecular weight is consistent with the prediction ( Fig. 2 a). After 24 h of dialysis against ultrapure water or artificial sea water (ASW), the soluble proteins begin to aggregate into insoluble amyloid fibrils. Circular dichroism studies ( Fig. S3 ) showed that both CsgA and CsgA-cp19k proteins dialyzed in pure water were enriched in β-sheet secondary structure with spectral minimums around 220 nm and maximums around 200 nm. Whereas a minimum around 200 nm can be observed in the cp19k spectrum, which is characteristic of random coil conformations. The shape of the CsgA-cp19k curve is slightly different from that of CsgA, possibly due to the unstructured features of the cp19k domain displayed outside the amyloid core. Thioflavin T (ThT) assay which probes for the β-sheet-rich structures of amyloid fibrils was used to monitor the in vitro self-assembly of proteins [ 23 ]. The self-assembly efficiencies of proteins in pure water and ASW were detected according to the Thioflavin T fluorescence signal, respectively. As shown in Fig. 2 , the assembly of cp19k occurs in ASW instead of pure water, indicating that cp19k requires a specific environment to trigger its amyloid formation tendency. This result is consistent with previous researchers who found that the recombinant cp19k can aggregate into typical amyloid fibrils in seawater analog [ 24 ]. In contrast, the ThT fluorescence curves of CsgA and CsgA-cp19k both rise rapidly and then reach a plateau after 12 h ( Fig. 2 b). In addition, the polymerization lag phase of Csga-cp19k is shorter than that of CsgA, which indicates that the fusion of the 19k domains accelerates the formation of amyloid. This phenomenon may be because unstructured cp19k domains enhance the efficiency of protein folding through the fly-casting mechanism, which assumes that relatively unstructured proteins can have a greater capture radius and increase protein folding rates by exploiting the available folding free energy [ 25 ]. Fig. 2 Expression of recombinant proteins and characterization of in vitro self-assembly by ThT assay and TEM observation. a ) SDS-PAGE gel verifies purification of the recombinant proteins purified by nickel-nitrilotriacetic acid (NI-NTA)-resin columns. Lane L, prestained protein ladder; lane 1, CsgA; lane 2, cp19k; lane 3, CsgA-cp19k. b ) and c ) The figures showing the relationship between Tht fluorescence (AU) of CsgA, cp19k and CsgA-cp19k proteins and time when incubated in ultrapure water or ASW. d) Transmission electron microscope (TEM) microscopic observation images of CsgA, cp19k and CsgA-cp19k proteins incubated in ultrapure water or ASW. Fig. 2 The morphological characteristics of in vitro purified proteins were further observed with transmission electron microscopy (TEM) ( Fig. 2 d). For cp19k, the protein appears as dispersed spherical particles in ultrapure water, while in ASW, short and slim fibrils (about 20–30 nm in diameter) are formed and converge into larger aggregates. Different from cp19k, CsgA and CsgA-cp19k produce a large number of fibrils with significantly higher diameter (about 40–60 nm in diameter) and fiber length, which hierarchically assemble into dense nanofiber networks in both ultrapure water and ASW. The TEM results are consistent with the ThT determination, which proves that the fusion with the CsgA domain improves the self-assembly ability of cp19k. 3.2 Characterization of surface binding ability of recombinant proteins In order to evaluate the surface binding ability of the purified proteins, a test was performed on quartz crystal microbalance with dissipation monitoring (QCM-D). The changes in frequency (ΔF) can reflect the amount of protein adsorbed on the silica surface. Compared with CsgA, the ΔF of cp19k and CsgA-cp19k dropped sharply in a short period of time, indicating that the latter two were absorbed on the silica surface and reached the adsorption equilibrium more rapidly ( Fig. 3 a). As reported, the biased amino acids in cp19k, including Thr, Ser, Ala, Val, and Lys, are particularly useful for coupling to silica surfaces via hydrogen bonding, hydrophobic interactions, and electrostatic interactions [ 26 ]. This explains the higher binding efficiency of CsgA-cp19k and cp19k to the silica surface in the test. Fig. 3 Evaluation of surface binding ability based on molecular interface interaction between protein and silica through QCM-D experiments. a) The frequency change (ΔF/n for overtone n = 3) and time curves of CsgA, cp19k and CsgA-cp19k proteins. b). The dissipation change (ΔD) and time curves of CsgA, cp19k and CsgA-cp19k proteins. c). The |ΔD/ΔF| value of CsgA, cp19k and CsgA-cp19k proteins at adsorption equilibrium before and after rinsing with pure water. Fig. 3 In addition, changes in dissipation (ΔD) can reflect viscoelasticity and stiffness of protein adsorption layers [ 27 ]. The |ΔD/ΔF| value were used to characterize the viscoelasticity of the three protein adsorption layers. As shown in Fig. 3 c, the |ΔD/ΔF| value of cp19k was the highest and the change before and after rising was the largest, indicating that the structure of the protein adsorption layer was relatively loose and most proteins were reversibly adsorbed. Relatively speaking, the adsorption layers of CsgA and CsgA-cp19k were close to rigid, indicating that proteins in form of amyloid fibrils were more likely to form a dense structure on the silica surface, and thus were more difficult to desorb. On the one hand, CsgA-cp19k can display more functional acid residues on the nanofibers by assembling into amyloid fibrils, which increases the contact area and promote non-covalent bonding with the substrate surface. On the other hand, the tape-like monomeric amyloids structure of CsgA and CsgA-cp19k may contribute to irreversible adhesion, because the formation of hinge during the desorption of nanofibers leads to an increase in peeling work [ 28 ]. This unique molecular design produces a synergistic effect to enhance adhesion performance. 3.3 Formation of recombinant curli and verification of the adhesion Utilizing the natural mechanism of curli biogenesis in E. coli , we successfully delivered CsgA-cp19k to the cell surface and verified the formation of recombinant curli by an immunofluorescence detection test ( Fig. 4 a), where green fluorescence was observed on cells expressing CsgA-cp19k and CsgA, indicating that the recombinant proteins were displayed on the cell surface and bound to fluorescent antibody via His-tag. Congo Red (CR) staining assay also confirmed the expression of curli ( Fig. 4 d). The protein composition of the curli was verified by western blotting using anti -histag antibody, and the molecular weights of the bands were consistent with the monomeric proteins, demonstrating that curli consisted of CsgA or CsgA-cp19k recombinant proteins, respectively ( Fig. 4 e). The yield of curli was quantified by a whole-cell Congo red depletion assay and the curve of curli yield versus time was plot in Fig. 4 f. Based on this curve, an induction time of 48 h was chosen to control the same yield of curli produced by CsgA-cp19k and CsgA in the following experiments. Furthermore, we morphologically observed the curli structures around E. coli cells with both transmission electron microscopy (TEM) and scanning electron microscopy (SEM) ( Fig. 4 b and c). In both TEM and SEM images, dense nanofibrous networks were observed around cells expressing CsgA-cp19k and CsgA, while no fibrous structures were found in the control group (empty plasmid transformation). These results confirm that CsgA-cp19k can be delivered to the cell surface and involved in the formation of curli. Fig. 4 Formation of recombinant curli was verified and the cell adhesion ability was confirmed by surface binding test. a ) Images of engineered E. coli cells conjugated with fluorescent antibody under fluorescence microscope, with 488 nm excitation light. b ) and c ) scanning electron microscopy (SEM) and transmission electron microscopy (TEM) microscopic images of BL21 (DE3) ΔCsgA cells with pET-22b/CsgA, pET-22b/CsgA-cp19k or empty plasmid. d ) Congo Red (CR) staining test on a LB plate. BL21 (DE3) ΔCsgA cells harboring pET-22b/CsgA, pET-22b/CsgA-cp19k or empty plasmid were coated on different areas of the plate. e ) Curli expression consisting of CsgA or CsgA-cp19k was confirmed by western blotting using anti -histag antibody. Lane L, prestained protein ladder; lane 1: empty plasmid control; lane 2, CsgA-cp19k; lane 3, CsgA. f ) The curve of curli yield versus time in BL21 (DE3) ΔCsgA cells harboring pET-22b/CsgA, pET-22b/CsgA-cp19k or empty plasmid by a whole-cell Congo red depletion assay. g ) The cells that were loaded on glass slides before washing and retained after washing in the glass binding assays. h ) Characterization of cell binding capacity by quantifying the remaining cells on the slide. (∗∗) P < 0.01. Error bars indicate s.d. Fig. 4 The interfacial adhesion ability of recombinant E. coli was verified by glass surface bonding experiment. As shown in Fig. 4 g, compared with the control group, the CsgA-cp19k and CsgA groups had significantly more recombinant cells remaining on the slide. Further quantitative analysis in Fig. 4 h showed that the cells expressing CsgA-cp19k had stronger surface binding capacity. The experimental results proved that the interfacial adhesion ability of E. coli cell can be effectively improved by delivering the recombinant protein to the cell surface and forming a fibrous network, which provides a basis for the manufacture of the bioadhesive. 3.4 Production and characterization of the living-cell bioadhesive As reported, there are hypothesized enzyme-catalyzed polymerization reactions between the proteins in barnacle cement, which enables self-assembled nanofibers to generate fibrous bulk cement through covalent cross-linking for further curing, thereby improving the cohesion of barnacle cement and the adhesion strength to the interface [ 29 ]. Here, we covalently conjugated E. coli cells and XG through an amidation reaction to form a denser cross-linked network, thus producing a living-cell adhesive ( Fig. 5 a). Through TEM and SEM observation in Fig. 5 , it was found that XG forms a dense network based on ordered double helix structure in aqueous solution. The network structure of XG was retained in the XG-cell adhesive, and a large number of XG chains were conjugated around the cell surface, which incorporate the cells into the cross-linked network. In order to verify the covalent conjugation of cells and XG, XG was directly mixed with cells as a control group, and the adhesives were repeatedly washed with pure water on the microporous membrane. As shown in Fig. 5 b, no residual XG was observed on cells after adequate washing in the directly mixed control group, while XG in the CsgA-cp19k and CsgA groups was still clearly bound to cells. In addition, no significant residual XG was observed in the empty plasmid control without curli. The SEM observation in Fig. 5 c also showed a similar phenomenon, indicating that curli plays a key role in the covalent conjugation of XG. Fig. 5 Synthesis and rheology analysis of XG-cell bioadhesive. a ) Synthesis of XG-cell bioadhesive through an amidation reaction. b ) Transmission electron microscopy (TEM) microscopic images of XG-cell bioadhesives before and after adequate washing with pure water with a directly mixed control. c ) Scanning electron microscopy (SEM) microscopic images of XG-cell bioadhesives before and after adequate washing with pure water with a directly mixed control. d ) Strain sweep of shear rheology of XG and XG-cell bioadhesives (XG-EP and XG-CsgA stand for the adhesives prepared with XG and E. coli cells with a wet weight of 0.5 g containing empty plasmid or pET-22b/CsgA, and XG-0.5 and XG-1.0 stand for the adhesives prepared with E. coli cells containing pET-22b/CsgA-cp19k with wet weight of 0.5 g or 1.0 g) measured at a constant frequency of 10 Hz and 25 °C. e ) Viscosity changes of XG and XG-cell bioadhesives from low shear rate (0.1 s −1 ) to high shear rate (500 s −1 ) at 25 °C. Fig. 5 Rheological analysis revealed the influence of XG and curli on properties of the bioadhesive. Due to the network structure, XG exhibit weak gel-like and shear-thinning in aqueous solution [ 30 ]. The rheological analysis results in Fig. 5 d showed that XG–cell adhesives retained the weak gel property of XG. The value of G′ in the figure is higher than the value of G”, indicating that the adhesives exhibit strong elastic properties. Significant differences in storage modulus were also observed between adhesives with different curli protein compositions. Compared to controls XG-EP and XG-CsgA, the G′ value of CsgA-cp19k adhesive was stronger, and it increased with the increase of cell content. This indicate that the rich amino acids of cp19k additional coupling sites and hydrogen bonds to XG, thereby enhancing the intermolecular association and increasing the cross-linking density ( Fig. 5 d). As shown in Fig. 5 e, XG-cell adhesives exhibit similar shear thinning behavior and good thixotropy as XG, which enables it to be pumped by injection under shear stress, thus facilitating practical applications. We further characterized the interface adhesion strength of XG-cell bioadhesive to aluminum or glass plates using a lap shear test on a universal tensile testing machine. As shown in Fig. 6 a, the bioadhesives prepared with CsgA-cp19k expressing E. coli cells exhibited significant higher adhesive strength than the control groups, and the adhesive strength increased with the increase of cell concentration, reaching a maximum of 278 kPa on the surface of the aluminum sheet, indicating that the cells displaying recombinant curli contributed to the surface coupling of the bioadhesive. Fig. 6 The shear adhesive strength of bioadhesives were determined by lap shear test on a universal testing machine and the self-regeneration ability of bioadhesive was verified. a ) Lap shear test of XG and XG-cell adhesive between aluminum or glass panels. XG-EP and XG-CsgA stand for the adhesives prepared with XG and E. coli cells with a wet weight of 0.5 g containing empty plasmid or pET-22b/CsgA. XG-0.25 to XG-1.0 stand for the adhesives prepared with E. coli cells containing pET-22b/CsgA-cp19k with wet weights from 0.25 g to 1.0 g. (∗∗) P < 0.01, (∗∗∗) P < 0.001. b ) The CFU number of viable E. coli cells in the bioadhesive over time when stored at 4 °C or 37 °C. c ) Detection of viable cell count and curli yield in 1–5 generations of regenerated adhesives. The CFU number of viable E. coli cells in the adhesives before and after regeneration detected by dilution coating plate method. d ) The shear adhesive strength of the 1–5 generations of regenerated XG-cell adhesives between aluminum or glass panels. Fig. 6 Rheological experiments proved that XG-cell gel exhibited a higher crosslinking density than XG gel, which helped to increase the cohesion of the adhesive, which may be one of the reasons for the enhanced adhesive strength of XG-cell gel. Interfacial coupling ability of cp19k is another important reason for the enhanced adhesive strength. cp19k possess a block copolymer-like sequence property, whose primary structure contains two alternating blocks which were believed to have a synergistic effect to promote interfacial adhesion force [ 31 ]. Among them, one block is dominated by Gly, Ser, Thr, Val and Ala residues containing large amounts of pendant amine or hydroxyl groups, which was believed to be favorable for the removal of surface-bound water layers and the interfacial couples to various foreign surfaces via hydrogen bonding, electrostatic interactions, hydrophobic interactions, etc. [ 11 ] The other block is rich in charged amino acids, which may participate in the formation of coordination bonds and help the adhesion of the metal interface. Moreover, the carboxyl and hydroxyl groups on the XG chain can also tightly bind to the interface, especially metals, through non-covalent interactions, which may also have a synergistic effect with cp19k [ 32 ]. The most attractive characteristics of living organism materials are viability and self-regeneration [ 33 ]. First, we proved that most of the E. coli cells can survive long-term in the bioadhesive when stored at low temperature ( Fig. 6 b). To examine regeneration performance, bioadhesives stored at 4 °C were passaged in LB broth supplemented with IPTG through five consecutive cycles of re-inoculation and growth. And the number of viable bacteria and curli yield in the adhesives before and after regeneration were detected ( Fig. 6 c). The results demonstrate the ability of live E. coli cells encapsulated in bioadhesive to self-regenerate and form curli under appropriate conditions. Further, we conjugated the regenerated cells with XG and used lap shear measurement to check the adhesive strength of the regenerated bioadhesive. The experimental results show that the 1–5 generations of regenerated adhesives still maintain a strong shear adhesive strength, proving the good reproducibility of the material ( Fig. 6 d)."
} | 6,975 |
34516898 | PMC8442903 | pmc | 4,340 | {
"abstract": "Biological hydroxylation, O-demethylation, and aryl side-chain oxidation were integrated in a cell for lignin valorization.",
"introduction": "INTRODUCTION Lignin, the most abundant renewable aromatic biopolymer in nature, makes up 10 to 35% of lignocellulosic biomass ( 1 ). It is estimated that there is approximately 300 billion tons of lignin source in the biosphere, and this number still increases annually, providing an enormous carbon pool ( 2 ). With the development of lignocellulose biorefineries, the cellulose and hemicellulose components of lignocellulosic biomass can be efficiently converted to bioethanol and other bioproducts. Nevertheless, most lignin components are designed to be burned for energy supply in the current biorefinery concept ( 3 ). Searching for a green and sustainable way to upgrade lignin is critical for the most complete utilization of lignocellulosic resources. Structurally, lignin is mainly composed of multiple phenylpropane derivatives. On the basis of the number of methoxy groups, these phenylpropane derivatives can be mainly divided into three units, namely, p -hydroxyphenyl (H), guaiacyl (G), and syringyl (S), which contain zero, one, and two methoxy groups in the aromatic ring. These units are linked with the chemical bonds of β- O -4, α- O -4, β-5, β-β, etc. ( 2 , 4 ). The diverse methoxy and other side-chain groups on the aromatic rings, as well as the multiple connected chemical bonds, endow lignin with a vast number of functional groups and heterogeneous properties. Plentiful functional groups in lignin—e.g., methoxy, phenolic and aliphatic hydroxyl, benzyl alcohol, noncyclic benzyl ether, and carbonyl groups—not only influence its reactivity but also lead to the different composition of monolignols in plant species ( 2 ). With the increasing focus on lignin valorization, multiple depolymerization methods have been developed to depolymerize lignin to multiple low–molecular weight fragments, which can be further converted to valuable products by chemical or biological approaches ( 5 – 9 ). Nevertheless, the heterogeneity of lignin commonly leads to multiple depolymerization products, which requires complicated processes for product separation and purification. For instance, there are generally p -coumarate, ferulate, p -hydroxybenzoate, vanillate, as well as some other low- and high-molecular lignin residues in alkaline-pretreated lignin liquors ( 10 ). It is desirable to convert multiple lignin components into single valuable products ( 11 ). Many notable studies have been performed for this purpose. For instance, lignin can be selectively transformed into terephthalic acid with combined oxidation, demethoxylation, and carbonylation of mixed lignin-derived monomers ( 12 ); into guaiacol with decarbonization and demethoxylation of lignin-derived alkyl-syringol and alkylguaiacol ( 13 ); and into phenol by removing the methoxy functionalities and alkyl side groups of lignin-derived aromatics ( 14 , 15 ). However, these conversions commonly contain several sequential reaction steps; thus, it requires more than one reactor to complete the conversion, leading to a time-consuming process. Furthermore, the diversity of lignin components often contributes to undesired side reactions if the conditions are not well controlled. Recently, the biological valorization of lignin has attracted increasing attention because some microorganisms can funnel multiple lignin components into a signal product by their elaborate metabolic pathways. As commonly recognized, lignin depolymerization in nature is mainly initiated with fungi (soft rot, white rot, and brown rot fungi, etc.), which can excrete powerful extracellular ligninolytic enzymes—including laccases, peroxidases, and some accessory enzymes—for the depolymerization of high–molecular weight lignin ( 2 , 6 , 16 , 17 ). Thereafter, some bacteria assimilate the generated low–molecular weight lignin components for carbon and energy supply, during which the lignin components are converted into carbon dioxide and some bioproducts ( 18 , 19 ). For instance, H-type monolignol p -coumarate can be converted to p -hydroxybenzoate with Fcs [feruloyl–coenzyme A (CoA) synthetase], Ech (enoyl-CoA hydratase/aldolase), and Vdh (aldehyde dehydrogenase), and the resultant p -hydroxybenzoate can be further hydroxylated to protocatechuate by PobA ( p -hydroxybenzoate hydroxylase); G-type monolignol vanillate can be O-demethylated to protocatechuate by VanAB in Pseudomonas strains ( 20 ). Compared to chemical catalysts, these biological enzymes not only have great potential in the selective conversion of lignin components but also can be integrated in a chassis cell, making biocatalysis a potentially efficient method for lignin upgrading. Here, Rhodococcus opacus PD630, a well-known strain that can use multiple lignin-derived aromatic compounds for cell growth ( 21 ), was selected as a chassis to selectively convert as many lignin components as possible into a single aromatic compound with integrated biocatalysis. Using gallate (3,4,5-trihydroxybenzoic acid)—a value-added compound that has been widely exploited in the food, pharmaceutical, and cosmetic industries ( 22 , 23 )—as a target product, we integrated three main biocatalytic reactions of hydroxylation, O-demethylation, and aryl side-chain oxidation in R. opacus PD630 by introducing exogenous biocatalytic pathways and enhancing endogenous biocatalytic systems ( Fig. 1 ). In particular, the constructed R. opacus PD630 biocatalyst can efficiently valorize alkaline-pretreated lignin and base-depolymerized ammonia fiber explosion (AFEX) lignin components to gallate with a signal bioreactor compared with traditional chemical approaches that require several reactors. This simple and efficient biocatalytic process can provide a reference for producing other aromatic products from lignin components, as well as producing target products from mixed substrates. Fig. 1. Catalyzing G-/H-/S-type lignin components into gallate with multiple-step reactions and the integrated one-pot biocatalytic route reported in this study.",
"discussion": "DISCUSSION Gallic acid is a phenolic acid widely exploited for its antioxidant, antimicrobial, anti-inflammatory, and anticancer activities ( 22 ). Currently, gallic acid is mainly obtained by hydrolysis of tannins with acids, bases, or microbial tannase ( 48 ). Nevertheless, no matter which hydrolysis method is applied, tannins are the most used substrate. The planting area and harvest time of relevant plants contribute to the fluctuation of tannin price and, further, gallic acid. Considering that lignin is the most abundant renewable aromatic resource on Earth, it is of great potential to prepare gallic acid from lignin. In this study, an efficient biocatalytic process was designed for converting lignin, the most abundant renewable aromatic in the atmosphere, to gallate with R. opacus PD630 mutants as biocatalysts. In detail, three main reactions of hydroxylation, O-demethylation, and aryl side-chain oxidation were integrated in a gallate degradation pathway–blocked R. opacus PD630 cell, and this developed biocatalyst can unify multiple G-lignin, H-lignin, and S-lignin–derived aromatics to gallate ( Fig. 9 ). Ultimately, when lignin extracted from corn stover by an alkaline method was applied as a substrate for gallate production, a gallate yield as high as 0.407 g/g of lignin was obtained. In another case, a gallate yield of 0.630 g/g of lignin was obtained when base-depolymerized AFEX lignin solution was applied as the substrate. The constructed R. opacus PD630-GA4 biocatalyst provided a sustainable and convenient approach for producing gallate from lignin resources. Fig. 9. Integrated biocatalytic reactions involved in gallate production from lignin by R. opacus PD630-GA4. Lignin was depolymerized to aromatic monomers and further catalyzed to gallate with the combined hydroxylation, O-demethylation, and oxidation systems. C23D, putative catechol 2,3-dioxygenase; P34D, protocatechuate 3,4-dioxygenase; PobA**, p -hydroxybenzoate hydroxylase (Y386F/T295A); PobA, p -hydroxybenzoate hydroxylase; Fcs, feruloyl-CoA synthetase; Ech, enoyl-CoA hydratase/aldolase; Vdh, aldehyde dehydrogenase; CalA, coniferyl alcohol dehydrogenase; CalB, coniferyl aldehyde dehydrogenase; VanA, vanillate demethylase; VanB, vanillate O -demethylase oxidoreductase; DesA, syringate O -demethylase; LigM, vanillate/3- O -methyl gallate O -demethylase; LigH, 10-formyltetrahydrofolate synthetase; MetF, 5,10-methylenetetrahydrofolate reductase; DesV, benzaldehyde-derivatives dehydrogenase; GlyA, glycine hydroxymethyltransferase; FolD, methenyltetrahydrofolate cyclohydrolase; Ru5P, ribulose 5-phosphate; Hu6P, d - arabino -3-hexulose-6-phosphate; RuMP, ribulose monophosphate pathway; PPP, pentose phosphate pathway; 3-MGA, 3- O -methylgallate; AMP, adenosine monophosphate. Numerous studies on lignin conversion have been conducted with R. opacus PD630 as a chassis cell, and some metabolic pathways for lignin-derived aromatics have been illustrated in R. opacus PD630 ( 21 , 27 , 40 – 43 , 69 ). However, the endogenous genome of R. opacus PD630 is still a valuable resource to be mined, especially for genes related to the biological degradation/conversion of aromatic and heterogeneous compounds, because this strain was originally isolated from soils of a gas-working plant. In this study, 2,3-cleavage or 5,6-cleavage of gallate was predicated in R. opacus PD630, instead of the reported 3,4-cleavage or 4,5-cleavage in Sphingomonas sp. ( 37 ), P. putida ( 38 ), and N. aromaticivorans ( 39 ). This verified gallate cleavage enzyme also exhibited extradiol dioxygenase activity on 4-methyl-catechol in our previous studies ( 27 ). With the understanding of the cleavage mode of gallate, the complete pathway for gallate biodegradation in R. opacus PD630 can be unraveled in future studies. With the gallate degradation pathway–blocked R. opacus PD630 as a chassis, a series of recombinant strains were constructed, and then multiple aromatic monomers, as well as two types of lignin, were applied as substrates to produce gallate with constructed R. opacus PD630 stains as biocatalysts. As a result, most tested aromatic monomers led to high gallate yields. In contrast, low gallate yields resulted from raw AFEX lignin samples. Nevertheless, alkali lignin and base-depolymerized AFEX lignin both led to high gallate yields. On the basis of previous studies and the process analysis in this study, lignin can be depolymerized to multiple aromatic monomers in alkaline conditions, which can be easily assimilated by R. opacus PD630. However, the inherent lignin-depolymerization ability of native R. opacus PD630 was weak, and, thus, many lignin oligomers and polymeric lignin remained unused. In nature, some fungi and bacteria can depolymerize lignin to low–molecular weight components by secreting laccase, peroxidase, dye-decolorizing peroxidase, etc. Consolidated simultaneous lignin depolymerization and product generation bioprocesses have been proven effective in some previous studies ( 16 , 70 ). In this study, we also tried to construct a consolidated lignin biological conversion process for gallate production by adding laccase into the culture broth. However, there was no obvious increase in gallate production. Instead, decreased gallate productions were observed in some cases. Further studies demonstrated that laccase not only depolymerized lignin but also functioned on gallate. Therefore, it is not a practical approach to produce aromatic compounds sensitive to oxygen from lignin by adding ligninolytic enzymes because these enzymes generally can also function on aromatic products. To produce these compounds from lignin resources, enhancing lignin depolymerization with chemical approaches for microbial utilization is straightforward and effective ( 28 ). In this study, the gallate yields were relatively high when low concentrations of substrates (both aromatic monomers and alkaline-pretreated lignin) were applied. In contrast, when high concentrations of substrates were applied, obvious decreases in gallate yields were observed. The decreased gallate yields might be due to a comprehensive result of several factors: (i) As the main supplier for carbon source, ATP, cofactors, etc., the initially added glucose was consumed up within 8 to 20 hours in most cases, resulting in a weak cell activity and hence a low substrate conversion efficiency. Maintaining a reasonable glucose level in culture broth will be beneficial to a high gallate yield. (ii) The mutant enzyme PobA** developed in this work exhibited high activity for protocatechuate at low concentrations. When the concentration of protocatechuate increased, the conversion of protocatechuate to gallate was inhibited, which was consistent with the accumulation of protocatechuate in some cases. Protein engineering on improving the activity of PobA** is essential for producing gallate from high concentrations of protocatechuate and other lignin-derived aromatics. (iii) Target genes were inserted into a free replicable plasmid for expression in this study, and a dose of kanamycin was added in the culture broth to maintain strain stability. A genome-integrated gene expression cassette may be helpful for both stable expression of target genes and reducing the expense on kanamycin, especially in industrial applications at a large scale. (iv) The autoxidation of gallate was observed with a high gallate titer. Controlling the dissolved oxygen at a low level may prevent autoxidation and lead to a higher gallate titer. Similarly, the pyrogallol titer was enhanced by adding a suitable oxygen-scavenging agent, ascorbic acid, to alleviate the autoxidation of produced pyrogallol ( 71 ). (v) The toxicity of the produced gallate toward host cells mainly limited the production of gallate. Converting toxic compounds to their nontoxic analogs is an alternative way to solve these issues ( 72 ). In some plants, gallate is converted to the β-glucogallin under gallate 1- O -galloyltransferase, which can provide a feasible approach for microbial gallate detoxification ( 73 ). Moreover, adaptive evolution is also a common strategy to improve microbial tolerance to multifarious aromatic compounds ( 72 ). The adaptive evolution for the constructed R. opacus PD630 will be beneficial for both improving its tolerance to the produced gallate and aromatic substrates. This work also provides valuable information for the valorization of other resources. For instance, the O-demethylation of methoxy aromatics also plays a vital role in C1 compound utilization ( 59 ); the β-oxidation of long aliphatic chain–substituted aromatics is an alternative method for acetyl-CoA generation. Converting lignin into biobased polymers or polymer building blocks has recently drawn enormous interest, during which the functionalization of phenol moieties was the main strategy for polymer production ( 7 , 64 ). Both of the mentioned biological hydroxylation and O-demethylation systems in this study can be useful for efficiently introducing phenolic hydroxyl groups into lignin-derived phenolics. In addition, this study can also provide a reference for the production of other aromatics from lignin resources, as well as for the utilization of other resources with mixed components."
} | 3,866 |
32648765 | null | s2 | 4,341 | {
"abstract": "Nanodiscs (ND) are soluble phospholipid bilayers bounded by membrane scaffold proteins; they have become invaluable in the study of membrane proteins. However, this multifunctional tool has been used individually, and applications involving multiple NDs and their interactions have fallen far behind their counterpart membrane model system: liposomes. One major obstacle is the lack of reliable methods to manage the spatial arrangement of NDs. Here we sought to extend the utility of NDs by organizing them on DNA origami. NDs constructed with DNA-anchor amphiphiles were placed precisely and specifically into these DNA nanostructures via hybridization. Four different tethering strategies were explored and validated. A variety of geometric patterns of NDs were successfully programmed on origami, as evidenced by electron microscopy. The ND ensembles generated in this study provide new and powerful platforms to study protein-lipid or protein-protein interactions with spatial control of membranes."
} | 250 |
35424317 | PMC8694029 | pmc | 4,342 | {
"abstract": "In this paper, experimental and theoretical studies of the piezoelectric effect of two-dimensional ZnO nanostructures, including straight nanosheets (SNSs) and curved nanosheets (CNSs) are conducted. The results show that the CNSs have a great advantage in piezoelectric property over the SNSs; the maximum output current of the NG based on CNSs was measured to be about 260 nA, much higher than that generated by SNSs. For comparatively analyzing the working mechanics of both NGs, the piezopotential distribution of both CNS and SNS structures was studied using the finite element method. The simulation result that the piezopotential generated by CNSs is always much larger than that generated by SNSs in the case of lateral bending, has more advantages for piezoelectric NGs than the SNSs. This work may provide guidance for structural optimization of piezoelectric nanogenerators and designing high-performance self-powered strain sensors.",
"conclusion": "5. Conclusions In summary, we have successfully grown straight nanosheets (SNSs) and curved nanosheets (CNSs) by a simply hydrothermal method. The electrical output signals of the NGs based on both structures were tested. The result shows that the maximum output current of the NG based on CNSs was measured to be about 260 nA, higher than that of NG based on SNSs (150 nA), which was sufficient to drive some micro/nano electronic devices. In order to explain the difference of output current caused by different structures, we used COMSOL software to simulate the piezopotential and displacement distribution of two nanosheets under the same conditions. The result shows that, in case of vertical compression, the piezopotentials generated by the two nanosheets have little difference in magnitude and distribution, while in the case of lateral bending, the piezopotential generated by CNSs is always much larger than that generated by SNSs. It can be seen from the displacement diagram that CNSs are prone to local strong deformation due to its unique curved structure in case of lateral bending. By comparing the same size CNSs and SNSs, we suggest that the CNSs have more advantage for designing high-performance piezo-nano/micro-devices. This investigation plays an important role in guiding the development of high-performance piezoelectric NGs and self-powered strain sensors.",
"introduction": "1. Introduction With the rapid development of science and technology, more and more electronic devices are moving towards miniaturization and portability. 1 Nowadays, multiple types of micro/nano robots and detectors which allow in situ , real-time safety monitoring and precise measurement, have been widely applied in many fields including national defense, energy, chemical engineering and electrical power, and the medical industry. 2,3 While such micro/nano devices still need an external power source, at present, there exists a lot of inconvenience in charging or changing batteries for electronic devices, owing to their short battery life, environmental pollution problems and so on. Therefore, providing continuous and self-powered electric energy has become a new target for scientists. As we all know, there is a lot of green renewable energy available in the environment including solar energy, wind energy, nuclear energy, heat energy, and mechanical energy. 4–6 Among them, mechanical energy is one of the most abundant and accessible energy resources in our daily life, 7 for example body movement, vibrations, acoustic/ultrasonic waves and so on. Since Wang and Song's pioneering work on the first nanogenerator (NG) prototype using ZnO piezoelectric nanowires (NWs) for converting mechanical energy into electricity, 8 NGs based on piezoelectric effect have emerged as one of the most promising approach to harvest ambient mechanical energies. 9–12 Among other piezoelectric materials, 11,13–16 ZnO with the direct wide band gap of 3.37 eV and large exciton binding energy of 60 meV, has aroused extensive research interests because of its excellent piezoelectric property as well as excellent electron transport property. 17–19 Recently, various approaches have been explored to manufacture piezoelectric NGs based on various ZnO nanostructures, such as ZnO nanowires, and nanotubes, nanosheets, nanotrees, and so on. 8,20–22 Especially, the emerging 2D nanostructures have many attractive properties such as high surface-to-volume ratio and good mechanical durability for applications in energy conversion and storage devices, etc. 23–26 In recent years, our group have presented DC power generation based on 2D ZnO nanosheets by a two-step low temperature growth method, and different 2D nanosheets (such as nanosheets, nanowalls, and ultra-thin nanosheets) were also compared and analyzed. 24 Our study shows that the appropriately tuning the morphology of the ZnO nanosheets is an effective way to increase the piezoelectric output performance. Inspired by this work, in this paper we have developed a low-temperature method to grow ZnO nanogrids on aluminum foil substrate which comprises the curved nanosheets (CNSs), and furthermore, comparative research between common straight nanosheets (SNSs) and the CNSs are also conducted in our experiment experimentally and theoretically.",
"discussion": "3. Results and discussion \n Fig. 1(a–d) demonstrates the SEM images of as-prepared CNSs and SNSs under different magnifications. We can see it clearly that CNSs look like nanogrids consisting of many curved nanosheets while SNSs are straight and interwoven. And they were both densely grown on the substrates. The arc length and thickness for the CNSs are about 1–2 μm and 50–100 nm in size. The average length of SNSs is 0.5–1 μm with a thickness in the range of 50–80 nm. The energy-dispersive spectrum (EDS) was also characterized for CNSs, as shown in Fig. 1(e) , from which we can find that only four elements: Zn, Al, O and C can be observed from the spectrum. To characterize the crystal structure of CNSs, XRD pattern of the CNSs on Al foil substrate was investigated in Fig. 1(f) . From Fig. 1(f) , we can see that the intensity of (100) and (101) peak are higher than that of (002) peak, indicating that not all ZnO grew along the c -axis, which is related to the formation of the 2D lamellar structure. This result is in good agreement with the previous analysis. 27 Fig. 1 (a and b) SEM images of CNSs under different magnifications; (c and d) SEM images of SNSs under different magnifications; (e) typical EDS of the as-prepared ZnO CNSs; (f) XRD pattern of ZnO CNSs on the Al foil substrate. For measuring the electric output performance generated from both NGs based on SNSs and CNSs, a manually applied pressure was periodically introduced to deform the device with a specially designed measuring setup, 28 so that the NGs could experience a cycling pressing-releasing process. The measuring setup is consisted of a fixation, a platform, and a screw to apply a certain amount vertical force along the z -axis on a thin nanofilm. When the devices were pressed and released repeatedly, the DC-type current outputs were observed. Fig. 2 illustrate the current output signal of the as-fabricated NGs, when we press and release the NGs by rotating the screw at half circle (∼1 kgf) and releasing the screw. We can discover that both NGs generated DC type current signals. And the amplitude of signal fluctuates were about 150 and 260 nA, respectively. Since both of our nanofilms have an area of 0.8 × 0.8 cm 2 , so here the comparison of current is reasonable when approximately considering the same density of the two nanosheets. The time intervals of recovery time artificially controlled are relatively short, which means that both NGs have high sensitivity. Fig. 2 The output performance for NGs based on SNSs (a) and CNSs (b)."
} | 1,949 |
39674656 | PMC11387356 | pmc | 4,344 | {
"abstract": "Graphical abstract",
"conclusion": "14 Conclusion Microalgae have drawn the attention of researchers for reasons other than fitness advantages and reevaluation. This study covers the most recent findings on organic hobby and function in addition to highlighting the bioactive components and organic habitats of marine microalgae. Polysaccharides, lipids, proteins, and pigments are examples of bioactive materials. In a few scientific and commercial domains, several physiological and pharmacological sports (such as nitrogen fixing, anticancer, antioxidant, antiviral, anti-inflammatory, and anticoagulant) are packaged. These marine microalgae bioactive compounds offer a fresh and substantial reassessment for enhancing commercial production and personal fitness care. Additionally, the extraction, production, and pricing of micro-algae active compounds are determined by their unique, powerful effects; similarly, the era of manufacturing optimization necessitates intensity control. A micro-algae business may be focused on the market, regardless of whether it sells nutrition and fitness products or cosmetics and pharmaceuticals. Certain demanding circumstances must be met to comply with this market trend. For example, early cultivation grade viable varieties must be chosen, and producers must begin working on improving their products. This pick will not be smooth at this time due to the range of microalgae. Furthermore, the extraction, production, and pricing of active substances derived from microalgae are determined by their unique and powerful effects; this intensity control is also necessary in the era of manufacturing optimization.",
"introduction": "1 Introduction The first members of the aquatic ecosystem and the progenitors of modern land plants are microalgae. Microalgae are unicellular microbes that can produce chemical force from solar energy and engage in photosynthesis. One of the main sources of oxygen produced in the atmosphere is microalgae. Unlike the superior flora, these organisms have a bloom and contribute additional biomass that is rich in bioactive chemicals. Microalgae may even domesticate in waste or saline water. They can even adapt to harsh environmental circumstances and cease to compete with farmers. Algal biomass has been utilized as a major food source and medication since the early 1500 BCE. The first microalgae were employed as a necessary food ingredient and pharmaceutical. 1 Numerous expensive bioactive substances, including proteins, carbohydrates, lipids, essential fatty acids, pigments, vitamins, antioxidants, and so forth, can be produced by microalgae. 2 The food, cosmetics, feed, biofuels, nutraceutical, and pharmaceutical industries have all made extensive use of microalgae due to these reasons ( Fig. 1 ). 3 Moreover, it has been applied to the production of biofuel, wastewater treatment, carbon dioxide mitigation, and bioremediation. 4 Fig. 1 Industrial applications of micro-algae. Nevertheless, producing strong derivatives from microalgae may not always be environmentally beneficial if it does not result in the production of an additional metabolite at a higher cost. The main risks associated with commercializing micro-algal products are their small market size, high manufacturing costs, low biomass, accumulation of products in large-scale cultivation, and overly restrictive regulations. 5 Nonetheless, recent studies in finance have demonstrated that the development of microalgal industries is a rapidly expanding industry in the market. By 2026, it is projected that the value of the global micro-algal product market will surpass seventy-five million US dollars. Bioactive micro-algae compounds have garnered increased attention and research efforts in the past few years. Extracted and refined from micro-algae, high-quality, expensive metabolites with nutritional and medicinal properties include proteins, polysaccharides, polyunsaturated fatty acids, polyphenols, vitamins, minerals, carotenoids, and more. 6 Metabolites generated from microalgae have been demonstrated to have potential use in the management and avoidance of several illnesses, including diabetes, heart disease, autoimmune diseases, rheumatoid arthritis, anemia, obesity, dementia, and a variety of neurodegenerative diseases. 7 To comprehend the function of synthetic goods, costly biosynthetic pathways from herbal components are artificially built in microalgae due to the rapid advancement of synthetic biology. 8 This manufacturing technique is more economical and quicker. For instance, students specifically understand that metabolic engineering increases the production of plant flora in the sentences about enhancing the productivity of herbal biosynthetic nutrition E. 9 The literature makes it evident that in the same historical circumstances, microalgae have been regarded as dietary supplements. Humans from Mexico first reported using spirulina as a food ingredient in 1300 CE It is also clear that African humans included spirulina in their regular diet. People ate Nostock as a supplemental meal from China, Mongolia, and South America. Blue algae were used by humans to create the dry cake known as “tequila” in Spain. As a historical illustration, spigyra has been a culinary element in places like Vietnam, Burma, and India. 10 Japanese people have been using cyanobacteria for a long time to cook their “Suizenjinori” home meal. Therefore, it is clear that the application of cutting-edge biotechnology features in the utilization of microalgae has existed for some time. A symposium on the collective lifestyle of algae, held in 1952 by Stanford University and Inside the United States, opened up a whole new avenue for the production and marketing of microalgae. A company named Nihon Chlorella was established in Japan in the 1960 s to cultivate chlorella in large quantities for packaged meals. Then, near Lake Texcoco, Mexico, a business by the name of Sosa Texcoco S.A. began cultivating Arthrospira in the 1970 s. More than 1,000 kg of microalgae are produced monthly by 46 large-scale microalgae cultivation plants in Asia, which were established in the 1980 s. Later, Dunaliella salina was widely cultivated to produce beta-carotene and eventually rose to prominence as a major producer of microalgae metabolites. Large-scale industries in India have begun to cultivate microalgae. 1 The algae biotechnology sector has grown rapidly in the last few years. With a turnover of almost $1.25 x 109 in a year, the micro-algae manufacturing market has grown to about 5,000 dry lots/12 months. 1 However, low biomass manufacturing and issues with product recovery—which will raise the cultivation's worth and the final product prize—are the main disadvantages of large-scale microalgal growing structures. Nonetheless, scientists concentrate on refining contemporary approaches to enhance the aesthetics of microalgae biomass industrialization and diminish the financial benefits of microalgae cultivation. 6 Expensive metabolites Micro-algae are also utilized as biofertilizers and to enhance nutrients in a variety of industries, including medicines, cosmetics, aquaculture, and poultry feed. 11 Therefore, the current paper outlines the current state of algal biotechnology, the expanding commercial applications for microalgae, the main challenges, and the future applications for microalgal products in the global market."
} | 1,847 |
31380089 | PMC6662555 | pmc | 4,345 | {
"abstract": "Abstract Anthozoans are a class of Cnidarians that includes scleractinian corals, anemones, and their relatives. Despite a global rise in disease epizootics impacting scleractinian corals, little is known about the immune response of this key group of invertebrates. To better characterize the anthozoan immune response, we used the model anemone Exaiptasia pallida to explore the genetic links between the anthozoan–algal symbioses and immunity in a two‐factor RNA‐Seq experiment using both symbiotic and aposymbiotic (menthol‐bleached) Exaiptasia pallida exposed to the bacterial pathogen Vibrio coralliilyticus . Multivariate and univariate analyses of Exaiptasia gene expression demonstrated that exposure to live Vibrio coralliilyticus had strong and significant impacts on transcriptome‐wide gene expression for both symbiotic and aposymbiotic anemones, but we did not observe strong interactions between symbiotic state and Vibrio exposure. There were 4,164 significantly differentially expressed (DE) genes for Vibrio exposure, 1,114 DE genes for aposymbiosis, and 472 DE genes for the additive combinations of Vibrio and aposymbiosis. KEGG enrichment analyses identified 11 pathways—involved in immunity (5), transport and catabolism (4), and cell growth and death (2)—that were enriched due to both Vibrio and/or aposymbiosis. Immune pathways showing strongest differential expression included complement, coagulation, nucleotide‐binding, and oligomerization domain (NOD), and Toll for Vibrio exposure and coagulation and apoptosis for aposymbiosis.",
"conclusion": "5 CONCLUSION Exposure to live Vibrio coralliilyticus for both symbiotic and aposymbiotic anemones had strong and significant impacts on gene expression, but their effects were independent or additive, not interactive. The pathways most affected by Vibrio exposure were the complement and coagulation cascades, NOD/Toll receptor signaling, and apoptosis. Despite the absence of canonical NOD and Toll receptors in the Exaiptasia genome, the downstream signaling indicates involvement of NOD and Toll pathways in the anthozoan immune response. Future studies will be required to determine if Exaiptasia possess some functional equivalent to NOD‐like and Toll‐like receptors, and if so, how such receptors interact with downstream signaling pathways. Aposymbiosis resulted in the up‐regulation of genes within the coagulation cascade and pro‐apoptotic P53 as well as down‐regulation of antiapoptotic BIRC5, indicating that menthol‐induced bleaching may involve apoptotic mechanisms similar to those involved in thermal stress‐induced bleaching. While we did not see the interaction that we expected between symbiotic state and response to a pathogen, this study provides additional data points to better understand both bleaching and pathogen response in anthozoans.",
"introduction": "1 INTRODUCTION Cnidarians represent one of the earliest animal groups (Steele, David, & Technau, 2011 ) and thus are ideal systems to study the origins of genetic processes like innate immunity (Bosch, 2013 ; Hemmrich, Miller, & Bosch, 2007 ; Lehnert, Burriesci, & Pringle, 2012 ). Initial characterizations of cnidarian immune genes indicated that they possess key components of the major innate immune pathways including Toll/TLR pathway, complement C3, membrane attack complex/perforin domains, and other components of innate immunity once thought to have evolved much later (Miller et al., 2007 ; Nyholm & Graf, 2012 ; Putnam et al., 2007 ; Shinzato et al., 2011 ); yet, it was not known whether cnidarians used these immune pathways to mount a response against pathogens. A number of groups have since used RNA‐Seq data to produce some of the first profiles of anthozoan innate immunity (Anderson, Walz, Weil, Tonellato, & Smith, 2016 ; Burg, Prentis, Surm, & Pavasovic, 2016 ; Fuess, Mann, Jinks, Brinkhuis, & Mydlarz, 2018 ; Fuess, Pinzón, Weil, Grinshpon, & Mydlarz, 2017 ; Libro, Kaluziak, & Vollmer, 2013 ; Libro & Vollmer, 2016 ; Pinzón et al., 2015 ; Poole & Weis, 2014 ; Vidal‐Dupiol et al., 2011 ; Weiss et al., 2013 ). To date, at least nine studies have profiled the immune response of corals and their anthozoan relatives, and the data suggest that the immune response varies across anthozoans and/or immune exposures. For example, Weiss et al. ( 2013 ) studied the response of the reef coral Acropora millepora to the bacterial cell wall derivative muramyl dipeptide (MDP) and observed the up‐regulation of GTPases of immunity‐associated proteins (GiMAPs), which are primarily associated with immunity in vertebrates (Wang & Li, 2009 ) and plants (Liu, Wang, Zhang, & Li, 2008 ). Vidal‐Dupiol et al. ( 2011 ) compared the transcriptomic responses of the reef coral Pocillopora damicornis to thermal stress and Vibrio coralliilyticus infection and observed that immune pathways—including Toll/TLR, complement, prophenoloxidase, and the leukotriene cascade pathways—were up‐regulated due to Vibrio exposure. Libro et al. ( 2013 ) compared the immune response of healthy and White Band Disease (WBD) infected Acropora cervicornis coral using RNA‐Seq and found that C‐type lectins, ROS production, arachidonic acid metabolism, and allene oxide production were strongly up‐regulated in diseased corals (Libro et al., 2013 ). Up‐regulation of C‐type lectins and ROS production are hallmarks of phagocytosis, and the metabolism of arachidonic acid via the allene oxide pathway has been linked to eicosanoid synthesis in wounded corals (Lõhelaid, Teder, Tõldsepp, Ekins, & Samel, 2014 ). Interestingly, Libro et al. ( 2013 ) did not identify strong up‐regulation of genes associated with the classic innate immune pathways such as Toll‐like receptor pathway or prophenoloxidase pathway. Reef‐building corals and other anthozoans like the symbiotic anemone Exaiptasia are also well known for their symbiotic relationship with the dinoflagellate Symbiodinium (also called zooxanthellae). This symbiosis presents a challenge with regard to the immune system because both pathogens and symbionts can elicit an allorecognition response, with the difference being that pathogens are typically eliminated, while symbionts are allowed to coexist within vacuoles in the endodermis of host cells (Kazandjian et al., 2008 ; Wakefield, Farmer, & Kempf, 2000 ) providing the anthozoan host up to 95% of its energy as translocated polysaccharides (Falkowski, Dubinsky, Muscatine, & Porter, 1984 ). Symbiosis requires clear communication between the host and symbiont. During the establishment of symbiosis, the anthozoan host must be able to recognize symbionts, engulf them in phagosomes, and shield these phagosomes from destruction (Davy, Allemand, & Weis, 2012 ). This suggests a clear link between symbioses and immunity wherein symbionts evade the immune response. Arrest of phagosomal maturation by Rab GTPases (Davy et al., 2012 ) and suppression of immune responses by transforming growth factor beta (TGF β ) (Detournay, Schnitzler, Poole, & Weis, 2012 ) have been identified as potential mechanisms by which symbionts are shielded from destruction by the immune system. Once symbiosis is established, the host must regulate the growth of the symbionts and remove dead or dying symbionts (Davy et al., 2012 ). Regulation of nutrients has been identified as one mechanism by which the host can prevent overgrowth of the dinoflagellates (Davy et al., 2012 ). Bleaching occurs when the symbionts are degraded or expelled by the coral host due to factors like thermal stress (Fitt, Brown, Warner, & Dunne, 2001 ), UV exposure (Gleason & Wellington, 1993 ), and disease (Libro et al., 2013 ). In addition to these naturally occurring stressors, chemical agents have been identified to deliberately induce bleaching in the laboratory for manipulative studies. These include menthol (Wang, Chen, Tew, Meng, & Chen, 2012 ) and photosynthesis inhibitors (Jones, 2004 ) that result in bleaching. Several mechanisms have been identified that result in the degradation and expulsion of the symbionts, including apoptosis, necrosis, and symbiont digestion via autophagy (symbiophagy; Dani et al., 2016 ), and the mechanisms vary depending on the type of stress. Apoptosis and necrosis predominate in heat‐stress bleaching, while symbiophagy predominates in menthol bleaching (Dani et al., 2016 ; Wang et al., 2012 ). Arrest of phagosomal maturation is required for the establishment of symbiosis, and Dani et al. ( 2016 ) suggest that a re‐engagement of phagosomal maturation is involved in the breakdown. A number of transcriptomic studies of anthozoan bleaching have shown varied immune responses. Mansfield et al., 2017 found that NF‐κβ protein levels increase after bleaching and decrease after re‐colonization in Exaiptasia . Pinzón et al. ( 2015 ) found that 1 year after a bleaching event in Orbicella faveolata colonies, 17 immune genes within tumor necrosis factor pathway, apoptosis, cytoskeleton, transcription, signaling, and cell adhesion and recognition were down‐regulated. Seneca and Palumbi ( 2015 ) found that the transcriptome response of Acropora hyacinthus exposed to heat varied widely between the initial heat exposure and the bleaching response 15 hr later, and the later response included up‐regulation of immune and apoptosis pathways including Toll‐like receptor and C‐type lectins. In this study, we explore whether breakdown of symbiosis triggered by exposure to menthol alters the subsequent immune response to the coral pathogen Vibrio coralliilyticus . The symbiotic anemone Exaiptasia has become a powerful model for studying symbiosis and immunity in symbiotic anthozoans because (a) it is a hardy animal that can be made aposymbiotic experimentally by exposing it to cold and heat stress (Lehnert et al., 2014 ), as well as by treating it with compounds like menthol (Matthews, Sproles, & Oakley, 2016 ), (b) it can be propagated clonally (Sunagawa et al., 2009 ), and (c) a well‐annotated genome for Exaiptasia now exists (Baumgarten et al., 2015 ). Limited gene expression data also exist for Exaiptasia comparing aposymbiotic and symbiotic anemones (Lehnert et al., 2014 ), Exaiptasia exposed to pathogens (Poole, Kitchen, & Weis, 2016 ), and Exaiptasia colonized by heterologous symbionts (Matthews et al., 2017 ). Lehnert et al. ( 2014 ) used RNA‐Seq to compare symbiotic and aposymbiotic anemones and identified 900 differentially expressed genes involved in metabolite transport, lipid metabolism, and amino acid metabolism. Poole et al. ( 2016 ) used qPCR to compare complement activity in response to colonization with Symbiodinium and the response to pathogen exposure ( Serratia marcescens ). Within the complement pathway, B‐factor 1 and MASP were up‐regulated and B‐factor 2b down‐regulated in response to both pathogen exposure and symbiont colonization. Matthews et al. ( 2017 ) used RNA‐Seq to profile immune and nutrient exchange activity in response to colonization with Symbiodinium trenchii versus its normal symbiont, Symbiodinium minutum . The expression pattern after colonization with the heterologous S. trenchii was intermediate between the aposymbiotic state and the normal ( S. minutum ) symbiotic state, with up‐regulation of innate immune pathways in response to heterologous colonization. In this study, we explore the genetic links between the anthozoan–algal symbioses and immunity in a two‐factor RNA‐Seq experiment using both symbiotic and aposymbiotic Exaiptasia exposed to the bacterial pathogen Vibrio coralliilyticus . Menthol bleaching was used to compare symbiotic (untreated) versus aposymbiotic (menthol‐treated) anemones where the hypothesized mechanism of menthol bleaching is thought to be the activation of autophagic digestion of Symbiodinium cells (symbiophagy) as part of host innate immunity (Dani et al., 2016 ). The bacterial pathogen Vibrio coralliilyticus was used to initiate the immune response of Exaiptasia 72 hr after exposure to menthol. The two‐factor design comparing Vibrio and aposymbiosis as factors allowed us to identify gene expression patterns that were due to Vibrio and/or symbiotic state as well as any interactions between pathogen exposure and symbiotic state.",
"discussion": "4 DISCUSSION Multivariate and univariate analyses of Exaiptasia gene expression demonstrated that exposure to live Vibrio coralliilyticus had strong and significant impacts on transcriptome‐wide gene expression for both symbiotic and aposymbiotic anemones, but interestingly, we did not see significant interactions between pathogen exposure and symbiotic state. In all, there were 4,164 DE genes for Vibrio, 1,114 DE genes for aposymbiosis, and 472 DE genes for the additive combinations of Vibrio and aposymbiosis. KEGG enrichment analyses identified 11 pathways—involved in immunity (5), transport and catabolism (4), and cell growth and death (2)—that were enriched due to Vibrio and/or aposymbiosis. Seven pathways were enriched for both Vibrio and aposymbiosis (complement and coagulation cascades, chemokine signaling, endocytosis, lysosome, peroxisome, apoptosis, P53 signaling), indicating overlapping genetic responses between pathogen infection and aposymbiosis. Four gene pathways were enriched for Vibrio alone (antigen processing and presentation, leukocyte transendothelial migration, NOD‐like receptor signaling, phagosome), demonstrating independent genetic responses underlying pathogen infection. Yet, over‐representation of DE genes was only significant for Vibrio exposure for complement and coagulation cascade pathway. Pathway level responses in gene expression are discussed further below. 4.1 Immune system response Among the immune pathways, there was strong evidence that the complement and coagulation cascade was responding to both Vibrio and aposymbiosis, whereas expression of NOD/TLR pathway, chemokine, and antigen processing was initiated primarily by Vibrio exposure. The stimulation of complement and coagulation cascade pathway and NOD/TLR pathway indicates that bacterial immune challenge by Vibrio involves two of the three primary innate immune pathways in invertebrates; we did not find significant evidence for stimulation of the prophenoloxidase (PPO) activating system (i.e., melanization). The absence of a transcriptomic PPO response was surprising, because enzymatic assays of Vibrio ‐infected Exaiptasia (10 6 cfu/ml) showed a tenfold increase in PPO enzymatic activity relative to controls (Zaragoza et al., 2014 ), and PPO has been shown to be up‐regulated in some hard and soft coral immune responses (Palmer & Traylor‐Knowles, 2012 ). Differences in immune responses would be expected between Exaiptasia and other symbiotic anthozoans, but the conserved immune features identified in the Exaiptasia genome (Baumgarten et al., 2015 ) support Exaiptasia as a model for anthozoan immune responses. 4.2 Complement and coagulation cascade Patterns of gene expression in the complement and coagulation cascade indicate that coagulation is initiated by Vibrio and aposymbiosis, whereas the complement alternative pathway is initiated primarily by Vibrio exposure. Out of the four highly expressed DE coagulation genes that differed due to Vibrio exposure and aposymbiosis, three genes (VFW, A2MG, TFPL1) have previously been associated with immune challenge in anemones (Rodriguez‐Lanetty, Phillips, & Weis, 2006 ; Stewart, Pavasovic, Hock, & Prentis, 2017 ) and corals (Libro et al., 2013 ; Libro & Vollmer, 2016 ; Oren et al., 2010 ). Von Willebrand factors (VWF) have also been observed to be up‐regulated in symbiotic Exaiptasia (Rodriguez‐Lanetty et al., 2006 ). Von Willebrand factor (VWF) is involved in cell adhesion and collagen binding (Ruggeri, 2007 ) and has been associated with allogeneic rejection, pathogen exposure, and symbiotic state in cnidarians. Oren et al. ( 2010 ) observed up‐regulation of VWF during allogeneic rejection in the coral Stylophora pistillata , Libro et al. ( 2013 ) observed up‐regulation of VWF in response to White Band disease infection in the coral Acropora cervicornis , and Rodriguez‐Lanetty et al. ( 2006 ) observed up‐regulation of VWF in symbiotic versus aposymbiotic forms of the anemone Anthopleura elegantissima . Our data demonstrate that VWF is up‐regulated due to aposymbiosis (menthol induced) and thus the difference between our data, and Rodriguez‐Lanetty et al. ( 2006 ) may reflect differences in VWF expression between expelling zooxanthellae versus a stable aposymbiotic state. Overall, our data coupled with published data indicate that up‐regulation of VWF may be a hallmark of anthozoan immunity, allorecognition, and breakdown of symbioses. Alpha‐2‐macroglobulin (A2MG) binds peptides including a wide range of proteinases (Borth, 1992 ). Its ability to bind the serine protease thrombin gives A2MG anticoagulant properties (Mitchell, Piovella, Ofosu, & Andrew, 1991 ), while its ability to inhibit proteins C and S gives it procoagulant properties (Cvirn et al., 2002 ). Many pathogen virulence factors act as proteases, and thus, A2MG's ability to inhibit proteases protects the host from these virulence factors (Armstrong & Quigley, 1999 ). A2MG has been associated with pathogen exposure in corals (Libro et al., 2013 ) and wound healing in anemones (Stewart et al., 2017 ), but had not been linked to aposymbioses. Tissue factor pathway inhibitor (TFPI) inhibits coagulation by inhibition of factor Xa and factor VIIa/tissue factor (Broze & Girard, 2012 ) and has been associated with coral immunity (Libro & Vollmer, 2016 ) and oxidative stress in anemones prior to bleaching (Richier, Rodriguez‐Lanetty, Schnitzler, & Weis, 2008 ). Plasminogen activator inhibitor 2 (PAI2) is a serine protease inhibitor associated with inhibition of fibrinolysis (Stump, Lijnen, & Collen, 1986 ) and negative regulation of apoptosis (Dickinson, Bates, Ferrante, & Antalis, 1995 ). Down‐regulation of PAI2 for both Vibrio and aposymbiosis suggests a net anticoagulant effect. Initiation of the complement cascade appeared to be stimulated only by pathogen exposure and not due to aposymbiosis. Within the complement cascade, six highly expressed DE genes were up‐regulated for Vibrio (C3, factor B, coagulation factor XII B polypeptide, membrane cofactor protein, decay‐accelerating factor, factor H). Complement C3 plays a central role in both the classical and alternative complement pathways. Up‐regulation of C3 has been observed in the coral Acropora millepora treated with bacterium Alteromonas (Brown, Bourne, & Rodriguez‐Lanetty, 2013 ) and in WBD‐infected Acropora cervicornis (Libro & Vollmer, 2016 ). Complement factor B (CFAB) is part of the alternative pathway. Poole et al. ( 2016 ) identified two variants of factor B in Exaiptasia , one of which was up‐regulated in response to both onset of symbiosis and treatment with Serratia marcescens . Up‐regulation of coagulation factor XII B chain (F13B), is associated with blood coagulation and hemostasis in vertebrates (Ivanov et al., 2017 ) and has also been observed in diseased Acropora cervicornis (Libro, 2014 ). In the complement pathway, the classical and lectin pathways require specific recognition molecules for initiation, but in the alternative pathway, C3b is deposited on all cells (host as well as pathogenic) exposed to activated complement (Ferreira, Pangburn, & Cortés, 2010 ). Three remaining highly expressed DE genes within complement (membrane cofactor protein, complement decay‐accelerating factor, factor H) that were up‐regulated by Vibrio exposure are all involved in protecting host tissues from attack by the alternative complement pathway (Elvington, Liszewski, & Atkinson, 2016 ; Ferreira et al., 2010 ). Up‐regulation of membrane cofactor protein (MCP/cd46), complement decay‐accelerating factor (DAF/cd55), and component factor H in Exaiptasia would limit C3b deposition on healthy Exaiptasia cells (Elvington et al., 2016 ; Ferreira et al., 2010 ). Neither DAF, MCP, or CFAH have previously been associated with anthozoan immunity. 4.3 NOD/TLR pathway \n Vibrio exposure resulted in the strong differential expression of six genes [three up and three down] in the NOD/Toll‐like receptor pathway. Myeloid differentiation primary response protein (MyD88), TNF receptor‐associated factor 3 (TRAF3), and Bcl‐2‐like 1 apoptosis regulator Bcl‐X (Bcl‐2) were up‐regulated, while TNF receptor‐associated factor 2 (TRAF2), receptor‐interacting serine/threonine protein kinase 2 (RIPK2), and calcium‐sensing receptor (CASR) were down‐regulated. MyD88, TRAF3, and RIPK2 are key regulators of the NOD and TLR pathways that lead to NF‐kappa‐β activation, cytokine secretion, and the inflammatory response (Deguine & Barton, 2014 ; Häcker, Tseng, & Karin, 2011 ; Nachbur et al., 2015 ), while Bcl‐2 inhibits caspases and suppresses apoptosis (Youle & Strasser, 2008 ). TRAF2 regulates activation of NF‐kappa‐β (Lin et al., 2011 ), and JNK (Brnjic, Olofsson, Havelka, & Linder, 2010 ) and CASR (Chakravarti, Chattopadhyay, & Brown, 2012 ) regulate calcium homeostasis. Out of the six DE genes in the NOD/Toll‐like receptor pathway, only MyD88 has previously been observed to be DE in cnidarians due to immune exposure. Libro et al. ( 2013 ) observed up‐regulation of MyD88 in WBD‐infected Acropora cervicornis . In humans (Wang, Dziarski, Kirschning, Muzio, & Gupta, 2001 ), mouse (Deguine & Barton, 2014 ), and fly (Horng & Medzhitov, 2001 ), stimulation of Toll‐like receptors (TLRs) causes MyD88 to associate with the intracellular domain of the TLR leading to downstream signaling of NF‐kappa‐β via IRAK and TRAF and production of pro‐inflammatory cytokines (Akira, Uematsu, & Takeuchi, 2006 ). TLR activation of MyD88 has been demonstrated in the anemone Nematostella vectensis in a reporter gene assay where Nematostella TIR domain of TLR activated human MyD88. Our results indicate that MyD88 interacts with TRAF3, but not IRAK, which is supported by a MyD88 knockdown study by Franzenburg et al. ( 2012 ) in the hydrozoan Hydra vulgaris , which resulted in the down‐regulation of TRAF3 but not IRAK. The expression patterns of the remaining 3 NOD/Toll‐like receptor pathway genes (Bcl‐2, RIPK2, and CASR) suggest that they are acting to prevent apoptosis in Exaiptasia exposed to Vibrio . The B‐cell lymphoma 2 (Bcl‐2) family of apoptosis‐regulating proteins includes both pro‐ and antiapoptotic members. Ainsworth et al. ( 2015 ) identified up‐regulation of the pro‐apoptotic Bcl‐2 family member Bak in Acropora hyacinthus tissues exhibiting white syndrome, and Pernice et al. ( 2011 ) proposed that up‐regulation of Bcl‐2 is a protective response to heat‐stress‐induced apoptotic activity in Acropora millepora . Down‐regulation of RIPK2 and CASR also suggests an antiapoptotic role in Exaiptasia exposed to Vibrio . RIPK2 activates NF‐kappa‐β and induces cell death (McCarthy, Ni, & Dixit, 1998 ). Up‐regulation of CASR leads to apoptosis in rat myocytes exposed to LPS (Wang et al., 2013 ). To our knowledge, we are the first to report differential expression of RIPK2 and CASR in anthozoans. Even though Exaiptasia shows strong evidence for TLR pathway activation, no Exaiptasia TLRs met our annotation criteria (best hit, e‐value < 1e −10 , coverage > 50%). Two Exaiptasia genes ( KXJ18603.1 , KXJ08560.1 ) annotated as a relaxin receptor 2 (RXFP2) and outer membrane protein OprM (OPRM) were up‐regulated for Vibrio and had had blast hits for a TLR with an e‐value < 1e −10 and coverage greater than 50%, but the TLR was not the best hit. While it is possible that these two genes represent TLRs, more data would be needed to confirm their putative functions. Toll‐like receptors (TLRs) are transmembrane proteins consisting of an extracellular leucine‐rich repeat region (LRR) involved in pathogen recognition, and an intracellular Toll–interleukin receptor (TIR) which initiates downstream activation of NF‐kappa‐β via MyD88 (Brennan et al., 2017 ). A single TLR has been identified in the model anemone Nematostella vectensis , and its activation and downstream signaling via NF‐kappa‐β have been demonstrated in response to Vibrio coralliilyticus (Brennan et al., 2017 ). We performed a Pfam domain search on the Exaiptasia predicted proteins using hmmscan and the Pfam‐A hidden Markov model Pfam‐A.hmm (Eddy, 2011 ) and found a number of LRR‐containing and TIR‐containing proteins up‐regulated for Vibrio , but none of the predicted proteins contained both domains as expected of TLRs; this is consistent with the findings of Baumgarten et al. ( 2015 ) who did not find any proteins containing both domains in the Exaiptasia genome. In contrast to TLRs, NOD‐like receptors (NLRs) are present in the Exaiptasia genome (Baumgarten et al., 2015 ), but as with TLRs, we observed up‐regulation of genes in the NOD pathway, but not up‐regulation of NLRs. NOD‐like receptors are intracellular pattern‐recognition proteins that when activated lead to activation of NF‐κB and MAPK, and production of inflammatory caspases (Franchi, Warner, Viani, & Nuñez, 2009 ). 4.4 Chemokine and antigen processing Chemokine and antigen processing pathways were also activated by Vibrio exposure. Chemokine pathway had three highly DE genes—signal transducer and activator of transcription 1‐alpha/beta (STAT1) was up‐regulated while C‐X‐C chemokine receptor type 4 (CXCR4), and guanine nucleotide‐binding protein subunit beta‐4 (GBB4) were down‐regulated. STAT1 mediates cellular responses to interferons (IFNs), cytokines, and other growth factors (Ramana, Chatterjee‐Kishore, Nguyen, & Stark, 2000 ), and up‐regulation of STAT in response to bacterial exposure has been reported in a number of invertebrates including Anopheles gambiae (mosquito; Barillas‐Mury, Han, Seeley, & Kafatos, 1999 ), Drosophila (Buchon, Broderick, Poidevin, Pradervand, & Lemaitre, 2009 ), and Fenneropenaeus chinensis (Chinese white shrimp; Sun, Shao, Zhang, Zhao, & Wang, 2011 ). Sinkovics ( 2015 ) proposed a Cnidarian origin of STAT based on genomic studies on Nematostella vectensis , but its role in immunity had not been confirmed by expression analysis. Two antigen processing genes were highly DE; proteasome activator subunit 2 (PSME2/PA28 beta) was up‐regulated for Vibrio , and regulator factor X‐associated ankyrin‐containing protein (RFXK) was down‐regulated for aposymbiosis. Proteasomes are involved in antigen processing (Michalek, Grant, Gramm, Goldberg, & Rock, 1993 ) and degradation of other intracellular proteins (Tanaka, 2009 ), including cytotoxic damaged proteins resulting from the oxidative stress of an immune response (Kammerl & Meiners, 2016 ). Traylor‐Knowles, Rose, Sheets, and Palumbi ( 2017 ) observed up‐regulation of proteasome components in Acropora hyacinthus exposed to heat stress. To our knowledge, we are the first to report up‐regulation of proteasomal proteins in response to bacterial immune challenge in cnidarians. 4.5 Transport and catabolism Within transport and catabolism (peroxisome, endocytosis, lysosome), there were more genes highly down‐regulated than up‐regulated with five up‐regulated (HRS, HSE1, CLH, PAOX, and AP3D) and eight down‐regulated (CXCR4, VPS4, PLD2, SOX, LCFB, DHRS4, GNPAT, and GALNS) for Vibrio and three up‐regulated (PAOX, EASC, and PAG15) and four down‐regulated (JUNO, GNPAT, BAAT, and BGLR) for aposymbiosis. Once a pathogen has been recognized, endosomes, lysosomes, and peroxisomes are involved in their engulfment, destruction, and clearance (Di Cara, Sheshachalam, Braverman, Rachubinski, & Simmonds, 2017 ). 4.6 Endocytosis Endocytosis pathway had six highly expressed DE genes; three genes were up‐regulated for Vibrio (HRS, HSE1, CLH), two genes were down‐regulated for Vibrio (VPS4, PLD2), and one gene was down‐regulated for aposymbiosis (JUNO). Following recognition by Toll‐like receptors, pathogens are engulfed by clathrin‐mediated endocytosis (Husebye et al., 2006 ). In Drosophila , endocytosis is required for activation of the Toll pathway, and endosomal proteins Mop and Hrs colocalize with the Toll receptor in endosomes (Huang, Chen, Kunes, Chang, & Maniatis, 2010 ). Although we observed more down‐regulation than up‐regulation of genes within the endocytosis pathway, those which were up‐regulated in response to Vibrio (hepatocyte growth factor‐regulated tyrosine kinase substrate HRS, clathrin heavy‐chain CLH, signal transducing adaptor molecule HSE1) are consistent with recognition by TLR pathway followed by clathrin‐mediated endocytosis. 4.7 Apoptosis—programmed cell death Apoptosis pathway had five highly expressed DE genes; two caspases were up‐regulated for Vibrio (CASP7, CASP9), TBA was down‐regulated for Vibrio , P53 was up‐regulated for aposymbiosis, and BIRC5 was down‐regulated for aposymbiosis. Apoptosis has been proposed as a means of removing Symbiodinium during thermal bleaching (Dunn, Schnitzler, & Weis, 2007 ; Kvitt, Rosenfeld, & Tchernov, 2016 ; Pernice et al., 2011 ; Rodriguez‐Lanetty et al., 2006 ) as well as clearing pathogens in the immune response (Fuess et al., 2017 ; Libro et al., 2013 ). Caspases are key initiators of apoptosis (McIlwain, Berger, & Mak, 2013 ). Up‐regulation of CASP7 and CASP9 have not been previously reported in anthozoans, but up‐regulation of caspase‐3 was documented in WBD‐infected Acropora cervicornis (Libro & Vollmer, 2016 ). Tubulin alpha (TBA) was also down‐regulated due to Vibrio exposure in Exaiptasia . Down‐regulation of tubulin beta has been observed for WBD‐infected Acropora cervicornis as well (Libro & Vollmer, 2016 ). Two DE apoptosis genes for aposymbiosis (P53, BIRC5) suggest apoptosis is a mechanism for menthol bleaching. Tumor protein P53 regulates a number of cell‐cycle functions including apoptosis, regulation of autophagy, cell‐cycle arrest, and senescence (Zilfou & Lowe, 2009 ). Lesser and Farrell ( 2004 ) observed up‐regulation of P53 in corals exposed to increased solar radiation, and Weis ( 2008 ) proposed activation of P53 by the reactive nitrogen species nitric oxide (NO) in thermally stressed corals as a mechanism of bleaching. The up‐regulation of P53 in aposymbiotic anemones may indicate that the mechanism of menthol‐induced bleaching is similar to the mechanisms of bleaching in thermal and solar radiation‐stressed corals. The second DE apoptosis gene for menthol BIRC5, also known as survivin, is an antiapoptotic caspase inhibitor (Li et al., 1998 ). The down‐regulation of BIRC5 for aposymbiotic anemones lends further support to apoptosis as a mechanism of menthol bleaching."
} | 7,713 |
34296001 | PMC8282963 | pmc | 4,346 | {
"abstract": "In this study, experimental results of hydrogen producing process based on anaerobic photosynthesis using the purple non-sulfur bacterium Rhodobacter capsulatus are scrutinized. The bacterial culture was carried out in a photo-bioreactor operated in a quasi-continuous mode, using lactate as a carbon source. The method is based on the continuous stirred tank reactors (CSTR) technique to access kinetic parameters. The dynamic evolution of hydrogen production as a function of time was accurately simulated using Luedeking-Piret model and the growth of R. capsulatus was computed using Gompertz model. The combination of both models was successfully applied to determine the relevant parameters (λ, μ max , α and β) for two R. capsulatus strains studied: the wild-type strain B10 and the H 2 over-producing mutant IR3. The mathematical description indicates that the photofermentation is more promising than dark fermentation for the conversion of organic substrates into biogas.",
"conclusion": "4 Conclusions The Luedeking-Piret model was used in this study to describe the bacterial growth, substrate consumption, and hydrogen gas production by using the purple non-sulfur bacterium Rhodobacter capsulatus . Hydrogen production by R. capsulatus during a quasi-continuous culture over a 150–200 h period was successfully controlled with various key parameters as inlet flow rate, feed volume, and dilution rate of substrate supply. The R. capsulatus over-producing strain IR3 displayed a greater hydrogen production yield correlated to the non-growth associated term (β). A good agreement was obtained between the experiments and simulations with high regression coefficient values (R 2 > 0.99). Fitting procedure allows to relevant parameters such as the lag time (λ), the maximum growth rate μmax and both non-growth (β) and growth-associated terms (α) of Luedeking-Piret model. This parameter access will be useful for comparison between strains or/and literature data and for the automation and control of bioprocesses for the continuous fermentation process.",
"introduction": "1 Introduction Economic development over the last few decades has been strongly dependent on fossil fuels as sources of energy. These resources are not unlimited in the long term, and environmental concerns have led to the search for clean, renewable energy sources. Urban/agro-industrial/agricultural wastes appear as relevant energetic resources in a sustainable energy mix. Biogas results from the anaerobic digestion of organic matter that, is the main constituent of these wastes. Indeed, the production of methane or hydrogen from the biomass follows the principle of the fermentation. Hydrogen is considered as the more sustainable energy carrier due to the high efficient end-use technologies such as the fuel cells [ 1 , 2 ]. Clean-hydrogen gas can be produced either by electrolysis [ 3 ], steam biomethane reforming [ 4 ], biomass gasification [ 5 ], or biological synthesis [ 6 , 7 ]. Biohydrogen production processes, are divided into two groups according to the dependency on light: dark and photo fermentations [ 8 , 9 ]. Under dark anaerobic conditions, the organic substrates and waste waters are metabolized to form hydrogen and lower molecular weight organic acids. Photofermentative hydrogen production, issued from oxidation of organic compounds, occurs under anaerobic, nitrogen-limited conditions, utilizing light as energy source. A wide range of photosynthetic bacteria has been reported to produce hydrogen, including Rhodobacter capsulatus , Rhodobacter sphaeroïdes , Rhodopseudomonas palustris, and Rhodospirillum rubrum [ 10 , 11 ]. Among them, Rhodobacter capsulatus is a favorable candidate for large-scale production due to its high energy and substrate conversion efficiencies and its ability to utilize a wide variety of substrates for growth and hydrogen production [ 12 , 13 ]. The rate and yield of hydrogen production is greatly dependent on the carbon source used, physiological growth conditions, such as light intensity [ 14 , 15 ], and bacterial growth mode. Fed-batch growth mode is recognized as the most suitable operation mode for H 2 production, in comparison with batch and continuous modes [ 16 ]. In fact, studies for H 2 production during fed-batch conditions by photosynthetic bacteria have been reported for Rhodopseudomonas palustris sp. and R. palustris 42OL (on acetate and malate as carbon source, respectively) [ 17 , 18 ], R. sphaeroïdes ZX-5 (on malate) [ 19 ], R. faecalis RLD-53 (on acetate) [ 20 ], and R. capsulatus DSM1710 (on acetate) [ 13 ]. Various kinetics models have been derived for biohydrogen production, as previously reviewed [ 21 , 22 ]. In particular, and because of its simple initial form, the Luedeking-Piret model, developed in 1959 to describe lactic acid production and others processes [ 23 , 24 ], has recently been applied to fermentative hydrogen production. In this widely used mathematical model, the rate of product formation (like hydrogen) can be related to both biomass concentration and microbial growth rate. All the studies concerned dark fermentation conditions with the exception of one devoted to the phot osynthetic bacterium, R. palustris [ 25 ]. Unfortunately, comparison between studies is quite complicated because the use of batch culture and different operating conditions makes it difficult to determine Luedeking-Piret model parameters from the published data. The estimation of not directly quantifiable compounds is essential in the development of biotechnology processes like photo-hydrogen production. However, a complex model involves several kinetics parameters and it is difficult to discuss the relevance of each one. It is also difficult to use traditional batch techniques, without complex model, to scrutinize useful parameters. This problem is theoretically overcome with continuous stirred tank reactors (CSTR) inside which the composition is uniform at any point. For a given mean residence time ( V Q ) and inlet and outlet concentrations, the mass balance is very simple and it is hence easy to calculate the kinetic constants. In practice, the stirring of bioreactor is well achieved with a good mixing but the inlet flow rate must be adapted to avoid wash-out of microbial biomass. The present study describes a mathematical approach to determine reliable kinetic parameters of hydrogen production by the photosynthetic bacterium, R. capsulatus , in a quasi-continuous photobioreactor. The developed mathematical/experimental approach is simple enough in terms of mathematical complexity and experimental procedure to be further used to attain reliable parameter values (α, β) of Luedeking-Piret model or maximum growth rate (μ max ) and lag time (λ). The present work combines Luedeking-Piret model and Gompertz model using the main assumptions of continuous stirred-tank reactor (CSTR) operation.",
"discussion": "3 Results and discussion 3.1 Experimental measurements of bacterial growth and substrate consumption Kinetics of bacterial production and substrate consumption for the wild type R. capsulatus B10 and the H 2 over-producing strain IR3 were shown in Figures 1 A-1B, respectively. Figure 1 Kinetics of the lactate consumption (S, ◊) and the bacterial production (X, ▲) during a quasi-continuous culture of R. capsulatus B10 (A) and IR3 (B). the arrows indicate the application of continuous flow. Figure 1 For the wild type B10 strain, the first feeding procedure started at the beginning of the batch stationary phase ( Figure 1 A), when lactate was almost completely metabolized (concentration close to 0 g L −1) . The application of continuous flow conditions led to an increase in lactate concentration to 0.7 g L −1 , concomitant with a decrease in bacterial protein concentration from 0.54 to 0.44 g L −1 . Interruption of continuous flow followed by batch culture then led to a stabilization at 0.7–0.8 g lactate L −1 and 0.44–0.48 g protein L −1 . This was repeated over 3 cycles of quasi-continuous operation. For the H 2 over-producing strain IR3, the feeding procedure started at the end of the exponential growth phase, after the consumption of 90% of lactate (residual concentration of 0.4 g L −1 ). The feeding of growth medium led to an increase in lactate concentration to 1 g L −1 and a decrease in bacterial protein concentration from 0.4 g L −1 to 0.35 g L −1 . Alternation of continuous and batch modes led to a stabilization at 0.34–0.38 g protein L −1 and 0.75–0.95 g lactate L −1 , and this was repeated over 5 cycles of quasi-continuous operation. The final, prolonged batch phase led to complete consumption of lactate and a final bacterial concentration of 0.45 g L −1 . In both case the decrease in bacterial concentration due to wash-out during the continuous mode operation was computed to be close to 20 mg of proteins per hour for an average flow rate < 1 ml min −1 . This is small compared to the potential range of biomass production in the reactor of 100–350 mg protein per hour. Taken together with the lack of substrate limitation (stabilization at 0.7–0.95 g lactate L −1 ), these results validate the assumptions of the Gompertz model. 3.2 Fitting procedure and parametric values The aim of the “data fitting” is to determine and compare parameter values of the model that describes the process. On the other hand, the parametric optimization is also a useful tool for validating the model. Nevertheless, the fitting procedure requires an appropriate attention to achieve relevant results. The Gradient methods are generally more efficient when the objective function is continuous in its first derivative. Gradient methods use information about the slope of the function to dictate a direction of search where the minimum is thought to lie. The simplest of these is the method of steepest descent in which a search is performed in an opposite direction of gradient of the objective function. This method is available with numerous commercial software. Gompertz function and ( Eq. 1 ) coupled Luedeking-Piret model and Gompertz function ( Eq. 12 ) are expedient mathematical functions to achieve the conventional numerical optimizations. Bacterial growth was modeled using the Gompertz function; and Figure 2 shows good agreement between the model and experimental values of biomass concentration. Table 1 compares the a, b and c values for the R. capsulatus strains. However, no direct interpretation is possible as these are raw parameters, and the Eqs. (5) , (6) , (7) , and (8) must be used in order to obtain meaningful physical parameters, such as μ max , λ, and Y xS . Figure 2 Bacterial growths in quasi-continuous culture of wild strain B10 (●) and over-producer strain IR3 (○): experimental data and Gompertz simulation. Figure 2 Table 1 Gompertz function parameters. Table 1 B10 IR3 a [-] 3.03 2.66 b [-] 1.77 2.18 c [h −1 ] 0.174 0.1581 R 2 0,9924 0.9968 In this context, λ is the intercept with x-axis of the tangent line as expressed in the Eq. (7) , μ max is given by the derivative function of Gompertz model ( Eq. (7) ) and Y xS is directly obtained from Eq. (9) and the initial substrate concentration. The growth-associated product coefficient, α, and the non-growth-associated product coefficient, β, were the determined by numerical optimization from experimental data and the Luedeking-Piret model Eq. (12) (i.e. data fitting). Experimental data for H 2 production were correctly modelled for both strains B10 and IR3, as shown in Figure 3 , and the fitted parameters are summarized in Table 2 . Figure 3 Simulated Luedeking–Piret's model coupled Gompertz equation to and H 2 -production by Rhodobacter capsulatus B10 (A) and IR3 (B) (35 mM lactate, 30,000 Lx). Figure 3 Table 2 Fitted parameters of Luedeking–Piret's model and associated growth parameters. Table 2 B10 IR3 Y XS 0.124 0.109 λ [h] 9.8 13.9 μ max [h −1 ] 0.195 0.166 β [ml H 2 (g L −1 ) −1 h −1 ] 17 72 α (ml (g L −1 ) −1 ) 4.5 36.6 R 2 0.9980 0.9975 β [ml H 2 (g L −1 ) −1 h −1 ] (linear regression) 14 75.6 R 2 0.9942 0. 9969 The corresponding physical parameters, μ max , Y xS , λ, α, β were summarized in Table 2 . μ max and λ parameters differed according to the basic growth observation of both bacterial strains. The value of the maximum specific growth rate, μ max , was higher for the wild type strain B10 (0.195 h −1 ) than the one observed for the over-producer strain IR3 (0.166 h −1 ). In addition, the wild strain B10 grew earlier than the mutant strain IR3 with a similar substrate utilization rate, Y XS : 0.124 and 0.109 g protein g −1 lactate, respectively B10 and IR3, this was correlated with a lower lag time observed for the wild type B10 (9.8 h) compared to the mutant IR3 (13.9 h). The Luedeking-Piret model implies that hydrogen production is associated with both non-growth and growth-associated terms. The growth-associated term indicates that hydrogen production (α) is proportional to the bacterial growth rate. On the other hand, the non-growth associated term (β) signifies that hydrogen production is linearly dependent on biomass concentration. In the case of strain IR3, both parameters were substantially higher than for the wild type strain B10 (see Table 2 ). It is noteworthy that the β parameter is main parameter to describe hydrogen production by photofermentation. Eq. (13) shows that it possible to model fermentation products using a linear relationship, assuming a negligible artificial mortality and biomass dilution during the continuous or quasi-continuous process: (14) P ( t ) = b + β X t ¯ Therefore, the non-growth associated term could be attained by direct linear regression and the parameters determined by the linear procedure ( equation 13 ) were very close to those obtained by the overall simulation of the Luedeking–Piret model. Both mathematical procedures can provide realistic β parameters, which is the most important parameter when selecting a productive strain. One cannot estimate α separately using batch culture data ( Eq. (12) ), except for t << tr, and concomitant α and β fitting is complicated by the necessary approximations of growth. Thus, in the case of batch cultures, realistic estimations of β and α require much additional data. 3.3 Discussion Using the Gompertz function to model bacterial growth and the Luedeking-Piret model to describe hydrogen production by R. capsulatus during quasi-continuous culture, good agreement was observed between the experimental data and the models, with high regression coefficient values (R 2 ) exceeding 0.99 in all cases ( Table 2 ). The β and α parameters were lower in the wild type strain B10 than in the H 2 over-producer mutant IR3, as previously observed in batch culture [ 28 ]. The authors showed that the specific hydrogen production rate (ml h −1 ) is proportional to the light intensity, which is the multiplying factor applied to non-growth -associated term (β) of Luedeking-Piret model. As above, in the present study, the wild type B10 exhibited a maximum specific growth rate, μ max , higher (0.195 h −1 ) that the over-producer strain IR3 (0.166 h −1 ), this latter converted more efficiently the substrate (lactate) in product (H 2 ) that the wild-type B10 strain. The maximum growth rate observed in indoor culture on synthetic medium with the wild-type R. capsulatus B10, on lactate-glutamate (35/5 mM as carbon and nitrogen source, is comparable with this obtained with the strain 37b4 on malate-(NH 4 ) 2 SO 4 (16/9.5 mM), 0.195 and 0.251 h −1 [ 32 ]. However, much lower maximum growth rates were observed in outdoor fed batch cultures on acetate-glutamate medium of R. capsulatus DSM1710 and the hydrogenase-deficient mutant YO3 (0.025 h −1 and 0.052 h −1 , respectively [ 13 , 33 ]. By contrast, in the present study both strains (IR3 and B10) exhibit similar values of the conversion rate of product/biomass. Hydrogen yields can be expressed according to different form like g or L hydrogen per g −1 biomass, or by lactate conversion rate. The latter was based on the theoretical total conversion (100%) of 1 mol of lactate leading to the formation of 6 mol of hydrogen. For both strains, the first step of culture corresponding to batch conditions led to a weak lactate conversion, 4.5 and 18.6 % of lactate conversion for B10 and IR3, respectively. During the quasi-continuous process, B10 strain exhibited a weak lactate conversion rate compared to the over-producer strain IR3, 26.7 % and 95 %, respectively. These values were equivalent to 0.36 and 1.1 L H 2 g −1 lactate, representing 27 mg and 110 mg H 2 g −1 lactate, for B10 and IR3 strains, respectively. It is essential to point out the assessments of Luedeking-Piret parameters in the literature in order to understand the values obtained in this study. Table 3 summarizes the H 2 production processes, based either on dark or light fermentation, and modeled by the Luedeking-Piret model. The modeling of dark fermentation processes only implied the hydrogen production associated to the biomass growth; β value was equal to zero. At the opposite, the H 2 -producing process based on photosynthetic biomass involved both a non-growth and a growth-associated hydrogen production. Our results are the first study devoted to the robust estimation of λ, μ max , α and β parameter. It shows that β parameter is main parameter to describe kinetic of gas production by photofermentation, therefore dark anaerobic conditions present lower potential of volatile fatty acid conversion into hydrogen than light anaerobic conditions. In addition, the production rates of hydrogen by R. capsulatus with respect to light intensity irrespective of cell concentration have been previously described by Obeid et al [ 28 ] and Androga et al [ 34 ]. Authors showed that the specific hydrogen production rate (ml h −1 ) is proportional to the light intensity, which is the multiplying factor applied to non-growth -associated term (β) of Luedeking-Piret model. Table 3 H 2 -producing processes described by the original or modified forms of the Luedeking-Piret (LP) model. Table 3 Biomass α value (ml g −1 VSS a ) β value (ml H 2 g −1 dw h −1 ) Reactor type substrate Fermentation/LP model Ref. R. palustris 6.85 b 0.41 b batch glycerol Light/modified [ 25 ] R. capsulatus B10 IR3 4.5 c 36.6 c 17 c 72 c Quasi-continuous Lactate Light/modified This study Mangrove sediments 11.04 0 batch glucose Dark/original [ 35 ] Local sewage sludge 224 0 continuous sucrose Dark/ e r H2 = DαX [ 36 ] Anaerobic sludge 759 0 batch glucose Dark/original [ 37 ] Anaerobic sludge 793 0 batch glucose Dark/original [ 38 ] Enterobacter cloacae 166 d 0 batch glucose Dark/original [ 39 ] Clostridium pasteuranium 918 876 0 0 batch xylose sucrose Dark/dP/dt = α(1/X.dX/dt) [ 40 ] a Volatile Suspended Solids. b undefined. c α = 1/Y XP expressed in ml H 2 g protein L −1 and β in ml H 2 (g L −1 ) −1 h −1 . d ml H 2 g −1 cell mass. e D = liquid phase dilution rate, X = biomass concentration. Since the first study [ 44 ] devoted to the outdoor H 2 production during a fed-batch culture of Rhodopseudomonas sphaeroïdes, many studies with PNS (purple-non-sulfur) bacteria have been carried out in different culture modes including fed-batch, repeated conditions or semi and continuous condition. They were summarized by Argun and Kargi [ 20 ], Androga et al. [ 41 ], Sagnak and Kargi [ 42 ], Basak et al. [ 10 ], and Uyar et al. [ 43 ]. Table 4 incorporates these and more recent data based on fed-batch and semi-continuous H 2 production processes using either pure PNS cultures or co-cultures with dark fermentative bacteria. Indoor conditions represented the majority of previous fed-batch studies, using different light source and intensity as tungsten, halogen, fluorescent, incandescent, and Na vapor light. Nature and use of different unities of illumination increased the difficulty to compare experimental conditions. Recently many outdoor operating system for H 2 production by R. capsulatus were carried out according to reactor design, effluent nature, bacterial strains, and feeding strategy conditions ( Table 5 ), principally with Rhodobacter capsulatus . Table 4 H 2 -producing processes based on fed-batch regime. These processes have been previously summarized by Argun and Kargi 20], Androga et al. [ 41 ], Sagnak and Kargi [ 42 ], Basak et al. [ 16 ], and Uyar et al. [ 43 ]. Table 4 Strains Carbon and nitrogen sources Photobioreactor characteristics and operating conditions Fed batch, semi continuous conditions Maximal H 2 productivity a (mM H 2 L −1 medium h −1 ) Substrate conversion efficiency (mol H 2 mol −1 carbon source) Ref. R. sphaeroïdes B6 (thermostable) Lactate (25 mM) Na Glutamate (5 mM) 6-L plate polyacrylate outdoor 15 % dilution/4 days 1.07–1.56 (fine weather) 0.61–1 (cloudy weather) 0.09–0.56 (rainy weather) 3.51 a [ 44 ] Rhodopseudomonas capsulata Glucose DFE b : acetate (8.5 mM) propionate (1.7 mM) butyrate (13.6 mM) Na Glutamate (3 mM) 1.5-L indoor HRT c : 72 h 0.65 1.6 (acetate) 2.8 (propionate) 4 (butyrate) [ 45 ] acetate (30.5 mM) propionate (2 mM) butyrate (9 mM) Na Glutamate (3 mM) 1.5-L indoor HRT c : 72 h 0.8 ND acetate (11.5 mM) propionate (3 mM) Na Glutamate (3 mM) 1.5-L indoor HRT c : 72 h 0.94 ND Rhodopseudomonas palustris 42OL Malic acid (24.3 mM) Glutamic acid (5.6 mM) 1.07 L cylindric glass pH 6.8 320 W m 2 408-h duration ND small concentrated stock solution volume replacing sampling volume 0.49 ND [ 46 ] Rhodopseudomonas palustris sp . Acetate (66.7 mM) Glutamic acid (5.6 mM) Cylindric glass (0.22 L) pH 6.8–7.2 4 W m 2 indoor 10% withdraw (experimental condition SC d 240h) 0.39 1.6 [ 17 ] Clostridium acetobutylicum DSM792 + R. sphaeroïdes O.U.O O 1 Non pretreated wheat starch from corn (C source)/yeast extract (0.5) e and peptone (1) e cylindrical reactor (0.25 L/0.12 L) f controlled pH 7 192 W/m 2 Light/Dark ratio g = 2 OLR h 1.5 g starch L −1 d −1 2.5 % volume of medium replaced every day 0.375 g OLR L −1 d −1 1.03 2.62 (average on 33 days) [ 47 ] Clostridium butirycum N1VLB-B-3060 + R. sphaeroïdes VKM-3050 Starch (4.5) c /yeast extract (0.04) c , Na Glutamate (0.9) c Hungate tube (16 ml/8 ml) d Microaerobic conditions pH7.5 30 W/m 2 Light/Dark ratio g = 2.28 95–96 % volume of medium replaced ND 5.2 [ 48 ] a calculated from data. b Dark Fermentation Effluent. c Hydraulic Retention Time. d Semi continuous, culture volume withdraw. e expressed in g L −1 . f working volume/total volume of reactor. g Light fermentative biomass/Dark fermentative biomass concentration ratio. h Organic Loading Rates. Table 5 Comparison of the H 2 productivity and substrate conversion ratio obtained by Rhodobacter capsulatus strains under fed-batch, continuous and semi-continuous conditions. Table 5 Carbon sources Photo-bioreactor Growth mode Location Studied biomasses Maximal H 2 productivity (mL H 2 L −1 medium h −1 ) Substrate conversion efficiency (mol H 2 mol −1 carbon source) Ref. Lactate (35 mM) Na Glutamate (5 mM) 1-L PMMA panel semi-continuous Indoor B10, wild-type 0.26 0.24 This study IR3, H 2 over producer 0.41 0.75 This study Acetate (40 mM) Na Glutamate (2 mM) 4L–8L PMMA panel fed-batch (10 L daily feeding) outdoor YO3, hup − mutant 0.51 0.53 [ 51 ] 80-L PMMA tubular fed-batch (10 L daily feeding) outdoor DSM1710, wildt type 0.31 0.60 [ 32 ] YO3, hup − mutant 0.40 0.35 [ 13 ] Lactate (4 mM) Acetate (23 mM) Na Glutamate (1.37 mM) 65-L PMMA panel continuous (daily replacement of 20 % volume reactor) outdoor DSM155, wild-type 0.36 ND [ 52 ] 4 ∗ 25-L PMMA tubular continuous (daily replacement of 20 % volume reactor) outdoor DSM155, wild-type 0.15 ND [ 16 ] Sugar beet thick juice DFE a (Acetate 42 mM, NH 4 + 2.2 mM, Total nitrogen 3.3 mM) b 4-L PMMA panel fed-batch (10% daily feeding v:v) outdoor YO3, hup − mutant 1.12 0.77 [ 49 ] Sugar beet thick juice DFE a (Acetate 31 mM, NH 4 + 2 mM, Total nitrogen 7 mM) c 4-L PMMA panel continuous (daily replacement of 10% v:v) indoor YO3, hup − mutant 1.01 0.48 [ 43 ] 4-L PMMA panel continuous (daily replacement of 10% v:v) indoor DSM1710, wild type 1.05 0.46 Molasse DFE a (Acetate 32.5 mM, lactate 2.5 mM, Formiate 2.5 mM) b 4-L PMMA panel fed-batch (10% daily feeding v:v) outdoor DSM1710, wildt type 0.5 0.5 [ 50 ] YO3, hup − mutant 0.67 0.78 a Dark fermentation effluent. b Supplemented with Fe-citrate and Na2MoO4,2H20, 0.1 mM and 0.16 μm, respectively. c Supplemented with Fe-citrate 0.1 mM. Efficiency of H 2 producing bioprocess based on PNS was habitually characterized by two principal parameters: the H2 productivity expressed as mM H2 produced L-1. h-1 and the substrate conversion ratio, which is the experimental H 2 yield divided by theoretical yield. H 2 productivity and substrate conversion ratio depend on experimental conditions as substrate nature, illumination, bacterial strain and feeding strategy conditions as indicated in previous summary tables extracted from recent reviews. In order to make the comparison more concise only results with Rhodobacter capsulatus strains were reported here. The production of H 2 by PNS totally depends on the enzymatic activity of nitrogenase. However, H 2 can be consumed by another enzymatic system known as uptake hydrogenase (hup). The productivity values for the wild type strain B10 and the over-producer strain IR3, extracted from Figure 3 ; were 0.26 and 0.41 mM H2 L-1 culture h-1, respectively. Substrate conversion ratio was calculated from oxidized lactate during semi-continuous of wild-type B10 and H 2 over-producer strain IR3. H 2 productivity value of wild-type B10 was 0.24 M M −1 lactate. This value was close to that of DSM155 (0.15–0.36 M M −1 carbon source), but lower that those obtained for the wild-type DSM1710 (0.5–0.78 M M −1 carbon source); irrespective to experimental conditions ( Table 5 ). The H 2 over-producer mutant IR3 exhibited a substrate conversion ratio of 0.75. This latter was larger than the wild-type B10, but close to the maximum H 2 substrate conversion ratio obtained during fed-batch culture of the hup-R. capsulatus mutant YO3 on dark fermented effluents of sugar beet thick huice [ 49 ] or molasse [ 50 ] ( Table 5 ). Therefore, our methodology provides a robust estimation of λ, μ max , α and β parameter in agreement with literature results (i.e. substrate conversion ratio and H 2 productivity) which achieved under fed-batch, continuous and semi-continuous conditions.\n\n3.3 Discussion Using the Gompertz function to model bacterial growth and the Luedeking-Piret model to describe hydrogen production by R. capsulatus during quasi-continuous culture, good agreement was observed between the experimental data and the models, with high regression coefficient values (R 2 ) exceeding 0.99 in all cases ( Table 2 ). The β and α parameters were lower in the wild type strain B10 than in the H 2 over-producer mutant IR3, as previously observed in batch culture [ 28 ]. The authors showed that the specific hydrogen production rate (ml h −1 ) is proportional to the light intensity, which is the multiplying factor applied to non-growth -associated term (β) of Luedeking-Piret model. As above, in the present study, the wild type B10 exhibited a maximum specific growth rate, μ max , higher (0.195 h −1 ) that the over-producer strain IR3 (0.166 h −1 ), this latter converted more efficiently the substrate (lactate) in product (H 2 ) that the wild-type B10 strain. The maximum growth rate observed in indoor culture on synthetic medium with the wild-type R. capsulatus B10, on lactate-glutamate (35/5 mM as carbon and nitrogen source, is comparable with this obtained with the strain 37b4 on malate-(NH 4 ) 2 SO 4 (16/9.5 mM), 0.195 and 0.251 h −1 [ 32 ]. However, much lower maximum growth rates were observed in outdoor fed batch cultures on acetate-glutamate medium of R. capsulatus DSM1710 and the hydrogenase-deficient mutant YO3 (0.025 h −1 and 0.052 h −1 , respectively [ 13 , 33 ]. By contrast, in the present study both strains (IR3 and B10) exhibit similar values of the conversion rate of product/biomass. Hydrogen yields can be expressed according to different form like g or L hydrogen per g −1 biomass, or by lactate conversion rate. The latter was based on the theoretical total conversion (100%) of 1 mol of lactate leading to the formation of 6 mol of hydrogen. For both strains, the first step of culture corresponding to batch conditions led to a weak lactate conversion, 4.5 and 18.6 % of lactate conversion for B10 and IR3, respectively. During the quasi-continuous process, B10 strain exhibited a weak lactate conversion rate compared to the over-producer strain IR3, 26.7 % and 95 %, respectively. These values were equivalent to 0.36 and 1.1 L H 2 g −1 lactate, representing 27 mg and 110 mg H 2 g −1 lactate, for B10 and IR3 strains, respectively. It is essential to point out the assessments of Luedeking-Piret parameters in the literature in order to understand the values obtained in this study. Table 3 summarizes the H 2 production processes, based either on dark or light fermentation, and modeled by the Luedeking-Piret model. The modeling of dark fermentation processes only implied the hydrogen production associated to the biomass growth; β value was equal to zero. At the opposite, the H 2 -producing process based on photosynthetic biomass involved both a non-growth and a growth-associated hydrogen production. Our results are the first study devoted to the robust estimation of λ, μ max , α and β parameter. It shows that β parameter is main parameter to describe kinetic of gas production by photofermentation, therefore dark anaerobic conditions present lower potential of volatile fatty acid conversion into hydrogen than light anaerobic conditions. In addition, the production rates of hydrogen by R. capsulatus with respect to light intensity irrespective of cell concentration have been previously described by Obeid et al [ 28 ] and Androga et al [ 34 ]. Authors showed that the specific hydrogen production rate (ml h −1 ) is proportional to the light intensity, which is the multiplying factor applied to non-growth -associated term (β) of Luedeking-Piret model. Table 3 H 2 -producing processes described by the original or modified forms of the Luedeking-Piret (LP) model. Table 3 Biomass α value (ml g −1 VSS a ) β value (ml H 2 g −1 dw h −1 ) Reactor type substrate Fermentation/LP model Ref. R. palustris 6.85 b 0.41 b batch glycerol Light/modified [ 25 ] R. capsulatus B10 IR3 4.5 c 36.6 c 17 c 72 c Quasi-continuous Lactate Light/modified This study Mangrove sediments 11.04 0 batch glucose Dark/original [ 35 ] Local sewage sludge 224 0 continuous sucrose Dark/ e r H2 = DαX [ 36 ] Anaerobic sludge 759 0 batch glucose Dark/original [ 37 ] Anaerobic sludge 793 0 batch glucose Dark/original [ 38 ] Enterobacter cloacae 166 d 0 batch glucose Dark/original [ 39 ] Clostridium pasteuranium 918 876 0 0 batch xylose sucrose Dark/dP/dt = α(1/X.dX/dt) [ 40 ] a Volatile Suspended Solids. b undefined. c α = 1/Y XP expressed in ml H 2 g protein L −1 and β in ml H 2 (g L −1 ) −1 h −1 . d ml H 2 g −1 cell mass. e D = liquid phase dilution rate, X = biomass concentration. Since the first study [ 44 ] devoted to the outdoor H 2 production during a fed-batch culture of Rhodopseudomonas sphaeroïdes, many studies with PNS (purple-non-sulfur) bacteria have been carried out in different culture modes including fed-batch, repeated conditions or semi and continuous condition. They were summarized by Argun and Kargi [ 20 ], Androga et al. [ 41 ], Sagnak and Kargi [ 42 ], Basak et al. [ 10 ], and Uyar et al. [ 43 ]. Table 4 incorporates these and more recent data based on fed-batch and semi-continuous H 2 production processes using either pure PNS cultures or co-cultures with dark fermentative bacteria. Indoor conditions represented the majority of previous fed-batch studies, using different light source and intensity as tungsten, halogen, fluorescent, incandescent, and Na vapor light. Nature and use of different unities of illumination increased the difficulty to compare experimental conditions. Recently many outdoor operating system for H 2 production by R. capsulatus were carried out according to reactor design, effluent nature, bacterial strains, and feeding strategy conditions ( Table 5 ), principally with Rhodobacter capsulatus . Table 4 H 2 -producing processes based on fed-batch regime. These processes have been previously summarized by Argun and Kargi 20], Androga et al. [ 41 ], Sagnak and Kargi [ 42 ], Basak et al. [ 16 ], and Uyar et al. [ 43 ]. Table 4 Strains Carbon and nitrogen sources Photobioreactor characteristics and operating conditions Fed batch, semi continuous conditions Maximal H 2 productivity a (mM H 2 L −1 medium h −1 ) Substrate conversion efficiency (mol H 2 mol −1 carbon source) Ref. R. sphaeroïdes B6 (thermostable) Lactate (25 mM) Na Glutamate (5 mM) 6-L plate polyacrylate outdoor 15 % dilution/4 days 1.07–1.56 (fine weather) 0.61–1 (cloudy weather) 0.09–0.56 (rainy weather) 3.51 a [ 44 ] Rhodopseudomonas capsulata Glucose DFE b : acetate (8.5 mM) propionate (1.7 mM) butyrate (13.6 mM) Na Glutamate (3 mM) 1.5-L indoor HRT c : 72 h 0.65 1.6 (acetate) 2.8 (propionate) 4 (butyrate) [ 45 ] acetate (30.5 mM) propionate (2 mM) butyrate (9 mM) Na Glutamate (3 mM) 1.5-L indoor HRT c : 72 h 0.8 ND acetate (11.5 mM) propionate (3 mM) Na Glutamate (3 mM) 1.5-L indoor HRT c : 72 h 0.94 ND Rhodopseudomonas palustris 42OL Malic acid (24.3 mM) Glutamic acid (5.6 mM) 1.07 L cylindric glass pH 6.8 320 W m 2 408-h duration ND small concentrated stock solution volume replacing sampling volume 0.49 ND [ 46 ] Rhodopseudomonas palustris sp . Acetate (66.7 mM) Glutamic acid (5.6 mM) Cylindric glass (0.22 L) pH 6.8–7.2 4 W m 2 indoor 10% withdraw (experimental condition SC d 240h) 0.39 1.6 [ 17 ] Clostridium acetobutylicum DSM792 + R. sphaeroïdes O.U.O O 1 Non pretreated wheat starch from corn (C source)/yeast extract (0.5) e and peptone (1) e cylindrical reactor (0.25 L/0.12 L) f controlled pH 7 192 W/m 2 Light/Dark ratio g = 2 OLR h 1.5 g starch L −1 d −1 2.5 % volume of medium replaced every day 0.375 g OLR L −1 d −1 1.03 2.62 (average on 33 days) [ 47 ] Clostridium butirycum N1VLB-B-3060 + R. sphaeroïdes VKM-3050 Starch (4.5) c /yeast extract (0.04) c , Na Glutamate (0.9) c Hungate tube (16 ml/8 ml) d Microaerobic conditions pH7.5 30 W/m 2 Light/Dark ratio g = 2.28 95–96 % volume of medium replaced ND 5.2 [ 48 ] a calculated from data. b Dark Fermentation Effluent. c Hydraulic Retention Time. d Semi continuous, culture volume withdraw. e expressed in g L −1 . f working volume/total volume of reactor. g Light fermentative biomass/Dark fermentative biomass concentration ratio. h Organic Loading Rates. Table 5 Comparison of the H 2 productivity and substrate conversion ratio obtained by Rhodobacter capsulatus strains under fed-batch, continuous and semi-continuous conditions. Table 5 Carbon sources Photo-bioreactor Growth mode Location Studied biomasses Maximal H 2 productivity (mL H 2 L −1 medium h −1 ) Substrate conversion efficiency (mol H 2 mol −1 carbon source) Ref. Lactate (35 mM) Na Glutamate (5 mM) 1-L PMMA panel semi-continuous Indoor B10, wild-type 0.26 0.24 This study IR3, H 2 over producer 0.41 0.75 This study Acetate (40 mM) Na Glutamate (2 mM) 4L–8L PMMA panel fed-batch (10 L daily feeding) outdoor YO3, hup − mutant 0.51 0.53 [ 51 ] 80-L PMMA tubular fed-batch (10 L daily feeding) outdoor DSM1710, wildt type 0.31 0.60 [ 32 ] YO3, hup − mutant 0.40 0.35 [ 13 ] Lactate (4 mM) Acetate (23 mM) Na Glutamate (1.37 mM) 65-L PMMA panel continuous (daily replacement of 20 % volume reactor) outdoor DSM155, wild-type 0.36 ND [ 52 ] 4 ∗ 25-L PMMA tubular continuous (daily replacement of 20 % volume reactor) outdoor DSM155, wild-type 0.15 ND [ 16 ] Sugar beet thick juice DFE a (Acetate 42 mM, NH 4 + 2.2 mM, Total nitrogen 3.3 mM) b 4-L PMMA panel fed-batch (10% daily feeding v:v) outdoor YO3, hup − mutant 1.12 0.77 [ 49 ] Sugar beet thick juice DFE a (Acetate 31 mM, NH 4 + 2 mM, Total nitrogen 7 mM) c 4-L PMMA panel continuous (daily replacement of 10% v:v) indoor YO3, hup − mutant 1.01 0.48 [ 43 ] 4-L PMMA panel continuous (daily replacement of 10% v:v) indoor DSM1710, wild type 1.05 0.46 Molasse DFE a (Acetate 32.5 mM, lactate 2.5 mM, Formiate 2.5 mM) b 4-L PMMA panel fed-batch (10% daily feeding v:v) outdoor DSM1710, wildt type 0.5 0.5 [ 50 ] YO3, hup − mutant 0.67 0.78 a Dark fermentation effluent. b Supplemented with Fe-citrate and Na2MoO4,2H20, 0.1 mM and 0.16 μm, respectively. c Supplemented with Fe-citrate 0.1 mM. Efficiency of H 2 producing bioprocess based on PNS was habitually characterized by two principal parameters: the H2 productivity expressed as mM H2 produced L-1. h-1 and the substrate conversion ratio, which is the experimental H 2 yield divided by theoretical yield. H 2 productivity and substrate conversion ratio depend on experimental conditions as substrate nature, illumination, bacterial strain and feeding strategy conditions as indicated in previous summary tables extracted from recent reviews. In order to make the comparison more concise only results with Rhodobacter capsulatus strains were reported here. The production of H 2 by PNS totally depends on the enzymatic activity of nitrogenase. However, H 2 can be consumed by another enzymatic system known as uptake hydrogenase (hup). The productivity values for the wild type strain B10 and the over-producer strain IR3, extracted from Figure 3 ; were 0.26 and 0.41 mM H2 L-1 culture h-1, respectively. Substrate conversion ratio was calculated from oxidized lactate during semi-continuous of wild-type B10 and H 2 over-producer strain IR3. H 2 productivity value of wild-type B10 was 0.24 M M −1 lactate. This value was close to that of DSM155 (0.15–0.36 M M −1 carbon source), but lower that those obtained for the wild-type DSM1710 (0.5–0.78 M M −1 carbon source); irrespective to experimental conditions ( Table 5 ). The H 2 over-producer mutant IR3 exhibited a substrate conversion ratio of 0.75. This latter was larger than the wild-type B10, but close to the maximum H 2 substrate conversion ratio obtained during fed-batch culture of the hup-R. capsulatus mutant YO3 on dark fermented effluents of sugar beet thick huice [ 49 ] or molasse [ 50 ] ( Table 5 ). Therefore, our methodology provides a robust estimation of λ, μ max , α and β parameter in agreement with literature results (i.e. substrate conversion ratio and H 2 productivity) which achieved under fed-batch, continuous and semi-continuous conditions."
} | 9,721 |
27734577 | PMC5362678 | pmc | 4,347 | {
"abstract": "Summary Microalgal neutral lipids [mainly in the form of triacylglycerols ( TAG s)], feasible substrates for biofuel, are typically accumulated during the stationary growth phase. To make microalgal biofuels economically competitive with fossil fuels, generating strains that trigger TAG accumulation from the exponential growth phase is a promising biological approach. The regulatory mechanisms to trigger TAG accumulation from the exponential growth phase ( TAEP ) are important to be uncovered for advancing economic feasibility. Through the inhibition of pyruvate dehydrogenase kinase by sodium dichloroacetate, acetyl‐CoA level increased, resulting in TAEP in microalga Dunaliella tertiolecta . We further reported refilling of acetyl‐CoA pool through branched‐chain amino acid catabolism contributed to an overall sixfold TAEP with marginal compromise (4%) on growth in a TAG ‐rich D. tertiolecta mutant from targeted screening. Herein, a three‐step α loop‐integrated metabolic model is introduced to shed lights on the neutral lipid regulatory mechanism. This article provides novel approaches to compress lipid production phase and heightens lipid productivity and photosynthetic carbon capture via enhancing acetyl‐CoA level, which would optimize renewable microalgal biofuel to fulfil the demanding fuel market.",
"introduction": "Introduction To replace traditional fossil fuels and develop sustainable energy production, identifying sources of biologically derived fuels is increasingly urgent. Microalgae is recognized as a promising alternative source as they can accumulate neutral lipid, mainly in the form of triacylglycerol (TAG), which can be converted into biodiesels readily (Hossain et al ., 2008 ). In recent years, many attempts have been undertaken for the enhancement of TAG overproduction in microalgae (Radakovits et al ., 2010 ). These approaches mainly focus on biochemical and genetic engineering of lipid biosynthesis pathways and blocking of competing pathways (such as carbohydrate formation), so as to increase the pool of metabolites available for TAG biosynthesis (Courchesne et al ., 2009 ; Sharma et al ., 2012 ). However, almost all these approaches led to TAG accumulation in the stationary growth phase at the expense of biomass accumulation (Chiu et al ., 2009 ; Wang et al ., 2009 ) and overall lipid productivity. Vigorous growth and TAG accumulation appear to be mutually exclusive as TAG is a secondary (storage) metabolite and the pyruvate to acetyl‐CoA (AcCoA) pathway is tightly regulated by the growth‐dependent pyruvate dehydrogenase complex activity (Li et al ., 2014 ; Oliver et al ., 2009 ). The oleaginous diatom Fistulifera solaris JPCC DA0580 was the first to be reported to have a temporal overlap of TAG accumulation and cell growth during the exponential growth phase (Satoh et al ., 2013 ). Such a feature that triggers TAG accumulation while maintaining high growth rate is a critical advantage in the large‐scale cultivation of oleaginous microalgae for TAG production. To further exploit this potential in microalgae, fast‐growing, TAG‐rich, easily cultivated Dunaliella tertiolecta was used as the experimental organism (Rismani‐Yazdi et al ., 2011 ; Shin et al ., 2015 ; Yao et al ., 2015 ). In microalgae, AcCoA, malonyl‐CoA and NADPH are the major substrates in the plastid supporting fatty acid synthesis. Malonyl‐CoA is also generated from carboxylation of AcCoA. Thus, AcCoA is the primary precursor for fatty acid synthesis (Garrett and Grisham, 2013 ). The AcCoA balance in an algal cell could be described by the following equation: \n [ AcCoA T ] − [ AcCoA B ] = [ AcCoA NL ] \n \n Total AcCoA ([AcCoA T ]) and reduced NADH are produced via glycolysis. In the exponential growth phase, NADH is mainly oxidized through respiration to yield ATP, and AcCoA is used predominantly for biomass growth ([AcCoA B ]), including that for structural lipid (glycerophospholipids) synthesis, while a minor fraction of AcCoA ([AcCoA NL ]) and reduced NAD(P)H is used for fatty acid synthesis to accumulate neutral lipids (TAG). When microalgal cells enter the stationary growth phase, carbon metabolism for biomass growth diminished, which leads to accumulation of AcCoA and reduced NADH. Thus, in stationary phases or growth hindering stress conditions, a conspicuous fraction of AcCoA and reduced NAD(P)H is channelled to fatty acids biosynthesis, resulting in TAG accumulation in cells (Carpinelli et al ., 2014 ). Recent studies also suggested that intracellular membrane remodelling contributed to TAG accumulation during stationary phase or nitrogen starvation (Simionato et al ., 2013 ; Urzica et al ., 2013 ; Yoon et al ., 2012 ). To accelerate TAG accumulation in exponential growth phase (TAEP) while maintaining cell growth, AcCoA ([AcCoA T ]) level should be elevated over a certain set point that is needed for biomass growth ([AcCoA B ]). There are three principal sources of AcCoA during growth phase, namely fatty acid oxidation, glycolysis pathway and amino acid degradation (Garrett and Grisham, 2013 ). Fatty acid oxidation is the reversal of fatty acid synthesis and does not generate de novo AcCoA. Instead, it is thought that AcCoA is largely derived from the glycolytic pathway via pyruvate. Pyruvate is converted to AcCoA by PDHC in mitochondria and chloroplasts, and this step has been suggested as the key rate limiting step (Garrett and Grisham, 2013 ; Oliver et al ., 2009 ; Tovar‐Méndez et al ., 2003 ). One approach to increase AcCoA production is to relieve pyruvate dehydrogenase kinase (PDK) control of pyruvate dehydrogenase complex (PDHC) resulting in the activation of PDHC. This would facilitate the bioconversion of pyruvate to AcCoA and enhance the metabolic flux towards both cell growth via the TCA cycle, and fatty acid biosynthesis in the growth phase. The third source of AcCoA, which derived from amino acid degradation, has largely been ignored as a relevant pathway for bioengineering. Despite the fact that it bypasses pyruvate and the highly controlled PDHC/PDK regulatory process, it was considered insufficient for fatty acid biosynthesis (Garrett and Grisham, 2013 ). We hypothesized that increase of AcCoA pool by multiple routes could trigger TAEP. In our study, from the activation of pyruvate to AcCoA reaction by addition of sodium dichloroacetate (DCA) to release the PDHC/PDK regulatory process, we achieved TAEP in the wild‐type (WT) D. tertiolecta . Besides this conventional de novo synthetic pathway, we questioned the contribution of amino acid degradation on TAEP, although it has largely been ignored. Through performing genetic engineering, we generated mutants, which exhibited pronounced TAEP with little compromise on growth rate. By employing transcriptomics and metabolomics, key phenotypic regulatory characteristics of lipogenesis in this microalga were uncovered, implying that a secondary contributor of AcCoA derived from amino acid catabolism, in particular branched‐chain amino acid catabolism, contributed to TAEP. Although no direct transport of AcCoA between subcellular compartments was reported in plant cells, a PDHC bypass pathway from activation of free acetate into AcCoA exists (Li‐Beisson et al ., 2013 ; Lin and Oliver, 2008 ). These two major approaches were proposed in our three‐step α loop model. The results highlight the complex interplay between microalgal cellular proliferation and carbon flux in lipogenesis and suggested that genetic and metabolic manipulations targeted at amino acid catabolism could be used to increase accumulation of fuel‐relevant molecules in microalgae in the exponential growth phase.",
"discussion": "Discussion The primary physiological purpose of amino acids is to serve as the building blocks for protein biosynthesis in eukaryotic cells, and as a consequence, the amount of free amino acids is trivial under most circumstances especially in cultures under high growth rate (Garrett and Grisham, 2013 ). Amino acids are derived from the TCA cycle, which provides carbon skeletons via 2‐oxoglutarate or oxaloacetate (Kanehisa et al ., 2016 ; Lane, 2009 ; Wagner, 2014 ). New amino acids can also be formed from transamination by transferring the amino group to a ketoacid (Booth, 2000 ). In our case, a significant proportion of free amino acids were degraded, driven by transcriptional up‐regulation of DtIVD , DtMCCB and DtACCA encoding for the key enzymes in the BCAA catabolic pathway, leading to AcCoA synthesis. This strategy shows the feasibility of using the aforementioned third source for AcCoA, which was previously ignored. The functional role of BCAA catabolic process in lipogenesis has been demonstrated in other various organisms. In the diatom, Phaeodactylum tricornutum , inhibition of MCCB expression by RNA interference disturbed the carbon flux, resulting in decreased TAG accumulation and impaired biomass growth (Ge et al ., 2014 ). Green et al . ( 2016 ) highlighted the contribution of BCAAs to adipocyte metabolism in mouse cell line (3T3‐L1 cells) and demonstrated that amino acids (BCAAs in particular) from both extracellular sources and protein catabolism were highly utilized by differentiated adipocytes. Inhibition of BCAA catabolism negatively influenced 3T3‐L1 adipogenesis. In the study of Peng et al . ( 2015 ), BCAA catabolic mutants defective in enzymes both upstream and downstream of IVD displayed enhanced senescence in prolonged darkness, showing that function of BCAA catabolism in providing TCA cycle substrates in energy‐limited conditions. It also demonstrated that IVD influences energy homeostasis in multiple ways, providing BCAA catabolic CoA intermediates to the mitochondrial electron transport chain, as well as catabolizing additional substrates such as phytanoyl‐CoA and aromatic amino acids (Araújo et al ., 2010 ; Ishizaki et al ., 2005 ). Interestingly, other amino acid catabolic pathways were found in concordant in contributing to the TCA cycle and AcCoA production. Lysine metabolism was previously demonstrated to interact with plant energy metabolism (Angelovici et al ., 2011 ). Vorapreeda et al . ( 2012 ) also reported that leucine and lysine degradation in oleaginous fungi provided the alternative substrate for AcCoA as the precursor for lipid production, by contrast to that in nonoleaginous fungi. This is confirmed by the free amino acid (leucine) uptake study reported here. Free amino acid uptake (transport, assimilation /accumulation) and excretion had been observed in microalgae (Flynn and Butler, 1986 ; Huo et al ., 2011 ). Three exogenous transamination and deamination cycles introduced by Huo et al . ( 2011 ) also reengineered carbon flux for fuel‐convertible amino acids and enabled protein hydrolysates to be used for fuel production. In conclusion, we have studied a TAG‐rich mutant strain of D. tertiolecta under controlled laboratory conditions to advance our understanding of lipid metabolic pathways in the growth phase. Our study revealed a ‘three‐step α loop’ model to elucidate lipogenesis in the exponential growth phase as shown in Figure 6 and summarized below. Figure 6 Regulation of metabolic pathways related to energy/carbon capture and conversion in Dunaliella tertiolecta mutant G11_7. The vertical path Under normal conditions, a number of key enzymes in microalgae supply carbon precursors for de novo fatty acid synthesis, which include those involved in PDHC, glycolysis and suites of specific transporters. They were found substantially up‐regulated under nitrogen deprivation conditions (Li et al ., 2014 ). This pyruvate‐dependent glycolysis pathway is tightly regulated by cell growth via PDHC/PDK cascade (Figure 6 , route 1, 2). AcCoA and NADPH produced from this pathway are readily used by the TCA cycle to produce amino acids, macromolecules and energy for biomass growth. During this process, a minor fraction of AcCoA and NADPH is used for fatty acids synthesis (Carpinelli et al ., 2014 ). The back loop While in G11_7 mutant, genes involved in the amino acid catabolism in mitochondria were enhanced. The overflow free amino acids and catabolized proteins were channelled into AcCoA. This provides an additional source of AcCoA for fatty acid synthesis (Figure 6 , route 5, 6). The recycling of amino acids at a moderate level led to a temporal increase in AcCoA concentration, with little compromise on the biomass growth rate (4%). This elevated AcCoA concentration (30%) exceeds the demand (set point) for biomass growth, which shunts the carbon and energy precursors to the fatty acid synthesis route (Figure 6 , route 7). Genes responsible for fatty acid synthesis were constantly overexpressed at the mRNA level, and this equilibrium leads to ultimate generation of TAG in the growth phase. The pull‐down In response to the drawn‐down of [AcCoA] for lipogenesis, a pull‐down of carbon flux from photosynthetic process takes effect, leading to the increase in photosynthetic rate (33%), and overexpression of genes participating in photosynthesis and glycolysis. The inclusion of DCA removed the regulation of PDHC, leading to an increase in AcCoA and accumulation in TAG (42%) (Figure 6 , route 8). In consistency with our observation, bioengineering manipulation, such as antisense knock‐down of PDK in the diatom P. tricornutum , was reported to promote TAG production by 82% (Ma et al ., 2014 ). The further increase in the TAG accumulation in the stationary growth phase in HL under the treatment of DCA (PDHC activated) could be explained by the following: in the stationary phase, as cells no longer drawn‐down AcCoA for growth, the carbon and energy generated from photosynthesis was channelled to TAG production. Moreover, the cell membrane lipid also serves as the source of BCAA, which is degraded and contributes to the AcCoA reservoir in the stationary growth phase. The proposed ‘three‐step α loop’ model suggests that the rate of lipogenesis in the growth phase is determined by the balance between the carbon/energy supply, biomass synthesis (growth) and amino acid catabolism. When the carbon metabolism is at high gears, a rapid catabolic rate would lead to accumulation of AcCoA and fatty acid synthesis, without compromise on biomass growth. However, an overrun amino acid catabolism would result in reduced growth rate, lower biomass concentration and ultimately lower TAG productivity of the culture. Collectively, TAEP in microalgae could be stimulated by elevated AcCoA level through multiple approaches. Besides aforementioned two major approaches, a balance for fatty acid oxidation is also of great interest to be investigated. To our knowledge, no genetic manipulation has been achieved in promoting TAG production in the growth phase. Collectively, the deliberated investigation provides targets for metabolic engineering of eukaryotic microalgae for efficient lipid production and may inspire novel biofuel production technology based on growth‐phase lipid‐producing oleaginous microalgae as alternative biofuel feedstocks."
} | 3,785 |
38496961 | PMC10938392 | pmc | 4,349 | {
"abstract": "The influence of enrichment of culturable microorganisms\nin in\nsitu coal seams on biomethane production potential of other coal seams\nhas been rarely studied. In this study, we enriched culturable microorganisms\nfrom three in situ coal seams with three coal ranks and conducted\nindoor anaerobic biomethane production experiments. Microbial community\ncomposition, gene functions, and metabolites in different culture\nunits by 16S rRNA high-throughput sequencing combined with liquid\nchromatography-mass spectrometry-time-of-flight (LC-MS-TOF). The results\nshowed that biomethane production in the bituminous coal group (BC)cc\nresulted in the highest methane yield of 243.3 μmol/g, which\nwas 12.3 times higher than that in the control group (CK). Meanwhile,\nMethanosarcina was the dominant archaeal genus in the three experimental\ngroups (37.42 ± 11.16–52.62 ± 2.10%), while its share\nin the CK was only 2.91 ± 0.48%. Based on the functional annotation,\nthe relative abundance of functional genes in the three experimental\ngroups was mainly related to the metabolism of nitrogen-containing\nheterocyclic compounds such as purines and pyrimidines. Metabolite\nanalysis showed that enriched microorganisms promoted the degradation\nof a total of 778 organic substances in bituminous coal, including\n55 significantly different metabolites (e.g., purines and pyrimidines).\nBased on genomic and metabolomic analyses, this paper reconstructed\nthe heterocyclic compounds degradation coupled methane metabolism\npathway and thereby preliminarily elucidated that enriched culturable\nbacteria from different coal-rank seams could promote the degradation\nof bituminous coal and intensify biogenic methane yields.",
"introduction": "1 Introduction The global energy revolution\nis accelerating, traditional oil and\ngas resources are close to depletion, 1 , 2 and unconventional\nnatural gas, represented by coalbed methane (CBM), has become an important\nenergy source for energy transformation due to its low pollution to\nthe environment. 3 , 4 According to the gas components,\nthe genesis of CBM can be categorized into biogenic and thermal genesis, 5 and according to the different formation times\nof CBM, the biomethane can be categorized into primary biogenic gas\nand secondary biogenic gas. 6 , 7 Secondary biogenic gas\nis a mixed gas produced by microorganisms, accounting for nearly 20%\nof coalbed methane. 8 The application\nof microorganisms to enhance biomethane production\nis a hot topic in the field of coalbed methane bioengineering, and\nsystematic and in-depth research is being carried out. 9 Secondary biogas production largely depends on the cometabolism\nbetween microbial communities. 10 , 11 Specifically, it is\ndifficult or impossible for methanogen to directly utilize the macromolecular\norganic matter in the coal in the polymerized state, 12 and it is necessary to rely on the bacterial community\nwith degradation function to degrade the macromolecular organic matter\nand generate the substrate (CO 2 , hydrogen, acetate, etc.)\nthat can be directly utilized by methanogenic archaea and finally\ncomplete the process of biomethane generation. 13 , 14 At present, researchers have explored various methods to stimulate\nthe potential of mine microbes to generate methane. Wang et al. 15 found that enrichment of in situ methanogen\nstrains and addition of exogenous microorganisms can promote in situ\nbiogenic methane formation in coal seams. Li et al. 16 found that altering actinomycete activity can directly\naffect the negative cohesion of microbial communities and, thus, the\npotential for methane production. Liu et al. 17 determined the feasibility that enrichment and cultivation of methanogenic\nbacteria from bituminous coal could be used to improve the production\nof biological CBM. Guo et al. 18 found that\nthe addition of exogenous carbon could improve the low bacterial activity\nin lignite and promote its biomethane potential. Davis et al. 19 suggested that activation of native microorganisms\nin coal seams by the addition of nutrients such as algae amendments\ncould promote biomethane production. These studies have shown that\nthe activation of in situ microorganisms in coal seams is the key\nto biogenic methane production. However, the environments of coal\nseams in different mining areas and at different depths are intricate,\nand the methanogenic potential of in situ microorganisms in coal seams\nof different coal qualities needs to be systematically analyzed. Coal is a condensed aromatic system composed of aliphatic compounds,\naromatic hydrocarbons, and heteroatom compounds. 20 During the anaerobic degradation of coal, the more bioavailable\naliphatic compounds are degraded first to methanogenic substrates,\nfollowed by water-soluble heterocyclic compounds. 21 However, aromatic compounds are difficult to biodegrade\ndue to the presence of benzene rings. Shi et al. 22 found that microbial reactors degrade difficult-to-degrade\nheterocyclic aromatic compounds in coal gasification wastewater and\ngenerate methanogenesis-related substances such as acetic acid and\nCO 2 . Fu et al. 23 observed the\ncomplete degradation of indole to acetate under sulfate-reducing conditions,\nand the degradation pathway was more similar to the methanogenic metabolic\nprocess. All previous studies have shown that degradation of organic\nmatter by functional bacteria is an important part of biomethane production.\nHowever, there has been a lack of studies exploring the degradation\nprocesses of these organics and related studies to identify, annotate,\nand categorize specific products. 24 Based on previous studies, in situ microorganisms from three different\ncoal qualities were selected in this study for enrichment and added\nto high-volatile bituminous coals for simulated methane production\nexperiments. First, this study compared the methanogenic potential\nof different enriched microorganisms, then determined their community\nstructure, gene function, and metabolites, and finally constructed\nthe coupled relationship between degradation metabolism and methanogenic\nmetabolism of bacteria through genomics and metabolomics to elucidate\nthe key process of bituminous coal biomethane formation under the\naction of enriched microorganisms. The results of the study help to\nreveal the metabolic mechanism of biomethane formation by enriched\nmicroorganisms and contribute to the exploration of biomethane formation\nin coal seams –and its exploitation and utilization.",
"discussion": "4 Discussion In this study, it was found\nthat the enrichment of target flora\nfrom the native flora of different coal seams could all significantly\nenhance the methane production potential of high-volatile bituminous\ncoals. In addition, the different strain compositions of different\nmicroorganisms resulted in differences in their methane production\npotential. Biomethane production is the result of cometabolism of\nmultiple functional flora in an anaerobic environment, which is subject\nto a variety of endogenous influences, including strain composition\nand diversity. 27 Currently, most biocoalbed\nmethane research focuses on increasing microbial diversity with the\naim of enhancing interactions between microorganisms to promote coal\ndegradation and provide more metabolic substrates for methanogen flora. 28 , 29 This study also found that enrichment of in situ microorganisms\nin coal seams significantly increased their diversity, which in turn\nenhanced bacterial biodegradation of bituminous coal and produced\nabundant aromatic, aliphatic, and alkane compounds, which in further\ndegradation provided methanogens with the necessary metabolic substrates\nand ultimately increased methane production. However, comparing the\nthree experimental groups, the methane production of the more diverse\nCC group was lower than that of the less diverse BC group, which could\nbe attributed to the fact that the enrichment targeting the methanogenic-associated\nmicrobial communities in the coal beds made the methanogenic metabolism\nfunctional flora become the main dominant flora in the culture system.\nIn addition, sulfate-reducing bacteria (SRB) were enriched in the\nCC group. Previous studies have shown that SRB and methanogen coexist\nin a symbiotic manner in the metabolic pathway of microbial methane\nproduction in coal seams. 30 These SRB can\nutilize casein as a terminal autoreceptor or shuttle to metabolize\nacetic acid or other simple fatty acids, an important step in coal\ndegradation. 31 However, some studies have\nfound that toxicants in the substrates of SRB metabolism, such as\nammonia nitrogen compounds, H 2 S, and sulfate, can stress\nthe succession of methanogenic bacterial communities. 32 The results of the present study also revealed changes\nin the structure of the methanogen community. In the sulfur metabolic\npathway, SRB forms APS by utilizing sulfate activated by sulfate adenylyltransferase,\nAPS reductase reduces APS to HSO 3 – , and\nfinally, sulfite reductase reduces HSO 3 – to H 2 S. 33 These S 2– ions, produced by SRB metabolism, further inhibit the metabolic\nactivity of methanogens in CC by inhibiting electron donors to the\nmitochondrial respiratory chain. 34 , 35 Methanogens\nare the true producers of biomethane and dominate the\nfinal segment of the pathway for biomethane formation, converting\nlow-molecular-weight intermediates (CO 2 , H 2 ,\nAcetate, etc.) to methane. 36 Previous studies\nhave classified methanogens into three major groups based on their\nmetabolic substrates, including hydrogenotrophic archaea that convert\nH 2 and CO 2 to methane, acetoclastic archaea\nthat convert acetate compounds to methane, and methylotrophic archaea\nthat convert methyl compounds to methane. 37 Dominant methanogenic archaea such as methanosarcina identified\nin this study are closely related to methylotrophic archaea. These\narchaea usually do not require organic growth factors and can produce\nmethane in a variety of environments through different pathways, including\ndisproportionation to CH 4 and CO 2 and NH 3 when methylamine and methanol are present in the substrate;\nwhen H 2 is present in the environment, methanol and methanol\nare reduced to CH 4 ; and also, the substrates that are available\ninclude H 2 + CO 2 or acetic acid analogues. 38 , 39 This diversity of metabolic pathways enables the strain to adapt\nto the time-changing culture environment in the laboratory system\nand has excellent methanogenic potential. The effect of microbial\nmetabolism and degradation on organic matter\nfractions has been an unelucidated research hotspot due to the complex\nstructure of the macromolecular skeleton in coal. In this study, genomic\nand metabolomic analyses revealed that the enrichment of enriched\nmicroorganisms directly promoted the degradation of macromolecules\nin bituminous coal to produce a large number of intermediate metabolites,\nsuch as heterocyclic compounds, benzene compounds, and aliphatic compounds,\nand further facilitated the fracture and hydrolysis of these intermediates\non branched chains to produce organic acids, lipids, and amides, which\nare the prerequisites for the production of central metabolic intermediates\n(acetate-CoA) of biomethane, and also important signaling molecules\nfor microbial anaerobic fermentation. 40 Furthermore, unlike in other studies, the degradation of nitrogen-containing\nheterocyclic compounds in bituminous coal analyzed in this study is\nlikely to be a key pathway for methane formation. During the degradation\nof nitrogen-containing heterocyclic compounds, 2-oxoisocaproate dehydrogenase\n[EC:1.2.4.4], uric acid oxidase [EC:1.7.3.3], and dihydropyrimidine\ndehydrogenase [EC:1.3.1.2] act on the reductive cleavage, and dihydropyrimidinase\n[EC:3.5.2.2] acted on hydroxylation reactions to open nitrogen-containing\nheterocyclic compounds in the closed state, yielding organonitrogen\ncompounds such as formiminoglycine, 3-ureidopropionate, and others. 41 Meanwhile, this also indicates that the acquisition\nof nitrogen is an important process in microbial metabolism, and the\nlow content of nitrogen in bituminous coal may limit microbial metabolism\nand growth. 42 Shi et al. 43 found that microbial degradation of organic pollutants\nwas related to nitrogen response and further determined that controlling\norganic nitrogen content could effectively influence the degradation\nof phenol, acetaminophen, and sulfamethoxazole in industrial wastewater.\nWild et al. 44 found that increasing nitrogen\nlevels in the soil can promote soil microbial metabolic activity at\nlow temperatures and further lead to the entry of soil carbon into\nthe carbon cycle. These studies indicate that nitrogen regulates the\nbioavailability of organic matter by microorganisms and further affects\ntheir potential for anaerobic fermentation to form methane. Another important finding is that low bioavailability of compounds\nin bituminous coal is an important factor affecting the rate of biomethane\nproduction. Hu et al. 45 found that in anaerobic\nenvironments, it is difficult to degrade most of the benzene ring\ncompounds in wastewater by microbial self-metabolism alone. Similarly,\nthe present study also revealed the presence of several benzene ring\nderivatives in the culture broth at the later stage of the experiment.\nThis suggests that the carbon ring structure of the benzene ring itself\nis relatively stable in anaerobic environments and that enrichment\nof microorganisms merely promotes the removal of functional groups\nsuch as alkyl side chains, carboxyl groups, or amino groups from the\nbenzene ring derivatives, and the breaking of aromatic compounds at\nthe branching, and leads to the accumulation of NH 3 , and\nthe benzene ring in the culture system. This may be the main reason\nfor the change in pH in the culture solution under anaerobic conditions.\nThe production of organic acids and carbonation of CO 2 during\ndegradation first reduced the pH value of the fermentation broth,\nand the metabolic utilization of organic acids and hydrolysis of NH 3 contributed to the alkaline tendency of the fermentation\nbroth. 46 To counteract the acidic conditions\nin the culture broth, functional strains had to devote more of their\nenergy to maintaining the expression drive of genes encoding ribosomal\nproteins, membrane transport, energy metabolism, and replication and\nrepair. In conclusion, this study confirmed that the enrichment\nof microorganisms\nfrom all three coal seams, bituminous, coking, and anthracite, could\npromote the anaerobic degradation of highly volatile bituminous coal\nand the production of biomethane. Degradation of nitrogen-containing\nheterocyclic compounds in bituminous coal may be a key reaction in\nthe methanogenesis process, and the degradation of such compounds\nproduces acetyl cofactor and lysine, which are important metabolic\nsubstrates for the two methanogenic pathways acetate ≥ methane\n(M00357) and CO 2 ≥ methane (M00567). This study\nis expected to provide functional microbial strains for coalbed methane\nextraction and enhance the development and utilization of coalbed\nmethane resources."
} | 3,772 |
27466140 | null | s2 | 4,352 | {
"abstract": "We have demonstrated in situ fabricated and acoustically actuated microrotors. A polymeric microrotor with predefined oscillating sharp-edge structures is fabricated in situ by applying a patterned UV light to polymerize a photocrosslinkable polyethylene glycol solution inside a microchannel around a polydimethylsiloxane axle. To actuate the microrotors by oscillating the sharp-edge structures, we employed piezoelectric transducers which generate tunable acoustic waves. The resulting acoustic streaming flows rotate the microrotors. The rotation rate is tuned by controlling the peak-to-peak voltage applied to the transducer. A 6-arm microrotor can exceed 1200 revolutions per minute. Our technique is an integration of single-step microfabrication, instant assembly around the axle, and easy acoustic actuation for various applications in microfluidics and microelectromechanical systems (MEMS)."
} | 225 |
25866055 | null | s2 | 4,355 | {
"abstract": "Genome data have created new opportunities to untangle evolutionary processes shaping microbial variation. Among bacteria, long-term mutualists of insects represent the smallest and (typically) most AT-rich genomes. Evolutionary theory provides a context to predict how an endosymbiotic lifestyle may alter fundamental evolutionary processes--mutation, selection, genetic drift, and recombination--and thus contribute to extreme genomic outcomes. These predictions can then be explored by comparing evolutionary rates, genome size and stability, and base compositional biases across endosymbiotic and free-living bacteria. Recent surprises from such comparisons include genome reduction among uncultured, free-living species. Some studies suggest that selection generally drives this streamlining, while drift drives genome reduction in endosymbionts; however, this remains an hypothesis requiring additional data. Unexpected evidence of selection acting on endosymbiont GC content hints that even weak selection may be effective in some long-term mutualists. Moving forward, intraspecific analysis offers a promising approach to distinguish underlying mechanisms, by testing the null hypothesis of neutrality and by quantifying mutational spectra. Such analyses may clarify whether endosymbionts and free-living bacteria occupy distinct evolutionary trajectories or, alternatively, represent varied outcomes of similar underlying forces."
} | 359 |
35541509 | PMC9078960 | pmc | 4,356 | {
"abstract": "The aquatic fern salvinia can retain an air layer on its hairy leaf surface when submerged under water, which is an inspiration for biomimetic applications like drag reduction. In this research, an electrostatic flocking technique is used to produce a hairy surface to mimic the air-trapping performance of the salvinia leaf. Viscose and nylon flocks with different sizes were selected. A volumetric method was established to analyze the air-retaining performance of the flocking samples, Salvinia molesta and lotus leaves as well. Through air volume change analyses, it is found that another factor that can affect the Salvinia molesta air-retaining ability is the curving of the leaf under water. A flocking sample fabricated by a kind of nylon flock is demonstrated to have a comparable air-retaining ability under static conditions as a Salvinia molesta leaf in its flat form.",
"conclusion": "4. Conclusions In this research, a hairy surface was fabricated by electrostatic flocking to mimic the air-trapping ability of the salvinia leaf. In a series of experiments a flocking sample marked as 5C which was fabricated with a kind of nylon fiber was found to retain air for more than 600 hours under water. A volumetric method to measure the air-volume change on the surface under water was established. This method was used to further analyze the air-retaining properties (total air volume, air bubble loss, air dissolved and air retention) of samples. Our result showed that the curving of the Salvinia molesta leaf assisted air retention by preventing air bubble loss, and decreasing the air diffusion area. Flocking sample 5C, whose hair density was much higher than the Salvinia molesta leaf, showed a comparable air-retaining ability to the salvinia-flat sample. The air lost as bubbles from sample 5C was more than from salvinia-flat, but the air diffusion of sample 5C was less, though biotic factors of plant samples should be considered for salvinia.",
"introduction": "1. Introduction Throughout millions of years, nature has developed and perfected itself in a magical way and it inspires us to create various materials, products and devices with special abilities. 1–8 The emerging field of biomimetics is of high scientific and economic interest. One of the classic examples of a biomimetic application is known as the ‘lotus effect’ due to the superhydrophobic and self-cleaning properties of surfaces represented by the lotus leaf. 9,10 The rolling water droplet on the leaf surface is in the so-called Cassie–Baxter state with air pockets between water and the micro- and nanostructures of the surface. 11 When lotus leaves are immersed in water, they display a silvery shine because of the air trapped between the hierarchical surface structures. This is another property associated with the formation of the Cassie–Baxter state on superhydrophobic surfaces – the ability to retain an air layer while submerged under water. 12–29 The underwater air layer has gained significant interest with regard to drag reduction, opening possibilities for biomimetic applications such as low-friction fluid transport and drag-reducing ship coatings. 14 For such applications the durability of the air layer is of the greatest importance. However, the prototype based on the lotus leaf surface is limited by the short time the air layer persisted. In contrast to species like the lotus leaf, other species such as aquatic ferns of the genus Salvinia were found to have high air layer persistence. The air-retaining properties of salvinia have been analyzed by Barthlott et al. 15–19 Other than the similar surface nanostructure of wax crystalloids like the lotus leaf, the salvinia leaf surface has a different microstructure: the leaf surface is covered with tiny hairs instead of papillose epidermal cells (as with the lotus leaf). Specifically, one of the Salvinia species, namely Salvinia molesta , shows that each leaf forms a concave shape underwater and the hierarchical architecture of the leaf surface is dominated by tiny hydrophobic eggbeater-shaped hairs coated with a nanostructure of wax crystalloids except for the terminal cells of each hair which form hydrophilic patches (as shown in Fig. 1 ). The three main features of the Salvinia molesta leaf, the elastic eggbeater-shaped hairs, the superhydrophobic leaf surface and the hydrophilic patches on top of the eggbeater-shaped hairs combine together to retain a layer of air when the leaf is submerged under water. Fig. 1 \n Salvinia molesta : (a) concave-shaped leaf, (b) a water droplet pinned on the hair of the leaf, (c) a SEM image of the eggbeater structure created by four hairs, (d) the hydrophilic patch, and (e) the nanostructure of wax crystalloids on the hair surface. Preliminary advances have been achieved in creating structures to mimic the behavior of salvinia in the past few years. Bhushan and Sung et al. 20,21 studied salvinia and lotus effects by fabricating micropillar structures on a silicon base through photolithography. Bhushan proved that a structure can be created in the lab to mimic the behavior of salvinia; Sung’s research indicated that the air-retaining properties can be greatly enhanced using the salvinia structure compared to the lotus one. Tricinci et al. 22 used a 3D laser lithography technique to fabricate a 3D patterned surface bio-inspired by salvinia and the sample showed interesting properties of hydrophobicity and air retention when submerged under water. Holscher et al. 23,24 fabricated a nanofur by a hot pulling method to mimic the salvinia effect cost-effectively. Methods were developed to measure the underwater performance of the bionic samples. Mayser et al. 17 and Gad-el-Hak et al. 25 detected and calculated the trapped air on superhydrophobic surfaces through the change of the buoyancy force. Drag reduction was characterized by Holscher et al. 24 measuring the pressure drop across the channel walls which were covered by the nanofur. Concerning the hairy structure on the salvinia surface and the elasticity of the salvinia hairs, utilizing textile fibers with flexibility in nature to form a hairy structure through an electrostatic flocking process might be an effective biomimetic method. The electrostatic flocking technique uses a high voltage electrical field to plant ‘flocks’ (short fibers which have been given a pre-electrostatic flocking treatment to improve separation and flying properties) into a thin layer of adhesive on the substrate. This technique has been studied as an anti-biofouling approach in aquaculture and navigation to decrease material wastage and friction on the hull. 30–33 In this research, a hairy surface was tailored by the flocking technique and then the flocking sample was treated with chemicals to obtain superhydrophobicity. In the meantime a method was established to evaluate the air-retaining ability by measuring the air bubble loss from the samples’ surfaces under water.",
"discussion": "3. Results and discussion Abundant flocking samples with different flock fibers and flock densities were obtained by varying the output voltage of the flocking machine and the flocking time. It was found that the underwater air layer persistence increased along with the flock density. During flocking, the flock density increased rapidly when the flocking time was below 10 s and increased slightly between 10–30 s. It was considered that at the flocking time of 30 s, the samples reached their maximum density. It had been found that when the flocking time exceeded 10 s, the flocking samples would show significant air layer persistence. The results of the CA and RA measurements together with the primary underwater tests for flocking samples with flocking times of 10, 20, and 30 s are listed in Table 2 . Those of Salvinia molesta and the lotus leaves are in Table 3 . As the length of the flock fiber increases, more fibers would lean or even lie on the surface during flocking, which was especially noticeable in sample 6. In order to enable longer fibers to stand more vertically, the idea of making a sample with hybrid flocking fibers emerged. Sample 7 was made by a two-step electrostatic flocking process using two kinds of flock fibers (first step using N1, second N4). In this situation the shorter fibers would support the longer ones. Another advantage for the hybrid sample was that the double structure may be helpful for air retention. 14,26,28 According to the CA result, it can be seen that the hydrophobic treatment was suitable as most of the flocking samples except samples 2B and 6 reached a CA of 140° to 150° which was similar to Salvinia molesta and lotus leaves. Sample 6 showed a similar shape of the water droplet in the CA test to the other samples, however, along with the increased flock fiber size, the average gap between the fibers increased. The gap (0.034 mm) between the flock fibers of sample 6 was much larger than those of other samples (in 5C for example, the gap was 0.017 mm). Moreover, with the leaning of the fiber, the water droplet sank into its surface and the reading of the contact angle decreased (in the sample 6 series, samples prepared with a flocking time below 30 s performed even worse, so were not included). Flocking parameters and air-retaining times of flocking samples Sample Flock fiber Height (mm) Diameter (μm) Flocking voltage (kV) Flocking time (s) CA (°) RA (°) Air-retaining time (h) 1A V1 0.30 ± 0.05 10.6 50 10 143.6 10 180–250 1B 20 145.7 6 200–240 1C 30 147.5 6 210–260 2A V2 0.50 ± 0.05 14 50 10 145.1 23 250–310 2B 20 128.4 21 250–330 2C 30 143.9 16 270–330 3A N1 0.50 ± 0.05 14 50 10 151.2 26 250–310 3B 20 147.6 23 250–300 3C 30 149.4 21 290–310 4A N2 0.70 ± 0.05 20 65 10 144.2 43 300–330 4B 20 143.8 42 320–350 4C 30 146.3 39 330–390 5A N3 0.90 ± 0.05 22 65 10 149.5 45 490–550 5B 20 145.8 39 520–600 5C 30 140.3 39 530–610 6 N4 2.90 ± 0.05 66 80 30 126.9 >90 210–300 7 N1 0.50 ± 0.05 14 50 30 145.4 >90 310–350 N4 2.90 ± 0.05 66 80 CAs, RAs and air-retaining times of Salvinia molesta and lotus leaves Sample CA (°) RA (°) Air-retaining time (h) \n Salvinia molesta \n 145.3 >90 >600 Lotus 146.3 3 <24 It was found that if the droplet volume was increased up to 30 μL, the droplet could still be supported by all of the flocking samples in Table 2 without penetration although the sunken meniscus increased. An increased contact angle hysteresis was shown to increase along with the flock size, or rather the flock diameter, as demonstrated in Table 2 where the roll off angle increased with the flock diameter. Samples 6 and 7 showed large RAs which were mainly caused by the water partially sinking into the surface made up of large fibers (for sample 7, the baseline of the water droplet was corrected to the N1 layer thus the CA was still above 140°). In addition, samples 1 to 5 were included in the evaporation test where the droplets revealed their receding contact angles. It can be seen from Fig. 3 that although a slight increase in the receding angle along with an increased flock diameter can be observed, all of the samples showed high CAs at the beginning and after 60 min, indicating the Cassie–Baxter state. 9 Fig. 3 Evaporation test of flocking samples, showing the water droplet change from the beginning to 60 min thereafter. It was seen from Table 3 that the air on the lotus leaf disappeared within 24 h while Salvinia molesta retained air on the surface for more than 600 h, in accordance with previous studies where although the lotus leaf is known for its great hydrophobicity, it does not have air-retaining abilities. The flocking series in Table 2 shared a similar phenomenon under water: during the sample submersion in water, several air bubbles formed quickly within seconds and escaped from the sample surface. This part of the air was not considered as part of the air layer. After a sample was placed on the bottom of the aquarium, it would retain a steady air layer for a long time, during which the air bubbles formed very slowly. From samples 1 to 5, the air-retaining time increased slightly when the flocking time increased from 10 to 30 s. For sample 6, as the length and diameter of the flock fiber are much larger than those of the others, the air layer on it was unstable. A large portion of air was lost in the first several minutes, and the silvery layer began to break along with it. Sample 7, with a double structure, as expected showed steadier air retention than sample 6. After 600 h of the primary underwater test, most of the flocking samples were considered to have lost most of the air as the silvery reflection was barely visible. However, on sample 5C a portion of air was still left, giving a noticeable scattered silvery reflection. The structural dimensions of salvinia observed were that the pillar average height was 1.8–2 mm and the density of the “eggbeater” as a whole was 1.5–2 mm −2 . The density of the flocking sample reached hundreds or even thousands per square millimeter which was much higher than salvinia. Probably because of the lack of help from the eggbeater structure and the hydrophilic tip, a high density of flock fiber was essential to achieve a satisfactory air retention. As for the flock material, there was no significant difference between the performances of viscose and nylon fibers. Although viscose fibers are relatively more hydrophilic, after the hydrophobic treatment, they showed a similar air layer persistence compared to nylon fibers of similar sizes (compare 2A to 2C with 3A to 3C). The samples were further subjected to air volume change tests and the results ( V total , V bubble , V dissolved and V left ) of 1C, 2C, 3C, 4C, 5C and 6 are demonstrated in Fig. 4 together with the air-retaining time results for comparison. It can be seen that with an increased flock size, or rather flock height (1C to 5C), V total and the air-retaining time increased, as well as V left ( Fig. 4b ), which was another indicator of air retention. However, when the flock size increased to a certain level, for example sample 6, where the effect of the enlarged gap surpassed the effect of the height increasing, water penetrated between fibers much more easily, leading to the air escaping as bubbles easily, and the air retention decreased. Fig. 4 Air volume change (test time 600 h) and air-retaining time of flocking samples, (a) air-retaining time and V total , (b) V bubble , V dissolved and V left . The best-performing flocking sample 5C ( Fig. 5 ) was tested for air volume changes together with salvinia and lotus leaves. When a Salvinia molesta sample was submerged under water, the two blades of the leaf tend to curve to form a concave shape and it was noticed that as a result air was trapped inside. Therefore a salvinia-flat sample (prepared by gluing the two blades flat on a glass slide to avoid curving) was involved in the test. The air volume changes of flocking sample 5C and salvinia and lotus leaves are demonstrated in Table 4 . Fig. 5 Flocking sample 5C, (a) silvery shine under water as the air layer is retained on its surface; (b) SEM image of the cross-section. Air volume change of flocking sample 5C and salvinia and lotus leaves a Sample \n V \n total \n \n V \n bubble \n \n V \n dissolved \n \n V \n left \n mL/25 cm 2 \n S % mL/25 cm 2 \n S % mL/25 cm 2 \n S % mL/25 cm 2 \n S % Salvinia 3.1 4 0 0 0.9 6 2.3 3 Salvinia-flat 2.1 8 0.5 8 1.2 7 0.4 4 Flocking sample 5C 1.8 4 0.7 6 0.7 12 0.5 7 Lotus 0.1 0 0 0 0.1 0 0 0 a Average value and standard deviation ( S ) of the four air volume parameters, V total , V bubble , V dissolved and V left were calculated from 5 tests (for the lotus, 3 tests), test time t = 600 h. From Table 4 it was seen that salvinia leaves, with a higher hair height and lower hair density, contained a greater total air volume than sample 5C and lotus leaf. It was also seen that V total and V left of salvinia were larger than those of salvinia-flat. This indicated that the curving of the salvinia leaf under water improved the retention of air. As illustrated in Fig. 6 , the curving of the leaf formed an air pocket under water, and as a result more air was retained on it than on the leaf in its flat form. This phenomenon would help when designing different biomimetic applications. Fig. 6 Salvinia leaf under water, (a) photo of salvinia leaf with the air pocket formed, (b) schematic diagram of the curved leaf to form an air pocket, and (c) schematic diagram of the air layer on the leaf in its flat form. As shown in Table 4 , V total of flocking sample 5C was 1.8 mL of air on 25 cm 2 . As the hairy surface of the flocking sample could be modeled as a surface covered with pillars, the theoretical volume of total air held by the sample could be calculated for comparison. The structure parameters, height, diameter and density of sample 5C observed by SEM are shown in Fig. 7 . Fig. 7 Scheme of the surface of flocking sample 5C. The maximum volume of air that can be contained on the flocking sample would be, 5 where L is the side length of the square sample (mm), h is the height of the flock fiber (mm), n is the density of the flock fiber on the sample (mm −2 ) and d is the diameter of the flock fiber (mm). Here the air layer height was arbitrarily considered as the same as the flock height measured where the sagging of the air–water interface and the tilting of the flock fiber from the ideal vertical position under water pressure were ignored. For sample 5C, the calculated air volume according to formula (5) was 1.65–1.83 mL/25 cm 2 , and the measured air volume V total ( Table 4 ) was 1.8 mL/25 cm 2 . The measured value agreed with the calculated result. It is clear that during the air volume change test, no air bubbles were observed to be lost from salvinia and lotus samples. These may stem from different reasons. The air contained on the lotus leaf was too little and scattered between micro papillae to form air bubbles. As for the Salvinia molesta sample, the curving of the leaf functioned as a pocket which would enclose the air so as to prevent air bubble loss. Air bubble loss was observed from the salvinia-flat sample and flocking sample 5C. By combining the time t and V bubble of 5 tests, air bubble loss trends were obtained ( Fig. 8 ). Fig. 8 Air bubble loss over time. In Fig. 8 , the test result points show a certain level of dispersion. If possible, a precise temperature control for this test would be ideal. Nevertheless, the air loss trends are clear, as sample 5C displayed rapid air bubble loss at the beginning which stabilized from around 60 h to the end of the test. This was different from the salvinia-flat sample which showed slow and steady air loss during the test. As the air lost as bubbles was highly dependent on the stability of the air–water interface, the steadier air layer on the salvinia-flat sample at the beginning might be due to the pinning effect on top of the eggbeater hairs. Another reason for the rapid bubble loss from sample 5C could be defects in electrostatic flocking or the hydrophobic coating leading to some areas being easily wetted by water. This was supported by the phenomenon that along with the air bubbles lost from the salvinia-flat sample, the silvery shine was intact on its surface for a much longer period of time compared to sample 5C, where the silvery layer began to break around 50 h, implying non-uniform fabrication. Aside from rapid air loss at the beginning, another feature of the trend of sample 5C was the later steady plateau, which could be explained by the Young–Laplace equation as the distance between fibers was quite small, there was capillary pressure ( P capillary ) caused by surface tension. Along with the sagging of water into the surface, the angle between the fiber and air–water interface would increase leading to the increase in P capillary . Equilibrium was reached when the hydrostatic and ambient pressures were balanced by P capillary and the air pressure inside the air layer. 12,27 Other than air loss with bubbles, air dissolving into water is another main factor of air loss 27,28 and this was also manifested in Table 4 . When under water, the air pressure inside the air layer is higher than atmospheric pressure because of the hydraulic pressure, so the air inside the air layer tends to dissolve into water and diffuses towards the atmosphere. Calculated by formula (4) , we got V dissolved, salvinia-flat > V dissolved, salvinia > V dissolved, 5C . The result that V dissolved, salvinia-flat > V dissolved, salvinia could be explained by air diffusion being proportional to surface area and the contact area of air–water on salvinia is smaller than it is on salvinia-flat. However, to compare sample 5C and the salvinia-flat sample, other than the hair density of sample 5C being much higher than the salvinia hair which leads to a smaller air–water interface area, it should be considered that biotic effects of the plant leaf such as metabolism or withering etc. might affect the real volume of diffused air from salvinia. When the test was ended at 600 h, the salvinia-flat sample and sample 5C had 18% and 25% air left, respectively, indicating good air-retaining ability. Moreover, with the help of curving, the air on the salvinia sample had even more than 50% left. Here, flocking sample 5C showed a similar air-retaining ability to the Salvinia molesta leaf in a flat form under static conditions, although biotic factors might affect the results on the leaf and the measured air layer persistence of the salvinia leaf should be considered as its minimum performance. Through the comparison between flocking samples and sample 5C and salvinia, it can be concluded that in order to obtain a proper air-retaining flocking surface, firstly, a relatively small fiber diameter and high density of fibers on a substrate which maximize the capillary pressure against water penetration would be preferred. Secondly, a higher flock height which would increase V total and extend the air diffusion time would benefit the air-retaining time. Flock fibers were commonly designed with synchronous diameters and heights in order to have enough strength to stand up and withstand suitable strain under pressure. This, to some extent, happened to fit with the demand to withstand underwater pressure and a proper balance between height and diameter would be critical. Furthermore, the air retention under dynamic conditions is still challenging, as the Salvinia molesta leaf has a hydrophilic tip to help stabilize the air–water interface. Further research into areas such as multi-layer structures and hydrophilic tip fabrication should be carried on in order to form an even steadier air layer. In addition, gas compensation methods 28,29 are potential ways to extend underwater air retention."
} | 5,725 |
22792045 | PMC3385630 | pmc | 4,357 | {
"abstract": "Under dark anoxia, the unicellular green algae Chlamydomonas reinhardtii may produce hydrogen by means of its hydrogenase enzymes, in particular HYD1, using reductants derived from the degradation of intercellular carbon stores. Other enzymes belonging to the fermentative pathways compete for the same reductants. A complete understanding of the mechanisms determining the activation of one pathway rather than another will help us engineer Chlamydomonas for fermentative metabolite production, including hydrogen. We examined the expression pattern of the fermentative genes PDC3, LDH1, ADH2, PFL1 , and PFR1 in response to day-night cycles, continuous light, continuous darkness, and low or high oxygen availability, which are all conditions that vary on a regular basis in Chlamydomonas ' natural environment. We found that all genes except PFL1 show daily fluctuations in expression, and that PFR1 differentiated itself from the others in that it is clearly responsive to low oxygen, where as PDC3, LDH1 , and ADH2 are primarily under diurnal regulation. Our results provide evidence that there exist at least three different regulatory mechanisms within the fermentative pathways and suggest that the fermentative pathways are not redundant but rather that availability of a variety of pathways allows for a differential metabolic response to different environmental conditions.",
"introduction": "1. Introduction Interest in the unicellular green alga Chlamydomonas reinhardtii (referred to here as Chlamydomonas throughout) has increased over the past decade in the hope that it may one day be possible to harness its exceptional capacity to produce hydrogen. Hydrogen may be used as a renewable energy carrier whose combustion does not release CO 2 into the atmosphere, rendering it more attractive than other potential renewable energies such as ethanol [ 1 ]. \n Chlamydomonas synthesizes hydrogen in a reaction catalyzed mainly by its hydrogenase1 (HYD1) enzyme [ 2 ]. Due to the hydrogenases' extreme sensitivity to oxygen [ 3 ], sustained H 2 synthesis occurs only under anoxia. In its natural environment, Chlamydomonas may be subjected to hypoxia or anoxia on a daily basis after sunset, when the absence of photosynthesis coincides with high rates of respiration [ 4 ]. In the lab, low oxygen cultures may also be obtained in the light in cases where photosynthetic O 2 evolution does not cover respiratory requirements, for example, by anaerobic gas influxation into liquid cultures or a number hours after being subjected to sulfur starvation [ 5 , 6 ]. A number of studies have shown that Chlamydomonas acclimates to anoxia by changing its metabolism from aerobic to fermentative [ 7 – 11 ]. In dark anoxia the source of electrons for the hydrogenases must come from the degradation of intercellular carbon stores, probably starch [ 7 ] as Chlamydomonas does not appear to assimilate extracellular sugars. In this path, starch is broken into glucose and metabolized to pyruvate via glycolysis, which is then converted to acetyl-CoA by either pyruvate-formate lyase (PFL1) or pyruvate-ferredoxin oxidoreductase (PFR1). One molecule of CO 2 is released in the PFR1 reaction, and ferredoxin is reduced. It is ferredoxin which passes electrons to the hydrogenase [ 12 ]. The acetyl-CoA generated may be further metabolized to acetic acid in two steps by phosphate acetyl-transferase (PAT1 or PAT2) and acetate kinase (ACK1 or ACK2) generating an ATP [ 11 , 13 ]. Alternatively, acetyl-CoA may be converted to ethanol by alcohol dehydrogenase1 (ADH1) reoxidizing two NADH and thereby allowing glycolysis to continue [ 14 ]. In Chlamydomonas , in addition to these H 2 /CO 2 /acetate/ethanol-generating pathways, a number of other fermentative pathways leading to a range of products are present, among which, malate, lactate, and succinate [ 13 , 15 , 16 ]. The relative production of each metabolite depends on culture conditions [ 7 , 17 ]. The capability of Chlamydomonas to vary its fermentation profile is known as “flexibility”, and it is of relevance for the development of strategies to engineer microorganisms for renewable energy production [ 18 ]. However we still lack a detailed understanding of factors determining relative contribution of each pathway in response to stress and the role of each pathway in sustaining cell viability during anaerobiosis. A number of factors that may function as switch between pathways have been suggested. Acidification of the cytoplasm, a characteristic feature of the response of many plant tissues to oxygen deprivation [ 19 , 20 ] activates pyruvate decarboxylase (PDC) and inhibits lactate dehydrogenase (LDH) in higher plants. The cell thereby reduces lactate production and redirects the metabolism towards ethanol fermentation [ 21 ]. The degree of anoxia may be another potential path switch determinant. The proportion ethanol : acetate generated has been shown to depend upon oxygen tension [ 22 , 23 ]. Different time points following anoxia induction may show different metabolic profiles. Such is the case when anoxia is induced by sulfur starvation in the light [ 16 ]. The activity of each pathway may influence the activity of the others in a type of cross-regulation [ 24 ] and this phenomenon is beginning to be studied by selectively blocking individual pathways using mutants or pharmaceuticals. A hydEF-deficient strain which produces no functional hydrogenase enzymes activated a new succinate fermentation pathway under anoxia which was not present in the background strain instead of upregulating a preexistent pathway [ 13 ]. A strain with an insertional mutation in the ADH1 gene which does not produce any ethanol in anoxia and in which no ADH1 protein can be detected also upregulates a new pathway, the synthesis and extracellular accumulation of glycerol, not present in the background strain [ 14 ]. A mutant strain unable to synthesize the PFL1 protein secretes no formate but produces more ethanol, D-lactate, and CO 2 than the wild type [ 24 ]. Interestingly, reduced levels of HYD1 transcript and HYD1 protein were found in pfl1 [ 24 ]. This was unexpected as PFL1 may be viewed as a competitor for electrons with HYD1, such that its elimination should have forced more electrons through PFR1 and on to the hydrogenase. It was suggested that the metabolite produced by PFL1, formate, could play a role in gene regulation and influence the ratio of fermentative products [ 24 ]. In Chlamydomonas the effects of formate have not been determined, but the plf1 mutant provides us with further insight into the complexity of the regulation of fermentative metabolism and proves that the activation/deactivation of one pathway does not necessarily influence metabolic fluxes the way we expect. Light availability (or absence) is also of relevance. Diverse conditions of light and dark affect starch degradation, fermentative gene regulation and enzyme activity [ 7 , 13 , 24 – 26 ], for example, the PFL1 gene is not upregulated in response to anoxia under light to the same extent as it is in response to dark anoxia [ 23 ]. Light is an input which regulates which “sets the time” of the circadian clock. In this context it is interesting to note that Chlamydomonas shows diurnal variation of intercellular starch reserves, with a peak starch content in the middle of the night [ 27 ]. A number of genes involved in carbohydrate metabolism and fermentation were found to be under circadian control including D-lactate dehydrogenase [ 28 ], a CoA-linked acetaldehyde dehydrogenase and iron-dependent alcohol dehydrogenase ( ADH1 ), a ferredoxin and the hydrogenase ( HYD2 ) gene itself [ 26 , 28 ]. In this work we examine the pattern of expression of selected fermentative gene in synchronous cells in a photoperiod and in response to continuous light/dark. Also, since the oxygen levels in the media of liquid cultures of Chlamydomonas depend on light conditions, we examine whether or not the observed patterns are determined by changes in oxygen levels. We uncovered the existence of three different gene expression profiles within the fermentative pathways, which might be an indication of a differential adaptive response of this green algae to different environmental inputs or changing metabolic factors which can coexist with low oxygen stress.",
"discussion": "4. Discussion Traditionally, fermentation is considered to substitute Krebs cycle and the electron transport elements of respiration in cases when oxygen is not available in order to regenerate NAD+ to allow glycolysis continue and produce at least some ATP. For organisms such as Chlamydomonas , which possesses a variety of fermentative pathways, a question arises as to which of these pathways will activate in the absence of oxygen, or perhaps will they all? Four of the genes ( LDH1 , ADH2 , PDC3 , and PFR1 ) studied in this paper fluctuated on a daily basis ( Figure 1 ), supporting results obtained in previous studies which take into examination the expression patterns of others fermentative genes [ 26 , 28 ]. However, LDH1 , ADH2 , and PDC3 differentiated themselves from PFR1 in that the first three show a day-time peak whereas the latter shows a night time peak suggesting that at least two different regulation profiles coexist in two different parts of the fermentative metabolism. PFL1 , in contrast to the aforementioned genes, did not show any regular daily fluctuations ( Figure 1 ), suggesting the existence of a third regulation profile within the fermentative pathways, or more likely, the insensitivity of PFL1 to the other two regulatory pathways. The different profiles of gene regulation identified in our experiments are represented in Figure 4 . In this paper we investigated some possible causes of these fluctuations, and in particular whether or not changes in light (dark) or low oxygen may influence them. Variations in light and oxygen availability are two regularly changing factors that occur over the day-night period in a natural context. The results suggest the existence of a double input in the regulation of LDH1, ADH2, PDC3 , and PFR1 : circadian cycle regulation which can be interrupted in continuous light, perhaps due to a loss of cell synchrony [ 26 ] as green algae have been reported to lose synchrony within 24 hours following exposure to continuous light [ 30 , 31 ]. The cell cycle has been shown to be under circadian regulation [ 32 ] though cell division is also gated by metabolic criteria (minimal volume and energy content), factors which in turn are influenced by light through photosynthesis [ 30 , 31 ]. The observed circadian fluctuations in fermentative gene expression might therefore be gated by other factors, among which continuous light due to its disruptive effect on the cell cycle. Interrupted circadian fluctuation of starch accumulation has been reported in Chlamydomonas in response to nutrient starvation, a condition which slows or stops the cell cycle [ 27 ]. In photosynthetic organisms there is a tight relationship between light and oxygen availability. We showed that ADH2 expression is not influenced by light (dark) or oxygen status over a 12-hour period (which is too short a period for cell synchrony to be lost) and that for LDH1 and PDC3 , only low oxygen tends to slightly reduce (or perhaps delay) the normal midafternoon peek ( Figure 3 ). Thus, variations in oxygen availability and light (dark), within the range of values used in our experiments (which simulate conditions that might occur in a natural context) do not determine the presence of the fluctuations we observed in the expression of these genes. \n PFR1 expression, contrasting with what was observed for LDH1, ADH2 , and PDC3 is completely eliminated in continuous light. Also, our experiments showed that PFR1 expression is upregulated by low oxygen, irrespective of light conditions ( Figure 3 ). When confronting PFR1 expression with that of HYD1 , the main gene responsible for hydrogen production [ 2 ], we found that they showed a similar response to combinations of light/dark and low/high oxygen ( Figure 3 ), further supporting the hypothesis the hydrogen producing branch of the fermentative pathway, at least at the mRNA level, is regulated differently than the branches that lead to ethanol and lactate. It is tempting to speculate that PFR1 expression is under exclusive regulation by low oxygen, and that the absence of expression under light is an indirect consequence of oxygen production by photosynthesis. This would allow a coordinated expression of both PFR1 and HYD1 to produce hydrogen exclusively under anoxia. \n PFL1 did not show regular fluctuations in any of the experimental conditions used in this paper suggesting the existence of a third expression profile within the fermentative metabolism. From our results, PFL1 does not appear to be modulated in response to stresses or environmental inputs at the RNA level. However the presence of a functional PFL1 protein is essential to the normal functioning of the other pathways, as its absence changes the fermentative product ratio and the change is not a simple increase of products produced by enzymes which compete for the same substrate [ 24 ]. PFL1 catalyzes a reaction that produces formate as one of its products. Formate is not a neutral metabolite, it has been suggested to play a role in gene regulation, possibly repressing the hydrogen metabolism [ 24 ]. The involvement of formate in the hydrogen metabolism has been demonstrated in Escherichia coli [ 33 ]. In Chlamydomonas formate is known to influence phothosyntesis by inhibiting electron and proton transfers in photosystem II [ 34 ]. We found that PFL1 is not influenced by light or oxygen leaving open the possibility of a feedback regulation by the product of its reaction, in support of what was suggested by Philipps et al. [ 24 ]. We could hypothesize that this feedback regulation may act to prevent the production of a metabolite which can downregulate of photosynthesis. Whether variations in RNA prove to be correlated with metabolic outcome or not, the existence of differential regulation in different branches of the path is indicative that different mechanisms are likely at work to determine different types of fermentation. Factors determining switches between fermentative pathways will be of particular interest in engineering Chlamydomonas for hydrogen production. Different branches of the pathways leading to different products may determine changes in pH (lactic acid, acetic acid), cause (or limit) accumulation of toxic metabolites (ethanol), or either the reoxidation of NADH over the synthesis of ATP (or vice versa), according to current physiological requirements [ 21 , 23 ]. In the natural environment, in addition to changes in light availability and the possible occurrence of low oxygen conditions, organisms may be simultaneously subjected to other stimuli which, above a certain level of intensity or length of time they may be perceived as stresses. Temperature typically varies on a daily basis and provides an input for setting the circadian clock [ 35 ]. Internal factors such as cell cycle status, and the circadian “time,” also vary at different times of day. We believe that the precise physiological response to a change in a specific environmental factor will vary according to the status of the other inputs. Therefore, under given circumstances, a particular type of fermentation maybe suitable, but in other conditions, a different fermentative path may be preferable. The possession of numerous fermentative options certainly could be viewed as advantages for a water dwelling yet aerobic organism so easily subjected to low oxygen on a regular basis. Uncovering the precise role of each of these paths will represent an interesting theme for future research."
} | 3,988 |
33806123 | PMC8037422 | pmc | 4,358 | {
"abstract": "Recent research works on intelligent traffic signal control (TSC) have been mainly focused on leveraging deep reinforcement learning (DRL) due to its proven capability and performance. DRL-based traffic signal control frameworks belong to either discrete or continuous controls. In discrete control, the DRL agent selects the appropriate traffic light phase from a finite set of phases. Whereas in continuous control approach, the agent decides the appropriate duration for each signal phase within a predetermined sequence of phases. Among the existing works, there are no prior approaches that propose a flexible framework combining both discrete and continuous DRL approaches in controlling traffic signal. Thus, our ultimate objective in this paper is to propose an approach capable of deciding simultaneously the proper phase and its associated duration. Our contribution resides in adapting a hybrid Deep Reinforcement Learning that considers at the same time discrete and continuous decisions. Precisely, we customize a Parameterized Deep Q-Networks (P-DQN) architecture that permits a hierarchical decision-making process that primarily decides the traffic light next phases and secondly specifies its the associated timing. The evaluation results of our approach using Simulation of Urban MObility (SUMO) shows its out-performance over the benchmarks. The proposed framework is able to reduce the average queue length of vehicles and the average travel time by 22.20% and 5.78%, respectively, over the alternative DRL-based TSC systems.",
"conclusion": "6. Conclusions and Future Work In this work, we have addressed the traffic signal control dual problem involving next phase determination and its duration. We aim to solve such a problem by leveraging the state of the art of a hybrid reinforcement learning variant. Specifically, we tailor the hybrid parameterized Deep Q-Networks, namely, Multi-Pass DQN, to dually control the TSC phase and its associated timing jointly. We conducted a simulation that allowed a series of controlled experiments for evaluating and demonstrating our framework performance. Moreover and for the sake of validity, we compared our framework to Deep RL benchmarks during training and taking decision at the intersection. The evaluation of the performance of our approach made use of the average travel time and the vehicle queue length as practical metrics. The results proved that our hybrid DRL variant outperformed the baselines in all the simulated experiments. A significant reduction of the average queue length of vehicles and the average travel time by 22.20% and 5.78%, respectively. The potential advantage of our framework is its hybrid nature, which allowed the TSC to control the phase selection as well as its duration. Our future works are twofold. Indeed, we would like to extend the scope of our hybrid DRL in order to cover more than one intersection in different ways, e.g., centralized and decentralized. In the second extension, we will direct our further simulations and experiments using real data from real world traffic intersections.",
"introduction": "1. Introduction Traffic congestion is one of the biggest issues in most of today’s cities causing significant delays and subsequent economic losses [ 1 ]. To tackle this issue, several research efforts in the transportation field attempted to develop intelligent transportation systems (ITS) aiming to overcome traffic congestion and improve traffic flow. Traffic signal control systems (TSCs) are one of the key research areas of intelligent transportation systems (ITS) made to control the traffic flow at intersections aiming to reduce traffic congestion [ 2 ]. Recently, various research works have leveraged reinforcement learning (RL) to replace the traditional traffic signal control systems [ 3 , 4 , 5 ]. In contrast with the standard traffic control approaches, RL and Deep RL (DRL) techniques can adapt to diverse traffic situations and conditions. In its recent application to TSC, DRL showed a higher performance over traditional traffic light management techniques [ 6 , 7 ]. In DRL-based traffic light controllers, the objective of the DRL agent is to decide the optimal action which yields improving the TSCs performance. Commonly, the action selection process is based on two strategies. In the first strategy, the DRL agent selects any phase from a finite set of phases without being limited to a predefined sequence of phases [ 8 ]. This strategy makes use of the discrete DRL architectures such as Deep Q-Nnetworks (DQN) [ 9 ], Double-DQN [ 10 ] and Dueling-DQN [ 11 ]. However, this strategy lacks the ability to predict the duration of the selected signal phase restricting it from choosing more optimal behavior. Whereas, in the second strategy, the agent’s actions are continuous instead, where the agent decides the duration of the next phase within a predefined cycle of traffic light phases [ 6 ]. The latter strategy belongs to the continuous type of DRL algorithms like Deep Deterministic Policy Gradient DDPG [ 12 ] and Normalized Advantage Function (NAF) [ 13 ]. Unfortunately, these two paths for controlling traffic signals lack flexibility and have not yet used jointly discrete and continuous DRL. Therefore, our ultimate objective, in this paper, is to bridge this gap and propose an approach that takes the potential advantage of combining the two strategies of applying DRL. Our approach is aimed to optimize traffic signal control by deciding simultaneously the proper phase and its associated duration. Hence, we propose a DRL based not only on employing discrete or continuous action spaces exclusively but combines them at the same time. Precisely, being inspired by DRL with parameterized actions [ 14 ], our contribution resides in tailoring a Parameterized Deep Q-Networks (P-DQN) architecture [ 15 ] that permits a hierarchical decision-making process that primarily decides the traffic light next phases and secondly specifies its associated timing. This design variant of DRL makes use of a hybrid architecture that combines discrete actions with continuous parameters. Subsequently, the learning agent within the DRL structure chooses at each decision step both the appropriate action and the parameter value associated with that action. The proposed framework is evaluated by establishing an experimental study that is conducted on the commonly used traffic Simulation of Urban MObility (SUMO) environment. The performance of the proposal as well as the benchmarks are assessed according to the common metrics used for TSC approaches evaluation such as the average travel time, the queue length and the average waiting time of vehicles [ 8 , 16 ]. Remarkably, the evaluation results of our proposed approach show considerable improvements of the TSC performance when compared to the benchmarks. The rest of this paper is organized as follows. in Section 2 , we review the works proposing DRL based solutions for TSC. Section 3 provides preliminaries and theoretical backgrounds needed by hybrid DRL-based TSC solutions. Our approach, as well as the proposed methodology behind it, are described in Section 4 . In Section 5 , we detail the experimental evaluation of our proposal and discuss the obtained results. Finally, in Section 6 , we draw the conclusion and present the potential future works.",
"discussion": "5.5. Results and Discussion We train the agent on the simulation setup using the training parameters discussed earlier. The resulting smoothed training curves of the proposed framework are illustrated in the Figure 6 . It can be noticed from the learning curves that the training undergoes what is known as a “cold start” [ 34 ] problem at early stages due to the exploration of the unfamiliar environment where the agent randomly applies decision actions. The agent subsequently optimizes its performance after grasping enough experience batches. Figure 7 shows the learning performance comparison against the Discrete and Continuous baselines. Remarkably, the Discrete approach exhibits fast initial learning but plateaus at lower performance than the Hybrid framework. It initially learns faster due to the fact that it already has a fixed phase timing and needs only to select the more suitable phase. The Continuous approach curve swings until it reaches a better performance but still worse than the rest. On the other hand, the Hybrid approach curve exhibits a linear-like decaying until it crosses the baselines’ curves where it outperforms the benchmarks’ performance. In Table 3 , we observe the average travel time scores of the Fixed-Time, Discrete and Continuous benchmarks versus the proposed framework with C1-C6 are the simulation configurations listed in Table 1 . Notably, the Fixed Time approach is far behind the other approaches due to its static behavior as opposed to the dynamic characteristics of the traffic flow. On the other side, one can remark that the deep reinforcement learning frameworks show noticeable results as they are more capable of dealing with dynamic conditions. Out of the DRL approaches, the proposed Hybrid actions framework outperforms the rest of the approaches in all simulated experiments. This is due to the fact that the Hybrid framework controls the TSC more flexibly by selecting the appropriate phase as well as its duration simultaneously. Further evaluations are shown in Figure 8 where we compare the queue length performance of the three deep RL approaches during one simulation episode. Similar to travel time performance results, the performance of the proposed approach surpasses the baselines by keeping the queue length lower throughout the traffic simulation."
} | 2,415 |
34669481 | PMC8528419 | pmc | 4,360 | {
"abstract": "A mesoscale plasticity rule improves AI learning.",
"introduction": "INTRODUCTION Activity-dependent synaptic modification is essential for learning in natural and artificial neural networks (ANNs). In ANNs, the idea that synaptic weights could be tuned to achieve the correct output has led to the development of a computational algorithm for supervised learning known as backpropagation (BP) ( 1 ). In BP, errors in the output of ANNs with respect to the differences between output values and expected values are used to adjust synaptic weights of upstream synapses layer by layer until the output meets the expectation. BP has been widely used in various types of ANNs, including convolutional and recurrent neural networks ( 2 ). There is now increasing interest in brain-inspired machine learning algorithms ( 3 – 6 ) that incorporate distinct brain features and could achieve problem solving with high efficiency and low computational cost. Spiking neural networks (SNNs) are considered to be the third-generation ANNs ( 7 ), in which information is conveyed by discrete events, called spikes or action potentials. While SNNs could be more potent in processing temporal information ( 8 ), there remains difficulty in developing learning algorithms with an efficiency comparable to that of BP. Methods commonly used to train SNNs to perform machine learning tasks can be roughly classified into five categories: recursive least square (RLS)–based, gradient-based, reward-based, conversion-based, and plasticity-based. First-order reduced and controlled error (FORCE) is a special RLS-based method, which has been well applied on tasks of sequence storage and replay in a recurrent SNN ( 9 ). Gradient-based methods are borrowed from BP, including spatial and temporal types, to train SNNs in a supervised manner. Spatial types such as surrogate gradient ( 10 ) and approximate gradient [pseudo-BP (p-BP)] ( 11 ) are used to circumvent the nondifferential nature of spikes ( 7 ), and temporal types like BP through time (BPTT) ( 12 ) and SpikeProp ( 13 ) are designed for gradient learning recursively and temporally (with spike latency). The reward-based methods, including eligibility propagation ( 14 ) and reward propagation ( 15 ), are more efficient on sequential machine learning tasks. Conversion-based methods that directly convert learned ANNs to SNNs are simpler but still limited by BP on both biological interpretability and plausibility ( 16 ). Substantial efforts have also been made in applying biologically plausible plasticity rules found in natural neural circuits into SNNs, including Hebb’s rule ( 17 , 18 ), short-term plasticity (STP) ( 19 , 20 ), long-term potentiation (LTP) ( 21 , 22 ), long-term depression (LTD) ( 23 ), and spike timing–dependent plasticity (STDP) ( 24 , 25 ). All these are local plasticity rules involving activity-dependent modification of synapses that induce the postsynaptic activity. Here, we focused more on plasticity-based methods and showed that introducing a novel form of nonlocal synaptic modification, termed “self-BP” (SBP) of synaptic potentiation and depression ( 26 – 30 ), helped to achieve a coordinated global propagation of synaptic modification, resulting in increased accuracy of SNNs and reduced computational cost of ANNs in performing supervised learning. The phenomenon of SBP, first discovered in cultures of hippocampal neurons ( 26 , 30 ), involves cross-layer BP of LTP and LTD from output synapses to input synapses of a neuron ( Fig. 1A ). Although other forms of nonlocal spreads of LTP and LTD in the pre- and postsynaptic neurons have been observed in natural networks ( 26 , 27 , 30 ), we here confined our study on SBP, because its existence was confirmed in developing retinotectal circuits in vivo ( 28 , 29 ). The phenomenon of SBP represents a form of nonlocal activity–dependent synaptic plasticity that may endow developing neural circuits the capacity to modify the weights of input synapses on a neuron in accordance with the status of its output synapses ( 27 ). In training SNNs, potentiation and depression of synapses on output neurons could result from STDP due to relative spike timing between network-generated and targeted spike trains at the output neuron, thus representing positive and negative weight modifications based on supervised learning. Further SBP of potentiation and depression signals to upstream synapses provided an efficient approach for backpropagating the “correct” and “error” signals, respectively. In training ANNs, traditional BP was used for synaptic weight adjustment at the output layer based on error signals. However, we found that replacing BP with SBP for weight adjustment of upstream synapses during a fraction of the learning period also produced beneficial effects. Fig. 1. Introducing biological SBP into SNNs. ( A ) Schematic diagram depicting the BP of potentiation (“+”) or depression (“−”) from the synapses at the output layer to those at the hidden layer in a three-layer network. The propagated synaptic modification had the same sign, consistent with the biological discovery of SBP ( 27 ). Similar configurations of fixed gradient mapping between neighborhood layers exist in artificial feedback alignment ( 48 ) and direct target propagation ( 49 ). ( B ) For a three-layer SNN, the induction of potentiation (+) or depression (−) occurred at the synapse W j , k on the output neuron by STDP, based on the timing of presynaptic spikes (in the hidden neuron) relative to the postsynaptic spikes in the output neuron, after updating by mean square error (MSE) of network-generated (Out 2 ) and teaching spike trains (Teaching 2 ). W i , j and W j , k represent synaptic weights of connections onto hidden and output neurons, respectively. The + and − signals created at synapses of hidden layer neuron (pink) onto an output neuron (blue) were allowed to spread to a percentage factor λ p (λ p ∈ [10%,100%]) of input synapses with a fraction factor λ f (λ f ∈ [0.1,1]) of the signals generated by the STDP (orange). V j and V k are membrane potentials at hidden and output layers, respectively. ( C ) The three-layer architecture of SNN, in which SBP and local plasticity (STP, STDP, and homeostatic V adjustment) were introduced at synapses at hidden and output layers, and the teaching spike train was given to the output LIF neurons. The diagram illustrates an output neuron inducing STDP (blue), a hidden neuron with the output synapse inducing STDP (pink), and input neurons with synapses receiving SBP (yellow).\n\nIntroducing SBP into three-layer SNNs and ANNs For training SNNs, we used a three-layer SNN ( Fig. 1C ). In the first (input) layer, neurons received spike trains as inputs encoded by comparing raw signals from datasets with a train of generated random numbers (see Materials and Methods for more details). The second (hidden) layer consisted of both excitatory and inhibitory leaky integrate-and-fire (LIF) neurons that exhibited the refractory period, nonlinear integration, and nondifferentiable membrane potential. The third (output) layer consisted of excitatory LIF neurons that received spiking signals from hidden layer neurons, and the supervised teaching signals were presented only in training procedures. The learning process used both local form of synaptic modification, i.e., STP ( 19 , 20 , 31 ) and STDP ( 32 , 33 ), and nonlocal SBP via sequential steps. First, feedforward processing of spiking signals was performed without introducing synaptic plasticity. Second, we introduced STP and homeostatic adjustments of membrane potential (homeo- V ) in hidden layer neurons to stabilize the spiking capability of the network (see Materials and Methods for details). Third, potentiation (“+”) or depression (“−”) of synaptic weights ( W j , k ) was produced by the STDP rule at all synapses made by hidden neurons onto output neurons ( Fig. 1B ). Last, potentiation or depression of latter synapses was allowed to spread retrogradely by SBP to synapses made by input neurons on hidden neurons. For introducing some specificity in the amount of synaptic modification via SBP, we set a percentage factor (λ p ∈ [10%,100%]) and a fraction factor (λ f ∈ [0.1,1]), allowing SBP to cover only a percentage of upstream neurons (see Materials and Methods for details) and a fraction of synaptic modifications to undergo SBP, respectively. The restricted Boltzmann machine (RBM) network ( 34 ) was used to examine the effect of introducing SBP into ANNs. The RBM contained three layers, artificial neurons with rectified linear unit (ReLU) activation functions, and fully connected feedforward connections ( Fig. 2A ). The learning procedure consisted of two phases: the unsupervised “sleep” phase and the supervised “wake” phase. During the sleep phase, only the energy function (see Materials and Methods for details) was used for calculating neuronal states toward the minimal energy ( Fig. 2C ). The wake phase included two types (I and II), interleaved by the sleep phase. In wake phase I, synaptic weights at the output layer ( W j , k ) were updated according to standard BP algorithm with both energy and cost functions (BP 1 ; see Materials and Methods for details), resulting in potentiation (+) or depression (−), and this was followed by the SBP of + or − ( Fig. 2D ). In wake phase II, BP was performed in a conventional manner, with BP 2 (at the hidden layer) following BP 1 ( Fig. 2E ). The learning procedure is summarized in the schematic diagram in Fig. 2B . The amount of self-propagated modifications was set as a fraction of ∆ W j , k , represented by the parameter θ sbp in E RBM and also the parameter λ f (corresponding to that in SNNs), and the percentage of upstream neurons receiving SBP was described by λ p as that in SNNs. Note that the replacement of BP 2 by SBP during wake phase I could notably reduce the computational cost normally required to perform the differential calculation of BP 2 during the learning process. Fig. 2. Introducing biological SBP into ANNs. ( A ) Schematic diagram depicting the architecture of the shallow ANN, represented by a three-layer restricted Boltzmann machine (RBM), with full connections between neurons in neighborhood layers. Neurons in hidden and output layers were artificial rate neurons with ReLU activation functions. Two network state indicators were used: the unsupervised energy function ( E RBM ; see Materials and Methods for details) describing the inner network state and the supervised cost function ( C RBM ; see Materials and Methods for details) describing network output state. ( B ) Schematic diagram depicting the learning process of RBM using SBP, in which wake phase I using BP (BP 1 ) and SBP and wake phase II using only BP (BP 1 + BP 2 ) interleaved by the sleep phase. ( C ) Unsupervised sleep phase, in which both W i , j and W j , k were tuned toward minimal energy function E RBM . ( D ) Wake phase I using both BP and SBP. The BP 1 produced potentiation (∆ W j , k > 0, +) or depression (∆ W j , k < 0, −) of W j , k between a hidden neuron (pink) and output neuron (blue), determined by differentiating the sum of C RBM and E RBM . The SBP induced + and − of W i , j based on ∆ W i , j , with a percentage factor λ p and a fraction factor λ f as described in Fig. 1B . ( E ) Wake phase II using only BP containing both BP 1 and BP 2 . W i , j and W j , k were updated on the basis of the minimization of both cost and energy functions with the chain rule of calculus. In this study, we examined the effects of introducing SBP on the accuracy and computational cost of SNNs and ANNs for three different learning tasks involving different extents of temporal information: (i) recognition of handwritten digits, using Modified National Institute of Standards and Technology (MNIST) dataset (fig. S1A) ( 35 ); (ii) phonetic transcription, using NETtalk dataset (fig. S1B) ( 36 ); and (iii) gesture recognition, using event-based dynamic vision sensor gesture (DvsGesture) dataset (fig. S1C) ( 37 ). The computational cost of networks during learning was defined by the product of the mean training epoch to achieve some defined accuracy levels ( Fig. 3A ) and algorithmic complexity per epoch ( Fig. 3B ). Our results demonstrated that introducing SBP into SNNs resulted in higher accuracies and lower computational costs in learning all three tasks. Furthermore, the combined use of SBP and BP in ANNs also resulted in similar benefits in all three tasks. These results underscored the usefulness of introducing a novel nonlocal plasticity rule found in natural neural networks into SNNs and ANNs. Fig. 3. The computational cost during learning. ( A ) Diagram depicting calculation of the mean epoch in N training epochs ( N = 5) for curves of f 1 ( x ) and f 2 ( x ) to achieve some defined error rate levels between an upper bound and a lower bound. The upper bound and lower bound represent the lowest and highest values of the error rate curves at the beginning and the end of learning epochs, respectively, among the algorithms under comparison (see Materials and Methods for more details). The computational cost was calculated by averaging the cost at five error rate levels (including upper and lower bounds). ( B ) Algorithmic complexity O ( ∙ ) in each epoch during learning. It includes feedforward propagation (FF) and feedback propagation (FB). m , n , and k are numbers of neurons in network’s input, hidden, and output layers, respectively. The compared algorithms include BP, STDP (or Hebb), direct target propagation (TP) ( 49 ), reward propagation (RP) ( 39 ), BPTT ( 12 ), and SBP.",
"discussion": "DISCUSSION In this work, we have introduced SBP of synaptic modification into SNNs and ANNs and examined its benefit for learning three benchmark tasks. For simplicity, we used three-layer feedforward networks comprising a variable number of neurons in the input, hidden, and output layers, depending on the task. The learning of the SNN consisted of two independent phases: first, the unsupervised learning phase of homeostatic adjustment of the membrane potential that maintained the firing capacity of the SNN and STP, which was found to be helpful in elevating the network efficiency (fig. S2, D to I), and second, the supervised learning phase that used STDP to initiate the correct and error signals in the form of potentiation and depression, respectively, and SBP for cross-layer synaptic weight adjustments. These SBP signals were generated by algebraic summation of synaptic changes based on the relative timing of all pairs of pre- and postsynaptic spikes using the standard pairwise STDP rule ( 42 ). Although not introduced in this study, additional constraints imposed by other STDP rules for natural spike trains in pre- and postsynaptic neurons ( 43 , 44 ) may further improve the network capability. Furthermore, other forms of nonlocal spread of synaptic modifications besides SBP, such as presynaptic lateral spread of LTP/LTD to synapses made by axon collaterals of the same presynaptic neurons ( 26 , 30 ) and to other converging inputs on the postsynaptic neuron ( 28 , 29 ), could be further explored for their potential benefits for SNNs. For ANNs, we have examined the benefit of introducing SBP into the training of RBM, using its special feature of separating the training into supervised and nonsupervised phases. We have also examined multilayer spiking-MLP models and found similar benefits (fig. S4D). In supervised wake phase of RBM, the standard synaptic weight update was mostly based on the BP of error signals toward the minimization of the global loss function. Adding SBP would disturb the supervised tuning of the direction of the BP-induced gradient. A similar situation was found when Hebb’s rule was added directly into BP ( 45 ). Perturbation of BP-induced synaptic weight updated by SBP could help drive the network modification toward an alternative direction, where the RBM may attain a higher accuracy with lower computational cost. Only simple three-layer ANNs were used in the present study for all benchmark tests. Our studies on the SNN with four to six layers, using SBP in all hidden layers, showed that the benefit of introducing SBP was greatly degraded to a level below that achieved by the three-layer SNN. Training of RBM with four to six layers, with the SBP replacing BP in all hidden layers during wake phase I training, yielded no improvement in accuracy beyond that achieved by the three-layer RBM, despite higher computational costs (fig. S4, A and B). The degradation of accuracy in SNNs with more than three layers may be attributed to excessive spread of potentiation or depression signals when SBP was allowed to occur beyond the neuron that generates the original synaptic modification. In addition, the failed learning of SNNs using SBP for higher layers might also be caused by the nonconvergence problem of synaptic modifications. The previous work has shown that the recurrent SNN contains exploding gradients ( 16 ). The SNNs using SBP also show a nonconvergence learning problem, especially for deeper ones (fig. S4A), where the synaptic modifications between input and hidden layers are dominated by the STDP in hidden and output layers, and the influence of SBP from the induction layer to backpropagated layers is progressively weaker. This hypothesis was further verified in fig. S4C, where the distribution of synaptic modifications in three (or four) layers was properly norm-distributed, while that in five (or six) layers is left the same as that in initialization. Biologically experimental results of SBP in a network containing hippocampal neurons are consistent with this phenomenon, where the SBP also fails to propagate beyond one layer to more upstream neurons ( 30 ). The biological interpretation of this failure is that the potentiation/depression at input synapses due to SBP is based on cellular mechanisms distinctly different from those underlying LTP/LTD at the output synapses, thus incapable of generating further SBP in more upstream neurons. Notably, in some regions of the nervous system, such as retina, hippocampus, or neocortex, information processing could largely be characterized as a three-layer network operation within the region. Our finding that three-layer ANNs appear to be the optimal network to implement SBP suggests that ANNs may benefit from the use of three-layer networks as relatively independent basic modules, and more sophisticated ANNs could be built via parallel and serial connections among them. In considering the efficiency of ANNs in performing standard benchmark tasks, previous studies using a variety of ANNs have largely focused on the accuracy in recognizing the test samples after network training. In this study, we have examined both the accuracy and the computational cost in learning tasks. Notably, the reduction of the computational cost represents the major benefit conferred by introducing SBP in both SNNs and ANNs. In estimating the computational cost, we used the product of the mean training epoch to achieve some defined accuracy levels ( Fig. 3A ) and algorithmic complexity per epoch ( Fig. 3B ) as an indicator. Other aspects of the cost, including the number of arithmetic operations and the number of bits required to specified synaptic weights and neuronal states within each iteration, were not included here but could be further considered in the future work. In addition, we compared the computational cost during training in reaching the same given accuracy levels, rather than those for attaining the final converged accuracy by each operation. In most operations, the small increment in the final accuracy often requires disproportional large amount of computation. For efficient performance of the network, relatively high rather than the highest accuracy could be sufficient. The notion of balanced computational cost and accuracy is in line with the efficient information processing of the brain, where the rapidity in computation (with low energy cost) is as relevant as the accuracy. Last, we note that the original experiment demonstrating SBP in cultured networks of hippocampal neurons ( 26 ) was inspired by the power of BP algorithm, although it seems to be biologically implausible ( 46 ). The SBP-associated information flow occurs in the neuronal cytoplasm, via retrograde fast axonal transport of molecular signals ( 28 , 29 ). The finding of SBP in natural networks has shown that an effective machine learning algorithm for ANNs can spur neuroscience discovery, and the present study further demonstrates that introducing algorithm-inspired biological discovery back to ANNs further elevates their efficiency. Such two-way interactions between neuroscience and artificial intelligence have much in store for the future."
} | 5,233 |
29844421 | PMC5974407 | pmc | 4,361 | {
"abstract": "Dense crossbar arrays of non-volatile memory (NVM) can potentially enable massively parallel and highly energy-efficient neuromorphic computing systems. The key requirements for the NVM elements are continuous (analog-like) conductance tuning capability and switching symmetry with acceptable noise levels. However, most NVM devices show non-linear and asymmetric switching behaviors. Such non-linear behaviors render separation of signal and noise extremely difficult with conventional characterization techniques. In this study, we establish a practical methodology based on Gaussian process regression to address this issue. The methodology is agnostic to switching mechanisms and applicable to various NVM devices. We show tradeoff between switching symmetry and signal-to-noise ratio for HfO 2 -based resistive random access memory. Then, we characterize 1000 phase-change memory devices based on Ge 2 Sb 2 Te 5 and separate total variability into device-to-device variability and inherent randomness from individual devices. These results highlight the usefulness of our methodology to realize ideal NVM devices for neuromorphic computing.",
"introduction": "Introduction Over several decades, the von Neumann architecture has enabled exponential improvements in system performance. However, as device scaling has slowed and demand to handle big data has soared, the time and energy spent transporting data across the physically separated memory and processing units have started to limit the performance and power efficiency. As potential alternatives, neuro-inspired non-von Neumann computing paradigms have become promising candidates to perform real-world tasks 1 , 2 . One avenue of research is referred to as in-memory computing or computational memory, which exploits the physical properties of non-volatile memory (NVM) devices for both storing and processing information 3 – 6 . Recently, a large-scale experimental demonstration of this concept using an array of one million phase-change memory (PCM) devices has been reported 7 . Another paradigm is hardware acceleration of deep neural network (DNN) 8 – 12 training via the use of dense crossbar arrays of NVM to perform locally analog computation at the location of the data. As shown in Fig. 1 , it is possible to use NVM devices with variable conductance states, such as resistive random access memory (ReRAM) 13 and PCM 14 to represent the synaptic weights and to perform vector-matrix multiplication using the basic electrical principles, i.e., Ohm’s and Kirchhoff’s laws, thus enabling local and parallel computation on a large scale. By making the conductance change of the NVM element bidirectional, backpropagation algorithm can be implemented. Such a crossbar array of NVMs is expected to achieve significant acceleration factors of DNN training and remarkable reduction in power and area 15 , 16 . Another active area of research is spiking neural networks (SNNs) motivated by the need to build more biologically realistic neural network models. Several neuromorphic computing platforms are being developed which are optimized for emulating spike-based computation. These SNNs are typically trained using certain local update rules, such as the spike-timing-dependent plasticity. NVM devices have recently found applications as both synaptic and neuronal elements of such SNNs 17 – 20 . Fig. 1 Neuromorphic computing system based on NVM. a Schematic illustration of one-layer neural network with synaptic weights ( W ) connecting an input layer to an output layer. b A synaptic weight is represented by a conductance value of an NVM element at each cross-point in a crossbar array structure. c Vector-matrix multiplication is performed by sensing the current ( I ) for each column, which is the product of the synaptic weight ( G ) and the input signal ( V ) The key technical challenge for these applications is to realize ideal NVM elements with continuous (analog-like) conductance tuning capability in response to electrical pulses with acceptable noise levels. For acceleration of DNN training, symmetric conductance change with positive and negative pulse amplitudes is another key requirement 15 , 16 . The device conductance should go up with a voltage pulse of one polarity and should go down by the same magnitude with a voltage pulse of the opposite polarity. In general, NVM elements do not show this symmetric switching behavior. Therefore, a differential approach is often used in which two conductance values are compared in a unit cell 14 . In this configuration, linearity in switching is required to ensure a symmetric differential signal. In reality, most NVM elements exhibit highly non-linear evolution of conductance as a function of the number of consecutively applied pulses. This results in significant errors in weight updates 13 . In addition, such non-linear conductance change makes separation of signal and noise extremely difficult. Most NVM elements show stochasticity related to the physical origins of switching. When incremental weight updates are performed for analog NVM devices, the magnitude of conductance change approaches the level of inherent randomness 21 , manifesting as significant noise components. Therefore, establishing a universally applicable methodology to evaluate signal-to-noise ratio (SNR) of non-linear and analog NVM devices is of paramount importance for neuromorphic computing applications. In this study, we first establish a practical methodology based on a machine learning algorithm to precisely separate signal and noise components from an analog NVM device with non-linear conductance changes. The methodology is agnostic to the device physics, enabling us to apply it to different types of NVM elements. First, the methodology is applied to HfO 2 -based ReRAM to understand the relationship between switching symmetry and SNR. Next, the methodology is applied to PCM devices based on doped-Ge 2 Sb 2 Te 5 (GST). We characterize 1000 devices and separate device-to-device variability and inherent randomness from individual devices.",
"discussion": "Discussion We established a practical methodology based on GPR to precisely separate signal and noise components from analog NVM elements with non-linear conductance changes. This solves key technical challenges for characterization of artificial synapses of neuromorphic computing system, namely extraction of switching symmetry and SNR. The methodology is agnostic to switching mechanisms and therefore applicable to various types of NVMs. We applied the methodology to HfO 2 -based ReRAM devices and found the tradeoff between switching symmetry and SNR. Using SF as a guideline, substantial improvement in switching symmetry was achieved compared to reported ReRAM devices in literature. By systematic analysis of 1000 GST-based PCM devices, we clearly demonstrated that a large portion of variability in weight update is attributable to inherent randomness from individual devices and this is the key component to be suppressed in order to achieve high classification accuracy. Finally, the proposed methodology helps neuromorphic system engineers in two ways depending on phases of technology development. In an exploratory phase, our methodology enables extraction of switching symmetry and SNR from individual devices and expedites search for ideal materials. The conventional methodology requires fabrication of many devices with tight device-to-device variability for extraction of SNR, which is difficult to attain in the early stage when exotic material options need to be screened. In a relatively mature technology phase, our methodology helps find the optimum input signals (e.g., duration and amplitude of pulses) that provide the best switching symmetry (linearity) and SNR within the tradeoff for the entire neuromorphic system."
} | 1,954 |
37447692 | PMC10346551 | pmc | 4,362 | {
"abstract": "Over the last couple of decades, numerous piezoelectric footwear energy harvesters (PFEHs) have been reported in the literature. This paper reviews the principles, methods, and applications of PFEH technologies. First, the popular piezoelectric materials used and their properties for PEEHs are summarized. Then, the force interaction with the ground and dynamic energy distribution on the footprint as well as accelerations are analyzed and summarized to provide the baseline, constraints, potential, and limitations for PFEH design. Furthermore, the energy flow from human walking to the usable energy by the PFEHs and the methods to improve the energy conversion efficiency are presented. The energy flow is divided into four processing steps: (i) how to capture mechanical energy into a deformed footwear, (ii) how to transfer the elastic energy from a deformed shoes into piezoelectric material, (iii) how to convert elastic deformation energy of piezoelectric materials to electrical energy in the piezoelectric structure, and (iv) how to deliver the generated electric energy in piezoelectric structure to external resistive loads or electrical circuits. Moreover, the major PFEH structures and working mechanisms on how the PFEHs capture mechanical energy and convert to electrical energy from human walking are summarized. Those piezoelectric structures for capturing mechanical energy from human walking are also reviewed and classified into four categories: flat plate, curved, cantilever, and flextensional structures. The fundamentals of piezoelectric energy harvesters, the configurations and mechanisms of the PFEHs, as well as the generated power, etc., are discussed and compared. The advantages and disadvantages of typical PFEHs are addressed. The power outputs of PFEHs vary in ranging from nanowatts to tens of milliwatts. Finally, applications and future perspectives are summarized and discussed.",
"introduction": "1. Introduction Energy harvesting is defined as the conversion of the ambient energies present in the environment in various forms into usable electrical energy for powering electronic devices, sensors, and circuits [ 1 ]. This technology has been developed rapidly in recent years, driven by the fact that the burning of fossil fuels releases a large amount of carbon dioxide and greenhouse gases into the air, leading to climate changes and global warming [ 2 ]. Another driving force is the local power sources for wearable sensors, portable electronics, health monitoring systems, and wireless devices. As the clean energy revolution is taking place, including solar, wind, water, geothermal, bioenergy, and nuclear energy [ 3 , 4 ], research on new energy sources accelerates. Among these new energy sources, mechanical motion/vibration is one of the most investigated types due to its abundance, accessibility, and ubiquity in the environment [ 5 , 6 ]. Mechanical energy, including kinetic energy and potential energy, could be obtained from industrial machinery, automotive, human motion, large-scale buildings, and ocean waves [ 7 ]. Energy harvesters are considered to be promising distributed power sources for low-power portable electronics and wearable sensors [ 8 , 9 ]. Unlike conventional chemical batteries, which present issues relating to limited lifespan, environmental pollution, and recharging [ 10 ], energy harvesting is largely maintenance free and environmentally friendly [ 11 ]. Human mechanical energy and environmental mechanical energy are intensively exploited due to their abundance in daily life [ 12 ]. For example, the mechanical energy from human walking and running can be collected by energy harvesters assembled in shoes. The most common methods for mechanical-to-electric conversion mechanisms are piezoelectric [ 13 , 14 , 15 , 16 , 17 , 18 , 19 ], electromagnetic [ 20 , 21 , 22 , 23 , 24 ], triboelectric [ 25 , 26 , 27 , 28 , 29 ], and their hybrid derivatives [ 30 , 31 , 32 , 33 ], each with its own advantages and disadvantages, as illustrated in Table 1 . In terms of efficiency, piezoelectric energy harvesting generally achieves good conversion efficiency in small volume space compared to electromagnetic systems, making it more suitable for low-power and low-profile applications where both energy and space are crucial [ 34 , 35 , 36 ]. Electromagnetic energy harvesting, while offering high power output, tends to require a large volume space [ 37 ]. Triboelectric energy harvesting can be scaled up or down, but its power output is generally low because of significantly high internal electrical impedance [ 12 , 38 , 39 ]. The choice between these approaches is application dependent, but the piezoelectric mechanism has been investigated predominantly, owing to the merits of its high energy density, high capacitance, low mechanical damping, easy shaping, and implementation [ 40 ]. Piezoelectric materials can generate electricity because the central symmetry of the crystal structure is broken under the action of the external force, forming a piezoelectric potential [ 7 ]. Among existing piezoelectric materials, lead zirconate titanate (PZT) and polyvinylidene fluoride (PVDF) are two of the most popular and cost-effective materials for energy harvesters mounted in shoes. Compared to PZT ceramics, PVDF has considerable flexibility, good stability, and is easy to handle and shape [ 41 ]. But PZT has the advantages of high mechanical-electric coupling factors, producing larger power, and easier integration with force amplification frames [ 40 , 42 ]. This paper gives a comprehensive review of the technology developments and research trends of piezoelectric footwear energy harvesters (PFEHs). The paper is organized as follows. The background and motivations for PFEHs research and developments are introduced in Section 1 . The fundamentals of piezoelectric properties for footwear energy harvesters are briefly presented in Section 2 . The force and dynamic energy distribution on the footprint, which includes the foot pressure, ground force reactions, and displacement, as well as acceleration during human walking, are reviewed in Section 3 . Section 4 discusses the energy flow from human walking to the harvested energy through the PFEHs and the methods to improve the energy conversion efficiency. Following on the fundamental knowledge learned from Section 2 , Section 3 and Section 4 , the major PFEH structures and mechanisms on how the PFEHs capture mechanical energy from human walking to piezoelectric structures are summarized in Section 5 . The main structures are classified into four types, including flat plate, curved, cantilever, and flextensional harvesters. The current applications and future perspectives of the PFEHs are presented in Section 6 . Finally, Section 7 briefly summarizes this work."
} | 1,705 |
37579148 | PMC10450648 | pmc | 4,363 | {
"abstract": "Significance Ectomycorrhizal fungi are important and ubiquitous symbionts of most temperate and boreal trees that form interaction networks with multiple hosts of the same or different species. Using DNA metabarcoding of root-associated ectomycorrhizal fungal communities, we show that soil moisture and tree host carbon status serve as critical ecological determinants of well-connected interaction networks between trees and their symbiotic fungi. Experimental warming and rainfall reduction led to shifts in taxonomic and functional composition of ectomycorrhizal fungal communities that corresponded with significant reductions in interaction network connectivity. We conclude that climate change scenarios that reduce soil moisture and decrease tree host performance have high potential to disrupt the interaction networks formed by trees and fungi in high-latitude forests.",
"conclusion": "Conclusions In this study, we have documented important changes in both ectomycorrhizal fungal community composition and interaction network structure associated with cooccurring temperate and boreal tree hosts in response to one of the most realistic tests of potential changes in temperature and rainfall. In particular, the climate change–induced effects on ectomycorrhizal fungal composition, which favored ectomycorrhizal fungal taxa with limited mycelial growth and shorter turnover times, likely disrupts the formation and fungal of shared mycelial networks among diverse forest trees. We hypothesize that this disruption will exacerbate negative effects on tree host performance and distribution, particularly those adapted to cooler and wetter climates (as documented at this site, 24). Since this study was positioned near the latitudinal boundaries of the tree hosts, however, it may not necessarily reflect their response at mid-range latitudes and future effort should be placed on further characterizing responses of ectomycorrhizal interaction networks to climate change in a variety of forest systems. Given the emerging recognition that plant-fungal networks are integral properties of terrestrial ecosystems ( 97 – 99 ), it is imperative that future research aims to understand how resilient they are to changing climate as well as their influence on ecosystem processes.",
"discussion": "Discussion Despite the widely recognized importance of ectomycorrhizal symbioses in forest ecosystems, how climate change will alter ectomycorrhizal fungal communities and their traits remains poorly understood, with the response and resilience to climate change of ectomycorrhizal interaction networks having gone largely unexamined to date. Our findings suggest that the effects of warming and reduced rainfall not only will lead to significant changes in ectomycorrhizal fungal community structure, but also will contribute to a disruption in the structure of ectomycorrhizal interaction networks at the boreal–temperate ecotone. Specifically, both warming and reduced rainfall significantly decreased the level of network generality inferred from the bipartite network analyses, resulting in less redundancy in the networks. This lowered redundancy was strongly correlated with reductions in soil moisture caused by the warming and reduced rainfall treatments and was most pronounced in the combined treatment ( Fig. 4 A ). Since adequate water is important for the growth and survival of ectomycorrhizal fungi ( 48 ), we suspect water limitation in particular may directly contribute to shifts in ectomycorrhizal community composition and the concordant changes in interaction network structure. Additionally, reduced soil moisture reduced host productivity for three of the four hosts, which could have led to indirect negative effects on ectomycorrhizal fungal communities and interaction network structure through reduced C allocation ( 25 , 49 ). Other studies have shown that ectomycorrhizal fungal biomass production is dependent on the amount of host C allocation ( 50 ) and that when water is limiting and photosynthesis is correspondingly reduced, EMM biomass production also declines ( 51 , 52 ). It appears that shifts in ectomycorrhizal fungal community composition were likely a primary factor undergirding the observed changes in interaction network structure in the warming and reduced rainfall treatments. Under ambient conditions, the ectomycorrhizal fungal communities of all the hosts studied were dominated by Tomentella, Russula , and Sebacina species. These genera are among the most common in forests globally ( 53 ) and share a suite of traits that may contribute to their dominance and the presence of highly redundant bipartite interaction networks. First, they have wide host ranges and colonize tree host species primarily via EMM and not from spores ( 54 , 55 ). The host generality of these taxa is a prerequisite to the formation of mycelial networks between different tree individuals ( 41 , 56 ) and their high abundance in ectomycorrhizal fungal communities suggests there is a high potential for the formation of mycelial connections between different tree individuals of the same or different species in the forests where they occur. Second, while there is variation in exploration type morphology within and among these genera, the species belonging to each have hydrophilic EMM that can be relatively extensive [i.e., medium distance exploration types; ( 57 )]. The extent of the EMM is an important feature not only for exploring the soil for nutrients, but presumably for colonizing fine roots and forming mycelial connections among different hosts ( 17 , 55 , 58 ). Since the hyphae of these taxa are hydrophilic, they may be more susceptible to water loss under reduced osmotic potential ( 45 ). This may explain their negative response to the warmed and reduced rainfall treatments, which had significantly reduced soil moisture. Tradeoffs between traits related to stress tolerance (e.g., moisture niche) and competitive ability (e.g., growth rate) have been documented among free-living fungi, where community dominants tend to have narrower niche breadth ( 59 ). While the application of this relationship to symbiotic fungi is complicated by dependence on host C provisions, it is reasonable to expect similar tradeoffs among ectomycorrhizal fungi. Finally, there is evidence suggesting that these dominant ectomycorrhizal genera are important in accessing organic N ( 60 – 62 ), facilitating soil N retention ( 63 , 64 ), and augmenting host nutrition ( 65 , 66 ). Together, the dominance of these taxa and the low level of interaction network specialization found in ambient plots indicates that these taxa are likely key components of shared mycelial networks in these high latitude forests. Under warming and reduced rainfall, the ectomycorrhizal fungal communities shifted in dominance toward members of the Ascomycota as well as specific basidiomycete genera such as Inocybe, Thelephora, Hebeloma, Laccaria, and Clavulina ( Fig. 1 ). Apart from Thelephora , all these EM fungi possess contact-short distance exploration strategies and their increased relative abundances with warming are consistent with findings of previous studies conducted at the B4WarmED sites ( 4 , 25 , 26 ) as well as other climate change experiments ( 67 – 70 ). Many of these taxa also possess traits consistent with a life history strategy that allocates resources to maximize reproduction and dispersal at the cost of an extensive EMM and colonizing roots almost exclusively from spores ( 54 , 71 , 72 ). Further, these fungi tend to be found in low abundance in the active EMF communities (i.e., colonized roots) of mature forests, but have resistant spores that lead to their accumulation and domination of spore banks ( 73 ). The colonization of roots from spores often results in numerous small genets that turn over quickly ( 12 , 74 ). Further, their ability to access and mobilize nutrients to support their hosts from organic pools is limited ( 54 , 75 – 78 ). While many are generalists in terms of host associations, their spatially limited and short-lived EMM may limit their capacity to connect tree hosts via shared mycelial networks in both space and over time ( 74 , 79 ). Ascomycete ectomycorrhizal fungi are common in both disturbed and early successional soils ( 80 – 83 ). They often possess traits likely favored in these habitats, including low-biomass short-distance exploration strategies ( 57 ), production of resistant propagules that dominate spore banks ( 84 ), frequent asexual spore production ( 85 ), and small genets that frequently turnover ( 86 ). However, some ectomycorrhizal ascomycetes are notably stress tolerant, with Cenococcum geophilum being perhaps the most notable ( 87 – 90 ), and may provide benefits to host nutrition during periods of water stress ( 65 , 91 – 93 ). At the same time, variation in tolerance to water stress and nutritional benefits to hosts under water stress has been found even among ectomycorrhizal fungal taxa in the same genus, making generalizations at higher taxonomic groupings challenging ( 65 , 94 ). Further, there may be limits to host benefits provided by ectomycorrhizal fungi if water stress becomes too extreme. For instance, in a warming and rainfall reduction experiment conducted in a semi-arid shrubland system, host nutrition declined significantly compared to the control despite the dominance of ascomycete ectomycorrhizal fungi ( 95 ). In our system, and in other temperate and boreal forest ecosystems, it remains unclear if ascomycete ectomycorrhizal fungi are involved in maintaining host nutrition during periods of water stress, colonizing host roots from resistant spore banks when other members fall out of the community, or are less carbon costly to stressed host plants. Both laboratory- and field-based experiments parsing these possibilities will be a key next step in better linking ectomycorrhizal fungal community structure to functioning under shifting environmental conditions. Interestingly, in addition to the increase in contact-short distance exploration types with warming and reduced rainfall, we did also observe a notable increase in Rhizopogon spp. and Suillus americanus ( SI Appendix , Fig. S7 ). These Suilloid taxa produce rhizomorphs, which are hydrophobic long-distance exploration structures that likely require significant host C investment to sustain ( 44 ). Since Suillus and Rhizopogon almost exclusively associate with Pinaceae hosts, the increase in their abundance under altered environmental conditions contributed to the increased network-level specialization metrics here. Their increased abundance, however, is surprising because Suilloid taxa are generally thought of as C-costly taxa, and we found that both pine hosts exhibited significant reductions in photosynthetic capacity with warming and reduced rainfall. Indeed, Fernandez et al. ( 25 ) showed that ectomycorrhizal fungal taxa that produce longer distance exploration strategies were positively correlated with photosynthetic capacity in Abies balsamea and Betula papyrifera in adjacent closed canopy B4WarmED plots. In this study, we observed a similar response except for Pinus -specific Suilloid taxa. Our Suilloid results are, however, consistent with a greenhouse study that grew Pinus pinaster seedlings under varying water availability and found that increasing drought conditions favored certain Suillus and Rhizopogon species. It is possible that due to their hydrophobic rhizomorphs, these taxa may facilitate greater water transport under reduced soil moisture ( 96 ). Alternatively, Bruns et al. ( 73 ) proposed that Suilloid ectomycorrhizal fungi may have adapted the ability to extract more C from their hosts compared to generalist taxa. In a scenario where competition with typically dominant ectomycorrhizal fungi is reduced by increased climate change–related stress, Suilloid ectomycorrhizal fungi may be able to extract relatively high amounts of C without necessarily increasing nutrient returns, effectively becoming parasitic under these conditions. While our study contributes to a mounting body of literature suggesting that mycorrhizal fungi are sensitive to changes in climate, we acknowledge several limitations that should be considered. First, the presence of mycelial connections between different tree individuals by the same fungal individual were not physically assessed, instead they were inferred based on sequence read abundance from metabarcoding of fine roots. This approach, while effective identifying shared taxa, cannot detect physical connections between plants via ectomycorrhizal fungi. As such, this study should be considered a first look at the potential response of ectomycorrhizal mycelial network structure to climate change, with future experiments aimed at validating the physical network responses using approaches that integrate metabarcoding, isotopic labeling, and visualization techniques (i.e., minirhizotron imaging, mesocosms) or quantify taxon-specific EMM (i.e., ingrowth bags) ( 17 ). Second, while we attempted to capture host representatives from both temperate and boreal latitudinal ranges as well as major phylogenies (broadleaf versus conifer), there were other ectomycorrhizal host species in the plots contributing to ectomycorrhizal fungal community composition that were not sampled. This could have led to sampling bias if those other tree host ectomycorrhizal fungal communities did not respond in a similar manner. That said, based on previous work from the B4WarmED experiment both above- ( 24 , 46 ) and below-ground ( 4 , 25 , 26 ), we expect similar negative responses in terms of host photosynthetic performance and changes in ectomycorrhizal fungal community composition. Third, while our results contribute to a mounting body of work suggesting that changes to host productivity under warming and drought stress are critical drivers of changes in ectomycorrhizal fungal communities, our analyses did not include root responses (i.e., production; specific root length), which may be equally important. Future studies should include root measures to explicitly test above- and below-ground host responses and linkages to ectomycorrhizal communities and interaction networks. Finally, because the tree hosts in the experiment are not mature trees and have relatively small root systems, it is plausible that they may be more susceptible to water stress than adults, and result in stronger ectomycorrhizal responses to the climate change treatments."
} | 3,642 |
24818897 | null | s2 | 4,364 | {
"abstract": "Swimming motility is a flagellum-dependent form of movement observed in the Gram-negative bacterium Pseudomonas aeruginosa. Swimming motility is defined as the movement in liquid or low-viscosity conditions (up to 0.3 % agar concentration). Unlike swarming motility, swimming motility requires a functional flagellum, but neither quorum sensing (QS) systems nor biosurfactants. While swimming motility can also be observed via microscopy, here we describe a reproducible plate-based method."
} | 122 |
36043065 | PMC9364158 | pmc | 4,367 | {
"abstract": "Electroactive aniline tetramer–spider silk composite fibers with high conductivity and mechanical strength were developed using a dip coating method. The fabricated spider silk composite fibers retain the high mechanical strength (0.92 GPa) and unique reversible relaxation–contraction behavior of spider dragline silks. The aniline tetramer modified on the silk surface imparted electroactive properties to the composite fibers. The color of aniline tetramer/spider silk composite fibers could be controlled by applying different pH values and voltages. Furthermore, the composite fiber's resistivity could reach 186 Ω m which can conduct electrical current to light LEDs. This study could provide a valuable guideline for developing highly-conductive electrochromic spider silks for use in E-textiles.",
"conclusion": "4. Conclusions We have shown the success of developing electroactive spider silk fibers with outstanding mechanical properties. Aniline tetramer was synthesized and subsequently incorporated to the surface of spider dragline silk fibers by using dip-coating technique. The aniline tetramer can impart electroactivity to the spider dragline silks and thus could reversibly switch their colors by applying electrical voltage. In addition to the electroactive/electrochromic behaviors, the developed composite silk fibers also can conduct electrical current to light the LED. Furthermore, the aniline tetramer can also enable the silk fibers to have proton doping/un-doping capacity so that they become pH responsive color changing fibers. For the mechanical properties, the resulting composite fibers have extensibility close to 25%, and breaking strength of 0.9 GPa, which cannot rivaled by most man-made fibers. Importantly, the spider dragline silk's unique relaxation–contraction response to humidity did not be affected by the aniline tetramer modification process. They could cyclically shrink and relax at low and high humidity surroundings, respectively. Thus, the aniline tetramer/spider dragline silk composite fibers developed in this study could provide a new avenue to fabricate multi-functional fibers with outstanding mechanical properties for applications in E-textiles.",
"introduction": "1. Introduction Web-weaving spiders are well-known for their capability of producing silks with unique mechanical properties. A female orb weaving spider is able to produce up to seven different silks, including major ampullate silk, flagelliform silk, aggregate silk, minor ampupllate silk, pyriform silk, aciniform silk, and cylindriform silk, for specific purposes and they possess a range of mechanical properties. 1 The most investigated spider silk is dragline silk (major ampullate silk), the main structural web silk, which is also used for the spider's lifeline, because of its unique combination of strength, extensibility, and toughness, which cannot be rivaled by most man-made fibers. For example, dragline silk has extensibility of 35% and toughness of 100 kJ kg −1 compared to extensibility of 5% and toughness of 30 kJ kg −1 for Kevlar fibers. 2 The primary amino acid sequences of silk proteins combined with highly sophisticated spinning process, including acidification, ion exchange, dehydration, shearing force, and elongational flow, contribute to the outstanding mechanical properties of a dragline silk fiber. 1 Furthermore, spider dragline can shrink or relax in response to the surroundings' humidity, which known as supercontraction. 3 The spider dragline silk supercontracts up to 50% of its original length at high humidity and thus can build substantial stress to deliver work. The report in literature has shown that simply using wet/dry air can drive the spider dragline silk to produce work density up to 500 kJ m −3 , which is 50 times higher than most biological muscles. 4 The mechanism for supercontraction is not clear yet, but it could be contributed to that water molecules break the hydrogen bonds within amorphous domains of the silk proteins and enable the silks relax to their less ordered and lower energy state. 3,5 In addition, spider dragline silk can also show a humidity-induced torsional deformation, which is related to the supercontraction behavior of spider dragline silk. Studies demonstrated that the dragline silk can generate a large twist deformation of more than 300° mm −1 , and the torsional actuation could be tailored by tuning the surroundings' humidity. 6 In addition to its unique mechanical behaviors, spider silk is a biodegradable, biocompatible, and hypo-immunogenic material so that it has great potential for biomedical and tissue engineering applications. 7 For example, the spider dragline silk has been utilized as nerve guidance conduits to reconstruct peripheral nerve injuries and showed success in supporting nerve regeneration. 8 In addition, the surface of the spider dragline could be modified with HfO 2 /ZrO 2 nanoparticles or carbon nanotubes for potential bio-applications in the fields of biosensing/bioimaging or electrophysiological detection. 9,10 Recently, Koop et al. revealed that the dragline silk of spiders might elicit granulomatous foreign body reaction in rat's spinal cord, yet, more researches are to be performed to elucidate the controversy. 11 Furthermore, silk proteins are with highly organized secondary structures, such as β-sheets and α-helices, formed through intra- and intermolecular hydrogen bonding between their amino acid sequences. These structures are tightly-bound and behave elastic-like so that electrical dipoles could be generated reversibly, which enable silk to have piezoelectric effect to convert mechanical elastic strain energy into electrical energy. Researchers have utilized spider dragline silk to develop bio-nanogenerators for harvesting green energy, which exhibiting high energy conversion efficiency, high output voltage and current through simple mechanical or biomechanical activities. 12,13 The fabricated device is biocompatible and ultra-sensitive towards physiological signal monitoring such as arterial pulse response which can be useful for potential smart biomedical applications. 13 Reports in literature have shown that spider silk can have various advanced applications, including artificial muscles, electro-tendons, biosensors, torsional actuators, and nanogenerators, originating from the primary and secondary structures of their constituent proteins. 6,10,12–14 In this study, we further explored the feasibility of spider silks for applications in highly demanding wearable electronic textiles (E-textiles). 15 Spider silks have significant potential to become a sustainable alternative to fossil-fuels-based materials not only because their remarkable mechanical properties and biocompatibility, but also their sustainability and biodegradability to meet the requirement of circular economy and green chemistry for fabricating E-textiles. 16 Polyaniline has great potentials for applications in electrochromic systems, anticorrosion coatings, electronic sensors, and biomedical fields because of their high intrinsic electrical conductivity, biocompatibility, and controllable oxidation/protonation states. 17,18 Several studies have reported that polyaniline could be utilized to increase the electrical conductivity of the spider silk. 19,20 However, the poor solubility in common solvents makes polyaniline difficult to process. 21 In this study, we utilized the aniline tetramer, which is the smallest aniline oligomer that can have the same oxidation-reduction/doping-undoping structures as polyaniline, 22 to develop electroactive spider silks with appreciable electrical conductivity for potential applications in E-textiles. We demonstrated the feasibility, for the first time, that a simple dip coating method can modify the surface of spider dragline silk with aniline tetramer to impart the conductivity and electroactivity to the composite fibers. The developed spider silk composite fibers not only remain the high mechanical strength (0.92 GPa), but also the unique reversible relaxation–contraction behavior of spider dragline silks. Furthermore, the color of the composite fibers could be controlled by the applied voltage or adjusting their pH values. The electrical conductivity of the spider dragline silk could be improved with aniline tetramer and thus allow the composite fibers to conduct electrical current to light LEDs.",
"discussion": "3. Results and discussion 3.1. Characterization of synthesized aniline tetramer Aniline tetramer has three specific oxidation-reduction states and their related doping/undoping states: fully reduced state (leucoemeraldine salt, LES and leucoemeraldine base, LEB), the half oxidized state (emeraldine salt, ES and emeraldine base, EB) and the fully oxidized state (pernigraniline salt, PNS and pernigraniline base, PNB) ( Fig. 2(a) 25,26 ). Consequently, it possess unique electrical, chemical, and optical properties and has numerous advanced applications including bioseparating materials, electrochemical sensors, supercapacitors, and mediators for enhancing differentiation of neural stem cells or facilitating the photothermal effect for cancer therapy. 21,27–30 Fig. 2 (a) Molecular structures of aniline tetramer with different redox states, (b) cyclic voltammetry measurement for synthesized aniline tetramer. LES-leucoemeraldine salt; ES-emeraldine salt; PNS-pernigraniline salt. The as-prepared aniline tetramer in this study was in proton-doping state because of synthesizing in HCl solution. Subsequently, it was washed with ammonium hydroxide solution to become un-doping state for spider silk modification. The synthesized aniline tetramer was characterized by a mass spectrometer. The results shows [M + H] + ions at m / z = 365.2, which indicates that it is in its most stable EB state C 24 H 20 N 4 = 364.3 (Fig. S2 † ). The chemical structure synthesized aniline tetramer was analyzed by FTIR (Fig. S3 † ). The peak of 1504 cm −1 and 1594 cm −1 could be assigned to benzenoid and quinoid rings stretching vibrations, respectively. There three peaks at 1309, 1161 and 851 cm −1 are from the C–N stretching vibration of a secondary aromatic amine, the aromatic C–H in-plane deformation, and the C–H out-of-plane deformation of 1,4-aromatic substituted benzene rings, respectively. 31 UV-Vis spectroscopy was utilized to investigate the UV-visible absorption behavior of the synthesized aniline tetramer. The as-synthesized emeraldine base (EB) of aniline tetramer exhibited two distinct absorption peaks ( λ max ) at 319 nm and 588 nm, which could contribute to the π–π* transition in the benzenoid ring and exciton absorption in the quinoid rings, respectively (Fig. S4 † ). 32,33 The cyclic voltammetry measurement could confirm the electroactive behavior of aniline tetramer, which showing two oxidation peaks at 0.32 V and 0.54 V ( Fig. 2(b) ). The first oxidation peak at 0.32 V and the second oxidation peak at 0.54 V contribute to the transition from fully reduced LES state to half oxidized ES state and half oxidized ES state to fully oxidized PNS state, respectively. These results shows electroactive aniline tetramer have been successfully synthesized. 3.2. Electroactive spider silks Typical orb-weaver draglines are made of major ampullate spidroin protein 1 and 2 (MaSp1 and MaSp2), which containing mainly a high content of glycine (G) (34.7–42.2%), alanine (A) (17.6–27.5%), and proline (1.7–15.7%) amino acid residues, 34 and their poly(A) and poly(GA) domains forms crystalline β-sheets, leading to the dragline silk as a semi-crystalline polymer. 2 Accordingly, the dragline silk of Nephila pilipes utilized in study is in semi-transparent white color as shown in Fig. 3(a) . In addition, a water drop can sit on the pristine dragline silk bundle without penetration, and the water contact angle is larger than 90° ( Fig. 3(a) ), indicating the hydrophobic behavior. The fact that dragline silks repel the high surface energy water is due to the hydrophobic lipid coating on the silk and the hydrophobic air surroundings. 35–37 Fig. 3 (a) Image of virgin Nephila pilipes dragline silk. (b) Image of aniline tetramer-coated Nephila pilipes dragline silk. The synthesized aniline tetramer could coat on the surface of the dragline silk after dip-coating procedure as evidenced by the pronounced peaks of aniline tetramer appear in the composite silk's FTIR spectrum (Fig. S5 † ). In addition, the color of the silk can be utilized to justify if the aniline tetramer coated on the silk surface. Fig. 3(b) shows the optical image of the aniline tetramer/silk composite fibers with the blueish color, indicating that the dragline surface is covered with EB aniline tetramer, compared to pristine silk in semi-transparent white color. The composite fibers still behaves hydrophobic as the water contact angle is larger than 90°. The aniline tetramer could have different colors, which depends on its energy gap resulting from the insertion/de-insertion of dopant ions. 38 The prepared aniline tetramer for dip-coating procedure in this study was in emeraldine base (EB) state (deprotonated emeraldine) so that the color of the resulting composite silk is blueish and its color became greenish when protonating with HCl solution (protonated emeraldine, emeraldine salt ES state). 39 The aniline tetramer concentration for dip-coating procedure significantly affected its amount modified on the silk's surface as shown in Fig. 4 . Apparently, there are not much aniline tetramer attached on the silk surface as using 0.01% and 0.05% aniline tetramer solution for coating process since it seems that the silk still exhibits colorless as shown in Fig. 4(a) and (b) . Investigating their SEM images confirms that there are only few aniline tetramer particles deposited on the silk surface, as shown in Fig. 5(a) and (b) , similar to the smooth without any distinct feature of pristine silks ( Fig. 1(b) ). More aniline tetramer particles attached on the silk surface as dip-coating concentration further increased to 0.1% ( Fig. 5(c) ). The silk surface could be completely covered by the aniline tetramer as the dip-coating concentration reached 1% ( Fig. 5(d) ). However, higher solution concentration of 10% not only allows the individual silk fiber covered with aniline tetramer, but also caused two silk fibers to connect together ( Fig. 5(e) ). In addition, some silk fibers' surface appears smooth, indicating no aniline tetramer on the silk surface ( Fig. 5(f) ). This result is due to that there is no strong bonding between the aniline tetramer layer and the silk surface. Therefore, the thick aniline tetramer layer can easily detach from the silk surface under disturbance such as preparation for SEM investigation. In contrast, Fig. S6 † shows that the composite fibers prepared by 1% aniline tetramer did not have any significant coating delamination even after stirring in water at speed of 1000 rpm for 7 days. In addition, surface-treatment method also performed to figure out whether the coating layer delamination issue could be improved. It is generally accepted that there is a lipid coating on the surface of silk fibers 35–37 and it can be stripped by organic solvents. 36 Therefore, dragline silks were washed in hexane for ten minutes to remove the lipid layer before dip-coating with aniline tetramer. Fig. S7(a) † shows the surface of the dragline silk is smooth without any distinct feature after hexane treatment. In contrast, aniline tetramer particles attached on the silk surface can be observed after subsequent 0.1% dip-coating procedure (Fig. S7(c) † ), and the silk surface could be completely covered by the aniline tetramer as the dip-coating concentration reached 1% (Fig. S7(d) † ). Increasing dip-coating concentration to 10% can also obtain silks coated with aniline tetramer (Fig. S7(e) † ), but coating delamination phenomenon still appeared (Fig. S7(f) † ). These results suggest that using hexane to remove the lipid layer on the silk surface did not improve the poor adhesion between the aniline tetramer coating and the silk surface. In summary, higher dip-coating concentration led to more aniline tetramer coated on the silk surface. However, excessive aniline tetramer could have adverse effect on the coating layer as found in the result of 10% aniline tetramer dip-coated composite fibers. The amount of aniline tetramer on the silk surface was evaluated using UV-Vis analysis and the result was summarized in Table S1. † Fig. 4 Optical images of spider dragline silks modified with various concentration of aniline tetramer solution. (a) 0.01%; (b) 0.05%; (c) 0.1%; (d) 1%; (e) 10%. (Left: emeraldine salt state of aniline tetramer. Right: emeraldine base state of aniline tetramer). Fig. 5 SEM images of spider dragline silks modified with various concentration of aniline tetramer solution. (a) 0.01%; (b) 0.05%; (c) 0.1%; (d) 1%; (e) 10%; (f) 10%. EDX analysis was also utilized to obtain the composition and the quantity of elements in the composite silks. The results were summarized in Table S2, † indicating only elements of C, O, and N in virgin silks and the atomic percent of these elements do not change significantly after aniline tetramer modification process (ANOVA analysis p > 0.05). These results are due to similar atomic composition of aniline tetramer coating and spider silk fibers, as well as element quantity provided by the EDX analysis was not only from the aniline tetramer layer, but also from the dragline silk. Furthermore, aniline tetramer, which having proton doping/un-doping capacity, could enable the spider silk to become pH responsive color changing fibers. The color of aniline tetramer in EB deprotonated and ES protonated state is blueish and greenish, respectively. The proton doping/un-doping procedure could be done not only by immersing the composite fibers in acidic/basic solutions, but also by exposing them in acidic/basic vapors. Fig. 6 shows the as-prepared aniline tetramer/dragline silk composite fibers could reversibly change their colors from blueish to greenish and then back to blueish simply by exposing to formic acid (HCOOH) and ammonia (NH 3 ) vapors, respectively. Fig. 6 Optical images of as-prepared aniline tetramer/dragline silk composite fibers cyclically exposed to formic acid (HCOOH) and ammonia (NH 3 ) vapors. (Exposure time: HCOOH for 1 minute and NH 3 10 minutes, respectively) (dragline silk is modified with 1% of aniline tetramer solution). In addition, the aniline tetramer enables the spider silk to be electroactive, which confirmed by the cyclic voltammetry measurement as shown in Fig. 7 . The virgin dragline silk did not exhibit any redox transition as the testing potentials ranged from −0.2 to 1.0 V. In contrast, there are distinct redox peaks present in the cyclic voltammetry test for all silk modified with aniline tetramer. The first pair of redox peaks close to 0.34 V oxidation peak was due to the aniline tetramer's transition from its fully reduced LES state to half oxidized ES state. The second pair of redox peaks with the oxidation potential close to 0.64 V contributes to the transition from its half oxidized ES state to fully oxidized PNS state. The pure aniline tetramer only need 0.54 V to switch its state from ES to PNS. However, the aniline tetramer/dragline silks required higher potentials due to their lower electrical conductivity compared to pure aniline tetramer. Furthermore, the redox current increased as increasing the added amount of aniline tetramer. These results confirm that aniline tetramer can impart electroactivity to the spider dragline silks. Moreover, the aniline tetramer/dragline silks could reversibly switch their colors by applying electrical voltages, which resulting from the electrochromic property of aniline tetramer. Fig. 8 demonstrates that the composite silk fibers showed bluish and greenish as the applied voltage was 0.9 V and 0.1 V, which relating to the aniline tetramer at PNS and LES state, respectively. These electrochromic behaviors are similar to the related polyaniline studies reported in literature since the aniline tetramer is the smallest oligomer unit that can fully represent the structure of polyaniline. 40–44 Fig. 7 Cyclic voltammetry measurement for spider dragline silks modified with various concentration of aniline tetramer solution. (a) 0%; (b) 0.01%; (c) 0.05%; (d) 0.1%; (e) 1%; (f) 10%. Note: potential applied is based on Ag/AgCl reference electrode. Fig. 8 Effect of applied voltage on the color of aniline tetramer/silk fibers (top: 0.1 V; bottom: 0.9 V). (dragline silk is modified with 1% of aniline tetramer solution). Aniline tetramer in the emeraldine base (EB) state can be acid doped/protonated to form the conducting emeraldine salt (ES) state, 39 which can lower the electrical resistivity of the spider dragline silk. The electrical resistivity of the dragline silks modified with different amount of aniline tetramer is shown in Fig. 9 . The silk composites exhibited high electrical resistivity at low incorporated aniline tetramer amount. However, the electrical resistivity significantly lowered to 186 and 2 Ω m as utilizing 1% and 10% aniline tetramer solution for dip-coating procedure, respectively (statistical significance: **** P < 0.0001). The resistivity is low enough so that it can conduct electrical current to light up LEDs, as demonstrated in Fig. 9 . Fig. 9 Electrical resistivity for spider dragline silk modified with different aniline tetramer dip-coating concentration (the statistical significance of electrical resistivity among varied treatments was analyzed by ANOVA (**** P < 0.0001)). 3.3. Mechanical properties of electroactive spider silks The dragline silk of orb-weaving spiders is known for its unique combination of strength, and extensibility, which cannot rivaled by most man-made fibers. Generally, its breaking strength ranges from 0.8 to 1.5 GPa and extensibility close to 30%. 45–48 The dragline silks from Nephila pilipes utilized in this study have Young's modulus of 8.2 GPa, extensibility of 0.22, and breaking strength of 0.92 GPa ( Fig. 10 ). Spider silk can be considered as a semicrystalline polymer, which its β-sheet crystalline region provides the strength, while the amorphous regions contributes to the elasticity of the silk. 49 The X-ray diffraction results show that there are two distinct 2 θ peaks at 11.9° and 13.3°, which are the β-sheet crystalline d -spacings of 4.8 and 4.3 angstroms, 50 respectively, present for the virgin dragline silk. These two peaks do not shift or disappear after the aniline tetramer coating procedure, indicating the original β-sheet crystalline did not significantly be affected. Furthermore, the mechanical testing for the aniline tetramer-modified silks also reveal that their Young's modulus, extensibility, and breaking strength are similar to the un-modified one, as shown in Fig. 10 . These results indicate that aniline tetramer do not penetrate the silk's microstructure and thus alter the spatial arrangement of the chain (secondary structure) so that the mechanical properties do not significantly be affected by the aniline tetramer modification process. on breaking strength ( P > 0.05), extensibility ( P > 0.05), and Young's modulus ( P > 0.05). Fig. 10 (a) XRD results for dragline silk modified with different aniline tetramer dip-coating concentration. Breaking strength (b), extensibility (c), and Young's modulus (d) are mechanical properties for dragline silk modified with different aniline tetramer dip-coating concentration. (Statistical analysis of breaking strength ( P = 0.8942), extensibility ( P = 0.0579), and Young's modulus ( P = 0.9507) revealed no significant differences among all aniline tetramer treatments). In addition to the mechanical properties, the aniline tetramer coatings neither interfere with the dragline silk's reversible relaxation–contraction response to humidity. “It is well-known that there are many hydrogen bonds within the secondary structure of the fibroins in virgin dragline silk. Water can penetrate silk and thus break hydrogen bonds between fibroins, thereby allowing the fibroins to re-arrange to lower energy levels and result in silk's irreversible supercontraction. 51 Removing water from the post-supercontraction silk allows the reformation of hydrogen bonding, which stiffening and contracting the dragline silk. When wetting the silk again, water can break the reformed hydrogen bonds, thus relaxing the silk and lowering its stiffness. The post-supercontraction silk's relaxation–contraction response to wetting and drying is highly reversible and has led to the development of humidity-driven actuators. 6,52,53 Our prepared aniline tetramer/silk threads also exhibited such humidity-driven reversible phenomenon as shown in Fig. 11 . The plastic ring (load) hung on the middle of the prepared composite bundle was in the relaxation state as saturated with water (wet state). Subsequently, the water within the silk was removed (dry state), which causing the reformation of hydrogen bonding and thus silk contracted so that the hung load was elevated. Again, the hydrogen bonds could be broken by water (wet state), leading to relaxation of silk and thus the position of the load lowered. This contraction and relaxation behavior is highly reversible and did not be affected by the aniline tetramer on the surface of the silk. Fig. 11 Two cycles of humidity-driven, reversible contraction and relaxation for aniline tetramer spider silk bundles subjected to a load. The red solid line indicates the position of the load after the silk contraction or relaxation. (Dragline silk is modified with 1% of aniline tetramer solution)."
} | 6,455 |
33437563 | PMC7790359 | pmc | 4,368 | {
"abstract": "As the fossil fuel reserves are depleting rapidly, there is a need for alternate fuels to meet the day to day mounting energy demands. As fossil fuel started depleting, a quest for alternate forms of fuel was initiated and biofuel is one of its promising outcomes. First-generation biofuels are made from edible sources like vegetable oils, starch, and sugars. Second-generation biofuels (SGB) are derived from lignocellulosic crops and the third-generation involves algae for biofuel production. Technical challenges in the production of SGB are hampering its commercialization. Advanced molecular technologies like metagenomics can help in the discovery of novel lignocellulosic biomass-degrading enzymes for commercialization and industrial production of SGB. This review discusses the metagenomic outcomes to enlighten the importance of unexplored habitats for novel cellulolytic gene mining. It also emphasizes the potential of different metagenomic approaches to explore the uncultivable cellulose-degrading microbiome as well as cellulolytic enzymes associated with them. This review also includes effective pre-treatment technology and consolidated bioprocessing for efficient biofuel production.",
"conclusion": "Conclusion and future prospects Despite being rich in lignocellulosic biomass, India still lags in the discovery of industrially viable lignocellulolytic enzymes for effective production of second-generation biofuels (SGB) at an industrial scale. In the last few years, advanced sequencing strategies like next-generation sequencing (NGS) have come up with the capability of generating large amounts of sequence data, unlike conventional methods. With the help of NGS and metagenomics, a lot of unexplored, uncultivable, novel lignocellulases can be identified and exploited for the acceleration of SGB production in India. In this review, few metagenomic outcomes were discussed to enlighten the importance of many unexplored habitats for novel cellulolytic gene mining. This review also highlights the potential of different metagenomics approaches as most of the uncultivable cellulose-degrading microbiome and their efficient enzymes remain unexploited. It is observed that the degree of success depends on the methodology opted for, as every methodology has its drawbacks. In the future, these drawbacks can be overcome and even different strategies can be developed in combination with various methodologies described in this review.",
"introduction": "Introduction The cognizance of fossil fuel depletion started in the early 1970s. Engineers then suggested that the consumption of more fuel than extracted will ultimately lead to its exhaustion. To prevent this, the demand for these fuels should be moderated along with the quest for alternate forms of energy that are capable of replacing fossil fuels. For a decade, there has been a constant increase in primary energy consumption (PEC) around the world. The average PEC growth rate (PECGR) for the years 2008–2018 globally was 1.6%, and remarkably in 2019, it decreased to 1.3%. The reasons for this decline are many and one of the chief reasons is the feeble economic growth of nations like Russia, the USA, and India. The decline of PECGR was observed in many nations except China. China stands as the highest individual country contributing about 24.3% of global PECGR, followed by the USA, India, and Russia with 16.2%, 5.8%, and 5.1% respectively. Though the global PECGR in 2019 has decreased, the individual PECGR by countries like China and India has increased from 135.33 and 33.30 exajoules (EJ) respectively to 141.70 and 34.06 EJ when compared to 2018 [ 1 ]. To control the surge in energy consumption, emerging countries like India and China have started to use the available fuels more efficiently. The oil consumption of countries like China and India has increased by 5% and 2.9% respectively. On the one hand, oil production is going down drastically, and on the other hand, oil consumption has been increasing [ 1 – 3 ]. Due to the prevailing COVID-19 pandemic, global energy review suggests that during the lockdown phase economies over the globe can expect a 20 to 40 % decline in economic output. In India, the national lockdown has reduced the energy demand by 30% [ 4 ]. Once this pandemic phase is over, it is expected the global energy demand would eventually increase and the statistics of the past 10 years suggest the need for alternate fuel like biofuels. In India, there is a 9.8% increase in the production of energy from renewable sources when compared with the previous year 2018. Global biofuel production increased from 1787 thousand barrels of oil equivalents per day (tboe/d) to 1842 tboe/d which accounts for about a 3% increase in global biofuel production. India has increased its biofuel production by 24.9% when compared with 2018 and stands in the second position after Indonesia (37.5%) in the Asian-Pacific countries [ 1 – 3 ]. The demand for bioenergy is expected to increase by up to 11% by 2040 [ 5 ]. In India, the contribution of bioenergy to total energy demand is gradually increasing. This increase can be accelerated as India has abundant reserves of biomass, which can be the raw material for producing various forms of biofuel. The pursuit is for efficient technology that can convert the biomass into bioenergy. According to the United Nations Food and Agriculture Organization, in India, there is an increase in the total area covered by forests in recent times. This signifies that there is no insufficiency of lignocellulosic raw material in India [ 6 ]. Plant biomass is a reliable source for sugars and when subjected to fermentation will yield biofuel. So, lignocellulose is a very significant source for the production of various biofuels [ 7 ]. Lignocellulose is chiefly composed of cellulose, hemicellulose, and lignin. Cellulose contributes about 40–50%, hemicellulose contributes 25–30% of lignocellulose and the rest is lignin [ 8 ]. Consequently, keeping present inferior lignocellulose separating approaches [ 9 ] in view, humongous attention has been given to improve the methods of lignocellulose hydrolysis facilitated by the novel, efficient, and engineered enzymes [ 10 ]. Combinations of different glycosyl hydrolases are necessary for the comprehensive breakdown of lignocellulose into a blend of different sugars. In a lignocellulose-degrading habitat, the microbiome produces different mixtures of glycosyl hydrolases which aid in the thorough degradation of lignocellulose [ 11 , 12 ]. So the quest for novel and efficient approaches for biofuel applications is still going on and one such new approach is metagenomics. Metagenomics is a novel approach for studying genomes of the entire microbiome residing at a given habitat. It helps in understanding the microbial composition of the habitat and gives a way to explore and exploit many novel genes from uncultivable/cultivable microbiome [ 13 ]. The increasing demand for steadfast and efficient lignocellulases and hemicellulases targeting biofuels may be met by this novel approach of metagenomics. In this review, the emphasis is to analyze most of the reported metagenomic-derived cellulases (endoglucanases in specific). Lignocellulosic biomass which is available as agricultural, industrial, and municipal solid waste and forest residues around the globe is a prospective raw material for bioethanol as well as other value-added biochemical production [ 14 – 16 ]. Production of bioethanol from lignocellulosic biomass consists of three important steps. (i) The pre-treatment process reduces the recalcitrant nature of lignin, thus allowing enzymatic hydrolysis of biomass converting it to fermentable sugars. Pre-treatment techniques can be categorized into physical/mechanical, physicochemical, chemical, and biological. Each pre-treatment method has its advantages and disadvantages (Table 1 ). Based on the composition of biomass and economics, appropriate pre-treatment method is employed. (ii) Saccharification and fermentation, the leftover biomass (rich in cellulose) after pre-treatment will be converted to monomeric glucose by using cellulase enzymes. Bacteria, fungi, and actinomycetes are major cellulase-producing microorganisms at the laboratory scale, for industrial and bioenergy applications (Table 2 ). (iii) Recovery, to separate/extract the bioethanol produced from the raw fermentation broth to obtain high-purity bioethanol. Even though several separation methods are available, either distillation or in combination with other processes remains the primary approach for bioethanol purification. Table 1 Advantages and disadvantages of different pre-treatment methods of lignocellulosic biomass (source: [ 17 ]) Type of pre-treatment Advantage Disadvantage Physical/mechanical pre-treatment • Reduce the crystallinity and degree of polymerization of cellulose and increase the surface and porosity • High energy requirement Physicochemical pre-treatment • Increases the surface area • Lignin transformation and hemicellulose solubilization • Less inhibitory compound • Easy recovery • High energy, power, and pressure requirement Chemical pre-treatment • High recovery of sugars • Disruption and solubilization of lignin • Partial/complete removal of hemicellulose • Inhibitory compounds formation • Corrosive catalysts (acid pre-treatment) • Biomass become greasy (alkali pre-treatment) • Costly (ionic liquids) Biological pre-treatment • Less inhibitory compound • Delignification • Partial hemicellulose hydrolysis • environmental friendly (no chemical requirements) • Reduction in the degree of polymerization of cellulose • Process rate is slow • Low treatment rate • Commercially not viable Table 2 Cellulases available in the market (all the price details are available in the website links except PCT1518*, the price of it was obtained from a local vendor) S.No. Cat.No Commercial name of the enzyme Company name Microorganism Pack Substrate Approximate price Website link USD INR 1 C2730 Cellulase Sigma-Aldrich Trichoderma reesei 50 ml Cellulose 138.89 10,378.86 https://www.sigmaaldrich.com/catalog/product/sigma/c2730?lang=en®ion=IN 2 SAE0020 Cellulase, enzyme blend Sigma-Aldrich Unknown source 50 ml Cellulose 138.44 10,344.86 https://www.sigmaaldrich.com/catalog/search?term=SAE0020&interface=All&N=0&mode=match%20partialmax&lang=en®ion=IN&focus=product 3 V2010 Viscozyme L Sigma-Aldrich Aspergillus sp. 50 ml Cellulose 140.23 10,478.33 https://www.sigmaaldrich.com/catalog/product/sigma/v2010?lang=en®ion=IN&cm_sp=Insite-_-rvRecBlock_recentlyViewed_userHistory-_-recentlyViewed5-2 4 C8546 Cellulase Sigma-Aldrich Trichoderma reesei ATCC 26922 10 KU Cellulose 208.11 15,550.88 https://www.sigmaaldrich.com/catalog/product/sigma/c8546?lang=en®ion=IN 5 C1794 Cellulase Sigma-Aldrich Trichoderma sp. 10 KU Cellulose 208.56 15,584.6 https://www.sigmaaldrich.com/catalog/product/sigma/c1794?lang=en®ion=IN 6 PCT1518* Cellulase Himedia Unknown source 100 KU Cellulose 541.77 41,674.5 http://www.himedialabs.com/intl/en/products/Plant-Tissue-Culture/Plant-Culture-Tested-Chemicals-Enzymes/Cellulase-Plant-Culture-Tested-PCT1518 7 C0615 Cellulase (≥ 5000 units/g solid) Sigma-Aldrich Trichoderma sp. 1 g Cellulose 262.33 19,602.54 https://www.sigmaaldrich.com/catalog/product/sigma/c0615?lang=en®ion=IN 8 65480 Cellulase “Onozuka” RS (16,000 U/g) SRL Chemicals Trichoderma viride 1 g Cellulose 139.36 10,414.00 http://www.srlchem.com/products/product_details/productId/143755/Cellulase%2D%2DOnozuka%2D%2DRS-ex%2D%2DTrichoderma-Viride%2D%2D16000U-g 9 32970 Cellulase “Onozuka” FA (2500 U/g) SRL Chemicals Trichoderma viride 1 g Cellulose 70.43 5263.00 http://www.srlchem.com/products/product_details/productId/143639/Cellulase%2D%2DOnozuka%2D%2DFA-ex%2D%2DTrichoderma-Viride%2D%2D2500U-g 10 24801 Cellulase “Onozuka” R-10 (10,000 U/g) SRL Chemicals Trichoderma viride 1 g Cellulose 73.66 5504.00 http://www.srlchem.com/products/product_details/productId/4231/Cellulase-Onozuka-R-10-ex%2D%2DTrichoderma-Viride%2D%2D10000U-g 11 95382 Cellulase ex. (Meicellase), (13,000 CMC U/g) SRL Chemicals Aspergillus niger 1 g Cellulose 92.43 6907.00 http://www.srlchem.com/products/product_details/productId/949/Cellulase-ex%2D%2DAspergillus-Niger%2D%2DMeicellase%2D%2D-13000CMC-U-g 12 SKU 08320961 CELLULASE Y-C MP Biomedicals Trichoderma viride 10 g Cellulose 869.05 64,939.5 https://www.mpbio.com/08320961-cellulase-y-c-from-trichoderma-viride 13 C0057 Cellulase TCI Chemicals Aspergillus niger 25 g Cellulose 184.74 13,804.64 https://www.tcichemicals.com/AU/en/p/C0057 14 22178 Cellulase (~ 0.8 U/mg) Sigma-Aldrich Aspergillus niger 100 g Cellulose 332 24,808.59 https://www.sigmaaldrich.com/catalog/product/sigma/22178?lang=en®ion=IN 15 SKU 0215058380 Cellulase MP Biomedicals Unknown source 100 g Cellulose 397.7 29,718.02 https://www.mpbio.com/us/cellulase-60-000-units-g 16 SKU 0215058380 Cellulase MP Biomedicals Unknown source 100 g Cellulose 558.85 41760 https://www.mpbio.com/in/cellulase-60-000-units-g In the orthodox method of producing bio-alcohol, processes like saccharification as well as fermentation are performed as a distinct individual process, involving their respective optimum parameters. This process is referred to as separate hydrolysis and fermentation (SHF) [ 18 ]. The chief limitation of SHF is the cellulase enzyme’s feedback inhibition, implicated by sugars liberated by the hydrolysis of the substrate [ 19 – 21 ]. To overcome this issue, simultaneous saccharification and fermentation (SSF) was recommended, which enhanced the enzyme consumption and efficiency of the process [ 22 – 25 ]. The major drawback of this process is incompatible temperatures of hydrolysis (45–60 °C) and fermentation (30 °C) [ 22 – 24 , 26 ]. To alleviate this issue, non-isothermal simultaneous saccharification and fermentation (NSSF) has been proposed, involving partial enzymatic hydrolysis at optimum temperature, and as soon as the culture media is inoculated, the optimum temperature for the microbial growth is set [ 20 , 27 ]. In the process of simultaneous saccharification, filtration, and fermentation (SSFF), membranes are used to obtain a clear sugar-rich filtrate from the hydrolysis liquid. The filtrate contains hydrolyzed sugars along with partially hydrolyzed lignocellulosic biomass. After glucose, xylose is the next abundant saccharide in many lignocellulosic materials. It would be apt to use the simultaneous saccharification and co-fermentation (SSCF) process to efficiently use the xylose part of the filtrate. In this process, xylose and glucose utilizing wild-type or engineered microorganisms are employed for ethanol production [ 28 – 31 ]. However, it is essential to gaze for new alternatives to the SSCF process, and one such alternative is consolidated bioprocessing (CBP) (Fig. 1 ). Fig. 1 Various strategies of bioprocessing for converting lignocellulosic biomass to biofuel. (CBP means consolidated bioprocessing; SSCF means simultaneous saccharification and co-fermentation; SSF means simultaneous saccharification and fermentation; SHCF means separate hydrolysis and co-fermentation; SHF means separate hydrolysis and fermentation) Consolidated bioprocessing (CBP) has been designed to evade the setbacks and expenses of orthodox biofuel production from lignocellulosic biomass. This involves the application of either pure culture or consortia depending on the output and the process parameters. CBP aims to associate the processes like production of enzyme, hydrolysis, and fermentation into a single step and also try to combine the pentose sugar utilization process into the same. This is expected to improve the efficacy of the processes by eliminating the dependency on various hydrolytic enzymes that are being supplemented exogenously and decreasing the cellulase feedback inhibition by sugars [ 32 , 33 ]. This leads to the reduction of unit operations involved in the total process, thus decreasing the process inclusive capital costs [ 32 ]. Further advances in CBP can evade the pre-treatment process of biomass by producing the biofuel from raw biomass [ 18 ]."
} | 4,039 |
39807652 | PMC11780727 | pmc | 4,369 | {
"abstract": "Simultaneously hydrophilic and oleophobic surfaces offer\nsubstantial\nadvantages for applications such as antifogging, self-cleaning, and\noil–water separation. It remains challenging to engineer such\nsurfaces without requiring polar functional groups. This study introduces\nHFIL, a novel ionic liquid (IL) coating that achieves simultaneous\nhydrophilic and oleophobic properties via a one-step dip-coating process\nwithout relying on polar functional groups. Key findings show that,\ndespite the bulk form of HFIL having a high hexadecane contact angle\n(HCA) of 74.1° and an even higher water contact angle (WCA) of\n87.6°, the IL forms a stable monolayer on high-energy surfaces\nexhibiting a much lower WCA of approximately 40° with minimal\nchange to the HCA. Washing tests demonstrate that, even without the\npolar functional groups, there is a non-zero bonded thickness upon\nwhich the oleophobicity is comparable to polytetrafluorethylene (PTFE).\nThese properties highlight HFIL’s potential for durable applications\nin antifouling, antifogging, and environmental separation technologies,\nwhere selective liquid interactions are essential. This work contributes\nto a broader understanding of IL-based surface modifications, advancing\nthe development of high-performance coatings.",
"conclusion": "Conclusion A novel imidazolium-based ionic liquid (HFIL)\nwith no polar functional\ngroups has been developed, which exhibits unique wettability properties.\nAlthough the bulk material is both hydrophobic and oleophobic, HFIL\nforms a simultaneously hydrophilic and highly oleophobic surface when\napplied as a nanometer-thick coating on high-energy substrates, such\nas silicon wafers. This behavior contrasts with previous findings\nthat polar functional groups are essential for bonding on polar substrates;\nHFIL demonstrates a stable, bonded thin film that achieves oleophobicity\ncomparable to PTFE even after washing. Attempts to coat HFIL onto\nPTFE revealed that this bonding effect is not solely due to confinement\nbut rather involves specific interactions between HFIL and polar substrates.\nFurther investigation into substrate–IL interactions may enable\nthe design of durable IL coatings tailored to enhance surface properties\nin applications, such as wetting control, antifouling, and lubrication.",
"introduction": "Introduction The study of ionic liquids (ILs) has expanded\nsignificantly over\nthe past 2 decades, establishing ILs as versatile components in advanced\nmaterial applications. ILs, characterized by their unique properties,\nlike low volatility, high thermal stability, and tunable characteristics,\nhave proven valuable for applications in fields ranging from energy\nstorage to surface engineering. 1 , 2 One of the most compelling\nareas of IL research involves their use in surface wettability modification,\nengineering surfaces to selectively interact with different liquids,\nwhich has far-reaching implications in areas such as medical instruments,\nenergy conservation, and environmental protection. 3 In particular, simultaneously hydrophilic and oleophobic\nsurfaces\noffer substantial advantages for applications such as long-term antifogging,\ndetergentless cleaning, and oil–water separation. 3 − 5 Traditional surface treatments achieve either hydrophilicity or\noleophobicity, yet it remains challenging to engineer surfaces that\ncombine both properties without requiring polar functional groups\nor complex coating procedures. 3 , 6 Previously, Li et al.\nexplored the unusual wetting behavior of nanometer-thick perfluoropolyether\n(PFPE) films, finding that they could achieve higher hexadecane contact\nangles (HCAs) than water contact angles (WCAs) on silicon substrates. 7 This unusual behavior, in which water readily\nwets the surface while oil is repelled, was attributed to a kinetic\nlimitation of the polymer coating. The study suggested that small\nstructural “voids” within the film, formed by molecular\npacking or dynamic rearrangements, permit water molecules to quickly\npenetrate the layer and interact with the substrate, whereas larger\nhexadecane molecules cannot penetrate easily and, thus, interact primarily\nwith the coating surface. These findings underscore the importance\nof fluorination and molecular packing in achieving simultaneous hydrophilic–oleophobic\nproperties. With further expansion on this concept, Wang et\nal. later investigated\nthe role of end groups in PFPE coatings and their influence on dual\nwetting characteristics. 6 By variation\nof the number of polar hydroxyl end groups, they demonstrated that\nonly specific amounts of polar functionality could establish the appropriate\ninterchain distances to allow water to penetrate while hindering oil.\nWhen there were no polar groups, the polymer chains remained loosely\npacked and mobile, allowing both water and oil to readily pass through.\nWhen there were too many polar groups, the polymer chains bonded to\nthe surface with more frequency, creating interchain gaps that were\ntoo tight to even allow water through. The study emphasized that the\nunique combination of oleophobicity and hydrophilicity is achieved\nnot only through fluorination but also by carefully controlling the\nend-group interactions with the substrate to ensure a stable, selective\nlayer. 6 In this study, we explore\na novel IL, HFIL, which exhibits dual\nhydrophilic and oleophobic properties without the presence of polar\nfunctional groups, a feature that challenges the previous understanding\nestablished with PFPEs. This work seeks to contribute to the growing\nbody of knowledge on ILs by examining HFIL’s unique wettability\nbehavior and its potential implications for practical applications\nin surface engineering and coating technologies.",
"discussion": "Results and Discussion Figure 1 shows the\nchemical structure of HFIL or 1–1 H ,1 H ,2 H ,2 H -perfluorohexyl-3-methylimidazolium\nbis(nonafluorobutanesulfonyl)imide. At room temperature, bulk HFIL\ntakes on a solid, largely crystalline structure with a melting point\nof around 85 °C. Figure 1 Chemical structure of HFIL. Figure 2 presents\nthe wettability of HFIL for water and hexadecane in both the bulk\nand thin-film states. As a bulk solid, HFIL exhibits a notably high\nHCA of 74.1 ± 6°, a value significantly greater than polytetrafluoroethylene\n(PTFE) surfaces, which typically yield HCAs of around 40–50°. 8 − 11 This unusually high oleophobicity suggests that HFIL’s molecular\nstructure repels oil more effectively than conventional oleophobic\nsurfaces. Figure 2 HCA and WCA of HFIL as a bulk material and as a nanometer-thick\ncoating on a Si wafer. With the transition to the thin-film regime, specifically\nto a\n0.75 ± 0.02 nm layer on native oxide silica, HFIL demonstrates\na marked reduction in the WCA relative to the bulk IL, dropping from\napproximately 90° to 40°, while the HCA remains stable across\nboth regimes. This observation aligns with the size-exclusion-based\npenetration mechanism for coatings with simultaneous hydrophilic–oleophobic\nproperties. Small structural “voids” within the coating,\nformed by molecular packing or dynamic rearrangements, permit small\nwater molecules to quickly penetrate the layer and interact with the\nsubstrate, whereas larger hexadecane molecules cannot penetrate easily\nand, thus, interact primarily with the coating surface. The surface\nof bare native oxide silica is both highly hydrophilic and highly\noleophilic, with very low contact angles (<10°) for water\nand hexadecane. 12 As the HFIL coating thickness\ndecreases, its surface coverage diminishes, rendering the water (but\nnot hexadecane) penetration easier and consequently reducing WCA (but\nnot HCA) significantly, as illustrated in Figure 2 . Prior studies indicate that polar\nfunctional groups are critical\nin achieving hydrophilic–oleophobic properties, as they enable\ncoatings to “bond” with polar substrates, stabilizing\nthe coating through specific packing arrangements. Coatings lacking\nthese polar groups typically remain mobile on the substrate, allowing\nboth water and hexadecane to penetrate with minimal resistance. 6 , 12 However, as shown in Figure 3 a, HFIL achieves a stable, sub-monolayer thickness of 0.30\n± 0.01 nm on native oxide silicon, even without polar functional\ngroups. Additionally, the washed surface maintains a HCA of 49.7 ±\n1.7°, as seen in Figure 3 b, comparable to PTFE, indicating significant oleophobicity.\nGiven this information and the fact that bulk HFIL is a solid and\nhighly crystalline structure at room temperature, it is likely that\nthe confined nanofilm on the substrate forms a fairly tight-packed\nsolid-like arrangement, contributing to its stability and resistance\nto washing. Figure 3 (a) Thickness of HFIL on Si wafer before and after washing and\n(b) resulting HCA at each state. Figure 4 illustrates\na notably different interaction when the HFIL is applied to a PTFE\nsubstrate. Bare PTFE exhibits a HCA of 49.8 ± 1.3° and a\nWCA of 104.2 ± 1.3°, values consistent with the literature. 8 − 11 , 13 − 16 Following immersion in HFIL solution,\nthe HCA rises slightly to 53.2 ± 1.6°, while the WCA falls\nslightly to 101.0 ± 1.7°. This simultaneous increase of\nHCA and decrease of WCA represent the added interaction between the\ntest liquids and the IL coating, shifting the contact angles toward\nthose of the bulk HFIL. However, both contact angles showed minimal\nchange, indicating that both hexadecane and water primarily interact\nwith the PTFE surface rather than the IL coating. This outcome suggests\na low bonded ratio of HFIL on PTFE, likely due to PTFE’s low\nsurface energy, and further implies that the bonded HFIL observed\non silica is not solely a result of confinement-induced ordering. 8 , 16 Instead, a stronger interaction between HFIL and high-energy surfaces\npromotes a densely packed nanofilm that restricts hexadecane penetration. Figure 4 HCA and\nWCA of Teflon with and without HFIL coating. In prior work, the surface energy of polymer substrates,\nsuch as\npoly(methyl methacrylate) (PMMA), polystyrene (PS), and polycarbonate\n(PC), has been increased using ultraviolet (UV)–ozone treatment\nto enhance bonding of functionalized coatings. 15 , 17 Here, a clean PTFE substrate was subjected to UV–ozone treatment\nfor 1 h before HFIL coating. Although slight shifts in contact angles\nwere observed, the changes were minimal, suggesting a limited impact\nof UV–ozone treatment on HFIL bonding. These slight angle adjustments\nmay instead reflect surface roughness changes induced by UV–ozone\ntreatment, as previously documented. 15 The HCA and WCA were used to calculate the surface free energy\nof the bulk HFIL as well as the HFIL-coated Si wafer based on the\nFowkes model shown in eq 1 1 where γ L is\nthe surface tension of the liquid, γ L d and γ L p are the dispersive (nonpolar) and polar\ncomponents of the liquid surface tensions, respectively, and γ S d and γ S p are the dispersive\nand polar components of the surface free energy, respectively. 18 The model assumes that the total surface energy\nis equal to the sum of the dispersive and polar components. Hexadecane\nis considered to be completely nonpolar; as such, the dispersive component\nis the only component of the surface tension, reported as 27.66 mN/m. 19 For water, the polar and nonpolar components\nfor water are reported as 43.7 and 29.1 mN/m, respectively. 20 The results for the surface energy of HFIL and\nthe calculated polar and nonpolar components are listed in Table 1 . The bulk HFIL has\na low surface energy of 20.2 mN/m, just slightly higher than that\nreported for PTFE. 11 The surface energies\nof the thin-film coatings increase significantly relative to the bulk,\nreaching 59.7 mN/m on the 0.75 nm coated Si wafer and 62.4 mN/m on\nthe 0.56 nm coated wafer. Table 1 Surface Free Energy of HFIL as a Bulk\nMaterial and as Nanometer-Thick Coatings on Silicon Wafer γ S d (mN/m) γ S p (mN/m) γ S total (mN/m) bulk HFIL 11.2 9.0 20.2 0.75 nm HFIL/Si wafer 11.0 48.7 59.7 0.56 nm HFIL/Si wafer 13.3 49.1 62.4"
} | 2,990 |
27609891 | null | s2 | 4,370 | {
"abstract": "A key aspect of bacterial survival is the ability to evolve while migrating across spatially varying environmental challenges. Laboratory experiments, however, often study evolution in well-mixed systems. Here, we introduce an experimental device, the microbial evolution and growth arena (MEGA)-plate, in which bacteria spread and evolved on a large antibiotic landscape (120 × 60 centimeters) that allowed visual observation of mutation and selection in a migrating bacterial front. While resistance increased consistently, multiple coexisting lineages diversified both phenotypically and genotypically. Analyzing mutants at and behind the propagating front, we found that evolution is not always led by the most resistant mutants; highly resistant mutants may be trapped behind more sensitive lineages. The MEGA-plate provides a versatile platform for studying microbial adaption and directly visualizing evolutionary dynamics."
} | 232 |
36855698 | PMC9888626 | pmc | 4,371 | {
"abstract": "Superomniphobic surfaces\nthat can self-repair physical damage are\ndesirable for sustainable performance over time in many practical\napplications that include self-cleaning, corrosion resistance, and\nprotective gears. However, fabricating such self-repairing superomniphobic\nsurfaces has thus far been a challenge because it necessitates the\nregeneration of both low-surface-energy materials and hierarchical\ntopography. Herein, a water-responsive self-repairing superomniphobic\nfilm is reported by utilizing cross-linked hydroxypropyl cellulose\n(HPC) composited with silica (SiO 2 ) nanoparticles (HPC-SiO 2 ) that is treated with a low-surface-energy perfluorosilane.\nThe film can repair physical damage (e.g., a scratch) in approximately\n10 s by regenerating its hierarchical topography and low-surface-energy\nmaterial upon the application of water vapor. The repaired region\nshows an almost complete recovery of its inherent superomniphobic\nwettability and mechanical hardness. The repairing process is driven\nby the reversible hydrogen bond between the hydroxyl (−OH)\ngroups which can be dissociated upon exposure to water vapor. This\nresults in a viscous flow of the HPC-SiO 2 film into the\ndamaged region. A mathematical model composed of viscosity and surface\ntension of the HPC-SiO 2 film can describe the experimentally\nmeasured viscous flow with reasonable accuracy. Finally, we demonstrate\nthat the superomniphobic HPC-SiO 2 film can repair physical\ndamage by a water droplet pinned on a damaged area or by sequential\nrolling water droplets.",
"conclusion": "Conclusions We have developed a\nwater-responsive self-repairing superomniphobic\nfilm by utilizing a cross-linked HPC-SiO 2 composite treated\nwith a low-surface-energy perfluorosilane. The HPC-SiO 2 film can repair a deep scratch and restore its inherent superomniphobic\nwettability and mechanical hardness upon exposure to water vapor.\nOur film’s self-repairing capability can be attributed to the\nreversible hydrogen bonds between the hydroxyl groups in the HPC-SiO 2 film which can be dissociated upon exposure to water molecules.\nConsequently, the HPC-SiO 2 film acquires sufficient mobility\nand demonstrates a viscous flow into a scratch cavity resulting in\nrepairing the damage. We demonstrated that a mathematical model based\non the glassy thin film equation can describe the experimentally measured\ntime-dependent evolution of a scratch profile with reasonable accuracy.\nWe also showed that a water droplet pinned on a scratch can depart\nand roll off the surface after repairing it. Finally, we demonstrated\nthat consecutive rolling water droplets can repair a deep scratch\nengraved on our superomniphobic HPC-SiO 2 film. We envision\nthat our water-responsive self-repairing superomniphobic surface can\noffer extensive utility in the marine, automotive, and aviation industries.",
"introduction": "Introduction A\nsurface that can repel liquids with both high (e.g., water) and\nlow (e.g., oil) surface tension (γ lv ) has demonstrated\npotential for a wide range of practical applications including oil–water\nseparation, 1 − 3 self-cleaning, 4 , 5 drag reduction, 6 , 7 and corrosion resistance 8 , 9 due to its extreme liquid\nrepellency. 10 Designing such a superomniphobic\nsurface [i.e., a surface that displays apparent contact angles (θ*)\ngreater than 150° and a roll-off angle (ω) less than 10°\nwith virtually all liquids] involves a low-solid-surface-energy (γ sv ) coating along with surface roughness. 6 Unlike a superhydrophobic surface (i.e., a surface exhibiting\nθ* > 150° and ω < 10° with water), we 6 , 8 , 11 , 12 and others 13 , 14 demonstrated that a superomniphobic\nsurface must possess a re-entrant surface texture (i.e., convex or\noverhang topography 10 ) which enables a\ncomposite solid–liquid–air interface even with a low-surface-tension\nliquid such as an oil. The composite interface can be further reinforced\nby a hierarchical surface topography (i.e., surface texture possessing\ntwo or more length scales 6 ) which often\ninvolves a microscopic coarse texture along with finer-scale textures\n(i.e., typically nanometric scale). 15 , 16 However, such\na re-entrant surface texture with a hierarchical topography often\nresults in poor mechanical durability. 6 For example, a superomniphobic surface can be readily compromised\nbecause a low-surface-energy coating may become delaminated upon mechanical\nabrasion. 7 , 17 Also, the surface textures are delicate\nand tend to easily undergo physical damage. 17 , 18 Thus, developing mechanically durable superomniphobic surfaces has\nthus far been an active area of research. A variety of strategies\nhave been employed to enhance the mechanical\ndurability of superomniphobic surfaces. For example, a low-surface-energy\ncoating has been grafted or cross-linked directly to the substrate. 19 , 20 This strategy enhances mechanical durability 19 by eliminating the need for an intermediate binding layer\nwhich can easily be compromised upon external stresses. Also, inherently\nhard materials (e.g., ceramics 21 , 22 and metals 23 , 24 ) have been employed in fabricating durable superomniphobic surfaces.\nAlthough these surfaces can maintain their extreme liquid repellency\neven after being subjected to mechanical stress, they are also vulnerable\nto physical damage upon cyclic loading or when being subjected to\nextremely high forces. 25 In a new\nvein, superomniphobic surfaces that can repair physical\ndamage and restore their inherent physicochemical properties have\nbeen reported. 2 Such self-repairing superomniphobic\nsurfaces require restoring both surface chemistry (e.g., low-solid-surface-energy\ncoating) and surface texture geometry (e.g., re-entrant and/or hierarchical\nsurface texture). 26 , 27 Conventionally, these surfaces\ncan initiate the repairing process in response to external triggers\nsuch as heat 8 , 28 − 30 and light irradiation. 31 , 32 For example, Zhou et al. 30 fabricated\nself-repairing superomniphobic coatings by using a mixture of low-surface-energy\nnanoparticles (e.g., polytetrafluoroethylene (PTFE) and DuPont Zonyl321)\nalong with a fluorocarbon surfactant. The coating restored its superomniphobic\nwettability upon heating due to the migration of fluorinated alkyl\nchains toward the surface to lower the overall surface free energy.\nZhao et al. 33 utilized a composite of epoxy\nresin and perfluorodecyl polysiloxane-modified silica nanoparticles\nto fabricate a self-healing superomniphobic coating on a metal substrate.\nThe coating can recover its liquid repellency after repairing physical\ndamage upon an increase in the temperature above its glass transition\ntemperature ( T g ). Dongli et al. 32 reported a self-healing superomniphobic coating\nprepared by using a combination of ultraviolet (UV) light-curable\npolyurethane acrylic resin and fluorinated alumina nanoparticles.\nThe coating demonstrated an accelerated self-healing of physical damage\nunder UV light irradiation. Recently, we 8 also reported a superomniphobic surface fabricated by reacting epoxidized\nsoybean oil, perfluorinated epoxy, citric acid, and silica nanoparticles.\nThe surface can repair a deep crack at a temperature above its glass\ntransition temperature. While heat or light has been extensively\nemployed as a trigger\nto initiate the self-repairing process, the usage of water as a self-repairing\ntrigger has been barely reported for these super-repellent surfaces\ndespite its versatility and immediate availability in many practical\napplications. For example, Bai et al. 34 fabricated a superhydrophobic surface by sequentially spraying an\nepoxy resin and a mixture of poly(methyl methacrylate), zinc stearate,\nand stearic acid. The surface can repair damage when submerged in\ndeionized (DI) water for 30 min and recover its original superhydrophobic\nwettability. Chen et al. 35 coated cotton\nfabric with layer-by-layer deposited branched poly(ethylenimine),\nammonium polyphosphate, and fluorinated-decyl polyhedral oligomeric\nsilsesquioxane. The coating can restore its compromised superhydrophobicity\nwhen exposed to an ambient environment with a relative humidity of\n35% for 4 h after being subjected to oxygen (O 2 ) plasma\nor mechanical rubbing. Fabricating a water-responsive superomniphobic\nsurface that can\nrepair physical damage requires the consideration of the following\ndesign criteria. First, the surface should acquire sufficient mobility\nwhen interacting with water to restore both a low-energy coating and\nrough texture. Also, the repairing process should be completed within\na short period of time. This becomes critical when water is applied\nin its vapor form instead of as a liquid which can significantly retard\nthe repairing process due to a limited number of available water molecules\nfor interacting with the surface at a given condition (e.g., volume).\nFinally, the repaired region should exhibit its inherent mechanical\nand chemical properties. This may be particularly challenging for\na surface that absorbs water molecules (e.g., hydrogel) because the\nabundant absorption of water often deteriorates the surface’s\nphysicochemical properties. By considering these design criteria,\nherein, a water-responsive\nself-repairing superomniphobic film is developed by utilizing cross-linked\nhydroxypropyl cellulose (HPC) composited with SiO 2 nanoparticles\nthat are treated with a low-surface-energy perfluorosilane. The film\ndemonstrates that it can repair a deep scratch upon exposure to water\nvapor for ≈10 s and restore its inherent superomniphobic wettability\nand mechanical hardness. This can be attributed to the reversible\nhydrogen bonds between the free hydroxyl groups of the HPC-SiO 2 film which can be readily dissociated upon exposure to water\nvapor. Consequently, the HPC-SiO 2 film acquires sufficient\nmobility and demonstrates a viscous flow into a scratch resulting\nin repairing the damage. A mathematical model based on the glassy\nthin film equation is utilized to describe the viscous flow of the\nHPC-SiO 2 film. Finally, the film demonstrates that it can\nrepair damage by a water droplet pinned on the damaged area or consecutive\nrolling water droplets at ambient conditions. We envision that our\nfilm can provide a viable solution for a protective coating against\nhostile environments in the marine, automotive, and aviation applications.",
"discussion": "Results\nand Discussion We fabricated a self-repairing superomniphobic\nfilm by spraying\na solution of HPC, glyoxal, and SiO 2 nanoparticles (average\ndiameter ≈250 nm) with a weight ratio of 98:1:1 (HPC:glyoxal:SiO 2 ) on a glass substrate. Subsequent heat treatment at 50 °C\nfor 15 min ( Experimental Section ) results\nin a cross-linked composite of HPC-SiO 2 due to the hydrolysis–condensation 36 , 37 of hydroxyl groups of HPC and SiO 2 with glyoxal (Supporting\nInformation, Section 1 ). Note that glyoxal\nwas used as a cross-linker while SiO 2 nanoparticles were\nused to create re-entrant surface texture. 8 The cross-linked HPC-SiO 2 film with a consolidated network\nenables the recovery of the original surface topography (i.e., re-entrant\ntexture) when the HPC gains sufficient mobility upon exposure to water.\nTo lower the overall solid surface energy, the HPC-SiO 2 film was vapor-deposited with 1 H ,1 H ,2 H ,2 H -perfluorodecyltrichlorosilane\n(F-silane) at 130 °C for 60 min ( Experimental\nSection ). Note that hydrolysis of F-silane molecules results\nin silanol groups that can undergo polycondensation to form a covalent\nbond with any unreacted hydroxyl groups on the HPC-SiO 2 film. 38 The resulting HPC-SiO 2 film exhibits a hierarchical\ntexture with re-entrant curvatures composed of micro- and nanoscale\nSiO 2 aggregates ( Figure 1 a). The HPC-SiO 2 film exhibits very high\ncontact angles for liquids with a broad range of surface tension (γ lv ) values (i.e., superomniphobic wettability) ( Figure 1 b and SI, Section 2 ). For example, the apparent advancing (θ* adv ) and receding (θ* rec ) contact angles of\nwater (γ lv = 72.1 mN m –1 ) were\nmeasured as θ* adv = 164° ± 3° and\nθ* rec = 163° ± 2°, respectively, while\nthose for n -heptane (γ lv = 20.1\nmN m –1 ) were measured as θ* adv =\n156° ± 3° and θ* rec = 145° ±\n3°, at a relative humidity of ≈9% ± 3%. We also measured\nthe apparent contact angles for water and n -heptane\nat a higher relative humidity (≈98%). Note that we acquired\nand maintained such a high relative humidity by water vapor with a\nflow rate of ≈0.5 mL min –1 at a distance\nof ≈30 cm in a custom-made humidifier ( Experimental\nSection ). The measured apparent contact angles for n -heptane were θ* adv = 155° ±\n2° and θ* rec = 143° ± 2°; those\nfor water were θ* adv = 159° ± 2° and\nθ* rec = 155° ± 2°. Such a slight decrease\nin the water apparent contact angles can be attributed to the condensation\nof water vapor on the surface which may partially replace the air\npockets trapped between the solid and the contacting water droplet. 39 Figure 1 (a) SEM image showing the surface of an HPC-SiO 2 film\nthat exhibits a hierarchical surface texture. The scale bar represents\n20 μm. The inset demonstrates a high-magnification SEM image\nof nanoscale SiO 2 particles with re-entrant curvatures.\nThe scale bar represents 1 μm. (b) Measured advancing and receding\napparent contact angles, as well as roll-off angles for liquids with\na broad range of surface tension values on an HPC-SiO 2 film.\n(c) Photographs of four droplets including ethanol (dyed blue, γ lv = 22.1 mN m –1 ), n -octane\n(dyed gray, γ lv = 21.2 mN m –1 ), n -heptane (dyed red, γ lv = 20.1 mN m –1 ), and n -hexane (dyed yellow, γ lv = 18.4 mN m –1 ) on an HPC-SiO 2 film fabricated on (i) stainless steel, (ii) polyester fabric, and\n(iii) ceramic-resin composite; (iv) freestanding HPC-SiO 2 film. The scale bar represents 3 mm. We also measured the roll-off angles (i.e., minimum tilting angle,\nω for a surface at which a contacting droplet starts to roll-off 40 ) for these liquids. The results show that even\na very low-surface-tension liquid such as n -heptane\ncan exhibit a roll-off angle of ω = 6° ± 1° on\nour HPC-SiO 2 film (see also Figure 1 b). We also demonstrated that a droplet of n -dodecane (γ lv = 25.3 mN m –1 , volume ≈ 5 μL) can bounce off our HPC-SiO 2 film two times before residing on the surface ( Movie S1 ). Further, the contact angle for a droplet of ethanol:water\nmixture (80:20 vol:vol, γ lv ≈ 24.3 mN m –1 ) 41 remained nearly unaffected\nwhile being evaporated, which corroborates that our HPC-SiO 2 film can form a robust solid–liquid–air composite\ninterface (SI, Section 3 ). Please note\nthat the measured apparent contact angles and the roll-off angles\nvaried on the HPC-SiO 2 films prepared with varied spraying\ntimes (SI, Section 4 ). The HPC-SiO 2 film can be readily fabricated on a variety\nof substrates including metal, fabric, composite, or even freestanding\n( Figure 1 c). This is\ndue to the dialdehyde chemistry of glyoxal that can strongly anchor\nto virtually any substrates via an acetalization reaction 42 , 43 which can form a covalent acetate linkage 44 with the underlying substrate (SI, Section 5 ). Physical damage such as deep scratches, cracks, and cavities\non\na superomniphobic surface can irreversibly compromise its liquid repellency. 17 , 18 We demonstrated that our HPC-SiO 2 film can repair physical\ndamage upon exposure to water vapor. First, a scratch (width ≈\n60.3 ± 2.1 μm) was engraved onto the HPC-SiO 2 film (thickness ≈ 100.1 ± 3.2 μm) by utilizing\na razor blade. Please note that the scratch was engraved such that\nthe underlying glass substrate was revealed ( Figure 2 ai). Upon exposure to water vapor with a\nflow rate of ≈0.5 mL min –1 at room temperature\n(≈22 °C), the scratch started to narrow down ( Figure 2 aii) and eventually\ndisappeared at t ≈ 10 s ( Figure 2 aiii). A movie showing the\nrepairing of physical damage on our HPC-SiO 2 film upon\nthe application of water vapor is provided in the Supporting Information\n( Movie S2 ). Figure 2 (a) Time-sequence optical\nmicroscopy images showing self-repairing\nof a scratch on the HPC-SiO 2 film upon introducing water\nvapor (flow rate ≈ 0.5 mL min –1 ). The scale\nbar represents 50 μm. The inset shows cross-sectional optical\nprofilometry images of a scratch during the repairing process. (b)\nTime-dependent plots of self-repairing efficiency (ζ) at different\nflow rates of water vapor. The inset shows time-dependent cross-section\nphotographs of a scratch (width ≈ 59 μm and depth ≈\n100 μm) engraved on the HPC-SiO 2 film undergoing\na repairing process when exposed to water vapor with a flow rate of\n≈0.5 mL min –1 . The scale bar is 50 μm.\n(c) Measured roll-off angles (ω) for liquids with a broad range\nof surface tension values placed on a repaired region. For a comparison,\nthe ω values on the scratch are also provided. The inset shows\ntime-sequence energy dispersive X-ray spectroscopy (EDS) maps for\nfluorine on a scratch. The scale bar is 50 μm. (d) Plot of the\nmeasured load–displacement on the repaired region on the HPC-SiO 2 film. For a comparison, the plot for the as-prepared surface\nis also shown. We calculated the self-repairing\nefficiency (ζ) as a function\nof water vapor exposure time which is defined as ζ (%) = (1\n– ( A t / A 0 )) × 100. Here, A t is the time-dependent cross-sectional area\nof the scratch during water vapor exposure while A 0 is that of the as-prepared scratch (SI, Section 6 ). Figure 2 b shows\nthe plots of self-repairing efficiency as a function of exposure time\nat different flow rates of water vapor. The values of ζ start\nto rapidly increase upon exposure to water vapor with a flow rate\nof ≈0.5 mL min –1 and reach ζ ≈\n100% at t ≈ 10 s, which indicates that a scratch\ndisappeared completely. When the water vapor was introduced at a lower\nflow rate of ≈0.1 or ≈0.2 mL min –1 , complete reparation (ζ ≈ 100%) was observed at t ≈ 36 s and t ≈ 24 s, respectively.\nWe showed that a scratch on our HPC-SiO 2 film can be repaired\nby ethanol vapor (SI, Section 7 ). To demonstrate that our HPC-SiO 2 film can recover its\ninherent superomniphobic wettability, we measured the roll-off angles\nfor liquids after the scratch was repaired. A liquid droplet was placed\non a region where the scratch was engraved and repaired. By comparing\nthe ω values with those measured on the as-prepared surface,\nwe verified that our HPC-SiO 2 film exhibited a nearly complete\nrecovery of its inherent superomniphobic wettability ( Figure 2 c). This is a result of restoring\nboth its hierarchical surface texture and low-solid-surface-energy\ncoating in the damaged region (see also the inset images in Figure 2 c). We also demonstrated\nthat our HPC-SiO 2 film can restore its superomniphobic\nwettability after repeated damage–repairing cycles (SI, Section 8 ). We conducted the nanoindentation\ntest 45 on the repaired region of the HPC-SiO 2 film ( Experimental Section ). The\nresults show that the\nmaximum indentation load ( F = 9000 μN) resulted\nin a displacement ( h ) value of h = 2185 nm on the repaired region which is close to that measured\non the as-prepared film ( h = 2070 nm) ( Figure 2 d). This clearly indicates\nthat our HPC-SiO 2 film can recover its inherent mechanical\nhardness after repairing physical damage. Our HPC-SiO 2 film’s ability to repair damage\nand recover its mechanical hardness can be attributed to the reversible\nhydrogen bonds between the hydroxyl groups in the HPC which can be\ndissociated upon exposure to water molecules. 46 This enables an HPC-SiO 2 film to exhibit a viscous flow\ninto the scratch cavity and repair it ( Scheme 1 ). Subsequently, the dissociated hydrogen\nbonds can be reformed in the absence of water vapor which enables\nthe HPC-SiO 2 film to demonstrate repeated damage–repairing\ncycles. Please note that the absorption of water by the HPC-SiO 2 film is negligible (SI, Section 9 ). Scheme 1 Illustration of the Proposed Self-Repairing Process of HPC-SiO 2 Film upon Water Vapor Exposure by Reversible Hydrogen Bonds\nbetween Free Hydroxyl Groups By assuming that the reversible hydrogen bond is a driving force\nfor the repairing process, a cross-linker (i.e., glyoxal) concentration\ncan directly affect the extent of repair. To prove this, we conducted\ndamage–repairing experiments on the HPC-SiO 2 films\ncross-linked with various glyoxal concentrations (e.g., 1.0%, 5.0%,\nand 10.0% of glyoxal with respect to HPC weight). Please note that\nthe SiO 2 concentration remains the same (i.e., 1.0%) with\nrespect to HPC and glyoxal weight. The HPC-SiO 2 films (thickness\n≈ 100 μm) were engraved with a deep scratch (width ≈\n60.5 ± 1.7 μm and depth ≈ 100.2 ± 2.1 μm)\nfollowed by exposure to water vapor with a flow rate of ≈0.5\nmL min –1 . Figure 3 a shows the measured self-repairing efficiency values\n(ζ) as a function of water vapor exposure time ( Experimental Section ). The results show that the HPC-SiO 2 film cross-linked with a 1.0 wt % glyoxal reached ζ\n≈ 100% at t ≈ 10 s whereas that prepared\nwith 10.0 wt % glyoxal exhibited ζ ≈ 100% at t ≈ 60 s. A more rapid repairing process on a film\ncross-linked with a lower glyoxal concentration is a consequence of\na reduced structural restraint and a larger free volume between the\nHPC chains which facilitates the penetration of water molecules. 46 This results in an accelerated increase in the\nchain mobility of HPC upon exposure to water vapor. Figure 3 (a) Measured time-dependent\nself-repairing efficiency (ζ)\nvalues on the HPC-SiO 2 films prepared with varying glyoxal\nconcentrations upon exposure to water vapor (flow rate ≈ 0.5\nmL min –1 ). The calculated ζ model values by using eq 1 match reasonably well with the experimental data. The inset shows\na zoomed-in plot during the first 2 s. (b) Plot of the measured highest\nζ values on the HPC-SiO 2 films prepared with varying\nglyoxal concentrations engraved with scratches of different aspect\nratios (α). Water vapor was applied with a constant flow rate\nof ≈0.5 mL min –1 until either the width or\nthe depth of a scratch remains unchanged. The time-dependent evolution of a vertical profile ( H ( x , t )) of a scratch can be described\nby the glassy thin film equation 47 , 48 which is given\nas 1 where H m , η (Pa s), and γ\n(mN m –1 ) are\nthe height (i.e., thickness), viscosity, and surface tension of the\nHPC-SiO 2 film, respectively. It has been demonstrated 47 that physical damage engraved on a film of glassy\npolymer (i.e., amorphous polymers that can demonstrate a glass transition\ntemperature 49 ) can exhibit time-dependent\nevolution in its vertical profile ( H ( x , t )) when it gains mobility. Please note that HPC\nhas a high degree of amorphous content 50 and shows a glass transition temperature (SI, Section 10 ). 51 By assuming H ( x , t ) as an infinitesimal\nstep function and by utilizing a Fourier transform, 52 eq 1 can be\nsolved for ( H ( x , t )). Please note that the values of η and γ utilized in\nthe equation are experimentally determined (SI, Section 11 ). The calculated values of H ( x , t ) were then utilized to determine the\ntheoretical self-repairing efficiency according to the relation ζ model = ( H ( x , t )/ H m ) × 100. The results show that\nthe calculated values of ζ model by using eq 1 and the experimentally\nmeasured ζ values match reasonably well (see Figure 3 a). Please note that the values\nof η and γ of our HPC-SiO 2 film showed insignificant\nchange upon varied water vapor exposure time (SI, Section 12 ) which can be attributed to a combinatorial effect\nof negligible absorption and rapid evaporation of water vapor by the\nHPC-SiO 2 film (see also Section 9 in the SI). Further, we demonstrated that the adhesion strength\nof the HPC-SiO 2 film to the underlying substrate was not\nsignificantly affected after being exposed to water vapor (SI, Section 13 ). Based on eq 1 , it\ncan be inferred that the viscous flow-driven reparation on our HPC-SiO 2 film can be limited when the width or depth of a scratch\nis too wide or deep. For example, when a scratch is too wide, the\ntwo edges of the HPC-SiO 2 film may not meet each other,\nwhich can result in incomplete repairing. To study our HPC-SiO 2 film’s repairing capability, we prepared HPC-SiO 2 films with varying glyoxal concentrations and engraved a\nscratch with different ratios of the depth to the width (i.e., aspect\nratio, α ≈ 0.11, 0.32, 0.51, 1.12, 1.65, and 2.06). Please\nnote that the thickness (i.e., depth of scratch) of all HPC-SiO 2 films was approximately 100 μm. All films were subjected\nto water vapor with a flow rate of ≈0.5 mL min –1 until either the width or the depth of a scratch remained unchanged.\nSubsequently, we calculated the self-repairing efficiency values based\non the cross-sectional area of the scratches (see also Section 6 in the SI). Figure 3 b shows the values of ζ as a function\nof α values. The results show that all HPC-SiO 2 films\nexhibited ζ ≈ 100% when α ≥ 1.12 (i.e.,\nα = 1.12, 1.65, and 2.06) while they demonstrated incomplete\nrepairing when α ≤ 0.51 (i.e., α = 0.51, 0.32,\nand 0.11). For example, when α = 0.11 (i.e., depth ≈\n100 μm and width ≈ 908 μm), the self-repairing\nefficiency values were measured as ≈93%, ≈65%, and ≈38%,\nrespectively, for HPC-SiO 2 films prepared with glyoxal\nconcentrations of 1.0, 5.0, and 10.0 wt %. Despite the fact that a\nscratch can be partially repaired when α ≤ 0.51, a liquid\ndroplet can still maintain the Cassie–Baxter state 53 if it forms a robust composite solid–air–liquid\ninterface. For example, when a liquid droplet touches the underlying\nsubstrate (e.g., a clean glass slide), it may transition to the “fully\nwetted” Wenzel state 54 which often\nresults in a loss of super-repellency. On the other hand, when the\ntwo edges of the dissected film come close enough, our film can exhibit\nits inherent superomniphobic wettability although self-repairing efficiency\nis not 100%. This is possible because our HPC-SiO 2 film\npossesses a monolithic configuration which enables the edges to show\nthe same surface chemistry and topography as those of the topmost\nsurface. We demonstrated that our HPC-SiO 2 film can\nrecover its\nsuperomniphobic wettability even for a pinned water droplet placed\non a deep scratch ( Figure 4 a). As a water droplet (volume = 5 μL) gradually evaporates\nin ambient conditions (i.e., temperature ≈ 22 °C and relative\nhumidity ≈ 9% ± 3%), the scratch (≈101.4 ±\n1.9 μm deep and ≈60.3 ± 2.5 μm wide) becomes\nnarrower. Eventually, a water droplet abruptly departs and rolls off\nthe surface at t ≈ 56 s ( Figure 4 b). Note that a water droplet\nremained pinned without spreading on the HPC-SiO 2 film\ntilted at an angle of ≈10° relative to the horizontal\nplane. A movie illustrating the in situ recovery of super-repellency\nof our HPC-SiO 2 surface by a stationary water droplet is\nincluded as Movie S3 . Figure 4 Schematic (a) and time-sequence\nsnapshots (b) illustrating a pinned\nwater droplet that repairs a scratch on an HPC-SiO 2 film\nand subsequently departs and rolls off after the completion of the\nreparation. The scale bar in part b represents 2 mm. The inset shows\ncross-sectional images of a scratch undergoing self-repairing by a\npinned water droplet. The scale bar in the inset is 100 μm.\n(c) Schematic illustrating the self-repairing of a scratch on an HPC-SiO 2 film by sequential rolling water droplets. The inset shows\noptical microscopy images of a scratch region in contact with rolling\nwater droplets. The scale bar in the inset is 150 μm. (d) Measured\nself-repairing efficiency of a scratch on an HPC-SiO 2 film\nby sequential rolling water droplets dispensed at varied time intervals\n(0.5, 1.0, and 1.5 s). Finally, we demonstrated\nthat consecutive rolling water droplets\ncan repair a deep scratch on our HPC-SiO 2 film 55 ( Figure 4 c). The water droplets with a constant volume (5 μL)\nwere consecutively introduced to the top of the film by using a syringe\npump with a constant interdroplet time interval of ≈0.5 s ( Experimental Section ). The film was tilted to 10°\nwith respect to the horizontal plane such that rolling water droplets\nare not pinned on the scratch (≈100.6 ± 2.0 μm deep\nand ≈60.1 ± 2.6 μm wide). We measured the contact\ntime of each rolling water droplet with the scratch region as ≈0.1\ns by using a high-speed camera image analysis. An optical microscopy\nimage confirms that the region of a scratch subjected to rolling water\ndroplets disappeared at t ≈ 72 s (see the\ninset in Figure 4 c).\nWe varied the interdroplet interval time and measured the self-repairing\nefficiency ( Figure 4 d). 55 The water droplets introduced at\na shorter interdroplet interval can result in a more rapid reparation.\nA movie illustrating the self-repairing of a scratch by consecutive\nrolling water droplets on our HPC-SiO 2 film is included\nas Movie S4 . Self-repairing by consecutive\nrolling water droplets is highly desirable in real-world applications.\nFor example, when a vehicle is coated with a water-responsive self-repairing\ncoating with super-repellency, physical damage can be readily repaired\nby rain droplets. Despite promising results, the reliance on\nenvironmental conditions\nis a shortcoming of our water-responsive self-repairing HPC-SiO 2 film. For example, when the film is subjected to an extremely\ncold environment, the applied water vapor can be deposited as ice\nor frost on the surface. As a consequence, the water molecules cannot\nbe readily accessible to the HPC-SiO 2 film, which can retard\nthe self-repairing process."
} | 7,402 |
21425769 | null | s2 | 4,373 | {
"abstract": "We present a simple and effective method to obtain refined control of the molecular structure of silk biomaterials through physical temperature-controlled water vapor annealing (TCWVA). The silk materials can be prepared with control of crystallinity, from a low content using conditions at 4 °C (α helix dominated silk I structure), to highest content of ∼60% crystallinity at 100 °C (β-sheet dominated silk II structure). This new physical approach covers the range of structures previously reported to govern crystallization during the fabrication of silk materials, yet offers a simpler, green chemistry, approach with tight control of reproducibility. The transition kinetics, thermal, mechanical, and biodegradation properties of the silk films prepared at different temperatures were investigated and compared by Fourier transform infrared spectroscopy (FTIR), differential scanning calorimetry (DSC), uniaxial tensile studies, and enzymatic degradation studies. The results revealed that this new physical processing method accurately controls structure, in turn providing control of mechanical properties, thermal stability, enzyme degradation rate, and human mesenchymal stem cell interactions. The mechanistic basis for the control is through the temperature-controlled regulation of water vapor to control crystallization. Control of silk structure via TCWVA represents a significant improvement in the fabrication of silk-based biomaterials, where control of structure-property relationships is key to regulating material properties. This new approach to control crystallization also provides an entirely new green approach, avoiding common methods that use organic solvents (methanol, ethanol) or organic acids. The method described here for silk proteins would also be universal for many other structural proteins (and likely other biopolymers), where water controls chain interactions related to material properties."
} | 483 |
35592261 | PMC9110805 | pmc | 4,374 | {
"abstract": "The Izhikevich (IZH) spiking neuron model can display spiking and bursting behaviors of neurons. Based on the switching property and bio-plausibility of the memristor, the memristive Izhikevich (MIZH) spiking neuron model is built. Firstly, the MIZH spiking model is introduced and used to generate 23 spiking patterns. We compare the 23 spiking patterns produced by the IZH and MIZH spiking models. Secondly, the MIZH spiking model actively reproduces various neuronal behaviors, including the excitatory cortical neurons, the inhibitory cortical neurons, and other cortical neurons. Finally, the collective dynamic activities of the MIZH neuronal network are performed, and the MIZH oscillatory network is constructed. Experimental results illustrate that the constructed MIZH spiking neuron model performs high firing frequency and good frequency adaptation. The model can easily simulate various spiking and bursting patterns of distinct neurons in the brain. The MIZH neuronal network realizes the synchronous and asynchronous collective behaviors. The MIZH oscillatory network can memorize and retrieve the information patterns correctly and efficiently with high retrieval accuracy.",
"conclusion": "7. Conclusion In this study, the experimental implementation of the MIZH spiking neuron model exhibited the typical biological neuron functions and remarkably improved the IZH spiking neuron model. The 23 spiking patterns of cerebral cortical neurons were successfully simulated by the MIZH spiking model. It performed good biological spiking frequency adaptability and high firing frequency. The exploration of the MIZH spiking model provided a pathway for efficiently emulating the spiking and bursting patterns in various cortical neurons. The firing patterns of the excitatory neurons, the inhibitory neurons, and other neurons were obtained in the MIZH spiking neuron model. To show the collective dynamical activities intuitively, we efficiently realized and analyzed the synchronization and asynchronization activities in the experimental simulations by appropriately regulating the synaptic weight. The MIZH spiking model showed rich collective dynamical behaviors. The further work was to use the MIZH oscillatory network to realize associative memory and correctly implement the storage and retrieval of distorted patterns. Meanwhile, it performed high retrieval accuracy by comparing the Hopfield, the IZH, and the MIZH oscillatory networks. In addition, the inherent characteristics of memristor will affect the construction of artificial neuron models. The nanometer size of the memristor is beneficial to the large-scale integration of the neuron model and the construction of the large-scale neural network. Low power consumption will actively promote the large-scale integration of the neuron circuit. The resistive switch characteristic makes the circuit simple, and there is no need to consider adding other electronic devices to realize the variable resistance property. Therefore, the memristor is introduced into the IZH model will bring potential research and application significance to the hardware implementation of the artificial neuron model and its network.",
"introduction": "1. Introduction The cortical neurons in the brain receive and process a large amount of the perceptual information and the behavior signals, and respond accordingly (Truong et al., 2018 ). It is necessary to pay attention to construct and improve various neuron models to mimic the functions of the cortical neurons and unveil the dynamical behaviors in the human brain (Hodgkin and Huxley, 1952 ; FitzHugh, 1961 ; Morris and Lecar, 1981 ; Rose and Hindmarsh, 1989 ; Ermentrout, 1996 ; Izhikevich, 2001 ). Due to the simple computation and rich spiking patterns (Izhikevich, 2003 ), the IZH spiking model has been widely studied. The multiplierless noisy IZH model can realize large-scale neural networks and possess the low-cost property (Haghiri et al., 2018 ). The Izhikevich neuron model incorporates the CORDIC algorithms to implement a neuromorphic system with high speed and accuracy (Heidarpur et al., 2019 ). The modified IZH model with mechanoelectrical and ultrasonic-magnetic effects can produce distinct spiking behaviors (Zhang et al., 2018 ). The bionic tactile sensor outputs are applied to the IZH model to simulate the spiking patterns and achieve the artificial touch (Rongala et al., 2017 ). The IZH spiking model with a simple structure performs biologically meaningful and rich spiking and bursting patterns. Nevertheless, there are many pivotal issues such as the realization of large-scale neural networks, implementation efficiency, power consumption, model structure, the discovery of novel materials, and the application of new devices which need to be solved. To efficiently construct the biology-inspired neuron model, we need to find a unique electronic device. The memristor has nonlinearity, non-volatility, low power consumption, nanoscale size, and is easily compatible with the CMOS. It shows excellent potential to emulate the synapse (Choi et al., 2018 ; Li et al., 2021 ) and neurons (Dev et al., 2020 ; Duan et al., 2020 ), and fabricate the neural networks (Wang et al., 2019 ; Joksas et al., 2020 ). Therefore, the memristor is believed to be a crucial device for artificial neuronal networks. This study reports the IZH spiking model integrated with a memristor. The MIZH spiking model is successfully built. The IZH spiking model is briefly introduced, and 23 spiking patterns are generated, which we present in Section 2. The MIZH spiking model with 23 firing patterns is described in Section 3. In Section 4, the firing patterns in the excitatory cortical neurons, the inhibitory neurons, and the other neurons are exhibited by the MIZH spiking model. The MIZH spiking model simulates brain-inspired collective dynamical activities. The patterns and mechanisms of synchronization and asynchronization are explored in Section 5. Section 6 primarily focuses on applying the MIZH spiking model to oscillatory associative memory, which can efficiently memorize the information pattern and accurately retrieve the distorted information patterns. The conclusion of the study is arranged in Section 7."
} | 1,557 |
31991593 | PMC7074653 | pmc | 4,376 | {
"abstract": "Microvalves are important flow-control devices in many standalone and integrated microfluidic applications. Polydimethylsiloxane (PDMS)-based pneumatic microvalves are commonly used but they generally require large peripheral connections that decrease portability. There are many alternatives found in the literature that use Si-based microvalves, but variants that can throttle even moderate pressures (1 bar) tend to be bulky (cm-range) or consume high power. This paper details the development of a low-power, normally-open piezoelectric microvalve to control flows with a maximum driving pressure of 1 bar, but also retain a small effective form-factor of 5 mm × 5 mm × 1.8 mm. A novel combination of rapid prototyping methods like stereolithography and laser-cutting have been used to realize this device. The maximum displacement of the fabricated piezoelectric microactuator was measured to be 8.5 μm at 150 V. The fabricated microvalve has a flow range of 0–90 μL min −1 at 1 bar inlet pressure. When fully closed, a leakage of 0.8% open-flow was observed with a power-consumption of 37.5 μW. A flow resolution of 0.2 μL min −1 —De-ionized (DI) water was measured at 0.5 bar pressure.",
"conclusion": "5. Conclusions A proportionally-controlled piezoelectric microvalve was designed, fabricated and characterized. DI water at room temperature was used for characterization. The uniqueness of this work is that a novel combination of rapid-prototyping methods like stereolithography for three-dimensional microchannels and laser-cutting for spacer and actuator fabrication were used. Using shim-steel as an accurate and easily modifiable spacing method for such a device has also been demonstrated for the first time. The fabrication methodology detailed above is cheaper than silicon Microelectromechanical System (MEMS) devices and is suitable for small batch production. These techniques can be extended to design other microfluidic devices like micropumps and flow sensors. The specifications of the final microvalve are given in Table 2 . A thorough characterization of the unimorph microactuator was performed and a good match was found between the analytical and measured results. A high central displacement of 7.5 μm was observed for an actuator dimension of 5 mm × 5 mm × 0.2 mm (PZT diameter = 4 mm). Due to limitations in the available flow sensor, the full potential of the microvalve could not be explored. The projected flow rate range for this microvalve is 0–750 μL min −1 (water). A low power consumption of 37.5 μW was measured.",
"introduction": "1. Introduction The need for techniques that more efficiently utilize chemical and biological reagents in chemical analysis systems led to the introduction of micro-total analysis systems ( μ TAS) [ 1 ]. The versatility of miniaturizing fluidics was realized and subsequently utilized in applications like drug-delivery [ 2 ], micro- and nano-spacecraft thermal cooling and propulsion systems [ 3 ], and lab-on-a-chip devices [ 4 ]. An essential component in integrated microfluidic devices is the microvalve. In conjunction with a pressure source, it is used to control, direct, or regulate the pressure or flow rate of media within these microfluidic circuits. Valves that operate in the μ m- to cm-length scales are generally classified as microvalves and they are usually fabricated using microfabrication techniques like (soft-)lithography and etching. Typically, microvalves contain a flow-channel that is obstructed by an active/passive element. Active microvalves, in contrast to their passive counterparts, have a controllable element within the device. These valves are usually classified by the working principle of the active element. One of the first reported microvalves used an electromagnetic solenoid actuator to control the separation between a valve-membrane and inlet-orifice [ 5 ]. A valve membrane or valve plate has since been one of the most used valving mechanisms to control fluid flow, with large variations in the actuation scheme. Principles like electrostatic actuation [ 6 ], piezoelectric actuation [ 7 ], pneumatic actuation [ 8 ], and thermopneumatic actuation [ 9 ] have been used with varying degrees of success. In this work, a proportional microvalve was designed, fabricated and characterized. It is based on a piezoelectric actuation method that uses a commercially available piezoelectric plate as the active element. As the active material functions without any further processing, all the advantages of piezoelectric actuation were retained. The valve deflection was controlled by applying a desired voltage across the piezoelectric material, resulting in proportional control of the fluid flow rate. The following sections detail the design, fabrication and characterization of the microvalve.",
"discussion": "4. Discussion 4.1. Actuation In Figure 7 a, it is clear that the up-sweep path is different from the down-sweep path of displacement behaviour. Measurements were made after 20–30 cycles of actuation due to which poling effects can be eliminated as the source of this difference in behaviour. This means that piezoelectric hysteresis is the main reason. Hysteresis in piezoelectric actuators is one of the major causes for positioning inaccuracies [ 17 ]. Hysteresis models exist that could be used to relate the voltage to displacement, but it is beyond the scope of this work. The displacement behaviour for UPMs with different PZT diameters was tested and has been shown in Figure 7 b. The error bars indicate the standard deviation of three cycles of actuation (One cycle = 0 V → 150 V → 0 V) of each UPM. The measured results show a slightly different trend than that predicted. It appears that deflection is increasing at a constant rate in the measurements, but the analytical values show a clear decrease in slope as the radius increases. Actuators with higher radii PZT need to be tested to find where the slope changes. A maximum central displacement of 8.5 μm at 150 V was measured using the 4.4 mm dia UPM. To test repeatability of the UPM fabrication procedure, three UPMs (PZT dia—4 mm) were fabricated in the same batch and subjected to 150 V. The central displacement measured was 7.41 ± 0.36 μm. The primary difference between the actuators was the adhesive thickness, which was measured to be 12.6 ± 3.7 μm. Automating the process of applying epoxy might be a good solution to improving repeatability. 4.2. Valving Figure 8 shows that the microvalve behaviour can be predicted with good accuracy using both the analytical and numerical models. This, of course, is dependent on our assumption that the spacing of the microvalve has decreased to 3 μm due to tightening of the holder bolts. This was verified with reasonable confidence by observing the displacement of the UPM without any fluidic pressure. The UPM did not move downwards after 3 μm meaning that it contacted the valve seat. This is the valving chamber height. It is seen from Figure 8 that the measured flow rate matches accurately with the numerical model at lower pressures, but diverges at 500 mbar. This is because the roof of the valve-chamber, which is the UPM, deforms outward at these pressures, allowing a higher flow rate to pass. The numerical model did not account for this expansion as the computing time was too high. The analytical formulation predicts a higher flow rate so it is a good conservative metric to use when designing future microvalves. Including fluidic capacitance in the analytical expression will allow the trends of the two curves to match. Proportional control of flow rate in the microvalve is clearly observed in Figure 9 a. A uniform upward shift in the flow rate curve is seen in relation to the pressure differential. The small standard deviation (error bars) shows that there is good repeatability of the valving behaviour. It can also be seen that, at all pressures, a bulk of the valving is done from 0–100 V. The slope changes rapidly from 90 V and the flow rate appears to asymptotically decrease to zero. In the 1000 mbar curve, flow rate decreases at an average rate of 0.95 μL min −1 V −1 till 100 V, but this changes to 0.04 μL min −1 V −1 in the latter half of the curve. It is hypothesized that the UPM membrane contacts the valve-seat at 100 V, but due to surface defects, a large leakage flow is still observed. Increasing the voltage further causes the membrane to flatten itself against the valve seat and close the chamber orifice more efficiently. The leakage behaviour of the microvalve is shown in Figure 9 b. Here, leakage rate is defined as the ratio of closed and open flow rates. Generally, leakage is measured using helium gas but this facility was not available [ 10 ]. A high leakage-rate is observed at low pressures; this is because flow rate at open condition is low, but a constant leakage is always present due to valve seat defects. As pressure increases, open flow rate increases while closed flow rate remains relatively constant, effectively decreasing the leakage-rate. Some methods to decrease leakage include decreasing the valve-seat area by using a knife-edge contact and introducing a soft material like PDMS or a parylene layer to the actuator or valve-seat. To investigate fine control of the flow, voltage increments of 0.5 V were applied to the microvalve at 500 mbar pressure. This was the voltage resolution of the power supply. The resulting curve is shown in Figure 10 a. In the linear regime, a flow rate change of 0.2 μL min −1 was measured. This is dependent on the pressure differential across the microvalve. Higher flow rate changes were measured at higher pressures and vice versa. An important characteristic of any device is its reproducibility. The flow rate behaviour of three different microvalves are shown in Figure 10 b. The dimensions of the microvalves were similar, but there is a clear difference in their behaviour. Although valves 1 and 2 show similar behaviour, valve 3 acts very differently. This is because lower torque was applied while tightening the valve holder bolts in valve 3. Decreasing the torque results in a lower clamping force on the microvalve. This means the valve chamber height is larger than valves 1 and 2. This allows more fluid to pass through the microvalve. Valve 1 and 2 have bolts that were tightened with higher torque, resulting in higher clamping force, leading to a lower valve chamber height and a lower flow rate. This was done so that the flow rate could be measured using the available flow sensor over the entire pressure range of the experiment. The decrease in valve-chamber height is due to the spacer being pushed into the microchannel material, thereby permanently deforming it. Lower clamping force is therefore preferred. The reproducibility is then linked closely to the clamping force applied to the microvalve. The valve chamber height was measured using the maximum UPM displacement at 150 V. The UPMs of valves 1 and 2 deformed by 2.9 μm and 2.8 μm respectively while valve 3 deformed by 5.1 μm. As this is a normally-open microvalve, for applications that require long periods of closed state it is essential that the power consumption be minimal. To obtain proportional behaviour, the microvalve is operated in quasi-static mode with the external power supply consuming 37.5 μW. Similar piezoelectric microvalves have shown a static power consumption of 2500 μW [ 18 ], and 3000 μW [ 10 ]. At least to the authors’ knowledge, no piezoelectric microvalve was found that consumes such low static power in this pressure range."
} | 2,888 |
29676145 | null | s2 | 4,377 | {
"abstract": "Microbially driven nitrate-dependent iron (Fe) oxidation (NDFO) in subsurface environments has been intensively studied. However, the extent to which Fe(II) oxidation is biologically catalyzed remains unclear because no neutrophilic iron-oxidizing and nitrate reducing autotroph has been isolated to confirm the existence of an enzymatic pathway. While mixotrophic NDFO bacteria have been isolated, understanding the process is complicated by simultaneous abiotic oxidation due to nitrite produced during denitrification. In this study, the relative contributions of biotic and abiotic processes during NDFO were quantified through the compilation and model-based interpretation of previously published experimental data. The kinetics of chemical denitrification by Fe(II) (chemodenitrification) were assessed, and compelling evidence was found for the importance of organic ligands, specifically exopolymeric substances secreted by bacteria, in enhancing abiotic oxidation of Fe(II). However, nitrite alone could not explain the observed magnitude of Fe(II) oxidation, with 60-75% of overall Fe(II) oxidation attributed to an enzymatic pathway for investigated strains: Acidovorax ( A.) strain BoFeN1, 2AN, A. ebreus strain TPSY, Paracoccus denitrificans Pd 1222, and Pseudogulbenkiania sp. strain 2002. By rigorously quantifying the intermediate processes, this study eliminated the potential for abiotic Fe(II) oxidation to be exclusively responsible for NDFO and verified the key contribution from an additional, biological Fe(II) oxidation process catalyzed by NDFO bacteria."
} | 395 |
40087150 | PMC11909004 | pmc | 4,379 | {
"abstract": "ABSTRACT Methanogenic archaea were likely among the earliest organisms to populate the Earth, perhaps contributing to the Archaean greenhouse effect; they are also widely discussed as analogues to any potential life on Mars. However, fossil evidence of archaea has been difficult to identify in the rock record, perhaps because their preservation potential is intrinsically low or because they are particularly small and difficult to identify. Here, we examined the preservation potential of a methanogen of the genus Methanobacterium , recently isolated from a low‐temperature serpentinizing system, an environment somewhat analogous to habitats on the early Earth and Mars. Notably, this organism has a cell wall composed of peptidoglycan‐like pseudomurein, which may imply a mineralisation potential similar to that of gram‐positive bacteria. Methanobacterium cells were placed in carbonate, phosphate, and silicate solutions for up to 3 months in order to assess the relative tendency of these minerals to encrust and preserve cellular morphology. Cells readily acquired a thick, uniform coating of silica, enhancing their potential for long‐term preservation while also increasing overall filament size, an effect that may aid the discovery of fossil archaea while hindering their interpretation. Phosphates precipitated from the medium in all experimental setups and even in parallel experiments set up with low‐phosphate medium, suggesting a hitherto unknown biomineralisation capacity of methanogens. Carbonate precipitates did not form in close association with cells.",
"conclusion": "5 Conclusion This study investigated the precipitation of three common minerals found on earth: carbonate, phosphate, and silicate, on a methanogen strain with a novel cell wall structure. The results showed that silica in particular can precipitate in close association with the cell wall and that it creates a close‐fitting coating of precipitate around the cells that increases preservation potential. However, the silica coating was more than three times the diameter of the original filament, and this taphonomic effect could mislead investigators. Care should be taken when interpreting silicified filamentous microfossils in the rock record, as their large size may be artefactual. Furthermore, depending on the stage of silicification, the methanogens had coatings of varying thickness throughout the different phases of mineralisation, with some remaining uncoated even at 3 months. If translated into the fossil record, this heterogeneity may create a misleading impression of morphological diversity in an originally monospecific population. Experiments with phosphate precipitation showed a strong affinity between the methanogen species and phosphates. In all experiments, even those carried out in low‐phosphate medium, phosphates formed on and around the methanogens. Often this was simply due to supersaturation, and the resulting precipitate was an amorphous precursor to apatite (Krajewski et al. 1994 ). However, in many samples, methanogens were closely coated with the phosphatic precipitate, and in particular the methanogens grown in low‐phosphate medium with no mineralising agent added had phosphate precipitates that formed very near to the cell surfaces. We suggest that the negatively charged polymers found in the pseudomurein cell wall, in particular the hydroxyl binding sites for phosphate found in the N‐acetyltalosaminuronic acid of the pseudomurein, act as nucleation sites for phosphate precipitation in addition to cations acting as bridges between the cell wall and anion. Further studies are needed to evaluate the role of methanogenic metabolites formed in solution in accordance with phosphate precipitation.",
"introduction": "1 Introduction Mineralisation is an important step for fossilising microorganisms at a cellular level with a high degree of fidelity. Rapid entombment of microbes in minerals is often essential for their long‐term preservation since it hinders decay and oxidation of the organic matter and can also protect against compaction, mechanical destruction, and molecular degradation (Li et al. 2013 ; Alleon et al. 2016 ). This may occur in pores or fractures bearing mineral‐supersaturated fluids, particularly around hydrothermal systems, or through biomineralisation (Li et al. 2013 ). The mineralisation processes of microbes have previously been studied in both modern environments like hot springs and through experimental taphonomy in the laboratory. Experiments are very useful for increasing our understanding of the mechanisms by which the mineralisation of organic structures can occur. Early taphonomic experiments on microbes were performed using cyanobacteria (Oehler and Schopf 1971 ; Oehler 1976 ). Since then, a whole host of microbes have been subjected to experimental mineralisation, such as gram‐negative bacteria including iron oxidisers, gram‐positive bacteria including \n Bacillus subtilis \n , thermophilic bacteria, and archaea (Ferris et al. 1988 ; Westall 1997 ; Lalonde et al. 2005 ; Schieber et al. 2008 ; Orange et al. 2009 , 2011 ; Kish et al. 2016 ). Mineralisation has been observed to be localised to outer sheaths (Francis et al. 1978 ; Schieber et al. 2008 ), extracellular polymeric substances (EPS) (Francis et al. 1978 ; Orange et al. 2009 ), S‐layers (Kish et al. 2016 ; Miot et al. 2017 ; Orange et al. 2009 , 2011 ), peptidoglycan cell wall polymers (Westall 1997 ), on particulate organic matter (Oehler 1976 ) and as complete permineralization (Oehler and Schopf 1971 ). The fossil evidence for the earliest life on earth might well be expected to include methanogenic archaea. Methanogens were the first archaea recognized as a group separate from other bacteria, and their metabolism was thought to be one of the first viable on the primordial Earth (Burggraf et al. 1991 ). Indeed, putative methanogens have been tentatively identified in ~3.42 Ga hydrothermal vent rocks from the Barberton Greenstone belt, South Africa (Cavalazzi et al. 2021 ) and there is isotopically light methane of presumed biogenic origin in ~3.5 Ga fluid inclusions from the Pilbara Craton, Australia (Ueno et al. 2006 ). This evidence, together with independent phylogenetic studies (Battistuzzi et al. 2004 ; Nitschke et al. 2013 ), suggests that methanogens existed on the Archaean Earth. One of the key differences between archaea and bacteria is their cell wall structure (Figure 1 ). Archaea lack peptidoglycan, a major structural polymer in bacterial cell walls. Many archaea have an outer crystalline layer composed of proteins or glycoproteins with various 2‐dimensional symmetries called an S‐layer, which acts as a molecular “sieve” that cloaks the outside of the microbe (Schultze‐Lam et al. 1996 ). The S‐layer is not unique to Archaea but has been found in some bacterial genera, nor are S‐layers ubiquitous within the archaeal domain. For example, the order Methanobacteriales lacks an S‐layer and instead possesses a layer of pseudomurein, similar in structure to the murein that makes up the peptidoglycan in the bacterial cell wall (Meyer and Albers 2020 ). Both pseudomurein and peptidoglycan have a polysaccharide backbone linked by peptides, but the former is made up of N‐acetylglucosamine (NAG) and N‐acetyltalosaminuronic acid (TAL) whereas the latter is made of NAG and N‐acetylmuramic acid (NAM). They also differ in the type of peptide links between sugars, where pseudomurein contains L‐amino acids and peptidoglycan D‐amino acids. It is these cell wall polymers found in peptidoglycan, pseudopeptidoglycan, and S‐layers that offer the carboxyl, phosphoryl, and hydroxyl sites on which minerals can often nucleate (Ferris et al. 1988 ; Schultze‐Lam et al. 1996 ; Frankel and Bazylinski 2003 ; Konhauser et al. 2004 ; Roberts et al. 2004 ; Benning et al. 2005 ; Orange et al. 2009 ). The differences in the chemical and physical structure of the cell walls may lead to varying degrees of mineralisation, potentially resulting in taphonomic biases (Francis et al. 1978 ; Peel 1988 ). For example, gram‐positive bacteria have been argued to silicify more readily than gram‐negative bacteria because of their thicker peptidoglycan layer, which results in a more negative surface charge (Westall 1997 ). Furthermore, both gram‐positive bacteria and archaea with pseudopeptidoglycan cell walls have the sugar polymer as the outermost layer of their cell wall, making it available to surrounding aqueous chemical reactions. On the other hand, gram‐negative bacteria have an outer membrane that constitutes an additional layer of protection against environmental stress and removes the peptidoglycan from direct contact with its surroundings. In fact, archaea with pseudomurein in place of murein also stain gram‐positive (Boone 2015 ) which could imply that a pseudomurein wall also mineralises differently than S‐layers. S‐layers are also often the outermost layer, but since they are composed of proteins and not saccharides, they will likely interact with minerals in a different way. FIGURE 1 Diagram of cell wall differences between bacteria and archaea. Modified with permission (Swoboda et al. 2010 ). Here we compared and contrasted the extent of mineralisation of carbonate, phosphate, and silicate on a strain of the filamentous methanogen Methanobacterium sp. (close resemblance \n M. oryzae \n ). This strain is of interest for several reasons. Its morphology superficially resembles many filamentous structures, both abiotic and biological, found in the rock record (e.g., McMahon 2019 ; Cavalazzi et al. 2021 ), as well as organic filamentous biomorphs formed in the lab (Cosmidis and Templeton 2016 ; Nims et al. 2021 ). In addition, it was isolated from a low‐temperature serpentinizing environment (Neubeck et al. 2017 ) which may have been important to early life and analogous to conditions on several other planetary bodies (do Nascimento Vieira et al. 2020 ). Furthermore, this genus of archaea belongs to the order Methanobacteriales and has a pseudomurein wall (Meyer and Albers 2020 ). This study aims to analyse the mechanisms of mineralisation on these surfaces as opposed to S‐layers that have been studied previously (Orange et al. 2009 ; Kish et al. 2016 ). The mineralising ions, carbonate, phosphate, and silicate were chosen due to the abundance of these minerals in nature and their association with fossilised microbes throughout Earth's history. Additionally, the serpentinizing environment from which Methanobacterium sp. was isolated is rich in phosphates and carbonates. Phosphatization experiments of various organisms have been carried out previously (Briggs 2003 ; Kish et al. 2016 ; Miot et al. 2017 ) but to our knowledge, none have been carried out on methanogens. Many previous taphonomic experiments have used silica, including studies of the silicification of methanogens (Orange et al. 2009 ), although not those with a pseudomurein cell wall. The analysis of calcification, phosphatization, and silicification in tandem allows for a direct comparison of three mechanisms in preserving microbial morphology and gives insights into how these precipitate in association with a pseudomurein cell wall, comparable in structure to that of bacterial peptidoglycan.",
"discussion": "4 Discussion The purpose of these experiments was to analyze how three common minerals found on Earth's surface would precipitate and interact in association with a methanogen of the genus Methanobacterium . Methanobacterium is especially interesting because it lacks an S‐layer and instead has pseudomurein, a compound similar to bacterial peptidoglycan (Meyer and Albers 2020 ). The pH of all experiments was well buffered and unlikely to change through time, despite the addition of external mineralising agents. Previous work has attempted to define the pH tolerance of this species of methanogen, without success, since the buffer is so strong (Stephens 2023 ). In all experiments, mineral precipitation onto cell walls began immediately on some filaments while others remained uncoated well into the third month. This has been noted before (Lalonde et al. 2005 ; Orange et al. 2009 ; Kish et al. 2016 ; Gaboyer et al. 2017 ) and it has been suggested that uncoated cells are still viable but perhaps in a dormant state in which mineralisation is suppressed (Orange et al. 2009 ). Dormancy may have been induced by nutrient starvation since neither headspace gases nor other nutrients were replenished after the first day of mineralisation. In the majority of samples, many filaments remained intact and turgid, further suggesting that they were still viable. Several had wrinkles on their surface indicating the beginnings of lysis (Figure 5a ) and the detachment of the cytoplasm from the cell wall could be seen in TEM (Figures 4g , 5g ). A study by Katsen‐Globa et al. ( 2016 ) on the methods of HMDS drying showed that around 3 min in HMDS does not cause any structural change in the cells. The cells here were only kept for 30–60 s, so artefacts of the TEM drying process do not account for the observed deformation. The minute filaments seen in several samples (Figures 4e , 5f ) are notably longer and thinner than the methanogens. These structures are most likely fimbriae. Fimbriae have several different possible functions, among them the formation of biofilms. Similarly to cyanobacteria, biofilm production may be a stress response to adverse conditions such as starvation or the presence of minerals in solution (Cassier‐Chauvat and Chauvat 2014 ). However, these structures were very abundant in the low phosphate samples with added phosphate mineralising agent where cells were likely less stressed than in other samples. Fimbriae have also been observed in other archaea (Thoma et al. 2008 ) where they produce large fimbrial networks used to adhere to solid surfaces. Additionally, they have been noted in the Methanobacterium genus (Boone 2015 ). 4.1 Mineralisation Mechanism and Morphology: Carbonate and Phosphate The introduction of a calcium chloride solution into the medium reacted with bicarbonate and phosphate already present in the form of buffering agents in the following way: \n (1) \n 2NaHCO 3 + CaCl 2 ↔ CaCO 3 ↓ + 2NaCl + CO 2 + H 2 O \n \n \n (2) \n 2KH 2 PO 4 + 3CaCl 2 ↔ 2KCl + Ca 3 PO 4 2 ↓ + 4HCl \n \n \n (3) \n Na 2 HPO 4 + CaCl 2 ↔ 2NaCl + CaHPO 4 ↓ \n \n The reaction that occurred due to the addition of the phosphate mineralising agent was the same as Equation ( 2 ) above, whereas the one from the carbonate mineralising agent was as follows: \n (4) \n Na 2 CO 3 + CaCl 2 ↔ CaCO 3 + 2NaCl \n \n Considering the high concentration of Ca 2+ added (3.3 g/L) coupled with the concentration of the bicarbonate (2.9 g/L) and phosphate ions (0.56 g/L) from the buffers, the resulting solution is supersaturated. This was evidenced by the fact that the medium turned cloudy immediately when calcium chloride was added. The EDX spot analysis of morphologically alike granular sheet‐like precipitates in all the phosphate and carbonate samples showed the presence of Ca, P, O, and C, as well as minor amounts of Na (Figure S1 ). The broad peaks in the XRD pattern show a poorly crystalline calcium phosphate precipitate (Figure 3b ) which is likely a precursor to apatite (Krajewski et al. 1994 ). Poorly crystalline globular precipitates have also been observed in precipitation experiments of phosphate associated with bacteria (Lucas and Prevot 1985 ) but never with methanogens. The EDX results from the large, elongated crystals in carbonate experiments (Figure 4c ) indicate C, Ca, and O, and the XRD results show a calcite lattice produced by these crystals (Figure 3a ; Figure S1 ). As previously mentioned, the (002) and (211) crystal lattice planes of hydroxyapatite are also visible. The less crystalline nature of the precipitate giving rise to these peaks is evidenced by the lower intensity when compared to the well‐defined peaks of the calcite diffractogram due to a more random scattering of X‐rays by materials lacking ordered crystal lattices. The mechanism of calcite versus apatite precipitation has been referred to as a “switch” with local pH changes driving one precipitation over the other (Briggs 2003 ). Methanogens are known to induce calcium carbonate precipitation due to the removal of CO 2 in their metabolic process, thus locally increasing the pH (Roberts et al. 2004 ; Visscher and Stolz 2005 ). However, since calcite also precipitated in control samples without methanogens, it is difficult to determine the effect that methanogenic CO 2 fixation may have had on the precipitation of calcite. Comparing the XRD patterns of the abiotic controls with those of the 1‐month samples containing methanogens (C1Mcontrol/P1M control and C1M/P1M, Figure 3a,b ), the phosphate signals (002) and (211) are more clearly defined in samples containing methanogens. This suggests that the presence of methanogens had an effect on the crystallinity of the phosphate precipitate, whereas in their absence, the supersaturation alone caused the precipitation of amorphous calcium phosphate. Based on the solubility constants of calcium phosphate and calcium carbonate, it is likely that the latter, being less soluble, would precipitate first (Aylward and Findlay 1974 ). In several experiments, calcite crystals are overlain by phosphate precipitates (Figure 4c ). On the other hand, the presence of biology can interfere with the normal chemical equilibrium, and phosphates may precipitate first due to the presence of nucleation sites on cell walls (Figure 4d ). Notably, methanogens grown in low‐phosphate medium with added carbonate as a mineralising agent still had phosphate precipitates around them (Figure 4d,e ). This could be because small amounts of phosphate inhibit the precipitation of calcite (Langerak et al. 1999 ), but calcite crystals were also present in these samples. The calcium from the added mineralising agent could have reacted with phosphate in the medium despite the low concentration. However, both normal medium and low‐phosphate experiments with only methanogens and no mineralising agent also showed extensive precipitation of phosphates (Figure 7c–f ; Figure S3 ). Precipitation began already in the control grown for just 2 months but was much more widespread in the sample grown for 1 year. In both, the phosphatic precipitates formed the classical spheres and also coated the methanogens (Figure 7d–f ). Blank samples with only medium produced no precipitates. Once again, this points towards a methanogen‐driven mechanism for the precipitation of phosphates. Methanogen‐induced phosphate precipitation has, to our knowledge, never been documented. Studies have indicated that the precipitation of phosphates can be controlled by prokaryotic cells and is not always merely passive (Benzerara et al. 2004 ). This latter conclusion is partly drawn on the orientation of apatite crystals and their close association within the cell wall and on the interior of bacteria, a phenomenon that was not observed here. The close‐fitting coating of phosphates on and around the methanogens in many samples (Figures 4a,d,e , 5e , 7e ) suggests a surface nucleation mechanism. In the case of bacteria, precipitation onto the cell wall is thought to be related to the presence of charged functional groups such as carboxyl, phosphoryl, and hydroxyl found in the murein (Beveridge and Murray 1980 ; Ferris et al. 1988 ). Additionally, the negative surface charge of bacterial and archaeal cell walls can attract cations such as calcium in solution that act as a bridge for the negatively charged phosphate anions, thus facilitating precipitation. Peptidoglycan in bacterial cell walls is known to bind covalently via phosphodiester linkages to wall teichoic acids (WTA) (Swoboda et al. 2010 ). These cell wall polymers are found in gram‐positive bacteria, and their function, though not completely understood, is vital to the survival of bacteria (Esko et al. 2009 ; Swoboda et al. 2010 ). The bonding site of the phosphodiester linkage is the C‐6 hydroxyl of the N‐acetyl muramic acid sugars. Similarly, the same hydroxyl site on the N‐acetyltalosaminuronic acid of pseudopeptidoglycan could provide a binding site for phosphates (Figure 8 ). Consequently, the pseudopeptidoglycan cell wall of archaea may well be primed to bind to phosphate groups, which explains why phosphates precipitated on and around the methanogens in all samples, including those grown in low‐phosphate medium and controls with only methanogens and no mineralising agent added. FIGURE 8 The wall teichoic acids (WTA) in bacteria bind to the C6 of the N‐acetylmuramic acid in peptidoglycan via a phosphate group in the linkage unit. Similar potential binding sites are found on the C6 of the N‐acetyltalosaminuronic acid in pseudopeptidoglycan. This hypothesis predicts that precipitation would not occur on fimbriae, which have no surface charge or cell wall structure. Our observations satisfy this prediction. Only once was it possible to see a fimbria coated in precipitate, but this sample was embedded in the calcite grain and not coated in the granular phosphatic precipitate often seen coating the methanogen cells. The possibility for acidophile and hyperthermophile archaea to passively induce precipitation of phosphates has previously been documented (Kish et al. 2016 ; Miot et al. 2017 ). Kish et al. ( 2016 ) outline a stepwise mechanism of the nucleation and precipitation of Fe‐phosphates, similar to what is observed here. The initial stage is nucleation onto the cell surface, forming small patches of precipitate (Figure 4h ) that multiply and, binding together, form a coating all the way around the methanogen (Figure 5h ). The archaea that Kish et al. ( 2016 ) worked with have an S‐layer, and it was possible to observe the hexagonal surface layer of this structure. In the present study, the interaction is presumably with the pseudomurein. Keighley et al. ( 2018 ) proposed a comparable mechanism for the formation of phosphatized globules in Eocene oil‐shales. Following the death or lysis of cells and the release of P into waters, an apatite precursor may precipitate in the vicinity of the cell wall. Diagenetic alteration will transform the precursor into an apatite mineral and leave either a hollow pseudomorph of the cell wall or a mould of its interior. Of these two scenarios, the hollow void was observed here (Figure 5h ) whereas later infilling of the void was not, though this may be just a question of time. Another possible mechanism for the precipitation of phosphates is through passive precipitation in association with metabolites. The methanogens may secrete a variety of products that can react with compounds and induce precipitation of minerals (Frankel and Bazylinski 2003 ). This requires further investigation. A review by Krajewski et al. ( 1994 ) concludes that cells do not constitute a preferential substrate for the precipitation of apatite but that microbes do influence the concentration of reactive phosphate in sediments. The precipitates coating the methanogens in carbonate and phosphate samples were very irregular, or globular, and did not outwardly reflect the morphology of the encapsulated microbe (Figures 4a and 5b ). Interestingly, in experiments carried out in low‐phosphate buffer medium with added carbonate, the methanogens were sometimes quite uniformly coated in phosphatic precipitate, as seen in Figure 4d,e , though the coating was very rough in texture. Some filaments were initially coated in this granular precipitate and then embedded in the larger calcite crystal that formed around them (Figure 4d ). The precipitation of apatite followed by the overgrowth of calcite crystals has been previously documented (Briggs 2003 ). Fossil microbes in calcite may also be coated in different minerals, such as iron oxides, before being encapsulated in calcite (see possible examples in Trewin and Knoll 1999 ). 4.2 Mineralisation Mechanism and Morphology: Silicate Small grains of silica precipitate could be seen on the surface of the methanogens already at 2 weeks, which could be the result of nucleation or passive deposition of nanometre‐sized grains. Many cells were already coated in a relatively thick outer layer (Figure 6a,d ) whereas others, including fimbriae, were somewhat embedded in a silica precipitate rather than coated. However, only the cells had precipitate coatings that covered the whole filament and reproduced the initial morphology. As with the phosphatic precipitates, the coating of cells was heterogeneous, with some covered in a full jacket (Figure 6a,d,e ), others only partly coated (Figure 6c ) and others completely uncoated, even at 3 months (Figure 6e ). Gaboyer et al. ( 2017 ) noted similar tendencies in their experiments on the gram‐negative bacterium Yersinia up to 6 months. Live/dead staining in their study indicated the presence of still‐living cells able to resist encrustation by bacteria. In our experiments, some methanogens showed signs of lysis already at 2 weeks (Figure 5a ) and 1 month (Figure 4g ), but others remained turgid and whole well into 3 months (Figures 4e and 6e ). Furthermore, Gaboyer et al. also observed silica precipitation in an amorphous phase only on the outside, never proceeding to the inside of the cells, something which also did not occur here. Previous experiments on the silicification of methanogens have demonstrated the possibility of silica precipitation on the inside of the cell as well, but these were carried out for up to a year (Orange et al. 2009 ). TEM images reveal that the silica precipitation was not in direct contact with the cell wall but rather slightly removed from it (Figure 6f ). This has previously been observed and was suggested as a possible repulsion effect of silica by the cell (Orange et al. 2009 ). Silica is formed as a colloidal precipitate due to the addition of sodium silicate in a buffered solution of pH 7.4 to a final supersaturated concentration. The silica therefore nucleates in solution, followed by the formation of an amorphous silica precipitate (Maryani et al. 2018 ). Protons on the surface of colloidal silica often dissociate into solution, leaving a high negative charge and where it comes into contact with the negatively charged cell wall, it is repelled. Phosphate anions, on the other hand, are more likely to nucleate directly on the surfaces of the cells due to the presence of calcium ions clustered around the cell wall that act as a bridge between the negatively charged surface of the cell and the phosphate (Orange et al. 2009 ; Kish et al. 2016 ). Considering the immediate formation of amorphous silica in solution, the cations do not act as nucleation sites. Precipitation of calcium silicate would, in any case, require a higher pH (10–12) (Ntafalias and Koutsoukos 2010 ). Furthermore, experiments with phosphate mineralising solution contain more calcium ions from the added calcium chloride than do silica experiments, where the only calcium is that present in the medium. Nevertheless, experiments with only methanogens and no added mineralising agent also produced calcium phosphates, possibly because of the cation bridge effect. The solubility of phosphate is lower than that of silicate, and at pH 7, phosphates can readily precipitate as calcium phosphate. There may also be another unknown mechanism with which methanogens remove phosphates from solution purposefully, since phosphorus has been shown to inhibit methanogenesis (Mancipe‐Jiménez et al. 2017 ). This remains to be researched in the future. The presence of methanogen‐shaped holes in some samples (Figure 7a,b ) shows the close association between microbes and precipitate, a phenomenon previously noted between bacteria and calcite (Banks et al. 2010 ). Several studies have noted that no significant differences could be discerned in the silica precipitation rate between samples containing methanogens (Orange et al. 2009 ) or bacteria (Yee et al. 2003 ) and sterile controls. In some cases, there may be a slight increase in rate with microbes or organic matter present, but that after the initial coating, the silicification proceeds abiotically (Benning et al. 2003 ; Orange et al. 2009 ). Evidently, at high silica concentration, there is such a strong drive for polymerisation, homogeneous nucleation, and precipitation that any effect of microbial catalysis is negligible. The Methanobacterium sp. strain used here resisted silicification for long enough to retain its morphology when it was eventually coated. Silicification was the most efficacious in forming uniform coatings around the methanogens, which can be particularly well seen in Figure 6c . The texture resembles that of silicified microbes in modern hot springs from New Zealand (Benning et al. 2005 ). The coating increases the overall diameter by a factor of > 2 (Figure 6a ) and ultimately even more, from 250 nm up to 1.4 μm (Figure 6e ). This result may inform the interpretation of microbial fossils: a mineral filament with a diameter of more than 1 μm can be produced by an organism almost 6 times smaller. Fossil specimens of cyanobacteria have demonstrated this bias, where the increased diameter of the mineral leads to misidentification of the taxon (Peel 1988 ). Varying degrees of size are also present in the same group (Figure 6e ) which could be interpreted as different species of microbes in a single colony, but here we show that the same microbe can create quite differently sized casts depending on the timing of initial mineral precipitation. The question remains whether or not the interior of the mineral coat better reflects the true size of the microbe. On measuring the size of cell cross‐sections in TEM images, the diameter is often around 250 nm, in keeping with the cell size of the original methanogen. However, occasionally the diameter of the methanogen cross‐section plus the halo of space created around it had a diameter of up to 450 nm. This occurred particularly commonly in the silica experiments (Figure 6f ) but some phosphatised samples had a similar internal cavity of up to 350 nm. Were the cells to vanish and leave behind only a cavity or mould in the precipitate (Figure 7a,b ), the diameter measured would falsely indicate a larger microorganism than a methanogen. TEM images that show a circular or oval‐shaped halo can also be seen in similar experiments on the silicification of gram‐negative bacteria (Westall 1997 ). In any case, most studies today do not rely solely on morphological characteristics as an indicator of microbial life. Detailed large‐ and small‐scale analysis of the geological context of fossils is necessary (Javaux 2019 ) as well as the presence of concrete geochemical evidence of life‐like isotopes (Lepot et al. 2013 ) or biomolecules (Alleon et al. 2019 ), among others. Perhaps similar structures are found in natural microbialites where the initial stages of fossilization have been observed (Benning et al. 2005 ; Couradeau et al. 2013 ; Li et al. 2013 ). In ancient materials, phosphate minerals often occur with both carbonates and early diagenetic silica, but we are unaware of reports that closely match our experimental observations of individual mineralised microorganisms. In the future, fieldwork is necessary to ascertain whether or not the mineralisation processes in contemporary natural environments may unfold along similar lines to those seen in the experiments presented here. Further work to explore diagenetic/low‐grade metamorphic alteration of the materials produced in our experiments would also be fruitful."
} | 7,966 |
38647921 | PMC10992134 | pmc | 4,380 | {
"abstract": "In modern societies, the accumulation of vast amounts of waste Li-ion batteries (WLIBs) is a grave concern. Bioleaching has great potential for the economic recovery of valuable metals from various electronic wastes. It has been successfully applied in mining on commercial scales. Bioleaching of WLIBs can not only recover valuable metals but also prevent environmental pollution. Many acidophilic microorganisms (APM) have been used in bioleaching of natural ores and urban mines. However, the activities of the growth and metabolism of APM are seriously inhibited by the high concentrations of heavy metal ions released by the bio-solubilization process, which slows down bioleaching over time. Only when the response mechanism of APM to harsh conditions is well understood, effective strategies to address this critical operational hurdle can be obtained. In this review, a multi-scale approach is used to summarize studies on the characteristics of bioleaching processes under metal ion stress. The response mechanisms of bacteria, including the mRNA expression levels of intracellular genes related to heavy metal ion resistance, are also reviewed. Alleviation of metal ion stress via addition of chemicals, such as spermine and glutathione is discussed. Monitoring using electrochemical characteristics of APM biofilms under metal ion stress is explored. In conclusion, effective engineering strategies can be proposed based on a deep understanding of the response mechanisms of APM to metal ion stress, which have been used to improve bioleaching efficiency effectively in lab tests. It is very important to engineer new bioleaching strains with high resistance to metal ions using gene editing and synthetic biotechnology in the near future.",
"introduction": "Introduction to biofilm It is shown that the long-distance interaction and adhesion between bacteria and mineral surface are very important for the formation of biofilm in bioleaching systems (Hall-Stoodley et al. 2004 ; Liu 2020 ). The adhesion of bacteria starts and strengthens the bioleaching process, which is closely related to the leaching efficiency (Li et al. 2019 ; Li et al. 2017 ; Diza et al. 2018 ; Zhu et al. 2003 ). For example, the efficiency of the bioleaching of chalcopyrite can be improved by directly strengthening the adhesion of A. ferrooxidans (Feng et al. 2020 ). When microbial cells attach to the surfaces of solid particles, they will secrete EPS, which include polysaccharides, proteins, extracellular DNA, fibrin, lipids, and complex metal ions (Govender and Gericke 2011 ). EPS will embed the bacteria to form a three-dimensional polymerization network, namely, biofilm (Besemer et al. 2007 ). Biofilm is a micro-ecological environment that can provide an enclosure and protection for microbial community (Flemming and Wingender 2010 ; Flemming et al. 2007 ). Thus, most microorganisms in bioleaching systems live in a biofilm (Schippers et al. 2013 ; Castro et al. 2016 ). Compared with planktonic cells, the biofilm provides a much higher volumetric biomass density and a much more acidic local pH, and the synergy between various strains in a biofilm consortium makes it able to tolerate and survive in harsh bioleaching environments better (Ruiz et al. 2008 ; Vera et al. 2003 ; Hall-Stoodley et al. 2004 ; Jasu et al. 2021 ). In an APM biofilm, the sessile bacteria have higher metabolic activities, and the volumetric density of sessile biomass can be 5–6 orders of magnitude higher than that of planktonic cells (Santegoeds et al. 1998 ). It has been reported that an APM biofilm can play a very important role in a WLIBs bioleaching process (Liu 2020 ). Therefore, it is necessary to study the key influencing factors to obtain efficient biofilms for WLIBs bioleaching process."
} | 946 |
38359900 | PMC10902815 | pmc | 4,381 | {
"abstract": "Metabolomics is a\npowerful tool for uncovering biochemical diversity\nin a wide range of organisms. Metabolic network modeling is commonly\nused to frame metabolomics data in the context of a broader biological\nsystem. However, network modeling of poorly characterized nonmodel\norganisms remains challenging due to gene homology mismatches which\nlead to network architecture errors. To address this, we developed\nthe Metabolic Interactive Nodular Network for Omics (MINNO), a web-based\nmapping tool that uses empirical metabolomics data to refine metabolic\nnetworks. MINNO allows users to create, modify, and interact with\nmetabolic pathway visualizations for thousands of organisms, in both\nindividual and multispecies contexts. Herein, we illustrate the use\nof MINNO in elucidating the metabolic networks of understudied species,\nsuch as those of the Borrelia genus,\nwhich cause Lyme and relapsing fever diseases. Using a hybrid genomics-metabolomics\nmodeling approach, we constructed species-specific metabolic networks\nfor three Borrelia species. Using these\nempirically refined networks, we were able to metabolically differentiate\nthese species via their nucleotide metabolism, which cannot be predicted\nfrom genomic networks. Additionally, using MINNO, we identified 18\nmissing reactions from the KEGG database, of which nine were supported\nby the primary literature. These examples illustrate the use of metabolomics\nfor the empirical refining of genetically constructed networks and\nshow how MINNO can be used to study nonmodel organisms.",
"conclusion": "Conclusions Here, we introduce MINNO, a new software tool that allows researchers\nto integrate genomic and empirical metabolomics data into a single\nsoftware environment in order to build and refine metabolic networks.\nWe illustrate the utility of this tool for identifying missing reactions\nwithin multiple metabolic pathways for Borrelia species. Using MINNO, we identified 18 missing reactions from the\nKEGG database, of which nine were supported by the primary literature.\nThe remaining reactions show good homology as in the NCBI-RefSeq database\n( Table S1 ). MINNO provides a tool that\ncan be applied to any organism to systematically refine or investigate\nmetabolic pathways. MINNO was designed to be inherently flexible for\nthese diverse applications and support a wide range of input formats.\nWe anticipate that it will be a useful asset for analyzing genome-wide\nknockouts, studying novel organisms that are divergent from typical\nmodel organisms, metabolic flux analysis, and visualization of metabolic\nnetworks.",
"discussion": "Results\nand Discussion Strategy: Network and Data Visualization\nUsing MINNO The MINNO visualization tool facilitates both\nthe investigation and\nunderstanding of the complex interplay between genotypic and phenotypic\nfeatures in omics data. MINNO is a JavaScript-based web application\nthat is compatible with Google Chrome and Mozilla Firefox browsers.\nIt uses the D3.js JavaScript library to create dynamic interactive\nvisualizations in web browsers. 30 The tool\ncan load files, such as network files and data files, in JSON, XML,\nand CSV file formats, while it exports data in JSON, XML, PNG, and\nSVG formats for multiple applications. It has numerous built-in features\nthat facilitate the creation of detailed network visualizations without\nthe need to switch from multiple editing software tools. More details\nabout MINNO can be found in the user manual that includes a tutorial\ndeveloped for users to familiarize themselves with many of the tool’s\nfeatures. MINNO is available open-source (under the MIT open-access\nlicense) at www.lewisresearchgroup.org/software . MINNO comes with\n66 base metabolic pathways from the KEGG database, 31 covering all primary metabolic pathways that can be combined\nto build large-scale metabolic networks that include user-added reactions\nand features. Users can then superimpose an organism’s known\nmetabolic pathway data from the KEGG database on these base metabolic\npathways without the need to rebuild a network from scratch for each\norganism studied. The tool can also access metabolic network models\nfrom the Biochemical, Genetic, and Genomic (BiGG) database and the\nNCBI-Reference Sequence (RefSeq) database (available at https://www.ncbi.nlm.nih.gov/refseq/ ). 32 The tool accepts multiomics data,\nsuch as metabolomics, proteomics, and fluxomics data, which can be\nintegrated and visualized on the nodes and edges of the metabolic\nnetwork. MINNO utilizes empirical data to facilitate the identification\nof missing reactions by providing users with the ability to investigate\nreactions pathway-by-pathway or by individual modules. The concept\nof modularity plays a crucial role in this process. Metabolic networks\nexhibit modularity as a network property, wherein a module or pathway\nconsists of densely interconnected nodes compared to connections between\ndifferent modules. 33 , 34 This modular structure enables\nthe detection of missing reactions within metabolic networks by ensuring\nthat nodes within each module are interconnected with either each\nother or the surrounding environment. The concept of modularity is\na fundamental aspect of metabolic networks and can be applied to metabolic\nnetworks of any species. However, except for a handful of model organisms,\nthere are thousands of understudied species that have poorly constructed\nmetabolic networks due to homology mismatch issues. The KEGG database\ncurrently includes over 8794 species along with their respective metabolic\npathways. 31 By providing access to this\nextensive information, MINNO allows users to refine metabolic networks\nand explore individual species or interactions among multiple species. Network Refinement Strategy In this example, we used\nMINNO for our metabolic network refinement strategy to understand\nmetabolic differences among related microbial pathogens ( Figure 1 ). In the context\nof microbial growth, our strategy involves first culturing microbes in vitro and then sampling the cultures over specific time\nintervals so that metabolite intensities can be recorded as a function\nof time ( Figure 1 A).\nThe MINNO visualization tool takes metabolic base/ortholog network\nlayout data from the KEGG database and genetic annotation data from\nNCBI Reference Sequence database (NCBI-RefSeq) to generate the organism’s\nspecific metabolic pathway ( Figure 1 B). This approach facilitates the identification of\npotential missing reactions in the organism’s metabolic network,\nshown as gray edges. The user can then incorporate temporal metabolite\nintensity profiles and intra- or extracellular data onto the network\nto infer missing reactions by considering boundary fluxes and, if\navailable, the isotope labeling pattern, without resorting to complex\nmathematical modeling ( Figure 1 C). In this figure, the dashed links represent the missing\nreactions inferred by the user. Figure 1 Schematic representation of the metabolic\nnetwork refinement method.\n(A) Omics data profiles are produced for species of interest. (B)\nUsing a genetic database (NCBI refSeq) and MINNO, organism-specific\nmetabolic networks are constructed. (C) Refined metabolic networks\nare created using (B) organism-specific metabolic networks and (A)\nexperimentally generated omics data. Users can then search for genes and proteins corresponding to missing\nreactions using experimental or bioinformatics methods. This approach\nallows users to refine the networks of under-studied species such\nas Borrelia spp. and find the necessary\nreactions to explain the metabolic profiles that were missing from\ntheir original annotated networks. MINNO can also be used for multiomics\ndata integration as it can incorporate gene, protein, metabolite,\nand flux data on the same metabolic network. MINNO User Interface Figure 2 highlights\nsome key features of the web-based\napplication MINNO. The built-in menu is located on the left side of\nthe screen, where users can select base metabolic pathways from the\nKEGG database. Later, users can customize the network by dragging\nand aligning nodes in the network. Additionally, the tool shows organism-specific\nmetabolic networks using genetic annotation information from NCBI-RefSeq\ndatabase, which show annotated reactions as dark nodes and edges,\nwhile unannotated/missing reactions are depicted by light gray nodes\nand edges. This enables the user to determine the potential missing\nreactions after experimental data are uploaded onto the network. Figure 2 MINNO\nweb-browser interface with key features highlighted in red. Refining Nucleotide Metabolic Pathways Using\nMINNO We used MINNO to perform a metabolic network refinement\nanalysis\nof three Borrelia species known to\ncause Lyme and relapsing fever diseases. By leveraging the modularity\nconcept of metabolic networks and employing boundary flux analysis,\nwe were able to identify the missing reactions from the KEGG database\nfor these species. The purine metabolism of B.\nburgdorferi in the KEGG database is fragmented, as\ndepicted by solid links in Figure S1A ,\nas it lacks the classic purine salvage pathway. 35 The consumption of both adenine and guanine by B. burgdorferi suggests the presence of purine transporters,\nwhich has recently been reported in the literature. 36 − 38 By analyzing\nthe boundary fluxes of cultured cells, we have identified missing\nreactions in purine metabolism from the KEGG database, indicated by\ndashed links in Figure S1A . The thickness\nof the edges represents boundary flux values, while the internal curved\narrows pointing toward the biomass indicate the flux directed to DNA\nand RNA synthesis. This accounts for 35% for adenine and thymine and\n15% for guanine and cytosine, based on the A, T, G, and C composition\nin DNA of B. burgdorferi . In contrast,\nthe pyrimidine pathway for B. burgdorferi is relatively less fragmented in the KEGG database, as shown by\nsolid links in Figure S1A . However, the\nboundary flux profile of this species suggests the presence of pyrimidine-nucleoside\nphosphorylase ( PnP ) based on the production of thymine\nfrom thymidine, as shown in Figure S1A ,\nwhich is missing in the KEGG database. Additionally, B. burgdorferi lacks ribonucleotide reductase, an\nenzyme responsible for converting ribonucleotides (for RNA synthesis)\ninto deoxyribonucleotides (for DNA synthesis). 35 According to our data, PnP salvages deoxyribose\nsugars from thymidine for DNA synthesis. In summary, we used MINNO\nand empirical metabolomics data to identify eight reactions that are\nmissing from the KEGG database. Subsequent publications have confirmed\nsix of these missing purine reactions ( Table S1 ). Furthermore, MINNO predicted four missing pyrimidine metabolism\nreactions from the KEGG database, all of which are supported by the\nprimary literature ( Table S1 ). Metabolic Distinction\nbetween Borrelia Species Causing Lyme\nDisease and Relapsing Fever To better\nunderstand metabolic differences between Borrelia species, we focused on refining the metabolic networks of Borrelia species associated with relapsing fever: B. parkeri and B. turicatae . These species share genetic similarities, and as expected, their\nboundary flux profiles exhibit similarities as well 24 ( Figures 3 and S1B,C ). Figure 3 Heatmap showing the temporal\nprofile of selected metabolite intensities\nacross three Borrelia species, bbu: B. burgdorferi , bpk: B. parkeri and btu: B. turicatae with respect\nto the growth medium at 0 and 72 h. The change in metabolite intensity\nacross two different time points (0 and 72 h) has a p -value <0.01, and the row z -score is shown through\nthe color legend. Similar to the purine\nmetabolism of B. burgdorferi , purine\nmetabolism of both B. parkeri and B. turicatae is fragmented, as\nshown as solid links in Figure S1B,C . However,\nunlike B. burgdorferi , both possess\nthe classic purine salvage pathway. They both consume adenosine and\nadenine, and any excess purine is excreted as hypoxanthine. Interestingly,\nneither of the isolates studied has an annotated ribonucleotide reductase\nin the KEGG database. However, based on their boundary flux profiles,\nwe anticipate that both B. parkeri and B. turicatae harbor a ribonucleotide reductase ( rnr ) ( Figure S1B,C ). Metabolic Similarities\nbetween Borrelia Species Causing Lyme\nand Relapsing Fever Diseases It is\nworth noting that the three Borrelia species studied here also shared metabolic similarities. One common\nfeature observed in all three species is the absence of the thyX gene in the KEGG database, as shown in Figure S1 . The thyX gene is\nessential in all three species for providing the necessary deoxyribonucleosides\nrequired for DNA synthesis. Another notable similarity is their deficiency\nin various biosynthetic pathways essential for the production of nicotinate\nand nicotinamide. This deficiency highlights their reliance on salvaging\nprecursors for NAD(P) synthesis from the host or the surrounding environment.\nOur observations revealed that all three Borrelia species consume nicotinamide while excreting nicotinate out of the\ncells, as shown in Figure 4 . Notably, the net excretion of nicotinate exceeds the level\nof nicotinamide consumed for each isolate. This suggests the possible\npresence of nicotinamide-nucleotide amidase ( pncC ). This salvage process also leads to the generation of essential\nmolecules like PRPP and ATP, as well as the accumulation of ammonia. Figure 4 Refined\nnicotinate and nicotinamide metabolism. Solid links are\nannotated in the KEGG database, while dashed links represent orphan\nreactions. Abbr. PRPP: phosphoribosyl pyrophosphate, PPi: diphosphate,\nand NAD+: nicotinamide adenine dinucleotide. The error bars represent\nstandard deviation (sample size n = 3). In summary, we used MINNO to predict six reactions in purine\nmetabolism\nfor B. parkeri and B.\nturicatae that were missing from the KEGG database,\nwith four of these predictions supported by the primary literature\n( Table S1 ). MINNO was also used to identify\ntwo missing reactions in the KEGG database from the pyrimidine metabolism,\nalthough none of them are currently supported by the primary literature.\nHowever, these predictions are supported based on homology matches\nthrough the PGAP pipeline from NCBI-RefSeq. For nicotinate metabolism,\nwe predicted one reaction shared by all three Borrelia species, which is missing from the KEGG database ( Table S1 ). Summary of Functionality and Applications Overall,\nMINNO enables users to refine metabolic networks and integrate multiomics\ndata to provide a system level view of metabolic homeostasis. MINNO’s\nmodular approach, whereby discrete metabolic pathway modules can be\neasily merged together, facilitates the creation of metabolic networks\nin diverse nonmodel organisms. It also allows users to visualize data\non these merged metabolic pathways quickly and easily, without any\ncoding required, facilitating a deeper understanding of complex multiomics\ndata in the context of the broader metabolic system. MINNO can support\na variety of applications, such as FBA visualization to model more\nsophisticated genome-scale behaviors, 39 mapping metabolic architecture in complex microbiome communities, 40 investigating interspecies “cross-talking”\ninteractions, 41 , 42 and determining the molecular\nmechanisms of novel antibiotics. 43"
} | 3,846 |
26226457 | null | s2 | 4,383 | {
"abstract": "In this study, the secondary structure of the major ampullate silk from Peucetia viridans (Green Lynx) spiders is characterized by X-ray diffraction and solid-state NMR spectroscopy. From X-ray diffraction measurement, β-sheet nanocrystallites were observed and found to be highly oriented along the fiber axis, with an orientational order, fc≈0.98. The size of the nanocrystallites was determined to be on average 2.5nm×3.3nm×3.8nm. Besides a prominent nanocrystalline region, a partially oriented amorphous region was also observed with an fa≈0.89. Two-dimensional (13)C-(13)C through-space and through-bond solid-state NMR experiments were employed to elucidate structure details of P. viridans silk proteins. It reveals that β-sheet nanocrystallites constitutes 40.0±1.2% of the protein and are dominated by alanine-rich repetitive motifs. Furthermore, based upon the NMR data, 18±1% of alanine, 60±2% glycine and 54±2% serine are incorporated into helical conformations."
} | 243 |
34730826 | PMC8752119 | pmc | 4,385 | {
"abstract": "Abstract Colonization of land from marine environments was a major transition for biological life on Earth, and intertidal adaptation was a key evolutionary event in the transition from marine- to land-based lifestyles. Multicellular intertidal red algae exhibit the earliest, systematic, and successful adaptation to intertidal environments, with Porphyra sensu lato (Bangiales, Rhodophyta) being a typical example. Here, a chromosome-level 49.67 Mb genome for Neoporphyra haitanensis comprising 9,496 gene loci is described based on metagenome-Hi-C-assisted whole-genome assembly, which allowed the isolation of epiphytic bacterial genome sequences from a seaweed genome for the first time. The compact, function-rich N. haitanensis genome revealed that ancestral lineages of red algae share common horizontal gene transfer events and close relationships with epiphytic bacterial populations. Specifically, the ancestor of N. haitanensis obtained unique lipoxygenase family genes from bacteria for complex chemical defense, carbonic anhydrases for survival in shell-borne conchocelis lifestyle stages, and numerous genes involved in stress tolerance. Combined proteomic, transcriptomic, and metabolomic analyses revealed complex regulation of rapid responses to intertidal dehydration/rehydration cycling within N. haitanensis . These adaptations include rapid regulation of its photosynthetic system, a readily available capacity to utilize ribosomal stores, increased methylation activity to rapidly synthesize proteins, and a strong anti-oxidation system to dissipate excess redox energy upon exposure to air. These novel insights into the unique adaptations of red algae to intertidal lifestyles inform our understanding of adaptations to intertidal ecosystems and the unique evolutionary steps required for intertidal colonization by biological life.",
"conclusion": "Conclusions In this study, an algal genome was isolated and completed along with those of complex epiphytic bacteria using Meta-Hi-C sequencing methods. These methods provide an ideal solution for future genome assembly of impure culture samples and help to unravel the complex relationships between algae and bacteria. Genome structure analysis of N. haitanensis indicated that it exhibited complex functions and evolutionary histories that might be driven by its specific lifestyle and ecology. Integration of genes from bacteria apparently helped compensate for the extensive gene loss in the red algal genomic background that was originally simplified in its ancient ancestor. Multiomics analysis indicated overall that intertidal desiccation resistance and subsequent rapid recovery are not simple adaptations. The limited time intervals of desiccation that occur each day allow cells to make a series of preparations for recovery and repair upon dehydration. To achieve this, N. haitanensis immediately shuts down its photosynthetic system upon water loss to reduce photo-damage and reduce energy consumption from carbon assimilation. Furthermore, cells increase degradation activities and slow the synthesis of macromolecules. Stable mRNA pools are maintained throughout the process, and ribosomal biogenesis and ribosomal genes remain active. In addition, heat shock elements and repair mechanisms ensure the continual synthesis of proteins, along with the stability of DNA and RNA. Moreover, SAM-methyltransfer activities cleverly combine the polyamine and GSH-AA-NADPH antioxidant systems to effectively reverse inevitable oxidative damage and avoid cell death by strengthening the entire cellular antioxidant system. These unique adaptations can achieve a positive balance during a narrow window of opportunity that occurs as a result of the episodic nature of dehydration events. Consequently, the shut down-prepare-protect-repair-recovery scheme is a systematic, effective, and evolutionarily unique strategy allowing N. haitanensis to optimize survival in the intertidal zone and might be a common adaptive feature of intertidal algae.",
"introduction": "Introduction The initial diversification of macroalgae appears to have occurred in the Mesoproterozoic ( Bykova et al. 2020 ). Subsequent fluctuations of Earth’s climate and geologic evolution impacted the formation of intertidal zone habitats. Many organisms would have then been passively dragged from shallow waters to intertidal zones. Those that survived in these zones would be required to overcome many challenges including desiccation, daily and seasonally fluctuating temperatures, high levels of irradiance, and severe osmotic stress ( Wang et al. 2020 ). Around this time, some red algae, which are considered among the earliest members of the Archaeplastida ( Brawley et al. 2017 ), were also introduced to intertidal zones and encountered the harsh conditions associated with them. The surviving red algae would have thus successfully competed in this dynamic and stressful environment for over the next billion years, while acquiring various intertidal adaptive traits. \n Porphyra sensu lato within the Bangiales order of Rhodophyta is a typical exemplar of the adaptive evolution of organisms that transitioned from marine to early intertidal zone habitats ( Yang et al. 2020 ). Evidence supporting this comes from the Bangia-like fossil Bangiomorpha pubescens (dated to 1,198 ± 24 My), which was inferred to live in intertidal zone habitats during ancient eras ( Butterfield 2000 ; Blouin et al. 2011 ). Several aspects of intertidal adaptation can be observed in modern Porphyra . First, Porphyra exhibit adaptations to intertidal recurrent desiccation/rehydration cycling. They have developed the ability to undergo nearly absolute dehydration during air drying (losing up to 95% of their water), but become metabolically active as soon as they are rehydrated from rising tides ( Brawley et al. 2017 ). Second, Porphyra exhibit unique two-generation-altered life histories of gametophyte thalli and sporophyte conchocelis ( Wang et al. 2020 ). The ancestors of Bangiales may have adopted a unique shell-borne conchocelis lifestyle to obtain a stable, aqueous environment to mediate the transition from a water-borne environment to intertidal zones. However, the demand for population diversity and expansion would have required them to enter the intertidal environment and initiate thallus generation. Therefore, thallus is a typical adaptive state for Porphyra under harsh intertidal conditions. Desiccation tolerance (DT) of intertidal algae is similar to the desiccation resuscitation behavior of some plants. They all have the remarkable capacity to survive complete dehydration and come alive when water becomes available ( Xiao et al. 2015 ). However, revival mechanisms differ considerably among organisms. Resurrection plants, including ferns, mosses, and a few angiosperm plants, can tolerate desiccation for days, months, or even years ( Lüttge et al. 2011 ). Responses to desiccation have been intensively studied in resurrection plants, and observed strategies include increasing compatible solutes, strengthening antioxidant activity as a protective mechanism, using xanthophyll cycling to dissipate energy, and activating the abscisic acid (ABA) pathway ( Lüttge et al. 2011 ). In contrast, intertidal macroalgae experience rapid desiccation and rehydration changes as a result of rises and falls in tides that occur once or twice a day. The responses of seaweeds to desiccation have mainly been characterized physiologically. Several studies have investigated the variation in photosynthetic parameters and antioxidant enzyme activities ( Lipkin et al. 2009 ; Lüttge et al. 2011 ; Wang et al. 2020 ), although comprehensive analyses of their recurrent rapid revival mechanisms remain understudied. The acquisition of DT by intertidal algae after encounters of drought stress may have evolved through genome modification and activation of some innate mechanisms that are ancestral to intertidal algae and may be conserved in land plants. Increasing numbers of high-quality Rhodophyte genome sequences have been generated in recent years ( Matsuzaki et al. 2004 ; Bhattacharya et al. 2013 ; Collen et al. 2013 ; Lee et al. 2018 ; Cao et al. 2020 ) and these genomic resources will allow for a better understanding of the evolution of red algae. Consequently, comparative genomic analyses of these species will enable a systems approach for understanding the adaptations of the earliest organisms to intertidal zone habitats. Here, we present a high-quality genome for Neoporphyra haitanensis (redefined from the synonym Pyropia haitanensis in May 2020) that is a typical upper intertidal species. Comparison of this genome against other red algae genomes was used to identify common evolutionary adaptations underlying their unique stress tolerance mechanisms. In addition, a comprehensive analysis of the changes in thalli during desiccation that were related to the revival mechanism was conducted using a multiomics approach. These data were used to propose an evolutionary scenario involving ancestral red algae that were exposed to ecological stress.",
"discussion": "Results and Discussion Metagenome-Hi-C-Assisted Genome Assembly of N. haitanensis and Removal of Assembly Microbial Contamination Marine seaweeds often harbor bacterial populations tightly that cannot be completely removed by physical or antibiotic treatments. These close associations can cause significant obstacles when assembling seaweed genomes ( Nakamura et al. 2013 ). The primary genome assembly of the N. haitanensis double haploid was generated using 19.32 Gb of PacBio long-read sequencing data, resulting in a 99.03 Mb genome comprising 15 scaffolds and 343 contigs ( supplementary table 1 , Supplementary Material online). Bacterial attachment was not found on the surface, inside of the cell wall or between the cell wall and plasma membrane by scanning and transmission electron microscopy (TEM) and staining observations ( supplementary fig. 1 , Supplementary Material online). However, previous research on 16S rRNA sequencing of antibiotic-treated thalli showed that bacteria were still present ( Gu et al. 2020 ). Thus, to increase the accuracy of our genome assembly, we used metagenome-Hi-C (Meta-Hi-C) to separate the contigs of the target species from those of the attached microorganisms. Meta-Hi-C assembly cross-links DNA molecules that are in close physical proximity within intact cells ( Lieberman-Aiden et al. 2009 ; Stewart et al. 2018 ). DNA interactions are stronger within the same DNA molecule than between DNA molecules, thereby allowing for the differentiation of host genome sequences from microbial genome sequences and resulting in species-level deconvolution ( Burton et al. 2014 ). The Hi-C library comprised a total of 27 Gb of clean reads that were paired based on interaction information for different contigs to generate the Hi-C scaffolding data set. The contigs were separated into four groups ( fig. 1 A ), with the first containing five scaffolds with strong intra-scaffold Hi-C interactions and minor inter-scaffold interactions. The five assembled scaffolds comprised the chromosomal region of N. haitanensis . The second group comprised ten scaffolds of strong intra-scaffold Hi-C interactions, and little or no interactions between scaffolds ( fig. 1 A ). Comparison of these sequences against GenBank indicated that they resembled bacterial sequences. The third region was composed of short N. haitanensis unplaced contigs that were not able to cluster based on Hi-C data. The fourth group contained contigs that did not belong to N. haitanensis or to bacterial species (“unknown”) based on our BLASTn strategy ( supplementary data 1, Supplementary Material online). The final N. haitanensis nuclear genome assembly, excluding contigs that were plastid, bacterial, and unknown, was 49.67 Mb, with contig N50 and scaffold N50 of 650 kb and 7.79 Mb, respectively ( supplementary table 1 , Supplementary Material online). Annotation and masking of repetitive elements resulted in 31.61% of the genome being masked ( supplementary tables 2 and 3 , Supplementary Material online). Combining ab initio prediction, homology-based prediction, and PacBio Iso-seq data, 9,496 protein-coding genes were predicted in the N. haitanensis genome ( supplementary table 4 , Supplementary Material online), of which 99.35% were supported by RNA sequencing (RNA-seq) data. Gene annotation completeness according to BUSCO analysis estimated that 85.8% of the core eukaryotic gene sets were found ( supplementary table 5 , Supplementary Material online). Comparing our results with Porphyra umbilicalis ( Brawley et al. 2017 ) and P. haitanensis PH40 ( Cao et al. 2020 ) genome assembly, we achieved a higher contiguity ( supplementary table 6 , Supplementary Material online). The smaller genome size and gene count also could be attributed to the use of Meta-Hi-C to separate non- N. haitanensis sequences. Fig. 1. \n Neoporphyra haitanensis genome features. ( A ) Genome-wide Meta-Hi-C interaction links between the assembled contigs. Purple dots represent Meta-Hi-C interactions between two loci. ( B ) Summary statistics for the N. haitanenssis genome and comparison of genome size, GC content, and protein-coding gene numbers among red algae species. Comparison of all available red algal genomic data, including genomes belonging to the Cyanidiophyceae, Bangiophyceae, and Florideophyceae groups, indicated that the N. haitanensis genome size was intermediate between those of Cyanidiophyceae and Florideophyceae, whereas the genome size and number of predicted genes were markedly lower for the Bangiophyceae genomes. The GC content in the N. haitanensis genome is 70.06%, which is higher than for other red algae ( fig. 1 B ). Compact genes and high GC% are typical for this group, consistent with previous observations in other red algae and thus are possibly an ancestral trait ( Brawley et al. 2017 ). Comparative Genomics Provides Evidence for Adaptive Intertidal Characteristics Phylogenetic analysis of the estimated divergence of Porphyra (Bangiophyceae) and the clade comprising Chondrus crispus and Gracilariopsis chorda (Florideophyceae) from their common ancestor was dated to about 738.5 Ma ( fig. 2 A ). Because all these taxa have intertidal characteristics, the adaptation of red algae to intertidal zones should have occurred before 738.5 Ma. To identify unique features of the Porphyra gene repertoire related to an intertidal lifestyle, the gene family content of N. haitanensis and P. umbilicalis was compared against that of other red algae. The common ancestor of these taxa is predicted to have gained 142 gene families that were associated with several categories ( fig. 2 A and supplementary data 2, Supplementary Material online), as follows: 1) gene families associated with unique shell-borne conchocelis life histories, including the presence of genes encoding carbonic anhydrase (CA), calcium-transporting ATPases, and calcium-dependent protein kinases; 2) genes associated with adaptation to intertidal environments, including redox-associated genes like catalase (CAT), glutathione S-transferase (GST), peroxiredoxin, and thioredoxin, in addition to stress-resistance-related genes, including those encoding germin-like protein, CBL-interacting protein kinase, disease resistance protein, subtilisin-like protease, and lipoxygenase (LOX); and 3) transporter genes, including ATP-binding cassette (ABC) transporters and copper-transporting ATPase. A total of 508 unique gene family gains were identified for N. haitanensis ( fig. 2 A ), including a large number of expansions in antioxidant genes encoding glutamate synthase, GST, and haloperoxidases ( supplementary data 3, Supplementary Material online). Fig. 2. Comparative genomics analysis. ( A ) Comparative evolutionary histories of Neoporphyra haitanensis and other eukaryotic algae genomes. The number of gene families that were acquired (blue) or lost (purple) at each time point in the tree were estimated using the Dollo parsimony principle. The brown-shaded box in the phylogenetic tree represents algae that live in the middle or high intertidal zone and have desiccation adaptation features. ( B ) Venn diagram representation of shared and unique gene families in the N. haitanensis genome in the context of those common to red algae genomes, including those of Porphyra umbilicalis , Porphyridium purpureum , Cyanidioschyzon merolae , and Chondrus crispus . We evaluated the genomes from several red algae with complete annotation information based on shared gene contents ( fig. 2 B ). The N. haitanensis and P. umbilicalis genomes contained many common gene families that are specific for these two species, including 1,229 gene families (1,573 genes) ( supplementary data 3, Supplementary Material online). Identification of Important Relationships between N. haitanensis and Associated Bacterial Populations Based on Meta-Hi-C Analysis Meta-Hi-C assembly resulted in ten superscaffolds with contig N50 3.73 Mb and scaffold N50 4.44 Mb ( supplementary table 7 , Supplementary Material online). CheckM analysis for each of the ten superscaffolds showed eight superscaffolds with more than 98% completeness, whereas Superscaffold7 and Superscaffold8 showed 84.6% and 70.9% completeness, respectively. The estimated contamination in these superscaffolds were less than 1% in six superscaffolds. GTDB-Tk and PATRIC annotation also indicated these superscaffolds were of bacterial origin ( supplementary data 4–6, Supplementary Material online). Whole-genome alignment showed that only some superscaffolds were highly homologous to genomes from known bacterial species ( supplementary fig. 2 A , Supplementary Material online). For example, Superscaffold1 aligned well to the genome of Stappia sp., Superscaffold4 was highly syntenic with the genome of Maribacter sp., and Superscaffold9 exhibited good synteny with the Candidatus Phaeomarinobacter ectocarpi genome ( supplementary fig. 2 A , Supplementary Material online). Other superscaffolds exhibited poor synteny with genomes of known bacterial species. The sequence location mapping of the genes in the superscaffolds included a large number of unannotated sequences ( supplementary fig. 2 B , Supplementary Material online), potentially indicating the presence of a significant amount of genetic variation within the epiphytic bacterial genomes or the potential presence of novel bacterial genomic diversity. We also used the Distilled and Refined Annotation of Metabolism (DRAM) tool to see whether we could improve the annotation of these superscaffolds. Nearly all superscaffolds had full complements of prokaryotic electron transport chain (ETC) complexes, as well as genes in glycolysis pathway, pentose phosphate pathway, citrate cycle, glyoxylate cycle, and dicarboxylate hydroxybutyrate cycle ( supplementary fig. 3 , Supplementary Material online). Together, these results suggested that high-quality genomic assemblies based on Meta-Hi-C methods could inform the discovery of new bacterial species in samples from complex environments. Increasing evidence has suggested that interactions between epiphytic bacteria and seaweeds play key roles in normal algal development ( De Clerck et al. 2018 ). Here, a total of 32,229 genes were annotated among the bacterial scaffolds, but over 68.42% of the bacterial sequences did not exhibit good matches to genomes within public databases. Some cytokinin biosynthesis-associated genes of Ca. P. ectocarpi were identified on Superscaffold9, in addition to numerous genes related to the biosynthesis of biotin, pantothenate, coenzyme A, pyridoxine, and thiamine ( supplementary data 7, Supplementary Material online). Ca. P. ectocarpi has been observed to associate with brown algal Ectocarpus cultures and to harbor numerous transporters that can help assimilate algal metabolites. Their genomes also encode several proteins likely to be involved in the synthesis of algal hormones like auxins and cytokinins, in addition to vitamins ( Dittami et al. 2014 ). Eleven types of phytohormones have been identified in N. haitanensis , including ABA, auxin (IAA), and jasmonic acid (JA) ( Song et al. 2017 ). Most of the genes involved in phytohormone synthesis, however, are not present in the N. haitanensis genome. Therefore, the synthesis mechanisms for these hormones remain unknown ( Mikami et al. 2016 ). Nevertheless, only some of the genes involved in phytohormone synthesis pathways were identified in the ten bacterial superscaffolds ( supplementary data 7, Supplementary Material online). We could not dismiss the fact that epiphytic microorganisms and seaweed hosts could cooperatively synthesize some phytohormones. We found common sequences existed among epiphytic bacterial genomes of N. haitanensis and other Porphyra sequencing data. The PacBio sequencing data from P. haitanensis PH40 and P. umbilicalis were aligned to the bacterial superscaffolds. A clear bias in alignment was observed primarily on Superscaffolds1, 2, 7, 9, and 10 ( supplementary fig. 4 A , Supplementary Material online). A total of 1,090 gene sequences were shared by the bacterial superscaffolds and the P. haitanensis PH40 sequencing data. Among these, 664 gene sequences aligned more than 90% of their length, indicating the presence of numerous complete genes among these shared sequences. In addition, 248 sequences that belonged to annotated genes were also shared among the bacterial sequences from the N. haitanensis assembly and the two Porphyra genome sequencing data ( supplementary data 8, Supplementary Material online). N. haitanensis and P. haitanensis PH40 were collected from different sites of coastal China, and both underwent aseptic treatments before genomic sequencing ( Wang et al. 2020 ), whereas P. umbilicalis had a different taxonomy and a different collection location or time ( Brawley et al. 2017 ); nevertheless, it still harbored similar bacterial sequences. One explanation for this observation is that these sequences are bacterial conserved sequences, or bacterial epiphytes with these sequences universally associated with seaweeds. It is likely, then, that some close interactions exist between them and seaweeds. Enrichment analysis identified significant over-representation ( P < 0.05) of genes within the common 248 bacterial sequences that were associated with the KEGG pathways “membrane transport,” “signal transduction,” and “vitamins.” Furthermore, a large number of transporter genes were identified as enriched that are involved in the transport of Fe 3+ , inorganic phosphorus, polyamines, heavy metals, sodium ions, urea, amino acids, and other compounds. These transporter genes included those within the ABC, BCCT, MFS, and TRAP transporter families. Of the 248 bacterial sequences, 28 were ABC transporter genes. In fact, we observed that among the 1,090 bacterial gene sequences shared between N. haitanensis and P. haitanensis PH40, 10% were ABC transporter genes, 61 of which were from the same bacterial species ( supplementary data 8, Supplementary Material online). The high proportion of ABC transporter genes among the bacterial sequences common to Porphyra and the large number of transporter genes observed in N. haitanensis ( supplementary fig. 5 , Supplementary Material online) and other red algal genomes, including those of P. umbilicalis and G. sulphuraria ( Schönknecht et al. 2013 ; Brawley et al. 2017 ), suggests that transporters may provide a mechanism to rapidly accomplish metabolic adaptation to handle complex environmental stresses. Rapid Adaptation of N. haitanensis to the Dehydration/Rehydration Cycle Once algae entered intertidal zone ecosystems, they were exposed to periods of desiccation/rehydration cycles. Here, we observed that thalli lost more than 95% of their water after desiccation for 30 min. Upon rewetting, thalli recovered 75% of their water content in 10 s, returning to almost normal in 2 min. Less water loss also led to faster recovery ( fig. 3 B ). Intercellular space decreased during severe water loss, leading to cell shrinkage that was accompanied by extensive folding of cell walls. After rehydration for 12 h, cells had recovered from anhydrobiotic states. Across the entire recovery cycle, cell membranes and cell walls were closely connected, and thylakoids were always intact ( supplementary fig. 6 , Supplementary Material online). Even with drying up to 5% relative water content (RWC, 0.17 g ± 0.01 H 2 O/dry weight), almost no Evans blue or TUNEL positive staining within cells were observed, suggesting that N. haitanensis cells do not die during desiccation ( supplementary fig. 7 , Supplementary Material online). Fig. 3. Phenotypic response of Neoporphyra haitanensis to intertidal desiccation and rehydration cycling. ( A ) Overview of desiccation and rehydration cycling of N. haitanensis . Phenotypic response to the desiccation and rehydration process of a single thallus. ( B ) Time course of RWC changes in thalli during a desiccation and rehydration cycle. Multiomics Data Provide Evidence for Adaptation of N. haitanensis to Intertidal Environment Stresses The transcriptional, proteomic, and metabolic responses of N. haitanensis to the dehydration/rehydration process was evaluated to understand their responses to intertidal environment stresses ( fig. 4 A ). A total of 11,471 transcripts, 1,117 proteins, and 597 metabolites were identified, and principal components analysis (PCA) was performed to investigate the global responses at each functional level ( fig. 4 B ). PCA revealed significant changes that occurred during water loss based on the three omics data sets, with great variation compared with the hydration (HD) (3.33 ± 0.10 g H 2 O/dry weight) group. Rehydration responses were much more rapid for the metabolites and proteins than for transcripts, and various levels of separation were observed between the 5% RWC (0.17 ± 0.01 g H 2 O/dry weight) and rehydration (RH)-5 min groups. However, transcripts separated very clearly after just 12 h of rehydration, suggesting that transcriptional recovery from dehydration was slowest among the other data sets. After 12 h of rehydration, proteins in RH-12 h clustered closely with the HD group, which indicated that proteins gradually recovered to a state close to that of HD group. Distances still existed between the RH-12 h and HD groups in transcriptomic PCAs, which indicated that some of the recovering reactions of the transcription process were occurring even after 12 h rehydration, such as “ribosome” and “aminoacyl-tRNA biosynthesis,” which meant that cells continued to prepare for the protein synthesis required for post-stress recovery, which is similar to some plants ( Lüttge et al. 2011 ). The metabolites did not fully recover to the state before desiccation ( fig. 4 C ). Various amino acid metabolism pathways in RH-12 h group were different from the HD group, such as tyrosine, lysine, glutathione, and phenylalanine metabolism, which also indicated that the intermediates of protein biosynthesis were still in an active state. Fig. 4. Multiomics analyses of the adaptations of Neoporphyra haitanensis to the desiccation and rehydration stress of intertidal zone habitats. ( A ) Overview of the multiomics study and the data sets that were used along with detailed descriptions of samples. ( B ) PCA of transcriptomic, proteomic, and metabolomic profiles of individual traits following shifts in dehydration and rehydration cycling. Circles around the same color points indicate biological replicates for each treatment. ( C ) Heatmaps of significantly enriched pathways for differentially variable genes, proteins, or metabolites across comparison groups. ( D ) Clusters of scaled transcript, protein, and metabolite profiles of N. haitanensis undergoing desiccation and rehydration cycling. The normalized DEGs, differentially expressed proteins, and differentially abundant metabolites were clustered into four groups. The corresponding heatmaps describe significantly enriched KEGG pathways for each cluster of the three omics data sets. After drying, transcriptional changes were rapid and decreased from HD to 20% RWC (0.67 g H 2 O/dry weight) in 10 min, with the existence of about 2,000 differentially expressed genes (DEGs). However, when the water content decreased to 5%, the transcriptional state was maintained (with only 282 DEGs), indicating that the transcriptional changes primarily occurred in the early stages of water loss. Transcriptional levels did not respond quickly upon rehydration, with only 65 differential genes between the 5% RWC and RH-30 s (2.37 g H 2 O/dry weight) groups. When considering the difference between RH-1 h to RH-12 h, 1,855 DEGs were observed ( supplementary fig. 8 , Supplementary Material online). A transcript KEGG pathway enrichment heatmap indicated that dehydration favored gene transcripts involved in photosynthesis (i.e., those encoding photosynthesis-antenna proteins, photosynthesis, carbon fixation, and carotenoid biosynthesis genes) and glycometabolism ( fig. 4 C ). Gene-co-expression analysis and clustering of the high-level DEGs into four major clusters was then conducted ( fig. 4 D ). The pathways described above were all enriched in cluster 1 ( P < 0.05) and significantly downregulated upon dehydration, indicating that photosynthesis shuts down quickly. The proteome exhibited a similar phenomenon, with proteins involved in photosynthesis being suppressed (cluster 2, P < 0.01). Dehydration from 20% RWC to 5% RWC led to the upregulation of transcripts involved in the “peroxisome,” “ascorbate and aldarate metabolism,” and “starch and sucrose metabolism” pathways indicating that antioxidant and osmoprotectant pathways began to change in the later stages of dehydration. Proteome differences recapitulated these same changes, wherein the “glutathione metabolism” and “ascorbate and aldarate metabolism” pathways were both upregulated ( P < 0.01). Furthermore, the “ribosome biosynthesis,” “ribosome,” “RNA polymerase,” and “aminoacyl-tRNA biosynthesis” pathways were strongly upregulated during dehydration and in early rehydration (clusters 2–4, P < 0.01). Between RH-1 h and RH-12 h, the downregulated pathways of photosynthesis and carbon fixation recovered to normal states. Of course, it is possible that these pathways would recover in a shorter time. The heatmap of metabolite KEGG enrichment revealed that separation of the HD from 5% RWC metabolite profiles was driven by several metabolites including those involved in “carbon fixation,” “amino acid biosynthesis,” “glutathione metabolism,” and “pentose and glucuronate interconversions.” Metabolite cluster 2 suggested that “carbon fixation” and pathways for the metabolism of several amino acids (e.g., Lys, Phe, or Tyr) were downregulated as RWC decreased, but increased upon rehydration ( P < 0.05). In addition, differences between the 5% RWC and RH-5 min treatments, extending to the 1 h treatment, were driven by several glycometabolism, “tricarboxylic acid cycle,” and “amino acid intermediate” metabolites, indicating increased metabolic activity after the initiation of hydration ( fig. 4 B and C ). Several amino acid biosynthesis genes were also upregulated during dehydration and in the early stage of rehydration ( fig. 4 D ). Horizontal Gene Transfer May Be Involved in Intertidal Evolution The evolution of red algal genomes has proceeded from complex to simple and back to gradual complexity again, with changes closely tied to environmental shifts ( Bhattacharya et al. 2018 ). One of the major contributor is frequent horizontal gene transfer (HGT) to introduce gene gains and to reconstruct a rich gene function repertoire that compensated for gene loss events ( Schönknecht et al. 2013 ; Lee et al. 2018 ). To examine possible HGT events in N. haitanensis , we aligned N. haitanensis gene sequences with refseq protein database and assessed using Alien Index (AI) ( Gladyshev et al. 2008 ; Fan et al. 2020 ). We found 522 genes with AI score > 0 and among them, 489 genes were residing on the chromosomes. By categorizing genes with AI score > 10 and clustering with bacterial sequences on a phylogenetic tree, 267 genes can be regarded as strongly evidenced horizontally transferred candidate genes (HGT candidates) ( supplementary data 9 and fig. 9, Supplementary Material online ). To analyze these candidates further, we constructed a genome-wide dehydration and rehydration response profile between DEGs ( P < 0.05) and HGT candidates. The DEGs and HGT-candidate genes did not reveal any obvious clustering in the genome because they were distributed across the five superscaffolds ( supplementary fig. 10 A , Supplementary Material online). However, the majority of these genes were predominantly located on scaffolds that were in a gene-rich region, suggesting that past HGT events were prevalent and not acquired by recent evolutionary events based on the assumption that sufficient time had allowed for dispersal of the genes throughout the genome. The scattered distribution of DEGs across the genome also reflected that the intertidal tolerance of N. haitanensis was not introduced by a restructuring event, but rather via retooling of existing genetic elements. Gene enrichment analysis of the HGT genes further revealed that N. haitanensis might have acquired the ability to adapt to environments through a series of HGT events from bacteria ( supplementary fig. 10 B and data 3, Supplementary Material online). For example, 25 HGT-candidate genes were involved in oxidative stress responses, including Cu/Zn superoxide dismutase (SOD), GST, dioxygenase, and haloperoxidase. Many of them were unique to the N. haitanensis and P. umbilicalis genomes, with some being upregulated during dried and rehydration stages. In addition, Cu/Zn SOD and GST genes have been significantly expanded within the Porphyra genomes ( supplementary data 3 and fig. 11 A and B , Supplementary Material online ). Further phylogenetic analysis revealed that the Cu/Zn SOD genes from species of Porphyra and other red algae clustered together and with that from the marine cyanobacterium Acaryochloris marina MBIC 11017. The Awaji stain of A. marina is an epiphyte of red algae ( Chan et al. 2007 ), suggesting that they might have evolved from a common marine bacterial ancestral donor ( supplementary fig. 11 C , Supplementary Material online). CAs are important enzymes that enable Bangiaceae seaweeds to concentrate carbon and maintain inorganic carbon assimilation capacity ( Wang et al. 2020 ). Seventeen CA genes were identified in the N. haitanensis genome, eight of which were apparently derived from HGT. Four α-CA genes were specific to N. haitanensis and P. umbilicalis , two of which were expanded in the ancestor of the two species ( supplementary data 3, Supplementary Material online). A comparison of CA expansion in red, brown, and green algae indicated that α-CA significantly expanded in Porphyra ( supplementary fig. 12 A , Supplementary Material online), consistent with the unusual characteristics of shell-borne conchoceli life histories. Phylogenetic analysis was then conducted for the α-CA genes ( supplementary fig. 12 B , Supplementary Material online), wherein those from green algae, Cyanobacteria, and other bacterial groups formed an independent clade that differed from the α-CAs from red algae. α-CAs presumed to be of an HGT origin clustered with other α-CAs from two Porphyra species in one large clade, suggesting they were derived from a common ancestor, and limited to the Bangiaceae family. Synteny analysis was then performed using CA genes of N. haitanensis , P. umbilicalis , and P. purpureum ( supplementary fig. 12 C , Supplementary Material online). The CA genes were distributed across the five N. haitanensis chromosomes. Thirteen of the CAs comprised conserved orthologous groups with their counterparts in the P. umbilicalis genome, with six of these being putatively derived from HGT. However, only two HGT-derived CA genes ( α-CA4-1 and β-CA7 ) formed conserved orthologous groups in both P. purpureum and P. umbilicalis genomes, suggesting that they were obtained from the ancestral red alga by HGT. Extensive gene transfer appears to have been key to the genomic evolution of the metabolically versatile intertidal red algae. Red algae contain unusually diverse oxylipins, but it has remained a mystery regarding how such complexity of oxylipin synthesis pathways arises given their limited gene inventory ( Andreou et al. 2009 ). For example, only two genes encoding LOXs have been identified in Chondrus ( Collen et al. 2013 ). Here, evidence was observed that N. haitanensis obtained a whole set of LOX gene family genes, with six LOX genes arising from HGT, and four of these were specific to N. haitanensis and P. umbilicalis ( Phlox1 , 2 , 3 , and 5 , supplementary data 3, Supplementary Material online). The LOX genes within the N. haitanensis genome exhibited significant expansion in addition to those of other intertidal red algae, including P. umbilicalis and G. chorda ( supplementary fig. 13 A , Supplementary Material online). Phylogenetic analyses of LOX genes revealed two independent clusters. Phlox1 and 2 formed an independent clade with the LOX genes from other red algae and exhibited high homology to those from the two marine bacteria, including Shewanella violaea , suggesting that they might actually be derived from marine bacterial donors ( supplementary fig. 13 B , Supplementary Material online). Syntenic analysis indicated that four Phlox genes exhibited conserved orthologous groups with their counterparts in the P. umbilicalis genome. Phlox5 and Phlox3 , in addition to Phlox2 and Phlox1 , corresponded to one gene in each of the two contigs of the P. umbilicalis genome, indicating that the two large fragments that evolved from the ancestor underwent duplication events in the N. haitanensis genome. However, conserved orthologous groups were not observed in the P. purpureum genome, indicating that all of the LOX genes in Porphyra were obtained after divergence from the mesophilic unicellular red alga P. purpureum ( supplementary fig. 13 C , Supplementary Material online). Many Phlox genes were upregulated in the dehydration and rehydration conditions ( supplementary fig. 13 D , Supplementary Material online). In addition, free polyunsaturated fatty acid (PUFA) and PUFA-containing membrane lipid contents increased upon dehydration, indicating that the oxylipin pathway is likely to be activated under these conditions ( supplementary fig. 14 , Supplementary Material online). We previously observed that some LOX proteins in N. haitanensis are multifunctional enzymes. For example, PhLox1 possesses unusually high hydroperoxidelyase, LOX, and allene oxide synthase catalytic activities based on one catalytic domain of the protein ( Chen et al. 2015 ). Consequently, these results suggest that N. haitanensis obtained the LOX gene family from marine bacteria by HGT, the genes underwent massive expansion, and were later modified to achieve multifunctional ability, potentially explaining the observed diversity of red algal oxylipins. Short-Term Shutdown of Photosynthesis to Avoid Photo-Damage Red algae in intertidal zones experience periodic dryness every day. Their simple genomes have forced them to adopt passive, simple, and rapid modes of adaptation during desiccation. N. haitanensis shut down their photosynthetic activity through quenching Fv / Fm activities within minutes of transfer from fresh to desiccation conditions ( fig. 5 A ). After rewetting, photosynthetic function was quickly regained within 30 s, with the Fv/Fm ratio recovering to 74% of that of fresh thalli and complete recovery within 5 min. The expression of numerous genes and proteins involved in photosynthesis were repressed during dehydration including nuclear-encoded components of the Lhca antenna complexes, PSI and PSII proteins, and F-type-ATPases ( fig. 5 B ). However, protein expression was rapidly regenerated within 30 s of rehydration, whereas transcriptional levels continued to be inhibited until after 12 h of rehydration. N. haitanensis and Neopyropia yezoensis exhibit similar photo-inhibition phenomena under dehydration conditions ( Wang et al. 2018 ). Huang et al. (2021) recently found that photosynthesis-antenna proteins, such as Lhca 1, were upregulated when N. haitanensis was under dehydration, and inferred that this is the strategy adopted by N. haitanensis to prevent or lower light-stress-induced damage. This finding is not consistent with our results and the speculation that N. haitanensis would shut down the photosystem to reduce the photo-damage. Fig. 5. Strategies of Neoporphyra haitanensis photosynthetic systems in response to intertidal stress. ( A ) Photochemical efficiency of N. haitanensis during desiccation and rehydration cycling. ( B ) Transcriptome (squares) and proteome (triangles) expression profiles of the N. haitanensis photosynthetic system under dehydration and rehydration conditions are shown near the pathway as heatmaps. ( C ) Overview of the Calvin cycle response to osmotic stress. The data show transcriptome and proteome profiles from the HD (hydration), 5% RWC, RH (rehydration)-30 s, RH-1 h, and RH-12 h groups (from left to right). The square heatmaps show gene expression levels (normalized FPKM) ( n = 3), the triangle heatmaps show the normalized peak areas of proteins ( n = 8), and the circular heatmaps show the normalized peak area of metabolites ( n = 8). Gray letters represent genes that have not been detected. ( D ) Changes in photosynthetic pigment content across dehydration and rehydration conditions ( n = 3). In association with the shutdown of photosynthesis, the carbon fixation reactions of photosynthesis were also strongly repressed during dehydration. The RuBP , Pgk , and Gapa genes in the carboxylation and reduction stages of the Calvin cycle were downregulated, whereas most genes involved in the regeneration of ribulose 1,5-bisphosphate were also downregulated. Decreased abundances of detected metabolic intermediates confirmed this observation for the Calvin cycle ( fig. 5 C ). Huang et al. (2021) also observed that carbon fixation pathways were significantly enriched and downregulated when RWC was decreased to 15%, and proposed that N. haitanensis responded to drought stress by reducing energy metabolism. We discovered that these changes almost completely reversed during the later period of rehydration, consistent with changes in Fv/Fm over the same period. Another inevitable problem of intertidal life is photo-protection. N. haitanensis did not exhibit evident effects of photo-damage under dehydration conditions ( supplementary figs. 6 and 7 , Supplementary Material online). A nonphotochemical quenching (NPQ) mechanism implemented by the xanthophyll cycle is widely found in green algae and land plants that can protect the photosynthetic apparatus from photo-damage ( Lüttge et al. 2011 ). Red algae are however generally not thought to have a xanthophyll cycle. For example, Cyanidioschyzon merolae lacks various mechanisms for dissipating excessive light energy ( Matsuzaki et al. 2004 ). Porphyra exhibits similar deficiencies. When N. haitanensis was dehydrated, NPQ decreased, similar to the Fv/Fm trends ( fig. 5 A ). Although several carotenoid derivatives were detected in N. haitanensis , including primarily zeaxanthin, β-carotene, and lutein ( supplementary fig. 15 , Supplementary Material online), most exhibited decreased concentrations. Violaxanthin, which participates in the xanthophyll cycle, was not detected, and interconversions of xanthophylls were not evident. Moreover, N. haitanensis , P. umbilicalis , and N. yezoensis all were observed to lack violaxanthin ( Yang et al. 2014 ; Koizumi et al. 2018 ). Analysis of the N. haitanensis genome (along with other red algal transcriptomic analyses) also indicated that red algae lost the de-epoxidase gene ( Vde ); this loss was speculated to have occurred in red algae ancestors ( Wang et al. 2018 ). Furthermore, evidence was not observed in the genome for some members of the PSII complex that are associated with NPQ, including Lhc SR, PsbS, Lhcx, and Lhcb4 ( Lüttge et al. 2011 ), suggesting that N. haitanensis lacks a xanthophyll cycle. In contrast, we observed that the periodic photo-protection in N. haitanensis seems to rely on reducing all the outer light harvesting antenna pigments to avoid photo-damage caused by excessive photon absorption. During dehydration, significant decreases in the content of light-harvesting phycobiliproteins (phycoerythrin [PE], phycocyanin [PC], and allophycocyanin [APC]) were observed. In addition, the concentrations of the accessory pigment β-carotene were also diminished ( fig. 5 D ). The core pigment chlorophyll a was retained throughout the hydration/dehydration process, which would help to maintain rapid recovery of photosynthesis upon rehydration. All content recovered 1 h after rewetting. Indeed, it is astonishing that photosynthetic system proteins can be decomposed and resynthesized in such rapid cycles. This is similar to the performance of poikilochlorophyllous (PDT) in plants, although PDT will dismantle their thylakoid membranes during dehydration and recovers slowly upon rewetting due to the time required for protein resynthesis ( Lüttge et al. 2011 ). In contrast, the thylakoid membranes of dried N. haitanensis cells were intact based on TEM observation ( supplementary fig. 6 , Supplementary Material online). The mechanism underlying the rapid reconfiguration of the photosynthetic mechanism requires further study. However, this mechanism may be a deficiency of Porphyra . For example, because the photosynthetic system cannot properly initiate during low tide, they are unable to adapt to excessively long periods of desiccation, thereby preventing them from colonizing land habitats. S-adenosyl-l-Methionine-Dependent Methyltransferase Coordinates with Antioxidant System to Promote Intertidal Adaptation S-adenosyl- L- methionine (SAM)-dependent methylation plays several crucial roles in ribosomal stability ( Wu et al. 2020 ); the biosynthesis of nucleic acids, proteins, and secondary metabolites ( Luo et al. 2019 ); and in epigenetic regulatory processes ( Sun et al. 2021 ). Numerous SAM-methyltransferase (SAM-MTase) genes were identified in the unique genes shared by N. haitanensis and P. umbilicalis and were also greatly expanded in the N. haitanensis genome. An additional 11 SAM-MTase genes were suspected to have been derived from HGT ( supplementary data 3, Supplementary Material online). Phylogenetic analysis indicated that the SAM-MTase genes of N. haitanensis clustered with those of other Porphyra and were mainly distributed in two relatively independent clusters. The two clusters also comprised genes from the intertidal red alga G. chorda , suggesting that the transmethylation function was a specific adaptation of intertidal algae ( fig. 6 A ). Many of these genes were upregulated during either early or late dehydration stages and during rehydration ( fig. 6 B and supplementary data 3, Supplementary Material online). Importantly, the unusually high SAM content increased by 6.07-fold in the 5% RWC group compared with their expression in fresh thalli ( P < 0.01), whereas S-adenosylhomocysteine (SAH) content increased by 15.95-fold during dehydration ( fig. 6 C , P < 0.01). Demethylated SAH can be hydrolyzed by S-adenosyl-homocysteine hydrolase (ahcY) to generate homocysteine (HCY) that can then be remethylated to methionine (MET) by methionine synthase (metH) ( Tehlivets et al. 2013 ). The ahcY and metH genes were upregulated under dehydration and rehydration conditions, indicating that SAH could be continuously recycled back to MET. Typically, 90% of SAM is involved in various methylation activities, whereas the remaining 10% is used for synthesizing polyamines ( Moffatt and Weretilnyk 2001 ). Genes encoding enzymes involved in polyamine spermidine (SPD) synthesis exhibited significant upregulation ( P < 0.01) that promoted the formation of SPD. Indeed, SPD exhibited significantly increased levels during dehydration (5.66-fold increase, P < 0.01). Importantly, we observed a particularly interesting SAM process coupling the glutathione (GSH)–ascorbic acid (AA) redox cycle to influence the antioxidant enzyme system in N. haitanensis . HCY accumulation can cause cell oxidative stress and is consequently recycled back to MET, but can also be converted to cysteine and then to GSH ( Pizzorno 2014 ). Significantly increased GSH levels were observed in N. haitanensis during dehydration, with 45.13-fold higher levels than in fresh samples ( P < 0.05). In addition, the entire GSH-AA-NADPH redox response system was highly activated during dehydration ( P < 0.01), with AA exhibiting 433.59-fold higher levels than in the HD group ( P < 0.01). Transient or sustained changes in all or some components of this system were all observed. Therefore, by cleverly combining the SAM-methyltransfer cycle, polyamine synthesis, and the GSH-AA-NADPH antioxidant system, a highly reducing intracellular context is created within cells that can help protect seaweeds from oxidative damage when exposed to falling tide environments. Fig. 6. \n Neoporphyra haitanensis adapts to intertidal environments by enhancing SAM-dependent methyltransfer activity and synergistic activities with antioxidant systems. ( A ) Phylogenetic tree of SAM-dependent methyltransferase homologs from different species. Phylogenetic trees were constructed using neighbor-joining methods in MEGA7 and node support was evaluated with bootstrap tests (1,000 replicates). Red branches represent red algal genes, green branches represent plant genes (including genes from green algae and Arabidopsis thaliana ), blue branches represent bacterial genes, and brown branches represent genes from Ectocarpus siliculosus . Homologs from the N. haitanensis genome are highlighted in red text, and members thought to derive from HGT are marked with a red star. ( B ) Heatmap showing differential expression of SAM-dependent methyltransferase genes and methyltransfer-related genes in N. haitanensis under desiccation and rehydration stress conditions. Values are fold-change (log2-ratio) values of transcript levels under dehydration and rehydration conditions relative to fresh conditions. ( C ) Summary of the SAM-polyamine-GSH-ascorbic acid metabolic network during desiccation and rehydration response by N. haitanensis . The network was reconstructed based on KEGG pathways. The squares in the heatmap show gene expression levels (normalized FPKM) ( n = 3), and the circles in the heatmap show normalized average values of peak area for metabolites ( n = 8) under hydration (HD), 5% RWC, RH (rehydration)-5 min, RH-1 h, and RH-12 h conditions (from left to right). Rapid Biosynthesis and Repair of Macromolecules Is an Important Mechanism of Intertidal Adaptation \n N. \n haitanensis responds to rehydration amazingly rapidly. The three different omics data sets analyzed here suggested that protein and metabolite levels rapidly responded to rehydration, but it is unclear how this is achieved. The basis for this rapid response is certainly due to adequate advance preparation for responses. This preparation may involve several mechanisms. First, gene transcription always remains in a state of preparation and storage, with over 80% of N. haitanensis gene inventory remaining in an active transcriptional state (FPKM > 0; >9,000 genes) across treatments, even in the 5% RWC condition. Oliver et al. (2005) demonstrated that novel transcripts are not recruited by protein synthetic machinery during bryophyte drying and that protein synthesis is rapidly lost. This result contrasts significantly with those in the present study. Second, ribosomal function and ribosomal biogenesis remain in a transcriptional readiness state to initiate protein synthesis, with the expression of genes involved in these pathways continually increasing from hydration to 5% RWC ( fig. 4 and supplementary fig. 16 , Supplementary Material online; P < 0.05). In addition, a significant increase occurred at RH-30 s. In contrast, proteomic analysis indicated that ribosomal protein expression did not actually increase during 5% RWC conditions, but instead significantly decreased ( P < 0.05). This result may be due to short-term carbon starvation, when the synthesis of various macromolecules would have ceased during dehydration ( Lallemand et al. 2019 ). However, ribosomal reserves could be leveraged upon rehydration, and proteins would be quickly synthesized to restore their levels. Huang et al. (2021) also noticed the importance of the ribosome in the stress response of N. haitanensis to desiccation, because in the protein interaction network, multiple ribosomal relative proteins exhibited the highest connectivity values. In addition, protein repair and correct folding after synthesis is always maintained, as indicated by the massive expansion of Hsp90 genes in the genome and by the tight transcriptional regulation of several Hsp genes via heat shock transcription factors (Hsf). Hsfs also interact with genes including the cell division control protein 48 gene (CDC48), which regulates the biogenesis of the 60S ribosomal subunit ( Kressler et al. 2010 ); the nuclear valosin-containing protein-like (NVL) gene involved in pre-rRNA processing ( Nagahama et al. 2004 ); mitogen-activated protein kinase (MAPK) genes ( Hamel et al. 2006 ); and the target of rapamycin (TOR) protein-coding genes involved in activating transcription, protein synthesis, and ribosomal biogenesis ( Martin et al. 2006 ). Some of the aforementioned genes were upregulated under dehydration or rehydration conditions, including Hsp90 genes ( supplementary fig. 17 , Supplementary Material online, P < 0.05). In addition, genes involved in DNA mismatch repair, base excision repair, and nucleotide excision repair were also upregulated during desiccation ( supplementary fig. 18 , Supplementary Material online). Some proteins play primary protective roles in desiccation-tolerant plants or green algae, such as the late embryogenesis abundant (LEA) protein family, which will expand and accumulate to high quantities during dehydration ( Lüttge et al. 2011 ; VanBuren et al. 2018 ). LEA proteins were not found in the genome of N. haitanensis , indicating that in the early eukaryotic red algae stage, LEA proteins have not yet evolved or functioned as part of the desiccation-tolerant strategy."
} | 13,825 |
26273237 | null | s2 | 4,387 | {
"abstract": "Interspecies interactions are the basis of microbial community formation and infectious diseases. Systems biology enables the construction of complex models describing such interactions, leading to a better understanding of disease states and communities. However, before interactions between complex organisms can be understood, metabolic and energetic implications of simpler real-world host-microbe systems must be worked out. To this effect, untargeted metabolomics experiments were conducted and integrated with proteomics data to characterize key molecular-level interactions between two hyperthermophilic microbial species, both of which have reduced genomes. Metabolic changes and transfer of metabolites between the archaea "
} | 183 |
39012831 | PMC11287280 | pmc | 4,389 | {
"abstract": "Significance One of two fundamental strategies for forming multicellular organisms relies on cell aggregation. To do so, cells must distinguish self from nonself to form a cooperative tissue where myxobacteria serve as model organisms for making these transitions. Here, we tested the role of the TraAB kin recognition system for a role in cell–cell cooperation because these cell surface receptors mediate the bidirectional exchange of proteins and lipids between cells. Strikingly, cells that adapted to environmental stresses shared their beneficial trait with naïve kin in a TraAB-dependent manner. Surprisingly, donor cells also benefited by apparently establishing harmony in the population when confronted with stress. We conclude that TraAB plays diverse roles in myxobacterial multicellular behaviors including their ability to cooperate.",
"discussion": "Discussion Prior studies established that OME plays a key role in kin discrimination, but a possible role in kin cooperation was unclear, though implicated by its bidirectional transfer mechanism. This study revealed that cells adapted to a particular stress, detergent resistance, transfer their beneficial trait to naïve kin by OME. Here, the actor adapted by a genetic change, a common mechanism of bacterial adaptation ( 51 ), while the recipient’s adaptation was phenotypically based and not genetic. This study reports transfer of a beneficial adaptation trait to kin cells. Consistent with our findings, previous work showed M. xanthus cells can rescue mutants defective in motility or LPS biogenesis by sharing or replenishing vital cell components via OME ( 11 , 26 ). Moreover, we further show here that in mixed populations, actor cells also benefit by preventing a lethal stress response in sensitive kin that deleteriously impacts actor cells ( Fig. 4 ). Additionally, twnR cells are resistant to SitA3 lipoprotein intoxication by inhibitor cells whereby they transfer their resistant trait to sensitive target cells by OME ( Fig. 5 ) ( 19 ). Taken together our results indicate one or more factors are transferred from twnR actors to kin as summarized in Fig. 7 . The nature of this transferable factor(s) is currently unknown, but in principle, it could be a signaling molecule to trigger a protective cellular response or the factor itself could be the resistant determinant. Similar to other gram-negative bacteria ( 28 – 30 ), M. xanthus presumably modifies and adapts their cell envelope composition to environmental stresses. Therefore, the resistant factor(s) is likely overexpressed or represents an altered OM protein and/or lipid. Because the factor is transferrable by OME, which is likely mediated by OM fusion, it must be diffusible (i.e., transferable) in the cell envelope. Consistent with this idea, TwnR is structurally similar to the BamA OM assembly factor ( SI Appendix , Fig. S2 ) ( 43 ), and thus could play a related, albeit minor, role in the protein composition of the OM. In turn, the absence of TwnR could signal a response that changes the composition of the cell envelope. Clearly, future work needs to elucidate the mechanism of Tween-20 resistance and hence trait transfer. Fig. 7. Adaptation transfer model. In response to stress adapted actor cell expresses modified OM with altered or overexpressed proteins and/or lipids (green circles/membrane). Homotypic recognition of kin by TraAB receptors results in OME and transfer of adaptation phenotype to sensitive cells. The resulting adapted kin is resistant to the stress and does not antagonize actor cells when confronted with an insult. Our work suggests a mechanistic and evolutionary framework for myxobacteria multicellularity and a role played by OME. In addition to functioning in kin discrimination, and hence selecting for self in aggregative collectives, our results are consistent with the hypothesis that OME plays an instrumental role in cell–cell cooperation in natural populations. In essence, this allows cells adapted to different microenvironments, or differentiated into different cell types, to interact constructively by sharing cellular components, a hallmark of multicellular organisms ( 52 ). In an evolutionary context, the traAB locus represents a greenbeard gene involved in social recognition and the transfer of beneficial traits to kin ( 3 , 53 ). Thus, traAB , as a single locus, is a relatively simple system that promotes self-recognition and social interactions that conceptually arose in one evolutionary step toward multicellularity. In contrast to twnR , the dcaR E128K strain did not share its adaption trait, presumably because those adaptation determinants are not competent for transfer by OME. This finding suggests those factors are not freely diffusible and/or not localized in the OM. Previous work on DcaR (aka Nla13; MXAN_3811) showed that a mutant had no discernible development or motility phenotype ( 37 ), but, interestingly, during predation was one of the most severely down-regulated genes (>eightfold) ( 49 ). Our findings suggest DcaR plays a role in stress responses, particularly detergent resistance. When confronted with changing environmental stresses, a common strategy bacteria employ to adapt is to alter their gene regulatory networks through mutations in transcription factors ( 51 ). Indeed, our genetic studies indicate that the E128K substitution in the DcaR transcription factor resulted in a gain-of-function change that allows broad adaptation to detergent stress. Consistent with this, our proteomics studies detected the DcaR E128K protein, while the WT DcaR protein could not be detected ( Dataset S1A ). This indicates that the E128K substitution increased the abundance of DcaR E128K by positive autoregulation and/or increased protein stability. Additionally, 19 other proteins were either present or absent between these isogenic strains and another 140 proteins were either up or down-regulated ≥twofold ( Dataset S1 A and B ). These results show the dcaR E128K gain-of-function allele globally alters protein expression and suggests that in WT cells DcaR + plays a role in detergent adaptation. Consistent with this, a separate report with a DcaR paralog, uncovered an unknown role in pilA gene regulation by analogous gain-of-function substitutions in the REC and adjacent domains of the NmpR enhancer binding protein ( 38 , 39 ). Thus, as found for NmpR, along with our findings ( Fig. 2 A ), suggests that gain-of-functions alleles in these σ 54 transcriptional activators uncovered new biological functions. The Tween-20 stress response induces lipid body production and the fatty acid β-oxidation degradation pathway. These responses show intriguing parallels to predation and developmental responses ( 45 , 47 , 49 ). Specifically, predation induces the fatty acid β-oxidation degradation pathway, likely triggered by prey membrane degradation and fatty acid release by myxobacterial lipases ( 54 ). Starvation-induced development activates the same pathway along with lipid body production, which arises from phospholipid removal and processing from membranes during the transition from rod-shaped cells to spherical spores. Moreover, the external addition of fatty acids to E-signaling mutants restores lipid body production and development ( 55 ). In the case of Tween-20, it likely releases lipids from membranes during vegetative growth that similarly induces lipid body formation and the fatty acid β-oxidation degradation pathway. Alternatively, or in addition, given that M. xanthus contains 36 lipases ( 54 ), including phospholipase MXAN_1969 ( Fig. 6 ), these enzymes could cleave and release free fatty acids from Tween-20 ( Fig. 1 B ). In either case, Tween-20 exposure results in lipid body production and an apparent signal for induction of the fatty acid degradation pathway. Curiously, in the absence of adaptation transfer, sensitive kin killed twnR ::Tn cells upon detergent exposure. From an evolutionary perspective, this antagonistic response provides a selective pressure on actors to cooperate with their less-fit kin for their own survival. However, in an analogous antagonistic response, triggered by metabolic disharmony, where starving and stressed auxotrophs missing an essential metabolite, were instead killed by their more fit prototroph kin in a T6SS-dependent manner ( 44 , 56 ). In our current work, the roles were reversed, where stressed cells killed their more fit kin, along with their own demise, by a T6SS-independent mechanism. Clues to the mechanism of killing might come from their detergent-stress response. Here, lipid body production in M. xanthus serves as a morphological marker of stress, and during stress induced by starvation, widespread program cell death concurrently occurs during development ( 46 , 55 ). This killing response involves autocides, including AMI that is composed of fatty acids, and, unlike other gram-negative bacteria, M. xanthus is very sensitive to a variety of lipids and fatty acids ( 57 – 60 ). Therefore, given the induction of the fatty acid degradation pathway and lipid bodies, a plausible mechanism for killing the twnR ::Tn cells is that their stressed sibling releases autocides like AMI when they lyse. Clearly, future studies need to test these possibilities. In our experiments, adaptation transfer by OME only protected a small fraction of kin (relative value of ~2%) from detergent toxicity. The reason for this is likely multifaceted, though some explanations are apparent. First, assay conditions were harsh where no kin survived in monoculture and approximately 95% of twnR cells also perished. Because of this, only small fractions of twnR cells were competent enough to transfer a resistant phenotype. Presumably, those cells that survived the initial shock of detergent exposure subsequently developed a higher level of tolerance following their stress response ( Fig. 6 and SI Appendix , Fig. S9 ). Second, the OME transfer efficiency of detergent resistance is likely partial and does not reach the full level found in twnR donor cells. Additionally, there may be a temporal lag for recipients to adapt following OME, involving altered gene expression and/or physiological changes in their cell envelope. Finally, there were experimental parameters that both enhanced and suppressed OME efficiency. Here, the recipient strain (DW2301) contains a second chromosomal copy of traAB at the Mx8 attB locus, resulting in traAB overexpression that enhances OME. However, in contrast, twnR is a nonmotile strain and when cell mixtures were on Tween-20 plates there was no added calcium (TraA and TraB contain calcium-binding domains), both of which inhibit OME. Future studies need to elucidate why OME-dependent survival rates were low and how various conditions enhance or impede transfer efficiencies. In conclusion, our findings reveal myxobacteria adapted to environmental stresses can recognize and transfer their trait to kin by OME. This behavior benefits the population at large whereby naïve kin are preadapted and protected against forthcoming stresses in these cooperative communities. In other work, we showed OME can phenotypically rescue genetic lesions of kin and promotes the assembly of homogeneous populations ( 9 , 11 , 26 , 27 ). In nature, we suggest OME provides a fitness advantage for cell–cell cooperation and our findings provide clues about those fitness gains. However, additional studies are required to delineate the role of OME in nature along with other mechanisms employed to facilitate cooperation and multicellularity in myxobacteria."
} | 2,911 |
36032730 | PMC9402969 | pmc | 4,390 | {
"abstract": "To disclose the net effect of light on microalgal growth in photobioreactors, self-shading and mixing-induced light–dark cycles must be minimized and discerned from the transient phenomena of acclimation. In this work, we performed experiments of continuous microalgal cultivation in small-scale photobioreactors with different thicknesses (from 2 to 35 mm): working at a steady state allowed us to describe the effect of light after acclimation, while the geometry of the reactor was adjusted to find the threshold light path that can discriminate different phenomena. Experiments showed an increased inhibition under smaller culture light paths, suggesting a strong shading effect at thicknesses higher than 8 mm where mixing-induced light–dark cycles may occur. A Haldane-like model was applied and kinetic parameters retrieved, showing possible issues in the scalability of experimental results at different light paths if mixing-induced light–dark cycles are not considered. To further highlight the influence of mixing cycles, we proposed an analogy between small-scale operations with continuous light and PBR operations with pulsed light, with the computation of characteristic parameters from pulsed-light microalgae growth mathematical modeling.",
"conclusion": "5 Conclusion In high-scale microalgal cultivation systems, self-shading and mixing-induced light–dark cycles heavily hinder the true effect of light interaction with cells. To minimize these phenomena, we designed, built, and operated ultra-thin continuous photobioreactors. Experiments showed that at very low light paths (2–8 mm), high inhibition occurred at low irradiance values. We used experimental data from different reactor thicknesses to retrieve parameters of a Haldane-like model, and the trend of the fitted Lambert–Beer constant shows that there could be issues in the capability of the model to scale up experimental results at ultra-thin light paths. We infer that this is due to the averaged nature of the model that does not consider the effect of mixing-induced light–dark cycles. Indeed, the computation of parameters from the theory of mathematical modeling of pulsed-light effects on microalgae suggested that only in ultra-thin light paths (2–8 mm), the “real” continuous light regimen occurs, while in thicker light paths (15–35 mm) cells are subjected to an effective pulsed light one due to mixing-induced light–dark cycles.",
"introduction": "1 Introduction Among all the factors that affect microalgae productivity and growth, light covers a major role since it provides all the energy required for metabolism. The effect of light on microalgal photosynthesis has been the subject of a number of studies in the past that identify three different light-dependent regions for microalgal growth: photolimitation, photosaturation, and photoinhibition. The experimental observation of these three regions in photobioreactors is influenced by two adverse phenomena that hinder the true photosynthetic response to light: self-shading and mixing-induced light–dark cycles ( Barbosa et al., 2003 ; Graham et al., 2015 ). Self-shading is light attenuation by microalgae cell absorption that reduces light availability in the inner parts of the photobioreactors ( Flynn, 2021 ). Moreover, microalgae are exposed to light–dark cycles because of the light gradient induced by the self-shading and the turbulent mixing in the reactor; due to these conditions, cells receive light intermittently ( Barbosa et al., 2003 ; Zarmi et al., 2013 ). A strategy to minimize these adverse phenomena is to decrease the culture light path ( Castaldello et al., 2019 ). Decreasing the scale of photobioreactors (to microphotobioreactors) can lead to multiple advantages, such as faster experiments in large numbers of replicas with the possibility to precisely control growth conditions ( Perin et al., 2016 ; Kim et al., 2018 ). Microphotobioreactors can reach scale volumes down to the pico-liter ( Kim et al., 2018 ) and have been used to cultivate microalgae in droplet-based flow systems ( Saad et al., 2019a ; 2019b ), single-cell layer microfluidic chips ( Luke et al., 2016 ; Westerwalbesloh et al., 2019 ), or fed-batch microwell systems ( Perin et al., 2016 ; Castaldello et al., 2019 ). Recently, microphotobioreactors have been especially used to investigate light effects on microalgal growth, that is, applications include microalgal growth under irradiance and nitrate stress conditions ( Saad et al., 2019a ; 2019b ) and the evaluation of light effects on microalgae culture under non-limiting CO 2 conditions ( Castaldello et al., 2019 ). A limitation of the aforementioned works is that due to the micro-scale, continuous operations are not possible. This could be achieved by operating on a small scale (instead of a micro-scale) that represents an intermediate between microphotobioreactors and traditional “high-scale” systems. Several attempts have been made in previous literature, such as using a flat plate photobioreactor of 2 cm thickness by Busnel et al. (2021 ) or Pfaffinger et al. (2019 ). Other studies researched steady-state reactors of 1.2–1.5 cm deep ( Sforza et al., 2014 , 2015a , 2015b , 2018 ; Barbera et al., 2017 ), finding remarkably high concentrations, confirming that self-shading is limited. However, it is not clear if a light path of 1.5–2 cm is actually able to avoid the self-shading effect. The objective of this work was to capture the net effect of light on microalgal growth in photobioreactors by minimizing phenomena of self-shading and mixing-induced light–dark cycles in acclimated cultures. For this purpose, we designed small-scale continuous closed photobioreactors with different reactor thicknesses of 2, 5, and 8 mm, where Tetradesmus obliquus was cultivated at a steady state. The employment of such a small scale (2–8 mm thickness of the light path) in a continuous mode represents a novelty of this work; to our knowledge, few attempts of microalgae cultivations in closed photobioreactors with these light paths are reported in the literature and none in the continuous mode. Thin-layer cascade (TLC) photobioreactors can exhibit light paths of a few millimeters, with the possibility of semi-continuous or continuous operation ( Grivalský et al., 2019 ). However, TLCs are generally operated via flow recirculation usually comprising a retention chamber ( Grivalský et al., 2019 ). This operation mode determines that microalgae are not continuously exposed to light and make TLCs significantly different from continuous ultrathin flat-plate photobioreactors. We compared our experimental data at 2–8 mm with those at 15 and 35 mm for the same species and similar cultivation systems ( Sforza et al., 2015b ; Barbera et al., 2017 ; Borella et al., 2021 ) to highlight the effect of self-shading and mixing-induced light cycles. A modeling approach was then used on all the experimental data to better assess the impact of self-shading with reference to the available light path.",
"discussion": "4 Discussion To understand the aforementioned limitations, we recall the two main adverse phenomena that hinder the net effect of light on microalgal growth: self-shading and mixing-induced light–dark cycles. In the model, self-shading is approached through light attenuation in the Lambert–Beer law ( Eq. 1 ), while the effect of light–dark cycles is not considered. To obtain a measure of this phenomenon, Richmond et al. (2003) introduced the concept of cell travel time, that is, the average time required for cells to move back and forth in the reactor thickness, where they are exposed to different light conditions along the light attenuation profile. Richmond et al. (2003) defined some characteristic cell travel times, based either on random, diffusion-like motions or on back and forth movement through the optical path. The latter, which is called regular motion time, is more relevant in the context of this work and is defined as follows: \n τ t = L / v c e l l , \n (6) \n where \n L \n (cm) is the optical path and \n v c e l l \n (cm s −1 ) is cell lateral velocity, assumed equal to the bubble velocity (30–50 cm s −1 , in accordance with Richmond et al. (2003 ). Cell travel time does not affect photosynthesis if it is one or two orders of magnitude higher than a characteristic PSU turnover time (about 10 ms for Richmond et al., 2003 ). Usually, this is not an issue. However, for very thin reactors, cell travel time starts to approach photosynthetic unit (PSU) turnover time, influencing light–cell interaction. Table 2 shows the cell travel times for the different reactor thicknesses used in our work. From 2 to 8 mm, travel times are very close to the PSU turnover time and are significantly smaller than travel times at 15 and 35 mm. This highlights that the behavior at 2–8 mm is different from that in 15–35 mm, and in the smaller scales, mixing-induced light–dark cycles could have a significant impact. TABLE 2 Cell travel times for all reactor thicknesses. 2 mm 5 mm 8 mm 15 mm 35 mm Regular motion time [ms] 6.7 16.7 26.7 50 116.7 In an ideal situation for efficient light exploitation, cells should be exposed to high light conditions in the photic zone for a duration required for light reactions to occur and then move to the dark zone, being replaced by other cells from the dark zone ready to receive incoming photons ( Richmond et al., 2003 ). However, if the time spent by cells in the dark zone is too low, dark reactions do not have enough time to occur and photosynthetic efficiency decreases ( Zarmi et al., 2020 ). As introduced before, cells in ultra-thin reactors are exposed to strong illumination but, due to the flatter light profile and the absence of a proper dark zone, they do not fully recover from this strong light absorption, causing the photosynthetic efficiency to decrease. This is not the case in thick reactors, where cells are able to migrate to a dark zone and recover; this explains why photoinhibition in these systems is observed only under very high light conditions (or never, in some cases). Previous considerations suggest that not only light conditions cells are exposed to but also the duration of exposure of light along the attenuation light profile is necessary for the comprehension of cell–light interaction phenomena. To provide a systematic description, we can use the theory of the mathematical modeling of pulsed-light effects on microalgae: when cells move between zones with different irradiation, they are exposed to an effective illumination regime close to a pulsed light one ( Zarmi et al., 2020 ). As in the study by Schulze et al. (2017) , pulsed-light models have different inputs: intensity of the pulse, time of the light phase ( \n t l \n ), and time of the dark phase ( \n t d \n ). In an experimental pulsed-light apparatus, these inputs are usually set by a lamp, while in our case, we must calculate them with the aid of the travel time hypothesis in Richmond et al. (2003) . The intensity of the pulse is here assumed equal to the light intensity of the light source. Here, we assume that the photic zone (expressed through the length coordinates \n z p h \n (m)) is the zone in the reactor in which light intensity is inhibiting, thus higher than \n I o p t \n . Length \n z p h \n can be computed as \n z p h = − 1 k a X o u t log I o p t I 0 . \n (7) \n \n Light phase time for cell regular motion is calculated as \n t l = z p h v c e l l . \n (8) \n \n Corresponding dark phase time is calculated as \n t d = W − z p h v c e l l , \n (9) \n where \n W \n is the the reactor thickness. With light and dark phase times, the duty cycle \n ϕ \n (-) can be calculated as \n ϕ = t l t l + t d . \n (10) \n \n In our case, \n ϕ \n represents in proportion the time a single cell is exposed to inhibitory conditions. We observed that parameter k \n \n a \n decreases until stabilizing at about W = 15 mm ( Figure 3A ), where its value is about 0.1 m 2 g −1 , which is consistent with other estimates in similar systems ( Barbera et al., 2020 ). If that value is used to compute the duty cycle \n ϕ \n , it can be observed that \n z p h > W \n for 2–5–8 mm, and therefore \n ϕ \n can be assumed equal to 1 ( Table 3 ). This means that in ultra-thin photobioreactors, cells are exposed to light continuously and cannot adequately recover from the high photon absorption in the high irradiance region. The consequent inhibition is artificially captured by the model by increasing \n k a \n , that is, by assuming less average light in the photobioreactor than there is actually. TABLE 3 Duty cycle (non-dimensional) for all reactor thicknesses, computed for \n k a \n = 0.1 m 2 g −1 . 2 mm 5 mm 8 mm 15 mm 35 mm Duty cycle [-] 1 1 1 0.2 0.2 Previous considerations also show a touchpoint between thin-milli reactors and pulsed-light experiments toward the understanding and modeling of the effect of light on cell growth and confirm the importance of considering light–dark mixing cycles to ensure the scalability of laboratory results to higher-scale systems. Indeed, Eqs. 7 – 10 may represent a starting point to integrate the effect of mixing cycles in traditional light modeling in microalgal growth."
} | 3,297 |
36913533 | PMC10061930 | pmc | 4,391 | {
"abstract": "Cooperation between comammox and anammox bacteria for\nnitrogen\nremoval has been recently reported in laboratory-scale systems, including\nsynthetic community constructs; however, there are no reports of full-scale\nmunicipal wastewater treatment systems with such cooperation. Here,\nwe report intrinsic and extant kinetics as well as genome-resolved\ncommunity characterization of a full-scale integrated fixed film activated\nsludge (IFAS) system where comammox and anammox bacteria co-occur\nand appear to drive nitrogen loss. Intrinsic batch kinetic assays\nindicated that majority of the aerobic ammonia oxidation was driven\nby comammox bacteria (1.75 ± 0.08 mg-N/g TS-h) in the attached\ngrowth phase, with minimal contribution by ammonia-oxidizing bacteria.\nInterestingly, a portion of total inorganic nitrogen (∼8%)\nwas consistently lost during these aerobic assays. Aerobic nitrite\noxidation assays eliminated the possibility of denitrification as\na cause of nitrogen loss, while anaerobic ammonia oxidation assays\nresulted in rates consistent with anammox stoichiometry. Full-scale\nexperiments at different dissolved oxygen (DO = 2 – 6 mg/L)\nsetpoints indicated persistent nitrogen loss that was partly sensitive\nto DO concentrations. Genome-resolved metagenomics confirmed the high\nabundance (relative abundance 6.53 ± 0.34%) of two Brocadia- like anammox populations, while comammox bacteria within the Ca. Nitrospira nitrosa cluster were lower in abundance\n(0.37 ± 0.03%) and Nitrosomonas -like ammonia\noxidizers were even lower (0.12 ± 0.02%). Collectively, our study\nreports for the first time the co-occurrence and cooperation of comammox\nand anammox bacteria in a full-scale municipal wastewater treatment\nsystem.",
"introduction": "Introduction Despite their ubiquitous detection in\nengineered and natural ecosystems, 1 − 7 the role of comammox bacteria in full-scale nitrogen removal remains\nto be established. Our previous work demonstrated that comammox bacteria\nare most prevalent in nitrogen removal systems treating wastewater\nwith an attached growth phase or long solids retention times, and\nthey often co-occur with strict ammonia oxidizing bacteria (AOB) and Nitrospira -NOB. 8 Further, a large\nproportion of comammox bacteria detected in wastewater systems, including\nthose in our past studies, 8 , 9 belong to clade A1 comammox\nbacteria and are affiliated with Ca. Nitrospira nitrosa-like populations. 10 − 12 However, kinetic parameters\nfor the majority of comammox bacteria are undetermined; data from\nonly one isolated species ( Ca. Nitrospira\ninopinata) 13 and one enrichment ( Ca. Nitrospira krefti) 14 have indicate a high affinity for ammonia. Assessment of ammonia\noxidation activity in wastewater treatment systems with coexisting\nstrict AOB and comammox bacteria has been done using metatranscriptomics,\nwhich suggested comammox bacteria were active and potentially metabolically\nflexible. 15 However, quantifying the nitrification\nrates of comammox bacteria in wastewater treatment systems would help\nbetter define their roles in nitrogen removal from wastewater and\ntheir ecological niche relative to other nitrifying bacteria. Recent literature has demonstrated the potential for comammox bacteria\nto cooperate with anammox bacteria for efficient nitrogen conversion\nto dinitrogen gas in laboratory-scale systems. 16 − 19 This cooperation between comammox\nand anammox bacteria could take on different modalities. For instance,\ncomammox bacteria could provide nitrite to anammox bacteria through\npartial nitrification of ammonia 16 , 17 or comammox\nbacteria could perform complete nitrification to nitrate, which is\nthen converted to nitrite by denitrifying bacteria for use by anammox\nbacteria. 18 Both modalities involving comammox–anammox\nbacterial cooperation could be potentially more beneficial as compared\nto traditional strategies involving ammonia oxidizing bacteria (AOB),\nas this would minimize the potential for biotic nitrous oxide (N 2 O) production. 20 Comammox bacteria\nhave a higher affinity for ammonia compared to strict AOB, 13 , 14 which could be important in ammonia-limited environments. Further,\nsuppression of nitrite oxidizers is necessary to ensure anammox bacteria\ndo not washout of the system. 6 It has also\nbeen suggested that the cooperation between comammox and anammox bacteria\ncould be symbiotic, in which comammox bacteria assist anammox bacteria\nby providing nitrite and consuming oxygen, while anammox bacteria\nmaintain nitrite concentrations at non-inhibitory levels. 16 Thus, comammox bacteria could limit nitrite\navailability to strict\nNOB under aerobic conditions if they perform complete nitrification\nto nitrate. Studies have demonstrated comammox bacteria-associated\npartial nitrification—anammox achieved 70% nitrogen removal\nunder low dissolved oxygen (DO) conditions with suspended biomass, 17 while systems with a biofilm phase demonstrated\nnitrogen removal under both high 16 and\nlow 18 DO conditions. Further, Cui et al.\nreported the enrichment of Ca. Nitrospira\nnitrosa-like comammox bacteria in a predominantly anammox system when\noperated under microaerobic conditions. 19 Though some studies suggest comammox bacteria prefer oxygen-limited\nconditions, 6 , 18 our previous survey of comammox\nbacteria in different wastewater treatment systems did not find any\nassociation between the prevalence/abundance of comammox bacteria\nand DO concentrations. 8 To date, all four\nstudies reporting comammox–anammox co-occurrence and cooperation\nfor nitrogen removal are laboratory-scale systems. There are currently\nno reports of comammox–anammox cooperation in mainstream full-scale\nnitrogen removal wastewater systems. In this study, we report\nthe co-occurrence and cooperation of comammox\nand anammox bacteria for nitrogen removal in a full-scale integrated\nfixed film activated sludge (IFAS) system. Our previous studies at\nthe Hampton Roads Sanitation District (HRSD) James River Treatment\nPlant (JRTP) in Virginia found high abundance of comammox bacteria\nin the attached growth phase, with their concentration routinely exceeding\nthose of canonical AOB. 8 , 9 Follow-up experiments to determine\nthe nitrification kinetics of comammox bacteria indicated the potential\nfor the presence of anammox bacteria, including nitrogen loss consistent\nwith anammox stoichiometry. Thus, we systematically characterized\nthe intrinsic and extant (i.e., in situ ) kinetics\nof aerobic and anaerobic ammonia removal and the microbial community\nusing genome-resolved metagenomics to identify the nitrifying populations\nresponsible for nitrogen removal at full scale.",
"discussion": "Results and Discussion Comammox Bacteria Are the Principal Active Aerobic Ammonia Oxidizers\nin the Attached Growth Phase Aerobic intrinsic kinetics assays\nindicated that the specific ammonia oxidation rate (sAOR) was 2.5\ntimes lower for the suspended solids (0.694 mg-N/g TS-h) ( Figure SI-1 ), compared to the attached phase\n(1.75 ± 0.08 mg-N/g TS-h) ( Figure 2 A). This indicates that approximately 71% of the specific\nammonia oxidation capacity (i.e., mg-N/g TS-h) was in the biofilm,\nwhich is consistent with prior work at JRTP 47 and Broomfield Wastewater Treatment. 48 However, the intrinsic kinetic rates measured in this study were\nsignificantly lower than those previously reported from the same system. 47 Specifically, the average specific NO x production rate (sNPR) measured in this study\nwere 1.22 ± 0.09 mg-N/g TS-h ( Figure 2 A and Table SI-1 ) as compared to 2.39–5.87 mg-N/g TS-h in previous work using\nIFAS media. 47 , 48 The differing rates between systems\ncould be a result of differences in the nitrifier community composition\nas well as methods used to determine total solids in the attached\nphase. Figure 2 (A) Intrinsic rates of ammonia oxidation, nitrite plus nitrate\nproduction (NO x ), nitrate production,\nand total inorganic nitrogen loss for aerobic (uninhibited and inhibited\nwith 4 μM 1-octyne) and anaerobic ammonia oxidation and aerobic\nnitrite oxidation assays conducted using biomass attached to IFAS\nmedia. (B) Rates of ammonia oxidation, nitrate production, and total\ninorganic nitrogen loss in the aerobic zone at DO setpoints estimated\nfor full-scale experiment. Error bars denote variation across replicate\nbatch assays (A) and replicate measurements in full-scale system (B). Addition of ATU, which inhibits ammonia oxidation\nby all ammonia\noxidizers including comammox and AOB, in the kinetic assay with suspended\nsolids resulted in the complete cessation of ammonia oxidation, whereas\nammonia oxidation and corresponding NO x production occurred in the assay spiked with 1-octyne, used for\nselectively inhibiting only AOB activity, comparable to the uninhibited\nassay. Addition of ATU also resulted in the complete cessation of\naerobic ammonia oxidation with no nitrite or nitrate accumulation.\nHowever, with 4 μM 1-octyne, the average sAOR was 1.69 ±\n0.06 mg-N/g TS-h, indicating aerobic ammonia oxidation was not substantially\ninhibited ( Figure 2 A) ( p > 0.05, unpaired t -test).\nThis occurred despite irreversible inhibition of strict AOB reported\nat 1-octyne concentrations as low as 1 μM 21 with no inhibition of either ammonia oxidizing archaea\n(AOA) or comammox bacteria. 22 Our prior\nstudy suggested comammox bacteria were the dominant aerobic ammonia\noxidizers in the attached phase, while strict AOB were comparatively\nlower in abundance. 8 Thus, taken together,\nthese results suggest that comammox bacteria were likely the principal\naerobic ammonia oxidizer in the attached phase. Considering this,\nwe focused our remaining work on the attached phase microbial community. Loss of Total Inorganic Nitrogen Occurs in Both Aerobic and\nAnaerobic Ammonia Oxidation Conditions Interestingly, we\nobserved substantial total inorganic nitrogen loss (∼8%) during\nboth the uninhibited and 1-octyne spiked ammonia oxidation assays\n(TIN loss rate: 0.57 ± 0.09 mg-N/g TS-h) ( Figure 2 A) for attached phase biomass assays. This\nloss was likely not due to denitrification since the DO concentrations\nin the aerobic batch assays were maintained at 6 mg/L. To confirm\nthis, we performed aerobic nitrite oxidation assays under identical\nconditions as aerobic ammonia oxidation assays which revealed a nearly\nclosed nitrogen balance (0.6–1.2% gap in nitrogen balance)\n( Figure 2 A). This prompted\nus to investigate anaerobic ammonia oxidation as a possibile mode\nof nitrogen loss in attached phase assays. Anaerobic ammonium oxidation\nassays revealed a total inorganic nitrogen loss rate greater than\nthe nitrate produced ( Figure 2 A) (TIN loss rate: 2.15 ± 0.55 g-N/g TS-h). Further,\nthe proportion of the ammonia to nitrite consumption rate (1:1.20),\nammonium consumption to nitrate production rate (1:0.32), and rate\nof nitrogen loss (1.84) were indicative of anammox bacterial activity. 49 The capacity for both aerobic and anaerobic\nammonia oxidation has been observed in low DO (∼0.5 mg/L) bench-scale\ndemonstrations established from wastewater. 17 , 18 However, we observed a loss of total inorganic nitrogen under both\naerobic (6 mg/L) and anaerobic ammonia oxidization conditions, suggesting\nthat anammox activity may not be completely inhibited by higher DO\nconditions. This could potentially be due to anammox bacteria existing\nin oxygen-limited parts of the IFAS biofilm (biofilm thickness ranged\nfrom 200 to 600 μm) and nitrite made available by aerobic ammonia\noxidation used by anammox bacteria to drive a loss of nitrogen. 16 Though nitric oxide (NO) and nitrous oxide (N 2 O) were not measured in batch assays as possible forms of\nnitrogen loss, stoichiometric evidence strongly supports loss of total\ninorganic nitrogen was due to anammox bacteria in the attached phase.\nWhile strict AOB can produce N 2 O via NO through nitrifier\ndenitrification, differential inhibition assays indicated that comammox\nbacteria were the primary aerobic ammonia oxidizers in the IFAS media.\nIt has been demonstrated that at least one comammox bacteria species\n(i.e., Ca. Nitrospira inopinata) cannot\ndenitrify to N 2 O and produces N 2 O comparable\nto AOA, which is substantially lower than that of AOB. 20 Thus, while direct N 2 O measurements\nwould have been ideal, the evidence of comammox bacteria dominating\nammonia oxidation and their inability to produce N 2 O 20 indicates that nitrogen loss was unlikely to\nbe due to N 2 O production. Dissolved Oxygen-Dependent Nitrification and Nitrogen Loss Occur\nin the Full-Scale IFAS System Since anammox activity was\nobserved in batch assays, experiments were performed to quantify this\nactivity in situ in the full-scale system by modifying\nthe DO concentration of the aerobic zone to setpoints ranging from\n2 to 6 mg/L and monitoring of suite of process parameters ( Figure 2 B, Section SI-3, and Figures SI-2 and SI-3 ). Interestingly, total\ninorganic nitrogen loss (∼10%) was still observed despite the\nhigh DO setpoint ( Figures 2 B and SI-4 ), with the two lowest\nsetpoints, 2 and 3 mg/L, demonstrating higher rates of total inorganic\nnitrogen loss (16% nitrogen loss) ( Figure SI-4 ). While we cannot eliminate the possibility that some of this loss\ncould be due to assimilatory processes, the increase in inorganic\nnitrogen loss with a decrease in DO concentrations suggests that anaerobic\nactivity mediated by anammox bacteria was the primary mechanism ( Figure SI-5 ). The highest ammonia oxidation\nrate (1.76 mg-N/g TS-h) was obtained at a DO setpoint of 6 mg/L; this\nrate was similar to what was observed in the batch assays. Ammonia\noxidation rates at DO setpoints 2, 3, and 4 mg/L were similar to each\nother (0.870–1.03 mg/g TS-h) and were 42–51% lower compared\nto the rate observed at 6 mg/L. Comparatively, the percent ammonia\nremoved in the aerobic zone was 31, 35, 41, and 63% at DO setpoints\n2, 3, 4, and 6 mg/L, respectively ( Figure SI-4 ). The sharp decrease in ammonia oxidation rates with lowering of\nDO setpoints could be due to oxygen limitation within the biofilm. 47 , 50 However, Zhao et al. 51 recently demonstrated\na similar dramatic decrease in ammonia oxidation rates in a comammox\nenrichment moving bed biofilm reactor dominated by two Ca. Nitrospira nitrosa-like populations. Specifically,\nthey report a 50% decrease in ammonia oxidation rate with a decrease\nin DO concentrations from 6 to 2 mg/L and attribute this to the low\napparent oxygen affinity of Ca. Nitrospira\nnitrosa-like bacteria ( K o = 2.8 mg O 2 /L). This would appear consistent with our observations in\nthe full-scale system, further suggesting that comammox bacteria were\nthe primary drivers of aerobic ammonia oxidation. In this study,\nthe portion of in situ ammonia\noxidized aerobically by comammox bacteria increased with DO concentrations,\nwhile the portion oxidized by anammox bacteria was higher at lower\nDO setpoints ( Figure 3 ), indicating that lower DO concentrations reduce the aerobic ammonia\noxidation rate of comammox bacteria while simultaneously favoring\nconditions for anaerobic ammonia oxidation. This could be due to the\nhigher ammonia oxidation rates of comammox bacteria relative to their\nnitrite oxidation rates as well as the competitive advantage of NOB\nover anammox bacteria for nitrite at higher DO levels. Further, this\ndemonstrates that comammox and anammox bacteria can cooperate at a\nlow enough DO such that comammox bacteria can still make nitrite available\nfor anammox bacteria, who in turn can drive a loss of total inorganic\nnitrogen. The implications of operating at a much lower DO, such as\nthose suggested in other studies 6 , 18 (less than 1 mg/L),\nmay limit comammox bacterial ammonia oxidation such that they are\nunable to produce nitrite for anammox bacteria. Figure 3 Portion of the total\nammonia oxidation rates attributed to aerobic\nand anaerobic ammonia oxidation. Low Abundance Comammox Bacteria Co-Occur with Highly Abundant\nAnammox Bacteria in IFAS Media No comammox or strict AOB\nMAGs were assembled from samples collected during this study, which\nis in contrast to our previous assembly of two comammox and nine Nitrosomonas MAGs from this IFAS system. 9 Two amoA gene sequences in the metagenomic\nassembly were aligned using BLAST with previously assembled MAGs obtained\nfrom the same IFAS system. These amoA sequences showed\ngreater than 99 and 97% sequence identity, respectively, with the amoA genes present in one comammox and one Nitrosomonas MAGs previously assembled 9 ( Figure 4 A). Further, contigs\nobtained from the metagenomic assembly in this study were aligned\nwith BLAST against previously assembled nitrifier MAGs associated\nwith comammox bacteria, Nitrospira -NOB, and Nitrosomonas . This revealed that several contigs in this\nstudy were fully aligned (zero mismatches, 100% sequence similarity)\nto contigs within these previously assembled nitrifier MAGs, suggesting\nthat the comammox bacteria and AOB were present in the samples, but\nat very low abundances, and thus their genomes were not successfully\nreconstructed. Figure 4 (A) Maximum likelihood phylogenetic tree of the amoA genes found in the assembly (red), along with Nitrospira -comammox and Nitrosomonas amoA gene references\n(black). Black references in bold are amoA sequences\nin MAGs recovered from our previous study (2017–2018). Phylogenetic\nplacement of (B) Brocadia , (C) Nitrospira , and (D) Nitrosomonas MAGs with 90, 85, and 65\nreference genomes, respectively. Branches that are not related to\nany relevant MAG are collapsed. The complete list of the reference\ngenomes used in the analysis is in Table SI-4 . Red labels are MAGs recovered from this study; black labels are\ngenome references downloaded from NCBI; and black bold labels are\nMAGs from samples taken from 2017 to 2018 in the same IFAS system.\nMAGs selected after dereplication and used to calculate the relative\nabundances of nitrifying bacteria are marked with a red asterisk. MAGs associated with Brocadia ( n = 2) and Nitrospira ( n = 3), were\nobtained from biomass attached to IFAS media, even though anammox\nbacteria were not found in our past study 8 , 9 ( Table 1 ). Nitrospira and Brocadia MAGs represented 6.53 ± 0.34\nand 6.25 ± 1.33% of total reads in the sample, respectively.\nPhylogenomic analysis associated Brocadia -like MAGs\nwith Ca. Brocadia sapporoensis and Ca. Brocadia pituitae ( Figure 4 B), while Nitrospira -like\nMAGs were all placed in lineage I associated with Nitrospira\ndefluvii ( Figure 4 C). Two Nitrospira MAGs were very\nsimilar to three Nitrospira lineage I MAGs assembled\nfrom samples taken between 2017 and 2018 in the same IFAS system (Cotto\n2023) (ANI = 99.95% for Nitrospira_bin.66 and JAMMSM_NOB_4, 99.29%\nfor Nitrospira_bin.465 and JAMMSM_NOB_3 and 96.40% for Nitrospira_bin.465\nand JAMMSM_NOB_1). However, these Nitrospira MAGs\nwere at much lower abundances in this study (8.35 ± 1.91 RPKM)\ncompared with the previous study (55.56 ± 14.71 RPKM). 9 The decrease in Nitrospira abundance could be\nthe reason why several of the previously assembled MAGs could not\nbe assembled in the current study, despite the fact that five out\nof seven of the previously assembled Nitrospira MAGs\nhad a breadth of coverage of 90% using reads from this study ( Figure 5 A). Therefore, the\nrelative abundance of all nitrifying groups was calculated from a\nset of dereplicated MAGs recovered from both studies ( Table SI-3 ). However, only MAGs with breadth\nof coverage (i.e., percent of the genome covered by reads from this\nstudy) higher than 50% and an average coverage of 10× ( Table SI-5 ) were selected for relative abundance\ncalculations ( Table SI-6 ). The results\nconfirm the presence of most previously assembled Nitrospira MAGs (including one Nitrospira lineage II MAG)\nin the system but at much lower abundances (17.26 ± 3.61 RPKKM)\ncompared with the samples from 2017–2018 (122.90 ± 30.69\nRPKM) ( Figure 5 B and Table SI-6 ). Further, the breadth of coverage\nof previously assembled comammox (JAMMSM_CMX_1) and Nitrosomonas (JAMMSM_AOB_1) MAGs was 80.6 ± 9.8 and 72.3 ± 1.0%, respectively\n( Table SI-5 ). In conjunction with the amoA ( Figure 4 A) and contig-level analysis, this confirms the presence of previously\ndetected comammox and Nitrosomonas genomes in the\nsystem. Thus, relative abundance estimates of comammox bacteria and Nitrosomonas were calculated by mapping the reads from the\nsix IFAS samples to these previously assembled MAGs ( Figure 4 C,D). Figure 5 (A) Average breadth of\ncoverage of dereplicated MAGs in six samples\n(IFAS media pieces) taken from the aeration tank. Error bars represent\nthe standard deviation across the six samples. MAGs with breadth of\ncoverage higher than 50% were considered present in the system and\nused to calculate the relative abundances of the nitrifying groups\n(i.e., Brocadia , Nitrosomonas , Nitrospira -comammox, and Nitrospira -NOB).\n(B) Cumulative relative abundances in reads per kilobase million (RPKM)\nof Brocadia , Nitrosomonas , Nitrospira -comammox (CMX), and Nitrospira -NOB obtained from the current study samples (December 2021) and\nsamples taken between September 2017 and June 2018. (C) Average relative\nabundance of each genome in the IFAS pieces taken on December 2021.\nError bars represent the standard deviation across the six samples. Comammox and Nitrosomonas- like\nMAG relative abundances\nwere about 0.90 ± 0.8 and 0.40 ± 0.05 RPKM, respectively\n( Figure 5 C) in this\nstudy. This differs from our prior work, where comammox and Nitrosomonas relative abundances were 22 ± 6.26 and\n21.04 ± 6.17 RPKM, respectively ( Figure 5 B). Thus, it is very likely that the low\nabundance of comammox bacteria and Nitrosomonas affected\nthe assembly and binning process, which did not allow for the reconstruction\nof these genomes even though they are still present in the system.\nDespite the decrease in both comammox and Nitrosomonas relative abundances in the system, the comammox: Nitrosomonas proportion is higher in this study relative to our previous work\nin the same system. 8 , 9 These results, coupled with the\ninhibition kinetic assays with 1-octyne and drop in in situ ammonia oxidation rate with decrease in DO suggests that comammox\nbacteria are the principal aerobic ammonia oxidizers in this system.\nThe abundance adjusted ammonia consumption rate for comammox bacteria\nwas 64.19 μmol-N/mg protein-h, which is within the range reported\nfor isolated Ca. Nitrospira inopinata\n(14 μmol-N/mg protein-h) 13 and enriched Ca. Nitrospira kreftii (83 μmol-N/mg protein-h). 14 Additionally, the adjusted rate for anammox\nbacteria was 2.37 μmol-N/mg protein-h which is similar to the\nreported rate for other anammox bacteria (3.27 μmol-N/mg protein-h). 52 Anammox bacteria outnumbered comammox bacteria\nand strict AOB despite high bulk DO of the IFAS system favoring aerobic\nammonia oxidizers. While recent studies have suggested that anammox\nbacteria are most likely oxygen tolerant rather than strictly anaerobic, 53 , 54 the comparatively high abundance of anammox in the attached phase\ncould also be due to anaerobic zones deeper in the biofilm. Further,\ntranscriptional activity of anammox genes associated with Brocadia was found in aquifers with anoxic-to-oxic conditions,\nsuggesting anammox bacteria are able to contribute to nitrogen loss\nin a diverse range of oxygen environments. 55 Co-operative Nitrogen Removal by Comammox and Anammox Bacteria Comammox-anammox co-occurrence has been previously demonstrated\nin synthetic community constructs and/or lab-scale reactors using\nattached growth phases. 16 , 18 , 19 Spatial organization as a contributor to comammox–anammox\ncooperation was highlighted by Gottshall 2020, where comammox bacteria\nform a protective outer layer where oxygen was most available, while\nanammox bacteria occupy inner biofilm layers. Cooperation could also\nbe aided by their differing affinities for nitrite since comammox\nbacteria have a lower affinity for nitrite than anammox bacteria. 13 , 52 Here, co-occurrence likely occurred in the attached growth phase\nas opposed to the suspended phase because both populations are slow-growing,\nthe suspended solids retention time is too short to maintain them,\nand the biofilm gradient could support anaerobic activity. At JRTP,\nnitrite made available from ammonia oxidation by comammox bacteria\nwas used by anammox bacteria along with residual ammonia to generate\na loss of total inorganic nitrogen. Quantifying the spatial distribution\nand extent of co-localization of comammox and anammox bacteria within\nthe biofilm would have been ideal to contextualize this comammox-mediated\nnitrite provision for anammox bacteria. However, the high level of\nsimilarity between the 16S rRNA gene sequences of comammox bacteria\nand Nitrospira -NOB precludes the utilization of assays\nsuch as fluorescent in situ hybridization (FISH) for this purpose.\nIn this IFAS system, the influent to the aerobic zone contained limited\nnitrite ( Figure SI-3 ). Therefore, comammox-driven\nammonia oxidization was likely the primary source of nitrite production\nin the aerobic zone, which occurs predominantly in the attached phase\nand not in the suspended phase. One potential reason for a comparably\nlower sAOR/sNPR could be explained by the low abundance and slower\nnitrification rates of comammox bacteria. For example, Onnis-Hayden\net al (2007) estimated the relative abundance of their nitrifying\ncommunity to be about 10 and 15–20% Nitrosomonas -like ammonia oxidizers and Nitrospira -like bacteria,\nrespectively, with sNPR rates approximately 3 times higher than the\nrates observed in this study. 48 Our full-scale\nresults show loss of total inorganic nitrogen at various DO concentrations,\nsuggesting anammox bacteria were shielded from complete DO inhibition\nin aerobic environments. However, aerobic nitrification was still\nthe dominant process at each tested DO setpoint ( Figure SI-5 ). While nitrate accumulation under full-scale\nconditions demonstrates that strict NOB or comammox bacteria used\nmajority of the produced nitrite, the estimated rates suggest that\nanammox bacteria used a portion of it to drive a loss of total inorganic\nnitrogen at each tested DO concentration. The nitrite affinities \nfor Nitrospira -NOB and anammox bacteria associated\nwith MAGs in this study are similar ( Nitrospira defluvii ( K s =9 μM), 56 and Ca. Brocadia sapporoensis\n( K s =5 μM) 57 ), while the reported value for the one isolated comammox bacteria Ca. Nitrospira inopinata ( K s = 449.2 μM) 13 is much lower. Thus, Nitrospira -NOB and anammox bacteria may outcompete comammox\nbacteria for nitrite, and the decrease in nitrogen loss with increase\nin ammonia oxidation and nitrate production rates at higher DO concentrations\nsuggests the competition was oxygen dependent. To our knowledge,\nthis is the first report of a full-scale, mainstream system with co-occurring\ncomammox–anammox populations. Here, we show the potential for\ncooperation between comammox and anammox bacteria in mainstream systems\nacross a range of DO concentrations. Our results suggest that a DO-dependent\nreduction in the ammonia oxidation rate of comammox bacteria maximizes\nnitrogen loss via anammox activity, while higher DO concentrations\nresult in nitrate accumulation not only due to lower anammox rates\nbut also due to higher ammonia oxidation rates of comammox bacteria.\nThe ability to design and operate low DO processes that leverage interactions\nbetween comammox and anammox bacteria to maximize nitrogen loss via\nanammox activity has the potential to significantly reduce greenhouse\ngas emissions (N 2 O) associated with nitrogen removal from\nwastewater as well as significantly reduce energy requirements associated\nwith aeration. While we report the possibility of such a system at\nthe full scale, the precise mechanisms for the enrichment of anammox\npopulations in an aerobic reactor (routine DO concentrations = 2–3\nmg/L) treating low-strength waste (influent ammonia concentrations\n= 20–40 mg/L) and the corresponding low abundance of other\nnitrifiers, including comammox bacteria and AOB, merit further research.\nThe ability to grow in the biofilm phase on the IFAS medium and be\nprotected from DO exposure within biofilms are certainly factors that\ncould favor anammox bacterial growth. Further, the significantly lower\nabundance of populations (i.e., comammox and AOB) that serve as essential\nsources of nitrite for anammox bacteria could either be a chance occurrence\nor may suggest that the low abundance (and thus lower overall activity)\nof aerobic ammonia oxidizers is essential to ensure sufficient ammonia\navailability for anammox bacteria."
} | 7,196 |
22486781 | null | s2 | 4,393 | {
"abstract": "Biofilms are increasingly recognized as being the predominant form for survival for most bacteria in the environment. The successful colonization of Vibrio fischeri in its squid host Euprymna tasmanica involves complex microbe-host interactions mediated by specific genes that are essential for biofilm formation and colonization. Here, structural and regulatory genes were selected to study their role in biofilm formation and host colonization. We have mutated several genes (pilT, pilU, flgF, motY, ibpA and mifB) by an insertional inactivation strategy. The results demonstrate that structural genes responsible for synthesis of type IV pili and flagella are crucial for biofilm formation and host infection. Moreover, regulatory genes affect colony aggregation by various mechanisms, including alteration of synthesis of transcriptional factors and regulation of extracellular polysaccharide production. These results reflect the significance of how genetic alterations influence communal behavior, which is important in understanding symbiotic relationships."
} | 266 |
39789196 | PMC11717919 | pmc | 4,394 | {
"abstract": "Agroforestry systems are multifunctional land-use systems that promote soil life. Despite their large potential spatio-temporal complexity, the majority of studies that investigated soil organisms in temperate cropland agroforestry systems focused on rather non-complex systems. Here, we investigated the topsoil and subsoil microbiome of two complex and innovative alley cropping systems: an agrosilvopastoral system combining poplar trees, crops, and livestock and a syntropic agroforestry system combining 35 tree and shrub species with forage crops. Increasing soil depth resulted in a decline of bacterial and fungal richness and a community shift towards oligotrophic taxa in both agroforestry systems, which we attribute to resource-deprived conditions in subsoil. At each soil depth, the microbiome of the tree rows was compositionally distinct from the crop rows. We detected a shift towards beneficial microorganisms as well as a decline in putative phytopathogens under the trees as compared to the crop rows. Finally, based on our results on community dissimilarity, we found that compared to an open cropland without trees, spatial heterogeneity introduced by the tree rows in the agrosilvopastoral system translated into a compositionally less homogeneous soil microbiome, highlighting the potential of agroforestry to counteract the homogenization of the soil microbiome through agriculture. Supplementary Information The online version contains supplementary material available at 10.1038/s41598-025-85556-4.",
"conclusion": "Conclusion Soil depth strongly affected bacterial and fungal richness and community composition in both the agrosilvopastoral and syntropic agroforestry system. In addition to a decline in microbial richness from topsoil to subsoil, we observed a community shift towards oligotrophic microorganisms with increasing soil depth, as only a limited number of taxa can thrive in the resource-deprived conditions in subsoil. At both soil depths, tree rows of both systems harboured compositionally distinct microbiomes from those of the crop rows. The soil microbiome modulated by the trees showed a shift towards beneficial microorganisms (potential plant growth-promoting bacteria and EMF) while putative phytopathogens were found in lower relative abundance as compared to the crop rows. Finally, we show that compared to an open cropland system without trees, spatial heterogeneity introduced by the tree rows in the agrosilvopastoral system translated into a compositionally less homogeneous soil microbiome that became increasingly homogenous as the distance from the trees increased. Therefore, we suggest that agroforestry can counteract the homogenization of the soil microbiome through agriculture.",
"introduction": "Introduction Agroforestry is an umbrella term that describes the fusion of agriculture and forestry on the same land that is practiced by millions of farmers around the globe and includes a large variety of transformative land-use systems. In the temperate zone, silvoarable systems that simultaneously grow woody perennials and crops are gaining increasing popularity 1 , 2 . Although the intentional spatial integration of tree rows into arable land (i.e. alley cropping) through cropland agroforestry can be associated with specific challenges (e.g. resource competition between trees and crops, labor-intensive management of trees, and lack of financial incentives), it offers numerous environmental benefits as compared to open croplands without trees 3 . For example, well-designed tree rows in alley cropping systems can effectively reduce wind speeds and the associated risk for soil erosion 4 and improve microclimatic conditions 5 . Further environmental benefits include sequestration of C in the biomass of the trees as well as in soil 6 , promotion of biodiversity and related ecosystem services 7 – 9 , and improvement of soil health 10 . Another key benefit of trees in alley cropping systems is that they can reduce nutrient leaching as their deep rooting system can act as a “safety-net” for leached nutrients that are inaccessible to crops and would otherwise be lost from the system 11 , 12 . Although the “safety-net”-role of trees was first mentioned in the scientific literature decades ago 13 , this mechanism still requires further investigation 14 , 15 . Overall, integrating trees into annual cropping systems through alley cropping increases multifunctionality as compared to open croplands 3 . Multiple ecosystem functions are mediated by above- and belowground biota including microorganisms in soil 16 , 17 . Soil microorganisms constitute a substantial part of global biodiversity 18 and contribute to key soil functions such as nutrient cycling 19 . Therefore, their abundance, diversity, and functions are an integral 20 but underrepresented 21 part of soil health assessments. Temperate alley cropping agroforestry systems have consistently shown to promote the abundance, diversity, and functions of soil microorganisms as reviewed by Beule et al. 22 . As the integration of tree rows into arable land through agroforestry increases the complexity of the systems and therefore introduces spatial heterogeneity, sampling efforts in agroforestry systems are greater than in open croplands in order to capture spatial heterogeneity. This was already recognized over a quarter of a century ago 23 and found implementation in the sampling designs of numerous recent studies on soil microorganisms in alley cropping agroforestry systems as reviewed recently 22 . Commonly, spatial gradients from the trees into the crops as well as different soil depths are investigated. Realizing such a study design, several studies reported that soil microbial population size and activity were not just promoted under the trees but that the beneficial effects of the trees gradually extend into the crop rows 24 , 25 . In addition, Beule et al. 14 found that microbial communities in subsoil were stronger promoted by trees in agroforestry systems than those in topsoil. The extent to which beneficial effects of tree rows on soil microbial communities expand into the crop rows of agroforestry systems is hypothesized to be mainly depending on the distribution of below- (rhizodeposits and root litter) and aboveground tree litter inputs (leaf litter) 22 . Despite an ever-growing body of literature on soil microorganisms in temperate cropland agroforestry systems, most research efforts focused on conventionally farmed alley cropping systems that combine a single tree species (mostly poplar or walnut) with a rather narrow crop rotation of annual crops (mostly small-grain cereals or maize). This ignores the large diversity in spatial and temporal configuration and management of agroforestry systems in the temperate zone. For instance, agrosilvopastoral systems, which integrate trees with crops and livestock, are highly understudied although such integrated crop-livestock systems offer several benefits 26 . A prominent example for such benefits is the grazing of cover crops grown between cash crops, which offers economic returns without compromising soil health 27 – 29 . Grazing is also known to influence microbial communities 30 . For instance, grazing of cover crops has recently been shown to increase microbial biomass 31 . Furthermore, carefully selected tree species can benefit animal welfare through, inter alia , provision of shelter, improved digestion, and enhanced parasite control 32 . Similar to agrosilvopastoral systems, syntropic agroforestry systems, in which complex tree components are designed to mimic natural succession and stratification in order to maximize synergies among plants, are not yet in the focus of research. This is surprising given that syntropic agriculture, also referred to as dynamic or successional agroforestry, is gaining increasing popularity and is perceived as an innovative approach to agroecology 33 . Syntropic agroforestry involves a complex and diverse planting scheme that aims to mimic a forest ecosystem to create a resilient, self-sufficient, and constantly evolving system that is supposed to produce long-term yields without external inputs (e.g. fertilizer and irrigation). Due to greater plant diversity and density compared to open croplands, syntropic agroforestry systems can be expected to have a strong impact on soil microbial communities, especially through increased quality and quantity of rhizodeposition and leaf litter inputs 34 – 36 . Finally, organic agriculture, which –unlike in conventional agriculture– abandons synthetic fertilizers and pesticides, has shown a significant increase during the last 25 years 37 . However, studies investigating soil microbial communities in organically farmed temperate cropland agroforestry systems are scarce. Recently, Rosati et al. 38 collected evidence that despite challenges, the adoption of agroforestry in organic agriculture can result in multiple environmental benefits. The authors further argued that organic farmers are expected to be more open to the adoption of agroforestry as compared to their conventional counterparts 38 . As highlighted above, despite their large potential spatial and temporal complexity, the majority of studies that investigated soil life in temperate agroforestry systems is at the lower end of the spectrum in terms of system complexity. Our study aims to investigate the soil microbiome in topsoil and subsoil of two temperate alley cropping agroforestry systems (an agrosilvopastoral and a syntropic system) under organic farming in Germany. We determined the population size of soil archaea, bacteria, and fungi using real-time PCR and assess the diversity and community composition of soil bacteria and fungi by amplicon sequencing. We hypothesized that (i) microbial population size decreases with soil depth due to resource limitations in deeper soil layers. Furthermore, we hypothesized that (ii) microbial populations size increases with decreasing distance to the trees due to increased tree litter inputs (rhizodeposition and leaf litter) in close proximity to the trees. We further expected that (iii) microbial communities within topsoil have greater community similarity than those within subsoil as a result of homogenization of topsoil communities due to soil management and that (iv) tree rows promote beneficial microorganisms rather than putative plant pathogens.",
"discussion": "Discussion Despite the large diversity of temperate alley cropping systems, so far most research focused on rather non-complex systems in terms of management and tree species composition. In this work, we investigated the topsoil and subsoil microbiome of two complex alley cropping systems, namely, an agrosilvopastoral and a syntropic system. Overall, we found that the richness of bacterial and fungal taxa declined from topsoil to subsoil in both agroforestry systems, which we attribute to resource-deprived conditions in subsoil. Accompanied with the decline in taxa richness, we detected a community shift for both bacterial and fungal communities from topsoil to subsoil. In both agroforestry systems, the tree rows harboured bacterial and fungal communities that were compositionally distinct from those in the crop rows. Furthermore, in the agrosilvopastoral system, the soil microbiome became compositionally less homogeneous with decreasing distance to the trees. Patterns of SOC in two young and differently managed alley cropping systems SOC in topsoil and subsoil of the agrosilvopastoral system remained largely unaffected by the integration of trees (Fig S1 A, B), which is likely due to the young age of the system (4 years). Indeed, recent literature suggests that increases in SOC stocks in alley cropping systems are usually detectable after one decade post their establishment 65 . In contrast, SOC under the trees in the syntropic system was greater than in the crop row although both systems are of the same age. The intensive mulching of tree rows most likely caused the rapid increase in SOC under the trees. Although we did not determine different SOC fractions, we expect that mainly labile SOC fractions increased through mulching as these fractions can respond rapidly to environmental changes. Furthermore, labile SOC fractions are considered indicators of long-term SOC trends 66 , however, their fast turnover and sensitivity towards environmental changes also makes them vulnerable to loss 67 . In a case study from 2015, Cardinael and others 68 suggested that the additional SOC storage under the trees of an 18-year-old walnut alley cropping system in France was mostly due to an increase in labile SOC fractions, raising questions regarding its stability. We therefore suggest that future research on SOC stocks in alley cropping systems should consider differentiating between labile and stable SOC fractions in order to infer information on SOC stabilization dynamics in alley cropping systems. Soil depth as a key driver of the abundance and community composition of bacteria and fungi Soil depth is a key driver of the assembly of the soil microbiome since multiple influencing factors such as the availability of oxygen and other resources generally decrease as soil depth increases. The scarcity of resources in subsoil serves as a reasonable explanation for the observed decline in population size of bacteria, fungi, and archaea with increasing soil depth (Fig. 2 ) that confirms our first hypothesis and is congruent with previous studies 69 – 72 . Furthermore, specialization to resource-deprived conditions in subsoil also explains the decline in bacterial species richness from topsoil to subsoil of the agrosilvopastoral system (Fig. 3 ) as only a limited number of taxa can thrive in subsoil 72 . In addition to changes in alpha diversity, soil depth led to compositionally distinct bacterial communities in both agroforestry systems. For instance, multiple phyla associated with an oligotroph lifestyle such as Chloroflexi , Verrucomicrobiota , Gemmatimonadota , Nitrospirota , and Latescibacterota showed greater relative abundance in subsoil than topsoil, highlighting their ability to adapt to resource-deprived soil environments. Similarly, in the agrosilvopastoral system, Gaiella and Nocardioides spp. as well as their affiliated phylum Actinobacteria showed a similar pattern of promotion in subsoil, which is in line with their presumed oligotrophic lifestyle. However, it should be noted that the generalized assignment of a single lifestyle to a broad taxonomic group may not match the lifestyles of all its members 73 , 74 . Although our results for certain phyla mentioned above ( Actinoabacteria , Chloroflexi , Nitrospirota , and Latescibacterota ) are in line with previous studies that investigated bacterial communities at different soil depths 69 , 75 , Lopes et al. 76 reported opposing results for Verrucomicrobiota and Gemmatimonadota (i.e. greater relative abundance in topsoil than subsoil). For instance, the authors found the genus Candidatus Udaeobacter ( Verrucomicrobiota ) to be depleted in deeper soil layers and attributed this finding to its aerobic lifestyle 76 , 77 . However, in the agrosilvopastoral system, relative abundance of Candidatus Udaeobacter was greater in subsoil than topsoil (Fig. 6 D). Such a discrepancy may partly be due to differences in sampling depth: while Lopes and co-authors 76 sampled up to 240 cm deep, subsoil samples in our study were collected at 30–60 cm depth, which likely did not cover anaerobic conditions. Likewise, in 2015, Navarette et al. 78 reported decreasing relative abundance of Verrucomicrobiota with increasing soil fertility, which coincides with the observed increased proportion of this phylum in resource-deprived subsoil (Fig. 6 A, B). In contrast to bacteria, only very few fungal phyla and genera were differentially abundant between topsoil and subsoil; however, in the agrosilvopastoral and syntropic system, members of Mortierellomycota and Mortierella showed greater relative abundance in subsoil than topsoil, respectively (Fig. 6 C, F). Although information on the vertical distribution of Mortierellomycota and Mortierella spp. in agricultural soils is scarce, Mortierella spp. are widely recognized as important decomposers as well as for their benefits as plant-growth-promoting fungi 79 , 80 . Another fungal genus that was dominant in subsoil of the agrosilvopastoral system was Filobasidium (Fig. 6 F). The genus Filobasidium harbours members with the ability to produce extracellular polymeric substances, which enable them to form biofilms 81 , a survival mechanism by which these organisms may cope with the resource-deprived conditions in subsoil. In addition to a distinct community composition between soil depths, community similarity of bacteria and fungi was greater within topsoil than subsoil (Fig. 5 ), revealing greater compositional homogenization of the topsoil than the subsoil microbiome and thereby confirming our third hypothesis. As our study sites are located in a glacial landscape that is characterized by small-scale spatial variability of soil properties, we expected a comparatively large dissimilarity in microbiome composition among samples. Although our study sites were managed under reduced tillage for five years, the previous decade-long tillage regime likely reduced the degree of spatial variability of soil properties in topsoil through mechanical homogenization while subsoil remained undisturbed. We therefore propose that spatial homogenization of topsoil through tillage translated into compositional homogenization of topsoil biota. Likewise, we argue that greater spatial variability of soil properties in deeper soil layers contributed to greater community dissimilarity in subsoil as compared to topsoil. Distance to trees shapes the soil microbiome in agroforestry systems The distance to the tree rows influenced bacterial and fungal community composition in both systems. For bacterial communities, most striking effects of the trees were observed in subsoil of the syntropic system: Actinobacteria and its affiliated genera Microbacterium , Microlunatus , and Streptomyces showed greater relative abundance under the trees than in the crop row. Members of these groups are often found in rhizosphere and include species that are recognized for their antimicrobial properties as well as their plant growth promoting effects 82 , 83 . We found similar effects for Proteobacteria and its affiliated genera Sphingomonas and MND1 that are also frequently recovered from rhizosphere with some strains showing beneficial effects on plant growth 84 – 86 . The syntropic system is characterized by high diversity of plants, which may explain the promotion of putative beneficial bacteria that are known for root interactions. We expect the aboveground plant diversity and density to be mirrored belowground through the formation of a complex rooting system along the soil profile that creates numerous habitats for soil organisms. In line with this assumption, we observed greater mean population size of bacteria, fungi, and archaea in the tree row than the crop row of the syntropic system, partly confirming our second hypothesis (Fig. 2 ). The promotion of microbial population size was likely due to increased root biomass and root exudates as plant diversity increased as shown previously for bacteria and fungi in a microcosm experiment 35 . In topsoil of the agrosilvopastoral system, trees specifically promoted ectomycorrhizal fungi (EMF) of the genera Hebeloma , Laccaria , and Tuber (Fig. 7 M). This finding is not surprising considering the symbiotic association that EMF form with poplar trees and agrees with our previous findings 87 . Furthermore, for Hebeloma and Tuber spp., the promotion through the trees was also detected in subsoil, indicating that the formation of beneficial tree–microbe relationships extended into deeper soil layers. In contrast, the genus Fusarium , which harbours numerous economically relevant plant pathogens, showed a pattern of lower relative abundance under the trees than in the crop row and open cropland in topsoil. While this pattern could simply emerge due to the absence of their host plants in the tree rows, increased microbial antagonism as well as enhanced biological control through improved soil faunal activity under the trees as proposed by Vaupel et al. 87 may have contributed to the observed pattern. Together with our results on potentially beneficial bacteria and EMF, lower abundance of Fusarium spp. under the trees than the crops confirms our fourth hypothesis that trees promote beneficial and symbiotic microorganisms rather than putative phytopathogens. The introduction of tree rows in arable land through agroforestry is frequently described as a diversification measure that increases spatial heterogeneity by enabling biological interactions between trees and crops 88 . Consequently, biological interactions within alley cropping systems can be expected to be spatially depended on the distance from the trees into the crop rows. In order to capture such spatial heterogeneity, sampling soil along linear transects spanning from the tree rows into the crop rows is advisable. Using transect sampling within the agrosilvopastoral systems, we here for the first time report that (i) tree rows in a temperate agroforestry system not only promote microbial community dissimilarity in both topsoil and subsoil but that (ii) community dissimilarity decreased with increasing distance from the trees into the crop rows and the open cropland (Fig. 5 ). In other words, compared to the open cropland system, spatial heterogeneity introduced by the tree rows in the agrosilvopastoral system translated into a compositionally less homogeneous soil microbiome that became increasingly homogenous as the distance from the trees increased. However, we did not observe this pattern in the syntropic system, which we expect to be due to the intensive mulching of the tree rows that has a strong impact on soil microbial communities 89 , 90 . Nevertheless, our finding on community dissimilarity in the agrosilvopastoral system is of particular interest given that land-use change from natural ecosystems to agriculture has recently been shown to result in taxonomic homogenization of soil microorganisms 91 . Agroforestry practice may counteract the homogenization of the soil microbiome through agriculture."
} | 5,635 |
35964442 | null | s2 | 4,395 | {
"abstract": "Understanding the microscopic driving force of water wetting is challenging and important for design of materials. The relations between structure, dynamics and hydrogen bonds of interfacial water can be investigated using molecular dynamics simulations. Contact angles at the alumina (0001) and (112‾0) surfaces are studied using both classical molecular dynamics simulations and experiments. To test the superhydrophilicity, the free energy cost of removing waters near the interfaces are calculated using the density fluctuations method. The strength of hydrogen bonds is determined by their lifetime and geometry. Both surfaces are superhydrophilic and the (0001) surface is more hydrophilic. Interactions between surfaces and interfacial waters promote a templating effect whereby the latter are aligned in a pattern that follows the underlying lattice of the surfaces. Translational and rotational dynamics of interfacial water molecules are slower than in bulk water. Hydrogen bonds between water and both surfaces are asymmetric, water-to-aluminol ones are stronger than aluminol-to-water ones. Molecular dynamics simulations eliminate the impacts of surface contamination when measuring contact angles and the results reveal the microscopic origin of the macroscopic superhydrophilicity of alumina surfaces: strong water-to-aluminol hydrogen bonds."
} | 339 |
39420145 | PMC11487260 | pmc | 4,396 | {
"abstract": "Diverse microbes in nature play an important role in ecosystem functioning and human health. Nevertheless, it remains unclear how microbial communities are maintained. This study proposes that evolutionary changes in the pH niche of bacteria can promote bacterial coexistence. Bacteria modify the pH environment and also react to it. The optimal environmental pH level for a given species or pH niche can adaptively change in response to the changes in environmental pH caused by the bacteria themselves. Theory shows that the evolutionary changes in the pH niche can stabilize otherwise unstable large bacterial communities, particularly when the evolution occurs rapidly and diverse bacteria modifying pH in different directions coexist in balance. The stabilization is sufficiently strong to mitigate the inherent instability of system complexity with many species and interactions. This model can show a relationship between pH and diversity in natural bacterial systems.",
"introduction": "Introduction How diverse organisms coexist in natural ecosystems remains a longstanding puzzle in ecology. Ecological theory suggests that communities with numerous interacting species are inherently unstable 1 , posing a challenge to the maintenance of species diversity. While efforts to address this “diversity–stability” problem have primarily focused on macro-organisms such as animals and plants 2 , 3 , microbial communities—comprising a vast array of species essential to ecosystem function 4 – 8 —present a unique yet underexplored avenue for understanding stability mechanisms 9 – 12 . Microbial communities exhibit distinctive features that differentiate them from macro-organisms. One notable feature is the feedback loops between their metabolic activities and environmental conditions 13 – 16 . This interaction is particularly evident in the modulation of pH, a crucial environmental factor influencing microbial physiology and community dynamics 17 – 22 . Some bacteria actively modify their surrounding pH through metabolic processes 17 , 23 , 24 , creating conditions that influence the growth and survival of coexisting species. At the same time, because pH is a key niche for bacteria, interspecific competition can occur if competing bacteria prefer a similar pH environment 25 , 26 . Another distinctive feature of microbial communities is their rapid adaptation rates 15 . Microbes have generation cycles much faster than those of most animals and plants, enabling them to swiftly evolve in response to environmental fluctuations 27 – 29 . This rapid evolutionary adaptation plays a pivotal role in shaping ecological population dynamics and promoting species coexistence 30 – 35 . Bacterial species can adapt their pH preferences, or “pH niches,” in response to environmental changes 36 – 40 , thereby maintaining ecological balance despite potential fitness costs 41 . Considering these two major features in bacteria—the feedback between environmental conditions and metabolic activities, and their rapid evolutionary adaptation—a hypothesis emerges: The dynamics of bacterial communities are driven by the changes in pH environment caused by themselves and their evolutionary adaptation in response to these pH changes. In this study, I delve into the eco-evolutionary dynamics of bacterial communities, focusing on two distinct groups of bacteria based on their different abilities to change pH (Fig. 1 ). Alkaliphilic bacteria, such as Clostridium perfringens , Corynebacterium ammoniagene , and Pseudomonas veronii , increase pH levels, while acidophilic bacteria, such as Bifidobacterium , Lactobacillus plantarum , and Serratia marcescens , decrease pH levels 42 – 44 . The interaction between these two types of groups, mediated by their pH-modifying capabilities and adaptive responses, forms the core of the investigation into bacterial community stability (“Methods”). The goal of this study is to reveal how the pH niche adaptation in response to pH change caused by themselves affects to the stability of complex microbial communities with alkaliphilic and acidophilic bacteria. The stability is evaluated by community persistence which is the probability that all species persist for a given time (“Methods”). By elucidating how pH niche adaptation influences the persistence and dynamics of complex bacterial communities, this study aims to contribute to our understanding of fundamental ecological principles governing microbial ecosystems. Through a theoretical approaches, I explore the mechanisms underlying community stability, providing insights into how bacterial communities maintain biodiversity and ecosystem function in diverse environments. Fig. 1 Schematic illustrations of the model. a Ecological relationships between bacterial species and pH. The circles, excluding the one representing pH, denote different bacterial species. Typical arrows indicate promotion, while hammerhead arrows indicate inhibition. Black arrows represent effects due to bacteria, and red arrows represent effects due to pH. Acidophilic bacteria lower pH (inhibit), while alkaliphilic bacteria increase pH (promote). pH impacts bacterial growth either positively or negatively, depending on the relationship between the current pH value and the bacterial pH niches. Even if bacterial species have similar pH niches, they may either compete for resources (depicted between pink and blue) or not. b Effects of key factors on the fitness of a bacterial species. The blue, orange, and green curves illustrate the cost, pH, and competition functions, respectively. Note that the cost and pH functions are positive, while the competition function is negative. The dashed lines in each color (blue, orange, green) represent higher cost, increased pH sensitivity, and narrower niches, respectively. Blue, orange, red, and green arrows indicate the physiological optimal pH, the environmental pH value, the niche position of the focal species (sp. 1), and the niche position of a competitor (sp. 2), respectively. The three red circles, from bottom to top, represent the effects of cost, pH, and competition on the fitness of species 1 at its pH niche. Bacteria can shift their pH niche if a mutant with a slightly different pH preference shows higher fitness compared to the wild type (“Methods”).",
"discussion": "Discussion Evolutionary changes in pH niches play a crucial role in maintaining communities with both alkaliphilic and acidophilic bacteria, especially when adaptation occurs rapidly. Faster adaptation helps uphold these communities, except when there is a significant imbalance in the composition or ratio of alkaliphilic to acidophilic bacteria. Communities tend to persist when species converge to adapt to the pH environment influenced by the majority of bacteria types. Conversely, divergent evolution towards vastly different pH niches undermines community persistence. The stabilizing effect of rapid evolution effectively mitigates the inherent destabilization associated with increased community complexity or species diversity. These findings underscore the self-sustaining nature of microbial communities with alkaliphilic and acidophilic bacteria, adapting to pH changes induced by the community itself. Divergent or convergent evolution in pH niches is influenced by the predominant bacterial types shaping the pH environment. When a majority of bacteria create a pH environment biased towards one direction, there is a tendency for other species to adapt their pH niches to match the majority. This matching can be advantageous initially, as it allows species to thrive in the prevailing pH conditions. However, it also intensifies interspecific competition by creating niche overlaps among bacterial species. Species that are weaker or have a suboptimal pH niche, characterized by a low growth rate or an optimal pH value distant from the majority, face challenges in surviving. If these species cannot adapt quickly enough to the pH environment set by the majority, they may fail to persist (Fig. S9 ). Conversely, even if they adapt to match the majority, they may still struggle due to strong competition from numerous competitors. In such a dilemma, species with faster adaptation rates can simultaneously match their pH niche to the majority and partition their niche space, even among weaker or less well-adapted species. This process promotes convergent evolution towards similar pH niches within the community. On the other hand, slow adaptation rates (noting that evolution in weaker or less well-adapted species progresses slower due to smaller fitness gradients) hinder species from effectively matching the pH niche of the majority and maintaining distance from majority pH niches. Slow adaptation makes it challenging to finely tune pH niches to avoid niche overlaps. Even with significant niche partitioning from the majority, weaker or less well-adapted bacteria struggle to survive in an unfavorable pH environment. This underscores the critical role of adaptation speed in determining whether species converge towards similar pH niches or maintain distinct, potentially divergent niches. Empirical works have shown that soil pH is a major driver of the soil communities 40 , 45 – 49 . The soil pH significantly explains the community composition and diversity of soil bacterial communities. Soil communities can show different patterns in the relationship between pH and diversity. In addition to peaking around nearly neutral pH 50 , 51 , soil communities shows an increasing community composition and diversity with an increase in pH 45 . The theory proposed here suggests that communities tend to be stable under nearly neutral pH conditions, particularly when the community composition is balanced between alkaline-producing and acid-producing bacteria, and there is high pH sensitivity and a substantial cost associated with trait changes (Fig. S2 ). Therefore, such conditions are crucial for the peak diversity observed around nearly neutral pH. Furthermore, this theory posits that when the community composition is biased towards one type of bacteria, the peak diversity shifts towards a particular pH direction (Fig. 5 , Fig. S2 ). Such biases may lead to another pattern where pH increases are associated with an increasing bias towards one bacterial type. The present study also predicts that under conditions of nearly balanced community composition and moderate pH sensitivity, two peaks of diversity can emerge (Fig. S2 ). Considering these complex interrelationships and feedback loops, the question arises: what drives the relationship between pH and diversity? The present study suggests that adaptation to pH changes plays a pivotal role in maintaining species diversity. Indeed, empirical evidence has shown that bacterial species capable of acid tolerance thrive in strongly acidic environments, providing them with a competitive advantage over other bacteria 45 . The ability of species to persist or decline is largely determined by their speed of adaptation to environmental changes. pH adaptation could play a key role in bacterial community dynamics. In a healthy state, the dental community hosts numerous bacteria capable of alkalinizing acids and maintaining health by neutralizing them. However, when there is a disturbance (resulting in a decrease in pH), acid-producing bacteria and those adapted to acidic conditions form a community that excludes others, creating a lower pH environment and preventing dental caries 52 , 53 . Such adaptations to pH changes in communities, along with alterations in their composition, are also observed in bacteria found in wastewater 54 , soil 55 , and coral reefs 56 , 57 . Moreover, studies investigating the stability of microbial communities in response to pH disturbances have indicated that microbial diversity is linked to stability 58 . However, in the present study, it was found that species diversity alone does not contribute to stability unless the rate of adaptation is rapid. Therefore, research that tracks adaptive changes 59 is crucial to test this theoretical prediction. It is possible to test this theory in both experimental and field studies. In a simple experimental study with various types of bacteria species on the metabolic properties of pH, each bacteria species largely changes the pH value by themselves in the environment. In addition, it also largely affects to the population dynamics, resulting in coexistence or competitive exclusion. However, the experiment is based on the ecological dynamics in which evolution does not occur 12 , 44 . Hence, if we can control the evolution of traits on pH niche (such as preference or tolerance) and trace the trait dynamics in diverse bacteria species, it can be made close to the present model situation. If it is possible, the destabilization effect of strong interactions among bacteria species (high value of pH change rates) shown by an experimental study 12 may be strongly weakened by rapid pH adaptation. On the other hand, in the field study, the situation is more complicated. In addition to a similar examination with the lab experiment, it needs to examine what causes pH change and how much the microbes affect to the pH dynamics in the broad environment. It also needs to know the metabolic properties of pH in each diverse species. A nonlinear time series analysis 60 can help to capture the effects of each species on pH dynamics. The model assumed a particular type of microbe. Bacterial growth tends to be favored within optimal growing conditions, and bacteria are capable of altering their surrounding pH. However, in nature, there are diverse types of microbes. For instance, under stressful environments, microbes can accelerate their growth rates 61 . Some microbes possess mechanisms to adapt to pH changes without altering their pH niche 41 . Furthermore, various types of interactions are present in microbial communities 62 . Understanding and integrating these diverse microbial characteristics and interactions for community stability will be a major goal in future studies. The present study shows that the speed of evolution and community composition (alkaline producing vs acid producing) play a key role in maintaining large bacterial communities. The speed of evolution is related to genetic diversity, and the community composition is related to the diversity of metabolic properties among species. The present findings suggest that species diversity is maintained by genetic diversity and metabolic diversity."
} | 3,632 |
37578829 | PMC10482373 | pmc | 4,397 | {
"abstract": "In Gram-positive bacteria such as \n \n Staphylococcus aureus \n \n and the coagulase-negative staphylococci (CoNS), the accessory gene regulator ( agr ) is a highly conserved but polymorphic quorum-sensing system involved in colonization, virulence and biofilm development. Signalling via agr depends on the interaction of an autoinducing peptide (AIP) with AgrC, a transmembrane sensor kinase that, once phosphorylated activates the response regulator AgrA. This in turn autoinduces AIP biosynthesis and drives target gene expression directly via AgrA or via the post-transcriptional regulator, RNAIII. In this review we describe the molecular mechanisms underlying the agr -mediated generation of, and response to, AIPs and the molecular basis of AIP-dependent activation and inhibition of AgrC. How the environment impacts on agr functionality is considered and the consequences of agr dysfunction for infection explored. We also discuss the concept of AIP-driven competitive interference between \n \n S. aureus \n \n and the CoNS and its anti-infective potential.",
"conclusion": "Concluding remarks The function and dysfunction via inhibition or mutation of the agr system play key roles in the adaptive behaviour and fitness of \n \n S. aureus \n \n in different host environments, in response to antibiotics and during encounters with other staphylococcal species especially on the skin. Considerable progress in understanding the molecular mechanisms underlying agr- dependent QS has been gained although there is still relatively little high-resolution information on the 3D structures of AgrC, AgrB and MroQ or on the nature of the respective AIP and AgrD substrate-binding sites on each of these transmembrane proteins. Whether MroQ processes its thiolactone substrate on the inner or outer face of the cytoplasmic membrane and how AIPs are exported remain to be elucidated. Given the diversity of AIPs leading to extensive competitive interference, it is perhaps surprising that there is little cross-activation within or between staphylococcal species. Whether AIP-driven interference occurs between staphylococci and other Gram-positive bacteria that harbour agr systems especially in complex microbial communities is not yet known. We also have little understanding of the selective pressures that led to the emergence of different agr groups or indeed why four agr groups evolved in \n \n S. aureus \n \n with no evidence of intermediates. Given the rise of multi-antibiotic-resistant MRSA, the agr system does offer multiple druggable molecular targets for inhibitors including AgrA, AgrB and AgrC [ 18 ]. However, despite the in vitro identification of highly active AIP analogues and non-peptidic agr inhibitors much more in vivo work will be required before any ‘hits’ with the appropriate pharmacological and pharmacokinetic properties can enter human clinical trials. These are likely to be complicated by the association between agr dysfunction, biofilm formation and chronic infections. Nevertheless, employing CoNS strains or antagonistic AIPs as a means of limiting \n \n S. aureus \n \n skin infections appears to offer a realistic therapeutic opportunity [ 141 ].",
"introduction": "Introduction Coagulase-negative staphylococci (CoNS) such as \n \n Staphylococcus epidermidis \n \n and \n \n Staphylococcus hominis \n \n are primarily skin commensals while the coagulase-positive \n \n Staphylococcus aureus \n \n is not only a commensal colonizing human nares and skin but also a major opportunistic pathogen [ 1, 2 ]. \n \n S. aureus \n \n infections can be minor or invasive and life-threatening [ 3 ]. They may be acute or chronic ranging from skin, soft tissue and medical device-related to bacteraemia, endocarditis, osteomyelitis, food poisoning, septic arthritis, scalded skin and toxic shock syndromes [ 3 ]. \n \n S. aureus \n \n is also on the WHO ‘ESKAPE’ list of multi-antibiotic resistant priority pathogens given that treatment of \n \n S. aureus \n \n infections has been compounded by the emergence of multi-antibiotic-resistant strains [ 4 ]. These include vancomycin-resistant (VRSA and VISA) and methicillin (MRSA)-resistant isolates, which can be further sub-divided into hospital-acquired MRSA (HA-MRSA) and community-acquired (CA-MRSA) strains. \n \n S. aureus \n \n has often been reported as a co-infecting microbe in polymicrobial infections where it may be co-operative or competitive [ 5 ]. In the context of difficult-to-eradicate infections such as non-healing diabetic foot ulcers and in the lungs of individuals with cystic fibrosis (CF), co-infections of \n \n S. aureus \n \n with P. aeruginosa are indicative of much poorer clinical outcomes than in those infected with either species alone [ 6 ]. As both commensal and pathogen, \n \n S. aureus \n \n is capable of rapidly responding and adapting to fluctuating host and inanimate environments and switching between colonization and pathogenic modes. Such adaptation depends on local extracellular signals that include oxygen availability, temperature, pH, nutrient limitation, bacterial cell population density, other co-localizing microbes, antimicrobials and is associated with specific host tissue interactions, implanted medical devices, abiotic surfaces encountered in households or hospitals as well as particulate air pollution [ 7–9 ]. The versatility of \n \n S. aureus \n \n as a pathogen revolves around diverse cell-wall colonization factors, immune modulating agents and secreted exoproducts [ 3 ], many of which are controlled via quorum sensing (QS), a mechanism for synchronizing gene expression via self-generated diffusible signal molecules in a population-dependent manner [ 10 ]. In Gram-positive pathogens including the staphylococci, clostridia, enterococci and listeria, QS is mediated by the accessory gene regulator ( agr ) system [ 11, 12 ]. In \n \n Staphylococcus aureus \n \n , agr reciprocally regulates multiple cell-wall proteins (e.g. immunoglobulin and fibronectin-binding proteins) and exotoxins (e.g. haemolysins, enterotoxins, leucocidins, toxic shock syndrome toxin, exoenzymes (nucleases, proteases, lipases) and the phenol soluble modulins (PSMs), a family of short amphipathic peptides with cytolytic activity similar to δ-toxin [ 3 ] ( Fig. 1 ). Staphylococcal surface proteins promote adherence to host tissues and aid immune evasion while exotoxins cause tissue damage and many function as super-antigens promoting the onset of shock-like syndromes [ 3, 13 ]. The S. aureus agr system is involved in endosomal escape, intracellular survival and replication [ 14–16 ]. With respect to biofilm development, agr contributes to initial attachment, structuring and dispersal [ 17 ]. Fig. 1. The agr QS system in \n \n S. aureus \n \n negatively (-) regulates the production of capsular polysaccharides and multiple cell-wall proteins involved in host protein interactions including immunoglobulin, fibronectin and fibrinogen binding proteins. agr positively (+) regulates the expression of diverse virulence factor genes including those coding for tissue degrading exoenzymes, haemolysins, enterotoxins, exfoliative toxins leukocidins and phenol soluble modulins. In acute animal models of skin, soft tissue, respiratory, arthritis and bone, S. aureus agr mutants are attenuated, highlighting a key role for QS-dependent regulation of virulence determinants at these infection sites [ 18 ]. Paradoxically, agr is required for skin colonization [ 19 ]. However, a functional agr system is dispensable for chronic, biofilm-related infections associated with, for example, implanted medical devices and cystic fibrosis [ 18 ]. Furthermore, allelic variation in agr genes contributes to intra- and inter-staphylococcal competition since the cognate agr signal molecule of one staphylococcal agr variant may inhibit agr in a strain possessing a different agr variant [ 20–22 ]. In this review we outline the intricate molecular mechanisms underlying the agr -mediated generation of, and response to, autoinducing peptide (AIP) signal molecules focusing primarily on \n \n S. aureus \n \n . We describe the specifics of AIP-activation and inhibition of the sensor kinase AgrC, how the environment impacts on agr functionality and explore the consequences of agr dysfunction for infection. We also discuss the concept of AIP-mediated competitive interference between \n \n S. aureus \n \n and the CoNS and its therapeutic potential for suppressing \n \n S. aureus \n \n skin disease and other infections. The S. aureus agr system In \n \n S. aureus \n \n the agr locus consists of two divergent transcriptional units, agrBDCA and the regulatory RNA effector, RNAIII controlled by the P2 and P3 promoters, respectively [ 20, 23 ]. ( Fig. 2a ). The P2 operon consists of four genes, agrBDCA , which are required for the activation of transcription from the P2 and P3 promoters while the P3 transcript, RNAIII, a 517-nucleotide transcript, is itself the primary effector for the agr response and also encodes δ-haemolysin [ 24 ]. AgrA and AgrC constitute a two-component system in which the transmembrane protein AgrC is the histidine sensor kinase and the cytoplasmic AgrA protein is the response regulator [ 23, 25 ]. The diffusible agr QS signal is an autoinducing peptide (AIP) ( Fig. 2 ) derived via the AgrB-dependent proteolytic processing of the ribosomally synthesized AgrD pro-peptide. AIPs induce the trans -auto-phosphorylation of AgrC, which transfers the phosphate to a conserved Asp on AgrA. This binds to the agrP2 promoter upregulating agrBCDA , conferring a positive-feedback loop that autoinduces AIP production and drives target gene expression directly via AgrA or via the AgrA-dependent agr P3 promoter and the post-transcriptional regulator, RNAIII [ 24 ] ( Fig. 2a ). There are four allelic variants (I–IV) of the S. aureus agr locus with respect to a hypervariable region contained in the agrB , agrD and agrC genes ( Fig. 2b, c ). These reflect the amino acid sequence variations in the four \n \n S. aureus \n \n AIPs and explains their specificity for AgrC as activators of their cognate receptors but competitive inhibitors of the other AgrC variants [ 20, 26–31 ]. For example, AIP-1 is a competitive antagonist of the AIP-2/AgrC2 and AIP-3/AgrC3 interactions whereas AIP-2 and AIP-3 antagonize AIP-1/AgrC1 interactions ( Fig. 2c ). AIP-4, which differs from AIP-1 by a single amino acid residue (Asp is replaced by Tyr; Fig. 2c ), is an agonist of both AgrC4 and AgrC1 but an inhibitor of AgrC2 and AgrC3 [ 28, 29 ]. Recently Raghuram et al . [ 32 ] developed a software tool (AgrVATE; github.com/VishnuRaghuram94/AgrVATE ) to interrogate S. aureus agr variability and evolution. Analysis of ~40 000 \n \n S \n \n . \n \n aureus \n \n genomes revealed that the distribution of the four agr groups is ~60 % group I, ~22 % group II, ~14 % group III and ~2.5 % group IV [ 32 ]. This extensive in silico analysis did not uncover any novel agr groups, intermediate AIPs or AIPs acquired from CONS. Potential relationships between certain infectious diseases and agr group have been highlighted such as between toxic shock and scalded skin syndromes and agr groups III and IV, respectively [ 33 ]. However, the number of strains examined was relatively small and others [ 34 ] found no associations between agr -specific groups and infection type. A recent major study found that the distribution of agr groups in over 10 000 strains from blood, skin and the nasal cavity was similar to the general distribution of agr groups [ 32 ]. agr acts as an AIP-dependent autoinducible system such that mutation of any of the corresponding agrBDCA genes results in the loss of activity [ 20 ]. As a QS-dependent master virulence gene regulator, it is also subject to control via by a sophisticated interconnected network of regulators, which integrates diverse environmental and host cues via two component sensor regulator systems such as SaeRS, SrrAB and ArlRS, sigma factors (e.g. SigB) and the SarA family (e.g. SarA, SarR, SarS, MgrA and Rot). A detailed consideration of the molecular nature and function of these gene regulatory elements in controlling agr is beyond the scope of this review and the reader is referred to recent reviews [ 23, 24 ]. Fig. 2. ( a ) Schematic of the staphylococcal agr quorum-sensing system. The agrBDCA locus is composed of two divergent transcripts, RNAII and RNAIII, driven by the agr P2 and agr P3 promoters, respectively. AgrD, the pro-peptide precursor of the autoinducing peptide (AIP) is processed at the cytoplasmic membrane by AgrB and MroQ such that AIPs are released extracellularly. AIPs bind to and activate the AgrC receptor, a membrane-bound histidine sensor kinase resulting in phosphorylation of the response regulator AgrA and activation of the agrP2 and agrP3 promoters. This drives the autoinduction circuitry to generate more AIP signal molecules and induces expression of virulence genes either indirectly via RNAIII or directly via the target gene promoters. ( b ) Schematic showing the hypervariable region (in green/cyan) of the agrBDCA locus incorporating agrD and giving rise to the different agr groups. Amino acid residues marking the beginning and ends of the variable regions are numbered (adapted from [ 142 ]). ( c ) Generalized AIP structure and summary table showing the amino acid sequences of the AIPs belonging to each of the four S. aureus agr groups and their cross-group activities. The brackets denote the amino acid residues within the macrocycle. In \n \n S. aureus \n \n , the AIP N-terminal tails have two, three or four amino acid residues. AIP identification and quantification In the staphylococcal agr system, AIP signal molecules are thiolactones, which have similar structures but different primary amino acid sequences. Each AIP has a common central Cys residue, the thiol of which is linked to the α-carboxyl group of the C-terminal residue forming a five residue, 16-membered macrocycle with an exocyclic N-terminus of variable length ( Fig. 2c ). These have been identified by liquid-chromatography (LC) mass spectrometry (MS) aided by the annotated AgrD amino acid sequence data [ 27, 28, 35, 36 ] or by chemo-selective trapping [ 37 ]. AIPs have been produced by solid-phase chemical synthesis [ 28, 38 ] and via protein engineering using mini-intein technology [ 39 ]. They can readily be detected at nanomolar concentrations in cell-free \n \n S. aureus \n \n culture supernatants using cell-based transcriptional reporter assays for AIP-dependent AgrC activation or inhibition. Several such reporter assays have been described in which the agrP2 or agrP3 promoter is fused to a reporter gene such as blaZ , gfp, lux or gluc to provide colorimetric, fluorescence or bioluminescence outputs [ 26, 31, 40–42 ]. The assays can be conducted using microtitre plates to quantify AIP levels, evaluate AIP structure activity relationships (SAR), pharmacological properties (agonist, inverse agonist, antagonist) or the functionality of mutations in AgrC receptor proteins. For example, Jensen et al . [ 31 ], deleted the chromosomal agr locus and replaced it with the luxCDABE operon under the control of the agrP3 promoter. The agrA and agrC genes were then introduced on a plasmid under the control of the agrP2 promoter. Since this reporter system is unable to produce AIPs and lacks the autoinduction pathway characteristic of the native agr QS system, it facilitates the in-depth pharmacological evaluation of AIP analogues without interference from an endogenously active agr system [ 31 ]. It also facilitates the introduction of the desired native or mutated agrC gene and requires no additional reagents. For example, using this reporter for AIP-1/AgrC1, an EC 50 of 6±1 nM was derived from a dose–response curve and an EC 50 of 9±1 nM for AIP-4/AgrC4. In contrast, AIP-1 is a very weak activator of AgrC4 (EC 50 3542±997 nM) [ 31 ]. Comparable nanomolar EC 50 s have been reported by others for activation of AgrC by the cognate AIP [ 28, 29, 38 ] using alternative transcriptional reporters. AgrD pro-peptide processing by the transmembrane protein, AgrB AIPs are derived from an internal fragment of an AgrD pro-peptide, which consists of 40–50 amino acid residues incorporating an N-terminal amphipathic leader (N-AgrD; 24–25 amino acids), a mid-region of 7–9 amino acid residues that constitutes the AIP and a charged C-terminal tail (AgrD-C; 14–15 amino acids) ( Fig. 3 ). The generation of an extracellular AIP requires at least four membrane-associated steps (i) removal of AgrD-C, (ii) formation of the thiolactone macrocycle, N-AgrD-AIP (iii) cleavage of N-AgrD and (iv) export of AIP and N-AgrD ( Fig. 3 ). Fig. 3. Schematic showing the processing of AgrD pro-peptides to generate the active cyclic AIP signal molecules. ( a ) Amino acid sequences of \n \n S. aureus \n \n AgrD1-D4 showing the AgrB cleavage site and the MroQ sites for AgrD1, D2 and D4. MroQ does not cleave AgrD3. ( b ) Processing of AgrD by AgrB and MroQ to release N-AgrD and the cyclic AIP. ( c ) Schematic showing the formation and release of the AIP and N-AgrD at the cytoplasmic membrane. Cleavage of AgrD by AgrB releases a 14 amino acid C-terminal peptide (AgrD-C), which is degraded in the cytoplasm. N-AgrD-AIP is cleaved by MroQ to release N-AgrD and the mature AIP. The mechanism by which the AIP and N-AgrD are exported is not known. \n S. aureus agrB deletion mutants are unable to activate agr as they fail to produce AIPs. This is because AgrB, a unique transmembrane cysteine protease is required for processing the AgrD pro-peptide [ 27 ] via steps (i) and (ii) ( Fig. 3 ). Although AgrB lacks homology with other proteins, the alignment of AgrB sequences from diverse Gram-positive bacteria has revealed multiple, highly conserved amino acid residues. However, the \n \n S. aureus \n \n AgrBs do exhibit some substrate specificity. AgrB1 is able to process AgrD1 and AgrD2 but not AgrD3 and vice versa. Chimeric AgrB1 and AgrB2 proteins have been constructed and regions putatively involved in AgrD group specificity identified [ 43 ]. Two of the invariant residues in AgrB homologues are Cys84 and His77. These appear to constitute a catalytic dyad that enables AgrB to cleave AgrD-C from AgrD. This is likely to result in the formation of an acyl-enzyme thioester intermediate, followed by peptidyl transfer to generate the thiolactone macrocycle via the internal Cys28 of AgrD so releasing N-AgrD-AIP ( Fig. 3 ). Furthermore, an N-terminal amphipathic helix, the length of the C-terminal tail and certain C-terminal residues (e.g. E34 and L41; Fig. 3 ) of AgrD all appear to be essential for substrate recognition and processing by AgrB in bacterial cell-based assays [ 44–46 ]. When purified, the \n \n S. aureus \n \n AgrB protein forms stable dimers and is enzymatically active only when embedded in lipid bilayers [ 47 ]. It clearly drives a proteolytic cyclization reaction removing the C-terminal 14–15 amino acid residues of AgrD while reversibly catalysing the formation of a thiolactone ring, which undergoes slow irreversible hydrolysis to the corresponding linear form. The N-AgrD-AIP thiolactone is protected from ring opening by association with membrane phospholipids and the reaction is driven efficiently by the rapid degradation of the AgrD-C fragment [ 47 ], possibly via ClpP or ClpX in staphylococcal cells [ 48 ] Topology predictions for AgrB initially proposed two different models; a six transmembrane domain model with both termini at cis [ 44, 49 ] and a four transmembrane domain with an additional half-transmembrane hairpin where both termini are at trans [ 50 ]. These contradictory models, which contain structurally implausible domains, have yet to be replaced by high-resolution crystallographic or NMR-derived structures. However, a recent molecular dynamics model supported by experimental circular dichroism data revealed a tightly packed six transmembrane domain helical topology for AgrB with both N- and C-termini on the same side and placing amino acid residues known to be important to function on the cis side near the polar/apolar interface [ 51 ]. These include the enzyme active site Cys84, which is located in the membrane interior providing access to the AgrD substrate. Additional molecular dynamics simulations of AgrB and AgrD in a membrane environment have provided a new model incorporating an AgrB dimer with crucially two non-equivalent AgrB sites in which one AgrB monomer facilitates insertion and positioning of AgrD in the correct orientation for catalytic processing by the second AgrB ( Fig. 4 ) [ 51 ]. Whether the proposed (AgrB) 2 AgrD complex exists requires further experimental confirmation. Fig. 4. Conformation of membrane-embedded ternary complex AgrB 2 /AgrD after molecular dynamics simulations. The two AgrB proteins are inequivalent with AgrB-I (cornflower blue) guiding substrate AgrD (fuchsia) to the active site involving catalytic AgrB-II (blue). AgrD residues C28 and M32 are close to each other and to catalytic AgrB-II C84. Key interactions stabilizing the complex include AgrB-I K139-D33 and AgrB-I R70-E34 contacts with AgrD, and AgrB-II H77-F30 via π-interactions. Sidechains are colour-coded aqua (C28, C84 AgrB-II), grey (M32 AgrD), pink (F30 AgrD), purple (H77 AgrB-II), fuchsia (E34 AgrD), green (D33 AgrD), yellow (K139 AgrB-I) and orange (R70 AgrB-I). All four AgrB termini are on the cytosolic side (top) consistent with six transmembrane domain topology of AgrB. The (AgrB) 2 /AgrD complex orientation here is shown with cytosol at the top offering a better view of the cytosol-accessible active site (adapted from [ 51 ]). N-terminal cleavage of AgrD requires a second transmembrane protease, MroQ The identity of the transmembrane protease required for cleaving the N-terminal region from the N -AgrD-AIP thiolactone to release the AIP [step (iii) ( Fig. 3 )] has been something of an enigma. Expression of agrB and agrD from agr group I in \n \n Escherichia coli \n \n surprisingly resulted in the formation of extracellular AIP [ 45 ] indicating that \n \n E. coli \n \n must also possess an AgrD N-terminal cleaving protease. Kavanaugh et al . [ 52 ] developed a fluorescence assay based on a linear synthetic peptide designed to identify \n \n S. aureus \n \n peptidases that could potentially cleave the AgrD N-terminal amphipathic peptide. Evidence was provided in support of signal peptidase B (SpsB) as the missing transmembrane protease, which included the ability of SpsB peptide inhibitors to reduce agr reporter expression and AIP production in \n \n S. aureus \n \n . These data were also consistent with AIP production in \n \n E. coli \n \n , which has only one, essential signal peptidase that, in a temperature-sensitive mutant, can be genetically complemented by SpsB. However, more recent work indicates that in \n \n S. aureus \n \n a different transmembrane protein is the primary enzyme involved in removing AgrD N-terminal peptides [ 53, 54 ]. Upstream of the agr locus, a gene termed mroQ , the mutation of which results in the loss of virulence and significantly reduced extracellular AIP levels was identified [ 53, 54 ]. The translated product is predicted to have eight transmembrane-spanning helices ( Fig. 5 ) and is a member of the CPBP (CAAX p roteases and b acteriocin-processing enzymes) family of putative membrane metalloproteases encompassing over 5000 members that incorporate the misnamed ‘Abortive phage infection’ (Abi) domain [ 55 ]. CPBP superfamily members share three signature motifs and are found in both prokaryotes and eukaryotes. In the latter, prenylation is crucial for the function and membrane protein targeting [ 55 ]. In bacteria, only a few CPBP proteins have been studied experimentally but their functions in general have remained relatively elusive. SkkI for example from \n \n Lactobacillus plantarum \n \n functions as a bacteriocin receptor and immunity protein [ 56 ], whereas the group B streptococcal CPBP protein AbxI interacts with the histidine sensor kinase, CovS to control virulence gene expression [ 57 ]. For SkkI but not for AbxI proteolytic activity is necessary for function. Several CPBP proteins are present in \n \n S. aureus \n \n including MroQ, which contains the EEXXXR and FXXXH motifs considered necessary for enzyme activity. The critical residues for metalloprotease activity include the predicted MroQ catalytic site Glu141 and Glu142 and the His residues 180 and His213 required for zinc co-ordination [ 53 ] ( Fig. 5 ). Replacement of either Glu or the His with Ala in these motifs variably reduced AIP production but \n \n S. aureus \n \n virulence was only attenuated in a mouse skin and soft-tissue infection model to the same level as an agr deletion mutant for Glu141Ala replacement [ 53 ]. In a cell-based assay, expression of agrD in an S. aureus mroQ mutant resulted in the accumulation of a membrane-associated AgrD intermediate but no AIPs were produced [ 58 ]. The amino acid sequence of MroQ does not vary with agr group suggesting that the same enzyme processes all four agr groups. However, in agr group III strains, deletion of mroQ did not prevent the production of an active AIP [ 58 ]. Whether AgrB and MroQ interact directly in \n \n S. aureus \n \n membranes to process AgrD in a manner analogous to that of the MroQ homologue SpdC (LyrA) with SagB is not known. SpdC and SagB (a membrane-bound N -acetylglucosaminidase) form a complex that functions as a peptidoglycan release factor. However, the conserved residues required for CAAX proteolytic activity are not required by SpdC [ 59, 60 ]. Fig. 5. Topological model of transmembrane endopeptidase MroQ. The homology model of MroQ (unpublished) was obtained from I-TASSER [ 143 ] and annealed in atomistic MD simulations using NAMD [ 144 ] within a membrane patch built in CHARMM-GUI [ 145 ]. The model reveals an eight TMD topology – lateral view (left); and axial view (right). Residues Glu141 and Glu142 and His180 and His 213 required for MroQ functionality are shown within the membrane region where they have access to the hydrophobic AgrD substrate. Biochemical confirmation that MroQ is able to remove the N-terminal peptide from N-AgrD-AIP was obtained by reconstituting the MBP-MroQ fusion proteins into proteoliposomes and incubating with N-AgrD-AIP thiolactones from agr groups I, II and III. The efficient generation of AIP-1 and AIP-2 was confirmed by LC MS/MS, although no cleavage of the agr group III substrate was observed. By including AgrB in the proteoliposomes with MroQ, complete reconstitution of AIP biosynthesis from AgrD was achieved [ 61 ]. These elegant biochemical studies together with the observation that SpsB is unable to cleave AIP biosynthetic intermediates in vitro confirmed that MroQ is the primary transmembrane protease involved in N-terminal cleavage of the AgrB processed AgrD thiolactone product. However, although the agr N-AgrD-AIP thiolactones from agr groups I and II were substrates for MroQ, it did not cleave the agr group III AgrD peptide thiolactone. Thus, there must be an alternative enzyme, which does not appear to be SpsB. Both the N-AgrD peptide and AIP are released from the cells into the extracellular environment. N-AgrD has PSM toxin-like properties and is also amyloidogenic, capable of forming amyloid fibrils in \n \n S. aureus \n \n biofilms [ 62 ]. Gonzalez et al . [ 63 ] identified both formylated and non-formylated peptide variants derived from N-AgrD in \n \n S. aureus \n \n culture supernatants, which were cytotoxic for mammalian cell lines, modulated neutrophil chemotaxis and increased the size of murine skin lesions induced by an S. aureus agr mutant. Whether MroQ processes N-AgrD-AIP thiolactones on the cytoplasmic or external face of the cytoplasmic membrane has not yet been established. In either case it is not clear whether export occurs via MroQ, AgrB, an MroQ/AgrB complex or via dedicated transport mechanisms such as the ABC transporters PmtCD and AbcA [ 64 ]. These can independently export PSMs from either membrane or cytosolic environments. The AgrCA two-component sensor regulator system Once exported, AIPs are sensed via a two component system (TCS) in which AgrC is the histidine kinase sensor and AgrA, response regulator ( Fig. 2a ). AgrC belongs to the minor HPK10 histidine kinase protein sub-family [ 65 ] and has a modular architecture with an N-terminal sensory module incorporating an AIP-binding site linked via a helical linker to a C-terminal cytoplasmic module containing two subdomains [ 25 ] ( Fig. 6 ). These are the dimerization and histidine phosphorylation (DHp) and the catalytic and ATP-binding (CA) subdomains [ 25 ]. The CA subdomain containing the G1 box lacks both the conserved DXGXG motif and the key Asp residue, which is replaced by Asn such that the binding affinity of AgrC for ATP and hence kinase activity is likely dependent on cellular energy levels [ 25, 66 ]. The interactions between AgrC and the activating AIP ligand are highly specific occurring at nanomolar affinities where the AIP non-cooperatively binds to AgrC in a 2 : 2 stoichiometry [ 25 ]. Fig. 6. Architecture of the AgrC and AgrA two-component sensor and response regulator proteins. Models of the homodimeric sensor kinase AgrC (left) (orange/cornflower blue) and response regulator AgrA (right) (cyan) associated with the agr P2 promoter site. Initial conformations of AgrC and AgrA were obtained from AlphaFold [ 146 ]. The AgrC dimer was built using ClusPro [ 147 ]. Lipid phosphates are shown to mark the membrane interface (P atoms are shown in cyan). His239, responsible for AgrC autophosphorylation, is shown in purple; Arg238 (yellow) and Gln305 (fuchsia) are important for the molecular ‘latch’ that stabilizes AgrC in the ‘off state’ [ 67 ]. The AgrA model was superimposed onto the structure of the AgrA C-terminal DNA-binding domain [PDB:3BS1] [ 71 ]. Asp59 (green) phosphorylation by AgrC is important for activation of AgrA. His169 (purple), Asn201 (fuchsia) and Arg233 (orange-red) are responsible for DNA recognition. AgrC behaves like a rheostat, where activation of its membrane spanning domain following the binding of an AIP results in the twisting of a helical linker relative to the cytoplasmic domain and subsequent dimerization that results in AgrA phosphorylation [ 25 ]. The binding of a classical competitive antagonist such as AIP-3 to AgrC1 blocks rotation of the helical linker whereas AIP-2, an antagonist that behaves as an inverse agonist of AgrC-1 drives rotation of the helical linker in the opposite direction to that of the agonist AIP-1 [ 25 ]. A key non-covalent interaction between Arg238 and Gln305 that stabilizes AgrC in the ‘off’ state has also been identified [ 67 ] ( Fig. 6 ). This ‘latch’ is proposed to lift following the binding of the cognate AIP to the extracellular AgrC sensor domain, thus mediating the structural changes that result in AgrC activation. Replacement of Arg238 with Ala renders AgrC constitutively active. The mechanism by which AIPs induce the structural changes in AgrC that lead to activation of kinase activity is not yet clear, although AIP-binding is likely to either induce or stabilize specific rotational conformations [ 67 ]. No full-length AgrC structure has yet been obtained and hence the nature of the AIP-binding site is currently not known. However, topology models predict that the N-terminal sensory module of AgrC contains six transmembrane-spanning helices and three extracellular loops ( Fig. 6 ). Since the S. aureus agr system has undergone significant evolutionary divergence, the retention of functionality requires that changes in AgrD, which modify the AIP structures should be accompanied by compensatory changes in the AgrC receptor protein. Since AIPs are not internalized, the AIP-binding site is likely to be on the outer face of the transmembrane AgrC protein and therefore likely to involve the extracellular loops. AIP-1 and AIP-4 differ by a single amino acid; when the corresponding AgrC1 and AgrC4 are compared, only two of their three predicted AgrC extracellular loops exhibit amino acid differences [ 31, 68 ]. In loop 1, 7 out of 19 and in loop 2, 3 out of 9 amino acid residues differ. Replacement by site-specific mutagenesis of these in AgrC4 either singly or in combination with those from AgrC1 revealed that while differential recognition of AIP-1 and AIP-4 depends primarily on three amino acid residues in loop 2, loop 1 was essential for receptor activation by the cognate AIP [ 31, 68 ]. Furthermore, a single mutation in the AgrC1 loop 2 resulted in conversion of (Ala 5 )AIP-1, a non-native AgrC1 inhibitor, to an activator, essentially resulting in the forced evolution of a ‘new’ AIP group [ 31 ]. Taken together, these data suggest that extracellular loop two may constitute the AIP macrocycle-binding site while the exocyclic N-terminal amino acids interact with loop 1 to facilitate receptor activation. However, there is no direct evidence to demonstrate that the key amino acid residues in AgrC loop 2 are directly involved in AIP-binding. It is conceivable that they could act indirectly on the conformation/presentation of direct contact residues elsewhere on AgrC [ 68 ]. The specificity of AgrC1 could also be further broadened by replacement of Ile at position 171 with Lys in the third predicted AgrC extracellular loop. This site-specific AgrC1 mutant was activated at nM EC 50 concentrations not only by the cognate AIP-1 but also by AIP-3, AIP-4 and (Ala 5 )AIP-1 [ 69 ]. Multiple AgrC mutants, which are constitutively active have also been isolated and map to both the last transmembrane helix of the sensor domain and to the histidine kinase domain [ 69 ]. Once AgrC has been trans -autophosphorylated, phospho-transfer to AgrA, a member of the LytTR class of transcriptional regulators, occurs via the AgrC DHp cytoplasmic phospho-transfer sub-domain. AgrA consists of an N-terminal receiver domain containing the conserved Asp residue that drives dimerization upon phosphorylation and a C-terminal DNA-binding domain [ 70 ] ( Fig. 6 ). Dimerization of AgrA enhances DNA binding to the LytTR domain-binding sites of which there are two, 9 base pair, high affinity sites in the agrP2 promoter region separated by 12 bp and both a high- and a low-affinity LytTR binding site in the agrP3 promoter region [ 71 ]. This enables differential expression of RNAII and RNAIII to facilitate QS via autoinduction and to avoid premature expression or degradation of RNAIII [ 46, 72 ]. Related agr promoter region motifs have also been identified in genes directly regulated by AgrA such as those coding for PSMs [ 73 ]. AIP-mediated activation and inhibition of AgrC Extensive AIP structure activity (SAR) studies using transcriptional reporters in \n \n S. aureus \n \n in combination with native AIPs and diverse AIP analogues have established the key molecular features for AgrC activation and inhibition [ 22, 28–30, 74–77 ] ( Fig. 7 ). In general, minor differences in the AIP peptide sequence result in the complete loss of agonist activity. However competitive inhibition is highly tolerant of AIP sequence variation [ 22 ]. The macrocyclic ring is essential for AgrC activation as the corresponding linear peptides are inactive [ 26, 27, 78 ]. Similarly changes to the size of the macrocycle are not tolerated [ 79 ]. Removal of the three exocyclic N-terminal amino acids in AIP-1 also results in the loss of agonist activity as does replacement of the thiolactone S with N or with O to form the corresponding lactam ( Fig. 7 ) and lactone, respectively [ 28, 29 ]. For AIP-1, replacement of each l -amino acid residue in turn with the corresponding d -isomer resulted in six out of the eight analogues exhibiting markedly lower activity. However, exchanging either the macrocycle Phe or the terminal Met residue with the corresponding d -isomer had little impact suggesting that the AIP-binding site in AgrC1 is able to accommodate these differences [ 28 ]. S-oxidation of the methionine thioether side-chain to form the methionyl sulphoxide derivative inactivated AIP-1 ( Fig. 7 ), as did replacement of the Met with norLeu, Ser, Glu, Lys or Pro but not Ile, further emphasizing the critical role of the C-terminal thioether side-chain for AgrC1 activation [ 28, 29 ]. AIP-4, which also has a Met in the same position as AIP-1, is the only other \n \n S. aureus \n \n AIP capable of being inactivated via formation of the methionyl sulphoxide [ 28 ]. Substituting each AIP-1 amino acid residue in turn with Ala ( Fig. 7 ), except for the central Cys, revealed that the only replacement showing increased activity (by ~2 fold) was for the exocyclic Ser [ 28 ]. The other AIP-1 Ala analogues were either inactive or exhibited reduced agonist activity, except that replacement of the endocyclic amino acid residue (Asp) located C-terminally to the central Cys with Ala converted AIP-1 from an activator to a potent low nanomolar IC 50 cross-group inhibitor [ 28, 29 ] ( Fig. 7 ) In AIP-4, the endocyclic Asp residue is replaced by Tyr ( Fig. 2 ) and is therefore the critical determinant of AIP specificity for agr groups I and IV [ 28, 29 ]. Detailed SAR studies of activation and inhibition of the cognate AgrCs have also been undertaken for AIP-2 and AIP-3 with broadly similar findings to those reported for AIP-1 and AgrC1 [ 22, 29, 30, 38, 74–77 ]. These SAR data suggested that the macrocycle was required for receptor recognition and binding while the exocyclic region was necessary for receptor activation [ 22, 80 ]. Further refinement of these data via solution NMR structural analysis of the native AIP-3 peptide and a series of analogues revealed the importance of a hydrophobic ‘bulge’ formed by hydrophobic endocyclic residues and exocyclic tail contacts [ 22, 75 ]. These findings have highlighted the contribution of 3D conformation and the orientation of the AIP exocyclic tail relative to the macrocycle with respect to AgrC activation and inhibition. Fig. 7. SAR for \n \n S. aureus \n \n AIP-1 showing how minor modifications to the peptide sequence including Ala-scanning influence activity. Substitution of the Asp residue ( d 5 \n ) with Ala to give (Ala 5 )AIP-1 converted AIP-1 from an activator of AgrC1 to a potent cross group inhibitor. IC 50 data from [ 29 ]. \n S. aureus agr heterogeneity and environmental inactivation The agr system plays a key role in reciprocally regulating planktonic exotoxin producing, colonization and biofilm-associated lifestyles of \n \n S. aureus \n \n . It functions as a positive feedback loop displaying bimodal, heterogeneous behaviour that leads to the emergence of distinct subpopulations of \n \n S. aureus \n \n cells [ 81 ]. These subpopulations are characterized by the presence or absence of agr activity and by their relative numerical sizes consistent with bet-hedging strategies where some cells behave as individuals while others act co-operatively [ 82 ]. Such phenotypic heterogeneity has been also observed in the QS populations of other bacterial species where it may be transient and restricted to the early stages of activation with the population subsequently becoming homogeneous or heterogeneity may persist resulting in a bimodal, heterogeneous population [ 82 ]. In \n \n S. aureus \n \n the ratio of the agr ‘on’ to agr ‘off’ sub-population may be modified in response to environmental signals such as higher Mg 2+ concentrations that increase cell wall rigidity by binding to teichoic acids and triggering the σ B -mediated down-regulation of agr [ 81 ]. Other cell-wall changes have been observed to modify agr activation. For example, in some, but not all, HA-MRSA strains [ 83, 84 ] expression of mecA, which codes for penicillin-binding protein 2A, reduced agr expression and virulence in a mouse infection model. Expression of agr could be restored by partially digesting the cell wall, suggesting that MecA-induced changes in cell wall architecture potentially reduce accessibility of the AIP to AgrC. Regulatory interdependence of mecA and agrA has also been noted for certain CA-MRSA strains in which methicillin resistance is agr -regulated [ 85 ]. HA-MRSA and CA-MRSA strains appear to have similar kinetics of agr activation but the latter achieve much higher magnitudes [ 85 ]. For activation of agr , AIP levels must reach a critical threshold concentration. This is not fixed but will vary according to the relative rates of AIP production, accumulation, diffusion and inactivation, which in turn will also depend on the prevailing growth environment. Little difference in the EC 50 s for AgrCs activated by their cognate AIPs (all low nM) are apparent [ 28, 29, 38 ]. Exogenous provision of the cognate AIP at the time of inoculation overcomes the ‘quorum’, prematurely activating agr although there is a window within the first few hours of growth after which \n \n S. aureus \n \n does not respond [ 86, 87 ]. Genotype versus agr locus-dependent differences in agr dynamics have been investigated in the context of agr group divergence by constructing congenic \n \n S. aureus \n \n strains (8325–4 and Newman) each with a different agr group allele and carrying an agrP3-blaZ fusion [ 86 ]. These revealed differences in the timing and magnitude of agr activation with S. aureus agr group III cells showing the most delayed induction and lowest level of agr expression, which in turn was reflected in cell-wall protein and exotoxin production. Whether such differences impact on virulence in infection models has not yet been established. The ability to switch agr on or off during different stages of infection of host cells and in different tissues is clearly advantageous when moving from an acute toxigenic to a chronic persistent lifestyle, where increasing production of surface adhesins, reducing exotoxin secretion and either taking up intracellular residence or forming a biofilm enables \n \n S. aureus \n \n to avoid host immune defences. In vivo evidence supporting this on/off agr switch was obtained by Wright et al . [ 88 ] using whole body luminescence imaging of \n \n S. aureus \n \n transformed with an agr P3- lux expression vector. This revealed early rapid activation of agr followed by several days without any agrP3 expression prior to renewed light output. While this switching of agr may have been due to transient agrP3 expression, it could alternatively have been a consequence of either exhaustion of the LuxAB and LuxCDE enzyme substrates FMNH 2 and a long-chain fatty aldehyde or reduced oxygen availability in deeper tissues resulting in a lack of luciferase activity [ 89 ]. Although AIPs are not known to be degraded or inactivated by endogenous staphylococcal enzymes, the C-terminal Met of \n \n S. aureus \n \n AIP-1 is S-oxidized during aerobic growth in laboratory media to form the corresponding methionyl sulphoxide [ 28 ] ( Fig. 7 ). This methionyl sulphoxide-containing compound is unable to activate or inhibit AgrC [ 28 ]. Whether its formation and accumulation impact on the timing of agr induction by reducing AIP-1 levels (or AIP-4 which is the only other \n \n S. aureus \n \n AIP with a terminal Met) below the activation threshold is not known. However, this S -oxidation reaction has in vivo relevance as it occurs in response to phagocyte derived reactive oxygen and nitrogen species such as hypochlorous acid (HOCl) and peroxynitrite (ONOO - ) and results in down-regulation of agr and a concomitant reduction of virulence in a mouse skin infection model [ 15 ]. Oxidative stress can also down-regulate agr via AgrA modification as a consequence of disulphide bond formation between the redox-reactive Cys119 and Cys288 leading to dissociation of the modified AgrA from DNA [ 90 ]. In addition, Cys119 can undergo ‘CoAlation’, i.e. the covalent modification by co-enzyme A, which, under nutrient deprivation or oxidative stress conditions also reduces the affinity of AgrA for the agrP2 and agrP3 promoters [ 91 ]. During in vitro competition and evolution experiments, oxidative stress was observed to drive the emergence of agr mutants, which possess a fitness advantage only under aerobic growth conditions due to the reactive oxygen species generating capacity of PSMs and RNAIII-regulated factors [ 92 ]. Consequently, agr imposes an oxygen-dependent fitness cost such that hypoxia favours maintenance of QS and increased exotoxin production. This oxygen-driven tuning of the agr system may therefore exert a major influence on tissue-dependent disease progression during infection. Exposure to air pollution and in particular particulates such as black carbon (BC) are associated with exacerbations of chronic respiratory disease [ 9 ]. Growth of \n \n S. aureus \n \n in BC prior to inoculation increased the adhesion to, and invasion of, human epithelial cells in vitro and murine respiratory tract colonization and pulmonary invasion in vivo [ 9 ]. Global transcriptional analysis revealed that numerous agr -regulated exoprotease, exotoxin and immune evasion genes were upregulated while certain adhesin and metabolic genes were repressed suggesting that that BC acts directly on the pathogen rather than exclusively on the host [ 9 ]. The mechanism by which BC controls this subset of the agr regulon has yet to be elucidated. Host factors known to impact on the agr -driven switch from a colonizing to an invasive phenotype include serum proteins such the low-density (LDL) and very low-density (VLDL) particles associated apo-lipoprotein B (ApoB), which interfere with agr -dependent QS by specifically and reversibly sequestering \n \n S. aureus \n \n AIPs [ 93 ]. This ApoB-mediated downregulation of agr in serum can be circumvented using constitutively active AgrC variants [ 94 ]. All four \n \n S. aureus \n \n AIPs as well as the inactive methionyl oxide derivative of AIP-1 bind to ApoB via an interaction that is likely to be dependent on the hydrophobic thiolactone macrocycle as the (less hydrophobic) linear peptide corresponding to AIP-1 does not bind. The in vivo relevance of these findings is supported by the greater susceptibility of mice deficient in ApoB to invasive infection with a wild-type \n \n S. aureus \n \n strain compared with an isogenic agr deletion mutant [ 93 ]. Although once considered to be an exclusively extracellular pathogen, \n \n S. aureus \n \n can internalize and survive in a variety of mammalian cells including endothelial, epithelial and professional phagocytic cells where they contribute to chronic and relapsing infections [ 95 ]. Quenching of agr -dependent QS in the bloodstream, for example, ensures that the \n \n S. aureus \n \n cell-wall proteins required for host cell attachment and internalization remain highly expressed [ 93 ]. Once inside an endosome or phagosome, \n \n S. aureus \n \n cells escape into the cytoplasm, kill the host cell, be killed or remain intracellular, protected from host defences and acting as a reservoir for persistent infections ( Fig. 8 ). Within these intracellular vesicles, AIPs accumulate rapidly activating agr and leading to the expression of exotoxins such as α-haemolysin and the PSMs, which lyse the endosomal or phagosomal membrane [ 14, 16 ]. With respect to neutrophil phagosomes, AIP-mediated staphylococcal escape can be blocked by inactivation of AIP-1 and AIP-4 by S-oxidation via NAPDPH-derived oxygen radicals [ 15 ]. This intracellular escape process has been termed confinement or compartment sensing rather than QS since agr activation occurs in a single trapped bacterial cell rather than in a population [ 10, 96 ]. Fig. 8. ( a ) Intracellular uptake of \n \n S. aureus \n \n into non-professional epithelial and endothelial cells is independent of agr expression. Once internalized, agr expression precedes endosomal escape by facilitating endosomal lysis via α-haemolysin or the PSMs. This enables \n \n S. aureus \n \n to replicate and persist within the cytoplasm protected from host immune defences and antibiotics or to lyse the host cells and establish further rounds of uptake and release leading to tissue destruction. ( b ) Fluorescence microscopy of \n \n S. aureus \n \n transformed with an agrP3-gfp reporter invading mammary epithelial cells and stained with DAPI (DNA; blue) and an anti-tubulin-Cy3 conjugate (red; microtubules). After 2 h incubation, intracellular staphylococcal cells are observed (white arrow) with red cell walls as the Cy3 antibody conjugate has bound to protein A showing that agr has not yet been activated. As the infection proceeds to 6 h agr expression has clearly been induced as observed by the high level of gfp expression (green) in the bacterial cells (adapted from [ 14 ]). \n agr dysfunction and infection Despite the relevance of agr to \n \n S. aureus \n \n virulence in animal infection models, agr defective mutants are commonly found in clinical samples from both asymptomatic nasal carriage and serious infections where the loss of agr functionality in both MRSA and methicillin-sensitive (MRSA) strains is associated with greater adaptability, persistent infections and poorer outcomes [ 87, 97–103 ]. In healthy individuals in the community, nasal carriage of agr mutants has been reported to be relatively low at ~9 % colonization. However, agr dysfunction has been strongly associated with hospitalization and antibiotic usage suggesting a trade-off between virulence and antibiotic resistance [ 100, 102 ]. S. aureus agr mutants exhibit enhanced survival in the presence of daptomycin because they shed phospholipids that neutralize the antibiotic extracellularly. In contrast, agr -functional strains released less phospholipid and secreted agr -regulated PSMs, which inhibited the phospholipid-daptomycin interaction [ 104 ]. Analysis of de novo mutations in >1000 \n \n S \n \n . \n \n aureus \n \n genomes from 105 infected patients with prior nasal colonization revealed that adaptive mutations in pathogenesis-associated genes including agr were enriched in infecting but not nasal-colonizing bacteria indicative of within host selection pressures [ 101 ]. Others have shown that events associated with agr inactivation result in agr -defective blood and nasal strain pairs that are enriched in mutations compared to pairs from wild-type controls [ 105 ]. These additional mutations outside the agr locus can contribute to diversification and adaptation during infection by agr mutants associated with poor patient outcomes [ 105 ]. In their analysis of >40 000 \n \n S \n \n . \n \n aureus \n \n genomes, Raghuram and co-workers [ 32 ] did not find any stable agr -defective strain lineages. This is consistent with previous suggestions that agr defective strains may be unable to establish and maintain circulating populations outside their original location [ 32, 99 ]. While the host factors responsible for driving the selection of agr dysfunctional mutants have not yet been identified, host resistance should not be overlooked as indicated by Thänert et al . [ 106 ] who compared susceptible (A/J) and resistant (C57BL/6) mouse strains with respect to \n \n S. aureus \n \n infection and virulence gene expression. Strains with mutations in agr are capable of prolonged intracellular survival. Although they cannot escape phagosomes, they induce less cell death and so survive in greater numbers within host cells [ 107 ]. agr mutants can also persist via an alternative intracellular pathway in endothelial cells involving LC3 + vesicles [ 107 ]. Furthermore, sub-populations of slow-growing \n \n S. aureus \n \n small colony variants (SCVs) able to survive within host cells are frequently recovered from chronic and recurrent infections [ 108, 109 ]. These SCVs may be genetically stable with characteristic mutations in metabolic pathways or unstable and exhibiting increased expression of negative agr regulators (e.g. SigB, ArlRS and CodY). In both cases, a common characteristic of SCVs that arise from either altered electron transport or global regulatory pathway changes is reduced Agr activity. These observations suggest that \n \n S. aureus \n \n populations progressing from colonization to infection at different body sites may be heterogenous with respect to agr expression, or consist of agr -functional cells or agr -defective mutants or mixed populations with the balance influencing the outcome. It is also apparent that agr mutant populations may contain a small fraction of phase variable cells capable of reverting to agr -functional cells [ 110 ]. This appears to arise at least in vitro via a mechanism involving a poly(A) tract alteration and a genetic duplication plus inversion event. This strategy has been suggested to act as cryptic insurance against host-mediated stress enabling the population to survive phagocytosis and sustain infection [ 110 ]. An explanation for why S. aureus agr mutants exhibit reduced virulence in animal models but are frequently isolated from clinical samples was suggested by Pollitt et al . [ 111 ] using a waxworm larvae infection model. They showed that agr -dependent QS is a beneficial social trait in which agr mutant ‘cheats’, which neither produce nor respond to AIPs exploit agr functional AIP-producing co-operators. However, while these data provide an explanation for mixed populations in toxigenic infections, they do not account for chronic, biofilm-associated clinical \n \n S. aureus \n \n infections where homogeneous agr -negative populations can rapidly emerge [ 112 ]. It should also be noted that \n \n S. aureus \n \n isolates exhibiting low levels of exotoxins are not necessarily agr mutants highlighting the likely existence of other, novel exotoxin regulators [ 113 ]. \n agr dysfunction – the molecular basis DNA sequence analysis of agr -dysfunctional \n \n S. aureus \n \n clinical isolates has revealed frameshifts, insertions, deletions and substitutions in the agrBDCA operon [ 87, 97, 98, 113, 114 ]. In their analysis of over 40 000 \n \n S \n \n . \n \n aureus \n \n genomes Raghuram et al . [ 32 ] found that >5 % had agr operon frameshifts. These were almost exclusively found in agrA and agrC with the latter having accumulated the greatest number of different mutations. The insertion of an extra adenine at the 3′ end of agrA was the most common frameshift and is known to result in delayed agr activation and haemolysin production [ 115 ]. Interestingly, when comparing the agr histidine kinase ( agrC ) and response regulator ( agrA ) genes with other \n \n S. aureus \n \n TCSs ( arl, kdp , nre , pho , srr and wal ), the frequency of mutations in agr is highly enriched [ 32 ]. The frequent occurrence of common but independently acquired convergent mutations may be an adaptive response to specific host selective pressures. Some of these naturally occurring changes have either been predicted or experimentally demonstrated to result in the complete loss of agr functionality. For example, Mairpady Shambat et al . [ 116 ] isolated an MRSA strain harbouring a single AgrC Y223C cytoplasmic domain substitution that switched the virulence phenotype from cytotoxic to colonizing that could be reversed by mutating back to C223Y. However, not all naturally occurring agr mutations are likely to be inactivating but may instead modify the timing and/or the strength of agr induction. Sloan et al . [ 87 ] identified a number of natural mutations associated with reduced cytotoxicity also linked to the cytoplasmic domain of AgrC. These delayed the onset and accumulation of AIPs and exhibited impaired AgrC-AIP responsiveness as revealed by the increased threshold for AgrC activation. Exotoxin production, in this case Panton–Valentin leucocidin production, could be restored by exogenous provision of the cognate AIP at the time of inoculation indicating that delayed activation of agr autoinduction and consequently failure to express RNAIII results in the lack of exotoxins. Molecular dynamics simulations from in silico engineered point mutations in the AgrC cytoplasmic domain revealed subtle changes that alter both domain conformation and relative domain orientation [ 87 ]. The efficiency of the rheostat-like behaviour of AgrC in which AIP binding induces the twisting of a helical linker relative to the cytoplasmic domain and subsequent dimerization [ 25 ] is likely to be impaired by these mutations. Consequently, a greater magnitude of ‘input’ into the rheostat-like mechanism will be required in order to produce the same response. In turn, this would impact on the efficiency of the agr autoinduction circuitry. Such conformational rearrangements of key functional subdomains in these AgrC cytoplasmic domain mutants highlight the cooperative responses of protein structures involving dimerization, ATP binding and phosphorylation, as well as sites involved in AgrA interactions. Whether increasing the threshold for agr activation offers a fitness advantage remains to be established. Intra- and inter-species agr -mediated competitive interference among the Staphylococci In S. aureus, strains belonging to one agr group produce an AIP that cross-inhibits each of the other three agr groups, giving rise to the concept of ‘competitive interference’ [ 27, 117 ]. In laboratory experiments in broth inoculated with mixed cultures, the agr group did not impact on competitiveness [ 117 ]. This is perhaps not surprising given that such interference is at the level of agr expression rather than growth. In vivo in a Manduca sexta (tobacco hornworm) infection model, despite the genetic diversity of the \n \n S. aureus \n \n strains tested, differences in the fitness of competing strains belonging to different agr groups were observed [ 117 ]. There have been very few documented reports of clinical samples containing mixed agr group isolates. In one case a patient was reported to have an S. aureus agr group 1 blood isolate and a group 2 wound isolate [ 97 ] while in another, a CF sputum sample contained genome sequences from both agr group 1 and 2 strains [ 32 ]. A comparison of consecutive and co-colonizing strains in healthy individuals or those with CF revealed that strain replacement was accompanied by a change in the agr group in 63 and 80 % of the two cohorts, respectively [ 118 ]. Co-colonizing strains from CF belonged to interfering agr groups in six of ten cases whereas for healthy individuals, nasal co-colonization with strains belonging to different agr groups was rarely observed [ 118 ]. However, in CF sputum, where agr is not expressed [ 119 ], no cross-group inhibitory AIPs will be synthesized making agr group competition unlikely. Competitive interference however does occur during interactions between \n \n S. aureus \n \n and the CoNS, which produce a broad range of AIPs [ 37 ]. They constitute a diverse group of at least 38 different staphylococcal species that are primarily found on the healthy skin and mucous membranes of humans and other mammals, alongside other bacterial genera including corynebacteria, streptococci and micrococci [ 120, 121 ]. In a screen of culture supernatants prepared from 52 staphylococcal isolates representing 17 different species obtained from dogs, cows, horses, mink, cats, pigs and birds, Canovas et al . [ 21 ] identified 17 different species that inhibited S. aureus agr . Of 54 CoNS isolates obtained from 25 pig nasal swabs, the eight different species capable of inhibiting S. aureus agr included Staphylococcus hyicus, Staphylococcus simulans, Staphylococcus arlettae, Staphylococcus lentus and \n \n Staphylococcus chromogenes \n \n [ 2 ]. However, \n \n Staphylococcus sciuri \n \n and \n \n Staphylococcus xylosus \n \n were unable to inhibit S. aureus agr [ 2 ]. The presence of specific CoNS species in pig nares has been associated with reduced MRSA colonization and of relevance to intensive pig farming where the transmission of livestock associated MRSA from pigs to humans is a potential health risk [ 2 ]. The most common CoNS skin species are \n \n S. epidermidis \n \n , \n \n S. hominis \n \n , \n \n Staphylococcus haemolyticus \n \n , \n \n Staphylococcus capitis \n \n , \n \n Staphylococcus lugdunensis \n \n and \n \n Staphylococcus warneri \n \n . Most CoNS are relatively harmless, beneficial commensals that actively contribute to shaping the skin microbiota and the cutaneous immune response, promoting tissue-repair and combatting the external threat posed by invading pathogens such as \n \n S. aureus \n \n and group A streptococci [ 1, 122–124 ]. However, certain CoNS species also have dual lifestyles as colonizers and opportunistic pathogens. S. epidermidis, regarded as a keystone skin commensal, is frequently responsible for blood stream infections associated with implanted medical devices including intravascular catheters, prosthetic vascular grafts and heart valves, cardiac devices and coronary stents [ 121 ]. Considering their genetic flexibility and integrity of the skin barrier, it has been proposed that the beneficial or harmful behaviour of \n \n S. epidermidis \n \n may vary depending on the specific strain and interactive context [ 123, 124 ]. Indeed, analysis of 1482 \n \n S \n \n . \n \n epidermidis \n \n genomes from the skin of five healthy individuals established that they had evolved from multiple founder lineages rather than a single colonizer [ 125 ]. \n \n S. saprophyticus \n \n , a ‘moderately’ pathogenic CoNS species found in the genito-urinary tract is the second most frequent cause of uncomplicated lower urinary tract infection in young women [ 121 ]. Among CoNS species, the agrBDCA genes are widespread but highly divergent and offer opportunities for competitive interference within and between species [ 126 ]. Given the diversity of CoNS colonizing the same environmental niche, there is considerable scope for competitive interference at both intra- and inter-species levels, promoting colonization and suppressing virulence factor production by both \n \n S. aureus \n \n and CoNS capable of adopting pathogenic lifestyles [ 1, 37 ]. Such control will ultimately depend not only on competitive agr interference but also on the production of antimicrobials by CoNS such as PSMs that selectively inhibit \n \n S. aureus \n \n and which may themselves be agr -regulated [ 122 ]. Based on their AgrD peptide sequences, most CoNS, in common with \n \n S. aureus \n \n can be divided into different agr subgroups; between two and six depending on the species ( Table 1 ). In \n \n S. hominis \n \n there are six agr groups [ 127, 128 ] while of the four S. epidermidis agr groups, healthy human skin is commonly dominated by a large proportion of S. epidermidis agr group I strains together with smaller sub-populations of agr -II and -III strains [ 125, 129 ]. To date, CoNS species where chemical structures have been confirmed, produce AIPs that, in common with \n \n S. aureus \n \n , are usually between seven and nine amino acid residues and incorporate a thiolactone ring. The exceptions are \n \n S. intermedius \n \n that produces a lactone [ 130 ] and \n \n S. epidermidis \n \n which makes three AIPs with extended seven amino acid residue exocyclic tails [ 125, 129 ]. The \n \n S. intermedius \n \n AIP lactone is a functional autoinducer in which the switch from Cys to Ser may have arisen via spontaneous point mutation [ 130 ]. Neither AgrB1 nor AgrB2 from \n \n S. aureus \n \n could enzymatically process a Ser containing AgrD pro-peptide to generate the mature AIP nor could the \n \n S. aureus \n \n AIP-1 or AIP-2 lactones activate the \n \n S. aureus \n \n AgrC1 or AgrC2 receptors [ 28, 130 ]. Interestingly the readily S -oxidized C-terminal Met present in \n \n S. aureus \n \n AIP-1 and AIP-4 is very rarely found in other staphylococci except for \n \n S. argenteus \n \n and \n \n S. schweitzeri \n \n , which, respectively, make the same AIPs as S. aureus agr groups I and IV [ 37 ]. Table 1. Competitive interference between AIPs produced by CoNS and the agr groups of \n \n S. aureus \n \n and \n \n S. epidermidis \n \n \n \n Coagulase-negative staphylococci \n \n \n \n \n S. aureus \n \n \n \n \n \n \n \n S. epidermidis \n \n \n \n \n \n Species \n \n \n \n Autoinducing peptide \n \n \n \n \n agrI \n \n \n \n \n \n agrII \n \n \n \n \n \n agrIII \n \n \n \n \n \n agrIV \n \n \n \n \n \n agrI \n \n \n \n \n \n agrII \n \n \n \n \n \n agrIII \n \n \n \n \n \n \n \n S. epidermidis \n \n \n \n \n \n \n AIP-1 \n \n \n \n DSV-[CASYF] \n \n \n \n 166 \n \n \n \n >1000 \n \n \n \n 13 \n \n \n \n >1000 \n \n \n – \n \n \n 9.64 \n \n \n \n 34.3 \n \n \n \n AIP-2 \n \n \n \n NASKYNP-[CSNYL] \n \n \n – \n \n – \n \n – \n \n – \n \n \n 13.9 \n \n \n – \n \n – \n \n \n AIP-3 \n \n \n \n NAAKYNP-[CASYL] \n \n \n – \n \n – \n \n – \n \n – \n \n \n 2.13 \n \n \n \n \n \n \n S. caprae \n \n \n \n \n \n \n AIP \n \n \n \n YST-[CSYYF] \n \n \n \n 0.6 \n \n \n \n 0.26 \n \n \n \n 0.2 \n \n \n \n 9.0 \n \n \n – \n \n – \n \n – \n \n \n \n \n \n S. hominis \n \n \n \n \n \n \n AIP-1 \n \n \n \n SYNV-[CGGYF] \n \n \n \n 13 \n \n \n \n 31 \n \n \n \n 5 \n \n \n \n 2910 \n \n \n \n NC \n \n \n \n 34 \n \n \n \n 16 \n \n \n \n AIP-2 \n \n \n \n SYSP-[CATYF] \n \n \n \n 15 \n \n \n \n 2109 \n \n \n \n 3 \n \n \n \n 1130 \n \n \n \n 20 \n \n \n \n 19 \n \n \n \n 62 \n \n \n \n AIP-3 \n \n \n \n TYST-[CYGYF] \n \n \n \n 11 \n \n \n \n 4 \n \n \n \n 6 \n \n \n \n NC \n \n \n \n 4 \n \n \n \n 3 \n \n \n \n 3 \n \n \n \n AIP-4 \n \n \n \n TINT-[CGGYF] \n \n \n \n 128 \n \n \n \n 140 \n \n \n \n 37 \n \n \n \n NC \n \n \n \n 237 \n \n \n \n 93 \n \n \n \n 28 \n \n \n \n AIP-5 \n \n \n \n SQTV-CSGYF] \n \n \n \n 43 \n \n \n \n 59 \n \n \n \n 4 \n \n \n \n 3809 \n \n \n \n 10 \n \n \n \n 22 \n \n \n \n 2 \n \n \n \n \n \n \n S. warneri \n \n \n \n \n \n \n AIP-1 \n \n \n \n YSP-[CTNFF] \n \n \n \n 10 \n \n \n \n 4 \n \n \n \n 13 \n \n \n \n 146 \n \n \n \n 3 \n \n \n \n 19 \n \n \n – \n \n \n AIP-2 \n \n \n \n ANP-[CAMFY] \n \n \n \n 2 \n \n \n \n 30 \n \n \n \n 2 \n \n \n \n 2 \n \n \n – \n \n – \n \n – \n \n \n \n \n \n S. simulans \n \n \n \n \n \n \n AIP-1 \n \n \n \n KYNP-[CLGFL] \n \n \n \n 2.2 \n \n \n \n 1.1 \n \n \n \n 3.5 \n \n \n \n 23 \n \n \n – \n \n – \n \n – \n \n \n AIP-2 \n \n \n \n KYYP-[CWGYF] \n \n \n \n 1.6 \n \n \n \n 1.5 \n \n \n \n 11.5 \n \n \n \n 40 \n \n \n – \n \n – \n \n – \n \n \n AIP-3 \n \n \n \n KYNP-[CWGYF] \n \n \n \n 1.7 \n \n \n \n 6 \n \n \n \n 3.2 \n \n \n \n 48 \n \n \n – \n \n – \n \n – \n \n \n \n \n \n S. schleiferi \n \n \n \n \n \n \n AIP \n \n \n \n KYPF-[CIGYF] \n \n \n \n 2.8 \n \n \n \n 86 \n \n \n \n 80 \n \n \n – \n \n – \n \n – \n \n – \n \n \n \n \n \n S. lugdunensis \n \n \n \n \n \n \n AIP-1 \n \n \n \n DI-[CNGYF] \n \n \n \n 384 \n \n \n \n 419 \n \n \n \n 36.6 \n \n \n \n >1000 \n \n \n – \n \n – \n \n – \n \n \n \n \n \n S. hyicus \n \n \n \n \n \n \n AIP \n \n \n \n KINP-[CTVFF] \n \n \n \n 3.3 \n \n \n \n 350 \n \n \n \n 4 \n \n \n \n 180 \n \n \n – \n \n – \n \n – \n \n \n \n \n \n S. chromogenes \n \n \n \n \n \n \n AIP \n \n \n \n SINP-[CTGFF] \n \n \n \n 15 \n \n \n \n 200 \n \n \n \n 60 \n \n \n \n 350 \n \n \n – \n \n – \n \n – \n \n \n \n \n \n S. haemolyticus \n \n \n \n \n \n \n AIP \n \n \n \n SFTP-[CTTYF] \n \n \n \n 340 \n \n \n \n IA \n \n \n \n 340 \n \n \n \n IA \n \n \n – \n \n – \n \n – \n \n \n \n \n \n S. vitulinus \n \n \n \n \n \n \n AIP \n \n \n \n VIRG-[CTAFL] \n \n \n \n 190 \n \n \n \n 800 \n \n \n \n 690 \n \n \n \n IA \n \n \n – \n \n – \n \n – \n The square brackets denote amino acid residues within the AIP macrocycle. IC 50 values compiled from [ 37, 128, 133–136 ]; NC, not calculated, likely to be a weak inhibitor; IA, not active at highest dose tested; – not available. In general, there is relatively little information on the target genes regulated via agr in different CoNS species. Given the difficulty in genetically manipulating CoNS, Severn et al . [ 128 ] used the AgrA inhibitor apicidin to transcriptionally profile genes regulated by an agr group-I \n \n S. hominis \n \n commensal and identified ~40 down- and ~7 up-regulated genes. Down-regulated genes coding for PSMs, a predicted lipase, acetoin production and multiple transcriptional regulators as well as agrB , agrD and RNAIII were identified [ 128 ]. The up-regulated \n \n S. hominis \n \n genes were of unknown function but likely to be involved in metabolism and sensing. In \n \n S. epidermidis \n \n , one of the most abundant CoNS skin colonizers in which a functional agr system enhances skin colonization [ 129 ], genes down-regulated in an RNAIII mutant included protease and lipase exoenzymes, PSMs, δ-haemolysin, and agr . The haemolytic CoNS species \n \n S. lugdunensis \n \n can cause severe human infections but its repertoire of agr -dependent virulence determinants have not been well characterized. In a recent study, Chin et al . [ 131 ] showed that the \n \n S. lugdunensis \n \n synergistic haemolysins (SLUSH), the metalloprotease lugdulysin, a urease and a number of ABC transporters were downregulated in an agr mutant as were unidentified factors required to protect the organism from host innate immune defences. Studies of AIP-mediated activation of CoNS agr systems or of competitive interference between CoNS species, or between CoNS and \n \n S. aureus \n \n are usually undertaken initially using spent culture supernatants and screened using agr P3 reporter gene fusions based on activation or inhibition of \n \n S. aureus \n \n or CoNS agr groups [ 2, 21, 128, 130 ]. While these highlight the diversity of CoNS capable of inhibiting one or more S. aureus agr groups, such data requires validation via synthesis of the predicted AIP. This then facilitates quantitative determination (EC 50 or IC 50 ) of the agonist or antagonist activities of a specific AIP. In experiments employing S. epidermidis agr P3-gfp reporter fusions and spent culture supernatants [ 129 ], interference between S. epidermidis agr group I and groups II and III but not between agr groups II and III were observed. These data were subsequently confirmed using the corresponding synthetic AIPs [ 129, 132, 133 ] ( Table 1 ). In contrast to AIP-1, both AIP-2 and AIP-3 have extended exocyclic tails that likely account for the differential activity. By quantifying δ-haemolysin as a read-out for agr inhibition in different \n \n S. aureus \n \n strains, \n \n S. epidermidis \n \n AIP-1 was reported to inhibit S. aureus agr groups I to III but not group IV. Conversely, \n \n S. aureus \n \n AIP-4 was the only \n \n S. aureus \n \n AIP found to inhibit S. epidermidis agr group I [ 132 ]. Using an agr P3 reporter, Yang et al ., [ 133 ] observed that \n \n S. aureus \n \n AIP-2 but not AIP-1, AIP-3 or AIP-4 inhibited S. epidermidis agr group I [ 133 ]. \n \n S. hominis \n \n AIP-1 to −6 all inhibited S. aureus agr groups I, II and III with IC 50 s ranging from 3 to 140 nM except for AIP-2 which had little activity against S. aureus agr group II [ 126 ] ( Table 1 ). The \n \n S. hominis \n \n AIPs were all active against the three S. epidermidis agr groups except for AIP-1, which lacked antagonistic activity towards S. epidermidis agr group I [ 128 ]. The strongest \n \n S. hominis \n \n intraspecies interactions were observed between agr groups I and II, which were the two most common S. hominis agr types identified on the skin of a group of 14 human volunteers and in database genome sequences [ 125, 128 ]. In common with \n \n S. epidermidis \n \n , none of the \n \n S. hominis \n \n AIPs were effective antagonists of S. aureus agr group IV [ 128 ] although the IC 50 s for the \n \n S. simulans \n \n AIPs were in the 20–40 nM range [ 134 ] ( Table 1 ) S. simulans agr types also displayed varying degrees of resistance to cross-inhibition from \n \n S. aureus \n \n AIPs, with S. simulans agr III being weakly susceptible to antagonism by \n \n S. aureus \n \n AIP-1 and AIP-4 [ 134 ]. Intraspecies interactions between S. simulans agr variants may also occur since agr group I is inhibited by AIPs-2 and 3 whereas agr groups II and III are not subject to cross-inhibition. Two potent \n \n S. aureus \n \n cross-group inhibitors are \n \n S. warneri \n \n AIP-2 and the \n \n S. caprae \n \n AIP [ 135 ] both of which are potent antagonists of all four S. aureus agr groups [ 136 ] ( Table 1 ). Although inter-species agr activation is rarely observed, the AIPs of \n \n Staphylococcus schleiferi \n \n and \n \n Staphylococcus hominis \n \n (AIP-3) both induce S. aureus agr group IV [ 37, 128 ] ( Table 1 ). Competitive interference and skin disease \n \n \n S. aureus \n \n and especially CA-MRSA strains are the most common causes of skin and soft tissue infections in humans [ 1, 137, 138 ]. The severity of atopic dermatitis (AD), a chronic disease of unclear aetiology is characterized by dry, itchy and inflamed skin and by dysbiosis of the skin microbiota. \n \n S. aureus \n \n colonizes AD patients’ skin lesions and exacerbates disease by promoting inflammation and degradation of the skin barrier function. These have all been linked with \n \n S. aureus \n \n exotoxins, superantigens, exoproteases and PSMs [ 137 ] and also with an overabundance of \n \n S. epidermidis \n \n strains producing the cysteine protease EcpA [ 123, 139 ]. Metagenomic analysis of the AD skin microbiome revealed that an increase in the relative abundance of \n \n S. aureus \n \n in patients with active AD correlated with a lower CoNS AIPs to \n \n S. aureus \n \n ratio, thus reducing the ability of the CoNS to inhibit the S. aureus agr system [ 127 ]. In a further study involving 268 Japanese infants aged 1 to 6 months, 121 were colonized with \n \n S. aureus \n \n at 1 month irrespective of AD outcome [ 19 ]. However, colonization with \n \n S. aureus \n \n at 6 months increased the likelihood of developing AD. Selection for dysfunctional agr mutations primarily in agrC occurred in \n \n S. aureus \n \n strains from 6-month-old infants who did not develop AD. In an epicutaneous mouse inoculation model that induces agr and skin surface inflammation, the expression of a functional S. aureus agr system was found necessary for skin colonization and the development of AD-like inflammation [ 19 ]. There is therefore considerable interest in the therapeutic potential of commensal CoNS for suppressing \n \n S. aureus \n \n in AD either by production of potent \n \n S. aureus \n \n selective antimicrobials [ 122 ], via competitive interference [ 127 ] or both since antimicrobials such as the PSMs are regulated via agr . The ability of a specific CONS species or the corresponding synthetic AIP to prevent or reduce skin colonization and damage by \n \n S. aureus \n \n CA-MRSA agr group I strain LAC has been extensively evaluated in murine epicutaneous (bacteria applied topically via gauze) and dermonecrosis (bacteria injected intradermally) skin models for \n \n S. hominis \n \n [ 127, 128 ], \n \n S. caprae \n \n [ 135 ], \n \n S. simulans \n \n [ 134 ] and \n \n S. warneri \n \n [ 136 ]. These in vivo investigations facilitated quantification of the dose-dependent efficacy of the CoNS AIPs or the corresponding CoNS strain with respect to reducing MRSA skin colonization, lesion morphologies and sizes, weight loss, bacterial burden and skin barrier integrity. \n \n S. hominis \n \n AIP-1 and AIP-2 both inhibited S. aureus agr group I activity on mouse skin and protected against epidermal damage by reducing skin lesion size, transepithelial water loss and inducing a productive host response without affecting \n \n S. aureus \n \n abundance [ 127, 128 ]. Co-challenge with \n \n S. hominis \n \n was not as effective as the \n \n S. hominis \n \n AIP-2 alone, which was suggested to be a consequence of lower levels of AIP-2 production on the skin by the human isolate employed or because mouse rather than human skin was used [ 128 ]. Similar results were obtained for MRSA co-infection with \n \n S. simulans \n \n [ 134 ], \n \n S. warneri \n \n [ 136 ] and \n \n S. caprae \n \n [ 135 ] and also in response to their respective AIPs. \n \n S. warneri \n \n AIP-2, which is more fivefold more potent in vitro than AIP-1, is one of the few naturally occurring AIPs that can inhibit S. aureus agr group IV [ 136 ]. The latter has been associated with scalded skin syndrome and the production of exfoliative toxins. In a dermonecrotic model, the \n \n S. warneri \n \n AIP-2 reduced skin lesion size and protected mice infected with an MRSA agr group IV strain from weight loss and skin damage [ 136 ]. Apart from S. caprae, none of the other CoNS AIPs reduced the MRSA agr group I bacterial burden during the course of infection to levels comparable to those observed for an MRSA agr deletion mutant [ 135 ]. This appears to be a consequence of the \n \n S. caprae \n \n AIP sensitizing MRSA to neutrophil-mediated clearance rather than a direct immunogenic effect of the AIP. Why this did not occur with the other CoNS AIPs is not clear, but may relate to the greater in vivo activity of the \n \n S. caprae \n \n AIP since significant protection was achieved at 10 µg per mouse [ 135 ] whereas comparable protection required 50 µg of the other CoNS AIPs [ 125, 128, 134, 136 ]. For each of the AIPs tested, significant protection was observed for a single AIP dose when administered at the time of infection although more subtle therapeutic effects were observed when AIPS were delivered later in the course of infection [ 127, 128, 134–136 ]. These findings provide the tantalizing prospect of employing CoNS or AIPs as a means of limiting skin infections caused by \n \n S. aureus \n \n especially in the context of multi-antibiotic resistant MRSA. Further work will need to consider asymptomatic mouse skin models, the differences between human and mouse microflora as well as innate immune and healing responses. Given that skin is colonized by diverse commensals producing many different AIPs and antimicrobials, their overall impact is likely to be complex [ 125 ]. As yet, there is little information on competitive interference when multiple CoNS strains and AIPs are involved or even other commensal bacteria such as the \n \n Corynebacterium \n \n species, which are capable of inhibiting S. aureus agr groups I to III via an unknown mechanism [ 140 ]. In a recent, double-blinded, randomized phase I clinical trial, the safety of an \n \n S. hominis \n \n strain delivered topically to forearm skin was evaluated over 7 days. Promisingly, the trial met its primary endpoint of safety with those treated with \n \n S. hominis \n \n experiencing fewer AD-associated adverse events and a reduction in \n \n S. aureus \n \n [ 141 ]."
} | 19,610 |
25374778 | PMC4217183 | pmc | 4,400 | {
"abstract": "Immunity is an important biological trait that influences the survival of individuals and the fitness of a species. Immune defenses are costly and likely compete for energy with other life-history traits, such as reproduction and growth, affecting the overall fitness of a species. Competition among these traits in scleractinian corals could influence the dynamics and structural integrity of coral reef communities. Due to variability in biological traits within populations and across species, it is likely that coral colonies within population/species adjust their immune system to the available resources. In corals, the innate immune system is composed of various pathways. The immune system components can be assessed in the absence (constitutive levels) and/or presence of stressors/pathogens (immune response). Comparisons of the constitutive levels of three immune pathways (melanin synthesis, antioxidant and antimicrobial) of closely related species of Scleractinian corals allowed to determine the link between immunity and reproduction and colony growth. First, we explored differences in constitutive immunity among closely related coral species of the genus Meandrina with different reproductive patterns (gonochoric vs. hermaphrodite). We then compared fast-growing branching vs. slow-growing massive Porites to test co-variation between constitutive immunity and growth rates and morphology in corals. Results indicate that there seems to be a relationship between constitutive immunity and sexual pattern with gonochoric species showing significantly higher levels of immunity than hermaphrodites. Therefore, gonochoric species maybe better suited to resist infections and overcome stressors. Constitutive immunity varied in relation with growth rates and colony morphology, but each species showed contrasting trends within the studied immune pathways. Fast-growing branching species appear to invest more in relatively low cost pathways of the immune system than slow-growing massive species. In corals, energetic investments in life-history traits such as reproduction and growth rate (higher energy investment) seem to have a significant impact on their capacity to respond to stressors, including infectious diseases and coral bleaching. These differences in energy investment are critical in the light of the recent environmental challenges linked to global climate change affecting these organisms. Understanding physiological trade-offs, especially those involving the immune system, will improve our understanding as to how corals could/will respond and survive in future adverse environmental conditions associated with climate change.",
"conclusion": "Concluding remarks Increasing interest in understanding the levels and mechanisms of resistance of scleractinian corals is due, in part, to the increased deterioration of coral health and coral reef environments. Combinations of biological and ecological traits have been used to classify and/or determine life-history strategies in corals, and predict coral species potential for disease susceptibility ( Díaz & Madin, 2010 ) and levels of tolerance ( Darling et al., 2012 ; Putnam et al., 2012 ). Another factor that has been considered in the survival of coral species in relation to immunity is the evolutionary history. Constitutive levels of immune-related proteins have been shown to have a phylogenetic signal among Caribbean coral species with older diverged coral groups better suited to prevent infections than coral lineages that diverged more recently ( Pinzón et al., 2014 ). However limited in species and locations, our approach takes these analyses one step further by examining how vital biological traits affect immune function among coral species. Ultimately, resistance and survivorship (fitness) of species are highly dependent on how species prioritize and distribute their resources among physiological traits. While our conclusion is restricted to two groups of Caribbean corals, the results suggest that coral immunity is an equally important variable and varies due to investment in other life-history traits such as reproductive strategy and colony growth and morphology. These findings highlight the importance of incorporating defense mechanisms (i.e., constitutive immunity levels and immune responses) to improve our ability to predict coral survival, biodiversity loss, and ecological composition of future reefs under changing conditions.",
"introduction": "Introduction The maintenance of an effective immune system is costly to organisms, but it is necessary to prevent and/or fight infections ( Lochmiller & Deerenberg, 2000 ; Rolff & Siva-Jothy, 2003 ). Interactions between physiological traits and their resulting trade-offs are highly influential in the strategies organisms use to maximize resources ( Leuzinger, Willis & Anthony, 2012 ). Limited resource availability forces trait competition, affecting the physiological functions, such as reproduction, growth and immunity ( Williams, 1966 ; Stearns, 1989 ; Hall & Hughes, 1996 ), potentially affecting species fitness ( Sandland & Minchella, 2003 ; Martin, Weil & Nelson, 2008 ; Graham et al., 2010 ; van der Most et al., 2010 ). Impacts on immunity due to differential resource allocation are common in nature. Temperate birds, for example, increase their body mass during winter affecting their immune functions ( Moreno-Rueda, 2011 ), a trade-off also observed during mating season. Environmental factors impact the distribution of energy across traits ( Nogueira, Baird & Soares, 2004 ; Finstad et al., 2010 ; Catalan et al., 2012 ), forcing immunity to be plastic to environmental changes ( Ardia, Parmentier & Vogel, 2010 ). Trade-offs, not involving immunity, have been observed in vertebrates ( Nebel et al., 2011 ; Horrocks et al., 2012 ) as well as terrestrial ( Fuller et al., 2011 ; Triggs & Knell, 2011 ; Catalan et al., 2012 ) and marine invertebrates ( Matranga et al., 2000 ; Samain, 2011 ; Coates et al., 2012 ). In scleractinian corals, the interplay between immunity and other biological traits remains elusive. Scleractinian corals are key in the development, construction and sustainability of coral reefs ( Bryant et al., 1998 ; Spalding, Raviliuous & Green, 2001 ). The success of corals as reef builders is mostly due to multiple reproductive strategies ( Harrison & Wallace, 1990 ; Harrison, 2011 ) and variable modular growth ( Chappell, 1980 ; Filatov et al., 2010 ). Sexual reproduction in corals is complex, with species showing different reproductive patterns (hermaphroditism and gonochorism), different developmental modes (broadcasting and brooding) and different gametogenic cycles and spawning timings ( Harrison, 2011 ). Colony growth and morphology vary from slow-growing in crustose and massive colonies to fast-growing in branching or columnar colonies ( Dullo, 2005 ; Díaz & Madin, 2010 ). Reproductive strategy, as well as growth rates and form, are costly and might be important in defining which species of corals ( McField, 1999 ; Knowlton, 2001 ; Baird & Marshall, 2002 ; Weil, Croquer & Urreiztieta, 2009a ; Weil, Croquer & Urreiztieta, 2009b ), and/or individuals within a species ( Van Woesik et al., 2012 ), have higher levels of resistance to environmental stressors and/or pathogenic infections. The overall efficiency of the immune system could then be influenced by the way corals allocate resources to different functions. Several coral immunity pathways, signaling and effector cascades have been characterized including the melanin synthesis cascade (i.e., prophenoloxidase, phenoloxidase and melanin deposits in cells), antioxidants (i.e., superoxide dismutase, peroxidase and catalase) ( Mydlarz et al., 2009 ; Mydlarz, McGinty & Harvell, 2010 ; Palmer, Bythell & Willis, 2010 ; Palmer et al., 2011 ) and antimicrobial compounds and proteins ( Mydlarz et al., 2009 ; Mydlarz, McGinty & Harvell, 2010 ; Palmer, Bythell & Willis, 2010 ; Palmer et al., 2011 ; Vidal-Dupiol et al., 2011a ; Vidal-Dupiol et al., 2011b ). The activity of these immune related proteins can be assessed in the absence (constitutive levels) and/or presence of pathogens (immune responses). In corals, constitutive levels and immune responses vary across species and exposure to stressors suggesting each species or population may allocate resources to immunity differently ( Palmer, Bythell & Willis, 2010 ; Palmer et al., 2011 ). Determining the effect of biological traits (i.e., energy/resource trade-offs) on immunity is critical to understand how species will thrive under adverse environments and more frequent and diverse pathogenic infections. Recent studies of ecological strategies in corals suggest immune traits, along with other biological factors, can define coral species susceptibility and/or competitive edge ( Côté & Darling, 2010 ; Darling et al., 2012 ; Putnam et al., 2012 ). In the present study, we take these analyses one-step further and test two hypotheses about the relationship between immunity and other biological traits. Our first hypothesis is that species with different sexual reproductive patterns (gonochoric vs. hermaphrodite) have similar levels of constitutive immunity. Each reproductive pattern, gonochorism and hermaphroditism, has its own advantages and disadvantages ( Hamilton & Axelrod, 1990 ; Bush, Newman & Koslow, 2004 ) influencing how resources are allocated to this or other biological traits (i.e., immunity or growth). The Caribbean genus Meandrina is ideal to test this hypothesis because these species show contrasting reproductive patterns; M. jacksoni is gonochoric while M. meandrites and M. danae are hermaphrodites ( Pinzón & Weil, 2011 ). Our second hypothesis is that constitutive immunity levels are similar in fast-growing branching corals compared to slow-growing massive species. Constitutive immunity levels were compared between the massive/crustose Porites astreoides and its fast growing branching sister species P. porites , to test the second hypothesis. While our focus is on Caribbean corals, the trends observed in this study could be extrapolated to similar corals in other locations although testing them is recommended.",
"discussion": "Results and Discussion Investment in immunological defenses to reduce risks associated with infections can challenge resources allocated to other important physiological functions such as reproduction and growth ( Lochmiller & Deerenberg, 2000 ; Sandland & Minchella, 2003 ; Simmons, 2011 ). Results of this study show that investment in immunity may be affected by investments in reproductive strategies and growth/morphology in Caribbean corals. The interaction between these traits could have an impact on coral’s survival and might be important in maintaining the integrity and persistence of coral reef ecosystems in the face of new disease outbreaks, bleaching, climate change and local environmental stressors. Comparing closely related species that differ in a given trait, like reproduction pattern and colony morphology in corals, is a good first approach towards understanding trends in coral immunity. Even though our findings are restricted to a few Caribbean corals, similar patterns might occur in other genera and/or geographic locations, where sister species have different reproductive patterns and/or morphologies. Immunity and reproductive strategy Comparisons across immune traits revealed significant differences in constitutive immunity between Meandrina species with different reproductive patterns (MANOVA Wilk’s Lambda F 12 = 8.14, p < 0.0001; Table 1 ). Melanin synthesis (putatively prophenoloxidase activity) was significantly lower in the hermaphroditic species ( M. meandrites and M. danae ) compared to the gonochoric species ( M. jacksoni ; Fig. 1 ). In contrast, both hermaphroditic species showed significantly higher superoxide dismutase activities compared to the gonochoric M. jacksoni ( Fig. 1 ; Table 1 ). Antimicrobial activity was also different, with M. danae showing the lowest level of inhibition, M. jacksoni intermediate values and M. meandrites higher inhibitions ( Fig. 1 ; Table 1 ). 10.7717/peerj.628/fig-1 Figure 1 Relation between immunity and reproduction in corals. Mean constitutive immunity among Meandrina species with different sexual patterns as determined by melanin synthesis, superoxide dismutase and antibacterial (doubling time and percent inhibition) activity. Letters on the bars indicate significant differences (Tukey post-hoc tests at p < 0.05). Data presented as mean ± standard error, for melanin synthesis as Δ absorbance 490 nm mg protein −1 min −1 , for superoxide dismutase as absorbance 450 nm mg protein −1 min −1 , and for doubling time as hours with the percentage of inhibition inside each bar. Antimicrobial data compares growth of Vibrio alginolyticus when exposed to coral extract with untreated controls. 10.7717/peerj.628/table-1 Table 1 Constitutive immune levels in phylogenetically close Caribbean corals with different reproductive patterns and colony morphologies. Comparisons of the levels of six constitutive immunity measures between coral species with different reproduction patterns ( Meandrina meandrites and M. danae vs. M. jacksoni ) and between species with different growth rates and colony morphology ( Porites astreoides vs. P. porites ). \n REPRODUCTION \n \n MORPHOLOGY \n Meandrina spp. Porites spp. \n F \n df \n p \n \n F \n df \n p \n MANOVA Wilk’s lambda \n 8.14 \n \n 12 \n \n 0.0001 \n \n 18.00 \n \n 6 \n \n 0.0001 \n Univariate tests Melanin synthesis \n 5.51 \n \n 2 \n \n 0.0101 \n \n 10.27 \n \n 1 \n \n 0.0037 \n Melanin concentration 0.70 2 0.5996 \n 58.58 \n \n 1 \n \n 0.0003 \n Superoxide dismutase \n 21.41 \n \n 2 \n \n 0.0003 \n \n 43.75 \n \n 1 \n \n 0.0003 \n Catalase \n 5.87 \n \n 2 \n \n 0.0100 \n \n 11.84 \n \n 1 \n \n 0.0024 \n Peroxidase 0.45 2 0.6390 0.11 1 0.7451 Antibacterial \n 10.47 \n \n 2 \n \n 0.0003 \n \n 16.50 \n \n 1 \n \n 0.0006 \n \n Notes. \n Bolded values indicating significant differences. \n F F statistic for the ANOVA and MANOVA analyses df degrees of freedom p p -values (corrected using False Discovery Rate correction) \n Evolutionary and ecological species success is in part driven by their sexual pattern ( Alvarez et al., 2005 ). Since hermaphrodites have to generate both, ova and sperm and prevent self-fertilization, they are likely to allocate more resources to reproduction than to other biological traits ( Michiels & Koene, 2006 ), such as immunity. Our data suggests that corals with different sexual reproductive patterns have different levels of constitutive immunity. As shown in our results, gonochoric corals show higher melanin synthesis activity and lower levels of antioxidants than their conspecific hermaphrodites. Similarly, in plants the immune related enzyme polyphenol oxidase ( Mayer, 2006 ), was more active in female flowers of Carica papaya than in hermaphrodite flowers of the same species ( Cano et al., 1996 ). Differences in immune responses across organisms can also be reflected in disease prevalence. Fungal infections in plants showed a lower load of pathogens in dioecious species than in hermaphroditic ones ( Williams, Antonovics & Rolff, 2011 ). Our findings of differences in immune levels between gonochoric and hermaphrodite Caribbean coral species may have implications to the survivorship of many species as infectious disease outbreaks increase. In the Caribbean, gonochoric species (i.e., Meandrina jacksoni , Montastraea cavernosa , Siderastrea siderea ) tend to have fewer diseases than hermaphroditic species ( Bruckner & Hill, 2009 ; Weil & Rogers, 2011 ). Furthermore, gonochoric species seem to be recruiting more successfully, becoming more common, and defining the new assemblages in declining Caribbean reefs ( Bruckner & Hill, 2009 ; Irizarry-Soto & Weil, 2009 ; Yakob & Mumby, 2011 ), which might be underscored by a higher investment in constitutive immunity. Additional factors outside the realm of reproduction may also contribute to the resistance levels of gonochoric species. Colony morphology and growth rate effects on constitutive immunity Results showed differences in baseline immunity between Porites spp. with different morphologies and growth rates (MANOVA Wilk’s Lambda F 6 = 18.00, p < 0.0001; Table 1 ). Overall, the branching P. porites seems to devote more resources to growth than to immunity given its significantly lower levels of immune protein activity compared to the massive P. astreoides ( Fig. 2 ). Nevertheless, there was some variation and opposing trends among the immune traits for each of these species. On one hand, P. astreoides seems to invest in the active elements of melanin synthesis (i.e., prophenoloxidase), rather than in the products of the pathway (melanin, 5.6 ± 0.56 mg melanin mg tissue −1 ). In contrast, P. porites seems to do the opposite by keeping constitutively higher melanin concentration in its tissues (20.38 ± 2.40 mg melanin mg tissue −1 ), instead of having high levels of the proteins involved in melanin synthesis. A similar trend to melanin concentration was found in bactericidal activity, where P. porites have higher levels of antimicrobials than P. astreoides. 10.7717/peerj.628/fig-2 Figure 2 Relation between immunity and colony morphology and growth rates in corals. Mean levels of constitutive immunity among Porites species with different growth rates and colony morphology as determined by melanin synthesis, superoxide dismutase and the antibacterial (doubling time and percent inhibition) activity. Letters on the bars indicate significant differences (Tukey post-hoc tests at p < 0.05). Data presented as mean ± standard error, for melanin synthesis as Δ absorbance 490 nm mg protein −1 min −1 , for superoxide dismutase as absorbance 450 nm mg protein −1 min −1 , and for doubling time as hours with the percentage of inhibition inside each bar. Antimicrobial data compares growth of Vibrio alginolyticus when exposed to coral extract with untreated controls. This differential investment among immune pathways could be attributed to trade-offs between immune traits ( Moret & Schimid-Hempel, 2000 ; Moret & Siva-Jothy, 2003 ; Schmid-Hempel, 2003 ; Adamo, Jensen & Younger, 2005 ; Boughton, Joop & Armitage, 2011 ) along with the different growth rates. The relatively less costly melanin cascade has a broader impact by contributing to various protective functions such as tissue repair, encapsulation, defense against microorganisms ( Mydlarz et al., 2008 ; González-Santoyo & Córdoba-Aguilar, 2011 ) and photo-protection ( Shick, Lesser & Jokiel, 1996 ). Antimicrobial activity, on the other hand, has a more specific function (i.e., exclusively killing pathogens) and is more costly to synthesize and use ( Moret & Schmid-Hempel, 2001 ). In modular organisms, like corals, growth is the result of continuous asexual production of genetically identical modules or polyps, an important characteristic that influences energy allocation ( Leuzinger, Willis & Anthony, 2012 ) and defines life-history strategies ( Darling et al., 2012 ). Colony growth in corals can be divided into fast-growing branching/foliose species and slow-growing crustose/massive species ( Dullo, 2005 ). Recent observations indicate that some coral species are more resilient and more likely to overcome the impacts associated with climate change. In the Caribbean, one of the most prominent members of this group is the massive slow growing P. astreoides ( Green, Edmunds & Carpenter, 2008 ; Edmunds, 2010 ; Palmer et al., 2011 ; Yakob & Mumby, 2011 ), which is increasing in numbers while more susceptible species, such as the branching Acropora palmata, A. cervicornis , and the columnar O. annularis and massive O. faveolata , are dying off ( Bruckner & Hill, 2009 ; Miller et al., 2009 ; Weil, Croquer & Urreiztieta, 2009a ; Weil, Croquer & Urreiztieta, 2009b ). Although no study comparing changes in abundance related to disease infections between Porites spp. exist to our knowledge, our results suggest a mechanism (investment in constitutive immunity) for the recent survival success of P. astreoides from intensive bleaching events and several white plague disease outbreaks throughout the Caribbean. Concluding remarks Increasing interest in understanding the levels and mechanisms of resistance of scleractinian corals is due, in part, to the increased deterioration of coral health and coral reef environments. Combinations of biological and ecological traits have been used to classify and/or determine life-history strategies in corals, and predict coral species potential for disease susceptibility ( Díaz & Madin, 2010 ) and levels of tolerance ( Darling et al., 2012 ; Putnam et al., 2012 ). Another factor that has been considered in the survival of coral species in relation to immunity is the evolutionary history. Constitutive levels of immune-related proteins have been shown to have a phylogenetic signal among Caribbean coral species with older diverged coral groups better suited to prevent infections than coral lineages that diverged more recently ( Pinzón et al., 2014 ). However limited in species and locations, our approach takes these analyses one step further by examining how vital biological traits affect immune function among coral species. Ultimately, resistance and survivorship (fitness) of species are highly dependent on how species prioritize and distribute their resources among physiological traits. While our conclusion is restricted to two groups of Caribbean corals, the results suggest that coral immunity is an equally important variable and varies due to investment in other life-history traits such as reproductive strategy and colony growth and morphology. These findings highlight the importance of incorporating defense mechanisms (i.e., constitutive immunity levels and immune responses) to improve our ability to predict coral survival, biodiversity loss, and ecological composition of future reefs under changing conditions."
} | 5,521 |
29074770 | null | s2 | 4,401 | {
"abstract": "Materials often exhibit a trade-off between stiffness and extensibility; for example, strengthening elastomers by increasing their cross-link density leads to embrittlement and decreased toughness. Inspired by cuticles of marine mussel byssi, we circumvent this inherent trade-off by incorporating sacrificial, reversible iron-catechol cross-links into a dry, loosely cross-linked epoxy network. The iron-containing network exhibits two to three orders of magnitude increases in stiffness, tensile strength, and tensile toughness compared to its iron-free precursor while gaining recoverable hysteretic energy dissipation and maintaining its original extensibility. Compared to previous realizations of this chemistry in hydrogels, the dry nature of the network enables larger property enhancement owing to the cooperative effects of both the increased cross-link density given by the reversible iron-catecholate complexes and the chain-restricting ionomeric nanodomains that they form."
} | 246 |
31333629 | PMC6621641 | pmc | 4,404 | {
"abstract": "Microbial community assembly in engineered biological systems is often simultaneously influenced by stochastic and deterministic processes, and the nexus of these two mechanisms remains to be further investigated. Here, three lab-scale activated sludge reactors were seeded with identical inoculum and operated in parallel under eight different sludge retention time (SRT) by sequentially reducing the SRT from 15 days to 1 day. Using 16S rRNA gene amplicon sequencing data, the microbial populations at the start-up (15-day SRT) and SRT-driven (≤10-day SRT) phases were observed to be noticeably different. Clustering results demonstrated ecological succession at the start-up phase with no consistent successional steps among the three reactors, suggesting that stochastic processes played an important role in the community assembly during primary succession. At the SRT-driven phase, the three reactors shared 31 core operational taxonomic units (OTUs). Putative primary acetate utilizers and secondary metabolizers were proposed based on K-means clustering, network and synchrony analysis. The shared core populations accounted for 65% of the total abundance, indicating that the microbial communities at the SRT-driven phase were shaped predominantly by deterministic processes. Sloan’s Neutral model and a null model analysis were performed to disentangle and quantify the relative influence of stochastic and deterministic processes on community assembly. The increased estimated migration rate in the neutral community model and the higher percentage of stochasticity in the null model implied that stochastic community assembly was intensified by strong deterministic factors. This was confirmed by the significantly different α- and β-diversity indices at SRTs shorter than 2 days and the observation that over half of the core OTUs were unshared or unsynchronized. Overall, this study provided quantitative insights into the nexus of stochastic and deterministic processes on microbial community assembly in a biological process.",
"introduction": "Introduction Microbial community is shaped by stochastic and deterministic processes ( Ofiţeru et al., 2010 ), but the extent to which these two processes influence community development is still a much debated topic ( Ayarza and Erijman, 2011 ; Valentín-Vargas et al., 2012 ; Zhou et al., 2013 ). For stochastic processes, neutral theory predicts that communities are randomly assembled through birth-death, drift, and speciation ( Sloan et al., 2006 ; Zhou and Ning, 2017 ). For deterministic processes, community assembly is thought to be governed by interspecies interactions (e.g., competition, predation, and syntrophy) and niche differentiation (e.g., operational conditions in bioreactors) ( Wang et al., 2013 ; Ju et al., 2014 ). The study of ecological succession can help understand the microbial community assembly mechanisms. Previous studies have focused on the effects of stochastic processes during microbial succession in natural ecosystems including soil ( Ferrenberg et al., 2013 ), groundwater ( Zhou et al., 2014 ), and salt marsh soil ( Dini-Andreote et al., 2015 ). Those studies consistently demonstrate that the succession of microbial community is initially driven by stochastic processes, and progressively determined by environmental factors such sodium concentration and soil organic matter ( Dini-Andreote et al., 2015 ). It is reasonable to hypothesize that stochastic and deterministic processes can interact with each other and together shape the community structure. The hypothesis is supported by the findings from natural ecosystems that deterministic disturbance may promote random community assembly ( Didham et al., 2005 ). Strong disturbances such as wildfire can reset assembly processes by exerting adverse impacts equally on all members in a microbial community, thus creating a pristine environment in which stochastic processes briefly govern community assembly ( Ferrenberg et al., 2013 ). Engineered biological systems allow the manipulation of individual deterministic factors and can be used as a model system to elucidate community assembly mechanisms ( Daims et al., 2006 ; Wells et al., 2009 ; Yang et al., 2011 ; Vanwonterghem et al., 2014 ). Among various deterministic parameters, sludge retention time (SRT, the time that microorganisms stay in a bioreactor) represents a strong driving force of the community dynamics ( Kim et al., 2011 ). Microbes with low growth rates will be washed out if short SRTs are imposed, leaving fast-growing microbes in the dominant populations. This was evidenced by the drastic decline in diversity and significant change in community composition when the SRT of a wastewater treatment plant (WWTP) was reduced from 30 to 3 days ( Vuono et al., 2014 ). The impacts of SRT was further inferred by an environment-species network constructed based on the communities from a full-scale WWTP ( Ju and Zhang, 2015 ). Modeling-based studies suggest that the combination of stochastic and deterministic processes can better explain the community structure variation in engineered systems ( Ofiţeru et al., 2010 ; Curtis et al., 2013 ). This incorporative theory is supported by experimental observations. For instance, a randomly assembled microbial community shifted to niche-based communities when the deterministic selection imposed by wastewater concentration became stronger ( Van Der Gast et al., 2008 ). The interactive effects of stochastic and deterministic processes on community assembly has not been reported in engineered biological systems, but is possibly present due to the ubiquitous impacts of stochastic processes. The objectives of this study are to characterize stochastic processes during succession at the start-up phase, to understand the deterministic effects of SRT on community dynamics and compositions, and to quantify the interactive impacts of stochastic and deterministic processes on microbial community assembly. To achieve these objectives, triplicate lab-scale activated sludge reactors were seeded with identical sludge inoculum and operated in parallel. The reactors were operated under eight different SRTs by step-wise decreasing the SRT from 15 days to 1 day over 270 days. The microbial community was characterized using 16S rRNA amplicon sequencing and the succession steps at the start-up phase were identified using hierarchical clustering. SRT-driven dynamics were further examined with K-means clustering, network analysis and the synchrony of the core populations across reactors. Finally, the Sloan’s Neutral Community Model and a null model were applied to determine the relative contribution of stochastic vs. deterministic processes.",
"discussion": "Discussion Various microbial processes have been developed in engineering systems to mimic the important microbial functions (e.g., nutrient removal, biogas production, etc.) occurring in nature to treat domestic and industrial waste streams ( Grady, Daigger et al., 2011 ). To establish the desired microbial communities, an acclimation or start-up phase is required prior to optimizing operation conditions. During this procedure, stochastic and deterministic processes are identified as two major mechanisms shaping community structure ( Ofiţeru et al., 2010 ), but it remains unclear to what extent stochastic and deterministic processes interact with each other and contribute to the selection. The present study addressed this question by characterizing the microbial communities in three lab-scale activated sludge reactors under a strong deterministic factor (i.e., SRT). In this study, the three reactors were fed with acetate as a readily degradable substrate. At the start-up phase under a long SRT of 15 days, microbial succession was observed ( Supplementary Figure S4 ). Many studies have investigated the microbial community succession in different ecosystems and reported that the succession is driven predominantly by deterministic factors ( Koenig et al., 2011 ; Podell et al., 2013 ; Datta et al., 2016 ; Lu et al., 2016 ). However, the operation conditions of the activated sludge reactors used here were less deterministic than in previously reported, and thus stochastic processes were likely playing an important role in structuring the community assembly. This can be explained by a conceptual model ( Dini-Andreote et al., 2015 ) that the microbial succession was initially governed by stochastic processes, and as the microbes altered their environment, the community was progressively shaped by interspecies interactions. The impacts of stochastic processes were demonstrated by the differentiated successional steps observed among the three reactors ( Supplementary Figure S4 ). The shift from stochastic to deterministic processes was also confirmed by the decreasing contribution of stochasticity in the null model analysis ( Figure 5C ). Collectively, the community assembly during the start-up phase was affected both by stochastic and deterministic processes. At the SRT-driven phase, the community assembly was governed mainly by deterministic factors such as niche differentiation (i.e., varied SRT) and interspecies interactions (i.e., competition for acetate and cross-feeding between the primary and secondary metabolizers). As the SRTs were sequentially reduced, a decrease in VSS ( Figure 1 ) and a clear shift in community structure ( Figure 3 ) were observed, leading to an increased F/M ratio. The results indicated the selection of slow-growing K-strategists at low F/M ratio and fast-growing r-strategists at high F/M ratio ( Andrews and Harris, 1986 ). It is reasonable to assume that the dominant K- and r-strategists at the respective SRT ranges are responsible for acetate utilization. These two types of primary metabolizers had different growth kinetics in acetate metabolisms that allowed them to dominate at different SRTs. For example, OTU 1807, which could be considered as a K-strategists, was abundant at medium SRT range ( Figure 3 ). Based on the MiDAS database ( Mcilroy et al., 2017 ), this OTU is found to be phylogenetically close to the genus Trichococcus , an aerobic heterotrophic generalist that have been widely found in activated sludge communities ( Vandewalle et al., 2012 ; Saunders et al., 2016 ). On the other hand, OTU 1397 as a r-strategist was abundant at short SRTs. It was phylogenetically related to Amaricoccus spp., which had also been observed in WWTPs around the world ( Maszenan et al., 1997 ). Findings of this study including the diversity indices, the composition of the core populations, and the modeling results all suggested that stochastic processes could exert significant effects on the community assembly at the SRT-driven phase. These supported our hypothesis that stochastic processes could be intensified by deterministic processes and the two processes could drive community assembly in an interactive way. A conceptual model predicts that when a deterministic disturbance occurs, a community will return to a pristine condition where all members having equal chance to grow, and consequently stochastic processes become the main driving force of the community assembly ( Dini-Andreote et al., 2015 ). In the present study, the interactions between stochastic and deterministic processes under shorter SRT could be explained by increased frequency of migration (i.e., higher volume of waste sludge discarded from the reactors), as evidenced by the higher estimated migration rate at shorter SRTs ( Figure 5B ). Such an interactive influence is proposed for the first time in engineered biological systems, and remains to be investigated in large-scale bioreactors. Although it has been widely accepted that stochastic and deterministic processes occur simultaneously during microbial community formation, disentangling their relative influence is still a compelling challenge facing microbial ecologists. By far, three major approaches have been developed to quantify the importance of ecological stochasticity: multivariate analysis, neutral-theory-based process models, and null modeling analysis ( Zhou and Ning, 2017 ). Variation partitioning analysis as one type of multivariate analysis has been applied to explain the different communities formed in parallel microbial electrolysis cells operated under identical conditions ( Zhou et al., 2013 ). In contrast to microbial electrolysis cells, the activated sludge reactors used here lack deterministic variables such as pH, gas composition and electricity generation, and thus a multivariate analysis may lead to biased results. Instead, we used neutral-theory-based modeling and null modeling to understand the relative influence ( Zhou et al., 2014 ; Burns et al., 2015 ). The results not only quantified the contribution of stochastic and deterministic processes to community establishment, but also confirmed the interactions between the two processes ( Figure 5 ), presenting a powerful tool to understand microbial community assembly mechanisms. It should be noted that acetate was used as the main carbon source in this study, and thus the microbial communities were shaped toward acetate metabolism. In contrast, real wastewater contains a variety of substrates, and the communities in full-scale WWTPs are composed of not only generalists (e.g., Trichococcus -related OTU 1807), but also diverse specialist guilds, including nitrifier, phosphate accumulating organisms, hydrolyzers, bulking and foaming bacteria, etc. It has been proposed that generalists are more likely to be assembled randomly, whereas specialists are selected mainly by deterministic factors ( Liao et al., 2016 ). The deterministic selection of specialists has been demonstrated in an activated sludge community, where the metabolic potential of micropollutant removal is impaired with decreased SRT ( Vuono et al., 2016a , b ). These observations suggest that, in order to better understand the assembly mechanisms in complicated communities, the interactive effects of stochastic and deterministic processes on different functional guilds should be quantified separately. Microbial immigration from sewers to WWTPs is another critical process to be considered when studying community assembly ( Frigon and Wells, 2019 ). Immigrating microbes may replace indigenous microbes in the receiving community, especially after disturbance, thereby promoting random community assembly. For example, the percentage of OTUs shared by activated sludge and influent was found to be less than 10% under low selection pressure (SRT = 10.5 days) ( Lee et al., 2015 ), but was as high as 62% under high selection pressure (SRT = 3 days) ( Callahan et al., 2016 ). The results can be explained using the abovementioned conceptual model ( Dini-Andreote et al., 2015 ), which describes that disturbance indiscriminately eliminates resident populations in a community and creates niche space for immigrants to colonize. The implication that stochastic processes can be further intensified in the presence of both strong deterministic factors and constant microbial immigration is of fundamental and practical significance to understand the interplay between these mechanisms and warrant further study."
} | 3,809 |
26039407 | PMC4454662 | pmc | 4,407 | {
"abstract": "Habitat loss and fragmentation are leading causes of species extinctions in terrestrial, aquatic and marine systems. Along coastlines, natural habitats support high biodiversity and valuable ecosystem services but are often replaced with engineered structures for coastal protection or erosion control. We coupled high-resolution shoreline condition data with an eleven-year time series of fish community structure to examine how coastal protection structures impact community stability. Our analyses revealed that the most stable fish communities were nearest natural shorelines. Structurally complex engineered shorelines appeared to promote greater stability than simpler alternatives as communities nearest vertical walls, which are among the most prevalent structures, were most dissimilar from natural shorelines and had the lowest stability. We conclude that conserving and restoring natural habitats is essential for promoting ecological stability. However, in scenarios when natural habitats are not viable, engineered landscapes designed to mimic the complexity of natural habitats may provide similar ecological functions.",
"introduction": "Introduction Coastal habitats host diverse ecological communities and provide numerous ecosystem services that affect the health, security and quality of life of human societies [ 1 , 2 ]. The degradation or loss of natural habitats is a ubiquitous problem for urbanized coastal regions and results from a multitude of anthropogenic stressors such as shoreline development and pollution [ 3 – 5 ]. Conservation scientists have made substantial efforts to understand the consequences of habitat degradation or loss, and have shown that the potential for recovering lost ecosystem functions and services exists if natural habitats are sufficiently protected and restored [ 6 ]. However, in urbanized coastal settings, restoring natural landscapes to their historical baselines is unrealistic in part due to the prominence of artificial and engineered shorelines implemented for coastal protection and erosion control [ 5 ]. For instance, largely featureless seawalls and bulkheads can degrade natural habitats, and typically support less diverse ecological communities than vegetated shorelines [ 7 , 8 ]. Conversely, structurally complex artificial structures mitigate some of these negative ecological consequences of urbanization along shorelines [ 9 , 10 ]. However, it is still unclear how the increased prevalence of engineered shorelines affects ecological communities in urbanized ecosystems at broader scales. The concepts of stability and resilience have been central foci of both fundamental ecology and applied conservation for at least the past half-century [ 11 – 14 ]. Stability and resilience are often characterized by the tendency of a system to fluctuate less [ 15 , 16 ] or its capacity to absorb perturbations and still maintain function [ 11 , 17 ]. However, the complexity of ecological communities and the inherent non-linearity of ecosystem functions complicate the study of resilience in dynamic ecosystems [ 18 – 20 ]. Studies that reveal critical properties that consistently promote the stability and resilience of communities exposed to heavy and dynamic disturbance regimes will contribute fundamentally to our understanding of how ecosystems function and help managers design strategies that ensure the maintenance of key ecosystem services. The utilization of engineered coastal structures such as vertical walls and revetments directly replaces natural shoreline habitats, disrupts land-water exchange, and alters the biophysical environment (e.g., wave climate, depth profile), potentially indirectly harming other natural habitats [ 8 , 21 , 22 ]. Only recently, and largely in response to major disasters such as Hurricanes Katrina and Sandy, have coastal protection initiatives focused on incorporating ecological and ecosystem processes alongside physical and engineering objectives [ 2 , 23 ]. Moreover, it is essential that we understand how the growing number of engineered landscapes impact the structure and resilience of ecological communities, which in turn will impact the delivery of ecosystem services. Although the societal and ecological costs of coastal habitat degradation are becoming increasingly recognized [ 23 – 25 ], coastal population size and development have continued to expand. However, very few studies to date have directly considered how the coastal protection structures that are currently replacing natural coastline features affect the stability or resilience of ecological communities [ 10 , 26 ], even though these communities when intact are highly productive and contribute to many valuable ecosystem services associated with coastal ecosystems. Here we couple high resolution shoreline condition data and an eleven-year time series of coastal fish abundances to examine how shoreline condition affects fish community stability and structure. We predicted that the communities associated with natural landscapes would fluctuate less than those near engineered shorelines, especially vertical walls that provide little to no habitat structure.",
"discussion": "Discussion We found that natural landscapes support more stable fish communities than engineered landscapes. However, not all engineered landscapes performed identically. For instance, the stability of coastal fish communities was significantly higher near engineered shorelines characterized by rubble and riprap revetments than vertical walls or vertical walls with riprap. This finding suggests that structural complexity can in some instances reduce the negative effect of engineered structures on community stability. Natural habitats such as saltmarsh, oyster reef and submerged aquatic vegetation are structurally complex and widely recognized for providing essential habitat and nursery grounds for a variety of coastal species [ 31 – 34 ]. On the other hand, vertical walls typically provide very little structural complexity, and their presence often destroys proximal natural habitats by reflecting wave energy and enhances erosive processes on adjacent shorelines [ 7 , 21 , 35 ]. Since riprap revetments appeared to promote greater fish community stability than vertical walls, our study provides further evidence that structurally complex alternatives may be less ecologically harmful when shoreline armoring is deemed necessary [ 9 , 10 ]. By analyzing community similarity and variability at annual and monthly intervals, we assessed typical fluctuations or trends of stability and evaluated the potential impacts of discrete events of disturbance. For both the sequential and overall annual time series analyses, communities associated with natural shorelines exhibited higher community similarity and fluctuated less than all engineered shoreline conditions. The global wavelet power analyses indicated that the high resilience of natural landscapes, which was observed at annual timescales, also applies at shorter time periods. During the 11 years that were examined in our study, the Alabama Gulf coast was impacted by several hurricanes and tropical storms including Allison in 2001, Ivan in 2004, Dennis, Katrina and Rita in 2005 and Ike in 2008. The Gulf of Mexico also experienced a massive oil spill following the explosion of the Deepwater Horizon drilling rig in 2010. However, studies of tidal marsh creeks following Ivan and seagrass meadows following Katrina found very little impact of the hurricanes on coastal habitats [ 36 , 37 ]. The 2010 oil spill appears to have had no detectable immediate and direct effect on the coastal habitats of Mobile Bay, but the long term and indirect effects of response actions including precautionary fishing closures remain unclear [ 38 ]. Our findings indicate that fish communities adjacent to natural shorelines were resilient to disturbance from each of these stressors, and communities associated with engineered shorelines exhibited higher temporal variability. Although engineered shorelines that mimic the complex structure of natural coastal habitats can partially restore community stability at local scales, preserving natural habitats may be important for community stability at both local and regional scales by “spilling over” via dispersal. Indeed, in fluctuating and interconnected metacommunities experiencing different environmental conditions (e.g., disturbance regimes, habitat types), connectivity can have a large impact on community stability across scales [ 39 , 40 ]. In the absence of connectivity, local communities will fluctuate asynchronously because of differences in local conditions. In such cases, local community stability will be low, but regional or metacommunity stability will be high because of the statistical averaging of asynchronously fluctuating local communities. Conversely, when connectivity is high, stability will be low at local and regional scales because dispersal will lead to large and synchronized fluctuations in community dynamics [ 39 ]. Hence, maintaining natural habitats and some level of connectivity may be critical for stability by supporting “spillover” into engineered habitats, and thereby promoting the persistence of the entire metacommunity. Determining the minimum level of connectivity and proportion of natural habitat required to promote stability without causing spatial synchrony is critical in order to preserve functions in increasingly altered ecosystems. Furthermore, understanding how the spatial extent and geographical distribution of natural habitats affect the sustainability of fisheries by controlling the delivery of larvae and adults into adjacent exploited ecosystems is critical for developing effective management programs [ 41 – 43 ]. The legacy and extraordinary degree of shoreline alteration in Mobile Bay, like many other coastal systems, dates back far longer than comprehensive ecological monitoring, making it quite challenging to understand how current fish communities adjacent to different shoreline types actually compare to a natural coastal community. However, the emergence of landscape ecology and the availability of longer term data series on ecosystem change have greatly improved our ability to understand how human activities have transformed the structure and function of natural landscapes [ 44 – 46 ]. For coastal ecosystems, these transformations have almost exclusively resulted in less desirable ecological conditions such as declining fisheries and water quality [ 6 , 47 ]. The transformation of coastal shorelines with artificial and engineered structures has been occurring for centuries but has rapidly increased in recent decades in part due to growing coastal populations and the cascading consequences of increasingly urbanized coastal ecosystems [ 25 ]. Further ecological studies in these increasingly urbanized settings are needed to resolve many uncertainties regarding the processes that mediate spatial and temporal variability in the habitat functioning of natural and engineered coastlines. Only recently have the potential impacts of coastal protection structures on ecosystems or human well-being been considered [ 10 , 18 , 23 , 26 ]. Our results indicate that conserving and restoring the integrity of natural habitats is the best approach for enhancing the resilience of coastal fish communities. In heavily developed systems and other settings where natural habitats may no longer be viable, our findings indicate that requiring coastal protection schemes provide structural complexity may mitigate some of the ecological impacts of coastal development. However, such structurally complex coastal protection features may only mimic this one function of natural habitats, and may not compensate for the loss of other ecosystem functions when natural habitats are degraded."
} | 2,956 |
23179111 | PMC3625420 | pmc | 4,408 | {
"abstract": "The capacity to discriminate between choice options is crucial for a decision-maker to avoid unprofitable options. The physical properties of rewards are presumed to be represented on context-dependent, nonlinear cognitive scales that may systematically influence reward expectation and thus choice behavior. In this study, we investigated the discrimination performance of free-flying bumblebee workers ( Bombus impatiens ) in a choice between sucrose solutions with different concentrations. We conducted two-alternative free choice experiments on two B. impatiens colonies containing some electronically tagged bumblebees foraging at an array of computer-automated artificial flowers that recorded individual choices. We mimicked natural foraging conditions by allowing uncertainty in the probability of reward delivery while maintaining certainty in reward concentration. We used a Bayesian approach to fit psychometric functions, relating the strength of preference for the higher concentration option to the relative intensity of the presented stimuli. Psychometric analysis was performed on visitation data from individually marked bumblebees and pooled data from unmarked individuals. Bumblebees preferred the more concentrated sugar solutions at high stimulus intensities and showed no preference at low stimulus intensities. The obtained psychometric function is consistent with reward evaluation based on perceived concentration contrast between choices. We found no evidence that bumblebees reduce reward expectations upon experiencing non-rewarded visits. We compare psychometric function parameters between the bumblebee B. impatiens and the flower bat Glossophaga commissarisi and discuss the relevance of psychophysics for pollinator-exerted selection pressures on plants.",
"introduction": "Introduction Decision-makers such as foraging animals or humans choosing between gambles are able to utilize information about different parameters of the choice options (i.e. probability of reward, amount of reward: Markowitz 1952 ; Kahneman and Tversky 1979 ; Wedell 1991 ; Kacelnik and Brito e Abreu 1998 ; Bateson et al. 2003 ; Cnaani et al. 2006 ; Bacon et al. 2011 ). Theoretical analyses of choice assume that different reward dimensions are integrated into some common currency, that is, “utility” (Chib et al. 2009 ; Kenrick et al. 2009 ). It is further assumed that behaviors maximizing the return currency are associated with fitness benefits and are the products of natural selection (Ritchie 1990 ; Kenrick et al. 2009 ). Underlying the capacity to make choices that maximize profitability is the ability to sense and evaluate differences among alternative options (Kacelnik and Brito e Abreu 1998 ; Livnat and Pippenger 2008 ; Shafir et al. 2008 ). Profitability maximization in the case of sequential sampling of multiple options relies on sensation (converting a physical stimulus into a neuronal firing pattern), memory (maintaining a representation of a physical stimulus over a period of time), and decision-making (comparing representations from different sources and performing a motor task based on the results of this comparison). Hereafter, we refer to the conjunction of these three processes as ‘information processing’. Since the inception of the field of psychophysics, researchers have been interested in the neural and cognitive representations of physical scales (Fechner 1860 ; Thurstone 1927 ; Stevens 1961 ). As direct observations and measurements of subjective sensations are not possible, scientists have instead focused on measuring behavioral output or neuronal activity. Psychometric analyses of scales such as sweetness, heaviness, brightness, and even abstract scales such as time and numerosity typically reveal a nonlinear correspondence between the original scale and the psychological scale (Stevens 1961 , 1969 ; Perez and Waddington 1996 ; Dehaene 2003 ; Toelch and Winter 2007 ; Billock and Tsou 2011 ; Nachev and Winter 2012 ). The logarithmic or weak power law compression of sensory information typically observed may result from the tuning properties of sensory neurons (Dayan and Abbott 2001 ) and has furthermore been suggested not only for sensory traces, but also for reactivated memories as well (Gallistel and Gelman 2000 ; Nieder and Miller 2003 ; Papini and Pellegrini 2006 ). This type of representational mechanism is robust against errors and arguably superior to alternative mechanisms (Sinn 2003 ; Portugal and Svaiter 2010 ), but it can influence choice behavior in a systematic way (Livnat and Pippenger 2008 ; Nachev and Winter 2012 ). For example, in a choice between two alternative magnitudes (e.g. numbers, sucrose concentrations, or volumes), discrimination performance is expected to improve as the difference between the options increases (distance effect) and decline as distance (the absolute difference between the two options) is kept constant but the average magnitude of the two options increases (magnitude effect, a consequence of the nonlinear compression of sensory information). A well-established tradition uses honeybees (Apinae: Apini) and more recently bumblebees (Apinae: Bombini) as model organisms for studying foraging behavior and decision-making (von Frisch 1927 ; Real 1981 ; Schmid-Hempel 1987 ; Schmid-Hempel and Schmid-Hempel 1987 ; Harder 1988 ; Waddington and Gottlieb 1990 ; Shafir et al. 2002 , 2008 ; Heinrich 2004 ; Waldron et al. 2005 ; Cnaani et al. 2006 ; Gil 2010 ). However, despite the investigations into the mechanisms of information processing in these insects (Waddington and Gottlieb 1990 ; Shafir 2000 ; Waddington 2001 ; Shafir et al. 2002 , 2008 ; Waldron et al. 2005 ; Gil 2010 ), the relationship between information processing and choice profitability remains unclear. It has been demonstrated that bees form reward expectations (Gil 2010 ) and it has been suggested that the differences between the expectation and the actual perceived reward shape the development of economic flower preferences (Waldron et al. 2005 ; Wiegmann and Smith 2009 ). An important question that still needs to be addressed is how well bees track differences along reward dimensions while foraging under conditions similar to the natural situation, where there is uncertainty whether a flower contains any nectar. In this study, we investigated the ability of the Common Eastern Bumblebee Bombus impatiens to discriminate between sucrose solutions with different sugar concentrations. Previous experiments have already shown that bumblebees are very sensitive to differences in sucrose concentration (Waddington 2001 ; Waldron et al. 2005 ; Cnaani et al. 2006 ; Wiegmann and Smith 2009 ). These studies suggest a nonlinear relationship between objective sucrose concentration (weight/weight percentage) and subjective evaluation (Waddington 2001 ) and indicate that foraging choices do not always conform to predictions based on net energy gain maximization (Schmid-Hempel 1987 ; Waldron et al. 2005 ; Cnaani et al. 2006 ). However, the precise functional relationship between discrimination performance and concentration has not yet been investigated. A traditional psychophysical method for estimating discrimination performance is fitting a psychometric function to data from n-alternative force choice tasks ( n -AFC: Treutwein and Strasburger 1999 ). The psychometric function takes a measure of the intensities of the presented stimuli as argument and gives the discrimination performance, for example, the probability with which an observer judges one stimulus to be larger in magnitude from another stimulus. In previous two-alternative choice experiments with nectar-feeding bats (Toelch and Winter 2007 ; Nachev and Winter 2012 ), the ratio of the linear difference of the stimuli to the average stimulus value was proposed as the appropriate intensity measure, because it captures the expectations that discrimination performance should increase with the difference (distance effect) and decrease with the mean magnitude of the two options (magnitude effect). The psychometric functions are typically assumed to have a sigmoidal shape and are modeled as the distribution functions of the normal, logistic, Weibull, or Gumbel distributions (Treutwein and Strasburger 1999 ; Kuss et al. 2005 ). Parameterization of the functions is preferably made so that the parameters have a meaningful biological interpretation, as is the case with the Weibull parameterization (Kuss et al. 2005 ; Fründ et al. 2011 ). The three parameters in the Weibull parameterization are the threshold, slope, and lapse rate. The threshold is the point on the curve that is halfway between the lower and the upper asymptote. In 2-AFC experiments, it usually corresponds to a discrimination performance around 75 %. The slope of the function is measured at the threshold and has been proposed as a reliability measure of sensory performance (Strasburger 2001 ). Finally, the lapse rate is seen as a measure of the frequency of errors due to motivational problems and other factors of non-perceptual nature. The lapse rate is a measure that depends on the particular task given and we suggest that in animal studies, lapsing can also result from exploratory behavior (or from competition avoidance). Foraging animals face the exploration–exploitation dilemma and will not necessarily always make choices based on expected values. In psychometric analyses, it is assumed that a forager has a constant lapse rate, that is, a constant probability to select an option not based on stimulus intensity. When a forager lapses during a specific choice in a 2-AFC experiment, its probability of selecting the correct option is at the chance level of 0.5 and equals the probability of selecting the incorrect option. Therefore, the lapse rate is calculated as one minus the upper asymptote of the psychometric curve (the estimated base rate of incorrect choices) multiplied by two. To the best of our knowledge, a psychometric function for sugar concentration discrimination performance has so far only been fitted for one species, the nectar-feeding bat Glossophaga commissarisi (Nachev and Winter 2012 ). The estimates for the lapse rate, threshold, and slope were 0.04, 0.50, and 3.41, respectively. In a recent dynamic modeling study of nectar extraction, the optimal sugar concentration for viscous dippers (animals that extract flower nectar by repeatedly dipping and retracting their tongues in the viscous liquid) was estimated at 52 % w/w (Kim et al. 2011 ). However, although both bumblebees and bats are classified as viscous dippers (Kim et al. 2011 ), typical bat-pollinated plants have nectars with much lower sugar concentrations (13–18 % w/w: Pyke and Waser 1981 ; von Helversen and Reyer 1984 ) than typical bee-pollinated plants (35 % w/w: Pyke and Waser 1981 ). This difference cannot be explained by differences in nectar-drinking style as modeled by Kim et al. ( 2011 ). On the other hand, differences in discrimination performance between the two groups of pollinators might influence the evolution of nectar concentrations in the plants they pollinate. Since bumblebees live in an ecological environment with higher nectar sugar concentrations than flower bats, bumblebees may be expected to have a better developed ability for concentration discrimination. This is because of the magnitude effect. At the higher end of a perceptive scale, that is, a higher sugar concentration, a higher sensitivity is required to discriminate between options that differ by a given distance in stimulus intensity. Here, we present the first psychometric analysis of sugar concentration discrimination performance in a nectar-feeding insect, based on two-alternative, free choice experiments with individually identifiable B. impatiens workers foraging on an array of computer-automated artificial flowers.",
"discussion": "Discussion Our bumblebees could choose between two types of sugar solutions that differed on different experimental days in their relative intensity to each other. Depending on relative intensity of difference between options, B. impatiens workers were either indifferent to differences in sucrose concentration or made more visits to the feeders with the higher concentration. Their discrimination performance can be described by the psychometric function presented in this study (Fig. 3 ). In general, the predicted relative visitation rate to the sweeter option of two concentrations (from the range 15–50 % w/w) with relative intensity x can be calculated with the following equation: 1 \\documentclass[12pt]{minimal}\n\t\t\t\t\\usepackage{amsmath}\n\t\t\t\t\\usepackage{wasysym} \n\t\t\t\t\\usepackage{amsfonts} \n\t\t\t\t\\usepackage{amssymb} \n\t\t\t\t\\usepackage{amsbsy}\n\t\t\t\t\\usepackage{mathrsfs}\n\t\t\t\t\\usepackage{upgreek}\n\t\t\t\t\\setlength{\\oddsidemargin}{-69pt}\n\t\t\t\t\\begin{document}$$ \\Uppsi (x,m,s,\\pi_{l} ) = \\frac{1}{2}\\left[ {\\pi_{l} + (1 - \\pi_{l} )\\left[ {2 - \\exp \\left( { - \\exp \\left( {\\frac{2sm}{\\ln (2)}(\\ln (x) - \\ln (m)) + \\ln (\\ln (2))} \\right)} \\right)} \\right]} \\right] $$\\end{document} where m is the threshold, s is the slope at the threshold, and π \n l is the lapse rate (from equations (1) and (11) in Kuss et al. 2005 ). For instance, the psychometric function predicts that for intensities higher than the threshold ( x > 0.25, Table 2 ), the options with the more concentrated nectars will receive at least 70 % of all visits. Because of the somewhat high estimated lapse rates (Fig. 3 ; Table 2 ), the psychometric function likely underestimates the perceptual capacity for sugar discrimination in bumblebees. Caution should also be taken when using concentrations higher than 50 % w/w, as viscosity and extraction costs are known to increase with concentration (Harder 1986 ; Kim et al. 2011 ) and may invalidate predictions based on the psychometric function. Whether that is the case could be tested by disassociating viscosity from sweetness using the inert polymer Tylose (Josens and Farina 2001 ; Borrell 2006 ; Köhler et al. 2010 ). Fig. 3 Psychometric curves for sucrose concentration discrimination. Sucrose concentration intensities are given on the abscissa and are calculated as the absolute value of the difference divided by the mean of two concentrations (see “ Methods ”). Black circles represent weighted average responses (proportion of visits to the higher sucrose concentration) over two presentations of the same pair of sucrose concentrations (Table 1 ), using number of visits as weights. The continuous curves represent the respective psychometric functions, and the dashed vertical lines indicate the psychometric function thresholds. The top three panels from left to right give data from three individually marked bumblebees. The bottom left panel gives the weighted average responses of marked bumblebees from both colonies that satisfied the minimum 800 visits per day criterion, but were not detected on a sufficient number of days for individual psychometric analysis. (Most of these data points are for single days only, rather than average values over 2 days.) The bottom middle panel gives the weighted average responses of all unmarked bumblebees from both colonies, and the bottom right panel gives the average responses ( circles ) and standard deviations ( whiskers ) calculated from pooling all data together (B20, B25, B30, miscellaneous, unmarked). The dashed curves in the bottom panels represent the psychometric function with parameters (lapse rate, threshold, and slope) averaged over the parameters of the three individually marked bumblebees \n When comparing the individually calculated psychometric functions with functions fitted on pooled data from unmarked or miscellaneous marked bumblebees (Fig. 3 ; Table 2 ), the different data sets yield similar estimates for the threshold (all in the range 0.22–0.26) and are consistent with respect to the lapse rate (all in the range 0.18–0.25). As shown in the results and in Fig. 4 , the slope is underestimated when pooled data from unmarked or miscellaneous marked bumblebees are analyzed instead of separately analyzing individual data. We conclude from this that if researchers are primarily interested in estimating the threshold rather than the slope, then similar psychometric studies (e.g. on nectar volume, or probability of reward) can be conducted without the individual transponder tracking used in this study. Fig. 4 Data pooling can cause underestimation of the psychometric function slope. The figure illustrates with a theoretical example how the averaging of individual data changes psychometric function parameters. We start with 7 “individuals” represented by psychometric functions (PFs, gray lines ) with different thresholds (mean ± SD: 0.25 ± 0.057), but equal lapse rates (0.15) and slopes (5). From the individual curves, we calculate the predicted discrimination performance values at relative intensities 0.05, 0.15, 0.25, 0.3, 0.4, 0.5, and 0.6 ( arrows ). We then average the predicted discrimination performances across animals using 200 visits per animal for each intensity value ( N = 200 visits × 7 animals = 1,400 visits per relative intensity value) and apply the algorithm for psychometric function fitting by Kuss et al. ( 2005 ). We use a flat prior for the slope, in order to exclude potential confounding effects of the prior and select all remaining parameters as described in the “ Methods ” section. The resulting psychometric curve ( dashed line ) has a slope (±95 % CI) of 4.07 ± 0.67, significantly lower than the actual value of 5 that was identical for all individuals in the initial theoretical functions ( p < 0.05). The estimates for the lapse rate (0.15 ± 0.02) and threshold (0.25 ± 0.01) do not differ from the average parameters. For comparison, the psychometric curve with parameters averaged across animals is also shown ( continuous black line ) \n Gustatory perception of sucrose concentration depends on chemoreceptors on bees’ glossae (Whitehead and Larsen 1976 ), and evaluation of this information is probably immediate. Yet bumblebees needed several hundred visits to reach asymptotic performance in their choice behavior (Fig. 2 ). The lower learning rates in comparison with the rates reported by Cnaani et al. ( 2006 ) may possibly reflect the difficulty of performing a spatial reversal task in our experiments. We interchanged the positions of higher and lower quality feeders in the experimental array daily. Impeded learning could also be explained by differences in salience of the sensory cues (visual vs. olfactory) or by a possible confounding effect of the 10-s delay rule (see “ Methods ”), which led to ca. 50 % unrewarded visits. The psychometric function predicts that bumblebee workers will be indifferent to sugar concentration differences below a relative intensity value of about 0.1. However, strong preferences for one feeder type over the other were detected in some marked bees even below this value (Table 1 , Colony 1, days 7 and 15; see also Fig. 3 , bottom left panel, points at 0.11 relative intensity). This discrimination performance may have been facilitated by a carryover effect from the previous day providing a learning phase with 2-day duration. On experimental days 7 and 15, in deviation from regular routine, there was no reversal with respect to the previous days, that is, the higher concentrations were in the same colored feeders for two consecutive nights (Table 1 ). It appears that in the absence of strong sugar concentration differences, some bumblebees did not update the remembered value of the lower concentration type as fast as others. It has been hypothesized that the difference between reward expectation and actual perceived reward drives the choice for more profitable food options in bees (Waldron et al. 2005 ; Wiegmann and Smith 2009 ). There is some field evidence that bumblebees employ a win-stay, lose-shift strategy: when they consecutively experienced low reward volumes (estimated by measuring flower handling time as proxy) at one flower species, they were more likely to switch to another species (Chittka et al. 1997 ; but see Bar-Shai et al. 2011 ). In addition to the difference between the two sucrose concentrations, the bumblebees in our experiment could also experience unrealized reward expectations when making a non-rewarded visit at each feeder type. One way to demonstrate a negative incentive contrast of this kind is to show that after experiencing two unrewarded visits at high concentration feeders (e.g. blue), bumblebees are more likely to sample a low concentration feeder (e.g. yellow) than after experiencing a reward followed by a non-rewarded visit at blue feeders (Prediction 1). (Hereafter, we refer to the high concentration feeders as blue and low concentration feeders as yellow for ease of explanation). Similarly, if the remembered value of a feeder is downgraded after a non-rewarded visit, then bumblebees should be more likely to sample a yellow feeder after making two unrewarded visits at blue feeders than after making two rewarded visits at blue feeders (Prediction 2). In order to test these predictions, we looked at the first 800 visits marked bumblebees made on days with relative intensity of 0.67 (the condition with the highest number of detected marked bumblebees). We excluded animals if they did not develop a preference above 90 % for blue feeders and performed paired t tests with probability to shift from blue to yellow as the dependent variable and the last two reward experiences (two rewards, or one reward followed by no reward, or two unrewarded visits) as the independent variable. Our results failed to support Prediction 1 (paired t (6) = −1.989, p = 0.09, N = 7 bumblebees) and Hypothesis 2 (paired t (6) = −2.454, p = 0.0495, N = 7 bumblebees). In both cases, the differences were in the opposite direction of the predicted, that is, bumblebees were more likely to shift to yellow after experiencing two rewards at blue feeders than after experiencing two non-rewarded visits at blue feeders. Our interpretation of these results is that bumblebees do not update the expected value of color marked feeders when experiencing non-rewarded visits. Despite the uncertainty and frequent changes in feeder quality, the psychometric function that describes the discrimination performance of B. impatiens workers is finely tuned, with a lower threshold (0.25) and a steeper slope (5.3) than the mean threshold (0.50) and slope (3.3) of G. commissarisi bats measured in a similar two-alternative free choice task (Nachev and Winter 2012 ). In other words, bumblebees seem to be better at discriminating small differences between sugar concentrations than nectar-feeding bats. As described in the introduction, bumblebee-pollinated plants have on average sweeter nectars than bat-pollinated plants. Here, we show that the groups also differ in their psychometric functions of sweetness perception. This raises the question how the evolution of plant nectar traits and pollinator information-processing mechanisms might be related."
} | 5,809 |
31908360 | null | s2 | 4,409 | {
"abstract": "Concern about the functional consequences of unprecedented loss in biodiversity has prompted biodiversity-ecosystem functioning (BEF) research to become one of the most active fields of ecological research in the past 25 years. Hundreds of experiments have manipulated biodiversity as an independent variable and found compelling support that the functioning of ecosystems increases with the diversity of their ecological communities. This research has also identified some of the mechanisms underlying BEF relationships, some context-dependencies of the strength of relationships, as well as implications for various ecosystem services that mankind depends upon. In this paper, we argue that a multitrophic perspective of biotic interactions in random and non-random biodiversity change scenarios is key to advance future BEF research and to address some of its most important remaining challenges. We discuss that the study and the quantification of multitrophic interactions in space and time facilitates scaling up from small-scale biodiversity manipulations and ecosystem function assessments to management-relevant spatial scales across ecosystem boundaries. We specifically consider multitrophic conceptual frameworks to understand and predict the context-dependency of BEF relationships. Moreover, we highlight the importance of the eco-evolutionary underpinnings of multitrophic BEF relationships. We outline that FAIR data (meeting the standards of findability, accessibility, interoperability, and reusability) and reproducible processing will be key to advance this field of research by making it more integrative. Finally, we show how these BEF insights may be implemented for ecosystem management, society, and policy. Given that human well-being critically depends on the multiple services provided by diverse, multitrophic communities, integrating the approaches of evolutionary ecology, community ecology, and ecosystem ecology in future BEF research will be key to refine conservation targets and develop sustainable management strategies."
} | 514 |
22258828 | PMC3272374 | pmc | 4,410 | {
"abstract": "In the study of neurosciences, and of complex biological systems in general, there is frequently a need to fit mathematical models with large numbers of parameters to highly complex datasets. Here we consider algorithms of two different classes, gradient following (GF) methods and evolutionary algorithms (EA) and examine their performance in fitting a 9-parameter model of a filter-based visual neuron to real data recorded from a sample of 107 neurons in macaque primary visual cortex (V1). Although the GF method converged very rapidly on a solution, it was highly susceptible to the effects of local minima in the error surface and produced relatively poor fits unless the initial estimates of the parameters were already very good. Conversely, although the EA required many more iterations of evaluating the model neuron’s response to a series of stimuli, it ultimately found better solutions in nearly all cases and its performance was independent of the starting parameters of the model. Thus, although the fitting process was lengthy in terms of processing time, the relative lack of human intervention in the evolutionary algorithm, and its ability ultimately to generate model fits that could be trusted as being close to optimal, made it far superior in this particular application than the gradient following methods. This is likely to be the case in many further complex systems, as are often found in neuroscience.",
"introduction": "Introduction Automated Procedures The study of complex systems , which comprise multiple interacting components with an overall behaviour that becomes hard to predict from any of the individual components, presents something of a challenge for many modern scientific disciplines. For example, having a good understanding of the response characteristics of individual neurons and their synapses, does not allow us to predict the behaviour of a network of such neurons. By the fact that these systems’ behaviour is hard to predict, our understanding of them typically requires the development of quantitative computational models. Systematic failures of the models to reproduce the behaviours of interest are useful in highlighting shortcomings in our understanding. To find such failures the typical approach is to try to fit a model to data by manipulating its parameters until its outputs are similar to those of the target system. There are two technical challenges to this endeavour. The first is how to quantify the goodness of a fit, a problem for which there are several possible measures. The second challenge is to find the best fit that the model is able to achieve, by maximising the goodness of fit measure(s). For simple problems, normally defined by convex optimisation in low number of dimensions, the choice of fitting algorithm may not be critical (Box 1966 ; Nash 1984 ). For complex systems, however, it is increasingly difficult to know whether the solution arrived at is actually the closest fit to the data that the model can achieve. That is, in cases where a model has failed to provide a good fit to the data it may not be that the model is unable to provide a good fit, merely that the fitting procedure has failed to find the optimal parameters. There is, therefore, an increasing need for powerful methods that consistently find the optimal fit to the data. Furthermore, there is also a great need for methods that are able to run with minimal need for manual human input. Here we consider the advantages and disadvantages of some of these methods as applied to a real-world problem; the fitting of a nonlinear, multi-stage, high-parameter model to somewhat noisy response data. In our case these data came from extracellular in-vivo electrophysiological recordings from primary visual cortex (V1) of the macaque, but the problem is one common to many other areas of neuroscience. Quantifying Fit Quality All fitting algorithms essentially come down to an optimisation problem in which the aim is to maximise a goodness-of-fit measure or, equivalently, minimise an error term, E , that quantifies the discrepancy between the model outputs and the data. There are several options to quantify the goodness-of-fit, including;\n (i) distance-based measures, e.g. sum of squared residuals ( SSQ ) (ii) quantifying the number of model outputs that fulfil a certain criteria, e.g. ratio that fall within the confidence interval ( RCI ), of the respective data point (iii) correlation-based measures, e.g. the percentage of the variance in the data that is explained by the model ( R \n 2 ) \n The most commonly used error terms in these optimisation problems are the distance-based measures. Usually these quantify the residual, the difference between the model and the data at each point for which data were collected. One could simply take the sum of the residual, but this would be a poor measure because positive and negative deviations between data and model outputs will cancel. The SSQ gets around the problem of cancellation between positive and negative deviations by squaring all elements of the residual and then summing them. The SSQ can be normalised by the number of data points used to give the mean squared residual ( MSQ ). This has the advantage of being independent of the number of data points used in fitting, but the fit is functionally equivalent. Similarly the root-mean-squared residual ( RMS ) can be used which has the advantage that it is then expressed in the same units as the original data. Although commonly used during optimisation routines, the distance-based measures have disadvantages. One of these is that the final value of the measure does not signal the fit quality in a general case; it is dependent on both the units and the amplitude of the data. By describing the fit quality in terms of the ratio or percentage of points falling within the confidence intervals of the data ( RCI ), less prior understanding of the data is required to judge the fit quality. For example, finding that the fitting of model neuronal outputs to data resulted in a RMS error of 40 impulses per second ( ips ), is hard to interpret without prior knowledge of typical responses, whereas the fact that 85% of the model responses were within the confidence intervals is more naturally informative. For this reason the RCI may be useful in reporting fit quality. For the purpose of the optimisation procedure the related error term is quantified as 1 − RCI . Another natural choice to quantify similarity between data and model outputs is to use the correlation between them. The correlation coefficient squared directly expresses the percentage of the variance in the data that is explained by the model ( R \n 2 ) (Howell 2004 ; Kent 1983 ). The error that needs to be minimised is 1 − R \n 2 . In the limit, where the distance-based error (e.g. SSQ) actually approaches zero, the correlation- and criteria-based measures are also naturally optimised, whereas the converse is not necessarily true. However, for noisy systems, the SSQ is not likely to approach zero, and minimising the SSQ might not optimise the other measures. An example of this can be seen in Fig. 1 . In this example synthetic data has been created and a pair of candidate model fits are presented. The curve with the lower SSQ (solid line) actually captures the data less well according to the ratio in the confidence interval ( RCI ) and the variance explained ( R \n 2 ). An advantage of using correlation-based measures is their strong affinity to the characteristic shape of the data. The disadvantage is that they take no account of the overall amplitude of the data. The RCI suffers from the problem that once a model output falls outside the confidence interval, it no longer matters how far it strays. As a result, some points may end up with extreme errors for single points in order to maximise the number of other points falling within the criterion region.\n Fig. 1 Example of two candidate model fits to sinusoidal data with noise. All data are synthetic. Note that, according to the SSQ measure of fit quality, the solid line would be considered a better fit, whereas the RCI and R \n 2 measures would both find the dashed line to be more representative of the data \n Gradient Following Algorithms For non-linear optimisation problems, in our case the minimisation of the SSQ -error, Gradient Following (GF) algorithms are the most commonly used. To visualise this optimisation problem we can imagine an error surface as a landscape in which we attempt find the lowest point. The principle of the GF method, is to follow the gradient of the error down towards this minimum. There are numerous practical approaches to accomplish this task. The Nelder–Mead method is a direct search scheme that performs geometrical manipulations of a simplex placed on the error surface (Heath 2002 ; Nelder and Mead 1965 ). This is a method that is effective in low dimensions where it is suitable to fit to non-smooth objective functions (Lagarias et al. 1998 ). In higher dimensions it is very computationally expensive compared to other GF methods. The Newton method for optimisation is an alternative and is guaranteed to converge to a minimum, at a quadratic rate (Heath 2002 ; Nash and Sofer 1996 ) for points on the error surface sufficiently close to the minimum. For more distant points, however, the convergence will be slower and it may fail altogether. Furthermore, the method requires construction of the Hessian matrix (the second derivative of the error surface) and, as with the Nelder-Mead method, it is computationally expensive for high-dimensional problems. A number of quasi-Newton methods have been developed, which approximate the Hessian, making them computationally cheaper, less sensitive to the starting parameters, or both. They typically differ in the way the Hessian approximation is made. The most common quasi-Newton schemes are secant updating, conjugate gradient, nonlinear least squares and truncated Newton methods (Heath 2002 ; Broyden 1967 ; Dixon 1972 ; Nash and Sofer 1996 ). One obvious problem with following the gradient of the error surface downwards to a minimum, is that we cannot know if this is the lowest point the surface ever is reached or just a local minimum. The choice of the length of step to be taken may be critical in this, since a large step size may allow local minima to be passed, but may of course also prevent the global minimum from being discovered. Ultimately the search is local and if we choose a starting point far from the global minimum there are no guarantees that this will be found. For a system where the parameters, and their effects on the overall behaviour of the system, are well-understood, finding a starting point close to the global minimum may be achievable. The very nature of complex systems, however, makes this task extremely difficult or impossible. An alternative might be to use multistart methods, whereby a number of GF searches are run in parallel, with different initial conditions (Bolton et al. 2000 ). Evolutionary Algorithms A different approach to optimisation is to use evolutionary algorithms (EAs). These algorithms are inspired by biological evolutionary processes, whereby a population consists of individuals from one or more generations that contribute in some form to future generations in a non-deterministic manner. As of lately this class of methods have become increasingly popular as a tool to fit complex models to neuronal data (Van Geit et al. 2008 ). Each instantiation of the model can be considered an individual whose characteristics (described by the model parameters) can be mutated and propagated to future instantiations. The probability that an individual contributes to future generations depends on its fitness . Depending on the EA being used, this contribution may take the form of parameters either being mutated and then passed directly to its ‘offspring’ (in a manner akin to cell division), or can be mutated and combined with the parameters of another individual (akin to breeding). Fitness of an individual is determined by its quality of fit relative to the rest of the population , which comprises other individuals of the current, and possibly previous, generations. The system is non-deterministic in that all mutations and breedings are made on a random basis and even individuals with poor fitness may have an opportunity to influence future generations. EAs can differ in; the evolutionary strategy (ES) governing the way in which future generations are formed, the structure of the population (e.g. the size and lifetime of each generation), and the calculation of fitness for each individual. There are numerous ESs of varying complexity which govern how the generations are produced based on the fitness evolution (Hansen et al. 1995 ; Hansen and Ostermeier 1997 ). For a given EA there are numerous choices of which ES to implement (Sbalzarini et al. 2000 ). In the simplest case only mutations occur and these mutations are drawn from a constant distribution. Alternatively the noise distribution from which the mutation is generated could be adaptive. More sophisticated strategies implement a combination of mutation and breeding, where there are also many potential ways to implement ‘breeding’. For example, the method might make use of correlations between the fittest individuals in different generations to determine the most effective way to generate following generations. The structure of the population can vary in the lifetime of each generation; in one extreme, individuals are removed from the population as soon as a new generation has been created such that only the new generation can contribute to future generations, whereas in the other extreme, all individuals in the history of the system can contribute to a new generation. In practice, the least fit individuals from past generations may be discarded since they are not likely to contribute anyway. This leads to a population structure containing the current generation and an elite pool of fit individuals from previous generations. The other consideration for the population structure is the size of each generation; if this is too small the parameter space of the model may not be well-explored, whereas if it is very large the computational cost of quantifying fit quality for each individual becomes large. Quantifying the fitness of an individual in an EA is more flexible than simply measuring the quality of fit through a single error measure, as in GF methods. In particular it is possible to perform multiple objective optimisation (MOO) (Druckmann et al. 2007 ; Vrugt and Robinson 2007 ) using EAs. If there are several possible error measures that we wish to minimise, and there may not be a unique solution that minimises all, we can track all these errors simultaneously in the fitness of the individual, which is not possible using GF methods. There are also multiple options to implement this MOO. One commonly-used solution (e.g. see Zitzler and Thiele 1999 ; Deb 1999 ; Druckmann et al. 2007 ) is to determine which individuals are on a Pareto front in the error space—the set of individuals that are not improved upon by any other individual on all error terms, they are non-dominated (see Fig. 2 )—and then calculate fitness based upon the distance of each individual from that Pareto front.\n Fig. 2 An example of a two dimensional Pareto front. Each point in the figure is a solution that gives values for two errors, E \n 1 and E \n 2 . When attempting to mimimise both errors we get a Pareto front. On the Pareto front the solutions are optimal in the sense that there is no other point for which both E \n 1 and E \n 2 are smaller, the filled dots connected with lines show the Pareto front in this figure \n Using the Algorithms in Complex Systems Due to the probabilistic nature of the EAs, there are no theoretical guarantees that a minimum will be found in finite time. On the other hand they may be less susceptible to converging on local minima, by the fact that the set of solutions that are non-dominated span several locations on the error surface and the fact that they are non-deterministic. In a system about which we have little prior knowledge, and where we expect local minima to be a problem, it may be more important that we are confident the algorithm has found something close to the true global minimum than that it has converged accurately and efficiently. In our case, we have a 9-parameter nonlinear multistage model of a neuron in primary visual cortex that we wished to fit to response data from a large set of recordings made in anaesthetised macaque monkey. By performing manual fitting we found that the GF methods that we initially used in this optimisation problem frequently converged on solutions that were clearly not optimal. This led us to consider an evolutionary approach and to explore its performance relative to GF methods for tackling our real-world optimisation problem. In particular we wished to uncover whether the EA could find a superior solution to that found by the GF method. We would also like to explore the class of multiple objectives best suited to capturing the data, and contrast the computational cost of using GFs to EAs. For the sake of this comparison we focus primarily on an implementation of the Broyden-Fletcher-Goldfarb-Shanno (BFGS) quasi-Newton GF method (Broyden 1967 ) and the Strength Pareto Evolutionary Algorithm (SPEA) (Zitzler and Thiele 1999 ; Sbalzarini et al. 2000 ; Zitzler et al. 2003 ). We found that, although the BFGS method generated slightly better fits for some cells, the SPEA was more consistent in producing good fits, such that it substantially outperformed the BFGS for some neurons. The BFGS typically took roughly an order of magnitude less computational time to perform its optimisation, but required substantially more human intervention in the choice of good starting parameters. Fitting synthetic cells revealed that the model has a very complex and flat error surface. BFGS turns out to stay in the vicinity of initial parameters while SPEA does a much more efficient global search. We also tested a combined method in which the SPEA was used to generate initial parameters for the BFGS. This memetic approach indicated that the SPEA did typically converge to a point close to the minimum even though this was not guaranteed. The SPEA also had the advantage of providing a family of fits that were, in some sense, equally good rather than giving the illusion of there being a single best fit.",
"discussion": "Discussion We have fitted a multi-stage filter-based model to electrophysiological data from macaque V1, using both traditional GF methods (in particular the BFGS) and evolutionary methods (SPEA). We have examined the computational cost of each method as well as the quality of solutions that they ultimately found for 107 neurons. The BFGS certainly has the lowest computational cost of the methods used here; converging roughly an order of magnitude faster than the SPEA. In terms of fit quality, however, it was highly variable. For many cells, fits generated by this algorithm were very good and sometimes slightly better than the SPEA fits, according to the SSQ. For the vast majority of cells SPEA reached solutions with lower SSQ , the improvement was not always very large but in 90% of the cells SPEA was better. In a number of neurons the BFGS fit was extremely poor and in these cases it was dramatically outperformed by the SPEA. Probably these very poor fits were caused by the fact that we were unable, despite a great deal of effort, to provide sufficiently good initial parameters for the model. The quasi-quantative criteria we used to pick initial parameters worked reasonably well for some neurons but in the cases where it did not BFGS was simply unable to recover a reasonable solution. This again highlights the need for expert knowledge of the system and man-hours needed to get any success with GF fitting. With an evolutionary approach this is greatly reduced as we have demonstrated. At points very far from the global minimum, the SSQ is typically very insensitive to changes in parameters; if the model response is extremely poor then small changes to the parameters will not improve it. The TNC is designed to be less sensitive to starting parameters, but carries much higher computational cost. For the dataset and model tested here we found the TNC to take roughly 30 times longer than the BFGS for the subset of neurons on which we tested it. For the SPEA the computation time was slower by roughly a factor 10, but there was much greater consistency in finding solutions that we considered to be plausible. A multi-start GF, in which several parallel GF searches are conducted from different start points, is one way to alleviate this problem. This has been shown to be superior to EAs in benchmark tests (Bolton et al. 2000 ) and comparable with EAs in certain real-world applications (Mendes and Kell 1998 ; Pettinen et al. 2006 ). This is surely dependent on the number of parameters required for the model being fitted. For a 9-parameter model, as we have here, to span the parameter space evenly, with only 2 points per parameter (which most likely would be insufficient to find the neighbourhood of the global minimum) we would need 2 9 start points, making the algorithm more than 500 times more costly than the original BFGS. That would make it much more expensive than the SPEA which does not require this large number of individuals to cover the parameter space adequately. The results from fitting on synthetic cells and the use of SPEA fits as initial data for BFGS also suggests that for this complex, high-dimensional system the result of a multi-start would be determined to a higher degree on how to pick the initial parameters than the BFGS performance. To get around the problem that the SPEA is not theoretically guaranteed to converge to a minimum, one could use a memetic approach in which the SPEA is used to find starting parameters for the BFGS. We examined this for a subset of neurons tested here and found relatively modest improvements in terms of the SSQ and RCI, and these often actually decreased the fit quality as measured by the R \n 2 . An advantage of the SPEA is its ability to optimise multiple objectives simultaneously (MOO). Different error measures emphasise different aspects of fit quality with SSQ, for example, being biased towards overall amplitude of response and R \n 2 being biased towards the shape of the tuning curves. A method capable of MOO allows the fitting to search for solutions that optimises both shape and amplitude. Methods using single objective optimisation could use a compound error term, such as a weighted average of these errors, to try to optimise both, but would always reveal a single solution and would not care about the respective size of the individual terms. For example a low compound error term might be reached by having a very low SSQ or a very high R \n 2 , or a moderately good fit in both terms. In reality there are likely to be an infinite number of solutions with identical compound errors, trading off one term against another. In SPEA a set of solutions that are, in some sense, equally good is represented by the Pareto front and this makes obvious to the user the reality that there are many good solutions, whereas single objective methods provide the illusion of there being a single ‘best’ solution to a given optimisation problem. That, of course, leaves the user with the task still of deciding which of the optimal set of solutions they wish to use or report. The biggest advantage of the method has been the fact that there was a relatively small need for human intervention for each cell. For GF algorithms the need to manually tune initial parameters, or to create sophisticated algorithms to find such parameters, requires substantial man-hours and expert knowledge of the system being modeled. In our case, even with that effort and knowledge, the initial parameters were still often insufficiently close for our GF method to find the global minimum. In contrast, although the SPEA needed a little extra coding effort initially because EAs currently need custom implementations for each problem. This effort was far outweighed by the much-reduced effort of determining initial parameters for the algorithm. In this instance the SPEA has been extremely beneficial in fitting the data to our neuron model. We fully expect that the same would be true in numerous other applications of forward models of complex systems in neuroscience. For instance, EAs have already been used to fit conductance-based single neuron models to spiking data (Druckmann et al. 2007 ), and is likely also to be applicable to computational modeling of EEG, LFP, fMRI and MEG data."
} | 6,231 |
22258828 | PMC3272374 | pmc | 4,410 | {
"abstract": "In the study of neurosciences, and of complex biological systems in general, there is frequently a need to fit mathematical models with large numbers of parameters to highly complex datasets. Here we consider algorithms of two different classes, gradient following (GF) methods and evolutionary algorithms (EA) and examine their performance in fitting a 9-parameter model of a filter-based visual neuron to real data recorded from a sample of 107 neurons in macaque primary visual cortex (V1). Although the GF method converged very rapidly on a solution, it was highly susceptible to the effects of local minima in the error surface and produced relatively poor fits unless the initial estimates of the parameters were already very good. Conversely, although the EA required many more iterations of evaluating the model neuron’s response to a series of stimuli, it ultimately found better solutions in nearly all cases and its performance was independent of the starting parameters of the model. Thus, although the fitting process was lengthy in terms of processing time, the relative lack of human intervention in the evolutionary algorithm, and its ability ultimately to generate model fits that could be trusted as being close to optimal, made it far superior in this particular application than the gradient following methods. This is likely to be the case in many further complex systems, as are often found in neuroscience.",
"introduction": "Introduction Automated Procedures The study of complex systems , which comprise multiple interacting components with an overall behaviour that becomes hard to predict from any of the individual components, presents something of a challenge for many modern scientific disciplines. For example, having a good understanding of the response characteristics of individual neurons and their synapses, does not allow us to predict the behaviour of a network of such neurons. By the fact that these systems’ behaviour is hard to predict, our understanding of them typically requires the development of quantitative computational models. Systematic failures of the models to reproduce the behaviours of interest are useful in highlighting shortcomings in our understanding. To find such failures the typical approach is to try to fit a model to data by manipulating its parameters until its outputs are similar to those of the target system. There are two technical challenges to this endeavour. The first is how to quantify the goodness of a fit, a problem for which there are several possible measures. The second challenge is to find the best fit that the model is able to achieve, by maximising the goodness of fit measure(s). For simple problems, normally defined by convex optimisation in low number of dimensions, the choice of fitting algorithm may not be critical (Box 1966 ; Nash 1984 ). For complex systems, however, it is increasingly difficult to know whether the solution arrived at is actually the closest fit to the data that the model can achieve. That is, in cases where a model has failed to provide a good fit to the data it may not be that the model is unable to provide a good fit, merely that the fitting procedure has failed to find the optimal parameters. There is, therefore, an increasing need for powerful methods that consistently find the optimal fit to the data. Furthermore, there is also a great need for methods that are able to run with minimal need for manual human input. Here we consider the advantages and disadvantages of some of these methods as applied to a real-world problem; the fitting of a nonlinear, multi-stage, high-parameter model to somewhat noisy response data. In our case these data came from extracellular in-vivo electrophysiological recordings from primary visual cortex (V1) of the macaque, but the problem is one common to many other areas of neuroscience. Quantifying Fit Quality All fitting algorithms essentially come down to an optimisation problem in which the aim is to maximise a goodness-of-fit measure or, equivalently, minimise an error term, E , that quantifies the discrepancy between the model outputs and the data. There are several options to quantify the goodness-of-fit, including;\n (i) distance-based measures, e.g. sum of squared residuals ( SSQ ) (ii) quantifying the number of model outputs that fulfil a certain criteria, e.g. ratio that fall within the confidence interval ( RCI ), of the respective data point (iii) correlation-based measures, e.g. the percentage of the variance in the data that is explained by the model ( R \n 2 ) \n The most commonly used error terms in these optimisation problems are the distance-based measures. Usually these quantify the residual, the difference between the model and the data at each point for which data were collected. One could simply take the sum of the residual, but this would be a poor measure because positive and negative deviations between data and model outputs will cancel. The SSQ gets around the problem of cancellation between positive and negative deviations by squaring all elements of the residual and then summing them. The SSQ can be normalised by the number of data points used to give the mean squared residual ( MSQ ). This has the advantage of being independent of the number of data points used in fitting, but the fit is functionally equivalent. Similarly the root-mean-squared residual ( RMS ) can be used which has the advantage that it is then expressed in the same units as the original data. Although commonly used during optimisation routines, the distance-based measures have disadvantages. One of these is that the final value of the measure does not signal the fit quality in a general case; it is dependent on both the units and the amplitude of the data. By describing the fit quality in terms of the ratio or percentage of points falling within the confidence intervals of the data ( RCI ), less prior understanding of the data is required to judge the fit quality. For example, finding that the fitting of model neuronal outputs to data resulted in a RMS error of 40 impulses per second ( ips ), is hard to interpret without prior knowledge of typical responses, whereas the fact that 85% of the model responses were within the confidence intervals is more naturally informative. For this reason the RCI may be useful in reporting fit quality. For the purpose of the optimisation procedure the related error term is quantified as 1 − RCI . Another natural choice to quantify similarity between data and model outputs is to use the correlation between them. The correlation coefficient squared directly expresses the percentage of the variance in the data that is explained by the model ( R \n 2 ) (Howell 2004 ; Kent 1983 ). The error that needs to be minimised is 1 − R \n 2 . In the limit, where the distance-based error (e.g. SSQ) actually approaches zero, the correlation- and criteria-based measures are also naturally optimised, whereas the converse is not necessarily true. However, for noisy systems, the SSQ is not likely to approach zero, and minimising the SSQ might not optimise the other measures. An example of this can be seen in Fig. 1 . In this example synthetic data has been created and a pair of candidate model fits are presented. The curve with the lower SSQ (solid line) actually captures the data less well according to the ratio in the confidence interval ( RCI ) and the variance explained ( R \n 2 ). An advantage of using correlation-based measures is their strong affinity to the characteristic shape of the data. The disadvantage is that they take no account of the overall amplitude of the data. The RCI suffers from the problem that once a model output falls outside the confidence interval, it no longer matters how far it strays. As a result, some points may end up with extreme errors for single points in order to maximise the number of other points falling within the criterion region.\n Fig. 1 Example of two candidate model fits to sinusoidal data with noise. All data are synthetic. Note that, according to the SSQ measure of fit quality, the solid line would be considered a better fit, whereas the RCI and R \n 2 measures would both find the dashed line to be more representative of the data \n Gradient Following Algorithms For non-linear optimisation problems, in our case the minimisation of the SSQ -error, Gradient Following (GF) algorithms are the most commonly used. To visualise this optimisation problem we can imagine an error surface as a landscape in which we attempt find the lowest point. The principle of the GF method, is to follow the gradient of the error down towards this minimum. There are numerous practical approaches to accomplish this task. The Nelder–Mead method is a direct search scheme that performs geometrical manipulations of a simplex placed on the error surface (Heath 2002 ; Nelder and Mead 1965 ). This is a method that is effective in low dimensions where it is suitable to fit to non-smooth objective functions (Lagarias et al. 1998 ). In higher dimensions it is very computationally expensive compared to other GF methods. The Newton method for optimisation is an alternative and is guaranteed to converge to a minimum, at a quadratic rate (Heath 2002 ; Nash and Sofer 1996 ) for points on the error surface sufficiently close to the minimum. For more distant points, however, the convergence will be slower and it may fail altogether. Furthermore, the method requires construction of the Hessian matrix (the second derivative of the error surface) and, as with the Nelder-Mead method, it is computationally expensive for high-dimensional problems. A number of quasi-Newton methods have been developed, which approximate the Hessian, making them computationally cheaper, less sensitive to the starting parameters, or both. They typically differ in the way the Hessian approximation is made. The most common quasi-Newton schemes are secant updating, conjugate gradient, nonlinear least squares and truncated Newton methods (Heath 2002 ; Broyden 1967 ; Dixon 1972 ; Nash and Sofer 1996 ). One obvious problem with following the gradient of the error surface downwards to a minimum, is that we cannot know if this is the lowest point the surface ever is reached or just a local minimum. The choice of the length of step to be taken may be critical in this, since a large step size may allow local minima to be passed, but may of course also prevent the global minimum from being discovered. Ultimately the search is local and if we choose a starting point far from the global minimum there are no guarantees that this will be found. For a system where the parameters, and their effects on the overall behaviour of the system, are well-understood, finding a starting point close to the global minimum may be achievable. The very nature of complex systems, however, makes this task extremely difficult or impossible. An alternative might be to use multistart methods, whereby a number of GF searches are run in parallel, with different initial conditions (Bolton et al. 2000 ). Evolutionary Algorithms A different approach to optimisation is to use evolutionary algorithms (EAs). These algorithms are inspired by biological evolutionary processes, whereby a population consists of individuals from one or more generations that contribute in some form to future generations in a non-deterministic manner. As of lately this class of methods have become increasingly popular as a tool to fit complex models to neuronal data (Van Geit et al. 2008 ). Each instantiation of the model can be considered an individual whose characteristics (described by the model parameters) can be mutated and propagated to future instantiations. The probability that an individual contributes to future generations depends on its fitness . Depending on the EA being used, this contribution may take the form of parameters either being mutated and then passed directly to its ‘offspring’ (in a manner akin to cell division), or can be mutated and combined with the parameters of another individual (akin to breeding). Fitness of an individual is determined by its quality of fit relative to the rest of the population , which comprises other individuals of the current, and possibly previous, generations. The system is non-deterministic in that all mutations and breedings are made on a random basis and even individuals with poor fitness may have an opportunity to influence future generations. EAs can differ in; the evolutionary strategy (ES) governing the way in which future generations are formed, the structure of the population (e.g. the size and lifetime of each generation), and the calculation of fitness for each individual. There are numerous ESs of varying complexity which govern how the generations are produced based on the fitness evolution (Hansen et al. 1995 ; Hansen and Ostermeier 1997 ). For a given EA there are numerous choices of which ES to implement (Sbalzarini et al. 2000 ). In the simplest case only mutations occur and these mutations are drawn from a constant distribution. Alternatively the noise distribution from which the mutation is generated could be adaptive. More sophisticated strategies implement a combination of mutation and breeding, where there are also many potential ways to implement ‘breeding’. For example, the method might make use of correlations between the fittest individuals in different generations to determine the most effective way to generate following generations. The structure of the population can vary in the lifetime of each generation; in one extreme, individuals are removed from the population as soon as a new generation has been created such that only the new generation can contribute to future generations, whereas in the other extreme, all individuals in the history of the system can contribute to a new generation. In practice, the least fit individuals from past generations may be discarded since they are not likely to contribute anyway. This leads to a population structure containing the current generation and an elite pool of fit individuals from previous generations. The other consideration for the population structure is the size of each generation; if this is too small the parameter space of the model may not be well-explored, whereas if it is very large the computational cost of quantifying fit quality for each individual becomes large. Quantifying the fitness of an individual in an EA is more flexible than simply measuring the quality of fit through a single error measure, as in GF methods. In particular it is possible to perform multiple objective optimisation (MOO) (Druckmann et al. 2007 ; Vrugt and Robinson 2007 ) using EAs. If there are several possible error measures that we wish to minimise, and there may not be a unique solution that minimises all, we can track all these errors simultaneously in the fitness of the individual, which is not possible using GF methods. There are also multiple options to implement this MOO. One commonly-used solution (e.g. see Zitzler and Thiele 1999 ; Deb 1999 ; Druckmann et al. 2007 ) is to determine which individuals are on a Pareto front in the error space—the set of individuals that are not improved upon by any other individual on all error terms, they are non-dominated (see Fig. 2 )—and then calculate fitness based upon the distance of each individual from that Pareto front.\n Fig. 2 An example of a two dimensional Pareto front. Each point in the figure is a solution that gives values for two errors, E \n 1 and E \n 2 . When attempting to mimimise both errors we get a Pareto front. On the Pareto front the solutions are optimal in the sense that there is no other point for which both E \n 1 and E \n 2 are smaller, the filled dots connected with lines show the Pareto front in this figure \n Using the Algorithms in Complex Systems Due to the probabilistic nature of the EAs, there are no theoretical guarantees that a minimum will be found in finite time. On the other hand they may be less susceptible to converging on local minima, by the fact that the set of solutions that are non-dominated span several locations on the error surface and the fact that they are non-deterministic. In a system about which we have little prior knowledge, and where we expect local minima to be a problem, it may be more important that we are confident the algorithm has found something close to the true global minimum than that it has converged accurately and efficiently. In our case, we have a 9-parameter nonlinear multistage model of a neuron in primary visual cortex that we wished to fit to response data from a large set of recordings made in anaesthetised macaque monkey. By performing manual fitting we found that the GF methods that we initially used in this optimisation problem frequently converged on solutions that were clearly not optimal. This led us to consider an evolutionary approach and to explore its performance relative to GF methods for tackling our real-world optimisation problem. In particular we wished to uncover whether the EA could find a superior solution to that found by the GF method. We would also like to explore the class of multiple objectives best suited to capturing the data, and contrast the computational cost of using GFs to EAs. For the sake of this comparison we focus primarily on an implementation of the Broyden-Fletcher-Goldfarb-Shanno (BFGS) quasi-Newton GF method (Broyden 1967 ) and the Strength Pareto Evolutionary Algorithm (SPEA) (Zitzler and Thiele 1999 ; Sbalzarini et al. 2000 ; Zitzler et al. 2003 ). We found that, although the BFGS method generated slightly better fits for some cells, the SPEA was more consistent in producing good fits, such that it substantially outperformed the BFGS for some neurons. The BFGS typically took roughly an order of magnitude less computational time to perform its optimisation, but required substantially more human intervention in the choice of good starting parameters. Fitting synthetic cells revealed that the model has a very complex and flat error surface. BFGS turns out to stay in the vicinity of initial parameters while SPEA does a much more efficient global search. We also tested a combined method in which the SPEA was used to generate initial parameters for the BFGS. This memetic approach indicated that the SPEA did typically converge to a point close to the minimum even though this was not guaranteed. The SPEA also had the advantage of providing a family of fits that were, in some sense, equally good rather than giving the illusion of there being a single best fit.",
"discussion": "Discussion We have fitted a multi-stage filter-based model to electrophysiological data from macaque V1, using both traditional GF methods (in particular the BFGS) and evolutionary methods (SPEA). We have examined the computational cost of each method as well as the quality of solutions that they ultimately found for 107 neurons. The BFGS certainly has the lowest computational cost of the methods used here; converging roughly an order of magnitude faster than the SPEA. In terms of fit quality, however, it was highly variable. For many cells, fits generated by this algorithm were very good and sometimes slightly better than the SPEA fits, according to the SSQ. For the vast majority of cells SPEA reached solutions with lower SSQ , the improvement was not always very large but in 90% of the cells SPEA was better. In a number of neurons the BFGS fit was extremely poor and in these cases it was dramatically outperformed by the SPEA. Probably these very poor fits were caused by the fact that we were unable, despite a great deal of effort, to provide sufficiently good initial parameters for the model. The quasi-quantative criteria we used to pick initial parameters worked reasonably well for some neurons but in the cases where it did not BFGS was simply unable to recover a reasonable solution. This again highlights the need for expert knowledge of the system and man-hours needed to get any success with GF fitting. With an evolutionary approach this is greatly reduced as we have demonstrated. At points very far from the global minimum, the SSQ is typically very insensitive to changes in parameters; if the model response is extremely poor then small changes to the parameters will not improve it. The TNC is designed to be less sensitive to starting parameters, but carries much higher computational cost. For the dataset and model tested here we found the TNC to take roughly 30 times longer than the BFGS for the subset of neurons on which we tested it. For the SPEA the computation time was slower by roughly a factor 10, but there was much greater consistency in finding solutions that we considered to be plausible. A multi-start GF, in which several parallel GF searches are conducted from different start points, is one way to alleviate this problem. This has been shown to be superior to EAs in benchmark tests (Bolton et al. 2000 ) and comparable with EAs in certain real-world applications (Mendes and Kell 1998 ; Pettinen et al. 2006 ). This is surely dependent on the number of parameters required for the model being fitted. For a 9-parameter model, as we have here, to span the parameter space evenly, with only 2 points per parameter (which most likely would be insufficient to find the neighbourhood of the global minimum) we would need 2 9 start points, making the algorithm more than 500 times more costly than the original BFGS. That would make it much more expensive than the SPEA which does not require this large number of individuals to cover the parameter space adequately. The results from fitting on synthetic cells and the use of SPEA fits as initial data for BFGS also suggests that for this complex, high-dimensional system the result of a multi-start would be determined to a higher degree on how to pick the initial parameters than the BFGS performance. To get around the problem that the SPEA is not theoretically guaranteed to converge to a minimum, one could use a memetic approach in which the SPEA is used to find starting parameters for the BFGS. We examined this for a subset of neurons tested here and found relatively modest improvements in terms of the SSQ and RCI, and these often actually decreased the fit quality as measured by the R \n 2 . An advantage of the SPEA is its ability to optimise multiple objectives simultaneously (MOO). Different error measures emphasise different aspects of fit quality with SSQ, for example, being biased towards overall amplitude of response and R \n 2 being biased towards the shape of the tuning curves. A method capable of MOO allows the fitting to search for solutions that optimises both shape and amplitude. Methods using single objective optimisation could use a compound error term, such as a weighted average of these errors, to try to optimise both, but would always reveal a single solution and would not care about the respective size of the individual terms. For example a low compound error term might be reached by having a very low SSQ or a very high R \n 2 , or a moderately good fit in both terms. In reality there are likely to be an infinite number of solutions with identical compound errors, trading off one term against another. In SPEA a set of solutions that are, in some sense, equally good is represented by the Pareto front and this makes obvious to the user the reality that there are many good solutions, whereas single objective methods provide the illusion of there being a single ‘best’ solution to a given optimisation problem. That, of course, leaves the user with the task still of deciding which of the optimal set of solutions they wish to use or report. The biggest advantage of the method has been the fact that there was a relatively small need for human intervention for each cell. For GF algorithms the need to manually tune initial parameters, or to create sophisticated algorithms to find such parameters, requires substantial man-hours and expert knowledge of the system being modeled. In our case, even with that effort and knowledge, the initial parameters were still often insufficiently close for our GF method to find the global minimum. In contrast, although the SPEA needed a little extra coding effort initially because EAs currently need custom implementations for each problem. This effort was far outweighed by the much-reduced effort of determining initial parameters for the algorithm. In this instance the SPEA has been extremely beneficial in fitting the data to our neuron model. We fully expect that the same would be true in numerous other applications of forward models of complex systems in neuroscience. For instance, EAs have already been used to fit conductance-based single neuron models to spiking data (Druckmann et al. 2007 ), and is likely also to be applicable to computational modeling of EEG, LFP, fMRI and MEG data."
} | 6,231 |
22258828 | PMC3272374 | pmc | 4,411 | {
"abstract": "In the study of neurosciences, and of complex biological systems in general, there is frequently a need to fit mathematical models with large numbers of parameters to highly complex datasets. Here we consider algorithms of two different classes, gradient following (GF) methods and evolutionary algorithms (EA) and examine their performance in fitting a 9-parameter model of a filter-based visual neuron to real data recorded from a sample of 107 neurons in macaque primary visual cortex (V1). Although the GF method converged very rapidly on a solution, it was highly susceptible to the effects of local minima in the error surface and produced relatively poor fits unless the initial estimates of the parameters were already very good. Conversely, although the EA required many more iterations of evaluating the model neuron’s response to a series of stimuli, it ultimately found better solutions in nearly all cases and its performance was independent of the starting parameters of the model. Thus, although the fitting process was lengthy in terms of processing time, the relative lack of human intervention in the evolutionary algorithm, and its ability ultimately to generate model fits that could be trusted as being close to optimal, made it far superior in this particular application than the gradient following methods. This is likely to be the case in many further complex systems, as are often found in neuroscience.",
"introduction": "Introduction Automated Procedures The study of complex systems , which comprise multiple interacting components with an overall behaviour that becomes hard to predict from any of the individual components, presents something of a challenge for many modern scientific disciplines. For example, having a good understanding of the response characteristics of individual neurons and their synapses, does not allow us to predict the behaviour of a network of such neurons. By the fact that these systems’ behaviour is hard to predict, our understanding of them typically requires the development of quantitative computational models. Systematic failures of the models to reproduce the behaviours of interest are useful in highlighting shortcomings in our understanding. To find such failures the typical approach is to try to fit a model to data by manipulating its parameters until its outputs are similar to those of the target system. There are two technical challenges to this endeavour. The first is how to quantify the goodness of a fit, a problem for which there are several possible measures. The second challenge is to find the best fit that the model is able to achieve, by maximising the goodness of fit measure(s). For simple problems, normally defined by convex optimisation in low number of dimensions, the choice of fitting algorithm may not be critical (Box 1966 ; Nash 1984 ). For complex systems, however, it is increasingly difficult to know whether the solution arrived at is actually the closest fit to the data that the model can achieve. That is, in cases where a model has failed to provide a good fit to the data it may not be that the model is unable to provide a good fit, merely that the fitting procedure has failed to find the optimal parameters. There is, therefore, an increasing need for powerful methods that consistently find the optimal fit to the data. Furthermore, there is also a great need for methods that are able to run with minimal need for manual human input. Here we consider the advantages and disadvantages of some of these methods as applied to a real-world problem; the fitting of a nonlinear, multi-stage, high-parameter model to somewhat noisy response data. In our case these data came from extracellular in-vivo electrophysiological recordings from primary visual cortex (V1) of the macaque, but the problem is one common to many other areas of neuroscience. Quantifying Fit Quality All fitting algorithms essentially come down to an optimisation problem in which the aim is to maximise a goodness-of-fit measure or, equivalently, minimise an error term, E , that quantifies the discrepancy between the model outputs and the data. There are several options to quantify the goodness-of-fit, including;\n (i) distance-based measures, e.g. sum of squared residuals ( SSQ ) (ii) quantifying the number of model outputs that fulfil a certain criteria, e.g. ratio that fall within the confidence interval ( RCI ), of the respective data point (iii) correlation-based measures, e.g. the percentage of the variance in the data that is explained by the model ( R \n 2 ) \n The most commonly used error terms in these optimisation problems are the distance-based measures. Usually these quantify the residual, the difference between the model and the data at each point for which data were collected. One could simply take the sum of the residual, but this would be a poor measure because positive and negative deviations between data and model outputs will cancel. The SSQ gets around the problem of cancellation between positive and negative deviations by squaring all elements of the residual and then summing them. The SSQ can be normalised by the number of data points used to give the mean squared residual ( MSQ ). This has the advantage of being independent of the number of data points used in fitting, but the fit is functionally equivalent. Similarly the root-mean-squared residual ( RMS ) can be used which has the advantage that it is then expressed in the same units as the original data. Although commonly used during optimisation routines, the distance-based measures have disadvantages. One of these is that the final value of the measure does not signal the fit quality in a general case; it is dependent on both the units and the amplitude of the data. By describing the fit quality in terms of the ratio or percentage of points falling within the confidence intervals of the data ( RCI ), less prior understanding of the data is required to judge the fit quality. For example, finding that the fitting of model neuronal outputs to data resulted in a RMS error of 40 impulses per second ( ips ), is hard to interpret without prior knowledge of typical responses, whereas the fact that 85% of the model responses were within the confidence intervals is more naturally informative. For this reason the RCI may be useful in reporting fit quality. For the purpose of the optimisation procedure the related error term is quantified as 1 − RCI . Another natural choice to quantify similarity between data and model outputs is to use the correlation between them. The correlation coefficient squared directly expresses the percentage of the variance in the data that is explained by the model ( R \n 2 ) (Howell 2004 ; Kent 1983 ). The error that needs to be minimised is 1 − R \n 2 . In the limit, where the distance-based error (e.g. SSQ) actually approaches zero, the correlation- and criteria-based measures are also naturally optimised, whereas the converse is not necessarily true. However, for noisy systems, the SSQ is not likely to approach zero, and minimising the SSQ might not optimise the other measures. An example of this can be seen in Fig. 1 . In this example synthetic data has been created and a pair of candidate model fits are presented. The curve with the lower SSQ (solid line) actually captures the data less well according to the ratio in the confidence interval ( RCI ) and the variance explained ( R \n 2 ). An advantage of using correlation-based measures is their strong affinity to the characteristic shape of the data. The disadvantage is that they take no account of the overall amplitude of the data. The RCI suffers from the problem that once a model output falls outside the confidence interval, it no longer matters how far it strays. As a result, some points may end up with extreme errors for single points in order to maximise the number of other points falling within the criterion region.\n Fig. 1 Example of two candidate model fits to sinusoidal data with noise. All data are synthetic. Note that, according to the SSQ measure of fit quality, the solid line would be considered a better fit, whereas the RCI and R \n 2 measures would both find the dashed line to be more representative of the data \n Gradient Following Algorithms For non-linear optimisation problems, in our case the minimisation of the SSQ -error, Gradient Following (GF) algorithms are the most commonly used. To visualise this optimisation problem we can imagine an error surface as a landscape in which we attempt find the lowest point. The principle of the GF method, is to follow the gradient of the error down towards this minimum. There are numerous practical approaches to accomplish this task. The Nelder–Mead method is a direct search scheme that performs geometrical manipulations of a simplex placed on the error surface (Heath 2002 ; Nelder and Mead 1965 ). This is a method that is effective in low dimensions where it is suitable to fit to non-smooth objective functions (Lagarias et al. 1998 ). In higher dimensions it is very computationally expensive compared to other GF methods. The Newton method for optimisation is an alternative and is guaranteed to converge to a minimum, at a quadratic rate (Heath 2002 ; Nash and Sofer 1996 ) for points on the error surface sufficiently close to the minimum. For more distant points, however, the convergence will be slower and it may fail altogether. Furthermore, the method requires construction of the Hessian matrix (the second derivative of the error surface) and, as with the Nelder-Mead method, it is computationally expensive for high-dimensional problems. A number of quasi-Newton methods have been developed, which approximate the Hessian, making them computationally cheaper, less sensitive to the starting parameters, or both. They typically differ in the way the Hessian approximation is made. The most common quasi-Newton schemes are secant updating, conjugate gradient, nonlinear least squares and truncated Newton methods (Heath 2002 ; Broyden 1967 ; Dixon 1972 ; Nash and Sofer 1996 ). One obvious problem with following the gradient of the error surface downwards to a minimum, is that we cannot know if this is the lowest point the surface ever is reached or just a local minimum. The choice of the length of step to be taken may be critical in this, since a large step size may allow local minima to be passed, but may of course also prevent the global minimum from being discovered. Ultimately the search is local and if we choose a starting point far from the global minimum there are no guarantees that this will be found. For a system where the parameters, and their effects on the overall behaviour of the system, are well-understood, finding a starting point close to the global minimum may be achievable. The very nature of complex systems, however, makes this task extremely difficult or impossible. An alternative might be to use multistart methods, whereby a number of GF searches are run in parallel, with different initial conditions (Bolton et al. 2000 ). Evolutionary Algorithms A different approach to optimisation is to use evolutionary algorithms (EAs). These algorithms are inspired by biological evolutionary processes, whereby a population consists of individuals from one or more generations that contribute in some form to future generations in a non-deterministic manner. As of lately this class of methods have become increasingly popular as a tool to fit complex models to neuronal data (Van Geit et al. 2008 ). Each instantiation of the model can be considered an individual whose characteristics (described by the model parameters) can be mutated and propagated to future instantiations. The probability that an individual contributes to future generations depends on its fitness . Depending on the EA being used, this contribution may take the form of parameters either being mutated and then passed directly to its ‘offspring’ (in a manner akin to cell division), or can be mutated and combined with the parameters of another individual (akin to breeding). Fitness of an individual is determined by its quality of fit relative to the rest of the population , which comprises other individuals of the current, and possibly previous, generations. The system is non-deterministic in that all mutations and breedings are made on a random basis and even individuals with poor fitness may have an opportunity to influence future generations. EAs can differ in; the evolutionary strategy (ES) governing the way in which future generations are formed, the structure of the population (e.g. the size and lifetime of each generation), and the calculation of fitness for each individual. There are numerous ESs of varying complexity which govern how the generations are produced based on the fitness evolution (Hansen et al. 1995 ; Hansen and Ostermeier 1997 ). For a given EA there are numerous choices of which ES to implement (Sbalzarini et al. 2000 ). In the simplest case only mutations occur and these mutations are drawn from a constant distribution. Alternatively the noise distribution from which the mutation is generated could be adaptive. More sophisticated strategies implement a combination of mutation and breeding, where there are also many potential ways to implement ‘breeding’. For example, the method might make use of correlations between the fittest individuals in different generations to determine the most effective way to generate following generations. The structure of the population can vary in the lifetime of each generation; in one extreme, individuals are removed from the population as soon as a new generation has been created such that only the new generation can contribute to future generations, whereas in the other extreme, all individuals in the history of the system can contribute to a new generation. In practice, the least fit individuals from past generations may be discarded since they are not likely to contribute anyway. This leads to a population structure containing the current generation and an elite pool of fit individuals from previous generations. The other consideration for the population structure is the size of each generation; if this is too small the parameter space of the model may not be well-explored, whereas if it is very large the computational cost of quantifying fit quality for each individual becomes large. Quantifying the fitness of an individual in an EA is more flexible than simply measuring the quality of fit through a single error measure, as in GF methods. In particular it is possible to perform multiple objective optimisation (MOO) (Druckmann et al. 2007 ; Vrugt and Robinson 2007 ) using EAs. If there are several possible error measures that we wish to minimise, and there may not be a unique solution that minimises all, we can track all these errors simultaneously in the fitness of the individual, which is not possible using GF methods. There are also multiple options to implement this MOO. One commonly-used solution (e.g. see Zitzler and Thiele 1999 ; Deb 1999 ; Druckmann et al. 2007 ) is to determine which individuals are on a Pareto front in the error space—the set of individuals that are not improved upon by any other individual on all error terms, they are non-dominated (see Fig. 2 )—and then calculate fitness based upon the distance of each individual from that Pareto front.\n Fig. 2 An example of a two dimensional Pareto front. Each point in the figure is a solution that gives values for two errors, E \n 1 and E \n 2 . When attempting to mimimise both errors we get a Pareto front. On the Pareto front the solutions are optimal in the sense that there is no other point for which both E \n 1 and E \n 2 are smaller, the filled dots connected with lines show the Pareto front in this figure \n Using the Algorithms in Complex Systems Due to the probabilistic nature of the EAs, there are no theoretical guarantees that a minimum will be found in finite time. On the other hand they may be less susceptible to converging on local minima, by the fact that the set of solutions that are non-dominated span several locations on the error surface and the fact that they are non-deterministic. In a system about which we have little prior knowledge, and where we expect local minima to be a problem, it may be more important that we are confident the algorithm has found something close to the true global minimum than that it has converged accurately and efficiently. In our case, we have a 9-parameter nonlinear multistage model of a neuron in primary visual cortex that we wished to fit to response data from a large set of recordings made in anaesthetised macaque monkey. By performing manual fitting we found that the GF methods that we initially used in this optimisation problem frequently converged on solutions that were clearly not optimal. This led us to consider an evolutionary approach and to explore its performance relative to GF methods for tackling our real-world optimisation problem. In particular we wished to uncover whether the EA could find a superior solution to that found by the GF method. We would also like to explore the class of multiple objectives best suited to capturing the data, and contrast the computational cost of using GFs to EAs. For the sake of this comparison we focus primarily on an implementation of the Broyden-Fletcher-Goldfarb-Shanno (BFGS) quasi-Newton GF method (Broyden 1967 ) and the Strength Pareto Evolutionary Algorithm (SPEA) (Zitzler and Thiele 1999 ; Sbalzarini et al. 2000 ; Zitzler et al. 2003 ). We found that, although the BFGS method generated slightly better fits for some cells, the SPEA was more consistent in producing good fits, such that it substantially outperformed the BFGS for some neurons. The BFGS typically took roughly an order of magnitude less computational time to perform its optimisation, but required substantially more human intervention in the choice of good starting parameters. Fitting synthetic cells revealed that the model has a very complex and flat error surface. BFGS turns out to stay in the vicinity of initial parameters while SPEA does a much more efficient global search. We also tested a combined method in which the SPEA was used to generate initial parameters for the BFGS. This memetic approach indicated that the SPEA did typically converge to a point close to the minimum even though this was not guaranteed. The SPEA also had the advantage of providing a family of fits that were, in some sense, equally good rather than giving the illusion of there being a single best fit.",
"discussion": "Discussion We have fitted a multi-stage filter-based model to electrophysiological data from macaque V1, using both traditional GF methods (in particular the BFGS) and evolutionary methods (SPEA). We have examined the computational cost of each method as well as the quality of solutions that they ultimately found for 107 neurons. The BFGS certainly has the lowest computational cost of the methods used here; converging roughly an order of magnitude faster than the SPEA. In terms of fit quality, however, it was highly variable. For many cells, fits generated by this algorithm were very good and sometimes slightly better than the SPEA fits, according to the SSQ. For the vast majority of cells SPEA reached solutions with lower SSQ , the improvement was not always very large but in 90% of the cells SPEA was better. In a number of neurons the BFGS fit was extremely poor and in these cases it was dramatically outperformed by the SPEA. Probably these very poor fits were caused by the fact that we were unable, despite a great deal of effort, to provide sufficiently good initial parameters for the model. The quasi-quantative criteria we used to pick initial parameters worked reasonably well for some neurons but in the cases where it did not BFGS was simply unable to recover a reasonable solution. This again highlights the need for expert knowledge of the system and man-hours needed to get any success with GF fitting. With an evolutionary approach this is greatly reduced as we have demonstrated. At points very far from the global minimum, the SSQ is typically very insensitive to changes in parameters; if the model response is extremely poor then small changes to the parameters will not improve it. The TNC is designed to be less sensitive to starting parameters, but carries much higher computational cost. For the dataset and model tested here we found the TNC to take roughly 30 times longer than the BFGS for the subset of neurons on which we tested it. For the SPEA the computation time was slower by roughly a factor 10, but there was much greater consistency in finding solutions that we considered to be plausible. A multi-start GF, in which several parallel GF searches are conducted from different start points, is one way to alleviate this problem. This has been shown to be superior to EAs in benchmark tests (Bolton et al. 2000 ) and comparable with EAs in certain real-world applications (Mendes and Kell 1998 ; Pettinen et al. 2006 ). This is surely dependent on the number of parameters required for the model being fitted. For a 9-parameter model, as we have here, to span the parameter space evenly, with only 2 points per parameter (which most likely would be insufficient to find the neighbourhood of the global minimum) we would need 2 9 start points, making the algorithm more than 500 times more costly than the original BFGS. That would make it much more expensive than the SPEA which does not require this large number of individuals to cover the parameter space adequately. The results from fitting on synthetic cells and the use of SPEA fits as initial data for BFGS also suggests that for this complex, high-dimensional system the result of a multi-start would be determined to a higher degree on how to pick the initial parameters than the BFGS performance. To get around the problem that the SPEA is not theoretically guaranteed to converge to a minimum, one could use a memetic approach in which the SPEA is used to find starting parameters for the BFGS. We examined this for a subset of neurons tested here and found relatively modest improvements in terms of the SSQ and RCI, and these often actually decreased the fit quality as measured by the R \n 2 . An advantage of the SPEA is its ability to optimise multiple objectives simultaneously (MOO). Different error measures emphasise different aspects of fit quality with SSQ, for example, being biased towards overall amplitude of response and R \n 2 being biased towards the shape of the tuning curves. A method capable of MOO allows the fitting to search for solutions that optimises both shape and amplitude. Methods using single objective optimisation could use a compound error term, such as a weighted average of these errors, to try to optimise both, but would always reveal a single solution and would not care about the respective size of the individual terms. For example a low compound error term might be reached by having a very low SSQ or a very high R \n 2 , or a moderately good fit in both terms. In reality there are likely to be an infinite number of solutions with identical compound errors, trading off one term against another. In SPEA a set of solutions that are, in some sense, equally good is represented by the Pareto front and this makes obvious to the user the reality that there are many good solutions, whereas single objective methods provide the illusion of there being a single ‘best’ solution to a given optimisation problem. That, of course, leaves the user with the task still of deciding which of the optimal set of solutions they wish to use or report. The biggest advantage of the method has been the fact that there was a relatively small need for human intervention for each cell. For GF algorithms the need to manually tune initial parameters, or to create sophisticated algorithms to find such parameters, requires substantial man-hours and expert knowledge of the system being modeled. In our case, even with that effort and knowledge, the initial parameters were still often insufficiently close for our GF method to find the global minimum. In contrast, although the SPEA needed a little extra coding effort initially because EAs currently need custom implementations for each problem. This effort was far outweighed by the much-reduced effort of determining initial parameters for the algorithm. In this instance the SPEA has been extremely beneficial in fitting the data to our neuron model. We fully expect that the same would be true in numerous other applications of forward models of complex systems in neuroscience. For instance, EAs have already been used to fit conductance-based single neuron models to spiking data (Druckmann et al. 2007 ), and is likely also to be applicable to computational modeling of EEG, LFP, fMRI and MEG data."
} | 6,231 |
22258828 | PMC3272374 | pmc | 4,411 | {
"abstract": "In the study of neurosciences, and of complex biological systems in general, there is frequently a need to fit mathematical models with large numbers of parameters to highly complex datasets. Here we consider algorithms of two different classes, gradient following (GF) methods and evolutionary algorithms (EA) and examine their performance in fitting a 9-parameter model of a filter-based visual neuron to real data recorded from a sample of 107 neurons in macaque primary visual cortex (V1). Although the GF method converged very rapidly on a solution, it was highly susceptible to the effects of local minima in the error surface and produced relatively poor fits unless the initial estimates of the parameters were already very good. Conversely, although the EA required many more iterations of evaluating the model neuron’s response to a series of stimuli, it ultimately found better solutions in nearly all cases and its performance was independent of the starting parameters of the model. Thus, although the fitting process was lengthy in terms of processing time, the relative lack of human intervention in the evolutionary algorithm, and its ability ultimately to generate model fits that could be trusted as being close to optimal, made it far superior in this particular application than the gradient following methods. This is likely to be the case in many further complex systems, as are often found in neuroscience.",
"introduction": "Introduction Automated Procedures The study of complex systems , which comprise multiple interacting components with an overall behaviour that becomes hard to predict from any of the individual components, presents something of a challenge for many modern scientific disciplines. For example, having a good understanding of the response characteristics of individual neurons and their synapses, does not allow us to predict the behaviour of a network of such neurons. By the fact that these systems’ behaviour is hard to predict, our understanding of them typically requires the development of quantitative computational models. Systematic failures of the models to reproduce the behaviours of interest are useful in highlighting shortcomings in our understanding. To find such failures the typical approach is to try to fit a model to data by manipulating its parameters until its outputs are similar to those of the target system. There are two technical challenges to this endeavour. The first is how to quantify the goodness of a fit, a problem for which there are several possible measures. The second challenge is to find the best fit that the model is able to achieve, by maximising the goodness of fit measure(s). For simple problems, normally defined by convex optimisation in low number of dimensions, the choice of fitting algorithm may not be critical (Box 1966 ; Nash 1984 ). For complex systems, however, it is increasingly difficult to know whether the solution arrived at is actually the closest fit to the data that the model can achieve. That is, in cases where a model has failed to provide a good fit to the data it may not be that the model is unable to provide a good fit, merely that the fitting procedure has failed to find the optimal parameters. There is, therefore, an increasing need for powerful methods that consistently find the optimal fit to the data. Furthermore, there is also a great need for methods that are able to run with minimal need for manual human input. Here we consider the advantages and disadvantages of some of these methods as applied to a real-world problem; the fitting of a nonlinear, multi-stage, high-parameter model to somewhat noisy response data. In our case these data came from extracellular in-vivo electrophysiological recordings from primary visual cortex (V1) of the macaque, but the problem is one common to many other areas of neuroscience. Quantifying Fit Quality All fitting algorithms essentially come down to an optimisation problem in which the aim is to maximise a goodness-of-fit measure or, equivalently, minimise an error term, E , that quantifies the discrepancy between the model outputs and the data. There are several options to quantify the goodness-of-fit, including;\n (i) distance-based measures, e.g. sum of squared residuals ( SSQ ) (ii) quantifying the number of model outputs that fulfil a certain criteria, e.g. ratio that fall within the confidence interval ( RCI ), of the respective data point (iii) correlation-based measures, e.g. the percentage of the variance in the data that is explained by the model ( R \n 2 ) \n The most commonly used error terms in these optimisation problems are the distance-based measures. Usually these quantify the residual, the difference between the model and the data at each point for which data were collected. One could simply take the sum of the residual, but this would be a poor measure because positive and negative deviations between data and model outputs will cancel. The SSQ gets around the problem of cancellation between positive and negative deviations by squaring all elements of the residual and then summing them. The SSQ can be normalised by the number of data points used to give the mean squared residual ( MSQ ). This has the advantage of being independent of the number of data points used in fitting, but the fit is functionally equivalent. Similarly the root-mean-squared residual ( RMS ) can be used which has the advantage that it is then expressed in the same units as the original data. Although commonly used during optimisation routines, the distance-based measures have disadvantages. One of these is that the final value of the measure does not signal the fit quality in a general case; it is dependent on both the units and the amplitude of the data. By describing the fit quality in terms of the ratio or percentage of points falling within the confidence intervals of the data ( RCI ), less prior understanding of the data is required to judge the fit quality. For example, finding that the fitting of model neuronal outputs to data resulted in a RMS error of 40 impulses per second ( ips ), is hard to interpret without prior knowledge of typical responses, whereas the fact that 85% of the model responses were within the confidence intervals is more naturally informative. For this reason the RCI may be useful in reporting fit quality. For the purpose of the optimisation procedure the related error term is quantified as 1 − RCI . Another natural choice to quantify similarity between data and model outputs is to use the correlation between them. The correlation coefficient squared directly expresses the percentage of the variance in the data that is explained by the model ( R \n 2 ) (Howell 2004 ; Kent 1983 ). The error that needs to be minimised is 1 − R \n 2 . In the limit, where the distance-based error (e.g. SSQ) actually approaches zero, the correlation- and criteria-based measures are also naturally optimised, whereas the converse is not necessarily true. However, for noisy systems, the SSQ is not likely to approach zero, and minimising the SSQ might not optimise the other measures. An example of this can be seen in Fig. 1 . In this example synthetic data has been created and a pair of candidate model fits are presented. The curve with the lower SSQ (solid line) actually captures the data less well according to the ratio in the confidence interval ( RCI ) and the variance explained ( R \n 2 ). An advantage of using correlation-based measures is their strong affinity to the characteristic shape of the data. The disadvantage is that they take no account of the overall amplitude of the data. The RCI suffers from the problem that once a model output falls outside the confidence interval, it no longer matters how far it strays. As a result, some points may end up with extreme errors for single points in order to maximise the number of other points falling within the criterion region.\n Fig. 1 Example of two candidate model fits to sinusoidal data with noise. All data are synthetic. Note that, according to the SSQ measure of fit quality, the solid line would be considered a better fit, whereas the RCI and R \n 2 measures would both find the dashed line to be more representative of the data \n Gradient Following Algorithms For non-linear optimisation problems, in our case the minimisation of the SSQ -error, Gradient Following (GF) algorithms are the most commonly used. To visualise this optimisation problem we can imagine an error surface as a landscape in which we attempt find the lowest point. The principle of the GF method, is to follow the gradient of the error down towards this minimum. There are numerous practical approaches to accomplish this task. The Nelder–Mead method is a direct search scheme that performs geometrical manipulations of a simplex placed on the error surface (Heath 2002 ; Nelder and Mead 1965 ). This is a method that is effective in low dimensions where it is suitable to fit to non-smooth objective functions (Lagarias et al. 1998 ). In higher dimensions it is very computationally expensive compared to other GF methods. The Newton method for optimisation is an alternative and is guaranteed to converge to a minimum, at a quadratic rate (Heath 2002 ; Nash and Sofer 1996 ) for points on the error surface sufficiently close to the minimum. For more distant points, however, the convergence will be slower and it may fail altogether. Furthermore, the method requires construction of the Hessian matrix (the second derivative of the error surface) and, as with the Nelder-Mead method, it is computationally expensive for high-dimensional problems. A number of quasi-Newton methods have been developed, which approximate the Hessian, making them computationally cheaper, less sensitive to the starting parameters, or both. They typically differ in the way the Hessian approximation is made. The most common quasi-Newton schemes are secant updating, conjugate gradient, nonlinear least squares and truncated Newton methods (Heath 2002 ; Broyden 1967 ; Dixon 1972 ; Nash and Sofer 1996 ). One obvious problem with following the gradient of the error surface downwards to a minimum, is that we cannot know if this is the lowest point the surface ever is reached or just a local minimum. The choice of the length of step to be taken may be critical in this, since a large step size may allow local minima to be passed, but may of course also prevent the global minimum from being discovered. Ultimately the search is local and if we choose a starting point far from the global minimum there are no guarantees that this will be found. For a system where the parameters, and their effects on the overall behaviour of the system, are well-understood, finding a starting point close to the global minimum may be achievable. The very nature of complex systems, however, makes this task extremely difficult or impossible. An alternative might be to use multistart methods, whereby a number of GF searches are run in parallel, with different initial conditions (Bolton et al. 2000 ). Evolutionary Algorithms A different approach to optimisation is to use evolutionary algorithms (EAs). These algorithms are inspired by biological evolutionary processes, whereby a population consists of individuals from one or more generations that contribute in some form to future generations in a non-deterministic manner. As of lately this class of methods have become increasingly popular as a tool to fit complex models to neuronal data (Van Geit et al. 2008 ). Each instantiation of the model can be considered an individual whose characteristics (described by the model parameters) can be mutated and propagated to future instantiations. The probability that an individual contributes to future generations depends on its fitness . Depending on the EA being used, this contribution may take the form of parameters either being mutated and then passed directly to its ‘offspring’ (in a manner akin to cell division), or can be mutated and combined with the parameters of another individual (akin to breeding). Fitness of an individual is determined by its quality of fit relative to the rest of the population , which comprises other individuals of the current, and possibly previous, generations. The system is non-deterministic in that all mutations and breedings are made on a random basis and even individuals with poor fitness may have an opportunity to influence future generations. EAs can differ in; the evolutionary strategy (ES) governing the way in which future generations are formed, the structure of the population (e.g. the size and lifetime of each generation), and the calculation of fitness for each individual. There are numerous ESs of varying complexity which govern how the generations are produced based on the fitness evolution (Hansen et al. 1995 ; Hansen and Ostermeier 1997 ). For a given EA there are numerous choices of which ES to implement (Sbalzarini et al. 2000 ). In the simplest case only mutations occur and these mutations are drawn from a constant distribution. Alternatively the noise distribution from which the mutation is generated could be adaptive. More sophisticated strategies implement a combination of mutation and breeding, where there are also many potential ways to implement ‘breeding’. For example, the method might make use of correlations between the fittest individuals in different generations to determine the most effective way to generate following generations. The structure of the population can vary in the lifetime of each generation; in one extreme, individuals are removed from the population as soon as a new generation has been created such that only the new generation can contribute to future generations, whereas in the other extreme, all individuals in the history of the system can contribute to a new generation. In practice, the least fit individuals from past generations may be discarded since they are not likely to contribute anyway. This leads to a population structure containing the current generation and an elite pool of fit individuals from previous generations. The other consideration for the population structure is the size of each generation; if this is too small the parameter space of the model may not be well-explored, whereas if it is very large the computational cost of quantifying fit quality for each individual becomes large. Quantifying the fitness of an individual in an EA is more flexible than simply measuring the quality of fit through a single error measure, as in GF methods. In particular it is possible to perform multiple objective optimisation (MOO) (Druckmann et al. 2007 ; Vrugt and Robinson 2007 ) using EAs. If there are several possible error measures that we wish to minimise, and there may not be a unique solution that minimises all, we can track all these errors simultaneously in the fitness of the individual, which is not possible using GF methods. There are also multiple options to implement this MOO. One commonly-used solution (e.g. see Zitzler and Thiele 1999 ; Deb 1999 ; Druckmann et al. 2007 ) is to determine which individuals are on a Pareto front in the error space—the set of individuals that are not improved upon by any other individual on all error terms, they are non-dominated (see Fig. 2 )—and then calculate fitness based upon the distance of each individual from that Pareto front.\n Fig. 2 An example of a two dimensional Pareto front. Each point in the figure is a solution that gives values for two errors, E \n 1 and E \n 2 . When attempting to mimimise both errors we get a Pareto front. On the Pareto front the solutions are optimal in the sense that there is no other point for which both E \n 1 and E \n 2 are smaller, the filled dots connected with lines show the Pareto front in this figure \n Using the Algorithms in Complex Systems Due to the probabilistic nature of the EAs, there are no theoretical guarantees that a minimum will be found in finite time. On the other hand they may be less susceptible to converging on local minima, by the fact that the set of solutions that are non-dominated span several locations on the error surface and the fact that they are non-deterministic. In a system about which we have little prior knowledge, and where we expect local minima to be a problem, it may be more important that we are confident the algorithm has found something close to the true global minimum than that it has converged accurately and efficiently. In our case, we have a 9-parameter nonlinear multistage model of a neuron in primary visual cortex that we wished to fit to response data from a large set of recordings made in anaesthetised macaque monkey. By performing manual fitting we found that the GF methods that we initially used in this optimisation problem frequently converged on solutions that were clearly not optimal. This led us to consider an evolutionary approach and to explore its performance relative to GF methods for tackling our real-world optimisation problem. In particular we wished to uncover whether the EA could find a superior solution to that found by the GF method. We would also like to explore the class of multiple objectives best suited to capturing the data, and contrast the computational cost of using GFs to EAs. For the sake of this comparison we focus primarily on an implementation of the Broyden-Fletcher-Goldfarb-Shanno (BFGS) quasi-Newton GF method (Broyden 1967 ) and the Strength Pareto Evolutionary Algorithm (SPEA) (Zitzler and Thiele 1999 ; Sbalzarini et al. 2000 ; Zitzler et al. 2003 ). We found that, although the BFGS method generated slightly better fits for some cells, the SPEA was more consistent in producing good fits, such that it substantially outperformed the BFGS for some neurons. The BFGS typically took roughly an order of magnitude less computational time to perform its optimisation, but required substantially more human intervention in the choice of good starting parameters. Fitting synthetic cells revealed that the model has a very complex and flat error surface. BFGS turns out to stay in the vicinity of initial parameters while SPEA does a much more efficient global search. We also tested a combined method in which the SPEA was used to generate initial parameters for the BFGS. This memetic approach indicated that the SPEA did typically converge to a point close to the minimum even though this was not guaranteed. The SPEA also had the advantage of providing a family of fits that were, in some sense, equally good rather than giving the illusion of there being a single best fit.",
"discussion": "Discussion We have fitted a multi-stage filter-based model to electrophysiological data from macaque V1, using both traditional GF methods (in particular the BFGS) and evolutionary methods (SPEA). We have examined the computational cost of each method as well as the quality of solutions that they ultimately found for 107 neurons. The BFGS certainly has the lowest computational cost of the methods used here; converging roughly an order of magnitude faster than the SPEA. In terms of fit quality, however, it was highly variable. For many cells, fits generated by this algorithm were very good and sometimes slightly better than the SPEA fits, according to the SSQ. For the vast majority of cells SPEA reached solutions with lower SSQ , the improvement was not always very large but in 90% of the cells SPEA was better. In a number of neurons the BFGS fit was extremely poor and in these cases it was dramatically outperformed by the SPEA. Probably these very poor fits were caused by the fact that we were unable, despite a great deal of effort, to provide sufficiently good initial parameters for the model. The quasi-quantative criteria we used to pick initial parameters worked reasonably well for some neurons but in the cases where it did not BFGS was simply unable to recover a reasonable solution. This again highlights the need for expert knowledge of the system and man-hours needed to get any success with GF fitting. With an evolutionary approach this is greatly reduced as we have demonstrated. At points very far from the global minimum, the SSQ is typically very insensitive to changes in parameters; if the model response is extremely poor then small changes to the parameters will not improve it. The TNC is designed to be less sensitive to starting parameters, but carries much higher computational cost. For the dataset and model tested here we found the TNC to take roughly 30 times longer than the BFGS for the subset of neurons on which we tested it. For the SPEA the computation time was slower by roughly a factor 10, but there was much greater consistency in finding solutions that we considered to be plausible. A multi-start GF, in which several parallel GF searches are conducted from different start points, is one way to alleviate this problem. This has been shown to be superior to EAs in benchmark tests (Bolton et al. 2000 ) and comparable with EAs in certain real-world applications (Mendes and Kell 1998 ; Pettinen et al. 2006 ). This is surely dependent on the number of parameters required for the model being fitted. For a 9-parameter model, as we have here, to span the parameter space evenly, with only 2 points per parameter (which most likely would be insufficient to find the neighbourhood of the global minimum) we would need 2 9 start points, making the algorithm more than 500 times more costly than the original BFGS. That would make it much more expensive than the SPEA which does not require this large number of individuals to cover the parameter space adequately. The results from fitting on synthetic cells and the use of SPEA fits as initial data for BFGS also suggests that for this complex, high-dimensional system the result of a multi-start would be determined to a higher degree on how to pick the initial parameters than the BFGS performance. To get around the problem that the SPEA is not theoretically guaranteed to converge to a minimum, one could use a memetic approach in which the SPEA is used to find starting parameters for the BFGS. We examined this for a subset of neurons tested here and found relatively modest improvements in terms of the SSQ and RCI, and these often actually decreased the fit quality as measured by the R \n 2 . An advantage of the SPEA is its ability to optimise multiple objectives simultaneously (MOO). Different error measures emphasise different aspects of fit quality with SSQ, for example, being biased towards overall amplitude of response and R \n 2 being biased towards the shape of the tuning curves. A method capable of MOO allows the fitting to search for solutions that optimises both shape and amplitude. Methods using single objective optimisation could use a compound error term, such as a weighted average of these errors, to try to optimise both, but would always reveal a single solution and would not care about the respective size of the individual terms. For example a low compound error term might be reached by having a very low SSQ or a very high R \n 2 , or a moderately good fit in both terms. In reality there are likely to be an infinite number of solutions with identical compound errors, trading off one term against another. In SPEA a set of solutions that are, in some sense, equally good is represented by the Pareto front and this makes obvious to the user the reality that there are many good solutions, whereas single objective methods provide the illusion of there being a single ‘best’ solution to a given optimisation problem. That, of course, leaves the user with the task still of deciding which of the optimal set of solutions they wish to use or report. The biggest advantage of the method has been the fact that there was a relatively small need for human intervention for each cell. For GF algorithms the need to manually tune initial parameters, or to create sophisticated algorithms to find such parameters, requires substantial man-hours and expert knowledge of the system being modeled. In our case, even with that effort and knowledge, the initial parameters were still often insufficiently close for our GF method to find the global minimum. In contrast, although the SPEA needed a little extra coding effort initially because EAs currently need custom implementations for each problem. This effort was far outweighed by the much-reduced effort of determining initial parameters for the algorithm. In this instance the SPEA has been extremely beneficial in fitting the data to our neuron model. We fully expect that the same would be true in numerous other applications of forward models of complex systems in neuroscience. For instance, EAs have already been used to fit conductance-based single neuron models to spiking data (Druckmann et al. 2007 ), and is likely also to be applicable to computational modeling of EEG, LFP, fMRI and MEG data."
} | 6,231 |
24386315 | PMC3875502 | pmc | 4,412 | {
"abstract": "Aerobic anoxygenic phototrophs (AAPs) as being photoheterotrophs require organic substrates for growth and use light as a supplementary energy source under oxic conditions. We hypothesized that AAPs benefit from light particularly under carbon and electron donor limitation. The effect of light was determined in long-term starvation experiments with Dinoroseobacter shibae DFL 12 T in both complex marine broth and defined minimal medium with succinate as the sole carbon source. The cells were starved over six months under three conditions: continuous darkness (DD), continuous light (LL), and light/dark cycle (LD, 12 h/12 h, 12 µmol photons m −2 s −1 ). LD starvation at low light intensity resulted in 10-fold higher total cell and viable counts, and higher bacteriochlorophyll a and polyhydroxyalkanoate contents. This coincided with better physiological fitness as determined by respiration rates, proton translocation and ATP concentrations. In contrast, LD starvation at high light intensity (>22 µmol photons m −2 s −1 , LD conditions) resulted in decreasing cell survival rates but increasing carotenoid concentrations, indicating a photo-protective response. Cells grown in complex medium survived longer starvation (more than 20 weeks) than those grown in minimal medium. Our experiments show that D. shibae benefits from the light and dark cycle, particularly during starvation.",
"introduction": "Introduction Aerobic anoxygenic phototrophs (AAPs) are widespread in marine habitats [1] – [3] but also occur in brackish [4] – [5] and fresh waters [6] – [7] , saline lakes [8] , soil [9] , and hot springs [10] . Environmental studies based on infrared microscopy [11] , pigment extraction [12] , or sequencing of the pufM gene [1] , [13] – [14] revealed high abundances and a heterogeneous distribution of AAPs world-wide [15] – [16] . It has been estimated that AAPs contribute up to 5% of the photosynthetic electron transport in oceanic surface waters [17] . Comparable to purple bacteria, AAPs are capable of light-driven and respiratory electron transport for their energy metabolism. However, this activity is different in several aspects. AAPs use light energy under oxic conditions and contain much less bacteriochlorophyll a (BChl a ), with sometimes higher carotenoid than BChl a concentrations [18] . Increased amounts of carotenoids may possibly be a protective response against reactive oxygen species (ROS), which are produced in the light under oxic conditions [19] . As the biosynthesis of BChl a is known to be sensitive to ROS, AAPs synthesize BChl a only in the dark [20] – [21] and require a day and night cycle to sustain their photosynthetic activity. In accordance with the low BChl content , light is only used as a supplementary energy source [22] – [23] . Furthermore, AAPs are not autotrophs, but heterotrophs. Their photosynthetic activity is not used for CO 2 fixation, but prevents the oxidation of organic substrates. Light-driven proton translocation and ATP formation by AAPs have been investigated in several studies [24] – [25] . Recently, Tomasch et al. [26] studied the transcriptional response of D. shibae under different light regimes. The substrate-saving effect of light energy was demonstrated by lower respiration rates in the light [23] , [25] . In chemostat cultures, the light-dependent increase of growth yields [27] , [28] was reversely correlated to the growth rate and increased at low rates up to 110% [29] . The purpose of the present study was to identify the conditions under which AAPs benefit the most from their photosynthetic capacities. We hypothesized that AAPs benefit from photon energy under conditions of carbon and electron donor limitation, as proteomic responses to starvation and light conditions among AAPs have been reported [30] . Here, we performed long-term starvation experiments under different light regimes, and investigated survival and physiological fitness of Dinoroseobacter shibae \n [31] a representative of the globally abundant marine Roseobacter clade [32] – [34] .",
"discussion": "Discussion Our study demonstrated that day and night cycle substantially increases the survival of Dinoroseobacter shibae during long term starvation. However, high light intensity may be detrimental and induce protective reactions. The Diurnal Light and Dark Regime Increases Cell Fitness All measured parameters show that Dinoroseobacter shibae benefits from exposure to a diurnal light and dark regime. Under this condition, the cells survived starvation with better physiological fitness. As cells were shown to preserve intracellular PHA, known to be the main storage material in D. shibae \n [28] , cells maintained increased dry mass to protein ratios. Cell pigmentation, morphology and motility of cells grown under the diurnal light cycle were less affected than those in the dark or under continuous light. Viability and fitness of cells decreased in similar manner both in the dark and under continuous light, which implies that light inhibits BChl a synthesis and prevents photosynthetic activity over longer periods. The increased fitness of LD-starved cells was also found by quantifying respiration, proton translocation and ATP regeneration. The beneficial effects became more pronounced with prolonged starvation. These observations could be explained by one underlying rationale: Light utilization appears rather as a mechanism to survive starvation than as a growth-promoting factor, which improves survival and physiological fitness several-fold. Light as a Stress Factor Not only continuous light, but also high intensities during day and night cycles had negative effects on the cells. The harmful light effects were also recognized from the pigment analysis, as cells increased their carotenoid contents at high light intensities likely as a protective response. By low pigment concentrations, AAPs might prevent formation of reactive oxygen species [19] . This was corroborated by the low optimum light intensity of 12 µmol photons m −2 s −1 , which might additionally minimize ROS exposure. The optimum light intensity of 12 µmol photons m −2 s −1 is much lower than mid-day intensities in, e.g., the North Pacific Gyre, which average 50–150 µmol photons m −2 s −1 at 20–150 m water depth [40] . Thus, low-light adaption of Dinoroseobacter shibae minimizes ROS exposure and might explain why none of the AAPs is able to grow purely phototrophic. Ecological Implications The light-supported survival of starvation of D.shibae has also been reported for other groups of bacteria. In a bacteriochlorophyll a -containing gammaproteobacterium, expression of the photosynthesis genes depended on the type of carbon source [41] . Also some proteorhodopsin-carrying bacteria such as the abundant Cand. Pelagibacter ubique \n [42] , Dokdonia sp. strain MED134 [43] and Vibrio sp. strain AND4 [44] , benefit from light during starvation. Considering the specific advantage of the AAPs in their natural environment, it becomes clear that light is not the overall limiting factor for their distribution. Cottrell et al. found that AAPs are distributed over the whole photic zone [40] . In a particle-rich estuary, the BChl a concentrations of AAPs varied in response to particles but not to light limitation [45] . In a recent study, Čuperová et al. showed that the DOC concentrations influenced the AAP abundance in alpine lakes [46] . In agreement with these findings, our study suggests that the limitation of organic substrates might promote the competitiveness of AAPs. Nutrient limitation is a predominating feature in most oceanic regions. Biomass of AAPs in the South Pacific Ocean was on average two-fold higher than that of other prokaryotic cells [11] , [16] . Obviously, AAPs can profit from light utilization by conserving instead of oxidizing organic substrates or their storage compounds."
} | 1,986 |
26829286 | PMC4877668 | pmc | 4,413 | {
"abstract": "Corynebacterium\nglutamicum is an important organism for the\nindustrial production\nof amino acids. Metabolic pathways in this organism are usually engineered\nby conventional methods such as homologous recombination, which depends\non rare double-crossover events. To facilitate the mapping of gene\nexpression levels to metabolic outputs, we applied CRISPR interference\n(CRISPRi) technology using deactivated Cas9 (dCas9) to repress genes\nin C. glutamicum . We then determined the effects\nof target repression on amino acid titers. Single-guide RNAs directing\ndCas9 to specific targets reduced expression of pgi and pck up to 98%, and of pyk up\nto 97%, resulting in titer enhancement ratios of l -lysine\nand l -glutamate production comparable to levels achieved\nby gene deletion. This approach for C. glutamicum metabolic engineering, which only requires 3 days, indicates that\nCRISPRi can be used for quick and efficient metabolic pathway remodeling\nwithout the need for gene deletions or mutations and subsequent selection.",
"conclusion": "Conclusion Given its simplicity of design and ease\nof deployment, the CRISPR/Cas system has enabled researchers to edit\nDNA or gene transcription levels (sgRNA/dCas9 or CRISPRi) across an\nextremely diversified range of organisms. 51 − 53 The goal of\nour study was to test the performance and suitability of CRISPRi for\npathway engineering in C. glutamicum . Our approach,\ninvolving the use of sgRNAs 42 and the dCas9\nenzyme to repress the expression of targeted genes, 54 allowed us to go from the initial cloning in E.\ncoli to the final engineered C. glutamicum strains ready for testing in as little as 3 days. The use of dCas9\ninstead of Cas9 omits the need to select for rare Cas9-mutated strains 55 or single- or double-crossover mutants obtained via suicide vectors. C. glutamicum plays a multibillion dollar role in the production of two amino\nacids with the largest market sizes, l -lysine for animal\nfeed and l -glutamate for food additives. The genes chosen\nfor the present study, pgi , pck and pyk , indirectly impact the production of these two amino\nacids. For pck and pyk , we found\nthat the repression level caused by the steric blockage of the progression\nof the RNA polymerase to be equally efficient whether dCas9 annealed\nonto the T or NT strand of the coding region. For all three genes,\nthe amino acid production measured after glucose depletion in early\nstationary phase increased in a statistically significant manner when\ngene repression was mediated by any of the sgRNAs in the presence\nof dCas9 ( Figure 1 – 3 ). In most cases, targeting the NT strand with the\nCRISPRi system resulted in greater transcriptional repression and\namino acid production than targeting the T strand. This observation\nis consistent with earlier publications that reported that template-strand\ntargeting can be inefficient, 40 , 42 even though other groups\nhave shown that template-strand repression can be achieved. 41 , 44 Because of its programmability and ease of implementation,\nwe envision\nthat CRISPRi-based gene repression could be used to identify putative\ngenes that should be targeted for enhancing bioproduction. Unlike\nknockout-based approaches to metabolic engineering, the likelihood\nof completely inhibiting gene expression with CRISPRi is low. One\ncould take advantage of this property to map how intermediate levels\nof gene expression translate to metabolic production. Furthermore,\nmultiplexed gene perturbation via the expression\nof multiple sgRNAs from barcoded constructs ( e.g. , those built using CombiGEM assembly 56 ) coupled with high-throughput characterization via liquid chromatography coupled to mass spectrometry could be used\nfor combinatorial metabolic pathway engineering. However, in other\ncases, even a low expression level of certain metabolic genes could\npotentially contribute enough enzyme activity to maintain flux at\na near wild-type level and thus mask the effect of targeting such\ngenes. In the future, dynamic metabolic pathways could be engineered\nby coupling the expression of CRISPRi system to promoters that are\nresponsive to metabolite concentration. The ability to tune metabolic\npathways in response to intracellular or extracellular conditions\nmay help to maximize production titers for the molecules of interest. 57 , 58 Table 1 Strains and Plasmids\nUsed in This\nStudy strain properties source C. glutamicum ATCC\n13032 Wild type, biotin auxotroph ATCC C. glutamicum DM1729 (DSM17576) DM1729 is an aminoethylcysteine-resistant mutant of ATCC 13032;\npyc(P458S) hom(V59A) lysC(T311I)- l -Lysine overproducer Evonik Degussa GmbH E. coli DH5α Cloning\nstrain Lab stock plasmids properties source pDSW204 E.\ncoli IPTG inducible expression vector,\nAmpR ( 59 ) pZ8–1 E. coli – C. glutamicum P tac constitutive expression shuttle\nvector, KanR Evonik Degussa GmbH pAL374 E. coli –Corynebacterineae\nexpression\nshuttle vector, SpecR ( 60 ) pZ8-T_dCas9 pZ8–1 plasmid carrying dcas9 , driven\nby the IPTG-inducible P tac promoter, KanR This study pZ8-P_dCas9 pZ8–1\nplasmid carrying dcas9 driven\nby the propionate-inducible prpD2 promoter (P prpD2) , KanR This study pZ8-Ptac IPTG inducible version of pZ8–1 plasmid,\nto which lacI q was added,\nKanR This study pZ8-Prp pZ8–1 with the P prpD2 propionate-inducible\npromoter instead of P tac , KanR This study pPP208 Donor of dcas9 ( 54 ) pAL-pgi_T pAL374 plasmid carrying the pgi (T) sgRNA,\ntargeting the template strand of pgi , SpecR This study pAL-pgi_NT pAL374 plasmid carrying the pgi (NT) sgRNA,\ntargeting the nontemplate strand of pgi , SpecR This study pAL-pck_T pAL374 plasmid carrying the pck (T) sgRNA\ntargeting, the template strand of pck , SpecR This study pAL-pck_NT pAL374 plasmid carrying the pck (NT) sgRNA,\ntargeting the nontemplate strand of pck , SpecR This study pAL-pyk_T pAL374 plasmid carrying the pyk (T) sgRNA\ntargeting the template strand of pyk , SpecR This study pAL-pyk_NT pAL374 plasmid carrying the pyk (NT) sgRNA\ntargeting the nontemplate strand of pyk , SpecR This study pRIM2 E.coli – C. glutamicum shuttle vector,\nintegrative in C. glutamicum downstream of ppc , KanR ( 61 ) pRIM3 GentR version of pRIM2 This study pRIM3-rfp pRIM3 carrying rfp GentR This study pBTK30 Donor of GentR ( 62 ) pZ8–1_rfp Donor of rfp , KanR Sinskey Lab (MIT) pAL-rfp_NT pAL374 plasmid carrying the rfp (NT) sgRNA\ntargeting the nontemplate strand of rfp , SpecR This study pAL-rfp_T pAL374 plasmid carrying the rfp (T) sgRNA\ntargeting the template strand of rfp , SpecR This study Table 2 Ratio of Changes in Amino Acid Production,\nmRNA Levels and Enzymatic Activity in Cultures of C. glutamicum when Targeting Genes with CRISPRi a amino\nacid ratio (after glucose depletion in stationary phase) gene targeted (ID) strand\ntargeted CRISPRi gene deletion (reference values) mRNA ratio (exponential phase) amino acid ratio (exponential phase) enzyme activity\nratio (exponential phase) pgi T 1.05 1.7 45 1.12 1.01 1.01 NT 2.14 0.02 1.31 0.05 pck T 2.18 4.5 46 b 0.31 1.28 0.51 NT 2.24 0.02 1.46 0.17 pyk T 1.92 1.25 24 c 0.16 1.97 0.74 NT 3.25 0.03 3.04 0.48 a Results are expressed as the ratio\nbetween (dCas9+sgRNA strain)/(dCas9+no sgRNA strain). b In the study referenced, Tween 60\nwas used to induce glutamate production, whereas ethambutol was used\nin our paper. c In the study\nreferenced, biotin\nwas used to induce glutamate production, whereas ethambutol was used\nin our paper. T = template; NT =\nnontemplate. Table 3 Primers Used for the Quantification\nof the mRNA in the Samples by qRT-PCR template strand primer sequence\n(5′–3′) pgi F TCATTGGTTTCGCTCGTCCA R AGCGTTCTTACCGAAAGCCA pck F ATTGGCTACAACGCTGGTGA R CCACTTCAGAACGCGAGAGT pyk F GATACCGCAAAGCGTGTGG R GACAGGTGGACACAGGAAGG 16S rRNA F TTACCTGGGCTTGACATGGAC R GCTGGCAACATAAGACAAGGG",
"discussion": "Results\nand Discussion CRISPRi-based rfp Repression To analyze\nthe utility of CRISPRi for gene repression in C. glutamicum , we first integrated the gene coding for a red fluorescent protein\n( rfp ) into its chromosome. We chose to target both\nthe template (T) and nontemplate (NT) strands of rfp as there are contradictory reports regarding the efficiency of T-targeting\nat repressing transcription. Some authors have reported that T-targeting\nis inefficient, 40 , 42 whereas others have found T-targeting\nto lead to effective transcriptional repression. 41 , 44 Inducing the expression of a plasmid-borne dcas9 with sodium propionate (using the P prpD2 promoter) in\ncells expressing sgRNAs targeted at the template (T) or nontemplate\n(NT) strand of the rfp gene resulted in reduced RFP\nproduction ( Supplemental Figure S1 ). Surprisingly,\nadding sodium propionate to the strain expressing dcas9 alone ( i.e. , no sgRNA) resulted in higher levels\nof rfp expression. CRISPRi-Based Regulation\nof Amino Acid Production by C. glutamicum Having seen that a heterologous gene\n( rfp ) could be repressed by CRISPRi in C.\nglutamicum , we used this tool to target three endogenous\ngenes of commercial relevance: pgi , pck and pyk . The deletion of pgi leads\nto NADPH overproduction through the pentose-phosphate pathway and\nthis results in increased l -lysine titers. 23 , 45 The deletion of pck or pyk indirectly\nincreases l -glutamate production. 24 , 46 The disruption of pyk is thought to result in increased l -glutamate production through an enhanced anaplerotic flux. 24 , 47 Disrupting pck also results in an accumulation\nof l -glutamate, via an enhanced flux toward oxaloacetate\nin the TCA cycle, due to the disruption of backward flux from oxaloacetate\nto phosphoenolpyruvate. 46 To repress\nthese genes, we built several sgRNAs that together with dCas9 would\nsterically block transcription. In initial experiments, we found that\nproducing dCas9 from an unrepressed P tac promoter resulted\nin no clones after transformation (data not shown). Previous work\nshowed that other bacterial hosts in which dCas9 was expressed grew\npoorly or not at all. 44 , 48 To overcome this issue, we repressed\nsgRNA and dcas 9 expression from P tac via\nthe LacI transcription factor, which can be induced by the addition\nof IPTG. These constructs were located on independent replicative\nplasmids: sgRNAs from pAL374 ( Table 1 ), and dCas9 from plasmid pZ8–1 ( Table 1 ). The following experiments\nwere performed in CgXII minimal medium with 2% (w/v) glucose as carbon\nsource. In this section, sampling for amino acid quantification was\nperformed immediately after glucose depletion with cells in stationary\nphase (the standard for quantification of maximal amino acid production 49 , 50 ). When the CRISPRi system targeted the NT strand of pgi , the titer of l -lysine ( p =\n4.65 ×\n10 –13 ) increased by a factor of 2.1 over its sgRNA-less\ncounterpart ( Table 2 ; Figure 1 ), suggesting\nstrong repression of Pgi production. A milder effect (a 1.05-fold\nincrease over the control strain with dCas9 but without the sgRNA)\nwas observed when the sgRNA targeted the T strand of the same gene\n( p = 0.048). The increase in l -lysine production\nwas much stronger when targeting the NT versus the\nT strand ( p = 2.31 × 10 –12 ). In the absence of dCas9, expression of the sgRNAs alone did not\ninterfere with the production of l -lysine ( Figure 1 c). Figure 1 CRISPRi efficiently represses pgi transcription,\nincreasing l -lysine production. (a) Central metabolic pathway.\nThe notation in red represents the reduced transcription of pgi due to CRISPRi-mediated repression. (b) Gene and sequences\ntargeted by dCas9. (c) Amino acid titers (g/L) of the control and\ntest strains ( N = 9 per strain) were determined upon\nglucose depletion. The error bars represent the standard deviation\nof samples. * p ≤ 0.05 and *** p ≤ 0.001. For pck , targeting either the T or NT strands\nresulted in increased glutamate production by ∼2.2-fold compared\nwith a similar strain that only lacked the sgRNA ( Figure 2 ; p T = 0.02; p NT = 0.006). No significant\ndifference was observed between targeting the T or NT strand ( Figure 2 ; p = 0.90). For pyk , targeting the T strand increased l -glutamate production by ∼1.9 fold ( p = 0.03) and targeting the NT strand increased it 3.2-fold compared\nwith the control strain carrying the dCas9 but no sgRNA ( Figure 3 ; p = 0.002). Similar to pgi , there was a statistically\nsignificant difference between the fold-induction achieved by targeting\nthe T versus the NT strand of pyk , but the magnitude of the difference was smaller ( Figure 3 ; p = 0.05).\nIn the absence of dCas9, expression of the sgRNAs alone did not interfere\nwith the production of l -glutamate ( Figure 2 c and 3 c). Figure 2 CRISPRi efficiently represses pck transcription,\nincreasing l -glutamate production. (a) Central metabolic\npathway. The notation in red represents the reduced transcription\nof pck due to CRISPRi-mediated repression (b) Gene\nand sequences targeted by dCas9. (c) Amino acid titers (g/L) of the\ncontrol and test strains ( N = 9 per strain) were\ndetermined upon glucose depletion. The error bars represent the standard\ndeviation of samples. “NS” represents nonsignificant\n( p > 0.05), ** p ≤ 0.01\nand\n*** p ≤ 0.001. Figure 3 CRISPRi efficiently represses pyk transcription,\nincreasing l -glutamate production. (a) Central metabolic\npathway. The notation in red represents the reduced transcription\nof pgi due to CRISPRi-mediated repression. (b) Gene\nand sequences targeted by dCas9. (c) Amino acid titers (g/L) of the\ncontrol and test strains ( N = 9 per strain) were\ndetermined upon glucose depletion. The error bars represent the standard\ndeviation of samples. * p ≤ 0.05, ** p ≤ 0.01 and *** p ≤ 0.001. The increased l -lysine\nproduction we observed by repressing pgi with CRISPRi\nis similar to that observed when pgi is deleted ( Table 2 ). 45 The repression of pck and pyk expression via CRISPRi achieved l -glutamate\nproduction ratios (targeted/nontargeted)\nthat exceed published results for the deletion of those genes ( Table 2 ). 24 , 45 , 46 However, the absolute amounts of amino acids\nproduced using the CRISPRi system were not as high as when the genes\nwere deleted, 24 , 45 , 46 which could be due to the use of different strains or conditions,\nor because of residual enzyme expression due to incomplete repression\nby CRISPRi. Quantification of the mRNA Levels of Genes\nTargeted by CRISPRi We next evaluated the levels of transcription\nof each gene targeted\nby CRISPRi ( Figure 4 ). Here, we sampled C. glutamicum cultures in midexponential\nphase (maximal growth rate) to estimate mRNA levels and concomitant\namino acid production. We normalized mRNA levels of the targeted genes\nto those of 16S rRNA and calculated grams of amino acid produced per\ngram of cell dry weight, to account for the slight variations between\nthe ODs of the strains and independent experiments, at the time of\nsampling. Figure 4 mRNA levels decrease with CRISPRi targeting, resulting\nin increases\nin amino acid levels. In order to measure the mRNA levels of pgi , pck and pyk in the\ncells, samples were collected in midexponential phase. Upon total\nRNA extraction, qRT-PCR was performed on each of the samples (three\nbiological replicates per strain, each with three technical replicates)\nand using 16S rRNA for data normalization. Amino acid titers were\nmeasured by HPLC and are reported as grams of amino acid per cell\ndry weight. (a) sgRNA/dCas9-mediated pgi targeting,\n(b) sgRNA/dCas9-mediated pck targeting and (c) sgRNA/dCas9-mediated pyk targeting. “NS” represents nonsignificant\n( p > 0.05), * p ≤ 0.05,\n** p ≤ 0.01 and *** p ≤\n0.001. No CRISPRi-mediated repression\nof transcription was observed during\nthe exponential phase when the T strand of pgi was\ntargeted; however, targeting the NT strand strongly repressed mRNA\nlevels ( Figure 4 a, Table 2 ). The relative levels\nof mRNA were reduced by nearly 98% ( p = 1.02 ×\n10 –13 ), resulting in a 1.31-fold increase in l -lysine/gCDW, versus the control strain, which\nexpressed dCas9 but lacked the sgRNA ( p = 3.40 ×\n10 –8 ). When the T strand of pck was targeted, a 69% ( p = 2.04 × 10 –7 ) reduction in its mRNA levels and a 1.28-fold increase in the levels\nof secreted l -glutamate/gCDW ( p = 0.02)\nwere observed over the control levels of the strain lacking the sgRNA\nonly ( Figure 4 , Table 2 ). In comparison,\ntranscriptional repression was much higher when dCas9 was directed\nagainst the NT strand of pck (98% reduction in mRNA versus the sgRNA-less counterpart, p =\n1.13 × 10 –14 ); this repression appeared to\nresult in 1.46-fold more l -glutamate/gCDW compared to the\ncontrol strain. Despite the strong transcriptional repression ( Figure 4 b), the variability\nof the experimental data indicated that the increase in l -glutamate/gCDW in the exponential phase was nonsignificant at the p = 0.08 level, which contrasts with the data obtained in\nstationary phase shown in Figure 2 c. Finally, targeting the T and NT strands of pyk with CRISPRi resulted in mRNA reductions of 84% ( p T = 4.35 × 10 –5 ) and\n97% ( p NT = 0.0002) and 1.97 ( p T = 0.016) and 3.04 ( p NT =\n0.0005) fold increases in secreted l -glutamate/gCDW, respectively\n( Figure 4 , Table 2 ), versus the control strain. In general, greater mRNA repression correlated\nwith higher levels of secreted amino acids per gram of cell dry weight. Quantification of Enzymatic Activity Given that sgRNA/dCas9\ndid not fully repress transcription, we decided to investigate enzyme\nactivity levels in midexponential cultures of C. glutamicum in which CRISPRi was present or not. These samples were identical\nto those taken for mRNA quantification in Figure 4 . The specific enzymatic activities of Pgi,\nPck, and Pyk reflect the amount of active enzyme available in the\ncell at the time of sampling. Pgi activity levels decreased by 95%\nin comparison with a control with no sgRNA when the NT strand was\ntargeted ( p = 1.25 × 10 –38 , Figure 5 a, Table 2 ). Targeting the T\nstrand of pgi had no impact on the enzyme activity\nlevels observed ( p = 0.80, Figure 5 a, Table 2 ), versus a control with no sgRNA.\nTargeting the NT strand in the absence of dCas9 had no effect on Pgi\nactivity (0.49 U/mg; p = 0.20 versus strain with the same sgRNA and dCas9). Pck activity levels decreased\nby 49% ( p = 0.005) and 83% ( p =\n9.08 × 10 –7 ) when the T and NT strands were\ntargeted by specific sgRNAs, respectively, compared with a control\nwith no sgRNA ( Figure 5 b, Table 2 ). For Pyk,\ntargeting the T or NT strands resulted in a 26% ( p = 5.38 × 10 –11 ) and 52% ( p = 1.85 × 10 –24 ) decrease in specific activity,\nrespectively, compared with the no-sgRNA control ( Figure 5 c, Table 2 ). As expected, there was a correlation between\nreductions in mRNA levels ( Figure 4 ) and the corresponding enzyme activity ( Figure 5 ), with sgRNAs directed against\nthe NT strand leading to lower mRNA and enzyme activity levels compared\nwith those directed against the T strand. Figure 5 The specific Pgi, Pck\nand Pyk activities in crude extracts decrease\nwhen the corresponding genes are targeted by the CRISPRi system. Crude\nextracts from midexponential growing strains (9 biological replicates\nper strain, each with three technical replicates) were obtained upon\ncell lysis and used for the quantification of the Pgi, Pck or Pyk\nactivity, as seen by a decrease in the NADH available. The specific\nactivity (U/mg) of (a) Pgi, (b) Pck and (c) Pyk when CRISPRi targets\nthe template (T) or nontemplate (NT) DNA strands is shown. “NS”\nrepresents nonsignificant ( p > 0.05), *** p ≤ 0.001. Conclusion Given its simplicity of design and ease\nof deployment, the CRISPR/Cas system has enabled researchers to edit\nDNA or gene transcription levels (sgRNA/dCas9 or CRISPRi) across an\nextremely diversified range of organisms. 51 − 53 The goal of\nour study was to test the performance and suitability of CRISPRi for\npathway engineering in C. glutamicum . Our approach,\ninvolving the use of sgRNAs 42 and the dCas9\nenzyme to repress the expression of targeted genes, 54 allowed us to go from the initial cloning in E.\ncoli to the final engineered C. glutamicum strains ready for testing in as little as 3 days. The use of dCas9\ninstead of Cas9 omits the need to select for rare Cas9-mutated strains 55 or single- or double-crossover mutants obtained via suicide vectors. C. glutamicum plays a multibillion dollar role in the production of two amino\nacids with the largest market sizes, l -lysine for animal\nfeed and l -glutamate for food additives. The genes chosen\nfor the present study, pgi , pck and pyk , indirectly impact the production of these two amino\nacids. For pck and pyk , we found\nthat the repression level caused by the steric blockage of the progression\nof the RNA polymerase to be equally efficient whether dCas9 annealed\nonto the T or NT strand of the coding region. For all three genes,\nthe amino acid production measured after glucose depletion in early\nstationary phase increased in a statistically significant manner when\ngene repression was mediated by any of the sgRNAs in the presence\nof dCas9 ( Figure 1 – 3 ). In most cases, targeting the NT strand with the\nCRISPRi system resulted in greater transcriptional repression and\namino acid production than targeting the T strand. This observation\nis consistent with earlier publications that reported that template-strand\ntargeting can be inefficient, 40 , 42 even though other groups\nhave shown that template-strand repression can be achieved. 41 , 44 Because of its programmability and ease of implementation,\nwe envision\nthat CRISPRi-based gene repression could be used to identify putative\ngenes that should be targeted for enhancing bioproduction. Unlike\nknockout-based approaches to metabolic engineering, the likelihood\nof completely inhibiting gene expression with CRISPRi is low. One\ncould take advantage of this property to map how intermediate levels\nof gene expression translate to metabolic production. Furthermore,\nmultiplexed gene perturbation via the expression\nof multiple sgRNAs from barcoded constructs ( e.g. , those built using CombiGEM assembly 56 ) coupled with high-throughput characterization via liquid chromatography coupled to mass spectrometry could be used\nfor combinatorial metabolic pathway engineering. However, in other\ncases, even a low expression level of certain metabolic genes could\npotentially contribute enough enzyme activity to maintain flux at\na near wild-type level and thus mask the effect of targeting such\ngenes. In the future, dynamic metabolic pathways could be engineered\nby coupling the expression of CRISPRi system to promoters that are\nresponsive to metabolite concentration. The ability to tune metabolic\npathways in response to intracellular or extracellular conditions\nmay help to maximize production titers for the molecules of interest. 57 , 58 Table 1 Strains and Plasmids\nUsed in This\nStudy strain properties source C. glutamicum ATCC\n13032 Wild type, biotin auxotroph ATCC C. glutamicum DM1729 (DSM17576) DM1729 is an aminoethylcysteine-resistant mutant of ATCC 13032;\npyc(P458S) hom(V59A) lysC(T311I)- l -Lysine overproducer Evonik Degussa GmbH E. coli DH5α Cloning\nstrain Lab stock plasmids properties source pDSW204 E.\ncoli IPTG inducible expression vector,\nAmpR ( 59 ) pZ8–1 E. coli – C. glutamicum P tac constitutive expression shuttle\nvector, KanR Evonik Degussa GmbH pAL374 E. coli –Corynebacterineae\nexpression\nshuttle vector, SpecR ( 60 ) pZ8-T_dCas9 pZ8–1 plasmid carrying dcas9 , driven\nby the IPTG-inducible P tac promoter, KanR This study pZ8-P_dCas9 pZ8–1\nplasmid carrying dcas9 driven\nby the propionate-inducible prpD2 promoter (P prpD2) , KanR This study pZ8-Ptac IPTG inducible version of pZ8–1 plasmid,\nto which lacI q was added,\nKanR This study pZ8-Prp pZ8–1 with the P prpD2 propionate-inducible\npromoter instead of P tac , KanR This study pPP208 Donor of dcas9 ( 54 ) pAL-pgi_T pAL374 plasmid carrying the pgi (T) sgRNA,\ntargeting the template strand of pgi , SpecR This study pAL-pgi_NT pAL374 plasmid carrying the pgi (NT) sgRNA,\ntargeting the nontemplate strand of pgi , SpecR This study pAL-pck_T pAL374 plasmid carrying the pck (T) sgRNA\ntargeting, the template strand of pck , SpecR This study pAL-pck_NT pAL374 plasmid carrying the pck (NT) sgRNA,\ntargeting the nontemplate strand of pck , SpecR This study pAL-pyk_T pAL374 plasmid carrying the pyk (T) sgRNA\ntargeting the template strand of pyk , SpecR This study pAL-pyk_NT pAL374 plasmid carrying the pyk (NT) sgRNA\ntargeting the nontemplate strand of pyk , SpecR This study pRIM2 E.coli – C. glutamicum shuttle vector,\nintegrative in C. glutamicum downstream of ppc , KanR ( 61 ) pRIM3 GentR version of pRIM2 This study pRIM3-rfp pRIM3 carrying rfp GentR This study pBTK30 Donor of GentR ( 62 ) pZ8–1_rfp Donor of rfp , KanR Sinskey Lab (MIT) pAL-rfp_NT pAL374 plasmid carrying the rfp (NT) sgRNA\ntargeting the nontemplate strand of rfp , SpecR This study pAL-rfp_T pAL374 plasmid carrying the rfp (T) sgRNA\ntargeting the template strand of rfp , SpecR This study Table 2 Ratio of Changes in Amino Acid Production,\nmRNA Levels and Enzymatic Activity in Cultures of C. glutamicum when Targeting Genes with CRISPRi a amino\nacid ratio (after glucose depletion in stationary phase) gene targeted (ID) strand\ntargeted CRISPRi gene deletion (reference values) mRNA ratio (exponential phase) amino acid ratio (exponential phase) enzyme activity\nratio (exponential phase) pgi T 1.05 1.7 45 1.12 1.01 1.01 NT 2.14 0.02 1.31 0.05 pck T 2.18 4.5 46 b 0.31 1.28 0.51 NT 2.24 0.02 1.46 0.17 pyk T 1.92 1.25 24 c 0.16 1.97 0.74 NT 3.25 0.03 3.04 0.48 a Results are expressed as the ratio\nbetween (dCas9+sgRNA strain)/(dCas9+no sgRNA strain). b In the study referenced, Tween 60\nwas used to induce glutamate production, whereas ethambutol was used\nin our paper. c In the study\nreferenced, biotin\nwas used to induce glutamate production, whereas ethambutol was used\nin our paper. T = template; NT =\nnontemplate. Table 3 Primers Used for the Quantification\nof the mRNA in the Samples by qRT-PCR template strand primer sequence\n(5′–3′) pgi F TCATTGGTTTCGCTCGTCCA R AGCGTTCTTACCGAAAGCCA pck F ATTGGCTACAACGCTGGTGA R CCACTTCAGAACGCGAGAGT pyk F GATACCGCAAAGCGTGTGG R GACAGGTGGACACAGGAAGG 16S rRNA F TTACCTGGGCTTGACATGGAC R GCTGGCAACATAAGACAAGGG"
} | 6,657 |
35893564 | PMC9330434 | pmc | 4,414 | {
"abstract": "Antimony (Sb) and arsenic (As) are two hazardous metalloid elements, and the biogeochemical cycle of Sb and As can be better understood by studying plant rhizosphere microorganisms associated with Sb mine waste. In the current study, samples of three types of mine waste—Sb mine tailing, waste rocks, and smelting slag—and associated rhizosphere microorganisms of adapted plants were collected from Qinglong Sb mine, southwest China. 16S rRNA was sequenced and used to study the composition of the mine waste microbial community. The most abundant phylum in all samples was Proteobacteria , followed by Bacteroidota , Acidobacteriota , and Actinobacteriota . The community composition varied among different mine waste types. Gammaproteobacteria was the most abundant microorganism in tailings, Actinobacteria was mainly distributed in waste rock, and Saccharimonadia , Acidobacteriae , and Ktedonobacteria were mainly present in slag. At the family level, the vast majority of Hydrogenophilaceae were found in tailings, Ktedonobacteraceae , Chthoniobacteraceae , and Acidobacteriaceae (Subgroup 1) were mostly found in slag, and Pseudomonadaceae and Micrococcaceae were mainly found in waste rock. Actinobacteriota and Arthrobacter are important taxa for reducing heavy metal(loid) mobility, vegetation restoration, and self-sustaining ecosystem construction on antimony mine waste. The high concentrations of Sb and As reduce microbial diversity.",
"conclusion": "5. Conclusions In the current study, the Sb and As content in the three types of mine waste from the Qinglong Sb mine were very high. The investigation of plant rhizosphere microorganisms in the three types of mine waste indicated that there were significant differences in the microbial communities they contained. At the class level, the most abundant species in tailing is Gammaproteobacteria , Actinobacteria is mainly distributed in waste rocks, and the abundance of Saccharimonadia , Acidobacteriae , and Ktedonobcteria is the highest in slag. At the family level, Hydrogenophilaceae are enriched in tailings, and the abundance of Ktedonobacterae , Chthoniobacacterae , and Acidobacteriaceae (subgroup 1) in slag is significantly higher than that of the other two kinds of mine waste, while Pseudomonadaceae and Micrococcaceae were mainly found in waste rock. Since they are closely related to the cycling of nutrients such as nitrogen and phosphorus, Actinobacteriota and Arthrobacter are considered to be important groups in the oligotrophic mine waste environment. We suggest that Actinobacteriota and Arthrobacter can be used as important groups to reduce heavy metal(loid) mobility, revegetation, and create self-sustaining ecosystems in antimony mining areas. The microbial diversity in tailing was slightly higher than that in slag and waste rock is the highest. Increased Sb and As content reduced microbial diversity. Low pH reduces the diversity of rhizosphere microbial communities associated with slag plants, but it also promotes the development of some microorganisms in slag. The present study further revealed the composition and diversity of plant rhizosphere microbial communities in Sb mine waste. From the perspective of geochemistry, the regulatory mechanisms involved in microbial community compositions in Sb mine waste were further clarified.",
"introduction": "1. Introduction Antimony (Sb) is a potentially toxic and carcinogenic metalloid element. Humans and animals can be exposed to Sb in the environment via water, air, food, skin contact, and respiration. Long-term skin contact with dust containing Sb can lead to Sb spots, and inhalation of low concentrations of Sb dust or Sb-containing fumes can induce pneumoconiosis, lung cancer, and other diseases [ 1 , 2 , 3 ]. Sb and its compounds are also listed as priority pollutants by the United States Environmental Protection Agency and the European Union [ 4 , 5 , 6 ]. Mining activity is a major cause of the release of anthropogenic Sb into the environment [ 6 ]. Mining, flotation, and smelting are indispensable components of the process of mineral production, and they generate large amounts of mining waste such as mining waste rock, tailings, and smelting slag. Low metal recovery in some mines leads to mine waste that is potentially risky to humans and the environment due to its high heavy metal(loid) content [ 7 , 8 ]. In addition, many mine wastes are directly disposed of in the mine site without any treatment, which further increases the risk of heavy metal(loid) pollution [ 9 ]. China’s Sb ore reserves and production are the largest in the world, and Sb mining and smelting are the main source of Sb pollution, which may cause Sb levels in the atmosphere, water, and soil to exceed the standard [ 4 ]. The Sb content in the soil around the Xikuangshan Sb mine in Hunan ranged from 527 to 11,798 mg/kg [ 10 ]. Fu et al. [ 11 ] reported that the Sb content in water around Xikuangshan ranged from 5.6 to 163 μg/L (mean 24.7 μg/L), and in soils it ranged from 141 to 8733 mg/kg (mean 1315 mg/kg). High Sb levels were also detected in tailings (68.0–417,196 mg/kg, mean 3789 mg/kg), fish (1.0–1112 μg/kg, mean 86.8 μg/kg), surrounding plants (0.1–609 mg/kg, mean 13.5 mg/kg), and vegetables in water (0.1–10.7 mg/kg, mean 2.3 mg/kg). Liu et al. [ 12 ] reported that Sb levels in the hair of residents from Xikuangshan Sb mine (0.250–82.4 mg/kg, mean 15.9 mg/kg) and Qinglong Sb mine (0.060–45.9 mg/kg, mean value 5.15 mg/kg) were significantly higher than those of residents of from Guiyang City (0.065–2.87 mg/kg, mean 0.532 mg/kg). The establishment of vegetation caps on mine wastes to reduce the mobility of heavy metals is considered an effective method to mitigate mining waste contamination. Plant covering also increases organic matter content, cation exchange capacity, and nutrient levels, further improving the chemical and biological properties of contaminated soils, thus creating a self-sustaining ecosystem [ 13 , 14 ]. However, plant growth in mining waste sites is frequently constrained by harsh field conditions, which are characterized by low nutrient and organic material contents, high levels of heavy metals, and/or low pH. Rhizosphere microorganisms have a strong influence on plant growth and survival, and synergistic interactions between plants and rhizosphere microorganisms facilitate the removal of heavy metals from soil [ 15 ]. Rhizosphere microbes are also considered an important pathway for nutrient uptake by plants, directly affecting plant productivity and soil ecosystem function [ 16 ]. Plant rhizosphere bacteria can mitigate the adverse effects of nutrient deficiencies and metal(loid) pollution, promote plant growth in the soil [ 17 ], fix atmospheric nitrogen, and dissolve minerals such as phosphate, providing essential nutrients to plants [ 15 , 18 , 19 ]. Rhizosphere microorganisms can immobilize heavy metals at the rhizosphere level, and can also influence the uptake of heavy metals by plants by affecting their speciation in the soil [ 20 ]. Functional groups on microbial cell walls, including hydroxyl, carboxyl, sulfhydryl, and amino groups, can adsorb metal ions [ 21 ]. Microorganisms can mediate the transformation of Sb and arsenic (As) in soil–plant systems, thereby modifying the toxicity and mobility of As and Sb [ 22 ]. For example, Sb-oxidizing bacteria can rapidly oxidize Sb(III) to Sb(V) and reduce the transport capacity of Sb, thus reducing Sb uptake by plants [ 23 , 24 , 25 ]. Shewanella oneidensis can immobilize Sb through adsorption and complexation [ 26 ]. In addition to bacteria, as a key component of soil microbial communities, fungi can release nutrients from decomposed dead organisms, and drive material cycles. As the main decomposer and carbon sequester in soil, fungi can maintain soil fertility and health, and play an important ecological role in the ecosystem [ 27 , 28 , 29 ]. Jia et al. [ 30 ] found that arbuscular mycorrhizal fungi changed the oxidation of heavy metals and bacterial community structure in rhizosphere soil, reducing the bioavailability of heavy metals. The synergistic evolution of plants and rhizosphere microorganisms in harsh environments makes the use of rhizosphere microbial composition a prerequisite for the remediation of heavy metal-contaminated vegetation at mine waste sites [ 31 ]. Soil properties are an important determinant of plant rhizosphere microbial community composition [ 32 ]. Microbial communities and their metabolic activities are also affected by extreme geochemical conditions [ 33 ]. Environmental factors such as nutrient elements, pH, temperature, and especially heavy metal content can control the distribution and abundance of microorganisms. Mining wastes (including tailing, smelting slag, and mining waste rock) with high metal(loid) content have important inhibitory effects on microorganisms (including fungi and bacteria) [ 34 , 35 , 36 , 37 , 38 ], and some studies [ 39 ] have shown that antimony and arsenic have stronger inhibitory effects on soil bacteria than fungi. Therefore, this study focuses on rhizosphere bacteria. Microbial community composition under heavy metal contamination conditions has become one of the hotspots of current research, but very little research has been conducted on the rhizosphere microorganisms associated with Sb mine waste-adapted plants. Only a few studies have been conducted on Sb mine tailing microorganisms [ 31 , 40 , 41 ]. Even less research has investigated plant rhizosphere microorganisms adapted to smelting slag and mining waste rocks. It is critical to investigate the microbial composition of mine waste-adapted plants’ rhizospheres in order to understand the biogeochemical cycle process of heavy metal(loid)s, and to prevent and control heavy metal(loid)s. In the current study, we characterized the rhizosphere microbial communities of native adapted plants from three types of mining waste from the Qinglong Sb mine in southwestern Guizhou, China: tailing, smelting slag, and mining waste rocks. High-throughput sequencing was used to characterize the rhizosphere microbial communities. The objectives of the study were to (1) understand Sb and As contamination in the three types of mine waste and the compositions of plant rhizosphere microbial communities, (2) compare plant rhizosphere microbial community compositions and diversity in the three waste types, and (3) investigate the underlying mechanisms responsible for differences in microbial community diversity in the three waste types.",
"discussion": "4. Discussion The geochemical parameters show that, after beneficiation and smelting, the Qinglong Sb mine waste is at great risk of contamination, and from another point of view, the Qinglong Sb mine contains considerable potential resources. The Sb and As content in slag and tailing was extremely high. Very high Sb content may be one of the reasons why there were few types of plants growing in slag. The smelting process causes the loss of As and S (sulfur) in slag and relative enrichment of Fe. In general, high concentrations of heavy metal(loid)s are toxic to microorganisms and inhibit their growth, which in turn affects microbial communities and microbial diversity. In comparisons of the relative abundance of microbial communities and α-diversity in the three mine waste types, the microbial diversity in waste rock with relatively low Sb and As content was higher than that of the other two waste types, and it was hypothesized that the two metalloids had an inhibitory effect on the development of microbial communities. The main microorganisms in the mine waste were Proteobacteria , Bacteroidota , Acidobacteriota , and Actinobacteriota , which is consistent with previous studies [ 34 , 56 , 57 , 58 , 59 ]. Proteobacteria are the major Sb-resistant bacteria in the microbial community [ 60 , 61 , 62 ], and 99 of 125 culturable Sb(III)/Cu(II)-resistant bacteria from 11 different types of mining waste are Proteobacterial species, including α- Proteobacteria (mainly Brevundimonas ) and γ- Proteobacteria (mainly Pseudomonas ). Actinobacteria (mainly Arthrobacter ) and Firmicutes are also considered to be highly resistant to Sb(III) [ 63 ]. Pseudomonas reportedly has the ability to dissolve heavy metals and form complexes with them, and they can resist heavy metals through bioaccumulation, and fix them via metabolic processes [ 64 ]. In addition to resistance to Sb, Proteobacteria , Actinobacteriota , Firmicutes , Pseudomonadales , Comamonadaceae , Acinetobacter , Arthrobacter , Bacillus , and Hydrogenophaga are among the species that have been reported to function in the oxidation of Sb(III) or to be involved in the biotransformation of Sb [ 61 , 65 , 66 , 67 ]. All these communities were present in considerable abundance in the samples tested in the current study ( Table S3 ). Compared with previous studies [ 68 , 69 , 70 ], the abundance of Actinobacteriota and Arthrobacter in plant rhizosphere microorganisms in antimony mining areas is significantly higher than that in non-mining areas. In the study of Sun et al. [ 71 ], the members of Actinobacteriota are considered to play a key role in the microbial community of active antimony tailing. Members of Actinobacteriota are usually identified as key groups in extremely oligotrophic environments. Oligotrophic bacteria are important contributors to various ecological cycles in nature. They are closely related to the cycles of nitrogen, carbon, sulfur, phosphorus, and trace elements in nature, and play a very important role in the self-purification process of environmental systems [ 72 ]. Arthrobacter is considered to have an important role in the ecological restoration of Pb–Zn tailings [ 73 ]. The study shows that Arthrobacter has the functions of phosphorus dissolution and nitrogen fixation [ 74 ]. These functions are essential for the growth of plants on oligotrophic mine waste. In addition, as an acidophilic and metal tolerant bacterium, Arthrobacter is considered to be an ideal plant growth-promoting bacterium (PGPB) in acidic mine waste [ 75 ]. Therefore, we suggest that Actinobacteriota and Arthrobacter are important taxa for reducing the heavy metal(loid) mobility, vegetation restoration, and self-sustaining ecosystem construction on antimony mine waste. Redundancy analysis of microbial communities and geochemical factors indicated that Mg, Sb, and pH had the strongest effects on microbial community composition. pH has a strong influence on the structure and diversity of microbial communities. It can indirectly affect microbial community composition by changing the physical and chemical characteristics of the environment, for example by influencing the leaching of heavy metals [ 76 , 77 ]. In addition, pH directly affects the physiology and growth of microorganisms. When the pH value is neutral, microbial diversity is highest. When the pH value exceeds the survival range of one microorganism, most microorganisms will not survive [ 37 , 78 ]. The pH of tailing and waste rock was neutral to weakly alkaline, which is suitable for microbial development and promotes community diversity. Conversely, the acidic pH of slag reduces community diversity to an extent, and a similar situation has been found in other mining areas [ 79 , 80 ]. Some microorganisms were detected in the slag area, however, such as Acidobacteriae and Ktedonobacteria . In contrast, their abundance in tailing and waste rock was very low, indicating that these microorganisms can adapt to low-pH environments ( Figure 6 ). In another study, Acidobacteriae was significantly enriched in acid soil that was severely polluted with heavy metals [ 81 ]. Sb was the main polluting element in the mining area, and its influence of it on the microbial community cannot be ignored. Studies have shown that Sb(III) can reduce the abundance of specific bacteria and bacterial diversity at the phylum level [ 82 ]. Adding Sb to the soil can significantly reduce substrate-induced resuscitation [ 83 ]. Sb can influence microbial diversity by affecting biological community-level physiological profile, and soil dehydrogenase activity [ 84 ]. In an Sb-polluted area, microbial composition and diversity were negatively correlated with Sb content [ 56 ]. Similarly, in the current study microbial diversity in slag—which had the highest Sb content—was lower than that in the other two types of mine waste. In previous studies, the abundance of some microorganisms resistant to As and Sb or involved in the biogeochemical cycling of As and Sb, such as Actinobacteria , Firmicutes , Nitrospirae , Tenericutes , and Gemmatimonadetes , was positively correlated with As and Sb content [ 85 ]. These microorganisms are considered to have functional genes and proteins related to As or Sb metabolism, including arsC, arrA, arsM, aioA, ArsB, and ACR3 [ 86 ]. In the present study, Spearman’s correlational analysis was performed by analyzing the relative abundance of several bacteria resistant to As and Sb or involved in As and Sb biogeochemical cycling processes such as Proteobacteria , Actinobacteriota , and Pseudomonadales , and the main geochemical parameters ( Figure 9 ). As and Sb content were not significantly positively correlated with the abundance of these bacterial communities, and in fact the correlation was negative. Conversely, Mg was positively correlated with the abundance of several bacterial species, and it was highly correlated with the abundance of Actinobacteriota and Arthrobacter . Mg is an important component of some bacterial enzymes and plays an important role in stabilizing and regulating bacterial membrane structure and ribosomes [ 87 , 88 , 89 ]. It has also been shown that Mg has a crucial role in ATPase, and with increased Mg 2+ concentration ATPase activity is enhanced [ 90 ]. ArsB, the transporter protein encoding Sb and As, is dependent on ATPase catalysis [ 64 , 91 ]. It is therefore hypothesized that the Mg deficiency in tailing and slag in the Sb mining areas has inhibited bacterial transport of Sb and As in vitro, resulting in reduced resistance to As and Sb. High levels of As and Sb are in turn harmful to bacteria, resulting in a negative correlation between As and Sb levels and the abundance of the corresponding bacterial communities."
} | 4,607 |
35509934 | PMC9058850 | pmc | 4,416 | {
"abstract": "Graphical abstract",
"conclusion": "8 Conclusion This paper integrates equipment to test the electrical and mechanical properties. Simultaneously testing the mechanical and electrical properties should elucidate the essential characteristics of stretchable conductive materials. Although our equipment is designed specifically for soft materials, it could be used for hard materials such as PVC, carbon composites, or nylon by modifying the gripping part. Additionally, our equipment has the potential to measure the properties of touch and pressure sensors by modifying the gripping part. Changing the linear stage controller from HSC-103 to SHOT 702 should reduce the overall price since SHOT 702 is $1000 cheaper than HSC103. We expect that our equipment will contribute to the development of soft robotics, especially soft and stretchable sensors."
} | 207 |
36134244 | PMC9417313 | pmc | 4,418 | {
"abstract": "Structural color materials that are colloidally assembled as inspired by nature are attracting increased interest in a wide range of research fields. The assembly of colloidal particles provides a facile and cost-effective strategy for fabricating three-dimensional structural color materials. In this review, the generation mechanisms of structural colors from colloidally assembled photonic crystalline structures (PCSs) and photonic amorphous structures (PASs) are first presented, followed by the state-of-the-art and detailed technologies for their fabrication. The variable optical properties of PASs and PCSs are then discussed, focusing on their spatial long- and short-order structures and surface topography, followed by a detailed description of the modulation of structural color by refractive index and lattice distance. Finally, the current applications of structural color materials colloidally assembled in various fields including biomaterials, microfluidic chips, sensors, displays, and anticounterfeiting are reviewed, together with future applications and tasks to be accomplished.",
"introduction": "1. Introduction Natural colors, as widely displayed by animals, plants, and other organisms, are mainly generated by pigments, structural colors, fluorescence, or their combinations. 1–7 Unlike pigment and fluorescence, a structural color is generated from light reflection, diffuse reflection, diffraction, and interference with spatially ordered nano- or microstructures in organisms. 8 Over the past 515 million years, organisms in nature have been using these nano- or microstructures to produce structural colors to embellish their appearances. Modern studies involving structural colors date back to 1665, when Hooke first proposed that the color of silverfish came from microstructures. In 1730, Newton explained the principle of structural color generation in peacock feathers. A detailed explanation of the structural colors was completed by Anderson and Richards in 1942, which was accomplished by the application of nanoscopic methods such as electron microscopy. 9 Biophotonic crystals were first reported by Parker in 2001. 10 Since then, research into biostructural color nanomaterials has rapidly developed. These studies have shown that rich colors in bird feathers and butterfly wings were mainly caused by the interference of light. For example, the blue color of the head and neck skin of turkeys and the blue color of facial, hips, and reproductive area of primates are both derived from light scattering by a large number of fine particles in the corresponding epidermal tissues. In addition, organisms with a dynamic structural color were found to be able to control their colors and patterns to transmit information about sexual desire, warnings, and camouflage. 11 These changing colors can be directly applied to the construction of tunable optical devices. The modulation of color at the subcellular level also provides a good prototype for the development of tunable color display devices. In the recent decades, scientists have devoted considerable effort to mimic the fantastic photonic materials present in natural biostructures. However, these structures are usually very subtle and complex, which are difficult to mimic using the widely popular top-down technologies, such as lithography. Alternatively, the self-assembly of colloidal particles as a facile and cost-effective bottom-up strategy for the fabrication of photonic materials has been rapidly developing in the last two decades, which has led to a wider range of applications. In the following sections, a detailed introduction on the color mechanism and colloidal assembly approaches for photonic crystalline structures (PCSs) and photonic amorphous structures (PASs) has been presented. The angle independence and optical modulation of PCSs and PASs are separately discussed. The important applications of these structural color materials are presented thereafter, followed by some perspectives on their future development and challenges toward their practical applications."
} | 1,018 |
32060424 | PMC7021801 | pmc | 4,419 | {
"abstract": "The microbiota colonizing the root endophytic compartment and surrounding rhizosphere soils contribute to plant growth and health. However, the key members of plant soil and endophytic microbial communities involved in inhibiting or assisting pathogen invasion remain elusive. By utilizing 16S high-throughput sequencing and a molecular ecological network (MEN) approach, we systematically studied the interactions within bacterial communities in plant endophytic compartments (stem and root) and the surrounding soil (bulk and rhizosphere) during bacterial wilt invasion. The endophytic communities were found to be strongly influenced by pathogen invasion according to analysis of microbial diversity and community structure and composition. Endophytic communities of the infected plants were primarily derived from soil communities, as assessed by the SourceTracker program, but with rare migration from soil communities to endophytic communities observed in healthy plants. Soil and endophytic microbiomes from infected plants showed modular topology and greater complexity in network analysis, and a higher number of interactions than those in healthy plants. Furthermore, interactions among microbial members revealed that pathogenic Ralstonia members were positively correlated with several bacterial genera, including Delftia, Stenotrophomonas, Bacillus, Clostridium XlVa, Fontibacillus, Acidovorax, Herminiimonas , and three unclassified bacterial genera, in infected plant roots. Our findings indicated that the pathogen invasion in the rhizosphere and endophytic compartments may be highly associated with bacteria that are normally not detrimental, and sometimes even beneficial, to plants.",
"discussion": "Discussion Several members of the genus Ralstonia , especially R. solanacearum , are well-known and important phytopathogens due to their ability to cause wilt symptoms and economic losses in many cultivated members the Solanaceous family of plants. 9 , 10 Meanwhile, the diversity of resident microbes could also affect the antagonistic and/or facilitative interactions between plants and pathogens. 27 , 36 Our results showed that the community diversity of infected roots and stems were higher than in healthy samples according to both Chao1 and PD indices (Fig. 1 ), which could be explained by the fact that the plant’s defense system was disrupted after bacterial wilt invasion, allowing more organisms from soil microbial communities to enter the plant. This result was also consistent with previous research that endophytes are believed to play important roles in priming host defenses against pathogen invasion 37 and high diversity might increase community invasion resistance due to interactive effects on community stability. 38 Through the species classification, we found the OTUs assigned to potential pathogenic Ralstonia (we could not technically affirm all those OTUs assigned to Ralstonia are pathogenic by just using 16S sequences) were rarely observed in all healthy samples (bulk soils, rhizosphere, stems and roots), but showed fairly high abundances in the infected plant samples, consistent with field observations of plant wilt. Correspondingly, the relative abundances of some bacteria were clearly altered after bacterial wilt infection. There was a decline in the relative abundances of Arthrobacter, Massilia, Nocardioides, Sphingobium, Gaiella , and Conexibacter in infected bulk and rhizosphere soils (Fig. 2a ). These compositional changes could be a consequence of pathogen invasion. For example, Arthrobacter (49.7% and 19.1% reduction in infected bulk and rhizosphere soil, respectively) is known to have pathogen suppression potential for Fusarium wilt. 39 In the endophytic compartments, the relative abundances of Lactococcus, Pseudomonas, Bacillus, Falsibacillus , and Leuconostoc , often considered to be plant-beneficial microbes, 40 – 46 showed significant decrease compared to the healthy samples. These decreases suggest that the normal endophytic taxa were either actively excluded by the host immune system or outcompeted by more-successful colonizers. 47 , 48 The genera Lactococcus, Enterococcus , and Leuconostoc are recognized as lactic acid bacteria, 40 with the ability to act as plant growth promoting bacteria, inhibiting wilt of tomato caused by Ralstonia solanacearum . 49 Members of the genera Bacillus , Enterococcus , and Falsibacillus are widely recognized as the biocontrol agents with the ability to secrete antibiotics or other antimicrobial proteins, 41 – 46 and have been applied to prevent and control bacterial diseases of alfalfa, tobacco, and cucumber. 50 – 52 Meanwhile, the relative abundances of Stenotrophomonas, Paenibacillus, Achromobacter , Rhizobium, Clostridium, Delftia, Acidovorax , and Microbacterium in infected samples showed a marked increase compared to healthy samples, suggesting they may be involved in the process of pathogen invasion and have mutualistic relationships with pathogenic members of Ralstonia , 53 or perhaps they are opportunists, able to take advantage of potential ecological niches opened by pathogen invasion. 47 These changes in microbiome composition and structure indicate a change in the root exudates caused by pathogen invasion or a sophisticated plant immune system, 54 – 56 which drives either differential recruitment of beneficial microbes and/or differential exclusion to enable wilt resistance in plant roots and stems. 12 Soil microorganisms likely affect plant immune defense and pathogen migration, therefore understanding how plant endophytes interact with the soil microorganisms may provide a “road map” to define the pathogen invasion process. The SourceTracker program has been used to estimate the proportions of contaminants in a given community that come from potential source environments 19 and has been used to analyze the relationship between human-associated microbial communities and home surfaces. 57 In this study, we utilized this program to track the source of plant rhizospheric and endophytic microbial communities during the process of pathogenic wilt invasion. A previous study had shown that R. solancearum invaded plants via the roots, multiplied, and then aggressively colonize the xylem elements in the vascular system, blocking water transport such that infected plants wilt and die. 58 Our results showed different sources for microbiota in infected plants, in which the bacteria communities in the rhizosphere soils were mainly (71.4%) derived from the bulk soil. This is consistent with the previous studies that the bulk soil was the main source of microbial species richness in plant rhizosphere. 2 , 3 , 59 However, root endophytic communities were mainly derived from the bulk soils (50.1%) rather than from the rhizosphere soil (11.7%) (Fig. 3 ). It was concluded that pathogen invasion may begin in the bulk soils, transfer to the plant roots, and in turn infect plant stems. In the recent years, visualization of interactomes from diverse organisms has led to great progress in network biology. 60 , 61 While several studies have established a positive correlation between community diversity and invasion resistance, it is less clear how interactions between members within resident communities are involved in this process. 62 From the perspective of resource utilization and competition, plants and soil microbes can have direct co-evolutionary relationships, such as those between plants and pathogens. 63 , 64 It is becoming more evident that pathogenic and mutualistic–symbiotic organisms influence plant microbial community diversity and succession. 65 , 66 In this study, we performed network analyses on soil and endophytic bacterial community interactomes of infected and healthy plants, and revealed their topological features (Supplementary Table 2 , Fig. 4a–d ). The soil (Fig. 4b ) and endophytic microbiota (Fig. 4d ) of infected plants exhibited more complex, and highly connected bacterial communities than the respective communities of healthy plants, (Fig. 4a, c ). In this sense, by changing soil community structure, invasive pathogenic microbes could generate positive feedback that enhances both their own competitiveness and subsequent interactions with their neighbors. Crucially, highly connected and modular microbiota could prime the plant immune system for accelerated activation of defense against the pathogen. 54 , 55 , 67 In addition, we found that there were more nodes (9 nodes) of pathogenic Ralstonia members and a greater number of links (38 links) among Ralstonia and other microbial members in the network of infected endophytic compartments (Fig. 4d ) than in the infected soil network (1 node, 12 links) (Fig. 4b ). A greater number of Ralstonia nodes (8) and links (69) between Ralstonia and other microbial members were observed in the infected root network (Fig. 4f ) as compared to infected stems (4 nodes, 6 links) (Fig. 4e ). Based on these network topological data and source tracking analyses results (Fig. 3 ), we predicted that the endophytic microbiota played important role in the suppression of plant pathogens and that, from the perspective of microbial interactions and source tracking, plant roots were the critical migration site during the process of tobacco bacterial wilt disease. Network analysis also revealed the relationships between pathogen and other associated bacteria species (Fig. 5 ). The highly connected and anomalously correlated nodes are either targets or helpers of diverse pathogens. 68 Microbes that positively interact with the pathogenic Ralstonia members were the preferred helpers for pathogen attack in tobacco bacterial wilt disease. 68 We identified previously unknown bacteria ( Delftia, Stenotrophomonas, Bacillus, Clostridium XlVa, Fontibacillus, Acidovorax, Herminiimonas , and three unclassified bacterial genera (family: 2 Burkholderiaceae , Rhizobiales )) that may have a positive effect on wilt disease invasion, and were enriched in infected roots. For instance, species belonging to the Rhizobiales are intriguing and extensively researched for including both bacteria with the ability to fix nitrogen when in symbiosis with leguminous plants and pathogenic bacteria to plants, 69 could colonize both below- and above-ground tissues of tobacco using a dynamic invasion process that involves both epiphytic and endophytic life styles. 70 These non-detrimental microbial members could closely collaborate with pathogens in the endophytic root compartment. This is consistent with our source tracking analyses results that the root was the key compartment for microbial community assembly from soil into endophytic communities during tobacco bacterial wilt invasion. Taken together, we infer that infection by pathogenic Ralstonia members may be highly associated with positive interactions between them and non-detrimental bacteria including Delftia, Stenotrophomonas, Bacillus, Clostridium XlVa, Fontibacillus, Acidovorax, Herminiimonas , and three unclassified bacterial genera (family: 2 Burkholderiaceae , Rhizobiales ), and that these non-detrimental bacteria could obtain benefits from promoting pathogen, which might lead to the migration of many additional bacterial genera into plant root and stems from bulk soils, eventually causing an outbreak of tobacco bacterial wilt disease. This discovery will provide potential ideas and a theoretical basis for controlling tobacco bacterial wilt disease. Further work is needed to confirm these findings."
} | 2,901 |
39100233 | PMC11291943 | pmc | 4,422 | {
"abstract": "Abstract Microbes are key drivers of global biogeochemical cycles, and their functional roles arey dependent on temperature. Large population sizes and rapid turnover rates mean that the predominant response of microbes to environmental warming is likely to be evolutionary, yet our understanding of evolutionary responses to temperature change in microbial systems is rudimentary. Natural microbial communities are diverse assemblages of interacting taxa. However, most studies investigating the evolutionary response of bacteria to temperature change are focused on monocultures. Here, we utilize high-throughput experimental evolution of bacteria in both monoculture and community contexts along a thermal gradient to determine how interspecific interactions influence the thermal adaptation of community members. We found that community-evolved isolates tended toward higher maximum growth rates across the temperature gradient compared to their monoculture-evolved counterparts. We also saw little evidence of systematic evolutionary change in the shapes of bacterial thermal tolerance curves along the thermal gradient. However, the effect of community background and selection temperature on the evolution of thermal tolerance curves was variable and highly taxon-specific,with some taxa exhibiting pronounced changes in thermal tolerance while others were less impacted. We also found that temperature acted as a strong environmental filter, resulting in the local extinction of taxa along the thermal gradient, implying that temperature-driven ecological change was a key factor shaping the community background upon which evolutionary selection can operate. These findings offer novel insight into how community background impacts thermal adaptation.",
"conclusion": "Concluding remarks Here, we provide novel evidence that the community background fundamentally alters the evolution of thermal performance and that this biotic effect has a more pronounced impact on thermal performance phenotypes than the evolution temperature (i.e., the abiotic driver). Furthermore, we show that the effect of the community background is highly idiosyncratic and taxon-specific. Therefore, attempts to predict how microbial communities will respond to global warming will require consideration of the specific community’s composition, how those community members interact, and how the interactions might respond to shifting thermal conditions. Although, broadly, we see an increase in growth rate across the TPC for community-evolved isolates, by design our study only allows us to observe the evolutionary trajectories of the lineages that persisted to the end of the experiment in the communities. It is likely that these strains persevered precisely because they were able to evolve higher growth rates, increasing their ability to compete with heterospecifics for space and resources. The finding that extinctions only occurred in communities and not in monocultures shows that even in our relatively simple artificial five taxon communities, there is a strong impact of the biotic environment on whether a taxon can persist in the face of shifts in temperature. It is intuitive that in natural communities with very high standing taxonomic richness and continuous exposure to dispersing novel strains, this biotic mediation of the thermal environment’s impact will only be more pronounced. As microbial systems respond to climate change through the evolution of individual community members and the shifting composition of communities, their contributions to biogeochemical cycling will be altered, potentially considerably. In this study, we provide novel evidence for the extent to which the biotic environment uniquely impacts the evolutionary trajectories and survival of community members undergoing warming. In light of these findings, we argue that the inferences gleaned from monoculture evolution experiments may be insufficient to understand how complex microbial ecosystems will respond to climate change.",
"introduction": "Introduction Microbial communities make key contributions to the functioning of ecosystems ( Falkowski et al., 2008 ) and play critical roles in global elemental cycles ( Davidson & Janssens, 2006 ; Giorgio & Duarte, 2002 ; Goericke & Welschmeyer, 1993 ; Philippot et al., 2013 ; Williams, 1981 ). Temperature exerts strong control over subcellular metabolic processes, which in turn determine the contribution of microbial communities to ecosystem functioning and biogeochemical cycling ( Duffy et al., 2021 ; Mahecha et al., 2010 ; Qi et al., 2002 ; Yvon-Durocher et al., 2010 , 2012 , 2014 ). Therefore, there is an urgent need to understand the mechanisms that shape how microbial communities will respond to rising global temperatures. Over short time scales, microorganisms exhibit acute responses to changes in temperature ( Antoniou et al., 1990 ; Bakermans & Nealson, 2004 ; Margesin, 2009 ; Pietikäinen et al., 2005 ). However, given their short generation time, natural microbial communities are predicted to respond to long-term global warming via evolutionary adaptation of individual community members ( O’Donnell et al., 2018 ; Padfield et al., 2016 ; Tian et al., 2022 ), ecological sorting of the standing community diversity ( Garcia et al., 2022 ), and invasion by taxa that are better preadapted to the novel thermal environment ( Vezzulli et al., 2012 , 2013 ). In monoculture thermal adaptation experiments, it has been shown that selection favors a shift in the thermal dependence of key metabolic traits, such as downregulation of respiration and photosynthesis in high-temperature evolved phytoplankton ( Padfield et al., 2016 ), a higher thermal optimum in high temperature evolved diatoms ( O’Donnell et al., 2018 ), narrower thermal performance curves (TPCs) in Escherichia coli evolved at a constant temperature ( Cooper et al., 2001 ), a higher thermal minimum in high temperature evolved bacteria ( Tian et al., 2022 ), and improved tolerance of temperature fluctuations in bacteria evolved in a fluctuating thermal environment ( Saarinen et al., 2018 ). However, wild microbes inhabit highly speciose communities ( Curtis et al., 2002 ), and owing to a notable lack of studies that utilize experimental evolution of thermal tolerance in a community context, it is unclear how the community background might impact the evolutionary trajectories of microbial thermal adaptation. Several studies have provided evidence that interspecific competition in microbial communities can impact environmental adaptation. Theoretical models have predicted that interspecific competition can constrain adaptive evolution in the face of environmental change via reduced population sizes and reduced selective pressure on tracking the optimum resource niche ( Johansson, 2008 ). Experimental studies have shown that community members have reduced rates of adaptation to novel CO 2 concentrations ( Collins, 2011 ), low pH ( Scheuerl et al., 2020 ), new food sources ( Lawrence et al., 2012 ), and even typical laboratory conditions ( Castledine et al., 2020 ), compared with their monoculture evolved counterparts. The existence of trade-offs between adaptation to abiotic conditions and community background is an intuitive and frequently hypothesized explanation for these findings. If such trade-offs exist between adaptation to the thermal environment and heterospecific imposed selection pressures, a microbial community member may experience more constrained adaptation to temperature than they would in a monoculture. Diverse communities can impact their members’ evolutionary response to increased temperature via mechanisms beyond competition-imposed selective trade-offs. Heterospecifics can be sources of beneficial mobile genetic elements. For example, the transmissible locus of stress tolerance (tLST) is a highly conserved and widely transmitted genomic island that codes for a number of gene products, including small heat shock proteins ( Kamal et al., 2021 ). Natural microbial communities can also contain nonprokaryotic organisms, such as bacteriophages (phages) that parasitize bacteria and archaea. As it has been shown that differences in the TPCs of phages and their hosts lead to acute shifts in the latter’s TPC ( Padfield et al., 2020 ), phage presence could also modulate thermal adaptation over longer timescales. Protist bacterivores are also widespread and functionally important members of microbial communities, and recent research has shown that the presence of predators interacts with temperature to regulate microbial community respiration ( Rocca et al., 2022 ). Despite the urgent need to understand how microbial ecosystems will adapt in response to warming, there remains considerable ambiguity surrounding how community context influences the evolution of thermal performance. Here, we address this gap by evolving five taxa across a range of temperatures in both monocultures and communities. Using a high-throughput growth assay, we quantified the TPCs for each evolved lineage, allowing us to determine how the community background shapes the evolution of thermal performance.",
"discussion": "Results and discussion All taxa analysis To investigate the influence of the community background on the evolution of thermal performance, we designed an experiment whereby five bacterial taxa were evolved in two treatment groups, “monoculture” and “community,” for ~100 generations. In the monoculture treatment group, each taxon was evolved without the presence of heterospecifics at eight different “evolution temperatures” (ranging from 15 °C to 42 °C) for ~100 generations. In the community treatment group, all five taxa were combined into communities and evolved at the same eight evolution temperatures as the monoculture treatment group. At the end of the experiment, each evolved lineage from every experimental unit was isolated, and their TPC was quantified by assaying the maximum growth rate across the thermal gradient (i.e., taxon A, evolved at 15 °C in the monoculture treatment group, would be grown at all eight temperatures ranging from 15 °C to 42 °C). We used a generalized additive mixed-effects model (GAMM) approach to analyze the TPCs of the isolates. In this experiment, our principal aim was to quantify whether evolving in a community context or in monoculture affects the evolution of thermal performance. Our target with the analysis of the TPCs was, therefore, to determine whether the shape—i.e., the nature of the relationship between growth rate and temperature—and the elevation—i.e., the average growth rate across the TPC—differed between monoculture and community evolved isolates. A GAMM allowed us to achieve these aims within a highly flexible statistical modeling framework. First, the GAMM captures the nonlinear shape of the TPCs using a smoothing function ( Venables & Ripley, 2002 ). Importantly, unlike mechanistic models of TPCs ( Schoolfield et al., 1981 ), the GAMM makes no assumptions about the shape of TPCs and, therefore, affords additional flexibility to capture differences among the taxa and treatment groups. Second, the GAMM facilitates modeling the hierarchical structure of the data—i.e., taxon-level TPCs nested with the broader aggregate response at the treatment group level. We found that there was a significant difference in the intercepts of the TPCs between isolates that evolved in a community context and isolates that evolved in monoculture ( Figure 1 ; Table 1 ). Bacteria isolated from community settings had higher maximum growth rates across the TPC (estimated marginal means [centered on mean of growth temperature] of growth rate for community evolved is 0.421 ± SE 0.011, and for monoculture evolved is 0.365 ± SE 0.011). Allowing the smoothing term on temperature to vary between monoculture and community-evolved isolates did not significantly improve the fit of the model ( Table 1 ), implying that there was no difference in the shape of the TPC between these groups. There was also no significant effect of evolution temperature or significant interaction between community context and evolution temperature on the intercept (see Table 1 “All taxa” for ΔAICc values for removal of each fixed effect and Table 2 for Tukey’s adjusted p -values and effect sizes). Table 1. The significance of fixed effects on maximum growth rate [ r (h −1 )] for the five taxa modeled separately as well as the model containing all taxa combined. \n Pseudomonas spp. \n Serratia spp. \n Aeromonas spp. \n Herbaspirillum spp. \n Janthinobacterium spp. All taxa Treatment intercept 20.0739 22.0284 41.5687 7.696 NS 14.501 Evotemp intercept 12.5241 8.2962 NS 7.62 NS NS Smooth varies by treatment NS 14.7586 9.9028 10.954 33.646 NS Treatment: Evotemp NS NS NS 8.728 NS NS All models are generalized additive mixed-effects models (GAMMs). Values are ΔAICc for the removal of the respective fixed effect from the lowest AICc model. “NS” is “nonsignificant” and denotes that the respective fixed effect was not present in the best model for that taxon. “Evotemp” refers to the evolution temperature an isolate was evolved at, “Treatment” denotes whether an isolate is community evolved, monoculture evolved, or ancestral (for the “All taxa” analysis, Treatment refers to only community evolved, and monoculture evolved). “intercept” denotes an intercept effect while “Smooth vary by Treatment,” denotes whether the smoothing term was allowed to vary between treatments groups. Table 2. Tukey’s adjusted p -values ( p ) and pairwise differences (effect size) for pairwise comparisons of all treatment-level intercepts. \n Pseudomonas spp. \n Serratia spp. \n Aeromonas spp. \n Herbaspirillum spp. \n Janthinobacterium spp. All taxa \n p \n Effect size \n p \n Effect size \n P \n Effect size \n p \n Effect size \n p \n Effect size \n p \n Effect size Ancestor—Monoculture \n <0.001 \n −0.0678 \n 0.019 \n −0.0907 0.911 0.0222 \n <0.001 \n −0.0822 NA NA NA NA Ancestor—Community \n <0.001 \n −0.1436 \n <0.001 \n −0.1262 \n <0.001 \n −0.2308 \n <0.001 \n −0.0783 NA NA NA NA Monoculture—Community \n <0.001 \n −0.0758 0.379 −0.0355 \n <0.001 \n −0.253 0.931 0.00387 NA NA \n <0.001 \n −0.056 Bold values denote statistically significant pairwise differences. Figure 1. The relationship between maximum growth rate [ r (h −1 )] and growth temperature (°C), including all five taxa evolved in monocultures (orange), communities (purple), and the ancestors to the evolved isolates (green). For evolved isolates, all evolution temperatures are combined. Lines denote the model-fitted values, while ribbons denote the standard errors. Evolved isolate and ancestor curves are produced from separate models. The relationship between maximum growth rate [r(h−1)] and growth temperature (°C), including all five taxa evolved in monocultures, communities, and the ancestors to the evolved isolates. For evolved isolates, all evolution temperatures are combined. Lines denote the model-fitted values, while ribbons denote the standard errors. Evolved isolate and ancestor curves are produced from separate models. We additionally modeled the relationship between temperature and growth rate for the ancestors, allowing us to make comparisons with the evolved isolates. We found that the ancestor had a similar estimated marginal mean growth rate to the monoculture-evolved isolates (0.364 ± SE 0.022). This suggests that the growth rate of the monoculture-evolved isolates changed relatively little compared to the ancestor, while by contrast, the community-evolved isolates evolved a higher growth rate (see Statistical analysis section for why the ancestor was not included as an additional level of treatment group). We hypothesize that higher growth rates in the community-evolved treatment group are driven by interspecific competition, which selects for more rapid resource acquisition and, thus, a higher maximum growth rate. Although a key prediction of ecological theory is that the strength of intraspecific competition will exceed that of interspecific competition ( Adler et al., 2018 ; Goldberg & Barton, 1992 ; Macarthur & Levins, 1967 ), the community-evolved isolates will have been subjected to both intra- and interspecific competition, while monoculture evolved isolates face only intraspecific competition. The additional effect of interspecific competition could have driven the evolution of a higher maximum growth rate in the community setting compared to that in monoculture. This result is consistent with previous research demonstrating the competitive advantage provided by a higher growth rate ( Ketola et al., 2016 ; Winkler et al., 2017 ). However, the evolution of a higher growth rate in a community context is not ubiquitous, with another study finding the opposite effect; community-evolved isolates had reduced maximum growth rates due to cross-feeding specialization (when one taxon feeds on waste products of another taxon) occurring in the community evolved isolates, a strategy that could not be utilized when growth assayed in monoculture ( Lawrence et al., 2012 ). It is important to note that although we observe higher growth rates in the community evolved isolates when they are assayed in monoculture, it is possible that their actual growth rates in community settings would be reduced by antagonistic interactions with other taxa, as observed in O’Donnell et al. (2018) . Within-taxon analyses As well as investigating the broad-scale effect of the community environment on TPC evolution, we also conducted within-taxon analyses ( Figure 2 ). Here, we used separate GAMMs for each taxon to model the maximum growth rate as a function of treatment group (ancestral, monoculture evolved, community evolved), evolution temperature, and growth temperature to explicitly assess taxon-level variability in TPC evolution. Figure 2. The relationships between maximum growth rate [ r (h −1 )] and growth temperature (°C) for isolates evolved in monocultures (orange), communities (purple), and the ancestors to the evolved isolates (green). Lines denote the model-fitted values, while ribbons denote the standard errors. Each row of plots has different evolution temperatures, and each column (A–E) is a different taxon. Empty plots denote that the respective taxon did not survive to the end of any of the community replicates for the respective evolution temperature. Pairs of evolution temperatures that differ significantly in their intercepts are denoted with X, Y, or Z to denote the pair, and *** p < 0.001, ** p < 0.01, and * p < 0.05. Note that the ancestors’ data are only included across all evolution temperatures to facilitate visual comparison to evolved isolates. The relationships between maximum growth rate [r(h−1)] and growth temperature (°C) for isolates evolved in monocultures, communities, and the ancestors to the evolved isolates. Lines denote the model-fitted values, while ribbons denote the standard errors. Each row of plots has different evolution temperatures, and each column (A–E) is a different taxon. Empty plots denote that the respective taxon did not survive to the end of any of the community replicates for the respective evolution temperature. Pairs of evolution temperatures that differ significantly in their intercepts are denoted with X, Y, or Z to denote the pair, and ***p < 0.001, **p < 0.01, and *p < 0.05. Note that the ancestors’ data are only included across all evolution temperatures to facilitate visual comparison to evolved isolates. Effects of community context Treatment group (ancestral, monoculture evolved, or community evolved) was present in the best model for all taxa as either an intercept effect or as a term allowing the smoothing effect on growth temperature to vary between treatment groups—meaning that the shape of the TPC differed between ancestral, monoculture, and community evolved isolates. For three out of five taxa, both of these terms were present in the best model ( Table 1 ). We conducted post-hoc significance testing of the pairwise differences between the intercepts of ancestral, monoculture evolved, and community evolved isolates (see Table 2 ). For all four taxa with an intercept effect of treatment in the best model ( Pseudomonas spp., Serratia spp., Aeromonas spp., and Herbaspirillum spp.) the community evolved isolates had a significantly higher growth rate than the ancestor, and for Pseudomonas spp. and Aeromonas spp. the community evolved isolates also had significantly higher growth rates than the respective monoculture-evolved isolates. For Serratia spp. and Herbaspirillum spp., monoculture-evolved isolates also had significantly higher growth rates than the ancestor but were not significantly different from their community-evolved counterparts. These findings suggest that for Pseudomonas spp. and Aeromonas spp. the community environment selects for higher growth than the monoculture environment, yet for Serratia spp. and Herbaspirillum spp., it does not. A fixed effect that allows the smoothing term on growth temperature to vary by treatment group was present in the best model for Serratia spp., Aeromonas spp., Herbaspirillum spp., and Janthinobacterium spp. This suggests that in addition to contributing to a blanket change in growth rate across the TPC, community context can also alter the shape of the TPC. These community context-mediated changes in TPC shape are variable and taxon-specific. For Serratia spp. and Janthinobacterium spp., there is a shift toward peak growth occurring at higher temperatures for community-evolved isolates compared to monoculture-evolved isolates. For Aeromonas spp., we see the difference in growth rates between monoculture and community-evolved isolates increases around the temperature at which peak growth occurs and decreases at more extreme temperatures. For Herbaspirillum spp., the effect is subtle, but community-evolved isolates have a seemingly flatter TPC than monoculture-evolved isolates. We propose the reason that monoculture and community TPCs did not differ in shape in the “All taxa” analysis is related to this variability; when summed, these idiosyncratic and taxon-specific effects pull the aggregate, treatment-level effects in opposite directions, meaning the treatment effects on TPC shape are only revealed by modeling each taxon separately. The taxon-specific effects of the biotic environment on evolutionary trajectories we observed in our experiments are consistent with previous observations made by Lawrence et al. (2012) in their work investigating resource use evolution ( Lawrence et al., 2012 ). When we consider the complexity of selective pressures imposed by the community background, the diversity in responses from different taxa is unsurprising. In the community treatment group, each taxon will be exposed to a unique set of pairwise interactions with other community members. These interactions will also vary across the time course of the evolution experiment as each community member evolves and relative abundancies shift. Moreover, consideration of pairwise interactions alone is insufficient for predicting evolutionary trajectories in community contexts ( Terhorst et al., 2018 ). Higher-order interactions and indirect effects are also known to be key factors shaping community structure, whereby pairwise interactions between two taxa are altered by the presence of additional taxa ( Strauss, 1991 ). Furthermore, our experiment involved evolution across a thermal gradient, which adds yet another level of complexity because the taxonomic composition, interaction structure, and strength of interactions will vary along the temperature gradient; the community members will also differ in their levels of preadaptation to different temperatures. This multitude of factors means that the complex selective landscape will also depend on the temperature at which the community evolved. The heterogeneity in responses we observe when taxa are evolved within a relatively simple community context highlights the difficulty in predicting evolution in speciose natural microbial communities and the inadequacy of using insights gleaned from monoculture experiments to do so. Effects of evolution temperature Evolution temperature was observed in the best model for three taxa ( Pseudomonas spp., Serratia spp., and Herbaspirillum spp.); however, post-hoc analysis reveals few significant pairwise differences ( Figure 2 ). For Pseudomonas spp., the inclusion of evolution temperature appears to be primarily driven by isolates evolved at 23 °C having higher growth than those evolved at 20 °C, 27 °C, and 30 °C ( Figure 2A ). Similarly, for Serratia spp., those that evolved at 23 °C had significantly higher growth rates than those that evolved at 27 °C and 30 °C ( Figure 2B ). For Herbaspirillum spp. there was a significant interaction term between the evolution temperature and treatment group, suggesting that the effect of the treatment group varies between evolution temperatures. However, this effect is subtle, with the greatest effect being the difference between community-evolved isolates and ancestral isolates at 27 °C and 37 °C ( Figure 2C ). It is notable that the differences in the TPC for Herbaspirillum spp. across evolution temperatures and between community evolved, monoculture evolved, and ancestral isolates are less pronounced than for other taxa, yet less variation in the growth rate of this taxon was observed in general. The lower error in the model fit likely allows for the detection of effects with smaller sizes in this case. The relatively limited effect of evolution temperature on the relationship between growth temperature and growth rate, compared with the varied and pronounced effects of the biotic environment, suggests that evolution temperature is less impactful on the evolution of thermal performance than the community context. This is somewhat surprising given how previous experiments utilizing monocultures have shown marked differences in the TPCs of higher and lower temperature evolved diatoms ( O’Donnell et al., 2018 ), phytoplankton ( Padfield et al., 2016 ), and bacteria ( Tian et al., 2022 ). However, evolution temperature was observed to be related to the local extinction of specific taxa at various temperatures in the communities ( Figure 3 ). The effect of evolution temperature on survival in communities was statistically significant (ΔAICc = 66.356), as was taxon identity (ΔAICc = 13.694). As we observed no extinctions in any monoculture replicates, we hypothesize that these extinctions in the communities are due to competitive exclusion. This temperature-mediated ecological sorting is consistent with previous studies demonstrating that in synthetic laboratory communities, warming alters community structure ( Garcia et al., 2022 ). This is also consistent with field studies showing that experimentally warmed and ambient mesocosms have differing phytoplankton ( Yvon-Durocher et al., 2015 ) and bacterial assemblages ( Aydogan et al., 2018 ). Our results, however, show that temperature-driven extinctions in monoculture are not a prerequisite for local extinctions in a community context. This suggests that in communities, it is the interplay between ecological dynamics imposed by heterospecifics and selection pressure imposed by the thermal environment that regulates whether specific taxa can persist. Furthermore, these findings add significant weight to the hypothesis that temperature change radically alters the ecological context upon which abiotic selection operates by changing the structure of communities. Figure 3. The count of community replicates where each taxon survived to the end of the community evolution experiment at each evolution temperature (°C) out of a total of three."
} | 6,964 |
35446625 | PMC9169928 | pmc | 4,423 | {
"abstract": "Significance The population ecology of microbial communities is still poorly understood and their notorious instability makes them impossible to control. Much of the instability is caused by the stochastic assembly of microorganisms, especially in highly diverse microbiomes where structural and hence functional changes occur rapidly due to the short generation time of their members. Usually, to maintain organismic proportions in communities, their niches are deterministically reinforced, but stochasticity strongly counteracts this. Based on metacommunity theory, a looped mass transfer was developed that uses the rescue effect to stabilize communities. This study fills a long-standing gap and enables continuous and proportionally equal growth of community members using an unprecedented operational design that addresses an acute need in the healthcare and biotechnology industries.",
"discussion": "Discussion Stochasticity and dispersal are essential processes in shaping the structure and function of (meta)communities in natural environments, thereby influencing the ecological capacity and impact of microbial communities [e.g., Zhou et al. ( 42 ) and Wu et al. ( 43 )]. Both random drift events, which influence the birth and death of organisms, and the extent to which organisms disperse are key factors in how communities are affected at local and temporal scales. According to Hubbell’s ( 44 ) neutral theory, many natural patterns of community assembly can only be explained by these neutral events, in which niche-based selection is accompanied by random immigration and emigration, resulting in ecological drifts and a metacommunity formed by dispersal ( 45 ). Through experimentation and analogy, ecological drifts have been shown to be greater when abiotic niches are shallow, competition is weak, and dispersal is low ( 46 – 48 ). The absence of dispersal results in communities that are isolated from each other, making them susceptible to stochastic variation ( 9 , 10 , 13 ). Low dispersal is common in macroecology and contributes to the presence and coexistence of multiple species within a metacommunity of multiple patches ( 49 , 50 ). In contrast, high dispersal is known to increase similarity between localities and thus the risk of global extinction ( 51 – 53 ). High dispersal has a quantitative effect on local community formation in accordance to the mass-effect paradigm, in which community composition in sink habitats tends to resemble that of the source habitat ( 44 , 54 ). Frequently, the abundance of species in source communities and their transfer rates influence the composition of sink habitats ( 55 ), and lost individuals are replaced by members of source communities ( 56 ). This underscores the assumption that a regional pool that floods local communities with high rates of mass transfer can lead to synchrony and stability. Usually, mass transfer occurs between ecologically distinct niches in natural and engineered systems. Even when mass transfer is high, different environments have specific community structures, e.g. discrete bacterioplankton communities in a lake and its inlets ( 32 ) or microbial communities in connected transects of wastewater treatment plants ( 26 , 37 ). In the metacommunity of this study, identical localities operated in the same way were also expected to develop disparity ( 9 ), but these were overcome by the loop design. The looped mass transfer between the five local reactors L1–L5 and the regional pool R minimized the ability to form independent and disparate microenvironments. Separate local niches that originally formed during the first 26 d of insular setup elapsed. After, the microbiomes in L1–L5 and R formed a mutual niche where the same core microbiomes always dominated ( SI Appendix , Fig. S12.2 ). These metacommunity niche-shaping effects were major events that caused the transitional loss of cell types that were previously dominant in the insular phase. Our study demonstrated that the invented loop-designed mass transfer significantly switched the community assembly from the dispersal limitation to a homogenous dispersal process ( SI Appendix , section S14 ; NST, 40; QPEN, 38; iCAMP, 39), thereby stabilizing the microbiomes and increasing the synchrony between the six localities. It should be noted that while experiments which are designed to compare different treatments often focus on effect size and significance and therefore require a certain number of replicates, in our case stabilization of community structure and function, as compared to isolated reactors as in Liu et al. ( 9 ), has been demonstrated by a wide range of metrics, addressing different aspects of the community, all based on single-cell data, building on an analysis of about 90 million cells. More extensive studies, either tracking thousands of generations, as shown for example by the Lenski group ( 57 ) for pure populations, or involving multiple independent and parallel metacommunities, could further verify the results due to even higher sample density and the repetitive experimental design. Nevertheless, studying a community for a minimum of 114 generations under balanced continuous growth conditions allowed us to determine the basic properties of the community under successive changes in mass transfer rates. These were stability properties, diversity metrics, net growth rates of subcommunities, proportions of stochastic and deterministic processes, and the rescue of cell types at high mass transfer, creating communities that remain unchanged in composition and function. The metric “stability” is described by various properties, out of which constancy, resistance, and recovery are the most essential ones ( 1 ). We found greater constancy and resistance values in L1–L5 at RC 50 and RC 80 , reflecting stability of the community to stochastic assembly processes and perturbation events (i.e., dilution rate). The source–sink relationship between L1–L5 and the regional pool R created a reduction of the species pool but at high cell abundance that provided less space for neutral forces. Long-term constancy was established when transfer rates at RC 50 and RC 80 prevented the extinction of local species through the rescue effect and when regional equality was achieved according to Chesson ( 58 ). The increase in resistance that was observed in the present study could also be the result of the gradual increase in mass transfer rates (i.e., the RC 10 phase selected already partially nested SCs that eventually dominated at RC 50 and RC 80 , such as G2; SI Appendix , Table S8.4 ). Recovery by definition describes the ability of a community to return to the constancy space after a temporary disturbance ( 1 , 2 ). We found that recovery values were always low because the microbiomes were intentionally steered toward a low-diverse but unchanging state through the use of the loop-designed metacommunity setup. Nonetheless, recovery values were always positive, despite permanent changes in mass transfer rates and because of reinforcement of the rescue effect, which, in addition to the high constancy and resistance values, demonstrates the high stability properties of the metacommunity at RC 50 and RC 80 . Nevertheless, we would like to point out that very long-lasting mass transfer rates could still lead to slow, transient changes in microbial community structure, as Francis et al. ( 59 ) predicted for macroecological systems, but we always would expect synchrony to be maintained in local communities. The functions of the wastewater community, i.e., high carbon, ammonium-nitrogen, and phosphorus removal efficiencies and biomass production, were always active in both L1–L5 and R and were not lost even as mass transfer rates increased ( SI Appendix , section S3 ). For carbon and ammonium we even found an increase in removal efficiency with mass transfer rates in the metacommunity setup due to repeated recycling of the compounds in the flow, suggesting that the functions of the wastewater community to remove carbon, nitrogen, and phosphorus were at least stabilized if not increased. Notably, relief from mass transfer restored efficiencies to near starting values and also led to a renewal of earlier diversity levels (Insular I phase vs. Insular II phase), indicating that rare species were still present to allow diversity to return. By using various diversity metrics we found that high mass transfer rates affected diversity values and reinforced specific cell types. Looped high mass transfer did not alter richness in terms of number of dominant SCs or individual cell production per reactor much (PC; SI Appendix , Table S10.1 ), but it reduced β-diversity between and within microbiomes and resulted in the loss of local niches, which in turn reduced γ-diversity ( Table 1 ). As a consequence of the lower number of cell types, high mass transfer resulted in lower stochastic drift events ( SI Appendix , Table S8.1 ) and thus narrower established constancy spaces ( Fig. 2 B ). Fewer or even no drift events were found in the balanced growth periods, indicating that neutral forces in the metacommunity were low, differing from insular environments ( 9 , 10 ). Random birth and death events were limited by the continuous inflow of source organisms from the regional pool R and by the increase in biomass with increasing mass transfer ( SI Appendix , Fig. S3.1 ). High biomass is known to lower susceptibility to demographic drift or disturbance ( 60 ). Another argument in favor of the power of mass transfer to prevent variation was the low degree of the exchange, β SIM (turnover), of SCs within and between microbiomes at high mass transfer rates, which was different for insular situations ( SI Appendix , Fig. S8.6 and Table S8.3 ). Conditions of strong mass transfer have been described to moderate the turnover of community structures and support the nestedness of species, e.g. in the process of biotic homogenization ( 61 ) or postglacial recolonization processes of northern biotopes ( 62 ). Most SCs that benefitted from mass transfer were also those that showed persistence, β NES (nestedness; SI Appendix , Table S8.4 ). Similar to core species, they could also play an important role in maintaining community traits. Core species are common and dominant, e.g. in gut microbiomes ( 63 ), in benthic octocoral associations in the form of symbiosis ( 64 ), and species in macroecology that cope with climate warming, which largely determines temporal stability of the total biomass in alpine grassland communities ( 65 ). A huge number of technologies are available to recognize core species and, moreover, to determine their functions, as recently reviewed by Hatzenpichler et al. ( 66 ). However, our study did not focus on the functional activity of individual cells. Nevertheless, we were able to decipher some of the most persistent cell types and their functional capacity through correlation analyses combined with cell sorting and 16S rRNA gene sequencing. In our loop design, dominant SCs indicated an increased number of correlations despite nutrient limitation that was caused by high cell density during RC 50 and RC 80 in L1–L5 and especially in the regional pool R ( SI Appendix , Table S13.1 ). Genera in nested SCs were able to survive in R ( SI Appendix , Table S8.4 ) because they could potentially cope with the low carbon and ammonium resources, such as Sphingobacteriaceae ( SI Appendix , Fig. S9.1 ) ( 67 – 69 ) or PeM15 ( 70 ). Nested Azospirillum (G33), Azospira (G18), and Pseudacidovorax (G2, G11; SI Appendix , Fig. S9.5 ) are nitrogen fixers. They may have displaced other genera because of ammonium self-sufficiency ( 71 – 73 ). Thus, the regional pool R might have acted as a “hotspot” for genus and function selection, which underscores the possibility of selecting desired functions by modifying the conditions of the regional pool. We also found that mass transfer supported slow-growing organisms through the rescue effect. Mass transfer is a source–sink relationship and contributes to the spread and survival of species in sinks that would otherwise go extinct ( 74 ). Source–sink relationships are generally very strong to ensure coexistence, and modeled environmental variations were found to not affect them ( 75 – 77 ). The rescue of species at a sink site is reflected in various biotechnological processes where biotechnologists have successfully used this source–sink principle for bio-augmentation to keep a desired species in a system ( 78 , 79 ). Most studies, however, ignore or cannot distinguish between the niche-specific competitive hierarchy of microbiome members in sources or sinks and the effects of emigration and immigration on these relationships. Recent studies have begun to track sink members within microbiomes by calculating their net growth based on the relative abundances of 16S rRNA gene sequencing data and bulk biomass to approximate their contribution to sink biomass formation ( 80 , 81 ). Our study goes a step further by implementing net growth metrics for all cell types that migrated back and forth within the loop-designed source–sink metacommunity, based on data from individual cells. We found that slow-growing or almost nongrowing cell types, which would typically be washed-out under continuous feeding conditions ( 9 ), remained part of the microbiome through the rescue effect ( Fig. 4 ). At the extreme, at RC 80 , when the metacommunity was fully mixed and when the final microbiome was established, the growth of some SCs in L1–L5 was especially low or nonexistent because of an increase in biomass and decrease in nutrient resources. However, these SCs overcame their nongrowth in one or two reactors by growing in one or more of the other five local reactors and, subsequently, by redistribution via R, which enhanced their persistence in the metacommunity. Similar rescue effects were modeled for a macroecology background in relation to a metafood web, suggesting that biodiversity can be buffered under global change ( 82 ). The source–sink relationship from local communities L1–L5 to the regional pool R and its reverse loop supported the growth of SCs in L1–L5 rather than in nutrient-limited R. Unknown is why cells were nested in R and rescued through R. The few SCs (e.g., multidominant G5 and PeM15 in G12, G14) showed small cell morphologies. It is conceivable that they might have been able to successfully utilize the limited resources and simultaneously take advantage of the high-exchange-rate ( D = 3.6 d −1 ) qualities of R. Our setup was thus able to protect slow-growing or even nongrowing microorganisms for at least 114 generations, which otherwise would not have survived without mass transfer and the rescue effect. The loop-designed metacommunity may thus be a tool to protect and preserve functionally valuable microorganisms even if their growth rate is slow and lower than the prevailing dilution rate. In summary, looped mass transfer is a means of stabilizing microbial communities over long periods of time. The degree of stabilization can be selected via the mass transfer rate R C . Mass transfer reduced local and temporal variations, and the stochastic behavior that is normally observed in insular setups was reduced. All microbiomes showed high constancy and increasing resistance as well as unaffected functions at high mass transfer rates. Mass transfer also synchronized structures of the microbiomes by the mechanism of homogeneous dispersal, resulting in the lowest intercommunity β-diversity at the highest mass transfer. The variation of β-diversity within communities ceased, and the persistence of particular SCs was highest at high mass transfer. High turnover of community structures was observed only when no mass transfer occurred. An increase in mass transfer also increased cell numbers, thereby decreasing net growth rates μ ′ . Subcommunities that showed no growth ( μ ′ S C x = 0 ) in one locality were rescued by growth at another locality and by their redistribution via the loop design. Thus, lost SCs, whose growth rate was below the dilution rate and that would normally go extinct, were fostered and replaced by members of the source community. The regional pool itself also served as a rescue site through the redistribution of SCs that accumulated specifically in R. The local reactor conditions that were used in this study ultimately selected our microbiome. It is conceivable to test other local and especially regional conditions that support other cell types in natural communities in the future and thus design other stabilized communities. In particular, when certain medically or biotechnologically relevant functions are desired that require organisms with different physiological properties, including different growth rates, our loop design provides a solution for long-term stabilization and thus the reliable functioning of microbiomes."
} | 4,245 |
38294911 | PMC10846377 | pmc | 4,424 | {
"abstract": "Summary Superhydrophobic surfaces face challenges in comprehensive durability when used in extreme outdoor environments. Here, we present a protocol for preparing nanocomposite bulks with hierarchical structures using the template technique. We describe steps for using hybrid nanoparticles of polytetrafluoroethylene and multi-walled carbon nanotube to fill inside and dip on the polyurethane (PU) foam. We then detail procedures for its removal by sintering treatment. The extra accretion layer on the PU foam surface was highlighted to construct hierarchical porous structures. For complete details on the use and execution of this protocol, please refer to Wu et al. 1"
} | 168 |
37303512 | PMC10248095 | pmc | 4,425 | {
"abstract": "Accurate reservoir inflow forecasting is crucial for efficient water management. In this study, different deep learning models, including Dense, Long short-term memory (LSTM), and one-dimensional convolutional neural networks (Conv1D), were used to build ensembles. Seasonal-trend decomposition using loess (STL) was applied to decompose reservoir inflows and precipitations into random, seasonal, and trend components. Seven ensemble models, namely STL-Dense, STL-Conv1D, STL-LSTM, STL-Dense-LSTM-Conv1D, STL-Dense multivariate, STL-LSTM multivariate, and STL-Conv1D multivariate, were proposed and evaluated using daily inflows and precipitation decomposed data from the Lom Pangar reservoir from 2015 to 2020. Evaluation metrics, such as Mean Absolute Error (MAE), Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE), and Nash Sutcliff Efficiency (NSE), were applied to assess model performance. Results showed that the STL-Dense multivariate model was the best ensemble among the thirteen models with MAE of 14.636 m 3 /s, RMSE of 20.841 m 3 /s, MAPE of 6.622%, and NSE of 0.988. These findings stress the importance of considering multiple inputs and models for accurate reservoir inflow forecasting and optimal water management. Not all ensemble models were good for Lom pangar inflow forecast as the Dense, Conv1D, and LSTM models performed better than their proposed STL monovariate ensemble models.",
"conclusion": "5 Conclusion Due to the stochastic nature of reservoir inflow, there is a need to forecast future inflows as this will help in optimal water management with applications to hydroelectricity production, irrigation, and many others. The limitation of classical methods to handle the study of reservoir inflows due to the stochastic nature of water gives room to advanced methods precisely deep learning ensembles to handle this problem. Seven ensemble deep learning models based on STL were proposed for reservoir inflow forecast. STL was chosen for decomposition over other methods due to its robust nature, its ability to handle all kinds of seasonality, and the parameter tuning properties. STL decomposition was applied to the decomposed reservoir inflows and precipitations data set to forecast a day inflow. The decomposed components namely the random component, the seasonal component, and the trend component of the inflow forecast using deep learning models. The final prediction obtained is the sum of the three decomposed components. The forecasting results obtained with application on the Lom Pangar reservoir with statistical parameters of the ensemble deep learning models were compared with the single deep learning models. Among the thirteen deep learning models used, the STL-Dense multivariate model was the best. The combining effect of using multiple models and the application of multivariate forecasting by considering precipitation as part of the input leads to this model outperforming the other models. Deep learning ensemble models for daily inflow forecast are proposed but we acknowledge their limitations due to uncertainty analysis. Also, better hyperparameter tuning can greatly improve the model's performance. The unavailability of larger data sets in training and testing the models is part of the limitations of the proposed models. Not all ensemble models were better than their single model for the Lom pangar reservoir inflow forecast. The monovariate STL decomposition ensemble models perform poorly as compared to their deep learning models. This gives room for future research for which other ensemble methods like bagging, boosting, random forest, and many others could be taken into consideration for better ensembles. Also, varying hyperparameters to produce an ensemble deep learning model to forecast each of the decomposed components constitute the perspective of this work.",
"introduction": "1 Introduction A model for optimal water resource management requires a robust and efficient water management system [ 1 ] In the study of efficient water use, statistical and machine learning methods have been employed [ [2] , [3] , [4] ] with interesting output results. Due to the random nature of water resources, there is difficulty in modeling these systems using classical methods. To solve this problem, machine learning methods have been applied to model them for efficient sustainability [ 5 ]. In an attempt to further increase these models' efficiency, deep learning neural network models have been applied to reservoir inflow forecasting with promising results [ [6] , [7] , [8] , [9] , [10] , [11] , [12] ] Recurrent Neural Networks (RNN), Gated Recurrent Unit (GRU), and Long-Short Tem Memory (LSTM) models have been used for reservoir inflow forecasting, with LSTM proving to be the most effective [ 13 ]. A hybrid framework using machine learning for reservoir inflow forecast has been proposed by Tian et al. [ 14 ] with interesting results as an outcome as compared to classical methods. Luo et al. [ 15 ] proposed an ensemble model combining the Deep believe network (DBN) and LSTM with the ensemble result better than that of the individual models. Ensemble modeling has been researched for reservoir inflow prediction due to its improved performance through the combination of the strengths from different base models. An ensemble model consisting of various models has been proposed for forecasting inflow into a reservoir [ [16] , [17] , [18] , [19] ]. The obtained result for the cited ensembles is better than their composition models. Hong et al. [ 20 ] utilized an ensemble of recurrent neural networks (RNN-LSTM) and convolutional neural network (CNN-LSTM) for reservoir inflow prediction, while Zhang et al. [ 21 ] proposed a fusion model of CNN, Partial Least Squares (PLS), and Extreme Gradient Boosting (XGBoost) for inflow prediction. Mendes et al. [ 22 ] applied an ensemble of High-Resolution Model (HRES) and the Ensemble Prediction System (EPS) products to forecast the inflow of the Aguieira reservoir in Portugal. These proposed ensembles outperform the single models based on evaluation criteria. Wang et al. [ 23 ] utilized a hybrid decomposition-based multi-model and multi-parameter ensemble method for stream flow forecasting of the Yalong River, resulting in increased accuracy and reduced uncertainty. Sushanth et al. [ 24 ] applied an explainable machine learning model with LSTM for real-time inflow and streamflow forecasting of Konar and Tenughat reservoirs with appreciable accuracy up to a 3-day lead. The shortcoming of the above-cited ensemble models is the proposal of a single ensemble to treat the problem. As a contribution, seven ensembles from different deep-learning models are proposed. The Lom Pangar reservoir is the largest in the Southern Interconnected Grid (SIG) of Cameroon, serving to regulate the flow rate of the Song bengue river for hydroelectricity production 466 km downstream in the Songloulou hydropower plant. The study of this reservoir inflow rate is crucial for electricity production as it is the largest reservoir in the watershed. Moreover, its outflow regulates the flow rate on the Song bengue river which serves as an inflow in the Songloulou hydropower plant downstream. Songloulou power plant as of now is the largest power plant on SIG, the largest grid in the nation, and supplies energy to six regions out of the ten regions. In this study, seven ensemble deep learning models are proposed from Seasonal-trend decomposition using loess (STL). These models include STL-Dense, STL-Conv1D, STL-LSTM, STL-Dense-LSTM-Conv1D, STL-Dense multivariate, STL-LSTM multivariate, and STL-Conv1D multivariate. STL is used to decompose the reservoir inflow and precipitation into random, seasonal, and trend components. STL was chosen over other decomposition methods such as classical decomposition, and SEATS (seasonal extraction in autoregressive integrated moving average (ARIMA) time series) due to its ability to handle different types of seasonality and user control of smoothness in trend-cycle interaction among others [ 25 , 26 ]. Deep learning models namely the dense model, LSTM model, and Conv1D have been implemented to forecast the STL decomposed components and the resultant reservoir inflow is the sum of the forecasted decomposed inflows of the reservoir. Thirteen deep learning models, namely the dense model, the Conv1D model, the LSTM model, the multivariate dense model, the multivariate Conv1D model, the multivariate LSTM model, the STL-dense model, STL-Conv1D model, STL-LSTM model, STL-Dense_LSTM_Conv1D model, STL-Dense multivariate model, STL-Conv1D multivariate model, and STL-LSTM multivariate model are evaluated using evaluation criteria for the best model. The mean absolute error (MAE), mean square error (MSE), root mean square error (RMSE), mean absolute percentage error (MAPE), and Nash Sutcliff Efficiency (NSE) are used as evaluation criteria. This manuscript is partitioned as follows: section 1 is the introduction, section 2 presents the study area, section 3 gives the methodology used, and section 4 summarizes the result and discussion. The conclusion of the work is in section 5 .",
"discussion": "4 Results and discussion 4.1 Results The result obtained when the models were applied to forecast a day reservoir inflow of the Lom pangar reservoir using a window size of the past seven days is shown in Fig. 9 , Fig. 10 , Fig. 11 , Fig. 12 . The deep learning models were executed in Python 3.8 on a personal computer with AMD A6-3410MX APU Radeon (tm) HD Graphics 1.6 GHz processor. 80% of the data was used to train the models while 20% was used to test the models. This division was done since it provides a better result empirically [ 36 ]. The elaborate comparative result is shown in Table 2 . The best result was obtained with the STL-dense-multivariate model with the least MAE of 14.636 m 3 /s. With the univariate deep learning models namely the dense model, conv1D model, and LSTM, the least error was obtained with the dense model followed by the LSTM model then conv1D. In general, there is not much variation in the NSE of the models. The values are between 0.986 and 0.988. Table 2 Evaluation criteria for the different models. Table 2 Deep learning Model MAE (m 3 /s) MSE (m 3 /s) 2 RMSE (m 3 /s) MAPE (%) NSE Dense model 14.961 474.579 21.785 6.597 0.987 Conv1D model 15.631 521.269 22.831 6.861 0.986 LSTM model 15.575 518.863 22.779 6.881 0.986 Multivariate dense model 15.072 488.127 22.094 6.592 0.987 Multivariate Conv1D model 15.579 516.560 22.728 6.855 0.986 Multivariate LSTM model 15.280 475.594 21.808 6.853 0.987 STL-Dense model 15.178 468.785 21.651 6.757 0.988 STL-Conv1D model 15.783 521.841 22.844 6.921 0.986 STL-LSTM model 16.227 543.184 23.306 7.303 0.986 STL-Dense-LSTM-Conv1D model 15.517 497.051 22.295 6.924 0.987 STL-LSTM multivariate 15.984 506.955 22.516 7.276 0.987 STL-Conv1D multivariate 15.349 489.776 22.131 6.808 0.987 STL-Dense multivariate model 14.636 434.353 20.841 6.622 0.988 Fig. 9 , Fig. 10 , Fig. 11 , Fig. 12 (a), represent the results obtained from forecasting the Lom pangar daily inflows using the thirteen deep learning models. Fig. 9 (a–d) represent the plot of the inflows with STL-Conv1D ensemble model, STL-Conv1D multivariate ensemble model, STL-Conv1D ensemble model, and STL-Dense multivariate ensemble model respectively. Fig. 10 (a–d) represent the plot of the inflows with the STL-LSTM ensemble model, STL-Dense-LSTM-Conv1D ensemble model, STL-Dense ensemble model, and Multivariate Dense ensemble model respectively. Fig. 11 (a–d) represent the plot of the inflows with the LSTM model, Conv1D model, Dense model, and Multivariate LSTM model. The detailed results from the evaluation criteria for the thirteen models are shown in Table 2 . Fig. 1 s(a–d) presents the results obtained from forecasting the decomposed components namely the trend, random, and seasonal components using the dense model. Table 3 presents the results obtained from the evaluation criteria. Fig. 9 Lom pangar inflows compared with (a) STL-Conv1D ensemble, (b) STL-Conv1D multivariate ensemble, (c) STL-Conv1D ensemble, and (d) STL-Dense multivariate ensemble. Fig. 9 Fig. 10 Lom pangar inflows compared with (a) STL-LSTM ensemble, (b) STL-Dense-LSTM-Conv1D ensemble, (c) STL-Dense ensemble, and (d) Multivariate Dense ensemble. Fig. 10 Fig. 11 Lom pangar inflows compared with (a) LSTM model, (b) Conv1D model, (c) Dense model, and (d) Multivariate LSTM model. Fig. 11 Fig. 12 Lom pangar inflows compared with (a) multivariate Conv1D model, (b) STL-trend inflow compared with its dense forecast, (c) STL-random inflow compared with its dense forecast, and (d) STL-seasonal inflow compared with its dense forecast. Fig. 12 Table 3 MAE for the different decomposition terms with Dense, conv1D, and LSTM models. Table 3 MAE (m 3 /s.) Random inflow Lom pangar Seasonal inflow Lom pangar Trend inflow Lom pangar Dense model 13.871 5.706 0.137 Conv1D model 13.781 8.639 0.129 LSTM model 14.892 6.811 0.135 4.2 Discussion Unlike Qi et al. [ 37 ] which implement Empirical mode decomposition to forecast the reservoir inflow for univariate time series, this work takes into consideration the reservoir precipitation and constitutes a multivariate problem. Again, Qi et al. [ 37 ] applied LSTM to forecast the decomposed variable while in addition to LSTM, this work takes into consideration the dense model and Conv1D model to forecast the decomposed variables. The addition of these two deep learning models and the proposal of seven ensembles gives the planner a wide range of methods to choose from. This lack of numerous ensembles is the limitation of Qi et al. [ 37 ]. Also, the findings of this work are per Ding et al. [ 38 ] who equally use STL decomposition with the application of LSTM to forecast hydroelectricity production. The findings of Ding et al. [ 38 ] suggest that the proposed ensemble performed better than the LSTM model. The limitation of [ 38 ] concerning this work is the same as that of [ 37 ] which includes the proposal of a single ensemble as compared to seven ensembles proposed. Again, the advantage of these ensembles permits us to notice that contrary to Refs. [ 37 , 38 ] in which their proposed methods use LSTM to build the ensembles, this study shows that the LSTM ensemble is good but the best ensemble was obtained with the STL-Dense multivariate ensemble in forecasting the reservoir inflow for Lom pangar reservoir. Again, the least MAPE of 6.622% was obtained from the multivariate dense model still strengthening the advantage of the multivariate model considering precipitation over the univariate model for reservoir inflow forecast at the Lom Pangar reservoir. From Table 3 , the three models namely Dense, Conv1D, and LSTM forecast the trend term with a very small MAE. Thus, the deep learning models used are suitable to forecast the trend term of the daily reservoir inflow of the Lom Pangar reservoir. For the univariate models, the dense model obtained the best MAE result for forecasting the seasonal inflow of Lom pangar. The Conv1D model forecast the random and trend inflow and obtains the minimum error among the three univariate deep learning models. These show the correlation between the decomposed data sets with the deep learning models, and also the strength of assembling other models to obtain a better fit. Surprisingly, from Table 2 , the Dense, Conv1D, and LSTM models perform better than their monovariate STL decomposed models. The MAE of the models are 14.961 m 3 /s, 15.631 m 3 /s, and 15.575 m 3 /s as compared to 15.178 m 3 /s, 15.783 m 3 /s, and 16.227 m 3 /s respectively. This result shows that not all assemble are better than the individual deep learning model for forecasting the daily reservoir inflow of the Lom pangar reservoir. It gives the application of a new domain of research like using other ensemble methods such as bagging, boosting, and many others on the data set to study the correlations for more efficient ensembles.\n\n4.2 Discussion Unlike Qi et al. [ 37 ] which implement Empirical mode decomposition to forecast the reservoir inflow for univariate time series, this work takes into consideration the reservoir precipitation and constitutes a multivariate problem. Again, Qi et al. [ 37 ] applied LSTM to forecast the decomposed variable while in addition to LSTM, this work takes into consideration the dense model and Conv1D model to forecast the decomposed variables. The addition of these two deep learning models and the proposal of seven ensembles gives the planner a wide range of methods to choose from. This lack of numerous ensembles is the limitation of Qi et al. [ 37 ]. Also, the findings of this work are per Ding et al. [ 38 ] who equally use STL decomposition with the application of LSTM to forecast hydroelectricity production. The findings of Ding et al. [ 38 ] suggest that the proposed ensemble performed better than the LSTM model. The limitation of [ 38 ] concerning this work is the same as that of [ 37 ] which includes the proposal of a single ensemble as compared to seven ensembles proposed. Again, the advantage of these ensembles permits us to notice that contrary to Refs. [ 37 , 38 ] in which their proposed methods use LSTM to build the ensembles, this study shows that the LSTM ensemble is good but the best ensemble was obtained with the STL-Dense multivariate ensemble in forecasting the reservoir inflow for Lom pangar reservoir. Again, the least MAPE of 6.622% was obtained from the multivariate dense model still strengthening the advantage of the multivariate model considering precipitation over the univariate model for reservoir inflow forecast at the Lom Pangar reservoir. From Table 3 , the three models namely Dense, Conv1D, and LSTM forecast the trend term with a very small MAE. Thus, the deep learning models used are suitable to forecast the trend term of the daily reservoir inflow of the Lom Pangar reservoir. For the univariate models, the dense model obtained the best MAE result for forecasting the seasonal inflow of Lom pangar. The Conv1D model forecast the random and trend inflow and obtains the minimum error among the three univariate deep learning models. These show the correlation between the decomposed data sets with the deep learning models, and also the strength of assembling other models to obtain a better fit. Surprisingly, from Table 2 , the Dense, Conv1D, and LSTM models perform better than their monovariate STL decomposed models. The MAE of the models are 14.961 m 3 /s, 15.631 m 3 /s, and 15.575 m 3 /s as compared to 15.178 m 3 /s, 15.783 m 3 /s, and 16.227 m 3 /s respectively. This result shows that not all assemble are better than the individual deep learning model for forecasting the daily reservoir inflow of the Lom pangar reservoir. It gives the application of a new domain of research like using other ensemble methods such as bagging, boosting, and many others on the data set to study the correlations for more efficient ensembles."
} | 4,762 |
35702894 | PMC9543831 | pmc | 4,428 | {
"abstract": "Abstract Microbial communities in many ecosystems are facing a broad range of global change drivers, such as nutrient enrichment, chemical pollution, and temperature change. These drivers can cause changes in the abundance of taxa, the composition of communities, and the properties of ecosystems. While the influence of single drivers is already described in numerous studies, the effect and predictability of multiple drivers changing simultaneously is still poorly understood. In this study, we used 240 highly replicable oxic/anoxic aquatic lab microcosms and four drivers (fertilizer, glyphosate, metal pollution, antibiotics) in all possible combinations at three different temperatures (20, 24, and 28°C) to shed light into consequences of multiple drivers on different levels of organization, ranging from species abundance to community and ecosystem parameters. We found (i) that at all levels of ecological organization, combinations of drivers can change the biological consequence and direction of effect compared to single drivers, (ii) that effects of combinations are further modified by temperature, (iii) that a larger number of drivers occurring simultaneously is often quite closely related to their effect size, and (iv) that there is little evidence that any of these effects are associated with the level of ecological organization of the state variable. These findings suggest that, at least in this experimental ecosystem approximating a stratified aquatic ecosystem, there may be relatively little scope for predicting the effects of combinations of drivers from the effects of individual drivers, or by accounting for the level of ecological organization in question, though there may be some scope for prediction based on the number of drivers that are occurring simultaneous. A priority, though also a considerable challenge, is to extend such research to consider continuous variation in the magnitude of multiple drivers acting together.",
"introduction": "1 INTRODUCTION Microbial communities are key components of many ecosystems and are often exposed to many anthropogenic global change drivers (Christensen et al., 2006 ; Jackson et al., 2016 ). Exposure to fertilizer (da Costa et al., 2021 ) and metal pollution (Xu, Chen, et al., 2018 ), pesticides like glyphosate (Relyea, 2009 ; Solomon & Thompson, 2003 ), antibiotics (Xu, Li, et al., 2018 ), and temperature increase (Wu et al., 2011 ) force populations and whole ecosystems to develop in different ways compared to less or unaffected ones (Tylianakis et al., 2008 ). Studies show that even slight changes in environmental conditions can lead to large taxonomic and functional microbial change, and even to regime shifts, with potential long‐term consequences for biogeochemical cycles and ecological function of the affected habitat (Gruber, 2011 ). While many studies have focused on one global change driver, recent studies (Christensen et al., 2006 ; Rillig et al., 2019 ; Suleiman, Choffat, et al., 2021 ) indicated that the combined effect of multiple drivers applied simultaneously is often not equal to the sum of the effects of each individual driver. Rather, they demonstrate that interactive effects can occur. These can be synergistic (combined effect greater than the sum of the effect of individual drivers) or antagonistic (combined effect less than the sum of the effect of individual drivers). Furthermore, not just the type of driver, but also the number of factors combined plays a crucial role (Rillig et al., 2019 ). Nevertheless, the number of studies testing three or more drivers remains very low (Rillig et al., 2019 ), which highlights the need for further investigations of combined effects. Combining this with the importance of understanding drivers of change in microbial communities is a critical aspect of the emerging research field of climate/global change microbiology (Hutchins et al., 2019 ). The limited number of studies with three or more drivers may be in part due to the logistical demands of the required experiments and the complexity of interpreting their results. For example, to assess all interactions a fully factorial experimental design is needed which quickly leads to large experiments. Regarding interpretation, it can be difficult to understand the meaning of interactions among more than three drivers, such that even if one would conduct a fully factorial experiment, traditional methods for interpreting interactions (such as interaction plots) may be insufficient. Some studies have instead focused on the effect of variation in the number of drivers and have applied only a subsample of all possible combinations (e.g. Brennan & Collins, 2015 ; Rillig et al., 2019 ). Another gap in understanding is how drivers and combinations of drivers act across levels of ecological organization from individuals to ecosystems (Galic et al., 2018 ; Simmons et al., 2021 ). For example, in a case study of a model of amphipod feeding behavior Galic et al. ( 2018 ) showed that responses to multiple drivers at the individual level were not consistent with those at higher levels of organization. In their case‐study, the nature of this inconsistency would lead to underestimation of effects at population and ecosystem levels if effects at the individual level were assumed to hold across levels of organization. These gaps in knowledge are problematic. One reason is that interactions can be source of ecological surprises, that is, if we assume additivity, we can be surprised if there are interactions (Christensen et al., 2006 ). Also, unless there are some generalities about interactions among drivers, we will never be able to predict the effect of a new (previously unstudied) combination of drivers, so interactions will also then be a surprise (Christensen et al., 2006 ). Hence, we and others (e.g., Rillig et al., 2019 ; Simmons et al., 2021 ) are interested in discovering if there are any general patterns that will allow us to predict the effects of combinations of environmental changes. This report is about our search for signals of general patterns in the effects of combinations of environmental drivers. For example, whether the relationship between a biological variable and the number of drivers is dependent on the level of ecological organization of the biological variable and whether the strength and nature of the interaction effects varies with level of ecological organization. Microbial communities in aquatic ecosystems are complex (Christensen et al., 2006 ; Davis et al., 2010 ; Faust & Raes, 2012 ), consisting of feedback among numerous abiotic and biotic interactions (Singh et al., 2009 ), among and within functional microbial groups (Bush et al., 2017 ; Richardson et al., 2018 ). Recent work indicates that these ecosystems are sensitive to environmental change (Christensen et al., 2006 ; Shade et al., 2011 , 2012 ; Suleiman, Choffat, et al., 2021 ; Suleiman, Pennekamp, et al., 2021 ), identifying them as appropriate systems for the study of the influences of global changes. In this work, we applied four different global change drivers (fertilizer, glyphosate, metal pollution, and antibiotics) together with increasing temperatures (20, 24, and 28°C) in all combinations possible, on a recently developed and highly replicable stratified aquatic microbial lab system. Our experiment includes the analysis of several abiotic (oxygen, total nitrogen [TN], total organic carbon [TOC], pH) and biotic variables (Shannon index, microbial community composition, genera abundances), since recent studies have shown that various levels of ecosystems can be affected (Shade et al., 2011 , 2012 ; Suleiman, Pennekamp, et al., 2021 ; Weithoff et al., 2000 ) and in order to examine if effects show predictable variation across levels of organization. We hypothesize that (i) combinations of drivers will have non‐additive effects on system properties, (ii) the non‐additive effects will be different depending on the temperature, and (iii) increasing the number of drivers applied will cause a systematic change in system properties.",
"discussion": "4 DISCUSSION Our study revealed compelling evidence of effects of the combination and the total number of drivers on taxa abundances, community composition, and ecosystem properties, and as such, that multiple drivers acting in combination can have important consequences across levels of ecological organization (Rillig et al., 2019 ). Specifically, the addition of a driver, and/or temperature increase, could change the biological consequences of already applied drivers, even when the added driver applied alone had no effect (Figure 3 ). We observed three trends here. The addition of a further driver could increase (e.g. F:A:T28), decrease (e.g. F:T28), or reverse the effect of the drivers which were already applied (e.g. F:A on Tychonema abundance, Figure 6 ). This demonstrates clearly the need of analysing multiple drivers in all combinations. Furthermore, our study also confirmed that the number of drivers occurring together can influence microbial communities and thereby provide some hope that number of drivers can be used as a useful predictor, though individual driver effects can greatly contribute to the effect of number of drivers (Rillig et al., 2019 ). Furthermore, the effect of the number of drivers can be negative (e.g., Tychonema abundance) or positive (e.g. TN) and of different magnitudes, which highlights again the need of analysing various ecological response variables of species, microbial communities, and the whole ecosystem. While there may be some patterns across levels of organization, e.g., the largest effects of number of drivers were observed at species and community levels of organization (greater than 0.5 or less than −0.5 standardized effect size), the absence of strong patterns limits the scope for explanation and prediction of effect of multiple drivers across levels of organization. FIGURE 6 Effects of fertilizer, antibiotics and their combination at 20, 24, and 28°C on amplicon sequence variant relative abundance, microbial composition, and ecosystem variables. Only significant treatments/treatment combinations are shown ( F ‐test p value <.05). Total nitrogen and total carbon are not listed in this figure because they were not significantly affected by F, A, and F:A It is possible that effects of number of drivers is caused by multiple drivers acting in a similar fashion as a single driver of greater intensity. That is, two drivers may be considered to typically have a higher effect because the overall intensity is higher. This would imply that drivers combine additively, or at least not completely substitutively. If this is the case, then the overall effect may partly be driven by the overall intensity of drivers. Given that responses can be non‐linear (and typically are), we can not exclude that some trends observed may simply be intensity effects. A clear pattern in the results is that of a less apparent variability at 0 and at 4 drivers applied than with 1, 2, or 3 applied. This is mainly because there are fewer combinations possible (1 combination possible to be precise: all or none), while at 1 to 3 drivers there are more combination of drivers possible, which leads to more variance. In contrast variance within driver combinations is roughly invariant regardless of number of drivers. We are certain if we had a 5th stressor applied, we would see much more variation at 4 stressors applied, again because then we would have (5 choose 4) = 5 possible stressor combinations. While fertilizer, metal pollution and antibiotics had detectable effects when applied as single drivers, glyphosate did not. This finding is likely quite specific to the concentration of treatments used and should be regarded with caution. Nevertheless, it is even more interesting that when adding glyphosate to other drivers it did have an effect (see e.g., Shannon index, Figure 3 ). Strong interactions were detected for antibiotics and fertilizer, and interestingly, these drivers revealed in combination different biological consequences compared to applied as a single driver, and temperature could also change the effects of the fertilizer‐antibiotic driver duo again (Figure 6 ). The effect of temperature is in our view important to note, since it predicts that observed biological consequences caused by global change can change with on‐going global warming. An increase of temperature can lead to new interactions among driver combinations (F:A with increasing temperature led again to significant affected variables) or can lead to significant effects that were not observed at 20°C (G:M, G:M:24 and G:M:28 for NMDS1). Our study also contributes to narrowing the gap of understanding of how drivers and their combination act across various levels of organization. Our results revealed that individual taxa can be affected differently compared to abiotic factors of the ecosystem or the microbial community composition. Therefore, our study is in line with the findings of Galic et al. ( 2018 ) who reported differences of the influence of multiple drivers across levels of organization. Interestingly, in our study, some treatment combinations, like fertilizer and antibiotics, showed significant influence across most variables and the three levels of organization (except TN and total carbon), though even then the sign of the interaction term was sometimes negative and sometimes positive. Effects of drivers, such as those we manipulated, are of course concentration‐dependent. In this study, we used only two levels of driver concentration and added the driver in a single pulse treatment. More information about specific concentration thresholds and press disturbances are important to study and may clarify some observed trends, as well as the timing of the driver applied. In addition, responses of functional (metabolic) aspects of the aquatic microbial communities should be investigated in future studies, which could shed light into activated and deactivated pathways and enzymes used in microbial communities to react to changing environmental conditions. One limitation of this study is that the background level of chemical stressor was not analysed prior to stressor addition. This information is important, since prior exposure to pollution would influence the responses to additional exposures. We expect that the small pond which was used to provide the water and sediment sample is low‐nutrient loaded, and is due to its location largely free from herbicide, metal‐, and antibiotic contamination. Nevertheless, this study showed clearly that combinations of stressors can lead to a new biological consequence, and this finding is very likely not impacted by any pre‐exposure of drivers of the small pond. One promising avenue for further research about species responses to multiple environmental drivers is to attempt to relate the patterns of response to features of the species, such as functional traits of the species. A relationship would reveal what traits of the species are determining how they respond to multiple drivers, and then give potential to predict those responses from species traits. This could be done for already present species, and also potentially invasive species. In studies such as ours, this would involve relating response patterns to characteristics of the microbial ASVs, and hence would require mapping of ASVs to trait information, which is not straightforward. Another option would be to use the 16S rRNA gene sequences to predict functional characteristics of the ASVs (e.g. (Ling et al., 2022 ). In sum, our study confirmed the need to research the effects of multiple drivers on microbial communities and indicates further that a broad range of levels of organization (species, community, ecosystem) should be analysed due to unique sets of effect and interactions across these and the different variables within them. Furthermore, our study showed that combination of specific drivers can change the biological consequence and direction compared to single drivers at all levels of organization and that the effects of driver combinations can be modified by temperature. Overall, the lack of patterns in effect sizes across levels of organization, and with respect to the number of drivers, represents lack of evidence for these to be useful in the prediction of the effects of combinations of global change drivers on ecological communities. Finally, it is unclear how transferable our findings are to non‐microbes or even to other types of microbial ecosystems. Microbes have some unique characteristics, such as mechanisms by which genetic material can move among individuals (horizontal gene transfer). There may be other important differences, such as strong interspecific interactions, particularly in closed experimental communities. Clearly, what is required is a sufficient number of experiments, across a diverse enough collection of ecosystem and organizm types, that involve at least four environmental drivers and observation of responses at population, community and ecosystem level, such that a formal meta‐analysis is possible. This could be achieved by a global network of such multi‐stressor factorial experiments studying responses across levels of ecological organization."
} | 4,342 |
37575387 | PMC10415191 | pmc | 4,431 | {
"abstract": "Due to technical limitations, research to date has mainly focused on the role of abiotic and biotic stress–signalling molecules in the aerial organs of plants, including the whole shoot, stem, and leaves. Novel experimental platforms including the dual-flow-RootChip (dfRC), PlantChip, and RootArray have since expanded this to plant-root cell analysis. Based on microfluidic platforms for flow stream shaping and force sensing on tip-growing organisms, the dfRC has further been expanded into a bi-directional dual‐flow‐RootChip (bi-dfRC), incorporating a second adjacent pair of inlets/outlet, enabling bi-directional asymmetric perfusion of treatments towards plant roots (shoot-to-root or root-to-shoot). This protocol outlines, in detail, the design and use of the bi-dfRC platform. Plant culture on chip is combined with guided root growth and controlled exposure of the primary root to solute changes. The impact of surface treatment on root growth and defence signals can be tracked in response to abiotic and biotic stress or the combinatory effect of both. In particular, this protocol highlights the ability of the platform to culture a variety of plants, such as Arabidopsis thaliana , Nicotiana benthamiana , and Solanum lycopersicum , on chip. It demonstrates that by simply altering the dimensions of the bi-dfRC, a broad application basis to study desired plant species with varying primary root sizes under microfluidics is achieved. \nKey features\n Expansion of the method developed by Stanley et al. (2018a) to study the directionality of defence signals responding to localised treatments. Description of a microfluidic platform allowing culture of plants with primary roots up to 40 mm length, 550 μm width, and 500 μm height. Treatment with polyvinylpyrrolidone (PVP) to permanently retain the hydrophilicity of partially hydrophobic bi-dfRC microchannels, enabling use with surface-sensitive plant lines. Description of novel tubing array setup equipped with rotatable valves for switching treatment reagent and orientation, while live-imaging on the bi-dfRC. \nGraphical overview\n \n \n Graphical overview of bi-dfRC fabrication, plantlet culture, and setup for root physiological analysis. (a) Schematic diagram depicting photolithography and replica molding, to produce a PDMS device. (b) Schematic diagram depicting seed culture off chip, followed by sub-culture of 4-day-old plantlets on chip. (c) Schematic diagram depicting microscopy and imaging setup, equipped with a media delivery system for asymmetric treatment introduction into the bi-dfRC microchannel root physiological analysis under varying conditions."
} | 660 |
37179628 | PMC10173426 | pmc | 4,432 | {
"abstract": "The controllable\nspontaneous transport of water droplets on solid\nsurfaces has a broad application background in daily life. Herein,\na patterned surface with two different non-wetting characteristics\nwas developed to control the droplet transport behavior. Consequently,\nthe patterned surface exhibited great water-repellant properties in\nthe superhydrophobic region, and the water contact angle reached 160°\n± 0.2°. Meanwhile, the water contact angle on the wedge-shaped\nhydrophilic region dropped to 22° after UV irradiation treatment.\nOn this basis, the maximum transport distance of water droplets could\nbe observed on the sample surface with a small wedge angle of 5°\n(10.62 mm), and the maximum average transport velocity of droplets\nwas obtained on the sample surface with a large wedge angle of 10°\n(218.01 mm/s). In terms of spontaneous droplet transport on an inclined\nsurface (4°), both the 8 μL droplet and 50 μL droplet\ncould move upward against gravity, which showed that the sample surface\npossessed an obvious driving force for droplet transport. Surface\nnon-wetting gradient and the wedge-shaped pattern provided unbalanced\nsurface tension to produce the driving forces in the process of droplet\ntransport, and the Laplace pressure as well is produced inside the\nwater droplet during this process. This work provides a new strategy\nto develop a patterned superhydrophobic surface for droplet transport.",
"conclusion": "Conclusions In\nsummary, non-wetting gradient surfaces with wedge-shaped patterns\nare fabricated through a combination method of chemical etching, gradient\nchemical modification, and selective UV irradiation. The superhydrophobic\nregion on the sample surface exhibits a gradient water-repellant characteristic,\nand the water CA ranged from 160.5° ± 0.2° to 149.2°\n± 0.3°. Meanwhile, the water CA on the hydrophilic region\ngradually drops as UV irradiation time increases and equals 22°\n± 0.1° after 1-hour UV irradiation. On this basis, self-transport\nprocess of 8 μL droplets is observed on a horizontal sample\nsurface, and a maximum transport displacement of 10.62 mm is achieved\non the sample with a wedge angle of 5°, while a maximum average\ntransport velocity of 218.01 mm/s is achieved on the sample surface\nwith a wedge angle of 10°. Besides, self-transport processes\nof both 8 μL and 50 μL droplets on the inclined surface\nare evaluated. Finally, theoretical force analysis revealed that driving\nforces of droplet transport could be divided into three parts: force\nfrom non-wetting gradient, unbalanced surface tension from a wedge-shaped\npattern, and Laplace pressure inside the droplet. When the three-phase\ncontact line crosses the superhydrophobic and hydrophilic regions,\nthe driving force is the resultant force of the three driving forces.\nHowever, when the droplet moves into the hydrophilic region completely,\nthe driving force from non-wetting gradient disappears. Large droplets\ncould\nnot enter the hydrophilic region completely and are subjected to large\ndriving force all over the transport process, increasing the transport\ndistance.",
"introduction": "Introduction Controllable spontaneous water droplet\ntransport has important\nsignificance in multiple fields, such as microfluidic devices, 1 − 3 water harvesting, 4 − 6 and water–oil separation. 7 − 9 Surfaces with\ncontrolled wetting properties are widely acceptable materials to achieve\nself-transport of droplets. 10 , 11 In recent years, researchers\nhave modified various surfaces with non-wetting gradient to realize\nthe facility of unidirectional water self-transport. 12 , 13 For instance, Chaudhury and Whitesides 14 experimentally achieved droplet transport on a wetting gradient\nsurface, and a 1 μL droplet moved upward on an inclined surface\nwith a transport velocity of 1–2 mm/s. There are some related\nliterature reporting the steering of droplets using various surfaces.\nFor instance, Leng et al. 15 summarize the\nmechanisms of water droplet manipulation on various biological surfaces.\nYang et al. 16 explored multi-bioinspired\nSLIPS-patterned surfaces that allow for directional droplet sliding,\nand precise steering of droplet friction was created by coordinating\nthe heterogeneous wettability of the back of the desert beetle, directional-dependent\narchitecture of the butterfly wing, and ultraslippery configuration\nof N. alata . Yang et al. 17 reported three-dimensional (3D) topological\nSLIPS with an anisotropic-slippery-Wenzel state; fabricated rice leaflike\ngrooved nanotextured SLIPS can properly shape the droplet footprint\nto achieve a sliding resistance anisotropy of 109.8 μN. In recent years, it has been proved that wettability pattern design\nis the key to achieve spontaneous directional transport, and the wedge-shaped\npattern is quite effective and simple to prepare. 18 − 22 Khoo and Tseng 23 prepared\na chemically patterned nanotextured surface with wedge-shaped non-wettability\ngradient on the aluminum substrate, wherein both microliter and nanoliter\ndroplets could be transported rapidly along the expected direction.\nBesides, Zheng et al. 24 formulated a wedge-shaped\nsuperhydrophobic copper surface with a poly-(dimethylsiloxane) (PDMS)\noil layer and systematically discussed the droplet movement on the\nsurface. It is concluded that non-wettability gradient and wedge-shaped\ngradient are both beneficial for self-transport of water droplets. 25 − 29 However, only a few research studies combine these two methods to\nsynergistically achieve spontaneous directional transport of water\ndroplets, and how to effectively enhance the transport velocity and\ndistance of droplet is still an unclear problem. Herein, gradient\nnon-wetting surfaces were prepared on an aluminum\nsubstrate, and the hydrophilic regions were obtained on the substrate\nby proper ultraviolet irradiation treatment. The surfaces were characterized\nbased on several aspects and the motion behaviors of water droplets\nwere recorded using a high-speed camera. On this basis, some parameters\nsuch as the angle of wedge shape, droplet size, and surface inclination\nangle were adjusted, and the motion characteristics of droplets were\nanalyzed. Finally, the motion law of droplets and the state of driving\nforce during droplet motion were summarized and revealed. The velocity\nand distance of droplet movement on the surfaces were enhanced to\ningratiate diverse applications.",
"discussion": "Results and Discussion Surface Morphology and\nGradient Non-wettability After\nsandblasting and chemical etching treatment, uniform micropit structures\nappear on the sample surfaces, and well-arranged nanosheet structures\nwith a diameter of 100 nm evenly cover the sample surfaces, as shown\nin Figure 2 a,b. Figure 2 c,d presents the\nwetting state after surface chemical modification, wherein the droplet\nvolume is exactly 6 μL. After superhydrophobic modification,\nthe CA of the surface is about 158° due to the rough structures\nand low surface energy, as shown in Figure 2 c, 30 , 31 which proves that element\nF is grafted on the sample surfaces (from FAS-17). In comparison,\na significant decrease of wettability occurs on the surface after\nUV irradiation for 1 h. Figure 2 d shows that the droplet could spread on the sample surface,\nand the WCA drops to about 22°. Ultraviolet light has strong\noxidizing ability and could oxidize fluoroalkyl at the end of fluorosilane\nmolecules on the surface. Fluoroalkyl gradually changes into hydrophilic\nhydroxyl with the increase of oxidation time, which greatly increases\nthe surface hydrophilicity. 32 − 34 Surface composition changes could\nbe found in Figures S1 and S2 in the Supporting\nInformation. After 1 h of UV irradiation, the content of element F\ncould not be detected in the modified areas. The relationship between\nUV irradiation time and WCA is presented in Figure 2 e. The surface hydrophobicity gradually decreases\nas the irradiation time increase from 0 min to 1 h. In this process,\nthe WCA of the treated regions drops from 158.5° ± 0.2°\nto 22° ± 0.1° and the huge difference between the superhydrophobic\nand hydrophilic regions form the unbalanced surface tension, which\nis advantageous for liquid transport. 33 Figure 2 Wettability\ncharacteristics of the sample surface. (a) Micro-scale\nstructure of surface (×200). (b) Nano-scale structure of surface\n(×50,000). (c) Non-wetting sample surface after FAS-17 modification.\n(d) Hydrophilic sample surface after UV irradiation. (e) Variation\nof the water contact angle with UV irradiation time from 0 to 60 min. Spontaneous Transport Features of Droplets The complete\ncontrolled transport processes of 8 μL droplets on different\nsurfaces are shown in Figure 3 a–d, and controlled transport processes of 50 μL\ndroplets are shown in Figure 3 e; details could be found in Supporting Videos ( S1 – S5 ). In\ngeneral, the transport distance and transport velocity are two vital\nfactors to evaluate the droplet transport process. Figure 3 a–c displays the images\nof droplet transport on different horizontal patterned surfaces. Water\ndroplets move along the anticipated direction quickly without any\nexternal force on the surface with 5° wedge-shaped patterns,\nas shown in Figure 3 a. As a result, water droplets could only be transported for a relatively\nshort distance of 6.1 mm. However, on the surface in Figure 3 b, the superhydrophobic regions\nalso have surface chemical gradient apart from the wedge-shaped pattern.\nWhen the droplet is released from the tip end of the wedge-shaped\npattern, it moves spontaneously toward hydrophilic regions from the\nsuperhydrophobic regions. In the above process, a part of the droplet\nbottom contacts the boundary between the hydrophilic and superhydrophobic\nregions on the patterned surface, and an unbalanced capillary force\ndrives the droplet motion. Meanwhile, as the droplet diameter is larger\nthan the width of the pattern tip, a part of the droplet bottom contacts\nthe gradient wetting areas on the sample surface and is also subjected\nto the additional lateral wetting gradient force in the direction\nof droplet movement. The moving behavior of the water droplet is more\nprominent on a patterned non-wetting gradient surface than on a sample\nsurface with only the wedge-shaped patterns because of the additional\ndriving force from non-wetting gradient. Furthermore, under the additional\ndriving force, the maximum value of the front section of the droplet\ndeviating from the original position is 10.62 mm, which is a significant\nimprovement in transport distance than that on the sample surface\nwith only wedge-shaped patterns. Figure 3 Spontaneous droplet transportation in\ndifferent situations. (a)\nOn a surface with only a wedge-shaped pattern (ω = 5°).\n(b) On a surface with both non-wetting gradient and wedge-shaped pattern\n(ω = 5°). (c) On a surface with both non-wetting gradient\nand wedge-shaped pattern (ω = 10°). (d) Transportation\nof an 8 μL droplet on an inclined surface (ω = 10°).\n(e) Transportation of a 50 μL droplet on an inclined surface\n(ω = 10°). The wedge angle ω\nis an important factor affecting the process\nof droplet transportation. 35 Thus, wetting\npatterns with different ω (ω = 5° and ω = 10°)\nare designed to investigate the role of ω on the droplet transport\nprocess. A larger ω means the increase in area of the hydrophilic\nregion, leading to the faster spread velocity of droplet in the transport\nprocess, which shortens the transport distance of the water droplet\ndue to the resistance from the hydrophilic region. The results in Figure 3 c reveal that the\nvalue of maximum transporting displacement on the surface with ω\nof 10° is 6.62 mm, which is only 62.3% of that value on the surface\nwith ω of 5°. The transport behavior of different\nsized droplets on an inclined\nsurface is demonstrated in Figure 3 d,e. It is significant to evaluate the sample surface\nperformance to resist gravity. The inclination angle β is fixed\nat 4°, and the volume of the small droplet and large droplet\nequaled 8 and 50 μL, respectively. Besides, the wedge angle\nis selected to be exactly 10° for this experiment. The 8 μL\ndroplet climbs up for a maximum displacement of 2.82 mm in 38 ms,\nand a 50 μL droplet moves up to 4.34 mm within 53.4 mm, as shown\nin Figure 3 d,e. These\nresults prove that for a certain wedge angle, the larger droplet may\nmove a larger distance. 32 Droplet Transport\nBehavior Transport distance and velocity\nof droplets on the surfaces are recorded and shown in Figure 4 a–d, in order to understand\nthe characteristics of droplet transport, and some notable values\nin the process ( V max , V ave , and V 10ms ) are also listed\nin Table 1 . Herein,\nthe transport distance and velocity of the front end of the droplet\nare measured during the whole process. It is noted that both transport\ndistance and velocity of the droplets on the patterned non-wetting\ngradient surface (Sample 2, Sample 3) are larger than those on the\nsample without non-wetting gradient (Sample 1), which suggests that\nthe non-wetting gradient provides additional driving force and could\nbe superimposed on the unbalanced capillary force caused by the wedge-shaped\npattern in the expected transport direction. Furthermore, the droplets\non the surface with a smaller wedge angle (ω = 5°) has\na clear advantage in transport distance, and the maximum instantaneous\nvelocity V max reached a high value of\n682.6 mm/s, which is larger than those in several aforementioned studies. 23 , 24 , 36 However, V max could not fully represent the transport velocity of the\ndroplets. As listed in Table 1 , the average velocities on the surface with a larger wedge\nangle of 10° (218.01 mm/s) are faster than that on the surface\nwith a smaller wedge angle of 5° (212.58 mm/s), which signifies\nthat a larger ω may be beneficial for an overall faster droplet\ntransport behavior. Figure 4 Position and velocity of droplets with different volume.\n(a) Relation\nbetween droplet transport distance and time on different horizontal\nsurfaces. (b) Relation between droplet transport velocity and time\non different horizontal surfaces. (c) Relation between transport distance\nand time of droplets with different volume on inclined surfaces (β\n= 4°). (d) Relation between transport velocity and time of droplets\nwith different volume on inclined surfaces (β = 4°). Table 1 Droplet Transporting Velocities and\nDisplacements on Different Samples wetting gradient wedge angle\n(°) S max (mm) V max (mm/s) V ave (mm/s) V 10ms (mm/s) Sample 1 no 5 6.1 408.7 204 204.38 Sample 2 yes 5 10.62 682.6 212.58 254.61 Sample 3 yes 10 6.61 429.2 218.01 224.82 As the results shown in Figure 4 b, the transport velocities of droplets exhibit\noscillating\ncharacteristics during the whole transport process, which indicates\nthe occurrence of multiple acceleration and deceleration circles.\nAt the initial time, a large driving force is produced when the droplet\ncontacts with both the wedge-shaped pattern region and wetting gradient\nregion, and the front part of the droplet rapidly moves under two\ncombined driving forces. Meanwhile, a major part of the droplet moves\nforward with a slower velocity under the pull of the front part of\nthe droplet, due to the uneven force inside the droplet. When the\nfront part of the droplet moves forward, the combined driving force\ngradually decreases due to the lack of pressure difference and reduction\nof the wetting gradient force, resulting in a significant velocity\ndecrease of the droplet. Subsequently, the major part of liquid flows\ninto the front part through a meniscus during the coalescence of the\ndroplet. 37 − 40 After the coalescence of the droplet, the volume of the droplet\nincreases, and the thrust produced during the coalescence of the droplet\nenhances the advancing CA, causing enough driving force for a second\nacceleration and farther droplet transport. The combined driving force\ngradually decreases with the motion of the droplet and gradually drops\nto zero after 2 or 3 acceleration-deceleration processes, resulting\nin the ending of the transport process. It is noted that the time\ninterval between oscillations is related to ω, as shown in Figure 4 b, and the time interval\nbetween oscillations is inversely proportional to ω. The time\ninterval between oscillations depends on the flowing velocity of the\ndroplet, and a larger ω cause a wider meniscus, which accelerates\nthe flowing velocity of the liquid inside the droplet. 41 Though the V max (682.6\nmm/s) occurs on the surface with a smaller wedge angle, the driving\nforce at the initial position is closely related to the front contact\nline length between the droplet and wedge-shaped pattern, which means\nthat surfaces with larger ω could create stronger driving forces\nfrom the wedge-shaped pattern. 42 However,\nthere is another driving force, non-wetting gradient force, which\nexists on the patterned sample surface, and the continuous driving\nforce from the non-wetting gradient in initial transport aids to accelerate\nthe movement of the water droplet. For water droplets with the same\ndiameter, the contact area between droplets and superhydrophobic regions\nis larger on the surface with smaller ω, causing the larger\nadditional driving force from non-wetting gradient. Consequently,\nunder the superposition of the two driving forces, the droplet on\nthe surface with ω of 5° obtains a higher initial velocity.\nIn general, the larger the wedge angle, the stronger the driving force\nobtained from the wedge pattern, but this does not mean that the droplet\nhas a strong self-transport capacity. Factors such as the driving\nforce provided by the wetting gradient surface and the droplet wetting\ncontact area also influence the droplet self-transport behavior. It has been proven that droplets have great motion ability on a\nhorizontal surface with both non-wetting gradient and a wedge-shaped\npattern; here, the droplet transport behavior is further analyzed\non inclined surfaces (Sample 4, Sample 5). Compared with the droplet\ntransport on the horizontal surface, additional resistance force during\nthe transport process increases significantly due to the presence\nof gravity component mgsinβ. Therefore, as demonstrated in Figure 4 c,d, the motion ability\nof droplets is partially weakened on the inclined surface. For a droplet\nwith a fixed volume, not only transport distance decreases but also\nthe transport velocity is also reduced. For an 8 μL droplet\non the inclined surface, the maximum instantaneous velocity V max only reaches a value of 210.8 mm/s, which\nis only about 50% of that on the horizontal surface. The transport\ncharacteristics of droplets on the inclined surface\nare similar to those on the horizontal surface, and different sized\ndroplets both underwent multiple acceleration and deceleration circles.\nIt is surprising that larger droplets could move to higher positions\nwith a larger average transport velocity ( V ave ). Although the 50 μL droplet needs to encounter higher resistance\nforce due to larger mass, it has a larger combined driving force for\ntransport. On the one hand, a larger droplet volume leads to higher\nadvancing CA, providing larger driving force for droplet acceleration.\nOn the other hand, a larger droplet has more contact areas with hydrophobic\nnon-wetting gradient regions, resulting in a larger non-wetting gradient\nforce during transportation. Besides, transport velocity is also related\nto the degree of wetting difference in superhydrophobic areas and\nwedge-shaped pattern areas, and the increase of droplet volume means\nthe larger length of contact line between the droplet and wedge-shaped\npattern. 43 In fact, the driving forces\nfrom non-wetting gradient and the wedge-shaped pattern are both important\nfor the transport of the droplets. The droplet is difficult to achieve\nself-transport if there is insufficient driving force from non-wetting\ngradient. The relevant analysis is described in detail in the following\nsection. Mechanical Analysis of Droplet Transport In general,\ndroplet movement is suffered from various resistances, such as viscous\nforce ( F viscous ), hysteresis force ( F hsyteresis ), friction force ( F friction ), and surface tension force ( F surface tension ). When these resistances are greater\nthan the driving forces, droplets could not achieve self-transport.\nBefore discussing the droplet transport behavior on a patterned non-wetting\ngradient surface, a simpler situation is considered first. If a water\ndroplet is placed on a homogeneous surface, it is subjected to balanced\nsurface tension ( F s ). Figure 5 a shows the three-phase contact\nline of a water droplet placed on a homogeneous surface, and the surface\ntension force component in the x -axis direction along\nthe contact line could be expressed as: 24 1 where γ is surface tension,\nθ is the CA at the position of d s , R is circular radius, and α is the angle between the\ndirection of d s and x -axis. Since\nthe CA θ on a homogeneous surface is unchangeable and the three-phase\ncontact line shows complete symmetry, the driving force for droplet\nmotion from the surface equals to zero. However, an unbalanced F s exists in the x -axis direction\non some surfaces such as wetting gradient surfaces. 44 , 45 This driving force is caused by the asymmetric wetting properties\nand asymmetric water adhesion force at different positions on the\nsample surface. In this case, the above equation could appropriately\nbe corrected as: 2 where\nθ A and\nθ R are the advancing CA and receding CA, respectively.\nAs shown in Figure 5 b, F s on a non-wetting gradient surface\ndrives small droplets to move from regions with high WCA toward regions\nwith low WCA on the surface, leading to the directional droplet transport\nbehavior. Figure 5 Schematic diagram of driving forces acting on droplets during spontaneous\ntransportation. (a) Three-phase contact line of a droplet on a homogeneous\nsurface. (b) Driving force F s from wettability\ngradient. (c) Diagram of three-phase contact line of a droplet on\npatterned non-wetting gradient surface. (d) Front view of liquid bulge\nphenomenon. A droplet self-transport behavior\nalso occurs on a wedge-patterned\nsurface. Under this circumstance, the obvious wettability contrast\nbetween the superhydrophobic regions and hydrophilic regions provides\na relatively large continuous driving force for the droplet transport.\nIf CAs in the superhydrophobic regions and hydrophilic regions are\nθ sup and θ phi , respectively, the\nnet surface tension could be expressed as: 46 , 47 3 where ϕ B and\nϕ F are the azimuthal angles at the back boundary\nand front boundary, respectively, for the hydrophilic and superhydrophobic\nregions, as shown in Figure 5 c. F sx in formula 3 is proportional to the wettability difference\nin two regions, and a larger difference could generate a greater driving\nforce. Besides, transport behavior of droplets on the wedge-patterned\nsurface is also impacted by the Laplace pressure as another driving\nforce, which is produced by the variations in the fluid surface curvature\nduring this process. When the water droplet moves toward wider hydrophilic\nregions, a liquid bulge is formed and the droplet spreads out, causing\nan unbalanced Laplace pressure, as shown in Figure 5 d. The net Laplace pressure as a driving\nforce in the moving direction could be expressed as: 33 4 where δ( x ) is the liquid bulge width, r ( x ) is the curvature of the liquid which\nsatisfies r ( x ) = δ( x )/[2sinθ( x )], and θ avg is the contact angle over\nthe bulge length. As shown in formula 4 , a higher wedge angle leads to a greater Laplace pressure\nas driving force, so that the droplet has larger transport velocity.\nHowever, a larger wedge angle also leads to the increase of contact\narea between the droplet and hydrophilic regions, causing more resistance\nand shorter transport distance. In summary, a larger ω on the\nwedge-patterned surface causes a faster droplet transport velocity\nand a shorter distance, which is consistent with the experiment results\nin Figure 3 . The whole transport process could be divided into two stages (stage\nI and stage II), and droplet transport behaviors depend on different\ndriving force during two stages. Stage I is the initial stage of droplet\nmovement. When the droplet is first placed on the tip end of the wedge-shaped\npattern, the circular-shaped three-phase contact line crosses both\nthe hydrophilic region and superhydrophobic region, as shown in Figure 6 a. The part of the\ndroplet on the superhydrophobic region is affected by the wetting\ngradient and moves toward the hydrophilic region rapidly under the\nimpact of chemical gradient force. Meanwhile, the other part of the\ndroplet on the hydrophilic region was influenced by the continuous\nsurface tension and moves toward the wider end of the hydrophilic\nregion. In this situation, the droplet moves with a combining driving\nforce, leading to a larger initial transfer velocity. The combining\ndriving force ( F d ) in Figure 6 a could be expressed as: 5 Figure 6 Schematic\ndiagram of droplet force analysis during transportation\nin different situations. (a) Combined driving force F d during stage I of droplet transportation. (b) Driving\nforce during stage II of droplet transportation. (c) Diagram of droplet\nmoving upward on an inclined plane. (d) Comparison of droplet transporting\ndistance and average velocity under different inclination angles and\ndroplet volumes on the same surface. The resistance forces ( F r ) could be\nexpressed as: 6 These four kinds of\nresistances have been mentioned in the\naforementioned\nwork, and the hysteresis force and the viscous force are main resistances\nduring this process. 48 The spontaneous\ndroplet transport occurs under the coaction of driving force and resistance\nforce. After all the droplets moved into the wider hydrophilic region,\nthe transport process enters stage II with a relatively large initial\nvelocity. In stage II, the driving force of the droplets is primarily\nunbalanced surface tension from the wedge-shaped pattern and Laplace\npressure inside the droplet, which is lower than that in stage I,\ndue to the disappearance of driving force from non-wetting gradient.\nAt the same time, water adhesion with the hydrophilic region is clearly\nlarger than that on the superhydrophobic regions; hence, the resistance\nforce also increases at stage II. Figure 6 c presents\na schematic diagram of the force analysis of the droplet on an inclined\nsurface. Droplet transport on an inclined surface is subjected to\ngravity component mgsinβ in the opposite direction. As a result,\nthe moving velocity and distance of the droplet were both reduced\ncompared to that on the horizontal surface. Figure 6 d displays the comparison of the droplet\ntransporting distance and average velocity under different inclination\nangles and droplet volumes on the same surface. A large droplet could\nstill achieve upward transport of water droplets in spite of a greater\ngravity effect. It could be inferred that a larger droplet possesses\nmore contact areas with the substrate surface, and the width of the\ndroplet contacting the substrate surface exceeds the width of the\nhydrophilic areas. Therefore, large droplets are driven under a combined\ndriving force of chemical gradient force in the superhydrophobic region\nand force from the wedge-shaped pattern all over the transport process,\nand the droplet is maintained in stage I. In summary, a larger droplet\nis affected by both a greater driving force and higher resistance\nduring the upward transport."
} | 6,893 |
40275784 | PMC12022743 | pmc | 4,433 | {
"abstract": "ABSTRACT Photosymbioses, the symbiotic relationships between microalgae and non‐photosynthetic eukaryotes, are sporadically found in many eukaryotic lineages. Only a few taxa, such as cnidarians and ciliates hosting algal endosymbionts, have been actively studied, which has hindered understanding the universal mechanisms of photosymbiosis establishment. In Amoebozoa, few species are reported as photosymbiotic, and how the photosymbioses are established is still unclear. To investigate the extent to which one of the photosymbiotic amoebae, Mayorella viridis , depends on their symbionts, the amoebae were treated with reagents known to induce the collapsing of photosymbioses in other species. We succeeded in removing algal symbionts from the hosts with 2‐amino‐3‐chloro‐1,4‐naphthoquinone. While the apo‐symbiotic amoebae grew to the same extent as the symbiotic state when they fed on prey, their survival rates were lower than those of the symbiotic ones during starvation, suggesting that the impact of the photosymbiosis on fitness is condition‐dependent. Furthermore, we showed that the photosymbiotic state was reversible by feeding two strains of the green alga Chlorella to the apo‐symbiotic amoebae. The efficiencies of ingesting algal cells significantly differed between algal strains. These results suggest that the photosymbiotic relationship in the amoeba is facultative and that different algal strains have discrete symbiotic abilities to the amoeba.",
"conclusion": "5 Conclusion We have established an experimental system to collapse and re‐establish photosymbiotic relationships between \n M. viridis \n and Chlorella . Our data showed that the photosymbiosis was not absolute but dispensable under the nutrient‐rich condition tested and had a positive effect on the host fitness under starvation, suggesting that the evolutionary impacts of the photosymbiosis depend on the environmental contexts. The bleaching of \n M. viridis \n also allowed the quantification of the symbiotic abilities of each algal strain on certain criteria, which revealed that the symbiotic abilities of algae to \n M. viridis \n varied depending on the algal strain. This bleaching method will provide an important model tool not only for the comparative analysis of hosts in symbiotic and non‐symbiotic states but also for the observation of the establishment process of photosymbiosis in live cell imaging.",
"introduction": "1 Introduction Aquatic photosynthesis by algae and photosynthetic prokaryotes is responsible for approximately half of net‐production on the earth (Field et al. 1998 ). Some microalgae, as endosymbionts, can establish symbiotic relationships with non‐photosynthetic eukaryotes as hosts, which here we call photosymbiosis. Photosymbiosis has been reported from various aquatic and terrestrial environments. For example, corals, sea anemones, and other cnidarians belonging to Opisthokonta are known to have photosymbiotic relationships with some microalgae, including dinoflagellates and green algae (Davy et al. 2012 ). Photosymbioses between ciliates and Chlorella have also been well studied (Fujishima and Kodama 2012 ). Some foraminifera species, which are members of Rhizaria, have photosymbiotic relationships with several lineages of algal symbionts, such as dinoflagellates, red algae, and diatoms (Johnson 2011 ; Dorrell and Howe 2012 ). The mechanisms of photosymbioses are studied from a molecular, physiological, and ecological perspective, particularly in cnidarians and ciliates (Davy et al. 2012 ; Fujishima and Kodama 2012 ). Importantly, researchers can artificially “bleach” or induce the apo‐symbiotic state, in which all the algal symbionts are expelled from their host bodies. Developing a reliable bleaching method in such organisms is a prerequisite for comprehensive host‐symbiont studies. There are several reports about artificial bleaching methods so far (Belda‐Baillie et al. 2002 ; Mihirogi et al. 2023 ). For example, the model sea anemone Aiptasia ( Exaiptasia diaphana ) can be bleached by exposure to a high temperature of 35°C for 20 days (Belda‐Baillie et al. 2002 ) or more easily by treating them with the herbicide 2‐amino‐3‐chloro‐1,4‐naphthoquinone (ACN), which is also known as quinoclamine (Mihirogi et al. 2023 ). Green hydra, also belonging to Cnidaria and having green algal endosymbionts, can be bleached as well by treating with glycerol (Whitney 1907 ). Similarly, some bleaching induction methods have been developed for ciliates. \n Paramecium bursaria \n , intracellularly possessing Chlorella endosymbionts, can be bleached by treatment with paraquat for 5 days (Hosoya et al. 1995 ). There are reports that cycloheximide and 3‐(3,4‐dichlorophenyl)‐1,1‐dimethylurea (DCMU) are also effective in bleaching \n P. bursaria \n (Kodama and Fujishima 2008 ; Reisser 1976 ). Using these bleaching techniques, we can compare and analyze the responses when the hosts and symbionts are in a symbiotic state or apo‐symbiotic state. In the case of the sea anemone \n E. diaphana \n , the changes in the metabolisms of the hosts in response to symbionts can be studied by comparing transcriptomes of the symbiotic state with the apo‐symbiotic state (Lehnert et al. 2014 ; Ishii et al. 2019 ). Similar studies have been conducted on green hydras and \n P. bursaria \n (Kodama et al. 2014 ; Ishikawa et al. 2016 ; He et al. 2019 ). Eukaryotic species possessing photosymbionts have been found sporadically within phylogenetically distant taxa, such as Opisthokonta and Alveolata (Dorrell and Howe 2012 ), and a limited number of model species have been studied. Thus, it is still unclear whether cellular or molecular features found in some photosymbiotic organisms are universal or lineage‐specific. Amoebozoans have predatory behavior suitable for observing the phagocytosis process, which has been known as the first and key step of photosymbiosis establishment (Davy et al. 2012 ). Among them, only a few species ( Mayorella viridis , Parachaos zoochlorellae and Hyalosphenia papilio ) are found as photosymbiotic organisms, and few studies have been conducted on their photosymbiosis establishments at molecular and cellular levels (Willumsen et al. 1987 ; Weiner et al. 2022 ). These have hampered understanding of the importance of the phagocytosis process in photosymbioses, and it is important to build an ideal system in which we can simply compare the predatory process and symbiosis process. \n \n M. viridis \n is a member of Amoebozoa in freshwater and intracellularly possesses microalgal endosymbionts from the green algae of the Chlorella species (Cann 1981 ). Because the amoeba cells are adhesive and their locomotion is slower than that of some other photosymbiotic protists, such as the ciliate \n P. bursaria \n , it is comparatively easy to observe the algal symbionts inside the amoeba host cells under a microscope. Taking advantage of this species, here we show that we have developed a bleaching method for the amoebae and made a quantitative comparison between different Chlorella strains.",
"discussion": "4 Discussion 4.1 How Photosymbiosis Stability Can Affect the Amoeba Host The establishment of photosymbiosis between \n M. viridis \n and algal symbionts does not seem to be correlated to the photosynthetic activity of the symbionts because DCMU suppressed the photosynthetic activity of free‐living symbionts (see Figure S2 ) but did not affect the photosymbiosis stability (Figure 2A ). A previous study suggested that the symbiosis establishment between Aiptasia and symbiotic algae is independent of the symbiont's photosynthesis (Jinkerson et al. 2022 ). Thus, the bleaching of \n M. viridis \n induced by ACN seems to be caused by factors other than the inhibition of the symbiont's photosynthesis. Considering that ACN also induces the bleaching of cnidarians (Mihirogi et al. 2023 ), the same mechanism likely causes the symbiotic collapse in \n M. viridis \n . Further studies on the molecular mechanisms by which ACN affects photosynthetic symbionts and host organisms will shed more light on how photosymbiosis is established. Our results showed that the algal symbionts affect the amoebae host's fitness depending on the environmental context. When there are enough food supplies for the amoebae hosts, the symbionts may not be a major factor for the host's survival; whereas, in conditions with insufficient food supplies, the symbionts may have a more significant effect on their host's fitness. A previous study on \n P. bursaria \n showed that intracellular algal symbionts could negatively affect their hosts' fitness when the cost of maintaining the symbionts exceeds the benefits provided by the symbionts' photosynthesis (Lowe et al. 2016 ). The balance between cost (symbiont load) and benefits derived from the symbionts may fluctuate, for example, depending on light and feeding conditions, as shown in ciliates (Lowe et al. 2016 ), potentially leading to differences in the fitness of \n M. viridis \n between the symbiotic and apo‐symbiotic states. 4.2 Symbiotic Ability of Algal Symbionts and Symbiont Recognition by Host Cells The difference in symbiotic ability between the two Chlorella strains suggests that there is a symbiont recognition mechanism in \n M. viridis \n . Partner specificity in photosymbiosis has also been found in cnidarians (LaJeunesse et al. 2018 ; Yorifuji et al. 2021 ). A previous study on coral photosymbiosis suggests that the host corals can recognize their symbionts and modulate their immune systems (Jacobovitz et al. 2021 ). Recent studies on other corals suggest that LePin (lectin and kazal protease inhibitor domain) proteins in hosts may be involved in phagocytosis and recognition of the algal symbionts (Hu et al. 2020 , 2023 ). Interplay between lectins and sugar residues on the algal cell walls recognized may be a key for \n M. viridis \n to distinguish between symbionts and other organisms. It is shown that amoebozoan predators that feed on photosynthetic prey change the frequency of their phagocytic uptake and speed of food digestion depending on light conditions (Uzuka et al. 2019 ). Considering that the number of strain A inside the amoebae cells remained at a low level but stable after one‐week cultivation, the population of strain A, in addition to strain B, may be maintained at a certain level within \n M. viridis \n cells/populations in natural environments. Regulation of the balance between uptake, digestion, and expulsion of the algae by the host may have affected the difference in symbiotic abilities between the Chlorella strains used in this study. Comparative analysis of phagosome maturation of the apo‐symbiotic \n M. viridis \n cells ingesting prey and symbionts may provide insights into this and some hints to address evolutionary issues, for example, transitions from heterotrophic lifestyles to mixotrophic lifestyles in initial processes of plastid acquisition through endosymbiosis. \n Chlorella Strain B with high symbiotic ability to \n M. viridis \n was closely related to algal symbionts of the heliozoan \n A. turfacea \n , suggesting that the photosymbiotic ability of Chlorella species may be universal across phylogenetically distant host species. Quantification of the symbiotic abilities of strain B to other host species may shed light on this issue. Genomic analysis and the development of genetic engineering methods will be useful in elucidating the molecular mechanisms that cause the strain specificity of algal symbiosis. RNAi methods have already been established for Acanthamoeba , and similar methods may apply to \n M. viridis \n (Lorenzo‐Morales et al. 2005 )."
} | 2,921 |
34021824 | PMC8141083 | pmc | 4,434 | {
"abstract": "The increasing global perception of the importance of microbial inoculants to promote productivity and sustainability in agriculture prompts the adoption of bio-inputs by the farmers. The utilization of selected elite strains of nitrogen-fixing and other plant-growth promoting microorganisms in single inoculants creates a promising market for composite inoculants. However, combining microorganisms with different physiological and nutritional needs requires biotechnological development. We report the development of a composite inoculant containing Bradyrhizobium diazoefficiens and Azospirillum brasilense for the soybean crop. Evaluation of use of carbon sources indicates differences between the microbial species, with Bradyrhizobium growing better with mannitol and glycerol, and Azospirillum with malic acid and maleic acid, allowing the design of a formulation for co-culture. Species also differ in their growth rates, and the best performance of both microorganisms occurred when Azospirillum was inoculated on the third day of growth of Bradyrhizobium . The composite inoculant developed was evaluated in five field trials performed in Brazil, including areas without and with naturalized populations of Bradyrhizobium . The composite inoculant resulted in symbiotic performance comparable to the application of the two microorganisms separately. In comparison to the single inoculation with Bradyrhizobium , co-inoculation resulted in average increases of 14.7% in grain yield and 16.4% in total N accumulated in the grains. The performance of the composite inoculant was similar or greater than that of the non-inoculated control receiving a high dose of N-fertilizer, indicating the importance of the development and validation of inoculants carrying multiple beneficial microorganisms. Supplementary Information The online version contains supplementary material available at 10.1186/s13568-021-01230-8.",
"introduction": "Introduction The global search for low-cost agricultural technologies that can help increase food offer under sustainable production models is becoming more relevant (Sá et al. 2017 ). Microorganisms are major players in this vision of agriculture of the future and, in fact, in the last decade changes in farmers´ perception have already been noticed, reflecting in increased adoption of microbial bio-inputs (Malusá and Vassilevde 2014 ; de Bruijn 2015 ; Fukami et al. 2018a ; Bellabarba et al. 2019 ; Santos et al. 2019 ). In the coming years, Brazil should consolidate its position as a major player in the production and commercialization of food, in addition to standing out in the production of fibers, biofuels and biomass. Indeed, the country has just become the world's largest soybean producer [ Glycine max (L.) Merr.] (USDA 2020 ). The effort in the development of research with microorganisms of agricultural importance has been a constant in the country, and a great example is the adoption of the process of biological N 2 fixation (BNF) for the economic viability of the soybean crop, making it independent of N-fertilizers. The savings provided by the BNF with the soybean crop in Brazil are estimated today at US $ 14.4 billion per crop. Noticeable is also the environmental contribution, lowering the emission of greenhouse gases and reducing the leaching of nitrate to rivers, water reservoirs, groundwater, and lakes (Hungria et al. 2013a ; Hungria and Mendes 2015 ; Sá et al. 2017 ; Hungria and Nogueira 2019 ). Considering the soybean crop, three major milestones in the recent history of BNF in Brazil can be highlighted. The first is the validation of the reinoculation technology, with an average 8% increase in grain yield by inoculation every year (Hungria and Mendes 2015 ; Hungria and Nogueira 2019 ; Hungria et al. 2006 , 2007 , 2020 ). The second came from the selection of the first commercial strains of Azospirillum brasilense for grasses, with the release of the first commercial product in 2009 (Hungria et al. 2010 ; Santos et al. 2019 , 2021 ). The third major milestone was the development of the technology of co-inoculation for the soybean in 2013, consisting of the combination of strains of Bradyrhizobium spp. highly efficient in fixing nitrogen and strains of A. brasilense with high phytohormone production capacity (Hungria et al. 2013b , 2015 ). In the short period of five years, co-inoculation has been adopted on average in 25% of the 36 million ha cropped with this legume (Santos et al. 2021 ). There is no doubt that the combination of microbial inoculants can provide excellent results, justifying their great potential of being employed worldwide (Bashan 1998 ; Juge et al. 2012 ; Malusá and Vassilevde 2014 ; Fukami et al. 2017 ; Bellabarba et al. 2019 ; Santos et al. 2019 , 2021 ). Farmers, however, demand more practicality at sowing, and are requesting commercial products that combine various microorganisms. However, it is not easy to combine microorganisms with different nutritional requirements and growth rates. A strategy is presented in this study, where the stages of development and field validation of a composite inoculant containing Bradyrhizobium and Azospirillum for the soybean crop are presented.",
"discussion": "Discussion Although sucrose, mannitol, and glycerol were the C sources that yielded higher cell concentrations, growth of Bradyrhizobium in sucrose as a primary source of C was not expected, since the genus reportedly lacks the invertase enzyme, which is necessary for cells to metabolize sucrose, rhamnose and trehalose (Vincent 1977 ; Martinez-Drets and Arias 1974 ). For that reason, sucrose has been used by some inoculant industries as a cell protector of soybean Bradyrhizobium , and not as a C source. One hypothesis to explain the good growth observed in our experiments could be the possible hydrolysis of sucrose by hydrogen ions (H + ) derived from the dissociation of the molecule at high temperatures, in the process of autoclaving the culture medium. Another hypothesis is that, even with low dissociation, the remaining sucrose would be able to provide protection to the cells, favoring survival and contributing to greater cellular concentration. It is also important to search in the genome of the two strains of Bradyrhizobium used in this study (Hungria et al. 2018 ) for possible alternative routes of assimilation of sucrose. Mannitol is broadly known as an appropriate source of C for Bradyrhizobium and the same occurs with glycerol, which is also the source of preference in several inoculant industries (Vincent 1970 , 1977 ; Lopreto et al. 1972 ; Balatti 1992 ; Balatti and Freire 1996 ; Hungria et al. 2016 ). The growth of A. brasilense was maximized in a culture medium containing malic acid and maleic acid; in fact, the preferred use of malic acid by this species is broadly known (Döbereiner et al. 1995 ). However, the high cost of maleic acid makes it unattractive for utilization in the commercial production of inoculants. There was also good growth in the presence of mannitol, which is justified by the presence of the enzyme mannitol dehydrogenase (Westby et al. 1983 ). According to Döbereiner and Pedrosa ( 1987 ), and Hartmann and Baldani ( 2006 ), sucrose would not be a good source of C for A. brasilense ; however, in our study, good growth was verified with this source. As for Bradyrhizobium , it is necessary to investigate whether the growth of A. brasilense is due to partial hydrolysis of sucrose, or to metabolic pathways for the use of sucrose. Brazil plays an increasingly prominent role in the international agricultural scenario due to the use of inoculants in soybean crops, with high economic and environmental returns for farmers and for the country (Hungria et al. 2006 , 2007 ; Hungria and Mendes 2015 ; Hungria and Nogueira 2019 ). Increasing success has been obtained after the deployment of the technology of co-inoculation with Bradyrhizobium spp. and A. brasilense in 2013, with doubled grain yield benefits culminating in the commercial launch of a technological package containing the two bacteria separately (Hungria et al. 2013b , 2015 ; Barbosa et al. 2021 ; Santos et al. 2021 ). In a recent a meta-analysis of 48 publications covering field trials at 38 different locations in Brazil, Barbosa et al. ( 2021 ) detected statistically significant increases in grain yield and other plant growth parameters due to the co-inoculation. In addition, the main published studies in Brazil were recently compiled by Santos et al. ( 2021 ), and as an example, the studies of Ferri et al. ( 2017 ) and Galindo et al. ( 2018 ) indicated increases in grain yield due to the co-inoculation of 20.3 and 11.2%, respectively, compared to the inoculation exclusively with Bradyrhizobium . In the co-inoculation, although there may be a contribution from A. brasilense via biological nitrogen fixation, the benefits provided by strains Ab-V5 and Ab-V6 have been mainly attributed to the production of phytohormones, resulting in expressive increases in several root parameters (Fukami et al. 2018b ; Rondina et al. 2020 ). Root growth increases also favor the uptake of fertilizers (Galindo et al. 2021 ). There are also reports of induction of plant tolerance to abiotic stresses (Fukami et al. 2017 ; 2018a ). In addition to scientific evidence, the success of a new agricultural technology depends on large-scale proof of benefits to farmers. In this context, field extension efforts have been applied for three growing seasons in the state of Paraná, Brazil, with the establishment of reference units (RU) and field days. In the first year, 2017/2018, 37 RU were installed in 23 municipalities, assisting 665 farmers. Co-inoculation with Bradyrhizobium spp. and A. brasilense resulted in average increase in grain yield of 228 kg ha −1 with profit of R$ 263.4 ha −1 (~ U$ 70 in 05/18) in relation to the single inoculation with Bradyrhizobium (Nogueira et al. 2018 ). In the following season (2018/2019), in 61 RU located in 46 municipalities assisting 925 producers, co-inoculation resulted in average yield increase of 259 kg ha −1 , and net profit of R$ 296.00 ha −1 (~ U$ 76 in 05/19) (Prando et al. 2019 ) and, in the third season (2019/2020), these numbers were of 63 RU, in 54 municipalities, assisting 636 farmers, average gain of 266 kg ha −1 and profit of R$ 348.23 ha −1 (~ U$ 64 on 05/20) (Prando et al. 2020 ). The large-scale confirmation of benefits explains the widespread adoption of co-inoculation in the country in a short time, estimated at 15% of the entire cultivated area in 2018/2019, increasing to 25%, or almost 9 million ha in 2019/2020 (Santos et al. 2021 ). The success of co-inoculation in Brazil finds limitations in the use of microorganisms packaged separately, with a great demand, mainly by medium and small farmers, for composite inoculants. This is just one example, as the demand for composite inoculants, containing microorganisms with different metabolic functions, grows internationally (Santos et al. 2019 ). However, the development of inoculants containing microorganisms with different metabolic needs and growth rates is challenging. In this study, the feasibility of combining microorganisms with different metabolic needs and growth rates was demonstrated, through the development of a proper formulation composed by viable C sources and definition of times of inoculation. In our study, the preferred C sources for Bradyrhizobium were glycerol and mannitol, and for A. brasilense malic acid. In addition, as the growth rates were different, the best results were obtained with the inoculation of the fast-growing A. brasilense on the third day of growth of Bradyrhizobium . In five field trials, the developed composite inoculant showed performance similar to that of co-inoculation with the two microorganisms provided separately, resulting in average increase in grain yield of 502 kg ha −1 , or 14.7% in relation to the liquid inoculant containing only Bradyrhizobium (Fig. 2 ). It should also be noted that there is great concern in the agribusiness sector about the low levels of protein in soybean grains and there are indications that the N from the BNF is more easily translocated to grains than the N-fertilizer (Hungria and Neves 1987 ; Kaschuk et al. 2010 ; Hungria et al. 2020 ). This was confirmed in the field trials performed in our study, with an average increment of 16.4% in the total N accumulated in the grains in response to inoculation and co-inoculation (Fig. 2 ). The development of composite microbial inoculants and extension activities with the farmers showing the benefits of microbial inoculants should be encouraged globally, in view of several reports, including this one, showing agronomic, economic, and environmental benefits by the replacement of chemical fertilizers."
} | 3,232 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.