Integrated use of electrochemical anaerobic reactors and genomic based modeling for characterizing methanogenic activity in microbial communities exposed to BTEX contamination
电化学厌氧反应器与基因组建模的整合应用用于表征暴露于 BTEX 污染下微生物群落的产甲烷活性

The closed circuit MEC consisted of two 1-liter glass-bottles with carbon cloth electrodes, serving as the anode and cathode chambers. The electrodes were connected to an 800 mV-powered potentiostat (PalmSens, The Netherlands), while the double-chamber control MEC operated under open circuit conditions. The 3-liter volume AB and MEC chambers were inoculated with medium and clean soil collected from Newe Yaar Research Center (32°42′18″N; 35°11′07″E; Organic plots, from the margins of the control blocks, Ramat Yishay, Israel). The medium (previously described in (Yanuka-Golub et al., 2019) and soil characterization are described in Tables S1 and S2 respectively. A BTEX-DMSO (Sigma-Aldrich, Israel) was used as the stock solution (solution contain benzene, toluene, ethylbenzene, p-xylene, m-xylene, o-xylene, 1 mg mL⁻¹ each) to ensure BTEX solubility before injection into the reactors. BTEX was injected directly into the reactors through PVC tubes to reach a final concentration of 15 mg L⁻¹ for each BTEX component, according to effective concentration values described in (Lueders, 2017). Immediately following the injection, the tubes were washed several times by injecting the anaerobic medium described above. Glucose was injected during the clean soil phases as a single carbon source, reaching a final concentration of 500 mg L⁻¹ (Fig. 2 a). The addition of glucose was found to be necessary because the carbon sources were very low (chemical oxygen demand (COD) ∼ 50 mg L⁻¹). During the BTEX exposure phases (Fig. 2), BTEX was added as the single carbon source. Samples were collected directly from the reactors into 2 mL amber glass vials using a syringe. For BTEX analysis, the samples were preserved in the glass vials, ensuring no headspace to minimize BTEX evaporation. The samples were kept et 4 °C until measurement. BTEX concentrations were analyzed by gas chromatography–mass spectrometry (GC-MS, Agilent 6890). The produced biogas in the anaerobic system was collected in 3 L Tedlar Air® sampling bags. The collected biogas volume was measured weekly by gas chromatography GC system (Agilent 7890B Gas Chromatograph). Overall, the AB-MEC system operated for 160 days. During the experiment, monitoring of chemical oxygen demand (COD), pH, gas emission and short-chain fatty acids (SCFA) and microbial community dynamics was conducted weekly. SCFA were measured by HPLC (Thermo Finnigan LLC Scientific Surveyor, California, USA), pH by combo water tester (MRC Labs IP67, Israel) and COD was measured according to the 18th edition of Standard Methods for the Examination of Water and Wastewater (American Public Health Association, 1926). For community structure analysis, 16S rRNA gene amplicon sequencing was performed for samples taken at different time points from each reactor.
该闭路微生物电解池(MEC)系统由两个 1 升玻璃瓶构成，分别作为阳极室和阴极室，内装碳布电极。电极连接至 800 mV 恒电位仪(PalmSens，荷兰)，而作为对照的双室 MEC 则在开路条件下运行。3 升容积的厌氧消化(AB)和 MEC 反应器接种了来自 Newe Yaar 研究中心(32°42′18″N；35°11′07″E；以色列 Ramat Yishay 有机试验区边界对照地块)的培养基与洁净土壤。培养基配方(详见 Yanuka-Golub 等人 2019 年研究)及土壤特性参数分别见表 S1 和 S2。使用 BTEX-DMSO(Sigma-Aldrich，以色列)作为母液(含苯、甲苯、乙苯、对二甲苯、间二甲苯、邻二甲苯各 1 mg/mL ⁻¹ )以确保注入反应器前 BTEX 的溶解性。通过 PVC 管将 BTEX 直接注入反应器，根据(Lueders, 2017)报道的有效浓度值，使各 BTEX 组分终浓度均达到 15 mg/L ⁻¹ 。 注射完成后，立即使用上述厌氧培养基对反应管进行多次冲洗。在清洁土壤阶段注入葡萄糖作为单一碳源，最终浓度达到 500 mg/L（图 2a）。由于初始碳源含量极低（化学需氧量 COD 约 50 mg/L），葡萄糖的添加被证实是必要的。在 BTEX 暴露阶段（图 2），BTEX 作为单一碳源加入。使用注射器将反应器中的样品直接采集至 2 mL 棕色玻璃样品瓶中。针对 BTEX 分析，样品在无顶空条件下保存于玻璃瓶中以减少挥发，并于 4°C 冷藏直至检测。BTEX 浓度通过气相色谱-质谱联用仪（GC-MS，安捷伦 6890）进行分析。厌氧系统产生的沼气收集于 3L 特德拉 Tedlar Air®采样袋中，每周通过气相色谱系统（安捷伦 7890B 气相色谱仪）测量沼气体积。整个 AB-MEC 系统持续运行 160 天。 实验期间，每周监测化学需氧量（COD）、pH 值、气体排放量、短链脂肪酸（SCFA）及微生物群落动态变化。SCFA 采用高效液相色谱仪（美国加州 Thermo Finnigan LLC Scientific Surveyor）测定，pH 值使用组合式水质测试仪（以色列 MRC Labs IP67）检测，COD 则依据《水和废水标准检验方法》第 18 版（美国公共卫生协会，1926 年）进行测定。针对群落结构分析，对各反应器不同时间点采集的样本进行了 16S rRNA 基因扩增子测序。

Genome-scale metabolic models (GSMMs) were constructed for 21 microbial species, selected to represent the microbial communities in the different reactors. A metabolic model includes a network that is described as a stoichiometric matrix Sm × n, where m represents the number of metabolites and n the number of reactions in the model that are inferred based on the gene content in the respective genome (Mahadevan and Schilling, 2003). Genome sequences representing the selected ASVs were retrieved from public depositories of fully sequenced genomes based on sequence similarity search. Representative ASVs and the respective genome sequences selected for constructing the metabolic networks models are listed in Table 1. With the exception of the two methanogens (Methanosarcina acetivorans and Methanobacterium formicicum), the initial automatic construction followed the protocol described by (Ofaim et al., 2020; Dhakar et al., 2021, 2022; Xu et al., 2019). Briefly, Model SEED was used for constructing the initial draft metabolic models from the genome sequence data (Faria et al., 2018). Annotations were done through RAST (Meyer et al., 2008). For the two methanogens, draft reconstruction relied on a high-quality reconstruction of a methanogen (Richards et al., 2016) that was modified according to the genome specificities of the selected representatives. As part of the automatic reconstruction, the reactions of each of the models were further updated according to additional available genome annotation schemes including KEGG (Kanehisa et al., 2014), UniProt (UniProt Consortium, 2018), JGI (Nordberg et al., 2014). Reactions added from additional annotation platforms were converted according to the KBASE (Arkin et al., 2018) rxn conventions, reaction stoichiometry and reversibility was validated, and futile loops were identified and eliminated. Following initial reconstruction, the models were manually curated through a systematic procedure: First, metabolic capacities and physiological growth conditions were determined based on a detailed literature survey (Table 2). Then, simulations were conducted to validate the is silico performances of each model. For microorganisms designated as BTEX degraders (based on amplification of functional genes), their selectivity to BTEX components was retrieved from the literature, also ensuring their ability to utilize these components as a sole carbon source. Relevant degradation pathways were added to the respective models based on the KEGG database (https://www.genome.jp/kegg/pathway.html). For methanogens, the full inclusion of reactions involved in acetoclastic and hydrogenotrophic methanogenesis was verified and missing reactions were added. Similarly, growth requirements of members of other functional groups were defined based on a comprehensive literature search and the models were modified accordingly (Table 2). To validate the performances of the different models we designed a simulation matrix testing biomass production and consumption and production of short chain fatty acids (SCFA), H₂, CO₂ and CH₄ for each model in 25 in silico media representing control (non-feasible conditions), and selective media that are relevant to specific microbial groups as defined based on their validated physiology from the literature (Table 2). All media are composed of essential trace elements and minerals (MMM: K⁺, Mn²⁺, CO₂, Zn²⁺, SO₄²⁻, Cu²⁺, Ca²⁺, HPO₄²⁻, Mg²⁺, Fe²⁺, Cl⁻) supplemented with alternative C and N sources, in accordance with species physiology. Simulations were carried out using flux balance analysis (FBA), following protocols described in (Ofaim et al., 2020; Dhakar et al., 2021, 2022; Xu et al., 2019). Briefly, FBA makes use of the cellular reconstructions for depicting cellular processes. This is done by following the law of mass conservation while considering the metabolic framework as a static stoichiometric matrix (metabolites × reactions) (Rana et al., 2020). FBA describes the predicted flux distribution through linear optimization by targeting specific cellular objectives (mainly growth). Here, the growth of the model was chosen as the objective function through the maximization of the biomass reaction. Specific fluxes for metabolites consumption and secretion were determined following fixation of the respective biomass reaction to its maximal value and then using Flux Variability Analysis (FVA), a linear programming method to find the minimal amount of metabolites needed to produce a preset amount of biomass (Mahadevan and Schilling, 2003). All model simulations were done on an Intel i7 quad-core server with 128 GB of memory, running Linux. The development programming language of our simulators was Java, and our linear programming software was IBM CPLEX. Simulation environments and running conditions, and simulation outcomes (text files and heatmaps) and code for constructing the simulation matrix are available in https://github.com/FreilichLab/Supplementary_methanogenesis/tree/main/BEvis. Models are available as Systems Biology Markup Language (SBML) (Hucka et al., 2003) files https://github.com/FreilichLab/Supplementary_methanogenesis/tree/main/MatrixViz.
本研究为 21 种微生物物种构建了基因组尺度代谢模型（GSMMs），这些物种选自不同反应器中的微生物群落代表。代谢模型包含一个由化学计量矩阵 Sm×n 描述的代谢网络，其中 m 代表代谢物数量，n 表示基于相应基因组基因内容推断的模型反应数量（Mahadevan 和 Schilling，2003 年）。通过序列相似性搜索从公共全基因组数据库中获取了代表选定 ASVs 的基因组序列。用于构建代谢网络模型的代表性 ASVs 及其对应基因组序列列于表 1。除两种产甲烷菌（乙酸甲烷八叠球菌和甲酸甲烷杆菌）外，初始自动构建流程遵循 Ofaim 等人（2020 年）、Dhakar 等人（2021 年，2022 年）及 Xu 等人（2019 年）描述的方案。简言之，采用 Model SEED 从基因组序列数据构建初始代谢模型草案（Faria 等人，2018 年），并通过 RAST 系统完成注释（Meyer 等人，2008 年）。 针对这两种产甲烷菌，草案重建工作基于一个高质量产甲烷菌模型（Richards 等人，2016 年研究）进行修改，并根据所选代表菌株的基因组特性进行调整。在自动重建过程中，各模型反应进一步根据其他可用基因组注释方案进行更新，包括 KEGG（Kanehisa 等人，2014 年）、UniProt（UniProt Consortium，2018 年）和 JGI（Nordberg 等人，2014 年）。从其他注释平台添加的反应按照 KBASE（Arkin 等人，2018 年）反应规范进行转换，验证了反应化学计量和可逆性，并识别消除了无效循环。初步重建完成后，通过系统程序对模型进行人工校验：首先基于详细文献调研确定代谢能力和生理生长条件（表 2），随后通过模拟验证各模型的计算机性能表现。 对于被指定为 BTEX 降解菌的微生物（基于功能基因扩增），其对各 BTEX 组分的选择性数据来自文献检索，同时确保这些微生物能以 BTEX 组分作为唯一碳源。根据 KEGG 数据库（https://www.genome.jp/kegg/pathway.html）将相关降解途径添加至相应模型。针对产甲烷菌，我们验证并补全了涉及乙酸营养型和氢营养型产甲烷途径的所有反应。同理，通过全面文献调研确定了其他功能菌群的生长需求，并相应调整了模型参数（表 2）。为验证不同模型的性能，我们设计了模拟矩阵：在 25 种计算机模拟培养基中测试各模型的生物量产出、短链脂肪酸（SCFA）代谢以及 H ₂ 、CO ₂ 和 CH ₄ 的生成情况，这些培养基包括对照组（不可行条件）和根据文献证实的特定微生物群生理特性配置的选择性培养基（表 2）。 所有培养基均根据物种生理特性，添加了必需微量元素和矿物质（MMM：K ⁺ 、Mn ²⁺ 、CO ₂ 、Zn ²⁺ 、SO ₄ ²⁻ 、Cu ²⁺ 、Ca ²⁺ 、HPO ₄ ²⁻ 、Mg ²⁺ 、Fe ²⁺ 、Cl ⁻ ）以及替代碳氮源。模拟采用通量平衡分析（FBA）进行，遵循（Ofaim 等人，2020；Dhakar 等人，2021，2022；Xu 等人，2019）所述方案。简言之，FBA 利用细胞重建来描述细胞过程，其通过遵循质量守恒定律，将代谢框架视为静态化学计量矩阵（代谢物×反应）（Rana 等人，2020）。FBA 通过线性优化针对特定细胞目标（主要是生长）来描述预测的通量分布。本研究选择模型生长作为目标函数，通过最大化生物量反应来实现。 通过将各生物量反应固定至最大值后，采用通量变异性分析（FVA）方法测定代谢物消耗与分泌的特异性通量。该线性规划方法用于确定产生预设生物量所需代谢物的最小量（Mahadevan 和 Schilling，2003）。所有模型模拟均在配备 128GB 内存的 Intel i7 四核服务器（运行 Linux 系统）上完成。模拟器开发采用 Java 编程语言，线性规划软件使用 IBM CPLEX。模拟环境与运行条件、模拟结果（文本文件与热图）以及构建模拟矩阵的代码详见 https://github.com/FreilichLab/Supplementary_methanogenesis/tree/main/BEvis。模型以系统生物学标记语言（SBML）（Hucka 等，2003）文件形式发布于 https://github.com/FreilichLab/Supplementary_methanogenesis/tree/main/MatrixViz。

The designed bioreactor system consisted of three chambers: Anaerobic Bioreactor AB chamber and two chambers of Microbial Electrolysis Cell (MEC) – anode and cathode chambers (Fig. 1). Each chamber was designed to selectively enrich a specific microbial functional population to spatially dissect a typical redox gradient (Fig. 1a). The anode-chamber served as an electron acceptor where iron reducers were expected to be enriched (yellow group, Fig. 1); proton transfer from the anode towards the cathode were expected to enrich hydrogenotrophic methanogens vs acetoclastic methanogens (green and brown groups, respectively, Fig. 1). Alternatively, in the AB reactor, no specific condition was applied, beyond the absence of oxygen resulting in a mixed methanogenic population, comprising a higher abundance of acetolactic compared to hydrogenotrophic methanogens (Dolfing, 2001; Thauer et al., 2008; Liu and Whitman, 2008). In parallel, an additional double-chamber MEC system operated under open circuit conditions and served as a control. Overall, the experiment was designed to compare the functionality and composition of the soil-derived microbial communities within the different reactors. The AB-MEC system operated for 160 days, where the period was divided into four distinct phases (Fig. 2): I) First "clean" phase (days 0–40): inoculation with non-contaminated soil and system initiation (days 0–40) with glucose addition on day 19. II) First BTEX exposure period (days 40–83) with two consecutive spiking additions of BTEX (days 40 & 63). III) Second "clean" phase (days 83–146) with no BTEX exposure allowing soil natural attenuation and soil remediation, followed by glucose addition (days 83, 91, 111). IV) Second BTEX exposure period (days 146–160): re-contamination of the soil by additional two consecutive spiking events (days 146 & 159). As the clean soil was poor in organic carbon glucose was added as a carbon source during the clean phases (Fig. 2). To ensure process stability and to estimate proper time for glucose/BTEX spiking, pH and chemical oxygen demand (COD) were measured weekly. In all bioreactors COD increased sharply with glucose addition (Fig. 3b). Nevertheless, BTEX addition did not induce such a sharp rise in COD. As expected, pH measurements remained significantly different between bioreactors' chambers. The pH in the anode closed circuit bioreactor chamber decreased (∼pH = 6) and the pH of the cathode closed circuit bioreactor chamber increased (∼pH = 9). Such measures indicate oxidation reactions and H₂ release in the cathode, due to reduction reactions in the anode. In comparison, in the AB and open circuit MECs bioreactors’ chamber the pH remained relatively constant (Fig. 3c). Fig. 3a shows that in all reactor chambers, the concentration of the four BTEX components compounds dropped rapidly. Such results may be due continuous mixing of soil slurry that increases carbon sources and water availability or absorption (Zanello et al., 2021). Rapid degradation of BTEX components is likely to be abiotic as efficient biodegradation needs a longer acclimation period (European Chemicals Bureau and European Chemicals Bureau, 2008; European Chemicals Bureau and European Chemicals Bureau, 2007; Carvajal et al., 2018). At the end of the experiment (day 159) BTEX compounds did not degrade completely: 30%, 0.2%, 10% of added benzene,17%, 0.5% and 3% of added toluene, 12%, 4% and 1% of added ethylbenzene and 14.7% 2% and 6% of added xylenes remained in AB, anode and cathode CC, respectively. Gas emission profiles indicated that methanogenic activity significantly differed in the bioreactor system (Fig. 3d). While in the AB chamber the measured biogas emission (CO₂:CH₄) at the end of the experiment was 50%:33% - consistent with the typical range previously reported for anaerobic bioreactors (Kapoor et al., 2020), the anode CC chamber displayed a much lower biogas percentage CH₄:CO₂ ratio of 12%:32% at the end of the experiment. An opposite pattern was detected in the cathode chamber where CH₄ reached 84% at phase 3 of the experiment with no apparent CO₂ emission. Such results suggest the selective enrichment of specific bacterial groups, possibly iron reducers in the anode and methanogens in the cathode. As expected, monitored biogas emission in open circuit MECs showed the same ratio as in AB bioreactor (Fig. 3d). The detected differences in the emission profile support that selective conditions induced a functional divergence in the microbial communities in the reactors (Fig. 1).
设计的生物反应器系统由三个腔室组成：厌氧生物反应器 AB 腔室和微生物电解池（MEC）的两个腔室——阳极室与阴极室（图 1）。每个腔室专门设计用于选择性富集特定微生物功能种群，以空间解析典型的氧化还原梯度（图 1a）。阳极室作为电子受体，预期富集铁还原菌群（黄色组别，图 1）；质子从阳极向阴极转移的过程预计将促进氢营养型产甲烷菌相对于乙酸营养型产甲烷菌的富集（分别为绿色和棕色组别，图 1）。而在 AB 反应器中，除缺氧条件外未施加特定限制，因此形成了混合型产甲烷菌群，其中乙酸营养型产甲烷菌的丰度高于氢营养型产甲烷菌（Dolfing，2001；Thauer 等，2008；Liu 和 Whitman，2008）。实验同时设置了在开路条件下运行的双室 MEC 系统作为对照。整体实验设计旨在比较不同反应器中土壤源微生物群落的功能与组成差异。 AB-MEC 系统运行了 160 天，实验周期划分为四个不同阶段（图 2）：I）首个"清洁"阶段（0-40 天）：接种未污染土壤并启动系统（0-40 天），第 19 天添加葡萄糖。II）首次 BTEX 暴露期（40-83 天），分两次连续投加 BTEX（第 40 天和第 63 天）。III）第二个"清洁"阶段（83-146 天），无 BTEX 暴露以促进土壤自然衰减和修复，期间分三次添加葡萄糖（第 83 天、91 天和 111 天）。IV）第二次 BTEX 暴露期（146-160 天）：通过两次连续投加（第 146 天和 159 天）对土壤进行再次污染。由于清洁土壤有机碳含量较低，在清洁阶段添加葡萄糖作为碳源（图 2）。为确保过程稳定性并确定葡萄糖/BTEX 的最佳投加时机，每周监测 pH 值和化学需氧量（COD）。所有生物反应器中 COD 均随葡萄糖投加而急剧上升（图 3b），但 BTEX 投加并未引起 COD 的类似骤增。正如预期，各生物反应器舱室的 pH 测量值始终存在显著差异。 阳极闭合电路生物反应器腔室中的 pH 值降低（约 pH=6），而阴极闭合电路生物反应器腔室的 pH 值升高（约 pH=9）。这些测量结果表明，由于阳极发生还原反应，阴极发生了氧化反应并释放 H ₂ 。相比之下，AB 和开路 MEC 生物反应器腔室中的 pH 值保持相对恒定（图 3c）。图 3a 显示，在所有反应器腔室中，四种 BTEX 组分的浓度均迅速下降。这一结果可能是由于土壤浆料的持续混合增加了碳源和水分的可利用性或吸收（Zanello 等，2021）。BTEX 组分的快速降解很可能是非生物性的，因为有效的生物降解需要更长的适应期（欧洲化学品管理局，2008；欧洲化学品管理局，2007；Carvajal 等，2018）。 实验结束时（第 159 天），BTEX 化合物未完全降解：在厌氧生物反应器（AB）、阳极室和阴极室中分别残留了添加苯的 30%、0.2%、10%，添加甲苯的 17%、0.5%、3%，添加乙苯的 12%、4%、1%，以及添加二甲苯的 14.7%、2%、6%。气体排放曲线表明生物反应器系统中的产甲烷活性存在显著差异（图 3d）。AB 反应室实验末期测得沼气排放比例（CO₂:CH₄）为 50%:33%——与既往报道的厌氧生物反应器典型范围一致（Kapoor 等，2020），而阳极室在实验末期显示的沼气比例 CH₄:CO₂显著降低至 12%:32%。阴极室则呈现相反模式，在实验第三阶段 CH₄占比达到 84%且未检测到明显 CO₂排放。这些结果表明特定菌群（阳极可能富集铁还原菌，阴极可能富集产甲烷菌）的选择性增殖。正如预期，开路微生物电解池（MECs）监测到的沼气排放比例与 AB 生物反应器相同（图 3d）。 检测到的排放特征差异表明，选择性条件诱导了反应器中微生物群落的功能分化（图 1）。

DNA samples were collected along the experimental phases (Fig. 2) and microbial community composition was assessed using 16S amplicon sequencing. A sharp decline in biodiversity was observed in the three reactors on the 3rd phase, after glucose addition suggesting enrichment and further domination of specific microbial species (Fig. 4a). Alternative alpha diversity measures, such as Chao1 or Faith PD indices, were inconclusive regarding the effect of BTEX exposure on community diversity, however all measures indicated a dramatic decline in community diversity in the 3rd phase (Fig. S1). The PCoA analysis showed that while initial microbial communities in the three chambers were similar at the beginning of the experiment (phase I) - a divergence occurred along the experiment time period (Fig. 4b). The most dramatic shift in the diversity distances of the bioreactor chambers was observed after glucose addition (day 99, phase 3) in parallel to the drastic decline in Shannon diversity indices. Following this shift, the communities in the cathode and anode bioreactors' chambers dispersed in opposite direction from the AB chamber. The divergence is congruent with the functional differentiation as a response to the chamber-specific conditions applied. Microbial classes dominating the reactors during the first phase (clean soil) of the experiment were Clostridia, Bacteroidia, and Bacilli, all of which exhibited high abundance in all three bioreactor chambers. Clostridia were significantly more abundant in AB than in MEC reactors (Fig. 5). At the same time period, Alphaproteobacteria and Gammaproteobacteria were found to be abundant in the MEC chambers but not in the AB chamber (p < 4.4x10⁻¹²). The high abundance (>7.5%) of six microbial classes in the MEC chambers vs four in the AB chamber possibly accounts for their higher biodiversity (Fig. 4a). BTEX addition significantly increased (p < 1.6x10⁻²²) the relative abundance of Clostridia and Bacilli in MEC reactors but not in AB (Fig. 5). In the third phase (second glucose addition) of the experiment, there was a significant and sharp increase in the relative abundance of Clostridia in all reactors' chambers which apparently became the most dominant class in all chambers. The noticeable and rapid rise in the relative abundance of Clostridia (Fig. 5) can be attributed to the decrease in biodiversity (Fig. 4a) and to the shift in community composition (Fig. 4b) detected in all reactors in this phase. In parallel, there was significant increase in the relative abundance of Bacteroidia (p < 5.46x10⁻⁴¹) versus a decline of Bacilli, Alphaproteobacteria and Gammaproteobacteria (p < 0.001) in all bioreactors’ chambers (Fig. 5). Throughout the experiment, we observed an enrichment of Anaerolineae in all bioreactors' chambers. In addition, the relative abundance of Thermotogae in the AB chamber enriched (p = 1.3x10⁻⁵⁷) in the fourth phase (second BTEX spiking). Finally, at the end of the experiment the relative abundance of Archaeal classes (compromising mostly methanogens) reached 6.4% and 2.3% in the AB and cathode chambers, respectively, compared with 0.8% in the anode chambers. The abundance pattern correlated with the high biogas emission profile observed in the late phases of the experiment (mainly 3rd and 4th) in these two chambers but not in the anode (Fig. 2b). The different biogas emission profiles observed in the third phase of the experiment – 50:33 vs 84:16 CO₂:CH₄ ratio in the AB vs cathode chamber, respectively - corroborates to significant changes within Archaeal class composition. Whereas in the AB reactor the most dominant Archaeal class are Methanosarcinia (mostly of acetoclastic methanogens), reaching 5% of the microbial community and 78% of the Archaea (in comparison to 19% in the cathode chambers), the most dominant class in the cathode are Methanobacteria (mostly hydrogenotrophic methanogens) reaching 1.8% of the microbial community and 79% of the Archaea (in comparison to 15% in the AB chamber). The high ratio of hydrogenotrophic methanogens in the cathode at the final stages of the experiment provided a potential explanation for the high methane rate observed in the cathode chamber.
实验各阶段均采集了 DNA 样本（图 2），并通过 16S 扩增子测序评估微生物群落组成。在第三阶段添加葡萄糖后，三个反应器均观察到生物多样性急剧下降，表明特定微生物物种的富集与进一步优势化（图 4a）。虽然 Chao1 指数和 Faith PD 指数等α多样性指标对 BTEX 暴露影响群落多样性的结论不一致，但所有指标均显示第三阶段群落多样性出现断崖式下降（图 S1）。主坐标分析（PCoA）表明，尽管实验初期（第一阶段）三个反应室的初始微生物群落相似，但随着实验推进出现了分化（图 4b）。在添加葡萄糖后（第 99 天，第三阶段），伴随着香农多样性指数的骤降，生物反应器各室的多样性距离发生了最显著变化。此后阴极和阳极生物反应器中的群落分别朝与 AB 反应室相反的方向离散。 这种差异与功能分化相一致，是对各反应室特定环境条件的响应。实验第一阶段（清洁土壤）中，主导反应器的微生物类群为梭菌纲（Clostridia）、拟杆菌纲（Bacteroidia）和芽孢杆菌纲（Bacilli），这三类微生物在所有三个生物反应室中均呈现高丰度。其中 AB 反应器中的梭菌纲丰度显著高于 MEC 反应器（图 5）。同期，α-变形菌纲（Alphaproteobacteria）和γ-变形菌纲（Gammaproteobacteria）在 MEC 反应室中丰度较高，而在 AB 反应室中未检测到显著存在（p < 4.4×10^-10）。MEC 反应室中六类微生物的高丰度（>7.5%）相较于 AB 反应室中的四类，可能是其更高生物多样性的原因（图 4a）。BTEX 添加使 MEC 反应器中梭菌纲和芽孢杆菌纲的相对丰度显著增加（p < 1.6×10^-11），但 AB 反应器中未观察到该现象（图 5）。在实验第三阶段（第二次葡萄糖添加）中，所有反应室中梭菌纲的相对丰度均出现显著急剧上升，显然成为各反应室中最优势的类群。梭菌纲相对丰度的显著快速升高（图 5）可归因于此阶段所有反应器中生物多样性的下降（图 4a）及群落组成的变化（图 4b）。与此同时，所有生物反应器腔室中拟杆菌纲（Bacteroidia）相对丰度显著上升（p < 5.46×10^-2），而芽孢杆菌纲（Bacilli）、α-变形菌纲（Alphaproteobacteria）和γ-变形菌纲（Gammaproteobacteria）则呈现下降趋势（p < 0.001）（图 5）。整个实验过程中，我们观察到所有生物反应器腔室中厌氧绳菌纲（Anaerolineae）的富集现象。此外，在第四阶段（第二次 BTEX 投加期间），AB 腔室中热袍菌门（Thermotogae）的相对丰度显著增加（p = 1.3×10^-3）。实验末期，AB 腔室和阴极腔室中古菌纲（主要为产甲烷菌）的相对丰度分别达到 6.4%和 2.3%，而阳极腔室仅为 0.8%。这种丰度分布模式与实验后期（主要是第三和第四阶段）在这两个腔室中观察到的高生物气排放特征相吻合，但在阳极腔室中未出现此现象（图 2b）。 实验第三阶段观察到的不同沼气排放特征——阳极生物膜反应器(AB)与阴极室的气体比例分别为 50:33 和 84:16（CO ₂ :CH ₄ ）——证实了古菌纲组成发生的显著变化。在 AB 反应器中，优势古菌纲为甲烷八叠球菌纲（主要为乙酸营养型产甲烷菌），占微生物群落的 5%和古菌的 78%（阴极室仅为 19%）；而阴极室优势菌纲为甲烷杆菌纲（主要为氢营养型产甲烷菌），占微生物群落的 1.8%和古菌的 79%（AB 室仅为 15%）。实验末期阴极室氢营养型产甲烷菌的高占比，为观测到该室甲烷高产率现象提供了合理解释。

To go beyond the descriptive level gained by genomic characterization towards a functional view of the degradation process as depicted in Fig. 1, we simplified the analysis by focusing on microorganisms representing the primary groups involved in the degradation processes, starting from BTEX degradation and ending in methane emission. Species selection aimed to focus on key taxonomic groups in each reactor, capturing both functional and taxonomic composition. For downstream analysis, representative species were such that have a closely related homolog with fully sequenced genome. Overall, the 24055 16S ASVs were mapped to 2680 unique taxa comprised of 815 genera, 471 families, 147 classes and 53 unique phyla (Fig. 6). To identify potential BTEX degraders in the reactors we have complemented the 16S rRNA amplicon sequencing with bamA gene sequencing. bamA is the gene encoding for 6-oxocyclohex-1-ene-1-carbonyl-coenzyme a hydrolase. This key enzyme carries out the ring cleavage step and is considered as biomarker for anaerobic BTEX degradation (Wu et al., 2022; Kuntze et al., 2008; von Netzer et al., 2016). bamA was successfully identified in all the reactors after BTEX addition (Fig. S2). To determine the taxonomic origin of bamA, the gene was amplified in the three reactors at four time points representing the four experimental phases. 29 dominant ASVs with minimum 10000 reads each were further mapped to eight species (Fig. 7). Eight out of the 29 ASVs were associated with Magnetospirillum magneticum (Table S4)), which is considered a key species involved in anaerobic BTEX degradation (Shinoda et al., 2005). The ASVs classified as 'M. magneticum' were highly dominant in all reactors and their relative abundance increased after BTEX exposure (Fig. 7). Whereas following BTEX exposure the AB reactor is almost exclusively dominated by M. magneticum ASVs, the anode and cathode chambers are co-dominated by ASVs associated with Hydrogenophaga aromaticivorans. H. aromaticivorans was also documented as central BTEX degraders (Banerjee et al., 2021). ASV associated with Geobacter metallireducens, a key iron reducer reported to degrade BTEX (Zhang et al., 2012, 2014), was dominant in the anode chamber. In the cathode, we detected a high abundance of two ASVs associated with Dechloromonas aromatica - reported for its BTEX degrading activity (Chakraborty et al., 2005; Coates et al., 2001). Overall, the sequence distribution points at four dominant taxonomic groups carrying bamA (M. magneticum, H. aromaticivorans, G.metallireducens, D. aromatica). All these species were associated (≥97%) with taxonomic groups that were detected in the reactors based on 16S rRNA amplicon sequencing (Table 1, Table S4). The distribution of the taxonomic groups varied across time points and between reactors (Fig. 7). Looking downstream at the degradation process, we screened for key species involved in acetoclastic and hydrogentrophic methanogenesis (Fig. 1). Features classified to the Methanosarcinaceae and Methanobacteriaceae archaeal families were identical to 16S rRNA sequences of Methanosarcina acetivorans and Methanobacterium formicicum, respectively (Table 1) – key acetoclastic and hydrogentrophic methanogens in anaerobic digestion (Wu et al., 2022; Galagan, 2002; Meng et al., 2010; Chellapandi et al., 2018; Vítězová et al., 2020). The other key steps in the biochemical process in the reactor - fermentation and iron reduction processes (Fig. 1) cannot be systematically inferred based on taxonomy. To fully cover all steps, we screened for representatives that will capture the taxonomic composition and dynamics observed in the reactors across the experimental phases. Following filtering families with a relative abundance of at least 5% in at least a single treatment, the community was composed of 22 families and 107 genera (Fig. 5). For each family, a representative ASV was selected based on the availability of a genome sequence in the NCBI depository for a closely related homolog for the most abundant feature. Overall, 21 representatives were assigned to 21 dominant families dominating together >80% of the entire community in each reactors' chamber (Table 1). Comparing the composition of the original community (2680 unique taxa) to the composition of the set of selected ASVs, indicates that the representatives form a subset that captures the taxonomic composition and dynamics of the original community (Fig. 6). The entire selected representatives displayed temporal variations in their differential abundance pattern (Table 1). Looking at class level dynamics (Fig. 5), representatives cover Clostridia (Clostridacea, Peptostreptococcales-Tissierellales and Thermincolaceae family's representatives) and Bacilli (Bacillaceae, Erysipelotrichaceae and Acholeplasmataceae), the dominant classes (Fig. 5). As observed for their respective classes (Fig. 4), the family representatives of the Alphaproteobacteria and Gammaproteobacteria - Magnetospirillaceae, Rhodocyclaceae, Comamonadaceae, Moraxellaceae and Pseudomonadaceae - were mainly abundant in MEC bioreactors' chambers in the early phase of the experiment. Slight reduction in the relative abundance of Gammaproteobacteria in the MEC chambers is consistent with the decline in the relative abundance of H. aromaticivorans – the representative of the Comamonadaceae family, also confirmed by relative prevalence of the bamA gene (Fig. 7). Notably, the drastic decline in the overall relative abundance of Alphaproteobacteria and Gammaproteobacteria in the third phase of the experiment, after glucose addition, can be related to the reduction in BTEX degraders, all belong to the Proteobacteria – the taxonomic group to which most degraders of aromatic compounds belong to (Wu et al., 2022; Lueders, 2017; Rabus et al., 2016). Along the experiment, dynamics in the relative abundance of the Anaerolineae representative from the Anaerolineaceae family were detected in correspondence with class level pattern. Consistent with the class Thermotogae, its representative from the Kosmotogaceae family is mainly detected in the AB chamber in the fourth phase. Finally, along the experiment Methanobacteria and Metanosarcina representatives increase in relative abundance in all bioreactors' chambers (p < 2.37x10⁻¹¹) with dominance of the acetolactic representative (M. acetivorans) in the AB chamber vs dominance of the hydrogenotrophic representative (M. formicicum) in the cathode chamber. For downstream analysis, for each of the selected ASVs a respective complete genome sequence was retrieved from the NCBI depository based on blast inferred similarity. 11 of the representatives were assigned based on homology of 97–100% identity between the selected feature from the reactors and NCBI accession; 5 based on homology of 94–97% identity and 5 based on homology of 90–94%. Groups for which no homolog above the 90% similarity threshold was detected were not represented. The relative promiscuity in representative selection is supported by reports of high-level taxonomic consistency in interactions pattern (Belcour et al., 2020; Kehe et al., 2021). Based on a literature survey, we inferred that all selected representatives are known to be detected in anaerobic communities (Lueders, 2017) and classified species into functional categories in the key anaerobic biodegradation process (Fig. 1). The representative community includes four BTEX degraders inferred based on bamA functional gene sequences, 14 fermenters and iron reducers inferred based on literature survey, and two methanogens whose functionality was inferred based on taxonomy (Table 1). The focus on selected representatives allows translating the taxonomic description of the community into a functional view and examine its distribution along experimental phases (Fig. 6). Overall, the enrichment pattern in the bioreactor system was consistent with patterns expected based on the original design of the reactors (Fig. 1). The relative abundance of iron reducers was higher in the anode chamber (13% at the begging of the experiment vs 53% in its end, p = 1.05x10⁻²⁰⁴, 2.31x10⁻⁹² of Clostridiaceae and Thermincolaceae representatives respectively). G. metallireducens, an iron reducer (Lovley et al., 1993; Weber et al., 2006), was observed only in anode chamber. A high fraction of hydrogenotrophic methanogens were detected in the cathode vs high fraction of acetoclastic methanogens in the AB chamber and a lower overall abundance of methanogens in the anode (Fig. 1). To further validate the comparison of methanogen prevalence across bioreactors' chambers as inferred from representative selection, we used PICRUSt2 pipeline (Douglas et al., 2020) and compared the predicted methanogenic activity across reactors for the original, full community. The PICRUSt2 analysis further confirmed that the relative abundance of Methyl‐Coenzyme M Reductase (MCR) - a functional marker for methanogens (Friedrich, 2005) was significantly higher in AB and cathode than in anode chambers. Consistent with the patterns observed in the full community and in the representative community, a higher fraction if methanogens were detected in the AB vs cathode chamber (Fig. S3). Overall, considering taxonomy, dynamics and functionality, the selected species provided a simplified representation of the original community. The respective complete genome sequences assigned to each species (Table 1) supported the conductance of detailed functional analysis of the functionality of each member.
为突破基因组表征的描述性层面，转向图 1 所示的降解过程功能解析，我们通过聚焦参与降解过程的主要微生物类群简化了分析流程，研究范围从 BTEX 降解起始延伸至甲烷排放终止。物种筛选旨在锁定各反应器中的关键分类群，同时反映功能与分类组成特征。下游分析所选代表物种均需具备基因组已完整测序的近缘同源种。总体而言，24055 个 16S ASV 被归类至 2680 个独立分类单元，涵盖 815 属、471 科、147 纲和 53 个独立门类（图 6）。为识别反应器中潜在的 BTEX 降解菌，我们在 16S rRNA 扩增子测序基础上补充了 bamA 基因测序。bamA 基因编码 6-氧代环己-1-烯-1-甲酰辅酶 A 水解酶，该关键酶催化开环反应步骤，被公认为厌氧 BTEX 降解的生物标志物（Wu 等，2022；Kuntze 等，2008；von Netzer 等，2016）。 在所有反应器添加 BTEX 后均成功检测到 bamA 基因（图 S2）。为确定 bamA 的分类学来源，我们在代表四个实验阶段的四个时间点对三个反应器中的该基因进行了扩增。将 29 个 reads 数均超过 10000 的优势 ASVs 进一步比对到八个物种（图 7）。其中 8 个 ASVs 与磁性螺菌（Magnetospirillum magneticum）相关联（表 S4），该菌被认为是参与厌氧 BTEX 降解的关键物种（Shinoda 等，2005）。被归类为"磁性螺菌"的 ASVs 在所有反应器中均占绝对优势，且其相对丰度在 BTEX 暴露后显著增加（图 7）。值得注意的是，BTEX 暴露后 AB 反应器几乎完全由磁性螺菌 ASVs 主导，而阳极室和阴极室则共同由嗜芳烃氢噬胞菌（Hydrogenophaga aromaticivorans）相关 ASVs 主导。嗜芳烃氢噬胞菌同样被证实是核心 BTEX 降解菌（Banerjee 等，2021）。在阳极室中，金属还原地杆菌（Geobacter metallireducens）相关 ASVs 占据优势，该铁还原菌已被报道可降解 BTEX（Zhang 等，2012，2014）。 在阴极区域，我们检测到与芳香族化合物降解菌 Dechloromonas aromatica 高度同源的两个 ASV（Chakraborty 等，2005；Coates 等，2001）。总体而言，序列分布表明携带 bamA 基因的四大优势分类群（M. magneticum、H. aromaticivorans、G.metallireducens、D. aromatica）。基于 16S rRNA 扩增子测序，这些物种均与反应器中检测到的分类群具有≥97%的关联性（表 1，表 S4）。不同时间点和反应器间分类群的分布存在差异（图 7）。针对下游降解过程，我们筛选了参与乙酸裂解型和氢营养型产甲烷的关键菌种（图 1）。经鉴定的甲烷八叠球菌科（Methanosarcinaceae）和甲烷杆菌科（Methanobacteriaceae）古菌特征序列，分别与 Methanosarcina acetivorans 和 Methanobacterium formicicum 的 16S rRNA 序列完全匹配（表 1）——这两类菌是厌氧消化过程中关键的乙酸裂解型和氢营养型产甲烷菌（Wu 等，2022；Galagan，2002；Meng 等，2010；Chellapandi 等，2018；Vítězová等，2020）。 反应器内生化过程中的其他关键步骤——发酵与铁还原过程（图 1）无法单纯基于分类学进行系统推断。为全面涵盖所有步骤，我们筛选了能够反映各实验阶段反应器内分类组成与动态变化的代表性菌群。经过滤处理（保留至少在单一处理组中相对丰度≥5%的菌科）后，群落由 22 个科和 107 个属构成（图 5）。针对每个菌科，我们根据 NCBI 数据库中与最丰富特征序列高度同源基因组序列的可获取性，选取了代表性 ASV。最终 21 个代表性 ASV 被分配至 21 个优势菌科，这些菌科在各反应室中共同占据群落总量的 80%以上（表 1）。将原始群落（2680 个独特分类单元）与筛选 ASV 集的组成进行对比表明，这些代表性菌群构成的子集能准确反映原始群落的分类学组成与动态特征（图 6）。 所有选定的代表性菌群均显示出丰度差异模式的时序变化（表 1）。从纲水平动态来看（图 5），代表性菌群涵盖了优势纲——梭菌纲（梭菌科、消化链球菌目-毛螺菌目及热脱硫杆菌科的代表菌株）和芽孢杆菌纲（芽孢杆菌科、丹毒丝菌科及无胆甾原体科）。正如各纲所观察到的（图 4），α-变形菌纲和γ-变形菌纲的代表性科——磁螺菌科、红环菌科、丛毛单胞菌科、莫拉克斯菌科和假单胞菌科——主要富集于实验初期微生物电解池（MEC）生物反应器的腔室中。MEC 腔室中γ-变形菌纲相对丰度的轻微下降与丛毛单胞菌科代表菌株 H. aromaticivorans 的丰度降低趋势一致，这一现象也通过 bamA 基因的相对流行度得到证实（图 7）。 值得注意的是，在实验第三阶段添加葡萄糖后，α-变形菌纲（Alphaproteobacteria）和γ-变形菌纲（Gammaproteobacteria）总体相对丰度的急剧下降可能与 BTEX 降解菌的减少有关——这些降解菌均属于变形菌门（Proteobacteria），该分类群包含了大多数芳香族化合物降解菌（Wu 等，2022；Lueders，2017；Rabus 等，2016）。实验过程中，厌氧绳菌科（Anaerolineaceae）中厌氧绳菌纲（Anaerolineae）代表的相对丰度动态变化与纲水平模式一致。与热袍菌纲（Thermotogae）类似，其科斯莫热袍菌科（Kosmotogaceae）代表主要在第四阶段的 AB 反应室中被检出。最终，所有生物反应器腔室中甲烷杆菌目（Methanobacteria）和甲烷八叠球菌属（Metanosarcina）代表的相对丰度均呈上升趋势（p < 2.37×10 ⁻¹¹ ），其中 AB 反应室以乙酸营养型代表（M. acetivorans）为主，而阴极室则以氢营养型代表（M. formicicum）占优势。 在下游分析中，针对每个选定的 ASV（扩增子序列变体），我们基于 BLAST 推断的相似性从 NCBI 数据库检索了相应的完整基因组序列。其中 11 个代表性菌株是根据反应器样本特征序列与 NCBI 登录号之间 97-100%的相似性同源性确定的；5 个基于 94-97%相似性同源性；另有 5 个基于 90-94%相似性同源性。未检测到相似性阈值超过 90%同源序列的类群未予纳入。这种代表性选择的相对包容性得到了关于互作模式具有高阶分类一致性的研究支持（Belcour 等，2020；Kehe 等，2021）。通过文献调研，我们确认所有选定代表菌株均已知存在于厌氧微生物群落中（Lueders，2017），并将其按关键厌氧生物降解过程中的功能类别进行分类（图 1）。该代表性群落包含：基于 bamA 功能基因序列推断的 4 种 BTEX 降解菌、基于文献调研推断的 14 种发酵菌和铁还原菌，以及通过分类学推断功能的 2 种产甲烷菌（表 1）。 通过选取代表性菌种，可将群落的分类学描述转化为功能视角，并考察其在实验各阶段的分布情况（图 6）。总体而言，生物反应器系统中的富集模式与基于反应器原始设计预期的模式一致（图 1）。阳极室中铁还原菌的相对丰度较高（实验初期为 13% vs 末期达 53%，p 值分别为 1.05×10 ⁻²⁰⁴ 和 2.31×10 ⁻⁹² ，对应梭菌科和热袍菌科的代表菌种）。仅在阳极室中观察到铁还原菌 Geobacter metallireducens（Lovley 等，1993；Weber 等，2006）。阴极室检测到较高比例的氢营养型产甲烷菌，而 AB 室则以乙酸裂解型产甲烷菌为主，阳极室的产甲烷菌总体丰度较低（图 1）。为验证通过代表性菌种选择推断的生物反应器各室产甲烷菌分布比较结果，我们采用 PICRUSt2 分析流程（Douglas 等，2020），对原始完整群落的产甲烷活性预测值进行了跨反应器比较。 PICRUSt2 分析进一步证实，甲基辅酶 M 还原酶（MCR）——产甲烷菌的功能标记物（Friedrich, 2005）在 AB 室和阴极室的相对丰度显著高于阳极室。与完整群落和代表性群落中观察到的模式一致，AB 室检测到的产甲烷菌比例高于阴极室（图 S3）。总体而言，从分类学、动态变化和功能角度考量，所选物种为原始群落提供了简化表征。分配给各物种的完整基因组序列（表 1）支持对每个成员功能进行详细的功能分析。

Following an initial semi-automatic construction of the draft models, a systematic procedure of manual curation was conducted. The manual curation process focused on ensuring that the models’ performances correspond with their published physiological properties. The growth media for each species/model were curated based on a literature review (Table 2). To validate the performances of the different models we designed a simulation matrix testing biomass production and consumption and production of SCFA, H₂, CO₂ and CH₄ for each model in 25 in silico media representing control (non-feasible conditions), and selective media that are relevant to specific microbial groups as defined based on their validated physiology from the literature (Table 2). Manual curation was done throughout an iterative process. The final version of the curated models was validated to be aligned with the published experimental data regarding the nutritional requirements and physiological performances of each species (Fig. 8). As expected, none of the species could grow in non-feasible conditions where no carbon and/or nitrogen source are provided; heterotrophs could grow on a minimal medium supplemented with manually defined specific organic compounds such as carbon source and only autotrophs (methanogens) could grow on C1 carbon sources. BTEX is consumed by specific primary degraders. Growth profile of BTEX degraders on media complemented with either benzene, toluene, ethylbenzene or xylene corresponded to their literature reports. Degraders secreted SCFAs; SCFAs served the production of methane by acetolactic methanogens; CO₂ together with H₂ served the production of methane by hydrogenotrophic methanogens.
在完成初步半自动构建的模型草案后，我们进行了系统化的手动修正流程。该人工修正过程着重确保模型性能与其已发表的生理特性相符。通过文献调研（表 2），我们对每个物种/模型的生长培养基进行了校准。为验证不同模型的性能表现，我们设计了模拟矩阵测试，在 25 种计算机模拟培养基中检测各模型的生物量产出、短链脂肪酸（SCFA）代谢以及 H2、CO2 和 CH4 的生成情况，这些培养基包括对照组（不可行条件）和根据文献中已验证的生理特性为特定微生物群定制的选择性培养基（表 2）。整个修正过程采用迭代式方法完成。最终版本的修正模型经验证，其营养需求和生理性能指标均与已发表的实验数据吻合（图 8）。 正如预期，所有菌种均无法在未提供碳源和/或氮源的不可行条件下生长；异养菌只能在添加了人工定义特定有机化合物（如碳源）的基础培养基上生长，而只有自养菌（产甲烷菌）能够利用 C1 碳源生长。BTEX 由特定的初级降解菌消耗。苯、甲苯、乙苯或二甲苯补充培养基中 BTEX 降解菌的生长曲线与其文献报道一致。降解菌分泌短链脂肪酸（SCFAs）；这些 SCFAs 被乙酸营养型产甲烷菌用于甲烷生成；而 CO ₂ 与 H ₂ 则被氢营养型产甲烷菌用于甲烷合成。

By interlinking the secretion-consumption fluxes we depicted chamber-specific networks of interactions (Fig. 9). The networks point at potential scenarios for mechanisms underlying the functional divergence resulting in the different emission profile of each reactor. For example, in the AB and the cathode, acetate is the dominant predicted fermentation product of C. fungisolvens, supporting downstream acetoclastic methanogenic activity (Fig. 9). In the anode, despite the similar abundance of this fermenter, it mainly secrets carbon as CO₂ rather than acetate as in the AB and the cathode (Fig. 9). This is consistent with previous reports on the enhanced full substrate mineralization under iron-reducing conditions (Lovley et al., 2004). The shift from acetate to CO₂ secretion can be explained by the scarcity of H₂ in the anode chamber due to its transfer to the cathode chamber. Such conditions – low hydrogen environment in MEC chambers, induce a transition of certain bacteria from acetate production to acetate consumption (and CO₂ production) that subsequently leads to unfavorable conditions for the methanogens (Hasibar et al., 2020), suggesting a possible explanation for their low abundance in the anode chamber. Another example for a network-based predicted chamber-specific interaction is the production of hydrogen by M. acetivorans on the last day of the experiment in the AB chamber but not in the cathode (Fig. 9). Production of H⁺ by M. acetivorans was reported to take place in low hydrogen environments (Valentine et al., 2000), consistent with conditions in AB at the end of the experiment. An additional explanation to the favorable conditions to hydrogenotrophic vs acetoclastic methanogenesis in the cathode vs the AB chamber (beyond the design-induced high levels of H₂) can be related to the cathode-specific dominant presence of Rikenellaceae (represented by T. charontis). A typical prevalence of Rikenellaceae in cathode chambers is consistent with previous reports (Jiang et al., 2022). Considering the predicted consumption profile, Rikenellaceae are predicted to compete with the acetoclastic methanogen M. acetivorans on acetate in the cathode but not in the AB chamber where their relative abundance is low. Finally, in both MEC chambers we predict an exchange of hypoxanthine (HYXN) molecule between C. fungisolvens (anode) and G. oleilytica (cathode) (as producers) and methanogens (as consumers). Hypoxanthine – a purine derivate, was previously reported as a common microbial exchange molecule (Klitgord and Segrè, 2010). Such molecule may serve as template for purine production in anaerobic environments (Ornelas et al., 2022). In methanogens, the role of purine degradation in methane biosynthesis and energy production was explored and reported to generate a favorable proton gradient and result in the production of amino acids (DeMoll, 1998; Worrell and Nagle, 1990). In addition, hypoxanthine is a central metabolite in purine salvage in AMP and GMP production (Miller et al., 2016). As G. oleilytica (hypoxanthine source) is highly abundant only in the cathode chamber, it might further support the beneficial conditions to M. formicicum found in this chamber. In addition to the cathode-specific high abundance of two taxonomic groups (Rikenellaceae represented by T. charontis and Peptostreptococcales-Tissierellales represented by G. oleilytica, Fig. 6) that can be related to positive interactions with methanogenic archaea (competition on acetate and being a source for hypoxanthine, respectively), the cathode-specific low abundance of Tannerellaceae (represented by M. fermentans) – a potential hypoxanthine consumer, can also support the beneficial conditions in this chamber. The resource competition of two fermenters (M. fermentans and N. niacini) on hypoxanthine might also contribute to a low abundance of methanogens and subsequent methanogenic activity in the anode chamber. Hence, the network view suggests a functional role to the two non-methanogenic bacteria that are predominantly specific to the cathode – Rikenellaceae and Peptostreptococcales-Tissierellales in supporting the prevalence of hydrogenotrophic methanogen and indirectly contributing to the high methane emission either by supporting the hydrogenotrophic or competing the acetoclastic methanogens.
通过关联分泌-消耗通量，我们绘制了反应室特异的相互作用网络（图 9）。这些网络揭示了导致各反应器排放特征差异的功能分化潜在机制。例如，在厌氧反应区（AB）和阴极室，C. fungisolvens 的主要预测发酵产物是乙酸，这支持了下游乙酸裂解产甲烷活动（图 9）。而在阳极室，尽管该发酵菌的丰度相近，但其主要分泌形式为 CO ₂ 而非 AB 区和阴极室的乙酸（图 9）。这与先前关于铁还原条件下底物完全矿化增强的研究结果一致（Lovley 等，2004）。从乙酸到 CO ₂ 的分泌转变可归因于阳极室 H ₂ 的稀缺性——因其持续向阴极室转移所致。 微生物电解池（MEC）反应室中的低氢环境会促使某些细菌从乙酸生成转变为乙酸消耗（同时产生 CO ₂ ），进而导致产甲烷菌生存条件恶化（Hasibar 等，2020），这可能是阳极室中产甲烷菌丰度较低的原因。另一个基于网络预测的腔室特异性相互作用案例是：实验最后一天，M. acetivorans 在 AB 反应室中产氢，而在阴极室中未观测到该现象（图 9）。研究报道 M. acetivorans 在低氢环境中会产生 H ⁺ （Valentine 等，2000），这与实验末期 AB 反应室的环境条件相符。除设计因素导致的高浓度 H ₂ 外，阴极室相对于 AB 反应室更有利于氢营养型而非乙酸裂解型产甲烷作用的另一个原因，可能与阴极室中 Rikenellaceae（以 T. charontis 为代表）的优势定植有关。Rikenellaceae 在阴极室的典型优势定植现象与既往研究报道一致（Jiang 等，2022）。 根据预测的消耗特征，Rikenellaceae 科细菌预计会在阴极室与乙酸营养型产甲烷菌 M. acetivorans 竞争乙酸底物，但在其相对丰度较低的 AB 室中则不会形成竞争。最后，在两个微生物电解池（MEC）反应室中，我们预测存在次黄嘌呤（HYXN）分子在 C. fungisolvens（阳极）和 G. oleilytica（阴极）（作为生产者）与产甲烷菌（作为消费者）之间的交换。次黄嘌呤作为一种嘌呤衍生物，先前已被报道为微生物间常见的交换分子（Klitgord 和 Segrè，2010）。该分子可能作为厌氧环境中嘌呤合成的模板（Ornelas 等，2022）。在产甲烷菌中，嘌呤降解在甲烷生物合成和能量产生中的作用已被研究证实，其能形成有利的质子梯度并产生氨基酸（DeMoll，1998；Worrell 和 Nagle，1990）。此外，次黄嘌呤是 AMP 和 GMP 合成过程中嘌呤补救途径的核心代谢物（Miller 等，2016）。由于 G. oleilytica（次黄嘌呤来源菌）仅在阴极室具有高丰度，这可能进一步支持该反应室中 M. formicicum 所需的有利生长条件。 除阴极室特有的两类高丰度菌群（可由 T. charontis 代表的 Rikenellaceae 科和 G. oleilytica 代表的 Peptostreptococcales-Tissierellales 目，见图 6）——它们分别通过与产甲烷古菌的乙酸竞争和提供次黄嘌呤来源产生正向互作外，阴极室中潜在次黄嘌呤消耗菌 Tannerellaceae 科（以 M. fermentans 为代表）的低丰度同样为该腔室创造了有利条件。两种发酵菌（M. fermentans 和 N. niacini）对次黄嘌呤的资源竞争也可能导致阳极室产甲烷菌丰度及后续产甲烷活性较低。因此，网络分析表明两种阴极特异性非产甲烷菌——Rikenellaceae 科和 Peptostreptococcales-Tissierellales 目具有双重功能：既通过支持氢营养型产甲烷菌的生存优势，又通过竞争乙酸裂解型产甲烷菌的生态位，间接促成了高甲烷排放。