Resilience, Dynamics, and Interactions within a Model Multispecies Exoelectrogenic-Biofilm Community

The overall aims of this study were to analyze the dynamics of community composition and metabolic processes of a controlled microbial community on anode surfaces, as well as to try to unravel the possible mutual effects within the biofilm community. In the first set of experiments, the dynamics of a biofilm composed of the three model organisms, S. oneidensis, G. sulfurreducens, and G. metallireducens, was analyzed over a time course of 7 days. The three strains were added with optical density (OD) values equal to those of the microbial electrochemical cells so that the final optical density was 0.1. The OD values from similar cell numbers of different species did not correspond due to the differences in cell size among these strains. The anode potential was adjusted to 0 mV. The compositions of anode biofilms, as well as the planktonic community, were analyzed in triplicate 1, 2, 4, and 7 days following inoculation, using quantitative PCR (qPCR). G. sulfurreducens clearly dominated the sessile community beginning on the first day (Fig. 1A). This was expected, as the strategy of G. sulfurreducens cells for electron transfer to insoluble electron acceptors relies solely on a direct interaction that is mediated by conductive pili and c-type cytochromes on the surfaces of the cells and within the extracellular matrix (25). Nevertheless, neither S. oneidensis nor G. metallireducens was outcompeted from the anode during the time course of the experiment. On the contrary, we observed active growth. Results from the planktonic phase (Fig. 1B) revealed a more dynamic development of the community than in the anode biofilm. The percentage of S. oneidensis cells increased within the first 4 days to 73%, while G. metallireducens cells were almost eliminated from the planktonic phase. This might be due to oxygen contamination within the reactor, as will be stated in later parts of this report. G. sulfurreducens cells accounted for 27% of the overall community. The flavin concentration in the medium increased slightly but significantly (1.34-fold) to 37.5 nM (P = 0.023, unpaired t test, heteroscedastic) compared with that at the start of the experiments. Unexpectedly, for an unknown reason, the flavin concentration decreased when the same experiment was conducted with either one of the three individual strains alone (see Fig. S1 in the supplemental material). Hence, it is possible that flavins were used as mediators for electron transfer by the two strains found in the planktonic phase of the coculture reactors.

Results from fluorescent in situ hybridization (FISH) experiments corroborated the qPCR-based results (Fig. 2). The micrographs also displayed the different sizes of Shewanella compared with those of Geobacter cells that result in starting compositions of equal OD values but different concentrations of the individual strains. After 2 to 3 days of cultivation, the multispecies biofilm generated stable currents that reached values between 0.60 to 0.64 mA/cm². This observation indicates that the community within the BESs achieved stable rates of carbon oxidation.

In the second set of experiments, the stability of the previously described multispecies biofilm was investigated under various electrochemical process parameters. Fifteen BES reactors were started at the same time under the following applied potentials: 0.44 V, 0.24 V, 0.04 V, −0.06 V, and −0.16 V. Interestingly, the current values produced were highly similar during the time course of the experiment, when the applied potential was in a range between 0.04 and 0.44 V (Fig. 3). Lower currents were observed only if the working electrodes were poised to potentials lower than 0.04 V.

The biofilm community compositions at the completion of the experiments were similar under each of the tested conditions and were comparable to the results shown in the previous section (Fig. 4). At the end of the 7 days of incubation, G. sulfurreducens accounted for 86% to 96% of all organisms. S. oneidensis cells represented 4% to 12% of the community, while G. metallireducens was only detectable in minor quantities. In line with the measured current densities, we saw a pronounced drop in cell density when potentials lower than 0 mV were applied. Working electrodes poised to a potential of −0.06 V had >80% fewer cells than anodes adjusted to a potential of 0.04 V. In comparison, all anodes poised to positive values showed only minor differences in terms of overall cell numbers. Planktonic cells in the reactors were also quantified. Here, S. oneidensis (57 to 73%) and G. sulfurreducens (26 to 43%) comprised the majority of organisms (Fig. 4). It seemed that working electrode potentials higher than −0.06 V sustained the maximum rate of electron transfer, which correlates to the highest biomass production rates. While S. oneidensis and G. sulfurreducens responded similarly to the different potentials, we did not observe as clear of a trend for G. metallireducens. It is possible that under the chosen conditions, the growth of G. metallireducens was limited by the substrate oxidation rate rather than by electron transfer to the anode surface.

Linear sweep voltammetry was conducted to determine the minimal potential at which an exoelectrogenic multispecies biofilm transfers respiratory electrons to the working electrode. An initial applied potential of 0.04 V was linearly decreased to −0.26 V, with a sweep rate of 0.75 μV s⁻¹. Thereafter, the sweep was repeated in the reverse direction until the starting potential of 0.04 V was reached. As indicated in Fig. 5, the cells stopped producing current when the applied potential was lower than approximately −0.2 V. Half-maximal electron transfer rates were achieved with an anode potential of −0.08 V. This potential is in line with typical midpoint potentials of outer membrane cytochromes, as well as with flavins, which are released by S. oneidensis and can be used as electron transfer shuttles (27, 36, 37). Immediately after reaching −0.26 V, the experiment was repeated in the reverse potential direction. The current-over-potential diagram showed a very similar progression. The consortium stability did not seem to be highly affected by variations of electron transfer rates. This is surprising, since the very low sweep rate resulted in an overall time course for the experiment of more than 9 days for the complete sweep from 0.04 V down to −0.26 V and back up to 0.04 V. The ability to control current production by exoelectrogenic bacteria might offer promising solutions for applied processes.

Transcriptomes and metatranscriptomes were analyzed to reveal metabolic changes in the individual strains as a result of the shift from solitary to multispecies growth on anodes, thereby illuminating the observation that multispecies biofilms often produce higher current densities than axenic cultures. Twelve BES reactors were started. They contained, in triplicates, either single cultures of S. oneidensis, G. sulfurreducens, and G. metallireducens or a mixture of all three strains. The applied working electrode potential in all cases was 0.04 V. The experiment was conducted over a 7-day time period. From the second day onward, medium was continuously pumped through the reactors to reach a hydraulic retention time of 160 min. The medium contained lactate and propionate as carbon and electron sources. To provide concentrations of electron donors that were similar to those of the coculture experiments, acetate (5 mM) was added in the experiments that were conducted solely with G. sulfurreducens. Summers et al. (38) showed that wild-type G. sulfurreducens grows on lactate (this is contrary to the results in a publication by Caccavo et al. [30]). According to Summers et al., growth rates were very low and increased to 8 h only after a spontaneous point mutation. Under coculture conditions, acetate was produced by S. oneidensis as an end product of lactate oxidation.

The concentrations of the different carbon sources were assessed in the media after day 3, as the growth experiments revealed rather stable cell numbers in the second half of the experiment under coculture conditions (Fig. 6). S. oneidensis metabolized 8 mM lactate on average and produced 7 mM acetate. Interestingly, the growth of S. oneidensis led to the concomitant production of 3.8 mM propionate, which most likely resulted from the consumption of amino acids from the medium. The latter was also corroborated by the results from the transcriptome analysis. Formate was not detected in the spent medium. In single cultures, G. sulfurreducens consumed 2.1 mM of the added 5 mM acetate on average. Lactate was not consumed, while the propionate concentration increased from 5 to 5.8 mM. G. metallireducens metabolized 0.8 mM propionate, while a decrease in lactate concentration was not detectable. In cocultures, we observed a decrease in the propionate concentration of 0.6 mM, while 7.3 mM lactate was consumed. The acetate concentration was 5.6 mM. Hence, the observed lactate consumption rate was very similar to that of the pure culture experiment with S. oneidensis alone. The available acetate concentration in the coculture experiment would be 6.6 mM if it is assumed that the same ratio of catabolic lactate consumption that was observed in the S. oneidensis single culture can be applied to the coculture experiment. Even though two Geobacter strains could potentially thrive using the produced acetate, the acetate consumption was lower than in the experiment with G. sulfurreducens alone. Regarding propionate consumption, it is hard estimate to what extent G. metallireducens cells metabolize propionate in the coculture. The concentration was lower than the added 5 mM, but we cannot say to what extent S. oneidensis produces propionate under coculture conditions. Nevertheless, the expression levels of the genes involved in branched-chain amino acid metabolism were downregulated under coculture conditions in S. oneidensis.

The bioelectrochemical reactors that were used in this study were constructed from polycarbonate, which is not gas tight. Moreover, the cathode compartment was separated from the anode by a Fumapem membrane, which is a possible source of contaminating oxygen in the anode. Therefore, the oxygen concentrations in all bioelectrochemical reactors were measured at the completion of the individual experiments (see Fig. S2). Oxygen reached a concentration of 19.9 μM in the reactors operated with G. metallireducens alone. In contrast, when the reactors were operated with S. oneidensis alone, G. sulfurreducens alone, or the mixture of all three strains, the oxygen concentration reached values of 4.3, 4.2, and 4.8 μM, respectively (Fig. S2). Hence, it seems as if the abilities of S. oneidensis and G. sulfurreducens to thrive using oxygen as an electron acceptor led to a depletion of oxygen in the reactors. Interestingly, we detected G. sulfurreducens cells only under coculture conditions in the planktonic phase. As the oxygen concentrations were similar when G. sulfurreducens was grown alone or together with S. oneidensis and G. metallireducens, it is unlikely that the toxic effects due to reactive oxygen species were the only reason that planktonic cells of G. sulfurreducens were observed only under coculture conditions. The higher flavin concentrations measured in the coculture experiment might lead to more pronounced electron shuttling via this mediator and might enable the growth of G. sulfurreducens in the planktonic phase, but this needs to be investigated in further studies.

During solitary growth, each of the strains produced less current than that from coculture conditions (0.62 mA cm⁻²). The highest current production in solitary growth was recorded for G. sulfurreducens (0.47 mA cm⁻²), whereas S. oneidensis produced only around 0.13 mA cm⁻² and G. metallireducens produced 0.06 mA cm⁻².

A comparison of the transcriptomes from solitary and coculture growths revealed that 1,124 of the 3,430 protein-coding genes of G. sulfurreducens showed log2-fold changes higher than ±1 with corresponding adjusted P values (P adj) of less than 0.05. Of the total 4,214 genes of S. oneidensis, 677 followed the above-described characteristics, while in G. metallireducens, only 23 genes were detected as differentially expressed (see Table S1). This rather low number of detectable candidate genes might be due to a reduced experimental sensitivity caused by limited growth of the G. metallireducens cells compared with the other two strains and its subsequently reduced representation in the coculture transcriptomic data.

As mentioned above, a multitude of genes were differentially expressed between cells grown in the absence or presence of other organisms that thrive on the anode surface. This study focused on the central metabolic changes that result from cocultivation of the strains. The data presented suggest a number of interactions that are summarized in Fig. 7.

Generally, it seems as if the central metabolism of G. sulfurreducens is upregulated under coculture conditions (Table 1).

Outer membrane cytochromes, components of the membrane-spanning porin-cytochrome complex that transfer electrons across the outer membrane, several periplasmic and inner membrane-associated cytochromes, and the major pilus component pilA are upregulated under coculture conditions (Table 1). This indicates a generally more active transfer of electrons to the anode by G. sulfurreducens cells within the mixed species biofilm. Detailed information about the upregulated cytochromes and pilus genes can be found in the supplemental material. The hypothesis that increased electron transfer results from cocultivation is corroborated by the upregulation of genes coding for central reactions involved in the uptake (atoA) and oxidation of acetate via the citric acid cycle, as well for the NADH dehydrogenase, Nuo (Table 1; average upregulation of nuo genes, 2.9-fold). The reason for this general response might be that the hydrogen-dependent growth transcriptional regulator HgtR is downregulated. HgtR has been shown to repress genes involved in biosynthesis and energy generation (39). Although HgtR is downregulated under cocultivation conditions, the typical uptake hydrogenase, HybAB (40), of G. sulfurreducens is strongly upregulated (average upregulation of hyb genes, 17.5-fold). One explanation is that G. sulfurreducens under coculture conditions may enter a state in which it uses hydrogen (presumably provided by S. oneidensis; see below) and acetate as electron donors, whereas in pure cultures, acetate is the main carbon source. The use of hydrogen might also enable a higher proportion of acetate to be utilized in gluconeogenesis. At least two genes of gluconeogenesis were discovered to be upregulated under coculture conditions, namely, eno (2.4-fold upregulation) and pgk (2.5-fold upregulation). Interestingly, the results from the analysis of acetate concentrations in the coculture experiment suggest a lower acetate consumption rate than in the experiment with G. sulfurreducens alone. This result might corroborate the hypothesis of an increased usage of hydrogen as an electron and energy source. The transcription levels of genes necessary for nitrogen fixation strongly decreased as a result of cocultivation. This might indicate that, although the medium contains 3.7 mM ammonium, the cells encounter a nitrogen limitation within the anode biofilm in pure culture. Both Geobacter strains showed downregulations of cydA and cydB as a result of cocultivation. These encode cytochrome bd ubiquinol oxidase, which was discussed as a mechanism for coping with oxidative stress (41). The downregulation might be explained by the reduced respiration from the 20 μM oxygen available in the anode chamber under cocultivation conditions shared by S. oneidensis and G. sulfurreducens.

In addition to the downregulation of genes involved with oxidative stress, G. metallireducens showed another adaptation to growth with G. sulfurreducens and S. oneidensis, namely, an upregulation of the prp operon (Table 1). The prp genes encode the proteins necessary for propionate oxidation (42). G. metallireducens can grow with acetate and propionate. These cells apparently do not adapt their metabolism for acetate as a carbon source but rather specialize in using propionate, for which there is no competitor in the community. Propionate is a challenging substrate under anoxic conditions, since its oxidation involves an energy-dependent activation to propionyl-CoA and an ATP-dependent oxidation of succinate to fumarate if electron acceptors are used, a reduction that is menaquinone dependent. This might also explain the rather low growth rate of G. metallireducens (43).

We observed that S. oneidensis cells grown alone in the anode medium produced propionate. Although the results from the transcriptomic analysis of the S. oneidensis cells (see below) predict that propionate production will be lower in the coculture experiment, it is possible that propionate produced by S. oneidensis is directly transferred to G. metallireducens within the anode biofilm. Propionate has antimicrobial activity. The mechanism through which it exerts this activity involves the uncoupling of the cytoplasmic membrane potential as it enters the cell as a protonated molecule and then its dissociation in the cytoplasm (44). Other lines of evidence suggest that the intracellular accumulation of volatile fatty acids like propionate blocks a number of metabolic pathways and consequently arrests cell growth (45). The parallel production and consumption of propionate by S. oneidensis and G. metallireducens, respectively, might lead to an overall fitness benefit for the biofilm by reducing the concentration of this antimicrobial compound.

S. oneidensis cells also showed distinct responses to growth in a coculture biofilm (Table 1). We observed an upregulation of the Mtr pathway, which is necessary for the transport of electrons through the outer membrane to an extracellular electron acceptor (mtrABC and omcA showed 2.3-fold upregulation on average). Furthermore, protein-coding genes involved in lactate transport and oxidation were also upregulated (SO_1518 to SO_1522; >4-fold). As was observed for G. sulfurreducens, cocultivation seemed to positively affect the substrate oxidation and electron transfer processes. It is possible that oxygen contamination in the reactors played a role. Although the oxygen concentrations in the reactors with S. oneidensis alone or under coculture conditions were similar, oxygen was depleted simultaneously by G. sulfurreducens and S. oneidensis in the coculture experiment. S. oneidensis also responded to cocultivation by upregulating a hydrogenase (hypBEF, >3-fold; hyaABCD, >6-fold; see the supplemental material for further information). It was reported that hydrogen formation in S. oneidensis occurs under anoxic conditions in the absence of an electron acceptor and that hydrogen may be derived directly from pyruvate or indirectly from formate as the intermediate (46).

There are a number of gene clusters that were upregulated in cells grown in the absence of G. sulfurreducens and G. metallireducens. Very prominent were genes that are involved in degrading amino acids (the liu and ivd gene clusters). The medium contained Casitone and was therefore a source of amino acids. The catabolic consumption of amino acids, such as valine and isoleucine, by the Ivd proteins is corroborated by the observed production of propionic acid in the pure-culture experiment (47). Currently, we can only speculate that this distinct regulatory response was due to the faster depletion of the pool of available amino acids when the three organisms thrived simultaneously in the same reactor.

In this study, the interaction of three exoelectrogenic strains within BESs was analyzed. The biofilms formed on the anode surface appeared to be highly stable against fluctuations of the anode potential, which steered their respiratory activity. Although G. metallireducens plays a minor role in terms of cell size, it appears to occupy a stable ecological niche, which is likely characterized by the organism's ability to use propionate as an electron donor. S. oneidensis and G. sulfurreducens alone and together decreased the oxygen contamination in the reactor down to roughly 5 μM. This might positively affect the metabolism of the G. metallireducens biofilm. A transcriptomic analysis revealed that the interaction with other exoelectrogenic strains generally led to an upregulation in the expression of genes involved in the central metabolisms of the organisms, thus supporting the hypothesis of a positive interaction. The interaction between G. sulfurreducens and S. oneidensis is, for instance, likely displayed by the presence of a hydrogen cycle with S. oneidensis as a producer and G. sulfurreducens as a consumer. G. sulfurreducens cells were observed in the planktonic phase only in the presence of S. oneidensis. Further experiments are needed to reveal whether this was due to the use of flavins as electron shuttles by G. sulfurreducens cells. This would be similar to cheater-like behavior, as was described for extracellular signals in quorum-sensing studies (48, 49). The results presented in this report at least partially explain the observation reported multiple times in the literature that multispecies biofilms tend to be more efficient in current production than single cultures.

S. oneidensis, G. sulfurreducens, and G. metallireducens were routinely cultured anaerobically at 30°C in medium that was developed according to Coppi et al. (50) and Holmes et al. (51). The growth medium contained 0.42 g/liter KH₂PO₄, 0.22 g/liter K₂HPO₄, 0.2 g/liter NH₄Cl, 0.38 g/liter KCl, 0.36 g/liter NaCl, 0.21 g/liter MgCl₂·6H₂O, 1.8 g/liter NaHCO₃, 0.5 g/liter Na₂CO₃, 60 mg/liter CaCl₂·2H₂O, and 2 g/liter Casitone and was further complemented with 1.0 ml/liter selenite-tungstate solution (0.5 g/liter NaOH, 3 mg/liter Na₂SeO₃, 4 mg/liter Na₂WO₄·2H₂O), 10 ml/liter NB trace mineral solution (15, 50), 10 ml/liter vitamin solution (medium 141; German Type Culture Collection, DSMZ), 0.2 mM sodium ascorbate, 1.0 mM cysteine, 0.2% (wt/vol) yeast extract, and 50 mM ferric citrate as an electron acceptor. Although cysteine could theoretically act as a shuttling component, Straub and Schink indicated that it serves as a reducing agent rather than as an electron shuttle for Geobacter strains (52). The medium contained sodium lactate (12.5 mM), sodium acetate (6.25 mM), and sodium propionate (5 mM) as electron donors. These concentrations were chosen to enable electron acceptor limitation of the system. Standard anaerobic techniques were used throughout the study (53).

All anaerobic media were prepared in bottles capped with butyl rubber stoppers. To remove oxygen, the media were boiled for 10 min and flushed with a mixture of 80% N₂:20% CO₂ for 30 min. The pH was then adjusted to 7.2. The optical density during anaerobic growth on ferric citrate was measured at a wavelength of 655 nm (OD₆₅₅) to avoid an influence by ferric iron. All experiments were carried out in at least three independent replicates. Error bars represent the standard deviations.

The electrochemical reactors used in this study were described by Kloke et al. (54). All experiments were carried out in a three-electrode setup. The individual chambers of the bioelectrochemical system (BES) were separated from each other by a Fumapem F-950 membrane (Quintech, Göppingen, Germany). A saturated calomel electrode (Sensortechnik Meinsberg GmbH, Ziegra-Knobelsdorf, Germany) was used as the reference electrode. Throughout the study, all potentials were corrected for the corresponding values against a normal hydrogen electrode (NHE). Two 2.25-cm² pieces of activated carbon cloth (C-Tex 13; MAST Carbon International, Ltd., Hampshire, United Kingdom) were used as an anode and cathode. The electrodes were connected to a potentiostat (Pine Instruments, Grove City, USA) using a titanium wire. All BES experiments were carried out at 30°C.

The medium for these experiments was composed as described above but did not contain a soluble electron acceptor. It contained 12.5 mM sodium lactate and 5 mM sodium propionate as electron donors. If G. sulfurreducens was cultured alone, 5 mM sodium acetate was added to the medium, since this strain cannot grow with propionate and only grows at a slow rate with lactate. In contrast, acetate is produced from lactate by S. oneidensis, which provides a constant acetate feed under coculture conditions. In all experiments, the anode chamber contained 22.5 ml of growth medium whereas the cathode chamber and the reference electrode compartment contained 25 ml and 5 ml, respectively, of medium without an electron donor or acceptor. Throughout the entire experiment, the anode chamber was flushed with a gas mixture of 80% N₂:20% CO₂ while an aerobic cathode served as a counter electrode.

Prior to all BES experiments, the cells were washed twice in a buffer containing 0.42 g/liter KH₂PO₄, 0.22 g/liter K₂HPO₄, 0.2 g/liter NH₄Cl, and 0.38 g/liter KCl, and then resuspended to an OD₆₅₅ of 1.0 in the same buffer. The initial cell densities in the BESs were set to an OD₆₅₅ of 0.1. The anode chamber was connected to a peristaltic pump, and after the first 24 h of the experiment, fresh medium was constantly pumped through the system at flow rate of 0.14 ml/min, resulting in a hydraulic retention time of 160 min.

Genomic DNA was extracted from the planktonic cells and the anode biofilm using the innuPREP stool DNA kit (Analytic Jena, Jena, Germany). Planktonic cells were treated as suggested by the manufacturer, while biofilm samples were additionally incubated in the lysis solution at room temperature for 10 min. Cell quantities were obtained by using quantitative PCR (qPCR) with primers specific for inserted barcode regions in the genome (Table 2 and reference 55). All experiments were conducted using independent biological triplicates and three technical replicates for each biological sample. To obtain the qPCR detection limits for each strain, a standard curve was constructed based on the threshold cycle (C_T) values for a dilution series of DNA from pure cultures (1, 1 × 10⁻¹, 2 × 10⁻², 1 × 10⁻², 2 × 10⁻³, and 1 × 10⁻³ diluted with DNase-free water). The negative-control samples contained DNase-free water instead of template DNA. To determine the initial concentrations, cells were counted using a Neubauer hemocytometer with a 0.01-mm depth (Marienfeld, Lauda-Knigshofen, Germany).

The amplification was carried out using the DyNamo Flash SYBR green qPCR kit (Biozyme, Hessisch Oldendorf, Germany) according to the manufacturer's instructions. The optimal annealing temperature was determined using a temperature gradient qPCR. All qPCRs were analyzed in a CFX96 cycler (Bio-Rad, Munich, Germany). The amplification was conducted using the following program: 5 min at 95°C for predenaturation and then 34 cycles of 10 s at 95°C for denaturation, 30 s at 56°C for annealing, and 30 s 60°C for extension reactions. As a last step, a melting curve from 60°C to 95°C was used to analyze the purity of the qPCR products. The amplified DNA samples were loaded on a 2.5% agarose gel for verification of the expected fragment length. The C_T values were automatically calculated with the Bio-Rad CFX Manager software. The standard curves were obtained by plotting the C_T values against the logarithm of the concentration of each dilution of template DNA. For S. oneidensis and G. sulfurreducens, primers described by Dolch et al. (29) were used. For G. metallireducens, specific primers were designed with the software tool Geneious (Biomatters, Auckland, New Zealand). All primer sequences are listed in Table 2.

FISH was used for the simultaneous visualization, identification, and localization of each strain in the multispecies biofilm community. Samples for FISH were taken from the planktonic phase and directly from the anode surface at the completion of the experiment. Liquid samples were fixed with 4% formaldehyde for 1 h at 4°C. Afterwards, they were washed twice in phosphate-buffered saline (PBS) [2.7 mM KCl, 1.76 mM KH₂PO₄, 137 mM NaCl, 10 mM Na₂HPO₄, 9 mM (NH₄)₂SO₄, 1 mM Mg₂SO₄, 0.1 mM CaCl₂, pH 7] via centrifuging at a relative centrifugal force (RCF) of 8,000 for 2 min. The samples were then stored in a 50% (vol/vol) ethanol solution at −20°C. Complete anodes were fixed with the same concentration of formaldehyde for 4 h at 4°C. Thereafter, they were washed in PBS for 30 min and then stored at 4°C. FISH analysis was conducted as described by Dolch et al. (29) with the probes specified in Table 3. Images were collected with a Leica DM 5500 B microscope using a 63× immersion lens and a digital color camera (DFC 300 FX; Leica, Wetzlar, Germany). The filter sets used are listed in Table 4.

RNA was extracted individually from the triplicates of four experimental BES setups, which were inoculated either with pure cultures of S. oneidensis, G. sulfurreducens, or G. metallireducens or with a coculture of the three strains. Prior to the extraction, all anodes with a biofilm were additionally preincubated in 3 ml of bacterial protection solution for 10 min at room temperature (RNeasy minikit; Qiagen, Hilden, Germany). Total RNA was isolated from the anode biofilms using the RNeasy kit provided by Qiagen according to the manufacturer's instructions. Residual DNA was removed with an overnight DNase I digestion at 37°C. Further purification of RNA from DNA was conducted using the Ambion DNA-free kit (Life Technologies, Carlsbad, CA, USA). Aliquots from treated samples were subjected to PCR amplification with specific primers (Table 2) to validate the absence of genomic DNA. The Ambion MICROBExpress bacterial mRNA enrichment kit (Life Technologies) was used for rRNA depletion. Strand-specific cDNA libraries were prepared from 50 ng of rRNA-depleted RNA following the TruSeq v2 RNA protocol (Illumina). The libraries were prepared using multiplex primers to enable simultaneous sequencing on a single lane, which was performed on a HiSeq1500 system using SBS v3 kits (Illumina), generating paired-end reads of 2 × 51 nucleotides (nt) (G. sulfurreducens and S. oneidensis) or 2 × 100 nt (coculture and G. metallireducens samples). Cluster detection and base calling were performed using RTAv1.13, and the quality of reads was assessed with CASAVA v1.8.1 (Illumina).

Sequence data were clipped for the samples with 2 × 100 nt to obtain a fixed read length of 51 nt in all samples and were mapped to a database containing the merged genomes of G. sulfurreducens (NCBI GenBank accession number NC_002939.5 ), G. metallireducens (NC_007517.1 ), S. oneidensis (NC_004347.2 ), and its megaplasmid (NC_004349.1 ) using Bowtie (56). Differential expression profiles were calculated individually for each species in pure cultures versus the coculture by comparing the gene expression levels from the mapped sequences using HTSeq (57) for counting and DESeq2 (58) for differential expression analysis. An adjusted P value (P adj) of 0.05 was used as the threshold to determine statistical significance for differential gene expression.

Organic acids that were consumed or produced during the experiments were quantified using high-performance liquid chromatography (HPLC). Organic acids were quantified as described by Sturm-Richter et al. (59). Cell densities were determined via the Thermo Scientific GENESYS 20 spectrophotometer.

Oxygen was detected with the fiber-optic oxygen meter Microx-TX3 and a needle-type housing oxygen microsensor according to the manufacturers' instructions (PreSens, Germany). For 2-point calibration, an anoxic 57.5 mM sodium dithionite solution and air-saturated water were used as minimum and maximum references for oxygen saturation, respectively. The anodic oxygen concentration was measured in each reactor at the completion of the experiments.

Liquid samples of the BES were collected at different time points. The samples were centrifuged (6,000 rpm, 5 min) and filtered through a 0.2-μm-pore-size filter. Samples 200 μl in volume were transferred to a UV-transparent microtiter plate and read in a Tecan Infinite M200 PRO plate reader (Tecan, Austria) at 440 nm excitation and 525 nm emission. Different concentrations ranging from 25 nM to 2.5 μM of a mixture of flavin mononucleotide (FMN) and riboflavin (80:20) were used for establishing a standard calibration curve.

Single-nucleotide resolution RNA sequencing of cocultures and biofilms yielded between 2.3 and 37.2 million read pairs that have been deposited into the NCBI GEO database under the accession number GSE79750 .