Increasing sulfate levels show a differential impact on synthetic communities comprising different methanogens and a sulfate reducer

Methane-producing microbial communities are of ecological and biotechnological interest. Syntrophic interactions among sulfate reducers and aceto/hydrogenotrophic and obligate hydrogenotrophic methanogens form a key component of these communities, yet, the impact of these different syntrophic routes on methane production and their stability against sulfate availability are not well understood. Here, we construct model synthetic communities using a sulfate reducer and two types of methanogens representing different methanogenesis routes. We find that tri-cultures with both routes increase methane production by almost twofold compared to co-cultures and are stable in the absence of sulfate. With increasing sulfate, system stability and productivity decreases and does so faster in communities with aceto/hydrogenotrophic methanogens despite the continued presence of acetate. We show that this is due to a shift in the metabolism of these methanogens towards co-utilization of hydrogen with acetate. These findings indicate the important role of hydrogen dynamics in the stability and productivity of syntrophic communities.


Introduction
All studied habitats, ranging from human and animal guts to the soil and ocean, are found to be inhabited by microbial communities composed of hundreds of different species [1]. Interactions among these species ultimately give rise to community-level functions, including metabolic conversions that enable animal and plant nutrition [2,3], and geo-biochemical cycles [4,5]. Understanding the biochemical and physical basis, and the ecological and evolutionary drivers of functional stability in microbial communities is thus a key open challenge in microbial ecology [1]. Achieving a better understanding of these drivers for stable community function can enable prediction of functional stability and collapse thereof [6,7], the design of interference strategies to shift community function [8,9] and the engineering of bespoke 'synthetic communities' [10][11][12][13].
Towards deciphering ecological and evolutionary drivers of function and functional stability in microbial communities, methanogenic anaerobic digestion (AD) offers an ideal model system, where the production of methane from complex organic substrates can be taken as a proxy for a community function. AD communities are found in many environments including ocean and lake sediments, soil and animal guts and are used in biotechnological re-valuation of organic waste [14]. It is well known that high substrate levels and limited availability of electron acceptors in the AD system can create thermodynamic limitations that can dominate functional stability and community dynamics [15], underpin the emergence and maintenance of diversity in the community [16] and drive evolution of metabolic interactions among different species [17,18]. A key reason for the importance of thermodynamic limitations in AD systems is that it forces a cooperative (i.e. syntrophic) metabolism of organic acids, whereby degradation of these compounds by one group of organisms can only be maintained (i.e. be thermodynamically feasible) by continuous removal of end-products by another [18,19]. This syntrophic degradation can be performed by a range of fermentative microbes including sulfate-reducing bacteria (SRB), while the end-product removal can only be performed by aceto-and hydrogenotrophic methanogens, which specialize in the consumption of acetate and hydrogen, respectively [18,20]. In the case where the syntrophic degradation step is disrupted, acetate and hydrogen can accumulate, leading to further thermodynamic inhibition, as well as acidification, ultimately causing the functional collapse of the AD system [21,22].
A key syntrophic interaction in AD systems is that between SRB and methanogens. This interaction can have a versatile metabolic basis, which has been studied before in controlled co-cultures, but mostly in either the absence or excess presence of sulfate. In the absence of sulfate, for example, certain SRB can ferment organic acids such as lactate and formate to produce H 2 and acetate, which can be used by aceto-and hydrogenotrophic methanogens [23][24][25][26]. In the presence of sulfate, co-cultures of SRB and acetoclastic methanogens show H 2 consumption and production by these two groups, respectively [27,28]. With sulfate present, it is also possible that SRB can assimilate acetate [29][30][31]. Based on these documented metabolic interactions, it can be expected that different levels of sulfate can potentially cause either competitive exclusion of methanogens by SRB or cooperative interactions between the two groups. Several studies show that both aceto-and hydrogenotrophic methanogenesis can still coexist with SRB in the presence of significant concentrations of sulfate [32][33][34] and can persist or adapt to sulfate perturbations [35,36].
It is possible that changes in sulfate levels can affect the stability and type of interaction between SRB and methanogens differently, when different methanogenic groups are involved. Methanogens are distinguished into two major groups through their respiratory and energy-conserving mechanisms, and in particular, whether they contain key respiratory cytochromes or not [20,37,38]. Most hydrogenotrophic methanogens lack cytochromes and are specialized on H 2 , while acetotrophic methanogens encoding cytochromes can grow on low molecular carbons including acetate, methanol and methylamines [20]. Thus, it is possible that hydrogenotrophic methanogens are more susceptible to sulfate perturbation (compared to acetoclastic methanogens) due to competition for H 2 with SRB. It is, however, also possible that competition for H 2 with SRB affects those acetoclastic methanogens that maintain an ability for hydrogenotrophic methanogenesis [20,38,39]. These hydrogeno/acetotrophic methanogens are common in the Methanosarcina genus [20], and their H 2 cycling and utilization dynamics are studied in the model organism Methanosarcina barkeri [38,40,41]. For obligate acetoclastic methanogens, sulfate perturbation can still be problematic in the presence of sulfate reducers, as some of these can assimilate acetate [29 -31].
Given these possible competitive interactions and metabolic versatilities of the involved organisms, it is unclear if and how the productivity and stability of syntrophic interactions between SRB and hydrogenotrophic versus hydrogenotrophic/acetotrophic methanogens differ under different conditions of sulfate perturbation. To answer this question here, we use synthetic communities comprising the model sulfate reducer D. vulgaris Hildenborough, the obligate hydrogenotrophic methanogen, Methanococcus maripaludis and the hydrogenotrophic/acetotrophic methanogen, Methanosarcina barkeri. The latter species is chosen as a representative organism due to its ease of culturability and relevance in a wide range of methanogenic conditions including soils/sediments and AD systems [40,42]. D. vulgaris Hildenborough does not mineralize organic carbon substrates and can use lactate to produce acetate as an end-product [43]. We construct synthetic co-and tri-culture communities of these species and evaluate their productivity and stability under sulfate perturbations. We find that tri-culture communities produce twice the amount of methane from lactate compared to co-cultures of the sulfate reducer with a single methanogen. With increasing sulfate availability, however, we find a differential impact on the two methanogenic groups. While M. maripaludis was lost from the community at sulfate levels that only allow full respiration of the available lactate, M. barkeri was lost readily at lower sulfate levels. This differential stability was also evident at the level of productivity in the tri-culture, where the contribution from M. barkeri reduced with increasing sulfate. These results could be explained through mass balance calculations, but only if we assumed a dependency of the M. barkeri on hydrogen. We have then verified this assumption experimentally using monocultures. Together, these results show that H 2 -based competition in the presence of strong electron acceptors can influence both aceto-and hydrogenotrophic methanogens, with the former being more prone to be lost from the system as a result. These findings are of high relevance to understand complex, natural AD communities, and to further engineer synthetic communities mimicking their functionality and optimized for specific applications.

Results
To better understand the functional role and stability of syntrophic interactions between SRB and methanogens in AD communities, we created here a set of synthetic microbial communities composed of two and three species. We used three key species to represent the roles of SRB (Desulfovibrio vulgaris Hildenborough; Dv), and hydrogenotrophic/acetotrophic (Methanosarcina barkeri; Mb) and hydrogenotrophic methanogens (Methanococcus maripaludis; Mm). The Dv-Mm pair has emerged in recent years as a model system to study syntrophic interactions [44] and was recently shown to be enabled by polymorphisms found in Dv [45]. Mb is one of the most well-studied methanogens capable of hydrogenotrophic/ acetotrophic methanogenesis and can be more abundant in AD systems compared with obligate acetotrophic methanogens [40,42,46]. We cultivated these organisms and created relevant synthetic communities composed of one, two and three species (see Material and methods). We initiated replicate synthetic communities using a chemically defined media with lactate (30 mM), as the sole organic carbon source, and cultivated them under different levels of sulfate (see Material and methods). Each constructed community was incubated, and sub-cultured twice, over three-week periods. These conditions mimicked a low-flow, chemostat-like system, while different levels of sulfate mimicked different availability of strong electron acceptors.
royalsocietypublishing.org/journal/rsif J. R. Soc. Interface 16: 20190129 2.1. All species coexist and community productivity increases in the absence of sulfate The presence of both methanogenesis routes through acetoand hydrogenotrophic species is expected to increase the productivity in AD communities due to a more complete conversion of the key fermentation products from sulfate reducers (figure 1a). We found this expectation to be fulfilled in the absence of sulfate; the synthetic Dv-Mm-Mb tri-culture produced close to twofold more methane compared with the Dv-Mm and Dv-Mb co-cultures (figure 1b). The tri-culture and the Dv-Mm co-culture achieved stable methane levels over three sub-cultures, while methane production in the Dv-Mb co-culture was highly variable. In line with these observations, the tri-culture and the Dv-Mm system displayed full lactate conversion, while there was significant lactate remaining in one replicate Dv-Mb system (figure 1c). Interestingly, both the tri-culture and the Dv-Mb co-culture displayed also significant levels of residual acetate (around 7.0-16.0 mM), which was well above the value expected (less than 0.5 mM) from the estimated half saturation coefficient of Mb for acetate (K ¼ 4.5-5.0 mM) [47,48]. Thus, Mb was not able to consume all of the acetate fermented by Dv ( figure 1d). This finding was replicated when we cultivated the cultures under a five-week sub-culturing regime (electronic supplementary material, figure S1), suggesting that lack of expected acetate consumption is not simply due to slow growth of Mb on this substrate.

2.2.
Increased sulfate availability shows a differential impact on the maintenance and productivity of Mb versus Mm In order to find out the impact of sulfate availability on the stable coexistence of Dv and different methanogens, we further analysed the dynamics of each co-culture and the tri-culture at different sulfate levels. In particular, we cultivated communities in sulfate concentrations that provide either half or full stoichiometric equivalence to lactate; i.e. 7.5 or 15 mM sulfate allowing either half or full respiration of lactate by Dv (these conditions are referred to as 'half-' and 'full-sulfate' from now on). We found that increased sulfate availability immediately impacted the Dv -Mb coculture and resulted in a loss of methane production already in half-sulfate treatments (figure 2). The Dv-Mm co-culture displayed stable coexistence at half-sulfate treatments, but methanogenesis was clearly showing a diminishing trend in the full-sulfate treatment (figure 2). Methanogenesis under increasing sulfate levels in the synthetic tri-culture behaved qualitatively similarly to the Dv -Mm co-culture, but methane royalsocietypublishing.org/journal/rsif J. R. Soc. Interface 16: 20190129 levels in the tri-culture during each culturing period were slightly higher (figure 2). We found that the impact on methane production by switching from individual co-cultures to a tri-culture also depends on sulfate availability (compare figure 1b and figure 2). In particular, we found that going from Dv-Mm co-cultures to tri-cultures, in the absence of sulfate, increased methane production by almost 100% by the end of the third three-week incubation. Instead, the same comparison shows only a 16.58% increase under the half-sulfate treatment. This suggests that Mb populations are either diminishing under the half-sulfate treatment or are not receiving enough acetate. We excluded the latter possibility by measuring lactate and acetate levels for all co-cultures and the tri-culture, and under each sulfate treatment (electronic supplementary material, figure S2). This showed that there are significant levels of acetate in the tri-culture under half-sulfate treatment (as well as full-sulfate treatment), suggesting that the observed smaller increase in productivity (from co-to tri-cultures) compared to the no-sulfate case is not due to acetate limitation.
To further corroborate these findings, we analysed community stability at the species level by enumerating the different populations using quantitative PCR (qPCR) of the targeted species gene copies at the end of the overall experiment (see Material and methods). In general, Dv populations accounted for a large fraction (greater than 80%) of the overall community in all treatments and displayed an increasing trend with sulfate addition ( figure 3a). An opposite trend is observed for the population sizes of Mm and Mb, as expected from the observed decrease in methane production. The Mb abundance showed high variance in most cases, except for the tri-culture with no sulfate, while Mm populations showed an increase in triculture (for all distinct sulfate treatments) compared with the same sulfate level co-culture (figure 3b). Taken together, these findings show that in the presented system, an increase in community complexity (i.e. extended syntrophic interactions) results in an increased stability of methanogen populations both under sulfate perturbation and without sulfate, and a lower stability of Mb populations compared to Mm, as sulfate becomes available.

Mb populations productivity from acetate shows a significant dependence on H 2
Why can the acetotrophic Mb contribute to methane production under no-sulfate treatment, but not under half-and full-sulfate treatments, even though there is enough acetate available for it to grow? As shown above, Dv contributes to a higher fraction of the population with increasing sulfate and can use H 2 , as well as lactate, under this condition [30]. This creates a competitive situation for Mm, but possibly also for Mb, if it relies also on H 2 for maintaining its population size. Indeed, we observed H 2 utilization by Mb both in control monocultures, with lactate as the sole carbon source (electronic supplementary material, figure S3), as well as in two replicates in the final sub-culturing of the Dv-Mb co-cultures under no-sulfate treatment (electronic supplementary material, figure S4). These observations, as well as previous indications of H 2 utilization of Mb [20,23,38,39,41], prompted us to more directly test the impact of H 2 on the growth of Mb with acetate, using its      The consumption of these substrates by Mb can proceed theoretically through aceto-and hydrogenotrophic pathways (reactions 1 and 2 in table 2), and their possible combination through H 2 oxidation with acetate reduction. If we use H 2Mb and Acetate Mb as given constraints, we can show that the theoretical overall methane output (CH 4calc. ) would always be equal to H 2Mb /4 þ Acetate Mb (see Material and methods). We find that the observed methane in the system (CH 4obs. ) was almost always below this theoretical maximum ( table 1). There were, however, two cases that result in more methane than theoretically possible, by 1% and 10% more. We find that these two cases present the lowest acetate consumption (no detectable consumption in the second case), and the highest H 2 consumption, indicating significant H 2 consumption by Mb to produce methane through reaction 1 (and possibly also a combination of reactions 1 and 2). This might have altered Dv's metabolism to shift from acetate fermentation into H 2 production [45,50] and/or its investment of reductive power into Table 1. Observed and calculated substrate levels and per cent consumption and production in the Desulfovibrio vulgaris (Dv)-Methanosarcina barkeri (Mb) coculture without sulfate. Observed lactate levels are obtained from initial and residual levels of this compound in the system, while observed CH 4 is that measured at the end of three-week cultivation period. Theoretical maximum of H 2 and acetate that are utilized by Mb (columns 4 and 5) is calculated from the theoretical amount available from an assumed full conversion by Dv through fermentation (i.e. reaction 5 in table 2), adjusted by the observed residual level of acetate and H 2 changes compared with its levels in the beginning in the system. Mb's consumption of these substrates and conversion into CH 4 (column 6) is based on the assumption of it utilizing reactions 1 -2, given in table 2 (see Material and methods and main text). The per cent production of CH 4 as that of possible maximum (column 7) is based on this and the observed CH 4 (column 3). Finally, Mb's per cent utilization of acetate (column 8) is based on the full conversion of lactate (shown on column 2) and theoretically available to Mb (column 5), based on observed residual acetate. The unit for chemicals is mM for organic acids and mmoles per l medium for gases.
royalsocietypublishing.org/journal/rsif J. R. Soc. Interface 16: 20190129 biomass production, which could explain the discrepancy with our theoretical calculation based on reaction 5. Overall, our results summarized in table 1 show that the methane production in the system cannot be explained solely by acetotrophic methanogenesis but requires involvement from reactions 1 and 2, or their combination. Note that this general conclusion would not be affected by possible investment into biomass by Dv or Mb, which we neglected in the calculations shown in table 1. Moreover, methane production as a percentage of the theoretical maximum (as calculated above) increases over the course of the three sub-culturing periods, while acetate consumption decreases (table 1). In other words, Mb seems to be shifting its metabolism in the presence of Dv in a way favouring increasingly H 2 (co)utilization. This trend, in turn, could explain the instability of Mb in the co-and tri-cultures under increasing sulfate conditions, where competition for H 2 would be higher (due to utilization both by Dv and Mm).

Discussion
We have developed here a set of co-and tri-cultures comprising three key functional populations found in AD systems, a sulfate reducer (Dv) and aceto-(Mb) and hydrogenotrophic (Mm) methanogens. These systems allowed us to study the syntrophic interactions among these species under a common ecological perturbation in the form of sulfate availability. Our results showed an increased productivity, in the form of methane production, and high stability, through species coexistence, in the tri-cultures with no sulfate addition. With an increasing availability of sulfate, the shift in Dv metabolism towards respiration created a disruption in the methanogen populations, and under non-limiting sulfate concentrations, we found both hydrogenotrophic/acetotrophic and hydrogenotrophic methanogenesis showing a strongly diminishing trend. At limiting levels of sulfate, the disruption to coexistence was also limited, but we found a differentially stronger impact on hydrogenotrophic/acetotrophic populations represented by Mb. Experiments on the monoculture of this species verified the strong influence of H 2 on its growth with acetate, suggesting that its observed instability in tri-and co-cultures could be due to competition with Dv and Mm for this compound.
Perturbation of methanogenic populations due to competition for H 2 with SRB has been postulated and studied in several complex communities [32][33][34][35][36]. The presented study, with its well-defined, simplified synthetic communities, provides a direct observation of this competition, and instability of methanogens, in the presence of a sulfate reducer and sulfate as an electron acceptor. More importantly, these synthetic communities reveal that hydrogenotrophic/acetotrophic methanogens are more prone to suffer from such sulfateinflicted instability despite their ability to use acetate. It would be very interesting to further evaluate this finding in the context of complex AD communities found in nature and in bioreactors. In particular, there is some evidence from the latter systems that hydrogen supplementation can lead to higher methane production [51] which, according to our findings, could be due to a reduction in the competition for H 2 and enhanced productivity (and possibly growth) of hydrogenotrophic/ acetotrophic methanogens.
The synthetic community approach presented here can and should be extended to other combinations of species. In particular, we note that while Mb is capable of growth on acetate, there are several other methanogens in nature that seem to have become obligate growers on this substrate, including those from the genus Methanotrix (formerly Methanosaeta) [37]. It would be very interesting to assess the stability of these obligate acetotrophic methanogens against SRB and hydrogenotrophic methanogens. To this end, a representative species (Methanosaeta concilii) from this functional group has already been studied using a synthetic community approach [28], resulting in the identification of both competitive and cooperative interactions with Dv and Mm. The biochemical underpinning of these interactions, both in that study and the current one, is the flexibility and efficiency of energy conservation mechanisms found in the methanogens [20]. Recent studies have shown that the ability to encode different cytochromes and hydrogenases allows Mb (and other methanogens encoding cytochromes) to channel electrons resulting from both the oxidation of one-carbon molecules and H 2 into the reduction of the key heterodisulfide CoM-S-S-CoB [37]. The resulting electron flow scheme allows Mb to perform both aceto-and hydrogenotrophic methanogenesis with higher ATP yield but causes a higher H 2 pressure requirement for the latter process compared to obligate hydrogenotrophic methanogens [20]. In addition, acetate and one-carbon consumption under this electron flow scheme are suggested to involve H 2 cycling, whereby H 2 is generated in the cytosol to then diffuse out of the cell and be re-used at membranebound hydrogenases [38]. Both its high H 2 requirement for hydrogenotrophic methanogenesis, and its possible reliance on H 2 cycling for aceticlastic methanogenesis, makes Mb vulnerable to ecological perturbances as we have shown here.
In this biochemical context, it would be very interesting to see if Mb can adapt to co-culturing with Dv under a nosulfate regime and become more tolerant to sulfate-based perturbances. We observe some indication of such possibility, where some of the Dv-Mb replicates shifted to significant H 2 consumption and produced high levels of methane under the no-sulfate treatment. In these cases, we also observe a higher acetate residual (figure 1; electronic supplementary material, figure S4), in line with a previous study of a Dv-Mb coculture under sulfate free conditions, where it was suggested that the presence of Dv inhibits acetate utilization by Mb [23]. Based on our mass balance calculation, however, the observed acetate residual could be explained by a complete switch of Mb metabolism to H 2 oxidation with acetate reduction, as shown in reaction 6. In this scenario, the production of 1 mole of methane only requires 0.5 moles of acetate, fully explaining the observed mass balance in some of the cultures (table 1, figure 1; electronic supplementary material, figure S4). In other words, the acetate consumption by Mb would be lower to obtain the same energy yield per reaction if Mb's metabolic pathway follows reaction 6 (as presented in Material and methods, §4.8). Even in the case of half-sulfate treatment, we found high variance in the Dv-Mb co-cultures in terms of productivity, indicating the ability of Mb to use H 2 . It would be interesting to further evaluate this possibility of Mb's adaptation into a hydrogenotrophic (H 2 /CO 2 ) or mixotrophic (H 2 /Acetate) metabolism, and whether the newly identified electron bifurcation mechanisms in hydrogenotrophic methanogenesis pathways of Mm [52] could also be present in Mb or other hydrogenotrophic/acetotrophic methanogens. While our combination of Dv, Mm and Mb is not a naturally occurring one and these species have not necessarily royalsocietypublishing.org/journal/rsif J. R. Soc. Interface 16: 20190129 undergone coevolution (except throughout these experiments), there is now increasing evidence that the interplay of evolutionary and ecological dynamics is important for the emergence and stability of microbial interactions [53]. For example, recent community coalescence studies find the dominance of entire AD communities over others [54], suggesting co-adaptation among community members being a key drive of productivity and stability. Supporting this view, enriched AD communities are shown to display additional metabolic interactions ( particularly auxotrophic interactions) on top of syntrophic interactions [17]. Evolutionary adaptations are also seen in the Dv-Mm co-culture used here; both species are found to accumulate beneficial mutations when co-evolved in the absence of sulfate [55], and Dv populations are found to harbour polymorphisms that directly influence the ability to form a syntrophic interaction with Mm [45]. Thus, natural communities might display evolutionary adaptations that render them more resilient to perturbations than our synthetic systems and might display auxiliary interactions on top of the metabolic syntrophies and cross-feeding interactions that we observed here.
Besides their value as experimental hypothesis-generating tools, synthetic communities are also suggested to have potential as specific biotechnological applications [1]. To this end, the co-and tri-cultures presented here can be further expanded with additional functional groups of microbes to attain biotechnologically relevant conversions. It has been suggested, for example, that energy-limited systems presenting thermodynamically driven syntrophic interactions, as well as cross-feeding can provide enhanced productivity compared to monoculture-based bioproduction [56]. Certain chemical conversions and degradations of complex biomaterials, such as cellulose, cannot be achieved by monocultures, and for the evaluation of these compounds, a synthetic community approach, as presented here, will be necessary. Therefore, it would be interesting to expand the tri-culture presented here with primary degraders to allow conversion of complex sugars into methane, as already attempted for cellulose [57]. We advocate the combined use of ecological, evolutionary and engineering approaches to the development and further engineering of such synthetic communities, to achieve robust new biotechnological applications and more representative model ecosystems.

Strains used
Desulfovibrio vulgaris Hildenborough (DSM644, Dv-WT), Methanosarcina barkeri (DSM800, Mb) and Methanococcus maripaludis S2 (DSM2067, Mm) were originally ordered from the public strain centre DSMZ (www.dsmz.de). The particular Desulfovibrio vulgaris Hildenborough strain (referred to as 'Dv' in this text) used in the present work was previously isolated in our laboratory and contains two key genetic polymorphisms that allow it to grow syntrophically with Methanococcus maripaludis without sulfate [45].

Growth media
A defined anaerobic medium, adapted from previous studies [45,50], is used to grow Dv, Mm and Mb in co-and tri-culture. This medium is created by mixing basal salt, trace metal and vitamin stock solutions in appropriate volumes (as explained below).
Different carbon and electron acceptor sources were then added to this main media composition, according to the culture used. For the co-and tri-cultures, 30 mM Na-lactate was added as the carbon source, and Na 2 SO 4 was added at different levels of 0, 7.5 and 15 mM as described in the main text. For Dv monocultures, 30 mM Na-lactate and 10 mM Na 2 SO 4 were added. For Mb monocultures, 100 mM Na-acetate was added, and for Mm monocultures, 10 mM Na-pyruvate and 682 mM NaCl were added. Furthermore, the Mm monoculture headspace was replaced with 2 bar 80%H 2 -20%CO 2 .
All media were prepared anaerobically. First, 10 ml of the 100Â trace element stock solution was added to 1Â concentrated 1 l basal salt media (with carbon and electron acceptor sources added as explained above). To this, 1 ml Resazurin stock (1 g l 21 ) solution was added, to act as an oxygen indicator. The resulting media was degassed in batches of 200 ml. Each batch was brought to the boiling point in a 500 ml conical flask and then maintained at 808C under a continuous flow of gas (80% N 2 þ 20% CO 2 ) at a flow rate of 0.5 LPM. The gassing line was a blue cannula (0.6 mm ID, Microlance, Beckton Dickinson, Franklin Lakes, NJ, USA) equipped with a sterile filter (Minisart, Sartorius, Göttingen, Germany). After 5 min degassing, 0.2 ml of 1000Â vitamin stock solution was added to the (200 ml) medium. To this, 2 ml of cysteine-HCl stock solution (0.2 M) was added to create a reductive environment. The media is then degassed for another hour (at the same flow rate of gas) while being stirred. The removal of oxygen was verified by the Resazurin colour-shift from pink (and occasionally by a redox measurement). All gases (BOC, UK) used for degassing are run through an oxygen scrubber column (Oxisorb, MG Industries, Bad Soden, Germany), to remove any residual oxygen.
After degassing, media were transferred into an anaerobic chamber station (MG 500, Don Whitley). This chamber is maintained according to the manufacture's instruction using N 2 , CO 2 and H 2 supplies with an actual gas fraction of 3.14% H 2 and 5.32% CO 2 , as measured by micro-gas chromatography (GC) (Agilent 490 micro-GC, Agilent Technologies). Before use, any empty culture tubes, rubber stoppers and other tools (i.e. glass baker, electronic dispenser (Eppendorf multipette E3x, Germany) and adaptor (Eppendorf Combtips advanced, Germany)) were degassed for at least 24 h in the anaerobic chamber to exclude any O 2 contamination. Within the chamber, culture tubes (Hungate anaerobic culture tubes, Chemglass Life Sciences, Vineland, NJ, USA) were filled with 5 ml media. They were then immediately sealed with a blue butyl rubber stopper (Chemglass Life Sciences, Vineland, NJ, USA), transferred out of the chamber, and crimp sealed with aluminium crimp caps (Scientific Laboratory Supplies, Nottingham, UK). Tubes containing the media were autoclaved for 15 min at 1218C in a desktop autoclave (ST royalsocietypublishing.org/journal/rsif J. R. Soc. Interface 16: 20190129 19T, Dixon, Wickford, UK). Before inoculation, a further 0.1 ml of 50Â concentrated Na 2 S stock solution was added into the medium to achieve a final concentration of 2 mM Na 2 S. Cultures were then inoculated into such prepared tubes.

Experimental design
Co-cultures of Dv -Mb and Dv-Mm and tri-cultures of Dv -Mb-Mm were constructed as shown in figure 5a and tested for the methane production in three batches of cultivation, each of three weeks duration. For co-and tri-culture communities, three treatments of 0, 7.5 and 15 mM sulfate were used, with the latter two treatments corresponding to the half and full theoretical amount required to respire 30 mM lactate (table 1 and figure 5a). The cultivations were performed in triplicate, with each set incubated at 378C for three weeks. The dilution level for sub-culturing was 5% (v/v). In addition, a single round of five weeks' incubation of co-cultures and tri-culture was also conducted. Individual monocultures were also incubated in the same way, to serve as live controls.
The construction of co-and tri-cultures was done using the inoculum from individual monocultures. Dv, Mb and Mm were cultivated until late lag phase for 4, 21 and 7 days, respectively, before inoculation into mixed cultures. For each multi-species culture, 200 ml of individual monocultures are inoculated into 5 ml medium (i.e. 4% v/v).
To test the ability of Dv, Mb and Mm to grow on lactate under the same conditions as co-and tri-cultures, individual monocultures of each strain were incubated with the same medium as those cultures (and under same headspace conditions); 30 mM Na-lactate as carbon source, 7.5 mM Na-sulfate and headspace air fraction same with the anaerobic chamber air.

Gas measurements
Prior to any sampling, overall headspace pressure was measured using a needle pressure gauge (ASHCROFT 310, USA) at the beginning and end of each culturing period (figure 5b). The measured pressure was corrected for dead volume of the measurement device, by performing a two-point pressure measurement. The headspace gas composition was analysed using a micro-GC  (equipped with a micro thermal conductivity detector and Molsieve 5A and PoraPlot 10 m columns, Agilent 490, Agilent Technologies) at the end of each culturing period. To sample the headspace, a gas-tight glass syringe (Cadence Science, Inc., Italy) was connected to an inert gas sampling syringe valve (Hamilton, USA), a hydrophilic 0.2 mm sterile filter (Sartorius Stedim Biotech GmbH, Gottingen, Germany) and a 23 G sterile needle (Becton Dickinson, S.A., Fraga, Spain). At first, 1 ml gas was extracted from the headspace for filling any dead volume in the sampling system. Then, 2 ml gas was extracted and supplied to micro-GC sampling loop, equipped with a Genie 170 membrane separator (A þ Corporation, LLC, LA, USA) to exclude any water contamination. The running parameters for micro-GC analysis were: 1008C column temperature, initial column pressure of 175 kPa (static pressure mode) and 1008C injector temperature for two channels. The injection time for the Molsieve column was 40 ms with 9 s backflush time, and that for the PoraPlot column was 100 ms without backflush.

Culture sampling and pH measurement
At the end of every three weeks culturing period, 1.5 ml culture was extracted using a 1 ml syringe inside the anaerobic chamber and centrifuged at 5500 rpm for 3 min. The biomass and supernatant were separated and stored at 2208C for further DNA extraction and ion chromatography (IC) analyses. After sampling, culture tubes were opened and the residual culture (approx. 3 ml) was pooled out for pH measurement (Mettler Toledo M300, Columbus, Ohio, USA).

Optical density and metabolite measurements
Optical density (OD) of the cultures at 600 nm was measured on a daily basis using a spectrophotometer (Spectronic 200E, Thermo Scientific). Each tube was vortexed for 5 s before each OD measurement using a vortex mixer (Stuart SA8, Stone, UK). Lactate, acetate, pyruvate and sulfate were measured using an IC (Dionex ICS-5000 þ DP, Thermo Scientific) equipped with a conductivity-based detector and supplied with MilliQ water (R . 18.2 V, measured and prepared using an Alto Ultrapure Water System, TripleRed Ltd, Buckinghamshire, UK) for eluent generation. Collected samples were prepared for measurement by centrifuging cell pellets down (5000 rpm for 3.0 min). After separating from the cell pellets, culture supernatants were filtered through a 0.22 mm pore nylon membrane using a Costar Spin-X centrifuge tube filter (Corning Inc., Salt Lake City, USA). The resulting samples were diluted 10 or 100 times (by 10 times dilution series) using MilliQ water (R . 18.2 V). Each sample (500 ml) was placed into specific IC sampling vials (cat no. 079812, Thermo Fisher Scientific) for analysis, and the IC was run with a sampling size set at 2.5 ml. The autosampler was primed at the beginning of each round of IC analysis according to the equipment manual and using a wash-rinse fluidic cycle. An analytical anion column (Dionex IonPac TM AS11-HC, Thermo Scientific, USA) with 4 mm ion exchange matrix beads was used according to the following separation conditions: 0.38 ml min 21 flow rate, 4300 psi pressure and 308C column temperature. The used eluent was KOH, applied over 37 min with the following gradient profile: 1.5 mM for 27-0 min (pre-run for equilibration), 1.5 mM for 0-8 min (isocratic), increased to 15 mM for 8-18 min, increased to 24 mM for 18-23 min, increased to 60 mM for 23-24 min and stayed at 60 mM for 24-30 min.

DNA isolation, PCR and qPCR
DNeasy Power Soil Kit (QIAGEN, Germany) was used for isolating genomic DNA according to the manufacturer's instructions. This genomic DNA isolation kit was formerly sold by MO BIO as PowerSoil DNA Isolation Kit and used for isolating DNA from bacterial-archaeal co-cultures [28]. Genomic DNA was quantified using NanoDrop spectrophotometer (N60, IMPLEN) and stored at 220 for further analyses.
Specific primers were designed for targeting dsvA gene of Desulfovibrio vulgaris Hildenborough (IMG gene ID: 637121620), mtaB gene of Methanosarcina barkeri (IMG gene ID: 637699633) and coenzyme F420 hydrogenase of Methanococcus maripaludis (IMG gene ID: 2563556008). The specificity of the developed primers was tested and verified by amplifying the DNAs from monocultures of Dv, Mb and Mm using the polymerase chain reaction (PCR).
PCR mixtures (in a total volume of 50 ml distilled water) contained 1 ml of 10 mM dNTPs (Bio Lab, USA), 4 ml of 25 mM MgCl 2 (Promega, USA), 2 ml of forward primer (10 mM), 2 ml of 10 mM reverse primer, 10 -20 ng template DNA, 10 ml of GoTaq Flexi Buffer (Promega, USA), 2 ml of 4 mg ml 21 bovine serum albumin (Bio Lab, USA) and 0.25 ml of 5 U ml 21 GoTaq G2 Flexi polymerase (Promega, USA). PCR mixtures were prepared in bulk volume each time (greater than 500 ml), to minimize preparation errors, and the working volume per sample was 25 ml.
PCR was conducted using a 96-well thermal cycler (Veriti, Applied Biosystems) with the following settings: 958C for 5 min, 35 cycles of 958C for 30 s, an annealing temperature of 608C for 30 s, followed by 728C for 1 min, and finally 728C for 10 min. All PCR products were electrophoresed in TAE buffer on 1.0% Hi-Res standard agarose gels (AGTC Bioproducts, UK) with 0.01% GelRed nucleic acid stain (BIOTIUM 10,000X, Hayward CA, USA). DNA band in the gels was visualized by a gel imaging system (U Genius 3, SYNGENE). Agilent Technologies Stratagene Mx3005P real-time PCR system and SYBR Green JumpStart Taq ReadyMix were applied (Sigma-Aldrich, USA) for qPCR analysis.
The genomic sequence lengths excluding plasmids (bp) were retrieved from NCBI for the use in the present work, which are 3570858 bp (NCBI ID: ASM19575v1), 4 533 209 bp (NCBI ID: ASM97002v1), and 1 746 697 bp (NCBI ID: ASM22064v1) for Dv, Mb and Mm, respectively. A standard DNA template for each strain was diluted using sterile water (10-fold dilution series) and tested with the unknown samples in one single qPCR run to generate a standard curve. Each standard sample and replicate in the above experimental design was tested in triplicate under qPCR assay with internal reference dye mode (ROX). The correlation coefficients (R 2 ) of the standard curves were 0.9987 (Dv), 0.9973 (Mb) and 0.9999 (Mm), and the qPCR efficiencies were 96.1% (Dv), 96.8% (Mb) and 94.4% (Mm).

Mass balance calculations
We performed mass balance calculations based on the assumption that Methanosarcina barkeri (Mb) and Desulfovibrio vulgaris Hildenborough (Dv) use only the compounded overall reactions 1 -2 and 3-5, shown in table 2, respectively. It is also possible that Mb might combine reactions 1 and 2 so to couple acetate reduction with H 2 oxidation; To calculate total methane production in the closed system, we first estimate the amount of acetate and H 2 available to Mb. These compounds can only be produced by Dv, through its fermentation pathway, i.e. in reaction 5 from : ð4:2Þ We can now use these values to calculate the estimated stoichiometric, theoretical methane production ([CH 4 ] calc ) by Mb, through reactions 1, 2 and 6. The actual amounts of acetate used in reactions 2 and 6, as well as the actual amounts of H 2 used in reactions 1 and 6, are unknown. If we assume a full conversion through the three reactions, we would have the following stoichiometric balances: