Syntrophomonas wolfei G311 (DSM102351), S. lipocalidus TGB-C1 (DSM12680), and M. maripaludis S2 (DSM14266) were purchased from German culture collection DSMZ (Braunschweig, Germany). M. conradii strain HZ254^T was isolated and was routinely maintained in our lab as described previously (Lu and Lu, 2012). The pure cultures of S. wolfei and S. lipocalidus were cultivated in medium containing 20 mM of sodium crotonate as described (Liu et al., 2014; Zhang et al., 2018). M. maripaludis was cultivated in a modified DSMZ141 medium containing 100 mM of NaCl, 7.87 mM of MgCl₂⋅6H₂O, and 0.007 mM of Fe(NH₄)₂(SO₄)₂ under 170 kPa of H₂/CO₂ (80:20, v/v) (Zhang et al., 2018).

Two defined cocultures, one pairing S. wolfei with M. maripaludis (hereafter refer to the WM coculture, which is mesophilic) and the other pairing S. lipocalidus with M. conradii (hereafter refer to the LC coculture, which is thermophilic), were constructed. The cocultures were prepared by mixing different volumes of the log phase S. wolfei or S. lipocalidus with the vigorously growing M. maripaludis or M. conradii in the medium of the respective methanogens. The headspace of 250-ml incubation bottles was flushed with N₂ for 5 min and then adjusted with N₂:CO₂ (80:20 [v/v]) to a final pressure of 172 kPa. All the bottles were sealed with butyl stoppers and crimped aluminum caps. Cultivation was carried out at 35°C for the mesophilic WM coculture and at 55°C for the thermophilic LC coculture under dark without agitation. The sodium butyrate was added as sole substrate to a final concentration of 10 mM. To create the treatments for different start conditions, the volumes of the syntrophs and methanogens for inoculation were adjusted so that high, medium, and low ratios of syntroph versus methanogen cells were established at the beginning. The exact ratios were different between the two defined cocultures, which was based on the pre-experiment growth of the cocultures.

Gasses were sampled with a Pressure-Lok syringe (VICI, Houston, TX, United States); and the concentration of methane and hydrogen was determined using a gas chromatograph (7890A, Agilent, Santa Clara, CA, United States) equipped with an 80/100-mesh Porapak Q column (Supelco; Sigma-Aldrich, St. Louis, MO, United States) and a thermal conductivity detector (Liu et al., 2014). The unit of CH₄ concentration was converted from partial pressure in headspace to mmol L^–1 in liquid medium by using Avogadro’s law. Liquid samples (0.5 ml) were collected periodically with a sterile syringe, centrifuged, and filtered through 0.22-μm filters. Concentrations of butyrate and acetate in culture medium were determined by high-performance liquid chromatography with a ZORBAX SB-Aq C18 organic acid column (250 by 4.6 mm; particle size 5 μm; Agilent) at a flow rate of 0.8 ml min^–1. The UV absorbance detector was set at 210 nm (Zhang et al., 2018).

qPCR was used to estimate the growth of syntrophs and methanogens during incubation. Cells were harvested periodically from the syntrophic cocultures. The cell suspensions were centrifuged at 20,817 × g at 4°C for 8–10 min (Avanti J-26XP, Beckman Coulter, Brea, CA, United States). The DNeasy^® Blood & Tissue kit (Qiagen, Hilden, Germany) was used to extract microbial DNA from the pellets following the procedure of the manufacturer. The purified DNA samples were stored in 50 μl of Buffer AE at –20°C. The qPCR was performed using ABI Prism 7500 Real-time qPCR Detection System (Applied Biosystems, Foster City, CA, United States) with the primer pairs of Arc364r/915r for methanogens and Bac338f/518r for syntrophs. The PCR mixture of 20 μl contained 10 μl of PowerUp SYBR Green PCR Master Mix (Applied Biosystems), 0.8 μl of each primer (stock concentration of 10 μM), 1 μl of template DNA, and 7.4 μl of ddH₂O. Each measurement was performed in triplicate. The thermal cycles and fluorescence signal acquisition followed the protocols as described previously (Ma et al., 2013; Hu and Lu, 2015). Data analysis was carried out with 7500 System SDS Software. Standard curves were obtained using serial dilutions (10⁸–10² copies μl) of linearized plasmids containing cloned fragments of either bacterial 16S rRNA or archaeal 16S rRNA (Liu and Lu, 2018). The cloned fragments were obtained through PCR from pure culture DNA using the primer pairs described above. One-way ANOVA and Duncan’s post hoc test was used for testing significant differences in qPCR data using SPSS software (version 17).

Fluorescence in situ hybridization (FISH) analysis was performed for the WM coculture on 4% paraformaldehyde-fixed samples according to a procedure described previously (Moter and Gobel, 2000). Oligonucleotide probes specific for bacteria (Cy3-labeled EUB338mix probes) and archaea (fluorescein isothiocyanate (FITC)-labeled ARC915 probe) were used. The details of the probes design are available in the probeBase (Loy et al., 2003). The labeled samples were visualized using epifluorescence microscopy (Axioimager D2, ZEISS, Oberkochen, Germany) (Zhang and Lu, 2016). The brightness of fluorescence images was optimized by ImageJ (Ciniselli et al., 2015). The cells in the lag phase, early exponential phase, mid-log phase, and late exponential phase were collected and observed. For scanning electron microscopy (SEM), cells were collected by a syringe, fixed with 2.5% (wt/vol) glutaraldehyde in phosphate-buffered saline, and sequentially dehydrated with ethanol [20, 40, 60, 80, 95, and 100% (v/v) ethanol, 10 min for each step]. The dried samples were coated with platinum and imaged using SEM (Axio imager D2, ZEISS) (Zhang and Lu, 2016).

The microfluidic experiment was conducted for the WM coculture. The microfluidic chips were fabricated following the protocol as described (Sun et al., 2016). A silicon substrate (100 mm) was heated at 190°C for 5 min. The positive lift-off photoresist SU8 3050 (MicroChem Corp, Newton, MA, United States) was sequentially deposited and was exposed to the photomask at a UV light. Then, the lift-off process was applied to remove the photoresist. The microfluidic channels were made using soft lithographic techniques. The elastoplastic material polydimethylsiloxane (PDMS) (RTV615, Momentive, Waterford, NY, United States) mixture was degassed and poured into the SU8 molds (Sun et al., 2016). The thickness of the PDMS was 170 μm. After the PDMS replica was cured and released, the cleaned glass microfluidic substrate and PDMS replica were surface treated by air plasma in a reactive ion etching system. Immediately, the PDMS replica was placed against the glass substrate. The clamped PDMS replica and glass substrate were subsequently placed in an oven at 70°C for 1 h.

Three batches of microfluidic experiments were conducted. The microfluidic platform was placed within an anaerobic glove box to maintain an anaerobic condition for all experiments. First, the log-phase S. wolfei in pure culture was pumped into the channels of the chip at the speed of 6,000 μl h^–1 until the channels were fully filled with the culture (Supplementary Figure 1). The entire chip was then tilted to make the cells move into the wells at one side of the chamber under gravity. Two hours later, the chip was placed back horizontally. The remaining S. wolfei in channel passages were washed out thoroughly with the fresh medium at the speed of 200 μl h^–1. Then the M. maripaludis were pumped into channels at the speed of 200 μl h^–1 for 2 h. Finally, the velocity of mass flow in channels was reduced to 10 μl h^–1 for maintaining the liquid environment in chip while reducing the disturbance of liquid flow for cell cultivation. For the second batch experiment, similar procedure was conducted except the reverse of order for cell injection; i.e., M. maripaludis was injected first and then followed by S. wolfei. The third batch was a single organism control experiment where only the S. wolfei cells were used in both steps of cell injection.

We investigated the growth synchronization between butyrate-oxidizing syntrophs and methanogens by setting different initial ratios of two populations in the coculture. The synchronization between S. lipocalidus and M. conradii was tested by setting the initial cell ratio of S. lipocalidus versus M. conradii to 24:1, 3:1, and 1:3, which were designated as the high, middle, and low ratios, respectively. The high ratio showed the shortest lag phase of CH₄ production followed by the middle ratio, whereas the lag phase nearly doubled in the low-ratio treatment (Figure 1A). Moreover, butyrate consumption and acetate accumulation were faster for the high and middle ratios than the low-ratio treatment (Figures 1B,D). H₂ accumulated transiently, with the maximal concentration being higher and earlier in the high- and middle-ratio treatments compared with the low-ratio treatment (Figure 1C).

Cocultures were evaluated by the relative growth and relative growth rate of each partner as determined by qPCR of subsamples collected during incubation. The relative growth was defined as the fold change of cell number at time T relative to that at time 0, while the relative growth rate was defined as the difference in cell number between time T2 and time T1 divided by the duration between two timepoints. For the high-ratio treatment where butyrate oxidation was more rapid, the cell number of M. conradii increased 4.5-fold in 3–4 days and 7.5-fold in 6 days and then leveled off, whereas the cell number of S. lipocalidus increased only 2.5-fold at day 6 and leveled off (Figure 2A, left). The relative growth rate of M. conradii reached the maximum of 38 cell numbers h^–1 at day 4; meanwhile, that of S. lipocalidus remained close to zero (Figure 2B, left). For the middle-ratio treatment, the relative growth displayed a similar pace for two populations, but reaching greater values for M. conradii than for S. lipocalidus (Figure 2A, middle). Importantly, the relative growth rate showed a shift from being significantly greater for M. conradii at day 4 to being significantly greater for S. lipocalidus at day 5 (Figure 2B, middle). For the low-ratio treatment, both the relative growth and the relative growth rate of S. lipocalidus significantly surpassed those of M. conradii. Only in the later stage (after 9 days) did the relative growth rate of M. conradii become greater than the S. lipocalidus (Figures 2A,B, right). We calculated the cell ratio (R) of S. lipocalidus versus M. conradii over the incubation, which showed that the log₂R value fluctuated in the early stage depending on the initial ratio treatment but finally converged at about R = 1.2 from day 5 forward for all the three treatments (Figure 2C). The total biomass and the Gibbs free energy changes available for S. lipocalidus and M. conradii are illustrated in Supplementary Figures 2, 3. The Gibbs free energy changes showed no significant difference among the cell ratio treatments, albeit the shift in timeframe of availability (Supplementary Figure 3). The highest biomass accumulation was detected in the middle-ratio treatment followed by the high-ratio treatment, while only about a half of biomass was formed in the low-ratio treatment compared with the middle ratio (Supplementary Figure 2).

A similar test was conducted for the mesophilic WM coculture, in which the initial cell ratio of S. wolfei versus M. maripaludis was set at 5:1, 3:2, and 1:5 for the high-, middle-, and low-ratio treatments, respectively (Figure 3). Like in the thermophilic LC coculture, the high-ratio treatment showed the shortest lag period of CH₄ production (10–12 days) followed by the middle-ratio treatment (15 days), while the low-ratio treatment had the longest lag phase (18–20 days) (Figure 3A, left). The transient H₂ accumulation in three treatments was correlated with the CH₄ production (Figure 3A, right). The growth of M. maripaludis initiated at day 10, increased by three-fold at day 15 and then leveled off (Figure 3B, left), while S. wolfei exhibited only minor growth at day 15. For the middle-ratio treatment, the growth of both M. maripaludis and S. wolfei initiated at day 15, displaying moderately greater values for M. maripaludis than for S. wolfei (Figure 3B, middle). By comparison, the relative growth was much greater for S. wolfei than for M. maripaludis in the low-ratio treatment (Figure 3B, right). Similar to the LC coculture, the cell ratio of the WM cocultures varied in the early stage but converged at around 1 from day 15 forward for all the three treatments.

Next, we tested the effect of physical disturbance on the formation and activity of microbial aggregates in the mesophilic WM coculture. The actively growing coculture was inoculated to fresh medium; and two disturbance treatments including the intermittent ultrasound and constant shaking (180 r min^–1) of the incubation bottles were applied, with static incubation serving as the control. The new cocultures presumably had the initial syntroph-to-methanogen cell ratio inherited from the inoculants. The results showed that the activity of CH₄ production was slightly repressed by both ultrasound and shaking treatments compared with the static control (Figure 4A). We calculated the maximum rate of CH₄ production during the periods when CH₄ concentration linearly increased in headspace, which revealed 10% reduction in the ultrasound and 25% reduction in the constant shaking treatment compared with the static control (Figure 4C). H₂ accumulated transiently to a concentration of 20–25 Pa with the higher values for the shaking treatment (Figure 4B).

Culture samples were collected four times from incubations for FISH observation. Microbial aggregates were detected at the first day, which were composed mainly of the S. wolfei cells (Figure 5A). At the fifth day, corresponding to the initiation of coculture growth (Figure 4A), the cells of M. maripaludis gradually increased surrounding the aggregates. At the eighth day, corresponding to the mid-exponential phase, the size of aggregates significantly enlarged with the increased cell numbers of both S. wolfei and M. maripaludis within aggregates. Further growth of syntrophic aggregates was observed at the 12th day in the stationary phase when most of butyrate was already consumed. The syntroph and methanogen cells were found interconnected densely within the mature aggregates (the eighth and 12th days in Figure 5). Notably, there was no significant difference in the tendency of microbial aggregation among the three treatments. The fluorescence intensity of the individual populations was estimated using the ImageJ software, which indicated a nearly linear increase of M. maripaludis cells over 2 weeks’ incubation (Figure 5B, left), while S. wolfei showed an obvious lag phase during the initial 5 days (Figure 5B, right). Consistent with the pattern of FISH images, the fluorescence intensity of two populations showed no significant difference among the three treatments. SEM observation made at the end of incubations revealed the extensive formation of extracellular polymeric substances (EPSs) that physically served as the matrix support for microbial aggregation (Figure 6). The influence of disturbance treatments, however, could not be explicitly identified in SEM images. These results suggest that the S. wolfei cells initiated the syntrophic aggregation and allowed for the attachment and gradual growth of M. maripaludis before the rapid activity of syntrophic coculture. Importantly, although the activity of the coculture was influenced by two disturbance treatments, the tendency of microbial aggregation was not.

Last, we performed the microfluidic chip experiments to observe the cell-to-cell interaction in the WM coculture. The chip contained two symmetric wells for culture incubation along the opposite sides of flow channels (Supplementary Figure 1). The flow medium contained 10 mM of butyrate and was used throughout all microfluidic experiments. Three alternate tests were conducted. First, S. wolfei cells were injected; and the chips were tilted to allow more cells shunted to the left-side well than the right-side well. The chips were then returned to horizontal position; and after the remaining cells were washed thoroughly away from chip channels, M. maripaludis cells were loaded for 2 h and further incubated for over 20 h. We found that more of M. maripaludis cells entered the left-side well than the right-side (Supplementary Video 1). Over the 20-h incubation, the coculture in the left-side well grew to a much larger population than in the right-side well. In the next test, we changed the order of cell loading so that the M. maripaludis cells were injected first with more cells loaded to the left-side well than the right-side. The following feeding of S. wolfei cells revealed that more of S. wolfei cells entered the left-side well where more of the M. maripaludis cells were already located (Supplementary Video 2); and the populations increased to a much greater extent in the left-side well. The third test was used to verify if the inertia of microfluidic flow influenced the cell movement. Here, we pumped the S. wolfei first and tilted the chip to the right side so that more S. wolfei cells were loaded to the right-side well at the beginning. After returning the chip to horizontal position, we injected the S. wolfei cells again. Within 2 h of the cell injection, we observed that the S. wolfei cells entered the wells of both sides without a preference (Supplementary Video 3). Further incubation was not continued as S. wolfei could not grow on butyrate alone in the media.