Controls on Interspecies Electron Transport and Size Limitation of Anaerobically Methane-Oxidizing Microbial Consortia

ABSTRACT About 382 Tg yr−1 of methane rising through the seafloor is oxidized anaerobically (W. S. Reeburgh, Chem Rev 107:486–513, 2007, https://doi.org/10.1021/cr050362v), preventing it from reaching the atmosphere, where it acts as a strong greenhouse gas. Microbial consortia composed of anaerobic methanotrophic archaea and sulfate-reducing bacteria couple the oxidation of methane to the reduction of sulfate under anaerobic conditions via a syntrophic process. Recent experimental studies and modeling efforts indicate that direct interspecies electron transfer (DIET) is involved in this syntrophy. Here, we explore a fluorescent in situ hybridization-nanoscale secondary ion mass spectrometry data set of large, segregated anaerobic oxidation of methane (AOM) consortia that reveal a decline in metabolic activity away from the archaeal-bacterial interface and use a process-based model to identify the physiological controls on rates of AOM. Simulations reproducing the observational data reveal that ohmic resistance and activation loss are the two main factors causing the declining metabolic activity, where activation loss dominated at a distance of <8 μm. These voltage losses limit the maximum spatial distance between syntrophic partners with model simulations, indicating that sulfate-reducing bacterial cells can remain metabolically active up to ∼30 μm away from the archaeal-bacterial interface. Model simulations further predict that a hybrid metabolism that combines DIET with a small contribution of diffusive exchange of electron donors can offer energetic advantages for syntrophic consortia.

sediments reduce emissions of the potent greenhouse gas methane (5) to the overlying water and the atmosphere. Due to the role of methane in atmospheric radiative forcing (6), it is important to understand the processes and mechanisms involved in AOM. Recent studies provide evidence that supports direct extracellular electron transfer, for example, in single (7,8)-and mixed (9)-species Geobacter biofilms. Furthermore, direct interspecies extracellular electron transfer (DIET) has been observed in cocultures (10,11) and microbial aggregates (12)(13)(14)(15)(16). There is also a growing body of evidence that DIET also takes place between methanotrophic archaea and syntrophic sulfate-reducing bacteria in AOM consortia (17)(18)(19), where it serves as an effective transport mechanism over long spatial distances (17). It overcomes limitations inherent in the diffusive exchange of dissolved electron-carrying molecules (mediated interspecies electron transfer, or MIET) that lead to the build-up of reaction products and the subsequent shutdown of metabolic activity (19,20).
DIET is thought to occur through a variety of mechanisms, including direct contact between cells (21), through electrically conductive pili (10,11,(13)(14)(15) and/or extracellular cytochromes (11,15,16). Genomic and transcriptomic data of enrichments with different types of AOM consortia (ANME-1a/HotSeep-1, ANME-1a/Seep-SRB2, and ANME-2c/Seep-SRB2) revealed that genes encoding flagella or type IV pili, and/or surface-bound or extracellular c-type cytochromes, were highly expressed (22). Notably, ANME-2 genomes encode large multiheme cytochromes containing putative S-layer domains (17) thought to be analogous to the Gram-negative porin-cytochrome conduits in that they can be used for electron egress through the outermost cell layer (23). Observations using transmission electron microscopy (TEM) showed staining consistent with heme-rich areas and pilus/ wire-like structures in the intracellular space in AOM consortia (17,18,22). These features suggest that DIET is the principal mechanism of sulfate-dependent AOM. While this hypothesis awaits direct experimental confirmation or indirect support through measurements that show the potential for conduction within the aggregates and is hampered by a lack of any pure cultures of microorganisms carrying out this metabolism, modeling efforts indicated that DIET can support cell-specific AOM rates and archaeal activity distributions that were consistent with observations from single-cell resolved fluorescent in situ hybridization-nanoscale secondary ion mass spectrometry (FISH-nanoSIMS) analyses (19).
Recently, a finite distance over which extracellular electron transport sustains metabolic activity was documented in biofilms of Geobacter sulfurreducens (24). These results suggest that the extent to which conductive biomolecules can support optimal cell growth away from an electrode surface is limited (24,25). Using a similar experimental approach, a drop in activity with distance between electron donors (archaea) and acceptors (bacteria) was not observed in AOM aggregates (17)(18)(19). However, the size of the microbial aggregates analyzed was much smaller than the Geobacter biofilm thickness, leading to short separation distances between the syntrophic partners within the aggregates (17). In this study, we target exceptionally large aggregates (radius, ;20 mm) in which bacteria and archaea were spatially segregated. We measured and analyzed the metabolic activity of individual cells using FISH-nanoSIMS. Measurements of 15 NH 4 1 incorporation are then used to validate a reactive transport model. Simulation results consistent with our empirical observations form the basis for three key novel aspects of this work. First, we investigate the mechanisms of potential losses associated with direct extracellular electron transport by accounting for ohmic resistance and activation loss that ultimately limits metabolic activity away from an archaeal-bacterial interface, an effect not apparent in small or well-mixed aggregates we have reported on earlier (17,19). Second, we investigate the potential for environmentally sourced electron donors used by the SRB, partially decoupling archaeal methanotrophy and bacterial sulfate reduction. Third, we consider the advantages of a hybrid DIET-MIET mechanism that can offer energetic benefits allowing for balanced microbial energetics for both syntrophic partners, particularly for large aggregates.

RESULTS AND DISCUSSION
Large, segregated aggregates display significant spatial variation in cellular activity. Previous experimental work measuring the activity of individual cells in syntrophic ANME-SRB aggregates demonstrated a lack of significant correlation between cellular activity and distance to syntrophic partner over short distances (a few cell diameters [17]). These observations were sufficient to rule out molecular diffusion as the major mechanism of electron transfer between the two partners but were limited in their spatial extent due to relatively small aggregate size as well as the complex three-dimensional structure of many AOM consortia that made it difficult to confidently assign distances to nearest partners that may lie above and below the plane when analyzing single two-dimensional cross-sections. We have occasionally observed exceptionally large AOM consortia in nanoSIMS analyses where significant variations in activity appear to be related to distance from their nearest partner (for example, see Fig. S1 at https://doi.org/10.6084/m9.figshare.13536086.v2). While these previous observations suggested that cellular activity is correlated with distance to nearest syntrophic partner over large distances, it was not possible to determine a precise magnitude of the activity gradients without additional information about the three-dimensional aggregate structure.
To overcome these challenges, we cut and analyzed parallel sections through a large, well-segregated ANME-2/SRB consortium after 15 NH 4 1 stable isotope probing, allowing us to roughly reconstruct the spatial distribution of both partners across the entire consortia ( Fig. 1A and B). Two features of this .50-mm AOM consortium made it ideal to study. First, the spatial organization of the syntrophic partnership is simple and well defined, with no incursions of bacteria into the ANME-dominated interior of the aggregate. Second, the bacteria form a crescent around the archaeal core instead of a complete shell. Had the bacteria formed a complete shell, there would be perfect correlation between ANME distance to nearest syntrophic partner and distance to the Top row, raw 14 N 12 C 2 secondary ion counts illustrating the position of cells. Bottom row, fractional abundance of 15 N calculated as 15 N 12 C 2 /( 15 N 12 C 2 1 14 N 12 C 2 ), all scaled to the same intensity. Note sulfate-reducing bacteria (SRB) assimilate significantly more 15 N, on average, than their ANME-2 counterparts, as has been previously shown (17). (C) Illustration of nanoSIMS data extraction and modeled geometry. From left to right, FISH image indicating phylogenetic identity of cells (green, general bacterial probe [Eub338mix]; red, ANME-2b-specific probe [ANME-2b-729]; blue, DNA stain [DAPI]); segmentation image showing SRB and ANME cells manually segmented based on observation of FISH and nanoSIMS data; individual segmented cells shaded by their total 15 N fractional abundance; SRB and ANME cells scaled by minimum and maximum values within the population; and illustration of modeled aggregate geometry (the dashed line represents axis of rotation). The yellow X marks the approximate minimum ANME cell activity. Note that additional sections were visually inspected to help verify aggregate structure. Only those analyzed by nanoSIMS analysis are shown. This figure contains tiled images that were stitched together to make these composite images. Black regions within the image are places where the square tiles did not overlap.
Electron Transport in AOM Consortia ® surface of the aggregate, making these two potential controls on cellular activity difficult to disentangle. With a crescent geometry, however, some ANME can be found at the surface of the consortia closest to the surrounding environment and at great distance from the nearest SRB, allowing us to distinguish between the effect of syntrophic distance and distance to the environment that supplies the growth substrates CH 4 and SO 4 22 and the tracer 15 NH 4 1 . Since the minimum ANME activity was observed to be near the aggregate surface, far from the SRB, we conclude that distance to partner is more significant than substrate limitation (Fig. 1C). This finding is consistent with the measurements in the second large aggregate we observed, one with a slightly less segregated distribution of archaea and bacteria, as shown in Fig. S1 at https://doi.org/10 .6084/m9.figshare.13536086.v2.
A unifying model across aggregate size. Observations of 15 N incorporation in single cells from a section cut approximately normal to the ANME-SRB interface revealed a decrease in the anabolic activity of both ANME and SRB with increasing distance to their nearest syntrophic partners ( Fig. 2A). This effect was highly significant and explained large portions of the variability of cellular activity in the two populations, with a slope of 20.0238 6 0.0009 fmol cell 21 day 21 mm 21 (R 2 = 0.69) and 20.0594 6 0.0083 fmol cell 21 day 21 mm 21 (R 2 = 0.27) for archaea and bacteria, respectively ( Fig. 2A). Our base model, in which 92.5% of the electrons produced in the oxidation of CH 4 are transferred to the bacteria via DIET and 7.5% of the electrons are transferred via MIET, provides the best fit of the activities observed in aggregates across a wide range of aggregate sizes (Fig. 2). Cell-specific activities decrease slightly with increasing distance from the nearest syntrophic partner in a simulated 20-mm radius aggregate, with slopes of 20.0267 6 0.0004 fmol cell 21 day 21 mm 21 (R 2 = 0.9954) and 20.0653 6 0.0017 fmol cell 21 day 21 mm 21 (R 2 = 0.9936) for archaeal and bacterial activity, respectively ( Fig. 2A). One-way analysis of covariance (ANCOVA) revealed that the slopes and intercepts of the regressions of model results and of observational data do not differ significantly, with P values of 0.30 and 0.71 for archaea and bacteria, respectively. Simulations for a small aggregate with the identical model parameterization retained good agreement between observed and modeled metabolic activity patterns (Fig. 2B), with a P value of 0.96 for both archaea and bacteria compared to observations.
How far apart can ANME and SRB cells be and remain active in AOM consortia? The metabolic activity of syntrophic AOM aggregates can be limited by the availability of electron donors and acceptors, as reflected by the thermodynamics (equation 6) of the overall reactions (termed Rxn3 and Rxn4 and described in Materials and Methods). Here, we investigate the internal and external constraints that potentially limit the metabolic activity within the context of the observed aggregate arrangement. All archaeal and bacterial cells remained active over a wide range of aggregate sizes in our model simulations (Fig. 3); however, the simulated activity of individual cells did decrease with increasing distance from their syntrophic partners. This effect is observed in FIG 2 Measured and modeled cell-specific activity in aggregates with a radius of 20 mm (A; this study) and 5 mm (B; 17), plotted against their distance from the closest syntrophic partner (interface). Data were fitted using linear regression with 95% confidence intervals. Note that the cell-specific rate constants were not retuned to match the activities in the small aggregate. model simulations for both archaea (Fig. 3A) and bacteria (Fig. 3B) and is slightly steeper for the latter. The shape and magnitude of the activity decrease curve were nearly identical between aggregates of different sizes, highly consistent with what we observed with anode-respiring G. sulfurreducens biofilms of different thicknesses under high and low anode potentials (24). We included in our model simulations segregated aggregates with radii of up to 100 mm (same spatial arrangement as that shown in Fig. 1C). In strongly segregated AOM aggregate and over sufficiently long distances, cell activity decreases with distance to the syntrophic partner even with electron transfer via DIET. Cellular activities in strongly segregated large aggregates experienced a .70% drop in activity as separation distances increase to 15 mm for bacteria and to 30 mm for archaea (Fig. 3). Thus, DIET allows for much bigger clusters than can be supported with MIET alone.
Simulations including molecular diffusion (MIET) of potential syntrophic intermediates, such as intermediate in addition to DIET, revealed that metabolic activity could become severely limited with large separation distances between partners (aggregate size [r agg ], 60 mm; see Fig. S9 at https://doi.org/10.6084/m9.figshare.13536086.v2), even though MIET only accounted for 7.5% of the electron transfer from archaea to bacteria. It is noted that even at this size extreme, the mass transport of substrates and metabolites, including CH 4 , SO 4 22 , HS 2 , H 1 , and HCO 3 2 , was not limiting due to the relatively high concentrations of methane and sulfate at the outer environmental boundary, varying by a factor of less than 1%, except for HS 2 , which varies by 10% across the aggregate (data not shown). These results suggest that this distance-dependent cellular activity pattern is a critical factor determining the size of monospecies clusters within AOM consortia. Thus, larger aggregates would be expected to have a more interspersed distribution of archaeal and bacterial partners to maintain high levels of single-cell activity or, once a segregated aggregate size limit is reached, larger consortia then separate into two or transform into a larger clustered morphology as bacteria grow into the archaeal core (26).
What controls the spatial distribution of activity? The spatial variation of the cell metabolic activity was found to depend on the usable electric potential (h net ), which is set by the available energy from the reaction (at approximately 0.0357 V for archaea and bacteria) minus the effect of losses. The activation loss was the main contribution to potential losses at distances of approximately #8 mm to the partner interface, while ohmic losses were important at larger distances (Fig. 4). This pattern was observed for both archaeal and bacterial cells. Activation loss was maximal at the archaeal-bacterial interface, with a value of 0.013 V, and decreased away from the archaeal-bacterial interface. In contrast, ohmic resistance loss increased from 0 to ;0.02 V as the distance from the archaeal-bacterial interface increased, leading to a maximum total potential loss at a value of ;0.023 V for archaea and bacteria. As the net available potential Electron Transport in AOM Consortia ® (h net ) approaches the minimum potential required for ATP synthesis (;0.013 V; equation 6), metabolic rates decrease due to energetic limitations, as indicated by the thermodynamic factor, F T , approaching 0 ( Fig. 4).
Voltage losses depend on a number of factors, including the concentration of redox-active molecules (M tot ), conductive network density (N nw,cell ), its conductivity (s ), cell surface redox activation factor (k act ÂA act ), and cell rate constants (k A , k B ). Activation loss was strongly impacted by k A , k B , N nw,cell , and k act ÂA act and less so by M tot and s (Fig. 5A). Increasing k act , N nw,cell , and k act ÂA act by a factor two or k A , k B by 1.5-fold reduced the activation loss by 6.4 6 0.9 mV, 6.3 6 1.3 mV, 6.4 6 0.9 mV, and 3.9 6 1.6 mV, respectively, while increasing M tot or s by a factor of two led to an increase of activation loss by 1.0 6 1.1 mV and 0.4 6 0.7 mV, respectively. k A , k B , and N nw,cell showed similar effects on activation and ohmic resistance losses, but changes in k act ÂA act , M tot , and s had opposite impacts, with an increase by a factor two of k act ÂA act , M tot , and s leading to a change in ohmic resistance losses by 0.5 6 0.7 mV, 2.4 6 1.9 mV, and 21.6 6 1.1 mV, respectively (Fig. 5B). In total, k A , k B , k act ÂA act , and N nw,cell exhibited substantial impact on net available potential, whereas M tot and s showed moderate effects, in part due to the counteracting effect on h act and h om for M tot and s ( Fig. 5A and B). It should be noted that these results FIG 4 Factors controlling cell activity as a function of the distance from the archaeal-bacterial interface at aggregate radius of 60 mm for archaea (A) and bacteria (B). The left axis reflects electric potential for activation loss (h act ), ohmic resistance loss (h om ), net available potential (h net ), potential from reaction (h rxn ), and minimum potential required for ATP synthesis (h ATP ). The right axis reflects the thermodynamic factor, F T . The shaded areas highlight the range of distances encountered in the observed aggregate with a 20-mm radius (Fig. 1).

He et al.
® are insensitive to changes in the electron conduction constant (k D ) and electric field associated rate constant (k EF ) ( Fig. S14 and S15 at https://doi.org/10.6084/m9.figshare.13536086 .v2). In agreement with results reported previously (19), we observed no significant difference between simulations with electric field as the sole driving force and simulations with redox gradient as the driving force. Note that changes in these parameters affect not only the overall energetics for the AOM consortium but also the distribution of cell activity. Changes in M tot , s , and k act ÂA act alter the shape of cell activity with distance between syntrophic partners, while k A , k B , and N nw,cell mostly affect the slope of a linear decrease of activity with distance ( Fig. S16 at https://doi.org/10.6084/m9.figshare.13536086.v2).
Because several experimentally poorly characterized model parameters impact the magnitude of activity and spatial patterns of modeled electric losses (Fig. 5, Fig. S11 to S13 at https://doi.org/10.6084/m9.figshare.13536086.v2), our work emphasizes important targets for future study and observation, such as an assessment of the number of pili/wire-like structures recently observed to be involved in extracellular electron transfer (EET) for some archaeal/bacterial syntrophic consortia (18,22). The accurate quantification of these connections is challenging, as not all such structures are necessarily conductive, and most observations are two-dimensional sections through a threedimensional matrix of extracellular material. However, the extent to which archaeal and bacterial cells are connected is important, because variations in the extent of conductive connections can substantially alter the metabolic activity pattern by influencing both activation loss, h act (Fig. 5A), and ohmic resistance loss, h om (Fig. 5B), and, hence, the net available potential, h net (Fig. 5C). Halving N nw,cell significantly limited the metabolic activity due to the reduced availability of h net (Fig. 5C), in agreement with Storck et al. (27), who reported that decreasing conductive network density (N nw,cell ) by a factor of 10 led to a 60% decrease of electron transport rate. Doubling N nw,cell resulted FIG 5 Changes of activation loss, Dh act (A), ohmic resistance loss, Dh om (B), and net available potential, Dh net (C), due to a change in total redox active molecules (M tot ), number of conductive connections (N nw,cell ), conductivity (s ), cell redox active factor (k act ÂA act ), and cell rate constants (k A and k B ). Error bars reflect that the impact is not exactly constant with distance for archaeal-bacterial interface (see Fig. S11 to S13 at https://doi.org/10.6084/m9.figshare.13536086.v2).
Electron Transport in AOM Consortia ® in a homogenous distribution of metabolic activity, similar to the finding in the study by Storck et al. (27), in which the electron transport rate increased slightly for a 10-fold increase in N nw,cell , suggesting a plateau was reached. Furthermore, while no data on AOM consortium conductivity, s , have been published yet, such measurements have been made in Geobacter biofilms (9,(28)(29)(30)(31), Geobacter pilin nanofilaments (28,32), Desulfovibrio desulfuricans nanofilaments (33), methanogenic aggregates from anaerobic wastewater reactor (12), and granules from anaerobic bioreactors (34), among others. The conductivity, s , has a significant impact, with a reduction by a factor of 10 to 10 23 S m 21 drastically reducing the metabolic activity ( Fig. S16C at https://doi.org/10.6084/m9.figshare .13536086.v2). By increasing conductivity to 10 21 S m 21 , metabolic activity reached a homogenous spatial distribution, owing to the increased h net at higher conductivity (Fig. 5C).
Type and strength of syntrophic coupling between archaea and bacteria. The model was used to assess potential advantages of a mechanism in which electron transport through both DIET and MIET is active. A hybrid DIET-MIET mechanism, as implemented in our baseline simulation, can lead to a higher energy yield than electron transfer by DIET alone, as it allows for more balanced microbial energetics for both syntrophic partners. The conditions for sulfate-reducing bacterial cells were slightly more energetically favorable, with a 92.5% DIET/7.5% MIET hybrid metabolism ( Fig. 6 (Fig. 6), the model simulations showed a narrow window with approximately 90 to 100% DIET that enabled energetically favorable conditions for both bacterial and archaeal cells (Fig. 6). Importantly, a hybrid mechanism can affect the balance of energy gains between the syntrophic partners, which results in improved energetic conditions for the partner most energetically constrained, thereby benefitting both archaea and bacteria (Fig. 6).
Potential for decoupling of archaeal and bacterial metabolisms. We considered metabolic decoupling between the ANME and SRB partners, where the bacteria may use electron donors derived from the external environment rather than be provided the syntrophic partner. We explored the impact of an externally sourced electron donor, DH, on bacterial metabolism by loosening the coupling between archaeal and bacterial metabolism (see Appendix A2 in the supplemental material at https://doi.org/10 .6084/m9.figshare.13536086.v2). Such decoupling has been observed in thermophilic AOM consortia, where it has been shown that the ANME-1 sulfate-reducing bacterial partner HotSeep-1 can utilize H 2 and grow independently of ANME (18). As MIET using H 2 is not thought to be an important form of syntrophic electron transfer (18), detectible hydrogenases are lacking in ANME (35) and SRB (36) genomes recovered from cold seeps, and experimental data demonstrated that excess hydrogen addition does not inhibit AOM activity in sediment incubations and enrichment cultures (37-39); for convenience, we continue to consider formate a soluble electron donor. Formate concentrations in marine sediments range from below the detection limit (0.37 mM) to 10.38 mM in Baltic Sea sediments (40), 2 to 18 mM in northern Gulf of Mexico sediments (41), up to 59.5 mM in Hydrate Ridge sediments (42), 12.1 mM in Aarhus Bay sediments (43), and 36 to 158 mmol/kg in fluid from the Lost City hydrothermal field (44). Thus, simulations were carried out for 1 to 100 mM formate in the environment. Increasing formate from 1 mM to 15mM led to a significant increase of bacterial activity at the aggregate surface while showing nearly no impact on archaeal cells (Fig. S8A at https:// doi.org/10.6084/m9.figshare.13536086.v2). At a lower HCOO 2 concentration (1mM), bacterial cells exhibited a slight shortage of HCOO 2 supply away from the archaeal-bacterial interface (Fig. S8B). At high formate concentrations (.15mM), carrying out archaeal CH 4 oxidation could become thermodynamically unfavorable due to the accumulation of HCOO 2 (not shown). Noticeably, the Gibbs free energy change for sulfate reduction [DG R(4) ] significantly decreased from ;-27.5 kJ mol 21 to 230.05 kJ mol 21 when changing formate from 1mM to 15mM (Fig. S8C), leading to a significant increase of bacterial thermodynamic constraint F T from 0.35 to 0.7 at the aggregate surface, while no significant changes were observed for archaea (Fig. S8D). Notably, the increased formate from 1 mM to 15mM did not significantly impact the total flux of HCOO 2 , although an increase of HCOO 2 concentration within consortium was observed (Fig. S8B).
Conclusions. We report on the metabolic activity distribution of individual cells in a large AOM consortium using FISH-nanoSIMS. A decline in cell activity with the increasing distance from the archaeal-bacterial interface was observed in a section through the center of the aggregate, cut approximately normal to the ANME-SRB interface. These results provide the first quantitative assessment of the growth penalty that exists over large separation distances between these syntrophic partners, an effect that is not apparent in small or well-mixed aggregates (17,19). A reactive transport model accounting for thermodynamic limitations on cell metabolism, as well as activation and ohmic resistance losses in the exchange of electrons between syntrophic microorganisms, successfully reproduced these novel observations. Direct interspecies electron transfer makes the observed spatially distributed cell activity possible, where at larger distances ohmic losses are predominantly responsible for constraining the interspecies syntrophic partner distance within ,30mm. The process-based model also revealed possible advantages of a hybrid DIET-MIET mechanism, allowing for balanced microbial energetics for both syntrophic partners but opening up the potential for decoupling of the sulfate-reducing bacterial partner from the methanotrophic archaea by utilizing electron donors from the environment. While this points to the possible benefit of versatile and adaptable use of diverse electron donors and modulating association strengths, the nature of such small redox-active molecules acting as electron shuttles remains unknown. Future work will Electron Transport in AOM Consortia ® help us answer these mechanistic questions by a careful comparison of ANME and SRB genomic potential and expression with their cellular activity patterns.

MATERIALS AND METHODS
Experimental data. (i) Sample collection. Methane seep sediments covered with white bacterial mats were collected from Jaco Scar, off Costa Rica, at 1,811-m water depth (lat 9.1163, long 284.8372). Samples were collected by push core (PC6) during dive number AD4912 on 27 May 2017 by DSV Alvin, launched from R/V Atlantis on research cruise AT37-13. The sediment core was processed shipboard into 3-cm-depth horizons that were placed in separate Whirl-Pak bags and stored under anoxic conditions in a large sealed Mylar bag flushed with Ar. These sediments were stored at 4°C until they were returned to the laboratory, where sediments were mixed with N 2 -sparged, 0.2-mm-filtered seawater collected above the sampling site and incubated in anoxic 1-liter Pyrex bottles with a secured butyl rubber stopper supplied with a 100% methane headspace (30 lb/in 2 ).
(ii) Stable isotope probing, incubation, and sampling. Stable isotope incubation experiments were conducted using slurried sediment from PC6, corresponding to the 3-to 6-cm-depth horizon. Sediment was mixed 1:3 with N 2 -sparged, 0.2-mm-filtered seawater from above the sampling site (28 mM sulfate) and amended with 1 mM NH 4 Cl with 99% 15 N abundance (Cambridge Isotope Laboratories, Inc.) and incubated at 4°C. Headspace composition was 100% methane at 30 lb/in 2 . After 7 days, subsamples were collected for analysis by first shaking the incubation bottle to resuspend the sediment slurry and then collecting an aliquot using an N 2 -flushed needle and syringe. A volume of 1 ml of sediment was chemically fixed by mixing with 1 ml of 4% paraformaldehyde in 3Â PBS and incubated for 1 h at room temperature. Sediments containing AOM aggregates were washed three times with 3Â PBS and finally resuspended in 50:50 PBS-ethanol (EtOH) and stored at 220°C.
(iii) Resin imbedding and FISH staining. Fifty microliters of fixed sediment slurry in 50:50 PBS-EtOH was mixed with 750 ml PBS in a 2-ml microcentrifuge tube and sonicated on ice with a microtip sonication probe (Branson), 3Â for 10 s at setting 3 (8 W). Aggregates were separated from sediment particles by density gradient centrifugation by underlaying the sonicated liquid with 1 ml of Percoll and spinning at maximum speed for 30 min in a tabletop microcentrifuge at 4°C. The top aqueous layer containing concentrated aggregates was removed and pelleted by spinning at 10,000 Â g at room temperature for 1 min. The pellet was gently removed and immobilized in molten 3% noble agar in PBS. Once solidified, agar was trimmed to a small cube around the pellet and imbedded in glycol methacrylate (Technovit 8100) resin by following the manufacturer's protocol. Semithin section (1 to 2 mm thick) were cut using a microtome and deposited on water droplets on polylysine-coated slides with Teflon-lined wells (Tekdon, Inc.). FISH hybridization on thin sections was conducted as described previously (17). ANME-2b-specific probe ANME-2b-729 with a dual 39/59 Cy3 label (45) and a universal bacterial probe EUB338mix (EUB338, -II, and -III) labeled with fluorescein isothiocyanate (FITC) were used at 35% formamide concentration (supplied by Integrated DNA Technologies). Sections were counterstained with 49,6-diamidino-2-phenylindole (DAPI) (5 mg/ml) in CitiFluor mounting medium and fluorescently imaged with a fluorescence microscope (Elyra 7; Zeiss) at Â100 magnification (Plan-APOCHROMAT 100Â objective).
(iv) NanoSIMS. Sections were rinsed with deionized water to remove DAPI and mounting medium, and then glass slides were scored with a diamond scribe, broken, and filed to fit into the nanoSIMS sample holder. Sections and slide fragments were sputter coated with 40 nm of gold (Cressington). Areas containing aggregates of interest were presputtered using a primary cesium ion beam at 90 pA (D1 = 1) until 14 N 12 C 2 ion counts stabilized (;5 min). NanoSIMS images were acquired in 10-mm by 10-mm rasters with 128 by 128 pixels with 0.3 pA (D1 = 3, ES = 3) Cs 1 ion beam with a 12-ms/pixel dwell time. Between 20 and 30 10-mm by 10-mm acquisitions were tiled across the aggregate with approximately 2-mm overlap, and the data were manually stitched together postanalysis to create final data products. In addition to the new FISH-nanoSIMS data generated for this study, we also incorporated select nanoSIMS data as a point of comparison from published studies with similar experimental designs (17,46). Regions of interest (ROIs) consisting of individual archaeal and bacterial cells within a consortium were identified and segmented (outlined) by hand using the nanoSIMS 14 N 12 C 2 ion images. Archaeal or bacterial identities for each cell were assigned based on comparison of the nanoSIMS image to the corresponding FISH image. Distances between cells were calculated based on the centroid of each segmented cell in MATLAB.
(v) Cell-specific activity calculation. Growth rates were calculated from nanoSIMS data by (47) m ¼ where m is the growth rate (encompassing both cell maintenance and generation of new cells), T incub is the length of the incubation (7 days), F label is the labeling strength of the nitrogen source provided, , F final is the nanoSIMS measurement, and F nat = 0.0036 is the natural 15 N abundance. The cellspecific metabolic rates (in mol CH 4 cell 21 day 21 ) were calculated as where r is the g cell dry weight per m 3 , B cell is the cell density in m 3 per cell, and Y CH4 is the growth yield in g cell dry weight per mol CH 4 oxidized. See Table 1 for values and sources. Modeling approach. Electron transfer between archaea and bacteria was implemented as a mixed DIET-MIET mechanism where electrons from the oxidation of methane are captured by either redox-active   (2), respectively] Rxn(1) where f M and f D represent the fraction of electron transfer via MIET and DIET, respectively, and Rxn (1) and Rxn(2) are the (unbalanced) overall metabolic reactions of archaea and bacteria. We chose formate (48) as a representative intermediate between ANME and SRB to establish the stoichiometry and thermodynamics, but we recognize evidence that suggests otherwise and note that other small molecules could also be considered the putative intermediates for AOM (19,20,37,(48)(49)(50)(51)(52)(53)(54). This choice affects the energetics and reaction stoichiometries, but, due to similarities arising from diffusion limitations (19,20,55), comparable results are obtained in the context of this study.
For a case where formate is identified as the dissolved electron donor, DH, the reactions [Rxn(3) and Rxn(4), respectively] become Rxn (4) where where F X k represents the reaction kinetics of reaction X and is the product of a cell-specific rate constant, k, the cell density, B cell , and the dependence on substrate availability (19): The thermodynamic factor (0#F X T #1) reflects that there must be sufficient free energy available from the reactions to fuel ATP synthesis and cell maintenance and is given by (56, 57) where n is the number of electrons per reaction, x , the number of ATP synthesized per reaction, is set to 1 (20), R gas is the universal gas constant (8.314 J K 21 mol 21 ), and T is temperature (277.15 K). h ATP represents the potential related to the energy required to synthesize ATP by h ATP = 2DG ATP /nF, where F is the Faraday constant and DG ATP = 210 kJ mol 21 (56,57). The net available potential is given by where h rxn,X is calculated from the Gibbs free energy, DG X , of reaction for archaea [X = R(1)] and bacteria [X = R(2)], and h act and h om are the voltage losses associated with activation and ohmic resistance, respectively. Here, we expand our earlier work (19) by taking into account ohmic resistance and activation loss that ultimately limit metabolic activity away from the archaeal-bacterial interface. Activation loss describes the energetic loss occurring during the electron transfer between cell and conductive pili/matrix. The voltage drops associated with the electron conduction between M and MH can be described by the Butler-Volmer equation assuming a one-step, single-electron transfer process (27). The activation loss, h act , is related to the current density: where I is the current produced by methane oxidation (I = f M R 1 N ANME F, where R 1 is the methane oxidation rate in fmol cell 21 day 21 , N ANME is the number of archaeal cells, and F is the Faraday constant), A act is the redox active surface area in m 2 per cell (27), k act is the activation loss-associated constant in m s 21 , b is the charge transfer coefficient, and M tot is the concentration of electron-carrying molecules (M tot = [M] 1 [MH]). N nw is the total conductive connections within an AOM consortium and can be described as N nw = Electron Transport in AOM Consortia ® M tot V agg k nw , where V agg is the volume of consortium and k nw is the constant associated with conductive network. Conductive network density can be described as N nw,cell = N nw /N ANME . The ohmic loss results from electronic resistance to the flow of electrons through the conductive pili/matrix. The corresponding voltage drop is proportional to current density and is given by (27) Here, R nw is electrical resistance (X), which can be further described as d/(s A nw ), where s is the electrical conductivity of pilus (S m 21 ), d is the distance from archaeal-bacterial interface, and A nw is the cross-section area of a single pilus.
Several of the above-described parameters are poorly constrained experimentally, including the characteristics and concentration of redox active molecules (M tot ), the conductive network density (N nw,cell ), its conductivity (s ), and the various constants (k A , k B , d, and k act ). Other physiological parameters, such as A act , are highly tunable by the cell (27). Thus, it should be noted that the same modeled activity levels and patterns can be achieved for different combinations of these parameters. For instance, decreasing N nw,cell 10-fold can be counterbalanced by increasing conductivity and cell redox active factor, k act ÂA act , by a factor of 10, as is evident from the expressions for activation loss (equation 8) and ohmic resistance (equation 9). To deal with these compensating effects, we identified the key combined parameters of the system and varied those in our simulations. The equations listed above are sensitive to changes in the combined independent parameters, the maximum metabolic activities, k A B A , and k B B B , the maximum cell-specific current, FA act k act M tot , the re- x RgasT . (ii) Implementation. A spherical AOM aggregate was implemented at the center of a domain that represents the surrounding environment with a radius 2.5Â that of the aggregate (r agg ). The spatial distribution of archaea and bacteria in the aggregate (Fig. 1A) was set to reflect the distribution patterns observed in the nanoSIMS analysis (Fig. 1B). A specific cell ratio of 1:1 was set to archaea and bacteria, with the same radii of 0.4 mm for both archaeal and bacterial cells (2,58,59). It is acknowledged that different AOM aggregates may have different cell radii and biovolumes (60), which would impact the estimates of cell-specific rates of the model results reported below.
The concentration fields of CH 4 , HCO 3 2 , CO 2 (aq), CO 3 22 , SO 4 22 , HS 2 , H 1 , OH 2 , HCOO 2 , HCOOH, and B(OH) 4 2 were simulated subject to diffusive transport and reaction, with aqueous diffusion coefficients listed in Table 1. The concentrations at the outer domain boundary were set to fixed concentrations reflecting environmental conditions (Table 1), which were also used as initial conditions. The distribution of MH depends on metabolic rate and electron hopping on conductive pili or matrix. This can be expressed as (61,62) @MH @t where D M = k D M tot d is an effective diffusion coefficient (61,62) that depends on the electron conduction constant (k D ), the distance between two redox-active molecules (d ), and the concentration of electroncarrying molecules, and r·J EF reflects the electron transfer rate driven by a local electric field adapted from (61,62). This flux is given by Rgas T 2e 2 12b ð Þ FEd Rgas T , where k EF is the electric field associated rate constant and E is the electric field strength (61,62).
Acid-base reactions govern the speciation of cell surface-associated immobile carboxy (R-COOH) and amino groups (R-NH 2 ). We considered the dissolved inorganic carbon (DIC) and borate system (63) to quantitatively calculate the carbonate system and dynamically simulate acid-base reactions, using the kinetic implementation described previously (63,64), with a total boron (T B ) concentration of 0.427 mM and total DIC (T DIC ) of 2.36 mM. Archaeal and bacterial cell density and cell size were held constant in all models, with cell numbers varying with AOM consortia radii. The model was implemented in COMSOL Multiphysics 5.4 (COMSOL Inc., Burlington, MA, USA), and simulations were run to steady state.
Baseline simulations presented below use the parameterization shown in Table 1. It was constrained by literature values where available and chosen to yield rates and rate distributions consistent with the observations. Statistical analysis. Data are represented as means 6 standard errors. The statistical difference between the observed and simulated cell-specific activity patterns was assessed by one-way analysis of covariance (ANCOVA) of the slopes of the regression of cell-specific activity versus distance from the archaeal-bacterial interface. P values of ,0.05 were considered statistically significant, whereas P values of .0.05 indicated no statistical significance for the slopes of the regression lines. The statistical analyses were performed using MATLAB 2018 (MathWorks, Natick, MA, USA).

ACKNOWLEDGMENTS
This work was supported by the U.S. Department of Energy, Office of Science, Office of Biological and Environmental Research, Genomic Science Program, under award numbers DE-SC0016469 and DE-SC0020373 (to C.M. and V.J.O.) and DE-SC0016469 (subaward S390693 to C.P.K.) and by a grant from the Simons Foundation collaboration on Principles of Microbial Ecosystems (PriME; to V.J.O.). Samples for this study were