Candidatus Alkanophaga archaea from Guaymas Basin hydrothermal vent sediment oxidize petroleum alkanes

Methanogenic and methanotrophic archaea produce and consume the greenhouse gas methane, respectively, using the reversible enzyme methyl-coenzyme M reductase (Mcr). Recently, Mcr variants that can activate multicarbon alkanes have been recovered from archaeal enrichment cultures. These enzymes, called alkyl-coenzyme M reductase (Acrs), are widespread in the environment but remain poorly understood. Here we produced anoxic cultures degrading mid-chain petroleum n-alkanes between pentane (C5) and tetradecane (C14) at 70 °C using oil-rich Guaymas Basin sediments. In these cultures, archaea of the genus Candidatus Alkanophaga activate the alkanes with Acrs and completely oxidize the alkyl groups to CO2. Ca. Alkanophaga form a deep-branching sister clade to the methanotrophs ANME-1 and are closely related to the short-chain alkane oxidizers Ca. Syntrophoarchaeum. Incapable of sulfate reduction, Ca. Alkanophaga shuttle electrons released from alkane oxidation to the sulfate-reducing Ca. Thermodesulfobacterium syntrophicum. These syntrophic consortia are potential key players in petroleum degradation in heated oil reservoirs.

In deep seafloor sediments, pressure and heat transform buried organic matter into complex hydrocarbon mixtures, forming natural gas and crude oil 1,2 . n-Alkanes (hereafter referred to as 'alkanes') constitute a major fraction of these mixtures 3 and become energy-rich substrates for microorganisms 4 in habitable anoxic zones. Sulfate-reducing bacteria (SRB) oxidize alkanes ≥ propane (C 3 alkane) 5,6 after activation via fumarate addition through alkylsuccinate synthases 7 . Archaea possess a different mechanism for anaerobic alkane degradation based on reversal of the methanogenesis pathway. This mechanism was first revealed in anaerobic methanotrophic archaea (ANME) 8,9 , which activate methane to methyl-coenzyme M (methyl-CoM) via the key enzyme of methanogenesis methyl-coenzyme M reductase (Mcr) 10 . Recently cultured archaea oxidize non-methane alkanes analogously to ANME, as a first step activating the alkanes to alkyl-CoMs via divergent variants of the Mcr, termed alkyl-CoM reductases (Acrs) 11 . Candidatus Argoarchaeum ethanivorans 12 , Ca. Ethanoperedens thermophilum 13 and Ca. Syntrophoarchaeum spp. 14 oxidize short-chain gaseous alkanes (C 2 -C 4 ), while Ca. Methanoliparum spp., enriched from oil-rich environments, oxidize long-chain alkanes (≥C 16 ) 15 . Similar to most ANME, the short-chain alkane-oxidizing archaea lack respiratory pathways and shuttle the electrons from alkane oxidation to partner SRB 13,14,16,17 . In contrast, Ca. Methanoliparum encodes a canonical Mcr in addition to the Acr, with which it couples alkane oxidation to methanogenesis in a single cell 15 . Article https://doi.org/10.1038/s41564-023-01400-3 slurries produced >10 mM sulfide. Sulfide production was accompanied by dissolved inorganic carbon (DIC) production and sustained during dilution steps ( Fig. 1 and Extended Data Fig. 2), yielding effectively sediment-free cultures after the third dilution. Cultures, except the considerably slower C 5 culture, doubled within 13-40 d (Supplementary Table 1).
According to the general formula the ratio of DIC production to sulfate reduction is ~1. 25-1.30 in case of complete alkane oxidation. In two representative cultures (C 6 and C 14 ), this ratio was slightly lower, with 1.21 ± 0.22 in the C 6 culture and 1.09 ± 0.04 in the C 14 culture. These values suggest that around 10% (C 6 ) and 35% (C 14 ) of the carbon released from alkane oxidation is assimilated into biomass (Supplementary Table 2).

Ca. Alkanophagales archaea are abundant in the cultures
We reconstructed two high-quality archaeal metagenome-assembled genomes (MAGs) from the cultures (Supplementary Table 3 Table 4). Both MAGs were rare (relative abundances ≤0.1%) in the original slurry (Extended Data Fig. 1e,f). The in situ temperatures of the studied sediment (Extended Data Fig. 1d), which captured only the upper sediment layer up to 30 cm depth, probably did not reach the optimal growth temperatures of the two organisms. Both MAGs recruited up to 39% (MAG 4) and 5% (MAG 1) of raw reads in deeper, hotter layers of the Guaymas Basin 26 (Supplementary Table 5). MAGs 1 and 4 represent two species within one genus (average nucleotide identity (ANI) 81.5%) and belong to the same genus as the previously published MAG ANME-1 B39_G2 reconstructed from Guaymas Basin sediments (ANIs: MAG 1-ANME-1 B39_G2 98.8% and MAG 4-ANME-1 B39_G2 80.8%) 27 . The name Ca. Alkanophagales was recently proposed for the clade represented by ANME-1 B39_G2 on the basis of its genomic content which hinted at a capacity for multicarbon alkane metabolism 11,27 . MAGs 1 and 4 form a clade diverging at the root of ANME-1 and next to Ca. Syntrophoarchaeum, together forming the class Syntrophoarchaeia (Fig. 2b).
Visualization of the organisms revealed mixed aggregates of archaea of the Ca. Alkanophagales clade and bacteria ( Fig. 2c-f). These associations resemble those of short-chain alkane-oxidizing cultures 13,14,16 , suggesting that archaea oxidize the alkanes and partner SRB perform sulfate reduction.

The enriched archaea activate alkanes with Acrs
Both Ca. Alkanophagales MAGs encode three Acrs (acrABG) (Extended Data Fig. 3). Currently, only the sister group Ca. Syntrophoarchaeum encodes a higher number of Acrs with four copies 14 . The six acrA sequences, which code for the catalytic subunit 28 , form three clusters of two highly similar sequences, one of each species (≥89% identity) in the acrA clade ( Fig. 3a and Supplementary Table 6) [12][13][14][15] . All clusters are highly similar to acrAs of Ca. Syntrophoarchaeum (Supplementary Table 6).
Both species highly expressed the acrA of the third cluster, placing it among the top 19 (C 8 ) to top 4 (C 5 ) expressed genes ( Fig. 3b and Supplementary Table 7). This cluster is phylogenetically closely related to acrAs that presumably activate long-chain alkanes, for instance in Ca. Methanoliparum 15 . In both MAGs, this acrA is spatially separated from the acrB and acrG subunits (Extended Data Fig. 3), which has been previously reported for Ca. Syntrophoarchaeum 14 .
As in other Acr-dependent alkane-degrading cultures 14,29 , a selective inhibitor of the Mcr/Acr, the CoM analogue Anaerobic archaea capable of petroleum alkane (C 5 -C 15 ) oxidation via Acrs were unknown. These alkanes are the major constituents of gasoline and kerosene 18,19 , and of high ecological relevance because of their toxicity 20,21 . Lately, many acr genes with unknown function have been recovered from environmental metagenomes, especially from hot springs [22][23][24] . We hypothesized that yet uncultured thermophilic archaea could activate petroleum alkanes via Acrs. We aimed to enrich such archaea from heated oil-rich sediment from the hydrothermal vent site Guaymas Basin (Gulf of California, Mexico) 25 . We obtained eight enrichment cultures thriving at 70 °C, in which alkanes from C 5 -C 14 were oxidized in combination with sulfate reduction. Analyses of these cultures via omics approaches and physiological tests revealed that a sister clade of ANME-1, Ca. Alkanophaga, was oxidizing the alkanes after activation via Acrs coupled to sulfate reduction by a partner Thermodesulfobacterium. Such consortia potentially contribute to souring in deeply buried, heated oil reservoirs.

Thermophilic microorganisms thrive on petroleum alkanes
Anoxic slurries produced from heated sediment collected at the hydrothermal vent complex Cathedral Hill in the Southern Trough of the Guaymas Basin (Extended Data Fig. 1a-d) were amended with petroleum alkanes (C 5 -C 14 ) as sole carbon and electron source and sulfate as electron acceptor, and incubated at 70 °C. Within 3-7 months, the   30 , suppressed sulfide production (Extended Data Fig. 4a,b), consistent with an Acr-based activation mechanism. Further, metabolite extracts of all cultures contained peaks pertaining to the masses of the corresponding alkyl-CoMs as indicative activation product (Fig. 3c,d and Extended Data Fig. 5).
While alkanes from C 5 -C 7 were activated at the first and second carbon atom in similar ratios ( Fig. 3c and Extended Data Fig. 5), we observed a shift to more subterminally activated alkanes with increasing alkane length (≥C 9 ) ( Fig. 3d and Extended Data Fig. 5). The longest alkanes C 12 and C 14 seemed to be activated predominantly to ≥3-alkyl-CoM ( Fig. 3d and Extended Data Fig. 5). An activation at multiple positions was previously observed in Ca. Syntrophoarchaeum 14 . The comparatively high activation rate at the terminal position for shorter alkanes is unexpected, because particularly in short alkanes, C-H bonds are stronger at terminal positions compared with subterminal positions 31 . Further degradation of non-terminally activated alkanes probably requires a rearrangement to 1-alkyl-CoM as described for bacterial alkane degradation 32 . We conclude that the archaea represented by MAGs 1 and 4 oxidize the petroleum alkanes. We propose the genus name Ca. Alkanophaga, consistent with the previously suggested name Ca. Alkanophagales 11 , and analogous to the closely related methanotrophs Ca. Methanophagales (ANME-1) 33 . The Ca. Alkanophaga MAGs share amino acid identities (AAIs) of 55-59% with ANME-1 MAGs (Supplementary Table 8), placing Ca. Alkanophaga within the ANME-1 family 34 . On the basis of apparent substrate preference in our enrichment cultures, we propose the names Ca. Alkanophaga volatiphilum for the archaeon represented by MAG 4 and Ca. Alkanophaga liquidiphilum   Fig. 6).

Ca. Alkanophaga completely oxidize the alkanes to CO 2
The oxidation of alkyl-CoMs generated by the Acr requires conversion to acyl-CoA (Fig. 4a,b) The expression of the other acrA genes was low. c,d, Extracted ion chromatograms (EICs) based on exact mass and a window of ±10 mDa of deprotonated ions of variants of C 6 -CoM (c) and C 14 -CoM (d) detected via liquid chromatography-mass spectrometry. In both c and d, the upper shaded panels show the culture extract, with isomers of alkyl-CoM standards below. In d, the shaded bottom panel shows the EIC produced with the exact mass of the C 14 -thiolate, a fragmentation product derived in MS/MS experiments from the precursor C 14 -CoM. Dashed vertical lines were added at retention times of peak maxima of standards (c) or standards and fragmentation products (d) for easier identification of peaks in the culture extracts. While C 6 is activated on the first and second carbon atom to a similar degree, C 14 is activated predominantly to ≥3-C 14 -CoM.
Article https://doi.org/10.1038/s41564-023-01400-3 enzymes have been proposed. The C 2 -oxidizing Ca. Ethanoperedens thermophilum may catalyse this step with tungstate-containing aldehyde:ferredoxin reductases (Aors). This archaeon encodes three aor copies located closely to genes of the Wood-Ljungdahl (WL) pathway and expresses them during ethane oxidation 13 . While both Ca. Alkanophaga encode complete aor gene sets, those genes were only moderately expressed (Supplementary Table 7), casting doubt on a crucial role of the Aor in this reaction in our cultures. A transfer of alkyl moieties to CoA via methyltransferases, as was hypothesized for Ca. Syntrophoarchaeum 14 , is equally unlikely because of the large alkanes consumed by Ca. Alkanophaga. In conclusion, the conversion of alkyl-CoM to acyl-CoA requires further investigation. Similar to Ca. Syntrophoarchaeum 14 , Ca. Alkanophaga probably processes acyl-CoA to acetyl-CoA units via the β-oxidation pathway 35 (Fig. 4b). Ca. Alkanophaga encode all genes for even-chain β-oxidation and expressed them during alkane oxidation (Figs. 4a and 5, Extended Data Fig. 7 and Supplementary Table 7). For odd-chain alkanes, three additional genes are required to degrade the potentially toxic C 3 -compound propionyl-CoA 36,37 , two of which are missing from Ca. Alkanophaga. We could not identify complete alternative pathways for the degradation of propionyl-CoA, for example the methylcitrate cycle 37 . Thus, the fate of the propionyl-CoA remains, for the moment, unclear.
Acetyl-CoA units from β-oxidation are shuttled into biomass production or completely oxidized. For the latter, the acetyl-CoA decarbonylase/synthase (ACDS) complex splits a methyl group from acetyl-CoA which is transferred to tetrahydromethanopterin (H 4 MPT) (  Table 7).
Unlike the closely related Ca. Syntrophoarchaeum and ANME-1, both Ca. Alkanophaga encode several 5,10-methylene-H 4 MPT reductase (mer) genes. This enzyme catalyses the oxidation of methyl-H 4 MPT (CH 3 -H 4 MPT) to methylene-H 4 MPT (CH 2 = H 4 MPT) in the first step of the oxidative WL pathway 38 . Two of these genes, OD814_001315 in Ca. A. volatiphilum and OD815_000385 in Ca. A. liquidiphilum, most probably code for a canonical mer because they are highly similar (>99%) to mer copies of Methanomicrobia. A phylogenetic analysis placed these two mer sequences next to each other and close to those of the hydrogenotrophic methanogens Methanocellales 39 (Extended Data Fig. 8a). We therefore hypothesize that Ca. Alkanophaga inherited mer vertically from the methanogenic ancestor of Methanocellales. Ca. Syntrophoarchaeum and ANME-1 seem to have replaced mer with methylene-tetrahydrofolate (H 4 F) reductase (metF) of the H 4 F methyl branch of the WL pathway 14,40 . Both Ca. Alkanophaga MAGs also encode metF copies, which are highly similar (70-80%) to those of Ca. Syntrophoarchaeum and cluster next to metF sequences of Hadarchaeota from Jinze hot spring (China) and Yellowstone National Park (USA) (Extended Data Fig. 8b). While both mer and metF were transcribed, mer was especially expressed by Ca. A. liquidiphilum in cultures oxidizing longer alkanes ≥C 10 ( Fig. 5b,d, Extended Data Fig. 7e,f,k,l and Supplementary Table 7).

Ca. Alkanophaga partner with a Thermodesulfobacterium
Ca. Alkanophaga lack the dissimilatory sulfate reduction (DSR) pathway and therefore require a partner organism. We identified a Thermodesulfobacterium represented by MAG 24, which was enriched in all cultures ( Fig. 2a and Supplementary Table 4) and rare in the original slurry, as the most likely syntrophic sulfate reducer. MAG 24 encodes and expressed the three DSR proteins ATP-sulfurylase (Sat), APS-reductase (Apr) and dissimilatory sulfite reductase (Dsr) 41 (Fig. 5, Extended Data Fig. 7 and Supplementary Table 7). We propose the name Ca. Thermodesulfobacterium syntrophicum for this bacterium, which is closely related to the hyperthermophilic sulfate-reducing Thermodesulfobacterium geofontis isolated from the Obsidian Pool in Yellowstone National Park (USA) 42 (Extended Data Fig. 9).  Table 7). However, the addition of hydrogen or formate did not accelerate sulfide production (Extended Data Fig. 4c,d). Moreover, cultures in which sulfate reduction was inhibited by the addition of sodium molybdate produced only miniscule fractions (max. 2.4% for C 6 and 0.9% for C 14 ) of the hydrogen concentrations that would be necessary were hydrogen the sole electron carrier (Supplementary Table 9). Thus, neither molecular hydrogen nor formate are probably primary electron carriers.

Etymology
Alternatively, alkane oxidation and sulfate reduction are coupled through DIET, as suggested for other alkane-oxidizing consortia 13,14,45 . DIET probably involves cell appendages, such as bacterial type IV pilin (PilA) or the archaeal flagellin B (FlaB), and multihaem c-type cytochromes (MHCs), forming conductive nanowires enabling electron transport 46,47 . Both components are present and strongly expressed in previously established alkane-oxidizing consortia 13,14,45 . Surprisingly, neither our nor the previously published Ca. Alkanophaga MAGs encode any MHCs, while the closest relatives of Ca. Alkanophaga, ANME-1 and Ca. Syntrophoarchaeum, encode multiple MHCs 14,33 . Ca. T. syntrophicum encodes six MHCs, only one of which was slightly enriched in all cultures (Supplementary Table 7). This implies a minor role of MHCs in the interaction of both organisms.
Both Ca. Alkanophaga encode several copies of pilA and flaB for the formation of cell appendages for DIET. These genes were among the most highly expressed genes of Ca. Alkanophaga in all cultures. Ca. T. syntrophicum encodes several pilA genes as well, some of which were strongly enriched in the C 10 -C 14 cultures (Supplementary Table 7). Transmission electron microscopy revealed diffuse filamentous structures in the intercellular space that might pertain to such nanowires (Extended Data Fig. 10), but further analyses are necessary to confirm the identity of these structures.
We predict that electron transfer in our cultures is based predominantly on DIET. The lack of MHCs in Ca. Alkanophaga might be compensated by MHC production in the partner bacterium similar to observations in syntrophic methane-oxidizing cultures, where only the bacterial partner expressed pilA genes 45 . Alternatively, DIET might be completely independent of MHCs, which has been observed before 48,49 . It remains possible that a small fraction of electrons are transferred via soluble intermediates such as hydrogen. Such a combination of Article https://doi.org/10.1038/s41564-023-01400-3 DIET with diffusion-based electron transport was recently shown to be energetically favourable for syntrophic consortia 50 .

Discussion
Petroleum-rich anoxic environments such as oil reservoirs, oily sludges and polluted sediments harbour oil-degrading microorganisms. Isolates from these environments that couple petroleum alkane oxidation to sulfate reduction are mostly bacteria active at temperatures ≤60 °C (ref. 51). With Ca. Alkanophaga, we enriched a thermophilic clade thriving on petroleum alkanes from C 5 to C 14 at temperatures between 65-75 °C (Extended Data Fig. 4e,f), which approach the suggested upper limit of microbial hydrocarbon degradation in petroleum reservoirs of around 80 °C (ref. 52). This temperature optimum is reflected by the high relative abundance of Ca. Alkanophaga in deep, heated sediment  Article https://doi.org/10.1038/s41564-023-01400-3 layers of the Guaymas Basin, inferring a crucial role of these archaea in thermophilic hydrocarbon transformation. Ca. Alkanophaga encode three Acrs for anaerobic alkane activation, one less than the closely related short-chain alkane oxidizer Ca. Syntrophoarchaeum 14 . Independent of alkane length, Ca. Alkanophaga strongly expressed only one of the Acrs, which is highly similar to the highest expressed Acr in Ca. Syntrophoarchaeum during C 4 oxidation 14 . Future studies may reveal functions or substrates of the other two lower expressed Acrs. Ca. Alkanophaga stand out among Acr-using archaea with their wide substrate range between C 5 and C 15 . Therewith, all alkanes between C 1 and C 20 are confirmed substrates of alkane-oxidizing archaea [12][13][14][15] . Our study implies that substrate flexibility of the Acr increases with increasing alkane length, which is presumably enabled by a wider catalytic cleft in the Acrs activating C 3 + alkanes 31 compared with the highly selective hydrophobic tunnel detected in the C 2 -activating Acr 29 . Crystallization efforts may resolve molecular and structural modifications of these Acrs that make use of such a wide substrate spectrum.
The three clades of the class Syntrophoarchaeia (Ca. Alkanophaga, Ca. Syntrophoarchaeum and ANME-1), share many metabolic features such as obligate syntrophic growth with partner SRB and presence of the β-oxidation and WL pathways. At the same time, they exhibit remarkable metabolic and genomic differences. For instance, ANME-1 encode the canonical Mcr for methane metabolism, which is missing in Ca. Syntrophoarchaeum and Ca. Alkanophaga, preventing them from oxidizing and producing methane. Instead, the latter two possess multiple multicarbon alkane-activating Acrs, which are in turn absent in ANME-1. Our study supports the previously established hypothesis that multicarbon alkane metabolism probably preceded methanotrophy in the Syntrophoarchaeia 11,53 because of the basal position of both multicarbon alkane oxidizers (Fig. 2b) and their similar metabolisms. The presence of the β-oxidation pathway in ANME-1 (ref. 11) supports this notion because this pathway is required for the oxidation of C 3 + alkanes but serves no purpose in the oxidation of methane. We propose that the common ancestor of the Syntrophoarchaeia was a multicarbon alkane-oxidizing archaeon with multiple Acrs. Ca. Syntrophoarchaeum   and Ca. Alkanophaga emerged from this ancestor, preserving a similar metabolism. Today, Ca. Syntrophoarchaeum thrives at much lower temperatures (50 °C) and seems incapable of oxidizing liquid alkanes 14 .
Thus, adaptation to different temperatures and substrates might have enabled Ca. Syntrophoarchaeum and Ca. Alkanophaga to occupy different ecological niches. Ca. Alkanophaga and ANME-1 also shared a common ancestor from which ANME-1 probably diverged after losing their Acrs 54 and acquiring an Mcr, potentially from a methanogen via lateral gene transfer 33,55 .
Ca. Alkanophaga differ from the two other groups of the Syntrophoarchaeia in two main aspects. First, Ca. Alkanophaga encode and expressed mer, an essential enzyme of the canonical methanogenesis pathway 56 . ANME-1, except for a putative methanogenic ANME-1 member 57 , and Ca. Syntrophoarchaeum lack mer and instead code for the phylogenetically widely distributed metF 14,33,58 , which is also present and expressed in Ca. Alkanophaga. We hypothesize that mer in Ca. Alkanophaga is a remnant from a methanogenic ancestor. Second, Ca. Alkanophaga lack MHCs, which are often considered essential for DIET between syntrophic partners 46 . All other syntrophic alkane-oxidizing archaea code for several MHCs 53 . However, an absence of MHCs in DIET-performing methanogens has been recognized before 48 . It is thus conceivable that MHCs aid in but are not essential for DIET and that MHCs were potentially lost by Ca. Alkanophaga without a substantial impact on the efficiency of electron transfer. The loss of all MHCs opens up questions as to the mechanisms that occurred. In a recent study, giant extrachromosomal elements named Borgs, many of which carried clusters of MHCs, were reconstructed from methane-oxidizing Methanoperedens (ANME-2d) archaea 59 . One could imagine that MHCs in the Syntrophoarchaeia ancestor were encoded on such a Borg, which was then lost by Ca. Alkanophaga. This could explain why all MHCs are absent in Ca. Alkanophaga. However, the presence of Borgs in other members of the Syntrophoarchaeia still needs to be examined.
Ca. Alkanophaga partner with the sulfate-reducing Ca. Thermodesulfobacterium syntrophicum. Previously enriched alkane-oxidizing archaea partner with a different bacterium, Ca. Desulfofervidus auxilii, which has an optimal growth temperature of 60 °C (refs. 13,14,16,17). We suspect that the higher incubation temperature of our study selected for a more thermophilic partner organism. Recently, another Thermodesulfobacterium species, Ca. Thermodesulfobacterium torris (ANI 84.0%, Extended Data Fig. 9), has been reported as syntrophic sulfate reducer partnering with thermophilic ANME-1c at 70 °C (ref. 60). Thus, Thermodesulfobacteria represent a new group of partner organisms for alkane-oxidizing archaea at high temperatures. In contrast to Ca. Alkanophaga, Ca. T. syntrophicum encodes and expressed several MHCs, which could support DIET for both partners.
All currently available Ca. Alkanophaga sequences originate from the Guaymas Basin, a thoroughly studied hydrothermal vent area hauling heated fluids rich in alkanes 61 . We suspect two main reasons for this apparent absence in other environments. First, until recently, microbial community studies have mostly focused on 16S ribosomal (r)RNA gene amplification and sequencing, a method depending heavily on primer choice 62 . We discovered a mismatch of the commonly used archaeal primer Arch915 (5′-GTGCTCCCCCGCCAATTCCT-3′ 63 , mismatch in bold) to the 16S rRNA gene sequences of Ca. Alkanophaga, which probably produces an artificial underrepresentation of Ca. Alkanophaga in public databases. Second, sequencing data from other environments similar to the Guaymas Basin, that is, heated oil reservoirs with sulfate supply, remains scarce. Many of these reservoirs, often buried kilometres deep within the subsurface, are extremely hard to access 64 . In addition, the risk of contamination from the upper biosphere during sampling increases with depth, which might conceal the native community 64 . Still, sampling technologies have greatly improved in recent years, and the focus has shifted from amplification-based 16S rRNA gene to shotgun metagenome studies, which should facilitate a more accurate molecular characterization of reservoir microorganisms.
Thus, future studies may disclose the coexistence and activity of Ca. Alkanophaga and Ca. T. syntrophicum in other heated, petroleumrich environments.

Methods
All chemicals were of analytical grade and obtained from Sigma Aldrich, unless otherwise stated. All incubations were done under gentle shaking (40 r.p.m.) in the dark.
Sulfide production was measured every 2-4 weeks using a copper sulfate assay 67 . Once sulfide concentrations reached 12-15 mM, cultures were diluted 1:3 with SRM and supplied with fresh substrate. Activity doubling times were determined from the development of sulfide concentrations during the first two dilutions. Sulfide concentrations over time were displayed using a logarithmic (base 2) y axis. An exponential trend line with the formula y = n × e mx was generated. Per definition, the doubling time equals

Quantitative substrate turnover experiment
Triplicate 100 ml dilutions with 20 ml headspace were prepared from C 6 -and C 14 -oxidizing cultures, supplied with substrate and incubated at 70 °C, complemented by a substrate-free negative control. Sulfate and DIC concentrations were measured from weekly subsamples until the cultures had reached sulfide concentrations of ≥15 mM. Samples were sterile filtered using a GTTP polycarbonate filter (0.2 µM pore size; Millipore). For DIC measurements, 1 ml filtrate was transferred into synthetic-air-purged 12 ml Exetainer vials (Labco) filled with 100 µl phosphoric acid (45%). After 10 h of equilibration, headspace DIC was measured by isotope ratio infrared spectroscopy (Thermo Fisher; Delta Ray IRIS with URI connect and Cetac ASX-7100 autosampler) with standards of known concentration. To determine sulfate concentrations, 1 ml of the filtrate was fixed in 0.5 ml 100 mM zinc acetate. The sample was centrifuged and the clear supernatant was diluted 1:50 in deionized water. Sulfate was measured by ion chromatography (930 compact IC, Metrohm) against standards with known concentrations.

DNA extraction and short-read sequencing
DNA was extracted from pellets of 25 ml culture samples collected after the third dilution, using a modified SDS-based extraction method as previously described 68 . Total DNA yield per sample, determined by fluorometric DNA concentration measurement, ranged from 0.9 µg to 3.6 µg. Samples were sequenced at the Max Planck-Genome-Centre (Cologne, Germany). C 6 -C 14 culture samples were sequenced as 2 × 250 paired-end reads on an Illumina HiSeq2500 sequencing platform. The C 5 culture sample was sequenced later because of slower growth, together with a sample of the sediment slurry before incubation, by which time the sequencing facility had changed their settings to 2 × 150 bp paired-end reads on an Illumina HiSeq3000 platform. Between 4,140,953 and 4,234,808 raw reads were obtained per culture sample. From the original slurry, 3,130,329 reads were gained. Article https://doi.org/10.1038/s41564-023-01400-3

Short-read DNA data analysis
Reads from short-read metagenome sequencing were quality-trimmed using BBDuk (included in BBMap v.38.79; https://sourceforge.net/projects/bbmap/; minimum quality value: 20, minimum read length: 50). Reads of the C 6 -C 14 samples were coassembled using SPAdes (v.3.14.0; https://github.com/ablab/spades) 69 , running BayesHammer error correction and k-mer increments (21,33,55,77,99 and 121) with default settings. The output scaffolds were reformatted using anvi'o (v.7; https:// github.com/merenlab/anvio/releases/) 70 , simplifying names and removing contigs shorter than 3,000 bps. Trimmed reads were mapped back to the reformatted scaffolds using Bowtie2 (v.2.3.2; http://bowtie-bio. sourceforge.net/bowtie2/index.shtml) 71 in the local read alignment setting. Sequence alignment map files were converted to binary alignment map (BAM) files with SAMtools (v.1.5; http://samtools.sourceforge. net/) 72 73 . The contigs database was inspected in the anvi'o interactive interface, which clusters the contigs hierarchically on the basis of sequence composition and differential coverage, thereby indicating their relatedness to each other 70 . Binning was performed manually in the interface by clicking branches of the dendrogram in the centre of the interface and using the GC content, mean coverage in the samples, gene taxonomy and real-time statistics on completion and redundancy based on single-copy core genes as guides. The dendrogram branches were followed systematically in a counterclockwise direction to obtain the maximum number of bins. Bin quality was assessed again with CheckM (v.1.1.3; https:// ecogenomics.github.io/CheckM/) 74 and only bins with completeness >50% and redundancy <10% were kept. Taxonomies were assigned to these metagenome-assembled genomes (MAGs) using GTDB-Tk (v.1.5.1; https://github.com/Ecogenomics/GTDBTk) 75 . All manually generated MAGs were refined with anvi'o to minimize contamination. We identified MAGs 1 and 4 as the likely alkane oxidizers and MAG 24 as the likely sulfate reducer based on their mean coverages and HMM hits. To increase the completeness of these three MAGs, an iterative reassembly loop (https://github.com/zehanna/MCA70_analysis/targeted_reas-sembly_loop.sh) was performed. Therein, the trimmed reads were repeatedly mapped to the refined MAG using BBMap with a minimum alignment identity of 97%. Mapped reads were then assembled using SPAdes. The assembly was quality-checked with CheckM and used as a new reference file to map the trimmed reads to. After performing 25 iterations of this loop, the assembly with the highest quality (that is, highest completeness, lowest contamination and lowest strain heterogeneity) was selected for further analysis. Final MAGs were annotated with Prokka (v.1.14.6; https://github.com/tseemann/prokka) 76 and the anvi'o-integrated databases NCBI clusters of orthologous genes (COGs) 77 , Kyoto Encyclopedia of Genes and Genomes (KEGG) 78 , Protein Families (Pfams) 79 and KEGG orthologues HMMs (KOfams) 80 . A bash script (https://github.com/zehanna/MCA70_analysis/CxxCH_scan.sh) was run to search for the haem-binding CxxCH amino acid motif 81 in the translated gene sequences of the three MAGs. Selected translated gene sequences were exported for gene calls from the contigs database with anvi'o and compared via the BLASTp 82 web interface (http://www.ncbi. nlm.nih.gov/blast).
Because of later sequencing, the original slurry and C 5 samples were treated separately from the previously sequenced samples and assembled individually. We could not obtain quality MAGs for the original slurry sample; therefore, we estimated the phylogenetic composition on the basis of reconstructed small subunit ribosomal RNAs (SSU rRNAs) mapped against the SILVA SSU reference database (v.138.1) 83 with phyloFlash (v.3.4.1; https://github.com/HRGV/phyloFlash) 84 . For the C 5 sample, the same procedure as for the previously sequenced culture samples was followed. The identity (ANI ≥ 95%; ref. 85) of the Ca. Alkanophaga volatiphilum and Ca. Thermodesulfobacterium syntrophicum MAGs from the C 5 sample, MAG 4_1 and MAG 24_1, respectively, to the previously reconstructed ones was confirmed via FastANI.
To estimate relative abundances of Ca. Alkanophaga and Ca. T. syntrophicum MAGs in the original slurry, the trimmed reads of the original slurry were mapped to the MAGs with CoverM.

Construction of phylogenomic trees for archaea and bacteria
The archaeal tree was constructed using 98 publicly available Halobacteriota and Thermoproteota genomes (Supplementary Table 10) from NCBI plus the Ca. Alkanophaga MAGs from this study. For the bacterial tree, 121 publicly available Desulfobacterota and Bipolaricaulota genomes (Supplementary Table 10) and the Thermodesulfobacterium MAG from this study were included. Trees were based on the concatenated alignment of 76 single-copy core genes (SCG) for archaea and 71 SCGs for bacteria. Alignments were generated with anvi'o, which uses the multiple sequence alignment tool MUSCLE 86 (v.5.1; https:// github.com/rcedgar/muscle). Trees were calculated with RAxML (randomized accelerated maximum likelihood) (v.8.2.12; https://cme.h-its. org/exelixis/web/software/raxml/) 87 using the PROTGAMMAAUTO model and autoMRE option, which required 50 iterations to reach a convergent tree for both alignments. Trees were visualized with the Interactive Tree of Life online tool (https://itol.embl.de/) 88 . To resolve taxonomic levels, the Ca. Alkanophaga MAGs were compared to the ANME-1 and Ca. Syntrophoarchaeales MAGs included in the tree by calculating average amino acid identities (AAIs) using the aai_wf feature of the CompareM software (v.0.1.2; https://github.com/dparks1134/ CompareM) with default settings.

Phylogenetic analysis of proteins involved in alkane oxidation in Ca. Alkanophaga
For the mcrA tree, the six full-length mcrA sequences of Ca. Alkanophaga were aligned with 347 publicly available mcrA sequences. For the mer and the metF trees, Ca. Alkanophaga sequences were added to publicly available alignments in ref. 33 92 . For the mcrA tree, sequences shorter than 450 amino acids were removed after trimming, after which 337 sequences remained (Supplementary Table 10). Trees were calculated with RAxML (v.8.2.4) using the PROTGAMMAAUTO model, which assigned LG with empirical base frequencies as amino acid model and the autoMRE option for bootstraps, which required 300, 550 and 400 iterations to reach a consensus tree for the mcrA, mer and metF alignments, respectively. Trees were visualized with the Interactive Tree of Life online tool (https://itol.embl.de/) 88 .

RNA extraction and short-read sequencing
For total RNA extraction, 10 ml of culture material collected after the third dilution at the exponential growth stage were filtered through an RNAse-free cellulose nitrate filter (pore size 0.45 µm; Sartorius). Immediately after filtration, filters were incubated with 5 ml RNAlater for 30 min. RNA was extracted from filters using the Quick-RNA miniprep kit (Zymo Research). DNA was digested without RNase inhibitor. No rRNA depletion step was performed. Between 0.3 and 1.3 µg of total RNA were obtained per sample as determined by fluorometric RNA concentration measurement. Samples were sequenced as 2 × 250 (C 5 : 2 × 150) paired-end reads at the Max Planck-Genome-Centre on the Illumina HiSeq2500 (C 5 : Illumina HiSeq3000) sequencing platform. Between 4,043,349 and 4,785,231 raw reads were obtained per sample.

Short-read RNA data analysis
Reads from metatranscriptome sequencing were quality-trimmed using BBDuk (included in BBMap v. 38.79). Trimmed reads were mapped to the concatenated Ca. Alkanophaga MAGs to minimize unspecific mapping because of the high similarity of the two MAGs and to the Ca. Thermodesulfobacterium syntrophicum MAG using BBMap (v.38.87) with minimal alignment identity of 98%. Mapped reads were counted using featureCounts (v.1.4.6-p5; http://subread.sourceforge.net/) 93 with minimum required number of overlapping bases and minimum mapping quality score of 10, counting fragments instead of reads.
Before normalization, rRNA reads were excluded. Fragments were first normalized to gene length, yielding fragments per kilobase (FPK).
The centred-log ratio (CLR) was calculated as the base-10 logarithm of read count C i of gene i normalized by gene length L i in kilobases and divided by the geometric mean of all read counts C 1 − C n normalized by their respective gene length L 1 − L n .

Test of a selective Mcr inhibitor on culture activity
Duplicates of C 6 -and C 14 -oxidizing culture were supplied with substrate and 5 mM (final concentration) BES. A control culture was supplied with substrate but not with BES. Cultures were incubated at 70 °C and sulfide concentrations were measured until the control cultures had reached >15 mM sulfide.

Metabolite extraction
Metabolite samples were collected at sulfide levels of 10-14 mM. An 80 ml culture sample of each substrate was pelleted via centrifugation (15 min, 3,100 × g, 4 °C). Supernatants were removed, pellets were resuspended in 1 ml of acetonitrile:methanol:water (2:2:1 v/v/v) and transferred to bead-beating tubes. Samples were agitated for 15 min on a rotor with vortex adapter at maximum speed. Samples were centrifuged for 20 min at 10,000 × g at 4 °C. Clear supernatants were stored at 4 °C.

Synthesis of authentic alkyl-CoM standards
Coenzyme M (sodium 2-mercaptoethanesulfonate) (0.1 g) was dissolved in 2 ml 25% (v:v) ammonium hydroxide solution and twice the molar amount of bromoalkane was added. We acquired 2-and 3-bromohexane from Tokyo Chemical, and 2-bromotetradecane from Alfa Aesar. Vials were incubated for 6 h at r.t. under gentle shaking on a rotor with vortex adapter. The clear upper phase (1 ml) was collected and stored at 4 °C.

Mass spectrometry of culture extracts and standards
Culture extracts and standards were analysed using high-resolution accurate-mass mass spectrometry on a Bruker maXis plus quadrupole time-of-flight (QTOF) mass spectrometer (Bruker) connected to a Thermo Dionex Ultimate 3000RS UHPLC system (Thermo Fisher) via an electrospray ionization (ESI) ion source. Sample aliquots were evaporated under a nitrogen stream and re-dissolved in a methanol:water (1:1 v/v) mixture before injection. A 10 µl aliquot of the metabolites was injected and separated on an Acclaim C30 reversed phase column (Thermo Fisher; 3.0 × 250 mm, 3 µm particle size) set to 40 °C using a flow rate of 0.3 ml min −1 and the following gradient of eluent A (acetonitrile:water:formic acid, 5:95:0.1 v/v/v) and eluent B (2-propanol:acetonitrile:formic acid, 90:10:0.1 v/v/v): 0% B at 0 min, then ramp to 100% B at 30 min, hold at 100% B until 50 min, followed by re-equilibration at 0% B from 51 min to the end of the analysis at 60 min to prepare the column for the next analysis. The ESI source was set to the following parameters: capillary voltage 4,500 V, end plate offset 500 V, nebulizer pressure 0.8 bar, dry gas flow 4 l min −1 , dry gas heater 200 °C. The QTOF was set to acquire full scan spectra in a mass range of m/z 50-600 in negative ionization mode. The C 14 culture extract was additionally analysed in tandem mass spectrometry mode, and mass spectra of the fragmentation products of m/z 337.1877 isolated in a window of 3 Da and fragmented with 35 eV were acquired. Every analysis was mass-calibrated to reach mass accuracy of 1-3 ppm by loop injection of a calibration solution containing sodium formate cluster ions at the end of the analysis during the equilibration phase and using the high-precision calibration algorithm. Data were processed using the Compass DataAnalysis software package v.5.0 (Bruker). Article https://doi.org/10.1038/s41564-023-01400-3

Substrate range tests
Cultures originally grown with C 6 and C 14 were diluted 1:10 in fresh SRM. Dilutions were supplemented with alkanes between C 5 and C 14 for which growth had not been confirmed yet, and with shorter (C 3 and C 4 ) and longer (C 16 -C 20 ) alkanes (Table 1).
A negative (inoculated culture without substrate) and a positive (inoculated culture supplied with substrate with which the culture was originally grown) control were also set up. Cultures were incubated at 70 °C and activity was tracked via sulfide measurements. Once sulfide concentrations reached >10 mM, cultures were diluted 1:3 with SRM. The procedure was repeated and incubations that showed sustained activity over two dilutions were considered successful.
Hydrogen production measurements C 6 and C 14 cultures were divided into two 20 ml aliquots in 156 ml serum bottles. One aliquot was left untreated, the other one was treated with 10 mM (final concentration) sodium molybdate. Hydrogen was measured by injecting 1 ml of headspace sample into a Peak Performer 1 gas chromatograph (Peak Laboratories). Measurements were taken in 1 h intervals up to 8 h after the start of the experiment. A final measurement round was conducted from 24 h to 30 h in 2 h intervals.

Test of the effect of addition of hydrogen and formate on culture activity
Two replicates of C 6-and C 14 -oxidizing cultures were supplied with substrate and with 10% H 2 in the headspace or 10 mM (final concentration) sodium formate in the medium. A control culture was supplied only with substrate. Cultures were incubated at 70 °C and sulfide concentrations were measured until the control cultures had reached ≥15 mM sulfide.
Transmission electron microscopy C 6 and C 14 culture (100 ml) were collected at 1,000 × g using a Stat Spin Microprep 2 table top centrifuge. Cells were transferred to aluminium platelets (150 µm depth) containing 1-hexadecene 94 . Platelets were frozen using a Leica EM HPM100 high-pressure freezer (Leica). Frozen samples were transferred to a Leica EM AFS2 automatic freeze substitution unit and substituted at −90 °C in a solution containing anhydrous acetone and 0.1% tannic acid for 24 h, and in anhydrous acetone, 2% OsO 4 and 0.5% anhydrous glutaraldehyde (Electron Microscopical Science) for a further 8 h. After further incubation over 20 h at −20 °C, samples were warmed to +4 °C and subsequently washed with anhydrous acetone. Samples were embedded at room temperature in Agar 100 (Epon 812 equivalent) at 60 °C for 24 h. Thin sections (80 nm) were counterstained using Reynolds lead citrate solution for 7 s and examined using a Talos L120C microscope (Thermo Fisher).

Temperature range tests
Aliquots of C 6 -and C 14 -oxidizing cultures were diluted 1:6, supplied with substrate and incubated at 60-90 °C in 5 °C increments. Sulfide production was tracked until the 70 °C cultures had reached >10 mM sulfide.

Availability of biological materials
Official culture collections do not accept syntrophic enrichment cultures, but G.W. will maintain the cultures. Non-profit organizations can obtain samples upon request.

Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Code availability
The workflow for metagenome and transcriptome analysis, and the scripts for targeted reassembly and for the search of CxxCH motifs are available under https://github.com/zehanna/MCA70_analysis. Further inquiries about bioinformatic analyses may be directed to the corresponding authors. Tested C x -n-alkanes Ca. Alkanophaga volatiphilum C 5 , C 6 , C 7 C 6 C 3 , C 4 , C 8 , C 9 , C 10 , C 11 , C 12 , C 13 , C 14 , C 15 , C 16 , C 18 , C 20 Ca. Alkanophaga liquidiphilum C 8 , C 9 , C 10 , C 12 , C 14 C 14 C 3 , C 4 , C 11 , C 13 , C 15 , C 16 , C 18 , C 20 Extended Data Fig. 1 | Sampling site in the Guaymas Basin and microbial community in the original sediment. a, Location of the Guaymas Basin in the Gulf of California. b, Bathymetry of the southern end of the Southern Trough of the Guaymas Basin with the location of the Cathedral Hill hydrothermal vent area. c, Sampling of the push core (4991-15) that was used for anoxic cultivations in an area densely covered by orange sulfur-oxidizing Beggiatoa mats. d, Depthtemperature profile in the sampling site. The temperature was measured using Alvin's heatflow probe. Push cores reached about 30 cm into the sediment, where the temperature approached about 60 °C (sampling site photograph and temperature data courtesy of the Woods Hole Oceanographic Institution, from RV Atlantis cruise AT42-05). e, f Microbial community in the anoxic sediment slurry prepared from core 4991-15 before starting anoxic incubations based on 16S rRNA gene fragments recruited from the metagenome. e, On the domain level, archaea and bacteria each make up around 50%. f, Taxonomic groups on order level. For groups with unknown order assignment marked with *, the next known higher taxonomic levels are indicated. An ANME-1 group is abundant within the archaeal fraction while the bacterial fraction is very diverse. Fig. 5 | Detection of alkyl-CoMs in C x -n-alkane-degrading cultures. Samples were separated by liquid chromatography and extracted ion chromatograms (EICs) based on the exact mass of deprotonated ions of the C x -alkyl-CoMs with a window of ±10 mDa were created. Panels show the EICs of culture extracts together with synthetic standards. Dashed vertical lines were added at the retention times of peak maxima of the standards for easier identification of peaks in the culture extracts. Peaks with mass-to-charge ratios (m/z) of the respective alkyl-CoM were detected in all cultures. All culture extracts show several peaks, indicating an activation at different carbon atoms. While shorter alkanes are activated to a similar degree at subterminal and terminal positions, longer alkanes are predominantly activated at non-terminal carbon atoms.