We chose halophilic enrichment cultures which were originally sampled from 2007 until 2013 during microbiological routine analysis at various European UGS and hydrocarbon storages in salt caverns and saline aquifers. Microbial communities from these samples were enriched and maintained with various electron donors and acceptors. The selection was based on microbial activity in general and their ability to grow at hypersaline and sulfate reducing conditions. The initial enrichment procedure is provided in Supplementary Table S1 of the Supplementary Material. These cultures were unified and cultivated for 90 days (Supplementary Table S2) for further enrichment. From the unified samples, three pre-cultures were subsequently prepared at three salinity levels (2.5, 3.4 and 4.4 M NaCl) and cultivated for another 90 days, before subsets with different C sources (acetate, lactate, methanol, no additional) were prepared. These subsets were used as inoculum for the microcosm studies which are described in 2.2. A modified mineral nutrient medium was used, and the composition per L was as follows: 0.5 g K₂HPO₄; 0.43 g MgCl₂∙6H₂O; 0.01 g CaCl₂; 0.5 g (NH₄)₂Cl; 2.8 g Na₂SO₄, 0.5 ml resazurin solution (0.2%), 150 (2.5 M), 200 (3.4 M) or 260 g NaCl (4.4 M) depending on the culture. After autoclaving, the following solutions were added from sterile anoxic stock solutions: 1 ml trace element solution (Widdel et al., 1983), 2 ml vitamin solution (Wolin et al., 1963), 2 ml selenite-tungstate solution (0.02 mM), 30 ml bicarbonate solution (1 M) and 10 ml sodium dithionite solution (0.01 M). The final pH value was set to 7.2 using either anoxic HCl or NaOH. Hundred ml were subsequently dispensed anoxically to 375 ml serum bottles in a N₂ atmosphere. Serum bottles were sealed with butyl rubber stoppers (Ochs Laborfachhandel e. K, Bovenden, Germany) and the headspace (275 ml) was changed to a gas mixture of H₂:CO₂ (8:2) set to 2000 mbar.

The microcosms (S4-S15) were cultivated at three salinity levels (low NaCl concentration with 2.5 M (S4, S7, S10, S13), medium NaCl concentration with 3.4 M (S5, S8, S11, S14), and high NaCl concentration with 4.4 M (S6, S9, S12, S15)), in minimal medium (see above) with 14 mM sulfate and a headspace gas mixture of 80% H₂ and 20% CO₂ (2000 mbar). Each enrichment culture was spiked with one organic C source: 10 mM acetate (S4, S5, S6), 8 mM lactate (S7, S8, S9), or 10 mM methanol (S10, S11, S12). One set was incubated without organic C source (S13, S14, S15). Acetate and lactate were prepared in 2 M stock solutions from their sodium salts at neutral pH value. Both stock solutions and methanol (≥99.8%) were degassed and subsequently sterile-filtered into 120 ml serum bottles using an anaerobic work bench. All cultures were kept in serum bottles, at a temperature of 30°C, and a pH value between 7.2–7.5 for an incubation period of 125 d. During this time, cell numbers, acetate, lactate and sulfate concentrations, as well as the composition of the gas phase were monitored in intervals of 20–25 d. After the incubation period, the microbial community composition was assessed by 16S rRNA gene-based amplicon sequencing. Once samples were subcultured, after a total incubation time of 1445 d, sulfate concentrations were measured again and four samples (S11, S12, S14 and S15) were selected for metagenome sequencing.

Anions (sulfate, acetate, and lactate) were analyzed throughout the experiment, at intervals of 20–25 d. Therefore, sub-samples of 1 mL were aseptically removed from the culture flasks using sterile, disposable syringes, filtered (13 mm syringe filter, PES, 0.22 µm) and stored at −18°C in 1.5 ml reaction tubes until measurement. They were analyzed using a Dionex ICS-1100 ion chromatograph equipped with a Dionex IonPac™ AS23 column (4 × 250 mm) and a conductivity detector (flow rate 1 ml min^-1, sample volume 25 µl). Separation was achieved with gradient elution in 4.5 mM/0.8 mM Na₂CO₃. Organic acids (lactate, acetate) were separated on the same ion chromatograph but using a Dionex IonPac™ AS1 (9 × 250 mm) column and elution with 0.5 mM heptafluorobutyric acid. The volume of culture liquid that was removed for analysis was not refilled.

The composition of the gas phase was monitored throughout the experiment, at intervals of 20–25 d using sub-samples of 0.1–1 ml from the headspace that were aseptically removed from the culture flasks using gastight syringes. Gases were analyzed using a Gach 21.3 gas chromatograph (VEB Chromatron, Berlin) equipped with a thermal conductivity detector. For CO₂ measurements, 100 µl were injected into a hand-packed Al₂O₃ (VEB Leuna-Werke) column (3 m) coated with 20% Oxidipropionitril, H₂ served as carrier gas with a flow rate of 0.8 ml s^-1. For analysis of CH₄ and H₂, a molecular sieve 5A column (3 m; Molsieb 5A 60/80 Mesh, Ziemer Chromatographie) was used with argon as carrier gas at a flow rate of 0.8 ml s^-1. The sampled headspace volume was replaced with an equal volume of the initial gas composition, and gas consumption was calculated accordingly.

Samples were examined microscopically (Zeiss Axioscope A.1, Jena, Germany) to determine the total number of cells and to study cell morphology. Cell counts were determined by means of phase contrast microscopy (magnification 1:1.200) using a THOMA counting chamber.

After an incubation time of 125 days, 2 ml aliquots of each of the 12 microcosms were filtered onto sterile 0.1 μm polyethersulfone filters (PES, Sartorius, Göttingen). Genomic DNA (gDNA) extraction was performed using a previously described phenol-chloroform method (Schwab et al., 2022). Archaeal and bacterial 16S rRNA gene-based Illumina MiSeq amplicon analysis was performed in duplicates at LGC Genomics GmbH (Berlin, Germany) with 2 × 300 bp paired-end sequencing according to manufacturer’s instructions. Briefly, gDNA was amplified using bacterial primer-set Bac-341-fwd/Bac-785-rev (5′-CCTACGGGNGGCWGCAG-3′ and: 5′-GACTACHVGGGTATCTAATCC-3′) with a temperature profile of 95°C for 3 min, followed by 30 cycles of 95, 55°C and 72°C, each step for 30 s, and a final elongation at 72°C for 5 min (Klindworth et al., 2013). In addition, gDNA was amplified in a nested PCR approach, firstly pre-amplifying archaeal sequences with primer-set 340F-fwd/1000R-rev (5′-CCCTAYGGGGYGCASCAG-3′ and: 5′-GGCCATGCACYWCYTCTC-3′) with a temperature profile including an initial denaturation at 96°C for 1 min, followed by 20 cycles of 96°C for 15 s, 50°C for 30 s and 70°C for 90 s. Amplicons (1 μl) of this reaction mixture were taken as a template for the second PCR, using the same temperature profile and the universal primer-set 341F-fwd/806R-rev (5′-CCTAYGGGRBGCASCAG-3′ and 5′-GGACTACNNGGGTATCTAAT- 3′) (Gantner et al., 2011; Sundberg et al., 2013).

Raw adapter-trimmed reads were processed using QIIME2 v2021.2.0 (Bolyen et al., 2019). Briefly, the quality of the reads was assessed with FastQC v0.11.9 (Andrews, 2010) and summarized with MultiQC v1.10.1 (Ewels et al., 2016), before and after primer removal. Primers were removed with cutadapt v3.2 (Martin, 2011), discarding reads shorter than 50 bp. Sequences were denoised, truncated to 270 bases (forward reads) and 240 bases (reverse reads), de-replicated and filtered for chimeras (chimera-method consensus, min-fold-parent-over-abundance 2) using the QIIME2 plugin DADA2 (version q2-dada2 v2021.2.0) (Callahan et al., 2016). Taxonomy was assigned to the resulting amplicon sequence variants (ASVs) with the QIIME2 plugin feature-classifier using a pre-fitted sklearn-based taxonomy classifier (Pedregosa et al., 2011; Bokulich et al., 2018). This classifier was trained against the 16S rRNA gene reference database SILVA (release 138, 27.08.2020). Results were unified in a biom-formatted file (McDonald et al., 2012) analyzed using RStudio with the package phyloseq to combine all reads on the family level. Taxa with a relative sequence read counts below 3% were summarized as “other”. Finally, the results were visualized using OriginPro (OriginLab Corporation, Northampton, MA, United States).

At all salinity levels, an increase of cell number was observed during the incubation period of 125 d (Figures 1A, E, I, M). Without added organic C substrate, acetate (formed by acetogenesis) accumulated at all tested salinity levels with no clear trend with regard to salinity: after 125 days, most acetate was produced at medium salinity, and fewest at low salinity level (Figure 1O). In turn, lactate was converted to acetate (Figures 1G,H), and no enhanced correlation between sulfate consumption and lactate oxidation was observable at medium and high salinity level (Figures 1F,H).

Sulfate was added as potential electron acceptor in all microcosms, and sulfate reduction was observed at all low and medium salinity levels within 125 days (Figures 1B, F, J, N). The rate of sulfate reduction seemed to be affected by both, cell number and type of C substrate (Table 1). In brief, after a lag phase of 20–60 d, sulfate was fully consumed in all low salinity microcosms within 93 d; faster sulfate reduction was observed with acetate and methanol as C sources, than with CO₂ and lactate (compare Figures 1B,J with Figures 1F,N). Complete and faster sulfate consumption was also achieved in medium salinity level with acetate and methanol as C sources. Incomplete sulfate reduction (final sulfate concentration 2 mM and 6 mM after 125 days of incubation) was achieved in medium saline level with CO₂ or lactate as C substrate (Figures 1F,N), while no sulfate reduction was observable during 125 d at high salinity. However, after a total of 1445 days, sulfate was fully consumed in the high salinity setups with organic C source (Table 2, S6, S9, S12). In S15, with CO₂ as sole C source and high salinity, the sulfate concentration was unchanged after 1445 days. Acetate production from CO₂ was observed at low and medium salinity level (Figure 1O), indicating homoacetogenic activity. Furthermore, acetate accumulated at low and medium salinity level, illustrating that acetogenesis was faster than a putative consumption of acetate by other microorganisms in the community. Likewise, acetate was produced at high salinity level but without sulfate reduction (Figures 1N,O). Acetate was also produced from lactate; the data show that lactate was oxidized to acetate prior to sulfate reduction at medium and high salinity level, while at low salinity level lactate oxidation co-occurred with sulfate reduction (Figures 1F–H). Methane was solely produced from methanol at low and medium salinity level (Supplementary Figure S1, Table 1).

The cell numbers increased at all salinity levels over time, and amplicon sequencing data imply that members of the class Halanaerobiia had the highest sequence read abundances (Figure 2); Halanaerobiaceae and Halobacteroidaceae were found at all salinity levels. This finding was underlined with results from metagenome sequencing, which further revealed three MAGs that were assigned at the order level to Halanaerobiales (Table 1). Cell numbers ranged from 1.9 to 4.0∙10⁸ cells ml^-1 at low salinity level, with lactate and methanol being most beneficial for cell proliferation. As expected for slow growing microorganisms, stationary phase was not reached with these C sources during the incubation period of 125 days (Figures 1A,E,I, M). Cell numbers in the medium saline level varied less, ranging from 2.5 to 3.5∙10⁸ cells ml^-1. At low and medium salinity level, halophilic archaea of the family Methanosarcinaceae contributed largely to the sequence read abundance, indicating a high cell number. At the highest salinity level, acetate and lactate clearly supported proliferation, contrary to the same salinity level with the C sources methanol and CO₂ (Figures 1A, E, I, M). Members of the family Desulfohalobiaceae were present at all salinity levels except at highest salinity and only CO₂ as C source. Desulfohalobiaceae phylotypes were not found in the metagenome sequencing approach. Instead, sulfate reducing bacteria (SRB) belonging to the family Desulfonatronovibrionaceae sp. were identified (Figure 2; Table 2).

Metagenome sequencing allowed for closer characterization of the ASVs assigned to Halobacteroidaceae sp.; MAG 2 was identified as Acetohalobium sp. Automatic DRAM annotation and manual tblastN analysis revealed a complete Wood-Ljungdahl pathway for MAG 2 (Table 2). An incomplete Wood-Ljungdahl pathway was identified in MAG 4, which belongs to the Halanaerobiales (Table 2; Figure 3A). Genetic information for autotrophy were further found for the methanogen Methanohalophilus euhalobius (MAG 5), which was dominant at low and medium salinity levels. The lack of the acetate kinase gene and genes associated with H₂ oxidation is in line with the result that methane was solely produced from methanol (Supplementary Figure SI2). In turn, MAG7, identified as a SRB due to the presence of the dissimilatory sulfate reduction pathway, did not fully encode genetic information for autotrophy or lactate-dehydrogenase (Table 2, Supplementary File S1, DRAM Energy), which is in line with the finding that the presence of lactate did not enhance sulfate reduction rates (Figures 1F, H). Lactate-dehydrogenase was found in MAGs 1, 2 and 6.

The proteomic pI value was analyzed to indirectly assess microbial fitness towards osmoadaptation (Figure 3B). Methanohalophilus euhalobius (MAG 5) had the lowest pI value (5.9), due to a small fraction of alkaline proteins (Figure 3B, olive). This low pI value indicates adaptation towards hypersaline conditions through K⁺ accumulation, however corresponding ASVs (Methanosarcinaceae) were absent at high salinity level. The SRB (MAG 7), in contrast, had the highest pI value of 7.1. The pI values of the other MAGs were within this range, but it has to be noted that a high count in acidic proteins was found in Acetohalobium (MAG 2), which exceeded the numbers of the other Halanaerobiia by more than 200 proteins and resulted in a pI value of 6.1.

Three MAGs assigned to the order Halanaerobiales (MAG 1, 3, 4) clustered on a separate phylogenetic branch, embedded between the Halanaerobiaceae and Acetohalobiaceae (Figure 4). Their pI values ranged from 6.7 to 7.0. A fourth MAG, MAG 8, was closely related to a recently described lineage from a saline offshore oil reservoir (Scheffer et al., 2021) and clustered with MAGs 6 and 9, all having similar pI values of 6.4–6.6.

In the microcosms, we observed a conversion of CO₂/H₂ to acetate (Figure 2; Table 2). This could be related to the presence of Acetohabium sp. as these species have been described to use the Wood Ljungdahl pathway to fix CO₂ (Ljungdahl and Wood, 1969; Zhilina and Zavarzin, 1990; Sikorski et al., 2010). Our hydrochemical data (Figure 1) show that acetate was continuously produced at all salinity levels, indicating that acetogenesis was independent from the salinity level. A possible reason for the tolerance of acetogens to highest salt concentrations is that they require less energy for osmoregulation than sulfate reducers. Apparently, less energy can be expected from the thermodynamic calculations with the reduction of CO₂ to acetate then from sulfate reduction (Thauer et al., 1977). In autotrophic homoacetogens, the net production of ATP from substrate level phosphorylation is zero and therefore has to be achieved through the activity of ATP synthase, either fueled by a H⁺ or Na⁺ gradient (Müller, 2003). So far, three types of energy-conserving complexes in acetogens coupled by ion pumping are known, either a ferredoxin:NAD⁺-oxidoreductase, an energy converting hydrogenase and the methyl viologen-reducing hydrogenase complex (Biegel and Müller, 2010; Schoelmerich and Müller, 2019; Kremp et al., 2022). Making use of an energy conserving complex, which at the same time regulates Na⁺ translocation, could be a reason for the efficient growth of Acetohalobium sp. The type-strain Acetohalobium arabaticum DSM 5501 was shown to be capable of producing 25 mM acetate when grown in pure culture at low salinity level, and the salinity range for growth was between 1.7 and 4.3 M NaCl (Zhilina and Zavarzin, 1990). The generation of a proton motive force is also necessary for the mode of osmo-adaptation. Some Halanaerobiales are known to belong to the salt-in strategists, which balance osmotic pressure with high intracellular concentrations of Na⁺, K⁺, Cl⁻ ions, while not exhibiting an acidic proteome pI (Elevi Bardavid and Oren, 2012; Oren, 2013). Our data confirm a similar proteome pI and imply this mode of osmo-adaptation for the Halanaerobiia found in the dataset. This enables these microorganisms to be active even at extremely high salinity levels. However, the cell numbers at the highest salinity level tested in this study was four times lower than observed at the medium salinity level.

All identified SRBs were affiliated as members of the order Desulfovibrionales (Figures 2 and 4), whose genetic information did not encode for a complete CO₂ fixation pathway; thus, they probably rely on organic C sources (Table 2). Both, acetate and methanol had a beneficial effect on sulfate reduction, a finding which is supported by the fact that sulfate was fully consumed after 1445 days also at the highest salinity level tested with these organic C sources. Generally, a possible utilization of acetate as electron donor or C source by SRBs in the cultures during the 125 days incubation period is difficult to evaluate since acetate accumulated, although in different amounts, at all conditions (Figure 1) due to H₂/CO₂ driven acetogenesis or lactate oxidation. Several SRBs are known to oxidize acetate as electron donor or to assimilate acetate when using H₂ as electron donor (Rabus et al., 2015). Lactate, being generally a favorable electron donor for SRBs (Santos et al., 2022), was consumed at all salinity levels, but this resulted in enhanced sulfate reduction rates only at low salinity level at which considerably less acetate accumulated compared to medium and high salinity level. High concentrations of acetate in the microcosms with lactate as C source at medium and high salinity level imply a conversion of lactate to acetate, which was not further used by SRBs. However, no DL-lactate dehydrogenases were encoded in the MAG of this SRB (Supplementary Figure S1), and no sulfate was reduced at high salinity level, hence lactate was seemingly not exclusively utilized by SRB, at least at high salinity level. A reduced sulfate reduction rate at high salt levels was previously shown, in which SRBs in anaerobic microbial co-culture belonged to Desulfonatronospira and underwent sulfate reduction at Na⁺ concentration up to 3.5 M and complete oxidation of 18 mM acetate during 300 d incubation (Sorokin et al., 2014). Other Desulfonatronovibrio spp. (D. thiodismutans, D. magnus and D. hydrogenovorans) grew at Na⁺ concentrations up to 3.0 M and with H₂ as electron donor for sulfate reduction (Sorokin et al., 2011b). Growth at moderate salt concentrations was reported with H₂ and formate or acetate, not CO₂ (Zhilina, 1997; Sorokin et al., 2011b). The proteome pI value of the SRB was the highest of all MAGs found in the microcosms. For halophilic SRB both osmoadaptation strategies are reported and a differentiation was not possible here. However, sulfidogenesis from other S-species than sulfate, such as thiosulfate and elemental sulfur is common at salt concentrations close to saturation (Sorokin et al., 2010). A possibility remains that the SRB can grow as an autotroph, as the Calvin and reverse citric acid cycle were partially found in MAG 7 and the latter being a confirmed C fixation pathway in SRBs (Schauder et al., 1987).

Methanogenesis was only observed when methanol was offered as a C source at low and medium salinity level, a fact known from other saline systems (L’Haridon et al., 2020), as this compound needs less reducing power and is an energy efficient substrate for methanogens. In fact, other halophilic methanogens are reported to lack genes allowing for hydrogenotrophic and acetoclastic methanogenesis (Guan et al., 2019). The presence of Methanosarcinaceae in the low and medium salinity level microcosms without methanol suggests that these microorganisms entered a state of survival rather than growth, while inoculation at high salinity levels during the enrichment process decreased the number of methanogens, whose salt range is reported to be optimal below 3.4 M (Guan et al., 2019; L’Haridon et al., 2020). A potential relationship between the co-occurence of methanogens and SRBs can be drawn as sulfate reduction was accelerated when methanol as C substrate was present. Syntrophy between SRB and methanogens or an unknown sulfate reduction pathway in the methanogens themselves, has already been suggested for other archaea (Milucka et al., 2012).