The soil samples were from the research stations in Vercelli, Italy, and the International Rice Research Institute (IRRI) in the Philippines. Sampling and soil characteristics were described before (Liu et al., 2018). The lake sediments (top 10 cm layer) were from the NE and SW basins of Lake Fuchskuhle, an acidic bog (pH 4.2–4.6) lake in northeastern Germany (Casper et al., 2003). The lake was artificially divided into four compartments in an ongoing process from 1987 to 1991, resulting in four nearly equal-sized compartments, each with a different catchment area. The NE basin is characterized by higher biomass and activity throughout all trophic levels in the water column than those of the SW basin. The lake sediments were sampled in July 2016 using a gravity core sampler as described before (Kanaparthi et al., 2013). The experiments with rice field soil were carried out in 2016, while the experiments with sediments of Lake Fuchskuhle were carried out in 2017.

The experimental setup was exactly the same as during previous studies of acetate consumption (Conrad et al., 2021) and propionate consumption (Conrad and Claus, 2023). For methanogenic conditions, paddy soil was mixed with autoclaved anoxic H2O (prepared under N2) at a ratio of 1 : 1 and incubated under N2 at 25 °C for 4 weeks. In a second incubation, for sulfidogenic conditions, paddy soil was mixed with autoclaved anoxic H2O at a ratio of 1 : 1, was amended with 0.07 g CaSO4 ⋅ 2H2O, and was then incubated under N2 at 25 °C for 4 weeks. These two preincubated soil slurries were sampled and stored at -20 °C for later molecular analysis (see data in Conrad et al., 2021). The preincubated soil slurries were also used (in three replicates) for the following incubation experiments. Two different sets of incubations were prepared. In the first set (resulting in methanogenic conditions), 5 mL of soil slurry preincubated without sulfate was incubated at 25 °C with 40 mL of 20 mM potassium phosphate buffer (pH 7.0) in a 150 mL bottle under an atmosphere of N2. The bottles were amended with (i) 5 mL H2O, (ii) 5 mL H2O + 4.5 mL CH3F, (iii) 5 mL 200 mM sodium formate, and (iv) 5 mL 200 mM sodium formate + 4.5 mL CH3F. In the second set (resulting in sulfidogenic conditions), 5 mL of soil slurry preincubated with sulfate was incubated at 25 °C with 40 mL of 20 mM potassium phosphate buffer (pH 7.0) in a 150 mL bottle under an atmosphere of N2. The amendments were the same as above but with the addition of 200 µL of a CaSO4 suspension corresponding to a concentration of 2.5 M (giving a final concentration of 10 mM sulfate).

For lake sediments under methanogenic conditions, 5 mL sediment was incubated in three replicates at 10 °C (which is close to the in situ temperature) with 40 mL of 20 mM potassium phosphate buffer (pH 7.0) in a 150 mL bottle under an atmosphere of N2. The bottles were amended with (i) 5 mL H2O, (ii) 5 mL H2O + 4.5 mL CH3F, (iii) 5 mL 200 mM sodium formate, and (iv) 5 mL 200 mM sodium formate + 4.5 mL CH3F. For sulfidogenic conditions, lake sediments were preincubated with sulfate by adding 0.1 g CaSO4 ⋅ 2H2O (gypsum) to 50 mL of sediment and incubating at 10 °C for 4 weeks. For sulfidogenic conditions, 5 mL of the preincubated sediment was incubated in three replicates at 10 °C with 40 mL of 20 mM potassium phosphate buffer (pH 7.0) in a 150 mL bottle under an atmosphere of N2. The bottles were amended as above but also with 200 µL of a CaSO4 suspension, giving a final concentration of 10 mM sulfate. Samples for later molecular analysis were taken from the original lake sediment and from the lake sediment preincubated with sulfate. The samples were stored at -20 °C (see data in Conrad et al., 2021).

Gas samples for the analysis of partial pressures of CH4 and CO2 were taken from the headspace of the incubation bottles after vigorous manual shaking for about 30 s using a gas-tight pressure-lock syringe which had been flushed with N2 before each sampling. Soil slurries were sampled, centrifuged and filtered through a 0.2 µm cellulose membrane filter and stored frozen at -20 °C for later fatty acid analysis. Chemical and isotopic analyses were performed as described in detail previously (Goevert and Conrad, 2009). Methane was analyzed by gas chromatography (GC) with a flame ionization detector. Carbon dioxide was analyzed after conversion to CH4 with a Ni catalyst. Stable isotope analyses of 13C/12C in gas samples were performed using GC-combustion isotope ratio mass spectrometry (GC-C-IRMS). Formate and acetate were measured using high-performance liquid chromatography (HPLC) linked via a Finnigan LC IsoLink to an IRMS. The isotopic values are reported in the delta notation (δ13C) relative to the Vienna PeeDee Belemnite standard, having a 13C/12C ratio (Rstandard) of 0.01118: δ13C = 103 (Rsample/Rstandard-1). The precision of the GC-C-IRMS was ±0.2 ‰, and that of the HPLC-IRMS was ±0.3 ‰.

Millimolar concentrations of CH4 were calculated from the mixing ratios (1 ppmv = 10-6 bar) measured in the gas phase of the incubation bottles, with 1000 ppmv CH4 corresponding to 0.09 µmol per milliliter of liquid. Note that this is the total amount of CH4 in the gas phase relative to the liquid phase.

Fractionation factors for reaction A→B are defined after Hayes (1993) as 1αA/B=(δA+1000)/(δB+1000), also expressed as ε≡1000(1-α) in permil. The carbon isotope enrichment factor εform associated with formate consumption was calculated from the temporal change in δ13C of formate as described by Mariotti et al. (1981) from the residual reactant 2δr=δri+ε[ln⁡(1-f)], where δri is the isotopic composition of the reactant (formate) at the beginning and δr is the isotopic composition of the residual formate, both at the instant when f is determined. fform is the fractional yield of the products based on the consumption of formate (0<fform<1). Linear regression of δ13C of formate against ln⁡(1-f) yields εform as the slope of best-fit lines. The regressions of δ13C of formate were done for data in the range of fform<0.7. The linear regressions were done individually for each experimental replicate (n=3) and were only accepted if r2>0.7. The ε values resulting from the replicate experiments were then averaged (± SE).

The rice paddy soils were submerged and preincubated to create methanogenic or sulfidogenic conditions. Samples of these soils were suspended in a buffer at pH 7 and amended with formate. In the Vercelli soil, formate was consumed after a lag phase of 4 d under methanogenic and 5 d under sulfidogenic conditions (Fig. 1a). During this time the pH increased from pH 7 up to pH 8 despite buffering. Formate consumption was not inhibited by CH3F (Fig. 1a). Similar results were obtained with IRRI soil (Fig. S1 in the Supplement). Acetate was produced concomitantly with formate consumption, again without effect by CH3F (Fig. 1b). The production of acetate under sulfidogenic conditions was smaller than under methanogenic conditions. Methane was also produced under both methanogenic and sulfidogenic conditions concomitantly with formate consumption (Figs. 1c; S1c). It is noteworthy that CH3F inhibited the production of CH4 (Figs. 1c; S1c). Finally, CO2 was produced under all conditions without lag phase and without effect by CH3F (Fig. 1d). In Vercelli soil, CO2 production was about 2 times larger under sulfidogenic conditions than under methanogenic conditions (Fig. 1d). In IRRI soil, it was only slightly larger (Fig. S1d). The accumulation of acetate plus CH4 was equimolar to the consumption of formate in terms of electron equivalents, while the accumulation of CH4 alone accounted for only <30 % and in the presence of CH3F even less (Figs. 2a; S2a). Hence, acetate was the more important product of formate consumption. Under sulfidogenic conditions, accumulation of acetate plus CH4 was less than equimolar, especially in Vercelli soil (Fig. 2b), probably since formate was instead converted to CO2. However, acetate formation was still substantial, accounting for 60 %–80 % of formate consumption (Figs. 2b; S2b).

The sediments from Lake Fuchskuhle were methanogenic in situ so that preincubation of the samples was not required. However, sulfidogenic conditions were created analogously to the paddy soils by preincubtion with sulfate (gypsum). Substantial formate depletion did not start before about 20 d of incubation in sediments from the NE basin (Fig. 3) and the SW basin (Fig. S3). Again, CH3F only inhibited the production of CH4 but not that of acetate or CO2 (Figs. 3; S3). The main difference to the paddy soils was that CH4 was not produced concomitantly with formate consumption but started right from the beginning. However, the amounts of CH4 produced were only small and were apparently due to the little formate that was consumed in the beginning of incubation (i.e., before day 20), as seen by the fact that CH4 production in the water control (not amended with formate) was negligible (Figs. 3c; S3c). Production of CO2 started without lag phase but accelerated together with formate consumption (Figs. 3d; S3d). In the lake sediments, CH4 accounted only for <10 % of formate consumption, while acetate was the main product when sulfate was absent (Figs. 4a, S4a). In contrast to the paddy soils, formate consumption in both lake sediments was much slower under sulfidogenic conditions than under methanogenic conditions (Figs. 3a; S3a). In the sediment from SW basin, formate consumption was very slow; therefore, less than half of the formate was consumed during 80 d of incubation and consumption was not completed until the end of the experiment (Fig. S3a). Very little acetate was produced and no CH4 was formed from formate in both lake sediments, when sulfate was present (Figs. 4b, S4b).

In the rice paddy soils, δ13C values of formate increased when formate was consumed, indicating discrimination against the heavy carbon isotope. This process was not affected by CH3F and was similar without and with sulfate (Figs. 1e; S1e). The same was the case with the sediment from the NE lake basin but only in the absence of sulfate (Fig. 3e). With sulfate, the δ13C of formate slowly decreased with time (Fig. 3e). In the sediment from the SW basin, δ13C of formate slowly decreased (without sulfate) or stayed constant with time (with sulfate) (Fig. S3e). Note that formate was not completely consumed in the SW sediment when sulfate was present (Fig. S3a).

Mariotti plots of δ13C of formate as a function of fform resulted in negative slopes (Figs. 4; S5). Hence, the enrichment factors (εform) for the paddy soils, both without and with sulfate, and for the sediments from the NE basin of Lake Fuchskuhle without sulfate showed that the light isotope of formate carbon was preferred. Values of εform were in the range of -8.5 ‰ to -2.5 ‰ (Fig. 6). Under sulfidogenic conditions, however, the Mariotti plots of the sediments from the NE basin (Fig. 5) did not show a negative slope and εform could not be determined. The same was the case for the sediments from the SW basin (Fig. 6).

The negative εform indicates that products of formate should be depleted in 13C. Indeed, the δ13C of acetate and CH4 was generally more negative than the δ13C of formate. This was the case in the paddy soils from Vercelli (Fig. 1f) and the IRRI (Fig. S1f), as well as in the sediments from the NE basin (Fig. 3f) and the SW basin (Fig. S3f) of Lake Fuchskuhle. In the sediment of the NE basin, the δ13C of acetate increased from very low level, at -95 ‰, to finally about -57 ‰ in parallel with formate consumption (Fig. 3f). CO2 was also produced during formate degradation to various extent (Reactions R1, R2 and R3). Since the pH was in a range of pH 7 to pH 8, CO2 was also converted to bicarbonate. The δ13C of bicarbonate is generally more positive than the δ13C of CO2 by about 10 ‰ (Stumm and Morgan, 1996). The δ13C of the gaseous CO2 was always close to the δ13C of formate or was more positive. In the paddy soils and the NE basin of Lake Fuchskuhle, the δ13C of CO2 increased in parallel with the increasing δ13C of formate (Figs. 1h, 3h; S1h). The δ13C of the gaseous CO2 produced from the formate-amended samples was initially more negative than that from the unamended samples, but eventually the δ13C increased above these values when formate was completely consumed (Figs. 1h, 3h; S3h).

The δ13C values of the initial formate were about -24 ‰ (Fig. 5). When formate was completely consumed, the δ13C values of the products acetate and CH4 were always more negative. The average δ13C values of the products after complete consumption of formate are shown in Fig. 7. In the absence of sulfate, δ13C of acetate was in a range of -51 ‰ to -49 ‰ and -70 ‰ to -63 ‰, in the paddy soils and lake sediments, respectively (Fig. 7). In the presence of sulfate, δ13C of acetate was in a range of -57 ‰ to -52 ‰ and -78 ‰ to -72 ‰, in the paddy soils and lake sediments (only NE basin), respectively (Fig. 7). The δ13C of CH4 was in a range of -70 ‰ to -54 ‰ and -60 ‰ to -54 ‰, in the absence and presence of sulfate, respectively (Fig. 7). The δ13C of gaseous CO2 (for bicarbonate plus 10 ‰) was in a range of -23 ‰ to -11 ‰ and -24 ‰ to -19 ‰, in the absence and presence of sulfate, respectively (Fig. 7).

In rice paddy soils, formate was consumed within <10 d. The absence of sulfate did not allow sulfidogenic (Reaction R3) degradation but allowed the operation of methanogenic (Reaction R1), homoacetogenic (Reaction R2) or syntrophic (Reaction R4) degradation. Syntrophic degradation is still disputed, since many microorganisms are able to enzymatically equilibrate H2 and formate and thus prohibit exploitation of the difference in the energy content (Montag and Schink, 2018; Schink et al., 2017). Syntrophic formate degradation is exergonic by only a few kilojoules of Gibbs free energy per mole and requires the coupling with methanogenesis or other efficient hydrogen (electron) scavengers. Although formate-driven CH4 production was observed in our study, the production was sensitive to inhibition by CH3F, indicating that CH4 was predominantly produced from acetate rather than from H2. Therefore, syntrophic formate oxidation coupled to CH4 production was probably not a major pathway.

Acetate was the most important product of formate degradation in the paddy soils and the lake sediments. Methane also was a product but was much less important than acetate. Furthermore, it was predominantly produced from acetate, as shown by the inhibition by CH3F and the concomitant decrease in δ13C of CH4, which is characteristic of hydrogenotrophic methanogesis that is not inhibited by CH3F (Conrad et al., 2010). Hence, formate was apparently primarily degraded by homoacetogenesis (Reaction R1). Only part of the produced acetate was immediately used by aceticlastic methanogenesis generating CH4 as secondary product. Although formate is a perfect substrate for homoacetogenic bacteria operating the Wood–Ljungdahl pathway (WLP) (Drake, 1994), the yield of Gibbs free energy per mole formate is less for homoacetogenic than for methanogenic degradation (Dolfing et al., 2008). Thus, it is surprising that formate-driven homoacetogenesis prevailed over methanogenesis. Nevertheless, simultaneous operation of homoacetogenesis and methanogenesis from formate has been observed before in a fen soil (Hunger et al., 2011). Homoacetogenesis prevailing over methanogenesis has also frequently been observed with H2/CO2 as substrate (Conrad et al., 1989; Nozhevnikova et al., 1994), indicating that homoacetogens can take particular advantage from low temperatures (Conrad, 2023) or the availability of secondary substrates (Peters et al., 1998). It is noteworthy that homoacetogens have to invest energy for fixation of formate, while methanogens are able to bypass this step (Lemaire et al., 2020). Perhaps it is such energy investment which makes the homoacetogens to competitive with formate utilizers.

Formate consumption was recorded upon the addition of formate to initial concentrations of about 15 mM, which was much higher than the in situ concentration being typically on the order of a few micromoles per liter (Montag and Schink, 2018). However, the increased concentration allowed stable isotope fractionation, which would not occur under formate limitation. The δ13C of the produced acetate was by about 24 ‰–33 ‰ lower than that of formate. This isotopic discrimination between formate and acetate is similar to that measured in a culture of the homoacetogen Thermoanaerobacter kivui (Freude and Blaser, 2016). However, this discrimination is much larger than the isotopic enrichment factors (εform of -8 ‰ to -2.5 ‰) determined from the change in δ13C during formate consumption. There are two conceivable explanations for this observation. (1) Formate is disproportionated to CO2 and acetate. In the WLP, three formate are oxidized to CO2, one formate is reduced to the methyl group of acetate and one of the produced CO2 is reduced to the carboxyl group of acetate. The disproportionation of formate to acetate and 2CO2 is possibly a branch point (Fry, 2003; Hayes, 2001), at which the carbon flow is split into the production of 13C-enriched CO2 and 13C-depleted acetate, which together result in the εform observed. (2) Formate first is completely converted to CO2 plus H2 (Reaction R5) or other electron equivalents. This reaction displays the εform determined by the Mariotti plots. Acetate is then produced via the WLP by the chemolithotrophic reduction of 2CO2 to acetate, of which the isotopic enrichment factor is typically on the order of about -55 ‰ (Blaser and Conrad, 2016). In any case, it is plausible to assume that acetate was formed via the WLP. In the WLP, oxidation of formate is catalyzed by a formate dehydrogenase, which provides CO2 to the carboxyl branch of the WLP. The methyl branch of the WLP normally starts with formate being converted to formyl-THF. However, it can also start with the reduction of CO2 to formate with a hydrogen-dependent carbon dioxide reductase (HDCD). Homoacetogens (e.g., Acetobacter woodii, T. kivui) contain such a HDCD, which allows the interconversion of formate and H2 plus CO2 (Jain et al., 2020; Schuchmann et al., 2018). The isotope discrimination in our experiments indicates that the CO2 produced from formate has been enriched in 13C rather than depleted, thus supporting the first explanation. The δ13C of CO2 produced from formate was initially lower than that of the unamended soil or sediment being on the order of -20 ‰ to -10 ‰ (Figs. 1h, 3h, S1h, S3h). Eventually, however, δ13C of CO2 reached values of -25 ‰ to -10 ‰ (Fig. 7). The δ13C of bicarbonate is 10 ‰ more positive than that of CO2. This mixed inorganic carbon would be the CO2 substrate for WLP, which together with formate generates the acetate having a δ13C of about -70 ‰ to -50 ‰ (Fig. 7).

Methane was a minor product of formate degradation in all soils and sediments. Since CH4 formation was strongly inhibited by CH3F, it was most likely produced from acetate by aceticlastic methanogens. Since CH4 production from the soils or sediments was much lower without formate amendment, the CH4 must have primarily been produced from the acetate that was generated from formate. The δ13C of CH4 in the soil incubations was more negative than that of acetate (Fig. 7). The difference between the δ13C of CH4 and the δ13C of acetate indicated an isotopic enrichment factor of εac-CH4=-10 ‰ to -8 ‰, which is close to the enrichment factor of aceticlastic Methanosaeta (Methanothrix) concilii (Penning et al., 2006). In the lake sediments, the δ13C of CH4 and acetate was not much different, indicating that acetate was instantaneously consumed by methanogens, as it was produced by homoacetogens so that carbon isotopes were not discriminated. Both paddy soils and lake sediments contained mcrA genes (coding for a subunit of methyl CoM reductase) of Methanosaetaceae (Methanotrichaceae) (Conrad et al., 2021).

In the rice paddy soils, formate was consumed within 10 d when sulfate was present and not quite as fast as without sulfate. In the lake sediments, however, sulfidogenic formate consumption was much slower. Formate degradation by sulfate reduction normally results in complete oxidation to CO2 (Reaction R3). In the lake sediments, CO2 was indeed the main degradation product. However, in the paddy soils, substantial amounts of acetate and even CH4 were also produced. The homoacetogenic bacteria in these soils apparently competed well with the sulfate reducing bacteria, although the soils had been adapted by preincubation in the presence of sulfate. The production of acetate and CH4 was dependent on formate degradation, since no production was observed in the unamended control. Production of CH4 was inhibited by CH3F, indicating that aceticlastic methanogenesis was the main process of CH4 production. The carbon isotope fractionation of formate was similar under non-sulfidogenic conditions, exhibiting a small εform of -8 ‰ to -3.5 ‰ (Fig. 5) and displaying a strong isotope effect with the formation of acetate (δ13C = -57 ‰ to -52 ‰) and CH4 (δ13C = -60 ‰ to -58 ‰). The mechanism of fractionation is probably the same (see above).

In the lake sediments, however, sulfidogenic degradation of formate was much slower than methanogenic/acetogenic degradation. In the sediment of the SW basin, formate was not even completely degraded within 80 d. In the sediments of both lake basins, neither acetate nor CH4 was a major product of sulfidogenic formate degradation. Hence, formate was apparently degraded according to Reaction (R3) forming CO2 as the main carbon product. This formation process displayed no depletion of the heavy carbon isotope, as the Mariotti plots of δ13C of formate did not exhibit a negative slope. The δ13C of the CO2 slowly decreased with an increasing fraction of formate consumed (Figs. 3h; 5c), probably involving isotope exchange between formate and CO2 (DeGraaf and Cappenberg, 1996). The little acetate which was formed displayed a δ13C of -77 ‰ (Fig. 7b), indicating that it was produced by a mechanism similar to the one found in the absence of sulfate, presumably via the WLP.

The strong differences between rice paddy soils and lake sediments were possibly caused by their different microbial communities (Conrad et al., 2021). The differences were seen in the composition of the mcrA and dsrB genes coding for methyl CoM reductase and dissimilatory sulfate reductase, respectively, as well as the gene coding for the bacterial 16S rRNA (data are shown in Conrad et al., 2021). The microbial community structures based on these genes were similar whether the soils and sediments were amended with sulfate or not. However, they were strongly different between soils or sediments were used (Conrad et al., 2021). Unfortunately, these data do not allow for the discrimination of particular taxa of homoacetogenic bacteria. Nevertheless, it is possible that formate-consuming homoacetogens were more prevalent in the soils than in the sediments and competed accordingly with the formate-consuming sulfate reducers, more or less.

Formate was found to be an excellent substrate for acetate formation in the paddy soils and in the lake sediments, confirming and extending similar observations in a fen soil (Hunger et al., 2011). In the anoxic soils, acetate was the major product even in the presence of sulfate, which would have allowed sulfate reduction. The acetate was strongly depleted in 13C relative to formate, but the consumption of formate itself displayed only a small isotopic enrichment factor. Therefore, it is likely that formate was disproportionated to 13C-depleted acetate and 13C-enriched CO2. The δ13C of CO2 was indeed slightly higher than that of formate. Acetate was most likely produced by homoacetogenesis via the WLP. The produced acetate was then used by aceticlastic methanogens (probably by Methanothrix) but only to a minor extent, resulting in further depletion of 13C. The homoacetogenic bacteria in the paddy soils apparently competed well with both methanogenic and sulfate-reducing microorganisms when formate was the substrate. The preference of homoacetogenesis as a degradation pathway is unexpected, since other substrates, such as acetate and propionate, are degraded in these paddy soils by methanogenesis or sulfate reduction (Conrad et al., 2021; Conrad and Claus, 2023). Only in the lake sediments was formate oxidation by sulfate reduction was more prevalent than homoacetogenesis.