Introduction

Methane dynamics in the seabed are linked to mineralization of organic matter, which is mediated by complex microbial communities and coupled to the utilization of oxidants in a sequence that follows a progressive decrease in energy yield [1]. In most continental margin sediments sulfate reduction is the main pathway of anaerobic organic C oxidation [2]. When sulfate and other available electron acceptors are depleted at depth, the mineralization of organic C proceeds to CH4 and CO2 [1, 3]. The flux of methane from the seabed to the overlying water column is controlled by sulfate-dependent anaerobic oxidation of methane (AOM) to CO2 catalyzed by a syntrophic consortium of anaerobic methane oxidizing (ANME) archaea and sulfate reducing bacteria [4, 5]. Most AOM takes place within the sulfate-methane transition (SMT), a sediment interval where overlapping sulfate and methane drive a kinetically sluggish and thermodynamically poor AOM process [6,7,8].

The geochemical zonation in marine sediments is evident from major pore-water constituents, such as sulfate, methane, and dissolved inorganic carbon (DIC), but microbial processes behind this zonation are not as simple as they appear. A recent study identified a global imbalance of methane and sulfate fluxes into the SMT [9], and it is suggested that organoclastic sulfate reduction extends far into the SMT [10, 11]. This additional sulfate consumption may explain the apparently nonstoichiometric oxidation of methane by sulfate. Further, extremely 13C-depleted methane at the base of the SMT has been proposed to be the result of concurrent AOM and methanogenesis [12], or of C isotope equilibration between CH4 and CO2 during AOM under low sulfate supply [13]. It thus appears that other microbial processes in addition to AOM may contribute to the carbon and sulfur transformations within the SMT.

Rates of microbial processes in sediments can be quantified by radiotracer experiments or mathematically from mass-balance and reaction–transport models, e.g., involving DIC and methane that are both released to the porewater during mineralization [14,15,16]. A combination of these approaches can resolve the spatial distribution and rate of individual microbial processes. Such methods have shown that rates of organic matter degradation decrease by a continuous power law function down through the sediment, irrespective of the prevailing redox zonation and terminal degradation pathway [17, 18].

Given the energy limitation that microorganisms experience under long-term conditions in the subseafloor [19, 20], the microbial communities may be balanced between the available energy flux and their minimum energy requirement [21]. Hence, the abundance of different functional groups could be directly related to their total metabolic activity. Analyses of specific gene marker distributions may thereby locate associated microbial processes, and even identify “cryptic” reaction zones that are not directly discernible from geochemical analysis [22, 23]. Microorganisms involved in methane cycling can be studied via their functional genes, e.g., encoding the methyl-coenzyme M reductase (mcr), an enzyme unique to methanogens and anaerobic methane oxidizers. Methanogens have been difficult to detect in marine sediments, however, which has been attributed to their low abundance and to yet undiscovered methanogen groups with mcr genes that are not covered by available primer sets [24]. Even the ANME archaea themselves, which are phylogenetically related to methanogenic archaea but typically assumed to be obligate methane-oxidizers, might also be involved in methane formation [25,26,27].

The primary goal of this study was to understand microbial processes that affect C- and S-transformations within the SMT. Specifically, we aimed to (i) constrain the quantitative relationship between AOM and other microbial processes that have major influence on CH4 and SO42− turnover in the SMT, and (ii) resolve the distribution of CH4 production and consumption, as well as the microorganisms involved in these processes. We performed synoptic analyses of the biogeochemistry, microbial activity, and molecular biology in sediment cores from Aarhus Bay and used new and previously reported data sets from the Baltic Sea to test the general validity of our results (Fig. 1a, Table S1).

Fig. 1
figure 1

a Sampling locations of Baltic Sea sites investigated in this study (Table S1). b Pore-water profiles of SO42− and CH4 concentrations in Aarhus Bay M5 sediment. The SO42− is also shown with log10 scale (inset). The gray shaded area indicates the sulfate-methane transition. Fluxes (in nmol cm−2 d−1) are reported next to the corresponding best-fit slope line used for calculation according to Fick’s first law (±standard error). Open circles indicate methane data points that are likely affected by degassing during sampling. c Total SO42− and CH4 fluxes into the SMT, as well as the fraction of those fluxes accounted for by AOM. Open circles represent data from the POS392 cruise (Fig. S1) [30]

Full size image

Materials and methods

Sediment coring and sampling

Cruises were undertaken with the R/V Aurora, to Aarhus Bay station M5 [28, 29] in 2015 and 2016. Additionally, we utilized sample material and published data from other organic-rich fine-grained sediments in the Baltic Sea, especially from the Bornholm Basin, for our analysis (Table S1, Fig. S1) [18, 30]. Coring and sampling procedures for Aarhus Bay, using a combination of Rumohr corer and gravity corer, were similar to those employed during cruises to Bornholm Basin and described in ref. [18].

Pore-water and sediment biogeochemical analyses

Methods for the quantification of pore water sulfate, methane, DIC, acetate, and sediment organic C (Corg) were performed as described in ref. [18]. Briefly, sulfate was measured by an IC-2500 ion chromatography system (Dionex Corporation); methane was measured using a Gas Chromatograph (GC, SRI Instruments 310C) with a flame ionization detector and the stable carbon isotopic composition of methane was determined using a coupled preconcentration GC/IRMS (isotope ratio mass spectrometer, Thermo Fisher Scientific DeltaV™) [31]; DIC was determined on a Finnigan GasBench II (Thermo Fisher Scientific) and the stable carbon isotopic composition was determined with an IRMS [11]; acetate was analyzed by two-dimensional ion chromatography-mass spectrometry [32, 33]. The concentration and stable carbon isotopic composition of Corg were determined by an elemental analyzer (Thermo Fisher Scientific Flash EA 1112) coupled to an IRMS. The δ13C values are reported vs. the Vienna Pee Dee Belemnite standard.

Stoichiometric ratios of DIC production during AOM and organic matter mineralization

AOM leads to DIC production in the SMT according to a 1:1 reaction stoichiometry: CH4 + SO42− → HCO3− + HS− + H2O. The reaction stoichiometry for organic matter mineralization, depends on the mean carbon oxidation state (ox): Corg + [(4 – ox)/8] SO42− → HCO3− + [(4 – ox)/8] H2S for organoclastic sulfate reduction; and Corg → [(4 + ox)/8] CO2 + [(4 – ox)/8] CH4 for methanogenesis. The ox values for the sediments investigated in this study were calculated from parameter plots of DIC vs. SO42− (dDIC/dSO42− = –[8/(4 – ox)] × [\({D}_{{\rm{SO}}_4^{2-}}\)/DDIC]) (Fig. S2) [34].

Diffusive fluxes

Diffusive fluxes of pore water solutes were calculated from their concentration gradients using Fick’s First Law:

$${J} = -\varphi \times {{D}}_{\mathrm{s}} \times d{C}/d{z}$$
(1)

where J is the diffusive flux, C the concentration, z the depth, Ds the whole-sediment diffusion coefficient, and φ the porosity. Ds was calculated from porosity and molecular diffusion coefficients of individual species, Di, at 8 °C in situ temperature and salinity of 27 for Aarhus Bay sediments (\({D}_{{\rm{CH}}_4}\) = 0.973 cm2 d−1, \({D}_{{\rm{SO}}_4^{2-}}\) = 0.554 cm2 d−1, \({D}_{{\rm{HCO}}_3^{-}}\)  = 0.597 cm2 d−1) or salinity of 15 for the Baltic Sea sediments (\({D}_{{\rm{CH}}_{4}}\) = 1.08 cm2 d−1, \({D}_{{\rm{SO}}_4^{2-}}\) = 0.566 cm2 d−1, \({D}_{{\rm{HCO}}_3^{-}}\)  = 0.609 cm2 d−1), according to Ds = Di/(1 – ln(φ2)) [15]. Note that the isotopic composition of carbon in a solute flux (DIC or CH4) past a given depth in the sediment column is not the same as the isotopic composition of the solute at that depth. The isotopic composition of the flux can, however, be deduced from the slope of mixing-reaction plots where (solute concentration × isotopic composition) is plotted as function of solute concentration (Figs. S3–S5) [10, 35]. All fluxes and their δ13C values used for the following isotope mass balance calculations are summarized in Table S2.

Calculation of DIC sources in the SMT from DIC-fluxes and isotopes

The isotopic composition of DIC produced within the SMT regardless of its source (δ13CDIC,JSMT) was calculated from the balance of isotopes diffusing into, and out of the SMT [10]:

$$\begin{array}{l}{{{\rm{\delta}}^{13}{\rm{C}}_{{\rm{DIC}},{\rm{JSMT}}}}} = \left({{{{\rm{\delta}}^{13}{\rm{C}}_{{\rm{DIC}},{\rm{JTop}}} \times {J}_{{\rm{DIC}},{\rm{Top}}}-{\rm{\delta}}^{13}{\rm{C}}_{{\rm{DIC}},{\rm{JBot}}} \times {J}_{{\rm{DIC}},{\rm{Bot}}}}}} \right)\\ {\mathrm{/}}\left( {{{J}_{{\rm{DIC}},{\rm{Top}}} - {J}_{{\rm{DIC}},{\rm{Bot}}}}} \right)\end{array}$$
(2)

where by JDIC,Top and JIC,Bot represent the DIC flux just above and below the SMT, respectively; δ13CDIC,JTop and δ13CDIC,JBot are δ13C values of the two DIC fluxes. The investigated sediments were carbonate-poor [11, 36, 37], and therefore the influence of authigenic carbonate precipitation on DIC budgets was assumed to be negligible. Once the isotopic composition of the mixed DIC from AOM and organoclastic mineralization within the SMT was calculated, then the relative contribution from the two sources that would give that mixture could be calculated from the known isotopic composition of DIC:

$${F}_{\mathrm{DIC,org}} = \hskip1pt \left({{{\rm{\delta}}^{13}{\rm{C}}_{{\rm{DIC}},{\rm{JAOM}}}}-{{{\rm{\delta }}^{13}{\rm{C}}_{{\rm{DIC}},{\rm{JSMT}}}}}} \right)\\ /\left({{{\rm{\delta}}^{13}{\rm{C}}_{{\rm{DIC}},{\rm{JAOM}}}}-{{{\rm{\delta}} ^{13}{\rm{C}}_{{\rm{DIC}},{\rm{Jorg}}}}}} \right)$$
(3)

where by FDIC,org is the fraction of DIC-production in the SMT that must come from mineralization of organic material. δ13CDIC,JAOM is the δ13C value of DIC produced by AOM assuming ~9‰ fractionation during oxidation of methane (δ13CDIC,AOM = \({\rm{\delta}}^{13}{\rm{C}}_{{\rm{SMT}},{{{\rm{CH}}_4}}}\)  – 9‰) [38, 39]. δ13CDIC,Jorg is the expected δ13C value of DIC produced by organoclastic activity (δ13CDIC,Jorg = δ13CSMT,Corg). Alternatively, FDIC,org can be calculated from the part of the sulfate flux that was not accounted for by methane oxidation:

$${{{J}_{{\rm{DIC}},{\rm{org}}}}} = \left[{8/\left( {4-{\it{ox}}} \right)} \right] \times \left({{{J}_{{\rm{SO}}_4^2 - }-{J}_{{\rm{CH}}_4}}} \right)$$
(4)
$${{{J}_{{\rm{DIC}},{\rm{AOM}}}}} = {{{J}_{{\rm{CH}}_4}}}$$
(5)
$${{{F}_{{\rm{DIC}},{\rm{org}}}}} = {{{J}_{{\rm{DIC}},{\rm{org}}}}}/\left({{{J}_{{\rm{DIC}},{\rm{org}} + }{J}_{{\rm{DIC}},{\rm{AOM}}}}} \right)$$
(6)

To test the data for internal consistency, Eq. (3) can be rearranged to isolate δ13CDIC,JSMT, and calculate which isotopic composition the calculated FDIC,org would lead to:

$$ {{{\rm{\delta}}^{13}{\rm{C}}_{{\rm{DIC}},{\rm{JSMT}}}}} = {{{\rm{\delta}}^{13}{\rm{C}}_{{\rm{DIC}},{\rm{JAOM}}}}} \\ \hskip30pt -\left[ {\left( {{{\rm{\delta}}^{13}{\rm{C}}_{{\rm{DIC}},{\rm{JAOM}}}}-{{{\rm{\delta}}^{13}{\rm{C}}_{{\rm{DIC}},{\rm{Jorg}}}}}} \right) \times {{{F}_{{\rm{DIC}},{\rm{org}}}}}} \right]$$
(7)

Measurements of gross rates with radiotracers

Radiotracer-based experiments to measure: (a) sulfate reduction rates using [35S]-labeled sulfate (SRR), (b) methanogenesis rates using [14C]-labeled DIC (MGRDIC), (c) methanogenesis rates, and (d) acetate oxidation rates using [2-14C]-labeled acetate were performed as described before [18]. C oxidation rates were calculated from SRR and MGR based on the estimated stoichiometry of organic matter mineralization during sulfate reduction and methanogenesis (\({r}_{{{\rm{C}}:{\rm{SO}}_4^{2-}}} = {r}_{{\rm{C}}:{\rm{CH}}_{4}}\) = 8/(4 – ox) = 1.37 and 1.46 for Bornholm Basin and Aarhus Bay M5 sediments, respectively) [34]:

$${\mathrm{COR}} = \left[ {\mathrm{SRR} \times r_{{{\rm{C}}:{\rm{SO}}_4^{2-} }}} \right] + \left[ {\left( {{{{\rm{MGR}}_{{\rm{DIC}}} + {\rm{MGR}}_{{\rm{Ac}}}}}} \right) \times r_{{{\rm{C}}:{\rm{CH}}_4}}} \right]$$
(8)

Modeling of net CH4 turnover from solute profiles

The depth distribution of net methane turnover in the sediment (ΣRCH4) was calculated from the balance of reaction and transport using the following equation [40]:

$$\Sigma R_{{{\rm{CH}}_4}} = \left( {\omega \times dC_{{{{\rm{CH}}_4}}}{\mathrm{/}}dz} \right)-\left( {\varphi \times D_{\mathrm{s}} \times d^2C_{{{\rm{CH}}_4}}{\mathrm{/}}dz^2} \right)$$
(9)

whereby ω is the sedimentation rate (0.14 cm y−1), \({C}_{{\rm{CH}}_{4}}\) is the methane concentration, z is depth, \({D}_{{\rm{CH}}_{4}}\) is the molecular diffusion coefficient of methane (0.973 cm2 d−1), and φ is porosity. The first- and second-order derivatives, corresponding to diffusive transport and production or consumption, respectively, were estimated by fitting a smoothing cubic spline function to the methane concentration data. The degree of smoothing was kept at the minimum needed to filter out noise in the data.

Thermodynamic and kinetic calculations

These are described in detail in SI Materials and Methods.

DNA extraction

Subsamples for DNA extraction were collected with sterile cut-off 5-ml syringes from the same core sections as those for process rate measurements, and immediately stored at –80 °C. Aliquots of frozen sediment (0.2 g wet weight) were used for total DNA extraction, as described in SI Materials and Methods.

Quantitative PCR

The gene abundance of dsrB, mcrA, and archaeal and bacterial 16S rRNA in sediment DNA extract was determined by quantitative polymerase chain reaction (qPCR) using primer mixtures (Table S3) dsrB-F1a-h/dsrB-4RSI1a-f [41], mcrA-uniF/mcrA-uniR, arc915Fmod/Arch958R [42, 43], and bac908F/bac1075R [44], as described in SI Materials and Methods.

Illumina MiSeq sequencing and sequence analysis

DNA was amplified with 25 PCR amplification cycles, using primers (Table S3) specific to the mcrA gene (mlasF/mlasR) [45], and the hypervariable regions (V3–V6) of the archaeal 16S rRNA gene (arc344Fmod/arc915R) [46, 47], according to protocols provided in these references. MiSeq sequencing of the PCR products was performed according to the Illumina® “16S Metagenomic Sequencing Library Preparation” protocol. Quality control, denoising and sequence analysis of paired-end MiSeq reads was performed with MOTHUR version 1.39.5 [48], fungene pipeline (http://fungene.cme.msu.edu) [49], and ARB version 6.0.4 [50], as described in detail in SI Materials and Methods.

Metagenome assembly, differential coverage binning, and metabolic reconstruction of the ANME-1 draft genome

Detailed information about the sampling procedure, DNA extraction, and sequencing of Aarhus Bay station M5 sediment metagenome samples can be found in ref. [51]. Metagenome reads were downloaded from the INSDC Sequence Read Archives (Bioproject ID PRJNA305566), and used for assembly, binning, and metabolic reconstruction of an ANME-1 draft genome, as described in detail in SI Materials and Methods.

Results

Concentration profiles and diffusive fluxes of sulfate and methane

Sulfate concentrations in Aarhus Bay M5 sediment decreased almost linearly from ~26 mM at the sediment-water interface to <100 µM in the SMT (Fig. 1b). Below the SMT sulfate concentrations were about 10 µM (inset in Fig. 1b). The sulfate profiles of other stations in the brackish Baltic Sea (Fig. S1) were similar in shape, but with lower values <16 mM in surface sediments. Methane concentrations were generally low (<0.01 mM) throughout the sulfate zone, and increased steeply down through the SMT and into the methane zone below (Fig. 1b and S1). In deeper parts of the methane zone scatter in the CH4 concentration data often suggests extensive degassing of the supersaturated CH4, which was also visible upon core retrieval.

Diffusive fluxes of sulfate and methane were calculated by applying Fick’s first law to linear gradients right above and below the SMT. In all investigated sediments, the sulfate flux down into the SMT exceeded the upwards methane flux, thereby deviating from the expected 1:1 ratio of AOM. The sulfate-to-methane flux ratio for the SMT in Aarhus Bay station M5 was 2.0, while the flux ratios in Bornholm Basin sediments ranged from 1.4 at BB03 to as high as 9.5 at BB02 (Fig. 1c). Comparison with highly depth-resolved sulfate and methane profiles obtained during the POS392 cruise to other sites in the Baltic Sea confirmed that this flux discrepancy was a general feature in the Baltic Sea. The relation between the sulfate and the methane flux could be described by the linear relationship \({J}_{{\rm{SO}}_4^{2-}}\) = 1.21 \({J}_{{\rm{CH}}_{4}}\) + 0.33 mmol m−2 d−1 (R2 = 0.99), suggesting that (a) AOM is not the sole sulfate sink in the SMT, and (b) lower methane fluxes are, in these sediments, associated with a higher relative contribution of sulfate removal that is not driven by methane oxidation.

δ13C budgets of DIC, CH4, and organic C for the SMT

We used two independent isotope mass balance approaches to identify the source of the DIC produced within the SMT (δ13CDIC,JSMT; Eqs. (2)–(7)). We thereby utilized the extreme differences in δ13C isotopic composition of Corg (–26.4‰ to –21.3‰ in Bornholm Basin and –17.2‰ in Aarhus Bay) and CH4 (–81‰ to –101‰ in Bornholm Basin sediments and –84‰ in Aarhus Bay M5; Figs. S3–S5) as the substrates for organoclastic activity and AOM, respectively. A comparison of the two independent ways to calculate the relative contribution of organoclastic activity to DIC production in the SMT (FDIC,org) is shown in Fig. 2a. The two approaches agree well, and suggest DIC that is significantly enriched in 13C as compared to the δ13C isotopic signature of the upwards diffusing CH4. Therefore, not only AOM, but also organoclastic activity appears to contribute substantially to C cycling within the SMT. The δ13C of produced DIC followed the trend of increasing influence of AOM in sediments with higher methane fluxes (Fig. 2b), consistent with the observed scaling of SO42−–CH4 flux discrepancies (Fig. 1c). However, even at station BB03a, with shallowest SMT (0.23–0.51 mbsf) and highest methane flux (155 nmol cm−2 d−1), δ13CDIC,JSMT values suggested that about 35% of the DIC produced in the SMT originated from Corg oxidation.

Fig. 2
figure 2

a 13C mass balance for the SMT in Aarhus Bay M5 and Bornholm Basin sediments. The diagonal lines show how the relative contributions of DIC produced from organoclastic activity (FDIC,org) and from AOM (1 – FDIC,org) shape the isotopic composition of the combined DIC product (δ13CDIC,JSMT). The isotopic composition of DIC produced within the SMT was either calculated from the produced DIC (circles, Eqs. (2) and (3)), or the weighted average of the consumed substrates CH4 and Corg (squares, Eqs. (6) and (7)). The closer the symbols plot the better the two independent mass balance models agree with each other. Lines and symbols are color coded similarly. Diffusive fluxes and isotopic values that were used for the mass balance calculations are summarized in Table S2. Biogeochemical profiles for the individual stations can be found in Figs. S3–S5. b Comparison of the CH4 flux (\({J}_{{\rm{CH}}_{4}}\)) at individual stations with the fraction of DIC that accounted for oxidation of CH4 or of Corg, as calculated from the isotopic composition of produced DIC (symbols as in a) or the SO42–CH4 flux discrepancy (white and gray areas as in Fig. 1c)

Full size image

Vertical profiles of microbial activity

We resolved the vertical rate distribution of organoclastic SRR and MGR by radiotracer-based experiments, and determined their contribution to C transformations in the SMT. SRR were highest in the uppermost 5 cm (>100 nmol cm−3 d−1 in Aarhus Bay M5 and >10 nmol cm−3 d−1 in Bornholm Basin) (Fig. 3b) [18], and dropped in a log–log linear relationship with sediment depth down toward the SMT. MGRDIC in the sulfate zone were only above detection limit beneath 40 cmbsf and were 2–3 orders of magnitude lower than SRR above the SMT (<10−1 nmol cm−3 d−1). Modeled rates of net CH4 consumption were also <10−1 nmol cm−3 d−1 throughout the sulfate zone (Fig. 3a).

Fig. 3
figure 3

a Depth profiles of sulfate and methane and of net methane consumption in Aarhus Bay M5 sediment modeled from a smoothing fit to the CH4 concentration data (red curve). b Depth profiles of radiotracer-based microbial activity measurements in Aarhus Bay M5 sediment. The black curves indicate the trend of anaerobic C oxidation rates (COR), as calculated from the sum of SRR and MGR (2016), corresponding to the following power law dependence on sediment depth (in cmbsf): COR = 2930 × depth−1.75 nmol C cm−3 d−1 (N = 46). The 95% confidence interval of the estimate of the exponent was –1.57 to –1.92 and the 95% confidence interval of the estimate of the constant was 1610–5540. The gray shaded areas indicate the SMT. The sulfate reduction rate data are also shown in ref. [86]

Full size image

In the SMT, both SRR and MGRDIC increased parallel to rates of net CH4 consumption, with a distinct peak in the lower part of the SMT where sulfate concentrations dropped to low values <0.1 mM (Fig. 3a, b). Here, MGRDIC were generally as high as or higher than corresponding SRR (Fig. S6). It is, therefore, unlikely that AOM-associated enzymatic back-reaction, leading to conversion of 14C-DIC to 14CH4, represented a major component in our measurements.

In the methane zone, MGRDIC continued to decrease along the same log–log linear depth trend that was found for SRR in the sulfate zone, whereas SRR dropped away from this trend. In all sampled depth intervals, MGRAC were extremely low (<10−3 nmol cm−3 d−1, Fig. 3b), and thereby accounted for <1% of the total CH4 production. The [2-14C]-acetate was instead oxidized to CO2 with rates similar in trend, but lower than, SRR in the sulfate zone, or MGR in the methane zone.

Profiles of prokaryotic abundance and methanogen diversity

Absolute and relative abundances of sulfate reducing and methane-cycling microbes in sediments of station M5 were investigated by qPCR of the genes encoding: (a) the β subunit of dissimilatory sulfite reductase (dsrB), (b) the α subunit of methyl coenzyme M reductase (mcrA) covering most methanogens and the closely related ANME clades (Table S3). In addition, we estimated their relative abundance among the total microbial community by qPCR of archaeal and bacterial 16S rRNA genes. For this comparison, we assumed that on average the number of 16S rRNA gene copies per archaeal and bacterial genomes are 1 and 1.4, respectively [20].

All investigated gene markers showed vertical decrease in abundance, with ~10 times higher abundance at the sediment surface than at 1.25 mbsf (Fig. 4a). Consistent with the vertical distribution of methanogenesis rates, the proportion of mcrA genes showed a pronounced peak in the SMT with >5 × 106 gene copies cm−3, accounting for up to ~10% of total archaea, and then dropped back to ~106 gene copies cm−3, accounting for ~1% of total archaea in the methane zone below (Fig. 4b). dsrB gene abundance was highest close to the sediment surface with >5 × 108 gene copies cm−3, accounting for ~20% of the bacterial 16S rRNA genes, and decreased throughout the core to ~1% of total bacteria.

Fig. 4
figure 4

a, b Depth profiles of absolute and relative prokaryotic abundances, c mean cell-specific metabolic rates, and d, e diversity of archaea potentially involved in methane cycling in Aarhus Bay M5 sediment. The vertical line in c depicts the average (±standard deviation shown as shaded areas) of mean cell-specific sulfate reduction rates in the sulfate zone (blue) or mean cell-specific methanogenesis rates in the methane zone (green). The gray shaded area indicates the SMT

Full size image

We compared the abundance of putative methanogenic and sulfate reducing microbes with associated radiotracer-based metabolic activity measurement to calculate potential mean cell-specific metabolic rates (Fig. 4c). Sulfate reducer and methanogen genomes were hereby assumed to each contain a single copy of dsrB or mcrA, respectively, as indicated by available genome sequence data in public databases (https://img.jgi.doe.gov and https://www.ncbi.nlm.nih.gov/genome/microbes/). However, we note that prokaryotes may be polyploid, containing multiple genome copies [52,53,54]. Taking this into consideration, potential mean cell-specific metabolic rates indicated that, on the community level, sulfate reducers operated at fairly constant metabolic rates of ~0.4 fmol dsrB−1 d−1 throughout the sulfate zone with an abrupt decline to low values <0.01 fmol dsrB−1 d−1 under low sulfate concentrations in the methane zone. Here, potential mean cell-specific methanogenesis rates increased to average values of ~0.5 fmol mcrA−1 d−1, and were thereby similar to the potential cell-specific metabolic rates of sulfate reducers in the sulfate zone.

The diversity of putative methanogens and anaerobic methane oxidizers was assessed by high-throughput sequencing of mcrA genes as well as archaeal 16S rRNA genes. Phylogenetic reconstruction (Figs. 4d, e, and 6) suggests that the diverse group of methanogens in the sulfate zone belonged to, e.g. Methanosarcinacea (44–68%) or Methanomicrobiaceae (up to 22%). At the SMT, the community shifted to a complete dominance of ANME-1 archaea. Within the methane zone other putative methanogens remained largely undetected with amplicons for either the mcr or 16S rRNA genes. This suggests that ANME-1 archaea were not only involved in sulfate-dependent methane oxidation but possibly also in methanogenesis.

Aarhus Bay Station M5 ANME-1 draft genome

In order to assess the metabolic potential of ANME-1 archaea, we extended our analysis of methanogenic marker genes to published metagenomic data sets that were generated from Aarhus Bay station M5 sediment [51]. Assembly and differential coverage binning [55, 56] of individual metagenomic libraries enabled the reconstruction of an ANME-1 draft genome (Fig. S7) that was, estimated from the abundance of single copy marker genes [57] to be >86% complete and with low potential for contamination (<3%) (Table S4). The draft genome contained single copies of a 16S rRNA and mcrA gene that confirmed the affiliation with ANME-1 archaea (Fig. S8, Table S4). Nonetheless, it is possible that the binned contigs also include genetic information of other co-occurring and closely related ANME-1 organisms. Highest ANME-1 genome coverage was found for metagenome reads from 175 cmbsf, deep within the methane zone, and, consistent with our initial diversity screening, all mcrA-related genes that were detected in the individual metagenomic libraries showed high similarity to those from ANME-1 archaea (Fig. S8). Metabolic reconstruction of the ANME-1 draft genome revealed features that could potentially enable both methane oxidation and production, as discussed in the following.

Discussion

“Cryptic” methane cycling in the SMT

Our study provides multiple lines of evidence that upwards diffusing CH4 is not the only electron donor driving SO42− reduction in the SMT. Instead, a significant proportion of SO42− removal in the SMT is driven by the oxidation of buried organic matter, and this can occur either (i) directly via organoclastic sulfate reduction, or (ii) indirectly via methanogenesis, which provides an SMT-internal CH4 source, thereby fueling additional AOM and generating a “cryptic methane cycle” (Fig. 5a).

Fig. 5
figure 5

a Proposed pathways of sulfate and methane turnover in the SMT. Co-occurrence of AOM and organoclastic activity results in an imbalance between SO42− and CH4 fluxes according the reaction stoichiometry of organoclastic sulfate reduction and methanogenesis (cf. Materials and Methods). Because the produced CH4 is consumed again by AOM, SMT-internal methanogenesis results in the same additional sulfate consumption as would organoclastic sulfate reduction. Gray arrows indicate the transfer of reducing equivalents, symbolized by electrons [e−]. b CH4 turnover in Aarhus Bay M5 sediment without and with SMT-internal methanogenesis. The gray shaded area indicates the SMT

Full size image

Availability of sulfate in the SMT is limited by diffusive transport, and our results suggest that the competition between AOM and organoclastic sulfate reduction or methanogenesis is controlled by a balance between OM reactivity within the SMT and CH4 flux into the SMT. The relatively higher contribution of AOM to overall SO42− reduction in the SMT at sites with high methane fluxes is consistent with the predicted kinetic control of sulfate-dependent AOM in sediments, related to the high Km for methane in the mM range [6]. High AOM rates appear to be restricted to a narrow zone in the lower part of the SMT where methane concentrations are high and SO42− has not been depleted yet [7]. Indeed, the kinetic drive of AOM in Aarhus Bay M5 sediment increased towards a sharp peak at the base of the SMT (Fig. S9). Here, also the abundance of ANME-1, CH4 consumption, and MGRDIC peaked, indicating a tight coupling between methane oxidation and production in the same depth interval.

Taking into consideration that experimental rate measurements tend to scatter within an order of magnitude, we estimate that an important fraction of carbon cycling in the SMT is fueled by SMT-internal methanogenesis, accounting for 20% (station BB02) to 60% (station M5) of the total organoclastic activity (SRR + MGR) in this interval. This has a pronounced influence on the methane dynamics, as for example at station M5 the estimated methane turnover times in the SMT with and without this additional internal methane production were on average about 280 days and 150 days, respectively (Fig. 5b). Our model of enhanced methane cycling by concurrent AOM and methanogenesis is consistent with observations of a distinct peak in experimentally measured methane oxidation rates at the base of the SMT and corroborates extremely 13C-depleted methane isotope signatures at the sulfate–methane transition [7, 12, 13]. The 13C-depleted DIC produced by AOM is hereby recycled back to methane with preferential use of 12C during CO2 reduction.

The directly measured methanogenesis in the SMT of M5 (56 nmol CH4 cm−2 d−1) was almost as high as the methanogenesis in the entire methane zone, as estimated from the methane flux (61 nmol CH4 cm−2 d−1), and from extrapolation of the log–log linear depth trend of organoclastic activity (95 nmol CH4 cm−2 d−1). Thus, a large fraction of the total methanogenesis in the sediment appears to take place through a “cryptic” methane cycle. This methane cycle has remained undetected by modeling of the net methane production from methane concentration data.

Metabolic potential of Aarhus Bay station M5 ANME-1 archaea

The evidence for concurrent methane production and oxidation in the same depth interval, presented above, together with amplicon and metagenomic data, suggest that one closely related group of organisms affiliated with ANME-1 are capable of both methane oxidation and methanogenesis. Previous studies that indicated the capability of ANME archaea to perform methanogenesis have been contemplated with doubt, however, because it remained unclear whether methane production was instead attributed to other co-occurring methanogen populations [25, 58,59,60]. If we assume that other methanogens did not evade detection by our employed methods and represented <1% of detectable methanogens, their mean cell-specific metabolic rates would be 1–3 orders of magnitude higher than those of methanogens in, e.g., pure culture, enrichments, or different wetland soils (~0.05–5 fmol cell−1 d−1) [61,62,63,64], which would be highly unlikely. Although our results suggest that nearly all potential methanogens within and beneath the SMT of Aarhus Bay M5 sediments belonged to ANME-1 archaea, we cannot exclude that undiscovered methanogen lineages were also active. However, we consider this unlikely because (1) consistent with stable mean cell-specific metabolic rates of sulfate reducers in the sulfate zone, the abundance of ANME-1 in the SMT and in the methane zone closely reflected measured MGR, presumably in the absence of AOM, and (2) even the deeply sequenced metagenomes did not indicate presence of mcr genes belonging to putative methanogens other than ANME-1.

How could it be possible that ANME-1 are responsible for both methanogenesis and methane oxidation in the same sediment? Although ANME are commonly referred to as conducting a metabolism of “reverse methanogenesis” [65], the two physiologies differ in their terminal electron acceptors and protein complexes, which conserve energy as electrons move towards electron acceptors [26, 27]. Since ANME and methanogens are polyphyletic, different AOM and methanogenesis phenotypes may be distinguishable by phylogenetic clustering, but it is also possible that both physiologies could be found within a cluster. Alternatively, the routing of electrons through different protein complexes could enable a switch between methanogenesis and methanotrophy in a single organism. The metabolic mode of ANME-1 cells could thereby be controlled by electron donor or acceptor availability, and also by syntrophic associations. Depending on whether the partner organisms are able to (a) accept reducing equivalents from the ANME-1 and use them to reduce sulfate, or (b) supply reducing equivalents that can be used by the ANME-1 to reduce CO2, the net reaction would be either methane production or methane consumption. Such a metabolic versatility could provide a competitive advantage over other methanogens and, for example, allow ANME-1 archaea to adjust to seasonal variations in the depth of the SMT and the associated sulfate supply [66].

In the ANME-1 genome analyzed here, all core enzymatic carbon transfer and redox reactions required by a methanogenesis pathway were represented, even a N5,N10-methylene-tetrahydromethanopterin reductase (Mer) homolog that has not previously been detected in ANME-1 [65, 67, 68]. During anaerobic methanotrophy (Fig. 6a), generation of a proton gradient and ATP synthesis could be driven by complete reversal of the core methanogenic steps, coupling oxidation of CH4 to CO2 with transfer of the electrons to a partner organism [69]. As in previous ANME-1 genomes, HdrE was not detected [65, 67, 68], though here HdrD was. Electron transfer from the ANME-1 to a syntrophic partner might function via direct cell-to-cell connections, as indicated by the presence of genes encoding multihaem cytochrome complexes [69]. Although speculative, archaeal flagella/pilus-related proteins (Fla) might also play a role in direct electron transfer [70, 71]. Direct electron transfer from a different syntrophic partner to the ANME-1 could, in contrast, enable reduction of CO2 to CH4 (Fig. 6b). Cytoplasmic proton consumption during CO2 reduction would result in a chemiosmotic potential that could be used for ATP generation. Alternatively, energy conservation may occur at the Mtr step. Although CO2 reduction via direct electron transfer has been identified for other methanogens, the intracellular route of electrons and associated energetic constraints remain largely unresolved [72,73,74].

Fig. 6
figure 6

Proposed metabolic versatility of Aarhus Bay M5 ANME-1. a During AOM (reversal of the methanogenesis pathway, red box), ANME-1 cells supply reducing equivalents (symbolized by electrons [e−]) to sulfate-reducing partner organisms via direct electron transfer (blue box). b During methanogenesis, partner organisms supply reducing equivalents from organic C oxidation to ANME-1 cells via hydrogen, different C substrates, and/or direct electron transfer (purple boxes). Abbreviations for enzymes and co-factors: H4-MPT, tetrahydromethanopterin; CoB-SH, coenzyme B; CoM-SH, coenzyme M; MP, methanophenazine Fd, ferredoxin; F420, coenzyme F420; Fmd, formyl-methanofuran dehydrogenase; Ftr, formylmethanofuran--tetrahydromethanopterin N-formyltransferase; Mch, methenyl-H4-MPT cyclohydrolase; Mtd, F420-dependent methylene H4-MPT dehydrogenase; Mer, F420-dependent methylene-H4-MPT reductase; Mtr, Na+-translocating methyl-H4-MPT:coenzyme M methyltransferase; Mrp, multicomponent Na+:H+ antiporter; Hdr, heterodisulfide reductase; FrhB, F420-reducing hydrogenase subunit B; Nuo, NADH/F420-H2 dehydrogenase-like membrane complex; Fla, flagella; Cyt × 4–10, c-type cytochromes with up to 10 haem binding motifs (CxxCH [69]); Mbh, membrane bound [NiFe] hydrogenase; Cdh, CO dehydrogenase; Acs, acetyl-CoA synthase; Fdo, formate dehydrogenase; Aor, aldehyde:ferredoxin oxidoreductases; GlcD, cytoplasmic FAD-containing dehydrogenase; Atp, V/A-type H+/Na+-transporting ATPase. Other genes detected in the ANME-1 draft genome are listed in Table S5 and the Supplementary Data Set S1

Full size image

We identified several other potential electron entry points that could allow a methanogenic pathway in these ANME-1 cells. Consistent with previous studies of ANME, hydrogenases associated with hydrogenotrophic methanogenesis were largely absent [65, 67, 68]. The ANME-1 draft genome encoded only one putative hydrogenase with homology to membrane-associated [NiFe]-Group 4d hydrogenases [75, 76], suggestive of a possible route to supply reduced ferredoxin necessary for CO2 reduction via H2 oxidation at the expense of a chemiosmotic gradient via proton import. The reduction of F420, an essential electron donor in methanogenesis from CO2, might proceed through FrhB, the F420-reducing hydrogenase subunit B. The FrhB would thereby facilitate electron exchange between ferredoxin and F420, either alone or in combination with other enzyme complexes [77,78,79]. We found multiple gene copies of FrhB with a F420 binding motif [79] in the ANME-1 draft genome that were co-located to genes encoding a cytoplasmic heterodisulfide reductase (HdrABC), and a putative membrane-bound Nuo-like oxidoreductase complex that shows structural similarity to F420-H2 dehydrogenase in Methanosarcinales [80].

We suspect that also other electron donors than H2 could be used for CH4 production, such as formate, carbon monoxide or acetate, as indicated by the presence of genes encoding a formate dehydrogenase (Fdo), and a CO dehydrogenase/acetyl-CoA synthase complex (Cdh/Acs), respectively. The Cdh/Acs may also function anabolically [81, 82]. Interestingly, we found genes that indicate utilization of more complex C substrates. For example, different aldehyde:ferredoxin oxidoreductases (Aor or For) could catalyze the conversion of an aldehyde to a carboxylic acid and thereby supply reduced ferredoxin [83]. A gene encoding a cytoplasmic FAD-containing dehydrogenase (GlcD) co-located to genes of HdrD might further allow the oxidation of α-hydroxy acids to be coupled to reduction of CoMS–SCoB [84, 85]. Altogether, a combination of electron donors could potentially be used during methanogenesis, including hydrogen, reduced carbon compounds, and electron transfer from syntrophic partners. The chemiosmotic potential that could be used for ATP generation and coupling methanogenesis to energy conservation, may therefore result both from the intracellular consumption of protons during direct electron transfer, the translocation of sodium ions at the Mtr step, and potentially also with proton translocation via the Nuo-like oxidoreductase complex.

Despite the presence in the ANME-1 draft genome of genes that could enable disproportionation of the central intermediate acetate (to CH4 and CO2) we detected only very low rates of acetoclastic methanogenesis. The supplied [2-14C]-acetate was instead converted to [14C]-CO2. There were no genomic indications that ANME-1 could completely oxidize acetate to CO2, e.g., via the acetyl-CoA pathway. We, therefore, suspect that acetate oxidation was done by other organisms and that this was kinetically or energetically more favorable. Without electron acceptors other than CO2, acetate oxidation leading to methane formation would require syntrophic coupling to ANME-1 utilizing the electrons. Our thermodynamic calculations suggest that throughout the SMT both methanogenesis from acetate and methane oxidation with sulfate were exergonic but close to thermodynamic equilibrium with energy yields of −10 kJ molacetate−1 and –15 to –20 kJ molCH4–1, respectively (Fig. S9).