Hydrogen–independent CO2 reduction dominates methanogenesis in five temperate lakes that differ in trophic states

Abstract Emissions of microbially produced methane (CH4) from lake sediments are a major source of this potent greenhouse gas to the atmosphere. The rates of CH4 production and emission are believed to be influenced by electron acceptor distributions and organic carbon contents, which in turn are affected by anthropogenic inputs of nutrients leading to eutrophication. Here, we investigate how eutrophication influences the abundance and community structure of CH4 producing Archaea and methanogenesis pathways across time–resolved sedimentary records of five Swiss lakes with well–characterized trophic histories. Despite higher CH4 concentrations which suggest higher methanogenic activity in sediments of eutrophic lakes, abundances of methanogens were highest in oligotrophic lake sediments. Moreover, while the methanogenic community composition differed significantly at the lowest taxonomic levels (OTU), depending on whether sediment layers had been deposited under oligotrophic or eutrophic conditions, it showed no clear trend in relation to in situ distributions of electron acceptors. Remarkably, even though methanogenesis from CO2-reduction was the dominant pathway in all sediments based on carbon isotope fractionation values, taxonomic identities, and genomes of resident methanogens, CO2-reduction with hydrogen (H2) was thermodynamically unfavorable based on measured reactant and product concentrations. Instead, strong correlations between genomic abundances of CO2-reducing methanogens and anaerobic bacteria with potential for extracellular electron transfer suggest that methanogenic CO2-reduction in lake sediments is largely powered by direct electron transfer from syntrophic bacteria without involvement of H2 as an electron shuttle.


Introduction
Lake sediments are globally important organic carbon (OC) sinks that, despite the much smaller surface area of lakes, have annual OC burial rates in the range of marine sediments [1,2].In addition, lake sediments are important sources of greenhouse gasses, such as carbon dioxide (CO 2 ) and methane (CH 4 ) [3,4].CH 4 release from lakes is on the rise due to human activities that increase OC loads (eutrophication), raise water temperatures, and promote bottomwater oxygen (O 2 ) depletion [5][6][7].It has been estimated that freshwater sediments are the biggest natural source of CH 4 to the atmosphere (∼122 Tg CH 4 yr −1 ) after freshwater wetlands (∼185 Tg CH 4 yr −1 ) and exceed total emissions from oceanic sources (∼14 Tg CH 4 yr −1 ) [8].
Most of the CH 4 in lake sediments is generated by anaerobic Archaea, also known as "methanogens".Methanogenic Archaea gain energy by converting end products of the fermentative degradation of organic matter, e.g.H 2 + CO 2 , acetate, and methanol, to CH 4 [9].The distribution of methanogenic activity is in part controlled by competition with other anaerobic microorganisms that respire nitrate, metal oxides (Mn(IV), Fe(III)), or sulfate.Because these electron acceptors typically provide higher energy gains from the same substrates [10], methanogens often only dominate respiration in deeper layers ("methanogenesis zones"), where competing electron acceptors have been depleted.
Multiple archaeal taxa have been linked to CH 4 production.Among these, the euryarchaeotal orders Methanomicrobiales, Methanosarcinales, Methanocellales, and Methanobacteriales are often considered to be the main CH 4 producers in lake sediments [11].Yet, other, more recently discovered groups such as the euryarchaeotal order Methanomassiliicoccales [12,13] and class Methanonatronarchaeia [14], as well as putative methanogens identified by metagenomic analyses [15][16][17] could also be important.In addition, several environmental clusters of mcrA, a well-studied marker gene of CH 4 -cycling Archaea that encodes the alpha subunit of methyl-coenzyme M reductase, remain phylogenetically unassigned [18].
Archaeal methanogenesis is known to proceed via four pathways: CO 2 reduction with H 2 or formate as electron donors ("hydrogenotrophic"), acetate disproportionation ("aceticlastic"), dismutation of methylated compounds, e.g.methanol, methyl amines, or methyl sulfides, with or without H 2 ("methylotrophic"; for review, see [9]), and O-demethylation of methoxylated compounds (methoxydotrophic; [19]).Among these, acetate disproportionation and CO 2 reduction are considered the dominant pathways on Earth with theoretical calculations predicting a 2:1 ratio of the former to the latter [11,20].Yet, carbon stable isotopic data paint a more complex picture, e.g.dominance of CO 2 reduction in marine sediments versus dominance of aceticlastic methanogenesis in freshwater sediments [21].However, even in freshwater lakes, isotopic data and methanogenic community composition indicate CO 2 reduction to dominate in some cases [20,22,23].
One potential driver of methanogenic pathways in nature is the chemical composition of organic matter from which methanogenic substrates are produced.Fermentation reactions, which are the main sources of methanogenic substrates, vary in the compositions of their end products based on the chemical structure of the compounds that are being fermented.For instance, methanol is known to be released during the degradation of aromatic acids and pectin from vascular plants [24,25], whereas acetate may mainly form from fresh or labile algal organic matter [26].Based on these past studies, one might speculate that the process of eutrophication, which promotes phytoplankton growth and sedimentation of labile algal organic matter, should favor aceticlastic methanogenesis, while high inputs of terrestrial OC should lead to elevated contributions of methylotrophic methanogenesis.Yet, the existing data do not necessarily support this view and suggest additional important roles of temperature and interactions with syntrophic bacteria [11].Adding to the complexity, laboratory studies show that certain aceticlastic Methanosarcinales (Methanosarcina and Methanothrix) can reduce CO 2 to CH 4 using electrons from electrogenic partner organisms [27].The potential for direct electron transfer between syntrophic bacteria and methanogens in lake sediments was subsequently shown by adding conductive particles to anoxic sediment slurries [28,29], but the environmental significance of this process remains unknown.
Understanding the drivers behind methanogenic pathways is key to understanding the lacustrine carbon cycle and how it will respond to future anthropogenic perturbations.Here, we investigate the long-term impact of eutrophication on dominant methanogenic pathways and on the abundance and community structure of methanogenic Archaea across timeresolved sedimentary records of five lakes in central Switzerland.Carbon isotopic values of CH 4 and dissolved inorganic carbon (DIC) and taxonomic compositions of methanogens indicate CO 2 reduction as the main methanogenic pathway.Yet, Gibbs energies of methanogenic CO 2 reduction with H 2 , based on measured substrate and product concentrations in natural sediments, suggest this pathway to be largely endergonic and unlikely to be important.Complementary metagenomic analyses, which indicate significant correlations between abundances of CO 2reducing methanogens and syntrophic bacteria with potential for extracellular electron transfer (EET), lead us to propose that electrotrophic CO 2 reduction is the dominant methanogenic pathway in the lakes studied.

Lake trophic histories
All five lakes are well-characterized with respect to their trophic histories.(Table 1; for details see [6]).After the era of peak eutrophication (∼1950 to 1970), P bans on detergents and improved wastewater management lowered P inputs significantly [6].As a result, Lake Zurich returned to a mesotrophic state in ∼1980, though the deep basin of this lake remains hypoxic today.By contrast, Lake Greifen, Lake Baldegg, and Lake Zug remain eutrophic due to retention of P that was introduced in the 20th century.Despite artificial water column mixing and aeration (Lake Baldegg: since 1982/83; Lake Greifen; since 2009), Lake Baldegg experiences strong water column oxygen decreases and Lake Greifen becomes hypoxic or anoxic below 10 m water depth every summer.By contrast, Lake Lucerne, despite slight increases in P concentrations in the 1960s and 1970s, always remained oligotrophic.

Sampling and in situ biogeochemical zonation
The long-term inf luence of eutrophication on the sedimentary carbon cycle in the five lakes has been studied as part of the research effort "Lake Eutrophication Impacts on Carbon Accumulations in Sediments".During a field campaign in June/July 2016, three stations in each lake that differed in water depths ("shallow", "medium", "deep") were sampled using UWITEC gravity corers with 15-cm liner diameters (Table 1; also see [6]).Detailed analyses of biogeochemical gradients, macrofaunal communities, and microbial community structure have been performed and complemented by high-resolution sediment age models, with which the trophic state at the time of deposition could be determined for sediment layers throughout these cores [6,30,31].
In all lakes, bottom water temperatures at the time of sampling ranged from 5 to 9 • C and the pH of sediment pore water was neutral to slightly alkaline (7.0 to 8.4).O 2 was depleted in the top 1 cm, with the shallowest penetrations (average ± standard deviation (SD)) in the hypoxic deep basin of Lake Zurich (no measurable O 2 ), Lake Baldegg (0.08 ± 0.02 cm), and Lake Greifen (0.17 ± 0.03 cm), and the deepest penetration in Lake Lucerne (0.73 ± 0.25 cm).Notably, the dominant anaerobic respiration reactions overlapped in all lakes, with methanogenesis setting in at sediment depths where the reduction of nitrate, sulfate, iron(III) and manganese(IV) were still occurring [6,32].As a result, the methanogenesis zone started at the surface in Lakes Baldegg, Greifen and Zug, at 2.6 ± 2.4 cm in Lake Zurich, and 3.0 ± 1.7 cm in Lake Lucerne [6].
Dihydrogen (H 2 ).Concentrations of dissolved H 2 were measured using the incubation method [35].Sediment cores were extruded where χ H 2 is the mol fraction of H 2 (in parts per billion), P the headspace gas pressure (1 atm), R the universal gas constant (0.008314 kJ mol −1 K −1 ), and T the incubation temperature (in K).Dissolved H 2 concentrations were calculated by multiplying [H 2 ] g by the Bunsen coefficient, β, for freshwater at +4 • C ( [36]; Volatile fatty acids: Pore water concentrations of volatile fatty acids (VFAs) were measured on a two-dimensional ion chromatography system as previously published [37] with minor modifications.The instrument used was a Dionex ICS6000+ system (Thermo Scientific).Separation was achieved with a 2-mm AS24 column for the first dimension and a 2-mm AS11HC column for the second dimension.The retention time window for collecting VFAs from the first chromatographic dimension was set to 3-6.8 min.Quantification was achieved by conductivity detection at detection limits of 0.1-0.4μM (for details see [37]).Prior to analysis, 1 mL of sample pore water was filtered through Milli-Q washed IC grade syringe filters (Acrodisc, Supor ® membrane, 13 mm diameter).The first 0.5 mL of the filtered pore water was discarded, while the second 0.5 mL was used for the analysis.

Thermodynamic calculations
In situ Gibbs energy yields (ΔG r ') of methanogenesis reactions from H 2 + CO 2 , acetate, methanol, and methanol+H 2 were calculated based on the equation: where ΔG r 0 (TP) is the Gibbs energy (kJ mol −1 of reaction) at standard concentrations (1 M per each reactant and product, pH 7.0) at in situ temperature T (K) and pressure P (bar), R is the universal gas constant (0.008314 kJ mol −1 K −1 ), and Q r the quotient of (except where noted otherwise) measured product and reactant activities.The standard Gibbs energy of the reaction ΔG r 0 was corrected to in situ temperature and pressure as outlined in [38].Standard Gibbs energies, enthalpies, and molal volumes of formation are shown in Table S1.Activities of aqueous species were calculated using the activity coefficients γ HCO3-= 0.832 (calculated using PHREEQC v3) and γ CH4 = 1.24 [39].The activity coefficient of H 2 was assumed to equal that of CH 4 , whereas activities of acetate and methanol were assumed to equal 1.0.All concentrations were measured as described above except those of methanol.For methanol we conservatively assumed a concentration of 1 nM (typical methanol concentrations in sediments are higher (0.1 to 1 μM); [40]).

Analysis of methanogenic community structure
DNA extraction.DNA was extracted following the modular extraction method of [41].For the exact protocol used, see [31].
Quantitative PCR.The primer combination mlas F (5 -GGT GTM GGD TTC ACM CAR TA)-mcrA-rev (5 -CGT TCA TBG CGT AGT TVG GRT AGT) was used for mcrA quantification by quantitative polymerase chain reaction (qPCR).Each reaction was run in triplicate and consisted of a final volume of 10 μL (5 μL SYBR Green I Master Mix (Roche Molecular Systems), 0.5 μL of each 10 μM primer solution, 1 μL 10 mg mL −1 bovine serum albumin, 2 μL DNA extract, 1 μL molecular grade water).qPCR assays were performed on a LightCycler 480 II (Roche Molecular Systems) and consisted of (i) enzyme activation: 95  C in 1 min intervals to check for primer specificity.qPCR standards consisted of a tenfold dilution series of complete mcrA genes of Methanocorpusculum parvum (∼6.59 × 10 8 copies μL).Contributions of CH 4 -cycling Archaea to total microbial communities were calculated based on ratios of mcrA copies to total 16S ribosomal ribonucleic acid (rRNA) gene copies from [31].
Amplicon libraries.Amplicon libraries were prepared as described in [31] and involved the same mcrA primer combination and amplication protocol as for qPCR.Sequencing was done on an Illumina MiSeq (Illumina Inc.).Raw sequence reads were quality-checked by FastQC (www.bioinformatics.babraham.ac.uk/ projects/fastqc), read ends trimmed using seqtk (github.com/lh3/seqtk), paired end reads merged into amplicons by FLASH [42], primer sites trimmed by usearch [43], and quality filtering was done by prinseq [44].Zero-radius Operational Taxonomic Units were generated using UNOISE3 [45] and clustered into OTUs using a 97% similarity threshold.These OTUs were taxonomically identified based on neighbor-joining phylogenetic trees in ARB using a publicly accessible mcrA database (mcrA4All; [46]).This database contained >2400 high-quality mcrA sequences from a wide range of published amplicon and whole-genome sequencing studies with manually optimized sequence alignments.Diversity calculations, sample ordination and statistical tests were performed in R using the vegan (v.2.6-4) package [47].Visualizations were done in R with the ggplot2 package [48] and esthetically edited in Inkscape vector graphics software.

Metagenome assembly and functional annotation
DNA extracted from five sediment layers of the deep station of each lake (25 in total) was shipped to the Joint Genome Institute (JGI), Walnut Creek, CA, USA, where it was sequenced and assembled according to the standard JGI pipeline (see Supplementary material for details).Criteria used to search for the reported genes are in SQL query log (File S1).Metagenomes from two additional sediment layers per lake were sequenced with the same technology at the Functional Genomics Center Zürich and used for read mapping to the annotated assemblies.For more detailed information, see the Supplementary material.

Metagenome-assembled genomes generation and classification
The 25 metagenome assemblies were subjected to differentialcoverage and tetranucleotide-frequency-based binning and summarized into one non-redundant set of bins.Bins were analyzed for size, completeness, redundancy, and contig length, and manually refined to reduce "contamination", i.e. duplications of single-copy genes (for details, see Supplementary material).After refinement, bins with a completeness of >50% and contamination <5% were considered for further analysis (434 in total).MAGs were taxonomically classified by GTDB-Tk [49].For MAGs of Methanomicrobia, the classification was confirmed by calculating a de novo phylogenetic tree based on amino acid positions conserved in at least 25% of the concatenated sequence alignments of 53 marker proteins [50] made by GTDB-Tk [49].The main metrics of the MAGs can be found in Table S2.

Identification of potential syntrophic partners of methanogens
Spearman rank correlations between coverage patterns of MAGs were calculated using cor.testfunction in R as implemented in [51].Only MAGs with a coverage of >0.5x that were present in >20% of the samples were included.Among methanogens, only six MAGs (all Methanoregulaceae) fulfilled this criterion.MAGs showing a significant (P < .05;ρ > 0.75) to at least one methanogenic MAG were considered potential partner organisms (Table S3).
The true abundances of microbial taxa were estimated by multiplying relative abundances of the MAGs by qPCRdetermined 16S gene copy numbers.Spearman rank correlations between estimated abundances of methanogens and other members of the microbial community were calculated as in [51].MAGs showing a positive correlation coefficient of >0.75 to at least one methanogenic MAG were considered potential partner organisms.

Potential for extracellular electron transfer in MAGs of methanogens and syntrophic partners
Encoded proteins containing multiple heme-binding sites (CxxCH patterns) were analyzed with InterPro Scan to confirm multiheme c-type cytochrome annotation (InterPro superfamily IPR036280).Phylogenetic analyses were conducted to test if the identified multi-heme cytochromes were closely related to proteins with known functions such as ammonia-forming nitrite reductase NrfA.The following selection of amino acid sequences was used for the tree calculation: (i) multi-heme cytochromes identified in the MAGs, (ii) 10 closest sequences from the Uniprot database, and (iii) members of the multi-heme cytochrome superfamily (IPR036280) with reviewed functional annotation.The sequences were aligned with MAFFT (e-ins-i method).A phylogenetic tree was calculated based on alignment sites conserved in ≥25% of the aligned sequences using FastTree [52] with the Le-Gascuel [53] substitution model.
The conductivity of pili was estimated based on PilA protein sequences.Pili were considered potentially conductive if gaps between aromatic amino acids were shorter than 40 amino acids and percentages of aromatic amino acids exceeded 9%, as suggested by [54,55].Flavin-dependent extracellular electron transfer (FLEET) proteins of Listeria monocytogenes, as described by [56], were identified in Uniprot KB to obtain information on the COG orthologous groups and Pfam families these proteins are assigned to.Genes encoding proteins belonging to the same COG orthologous groups and exhibiting the same Pfam functional domains were searched in the MAGs to identify potential homologs of Listeria FLEET genes.

Dissolved inorganic carbon and methane concentrations and methanogen abundances
DIC and CH 4 concentration trends ref lect the stimulation of organic matter mineralization-in particular methanogenesisrates in sediments of the eutrophic lakes, which receive higher organic matter inputs [6].DIC concentration profiles follow typical concave-down trends, with steepest increases in surface sediments of eutrophic lakes, where organic matter mineralization rates are the highest (Fig. 1, 1st row).Thus, DIC concentration gradients increase from oligotrophic Lake Lucerne to the highly eutrophic Lake Baldegg and Lake Greifen.CH 4 concentrations show similar trends in relation to trophic state (Fig. 1, 2nd row), and moreover indicate significant CH 4 production all the way to the sediment surface in eutrophic lakes (also see [32]).Within most lakes, CH 4 concentrations reached their highest values at the deepest station.This trend was not seen for DIC.
McrA copy numbers, used as a proxy for CH 4 -cycling archaeal abundance, were in the range of 10 4 -10 7 copies g −1 wet sediment (Fig. 1, 3rd row).Copy numbers were generally higher in Lake Zurich and Lake Lucerne than the three eutrophic lakes, a trend that was also observed for total archaeal and bacterial 16S rRNA gene copies [31] and total cell counts [6].This trend was opposite to the observed methanogenic activity increase with trophic state.Station-specific gene abundance profiles varied from near-constant abundances with depth (Lake Baldegg, deep and medium station in Lake Greifen) to increasing in the top ∼10 cm and constant below (Lake Zurich), to decreasing with depth (shallow station in Lake Greifen, shallow and medium stations in Lake Zug).Some stations also had an increase in the top ∼10 cm and decrease below (Lake Lucerne, deep station in Lake Zug).
We also calculated the contributions of CH 4 -cycling Archaea to total microbial populations based on ratios of mcrA to total 16S rRNA gene copy numbers from Han

Concentrations of methanogenic energy substrates
Concentrations of the methanogenic substrates H 2 and acetate (Fig. 1B) differ by three orders of magnitude (H 2 : mainly <2 nM; acetate: mostly 0.5 to 20 μM), but otherwise show several similarities.Both substrates are low in bottom water but have concentration peaks in the uppermost sediment layer (0-1 cm) of eutrophic lakes, which corresponds to the oxic-anoxic transition.These surface peaks are absent from the mesotrophic Lake Zurich and oligotrophic Lake Lucerne, where sedimentary concentrations of both substrates are lower (H 2 : mostly <1 nM; acetate: <4 μM).While both substrates show general concentration increases with depth in all lakes, there is a higher degree of subsurface variability in the three eutrophic lakes.Eutrophic lakes, moreover, display local peaks in H 2 concentrations in the interval from 10-20 cm.Notably, sediments from this interval were deposited during the era of peak eutrophication, and remain highly enriched in OC content today [6].

Stable carbon isotopic values of methane and dissolved inorganic carbon
Carbon isotopic analyses of DIC and CH 4 show similar trends across all lakes (Fig. 2, 1st and 2nd row).The δ 13 C-DIC values (range: −10 to +10 per mil) follow concave-down profiles in the top sediment, with the most negative values in surface sediment.The δ 13 C-CH 4 values (range: −75 and − 95 ) show the opposite trend and decrease with depth.When the same depths are compared, δ 13 C-CH 4 values are lower in Lake Zurich and Lake Lucerne than in the eutrophic lakes -a trend that was also seen in δ 13 C-DIC values and underscores the higher contribution of methanogenesis to organic matter mineralization rates in the eutrophic lakes [6].The α C values were generally >1.065 in sediments and increased with depth, showing no systematic trend in relation to trophic state (Fig. 1, 3rd row).Assuming mainly CH 4 production from CO 2 and acetate, these values indicate dominance of CO 2 reduction.

Thermodynamic calculations of methanogenesis
To examine the energetic potential of different methanogenesis reactions, we calculated the Gibbs energies of hydrogenotrophic and aceticlastic methanogenesis (Fig. 2B, 1st and 2nd row), as well as methanogenesis from methanol (methanol+H 2 : Fig. 2B, 3rd row; methanol only: Fig. S1).With exception of the sediment surface, G' CO2 + H2 values were > 0 kJ mol −1 , indicating that CO 2 reduction with H 2 was thermodynamically unfavorable.This result was surprising given that the δ 13 C-CH 4 and α Cvalues (Fig. 2A) indicated predominantly methanogenesis by CO 2 reduction, and that H 2 is often assumed to be the main electron donor for methanogenic CO 2 reduction in sediments.By contrast, aceticlastic and both methylotrophic reactions were exergonic.
G' acetate values were slightly more exergonic than the theoretical minimum energy gain for energy conservation (biological energy quantum (BEQ); −10 kJ mol −1 ; [57,58]) in surface sediment (0-5 cm) and stabilized at BEQ values below.The Gibbs energies of both methylotrophic reactions were even more exergonic (−50 to −30 kJ mol −1 ; assuming [methanol] = 1 nM) and in a similar range compared to each other, with on average slightly lower values for the reaction with methanol only.These highly similar Gibbs energy values of both methylotrophic reactions are independent of methanol concentrations.For instance, methanol concentrations of 1 μM would lower the Gibbs energies of both reactions by ∼16 kJ mol −1 .

Community zonation of methane-cycling Archaea
McrA sequences were clustered into 1379 OTUs and taxonomically classified based on a phylogenetic tree (Fig. S2).Putative methanogens of the CO 2 -reducing order Methanomicrobiales made up the majority of mcrA sequences in all samples (Fig. 3; Fig. S3).Within the Methanomicrobiales, the genus Methanoregula accounted for 30 to 88% (average ± standard deviation (SD): 67 ± 12%) of total reads per sample.Other locally abundant Methanomicrobiales included the families Methanospirillaceae and Methanomicrobiaceae, the genus Methanolinea, and unclassified genus-level groups (Methanolinea-related, Methanoregula-related). Deeper layers in Lake Lucerne, and from the oligotrophic period or oligotrophic-eutrophic transition in Lake Greifen and Lake Zurich, harbored significant percentages of novel, family-level clusters of Methanomicrobiales (Environmental Clusters I and II).In addition, sequences of the CO 2 -reducing orders Methanocellales and Methanobacteriales were locally abundant in samples from Lake Lucerne.
Other groups included the aceticlastic genus Methanothrix (formerly Methanosaeta), which accounted for up to 36% (average ± SD: 12 ± 8%) of reads per sample (Fig. 3), and the methanol+H 2 -utilizing order Methanomassiliicoccales (up to 14% in Lake Baldegg; average ± SD: 5 ± 5%).The generalistic genus Methanosarcina, members of which can catabolize H 2 + CO 2 , acetate, and methylated substrates, was also detected but rare, except in surface sediments and one bottom water sample from Lake Lucerne.The only methanotrophic Archaea detected belonged to the Methanoperedenaceae.This family was rare except in deep sediment layers of Lake Lucerne (up to 51%).

Environmental factors shaping CH4-cycling archaeal communities
The diversity of CH 4 -cycling archaea was not significantly affected by the trophic status based on a comparison of species richness, alpha-diversity, or community evenness across lakes (Fig. S4).Similarly, a Non-linear Multidimensional Scaling (NMDS) of the samples based on their OTU-level community composition did not reveal a clear separation of samples based on trophic status.However, oligotrophic samples clustered toward one side of the plot and were more dispersed than the eutrophic samples, indicating higher community variation in sediments that were deposited under oligotrophic conditions (Fig. 4A).
To investigate whether certain taxa (OTUs) are more abundant in sediments deposited under a specific trophic state, we performed a SIMilarity PERcentage analysis (SIMPER; Fig. 4B).We identified 253 OTUs with a significantly higher average relative abundance in oligotrophic samples (12 to 97% of reads in any oligotrophic sample).The proportion of these "oligotrophic OTUs" increased with depth in Lake Lucerne and reached similarly high values in the deepest sediments from eutrophic lakes, which were deposited before these lakes became eutrophic.We also identified 68 and 205 OTUs with significantly higher percentage contributions in eutrophic (9 to 57%) and mesotrophic samples (33 to 74%), respectively.By comparison, most OTUs (852 out of the total of 1379) showed no significant differences in read percent contributions between trophic categories (1.7 to 67% of reads in any given sample).

Methanogenesis pathways based on metagenome-assembled genomes (MAGs)
The isotopic indications that methanogenesis is dominated by CO 2 reduction (Fig. 2A) appear at odds with thermodynamic calculations, which indicate that CO 2 reduction with H 2 , which is often considered the main electron source for CO 2 reduction, is energetically unfavorable (Fig. 2B).To address this apparent contradiction, we analyzed genes indicative of methanogenesis pathways in 16 CH 4 -cycling archaeal MAGs completeness: >50%; genome contamination: <5%; Fig. 5, Fig. S5; Table S2) obtained from metagenomes of 35 sediment samples from the studied five lakes.
Based on phylogenomic analyses and complementary phylogenetic analyses of mcrA sequences, which were recovered from all but 3 MAGs (Fig. 5), 14 MAGs belonged to CO 2 -reducing Methanomicrobiales (12 Methanoregulaceae, 2 unassigned) and two to the aceticlastic family Methanotrichaceae (Fig. S5).At least four genes involved in hydrogenotrophic methanogenesis were present in all of these MAGs (Fig. 5).Consistent with past analyses [59,60], these also included MAGs of Methanotrichaceae.Despite not using H 2 as an electron source, members of this family express these genes during CO 2 -reduction via DIET [60].All Methanotrichaceae MAGs also contained genes for aceticlastic methanogenesis via the high-affinity enzyme acetyl-CoA synthetase.Although all Methanomicrobiales MAGs also had an acetyl-CoA synthetase gene, the encoded enzyme belonged to a phylogenetically distinct group that is not involved in methanogenesis (Fig. S7).Consistent with the ability of certain Methanomicrobiales to oxidize formate to obtain H 2 for CO 2 reduction, eight MAGs included genes encoding H 2 -evolving [NiFe]-hydrogenase 4a and nine included genes  In addition, we explored alternative pathways of CO 2 reduction, e.g.via EET involving conductive structures (multi-heme cytochromes, conductive pilins).It should be noted here that although archaea are able to receive extracellular electrons, molecular mechanisms of EET on the archaeal side remain largely unknown and are therefore hard to identify via databasedriven genome annotation.While no EET cytochromes were found, we identified genes encoding archaellum components in 14 MAGs and the gene encoding the archaellin monomer FlaB in eight MAGs.Archaellum of a hydrogenotrophic methanogen (Methanospirillum sp.) was recently shown to be electrically conductive [61].

Potential syntrophic partner organisms of methanogenic archaea
We searched our metagenomes for potential partner organisms, that could supply electrons to methanogens by direct interspecies electron transfer (DIET), by identifying MAGs whose abundance was positively correlated to methanogenic MAGs (ρ > 0.75, P < .05;Table S3).A total of eight MAGs, all belonging to Bacteria, met these criteria, of which six belonged to the phylum Desulfobacterota (4 Syntrophales, 1 Desulfatiglandales, 1 SM23-61), one to Chlorof lexota (GIF9 Dehalococcoidia), and one to Eisenbacteria (formerly Latescibacterota) (Fig. 5).Seven MAGs encoded genes for bifurcating and hydrogen evolving hydrogenases, i.e., ([NiFe] group 3d, [FeFe] group A3), and formate dehydrogenases, indicating both hydrogen-and formate-producing fermentation Figure 3. Community composition of CH 4 -cycling archaea based on mcrA gene amplicon sequences (bw = bottom water; sampled 5 to 10 cm above the sediment surface in each core).OTU-representative sequences were classified based on their placement in the phylogenetic tree of mcrA gene sequences (Fig. S2).Horizontal dashed lines in Lakes Greifen, Baldegg, and Zug indicate the timing of past shifts from oligotrophic to eutrophic, while in Lake Zurich they indicate changes from oligotrophic to eutrophic (lower line) and from eutrophic to mesotrophic (upper line).as possible catabolisms.In addition, indications for respiratory capabilities were found.Four MAGs (one of each Desulfobacterota SM32-61, Desulfatiglandales, Dehalococcoidia, and Syntrophales) also encoded genes for nitrate and sulfate reduction, whereas four others (three Syntrophales, one Eisenbacteria) had genes encoding NADH:ubiquinone oxidoreductase (respiratory complex I) subunits ( Fig. S6).
Within these MAGs, we found several genes encoding membranebound multi-heme cytochromes with 3-8 heme binding sites (Fig. 5).All MAGs, moreover, contained genes involved in pilus assembly, and three had pilA genes encoding the monomer constituting the pilus itself.Besides pilA sequences, we identified homologs of genes that are believed to be involved in FLEET in Listeria monocystogenes [56] in the MAGs of potential syntrophs.The presence of these genes in bacterial MAGs that positively correlate in abundance with methanogenic MAGs confirms the potential for these bacteria to engage in DIET with methanogenic partners.

Discussion
Increases in the supply of algal organic matter to lake sediments due to eutrophication have been shown to increase sedimentary CH 4 production and atmospheric CH 4 emissions from lakes [6,[62][63][64].Less is known about how these increases in OC supply and methanogenesis rates affect sedimentary communities of methanogens.Here we use time-resolved sedimentary records of five temperate lakes in central Switzerland to examine how changes in trophic state over the last century have affected the abundance, community structure, and metabolic pathways of methanogens.

Eutrophication impacts on abundance of methanogenic Archaea
We show that, despite increased sedimentary input and burial of algal OC in the three eutrophic lakes [31,65], which coincides with elevated sedimentary CH 4 concentrations and methanogenesis rates [6], the population size of CH 4 -cycling Archaea was higher in the meso-and oligotrophic lakes (Fig. 1).This trend is opposite to that previously documented in two lakes on the Yunnan plateau, China [64].However, in the lakes studied, the same trend was also observed for total cell numbers [6], total 16S rRNA gene copy numbers [30,31], and copy numbers of genes of aerobic methanotrophy [32], all of which were lower in eutrophic lake sediments.The fact that the eutrophic lake sediments have overall lower microbial population densities than oligo-and mesotrophic sediments with lower OC inputs and metabolic activities has been explained with higher grazing pressure by deposit-feeding Figure 4. Impact trophic status on CH 4 -cycling archaeal community composition based on mcrA gene amplicon sequences.(A) NMDS of samples based on Bray-Curtis dissimilarities at the OTU-level.Sample labels were reduced for legibility.Although no clear separation can be observed, most oligotrophic samples (circles) cluster separate from other trophic categories.Furthermore, a trend of ordering by increasing sediment depth, from left to right, can be observed for oligotrophic samples.(B) Cumulative read percentages of OTUs that showed significantly higher relative abundances in one of the three trophic categories by SIMPER (P < .05).The majority of reads belonged to OTUs that occurred at significantly higher percentages in one of the three trophic categories.Dashed horizontal lines in several plots indicate past changes in trophic state.In lakes Greifen, Baldegg, and Zug, these lines indicate past shifts from oligotrophic to eutrophic, while in Lake Zurich they indicate changes from oligo-to eutrophic (lower line) and eu-to mesotrophic (upper line).oligochaetes [ 6], which occur at high densities (∼1000 to 9000 specimen m −2 ) in the upper 10 to 20 cm of sediments in these lakes [30].Yet, differences in microbial population size do not necessarily translate into differences in activity, as cell-specific metabolic rates can vary by orders of magnitude among microorganisms in response to energy availability [57].Recent research has, moreover, demonstrated that even under intense depositfeeding with constant microbial biomass removal by macrofauna, total microbial activity can exceed that in sediments where these macrofauna are absent [66].We thus propose that the elevated input of labile organic matter in the three eutrophic lakes supports significantly higher cell-specific and total methanogenesis rates, despite (grazing-induced) smaller populations of microorganisms, including methanogens.

Impacts of trophic state on methanogenic community structure
Independent of the trophic state at the time of sediment deposition, mcrA amplicon and metagenome analyses indicate that all samples were dominated by CO 2 -reducing methanogens Figure 5. Genes indicative of methanogenesis, fermentation, and potential EET (rows) found within the MAGs of methanogens and potentially syntrophic microorganisms (columns).Numbers stand for differently annotated genes in a given category, e.g.encoding different subunits of an enzyme (e.g.methyl-CoM reductase), or different enzymes involved in a pathway (e.g.mixed-acid fermentation).For presence-absence of metabolic capabilities see Fig. S5; for the full list of annotations see Fig. S2.
(83% ± 9%), with the majority of reads belonging to the genus Methanoregula and other closely related Methanoregulaceae (order Methanomicrobiales; Fig. 3).Although the type species Methanoregula boonei was originally isolated from an acidic oligotrophic peat bog [67], other uncultured members of Methanoregulaceae have since been shown to be widespread in lake sediments [22,68,69].Phylogenetic (Fig. S2B) and phylogenomic analyses (Fig. S5) show that the Methanoregula in our samples belong to previously unsequenced species.Other widespread, putatively CO 2 -reducing Methanomicrobiales include members of Methanospirillaceae (mainly bottom water and surface sediment), Methanolineaceae (no clear pattern in distribution), and five uncultivated environmental clusters (Fig. S2B).Among the latter, the two dominant environmental clusters (clusters I and II) are closely related to Methanoregulaceae (Fig. S2B).Notably, these two clusters are rare in sediments deposited under eutrophic conditions and are the only putatively CO 2 -reducing taxa with an apparent trophic preference.
Aceticlastic Methanosaetaceae (Methanotrichaceae) and methylotrophic methanogenic taxa (methanol+H 2 ; Methanomassiliicoccales) were also widespread, but consistently less abundant than CO 2 reducers and show no noticeable trends in relation to present or past trophic state (Fig. 3).Among the aceticlastic methanogens, the obligately aceticlastic Methanotrichaceae dominated over facultatively aceticlastic Methanosarcina consistent with previous studies on lake sediments [70,71] and on the physiologies of both groups [72,73].The latter have shown that Methanothrix outcompete Methanosarcina at acetate concentrations in the low micromolar range, as in the sediments studied (Fig. 1B).This is due to the higher affinity to acetate of acetyl-CoA synthetase, an enzyme that activates acetate in Methanothrix, compared to acetate kinase, which is used by Methanosarcina [72,73].Matching these studies, MAGs of Methanotrichaceae in our study also contained genes encoding high-affinity acetyl-CoA synthetases (Fig. 5).
As in other recent studies on lake sediments [22,74], methylotrophic methanogens consisted largely of Methanomassiliicoccales (class Thermoplasmata), which reduce methanol to CH 4 using H 2 .The presence of significant percentages of Methanomassiliicoccales in almost all samples, and at much lower percentages the equally hydrogenotrophic methanol-reducing Methanomethylicales (Verstraetearchaeota; Fig. 3), indicates that methanol is the main methylotrophic methanogenic substrate in the lakes studied.Noticeably absent were potentially competing, methanoldisproportionating genera of Methanosarcinaceae (e.g.Methanococcoides, Methanolobus, Methanomethylovorans), which are widespread in marine sediments [68,75].Our calculations indicate methylotrophic reactions with and without H 2 to have similar Gibbs energy yields per mol methanol in the lakes studied (Fig. 2, Fig. S1; also see Results).Thus, factors other than in situ energy yields, e.g.differences in substrate affinity of essential enzymes, presence of syntrophic partner organisms, or availability of essential metal catalysts and cofactors, may explain the observed dominance of methanol-reducing over methanol-disproportionating methanogens.
Different from higher taxonomic levels, methanogen communities at the OTU-level show clear structuring in relation to the trophic state at the time of sediment deposition (Fig. 4).This pattern is similar to that previously observed for CH 4 -oxidizing bacteria in the same samples [32] and indicates that sediments of the five lakes are dominated by closely-related but nonetheless distinct species (or subspecies).We propose that, despite having the same basic ecological function, slight differences in ecophysiologies select for distinct methanogenic species assemblages in sediments that were deposited under different trophic regimes.

What drives the contributions of aceticlastic and methylotrophic methanogenesis?
Yet, a growing number of studies, including this study, is showing that this trend is not universal, and that methanogenesis by CO 2 reduction dominates in certain lakes [76,77].
The reasons for the uniform dominance of CO 2 reduction over other methanogenic pathways in the lakes studied here remain elusive (also see next section), as does the fact that relative abundances of aceticlastic and methylotrophic methanogens do not follow documented trends.For instance, while some studies have proposed that higher algal input of fresh organic matter favors aceticlastic methanogenesis [11,26,78], we do not observe an increase in the contribution of aceticlastic methanogens (Fig. 3) or decrease in α C -values in surface sediments of eutrophic lakes (Fig. 2A), where algal biomass inputs and contributions would be the highest.Similarly, even though methanol is produced during the breakdown of vascular plant polymers, e.g.xylan, pectin, and lignin [24,25,79], the percentage of methylotrophic methanogens does not increase in sediments of Lakes Zug and Lucerne, which have elevated contributions of terrestrial plant-derived organic matter [31].
In the case of acetate, it is possible that only small amounts of labile organic matter reach the methanogenesis zone.Sedimentation rates at the stations in the three eutrophic lakes are in the range of 0.22 to 0.37 cm yr −1 [6] and macrofaunal sediment mixing (reworking) was shown to be minimal both in the field [33] and laboratory [30], despite high abundances of deeply burrowing tubificids.In the absence of vertical mixing, organic matter will have undergone months of aerobic degradation, including depletion of its most reactive fractions, by the time it reaches the anoxic zone at 1-2 mm sediment depth.In the case of methylotrophic methanogenesis, the reason could be related to the fact that it is now known that pectin and lignin are not only produced by vascular plants, but also by microalgae [75,80].Fermentative breakdown of phytoplankton biomass could thus be an additional methanol source and explain why percentages of methylotrophic methanogens do not follow terrestrial contributions of organic matter.
While the above arguments offer reasons for the absence of aceticlastic or methylotrophic pathway trends in relation to trophic state or organic matter source, they per se do not explain why neither are quantitatively more important.One explanation is that the relative contributions of both pathways is controlled upstream by fermentative organisms.If only a minor portion of fermentation reactions yields acetate as an end product in the lakes studied, then this would explain why acetate is not the dominant methanogenic substrate.Aceticlastic methanogens (Methanothrix) may still control in situ acetate concentrations, driving these concentrations down until the apparent thermodynamic energy threshold for the reaction is reached (−10 kJ mol −1 ; Fig. 2B).Yet, this does not require acetate to be a dominant fermentation product or methanogenic substrate.In addition, competing anaerobic acetate oxidizing microorganisms may divert acetate away from aceticlastic methanogens.Indeed, thermodynamic calculations indicate that in all lakes in situ energy yields of anaerobic acetate oxidation are higher (∼0 to 10 cm sediment depth) or similar (below ∼10 cm) to those of aceticlastic methanogenesis (Fig. S8).In the case of methanol, the high energy yield of methanogenesis from methanol+H 2 suggests that methylotrophic methanogens, unlike aceticlastic methanogens, do not reach minimum energy thresholds.For these methanogens, substrate availability may be the main limiting factor.This would match past observations that methanol is not a major fermentation product in anoxic sediments [81][82][83].

What is the electron source for methanogenic carbon dioxide reduction?
Lastly, and most importantly, why do C-isotopic compositions (Fig. 2A) and methanogenic community structure indicate dominance of CO 2 reduction, while Gibbs energies of hydrogenotrophic CO 2 reduction are largely endergonic?Potential explanations include the presence of chemical microenvironments [84,85] or involvement of electron sources other than H 2 .Neither scenario can be ruled out, however, the latter appears more likely.
If chemical microenvironments are the explanation, then, according to our calculations, H 2 concentrations within these microenvironments would need to be 10-to 50-fold higher than measured H 2 concentrations in bulk sediments for Gibbs energies to reach the minimum energy threshold of −10 kJ per mol of CH 4 produced.Labile organic particles that support locally high rates of H 2 production by fermentation could generate such H 2 concentration peaks.Yet, it is unclear how these concentration peaks would be maintained in the face of rapid diffusive equilibration with much lower H 2 concentrations in surrounding sediments.Immediate efficient scavenging of produced H 2 can only decrease the H 2 concentration down to the "minimum biologically useful level" resulting in a Gibbs energy near −10 kJ per mol [58], while bulk H 2 concentrations measured in the studied sediments are far below this level.Moreover, most of our samples were from buried subsurface layers that have experienced decades of organic matter degradation and are unlikely to be rich in labile organic particles.More generally, while microniches can explain methanogenesis in environments where methanogenesis is a background process, e.g.oxic water columns or sulfate-reducing marine surface sediments [85,86], it is difficult to envision how they would be the main sites of CH 4 production in sediments where methanogenesis is the dominant respiration reaction [6].
By the same reasoning, CO 2 reduction involving formate as an electron source does not provide a plausible explanation either.This process on the intracellular conversion of formate to CO 2 and H 2 , after which the resulting H 2 is used to reduce CO 2 to CH 4 [87].However, microbial cells cannot contain H 2 , which-unless enzymatically metabolized-diffuses rapidly through microbial cell walls to the surrounding sediment [88].Taking into account the additionally very low cell-specific catabolic activity of subsurface microorganisms in the sediments studied [6], it seems unlikely that methanogens maintain intracellular H 2 concentrations that are 10-to 50-fold higher than those in surrounding sediments.Moreover, previous multiday incubation experiments involving formate addition (100 μM) to anoxic sediments from Lake Lucerne resulted in no significant stimulation of methanogenesis [89].
Support for DIET comes from the positive correlations between relative abundances of methanogenic MAGs and MAGs of certain anaerobic bacteria, as well as the genetic repertoire of these bacterial MAGs.The majority of the latter were classified as Deltaproteobacteriota, of which most belong to uncultured families of Syntrophales.Cultured representatives of the genus Syntrophus are known to form syntrophic associations with methanogens in co-cultures and anaerobic digesters (e.g.[94][95][96]) and can produce conductive pili [55].It has even been suggested that members of this genus prefer DIET over H 2 or formate exchange [55].While Walker et al. [55] suggested that S. aciditrophicus relies solely on conductive pili for DIET and does not encode multiheme cytochromes, we found c-type cytochromes with three to eight heme domains encoded in our Syntrophales MAGs and in the published genome of S. aciditrophicus ([97]; also see Supplementary material).Additionally, we found pilus assembly genes in all potentially syntrophic bacterial MAGs, although-perhaps due to MAG incompleteness-several were missing the gene for the pilin monomer.The pilin genes we found in our MAGs only partially fulfill the conductivity criteria [55], i.e. while gaps between aromatic amino acids are sufficiently short the proportion of aromatic amino acids is too low to safely conclude conductivity.Yet, pilin genes of Syntrophales that fulfill both criteria were found in the unbinned fraction of the metagenome.Lastly, we found genes homologous to the ones involved in EET mediated by extracellular f lavin shuttles [56] in MAGs of all potential syntrophs.
Identifying extracellular electron uptake genes in the archaeal MAGs is more challenging due to the lack of molecular understanding and genetic markers for DIET or EET in Archaea [27,98].Most Archaea do not have multi-heme cytochromes and only one conductive archaellin has been described so far [61].While we detected archaellin genes in most of our MAGs, the conductivity of these genes remains unclear.

Conclusions
Our combination of isotopic, community compositional, and genomic approaches indicates that methanogenesis by CO 2 reduction is the dominant methanogenic pathway in five temperate lakes irrespective of past or present trophic state, organic matter inputs, or electron acceptor distributions.At the species-or subspecies-level, the methanogenic community appears structured in relation to lake trophic state at the time of sediment deposition; however, this structuring is not evident at the genus-level and above.Similarly, the methanogenic community is not organized in relation to electron acceptor distributions.Our findings contrast with previous studies on total microbial community structure [34] and methane-oxidizing bacterial community structure [36] within sediments of the same lakes, both of which showed clear zonations in relation to trophic history and electron acceptor distributions.A potential explanation is that, despite major differences in past and present environmental conditions, anaerobic organic matter decomposition in these lakes produces a small set of products at ratios that are similar and consistently favor CO 2 reduction over other methanogenic pathways.This explanation is at odds with past research that concluded dominance of aceticlastic methanogenesis in freshwater sediments [21].To add further oddity, thermodynamic calculations rule out a predominance of H 2 -dependent CO 2 reduction.Based on genomic analyses and environmental correlations of CO 2 -reducing methanogens with syntrophic bacteria that are known to engage in EET, we propose that CO 2 reduction via DIET is the dominant methanogenic pathway in the lakes studied and a widespread process in sedimentary environments.Identifying the mechanisms by which electrons are transferred in situ, e.g.directly from partner organisms or via electrically charged chemical structures, is a priority for future research.

Figure 1 .
Figure 1.(A) Depth profiles of DIC and CH 4 concentrations, absolute abundances of methanogens based on mcrA gene copy numbers, and relative abundances of methanogens ((mcrA copy numbers / total 16S rRNA copies) × 100).Notes: (i) f luctuations in CH 4 concentrations in deeper layers of eutrophic lakes are presumably due to variations in outgassing intensity during sampling.(ii) The higher CH 4 concentrations in eutrophic lakes were not ref lected in higher mcrA copy numbers.(B) Depth profiles of methanogenic substrate concentrations (H 2 , acetate).DIC and CH 4 concentrations were previously published in Fiskal et al. (2019) [6].

Figure 2 .
Figure 2. (A) Depth profiles of δ 13 C-CH 4 , δ 13 C-DIC, and α c .(B) Gibbs energies ( G' in kJ Mol −1 ) of methanogenesis reactions with different energy substrates.Red dashed lines indicate the biological energy quantum (−10 kJ Mol −1 ).For calculations involving methanol, we assumed an in situ methanol (MeOH) concentration of 1 nM.Note: Though the high α c values in (A) indicate methanogenesis via CO 2 reduction as the dominant pathway, CO 2 reduction with H 2 was endergonic in most samples (B).

encoding a formate dehydrogenase ( 6
MAGs encoded both).Genes indicative of methylotrophic methanogenesis were not detected.

Table 1 .
Water depths of sites and trophic histories of the five investigated lakes.