Lake Joux is a perialpine lake in the Joux Valley in the Swiss Jura Mountains (Fig. 1A). The valley developed in a Jura syncline, marked by glacial erosion and Quaternary deposits, and it lies mainly on Upper Jurassic and Tertiary limestones. The lake has an average depth of 32 m, a surface area of approximately 9 km2 (maximum 9 km in length and 1 km in width), and a watershed covering around 211 km2. Situated at an altitude of 1183 m, the lake is subject to intense seasonal variations and meteorological events, which drive the runoff of both natural and anthropogenic materials.

Human activity in the watershed dates back over 6850 years (Lavrieux et al., 2017; Mitchell et al., 2001; Monchamp et al., 2021). By the 16th century, the area around Lake Joux became more densely populated, leading to land drainage and deforestation for livestock farming (Piguet, 1946). Frequent crop failures and food shortages during the late 17th century spurred the growth of glassmaking and lapidary industries. Horology, introduced in the 18th century, became the region's dominant economic activity by the 19th century. This period of industrialization resulted in a transition from cultivated farmland to pastures and fallow fields (Lavrieux et al., 2017).

Agricultural intensification and urban expansion during the 20th century significantly increased nutrient inputs to Lake Joux, resulting in pronounced eutrophication. Phosphorus levels peaked at 35 µgL-1 in 1979 (Lods-Crozet et al., 2006), triggering major ecological changes, including rapid shifts in phytoplankton communities as eutrophication-adapted taxa outcompeted the lake's original species (Monchamp et al., 2021). A re-oligotrophication phase began in 1988–1989 following improved nutrient management and mitigation efforts. However, despite these reductions in external nutrient loading, the lake has not returned to its pre-eutrophication conditions. More than 70 years after the documented episode of eutrophication, the water column remains altered, suggesting that the system has shifted to an alternative stable state with a biological configuration resistant to reversal (Lods-Crozet et al., 2006; Monchamp et al., 2021).

In May 2023, three gravity cores (45–55 cm long) were recovered from the lakebed of Lake Joux using a Uwitec gravity corer. The cores were taken from one of the deepest parts of the lake (46°38^′12^′′ N, 6°17^′00^′′ E; 28 m depth) and sealed with rubber caps. One core was pre-drilled and taped at 3 cm intervals to facilitate rapid CH4 sampling using cut-off 3 mL syringes on shore. The other two cores, one for porewater and the other for sediment chemistry and microbiome analyses, were processed in the laboratory within 24 h. Porewater was extracted via N2 flushed syringes attached to 0.2 µm Rhizons (Rhizosphere), inserted every 3 cm along the core, stored at 4 °C and analyzed within 48 h. The third sediment core was opened with a handheld saw and sectioned every 3 cm. Each solid sample was split into an acid-cleaned vial and a sterile vial, then frozen at -20 °C.

Porewater samples for dissolved anions (PO43-, NO3-, NO2-, SO42-) were transferred to plastic vials while flushing with N2, capped, and analyzed using an ion chromatograph (DX-ICS-1000, DIONEX) equipped with an AS11-HC column. For dissolved inorganic carbon (DIC) porewater samples were filled into 1.5 mL borosilicate vials and capped headspace-free to prevent CO2 degassing. DIC concentration (mmolL-1) was obtained from the CO2 yield after acid conversion of aliquots transferred to helium-flushed Exetainers containing 200 µL of 99 % H3PO4, and the resulting CO2 peak areas were quantified on a GasBench II (Thermo Fisher Scientific). A response factor (µmolCO2 per peak-area unit) was derived from identically processed Carrara Marble (CM) standards and applied to the second CO2 peak of each sample; moles of CO2 were converted to DIC using the injected sample volume. External uncertainty on DIC concentration, based on CM reproducibility, was < 7 %. For δ13CDIC, the headspace CO2 was analyzed by GasBench II coupled to a Delta V Plus IRMS; each sample was measured 6 times, and we report the mean with 1σ (typically < 0.10 ‰). Values were normalized to the in-house CM standard (δ13C = +2.05 ‰) calibrated against NBS-19 and NBS-18; external reproducibility from CM replicates (n = 12) was < 0.05 ‰ (1σ). Carbon isotopes are reported in delta (δ) notation relative to the Vienna Pee Dee Belemnite (VPDB) standard, defined as δ=((Rsample\Rstandard)-1) × 1000 ‰, with R the 13C/12C ratio; positive δ indicates enrichment in 13C relative to Vienna Pee Dee Belemnite (VPDB) standard. For dissolved sulfide analysis, porewater was fixed with 0.05 M Zn-acetate (Zn(CH3COO)2 ⋅ 2H2O) solution at a 1:2 ratio immediately after extraction, and dissolved sulfide was quantified photometrically using the methylene blue method (Cline, 1969).

Sediment from the opened core was visually assessed (using standard charts) for color and granulometry based on observable differences in particle size, texture, and sorting within the sediment layers. Lithological boundaries in our core were aligned to the dated Lake Joux record of Lavrieux et al. (2017) using their carbonate sediment interval (whiter sediments) as reference. Ages were transferred from their 210Pb/137Cs model; uncertainties are those reported therein.

For total phosphorus (P), ∼ 1 g of wet sediment was digested in 9 mL of 4:2 HNO3:HCl using an Anton Paar microwave system, filtered (0.45 µm glass fiber), and analyzed by inductive coupled plasma–optical emission spectrometry (ICP-OES, Agilent 5800). Calibration used a multi-element standard, with certified reference materials yielding 85 %–102 % recovery.

Elemental C, N, H, and S were measured on 1–3 mg of freeze-dried sediment using a UNICUBE (Elementar^®) at EPFL's ISIC-MSEAP. Total organic carbon (TOC) and total inorganic carbon (TIC) were estimated by loss on ignition (500 and 1200 °C). δ13Corg was determined by Elemental Analyzer–Isotope Ratio Mass Spectrometry (EA-Isolink IRMS, Thermo Fisher) after 48 h treatment with 6 N HCl to remove carbonates. Results are reported in delta notation related to VPDB, as described above for δ13CDIC), with a reproducibility better than 0.2 ‰.

Acid-volatile sulfur (AVS) and chromium-reducible sulfur (CRS) were extracted from 1–2 g of frozen sediment as per Spangenberg and Bosco-Santos (2024). Sulfide in AVS and CRS fractions was measured colorimetrically (Cline, 1969) and CRS sulfur isotopic composition (δ34Scrs) by IRMS (Spangenberg and Bosco-Santos, 2024). These measurements help distinguish easily mobilized sulfide pools (AVS) from more stable sulfur forms (CRS) in sediments.

Oxygen concentrations were measured using a 200 µm-tip glass microsensor (Unisense) after 2-point calibration in Na-dithionite and air-saturated water. Seven vertical profiles from the same core were recorded at 250 µm steps with a motorized controller and Field Multimeter (Unisense).

For CH4 analysis, 3 cm3 of sediment was transferred into 100 mL serum bottles with 5 mL of 10 % NaOH, sealed, and homogenized. Dissolved CH4 was extracted by headspace displacement and quantified via gas chromatography (Joint Analytical Systems) equipped with an FID at the Eawag (Khatun et al., 2024). δ13CCH4 was measured using Gas Chromatography–Combustion–Isotope Ratio Mass Spectrometry (GCC-IRMS, Agilent 6890N with Thermo Finnigan IRMS) and analyzed with IonVantage software (Khatun et al., 2024). Results are reported in delta notation relative to VPDB with an analytical error < 1.1 ‰.

Carbon isotopic fractionation factors (α) between organic carbon (δ13Corg, substrate) and methane (δ13CCH4, product, were calculated as: α=(δ13Corg+1000)/(δ13CCH4+1000)). The corresponding isotopic fractionation (ε, ‰) was then determined by the relationship ε=(α-1)×1000, allowing interpretation of trends in dominant methanogenic pathways.

In order to determine sediment zonation by environmental variables, we performed non-metric multidimensional scaling (NMDS) on a Euclidean distance matrix of z-scored environmental data for samples between 0.5 and 43.5 cm sediment depth, using the function metaMDS() in the R package vegan. The NMDS stress value was 0.04. Differences among depth-defined clusters based on environmental variables were tested using a permutational multivariate analysis of variance (PERMANOVA) on smoothed Euclidean distance matrices.

DNA was extracted from Lake Joux sediments using the PowerSoil Pro Kit (Qiagen). Extraction, sequencing, and raw data processing were conducted at the Joint Microbiome Facility (Medical University of Vienna and University of Vienna; project ID JMF-2310-14). The V4 hypervariable region of the 16SrRNA gene was amplified and sequenced to assess the total microbial diversity in the collected samples. Amplification was performed with linker-modified 515F and 806R (Apprill et al., 2015; Parada et al., 2016) primers, and amplicons were barcoded, multiplexed, sequenced on an Illumina MiSeq (v3 chemistry, 2x 300 bp), and extracted from the raw sequencing data as described in detail in Pjevac et al. (2021). Amplicon Sequence Variants (ASVs) were inferred using the DADA2 R package v1.42 (Callahan et al., 2016b), applying the recommended workflow (Callahan et al., 2016a). FASTQ reads 1 and 2 were trimmed at 220 and 150 nt with allowed expected errors of 2. ASV sequences were subsequently classified using DADA2 and the SILVA database SSU Ref NR 99 release 138.1 (Quast et al., 2012; Mclaren and Callahan, 2021) using a confidence threshold of 0.5. ASVs without classification or classified as eukaryotes, mitochondria, or chloroplasts, as well as well-known buffer contaminations, were removed. After filtering, only samples with at least 7000 read pairs were kept for further analyses, and relative abundances of ASVs grouped at higher taxonomic levels were calculated in relation to all remaining data. The relative abundance of chloroplast sequences, which were removed from the microbial community dataset, was examined separately to assess phytoplankton debris abundance across the sediment profile.

Downstream analyses were performed using R v4.3.2 and Bioconductor v3.16 packages SummarizedExperiment v1.32, SingleCellExperiment v1.24, TreeSummarizedExperiment v2.8 (Huang et al., 2021), mia v1.8 (https://github.com/microbiome/mia), LMdist (Hoops and Knights, 2023), vegan 2.6-8, phyloseq v1.44 (Mcmurdie and Holmes, 2013) (Vegan R package; phyloseq R package), microbiome v1.22 (http://microbiome.github.io, last access: 2 February 2026), microViz v0.10.8 (Barnett et al., 2021), and corrplot (Wei and Simko, 2024). Microbial community alpha diversity indices were calculated on rarified 16SrRNA gene amplicon data using R packages vegan and mia. For community dissimilarity analysis, microbial 16SrRNA gene amplicon sequence count data was centered log ratio (CLR) transformed, a pairwise Aitchison distances matrix was computed, and oversaturated distances in the dissimilarity matrix were corrected and smoothed using LMdist with default settings prior to ordination using principal coordinates analysis (PCoA). Differences in 16SrRNA gene amplicon community composition among three depth-defined clusters were tested using permutational multivariate analysis of variance (PERMANOVA) on an Aitchison distance-based dissimilarity matrix.

To identify the environmental variables that significantly contributed to the variation in microbial community structure, correlations between microbial community composition and environmental variables were assessed using Mantel tests, based on Euclidean distances calculated from Z-score standardized environmental variables and LMdist corrected and smoothed Aitchison distances of 16SrRNA gene amplicon sequencing data. Prior to correlation analysis, five samples from the deep eutrophic layer without corresponding environmental data were excluded. Mantel tests were performed using Spearman's rank correlation as implemented in the R package vegan. The resulting p values were adjusted for multiple testing using the false discovery rate (FDR) method. Highly correlated environmental variables (Spearman's r > 0.8, Fig. S1 in the Supplement), as assessed by the function cor() in the R package corrplot, were removed before the Mantel tests.

The 55 cm deep sediment record of Lake Joux could be classified into three main intervals based on distinct lithological and chemical features (Figs. 1 and S1 in the Supplement). The “deep eutrophic” interval, from 55 to 30 cm, comprises black and silty sediments, indicating a period of higher lake productivity and low oxygen conditions. Occasional fine sand and organic fibers are also present. In the “middle carbonate” interval, from 30 to 11 cm, the sediments transition from a murky gray with heterogeneous brownish features, suggesting changes in organic matter quality and oxidation states (Fig. 1) to whitish silty-sandy sediments, with the contribution of shells above 13.5 cm, indicating the dominant deposition of carbonates (Fig. 1). The “upper eutrophic” interval, from 11 to 0 cm, contains intensely black sediment with frequent plant debris, reflecting recent environmental changes.

To assign approximate ages to our sedimentary profile, we correlated our lithological intervals to other Lake Joux sedimentary sequences previously published and dated using 237Cs and 210Pb (Lavrieux et al., 2017; Magny et al., 2008) (Fig. 1). The middle carbonate unit (30–11 cm) in our core aligns with their U3–U4 carbonate interval, including the distinctive pale “white” boundary at ∼ 16–11 cm. The overlying upper eutrophic black, organic-rich sediments (11–0 cm) correspond to U5, which spans the 20th-century eutrophication phase and includes the 137Cs markers at ∼ 1954 and ∼ 1963 in the Lavrieux record. By transfer of their age–depth model, the base of our deep eutrophic interval (below 30 cm) falls in the late 16th–early 17th century. Reported sedimentation rates in Lavrieux (≈ 0.04–0.11 cmyr-1 before the 18th century, a short-lived peak ≈ 0.83 cmyr-1 in the late 18th century, and ≈ 0.18 cmyr-1 over recent decades) are consistent with the thicknesses of our corresponding units (Lavrieux et al., 2017).

Sulfate (SO42-, between 0.35 and 4.5 µM) and dissolved sulfide (H2S, between 2 and 14.3 µM) were measurable throughout the entire sedimentary profile. In the upper eutrophic interval (0–11 cm), opposing gradients of SO42- and H2S from the surface to 7.5 cm are evidence of sulfate reduction (Fig. 2A and B). Below this depth, sulfate is absent, but a broad H2S maximum in the deep eutrophic sediments (∼ 40 cm) could be associated with organic sulfur degradation.

Dissolved nitrate (NO3-) and phosphate (PO43-) first appeared at 19.5 and 16.5 cm, respectively, with concentrations progressively increasing toward the surface, reaching maximum values of 27 µM for NO3- and 0.62 µM for PO43- (Fig. 2C and D). Nitrite concentrations were close to the detection limit (0.03 µM) and uniformly low (range 0.038–0.087 µM, median ≈ 0.043 µM) with no systematic depth trend (Fig. S2 in the Supplement).

Dissolved inorganic carbon (DIC) concentrations with isotopically heavier composition (δ13CDIC) were highest in the middle carbonate interval between 16.5 and 31.5 cm (Fig. 2E and F). Above 7.5 cm depth, δ13CDIC values became progressively lighter toward the surface, reaching a minimum of -12 ‰ (Fig. 2F).

The intensely black sediments abundant in chloroplast-related sequences and elevated TOC content with low C/N ratios and light δ13Corg in the deep eutrophic interval (55–30 cm, Fig. 1) denote predominantly autochthonous organic matter, derived from phytoplankton blooms (Lamb et al., 2006; Morales-Williams et al., 2017). Similar patterns recorded in other Lake Joux sediment profiles (Dubois, 2016; Lavrieux et al., 2017; Magny et al., 2008) correspond to a period of intensified deforestation and settlement expansion between 1525 and 1790 CE (Dubois, 2016; Lavrieux et al., 2017; Magny et al., 2008). While no official records confirm eutrophication during this period, these anthropogenic activities likely led to increased erosion and sediment/nutrient transport, stimulating phytoplankton productivity (Fig. 1). Notably, these sediments also exhibit high relative H2S and AVS concentrations (Fig. 2B and K), which are characteristic of late sediment diagenesis under eutrophic depositional conditions (Holmer and Storkholm, 2001).

The middle carbonate layer (30–11 cm depth) reflects a shift towards more oligotrophic conditions, likely linked to the abandonment of land-intensive activities and the switch to manufacturing in the 18th century. In addition, the 1777 construction of a dike between Lake Joux and Lake Brenet lowered the lake level by 3.6 m, mobilizing limestone-rich sediments (high TIC) and terrestrial plant material with heavier δ13Corg and higher C/N ratios (Lavrieux et al., 2017; Magny et al., 2008; Monchamp et al., 2021) (Fig. 2). The sedimentological transition that marks the beginning of this interval at 30 cm depth aligns with a shift from methylotrophic to hydrogenotrophic methanogens (Fig. 4), likely responsible for the lighter δ13CDIC values (Fig. 2F). Furthermore, the white-colored boundary of the middle carbonate layer (16–11 cm) coincides with warmer post-Little Ice Age conditions, promoting calcium carbonate precipitation and TIC enrichment (Lavrieux et al., 2017).

The upper eutrophic sedimentary interval (11–0 cm) consists of black sediments rich in TOC, lighter δ13Corg values, low C/N ratios, and abundant chloroplast- and cyanobacteria-related sequences reflecting a well-documented 20th century eutrophication phase (Lavrieux et al., 2017; Magny et al., 2008; Monchamp et al., 2021) (Figs. 2 and 3B). Elevated nutrient levels in this interval could result from external nutrient inputs trapped in porewater or from organic matter remineralization. Porewaters are strongly reducing, reflected by elevated H2S and CRS, and the absence of O2 below 0.5 cm. Downward diffusing SO42- meets upward-diffusing CH4, and in the 7.5–0 cm horizon, CH4 concentrations fall sharply while δ13CCH4 becomes heavier and δ13CDIC lighter (Fig. 2A, B and K, M). Aerobic methanotrophy fractionates carbon, preferentially consuming ¹²CH4 while leaving behind heavier CH4 (enriched in 13C) and can produce lighter (relatively less 13C) co-localized DIC depending on mixing and available electron acceptors. Together with the dominance of MOB and the absence of ANME-related 16SrRNA gene amplicon sequences, the paired isotopic shifts observed indicate methanotrophy dominated by MOB as the main CH4 sink in these anoxic, nutrient-replete surface sediments.

Changes in organic matter sources to Lake Joux over the last four centuries appear closely tied to shifts in dominant methanogenic groups within its sediments. Deep eutrophic sediments, characterized by the highest CH4 concentrations, are dominated by Methanomassiliicoccales, which are hydrogen-dependent methylotrophic methanogens, meaning that they use H2 as the electron donor and methylated one-carbon compounds (e.g., methanol, methylamines, methylated S compounds) as electron acceptors, rather than reducing CO2 (Bueno De Mesquita et al., 2023; Ellenbogen et al., 2024; Söllinger and Urich, 2019; Sun et al., 2019; Wang and Lee, 1994). The decomposition of algal and cyanobacterial biomass can release methylated sulfur compounds (including DMS and dimethylsulfoxide) and methylated amines, which have stimulated methylotrophic methanogenesis in laboratory experiments and natural environments (Bose et al., 2008; Chistoserdova, 2011; Chistoserdova et al., 2009; Huang et al., 2018; Singh et al., 2005; Tebbe et al., 2023; Whiticar, 1999; Zhou et al., 2022).

Indirect evidence for the presence of methylated sulfur compounds comes from relatively higher concentrations of H2S, AVS, and CRS at depth (Fig. 2B and K), indicating active sulfur cycling despite limited SO42- availability. Furthermore, δ34S values measured in CRS (primarily pyrite) consistently show positive isotopic signatures (7 ‰ to 10 ‰) in both the deep eutrophic and middle carbonate zones. While microbial SO42- reduction typically produces 34S-depleted sulfides (δ34S < 0 ‰) under open-system or moderately sulfate-limited conditions (Bradley et al., 2016; Canfield, 2001; Habicht and Canfield, 1997), the isotopic enrichment observed here is more consistent with either the degradation of sulfurized organic matter or methylated sulfur compounds (Phillips et al., 2022; Raven et al., 2019; Werne et al., 2004). These could simultaneously fuel methylotrophic methanogenesis and pyrite formation. This interpretation warrants confirmation through direct measurements of methylated sulfur species in future studies. Alternatively, the 34S enrichment could reflect near complete consumption of a limited SO42- pool so that 34S sulfide reflects positive values of the original sulfate (Bernasconi et al., 2017)(Fig. 2).

It is important to note that methylotroph distributions could also be influenced by competition for H2 with CO2-reducing hydrogenotrophs. In sulfate-poor anoxic sediments, H2 is typically buffered at low steady-state levels by continuous fermentative supply and rapid consumption – reflecting thermodynamic control rather than chronic scarcity (Conrad, 1999; Schütz et al., 1988; Kessler et al., 2019). Obligately methyl-reducing methanogens have very low H2 thresholds and are predicted to outcompete hydrogenotrophs for H2 when methyl groups are available. Thus, their activity is primarily considered to be limited by the availability of methylated substrates (Borrel et al., 2023; Bueno De Mesquita et al., 2023; Feldewert et al., 2020; Söllinger and Urich, 2019; Speth and Orphan, 2018). Given the dominance of Methanomassiliicoccales at depth, we infer that methylated-substrate supply rather than H2 limitation is the primary factor structuring the methanogenic community in the deep eutrophic interval. This interpretation is consistent with isotope patterns, as we have recorded comparatively heavier δ13CDIC in the deep eutrophic layer and a shift to lighter δ13CDIC above ∼ 30 cm where the relative abundance of CO2-reducing hydrogenotrophic methanogens increased (Fig. 2F).

Methylotrophic methanogenesis is typically a minor pathway in freshwater sediments because methylated substrates are scarce (Borrel et al., 2011; Bueno De Mesquita et al., 2023). However, in the deep eutrophic layer, prolonged algal biomass degradation likely generated a reservoir of recalcitrant methylated compounds (Achtnich et al., 1995; Rissanen et al., 2018), favoring methylotrophic methanogens. In contrast, hydrogenotrophic (using CO2 + H2) and acetoclastic (using acetate) methanogens primarily depend on fresh, labile organic matter, which rapidly becomes limited with burial (Achtnich et al., 1995; Meier et al., 2024; Rissanen et al., 2023; Rissanen et al., 2018). Thus, methylotrophs gain a selective advantage in these older, more refractory sediments. Above ∼ 28.5 cm, concurrent with a shift toward more terrestrial organic matter, methylotrophic methanogens decline and hydrogenotrophs progressively dominate (Fig. 4A). We interpret this pattern as consistent with a reduced supply of methylated substrates typically derived from algal organic matter although these compounds were not directly measured.

To further support the interpretation of distinct methanogenic pathways, we analyzed the εCorg-CH4, reflecting the isotopic discrimination during CH4 formation from Corg (Fig. 2O). Methanogenesis discriminates against the heavier 13C isotope (Conrad, 2005). In theory, when CH4 is produced from CO2 + H2 (hydrogenotrophy), microbes selectively withdraw 12C from the DIC pool, leaving residual DIC relatively 13C-enriched (less negative δ13CDIC); when methylotrophy dominates, CH4 carbon is drawn from methyl pools and δ13CDIC  is affected less. We observed lower ε values in deeper eutrophic sediments compared to shallower zones. Although interpreting specific metabolic pathways from isotopic fractionation is challenging in mixed microbial communities, the contrast in εCorg-CH4 (Fig. 2O) indicates distinct CH4-producing processes dominate at different sediment depths.

Our results support the view that eutrophication leaves a distinct imprint on methanogen stratification in sediments. In contrast to earlier studies that reported either weak vertical structuring (Meier et al., 2024) or only subtle shifts in methanogen dominance (Rissanen et al., 2023), we found clear zonation with Methanomassiliicoccales prevailing in the deepest eutrophic interval, Methanomicrobiaceae in the carbonate-rich middle section, and Methanobacteriaceae dominating the upper eutrophic sediments. Considered alongside these previous observations in other lakes, our findings question the usefulness of broad generalizations and suggest that methanogen communities are primarily shaped by habitat-specific conditions – such as lithology, organic-matter quality, and redox context – rather than exhibiting universal hydrogenotroph dominance. By comparison, a pronounced vertical structuring of methane-oxidizing bacteria appears more consistent across systems (Mayr et al., 2020; Rissanen et al., 2018; Van Grinsven et al., 2022).