Sources of solutes and carbon cycling in perennially ice-covered Lake Untersee, Antarctica

Perennially ice-covered lakes that host benthic microbial ecosystems are present in many regions of Antarctica. Lake Untersee is an ultra-oligotrophic lake that is substantially different from any other lakes on the continent as it does not develop a seasonal moat and therefore shares similarities to sub-glacial lakes where they are sealed to the atmosphere. Here, we determine the source of major solutes and carbon to Lake Untersee, evaluate the carbon cycling and assess the metabolic functioning of microbial mats using an isotope geochemistry approach. The findings suggest that the glacial meltwater recharging the closed-basin and well-sealed Lake Untersee largely determines the major solute chemistry of the oxic water column with plagioclase and alumino-silicate weathering contributing < 5% of the Ca2+–Na+ solutes to the lake. The TIC concentration in the lake is very low and is sourced from melting of glacial ice and direct release of occluded CO2 gases into the water column. The comparison of δ13CTIC of the oxic lake waters with the δ13C in the top microbial mat layer show no fractionation due to non-discriminating photosynthetic fixation of HCO3– in the high pH and carbon-starved water. The 14C results indicate that phototrophs are also fixing respired CO2 from heterotrophic metabolism of the underlying microbial mats layers. The findings provide insights into the development of collaboration in carbon partitioning within the microbial mats to support their growth in a carbon-starved ecosystem.

www.nature.com/scientificreports/ and 6.5 km long, making it the largest freshwater lake in central Dronning Maud Land (DML) 10 . Untersee Oasis is part of a polar desert regime, the climate subjecting the area to intense ablation limits surface melt features due to cooling associated with latent heat of sublimation (e.g., Refs. 13,21,22 ). Lake Untersee has two sub-basins; the largest, 160 m deep, lies adjacent to the Anuchin Glacier and is separated by a sill at 50 m depth from a smaller, 100 m deep basin to the south. The northern deep basin is wellmixed due to buoyancy-driven convection caused by melting of the ice-wall at the glacier-lake interface 11,23 ; however, the smaller southern basin is chemically stratified below the sill depth (c. 50 m) and its higher density prevents mixing with the overlying oxic water column 24,25 . Lake Untersee loses c. 1% of its water annually from the sublimation of the 2-4 m thick ice cover. To maintain hydrological balance the lake must be recharged by an equal inflow (i.e., Refs. 13,26 ). The lake is dammed at its northern sector by the Anuchin Glacier and mass balance calculations suggest that subaqueous melting of terminus ice contributes 40-45% of the annual water budget with subglacial meltwater contributing the remainder 13 . Based on δD-δ 18 O of the water column, the lake has not developed a moat for at least the past 300-500 years 17 . The oxic water column has a Na(Ca)-SO 4 geochemical facies with uniform temperature (0.5 °C), pH (10.5), dissolved oxygen (c. 150%), specific conductivity (c. 505 µS cm -1 ) and chlorophyll (0.15 µg L −1 ) 11,12 . The nutrients are very low with TIC (0.14 mg C L −1 ), NO 3 (0.2 mg L −1 ), NO 2 (0.11 mg L −1 ), NH 4 (40 µg L −1 ) and phosphate (< 1 µg L −1 ) 15,27 . Approximately 5% of incident irradiance is transmitted through the lake ice, with a vertical extinction coefficient for scalar photosynthetically active radiation (PAR) of 0.033 m −1 , resulting in a 0.1% surface irradiance depth of ~ 135 m 12 . Lake Untersee supports an exclusively microbial ecosystem along its lake bed with photosynthetic microbial mat communities observed to depths of at least 130 m 12,15 . The mats are composed of filamentous cyanophytes that form mats, cm-scale cuspate pinnacles, and large conical stromatolites that rise up to 70 cm above the lake floor. The mats are characterized by lamina of organic material and silt-and clay-sized sediments with occasional blocky plagioclase crystals. The finegrained sediment enters the lake as glacial flour through subglacial melting and deposited on the microbial mats.

Results and discussion
Source of major solutes to Lake Untersee. Lake Untersee has a Na(Ca)-SO 4 2+ ] of 166 ± 0.7, 61.8 ± 1.2, 45.7 ± 1.2 mg L −1 , respectively (Fig. 2). The δ 34 S SO4 values range from 7.6 to 8.7‰. The water column is also characterized by Sr + (18 ± 0.2 µg L −1 ) with very radiogenic 87 Sr/ 86 Sr ratios (0.71804-0.71832; Table 1). To gain insight into the source of solutes to Lake Untersee, the major solutes and their isotopes are first compared to firn cores from the DML region. The nearby Core Epica in western DML is characterized by a similar Na-SO 4 facies but with solute concentration 3-4 orders of magnitude lower than in the lake 28 . The water column in Lake Untersee undergoes cryo-concentration of solutes as water is frozen onto Figure 2. (A) Geochemical profiles in North and South basins of Lake Untersee, Antarctica. Ionic content, δ 34 S SO4 , δ 13 C TIC and F 14 C TIC . (B) Molar ratios (Ca/Cl, Na/Cl, SO 4 /Cl and DIC/Cl) in Lake Untersee compared to DML Core Epica and weathering of plagioclase minerals. Weathering simulation were performed with PHREEQC hydrogeochemical modeling. See Table 2  www.nature.com/scientificreports/ the bottom of the ice cover and the solutes remain mostly in the water column (i.e., Refs. 29,30 ). The effective segregation coefficient (Keff) during freezing varies between solutes as it is dependent on solubility, molecular symmetry and the structure of the crystal lattice 29 . A recent study of lake ice formation calculated from their freezing experiments that the segregation coefficient is similar for Ca-Na-Cl (Keff: 0.133-0.177), however, it is much higher for SO 4 (Keff: 0.299 ± 0.025) indicating that more SO 4 is incorporated in the ice 30 . If we first correct the solute concentration for ionic segregation during freezing using the Keff for respective solutes, the effect of solute enrichment by cryo-concentration can then be removed by normalizing the solutes to Cl − , a conservative tracer. Lake Untersee has significantly higher SO 4 /Cl ratios (1.765 ± 0.002) compared to those in Core Epica (0.89 ± 0.61) (one-way ANOVA, F = 25.88; p < 0.001). The lake also has significantly lower δ 34 S SO4 (7.6 to 8.7‰) than the δ 34 S tot values from nearby CV and CM (14.5-15‰; Ref. 31 ). The difference in SO 4 /Cl ratios and δ 34 S SO4 indicates the input of SO 4 to Lake Untersee is c. 50% higher with a depleted δ 34 S composition relative to the firn cores from western DML. The western DML cores are situated 400-600 km from the Untersee Oasis and span only the past 50-1,500 years, whereas Lake Untersee has been accumulating solutes since its formation c. 12-10 ka years ago. Despite these spatio-temporal differences in the SO 4 data, and in the absence of a deep ice core with SO 4 and δ 34 S measurements from DML, inferences on possible SO 4 sources to Lake Untersee can still be made. First, a local source of SO 4 can be ruled out since sulfate evaporites are not found in the vicinity of the Untersee Oasis. Secondly, volcanic eruptions are identifiable as distinct SO 4 peaks in ice cores and have δ 34 S values in the 0 to 5‰ range 32 . Core Epica shows five SO 4 volcanic peaks between 1932 and 1991 (representing c. 2% of total SO 4 in the core); however, the EDC96 SO 4 record shows that volcanic SO 4 contributed ~ 6% of the total SO 4 during the Holocene 33 . Finally, the concentration of SO 4 and δ 34 S SO4 from Dome C and Taylor Dome ice cores are 2-3× higher and c. 2‰ lower (9.5-12.6‰), respectively, during the last glacial maximum than during the Holocene as a result of increased meridional transport of marine biogenic SO 4 and terrestrial dust [34][35][36] . Therefore, despite the differences in SO 4 /Cl and δ 34 S SO4 between Lake Untersee and the nearby DML firn cores, it appears that SO 4 in the lake is sourced from glacial meltwater. Considering that Lake Untersee formed c. 12-10 ka years ago during deglaciation, the initial late Pleistocene glacial meltwater filling the basin would likely have had a higher SO 4 /Cl with depleted δ 34 S signature relative to shallow firn cores from western DML. Then, continuous recharge from glacial meltwater during the Holocene with likely higher volcanic SO 4 and SO 4 -bearing terrestrial dust load (both with low δ 34 S; i.e., Refs. 32,37 ), as evidenced from the EDC96 SO 4 record, also contributed to the SO 4 in Lake Untersee.
Lake Untersee also has significantly higher Na/Cl and Ca/Cl molar ratios (2.468 ± 0.0483 and 1.048 ± 0.027, respectively) compared to those in Core Epica (0.494 ± 0.213 and 0.0336 ± 0.0365, respectively (one-way ANOVA, Na/Cl: F = 7,013.3, p < 0.001; Ca/Cl: F = 770.2, p < 0.001) (Fig. 2). The higher ratios can be explained in part to the higher [Na + ] and [Ca 2+ ] in glacial ice during last glacial maximum 34 , however the Na/Cl at Dome C only averages 1.83 ± 1.80 over the past 44 ka 34 . References 10,27 suggested that the enrichment in Na + and Ca 2+ can be www.nature.com/scientificreports/ attributed to weathering of silicate minerals (i.e., plagioclase: albite, anorthite and kaolinite). X-ray diffraction analysis revealed that the mineralogy of the fine particles in the mats includes anorthite (40-58%) and kaolinite (7-10%), with a few discrete sand layers having higher quart/tremolite. Here, we test the effect of plagioclase weathering in the water column under closed-system condition, which simulates a well-sealed lake with no exchange of gases. Geochemical modeling using PHREEQC hydrogeochemical software 38 predicts that closedsystem (P CO2 = -10) weathering of plagioclase minerals from glacial meltwater would generate water with a pH of 7.7-11.7 with a range of Na/Cl and Ca/Cl ratios that depend on the mineral assemblage: in all modeled scenarios Ca/Cl ratios (18 to 542) were higher than values in Lake Untersee; however only weathering of the mineral assemblage with albite yielded Na/Cl ratios higher than the lake (Table 2). Therefore, a two-component mixing calculation between the modeled Ca/Cl and Na/Cl ratios and glacial meltwater composition from Core E suggests that weathering of plagioclase contributes a minor portion of the Na + and Ca 2+ load (< 5%) relative to solutes contributed from glacial melt. Strontium in Lake Untersee is considerably more radiogenic (0.71804-0.71832) than the local anorthosite and norite (0.7079-0.7118), dust in glacial ice (0.708047-0.711254) and modern sea-water (0.70918) ( Table 1). Therefore, a more radiogenic source of Sr is needed to explain the Sr isotope ratios in the lake. Refs. 39,40 demonstrated the preferential solubilization of 87 Sr in the early stages of weathering of glacial deposits due to the rapid weathering of biotite relative to plagioclase. Although the regional basement rocks are dominantly anorthosite, norites and some dykes, the morainic material contains boulders of varied lithology including biotite-bearing gneisses. Thus, weathering products from glacial scouring and/or subglacial weathering of the biotite-bearing gneiss may enrich the 87 Sr/ 86 Sr in the lake waters with respect to the signature in the local Eliseev anorthosite complex.
The hydrological conditions of Lake Untersee play a key role on determining the major solute chemistry of the oxic water column. The closed-basin lake loses water only by sublimation of its ice cover and is recharged by subaqueous melting of glacial ice and subglacial meltwater with no summer moating 13 . The major solutes in Lake Untersee reflect a heritage of 12-10 ka of recharge from glacial meltwater and cryo-concentration of the ionic load in the water column. Early in its existence, the lake would have received a higher ionic concentration from late Pleistocene glacial meltwater that shifted to lower ionic load from glacial meltwater during the Holocene (i.e., Refs. [34][35][36] ). Weathering of local plagioclase contributes a small fraction of ions and Sr isotopes suggest weathering from biotite-bearing gneisses within the drainage basin. However, it is still unclear if weathering occurs primarily while glacial flour is suspended in the lake, within the mats on the lake floor where a higher pCO 2 from respiration would enhance weathering, or in the subglacial meltwaters along the basement rocks.
Source, fixation and cycling of carbon in Lake Untersee. The oxic water of Lake Untersee has TIC concentration near detection limit (0.3-0.4 mg C L −1 ) with δ 13 C TIC ranging from -6.8 to -9.8‰ (average = -9.1 ± 0.4‰) (Fig. 2). The measured TIC concentration are similar to those measured by Ref. 15 (0.14 mg C L −1 ), but significantly lower than values reported by Ref. 25 (DIC: 2-8 mg C L -1 ; δ 13 C DIC : 4.3‰). The higher DIC and δ 13 C DIC measured by Ref. 25 is likely due to atmospheric contamination during sampling and filtering of samples in the field. When exposed to air during filtering, the poorly-buffered high pH lake water (pH 10.5) quickly exchanges with atmospheric CO 2 ; PHREEQC modeling calculates DIC of 1.5 mg C L −1 and δ 13 C DIC 3‰ for lake water equilibrating with atmospheric CO 2 . The TOC in the oxic water was below the detection limit (< 0.3 mg C L -1 ) in most samples (one sample at 40 m depth in the south basin yielded DOC of 0.4 mg C L −1 with δ 13 C DOC of -27.5‰). However, two composite 1L samples analyzed for 14 C TOC yielded 0.27 mg C and 0.22 mg C, suggesting that TOC is c. 0.12 mg C L -1 . The TOC is about one order of magnitude lower than in other perennially Table 2. Results of numerical simulations using PHREEQC hydrogeochemical software 38 . Simulations were conducted using as input the average chemical composition of glacial meltwater (Core Epica, western Dronning Maud Land 28 ) and were equilibrated with atmospheric CO 2 (P CO2 = -3.43) or we simulated weathering of plagioclase mineral under closed-system (P CO2 = -10), which simulates a well-sealed lake with no exchange of gases. Calcite is allowed to precipitate if saturation is reached. al albite, an anorthite, ka kaolinite, pl plagioclase.  41 ), but is closer to the average compiled for Antarctic glacial ice determined from a range of ice sheet and valley glaciers (0.14 mg C L −1 ; Ref. 42 ). Tritium ( 3 H) and radio-iodine ( 129 I) measurements were below detection limit in the water column (< 0.8 TU and < 7.5 × 10 5 atoms L -1 , respectively). Radiocarbon analysis of TIC yielded similar F 14 C TIC (0.414 to 0.436), which is equivalent to apparent ages of 6,666 ± 98 to 7,054 ± 88 years BP (Table 3). For 14 C TOC , two pairs of samples were combined to obtain sufficient carbon for analysis (NB-10 and NB-40; NB-60 and NB-120) and yielded F 14 C TOC of 0.42 ± 0.02 (6,906 ± 610 years BP) and 0.55 ± 0.02 (4,734 ± 606 years BP), respectively ( Table 3).
The TIC (0.35 mg C L −1 ) and δ 13 C TIC (-9.1 ± 0.4‰) in Lake Untersee are first compared to six oligotrophic closed-basin lakes in the MDV. The wide range in DIC (2 to 1,000 mg C L −1 ) and δ 13 C DIC values (-5 to 10‰) of the MDV lakes are a product of current and past recharge mechanisms, conditions of the ice cover, and biogeochemical processes occurring in the lakes, including mineralization of organic matter and dissolution of CaCO 3 . The DIC is brought into the MDV lakes by ephemeral proglacial streams and its concentration is largely influenced by ice cover phenology: DIC is higher in lakes that did not lose their ice cover over the past few millennia and the effect of cryo-concentration (i.e., Bonney, Vanda, Fryxell); DIC is lower in lakes that lost their ice cover due to CO 2 degassing and calcite precipitation as the pH decreased (i.e., Hoare, Joyce; Refs. 3,8,43 ). The δ 13 C DIC of the MDV lakes also shows that DIC is not a limiting factor to primary productivity of photosynthetic mats. When abundant CO 2 is available for carbon fixation by the Rubisco enzyme, photosynthetic activity preferentially utilizes 12 C, producing microbial mats with depleted δ 13 C and leaving the residual DIC enriched in 13 C 7,44,45 . The maximum fractionation factor by cyanobacteria ranges from 1.023 to 1.018 46,47 , which is the difference in δ 13 C DIC and δ 13 Cmats in many MDV lakes 7,45,48 . Conversely, re-mineralization of δ 13 C depleted organic matter would produce δ 13 C DIC with similarly low values, as seen near the bottom of Lake Vanda 7 . Based on the findings of MDV lakes, a low TIC in Lake Untersee is unexpected considering that Untersee has been well-sealed for the past few hundred years 17 and likely throughout most of the history of the lake due to the paucity of calcite in the mats 12 . The mats in Lake Untersee are active, albeit with a lower gross photosynthesis and sequestration rates of organic carbon than mats in MDV lakes 15 , and as such the δ 13 C DIC would be expected to be similar to those in the MDV lakes. A significant question then is, why are the TIC and δ 13 C TIC in Lake Untersee so different from those in the MDV lakes?
The much lower DIC concentration in Lake Untersee can be attributed to the absence of ephemeral surface streams feeding the lake and its rather unique recharge mechanism: glacial meltwater that does not come in direct contact with the atmosphere. The subaqueous melting of the Anuchin Glacier releases occluded air directly into the water column and the occluded atmospheric CO 2 gas in the ice then becomes a contributor to the TIC pool. Concentrations of TIC and δ 13 C TIC in the subaqueous meltwater can be estimated from the occluded gas volume and its age, firn porosity at the depth of closure and CO 2 concentration. The Anuchin Glacier at the lake-glacier interface is c. 160 m thick and likely contains occluded gases with an average age of 1,800 years (i.e., based on age-depth relation of occluded gases from nearby EDML ice core; 17 ). Assuming an average gas volume of 0.11 cm 3 g -1 in ice 49,50 and an average CO 2 concentration of 300 ppm for the late Holocene 51 , a TIC concentration of 0.0165 mg C L -1 is estimated for the contribution of subaqueous meltwater; this value is 2-3 orders of magnitude lower than the [DIC] in streams feeding the MDV lakes 43 . The direct release of atmospheric CO 2 gases into the water column is not a fractionating process, and the δ 13 C TIC would preserve the δ 13 C of occluded CO 2 gas (δ 13 C CO2 from Dome C, and Talos Dome range from -7.3 to -6.3‰; Refs. 52,53 ). Lake Untersee is also recharged by subglacial meltwater, and the DIC in this environment can originate from three sources: melting of basal ice, dissolution of carbonates or oxidation of organics in the bedrock; all three sources have δ 13 C and 14 C composition that differ substantially. In the Precambrian anorthosite bedrock, the carbonate and organics would be 14 C dead with δ 13 C values near 0 and c. -25‰, respectively 54 . This which would result in much lower the F 14 C TIC of the water column and vastly different δ 13 C TIC . Therefore, a major contribution from subglacial carbonate dissolution or oxidation of organic carbon (i.e., Refs. 55,56 ) can be ruled out, and the DIC in the subglacial meltwater is likely the result of basal melting with a similar TIC and δ 13 C TIC as the direct melting of the Anuchin Glacier. Table 3. Radiocarbon results of total inorganic carbon (TIC) and total organic carbon (TOC) in Lake Untersee, Antarctica. F 14 C, fraction modern carbon; 14 C year BP, years before 1950; a, samples NB-10 and NB-40; b, samples NB-80 and NB-120 were combined to yield sufficient carbon for analysis. www.nature.com/scientificreports/ Dissolution of occluded CO 2 undergoes hydrolysis to hydrolytic acid (H 2 CO 3 ) in the lake, which is then utilized in weathering reactions of plagioclase: the resulting HCO 3 and thus DIC would have the same δ 13 C and 14 C composition as the source CO 2 57 . The DIC/Cl cannot be used to infer the contribution of other sources of DIC to the lake because DIC/Cl in lake is much lower than potential inputs due to significant sequestering of carbon by the mats. The δ 13 C TIC (-9.1 ± 0.4‰) is slightly lower to the δ 13 C of occluded atmospheric CO 2 in Antarctic ice cores (-7.3 to -6.3‰; Refs. 52,53 ) and the minor depletion can be attributed to oxidation of the trace amount of DOC with low δ 13 C in the water column. However, the fact that δ 13 C TIC is not enriched over δ 13 C CO2 , like most lakes in the MDV, suggest that photosynthesis is not fractionating the DIC reservoir. This can occur in an environment where the phototrophic activity is carbon-starved, and isotope fractionation by cyanobacteria becomes negligible (i.e., Ref. 44 ).
The δ 13 C and 14 C of the microbial mats provide evidence that Lake Untersee is an isotopically indiscriminate carbon-starved ecosystem that utilizes HCO 3 and CO 2 sources for phototrophic carbon fixation. The organic C abundances and δ 13 Corg of the microbial mats range from 1.0 to 5.8 wt% and -6.9 to -18.1‰, both with a general decrease with depth (Fig. 3). Several mat laminae from the three cores were radiocarbon dated, and the top laminae reported radiocarbon ages of 10,875 ± 37 years BP (Core 1), 10,052 ± 56 years BP (Core 2) and 9,524 ± 48 years BP (Core 3) ( Table 4). The bottom mat layers were dated at 16,208 ± 65 years BP (Core 1), 12,031 ± 68 years BP (Core 2) and 13,049 ± 90 years BP (Core 3). Despite some age reversals, a general increase in 14 C age is observed with depth, indicating that mats are slowly building biomass at a rate of c. 2.5 mm per 100 years. The δ 13 C TIC in lake waters (-9.1 ± 0.4‰) is similar to the δ 13 C of the top layer of microbial mats (Core 1 = -11.5‰, Core 2 = -9.2‰ and Core 3 = -9.2‰), indicating that little to no 13 C fractionation is occurring between the DIC pool and the cyanobacteria during carbon fixation. The pH of Lake Untersee is high (~ 10.5) and the dominant DIC species are HCO 3 and CO 3 2-. The cyanobacteria are therefore actively transporting HCO 3 into their cells where carbonic anhydrase subsequently catalyzes the production of dissolved CO 2 used for carbon fixation (e.g., Refs. 44,58 ). Considering that this process is energetically costly, the cyanobacteria use all of the transferred HCO 3 and no 13 C fractionation ensues. The 14 C results indicate that the cyanobacteria are also fixing respired CO 2 from heterotrophic metabolism in the underlying microbial mats layers. The top microbial mat layer (10,875 ± 37 years BP to 9,524 ± 48 years BP) are c. 3,000 years older than the TIC pool in the lake waters ( 14 C TIC = ~ 7,000 years www.nature.com/scientificreports/ BP). Since the mats are accumulating biomass at a rate of c. 2.5 mm per 100 years, the rate of carbon sequestration is not an explanation to the offset. The 14 C difference between the TIC and top mat layer is attributed to the cyanobacteria fixing carbon from two sources: (1) the HCO 3 in the water column with δ 13 C TIC of -9.1 ± 0.4‰ and 14 C TIC near 7,000 years BP, and (2) heterotrophically respired CO 2 from the decomposition of "older" buried organics with slightly more depleted δ 13 C. Metagenomic analysis for predictive metabolic functions indicates that the heterotrophic communities are capable of metabolic activities typical of recycling of organic material, such as amino-acid metabolism and carbohydrate metabolism 16 .

concluding remarks
Stromatolites may be Earth's oldest macroscopic fossils and provide plausible evidence of a biosphere on Earth 3.45 GYA 59,60 . Untersee is, at present, the only reported lake hosting the formation of modern, large conical stromatolites morphologically similar to those earliest examples 61 . The lake and its microbial ecosystem provide a glimpse into early Earth and is regarded as an analogue for Enceladus and potential ice-covered water bodies Table 4. Radiocarbon results of microbial mats in Lake Untersee, Antarctica. F 14 C, fraction modern carbon; 14 C year BP, years before 1950. *Sediments underlying mats. www.nature.com/scientificreports/ on early Mars 18 . Lake Untersee is recharged by subaqueous melting of glacial ice and subglacial meltwater and plagioclase weathering is a minor contributor of Ca 2+ and Na + . The lake remains ice-covered with no open water along the margin that would provide access to nutrients, CO 2 and enhanced sunlight [12][13][14] ; yet its floor is colonized by active cyanobacteria forming mats, pinnacles and large conical structures growing in a carbon-starved and light-limited waters 12 . The study highlights the role of carbon cycling within the laminated microbial mats where cyanobacteria utilizes both HCO 3 and CO 2 for phototrophic carbon fixation. To survive in this carbonstarved system, there is substantial DIC utilization between the cyanobacteria in the top layer of the mats and the underlying heterotrophs.
Methods field sampling. Lake water samples were collected at 5-10 m intervals in 2017-2019 using a clean 2.5 L Niskin bottle from holes drilled through the ice cover in the North and South basins. The water samples were immediately transferred into sampling bottles unfiltered due to the sensitivity of the high pH waters to rapid absorption of atmospheric CO 2 that would affect pH and DIC. Water samples for major and trace ions, were collected in HDPE bottles; waters for TIC, TOC, δ 13 C TIC analyses were collected in 40 mL amber glass vials with a butyl septa cap. Radiocarbon samples were collected in pre-baked 1 L amber glass amber bottles. Benthic microbial mats along the lake bottom were sampled for organic C content, δ 13 Corg and 14 C analyses. Three cores were collected near the eastern edge of the push-moraine at depths of 18.5 m (Core 1), 13 m (Core 2) and 17 m (Core 3) by a scientific diver using SCUBA and pushing a 4 mm diameter polycarbonate tube vertically into the mats. Sodium polyacrylate was added to the waters above the mats in the tube to form a gel seal prior to sealing with a rubber stopper. The gel seal preservation method minimizes disturbances to sediment cores during transport and shows no detectable effects on measurements of organic carbon 62 .
Lake water analyses. Major ions were measured at the Geochemistry Laboratory at the University of Ottawa. Major cations (Na, Ca, Mg, K, Sr) were acidified upon receipt at the university using 2 µL of trace metal grade 10% nitric acid and measured by an Agilent 4,200 inductively coupled plasma atomic emission spectrometer. Major anions (SO 4 , NO 2 , NO 3 , Cl) were measured unacidified using a DIONEX ion-chromatograph. Analytical precision is ± 5% for cations and anions. Charge balance error was < 2.1% when calculated using measured cations-anion and [H + ] and [OH -] for an average pH of 10.6.
Barium sulfate (BaSO 4 ) was precipitated from the oxic water samples for stable sulfur isotope ( 34 S/ 32 S) analysis of total and dissolved sulfate to investigate the source of sulfate in the lake waters. The dried and rinsed precipitates were weighed (~ 0.5 mg) into tin capsules with ~ 1 mg of WO 3 , loaded into the Isotope Cube EA to be flash combusted at 1,800 °C. The released gases were carried by helium through the EA to be cleaned, then separated. The resulting SO 2 gas was carried into the Delta Plus XP isotope ratio mass spectrometer (ThermoFinnigan, Germany) via a conflo IV interface for 34 S/ 32 S determination. All δ 34 S results are reported as permil deviation relative to VCDT and expressed using the delta-notation. The 2σ analytical precision is ± 0.4‰.
Strontium isotope ratios ( 86 Sr/ 87 Sr) were measured in lake waters to investigate potential sources of weathering. The Sr isotope ratios were measured at Queen's University Facility for Isotope Research (QFIR) using a ThermoFinnigan Neptune MC-ICP-MS with all ratios normalized to 86 Sr/ 88 Sr ratio of 0.1194 to account for mass fractionation. Results are reported as the mean of 63 consecutive measurements and corrected to internally normalized NIST standard NBS 987 ( 87 Sr/ 86 Sr = 0.710291). Reported error (2σ) is ± 0.0002 to 0.005. The replicate samples were within the analytical uncertainty.
Tritium ( 3 H) was measured by electrolytic enrichment and liquid scintillation counting. Samples were pretreated prior to enrichment by de-ionizing the waters using an ion-exchange resin and shaken vigorously for 4 h. The deionized waters were mixed with sodium peroxide (Na 2 O 2 ) and electrolytically enriched in metal cells. Samples were enriched at ~ 5.8 amps for 5 days, reducing the volume from 250 mL to approximately 13 mL. The enriched samples were then decay counted on a low-background Quantilus liquid scintillation counter. Results are reported in tritium units (TU), where1 TU = 0.11919 Bq/L; 2σ analytical precision is 0.8 TU.
Radiodine ( 129 I) was extracted from aqueous samples using a NaI carrier-addition extraction method and precipitated as AgI, using methods described in Marsh (2018). Targets were prepared by mixing 1-2 mg of AgI with high-purity niobium powder, and pneumatically pressed into a stainless-steel cathode for analysis of 129 I by AMS (HVE 3MV Tandetron). Measurements were normalized to the ISO-6II reference material by calibration with the NIST 3,230 I and II standards. Results are reported as the measured 129 I/ 127 I ratio 10 -14 (with carrier), the calculated ratio (without carrier) and the calculated sample 129 I concentration (atoms/L). The 129 I concentrations without carrier are calculated based on total iodine measured by ICP-MS, amount of I-carrier added, and the AMS measured 129 I/ 127 I ratio.
The TIC-TOC concentration and their stable isotope ratios (δ 13 C TIC , δ 13 C TOC ) in lake waters were measured by a wet TOC analyzer interfaced with a Thermo DeltaPlus XP isotope-ratio mass spectrometer using methods described by Ref. 63 at the Ján Veizer Stable Isotope Laboratory, University Ottawa. The isotope ratios are presented as permil deviation relative to VPDB and expressed using the delta-notation. The 2σ analytical precision is ± 0.5 ppm for TOC and TIC concentrations and ± 0.2‰ for δ 13 C TIC and δ 13 C TOC . The detection limit of TIC and TOC is < 0.3 ppm.
Radiocarbon analysis of waters was performed at the A.E. Lalonde Accelerator Mass Spectrometry Laboratory, University of Ottawa. Sample preparation, extraction of inorganic and organic from waters and graphitization is described by Murseli et al. 64 . Graphitized samples were analyzed on a 3MV tandem mass spectrometer and the 14 C/ 12 C ratios are expressed as fraction of Modern Carbon (F 14 C) and corrected for spectrometer and preparation fractionation using the AMS measured 13 C/ 12 C ratio 65  www.nature.com/scientificreports/ and reported in 14 C year BP (BP = AD1950) as described by 66 . The 2σ errors are reported with the results below and are below 0.013 F 14 C (or < 190 years).
Microbial mat analyses. The three microbial mat cores were first extruded from the polycarbonate tube onto a clean surface, and each core was divided in half lengthwise using a stainless-steel spatula. The core material consisted of 6-10 cm thick laminated microbial mats that transitioned sharply to sediments. The mat laminae were 0.2-1 mm thick and composed of organic material intermixed with silt and clay-sized particles. In the uppermost section of the cores, the microbial mat laminae were peeled individually, and sampling resolution increased to several laminae (1-10 mm resolution) before the mats transitioned to the underlying sediments which were sampled at 0.5-2 cm resolution. All samples were dried in aluminum trays at 55 °C, weighed, crushed to a powder using a mortar and pestle and acidified with 10% HCl to remove inorganic carbonates. The mineralogy of sediments in the mats was determined on powdered samples using a Rigaku PXRD using a high intensity 10 mm slit and a scan speed of 0.2° per minute for optimal peak resolution. The resulting intensities were peak matched to mineral phases known to exist in the region and Rietveld Refinement was done to provide relative phase abundances.
Samples were analyzed for organic C content and δ 13 C org a by purge and trap chromatography on an elemental analyzer (Vario Isotope Cube, Elementar) coupled via a Conflo III interface to a Thermo DeltaPlus isotope ratio mass spectrometer. All samples were weighed in tin capsules, mixed with tungsten oxide and flash combusted at 1,800 °C with the resulting CO 2 gases measured separately by the thermal conductivity detector. The δ 13 C values are reported against VPDB using internal standards calibrated to international standards. The analytical precision of organic C is ± 0.1% and for δ 13 C is ± 0.2‰. Results from duplicate samples were within the analytical uncertainty.
Sub-samples of individual laminae of microbial mats in the three cores were submitted to the AMS laboratory for 14 C analysis where the samples were pretreated (e.g., freeze-dried, acid-treated to remove inorganic C) according to methods described by Ref. 65 . For organic C extraction, the pretreated and dried sediments were combusted using a Thermo Flash 1,112 elemental analyzer (EA) in CN mode interfaced with an extraction line to remove non-condensable gases and trap the pure CO 2 in a prebaked 6 mm Pyrex breakseal 65 . The CO 2 in the breakseal was graphitized, measured, background corrected and reported using the same methods described for the lake water samples.
Description of ice core datasets. Lake Untersee is recharged by glacial meltwater (subaqueous melting of the Anuchin Glacier and subglacial meltwater), however deep ice core from the Anuchin Glacier and nearby ice field do not exist. To gain insight into solutes content from the glacial meltwater, the chemistry of Lake Untersee is compared to the following ice core datasets.