The well transect of the Hainich CZE passes through different surface land use types (through five sites H1 to H5), from forest at the top of the hill slope to grassland and cropland agricultural areas at sites H4 and H5. Two superimposed aquifer assemblages are found in the limestone bedrock, partially separated by aquitards, which leads to important differences in their groundwater chemistry (Figure 1). Indeed, the availability of nitrogen compounds, dissolved oxygen, and the distance to the surface were shown to be the main drivers of community composition in the Hainich groundwater systems (Yan et al., 2020). The lower aquifer assemblage (HTL) shows intensive karstification with decreasing oxygen concentration along the flow path. The HTL recharge area is mainly located on the Hainich Hill slope in the forest area. The HTL groundwater samples were obtained from wells H41 and H51 at sites H4 and H5 (Figure 1). Well H51, located down the slope of the groundwater transect, is characterized by a higher sulfate but lower oxygen concentration, reflecting ascending sulfate-rich groundwaters and a longer water flow path. The upper aquifer assemblage (HTU) contains several minor aquifer stories with very little to no dissolved oxygen in the groundwater. Recharge occurs through a long path of ascending or lateral flow, causing a strong decrease in oxygen and surface-derived labile organic matter (Küsel et al., 2016). From the HTU, we sampled wells H43 and H52, also at sites H4 and H5, with well H52 particularly confined and isolated from the surface (Figure 1).

For each sampled well, 10,000 L of groundwater was pumped and filtered on-site using a submersible pump (Grundfos SQE 5-70, Grundfos, Denmark) connected to a stainless-steel filter device (293 mm, Millipore Corp, USA) equipped with a removable pre-combusted (5 h at 500°C) glass fiber filter (Sterlitech Corp., USA) with a 0.3 μm pore size (Schwab et al., 2017). Groundwater chemistry (e.g., O₂, NO₃, temperature, and salinity) was monitored continuously during the large volume pumping to ensure that groundwater characteristics did not change over time. The filters with concentrated microbial biomass were cut into different pieces and placed in incubation bottles containing different labeled substrates.

We used a culture-based approach based on stable isotope probing (SIP) incubations to identify heterotrophic bacteria involved in the assimilation of lignin-like molecules and in situ cell-derived biomass. For this, we incubated groundwater, together with concentrated in situ biomass on a filter piece (of ca. 1,000 L), with ¹³C-veratric acid, targeting bacteria able to metabolize lignin-like compounds. In addition to these single-labeling incubations, we also carried out dual-labeling incubations using both D-VA and ¹³CO₂ as substrates to assess if the stimulation and growth of autotrophic bacteria would have an influence on the activity and diversity of the heterotrophic community. We used ¹³C and D as isotope labels of the supplemented substrates to distinguish between the label assimilation of both heterotrophs and autotrophs in the DNA, PFLAs, and metabolites produced (Atzrodt et al., 2018). However, it was not possible to distinguish between labeling with either ¹³C or D in the DNA. We also carried out single labeling incubations with ¹³CO₂, in order to compare with the dual labeling experiments.

Two-liter borosilicate glass bottles containing 1 L of filtered groundwater for use as a medium were prepared as presented in Table 1. For each of the four studied aquifers (H43, H41, H52, and H51), 500 mg of ¹³C-labeled sodium bicarbonate (99 atom% ¹³C, Sigma-Aldrich, USA) was added in two bottles (duplicates 1 and 2), and 500 mg of unlabeled (¹²C) sodium bicarbonate (Sigma-Aldrich, USA) was added in one (control) bottle. Labeled VA (with ¹³C and D) was synthesized in-house (for details on preparation, see Supplementary Annex 1). For three of the four studied aquifers (H41, H52, and H51, we were unable to collect samples from H43), 70 mg of ¹³C-labeled veratric acid (VA) was added in two bottles (duplicates 1 and 2), and 70 mg of unlabeled (¹²C) VA (Sigma-Aldrich, USA) was added in one (control) bottle. Finally, for each of the four studied aquifers (H43, H41, H52, and H51), 500 mg of ¹³C-sodium bicarbonate + 70 mg of D-VA were added to three bottles (replicates 1, 2, and 3), and 500 mg of unlabeled ¹²C-sodium bicarbonate + 70 mg of unlabeled VA were added to one (control) bottle.

Bottles were sealed with pre-autoclaved 23.7-mm rubber stoppers with folding skirts (VWR, Germany). After filtration, the filter pieces from each aquifer were cut evenly on a sterile plate, and each filter piece was put in the aforementioned prepared bottles (Lazar et al., 2017). Before adding the filter piece to the bottles, 120 mL of ¹²CO₂ was added to the headspace of each bottle containing added sodium carbonate. This was done in order to decrease the pH to neutral values (in situ measurements indicated values of 7.20–7.35 in both aquifers). After 5 h, the pH was measured and found to be 7.2. For bottles prepared for samples from the anoxic aquifers (H43 and H52), the 1 L groundwater with the labeled and/or unlabeled substrates was transferred to 1 L borosilicate glass bottles, which were sealed with rubber stoppers with folding skirts and flushed with Argon for 30 min. After adding the filter pieces to the anoxic bottles, the headspace was re-flushed with argon for 5 min. All bottles were covered in aluminum foil to keep them dark, transported in coolers with ice packs (in situ temperature was on average 10°C), and upon return to the laboratory, the bottles were stored on an agitator (60 rpm) at 15°C for 12 weeks for samples from site H4. Because the measurement of D-PLFA indicated an extremely high amount of labeling after these 12 weeks, we decided to run the mesocosms using groundwater from site H5 for 5 weeks rather than 12. pH and oxygen concentrations in the headspace were monitored daily for the first 2 weeks and subsequently once a week (data not shown). One filter piece from each well was immediately put in dry ice in the field and subsequently stored at −20°C for the characterization of the initial microbial community (T0). After the incubation period, the groundwater medium was filtered using a 0.3 μm glass fiber filter, and all filter pieces were stored at −20°C.

Total DNA was extracted from the filter pieces (initial glass fiber filter pooled with the filtered medium after the incubations were stopped) using the RNA PowerSoil^® Total Isolation kit followed by the RNA PowerSoil^® DNA elution accessory kit (QIAGEN, Hilden, Germany). Ultracentrifugation, fractionation of the total extracted DNA, and identification of the labeled DNA were carried out as indicated in Lazar et al. (2017). The extracted DNA was separated using cesium chloride gradient ultracentrifugation. Following this, 11 to 12 fractions of 400 uL were collected by injecting sterile deionized water with a syringe into the top of the tube and collecting the drops from a hole created with a needle at the bottom of the tube in sterile Eppendorf tubes. The density of each fraction was measured using a refractometer. The collected DNA from each fraction was then precipitated by adding 1 μL of glycogen (20 mg ml⁻¹) and 2 volumes of PEG (30% polyethylene glycol 6000). DNA concentration from each fraction was then measured using PicoGreen (Invitrogen, CA, USA) and fluorescence measurements for the labeled and control incubation samples (Supplementary Figures S1–S11). A denaturating gradient gel electrophoresis (DGGE) for each fraction and the DNA concentration graphs comparing the control non-labeled and labeled incubations were also used to screen the different fractions. DNA samples were shipped to LGC Genomics GmbH (Berlin, Germany) for Illumina MiSeq sequencing with the 341F/85R primer pair. The Illumina sequence datasets were analyzed using mothur v.1.47.0 (Schloss et al., 2009). Pair ends obtained after sequencing were merged, and after processing, the sequence reads were 445 bp. Bacterial taxonomy was assigned using the Silva reference database v.138.1 (Quast et al., 2013). All sequences were deposited on the National Center for Biotechnology Information platform (NCBI) under the BioProject ID PRJNA916526.

Phospholipid fatty acids (PLFAs) were extracted from the filter piece using a method slightly modified by Bligh and Dyer (1959) and Schwab et al. (2017). Details on the methods are given in the Supplementary Annex 2. In brief, the filter pieces were cut into small parts and extracted in a phase solution of chloroform–methanol (2:1; v/v) with 0.005 M phosphate buffer. After separation into neutral lipids (NLs), glycolipid (GL), and phospholipid (PL) fractions, the phospholipids were converted to FAMEs using mild-alkaline hydrolysis and methylation (White et al., 1979). The different fatty acids were then separated using an NH₂ column before analysis on a gas chromatograph (Trace 1310 GC) coupled with a triple quadrupole mass spectrometer (TSQ-8000; Thermo Fisher Scientific, Bremen, Germany). Compounds were assigned by comparison with standards, published mass spectra (Lipski et al., 2001; Sinninghe Damsté et al., 2005), and relative retention times. Standard nomenclature was used to describe PLFAs. The number before the colon refers to the total number of C atoms; the number(s) following the colon refers to the number of double bonds and their location (after the “ω”) in the fatty acid molecule. The prefixes “Me,” “cy,” “i,” and “a” refer to the methyl group, cyclopropane groups, and iso- and anteiso-branched fatty acids, respectively.

The carbon and hydrogen stable isotope compositions of pre-purified PLFAs were determined using a GC-C-IRMS system (Delta Plus XL, Finnigan MAT, Bremen, Germany). Isotope values, expressed in the delta notation (‰), were calculated with ISODAT version software relative to the reference gas and reference standard mixture. The PLFA isotope composition was corrected for the offset due to the addition of the methyl group during methylation.

The eluate obtained after the filtration of the groundwater mentioned above, was passed through a polymer-based resin-filled cartridge (500 mg Strata X, Phenomenex, Aschaffenburg, Germany). Elution was carried out as described previously (Mori et al., 2017). Details can be found in the Supplementary Annex 1.

Labeled bacterial taxa were analyzed based on the ratio of odds ratio (RoOR) calculations for each taxon (Lazar et al., 2017). A RoOR >1 for a given taxon indicates that it is more enriched in the heavy fraction compared to the light fraction in the labeled ¹³C or ¹³C/D bottles than in the control bottles incubated with unlabeled substrates (¹²C or H), suggesting labeling of this taxon. Only labeled taxa present in both duplicate bottle samples for the single labeling experiments and in two out of the three replicates for the dual labeling experiments were used to calculate the RoOR. Furthermore, we only took into consideration taxa, which were represented by a minimum of five reads in the heavy and light fractions (labeled and control mesocosms).

The PLFA ¹³C incorporation rates evidence the activity of autotrophic bacteria. The lipid D-enrichment throughout the incubations was used to track total lipid production rates (from both autotrophs and heterotrophs). The amount of ¹³C or D incorporated into each PLFA (pmole isotope L⁻¹ d⁻¹) was calculated as follows:

Here Δ^13C/DF produced indicates the increase of the incorporated label in the PLFA between labeled samples (Tend) and unlabeled samples (T0), respectively. C produced is the amount of carbon produced based on the PLFA concertation between Tend and T0.

In the dual-labeling experiments, the rate of total lipid production (Rt) and assimilation of inorganic carbon into lipids (Ra) were calculated from the production rate using D and ¹³C, respectively. The resulting ratio Ra/Rt was used to discriminate predominantly heterotrophic (≤0.3) from completely autotrophic (~1) lipid production.

Liquid chromatography coupled with high-resolution (orbitrap) mass spectrometry (LC-HRMS) is able to distinguish between carbon [M+1.0034] and deuterium [M+1.0063] isotopologues in a dual labeling approach (Baumeister et al., 2018). In a dual-SIP experiment, this technique allows determining the exometabolites with incorporated ¹³C and/or D isotopes and thus evidencing bacterial activity on the different carbon sources. In particular, exometabolites that have a number of deuterium atoms divisible by three suggest direct incorporation of a methyl group with three deuterated atoms (CD₃) of VA, i.e., they are evidence of carbon assimilation with pathways using methyl transferase.

The concentrations of oxygen, NO3-, and SO42- were measured as detailed in Lazar et al. (2017). Changes in the oxygen concentrations were measured in the headspace of the oxic incubations using a 5890A gas chromatograph (Hewlett-Packard, USA) with a thermal conductivity detector. The different anions in the groundwater were measured on an ion chromatography system DX-500 (Thermo Fisher Scientific GmbH, Dreieich, Germany) or ICS-5000 (Thermo Fisher Scientific GmbH, Dreieich, Germany). The hydrogen isotope compositions of the water were analyzed using a high-temperature reactor (TC/EA) coupled online via a ConFlo III interface to a Delta+ XL isotope ratio mass spectrometer (all units from Finnigan MAT). The results were calibrated using in-house standards and the Vienna Standard Mean Ocean Water (VSMOW) and Standard Light Antarctic Precipitation (SLAP), according to Gehre et al. (2004).

Stimulation of the heterotrophs using the supplementation of ¹³C-VA and the in situ filter-concentrated biomass caused an increase in the PLFA concentration and label incorporation, supporting growth on VA. Both sites of the lower aquifer highlight the same shared taxon, Hyphomicrobium, which is not known to degrade VA. Known VA-degraders at both sites differ, though, with Sphingobium identified at H41 and Microbacterium at H51, both using distinct pathways for VA-degradation (Taubert et al., 2018). Only one well of the upper aquifer was incubated with ¹³C-VA (H52), and in this study, we observed one taxon with the metabolic capability to degrade VA (Pseudomonas), representing <1% of the total sequences.

When comparing relative abundances of the heterotrophic taxa and ¹³C-label incorporation between the lower and upper aquifer at site H5 (H51 with H52), the heterotrophs are more active in the oxic wells (lower aquifer). When faced with VA in excess in the lower aquifer, the stimulated bacteria are mainly, if not all, heterotrophs. Recharge waters of the lower aquifer are thought to flow rapidly from the forest area located uphill, and the lower aquifer is recharged with groundwater richer in lignin than the upper aquifers (Benk et al., 2019). The presence of active VA-degraders in the lower aquifer thus established a strong potential for the consumption of surface organic matter of the lignin type organic matter in contrast to the microbial community of the upper aquifer, which shows a higher degradation potential of complex biopolymers such as those found in bacterial cells. These results show that the type and amount of surface organic matter fuelling the groundwater determine the growth potential and the metabolic activity of the groundwater bacteria. Furthermore, the active VA-degraders differ between the wells H41 and H51 of the lower aquifers, indicating local adaptations of these heterotrophs, likely due to differences in local water geochemistry and the availability of organic matter. For instance, dissolved organic matter (mean 2.1 mg/L and 1.8 mg/L at sites H41 and 51, respectively) and O₂ (mean O₂ of 5 mg/L and 3 mg/L at sites H41 and 51, respectively) decrease along the groundwater flow path (Küsel et al., 2016).

For the lower aquifer, we mainly identified heterotrophs based on the DNA-labeled sequences. For both sites (wells H41 and H51), the Ra/Rt ratios of the PLFAs further supported a heterotrophic-dominated growth of the bacterial community when stimulated with both D-VA and ¹³CO₂. Furthermore, at the H4 site (well H41), the occurrence of labeled exometabolite with only a factor of 3D atoms indicated a direct incorporation of CD₃ derived from the aromatic methyl ether group into the exometabolites. This shows that the serine pathway is an important carbon assimilation pathway, that the lignin-derived methoxyl groups are essential carbon donors, and that the microbial organisms initially present in the groundwater have the physiological capabilities to carry out this metabolism.

From a hydrogen isotopic point of view, this indicates that in groundwater habitats, organic matter, in addition to water, can be an important source of hydrogen in the newly biosynthesized lipids. During lipid synthesis, the H isotope fractionation between water and the lipids appears to vary from −200 to +200‰ (Sessions et al., 1999). Small changes in this fractionation are induced by variations in the H isotope composition of cellular water, acetate, and the co-factor NADPH, and by different lipid biosynthetic pathways or exchanges between different cell compartments (Zhang et al., 2009). The fractionation we estimate by comparing lipid and water δD values in our experiments is inconsistent with this range. This result suggests that fatty acid elongation in the serine (C1) pathway introduces VA directly derived from CD₃ into lipids as well as metabolomes, thereby profoundly altering their δD compositions. In oligotrophic environments, reusing the CH₃ group instead of transforming it could be a way to save energy. This has a potential impact on the interpretation of δD values in lipid PLFA in heterotrophic organisms.

Another notable result is the shift of bacterial community taxa involved in VA-degradation when only faced with excess VA (single labeling experiments) and when faced with both CO₂ and VA in excess, at well H41. When only VA is present, the lower aquifer at site H4 is dominated by Sphingobium and by Microbacterium when both substrates are added. We can assume here that the other organisms stimulated by the CO₂ addition either changed the geochemical surroundings or that they worked very closely with the VA-degraders, leading to the observed diversity switch. Incubation time could also be a factor since unclassified Microbacteriaceae dominated both the single VA and dual labeling experiments from site H5 (well H51). The fact that the labeled chemoautotrophs we observed in the single ¹³CO₂ labeling experiments were not identified in the dual labeling experiment indicates that there is potential for CO₂ fixation in the lower aquifer, but that when in the presence of excess organic matter presumably originating from the forest soils, VA-degradation is a key metabolic pathway.

When VA and CO₂ were added, heterotrophic growth of the acetogens was the major pathway occurring at site H4 (well H43), supported by the observation that all initial VA was consumed and only up to 40% of the carbon in the PLFAs derived from the inorganic carbon source. High production rates in anoxic zones as anaplerotic carbonate incorporation may play an important role in compensating oligotrophic conditions and promoting cell activity maintenance and/or longer survival (Alonso-Sáez et al., 2010). Alternatively, the acetate produced stimulated the activities of many heterotrophs, as observed during our incubations (e.g., Ferribacterium or Curvibacter).

At site 5 (well H52), we observed the same two major taxa during the ¹³CO₂ + D-VA labeling experiment as with the single ¹³C-VA labeling experiment (unclassified Thermodesulfovibrionia and BSV13 Prolixibacteraceae). Thiosulfate-oxidizing autotrophs (Sulfuricella and Sulfuritalea) and ammonia-oxidizing autotrophs (unclassified Nitrosomonadaceae) were detected in the dual labeling experiment but not in the single ¹³CO₂ labeling experiment. This indicates that these autotrophs need the presence of bacteria stimulated by the presence of VA. This and the PLFA Rt/Ra ratio showed a higher potential for CO₂ fixation at H52. Metatranscriptomic datasets have previously highlighted H52 as a hotspot for CO₂ fixation using sulfur and ammonia as energy sources (Wegner et al., 2019; Overholt et al., 2022). Finally, if we compare the dual labeling experiments in the upper aquifer between sites 4 and 5, we can observe extremely dissimilar communities, showing very localized activities. Because H52 is thought to be fuelled by sulfur-rich water ascending from the well H51, a sulfur-based community may predominate, supported by the detection of sulfur-based reducers and oxidizers.