Sediment cores were obtained from three different water depths in the oligotrophic Lake Lucerne, the mesotrophic Lake Zurich, and the eutrophic Lake Zug, Lake Baldegg, and Lake Greifen in central Switzerland in June and July 2016 (Fig. 1, Table 1; for further information on trophic histories, please see Fiskal et al., 2019). Sediment cores were taken using gravity cores with 60 cm long liners that had an inner diameter of 150 mm (UWITEC, AT) from boats or motorized platforms. The four sediment cores per station were used as follows: the most undisturbed core was used for microsensor measurements (O2, pH) and afterwards for macrofaunal community sampling. The second core was used for analyses of DNA sequences, methane concentrations, δ13C-methane, TOC content, and δ13C-TOC. The remaining cores were used for porewater sampling using rhizons (0.2 µm pore size, Rhizosphere), with DOC and δ13C-DOC sampling being done on a separate core than that for dissolved inorganic carbon (DIC) and δ13C-DIC sampling. A wide range of additional porewater geochemical analyses were performed on the core used for DIC sampling (including concentrations of nitrate, sulfate, hydrogen sulfide, Fe2+, Mn2+, and ammonium; for further details, see Fiskal et al., 2019). In all cores, the top 4 cm was sampled in 0.5 to 1 cm depth intervals, samples from 4 to 20 cm sediment depth in 2 cm depth intervals, and all deeper layers in 4 cm depth intervals. Cores were typically ∼40–50 cm long; however, the lowermost 5–10 cm was discarded due to contamination with lake water during core retrieval. An additional, narrow core 6 cm in diameter was obtained for radionuclide (210Pb and 137Cs) analyses (for analytical details, see Fiskal et al., 2019). Cores for macrofaunal community analyses were extruded and macrofauna collected by sieving sediments through 400 and 200 µm mesh sieves. Three stations (two in Lake Lucerne, one in Lake Baldegg) were revisited in November 2017 and October 2018 to collect additional macrofaunal specimens and chironomid larval tubes for DNA analyses.

For each depth interval, specimen numbers of oligochaetes and chironomid larvae were carefully picked with tweezers, counted, and preserved in 70 % ethanol for taxonomic and 13C-isotopic analyses or frozen on dry ice for DNA extractions. Detailed taxonomic analyses to the genus and, when possible, species level were performed on a subset of oligochaetes and chironomid larvae. Oligochaete specimens were sent to AquaLytis (Wildau, Germany), where they were embedded in epoxy resin and identified by light microscopy. Chironomid larvae were microscopically identified by AquaDiptera (Emmendingen, Germany).

Carbon isotope analyses were performed on DOC, methane, TOC, macrofaunal specimens, or separately on guts and remaining bodies of macrofaunal specimens. Values are given in the δ notation, i.e., δ13C=13C/12Csample/13C/12Cstandard. For δ13C-DOC, porewater samples were analyzed as described in Lang et al. (2012). Briefly, 2–7 mL of sample was added to 12 mL Vacutainers^®. After removal of dissolved inorganic C by addition of 85 % phosphoric acid and bubbling with high purity He, DOC was oxidized to CO2 using persulfate (1 h at 100∘C). The evolved CO2 was analyzed on a GasBench II coupled to a Delta V mass spectrometer (Thermo Fisher Scientific, Bremen). Water soluble organic standards of known isotope composition (phthalic acid and sucrose) were used as standards.

For δ13C-Methane, methane was extracted by creating a sediment slurry with Milli-Q water under saturating NaCl concentrations (∼6.3 M). A total of 2 cm3 of sediment was transferred to 20 mL crimp vials containing 2.514 g NaCl and 5 mL Milli-Q water, crimped, mixed, and stored on ice or at +4 ∘C until analysis using a trace gas (Isoprime) coupled to an isotope ratio mass spectrometer (GC-IRMS; Isoprime, Manchester). Separation was performed through a GC-column (PoraPLOT Q 30 m column). The precision of the method was ±0.7 %. After every sixth sample we included a standard with a known δ13C value (standards: L-iso1 with 2500 ppmv CH4 at -66.5 % δ13C-methane and T-iso3 with 250 ppmv CH4 at -38.3 % δ13C-CH4; Air Liquide).

For δ13C-TOC analyses, 5–10 g of frozen sediment was freeze-dried in glass vials and quantified using an elemental analyzer (Thermo Fisher Flash EA 1112) coupled to an isotope-ratio mass spectrometer (Thermo Fisher Delta V Plus) (EA-IRMS) as outlined in Fiskal et al. (2019).

For δ13C-Macrofauna, δ13C-analyses were performed on macrofaunal biomass according to the same method used for δ13C-TOC. Single specimens were cleaned with molecular grade water to remove sediment. Whole organisms, or separated guts and residual bodies, were placed in tin foil capsules, which were mounted to 96-well plates. The 96-well plates were sealed using plastic seal foil, the foil above each well was pierced, and the whole plate was freeze dried. Afterward, the foil was removed, the tin foil capsules were closed, and the δ13C of macrofaunal biomass was measured.

Assuming TOC and methane as the only carbon sources, a two end-member mixing model was used to estimate the contribution of methane to biomass C of macrofauna. CH4-Contribution (%)=(100-δ13Cfauna-δ13CCH4/δ13CTOC-δ13CCH4×100)

DNA was extracted according to lysis protocol II of the modular DNA extraction method of Lever et al. (2015) following the exact procedure outlined in Han et al. (2020). While we used existing sediment DNA extracts from the latter study, DNA from empty larval tubes and from macrofauna were newly extracted. To remove sediment, macrofaunal specimens were rinsed with molecular grade water. DNA was then extracted from entire specimens, or separately on guts and the remaining body, after being cut into three to four pieces to increase extraction efficiency using a sterile scalpel. All DNA extracts were stored at -80 ∘C.

Quantitative polymerase chain reaction (qPCR) was performed to quantify bacterial and archaeal 16S rRNA genes, as well as genes encoding particulate methane monooxygenase of aerobic methanotrophic bacteria (pmoA) and methyl coenzyme M reductase of methanogenic and anaerobic methane-oxidizing archaea (mcrA) (Table S1 in the Supplement). Standards consisting of plasmids containing 16S rRNA, pmoA, or mcrA genes from specific organisms (Table S1 in the Supplement) were run in 10-fold dilutions of ∼101 to ∼107 gene copies per qPCR reaction. All sample DNA extracts and standard dilutions were run in duplicate.

The qPCR protocols are shown in Table S2 in the Supplement. For each qPCR reaction, 2 µL of DNA extract was mixed with 1 µL of molecular grade water, 1 µL of bovine serum albumin (10 mgmL-1; New England Biolabs, USA), 0.5 µL each of forward and reverse primers (10 µM), and 5 µL LightCycler^® 480 SYBR Green I Master Mix (Roche, Switzerland). All standards and samples were kept on ice throughout the preparations and run immediately after in transparent 96-well plates on a Roche LightCycler^® 480.

Libraries of bacterial and archaeal communities were produced using the universal 16S rRNA primer pair Univ519F and Univ802R (Claesson et al., 2009; Wang and Qian, 2009). After library preparation DNA was pooled and sequenced using a MiSeq (Illumina Inc., USA). Library preparations and subsequent data processing, including 97 % zero-radius operational taxonomic unit (ZOTU) clustering, were done as outlined in Han et al. (2020) (for polymerase chain reaction mixtures and cycler conditions, see Table S3 in the Supplement). Briefly, raw sequences were initially quality trimmed using seqtk (https://github.com/lh3/seqtk) and paired-end reads were merged using flash (Magoč and Salzberg, 2011). This was followed by a final quality filtering using prinseq (Schmieder and Edwards, 2011). Sequences were then used to generate ZOTUs with USEARCH unoise3 using a 97 % clustering identity (Edgar, 2016).

Statistical differences between C isotope signatures of macrofauna and C pools, as well as of percentages of bacterial 16S rRNA, mcrA, and pmoA gene copy numbers relative to total 16S rRNA gene copy numbers across macrofauna, larval tubes, and sediment, were determined using Wilcoxon signed rank tests for paired data. All tests were performed in R (Team, 2018) using the following command: wilcox.test (A, B, paired = TRUE, alternative = “two.sided” for (a), “greater/less” for (b), mu = 0.0, exact = TRUE, correct = TRUE, conf.int = TRUE, conf.level = 0.95). Principal coordinates analysis (PCoA) on bacterial communities at the phylum, class, order, family, and genus levels was performed using Bray–Curtis distances in R (Team, 2018).

Macrofauna are present at all stations except the hypoxic deep station of Lake Zurich and are dominated by oligochaetes and chironomid larvae. While oligochaetes are present in all lakes, no chironomid larvae were found in Lake Greifen. Oligochaete densities increase with trophic state, from 75±86 individuals per square meter in Lake Lucerne to 4849±4443 individuals per square meter in Lake Baldegg (number of individuals are expressed as averages per lake with standard deviations of three stations). Numbers of chironomid larvae show the opposite trend, decreasing from 641±346 individuals per square meter in Lake Lucerne and 849±160 individuals per square meter in Lake Zurich to less than 75±86 individuals per square meter in the three eutrophic lakes (Fig. 2, Table S5 in the Supplement). Other macrofauna, e.g., copepods, Daphnia, and leeches, were only occasionally found and will not be discussed further.

The depth distributions of oligochaetes and chironomid larvae follow different trends (Fig. 3). Chironomid larvae are most abundant in surface sediment (0–5 cmblf, centimeters below lake floor), while oligochaetes occur over a greater depth interval (Fig. 3). In Lake Greifen and Lake Baldegg, oligochaetes are present in high numbers to 12 and 15 cm sediment depth, respectively, including layers that are distinctly laminated (see horizontal lines in Fig. 3). In Lake Zug, oligochaetes are present to even greater depths (22 cm). In sediments of Lake Zurich, where oligochaetes and chironomids occur in similar abundances, chironomids dominate the top ∼2–3 cm, whereas oligochaetes dominate below. Despite depth ranges extending significantly below the sediment surface, macrofaunal sediment reworking is minimal based on radionuclide measurements. These show 137Cs peaks that match the 1986 (Chernobyl) and 1963 (bomb test) time markers, and clear 210Pbunsupported decreases from the top 2 cm downward at all faunated stations (Fig. S6 in the Supplement; data analyzed but not shown in Fiskal et al., 2019). Light microscopic images of the two dominant macrofaunal groups and depth distributions of individual macrofaunal species can be found in Fig. S1 in the Supplement.

Oligochaetes and chironomid larvae were assigned to 9 and 14 different taxonomic groups, respectively (Fig. 4; for station-specific data, see Fig. S2 in the Supplement). All oligochaetes belong to the family Naididae (Syn. Tubificidae) and all chironomid larvae to the family Chironomidae. Two oligochaete morphotypes, Tubificidae + bristles and Tubificidae - bristles, could not be assigned to a known genus.

For Oligochaete group overlap between lakes, four of the nine groups (Tubificidae + bristles, Tubificidae - bristles, P. hammoniensis, L. hoffmeisteri) occur in four of the five lakes. E. velutinus (Lake Zurich), L. profundicula (Lake Baldegg), and P. vejdovskyi (Lake Lucerne) were the only species that were only found in one lake. Comparing the dominant oligochaete groups reveals the dominance of uncharacterized Tubificidae (+ bristles) in Lake Zurich, Lake Zug, and Lake Greifen but very different communities in Lake Baldegg, which is dominated by uncharacterized Tubificidae (- bristles) and L. hoffmeisteri. All identified tubificids except E. velutinus are subsurface deposit feeders that are believed to mainly feed on sedimentary bacteria, whereas E. velutinus is a surface deposit feeder (Table S6 in the Supplement).

Chironomid larval communities in Lake Zurich and Lake Lucerne share many members, but the dominant groups only partially overlap. Lake Zurich sediment is dominated by Micropsectra sp., Tanytarsus sp., Chironomus riparius, Chironomus piger gr., and Sergentia coracina, whereas Lake Lucerne is dominated by Procladius sp., Micropsectra sp., Macropelopia fehlmanni Kieffer 1911, Tanytarsus sp., and S. coracina. Micropsectra sp., Tanytarsus sp., and S. coracina are mainly sedimentary detritus feeders, whereas Chironomus riparius and C. piger gr. are known to mainly filter feed. Both Procladius sp. and M. fehlmanni are predators (Table S6 in the Supplement).

Average C isotope compositions of macrofaunal specimens are displayed with those of the potential C sources methane, TOC, and DOC in Fig. 5 (for depth profiles, see Fig. S3 in the Supplement). Macrofaunal values are lowest in Lake Baldegg (oligochaetes: -36.7±3.3 %, N= 14; larvae: -37.6±1.9 %, N= 4) and Lake Greifen (oligochaetes: -37.6±2.5 %, N= 12; no larvae found) and highest in Lake Lucerne (oligochaetes: -31.7±0.4 %, N= 2; larvae: -31.5±2.2 %, N= 24) and Lake Zurich (oligochaetes: -32.8±0.9 %, N= 5; larvae: -32.5±2.1 %, N= 24). There was no apparent trend between δ13C values of macrofauna and sediment depth (Fig. S3 in the Supplement).

Average δ13C-methane values are in all cases ∼35 % to 50 % more negative than those of macrofauna. The most negative methane values are present in Lake Lucerne (-78.8±4.3 %, N= 18) and Lake Zurich (-76.7±2.4 %, N= 25), followed by Lake Baldegg (-74.3±2.6 %, N= 20), Lake Greifen (-73.6±3.7 %, N= 21), and Lake Zug (-70.1±4.5 %, N= 23). All stations except the middle station in Lake Baldegg have 13C-methane increases indicative of methane oxidation in surface layers (Fig. S3 in the Supplement).

The δ13C values of TOC are much closer to those of macrofauna (Fig. 5; Fig. S3 in the Supplement), with averages ranging from equal (Lake Zurich) to ∼5 % higher (Lake Baldegg). The lowest average δ13C-TOC was measured in Lake Greifen (-34.5±1.5 %, N= 35), followed by Lake Baldegg (-32.4±1.2 %, N= 37), Lake Zurich (-32.2±1.9 %, N= 29), Lake Zug (-30.8±1.3 %, N= 35), and Lake Lucerne (-29.7±1.2 %, N= 32). Isotopic values of TOC increase by 4 %–6 % with sediment depth at all sites (Fig. S3 in the Supplement). Despite the small differences between δ13C-TOC and δ13C-macrofauna, δ13C-TOC values are significantly higher than those of oligochaetes and larvae in all lakes except Lake Zurich (Fig. 5). Average δ13C-DOC is slightly higher than δ13C-TOC in all lakes and significantly higher than the δ13C of macrofaunal biomass (Fig. 5). Additional analyses on water column algal material and algal bloom layers in sediment (Fig. S3 and Table S4 in the Supplement) suggest δ13C values similar to those of TOC.

A two end-member mixing model suggests that on average ≥88 % of macrofaunal biomass-C can be explained with assimilation of detrital organic C (TOC) (Table 2). By contrast, methane-derived carbon accounts for ≤12.1 % or ≤6.3 % of biomass-C depending on the assumed isotopic fractionation factor during aerobic methane oxidation (for further details, see Table 1 caption). Chironomid larvae and oligochaetes from the same lakes have highly similar average methane-derived carbon contributions to biomass. Consistent with past studies (Hershey et al., 2006; Jones and Grey, 2011), the contribution of methane-derived carbon to macrofaunal biomass increases with trophic state, with the lowest contributions in Lake Zurich and Lake Lucerne and highest contributions in Lake Baldegg, followed by Lake Zug and Lake Greifen.

To investigate the nature of macrofauna–microbiota associations, e.g., with respect to microbial gardening or grazing of methane-cycling microorganisms or symbiotic relationships, we studied 16S rRNA gene sequences of macrofauna (whole organisms, guts, residual body without guts) and chironomid larval tubes and compared these to those in surrounding sediments (Fig. 6).

To further investigate potential interactions between macrofauna and microorganisms in general, and methane-cycling microorganisms in particular, we compared the contributions of Bacteria, methane-cycling archaea, and methane-oxidizing bacteria across sample types. Trends related to lake trophic state and gardening of or preferential grazing on methane-cycling microorganisms are largely absent, but we observe other trends.

Bacteria account for >80 % of total 16S gene copies in all samples (Fig. 8, left panel). Significantly higher proportions are present in oligochaetes, larvae, and tubes relative to sediments (Table 3). The contribution of Bacteria decreases from 94 %–98 % in surface sediments to 82 %–86 % below 12 cmblf. By comparison, Bacteria contribute ≥99 % in most macrofauna samples. The lowest bacterial contributions are ∼98 % in chironomid larvae, 90 % in oligochaetes, and 96 % in tubes.

In the vast majority of samples, mcrA gene copy numbers are ≥100 times lower than total 16S rRNA gene copy numbers (range: below the detection limit of ∼0.0001 % to 2 %) (Fig. 8, middle panel), suggesting very low contributions of methanogenic and/or anaerobic methanotrophic archaea. The mcrA contributions are significantly higher in sediments compared to oligochaetes, larvae, and tubes (Table 2) and are even below qPCR detection in all but one larval specimen. While the contribution of mcrA increases with depth in larval tubes, oligochaetes and sediments show no depth-related trends. The 16S rRNA genes of methane-cycling Archaea were found in sediments (mainly Methanobacteria and M. fastidiosa) and at very low read numbers in a few tubes (M. fastidiosa) and oligochaetes (M. fastidiosa, M. peredens) but not in larvae.

The pmoA contributions range from below detection (≤∼0.001 %) to ∼15 % (Fig. 8, right panel) and are – compared to sediments – significantly elevated in oligochaetes but not in larval specimens or larval tubes (Table 3). This suggests the potential for preferential grazing by, or elevated populations of symbiotic aerobic methanotrophic bacteria within, oligochaetes. Nonetheless, it is worth mentioning that the median calculated pmoA percentage in oligochaetes was only ∼1 % and that based on the maximum calculated value of 15 % methane-oxidizing bacteria in no case dominated oligochaete bacterial communities. As for mcrA, pmoA was only detected in very few (2) larval samples. While pmoA contributions decrease with depth in sediments, there is no clear depth trend in oligochaete or larval tube samples. The 16S rRNA gene sequences indicate that all methane-oxidizing bacteria are γ-Proteobacteria, dominated by Crenothrix (Methylococcales). Crenothrix are moreover the only methane-oxidizing bacterium detected in oligochaetes, whereas low read percentages of Methylococcaceae (Methylobacter, Methylocaldum, Methylococcus, and Methyloparacoccus) were detected in larvae, larval tubes, and sediments. In addition, the denitrifying methanotroph Methylomirabilis (candidate phylum NC10) was detected in low read numbers in several tube and sediment samples (mostly from Lake Lucerne). Despite the significantly higher calculated abundance of methane-oxidizing bacteria in oligochaetes based on ratios of pmoA to total 16S rRNA gene copy numbers, we did not detect significantly different 16S read percentages between larvae, tubes, oligochaetes, or sediments (data not shown).

Oligochaete abundances follow the environmental index proposed previously by Milbrink (1983) which predicts a strong rise in worm abundance with increasing trophic state. Chironomid abundances are also within the range previously reported for lakes (Mousavi, 2002). While chironomid larvae show typical depth distributions (e.g., Panis et al., 1996), oligochaetes have unusually deep ranges, with high abundances to 10–14 cm in eutrophic lakes. By contrast, most publications report that oligochaetes are mainly present at 2–8 cm sediment depth (reviewed in McCall and Tevesz, 1982).

The observed shift in dominance from chironomid larvae to tubificids with increasing trophic state (Fig. 2, Fig. 3) matches past studies reporting the dominance of oligochaetes in eutrophic lakes (Saether, 1980; Lang, 1985; Timm, 1996; Bürgi and Stadelmann, 2002) and changes from chironomid-larva- to oligochaete-dominated communities as the first signs of eutrophication (Saether, 1979). This dominance of oligochaetes in eutrophic lakes is possibly related to an overall higher tolerance of low O2 conditions as many oligochaetes feed in anoxic parts of sediments (McCall and Tevesz, 1982) and efficiently exchange gases through their body walls (Martin et al., 2008). Longer survivorship of anoxic conditions among oligochaetes is also possible (Hamburger et al., 1998), though anaerobic respiration and tolerance of extended anoxic periods is also known for certain species of chironomid larvae (Pinder, 1995). Additional reasons could be the superior ability of oligochaetes to exploit high organic matter supplies or that deeper burrows of oligochaetes provide better protection from benthic predators, such as bottom-feeding fish, which are abundant in eutrophic lakes (Scheffer et al., 1993).

While most oligochaete specimens could only be classified to the family level – Tubificidae (+bristles); Tubificidae (- bristles); Fig. 4; Table S7 in the Supplement – distributions of those that were taxonomically classifiable to the species level match published distributions. On one hand, subsurface deposit feeders known to rely on bacteria and algae as food sources dominated eutrophic lakes. L. hoffmeisteri, an indicator species of eu- to hypertrophic lakes (Brinkhurst, 1982), occurs in high abundances in Lake Baldegg (Table S7 in the Supplement). P. hammoniensis and T. tubifex, which frequently co-occur in high abundances in mesotrophic to eutrophic lakes (Lang, 1990; Timm, 1996), dominate Lake Zurich, Lake Zug, and Lake Greifen. On the other hand, surface-deposit-feeding E. velutinus, which indicates oligo- to mesotrophic conditions (Martin et al., 2008), was only found in Lake Zurich.

Even though many tubificids are subsurface conveyor feeders, the lakes investigated show little evidence of sediment mixing. We observed clear laminations at the deep station in Lake Baldegg and the deep and middle station in Lake Greifen in sediments that were being deposited until the mid-1980s and ∼2010, respectively (Fig. 3, Fig. S7 in the Supplement; Fiskal et al., 2019), so until the onset of artificial water column mixing and oxygenation in these lakes (Lake Baldegg in 1984 and Lake Greifen in 2009; Fiskal et al., 2019). While the subsequent disappearance of laminae suggests rapid re-colonization by macrofauna, it appears that mixing has remained limited to surface sediments even though burrows of tubificids extend far into laminated layers. Depth profiles of radionuclides confirm this interpretation and even indicate minimal sediment mixing in the presence of macrofauna (Fig. S7 in the Supplement). Independent of faunal presence, 137Cs peaks that match the 1986 (Chernobyl) and 1963 (bomb test) time markers, and clear 210Pbunsupported decreases from the top 2 cm downward, are present at all stations. These findings contrast with the rapid sediment homogenization to 10 cm by tubificids in the laboratory (Fisher et al., 1980; Matisoff et al., 1999) and homogeneous radionuclide profiles to 6 cm in tubificid-dominated natural lake sediments (Robbins et al., 1977; Krezoski et al., 1978). Similar to tubificids, chironomid larval communities change in relation to trophic state (Fig. 4; Table S7 in the Supplement). Large free-living and predatory larvae account for half of the specimens in Lake Lucerne, whereas tube-building herbivorous, surface detritus-feeding, and gardening larvae dominate Lake Zurich and the small sample sizes in eutrophic lakes. The shift in diet at higher trophic levels matches the higher input of algae and algal detritus (Fiskal et al., 2019), whereas the potential increase in microbial gardening matches observed increases in gardening by C. riparius and other Chironomus spp. under hypoxic or eutrophic conditions (Stief et al., 2005; Yasuno et al., 2013). By contrast, the reasons for the high abundances of predatory larvae in Lake Lucerne are unclear. Possible reasons are the low hypoxia tolerance of large predatory Macropelopia and Procladius spp. (Hamburger et al., 1998; Brodersen et al., 2008), higher availability of zooplankton food in oligotrophic lakes (Jeppesen et al., 1990; Jeppesen et al., 1999), and/or stronger predation pressure in mesotrophic and eutrophic lakes, which often have high populations of bottom-feeding fish (Scheffer et al., 1993).

Similar to previous studies (e.g., Grey et al., 2004; Jones et al., 2008) we calculate an increase in the contribution of methane-derived carbon with increasing trophic state (Fig. 5; Table 2). Yet, this contribution is at most 12 %, even in the highly eutrophic lakes. Other studies have estimated methane-derived carbon contributions of >40 % for chironomid larvae in eutrophic lakes (e.g., Deines and Grey, 2006; Eller et al., 2007; Jones et al., 2008) and reported strong δ13C depletions in oligochaete specimens from profundal sediment (Premke et al., 2010). Yet, minor contributions of methane-derived carbon to the biomass of benthic invertebrates are not new. A survey of 87 lakes suggested that marked 13C depletions were only present in chironomid larvae from lakes with seasonal stratification and bottom water anoxia (Jones et al., 2008). Moreover, the limited published δ13C data on lake oligochaetes are mostly similar to those of TOC (Kiyashko et al., 2001; Premke et al., 2010).

In support of C-isotopic interpretations, DNA-based analyses indicate that neither methane-oxidizing bacteria nor methanogens are dominant microorganisms in surface sediments or chironomid larval tubes. Thus, strong enrichment or gardening of methane-oxidizing bacteria or methanogens as observed elsewhere in chironomid tubes (e.g., Kajan and Frenzel, 1999; Kelly et al., 2004) or surface sediments (e.g., Eller et al., 2005; Deines et al., 2007a) is absent for reasons that are unclear. Despite being artificially oxygenated, bottom water in Lake Baldegg and Lake Greifen experiences seasonally low O2 conditions (0.5–4 mgL-1) or hypoxic conditions (<0.5 mgL-1), respectively (Fiskal et al., 2019). These values are within or below the seasonal O2 threshold (2–4 mgL-1) that is characteristic of lakes with marked 13C depletions in chironomid biomass (Jones et al., 2008). Jones et al. (2008) argued that the contribution of methane-derived carbon increases inversely with the depth of the oxic–anoxic interface. In June 2016, this interface was ≤1 mm at all stations in Lake Baldegg and ≤2 mm at those in Lake Greifen, while methanogenesis occurred in the top 1 cm of sediment (Fiskal et al., 2019). Thus, conditions were potentially well-suited for the strong enrichment of methane-cycling microorganisms. It is possible that the growth of methane-oxidizing bacteria is mainly promoted at narrow oxic–anoxic (high O2-methane) interfaces produced by ventilating and tube-building chironomid larvae (Brune et al., 2000). Tubificids, which dominated our eutrophic lakes, do not produce such stable oxic–anoxic interfaces and also perform less burrow ventilation than chironomid larvae (Gautreau et al., 2020 and references within). Yet, the fact that all three identified larvae from Lake Baldegg belong to tube-building taxa and that the four isotopically analyzed larvae from this lake only had minor methane-derived carbon contributions suggests that yet unknown factors contribute to the enrichment of methane-oxidizing bacteria by tube-building chironomids in surface sediment.

Instead of methane-derived carbon, our C-isotopic data indicate that algal or detrital organic carbon, or microorganisms that have assimilated the isotopic signatures of algal or detrital organic carbon, is the main food source of dominant macrofauna (Table 2). Rather than methane-derived carbon, selective feeding on isotopically depleted subportions of the TOC pool could even, in principle, explain the minor isotopic depletions of oligochaete and chironomid larval biomass in eutrophic lakes. Yet, our limited data on algal bloom layers in sediments and phytoplankton from overlying water indicate similar 13C values relative to TOC (Fig. S5 in the Supplement). Preferential feeding on organic C from surface sediments, which in many cases has the lowest C-isotopic values, or isotopic fractionations during C-assimilation and biosynthesis are also not plausible. As bottom-up conveyor feeders, tubificids feed mostly at several centimeters depth (McCall and Tevesz, 1982), and C-isotopic fractionation during biosynthesis of bulk animal biomass is typically low (Fry and Sherr, 1989).

Minimal sediment reworking and deep sedimentary distributions of tubificids suggest that shallow subsurface deposit feeding may not be the main dietary mode of these worms in the lakes studied, raising questions concerning their main foraging strategy. One possibility is that oligochaetes selectively graze on microbial biofilms inhabiting the walls of their deep and extensive gallery-type burrow networks. Under this scenario one might expect large amounts of DNA of sediment microorganisms in oligochaete intestines. This is not the case, however, suggesting that grazed communities are very different from those in sediments or their DNA is rapidly digested. Another foraging strategy may not involve ingestion via the oral cavity but diffusive uptake. T. tubifex can actively take up short-chain organic acids, such as acetate and propionate, through their body wall (Hipp et al., 1985; Sedlmeier and Hoffmann, 1989). The subsequent respiration of these organic acids can account for up to 40 % of T. tubifex energy turnover (Hipp et al., 1986). Other species of tubificids take up amino acids through the body wall (Brinkhurst and Chua, 1969). Tubificid body walls are also permeable to dissolved gases, which is why tubificids acquire O2 by undulating movements of their tail ends in oxic water above sediments (Brinkhurst, 1996). Permeability to gases could also provide energy if, for example, methane or H2 diffusing from pore water into worms supports symbiotic microorganisms.

Matching the slight increase in methane-derived carbon in oligochaetes from eutrophic lakes, we observe higher contributions of pmoA in oligochaetes compared to surrounding sediment. Assuming that these pmoA belong to living methane-oxidizing bacteria, movement of oligochaetes between methane-rich, deeper layers and the oxic sediment surface could favor their growth and result in an endosymbiotic relationship. The potential for annelid hindguts to make excellent microbial habitats was previously demonstrated in the polychaete Abarenicola vagabunda (Plante et al., 1989). How this methane-derived carbon would be assimilated is unclear, however. Potential mechanisms include uptake of organic intermediates of methane oxidation, e.g., methanol, through the hindgut or ingestion of faeces that are enriched in methane-oxidizing bacteria.

The overwhelming majority of microbial DNA from oligochaetes, however, belongs to Bacteria, often single ZOTUs, which are not linked to methane oxidation. In 22 of the 30 specimens sequenced, a single ZOTU accounted for >50 % of the total reads (Table S8 in the Supplement). In 15 specimens, this dominant ZOTU belonged to a single genus-level cluster of unclassified Fusobacteriaceae (Fusobacteriaceae Cluster I) that was previously found in earthworm and aquatic vertebrate intestines, anaerobic sediments, bioreactors, soil, and diverse water samples (Fig. S7A in the Supplement). The high percentages of this cluster are striking considering that Fusobacteriaceae account for on average only 0.01±0.02 % of total 16S reads in the surrounding sediments. All cultivated members of Fusobacteriaceae are anaerobes that fermentatively degrade polymeric organic compounds, in particular proteins and carbohydrates, with acetate, butyrate, and other short-chain organic acids as main end products (Olsen, 2014). Given previous evidence for the preference of proteinaceous organic matter by tubificids (de Valk et al., 2017), these Fusobacteriaceae could be primary degraders of proteins within the digestive tracts of oligochaetes. This relationship could be mutually beneficial, commensal, or parasitic. A mutually beneficial relationship could entail symbionts gaining energy by fermenting proteins that are not digestible by the host and through the host respiring the resulting fermentation products.

The remaining seven dominant ZOTUs belong to the phyla Proteobacteria (α, β, and ε classes), Bacteroidetes, and Parcubacteria (Fig. S8, Table S8 in the Supplement). ZOTU18 falls into the anaerobic ε-proteobacterial genus Wolinella (order Campylobacterales), isolates of which use H2 or formate as electron donors and fumarate and nitrate as electron acceptors (Tanner and Paster, 1992). Succinate is the main end product of fumarate reduction by Wolinella and could benefit hosts under low O2 conditions given that succinate is the main intermediate during anaerobic metabolism of tubificids (Seuß et al., 1983). ZOTU8 falls into the facultatively aerobic β-proteobacterial genus Deefgea (order Neisseriales), members of which ferment carbohydrates to organic acids (Stackebrandt et al., 2007) and could benefit hosts as proposed for Fusobacteriaceae. The α-proteobacterial ZOTU4 falls into the family Holosporaceae, members of which are obligately intracellular, potentially parasitic symbionts of ciliates (Santos and Massard, 2014). ZOTU4 could derive from commensal ciliates, which often inhabit guts of freshwater oligochaetes (Falls, 1972). Alternatively, given the high percentage of 16S reads in one specimen (93 %), a novel form of (intracellular) symbiosis with tubificids cannot be discounted. Similarly unclear is the host relationship with ZOTU199, which belongs to the candidate phylum Parcubacteria of the Candidate Phyla Radiation (Brown et al., 2015). Members of this phylum have been retrieved from diverse, mostly anoxic habitats, have genes linked to carbohydrate fermentation, and have been implicated in ectosymbiotic or parasitic lifestyles (Wrighton et al., 2012; Nelson and Stegen, 2015). The remaining ZOTUs fall into an unclassified genus-level subcluster of β-Proteobacteria (ZOTU6; order Rhodocyclales) and an unclassified order-level cluster of ε-Proteobacteria (ZOTU9). Based on existing knowledge, it is not possible to infer the potential roles of these ZOTUs within their hosts.

Similar to tubificids, most chironomid larvae (12 of 19 sequenced specimens) are dominated by single ZOTUs (Table S8 in the Supplement). Interestingly, more specimens are dominated by single ZOTUs in Lake Lucerne (9 of 10) than in Lake Zurich (three of seven) or Lake Baldegg (zero of two), suggesting that the frequency and/or importance of these associations is linked to trophic state. These single dominant ZOTUs are mostly Proteobacteria (10 of 12; α, β, and γ classes). In addition, single specimens were dominated by the same unclassified Fusobacteriaceae (Fusobacteriaceae Cluster I) that dominate tubificids and an unclassified sister group of Bacteroides, Bacteroidetes), which we call “Unclassified Wastewater and Gut Group” based on reported occurrences.

Two proteobacterial groups most commonly dominate chironomid larvae. ZOTU2 of the α-proteobacterial genus Wolbachia (Rickettsiales) dominates four specimens. Members of this genus are widespread intracellular symbionts of insects whose relationships with their hosts range from parasitic to mutualistic (Correa and Ballard, 2016), though, to our knowledge, dietary contributions have not been demonstrated. ZOTU3 and ZOTU21 of the γ-proteobacterial genus Aeromonas dominate three specimens. Members of this facultatively anaerobic genus are widespread in aquatic habitats (Huys, 2014) and were previously found in aquatic invertebrates, including chironomid larvae (Eller et al., 2007). Aeromonads can ferment carbohydrates to organic acids, which might supplement the diet of chironomid larvae, but they have also been shown to degrade the egg masses of chironomids (Senderovich et al., 2008). Similar functions, ranging from mutualistic to detrimental, are likely for the γ-proteobacterial genus Serratia (γ-Proteobacteria) and an unclassified cluster of Moraxellaceae (Pseudomonadales) and for the Unclassified Wastewater and Gut Group. ZOTUs of these groups each dominate one larval specimen (ZOTU11, ZOTU26, and ZOTU28, respectively). All three groups degrade carbohydrates anaerobically (Serratia, Bacteroidetes) or aerobically/facultatively anaerobically (Moraxellaceae) to organic acids, which may provide energy to larvae but can also be pathogenic or mutualistic in ways unrelated to diet (Grimont and Grimont, 2006; Sabri et al., 2011; Teixeira and Merquior, 2014; Wexler, 2014). The remaining ZOTUs belong to unclassified genus-level subclusters of β-proteobacterial Rhodocyclales (ZOTU6) and Burkholderiales (ZOTU12). Due to the very diverse ecophysiologies of Rhodocyclales and Burkholderiales the potential roles of these ZOTUs within their hosts are highly uncertain.