Approval of the present study by a review board institution or ethics committee was not necessary because all fish were caught under permits issued by the local fisheries administration (Regierungspräsidium Tübingen), by qualified (license-holding) personnel subject to regular checks by the local fisheries administration (Regierungspräsidium Tübingen). All fish were caught according to the German Animal Protection Law (Tierschutzgesetz § 4) and the ordinance on slaughter and killing of animals (Tierschutzschlachtverordnung § 13).

Lake Constance is part of the Rhine drainage basin and is bordered by Austria, Germany and Switzerland (47°38'N, 9°22'E). The total surface area of 536 km² is divided between the large (472 km²), deep (>250 m) Upper Lake (ULC) and the smaller (63 km²), shallower Lower Lake (LLC). Due to missing data and lack of knowledge about the stickleback situation in LLC and different type of lake (warm, mesotrophic), this basin was excluded in the present study. Thus, the current study only focusses on warm monomictic, large oligotrophic pre-alpine basin of ULC. The fish community of ULC comprises a minimum of 30 species (Eckmann and Rösch 1998) of which about 10 are targeted by professional fishermen (Rösch 2014). Of these, whitefish (Coregonus spp.) are the most economically important, and fisheries management is based on routine monitoring of this important group (www.ibkf.org). An overview of the fisheries situation is given by Baer et al. (2017).

Stickleback sampling of ULC was conducted monthly, from March 2017 until November 2018, using littoral and pelagic gillnets with mesh sizes of 10–12 mm. All nets had a height of 3 m, while length varied with mesh size: 30 m for nets with 10 mm mesh and 15 m for the 12-mm mesh net. All pelagic nets were deployed to drift freely behind the nets used in the monthly monitoring of whitefish (mesh sizes 36–44 mm), at depths of 3–15 m according to the areas of greatest stickleback abundance recorded during hydroacoustic surveys (Gugele et al. 2020). Benthic nets were set at depths from 6 to 20 m. All nets were set overnight, with a soak time of about 15 h. The overall catch in the pelagic gillnets (number per unit effort, NPUE, as n/m² per net) was low from January to September (NPUE 0.03–0.25) and peaked between November and December (NPUE 1.1–7.7), the spawning season of whitefish (Baer et al. 2022b). Catches in the benthic gillnets were highest during the stickleback spawning season between May and July (NPUE 11.8–59.0) and a second peak was again observed during November and December (NPUE 6.0–79.0; Baer et al. 2022b).

10 to 34 samples of stickleback white muscle and liver tissue were taken from each monthly catch. Catches of fewer than 10 individuals (recorded in August and September each year, plus April 2017, July 2017 and October 2018) were excluded from analysis. C and N stable isotope analysis was run on 275 sticklebacks. Of these, 193 were caught in the littoral zone and 82 in the pelagic zone. All fish were euthanised with an overdose of clove oil (1 mL L⁻¹) and a gill cut. They were measured post-mortem (total length (TL) to the nearest mm), weighed to the nearest 0.01 mg and sex was recorded. Some sticklebacks were infested with the pseudophyllidean cestode Schistocephalus solidus, and because it is known that the health status of a fish can have direct effects on the stable isotope values (Karlson et al. 2018), a combined parasite-to-host biomass ratio (parasite index, PI) was calculated as an indirect measure of the severity of infestation (Baer et al. 2022b). All parasites were counted per host, blotted, and weighed to the nearest 0.01 mg and PI was determined using the formula

PI = P/H (1)

where P is the total weight of the parasites and H is the mass of the host without the parasite.

Due to internal procedures, gastrointestinal tracts (stomach and intestine) were analysed from a subsample of 109 sticklebacks; 69 caught in the pelagic zone and 40 caught in the littoral zone (TL 68 mm ± 6 mm standard deviation SD). Samples were taken during all four seasons (autumn 2017, winter 2017, spring 2018, and summer 2018), and for each season and each habitat, the gut contents of at least 10 individuals were analysed with the exception of some sampling dates (20 during winter and summer in both pelagic and littoral zone, and 19 during spring in the pelagic zone). Food items were identified and counted in a zooplankton counting chamber and categorised into five groups, namely copepods (nauplii, copepodites and copepods of Cyclopoida, Calanoida and Harpacticoida); Bosmina (all members of the genus Bosmina); other herbivorous/detritivorous cladocerans (Daphnia spp., Diaphanosoma brachyurum and Chydoridae); predatory cladocerans (Bythotrephes longimanus and Leptodora kindtii); fish (eggs and larvae) and other (Chironomidae, Annelida, Bivalvia, Collembola, Ceratopogonidae, Ephemeroptera, adult Heteroptera, Hydrachnidia, adult Mysidae, Nematoda, Ostracoda, Plecoptera, Simuliidae and Trichoptera). Diet quantification of sticklebacks followed the use of the numerical method and diet was calculated as a percentage of the total number of prey items eaten per stickleback (Amundsen and Sánchez-Hernández 2019). Furthermore, we calculated for each season and habitat the mean number (± standard deviation SD) of consumed food items per category.

Although stable isotope analyses on fish C and N are generally performed using muscle tissue alone, the more rapid turnover of liver tissue means isotope signatures there reflect more recent feeding (Boecklen et al. 2011). Repeated at suitable intervals, in this case monthly, these tissue-based differences in signal lag can help resolve the timing of hard-to-predict peaks in seasonal prey availability, such as that of whitefish eggs and larvae.

Tissue samples from sticklebacks caught in 2017–18 were prepared for analysis by drying them in an oven at around 60 °C for 48 hr and grinding them into a fine powder. Lipid extraction was performed on the samples because some studies have shown that in tissues with C:N ratios greater than 3.5, such treatment reduces bias in δ¹³C values (Skinner et al. 2016). Therefore, lipid extraction of samples was conducted by adding 200 µL of 2:1 Chloroform:Methanol mixture to the powdered tissue. Afterwards, samples were vortexed and centrifuged for two minutes at 4000 rpm. The excess sample was discarded, and centrifugation was repeated 2–4 times until the sample colour changed from yellow to colourless. Samples were then washed in 200 µL Milli-Q water, followed by further vortexing and centrifugation for 2 min (4000 rpm). Again, the excess sample was discarded, and washing was repeated multiple times during the lipid extraction. Next, samples were dried in a fume hood for 48 hr, then ground again to a fine powder and weighed (ca. 0.7 mg) to the nearest 0.001 mg in tin capsules, using a microbalance (Chyo Balance Corporation, Kyoto, Japan).

To measure isotopic ratios samples were combusted in a vario MICRO cube elemental analyser (Elementar Analysensysteme GmbH, Hanau, Germany). The emerging gases were separated via gas chromatography and passed into a Micromass Isoprime isotope mass spectrometer (Isoprime Ltd., Cheadle Hulme, UK) for determination of the 13C/12C and 15N/14N ratios (R). Measurements are reported in δ-notation (δ¹³C, δ¹⁵N) in parts per thousand deviations (‰), where δ = 1000 × (R_sample/R_standard – 1) relative to the Pee Dee Belemnite (PDB) for carbon and atmospheric N₂ for nitrogen. Two sulphanilamide (Isoprime internal standards) and two casein samples were used as laboratory standards for every 10 unknowns in the sequence. Replicate assays of internal laboratory standards indicated measurement errors (SD) of ± 0.05% and 0.15% for δ¹³C and δ¹⁵N, respectively.

To compare the isotopic values of sticklebacks with those of other species of fish in ULC, 128 additional sampled of muscle tissue were analysed from bleak (Alburnus alburnus L., 1758; n = 18; mean TL 60 mm ± 6 mm SD) roach (Rutilus rutilus L., 1758; n = 34; mean TL 263 mm ± 71 mm SD); rudd (Scardinius erythrophthalmus L., 1758; n = 12; mean TL 217 mm ± 40 mm SD); tench (Tinca tinca L., 1758; n = 19; mean TL 226 mm ± 147 mm SD); pelagic whitefish (n = 21; mean TL 308 mm ± 55 mm SD); burbot (Lota lota L., 1758); n = 19; mean TL 374 mm ± 38 mm SD); and pike (Esox lucius L., 1758; n = 5; mean TL 289 mm ± 44 mm SD). All were sampled with gill nets during August and September 2020. All fish were euthanised with an overdose of clove oil (1 mL L⁻¹) and a gill cut. Using data from stomach content analyses carried out in prior studies, all fish species were divided into different feeding guilds: whitefish were categorised as zooplanktivorous; burbot as partly piscivorous (Hansen et al. 2022); pike as piscivorous; roach, bleak, and tench as benthivorous/insectivorous; and rudd as herbivorous (Baer et al. 2022a). Due to the stomach content analysis from this study, sticklebacks were categorised as omnivorous. Samples of white muscle were excised and frozen (-20 °C) until further processing.

For calculating the trophic position of sticklebacks, pike and burbot, faucet snails (Bithynia tentaculata, n = 10) were collected in August 2020 from the littoral habitat and used for the estimation of the littoral baseline (=δ¹⁵N_{lit. base}). Quagga mussels (Dreissena rostriformis bugensis, n = 200) were collected from free-standing piles in the pelagic zone 0.5–2 m depth in the upper mixed layer of Lake Constance and used for the estimation of the pelagic baseline (=δ¹⁵N_{pel. base}).

To gain more insight into the isotopic signatures of potential stickleback prey during winter and spring, five samples of zooplankton (wet weight (g): mean = 2.34, SD = 2.53) were netted with 300 μm mesh in the epilimnion of ULC, first in October and December 2021, then in February, March, and early May 2022. An abundance of pollen in the lake epilimnion during April 2022 prevented an uncontaminated sample being taken during that month. In addition, in December 2021, 36 females of C. wartmanni (pelagic whitefish) and 42 C. macrophthalmus (benthic whitefish) were caught during spawning at their spawning grounds in ULC as part of routine sampling conducted by the Fisheries Research Station of Baden-Württemberg. To get the isotopic signature of whitefish eggs and larvae, a small sample of eggs was taken from each individual and larvae hatched from the eggs of pelagic whitefish (kept at a hatchery facility in Langenargen, Baden-Württemberg) were also sampled. After hatching, larvae were held in rearing vats until the yolk sac was partly absorbed and larvae had begun to exhibit normal swimming behaviour. From these non-fed, free-swimming larvae, four subsamples of multiple individuals (n: mean = 172, SD = 100) were taken and euthanised with an overdose of carbonated water. Clove oil was avoided in this instance as it may have biased isotopic readings, and unlike larger fish, the delicate larvae cannot be easily washed without damage. To remove potential biases due to the length of time between the main stickleback sampling (2017–2018) and the sampling of the zooplankton, whitefish eggs and larvae (2021–2022) (Fig. 1), a further 50 sticklebacks were caught for additional stable isotope analysis in January, February and April 2022 (25 each in the pelagic and littoral zone, in total 150) using gill nets.

Figure 1.

Timeline of sampling in the present study.

Samples of sticklebacks caught in 2022 and other fish were prepared for stable isotope analysis via freeze drying at -50 °C under pressurisation (<1 mbar), and ground to homogenous powder using a mixer mill. Whitefish egg samples were dried at 60 °C in a drying oven, before being stored in a glass desiccator filled with silica desiccator beads; the desiccator was stored in a cool, dark environment. Samples of plankton and pelagic whitefish larvae were dried overnight in a drying oven (60 °C), then stored in freezers at -20 °C. Each individual dried sample of plankton, whitefish larvae and whitefish eggs were then separately homogenised using a BeadRupture Homogenizer (Omni International, Kennesaw, Georgia, United States) by dispensing the sample into a plastic microtube, with a number of sterilised metal beads (<0.5 mL), and processing the sample into a fine, homogenous powder. The times and speeds used in the homogenisation process were adapted according to the individual condition of the samples. After homogenisation, samples were stored in freezers at -20 °C. Sample powder (0.3–0.4 mg) was weighed into tin capsules and combusted in an isotope ratio mass spectrometer (Delta plus, Finnigan MAT, Mas-Com GmbH, Bremen, Germany), interfaced (viaConFlo II, Finnigan MAT, MasCom GmbH, Bremen, Germany) with an elemental analyser (EA 1108, CarloErba, Thermo Fisher SCIENTIFIC, Milan, Italy). Because the mean C:N values (±SD) of all fish samples were below 3.5, lipid extraction of fish muscle tissue was not conducted (Matthews et al. 2010; Skinner et al. 2016). Measurements are reported in δ -notation (δ¹³C, δ¹⁵N) in parts per thousand deviations (‰), where δ = 1000 × (Rsample/Rstandard – 1) relative to the Pee Dee Belemnite (PDB) standard for carbon and atmospheric N₂ for nitrogen. Finely ground animal horn (keratin) was used as a laboratory standard for every 10 unknowns in sequence. Replicate assays of internal laboratory standards indicated measurement errors (SD) of ± 0.05% and 0.15% for δ¹³C and δ¹⁵N, respectively.

The trophic position of sticklebacks was calculated according to the protocol established by Post (2002) for a two-source food web:

Tophic position = λ_Base + (δ¹⁵N_stickleback– [δ¹⁵N_{lit. base}* α + δ¹⁵N_{pel. base} * (1 – α)])/Δ_n (2)

$\alpha =\frac{{\delta }^{13}{\text{C}}_{\text{stickleback}}-{\delta }^{13}{\text{C}}_{\text{pel.base}}}{{\delta }^{13}{\text{C}}_{\text{lit.base}}-{\delta }^{13}{\text{C}}_{\text{pel.base}}}$ (3)

where λ_Base denotes the trophic position of the consumer (λ_Base = 2) used for the estimation of the littoral (=δ¹⁵N_{lit. base}) and pelagic (=δ¹⁵N_{pel. base}) baseline. The isotope values of faucet snails and quagga mussels were used for δ¹⁵N_{pel. base} and δ¹³C_{pel. base}. As filter feeders, quagga mussels are an ideal integrator species for representing the consumer base of the pelagic food web, and are favoured over bulk seston or plankton samples, which may include non-consumer material and undifferentiated detritus and thus bias stable isotope ratio signatures. The isotopic ratios of quagga mussels and faucet snails were assessed using the same method applied for plankton, whitefish eggs and larvae (see above). δ¹³C_stickleback is the measured δ¹³C value of sticklebacks muscle. δ¹⁵N_stickleback is the measured δ¹⁵N value of sticklebacks muscle, Δ_n is the enrichment in δ¹⁵N per trophic level (Δ_n = 3.4 (Minagawa and Wada 1984; Vander Zanden and Rasmussen 1999; Post 2002; Reimchen et al. 2008)) and α denotes the proportion of sticklebacks carbon derived ultimately from littoral sources. A Δ_n value of 3.4 is commonly used for sticklebacks (Post 2002; Matthews et al. 2010), but in the light of recent reviews suggesting a lower level of δ¹⁵N enrichment for carnivorous fish (Vanderklift and Ponsard 2003; Boecklen et al. 2011; Blanke et al. 2017; Kambikambi et al. 2019), a further analysis as included using Δ_n = 2.0 as a value for the trophic enrichment of δ¹⁵N in sticklebacks.

To compare the trophic position of sticklebacks with piscivorous fish species, the trophic position for pike and burbot was then calculated for each pike and burbot using the formula: Trophic position = [(piscivorous fish δ¹⁵N – δ¹⁵N_{lit. base})/ Δ_n] + λ_Base (Nyqvist et al. 2018).

To test the effects of covariates on the δ¹⁵N or δ¹³C of muscle or liver tissue and the trophic position of sticklebacks, the following general linear model (GLM) (Sachs 1997) was used:

Y_ijklmno = μ + α_i + β_j + (αβ)_ij + γ_k + δ_l + ε_m + ζ_n + η_o + θ_ijklmno (4)

where Y_ijklmno is δ¹⁵N or δ¹³C in muscle or liver tissue or the trophic position of sticklebacks; µ is the overall mean, α_i denotes month, β_j is total length, (αβ)_ij is the interaction between month and total length, γ_k represents year and was added to the model as a random factor, δ_l is habitat (pelagic or littoral zone), ε_m is sex (male or female), ζ_n denotes the infection state (yes/no), η_o is parasite index and θ_ijklmno is the random residual error. Model requirements, i.e. residuals not violating linearity, normality or non-independence were checked by inspecting residuals (predicted vs. expected plots) and multicollinearity by inspection correlation of independent variables. Single outliers with extreme values were excluded from the dataset (selection criteria: more than eight times standard deviation). Student’s t-test was used for post hoc comparisons between habitat and sex after testing for homoscedasticity (Levene test) and by building contrasts (Sokal and Rohlf 2003). The GLM for trophic position of sticklebacks was also run with Δ_n set at 2.0.

Differences in mean δ¹⁵N and δ¹³C values among Lake Constance fish species were examined using Tukey-Kramer HSD-tests.

The contribution rate of potential food sources for sticklebacks’ diet was estimated using the Bayesian mixing model in the SIMMR package (Parnell and Inger 2019), which was based on the SIAR package (Parnell et al. 2010), and implemented in the R 4.04 software (R Core Team 2020). We run separated mixing models for isotopic signatures of the liver samples of sticklebacks collected in winter (December to March) and in the summer (June and July). We choose to use only liver values because isotope values of the liver react faster than those of the muscles (time lag of only one month, Perga and Gerdeaux 2006). We include as potential food sources the isotopic values for whitefish larvae and eggs of both whitefish species, plankton, burbot, pike, and chironomids from both pelagic and littoral habitats. Whitefish eggs and larvae are only available for stickleback during winter (Roch et al. 2018; Baer et al. 2021); as such, they were not included in the model for summer samples. Furthermore, we used the isotopic values from burbot and pike as a proxy for the food source “fish”, because eggs and larvae of pike are known to be a food source for sticklebacks (Bergström et al. 2015) and the larvae of burbot are small and are available for sticklebacks the whole year round (Probst and Eckmann 2009). The δ¹⁵N and δ¹³C values of chironomids from both habitats were taken from previous samplings in Upper Lake Constance, made in 2015–2017, representing mean yearly values (kindly provided by M. Sabel and D. Straile). As correction factors, we used the widely accepted trophic enrichment factors (TEFs) (Cui et al. 2021) to estimate the direct contribution of food sources to detritivores (chironomids) and planktivores (whitefish) and omnivores (stickleback), and the indirect contribution of food sources to sticklebacks caught in the pelagic and littoral zone during winter and summer. In order to simplify the model and to enhance interpretability, we performed a-posteriori combination of some of the food sources that fell on similar regions of the isospace plot (Suppl. material 1: fig. S1). We combined the isotopic values of both species of whitefish eggs and larvae to category “Whitefish eggs” and “Whitefish larvae”, the chironomids from both habitats to “Chironomids”, and the burbot and pike to new category “Fish”. Furthermore, to get more insight into the importance of the prey category “Fish”, we also ran the model without this category. SIMMR relies on a Markov Chain Monte Carlo to find possible solutions and disregards those not probabilistically consistent with the data. The iterations run were 10⁴, the burn-in was 10³, the posterior was thinned by 10, and the number of chains fit was 4.

Furthermore, data from all examined fish species were pooled according to feeding guild in order to calculate a standard ellipse area corrected for small samples (SEA_c). The SEA_c represents the core isotopic niche of each guild after factoring in maximum likelihoods. It comprises around 40% of data and resembles a two-dimensional measurement of standard deviation (Jackson et al. 2011). The small sample size-corrected standard ellipse area (SEAc) represents the core isotopic niche area of the individuals sampled. The isotopic niche overlap of omnivorous sticklebacks with the remaining feeding guilds was subsequently calculated as a proportion of the sum of non-overlapping areas of the SEA_cs. All analyses were performed using the SIBER package (Stable Isotope Bayesian Ellipses in R, v. 2.15; (Jackson et al. 2011)) in R (v. 4.04, (R Core Team 2020)).

To compare the trophic position of sticklebacks in Lake Constance to that of conspecifics in similar ecosystems (to see if the position in ULC is common), the trophic positions of sticklebacks from Lake Constance (here: mean value of all sticklebacks, independent of habitat) were compared to lake populations from North America (Matthews et al. 2010) and Norway (Østbye et al. 2016) using a z-test on arithmetic means and standard deviation with post hoc Bonferroni correction. We used for this comparison all lakes cited by Matthews et al. (2010) and only lake populations from Norway with a clear assignment to the isotopic values of sticklebacks (Einletvatn and Farstadtvatn lake). The same trophic position calculation (according to Post 2002) and the same trophic enrichment factor (Δ_n = 3.4) used in this study also applied to analyses of these lake systems.

Unless further specified, all statistics were performed in JMP Pro 15.1 (64 bit, SAS Institute).

For sticklebacks sampled between 2017 and 2018 (mean TL 66 mm ± 5 mm (± SD)), the mean δ¹⁵N value was 14.9 ± 1.2‰ (± SD) for muscle and 14.6 ± 2.2‰ (± SD) for liver tissue. The lowest δ¹⁵N values for stickleback muscle tissue were recorded in summer (June-July), with mean values between 14.1‰–14.3‰ (Fig. 2). During autumn and winter (October-February) intermediate values of 14.5‰–15.2‰ were measured, and the highest values of between 15.3‰–15.7‰ occurred during spring in March and April (Fig. 2). Similar trends were observed for δ¹⁵N in stickleback liver tissue, with lowest mean values occurring during the summer months (12.3‰–12.4‰), intermediate values from October to December (14.6‰–15.7‰) and highest values in March and April (16.1‰–16.6‰) (Fig. 2). δ¹⁵N in the liver showed more monthly variation than muscle tissues in both littoral and pelagic habitats (F-tests on within-group variations; 3.88_185,180 and 3.25_82.70 respectively, both P ≤ 0.001). Month had a significant influence, with the effect strength greatest in muscle tissue δ¹⁵N (GLM, r² = 0.23, n = 252, P < 0.0001, Table 1). No other parameters had a significant effect (Table 1). The GLM for δ¹⁵N values of liver tissue (r² = 0.60, n = 266) showed a similar result for month and year effects (P < 0.0001 and P < 0.01, respectively). The interaction between TL and month also significantly affected liver δ¹⁵N, while no other parameters showed any significant effect (Table 1).

Figure 2.

Fitted spline intervals of δ¹⁵N and δ¹³C in the muscle and liver of sticklebacks sampled in 2017–2018. Solid lines are the mean values, and shaded areas represent upper and lower 95% confidence intervals during the course of the year in Lake Constance.

Mean δ¹³C values averaged -30.5 ± 0.8‰ (± SD) for stickleback muscle tissue and -31.2 ± 1.5‰ (± SD) for liver tissue. Muscle δ¹³C was lowest during July (-29.5‰) and fluctuated slightly between -30.1‰–-31.0‰ (Fig. 2) during all other sampling months. Stickleback liver δ¹³C values were highest during July (-28.5‰) and lowest during March (-32.2‰) and varied across other sampling months between -31.0‰–-32.0‰ (Fig. 2). The model testing effects on δ¹³C muscle values (GLM, r² = 0.27, n = 251, P < 0.0001) revealed that only month and sex had a significant influence (P < 0.0001). Post hoc comparison found that female fish had significantly lower muscle δ¹³C values than males (Student-t, P < 0.05), but the effect strength of sex was 5× smaller than for month (Table 1). All other parameters had either no impact or exhibited only weak effects (Table 1). Month and sex had a significant influence (P < 0.0001) on liver δ¹³C (GLM, r² = 0.46, n = 255, P < 0.0001), and as with muscle δ¹³C, females exhibited lower values than males (Student-t, P < 0.05). In contrast to the δ¹³C level in muscle tissue, however, habitat had a significant effect (P < 0.05) on liver δ¹³C, with significantly lower values in sticklebacks from the pelagic zone compared to littoral specimens (Student-t, P < 0.05). Overall, the effects of sex and habitat on the whole model were low (Table 1). There were no significant differences in the isotopic signature of sticklebacks caught in the pelagic and littoral habitat, except for the liver δ¹³C values.

The δ¹⁵N value of zooplankton increased from October (10.6‰) until December (13.4‰), showed the highest peak in February (14.4‰) and decreased until April to a value of 7.9‰ (Fig. 3). Eggs of pelagic whitefish showed a mean δ¹⁵N value of 13.9 ± 0.5‰ (± SD) and did not differ significantly from those of benthic whitefish, which had a mean δ¹⁵N value of 14.1 ± 1.1‰ (± SD) (t-test, P > 0.05). The combined mean δ¹⁵N value for eggs of both species was 14.0 ± 0.8‰ (± SD), significantly higher than that of zooplankton sampled in the same months (Fig. 3). The mean δ¹⁵N value of sampled whitefish larvae, 15.3 ± 0.2‰ (± SD), was also higher than that for zooplankton sampled in the same months (Fig. 3). Sticklebacks sampled in 2021–2022 exhibited a mean δ¹⁵N value of 15.4 ± 0.5‰ (± SD) and a mean δ¹³C value of -31.0 ± 0.6‰ (± SD) and were close to results from winter 2017–2018. No differences were observed in mean δ¹⁵N and δ¹³C values between month and habitats (Tukey-Kramer-HSD, P > 0.0.5). The δ¹⁵N values of sticklebacks showed no difference from those of whitefish larvae but were notably higher than those for zooplankton (Fig. 3). In March and April, when the zooplankton showed decreasing δ¹⁵N values, the δ¹⁵N values of sticklebacks continued on the same level as the month before (Fig. 3).

Figure 3.

Arithmetic mean δ¹⁵N values with standard deviation of zooplankton, whitefish eggs and larvae and sticklebacks from the pelagic and littoral zone of Upper Lake Constance sampled in winter 2021–2022.

The δ¹³C values of sampled zooplankton showed no clear temporal trend (-32.9‰ in October, -34.2‰ in December, -34.0‰ in February, -35.9‰ in March, and -34.8‰ in April). The mean δ¹³C value of pelagic whitefish eggs was -33.7 ± 0.4‰ (± SD), significantly different to that of benthic whitefish eggs at -33.1 ± 0.6‰ (± SD) (t-test, P < 0.05). The mean δ¹³C value of whitefish larvae was -33.2 ± 0.5‰ (± SD).

Nearly all analysed sticklebacks had food in their digestive tracts. Only one individual sampled in the pelagic zone during spring had an empty stomach. The numerically dominant food source for sticklebacks during spring and winter, independent of sampled habitat, were copepods (Table 2). Interestingly, in 4 out of 10 sticklebacks (40%) sampled during spring in the littoral zone, fish eggs of unknown taxa were recorded, and in winter, the stomachs of two out of 20 (10%) pelagic sticklebacks contained fragments of fish eggs and partially digested fish larvae. Differences between the stomach contents of sticklebacks sampled in the pelagic or littoral zone became more obvious during other seasons. In summer, pelagic sticklebacks consumed mostly Daphnia (Table 2) while littoral sticklebacks fed mostly (73%) on benthic macroinvertebrates, mainly chironomids (here: other, Table 2). One stickleback caught in the littoral zone during summer had consumed 14 fish eggs of unknown taxa, amounting to 38% of all food items present in that individual’s digestive tract. In autumn, copepods remained the most common prey category (81%) consumed by pelagic sticklebacks, whilst Bosmina (74%) were the most frequent prey category in littoral. Autumn was the only season when no fish (larvae or eggs) were found in either pelagic or littoral stickleback digestive tracts (Table 2).

The results of the SIMMR mixing model suggested a clear seasonal distinction in the contribution of food sources to sticklebacks (Fig. 4). During winter, whitefish larvae and whitefish eggs contributed a mean proportion of 30–40% (the highest contribution) to the diet, while zooplankton and chironomids are of lesser importance (mean values between 10–15%), independent of sampling habitat (littoral or pelagic zone) (Fig. 4). During summer, the importance of chironomids and zooplankton increased to proportions of 25–30% (chironomids) and 35–40% (zooplankton). During summer, neither whitefish eggs nor whitefish larvae are available. However, other protein-rich sources, i.e. fish larvae from later spawning species, increased in importance during that time to provide a mean value of 30% (pelagic zone) and 40% (littoral zone), compared to winter, when fish (non-whitefish) contribute a proportion of around 15% to the diet of sticklebacks (Fig. 4). When we excluded the category “Fish” from the model, the importance of chironomids and plankton increased during summer to proportions of 30–35% for chironomids and 65–70% for zooplankton (Suppl. material 1: fig. S2), however, during winter, the importance of whitefish eggs and larvae was (with a mean contribution of 30–45% to the diet) similar to the outcome of the model which included the category “Fish” (Suppl. material 1: fig. S2). The detailed summary of both Bayesian mixing models (SIMMR) outputs and matrix plots of source contribution proportions are given in the supplements (Suppl. material 1: tables S1–S4).

Figure 4.

Posterior distribution of dietary proportion estimates of different food sources from sticklebacks from the pelagic and littoral zone of ULC during summer and winter, according to Bayesian modelling, expressed as Box-and-Whisker plots with median values and interquartile range (IQR), and minimum and maximum if it doesn’t extend the IQR value beyond 1.5. Data outside this range are plotted individually.

For the other fishes species examined from Lake Constance, mean δ¹⁵N values for muscle tissue varied by up to 4.0‰, with a minimum of 9.6 ± 1.1‰ for herbivorous rudd and a maximum of 13.6 ± 0.1‰ for piscivorous pike with whitefish, bleak, roach, burbot, and tench exhibiting intermediate values (Fig. 5). At the time of sampling (summer), mean stickleback muscle δ¹⁵N (14.3 ± 1.2‰) did not differ from that of exclusively piscivorous pike (Tukey-Kramer-HSD, P > 0.05) and was significantly higher than that of all other fish species (Tukey-Kramer-HSD, P < 0.0.5). The δ¹⁵N values of quagga mussels and faucet snails were 5.6 ± 0.1‰ and 6.9 ± 0.2‰ respectively (Fig. 5).

Figure 5.

δ¹⁵N versus δ¹³C bi-plot showing the mean isotope values of aquatic consumers in Lake Constance during summer (August and September). Horizontal and vertical bars represent ± SD of total pooled data. The standard ellipse areas (SEAc) represent the core isotopic niche for each trophic guild (comprising ~ 40% of the data; (Jackson et al. 2011)).

The mean muscle δ¹³C of all analysed fish species, except sticklebacks, ranged from -29.6 ± 1.4‰ for whitefish to -21.4 ± 1.9‰ for rudd, while mean values for bleak, roach, tench, pike and burbot ranged between -28.4‰ and -25.1‰ (Fig. 5). At the time of sampling, in summer, the mean muscle δ¹³C of sticklebacks (-30.5 ± 1.0‰) was significantly lower than that of all other fish species (Tukey-Kramer-HSD, P < 0.0.5). The δ¹³C values of quagga mussels and faucet snails were -30.1 ± 0.3‰ and -24.3 ± 0.9‰ respectively (Fig. 5).

Fig. 5 shows the core isotopic niche of each feeding guild. SEAc values range from 2.43 (zooplanktivorous) to 6.56 (benthivorous). The niche overlap between omnivorous sticklebacks and other feeding guilds was limited to a 9% overlap with zooplanktivorous whitefish.

The statistical model testing effects on the stickleback trophic position (GLM, r² = 0.29, n = 249, P < 0.0001) identified a significant influence of month (P < 0.0001), which comprised > 64% of total effect strength revealed the model (Table 3). When using a Δ_n value of 3.4, the mean trophic position of sticklebacks in Lake Constance was 4.7 ± 0.6 (± SD). The reduced Δ_n value of 2.0 yielded stickleback trophic positions of 6.7 ± 1.0 (± SD). This is unrealistic given the food web and feeding guilds of Lake Constance. Consequently, further analyses of trophic position are based solely on results from the Δ_n = 3.4 iterations of the model. The lowest trophic positions for sticklebacks were calculated for the summer months of June and July (4.4–4.6) and the highest for the spring months of March and April (4.9–5.0). No other tested parameters had a significant effect on trophic position (Table 3).

The trophic position for piscivorous pike in ULC was 4.2 ± 0.2 (± SD) and for partly piscivorous burbot 4.2 ± 0.4 (± SD).

A seasonal trend in stickleback trophic position was apparent, with lowest values during the stickleback spawning season (summer months), and increasing in the spawning season of whitefish during autumn and winter (Fig. 6).

Figure 6.

Fitted spline intervals of the trophic position of sticklebacks in Upper Lake Constance. Solid lines are the mean values, and shaded areas represent upper and lower 95% confidence intervals during the course of the year.

Sticklebacks from Lake Constance have a significantly higher trophic position than investigated populations in North America and Norway (z-test, P < 0.001) (Fig. 7). The proportion of littoral carbon in the diet of sticklebacks (α) from Lake Constance (mean = 0.06, standard error = 0.01) is comparable with that observed in populations with a “limnetic-like” phenotype from Norway and North America (α ≤ 0.4; Fig. 7) but far from that documented for benthic populations (α ≥ 0.6; Fig. 7).

Figure 7.

Comparison of trophic position and littoral carbon in the diet of three-spined sticklebacks. Values for sticklebacks from Lake Constance in black (mean value of all analysed sticklebacks, independent of habitat), from North America in blue and Norway in red; error bars indicating standard deviation. The eight North American stickleback populations (Matthews et al. 2010) (PRL = Priest Lake, KL = Kennedy Lake, PAL = Paxton Lake, CL = Cranby Lake, DL = Dugout Lake) and the two from Norway (Østbye et al. 2016) (EP = Einletvatn pond, FL = Farstadtvatn lake) are subdivided into limnetic (n = 4), benthic (n = 4) and intermediate (n = 1) ecophenotypes.