Introduction:

Daphnia, and especially the pond species Daphnia magna, live in habitats with extensive variation in biotic and abiotic parameters. Much of this variation is seasonal. According to the PEG-model, abiotic (e.g. temperature, pH, light intensity, O2-concentration) and biotic factors (e.g. food quality and quantity, predation, competition) lead to the seasonal succession of phyto- and zooplankton (Sommer et al. 1986). However, not only a succession of zooplankton species (DeMott 1983) but also a seasonal change in clonal composition of zooplankton (Carvalho and Crisp 1987; King and Serra 1998; Stibor and Lampert 2000) has been found. Some changes of environmental parameters are of minor importance, and zooplankton can adjust to these by physiological acclimatization or changes in behaviour. Other environmental challenges are more severe and lead to natural selection of a zooplankton population, resulting in a seasonal shift in genotype composition and a change in genotype frequency. In the case of Daphnia, clonal succession has been found to be related to temperature and population density (Carvalho and Crisp 1987), the loss of deep water habitat due to seasonal anoxia (Geedey et al. 1996) and the presence of vertebrate and invertebrate predators (Stibor and Lampert 2000).

Another factor, which has been shown to strongly influence Daphnia population structure, is food quality. Cyanobacteria are of particularly low food quality for Daphnia. This is due to several factors. Cyanobacterial filaments interfere with the filtering apparatus of Daphnia (DeMott et al. 2001; Gliwicz and Lampert 1990), cyanobacteria lack many essential lipids, i.e. polyunsaturated fatty acids (Von Elert and Wolffrom 2001) and sterols (Martin-Creuzburg et al. 2008; Von Elert et al. 2003), and cyanobacteria often contain toxic secondary metabolites (Gademann and Portmann 2008).

Especially during the last decades, cyanobacterial mass developments, so-called blooms, have become common in eutrophic lakes; it has been claimed that these blooms are a major factor leading to the summer decline of Daphnia biomass (de Bernardi and Giussani 1990; Gilbert 1990). Protease inhibitors have been found in almost 60 % of 17 investigated Microcystis blooms (Agrawal et al. 2001) and a seasonal dynamic of sestonic chymotrypsin inhibitors has only recently been observed (Kuster et al. 2012). Protease inhibitors are thus among the most wide-spread secondary metabolites of cyanobacteria. Cyanobacterial strains containing protease inhibitors have been shown to negatively affect Daphnia by reducing growth and decreasing ingestion rates (Lürling 2003; Rohrlack et al. 1999; Schwarzenberger et al. 2010).

Although Daphnia are affected by protease inhibitors in vitro (Agrawal et al. 2005; Von Elert et al. 2012) and in situ (Schwarzenberger et al. 2010; Von Elert et al. 2012), different D. magna clones have been shown to differ in their tolerance to dietary protease inhibitors (Schwarzenberger et al. 2012). Such variability in tolerance could be the basis for selection by cyanobacterial protease inhibitors and might lead to microevolution of Daphnia, resulting in locally adapted populations. Such a local adaptation of Daphnia to protease inhibitors has been demonstrated in vitro by Blom et al. (2006).

In this study, we sampled Daphnia from a eutrophic Swedish lake that frequently experiences cyanobacterial blooms. The cyanobacterial composition of lake seston was determined monthly with denaturing gradient gel electrophoresis (DGGE). We investigated whether there was seasonal succession of D. magna clones by genotyping natural monthly subpopulations with neutral genetic markers. Simultaneously, we related several abiotic and biotic lake parameters to the resulting haplotype frequencies of the sampled D. magna clones in order to identify the possible cause of this clonal succession. By comparing clones from before and after a cyanobacterial bloom that contains protease inhibitors, we also investigated whether or not protease inhibitors from the seston were a selection factor, leading to seasonally adapted Daphnia. For this purpose, we investigated the direct targets of the protease inhibitors, i.e. digestive proteases from Daphnia, by activity stained SDS-PAGE and by measurement of IC50 values in photometric inhibition assays.

Material and methods

Study site and lake parameters

From April to September 2010, Lake Bysjön (situated in Southern Scania, Sweden: N 55.675399 E 13.545805), a lake known to frequently develop Aphanizomenon blooms (Gustafsson 2007), was sampled monthly for zooplankton and seston. Several lake parameters also were determined every month: temperature, pH, conductivity and O2-concentration. The latter two were measured with probes. Measurement of lake parameters and sampling of zooplankton and 80 l seston/month took place in the euphotic zone of the lake.

Zooplankton

Each month, living D. magna were sampled at a depth of 1–2 m with a plankton net (200 μm) and identified according to Scourfield and Harding (1966). The body length (measured as the distance from the eye to the turning point of the animals’ carapax between the furca claw and spina) and clutch size were determined on the sampling day. The animals and the reference clone P6 were cultivated in clonal lines in 250 ml jars with a food concentration of 2 mg C/l of Chlamydomonas klinobasis (strain 56, culture collection of the Limnological Institute at the University of Konstanz) for several generations. The reference clone P6 originated from a Polish lake near Warsaw that never experienced cyanobacteria.

Seston

Aliquots of filtered near-surface seston (<55 μm, which is equivalent to the edible fraction for Daphnia (Burns 1968)) were filtered onto precombusted glass fibre filters (Whatman GF/F), dried and analyzed with the Flash 2000 organic elemental analyzer (Thermo Scientific) for particulate organic carbon (POC) and particulate organic nitrogen (PON). For determination of particulate phosphorus, aliquots of the filtered seston were collected on rinsed polysulfone filters (Pall Corporation) and digested with a solution of 10 % potassium peroxodisulfate and 1.5 % sodium hydroxide in an autoclave for 60 min; soluble reactive phosphorus (SRP) was determined using the molybdate-ascorbic acid method (Greenberg et al. 1985).

About 80 l of near-surface seston were filtered through a sieve (<55 μm), concentrated to 1 l by hollow-fibre-filtration (A/G Technology Corp., Needham, CFP-1-D-6a, 0.1 μm mesh size), frozen at −20 °C and freeze-dried (Christ LOC-1 m, LPHA 1-4). The freeze-dried seston was pestled, and 50 mg of the powder was dissolved in 1 ml of 100 % methanol, sonificated for 10 min and centrifuged at 14,000×g for 3 min. The supernatant was used to inhibit the chymotrypsin and trypsin activity of the homogenate of 6-day-old, third clutch animals of the reference D. magna clone P6.

DGGE analysis of the cyanobacterial composition of monthly seston samples

Every month, 60–180 ml of the fraction >55 μm and the edible seston fraction (<55 μm) of near-surface water were filtered using a 0.22 μm MF-Millipore MCE filter (diameter 47 mm; Millipore). The filter was subsequently used for DGGE as according to Degans et al. (2002). DNA was extracted according to Zwart et al. (1998) and cleaned with the Wizard DNA clean-up system (Promega). The 16S rDNA was amplified in a nested PCR with two primer pairs (cyaF371/cyaR783 and F357-GC/R518; Zwart et al. 1998). DGGE was performed using the DCode system for DGGE (Bio-Rad) with a 35–70 % denaturing gradient (100 % denaturant corresponded to 7 M urea and 40 % formamide). In every lane, equal amounts of PCR product were applied. Electrophoresis was performed for 16 h at 75 V. The temperature was set at 60 °C. DGGE gels were stained with Sybr gold solution in 1× TAE buffer.

After the run, digitalized DGGE images were analysed using the software BioNumerics 4.5. A matrix was compiled based on band intensities, and absolute values were converted into relative values according to Van Gremberghe et al. (2009). In every monthly, sample DNA bands were cut from the gel and left overnight in 30 ml 1× TE-buffer. Afterwards, a PCR with a third primer pair (357F/518+; Zwart et al. 1998) was performed for each cut band, and the amplicon was sequenced. All sequences were analyzed with blastn (http://www.ncbi.nlm.nih.gov/).

Proteolytic activity and IC50 values

D. magna clones isolated from the Swedish population between April and June were assigned to two subpopulations: April/May and June. From each of the D. magna clones (April/May 2/7 clones, June 9 clones), 10 six-day-old animals were homogenized and centrifuged for 3 min at 14,000×g. The protein concentration of the supernatant (i.e. Daphnia homogenate) was analyzed using a Qubit fluorometer and the appropriate Quant-iT™ Protein Assay Kit (Invitrogen) according to the manufacturer’s instructions.

Chymotrypsin and trypsin activity of each Daphnia homogenate was immediately adjusted to the same activity prior to inhibition. This ensured that the proteolytic measurements between clones were comparable. Chymotrypsin activity of the Daphnia-homogenate was measured photometrically according to Von Elert et al. (2004) using the artificial substrate N-Succinyl-alanine-alanine-proline-phenylalanine-para-nitroanilide (Sigma). Ten microlitres Daphnia-homogenate were mixed with 980 μl 0.1 M potassium-phosphate-buffer, pH 7.5. The buffer contained 125 μM S(Ala)2ProPhepNA and 1 % DMSO. The change in absorption was measured at a wavelength of 390 nm at 30 °C continuously over 10 min. Trypsin activity was measured according to Von Elert et al. (2004) using the artificial substrate N-Benzoyl-arginine-para-nitroanilide (Sigma). Ten microlitres Daphnia-homogenate were mixed with 895 μl 0.1 M potassium-phosphate-buffer, pH 7.5. The buffer contained 1.88 mM BApNA and 7.5 % DMSO. The change in absorption was measured at a wavelength of 390 nm at 30 °C continuously over 10 min. The Daphnia-homogenate of the reference clone P6 was assayed after the addition of 20 μl of either methanol, methanolic seston extracts from each month, or methanolic extract from the lyophilized cyanobacterium Microcystis sp. BM25, which was isolated from Lake Bysjön in 2009 (provided by Luc DeMeester). The methanolic extract of the BM25 powder was processed in the same way as the methanolic seston extracts.

The extract of the May seston caused maximum inhibition of the reference clone. Thus, the May extract was used to compare the different D. magna clones isolated from Lake Bysjön before, during and after the cyanobacterial bloom with regard to their tolerance to naturally occurring protease inhibitors. Five to six different concentrations of the methanolic seston extract from May were tested for inhibition of D. magna trypsin and chymotrypsin activity. The concentration by which 50 % of protease activity was inhibited (IC50) was calculated fitting a sigmoidal dose–response curve using the software Graph Pad Prism, version 4.0 (GraphPad Software, Inc.).

Activity stained SDS-PAGE

For the activity stained SDS-PAGEs, Daphnia (20 μg protein each) of each homogenate of the used D. magna clones from April/May (2/7 clones) and June (9 clones) were mixed with 5 μl 4× Laemmli-buffer (Laemmli 1970), loaded onto a 12 % SDS–polyacrylamide gel (8 × 7 × 0.075 cm) and run at 200 V. After the run, the gels were activity stained for proteases according to Von Elert et al. (2004).

Microsatellites

Six polymorphic microsatellite primer pairs were chosen for the analysis of genetic differentiation of the D. magna clones (Table 1). The clones’ haplotypes were determined via microsatellites and resulted in several genotypes. Here, “Genotypes” was not identical with the term “clones”. “Clones” always referred to individual animals isolated and cultivated as single lines in the laboratory. Every month, up to 35 clones of D. magna were sampled and analysed. DNA from the D. magna clones was extracted using the peqGold Tissue DNA Mini Kit (peqlab) according to the manufacturer’s instructions. Each subsequent PCR reaction contained 1 μl of DNA template, 5 μl 10× Taq Buffer advanced (5Prime), 0.2 μM dNTPs, 2.5 mM of each primer (fluorescence-labelled forward primers) and 2.5 U Taq-Polymerase in a final volume of 50 μl. Cycling parameters were 95 °C for 15 min to activate the DNA polymerase, followed by 30 cycles of 94 °C for 30 s, the specific annealing temperature for the microsatellites (Table 2) for 30 s and 72 °C for 10 min. Six microlitres of PCR product was mixed with 9 μl of a 1:300 dilution of Gene Scan 500 Rox Size Standard (ABI). Allele sizes were measured with the ABI 48-capillary 3730 DNA Analyzer and analysed with the software GeneMarker 1.8 (SoftGenetics). FST values and haplotype frequencies were calculated using the program Arlequin 2 (Schneider et al. 2000). Canonical correspondence analysis (CCA, ter Braak 1986) was performed to test for possible effects of any of these nine biotic and abiotic environmental variables on the seasonal change in frequency of the 11 most common haplotypes (i.e. haplotypes present during at least 2 months).

Table 1 Forward and reverse primer sequences, accession numbers and annealing temperature (Ta) of six microsatellite loci used for the genotyping of the D. magna clones
Full size table
Table 2 Pairwise FST values estimated for each pair of D. magna sampled from April to August 2010
Full size table

Statistics

Statistical analyses were conducted with the program Statistica 6.0. Data were analyzed using one-way ANOVA and a post hoc analysis (Tukey HSD). A Levene’s test was conducted to ensure homogenous variances. The level of significance was p < 0.05. Canonical correspondence analysis (CCA) was calculated using the program Canoco 4.5; a Monte Carlo significance test (499 permutations) was calculated after log-transformation of haplotype frequencies.

Results

Seasonal succession within the D. magna population in Lake Bysjön

The number of D. magna haplotypes increased from three in April to up to 20 in July (Fig. 1). Due to the low abundance of D. magna during September, only one individual could be sampled during this month; this sample was excluded from further analyses. The number and relative abundances of haplotypes changed over the seasons (Fig. 1). Furthermore, a seasonal succession of clones was observed. Although a total of 43 different haplotypes were detected, many of them were found only on a single date, and only 11 haplotypes were present for at least 2 months. This succession and the changes in number and frequency of haplotypes in the subpopulations resulted in a continuous increase of pair-wise FST values over the season (Table 2). However, only the pair-wise FST values between the early summer months and August were significantly different from those of the other samples (Table 2).

Fig. 1
figure 1

Haplotype frequencies of all D. magna clones sampled from April to September 2010 in Lake Bysjön, Sweden. The haplotypes are based on six polymorphic microsatellite loci. Every shade represents a single haplotype (average number of haplotypes 43). Since only one animal could be sampled in September, an overestimation of the frequencies of the two haplotypes of this animal was observed here. Therefore, all samples and environmental variables from September were excluded from further analyses. (#gen = number of genotypes sampled per month, #hap = number of haplotypes underlying the genotypes sampled per month)

Full size image

Cyanobacterial composition of monthly seston samples

On the DGGE gel, following the amplification of 16S rDNA, bands were visible in every month. In most of the whole seston samples (>55 μm), the same or a higher number of bands were present than in the seston fraction smaller than 55 μm (Table 3). In every edible fraction of the monthly seston samples (<55 μm), the DNA sequences could be assigned to cyanobacteria (Tables 3, 4; Suppl. 1 and 2). In every month, cyanobacteria were present and ingestible for D. magna in Lake Bysjön with a heavy cyanobacterial bloom occurring in May. In this bloom, 61.3–66.13 % of the hits were assigned to filamentous cyanobacteria of the genus Anabaena (Table 3), according to the 16S rDNA sequence analysis,

Table 3 Cyanobacterial composition (%) of monthly seston samples resulting from DGGE analysis
Full size table
Table 4 Assignment of the bands (depicted as % of denaturant) to the first three hits for cyanobacteria with the highest score in the blastn search after sequencing of the DGGE bands; accession numbers are given in parentheses
Full size table

Seasonal changes of abiotic and biotic parameters in Lake Bysjön

Monthly samples of the natural phytoplankton were tested for their inhibitory effects on digestive proteases of the reference clone. When extracts of the same biomass of the edible phytoplankton fraction were tested, the sample from May inhibited Daphnia chymotrypsins strongly, whereas most of the other samples did not (Fig. 2). The same phytoplankton extracts had no effect on Daphnia trypsins (data not shown).

Fig. 2
figure 2

Relative inhibition (n = 3, mean + SD) of the digestive chymotrypsin activity of the reference clone D. magna P6 by methanolic extracts of the edible fraction (<55 μm) of natural phytoplankton from 2010 from Lake Bysjön, Sweden, and by a methanolic extract from the cyanobacterium Microcystis sp. BM25

Full size image

Every month, several abiotic and biotic parameters were measured and investigated for seasonal variability. Conductivity and pH proved to be stable over time (Fig. 3a, b). Dissolved oxygen concentrations were variable (Fig. 3c), and also temperature changed substantially over the season (Fig. 3d). The abundances of Chaoborus sp. larvae showed strong seasonal variation (Fig. 5a). The densities from April to June ranged between 0 and 0.075 Ind./l, whereas abundances reached a maximum of 1.33 Ind./l in July. Determination of fish larvae abundances was not feasible within this study. Here, we used the body size of egg-bearing, i.e. reproducing, D. magna as surrogate parameter for fish predation. The average body size remained constant throughout the season. Food quantity (Fig. 4a) and quality (Fig. 4b, c) changed over the season.

Fig. 3
figure 3

Environmental parameters from Lake Bysjön from April to September 2010. Conductivity (a), pH (b), oxygen (c) (one missing data in August) and water temperature (d) were measured in the euphotic zone

Full size image
Fig. 4
figure 4

Particulate organic carbon (POC, a), molar ratios of carbon:phosphorus (b) and carbon:nitrogen (c) of the edible fraction (<55 μm) of the phytoplankton from Lake Bysjön from April to September 2010. At a C:P > 300 or a C:N > 15 in the phytoplankton, Daphnia, which have to maintain a constant elemental ratio, may become P- and N-limited (Urabe and Watanabe 1992). All food parameters showed a conspicuous seasonality, with the highest POC during the cyanobacterial bloom. High C:P ratios in summer indicate potential P-limitation of D. magna in Bysjön, whereas N-limitation is improbable

Full size image
Fig. 5
figure 5

Density of Chaoborus sp. larvae and size of reproductive D. magna in Lake Bysjön in 2010. The abundance of Chaoborus sp. larvae (a) was determined by counting monthly zooplankton samples. b Body length (mean + SD) of up to 38 egg-bearing living D. magna mothers was measured monthly

Full size image

Canonical correspondence analysis of haplotype frequencies and lake parameters

The possible effects of the 9 biotic and abiotic environmental variables on the seasonal change in frequency of the 11 most common haplotypes were analyzed in a canonical correspondence analysis (Fig. 6). The frequencies of each haplotype per month are shown in Table 5. The eigenvalues, which measure the importance of an ordination axis, were 0.699 (x-axis) and 0.409 (y-axis). The variance explained by CCA was 54 % for axis 1 and 31.6 % for axis 2. Axes 3 and 4 explained only 14.4 % and were thus negligible. Hence, all environmental variables that lie within the effective area of both axes together (grey shades in Fig. 6) had the highest influence on the haplotype-frequency distribution. Since the length of the vector arrows indicates the strength of the influence of these environmental variables as well, we found the C:P ratio and the C:N ratio of the edible fraction of natural phytoplankton, temperature and O2-concentration to have the greatest influence on the seasonal haplotype frequencies. A Monte Carlo permutation test for the first axis alone is stricter than the test for all axes together. Both tests were not significant (significance of the first canonical axis alone: F value < 0.001; p = 1). Thus, the effect of the environmental variables on the distribution of haplotype frequencies was not significant. In conclusion, no effect of a single environmental variable on the haplotype frequencies over the summer was detected by CCA, which indicates that the temporary presence of dietary chymotrypsin inhibitors (May–June) had no effect on haplotype frequencies of D. magna.

Fig. 6
figure 6

Canonical correspondence analysis (Canoco 4.5; Monte Carlo significance test; log-transformation of haplotype frequencies) showing microsatellite-based haplotype frequencies of the 11 most common haplotypes present during at least 2 months (triangles) in Lake Bysjön in April–August 2010 and influence of several environmental variables (vector arrows). The length and the direction of the arrows explain the strength of the influence of the environmental variables on the haplotype frequencies (C:N and C:P: molar particulate carbon–nitrogen rsp. carbon-phosphorus ratio of the edible fraction of natural phytoplankton; O2: O2-concentration (mg/ml) of near-surface water; CT-Inhib: strength of inhibition of chymotrypsin activity of D. magna homogenate (%) by extracts of the monthly samples of natural phytoplankton in comparison to not-inhibited CT-activity; Magna: monthly abundance of D. magna (Ind./l); pH: pH of near-surface lake-water; Conducti: conductivity (μS/cm) of near-surface lake-water; Chaoboru: monthly Chaoborus sp. abundance (Ind./l); T: temperature (°C) of near-surface lake-water; Hap: haplotype). Eigenvalues: axis 1 0.699, axis 2 0.409. Grey shades effective area of both axes together

Full size image
Table 5 Haplotype frequencies of the 11 D. magna haplotypes (hap1 to hap11) included in CCA
Full size table

Comparison of two Daphnia subpopulations before and after a bloom with protease inhibitors

We investigated individual clones isolated in April/May (before the protease-inhibitor-containing bloom) and June (after the bloom). The direct targets of protease inhibitors, i.e. digestive proteases of Daphnia, were visible as nine protease bands on activity stained SDS-PAGE for each clone (Fig. 7). This pattern has already been shown in a D. magna clone, and the upper five bands were assigned to trypsins (T; Agrawal et al. 2005) while the lower four bands are chymotrypsins (CT; Schwarzenberger et al. 2010; Von Elert et al. 2012). All bands were previously assigned to six digestive protease genes [T152, T208, T610, CT383, CT448, CT802; (Schwarzenberger et al. 2010)]. The nine clones from the “June” subpopulation all showed the same pattern (Fig. 7, not all clones delineated), whereas the patterns of the nine clones from “April/May” differed. In comparison to the “June” subpopulation, differences in the apparent molecular weight of two chymotrypsin bands (CT383 in clone M19, CT448 in M24; indicated as arrows, Fig. 7) were found. Most clones (6 out of 9) showed the protease band pattern of clone M7.

Fig. 7
figure 7

Activity stained SDS-PAGE of the homogenate of five different D. magna clones from April/May and June. White bands indicate active proteases; their corresponding genes are depicted on the right side of the gel [trypsins: T152, T208 and T610, chymotrypsins: CT383, CT448 and CT802; (Schwarzenberger et al. 2010]. Since all clones from June showed the same band pattern, only two clonal homogenates (J 15 and J 17) are shown here. In April/May different band patterns were found as exemplified here by three clones (M7, M19 and M24). Arrows indicate bands of the clones from April/May that differ from the homogenous band pattern of the clones from June

Full size image

We used IC50 values of D. magna gut homogenate of 18 clones isolated in April/May and in June as a measure of tolerance of the digestive chymotrypsins to dietary chymotrypsin inhibitors from the May phytoplankton, which had proven to have the highest strength of inhibition against chymotrypsins (Fig. 2). Between clones from April/May and June, no difference was detectable with respect to the IC50 values (April/May 34.79 ± 6.45 μg DW/ml, June 29.32 ± 5.21 μg DW/ml one-way ANOVA, p = 0.067).

Discussion

The dominance of cyanobacteria in lakes has been reported to be one of the reasons for the observed summer decline of Daphnia abundance across and within lakes (Ghadouani et al. 2003; Sommer et al. 1986; Threlkeld 1979). One of the reasons for this decline might be the well-known content of toxins and of an array of other biologically active secondary metabolites produced by cyanobacteria (Carmichael 1992; Carmichael 1994; Gademann and Portmann 2008). Among these bioactive metabolites are cyanobacterial protease inhibitors, which have been found in different cyanobacterial genera (Weckesser et al. 1996) as well as in different strains of the same species (Martin et al. 1993; Von Elert et al. 2005). Protease inhibitors are widespread in nature. For example, they have been found in almost 60 % of 17 investigated Microcystis blooms (Agrawal et al. 2001) and only recently has the seasonal dynamic of chymotrypsin inhibitors been reported (Kuster et al. 2012).

Cyanobacterial protease inhibitors often inhibit serine proteases. These include trypsins and chymotrypsins, which represent the most important digestive enzymes in the gut of D. magna (Von Elert et al. 2004). Total trypsins and chymotrypsins of D. magna have been shown to be specifically inhibited by cyanobacterial protease inhibitors in vitro (Agrawal et al. 2005; Von Elert et al. 2012) and in situ (Schwarzenberger et al. 2010).

Dietary cyanobacterial protease inhibitors have been shown to trigger specific physiological responses in Daphnia, i.e. changes in the activity of digestive proteases, changes in protease gene expression and induction of protease isoforms (Schwarzenberger et al. 2010). These specific physiological responses have also been shown to occur when the pure inhibitors were fed to Daphnia via liposomes (Von Elert et al. 2012). This demonstrates that these specific physiological responses are actually due to protease inhibitors and are not caused by other unknown cyanobacterial metabolites. These specific physiological responses are most probably costly, since they only occur when protease inhibitors are present. Furthermore, a maternal transfer of induced protease gene expression has been reported and was shown to lead to higher tolerance of the offspring generation (Schwarzenberger and Von Elert 2012). These findings also make it reasonable to assume that these responses have been shaped by selection by cyanobacterial protease inhibitors.

It has been shown that the seasonal dynamics of phytoplankton succession, with cyanobacteria occurring in summer, leads not only to a decline in abundance of Daphnia but to a seasonal change of zooplankton composition as well (Sommer et al. 1986). It is quite likely that the seasonal dynamic of phytoplankton also leads to a shift in genotype composition of daphnids, since clonal succession of Daphnia has been reported with regard to several other environmental parameters (Carvalho and Crisp 1987; Geedey et al. 1996; Stibor and Lampert 2000). Such a clonal succession of Daphnia might therefore also be caused by the seasonal dynamics of the content of protease inhibitors in the seston: The occurrence of protease inhibitors as strong selection factors in a natural environment should lead to positive selection of Daphnia clones with a higher tolerance to these inhibitors and should result in a shift in clonal composition. We chose the Swedish Lake Bysjön to test this hypothesis for two reasons: Lake Bysjön is known to frequently develop cyanobacterial blooms (Gustafsson 2007), and it is relatively small and thus more likely to show seasonal succession of Daphnia clones. This is due to the fact that marked genetic changes are typical of Daphnia populations, which inhabit small or shallow ponds (Hebert 1974; Young 1979), whereas a relatively stable clonal composition over the seasons has been found in Daphnia populations in large lakes (Mort and Wolf 1985).

A clonal succession of D. magna haplotypes was observed in Lake Bysjön in 2010, with only 11 of 43 haplotypes present in more than 1 month. In order to investigate if a seasonal occurrence of dietary protease inhibitors might have caused these changes in haplotype composition, monthly samples of the natural phytoplankton of Lake Bysjön were tested for their inhibitory effects on digestive proteases of a reference clone of D. magna. DGGE analyses indicated that in monthly samples of lake seston, cyanobacteria were present throughout the whole sampling period; in May, a heavy cyanobacterial bloom was observed in Lakes Bysjön. DGGE analyses suggested the presence of filamentous cyanobacteria most probably Anabaena. Interestingly, from strains of the genus Anabaena serine protease inhibitors have been isolated (Fujii et al. 1995). When extracts of this phytoplankton biomass were tested, the sample inhibited Daphnia chymotrypsins very strongly, whereas most of the other samples did not. The strength of chymotrypsin inhibition by the natural seston extracts from Lake Bysjön was comparable with that of an extract of a pure cyanobacterial strain that produces protease inhibitors. This finding suggests that the cyanobacteria of Lake Bysjön in May 2010 produced a high amount of sestonic chymotrypsin inhibitors. Here, the same phytoplankton extracts that were tested against digestive chymotrypsins had no effect on Daphnia trypsins (data not shown), demonstrating that selective pressure on digestive proteases of Daphnia in Lake Bysjön can be expected to occur in chymotrypsins but not in trypsins.

However, besides the seasonal change in strength of chymotrypsin inhibition, other factors also might have been responsible for the change in D. magna haplotype composition. It has been shown that a seasonal change in a variety of abiotic factors also might cause succession in Daphnia (DeMott 1983). However in Lake Bysjön, conductivity and pH proved to be stable over time, with the observed pH values of as high as 9 to be known to favour the dominance of cyanobacteria (Shapiro 1990). Here, O2 concentrations always exceeded 10 mg/l, rendering seasonal effects of O2 unlikely, since the coexistence of Daphnia species is known to be affected by low oxygen levels (Sell 1998). Temperature, which is assumed to affect community composition of zooplankton (Moore et al. 1996); changed substantially over the season.

Biotic parameters also have been demonstrated to affect Daphnia composition. Major parameters are predation, food quantity (Gliwicz 1990) and food quality (Weider et al. 2008). Predation may be due to invertebrate larvae (mainly Chaoborus sp.) or planktivorous fish (mainly young-of-the-year fish in early summer). In an earlier study (Voss and Mumm 1999), densities of Chaoborus larvae ranged between 78 and 335 Ind./m2 in a water column of 5 m, which translates into a density of 0.016–0.067 Ind./l. These densities are comparable with our data from April to June (range 0–0.075 Ind./l), whereas abundances in Lake Bysjön reached a maximum of 1.33 Ind./l in July, which was much higher than the data given in the literature. These findings suggest that predation by Chaoborus is weak during the cyanobacterial bloom in early summer (May), but that it is very strong in August.

Fish larvae strongly select for large-sized zooplankton, so that the highly synchronized hatching of fish larvae in early summer leads to a pronounced reduction in body size of Daphnia (Gliwicz and Wrzosek 2008). It has been shown that a seasonal shift of the composition of a Daphnia population led to summer clones that reproduced at a smaller size than clones from spring; this was most probably due to fish predation (Brzezinski et al. 2010). However, body size of egg-bearing, i.e. reproducing, D. magna remained constant throughout the season, indicating that fish predation remained on the same level throughout the whole season in Lake Bysjön.

Food quantity and quality changed over the season. However, food quantity clearly did not affect the composition of D. magna genotypes, as it was continuously far above the incipient limiting level of 0.5 mg C/L (Lampert 1977). Food quality as determined by the C:P or the C:N ratio of the food may affect coexistence of Daphnia (Weider et al. 2008). In Lake Bysjön, the relative content of phosphorus but not that of nitrogen in the edible fraction of natural phytoplankton was temporarily so low that it might have affected Daphnia coexistence (Fig. 4).

Cyanobacterial protease inhibitors were not the only parameters that varied seasonally in 2010. However, none of these parameters were found to be responsible for the seasonal shift in haplotype composition of D. magna in Lake Bysjön in 2010. Although no correlative evidence that the change in protease inhibitor content of the seston affected the clonal D. magna succession was found, a temporal exposure to the protease inhibitors in May might have exerted a strong selection pressure on the D. magna population from Lake Bysjön. Thus, if the dietary chymotrypsin inhibitors favoured more tolerant genotypes, D. magna clones isolated in April/May, i.e. before and during the bloom, should not have been affected by protease inhibitors, whereas genotypes in June could already have undergone selection by these inhibitors.

Since digestive proteases of Daphnia are the direct targets of protease inhibitors, we investigated the two D. magna subpopulations (April/May and June) for differences in these proteases due to possible selection. Von Elert et al. (2004) observed nine protease bands in D. magna gut homogenate on an activity stained SDS-PAGE. Agrawal et al. (2005) found 5 of the bands between 28 and 75 kDa to be trypsins, whereas the 4 others were assigned to chymotrypsins (Schwarzenberger et al. 2010; Von Elert et al. 2012). Here, a difference in the band pattern of these target proteases was observed among clones from the subpopulation from before and during the cyanobacterial bloom (April/May). However, after the bloom a homogenous band pattern was found in all clones of the Daphnia subpopulation from June (data not shown). It was conspicuous that this difference in the protease pattern occured for chymotrypsins but not in trypsins, which suggests a selection of chymotrypsin isoforms by the chymotrypsin inhibitors present in the cyanobacterial bloom. Since no other biotic or abiotic parameters showed a peak during the cyanobacterial bloom in May, we can exclude a bottleneck event on the Daphnia subpopulation from April/May. We thus assumed that the exclusion of clones with different protease band patterns was not randomly caused by one of these parameters, but was rather due to the presence of cyanobacterial chymotrypsin inhibitors. We therefore expected the chymotrypsin isoforms of the June subpopulation to have a higher tolerance to the protease inhibitors from May than the mix of isoforms of the subpopulation from April/May. IC50 values have been shown to determine the variation in clonal fitness and are thus a measure of tolerance (Schwarzenberger et al. 2012). However, IC50 values for inhibition of Daphnia chymotrypsins with the extract of the May seston did not differ between clones from April/May and June. This indicates that the cyanobacterial bloom in May did not select for clones with a higher tolerance of digestive chymotrypsins to the natural protease inhibitors.

Alternatively, the frequent observations of cyanobacterial blooms in Lake Bysjön for several decades (Gertrud Cronberg, personal communication) suggest that the D. magna population is locally adapted. If these blooms have frequently contained protease inhibitors, they should thus have selected for a more tolerant D. magna population. This reasoning is supported by the finding that the genetic structure of the D. magna population has apparently been quite stable at least over the last 3 years, since six of the most abundant haplotypes in 2010 were also among the most abundant haplotypes of D. magna clones that had been isolated in 2008 from the same lake (data not shown).

Further investigations are required to test for a possible local adaptation of the Lake Bysjön D. magna population to chymotrypsin inhibitors by comparing the results with those from a population that stems from a lake without cyanobacterial protease inhibitors. A local adaptation of a Daphnia population to microcystin-containing cyanobacteria has already been demonstrated (Sarnelle and Wilson 2005); such an adaptation was assumed by (Hairston et al. 1999) to be due to microevolution. With regard to cyanobacterial protease inhibitors, an in vitro study has reported local adaptation of Daphnia to a trypsin inhibitor (Blom et al. 2006). In this study, clones from two Daphnia populations were treated with the dissolved inhibitor in acute toxicity test. However, local adaptation to protease inhibitors still remains to be demonstrated by subjecting Daphnia to the natural way of exposure, i.e. by feeding them cyanobacteria that contain protease inhibitors.