Chytrid parasitism facilitates trophic transfer between bloom-forming cyanobacteria and zooplankton (Daphnia)

Parasites are rarely included in food web studies, although they can strongly alter trophic interactions. In aquatic ecosystems, poorly grazed cyanobacteria often dominate phytoplankton communities, leading to the decoupling of primary and secondary production. Here, we addressed the interface between predator-prey and host-parasite interactions by conducting a life-table experiment, in which four Daphnia galeata genotypes were maintained on quantitatively comparable diets consisting of healthy cyanobacteria or cyanobacteria infected by a fungal (chytrid) parasite. In four out of five fitness parameters, at least one Daphnia genotype performed better on parasitised cyanobacteria than in the absence of infection. Further treatments consisting of purified chytrid zoospores and heterotrophic bacteria suspensions established the causes of improved fitness. First, Daphnia feed on chytrid zoospores which trophically upgrade cyanobacterial carbon. Second, an increase in heterotrophic bacterial biomass, promoted by cyanobacterial decay, provides an additional food source for Daphnia. In addition, chytrid infection induces fragmentation of cyanobacterial filaments, which could render cyanobacteria more edible. Our results demonstrate that chytrid parasitism can sustain zooplankton under cyanobacterial bloom conditions, and exemplify the potential of parasites to alter interactions between trophic levels.

Scientific RepoRts | 6:35039 | DOI: 10.1038/srep35039 toxins 14,15 . Cyanobacteria display high resistance to grazing, which often leads to the decoupling of primary and secondary production and inefficient carbon transfer to zooplankton 16 . The inability of zooplankton to exert effective top-down control on cyanobacterial populations has traditionally been linked to the poor edibility of cyanobacteria with colonial or filamentous morphologies, the production of toxic metabolites, and their low nutritional value 17 . Meta-analyses of experimental data have shown that, although grazing resistance cannot be generalised, cyanobacteria in fact constitute a poor food resource, as they lack essential nutritional compounds for zooplankton, such as sterols and polyunsaturated fatty acids (PUFAs) 18,19 . However, field observations often report a high biomass of grazers during bloom events [20][21][22] , suggesting alternative sources of nutrition capable of sustaining zooplankton growth.
Besides grazing, parasitism can act as an additional top-down control on phytoplankton 23,24 . In addition to virus and prokaryotic parasites (e.g. lytic bacteria), a so-far hidden diversity of small microeukaryotes has been revealed by metagenomics surveys, many of which display parasitic lifestyles 25 . Among these eukaryotic parasites, phytoplankton is particularly affected by chytrids, a group of primitive fungi characterised by a free-swimming zoosporic life stage 26 . Chytrid infection is lethal and has the potential to act as a controlling agent on cyanobacterial populations, often reaching epidemic proportions 24 . Zoosporic fungi can be an important food source for zooplankton, as they provide sterols and long-chain PUFAs that are lacking in prokaryotic prey [27][28][29] . Arthropods cannot synthesise these compounds de novo, so that they need to obtain these essential lipids from their diet 30,31 . Experimental work has shown that the cladoceran Daphnia, a keystone crustacean herbivore that drives much of the secondary production in pelagic ecosystems 32 , is able to feed on chytrid zoospores infecting the large inedible diatom Asterionella formosa. Thus, constituents may potentially be channeled from inedible algae to zooplankton via a trophic link termed the "mycoloop" 33,34 . In a similar way, chytrids may enable nutrient transfer from A. formosa to copepods 35 . However, due to its size, A. formosa cannot be grazed upon either by Daphnia or copepods, leaving open the question of whether chytrids can enhance the coupling between zooplankton and smaller, yet nutritionally suboptimal phytoplankton, like cyanobacteria.
Here, we experimentally address the interrelation between predator-prey and host-parasite interactions. To do so, we used a host-parasite system based on the filamentous, bloom-forming cyanobacterium Planktothrix agardhii and its obligate chytrid parasite Rhizophydium megarrhizum. A laboratory experiment was conducted in which four genotypes of Daphnia galeata were maintained under quantitatively comparable diets of infected and uninfected cyanobacteria, together with additional experimental treatments consisting of chytrid zoospores and heterotrophic bacteria suspensions. We hypothesise that chytrid infection on phytoplankton can sustain zooplankton growth and improve its fitness under cyanobacterial dominance. Table 1 shows particulate organic carbon (POC) concentrations and counts of cyanobacteria, zoospores and heterotrophic bacteria in the respective treatments. In both infected and uninfected cyanobacteria treatments, POC concentrations were well above Daphnia requirements (> 0.6 mg C l −1 ; refs 36,37), although about 20% lower in the infected compared to the uninfected treatment. Additional treatments consisting of zoospores or heterotrophic bacteria provided comparable food quantities, yet supplied only about 10% of POC relative to cyanobacteria treatments. The density of heterotrophic bacteria in the uninfected cyanobacteria treatment was about 80% lower compared to other treatments. Infected cyanobacterial filaments were halved in length relative to conditions of absence of infection (Fig. 1).

Results
Two-way ANOVAs showed a significant effect of diet (i.e. treatment) on all measured fitness parameters. Daphnia genotype had an effect for all parameters except fecundity ( Table 2). Significant treatment by genotype interactions were found for fecundity and body size of Daphnia, both of adults and offspring. For the parameters age at maturity and body size of adults (which presented non-normal distributions), additional two-way ANOVAs were performed on aligned rank transformed data resulting in the same significance levels for adults body size, and an additional significant treatment by genotype interaction for age at maturity (p < 0.001).
Within-genotype comparisons revealed that, for all fitness parameters except fecundity, at least one Daphnia genotype performed better on an infected compared to an uninfected cyanobacteria diet (Fig. 2, Table 2). Two genotypes matured significantly earlier (Fig. 2a) and larger body sizes in adults were found for one genotype (Fig. 2d). All genotypes displayed offspring with bigger size (Fig. 2e), whereas two genotypes showed higher growth rates when fed with infected cyanobacteria (Fig. 2c). There were no significant differences in total offspring number (Fig. 2b) or in the proportion of individuals that survived until the third reproductive cycle between infected and uninfected cyanobacteria treatments (data not shown).

Treatment
Carbon (mg l −1 ) Filaments (ml −1 ) Zoospores (ml −1 ) Bacteria (10 6 ml −1 ) Cyn Zoospore and heterotrophic bacteria diets induced a strong reduction in total offspring number and growth rates relative to treatments supplying cyanobacteria (Fig. 2b,c, Table 2). However, despite lower POC concentrations supplied (about 90% lower), Daphnia on a zoospore diet performed equal or better in all other fitness parameters compared to the uninfected cyanobacteria treatment. In all cases, Daphnia displayed bigger offspring on zoospore vs. bacteria (Fig. 2e). On a zoospore diet significant differences in size of adults, total offspring and growth rates were found for one out of four genotypes when compared with a heterotrophic bacteria diet. Also, a higher proportion of replicates survived until the third reproductive cycle when Daphnia were fed with zoospores ( Fig. 3), although food quantity in both treatments was comparable (Table 1).

Discussion
In aquatic pelagic systems, primary and secondary production is subject to severe decoupling when poorly edible colonial or filamentous cyanobacteria dominate primary production 16 . Here, we addressed the question of whether parasites are able to prevent such decoupling by providing alternative trophic links between cyanobacteria and zooplankton. Chytrid parasites of cyanobacteria, although typically neglected, are ubiquitous and often burst into epidemics, reaching infection prevalence over 90% 38,39 . Our study shows that Daphnia fitness can be significantly improved when cyanobacteria are attacked by parasites. Upon chytrid infection, filamentous cyanobacteria get fragmented, potentially increasing their edibility, and alternative food sources are made available for grazers in the form of chytrid zoospores and increased abundances of heterotrophic bacteria.
In our experiment, we observed for at least one genotype a significant improvement in fitness in terms of age at maturity, growth rates, and body size of adults and offspring when Daphnia was fed with infected cyanobacteria diet compared to an uninfected cyanobacteria diet. We measured five different parameters to better characterize Daphnia fitness. Under natural conditions, earlier born offspring can imply strong contributions to the establishment of the population, especially under food limitation and in the presence of predators 40 . Body size is another important fitness trait in Daphnia. Bigger size at birth leads to higher resistance to starvation and larger body size of adults 41,42 . Adult size in turn is positively correlated with clutch size 42,43 ; hence, larger animals potentially contribute to the population with higher amounts of offspring. In our study, no differences between infected and uninfected diets were detected concerning total number of offspring. However, in natural settings, the time needed to produce offspring can be another crucial factor. We thus compared growth rates between genotypes. Two genotypes displayed increased growth rates in the infected diet compared to healthy cyanobacteria, indicating that an infected cyanobacteria diet contributes to Daphnia fitness by accelerating reproduction, rather than increasing total offspring.
To disentangle the contribution of the aforementioned additional food sources to Daphnia fitness, zoospores and heterotrophic bacteria were separated from the infected cultures and used as food in two additional experimental treatments. The resulting reduction in POC in these treatments (about 90%) compared to cyanobacterial diets induced a strong reduction in the production of offspring in all genotypes. However, these adverse effects were not observed in other fitness parameters. Notably, the zoospore diet yielded higher or otherwise not significantly different fitness in other parameters compared to the uninfected cyanobacterial diet, despite a 90% reduction in supplied POC. This strongly suggests enhanced carbon transfer efficiency when conveyed to grazers via chytrid zoospores. Zoospores are within the optimal size range for ingestion by Daphnia, and display high cellular contents of PUFAs and sterols (in particular, cholesterol and its precursors sitosterol and stigmasterol 34,44 ) that prevent the lipid limitation typically observed when cladoceran grazers feed on cyanobacteria 45 . By extracting, transforming and repacking nutrients from cyanobacteria into higher quality, more readily ingestible zoospores, chytrids upgrade cyanobacterial carbon for zooplankton, thereby increasing the coupling between primary and secondary production under cyanobacterial dominance or bloom conditions. A chytrid-mediated trophic link between Daphnia and inedible diatoms has been proposed previously, giving rise to the concept of the "mycoloop". Kagami et al. 34 undertook a 6-day life-table experiment in which Daphnia were fed with chytrid-infected and uninfected cultures of the large inedible diatom Asterionella formosa. Daphnia displayed significantly larger body size when fed with infected diatoms, as a result of feeding on chytrid zoospores. However, in that study reproduction was not assessed, raising questions as to whether zoospores alone can sustain Daphnia populations. The present experiment was conducted over three reproductive cycles, which not only allowed testing for population viability, but also minimised the potential effects of maternal nutrient reserves, which might mask nutrient limitations imposed by the different diets tested 46 . Moreover, in the former study, limitations in the experimental design (6 replicates per treatment with only one replicate left in the uninfected diet treatment) complicate the interpretation of the results. Whereas that study tested only a single Daphnia genotype, the present experiment included four different genotypes. Our results showed significant effects of genotype and/ or genotype by diet interactions for all measured fitness parameters and support the importance of addressing inter-clonal variability in Daphnia populations 47,48 . Finally, whereas Asterionella represents an inaccessible carbon source that cannot be exploited by Daphnia, cyanobacteria constitute a suboptimal, yet edible food source. This difference uncovers a new facet of the mycoloop whereby chytrid parasites act as trophic upgraders of suboptimal prey, illustrating their potential to operate in ways other than making carbon available from inedible sources.
Our experiment shows that chytrid zoospores alone are sufficient to sustain not only Daphnia growth, but also reproduction, and can enhance fitness relative to parasite-free conditions. This demonstrates that natural zooplankton populations can, in principle, be sustained solely by mycoloop contributions. The extent of the mycoloop largely depends on the range of naturally-occurring zoospore densities in the water column. Direct quantifications of zoospores from environmental samples, although seldom undertaken, have shown maximal zoospore densities of up to 500 zoospores ml −1 treatments 49 , which are about one quarter of those provided in our experimental treatments. However, much higher densities can potentially be reached. For example, Rasconi et al. 39 reported sporangia (i.e. sessile reproductive structures that release new zoospores upon maturation) densities of 3 × 10 4 ml −1 under epidemic conditions. Considering a mean release of 4-25 zoospores per sporangium 50 , free-swimming zoospores could reach densities two or even three orders of magnitude higher than provided in our assay, suggesting that the contribution of zoospores to zooplankton diet under natural conditions may be greater than shown by this experiment.
In addition to chytrid zoospores, our results suggest that heterotrophic bacteria cannot be neglected as an extra food source for Daphnia. In contrast to the study of Kagami et al. 34 , chytrid infection caused a 5-fold increase in bacterial densities compared to uninfected conditions. Such increase is attributable to the decay of  infected cyanobacteria and the subsequent release of dissolved organic carbon, which can be readily used by heterotrophic bacteria for growth. Observed increase in bacterial biomass was enough to sustain Daphnia growth and reproduction. However, on a purified bacterial diet, fewer individuals reached the third reproductive cycle (Fig. 3), and their offspring were smaller relative to the zoospore diet in all cases (Fig. 2e). This indicates that bacteria constitute a food source of lower quality than chytrids, which is consistent with the general lack of essential lipids in prokaryotes 51−53 . Still, increased shares of bacterial food sources during the decay of inedible phytoplankton can act as a conveyor of dietary energy to Daphnia. Bacteria repack otherwise poorly ingestible cyanobacteria into smaller, easily ingestible particles and potentially detoxify cyanobacterial carbon. Moreover, bacterial proliferation under natural conditions activates the microbial loop 54 , promoting the growth of heterotrophic nanoflagellates and ciliates, which in turn have the potential to upgrade the biochemical quality of prokaryotic carbon by de novo synthesis of essential lipids 55,56 . Thus, increased bacterial densities may have more profound effects under natural conditions than our experiment can show. Beside the aforementioned effects of zoospores and increased densities of heterotrophic bacteria, cyanobacterial filaments underwent fragmentation upon chytrid infection. Although the present experimental design did not allow a direct comparison of Daphnia grazing rates on shorter (infected) and longer (uninfected) filaments, a halving in length arguably reduces mechanical feeding interference and could facilitate grazing 20 . Filament  Table 2. fragmentation upon infection by chytrids has been documented in nature 57 , but its implications for zooplankton grazing remain to be assessed.
All in all, our experiment shows how the decoupling of primary and secondary production, typically assumed in cyanobacteria-dominated aquatic ecosystems, can be circumvented by the effect of parasites, which can establish new, alternative trophic links and enhance existing ones, facilitating the transfer of carbon up the food web.

Methods
Daphnia, cyanobacteria and chytrid cultures. Four D. galeata genotypes (Mugg6b, Mugg7a, Mugg11c, Mugg13c) were isolated from the eutrophic Lake Müggelsee in eastern Germany. Clonal lines were kept in jars containing 200 ml of medium (five individuals per jar) consisting of a mixture of 95% synthetic Daphnia medium (based on ultrapure water, trace elements and phosphate buffer) and 5% Z8 medium 58 . Pre-experimental conditions included: constant temperature of 20 ± 1 °C, 12:12 h light:dark cycle, and feeding every two days with 1 mg C l −1 of the green algae Scenedesmus obliquus. The filamentous Planktothrix agardhii strain NIVA CYA630, isolated from Lake Lyseren (Norway), was maintained in Z8 medium as non-axenic semi continuous cultures under 20 °C and 20 μ mol photons m −2 s −1 . The obligate chytrid parasite strain Chy-Kol2008, isolated from Lake Kolbotnvatnet (Norway) and identified as Rhizophydium megarrhizum 59 , was used to infect cultures of the cyanobacterial strain NIVA CYA630.
Preparation and characterisation of feeding treatments. Daphnia were fed with two diets consisting of uninfected or chytrid-infected cyanobacteria (Planktothrix agardhii NIVA CYA630), respectively, both providing similar POC concentrations above Daphnia requirements (Table 1). To determine feeding volumes, optical density at 750 nm (measured with 5 cm cuvettes) was correlated with POC concentrations. For the preparation of chytrid-infected cyanobacterial feeding suspensions, a standard infection protocol was used: seven days before each feeding occasion, chytrid zoospores (final conc. 2000 ml −1 ) were added to exponentially growing cyanobacterial cultures (density 4 × 10 5 filaments ml −1 ) and incubated at 20 °C and 20 μ mol photons m −2 s −1 . After incubation, infected cultures were used to feed Daphnia providing a final POC concentration of ~1 μ g C ml −1 . Adequate feeding volumes were estimated by independent POC determinations from five standard infected replicate cultures before the start of the experiment.
In order to disentangle the contribution of chytrid zoospores and heterotrophic bacteria to changes in Daphnia fitness under infected and uninfected cyanobacterial diets, two additional feeding treatments were included, consisting of chytrid zoospores (and heterotrophic bacteria) and heterotrophic bacteria only, respectively. At each feeding occasion, zoospore suspensions were obtained by sequential filtration of the cultures used for the infected cyanobacterial treatment through a 10 μ m nylon mesh, followed by 5 and 3 μ m polycarbonate filters. The filtrate was microscopically checked for the absence of cyanobacterial filaments and used as a zoospore feeding suspension. The remaining volume was filtered through a 1 μ m filter. The zoospore-free filtrate was microscopically checked for the absence of zoospores and used as a heterotrophic bacteria feeding suspension. When feeding with zoospores and heterotrophic bacteria, the same feeding volume as for the infected cyanobacteria diet was used. For all four treatments and feeding occasions, acid Lugol and formaldehyde-fixed (2% final concentration) samples were collected. Cyanobacteria and zoospore densities were determined by Utermöhl's technique 60 . Heterotrophic bacteria were counted in a haemocytometer under an epifluorescence microscope after staining with 4′ , 6-diamidino-2-phenylindole (DAPI; 1 μ g ml −1 final conc.). For each feeding occasion, the lengths of at least fifty Planktothrix filaments in both infected and uninfected treatments were measured under a Nikon Ti Eclypse inverted microscope using the NIS-Element BR 4.5 software. Actual POC concentrations of all treatments at each feeding occasion (Table 1) were determined by filtering aliquots of each feeding suspension through pre-combusted and pre-weighted GF/F filters (pore size approx. 0.7 μ m), dried for at least 24 h, weighed and analysed using an Elementar Vario EL analyzer. POC concentrations in heterotrophic bacteria suspensions were estimated from bacterial counts, assuming an average carbon content of 20 fg cell −1 61 . Experimental setup. Daphnia neonates of the 3 rd generation, all born within 24 h, were transferred individually to 50 ml vials containing 40 ml synthetic Daphnia medium. 15 replicates per treatment and genotype were set up. Clonal line identities were blinded. Daphnia were fed every 2 days with the appropriate diet (cyanobacteria, infected cyanobacteria, zoospores, heterotrophic bacteria) and media was exchanged every 4 days. All replicates were checked daily for survival and offspring production. Offspring were counted and frozen at − 20 °C for subsequent analysis. Offspring found to be dead were excluded from body size, fecundity and growth rate analyses. Adult Daphnia were taken out of the experiment after the release of the 3 rd clutch and frozen. After 24 days, the experiment was stopped for all remaining replicates that had not reached the 3 rd clutch (n = 12) or had not reproduced at all (n = 3); in the latter case Daphnia were checked microscopically for the presence of males. Body size (length from top of head to base of tail spine) of offspring from the first clutch and of adults that had released three clutches was measured using a Nikon SMZ 25 stereomicroscope and NIS-Element BR 4.5 software.
Statistical analysis. Three replicates were excluded from the overall data set: two that were found to be males and one that lay at the bottom of the vial for several days in the third week, barely moving, before recovering and reproducing toward the end of the experiment. The following Daphnia fitness parameters were assessed: maturity (age at first reproduction), fecundity (total number of living offspring), growth rate (number of living offspring per day), body size of adults and body size of offspring from the first clutch. To assess differences in age at maturity and offspring size, replicates that reproduced at least once were included in the analyses. For fecundity and growth rates, only replicates that released a third clutch were included in the analyses. If multiple offspring were born, body size of all living individuals was measured and averaged for the analysis. Occasionally, Daphnia individuals showed deformed bodies upon thawing; these individuals were excluded from body size analyses. Two-way ANOVAs were performed for each fitness parameter, including diet (i.e. treatment) and genotype as fixed factors, followed by a contrast test for the effect of diet within genotypes (least-squares means test with Holm's p-value adjustment). Data of the parameters fecundity and body size of offspring were transformed (see Table 2). For two parameters (maturity and body size of adults), data were not normally distributed, even after transformation. Results shown in the main text stem from parametric ANOVAs, given that both parameters showed equal variances of the residuals and that ANOVA is reasonably robust to non-normality. However, in these two cases, a non-parametric two-way ANOVA on aligned rank transformed data 62 was performed in parallel. The number of replicates per genotype that reached the third reproductive cycle (within the 24-day experiment) under each treatment was compared using Fisher´s exact test (uninfected vs. infected and zoospores vs. bacteria). All statistical analyses were performed in Rstudio (v.0.99.903).