Increasing frequency, duration, and range of cyanobacteria blooms reduce water quality and disrupt the flow of energy from primary production to higher trophic levels due to their poor nutritional quality, toxicity, and morphological defenses for zooplankton grazers (Dickman et al., 2008; Rastogi et al., 2015; Heathcore et al., 2016). While elevated nutrient concentrations and warm temperatures are recognized as key drivers of blooms (Smith and Schindler, 2009; Paerl and Otten, 2013), the trophic interactions in eutrophic systems, and especially the question of “who grazes the bloom,” is less understood (Urrutia-Cordero et al., 2015; Ger et al., 2016c). This is partly because of the bias toward studies in cooler temperate climates with relatively short bloom durations, where large and tolerant generalist grazers like Daphnia make top down control a possibility even under ideal abiotic conditions for cyanobacterial growth (Sarnelle et al., 2010; Chislock et al., 2013). Yet, in warmer climates with longer duration blooms, large generalist grazers are replaced with smaller and more selectively grazing zooplankton (Fernando, 1994; Hansson et al., 2007). Warmer, more eutrophic conditions are proliferating and predicted to expand globally (Paerl and Huisman, 2009; O'Neil et al., 2012). Consequently, trophic interactions in warmer and bloom-dominated waters deserve more attention, not only to better understand the ecology of cyanobacteria in tropical regions but also because they can be used as models for predicting future changes in temperate lakes (Ger et al., 2014; Paerl, 2017).

While cyanobacterial traits like toxicity and colonial morphology act as grazer defenses, grazing pressure on cyanobacteria also depends on the functional traits of zooplankton co-existing with blooms. Size, grazing behavior, and physiological tolerance to ingested cyanobacterial toxins are key traits controlling the degree of grazing pressure on blooms (DeMott et al., 1991; Kirk and Gilbert, 1992; Litchman et al., 2013). While generalist and tolerant grazers may reduce cyanobacterial blooms (Chislock et al., 2013), selective and tolerant grazers are expected to facilitate cyanobacteria by grazing on their eukaryotic phototrophic competitors (Scotti et al., 2015). Indeed, zooplankton that tolerates a limited ingestion of cyanobacteria while selectively grazing on their eukaryotic phytoplankton competitors has long been hypothesized to promote blooms (Kirk and Gilbert, 1992; Mitra and Flynn, 2006; Hong et al., 2013). This prediction, however, is largely based on short-term (<4 h) grazing experiments or modeling studies and remains untested experimentally at time scales above a few hours. Longer incubation periods would improve our understanding of the effects of selective grazers on cyanobacterial dominance. Specifically, longer incubations can be designed to account for phenotypic plasticity in grazing pressure (Ger et al., 2011) or cyanobacterial defenses (Jang et al., 2007); as well as effects due to starvation (e.g., prey switching) or mortality (e.g., cyanobacterial toxicity); and phytoplankton nutrient limitation or regeneration by zooplankton excretion.

High zooplankton biomass, especially calanoid copepods >50 ind.L⁻¹, often co-exists with permanent blooms across tropical and subtropical lowland eutrophic lakes and reservoirs (Bouvy et al., 2001; Rietzler et al., 2002; Sousa et al., 2008). In tropical South America, calanoid copepods from the genus Notodiaptomus often dominate zooplankton communities, and their abundance can be positively correlated to toxic cyanobacteria blooms (Eskinazi-Sant'Anna et al., 2013; Rangel et al., 2016). This makes Notodiaptomus species ideal candidates to study plankton interactions in eutrophic waters characterized by long-duration cyanobacteria blooms. Moreover, recent experiments showed that ingestion of microcystin producing Microcystis had no effect on survival or egg production of the copepod Notodiaptomus iheringi in the presence of alternative nutritious prey, and identified both physiological tolerance and selective avoidance of Microcystis as key traits for co-existing with toxic cyanobacteria during blooms (Ger and Panosso, 2014; Ger et al., 2016b). Hence, the copepod N. iheringi exhibits zooplankton traits that can be expected to facilitate cyanobacteria.

To experimentally evaluate the facilitation of cyanobacteria by grazers across periods longer than a few hours, however, it is necessary to account for all factors controlling the competition between cyanobacteria and eukaryotic phytoplankton. Thus, in addition to herbivory (i.e., selective removal of a competitor), potential effects due to interference competition (e.g., allelopathy) must also be accounted for. Allelopathy is the release of chemical cues into the environment by an organism that inhibits the growth of a competitor (Rice, 1984; Leflaive and Ten-Hage, 2007; Dias et al., 2017).

Cyanobacteria can have allelopathic effects on other groups of phytoplankton including cryptophytes (B-Béres et al., 2012), chlorophytes (Bar-Yosef et al., 2010; Leão et al., 2010; Bittercourt-Oliveira et al., 2015), and other cyanobacteria (Mello et al., 2012; Zhai et al., 2013). Yet, eukaryotic phytoplankton can also release allelopathic compounds, and an established community of chlorophytes might also suppress the growth of cyanobacteria (Bittercourt-Oliveira et al., 2015). As a result, understanding eutrophic plankton communities and the trophic interactions controlling cyanobacterial dominance requires experiments that account for multiple biotic interactions such as grazing and allelopathy (Granéli et al., 2008).

Accordingly, we tested the hypothesis that selective grazing would facilitate the dominance of cyanobacteria by direct elimination of their eukaryotic phytoplankton competitors. For this, we evaluated the trophic interactions between the tropical copepod N. iheringi and two competing phytoplankton prey: the prokaryotic cyanobacteria Microcystis and the eukaryotic Cryptomonas. The effect of this grazer on Microcystis dominance was evaluated in a 6-day incubation experiment, where we compared the abundance and growth of co-cultured phytoplankton with and without copepods with a 6:1 initial dominance of Cryptomonas biomass. Although the main objective was to test grazer facilitation of cyanobacteria, we also performed a secondary experiment in order to rule out any allelopathic effects in the absence of grazers. This experiment tested if the presence and initial proportion of a competitor inhibited growth of either phytoplankton relative to single species cultures by comparing growth rates of each phytoplankton (Microcystis and Cryptomonas) in single and co-cultures across various initial biomass proportions.

We evaluated the interaction of a common bloom-forming cyanobacterium with a eukaryotic phytoplankton competitor in the presence and absence of a copepod grazer to test whether zooplankton could facilitate cyanobacterial dominance. Adding the copepod N. iheringi to the competing phytoplankton increased the relative abundance of Microcystis co-cultured with Cryptomonas compared to no-grazer controls. Strong grazing pressure reduced the abundance of Cryptomonas, which resulted in negative growth rates for this phytoplankton. In contrast, neither the abundance nor the growth of Microcystis was affected significantly by N. iheringi. Overall, these results indicate that N. iheringi, a copepod commonly associated to cyanobacterial blooms in nature, facilitated the dominance of Microcystis by selectively consuming Cryptomonas. Hence, our results provide unique experimental evidence that zooplankton co-occurring with blooms may promote cyanobacterial dominance by selective grazing on their eukaryotic phytoplankton competitors. We believe these are novel results with implications for the role of selective grazers on cyanobacterial bloom dynamics and plankton ecology in general, especially given the future scenarios of higher temperatures and increased eutrophication (Paerl and Huisman, 2009).

The results emphasize the role of biotic drivers for facilitating the dominance of cyanobacteria and possibly other taxa of inedible phytoplankton. Top-down regulation of the phytoplankton community composition by zooplankton, especially by large generalist grazers like Daphnia, has been recognized for decades (Bergquist et al., 1985; Sterner, 1989). Increased grazer pressure, especially by large cladocerans, can shift the phytoplankton community to inedible chlorophytes or cyanobacteria with a colonial gelatinous morphology (Shapiro and Wright, 1984; Carpenter et al., 1987). Considerable research has also focused on the ability of zooplankton to control phytoplankton biomass and especially cyanobacterial blooms (e.g., Jeppesen et al., 2007). Yet little is known about the potential for zooplankton to positively affect cyanobacteria. Indeed, to our knowledge there is no previous experimental evidence besides short-term grazing experiments (Jiang et al., 2013; Ger et al., 2016b), though modeling studies (Mitra and Flynn, 2006) and associations from mesocosms and monitoring data (Hong et al., 2015; Rangel et al., 2016) have suggested positive interactions among selective grazing zooplankton and cyanobacteria. Hence, by scaling up from short-term grazing assays to using artificial foodwebs, our results provide further evidence for a plausible mechanism explaining the co-existence of copepods and cyanobacterial blooms. The top-down effect of smaller, selectively grazing zooplankton on the structure and function of phytoplankton community, however, remains poorly understood (Ger et al., 2016c), which limits predictions for the ecology of a warmer and more eutrophic world where these zooplankters are expected to dominate.

Selective grazing on Cryptomonas and avoidance of Microcystis by the copepod N. iheringi was likely due to a combination of toxicity and poor nutrition. Such behavior relies on cues to detect the size, shape, chemical composition, and toxicity of potential prey to maximize nutritional gain and minimize ingestion of harmful prey (DeMott and Moxter, 1991; Kleppel, 1993; Heuschele and Selander, 2014). The nutritional quality of phytoplankton for zooplankton grazers is largely based on sterols and fatty acids (Muller-Navarra et al., 2000; von Elert et al., 2003; Ravet et al., 2010). Cyanobacteria, including Microcystis, are low in nutrition while Cryptomonas is nutritious with a high concentration of highly unsaturated fatty-acids (Brett and Müller-Navarra, 1997). Copepods can avoid ingesting cyanobacteria by sensing chemical cues (e.g., toxins), while maintaining uninhibited ingestion on alternative prey (DeMott, 1986). Indeed, phytoplankton anti-grazer defenses also regulate zooplankton grazing and copepod prey selection (Pohnert et al., 2007). Both toxins and colonial morphology have been shown to defend phytoplankton against zooplankton grazers (Fulton and Paerl, 1988; Selander et al., 2006). In this study, the toxicity of the Microcystis strain likely explains copepod avoidance as the single or bi-cellular nature of the cultured Microcystis rules out colonial morphology as a defensive trait. Indeed, previous studies have suggested that microcystins may act as anti-grazer defenses for copepods including N. iheringi (Ger et al., 2016a,b). Moreover, although copepods are known to select for larger sized particles (DeMott et al., 1991), N. iheringi ingestion of a smaller (3 μm) non-microcystin producing Microcystis strain was 3-fold higher than a larger (5 μm) microcystin-producing strain (Ger et al., 2016b). Thus, the reason why Microcystis was avoided in this study was likely prey chemistry (i.e., toxicity), not size. In contrast with our study, however, Microcystis and other bloom-forming cyanobacteria exist as larger colonies or filaments in nature. Future work on grazer facilitation of cyanobacteria would therefore benefit from accounting for morphological defenses in similar incubations.

Rates of Microcystis clearance by N. iheringi in our study were similar to previous short-term (i.e., 2 h) grazing assays with the same phytoplankton prey species (Ger et al., 2016b). Clearance rates were also similar to other short-term experiments with copepods and bloom-forming cyanobacteria measured using a variety of methods, zooplankton density, and volumes ranging from 2.5 mL to 0.5 L (DeMott and Moxter, 1991; Ger et al., 2011; Hong et al., 2013). While these results indicate that grazing rates obtained from short-term assays may also reflect longer-term dynamics, it is likely that multiple factors affect longer term grazing in this current experimental setup. The relative concentration of prey to grazers should be a key factor regulating prey switching (Kiørboe et al., 1996) and extreme depletion of one prey (e.g., Cryptomonas) may result in increased grazing on Microcystis. Yet, prey specific clearance or ingestion rates measured over 2, 4, and 6 days were similar, suggesting that N. iheringi grazing on each prey type was relatively constant during our experiment. Hence, although the abundance of Cryptomonas was declining and that of Microcystis was increasing, N. iheringi maintained higher clearance or ingestion rates for the former and we found no evidence for prey switching during the 6-day incubation. Longer-term incubations, however, would help understand whether copepods completely deplete edible phytoplankton and how rapid and persistent prey switching is given the shifting dominance of poor quality prey such as toxic cyanobacteria.

The fact that copepods facilitate the dominance of cyanobacteria by ingesting edible phytoplankton raises the question how these grazers maintain high fitness during blooms when edible phytoplankton is scarce. If the nutritious edible prey were completely depleted, zooplankton fitness would be expected to decline as a result of food limitation or toxicity from switching to cyanobacterial prey (Sommer et al., 1986). Indeed, the role of nutritious prey for controlling copepod fitness was shown previously and although N. iheringi tolerated and produced eggs in diets dominated by toxic Microcystis, fitness was a function of nutritious food (Ger et al., 2016b).

The question of how copepods such as N. iheringi reach high biomass during blooms of toxic cyanobacteria in nature (e.g., Rangel et al., 2016), however, was outside the scope of this study. One plausible explanation is that copepods are omnivores and efficiently select for optimal prey including heterotrophic protists and fungal parasites, which are known to graze on and upgrade the food quality of toxic cyanobacteria (Engström-Öst et al., 2002; Park et al., 2003; Bec et al., 2006; Agha et al., 2016). Hence, even in a complete absence of edible and nutritious phytoplankton during long-duration cyanobacteria blooms, microzooplankton grazers can be expected to provide a sufficient and nutritious food source to explain high copepod fitness. Given that heterotrophic protists are expected to graze on cyanobacteria and be grazed by crustacean zooplankton (Canter et al., 1990; Engström-Öst et al., 2002; Bec et al., 2006; Combes et al., 2013), their intermediary role in fueling zooplankton production in bloom dominated waters merits more attention.

The current results emphasize how selective grazers may facilitate cyanobacterial dominance, at least in the absence of resource limitation. Indeed, the current study was designed to test the top-down effects of selective grazers under no resource limitation. We found no evidence for allelopathy, nutrient or light limitation as the growth rate of either Microcystis or Cryptomonas was unaffected by the presence or proportion of their competitor during the experiment and growth rates were constant or increasing for both phytoplankton in the absence of grazers. The difference in the growth rates of co-cultured phytoplankton in the allelopathy experiment and the no grazer control of the facilitation experiment were likely due to differences in the experimental setup (e.g., 250 mL vs. 1 L flasks, cotton wool vs. silicone stoppers with bubbling, see methods for details). Despite such differences, there was no evidence for allelopathy or another type of resource competition that could explain the grazer effect found in the current setup. Yet, top-down grazer effects on cyanobacterial dominance in nutrient or light limited waters may be different than what the current results suggest. Allelopathy is commonly observed among cyanobacteria and eukaryotic phytoplankton, including Microcystis and Cryptomonas (B-Béres et al., 2012; Mello et al., 2012; Rzymski et al., 2014; Bittercourt-Oliveira et al., 2015; Ma et al., 2015), and nutrient limitation can induce the production of allelopathic compounds by cyanobacteria and reducing the fitness of competing phytoplankton (Sukenik et al., 2002; Xu et al., 2017). Resource limitation can also result in indirect grazer effects on phytoplankton communities such as nutrient regeneration by zooplankton excretion (Elser et al., 1988). Zooplankton excretion may alter the concentration and ratio of nutrients available to phytoplankton due to phylum or species specific differences in the elemental composition (Sterner et al., 1992). Consequently, future work would benefit from considering both bottom-up and top-down effects simultaneously to evaluate the role of selective grazers on cyanobacterial abundance across a variety of abiotic conditions.

In addition to resource competition, future work would also benefit from accounting for indirect grazer effects on phytoplankton competition by inducing chemical or morphological changes in their prey. Chemical grazer cues may result in increased toxin formation or morphological defenses in phytoplankton (Jang et al., 2007; Yang et al., 2008). Since Microcystis morphology did not change in the presence of grazers in this experiment, it is not a likely factor explaining the results. Yet it is possible that toxin production in Microcystis increased due to grazing (Jang et al., 2007). Although the role of grazer induced chemical defenses on the ecology of phytoplankton is poorly understood (Pohnert et al., 2007), increased production of microcystins reduced copepod grazing pressure on Microcystis in a recent study (Ger et al., 2016a). Thus, future studies would benefit from accounting for chemical (i.e., stoichiometry, cellular and dissolved toxins) and morphological changes when studying the effect of zooplankton on phytoplankton competition.

Our results emphasize the potential for top-down regulation of phytoplankton communities by selective grazers in warmer and more eutrophic waters. Given that the bulk of freshwater zooplankton work is based on large generalist grazers (i.e., Daphnia), which become more rare in warmer and more eutrophic water bodies (Elser and Goldman, 1991; Fernando, 1994; Auer et al., 2004; Jeppesen et al., 2011), the role of selective grazers on phytoplankton dynamics is still mostly unexplored. Thus, important questions remain before further predictions can be made regarding the ecological relevance of our results for a warmer and more eutrophic future world. For example, there are several distinct functional traits represented within selectively grazing zooplankton (e.g., raptorial feeders, feeding current, cruise feeding) (Litchman et al., 2013). Each group of selectively feeding zooplankton is further differentiated in terms of their nutritional requirements and tolerance to ingested cyanotoxins (Litchman et al., 2013; Ger et al., 2016c). Moreover, variability in cyanobacterial traits, including toxin production and morphology, is equally large (Kruk et al., 2010; O'Neil et al., 2012), which further limits predictions based on the results of this study. Hence, while the current results suggest a strong positive effect of selective grazing zooplankton on the dominance of Microcystis, we urge future work to account for different functional groups within zooplankton and cyanobacteria—especially those that co-exist in nature—when evaluating the top-down regulation of cyanobacterial blooms in a more eutrophic world.

In conclusion, our results suggest that selective avoidance of cyanobacteria by zooplankton grazers is likely an important mechanism regulating the dynamics and dominance of cyanobacterial blooms and phytoplankton community structure in general, at least in nutrient rich waters where bloom-forming cyanobacteria typically occur. Since smaller and more selective grazing zooplankton dominate in eutrophic waters, the potential for grazer facilitation of blooms may operate as a positive feedback that stabilizes blooms. Bloom facilitation by grazers may be a major factor explaining the association of high zooplankton biomass during long duration cyanobacterial blooms, especially in warmer eutrophic waters. However, as detailed above, the large variability in the functional traits of both grazers and cyanobacteria, as well as the intermediary role of heterotrophic protists and parasites as both grazers of cyanobacteria and prey to larger crustacean zooplankton highlight the need for considering more species interactions. Ultimately, while selective grazers may facilitate blooms, what sustains their elevated biomass during blooms of toxic phytoplankton will determine the structure and function of bloom-dominated waters (Ger et al., 2014).

All authors designed the experiment and analyzed the data. All authors contributed to writing the experiment. EL and KG performed the experiment.