Eutrophication of freshwater ecosystems, driven primarily by over enrichment of nitrogen (N) and phosphorus (P) (Carpenter et al., 1998), is a serious threat to water quality, biodiversity and other key ecosystem functions (Smith et al., 2006; Cook et al., 2018; Liu et al., 2021). Nutrient enrichment promotes the appearance and persistence of harmful algal blooms (Heisler et al., 2008; Conley et al., 2009) and the decline of submerged macrophytes (SMs) (Zhang et al., 2017), altering the food web structure (Fujibayashi et al., 2018; Briland et al., 2020). Occurrences of eutrophication are expected to increase with climate and land-use changes (Jeppesen et al., 2010; Bergström and Karlsson, 2019; Le Moal et al., 2019), inducing regime switches from a macrophyte‐dominated clear state to phytoplankton‐dominated turbid state (Jeppesen et al., 2007a). Considering that human activity is the primary cause of the eutrophication, it is crucial to reduce anthropogenic contributions to aquatic ecosystems and to find effective approaches to control cyanobacterial blooms that usually dominate eutrophic waterbodies.

The restoration of SMs is considered a crucial measure for the rehabilitation of shallow eutrophic lakes (Liu et al., 2020; Li et al., 2021a), as SMs display certain functional traits, that they use to stabilize the clear-water state (Puijalon et al., 2011; Su et al., 2019; Rao et al., 2020). For example, SMs could suppress algal growth by competing for light and nutrients (Lürling et al., 2006), producing algae-inhibiting allelochemicals to interfere the photosynthetic activities (Zhu et al., 2010) and change other physiological and biochemical processes (Zhu et al., 2021; He et al., 2023), and providing grazing zooplankton with a daytime refuge against fish predation (Burks et al., 2001). In addition, SMs can facilitate nutrient uptake from the water column and sediment (Sand-Jensen and Borum, 1991) and reduce sediment resuspension (Horppila and Nurminen, 2003). Earlier studies involving small-scale experiments (Barrow et al., 2019; Amorim and Moura, 2020) and natural aquatic ecosystems (Chao et al., 2022; Peng et al., 2022) have repeatedly reported that the restoration of SMs decreases the phytoplankton abundance and increases water clarity. Thus, usage of SMs is a prospective tool for the elimination of algal blooms (Jeppesen et al., 2007b). The submerged macrophyte restoration is, therefore, expected to prevent or mitigate the expansion of cyanobacterial blooms.

Another restoration technique to improve water quality is the biomanipulation of filter-feeding freshwater animals, such as mussels; however, its effectiveness remains debatable. For example, grazing studies involving filter-feeding mussels, such as zebra mussels and triangle sail mussels, in Europe and China demonstrated that they can efficiently consume pelagic algae and detritus (e.g. MacIsaac et al., 1992; Gao et al., 2017). Further, mussels, as grazers, can reduce or even prevent algal blooms (Gulati et al., 2008). Furthermore, Wu and Culver (1991) found abundant zebra mussels in Lake Erie, and noticed that filter-feeding Daphnia were able to reduce edible algal density and enhance water transparency. Interestingly, some mussels display food selectivity and avoid consuming cyanobacteria resulting in dominance of cyanobacteria over other forms (Hwang et al., 2004; Colvin et al., 2015). Contrary to Hwang et al. (2004) and Baker et al. (1998); Colvin et al. (2015) reported that the invasion of zebra mussels led to a decline in Microcystis biomass in the Hudson River.

Numerous combined technologies for controlling lake eutrophication have been developed, demonstrating that the combined effect of the two technologies was better than the technology alone. For example, the combination of large herbivorous zooplankton and submerged macrophytes proved to be more efficient at controlling the biomass of cyanobacteria (Amorim and Moura, 2020). In addition, the successful restoration of submersed macrophytes improved the water quality in a eutrophic lake after the removal of common carp (Knopik and Newman, 2018). Despite recent advances on biological restoration methods related to eutrophication, little is known about whether and how the combined manipulation of filter-feeding bivalves and SMs control eutrophication and influence phytoplankton assemblages. Given the complexities of climatically, thermally, ecologically, and hydrologically induced change in natural lakes (Richardson et al., 2019), mesocosm studies have been heralded as a useful means to investigate the effects of multiple factors under manipulated or controlled environmental conditions while supporting realistic levels of biocomplexity (Stewart et al., 2013; Fordham, 2015).

In this study, we designed a 32-day nutrient-enriched freshwater mesocosm experiment to explore the potential interactions between the filter-feeding bivalves and SMs and their impact on the biomass and composition of phytoplankton assemblages. We reasoned that categorizing cyanobacteria based on the adaptations to avoid predation (e.g. colonial and filamentous cyanobacteria) may lead to greater insights into the combined effects of filter-feeding bivalves and SMs on the restoration of eutrophic water bodies. We hypothesized that: (i) biomanipulation via addition of filter-feeding bivalves and restoration of submerged macrophyte, under nutrient enrichment, will likely affect phytoplankton assemblages and control the growth of cyanobacteria; (ii) the interactive effects are likely to be superior to either alone for controlling eutrophication.

The development of SMs is considered as an important restoration strategy in eutrophic shallow lakes. Often restoration experiments do not capture the intricacies related to increased nutrient loading and the amount of filter-feeding animals, instead they primarily focus on assessing the effects of SMs (e.g. Bakker et al., 2013). The use of an experimental mesocosm approach is important for investigating the complexity observed in the field and to gain a mechanistic understanding about the single and interactive effects of multiple restoration measures (Amorim and Moura, 2020; Zhang et al., 2021; Boucher-Carrier et al., 2022). During the loading experiments we noticed that restoration of eutrophic waterbodies by manipulation, i.e., addition of filter-feeding bivalves and SMs, had a marked influence on water quality and algal biomass and community composition.

Our results of TP and TN reduction in the restoration treatments demonstrated that the addition of the filter-feeding bivalve C. plicata and the recovery of submerged macrophyte can significantly alleviate eutrophication. Although some species of rooted SMs are sensitive to relatively high nutrients and consequently get suppressed under eutrophic conditions (Søndergaard et al., 2010), in our study, the use of H. verticillata reduced the nutrient levels. As a rooted submerged macrophyte, H. verticillata can obtain nutrients from sediments via root uptake and from the water column via foliar uptake (Barko, 1982), thus acting as a major nutrient sink. Moreover, filter-feeding bivalves transfer nutrients (especially P) from the water column to the bottom, through excretion as well as biodeposition of faeces and pseudofaeces (Vaughn and Hoellein, 2018).

As expected, changes in nutrient concentrations and stoichiometry, which were influenced by filter-feeding bivalves and SMs, may have altered the phytoplankton assemblage composition during the mesocosm experiment. In agreement, a previous study observed suppressed Cyanobacterial taxa in lakes under P limitation (Havens et al., 2003). At the end of the experiment, the relatively high TN: TP ratios in the macrophyte (average of 37: 1) and the bivalve + macrophyte (average of 42: 1) treatments likely led to the competitive exclusion of cyanobacteria. These results concur with a previous study by Smith (1983) who noticed suppressed cyanobacterial blooms when the TN: TP ratio exceeded 29 to 1. Algae that are incapable of nitrogen fixation are reported to dominate under P-limited conditions (Schindler, 1977; Amano et al., 2010).

High aquatic N: P ratios in lakes are reported in agricultural regions (Arbuckle and Downing, 2001), suggesting that P is the principal production-limiting nutrient. In our study, relatively high TN: TP ratios (≈29) at the beginning indicated a phosphorus-limitation situation. During the loading experiments addition of N and P close to the Redfield ratio (N: P of 16: 1), relieves nutrient limitation, and promotes a much higher phytoplankton biomass development. Phytoplankton biomass showed a gradual increase under nutrient enrichment (the control treatment). Previous studies have reported a linear relationship between chlorophyll a concentration and TP for the lower nutrient ranges (TP < 5–100µgL^-1) and asymptotic behavior at higher ranges (TP >100 µgL^-1) (Canfield et al., 1984; Phillips et al., 2008; Borics et al., 2013). Noticeably, at the end of the experiment, TP concentration exceeded 100 µgL^-1, we noticed an increase in phytoplankton biomass with nutrient enrichment, suggesting nutrients control phytoplankton biomass in nutrient‐rich waters (Richardson et al., 2019).

Even in a nutrient-enrichment scenario, as expected, responses to the addition of filter-feeding bivalves (Cristaria plicata) alone contributed to the decline in cyanobacterial and total phytoplankton biomass. The overall decline in cyanobacterial and phytoplankton biomass in these mesocosm can be explained by the direct grazing impacts of the filter-feeding bivalves (Gulati et al., 2008). The direct effects of grazing by C. plicata led to statistically significant decrease in filamentous or colonial taxa such as the genera Dolichospermum (formerly Anabaena), Microcystis, and Planktothrix (Bastviken et al., 1998; Dionisio Pires et al., 2005) which usually form harmful algal blooms. These results are in line with previous studies indicating that cyanobacteria are directly grazed by filter-feeding bivalves (Hwang et al., 2004; Gao et al., 2017), and support the possibility that addition of filter-feeding bivalves will alleviate algal blooms that are associated with eutrophication. Further, several studies have also shown that filter-feeding bivalves are selective feeders and the filtering rate can vary depending on food particle size and bivalve species. Some functional traits of cyanobacteria, including cell size, life form (e.g. single celled, colonial, or filamentous), nutritional deficiency, and toxin production, can prevent them from being grazed by mussels (White and Sarnelle, 2014; Boegehold et al., 2019). For example, Corbicula fluminea selectively filtered particles in the range of 0.2–2 µm (Rong et al., 2021), Dreissena Polymorpha preferred food particles from 5 to 40 μm (Sprung & Rose, 1988), and Venerupis corrugatus filtered out particles of 5 to 13 μm (Stenton-Dozey and Brown, 1992). Although we did not consider selective grazing effects of C. plicata on phytoplankton the composition of phytoplankton assemblages significantly differed between the control and the bivalve treatments at the end of the experiment. This result indicates that once nutrient limitation is alleviated, selective grazing would likely be the main factor affecting the structure of the phytoplankton assemblage. Additionally, the phytoplankton assemblages in the filter-feeding bivalve addition treatments appear to have adaptive responses to selective grazing pressure by C. plicata, as larger diatoms (e.g. Cymbella, Fragilaria, Thalassiosira) dominated the phytoplankton assemblages.

SMs can also suppress algal growth via allelopathic controls and nutrient competition (van Donk and van de Bund, 2002; Mohamed, 2017; Zhu et al., 2021). However, in our experiments, at the end of the experiment a slight, but not significant increase in overall phytoplankton assemblage biomass (chlorophyll a) was noticed in the Macrophyte treatment. The phytoplankton blooms occurred when SMs were absent over the course of the nutrient loading experiment (the control treatment). While phytoplankton biomass initially decreased from day 0 to day 4, it then began to slightly increase until day 32, suggesting that SMs did cause a reduction in phytoplankton biomass. The submerged macrophyte H. verticillate, has been reported to produce and release allelochemicals that has inhibitory effects on Chlorella cell membrane (Zhang et al., 2012) and cyanobacteria (Wang et al., 2006; Gao et al., 2011). Over 32 days, SMs suppressed overall cyanobacteria biomass ( Table 2 ), concurring with findings from other experimental studies (Lürling et al., 2006; Barrow et al., 2019; Amorim and Moura, 2020). Further, factors such as light and nutrient competition likely interacted with allelopathic controls in the mesocosms and led to the competitive exclusion of cyanobacteria.

Although unexpected, we found that filter-feeding bivalves in combination with SMs reduced the biomass of cyanobacteria, and noticeably the effect size of this interaction was less than the sum of their individual effects (i.e., an antagonistic interaction). This result is in line with the widely observed antagonistic interactions in freshwater ecosystems (Jackson et al., 2016; Segurado et al., 2018; He et al., 2021), such as nutrient-pesticide effects on benthic invertebrate richness (Chará-Serna et al., 2019), and fish-shrimp effects on zooplankton biomass (He et al., 2021). The mechanism for the antagonistic interactions is largely unknown, however, a possible explanation could be the asymmetry of mean effect size. In our study, the larger magnitude of the submerged macrophyte restoration may have overridden the effect of stock filter-feeding bivalves ( Table 2 ), thereby negating its contribution to their net impact on the overall biomass of cyanobacteria (Sala et al., 2000; Barrow et al., 2019). An antagonism between filter-feeding bivalves and SMs was only detected for total cyanobacteria. For filamentous or colonial taxa, there was no significant interactive effects; rather filamentous cyanobacteria were more sensitive to macrophytes and colonial cyanobacteria to bivalves. The differential sensitivity of cyanobacterial taxa to different biomanipulation approaches has been noticed previously (Gazulha et al., 2012; Amorim and Moura, 2020), the reason being that cyanobacteria are a diverse and morphologically complex group of prokaryotes with different key ecological traits thus eliciting disparate responses (Mantzouki et al., 2016; Rangel et al., 2020).

Our first hypothesis that manipulation via addition of filter-feeding bivalves and restoration of submerged macrophyte will likely affect phytoplankton assemblages was confirmed, as this manipulation efficiently decreased water nutrient concentrations and the overall phytoplankton biomass. Since, phytoplankton were dominated by symmetrical desmids (e.g., Cosmarium, Micrasterias) and the large diatoms (e.g., Cymbella, Fragilaria, Thalassiosira) in the restoration treatments, with the competitive exclusion of cyanobacteria, our results also supported that manipulation could control the growth of cyanobacteria. Contrary to the second hypothesis, an antagonism between filter-feeding bivalves and SMs was detected but only for total cyanobacteria, demonstrating that the larger magnitude of the submerged macrophyte restoration may override the effect of stock filter-feeding bivalves. However, we should also noticed that the addition of bivalves combined with SMs was more efficient at decreasing nutrient concentrations than the isolated addition of bivalves, and at controlling total algal biomass than the isolated restoration of SMs. Overall, our results suggest that manipulation, through introduction of the stock of bivalves as grazers, associated with the restoration of SMs, is an efficient approach for reducing cyanobacterial blooms and alleviating eutrophication.

The original contributions presented in the study are included in the article/ Supplementary Material . Further inquiries can be directed to the corresponding author.

XD: Conceptualization, Methodology, Investigation, Formal analysis, Writing – original draft, Writing – review & editing, Funding acquisition. DS: Conceptualization, Investigation, Formal analysis, Writing – original draft, Writing – review & editing. HW: Investigation. JY: Resources. HL: Writing – original draft, Investigation. TH: Funding acquisition. All authors contributed to the article and approved the submitted version.