Warming and Predation Alter Consumer Coexistence

Understanding how species coexist is a key question in ecology, with implications for conserving biodiversity. Species coexistence is affected by multiple factors, including climate warming and the presence of predators, yet most studies consider these ‘stressors’ in isolation. The interactive effects of warming and predation on species coexistence have not been explored. Here, we constructed simplied shallow lake food webs in 24 large mesocosm ponds and applied a crossed design to simulate warming (+4.5°C) and predation (by crucian carp Carassius auratus ). We monitored population responses of two common snails, Bellamya aeruginosa and Radix swinhoei over 200 days. We predicted that warming will alter the phenology of the snails, and that this will combine with predation to reduce their abundance. Warming advanced the reproduction of R. swinhoei by 21.5 days and reduced biomass and density of both snails. The advance in R. swinhoei phenology suppressed the slower growing B. aeruginosa population, reducing coexistence. Warming and predation combined in an additive manner to decrease the biomass of B. aeruginosa . In contrast, the two drivers synergistically decreased the biomass and density of R. swinhoe i because of enhanced pressure from the sh on their preferred snail prey under warming, due to rising metabolic demands. Therefore, snail coexistence was further reduced with sh presence under warming. In conclusion, future continuous warming and strong predation might undermine the coexistence of the gastropods, thereby changing food web structures in shallow freshwater ecosystems.


Introduction
Understanding factors and processes that shape species coexistence is a central question in community ecology. The framework of modern coexistence theory has been widely used to clarify the conditions for species coexistence (Barabas et al. 2018;Chesson 2000Chesson , 2018Pande et al. 2020), which points out the mechanisms of promoting species coexistence via minimizing average tness differences (equalizing) and increasing niche differences (stabilizing) (Letten et al. 2017). Without niche differences, the species with higher average tness will eventually exclude the others. However, all species coexistences occur in changing environments, temporally and spatially, as environmental variation promotes species coexistence (Chesson 1994; Kuang and Chesson 2008). To coexist, different species might respond nonlinearly to uctuating limiting factors, or via storage effects, such as resource partitioning (Levine and HilleRisLambers 2009), differential vulnerability to predators  or pathogens (Bagchi et al. 2014), occupation of different microhabitats (Schlägel et al. 2020;Silvertown 2004), or phenological separation (Usinowicz et al. 2017). However, our planet has been experiencing unprecedented changes due to anthropogenic activities (Steffen et al. 2015), yet it remains to be explored how this affects species coexistence and what the consequences of changes in coexistence are for entire ecosystems ).
Previous studies have demonstrated that the warming associated with climate change will alter phenology (Thackeray et  . Climate warming can also affect species coexistence directly through physiological impacts or indirectly via alterations in food web interactions (Reuman et al. 2014). Indirect cascading effects via species interactions (e. g. phenological mismatch, changing in resource availability or predation) generally have greater impacts than direct ones (Ockendon et al. 2014). Temperature plays a key role because warming will change the vital rates of individuals, and species usually have limited thermal tolerance (Rohr et al. 2018; Sunday et al. 2012). There are two key mechanisms by which warming will alter coexistence: (i) shifting phenology and (ii) increasing metabolic demands. The rst predicts that populations will differ in their temporal responses to warming: acclimation capacities to the changing temperatures will differ among (Pinsky et al. 2019;Rohr et al. 2018) and within (Dahlke et al. 2020) species. For instance, stenothermal species might be more sensitive to warming than eurythermal species (Dahlke et al. 2020;Pörtner et al. 2005), and early-season species might shift more in time than laterseason species (Wolkovich et al. 2012). Therefore, warming might increase the phenological mismatch between coexisting species, thus reducing the degree of competition strength between them (Chmura et al. 2019) (Fig. 1, mechanism i). The latter mechanism refers to the greater metabolic demands which individuals experience at higher temperatures.
For ectothermic herbivores which share the same resource, warming should enhance competition, because ingestion rates of heterotrophs generally increase more rapidly with rising temperature than the . Therefore, effects of warming on coexistence of two competing species might not only be caused by increasing phenological mismatch, but also via enhancing competition on basal resource (Fig. 1, mechanism ii).
Predators are generally assumed to have negative effects on prey abundance, and the strength of this effect can differ greatly among prey species (Karakoc et al. 2020;Paine 1966). Coexistence of competing prey species will be promoted when the predators are either specialists on the more competitive species, or generalists via density or frequency-dependent predation (Ishii and Shimada 2012; Karakoc et al. 2020;Saleem et al. 2012). However, studies also showed that predation can undermine coexistence if it is particular strong (Bonsall and Hassell 1997;Chase et al. 2002), or if the rarer species is ine cient in resource consumption (Holt 1977). Eventually, species with the highest tolerance of predation will outcompete the others. Therefore, effects of predation on coexistence of competing species might be caused by the selective feeding by the predator (Fig. 1, mechanism iii).
Warming can affect the phenology of both predators and prey with implications for the effect of the predator on prey abundance ). Furthermore, warming will elevate the metabolic demands of predators which might strengthen the predator-prey interactions (Thakur et al. 2018;Thakur et al. 2017) and undermine prey coexistence . Therefore, in addition to shifts in phenology and selective feeding, warming and predation might interactively enhance the pressure of a predator on its prey (Fig. 1, mechanism iv), with implications for prey coexistence.
Here we test the predictions that (1) Warming will alter the phenology and competition of two primary consumers; (2) Predation will reduce the abundance of the primary consumers; (3) Warming and predation will interactively increase the competition between primary consumers, reducing coexistence.
To test these hypotheses, we used freshwater gastropods and their sh predators. Gastropods play key roles in aquatic ecosystems by contributing to nutrient cycling and water quality, particularly due to their role as algal grazers (Böhm et al. 2020;Strong et al. 2008). However, to our knowledge, no studies have investigated how warming and predation interact to alter their coexistence.

Study species
The targeted two gastropods Bellamya aeruginosa and Radix swinhoei are widespread in China, and commonly found to coexist in shallow aquatic ecosystems (Cai et al. 2018). B. aeruginosa is a branchiate gastropod (Gu et al. 2015), which feeds on sediment organic detritus, algae and bacteria (Liu et al. 1993). B. aeruginosa is a food resource for some sh (e.g. black carp, Mylopharyngodon piceus) and is also consumed by humans (Ma et al. 2010). The species is gonochoristic and ovoviviparous. The minimum temperature for the breeding of the snail is 16-18°C, and the optimal temperature is around 26°C (Chen and Song 1975). The reproductive cycle is about 6 months, and the snails can live for a few years. Adult females are gravid all year round and can release their newborn continuously in the breeding season (Ma et al. 2010).
R. swinhoei is a common pulmonated gastropod, which prefers habitat with macrophytes, and feeds on detritus, macrophytes, algae and even dead animal bodies (Li et al. 2005;Qi 1998; Zhi et al. 2020). It is hermaphrodite and allogamous, and it can breed multiple times a year (Qi 1998). Radix genus are generally cold-adapted gastropods (Pfenninger et al. 2006). Such as R. balthica, which prefers an optimum temperature of 16-20°C, and fails to reproduce over 24°C (Johansson et  We used a fully-factorial design with two factors crossed -warming and predation -resulting in four treatments: Control (C), normal temperature with no sh; warming treatment (W), + 4.5°C higher than the temperature in the control; predation treatment (F), with a 60 g common carp (Cyprinus carpio); and warming and predation treatment (WF). The temperature increase + 4.5°C was chosen to simulate the possible warming that could happen in this region at the end of this century (IPCC 2014), and temperature was manipulated automatically using two temperature sensors and a heating element controlled by a computer (Wang et al. 2020). Common carp is a typical omnivorous species which coexist with the two snails in water bodies in China, and the biomass of the sh in the mesocosms (equal to 340 kg ha − 1 ) is within the range found in the lakes at this region (Gao et  inserted into the sediment. The tube was taken out once a week, collected all snails, blotted dry, counted and weighted the snails to 0.0001g, and then released them back to the mesocosms.

Analyses
To analyze effects of warming, sh predation and their interaction on snail biomass, water quality, Chl. a concentrations in the water column and periphyton, multiple linear mixed effect models (LME) were deployed, with sampling date as a random factor. These dependent variables were log-transformed to increase the normality of the distribution. Normality and homoscedasticity of the dependent variables were assessed visually by plotting model residuals versus tted values and in quantile-quantile plots of the model residuals. As the densities of both snails were count data, and as such a Poisson link function was applied to the generalized linear mixed effect model (GLMM). The effects of warming and predation were considered additive if models excluding their interaction term had lower AIC values. For the nonadditive interactions, interaction effect sizes (F-values) were subtracted from the sum of the main effect sizes producing values representing the difference between additive and interactive effects, following previous studies (Jackson et al. 2020; Lakeman-Fraser and Ewers 2014). If the value was positive, then the relationship was considered to be antagonistic (the effect of the drivers were reduced when acting interactively); if however, the value was negative, then the relationship was considered to be synergistic (the effect of the drivers were increased when acting interactively) (Table S1). All linear mixed models were performed using the lme4 package (Bates et al. 2015).
To assess the distribution patterns of the biomass and density of the snails over time, data from each mesocosm were tted with a six parameters Weibull function using the cardidates package, which is designed for peak curve tting (Rolinski et al. 2007). Not all peak curve ttings were trustable. Only the results of the biomass and density of R. swinhoei and some of B. aeruginosa in the control and warming treatments without sh present tted well, as the tted results for the rest did not match their trends well (Table S2). The heights and timing of peaks between control and warming treatments of R. swinhoei were compared using independent t-tests. Normality and homogeneity of variance were checked by shapiro test and F-test, respectively. To indicate the relative competition between the two snails in all treatments, normalized data (data were divided by the maximum value of each snail in each mesocosm to diminish their different scales) of the two snails were plotted together. All statistics were performed in R version 3.6.2 (Team 2019).

Treatment manipulation
Daily averaged temperature was 4.25 ± 0.42 (n = 223, total number of days, mean ± sd) higher in the warmed mesocosms (Fig. 2a). Warming decreased the concentration of Chl. a in the water column, but increased concentration in periphyton samples (Table 1). Adding sh predators to the systems signi cantly increased turbidity, and concentrations of Chl. a in the water column and periphyton (Table 1 and Fig. 2b,c,d). In the warming treatments, adding sh predators increased the turbidity more than warming alone. Warming and sh predation also interactively affected the Chl. a concentration in the water column (Table 1 and Fig. 2c). Data of TN and TP can be found in Fig. S1. No signi cant increase of biomass of sh was found at the end of the experiment.

Snail response
In all treatments, snail biomass was dominated by B. aeruginosa. Snail density was dominated by R. swinhoei, but only when no sh were present (Fig. 3). Both warming and predation signi cantly decreased the biomass and density of the two snails, and the effects of predation was stronger than that of warming (Table 1 and Fig. 3). The combined effects of warming and predation were additive for the biomass of B. aeruginosa (AIC value was smaller in the model without interactive factor, Table S1), while the effects were antagonistic for density of B. aeruginosa (∆F = 20.4). The effects of warming were ampli ed under sh predation on both biomass (∆F = -4.0) and density (∆F = -220.4) of R. swinhoei (Table S1).
The average timing of peak density in R. swinhoei was 21.5 days earlier in the warming treatment (without sh) than in the control (Welch t-test, Table S3). However, the peak densities of R. swinhoei were not signi cantly different between warming treatment and control. The peak biomass and average time of peak biomass of R. swinhoei did not differ signi cantly under control and warming treatment without sh (Table S3). The time of peak biomass and density of B. aeruginosa varied largely within treatments, and the variations increased with warming (without sh) compared to control, some were advanced (n = 4) and one was delayed (n = 1) (Table S2 and Fig. S2). No clear peaks of the two snails could be detected in the treatments with sh present.
The overlaps between the two snails shifted with warming and predation. Peaks of biomass and density of B. aeruginosa were found reduced and shifted under warming without sh present, and clear pits were observed in biomass and density of B. aeruginosa when these peaks reached for R. swinhoei in the warming treatment without sh (Fig. S3). In the ambient treatment with sh present, peaks of biomass and density of B. aeruginosa were advanced, while the biomass and density of R. swinhoei increased over time and the highest values occurred at the end of the experiment. In the warming treatment with sh present, the biomass and density of B. aeruginosa decreased over time, and the highest values occurred at the beginning of the experiment, while the biomass and density of R. swinhoei were always low and increased gradually (Fig. S3). These imply that both snails were suppressed more under warming with sh present, particularly for R. swinhoei.

Discussion
Our results show that warming and sh predation interact to alter the phenology and demography of aquatic snails. The biomass and density of both snails declined with warming and adult B. aeruginosa dominated in the sh predation treatment. Warming and sh predation combined in an additive manner to decrease the biomass of B. aeruginosa, while its density declined less than expected (i. e. there was an antagonistic interaction between the stressors). Warming and sh predation interacted to amplify their independent effects on the biomass and density of R. swinhoei, causing a decline larger than expected by an additive response (i. e. a synergistic interaction). Both warming and predation induced higher periphyton biomass due to a trophic cascade. We discuss the underlying mechanisms and implications for freshwater ecosystems below.
Warming effects on coexistence of snails In our study, warming signi cantly advanced the peak of reproduction of R. swinhoei by 21. sh (Tao et al. 2018), with the phenology of primary consumers generally advanced more than other trophic levels (Thackeray et al. 2016). In our study, the advanced days of reproduction for R. swinhoei was almost the same amount of days earlier in warming treatments than in the ambient when temperature reached > 24°C. Species in this genus (Radix spp.) are normally cold-adapted species (Pfenninger et al. 2006) and, therefore, when temperature reached > 24°C, the snails fail to reproduce (Johansson et al. 2016;Johansson and Laurila 2017). In contrast, no consistent peaks of biomass and density could be detected for B. aeruginosa, which could be attributed to the different life-history traits of the snail. B. aeruginosa has a much higher upper-limited thermal tolerance, as its optimal reproduction temperature is ~ 26°C (Chen and Song 1975), and the snail has a life span which can last for a few years.
Furthermore, the female can be gravid all year round and release offspring continuously in the breeding season (Ma et al. 2010). Field investigations have also found multiple different peaks of abundance and biomass for B. aeruginosa throughout the year (Chen 1987;Gong et al. 2009;Yan et al. 2000). However, the time of peak biomass and density of B. aeruginosa did show differences between the ambient and warming treatment without sh present, which indicates that warming still altered the phenology of the snail, while the effect was inconsistent, could either be advanced or delayed. This might be because the effects of warming on the snail were different among ages.
Warming decreased the biomass and density of both snails, and there are two key reasons which can explain these changes linked to phenology and feeding rates. Firstly, warming altered the phenology of both snails. R. swinhoei is hermaphrodite and its reproductive cycle can be completed in eight weeks, whereas B. aeruginosa is gonochoristic and ovoviviparous, and its reproductive cycle is about 6 months (Ma et al. 2010). During the growing season of R. swinhoei, this species can reproduce faster and dominate in abundance, while the B. aeruginosa will respond slower to rising temperatures from spring to summer. In the warming treatments, the advancement of reproduction by R. swinhoei strongly suppress the population of B. aeruginosa, as clear pits were observed in biomass and density of B. aeruginosa when these peaks reached for R. swinhoei in the warming treatment (Fig. S3). This might be because that R. swinhoei depleted algal food resources in the growing season, thus reducing fecundity of B. aeruginosa. In our study, the standing biomasses of periphyton were very low in both ambient and warming treatment in the early growing season. These indicate that a strong grazing pressure on periphyton by the snails in the early growing season. Furthermore, the phenology of periphyton does not advance at the same rate as consumers with warming (Thackeray et al. 2016) and, therefore, there are less resources available when the snails reproduce in a warmer world. The competition between the two snails was strong, and R. swinhoei was more competitive in the growing season. In addition, top-down pressure might be enhanced with warming (Schaum et al. 2018), which means that an even stronger competition could happen between the two snails due to limited food source. Therefore, snail density and biomass both decreased due to warming.
Predation effects on coexistence of snails As expected, the biomass and density of the two snails were suppressed in the presence of sh (Karakoc et al. 2020;Paine 1966). The predation effects were particularly strong on R. swinhoei and the juveniles of both snails. This is because crucian carp can only prey on snails smaller than their jaw gape, hence, they substantially eliminated the juveniles of both snails. Furthermore, the adult individual biomass was larger in B. aeruginosa (can reach 5.77 g) than in R. swinhoei (hardly larger than 1.5 g), and shell toughness was much higher in B. aeruginosa than in Radix spp. snail (Zhu et  Therefore, we observed adult B. aeruginosa dominated in the predation treatment, particularly in the beginning, since R. swinhoei was preferentially consumed by the sh. These indicate that predation increased the relative competence of B. aeruginosa over R. swinhoei at the early time and the competence shifted over time. The presence of sh strongly disturbed the water promoting the growth of periphyton. We expected this higher turbidity and periphyton biomass to partly alleviate the predation effect from the sh, due to both declining visibility and increase food resources for the snail. However, in our study, the density of sh added to the system was high (Gao et al. 2014), and no such effects were observed. Under a strong predation pressure, B. aeruginosa might exclude R. swinhoei in the ecosystem eventually by undermining the balance of coexistence in the competing species (Bonsall and Hassell 1997;Chase et al. 2002).

Interactive effects of warming and predation on snails
Warming and predation combined additively to decrease the biomass of B. aeruginosa, and antagonistically to decrease its density. This suggests warming and predation combined to mitigate one another's independent effects on this species. This might be because warming increased the availability of algal resources for the snails with sh presence, which partly alleviate the predation effect from the sh. In contrast, there was a synergistic interaction between warming and predation on R. swinhoei, resulting in large declines in its biomass and density. Top-down control by predators on prey is enhanced by warming (Hansson et al. 2013). Here, the crucian carp exerted stronger pressure on R. swinhoei, the preferred food (Zhu et al. 2017) under warming. This caused a signi cant ampli ed interaction of warming and predation on the growth and reproduction of R. swinhoei. Warming has also shown to affect attack rate, handling time and predation success (Twardochleb et al. 2020). However, in our study, the prey are rather slow-moving species, the impacts of warming on the predation was mainly determined by the increased feeding rate, and the prey are unlikely to develop strongly avoidant mechanisms against predation in one or two generations. In a longer period, continuous warming and strong predation might eventually undermine the coexist of the two snail species, excluding R. swinhoei.

Implications for aquatic ecosystems
In our study, warming and predation interactively enhance the competition between the two snails, decreasing the biomass and density of both snails, resulting in higher growth of periphyton. The trophic cascading effects could have substantial impacts on aquatic food webs. With the decrease of gastropods, sh population might decrease over time (Nagelkerken and Connell 2015), but periphyton growth increase, resulting in bottom heavy trophic pyramid. This might hinder the nutrient cycle and energy transfer in the aquatic ecosystem (Nagelkerken et al. 2020). Furthermore, continuous warming and strong predation might undermine the coexistence of the two snails, eventually causing local extinction of the less dominate species (Fig. 4). The loss of key primary consumer not only decreases biodiversity, but might also weaken the stability of aquatic food webs (Ives and Cardinale 2004;Nagelkerken et al. 2020). Particularly in shallow lakes with strong anthropogenic activities, as eutrophication, chemical pollution and habitat destruction are common in these areas (Birk et al. 2020; Steffen et al. 2015). These might interactively alter the shallow aquatic ecosystem from a health state to a nuisance state, leading to loss in ecosystem functions and services (Scheffer et al. 2001). Future studies should investigate more on how warming could affect species coexistence in other model systems in longer time, and how these will further alter food web structures in ecosystems.   Daily averaged temperature, water turbidity, and Chl. a concentrations in the water column and periphyton changed over time in different treatments. In panel a, the horizontal dotted and dashed lines indicate temperature of 16°C and 24°C, respectively. When the averaged temperature started to go above 24°C, the days were 113 and 134 for the warming and ambient treatments, and these are indicated by vertical red and blue lines, respectively. Panel b, c and d share the same legend. Vertical bars are standard errors. C for control, W for warming, F for predation and WF for warming and predation. peak for the warming treatment without sh was 21 days earlier than the control. C for control, W for warming, F for predation and WF for warming and predation. Please note that the y-axis is on different scales for each species.