Physiological mortality of planktonic ciliates: Estimates, causes, and consequences

Contrasting physiological mortality with predator‐induced mortality is of tremendous importance for the population dynamics of many organisms but is difficult to assess. This is especially true for tiny organisms, such as most protists, which do not leave any carcasses behind. I performed a meta‐analysis using planktonic ciliates as model organisms to estimate the maximum physiological mortality rates (δmax) across pelagic ecosystems in relation to environmental and biotic factors. The median δmax of planktonic ciliates was 0.62 d−1 and did not differ between marine and freshwater species. Maximum ciliate mortality rates were species‐specific and affected by their specific growth rates (rmax), cell volume, and ability to encyst. Cyst‐forming species had, on average, higher δmax than species unable to encyst. Maximum mortality rates were positively related to rmax, but, in contrast to the latter, δmax appeared unaffected by temperature. I conclude that (1) in the ocean, physiological mortality is more critical for controlling ciliate population size than ciliate losses imposed by microcrustacean predation, but (2) in many lakes, the opposite holds; (3) cyst formation is an effective ciliate trait to cope with the high mortality of motile cells upon starvation. The lack of a temperature effect on δmax deserves further study; if correct, planktonic ciliates may take advantage of rising ocean and lake temperatures, with important implications for the pelagic food web.

Ciliates are significant players in virtually all planktonic food webs in the sea and inland waters (Pierce and Turner 1992;Finlay and Fenchel 1996).Like all other organisms, their population sizes are controlled bottom-up by their resource-dependent growth rates and top-down by losses imposed by predators and parasites (Coats and Bachvaroff 2013;Weisse and Montagnes 2022).A recent meta-analysis demonstrated that top-down control is common in many lakes, but predation pressure in the central parts of the ocean is too low to keep ciliate growth rates in check (Lu and Weisse 2022).In offshore regions of the sea and (ulta)oligotrophic lakes, resources are too scarce to promote fast ciliate growth.Importantly, ciliate growth rates become negative, and the populations decline if the food supply falls below a species-specific critical threshold (Lynn et al. 1987;Minter et al. 2011).Maximum mortality rates (δ max ) at zero food levels, also known as physiological mortality or physiological death rates (Weisse et al. 2023b), are the counterpart of specific growth rates (r max ) achieved under optimal, foodreplete conditions (Müller and Geller 1993;Montagnes et al. 1996).
In contrast to r max , δ max received relatively little attention in recent literature for ciliates and other planktonic organisms (Minter et al. 2011).Threshold food levels (where growth rates are zero) generally increase with temperature (Weisse et al. 2002).Therefore, rising water temperatures in many aquatic habitats caused by global warming may significantly enhance the mortality rates of marine and freshwater ciliates.Due to the ciliates' central position at the interface of the classical and the microbial food web (Worden et al. 2015;Weisse and Montagnes 2022), shrinking ciliate population sizes and, accordingly, reduced ciliate feeding and production rates may affect both the lower and higher trophic levels.Reduced ciliate production would provide less food for microzooplankton (mainly copepods and cladocerans; Montagnes et al. 2008;Weisse and Montagnes 2022).Because of trophic cascades, the net result of enhanced ciliate mortality rates and reduced ciliate grazing (everything else being equal) would be shifting toward the smallest size classes of the planktonic spectrum (Wickham et al. 2015).
From the preceding, it is evident that parameterizing ciliate physiological mortality rates and their temperature sensitivity is significant for predictive food web models (Minter et al. 2011;Davidson 2014).To this end, my primary aim was to estimate the maximum ciliate mortality rates across pelagic ecosystems (i.e., marine vs freshwater).I assumed that ciliate mortality rates are species specific.I hypothesized that (1) δ max would increase with temperature, (2) decline with ciliate cell volume, (3) be positively related to r max , and (4) would be higher in cyst-forming species than in ciliates that are unable to form resting cysts.The 1 st and 2 nd hypotheses are a corollary of the metabolic theory of ecology (Brown et al. 2004).The 3 rd hypothesis is a consequence of r/K selection predicting higher growth rates for small species and lower mortality rates for larger species (MacArthur and Wilson 1967;Pianka 1970).For ciliates, the r/K theory and the respective life strategies have only been applied for terrestrial species thus far (Lüftenegger et al. 1985).The 1 st and 4 th hypotheses were supported by experimental evidence obtained with sympatric freshwater ciliates (Weisse et al. 2023b).Similarly, following up on a recent analysis of the thermal response of planktonic ciliates (Luki c et al. 2022), I also postulated that (5) marine ciliates would have lower δ max than their freshwater counterparts.
This study is the last of three articles investigating the effects of starvation on planktonic ciliates in response to food shortage and temperature.We use ciliates as model organisms because they play a pivotal role in aquatic food webs (see above) and can easily be experimentally manipulated.The 1 st article presented the three main methods currently available to estimate ciliate mortality rates (Weisse et al. 2023a).The 2 nd article demonstrated that different traits and life strategies of contrasting sympatric freshwater ciliates affect their survival rates and temperature sensitivity under food-depleted conditions (Weisse et al. 2023b).The current work takes advantage of the results reported in the two previous, more specific articles but presents a global perspective on ciliate mortality rates in the ocean and freshwater.
Contrasting physiological mortality with predator-induced mortality is of tremendous importance for the population dynamics of many organisms but is difficult to assess.This is especially true for tiny, obligately, or primarily asexually reproducing organisms, such as most protists, which do not leave any carcasses behind.Therefore, this study may serve as a model for future work with other aquatic or terrestrial protists and small metazoans (e.g., rotifers).

Data compilation
I searched the ISI Web of Science and Google Scholar for experiments that measured growth and mortality rates of ciliates as a function of prey concentration (i.e., numerical responses [NRs]).The search terms were "growth (rate)" or "numerical response" in combination with "ciliate*" to search for NR experiments and "starvation" or "starved" in combination with "ciliate*" to search for mortality experiments.In addition, I searched the literature cited in these publications for further datasets.I considered only planktonic ciliates.
Mortality rates were estimated from two different methods, that is, NR experiments and experiments measuring the rate of decline (ROD) of ciliate cell numbers (Weisse et al. 2023a).NR parameters were estimated with a modified Michaelis-Menten equation that included the threshold prey concentration (P 0 ) as an additional parameter where r is the growth rate (d À1 ), r max the maximum or specific growth rate (d À1 ), P is the prey concentration (prey mL À1 ), P 0 is the threshold prey concentration where r is zero (prey mL À1 ); k is a constant (prey mL À1 ).Maximum mortality rates (δ max , d À1 ) were calculated from the above equation, setting P to zero.Since mortality represents negative growth rates (i.e., the slope of the least-squares linear regression in ROD studies and r at subthreshold food levels in NR experiments [Eq.1] are negative, Weisse et al. 2023a), I multiplied δ max by À1 to report positive values in the following.Therefore, negative estimates of δ max (after multiplication by À1) were discarded from the analysis because they represented positive mortality rates, which is impossible.
If conversion of units was needed for ciliate cell volume, I used the cell carbon to volume ratio provided by Menden-Deuer and Lessard (2000) (pg C cell À1 = 0.216 Â volume 0.939 ).I converted lorica volume to cell volume for tintinnids, assuming that cells occupied 30% of the lorica (Gilron and Lynn 1989a;Rychert 2011).When neither volume, dimensions, or carbon content were reported, I searched the dataset or secondary literature for cell volume of the same species, using preferentially publications by the same authors.
When studies did not report all parameters of the NR curve, the data were extracted from figures with DataThief III or WebPlotDigitizer (Version 4.6) and fitted using Eq. 1.Similarly, when studies used a Michaelis-Menten equation without threshold prey concentration to estimate NR parameters, I extracted the data from the publication and fitted Eq. 1 for consistency with the remaining dataset.
Mortality rates obtained by ROD experiments used the δ max reported in the respective study or calculated δ max from the maximum rate of decline after digitizing the data from the original curves, as described above.

Statistical analyses
I analyzed the dataset with the R Statistical Software (v 4.0.5, R Core Team 2021).The packages "lme4" (Bates et al. 2015) and lmerTest (Kuznetsova et al. 2017) were used to compute linear mixed models (MEM) and ANOVA procedures.The package AICcmodavg (Mazerolle 2023) was applied for model selection based on Akaike's information criterion corrected for small sample size (AICc).With the package MuMIn (Barton 2022), I calculated marginal and conditional R 2 (pseudo-R 2 ); the "r.squaredGLMM function" returns the percent variance explained by the combined fixed effects and the variance explained by the entire model, including both fixed and random effects (Nakagawa and Schielzeth 2013).Outliers were identified with the package "outliers" (Komsta 2022).Because the raw data were not normally distributed, ciliate volume, δ max , and r max were ln-transformed prior to analyses.
I first tested the effect of species on mortality rates using a linear model (LM).Because I was primarily interested in the environmental impact (i.e., temperature, habitat) and biotic effects (ciliate growth rates, volume, and cyst formation) on δ max , I discarded "species" from the following analyses.However, I included "study" as a random effect to account for the nonindependence of data points from the same study or species.I also included "method" as a fixed factor because two different methods were used to estimate ciliate mortality rates.I used the "visreg" function (package visreg; Breheny and Burchett 2017) to construct partial linear regressions between the dependent variable (δ max ) and the three predictor variables.

Results
The maximum mortality rates (δ max ) reported (see Data Availability Statement, below) were obtained from 41 studies investigating 56 species or strains in 81 NR experiments and 19 ROD experiments.I discarded NR studies that yielded unrealistic model parameters, that is, negative estimates of δ max (if P 0 > k) or because δ max could not be calculated from Eq. 1 (i.e., if P 0 ≥ k or P 0 = 0).Six other NR studies (with δ max > 6.2 d À1 ) and one ROD study (with r max > 4.7 d À1 ) were identified as outliers.The final dataset (n = 77) included 37 studies and 48 species.Most data (61) were obtained from NR experiments, compared to 16 from ROD studies.For the variables habitat (FW = 41, marine = 36) and cyst formation (42 species encysted, 35 did not), the distribution within the two levels of the individual factors was more balanced (Table 1).I did not consider the nutritional mode of ciliates in the analyses because classification into broad categories such as omnivorous, algivorous, and bactivorous is vague and mixotrophic ciliates represent a suite of different functional groups (Mitra et al. 2016).The food organisms used for the NR experiments are listed in the dataset (see Supporting Information).The maximum mortality rates (δ max ) selected for the final dataset ranged from 0.10 to 6.07 d À1 (Table 1).The median mortality was 0.62 d À1 , that is, lower than r max (median = 0.99 d À1 ).After ln-transformation, the normally distributed mean of r max was significantly higher than that of δ max ( p = 0.04, Welch's t-test).
I first tested the effect of species on mortality rates.Species as fixed factor accounted for 48% of the variance in δ max (adj.R 2 = 0.483).Species was not included as factor in the following analyses because it comprised too many levels for ANOVA procedures.
I then fitted a full linear mixed model with δ max as the dependent variable and r max , ciliate volume, temperature, method, habitat, and cyst formation as fixed effects and study as a random effect (see Supporting Information Table S1).The variance within a study was higher than the variance between studies (residual variance); the fixed effects explained 19% of the variance in ciliate mortality rates and 64% if the random factor was included (pseudo-R 2 = 0.642).
Model selection using the AICc criterion (see Supporting Information Table S2) yielded the final LM with r max , ciliate volume and cyst formation as fixed factors (Table 2).Cyst formation and r max affected ciliate mortality significantly; the effect of volume was marginally significant (p = 0.07, Table 2).These three factors accounted for 19% of the variance of ciliate mortality rates in the regression model (adj.R 2 = 0.192), with each factor contributing 6-7% to the explained variance of δ max .Ciliate mortality increased linearly with r max (Fig. 2B) and declined with ciliate volume (Fig. 2C).

Ciliate mortality as a function of food availability-Evidence and open questions
My primary aim was to estimate the maximum (physiological) ciliate mortality rates across pelagic ecosystems.I found no difference between ciliate mortality in the ocean and lakes, rejecting the 5 th hypothesis.The median maximum mortality of planktonic ciliates was 0.62 d À1 , that is, an average ciliate population of 1000 cells L À1 would be reduced to 13 cells L À1 after 1 week of severe starvation (i.e., without any food).Since some food will always be available in situ, the actual ciliate mortality (δ) should be lower than the maximum mortality (δ max ).The key question is, how common is ciliate mortality induced by insufficient food availability?Importantly, food availability refers to suitable ("palatable") food of adequate quality.Prey quality can considerably affect ciliate growth rates (Müller and Schlegel 1999;Chen et al. 2010).
Direct support for starvation of natural ciliate populations from experimental approaches is scarce.Methods for measuring grazing, growth, and production rates of ciliates in situ have been available since the 1980s (reviewed by Weisse and Montagnes 2022).Some studies using incubation experiments under quasi-natural conditions reported negative growth rates for ciliate species in the absence of predation in fresh and marine waters (Stoecker et al. 1983;Carrias et al. 2001;Weisse 2006).In other cases, it is not clear if the reported estimates of r = 0 meant zero population growth (i.e., without any change in ciliate abundance during the incubation period) or if "no growth" also included negative growth rates (e.g., Verity 1986;Gilron and Lynn 1989b).
Severe food limitation may be encountered in (ultra-)oligotrophic lakes and the central gyres of the ocean, where chlorophyll a (Chl a) levels in the mixed layer typically range from  (Behrenfeld et al. 2005), yields a phytoplankton carbon biomass ranging from 10 to 100 mg C m À3 .This biomass matches the threshold food levels (at which specific growth rates are zero) of oligotrich and choreotrich ciliates (Weisse 2006;Montagnes 2013), which prevail in the ocean (Sherr and Sherr 1987;Pierce and Turner 1992;Montagnes 2013).Since algal biomass in the deep sea is dominated by small cells (Li 2002;Barton et al. 2010), many heterotrophic ciliates can only survive at the low food availability if they efficiently feed on picoplanktonic algae (< 2 μm) and/or bacteria (Sherr and Sherr 1987).In the mixed layer, bacterial standing stocks are similar to phytoplankton biomass at low chlorophyll concentrations, typically reaching 10 mg C m À3 (reviewed by Ducklow 2000).
Furthermore, considering that heterotrophic ciliates compete with other microzooplankton for the same food, their temporal starvation appears likely in the central gyres and highly oligotrophic lakes.However, although phytoplankton biomass and species diversity are low and relatively homogenous in the open ocean at lower latitudes (Alvain et al. 2008;Gregg et al. 2017), its spatial structure and algal diversity are not well understood.Ciliates might take advantage of micro-scale food patches (Durham et al. 2013;Priyadarshi et al. 2019).Using high-resolution Chl a fluorescence sensors, high degrees of mm-scale phytoplankton patchiness has become apparent (Doubell et al. 2014;Foloni-Neto et al. 2016), increasing along a gradient from estuarine to open ocean waters (Priyadarshi et al. 2019).
Patchiness has not only been demonstrated for phytoplankton, the primary ciliate food, but is also known for ciliate populations (Grattepanche et al. 2016 and references therein).Using morphological and molecular tools for identification, mesoscale differences in the taxonomic composition and the feeding type have been reported recently from the South China Sea (Liu et al. 2021), the Gulf of Mexico (Snyder et al. 2021), and along the New England Shelf in the northwest Atlantic (Grattepanche et al. 2016).Ciliate patchiness seems to be generally higher in coastal areas than offshore (Grattepanche et al. 2016).Even so, the microscale spatial and temporal variability (in the order of days) in food availability for ciliates in the central gyres is virtually unknown.
Unlike obligate heterotrophs, mixotrophic species may have a refuge from starvation (Stoecker 1998;Stoecker et al. 2017).Evidence accumulated in the past decades that mixotrophy is widespread among planktonic ciliates (reviewed by Esteban et al. 2010) and may be the dominant mode of nutrition in the open ocean (Mitra et al. 2014;Stoecker et al. 2017).Direct observations and mechanistic plankton models revealed that mixotrophy contributes significantly to the overall productivity of the marine planktonic food web, shapes the biological carbon pump, and may enhance primary production in oligotrophic waters (Mitra et al. 2014;Stoecker et al. 2017).Plankton models have also been applied to explore the consequences of food quality Â quantity interaction for the pelagic carbon flow and food web structure (Mitra and Flynn 2007;Gaedke 2022).With the accumulating evidence from laboratory and field investigations, models may help to reconcile the estimates of physiological mortality reported here with observations from the natural environment.

Physiological mortality vs. predator-induced mortality
How does the estimated physiological mortality rate compare to losses imposed by microcrustacean predation on the ciliates?In the central ocean, copepods, the main ciliate predators, collectively graze ciliates at a rate of 0.018-0.055d À1 (Lu and Weisse 2022).That is, the effect of predation on ciliate population dynamics is about tenfold lower than physiological mortality.This comparison corrobarates the conclusion that the population dynamics of planktonic ciliates in the ocean is primarily bottom-up controlled by resource limitation (Lu and Weisse 2022).
The situation is more complex in lakes because their trophic status is highly variable, and the ciliate microzooplankton predators are more diverse than in the sea.Besides calanoid and cyclopoid copepods, cladocerans (mainly Daphnia), rotifers, and some large carnivorous ciliates prey upon the dominant small and medium-sized freshwater ciliates (Weisse and Montagnes 2022).At moderate temperatures ($ 15 C), it takes about 15 microcrustaceans L À1 to graze ciliates at a rate of 0.3 d À1 (Lu and Weisse 2022).Such a predator abundance is reached in many mesotrophic and moderately eutrophic lakes (Wetzel 2001;Sommer et al. 2012).With increasing lake trophy, ciliates' predator density and food resources increase, reducing their δ.Therefore, I conclude that bottom-up control  is minor relative to predator-induced top-down control in many lakes.
Which factors determine ciliate mortality rates?Ciliate mortality rates are highly variable and primarily species specific.My analysis further revealed that δ max appears unaffected by temperature (Fig. 2A), seemingly rejecting the 1 st hypothesis.This result is surprising because the analysis covered a wide temperature range (5-27.5 C).However, in a recent study with three freshwater ciliates performed under comparable laboratory conditions, the species' mean mortality rate increased with temperature by 0.09 C À1 (Weisse et al. 2023b).Temperature explained 46% of the variance of δ max in this previous study.Therefore, the conclusion that temperature does not affect ciliate mortality may change if the temperature response of δ max has become available for more ciliate species.
The data appeared to support the 2 nd hypothesis that δ max would decline with ciliate cell volume (Fig. 2C).However, the effect of ciliate cell volume on δ max was only marginally significant (p = 0.07) and moderate.Accordingly, larger cell size would offer only limited shelter against starvation, confirming an earlier conjecture (Jackson and Berger 1984).
I found that δ max and r max were significantly correlated but that the latter, in contrast to the former, was unrelated to ciliate cell volume (data not shown).Similarly, a recent analysis of ciliate-specific growth rates reported that r max did not decline with increasing cell volume (Luki c et al. 2022).Accordingly, the available data on ciliate mortality rates and specific growth rates provide little support for the tenets of the metabolic theory of ecology that all metabolic rates increase with temperature and scale allometrically with size (Brown et al. 2004).Exceptions to the general rule that rising temperatures reduce organism size and increase growth rates are known for some small aquatic ectotherms (Atkinson 1995). Luki c et al. (2022) concluded that the lack of an allometric relationship between ciliate cell volumes and growth rates results from the relatively narrow size range of the ciliates studied thus far.Yet, this does not apply to the present study and may explain why I found a marginally significant volume effect on δ max .In the dataset I used, ciliate cell volumes spanned more than three orders of magnitude, from 517 μm 3 (Cyclidium glaucoma) to 859,000 μm 3 (Didinium nasutum; see dataset).
My meta-analysis supported the 3 rd and 4 th hypotheses.The analysis confirmed a positive relationship between ciliate growth and mortality rates (Fig. 2B).Similarly, as postulated, mortality rates of cyst-forming species were, on average, higher than those of species unable to encyst.Encysting species can sustain relatively high mortality rates because those refer to trophonts (i.e., motile cells) only (Weisse et al. 2023b).Encysted specimens may survive starvation and other unfavorable conditions for a long time (Verni and Rosati 2011;Li et al. 2022).Therefore, relatively high mortality rates of the trophonts of cyst-forming ciliates may reduce the species' fitness only moderately.

What do physiological mortality rates tell us about ciliate life strategies?
Taken together, my findings that ciliate mortality is affected by r max , cell volume and cyst formation characterize a ciliate life strategy ("grow fast, die fast" or "boom and bust ciliates"; Weisse et al. 2013) in line with the r/K continuum.Fastgrowing ciliates (r-strategists) are superior competitors under food-replete conditions.However, if food is depleted, fastgrowing ciliates run a higher risk of local extinction than their slower-growing congeners.Cyst formation is an effective trait to overcome the disadvantage of high mortality rates of motile cells.Slowly growing ciliate species (K-strategists) have reduced mortality rates and may be superior competitors under moderate food supply.In conclusion, the present evidence suggests that not only the growth rates of ciliates but also their resistance to starvation is a vital feature for selection.
Approximately 80% of the variance of δ max remained unexplained by my analysis, mainly because ciliate species, the most critical factor, was not considered in the final model.Ciliate mortality rates are species-specific, but many species cannot be classified adequately into broad categories such as r/K strategies.The life strategies of planktonic ciliates are highly versatile and likely also affected by factors other than the seven factors (including species) that the present analysis considered.

A note of caution: Taxonomic bias and methodological constraints
The estimates of δ max that I used for the final analysis comprise 47 species from 11 orders (see dataset).The dataset is heavily biased for the closely related choreotrich ciliates (18 species) and oligotrich ciliates (11 species), that is, my analysis represents only a tiny fraction of the ciliate species diversity (Foissner et al. 2008;Lynn 2008) and, most likely, functional diversity.Accordingly, the finding that ciliate mortality rates do not differ between marine and freshwater species should be interpreted with some reserve.
Mortality rates analyzed in this study were based on two different methods (NR and ROD).The pros and cons of these methods have been recently discussed (Weisse et al. 2023a).My finding that results obtained with the NR approach did not differ from those achieved using the ROD method is crucial for future studies.Researchers can choose between these methods to reliably estimate ciliate mortality rates depending on the circumstances.
In contrast to the NR method, the ROD method does not require pure ciliate cultures, that is, it can be used more efficiently in (simulated) in situ studies.This is an important application because more estimates of δ max are needed from oceanic ciliates.Apart from some laboratories located on remote islands, such data may be obtained aboard ship during cruises to the central gyres.
Both conventional methods used in this study to estimate ciliate mortality rates target the population level, not the individual cell.Even if clonal populations are used, intraspecific variation (Montagnes et al. 1996;Chen et al. 2023) affects the mean estimate of δ max .However, the extent of this variation remains unknown.Single-cell analysis using imaging cytometry (Weisse et al. 2023a) and transcriptomic gene expression analyses of key enzymes of ingestion and digestion of ciliates and other phagotrophic protists (Caron et al. 2017;Zou et al. 2020) may overcome this restriction.Applying single-cell analyses for estimating protist mortality rates is an avenue for future research.

Conclusions
Assessing the spatial and temporal variability in suitable ciliate food in the open ocean is crucial for estimating the significance of ciliate starvation on a global scale.The available evidence suggests that in the sea, physiological mortality is more critical for heterotrophic species than ciliate losses imposed by microcrustacean predation.In many lakes, the opposite holds.Cyst formation is an effective ciliate trait to overcome the high mortality of motile cells under fooddeplete conditions.Temperature did not affect the ciliate mortality rates, but this finding requires more research with contrasting ciliate species.Because ciliates' specific growth rates increase with temperature, the lack of a temperature effect on their mortality rates would have important implications in the context of global warming.Planktonic ciliates may take advantage of rising ocean and lake temperatures if the temperature is unimportant for ciliate mortality.
0.1 to 1 mg m À3 (Cho and Azam 1990; Dasgupta et al. 2009; Cornec et al. 2021).In vast areas of the open ocean, from 50 S to 50 N, even at the depth of the deep chlorophyll maximum, Chl a levels are ≤ 1 mg m À3 (Cornec et al. 2021).Assuming a relatively high carbon-to-Chl a ratio of 100 typical of highlight, low-nutrient conditions encountered in the open sea

Fig. 1 .
Fig. 1.Box-and-whisker plots of ln-transformed ciliate mortality rate vs. method (A), habitat (B), and cyst formation (C).The hinges of the boxes correspond to the lower (25%) and upper (75%) quartile; the whiskers indicate the largest and smallest observations falling within a distance of 1.5 times from the nearest hinge; the solid line marks the median.In (C), 0 = no cysts, 1 = cysts formation.

Table 1 .
Mortality rates (raw data, d À1 ) for all factors investigated.
Df, degrees of freedom; Estimate, regression coefficient; SE, standard error.