Predators modify the temperature dependence of life‐history trade‐offs

Abstract Although life histories are shaped by temperature and predation, their joint influence on the interdependence of life‐history traits is poorly understood. Shifts in one life‐history trait often necessitate shifts in another—structured in some cases by trade‐offs—leading to differing life‐history strategies among environments. The offspring size–number trade‐off connects three traits whereby a constant reproductive allocation (R) constrains how the number (O) and size (S) of offspring change. Increasing temperature and size‐independent predation decrease size at and time to reproduction which can lower R through reduced time for resource accrual or size‐constrained fecundity. We investigated how O, S, and R in a clonal population of Daphnia magna change across their first three clutches with temperature and size‐independent predation risk. Early in ontogeny, increased temperature moved O and S along a trade‐off curve (constant R) toward fewer larger offspring. Later in ontogeny, increased temperature reduced R in the no‐predator treatment through disproportionate decreases in O relative to S. In the predation treatment, R likewise decreased at warmer temperatures but to a lesser degree and more readily traded off S for O whereby the third clutch showed a constant allocation strategy of O versus S with decreasing R. Ontogenetic shifts in S and O rotated in a counterclockwise fashion as temperature increased and more drastically under risk of predation. These results show that predation risk can alter the temperature dependence of traits and their interactions through trade‐offs.

them-change in response to the joint effects of temperature and ecological interactions.
However, phenotypically plastic responses to predation risk are often context specific whereby trait responses are governed by the nature of the threat (Beckerman, Rodgers, & Dennis, 2010;Benard, 2004;Bourdeau, 2009;Relyea, 2001;Riessen, 1999). For example, when predation risk is negatively size dependent and offspring mortality is relatively high compared to that of adults, freshwater snails (Physella virgate) delay reproduction to grow to a size refuge from predation (Crowl & Covich, 1990). Similarly, when exposed to cues of predators that selectively forage on larger prey (positively size-dependent predation), cladocerans change their life-history strategies to increase reproductive output, reproduce earlier, and at a smaller size (Beckerman et al., 2010;Stibor, 1992). Thus the effects of predation on traits (e.g., changes in time to reproduction, offspring size, number of offspring, reproductive investment) can be counter to or complement the effects of temperature depending on the nature of the risk posed by a predator. Regardless, the effects of temperature on rates (e.g., maturation, metabolism, growth), traits (e.g., body size, total reproductive investment) and links between traits (e.g., trade-offs) can alter the underlying ability of organisms to respond to predation.
Trade-offs result from constraints on life-history traits such that individuals must allocate finite resources among competing priorities (Davison, Boggs, & Baudisch, 2014;de Jong & van Noordwijk, 1992;Luhring & Holdo, 2015;van Noordwijk & de Jong, 1986;Smith & Fretwell, 1974;Stearns, 1989). For example, allocating resources to growth may reduce reproduction (Black & Dodson, 1990), and allocating resources to reproduction may reduce survivorship (Kirkwood & Rose, 1991). Similarly, parsing a fixed total reproductive investment among offspring results in a central life-history trade-off whereby increasing the number of offspring requires a reduction in offspring size (Fox & Czesak, 2000;Lim, Senior, & Nakagawa, 2014;Rollinson & Rowe, 2015;Smith & Fretwell, 1974). This trade-off arises because a fixed reproductive investment (R) in offspring biomass is given by the product of offspring size (S) and offspring number (O) (4), such that O and S are simultaneously determined by allocation strategy (location on the trade-off curve) as well as R (location of the trade-off curve). Furthermore, many traits that are expected to be locked in trade-offs and negatively correlated (e.g., O, S) are often positively correlated because of changes in the underlying R being partitioned (van Noordwijk & de Jong, 1986).
The size-number trade-off may constrain options for responding to changes in the thermal environment because O, S, and R vary with environmental temperature (Atkinson et al., 2001;Berger, Walters, & Gotthard, 2008;Ernest et al., 2003;Perrin, 1988) while simultaneously responding to ecological interactions such as the presence and types of predators in the environment (Riessen, 1999). R can increase with increasing temperatures because of greater resource uptake rates or resource productivity (Burnside, Erhardt, Hammond, & Brown, 2014;Englund, Öhlund, Hein, & Diehl, 2011;Ernest et al., 2003;Kerkhoff, Enquist, Elser, & Fagan, 2005) but only up to a point (Hammond & Diamond, 1997), thus moving the trade-off curve up and to the right at warmer temperatures ( Figure 1b). However, temperature accelerates reproductive schedules which generally leads to smaller adult size at reproduction, less time to accrue R, and would thus require a decrease in S and or O (Kingsolver & Huey, 2008;Perrin, 1988;Walls & Ventelä, 1998;Figure 1c). How temperature will affect the offspring size-number trade-off is thus contingent on how suites of interdependent traits jointly respond to temperature.
In this study, we assess the joint effects of temperature and predation risk on the plasticity of the offspring size-number trade-off in Daphnia magna. Daphnia show phenotypically plastic changes in O and S with changes in temperature and predation risk (Riessen, 1999;Walls & Ventelä, 1998 temperature on Daphnia suggests that adult size at first reproduction, clutch size, and offspring size should decline with increasing temperature (Giebelhausen & Lampert, 2001). Like increasing temperature, size-independent predation decreases adult size at and time to reproduction, however, it also increases offspring number (Riessen, 1999 Third brood neonates were collected within 24 h of birth and randomly assigned to experimental treatments.

| Predator cue production
Odonate larvae (mostly libellulids) were collected from a freshwater pond on the Spring Creek Prairie Audubon Center in southeastern Nebraska (Luhring & DeLong, 2016;Novich, Erickson, Kalinoski, & DeLong, 2014 odonates (N = 17). This was repeated across six containers for each night of cue production. Predators were allowed 24 hr to consume prey and produce a variety of kairomone sources (feces, excretion, etc.). After the 24 hr, the predator cue water was filtered through 63 μm sieves and predators saved for subsequent cue production.
The cue water from all containers within a night (hereafter "batch") was combined, mixed and then immediately frozen in 50-200ml increments to prevent cue degradation (Crawford, Hickman, & Luhring, 2012;Hickman, Stone, & Mathis, 2004). Three total batches of predator cue were prepared in this manner. For each water change during the experiment, equal amounts of predator cue water from each batch (1 L) were slowly thawed in lukewarm water baths and then combined to produce a master mix (3 L) of predator cue so that all batches were equally represented within and across all water changes.

| Husbandry and measurements during experiment
Water changes were conducted every Monday, Wednesday, and Friday starting with day 1 (Monday). During each water change, control treatments received 25 ml of fresh COMBO mixed with algae at 0.01 mg C/ml, while predator cue treatments received 25 ml of thawed predator cue water mixed with algae at 0.01 mg C/ml. All vials were acid-washed and oven dried to prevent unintentional transfer of predator kairomones. We controlled food availability by running all experiments in 24 hr dark which prevented algal growth (Cressler, Bengtson, & Nelson, 2017).

| Experimental design
During the first 7 days of the experiment (hereafter "natal" period), all Daphnia were maintained at 17°C (historic colony temperature) across seven environmental chambers (Percival Intellus Ultra Control System). After water changes on day seven, the "thermal performance curve" period (hereafter "TPC" period) began and temperatures in the seven environmental chambers were changed to 11, 17, 23, 27, 29, 31, or 33°C (chambers were randomly assigned temperatures). This temperature range encompasses both a decrease in temperature from the natal environment and a realistic increase in temperatures experienced by mobile plankton in freshwater systems (Kremer, Fey, Arellano, & Vasseur, 2018).
The 7 day acclimation period was used because it allowed us to study the effects of temperature on life history across a wider range of temperature; individuals exposed from birth to the lowest temperature would not have reached sexual maturity within the experimental time horizon, whereas individuals exposed from birth to the highest temperatures have very low survivorship.
Moreover, the acclimation period allowed us to isolate the effects of predation risk on age and size at maturity, key life-history traits known to respond to predation and to subsequently influence offspring size and number (Riessen, 1999 (Ebert, 1993):

| Data collection
Three offspring from each clutch were measured and the average size was used for a clutch estimate. Total clutch biomass was calculated as the product of clutch size and average offspring biomass for that clutch. Adults were measured in the same manner and on the same days as offspring with a Leica IC80 HD camera attached to a Leica M165C dissecting microscope. Because many births occurred on days when adults were not measured, we estimated adult size on these days by interpolating between adult sizes on days immediately before and after the clutch date.

| Curve fitting
To understand the effects of temperature and predation risk on individual traits in the size-number trade-off, we analyzed offspring size, clutch size, adult size at reproduction, and time to reproduction across temperature within each treatment. Adult size at each clutch was incorporated in statistical models to control for body size variation in resource accumulation (Cressler et al., 2017;van Noordwijk & de Jong, 1986) and packing constraints (Glazier, 2000). Because temperature-dependent biological phenomena are often nonlinear (Amarasekare & Savage, 2012;DeLong et al., 2017;Kingsolver, 2009) we fit all temperature-dependent processes with generalized additive models (GAMs) with the 'gam' function ('mgcv' package;R Core Team, 2017;Wood, 2006Wood, , 2015. Preliminary models indicated that three knots were optimal for all response variables and that temperature had strong nonlinear effects (significant smoothers We only analyzed first clutches that were produced prior to day 11 (4 days after TPC performance period started; N = 124 replicates) because we were interested in the signature of conditions of the natal environment (variation in predation regimes at the colony temperature of 17°C) on life-history traits. We chose day 11 as it was the earliest day for which first clutches were produced in all treatment-by-temperature combinations except 11 and 33°C. Not all treatments produced offspring at 11 and 33°C, and therefore these temperatures were removed from curve fitting analyses.
Second and third clutches developed entirely within the TPC period and were not restricted by experimental day prior to analysis.
Clutch size and offspring size analyses were restricted to clutches with more than one offspring, as clutches of size one were generally partial clutches from unhealthy individuals and were outliers relative to other replicates within treatment-by-temperature combinations.

| Depicting changes in offspring size and number
To track the effects of temperature and predation risk on offspring size and number, we plotted the offspring size and number aver-

| Temperature and predation effects on offspring size and number
Temperature (either alone or in combination with another predictor) had nonlinear effects on S, O, adult size at reproduction, and time to reproduction for all three clutches (Table 1). Clutch size (O) generally decreased with temperature and increased with predation risk in all three clutches ( Figure 2, Table 1; Supporting information Figure S1). Offspring size (S) showed a positive relationship with temperature for the first clutch, but shifted to a negative size-temperature relationship (Kingsolver & Huey, 2008) once temperature exposure of the adults was more chronic (clutches 2, 3; Figure 2, Table 1). Predation effects on S were inconsistent and largely absent except for smaller first clutch offspring in the early exposure treatment (Supporting information Figure S2). There were also treatment-specific temperature curves for S in the first clutch control and third clutch constant treatments ( Figure 2, Table 1).
Our results also indicate that the timing and duration of exposure to predation risk change the temperature dependence of lifehistory traits. Exposure to predation cues during the natal period (early and constant cue treatments) elevated clutch number (O) in the first clutch (first row Supporting information Figure S1). By the second clutch, recent exposure appeared to be more impactful as late and constant exposure treatments (but not the early exposure treatment) had higher O at warmer temperatures than the control (second row Supporting information Figure S1). By the third clutch, exposure to predation cue at any phase of ontogeny appeared to affect the temperature dependence of O by elevating it at intermediate temperatures (third row Supporting information Figure S1).

| Effects of temperature and predation on sizenumber trade-offs
The significant effects of temperature and predator cues on S and O led to shifts both along (constant R) and across size-number trade-off curves (changing R; Figure 3, Table 1). Two broad patterns emerged, separated by first and later clutches. In the first clutch, the negative effects of temperature on R (not reaching the ½ isocline until 27°C and not reaching the ¼ isocline at any temperature). In other words, although increased temperature decreased R overall, D. magna experiencing predation risk lowered R more gradually than controls and maintained higher O relative to controls as temperature increased.
The downward movement of trade-off curves in space with increasing temperature was consistent with accelerated reproductive schedules, and thus a reduction in R. The effects of temperature and predation risk on adult time and size of reproduction were prevalent across all clutches (Table 1). Daphnia adults reproduced earlier on average with the presence of predation cues and with increasing temperature (Table 1, Figure 4). The difference between control and constant cue treatments was most pronounced by the third clutch for both time to and size at reproduction (Figure 4), consistent with an accumulated effect of earlier clutch production over time. This increased departure from the control treatment at warmer temperatures was also seen in the early and late exposure treatments for age at reproduction (Supporting information Figure S3), indicating that reproductive schedules were influenced by both current and historical exposure to predation risk. While the temperature-dependent nature of adult size was often nonlinear (smooth terms, Table 1), only adult size in the constant cue treatment deviated on average from Daphnia in the control across temperatures (Figure 4; Supporting information Figure S4).
Only control D. magna raised at their natal colony temperature (17°C) throughout the experiment increased R, O, and S as they progressed from their first to latter clutches (black line Figure 5).  Figure 5).
We used the size-number trade-off in a clonal population of D.
magna to evaluate how a suite of linked life-history traits responded plastically to temperature and predation risk across ontogeny (first three clutches). The size-number trade-off helps to frame how offspring size (S) and number (O) change together given the shifting constraint of resource allocation to reproduction (R). All aspects of the trade-off (O, S, R) simultaneously responded to predation and temperature (Figure 3, Table 1) with trait responses showing largely interactive rather than additive effects of predation and temperature (10 of 12 clutch by trait combinations showed significant treatment-by-temperature smoother terms in Table 1). In contrast to most previous work, we show how both the strategy (moving along the trade-off curve; Figure 1a) and the overall allocation   (Figures 3, 5).
Life-history strategies in our study shifted across ontogeny and the magnitude and direction of these shifts were strongly affected by temperature and predation (Figures 3, 5). Ontogenetic shifts in size-number strategy moved in a counterclockwise fashion with increasing temperature ( Figure 5). However, predation risk altered 1)  Figure 5) are consistent with other studies lacking these added stressors (Glazier, 1992 Our results highlight the temperature dependence of phenotypically plastic traits in response to predation risk and how temperaturedependent shifts in constraints that underlie key life-history trade-offs shape trait space. These results serve as yet another example of the importance of incorporating multiple traits, their interactions (e.g., trade-offs), constraints, and responses to ecologically relevant pressures (e.g., predation) into projections of how organisms will respond to climate change. Life-history traits coevolve (Endler, 1995;Ghalambor, Walker, & Reznick, 2003;Protas et al., 2008) and respond to shifting environmental conditions through rapid evolution (Hairston, Ellner, Geber, Yoshida, & Fox, 2005;Padfield, Yvon-Durocher, Buckling, Jennings, & Yvon-Durocher, 2016;Thompson, 1998) and phenotypically plastic trait change (Kremer et al., 2018).
The manner in which suites of life-history traits will coevolve in response to shifting thermal clines remains is poorly understood as is the manner which these trait changes manifest across ontogeny and trade-offs (Angilletta, 2009). Furthermore, because aquatic habitats can show strong spatiotemporal temperature variance (Kremer et al., 2018), changing location in the water column to follow food or reduce predation risk (Burks et al., 2001)  Urbauer, S. Uiterwaal, and R. Vetter for transferring, counting, and measuring over 2,000 daphnia babies. We thank J.P. Gibert for commenting on an earlier draft of this manuscript. TML thanks the University of Nebraska's Program of Excellence in Population Biology.

CO N FLI C T O F I NTE R E S T
The authors declare no competing interests.