Male and female reproductive fitness costs of an immune response in natural populations

Parasites can mediate host fitness both directly, via effects on survival and reproduction, or indirectly by inducing host immune defense with costly side-effects. The evolution of immune defense is determined by a complex interplay of costs and benefits of parasite infection and immune response, all of which may differ for male and female hosts in sexual lineages. Here, we examine fitness costs associated with an inducible immune defense in a fish-cestode host-parasite system. Cestode infection induces peritoneal fibrosis in threespine stickleback (Gasterosteus aculeatus), constraining cestode growth and sometimes encasing and killing the parasite. Surveying two wild populations of stickleback, we confirm that the presence of fibrosis scar tissue is associated with reduced parasite burden in both male and female fish. However, fibrotic fish had lower foraging success and reproductive fitness (reduced female egg production and male nesting success), indicating strong costs of the lingering immunopathology. We show that these substantial sexually-concordant fitness effects of immune response act to align multivariate selection across the sexes, masking the signature of sexual antagonism that acted on morphology alone. Although both sexes experienced costs of fibrosis, the net impacts are unequal because in the two study populations females had higher cestode exposure. To evaluate whether this difference in risk should drive sex-specific immune strategies, we analyze a quantitative genetic model of host immune response to a trophically transmitted parasite. The model and empirical data illustrate how shared costs and benefits of immune response lead to shared evolutionary interests of male and female hosts, despite unequal infection risks across the sexes.


Introduction
Organisms' immune systems evolve to prevent or mitigate the fitness costs that parasites impose 48 on their hosts. But, these immune systems can impose their own costs by consuming energy 49 (Sheldon and Verhulst 1996) or attacking the host's own tissues (Billi et al. 2019). can be viewed and studied as a life history trait (Sheldon and Verhulst 1996, Schmid-Hempel 55 2003), whose expression may trade off with different fitness components and whose evolution 56 depends upon these tradeoffs (Svensson et al. 1998, Lochmiller andDeerenberg 2000). Thus 57 rather than simply being maximized, host immunity is expected to evolve to optimize these 58 tradeoffs ( Evolution of immune defense becomes potentially more complicated when hosts 61 reproduce sexually. Due to divergent gamete investment strategies and pervasive differences in 62 the strength of sexual selection, males and females have sex-specific life history priorities (Rolff 63 2002, Stoehr and Kokko 2006). Shifts in the relative importance of longevity versus success in a 64 single reproductive bout can alter the relative benefits of investing in immunological traits and 65 mounting a response to a parasite or pathogen (Williams 1966, Zuk andStoehr 2002, Gipson and 66 Hall 2016, Hall and Mideo 2018). These strategic differences have been invoked to explain sex 67 differences in the intensity and side-effects of male versus female immune response. There is 68 some empirical evidence suggesting costs of immunity may be sex-specific: for example in humans there are numerous examples of sex specific immunopathology, including heightened 70 prevalence of autoimmune disease in females (Fish 2008, Billi et al. 2019) and greater severity 71 and frequency of parasitic diseases in males (Klein 2000, Roberts et al. 2001). These differences 72 in optimal immune investment can be further exaggerated because males and females often differ 73 ecologically (Shine 1989). Differences in diet, habitat use, or activity levels can impart distinct 74 risks of infection , and consequently sex differences in parasite infection 75 rates in wild animal populations are commonplace but variable (Poulin 1996 transmitted trophically when a stickleback consumes an infected cyclopoid copepod; sex 94 differences in Schistocephalus infection rates appear to be related to differences in male and 95 female diet (Reimchen and Nosil 2001). Schistocephalus complete their lifecycle when the host 96 fish is consumed by a bird, which is their definitive host where the cestode mates. S.solidus is 97 well known for its capacity to manipulate host color, behavior, and buoyancy to increase 98 sticklebacks' risk of avian predation. In addition to aiding host mortality, Schistocephalus 99 infection is associated with reduced fecundity in females (Heins et al. 2010 We predict that variation in fibrosis immune response across populations and cestode infection 120 rates across sexes may be the result of differences in the cost/benefit trade-offs that dictate the 121 optimal immune strategy. To test this prediction, we surveyed two wild stickleback populations 122 that naturally differ in parasite infection rates. Both populations exhibit moderate rates of both 123 infection and fibrosis. Crucially, there is an imperfect association between infection and fibrosis 124 in both populations: not all individuals initiate fibrosis when infected. Also, fibrosis persists at 125 least 3 months after the parasite is eliminated, so we find individuals with fibrosis but no 126 surviving infection. These facts allow us to statistically separate the costs of infection from the 127 costs of fibrosis, in both populations. We find massive fitness effects of both parasite infection 128 and of the fibrotic immune response, in both sexes and across both populations. We show that 129 these concordant fitness effects act to align otherwise-antagonistic multivariate selection across 130 the sexes. These results leave us with a puzzle: there are persistent sex differences in infection 131 rates, but no corresponding difference in the probability of fibrosis or its costs. To explain this 132 apparent contradiction, we construct and analyze an optimality model of immune response 133 evolution. Our analysis indicates that immune response optima may often be shared across sexes 134 that differ in parasite encounter/infection rates when costs of infection and immune response are 135 concordant and high. 136

Methods 138
Fish capture and measurement 139 We captured adult threespine stickleback from two lakes (Boot Lake and Roselle Lake), on To compare reproductive success for males with versus without fibrosis (or, cestode 148 infection), we compared the rates of fibrosis (infection) in randomly caught males, versus males 149 that had successfully nested. We snorkeled in the littoral zone to search for nesting males; we 150 identified nesting males based on their behavior (territory defense) and presence of a nest (e.g. 151 arrangement of vegetative debris). Males exhibiting these qualities were observed until egg 152 fanning behavior, or hatched fry, were observed, upon which the male was deemed to be a 153 successful nester and was captured, immediately euthanized, and placed on ice and frozen. 154 During this period we also placed minnow traps nearby to capture a random sample of males and 155 females; this allowed us to obtain a random sample of unmated males for comparison to those 156 that were successfully defending nests. Thus we obtained data required to measure total sexual 157 selection acting via variance in male mating success (defining sexual selection sunsu Arnold and 158 Wade 1984b, Arnold and Wade 1984a). Traps were placed at varying depths and distance from 159 shore, and we avoided placing traps directly in areas of high nest density so most trapped males 160 are unlikely to be nesting. Trapped fish were euthanized and frozen. We sampled both lakes 161 until at least 50 nesting males had been captured. This number was chosen to avoid excessive 162 impact on the populations studied while still presenting a reasonable sample size; we note 163 however that this sampling design precludes accurate estimation of population mean mating rate 164 because we have controlled the number of nesting males captured. We also retained a random 165 sample of trapped females in each lake, using ovary mass as a metric of reproductive stage. 166 All fish were kept frozen until later laboratory dissection, upon which fish were thawed, 167 measured, and dissected. We measured seven external morphological traits: standard length, 168 head length (measured from the snout to the distal end of the operculum), snout length (measured 169 from the snout to the orbital), eye width, body depth, body width at the pelvic girdle, and middle 170 spine length. These traits were measured to two decimal places with digital calipers. We then 171 dissected the left gills and counted gill raker number, in addition to photographing the gill rakers 172 under a dissection microscope at fixed magnification to measure length of the longest raker. 173 Following these measurements fish were dissected and all Schistocephalus counted and weighed. 174 We note that Schistocephalus was the only abundant macroparasite found in our sample; we also 175 examined internal organs, eyes, and the digestive tract for other parasite taxa. Gonads were 176 removed and weighed. Stomach contents were removed and identified to (at least) order for all 177 individuals, with the exception of samples from nesting males from Roselle, whose stomachs 178 were lost during shipping. 179 We scored fibrosis on an ordinal scale, where a value of 0 corresponds to no apparent 180 fibrosis (organs move freely), a value of 1 corresponds to fibrosis between organs, a value of 2 181 corresponds to fibrotic connection between organs and the peritoneal tissue, and a value of 3 182 corresponds to excessive fibrosis across the entire body cavity. This scale is modified from that previously been shown to be highly repeatable between independent observers blind to 186 experimental vaccination treatment (Goldzmid and Trinchieri 2012 accommodates the ordinal nature of the variable), assuming exponential error. We avoided 194 including fish body mass in this model because Schistocephalus mass is a direct component of 195 total fish body mass and fibrosis is unrelated to mass, although results were unchanged when 196 correcting for mass. Because our sample size for this analysis was limited to the subset of fish 197 that were infected, we pooled lakes and sexes and did not model higher order interactions. We 198 used a separate linear model with infection probability as a binomial response (cestode present or 199 absent) to host fibrosis score, lake, and host sex; interaction terms were not significant and were 200 dropped. 201 202

Statistical Analysis-Costs of fibrosis and infection 203
We considered the fitness costs of fibrosis and infection on male and female reproductive 204 success (nesting or not; ovary mass). We assessed component fitness costs for females in Roselle 205 lake using a linear model with ovary mass as a response, fibrosis score and infection status as 206 fixed effects, and exponential error. Small sample size of gravid females precluded such an 207 analysis for Boot fish. We obtained qualitatively equivalent conclusions including body mass in 208 the model, although present the results without size correction. A caveat with our analysis of 209 female ovary mass is that we cannot control for females that may have already laid a clutch and 210 are in the early stages of developing a second. We repeated this analysis for males from Boot 211 and Roselle lakes, using binomial nesting success as the response variable and fitting separate 212 models for each lake. Male testes mass is less clearly related to reproductive success than is 213 female ovary mass and so we focus on male nesting success. 214 The above analysis of male nesting success indicated strong sexual selection against 215 fibrosis, and we were interested in if and how this selective force acts in conjunction with sexual 216 selection on morphology. To do this, we estimated the bivariate fitness surface for morphology 217 and fibrosis for males separately for Boot and Roselle lakes using binomial linear models with 218 nesting success as the response variable and fibrosis score and score on the multivariate 219 morphological selection gradient as fixed effects. Score on the morphological selection gradient 220 was obtained as discriminant function scores calculated in discriminant function analysis of 221 nesting success and morphology, performed separately for each lake; this is equivalent to 222 calculating individual scores on the vector of multivariate directional sexual selection on 223 morphology (Mitteroecker and Bookstein 2011). This approach essentially estimates individual 224 score on the morphological sexual selection gradient and then uses this score as a predictor in a 225 linear model with fibrosis as an additional predictor of fitness. This approach allowed us to 1) 226 estimate the effects of fibrosis on fitness accounting for effects of morphological traits, where 227 morphology has reduced dimensionality yielding increased power and 2) allowed us to plot the 228 corresponding bivariate fitness surface. We obtained qualitatively equivalent conclusions on the importance of fibrosis using a full multivariate model; we explore such a multivariate model 230 below (Statistical Analysis -Alignment of multivariate selection across the sexes). 231 To assess differences in diet, including potential costs of immune response and infection 232 on resource acquisition, we used a series of uni-and multivariate linear models to test for 233 associations between diet and fibrosis or infection. Because of the sparse multivariate nature of 234 diet data (i.e., many zeros for rare taxa), we constructed our models to test three specific 235 hypotheses, in all cases avoiding higher order interactions wherever appropriate. First, to 236 evaluate variation in overall prey intake, we used a univariate linear model with total prey counts 237 as the response and fibrosis, lake, and infection status as

Costs of fibrosis and infection 310
Countering these benefits, we also see evidence for costs of fibrosis. In both Boot and 311 Roselle lakes, males defending active nests exhibited reduced fibrosis levels compared to 312 randomly sampled males (Boot, odds ratio of unit offset at the mean 0.56, F 1,151 = 6.87, P = 313 0.0097; Roselle, odds ratio of unit offset at the mean 0.41, F 1,153 = 12.57, P = 0.0005) and 314 reduced cestode infection in Boot lake (odds ratio 2.45, F 1,151 = 4.3, P = 0.039; Roselle lake 315 odds ratio 1.89, F 1,153 = .57, P = 0.45). This is reflected in significant (Roselle, F 1,152 = 6.24, P 316 = 0.0135) or nearly significant (Boot, F 1,147 = 3.52, P = 0.062) negative effects of fibrosis on 317 mating probability in models including morphological discriminant function score, that is the 318 vector of sexual selection on morphology, as a predictor of mating success (P < 0.0001 for both lakes). Thus, for both lakes, nesting success was determined by the independent effects of 320 ecomorphology and fibrosis (Figure 3). 321 Although we lack data on mating success for females, we did obtain data on ovary mass, 322 which is expected to be directly related to fecundity. For females in Roselle lake, we found that 323 fibrosis was associated with a significant reduction in ovary mass (F 1,67 = 10.65, P = 0.0017; 324 In our analysis of diet content, we found individuals with high fibrosis had fewer total 331 prey items in their stomach (F 1,341 = 7.75, P = 0.0057; Figure 5A) controlling for lake effects 332 (F 1,341 = 43.11, P < 0.0001); this model indicated no difference in total prey items in infected 333 versus uninfected individuals (F 1,341 = .79, P = 0.37). In a multivariate model, we find no 334 evidence of sex differences in the total number of prey items in an individual's diet (sex effect, 335 F 1,343 = 0.06, P = 0.81), although the taxonomic content of individual diet varied with sex 336 (sex*prey taxon effect, F 7,2393 = 2.39, P = 0.019) while controlling for lake effects (lake*prey 337 taxon effect, F 7,2393 = 4.04, P = 0.0002). This sex effect was primarily driven by differences in 338 the abundance of dipteran larvae, fish eggs, and zooplankton ( Figure 5B), with females tending 339 to have more limnetic prey. For the subsample of males from Boot lake, for which we had diet 340 data for both nesting and non-nesting males, we found a main effect of nesting status in a 341 multivariate model (F 1,145 = 7.15, P = 0.0084) with nesting males having more prey items in their stomachs than a random sample of males in the population ( Figure 5C). We found no evidence 343 of a difference in the relative amounts of prey taxa in the diet of nesting and non-nesting males 344 (mate status*prey taxa effect, F 3,431 = 7.15, P = 0.35). 345 346

Alignment of multivariate selection 347
Given that fibrosis appears to confer similar component-fitness costs in males and 348 females, we can examine the effects of fibrosis on geometric alignment of multivariate selection 349 across the sexes. We find weak or even antagonistic alignment between multivariate selection 350 on morphological traits in males and females ( Figure 6), with evidence of sexually antagonistic 351 selection on body shape independent of body size. This is reflected in a negative estimate of the 352 correlation between male and female multivariate selection gradients ( Figure 6B is a latent trait describing the degree to which 399 fibrosis is induced in response to a parasite exposure, and is thus the key variable of interest in 400 our model, whose evolution we seek to predict. We borrow our notation here from the literature 401   The main conclusion from this model is that although optimum immune sensitivity can 431 indeed be governed by encounter rate, when costs and benefits of infection and immune response 432 are high these encounter rates play less of a role in determining the optimum immune response. 433 We find evidence of sex and population differences in parasite encounter rates, for each subpopulation sampled are provided in Table 2. 450  fibrosis was associated with a reduction in the total number of prey items found in an 619 individual's stomach (panel A). Controlling for differences among lakes, males and females 620 differed in their multivariate diet content, and this effect was driven by females consuming more 621 dipteran larvae and zooplankton and males consuming more fish eggs (panel B; least square 622 means and standard errors). In males from Boot lake, from which stomach contents of nesting 623 males were available, we find that successfully nesting males had consumed more total prey 624 items than a random sample of males captured in minnow traps (panel C; least square means and 625