A synthesis of senescence predictions for indeterminate growth, and support from multiple tests in wild lake trout

Senescence, or the deterioration of functionality with age, varies widely across taxa in pattern and rate. Insights into why and how this variation occurs are hindered by the predominance of lab-focused research on short-lived model species with determinate growth. We synthesize evolutionary theories of senescence, highlight key information gaps, and clarify predictions for species with low mortality and variable degrees of indeterminate growth. Lake trout are an ideal species to evaluate predictions in the wild. We monitored individual males from two populations (1976-2017) longitudinally for changes in adult mortality (actuarial senescence) and body condition (proxy for energy balance). A cross-sectional approach (2017) compared young (ages 4-10 years) and old (18-37 years) adults for (1) phenotypic performance in body condition, and semen quality - which is related to fertility under sperm competition (reproductive senescence), and (2) relative telomere length (potential proxy for cellular senescence). Adult growth in these particular populations is constrained by a simplified food web, and our data support predictions of negligible senescence when maximum size is only slightly larger than maturation size. Negative senescence (aka reverse senescence) may occur in other lake trout populations where diet shifts allow maximum sizes to be much larger than maturation size.

condition (proxy for energy balance). A cross-sectional approach (2017) compared young (ages 26 4-10 years) and old (18-37 years) adults for (1) phenotypic performance in body condition, and 27 semen quality -which is related to fertility under sperm competition (reproductive senescence), 28 and (2) relative telomere length (potential proxy for cellular senescence). Adult growth in these 29 particular populations is constrained by a simplified food web, and our data support predictions 30 of negligible senescence when maximum size is only slightly larger than maturation size. 31 Negative senescence (aka reverse senescence) may occur in other lake trout populations where 32 diet shifts allow maximum sizes to be much larger than maturation size. 33 34 KEYWORDS: ageing, disposable soma, sperm senescence, life history theory, sexual 35 selection, Salvelinus namaycush 36 37

38
Senescence is a decline in individual biological function with age, and is typically 39 quantified as an increase in adult mortality rate or reduced 'fertility' [1], but can be applied to any 40 decline in phenotypic performance. Tremendous variability exists among species in the shape 41 (direction) and speed (rate) of senescence [2][3][4][5], and many authors seek to explain such patterns 42 [e.g., 3, 6, 7]. The contention that the strength of selection declines with age is a common 43 explanation of senescence [8]. The premise being that few individuals reach old age, and many 44 have already reproduced at younger ages, therefore, selection cannot remove problematic traits 45 that arise only at old age. An hypothesis that "low adult death rates should be associated with low 46 rates of senescence, and high adult death rates with high rates of senescence" [9], has empirical 47 support. However, the nuances of the hypothesis and its predictions are debated [6,10,11]. 48 Relative adult to juvenile mortality appears critical [6], but asymmetry between parent and 49 offspring [7] can differ widely between determinate and indeterminate growers and 50 generalizations can be problematic. An example with bivalves provides a useful illustration [see 51 6, page 527], which would also apply to most fishes. 52 Our manuscript has three primary goals: 1) synthesize existing senescence theories, 53 showing the importance of growth pattern, and highlight types of data needed to fill key voids, 2) 54 considered in two phases [42,45]: pre-meiotic (how the age of the male influences sperm) and 148 post-meiotic (both before and after ejaculation). Sperm are particularly vulnerable to oxidative 149 damage [31], and the male mutational bias [42,46], has led to interest in human fertility and 150 paternal effects. Male fitness is a function of mating opportunities, sperm performance and 151 offspring viability [33,44], which can be separated under experimental conditions [e.g., 47, 48]. 152 Older males generally produce sperm with reduced fertilization ability [27,29,33] and lead 153 higher rates of developmental abnormalities among offspring [29]. 154 155 2. LAKE TROUT 156

Desirable attributes 157
Lake trout present an ideal indeterminate growth model for studies of senescence in 158 nature, with low adult mortality being a key attribute. They inhabit the hypolimnion of lakes [49], 159 where there are functionally no predators on adults (contrasts greatly to marine predation on 160 anadromous salmonids) and spawn on lake shoals at night [49,50], where they are not exposed to 161 terrestrial predators (unlike stream spawning salmonids). 162 Reproductive quality and investment can be accurately estimated from gametes. Lake 163 trout do not typically migrate to spawn, show few secondary sexual characteristics, no sexual 164 dimorphism, have no energetically costly courtship, and provide no parental care [49,50]. 165 Fertility increases with size (age), as larger females produce more eggs.  [e.g., 15, 16, 26, 28, 32]) in variables such as growing season, prey resources, and 176 juvenile predators. 177 178

Support for theories of ageing 179
If senescence is optimized (Figure 1) between fitness benefits early in life at a cost to 180 either hyperfunctioning genes (DFT) or somatic maintenance (DSH), then selection against a 181 decline in performance with age is predicted to be relatively high in lake trout, as fitness potential 182 increases dramatically with size (age), given adult mortality rates decline while fertility increases. 183 We are unaware of any published data that can shed specific light on DFT in lake trout. However, 184 low allocation in reproduction is predicted to plastically tradeoff with high investment in somatic 185 maintenance under DSH [55]. Possibly supporting this, lake trout have relatively low secondary 186 sexual characteristics/migration/courtship/fecundity (resulting in low annual reproductive effort) 187 and a predictably high incidence of iteroparity [49] ). This is critically important, as diet is known to affect gamete quality in 207 fishes [e.g., 58] and would bias age (size) comparisons in most systems. Sampling over the 208 course of 40+ years has shown that young and old adult male lake trout co-occur on the spawning 209 shoals at the same time (Rennie, unpublished), thus our comparisons of age are not confounded 210 by differential spawn timing. 211 212

213
In polyandrous mating systems like lake trout, male "fertility" is influenced by the ability 214 to achieve fertilizations under sperm competition [33,44]

(A) Fish collection 247
We collected fish on spawning shoals at night from 11 to 16 October 2017 and sampled 248 the next morning following previous procedures [66]. Ages of recaptured fish were determined in 249 the field by cross-referencing a database of tag IDs. Younger adult trout were more abundant than 250 older individuals. To avoid potential confounding variables associated with date of sampling 251 (e.g., weather, transport time to laboratory), we grouped fish as either being young (ages 4-10) or 252 old (18-37) and processed them in a 'group design' (i.e., the same number of young and old fish 253 were sampled each day). We analyzed 15 groups in each lake (60 total; Supplemental Methods). 254

(B) Sample collection 256
Eggs were extruded from one female each day and later separated from ovarian fluid 257 through a fine meshed net [67], which was used in sperm swimming performance trials [68], to 258 avoid neutral sperm swimming environments when post-ejaculatory sexual selection occurs [29]. 259 From each male, blood was taken from the caudal peduncle and semen was expressed by gentle 260 abdominal massage. All samples were immediately immersed in ice, and transported to the lab 261 for further processing (completed within 8 hours of collection). 262 Aliquots of blood and semen were removed from ice and centrifuged (5000 × g at ~15 o C 263 for 5 mins). Prior to freezing in liquid nitrogen, plasma was separated from blood cells. A 264 separate semen aliquot was centrifuged in hematocrit tubes, and spermatocrit was computed [69]. We measured relative telomere length from DNA recovered from red blood cells and 278 sperm pellets using a qPCR-based approach that produces a telomere repeat (T) to single gene (S) 279 copy number ratio (T/S). The assay was performed with two single copy genes, orexin (Ox) and Body condition, spermatocrit, and relative telomere length were evaluated as a function of 287 fish age (young vs. old) crossed with lake of origin. Sperm swimming declines rapidly after 288 activation, with most successful fertilizations occurring in the few seconds after release. As such, 289 we quantified sperm swimming using two approaches. First, to assess maximum swimming 290 speed, we measured sperm at 6 s post-activation as a function of fish age (continuous variable: 4-291 37 years) crossed with lake, including tag ID (random intercept) to account for the four technical 292 replicates per male. We also tested for changes in sperm swimming speed over time post-293 activation (continuous: 6-30 s) crossed with age (young vs. old) and lake. Tag ID (random slope 294 and intercept) and technical replicate (random slope and intercept) were included. In all cross-295 sectional analyses the interaction between age and lake was not significant (P > 0.23), indicating 296 that the effect of age was similar in both populations. We removed these non-significant 297 interactions prior to reporting final model results. 298 299

ACTUARIAL SENESCENCE 301
Annual mortality probability estimates of adult male lake trout were low (< 0.20) across 302 all ages in both lakes, and suggest a modest increase with age ( Figure 2b, c). This effect of age 303 was clearer in Lake 224 compared to Lake 223 (99.8% and 80.5% of the posterior distributions of 304 the slope parameter were positive, respectively). 305

Longitudinal condition 308
Accounting for random individual (194 fish, 1608 observations) and annual variation, 309 there was a significant change in adult body condition with age in Lake 224 (t216.2 = -2.6, P = 310 0.009; Figure 2d). The rate of decline was negligible at 1.4 units per decade, which is well within 311 the variation among fish and years (most observations between 70-105 units). this genetic potential. However, realized growth in lake trout varies depending on diet 342 availability, with fish achieving enormous sizes in some lakes, but are stunted in others. We 343 exploited this scenario to make age comparisons among male trout that were not confounded by 344 diet differences. Adult trout in our study lakes have functionally determinate growth due to a 345 simplified foodweb. Despite this, and as predicted by both DSH and DFT, they show little to no 346 senescence in many traits measured here, and while we conclude that overall senescence is 347 negligible in these populations, we argue that there may be negative (reverse) senescence in other 348 populations that are not growth constrained. 349 Although we have no molecular data to underpin endorsement of DFT, the DSH is clearly 350 supported in our lake trout model. That low adult mortality [relative to juveniles, 6] should be 351 associated with few negative effects of ageing [9,11]  Observed variation in senescent patterns among species [5,18,19] suggests contrasting 365 selection pressures as an ultimate cause. Indeterminate growth is predicted [20] to increase 366 selection against senescence when adult individuals experience reduced mortality and increased 367 fertility with age (increasing size). Testing this prediction in wild populations has been 368 challenging due to often confounding variables. Lake trout from our particular populations enable 369 unique opportunities to control such problems, including diet. However, a stable diet likely 370 results in old fish having inferior performance than their inherent potential. Negative senescence 371 is predicted when size at maturity is much smaller than maximum size [20]. In lake trout 372 populations where adults can switch to larger or more energy dense prey as they grow, old fish 373 achieve much larger sizes than young adults [52]. Very large adults would have high sperm 374 quantity (predicted to win under sperm competition = fertility), and high sperm quality (predicted 375 to win under sperm competition = fertility) if under suitable diet (and females would have much 376 more egg production). Such data would not only support negligible senescence, but also show the 377 aptitude for negative (reverse) senescence in this species. 378 Our study suggests lake trout have at most negligible senescence, with the potential to 379 exhibit negative (reverse) senescence in populations where adults can attain maximum sizes that 380 are much larger than those at maturity, due to prey availability. These data provide support of 381 evolutionary theories of ageing, from rarely studied long-lived indeterminate growing animals in 382 the wild. Our data are unique in that they coalesce information on 1) actuarial senescence using 383 mortality rates from mark-recapture, with 2) measures of phenotypic performance including 384 reproductive senescence. Furthermore, blood and sperm cell telomere lengths did not decline 385 with age, 3) indicating that, at least under the conditions of this study, telomere maintenance 386 through adulthood may in part underpin the lack of apparent senescence. Our age comparisons 387 combine longitudinal (same individuals across decades) with cross-sectional data (difference 388 aged individuals at the same time), which is an infrequent approach. These age assessments are 389 strengthened by the unique characteristics of the study populations that control for confounding 390 variables that are profuse in most natural situations.  conflict and the evolution of ageing and life span. Functional Ecology 22(3), 443-453. 518 ( The effects of male age, sperm age and mating history on ejaculate senescence. Functional 541 Ecology 33 (7), 1267-1279. (doi:10.1111/1365-2435.13305

Actuarial senescence
Using all individual adult males with known ages (Lake 223 = 385, Lake 224 = 422), we treated age as a continuous variable and estimated the probability of recapture and the probability of survival (1 -mortality) of each individual adult male in each year (sampled from posterior distributions). As extrinsic adult mortality is low, and young/old adults experience the same conditions, any changes in mortality with age are assumed to be attributed to intrinsic processes.
In order to test for an increase in adult mortality with age, we fitted a Cormack Jolly Seber model (Lebreton et al., 1992) with separate survival (1 -mortality) and capture probabilities as linear regression functions on the logistic scale. We fitted a first order autocorrelation structure for annual random effects on survival and recapture probability, as we expect that factors affecting these parameters (especially survival) are likely to be similar from year to year. The likelihood structure of the basic process model is thus defined as . Any individual not alive ( , , = 0) cannot be observed, and those that are alive may be observed with probability , , , provided that sampling was conducted in year ( = 1 if sampling was conducted; = 0 otherwise). As in the process (survival) part of the model, the probability of capture of live individuals is modelled as a logistic regression, with separate intercepts for each age, and an autoregressive structure for annual variation. The structure for annual variation is directly analogous to that described above for the survival part of the model.
The model was sampled by Gibbs sampling using jags (Plummer, 2010) in R version 3.6.2. We used diffuse normal priors on all fixed effects and autoregression parameters, and diffuse gamma priors on the precision (inverse of the variance) of the disturbances of the survival and capture parts of the model.

Additional parameters
In addition, we tested models that included combinations of a quadratic term for the rate of mortality change and a parameter that varied the minimum mortality rate. However, these models did not converge well and were not numerically stable. This suggests that a quadratic term did not fit our data, and therefore, we did not use these parameters in the final model described above. However, we have left these as options within our scripts for others to see how they were included, alongside the scripts for plots we used to check for convergence.

Body condition
Length-based body condition was estimated as a percentage of standard weight (1993). Fish that were recaptured at least 6 times during their adult life were used to determine if condition declined with age, and were analyzed with a mixed effects modelling framework using the lme4 package (Bates et al., 2014) in R. Condition was evaluated as a function of fish age (fixed effect), and repeated measures on the same individuals (modelled as a random slope), and the year sampled (random intercept). Significance of fixed effects was assessed using the Satterthwaite approximation for degrees of freedom with the lmerTest package (Kuznetsova et al., 2017). Random effects were retained if found to be significant in log-likelihood ratio tests using the anova() function in R. Assumptions of normality and homogeneity of variance were verified using model residuals. Only fish from Lake 224 were used in this analysis as exclusion of data prior to 1990 (Mills et al., 2000) limited sample sizes in Lake 223.