Effects of early nutritional stress on physiology, life histories and their trade-offs in a model ectothermic vertebrate

Early-life experiences can have far-reaching effects into adulthood and can ultimately impact fitness (Lemaitre et al., 2015; Lindström, 1999; Metcalfe and Monaghan, 2001). Early-life experiences can influence such life-history traits as growth, time to maturation, reproduction and survival, and trade-offs among them (Roff and Fairbairn, 2007). Much of the work on the fitness effects of early-life experiences has focused on resource availability, but some studies have also focused on the effects of heightened glucocorticoid levels (Grace et al., 2017) and thermal conditions (Lee et al., 2012). When considering resource availability in particular, poor nutrition in young animals can reduce fitness (Birkhead et al., 1999) or increase fitness (Ozanne and Hales, 2004; Shanley and Kirkwood, 2000), suggesting that the mechanisms by which early-life experiences either positively or negatively impact fitness may depend on the environmental context or the homeostatic state of the individuals within a population. In general, food scarcity per se impacts growth, reproduction and survival in a variety of vertebrates, possibly through moderating energy allocation decisions to competing traits (Kubička and Kratochvíl, 2009). Specifically, early-life nutritional stress can have significant long-term consequences for adult traits, such as reduced body mass in zebra finches (Birkhead et al., 1999); impaired neural development in sparrows (MacDonald et al., 2006); reduced reproduction in lizards (Kubička and Kratochvíl, 2009); increased mortality in fish (Inness and Metcalfe, 2008); and delayed age and reduced size at maturation in guppies (Auer, 2010). Even when nutritional stress is alleviated, lasting effects on adult phenotypes and ultimately fitness may occur (Marcil-Ferland et al., 2013).

Animals deploy different strategies to mitigate early-life nutritional stress on growth and maturation (Auer, 2010; Mueller et al., 2012; Vega-Trejo et al., 2016). Growth rates are plastic, particularly in species with indeterminate growth, and are usually regulated at optimal rather than maximal rates (Arendt, 1997). Growth can be increased when selective pressures favor an increase in overall size (Bronikowski, 2000; Metcalfe and Monaghan, 2003) and such plasticity can be important in environments where resources vary seasonally or annually. Moreover, in many species, adult size is a strong determinant of fitness (Choudhury et al., 1996; Gaillard et al., 2000) and the ability to speed up or slow down growth becomes an important mechanism linking ecological conditions with fitness (Bize et al., 2006; Bjorndal et al., 2003). Ultimately, individuals should balance the costs and benefits of obtaining a larger body size (Charlesworth, 1994). Three distinct patterns of growth and sexual maturation are possible after alleviation of early-life nutritional stress. Individuals may mature at a standard maturation size (but at a later age; Fig. 1A); they may mature at a standard maturation age (but smaller size; Fig. 1B), or they may increase their growth rate to mature at the standard size and age (Fig. 1C). In the first case, delaying maturation may be costly as it increases generation time and may decrease the overall reproductive lifespan of an individual (Roff, 1992). In the second case, maturing at a smaller body size may be costly if body size is positively correlated with survival and reproductive success (Roff, 1992). Given these potential fitness costs, selection might instead favor increasing growth to mature at normal age and size, provided the costs of doing so are not too high (Metcalfe and Monaghan, 2001). Compensatory growth has been reported across vertebrates: in mammals (Ryan, 1990), birds (Criscuolo et al., 2011), fish (Ali et al., 2003), amphibians (Hector et al., 2012) and reptiles (Radder et al., 2007). Compensatory growth is often achieved through a hyperphagic response to increases in resources (Ali et al., 2003; Gurney and Nisbet, 2004; Morgan and Metcalfe, 2001), but can also be achieved by allocating more energy into growth at the expense of other traits such as locomotor performance (Criscuolo et al., 2011; Lee et al., 2010), immune function (Norris and Evans, 2000), cognition (Fisher et al., 2006), maturation (Auer et al., 2010; Hector et al., 2012; Vega-Trejo et al., 2016), adult body size (Auer, 2010), lifespan (Birkhead et al., 1999; Lee et al., 2013) and litter size (Auer et al., 2010).

We tested among the competing hypotheses outlined in Fig. 1 to address how early-life nutritional environment influenced patterns of growth, maturation size, physiology and reproduction in the checkered garter snake Thamnophis marcianus (Baird and Girard 1853). To date, much of the literature on early-life experiences and the resultant effects on adult phenotypes has focused on birds and fishes (Ali et al., 2003; Lemaitre et al., 2015; Metcalfe and Monaghan, 2001). Our study species, the checkered garter snake, has indeterminate growth (in contrast to birds but in agreement with fishes), sexual size dimorphism, with females larger than males, and a positive correlation between body size and reproductive fitness for both males and females. We were further interested in whether measures of immune function and physiological stress would similarly reflect growth rate changes before and after the switch to normal food availability, as well as whether an effect of early-life nutritional stress would be evident on mating success and first reproductive effort. Finally, we tested whether any of these effects, and trade-offs among traits, differed between the sexes.

We studied the effects of early-life food availability in checkered garter snakes. This species has a broad distribution throughout southwestern North America (Rossman et al., 1996; Seigel et al., 2000). Checkered garter snakes have female-biased sexual size dimorphism [mean snout-to-vent length (SVL) for females: 579 mm; males: 477 mm]. This species is viviparous and have litters that range from 5 to 31 neonates (Ford and Karges, 1987; Rossman et al., 1996). Study subjects were offspring from paired breeding adults maintained at the Ophidian Research Colony (University of Texas at Tyler) where multiple generations have been bred from wild progenitors, subject to non-sibling matings. Seventy neonates from eight litters (i.e. eight different families) were randomly assigned to two experimental feeding groups with litters and sex split nearly evenly (‘normal-diet female’, N=15; ‘normal-diet male’, N=18; ‘low-diet female’, N=16; ‘low-diet male’, N=21). Neonates were weighed and measured at birth and housed in individual cages (35×21×13 cm translucent plastic storage boxes) each filled with 3–5 cm of aspen bedding with water available ad libitum and maintained on a 14 h:10 h light: dark cycle at 28±1°C to simulate a photoperiod and temperature regime commonly experienced throughout their range during the active season (Rossman et al., 1996).

Twice a week, for the first two weeks of life, all T. marcianus neonates were offered size appropriate pieces of tadpole tail to help ensure regular eating patterns, as their diet in the wild consists mainly of tadpoles, frogs and fish (Rossman et al., 1996). At two weeks of age, snakes were started on their diet protocol for either ad libitum or low-food availability. Each individual in the ad libitum early-life diet treatment (hereafter ‘normal’ diet) was offered 60% of its body mass weekly in newborn mice over two feedings as established by an earlier pilot study (N.B.F., unpublished data). Each individual on the early-life low-diet treatment (hereafter ‘low’ diet) was offered 20% of its body mass in newborn mice weekly in one feeding. Previous work had shown that this amount results in reduced size, compared to unrestricted conspecifics, but is still adequate to facilitate growth and maturation (N.B.F., unpublished data). Feeding amounts were adjusted every 15 days to track increasing body weight.

Voluntary food consumption was measured at each feeding by weighing food before and after feeding. Snakes were maintained on these diet treatments until both (i) a significant difference in mass between the two diet treatments occurred, and (ii) the low-diet individuals had doubled their birth mass, so as to ensure individuals were large enough to obtain an adequate blood sample. These criteria were met at 15 weeks of age and animals on the low diet were then switched to the normal diet for 15 weeks (see Fig. 2 for experimental timeline). Animals were maintained on the normal-diet protocol for an additional 10 weeks (i.e. 40 weeks of age), fasted for 2 weeks, and placed into hibernation at 4°C (Fig. 2). Over these first 38 weeks of the experiment (i.e. age 2–40 weeks) weekly food consumption was measured (g eaten), and animals were weighed (g) every 15 days, and measured (SVL, mm) every 30 days.

To test for the effects of normal and low diets on immune function, blood was drawn at two time points for all animals: at 15 weeks of age (just prior to switching all animals to ad libitum food availability) and again at 30 weeks of age. Approximately 100 μl whole blood was collected from the caudal vein in heparin-rinsed syringes, centrifuged to separate blood components, and plasma was snap-frozen and stored at −80°C for subsequent physiological and immune assays. We conducted three constitutive innate immunity assays: (i) natural antibodies (NAbs), (ii) complement-mediated cell lysis and (iii) bactericidal competence of blood plasma to characterize and quantify non-specific and rapid responses to invading pathogens as a first line of defense. Both NAbs and complement-mediated cell lysis were measured using a hemolysis–hemagglutination assay (Matson et al., 2005), with modifications for use in garter snakes (Sparkman and Palacios, 2009). A series of eight serial two-fold dilutions of 10 μl of plasma were made with phosphate buffered saline (PBS) in a 96-well plate. Each well then received 10 μl of a 2% heterologous sheep red blood cell (SRBC) suspension. All samples were run in duplicate; each plate had both positive (anti-SRBC) and negative (SRBC with no plasma sample) controls. Plates were incubated for 90 min at 28°C and then scored. For both hemagglutination and lysis, titers were estimated as the negative log₂ of the highest dilution factor of plasma. Half scores were given for titers that appeared intermediate. Sheep red blood cells were utilized as the foreign mitogen, as previously demonstrated in snake/reptile studies (Kawaguchi et al., 1978; Sparkman and Palacios, 2009).

Bactericidal competence of snake plasma was assessed (Matson et al., 2006) with minor modifications for use in garter snakes (Sparkman and Palacios, 2009). A pellet of lyphophilized Escherichia coli (Microbiologics, cat. no. 0483E7) was reconstituted using 40 ml PBS. This was further diluted with PBS to produce working solutions that yielded roughly 200 colony-forming bacteria per 10 μl. Plasma samples were diluted 1:10 with PBS. Sample reactions were prepared by adding 10 μl bacterial working solution to 100 μl of the diluted plasma samples. Replicate controls were prepared (one control per every 12 samples) by adding 10 μl of the bacterial working solution to 100 μl PBS. Sample reactions were incubated for 20 min at 28°C, (i.e. the animal maintenance temperature) to provide adequate time for bacterial killing to occur. Duplicate controls and sample reactions were plated in 50 μl aliquots on 4% tryptic soy agar and incubated for approximately 24 h at 28°C, additional incubation at 37°C was used if bacterial colonies were not clearly visible. The number of bacterial colonies on each plate was counted and the percentage of colonies on each plate relative to the mean number of colonies in the control plates calculated. This percentage was subtracted from 100 to obtain the percentage of bacterial colonies killed.

We measured two independent assessments of physiological state: (i) plasma concentration of corticosterone (Greenberg and Wingfield, 1987) – a hormone that is upregulated in both the fight-or-flight response and also under normal food conditions to facilitate foraging, and (ii) the ratio of heterophils-to-lymphocytes (H:L) in whole blood smears as a measure of stress. An increased H:L ratio is indicative of a higher number of heterophils in circulation and an increase in the movement of lymphocytes into peripheral tissues (Dhabhar et al., 1996, 1994).

Circulating CORT (ng ml⁻¹) was quantified in blood plasma to characterize the relationship between baseline hormone levels and early-life condition. CORT levels were quantified with a radioimmunoassay (MP Biomedicals ImmuChem Double Antibody Corticosterone I-125 RIA kit, Irvine, CA; lowest detectable concentration for this kit is 7.7 ng ml⁻¹) following methodology previously adapted for use with snakes (Palacios et al., 2012; Robert et al., 2009; Sparkman et al., 2014). All measurements were made using a single kit with the provided control used to assess intra-assay variability. The control was included in each run and yielded a coefficient of variation (%CV) of 13.0%. All samples were run in duplicate with a %CV<10%. Time between first contact and bleeding was included as a covariate in the statistical analysis of CORT levels, as increases in CORT above baseline measures have been observed in garter snakes after 10 min of handling (Palacios et al., 2012).

To measure H:L, blood smears were prepared from a drop of freshly drawn blood from each animal at the time of bleeding. Smears were fixed in methanol and stained with Wright–Giemsa stain (Fisher Scientific, cat. no. SDWG80). The number and type of leukocytes – lymphocytes, heterophils, eosinophils, basophils or monocytes – were classified by scanning blood smears under 1000× magnification and classifying the first 100 leukocytes encountered. If a heterophil was not viewed within classification of the first 100 leukocytes, the count was extended until either a heterophil was viewed or 200 leukocytes were examined.

After hibernation for 12 weeks, animals were warmed up gradually over 2 weeks, offered food twice (once per week, ad libitum) and mating trials were started (Fig. 2). To test for an effect of normal and low diet on both female and male reproductive output, females and males were randomly paired (subject to non-sibling matings) to construct all pairwise crosses of early-life diet treatment (Table 1). Each female and her designated male mate were placed in individual ten-gallon glass aquaria and observed for 1 hour. If copulation was not observed, females were paired with the same male 5 days later. If both mating opportunities were unsuccessful, a second non-sibling male of the same diet treatment group was paired with the female. If this third pairing did not result in copulation, we stopped mating attempts with this female. Data on successful (i.e. resulted in gravidity) and unsuccessful matings for each treatment combination were recorded and used to test for effects of early-life food availability on male and female mating success. Gravid females were monitored through parturition, and each female's reproductive effort (counts and masses of liveborns, stillborns and unfertilized yolks) and reproductive success (counts and masses of liveborns) were measured. These data were used to test for effects of early-life food availability on maternal reproduction.

For the analysis of body size (SVL, mm), birth SVL was used as a covariate to account for the potential effect of birth size on subsequent body size. For the analysis of growth (change in SVL, mm), we used two time-varying covariates; food consumption during each interval between the two measurements and SVL at the start of each interval to account for slowing of growth with increasing length. For the repeated-measures analyses of physiology, we included the covariate of body condition at each measurement date (15 and 30 weeks). Body condition, calculated as the residuals of the regression of log body weight on log SVL (Weatherhead and Brown, 1996), is known to influence immunological responses of vertebrates, including snakes (Gangloff et al., 2017a; Palacios et al., 2011). Additionally, time between first contact and bleeding was included as a covariate in the statistical analysis of CORT levels, as increases in CORT above baseline measures have been observed in garter snakes after 10 min of handling (Palacios et al., 2012). Samples where blood collection occurred more than 30 min from time of initial handling (N=3) were removed. As in previous studies (Palacios et al., 2011; Sparkman and Palacios, 2009), natural antibodies and complement-mediated lysis were highly correlated (Pearson r=0.921, P<0.0001, N=136), thus only the results for natural antibodies are shown (see Table S2 for complement-mediated lysis results).

Where diet is normal- versus low-food availability and SVL is female body size prior to parturition. Dependent variables included measures of reproductive effort (total litter mass, and a count of total litter size), reproductive success (total mass of liveborn, and liveborn litter size) and individual offspring body mass and size.

Experimental animals consumed food voluntarily, so our first goal was to understand whether animals in the two early-life diet treatments were consuming different amounts of food. Food consumption (analyzed as grams eaten) was significantly affected by the interaction of diet×sex×time (Table 2, plotted as percentage of body mass eaten in Fig. 3A). Animals on the normal diet consumed more food than low-diet animals did during the 15 weeks of the experiment where low-diet animals were given less food. From ages 16–40 weeks, all animals were offered 60% of their body mass in food; low-diet animals initially increased their food consumption to surpass that of normal-diet animals, but by 40 weeks of age, all animals had slowed their consumption (Fig. 3A).

Repeated measures of body size (SVL) was also significantly affected by the interaction of diet×sex×time (Table 1, Fig. 4). Females and males with early-life normal food availability had significantly larger body sizes across time beginning with week 5 and continuing through the end of the experiment at week 40. Growth (change in SVL) was significantly affected by the interaction of diet×sex×time (Table 1, Fig. 3B); the shapes of the body size trajectories varied by sex and by diet over the course of the experiment. During the diet treatment, normal-diet males and females exhibited faster growth compared with levels in low-diet individuals. From age 16 weeks onward (with all animals on the normal diet of 60% body mass), snakes previously on the low diet increased their growth rates to equal or surpass those of the early-life normal-diet animals. Interestingly, they did not quite catch up in size; at hibernation, low-diet animals were still significantly smaller. In addition, following this experiment, animals were housed in standard lab conditions where we subsequently found long-term, sex-specific, effects of early-life nutritional stress on survival (median survival of low-diet females=27.8 months; normal-diet females=40.9 months; all males, regardless of early-life diet=70+ months; see Fig. S1 for survival curves).

Early-life diet and time interacted to affect natural antibodies, bactericidal competence and levels of plasma CORT (Table 1, Fig. 5). For measures of natural antibodies, at the end of the diet restriction (15 weeks of age) agglutination titers were higher in normal-diet animals than in low-diet animals. Whereas, at 30 weeks of age (all animals having been on normal diets for 15 weeks) normal-diet animals decreased their NAbs such that there was no difference between early-life treatments (Fig. 5A). For bactericidal competence, early-life normal-diet animals displayed higher killing capacity at 15 weeks of age. By 30 weeks of age, previously low-diet animals had increased their bactericidal killing capacity such that there were no differences between diet treatments (Fig. 5B). For measures of CORT, at 15 weeks of age, low-diet animals had higher baseline levels of plasma CORT than normal-diet animals. By 30 weeks of age, previously low-diet animals decreased their plasma CORT levels such that there was no difference between diet treatments (Fig. 5C). H:L ratio was affected by time; all animals (regardless of early-life diet treatment) increased their H:L ratio between age 15 and 30 weeks. Additionally, diet and sex interacted such that normal-diet females had higher H:L ratios than all other diet×sex combinations at both time points (Table 1, Fig. 5D).

Of the 30 females that were used for the reproduction trials (one animal died during hibernation), six females did not mate with either their first assigned mate or their second assigned mate (Table 2). Of the 53 unique female–male pairings, 24 resulted in copulation (Table 2). Regarding male mating success, we found a significant effect of early-life diet, such that normal-diet males successfully mated more often than previously low-diet males (G=4.797, d.f.=1, P=0.029). We found no significant difference between early-life normal- and low-diet treatment for female mating success (G=0.007, d.f.=1, P=0.932).

Of the 24 females that copulated with males, normal-diet females were on average larger than low-diet females (Fig. 4). After accounting for this diet-associated variation in body size of females that successfully mated, we found no additional effect of early-life diet on reproductive effort (total litter mass, total litter size) or on reproductive success (total liveborn mass, total number of liveborn). However, early-life diet (after accounting for female SVL) significantly affected both individual offspring mass and length, with normal-diet females giving birth to heavier and longer offspring (Table 3, Fig. 6).

We tested whether early-life nutritional stress leaves its mark on growth, body size, immune function, physiological stress, mating success and female first reproduction. We found effects of early-life diet on growth, body size, mating success and offspring size. We found that immune function and physiological state responded in an immediate manner to food limitation, but values returned to normal as animals were placed on normal diets, with limited support for a longer-term trade-off with growth and reproduction. For most life-history traits, sex significantly interacted with diet and age, which suggests a sex-specific effect of early-life nutritional stress in this species.

We found that an individual's early-life nutritional environment influenced their rate of growth, not only during periods of reduced food availability, but also once resources were increased. Food consumption, growth rate and body size were lower in low-diet snakes for the duration of the diet treatment. When nutritional environment improved, previously limited males and females increased their food consumption and growth rates such that they either matched or surpassed that of their non-stressed conspecifics (Fig. 3). This pattern of increased growth rate in the low-diet treatment group most closely reflects the strategy of compensatory growth (Fig. 1C).

Although low-diet individuals were able to increase their rate of growth above that of normal-diet individuals shortly after alleviation of nutritional stress, we found that low-diet animals of both sexes did not fully catch up in size to normal-diet individuals by the end of the diet experiment (40 weeks of age). Although this absence of complete compensation could be an artifact of the length of the experiment, compensatory growth is not always accompanied by a convergence in growth trajectories. Compensation is often incomplete, with individuals never fully attaining the size of their non-restricted conspecifics (Ali et al., 2003; Jobling, 2010; Mangel and Munch, 2005; Metcalfe and Monaghan, 2001). This plasticity in growth rates suggests that growth is typically regulated at optimal rates, below their potential maxima, such that when selective pressures favor a larger overall size, growth rates can be increased (Arendt, 1997; Dmitriew, 2011; Metcalfe and Monaghan, 2001, 2003; Nylin and Gotthard, 1998).

The ability to increase growth rates can potentially mitigate effects of poor early-life environment, as obtaining a larger size may minimize fitness loss (Mangel and Munch, 2005; Metcalfe and Monaghan, 2001), be advantageous in resource acquisition (Arendt and Wilson, 1997), aid in the avoidance of size specific predators or in buffering individuals against harsh or variable environmental conditions. Although this is not always the case (e.g. advantages of smaller size discussed in Blanckenhorn, 2000), fitness benefits have been extensively studied in snakes (Ford and Killebrew, 1983; Ford and Seigel, 1989) and specifically, in our study species, it has been shown that large body size is positively correlated with measures of fitness (Ford and Karges, 1987; Ford and Seigel, 2015; Seigel et al., 2000). Individuals may also exhibit ontogenetic shifts in their energy allocation such that they preferentially allocate resources to somatic growth during early life and to reproduction when sexually mature (Bronikowski and Arnold, 1999; Norris and Evans, 2000; Stearns, 1989). This may explain why we found patterns of increased food consumption and growth in low-diet animals immediately after resources became readily available followed by all individuals (regardless of sex or early-life treatment) exhibiting a slowing in food consumption and growth as they reached maturation size (Fig. 3).

We found that nutritional restriction during early life had consequences on physiological state and constitutive innate immune function during the time of restriction; low-diet animals had lower immune function and higher circulating CORT levels (Fig. 5, Table 1). In wild garter snakes, we found the same patterns in years where food is scarce: lower innate immune function and higher CORT levels (Palacios et al., 2011, 2012; Sparkman and Palacios, 2009). Interestingly, this is true across ages, sexes and season. Studies examining immunocompetence in nestling birds (Birkhead et al., 1999; Brzek and Konarzewski, 2007; Hoi-Leitner et al., 2001) have also found reductions in various aspects of immune function under conditions of low quantity or quality resources.

In our study, once nutritional stress was alleviated (all animals on normal-diet), we found that as animals increased their growth rates, there was no concomitant reduction in immunocompetence. This lack of a trade-off between investment in compensatory growth and investment into immune function is in contrast to results in damselflies (Stoks et al., 2006) and poultry (van der Most et al., 2011), but is in accordance with work conducted on zebra finches (Killpack et al., 2014). In garter snakes (Palacios and Bronikowski, 2017), measures of adaptive but not innate immune function have been shown to trade off with reproduction. This may suggest that the immune components measured here are maintained at the expense of other more energetically expensive components of immune function. Additionally, the lack of an association between compensatory growth and immune function, such as natural antibodies, may not be altogether surprising as it has been suggested that their production is independent of internal and external stimuli (Ochsenbein and Zinkernagel, 2000). Moreover, a reduction in immunocompetence could negatively affect an individual's fitness and thus resource allocation to immune function may be maintained at the expense of other traits (Norris and Evans, 2000; Palacios and Bronikowski, 2017).

We found little evidence to suggest a trade-off between energy allocation to growth (i.e. compensatory growth) and the physiological parameters measured. In vertebrates, glucocorticoid hormones such as CORT mediate daily and seasonal metabolic processes such as energy acquisition, storage and utilization (Landys et al., 2006; Sapolsky et al., 2000). Upon exposure to adverse events such as increased risk of predation, changes in thermal conditions or food deprivation, activity of the hypothalamic–pituitary–adrenal (HPA) axis is upregulated (Greenberg and Wingfield, 1987), increasing the secretion of CORT into circulation (Gangloff et al., 2016, 2017b; Palacios et al., 2012). Stress-induced levels of CORT mediate energy balance, such that self-maintenance and survival are prioritized over processes such as immunity and reproduction in the short-term (Greenberg and Wingfield, 1987; Romero et al., 2009). Chronically stressed individuals may show sustained increases in baseline CORT levels or changes in the magnitude of response to an acute stress which may impair reproduction or survival long-term (Angelier and Wingfield, 2013; Wingfield, 2013). In this study, after the period of food-restriction, low-diet animals exhibited higher circulating levels of CORT but once nutritional stress was alleviated, we found no association between increased growth rates and CORT levels.

Additionally, glucocorticoids can act to redistribute leukocytes, with heterophils released into circulation and an increase in the movement of lymphocytes into peripheral tissues (Dhabhar et al., 1996, 1994; Vleck et al., 2000). H:L ratios have been shown to positively correlate with various environmental stressors (increases in H:L with increasing magnitude of a stressor; reviewed by Davis et al., 2008) and with levels of circulating glucocorticoids (Gangloff et al., 2017b; Goessling et al., 2015). However, levels of circulating CORT and H:L ratios are not always correlated within individuals (Davis and Maney, 2018; Goessling et al., 2015; Sparkman et al., 2014). We found no effect of early-life diet on H:L ratios and no trade-off with increasing growth rates in low-diet animals. This may be attributed to leukocyte numbers changing more slowly in response to stressors than do CORT levels; these changes are also less variable, longer lasting and often multiple stressors tend to have an additive effect (Vleck et al., 2000), thus changes in H:L ratios might better be considered a ‘downstream’ reaction to more chronically adverse environments (Davis and Maney, 2018).

Poor early-life environments can have negative consequences for fitness, with low resource availability leading to slower rates of growth and resulting in maturation at older ages or smaller sizes (as in Fig. 1A,B) (Auer, 2010). Delaying maturation may decrease fitness because it increases generation time and can also lead to a reduced reproductive lifespan (Roff, 2002) or a reduction in fecundity over time with subsequent reproductive events (Auer, 2010). When considering mating success, we saw no difference between low-diet females and those that had normal diets throughout the experiment. However, low-diet males were less likely to successfully mate (Table 2). This reduction in mating success for low-diet males may suggest that the allocation of resources to growth, following a period of growth depression, differentially trades off with age of maturation for males. This is supported by Morgan and Metcalfe (2001), who found that compensatory growth reduced the occurrence of sexual maturation in male salmon. Alternatively, this reduction in mating success may be due to a reduction in attractiveness, such that following a period of low food availability, males exhibiting smaller size or a suppression in other sexually selected traits were not desirable to females (e.g. Kahn et al., 2012; Livingston et al., 2014; Ohlsson et al., 2002).

When considering the reproductive effort and success of females we found no effect of early-life diet on total litter size or mass; normal- and low-diet mothers had equal litter sizes of equal mass. This is in accordance with findings in fish (Inness and Metcalfe, 2008) and in birds (Criscuolo et al., 2011), where compensatory growth did not affect the latency to produce the first egg or mean clutch size (although it did result in lower mean clutch mass). In our study, early-life diet was strongly correlated with female size (despite compensatory growth, normal-diet females were larger than low-diet females at first hibernation, Fig. 4). Variation in total litter size and litter mass was accounted for by the size of the mother (Table 3), thus we see an effect of early-life diet on female size at maturation and size in turn is a determinate of reproductive output. However, when accounting for differences in body size, we still find a signature of early-life diet on offspring mass and SVL with normal-diet mothers giving birth to larger offspring (both longer and heavier, Table 3, Fig. 6). This is in agreement with work done on Madagascar ground geckos (Paroedura picta), where food-limited females laid smaller eggs more infrequently (Kubička and Kratochvíl, 2009). Additionally, a large number of studies in both avian and non-avian reptiles (Janzen et al., 2001; Krist, 2011), including Thamnophis species (Addis et al., 2017; Gangloff et al., 2017a), show that larger body size at birth or hatching increases early-life survival. Our results suggest that, in females, fitness is negatively affected by trade-offs between increased rates of growth and reproduction.

We demonstrated a compensatory growth response with individuals increasing their food consumption and rate of growth following a period of food restriction. Low-diet animals did not quite catch up in size with their normal-diet conspecifics. We found immediate, but not lasting, effects of early-life diet on immune function and physiological state. Traits in which the two sexes responded differently to early-life stress included mating success and long-term survival. Males differed in reproductive success in relation to early-life experiences such that low-diet males exhibited reduced success. Although females did not differ in their propensity to mate, low-diet females gave birth to smaller offspring. Early-life food restriction had long-term effects not only on female reproduction, but also on survival, with low-diet females exhibiting lower survival than their normal-diet conspecifics. Resource availability, especially during periods of development, is important in determining the life-history trajectory of an organism. It has been widely hypothesized that accelerated growth rates, particularly when following periods of reduced resource availability, may trade off with longevity (Metcalfe and Monaghan, 2001, 2003). Our results suggest that, at least for females, current reproduction may be prioritized over future survival when exposed to adverse early-life conditions such as resource limitation.

We thank J. Bilyea, C. Cates, D. Ford, W. Krogman, and A. Walters for experiment assistance and animal care; D. Byers for technical assistance; E. Gangloff, J. Judson and W. Clark for manuscript feedback and J. Judson for assistance in generating manuscript figures. Animals were maintained at the University of Texas at Tyler's Ophidian Research Colony following IACUC protocol #UTT-006 under Dr Neil Ford.