Maternal provisioning and fluctuating thermal regimes enhance immune response in a reptile with temperature-dependent sex determination

The environment experienced in early life can have important and irreversible effects on phenotype and fitness (Lindström, 1999). In a variety of taxa, for instance, both stress and parental care in early life can influence adult fitness components (Burton et al., 2020; Cohen et al., 2013; Lee et al., 2013). Among ectotherms, it is well established that temperature experienced during embryonic development exerts pervasive effects on phenotypes (Booth, 2006; Deeming, 2004; Noble et al., 2018), and these effects are particularly well studied in reptiles (Noble et al., 2018). The widespread study of thermal traits in the early lives of reptiles arises, in part, because many reptiles exhibit temperature-dependent sex determination (TSD), whereby incubation temperature experienced by an embryo permanently affects its sex (Charnier, 1966).

Theory suggests the adaptive significance of TSD may be related to a differential effect of incubation temperature on male versus female fitness (Charnov and Bull, 1977). Specifically, females are produced at a temperature that confers a fitness advantage relative to males, and vice versa, indicative of a sex-by-environment interaction for fitness; this is known as the Charnov–Bull model (Charnov and Bull, 1977; Warner and Shine, 2008). Because TSD is phylogenetically widespread in reptiles, the Charnov–Bull model has been explored broadly (Valenzuela and Lance, 2004). A handful of studies have shown that the model is supported for fitness-related traits such as body size (Warner et al., 2020) and locomotor ability (Webb et al., 2001), but it is rare that a sex-by-environment interaction for fitness is observed (Warner and Shine, 2008). Further, despite support for the Charnov–Bull model in a few isolated studies, there is a dearth of hypotheses that outline a general mechanism underpinning a sex-by-environment interaction for fitness in reptile species (Bókony et al., 2019; Schwanz et al., 2016; but see Lawson and Rollinson, 2020 preprint; Warner et al., 2009).

Early-life physiological traits that are closely related to lifetime fitness might provide a framework for generalizing the Charnov–Bull model across reptiles (Hulbert et al., 2017). One such trait is the strength of the early-life immune response, which is strongly linked to early-life survival and lifetime fitness (Cichon and Dubiec, 2005; Møller and Saino, 2004; Schneeberger et al., 2014; Watson et al., 2016). Indeed, recent work underlines that the immune response is context dependent and sex specific, highlighting that it is a plausible candidate for a sex-by-environment interaction under the Charnov–Bull model. For instance, a sex-specific immune response in early life has been documented in non-reptilian taxa (Kraaijeveld et al., 2008; Laughton et al., 2011; Love et al., 2008; Tschirren et al., 2003), and sex has been found to influence hatchling immune function in a reptile with genetic sex determination (GSD), and adult immune function in a reptile with TSD (Judson et al., 2020; Palacios et al., 2020). Yet, despite the promising candidacy of immune strength underpinning a sex-by-environment interaction for fitness, it remains unclear how temperature affects immune strength in reptiles. For instance, immune strength is negatively associated with constant incubation temperature in two turtle species (Dang et al., 2015; Freedberg et al., 2008), but it is not clear whether or how this result is relevant under natural temperature regimes where temperatures fluctuate diurnally and vary across seasons (Les et al., 2009; Paitz et al., 2010). Indeed, the Charnov–Bull model, more broadly, is usually tested by comparing fitness from constant-temperature incubations (Janzen, 1995; Warner and Shine, 2005, 2008), despite recognition that these are evolutionarily novel environments and poorly reflect natural environments where temperature fluctuates (Kingsolver et al., 2015; Massey and Hutchings, 2020; Niehaus et al., 2012). Thus, investigating the immune response under incubation regimes that are at least partly naturalized may provide new insight into the origin of a sex-by-temperature interaction for fitness in reptiles.

The immune system may have been largely overlooked in the context of TSD because it is not a phenotype that is readily visible or easily measured. Additionally, many traits affect immune function, making it difficult to directly link immune function, sex and temperature. For instance, because the immune system is energetically costly (East et al., 2015; Leivesley et al., 2019; Lochmiller and Deerenberg, 2000; Sheldon and Verhulst, 1996), immune function is condition dependent, where greater energetic reserves (i.e. high condition) are associated with improved immune strength (Garbutt and Little, 2017). If immune function is also condition dependent in early life (Martínez-Padilla, 2006; Rubolini et al., 2006; but see Krist, 2011), then the amount of maternal provisioning that occurs before sex is determined should exert a strong positive effect on early-life immune strength. Maternal provisioning must therefore be carefully controlled in TSD studies in order to directly link immune function, sex and temperature. Fortunately, unlike for birds and mammals, the contents of reptile eggs usually reflect the entirety of offspring provisioning (Congdon and Gibbons, 1985). Egg mass can therefore be used as an unambiguous proxy for offspring provisioning, which can be leveraged to provide a clear test of whether provisioning per se affects immune strength while also controlling any potentially confounding effects of condition on immune function.

In the present study, we used a reptile with TSD to test whether immune strength exhibits a sex-by-environment interaction that is consistent with the Charnov–Bull model. We incubated snapping turtle, Chelydra serpentina (Linnaeus 1758), eggs at two mean temperatures, with and without thermal variance, using immune strength as a proxy for fitness. Critically, we also decoupled sex and temperature using a hormonal manipulation, allowing us to assess the independent effects of sex and temperature on immune strength. Applying the Charnov–Bull model to our study system, we predicted a relatively greater immune response (more bacteria killed in a bactericidal assay) for males at 24°C than females, and a greater response of females at 28°C than males. We also controlled for maternal provisioning across all environments, thereby holding constant any effect of hatchling condition while also providing an unambiguous test for a role for maternal provisioning in early-life immune function.

The present study is part of a long-term study on snapping turtles at the Algonquin Wildlife Research Station (45°35′N, 78°31′W) in Algonquin Provincial Park, Canada. In 2019, we collected and weighed 302 eggs (9 nests) soon after laying. Eggs were kept in containers with moistened vermiculite (1:1 vermiculite and water) and during the experiment, we added water regularly to bring containers back to their initial weight. We randomly assigned clutches to one of four temperature regimes: 24°C constant, 24±4°C, 28°C constant and 28±4°C in a split clutch design. At 24 and 28°C constant temperatures, we would expect the proportion of males produced to be 0.98 and 0.20, respectively (Massey et al., 2019). To produce male hatchlings at female-producing temperatures (28°C/28±4°C), we applied 95 µg of an aromatase inhibitor (fadrozole, Sigma-Aldrich, St Louis, MO, USA) in 100% ethanol to half of the eggs in each experimental group (Table 1) at the start of the thermosensitive period (day 13 and 22 for the 28 and 24°C regimes, respectively). We estimated the timing of the thermosensitive period using a thermal performance curve for embryo development in Algonquin snapping turtle populations (Rollinson et al., 2018). On the same days, we also applied 5 µl of 100% ethanol to the rest of the eggs as a control. Overall, we created eight experimental groups: either oestrogen-inhibited or control treatments in each of four incubation regimes. All animal work was done in accordance with the University of Toronto Animal Care Committee protocols.

Within 48 h of hatching, we euthanized hatchlings by decapitation and pithing, in accordance with animal use protocol #20011948 from the University of Toronto. On decapitation, we collected trunk blood in a sterile microcentrifuge tube stored over ice. To limit the effect that stress could have on the bactericidal capacity of blood, we collected blood within 10 min of disturbing the individuals (Beck et al., 2017). We centrifuged blood samples at 5000 rpm for 5 min, removed the plasma fraction and froze this at −80°C until we carried out the bactericidal assay. Plasma samples can be frozen at −80°C for up to 6 weeks without affecting plasma complement (Beck et al., 2017). We determined hatchling sex upon dissection and took photos of the gonads. Three researchers determined sex individually and we assigned sex by at least 2 of 3 researchers reaching consensus.

We completed a bacteria killing assay (BKA) following previously reported techniques (Beck et al., 2017). Optimization for hatchlings was necessary to determine the dilution of plasma and concentration of bacteria needed to achieve 50% bacteria inhibition, thereby maximizing the variation detectable. We used a pooled plasma sample from eight hatchlings (one from each treatment) to optimize the BKA. We created three plasma dilutions using sterile phosphate-buffered saline (PBS): 1:3, 1:5 and 1:10 (e.g. 1:10=one part plasma to nine parts PBS); and three concentrations of Escherichia coli (ATCC 8739, E^power microorganisms, Microbiologics, St Cloud, MN, USA); 10⁵, 10⁶ and 10⁷ colony forming units (CFU) per ml. We mixed 72 µl of each plasma dilution with 25 µl of each bacteria concentration in sterile microcentrifuge tubes (i.e. 9 samples – 3 plasma dilutions×3 bacteria concentrations) and included positive controls for each bacteria concentration (25 µl of bacteria mixed with 72 µl of PBS). All samples were incubated at 26°C, the mean incubation temperature in the study, for 30 min to complete the immune challenge. Next, we placed 20 µl of the bacteria mix in 96-well microplates in triplicate and added 140 µl of sterile tryptic soy broth (TSB) to each well; this serves as a food source for E. coli. Positive controls were also placed in triplicate on each plate, with 140 µl of sterile TSB. We incubated microplates at 37°C and measured absorbance at 600 nm at 6, 8, 10 and 12 h of incubation with a microplate reader (Synergy HTX, BioTek, Winooski, VT, USA). Absorbance is positively related to bacterial concentration; thus, we averaged the absorbance of each sample and calculated the proportion of bacteria killed as 1−(sample absorbance/positive control absorbance).

We followed the above method using 1:3 dilution of plasma and 10⁶ CFU ml⁻¹ of E. coli (see Results). Briefly, we incubated a mixture of 72 µl diluted plasma and 25 µl of E. coli at 26°C for 30 min. Then, we placed 20 µl of the bacteria mix in 96-well microplates in triplicate and added 140 µl of TSB to each well, including positive controls. We incubated microplates at 37°C and measured absorbance at 600 nm after 8 h of incubation with a microplate reader (Synergy HTX, BioTek). We averaged the absorbance of each sample and calculated proportion of bacteria killed as 1−(sample absorbance/positive control absorbance), as above. All BKAs were run on the same day.

Overall, 180 out of 302 eggs successfully hatched (Table 1), and 167 had sex determined. Egg mortality was not related to incubation temperature (t_4.062=−1.639, P=0.175), temperature type (t_3.305=−2.329, P=0.094) or treatment (t_5.534=−0.187, P=0.858). Ninety-four males hatched across the four treatments, with a mean±s.d. body mass 10.9±0.94 g and carapace length 28.3±1.75 mm. Seventy-three females hatched and had a mean±s.d. body mass of 10.8±1.1 g and carapace length of 27.9±2.26 mm. The sex manipulation was effective, as the sex ratio was male biased in three of the four hormonally manipulated temperature treatments, and in the fourth, the number of males was doubled relative to the control (Table 1).

Three combinations of bacteria concentration and plasma dilution produced close to 50% bactericidal capacity. These were 10⁵ CFU ml⁻¹ and 1:3 dilution at 6 h, 10⁵ CFU ml⁻¹ and 1:5 dilution at 6 h, and 10⁶ CFU ml⁻¹ and 1:3 dilution at 8 h, which resulted in a bactericidal capacity of 40%, 41% and 62%, respectively (Fig. 1). The 10⁶ CFU ml⁻¹ and 1:3 dilution at 8 h was chosen for the BKA as this allowed sufficient time for bacterial growth to take place. Other studies employing similar methods have also allowed 8 h of incubation for E. coli to grow after an immune challenge (Beck et al., 2017).

We successfully extracted blood from 119 hatchlings for use in the BKA. Intra-assay coefficient of variation (CV%) was 19.2, while inter-assay CV% was 22.7. Hormonal manipulation did not significantly affect the proportion of bacteria killed (β_Control=−0.049, s.e.=0.224, z=−0.216, P=0.829) and was not included in further analyses. Egg mass was significantly positively associated with the proportion of bacteria killed (Table 2, Fig. 2A); however, in the model replacing egg mass with maternal carapace length, maternal carapace length did not explain variation in the proportion of bacteria killed by hatchling plasma (z=0.812, P=0.417). Neither mean temperature nor sex was significantly associated with the proportion of bacteria killed, but temperature variance was positively associated with the proportion of bacteria killed (Table 2; z=3.00, P=0.003). Hatchlings incubated under fluctuating conditions had an immune response that was nearly 2 times stronger (mean±s.e. 0.453±0.044 bacteria killed) than that of those incubated under constant regimes (0.283±0.048 bacteria killed; Fig. 2B).

There are three major findings from the present study. First, we did not find support for the Charnov–Bull model of TSD evolution and maintenance using a single measure of functional immune response. Second, snapping turtles raised in constant temperature environments had a lower functional immune response than those from more naturalized environments, independent of mean temperature, thereby questioning the relevance of studies that use constant-temperature incubation to understand ecologically relevant phenotypes. Finally, the study demonstrates a clear and positive association between egg mass, a measure of maternal provisioning, and a measure of early life functional immune response in an ectothermic vertebrate.

A major goal of the present study was to explore whether the evolution of TSD may be related to early-life functional immune response, a phenotype that is rarely measured in TSD reptiles. Under the Charnov–Bull model, we predicted a relatively greater immune response for males at 24°C than females, and a greater response of females at 28°C than males (Charnov and Bull, 1977). Yet, despite a reasonable sample size (N=11–62 per treatment; Table 1), no such relationship was observed, nor was a main effect of sex observed. It is therefore possible that the immune response is an important aspect of an individual's early-life survival (Cichon and Dubiec, 2005; Møller and Saino, 2004), regardless of the sex of hatchlings, and that any sex specificity in functional immune response emerges later in life.

One version of the Charnov–Bull model, the ‘survival to maturity’ hypothesis, does not depend on a sex-by-temperature interaction in immune function. Instead, it suggests that temperature affects annual survival (i.e. immune function) equally in the two sexes and that the sex-by-temperature interaction on lifetime fitness and selection for TSD arises if temperature alone impacts immune function and the sexes differ in age at maturity (Schwanz et al., 2016). The logic is that the sex maturing later will experience relatively more mortality prior to maturity and will therefore experience a greater fitness benefit (greater juvenile survival) from developing at temperatures that confer high survival rates. The survival to maturity hypothesis has previously received some empirical support (Bókony et al., 2019), and it is applicable to snapping turtles, where males mature later than females (Steyermark et al., 2008). Further, the hypothesis is indeed in line with our finding of no difference in functional immune response between male and female hatchlings. However, under this hypothesis, we might expect to observe an effect of mean temperature on functional immune response, which we found no evidence for. Interestingly, we did find that temperature fluctuations enhanced the bactericidal capacity of blood compared with constant temperatures. It follows that the survival to maturity hypothesis would be supported if temperature fluctuation is associated with male-producing temperatures in the wild, but whether this is true is unknown. More broadly, we did not find any evidence that functional immune response is related to models of TSD evolution, but we underline that we used only one measure of immune response. It may therefore be premature to discount the possibility that immune response plays a role in the maintenance of TSD, and future studies on TSD may benefit from the use of multiple markers of immune response.

Two previous studies investigated how temperature fluctuations alter immune response compared with constant temperatures (Les et al., 2009; Paitz et al., 2010). Both used delay-type hypersensitivity tests to quantify the swelling response of painted turtle (Chrysemys picta) hatchlings, but the studies found mixed results. Les et al. (2009) found that a ±3°C temperature fluctuation did affect the swelling response (which is a measure of immune activity at a site of potential infection). However, the mean temperatures compared by Les et al. (2009) differed by 8°C and appeared to approach the thermal limits of their study species. In contrast, Paitz et al. (2010) found no difference in swelling response between constant temperature, and ±4°C and ±8°C fluctuating treatments. Our study may provide insight into these contradictory results. Specifically, while limiting the degree of mean thermal difference between treatments, we found support for the notion that fluctuating temperature enhances the immune response. We therefore suggest that temperature fluctuations may have a transient effect on the immune response: Paitz et al. (2010) measured immune strength at 90 days post-hatching, whereas Les et al. (2009) measured it at 21 days post-hatching and beyond, and the present study measured it immediately after hatching. Thus, our study provides support for Les et al. (2009) and suggests the effect is transient, which may still cause important variation in early-life survival.

Better-provisioned offspring have advantages such as starvation resistance, increased gape size and foraging efficiency, and increased ability to avoid predation (Marshall et al., 2018; Rollinson and Hutchings, 2013a,b), all of which translate into enhanced survival (Rollinson and Rowe, 2015). As the immune system is energetically costly (East et al., 2015; Leivesley et al., 2019; Lochmiller and Deerenberg, 2000; Sheldon and Verhulst, 1996), greater energetic reserves should allow better-provisioned offspring to invest more in immune system development and maintenance. Indeed, we found egg mass to be significantly positively associated with bactericidal capacity. As no parental care is provided in snapping turtles, egg provisioning represents total maternal investment in offspring, and thus we have arguably provided a relatively direct test of whether maternal investment in offspring affects the immune response. However, we did not measure maternal antibodies in the eggs or maternal hormones in the eggs, nor did we experimentally manipulate egg size, so we cannot discount the possibility that egg mass is confounded with an unmeasured physiological variable. Nevertheless, there are two reasons we believe it is the energetic reserves in the egg driving the association between egg size and immunity. First, there is no evidence to date that egg mass covaries with maternal antibody provisioning. For instance, female flycatchers that are in greater general health may provision eggs with more antibodies, although these females do not produce relatively larger eggs (Hargitai et al., 2006). Further, across 16 wild populations of flycatchers, egg mass and antibody levels were uncorrelated (Ruuskanen et al., 2011), suggesting antibody provisioning is unrelated to egg size per se. Second, it is possible that variation in yolk hormone concentrations (especially in androgens) could have influenced our results, as yolk hormones have previously been linked to differences in offspring size, growth, behaviour and immunity (Bowden et al., 2004; Schwabl, 1993, 1996; Uller et al., 2007). In particular, recently matured female painted turtles (Chrysemys picta) produce relatively small eggs with high testosterone concentration (Bowden et al., 2004), suggesting that hormonal differences rather than provisioning differences can explain the positive association between egg size and immune response. It seems more likely that our result is explained by provisioning, however, as adult body size did not explain immune strength in our study. Our results are therefore in line with theory (Smith and Fretwell, 1974), and suggest that immune strength is an additional axis on which maternal investment in egg size influences neonate quality (Garbutt and Little, 2017; Krist, 2011; Martínez-Padilla, 2006; Pelayo and Clark, 2002; Rollinson and Rowe, 2015; Rubolini et al., 2006).

In sum, the present study does not support a role for early-life immune response in the evolution and maintenance of TSD, although future studies could be strengthened by using multiple markers of immunity to explore the temperature sensitivity of the immune response. We do, however, show that small fluctuations around mean temperatures likely have beneficial effects on hatchling fitness, and we suggest that these effects of temperature fluctuation on immunity are important but transient. This finding is interesting to consider given that studies of reptilian embryos still overwhelmingly use constant temperature incubation regimes to study reptile ecology, or conserve reptiles through artificial egg incubation and hatchling release. Finally, we also show that variation in maternal investment per offspring, through egg provisioning, is positively associated with variation in functional immune response. Importantly, maternal provisioning is not confounded with other sources of information transfer between parent and offspring, and we suggest our result may be generalizable to other groups without parental care.

We are grateful to Megan Frederickson and Marla Sokolowski for the use of lab space and equipment, and to Hayley Vleck, Mariel Terebiznick, Jared Connoy and Pallavi Pal for helping with egg incubation and hatchling sexing. We thank Anne Bronikowski and an anonymous reviewer for excellent comments that greatly improved a previous version of the manuscript. Accommodation was provided by the Algonquin Wildlife Research Station.