Additive effects of temperature and water availability on pregnancy in a viviparous lizard

Predicting how organisms will be impacted by global climate change is frequently viewed in the context of thermal effects on biological and ecological processes (Huey et al., 2012; Kearney et al., 2009; Nowakowski et al., 2018; Parmesan, 2006). However, it is equally important to consider the effects of water availability on biodiversity because water is a fundamental resource, rainfall events are forecasted to be less reliable, resulting in reduced water availability for many species (Marvel et al., 2019; Schlaepfer et al., 2017), and predicted changes in ambient temperatures are inextricably linked to changes in precipitation (Alexander et al., 2006). For example, during periods of prolonged drought, increasing temperatures and decreasing water availability frequently covary (Rahmstorf and Coumou, 2011), which can have a pronounced effect on hydration and survival (McLaughlin et al., 2002; Riddell et al., 2019). These combined effects are especially relevant in ectotherms, which rely on behavioral or physiological selection of optimal body temperatures that maximize performance, survival and reproduction (Huey, 1982; Angilletta et al., 2002; Huey et al., 2012; Sinervo et al., 2010), but risk losing greater volumes of water when maintaining higher body temperatures (Claussen, 1967; Munsey, 1972). Even though dual changes in temperature and water availability are predicted across the globe, it remains largely unknown if and how fundamental life-history stages, such as reproduction, will be affected by these combined changes (Rozen-Rechels et al., 2019).

In many environments, precipitation is already unevenly distributed throughout the year or water is rare, and water availability can therefore be limited during periods of reproductive investment (Hao et al., 2018). Previous research has found that changes in rainfall patterns, independent of temperature shifts, are currently challenging reproductive strategies (Visser and Both, 2005) and can have negative impacts on entire ecosystems (McCluney et al., 2012). In juxtaposition, shifts in temperature, independent of altered rainfall patterns, can drastically reduce fitness (Deutsch et al., 2008), and unpredictable temperature regimes have been implicated in large-scale extirpation events (Valladares et al., 2014). The negative effects of fluctuating temperature and water availability can have interactive, detrimental consequences on reproductive events. The physiological impacts of maternal hydration before and after pregnancy have previously been explored (Nelson et al., 1989; Hanson et al., 1994; Gesquiere et al., 2008; Bukovetzky et al., 2012), but far less is known about the direct consequences of hydration during gestation on maternal physiology and reproductive output. Some recent studies have found that reproductive females with limited access to water are able to tolerate dehydration (i.e. hyperosmolarity; Brusch et al., 2017; Dupoué et al., 2018a) or can utilize alternative hydric resources during gestation (e.g. bound water; Brusch et al., 2018) to buffer themselves from changes in water availability. However, these past studies focused on a single physiological challenge, water deprivation, and it is unclear how multiple, concurrent challenges, as predicted under climate change scenarios, may influence maternal physiology during reproduction. Recent work has started to emphasize potential interactions between temperature and water availability in ectotherms (Dupoué et al., 2020a; Rozen-Rechels et al., 2019). However, a combined, experimental approach exploring the relationship between thermoregulation and hydroregulation during reproduction is currently lacking.

To further explore how temperature and water availability impact maternal physiology and reproductive output, we used the viviparous form of the European common lizard (Zootoca vivipara) in a two-by-two factorial design manipulating jointly hydric and thermal conditions, for the first time. We investigated how water availability (±water restrictions), temperature (hot or cold daily cycle) or a combination of the two throughout gestation influence maternal physiology and reproductive output. Unique to our experimental design relative to most thermal biology studies, we used similar water vapor deficits between temperature treatments to eliminate the possibility of increased desiccation at higher temperatures, and therefore only temperature was manipulated. The common lizard is a wide-ranging, cold-adapted squamate that is sensitive to water deprivation (Lorenzon et al., 1999; Reichling, 1957). Higher temperatures can accelerate embryonic development, as in other viviparous lizards (Du and Shine, 2015), while simultaneously increasing metabolic and subsequent water loss rates of gravid females and consequently negatively impacting maternal physiological condition if the gravid females are not able to compensate with greater food intake (Dillon et al., 2010; Lourdais et al., 2017). Conversely, if lower temperatures delay embryonic development, the prolonged gestational period can exacerbate the effects of water restrictions and similarly reduce maternal condition. To ground truth our study in ecological relevance, we used contrasting water availability treatments that mimic the natural conditions of wild populations with either limited or permanent access to water (Dupoué et al., 2017; Lorenzon et al., 1999). We also selected temperature treatments that reflect either warm, favorable daily conditions for embryonic development (long access to preferred temperature without thermal stress) or cold, suboptimal daily conditions (short access to preferred temperature), and which correspond to the thermal environments experienced by this species in their typical range, excluding extremely hot or cold environments. We tested the hypothesis that contrasted temperature regimes and water availability will differentially impact aspects of maternal physiological condition and reproductive output. We used body mass and condition, tail width, plasma osmolality, oxidative stress, antioxidant capacity and food consumption as measures to evaluate maternal physiological condition. We used gestation length to evaluate reproductive phenology. Finally, we used total litter size and mass, and individual offspring size and mass as measures for reproductive output. We made the following predictions: (1) water restrictions will negatively affect maternal traits but will not impact reproductive output; (2) high temperatures will shorten pregnancy duration but negatively impact maternal physiological condition; and (3) a combination of high temperature and water restrictions during pregnancy will negatively affect both maternal traits and reproductive output.

Zootoca vivipara (Lichtenstein 1823) is a small-bodied lizard species [adult snout–vent length (SVL), 50–75 mm] in the family Lacertidae with the widest geographic range of all terrestrial reptiles (Dely and Böhme, 1984). The species is predominately found in cooler, mesic habitats with permanent access to water across northern Eurasia and occurs in both viviparous and oviparous forms. At our study sites, all females were viviparous. As lecithotrophic viviparous species, the majority of maternal energetic investment occurs during vitellogenesis (Blackburn, 2015). After ovulation, embryos remain enveloped in a thin membrane throughout gestation (Heulin, 1990; Panigel, 1956), which supports important maternal water transfers (Dauphin-Villemant et al., 1986) and also transfers calcium and some micronutrients to the developing embryos (Stewart, 1992; van Dyke and Beaupre, 2012). The young of viviparous females are usually born in the membranes and hatch within 1–2 h of parturition (Lorenzon et al., 2001). Females emerge from overwintering in late April to early May, with mating and fertilization occurring shortly thereafter (Bleu et al., 2013). Parturition occurs from late June through July, and females typically give birth to six neonates (range, one to 12), depending on their body size (Dupoué et al., 2018a; Foucart et al., 2014).

Females used for this study were captured in the first week of June 2018 from five different sites on the Plateau de Millevache (Limousin, France) separated by a maximum distance of 30 km (Table S1). Gravid females (N=97) were captured by hand and reproductive status was confirmed using abdominal palpation. Non-reproductive females were immediately released at the site of capture. Within 72 h of capture, all females were moved to the Centre d'Etudes Biologiques de Chizé (Villiers-en-Bois, France). Mass and SVL of lizards were recorded, after which all visible ectoparasites were removed with forceps and females were treated with an anti-parasite spray (Frontline, Merial Inc., Duluth, GA, USA). Females were then housed individually with a 12 h light:12 h dark regimen in opaque containers (30×18×10 cm) filled with ∼1 cm of soil and two different shelters, and provided access to food (house crickets, Acheta domestica) and water ad libitum. During the 8 h light cycle, a subsurface heating element was provided below one end of each cage to create a thermal gradient (20–40°C). All females were held in these conditions for ∼7 days prior to the start of the experiment, after which an initial blood sample and scanned image were collected (see below). Procedures were performed in accordance with laws relative to the capture, transport and experimental use of Z. vivipara (DREAL permit #13016_19042018) and approved by an independent ethical committee (Apafis #2018060111033048_v5).

Experiments were performed from 15 June 2018 until parturition for all females so as to mimic stable and consistent differences in temperature and water availability throughout the majority of gestation. Pregnant females maintain stable body temperatures during the day (range, 29–33°C) by selecting appropriate micro-climates (Le Galliard et al., 2003; Dupoué et al., 2017, 2018b). We assigned lizards of similar mean body size to one of two water regimens with permanent access to water or a water restriction (hereafter referred to as wet and dry, respectively), following similar protocols used in previous studies (Dupoué et al., 2018a; Lorenzon et al., 1999). Females in the wet group (N=49) were provided with water ad libitum, and water was sprayed twice per day (morning and afternoon) on one wall of each container so that drops could be used for drinking up to 1 h after spraying. Females assigned to the dry group (N=48) were only provided with water once per day (morning) by similarly spraying one wall of each container. Lizards were further divided into one of two temperature regimens, hereafter referred to as ‘cold’ or ‘hot’. Each day of the experiment, females in the cold group (N=49) were held overnight (17:00 h to 09:00 h; 16 h) at 15°C, during a portion of the day (11:00 h to 15:00 h; 4 h) at 31°C, and spent 4 h where temperatures were either gradually increasing or decreasing to avoid abrupt changes. Females in the hot group (N=48) were held overnight (19:00 h to 07:00 h, 12 h) at 22°C instead of 15°C, during a majority of the day (09:00 h to 17:00 h, 8 h) at 31°C, and similarly were held for the remaining 4 h with temperatures either gradually increasing or decreasing (Fig. 1). Final treatment groups of the factorial design were thus hot and dry (N=24), hot and wet (N=24), cold and dry (N=24), or cold and wet (N=25). Stable temperatures were maintained by housing females of similar treatment groups in climatic chambers (Vötsch VP 600, Balingen, Germany). The relative humidity in each chamber was adjusted depending on the temperature so that the water vapor deficit remained constant at ∼1 kPa (equivalent to 60% relative humidity at 20°C).

During the study, each female was provided every other day with live crickets (400±20 mg), which were dusted with vitamin powder containing calcium and vitamin D3. During feedings, any remaining crickets were removed and counted in order to quantify the total number and ratio of crickets consumed (i.e. number of crickets consumed/crickets offered) throughout the study. Females were weighed once per week and each cage was visually inspected twice a day for parturition. After parturition, females and their litter (total mass of living and stillborn offspring) were weighed, and a final blood sample and scanned image were collected. Offspring were classified as either living or stillborn in order to quantify the proportion of stillborn (i.e. number of stillborn/total offspring) for each female. Living neonates were also individually weighed and scanned. Females and neonates were placed into separate individual containers, and provided a thermal gradient (20–40°C) and access to food and water ad libitum, before being released at the maternal site of capture.

We used 20 µl hematocrit tubes to collect a ∼50 µl blood sample from the postorbital sinus of all reproductive females. Total time for capture, restraint and blood collection did not exceed 3 min. We immediately centrifuged the blood samples at 4000 g for 5 min at 20°C to separate plasma from blood cells, and then aliquoted plasma (∼20 µl) into separate vials, which were frozen at −30°C until we used them to measure plasma osmolality and oxidative status. We determined plasma osmolality using a vapor pressure osmometer (±3 mOsm kg⁻¹; model 5100C; Wescor Inc., Logan, UT, USA). To allow for duplicate readings, plasma samples were diluted (1:1) in reptile Ringer's solution (300 mOsm kg⁻¹) following methods from Secor et al. (1994). Prior to use, we calibrated the osmometer using sealed osmolality standards (100, 290 and 1000 mOsm kg⁻¹), in accordance with factory recommendations. We ran samples in triplicate and used 290 mOsm kg⁻¹ standards to check the osmometer for variation after every sample. If the standard varied more than the limits of the osmometer (±3 mosmol kg⁻¹), the osmometer was recalibrated. If it continued to vary more than the limits, the osmometer head was cleaned and the machine was recalibrated before continuing duplicate analysis, beginning with the last sample prior to calibration/cleaning to verify correct measurement of the sample.

In order to examine oxidative status, we performed two plasma-based assays within 5 months of blood collection. To evaluate the activity of organic hyperoxides (an index of oxidative damage; Costantini, 2016), we measured the concentration of reactive oxidative metabolites (ROMs) using d-ROMs colorimetric kits (MC003, Diacron International, Italy). Antioxidant capacity (OXY) was also assessed to determine the non-enzymatic ability of diluted female plasma samples (1:100) to neutralize an oxidant attack from hypochlorous acid (Costantini, 2011) using OXY-absorbent test kits (MC435, Diacron International, Italy). For all assays, we used a pooled plasma sample of randomly selected females (N=8) three times in each 96-well plate to measure coefficients of variation in ROMs (intra-plate, 2.8%; inter-plate, 6.0%) and OXY (intra-plate, 3.4%; inter-plate, 13.8%).

Prior to the experimental manipulation and just after parturition, the ventral surface of all females was scanned at 600 dots per inch using a flatbed scanner (Hewlett-Packard Co., ScanJet 3670). From these images, we measured tail width at the seventh subcaudal scales using Inkscape (v.0.92.3). We also scanned each neonate just after parturition to accurately measure their SVL. Animals were gently restrained by hand behind the neck and the tip of the tail during image collection.

Before testing the effects of water availability and temperature on maternal traits and reproductive output, we identified a set of explanatory variables to model data from reproductive females. Larger females typically emerge from overwintering and ovulate, and females with earlier ovulation dates give birth earlier in the year (Bauwens and Verheyen, 1985). Therefore, we included parameters on relative body size (individual SVL – mean SVL of all females) and relative duration of treatment (individual time in treatment – mean time in treatment of females within respective treatments). We did not directly manipulate litter size or mass (i.e. reproductive effort) and therefore included a variable describing individual differences in reproductive effort (standardized residuals from a linear regression of litter mass against SVL) to control for potential allocation differences between females (Dupoué et al., 2020b). We also explored potential interactions between reproductive effort and treatments to test for different allocation trade-offs depending on temperature and water availability. However, these interactions were never significant (all P>0.05) and therefore removed from all models. Starting with a full model, we used stepwise removal of insignificant variables (Arnold, 2010; Zuur et al., 2010).

We used linear models to examine the effects of water availability (wet or dry) and temperature (hot or cold) on mass and tail width changes and changes in plasma osmolality in reproductive females throughout the experiment. We used temperature, water availability and their interaction as fixed effects, and relative body size and reproductive effort as covariates. To model the duration that females were in the experiment and differences in litter size and mass, we used models with only relative body size as a covariate. Similarly, we used linear models to compare differences in initial and final OXY and only used relative duration as a covariate. Models used to investigate differences in initial and final ROMs included the same fixed effects and interactions but did not include any covariates. We used general linear models for ratio and proportion data (crickets consumed and number of stillborn), with temperature, water availability and their interaction as fixed factors. For ratio of crickets consumed, we used relative body size, relative duration and reproductive effort as covariates, and for proportion of stillborn, only relative duration was used as a covariate. For offspring data (mass and SVL), we used linear mixed-effect models with maternal treatments (temperature, water and their interaction) as fixed factors, SVL or mass as covariates, and maternal identity as a random factor to control for non-independence among offspring from the same litter. All analyses were performed using R software (R Development Core Team, version 3.5.2) using the packages ‘nlme’ (https://CRAN.R-project.org/package=nlme), ‘lattice’ (https://CRAN.R-project.org/package=lattice), ‘car’ (https://cran.r-project.org/package=car), ‘lsmeans’ (https://cran.r-project.org/package=lsmeans) and ‘MASS’ (Fox and Weisberg, 2011; Sarkar, 2008; Venables and Ripley, 2002). Data are presented as mean±s.e.m. and differences were accepted as significant at the level of P<0.05.

When comparing the duration that females were in their experimental treatments (time until parturition), we found a significant main effect of temperature regime (F_1,92=172.04, P<0.001), but no main effect of water availability or interaction between the two kinds of treatment (P>0.05). Irrespective of water availability, reproductive females exposed to hot temperatures gave birth sooner than those exposed to cold temperatures. Comparing duration of treatment and the relative body size of females further revealed a significant interaction between relative body size and treatment (F_1,93=13.31, P<0.001). The relationship between treatment and relative body size was more pronounced in reproductive females exposed to cold (β=−1.68, s.e.=0.24, R²_adj=0.497) than in those exposed to hot (β=−0.73, s.e.=0.10, R²_adj=0.499; Fig. 2) temperatures.

Females at the beginning of the experiment had similar body mass between treatments (F_3,93=0.16, P=0.93; Table S2). Mass change during the experiment (pre-parturition mass – initial mass) was additively influenced by water availability (F_1,91=8.39, P=0.004) and temperature regime (F_1,91=32.48, P<0.001), but we found no interaction between temperature and water availability (F_1,91=0.82, P=0.366). Females exposed to hot and wet conditions gained less mass during gestation compared with the females in all other treatment groups (Fig. 3C). Similarly, post-partum body condition (body mass adjusted to the mean body size of females: F_1,91=186.75, P<0.001) was influenced by water availability (F_1,91=4.12, P=0.045) and temperature treatment (F_1,91=17.57, P<0.001), with females exposed to hot and dry conditions having the lowest post-parturition body condition (3.21±0.12 g), and females exposed to cold and wet conditions having the highest post-parturition body condition (3.73±0.17 g). We found no interaction between temperature and water availability for post-parturition body condition (F_1,91<0.01, P=0.94).

There was no initial difference in tail width among the treatment groups (F_3,93=0.11, P=0.95; Table S2). Most females had reductions in tail width during the experiment, and these decreases were additively influenced by water availability (F_1,91=17.23, P<0.001) and temperature treatment (F_1,91=5.75, P=0.018), with females exposed to hot and dry conditions losing the most tail width and females exposed to cold and wet conditions losing the least tail width (Fig. 3A). We found no interaction between temperature and water availability (F_1,91=0.49, P=0.486).

Plasma osmolality change during the experiment was additively influenced by water availability (F_1,74=16.26, P<0.001) and temperature treatment (F_1,74=7.62, P=0.007), with females exposed to hot and dry conditions having the largest increases in plasma osmolality (308±2 mosmol kg⁻¹ to 323±3 mosmol kg⁻¹ for mean initial and final values, respectively); females in hot and wet, and in cold and dry, conditions having intermediary increases in plasma osmolality (305±2 mosmol kg⁻¹ to 314±2 mosmol kg⁻¹ and 311±2 mosmol kg⁻¹ to 318±1 mosmol kg⁻¹, respectively); and females exposed to cold and wet conditions having decreased plasma osmolality (323±3 mosmol kg⁻¹ to 315±3 mosmol kg⁻¹; Fig. 3B). We found no interaction between temperature and water availability (F_1,74=2.04, P=0.157).

There was no initial difference in oxidative stress (ROMs; F_3,93=0.40, P=0.75) or non-enzymatic antioxidant capacity (OXY; F_3,93=1.32, P=0.27) among the treatment groups. After parturition, females in dry conditions throughout pregnancy had significantly higher ROMs (F_1,93=8.34, P=0.005), but we found no significant interaction or main effect of temperature regime (P>0.05; Fig. 4A). When examining OXY values after parturition, we found a significant interaction between temperature and water availability (F_1,92=6.39, P=0.013). Within the hot treatments, females in wet conditions had higher OXY values compared with those in dry conditions. In juxtaposition, females exposed to cold and wet treatments had slightly lower OXY values compared with those in cold and dry treatments (Fig. 4B).

When evaluating the ratio of crickets consumed during the same period, we found a significant main effect of water availability (F_1,94=4.16, P=0.044) and temperature regime (F_1,95=36.43, P<0.001) but no interaction between the two (F_1,90=2.76, P=0.100). A Tukey's honestly significant different (HSD) post hoc test revealed that females in the hot temperature regimes ate a greater ratio of crickets than those in the cold regimes (P<0.01), irrespective of water regime, and within each temperature regime, females in wet conditions consumed a higher proportion of crickets, although not significantly (P>0.05; Fig. 5).

We did not detect any significant differences in litter size (F_3,93=0.49, P=0.688), total litter mass (F_3,92=0.49, P=0.688), individual offspring mass (F_3,92=1.08, P=0.363) or SVL (F_3,92=0.35, P=0.786) among the treatment groups. When comparing the proportion of stillborn offspring among treatment groups, we found a significant effect of temperature (F_1,95=6.69, P=0.011), where females exposed to cold temperatures had three times more stillborn offspring compared with females exposed to hot temperatures, irrespective of water regime (P<0.05; Table 1).

Understanding how animals cope with both altered temperature regimes and restricted resources, such as water, will enable us to better predict how they might be affected by anticipated climate change scenarios. This is especially important because in many environments climate warming induces concurrent changes in mean temperature and water availability, and increases the occurrence of extreme weather events (Fischer and Knutti, 2015; Mann et al., 2017; Rahmstorf and Coumou, 2011). Ours is the first study to explore the independent and combined effects of water availability and thermal conditions during gestation in a vertebrate. Our results regarding water availability are mostly consistent with previous work in squamate reptiles: mild dehydration during gestation negatively affects maternal physiology but has little effect on reproductive output (Dupoué et al., 2015; Lourdais et al., 2015). For all traits except antioxidant capacity of females, we also found that these effects are additive to temperature regimes and therefore similar in warmer and colder environments, despite very significant effects of thermal conditions on the length of gestation and maternal physiology. Overall, our study adds to a growing body of evidence that water is a fundamental currency used during reproduction. Our study also emphasizes the importance of considering multiple, concurrent physiological challenges when investigating organismal responses to climate changes, especially during fundamental life-history stages such as reproduction.

We found that temperature, but not water availability, affected gestation length, with higher temperatures resulting in a shorter period of pregnancy (Fig. 2). In both temperature regimes, smaller females had a longer duration of pregnancy, although this size-dependent gestation length was far more pronounced in the cold than in the hot treatments. Maternal body size is often an important factor influencing reproductive output and timing (Blueweiss et al., 1978; Díaz et al., 2012; Iverson et al., 1997; Kiefer et al., 2008), including in this species (Bauwens and Verheyen, 1985). Higher temperatures in our study appear to have accelerated gestation and partially outweighed the size-dependent timing of pregnancy, suggesting that small females caught up for initial differences in the timing of reproduction when provided with warmer environmental conditions. These results are in accordance with previous research showing that higher maternal temperatures result in shorter gestation length and faster embryonic development, both in oviparous and viviparous squamates (Dubey and Shine, 2011; Foucart et al., 2018; Lorioux et al., 2012, 2013a). They also confirm comparative findings among wild populations showing that climate warming induces phenological acceleration (e.g. shorter gestation time and lower variance in laying date) and significant demographic changes (Massot et al., 2017; Rutschmann et al., 2016).

Previous field and laboratory studies further suggested that increased gestational temperatures can also impact reproductive output, offspring phenotypes and future fitness (Chamaille-Jammes et al., 2006; Lorioux et al., 2013b; Lourdais et al., 2004; Ma et al., 2014; Marquis et al., 2008). Although we did not investigate effects on future reproduction and survival, we detected no significant differences in total litter size and mass, and offspring metrics between any of the experimental groups, but found that females had a poorer embryonic success in the cold compared with hot treatments. In ectotherms, embryonic metabolism is temperature dependent, and mothers often behaviorally thermoregulate to maintain optimal temperatures for embryonic development (Farmer, 2003; Gillooly et al., 2002; Lorioux et al., 2013a; Lourdais et al., 2013; Mettouris et al., 2017). Extreme temperatures are often associated with decreased reproductive success in ectotherms because they compromise behavioral thermoregulation and may increase physiological stress and decrease performance when they cannot be avoided behaviorally (Clusella-Trullas et al., 2011; Sinervo et al., 2010; Wang et al., 2016). In our study, females were not allowed to thermoregulate but had longer access time to preferred temperature (31°C) in the hot than in the cold treatments. The hot treatment temperature (31°C) was within the range of preferred temperature during gestation and was not meant to induce thermal stress or heat-related embryonic mortality (Le Galliard et al., 2003). Thus, embryonic success was maximized in hot conditions probably because gravid females were given better opportunities to maintain their preferred body temperatures during the prolonged daytime period (Foucart et al., 2018; Le Galliard et al., 2010; Le Henanff et al., 2013; Lorioux et al., 2012, 2013a). Although females in the cold treatments were provided access to preferred temperatures for 4 h each day, that window does not appear to have been large enough to effectively maintain embryonic development in these cold-adapted lizards, resulting in a greater proportion of stillborn offspring, as seen in wild populations (Le Galliard et al., 2010).

Higher temperatures and dry conditions resulted in additively higher tail width loss (Fig. 3A) and lower total body mass change during gestation (Fig. 3C). Tail reserves are a very important site for storing accumulated resources in many species of lizards (reviewed in Bateman and Fleming, 2009). However, our study is the first to demonstrate the negative impact of dehydration, enhanced by higher temperatures, on tail morphology. The tails of Z. vivipara are composed of both lipids and proteins (Avery, 1974), and increased muscle catabolism at this site could release bound water to offset physiologically damaging levels of dehydration (Brusch et al., 2018). Additionally, altered tail reserves likely explain lower body mass after parturition in females exposed to hot and dry treatments. Decreased maternal body condition after reproduction has been shown to challenge future survival, owing to increased starvation or predation risks if food is not readily available or if weaker mothers are less able to avoid predation (Bleu et al., 2013). These females also had higher plasma osmolality (Fig. 3B), suggesting that both high temperatures and water shortage challenge hydric state during pregnancy and lead to dehydration. Surprisingly, we did not detect an interactive effect. That is, gravid females exposed to hot and dry conditions did not have significantly higher plasma osmolality compared with females exposed to cold and wet, or hot and wet, conditions (Fig. 3B). Females may have increased muscle catabolism to release bound water and support some of the water demands of reproduction (Brusch et al., 2018), as females in dry conditions had higher tail width loss compared with females in wet conditions (Fig. 3A). Alternatively, females could have adjusted their behavior to maximize drinking when water was available in the morning, as has been shown in this species (Rozen-Rechels et al., 2020).

In addition to being more dehydrated, females in dry conditions throughout pregnancy also had higher markers of oxidative damage after parturition, regardless of temperature treatment (Fig. 4A). Unlike all the other dependent variables measured in our study, we found a significant interactive effect of temperature and water availability on non-enzymatic antioxidant levels. Dehydrated females in the hot treatments had the lowest non-enzymatic antioxidant levels (Fig. 4B), which, coupled with high levels of oxidative damage, equate overall to the highest levels of oxidative stress. These results suggest that the production of potentially damaging oxygen species is linked to hydric resources, and it appears that females were able to protect themselves with increased antioxidant production, as shown previously (Dupoué et al., 2020a; Stier et al., 2017). These results further complement recent comparative findings between wild populations, showing that warmer climates may be associated with lower OXY levels in pregnant Z. vivipara (Dupoué et al., 2020b), although, as demonstrated here, thermal impacts were conditioned by water resources. Importantly, the magnitude of alterations in oxidative status during pregnancy can have serious and delayed consequences on offspring mortality rate, which will need further examination (Dupoué et al., 2020a). Our study did not explicitly measure such reproductive trade-offs, however, and females in the hot treatments were exposed to water restrictions for ∼3 weeks, half the time that females from the cold treatments were exposed to the restrictions. It is worth noting that kinetic changes in oxidative stress are not stable during gravidity (Kouyoumdjian et al., 2019; Speakman and Garratt, 2014), so this might partly explain observed differences post-parturition. Additionally, antioxidants serve a myriad of functions and their production is frequently temporarily mismatched with the creation of reactive oxygen species over acute timescales (Costantini, 2019; Halliwell and Gutteridge, 2015). Females in the hot treatments may not have had enough time or hydric resources to balance their oxidative status, and a mechanistic understanding of these results is beyond the scope of this study but deserves further exploration.

When free-standing water is limited, some organisms are able to rely almost entirely on metabolic (as a by-product of metabolism) or dietary (that present in food) water (Karasov, 1983; Nagy and Gruchacz, 1994; Ostrowski et al., 2002). Because our experimental manipulations began after vitellogenesis, which is considered the main energy allocation period in this species (Foucart et al., 2014), and because metabolism is relatively low in squamate reptiles (Andrews and Pough, 1985), metabolic water was likely negligible (Shoemaker and Nagy, 1977). Additionally, temperature and metabolism are tightly linked in ectotherms (Dillon et al., 2010), but females in the hot treatments, regardless of water availability, had no significant differences in plasma osmolality changes (Fig. 3B), which further supports that metabolic water was negligible in our study. This still leaves the possibility that females in our experiment were able to partially compensate the hydric demands of reproduction through increased food consumption and dietary intake of water. Females in hot treatments ate a greater ratio of crickets (Fig. 5). If dietary water was an important component of hydration, we might expect lizards consuming a greater ratio of crickets to have decreased plasma osmolality, which was not the case in our study. Because we did not detect a significant difference in plasma osmolality change between females in hot and wet treatments and cold and dry treatments, it is difficult to conclude whether females were gaining any hydric benefit from food consumption. Overall, these results suggest that free-standing water is the most important source during pregnancy and essential for maintaining maternal hydration, especially considering that previous research has found no net osmotic advantage from eating in non-reproductive squamates (Lillywhite, 2017; Murphy and DeNardo, 2019; Wright et al., 2013). Higher temperatures still had an important effect on food consumption, and females with ad libitum access to water consumed marginally higher ratios of crickets, indicating a much more complex relationship between temperature, water availability and food consumption that merits future investigation.

Previous studies examining limited hydric resources during periods of reproductive investment have similarly found that mothers are burdened with the majority of water imbalance in favor of embryonic hydration (Brusch et al., 2018; Dupoué et al., 2015, 2018a). Although developing embryos often depend on maintaining water balance for survival (Cagle et al., 1993; Warner and Andrews, 2002), water imbalance is not necessarily physiologically detrimental in all cases. For example, immune enhancement as a result of dehydration has been documented in multiple taxa (Brusch and DeNardo, 2017; Hoang, 2001; Moeller et al., 2013), including mothers and embryos, during conflicts for limited hydric resources (Brusch and DeNardo, 2019; Brusch et al., 2017). Future studies should focus on the interplay between osmolality and physiological functions in both mothers and embryos, especially considering that previous explorations of more cryptic physiological traits have elucidated that both mothers and embryos are impacted by dehydration (Brusch et al., 2019).

Warmer and drier environmental conditions can challenge reproductive modes in several ways, and these lines of research deserve further attention. Future studies should continue to consider the interaction between temperature and water availability and further explore the impacts they have on maternal physiology, reproductive output and offspring physiology. In particular, developmental temperatures and hydric conditions outside optimal ranges, which could become more frequent under expected climate change scenarios, may lead to increased embryonic mortality (Gibbons et al., 1983; Madsen et al., 2006; Packard et al., 1977; Semlitsch, 1987; Sperry and Weatherhead, 2008). Therefore, considering both hydroregulation and thermoregulation under joint conditions of thermal and hydric stress will be essential for understanding ectothermic responses to predicted shifts in temperature and water availability (Rozen-Rechels et al., 2019; Smith et al., 2019; Wake and Vredenburg, 2008).

We are grateful to M. Leroux-Coyau, M. Dezetter, L. Lange and B. Michaud for assisting with experimental procedures. We also thank P. Blaimont for helping with lizard capture.