Food and light availability induce plastic responses in fire salamander larvae from contrasting environments

Sampling and experimental design

We collected 260 S. salamandra larvae corresponding to the developmental stage 1 (Juszczyk & Zakrzewski, 1981) from three populations in Romania: one inhabiting a surface habitat and two using subterranean habitats for breeding (Fig. 1). The surface habitat (Iconie, 44.99293°N, 22.78105°E, 325 m a.s.l.) consisted of a deciduous forest crossed by a stream and was sampled on two different occasions: once in April 2016 (Iconie₂₀₁₆, n = 128 larvae), and once in April 2017 (Iconie₂₀₁₇, n = 56 larvae). The two populations using subterranean habitats were sampled only once each, in April 2017, from a cave (Gaura cu Muscă, 44.66472°N, 21.69916°E, 100 m a.s.l., n = 47 larvae) and a tunnel (Buzău, 45.46981°N, 26.28086°E, 475 m a.s.l., n = 29 larvae). The non-permanent nature of the water bodies in the sampled subterranean habitats prevents the larvae from overwintering. The sample size was determined by the availability of stage 1 larvae in natural habitats.

We conducted an experiment with a 2  × 2 factorial design that resulted in four treatments: (1) high food availability and 8-hour light (high-light), (2) low food availability and 8-hour light (low-light), (3) high food availability and 0-hour light (high-dark) and (4) low food availability and 0-hour light (low-dark). The experiment was conducted in the cave-laboratory of “Emil Racoviţă” Institute of Speleology (Cloşani Cave, 45.073064°N, 22.800675°E, 433 m a.s.l.). A thorough description of this cave is provided by Povară, Drăguşin & Mirea (2019). The experimental setup was identical in both years and was distributed among two separate concrete cubicles within the cave: the first cubicle was equipped with a light source which kept the light intensity constant at 50 lx, 8 h a day, while larvae in the second cubicle were kept under constant darkness conditions. The 8-hour photoperiod and 50-lux light intensity were selected to mimic the conditions of natural shaded habitats at the surface, where salamander larvae are usually found. Humidity and temperature in both cubicles were constant throughout the experiment (close to 100%; 13 ± 1 °C).

The initial wet body mass and body size of larvae were measured (initial size-TL₀ and BM₀, see below for details) and then the larvae from each population and habitat type were randomly allocated to each of the experimental treatments (Table S1), but with the additional constraint that each treatment had an equal number of larvae (i.e., 65 larvae per treatment). The larvae were housed individually in plastic containers (length × width × height: 21  × 14  ×  9.5 cm) with a unique ID, with two cm water depth, following Warburg (2009). All larvae were fed live Tubifex sp. every third day and the water was changed after each feeding event. The larvae in high food availability treatments were fed six live prey items per feeding throughout the first 3 weeks, and ten prey items per feeding from the fourth week. The larvae in low food treatments received half the amount of food compared to larvae in the high food treatments.

The effects of food availability and light conditions on larva size were determined based on measurements of individual total length, i.e., the length from the tip of the snout to the tip of the tail (TL) (0.01 mm accuracy), and wet body mass (BM) (0.01 g accuracy). Total length measurements were made from photographs taken with a digital camera (Nikon D3300), using the software ImageJ v. 1.50i (Schneider, Rasband & Eliceiri, 2012). Measurements were calibrated using a standardized scale present in each digital image. Wet BM was measured with an electronic Pesola scale, after the larvae were blotted on a wet paper towel to remove excess water (Beachy, 1995). Both TL and BM were measured once at the beginning of the experiment (TL₀ and BM₀), and once at metamorphosis (TL_met and BM_met).

The experiment commenced in April and was terminated in October (after 170 days), both in 2016 and 2017. For each individual, we recorded the time needed to reach metamorphosis (Time_met) as a measure of developmental rate, defined as the time (days) necessary for the complete resorption of gills and tail fin. Over the 170-day experimental period, 75% of larvae (195 of 260) reached metamorphosis, the fastest after 67 days and the slowest after 169 days. The number of larvae that successfully reached metamorphosis in each treatment was the following: high-dark: subterranean n = 17; surface n = 32; low-dark: subterranean n = 16; surface n = 30; high-light: subterranean n = 14; surface n = 35; low-light: subterranean n = 15; surface n = 36. The 170-day timeframe was mainly dictated by logistical constraints related to maintaining the larvae; we considered that the threshold of 75% metamorphosed individuals was sufficient for the data analyses. The 65 larvae which did not metamorphose within the 170-days time frame were excluded from the analyses. All larvae and metamorphosed individuals were released in their original habitats at the end of the experiment. There was no mortality in our study.

The collection and rearing of the larvae complied with the Directive 2010/63/UE of the European Parliament and of the Council of 22 September 2010 on the protection of animals used in scientific purposes. Our research was carried out under permit no. 250/20.04.2016 obtained from the Administration of the Domogled - Valea Cernei National Park and permit no. 78/10.02.2016 from the Speleological Heritage Commission.

Data analyses

All analyses were performed with an a priori level of significance of 0.05, in R version 3.6.0 (R Core Team, 2019). The dataset comprising the 195 larvae used for the analyses is available as a supplementary file (Table S2). The two samples collected in 2016 and 2017 from the surface site were pooled for the analyses. Growth rates in TL and BM (Growth_TL and Growth_BM) were calculated following Alcobendas, Buckley & Tejedo (2004) as the difference between size at metamorphosis (TL_met and BM_met) and the initial size (TL₀ and BM₀), divided by the time to metamorphosis (Time_met). We tested the data for normality using the Shapiro–Wilk test and chose the subsequent statistical tests accordingly.

To check for differences in the initial size of larvae between habitats or among populations we used Mann–Whitney and Kruskal–Wallis tests. To test the effects of food availability (low versus high), photoperiod (darkness versus 8-hour light) and habitat of origin (subterranean versus surface), and the two-way interactions of these three factors on each response variable, i.e., Growth_TL, Growth_BM, Time_met, TL_met and BM_met we used linear mixed models (LMMs) or generalised linear mixed models (GLMMs). We included sampling identity (i.e., Iconie₂₀₁₆, Iconie₂₀₁₇, Buzău and Gaura cu Muscă) as a random effect in the models. We considered the identity of samples collected in 2016 (Iconie₂₀₁₆) and 2017 (Iconie₂₀₁₇) from the surface habitat separately in order to account for the potential yearly variation caused by differences in biotic (maternal or litter effects) or abiotic factors. We fitted the models with Gaussian, Poisson or Gamma distribution, and an identity or log link function, and evaluated the fit of competing models using the anova function of the “stats” package. Finally, GLMMs with Gamma distribution and identity link were employed for Time_met, BM_met and TL_met, while LMMs were used for Growth_TL and Growth_BM.

Adult body size may be strongly influenced by size at birth, i.e., initial size, so that larger offspring will result in larger metamorphs and adults (Alcobendas, Buckley & Tejedo, 2004). To account for this potential effect we included TL₀ and BM₀ as covariates in the models. Since TL₀ and BM₀ were highly correlated (Spearman’s rho = 0.784, p  <  0.001), we included them separately in models with Time_met as a response variable. Subsequently, we compared the resulting models to select the best fitting model. In addition, models with TL_met and Growth_TL as response variables were fitted with TL₀, while the ones with BM_met and Growth_BM as response variables were fitted with BM₀ as covariates.

We used the R package “lme4” version 1.1-21 (Bates et al., 2015) for fitting the models. We selected the best fitted models based on analysis of variance tests. We used the function ANOVA from the “car” package (Fox & Weisberg, 2019) to assess the significance of model predictors based on Wald and likelihood-ratio chi-square tests (type III analysis of deviance). We performed least square means post-hoc pairwise comparisons with Bonferroni correction for multiple tests, with the packages “multcomp” (Hothorn, Bretz & Westfall, 2008) and “emmeans” (Russell, 2021). All graphics were done with the R package “ggplot2” (Wickham, 2016).

Initial size and its effects on traits at metamorphosis

The initial size (BM₀ and TL₀) of the larvae varied significantly across samples (Kruskal–Wallis test, BM₀: H = 31.842, df = 3, p  <  0.001; TL₀: H = 36.114, df = 3, p  <  0.001) and between habitat types (Table S3, Mann–Whitney test, BM₀: W = 5232, p = 0.002; TL₀: W = 5723.5, p < 0.001), with larvae from the subterranean habitats being larger compared to those from the surface habitat (mean ± SE, subterranean: BM₀ = 0.315 ± 0.015 g; TL₀ = 38.73 ± 0.66 mm; surface: BM₀ = 0.257 ± 0.04 g; TL₀ = 35.79 ± 0.27 mm). However, the initial size of the larvae did not differ significantly across treatments (Kruskal–Wallis test, BM₀: H = 2.2 09, df = 3, p = 0.530; TL₀: H = 2.466, df = 3, p = 0.481, Fig. S1).

Growth rates were influenced significantly by the initial size (Table 1), larger and heavier larvae exhibiting lower growth rates in terms of both total length and body mass, respectively (Table 1, Fig. 2). Time to metamorphosis was best explained by a model that included BM₀ rather than TL₀. BM₀ had a significant negative effect on time to metamorphosis, thus larvae with a lower BM₀ attained metamorphosis later (Table 1, Fig. 3). Total length at metamorphosis was significantly affected by TL₀, larger larvae attaining larger total lengths at metamorphosis; there was no significant effect of the initial body mass on body mass at metamorphosis (Table 1).

Hypothesis 1: low food availability and absence of light decrease larval growth and delay metamorphosis

We found a significant main effect of food availability on only one of the five metamorphosis traits studied, specifically time to metamorphosis (Table 1). Larvae raised in low food availability treatments reached metamorphosis significantly later compared to those raised with high food availability (mean ± SE, low = 118.2 ± 3.03 days, high = 96.6 ± 1.90 days, p = 0.003; Fig. 4). Regarding photoperiod, we found a significant main effect on body mass growth rate (Table 1). Thus, larvae raised in darkness exhibited a significantly slower growth rate compared to those raised in the 8-hour photoperiod treatments (mean ± SE, dark = 0.0044 ± 0.0001 g/day, light = 0.0046 ± 0.0001 g/day, p = 0.040) (Table 1 and Fig. 5).

Hypothesis 2: light conditions interact with food availability, mediating its effects on growth and time to metamorphosis

The interaction between food availability and photoperiod did not significantly affect either growth rates or time to and size at metamorphosis (Table 1).

Hypothesis 3: larval response to contrasting light and food conditions is shaped by the habitat of origin

Habitat type and food availability had a significant interactive effect on growth rates (Table 1). Total length growth rate was significantly higher in larvae from the surface habitat raised in high compared to those in low food availability treatment (mean ± SE, surface-low = 0.116 ± 0.004 mm/day, surface-high = 0.155 ± 0.005 mm/day, p  <  0.001, Table 2, Fig. 6A). Instead, larvae from subterranean habitats showed similar total length growth rates between low and high food availability (Table 2, Fig. 6A). Body mass growth rate was significantly higher in larvae from the surface habitat raised in high food availability treatments (mean ± SE, surface-high = 0.0055 ± 0.0001 g/day) compared to those from subterranean and surface habitats raised in low food availability treatments (mean ± SE, subterranean-low = 0.0035 ± 0.0002 g/day, p = 0.039; surface-low = 0.0042 ± 0.0001 g/day, p  <  0.001, Table 2, Fig. 6B). Habitat type and food availability also had a significant interactive effect on time to metamorphosis (Table 1). Larvae from both surface and subterranean habitats raised in high food availability underwent metamorphosis significantly faster than larvae in low food availability treatments (Table 2, Fig. 7).

The interaction between habitat type and photoperiod did not significantly affect either growth rates or time to metamorphosis (Table 1). However, habitat type had a significant main effect on body mass growth rate (Table 1). Thus, larvae from subterranean habitats increased in body mass significantly slower compared to those from the surface habitat (mean ± SE, subterranean = 0.0038 ± 0.0001 g/day, surface = 0.0048 ± 0.0001 g/day, p = 0.029) (Table 2 and Fig. 8).

Although body mass and total length at metamorphosis were on average higher in larvae from the surface compared to those from the subterranean habitat (mean ± SE, surface: BM_met = 0.79 ± 0.01 g; TL_met = 50.63 ± 0.34 mm; subterranean: BM_met = 0.71 ± 0.01 g; TL_met = 48.23 ± 0.52 mm), none of the tested main factors (i.e., food availability, photoperiod or habitat type), nor their two-way interaction showed any significant effect on these traits (Table 1, Fig. S2).

Food availability and photoperiod act separately on metamorphosis traits

Previous studies have indicated that food availability affects growth and development in several amphibian species. Most of the studies showed that foraging shifts to higher food quantity or quality results in increased growth, which translates to faster developmental rates (Leips & Travis, 1994; Beachy, 1995; O’Laughlin & Harris, 2000; Álvarez & Nicieza, 2002; Hickerson, Barker & Beachy, 2005; Jones et al., 2015) and/or an increase in size at metamorphosis (Pandian & Marian, 1985). A similar study (Manenti et al., 2023) on European fire salamander larvae from different altitudes showed that rich food induced higher growth rates and earlier metamorphosis at smaller sizes, and highlighted that both environmental conditions and local adaptations determine the plastic response of larvae. However, it has been shown that acceleration or deceleration of development rate, depends on the developmental stage at which the foraging shift occurs (e.g., Crump, 1981; Alford & Harris, 1988; Newman, 1994; Tejedo & Reques, 1994; Álvarez & Nicieza, 2002). According to the Wilbur-Collins model, in the case of deceleration of development rate, the predictions are that larvae delay the metamorphosis either to attain a critical size at metamorphosis or to capitalize on the rapid growth opportunity (Wilbur & Collins, 1973). In our study, low food availability increased time to metamorphosis in both surface and subterranean salamanders, but only surface larvae displayed enhanced growth rates in high food conditions, suggesting that larvae adapted to high food conditions may capitalize better (Manenti et al., 2023). These differences in growth rates among larvae of different habitat origin suggest an interplay between plasticity and adaptation, favoring the exploitation of contrasting environments in this species. However, it is important to consider the potential carryover effects from conditions experienced at early stages, such as size at metamorphosis and survival at metamorphosis, as these can have significant impacts on the survival and fitness of individuals in their adult life stages (Melotto, Manenti & Ficetola, 2020). Our study revealed that the habitat of origin and trophic conditions did not induce differences in size at metamorphosis, suggesting more complex outcomes than those generally supported by the Wilbur-Collins model for amphibian metamorphosis. We showed that fire salamanders can compensate for ecological differences between contrasting habitats during early life stages and, similar to previous research (Manenti, Denoël & Ficetola, 2013), our results highlighted their ability to colonise and adapt to new habitats.

The role of the photoperiod in shaping growth and development patterns of amphibian larvae in natural populations is largely unknown. Several physiological studies conducted in the laboratory provided evidence that amphibian larval growth and developmental rates are altered by photoperiod (e.g., Guyetant, 1964; Wright et al., 1988; Crawshaw et al., 1992). These studies, where photoperiod was altered to an extreme degree (i.e., continuous light or darkness), found that amphibians generally increase growth and developmental rates in response to light. However, the response might also be determined by the duration of the photoperiod (Laurila, Pakkasmaa & Merilä, 2001). In our study, photoperiod affected only body mass growth rate, which was significantly higher in the 8-hour light treatments.

Interplay between photoperiod and food availability

A number of studies have documented the impact of foraging shifts interacting with other environmental factors and the induced changes in amphibian growth and development. These include interactions between foraging shifts and temperature, predation risk and pond desiccation, respectively (Nicieza, 2000; Álvarez & Nicieza, 2002; Enriquez-Urzelai et al., 2013). The results suggested that interactive effects are difficult to predict (Álvarez & Nicieza, 2002). As food availability and photoperiod both play a role in growth and development, it is unlikely that they operate independently. Therefore, in our study, we aimed to examine how they interact to affect growth and development rates. Specifically, we hypothesized that photoperiod may mediate the influence of food availability on these rates. Our results did not support this assumption, as the traits we investigated were responsive to food availability and light condition, but no interaction between these two factors was detected. However, a previous study (Băncilă et al., 2021) showed that these two factors have an interactive effect on the behavioural responses of fire salamander larvae; more specifically, larvae exposed to low food availability take more risks in exploring the environment, especially in the absence of light, when predator pressure would be lower. Manenti, Denoël & Ficetola (2013) found that behavioural plasticity is higher in larvae from subterranean environments. In addition, environmental factors, such as darkness, predator absence, and resource depletion, can interact with other factors, such as prey mobility and conspecific presence, to influence the behavior of cave-dwelling animals (Uiblein et al., 1992; Melotto, Ficetola & Manenti, 2019). Together, these findings suggest that the adaptive responses to environmental cues are mediated both physiologically and behaviorally during the early development of fire salamanders. As such, in natural conditions, a higher risk-taking behaviour may pay-off in terms of access to resources. We suggest that this, together with lower predation risks, would be an advantage in subterranean environments, even if larval growth rate is slower in the absence of light and/or food resources are limited.

Plasticity may not be selected for in relatively stable environments (Merilä et al., 2000; Kulkarni et al., 2011), and may not be present in all species because maintaining plasticity is costly (DeWitt, Sih & Wilson, 1998; Jannot, 2009; Reed, Schindler & Waples, 2010). In amphibians, plastic developmental rates have been associated with trade-offs in size at metamorphosis (Denver, Mirhadi & Phillips, 1998; Merilä et al., 2000). Metamorphosis at a smaller size was associated with reduced adult size and fecundity (Márquez-García et al., 2009), lowered immune system function (Terentyev, 1960) or increased mortality (Rudolf & Rödel, 2007). Larger size at metamorphosis requires additional time for growth, but it is associated with enhanced survival and performance in metamorphs (Cabrera-Guzmán et al., 2013). In our study, size at metamorphosis (i.e., total length and body mass) was unaffected by food availability or photoperiod. However, larger larvae (i.e., initial total length) resulted in larger metamorphs (i.e., total length at metamorphosis), while also reaching metamorphosis faster, at slower growth rates, in line with previous studies (e.g., Alcobendas, Buckley & Tejedo, 2004). Interestingly, while the initial size at the beginning of the experiment was significantly higher in larvae from subterranean habitats, size at metamorphosis was overall higher in larvae from the surface (although not significantly). This suggests that developmental plasticity can promote development with limited effects on size at metamorphosis.

Plasticity can serve as a potential mechanism that promotes the colonization process, especially in the early stages of facing novel ecological pressures. At this stage, plasticity allows for flexible responses that can increase the chances of survival and establishment. However, as selective pressures in the new environment persist, adaptation becomes crucial for successful colonization. The fixation of adaptive traits or canalization of plasticity may occur to ensure better fitness in the new environment (Romero, 2009; Levis & Pfennig, 2019). These mechanisms can coexist in situations where gene flow allows for exchange between populations adapted to the ancestral habitat and populations colonizing the new environment (Storfer, 1999; Richter-Boix et al., 2010). This idea is consistent with previous studies that have shown how phenotypic plasticity can help species to overcome altered environmental conditions and buy time for population persistence that may be followed by adaptive trait refinement and evolution (Diamond & Martin, 2021). However, there are still unanswered questions, such as which are the conditions under which plasticity can allow or limit subsequent evolutionary changes. Subterranean habitats might act as environmental filters and species that occur both in surface and subterranean habitats can be used for comparative analysis with respect to variation of plasticity. Surface habitats, in particular streams, are thought to be the ancestral habitats for fire salamanders (Seifert, 1991; Steinfartz, Weitere & Tautz, 2007). In the ancestral habitats, more plastic genotypes are expected compared to populations colonizing novel environments (Diamond & Martin, 2021). Our results support this prediction, surface larvae showing more plasticity in their growth depending on food availability.

It is important to interpret the results of our study with caution, as the limited number of populations sampled may not fully represent the overall variation in growth rates and time to metamorphosis between surface and subterranean habitats. Specifically, the study only sampled a single surface population, which may limit the generalization of the findings to other surface populations. In this regard, we cannot dismiss the possibility that the genetic component can have a large impact on life-history traits, along with raising conditions (e.g., Caspers, Steinfartz & Krause, 2015).

Overall, our study showed that S. salamandra is an excellent target species for testing hypotheses regarding populations persistence via plasticity and their potential to evolve when confronted with changing environments.