Ecosystem effects of intraspecific variation in a colour polymorphic amphibian

Abstract

An emerging consensus suggests that evolved intraspecific variation can be ecologically important. However, evidence that evolved trait variation within vertebrates can influence fundamental ecosystem-level processes remains sparse. In this study, we sought to assess the potential for evolved variation in the spotted salamander (Ambystoma maculatum) to affect aquatic ecosystem properties. Spotted salamanders exhibit a conspicuous polymorphism in the colour of jelly encasing their eggs—some females produce clear jelly, while others produce white jelly. Although the functional significance of jelly colour variation remains largely speculative, evidence for differences in fecundity and the morphology of larvae suggests that the colour morphs might differ in the strength or identity of ecological effects. Here, we assessed the potential for frequency variation in spotted salamander colour morphs to influence fundamental physiochemical and ecosystem properties—dissolved organic carbon, conductivity, acidity and primary production—with a mesocosm experiment. By manipulating colour morph frequency across a range of larval densities, we were able to demonstrate that larva density and colour morph variation were ecologically relevant: population density reduced dissolved organic carbon and increased primary production while mesocosms stocked with white morph larvae tended to have higher dissolved organic carbon and conductivity. Thus, while an adaptive significance of jelly coloration remains hypothetical, our results show that colour morphs differentially influence key ecosystem properties—dissolved organic carbon and conductivity.

1. Introduction

The realization that evolutionary processes can operate in contemporary time and across small distances has had a broad and profound influence on ecological understanding [1]. Early evidence that evolved phenotypic variation among and within populations could have a wide range of measurable ecological effects is now largely confirmed [2–4]. These findings are complemented by recent experiments that help dissect the density- and trait-mediated components of eco-evolutionary feedbacks [5–10]. However, evidence for strong linkages between microevolution and ecosystem dynamics has been slow to emerge [11,12]. Indeed, few studies demonstrate how contemporary evolution can affect ecosystem dynamics [13]. This problem is not new or unique to evolutionary ecology. Despite some important successes, evidence demonstrating a role for vertebrates as drivers of ecosystem processes is also sparse [14]. But by demonstrating connections between vertebrates and ecosystem ecology, initial efforts have made a strong case that fundamental processes such as nitrogen and carbon cycling are influenced by the ecology of vertebrates (e.g. [15]). Characterization of these linkages remains an important challenge for ecologists [16,17] and an emerging consideration of evolutionary biologists, who must now assess whether intraspecific variation is relevant for ecosystem ecology [18–20].

Here, we designed a project to probe the ecological significance of evolved intraspecific variation in spotted salamanders (Ambystoma maculatum; figure 1a). Spotted salamander females exhibit a conspicuous polymorphism; producing egg masses that differ in colour (figure 1b–d). Clear morph females are characterized by the production of transparent egg jelly—typical of many amphibians, including every other Ambystoma species. However, white morph females produce jelly that is practically opaque to visible and ultraviolet light [21]. Although egg masses with intermediate coloration are occasionally reported, evidence for an intermediate morph is limited [22,23]. Indeed, this polymorphism is most commonly treated as a discrete dimorphism comprising clear and white morphs [21,24–27], an approach we follow here. Polymorphic populations are common and can be found across the large geographical range of the spotted salamander [21–23,25]. Within this extensive geographical range, morph frequencies exhibit striking among-population variability—even reaching down to microgeographic spatial scales [21,23,27,28]. As with many polymorphisms, there has been little work to explore whether this extensive variation has any ecological relevance.

Figure 1.

Spotted salamanders (adult female in (a)) lay egg masses that vary in coloration. Clear egg masses are characterized by a transparent outer jelly matrix. White egg masses are characterized by an opaque jelly matrix that blocks the transmission of light. Spotted salamander populations vary in the frequency of each morph—from monomorphic clear to monomorphic white (b–d)—often at microgeographic scales. Photo credits: (a,c,d) Sean Giery; (b) Dana Drake.

In this experiment, we investigated whether morph frequency variation could affect the ecology of spotted salamander breeding ponds. Several factors suggest colour morphs could differ in their ecological effects. First, colour morphs appear to differ in the number of larvae produced per egg mass. This difference could arise through fecundity differences between morphs [28,29]. Or, morphs could vary in morph-specific hatching rates owing to differential susceptibility to predation or the light environment [23,26]. Thus, any density-dependent ecological effect could be influenced by morph frequency owing to differences in fecundity and hatching success. Second, the larval offspring of different morphs might exhibit trait variation that influences their ecological interactions. For example, a difference in hatchling size between colour morphs is often reported [26–28]. Because Ambystoma larvae are gape-limited predators, even minor size differences could influence their per capita effects on prey [30–32] and interactions with conspecifics [33,34]. Given well-demonstrated size-dependent top-down effects of Ambystoma larvae on temporary pond communities [35–40], we hypothesized that spotted salamanders could generate top-down effects that propagate through the food web and generate morph-dependent indirect effects on basal resource pools.

To evaluate our hypothesis, we raised embryos from white and clear morphs to hatching in a common garden, then stocked the newly hatched larvae from each morph into mesocosms across a range of densities. Because basal resources in temporary pond ecosystems consist of algal and detrital pools [41,42], we focused on assessing change in benthic algae and dissolved organic carbon—as well as two abiotic parameters, pH and conductivity. We tracked these ecosystem-level variables over a three-month period with a general expectation that a morph effect would emerge. But while direct top-down effects of spotted salamanders are established well enough to be expected in an experiment such as this, the evidence for indirect trophic effects remains limited [43]. In combination, poorly resolved indirect effects and seemingly idiosyncratic phenotypic variation among morph offspring strain the formulation of well-justified and specific predictions for how a morph effect would manifest. Density effects, in contrast, are presumably more predictable [44]. For example, algae-based food chains in temporary ponds are short, characterized by Ambystoma larvae that prey on consumers of phytoplankton and benthic algae [35,37,42,45]. In these three-level food chains, indirect trophic effects should propagate predictably, with algal biomass increasing with the magnitude of herbivore reduction [46]. However, the effect of Ambystoma larvae on longer, detritus-based food chains is more difficult to predict, or even assess. This difficulty arises from the notorious complexity of detrital food webs attributable to extensive omnivory, variable decay rates and diverse communities of decomposers and consumers—from invertebrate shredders to planktonic bacteria that metabolize dissolved organic compounds [47–49]. Thus, while recent data show that Ambystoma larvae can generate indirect trophic effects within detrital food chains, the mediating interactions remain undefined [40].

2. Methods

(a) . Study system and natural history

Adult spotted salamanders (figure 1a) migrate to ponds and wetlands in late winter and early spring to mate. After mating, females deposit one or more globular egg masses on the pond bottom or attached to submerged objects [50]. Each egg mass can include up to several hundred eggs encased in a jelly matrix [50] (figure 1b–d). Larvae hatch a few weeks later and spend several months as aquatic larvae before metamorphosing in mid-to-late summer. Spotted salamander larvae are predators of small animals and can have strong direct effects on aquatic communities—reducing total prey biomass and restructuring community composition [36,39]. Yet, their indirect effects on primary production, decomposition and nutrient cycling are essentially unknown [51,52]. But given the strong direct effects of Ambystoma larvae on prey, spotted salamander larvae have high potential to influence ecosystem dynamics via trophic cascades reaching down to detrital and algal resource pools.

Variation in morph frequency among populations is well documented. Multiple adaptive hypotheses have been proffered to explain this variation. To date, all of these hypotheses include some component of morph-dependent susceptibility of spotted salamander embryos to ultraviolet radiation [53,54], disease [55], oxygen limitation [26] or predation [23,56]. Several studies also report relationships between morph frequency and water chemistry; however, an adaptive mechanism has not been identified in these cases [22,27]. Overall, despite multiple attempts, unequivocal evidence for a distinct adaptive value of jelly coloration remains elusive [21]. By contrast, although the evidence is not definitive, multiple lines of evidence suggest that jelly coloration is under genetic control: ovisac anatomy varies by polymorphism [25], individual females lay the same egg mass morph through time [22], morph-specific protein profiles are conserved across 1600 km [25], and temporal patterns of morph frequency fluctuation conform neatly to expectations derived from evolutionary dynamics [21]. Further, the demonstration of complementary patterns of morph-specific mortality and microgeographic variation is consistent with evolution by natural selection [23].

Over the last several decades, multiple studies have investigated the fitness and phenotypic effects of jelly coloration. Egg number, embryo survival, nitrogen content of jelly, offspring size and others are typical targets of investigation. Studies often reveal differences between the colour morphs in one or more of these variables. However, there is often substantial variation reported among studies. As a whole, it appears that many of these morph differences are inconsistent, challenging succinct characterization of between-morph variation. Our suspicion is that many of these colour morph effects or differences (e.g. embryo survival, or body size) are highly sensitive to the developmental environment. Without better accounting for light, temperature and nutrient environments, results appear idiosyncratic. This is exemplified by results of Pintar & Resetarits [27] showing that variation in nutrient environment—and likely light environment—can invert the direction of morph effects on the body size of larvae. To provide additional context for our study and findings, we provide a summary of several parameters relevant for our study and others (electronic supplementary material, S1). We also provide quantitative information on the light, carbon and temperature environment of our common-garden experiments.

(b) . Common-garden experiment

Egg masses (n_white = 14, n_clear = 18) were collected on 4 April 2022 from a population near Linville, North Carolina (36.0588 N, −81.8812 W; 1093 m elevation). Egg masses were collected within 48 h of being laid and immediately moved 25 km northeast to an outdoor experimental aquatic facility in Boone, North Carolina (36.2145 N, −81.6925 W, 1070 m elevation), where we reared embryos to hatching in a common garden and subsequently conducted our mesocosm experiment. The aquatic facility where our common-garden and mesocosm experiments were conducted is exposed to full sun throughout the day.

The primary purpose of the common-garden rearing experiment was to ensure that developing embryos were exposed to similar light and temperature regimes—and help minimize environmental contributions to phenotypic variation. Monthly mean air temperature for Boone, North Carolina in April 2022 was 9.6°C (National Weather Service) with a photoperiod of approximately 12–13 h. Egg masses were divided among six 76 l open-top tubs filled to a depth of 20 cm with tap water treated with conditioner (AmQuel by Kordon) to eliminate chloramine and chlorine. Tubs included four (n = 2 tanks) or six (n = 4 tanks) egg masses each with white and clear egg masses in equal proportion. Leaf litter was not provided, and the water used was naturally low in dissolved organic carbon (less than 0.5 mg l⁻¹). Owing to low dissolved organic carbon, light transmission—including ultraviolet—is relatively high at these water depths (electronic supplementary material, S2). These water depths, dissolved organic carbon concentrations (DOC) and levels of ultraviolet exposure (l_UVB) are representative of natural conditions for spotted salamander egg masses in the study region (maximum depth = 25.3 cm; DOC = 3.1 mg l⁻¹, I_UVB at 10 cm = 39%; electronic supplementary material, S3). Egg masses were checked weekly to identify dead embryos and for embryonic staging (following [57]). Once hatching began, egg masses were moved to the laboratory to count and prepare larvae for release into mesocosms. The egg jelly was discarded after larvae hatched.

(c) . Experimental treatment

Mesocosms were 1100 l stock tanks filled with tap water. Water conditioner (AmQuel by Kordon) was added to eliminate chloramine and chlorine. All mesocosms were seeded with 2 kg of air-dried deciduous leaf litter collected from the adjacent forest floor (primarily, Quercus rubra and Acer rubrum). A total of 32 mesocosms were arranged into four spatial blocks of eight. One week after filling, mesocosms were inoculated with 2 l of pond water from a nearby (approx. 5 km away) vernal pond. This source pond hosts diverse invertebrate and amphibian assemblages, including spotted salamanders. By the time salamander larvae were stocked, phytoplankton and benthic algae were established and abundant, as were zooplankton (cladocerans and copepods) and benthic macroinvertebrates (e.g. chironomid larvae). Trait (larvae from white or clear egg masses) and density treatments (4, 8, 16, 32 larvae per mesocosm) were randomly assigned within each block. Stocking densities (mesocosm: 1.5–12.2 individuals m⁻²) correspond to low–moderate larva densities in regional ponds (1–49 individuals m⁻²; electronic supplementary material, S4). Recently hatched larvae were pooled, by morph, to minimize possible ‘family’ effects. Larvae were then drawn from these mixed pools and stocked into mesocosms on 4 May 2022. After stocking, all mesocosms were covered with 60% shade cloth to prevent colonization of predators and escape of salamanders. During the mesocosm experiment, monthly air temperatures for Boone ranged from 15.1°C in May to 21.3°C in July 2022.

(d) . Data collection

Immediately before larvae were stocked, three ceramic tiles (48 × 48 mm) were submerged in each replicate mesocosm. To avoid shading by leaf litter, tiles were elevated 7 cm above the bottom by placing them on a brick positioned in the mesocosm centre. Tile removal coincided with monthly water sampling. Using a razor blade, attached algae was scraped from each tile onto pre-weighed filters and dried at 40°C for 48 h before reweighing to obtain algal dry mass.

We collected water samples mid-way through the mesocosm water column (approx. 20 cm below the surface). Samples were kept dark and refrigerated for 1–24 h until filtered through Whatman GF/F filters and read on a Turner Aquaflor fluorometer. Following previously used methods [21,58,59], we converted relative fluorescence units (RFUs) to dissolved organic carbon in mg l⁻¹ with field-collected samples measured using Environmental Protection Agency method 415.1 [60]. Additional methodological details, including conversion formulae, are provided in electronic supplementary material, S5. Conductivity and pH were measured on filtered samples with an Oakton PCTS 35 probe.

The first metamorphosed salamanders were observed on 22 June. Afterwards, mesocosms were checked daily. Any metamorphosed salamanders were collected and measured for wet mass and body length. Wet body mass was measured after gently blotting any excess water from individual salamanders. Body length was measured with digital calipers from the tip of the snout to the posterior end of the cloaca. We calculated the residual mass index (R_i) for each metamorph [61]. An estimate of mass-for-length, R_i is a morphology-based index used to quantify variation in body condition—often used as a proxy of ecological performance or fitness (electronic supplementary material, S6).

(e) . Data analysis

Count data such as number of eggs and larvae per mass were analysed with Poisson regression. We used a generalized linear mixed-effects model to test for morph differences in survival probability. We specified the tub as the random effect in our common-garden setup. We examined the effects of density and morph on the body size and condition of newly metamorphosed salamanders using linear mixed models and performed follow-up analyses with the emmeans package in R [62]. We included ‘block’ as a random effect in these mixed-effects models and examined a set of candidate mixed models differing in complexity. For each focal ecosystem response variable (e.g. dissolved organic carbon, primary production), we constructed a series of linear mixed-effects models to examine colour morph and density treatment effects. We accounted for any among-replicate variation at the start of the experiment by including initial measures of dissolved organic carbon, pH and conductivity as covariates in each model. Within our mixed-model framework, we used a repeated-measures approach in which temporal samples within each replicate mesocosm were included. The temporal variable was expressed as weeks since the start of the experiment and was included as a fixed effect. For each response variable (pH, dissolved organic carbon, conductivity and algal biomass), we constructed a set of candidate models differing in complexity. Candidate models were evaluated by finding the models with lowest Akaike information criterion (AIC). In cases where ΔAIC was less than 2, we present results from the simpler model (fewer numerator degrees of freedom). Population density was log-transformed prior to analysis. Unless otherwise noted, models were fitted in R using the lme4 package [63].

3. Results

(a) . Common-garden experiment

We found that clear morph egg masses had 20% more eggs than the white morph egg masses (clear = 70.1, white = 58.6; z_1,29 = −4.07, p < 0.0001; figure 2). Embryos from each colour morph exhibited similar developmental rates; after 18 days they averaged Harrison stage 37 on 21 April (estimate = 0.23, F_1,25.1 = 0.66, p = 0.4241). After moving egg masses to the laboratory, larvae emerged from egg masses over a period of 12 days (21 April 2022–3 May 2022). Although embryo survival to hatching was high, survival was significantly lower for clear egg masses (94% versus 99%: z_1,29 = 7.42, p < 0.0001; figure 2). Despite the higher survival of white morph embryos, clear morph egg masses still produced more larvae (clear = 64.7, white = 56.2, z = −3.08, p < 0.0021; figure 2).

Figure 2.

Colour morphs differed in the initial number of eggs per mass (a). Over the duration of the common-garden experiment, survival probability of embryos from clear morphs was significantly lower than of those from the white morphs (b). Variation in embryo survival reduced the magnitude of initial fecundity differences. However, the number of larvae hatched per egg mass remained higher for the clear morph (c). Data are reported as morph-specific means ± s.e.

(b) . Mesocosm experiment

Survival to metamorphosis was high (85%) with no significant effects of colour morph or stocking density on the proportion of larvae surviving to metamorphosis (electronic supplementary material, S7). However, body size and condition at metamorphosis were highly variable across treatments. For example, body mass ranged from 0.38 to 2.65 g and body length ranged from 24 to 43 mm. These differences correspond to a 150% difference in mass and 57% difference in length. Body condition (R_i) was also highly variable, ranging from −0.28 to 0.40—a 200% difference in body condition.

Density and colour morph treatments had significant, independent effects on body mass and length. Density had a negative effect on mass and length at metamorphosis (table 1 and figure 3). White morph offspring were significantly heavier (7.6%) and longer (2.5%) at metamorphosis (table 1 and figure 3). Although the relationship between density and body size was best described by quadratic fits (lower AIC), linear fits also performed well and were comparable (electronic supplementary material, S8). For condition (R_i), a significant interaction term suggests that the effect of density differed between morphs (table 1). Comparison of linear slopes for each morph showed that clear morph offspring exhibited a significant negative relationship with density (clear: −0.066, CI [−0.09, −0.04] while the white morph offspring did not (white: −0.014, CI [−0.04, 0.01]; figure 3). Despite evidence for morph-specific slopes, a decomposition of the interaction term showed that population density still had an overall negative effect on body condition (overall: −0.040, CI: [−0.06, −0.02], with white morph offspring tending to have higher condition (t = −2.8, p = 0.0060). These body condition results correspond to the independent effects of density and colour morph on body size.

Table 1.

Effects of colour morph and density on the mass, length and body condition of newly metamorphosed spotted salamanders. Bold numbers indicate a significant effect (p = 0.05) of focal model terms: colour morph and density.

response	term	estimate	s.e.	d.f.	t	p-value
mass	intercept	1.46	0.23	162.0	6.5	<0.0001
	morph (white)	0.11	0.03	402.3	4.2	<0.0001
	log(density)	0.23	0.17	402.1	1.3	0.1824
	log(density)2	−0.12	0.03	402.1	−3.7	0.0002
length	intercept	34.5	2.43	186.4	14.2	<0.0001
	morph (white)	1.15	0.28	402.3	4.1	<0.0001
	log(density)	3.23	1.84	402.1	1.8	0.0795
	log(density)2	−1.32	0.35	402.1	−3.8	0.0002
condition (Ri)	intercept	0.18	0.04	388.4	4.6	0.0093
	morph (white)	−0.13	0.05	403.0	−2.3	0.0195
	log(density)	−0.07	0.01	403.5	−5.1	<0.0001
	morph × log(density)	0.05	0.02	403.1	2.9	0.0043

Figure 3.

Effects of larva density and colour morph treatments on mass (a), length (b) and body condition (c) of metamorphosed spotted salamanders. Individual data (small points) and treatment means (large points) are plotted for each colour morph. Asterisks indicate significant effects in linear mixed models (***p < 0.001, **p < 0.01, *p < 0.05). A non-significant effect of density on condition for white morph larvae is plotted as a dashed line in the right-most panel. ind, number of individuals m⁻².

Variation in larva density and colour morph affected multiple components of water chemistry within our mesocosms (electronic supplementary material, S9). All comparisons of models with and without colour morph × density interaction terms showed that models without interaction terms were a better fit to the data—thus, the effects of colour morph and density variation were independent (electronic supplementary material, S9). Our general expectation that colour morph offspring would differentially affect aquatic ecosystems was supported; mesocosms stocked with white morph salamander larvae had higher dissolved organic carbon (p < 0.0001) and conductivity (p = 0.0046; table 2 and figure 4). There were no effects of colour morph on algal production or acidity (p > 0.18; table 2 and figure 4). Our expectation that population density would affect mesocosms was also supported. Density negatively affected dissolved organic carbon (p = 0.0038) and positively affected algal biomass (p = 0.0458) but had no effect on pH (p = 0.1828) or conductivity (p = 0.6367; table 2 and figure 4).

Table 2.

Effects of colour morph and population density treatments on water chemistry in mesocosms assessed using repeated measures (time). Bold numbers indicate a significant effect (p = 0.05) of focal model terms: colour morph and density. DOC, dissolved organic carbon.

response	term	estimate	s.e.	d.f.	t	p-value
pH	intercept	4.40	0.776	89.94	5.67	<0.0001
	initial	0.27	0.128	89.95	2.10	0.0390
	time (weeks)	0.02	0.000	88.09	8.12	<0.0001
	morph (white)	0.03	0.021	88.17	1.35	0.1791
	log(density)	0.02	0.013	88.09	1.34	0.1828
DOC (mg l−1)	intercept	−0.01	0.478	54.57	−0.02	0.9855
	initial	1.18	0.129	88.60	9.19	<0.0001
	time (weeks)	0.13	0.013	88.10	10.32	<0.0001
	morph (white)	0.43	0.09	88.23	4.81	<0.0000
	log(density)	−0.17	0.058	88.48	−2.98	0.0038
conductivity (μS)	intercept	88.1	11.69	91	7.53	<0.0001
	initial	0.09	0.095	91	0.92	0.3615
	time (weeks)	0.48	0.114	91	4.20	<0.0001
	morph (white)	2.26	0.779	91	2.91	0.0046
	log(density)	−0.24	0.497	91	−0.47	0.6367
algae (mg tile−1)	intercept	6.49	1.916	92.00	3.39	0.0010
	time (weeks)	−0.43	0.131	92.00	−3.25	0.0016
	morph (white)	−0.36	0.889	92.00	0.41	0.6862
	log(density)	1.16	0.574	92.00	2.03	0.0458

Figure 4.

Effects (standardized coefficients and 95% CI) of colour morph and density of larvae on aquatic chemistry and primary production. Asterisks indicate significant (p < 0.05) effects in linear mixed models. DOC, dissolved organic carbon.

4. Discussion

Here, we explored the ecological significance of phenotypic variation in the spotted salamander. Our study generated three general findings. First, colour morphs differed in vital rates; clear egg masses contained more eggs, but exhibited lower hatching rates (figure 2). Second, larvae hatched from white egg masses were larger and in better body condition at metamorphosis (figure 3). Third, our mesocosm experiment demonstrated density and colour morph-mediated effects on fundamental components of aquatic ecosystems (figure 4). Though we did not expect substantial colour morph-dependent variation in vital rates, body condition, or morphology, the ecosystem-level effects matched our broad expectation that spotted salamander larva effects would be density- and trait-mediated. Below, we discuss these primary findings, propose possible mechanisms underpinning these ecosystem effects, and speculate on the broader significance of jelly coloration for salamander populations and aquatic ecosystems.

(a) . Colour morph-dependent vital rates

The primary purpose of the common-garden experiment was to ensure similar rearing environments for all egg masses—a strategy that presumably helped isolate a phenotypic effect of jelly coloration that would later manifest in the mesocosm stage. However, it also provided insights into some basic biological differences between the colour morphs. Indeed, the colour morphs differed in several aspects. First, clear egg masses contained more eggs and produced more larvae (figure 2). A second, interesting outcome of the common-garden portion of our study was a significant survival difference (figure 2). Under the conditions of our common-garden experiment, this survival advantage was too small to equalize the apparent fecundity advantage of the clear morph. But assuming that morphs produce similar numbers of egg masses, these differences could determine, in part, whether and how this polymorphism is maintained in nature. To date, evolutionary models exploring the mechanisms underlying polymorphism maintenance have assumed equal fecundity [21,27]. Yet, in an evolutionary context, differences in vital rates such as these are non-trivial and worth exploring in future analyses of polymorphism evolution.

What might underlie variation in vital rates is unknown. Previous studies have examined whether a colour morph difference in female size might lead—through allometry—to fecundity differences between them [28,64]. Multiple studies have also explored whether a difference in jelly nitrogen content exists and could generate nutritional trade-offs between jelly and egg production [28,65]. Neither hypothesis has been sufficiently evaluated and both remain viable explanations for the fecundity differences we observe here. However, the colour morph difference in embryo survival we observed here (figure 2) matches the results of Pintar & Resetarits [27], who also reared egg masses in shallow outdoor mesocosms (electronic supplementary material, S1). Given the sensitivity of developing Ambystoma embryos and larvae to ultraviolet radiation [53,66–70], we suspect that the lower survival of clear morphs under these conditions is explained by the low transmission of harmful ultraviolet light through white jelly [21]. To date, a photoprotective function for white jelly has only been evaluated twice [53,54]. Neither study reports a significant difference in embryo mortality between white and clear morphs. However, the morph-dependent impact of ultraviolet radiation on embryonic deformities in Starnes et al. [54] is consistent with a photoprotective effect of white jelly (electronic supplementary material, S10). We look forward to additional efforts that help resolve the mechanisms underlying the substantial effects of this peculiar colour polymorphism on embryonic survival and the morphology of larvae.

(b) . Ecosystem-level effects of density and trait variation

Variation in the density of spotted salamander larvae had effects on algal and detritus food chains. The significant effect of increasing population density on benthic algae was expected and consistent with a gradual strengthening of an indirect trophic interaction. Given broad evidence for predator-driven trophic cascades along algal food chains, we presume that the density effect observed in our experiment is mediated by density-dependent depletion of herbivorous invertebrates (table 2 and figure 4). Why increasing larva density has an effect on dissolved organic carbon is less clear. The effects of density on dissolved organic carbon and algae are similar in size; however, the connections between salamander larvae and dissolved organic carbon are not readily characterized—even with data on intermediate trophic levels (e.g. [71]). Nevertheless, there is clear evidence that predators can and do affect detrital resources—including dissolved organic carbon pools—and that these effects are driven by trophic interactions. For example, in mesocosm studies similar to ours, predator effects of small, invertivore fish on dissolved organic carbon concentrations do occur [71–74]. However, because dissolved organic carbon pools are the sum of complex and dynamic decomposition pathways, the specific effects of predators are difficult to assess. Sometimes predators increase dissolved organic carbon; other times, they reduce it. But while outcomes appear idiosyncratic, the experimental responses to predator manipulation exhibit features consistent with indirect trophic effects. Thus, we speculate that the negative relationship between dissolved organic carbon and larva density reflects variation in the strength of an indirect effect initiated by differential top-down control of benthic invertebrates. Undoubtedly, additional work is needed to resolve these complexities. But considered together, our observation of density-mediated effects on algal and detrital resources provides clear evidence that spotted salamanders are broadly important components of aquatic food webs with the ability to influence carbon cycling within temporary ponds through predation.

Evolutionary trait-mediated effects on ecosystem attributes and processes are increasingly well documented [2–4,75]. In our case, intraspecific variation in jelly coloration of spotted salamanders led to significant differences in mesocosm water chemistry—dissolved organic carbon concentrations and conductivity, specifically (figure 4). Given the density-mediated effects discussed above, we believe that these morph effects could be driven by indirect trophic interactions. Again, while further study is needed to understand what is driving these ecosystem-level effects, there is reason to suspect that morphs could vary in their top-down effects owing to morphological differences. Indeed, multiple lines of evidence suggest that the larvae of white morph females are larger. First, Pintar & Resetarits [27] found that larvae from white egg masses were larger at hatching and maintained this size difference through weeks of development and growth. Second, in this study, metamorphic salamanders from white egg masses were significantly larger (figure 2). Third, an ongoing study of the population used in this experiment shows significant size differences between colour morphs at hatching, with larvae from white morphs being longer with proportionately wider heads (S. T. Giery 2023, unpublished; electronic supplementary material, S11). In combination, these results suggest that morphological differences between colour morphs are maintained throughout the larval state to metamorphosis—and perhaps beyond. But despite an increasingly consistent signature of morphological difference, the larger size of white morph larvae might not be general. Indeed, other studies have not found phenotypic differences between colour morphs [24,26]. These findings suggest the possibility of some context dependency; but understanding why morphs might differ in these cases will require additional investment in estimating the combination of direct and indirect genetic effects that contribute to larva phenotypes. Nevertheless, given that trophic interactions in spotted salamanders are size- and gape-limited [32], we suspect that these trait differences underlie the colour morph effects observed during our mesocosm experiment.

An important caveat to our inference of top-down effects in our mesocosms is that we do not account for linkages between the detrital and algal food chains. The varied processes that mediate decomposition and primary production in freshwaters are also linked by a variety of indirect, non-trophic interactions. For example, despite weak nutritional linkages with detritus, herbivores can mobilize dissolved organic carbon and accelerate the degradation of coarse particulate matter while grazing periphyton [49,76]. Thus, in a three-level algal food chain, predation might suppress decomposition rate and the release of dissolved organic carbon. This indirect effect of predation may explain why predator density had a negative effect on dissolved organic carbon in our mesocosms (figure 4). Another indirect effect might arise from shading of benthic algae by dissolved organic carbon, which strongly absorbs photosynthetically active radiation [77,78]. Even small increases in dissolved organic carbon can limit primary production. And in our mesocosms, interactions that increase dissolved organic carbon are likely to dampen any effects that favour primary production—including a reduction of herbivore abundance. Although these hypothetical linkages were not examined here, they are likely to shape predator effects on ecosystem processes.

5. Conclusion

Our experiment demonstrates the influence of a widespread and conspicuous colour polymorphism on aquatic ecosystems. The connections leading from variation in jelly coloration to ecological variation are, however, not obvious. As discussed throughout, a substantial share of the uncertainty is attributable to a fragmentary understanding of why larvae differ between jelly colour morphs and how this phenotypic variation is translated to ecosystem-level variation. Nevertheless, the trait-mediated effects we uncover here are broadly relevant to our understanding of aquatic food webs. The demonstrated trait-mediated effects on dissolved organic carbon are particularly intriguing, in part because spatial variation in dissolved organic carbon is largely shaped by topographic and climatic features [79–81]. However, mesocosm experiments demonstrate that dissolved organic carbon can also be altered, to some degree, by local variation in food-web structure and consumer abundance [71,73,82]. Because dissolved organic carbon effects are so potent, even small differences—a few mg l⁻¹ of carbon—can substantially alter aquatic physiochemistry [83] and influence population, community and ecosystem dynamics [84,85]. Although limited in many ways, small-scale experiments such as ours are crucial demonstrations of how evolved intraspecific variation in consumers can alter ecosystem properties.

Finally, spatial and temporal variation in colour morph frequency is a hallmark of this conspicuous and widespread polymorphism (e.g. [21]). The genetic underpinnings of jelly colour remain unknown—as discussed in the Introduction and elsewhere [21,27]. Yet, colour morph frequency appears to evolve readily—on contemporary timescales and microgeographic spatial scales [21,23,27]. Given our results, we speculate that the trait-mediated effects revealed here might constitute part of a broad-sense eco-evolutionary feedback (sensu [13]) linking morph frequency evolution to the profound ecological and evolutionary effects of dissolved organic carbon [58,86–88]. Although plausible, additional research is needed to assess this hypothesis. Insights such as these are key contributions to an emerging realization that evolved intraspecific trait variation can have ecological effects [2,89,90] and support a role for evolution in regulating fundamental ecosystem properties such as the carbon cycle.

Ethics

All work was done under IACUC no. 22-0029 and NC Wildlife Permit no. 22-SC01312.

Data accessibility

Data are available in the Dryad Digital Repository: https://doi.org/10.5061/dryad.73n5tb342 [91].

Supplementary material is available online [92].

Declaration of AI use

We have not used AI-assisted technologies in creating this article.

Authors' contributions

S.T.G.: conceptualization, data curation, formal analysis, investigation, methodology, resources, supervision, visualization, writing—original draft, writing—review and editing; R.K.S.: investigation, writing—review and editing; J.W.: investigation, writing—review and editing; A.G.: investigation, writing—review and editing; J.M.D.: conceptualization, investigation, methodology, project administration, resources, supervision, writing—review and editing.

All authors gave final approval for publication and agreed to be held accountable for the work performed herein.

Conflict of interest declaration

We declare we have no competing interests.

Funding

S.T.G. was supported by an Eberly Research Fellowship of the Eberly College of Science at The Pennsylvania State University.