Introduction

Islands have long been recognized as natural laboratories for the study of evolution (e.g., Darwin 1859; Mayr 1967) due to their restricted geographic boundaries, generally small size, and often traceable geologic history. Oceanic islands of volcanic origin, in particular, are well suited for uncovering patterns and processes of ecological diversification since they emerge as empty spaces available to be colonized by organisms, whereas mainland systems are generally saturated by diversified and structured communities (e.g., Gübitz et al. 2005; Losos and Ricklefs 2009). Morphological traits are among the best-studied characters that have been used to understand patterns and causes of variation on islands. Large morphological variation can occur on islands as a consequence of the exploitation of free niches in the newly colonized environment or from multiple distinct colonization events from diverse source populations (e.g., Van Valen 1965; Schluter and Grant 1984; Robichaux et al. 1990; Gillespie et al. 1997; Losos et al. 1998; Millien 2006; Losos and Ricklefs 2009; but see also Bolnick et al. 2007). The Galápagos archipelago, which consists of 19 major islands located approximately 960 km from the mainland (Snell et al. 1996), is among the best-known examples of volcanic islands with organisms exhibiting high levels of morphological variation (e.g., Parent et al. 2008). This archipelago has a relatively well-known geologic history, aiding studies of the patterns and processes of island colonization and ecological and morphological diversification (e.g., Grant and Grant 2002; Parent et al. 2008; Ali and Aitchison 2014).

The marine iguana, Amblyrhynchus cristatus, is one of the endemic organisms inhabiting the Galápagos. This species diverged from its sister group, the terrestrial iguanas of the genus Conolophus, approximately 8.25 million years ago. Divergence among currently existing marine iguana populations dates at less than 50,000 years (MacLeod et al. 2015). Molecular data suggest that the colonization of existing islands went from east to west, from older to younger islands of the archipelago, followed by some episodes of gene flow after populations were established on different islands (Steinfartz et al. 2009).

Morphological studies conducted on a few populations of marine iguanas have highlighted the existence of a wide range of body size variation in this species. Body size appears to be quite plastic in these animals, and changes in body length have been observed during periods of food shortage. For example, during El Niño events, marine iguanas may shrink in body length within just a few months (Wikelski and Thom 2000). Body size variation in marine iguanas has been attributed to natural selection—including trade-offs between thermoregulation, food intake, and optimal food digestion—and sexual selection (e.g., Wikelski and Trillmich 1994, 1997; Wikelski 2005). Marine iguanas possess a lek mating system, with a tight clustering of males defending small territories and attracting females (Wikelski et al. 1996, 2001). It has been suggested that maximum achieved body size reflects a balance between attracting mates and thermoregulatory and digestive performance (Wikelski and Trillmich 1997; Wikelski 2005). The extent of body size variation across the species distribution, however, is largely unknown, as is the impact of evolutionary history on this variation. Furthermore, the magnitude and causes of body shape variation are not well understood. Body shape variation may be functionally important for fitness-related activities and behaviors, as observed for other lizards (e.g., Garland and Losos 1994; Losos and Miles 2002; Herrel et al. 2008; Kaliontzopoulou et al. 2012; Scharf and Meiri 2013).

The evolutionary history and ecology of marine iguanas have been studied for many populations, permitting investigation of the factors influencing morphological variation in this species. Examination of correlations between evolutionary history, environmental conditions, and phenotypic variation has been used successfully in other organisms to understand the contribution of these factors to morphological variation and evolution within and among species (e.g., Malhotra and Thorpe 2000; Chiari et al. 2004, 2009). Building on previous studies of body size variation conducted on only a few marine iguana populations (e.g., Wikelski and Trillmich 1994, 1997), we assess the geographic patterns and environmental parameters influencing body size and shape variation on this species across its entire distribution. Furthermore, we evaluate the influence of evolutionary history, based on previously described phylogeographic patterns of marine iguanas from across the archipelago (Steinfartz et al. 2009), on morphological variation (size and shape). We also tested whether variation in morphology could be related to different environments. Our hypothesis is that body size is mostly dependent on local factors such as island perimeter (possibly related to food availability, see below) and environmental conditions (e.g., temperature and productivity), and not strongly dependent on evolutionary history. This would support previous indications that body size variation in marine iguanas is plastic (Wikelski and Thom 2000) and temperature dependent (Wikelski 2005; Walters and Hassall 2006). Conversely, as there is currently no indication that mating, feeding strategy, or locomotion differ among localities and therefore show local adaptations, we predict a stronger relationship between evolutionary history, genetic divergence, and shape variation among populations, suggesting little influence of plasticity on shape. Finally, due to the lek mating system of this species, we expect to find sexual dimorphism in both size and shape to occur.

Materials and methods

Samples

Fieldwork was carried out in the Galápagos Islands in 1993. Sampling localities were recorded as global positioning system coordinates and included 16 sampling sites (henceforth referred to as “populations”) across 12 islands (Fig. 1; Supplementary material S1).

Fig. 1
figure 1

Map of the Galápagos Islands with populations and mitochondrial clades indicated (modified from Steinfartz et al. 2009). Population abbreviations (EPC, FMO, FPE, FPM, GCA, ICW, IPA, MBN, PCI, PDL, SCC, SFN, SFS, SJB, SRL, SRP) are as in Supplementary material S1

Full size image

We only considered adults in this study, i.e., individuals for which sex could be positively identified based on external morphology (Dellinger and von Hegel 1990). Eight morphological variables were obtained with a calliper and a tape ruler, which included: snout-vent length (SVL); tail length (TL); length of the front limb (LP); length of the third digit of the manus (LD); height, length and width of the head (HH, HL, and HW, respectively); and jaw length (JL) (Supplementary material S2). These variables are commonly sampled in lizards for morphological and functional studies, as SVL, TL, LP, and LD are known to be correlated with locomotor performance, while HH, HL, HW, and JL are involved in feeding, courtship, and mating (e.g., Vleck et al. 1981; Garland and Losos 1994; Miles et al. 1995; Wikelski and Trillmich 1997; Wikelski and Thom 2000; Losos and Miles 2002; Kaliontzopoulou et al. 2012; Gomes et al. 2016). Furthermore, in marine iguanas, the fingers of the manus are important for gripping to resist waves and breaking surf while feeding on rocks on the shore and intertidal zone (Y. C. and S. G., personal observation)

Genetic data for all individuals, including mitochondrial (D-loop) sequences and genotype data for 13 microsatellite loci, were obtained from Steinfartz et al. (2009). Each individual was previously assigned to one of three mitochondrial clades (A, B, or C) by Steinfartz et al. (2009) based on D-loop sequence data (Fig. 1; Supplementary material S1); these clade assignments were also used in the current analysis. We used both D-loop DNA sequences and microsatellite loci because they reflect distinct evolutionary time frames, due to differences in coalescence times (Avise 2000). In this study, we examined only individuals for which both genetic (mitochondrial and/or microsatellite) and morphological data were available (Supplementary material S1). For the morphological analyses alone, and for the morphological and mitochondrial DNA data, we used 343 individuals, while 347 individuals were used for the comparison of morphological and microsatellite data.

Morphometric and genetic analyses

All morphometric analyses were run in R version 3.0.1 (R Core Team 2013). Analyses were run on size and shape independently. Size and shape information were extracted from linear measurements using the log-shape ratio approach described in Mosimann (1970). Following this approach, size was estimated as the geometric mean from all the measurements of each individual, while shape corresponded to the log-shape ratio (here indicated simply as “log-shape”) of linear measurements on size (Mosimann 1970). Mosimann’s approach avoids arbitrary choices of which variables should represent size and is the best available method to accurately depict variation in size and shape, especially when individuals of similar shape have different sizes and vice versa (Jungers et al. 1995). The use of log data allows linearizing relationships with covariables and to easily analyze the presence of allometries—the relationship between changes in shape due to changes in size during growth. The use of the geometric mean helps account for the fact that size information is carried by every linear measurement. Due to the way they are calculated, log-shape ratios could not be considered independent. Therefore, prior to running the multivariate explanatory analyses on shape data, we performed a principal component analysis on log-shape to remove the last null component of variation (Claude 2008, 2013). In addition, the effect of explanatory variables was also assessed on every single log-shape ratio to identify which parts of shape were affected (see below) (Claude 2008, 2013).

Congruence in how individuals are grouped by morphology (log-size and principal components of log-shape), genetics (microsatellite loci), geography (geographic coordinates), or a combination of these (Supplementary material S3) was tested using the program Geneland version 4.0.0 (Guillot et al. 2012) in R. Geneland uses a Bayesian approach to detect if individuals can be grouped based on similarity in the data, and how many clusters (groups) of individuals occur, without having a priori information on the data. Datasets were run with ten independent runs, uncorrelated allele frequencies, maximum rate of Poisson process set to one hundred, 106 iterations, and thinning of 100. When analyses included geographic coordinates, we used an uncertainty in the coordinates of 0.02 degrees [about 2-km coastline, which corresponds to the approximate maximum migration estimate of marine iguanas (Lanterbecq et al. 2010)] to account for individual dispersal.

To test for differences in size among populations, sexes, and mitochondrial clades, a three-way ANOVA was run on log-size data using the F-test and type II sum of squares with population, sex, and mitochondrial clade as factors [factors are unbalanced within each category (Chiari et al. 2009; Claude 2013)]. To estimate possible differences in sexual size dimorphism among mitochondrial clades and populations, the interactions between factors were also taken into account.

Differences in shape between populations, sexes, and mitochondrial clades taking into account effects due to allometric growth (relationship between log-shape and log-size) were estimated through a multifactorial multivariate analysis of covariance using type II sum of squares and products of the seven non-null PCs shape variables, with population, sex, and mitochondrial clades as factors and log-size as covariate. To estimate possible differences in sexual shape dimorphism among mitochondrial clades and populations, the interactions between factors was also taken into account. To test for differences among populations and sexes for each log-shape ratio, the effect of the explanatory variables was also assessed on every single log-shape ratio alone. In these analyses, effects may be nested (e.g., clades within populations; Supplementary material S1). In this case, the df for analyses with nested and interaction effects take into account that some combinations of categories are absent in our sample design (e.g., missing one or two clades for some populations; Supplementary material S1). To test for differences in pairwise comparisons among mitochondrial clades and sexes for each log-shape ratio a Tukey test was run with mitochondrial clades and sex as factors.

Environmental data

To analyze the relationship between environmental parameters and size and shape variation among populations, we obtained oceanic data on intertidal productivity estimated from chlorophyll a concentration (mg m−3) and sea surface temperature (SST; °C) from moderate resolution imaging spectroradiometer data in the NASA Ocean Color site (oceancolor.gsfc.nasa.gov/) representing the period between 2000 and 2011, since data for the sampling year (1993) were not available. Although we did not have images of the year of sampling (low satellite coverage was available in 1993), we assumed that a decade of data would adequately represent the overall temporal and spatial pattern (see “Results”). We used chlorophyll a concentration as an indicator of algal productivity among populations because it is directly linked to marine iguana feeding requirements since they primarily eat algae (Drent et al. 1999). We used SST instead of deep-water temperature since marine iguanas feed and utilize the marine environment near the shore, in the intertidal and subtidal zones (Trillmich and Trillmich 1986). Often, marine iguanas feed with most of their bodies just barely covered by sea water (Y. C. and S. G., personal observation). Although males may swim to deeper water to feed, they typically stay within a few meters depth; therefore SST can be considered a good proxy for the water temperature experienced by marine iguanas, while sea depth would not be representative of the habitat experienced by these animals. We also obtained average monthly land data on precipitation (mm), and minimum, maximum and mean air temperature (°C) from WorldClim, which compiles weather station data from the period between 1960 and 1990 (Hijmans et al. 2005). The value for each sample location was assumed to be represented by the nearest pixel with a 2.5 arcmin spatial resolution (~4 km) for all variables. Remote sensing allows obtaining images of different wavelengths for specific times and localities. These data were related to variables of interest such as chlorophyll a. In situ validation of the resulting data is frequently updated (Werdell et al. 2003). ANOVA was used to assess whether the variation at each environmental parameter among populations followed the same trend over time, with population and year as factors. Island perimeter (m), which represents the coastline and therefore the actual space mostly used by marine iguanas, was measured from coastline geographic information system data (global self-consistent, hierarchical, high-resolution geography database; www.ngdc.noaa.gov/mgg/shorelines/gshhs.html) that was previously transformed to the Albers equal-area projection.

We ran a correlation analysis and a bivariate regression with population mean values for each environmental parameter calculated across the monthly or yearly time spans versus the population mean log-size and mean log-shape and mean log-size and mean log-shape of females and males separately. We also ran a regression between island perimeter and chlorophyll a to estimate if resource productivity may be correlated to island perimeter. Regressions were run in R using t-test statistics (H 0 = no relationship) and two-sided p-value estimates.

Results

Size variation

Mean log-size differed significantly among populations (Table 1; Supplementary material S4). The smallest mean log-size (independently of sex) occurred on Genovesa (GCA; log-size = 4.1), while the largest mean log-sizes were found in populations from Fernandina (FPE), Isabela (ICW and IPA), Floreana (FMO), and San Cristóbal (SRL and SRP) (Supplementary material S4). Sexual size dimorphism occurred in all populations (Table 1) except Genovesa (GCA), with the most pronounced size difference between males and females occurring in one of the populations from Isabela (ICW) (Fig. 2; Supplementary material S4). Although males were generally bigger than females (Fig. 2; Supplementary material S4), the degree of sexual size dimorphism differed among populations (Table 1).

Fig. 2
figure 2

Box plot of size (not log transformed) considering the distinct populations and each sex separately. Horizontal black bar represents median value; whiskers represent most extreme values (maximum and minimum values). Populations are ordered from the maximum to the minimum body size and indicated with abbreviations as in Supplementary material S1. For each population the box plot on the left represents males and the box plot on the right corresponds to females

Full size image
Table 1 Influence of population (Pop.), sex, and mitochondrial clade, and their interactions, to explain variation in mean log-size in Galápagos marine iguanas
Full size table

ANOVA results indicated that the three major mitochondrial clades previously described for marine iguanas (Steinfartz et al. 2009) did not explain log-size variation and sexual size dimorphism (Table 1). When Geneland was run with log-size data only, four clusters were recovered that did not correspond to any geographic grouping of sampling localities or to the microsatellite data clusters (11) previously identified by Steinfartz et al. (2009) and confirmed here (Supplementary material S3), indicating that individuals within specific populations or microsatellite clusters were not more similar in log-size than between populations or microsatellite groups. Only individuals from Genovesa (GCA) formed a distinct size group, while the remaining three clusters comprised individuals from multiple populations (Supplementary material S3). When Geneland was run on log-size data taking into account microsatellite data, the clustering was similar to that obtained with microsatellite data alone (Supplementary material S3). When Geneland was run on log-size data taking into account geographic coordinates or using genetic data and geographic coordinates together, the results were unreliable, as different numbers of clusters were obtained for the different runs, suggesting a lack of convergence of these runs (data not shown).

Shape variation

Mean log-shape differed significantly across populations, sexes (i.e., sexual dimorphism), and mitochondrial clades (Table 2; Supplementary material S5). Shape allometries—defined as changes in shape relative to changes in size—occurred and differed among populations (Table 2) and among sexes in distinct populations (Table 2). Each log-shape ratio differed significantly among populations (Supplementary material S5), with differences in shape allometries among populations only occurring for LP and HL (Supplementary material S5). Mitochondrial clades differed significantly in LP (Supplementary material S5), with individuals from clades C and A having the longest and shortest legs, respectively (data not shown). There was a significant interaction between mitochondrial clade and population, suggesting that evolutionary history, as defined by the phylogeographic distribution of the three mitochondrial clades, may not be the main factor influencing variation in LP among populations (Supplementary material S5). SVL and head measurements (HH, HL; Supplementary material S5) differed among sexes, indicating sexual dimorphism. According to the results of the Tukey test, females have longer heads than males across populations (p-value = 0.0003, data not shown), while males have higher heads than females (p-value < 0.0001, data not shown).

Table 2 Mean log-shape differences taking into account sex, population (Pop.), mitochondrial clade and allometries (Log-size) as factors
Full size table

When individual log-shape data were analyzed alone to look for possible structure in how individuals are grouped, six clusters were obtained in Geneland, with individuals in each cluster being distributed across multiple populations and only individuals from Genovesa (GCA) forming a single group distinct from the rest (Supplementary material S3). When microsatellite data were included in the Geneland analysis, clustering was primarily determined by microsatellite data structure, as shown by the correspondence between microsatellite clustering alone and the pattern obtained by analyzing log-shape data and microsatellites together (Supplementary material S3). Results on the influence of sampling locality (geographic coordinates) on log-shape could not be obtained due to a lack of convergence of the runs in Geneland (data not shown).

Influence of environmental parameters on size and shape variation

With the exception of chlorophyll a, the environmental parameters considered in this study generally showed a similar trend of variation among populations across the 12-year period or across months over a 30-year period. ANOVA interactions of year/month and population were not significant except for chlorophyll a (p-value <0.001, data not shown). These results suggest that environmental data exhibit consistent patterns across years and that the lack of environmental data for the particular year of morphological and genetic sampling for this study should not bias the results, as it is unlikely that there would be large variation in these parameters among the populations over the study period. We found no relationship between population values of chlorophyll a and island size (data not shown).

We found a significant negative relationship between mean log-size variation among populations and SST, and a positive relationship with island perimeter (Table 3). SST and island perimeter were also negatively and positively correlated, respectively, to mean log-size variation across populations in both females and males (Table 3). No significant relationship was observed between mean log-size and any other environmental parameter considered in these analyses (Table 3). We did not observe any significant relationship between log-shape or log-shape for females and males analyzed separately and any of the environmental parameters or island size (data not shown).

Table 3 Pearson correlation coefficient and correlation analysis between population mean log-size, population mean log-size per sex and environmental parameters
Full size table

Discussion

Size and shape variation

In animals, heat loss is proportional to the surface area to volume ratio, with larger organisms having a smaller ratio and thus reduced heat loss. On this physical basis, many ecological rules have been proposed to explain the relationship between body size and external temperature. Marine iguanas, like all reptiles, are ectotherms; thus, their body temperature depends on the external climate. In ectotherms in particular, a relationship between variation in body size and temperature (temperature-size rule) has been proposed, with colder temperatures leading to an increase in body size (Angilletta and Dunham 2003; Walters and Hassall 2006 and references therein; Kingsolver and Huey 2008).

Our results reveal significant body size variation among populations of Galapágos marine iguanas, which is not correlated to evolutionary history (i.e., genetic divergence in microsatellites or phylogeographic patterns based on mitochondrial DNA) among populations. This supports previous indications that body size in marine iguanas may be highly plastic (Wikelski and Thom 2000). Mean body size is found to be influenced by SST and island perimeter. For adult marine iguanas, which do not have natural predators and for whom aggressive inter-specific competition is largely absent, maximum body size is mainly constrained by the amount of algal pasture available, the time spent feeding, and thermoregulation (Wikelski and Carbone 2004). Our results, based on a larger sample size and a more widespread sampling than previous work (e.g., Wikelski and Carbone 2004), support the importance of thermoregulation, but not productivity, in influencing mean body size in this species.

Marine iguanas feed on algae either in the intertidal costal area or by swimming and diving into the cold oceanic water offshore to feed on submerged algae (Drent et al. 1999). Although energy expenditure between these two foraging modes is similar (Drent et al. 1999), the warming up rates can be quite different, thereby influencing foraging and digestive efficiency (Wikelski and Carbone 2004). The observed relationship between SST and mean body size suggests that temperature may constrain body size in these animals by determining the time an animal can spend in the water feeding, either in the intertidal area or in the offshore waters. Because larger body mass corresponds to decreased body surface area to volume ratio, larger individuals are generally less sensitive to heat loss (e.g., Pincheira-Donoso et al. 2008). Therefore, having a larger body size and mass may allow marine iguanas to be less sensitive to heat loss due to exposure to cold water during underwater feeding, where food is more abundant, or in localities where water is colder.

Although the results presented here indicate a role for thermoregulation in influencing mean body size in marine iguanas, it is also possible that colder waters may be more productive and may harbor more abundant and higher quality food, which would enable individuals to grow larger. In this study, we used chlorophyll a as a measure of productivity, but future work could examine other measures (e.g., algal height or type of algae) to further test the relationship between food availability, temperature, and body size. It is also plausible that the association between island perimeter and body size found in this study may be related to the number of sites available for feeding, and thus related to the amount of algal pasture available.

It is also possible that differences in longevity among populations may explain some of the variation in mean size between sampling localities, as populations with older, and thus larger, individuals may have higher mean size than populations with younger, smaller individuals. There are no basic data available on age structure and longevity of marine iguanas across the archipelago to test this question directly. However, marine iguanas, like reptiles in general, experience indeterminate growth, so that once sexual maturity is reached, growth slows down and largely depends on available resources. Since we only sampled adults in our study, observed size differences are not expected to be a function of age, but rather available resources.

Our results indicate that populations of marine iguanas differ both in mean shape and at each of the morphological variables analyzed. Mitochondrial clades, but not microsatellite loci grouping, are associated with variation in shape and leg length. This could suggest that evolutionary history, more than recent population differentiation (as depicted by microsatellites), has influenced shape variation across populations. In contrast to what was observed for size variation among populations, environmental parameters do not have an effect on shape, suggesting that local adaptation for shape variables may not have occurred.

Size and shape sexual dimorphism

Size and shape sexual dimorphism occur in all populations, with size sexual dimorphism differing among populations. Males are larger than females on all islands except Genovesa. Adaptive plasticity and/or genetic adaptation of sexual size dimorphism may therefore occur in this species. These results confirm previous work on a small number of populations (e.g., Wikelski and Trillmich 1997). Both natural and sexual selection are likely to influence sexual size dimorphism in these animals (Wikelski et al. 1996; Wikelski and Trillmich 1997; Wikelski and Romero 2003). Larger male size would offer an advantage during male fighting for territory and positioning in the lek. In addition, larger males are favored for mating (Wikelski et al. 1996), with male mating success highly skewed toward a few successful males due to female choice (Wikelski et al. 2001). Females generally mate with the males showing more active mating behavior (Wikelski et al. 1996, 2001). However, larger males would also be more sensitive to food availability (size-dependent mortality), requiring a larger amount of food overall (Wikelski and Trillmich 1997).

Mean body shape and individual morphological variables also differ between the two sexes. Males have higher and shorter heads than females. These traits could be associated with head use in male–male competition and mating behavior (e.g., Herrel et al. 2001; Scharf and Meiri 2013). Mating behavior involves lek formations, with the most attractive males occupying the center of the lek, and consists of active male–male fights with head-push or head-bob interactions with competing males and head-bobs or copulation attempts with females (Wikelski et al. 2001; Partecke et al. 2002).

Conclusion

Our results indicate that body size and shape differ among populations of marine iguanas. Variation in mean body size among the study locations is correlated with SST and island perimeter. Variation in body shape among populations could instead be partially explained by evolutionary history as depicted by species phylogeography. Sexual dimorphism occurs in all populations except one, Genovesa. Future work should address the influence of plasticity versus adaptation in relationship to the observed differences in body size and shape among sex, location, and island. In particular, our results indicate that Genovesa may represent a unique system among the Galapágos islands and that local adaptation may be occurring for body size and shape on this island. Studies focusing on the possible association between island size, population size, individual competition for basking sites, and distance among lekking groups could provide further understanding of the observed differences in body size among populations. Ecological and behavioral work will help address whether differences in body shape among populations may be associated with diverse functional performance (e.g., swimming, gripping onto rocks), different population densities, mating behavior, or diverse habitats.