Development of Distinct Morphotypes in Captive Seychelles–Aldabra Giant Tortoises
Abstract
Growth patterns of captive-bred Seychelles–Aldabra giant tortoises (Aldabrachelys/Dipsochelys) were studied. This enabled the comparison of the development of 3 distinct morphotypes variously ascribed by different authors to 3 distinct species or a single, variable species (A./D. dussumieri/gigantea, arnoldi, and hololissa). Geometric morphometric analyses identified differences in growth pattern between the 3 morphotypes of 234 juvenile tortoises reared under identical conditions. In plastral characters, all 3 morphotypes could be distinguished from hatching. Initially, hatchlings were very similar in dorsal view but by 30-cm straight carapace length were distinguishable by relative warp analysis. The arnoldi morphotype is the most distinctive, with constriction in the center of the carapace resulting in the beginning of development of saddle-backed morphology from 30 cm. All 3 morphotypes show 2 distinct growth patterns, one from hatching to 20–30 cm and a different pattern above 30 cm. As these morphotypes were reared under identical, largely constant conditions, this change in growth patterns appears to be a result of ontogeny rather than environment. Differences in development of the morphotypes cannot be explained by environmental factors alone and may be the result of differences in gene expression or of small differences in genes associated with skeletal development.
Tortoise external morphology is variable in captive populations but apparently less variable in wild populations, although there are few quantitative studies to demonstrate this (e.g., Juvik et al. 1991). However, distinct local morphotypes are known for many tortoise species, often described as distinct taxa despite a lack of identifiable genetic differentiation. This is most apparent in the genus Testudo, where some 25 geographical forms have been described based on morphology (Pieh and Perälä 2002; Guyot-Jackson 2004), but only about 10 genetic groups have been considered significantly distinctive (Parham et al. 2006; Fritz et al. 2007, 2009a, 2009b); a recent checklist tentatively recognizes 14 separate taxa in Testudo (including Chersine but not Agrionemys; Turtle Taxonomy Working Group 2009).
These differences can be hypothesized to be due to local ecological factors or population differences in allele frequencies influencing carapace development. Some morphotype differentiation does have a genetic element, as is recognized in several “saddle-backed” and domed taxa of Galapagos giant tortoises (Caccone et al. 2002; Chiari et al. 2009). Similar morphotype differentiation has been reported for Indian Ocean giant tortoises (Aldabrachelys/Dipsochelys) since the early taxonomic studies of this confusing group (Gunther 1877; Rothschild 1915; Bour 1982; Gerlach and Canning 1998).
Recognizing these morphotypes and determining whether this variation is individual, pathological, ecophenotypic, or genetic has been confounded by extinction of all wild Indian Ocean giant tortoises by the mid-1800s with the exception of 1 population on Aldabra atoll—the Aldabra giant tortoise (known variously as Aldabrachelys gigantea, Dipsochelys dussumieri, or Dipsochelys elephantina; nomenclature currently under review by the International Commission on Zoological Nomenclature). Almost all museum specimens lack reliable provenance data, and most, if not all, were derived from captive animals of unknown history (Gerlach and Canning 1998; Gerlach 2004b). Living captive adults similarly lack data on origins and captive history.
Three living morphotypes of Aldabra–Seychelles giant tortoises have been recognized in recent years and attributed to 3 distinct species (Gerlach and Canning 1998): the Aldabra tortoise (referred to A. gigantea or D. dussumieri) and 2 purported species from the granitic Seychelles islands, now extinct in the wild (referred to as Dipsochelys arnoldi and Dipsochelys hololissa). These forms are distinguished by general shape (flattened to saddle backed in arnoldi, evenly domed in gigantea/dussumieri, and flattened domed in hololissa). Shape is variable, and exact scute proportions (costals, vertebrals, and plastral scutes) and skeletal characters provide more accurate diagnoses (Gerlach and Canning 1998; Gerlach 1999a; Gerlach and Bour 2003).
Although the 3 morphotypes are distinguishable morphometrically, genetic studies have failed to find any consistent genetic grouping (Austin et al. 2003; Palkovacs et al. 2003; Gerlach 2005b). This may suggest that all giant tortoises from Aldabra and the granitic Seychelles islands form a single, variable species (Fritz and Havas 2007) or possibly morphologically defined subspecies as provisionally recommended by the Turtle Taxonomy Working Group (2009). However, there remains considerable uncertainty over the use of the most appropriate genetic markers for tortoise species differentiation, and further research into the most appropriate genes for analysis in this group is needed (Shaffer et al. 2007).
Geometric morphometric methods and multivariate statistics allow shape changes through ontogeny to be quantified (Monteiro 1999; Monteiro et al. 2000) provided that data are not confounded by uncertainty over developmental history. Captive breeding of all 3 morphotypes on Silhouette Island, Seychelles, provides the opportunity to investigate the development of the shell morphology based on juveniles of known parentage, raised under identical conditions. This should allow the main ecophenotype and genetic aspects of morphotype development to be distinguished. The results of this investigation are described here.
METHODS
The tortoises used in the present study were all captive bred on Silhouette Island, Seychelles. Breeding stock comprises 6 adults of each morphotype: 4 males and 2 females of the arnoldi morphotype (all long-term captives, captive bred), 2 males and 4 females of typical dussumieri/gigantea morphology (5 imported from Aldabra, 1 formerly free range on North island, all captive for at least 20 years), and 4 males and 2 females of hololissa morphology (4 long-term captives of unknown origin, 2 free range from Cerf Island). Tortoises are kept in large, naturally vegetated enclosures (Gerlach 2004a, 2005a). In these conditions, breeding has been successful with a total of 234 juveniles produced between 2002 and 2009. Most eggs have been incubated artificially in incubators kept at 26°–30°C (female incubator) and 29°–32°C (male incubator) at a relative humidity of 85%–95%.
Captive-bred juveniles are kept in secure boxes with a paper substrate and natural lighting until they reach 100 g. At this weight, they are transferred into outdoor enclosures. Food and water are provided ad libitum. Food comprises soft leaves of native coastal plants (Canavalia cathartica and Vigna marina) supplemented with fruit of Artocarpus altilis for the 30–100 g juveniles. Larger juveniles are provided with a more varied diet, with an increasing proportion of Ipomoea pes-caprae, the main plant species fed to the adults. Juveniles are individually identified with a painted number on the carapace but are not separated. As a result, juveniles of all 3 morphotypes have been reared under identical conditions. The first successful breeding occurred in 2002, with 234 tortoises hatching by 2010 (137 arnoldi, 32 dussumieri/gigantea, 40 hololissa), providing 7 years of growth data (Table 1).
For the present study, photographs were taken of 3 views of each tortoise in 2009: dorsal, lateral, and ventral. Size was recorded as straight carapace length (to the nearest 0.5 mm). The parents of the captive-bred juveniles were included in the analysis. The following measurements were recorded to the nearest millimeter using calipers or a flexible tape (for scute measurements): straight carapace length (CL); width; maximum height; height to the highest point of the costals; length and width of all vertebral, costal, and plastral scutes; and depth of the anal notch. These are all measurements that have been considered to be useful in distinguishing the morphotypes at a range of sizes (Gerlach and Canning 1998; Gerlach and Bour 2003).
Images were digitized and landmarks defined using the program tpsDig (Rohlf 2008a). Landmarks were selected to describe the size and shape of the shell and individual scutes in dorsal (50 landmarks), ventral (34), and lateral (29) views (Fig. 1). The landmark coordinates were used to perform a generalized Procrustes analysis using TpsRelw (Rohlf 2008b), superimposing the specimens onto a common system of coordinates and removing the effects of orientation and scale (Rohlf and Slice 1990). Relative warp analysis (a principal component analysis on the residuals from the superimposition of the landmarks) produced plots of each specimen on variability axes (relative warps). Relative warp analysis quantifies shape variation independently of size, translation, and orientation by superimposing landmark coordinates from all specimens to produce a consensus set of coordinates around which the principal component analysis is carried out. For visual interpretation, thin plate spline deformation grids were prepared for groups or axes of interest; these represent the transformations of each relative warp away from the consensus. Statistical analysis was carried out using multiple analysis of variance on the scores of each specimen in the first 4 warps (these explained at least 90% of the variation; Table 2). For analysis, tortoises were grouped by parental morphotype, sex (assumed from incubator temperature and confirmed with tail scale counts), and size (in 10-cm CL categories). Parental individuals were not used in the initial analysis; these were added subsequently to provide a visual indication of the relationship between adult and juvenile morphologies. Regressions of individual measurements against CL were carried out for regions of the shell with notable growth patterns highlighted by the relative warp analysis.
Citation: Chelonian Conservation and Biology 10, 1; 10.2744/CCB-0828.1
RESULTS
The relative warp analysis for each size group separates the morphotypes to a varying extent. For some orientations and size groups, overlap between morphotypes was high (e.g., 100% for hatchlings in dorsal view), but at other stages complete separation occurred (Figs. 2–4). Analysis of variance found no significant effect of sex but did identify morphotype and size as significant grouping variables, with the strongest effect being for the interaction of morphotype and size (dorsal size F5219 = 2.66, p < 0.05; dorsal size × morphotype F10,219 = 4.14, p < 0.001; lateral size F5219 = 2.69, p < 0.05; lateral size × morphotype F10,219 = 6.99, p < 0.001; plastral size F5219 = 2.99, p < 0.05; plastral size × morphotype F10,219 = 5.78, p < 0.001). Clearest separation was found for the 21–50 cm size group in dorsal view, 11–20 cm and 41–50 cm in lateral view, and hatchlings and 31–50 cm in plastral view. The 3 morphotypes exhibited notably different patterns of morphological change; these are summarized in Figs. 2–4. In these figures, the pattern of change for each species is represented by an arrow that passes through the center of each of the 95% ellipses for successive size categories. This provides a visual (but not statistical) representation of the developmental trajectory.
Citation: Chelonian Conservation and Biology 10, 1; 10.2744/CCB-0828.1
Citation: Chelonian Conservation and Biology 10, 1; 10.2744/CCB-0828.1
Citation: Chelonian Conservation and Biology 10, 1; 10.2744/CCB-0828.1
In dorsal view, all 3 hatchling morphotypes were initially indistinguishable in this morphometric analysis. A previous analysis did distinguish hatchlings, but in dorsal view this was significant only on the basis of coloration, distinctions being mainly in lateral and ventral views (Gerlach and Bour 2003). The dussumieri/gigantea morphotype became increasingly rounded throughout its growth, initially becoming broader at the center (increase in the first relative warp axis [RW1]), with increase in marginal expansion from the 31–40 cm category on. The hololissa and arnoldi morphotypes exhibited very similar development patterns, becoming broader in the center and then flaring at the anterior and posterior marginals from the 21–30 cm category. The flaring was slightly more apparent in hololissa, with a greater tendency toward constriction at the center of the carapace (mainly at the suture between vertebrals 2 and 3) in arnoldi.
In plastral view, the 3 morphotypes were clearly separated at hatching, primarily in RW1 (elongation of the anterior and posterior lobes of the plastron), with the hololissa morphotype being distinctive in constriction of the posterior width (RW2). At the < 10 cm stage, there was a convergence of morphology, but by the 21–30 cm stage, hololissa were distinct with a relatively long plastral lobes; arnoldi and dussumieri/gigantea were not distinctive until the 31–40 cm stage, when they were separated by arnoldi developing a relatively narrow posterior lobe and dussumieri/gigantea retaining largely allometric growth, developing a narrower anterior lobe. These differences are most obviously manifest in the posterior lobe of adults: narrow with a shallow anal notch in arnoldi, broad with a shallow notch in hololissa, and moderate but deeply notched in dussumieri/gigantea. The relatively constricted anterior lobe of dussumieri/gigantea results in a curvature of the humero-pectoral suture; in the other 2 morphotypes, a more pronounced angle is present at the midpoint of the suture, providing an obvious visual character in adults.
In lateral view, the dussumieri/gigantea morphotype is clearly distinct at hatching in RW1, with a low carapace and posterior shortening. This develops largely allometrically with an increase in the height of the dorsal part of the carapace and postero-dorsal expansion. In contrast, hololissa retains an approximation to the consensus morphology until the 31–40 cm stage, when it increases in dorsal height, and arnoldi has a highly distinctive growth pattern, initially increasing in height in the dorsal half of the carapace and in width at the center of the view, until the 31–40 cm stage, when dorsal height growth decreases and ventral height increases, combined with a constriction of the center of the dorsum (at the second and third vertebral suture). At all stages the 3 morphotypes remain largely separated (no more than a 25% overlap between any 2 morphotypes).
Regression of vertebral 2 and vertebral 3 and costals 1 and 2 and depth of anal notch against CL produced significantly different results in growth patterns for the different morphotypes (Table 3; Fig. 5). Results from anterior plastral lobe width and vertebral 4 were not significant. These confirm that the arnoldi morphotype develops significantly differently from hololissa and dussumieri/gigantea in the region of the third vertebral and that dussumieri/gigantea differs from arnoldi and hololissa in posterior plastral lobe development.
Citation: Chelonian Conservation and Biology 10, 1; 10.2744/CCB-0828.1
DISCUSSION
The captive-bred tortoises analyzed here differ in growth patterns according to their parental morphotype. Specifically, in comparison to the dussumieri/gigantea morphotype, the arnoldi morphotype is associated with constriction in the region of the third vertebral scute in individuals over 30 cm CL, and both arnoldi and hololissa differ from dussumieri/gigantea in the development of the posterior plastral lobe. These results are in accordance with previous analyses of the osteology of adults (Gerlach 1999a) that identified the exceptionally reduced size of neural bones 3 and 5 and expansion of neural 2 in the arnoldi morphotype as major contributors to the development of its distinctive carapace morphology. This was suggested to be associated with the relatively small size of vertebral scute 3 and the relatively large size of costal scute 1. As is demonstrated in the present study, constriction of the growth of the center of the carapace (corresponding to neural bones 3–5) will tend to draw up the anterior margin of the carapace (see the distortion of the lateral view in Fig. 4g), resulting in the flattened or “saddle-backed” appearance. For hololissa, a reduction in ossification compared to dussumieri/gigantea was suggested to result in a reduction in carapace strength and a lower dome, producing a low, flared carapace (Fig. 3g).
The external appearance of the chelonian carapace has been the principal characteristic used in their identification and classification (e.g., Duméril and Bibron 1835; Wermuth and Mertens 1961), although its form has been reported to be affected by diet (Medica et al. 1975; Jackson et al. 1976; Kirsche 1984; Lambert et al. 1988). Its development is the result of the interaction of multiple (largely unidentified) genes (Loredo et al. 2001; Cebra-Thomas et al. 2005; Kuraku et al. 2005). The precise genetic controls of shell development are not known; it is possible that relatively small mutations in some of these genes could produce the changes described here. In the case of the distinctive arnoldi morphotypes, this may be associated with small changes to the pattern of carapace development. The hololissa morphotype may be associated with reduced ossification, possibly in association with calcium metabolism. In both cases, distinctive morphology may result from small changes in functional genes or differences in population allele frequencies; if these are recent in origin, they will not be reflected in the neutral or conservative regions used in phylogenetic analysis. Alternatively, it is possible that there may be maternal effects on development; these have not been investigated thoroughly in chelonians to date, but a study of plastron morphology in the red-eared slider Trachemys scripta found no evidence of significant maternal effects contributing to differences between clutches (Myers et al. 2006).
In the tortoises studied here, there are relatively small differences between the morphotypes on hatching, principally in plastron shape and in height (as previously reported using principal components analysis; Bour and Gerlach 2003). Juvenile morphology appears to change most significantly from 30 cm CL and particularly from 60 cm CL (arnoldi; Fig. 6e, f). This corresponds to the change detected in assimilation data from wild Aldabra tortoises at 2–7 kg weight (Hamilton and Coe 1982); based on captive data, this corresponds to a CL of at least 21 cm (age 3–4 years), suggested to result from a dietary shift at this size (Gerlach 1999b). A shift in behavior, diet, and energetics at this size in the Aldabra tortoise has been suggested to be due to their having outgrown food-rich, sheltered crevices by 3 years (Swingland and Coe 1979; Coe and Swingland 1984) and subsequently being exposed to different growth conditions (particularly diet). The captive groups in the present study were all provided with the same diet, with changes in composition only at transitions from hatchling to < 10 cm CL categories. This suggests that changes in energetics and morphology result more from ontogenetic changes in gene expression than in behavioral or microhabitat changes. Although there are several reports of changes in growth rate with age in tortoises (Aresco and Guyer 1999; Hailey and Coulson 1999; Chen and Lue 2002; Znari et al. 2005) there have been few studies of the pattern of growth. Galapagos tortoise morphotypes are reported to result from ontogenetic changes in growth patterns (Fritts 1984), but this has not been fully quantified. Differences in growth patterns between juvenile and adult tortoises may result from known differences in their diets or from ontogenetic changes as reported here. The present study demonstrates the value of captive breeding in providing comparative data on growth without the confounding factor of microhabitat selection.
Citation: Chelonian Conservation and Biology 10, 1; 10.2744/CCB-0828.1
Landmarks used for morphometric analysis, dorsal (a), plastral (b), and lateral (c) views.
Changes in relative warps 1 and 2 for dorsal view. a–f) RW scores with 95% ellipses shown for each morphotype: a) hatchlings; b) < 11 cm straight carapace length; c) 11–20 cm; d) 21–30 cm; e) 31–40 cm; f) > 41 cm and adults. g) Patterns of change in RW scores shown by arrows for the 3 morphotypes (fitted by eye) and thin-plane spline images for the extremes of each axis. Open circles and arrow: arnoldi; closed circles and solid arrow: dussumieri/gigantea; shaded circles and arrow: hololissa.
Changes in relative warps 1 and 2 for lateral view. Details as in Fig. 2.
Changes in relative warps 1 and 2 for plastral view. Details as in Fig. 2.
Relationships between morphological variables. a) Vertebral 2 growth; b) vertebral 3 growth; c) anal notch growth; d) relationship between costals 1 and 2; e) relationship between vertebral 2 and 3; f) relationship between vertebral 3 and 4. Open circles: arnoldi; closed circles: dussumieri/gigantea; shaded circles: hololissa.
Morphological change in the 3 morphotypes. a) Hatchlings; b) < 11-cm straight carapace length; c) 21–30 cm CL; d) > 51 cm CL; e) adults. Photos of arnoldi in d and e show the same individual at 62- and 78 cm CL.
