Intraspecific variation in digit reduction in Testudo: the case of the Hermann’s tortoise

Phalangeal reduction is a common and widespread phenomenon among tortoises that has been associated with the adaptation to terrestrial life. While reduced manual digit 1 appears characteristic in almost all Testudo species, it is uncertain why the metacarpal I and distal carpal of the same digit are completely missing in some individuals of Hermann’s tortoise (Testudo hermanni hermanni). To clarify this issue, we investigated the number of manual claws in six populations of Hermann’s tortoise (one from the Ebro Delta in the Iberian Peninsula and five from Minorca Island), their age, sex, genetic lineage, and the substrate type that they inhabit. The number of claws was ascertained based on direct counts (n > 1500 individuals) and by X-rays (n = 32 individuals), obtaining three different phalangeal formulae: (1-2-2-2-1, D-2-2-2-1, 0-2-2-2-1). Thus, claw counts through both methodologies (direct count and X-ray) further confirm that the observed claws serve as a good proxy to assess the actual number of digits. Our results show no loss of phalanges, metacarpal and carpal bones in digit 1 associated with age, sex, or substrate, contrary to some previous authors who hypothesized a relationship between this loss and sexual dimorphism. Therefore, variations in the number of manual digits and the loss of metacarpal I and distal carpal in digit 1 in Hermann’s tortoise are related to population and genetic lineage. More detailed comparisons with other Testudo hermanni populations from elsewhere in Europe would be required to understand the evolutionary significance concerning the intrapopulation variability in the number of digits remaining.


Introduction
Turtles have successfully diversified into a variety of forms: adapted to aquatic, marine and terrestrial habitats, despite the fact that they have been structurally constrained by the presence of the shell.Thus, this great ecological diversity correlates with different locomotor patterns that are reflected, among others, in different limb anatomies (Delfino et al. 2010).Testudinids (Testudinidae or tortoises) are an exclusively terrestrial group of turtles with almost worldwide distribution (except Oceania and Antarctica) that comprise 18 genera and 65 species (Marmi and Luján, 2012;TTWG, 2017).In particular, the high-domed shell and unique limb anatomy are what mainly characterize tortoises, which are adapted for terrestrial life (Ernst et al. 2000).Testudinid limbs are thick and robust, serving as forelimbs that are rather flat and well bent, whereas hindlimbs are short and cylindrical (Ernst and Barbour, 1998).All digits in both manus and pes are short, and show short and rounded claws (Zug, 1971).
Currently, the genus Testudo includes the five extant Western Palaearctic species, and according to the more recent molecular phylogenies (Fritz and Bininda-Emonds, 2007;Vasilyev et al. 2014) and phylogenetic analysis of morphological traits (Luján et al. 2016), three subgenera are recognized (Agrionemys, Chersine and Testudo).
Specifically, Chersine subgenus is represented by a single extant species, the Hermann's tortoise (Testudo hermanni), which is a medium-sized species ranging in average carapace length between 13-21 cm (Cheylan, 2001;Bertolero et al. 2011a).Hermann's tortoise populations are distributed throughout most of the European Mediterranean region, usually below 500 m a.s.l.Nowadays, two subspecies of Hermann's tortoise are distinguished: T. h.hermanni in Western Europe, which is in strong decline and has very restricted distributions; and T. h.boettgeri in Eastern Europe, which seems to be more stable and displays rather continuous distribution; however, a recent assessment has shown a worse conservation situation (see Fritz et al. 2006a;Bertolero et al. 2011a;Nikolić et al. 2018).Moreover, human impact has been reducing the favorable habitats for this taxon, especially agricultural practices and urbanization (Cheylan, 1981(Cheylan, , 2001;;Vetter, 2006), human-related forest fires and poaching (Cheylan, 2004;Couturier et al. 2011;Santos and Cheylan, 2013) and illegal animal trade (Bertolero et al. 2011a).
Digital reduction (i.e. when species show fewer digits than their related species) is a common characteristic that often occurred throughout the evolution of tetrapods (e.g.Shapiro et al. 2003;McHorses et al. 2017).Although in Hermann's tortoise, either five or four digits can occur in both forelimbs (Fig. 1 and Fig. 2a-d) and hindlimbs (Bertolero, 2010;2014), in the former the digit that tend to reduce is the first and in the latter is the fifth (see Crumly and Sánchez-Villagra, 2004).We defined here the condition of a "full digit" when the members of the genus Testudo retains two phalanges in manus (being the distal phalanx the claw), which always occurs in the digits two, three and four (Fig. 2a).Thereby, a digit one or five is defined as a "reduced digit" when this digit lacks the proximal phalanx but still retain the distal one (Fig. 2a).The latter occurs in all Testudo spp. with the exception of T. marginata that retains two phalanges in digit five and thus has a "full digit" (see Crumly and Sánchez-Villagra, 2004;Hitschfeld et al. 2008).Similarly, digit one in at least three of five Testudo species (T.graeca, T. hermanni and T. horsfieldii) is reduced, but also it may be: a "lost digit", when the digit is completely missing (i.e. both phalanges are missing: Fig. 2b); a "lost digit and metacarpal I" which implies the additional loss of the metacarpal I (Fig. 2c); and a "lost digit, metacarpal I and distal carpal" when the distal carpal is also missing (Fig. 2d).
The first studies of the ontogeny of turtle limbs started in the late 18th century, but the examinations of the phalangeal formula in testudinids were restricted to anecdotal reports or were merely concerning a single species (e.g.Baur, 1888;Boulenger, 1889).Auffenberg (1966) was the first author to provide a functional interpretation of the adult carpus of several tortoise species.Then, the phalangeal formulae of the pes were analyzed in testudinids using a phylogenetic approach, although only four species of tortoises were examined by Zug (1971).He concluded that tortoises always have four clawed full digits in each manus; however, a reduced or lost digit occurs sometimes.Shortly after, detailed manus illustrations of both extant and fossil gopher tortoises (genus Gopherus) were provided by Bramble (1982).The first mention of the number of digits in Testudo hermanni is found in Highfield (1988), who points out that all individuals (both males and females) of T. h.boettgeri that he found in Bulgaria showed four full digits in their manus.However, preliminary results published by Eendebak (2001) claimed that the number of digits on T. h.boettgeri was strongly determined by the maternal genetic characteristics (i.e.offspring of mothers with four digits on their manus were four times more likely to have four digits themselves).Although Eendebak (2001) announced the publication of the complete study clarifying the later results on a genetic basis, this study has not been published so far; therefore, these explanations are still missing.There are remarkable results in a study focused on Hermann's tortoise from Corsica (T.h.hermanni) conducted by Vetter (2006), as he found sex differences related to the number of digits (44.2% of males and 59.4% of females possessed five digits), although the sample size was not provided.Of all the sampled individuals with five digits, only 4.2% of males and 12.5% of females showed a claw well developed.
Moreover, adult individuals with four digits often present the fifth digit also reduced, which is clearly recognized because the latter does not possess a claw (distal phalanx).
However, Vetter (2006) did not provided any hypothesis to explain these differences between sexes regarding to variations of digit one or five.In a study of carpal morphology and phalangeal formulae, Crumly and Sánchez-Villagra (2004) examined in testudinids the ontogenetic stage of the individuals, as well as the phylogenetic context.Fritz et al. (2006b) proposed that the phalangeal reduction present in terrestrial and semiterrestrial geoemydids (Cuora) and emydids (Terrapene) is related with their mode of life, mainly as a consequence of adapting to walk on land (see further Minx, 1992;Ludwig et al. 2007).In particular, in some highly aquatic geoemydids (e.g. Malayemys, Morenia, among other genera), there is no loss of phalanges; they even display an additional phalanx in the digit five of the manus and pes, which apparently seems more favorable for swimming (Ludwig et al. 2007).Finally, a unique study that was focused on carpal morphology and phalangeal reduction both in Testudo hermanni and T. horsfieldii (Hitschfeld et al. 2008) found more variation in phalangeal formula than Crumly and Sánchez-Villagra (2004).This is easily explained by the higher sample (29 individuals) studied by the former in comparison to just three specimens analyzed by the latter.Therefore, the loss of digit one (both phalanges), metacarpal I and distal carpal, together with the extensive fusion of carpal elements in T. horsfieldii, could be linked with its burrowing mode of life (Hitschfeld et al. 2008).
The main aim of this study is to characterize variation in both the number of digits and phalangeal formula in the manus and to test if this variation is related to other variables (age, sex, habitat among other), using a large sample of Hermann's tortoise obtained in wild populations.First, we investigated whether the number of visible digits corresponds with the phalangeal formula; X-rays were used because digit five is often atrophied, and/or digit one may be partially preserved as stated above, both inside the manus (see Vetter, 2006).Second, we tested whether the number of digits in the manus differs between the sexes, as other authors have claimed (Eendebak, 2001;Vetter, 2006).
Third, given that digit one may be atrophied, and it is more recognizable in individuals larger than 45 mm (approximately from one year of age), we tested whether the number of digits differs between age groups (juveniles vs. sub-adults and adults).Finally, we tested whether the number of digits in the manus differs between populations, genetic lineages, or the types of substrates that Hermann's tortoises inhabit.
Therefore, the most important contribution of this paper is that the number of individuals is significantly larger than that of previous works (here, 1,669 individuals compared to three in Crumly and Sánchez-Villagra, 2004;or 29 in Hitschfeld et al. 2008), allowing us to carry out reliable statistical analyzes.Secondly, our sample is very representative, including wild individuals of various ages, sexes and populations (i.e.six).This extraordinary amount of data is very important since the previous works used only museum specimens.Therefore, in most cases, it was nearly impossible to assess with confidence their subspecies, population, or sometimes even their age and sex.

Hermann's tortoise populations
This study was conducted on six wild populations of Hermann's tortoise, one in the Iberian Peninsula and five on Minorca Island (Tables 1 and 2).These populations inhabit a bioclimatic zone of evergreen oak trees, characterized by long dry summers, and wet warm winters (Cheylan, 2001).The mainland Ebro Delta (DE) population is situated in the Punta de la Banya Reserve (Delta de l'Ebre Natural Park), a flat sandy salt marsh peninsula of 2,514 ha, which is formed by dunes with small slopes fixed by psammophile and halophile vegetation (Bertolero, 2002;Bertolero et al. 2007).On Minorca Island, Herman's tortoise occupies most of the island with a patchy distribution, mainly concentrated on the coast around the island, except in the most arid areas (Bertolero and Pretus, 2012).We studied five populations near the towns of Ferreries (one population: NF), es Mercadal (three populations: RSM, DSM and DNM) and Alaior (one population: RSA).The exact locations are not specified owing to poaching concerns (Table 2).The NF population inhabits a hill formed by a consolidated sand dune with holm oak (Quercus ilex) and pine (Pinus halepensis) forest.
The DNM and DSM populations inhabit dune vegetated systems.The RSM and RSA populations are located in calcareous ravines.The proportion of juveniles and sub-adults relative to adults varies between the populations and years, but all show high tortoise densities (>20 tortoises/ha in Minorca Island and 7 tortoises/ha at the Ebro Delta; Bertolero et al. 2018, AB unpublished data).Juveniles included in this study had a minimum shell length of 45 mm, to avoid any problems with visual digit counting due to the small size of the individual (hatchlings were not included because in some individuals, the claw of the first digit is very small, and needs careful inspection, which is difficult to see during field work).
These populations are from two distinct genetic lineages (Zenboudji et al. 2016)

Field methods and data collection
The number of claws of both forelimbs have been recorded since 2000 in the Ebro Delta population and from 2003 to 2006 in the Minorca Island populations, within the framework of a long-term research program of these populations (Bertolero, 2002;Bertolero et al. 2018).Once captured, all tortoises were sexed (only sub-adults and adults), age-class classified according to the type of growth rings (juveniles of unknown sex, sub-adult and adult; Bertolero et al. 2011b) and marked individually with notches on the carapace.Additionally, from a previous study where females from the Ebro Delta were X-rayed to assess their reproductive status (Bertolero et al. 2007), radiographs of 32 individuals were carefully reviewed to assess their phalangeal formula (Table 1) to test if the number of claws could be a good proxy of the number of digits.Since the number of claws is equivalent to the number of digits (see results section) we use the term number of digits throughout the text.The coding of phalangeal formulae follows Crumly and Sánchez-Villagra (2004), where, for example, 1-2-2-2-1 indicates that one phalanx is present in the digits one and five, whereas two phalanges occur in digits two, three and four.The fieldwork conducted complies with Spanish laws.Handling and sampling of Herman's tortoises were authorized by the regional governments of Catalonia and the Balearic Island (Generalitat de Catalunya permit SF/058 and Govern de les Illes Balears permit CEP 08/2017).

Statistical analysis
Firstly, to test whether the number of digits that are externally displayed and the phalangeal formulae seen on the X-rays (Table 1) are equivalent, a paired Sign rank test was calculated with a total of 32 females from both right and left forelimbs (i.e.N = 64).
Secondly, to determine whether the number of digits depended on sex, age, population, genetic lineage or substrate type, three generalized linear mixed models (GLMM) were constructed with the number of the observed digit forelimbs (coded as 0 for four digits and coded as 1 for five digits) as the dependent variable.To investigate the effect of sex and population (model 1), this first model was constructed with sex (male vs. female) and population (DE, NF, RSM, DSM, DNM and RSA) as fixed factors.To investigate the effect of age and population (model 2), a second model was constructed with age-class (juveniles vs. sub-adult + adult) and population as fixed factors.These two models were built separately since sex data was not available for juveniles, and thus not allowing to include the effect of sex, age and population in a single model.Thus, the sample size of model 1 was lower (2,264 forelimbs and 1,132 individuals) than model 2 (3,338 forelimbs and 1,669 individuals).To investigate the effect of substrate types and genetic lineages, a third model (model 3) was conducted with the same dataset used in model 2 but with substrate (sand or calcareous and clay) and genetic lineage (mixed, Minorca 1 and Minorca 2) as fixed factors (3,338 forelimbs and 1,669 individuals) instead of the population factor.This third model was built since population (included in models 1 and 2) and genetic lineage are highly correlated and cannot be introduced in the same model.For all models, tortoise identity was included as a random factor (to count the right and left forelimbs).Tukey's HSD test was used for post-hoc comparisons.The GLMMs were fitted with the maximum likelihood method, using the logit link function and with binomial error distribution.Analyses were carried out with the R 3.2.2software (R Development Core Team, 2015), with BSDA (Arnholt and Evans 2017), lme4 (Bates et al. 2015) and multcomp packages (Hothorn et al. 2008).

RESULTS
We studied 3,338 forelimbs belonging in 1,669 individuals of six populations (Tables 1   and 2).

Does the number of digits displayed externally correspond to the phalangeal formulae?
The number of digits that are externally displayed and the phalangeal formulae seen on the X-rays did not significantly differ (S = 0; n = 64, P = 1) and thus the number of claws could be a good proxy of the number of digits.In the 32 females studied by Xrays, we observed that all the individuals with five or four visual digits corresponded with what we observed in the radiographies (Tables 1 and 2).Indeed, individuals with five visual digits had the digits 2, 3 and 4 full (two phalanges:) and digits 1 and 5 reduced (one phalanx = claw) and therefore displaying the following phalangeal formula (1-2-2-2-1: Fig. 2a).However, individuals with four visual digits had entirely lost digit one (both phalanges) but also the metacarpal I and distal carpal (0-2-2-2-1: Fig. 2d), with one exception only (right manus of the individual 1235: D-2-2-2-1).In this way, no individual with four digits of our sample showed a manus with the metacarpal I and distal carpal (M-2-2-2-1: Fig. 2b).We found only one specimen with four visual digits but presenting five digits in the X-rays.In this case, digit one was in an unnatural anatomical position (entirely blended in dorsal direction inside the manus) and therefore it was a pathological individual.

Does the number of digits depend on sex, age or population?
Although males tended to have a lower number of digits than females (61,5% of males and 65,6% of females had five phalanges), the gender factor was not significant (Table 3).As expected, the number of digits was not significantly different between juveniles and adults (61,3% of juvenile and 63,5% of adults had five phalanges; Table 4).Finally, the number of digits significantly differed between populations (Tukey's HSD tests, DNM vs. RSM: p = 0.002; DNM vs. RSA: p = 0.002; Fig. 3, Tables 3 and 4).

Does the number of digits depend on substrate or genetic lineage?
The number of digits was not significantly different between substrates (see Table 5).
However, the number of digits significantly differed between genetic lineages (Table 5, Fig. 4 ).The mixed group showed the lowest proportion of individuals with five digits, Minorca 1 the lineage with a higher proportion with five digits, and Minorca 2 the one with an average proportion between Minorca 1 and the mixed group (Table 5, Fig. 4: Tukey's HSD tests, Minorca 1 vs. Mixed: p<0.001;Minorca 2 vs. Mixed: p = 0.06; Minorca 1 vs. Minorca 2: p = 0.002).

Discussion
According to the evolutionary history of reptiles, the number of digits for a species seems to be a strongly stable feature, with the only known exceptions present in the lizard genus Hemiergis (Choquenot and Greer, 1989) and in both tortoises of the Testudo and Kinixys genera (Crumly and Sánchez-Villagra, 2004;Hitschfeld et al. 2008).In particular, the number of digits in the lizard H. peronii differs among populations of this species (Choquenot and Greer, 1989).However, for tortoises no detailed information is available in the literature regarding the population origin of the samples used in previous studies (Crumly and Sánchez-Villagra, 2004;Vetter, 2006;Hitschfeld et al. 2008).Thus, the present study, which focuses on T. h.hermanni, is the first to report intrapopulation variability in the number of visible digits in a reptilian species.
Several authors have reported that T. hermanni can display four or five digits in the forelimbs (Highfield, 1988;Eendebak, 2001;Vetter, 2006;Hitschfeld et al. 2008;Bertolero, 2014).Nevertheless, so far, the extent of this individual variation among and within populations has not been evaluated rigorously or in detail.Our results denote that the proportion of individuals with four visible digits varies according to populations and genetic lineage only (see below other relevant variables considered such as age, sex or type of substrate).
Digital reduction can be achieved by two ontogenetic processes: complete loss of the digit (i.e.loss of all phalanges and related metacarpal and carpal bones); or a severe reduction (loss of all phalanges; Shapiro et al. 2007).In any case, both cases externally produce the same result in the Hermann's tortoise: showing four digits in the forelimbs.
Given that the number of digits counted externally in the forelimbs of T. h.hermanni fits well with their real number of digits, the number of observed claws can be used as a good proxy of the real number of digits.Overall, our results confirm that 37% (n = 3,338 examined extremities) of T. h.hermanni individuals have only four digits in their forelimbs, but this percentage varies greatly among populations (range: 28.1% -49.5%).
Even though our sample size for the phalangeal formulae is much more limited and for one population only, this reduction may merely be due to the entire loss of digit one (0-2-2-2-1: 31.3%, n = 64 extremities), or to the loss of the phalanges and/or the metacarpal I (D-2-2-2-1: 1.6%, n = 64 extremities).Apart from the two later variations, Hitschfeld et al. (2008) further reported a third case produced by the loss of the phalanges (M-2-2-2-1) for this subspecies but retaining the metacarpal I and distal carpal; however, it was absent in our population (DE).
Concerning digit variation, we did not find differences between ages, as expected, since digits and claws are fully developed when tortoises hatch.To date, no other works analyze age differences in the number of digits in extant and extinct Testudines.
Similarly, we did not find differences between the sexes in the number of external digits, contrary to Vetter (2006) for Corsican tortoises (T.h.h.).Unfortunately, Vetter (2006) did not provide sample size or a statistical test; consequently, it is not possible to know if his observed frequencies are significantly different.The author also reported that only 44.2% of males showed five claws; however, our results denote that the five digits were present externally in greater proportion, although non significantly different from the females, in all our studied populations (61.5% of males with five digits, all populations pooled).Furthermore, Eendebak (2001) indicated that the number of external digits has a clearly genetic component, since he observed in their enclosures that females with four digits produced 80% of hatchlings that also had four digits.
Despite omitting to provide information about the sires and no further data were published to quantify his observations, our results indicate that a genetic component could be operating, since we found that lineage and population factors were the only significant ones in our analyses.However, we are not able to clearly disentangle both factors because some pair-wise comparisons between the populations of the two pure lineages did not show significant differences (e.g.no significant differences between populations NF from lineage Minorca 1 and RSM from lineage Minorca 2: Table 4).
Finally, the population factor does not encompass genetic characteristics only, but further ecological, environmental characteristics (e.g.substrate, but this was not relevant to explain the number of digits) and genetic drift.
Interestingly, the population with a high percentage of forelimbs with four external digits (49.5% at RSA) was also the population that showed a mixed genetic lineage (Massana et al. unpublished data).In contrast, populations of both pure lineages showed a lower frequency of four digits (30.2% for lineage Minorca 1 and 39.8% for lineage Minorca 2).Thereby, the RSA population does not on average show an intermediate percentage of forelimbs with four external digits between the two pure lineages.However, it is known that hybridization can produce phenotypic diversity (Grant and Grant, 1994), and hybrids can show phenotypes that exceed those of either parental line (i.e.transgressive segregation; Rieseberg et al. 2003).Therefore, the RSA high percentage of forelimbs with phalanx reduction could be a case of phenotypic diversity resulting from the hybridization of the two well differentiated genetic lineages (Zenboudji et al. 2016).
It has been suggested that the reduction in the number of digits in some groups of turtles (geoemydids and emydids) is related to the shift from aquatic to terrestrial locomotion (Zug, 1971), since the reduction of phalanges has only been observed in terrestrial or semi-terrestrial testudinoids (see Minx, 1992;Fritz et al. 2006b;Ludwig et al. 2007;Hitschfeld et al. 2008).Nevertheless, Hitschfeld et al. (2008) argued that the burrowing activity is not necessarily linked to the loss of digit one, because the other species of Testudo can also lose this digit, but they do not show significant digging behavior.Despite this, it is known that the digging behavior in Hermann's tortoise can be important in dune environments, as in the Ebro Delta and Minorca populations, where tortoises dig cavities and short tunnels to rest, hibernate, or shelter from extreme weather conditions (A.B.personal observations).Our results support the Hitschfeld's hypothesis that the loss of digit one is not related to digging behavior since no differences between populations inhabiting in different substrates were found (sandy soil vs. calcareous and clay soils).According to Hitschfeld et al. (2008), the differences in carpal fusion patterns between T. horsfieldii and T. hermanni are actually related with their different modes of life.A more rigid hand skeleton could explain why the carpal and metacarpal bones fuse more extensively in T. horsfieldii (useful for excavating burrows).Conversely, fused bones in T. hermanni occur in the part of the hand that is exposed to the greatest mechanical strain during walking; therefore, congruent with species without extensive digging behavior.
Several studies have corroborated that many genes are involved in digit development (e.g.Hedgehog, BmP, Fgf and Hox families: Merino et al. 1998;Litingtung et al. 2002;Sheth et al. 2012).Within these genes and regulatory factors, the Sonic hedgehog (SHH) gene seems to be most critical concerning the determination of the number of digits (see Litingtung et al. 2002;te Welscher et al. 2002).Even though the expression of the SHH varies between different lizard species, there is no doubt that it is more prolonged during ontogeny in lizards with more digits (Shapiro et al. 2003).The differences found in this study between both genetic lineages of tortoises and their hybrid population Moreover, the evolutionary significance of this intraspecific and intrapopulation variability in the number of digits remains an open question that merits further research.

Conclusions
Our study firstly reports on the intrapopulation variability in the number of visible digits in Testudo hermanni hermanni, the first tortoise species that this study has observed.
Moreover, this work also shows that the number of observed claws can be used as a good proxy of the real number of digits.Contrary to previous studies, no digit variation could be attributed to sex.Neither age nor the substrate that the Hermann's tortoises inhabit could explain the digit variation.On the contrary, phalangeal reduction varied according to the studied populations and genetic lineages.
Our knowledge of T. h.hermanni individuals in their natural environment prove that digging behavior can be significant in the Ebro Delta and Minorca populations, and more specifically in dune environments.However, Hermann's tortoise populations in sandy substrates showed a lower percentage, although not statistically significant, of individuals with four digits than populations in calcareous or clay soils, therefore supporting the hypothesis that the loss of digit one is not related to digging behavior.
In particular, the RSA population from Minorca was the only one that displayed a mixed genetic lineage and the highest percentage of forelimbs with phalanx reduction.
This phenotypic diversity could be explained through a hybridization process of two well-differentiated genetic lineages.Apparently, the differences obtained between both genetic lineages of tortoises and their hybrid population (RSA) seems to indicate differences in the gene expression that regulates digit formation.However, gene expression involved in the development of the digits needs further investigation in order to better explain the intraspecific and intrapopulation variability observed in T. h.hermanni.

Table 1 Comparison of the number of visible digits in adult manus individuals of
Testudo hermanni hermanni and the number of digits in the radiographs.The phalangeal formula is a sequence of numbers and/or letters.The formula 1-2-2-2-1 indicates that the digit one and five possess only a single phalanx, and digits 2-4 possess two phalanges; in D-2-2-2-1, digit one possesses only the distal carpal (the phalanx and metacarpal I have been lost); and in 0-2-2-2-1, no related bone of digit one is present (i.e.phalanges, metacarpal I and distal carpal).

Number of digits on radiographies
Phalangeal formula : 1, populations NF and DNM are of continental lineage, which includes populations from Spain (Albera), France (Var) and Italy (called Minorca 1); 2, populations DE, RSM and DSM are of insular lineage found only on Minorca and the Ebro Delta (called Minorca 2).The RSA population is an interbreed population of both lineages embedded in the area of the genetic group Minorca 2 (Massana et al. unpublished data).In the latter population, 1,615 Minorcan tortoises of unknown genetic lineage were released by the NGO GOB-Menorca for conservation purposes between 1995 and 2004 (GOB-Menorca com.pers.).

(
RSA) in relation to individuals with only four digits may indicate gene expression differences that regulate digit formation.Only new research focused on the expression of genes involved in the development of the digits could clarify the causes of the variability observed in T. h.hermanni, which is beyond the scope of the present work.

Fig. 1
Fig. 1 External manus morphology in Testudo hermanni hermanni with four (a, b) or five digits (c, d): left adult manus in ventral (a) and dorsal (b) views; adult (c) and juvenile (d) left manus in ventral views.The number of the digit is denoted by numerals from one to five.

Fig. 3
Fig. 3 Visible proportions of manus with five digits between the six studied populations (N = 3,338 forelimbs belonging to 1,669 individuals; 682 in RSM; 380 in RSA; 432 in NF; 836 in DE; 166 in DSM; and 842 in DNM).

Table 2
Number of tortoises analyzed by the population, including ecological and genetic characteristics of each population.

Table 3
Generalized linear mixed model showing the effects of sex and population on the number of digits in right and left manus (coded as 0 for four digits and as 1 for five digits) in Testudo hermanni hermanni.N = 2,264 forelimbs belonging to 1,132 individuals, including 1,196 males and 1,068 females.Abbreviations: Sex coded as 0 for females and 1 for males; pop is short for population.

Table 4
Generalized linear mixed model showing the effects of age and population on the number of digits in right and left manus (coded as 0 for four digits and as 1 for five digits) in Testudo hermanni hermanni.N = 3,338 forelimbs belonging to 1,669 individuals; 1,074 juveniles and 2,264 adults.

Table 5
Generalized linear mixed model showing the effects of substrate and genetic lineage on the number of digits in right and left manus (coded as 0 for four digits and as 1 for five digits) in Testudo hermanni hermanni.N = 3,338 forelimbs.