Intraspecific morphological tooth variability and geographical distribution: Application to the Savi's vole, Microtus (Terricola) savii (Rodentia, Arvicolinae)

The Italian peninsula exhibits a very interesting and contrasted geographical pattern offering a good opportunity to study the relationship between morphological variability and environment. Savi's vole, Microtus (Terricola) savii, occupies the main part of the Italian peninsula from Ticino in the north southwards to Sicilia. The material studied comprises 55 local populations distributed over the whole Italian peninsula and characterising different types of environments and different types of conditions, e.g. natural or anthropic. Morphological variability is analysed using the first lower molar by taking 27 measures on the occlusal surface. The analysis has shown a clear morphological differentiation between north–central and south populations. Morphological tooth variability highlights two geographical scales. At a regional scale, morphological variability is clearly in relation to geography and consequently to the climate and environment. On the contrary, at a local scale, morphological variability can be due to the fragmentation of some particular areas leading to isolation of small populations characterised by different morphotypes.


Methods
Twenty-seven linear-distance measurements, noted 1 to 27, were taken on the occlusal surface of M 1 (Figure 2) according to Brunet-Lecomte (1988) and Laplana et al. (2000). The six first variables describe the global morphology of the tooth, and the other variables focus on the anterior part which is the most variable region. Measurements were taken with a Nikon Measuring Microscope MM-60 (preci-sion51 mm).

General morphometry
The general morphometry of the M 1 is studied by canonical discriminant analysis based on 27 measures taken on the occlusal surface of the tooth. Statistics are performed using Statistica version 6.1 (Statsoft, Tulsa, OK, USA).

Analysis of dental criteria
Five dental criteria were calculated (V is for variable): 1. Relative length of the anterior part of the M 1 : LRPA5(V6 2 V3)/V66100. 2. Tilt of the pitymyan rhombus: RP5V4 2 V3. 3. Closure of the anterior loop: BA5(V20 2 V18)/V21)6100. 4. Ratio of length to width: V6215V6/V21. 5. Ratio of relative width of pitymyan rhombus: V25215V25/V21. The total length of the M 1 (LT) and the five dental criteria were compared between groups of populations using analysis of variance, completed by a Bonferroni test or test of contrasts when necessary.

General morphometry of the first lower molar
A first canonical discriminant analysis was carried out between the 55 local populations of M. (T.) savii in order to verify the relationship between the geographical origin of populations and the distribution of centroids of these populations in the canonical plan 1-2. Plan 1-2 explains 34% of the variance, 21% and 13% for axes 1 and 2, respectively. The distribution of centroids in plan 1-2 ( Figure 3) confirms this relation except for four populations: 1. One population from Emilia Romagna (population 8) is located on the negative sides of both axes whereas the other Emilia Romagna populations are located on the positive side of axis 1 and the negative side of axis 2. 2. One population from Lazio (population 5) is very positively situated on axis 2. 3. One population from Piemonte (population 9) is negatively situated on the two axes in contrast to the other nine populations which are all located close to the centre of plan 1-2. 4. One population from Sicilia (population 1) is located on the positive side of axis 2 whereas the other three are on the negative side.
All the other populations can be clustered in 11 homogeneous geographical groups: group 1 (denoted AB) composed of all seven populations from Abruzzo, Toscana, and Umbria; group 2 (denoted C1) with four populations from Campania (populations 1, 3, 6, and 8); group 3 (denoted C2) with the four other populations from Campania (populations 2, 4, 5, and 7); group 4 (denoted ER), with five populations from Emilia Romagna (populations 1, 2, 5, 7, and 9); group 5 (denoted LA), all populations from Lazio except population 5; group 6 (denoted PI), all populations from Piemonte except population 9; group 7 (denoted LO), populations from Lombardia; group 8 (denoted VE), all populations from Veneto and three populations from Emilia Romagna (populations 3, 4, and 6); group 9 (denoted P1) with three populations from Puglia (populations 1, 2, and 3) and the population from Basilicata; group 10 (denoted P2) with the other two populations from Puglia (populations 4 and 5); and group 11 (denoted SI), all populations from Sicilia except population 1. The second canonical discriminant analysis was carried out between the 11 groups defined above. All Mahalanobis distances between groups are significant (P,0.0001 in all cases). Plan 1-2 explains 58% of the variance, 40% and 18% for axes 1 and 2, respectively. The distribution of centroids of the groups in plan 1-2 is given in Figure 4. Axis 1 separates the five groups of south Italy (Mezzogiorno) located on the positive side, and corresponding to Campania 1 (C1) and 2 (C2), Puglia 1 (P1) and 2 (P2), Sicilia (SI), from the six groups of north-central Italy located on the negative side and corresponding to Veneto (VE), Emilia Romagna (  the groups of Campania 2 (C2), Puglia 1 (P1), and Sicilia (SI), on the negative side, from the groups of Campania 1 (C1) and Puglia 2 (P2), on the positive side. The groups on the negative side of axis 2 are characterised by a total length of M 1 smaller than the groups on the positive side.

Analysis of the first lower molar criteria
The description and the comparison of the M 1 criteria are given in Table I Within south Italian populations, one criterion distinguishes the Campania 1 group from the Campania 2 group: the total length of the M 1 is greater in the Campania 1 group (2.654¡0.171 mm) than in the Campania 2 (2.484¡0.134). One criterion distinguishes the Puglia 1 group from the Puglia 2 group: the total length of the M 1 is greater in the Puglia 2 group (2.616¡0.119 mm) than in the Puglia 1 (2.510¡0.117). Four criteria characterise the Sicilia group, a very open anterior loop (35.71¡0.8%), a reduced anterior part of the M 1 (50.0¡1.6%), a pitymyan rhombus poorly tilted (20.0045¡0.041 mm) and a great ratio V25/V21 (0.175¡0.037).
Morever, population 5 from Lazio is characterised by a greater total length of the M 1 (t-test, P,0.0001) (2.728¡0.175 mm) in comparison to the other Lazio groups (2.592¡0.123 mm). The sample size of three other excluded populations (population 8 of Emilia Romagna, population 9 of Piemonte, and population 1 of Sicilia) from the analyses of groups is too small (n,10) to express conclusions.

Discussion and conclusion
The general dental morphology of arvicolines can be considered as adaptative in relation to their diet compared to other groups of rodents such as murines or cricetines. Indeed, this particular outline with alternated triangles combined with the hypsodonty of the teeth is well adapted to a diet with high content of silica which is a very abrasive. Within the arvicoline sub-family, tooth outline is extremely variable. Two main hypotheses can be made to explain this morphological variability. Firstly, morphological variability can result from selection pressure of climatic conditions. Secondly, this variation can be due to some isolation phenomena leading to very small populations which may be more sensitive to a bottleneck effect. This last hypothesis has been previously noted within the Microtus (Terricola) complex in relation to fragmentation of habitat (Contoli et al. 1992). Moreover, some authors have observed morphological variation between different generations (Krapp and Winking 1976).  (Fatio, 1905) (Brunet-Lecomte 1988, 1990. A similar observation can be noted for the relative length of the anterior part of M 1 , which is less developed in southern species, such as M. (T.) duodecimcostatus, or southern populations of M. (T.) savii.  Another fact highlighted by canonical discriminant analyses is the isolated position of the populations from Veneto and Emilia Romagna, two regions of the Padana plain. Two hypotheses can be proposed to explain this morphometrical divergence: (1) the important change of this region due to human activity; (2) the complex hydrography of this plain. Concerning the first hypothesis, since the beginning of historic times, the extensive forest was progressively transformed into an agricultural plain. At the start of this deforestation, M. (T.) savii colonized a new and a dislocated area. Firstly, this dislocated area perhaps has lead to the emergence of small populations characterized by a different morphotype of the M 1 . Secondly, the development of agricultural activity in the Padana plain has permitted the colonisation of a great part of the Veneto and the Emilia Romagna by M. (T.) savii populations characterised by this new morphotype. Regarding the second hypothesis, the complex hydrography of the Padana plain can be a favourable factor in the process of the isolation of populations, by playing a barrier role. These two hypotheses also can act together.
Size variation is noted within three geographical groups, Veneto-Emilia Romagna, Campania, and Puglia. In the Veneto-Emilia Romagna groups, five populations from Emila Romagna (Emilia Romagna group) are characterised by smaller size. The explanation can be probably found in a secondary role of the above factors. The fragmentation of habitat usually acts as a stress on individuals which are often smaller than in more optimal ecological conditions (Marchand et al. 2003). In the Campania groups, the small-size group (C2) is composed of four populations from inland localities characterised by cold winters (Grassi and Milone 1985), a climatic condition which can reduce the size development of specimens. On the contrary, the large-size group (C1) is composed of four populations from coastal localities with Mediterranean conditions. In the case of the Puglia groups, no climatic, hydrographic, or human activity factors seem to be able to explain the observed size variations. Moreover, the size difference is lesser within groups from Puglia (0.116 mm) than within groups from Campania (0.170 mm).
In our study, on a regional scale, morphological variability corresponds to the first hypopthesis and therefore is related to geography, and consequently to climate and environment. A clear separation of the general morphology of M 1 between populations from south Italy and populations from central and north Italy has been observed. The second hypothesis does not seem to act at the regional level.
On the contrary, on a local scale, the second hypothesis cannot be rejected. As observed in the Padana Plain, morphological variability can be due to the fragmentation of this particular area leading to the isolation of populations characterised by different morphotypes.
In conclusion, the same morphological pattern is observed at an intraspecific level, within Microtus (Terricola) savii between north-central and south populations in Italy, as at an interspecific level within the sub-genus Microtus (Terricola), between northern and southern species in western Europe. Two morphological criteria, the anterior loop (BA) and the relative length of the anterior part (LRPA) of the first lower molar, distinguish species living in north-western Europe such as M. The morphological variability of these two criteria seems to be in relation to the geography at both intra-and inter-specific levels.