INTRODUCTION

Madagascar is famous for its organismal diversity and high degree of endemism. Among the most speciose vertebrate clades are the mantellines. This lineage, including Mantella and Mantidactylus (Glaw and Vences, 1994; Glaw et al., 1998) has been considered as subfamily Mantellinae of the cosmopolitan frog family Ranidae (Blommers-Schlösser, 1993) or as separate family Mantellidae (Dubois, 1992). Mantellines are characterized by a specialized mating behavior involving absence of a strong mating amplexus (Blommers-Schlösser, 1993; Glaw et al., 1998). They are a monophyletic group as supported by morphological and molecular studies (Glaw et al., 1998; Richards and Moore, 1998; Richards et al., 2000).

Currently more than 65 nominal species of Mantidactylus are known (Glaw and Vences, 1999, 2000; Glaw et al., 2000). They are highly diverse, ranging from large and semiaquatic to minute and scansorial, and from species with fairly generalized tadpoles to highly specialized species with direct development (Blommers-Schlösser and Blanc, 1991; Glaw and Vences, 1994). Molecular data demonstrated that Mantidactylus is not monophyletic: the molecular study of Richards et al. (2000) supported relationships of Mantella to species of Mantidactylus belonging to the subgenera Blommersia, Guibemantis, and Pandanusicola.

In contrast, the genus Mantella is a well defined monophyletic unit (Vences et al., 1998a, b) containing about 17 morphologically poorly differentiated species (Vences et al., 1999). Mantella are attractive, small diurnal frogs which accumulate skin alkaloids, most probably by uptaking arthropod prey (Daly et al., 1996; 1997), and characterized by aposematic coloration. Hypotheses of intrageneric relationships have been proposed based on a number of different character sets (Pintak et al., 1998; Vences et al., 1998b, c).

The chromosomes of mantelline frogs have been described thus far mainly by Blommers-Schlösser (1978). She described general chromosome morphology in 24 species of Mantidactylus and Mantella aurantiaca, M. betsileo, and M. haraldmeieri. In addition, Pintak et al. (1998) provided data on Mantella aurantiaca, M. baroni, M. betsileo, M. expectata, M. haraldmeieri, M. laevigata, and M. viridis. The present study complements these earlier contributions by adding new species of Mantella to the data set. Besides general chromosome morphology, we also studied the distribution and composition of heterochromatin and the location of nucleolus organizer regions (NORs). Our goals were (1) to test recent Mantella classification by searching for chromosomal differences between closely related taxa, and (2) to obtain a new set of data to test hypotheses of Mantella phylogeny.

MATERIALS AND METHODS

We examined a total of 36 specimens of Mantella belonging to 11 different species (see appendix for voucher specimens), which belong to the M. aurantiaca group (M. aurantiaca), M. betsileo group (M. betsileo, M. cf. betsileo, M. expectata, M. viridis), M. cowani group (M. baroni, M. cowani, M. nigricans), M. laevigata group (M. laevigata) and M. madagascariensis group (M. madagascariensis, M. pulchra) as defined by Vences et al. (1999). Six specimens of three species of Mantidactylus were used for outgroup comparisons. Voucher specimens have been deposited at the Museo Regionale di Scienze Naturali, Torino (MRSN) and the Zoologisches Forschungsinstitut und Museum Alexander Koenig, Bonn (ZFMK). Each specimen was treated with 0.01 ml per g body weight of a 0.5 mg/ml colchicine solution. Four hr later, animals were sacrificed using 0.1% tricaine metasulfonate (MS-222). Intestines, spleens, lungs, and gonads were removed and incubated for 30 min in a 0.7% sodium citrate solution. Chromosomes were obtained by air drying and scraping as described by Olmo et al. (1986). Besides conventional staining (5% Giemsa at pH 7), the following techniques were applied: (1) AgNO₃-banding of NORs following Howell and Black (1980); (2) staining with the C+G specific fluorochrome chromomycin A₃ (CMA₃) according to Sahar and Latt (1980), with a reduced exposure (a few seconds) to the non fluorescent dye, methyl green; (3) the A+T specific fluorochrome, DPI/distamycin (DA), and CMA₃/DA (DAPI: Schweizer 1976); (4) C-banding as described by Sumner (1972), incubating the slides for 5 min at 45°C in Ba(OH)₂; (5) in situ digestion with Alu I endonucleases (compare Mezzanotte et al. 1983). Suitable results were achieved by staining, either separately or sequentially, with CMA₃ and DAPI after hydrolysis in Ba(OH)₂ or digestion with Alu I.

Metaphase chromosomes were stained with Giemsa; AgNO₃ and C-banding/Giemsa were viewed on a Zeiss PHOM III phase contrast microscope, whereas the fluorochrome-stained metaphases (CMA₃ and DAPI) were viewed on a Leitz epifluorescent microscope. Of each taxon, at least four Giemsa-stained metaphases and two metaphases stained with each of the banding methods used were studied. Images were digitized using a scanner. Karyotypes were constructed using Adobe Photoshop 3.0. Measurements to determine relative chromo-some length (rl; percentage ratio between the length of each chromosome to the total length of all the chromosomes) and centromer index (ci; ratio between short arm and total length of each chromosome) were carried out using the digitized images. In the following accounts, DAPI-positive bands are referred to simply as “DAPI+”, CMA₃-positive bands as “CMA+”.

Cladistic analysis was carried out using PAUP*, version 4 beta (Swofford, 1998). We calculated Maximum parsimony and Neighbor-joining (NJ) trees based on total character differences, and tested the trees by running 2000 bootstrap replicates. An interspecific principal component analysis (PCA) of mean rl and ci values was performed with SPSS for Windows, version 6.1.2.

RESULTS

All species of Mantella and Mantidactylus examined had a karyotype of 2n=26 chromosomes with five larger and eight smaller chromosome pairs. All chromosomes were metacentric except for the submetacentric third and sixth pair (Table 1 and 2; Fig. 1). The only deviations from this pattern were exhibited by M. betsileo which showed a metacentric sixth pair, and M. nigricans in which the 11th pair was acrocentric (Table 1 and 2). All Mantella species as well as Mantidactylus grandisonae had a secondary constriction near the centromere on the short arm of the 2nd chromosome pair; this secondary constriction was selectively stained by the AgNO₃ and the CMA₃/MG staining, indicating that it corresponds to the NOR. In Mantidactylus cf. punctatus and M. bicalcaratus, the NOR was interstitial on the short arm of the 1st chromosome pair (Fig. 1).

Table 1

Relative chromosome lengths (rl) of chromosomes 1–13 in the species studied. Data are mean values with standard deviations.

Table 2

Centromer indices (ci) of chromosomes 1–13 in the species studied. Data are mean values with standard deviations.

Fig. 1

Giemsa stained karyotype of Mantella taxa studied: M. aurantiaca (A), M. baroni (B), M. betsileo (C), M. cf. betsileo (D), M. cowani (E), M. expectata (F), M. laevigata (G), M. madagascariensis (H), M. nigricans (I), M. pulchra (J), M. viridis (K). The AgNO₃ stained NOR bearing 2nd chromosome pair is also reported.

Eight principal component factors with an Eigenvalue >10 were obtained by PCA. The first and second principal component factors together explained 56.8% of the observed total variation. The first factor was mainly influenced by relative chromosome lengths: although the highest principal component loading was that of the centromer index of chromosome 2, the four next highest loadings were those of the relative lengths of chromosomes 2, 3, 10, and 13. The second factor, on the other hand, was mainly influenced by the centromer indices; the five highest loadings were those of the ci values of the chromosomes 9, 4, 13, 8, and 5. In the corresponding scatterplot, the three species of Mantidactylus appeared widely separated from all Mantella species (Fig. 2). Among Mantella, the analysis clustered the species of the M. betsileo group away from the remaining species (along factor 1), while the other Mantella species did not show a clustering pattern consistent with their attribution to species groups.

Fig. 2

Scatterplot of first and second factors of a PCA of morphometric chromosome data as given in Table 1 and 2 (ci and rl).

Fig. 3

C-banded karyotype of the 11 Mantella taxa studied: M. aurantiaca (A), M. baroni (B), M. betsileo (C), M. cf. betsileo (D), M. cowani (E), M. expectata (F), M. laevigata (G), M. madagascariensis (H), M. nigricans (I), M. pulchra (J), M. viridis (K).

Fig. 4

C-banded matephase plates of M. pulchra (a and a′), M. aurantiaca (b and b′), M. madagascariensis (c and c′), M. nigricans (d and d′), M. baroni (e and e′), M. cowani (f and f′), M. betsileo (g and g′); M. cf. betsileo (h and h′), M. expectata (i and i′), M. viridis (j and j′) and M. laevigata (k and k′) successively stained with Chromomycin A₃ (simple cases) and DAPI (marked cases).

Fig. 5

AgNO₃ stained metaphase plates (left row), and the same metaphase plates successively stained with Chromomycin A₃ (medium row) and DAPI (right row) of Mantidactylus bicalcaratus (a, b, and c), M. punctatus (d, e, and f) and M. grandisonae (g, h, and i). The arrows point to NORs.

Heterochromatin staining resulted in a wide array of band distribution and staining patterns. Beside the centromeric, telomeric and peritelomeric heterochromatin bands summarized in Table 3, paracentromeric bands were found in M. expectata on chromosomes 1, 3 and 5, and in M. laevigata on chromosomes 1 and 3. These consisted of DAPI-positive bands bordered by CMA₃-positive bands.

Table 3

Distribution of centromeric, telomeric, and peritelomeric heterochromatin in Mantella and Mantidactylus species studied. For each species, we listed numbers of chromosomes on which a respective band was observed, followed by C and/or D if the bands stained positively with MA₃ and/or DAPI, respectively. Chromosome numbers in brackets refer to faint staining. The telomeric heterochromatin of M. betsileo was located on the long arm of the first chromosome and the short arm of the second chromosome; the peritelomeric band of M. laevigata was located on the short arm of the second chromosome.

In order to use the karyological data to assess phylogenetic relationships, we defined 15 characters based on the results presented above: Maximum parsimony analysis of the data summarized in Table 4 (using the three Mantidactylus species as outgroups; all characters unordered) failed to resolve relationships among Mantella species. A strict consensus of the most parsimonious cladograms resulted in a basal polytomy in which Mantidactylus grandisonae clustered together with the ingroup species. In the Neighbor-joining analysis (Fig. 6), bootstrap support >50% was found for the following groupings: a clade containing Mantella laevigata and M. expectata, the sister group relationship of this clade to M. viridis, and a clade containing M. pulchra and M. madagascariensis.

Table 4

Karyological character states used for phylogenetic analysis: (1) Configuration of 6th chromosome. 0 submetacentric; 1 metacentric. (2). Configuration of 11th chromosome. 0 submetacentric; 1 acrocentric; 2 metacentric. (3) Intensity of centromeric C-bands. 0 distinct; 1 faint or very faint. (4) Presence of centromeric C-bands. 0 present on all chromosomes; 1 present on all chromosomes except 10th and 11th; 2 present on chromosomes 1–5; 3 present on chromosomes 1-3; 4 present on chromosomes 1, 3, 11; 5 present on chromosomes 6, 12. (5) DAPI-staining of centromeric C-bands. 0 DAPI-negative; 1 DAPI-positive. (6) CMA-staining of centromeric C-bands. 0 CMA-negative; 1 CMA-positive. (7) Presence of telomeric C-bands. 0 absent; 1 scattered on only a few (up to three) chromosomes; 2 mainly present on chromosomes 1–5; 3 mainly present on chromosomes 6–8; 4 present on all or almost all chromosomes. (8) DAPI-staining of telomeric chromosomes. 0 DAPI-negative; 1 DAPI-positive. (9) CMA-staining of telomeric chromosomes. 0 CMA-negative; 1 CMA-positive. (10) Large peritelomeric or telomeric C-band present on long arm of 6th chromosome. 0 absent; 1 present. (11) Large peritelomeric C-bands on 10th chromosome. 0 absent; 1 present. (12) Large peritelomeric C-band on 11th chromosome. 0 absent; 1 present. (13) Paracentromeric bands on 1st and 3rd chromosome. 0 absent; 1 present. (14) NOR localization. 0 interstitial on short arm of the 1st chromosome pair; 1 on short of the 2nd chromosome pair. (15) Peritelomeric band on the long arm of the 3rd chromosome pair: 0 absent; 1 present.

Fig. 6

Neighbor-joining tree based on analysis of data in Table 4. Numbers are bootstrap values in percent (2000 replicates).

DISCUSSION

The results largely support previous data on Mantella karyology (Blommers-Schlösser, 1978; Pintak et al., 1998). Generally, chromosome morphology in Mantella is rather uniform (all species have 2n = 26, with five larger and eight smaller chromosome pairs). However, several of the interspecific differences observed may bear taxonomic relevance.

The presence of an acrocentric chromosome pair in Mantella nigricans, unique in the genus, supports the hypothesis (Vences et al., 1999) that this taxon stands on its own at the species level. The differentiation found between Mantella betsileo and M. cf. betsileo regarding the morphology of the sixth chromosome (metacentric vs. submetacentric) and the important differences in heterochromatin distribution support the hypothesis that these two forms may be not conspecific. The affinities of the species of the M. betsileo group to each other and to M. laevigata is supported by their general chromosome morphology (Fig. 2).

Heterochromatin distribution is quite different among Mantella species. Actually, at least faint differences were observed between each species included in our study. Hence, the results have to be interpreted with some caution, as we studied only a limited number of specimens of each species, and too few data are available on differentiation between con-specific populations of Mantella. However, as we studied both males and females of most species (see appendix), we can exclude at least that the differences described here are erroneous interpretations of sexual dimorphism.

According to our analyses of Mantella, the same heterochromatic band can be DAPI-positive (thus rich in A+T) or CMA₃-positive (thus rich in G+C). So far, in different plant and animal species, the positive staining of a heterochromatin band has been observed to be limited to either DAPI or CMA₃ (John, 1988; Schmid and Guttenbach, 1988). Molecular studies of the organization of sex-linked satellite DNA in chicken W chromosomes has shown that specific satellite families are included in different chromomeres on the W lampbrush chromosome (Solovei et al., 1998). The arrangement in different chromomeres is a known character in heterochromatic bands (Okada and Comings, 1974). We therefore hypothesize the presence of different families of highly repeated DNA sequences (one rich in A+T and another one in G+C) in Mantella, which are each arranged in distinct chromomere units. In different species, selective amplification of these units may lead (1) to bands containing either mainly A+T rich chromomere units or G+C rich chromomere units (and thus staining either DAPI+ or CMA+, as in M. betsileo, M. madagascariensis, and M. viridis); (2) to bands containing both types of units in comparable proportion (and thus staining DAPI+ and CMA+ as in M. aurantiaca, M. cf. betsileo, and M. pulchra). A similar situation was observed in other Malagasy anuran genera belonging to the superfamily Ranoidea (Mantidactylus, Boophis, Heterixalus), and may therefore be widespread among ranoid anurans (Odierna, pers. obs.).

Taxonomic and limited phylogenetic relevance within Mantella may be attributed to the existence of specific locations in which the accumulation of heterochromatic material is possible or excluded. Indeed, within Mantella, two main groups of species can be distinguished regarding the heterochromatin distribution. In one, heterochromatin was mainly found in the telomeric or peritelomeric regions of the chromosomes 6-13 and was largely absent or scarce in the centromeric regions of these elements (M. aurantiaca, M. baroni, M. cowani, M. madagascariensis, M. nigricans, M. pulchra, and M. viridis). In the other, the centromeric regions of the chromosomes 6-13 were richer in heterochromatin than their telomeric and peritelomeric regions (M. betsileo, M. cf. betsileo, M. expectata, and M. laevigata).

Several further groupings are possible based on the heterochromatin distribution patterns. Mantella aurantiaca, M. madagascariensis, and M. pulchra share the presence of distinct peritelomeric bands on the 6th, 10th, and 11th chromo-some. Mantella aurantiaca additionally has a peritelomeric band on the 9th chromosome. This largely corresponds to the data of Pintak et al. (1998) who have found peritelomeric bands on the 6th, 11th, and 12th chromosome in M. aurantiaca (correct ordering of the small chromosomes is often difficult and may vary according to the method used). Pintak et al. (1998) have further observed peritelomeric bands on the 6th and 11th chromosome of M. crocea, indicating that it is also closely related to M. aurantiaca, M. madagascariensis, and M. pulchra. This is also corroborated by allozyme data which strongly support a monophyletic group containing these four species (Vences et al., 1998c). A peritelomeric C-band on the 6th chromosome also occurred in M. baroni, as indicated by the present data and data in Pintak et al. (1998). In contrast, allozyme data and osteology clearly indicate that M. baroni is part of a monophyletic group containing also M. cowani and M. nigricans (Vences et al., 1998b, c). Thus, the band on the 6th chromo-some may be not homologous in M. baroni and in the M. madagascariensis and M. aurantiaca groups (as also supported by the different reaction to CMA₃ and DAPI stainings of the band in the three groups). Alternatively, the band may have been lost in other members of the M. baroni group.

Another species pair with similar heterochromatin distribution comprised M. laevigata and M. expectata. These are the only species which showed adjacent separate bands in pericentromeric areas which were either CMA+ or DAPI+. Mantella expectata belongs to the M. betsileo group which also contains M. viridis, whereas M. laevigata has an isolated position within the genus (Vences et al., 1998c, 1999). Both M. laevigata and the M. betsileo group are thought to be basal groups within Mantella (Vences et al., 1998b, c), but a sister-group relationship of M. laevigata and M. expectata is contradicted by allozyme data (Vences et al., 1998c).

The location of the NOR has been demonstrated to be of phylogenetic and taxonomic validity in many animal groups including amphibians, reptiles, and fish (Amemiya and Gold, 1990; King, 1990; Olmo et al., 1993). In the genus Mantidactylus, Aprea et al. (1998) have found variability of the NOR location among different species groups and subgenera but, on the other hand, a constant state within these groups. The state occurring in Mantella is similar to Mantidactylus grandisonae (subgenus Blommersia) following the data presented herein. However, it differs from the states in the subgenera Brygoomantis (Mantidactylus alutus, Aprea et al., 1998) Gephyromantis (Mantidactylus luteus, Aprea et al., 1998; M. silvanus, Odierna, unpubl.), Pandanusicola (Mantidactylus bicalcaratus, M. cf. punctatus, data herein), and Phylacomantis (Mantidactylus redimitus, Aprea et al., 1998). Vences et al. (1998b) have hypothesized that the subgenera Guibemantis, Blommersia, and Pandanusicola may be the closest extant relatives to Mantella. The chromosomal data indicate that Blommersia probably is a better candidate for the sister group of Mantella than Pandanusicola. No NOR data are available so far for Guibemantis.

The analysis of intrageneric Mantella relationships based on karyological characters did not provide adequate phylogenetic resolution (Fig. 6). The karyological characters studied may have evolved either too slow (NOR) or too fast (heterochromatin distribution) to match the main cladogenetic events within Mantella. The NOR data appear to be more informative for rather old splits (e.g., the separation of the Blommersia Mantella clade from other Mantidactylus) whereas the heterochromatin is so quickly re-distributed that only presumably very young groups (such as the clade containing Mantella aurantiaca, M. madagascariensis, and M. pulchra which are genetically extremely similar according to Vences et al., 1998c) conserve some common, slightly informative patterns.