Phylogeny of Thylamys (Didelphimorphia, Didelphidae) species, with special reference to Thylamys karimii

The genus Thylamys Gray, 1843 lives in the central and southern portions of South America inhabiting open and shrub-like vegetation, from prairies to dry forest habitats in contrast to the preference of other Didelphidae genera for more mesic environments. Thylamys is a speciose genus including T. elegans (Waterhouse, 1839), T. macrurus (Olfers, 1818), T. pallidior (Thomas, 1902), T. pusillus (Desmarest, 1804), T. venustus (Thomas, 1902), T. sponsorius (Thomas, 1921), T. cinderella (Thomas, 1902), T. tatei (Handley, 1957), T. karimii (Petter, 1968), and T. velutinus (Wagner, 1842) species. Previous phylogenetic analyses in this genus did not include the Brazilian species T. karimii, which is widely distributed in this country. In this study, phylogenetic analyses were performed to establish the relationships among the Brazilian T. karimii and all other previously analyzed species. We used 402-bp fragments of the mitochondrial cytochrome b gene, and the phylogeny estimates were conducted employing maximum parsimony (MP), maximum likelihood (ML), Bayesian (BY), and neighbor-joining (NJ). The topologies of the trees obtained in the different analyses were all similar and pointed out that T. karimii is the sister taxon of a group constituted of taxa from dry and arid environments named the dryland species. The dryland species consists of T. pusillus, T. pallidior, T. tatei, and T. elegans. The results of this work suggest five species groups in Thylamys. In one of them, T. velutinus and T. kariimi could constitute a sister group forming one Thylamys clade that colonized Brazil.

Opossums of the genus Thylamys are small in size and have long, soft, hairy coats with a brown-gray threecolor pattern on the shoulders.They are pouchless, have exposed teats in the abdominal region, and a prehensile tail with storage capacity for a reserve substance at the base (GARDNER & CREIGHTON, 1989;HERSHKOVITZ, 1992;EISENBERG & REDFORD, 1999).Specimens of this genus are exclusive to South America and are found in Brazil, Peru, Bolivia, Chile, Paraguay, Uruguay, and Argentina in open, dry, semi-arid habitats mainly from sea level to heights of over 3,500 meters (CREIGHTON, 1985;PALMA, 1995).GARDNER (2005) lists three Thylamys species in Brazil: T. karimii, found only in the states of Pernambuco and Mato Grosso; T. macrurus, found in the southern region; and T. velutinus, found in the southeastern region.
However, ecological studies conducted in Central Brazil pointed to the presence of T. velutinus in the central portions of the Cerrado biome, therefore extending the geographic distribution of this species (PALMA & VIEIRA, 2006).
In the review of the genus conducted by CARMIGNOTTO & MONFORT (2006), the geographic distribution of T. karimii is considerably extended, including the glades in the ''Cerrado' ' and ''Caatinga'' biomes in northeastern, southeastern, and central Brazil.The genus is currently known to live in the Brazilian states of Rondônia, Mato Grosso, Tocantins, Piauí, Pernambuco, Bahia, Goiás, and Minas Gerais.In central Brazil, T. karimii lives sympatrically with T. velutinus.CARMIGNOTTO & MONFORT (2006) established that the three Brazilian species of Thylamys are different from one another and from the other members of the genus in terms of the combination of physical and craniodental characteristics.Thylamys karimii presents morphological traits adapted to terrestrial habits with an inconspicuous three-color dorsal pattern.The tail length is shorter than the sum of the head and the body length.The tail end is not prehensile and the paws and fingers are very small, with dermatoglyph-bearing plantar pads either too small or absent.The species that most resembles T. kariimi is T. velutinus, though in the former the dorsal region is brown and the sides are whitish, while the latter presents a reddish-brown dorsal region and grayish sides.Like the other species of the genus, T. karimii presents a chromosome diploid number of 2n = 14 with fundamental number variation: FN = 20 or 24 (CARMIGNOTTO & MONFORT, 2006;CARVALHO et al., 2002, respectively).
Molecular data published by PALMA & YATES (1998) and PALMA et al. (2002) suggest that there is a close relationship between T. macrurus and T. pusillus, which makes these species a separate group from the other species of the genus.These proposals led SOLARI (2003) to propose three species groups in Thylamys: the Andean group (T.elegans, T. pallidior, T. tatei, and T. venustus), the Paraguayan group (T.macrurus and T. pusillus), and the Brazilian group (T.karimii and T. velutinus).SOLARI (2003) and CARMIGNOTTO & MONFORT (2006) considered that the Brazilian group derived from the Paraguayan group, spreading across the Brazilian glades.Nevertheless, BRAUN et al. (2005) obtained arrangements that support four different groups: the Paraguayan (T.macrurus), Yungas (T.venustus and related taxa), Chacoan (T.pusillus), and Andean (T.elegans and related taxa and T. pallidior).In this proposal, the "forest species", represented by the Yungas and Paraguayan groups, are the most basal and the "dryland species", which are highly adapted to arid and semi-arid environments (Chacoan and Andean groups), are the most derived assemblage.PALMA et al. (2002) and BRAUN et al. (2005) used cytochrome b sequences to estimate the phylogenetic relationships between the species of the Thylamys genus, but neither completely clarified the issue because in both investigations the Brazilian T. velutinus and T. karimii species (this one widely distributed in South American territories) were not included.The object of this study is, therefore, to re-analyze the relationships within the Thylamys genus using the same molecular markers, including T. karimii, to assess its position.
Since the localities coordinates of the exemplars collected by PALMA et al. (2002) and BRAUN et al. (2005) were not referred, they were estimated using the Global Gazetteer Software Version 2.1.(available at: http:// www.fallingrain.com/world/).
Nucleotide acid sequence analysis.DNA was extracted from the kidney, liver, heart, or muscle (stored at -20 o C or in ethanol 70 % purity) using the standard protocol described in MEDRANO et al. (1990).The partial mitochondrial cytochrome b gene sequences were isolated via polymerase chain reaction (PCR) using the primers MVZ 05 (light-strand) -CGA AGC TTG ATA TGA AAA ACC ATC GTT G with MVZ16 (heavy-strand) -TAG GAA RTA TCA YTC TGG TTT RAT, as suggested by SMITH & PATTON (1993), and the following primers were used to amplify the complete cytochrome b sequence MVZ 05 (as mentioned above) with Mus 15398 (heavystrand) -GAA TAT CAG CTT TGG GTG TTG RTG in accordance with ANDERSON & YATES (2000).PCR products were purified with exonuclease I and shrimp alkaline phosphatase (Amersham Biosciences).The specimens were sequenced directly from purified PCR products using the primers cited above and the ABI Prism BigDye Terminator Cycle Sequencing Ready Reaction Kit (Applied Biosystems) according to the manufacturer's instructions.Sequencing of both strands was done using an ABI Prism 3100 Genetic Analyser (Applied Biosystems).The complete (MN36926) and partial (MN36405 and MN36285) cytochrome b sequences of the voucher specimens are available in GenBank as shown in table I.
Data analysis.The sequences obtained were read using the Chromas 1.45 program, and aligned using the Phylogeny of Thylamys (Didelphimorphia, Didelphidae) species... Clustal X 1.81 program (THOMPSON et al., 1997) under the default setting costs, and were manually refined with the aid of the GeneDoc program (NICHOLAS & NICHOLAS, 1997).The composition of bases and Kimura 2-parameter distance (KIMURA, 1980) were obtained with the Molecular Evolution Genetics Analysis software, Version 4 (MEGA 4;TAMURA et al., 2007).
The phylogenetic analysis was performed using neighbor-joining (NJ), maximum likelihood (ML), and maximum parsimony (MP) algorithms using PAUP* v.4.0b10, (SWOFFORD, 2001).Prior to the analyses, the most appropriate model of DNA sequence evolution was evaluated using the ModelTest 3.7 (POSADA & CRANDALL, 1998).The ModelTest chose the HKY+I Ã model as the best fit for our data.For ML tree estimation, heuristic searches with as-is and tree bisection-reconnection (TBR) branch swapping were selected.The support estimates for the ML trees branches by bootstrap analysis were obtained as described in XIANG et al. (2002).MP analysis was performed using a heuristic search with TBR branch swapping with the MULPARS option in effect (this option requests the saving of all equally most parsimonious trees; without this option in effect, only one shortest tree is saved in each replicate) and 100 random-addition replicates.Bootstrap statistical support (FELSENSTEIN, 1985) was carried out with 10,000 replications of the heuristic search and simple taxon addition, with the "all trees saved" option.
Bayesian analyses of the data were performed using MrBayes 3.0b4 (HUELSENBECK & RONQUIST, 2001) to generate a posterior probability distribution using Markov Chain Monte Carlo (MCMC) methods.No a priori assumptions about the topology of the tree were made and all searches were provided with a uniform prior.The MCMC processes were set so that four chains were run simultaneously for one million generations, with trees being sampled every 100 generations for a total of 10,000 trees.We excluded the first 100,000 generations as the "burn-in" period.
For the distance analysis, trees were constructed using the NJ method (SAITOU & NEI, 1987) with Kimura two-parameter distances.Reliability of the trees was tested using 10.000 bootstrap replications (HEDGES, 1992).

RESULTS
Two partial (727bp and 758bp) and one complete cyt b gene sequences (1149bp) were obtained for the T. karimii specimens analyzed in this study.However, as the majority of the taxa was studied using a 402-bp cyt b gene sequence fragment (BRAUN et al., 2005), we performed the phylogenetic analyses using this fragment length to gather information about the majority of the Thylamys species.
In the analysis, 156 variable sites were observed, of which 30 (19.23 %) were in the first position, 13 (8.34 %) were in the second position, and 113 (72.43 %) were in the third codon position.The average transition/ transversion rate observed among the taxa analyzed was 2.6 and the CT transition occurred most frequently.Generally, the tree topologies obtained by these different methods of analysis were similar in that all bootstrap values of the nodes generated were higher than 50 %, the majority being higher than 90 %.
The tree generated by the ML analysis is shown in figure 2, where T. macrurus is recovered as the most basal taxon sister group of clades with high support (91 %).In the first clade (grouping the "dryland species"), T. karimii exemplars are positioned as the sister group of the clade T. pusillus + (T.pallidior + T. tatei + T. elegans) with a bootstrap value of approximately 70 %.The second clade, named Yungas group, includes T. cinderella and T. venustus recovered as reciprocally monophyletic sister groups with bootstrap support of 54.The most parsimonious tree (not shown) presented 332 steps, a consistency index (CI) of 0.608, a retention index (RI) of 0.789, and the homoplasy index (HI) = 0.391.In the parsimony analysis, 130 sites were informative.The T. macrurus exemplar was recovered as basal to the other Thylamys taxa with a bootstrap value of 100.As seen in the ML tree, in this analysis (MP) the same two clades (dryland and Yungas) were observed, also with high support.In the clade that comprises the T. karimii specimens, it was also the most basal member but with a lower bootstrap value.The relationship among the other specimens of this clade (T.pusillus, T. pallidior, T. tatei, and T. elegans) was an unresolved politomy.The other clade comprised T. cinderella and T. venustus as sister groups.
In the NJ and BY analysis (not shown) the topologies generated in other analyses were in general maintained.In both analyses, T. karimii specimens positioned as the sister group of T. pusillus, T. pallidior, T. tatei, and T. elegans with 86% support (NJ) and 0.92 of probability (BY).T. venustus and T. cinderella, as observed in the ML tree, presented as the sister group of the other species of Thylamys with a bootstrap of 99 % (NJ) and a posterior probability of 1.0 (BY).
Percentage sequence divergences based on Kimura 2-parameters corrected distances were calculated among the specimens (Tab.II).An average genetic distance of 16.29 % was observed (ingroup only).With the exception of T. macrurus, which presented considerable genetic distances (the highest observed was 29.0 % when compared with T. pusillus from Argentina), the remaining species of the genus had smaller average distances between them (ranging from a minimum of 9.7 % between T. pallidior and T. pusillus to a maximum of 20.4 % between T. venustus and T. karimii).The identification of the T. macrurus specimen sequenced by PALMA et al. (2002) could not be verified by us, so we relied on the authors voucher identification.The three T. karimii specimens we sequenced presented intraspecific variations of 0.53 % and interspecific divergences from 15.8 % to 26.0 % when compared with the sequences of T. cinderella and T. macrurus, respectively.

DISCUSSION
Several hypotheses were suggested regarding the relationships of the three Brazilian Thylamys species with the other taxa of the genus.However, they were inconclusive because they did not include the species T. karimii, which is largely distributed in Brazilian territory.These relationships could be better clarified in this study with the inclusion, in a molecular analysis, of the Brazilian species T. karimii, which inhabits dry regions.
Previously, SOLARI (2003) and CARMIGNOTTO & MONFORT (2006), proposed three species groups in Thylamys as the first approach to natural groups: the Andean (T.elegans, T. pallidior, T. tatei, and T. venustus), the Paraguayan (T.macrurus and T. pusillus), and the Brazilian groups (T.karimii and T. velutinus).PALMA et al. (2002), andSOLARI (2003) suggested that T. karimii would present a basal position, possibly derived from T. macrurus and related to T. pusillus.BRAUN et al. (2005), through molecular data, obtained arrangements that were incongruent with those suggested by the authors mentioned earlier.The results obtained by BRAUN et al. (2005) supported four different groups: Paraguayan (T.macrurus), Yungas (T.venustus and related taxa), Chacoan (T.pusillus), and Andean (T.elegans and related taxa and T. pallidior).Our study showed that T. karimii is the sister taxon of the Andean species-group of BRAUN et al. (2005).In all topologies generated by the different analyses methods, we observed T. karimii positioned in the base of the Andean group (T.pusillus, T. pallidior, T. tatei, and T. elegans, sensu BRAUN et al., 2005), constituting a clade that lives in dry habitats (the dryland species).PALMA et al. (2002) suggest that this preference for dry environments may be the result of past dispersion

Figure 2 .
Figure 2. Maximum likelihood (ML) consensus bootstrap tree obtained from the partial sequences of the cytochrome b mitochondrial gene for Thylamys species.(maximum-likelihood/Bayesian/maximum-parsimony).Hyphens indicate bootstrap support below 50 %.

Table I .
Localities and GenBank accession numbers of the analyzed specimens of the genus Thylamys.