Molecular phylogenies , chromosomes and dispersion in Brazilian akodontines ( Rodentia , Sigmodontinae )

A new molecular phylogeny for akodontine rodents from Brazil was proposed. The phylogenetic tree was enriched with the area of occurrence and with information on the karyotype of the samples. Based on this enriched tree, and with a described methodology, hypotheses were proposed on the karyotype and area of occurrence of the ancestors of each Clade. Thus it was possible to discuss hypotheses on chromosome evolution of the group, and on dispersion events from the “area of original differentiation” of akodontines in the Andes. Chromosome evolution started with high diploid numbers (2n=52) and showed a tendency to reduction (until 2n=14 in more recent clades). Independent side-branches of the tree showed 2n reduction and in one case the 2n increased. At least four dispersion events from the Andes down to South-eastern Brazil were proposed. The results should suggest the direction of new studies on comparative karyology.

Most of the South American cricetid rodents belong to an assemblage of genera usually grouped under the subfamily Sigmodontinae.According to the view originally proposed by HERSHKOVITZ (1966HERSHKOVITZ ( , 1972)), and supported by SAVAGE (1974), REIG (1975REIG ( , 1978REIG ( , 1980REIG ( , 1981) ) and ENGEL et al. (1998), the North American ancestor entered South America through its northwestern corner by over water dispersal prior to the establishment of the Panamanian bridge, most probably between 5 and 9 MYA.This is the usually accepted hypothesis although there are some alternatives (see PARDIÑAS et al., 2002).
According to REIG (1984) the original stock ancestral to Sigmodontinae may be found within the tribe Oryzomyini from which the Akodontini differentiated probably in the central Andean region.REIG (1986) considers that the ancestral akodontine may have been a generalized Akodonlike form of boreal origin, which colonized the area of the Puna from the north in Late Miocene or Early Pliocene, before the Altiplano reached considerable heights, in the Middle Pliocene (AHLFELD, 1970).REIG (1984) assumes that the evolution of Akodontini is a history of successful dispersion, certainly promoted by a trophic and habitat versatility of the most speciose genera such as Akodon Meyen, 1833 which became distributed over the greater part of South America, but prevailing in the Andean area.
To better understand the geographic history of the Sigmodontinae Reig (1984) introduced the concept of Area of Original Differentiation (AOD) as a tool to infer the patterns of geographical origin and of dispersion of the different groups of high Andean rodents.The AOD is defined as "the geographical space inside of which a certain taxon suffered its main differentiation (cladogenesis) in the subordinate taxa that compose it."In that sense, an AOD is not a "center of origin", since the ancestral lineage of a taxon that has a certain AOD does not have to be originated in the same area in which the taxon suffered its main differentiation processes.At the same time, the AOD of a taxon is the area where a new derived taxon originates by cladogenesis and that thereafter invades another area in which it suffers its main cladogenesis.This new area constitutes the AOD of the derived taxon.For the determination of the AOD of Akodontini, REIG (1984) identified the area of greatest overlapping of the geographical distributions of the different species of this tribe.The data used were the records of distribution of each species, the frequency of a given taxon in a certain geographical unit, the analysis of the overlapping of the distribution of the genera in the area and also the percentage of endemic genera.REIG (1984) concluded that the South Central Andean Unit (SCA) and the North Southern Andean unit (NSA) are the regions of greatest occurrence of akodontine genera and subgenera (see a definition of these units in SIMPSON (1975) andREIG (1986)), suggesting that the akodontines probably differentiated in the intermediate zone between the SCA and the NSA.REIG (1986) considers that to explain the present distribution of members of the Akodontini it may be necessary to assume that most of the taxa originated in what is now the southern Puna region (AOD), and that they migrated northwards, southwards and to the eastern lowlands and experienced there further differentiation.
We assume that akodontines of the Atlantic Forest or their ancestrals came from the AOD.The question rises: was it a single dispersion event with differentiation in the Atlantic Forest or several events?And when did it happen?
To understand the karyotype evolution of akodontines it is necessary to know the direction of the chromosome rearrangements, toward fusion or fission.That is, if evolution proceeded from high to low diploid numbers by centric fusion or in the other direction by centromeric dissociation (BIANCHI et al., 1971;FAGUNDES et al., 1997;GEISE et al., 1998).
To determine this direction as well as plesiomorphic or apomorphic character states in karyotype, a reliable external phylogeny, morphological or molecular, is necessary.
SMITH & PATTON (1993, 1999), ENGEL et al. (1998), GEISE et al. (2001) and D'ELÍA (2003) addressed the question of the phylogeny of Sigmodontinae using sequences of the Cytochrome b gene of MT DNA.Fourteen species of Akodon of different geographical areas appeared in these author's cladograms as a monophyletic group with high bootstrap support.
In this paper we present a phylogenetic tree based on Cytochrome B sequences using new tissue samples and sequences taken from GenBank.Starting from Reig's theory of AODs (REIG, 1986), and with help of a molecular phylogeny, our aim is to propose new hypothesis on historical biogeography and chromosome evolution of eastern Brazilian species of Akodontini.This is probably the first study combining data from molecular phylogenies with information on karyotype and geographic distribution to understand aspects of karyotype evolution and geographic history of Brazilian akodontine rodents.

MATERIAL AND METHODS
Total DNA was obtained from grounded tissues of 20 samples belonging to eight species of the genus Akodon (Tab.I).Another 28 sequences were obtained from GenBank (Tab.I).Tissues were stored at the Laboratório de Citogenética de Mamíferos of the Departamento de Sistemática e Ecologia, Universidade Federal da Paraíba, and processed for DNA extraction (SAMBROOK et al., 1989) at the Laboratório de Genética e Biologia Molecular, Campus Bragança of the Universidade Federal do Pará.For each DNA sample a fragment of about 800 base pairs of the mitochondrial Cytochrome B gene was amplified and sequenced using primers described by SMITH & PATTON (1993).Further details of the technique used may be found in BARROS et al. (2003).
The Cytochrome b gene fragment sequenced in this study was aligned to homologous sequences obtained from the GenBank using the ClustalX program (THOMPSON et al., 1997) with default parameters and the data converted to FASTA and NEXUS formats.Minor modifications in the alignment were made using the BIOEDIT sequence editor (HALL, 1999).Sequences were deposited in GenBank, accession numbers pending.Nucleotide saturation was assessed by plotting transitions and transversions against K2P distances (KIMURA, 1980) et al., 2001).As outgroup we used Oryzomys (Hylaeamys) megacephalus (Fischer, 1814), a member of the sister group of the akodontines in the tree of D'ELÍA (2003).
The procedure used to reconstruct the chromosome evolution and the biogeographic history of Akodontini started from the concept of AOD, and from the need to polarize the chromosome rearrangements with help of an external molecular phylogeny.We formulated hypotheses to determine the diploid number and the geographic distribution of the most recent ancestral of the terminal units of the cladogram, repeating thereafter the procedure with the nearest ancestral toward the root of the tree.This information is than incorporated in the "enriched phylogenetic tree".
Once a phylogenetic tree was obtained we established for convention that the change of the character state, that is the anagenesis, occurred between the nodes of the tree and the nodes represent the ancestrals from which originated new taxa with different characteristics.The characters presented in table I and on the right side of figure 1, correspond to the terminal taxa, and were used as a base to infer the characteristics of ancestrals.Since the characters of the terminal taxa are known, the characters of its more recent ancestral can be inferred.If all the terminal taxa of the clade considered possesses the same character state this will be attributed to its closest common ancestral.The same procedure was followed to determine the geographic distribution of the ancestral.When one of the terminal units of a certain clade occurs in the AOD, this distribution will be attributed to the ancestral of the whole clade because the basic hypothesis states that the dispersion is unidirectional starting from AOD.In that case, a dispersion event originated in the AOD may have occurred within this clade.Even if the collecting place of the studied specimen is not the AOD but the distribution of the species includes AOD we plotted for him AOD in the tree.
The hypothetical reconstruction of the ancestral will be made starting from the top of the tree.When the enriched tree is finished, reading the tree from the root allows to follow the evolutionary history of the group.
In D'ELÍA'S ( 2003) study, Necromys groups with our clade IV while in our analysis Necromys is a sister group of all forms of Akodon s.s.but in both topologies support is low.
In our topology the basal genera, Brucepattersonius and Blarinomys Thomas, 1896 form a sister group of Oxymycterus (clade I).Scapteromys is a sister group of Necromys and Akodon with low support (BS=32, node 17).In D'ELÍA ( 2003) Scapteromys (clade A) occupies a more basal position being the sister group of Brucepattersonius and Blarinomys (clade B) but also with low support (JK<50, IB=1).In his cladogram the genus Oxymycterus (clade C) is placed closer to Bibimys Massoia, 1979 (clade D) and Akodon (clade E) and not to the genera Brucepattersonius and Blarinomys as seen in our topology.In support of our topology of these basal clades is the agreement with hypotheses of high chromosome numbers in basal forms.
Karyotype evolution.As seen above, to understand the evolution of the karyotype and establish the direction of the evolution we compared karyotype data with the molecular phylogenetic tree.
Since most rearrangements of karyotype are consequence of Robertsonian changes and inversions the 2n and NA are useful tools to trace the chromosome evolution in a certain group.Some caviomorphs, for instance, show a remarkable constancy of NA with changes in 2n.In sigmodontines both 2n and NA change.We work in our reconstruction of karyotype evolution with the diploid number alone but considering also, although not always explicitly, the size and number of arms in the whole karyotype.Primitive karyotypes in akodontines have large number of acrocentrics and derived karyotypes are formed in general by a smaller number of metacentrics usually consequence of centric fusions.
This tendency in akodontine rodents was suggested by several authors (VITULLO et al., 1986;REIG, 1987;GEISE et al., 1998) but changes may occur in both directions and only knowledge of the phylogeny of the group can establish the direction of change.Vitullo et al. (1986), based on other evidence, suggested that the ancestral 2n of akodontines was 58.
In a general overview of the literature on akodontine karyotypes we observed that the differences among them may be the consequence of rearrangements.These rearrangements may be due to inversions or fusion/fission events of the whole karyotype or of large number of chromosomes ("chromosomal revolution") or to just a few Robertsonian changes in a few chromosomes ("low level chromosome diversification").In establishing the ancestral karyotype of a clade we do not considered low level chromosome diversification.For instance 2n of 44, 42 and 40 were all considered as 2n=40.In the following paragraphs we will comment the enriched tree (Fig. 1), clade by clade, starting from the root.
Clade I is the most basal in our phylogenetic tree with ancestral in node 19.It includes three ecologically similar oxymycterine genera inhabitants of the Atlantic Forest floor.Most of them share 2n>52 in agreement with VITULLO et al. (1986) hypothesis that larger 2n is primitive.On the line that leads to Blarinomys a chromosome revolution occurred resulting in strong reduction of 2n.
Clade II includes Kunsia tomentosus (Lichtenstein, 1829) as a sister group of Scapteromys aquaticus Thomas, 1920 andS. tumidus (Waterhouse, 1837) according to D' ELÍA (2003).This clade derived from ancestral in node 17, and suffered a chromosome revolution reducing the 2n to 44 in K. tomentosus (ANDRADES-MIRANDA et al., 1999) and, later, a further reduction to 2n=32 and 24 in Scapteromys by low level diversification.It includes swamp and fossorial rats inhabitants of open country.As shown by ANDRADES-MIRANDA et al. (1999) no autosomes are shared by Kunsia Hershkovitz, 1966 and Scapteromys.Further, Scapteromys differed more widely from the ancestor scapteromyine than Kunsia.
Clade III, derived from ancestral 2n=52 in node 13, suffered also a chromosome revolution reducing the 2n to 34, a rather stable karyotype of a genus of ample distribution in South America occupying open country from Venezuela to Argentina and Brazil.Its area of origin certainly was the AOD, because Necromys occurs today in this area.In D' ELÍA (2003) tree Necromys is a sister group of the genus Thalpomys that has a similar karyotype and is also inhabitant of open country in Brazil.
The clade IV is formed by two species of forest dwellers of the Atlantic Forest and has the ancestral number 11 with an hypothetical karyotype of 2n=52.The 2n=52 of T. nigrita is the old karyotype of node 19.The 2n=46 karyotype of A. serrensis Thomas, 1902 can be derived from the 2n=52 of T. nigrita by low level chromosome diversification.
The area of occurrence of ancestral 10 is the Atlantic Forest of Southeast Brazil.A similar conclusion may be reached interpreting the tree of D'ELÍA (2003).
Clade V derived from ancestral 9, includes a group with around 2n=40, with tendency to reduction in some members attaining a 2n=22.We suggest the 2n=40 for ancestral 8 because this number is also present in the sister clade (ancestral 7).The area of occurrence of ancestrals 9 and 8 would be the region of Peru and Bolivia included in the AOD since most species of the clade V occur in Peru.
The clade VI derives from ancestral 7 with 2n=40.This hypothesis is confirmed by the presence of karyotype 2n=40 in the neighboring clades.Members of the clade VI occur in lowlands of Brazilian Southeast, and Northern Argentina, Paraguay and Uruguay which are the probable areas of occurrence of the ancestral.In the clade VI the 2n vary from 38 to 44 numbers derived by low level diversification from the ancestral 6.
The clade VII includes A. mystax with 2n=42.This karyotype may be derived by low level diversification from an ancestral 2n-40 in node 20 To check the position of A. mystax and A. reigi we prepared another tree including sequences in the GenBank used by SMITH & PATTON (1993) Except for the first clade the remaining clades had BS over 90.We checked the identification of the "mystax" specimen MN 48070 (D'ELÍA, 2003) in the collection of the Museu Nacional do Rio de Janeiro, Brazil, it belongs to A. reigi.Akodon mystax MN 48041, from Rio de Janeiro State AF184054 GEISE et al. (2001) grouped with high bootstrap with our A. reigi, it was probably misidentified.Akodon paranaensis Christoff, Fagundes, Sbalqueiro, Mattevi & Yonenaga-Yassuda, 2000 is on morphological grounds e junior subjective synonym of A. reigi.The degree of genetic divergence sometimes considered "moderate" and others "deep" by GONÇALVES et al. (2007) is not a taxonomic criterion to separate species.
As a result, the true A. mystax form a sister group of a clade formed by A. montensis and A. reigi.The A. lindberghi Hershkovitz, 1990from GEISE et al. (2001) groups with 100% BS with our specimen of A. mystax.According to GONÇALVES et al. (2007) A. mystax and A. lindberghi are sister species that can usually be separated morphologically.
Clade VIII includes A. reigi 2n=44 and A. montensis 2n=24.The phylogenetic line originated in node 2 suffered a karyotype revolution reducing drastically the 2n.See GEISE et al. (1998) for details of chromosome shuffling.
Clade IX includes a group of Peruvian species with a shared 2n=40, that is a sister group of the A. reigi -A.montensis assemblage of the Atlantic forest Akodon.
The clade X contains two parapatric sibling species A. cursor (Winge, 1887) 2n=14 and A. aff.cursor 2n=16 that form two monophyletic groups.Akodon cursor 2n=14 showed an extensive polymorphism (YONENAGA, 1972;FAGUNDES et al., 1998) and is distributed over the Southeastern part of the Brazilian Atlantic Forest.Akodon aff.cursor 2n=16 show a restricted polymorphism (MAIA & LANGGUTH, 1981) and is distributed over the Northeastern part of the Brazilian Atlantic Forest with an enclave in the State of Paraná.
If the tendency was toward the reduction of the 2n, the ancestral 1 of the clade X would have 2n=16 and may have occurred in the Atlantic forest of Southeastern Brazil.
The sister group of clade IV underwent between nodes 11 and 9 another chromosomal revolution reducing the 2n to around 40 chromosomes (38)(39)(40)(41)(42)(43)(44).This group suffered subsequent chromosome revolutions reducing independently the 2n to 14 (A.cursor) between nodes 1 and 3 on the one side and to 24 (A.montensis) between nodes 2 and 14 on the other side.These sister groups are genetically close (YONENAGA et al., 1975;FAGUNDES et al., 1997) being sibling species but are not part of the same lineage, that is, the 2n=14 species do not derived from the 2n=24 form.Besides this, most Akodon clades showed low level differentiation the 2n varying from 38 to 44 and in a few cases reducing gradually to 2n=22.
In the tree shown by D'ELÍA (2003) andD'ELÍA et al. (2005) there is a clade below the T. nigrita-A.serrensis branch that includes the genus Bibimys, a sequence not available to us.Bibimys is very specialized morphologically and with an equally odd karyotype 2n=70, NF=76 or 80 (DYZENCHAUZ & MASSARINI, 1999;GONÇALVES et al., 2005).This form probably derived by chromosomal revolution from an ancestral with 2n=52.The tendency of karyotype evolution was in this case not to reduction but to increase of 2n.This is the only plausible hypothesis.It will be interesting to study the mechanisms that lead to this higher chromosome number in Bibimys comparing his chromosome banding patterns with Brucepattersonius and T. nigrita.Certainly pericentric inversions followed by centric fissions worked out this karyotype, as is usually the case in species that increase de diploid number.
The ancestral 19 common to the group of the Oxymycterus-like mice and to the Akodon-like mice had probably a diploid number 52 in agreement with the hypothesis of VITULLO et al. (1986), this number (52) is present in the clade I.
Members of the clade I have a karyotype with diploid number 52-54 and its ancestral an Oxymycterus like form occupied the AOD.In summary within the universe of species studied, chromosomal evolution started from high diploid numbers 2n=58 according to VITULLO et al. (1986) or FN=64 as found in Oxymycterus, followed an evolutionary line of basic 2n=52 from which derived at different times branches which reduced 2n.This 2n reduction was probably the result of drastic chromosome repatterning events (chromosomal revolution).
Derived lines lead independently to reduction of 2n in Scapteromys, in Necromys, in ancestral Akodon, (node 40) in Akodon montensis, and in Akodon cursor.In these cases reduction of number was not simply consequence of centric fusion but involved also pericentric inversions.The general tendency to reduction was apparently reverted in the case of Bibimys.
Dispersion events from the AOD.Starting from the AOD, as defined by REIG (1986), several dispersion events may be suggested.The oldest one corresponds to the arrival of oxymycterines in Brazil (clade I).This was followed by the Scapteromys dispersal to eastern lowlands (clade II).The next dispersion event (clade III, Necromys) occurred over the lowlands of Argentina, Paraguay, Uruguay and Southern and Central Brazil as well as northwards to Venezuela.A new dispersal refers to members of clade IV (A. serrensis and A. nigrita), to the Atlantic Forest of Brazil.The next dispersion event refers to clade VI [A. azarae (Fischer, 1829) and A . molinae Contreras, 1968 etc.] over the lowlands of Argentina, Uruguay and Brazil.Ancestral of clades VII and VIII (A. mystax, A. reigi and A. montensis) dispersed to the Atlantic Forest of Brazil.The last event refers to clade X formed by A. cursor and A. aff.cursor whose ancestral dispersed independently from the AOD to the Atlantic Forest.
Dating dispersion events.We used in our calculations the age of 3,55 My for the separation of Akodon from Necromys (node 13) as suggested by SMITH & PATTON (1993) to calibrate rates of MT DNA sequence divergence at third position transversions.The estimates we obtained show that: ancestral 15 (Fig. 1) started a dispersion event of N. urichi (J. A. Allen & Chapman, 1897) towards Venezuela 3,51 My ago.The ancestral 11 that gave origin to the first dispersion of Akodon to the Atlantic Forest has an age of 3,11 My.The ancestral 7 that originated the clade VI (A. toba Thomas, 1921, A. molinae and A. azarae) in Paraguay and Argentina has an age of 2,52 My.The ancestral 20 that gave origin to the second dispersion of Akodon to the Atlantic forest have an age of 1,82 My.The ancestral 5 that gave origin to the A. reigi -A.montensis group of the Atlantic forest of Brazil has an age of 1,77 My.The ancestral 3 that originated the A. cursor clade has 2,13 My.
These dates confirm that the events of dispersion and differentiation of the akodontines occurred mainly during the Pliocene.
As any hypothesis in science the picture of the evolution of Brazilian akodontines presented here will change with time.We hope that this paper stimulates further studies in comparative karyology, as well as sequencing of other genes or other taxa that will falsify present hypothesis.
Three phylogenetic trees were constructed by the methods of Neighbor-Joining (NJ) (Fig. 1), Minimum Evolution (ME) and Maximum Parsimony (MP).The topology of the taxa was similar in all of them in spite of the low bootstrap values for the basal clades.They differed however in the positions of two clades: Scapteromys Waterhouse, 1837 (clade II, Fig. 1) and A. mystax Hershkovitz, 1998 (clade VII, Fig. 1).

Figure 1 .
Figure 1.Phylogenetic tree obtained by the method of "Neighbor-Joining" considering genetic distances corrected by Kimura 2parameters.The Roman numbers at right identify clades.For each sample is given: the species name followed by sample number when necessary; the diploid number between bars; the country of origin.The bootstrap values are indicated in italics below the branches.Bold numbers identify the ancestral of the clade (node).Inside the square the 2n is given.(AOD, area of original differentiation; MA, Atlantic Forest; BC, Central Brazil; //, chromosome revolution.Broken lines, lineages with ancestral karyotype).
described the chromosomes of B. griserufescensHershkovitz, 1998as having this diploid number).