Delomys dorsalis ( Hensel 1872 )

Gonçalves, Pablo Rodrigues & Oliveira, João Alves De, 2014, An integrative appraisal of the diversification in the Atlantic forest genus Delomys (Rodentia: Cricetidae: Sigmodontinae) with the description of a new species, Zootaxa 3760 (1), pp. 1-38 : 26-29

Delomys dorsalis ( Hensel 1872)

Hesperomys dorsalis Hensel (1872): 42

Hesperomys dorsalis obscura Leche (1886): 696

Akodon dorsalis: Trouessart (1898): 537 (new name combination) Akodon dorsalis obscura: Trouessart (1898): 537 (new name combination) Akodon dorsalis lechei Trouessart (1904): 434

Hesperomys (Oryzomys) dorsalis: Miranda-Ribeiro (1905): 187 (new name combination) Thomasomys dorsalis: Thomas (1906): 443 (new name combination) Delomys dorsalis: Thomas (1917): 196 (new name combination) Delomys dorsalis collinus Thomas (1917): 197

Delomys dorsalis dorsalis: Gyldenstolpe (1932): 60 (new name combination) Delomys dorsalis lechei: Gyldenstolpe (1932): 61 (new name combination) Thomasomys dorsalis dorsalis: Ellerman (1941): 368 (new name combination) Thomasomys dorsalis collinus: Ellerman (1941): 369 (new name combination) Thomasomys dorsalis lechei: Ellerman (1941): 369 (new name combination) Thomasomys lechei: Moojen (1952): 59

Thomasomys collinus: Moojen (1952): 60

Delomys collinus: Avila-Pires (1960): 32 (new name combination)

Emended diagnosis. A medium-sized species of the genus Delomys with soft and dense dorsal pelage, predominantly dark cinnamon-brown in color and without a bright yellow lateral line or patch; tail as long as head and body; skull with pronounced rostrum and elongated nasal tube, formed by most of premaxillary length and projected beyond the gnathic process; supraorbital region exhibiting the fronto-parietal (coronal) suture discontinuous with the fronto-squamosal suture, forming an area of dorsal contact between the squamosal and the frontal; mandible with wide, deep, and asymmetrically excavated sigmoid notch, robust coronoid process and wellpronounced angular process; karyotype with 2n=82 and FN=80.

Distribution. The samples recognized here as D. dorsalis are distributed from the southernmost limits of the Brazilian Atlantic forest to the northernmost limits of the Serra do Mar, at the center of the Rio de Janeiro state. The distribution of D. dorsalis extends westward to the Misiones province in Argentina, but few records are available in the interior parts of Paraná and São Paulo states. Most collecting localities are situated along the coastal mountains covered by extensive humid montane and submontane forests, which in the Southern Highlands are frequently associated with the austral conifer Araucaria angustifolia . The submontane and montane forests in coastal Brazil occur at elevations of 30 to 1000 m between latitudes of 24°S and 32°S in the southernmost limit of the Brazilian Atlantic forest, but tend to occupy higher elevational intervals of 50 to 1500 m between the latitudes of 24°S and 14°S in Southeastern Brazil ( Veloso et al. 1992). Likewise, the altitudinal distribution recorded for D. dorsalis in Southern Brazil varies from 57 m (locality 66) to 1700 m (locality 63), while in Southeastern Brazil, northern to the latitude of 24°S it varies from 650 (locality 44) to 1950 m (locality 27).

Variation in Delomys dorsalis . As previously evidenced by the molecular phylogenetic analyses, D. dorsalis exhibits an extensive molecular variation in relation to both D. altimontanus and D. sublineatus . Sixty-five haplotypes were identified among the 78 sequenced individuals of D. dorsalis sampled from 18 collecting localities, and their relationships suggest the recognition of three geographically defined intraspecific clades concordant with the distribution of the Mantiqueira, Serra do Mar and Southern mountain ranges in coastal Brazil ( Fig. 13 a). The quantitative results of the AMOVA analysis add further support to this spatial structure, indicating that a substantial component of the variation among haplotypes in D. dorsalis (80.9%; Φ CT= 0.8091, p<0.001) is due to differentiation among the three regions rather than within these regions or within populations (7.64% and 11.45%, respectively; Φ SC= 0.4002, p<0.001), with a significantly large fixation index among populations (Φ ST= 0.8855, p<0.001).

The molecular diversity in the Mantiqueira and Serra do Mar clades is relatively reduced, as haplotypes within these regions average 0.45% and 0.64% pairwise genetic distance, respectively. The Southern clade, on the other hand, exhibits larger polymorphism, with haplotypes averaging 0.91% of pairwise genetic distance. The larger intraclade genetic distance in the Southern group is due to comparisons between haplotypes from the northernmost (states of São Paulo and Paraná) and southernmost localities (states of Santa Catarina and Rio Grande do Sul), which diverge by 1.47% pairwise genetic distance in average. However, the positive and significant correlations between genetic and geographic distances of populations assessed by a Mantel test (r = 0.91, p = 0.025) suggest that the artifactual sampling gap of collecting localities cannot be ruled out as the main cause of the genetic discontinuity between northernmost and southernmost populations in the Southern clade (de Queiroz & Good 1997).

Delomys dorsalis is also morphologically variable and, as previously mentioned, the number of mammary pairs does exhibit a discrete pattern of geographic variation within the species ( Fig. 13 b). The adult females examined from the Mantiqueira and Serra do Mar clades (n = 35) exhibit six mammae, while those from the Southern clade exhibit eight mammae (n = 19, including the Argentinean sample from Misiones), corroborating the pattern of variation described by Thomas (1917) and Voss (1993). The geographic transition between the two mammary conditions occurs around 23°50’S latitude, in the São Paulo state, and no population sampled in this region exhibits polymorphism of this trait (localities 36, 37 and 48). This transition is coincident with the latitudinal location of populations, but not fully congruent with the phylogenetic hierarchy within D. dorsalis , as the 6-mammae and 8-mammae lineages are not recovered as reciprocally monophyletic groups ( Fig. 13 c). A parsimony optimization (accelerated transformation) of mammary variation onto the molecular phylogeny of the genus, assuming the 8-mammae condition as primitive ( Jansa & Weksler 2004; Weksler 2006), suggests two mammary number transformations along the evolutionary diversification of Delomys . The first would involve a loss of the pectoral pair in the ancestor of D. altimontanus and D. dorsalis , and the second would consist in a unique reacquisition of the pectoral pair (reversal) in the Southern clade of D. dorsalis . Nevertheless, since the close relationship between D. altimontanus and D. dorsalis has weak statistical support (<50% bootstrap values and 0.69 posterior probability), the mammary number transformations in Delomys are still uncertain and a better resolved phylogeny of the three species is needed for more conclusive character reconstructions.

A CVA including representatives of the Mantiqueira, Serra do Mar and Southern clades of D. dorsalis , including topotypes of D. dorsalis and the holotype of D. d. collinus , was carried out to test whether the craniometric variation among these three clades was congruent with their molecular or mammae number variation patterns. The three clades are depicted as three broadly overlapping clusters segregating along both CV1 and CV2, which account, respectively, for 70.2% and 29.8% of the among-group variation ( Fig. 14 a). The Mantiqueira clade diverges from Serra do Mar and Southern clades along CV2 by exhibiting longer molar toothrows (LM) and larger zygomatic plates (BZP) ( Fig. 14 b). Serra do Mar and Southern clades diverge along CV1 mostly due to variations in the interobital breadth (LIB), which is relatively wider in the Serra do Mar clade. Unlike the among-species CVA ( Fig. 8 ) or the molecular phylogenies, however, craniometric discrimination among the D. dorsalis clades is less clear cut, with discriminant functions correctly allocating respectively 84.1%, 90.7% and 77.6% of the Mantiqueira, Serra do Mar and Southern clades representatives. Indeed, bivariate plots combining the most discriminant craniometric characters confirm a large overlap between Mantiqueira and Serra do Mar clades and between Mantiqueira and Southern clades ( Figs. 14 c–d). Therefore, craniometric variation among the three intraspecific lineages of D. dorsalis is too subtle to identify exclusive clusters in multivariate or bivariate analyses.

Despite the craniometric continuum among the three intraspecific clades, samples without genetic data could be allocated to one of the three geographic groups with moderate confidence levels. The D. dorsalis topotypes from Taquara (locality 66) were allocated to the Southern clade in 81% of the bootstrap iterations, while the holotype of D. dorsalis collinus and specimens from adjacent localities 20, 30 and 33 ( Fig. 1 ) in Minas Gerais and São Paulo states were allocated to the Mantiqueira clade in all bootstrap iterations. Finally, a female of D. dorsalis with 8 mammae from a geographically distant locality in Argentina (locality 62) could also be confidently allocated to the Southern clade in all bootstrap iterations.

Thomas (1917) remarked on the length of upper molars as a distinctive trait for D. d. collinus (measured 5.0 mm by Thomas and 4.98 mm by us) in relation to Southern samples of D. dorsalis (measured 4.5–4.7 mm by Thomas). However, as shown in the bivariate plot ( Fig. 14 d), the Southern samples of D. dorsalis display a wider variation than that portrayed by Thomas (1917), weakening the value of dental measurements to diagnose D. d. collinus . In general, the craniometric variation among the three clades is subtle and no exclusive clusters representing the three D. dorsalis clades could be identified in the multivariate and bivariate combinations of characters ( Fig. 14 ).

In the light of the current molecular and morphological evidence, applying the name D. dorsalis collinus to the 6-mammate specimens from Mantiqueira and Serra do Mar, as originally envisioned by Thomas (1917) and Voss (1993), would render this subspecies paraphyletic or dyphyletic respective to 8-mammate D. dorsalis dorsalis . An alternative would be to restrict D. dorsalis collinus to the Mantiqueira clade and D. dorsalis dorsalis to the Southern clade, describing the Serra do Mar clade as a distinct subspecies. Although more congruent with the number of lineages in D. dorsalis , this last option does not seem practical, since the recognition of subspecies would be fully dependent on molecular data in the absence of diagnostic morphological characters between the Serra do Mar and Mantiqueira populations. Voss (1993) reported a high frequency of specimens from Boracéia and Casa Grande, in São Paulo state, with white tipped tails. This condition was not found in other samples from Serra do Mar examined by us, such as those from Teresópolis or Parati, in the Rio de Janeiro state, and may rather be a locally polymorphic character. The Serra do Mar clade is also polymorphic for tail color pattern, with a higher frequency (57%) of weakly bicolored tails in samples from Alto da Serra, Boracéia and Casa Grande, while other samples of this clade exhibit higher frequencies (>60%) of strongly bicolored tails. Given the current uncertainty in the morphological discrimination among the three clades of D. dorsalis , we opted not to formally recognize subspecies based on them, and regard D. dorsalis collinus as a junior synonym of D. dorsalis .

Figures

FIGURE 1. Collecting localities for specimens of Delomys sp. nov. (triangles) and D. dorsalis (circles) examined in this study. Localities from Mt. Itatiaia are highlighted in the inset. Acronyms refer to the Brazilian states of (ES) Espírito Santo, (MG) Minas Gerais, (RJ) Rio de Janeiro, (SP) São Paulo, (PR) Paraná, (SC) Santa Catarina and (RS) Rio Grande do Sul, and to the Argentinean province of Misiones (MI). All localities are listed in the Appendix.

FIGURE 8. Craniometric discrimination among species of Delomys: (a) CVA analysis among samples with karyotypic and / or molecular data, and (b) respective vector correlations among characters and CVA axes. (c) CVA analysis with sympatric samples of Delomys sp. nov. and D. dorsalis from Itatiaia Mt. highlighted, showing the allocation of the D. d. collinus holotype. (d) Bivariate plot of most discriminant craniometric characters to discriminate between Delomys sp. nov. and D. dorsalis with the measurements of D. d. collinus holotype included.

FIGURE 13. Distribution of (a) the intraspecific mitochondrial clades and (b) geographic variation in the number of mammae in Delomys dorsalis (white circles ─ 6 mammae; black circles ─ 8 mammae). The inset (c) shows the pectoral mammary pair variation mapped on the molecular phylogeny of Delomys.

FIGURE 14. Craniometric discrimination among intraspecific clades of D. dorsalis: (a) CVA analysis among the three clades and (b) respective vector correlations among characters and CVA axes. Bivariate plot of the two most discriminant craniometric characters (c) between the Mantiqueira and Serra do Mar clades, and (d) between the Mantiqueira and Southern clades, also depicting the allocation of taxonomically important samples.