A new species of nectar-feeding bat, genus Lonchophylla, from the Caatinga of Brazil (Chiroptera, Phyllostomidae)

Abstract We describe Lonchophylla inexpectata sp. n. from the Caatinga of Brazil. This new species can be distinguished from all known species of Lonchophylla that occur in Brazil by dental traits, cranial size, and fur colour. Specimens of Lonchophylla inexpectata have been misidentified as Lonchophylla mordax; but Lonchophylla inexpectata is a pale-venter species, similar in external appearance to Lonchophylla dekeyseri. We have found Lonchophylla inexpectata in the Caatinga of North-eastern Brazil; Lonchophylla mordax along the eastern border of the Caatinga and in the Atlantic Forest–Caatinga ecotone in North-eastern Brazil; and Lonchophylla dekeyseri in the Cerrado of Mid-western Brazil, in the Brazilian Cerrado–Caatinga ecotone, and as far west as the Cerrado of Bolivia.


Introduction
Lonchophylla Thomas, 1903 (Phyllostomidae) comprises 12 species of nectar-feeding bats restricted to the Neotropics (Griffiths andGardner 2008, Parlos et al. 2014). Parlos et al. (2014) revised the Lonchophyllinae and established Hsunycteris as a new genus to include the smaller species formerly known as the Lonchophylla thomasi complex.
we determined that the pale-venter Lonchophylla from Barra could be distinguished from L. mordax, and represented an undescribed species. Among distinctive traits distinguishing the Barra specimen from L. mordax are the paler colour of the ventral fur and the smaller skull that has a narrower and more delicate rostrum.
To test this hypothesis and further understand the geographic distribution of Brazilian species, we examined series of Lonchophylla from localities in the Caatinga, Cerrado, and Atlantic Forest, as well as from transitional zones between these habitats. The material used in our comparisons represents all Lonchophylla species known to occur in Brazil. During this process we found additional features that support our hypothesis that the pale-venter Lonchophylla from the Caatinga represents a new species, which we describe below.
Measurements in this report are from adults, and are either in millimetres (mm) or grams ([g] body mass). The body mass was recorded from skin labels. Other dimensions include: the forearm length (FA), from the elbow to the distal end of the forearm including carpals, measured with the wing partially folded; greatest length of skull (GLS), from the posteriormost point of the occiput to the tips of the upper inner incisors; condylo-incisive length (CIL), from the line connecting the occipital condyles to the tips of the upper inner incisors; basal length (BAL), from the anterior margin of the foramen magnum to the tips of the upper inner incisors; maxillary toothrow length (MTL), from the anterior surface of the upper canine, including the cingulum, to the posterior surface of M3; molariform toothrow length (M1M3), from the crown of M1 to the crown of M3; breadth across canines (BAC), greatest breadth across outer surface of the crowns of upper canines, including cingulae; breadth across molars (BAM), greatest breadth across outer edges of the crowns of upper molars; postorbital breadth (POB), least breadth across frontals posterior to the postorbital bulges; braincase breadth (BCB), greatest breadth of the globular part of the braincase; mastoid breadth (MAB), greatest breadth across the mastoid region; mandibular length (MAL), from the mandibular symphysis to the condyloid process; and the mandibular toothrow length (MAN), from the anterior crown of the lower canine, including cingulum, to the posterior crown of m3. Craniodental measurements were taken under binocular dissection microscopes with low magnification (usually 6×). Dimensions were taken by only one of us, using digital callipers accurate to 0.02 mm. Measurements were recorded and analysed to the nearest 0.01 mm, but values were rounded off to 0.1 mm throughout the text because this is the smallest unit that allows accurate repeatability with callipers (Voss et al. 2013). Descriptive statistics (mean and range) were calculated for all dimensions. The statistical significance of differences among samples was assessed by single analyses of variance (one-way ANOVA). This statistics was performed in PAST (Hamer et al. 2001).
Discriminant Function Analysis (DFA) was used to compare taxa. For the analysis, we selected a subset of the cranial dimensions (GLS, CIL, MAB, BCB, POB, BAC, BAM, M1M3, MTL, MAL) to represent different axes of length and width of the skull. As multivariate procedures require complete datasets, missing values (< 3% of the total dataset) were substituted by means. Measurements were transformed to natural logarithms and the covariance matrices were computed considering all variables. DFA was performed in SPSS.
Nomenclature of tooth morphology follows Phillips (1971). Capitalized colour nomenclature follows Ridgway (1912). Holotype. An adult male, USNM 238008, with skin and skull (Figures 1, 2), including mandible, collected by E. Garbe at Barra (12°42'S, 41°33'W), Bahia, Brazil, on January 1908. Skull and mandible are in good condition except for the minimally damaged anteriormost portion of the foramen magnum. The body is prepared as dry skin. Woodman and Timm (2006: 450) described USNM 238008 as a faded skin, but after comparison of its pelage colour with those from other specimens, only mem-  branes seem to be faded. External and craniodental measurements for the holotype and paratypes are in Table 1.

Lonchophylla inexpectata
Paratypes. The paratype series comprises 46 vouchers. Three paratypes are from the type locality in Barra, Bahia (AMNH 235608, FMNH 21077, 21078) Other specimen. One additional specimen (ALP 3686) from the Caatinga of Andaraí, Bahia may represent L. inexpectata. The specimen is preserved in spirit, and the dentition is partially worn, preventing its unambiguous identification.
Diagnosis. Lonchophylla inexpectata can be distinguished from all South American species that occur east of the Andes by the following set of traits: presence of a lingual cusp in the P4, absence of a lingual cusp in the P3, absence of a deep longitudinal groove in the posterior face of the upper canine, proximal portion of the dorsal surface of the forearm not furred, and ventral fur pale.  L. peracchii occurs in the Atlantic Forest, from Espírito Santo southward to São Paulo. Lonchophylla inexpectata can be distinguished from these two species by the presence of a well-developed lingual cusp in the P4, with lingual root in the median portion of the tooth; absence of a groove along the anterior surface of the upper canines; and proximal portion of the dorsal surface of the forearm not covered with fur. Based on the samples we have available, L. inexpectata resembles L. dekeyseri in the pale ventral fur, and L. mordax in the dental morphology. These three species overlap partially in external and cranial size, but in general, cranial measurements for L. inexpectata average significantly larger than those for L. dekeyseri and smaller than those for L. mordax (Table 1).
Lonchophylla mordax has been reported in the literature as a pale-venter species (e.g., Lima 1926, Vieira 1942, Taddei et al. 1983, Nogueira et al. 2007, and subsequent to the description of L. dekeyseri, these taxa have been considered the two paleventer species from NE Brazil (see Taddei et al. 1983, Nogueira et al. 2007, Dias et al. 2013). However, after examining part of the type series of L. mordax (BM 1903.9.5.34 [holotype], USNM 123392 [paratype]), along with one other specimen from the same locality of the type series (MHNG 667.13 [identified as L. mordax by Thomas]), and samples from a nearby locality having similar habitat (Itabaiana, Sergipe)-whose external and skull morphology fit with those of the type series of L. mordax (ALP 8768-8770, 8812-8819)-we concluded that L. mordax has a light-brown ventral pelage, which is consistently darker than the paler ventral pelage of the type material of L. dekeyseri and other samples of this species. The ventral pelage of specimens from Barra, Bahia (L. inexpectata) is similar to that of L. dekeyseri. Under "historical remarks" we discuss the reasons for previous assignments of pale-venter samples from the Caatinga of NE Brazil (= L. inexpectata) to L. mordax.
Lonchophylla inexpectata resembles L. dekeyseri in the pelage colour, but these species can be distinguished by qualitative and quantitative cranial characteristics. Lonchophylla inexpectata is significantly larger than L. dekeyseri in all length measurements of skull and rostrum (GLS, CIL, BAL, MTL, M1M3, MAL, MAN), but L. dekeyseri averages slightly larger in those measurements of the width of skull and rostrum (BAC, POB, BCB, MAB), indicating a longer but narrower skull in L. inexpectata (Table  1). L. inexpectata can be distinguished from L. dekeyseri by the narrower first upper premolar (P3) in occlusal view, with lingual lobe absent or obsolete (in contrast with the usually more robust P3, which has a small or moderately developed inner lobe in dekeyseri [ Figure 6]); absence of a deep longitudinal groove in the posterior surface of the canine; narrower and uninflated rostrum, with more widely projecting lacrimals (wider and more inflated rostrum, and lacrimal region almost indistinguishable in dekeyseri); upper molars (M1 and M2) with low crowns in lateral view (molars with higher crowns in dekeyseri); parastyle of M1 projecting labially over the posterior labial margin of the last upper premolar (P4); mesostyle of M1 shorter; metastyle of M1 well developed (reduced or absent in dekeyseri [ Figure 6]); parastyle of M2 well developed but slender (well developed and more rounded in dekeyseri); mesostyle of M2 shorter; metastyle of M2 distinct, moderate or well developed (reduced or absent in dekeyseri).
Multivariate analysis. To test the results obtained from the morphological analyses, we performed a discriminant function analysis including samples we confidently assigned to L. dekeyseri (three groups from the Cerrado of Mid-western Brazil), L. inexpectata (two groups from the Caatinga of NE Brazil), and L. mordax (one group from the Caatinga of NE Brazil, and one group from the Atlantic Forest-Caatinga ecotone in NE Brazil). The first two discriminant functions (DF1, DF2) summarized 47% and 40% of the total variation, respectively ( Table 2). All samples grouped as expected, confirming the cohesive pattern retrieved from the morphological analysis. Centroids for samples assigned to L. inexpectata were distinct from those of L. dekeyseri and L. mordax across the first two axes, and only a few scores of L. inexpectata are within the dispersal cloud of L. mordax (Figure 7). The three species overlap partially across the first axis, but L. inexpectata distinguishes from L. dekeyseri and L. mordax

Discussion
Historical remarks. Previous assignments of L. inexpectata to L. mordax seem to have originated with Lima (1926: 36) who based his account of L. mordax on the series from Barra, which was collected by Garbe and deposited in the Museu de Zoologia da Universidade de São Paulo. Barra is in the sertão of Bahia (Caatinga), ca. 450-500 km west of Lamarão, which is in the agreste of Bahia (transition between Atlantic Forest and Caatinga; type locality of L. mordax). Thomas (1903: 459) described L. mordax as follows: General external appearance, so far as can be judged by skins, exactly as in Glossophaga soricina, except that the colour averages paler. The type is near "cinnamon-brown" above, the bases of the hairs whitish, and "wood-brown" below, but there is some variation in tone, and the darker specimens are quite as dark as the paler examples of Glossophaga obtained at the same place. Lima (1926) seems to have misinterpreted Thomas (1903) where he reported that "darker specimens [of L. mordax] are quite as dark as the paler examples of Glossophaga obtained at the same place." Lima's conclusion might be biased by the series he had at hand, primarily composed by pale-venter specimens from Barra, Bahia. However, at that time, L. mordax was unquestionably the closest species-geographically and morphologically. Although Lima had identified this series from Barra as L. mordax, the label of the USNM 238008 bears the notation "Subsp. n. ?" Vieira (1942: 321) followed Lima (1926) and based his account of L. mordax on the same specimens collected by Garbe. Both recognized L. mordax as a pale-venter species. This was followed by Taddei et al. (1983) who compared the species they were describing (L. dekeyseri) with "L. mordax"-the other pale-venter species from NE Brazil, according to those authors. However, according to Thomas (1903), the ventral pelage of L. mordax is "wood-brown", but with some variation, with darker specimens almost as dark as paler specimens of Glossophaga from the same area. Glossophaga soricina (Phyllostomidae)-the only species of the genus that occur in the region-has ventral pelage varying from "buffy to fuscous" (Alvarez et al. 1991).
Taxonomic remarks. Molecular and morphological analyses have recovered Lonchophylla (sensu Griffiths and Gardner 2008) as a paraphyletic assemblage (Dávalos and Jansa 2004, Woodman and Timm 2006, Woodman 2007. Combining evidence from nuclear and mitochondrial genes, karyotypes and skull morphology, Parlos et al. (2014) also retrieved Lonchophylla as paraphyletic. Based on their findings, Parlos et al. (2014) described Hsunycteris and moved three species into this new genus-thomasi J. A. Allen, 1904;cadenai Woodman & Timm, 2006;and pattoni Woodman & Timm, 2006. As a result, Lonchophylla comprised 12 South and Central American species (Parlos et al. 2014). However, several species were not assessed, including L. mordax-the type species of Lonchophylla. According to Parlos et al. (2014), the two genera can be distinguished by size (with species in Lonchophylla larger than those in Hsunycteris), qualitative cranial features, and karyotypes (Lonchophylla spp The samples we have available show that L. dekeyseri and L. mordax are in parapatry with L. inexpectata: L. dekeyseri occurs in the Cerrado of Brazil and possibly in the Bolivian savannah (USNM 584472, 584473) and the Cerrado-Caatinga ecotone in NE Brazil (DZSJRP 11459); and L. mordax occurs in the Atlantic Forest-Caatinga ecotone (agreste), and along the eastern border of the Caatinga (sertão). We are not convinced that L. dekeyseri occurs in the Bolivian savannah and in the Cerrado-Caatinga ecotone in NE Brazil. One of the specimens supporting these records was examined a long time ago (DZSJRP 11459), and the other two (USNM 584472, 584473) are distinct from other samples of L. dekeyseri as determined in a previous discriminant function analysis. These specimens are not included in this analysis because we were not able to compare them with samples from other localities. Records previously assigned to L. mordax from N Brazil are based primarily on Handley (1967) and Piccinini (1974), and those identifications were not confirmed in subsequent surveys. We speculate that they are misidentifications of L. thomasi, now Hsunycteris thomasi. Similarly, previous unvouchered records of L. mordax from the Atlantic Forest of SE Brazil apparently represent L. peracchii based on the identity of material we have examined from nearby localities.
After Parlos et al.'s (2014) assignment of L. thomasi J. A. Allen, 1904 to Hsunycteris, L. inexpectata is the fifth Lonchophylla reported from Brazil-all pending phylogenetic positioning. There are several specimens pending verification of identity, particularly those from the Caatinga. Additional material, particularly from NE and Mid-western Brazil, will be important to a clearer understanding of the taxonomic diversity, and the geographic distribution of Brazilian species of Lonchophylla.