INTRODUCTION

On the way to extend the knowledge of biodiversity, scientists often come across shortfalls on local-scale data. Those shortfalls generate uncertainty in all analyses of biodiversity, compromising the generality and validity of theoretical knowledge and the quality of conservation assessments and actions (^{Hortal et al. 2015}). One of those recognized shortfalls is the Wallacean, which defines that, for the majority of taxa, geographical distributions are still poorly understood and contain many gaps, being frequently inadequate at all scales (^{Lomolino 2004}; ^{Whittaker et al. 2005}). Such shortfall is particularly true for tropical species, especially for those hardly trapped and/or not abundant in communities, even though many of them possess high conservation value (^{Beck et al. 2018}).

The Neotropical region is home to one of the richest mammal faunas in the world (^{Antonelli & Sanmartín 2011}; ^{Patterson & Costa 2012}). Nevertheless, much of the diversity it houses is still unknown and there are huge sample gaps, even in regions with a long history of human settlement and research, as is the case of the Atlantic Forest (^{Bovendorp et al. 2017}). Regarding the knowledge about Neotropical marsupials, within the last years the number of Didelphidae species has increased from 91 (^{Gardner 2008}) to 103 recognized species (^{Astúa 2015}). However, 15% of these are currently considered as Data Deficient based on IUCN designations (^{IUCN 2019}). Although research, and con- sequently, the knowledge on the Neotropical fauna has increased in recent years, gaps in taxonomic and biogeographic knowledge of specious and yet elusive groups, such as the Neotropical small mammals, hinder conservation initiatives. Effective biodiversity conservation requires minimal knowledge about the targets of protection (^{Brito 2004}), making it essential that efforts be devoted to cataloging, quantifying and mapping such biodiversity.

Cryptonanus is a didelphid genus that was previously recognized as belonging to the genus Gracilinanus, until morphological characters and molecular markers warranted its differentiation as a new genus (^{Voss et al. 2005}). Currently, four species of Cryptonanus are recognized: C. agricolai (^{Moojen, 1943}), C. chacoensis (Tate, 1931), C. guahybae (Tate, 1931) and C. unduaviensis (Tate, 1931), which are distributed throughout open habitats in tropical and subtropical ecoregions east of the Andes and south of the Amazon River, including the Caatinga, Chaco, Cerrado, Pampa, Yungas and Atlantic Forest (^{Díaz et al. 2002}; ^{Voss et al. 2005}; ^{D’Elía & Martínez 2006}; ^{Voss & Jansa 2009}; ^{Garcia et al. 2010}; ^{Quintela et al. 2011}).

C. agricolai, is mostly found in xeric habitats in the Caatinga and open formations of the Cerrado biomes in east-central Brazil, from 400 to 760 m (^{Gardner 2008}), but also occurs in contact zones with the Amazon (in northern Mato Grosso state, ^{Bezerra et al. 2009}) and the Atlantic Forest limits (seasonal tropical forests of Minas Gerais states; ^{Gardner 2008}).

Although there is few information available about its locomotion habits, most faunal surveys report captures in pitfall traps, suggesting a ground-dwelling locomotion (^{Astúa 2015}), and probably an insectivore/omnivore diet. The conservation status for such species is assessed as Data Deficient based on IUCN criterion, since there are few and scattered records over a wide area, and little is known about its habitat requirements, thus continued efforts to improve its geographical distribution is greatly needed (^{Carmignotto et al. 2016}).

Here we report the first records of C. agricolai in the Atlantic Forest core in southeastern Brazil, based on the identification via both morphological and molecular analysis of a recently collected specimen from Espírito Santo State. Additionally, we identify a previously reported Cryptonanus sp. from Rio de Janeiro State (^{Delciellos et al. 2016}) as C. agricolai. Our findings reveal an estimated species distribution of 1 773 398 km² and set its lowest altitudinal range limit to 108 m, in addition to confirm its presence in the Atlantic Forest.

MATERIALS AND METHODS

The Córrego do Veado Biological Reserve comprises a forested area of 2 357 ha in the municipality of Pinheiros, state of Espírito Santo, southeastern Brazil, where sec- ondary vegetation predominates (^{Ibama 2000}). The surroundings of Córrego do Veado Biological Reserve are characterized by anthropic activities, such as cattle ranches and coffee, papaya, eucalyptus and rubber tree plantations (^{Moscal 2012}). On October 17th, 2015, a specimen of a didelphid marsupial (LPC1636; collector number of Leonora Pires Costa) was caught in a pitfall trap at Córrego do Veado Biological Reserve (18°19’ S, 40°07’ W, altitude = 108 m; geographic coordinates were obtained with a Garmin 60CSx) in an area of Mussununga (i.e. “soft and wet white sand” in Tupi-Guarani; ^{Meira-Neto et al. 2005}), characterized by shrub and bush vegetation that develops in sandy soils (Fig. 1; ^{Saporetti-Junior et al. 2012}). The animal was secured as a museum voucher and deposited in the Mammal Collection at Universidade Federal do Espírito Santo under the number UFES-MAM 3075. Liver tissue samples were preserved in 70°GL ethanol and deposited in the Animal Tissue Collection at Universidade Federal do Espírito Santo under the number UFES-CTA 4116.

Biological information obtained from the collected individual included: reproductive stage, dental age class, body mass (BM, in grams) and external body measurements (in mm), including head-and-body length (HB), tail length (TL), hindfoot length including claw (HL) and ear length from notch (E). Skull was removed, cleaned, and measured using a digital caliper accurate to 0.01 mm. The following cranial and dental measurements were taken following ^{Voss et al. (2005)}: condylobasal length (CBL), nasal breadth (NB), least interorbital constriction (LIB), zygomatic breadth (ZB), palatal length (PL), palatal breadth (PB), maxillary toothrow length (MTR), length of M1 to M4 (LM), and length of M1 to M3 (M1–M3).

Additionally, 801 bp of the mitochondrial gene cytochrome b (Cyt-b) were used for molecular analyses, as these have been shown to easily distinguish sister species of mammals (e.g., ^{Bradley & Baker 2001}; ^{Agrizzi et al. 2012}) and also because there are several Cyt-b sequences available for a considerable number of didelphid species in GenBank. A sequence from a specimen referred as Cryptonanus sp. from Rio de Janeiro (MZUSP 35409; catalog number of Museu de Zoologia da Universidade de São Paulo, ^{Delciellos et al. 2016}) was also included in the molecular analyzes. Sequence alignment was performed using CLUSTALW algorithm implemented in MEGA X (^{Kumar et al. 2018}), with posterior manual edition. We performed BLAST searches in GenBank (<http://blast.st-va.ncbi.nlm.nih.gov/Blast.cgi>;) in order to determine the approximate association of the obtained sequences with published records. Phylogenetic inference of Maximum Likelihood (ML), Maximum Parsimony (MP) and Neighbor Joining (NJ) was generated in MEGA X using ^{Tamura & Nei (1993)} model of nucleotide substitution and variable sites following a gamma distribution (TN93+G). Bayesian analysis was performed in Mr. Bayes v3.2.6 (^{Ronquist & Huelsenbeck 2003}). Phylogenetic trees were later edited in FigTree v1.4.3 (^{Rambaut & Drummond 2012}). We also conducted systematic searches in online academic databases (Google Scholar, Scielo, Scopus, Web of Science) for occurrence localities of C. agricolai, in order to elaborate an updated distribution map of the known geographic range of the species– based on a minimum convex polygon that assemble our brand new record and literature records. We only considered as reliable those records that included a detailed description of how the specimen was identified.

RESULTS

The specimen (Fig. 2) UFES-MAM 3075 captured in a pitfall trap on 17th October 2015 was an adult male with dental age class 5 (all molars functional and permanent third premolar completely erupted; ^{Tribe 1990}). External, cranial and dental measurements are shown in Table 1. A combination of morphological aspects related to the upper premolars (P3 > P2) and fenestrae in the palate (large maxillopalatine fenestra, presence of palatal fenestra, posterolateral foramen, incisive foramen and absent of maxillary fenestra) led us to identify the specimen as belonging to the genus Cryptonanus (Fig. 3), according to ^{Voss & Jansa (2009)} and ^{Rossi et al. (2012)}.

For accurate identification at the species level, we also used molecular approaches. An initial BLAST search revealed a 98% identity match with published specimen sequences of C. agricolai. The obtained phylogenetic inference grouped UFES-MAM 3075 (GenBank accession number MW477899) in a highly supported clade of C. agricolai (Bayesian posterior probability = 1). Maximum likelihood and Bayesian inference trees depicted the same topology - only the latter is showed (Fig. 4). We recovered, with strong support, C. agricolai and C. chacoensis as sister taxa, and these two species formed a marginally supported clade including C. guahybae, with C. unduaviensis as the outermost clade, and all together supporting the monophyly of the genus Cryptonanus. The literature review recovered various, unambiguous records of C. agricolai (Table 2), and thereafter a minimum convex polygon estimation of those locations suggests an extension of occurrence area of 1 773 398 km² (Fig. 5).

DISCUSSION

The specimen UFES-MAM 3075 here identified as C. agricolai represents the first record of the species in the Central Atlantic Forest Corridor (^{Brazil 2006}; Câmara & Galindo-Leal 2009), a region formerly considered outside its distributional range (^{Carmignotto et al. 2016}). Previously, two specimens of C. agricolai had already been registered in the Atlantic Forest; however, one of them was tentatively assigned as agricolai (^{Souza et al. 2010}) and the other was identified only up to the genus level (^{Delciellos et al. 2016}). Therefore, the present study provides an accurate identification of specimens suspected to belong to C. agricolai, confirming its occurrence in the core of the Atlantic Forest biome, expands the extent of occurrence of the species by 324 452 km² (Fig. 5; hatched area), and increasing its known distribution to approximately 1 773 398 km². Additionally, the specimen UFES-MAM 3075 set the lower altitudinal limit for C. agricolai at 108 m above sea level (Table 2). Moreover, according to the geographical coordinates of a trap site in the Cerrado region mentioned in ^{Bonvicino et al. (2012)}, we inferred the potential highest altitudinal limit of the species as 850 m, an increase of almost a 100 m of that previously suggested (400 to 760 m; ^{Gardner 2008}). Hence, we can now establish the altitudinal range of C. agricolai from 108 to 850 m.

Table 1 External, cranial and dental measurements (in millimeters) for samples of the species of Cryptonanus recorded in Brazil, based on data provided by ^{Voss et al. (2005)}, ^{Rossi et al. (2012)}, and ^{Dias et al. (2015)}, and the specimen from Reserva Biológica do Córrego do Veado, Espírito Santo, here reported (UFES-MAM 3075). Number of measured individuals are indicated between parentheses. 

Table 2 Localities of Cryptonanus agricolai obtained from the literature and used as points to elaborate the map of extension of occurrence for the species (Fig. 5). Acronyms after hyphens indicate Brazilian states; BA = Bahia, CE = Ceará, ES = Espírito Santo, GO = Goiás, MG = Minas Gerais, MT = Mato Grosso, PE = Pernambuco, PI = Piauí, RJ = Rio de Janeiro, SE = Sergipe, TO = Tocantins. Asterisks (*) indicate values inferred from the geographic coordinates provided by authors. 

Fig. 1 Example of Mussununga formations in the Atlantic Forest of northern Espírito Santo at Reserva Natural Vale – 19°13’31” S, 39°58’46” W (photo: Geovane S. Siqueira). 

Besides changes in distributional and altitudinal ranges of C. agricolai, our work confirms the occurrence of the species, often associated with xeric habitats and open vegetation, in a central region of the Atlantic Forest domain, with points of occurrence far from the contact zones of this biome with either the Cerrado or Caatinga domains. Although the individual was captured in an ecoregion distinct from that usually related to C. agricolai, the species can still be primarily associated with xeric-open habitats, since the sampling site in Córrego do Veado Biological Reserve is dominated by patches of vegetation called Mussununga. Such formations are characterized by phytophysiognomies ranging from grasslands to woodlands over sandy soils that have high water retention, with a consistently hard and impermeable cementation layer, which causes flooding stress in the rainy season and drought stress in the dry season (^{Mecke et al. 2002}; ^{Horbe et al. 2004}; ^{Saporetti-Junior et al. 2012}; ^{Buso Junior et al. 2019}). Despite the fact that Mussununga formations are threatened by many factors (e.g., fire, logging, road construction, and biological invasion), such kind of vegetation is still underrepresented in the context of studies conducted in the Brazilian Atlantic Forest (^{Eisenlohr et al. 2015}; ^{Heringer et al. 2019}). This record of C. agricolai in Mussununga formations in the core of the Atlantic Forest opens the possibility that this species also occurs in other xeric phytophysiognomies in the ecoregion, such as dry restinga forests. It is worth mentioning that other species of mammals associated with open areas have been recorded in the Atlantic Forest, such as the canid Chrysocyon brachyurus (^{Xavier et al. 2017}) and the sigmodontine rodent Calomys cerqueirai (^{Colombi & Fagundes 2015}). However, it is uncertain whether these occurrences are due to recent changes on species distributions, following the opening of vast regions previously occupied by forests, or if these mammals already occurred in open phytophysiognomies of Atlantic Forest and are only now being registered.

In addition to C. agricolai, another species of the genus occurs in the South America diagonal of open formations (e.g., Chaco, Cerrado and Caatinga domains; ^{Prado & Gibbs 1993}). Cryptonanus chacoensis, is widely distributed in Gran Chaco, but there is also records for this species in southwest areas of the Cerrado Domain, as far west as the Yungas and into the interior Atlantic Forest of eastern Paraguay (^{De La Sancha & D’Elía 2015}; ^{Teta & Díaz-Nieto 2019}). C. agricolai and C. chacoensis share many morphological similarities, there are strong evidence that each species may constitute a species complex, pending a formal revision (^{Astúa 2015}), and they are possibly sympatric in the southwestern portion of Cerrado. Therefore, we did not include in the updated occurrence map (Fig. 5) records of C. agricolai provided by articles that report its occurrence in this area of potential superposition, without providing a detailed morphological description or the diagnosis reasoning (^{Cáceres et al. 2008}; ^{Paise 2010}; ^{Martin et al. 2012}; ^{Hannibal & Neves-Godoi 2015}; ^{Gonçalves et al. 2018}). Although delimiting the western limits of C. agricolai is somewhat outside the scope of our work, we emphasize the importance of collecting and accessioning specimens (^{Patterson 2002}), meticulous and comprehensive identifications, as well as descriptions of how species were diagnosed when distribution records are given, especially in areas with possible sympatry of cryptic genus and species.

Fig. 2 Museum skin of Cryptonanus agricolai (UFES-MAM 3075) from Córrego do Veado Biological Reserve, Espírito Santo State, Brazil, in dorsal (top), ventral (middle), and right lateral (bottom) views. Scale = 10 mm (photo: Heitor Bissoli). 

Fig. 3 Dorsal (left), ventral (middle), and lateral (right) views of the cranium of Cryptonanus agricolai from Reserva Biológica do Córrego do Veado, Espírito Santo, Brazil (UFES-MAM 3075). Scale = 10 mm (photos: Heitor Bissoli). 

When the genus was described, authors mentioned that significant range extensions of Cryptonanus could be expected by surveys in extralimital savanna landscapes, specially by pitfall trapping (^{Voss et al. 2005}; ^{Dias et al. 2015}). In the Central Atlantic Forest Corridor (^{Brazil 2006}; ^{Câmara & Galindo-Leal 2009}), small mammal surveys using pitfall traps are still scarce (^{Bovendorp et al. 2017}), despite their obvious and demonstrated efficiency to trap elusive species (^{Rocha et al. 2015}). Protocols that include this and other alternative kinds of trap methods could contribute largely to fill in many gaps on species distribution, as showed here. The records of C. agricolai presented here expanded the geographic distribution of the species approximately 430 km eastward (UFES-MAM 3075) and 360 km southward (the recognition of MZUSP 35409 as C. agricolai), throughout an area of 324 452 km² (hatched area; Fig. 5). This corresponds to a 22% of increase in the extent of occurrence previously known for the species and indicates that C. agricolai range is significantly larger, extending over a substantial portion of the Atlantic Forest ecoregion.

Fig. 4 Phylogenetic inference of four species of Cryptonanus based on Bayesian Analysis of 801 bp of mitochondrial Cytochrome b sequences, performed using the TN64 + G model. Values at the nodes refer to Bayesian posterior probabilities. In bold, the specimen reported in the present study (UFES-MAM 3075). Gracilinanus peruanus and Thylamys citellus were used as outgroups. 

Fig. 5 Updated extent of occurrence and range expansion of Cryptonanus agricolai to the Atlantic Forest, according to new records.