Craniometrical differentiation of gray brocket deer species from Brazil

The genus Mazama (brocket deer) constitutes successful adaptive radiation, with a wide distribution in the Neotropical region. However, the taxonomy and systematics of its species and subspecies is still controversial. The objective of this contribution was to carry out a comparative craniometric study of specimens deposited in Brazilian museums of M. gouazoubira (Mg) and M. nemorivaga (Mn), in order to characterize these cryptic species. We performed 36 measures on 87 skulls of adult specimens (Mg = 32 females and 27 males; Mn = 14 females and 14 males). The sample was compared by analysis of variance and multivariate analysis of principal components and discriminant. In most cranial measurements, males and females of M. gouazoubira were 5% larger, with the exception of 3 measures: premolar-prosthion, basifacial axis, and least breadth between the orbits, that were larger in males and females of M. nemorivaga. This study showed significant differences between the two taxa that would validate the distinction of both species. To examine in more detail the dynamics of the variation of these two taxa it is necessary to increase the sample size in order to analyse them in a geographical and genetic context.

M. nemorivaga was firstly described by Cuvier in 1817, from a specimen from Cayenne, in French Guiana.Subsequently, Allen (1915) recognized this taxon within his gray brocket deer species group and restricted the geographical range of M. nemorivaga for individuals from the region of French and British Guyanas.Miranda-Ribeiro (1919), in his extensive taxonomic study of Brazilian deer, described a small species of gray brocket deer in the state of Rondonia and named it M. rondoni.However, Cabrera & Yepes (1960) and Cabrera (1960) not recognized this species, proposing it as a synonym of M. gouazoubira superciliaris.In 1987, Czernay updates the taxonomy of Mazama and recognized six species: M. americana, M. gouazoubira, M. rufina, M. nana, M. chunyi and M. bricenii.

Mazama gouazoubira
The "gray brocket or brown brocket deer" (Mazama gouazoubira) is the most abundant deer in South America (Duarte, 1996), being classified by the IUCN Red List (2009) as a least concern.However, in Brazil, the regional classification of the species is variable, being considered as threatened in some states (e.g., Rio de Janeiro; Bergallo et al., 2000).
It is a small to medium sized species, with an average weight ranging from 11-25 kg and an average height of 50 cm (Duarte, 1998;Rossi, 2000).The general coloration of individuals varies from dark gray to reddish brown (Pinder & Leeuwenberg, 1997).The submandibular and periophthalmic regions are lighter, light cream coloured.Most individuals have a white spot above the eyes, which characterizes the species and below the tail.They have large and rounded ears and non-branched antlers, characteristic of all species of the genus Mazama (Duarte, 1996;Duarte, 1998).The antlers are present only in males and have the form of a spike of 56.3 to 117.3 mm in length.They are conical or laterally compressed, straight or slightly curved medially, and parallel or slightly divergent from each other, and positioned at a sharp angle with respect to the dorsal surface of the skull.The burr contains elongated and more or less deformed tubercles of different sizes (Black-Decima et al., 2010).
The skull has a long rostrum that occupies 47-51 % of the total length of the skull (Fig. 1).The premaxillaries are punctually united to nasals or completely separated from them by an anterior projected bar of the maxillaries.The lacrymal pit is shallow or moderately deep.The orbit is rounded or more or less squared, with a vertical diameter between 25.6 and 31.9 mm.The auditory bulla is small but wide, with its greatest width ranging from 10.4 to 13.3 mm; its ventral surface is adorned with crests or projections.The dental formula is 0/3i 0/1c 3/3pm 3/3m x 2 = 32.Small upper canines can be seen in some adult specimens.The length of the maxillary tooth row ranges from 45.2 to 55.9 mm (Black-Decima et al., 2010).
This species occupies dense and continuous forests to open savannahs with small patches of forests.It is an easily found species, occupying the Cerrado to the south of the Amazon River up to the north of Argentina and Uruguay (Ávila-Pires, 1959;Duarte, 1996;González & Elizondo, 2010).This diversity of ecosystems and seasonal regimes can influence several aspects of their ecology (selection, habitat, diet, density and reproduction), in addition to their social behaviour, causing certain variations within and between populations.This species is relatively resilient; being capable to live in agroecosystems intermixed with small forest fragments (Pinder & Leeuwenberg, 1997;Rodrigues et al., 2014Rodrigues et al., , 2017)).In the Amazonian region, it is probably replaced by Mazama nemorivaga, a hypothesis that has not been fully tested (Duarte, 1996;Rossi, 2000;Black-Decima et al., 2010;Rossi et al., 2010)

Mazama nemorivaga
The "Amazonian brown brocket deer" (Mazama nemorivaga) has a wide distribution and inhabit in many protected areas of Amazonian basin forests.Although this species is threatened by loss of habitat and local hunting (mainly for subsistence), it had been classified by the IUCN Red List (2016) as a least concern.However, we infer that it must be suffering a certain degree of threat since its habitat is being extensively altered by deforestation and agricultural activities, in addition to the hunting pressure that is the object (Rossi & Duarte, 2016).
The Amazonian brown brocket deer is a small deer, about 15 kg average weight and 48 cm height (Rossi et al., 2010).The pelage of the head exhibits a dark brown frontal tuft, tipped with black or not; small and barely defined buff spots on the edges of the upper lip; a whitish and not sharply defined spot on the anterior border of the chin, followed by dark buff and barely defined bands; and orange-brown inferior and superior orbital bands, not sharply defined, with yellow hairs mainly in the lateral areas.A diagnostic characteristic is the absence of the tarsal tuft.The adult males have spike-like antlers, ranging from 27.7 to 108.0 mm in length.Antlers are conical, usually slender, straight or slightly curved medially, and quite bent in relation to the dorsal surface of the skull; they may be parallel or slightly divergent between each other.The burr contains small spherical or elongated and more or less deformed tubercles (Rossi et al., 2010).
The skull has a long rostrum that corresponds to 47-52 % of the total length of the skull (Fig. 2).The premaxillaries are usually separated from the nasals by an anterior projected bar of the maxillaries.The depth of the lachrymal pit varies among individuals; in some it is shallow, while in others it is deep.The orbit is large, with vertical diameter between 28.0 to 34.0 mm, and may be rounded or more or less squared in shape.The auditory bulla is small and narrow, with its greatest width ranging from 8.2 to 11.4 mm; its ventral surface is adorned with crests and small projections.Small upper canines are rarely seen in adult specimens.The dental formula is 0/3i 0/1c 3/3pm 3/3m x 2 = 32.The length of the maxillary toothrow ranges from 46.2 to 53.2 mm.The length of the maxillary toothrow ranges from 46.2 to 53.2 mm (Rossi et al., 2010).
Its geographical range is extended from Brazil, French Guiana, Suriname, Guyana, Venezuela, Colombia, Panama (Isla San José), Ecuador, Peru, and probably northern Bolivia (Rossi & Duarte, 2008).In Brazil it is present in the states of Amazonas, Pará, Rondônia, Amapá, north of Mato Grosso, and northwest of Maranhão (Rossi, 2000).This species seems to be restricted to the Dense Ombrophilous Forests of the Amazonian Region and particularly in the State of Maranhão.It may also occur in the transitional areas between the Cerrado and Seasonal Forest and possibly in the Deciduous Seasonal Forest (Rossi, 2000;Rossi et al., 2010).Duarte (1998), based on significant morphological and karyotypic differences, reconsidered the existence of M. rondoni as a separate species from M. gouazoubira, as originally was proposed by Miranda-Ribeiro (1919).Later, Rossi (2000) performed a taxonomic revision based on morphological analysis of specimens from seven northern Brazilian states and considered that the species M. rondoni is a synonym of M. nemorivaga.Currently, M. nemorivaga is recognized as a valid species, occupying the Amazonian region (Duarte et al., 2008;Duarte et al., 2010;Duarte & González, 2010;Fiorillo et al., 2013;Rossi & Duarte, 2016).

Comparative craniometrical study
Previous craniometrical studies on South American deer were reported by Wemmer &Wilson (1987), andMerino et al. (2005), which analysed interspecific variation and phylogenetic relationship.Skull morphometry can be a useful tool for analysing intraspecific variation as has been reported by Cabrera (1943), González et al. (2002), andRossi (2000) from several Brazilian Mazama species.The craniometric study of Rossi (2000), that included an extensive sampling of brocket deer species failed to distinguish two different species of red brockets deer (M.americana and M. bororo) but discriminated by canonical analysis the two gray brockets (M.gouazoubira and M nemorivaga).As a part of an integrative project, Carranza et al. (2017) found evidence revealing weak premating behaviour isolation between M. gouazoubira and M nemorivaga, being more common at intraspecific interactions and suggesting discrimination between species.
Cursino & Duarte (2017) identified significant differences on the morphology of the spermatozoa and the ejaculate of both species.One of the characteristics that most clearly differentiated between them was the colour of the ejaculate (white for M. gouazoubira and reddish for M. nemorivaga), and sperm head dimensions.
Our aim was to carry out a comparative craniometric study between M. gouazoubira y M. nemorivaga, based mostly on specimens depos-deposited in Brazilian museum's collections, in order to analyse the intra and interspecific variation between these two cryptic species.Our purpose was to determine if there are differential cranial characteristics between these two taxa, to initiate a morphological database that allows examining in more detail the dynamics of the geographic and genetic variation to determine priority conservation units.

Collections analysed and measurements performed
We reviewed the mammal collections of the following Brazilian museums: National Museum of Rio de Janeiro (MNRJ), Museum of Zoology of the University of São Paulo (MZUSP), Museum of Biology Professor Mello Leitão (MBML), We measured 106 skulls, but we only included in the comparative analyses 87, because they were adults and almost complete specimens of M. gouazoubira (Mg: 32 females and 27 males) and M. nemorivaga (Mn: 14 females and 14 males).Individual specimens were classified mostly based on its geographical origin and overall morphology, following Miranda Ribeiro (1919).
We took 36 measurements using digital cali-caliper (precision 0.05 mm) following the criteria of the standard measurements for cervids of Von den Driesch (1976; see Appendix I).We also con-sidered the data about age, sex and geographical location.(Appendix II).We plotted coordinates on a map using ArcGIS 10.1 (ESRI 2011) based on geographical information from each specimen label (Fig. 3).

Statistical analysis
We analysed the morphological variation of the sample, considering separately the effects of sexual dimorphism and taxonomic information.We classified the sample by sex and species, obtaining the averages, standard deviation and the variance of each variable.First, we conducted a Cluster analysis using the linkage model and Euclidean distances.To avoid the effect of sexual dimorphism, we analysed adult males and females separately.
We performed a factorial analysis in order to have an overview of the variation to simplify and reduce the 36 variables in 5 Factors.The factors were extracted using the method of the principal components (correlation matrix).Main components were log-transformed and rotated with the standard VARIMAX method.The projection of the individuals on the most informative fac- tors was conducted by graphic means to detect the possible groups of specimens belonging to the same species and sex.
Analysis of variance (ANOVA) was used to test whether the factors presented statistically significant differences in relation to the sex and species (Sokal & Rohlf, 1969).The 36 original variables were also analysed using 2-way ANOVA.The advantage of the ANOVA is that this technique allows us to test if there are significant differences between groups and classified according to a specific criterion (e.g., species) after eliminating variation that could be attributed to other classification criteria (e.g., sex).The employed software to carry out all analyses was Statistica (StatSoft 1995).
We performed a stepwise discriminant function to analyse the patterns of differentiation and to classify individuals by sex and species (Sneath & Sokal, 1973;Sokal & Rohlf, 1969).Using the Statistica software, we build a step wise step-by-step model of discrimination to review and evaluate all variables that contribute most to the discrimination between sexes and species (using the same probability model for each class of classification).Next, that variable will then be included in the model, and Statistica software will proceed to the next step.

RESULTS AND DISCUSSION
Our results showed that gray brocket deer (M.gouazoubira) exhibited a wider intraspecific variation than the Amazonian brown brocket deer (M.nemorivaga).The averages of 36 measures are summarized in Appendix III, discriminated by sex and species.A wide intraspecific variability is observed as shown by the wide range of standard deviation of most measurements.In addition, most specimens belonging to M. gouazoubira were larger in most of the variables, being approximately 5% larger in both sexes.The exceptions were 3 measures: premolar-prosthion (PREPO), basifacial axis (BACR), and least breadth between the orbits (LBBO), which were larger in M. nemorivaga in both sexes.These results are indicative of differences in the skull shape between M. gouazoubira and M. nemorivaga.
When we analysed the behaviour of the variables using the Euclidean distance for performing a cluster analysis, we observed a diffuse morphometric differentiation pattern.To avoid the effect of sexual dimorphism, we considered separately the samples of adult males and females, retrieving not clear cluster aggrupation in any case.This would be explained by the wide intraspecific variability, also seen in standard deviation values (Appendix III).Alternatively, this situation would be a reflect of the large intra and interspecific variability of both species, that also have been evidenced by cytogenetic and molecular data (Duarte et al., 2008;González et al., 2010;Fiorillo et al., 2013).
Because cluster analysis was not efficient to discriminate between the two taxa, we proceed to perform a Factorial Analysis, that allows us to simplify and summarize the 36 morphometric variables.As a result, we reduced them to 5 factors that account for 70% of the total variance of cranial measurements (Table 1).When individuals were discriminated by sex, we observed that   sexual dimorphism exists in most of the variables (Fig. 4 and 5), as often occurred in other cervid species, including Neotropical ones, such as the pampas deer (Ozotoceros bezoarticus; González et al., 2002).The factor 1 was associated with 10 variables; while five represent the skull length, the other five were related to the width in the posterior part of the cranium.However, it should be noted that the skull length (factor 1) did not differ significantly between sexes and species.We found that values for females of M. gouazoubira, associated to this factor, were greater than those of the males.The factor 2 was also associated with seven variables (Table I detailed), being three of them related to cranial length and to the length of the dental series.These variables discriminated significantly among species (Fig. 4).As this factor was associated with the dental series measures would be indicating that the two species were adapted to different types of forest habitats and food, as had already been emphasized by Duarte (1998) and Rossi (2000).In contrast, factor 3 was correlated with four of the measures that define the length of the face (Table 1).The factor 4 was associated with the unique three measures that were larger in the M. nemorivaga specimens (i.e., PREPRO, BACR and LBBO).In addition, it was the only one with highly significant differences between both sexes and species (Fig. 5).Finally, factor 5 grouped the variables that exhibited low correlation than 70% and only discriminated between sexes.
In order to investigate whether the observed differentiation was statistically significant, we performed an ANOVA using the 36 original variables and the five factors obtained by grouping individuals according to sex and species (Table 2  and 3).Significant differences were observed in the two data sets (Table 2 and 3).Of the 36 variables analysed, 15 did not show significant differences between species and sexes.The 21 variables that discriminated between the two species and that exhibited high significant F values (p< 0.05) were those related to the length of the skull and the shape of the face.The ANOVA performed with the five extracted factors also showed highly significant values in 3 of them (Table 3).The variables associated with factors 1 and 3 did not show significant differences.Factors 2, 4 and 5 showed statistically significant differences (Table 3), being a clear indicator that the individuals come from two morphometrically different species.
Subsequently, we performed a discriminant function analysis using 36 skull variables under a step wise forward model to classify the sample according to sex and species.When the en-

CONCLUSIONS
This study supports previous findings reported by Miranda Ribeiro (1919) and Rossi (2000) that there are two craniometrically distinct species of gray brockets in Brazil: M. gouazoubira and M. nemorivaga.However, a more detailed analysis of geographical patterns of morphological differentiation need to be conducted in the future, using a larger sample than those presented here.It would be important to extend the sampling to collections outside Brazil for further examination of the morphological variation and taxonomy within gray brocket deer species, in order to determine priority conservation units among them.

ACKNOWLEDGMENTS
We acknowledge to Guillermo Cassini and Pablo Teta for inviting us to participate of this special volume to honour Dr. José Yepes , a prominent Argentinean mammologist that highlight the importance of considering the geographical distribution for interpretations of modern systematics (García, 2018).Also, to two anonymous reviewers who performed valuable editorial comments that greatly improve this manuscript.We are grateful to the following agencies that funded our research: Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP), Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq), Comision Sectorial de Investigación Cientifica de Universidad de la República (CSIC-UDELAR), Programa de Desarrollo de Ciencias Básicas (PEDECIBA), Agencia Nacional de Investigación e Innovación (ANII), and Lóreal-Unesco from Uruguay.We also want to thank to all curators for authorized to museum  tire sample was considered, the gender discrimination found was significant (F = 10.86,p < 0.0001), confirming that there is a sexual sharp dimorphism in both species (Table 4).When we considered the whole sample (not separated by sex), we found significant differences between the 2 species (F = 6.9201, p < 0.0001), with all the individuals correctly classified (Table 5).Our re-Our results were also corroborated in a limited sample of gray brockets, that was further analysed using others evidences, such as cytogenetics and biometrics (Rossi, 2000).Duarte et al. (2008) analysed a sample of Amazonian brown brocket deer using molecular markers and suggested the existence of two genetically differentiated groups: one would be occurring in the western region of Amazonia and the other in the east.However, it is necessary to increase the sample size to confirm this hypothesis and to evaluate if this genetic difference is

Fig. 1 .
Fig. 1.Skull of an adult female (NPC 003) and adult male (NPC014) of Mazama gouazoubira.A) Dorsal view of adult female, B) Lateral view of adult female, C) Dorsal view of adult male, D) Lateral view of adult male.

Fig. 2 .
Fig. 2. Skull of an adult female (NPC 010) and adult male (NPC053) of Mazama nemorivaga.A) Dorsal view of adult female, B) Lateral view of adult female, C) Dorsal view of adult male, D) Lateral view of adult male.

Fig. 3 .
Fig. 3. Individuals measured and mapped with ArcGIS 10.1 according with the information recorded on the museum collections detailed in Appendix II.
Mastozoological col-collections: Museum of Rio de Janeiro (MNRJ), Museum of Zoology of the University of São Paulo (MZUSP), Museum of Biology Professor Mello Leitão (MBML), Museum of Zoology of the Pontifical Catholic University of Rio Grande do Sul (PUC-RS), Museum of Zoology of the Federal University of Pernambuco (UFPE), Emilio Goeldi Museum Paraense, Museum of Natural History Capão da Imbuia (MHNCI), Museum of Zoology of the Federal University of Paraíba (UFPB) and Museum of the Deer Research and Conservation Center of UNESP (NUPECCE).

TABLE 1 .
Factors extracted from 36 cranial variables, using the principal components method and rotated to normalize them with the procedure VARIMAX (see Appendix I by the abbreviation of the variable names).The values correspond to the correlation coefficients between the variables and each factor.The red colour indicates if the variables that have a strong association (| r |> 0.7) with the factor.

Table 2 .
Analysis of Variance of 36 skull measurements between two gray brocketspecies.For abbreviations see the Appendix I.

Table 5 .
Classification Matrix (Rows: Observed Classification; Columns: Expected Classification) Summary statistics of the analysed variables discriminated by species, sex and all groups pooled together.Abbreviations: n= number of specimens; s.d.= standard deviation; measures are detailed in Appendix I.