PHYLOGEOGRAPHY, POPULATION STRUCTURE, AND IMPLICATIONS FOR CONSERVATION OF WHITE-TAILED DEER (ODOCOILEUS VIRGINIANUS) IN VENEZUELA

Abstract White-tailed deer, Odocoileus virginianus, is the most important game species in Venezuela. Some populations are currently threatened by overhunting and habitat loss, making it necessary for implementation of conservation programs. In this study, we employed molecular phylogenetics and population genetics principles to identify conservation units of Venezuelan populations of this species and to provide recommendations for its management. We analyzed DNA sequences—730 base pairs of the mitochondrial control region—in 26 individuals sampled from the 3 subspecies present in Venezuela. Results revealed moderate levels of genetic polymorphism. In addition, evidence of significant population structure was found. Phylogeographic analyses showed 4 lineages with the nominal subspecies O. v. gymnotis appearing to be polyphyletic. A remarkable divergence among haplotypes from Venezuela and North America was revealed in phylogenetic analyses, the former comprising a monophyletic group. The observed divergence among haplotypes from Venezuelan and North American populations was, in most cases, higher than that observed among the latter and haplotypes of O. hemionus (black-tailed deer). This result suggests that Venezuelan white-tailed deer could be considered an evolutionarily significant unit. We interpreted the results obtained within the context of climatic changes since Late Pleistocene. In addition to ecological and morphological evidence, our data suggest that O. v. margaritae and O. v. goudotii populations, the 2 subspecies considered endangered, are clearly differentiated and should be recognized as geminate evolutionary units and protected as distinct groups, even though there is no clear support for elevating these subspecies to species rank, as proposed recently.

White-tailed deer, Odocoileus virginianus, is the most important game species in Venezuela. Some populations are currently threatened by overhunting and habitat loss, making it necessary for implementation of conservation programs. In this study, we employed molecular phylogenetics and population genetics principles to identify conservation units of Venezuelan populations of this species and to provide recommendations for its management. We analyzed DNA sequences-730 base pairs of the mitochondrial control region-in 26 individuals sampled from the 3 subspecies present in Venezuela. Results revealed moderate levels of genetic polymorphism. In addition, evidence of significant population structure was found. Phylogeographic analyses showed 4 lineages with the nominal subspecies O. v. gymnotis appearing to be polyphyletic. A remarkable divergence among haplotypes from Venezuela and North America was revealed in phylogenetic analyses, the former comprising a monophyletic group. The observed divergence among haplotypes from Venezuelan and North American populations was, in most cases, higher than that observed among the latter and haplotypes of O. hemionus (black-tailed deer). This result suggests that Venezuelan white-tailed deer could be considered an evolutionarily significant unit. We interpreted the results obtained within the context of climatic changes since Late Pleistocene. In addition to ecological and morphological evidence, our data suggest that O. v. margaritae and O. v. goudotii populations, the 2 subspecies considered endangered, are clearly differentiated and should be recognized as geminate evolutionary units and protected as distinct groups, even though there is no clear support for elevating these subspecies to species rank, as proposed recently.
Viana 1991) with the 3 recognized subspecies, O. v. goudotii in the Andes (Mérida state), O. v. gymnotis throughout most of the country, and O. v. margaritae endemic to Margarita Island (Danields 1991;Rodríguez and Rojas-Suárez 1995;Smith 1991). Many populations currently are considered threatened, especially those belonging to the subspecies O. v. goudotii and O. v. margaritae, which are listed in the Red Book of the Venezuelan Fauna (Rodríguez and Rojas-Suárez 1995). Even though hunting of this species in Venezuela has been prohibited since 1979 (Carrillo-Batalla 1991), furtive hunting, in addition to habitat loss, still represents the main threat for whitetailed deer populations, especially those from Margarita Island and the Andes (Danields 1991; Rodríguez and Rojas-Suárez 1995).
Recently, it has been proposed that the Venezuelan subspecies of white-tailed deer should be considered separate species (Molina and Molinari 1999). The conclusion was based on morphometric analyses of craniomandibular traits. These results require further evaluation, not only to improve our understanding of taxonomic status of subspecies, but also to provide fundamental information for policymakers working on management and conservation issues (Avise 1989;Haig 1998).
The use of molecular markers to estimate population genetics parameters of relevance to conservation biology is becoming more common (Avise 1995;Frankham 1995). Perhaps 1 of the major contributions of molecular evolutionary biology to conservation biology has been in defining conservation units and in the inference of population processes in evolutionary time (Moritz 1995).
Conservation units define genetically divergent groups at several levels, providing alternative biological criteria for conservation (Bowen 1998;Moritz 1994aMoritz , 1994b. Three conservation units are recognized: evolutionarily significant units, management units, and geminate evolutionary units. Evolutionarily significant units are historically isolated sets of populations that are reciprocally monophyletic for mtDNA alleles and significantly differentiated at the nuclear level (Moritz 1994b;Ryder 1986). Populations with significant divergence of allele frequencies at nuclear or mitochondrial loci, regardless of phylogenetic distinctiveness of the alleles, are considered management units (Moritz 1994b). Geminate evolutionary units are populations that show differentiation at morphological, behavioral, or biogeographical levels, even though they do not have genetic divergence at neutral loci (Bowen 1998).
Mitochondrial DNA (mtDNA) has intrinsic characteristics (e.g., it is maternally inherited) that make it an ideal molecular target for studying evolutionary processes in populations (Avise 2000;Avise et al. 1987;Moritz 1994a). The mitochondrial genome has a relatively short noncoding region known as the control region or d-loop. The d-loop has major regulatory elements for mitochondrial replication and expression (Avise et al. 1987;Saccone et al. 1991;Sbisà et al. 1997). The high substitution rates observed in peripheral domains of the dloop have made it a widely used target in evolutionary studies (Avise 2000;Douzery and Randi 1997;Sbisà et al. 1997). Additionally, phylogenies based on mtDNA provide valuable information on population history and suggest boundaries of genetically divergent groups (Moritz 1994a(Moritz , 1995. Applications of mitochondrial DNA analysis in conservation have been reviewed by Moritz (1994a).
The white-tailed deer is perhaps one of the most studied of deer species (Eisenberg 1989); however, most of our knowledge about this species comes from North American populations. Particularly, the analysis of mtDNA has been employed to study historical biogeography in white-tailed deer from the southeastern United States (Ellsworth et al. 1994a), to asses spatial heterogeneity among populations of white-tailed deer and mule deer Odocoileus hemionus  Venezuela, after Eisenberg (1989) and Rodríguez and Rojas-Suárez (1995), and approximate locations of specimens sampled (black circles  , to evaluate genetic variation in restored populations of whitetailed deer in the southeastern United States (Ellsworth et al. 1994b), to study female philopatry and genetic heterogeneity in white-tailed deer (Purdue et al. 2000), to evaluate introgressive hybridization between white-tailed deer and mule deer (Carr and Hughes 1993;Cathey et al. 1998), and to asses phylogenetic relationships among Cervidae (Cronin 1991).
The specific aims of this study are to assess genetic diversity and differentiation at the mitochondrial level of the 3 subspecies of white-tailed in Venezuela, as well as to study their phylogeographic relationships in order to identify conservation units. This work represents the 1st molecular genetic study of white-tailed deer in South America.

MATERIALS AND METHODS
DNA sequencing.-Twenty-six tissue samples were taken from deer of the 3 subspecies of O. virginianus in Venezuela,including 16 O. v. gymnotis,4 O. v. goudotii,and 6 O. v. margaritae ( Fig. 1; Appendix I). These samples consisted of hair or blood of live animals kept in captivity by zoos or as pets, muscle of carcasses collected opportunistically in the field, and skin or bone of museum specimens. Blood samples were collected in Vacutainer tubes with edatic acid and preserved in 2 volumes of 80% ethanol; each tube was covered with aluminum foil and stored at Ϫ20ЊC. Hair, muscle, and skin samples were preserved in opaque containers in 100% ethanol and stored at Ϫ20ЊC. Fresh and preserved tissues were taken only from specimens of known localities.
DNA was extracted from blood samples following the standard protocol with phenol-chloroform (Maniatis et al. 1982). DNA was extracted from skin, muscle, hair, and bone samples following the protocol of the silica method (Höss and Pääbo 1993) modified by E. Randi (Hagelberg and Sykes 1989). In each amplification, a PCR negative control was included (PCR preparation without DNA). The 1st PCR amplification used a pair of external primers that matched tRNA Pro and t-RNA Phe genes, which flank the mtDNA control region of vertebrates (Douzery and Randi 1997;Saccone et al. 1987Saccone et al. , 1991. Primers used in this amplification were AL3237 (5Ј CGT CAG TCT CAC CAT CAA CCC CCA AAG C 3Ј), which matched the DNA light strand (tRNA Pro ), and AL3238 (5Ј GGG AGA CTC ATC TAG GCA TTT TCA GTG 3Ј), which matched the DNA heavy strand (tRNA Phe -Douzery and Randi 1997).
Downloaded from https://academic.oup.com/jmammal/article-abstract/84/4/1300/892004 by guest on 26 July 2018 Amplification cycles were performed in a programmable thermal cycler (model PTC-100, MJ Research, Inc., Waltham, Massachusetts). Each amplification consisted of an initial denaturation phase at 94ЊC for 2 min, followed by 30-35 cycles of denaturation at 94ЊC for 30 s, annealing at 48ЊC for 45 s, extension at 72ЊC for 1 min, and a final extension at 72ЊC for 10 min.
The nested PCR was carried out following the same protocol as above, except in this case, the DNA template was an aliquot of ϳ50-100 ng from the 1st PCR. Internal primers in the nested PCR were AL3448 (5Ј GCT CCA TAA AAT CCA AGA GC 3Ј), which matches the beginning of the control region light strand of whitetailed deer from North America, and AL3453 (5Ј GCG CCT ATA TAC TTA CCT CGC CC 3Ј), which matches the heavy strand at the end of this region.
Polymerase chain reaction products were purified directly with the product kits Concert Rapid PCR purification system (Gibco BRL, Carlsbad, California) and Wizard PCR preparation DNA purification system (Promega, Madison, Wisconsin) according to the manufacturers' instructions. In other cases, PCR products were purified in 1% agarose TBE 1ϫ gel and extracted from the agarose with gene elute agarose spin column (Sigma, St. Louis, Missouri) according to the manufacturer's instructions.
Purified PCR products were sequenced directly in the Centro de Secuenciación y Análisis de Á cidos Nucléicos (CeSAAN, Instituto Venezolano de Investigaciones Científicas, Venezuela). Sequencing reactions were performed as described in Applied Biosystems BigDye Terminator kit (Perkin Elmer, Wellesley, Massachusetts) on a programmable thermal cycler (PE Applied Biosystems GeneAmp model 9700, Perkin Elmer, Wellesley, Massachusetts). Sequencing was performed on an automated sequencer (model PE ABI377, Perkin Elmer, Wellesley, Massachusetts). Chromatograms were proofed and sequences were assembled visually.
Population structure and phylogenetic analyses.-Sequences were aligned with Clustal X software (Thompson et al. 1997) with the use of orthologous sequences of other cervids (Appendix I) as references for the alignment. Aligned sequences were then edited manually to fit them to the same length.
To examine population structure, haplotype samples were divided into 3 populations accord-ing to distribution of the 3 recognized subspecies of white-tailed deer (goudotii, gymnotis, and margaritae) in Venezuela (Fig. 1). Population polymorphism was estimated with 2 genetic diversity measures: nucleotide diversity (⌸), and haplotype diversity (h). ⌸ is defined as the average number of nucleotide differences per site between 2 sequences of DNA, whereas h is a measure of the frequency of haplotypic variants in a population (Nei 1987). These parameters were estimated following Nei (1987) with DNAsp 3.5 software (Rozas and Rozas 1999).
Genetic population structure was assessed by an analysis of molecular variance (AMOVA), which implements the approach of Weir and Cockerham's (1984) analysis of variance in which the ⌽ ST statistic is analogous to Wright's F ST (Excoffier et al. 1992;González et al. 1998). From this analysis, a measure of population differentiation and gene flow among populations was obtained. The significance of ⌽ ST was assessed through a permutation test with 2,000 repetitions.
Gene flow (M) was approximated to the Nm product, under the assumptions of Wright's (1969) island model, where N is the effective number of females and m is the migration rate; this expression represents the effective number of migrant females. This parameter was estimated from F ST ϭ 1/(1 ϩ 2Nm), which is a corrected Wright's equation for extranuclear genomes (Takahata and Palumbi 1985). Analyses of genetic population structure and gene flow were performed with the program Arlequin version 2000 (S. Schneider et al., in litt.). In addition, genetic distances between and within populations were estimated to evaluate genetic differentiation between populations (Nei 1987). These distances were estimated following Kimura's 2-parameter model (K2P- Kimura 1980) with MEGA version 2.0 (Kumar et al. 2001).
Phylogenetic relationships were inferred by both neighbor-joining and maximum parsimony analyses. Both analyses were conducted using MEGA version 2.0 (Kumar et al. 2001). The K2P model was used in the neighbor-joining analysis. In the maximum parsimony analysis, a close neighbor interchange algorithm with a search level of 3 was used as the heuristic tree search method with 100 replications of random addition trees. In both cases, haplotype regions with gaps, missing data sites, or both were not included in analyses. Clade support was as-  sessed by 2,000 bootstrap replicates (Felsenstein 1985). Haplotypes of additional cervid species were employed as outgroups (Appendix I).
Sequence data have been submitted to GenBank (accession numbers AF421829-AF421854; Appendix I). The alignment of sequences is available from the authors by request.

RESULTS
Lengths of PCR products ranged between 730 and 887 base pairs (bp). After all gaps were excised manually, all sequences consisted of 730 bp. The 26 sequences included 23 haplotypes with 72 polymorphic sites. No haplotypes were shared by different subspecies. Estimated h values were around 1. The ⌸ values ranged between 0.029 and 0.005, with goudotii being the population with the lowest value (Table 1).
Population structure and gene flow.-Results from AMOVA revealed significant differentiation among gymnotis, goudotii, and margaritae populations (Table 2), although intrapopulation variance was Ͼ70% of the observed total variance. Inter-and intrapopulation genetic distances (Table 3) indicate that intrapopulation genetic dis-tance in the gymnotis population is as high as the interpopulation genetic distance between this population and any of the others (goudotii or margaritae). Thus, an interpopulation net genetic distance, corrected by intrapopulation distances, was calculated ( Table 3). The complete matrix of pairwise genetic distances is available from the authors by request.
Pairwise values of ⌽ ST between populations also were significant (P Ͻ 0.01). From these, values of M between populations were obtained (Table 4). Both goudotii and margaritae populations seem highly differentiated (⌽ ST ϭ 0.61570), with little gene flow between them (Ͻ1 female in each of 3 generation). Gene flow values between gymnotis and goudotii and between gymnotis and margaritae were Ͼ1 female per generation in both cases. It is important to note that the Nm product as an indirect estimator of M has been subject to serious criticisms (Whitlock and McCauley 1999); however, this expression could provide a measure of relative strength of genetic drift and migration in a population (Neigel 1997;Slatkin 1987   Phylogeographic relationships among haplotypes.-Both maximum parsimony and neighbor-joining analyses of the 23 haplotypes of white-tailed deer from Venezuela showed similar patterns (Figs. 2 and 3), with the exception that clades 3 and 4 are interchanged. However, regarding only phylogeographic structure, it is possible to distinguish 4 lineages, identified as clades 1-4, respectively.
The 2nd clade clustered all haplotypes from Margarita (subspecies O. v. margaritae) and the haplotype from Sucre (subspecies O. v. gymnotis), which are geographically proximate (Fig. 1). Despite low bootstrap support for this clade (Ͻ50% in the maximum parsimony analysis), we find it repeatedly in our analyses, suggesting the distinctiveness of Margarita Island's deer population.
The 3rd clade contains haplotypes from Guárico and Apure, with bootstrap support Ͼ95%. These haplotypes, belonging to the subspecies O. v. gymnotis, are located geographically in the Venezuelan llanos (Fig.  1). However, this clade is poorly represented in our sample.
The 4th clade is represented by haplotypes from Miranda, Zulia, Falcón (Ovi 8 and 30), and Anzoátegui, with bootstrap support Ͼ75%. All of these haplotypes belong to the subspecies O. v. gymnotis and are distributed along the Venezuelan coastline (Fig. 1).
Phylogenetic analyses conducted in this work suggest that white-tailed deer populations from Venezuela form a monophyletic group supported by bootstrap values Ͼ83%. The estimated genetic distance among white-tailed deer from North America and Venezuela ranged from 5.6% to 7.7%. These genetic distances were, in most cases, greater than those observed between haplotypes of white-tailed deer from North America and black-tailed deer, which ranged from 3.6% to 9.3%.

DISCUSSION
Population structure and phylogeography.-Sample size is an important limitation of our study, especially for O. v. goudotii and O. v. margaritae populations. Samples from these subspecies were difficult to obtain. Nevertheless, results obtained through analysis of mtDNA remain useful, even under reduced sample sizes, a common situation when working with threatened species, because differences between populations can be detected readily at the mitochondrial level (Moritz 1994a).
Our results on haplotype and nucleotide diversity of d-loop sequences of the whitetailed deer in Venezuela suggest moderate levels of genetic variability. Observed val- ues of h were consistently high, whereas those of ⌸ were variable. The lowest observed value of ⌸ was in the goudotii population, whereas the values for other populations were similar to those reported for Pampas deer Ozotoceros bezoarticus (González et al. 1998) and higher than those for roe deer Capreolus capreolus (Randi et al. 1998;Wiehler and Tiedemann 1998) and C. pygargus (Randi et al. 1998). Values of h obtained in this study were much higher than those reported for North American white-tailed deer (Ellsworth et al. 1994b;Purdue et al. 2000).
The result of the AMOVA showed that the intrapopulation component of molecular variance was much greater than the interpopulation component, even though populations are significantly structured (⌽ ST ϭ 0.61570, P K 0.0001; Table 2). Similar results have been reported for Cervus nippon (Nagata et al. 1998), Capreolus capreolus (Wiehler and Tiedemann 1998), and human populations (Excoffier et al. 1992). Excof-  (1992) pointed out that this observation could be an artifact resulting from an arbitrary or inadequate selection of populations, the occurrence of isolation by distance in the studied sample, or the existence of zones with low gene flow. Results obtained in this work regarding genetic distances and phylogenies suggest that the first 2 alternatives could explain the observed result. The large genetic distances observed among haplotypes of the gymnotis population and its broad geographic distribution could explain the component of intrapopulation variance in AMOVA. However, future studies must confirm the possibility of isolation by distance and the presence of localized historical barriers for gene flow as potential explanations for the observed patterns.
The most important results obtained in this work indicate that 4 lineages define gymnotis populations as polyphyletic with respect to goudotii and margaritae; that populations are significantly structured (Table 2); that levels of genetic polymorphism are moderate when considering the whole species distribution in Venezuela, but low in the goudotii population in particular; and that white-tailed deer haplotypes from Venezuela and North America diverge remarkably. These observations can be integrated and explained within a historical context.
The last glacial maximum (18 ϫ 10 3 years ago-CLIMAP Project Members 1976) in Venezuela began 19-16 ϫ 10 3 years ago and ended approximately 13 ϫ 10 3 years ago (Rull 1998;Schubert 1974). Vegetation belts in the Venezuelan Andes were at lower elevations during that period, enabling deer populations inhabiting the páramos (Andean highlands) to expand their distributional range and form a continuum with deer populations in the lowlands (Vuilleumier 1971). At the same time, Margarita Island was connected to the mainland (Rull 1999). Given this scenario, it is likely that populations of white-tailed deer of that period were broadly distributed. Subsequent climatic changes would have created geographic barriers that limited the interconnection among populations, which, along with isolation by distance and changes in ecological factors, could have favored the differentiation of the 4 observed lineages.
Among these 4 lineages, populations from Margarita Island and Mérida represent particular cases for discussion because these populations currently are geographically isolated. In both cases, it is possible to suggest distinct mechanisms of differentiation by vicariance: in the former, due to an extrinsic barrier (sea), and in the latter, to founder effect, suggested by the low value of ⌸ observed in the Andean population (Kliman and Hey 1993;Wiehler and Tiedemann 1998). Nucleotide diversity observed in the goudotii population is of comparable magnitude with that observed in populations that are supposed to have experienced a population bottleneck (Wiehler and Tiedemann 1998). This observation suggests that ancestral populations of the Mérida deer might have been small and that the effects of genetic drift in small populations led to a loss in genetic diversity (Kliman and Hey 1993).
Assuming that current populations from both Margarita and Mérida were derived from an ancestral, broadly expanded population (probably similar to the gymnotis population), the observed phylogeographic structure could be explained as a consequence of ancestral polymorphism. This pattern suggests that isolation of these populations is relatively recent, and differentiation could still be occurring (category III of phylogeographic patterns-after Avise 2000; Avise et al. 1987), which would be coherent with the proposal of a late Pleistocene origin of these lineages. Arguments given by Neigel and Avise (1986) also support our conclusion. The apparent relationship among haplotypes from unrelated geographic localities in clades 1 and 2 could be evidence of this hypothetical widespread ancestral population (see Fig. 1). Additional evidence is provided by gene flow values Ͼ1 between both goudotii and gymnotis, and margaritae and gymnotis (Table 4). This result should not be interpreted as current gene flow, given that the Margarita population is isolated from the mainland by an insurmountable geographic barrier, whereas the Mérida population inhabits elevations of Ͼ3,000 m, without any contact with lowlands populations. We do not consider the possibility of migration mediated by humans, at least in our sample, because no haplotype was shared among populations and Mérida and Margarita clades appear well differentiated in the phylogenetic trees. This result could be interpreted as recent events of gene flow.
It has been suggested that the Mérida and Margarita populations are threatened due to dramatic decreases in their respective population sizes (Rodríguez and Rojas-Suárez 1995). In agreement with this statement, the estimated values of genetic diversity for the Margarita population (Table 1) suggest that the decrease in its population size is relatively recent because genetic variability would be rapidly lost in populations that have maintained a small size for a long time (González et al. 1998;Kliman and Hey 1993). A recent decline in the Margarita population would be consistent with the loss of habitats due to a remarkable increase in the human population size and tourism development in this island in the last few decades.
In the case of the Mérida population, 2 observations lead us to hypothesize that the low genetic diversity observed in this population is due to a small ancestral population (Kliman and Hey 1993), rather than to a recent loss of genetic variability following reduction in population size. First, Mérida appears as a monophyletic clade within lineage 1 (Figs. 2 and 3). Second, the low value of ⌸ is accompanied by a high value of h (Table 1), suggesting a population expansion after a population bottleneck (Eizirik et al. 2001) because it is expected that haplotype diversity recovers faster than nucleotide diversity (Wiehler and Tiedemann 1998). These 2 observations imply that sufficient historical isolation has occurred in this population, causing it to reach a monophyletic condition (Avise et al. 1987) and to lose genetic variability (González et al. 1998;Kliman and Hey 1993).
Our results do not support the proposal of Molina and Molinari (1999) that there are 3 white-tailed deer species in Venezuela. The genetic distances observed among haplotypes of currently accepted subspecies (equivalent to the species proposed by Molina and Molinari [1999]) were Ͻ3%, which lies in the range of values usually observed among subspecies of deer species (Cronin 1991(Cronin , 1992Kuwayama and Ozawa 2000;Nagata et al. 1999;Tamate et al. 1998). Great morphological variation commonly is observed in cervids, which could be explained by the effect of environmental factors and selection (Avise 2000;Cronin 1992;). Nevertheless, the phenotypic differentiation of Andean and Margarita populations of white-tailed deer should be studied further because these populations might represent an interesting case of ecological speciation (Orr and Smith 1998).
Genetic distances observed among North American and Venezuelan white-tailed deer haplotypes are comparable to distances reported for different cervid species (Cronin 1991;Kuwayama and Ozawa 2000;Polziehn and Strobeck 1998). The distinction of North and South American white-tailed deer as different species has been suggested by different authors (Geist 1998). Smith et al. (1986) found an appreciable genetic divergence (14.8%) between North American and Suriname subspecies of white-tailed deer through isoenzyme analysis, in addition to morphological differentiation. When isoenzyme results by Smith et al. (1986) are compared to results obtained by Baccus et al. (1983), it is apparent that North American white-tailed deer are genetically more similar to black-tailed deer than to their South American conspecific. Our data support this statement. In addition, a significant morphological differentiation between North American and Venezuelan whitetailed deer is reported by Molina and Molinari (1999). These observations suggest that North and South American white-tailed deer comprise separate lineages. However, more extensive studies with Central American samples and other molecular markers are needed to clarify the taxonomic status and phylogeographic relationships.
Implications for conservation.-Several authors have demonstrated and discussed the limitations of using either taxonomic classifications or genetic data, only, to establish conservation priorities (Avise 1989;Bowen and Karl 1999;Karl and Bowen 1999;Lande 1988;Riddle and Hafner 1999;Ryder 1986). Conservation units, i.e., evolutionarily conservation units, management units, and geminate evolutionary units, provide an alternative to identifying groups of organisms that require attention for conservation. Based on the concept of conservation units, we propose the following designations. First, Venezuelan whitetailed deer could be considered an evolutionary conservation unit. This proposition is based on observed levels of morphological (Molina and Molinari 1999) and mitochondrial differentiation. Also, Venezuelan and North American haplotypes are clustered in different reciprocally monophyletic clades. In addition, it seems likely that results in genomic differentiation similar to those of Smith et al. (1986) could be ob-served between Venezuelan and North American subspecies. This implies that endangered populations of white-tailed deer in Venezuela deserve attention from international agencies for wildlife conservation. Second, Margarita and Mérida populations could be considered as geminate evolutionary units because these populations are isolated, show morphological differentiation (Molina and Molinari 1999), represent independent lineages, and, in the case of Mérida populations, inhabit a very different ecosystem than that of their conspecifics. This designation underscores that these populations are possibly undergoing speciation and thus require special management schemes. Third, the populations corresponding to O. v. gymnotis could be considered as management units because there is little genetic differentiation at the mtDNA level and there is no evidence of ecological differentiation or geographical or behavioral isolation. Our results suggest that populations from the Venezuelan coast, the Andean lowlands, and the llanos could be considered as management units. Therefore, these conservation units provide a guideline to managing Venezuelan populations of white-tailed deer.
Our results did not support that Venezuelan white-tailed deer should be considered 3 different species; however, we found evidence indicating that threatened subspecies should be protected. Additionally, our results suggest the distinctiveness of Venezuelan white-tailed deer. Future studies should consider other molecular markers, such as nuclear and Y chromosome genes, as well as a wider sampling of localities along the whole distribution of white-tailed deer, in order to clarify the phylogeographic relationships among subspecies clustered in what might be a supraspecific complex.