Rapid and Parallel Chromosomal Number Reductions in Muntjac Deer Inferred from Mitochondrial DNA Phylogeny

Abstract

Muntjac deer (Muntiacinae, Cervidae) are of great interest in evolutionary studies because of their dramatic chromosome variations and recent discoveries of several new species. In this paper, we analyze the evolution of karyotypes of muntjac deer in the context of a phylogeny which is based on 1,844-bp mitochondrial DNA sequences of seven generally recognized species in the muntjac subfamily. The phylogenetic results support the hypothesis that karyotypic evolution in muntjac deer has proceeded via reduction in diploid number. However, the reduction in number is not always linear, i.e., not strictly following the order: 46→14/13→8/9→6/7. For example, Muntiacus muntjak (2n = 6/7) shares a common ancestor with Muntiacus feae (2n = 13/14), which indicates that its karyotype was derived in parallel with M. feae's from an ancestral karyotype of 2n ≥ 13/14. The newly discovered giant muntjac (Muntiacus vuquangensis) may represent another parallel reduction lineage from the ancestral 2n = 46 karyotype. Our phylogenetic results indicate that the giant muntjac is relatively closer to Muntiacus reevesi than to other muntjacs and may be placed in the genus Muntiacus. Analyses of sequence divergence reveal that the rate of change in chromosome number in muntjac deer is one of the fastest in vertebrates. Within the muntjac subfamily, the fastest evolutionary rate is found in the Fea's lineage, in which two species with different karyotypes diverged in around 0.5 Myr.

Introduction

Muntjac deer (Muntiacinae, Cervidae) are distributed throughout Southeast Asia, South China, and India. They are of great interest to evolutionary biologists and cytogeneticists because of the considerable diversity of their karyotypes, despite their morphological similarity (Fontana and Rubini 1990 ). The Indian muntjac, Muntiacus muntjak, possesses the lowest diploid chromosomal number in mammals (2n = 6 for females [F] and 7 for males [M]) (Wurster and Benirschke 1970 ; Wurster and Atken 1972 ; Shi 1976 ), whereas the Chinese muntjac (Muntiacus reevesi) has a 2n number of 46 in both sexes (Wurster and Benirschke 1967 ). These two species, however, can produce viable F₁ hybrids (2n = 27) in captivity, and partial spermatogenesis was observed in hybrids (Shi, Ye, and Duan 1980 ; Shi and Pathak 1981 ; Neiztel 1987 ). Other karyotyped species have intermediate numbers of chromosomes; for example, 2n = 8 F, 9 M in Muntiacus crinifrons (Shi 1983 ), 2n = 8 F, 9 M in Muntiacus gongshanensis (Shi and Ma 1988 ), and 2n = 13 F (Soma et al. 1983, 1987 ), 14 M (L. M. Shi, personal communication) in Muntiacus feae. The tufted deer (Elaphodus cephalophus), which is the sole species in the other genus of the Muntiacinae subfamily, has polymorphic karyotypes with three different diploid numbers, 46, 47, and 48, observed in natural populations (Shi 1981 ; Shi, Yang, and Kumamoto 1991 ). This considerable karyotypic diversity makes muntjac deer species excellent models for the study of chromosomal evolution and speciation.

Hsu, Pathak, and Chen (1975) hypothesized that the large chromosomes in M. muntjak result from multiple tandem and centromeric fusions of small ancestral acrocentric chromosomes similar to those retained in M. reevesi. Comparative cytogenetic studies, especially those recent works using chromosome painting, strongly support this hypothesis and find tandem fusion to be the major process in chromosome evolution of muntjacs (Shi, Ye, and Duan 1980 ; Brinkley et al. 1984 ; Lin et al. 1991 ; Yang et al. 1995, 1997a, 1997c ). However, the scenario of chromosome evolution in muntjacs, such as the chronology of the reduction events, is still unclear. A comprehensive phylogenetic analysis for muntjac deer would be helpful in addressing this question.

Meanwhile, muntjac deer are also a fascinating subject to mammalogists. Although the discovery of new large mammal species is very rare nowadays, a number of new muntjac species have been discovered since the late 1980s. In 1988, the Gongshan muntjac (M. gongshanensis) was discovered on Gongshan Mountain in Southwest China (Shi and Ma 1988 ; Ma, Wang, and Shi 1990 ). In 1994, the giant muntjac (Megamuntiacus vuquangensis or Muntiacus vuquangensis) was found in the Annamite mountains along the border of Laos and Vietnam (Evans and Timmins 1994 ; Touc et al. 1994 ) and was later confirmed (Schaller and Vrba 1996 ; Timmins et al. 1998 ). In 1998, Muntiacus truongsonensis was reported from the Annamite mountains of Laos (Giao et al. 1998 ). In 1999, Muntiacus putaoensis was discovered in northern Myanmar (Amato, Egan, and Rabinowitz 1999 ). Some preliminary observations even indicate that there may exist more new muntjac species in the Annamite mountains (Giao et al. 1998 ; R. J. Timmins, personal communication).

Although much attention has been paid to muntjac deer, the taxonomy of this family is controversial and the phylogeny is still an open question. It has been observed that sequences of mitochondrial ND4L and ND4 genes can supply good phylogenetic information for resolving relationships from subspecies to genus level in vertebrates (Forstner, Davis, and Arevalo 1995 ; Wang et al. 1997 ). In this paper, we present a phylogenetic analysis based on mitochondrial ND4L and ND4 gene sequences for the seven generally recognized Muntiacinae species: E. cephalophus, M. reevesi, M. muntjak, M. feae, M. crinifrons, M. gongshanensis, and M. vuquangensis. We then analyze the evolution of karyotype based on the mtDNA phylogeny obtained.

Materials and Methods

Sample Sources

Table 1 lists all of the species or putative species in Muntiacinae. Asterisks indicate those taxa used in this study. The Chinese water deer (Hydropotes inermis) is used as an outgroup because it retains the hypothetical ancestral Cervidae karyotype (2n = 70) and is believed to be closely related to the ancestral lineage (Groves and Grubb 1987 ; Neiztel 1987 ; Fontana and Rubini 1990 ; Yang et al. 1997c ). The Chinese water deer used to be widely distributed in East Asia but is now found only in east China and Korea (Sheng and Ohtaishi 1993 ). The DNA sample of H. inermis was a gift from Dr. F. Yang at Cambridge University. The blood samples of M. vuquangensis and M. muntjak annamensis were kindly provided by Dr. R. J. Timmins (Wildlife Conservation Society) with the transportation help of Dr. W. Bleisch in 1994. All of the other samples were from the collections of the Kunming Cell Bank of Type Culture Collection of the Chinese Academy of Sciences.

PCR and Sequencing

DNA was extracted from blood or tissue following the protocol of Sambrook, Fritsch, and Maniatis (1989) . The mtDNA fragment between the tRNA^Arg and tRNA^Leu genes, which contains the ND4L, ND4, tRNA^Ser, and tRNA^His genes and part of the tRNA^Arg and tRNA^Leu genes, was PCR amplified for each individual marked with an asterisk in table 1 by two pairs of primers: ND4 against Leu (Forstner, Davis, and Arevalo 1995 ), and Arg2 (5′-ACT CAA AAA GGA CTA GAA TGA-3′) against MuntND4R3 (5′-GCT CGC TAA GAG TCA TCA GGT GGC-3′), respectively. Additional oligonucleotides were designed to be used as sequencing primers to proof each nucleotide at least twice. The PCR reaction was performed on a Robocycler Gradient 40 temperature cycler (Stratagene). Cycle sequencing was carried out on a thermal cycler (Amplitron I, Barnstead/Thermolyne) using an FS-DNA sequencing kit (ABi, number 402079) following the supplied protocol. Sequences were scored on a 377 PRISM automated sequencer (ABi).

Data Analysis

Sequences were aligned manually in the interface of PAUP, version 3.1.1 (Swofford 1993 ), using the bovine mtDNA sequence (Anderson et al. 1982 ) as a reference. Sequence divergence between mtDNA types and neighbor-joining (NJ) trees (Saitou and Nei 1987 ) were obtained using MEGA (Kumar, Tamura, and Nei 1994 ) assuming the Kimura (1980) two-parameter model. The base composition of each DNA sequence was also analyzed using MEGA. Maximum-parsimony trees were constructed using the branch-and-bound option in PAUP, version 3.1.1 (Swofford 1993 ). The sequences of Bos taurus and H. inermis were defined as outgroups in all tree searches. One thousand bootstrap replications were performed for both phylogenetic analyses in order to test the robustness of branches. The ratio of transitions (Ts) over transversions (Tv) was estimated based on the consensus tree using MacClade, version 3.0 (Maddison and Maddison 1992 ). Uniform distribution analysis of substitutions along the DNA segment was also conducted using MacClade, version 3.0. To examine the phylogenetic information content of the final data set, we used the method of Hillis and Huelsenbeck (1992) to study the skewness in the distribution of trees, using 10,000 random trees generated with PAUP, version 3.1.1. In addition, a relative-rate test (Wu and Li 1985 ) was conducted using K2WuLi, version 1.0 (Jermiin 1996 ).

Results and Discussion

Phylogeny of Muntiacinae

We obtained 1,844 bp of unambiguous sequences for muntjac species but 1,846 bp for H. inermis, which contains ND4L, ND4, tRNA^Ser, and tRNA^His genes and part of the tRNA^Leu gene. Muntiacus muntjak yunnanensis and Muntiacus muntjak nigripes share the same sequence. GenBank accession numbers for these sequences are AF190673–AF190685. Base compositions are rather homogeneous across these taxa, with the A+T content ranging from 62.2% (M. feae) to 63.2% (M. vuquangensis). Uniform distribution analysis shows that substitutions are quite homogeneously distributed along the fragment (data not shown). Including the bovine sequence, we summarize the sequence variations of the three gene fragments (tRNA genes were pooled) in table 2 .

The distribution of 10,000 random trees is strongly left-skewed (g1 = −0.8293, P < 0.01), indicating that the data set is significantly structured and contains a strong phylogenetic signal. A single shortest tree with a length of 937 steps was obtained using the unweighted maximum-parsimony method (fig. 1a ). When a single gene or gene cluster (all tRNA genes) was used under an unweighted scheme, the topologies of resulting trees were more or less different from the tree in figure 1a. Nevertheless, if we assign more weight to transversions than to transitions (e.g., 7:1, as shown in table 2 ), the trees obtained from a single gene become quite consistent with the tree in figure 1a, and the tree obtained from the combined data set remains the same as that in figure 1a. The NJ tree, shown in figure 1b, was constructed based on the Kimura two-parameter distance matrix (table 3 ). This NJ tree has the same topology as the parsimony tree (fig. 1 ).

Some phylogenetic information has previously been provided for the genus Muntiacus based on morphological data (Ma, Wang, and Xu 1986 ; Groves and Grubb 1990 ). Ma, Wang, and Xu (1986) proposed that crinifrons and rooseveltorum were derived from the reevesi lineage, whereas feae and muntjak were derived from a different lineage. Groves and Grubb (1990) thought that rooseveltorum was a synonym of feae and that feae was a sister group to muntjak and crinifrons. Molecular markers have also been used to study the phylogeny of genus Muntiacus (Lan, Shi, and Suzuki 1993 ; Lan, Wang, and Shi 1995 ; Giao et al. 1998 ). The results of these previous studies, however, could not supply a sound phylogeny for muntjac deer due to either poor representation of species or lack of informative characters.

In this study, a more inclusive phylogenetic tree is generated for the muntjac subfamily. The proposed tree topology appears quite robust; the same consensus tree is always obtained using different phylogenetic inference methods (parsimony or NJ) or using different gene segments (single gene or combined data set; fig. 1 ). All branches in the trees are well supported by bootstrap analyses (>70%), with one exception for the vuquangensis-reevesi clade (53%) in the parsimony tree. However, this branch is well supported by the NJ tree with a probability of 0.82. In our phylogenetic trees (fig. 1 ), the basal split separates Elaphodus and Muntiacus. There are two major clades within the genus Muntiacus; one includes vuquangensis and reevesi, and the other includes feae, gongshanensis, crinifrons, and muntjak. In the latter clade, muntjak is positioned in a distinct branch as sister to the Fea's muntjac lineage. Within the Fea's muntjac lineage, gongshanensis is most closely related to crinifrons, with feae as their sister species. The four subspecies of muntjak are monophyletic. The newly discovered giant muntjac (M. vuquangensis) is relatively closer to M. reevesi. When it was initially described, the giant muntjac was granted generic status, Megamuntiacus (Touc et al. 1994 ). This classification, however, was subsequently questioned by Schaller and Vrba (1996) and Timmins et al. (1998) . Our result supports it as a member of Muntiacus. The sequence divergences between M. vuquangensis and other muntjacs are at the same level as those among other distinct Muntiacus species (table 3 ). Given the constant evolutionary rate of this mtDNA segment in muntjacs as shown below by the molecular-clock test, this result indicates that the giant muntjac has not diverged much from other Muntiacus species.

Molecular-Clock Test

The concept of a constant substitution rate along each muntjac lineage was tested with Wu and Li's (1985) relative-rate test. Using H. inermis as the outgroup, all pairwise comparisons between ingroup taxa were performed. None of the comparisons could reject the molecular-clock hypothesis (data not shown, available on request). Therefore, the molecular clock has been ticking well in muntjac mtDNAs. Assuming this rate constancy, by referring to the fossil records for some extant muntjac species, we tentatively estimated the evolutionary rate of this mtDNA fragment (mainly ND4L and ND4). Fossils of M. reevesi first appeared in the early Pleistocene (∼1–2 MYA), while M. muntjak and M. feae appeared in the middle Pleistocene (∼0.5–1 MYA) (Ma, Wang, and Xu 1986 ; Dong 1993). Based on the genetic distance in table 3 , we reach two estimations for the mtDNA evolution: 3.8%–7.5%/Myr and 3.8%–7.6%/Myr, which are quite compatible with each other. According to these rates, the following chronology of muntjac speciation is inferred: the two muntjac genera, Elaphodus and Muntiacus, diverged about 1.9–3.7 MYA; vuquangensis separated from reevesi about 0.9–1.8 MYA; feae and muntjak shared a common ancestor until 0.8–1.5 MYA; and the divergence between crinifrons and gongshanensis was a very recent event (0.3–0.5 MYA). Of course, because one species could have existed for a while before the earliest known fossil was formed, these datings could turn out to be underestimates, as all fossil-based molecular calibrations tend to be. Attempts (data not shown) to use rates calibrated using the well-known artiodactyl-cetacean divergence (Arnason and Gullberg 1996 ) led to divergence times that were clearly contradicted by the fossil data, indicating that either the muntjac clock might tick more rapidly than that of mammals in general or artiodactyls and cetaceans have diverged from each other so much that underestimation of multiple hits become inevitable for these mtDNA genes.

Rapid Reduction of Chromosomal Numbers

Considerable diploid number variations among related species have been reported from grasshoppers and some other insects (reviewed in White 1978 ), gibbons (2n = 38, 44, 50, and 52) (Jauch et al. 1992 ), and mice (Nachman and Searle 1995). In particular, there was a recent report of a group of island mice in which chromosome number has nearly halved in less than 500 years (Britton-Davidian et al. 2000 ). However, tandem fusions are rarely involved in the chromosome number changes in either gibbons or mice.

As mentioned above, tandem fusion is the major cytological change during the karyotypic evolution of muntjacs. However, the rate of reduction is unknown. In this study, with the help of molecular phylogenetic data, we find a remarkably rapid chromosomal reduction rate in muntjac deer. Bush et al. (1977) calculated chromosome evolution rates in vertebrates. They found rapid chromosomal evolution and speciation in mammals, and the fastest rate they observed was in horses, in which the rates of chromosome number and fundamental number (FN) changes were 0.61 and 0.79 per Myr, respectively. Based on the diploid numbers and FNs observed in muntjacs (Fontana and Rubini 1990 ), using the formula of Bush et al. (1977) , we obtained the result that, in the genus Muntiacus, the chromosome number changes by 1.08–2.11 per Myr and the FN changes by 0.97–1.91 per Myr. Even the conservative estimates of these two rates are faster than those for horses. Therefore, the evolutionary rate of chromosomal number in muntjac deer is among the fastest in vertebrates.

Within muntjacs, the most drastic reduction process occurred in the Fea's lineage. According to our divergence time estimates, the karyotypes of M. gongshanensis and M. crinifrons, which have been found apparently different (Yang et al. 1995, 1997b ), were derived from the ancestral karyotype less than 0.5 MYA. In fact, the genetic distances among the three species of the Fea's lineage are very small compared with the divergences among other deer. Cronin (1991) has shown that the mtDNA divergence within deer species and between species of the same deer subfamily are usually <3% and 4%–12%, respectively. In a phylogenetic study on musk deer (Moschus) based on the sequences of mitochondrial cytochrome b genes, Su et al. (1999) also reported that around 3% sequence divergences were found only between very close species or subspecies. In table 3 , the sequence divergences between crinifrons, gongshanensis, and feae are less than 4%, and the divergence between crinifrons and gongshanensis is as small as 2.2%. These data show that the Fea's lineage has apparently experienced extremely rapid chromosome evolution.

Parallel Chromosomal Number Reduction

Our phylogenetic results reveal that karyotypic evolution in muntjac deer has proceeded via reduction in diploid number, supporting the fusion hypothesis of muntjac chromosome evolution. Based on the phylogeny (fig. 1 ), together with all available chromosome data, six chromosomal reduction episodes can be deduced in these species. From the hypothetical ancestral karyotype of deer (2n = 70), the first two reductions result in the karyotypes of the tufted deer (E. cephalophus, 2n = 46, 47, 48) and the Chinese muntjac (M. reevesi, 2n = 46), respectively. The reevesi-like muntjac ancestor subsequently experienced further chromosomal reduction, and thereby the karyotypic array of Muntiacus species was derived. One clade of the progeny leads to the lineage of Fea's muntjac (M. feae, 2n = 13 F, 14 M). From this lineage, the black muntjac (M. crinifrons) and the Gongshan muntjac (M. gongshanensis), both with 2n = 8 F and 9 M, were independently derived within a short time (probably <0.5 Myr).

However, the reduction in number is not always linear, i.e., not strictly following the order 46→14/13→8/9→6/7. The 2n = 6/7 karyotype in the Indian muntjac (M. muntjak) was not sequentially derived from the 2n = 8/9 karyotype existing in crinifrons or gongshanensis. Rather, M. muntjak may share a common ancestor with feae (2n = 14/13), suggesting that it results from a reduction pathway parallel to the Fea's lineage. Since the karyotype of 2n = 8 has been found in one individual of Muntiacus muntjak muntjak (Wurster and Atken 1972 ), it is possible that the 2n = 6/7 karyotype held by other M. muntjak subspecies was derived from an intermediate 2n = 8/9 ancestor represented by M. m. muntjak. Nevertheless, this 2n = 8 karyotype is still a descendant of the reduction pathway parallel to the Fea's lineage. It is even possible that the 2n = 6/7 karyotype was independently derived from the common 2n ≥ 14 ancestor of M. muntjak and the Fea's lineage.

Evidence below supports the above interpretations. Chromosome painting data have suggested that the 2n = 6/7 karyotype of M. muntjak cannot be derived simply from the 2n = 8/9 karyotypes of crinifrons or gongshanensis. In fact, an unpublished chromosome tree based on chromosomal painting for reevesi, gongshanensis, crinifrons, muntjak, and another 2n = 70 deer (Mazama gouazoubira) reveals phylogenetic relationships for these species similar to those of the mtDNA tree (F. Yang, personal communication). In addition, the large population size, wide range, and well-defined subspecies differentiation of M. muntjak all suggest that M. muntjak is a species with a relatively long history, unlikely to be derived from the young feae lineage.

More reduction episodes and parallel events may be identified after the determination of karyotypes of those newly reported species. For example, M. vuquangensis may represent another chromosomal reduction pathway from the ancestral 2n = 46 karyotype, as this species is a member of Muntiacus and holds more derived characters than M. reevesi does (Schaller and Vrba 1996 ; Timmins et al. 1998 ). The most recently reported two species might be closely related to the giant muntjac based on the overlapping geographic distribution. If this is the case, then those two species could be the progeny of the reduction descent represented by the giant muntjac.

Chromosomal Number Reduction and Speciation

The inferred process of chromosome evolution in muntjac deer implies that there is a general, constant tendency toward chromosome reduction, with tandem fusion as a major factor, in the speciation process of this group of species. The correlation between chromosomal rearrangement and speciation has been discussed extensively (White 1978 ; Patton and Sherwood 1983 ; Sites and Moritz 1987 ), although tandem fusion has seldom been taken into account, probably due to its rarity in nature. Compared with other types of chromosome rearrangements, multiple tandem fusions would lead to high frequency of unbalanced gametes in heterozygotes and hence can serve effectively in erecting a reproductive isolation barrier once new multiply fused chromosomes are fixed. However, how to overcome the negative heterotic effects prior to the fixation is a problem (Patton and Sherwood 1983 ). As suggested by other authors, this kind of stasipatric speciation (sympatric speciation caused by chromosome rearrangements) most likely occurs in taxa with small demes (White 1978 ; Patton and Sherwood 1983 ; Neiztel 1987 ; Sites and Moritz 1987 ). Patton and Sherwood (1983) further pointed out that in the case of multiple rearrangements, in order for the new karyotype to be fixed, the fixation has to be rapid. Also, the chance to overcome the negative heterosis will increase if the specific chromosomal rearrangement is not unique. Alternatively, this problem can also be overcome by successive accumulation of slightly deleterious rearrangements. In other words, individual tandem fusions in muntjacs may not be so deleterious as to cause a rapid elimination of fused karyotype. After more fusions accumulate, postmating isolation may have been established between individuals with severely different karyotypes.

Interestingly, several pieces of evidence from muntjac deer suggest that these mammals may have succeeded in stasipatric speciation, with chromosomal tandem fusion as the major genetic mechanism. First, muntjacs have been found to have the smallest breeding group in Cervidae (Clutton-Brock and Albon 1980 ), while many Muntiacus species are sympatric with one or more other species (see table 1 ). Second, the muntjac chromosomes appear to have “sticky ends” that increase fusion ability. A number of satellite DNAs have been found at or near the centromeric and telomeric regions of muntjac chromosomes (Bogenberger, Neumaier, and Fittler 1985 ; Benedum et al. 1986 ; Lin et al. 1991 ; Scherthan 1991 ). Some of these repetitive elements, such as C5 and the mammalian telomere sequence, have been found at the putative fusion points (Lee, Sasi, and Lin 1993 ; Fronicke and Scherthan 1997 ; Yang et al. 1997a ). It has been proposed that homologous repetitive elements at the centromeres and telomeres are able to cause illegitimate recombination between nonhomologous chromosomes and thus result in tandem fusion (Ferguson-Smith 1973 ; Yang et al. 1997a ). The stickiness of the muntjac chromosome ends might have caused chromosome fusion at a higher rate than that of other mammals, as discussed above. In fact, the occurrence of tandem fusion has been observed in an established cell line of M. muntjak (Neiztel 1987 ). We can imagine that while the karyotypic lability continuously creates variations, some of the new karyotypes can be fixed in the small breeding populations by virtue of either successive accumulations of slightly deleterious fusions or coexistence of similar fusions. Each of the surviving karyotypes would represent a new Muntiacus species. Species with a recently survived karyotype would not have had enough time to disperse to a large area, which is compatible with the distribution pattern of Muntiacus species: most are limited to small areas, except for M. reevesi and M. muntjak (table 1 ).

The molecular mechanism whereby the muntjac telomere and centromere repetitive sequences induce frequent tandem fusions is unknown. Further study on muntjac repetitive DNAs, as well as on the mutation mechanism of these repeats, will bring us new insight into the essential nature of the evolution of muntjac chromosomes. By elucidating the driving force behind the tandem fusions, we may one day be able to reconstruct the reduction process in laboratories.

Ross Crozier, Reviewing Editor

1

Present address: Department of Ecology and Evolution, University of Chicago.

2

Present address: Department of Oncology, University of Wisconsin–Madison.

1

Keywords: muntjac deer (Muntiacinae) mitochondrial ND4L ND4 genes phylogeny chromosomal evolution

2

Abbreviations: MYA, million years ago; Myr, million years; mtDNA, mitochondrial DNA; ND4, subunit 4 of nicotinamide adenine dinucleotide dehydrogenase; ND4L, subunit 4L of nicotinamide adenine dinucleotide dehydrogenase; NJ, neighbor joining.

3

Address for correspondence and reprints: Wen Wang, Department of Ecology and Evolution, University of Chicago, Chicago, Illinois 60637. E-mail: [email protected].

Table 1 Information of Diploid Number and Range for All Extant Muntjac Species

Table 2 Summary of Sequence Variations

Table 3 Pairwise Kimura Two-Parameter Distances (below the diagonal) and Transition/Transversion Numbers (above the diagonal)

Fig. 1.—(a) Parsimony tree and (b) neighbor-joining tree. The supporting percentages of 1,000 bootstrap replicates are shown for each internal branch. Tree length is shown in the neighbor-joining tree. Diploid numbers are listed in parentheses

We dedicate this paper to our mentor, the late Professor Liming Shi, who was one of the pioneers in studying muntjac chromosome evolution. Special thanks to Dr. Chung-I Wu at the University of Chicago for his great help and encouragement during the preparation of the manuscript, and to Dr. Fengtang Yang at Cambridge University for his helpful discussion and generosity in showing us his unpublished chromosomal painting data. We also thank Drs. James Patton, Manyuan Long, Janice Spofford, A. Hon-Tsen Yu, and Esther Betran and Ms. Haijing Yu for their critical review of the manuscript. We appreciate Mr. Shikang Gou's technical assistance. We further thank Drs. Rob Timmins and William Bleisch from the Wildlife Conservation Society (New York) for providing the blood samples of M. vuquangensis and Muntiacus muntjak annamensis, Dr. Fengtang Yang for the DNA sample of H. inermis, and the Kunming Cell Bank of Type Culture Collections of the Chinese Academy of Sciences for samples of the other species used in this study. This study was supported by various fundings from the Chinese Academy of Science and the Chinese Science Foundation granted to Liming Shi, W.W., and H.L. Both authors contributed equally to this work.