Phylogeography and allopatric divergence of cypress species (Cupressus L.) in the Qinghai-Tibetan Plateau and adjacent regions.

BACKGROUND: Although allopatric speciation is viewed as the most common way in which species originate, allopatric divergence among a group of closely related species has rarely been examined at the population level through phylogeographic analysis. Here we report such a case study on eight putative cypress (Cupressus) species, which each have a mainly allopatric distribution in the Qinghai-Tibetan Plateau (QTP) and adjacent regions. The analysis involved sequencing three plastid DNA fragments (trnD-trnT, trnS-trnG and trnL-trnF) in 371 individuals sampled from populations at 66 localities.
RESULTS: Both phylogenetic and network analyses showed that most DNA haplotypes recovered or haplotype-clustered lineages resolved were largely species-specific. Across all species, significant phylogeographic structure (N(ST) > G(ST), P < 0.05) implied a high correlation between haplotypes/lineages and geographic distribution. Two species, C. duclouxiana and C. chengiana, which are distributed in the eastern QTP region, contained more haplotypes and higher diversity than five species with restricted distributions in the western highlands of the QTP. The remaining species, C. funebris, is widely cultivated and contained very little cpDNA diversity.
CONCLUSIONS: It is concluded that the formation of high mountain barriers separating deep valleys in the QTP and adjacent regions caused by various uplifts of the plateau since the early Miocene most likely promoted allopatric divergence in Cupressus by restricting gene flow and fixing local, species-specific haplotypes in geographically isolated populations. The low levels of intraspecific diversity present in most species might stem from population bottlenecks brought about by recurrent periods of unfavorable climate and more recently by the negative impacts of human activities on species' distributions. Our findings shed new light on the importance of geographical isolation caused by the uplift of the QTP on the development of high plant species diversity in the QTP biodiversity hotspot.

Background

Geographic isolation between populations within species is a major precursor of allopatric speciation [1,2]. In the absence of gene flow, different alleles become fixed in different populations through the action of selection and/or genetic drift potentially leading to the origin of new species [3]. It is expected, therefore, that in regions where geographic isolation is common, for example, in areas where high mountains separate populations occurring in deep valleys, allopatric divergence will be promoted and accelerated due to the formation of locally adapted gene pools and possibly the origin of intrinsic postzygotic isolation according to the Bateson-Dobzhansky-Muller model [3-5]. A phylogeographic analysis of closely related species that occur in such areas can be instructive on the importance of geographic isolation and allopatric divergence in their origin and subsequent evolution [6].

The Qinghai-Tibetan Plateau (QTP) and adjacent regions comprise an excellent mountainous system for studying plant diversification and evolution. Levels of plant diversity and endemism in these regions are very high and some core areas are regarded as biodiversity hotspots (e.g. [7]). A few studies in this region have begun to elucidate the process of species diversification and/or trace shifts in the ranges of single species during the Quaternary climatic oscillations (e.g. [8,9]). However, no previous studies have attempted to examine genetic differentiation in this region among species of a closely related group, and in so doing shed light on their respective evolutionary histories. Here we report such a case study of a group of cypress species (Cupressus L.) that occurs on the QTP and adjacent areas of China and the Himalayas. Cypresses are trees or shrubs native to scattered localities in temperate regions in the northern hemisphere. Because of their attractive morphology they are important in horticulture and are widely cultivated in many parts of the world [10]. In several mountainous regions of the QTP, cypresses form natural forests [11]. From the results of a recent molecular phylogenetic analysis of cypresses, Little [12] concluded that True Cypresses only encompass species in the Old World, and that those in the New World should be transferred to Callitropsis. As a result, the genus Cupressus has been re-circumscribed to comprise 12 Old World species, distributed respectively in the Mediterranean region and Africa (C. sempervirens, C. dupreziana and C. atlantica) and Asia (C. torulosa, C. gigantea, C. cashmeriana and C. austrotibetica in the high altitude region of the QTP and west Himalayas, and C. duclouxiana, C. chengiana, C. funebris, C. jiangeensis and C. tonkinensis in the low altitude region of the eastern QTP, central China and Vietnam [12]). Morphological distinction between some of these species is marginal causing taxonomic delimitation to be disputed, especially between Asian species [10,12-19]. We followed Little [12] and treated all taxonomic units recognized by him as separate species. Most of these Asian species occur allopatrically in the QTP and adjacent regions, and only a few are parapatric in parts of their range (Fig. 1) (Additional File 1) [10,15].

Figure 1

The distribution of eight cypress species and frequencies of plastid DNA haplotypes in each population surveyed. Upper left panel, a, b, c and d: Geographic ranges of eight cypress species and locations of populations sampled. The upper right panels a, b, c and lower panel d: plastid DNA haplotypes identified and the proportion of each present within populations of each species examined (circle size is proportional to the number of individuals surveyed per population). Further details of size and haplotype diversity of populations are provided in Additional File 1.

We used plastid (mainly chloroplast) DNA sequence data to examine interspecific divergence and to assess intraspecific diversity of these species. Plastid DNA is paternally inherited and dispersed mainly by pollen in all species of the Cupressaceae investigated [20-23]. Recent studies suggest that paternally inherited plastid DNA sequences in conifers may be species-specific and therefore effective in tracing lineage sorting between species [24-27]. Plastid DNA sequence variation has been used previously in phylogenetic studies of cypresses, but only single sample from most species were examined [12]. In addition, surveys of RAPD variation [28] and nuclear ITS variation [29] have been conducted within and among cypress species including those examined in the current study. However, the results of these studies were inconsistent with regard to interspecific relationships [12]. In this study, we aimed to address the following questions based on surveys of sequence variation for three plastid DNA fragments within and between populations at the species level: (1) Do plastid DNA haplotypes or clusters of such haplotypes correspond with morphological differentiation and the allopatric distributions of the Cupressus species examined? (2) Does intraspecific DNA diversity correlate with the natural population size of a species and, in turn, provide guidance for conservation management?

Results

Haplotype distribution and relationships

In total, 30 haplotypes were identified (Additional Files 1, 2 and 3, Fig. 1), of which 17 occurred in C. chengiana (N = 89), seven in C. duclouxiana (N = 86), two in C. funebris (N = 89) and four among C. gigantea (N = 58), C. jiangeensis (N = 1), C. austrotibetica (N = 22), C. cashmeriana (N = 4) and C. torulosa (N = 22). No intraspecific cpDNA variation was recorded in C. gigantea, C. jiangeensis, C. cashmeriana, C. austrotibetica and C. torulosa. Cupressus cashmeriana and C. austrotibetica shared the same haplotype, which differed by a single nucleotide from the haplotype possessed by C. torulosa. These two sequences were very different from the haplotype possessed by C. gigantea. A neighbour joining analyses showed that the 30 haplotypes clustered into six different lineages (A-F) with most haplotypes being species-specific (Fig. 2). Although all six lineages were recovered in the 50% strict consensus tree of 51267 most parsimonious trees (CI = 0.77, RI = 0.87, Length = 141), the pattern of relationships between lineages received low bootstrap support (Fig. 2). Lineage A comprised nine haplotypes possessed by individuals of C. chengiana sampled from Gansu province, China. In contrast, the other eight haplotypes detected in this species were present only in individuals sampled from Sichuan, China, which together with the single haplotype (H5) recovered from C. jiangeensis, comprised lineage E. The sister relationship between lineages A and E was poorly supported. Lineage D was composed of the single haplotype present in C. torulosa plus the haplotype shared by C. austrotibetica and C. cashmeriana, and was sister to lineages A and E (Fig. 2) of C. chengiana. The seven different haplotypes detected in C. duclouxiana formed a monophyletic lineage C, which was sister to lineage B that consisted of the single haplotype possessed by C. gigantea. Lineage F comprised the two haplotypes specific to C. funebris and was sister to the five other lineages (A to E) and to two cypress species occurring out of Asia. To evaluate the origins of these lineages (species), we selected the commonest haplotype representing each lineage and dated their origins. According to the estimated nucleotide mutation rate for Cupressus (Additional File 4), C. funebris was estimated to have diverged from the other lineage/species approximately 26 million years ago (Mya) (Additional File 4). C. duclouxiana in the eastern QTP diverged from the western C. gigantea around 10 Mya, while the eastern C. chengiana diverged from the western C. torulosa/C. austrotibetica/C. cashmeriana lineage around 13 Mya. The intraspecific divergence that gave rise to the two different lineages of C. chengiana identified was dated to approximately nine Mya.

Figure 2

A Neighbor-Joining tree showing phylogenetic relationships between plastid DNA haplotypes. Numbers associated with branches are maximum parsimony bootstrap supports. Circle sizes in network are proportional to haplotype frequencies over all populations; with the largest circle representing the most abundant haplotype and the colours of circles indicating different haplotypes as shown in Fig. 1.

Because the number of mutations distinguishing any two lineages was always greater than six, we constructed separate haplotype networks for the three lineages that contained most haplotypes (i.e. lineages A, C and E). Within lineage E, two haplotypes H10 and H3 occurred in a small sub-lineage that was separated by four mutations from other haplotypes (Fig. 2). However, H10 co-occurred with the most frequent haplotype in this lineage, H4, in most sampled populations. The single haplotype (H5) recovered from C. jiangeensis was also present in lineage E and is derived from H4. In lineage A, H17 was the most distantly related haplotype relative to other haplotypes (Fig. 2).

Diversity among and within species

Genetic differentiation across all species was high (Table 1, 2) and showed significant phylogeographic structure (NST (0.962) > GST (0.791), P < 0.05). Mantel tests revealed a significant association (r2 = 0.2285, P < 0.001) between genetic differentiation and geographic distances between populations over all species. For C. chengiana and C. duclouxiana, total intraspecific diversity (HT = 0.791 and 0.659, respectively) across populations was higher, but not substantially, compared with the average within-population diversity (HS = 0.536 and 0.443, respectively). Within C. duclouxiana, interpopulation variation accounted for a high proportion of intraspecific variation (23.3%, Table 2), indicating strong differentiation between populations. The significant phylogeographic structure of C. chengiana reflected the presence of two distinct lineages (A and E) within this species that were located in Gansu and Sichuan provinces, respectively. This regional divergence accounted for 44.2% of the total variation recorded within C. chengiana (Table 2). Variation within each of these lineages was almost entirely attributed to variation within populations (Table 2, note that H5 of C. jiangeensis was included in the calculations for lineage E). Rare haplotypes were frequent in many populations that were representatives of lineages A or E. In C. funebris, two haplotypes were recovered, but no variation was recorded among populations.

Table 1

Estimates of average genetic diversity within populations (HS), total gene diversity (HT), interpopulation haplotype differentiation (GST), and interpopulation haplotype differentiation taking into account sequence difference (NST) (mean ± SE in parentheses)

Species	H S	H T	G ST	N ST	NST - GST
C. chengiana	0.536 (0.079)	0.791 (0.051)	0.322 (0.074)	0.558 (0.056) *	0.236
C. chengiana (lineage E,Sichuan populations)	0.471 (0.136)	0.489 (0.117)	0.037 (0.093)	-0.007 (0.075) ns	-0.044
C. chengiana (lineage A,Gansu populations)	0.618(0.052)	0.597(0.060)	-0.035(0.053)	-0.022(0.083)ns	0.013
C. funebris	0.254 (0.060)	0.249 (0.050)	-0.021 (0.055)	-0.021 (0.055) ns	0
C. duclouxiana	0.443(0.095)	0.659(0.098)	0.328(0.116)	0.488 (0.092)ns	0.160
C. gigantea	NC	NC	NC	NC	NC
C. torulosa	NC	NC	NC	NC	NC
C. cashmeriana &C. austrotibetica	NC	NC	NC	NC	NC
C. torulosa s.l. & C. gigantea	0 (0)	0.665(0.039)	1 (NC)	1 (NC)	0
All species	0.190(0.034)	0.908(0.012)	0.791(0.036)	0.962(0.013)*	0.171

*indicates that NST is significantly different from GST (0.01

0.05); NC, not computed due to small sample size; C. torulosa s.l. includes C. torulosa, C. cashmeriana and C. austrotibetica

Table 2

Analysis of molecular variance (AMOVA) for cpDNA haplotypes of cypress species

Species/lineage	Source of variation	d.f.	SS	VC	Variation (%)	Fixation index
C. chengiana (lineage E,Sichuan populations)	Among populations	5	1.827	0.017	6.8	FST = 0.068
	Within populations	44	10.213	0.232	93.2
	Total	49	12.040	0.249
C. chengiana (lineage A,Gansu populations)	Among populations	3	0.925	-0.001	-0.2	FST = -0.002
	Within populations	36	11.325	0.315	100.2
	Total	39	12. 250	0.314
C. chengiana	Among lineages	1	10.066	0.218	43.9	FCT = 0.439*
	Among populations within groups	8	2.752	0.009	1.8	FSC = 0.032
	Within populations	80	21.538	0.269	54.3	FST = 0. 457*
	Total	89	34.356	0.496
C. funebris	Among populations	10	0.951	-0.004	-3.5	FST = -0.035
	Within populations	78	10.150	0.130	103.5
	Total	88	11.101	0.126
C. duclouxiana	Among populations	10	6.707	0.061	23.3	FST = 0.233*
	Within populations	75	15.142	0.202	76.7
	Total	85	21.849	0.263
C. gigantea	Among populations	5	0	0	0	FST = 0
	Within populations	52	0	0	0
	Total	57	0	0
C. torulosa	Among populations	11	0.952	0.053	100.0	FST = 1.000
	Within populations	9	0	0	0
	Total	20	0.952	0.053
C. cashmeriana &C. austrotibetica	Among populations	7	0	0	0	FST = 0
	Within populations	18	0	0	0
	Total	25	0	0	0
C. torulosa s.l. andC. gigantea	Among groups	3	30.659	0.469	97.6	FCT = 0.976*
	Among populations within groups	22	0.955	0.012	2.4	FSC = 1.000
	Within populations	80	0	0	0	FST = 1.000*
	Total	105	31.613	0.481
All species	Among species	7	100.946	0.319	63.5	FCT = 0.635*
	Among populations within species	58	20.258	0.039	7.7	FSC = 0.210 *
	Within populations	305	44.123	0.145	28.8	FST = 0.712 *
	Total	370	165. 326	0.502

d.f., degrees of freedom; SS, sum of squares; VC, variance components; * P < 0.001.

*indicates that NST is significantly different from GST (0.01