Molecular phylogeography and evolutionary history of Picea likiangensis in the Qinghai–Tibetan Plateau inferred from mitochondrial and chloroplast DNA sequence variation

Authors


E-mail: liujq@nwipb.ac.cn. Tel.: 86-931-8914305. Fax: 86-931-6143282.

Abstract

Abstract  The aim of the present study was to examine the phylogeographic and evolutionary history of Picea likiangensis, a dominant species of the conifer forests in the eastern declivity of the Qinghai–Tibetan Plateau. We collected 422 individuals from 42 natural populations of three major varieties classified under this species. In conifers, mitochondrial (mt) DNA and chloroplast (cp) DNA dispersed by seeds or pollen experience very different levels of gene flow. To this end, we examined the sequence variation of two mtDNA fragments (nad5 intron 1 and nad1 intron b/c) and three cpDNA fragments (trnL–trnF, trnS–trnG and nadhK/C). We found that cpDNA probably introgressed from P. purpurea into remote populations of P. likiangensis through long-distance dispersal. Multiple refugia seem to have been maintained for P. likiangensis during the Last Glacial Maximum because the cpDNA and mtDNA haplotypes recovered were fixed in the different regions. Postglacial expansions were only detected at the distributional edges of this species where a single cpDNA or mtDNA haplotype was fixed in adjacent populations. However, genetic imprints of postglacial expansions from these two sets of markers were different in the western and southeastern regions, which may result from the long-distance dispersal of the cpDNA, as well as its fast lineage sorting during intraspecific divergences. Analysis of molecular variance further suggested that genetic differentiation between the three varieties is higher at cpDNA markers than at mtDNA markers, which supports the previous viewpoint that cpDNA markers with a high rate of gene flow may be more effective in delimitating closely related taxa. Together, the results of the present study highlight the evolutionary complexity of a widely distributed species owing to interactions among local and edge expansion, long-distance dispersal, and intraspecific divergences at two sets of DNA genomes with different rates of gene flow.

The Qinghai–Tibetan Plateau (QTP), the highest region in the world, has been one of the ‘hotspots’ for phylogeographic studies for two reasons. First, this region is sensitive to climatic changes (Zhang et al., 2002), with those changes that have occurred in the past having significant impacts on the range distributions of both the plants and animals in the QTP. Changes in the ranges of those taxa may have left distinct genetic imprints in current populations (Hewitt, 2004). Second, because of the diversity of species distributed across the QTP, this region and its adjacent regions comprise one of the world's ‘biodiversity hotspots’ (Myers et al., 2000). Species with different habits or adaptive traits may differ in their evolutionary histories in response to past climatic changes. Recent phylogeographic studies of alpine plant species from this region have confirmed these predictions. For example, the extant population on the plateau platform derived from the refugia at the plateau edge after the Last Glacial Maximum (LGM) and/or during previous interglacial periods (Zhang et al., 2005; Meng et al., 2007; Chen et al., 2008; Yang et al., 2008; Cun & Wang, 2010). However, other species seem to have persisted at high-altitude at least during the LGM, although the similar range expansions did occur on a local scale (e.g. Opgenoorth et al., 2010; Wu et al., 2010). In addition, deep intraspecific divergences have been found in a few species and some intraspecific lineages survived in the central QTP through the total Quaternary climatic oscillations (e.g. Wang et al., 2009; Jia et al., 2011, 2012). Hybridizations or introgressions between different intraspecific lineages are common within these species. However, phylogeographic and evolutionary histories of other species remain unknown, especially those with a significant role in maintaining the stability of the local ecosystem as the dominant or key species, which should be the most sensitive to climate changes (Hewitt, 2004).

The aim of the present study was to clarify phylogeographic and evolutionary history of Picea likiangensis (Franch.) Pritzel. As a dominant species of the conifer forest ecosystems in western China, this species has a wide distribution, extending from west Sichuan to Tibet and from Yunnan to Qinghai, spanning a wide range of altitudes between 2500 and 4000 m. Five intraspecific varieties have been recognized (Fu et al., 1999); however, only three have wide distribution and discernible characters, namely var. likiangensis, var. rubescens Rehder & E. H. Wilson, and var. linzhiensis Cheng & L. K. Fu. These three varieties differ from one another in terms of the stomatal lines on the leaves, as well as in the shapes and hairs of the first-year branchlets. For example, var. likiangensis usually has slender and sparsely pubescent branchlets and leaves with two to four stomatal lines along each abaxial surface, whereas the first-year branchlets of var. rubescens are stout and densely pubescent with short nodes and leaves that have three or four stomatal lines along each abaxial surface. In var. linzhiensis, the first-year branchlets are usually glandular and hairy, whereas the stomatal line is lacking along the abaxial surface of the leaves. These three varieties have different distributions, with var. rubescens found in the north of the QTP, var. linzhiensis found in the west, and var. likiangensis found in the southeast, although the distribution of two of the three varieties overlaps in western Sichuan (Fig. 1). Owing to its wide distribution, this species is ideal to examine whether the distribution of plant species changed in response to past climatic changes in this region. Changes in distributional ranges should have left genetic imprints in current populations (e.g. Hewitt, 2004). Our previous study based on randomly amplified polymorphic DNA markers suggested that this species may have persisted in multiple refugia during the glacial stages (Peng et al., 2007). This was confirmed by a recent study of a limited number of populations based on both chloroplast (cp) DNA and mitochondrial (mt) DNA sequence variations (Du et al., 2011). In addition, gene flow was found to be frequent between this species and another spruce species (P. purpurea) in their contact zones. These interspecific introgressions may have occurred in the recent past, with more introgressions occurring from the local P. likiangensis to the invading P. purpurea rather than in the other direction and at mtDNA markers rather than at cpDNA markers (Du et al., 2011). In most conifers, two uniparentally inherited organelle genomes (mtDNA and cpDNA) have contrasting modes of inheritance and rates of gene flow (Neale & Sederoff, 1988; Mogensen, 1996; Petit & Vendramin, 2007). In addition, cpDNA markers with a high rate of gene flow have been found to show faster lineage sorting during speciation and genetic differentiation (Du et al., 2009; Zhou et al., 2010). In the present study, we sampled 422 individuals from 42 natural populations across the range of the species using the same cpDNA and mtDNA fragments as in our previous study (Du et al., 2011). The major goals of the present study were to: (i) examine whether exotic introgressions from P. purpurea had extended to P. likiangensis in other regions in addition to the contact zones between two species; (ii) reveal the range dynamics of P. likiangensis in history based on these two sets of markers; and (iii) compare the genetic divergence between varieties at these two sets of markers with different rates of gene flow.

Figure 1.

A, C, Distributions and B, D, networks of chlorotypes and mitotypes for Picea likiangensis and P. purpurea. The map is copied from a sample map (code: GS (2008) 1156) approved by National Administration of Surveying, Mapping and Geoinformation of China. Circles with different colors correspond to chlorotypes (A–G) in B and mitotypes (M1–M10) in D. Different types of lines indicate locations of the allopatric populations sampled for the three varieties.

1 Material and methods

1.1 Population sampling

Needle samples of Picea likiangensis were collected from 422 individuals in 42 natural populations, covering the entire distribution range of this species (Table 1). At least five pure populations based on morphological identification were collected for each of the three varieties and, in western Sichuan, more than 20 mixed populations with two or three varieties identified in each population were sampled (Fig. 1). In these mixed populations, we found that taxonomic definition of some individuals remained difficult because their morphological characters are intermediate between two varieties. However, we tentatively placed these individuals into one of three varieties according to their major morphological appearance. We also excluded all populations that are adjacent to the distribution range of P. purpurea (Du et al., 2011). However, we included three pure populations of this species in the present analyses. Between two and 20 individuals were collected for each population, and all individuals were at least 100 m apart. The location, elevation, and sample size for each population are given in Table 1.

Table 1.  Sampling sites, sample size, and haplotype distribution for 42 populations of Picea likiangensis and three populations of P. purpurea
Population codeSample location (All in China)Longitude (°E)Latitude (°N)Elevation (m)Sample sizeChlorotypesMitotypes
  1. GS, Gansu; QH, Qinghai; SC, Sichuan; TB, Tibet; YN, Yunnan.

P. purpurea
 1Langmusi, GS102°44.11′34°01.50′360012C (12)M10 (12)
 2Ruoergai, SC103°14.72′33°37.82′300012C (12)M10 (12)
 3Jiuzhaigou, SC103°39.48′32°52.74′315012C (12)M10 (12)
P. likiangensis
 4Xiangtang, SC100°58.55′32°16.05′3290 5A (4), B (1)M1 (2), M2 (3)
 5Seda, SC100°41.39′31°52.04′356020A (19), E (1)M2 (14), M3 (2), M4 (4)
 6Xiaojin, SC102°19.25′31°41.27′3830 1B (1)M2 (1)
 7Luhuo, SC100°19.20′31°36.67′358011A (11)M2 (3), M3 (8)
 8Daren, SC100°52.78′31°09.95′3120 9A (7), E (1), G (1)M2 (3), M3 (4), M4 (2)
 9Tagong, SC101°31.27′30°16.15′355010A (6), B (4)M2 (8), M3 (2)
10Daofu, SC101°36.29′30°31.55′3650 4B (4)M2 (4)
11Daofu, SC101°16.47′30°49.55′351012A (10), B (1), E (1)M1 (4), M2 (3), M3 (5)
12Yajiang, YN100°42.71′30°07.26′422011A (11)M1 (1), M3 (4), M4 (6)
13Yajiang, YN101°16.71′30°02.65′3470 5A (1), B (4)M2 (5)
14Kangding, SC101°55.11′29°52.04′3100 4A (2), B (2)M1 (1), M2 (2), M4(1)
15Yajiang, YN100°54.37′29°59.18′3610 9A (1), B (8)M2 (6), M4 (3)
16Litang, YN100°19.52′29°36.21′4180 7A (2), B (5)M2 (2), M3 (2), M4 (3)
17Sangdui, YN100°05.32′29°13.90′400012A (7), B (5)M2 (2), M3 (4), M4 (6)
18Xiangcheng, YN99°55.70′29°08.52′358011B (11)M2 (7), M4 (4)
19Zhongdian, YN99°44.99′28°18.91′384012B (12)M2 (10), M4 (2)
20Deqing, YN99°07.17′28°19.03390012B (12)M2 (1), M4 (5), M6 (2), M9 (4)
21Honglashan, TB98°39.79′29°12.47′372012A (5), B (6), G (1)M3 (11), M6 (1)
22Mangkang, TB98°09.56′29°36.49′407012A (10), B (2)M3 (1), M6 (11)
23Batang, TB98°37.73′29°43.75′400012A (10), B (2)M6 (12)
24Haizishan, TB99°33.13′30°18.28′446012A (9), B (2), E (1)M3 (12)
25Gugong, TB97°53.57′29°38.48′380012A (9), B (3)M7 (12)
26Chayu, TB97°11.09′29°17.45392011B (5), D (6)M7 (11)
27Ranwu, TB96°39.79′29°29.06′3476 9A (7), D (1), E (1)M1 (9)
28Galongla, TB95°41.89′29°48.15′361012B (5), D (7)M1 (12)
29Bomi, TB94°45.99′29°51.72′272012B (1), D (5), F (6)M1 (12)
30Sejila, TB94°43.67′29°41.96′320012B (5), D (7)M1 (12)
31Sejila, TB94°32.73′29°33.93′321010B (6), D (4)M1 (12)
32Linzhi, TB94°16.96′29°45.90′317011B (7), D (4)M1 (11)
33Milin, TB94°10.76′29°11.37′304012B (1), D (5), F (6)M1 (12)
34Milin, TB93°58.58′29°11.02′3000 9B (6), D (3)M1 (9)
35Leiwuqi, TB96°26.65′31°56.51′4303 5A (5)M1 (5)
36Langlashan, TB97°15.51′30°40.96′4331 7A (7)M1 (7)
37Nangqian, QH96°37.91′31°59.59′363310A (9), E (1)M1 (10)
38Yushu, QH96°55.47′32°15.84′3656 8A (8)M1 (8)
39Dege, TB98°49.31′31°55.59′3741 9A (6), B (2), E (1)M1 (9)
40Muli, SC101°19.87′28°07.27′355010B (10)M2 (3), M4 (7)
41Yanyuan, SC101°13.76′27°40.48′314011B (11)M4 (9), M5 (2)
42Diqin, YN100°01.16′27°35.74′300013B (13)M4 (1), M5 (2), M8 (10)
43Lijiang, YN100°16.19′27°12.95′320012B (12)M5 (12)
44Yulong, YN100°14.17′27°08.28′317017B (17)M4 (1), M5 (16)
45Baisha, YN100°12.17′27°02.31′3320 7B (7)M5 (7)

1.2 DNA extraction, amplification and sequencing

Genomic DNA was isolated from approximately 20 mg silica gel-dried needles using either a QIAGEN DNeasy Plant Mini Kit (QIAGEN, Valencia, CA, USA) or the modified cetyltrimethylammonium bromide (CTAB) procedure (Doyle & Doyle, 1990). We amplified and sequenced two mtDNA fragments using the previously reported primers nad1 intron b/c and nad5 intron 1, which have been shown to be polymorphic in other spruce species (Meng et al., 2007). Similarly, three cpDNA fragments were amplified and sequenced (i.e. trnL–trnF, trnS–trnG and ndhK/C) using primers suggested in previous studies (Taberlet et al., 1991; Hamilton, 1999; Anderson et al., 2006). Polymerase chain reactions (PCR) were performed in a volume of 25 μL, with each reaction containing 20–50 ng DNA, 50 mmol Tris-HCl, 1.5 mmol MgCl2, 0.5 mmol dNTPs, 2.5 μmol of each primer and 0.75 units Taq polymerase. The amplification and sequencing conditions were the same as those described previously (Meng et al., 2007; Du et al., 2011). All PCR products were sequenced with an ABI Prism Big Dye Terminator Cycle Version 3.1 sequencing Kit and an ABI 3130 × 1 or 3730 × 1 Genetic Analyzer (Applied Biosystems, Foster City, CA, USA).

Singletons were verified by repeated amplification and resequencing from the same DNA. Sequences were aligned using Clustal X (Thompson et al., 1997) or Clustal W implemented in MEGA 5.0 (Kumar et al., 2008).

1.3 Data analysis

Phylogenetic relationships were constructed among mtDNA haplotypes, as well as among cpDNA haplotypes, using median-joining networks with NETWORK version 4.6.0.0 (Bandelt et al., 1999; available at http://www.fluxus-engineering.com, accessed 11 March 2012). Estimated hierarchical partitioning of diversity between species, varieties, populations, and individuals was based on analysis of molecular variance (AMOVA; Excoffier et al., 1992) using ARLEQUIN version 3.0 (Excoffier et al., 2005), with significance tests based on 10 000 permutations. The proportion of genetic diversity due to allele frequency differences among populations (FST) was used to measure population differentiation within and between species. The average gene diversity within populations (HS), total gene diversity (HT), and the coefficients of differentiation GST and NST were estimated for each species for both mtDNA and cpDNA markers using PERMUT software (available at http://www.pierroton.inra.fr/genetics/labo/Software/Permut/, accessed 11 March 2012). We compared GST and NST using the U-statistic, which is approximated by a Gaussian variable by taking into account the covariance between GST and NST, and a one-sided test (Pons & Petit, 1996). The former considers only haplotype frequencies, whereas NST also takes into account differences between haplotypes. When NST is larger than GST, phylogeographic structure is obvious, which indicates that closely related haplotypes are found more often in the same area than less closely related haplotypes (Pons & Petit, 1996).

2 Results

2.1 Variations in cpDNA and haplotype distribution

One indel and 11 substitutions were detected across the three cpDNA fragments trnL–trnF, trnS–trnG, and nadhK/C (Table 2). These three sequences were combined into a total length of 1945 bp, and seven chlorotypes (A–G) could be identified among all individuals sampled (Table 2). Two distinct clades with six mutations were discerned in the chlorotype network (Fig. 1). The first clade included four chlorotypes (A, B, D, and E). Chlorotype A was fixed in 23 populations: at high frequency in five pure populations of Picea likiangensis var. rubescens and the remaining 18 mixed populations. Chlorotype B was widespread and shared by three varieties. However, all pure populations of var. likiangensis in the southeastern QTP were fixed by this chlorotype (Populations 40–45). This chlorotype was also found in var. linzhiensis at high frequency in the western region. Chlorotype E with the only step to B was found in var. rubescens (Populations 37 and 39) and five mixed populations. Chlorotype D was mainly present in seven populations of var. linzhiensis and two neighboring mixed populations (Populations 26 and 27). The other three chlorotypes (C, F, and G) belonged to the second clade. Chlorotype C was widely found in three populations of P. purpurea and Chlorotype G was present at low frequency in a mixed population near the contact zone of P. purpurea and P. likiangensis. However, Chlorotype F was found in two populations of var. linzhiensis (Populations 29 and 33) with extremely distant distributions to P. purpurea.

Table 2.  Variable sites of the aligned sequences of three chloroplast DNA fragments in seven haplotypes
 Nucleotide variable positions
trnS–trnG trnL–trnF nadhK/C
  1. *, GATTT.

 464477011123
 652924324675
Chlorotype463605419701
AAGATCCGACTA
BAGATCAGACTA
CCAC*TTCGGGGA
DAGAGCCGACTA
EAGATCAGACTG
FCAC*TTCAGGGA
GCAC*TCCGGGGA

2.2 Variations in mtDNA and haplotype distribution

Ten mitotypes (M1–M10) were identified across all samples from the examination of the sequence variation of two mtDNA fragments nad5 intron 1 and nad1 intron b/c (total 1557 bp in length; Table 3). Similarly, these mitotypes clustered into two distinct clades with more than 10 steps. One clade contained only M10, which is exclusively restricted to P. purpurea (Fig. 1). The remaining nine mitotypes of the other clade were found in P. likiangensis, and the network showed that M1 and M4 were located at the central position. The M1 mitotype was widespread in both var. rubescens and var. linzhiensis with extremely high frequency (100%) on the plateau platform. The pure populations of var. likiangensis in the southeast were fixed by a group of closely related mitotypes (M5, M8, M2, and M4). However, the mixed populations in western Sichuan and adjacent regions harbored seven of the 10 mitotypes with high diversity, and more than half of the populations in this regions fixed at least two mitotypes, especially Population 20, in which four mitotypes were found.

Table 3.  Variable sites of the aligned sequences of two mitochondrial DNA fragments in 10 haplotypes
 Nucleotide variable positions
nad5 nad1
  1. *, AGGCT; #, TTAAT; +, TGTCTAAAG; •, ATTTA; ×, CATA; Δ, CATT; ◊, AAATTA; □, AAAG; ○, TGTCT; ▪, AACGTAGTCGCTCGACCATAAGGGAGAGG.

 11440022222223366
 89001824556771415
Mitotype21068604373499081
M1TC+AC×GGCT
M2TC+AC×GTCT
M3TC+AAGGCT
M4TC+AA×GGCT
M5#TC+AA×GGCT
M6*#TC+AA×GGCT
M7TC+AC×GGCG
M8TA+AA×GGCT
M9*#TC+AC×GTCT
M10#GCCCΔTGAT

2.3 Genetic variations within and between varieties and species

The genetic diversity revealed by mtDNA markers was much higher than that found by cpDNA markers (Table 4). Among three varieties of P. likiangensis, HT and HS were highest for var. linzhiensis at the cpDNA markers (Table 4). However, both HT and HS were highest for var. likiangensis at the mtDNA markers. The highest between-population differentiation (GST) was found for var. likiangensis at cpDNA markers and for var. linzhiensis at mtDNA markers (Table 4). This was supported by the results of AMOVA, which showed that more cpDNA variations were partitioned between populations of var. likiangensis and more mtDNA variations were partitioned between populations of var. linzhiensis (Table 5). Within both species and varieties, FST was larger at mtDNA markers than at cpDNA markers (Table 5). However, phylogeographic structure is not obvious at either the variety or species level for both sets of genetic markers because all comparisons failed to detect larger NST than GST values (Table 4).

Table 4.  Estimates of average gene diversity within populations, total gene diversity, interpopulation differentiation, and number of substitution types
  H S H T G ST N ST
  1. H S, average gene diversity within populations; HT, total gene diversity; GST, interpopulation differentiation; NST, number of substitution types; cpDNA, chloroplast DNA; mtDNA, mitochondrial DNA.

cpDNA variation
Picea likiangensis 0.2850.6250.5440.400
 var. likiangensis0.1800.5180.6530.343
 var. rubescens0.2360.3780.3760.171
 var. linzhiensis0.3910.6620.4100.397
mtDNA variation
Picea likiangensis 0.2470.8110.6960.727
 var. likiangensis0.3020.8600.6490.796
 var. rubescens0.1710.8180.7900.811
 var. linzhiensis0.1030.5310.8050.685
Table 5.  Analysis of molecular variance of chloroplast DNA and mitochondrial DNA for Picea likiangensis
SpeciesSource of variation d.f. SS VC V% F-statistics
  1. *0.01 ≤P < 0.05; **0.001 ≤P < 0.01; ***P < 0.001.

  2. d.f., degrees of freedom; SS, sum of squares; VC, variance component; V, variation; FCT, correlation of species relative to total; FST, correlation within populations relative to total; FSC, correlation within populations relative to species.

Chloroplast DNA
 P. likiangensis Among all varieties245.550.149017.58 F CT= 0.1758***
 Among populations within varieties64118.380.221626.15 F SC= 0.3173***
 Within populations355169.290.477056.27 F ST= 0.4373***
 var. likiangensisAmong populations2933.910.159443.90 F ST= 0.4390***
 Within populations15431.370.203756.10 
 var. rubescensAmong populations229.270.035113.67 F ST= 0.1367**
 Within populations10924.200.222086.33 
 var. linzhiensisAmong populations1375.200.616433.27 F ST= 0.3327***
 Within populations92113.721.236166.73 
Mitochondrial DNA
 P. likiangensis Among all varieties275.0280.13874.98 F CT= 0.0498*
 Among populations within varieties64865.702.094275.14 F SC= 0.7908***
 Within populations355196.700.554319.88 F ST= 0.8012***
 var. likiangensisAmong populations29479.942.629780.88 F ST= 0.8088***
 Within populations15495.760.621819.12 
 var. rubescensAmong populations22302.772.305277.26 F ST= 0.7726***
 Within populations10973.970.678622.74 
 var. linzhiensisAmong populations1382.980.825373.80 F ST= 0.7380***
 Within populations9226.950.292926.20 

Analyses using AMOVA revealed that the highest genetic variation partitioned between var. likiangensis and var. rubescens (33.07%) at cpDNA markers, whereas that between var. likiangensis and var. linzhiensis (8.67%) occurred at mtDNA markers (Table 6). In addition, AMOVA analyses suggested that genetic variation partitioned between three varieties of P. likiangensis is higher at cpDNA markers (17.58%) than at mtDNA markers (4.98%; Table 5). However, genetic variation partitioned between P. likiangensis and P. purpurea is lower at cpDNA markers (76.85%) than at mtDNA markers (91.73%; Table 7). This trend can also be seen from the networks of both markers. The mtDNA variations separated two species distinctly; however, two clades of cpDNA were found for P. likiangensis, but only one was found for P. purpurea (Fig. 1).

Table 6.  Analysis of molecular variance of chloroplast DNA and mitochondrial DNA between three varieties of Picea likiangensis
Source of variation d.f. SS VC V% F-statistics
  1. *0.01 ≤P < 0.05; **0.001 ≤P < 0.01; ***P < 0.001.

  2. d.f., degrees of freedom; SS, sum of squares; VC, variance component; V, variation; FCT, correlation of species relative to total; FST, correlation within populations relative to total; FSC, correlation within populations relative to species.

Chloroplast DNA
 Between var. likiangensis and var. rubescens125.300.157733.07 F CT= 0.3308***
   Among populations within varieties5143.180.107822.61 F SC= 0.3379***
   Within populations26355.570.211344.31 F ST= 0.5569***
   Total315124.030.4768  
 Between var. likiangensis and var. linzhiensis119.720.120511.80 F CT= 0.1180*
   Among populations within varieties42109.110.310530.42 F SC= 0.3449***
   Within populations246145.100.589857.78 F ST= 0.4222***
   Total289273.931.0208  
 Between var. rubescens and var. linzhiensis123.170.170815.10 F CT= 0.1510***
   Among populations within varieties3584.470.273824.22 F SC= 0.2852***
   Within populations201137.920.686260.68 F ST= 0.3932***
   Total237245.561.1308  
Mitochondrial DNA
 Between var. likiangensis and var. rubescens135.720.10423.21 F CT= 0.0321
   Among populations within varieties51782.722.495176.90 F SC= 0.7945***
   Within populations263169.730.645419.89 F ST= 0.8011***
   Total315988.173.2446  
 Between var. likiangensis and var. linzhiensis151.200.23688.67 F CT= 0.0867*
   Among populations within varieties42562.921.995673.07 F SC= 0.8000***
   Within populations246122.710.498818.26 F ST= 0.8174***
   Total289736.822.7313  
 Between var. rubescens and var. linzhiensis123.220.06893.08 F CT= 0.0308
   Among populations within varieties35385.751.667574.49 F SC= 0.7686***
   Within populations201100.920.502122.43 F ST= 0.7757***
   Total237509.882.2385  
Table 7.  Analysis of molecular variance of chloroplast DNA and mitochondrial DNA between Picea likiangensis and P. purpurea
Source of variation d.f. SS VC V% F-statistics
  1. *0.01 ≤P < 0.05; **0.001 ≤P < 0.01; ***P < 0.001.

  2. d.f., degrees of freedom; SS, sum of squares; VC, variance component; V, variation; FCT, correlation of species relative to total; FST, correlation within populations relative to total; FSC, correlation within populations relative to species.

Chloroplast DNA
 Between P. likiangensis and P. purpurea  1301.064.436376.85 F CT= 0.7685***
 Among populations within species 43251.450.49518.58 F SC= 0.3705***
 Within populations414347.440.841314.57 F ST= 0.8543***
 Total458899.955.7727  
Mitochondrial DNA
 Between P. likiangensis and P. purpurea  11921.1728.603491.73 F CT= 0.9173***
 Among populations within species 43853.531.89086.06 F SC= 0.7334***
 Within populations414284.540.68732.20 F ST= 0.9780***
 Total4583059.2431.1814  

3 Discussion

In the present study, we detected cpDNA introgressions from Picea purpurea into the distant distributions of P. likiangensis, whereas no interspecific introgression was found at mtDNA markers in the allopatric distributions of these two species. Geographic distributions of both cpDNA and mtDNA haplotypes suggest that P. likiangensis survived in multiple refugia throughout its range during the LGM and that the following postglacial expansions may have occurred mainly in limited parts along the distributional edges of this species, where a single chlorotype or mitotype was fixed in the adjacent populations. In addition, we found that genetic differentiation between the three varieties is higher at cpDNA than mtDNA markers, possibly due to incomplete lineage sorting within the mtDNA. These results seem to suggest that different evolutionary forces and dispersal biology together may have shaped the genetic architecture of P. likiangensis at cpDNA and mtDNA genomes across its current distributions.

3.1 Interspecific introgressions

As suggested in our previous study (Du et al., 2011), two deep lineages detected in P. likiangensis and P. purpurea at both cpDNA and mtDNA markers should have originated from introgressions between species rather than by retention of ancestral polymorphisms. In their contact zones, although introgressions at both markers are bidirectional, more introgressions occurred from the expanding P. purpurea to the local P. likiangensis. In addition, massive introgression had taken place at mtDNA markers with low gene flow (Du et al., 2011). All these findings in the contact zones are largely consistent with neutral modeling results (Currat et al., 2008). However, in the present study, an extensive examination of all allopatric populations of P. likiangensis across its distribution range, we only detected two deeply diverged cpDNA lineages, with one of them clustered with the chlorotype of P. purpurea (Fig. 1). Two chlorotypes (G and F) of P. likiangensis in this lineage were distributed in two distant populations (29 and 33, respectively) in Tibet and a relatively nearby population (8) to P. purpurea. In addition, AMOVA analyses suggested lower genetic differentiation between P. likiangensis and P. purpurea at cpDNA markers (76.85%) than at mtDNA markers (91.73%; Table 7). These results are in contrast with the introgressions found in the contact zone (Du et al., 2011). We suggest that long-distance dispersals resulted in these recovered interspecific introgressions at cpDNA markers for two reasons. First, at the contact zone, both mtDNA and cpDNA introgressions can take place freely without spatial restrictions. However, in the case of allopatric distributions, only cpDNA can be infrequently dispersed by pollen into distant ranges in conifers. Second, the high intraspecific cpDNA gene flow will decrease introgressions by reducing genetic drifts in the contact zones (Currat et al., 2008). Such ‘rescuing’ roles are obviously reduced in remote populations of the hybridizing species. In fact, it has been reported that long-distance dispersal reduces intraspecific differentiation caused by genetic drift in other conifers (Liepelt et al., 2002). Our results further suggest that such long-distance dispersal may have blurred interspecific delimitations in allopatrically distributed populations. In addition, our findings indicate that the direction and extent of introgression predicted by the recent neutral model (Currat et al., 2008) should only be fitted to the contact zones of hybridizing species (Du et al., 2011) and that these predications cannot be extended to long-distance introgressions.

3.2 Glacial refugia and postglacial expansion

Except for two previously mentioned chlorotypes (G and F; Du et al., 2011) possibly introgressed from P. purpurea, we further recovered four chlorotypes (A, B, D, and E) specific to P. likiangensis. Chlorotype A was found mainly in the northern range of this species, mostly in populations identified as var. rubescens, whereas Chlorotype B was found in southeastern and western ranges, mostly in var. likiangensis and partly in var. linzhiensis. Chlorotype D occurred mostly in var. linzhiensis and two mixed populations at the western range, whereas Chlorotype E was restricted to one mixed population of three varieties. However, nine mitotypes were recovered for P. likiangensis. Because mitotypes M1, M2, and M4 were located at the central position of the network and because they were the most common mitotypes, they probably represented the most ancient haplotypes (see Donnelly & Tavaré, 1986; Crandall & Templeton, 1993). Furthermore, as expected, these mitotypes show the broadest distribution. The M1 mitotype was fixed in the northern and western ranges, at a frequency of 100% in most populations of var. rubescens and var. linzhiensis, whereas the M2 and M4 mitotypes occurred in multiple populations in the eastern and southeastern distributions of the species, respectively, identified as var. rubescens and var. likiangensis. The M3 mitotype, which is also frequent, occurred in central populations (var. rubescens and var. likiangensis). All other mitotypes had very limited distributions. The M5 and M8 mitotypes occurred only on isolated mountain massifs in the southeastern periphery of the QTP (var. likiangensis). The remaining three mitotypes (M9, M6, and M7) occurred in one or two populations in central distributions of the species. Because of the low mutation rates of cpDNA and mtDNA in conifers (Wolfe et al., 1987; Willyard et al., 2007), it is likely that most of the mitotype and chlorotype variation recorded across both species originated before the LGM. The given relief matrix and the temperature ranges of Picea lend support to the proposal that the distribution range of this species was not subjected to the extinguishing effects of the Ice Ages. Instead we can assume ecological stability in this region with moderate temperature changes (see Schmidt et al., 2011; Miehe et al., 2011). Thus, the recovery of chlorotypes and mitotypes in different regions or populations indicates that multiple glacial refugia existed for this species throughout its current range during the LGM (Hewitt, 2004; Petit & Vendramin, 2007). Most telling in this respect are the patterns of the M5 mitotype. Because mtDNA is maternally inherited, seed dispersal is responsible for this pattern. It is likely that birds founded the Picea forests on the island-like mountain massifs. Because M5 and M8 mitotypes seem to be derived from the M4 mitotype and do not occur outside the mountain massifs, it is likely that they evolved as separate lineages in situ. Because they have reached frequencies of up to 100% in these populations, it is likely that they represent a fairly old pattern. Similarly, M6 and M7 are mitotypes fixed in one or two populations, reaching up to 100% frequency in these populations. This pattern is similar to the results of phylogeographic analyses reported for other species on the QTP (e.g. Opgenoorth et al., 2010; Wu et al., 2010). However, a history of a range-wide colonization since the LGM appears to occur for some other conifers in the northeastern QTP (Zhang et al., 2005; Meng et al., 2007). For P. likiangensis, we found that such a colonization that led to the fixture of the same haplotype in the adjacent populations occurred only along the distributional edge of the species. This local and edge colonization differs in terms of the genetic imprints of chlorotype and mitotype. For chlorotype distribution, northern and southeastern regions were fixed by a single haplotype, indicating a likely recent colonization in these two regions. However, a common postglacial expansion appears to have occurred in northern and western regions, which were exclusively fixed by a single mitotype (M1). In the southeastern region, mitotype colonization seems to be smaller because only three southern populations were fixed by a single mitotype (Fig. 1). This difference may have originated from an interaction between colonization mediated mainly by seeds and dispersal mediated mainly by pollen. It is likely that only mitotype imprints may have mirrored the true phylogeographic process in the past, whereas wind-dispersed pollen may have blurred these genetic signatures, as in other conifers (Liepelt et al., 2002).

3.3 Genetic differentiation between varieties

Although long-distance dispersal by pollen may blur the interspecific delimitation (e.g. between P. likiangensis and P. purpurea, as discussed above) and phylogeographic process within the species, it is interesting that genetic divergence between three varieties is higher at cpDNA than mtDNA markers. This finding seems to be consistent with previous studies on other conifers based on these two sets of markers (Du et al., 2009; Zhou et al., 2010). Because gene flow is lower for mtDNA than for cpDNA, variety (or species)-specific variations may take longer to be extensively fixed within a certain taxon for mtDNA markers than cpDNA markers (Wright, 1943; Hoelzer, 1997; Petit & Vendramin, 2007). However, a faster DNA mutation rate may also result in rapid fixation of variety (or species)-specific alleles through genetic drift (Sloan et al., 2008). Until now, although it has been difficult to obtain accurate mutation rates for cpDNA and mtDNA in these two spruce species, more variations have been found for mtDNA than for cpDNA and the sequenced fragments are shorter for mtDNA than cpDNA. Therefore, our results obviously suggest that DNA markers with a high rate of gene flow are often effective in delimitating closely related species (Du et al., 2009; Zhou et al., 2010; Hollingsworh, 2011; Wang et al., 2011). If this is the case, some cpDNA variations (e.g. Chlorotype D) found for var. linzhiensis may represent effective barcoding markers for this evolutionary unit. Further population studies of genetic variation based on nuclear loci (e.g. Chen et al., 2010; Li et al., 2010) are needed to confirm whether three morphological varieties with specific chlorotypes comprise independent evolutionary or taxonomic units.

Acknowledgements 

The authors thank Richard ABBOTT (School of Biology, University of St Andrews, St Andrews, UK) and Richard MILNE (Institute of Molecular Plant Sciences, The University of Edinburgh, Edinburgh, UK) for comments on a previous version of the manuscript. The field explorations were aided by Qin WANG (State Key Laboratory of Grassland Agro-Ecosystem, College of Life Science, Lanzhou University, Lanzhou, China) and Shi-Long CHEN and Shen-Yun CHEN (Laboratory of Qinhai–Tibet Biological Evolution and Adaptation, Northwest Plateau Institute of Biology, Chinese Academy of Sciences, Xining, China). This research was supported by grants from the National Natural Science Foundation of China (30930072 and 40972018), the Key Project of International Collaboration Program, the Ministry of Science and Technology of China (2010DFB63500), and the International Collaboration ‘111’ Project (to JQL).

Ancillary