The Population Genetics of Cultivation: Domestication of a Traditional Chinese Medicine, Scrophularia ningpoensis Hemsl. (Scrophulariaceae)

Background Domestic cultivation of medicinal plants is an important strategy for protecting these species from over harvesting. Some species of medicinal plants have been brought into cultivation for more than hundreds years. Concerns about severe loss of genetic diversity and sustainable cultivation can potentially limit future use of these valuable plants. Genetic studies with comprehensive sampling of multiple medicinal species by molecular markers will allow for assessment and management of these species. Here we examine the population genetic consequences of cultivation and domestication in Scrophularia ningpoensis Hemsl. We used chloroplast DNA and genomic AFLP markers to clarify not only the effects of domestication on genetic diversity, but also determine the geographic origins of cultivars and their genetic divergence from native populations. These results will allow both better management of cultivated populations, but also provide insights for crop improvement. Results Twenty-one cpDNA haplotypes of S. ningpoensis were identified. Wild populations contain all haplotypes, whereas only three haplotypes were found in cultivated populations with wild populations having twice the haplotype diversity of cultivated populations. Genetic differentiation between cultivated populations and wild populations was significant. Genomic AFLP markers revealed similar genetic diversity patterns. Furthermore, Structure analysis grouped all wild populations into two gene pools; two of which shared the same gene pool with cultivated S. ningpoensis. The result of Neighbor-Joining analysis was consistent with the structure analysis. In principal coordinate analysis, three cultivated populations from Zhejiang Province grouped together and were separated from other cultivated populations. Conclusions These results suggest that cultivated S. ningpoensis has experienced dramatic loss of genetic diversity under anthropogenic influence. We postulate that strong artificial selection for medicinal quality has resulted in genetic differentiation between cultivated and wild populations. Furthermore, it appears that wild populations in Jiangxi-Hunan area were involved in the origin of cultivated S. ningpoensis.


Introduction
Plant domestication is one of the great milestones in human history. The process of domestication represents a continuum of increasing codependence between plants and people [1]. Crop species have been derived from their wild progenitors as a result of artificial selection for desirable traits by early farmers. In the light of crop domestication, a number of agencies now are recommending that wild medicinal plants be brought into cultivation systems [2,3]. Currently, more than 80% of the world's population in developing countries depends primarily on herbal medicine for basic healthcare needs [4,5]. Given the demand for a reliable and uniform supply of medicinal plants and the accelerating depletion of forest resources, cultivation of medicinal plants species would be an important strategy and a viable alternative to harvesting of natural populations [6,7].
At present some medicinal plants are grown in home farms, and some are cultivated as field crops. But the number of medicinal plants species currently in high intensity cultivation for commercial production are varied world-wide. In India, 20 species are currently under cultivation, in Hungary 40 and in Germany 100 [6,8,9,10]. China, with its history of thousands years of medicinal plant use cultivates only about 250 species [11,12]. Considering the harvest pressure on wild populations and the increasing demands, the most popular medicinal plants are cultivated under the supervision of Chinese government on large scale reaching almost 1 million acres [13].
As the demand for medicinal plant cultivation increased, many studies examined the impact of cultivation on genetic diversity, species such as Artemisia judaica [14], Scutellaria baicalensis [15], Corydalis yanhusuo [16], Coptis chinensis [17], Magnolia officinalis subsp. biloba [18], Fritillaria cirrhosa [19]. However, all these studies only used single molecular marker or local sampling which just revealed genetic diversity patterns in limited populations. Since genetic diversity underlies the plasticity of secondary metabolism and hence the production of medicinal compounds [20], understanding the population genetics across the range of wild populations is essential for utilizing native resources for improvement of cultivated species. A study with comprehensive sampling of both cultivated and wild populations of medicinal species, using both chloroplast and nuclear DNA markers, can clarify not only changes in genetic diversity, but also provide information on domestication including the geographic origins, ancestor populations and the impact of domestication on population genetics.
Plant secondary metabolites are responsible for phytochemical diversity and therapeutic efficacy of medicinal species. The production of secondary compounds is influenced by both environmental and genetic factors [20,21]. Thus geographically distinct populations even of the same species may have vastly different medical qualities [22]. Modern methods of phytochemical methods are used to discriminate geographically dispersed cultivars within a medicinal species [23,24,25,26,27]. Yet little is known about the genetic differences among the wild populations of a medicinal species and whether cultivated populations with high pharmaceutical quality have already significantly diverged from wild populations. Additionally, understanding genetic changes associated with cultivation not only helps control the quality of medicinal herbs, but also provides practical information essential for formulating appropriate conservation and cultivation management strategies [28,29].
Scrophularia ningpoensis Hemsl. is a perennial herb native to Southeastern China. Wild populations are distributed in forests along streams, thickets, and tall grasses below alt. 1700 m and reproduce from seed [30]. Roots of this medicinal herb have a long history of widespread use in China to treat inflammation, laryngitis, tonsillitis, abscesses and constipation [31,32,33]. The first recorded use in ancient Chinese literature is from ca. 100 BC [34]. Harpagoside, angroside C, acteoside and cinnamic acid are the main bioactive components [35,36,37,38,39]. According to the Chinese Pharmacopoeia, only the root of S. ningpoensis is listed as the Radix Scrophulariae [40]. Our recent field investigations indicated that wild populations of this herb have suffered rapid declines, and the species is extirpated in many locations due to over exploitation and deterioration of habitats. An analysis of the population genetics of wild S. ningpoensis populations is important and urgent for appropriate utilization, conservation and preservation of the wild resources.
Today, S. ningpoensis is cultivated for its roots on a large scale in several regions of Central and Southeastern China. This species can trace its cultivation history back to Song dynasty (1000 years ago) beginning in backyard gardens [41]. In spite of this long written history, the geographic origin of cultivation has not been recorded. Formal cultivation programs were initiated in China only since the 1950's [42]. Unlike wild populations, cultivated S. ningpoensis is propagated vegetatively from rhizome. During the cultivation, large rhizomes are often selected for propagation; flowers are removed before opening so that resources are allocated to vegetative growth, in particular the rhizomes [43,44]. Our previous study used HPLC revealed that among cultivated S. ningpoensis, accessions from Zhejiang Province produced higher concentrations of three main bioactive compounds, suggesting that the best quality of cultivated S. ningpoensis is from Zhejiang [45].
These results are consistent with the general perspective that S. ningpoensis from Zhejiang has the best quality for medical use [46]. Besides the phytochemical differences, the unresolved question of whether cultivated S. ningpoensis from Zhejiang is genetically differentiated for others is also important and indispensable to sustainable cultivation of S. ninpoensis.
Here we used two chloroplast DNA fragments and AFLP markers to study the population genetics and phylogeography of wild and cultivated S. ningpoensis with the aims of: (1) evaluating the genetic erosion of cultivated populations and the change in the pattern of genetic diversity under artificial selection; (2) unraveling the geographical origins and ancestral populations of cultivated S. ningpoensis; (3) clarifying the genetic divergence between cultivated populations from Zhejiang Province and the rest of cultivated S. ningpoensis; and (4) determining which wild populations are valuable and can be introduced for improvement of cultivation.

Ethics Statement
Management Bureau of Mt. Jinggang National Nature Reserve issued the permit for Jinggang Mountain (JGW); Management Bureau of Mt. Tianmu National Nature Reserve issued the permit for Tianmu Mountain (TNW, TM1W, TM2W); Management Bureau of Mt. Dapan National Nature Reserve issued the permit for Dapan Mountain (TWW). No specific permissions were required for other locations which are neither privately owned nor protected and the field study did not involve endangered or protected species.

Sampling and DNA extraction
Field studies were conducted throughout distribution range of Scrophularia ningpoensis. Specific locations (GPS coordinates) of our field work are indicated in Table 1. A total of 364 individuals representing 13 cultivated and 14 wild populations were sampled with nine to twenty-two individuals collected randomly for each population (Table 1, Figure 1). Leaf material was dried in silica gel and stored at 4uC. Voucher specimens are deposited in the Herbarium of Zhejiang University (HZU), China. Total genomic DNA was extracted using a modified (CTAB) method [47,48].

Chloroplast (cp) DNA sequences analysis
After preliminary screening for variation in the intergenic spacer (IGS) regions of chloroplast DNA, two regions: trnL-trnF [49] and psbA-trnH [50] were chosen for analysis because they contained sufficient levels of informative polymorphic sites. PCR amplification was performed in an ABI 9700 thermocycler (Applied Biosystems). Cycling conditions were 95uC for 2 min followed by 35 cycles of 94uC (45 s), 57uC (45 s), 72uC (1 min), and concluding extension at 72uC (10 min). The reaction mixture (25 mL) contained 2.5 mm/L MgCl 2 , 0.5 mm/L dNTP, 2.5 ml 106buffer, 2.5 mm/L primer, 1 U Taq DNA polymerase (Takara Bio Inc.) and 20 ng DNA template. PCR products were identified on 1.5% agarose gels in a 0.56TBE buffer, stained with ethidium bromide (0.5 mg/ml), visualized with ultraviolet light and photographed. DNA Marker DL2000 (Takara Bio Inc.) was used. The PCR products with a single band were cleaned by a GENE-CLEAN II Kit (BIO 101, Inc. Carlsbad, USA) for direct sequencing. Samples were sequenced in forward and reverse directions using the dideoxy chain termination method with fluorescent labeling with Big Dye Terminator (version 1.1). Sequences were determined with an ABI 377XL DNA sequencer and edited using Sequencher version 4.0 (Gene Codes Corp., Ann Arbor, MI, USA). Sequences of the two IGS regions were combined and aligned by Clustal W version 1.8 [51].
Sequences from all individuals were characterized for their cpDNA haplotypes. Chloroplast DNA haplotype diversity (h) and nucleotide diversity (p) [52] were calculated for each population (h S , p S ) and overall (h T , p T ) using DNASP 5.10 [53]. A haplotype network rooted by Scrophularia buergeriana that contains all linkages with .95% probability of being most parsimonious was drawn by TCS version 1.21 [54]. In this analysis, indels and one inversion were treated as single mutation events. Phylogenetic relationships among cpDNA haplotypes of S. nignpoensis were assessed by Bayesian inference (BI), maximum likelihood (ML), and maximum parsimony (MP) methods with S. buergeriana, S. spicata and S. dentate as the outgroup. For Bayesian inference, the substitution model was determined by MrModeltest 2.2 [55] and the best-fit model (F81+I+G) was selected by Akaike information criterion (AIC). Then Bayesian analysis was performed by MrBayes 3.1.2 [56]. Two independent runs of four Markov chains each starting with a random tree were processed in ten threads simultaneously for 50 million generations, sampling trees *In AFLP analysis, four populations RCC, YCC, SHC and DPC are combined as one population coded PAC, and in Figure 1C these four populations are also indicated as PAC. doi:10.1371/journal.pone.0105064.t001 at every 1000 generation. The first 12500 sampled trees (25%) were discarded as burn-in samples. Maximum likelihood [57] tree and ML bootstrap searches were conducted using RAxML V.7.2.8 [58] on the CIPRES [59]. Final tree was evaluated and optimized under GAMMAGAMMA Model. ML bootstrap values were estimated from 1000 bootstrap replicates. Maximum parsimony analyses were conducted using PAUP* version 4.0b10 [60]. Support values for the relationships were calculated by performing bootstrap analyses of 1000 heuristic search replicates using the TBR branching swapping algorithm with 100 random additions per replicate. The geographical distribution of haplotypes was plotted on a map of China using Mapinfo (Pitney Bowes, Inc.). To test for significant difference in haplotype frequencies between wild and cultivated populations, total diversity (h T ), withinpopulation diversity (h S ), U test (Nst/Gst) and population differentiation (G ST ) were calculated in HAPLONST [61]. Hierarchical structure of genetic variation was estimated by an analysis of molecular variance (AMOVA) on three levels: among groups (cultivated and wild groups); among-population within groups; and within-population. All AMOVAs were performed with ARLEQUIN [62]. Sequence variation was tested for deviations from neutrality by Tajima's D statistic [63], and by Fu and Li's D* and F* statistics [64] using DNASP 5.10 [65].
Raw AFLP data were scored and collected as presence/absence matrix firstly by Genemarker v1.7 (SoftGenetics LLC) and then each locus was manually inspected. Only unambiguously detectable fragments in the size range 75-450 bp were scored. To test the reproducibility of the AFLP data, the complete AFLP procedure (digestion, ligation, pre-selective, selective amplification) was repeated for 15 individuals (5% of the samples) and scored independently for each primer pair.
Assuming a set of measures of wild and cultivated populations, population genetic statistics were generated by the program POPGENE version 1.31 [68] including: the number of polymorphic fragments (Fp), the percentage of polymorphic fragments (PPF), Nei's [67] gene diversity (h), Shannon's diversity index (H), coefficient of gene differentiation (G ST = total genetic diversitygene diversity within population/total genetic diversity), and the level of gene flow: N m = 0.5(12G ST )/G ST . To explore the relationships among individuals Principal Coordinate Analysis (PCoA) was performed by software MVSP version 3.1 (Kovach Computing Services, Anglesey, Wales) from pairwise Euclidian distances between individuals genotypes. A neighbor-joining (NJ) tree rooted by the populations of Scrophularia buergeriana was constructed by the PHYLIP 3.62 package [69] and group support was assessed by a bootstrap analysis with 1000 replicates. Genetic admixture analysis was implemented in STRUCTURE version 2.2.3 [70] with ''no admixture model'' and assumed ''uncorrelated allele frequencies'' [71]. The main modeling assumptions of STRUCTURE are Hardy-Weinberg equilibrium within populations [70]. But the cultivated populations are Hardy-Weinberg disequilibrium because of extensively clonal reproduction by rhizome within population. So instead of using all cultivated populations, only one individual from each different cultivated population (total ten individuals) were used for STRUCTURE analysis. The number of clusters (K) was set from 2 to 14 (number of all wild populations) with 10 replicate runs for each K, a burn-in period of 100000, and additional 1000000 replicates of the MCMC chains after burn-in. The structure computations output files were carried out by the freely accessible STRUCTURE HARVESTER [72] to calculate similarity coefficients between the replicate runs and to plot the mean logarithmic likelihood of K values (the means of the estimated log posterior probability of the data over the replicate runs for each K value). All populations were divided into two groups (cultivated and wild) and analysis of molecular variance (AMOVA) [73] was conducted by ARLE-QUIN [62] to quantify genetic differentiation at different hierarchical levels. ''Frequency-down-weighted marker values'' (DW) [74] quantified by the amount of particular AFLP markers in the total dataset were calculated by AFLPdat [75].  (Table 2); 21 were single-site mutations, four were length polymorphisms (4 bp, 9 bp, 7 bp and 2 bp, respectively) and one was an inversion (22 bp). Considering that sequencing poly-N regions could easily cause homoplasies due to polymerase error, length variations in mononucleotide repeats (poly A or T stretches) were not treated as polymorphisms. The trnH-psbA region (21 polymorphic sites detected in 466 aligned positions; 4.5%) was more variable than the trnL-trnF region ( Haplotype diversity and within-population diversity declined under anthropogenic influence during the course of cultivation (Table 3). Wild populations contained all haplotypes, whereas cultivated populations only had three haplotypes (A, B and E). Moreover, in wild S. ningpoensis, eleven out of fourteen populations (78.6%) were polymorphic for haplotypes. In contrast, every cultivated population was fixed for a single haplotype. Wild populations had much higher estimates of haplotype diversity (h T = 0.919), the within-population component of diversity (h S = 0.444) and nucleotide diversity (p T = 0.00301) than do cultivated populations (h T = 0.399 h S = 0, p T = 0.00076). In wild S. ningpoensis, haplotype diversity varied: populations LSW, TM2W had the highest haplotype diversity whereas populations TWW, PTW, THW from Zhejiang were each fixed for a single haplotype ( Figure 1C). An AMOVA indicated that 41.02% of the total genetic variation occurred between cultivated and wild groups (Table 4), which is consistent with the high observed genetic differentiation (F ST = 0.71988) between cultivated and wild groups. And in cultivated S. ningpoensis all genetic variation was among populations, but in the wild populations, more variation occurred within populations (58.54%) than among wild populations (41.46%).

CpDNA sequence data
Maximum parsimony analysis of these 21 haplotypes, with S. buergeriana, S. spicata and S. dentata as the outgroup, resulted in 501 most parsimonious trees with a length of 87, a consistency index (CI) of 0.828, and a retention index (RI) of 0.583. Maximum likelihood optimization resulted in a final optimization likelihood of 22020.22, with the alpha parameter being 0.020000. The bestscoring ML tree had a length of 0.073087. The Maximum likelihood tree with statistical support indicated above branches was presented in Figure 1A, haplotypes of A, B, C, C1, C2, C3, D, D1 and D2 were monophyletic but with very low bootstrap support. The MP/BI/ML analyses yield mostly congruent topologies. Haplotypes (A, B, E) which cultivated populations harbored occur in different clades. The cpDNA haplotype network for the 27 populations of S. ningpoensisis is shown in Figure 1B. The rooted TCS network ( Figure 1B) showed a similar structure to the phylogenetic tree ( Figure 1A), which grouped 21 haplotypes into two distinct clades. The outgroup S. buergeriana is linked with the center of the whole network by two missing haplotypes. Two most common haplotypes A and B are on tip position linked with haplotype C and D to form one clade, while the majority of rare haplotypes forms a ''star'' shape dominated by haplotype E. The geographic distribution and haplotype frequencies within populations are shown in Figure 1C. The most frequent haplotypes (A, B and E) were shared by both wild and cultivated groups, while other remaining haplotypes were found only in wild populations ( Figure 1C). The most common and widest geographically distributed haplotype in cultivated populations was A which was unique to the wild population LSW located in Lenshui County in northeastern Jiangxi Province. In wild populations, there was no clear geographical pattern of withinpopulation diversity ( Figure 1C). Moreover, U test (Nst/Gst) indicated no phylogeographic structure in wild S. ningpoensis

AFLP data
After exclusion of ambiguous or irreproducible fragments, 289 fragments were consistently scored for 24 populations (306 individuals) of which 261 fragments (90.31%) were polymorphic. The test for reproducibility yielded a mean error rate of 1.7% and confirmed the reproducibility of the AFLP fragments. Wild populations had much higher estimates of genetic diversity (h = 0.0791-0.1614) than did cultivated populations (h = 0.0076-0.0875) ( Table 5). In wild S. ningpoensis, the percentage of polymorphic fragments varied from 22.15% in population TWW to 50.87% in population LAW; in cultivated S. ningpoensis, the percentage of polymorphic fragments varied from 3.11% in population SCC to 27.68% in population NSC ( Table 5).
The variation between cultivated and wild groups accounted for 19.62% of the overall genetic variation (Table 6). In the cultivated group, most variation (76.97%) existed in among-population with only 23.03% due to the within-population (Table 6). But the variation distribution in the wild group was opposite that most variation (60.40%) existed in within-population with only 39.60% due to the among-population (Table 6).
Principal coordinate analysis also showed a clear differentiation between cultivated and wild S. ningpoensis. All populations were grouped into three main clusters ( Figure 2): cluster I included most wild populations; cluster II included most cultivated populations and two wild populations LSW & HNW; and cluster III included three cultivated populations: PAC, XJC and GZC. Populations PAC and XJC were from Zhejiang province. While GZC is in Fujian Province, local farmers indicated that it was directly introduced from Zhejiang cultivated populations. Genetic admixture analysis performed by STRUCTURE divided genetic variation into two gene pools (k = 2, Figure 3B): (1) Cultivated individuals, wild population HNW and part of wild population LSW; (2) Other wild populations. In STUCTURE analysis K values ranged from 2 to 14 and the one with the highest likelihood (k = 2) was chosen (Figure 4).
The Neighbor-Joining analysis identified two wild populations (HNW & LSW) that had a close relationship with cultivated S. ningpoensis. The wild population HNW was sister to cultivated populations with a bootstrap value of 84%; LSW clustered with HNW and all cultivated populations supported by 100% bootstrap value ( Figure 3A). Although three cultivated populations, PAC, XJC and GZC, grouped together and are distinct from other S. ningpoensis in PCoA analysis, in the NJ tree they located in one clade with hundred percent support and were associated with other cultivated populations.  Table 4. Hierarchical analysis of molecular variance for 28 populations of Scrophularia ningpoensis based on chloroplast DNA sequences.

Change of genetic diversity under domestication
For many crop species the domestication process has left a signature in the pattern of genetic diversity as a result of founder effects, genetic bottlenecks and artificial selection [76]. Many crops show evidence of genome-wide reductions in variation [77,78,79]. In our study, the pattern of genetic diversity between cultivated and wild populations of Scrophularia ningpoensis is consistent with other domesticated crop species, even though the domestication history of S. ningpoensis is only one thousand years. Out of twenty-one cpDNA haplotypes in the species, only three haplotypes were detected in cultivated populations. Moreover, cultivated populations had little within-population genetic polymorphisms ( Table 3). The cpDNA nucleotide diversity (p T ) and haplotype diversity (h T ) of cultivated populations were much less than that of the wild group (Table 3). AFLP data also support these patterns (Table 5). These results are consistence with many crop-relative systems of domestication [80,81] although the degree of decline in diversity of S. ningpoensis with such a short history of domestication is surprising. This medicinal plant has lost genetic diversity as a result of domestication which clearly shows that a species can have large genetic alterations even before there are significant changes in morphology, phenology or breeding system. Other species native to China, such as Corydalis yanhusuo (traditional Chinese medicine) [16], Metasequoia glyptostroboides [82], Malus pumila [83] and Zizania latifolia [84] also revealed the same genetic pattern.
These large changes in genetic diversity are most likely the result of both founder effects and selection for medicinal quality, and importantly, the clonal mode of reproduction. The extent of the loss of genetic diversity depends in part on population size during domestication and time over which domestication occurred [77]. At the beginning of cultivation, S. ningpoensis was gathered from Table 6. Hierarchical analysis of molecular variance for 24 populations of S. ningpoensis based on AFLP.

Origin of cultivated Scrophularia ningpoensis
Locating the geographical site of domestication and assessing the demographic consequences of wild-domesticated species evolution can help to design strategies of use and management of genetic resources [85]. In the current study, three cpDNA haplotypes (A, B and E) found within cultivated populations are scattered on the haplotype network and shared with different wild population gene pools. But the most cultivated populations harbored haplotype A, which is only found in the wild population LSW. Two other haplotypes B & E, distributed in two cultivated populations (YLC and SCC) and one cultivated population (NSC) respectively, are shared with three and four different wild populations respectively. It is interesting that wild populations harboring haplotype B or E didn't have haplotype A. This cpDNA pattern indicates that the wild population LSW in Jiangxi Province contributed substantially to the origin of cultivated S. ningpoensis, while other six wild populations (JGW, LAW, JHW, TNW, TM2W, PTW) located in geographically separated places were only involved in origin of three cultivated populations. Moreover, the wild population JGW harboring both haplotype B and E is also in Jiangxi Province as the wild population LSW. Thus, the Jiangxi region is a critical site for the origin of cultivated S. ningpoensis.
However, results of AFLP analysis suggested that the origin of cultivated populations is more restricted to only two wild populations, HNW and LSW. In the neighbor-Joining analysis (Figure 3), HNW and LSW clustered with other cultivated populations by 100% bootstrap value and the STRUCTURE analysis was consistent, placing the two wild populations, HNW and LSW, into the same gene pool as the cultivated individuals. In principal coordinate analysis (Figure 2), wild population HNW and LSW grouped with majority of cultivated populations in one cluster. Thus, combined with cpDNA data, we infer that wild populations of the Jiangxi region may have played a major role in the origin of cultivated S. ningpoensis. Although three cultivated populations, PAC, XJC and GZC, grouped to another cluster in PCoA analysis (Figure 2), they still located in the cultivated clade identified in the Neighbor-Joining analysis ( Figure 3A) and were assigned to the same gene pool as the other cultivated S. ningpoensis ( Figure 3B). The wild population LSW located in eastern Jiangxi and HNW population is on the border between Hunan and Jiangxi, so AFLP data strongly supported that Jiangxi area of Eastern China is the geographic origin of cultivated S. ningpoensis.
Considering the discordance between chloroplast and nuclear markers,the origin of cultivated S. ningpoensis must be interpreted with caution. Firstly, patterns of cpDNA indicated broader geographic origin of cultivated S. ningpoensis compared with AFLP markers, which can be explained by the characteristics of chloroplast genome evolution. CpDNA evolves slowly and is uniparentally inherited in plants, so it usually only represents the maternal genome and often has low levels of variation making cpDNA less informative for intraspecific studies [86]. Nuclear genomic AFLP markers provided much higher resolution both because of rapid evolution and many independent markers, thus providing a more nuanced interpretation of the origin of cultivated S. ningpoensis. Revealed by AFLP markers, cultivated populations originated from Jiangxi area where wild populations were involved in the event of origin. Secondly, if AFLP data has revealed the more resolution, these two wild populations which were involved But the wild population HNW didn't share any cpDNA haplotype with cultivated populations; only the native population LSW harbored haplotype A. These data suggest that this entire region of China may have served as the source of medicinal S. ningpoensis rather than a specific population. The ancestors of cultivated populations YLC, SCC and NSC were probably introduced from wild population similar to the extant HNW; harvesting and cultivation may, by chance, collected all individuals which had the same cpDNA haplotypes with cultivated populations. In addition, the nearest wild population to HNW is JGW, which shared two cpDNA haplotypes, B&E, with cultivated populations. But the very high value of Dw in HNW (21.92) and JGW (24.79) suggests a long-term history of isolation [74]. The long term use, and most likely, trading of rhizomes of S. ningpoensis lead to a complicated domestication history for this species.

Crop improvement and sustainable cultivation of S. ningpoensis
The therapeutic efficacy of a medicinal plant often varies by geographical location among the populations of a species [22]. Cultivated S. ningpoensis from Zhejiang Province is recognized to have the best medicinal quality [87]. Our previous study on HPLC fingerprints of S. ningpoensis has also revealed that materials cultivated in Zhejiang Province produce the highest content of bioactive compounds compared to other cultivated populations [45]. In this study, analysis of AFLP markers group cultivated populations from Zhejiang into one cluster which is divergent from other cultivated populations and other wild populations ( Figure 2). Moreover, cultivated populations from Zhejiang gather in one clade associated with other cultivated populations in NJ tree ( Figure 3A). This evidence strongly supports the genetic distinct nature of cultivated S. ningpoensis from Zhejiang ( Figure 2). Under strong artificial selection, cultivated S. ningpoensis with high pharmaceutical quality has genetically diverged from wild populations. It is generally believed that different medical effects due to secondary compounds are caused by various environmental conditions and habitats where the medicinal plants are grown and harvested [88]. But our study combined with previous phytochemical analysis indicates that under artificial selection, cultivated S. ningpoensis from Zhejiang has become genetically differentiated from other cultivated populations and chemical diversity may also be influenced by genetic factors. Thus, this result can help design strategies of crop improvement that consider both genetic and ecological factors for the cultivation of medicinal plants.
Genetic diversity underlies the plasticity of many secondary metabolites [89] and results in phytochemical diversity on which the pharmaceutical quality of medicinal plants relies [90]. Study on S. ningpoensis by HPLC and ISSR fingerprinting have already revealed a strong association between chemical and genetic variation of S. ningpoensis [45] suggesting that higher genetic diversity can account for higher phytochemical diversity that is critical for the pharmaceutical quality. This study that compares cultivated populations to the range of extant wild populations indicates that wild S. nignpoensis has much higher genetic diversity than does the cultivated plant ( Figure 1C, Table 5). These wild populations represent an infrequently used source of genetic diversity that is conspecific with the cultivated gene pool. Cultivated and wild plants generally do not exhibit reproductive isolation [79]. Hence, wild resources can be utilized to improve the cultivated S. ningpoensis by introducing genetic diversity by sexual reproduction and then selection for desired traits.
The results of this study can be used to select the most diverse natural populations for such a breeding program. Using the AFLP data, the values of Nei's [67] gene diversity (h) of most wild S. ningpoensis were higher than 0.1 (Table 5), except for PTW, TWW and TM1W. For chloroplast data, three wild populations, TM2W, LSW and JHW, harbored more cpDNA haplotypes than others ( Figure 1C). Thus, TM2W, LSW and JHW populations have high genetic diversity for both the cpDNA and nuclear genomes. These native populations have the greatest reservoir of genetic diversity and could be used for further improvement of cultivated S. ningpoensis to ensure that it provides an adequate supply of the medicinal plant and the quality of the plant is enhanced. Furthermore, the utilization of wild populations with appropriate guidelines for collection and for regulating management could be an efficient way to find, conserve and deploy desirable agronomic traits which would be beneficial for Radix Scrophulariae cultivation.