|Home | About | Journals | Submit | Contact Us | Français|
Life form, mating system and seed dispersal are important adaptive traits of plants. In the first effort to characterize in detail the population genetic structure and dynamics of wild Medicago species in China, a population genetic study of two closely related Medicago species, M. lupulina and M. ruthenica, that are distinct in these traits, are reported. These species are valuable germplasm resources for the improvement of Medicago forage crops but are under threat of habitat destruction.
Three hundred and twenty-eight individuals from 16 populations of the annual species, M. lupulina, and 447 individuals from 15 populations of the perennial species, M. ruthenica, were studied using 15 and 17 microsatellite loci, respectively. Conventional and Bayesian-clustering analyses were utilized to estimate population genetic structure, mating system and gene flow.
Genetic diversity of M. lupulina (mean HE = 0·246) was lower than that of M. ruthenica (mean HE = 0·677). Populations of M. lupulina were more highly differentiated (FST = 0·535) than those of M. ruthenica (FST = 0·130). For M. lupulina, 55·5 % of the genetic variation was partitioned among populations, whereas 76·6 % of the variation existed within populations of M. ruthenica. Based on the genetic data, the selfing rates of M. lupulina and M. ruthenica were estimated at 95·8 % and 29·5 %, respectively. The genetic differentiation among populations of both species was positively correlated with geographical distance.
The mating system differentiation estimated from the genetic data is consistent with floral morphology and observed pollinator visitation. There was a much higher historical gene flow in M. ruthenica than in M. lupulina, despite more effective seed dispersal mechanisms in M. lupulina. The population genetic structure and geographical distribution of the two Medicago species have been shaped by life form, mating systems and seed dispersal mechanisms.
Life form, mating systems and seed dispersal are important adaptive traits shaping genetic structure and geographical distribution of plant populations (Levin, 1981; Loveless and Hamrick, 1984; Ennos, 1994; Hamrick and Godt, 1996; Bohonak, 1999; Clauss and Mitchell-Olds, 2006; Song et al., 2006; Mable and Adam, 2007). Analyses of phenotypic variation of these traits together with population genetic variation should provide insights into the evolutionary history and processes of plant species (Barrett et al., 1996; Juan et al., 2004), which in turn will help determine evolutionary potentials and conservation strategies for natural populations.
Here, a population genetic study of two wild Medicago (Fabaceae) species, Medicago lupulina and Medicago ruthenica, which differ markedly in life form, mating systems, seed dispersal mechanisms and distribution ranges, is reported. The genus Medicago is distributed worldwide and consists of approx. 83 species, including two forage crops, M. sativa and M. truncatula (Small and Jomphe, 1989). Thirteen wild species of Medicago are found in China (Wei and Huang, 1998). They are adapted to a diverse range of habitats located in different geographical regions of China, from cold northern desert to warm and humid southern and central China, and from near the sea level in eastern China to high mountains in the Himalayas. These wild species hold a rich source of natural variation for the better understanding of plant population dynamics and for the improvement of Medicago cultivars. With rapid urbanization and overgrazing in China, however, these wild Medicago populations are threatened by severe reductions in number and size (J. Yan and H.-J. Chu, pers. obs.). Thus, there is an urgent need to investigate the population genetics and evolutionary dynamics of wild relatives of important forage crops.
Medicago truncatula has been chosen as one of the two representatives of the Fabaceae to have its entire genome sequenced. Taking advantage of the availability of genomic information for M. truncatula (http://medicago.org/), the population genetic structures of M. lupulina and M. ruthenica were studied using microsatellite markers. Medicago lupulina is annual, biennial or occasionally short-lived perennial, predominantly self-fertilizing, and widely distributed, whereas M. ruthenica is long-lived perennial, outcrossing, and much more narrowly distributed (Wei and Huang, 1998). Medicago lupulina has small indehiscent pods that facilitate long-distance seed dispersal by biotic and abiotic agents, whereas M. ruthenica has dehiscent pods and lacks effective mechanisms for seed dispersal.
Medicago lupulina, although never cultivated, has been grown as a green fodder or manure (Turkington and Cavers, 1979). The species is characterized as a water-saving and easily-maintained turf legume that contains a high level of protein and a low level of fibre suitable for grazing (Cao et al., 2003). However, the biomass of the forage legume is relatively low. Medicago ruthenica is adapted to dry, stony habitats or desert with extremely low snowfall and very cold winters (Campbell et al., 1999). It was considered to be superior to the cultivated species M. sativa in soil nutrient-use efficiency and thus might be more suitable for low-input systems (Campbell et al., 1999). Both wild species are adapted to a much wider range of habitats than the cultivated species and are valuable genetic resources for developing better grazing legumes especially in drier and colder regions. Although the morphology, physiology, phenology, chromosomal variation and in-breeding of the two species had been studied previously (Lammerink, 1968; Sidhu, 1971; Hébert et al., 1994; Mariani et al., 1996; Qi, 1996; van Berkum et al., 1998; Campbell et al., 1999; Wilson, 2005; Li and Shi, 2006), their population genetic structures had not been characterized.
In the present paper, the genetic variation was investigated in a sample consisting of 16 M. lupulina populations and 15 M. ruthenica populations using 15 and 17 microsatellite markers, respectively. These included five populations of each species occurring sympatrically with those of the other species, and five microsatellite markers shared between the two species. This made it possible to evaluate to a certain extent the genetic differentiation between M. lupulina and M. ruthenica in the common environments and at the same loci. In characterizing the population genetic structures of the two species, the aim was to (a) quantify genetic variability; (b) estimate gene flow; (c) infer the correlation between the genetic relationships and geographical distributions; and (d) integrate the genetic information with phenotypic variation in mating system, life form and seed dispersal to understand the population dynamics and evolutionary processes of the two Medicago species.
Medicago ruthenica Trautv. is distributed from central China to Mongolia and Siberia (Small and Jomphe, 1989). Our population sampling covers its entire distributional range in China (Fig. 1). Medicago lupulina L. has a broader distribution, occurring natively in temperate and subtropical Eurasia and North Africa (Dunbier, 1972). To compare its population genetic structure with M. ruthenica, populations of M. lupulina were sampled from the distributional range of M. ruthenica, including five sympatric populations of each species. In addition, populations of M. lupulina were sampled from north-western China. A total of 328 individuals were sampled from 16 populations of Medicago lupulina and 447 individuals were sampled from 15 populations of Medicago ruthenica between August and October in 2004 and 2005 (Table 1). Populations were recorded by GPS co-ordinates. Leaf samples were collected from randomly selected individuals in each population and immediately dried using silica gel for DNA isolation.
DNA was isolated from approx. 0·5 g dried leaves using a modified cetyltrimethy lammonium bromide (CTAB) protocol (Doyle and Doyle, 1987). DNA quality was tested on 0·8 % agarose gels. After measuring DNA concentration with an Eppendorf BioPhotometer, samples were diluted to 5 ng μL−1.
For microsatellite analysis, 92 pairs of primers were selected from previous publications (Diwan et al., 2000; Baquerizo-Audiot et al., 2001; Julier et al., 2003; Eujayl et al., 2004) and the Medicago genome website (http://medicago.org/genome/downloads.php). These primers were tested using 32 individuals from each of M. lupulina and M. ruthenica. Fifteen and 17 primers that detected a suitable level of polymorphism in M. lupulina and M. ruthenica, respectively, were identified (Table 2). The two species shared five microsatellite loci, MTIC14, MTIC188, MTIC189, MTIC432 and AFct45.
Microsatellites were amplified by polymerase chain reaction (PCR). A 10-μL reaction contained 10–50 ng of template DNA, 1 × PCR reaction buffer, 1·5 mm of MgCl2, 0·2 mm of dNTP mix, 0·5 µm of each primer, and 0·5 U of Taq polymerase (Invitrogen). Reactions were performed in a Gene Amp PCR system 9700 (PE Applied Biosystems) with the following programme: 5 min at 95 °C, then 30 cycles of 45 s at 95 °C, primer-specific annealing temperature for 30 s at 50–60 °C, and 72 °C for 1 s, followed by a final extension of 7 min at 72 °C. PCR products were denatured for 5 min at 95 °C and run on a 6 % polyacrylamide denaturing gels that were made 0·4 mm thick. A 25-bp DNA ladder (Promega, Madison, WI, USA) was used as the size standard. The gels were treated with silver stained to visualize DNA bands and score microsatellite alleles. The gel could reliably resolve alleles with 2-bp length difference.
For each population, genetic diversity was estimated across all loci using the observed number of alleles (na), effective number of alleles (ne), HE, HO, FIS, and the number of private alleles. For each microsatellite locus, genetic polymorphism was assessed by calculating the total number of alleles (A, allelic diversity), the expected and observed heterozygosity (HE and HO). The inbreeding coefficient, FIS, was calculated by FSTAT 2·9·3 (Goudet, 2001). GenAlEx 6 software was used to estimate ne, na, A, HO, HE (Peakall and Smouse, 2006). Selfing rate was estimated as s = 2 FIS/(1 + FIS) (Ritland, 1990). Deviation from the Hardy–Weinberg equilibrium and linkage disequilibrium were tested using the FSTAT program (Goudet, 2001). The significant values for the linkage disequilibrium were corrected for multiple comparisons by Bonferroni correction (Rice, 1989)
STRUCTURE 2·2 (Pritchard et al., 2000; Falush et al., 2003) was used to test whether M. lupulina and M. ruthenica were genetically differentiated without a priori classification of individuals. The program STRUCTURE implements a model-based clustering method to demonstrate the presence of population structure, assign individuals to populations and identify migrants and admixed individuals (Pritchard et al., 2000), by assuming that the markers are unlinked and at linkage equilibrium within populations. The program calculates an estimate of the posterior probability of the data for a given K, Pr(X/K) (Pritchard et al., 2000). The range of possible K was from 1 or 2 to the true number of populations plus 3 (Evanno et al., 2005). Therefore, the number of populations (clusters), K, was set from 1 to 18. Under the assumption that admixture model and allele frequencies correlated, each K was replicated 3–5 times with different probability (v) for 100 000 iterations after a burn-in period of 50 000 without prior information on the population of origin. Additionally, population divergence was quantified using θ, and an unbiased estimator of FST (Weir and Cockerham, 1984; Slatkin, 1995) was estimated using FSTAT version 2·9·3 (Goudet, 2001). Genetic differentiation within and among populations was further measured by analysis of molecular variance (AMOVA) using ARLEQUIN 3·1 (Excoffier et al., 2005).
The POPULATION 1·2 software (Langella, 2000) was used to calculate the Nei's genetic distance (DA) (Nei, 1983)among individuals within each species. Unrooted neighbor-joining trees based on Nei's genetic distance were constructed and visualized with the TREEVIEW software (Page, 1996). To investigate spatial genetic structure, the relationship between the matrix of pairwise genetic distance [FST/(1 – FST)] and the matrix of the logarithm of geographical distances was analysed via a Mantel's test (Mantel, 1967) with 100 000 random permutations using the program IBD (Bohonak, 2002). Geographic distances between pairs of populations were calculated from linear distances between latitude and longitude positions (http://jan.ucc.nau.edu/~cvm/latlongdist.html).
Individual-based assignment tests have been used to estimate contemporary rates of gene flow and dispersal (Berry et al., 2004; Mix et al., 2006). A partial exclusion Bayesian-based individual assignment test (Rannala and Mountain, 1997), implemented in the GeneClass 2·0 (Piry et al., 2004), was used to assess the recent gene flow between pairs of populations. Assignment probabilities are computed based on the resampling method of Paetkau et al. (2004). A total number of 1000 individuals was simulated and a threshold of 0·01 was used. The limitation of this method is that the migration rate observed during a short study might not accurately reflect long-term patterns of gene flow (Manel et al., 2005). Therefore, FST was used to estimate historical rates of gene flow (Nm) according to Wright's island model of population genetic structure, where FST ≈ 1/(1 + 4Nm) (Wright, 1951; Slatkin and Barton, 1989; Gaggiotti et al., 1999; Sork et al., 1999).
The level and pattern of estimated genetic variation differed substantially between populations of M. lupulina and M. ruthenica (Table 1 and Fig. 2). For M. lupulina, a total of 328 individuals from 16 populations was surveyed, in which 112 alleles were identified at 15 microsatellite loci. The average of observed heterozygosity (HO) within populations was 0·017, ranging from 0 to 0·044. The values are considerably lower than expected heterozygosity (HE) assuming the Hardy–Weinberg equilibrium, which averaged 0·246.
In comparison to M. lupulina, M. ruthenica has a higher level of genetic variation. In 447 individuals from 15 sampled populations of M. ruthenica, 278 alleles were found at 17 loci. The average HO of each populations is 0·574, ranging from 0·413 to 0·641. This is slightly lower than the average HE of 0·677. Medicago ruthenica populations also possessed a larger average number of private alleles than M. lupulina populations (3 versus 1·9). However, the variation in the number of private alleles is greater among M. lupulina populations (range 0–9) than that of the M. ruthenica populations (range 1–5).
When the genetic variation was compared between the two species at the five shared microsatellite loci, M. ruthenica accessions also showed a higher allelic diversity than M. lupulina (Table 2). At these five loci, 17·4 alleles per locus were found for M. ruthenica and 7·2 alleles per locus for M. lupulina. This is similar to the averages of all sampled loci, which was 16·4 and 7·5 alleles per locus for M. ruthenica and M. lupulina, respectively.
The test for the Hardy–Weinberg equilibrium found that of 240 locus–population combinations in M. lupulina, 165, 128 and 109 or 68·8 %, 53·3 % and 45·4 % showed significant deviation at P = 0·05, 0·01 and 0·001, respectively. For M. ruthenica, of 255 locus-population combinations, 112, 79 and 48 or 43·9 %, 30·9 % and 18·8 % showed significant deviation at P = 0·05, 0·01 and 0·001, respectively. The test for the genotypic disequilibrium in all samples within each species found that 80 of 105 locus pairs in M. lupulina and 36 of 136 locus pairs in M. ruthenica showed significant deviation at the P = 5 %, but none of locus pairs was found to be in significant genotypic disequilibrium after the Bonferroni-type correction.
The average FIS values were 0·920 (range 0·741–1·000) and 0·173 (range 0·121–0·292) for M. lupulina and M. ruthenica, respectively. From these values, the rates of self-fertilization of M. lupulina and M. ruthenica are calculated at 95·8 % and 29·5 %, respectively.
The analysis of molecular variance (AMOVA, Table 3) revealed that the majority of genetic variation occurred among populations in M. lupulina (55·5 %) and within individuals in M. ruthenica (76·06 %). In contrast, the minimum partitions of genetic variation resided within individuals in M. lupulina (3·07 %) and among populations in M. ruthenica (10·81 %). Bayesian clustering without prior information about geographical origin of populations showed that the highest likelihood value (Ln PrX/K) occurred at K = 16 in M. lupulina and K = 15 in M. ruthenica (Fig. 3), where the number of clusters (K) was consistent with the natural populations sampled in this study. The result held for different values of v (the probability that an individual was an immigrant to a given population; v = 0·01, 0·05 and 0·1).
Substantial genetic differentiation was found among populations for both M. lupulina (FST = 0·535 P < 0·001) and M. ruthenica (FST = 0·130, P < 0·001; Table 2). The Mantel's test showed that the genetic distance [FST/(1 – FST)] and the geographical distance of the populations of each species are positively correlated (M. lupulina: r = 0·27, P = 0·0094; M. ruthenica: r = 0·41, P = 0·0018; Fig. 4). The cluster analyses of population relationships based on genetic distance showed the populations from closely situated regions were grouped together (Fig. 5). Nevertheless, some exceptions are noteworthy. Populations MR-10 and MR-12 from central China were clustered with populations from north-eastern China. ML-7 in north-eastern China was clustered with populations from north-western China, ML1-5, rather than clustered with the nearest population ML-8. Similarly, populations ML-8 was clustered with ML-6 in western China.
Based on estimated FST, historical gene flow among the sampled populations (Nm) was calculated at 0·218 in M. lupulina and 1·672 in M. ruthenica (Table 2). With regard to the contemporary gene flow, the assignment tests showed that an average of 80·8 % (P < 0·01) and 82·6 % (P < 0·01) of individuals were correctly assigned to their own source populations in M. lupulina and M. ruthenica, respectively (Fig. 6). For the remaining individuals, 3·4 % of M. lupulina and 1·4 % of M. ruthenica individuals did not belong to any of the populations sampled, and 15·7 % of M. lupulina and 16 % of M. ruthenica individuals were assigned to a population different from which they were collected.
This study provides the genetic estimate of the mating systems of the two Medicago species. From FIS values, selfing rates were estimated to be higher than 95 % for M. lupulina but lower than 30 % for M. ruthenica. Together with previous studies (Mulligan, 1972; Bena et al. 1998; Campbell et al., 1999) and the observed heterozygosity (HO), this suggests that M. lupulina is predominantly selfing whereas M. ruthenica is highly outcrossing.
The estimated mating system differentiation is consistent with differences in floral morphology, similar to the correlation between floral morphology and pollinator attraction found in other plant groups (e.g. Juan et al., 2004; Gómez et al., 2008). Medicago lupulina has relatively small flowers with yellow papilionaceous corollas of 2–4 mm long (Fig. 7A). Medicago ruthenica has larger and more showy flowers with yellow corollas of approx. 8 mm long tinged with dark purple on the outside of the petals and on the inside toward the base (Fig. 7C). It was observed that M. ruthenica was able to attract substantially more insect pollinators such as bees, bumblebees and butterflies.
Mating system and life form play important roles in shaping population genetic structure and distribution of plants (e.g. Loveless and Hamrick, 1984; Hamrick and Godt, 1989, 1996; Stenøien et al., 2005; Clauss and Mitchell-Olds, 2006; Drummond and Hamilton, 2007; Mable and Adam, 2007; Michalski and Durka, 2007). The substantially higher selfing rate in M. lupulina could have contributed to a lower overall level of estimated heterozygosity (HE = 0·246) than those from M. ruthenica (HE = 0·677). A low level of heterozygosity (HE = 0·348–0·476) was also found for the selfing species M. truncatula in the French Mediterranean region (Bonnin et al., 2001; Ellwood et al., 2006), whereas a higher level of heterozygosity (HE = 0·665–0·717) was observed for an outcrossing species Medicago sativa (Flajoulot et al., 2005). These values are also comparable to the genetic diversity previously reported for other plant species with the similar mating system and life form. For instance, Arabidopsis thaliana, a selfing annual plant, had a much lower level of the genetic diversity (HE = 0·01, 0·06) than the self-incompatible perennial relatives A. halleri (HE = 0·31) and A. petraea (HE = 0·41) (Clauss et al., 2002; Stenøien et al., 2005). The present results are generally consistent with the trends of genetic variation observed in many flowering plant groups based on microsatellite data, with average HE values being 0·41 for inbreeding populations versus 0·65 for outcrossing populations and 0·46 for annuals versus 0·68 for perennials (Nybom, 2004).
The estimate of population differentiation showed a much higher population genetic differentiation in M. lupulina (FST = 0·535) than in M. ruthenica (FST = 0·130). This fits well with the general estimates of approx. 5-fold higher FST in selfing, annual species than outcrossing, perennial species for flower plants (Hamrick and Godt, 1996).
The significantly positive correlation between genetic and geographical distances detected in M. lupulina (r = 0·2703, P = 0·0094) and M. ruthenica (r = 0·4113, P = 0·0018) indicates that spatial separation has played a role in shaping the population genetic structure of the species. There is a general tendency that closely situated populations are genetically more similar. However, a close relationship between two disjunct populations, ML-6 and ML-8, suggests that there exists a dispersal or gene flow corridor connecting these regions across the Mongolian grasslands. Although it was not possible to sample M. lupulina populations from Mongolia, finding genetically similar populations flanking the width of Mongolia implies that the gene pool from this broad region may be represented in the present samples.
While geographical isolation has played a role in population genetic differentiation, there is evidence for gene flow between populations of each species. Among individuals of M. lupulina and M. ruthenica, 3·4 % and 1·4 %, respectively could not be assigned to sampled populations, suggesting that they were immigrants from outside the areas sampled (Fig. 6). In addition, the assignment of 15·7 % and 16 % of individuals to populations different to their sampled populations may be a result of gene flow between these populations.
Plants and their genes migrate through seed and pollen dispersal. The study of the interplay between seed dispersal and pollination is essential for understanding plant population structure and distribution (Govindaraju, 1988; Ennos, 1994; Bohonak, 1999; McCauley, 1997; Heuertz et al., 2003; Juan et al., 2004; Otero-Arnaiz, 2005). Seed dispersal of M. lupulina and M. ruthenica may be substantially affected by the distinct fruit morphologies (e.g. Janson, 1983; Gautier-Hion et al., 1985; Willson and Traveset, 2000). Pods of M. lupulina are rough-ridged, indehiscent, up to 3 mm long and 1 mm wide and each contains one seed (Fig. 7B). Rough-surfaced, indehiscent pods are effectively dispersed by birds and animals through adhesion or ingestion (Lammerink, 1968; Dunbier, 1972; Sorensen 1986). In addition, indehiscent pods of M. lupulina can float in water for up to 12 d (Turkington and Cavers, 1979).
Medicago ruthenica, however, lacks an effective mechanism of seed dispersal. Its pod is up to 15 mm long and 5 mm wide with an oblong-falcate suture and contains three to five seeds (Fig. 7D). The dorsal suture is strongly convex and the ventral suture weakly convex to straight, facilitating dehiscence when pods mature. Pods of M. ruthenica dehisce and release seeds near the mother plant. Seeds have no wings and are dispersed mainly by gravity. This is similar to wild soybean which was reported to disperse seeds within 4·5 m of the mother plants (Oka, 1983; Kuroda et al., 2006).
The difference in current distributional ranges of the two species could be attributed at least partly to seed dispersal abilities. Medicago lupulina, with more effective seed dispersal mechanisms, occurs in a much broader geographical range than M. ruthenica. The narrower distributional range of M. ruthenica does not seem to have been a result of population extinction. Medicago ruthenica populations that were studied in the field appeared healthy, and the level of genetic variation currently maintained within and among populations does not provide any indication that this outcrossing species has experienced a severe reduction in genetic variation.
Despite more effective seed dispersal of M. lupulina, the present genetic data showed that there has been a much higher level of historical gene flow between M. ruthenica populations. This is not surprising, given a greater pollen dispersal capacity of the outcrosser, M. ruthenica than the selfer, M. lupulina. In predominantly selfing species, individuals migrating into other populations may not effectively incorporate their private alleles into the local populations through cross-pollination. As a result, these alleles may easily get lost through drift if they are not favoured by selection. New alleles from other populations are likely to be less fit than the alleles of the native populations that have been selected by local ecological factors. Furthermore, migrating alleles are especially susceptible to loss through drift in annual species. For an out-crossing species, on the other hand, migration into new populations via seed dispersal allows alleles to be much more easily integrated into the gene pool of the recipient populations through cross-pollination. Less-fit alleles may be maintained in recipient populations at the heterozygous loci as long as they are at least partially recessive, and may persist for a relatively long period of time in a perennial species even though they are not postively selected into the local gene pool.
Taken together, this study between the two Medicago species with a combination of several distinct biological features has allowed us to gain a better understanding of population processes of plant evolution. As a perennial species, M. ruthenica benefits from an outcrossing mating system for the maintenance of genetic variation within populations. As a result, the populations can be more stable and less susceptible to pathogen and environmental changes. Consequently, there might have been relatively little pressure for the development of effective seed-dispersal mechanisms. In M. lupulina, on the other hand, effective seed-dispersal mechanisms are essential for a predominantly annual species. Indeed, effective seed-dispersal mechanisms could have broadened its distribution. For an annual species possessing effective seed dispersal, self-pollination provides reproductive assurance so that a single or a few seeds could potentially establish a new population in a new locality (Holsinger, 1996). Characterization of the population genetic structure of M. lupulina and M. ruthenica has provided an understanding of historical population dynamics of the two species and their current distribution.
We thank M. Kang and X.-H. Yao for technical assistance. This study was partially supported by the Chinese Academy of Sciences.