|Home | About | Journals | Submit | Contact Us | Français|
The Asian genus Vigna, to which four cultivated species (rice bean, azuki bean, mung bean and black gram) belong, is suitable for comparative genomics. The aims were to construct a genetic linkage map of rice bean, to identify the genomic regions associated with domestication in rice bean, and to compare these regions with those in azuki bean.
A genetic linkage map was constructed by using simple sequence repeat and amplified fragment length polymorphism markers in the BC1F1 population derived from a cross between cultivated and wild rice bean. Using this map, 31 domestication-related traits were dissected into quantitative trait loci (QTLs). The genetic linkage map and QTLs of rice bean were compared with those of azuki bean.
A total of 326 markers converged into 11 linkage groups (LGs), corresponding to the haploid number of rice bean chromosomes. The domestication-related traits in rice bean associated with a few major QTLs distributed as clusters on LGs 2, 4 and 7. A high level of co-linearity in marker order between the rice bean and azuki bean linkage maps was observed. Major QTLs in rice bean were found on LG4, whereas major QTLs in azuki bean were found on LG9.
This is the first report of a genetic linkage map and QTLs for domestication-related traits in rice bean. The inheritance of domestication-related traits was so simple that a few major QTLs explained the phenotypic variation between cultivated and wild rice bean. The high level of genomic synteny between rice bean and azuki bean facilitates QTL comparison between species. These results provide a genetic foundation for improvement of rice bean; interchange of major QTLs between rice bean and azuki bean might be useful for broadening the genetic variation of both species.
The genus Vigna subgenus Ceratotropis consists of 21 species that are distributed across a wide region of Asia. Six cultivated species belong to this subgenus (Tomooka et al., 2002). Among these, mung bean (Vigna radiata), rice bean (V. umbellata), black gram (V. mungo) and azuki bean (V. angularis) are economically important in Asian countries. Taxonomically, wild forms of species are usually recognized as varieties below the rank of species of the Asian Vigna. However, numerous differences in morphological and physiological traits associated with domestication are observed between the cultivated and wild forms. These differences, collectively called the domestication syndrome, result from selection over several thousands of years of adaptation to cultivated environments, human nutritional requirements and preferences (Hawkes, 1983). The four cultivated species listed above are therefore ideal material for improving our comparative genomics-based understanding of the gene evolution related to domestication within and among Vigna species and for characterizing useful traits as quantitative trait loci (QTLs) for use in breeding.
To compare the genomic structures and genomic regions associated with domestication among these four species, genetic linkage maps of azuki bean (Han et al., 2005) and black gram (Chaitieng et al., 2006) were constructed previously using simple sequence repeat (SSR), restriction fragment length polymorphism (RFLP) and amplified fragment length polymorphism (AFLP) markers; the QTLs for domestication-related traits of azuki bean have been identified (Isemura et al., 2007) in populations derived from crosses between cultivated and wild forms. The order of common markers on the linkage groups (LGs) is highly conserved between the azuki bean map (Han et al., 2005) and the black gram map (Chaitieng et al., 2006), although differences, such as inversions, deletions, duplications and a translocation, were observed between the two genomes in several LGs.
This paper focuses on the rice bean genome as part of a comparative genome analysis among members of the subgenus Ceratotropis. Rice bean is a traditional crop grown across south, south-east and east Asia, and its wild form is distributed across a wide area of the tropical monsoon forest climatic zone from eastern India, Nepal, Myanmar (Burma), Thailand, Laos and southern China to East Timor (Arora et al., 1980; Tomooka et al., 2002; Bisht et al., 2005; Seehalak et al., 2006; Gautam et al., 2007; Tomooka, 2009). The dried mature seeds of rice bean are usually eaten with rice or in soups, and the leaves, flowers, shoots and young pods are eaten as a vegetable in the upland areas of south-east Asia and southern China. Rice bean is also important as a fodder and a green manure (Tomooka et al., 2002). Recently, with the aim of using these genetic resources efficiently, researchers studied genetic diversity in cultivated and wild rice bean from Thailand, India and Nepal using molecular markers (Seehalak et al., 2006; Bajracharya et al., 2008; Muthusamy et al., 2008). The domesticated accessions from south-east Asia showed the largest genetic variations in AFLP (Seehalak et al., 2006) and SSR (J. Tian et al., Institute of Cereal and Oil Crops, Hebei Academy of Agricultural and Forestry Sciences, China, unpubl. res.) markers and the largest phenotypic variations in seed coat colour and seed size (Tomooka, 2009). These results suggest that south-east Asia was the centre of the origin and genetic diversity of this species (Tomooka, 2009).
To date, rice bean has not been subjected to systematic breeding, despite the species' many useful characteristics, including bruchid resistance (Tomooka et al., 2000b; Kashiwaba et al., 2003; Somta et al., 2006); disease resistance, particularly to yellow mosaic virus (Arora et al., 1980; Borah et al., 2001) and bacterial leaf spot (Arora et al., 1980), and the highest potential grain yield among Ceratotropis species (Smartt, 1990). Therefore, rice bean can best be described as a scientifically neglected crop of great potential.
Genetic linkage maps using rice bean as one parent have been constructed for populations derived from interspecific crosses [rice bean × azuki bean (Kaga et al., 2000) and rice bean × V. nakashimae (Somta et al., 2006)]. These linkage maps have been used to localize genes for several qualitative traits (Kaga et al., 2000) and QTLs for resistance to bruchid beetles and for seed weight (Somta et al., 2006). However, a precise genetic linkage map from a large population derived from an intraspecific cross has not been constructed.
The objectives in this study were (a) to construct a precise genetic linkage map of rice bean using SSR markers from related grain legumes in an intraspecific population derived from cultivated and wild forms; (b) to examine genomic synteny between rice bean and azuki bean on the basis of the linkage map; (c) to identify QTLs for domestication-related traits in rice bean; and (d) to compare the QTLs detected in rice bean with QTLs in azuki bean.
Plant materials were obtained from the Genebank (http://www.gene.affrc.go.jp/index_en.php) collection of the National Institute of Agrobiological Sciences (NIAS), Tsukuba, Japan. A single F1 hybrid between a cultivated rice bean (JP217439, Col. No. 2002M21) and a wild rice bean accession (JP210639, Col. No. CED99T-2) was backcrossed to cultivated rice bean (JP217439) as a male parent to construct a BC1F1 population consisting of 198 plants. The wild parent accession was chosen because AFLP analysis had revealed that it was highly differentiated from other accessions (Seehalak et al., 2006). It was collected by one of us (N.T.) from Kanchanaburi province in Thailand; its habitat was a moist deciduous forest along a river (Tomooka et al., 2000a). The cultivated parent accession, a landrace from Myanmar (Tomooka et al., 2003), was selected because it had the largest seeds among the germplasm conserved in the NIAS Genebank.
Thirty-one traits related to domestication were evaluated (Table 1). Of these, 30 were treated as quantitative traits and one (hilum colour) as a qualitative trait. The BC1F1 population of 198 plants, together with ten plants of each parent, were grown in a vinyl house at NIAS, Tsukuba, Japan (36 °2′N, 140 °8′E), from September 2006 to March 2007 under natural day length. The natural day length gradually decreased from 12 h 28 min (14 September 2006) to 9 h 42 min (22 December) and thereafter increased to 12 h 18 min (25 March). From October, the vinyl house was heated to maintain a minimum temperature of 15 °C.
Seedling traits – leaf primary length (LFPL), leaf primary width (LFPW) and epicotyl length (ECL) – were recorded when the first trifoliate leaf opened, and vegetative traits – leaflet maximum length (LFML), leaflet maximum width (LFMW), stem internode length (1st to 10th: ST1I to ST10I), stem length (STL) and stem thickness (STT) – were recorded when the 10th trifoliate leaf was fully developed. After all pods had been harvested, branch number (BRN) and position of the first branch on the main stem (BRP) were recorded. These traits were evaluated in the BC1F1 generation.
Seed-related traits were investigated by using the seeds from BC1F1 plants. Seed water absorption (SDWA) was investigated as an index of seed dormancy. Fifteen unscarified seeds were placed on wet filter paper and incubated in the dark at 15 °C for 24 h; the number of seeds that absorbed water was then recorded. Seed dimensions – seed length (SDL), seed width (SDW) and seed thickness (SDT) – were averages of ten seeds. The 100-seed weight (SD100WT) was evaluated using intact seeds.
Pod traits – pod length (PDL), pod width (PDW) and number of twists along the length of the dehisced pod when kept at room temperature (PDT) – were based on ten pods. PDT was used as an index of pod dehiscence.
The number of days from sowing to first flowering (FLD) and number of days from first flowering to harvesting of first pod (PDDM) were recorded in the BC1F1 population. The number of seeds per pod (SDNPPD) was measured in ten pods.
Seed hilum colour (SDHC) was evaluated in seeds from BC1F1 plants, and the seed hilum of each plant was classified as either cultivated parent type (white) or wild parent type (pale red).
Total genomic DNA in BC1F1 plants was extracted from 200 mg of fresh leaf tissue using a DNeasy plant mini kit (Qiagen, Valencia, CA, USA). The DNA concentration was adjusted to 5 ng μL−1 for the SSR analysis and to 25 ng μL−1 for the AFLP analysis by comparison with known concentrations of standard λDNA in 1 % agarose gel.
The SSR analysis in the BC1F1 population was performed by the method of Han et al. (2005). Three hundred and twenty-five SSR primer pairs were screened from azuki bean (Wang et al., 2004), 163 from cowpea (Vgina unguiculata; Li et al., 2001) and 40 from common bean (Phaseolus vulgaris; Yu et al., 2000; Gaitan-Solis et al., 2002; Blair et al., 2003; Guerra-Sanz, 2004) to determine whether there was polymorphism between the two parents. By using the read2Marker program (Fukuoka et al., 2005), another 156 cowpea SSR primer pairs (Table S1 in Supplementary Data, available online) on cowpea genomic sequences from the Cowpea Genomics Knowledge Base (Chen et al., 2007; http://cowpeagenomics.med.virginia.edu/CGKB/) were designed and screened for parental polymorphisms.
AFLP analysis was performed with an AFLP Core Reagent Kit (Invitrogen, Carlsbad, CA, USA). The steps of DNA digestion, ligation of adaptor and pre-selective amplification were performed in accordance with the method of Han et al. (2005). Selective amplification and detection of AFLP bands were performed using the method of Kaga et al. (2008). On the basis of information on the AFLP primer combinations used to construct several linkage maps of Asian Vigna species (Han et al., 2005; Chaitieng et al., 2006; Kaga et al., 2008), 28 primer combinations were selected for the AFLP analysis in the BC1F1 population.
The linkage map was constructed using the method of Han et al. (2005) using JoinMap v. 4·0 (Van Ooijen, 2006). Marker segregation was analysed by a chi-squared test for goodness-of-fit to the expected Mendelian ratio (1 : 1). The recombination frequencies were converted into map distance (cM) by using the Kosambi mapping function (Kosambi, 1944). After a framework map had been built using codominant SSR markers, dominant SSR and AFLP markers were integrated into the framework. Numbering of the LGs followed that for azuki bean (Han et al., 2005).
For the quantitative traits, the mean, standard error minimum value, maximum value and broad-sense heritability were calculated, and the frequency distributions of phenotypes in the BC1F1 (or BC1F1:2) population were examined for each. The correlations between pairs of traits were also calculated. The hilum colour segregation pattern was investigated in each BC1F1 individual and the frequency distribution was analysed by chi-squared test for goodness-of-fit to the expected Mendelian ratio (1 : 1). The segregation pattern data were used to identify the map position of the gene controlling this trait.
Mean data for each trait (SDWA data were arcsine-transformed first) were used in the QTL analysis. QTLs were identified by means of multiple interval mapping using the software package MultiQTL v. 3·0, as described by Kaga et al. (2008). QTL nomenclature followed Somta et al. (2006). To test for randomness of the genomic distribution of QTLs for domestication-related traits, chi-squared tests were calculated as described by Isemura et al. (2007). To test whether or not QTLs were randomly distributed along an LG, a Poisson distribution function was calculated as described by Isemura et al. (2007).
SSR primer pairs developed from azuki bean, cowpea and common bean were used to construct the genetic linkage map of rice bean. The percentage fragment amplification in rice bean was high, with values between 73·6 % (cowpea) and 96·6 % (azuki bean) (Table 2). Of the 325 azuki bean, 163 cowpea and 40 common bean SSR primer pairs screened, 167 azuki bean (51·4 %), 44 cowpea (27·0 %) and 7 common bean (17·5 %) primer pairs showed clear polymorphism between the cultivated and wild rice bean parents (Table 2). All 223 marker loci derived from the 218 polymorphic SSR primer pairs were used for mapping; five azuki bean SSR primer pairs (CEDG018, CEDG081, CEDG154, CEDG251 and CEDG302) detected two loci.
In the AFLP analysis, 1380 bands were detected by 28 primer combinations. The total number of bands per primer pair ranged from 39 to 59, and the average number was 49·3. Out of 1380 bands, 103 (range 1–9 per primer combination, average 3·7) showed clear polymorphism between the cultivated and wild parents (Table 3).
A total of 326 loci (172 azuki bean SSR loci, 44 cowpea SSR loci, 7 common bean SSR loci and 103 AFLP loci) could be assigned to 11 LGs covering a total of 796·1 cM of the rice bean genome at an average marker density of 2·5 cM (Fig. 1 and Table 4). The genomic region in this map covered 95·7 % of the azuki bean genome (832·1 cM) reported by Han et al. (2005). The number of markers on each LG ranged from 14 (LG7) to 42 (LG8). The length of each LG ranged from 105·8 (LG1) to 39·4 cM (LG11). The average distance between two adjacent markers ranged from 1·58 (LG11) to 5·04 cM (LG7). All LGs except LG5 had gaps greater than 10 cM between markers. LGs 1 and 3 had gaps greater than 15 cM. Sixty-one markers (18·7 %) showed segregation distortion (P < 0·05). The ratio of plants with heterozygous genotypes was high for all 61 markers which were found on LGs 5, 7, 8 and 11. Interestingly, all of the markers on LG11 showed segregation distortion at significant levels.
The means, minima, maxima, standard errors and broad-sense heritabilities of traits in the parental lines and BC1F1 or BC1F1:2 populations were determined (Table 5). The mean SDWA of the cultivated parent was higher than that of the wild parent. PDT of the wild parent was greater than that of the cultivated parent. The sizes of the leaf, stem, seed and pod were larger in the cultivated parent. Among the stem-length-related traits, ECL, ST1I to ST10I and STL were longer in the cultivated parent.
FLD in the cultivated parent was 91 d and PDDM was 64 d. Because only one plant of the wild parent flowered, and it produced only one pod, it was not possible to evaluate the traits of wild parent precisely. In December 2007, pods and seeds were obtained from wild plants sown again in April 2007 and were used to evaluate the pod- and seed-related traits of the wild parent. SDNPPD and BRN were greater in the cultivated parent than in the wild parent.
The BC1F1 plants and BC1F1:2 lines showed a high degree of morphological and physiological variation (Table 5). Seed- and pod-size-related traits showed high heritability (>80 %), whereas stem- and leaf-related traits showed low heritability (<70 %) except in the case of ST4I, ST10I and STL. The means of the BC1F1 plants and BC1F1:2 lines fell between the means of the cultivated and wild parents for all traits except SDNPPD, PDT and BRP. Most traits in these populations showed a nearly normal distribution among parents (Fig. S1 in Supplementary data, available online). Transgressive segregation was observed in SDNPPD, PDT, FLD, STT and BRN.
In general, there were significant positive correlations (P < 0·05) between related traits, such as between stem length and each internode length, and between seed-size-related traits and pod-size-related traits (Table S2 in Supplementary Data). PDT was negatively correlated with seed-size-related traits, each internode length (ST1I to ST10I) and stem length (STL). Water absorption by seeds (SDWA) was positively correlated with epicotyl length and lower internode length (ST1I and ST2I). SDNPPD was highly correlated with seed-size-related traits and PDL. Seed- and pod-size-related traits and PDDM were positively correlated.
QTL analyses were performed for each trait in the population (Table 6 and Fig. 2; for details see Fig S2 in Supplementary Data). In total, 73 QTLs and one morphological marker gene are reported for the 31 domestication-related traits. The number of QTLs may have been overestimated as a result of measurement of related traits such as 100-seed weight, seed length, seed width and seed thickness. Generally, one to seven QTLs were detected for each trait at a significant level (P < 0·001), except in the case of FLD, LFPL, LFPW, ST3I and BRP, for which no QTLs were detected.
One of the most important changes that occurred during domestication was a reduction in seed dormancy. The SDWA assay revealed that the water absorption rate of the cultivated parent (98 %) was higher than that of the wild parent (33 %) (Table 5). Five QTLs were detected, on LG2, LG3, LG4, LG8 and LG10, and QTL Sdwa4.4.1 on LG4 explained 25·1 % of the phenotypic variation. The alleles of the cultivated parent increased SDWA at Sdwa4·3.1, Sdwa4.4.1 and Sdwa4·8.1 and decreased water absorption at Sdwa4.2.1 and Sdwa4.10.1. The allele of the cultivated parent at Sdwa4.4.1 had an especially large effect, increasing the SDWA by 10·5 %.
Loss (reduction) of pod dehiscence is advantageous for harvesting seeds. The number of twists along the length of the shattered pod was used as an index of pod dehiscence. The cultivated parent had fewer twists. One QTL, Pdt4·7.1, with a relatively high contribution to this trait (42·4 %) was found on LG7.
Domestication of rice bean has resulted in a 15-fold increase in seed weight (Table 5). Six or seven QTLs were detected for seed-size-related traits (SD100WT, SDL, SDW and SDT) on LG1, LG2, LG3, LG4, LG5, LG7 and LG11. The QTLs with the highest contribution to these traits (25·2–33·4 %) were found on LG4. The QTLs on LG2 explained a further 14·9–17·6 % of the phenotypic variation.
Domestication of rice bean has resulted in a 2·5-fold increase in pod length (Table 5). Three QTLs for PDL and six QTLs for PDW were found (Table 6). Those affecting these two traits were found mainly on LG2, LG3 and LG4. The QTLs with the highest contribution for these traits (23·1 % for PDL and 27·6 % for PDW) were found on LG4. As expected, the alleles from the cultivated parent increased both PDL and PDW at all QTL positions. The QTLs for both PDL and PDW on LG2, LG3 and LG4 were located close to the QTLs for seed-size-related traits on each respective LG (Fig. 2).
The primary leaf length in many BC1F1 plants was close to that in the cultivated parent. Primary leaf width of most of the BC1F1 plants was distributed within a narrow range (about 0·9 cm), as determined by the frequency distribution (Fig. S1 in Supplementary Data). The values of heritability for both traits were low (<40 %; Table 5). As a result, no QTLs for these traits were detected.
Major QTLs for maximum leaf size (LFML, LFMW) were detected on LG4. These QTLs were located close to QTLs for seed- and pod-size-related traits. However, in contrast to the latter, the alleles from the cultivated parent had a negative effect on leaf size at these QTLs. The QTLs for these traits on LG4 explained 31·8 % (LFML) and 25·6 % (LFMW) of the phenotypic variation.
In the cultivated parent, the epicotyl, all internodes and the stem were longer, and stem was thicker, than in the wild parent (Table 5). Three QTLs (Stl4.1.1, Stl4.5.1 and Stl4.7.1) for STL were found, on LG1, LG5 and LG7 (Table 6). Alleles from the cultivated parent at all QTLs had the effect of increasing stem length. However, stem growth at the early, middle and late stages was controlled by different QTLs. Stem length from the primary to the 10th node QTLs (Stl4.1.1 on LG1, Stl4.7.1 on LG7) and from the middle to the upper internode length QTLs (4th to 10th; St7i4.1.1 to St10i4.1.1 on LG1 and St4i4.7.1 to St10i4.7.1 on LG7) were consistently located in similar map positions on LG1 and LG7. On the other hand, QTLs for the lower (1st and 2nd) internode lengths (St1i4.7.1 and Stl2i4.7.1) in the early stem-growth stage were found mainly on LG7 (at different QTL positions from those of the middle and upper internode lengths on LG7). Four QTLs for epicotyl length (Ecl4.4.1, Ecl4.7.1, Ecl4.8.1 and Ecl4.11.1) were found, on LG4, LG7, LG8 and LG11. Among these, a QTL on LG7 was located close to QTLs for the first and second internode length (St1i4.7.1 and St2i4.7.1). QTLs for epicotyl length (Ecl4.4.1 on LG4 and Ecl4.7.1 on LG7) and first and second internode length (St1i4.7.1 and St2i4.7.1 on LG7) were detected relatively close to the QTLs for seed-size-related traits on LG4 and LG7 (Fig. S2 in Supplementary Data). Only one stem diameter QTL, Stt4.9.1, was detected, on LG9. The alleles from the cultivated parent increased stem thickness. This QTL explained 13·0 % of the phenotypic variation.
The cultivated parent produced more branches on the main stem and developed the first branch at a higher internode than did the wild parent (Table 5). Only one significant QTL, Brn4.4.1, was identified for BRN. This QTL was detected close to the QTLs for seed- and pod-size-related traits. No significant QTL was found for BRP.
The frequency distribution in BC1F1 plants revealed that most of the plants flowered within a short period (about 10 d; Fig. S1 in Supplementary Data). As a result, no QTL was identified.
As has been shown in azuki bean (Kaga et al., 2008), it was expected that in rice bean it would take more time to maturity (translocation of dry matter to the seed) in the cultivated parent than in the wild parent, presumably because of the difference in seed size between the two parents. Although it was not possible to evaluate the PDDM of the wild parent precisely, one QTL was found, on LG4. This QTL explained 17·7 % of the phenotypic variation and was found near the QTLs for seed- and pod-size-related traits. The alleles from the cultivated parent delayed maturity.
The cultivated parent produced slightly more seeds per pod than the wild parent (Table 5). Three QTLs with small effects (5·4–9·3 %) on SDNPPD were identified, on LG4 and LG9 (two of them on LG9). The QTL on LG4 was detected close to the QTLs for seed- and pod-size-related traits. At all QTLs, alleles from the cultivated parent had a positive effect on SDNPPD.
The pale red hilum from the wild parent was dominant to the white hilum. The segregation ratio for this trait fitted the expected ratio (1 : 1). The recessive gene for white hilum from the cultivated parent was tentatively named sdhc4.5.1. This gene was mapped between AFLP markers E63M55-097 and E93M93-128 on LG5.
QTLs with large effect were found on LGF2, LG4 and LG7 (Fig. 2) as described below.
QTLs for seed- and pod-size-related traits were found in a narrow region. In the region near cowpea marker cp08299, QTLs for seed size (Sd100wt4.2.1, etc.) and pod size (Pdl4.2.1 and Pdw4.2.1) were found. At about 15 cM away from this region, a QTL for seed coat permeability (Sdwa4.2.1) was located. Interestingly, the cultivated parent alleles decreased seed coat permeability.
Between SSR markers CEDG103 and CEDG175, QTLs with a strong effect on seed, pod and leaf sizes were found. QTLs for seed number per pod and branch number were also found in this region. At about 20 cM away from this region, a QTL strongly associated with seed coat permeability (Sdwa4.4.1) was located. QTLs for pod maturity (Pddm4.4.1) and epicotyl length (Ecl4.4.1) were also linked to this area.
QTLs were distributed in two regions. Near SSR marker CEDG064, QTLs for pod dehiscence (Pdt4.7.1), middle to upper internode length (St4i4.7.1, etc.) and stem length (Stl4.7.1) were found. Between SSR marker CEDG215 and AFLP marker E91M91-185 (an interval of about 20 cM), QTLs for seed size (Sdwt4.7.1, etc.), epicotyl length and lower internode length (Ecl4.7.1, St1i4.7.1 and St2i4.7.1) were detected.
Although it was unknown whether all 73 QTLs identified for domestication-related traits had independent gene actions on each trait, the observed number of QTLs was compared with the expected number of QTLs based on each LG length (Table 7). The χ2 value was 47·6; this value is significant at P = 0·001, suggesting a departure from random distribution across the rice bean genome. The numbers of QTLs on LG5 and LG7 were significantly higher than expected, whereas the number of QTLs on LG6 was significantly lower than expected. Furthermore, χ2 tests of the number of QTLs in each 10-cM interval indicated a non-random distribution on LG1, LG2, LG3, LG4 and LG7 (Table 8).
The present rice bean linkage map is the first to be constructed from a large population derived from an intraspecific cross. The number of LGs corresponded to the basic number of chromosomes in rice bean. Many codominant SSR markers from azuki bean, cowpea and common bean were integrated into the genome, and no gap was >20 cM. Therefore, this rice bean map is useful for understanding genomic synteny among species within Ceratotropis, as well as for identifying QTLs for useful traits.
Two linkage maps had been constructed for populations derived from interspecific crosses between rice bean and azuki bean (Kaga et al., 2000) and between rice bean and V. nakashimae (Somta et al., 2006). Markers showing segregation distortion were observed on some LGs on both maps, suggesting some interspecific genetic barriers. Although the present rice bean linkage map was based on an intraspecific BC1F1 cross population (rice bean/wild rice bean//rice bean), 61 markers still showed segregation distortion (P < 0·05, 18·7 %). Notably, all 26 markers located on LG11 were highly skewed towards a heterozygous genotype, making the number of wild rice bean alleles higher than expected. A similar phenomenon was observed in a BC1F1 population of rice (Oryza sativa/O. glaberrima//O. sativa): marked segregation distortion towards a heterozygous genotype (increased number of O. glaberrima alleles) was recorded (Doi et al., 1998; Koide et al., 2008b). It was proposed that the gamete eliminator gene S1 (on rice chromosome 6) induces abortion of gametes carrying the opposite allele only in heterozygotes. The wild rice bean accession used here might possess a similar genetic factor that causes partial elimination of the opposite allele in heterozygotes.
A high level of marker segregation distortion on LG11 has also been reported on a rice bean × V. nakashimae F2 map (Somta et al., 2006). On rice chromosome 6, several genes affecting segregation distortion, such as cim (the cross-incompatibility reaction in the male), Cif (the cross-incompatibility reaction in the female) and S1 (gamete eliminator in O. rufipogon) have been detected (Koide et al., 2008a). Because all the markers distributed along a distance of about 40 cM on LG11 of rice bean showed segregation distortion, several genes existing on LG11 might have played an important role in genetic differentiation, both within the rice bean and among Vigna species.
The rice bean linkage map was compared with an azuki bean linkage map (Han et al., 2005) by the distribution of common azuki bean SSR markers (Fig. 3). Out of 172 azuki bean SSR markers mapped on the rice bean linkage map, 129 were common to those used in the azuki bean map. The common SSR markers were present on all LGs, and the number of common markers per LG ranged from 7 (LG3 and LG7) to 15 (LG1). There was a high level of co-linearity of marker order between rice bean and azuki bean.
However, a few differences, exemplified by genetic distance, inversions and duplications, were observed between the two linkage maps (Fig. 3). Of the 11 LGs, LG9 exhibited the most differences. The upper part of rice bean LG9 is considerably longer (about 1·4 times) than that in azuki bean. Three SSR markers, CEDG080, CEDG290 and CEDG172, on azuki bean LG9 are scattered to different rice bean LGs (LG3, LG5 and LG11, respectively). Inversions were usually found near the ends of LGs (on LG1, between CEDG133 and CEDG149; on LG3, between CED176 and CEDG010; on LG4, between CEDC028 and CEDC055 and between CEDG181 and CEDG011). In contrast, duplications of marker loci were found near the centres of LGs. The single azuki bean marker loci CEDG018 (LG5), CEDG081 (LG10), CEDG154 (LG4) and CEDG251 (LG8) were, respectively, duplicated on rice bean LG6 or in different positions on LG10, LG6 and LG1, whereas the rice bean single marker loci CEDG116 (LG10) and CEDG290 (LG5) were duplicated on azuki bean LG6 and LG9, respectively.
Even though the model legume Lotus japonicus has a small genome, large-scale segmental duplication at the sequence level has occurred within the genome (Sato et al., 2008). Meiosis in hybrids between rice bean and azuki bean is normal (Ahn and Hartmann, 1978), and chromosomes of both species have a similar C-banding pattern (Zheng et al., 1991). However, the fluorescent banding patterns of rice bean chromosomes, indicating the distributions of GC- and AT-rich genomic regions, have no similarity with those in azuki bean (Zheng et al., 1993). The differences between rice bean and azuki bean observed here can be explained by the occurrence of small chromosomal rearrangements or translocations during the evolutionary divergence of these species.
Commonly, domestication-related traits are controlled by several major genes that are not randomly distributed across crop genomes (Gepts, 2004) with the only one reported exceptional example of sunflower domestication (Burke et al., 2007; Wills and Burke, 2007). This general rule, which is also applicable to rice bean domestication, may be related to the phenomenon called ‘cultivation magnetism’ and should be considered under ‘protracted transition paradigm’ of crop domestication (Allaby, 2010).
Domestication QTLs of rice bean for various organs co-localized to several narrow genomic regions on LG1, LG2, LG3, LG4, LG5 and LG7 (Tables 7 and and88 and Fig. 3 and Fig. S2 in Supplementary data). In particular, the distribution of QTLs with large effects was limited to three LGs (LG2, LG4 and LG7). LG2 was associated with changes in seed and pod size. LG4 was associated with changes in seed and pod size and water absorption by seeds. LG7 was associated with changes in pod dehiscence and stem length. The remainder – LG1, LG3 and LG5 – were associated with changes in plant size.
Clustering of QTLs may be due to pleiotropy or close linkage of QTLs. Single mutations can have pleiotropic effects on various organs. In maize, QTLs related to change in inflorescence sex and number and length of internodes in lateral branches and inflorescences are distributed within a narrow genomic region (Doebley et al., 1995). Change in these traits is explained by the pleiotropic effect of a single tb1 gene. In arabidopsis, the seed testa colour mutant gene tt causes a reduction in seed weight (Debeaujon et al., 2000). Similarly, the seed testa colour mutant gene bks in tomato decreases seed weight and increases fruit pH (Downie et al., 2003). Further studies are required to determine whether pleiotropy or close genetic linkage is responsible for the clustering of QTLs in rice bean. Near isogenic lines for 2 clustering QTLs on LG4 are currently being developed to elucidate this question.
In contrast to the high degree of conservation of the genomic structure between rice bean and azuki bean (Isemura et al., 2007), the distribution of the main QTLs showed a marked difference (Table 9 and Fig. 4). For 28 domestication traits, 69 QTLs were found in rice bean and 76 in azuki bean, of which only 15 were common (Table 9 and Fig. 4). Most of the common QTLs were associated with seed-size-related traits (SD100WT, SDL, SDW and SDT) and pod dehiscence (PDT).
In azuki bean, QTLs with a large effect [phenotypic variation explained (PVE) > 20 %] were found on five LGs (LG1, seed dormancy and internode length; LG2, seed size and epicotyl and lower internode length; LG7, pod dehiscence and pod size; LG8, pod length; LG9, twining habit) (Fig. 4). In rice bean, in contrast, they were detected on only two LGs (LG4, seed and pod size and seed dormancy; LG7, pod dehiscence) (Fig. 4). In mung bean (T. Isemura et al., unpub. res.), 79 QTLs were found for 28 traits, and QTLs with a large effect (PVE > 20 %) were found on seven LGs. These findings reveal that mutations that cause large changes in domestication traits have occurred on fewer LGs in rice bean than in the two other species. Marked differences between rice bean and azuki bean were found on LG4 and LG9. Many QTLs with large effect were detected on LG4 of rice bean, whereas few QTLs were detected on this LG in azuki bean. In contrast, major QTLs for domestication-related traits were abundant on LG9 in azuki bean, whereas only a few QTLs were detected on LG9 of rice bean.
The accession with the largest seed among germplasm conserved in the NIAS Genebank was selected as the cultivated parent of rice bean. The 100-seed weight of this accession was 27·4 g (Table 5). This value is much larger than those of commonly cultivated rice bean [4·89–6·68 g (Bisht et al., 2005); 3·9 g (Somta et al., 2006)]. There was a 15-fold weight difference between the cultivated and wild rice bean (1·8 g per 100 seeds) used here, and seven QTLs were detected for 100-seed weight (Table 6). Two QTLs were involved in the 10-fold weight difference between cultivated (16 g per 100 seeds) and wild (1·7 g per 100 seeds) azuki bean (Isemura et al., 2007). These two 100-seed weight QTLs were common to the two species and were identified on LG1 and LG2 (Table 9 and Fig. 4). In addition, a rice bean-specific 100-seed weight QTL with the largest effect (PVE = 33·4 %) was detected on LG4 (Table 6).
Physical seed dormancy is generally caused by the presence of water-impermeable layers of palisade cells in the seed coat (Finch-Savage and Leubner-Metzger, 2006). The water absorption occurs through the structure of the strophiole (parenchymatous tissue) adjacent to the hilum in Vigna species (Gopinathan and Babu, 1985).
Five seed-dormancy-related QTLs were detected in rice bean (Table 6). Five were detected also in azuki bean (Isemura et al., 2007). However, none is common to both species. The seed-dormancy-related QTL with the largest effect was detected on LG4 in rice bean and on LG1 in azuki bean.
A dominant gene controlling pale red hilum in the wild parent was mapped on LG5. However, no QTL for seed dormancy was found in this region. This result suggests that the mutation for change of hilum colour had no effect on physical dormancy in rice bean.
The weak but positive effects on water absorption of the wild alleles for seed dormancy QTLs on LG2 and LG10 might reflect the modest degree of seed dormancy in the wild parent (Table 5). Passport data revealed that the natural habitat of the wild parent was near a river in Kanchanaburi province, Thailand, where the soil may be moist and the temperature high enough for growth throughout the year (Tomooka et al., 2000a). Therefore, strong seed dormancy during the dry season may not be important for this wild accession, and this may suggest why the wild accession possesses a QTL that reduces seed dormancy.
A single QTL for pod dehiscence was detected in rice bean and azuki bean. In rice bean, a pod dehiscence QTL with a relatively high contribution to this trait (42·4 %) was found on LG7. In azuki bean, a QTL with a higher contribution (90·5 %) was detected at a similar map position (Isemura et al., 2007) and was considered common to the two species (Table 9 and Fig. 4).
The differences in the large-effect domestication QTLs suggest that rice bean and azuki bean harbour specific useful genes and can play important roles as new gene resources for each other. For example, a QTL for twining habit with large effect on LG9 is specific to azuki bean, whereas the 100-seed weight QTL with large effect on LG4 is specific to rice bean (Fig. 4). The seed-dormancy-related QTL with the largest effect was detected on LG4 in rice bean and LG1 in azuki bean. Similarly, the pod-size-related QTL with the largest effect was detected on LG4 in rice bean and LG7 in azuki bean, and no common QTL was found.
Hybrids can be produced between rice bean and azuki bean when rice bean is used as the seed parent (Kaga et al., 2000). Therefore, it seems that the seed weight QTL on LG4 in rice bean and the QTL for loss of twining habit on LG9 in azuki bean are particularly useful for the breeding of larger-seeded azuki bean cultivars and determinate rice bean cultivars, respectively.
This is the first report of a genetic linkage map and QTLs for domestication-related traits in rice bean. A genetic linkage map of rice bean was constructed using 223 SSR markers from related legume species and 103 AFLP markers. A total of 326 markers converged into 11 LGs, corresponding to the haploid number of rice bean chromosomes. Using this map, 31 domestication-related traits in rice bean were dissected into 69 QTLs, but the differences between cultivated and wild parents were found to be controlled by only a few major QTLs. Major QTLs for unrelated organs were distributed as clusters on LGs 2, 4 and 7. The inheritance of domestication-related traits was so simple that the few major QTLs explained the phenotypic variation between cultivated and wild rice bean. The high level of genome synteny between rice bean and azuki bean facilitated QTL comparison between the species. Major QTLs in rice bean were found on LG4, whereas major QTLs in azuki bean were found on LG9. The results suggest that the degree of domestication or divergence between wild and domesticated rice bean was lower than in azuki bean. The results provide a genetic foundation for improvement of rice bean. Interchanges of major QTLs between rice bean and azuki bean might be useful to broaden the genetic variation in both species.
Supplementary data are available online at www.aob.oxfordjournals.org/. Table S1: Information for cowpea SSR primer pairs designed in the present study. Table S2: Correlation coefficients between traits in the BC1F2 (or BC1F1:2) population of the cross between cultivated rice bean and wild rice bean. Fig. S1: Frequency distributions of all traits examined in the BC1F2 (or BC1F1:2) population. Fig. S2: Distributions of the QTLs found on each LG in the populations analysed and the effects of the cultivated rice bean parent.
We thank the following staff of the National Institute of Agrobiological Sciences for technical support: T. Nobori, T. Yoshida, N. Karino, Y. Ito, S. Hirashima, T. Nemoto, K. Sugimoto, H. Uchiyama, T. Taguchi, M. Akiba, J. Inoue, T. Misawa and H. Tomiyama. This research was supported by a Grant-in-Aid for Scientific Research Japan (KAKENHI No. 18390009) from the Japan Society for the Promotion of Science and by the Genebank project of the Ministry of Agriculture, Forestry and Fisheries of Japan.