Analysis of genetic diversity of rapeseed genetic resources in Japan and core collection construction.

Diversity analysis of rapeseed accessions preserved in the Japanese Genebank can provide valuable information for breeding programs. In this study, 582 accessions were genotyped with 30 SSR markers covering all 19 rapeseed chromosomes. These markers amplified 311 alleles (10.37 alleles per marker; range, 3-39). The genetic diversity of Japanese accessions was lower than that of overseas accessions. Analysis of molecular variance indicated significant genetic differentiation between Japanese and overseas accessions. Small but significant differences were found among geographical groups in Japan, and genetic differentiation tended to increase with geographical distance. STRUCTURE analysis indicated the presence of two main genetic clusters in the NARO rapeseed collection. With the membership probabilities threshold, 227 accessions mostly originating from overseas were assigned to one subgroup, and 276 accessions mostly originating from Japan were assigned to the other subgroup. The remaining 79 accessions are assigned to admixed group. The core collection constructed comprises 96 accessions of diverse origin. It represents the whole collection well and thus it may be useful for rapeseed genetic research and breeding programs. The core collection improves the efficiency of management, evaluation, and utilization of genetic resources.

Introduction

Rapeseed (Brassica napus L.; genome AACC, 2n = 38) is one of the most economically important oilseed crops worldwide. It is thought to have originated from multiple independent natural hybridization events between B. rapa (AA, 2n = 18) and B. oleracea (CC, 2n = 18) (U 1935 ). The oldest description of rapeseed cultivation was found in the Indian literature of 2000 to 1500 BCE (Röbbelen ). Nowadays, over 450 million tonnes of seeds are harvested worldwide every year, which accounts for about 20% of world grain production (Carré and Pouzet 2014). In 2012, the EU led the worldwide rapeseed production (~20 Mt), followed by Canada (15 Mt), China (12 Mt), and India (6 Mt) (ISTA Mielke GmbH 2012). In Japan, since the Meiji period (~150 years ago), many crops, including rapeseed, were imported and cultivated. Modern rapeseed breeding in Japan began in the 1930s; selective breeding of overseas rapeseed cultivars and crossing between rapeseed and B. rapa were implemented to breed cold-resistant and early maturing cultivars suitable for the Japanese climate and cultivation methods (Matsuo 1954, Shiga 1970, Yamamori 2006). During the second half of the 20th century, crossbreeding between Japanese and new European cultivars was performed to obtain single-zero (low erucic acid content) and double-zero (single-zero plus low glucosinolate content) cultivars, and many such cultivars were released (Ishida 2003).

Genetic resources are key to progress in breeding, contributing to sustainable agriculture and associated industries. More than 600 accessions, including >300 Japanese landraces and cultivars, are maintained at the Genetic Resources Center of the National Agriculture and Food Research Organization (NARO Genebank). A collection of this size is expected to contain a wide range of genetic variation and offer opportunities for trait improvement. However, screening for target traits in the whole collection is often time-consuming, laborious, and costly. To efficiently use germplasm collections in breeding, knowledge of the genetic diversity and population structure among the germplasms is vital. In addition, establishing a representative subset of the whole collection, called a core collection, will be a feasible way for management, evaluation, and utilization of genetic resources.

A core collection should represent most of the diversity in the whole collection and should cover at least 70% of the alleles present in the whole collection (Brown 1989). Phenotypic traits have been used for constructing core collections, but they are easily affected by environmental conditions and are sometimes difficult to measure accurately. Recently, many core collections have been established by using DNA marker data, because DNA markers can be easily obtained and are highly stable and polymorphic (Taniguchi , Xu , Zhao ). In rapeseed, several studies revealed genetic diversity of different collections using molecular markers such as RFLP (Song and Osborn 1992), RAPD (Ma ), SRAP (Riaz ), AFLP (Seyis ), and SSR (Hasan ). SSR markers are superior because they are able to detect multiple alleles per locus, are codominant at a single locus, and are relatively evenly distributed across genomes (Zalapa ); they have been widely used in recent diversity analyses of Brassica species (El-Esawi , Guo , Yao ).

Over the last five decades, effort devoted in Japan to rapeseed breeding has resulted in considerable progress. Despite a large number of germplasms in NARO Genebank, only a few studies analyzed the small number of either Japanese or overseas accessions in this collection (Diers and Osborn 1994, Gyawali , Hasan , Ma ). The number of rapeseed germplasms in the NARO Genebank has increased owing to both past studies and exploration of genetic resources. The aims of this study were to examine the genetic diversity and population structure of the NARO rapeseed collection by using single-locus SSR markers, and to establish a core collection with good representation of the genetic diversity present in the whole collection.

Materials and Methods

Plant materials

The rapeseed collection in NARO Genebank hold 639 germplasms, the seeds are preserved in cool dry storage. We excluded accessions from our analysis if both parental lines were in the collection. The total number of accessions analyzed was 582 (Supplemental Table 1), including 305 Japanese accessions (landraces, breeding lines, and cultivars) and 277 overseas accessions from all inhabited continents except Africa.

Genotyping

Seeds were germinated on moistened filter paper in 9-cm Petri dishes and then transplanted into a 72-cell tray filled with granular culture soil (Nippi-Engei-Baido-1gou soil, Nihon Hiryo Co., Ltd., Tokyo, Japan). Seedlings were grown in a climate chamber (25/22°C 12 h/12 h) until the first true leaves had fully expanded. Genomic DNA was extracted from the first true leaf of each seedling by using the DNeasy Plant Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer’s protocol with some minor modifications.

A total of 502 Brassica SSR markers (Supplemental Table 2; AAFC Consortium 2016, Cheng , Hatakeyama , Iniguez-Luy , Kim , Li , 2013, Lowe , Piquemal , Suwabe , Wang , Xu ) were prescreened against 8 representative accessions that were chosen on the basis of their origins and breeding histories: OOMINATANE (Japan), MICHINOKUNATANE (Japan), ASAHINATANE (Japan), CASCADE (USA), HAMBURG 1 (Germany), WESTAR (Canada), PROTA (Germany) and RAPORA (Korea). SSR markers with clear reproducible polymorphic amplification products were then applied to all other accessions. PCR mixtures (10 μl) contained template DNA (10 ng), 1× KAPA 2G buffer A, 200 nM dNTP, 0.5 mM MgCl2, 0.1 U KAPA 2G Fast DNA polymerase (KAPA Biosystems Inc., Woburn, MA, USA), 2 pmol reverse primer, and 0.5 pmol forward primer. The forward primers were 5′-labeled with the fluorescent dyes 6-FAM, VIC, NED, or PET (Shimizu and Yano 2011). PCR was performed in a C1000 Thermal Cycler (Bio-Rad Laboratories, Inc., Hercules, CA, USA) as follows: initial denaturation at 94°C for 3 min; 30 cycles of 94°C for 20 s, 54°C for 30 s, and 62°C for 30 s; 3 cycles of 94°C for 20 s, 49°C for 10 s, and 72°C for 5 s; and a final extension at 72°C for 10 min. The size of the amplified fragments was estimated by using an automated DNA analyzer (model 3130xl) with a GeneScan-600LIZ size standard and GeneMapper v. 4.0 software (all Thermo Fisher Scientific, Inc., Waltham, MA, USA).

Genetic diversity and population structure analysis

The number of alleles, major allele frequency, number of rare alleles (frequency <5%), and polymorphism information content (PIC) were calculated for the whole collection and for each geographic group in PowerMarker v. 3.25 software (Liu and Muse 2005). Observed heterozygosity (Ho), expected heterozygosity (He), Shannon’s information index (I), pairwise F statistics (FST), and Nei’s genetic distance (Nei ) were calculated in GenAlEx v. 6.502 software (Peakall and Smouse 2012). The components of variance between Japanese and overseas accessions, between different groups and among individuals within groups were estimated from the genetic distance matrix, as specified in the analysis of molecular variance (AMOVA) procedure using Arlequin 3.5.2.2 software (Excoffier and Lischer 2010). A nonparametric permutation procedure with 1000 permutations was used to test the significance of variance components associated with the different possible levels of genetic structure in this study (Excoffier ).

Genotyping data for the SSR markers were analyzed by using the model-based STRUCTURE v. 2.3.4 software (Pritchard ) to determine the most probable number of clusters (K value) and to assign rapeseed accessions to different clusters. The K value was determined by running an admixture and related frequency model with K = 1 to 20 (10 replications per K value); the burn-in period of each run and the Monte Carlo Markov Chain (MCMC) lengths were both set to 100,000. The website program STRUCTURE HARVESTER was used to estimate the optimal number of K value (Earl 2012). This program follows the ΔK method of Evanno . The same set of genotyping data was used to perform principal coordinates analysis (PCoA) in GenAlEx 6.502.

Construction of core collections

Core collections were constructed using four different methods: two maximization (M) strategy methods with different algorithms, the Core Hunter method, and the random sampling method. Both M strategy methods select a core collection to maximize the number of alleles of the SSR markers. CoreFinder v. 1.1 (Cipriani ) is based on the NP-complete set-covering problem and uses a Las Vegas–style randomized algorithm. PowerCore v. 1.0 (Kim ) uses an advanced M strategy with a modified heuristic algorithm (A*) that can find the optimum path for sample selection. Core Hunter v. 2.0 (De Beukelaer ) is a fast core-subset-selection program based on multiple genetic diversity measures and using a Mixed Replica search algorithm. The software allows choosing sampling intensity and the genetic measures to be used as selection criteria. The Core Hunter software was run in R (http://www.r-project.org/); the modified Roger’s distance (weight 0.7) and Shannon’s information index (weight 0.3) were chosen to define a core collection comprising about 20% of the entry collection. In the random sampling method, before sampling, a neighbor joining (NJ) tree based on Nei’s genetic distance is constructed in MEGA7 software (http://www.megasoftware.net/). Sampling was performed according to Hu : initially, an accession is selected at random from each lower-order level subgroup, if there is just one accession in the subgroup, it is sampled; if there are two accessions, one is randomly selected. A new dendrogram was then generated, and the process was repeated until the number of selected accessions was reduced to 20% of the initial collection. To evaluate the representativeness of the different core collections, genetic diversity indices (allele number, allele retention ratio, number of effective alleles, Ho, He, I, and PCI) were calculated in PowerMarker v. 3.25 and GenAlEx 6.502 software. The differences between the whole collection and each core collection were tested by t-test.

Results

Genetic diversity of the NARO rapeseed collection

Prescreening selected 30 of the 502 SSR markers. The selected markers amplified a total of 311 alleles in the 582 accessions, ranging from 3 (BoGMS0660 and BoEMS0049) to 39 (BrGMS0070) per marker, with a mean allelic richness of 10.37 (Table 1); 214 alleles (68.8%) with low frequency (<0.05) were regarded as rare alleles. In addition, the selected 30 SSR markers amplified only one or two alleles in each accession. Average genetic diversity indices were 0.56 for major allele frequency, 0.57 for He, and 0.52 for PIC. The average Ho value was 0.05 (range, 0.01–0.34), indicating that almost all accessions in the whole collection were highly homozygous.

Table 1

Statistics of the 30 SSR markers used for genotyping of 582 rapeseed accessions

Marker name	Linkage group	Number of alleles	Number of rare alleles	Major allele frequency	Ho	He	PIC
BrGMS4028a	A1	9	5	0.46	0.34	0.71	0.67
BrGMS4031a	A1	5	3	0.86	0.04	0.24	0.22
BRAS084b	A1	22	17	0.30	0.10	0.83	0.81
BrGMS1411a	A2	7	4	0.58	0.05	0.59	0.55
BrGMS0667a	A2	15	8	0.51	0.06	0.71	0.69
BrGMS2498a	A3	4	2	0.52	0.04	0.52	0.40
sN2025c	A4	10	5	0.44	0.04	0.70	0.65
BrGMS2252a	A5	4	2	0.82	0.03	0.32	0.29
BrGMS0070a	A5	39	36	0.28	0.06	0.87	0.86
BnEMS0753d	A6	7	4	0.63	0.02	0.52	0.46
BrGMS3750a	A6	6	2	0.60	0.02	0.57	0.53
BrGMS3837a	A7	8	6	0.60	0.01	0.51	0.41
BnEMS0620d	A7	17	14	0.54	0.08	0.64	0.60
BrGMS0742a	A8	11	8	0.61	0.03	0.55	0.50
BnGMS0281e	A9	8	4	0.43	0.06	0.71	0.66
BrGMS3857a	A10	5	2	0.61	0.06	0.55	0.49
BrGMS3688a	A10	6	2	0.41	0.06	0.71	0.66
BrGMS0086a	A10	12	7	0.41	0.06	0.76	0.73
BnGMS271e	C1	9	6	0.48	0.02	0.59	0.50
BoGMS2016f	C2	15	9	0.30	0.03	0.84	0.82
BoEMS0016g	C2	9	6	0.50	0.05	0.65	0.60
BoGMS0660f	C2	3	1	0.61	0.01	0.48	0.37
BnGMS0289e	C3	21	18	0.44	0.03	0.62	0.55
BnGMS347e	C4	10	8	0.44	0.04	0.62	0.55
BoGMS0037f	C5	9	7	0.85	0.01	0.26	0.23
BoGMS1909f	C6	13	10	0.78	0.03	0.37	0.36
BnGMS0353e	C6	8	5	0.54	0.04	0.62	0.57
BoEMS0049g	C7	3	1	0.53	0.01	0.50	0.38
BnGMS0336e	C8	4	1	0.81	0.17	0.33	0.31
BoGMS0525f	C9	12	11	0.89	0.07	0.21	0.21
Average		10.37	7.13	0.56	0.05	0.57	0.52

Ho, observed heterozygosity; He, expected heterozygosity; PIC, polymorphism information content.

Detailed marker information is available in

Xu ,

Piquemal ),

AAFC Consortium (2016),

Wang ,

Cheng ,

Li ,

Li .

The genetic diversity indices for geographical groups are summarized in Table 2. The average number of alleles per maker was 8.50 in Japanese germplasms and 9.37 in overseas germplasms. Among Japanese geographical groups, the average number of alleles per marker ranged from 2.90 (Chugoku-Shikoku area) to 5.43 (Chubu area), whereas overseas geographical groups had a wider variation, from 3.13 (Oceania) to 8.07 (Europe). The values of all other genetic diversity indices (Ho, He, I, PIC) were lower in Japanese than in overseas geographical groups. Hierarchical analysis of molecular variance revealed significant differences at all hierarchical levels (Table 3); the percentage of variation was highest among individuals within geographical groups (67.18%), followed by that between Japanese and overseas accessions (17.71%).

Table 2

Genetic diversity indices for the NARO rapeseed collection and variations among geographic groups

Geographical group	Number of accessions	Average number of alleles	Major allele frequency	Ho	He	I	PIC
Whole Collection	582	10.37	0.56	0.05	0.57	1.18	0.52
Japan	305	8.50	0.67	0.04	0.44	0.90	0.41
Hokkaido	9	3.30	0.58	0.03	0.52	0.91	0.47
Tohoku	69	5.27	0.65	0.06	0.46	0.88	0.44
Kanto	33	4.77	0.68	0.06	0.42	0.82	0.41
Chubu	40	5.43	0.65	0.04	0.46	0.90	0.44
Kinki	72	5.17	0.73	0.03	0.37	0.73	0.37
Chugoku-Shikoku	11	2.90	0.70	0.07	0.37	0.65	0.36
Kyushu	65	4.63	0.72	0.03	0.37	0.71	0.36
Unknown	6	–	–	–	–	–	–
Overseas	277	9.37	0.50	0.06	0.59	1.22	0.58
Europe	202	8.07	0.53	0.06	0.56	1.14	0.54
Asia	30	5.37	0.60	0.07	0.49	0.97	0.49
Oceania	8	3.13	0.61	0.05	0.48	0.84	0.45
America	27	4.40	0.63	0.05	0.47	0.88	0.45
Unknown	10	–	–	–	–	–	–

Ho, observed heterozygosity; He, expected heterozygosity; I, Shannon’s information index; PIC, polymorphism information content.

Table 3

Hierarchical analysis of molecular variance in the NARO rapeseed collection

Source of variation	d.f.	Parcentage of variation (%)
Between Japanese and overseas accessions	1	17.71 ***
Among geographical groups	9	7.01 ***
Among individuals	555	67.18 ***
Within individuals	566	8.10 **

significant at the 0.1% and 1% levels, respectively, for 1000 permutations.

There was significant genetic differentiation between geographic groups in Japan, except between Kanto and Tohoku and between Kanto and Chubu. Pairwise FST ranged from 0.017 (Kando and Chubu) to 0.192 (Hokkaido and Chugoku-Shikoku) (Table 4). The FST values between Hokkaido and western Japanese geographic groups (Kinki, Chugoku-Shikoku, and Kyushu) showed relatively high differentiation. Nei’s genetic distance among Japanese geographic groups ranged from 0.044 (Kinki and Kyushu) to 0.173 (Hokkaido and Chugoku-Shikoku). Similar to pairwise FST, the genetic distances tended to be higher between the Hokkaido group and western Japanese geographic groups.

Table 4

Pairwise FST (below diagonal) and Nei’s genetic distance (above diagonal) among geographic groups in Japan

	Hokkaido	Tohoku	Kanto	Chubu	Kinki	Chugoku-Shikoku	Kyushu
Hokkaido		0.105	0.125	0.110	0.148	0.173	0.143
Tohoku	0.059*		0.055	0.056	0.063	0.091	0.068
Kanto	0.102**	0.019		0.053	0.048	0.099	0.058
Chubu	0.122**	0.031**	0.017		0.056	0.091	0.049
Kinki	0.101**	0.039**	0.047**	0.058**		0.114	0.044
Chugoku-Shikoku	0.192**	0.084**	0.067*	0.076*	0.088**		0.075
Kyushu	0.138**	0.034**	0.063**	0.092**	0.055**	0.100**

significant at the 1% level;

significant at the 5% level.

Population structure of the NARO rapeseed collection

The STRUCTURE analysis suggested K = 2 as the real value of K even though a secondary peak of ΔK exists at K = 4 (Supplemental Fig. 1), indicating there are two genetic clusters in the NARO rapeseed collection (Fig. 1). With the membership probabilities (Q) threshold of 0.80, 227 accessions, mostly originating from overseas, were assigned to subgroup 1, and 276 accessions, mostly originating from Japan were assigned to subgroup 2 (Table 5). The remaining 79 accessions are assigned to admixed group, i.e., Q < 0.8. The result of PCoA was similar to that of STRUCTURE analysis (Fig. 2). The first and second factors of the PCoA explained approximately 19.1% and 7.3% of the variation in the genetic distance matrix, respectively. A bi-dimensional PCoA scatter plot indicated differentiation between Japanese and overseas accessions (Fig. 2).

Fig. 1

Population structure of 582 rapeseed accessions determined (at K = 2) by using 30 single-locus SSR markers.

Table 5

Geographic origin of rapeseed accessions assigned by STRUCTURE to two subgroups

Geographical group	Number of accessions	Subgroup 1a	Subgroup 2a	Admixed
Whole Collection	582	227	276	79
Japan	305	19	240	46
Hokkaido	9	4	3	2
Tohoku	69	5	46	18
Kanto	33	2	27	4
Chubu	40	4	27	9
Kinki	72	0	66	6
Chugoku-Shikoku	11	1	9	1
Kyushu	65	1	60	4
Unknow	6	2	2	2
Overseas	277	208	36	33
Europe	202	171	9	22
Asia	30	3	22	5
Oceania	8	5	0	3
America	27	25	1	1
Unknow	10	4	4	2

Accessions are considered belonging to either of two subgroup when membership probabilities of ≥0.8.

Fig. 2

Principal coordinates analysis (PCoA) of the 582 rapeseed accessions. Open symbols indicate Japanese accessions, and solid symbols indicate overseas accessions. Squares indicate accessions assigned to group 1 and circles indicate accessions assigned to group 2 by STRUCTURE analysis.

Construction of a core collection

Four different core collections were developed by different methods. Sample size and genetic diversity indices of each core collection are shown in Table 6. The random sampling strategy and CoreHunter selected the largest number of accessions (116; 19.93% of the whole collection), whereas the CoreFinder method selected the smallest number (96; 16.49%). The core collections constructed by the two M strategy methods (PowerCore and CoreFinder) had the highest allele retention ratio (100%), indicating that they had all alleles observed in the whole collection, whereas approximately 30% of alleles were lost in the two other core collections. Other indices (except Ho) were higher in the core collections constructed by the two M strategy methods than by the other two methods, although there were no significant differences among the core collections or between each core collection and the whole collection. Comparison of genetic diversity indices indicated that the core collections developed by PowerCore and CoreFinder were highly representative of the whole collection. Since the core collection by CoreFinder contains fewer accessions (96) than that by PowerCore (103), the former should be used as the core collection of the NARO rapeseed collection (Table 6). This core collection includes 39 Japanese accessions from seven geographic areas (Hokkaido, 2; Tohoku, 7; Kanto, 7; Chubu, 7; Kinki, 8; Chugoku-Shikoku, 1; Kyushu, 5; unknown, 2) and 57 overseas accessions (America, 6; Asia, 6; Europe, 41; Oceania, 2; unknown, 2) (Supplemental Table 1). A bi-dimensional PCoA scatter plot also demonstrated that this core collection maintains genetic diversity, although it included fewer than 20% of accessions present in the whole collection (Fig. 3).

Table 6

Genetic diversity indices of four core collections constructed by different methods

	Number of accessions	Number of alleles	Allele retantion ratio	Number of effective alleles	Ho	He	I	PIC
PowerCorea	103	10.37	100.0	3.77	0.05	0.64	1.41	0.59
CoreFindera	96	10.37	100.0	3.57	0.06	0.62	1.39	0.58
CoreHunter	116	7.50	72.3	3.17	0.06	0.60	1.23	0.55
Random sampling strategy	116	7.20*	69.5	2.89	0.07	0.57	1.17	0.52

significant at the 5% level between core collection and whole collection.

Ho, observed heterozygosity; He, expected heterozygosity; I, Shannon’s information index; PIC, polymorphism information content.

Maximization strategy methods.

Fig. 3

A principal coordinate plot of the core collection constructed using CoreFinder and the whole collection. Solid circles indicate the accessions included in the core collection, and open circles indicate accessions not included.

Discussion

We evaluated the genetic diversity of the NARO rapeseed collection by using single-locus SSR markers. In allopolyploid plants such as rapeseed, SSR primers usually amplify several alleles from multiple loci, which makes it difficult to assign these alleles to individual loci. Li found a set of 230 single-locus rapeseed SSR markers. However, they used only 6 inbred lines and noted that not all of the markers might be true single-locus markers. In this study, we applied a total of 502 SSR markers, including the 230 markers of Li , to 582 rapeseed accessions, and found that 30 SSR markers (Table 1), distributed across all linkage groups, always amplified single alleles with high polymorphism. This finding indicates that these 30 SSR markers used in this study are useful for research on rapeseed genetic resources.

We identified 311 alleles in the whole collection, with an average of 10.37 alleles per locus and a PIC value of 0.52. Several studies calculated genetic diversity indices of diverse rapeseed collections on the basis of SSR markers. The average numbers of alleles per locus were 4.31 in 96 European rapeseed genotypes (Hasan ), 3.4 in 192 inbred lines of various origins (Xiao et al 2012), and 7.3 in 169 worldwide rapeseed lines (Gyawali ). The PIC value in the 192 lines analyzed by Xiao was 0.37. The differences in allelic richness between the NARO collection and other rapeseed collections are due partly to the differences in the SSR markers and the number of accessions used in this study. Nevertheless, the higher values of the genetic diversity indices in this study indicate that the NARO rapeseed collection has moderately high genetic diversity. In contrast, the PIC value of the NARO rapeseed collection was lower than those reported in other Brassica species: 0.57 or 0.64 in B. oleracea (El-Esawi , Louarn ) and 0.90 in B. juncea (Yao ); this would reflect a narrower genetic base in rapeseed than in other Brassica species.

Rapeseed originated and was first domesticated in Europe, and then spread throughout the world. In Japan, rapeseed cultivation is considered to have spread after the Meiji period (from 1868 to 1912) (Matsuo 1954); since then, many landraces have been developed throughout the country. Modern rapeseed breeding in Japan began in the 1930s; government-led large-scale breeding projects initially used a limited number of accessions (Matsuo 1954), which may have restricted the genetic diversity of Japanese accessions. As expected, we found a significant genetic differentiation between Japanese and overseas accessions (Fig. 2, Tables 3, 4), and comparison of the genetic indices indicated that genetic diversity was lower in Japanese accessions than in overseas accessions (Table 2). However, the values of genetic diversity indices were not very different between Japanese and overseas accessions. Over the last five decades, repeated introduction of overseas germplasms and subsequent breeding may have increased genetic diversity.

In general, differences in breeding materials and targets among different breeders and regions may lead to genetic differentiation. In Japan, we found small but significant differences among geographical groups (Table 4). Genetic differentiation tended to increase with increasing geographic distance. For example, we found the highest FST values and Nei’s genetic distances between the Hokkaido group (northern Japan) and three groups in western Japan. The genetic differentiation might be caused not only by the diversity of germplasms used as breeding materials, but also by their adaptation to local environments. The genetic diversity indices PIC and I were higher in eastern than in western Japan. The genetic diversity of eastern Japanese accessions may have been further increased by recent active rapeseed breeding there.

Our STRUCTURE analysis indicated the presence of two main genetic clusters in the NARO rapeseed collection and demonstrated differentiation between Japanese and overseas accessions: with a membership probability threshold of 0.80, most of the overseas accessions were assigned to the subgroup 1, and the Japanese accessions were assigned to the subgroup 2 (Table 5). Yet the presence of several accessions in each subgroup is not consistent with their geographical origins (Table 5), as similarly observed in previous studies (Chen , Hu , Li , Xiao ). All of the Canadian cultivars, known by the trade name ‘Canola’ and bred in the 1970s or later, were classified into the subgroup 1. Many accessions from Asian countries other than Japan belonged to the subgroup 2. The principal coordinate analysis (PCoA) further confirmed the STRUCTURE results (Fig. 2). These results indicate that Asian rapeseed may have genetically differentiated from that in other regions of the world, and that Japanese, Chinese, and Korean landraces and cultivars may be closely genetically related (Fig. 1). Some studies have revealed a wide diversity of Chinese and Australian rapeseed accessions (Chen , Guo , Wang ). The NARO collection holds only 7 accessions from China and none from Australia. Obtaining accessions from these countries would further expand the diversity of the collection.

The most important factor in core collection construction is sample selection strategy, and many strategies based on stratified sampling and clustering methods are available (Zhang ). Because different strategies result in different core collections (Thachuk , Wang ), we used four different methods. As the M strategy is based on maximizing the number of alleles, it can automatically generate a sampling ratio on the basis of the genetic diversity of the species; this strategy has been widely used in recent years (Belaj , Liu , Zhang ). In this study, both PowerCore and CoreFinder, which are based on the M strategy, constructed core collections that retained 100% of alleles, and in this respect were superior to the other two methods used. The genetic diversity indices tended to be nearly same in both the PowerCore collection and the CoreFinder collection (Table 6). The number of accessions are fewer in the CoreFinder collection (Table 6), so we finally recommend it as the core collection of the NARO rapeseed collection. The core collection comprised 57 overseas accessions (20.6% of total) and 39 Japanese accessions (12.8%). The higher percentage of overseas accessions reflects their high allelic richness. However, the fact that approximately 40% of accessions in the core collection originated in Japan indicates their relatively high genetic diversity.

In conclusion, our study revealed the genetic structure of the NARO rapeseed collection and genetic relationships among accessions on the basis of single-locus SSR markers. Some accessions were genetically identical or closely related to each other. This information is important for decreasing redundancy in the collection, thereby reducing the management cost and avoiding unnecessary distribution of such accessions to breeders. Further integration of the data from other collections maintained in different countries will make it possible to exploit and preserve the whole rapeseed gene pool and to retain the largest number of allelic variants for genes controlling the most important agronomic traits in the NARO collection. Our candidate core collection can be used in further research such as genome-wide association studies to identify genomic regions controlling important agronomic traits. We did not analyze the phenotypic data recorded in the NARO rapeseed collection because they have been obtained by different investigators and methods under various environmental conditions. Therefore, it will be necessary to evaluate the phenotypes of all the accessions under the same environmental conditions, and integrate the data on genotypic and phenotypic diversity. This will make the core collection, which comprises accessions that are genetically diverse at both genotypic and phenotypic levels, available to breeders for enhancing the genetic potential of this crop.