Abstract
The chestnut (genus Castanea) has a long juvenile phase, and breeders have to wait three years or more to evaluate nut traits. Therefore, molecular markers associated with genes of interest are required to speed the selection process in chestnut breeding programs. Genetic linkage maps of the Japanese chestnut were constructed using two breeding populations derived from crosses between ‘Kunimi’ and breeding line ‘709-034’ (Kx709), and between ‘Porotan’ and ‘Tsukuba-43’ (Px43). Maps of the four parents and two integrated maps (one representing each cross) were constructed using 443 simple sequence repeat markers (SSRs) and 554 single-nucleotide polymorphism markers. In the Kx709 integrated map, which was the most saturated of the six maps, 12 linkage groups were identified that covered 668.1 cM with an average distance of 0.8 cM between loci. Using anchor SSRs, these six maps were successfully aligned to the Chinese chestnut consensus map. We evaluated eight important traits, including several nut traits, to identify molecular markers associated with these traits. At least one significant quantitative trait locus (QTL) was detected for each of the eight traits (21 in total). Logarithm of odds (LOD) values and phenotypic variance explained by these QTLs ranged from 2.60 to 7.90 and from 11.6% to 29.1%, respectively. In the Kx709 population analysis, several QTLs for nut harvesting date (HARVEST) and pericarp splitting (SPLIT) were detected. Under the assumption that the effects of these QTLs are additive, the percentage of total phenotypic variance explained by the combination of QTLs was high for both HARVEST (47.5%–60.8%) and SPLIT (33.4%–41.7%). Because these mapping populations and their parents are essential materials for Japanese chestnut breeding programs, these QTLs will soon be used for marker-assisted selection to improve breeding efficiency.
Introduction
The chestnut belongs to the genus Castanea in the family Fagaceae, which also includes Quercus, Fagus, and Castanopsis. At least seven Castanea species are recognized (Pereira-Lorenzo et al., 2012). The Japanese chestnut (C. crenata Sieb. et Zucc.), Chinese chestnut (C. mollissima Bl.), and European chestnut (C. sativa Mill.) are commercially grown in temperate areas in Japan and the Korean peninsula, in China, and in Europe and Asia Minor, respectively. The American chestnut (C. dentata Borkh.) had been used as a source of tannins and many wood products in North America, but was devastated by chestnut blight in the early 20th century (Wheeler and Sederoff, 2009). The seguin chestnut (C. seguinii Dode.) and the Henry chestnut (C. henryi Rehd. et Wils.), which originated in China, and the Allegheny chinkapin (C. pumila Mill.), which originated in the USA, are recognized as wild species.
The Japanese chestnut was used as an edible nut and for wood in Japan from ancient times until the mid-20th century (Motoki, 2007; Sawamura, 2006). These days, it is grown only for food; in 2013, chestnut orchards covered more than 20600 ha (FAOSTAT). An organized Japanese chestnut breeding program was started in 1947 at the National Institute of Fruit Tree Science in Japan (Kotobuki et al., 1999; Pereira-Lorenzo et al., 2012) and continues at the Institute of Fruit Tree and Tea Science, National Agriculture and Food Research Organization (NIFTS). The breeding objectives of the program have included an easily peeled pellicle with a heavy nut weight (WEIGHT); absence of pericarp splitting (SPLIT); freedom from infestation by insects (INSECT), specifically the peach moth, Conogethes punctiferalis (Lepidoptera: Crambidae); and high yield. In addition, it is desirable to broaden the range of nut harvest dates so that new cultivars will have different maturation dates and can thereby provide ripe chestnuts over a wider period. Molecular markers associated with the easy-peeling pellicle trait have already been used in our chestnut breeding program (Nishio et al., 2013). On the other hand, only a few genetic studies of other traits have been conducted in Castanea. These traits were found to be controlled by quantitative trait loci (QTLs), and broad-sense heritabilities in major Japanese chestnut cultivars were estimated as 0.84 for nut harvesting day (HARVEST), 0.27 for WEIGHT, 0.48 for SPLIT, and 0.17 for INSECT (Nishio et al., 2014). Nut tree species have a long juvenile phase, so breeders have to wait three years or more to evaluate nut traits. Therefore, molecular markers associated with these traits will speed up the selection process.
Although more than 300 simple sequence repeat markers (SSRs) have been developed in the Japanese chestnut and their high transferability to other Castanea species has been confirmed (Nishio et al., 2011), no genetic linkage map has been constructed for the Japanese chestnut. On the other hand, several genetic linkage maps have been constructed for other Castanea species. The first map of the American chestnut × Chinese chestnut was constructed using random amplified polymorphic DNA markers (Kubisiak et al., 1997). Subsequently, maps of the European chestnut were constructed using an intraspecific cross (Barreneche et al., 2004; Casasoli et al., 2001, 2006). Recently, a highly informative genetic map of the Chinese chestnut was constructed with 329 SSR markers and 1064 single-nucleotide polymorphism (SNP) markers using an expressed sequence tag (EST) database created by next-generation sequencing (Kubisiak et al., 2013). This consensus map consists of 12 linkage groups, each ranging from 50.6 to 90.4 cM, and encompasses 742.4 cM with an average distance of 0.64 cM between loci. These information sources and markers are effective tools for constructing a Japanese chestnut genetic map and developing molecular markers linked to genes of interest.
A new elite Japanese chestnut, ‘Porotan’, with an easy-peeling pellicle was released by NIFTS in 2006 and it promises to become a leading cultivar (Pereira-Lorenzo et al., 2012; Saito et al., 2009). The easy-peeling pellicle trait is controlled by a single recessive gene, but few Japanese chestnut cultivars carry this gene, and ‘Porotan’ is the only commercial cultivar with an easy-peeling pellicle (Nishio et al., 2013). For these reasons, our recent breeding populations were based on the use of ‘Porotan’ and its relatives as parents to obtain homozygous genotypes at the easy-peeling locus. Developing molecular markers linked to genes of interest from these cultivars and selections is essential to accelerating marker-assisted selection (MAS) for chestnut breeding.
In this study, we used two breeding populations derived from ‘Porotan’ and its relatives for QTL analyses of agronomic traits. The objectives were to construct genetic maps of the Japanese chestnut, to combine maps of different Castanea species, and to obtain molecular markers associated with important traits for chestnut breeding. This is the first study focused on detecting QTLs for agronomic and nut traits in Castanea.
Materials and Methods
Plant materials and extraction of nucleic acids
Two mapping populations, one from a cross between the ‘Kunimi’ and breeding line ‘709-034’ (Kx709, 110 F1 seedlings) and one from a cross between ‘Porotan’ and ‘Tsukuba-43’ (Px43, 99 F1 seedlings), were used to construct genetic maps and perform QTL analyses. ‘709-034’ and ‘Tsukuba-43’ are F1 progeny of ‘Tsukuba’ × ‘Porotan’ and ‘Kunimi’ × ‘Mikuri’, respectively. The pedigrees of these mapping populations are shown in Figure 1. ‘Porotan’ is homozygous recessive at the easy-peeling locus, and ‘Kunimi’, ‘709-034’, and ‘Tsukuba-43’ are heterozygous. MAS for this trait was conducted with the Px43 population, and only nuts with the easy-peeling genotype were planted. On the other hand, we could not conduct MAS for this trait with the Kx709 population because molecular markers associated with the easy-peeling trait had not yet been developed when this population was at the seedling stage.
Genomic DNA was extracted from young leaves or young buds using a DNeasy Plant Mini Kit (Qiagen, Hilden, Germany) following the manufacturer’s instructions.
Evaluation of phenotypic traits
The Kx709 and Px43 populations were sown in the winters of 2010 and 2011, respectively. Phenotypic traits for Kx709 were recorded in 2013 and 2014, and those for Px43 in 2014 and 2015. We evaluated eight important traits for chestnut breeding (HARVEST, WEIGHT, SPLIT, INSECT, number of burs per tree [BUR], number of nuts per tree [NUT], trunk diameter [TRUNK], and yield per tree [YIELD]; Table 1).
From each tree in the two populations, each bur was harvested when it had changed from green to brown and had begun to split open or dropped. After removal of the burs, the number of nuts harvested on a given day was recorded. Burs and nuts were harvested every 3 or 4 days in late August to late September. HARVEST of each nut was expressed as the number of days after August 1, and the average value of HARVEST for each tree was used for QTL analyses. WEIGHT (g) per nut was measured on a digital scale on each harvest date. SPLIT and INSECT were evaluated as the proportions of nuts with a split pericarp and of nuts infested by the peach moth (Conogethes punctiferalis), respectively, out of all nuts harvested from each tree. BUR was counted by visual assessment in early July. NUT was expressed as the total number of nuts harvested in a season. TRUNK (mm) was measured as the trunk diameter of each tree in December. YIELD was expressed as the total nut weight per tree in a season. Before QTL analysis, the WEIGHT values were log10-transformed, SPLIT values were log10-transformed and then converted to absolute values, and INSECT values were square-root-transformed to improve the normality of the distribution of the residual estimates, according to Nishio et al. (2013). The Kolmogorov–Smirnov test was used to check the normality of the data distribution for each trait, with P > 0.05 indicating a normal distribution.
Next-generation sequencing and SNP calling
To develop a set of novel SNP markers, plant materials for the re-sequencing of the Japanese chestnut (‘Ganne’, ‘Ginyose’, ‘Hayadama’, and ‘Obuse 2’) were obtained from trees grown in a field at NIFTS (Tsukuba, Japan). Frozen young leaves were homogenized for 20 s in a Shake Master Auto (Bio Medical Science, Tokyo, Japan). Genomic DNA was extracted with a DNeasy Plant Mini Kit (Qiagen) according to the manufacturer’s instructions, except that Buffer AP1 contained 2% 2-mercaptoethanol.
The size-selected DNA (average length, 500 bp) was purified, and then a standard paired-end sequencing library was built by using a TruSeq DNA Sample Preparation Kit (Illumina Inc. San Diego, CA, USA). Short reads (100 bp) at each end were generated from base calling (Illumina pipeline CASAVA v. 1.8.2).
Low-quality bases and adapter sequences were trimmed, and then reads shorter than 36 bp or unpaired reads were discarded by Trimmomatic v. 0.32 software (Bolger et al., 2014). The preprocessed reads were mapped to the draft sequences of the Chinese chestnut whole genome v. 1.1 (Cm_v1.1_scaffolds.fasta; http://www.hardwoodgenomics.org/chinese-chestnut-genome) in the Burrows–Wheeler Alignment Tool v. 0.7.12 (Li and Durbin, 2009), and then read pairs with high mapping quality (>Q20) that were properly aligned were selected by SAMtools v. 1.2 software (Li, 2011). Local realignment around indels was performed with Genome Analysis Toolkit v. 3.3 software (DePristo et al., 2011). Finally, PCR duplicates were removed by using MarkDuplicates software (Picard tools v. 1.119; http://sourceforge.net/projects/picard/; now available at http://broadinstitute.github.io/picard/). SNP variants among the four Japanese chestnut cultivars and biallelic variants were detected by using the “mpileup” option of SAMtools.
To design reliable SNP markers, we discarded SNPs meeting any of the following criteria: (1) Other SNPs or indels were detected within the 80-bp flanking regions on either side of the SNP. (2) The 80-bp flanking sequences were aligned to two or more regions in the reference genome of the Chinese chestnut by a BLASTN search. (3) Either of the 80-bp flanking regions included a repeat sequence. (4) The read depth was extremely low or high (<13 or >56). Among the selected SNPs, those that were unsuitable for probe design because they had a SNP score under 0.7 calculated by the Illumina Assay Design Tool were discarded. The final 768 SNPs were selected from different scaffolds (Table S1). All sequence data obtained in this study are available from the DDBJ Sequence Read Archive under accession number DRA004605.
SNP genotyping
SNPs were genotyped using the Illumina GoldenGate Genotyping Assay (Illumina Inc.). The scanned data were analyzed by the Genotyping module (v. 1.9.4) of Illumina GenomeStudio v. 2011.1 software to generate genotype data for individuals. Clustering of SNPs was adjusted by eye when necessary. Acceptable SNPs had scores of “GenTrain score” ≥ 0.4, “call freq” ≥ 0.85, “P-P-C errors” ≥ 2, and “minor freq” ≥ 0.01.
SSR genotyping
Previously described Chinese, Japanese, and European chestnut SSRs were used for genotyping (Table S2). PCR amplification was performed in 10 μL containing 5 μL of 2× Green GoTaq reaction buffer (0.4 mM each dNTP, Taq DNA polymerase, and 3 mM MgCl2, pH 8.5; Promega, Madison, WI, USA), 20 pmol of each forward primer labeled with a fluorescent chemical (5-FAM or 5-HEX) and unlabeled reverse primer, and 2.5 ng of genomic DNA. Amplification was performed in 35 cycles of 94°C for 1 min, 55°C for 1 min, and 72°C for 2 min. PCR products were separated and detected with a 3130 xl Genetic Analyzer (Life Technologies Co., Carlsbad, CA, USA). The size of each amplified band was determined by comparison with a set of internal standard DNA fragments (400HD-ROX; Life Technologies Co.) in GeneMapper v. 5.0 software (Life Technologies Co.).
Construction of genetic linkage maps
To construct integrated maps of Kx709 and Px43, we used the cross-pollination (CP) mode of JoinMap v. 4.1 software (Van Ooijen, 2006). All of the markers that showed polymorphism in at least one parent were used (in the JoinMap format for CP mode, this includes marker configurations ab × cd, lm × ll, nn × np, ef × eg, and hk × hk). The markers were grouped with a minimum logarithm of odds (LOD) score of 4.0 and a recombination frequency of 0.45. A regression mapping algorithm was used to build the linkage maps, and map distances were calculated according to the Kosambi mapping function (Kosambi, 1944). The linkage group names were assigned according to the Chinese chestnut genetic linkage maps (Kubisiak et al., 2013). The genetic maps were drawn using MapChart v. 2.2 software (Voorrips, 2002).
JoinMap was used to construct maps of ‘Kunimi’, ‘709-034’, ‘Porotan’, and ‘Tsukuba-43’ with the pseudo-testcross mapping strategy in the BC1 mode (Grattapaglia and Sederoff, 1994). The LOD threshold for mapping was set at 3.0 and the recombination frequency at 0.45. The marker configurations ab × cd, lm × ll, and ef × eg were used for the maternal maps (‘Kunimi’ and ‘Porotan’), and configurations ab × cd, nn × np, and ef × eg were used for the paternal maps (‘709-034’ and ‘Tsukuba-43’). Within each parent, the members of each pair of heterozygous alleles were designated as “A” and “H”. Since the linkage phase of each marker locus (relative to other loci) is unknown in the pseudo-testcross model, the dataset for each parent was duplicated and the allelic designations were reversed, i.e., those previously designated as “A” were designated “H” and vice versa. After the linkage phase of each marker was inferred and the pairs of linkage groups were deduced, one linkage group from each pair was chosen to create the linkage map of that parent.
QTL analysis
MapQTL v. 6.0 software (Van Ooijen, 2009) was used for interval mapping. A genome-wide LOD significance threshold was determined for each trait based on a permutation test with 1000 replications, and then QTLs with an LOD that was significant at P < 0.05 were identified. For HARVEST, WEIGHT, SPLIT, BUR, NUT, TRUNK, and YIELD, we classified only QTLs showing significant LOD scores in both years of evaluation as reliable QTLs. For INSECT, we identified QTLs showing significant LOD scores in only a single year as candidates for meaningful QTLs (2013 for the Kx709 population and 2015 for the Px43 population), because there was almost no insect infestation in the breeding field in 2014.
Results
Phenotypic trait evaluation
The distributions of HARVEST, WEIGHT, BUR, NUT, TRUNK, and YIELD were not significant at P = 0.05 by the one-sample Kolmogorov–Smirnov test, indicating that they were normal (Fig. S1). However, the distributions of INSECT (2014) in both populations and SPLIT (2014 and 2015) in the Px43 population were significant, indicating deviation from normal. As noted above, there was almost no insect infestation in the chestnut breeding field at NIFTS in 2014. Also, both parents and seedlings in the Px43 population showed extremely low rates of pericarp splitting (SPLIT). For these reasons, we omitted these data that showed abnormal distributions from further QTL analyses. The distribution of the residual estimates, except for INSECT (2014) in both populations and SPLIT (2014 and 2015) in the Px43 population, was not significant at P = 0.05 by the one-sample Kolmogorov–Smirnov test.
Genetic linkage maps
We constructed genetic linkage maps of ‘Kunimi’, ‘709-034’, ‘Porotan’, and ‘Tsukuba-43’ and integrated maps of Kx709 and Px43 (Table 2; Fig. S2). In the most saturated of the six maps, that of Kx709, 12 linkage groups were identified that covered 668.1 cM with an average distance of 0.8 cM between loci, with 409 SSRs and 537 SNPs (Table 2; Fig. 2). There were four large genetic intervals, each with a gap of >10 cM, at the bottom of LG-C, -H, -I, -K, and -L.
In the Px43 population, MAS for the easy-peeling pellicle locus was done using the molecular markers PRD52 and PRD58, which were mapped on LG-A in the Kx709 population. As the alleles linked to the easy-peeling gene of ‘Tsukuba-43’ were artificially selected before planting, the calculated genetic distances between the markers around the easy-peeling pellicle locus on LG-A of ‘Tsukuba-43’ and Px43 were inaccurate. Thus, we omitted the groups constructed by markers on LG-A from maps of ‘Tsukuba-43’ and Px43.
The six genetic maps were successfully aligned to the Chinese chestnut genetic map by using SSRs developed by Kubisiak et al. (2013) as anchor markers. In total, 997 molecular markers (281 Japanese chestnut SSRs, 145 Chinese chestnut SSRs, 17 European chestnut SSRs, and 554 SNPs) were mapped on the six genetic maps (Fig. S2; Table S2). The total number of loci differed greatly among the maps, ranging from 226 (‘Tsukuba-43’) to 839 (‘Kunimi’ × ‘709-034’). The total map length ranged from 452.7 cM (‘Tsukuba-43’) to 668.1 cM (‘Kunimi’ × ‘709-034’).
As none of the markers were mapped on LG-I of ‘Porotan’ or LG-L of ‘Tsukuba-43’, maps of those genomic regions were not constructed. Also, only four markers were mapped on LG-C of ‘Tsukuba-43’, and no markers were mapped on the lower side of LG-G of ‘709-034’. On LG-E of Kx709 and LG-H of Px43, few markers were mapped in the parental cultivars (6 in ‘Kunimi’ and 5 in ‘709-034’; 17 in ‘Porotan’ and 16 in ‘Tsukuba-43’), but more markers were mapped on the integrated maps (57 in Kx709 and 68 in Px43). Many markers showed distorted segregation in all six maps (Fig. S2); the proportion ranged from 0.14 (‘Porotan’) to 0.33 (‘Tsukuba-43’). Markers showing distorted segregation were present in all linkage groups, with no clear differences in the frequency among the linkage groups.
QTL analysis of traits
In the Kx709 population analysis, nine significant QTLs were mapped to linkage groups in ‘Kunimi’ and ‘709-034’ (Table 3; Fig. S3). For HARVEST and SPLIT, two significant QTLs were detected on different linkage groups (LG-D and LG-F for HARVEST, LG-D and LG-J for SPLIT). For each of WEIGHT, BUR, NUT, TRUNK, and YIELD, only one significant QTL was detected. The LOD values and percentages of phenotypic variance explained by these QTLs ranged from 2.94 to 6.93 and from 11.9% to 26.0% (qtl-BUR-Ku-L in 2014 and qtl-SPLIT-709-J in 2014, respectively).
In the Px43 population analysis, two significant QTLs for HARVEST were detected on different linkage groups on the map of ‘Tsukuba-43’ (LG-D and LG-H; Table 3; Fig. S4). For INSECT, significant QTLs were detected in approximately the same region of LG-F on maps of both ‘Porotan’ and ‘Tsukuba-43’. For each of WEIGHT and NUT, one significant QTL was detected. We did not detected significant QTLs for SPLIT, BUR, TRUNK, or YIELD in this population.
From the integrated map of Kx709, two QTLs for HARVEST (on LG-A and LG-D) and one QTL each for WEIGHT (LG-A) and SPLIT (LG-J) were detected (Table 4; Fig. S3). From the integrated map of Px43, one significant QTL each for HARVEST (LG-H) and INSECT (LG-F) were identified (Table 4; Fig. S4). The LOD values and phenotypic variance accounted for by QTLs from the integrated maps were higher than those from maps of one parent and ranged from 4.44 to 7.90 and from 19.0% to 29.1%, respectively. Out of the six significant QTLs from the integrated maps, five (qtl-HARVEST-Kx709-D, qtl-WEIGHT-Kx709-A, qtl-SPLIT-Kx709-J, qtl-HARVEST-Px43-H, and qtl-INSECT-Px43-F) were also significant at the same position on the maps of one parent (qtl-HARVEST-Ku-D, qtl-WEIGHT-709-A, qtl-SPLIT-709-J, qtl-HARVEST-T43-H, and qtl-INSECT-T43-F, respectively), meaning that only one (qtl-HARVEST-Kx709-A) was newly detected from the integrated maps (Tables 3 and 4; Figs. S3 and S4).
In total, 21 significant QTLs were detected from the six genetic maps. At least one significant QTL was detected for each of the evaluated traits.
Several significant QTLs for HARVEST and SPLIT were identified in the Kx709 population analysis (qtl-HARVEST-Ku-D, qtl-HARVEST-Ku-F, qtl-HARVEST-Kx709-A, qtl-SPLIT-Ku-D, and qtl-SPLIT-709-J) and the Px43 population analysis (qtl-HARVEST-T43-D and qtl-HARVEST-T43-H). Under the assumption that the effects of these QTLs are additive, the percentage of total phenotypic variance explained by the combination of three markers for HARVEST (qtl-HARVEST-Ku-D, qtl-HARVEST-Ku-F, and qtl-HARVEST-Kx709-A) was 60.8% in 2013 and 47.5% in 2014. The percentage explained by the combination of qtl-SPLIT-Ku-D and qtl-SPLIT-709-J was 33.4% in 2013 and 41.7% in 2014. In the Px43 population, the percentage explained by the combination of qtl-HARVEST-T43-D and qtl-HARVEST-T43-H was 36.7% in 2014 and 29.8% in 2015.
Discussion
Six genetic linkage maps of the Japanese chestnut were constructed using SSRs and SNPs. In the Kx709 population analysis, maps of ‘Kunimi’, ‘709-034’, and Kx709 consisted of 12 linkage groups, which corresponds to the basic chromosome number in Fagaceae. On the other hand, in the Px43 population analysis, each of the parental maps and the integrated map had 10 or 11 linkage groups. Using 145 Chinese chestnut SSRs developed by Kubisiak et al. (2013) as anchor markers, we found that each linkage group in the new maps showed a one-to-one correspondence with one of the groups in the Chinese chestnut consensus map (Fig. S2). The order of anchor markers in each group was very similar in all maps, and few markers mapped to different linkage groups on different maps (Table S2), suggesting that these maps show high colinearity. The most saturated map among the six was the integrated map of Kx709. The total length of the map (668.1 cM) was close to that of the Chinese chestnut consensus map (742.4 cM; Fig. S2). The longest LG among the 12 LGs was LG-A (82.4), as Kubisiak et al. (2013) found. However, there were 19 large gaps of >5 cM, including 5 of >10 cM, most of them at the top and bottom of the LGs. Many more markers will be needed to fill these gaps.
There were several specific regions within a single cultivar or selection in which no markers could be mapped (LG-I of ‘Porotan’, LG-L of ‘Tsukuba-43’, and parts of LG-G of ‘709-034’ and LG-C of ‘Tsukuba-43’; Fig. S2), although maps of the other Japanese chestnut cultivars and the Chinese chestnut consensus map cover these regions. These specific regions can be explained by fixation due to breeding, as excellent genotypes such as ‘Tanzawa’ and ‘Ganne’ have repeatedly been used as parents in the Japanese chestnut breeding program (Fig. 1). Since the pedigree-based inbreeding coefficients were 0.156 for ‘709-034’, 0.125 for ‘Porotan’, and 0.141 for ‘Tsukuba-43’, it is understandable that those cultivars would have homozygous regions. Also, on LG-E of Kx709 and LG-H of Px43, few markers were mapped in each individual cultivar, while an adequate number of markers was mapped on the integrated maps. In these regions, the genotypes of most markers were heterozygous for the same two alleles in both parents, making them uninformative for constructing maps of either parent in that region. These fixed regions represent a loss in genetic diversity, but probably represent an accumulation of favorable alleles in recent cultivars and selections. We did not find obvious inbreeding depression in these populations; however, it is necessary to pay close attention to any further increase in the amount of fixation.
The linkage groups of the Chinese chestnut consensus map were connected to the map of Chinese chestnut × American chestnut (Kubisiak et al., 2013) by using anchor SSRs. High colinearity was confirmed between Japanese and Chinese chestnuts (this study) as well as between Chinese and American chestnuts (Kubisiak et al., 2013), so the genetic maps of Japanese, Chinese, and American chestnuts could be aligned. On the other hand, maps of the European chestnut have not been sufficiently connected to these other maps, because the number of anchor SSRs mapped in European chestnut is limited. The maps of the European chestnut, with 12 linkage groups (named C1 to C12), have been connected to those of Quercus, however (Barreneche et al., 2004; Casasoli et al., 2006). We mapped five SSR markers that were mapped in both European and Japanese chestnut genetic maps (CsCAT13, CsCAT14, CsCAT15, EMC2, and EMC13). On the basis of this mapping, LG-C and LG-C8 (CsCAT15), LG-H and LG-C6 (EMC2 and EMC13), LG-J and LG-C12 (CsCAT13), and LG-K and LG-C2 (CsCAT14) are homologous linkage groups.
Many fruit breeders have estimated environmental variance to improve breeding efficiency (Leon et al., 2016; Nishio et al., 2014; Sato et al., 2000; Thaipong and Boonprakob, 2005; Yamada et al., 1993). Increasing the number of years of evaluation and/or the number of tree replicates improves the evaluation of genetic marker effects. As tree replication requires a large space and an enormous effort, and given the time and cost of maintaining trees over the long juvenile period, increasing the number of years of evaluation is more efficient than increasing the number of tree replicates (Nishio et al., 2014). Thus, we evaluated the traits over 2 years and selected QTLs showing significant LOD scores in both years as reliable QTLs, with the exception that a single year of data was used to select QTLs for INSECT (Tables 3 and 4).
QTLs for HARVEST were identified at four loci (LG-A, -D, -F, and -H) across the two populations. In the middle of LG-D, significant QTLs (qtl-HARVEST-Ku-D and qtl-HARVEST-T43-D) were detected in both the Kx709 and Px43 populations, and their LOD scores and percentages of phenotypic variance explained were similar, suggesting that these QTLs were controlled by the same gene and probably had high reliability and general versatility. Many QTLs for fruit harvesting day have been identified in several fruit tree species (Eduardo et al., 2011; Kenis et al., 2008; Kunihisa et al., 2014; Liebhard et al., 2003; Yamamoto et al., 2014), and broad-sense heritability of fruit harvesting day was higher than that of other fruit traits (Kunihisa et al., 2014; Nishio et al., 2014; Yamada et al., 1993). Thus, both the QTLs for fruit harvesting day identified in earlier studies and the HARVEST QTLs identified here may have been identified owing to their high heritability. On the other hand, although INSECT showed low broad-sense heritability in Japanese chestnut cultivars (Nishio et al., 2014), significant QTLs for INSECT were identified on LG-F in 2015 from the Px43 population (Tables 3 and 4). Because insect infestation was exceptionally rare in the breeding field in 2014, these QTLs should be validated in another year.
In the Kx709 population, the percentage of total phenotypic variance explained by the combination of three markers for HARVEST (qtl-HARVEST-Ku-D, qtl-HARVEST-Ku-F, and qtl-HARVEST-Kx709-A) was 60.8% in 2013 and 47.5% in 2014, under the assumption that the effects of these QTLs are additive. The percentage explained by the combination of qtl-SPLIT-Ku-D and qtl-SPLIT-709-J was 33.4% in 2013 and 41.7% in 2014. In the Px43 population, the percentage explained by the combination of qtl-HARVEST-T43-D and qtl-HARVEST-T43-H was 36.7% in 2014 and 29.8% in 2015. These values are high enough to improve chestnut breeding efficiency by MAS.
One of the urgent objectives of our breeding program is to release cultivars having an easily peeled pellicle and a different nut harvesting day from ‘Porotan’, which is the only commercial cultivar with this trait. ‘Porotan’ is an early-ripening cultivar; thus, we need to develop mid- to late-ripening cultivars with an easily peeled pellicle. We could probably obtain mid-ripening genotypes using the markers developed in this study because the four parent cultivars were early- to mid-ripening cultivars. To develop late-ripening cultivars with an easy-peeling pellicle, however, we need to introduce late-ripening cultivars into our breeding program. Genome-wide association studies based on a diverse cultivar collection that includes late-ripening genotypes would enable us to identify additional QTLs associated with HARVEST.
The molecular markers PRD52 and PRD58, which are linked to the easy-peeling locus (Nishio et al., 2013), were mapped on LG-A in the Kx709 population. Because we also detected QTLs for HARVEST and WEIGHT on LG-A, MAS for the easy-peeling trait using these markers would reduce some of the genetic variation for HARVEST and WEIGHT in the Kx709 population. Moreover, individuals showing the easy-peeling genotype in this population have earlier ripening and a lower nut weight than individuals showing the difficult-peeling genotype. However, the QTL peaks for HARVEST and WEIGHT were located about 15 cM from the easy-peeling locus, so it should be possible to identify recombinants by using molecular markers associated with each of the traits.
In this study, we constructed six genetic linkage maps using a leading Japanese chestnut cultivar, ‘Porotan’, and its relatives (‘Kunimi’, ‘709-034’, and ‘Tsukuba-43’), which are essential materials to obtain easy-peeling genotypes in the NIFTS Japanese chestnut breeding program. We identified 21 significant QTLs from the two breeding populations. In particular, several significant QTLs were identified for HARVEST and SPLIT. The percentage of total phenotypic variance explained by the combination of QTLs was high for both HARVEST (47.5%–60.8%) and SPLIT (33.4%–41.7%). Because ‘Porotan’ and its relatives were used as parents, the detected QTLs can be easily applied to Japanese chestnut breeding. We believe that MAS using these QTLs will greatly improve Japanese chestnut breeding efficiency.
Acknowledgments
We are deeply indebted to all the people involved in the Japanese chestnut breeding program at the Institute of Fruit Tree and Tea Science, NARO, Japan.
Literature Cited
- Barreneche, T., M. Casasoli, K. Russell, A. Akkak, H. Meddour, C. Plomion, F. Villani and A. Kremer. 2004. Comparative mapping between Quercus and Castanea using simple-sequence repeats (SSRs). Theor. Appl. Genet. 108: 558–566.
- Bolger, A. M., M. Lohse and B. Usadel. 2014. Trimmomatic: a flexible trimmer for Illumina sequence data. Bioinformatics 30: 2114–2120.
- Casasoli, M., J. Derory, C. Morera-Dutrey, O. Brendel, I. Porth, J. M. Guehl, F. Villani and A. Kremer. 2006. Comparison of quantitative trait loci for adaptive traits between oak and chestnut based on an expressed sequence tag consensus map. Genetics 172: 533–546.
- Casasoli, M., C. Mattioni, M. Cherubini and F. Villani. 2001. A genetic linkage map of European chestnut (Castanea sativa Mill.) based on RAPD, ISSR and isozyme markers. Theor. Appl. Genet. 102: 1190–1199.
- DePristo, M. A., E. Banks, R. Poplin, K. V. Garimella, J. R. Maguire, C. Hartl, A. A. Philippakis, G. del Angel, M. A. Rivas, M. Hanna, A. McKenna, T. J. Fennell, A. M. Kernytsky, A. Y. Sivachenko, K. Cibulskis, S. B. Gabriel, D. Altshuler and M. J. Daly. 2011. A framework for variation discovery and genotyping using next-generation DNA sequencing data. Nature Genet. 43: 491–498.
- Eduardo, I., I. Pacheco, G. Chietera, D. Bassi, C. Pozzi, A. Vecchietti and L. Rossini. 2011. QTL analysis of fruit quality traits in two peach intraspecific populations and importance of maturity date pleiotropic effect. Tree Genet. Genomes 7: 323–335.
- Grattapaglia, D. and R. Sederoff. 1994. Genetic linkage maps of Eucalyptus grandis and Eucalyptus urophylla using a pseudo-testcross: mapping strategy and RAPD markers. Genetics 137: 1121–1137.
- Kenis, K., J. Keulemans and M. W. Davey. 2008. Identification and stability of QTLs for fruit quality traits in apple. Tree Genet. Genomes 4: 647–661.
- Kosambi, D. D. 1944. The estimation of map distances from recombination values. Ann. Eugen. 12: 172–175.
- Kotobuki, K., T. Saito, T. Kashimura and M. Shoda. 1999. Chestnut breeding program in National Institute of Fruit Tree Science, Japan. Acta Hortic. 494: 323–325.
- Kubisiak, T. L., F. V. Hebard, C. D. Nelson, J. Zhang, R. Bernatzky, H. Huang, S. L. Anagnostakis and R. L. Doudrick. 1997. Molecular mapping of resistance to blight in an interspecific cross in the genus Castanea. Phytopathology 87: 751–759.
- Kubisiak, T. L., C. D. Nelson, M. E. Staton, T. Zhebentyayeva, C. Smith, B. A. Olukolu, G. C. Fang, F. V. Hebard, S. Anagnostakis, N. Wheeler, P. H. Sisco, A. G. Abbott and R. R. Sederoff. 2013. A transcriptome-based genetic map of Chinese chestnut (Castanea mollissima) and identification of regions of segmental homology with peach (Prunus persica). Tree Genet. Genomes 9: 557–571.
- Kunihisa, M., S. Moriya, K. Abe, K. Okada, T. Haji, T. Hayashi, H. Kim, C. Nishitani, S. Terakami and T. Yamamoto. 2014. Identification of QTLs for fruit quality traits in Japanese apples: QTLs for early ripening are tightly related to preharvest fruit drop. Breed. Sci. 64: 240–251.
- Leon, L., R. Arias-Calderon, R. de la Rosa, B. Khadari and E. Costes. 2016. Optimal spatial and temporal replications for reducing environmental variation for oil content components and fruit morphology traits in olive breeding. Euphytica 207: 675–684.
- Li, H. 2011. A statistical framework for SNP calling, mutation discovery, association mapping and population genetical parameter estimation from sequencing data. Bioinformatics 27: 2987–2993.
- Li, H. and R. Durbin. 2009. Fast and accurate short read alignment with Burrows-Wheeler transform. Bioinformatics 25: 1754–1760.
- Liebhard, R., M. Kellerhals, W. Pfammatter, M. Jertmini and C. Gessler. 2003. Mapping quantitative physiological traits in apple (Malus × domestica Borkh.). Plant Mol. Biol. 52: 511–526.
- Motoki, Y. 2007. Association between chestnut and human in Japanese archipelago. The Quarterly Report of Economics 1/2: 105–125 (In Japanese).
- Nishio, S., N. Takada, T. Yamamoto, S. Terakami, T. Hayashi, Y. Sawamura and T. Saito. 2013. Mapping and pedigree analysis of the gene that controls the easy peel pellicle trait in Japanese chestnut (Castanea crenata Sieb. et Zucc.). Tree Genet. Genomes 9: 723–730.
- Nishio, S., M. Yamada, N. Takada, H. Kato, N. Onoue, Y. Sawamura and T. Saito. 2014. Environmental variance and broad-sense heritability of nut traits in Japanese chestnut breeding. HortScience 49: 696–700.
- Nishio, S., T. Yamamoto, S. Terakami, Y. Sawamura, N. Takada, C. Nishitani and T. Saito. 2011. Novel genomic and EST-derived SSR markers in Japanese chestnuts. Sci. Hortic. 130: 838–846.
- Pereira-Lorenzo, S., A. Ballester, E. Corredoira, A. M. Vieitez, S. Agnanostakis, R. Costa, G. Bounous, R. Botta, G. L. Beccaro and T. L. Kubisiak. 2012. Chestnut. p. 729–769. In: M. L. Badenes and D. H. Byrne (eds.). Fruit Breeding. Springer, New York.
- Saito, T., K. Kotobuki, Y. Sawamura, K. Abe, O. Terai, M. Shoda, N. Takada, Y. Sato, T. Hirabayashi, A. Sato, T. Nishibata, Y. Kashimura, T. Kozono, H. Fukuda, K. Kihara, K. Suzuki and M. Uchida. 2009. New Japanese chestnut cultivar ‘Porotan’. Bull. Natl. Inst. Fruit Tree Sci. 9: 1–9 (In Japanese with English abstract).
- Sato, A., M. Yamada, H. Iwanami and N. Hirakawa. 2000. Optimal spatial and temporal measurement repetition for reducing environmental variation of berry traits in grape breeding. Sci. Hortic. 85: 75–83.
- Sawamura, Y. 2006. Chestnut. p. 82–85. In: Japanese. Soc. Hort. Sci. (ed.). Horticulture in Japan 2006. Shoukadoh, Kyoto.
- Thaipong, K. and U. Boonprakob. 2005. Genetic and environmental variance components in guava fruit qualities. Sci. Hortic. 104: 37–47.
- Van Ooijen, J. W. 2006. JoinMap® 4.0. Software for the calculation of genetic linkage maps in experimental populations. Kyazma BV, Wageningen.
- Van Ooijen, J. W. 2009. MapQTL® 6.0. Software for the mapping of quantitative trait in experiment populations of diploid species. Kyazma BV, Wageningen.
- Voorrips, R. E. 2002. MapChart: software for the graphical presentation of linkage maps and QTLs. J. Hered. 93: 77–78.
- Wheeler, N. and R. Sederoff. 2009. Role of genomics in the potential restoration of the American chestnut. Tree Genet. Genomes 5: 181–187.
- Yamada, M., H. Yamane, K. Yoshinaga and Y. Ukai. 1993. Optimal spatial and temporal measurement repetition for selection in Japanese persimmon breeding. HortScience 28: 838–841.
- Yamamoto, T., S. Terakami, N. Takada, S. Nishio, N. Onoue, C. Nishitani, M. Kunihisa, E. Inoue, H. Iwata, T. Hayashi, A. Itai and T. Saito. 2014. Identification of QTLs controlling harvest time and fruit skin color in Japanese pear (Pyrus pyrifolia Nakai). Breed. Sci. 64: 351–361.
