Development of SSR markers and identification of major quantitative trait loci controlling shelling percentage in cultivated peanut (Arachis hypogaea L.)

Key message A total of 204,439 SSR markers were developed in diploid genomes, and 25 QTLs for shelling percentage were identified in a RIL population across 4 years including five consistent QTLs. Abstract Cultivated peanut (Arachis hypogaea L.) is an important grain legume providing edible oil and protein for human nutrition. Genome sequences of its diploid ancestors, Arachis duranensis and A. ipaensis, were reported, but their SSRs have not been well exploited and utilized hitherto. Shelling percentage is an important economic trait and its improvement has been one of the major objectives in peanut breeding programs. In this study, the genome sequences of A. duranensis and A. ipaensis were used to develop SSR markers, and a mapping population (Yuanza 9102 × Xuzhou 68-4) with 195 recombinant inbred lines was used to map QTLs controlling shelling percentage. The numbers of newly developed SSR markers were 84,383 and 120,056 in the A. duranensis and A. ipaensis genomes, respectively. Genotyping of the mapping population was conducted with both newly developed and previously reported markers. QTL analysis using the phenotyping data generated in Wuhan across four consecutive years and genotyping data of 830 mapped loci identified 25 QTLs with 4.46–17.01% of phenotypic variance explained in the four environments. Meta-analysis revealed five consistent QTLs that could be detected in at least two environments. Notably, the consistent QTL cqSPA09 was detected in all four environments and explained 10.47–17.01% of the phenotypic variance. The segregation in the progeny of a residual heterozygous line confirmed that the cpSPA09 locus had additive effect in increasing shelling percentage. These consistent and major QTL regions provide opportunity not only for further gene discovery, but also for the development of functional markers for breeding. Electronic supplementary material The online version of this article (doi:10.1007/s00122-017-2915-3) contains supplementary material, which is available to authorized users.


Introduction
Cultivated peanut (Arachis hypogaea L.), also known as groundnut, is an allotetraploid (AABB, 2n = 4x = 40) grain legume native to South America, but now grown in diverse environments in six continents between latitudes 40°N and 40°S (Sharma and Bhatnagar-Mathur 2006). It provides edible oil and protein for human nutrition. In 2014, the annual production of peanut (pods without shelling) was around 42.32 million tones throughout the world (FAOSTAT 2014). Cultivated peanut was formed through the natural hybridization of its two diploid ancestors, A. duranensis (AA, 2n = 2x = 20) and A. ipaensis (BB,2n = 2x = 20). Because the assembly of chromosomal pseudomolecules of cultivated peanut is very challenging, the genome sequences of its diploid ancestors were reported recently, providing a foundation in understanding the genome of cultivated peanut (Bertioli et al. 2016).
Shelling percentage (SP) is an important economic trait in peanut production. Peanut pod has two parts: kernel and hull (Fig. 1a). Kernels (seeds) contain rich edible oil, 1 3 proteins, amino acids, and vitamin E, and are consumed worldwide as edible nut, peanut butter, or candy, and peanut oil extracted from the seeds (Bertioli et al. 2016;Ozudogru et al. 2013). The shelling of peanut hull is the first step needed to transform the peanut materials into a product (Guzel et al. 2005). Shelling percentage (weight of kernels/weight of pods) significantly varied among peanut varieties. For example, Jiang et al. (2013) reported that the shelling percentages of the Chinese core collection of peanut (574 accessions) ranged from 59.9 to 81.0%. Therefore, there is a great potential in the genetic improvement of shelling percentage in peanut breeding.
Quantitative trait locus (QTL) mapping has been widely conducted to identify the genomic regions associated with 1 3 economically important traits. Molecular markers tightly linked to QTLs can be developed and further deployed in marker-assisted breeding (Janila et al. 2016;Sukruth et al. 2015;Varshney et al. 2014). Because of abundance, easy to use, and highly polymorphic, SSR markers were developed (Cuc et al. 2008;Ferguson et al. 2004;Gimenes et al. 2007;Guo et al. 2009;He et al. 2003;Hopkins et al. 1999;Moretzsohn et al. 2005;Shirasawa et al. 2012;Zhou et al. 2016) and widely used in the QTL mapping of disease resistance (Leal-Bertioli et al. 2015;Shoha et al. 2013), drought tolerance (Gautami et al. 2012;Ravi et al. 2011), quality traits (Mondal et al. 2015;Pandey et al. 2014), agronomic and yield traits (Faye et al. 2015;Huang et al. 2015) in cultivated peanut. However, SSR markers have not been screened on whole-genome level in the Arachis genus. Although genome sequences of A. duranensis and A. ipaensis were reported, their SSRs have not been well exploited and utilized hitherto.
Limited efforts were made in identifying QTLs controlling shelling percentage in cultivated peanut. Faye et al. (2015) detected two QTLs for shelling percentage with 5.74-6.97% phenotypic variant explained (PVE) under water stress condition in a RIL population. Huang et al. (2015) identified three QTLs for shelling percentage in an F 2:3 population, including qSPA5 (6.08% PVE), qSPA7 (11.78% PVE), qSPA9 (2.00% PVE). These QTLs were detected only in single environment. Jiang et al. (2014) reported five, two and two significant associated alleles for shelling percentage in three field trials, respectively, through association analysis in Chinese peanut mini-core collection of 298 accessions. Only one SSR allele, 9B4-260, was associated with shelling percentage in all three trials (1.49-2.98% PVE) (Jiang et al. 2014). No major QTLs were reported to be consistently expressed so far. Therefore, it is necessary to identify major and consistent QTLs of shelling percentage in order to accelerate the process of genetic improvement in peanut breeding programs.
In this study, SSRs markers were developed in the A. duranensis and A. ipaensis genomes, and a mapping population (Yuanza 9102 × Xuzhou 68-4) with 195 recombinant inbred lines (RILs) was used to map QTLs controlling shelling percentage in Wuhan, China, in four consecutive years. The peanut cultivar Yuanza 9102 showed significantly higher shelling percentage than Xuzhou 68-4 in previous screening of the Chinese peanut core collection.

Plant materials and phenotyping
A mapping population comprising 195 recombinant inbred lines (RILs) was developed by crossing peanut cultivar Yuanza 9102 and Xuzhou 68-4 and advanced to the F 5 generation by single seed descent method (Luo et al. 2017). The female parent, Yuanza 9102, belongs to A. hypogaea subsp. hypogaea var. vulgaris and is derived from interspecific hybridization between the cultivated peanut Bai-sha1016 and wild species A. diogoi. The male parent, Xuzhou 68-4, belongs to A. hypogaea subsp. hypogaea var. hypogaea and has larger pods but significantly lower shelling percentage than the female parent, Yuanza 9102 (Fig. 1a). The RIL population was used as mapping population to validate the quality of newly developed SSR markers, to construct a dense genetic map and to conduct QTL analysis for shelling percentage in this study. Generations F 5 -F 8 of the RIL population were used in the present study for generating phenotyping data followed by QTL analysis.
The RIL population and the two parents were planted in the experimental field in OCRI-CAAS, Wuhan, China, in four consecutive years from 2013 to 2016. These experiments were treated as four environments and designated as Wuhan2013, Wuhan2014, Wuhan2015 and Wuhan2016 in this study. In each environment, the 195 RILs and the two parents were planted in a randomized complete block design with three replications. Each plot contained one row, with 12 plants in each row, 20 cm between plants and 30 cm between rows. Field management followed the standard agricultural practices. Eight representative plants in the middle of each row were harvested to investigate shelling percentage (SP = Weight of kernels Weight of pods × 100%), according to previously described standard procedures (Huang et al. 2015;Jiang et al. 2006).
Statistical analysis for the phenotypic data of shelling percentage was conducted using IBM SPSS Statistics Version 22 software. Treating the year as a single environment, the univariate variance analyses were performed with standard GLM method and variance components were estimated by restricted maximum likelihood (REML) method. The broad-sense heritability across the four environment trials was calculated based on the estimated variance components with the following formula: H 2 = σ 2 g /(σ 2 g + σ 2 g×e + σ 2 e ) based on plot mean and H 2 = σ 2 g /(σ 2 g + σ 2 g×e /r + σ 2 e /rn) based on entry mean, where σ 2 g is the genotypic variance component among RILs, σ 2 g×e is the RILs × environment interaction variance component, σ 2 e is the residual (error) variance component, and r is the number of environment trials, n is the number of replications in each field trials (Holl and Nyquist 2010).

Development of SSR marker in the genome sequences of A. duranensis and A. ipaensis
Genome sequences of A. duranensis and A. ipaensis were downloaded from the PeanutBase (Bertioli et al. 2016). 1 3 SSR motifs were identified using the MISA script (Thiel 2014). For normal microsatellites, a minimum of 10, 6, 5, 5, 5 and 5 repeats were required for detecting mono-, di-, tri-, tetra-, penta-and hexa-nucleotide motifs, respectively. Compound microsatellites were interrupted by less than 100 base pairs. Primer3 software (https://sourceforge.net/ projects/primer3/) was used to design SSR markers with the following parameters: minimum, maximum, and optimal sizes were 18, 27, and 20 nt, respectively; minimum and maximum GC content were 20 and 80%, respectively; minimum, maximum, and optimal Tm were 57, 63, and 60 °C, respectively; and product size range was from 100 to 300 bp. These markers were referred as newly developed markers in this context and designated with an initial letter 'Ad' and 'Ai' for A. duranensis and A. ipaensis, respectively, followed by the chromosome number, the corresponding subgenome character and an identifier.

Genotyping of mapping population and construction of genetic map
A total of 2240 newly developed markers (Table S3) as well as 7200 previously reported markers Cuc et al. 2008;Ferguson et al. 2004;Gimenes et al. 2007;Guo et al. 2009Guo et al. , 2012He et al. 2003;Hopkins et al. 1999;Hoshino et al. 2006;Huang et al. 2016b;Koilkonda et al. 2012;Leal-Bertioli et al. 2009;Macedo et al. 2012;Moretzsohn et al. 2005Moretzsohn et al. , 2009Moretzsohn Mde et al. 2004;Nagy et al. 2010;Naito et al. 2008;Shirasawa et al. 2012;Wang et al. 2012a;Zhou et al. 2016) were used to screen polymorphism between the two parental genotypes. Polymorphic markers were used to genotype individual RILs. Based on known genomic positions, the 2240 newly developed markers were selected to validate the quality of newly developed SSR markers and to improve the quality of previous genetic map (Luo et al. 2017) for the identification of QTLs controlling shelling percentage. The polymorphism of the 2240 newly developed markers was compared to that of the 7200 previously reported markers. Genomic DNA was extracted from young leaves collected from RILs in F 5 generations using a modified CTAB method (Doyle 1990). PCR amplification was conducted as described in Luo et al. (2017). The PCR products were separated on a 6% polyacrylamide gel and visualized by silver staining (Fountain et al. 2011).
Pearson's Chi-square test was used to assess the goodness of fit to the expected segregation ratio 15:2:15 for codominant marker or 17:15 for dominant marker (P < 0.001). A genetic linkage map was constructed using the JoinMap 4.0 software (Van Ooijen 2006). The recombination ratio was converted to map distance using the Kosambi function (Kosambi 2011). The graphical presentation of genetic linkage map was generated with the MapChart 2.3 software (Voorrips 2002). The linkage groups (LGs) were designated as A1-A10 and B1-B10 by aligning the markers to the integrated consensus genetic map (Shirasawa et al. 2013) and the genome sequences of A. duranensis and Arachis ipaensis (Bertioli et al. 2016). This consensus genetic map was integrated based on 16 genetic maps and used as reference in other publications Huang et al. 2015;Zhou et al. 2014).

QTL and meta-analyses
Genome-wide QTL mapping was performed using the mean value of shelling percentage in each environment. The QTLs were scanned with the Windows QTL Cartographer 2.5 software (Wang et al. 2012b) through composite interval mapping (CIM). The threshold of LOD for declaring the presence of a QTL was determined by 1000 permutation tests at P < 0.05. When separated by a minimum distance of 20 cM, two peaks on one chromosome were considered as two different QTLs (Ravi et al. 2011). Otherwise, the higher peak was chosen to more closely approximate the position of the QTL. QTLs are designated with an initial letter 'q' followed by the abbreviation of trait name (SP), and the corresponding linkage group, similar to the previously described nomenclature (Udall et al. 2006). After the linkage group, the codes 1, 2, 3 and 4 were added for QTLs detected in 2013, 2014 2015 and 2016, respectively. Alphabetical letters were added if two or more QTLs were identified in the same linkage group in the same year. For example, if two QTLs for shelling percentage were detected on chromosome A09 in 2013, they were names as qSPA09.1a and qSPA09.1b, respectively. In addition, QTLs with more than 10% PVE were considered as major QTLs while other QTLs were considered as minor QTLs. If QTLs detected in different environments had overlapping 2-LOD support intervals, they were considered to be a consistent QTL and subjected to meta-analysis to estimate its position using the BioMercator software (Sosnowski et al. 2012). Consistent QTLs were designated with initial letters 'cq'.

Phenotypic variation of shelling percentage
Phenotypic evaluation of shelling percentage of two parental genotypes and RILs showed significant variation across four environments, i.e., Wuhan2013, Wuhan2014, Wuhan2015 and Wuhan2016 ( The values of broad-sense heritability for shelling percentage was estimated to be 0.7769 based on plot mean and 0.9520 based on entry mean, indicating strong control by genetic factors. Variance analysis across the four trials also revealed that genetic, environmental effects and genotype × environment interactions significantly influenced shelling percentage ( Table 2).

The abundance of SSRs in the genomes of Arachis duranensis and A. ipaensis
The availability of the pseudochromosomes of A. duranensis and A. ipaensis, the diploid ancestors of cultivated peanut, provides physical maps for genetic studies in the Arachis genus. A total of 264,135 and 392,107 SSR loci were identified by searching through the genome sequences of A. duranensis and A. ipaensis, respectively, with the MISA script. The average intervals of SSR loci were estimated as 4.10 and 3.45 kb in A. duranensis and A. ipaensis, respectively, indicating the high abundance of SRRs in their genomes. Of the repeat motifs observed, the mononucleotide motif was the most abundant, followed by di-, tri-, tetra-, penta-and hexa-nucleotide motifs ( Table 3). The number of SSRs presented in compound formation in the A. duranensis and A. ipaensis genomes were 22,125 and 37,381, respectively.

Development and validation of SSR markers
SSR markers were designed for di-, tri-, tetra-, penta-and hexa-nucleotide motifs as well as compound microsatellites using the Primer3 software. A total of 84,383 and 120,056 SSR markers were finally developed in the A. duranensis and A. ipaensis genome, respectively (Tables 3, S2). There were 15,251 and 25,822 SSR markers with motifs in compound formation in the A. duranensis and A. ipaensis genome, respectively, while the remaining markers amplified single motif. In the A. duranensis genome, SSR makers with di-, tri-, tetra-, penta-and hexa-nucleotide motifs accounted for 40.87, 34.47, 4.46, 1.63 and 0.49%,   (Table 3). Similarly, SSR makers with di-, tri-, tetra-, penta-and hexa-nucleotide motifs in the A. ipaensis genome accounted for 45.14, 26.86, 4.37, 1.68 and 0.44%, respectively (Table 3). In order to validate the quality of the newly developed SSR markers, 2240 primer pairs were synthesized to screen polymorphic markers in both the parental genotypes and the RIL population. Among the 2240 newly developed markers, 365 markers amplified polymorphic bands while 1706 markers amplified same bands between the two parents (Table S2). A total of 180 newly developed SSR markers (accounted for 8.04%) amplified 185 polymorphic loci in the RIL population. In comparison, 682 previously reported markers amplified 693 polymorphic loci, which accounted for 9.47% of the 7200 previously reported markers screened.

Construction of genetic map
Polymorphic loci of both the 180 newly developed and the 682 previously reported markers were used to construct genetic linkage map with the JoinMap 4.0 software. Among the 862 polymorphic markers (Table S4), one maker (AHGS0729) amplified three genetic loci and 14 markers amplified two genetic loci, while the remaining 847 markers amplified single locus. Among these 878 genetic loci, 784 loci were co-dominant and 94 loci were dominant. Finally, a genetic linkage map containing the 830 loci was constructed spanning 1386.19 cM with an average inter-marker distance of 1.67 cM (Table 4; Fig. 2). The 830 loci were assigned to 20 LGs designated as A01-A10 for A subgenome and B01-B10 for B subgenome by aligning the markers to the integrated consensus genetic map (Shirasawa et al. 2013) and the genome sequences of A. duranensis and A. ipaensis (Bertioli et al. 2016). There were 371 loci for the A subgenome and 459 loci for the B subgenome with the map length of 588.48 and 797.71 cM, respectively. The length of LGs varied from 13.78 cM (A10) to 125.04 cM (B04) and the number of mapped loci ranged from 3 to 110 markers (Table 4; Fig. 2). The Chi-square analysis identified 258 loci (31.09%) that significantly deviated from expected  (Tables 4, S5). The skewed loci on LG A05 favored the female parent "Yuanza 9102" allele, while LG A01, A06, A07, A09, B01, B02, B04, B05 and B10 contained loci favoring the male parent "Xuzhou 68-4" allele. In addition, no more than five skewed loci were mapped on LG A02, A03, B03, B08 and B09. The most significant segregation distortion was observed on LG B05 whose percentage of skewed loci was 93.18% (Table 4).

Detection of QTLs for shelling percentage
Genome-wide QTL analysis was performed using the genetic map and phenotypic data of shelling To further dissect the QTLs controlling shelling percentage, meta-analysis was conducted to integrate QTLs detected in multiple environments, whose confidence intervals were overlapped, into five consistent QTLs (  Fig. 3b). The consistent QTLs cqSPB04 was detected in 2013 and 2016 (4.68-5.32% PVE), while cqSPB05 was detected in 2013 and 2015 (4.82-5.59% PVE).
In order to validate the consistent QTL cpSPA09, flanking marker Ad09A7577 and Ad91I24 (Fig. 3b) were used to screen the RIL population in F 7 generation grown in Wuhan in 2015, and a plant named RIL 15-71126 was found to be heterozygous at the cpSPA09 locus. Seeds of RIL 15-71126 were grown in Wuhan in 2016 and shelling percentages of 31 plants were measured. Among these residual heterozygous lines, five plants with homologous alleles from Yuanza 9102 had an average shelling percentage of 78.45%, while nine plants with homologous alleles from Xuzhou 68-4 had an average shelling percentage of 72.82% (Fig. 3c; Table S7). The shelling percentages of remaining 17 heterozygous plants averaged at 75.62%. Chisquare test revealed that the segregation of the cqSPA09 locus among the 31 plants fitted the expected 1:2:1 segregation ratio (P = 0.52). Variance analysis and multiple comparisons revealed significant differences (P < 0.05) between each pair of the three genotypic groups ( Fig. 3c; Table S7), which were congruent with the finding that the allele came from the parent Yuanza 9102 at the cpSPA09 locus had additive effect in increasing shelling percentage.

Discussion
SSRs were found to be abundant and dispersed throughout the A. duranensis and A. ipaensis genomes in this study. The recently completed genome sequences of the diploid ancestors of cultivated peanut provide physical maps of the highest resolution (Bertioli et al. 2016). A total of 264,135 and 392,107 SSR loci were identified from them, respectively. This number is much larger than the 375,180 loci found in another genome-derived SSR identification by genomic survey sequencing of the cultivated peanut Zhonghua 16 (Zhou et al. 2016) and other previous reports (Cuc et al. 2008;Ferguson et al. 2004;Gimenes et al. 2007;Guo et al. 2009;He et al. 2003;Hopkins et al. 1999;Huang et al. 2016b;Moretzsohn et al. 2005;Shirasawa et al. 2012). The average intervals of SSR loci were estimated as 4.10 and 3.45 kb in A. duranensis and A. ipaensis, respectively, which are neither the highest nor the lowest in plant (Shi et al. 2014;Yu et al. 2016). The mono-, di-and tri-nucleotide SSR motifs were more abundant than tetra-, penta-, and hexa-nucleotide motifs. Moreover, we found that the A/T (99.43%), AT/AT (48.97%) and AAT/ ATT (41.32%) repeats were the most abundant mono-, diand tri-nucleotide SSRs, respectively, in both genomes, which were congruent with previous reports in peanut (Zhou et al. 2016) or other species such as Brassica, rice and Arabidopsis (Katti et al. 2001;Shi et al. 2014;Temnykh et al. 2001).
A total of 84,383 and 120,056 SSR markers were finally developed in the A. duranensis and A. ipaensis genomes, respectively. Among the 2240 newly developed markers used in the genotyping of the RIL population in this study, 1706 markers amplified same bands and 365 markers amplified polymorphism bands in the two parents (Table  S2). The quality of the newly developed markers was similar to recent reported SSR markers (Huang et al. 2016b;Zhou et al. 2016). Subsequently, around 8.04% of the 2240 newly developed markers were polymorphic in the RIL population, which was similar to percentage of the previously reported markers (9.47%). These results verified the validity and reliability of the newly developed SSR markers. Compared to the previous genetic map of the same RIL population (Luo et al. 2017), the mapped loci of the genetic map constructed in this study were improved from 743 to 830, and more importantly, the designated chromosomes were improved from 16 to 20. With known genomic positions, the newly developed SSR markers could be easily selected to improve the quality of genetic map by filling uncovered chromosomes or increasing the densities of covered chromosomes. Note that some of the newly developed markers were not mapped back to their original chromosomes in the constructed linkage map (Table S6), owing to the fact that there might be some segment exchanges among peanut chromosomes (Huang et al. 2016a). Collectively, SSR markers identified in this study should be useful in a variety of applications, such as studying of population structures, genetic map construction and mapping genes for important traits.
The broad-sense heritability estimated in this study was relatively high for shelling percentage in cultivated peanut, indicating that genetic factors play a major role in the determination of this trait. In this study, a RIL population was used to construct a dense genetic linkage map and conducting QTL analysis for shelling percentage. Because of a lack of polymorphism at the DNA level, the first SSR-based genetic linkage map for peanut only had 135 SSR loci (Varshney et al. 2009). However, a genetic linkage map containing 830 loci and covering a total length of 1386.19 cM with an average inter-marker distance of 1.67 cM was constructed in this study. The loci number and density of our map were relatively higher than that of previous reports Huang et al. 2015;Qin et al. 2012;Ravi et al. 2011), except for the integrated consensus map (Shirasawa et al. 2013) and a recent report (Huang et al. 2016a), indicated a high quality of the linkage map constructed in this study. A total of 25 QTLs with 4.46-17.01% PVE were identified to be associated with shelling percentage across four environments. The LOD values of these QTLs ranged from 3.3 to 11.1 and were higher than

(c)
the threshold of LOD for declaring the presence of a QTL that was determined by 1000 permutation tests. The 25 QTLs were mapped on chromosomes A05, A09, B02, B03, B04, B05 and B10. Because no QTL for shelling percentage had been reported on subgenome B, the 17 QTLs on chromosomes B02, B03, B04, B05 and B10 were novel. The chromosomes A05 and A09 that might harbor important genes for shelling percentage as two QTLs from this study and one QTL from an earlier study (Huang et al. 2015) were mapped on A05, and six QTLs from this study and one QTL from a previous study (Huang et al. 2015) were identified on A09. Collectively, shelling percentage was controlled by many QTLs on multiple chromosomes and environments often affected their effects. All the linked markers after validation can be deployed in marker-assisted selection (MAS) for the improvement of shelling percentage in peanut breeding.
Although parents had merely 5% difference in their trait means, this did not limit segregation variance and the power to detect QTLs in the RIL population. Kalih et al. (2014) successfully identified QTLs for plant height, heading stage, and Fusarium head blight in triticale even if parents had similar trait means. Miedaner et al. (2012) also identified QTLs for several agronomic traits in a segregating population whose parents did not differ much. In this study, significant variances and transgressive segregations were observed in the RIL population (Table 1; Fig. 1), indicating that the parents carry complementary alleles at several loci that were newly combined in the progeny (Tanksley 1993 This is in accordance with the fundamental rule of quantitative genetics for complex traits. The QTLs on chromosomes A05, A09, B02, B03 and B04 and two QTLs on chromosome B05 (qSPB05.1a and qSPB05.4) had negative additive genetic effects (Table 5), which revealed that maternal parent Yuanza 9102 as the source of alleles improving the shelling percentage. However, the QTL qSPB05.3 and those on chromosome B10 had positive additive genetic effects ( Table 5), suggesting that the alleles for increasing shelling percentage came from the parent Xuzhou 68-4. These QTLs explained the transgressive segregation of shelling percentage in the RIL population (Fig. 1).
Because of the fact that the identification of QTLs for shelling percentage was highly affected by environment, it is very important to assess their consistent performance across varied environments. Faye et al. (2015) detected two QTLs for shelling percentage with 5.74-6.97% PVE in a RIL population only under water stress condition. Huang et al. (2015) detected three QTLs for shelling percentage with 2.00-11.78% PVE in an F 2:3 population in single environment. None of them were reported to be consistently expressed so far. Despite the significant G × E interactions (P < 0.001) present in the four trials conducted in this study, a consistent and major QTL, cqSPA09, has shown stable performance across all four environments. It was integrated from qSPA09.1b, qSPA09.2, qSPA09.3b and qSPA09.4 by meta-analysis. The consistent QTL cqSPA09 provided a significant level of consistent contribution to shelling percentage (13.75-26.82% PVE) in the four environments, and therefore may be an important interval for improving shelling percentage in peanut breeding. The segregation in the progeny of a residual heterozygous line, RIL 15-71126, confirmed that the allele came from the parent Yuanza 9102 at the cpSPA09 locus had additive effect in increasing shelling percentage. Further studies, for example fine mapping, should be conducted to investigate its candidate genes. In addition, four consistent and minor QTLs namely cqSPB02, cqSPB04, cqSPB05 and cqSPB10, were detected in two or three environments. Such QTLs with consistent performance for shelling percentage have been identified for the first time in peanut and will be very useful for further fine mapping of these QTL regions and development of diagnostic markers for peanut breeding.
Author contribution statement HL, XZ, YC, YL, BL and HJ conceived and designed the experiments. HL, JY, XR, LH and HJ performed SSR marker development. XR and HJ developed the RIL population. XR, XZ, YC and HJ conducted field trails. ZX, ZL, XL, JL, LH and WC performed genotyping. HL and HJ constructed the genetic linkage map and performed QTL analysis. HL, BL and HJ wrote the manuscript.