QTL analysis and development of HRM markers associated with fruit shape in interspecific pears (Pyrus pyrifolia × P. bretschneideri)

Fruit shape is one of the important quantitative traits in pear (Pyrus spp.) breeding program, thus genetic study related to fruit shape could be beneficial to pear breeding. Quantitative trait loci (QTL) analysis was carried out using ‘Whangkeumbae’ (P. pyrifolia, round) × ‘Yali’ (P. bretschneideri, pyriform) and high-resolution melting (HRM) markers were developed. The genetic linkage map of ‘Whangkeumbae’ × ‘Yali’ was constructed using single nucleotide polymorphisms (SNPs) derived from Axiom Pear 70 K Genotyping Array and simple sequence repeats. The integrated genetic linkage map of ‘Whangkeumbae’ × ‘Yali’ showed ~ 90% of genome coverage, a total genetic distance of 998.2 cM, and a marker density of 1.6 cM. F1 progenies of ‘Whangkeumbae’ × ‘Yali’ showed normal distribution of fruit length (L), diameter (D), and L/D ratio. Three QTLs located in linkage group (LG) 6, 7, and 12 were identified with LOD thresholds of 2.8–3.0. Six HRM markers were developed using array-SNPs anchored in the QTLs and predicted fruit shape with 28.6–65.3% accuracy. Notably, accuracy was increased by ~ 90% using an HRM marker combination consisting of CBp06sn02, CBp07sn01, and CBp12sn03. These results could provide a better understanding of the genetic mechanism of fruit shape development and reducing pear breeding period.

in Europe, America, and Australia. There are diverse pear fruit shapes such as oblate, round, pyriform, and obovate. Most of the European pears represented by P. communis bear pyriform-shaped fruits, while Asian pears have diverse fruit shapes (White et al. 2000).
Pear fruit shape is one of the important quality indices because it affects the consumers' visual preferences. After pollination and fertilization, the receptacle of the pear begins longitudinal and horizontal development. In case of the round-shaped pears, fruit longitudinal and horizontal development proceeds simultaneously from 25 to 145 days after anthesis (Xie et al. 2013). On the other hand, during the whole fruit developmental period of pyriform pear fruit, longitudinal growth proceeded faster than horizontal growth (Bain 1961).
High-density genetic linkage maps constructed with genome-wide molecular markers are essential for genetic studies on quantitative traits (Collard et al. 2005). Pear fruit shape is also one of the quantitative traits that is affected by environmental factors (Grandillo et al. 1999;Sadrnia et al. 2007). Thus, it is possible to identify the quantitative trait loci (QTL) controlling pear fruit shape by constructing a highdensity genetic linkage map.
The development of sequencing technologies has promoted genetic study in diverse plant species by allowing single nucleotide polymorphisms (SNPs) to be produced in large quantities. SNPs are the most abundant variations in plant genomes thus, they have been widely applied in plant genomic and genetic analyses. Sequencing-based technologies have been applied in the production of large numbers of SNPs, but it requires several bioinformatics analyses for genotyping. In addition, if the sequencing is not run at the same time, the SNP calling will be different for each run. However, SNP array could accurately generate millions of specific SNPs and save time and money for the genotyping of plant populations (Rasheed et al., 2017). Currently, SNP arrays have been developed in Rosaceae family, including pear (Montanari et al. 2019), apple (Bianco et al. 2016), peach (Verde et al. 2012), and strawberry (Bassil et al. 2015).
High-resolution melting (HRM), which analyzes the temperature differences via the melting curve of PCR amplicons, can detect a single nucleotide variation. HRM analysis does not require post-PCR separation such as gel electrophoresis (Simko 2016). HRM markers have been developed based on SNPs associated with a trait of interest. For instance, Lee et al. (2012) developed HRM markers linked to genetic male sterility in pepper. Wang et al. (2016) distinguished dwarf and standard growth types of pears through HRM analysis based on SNPs associated with the PcDw.
The pear fruit growth process comprises two stages: cell division (stage I) and expansion period (stage II) (Bain 1961;Zhang et al. 2008). The morphological changes of pear fruit proceed rapidly during stage I, and the mature pear fruit shape can be determined before stage II. Although fruit shape is one of the target traits in pear breeding, a marker that can identify fruit shape has not been developed. Therefore, in the present study, QTL analysis was performed using a genetic linkage map and fruit shape phenotype data of immature fruits after cell division was completed in order to develop HRM markers that determine fruit shape of pears.
Fresh young leaves of 'Whangkeumbae', 'Yali', and the F 1 progenies were collected at full blooming stage and preserved at −70 °C until DNA extraction. Total genomic DNA was extracted from the leaves of both parents and F 1 progenies. DNA was extracted with DNeasy Plant Mini Kit (Qiagen, Hilden, Germany) according to the instruction of manufacturer. Then, the extracted DNA samples were quantified using a DS-11 Spectrophotometer (DeNovix, DE, USA) and were qualified by 1.5% agarose gel electrophoresis.

Page 3 of 11 122
Vol.: (0123456789) Fruit shape investigation Immature fruits of 'Whangkeumbae', 'Yali', and their 49 F 1 progenies were collected at 93 days after full bloom (DAFB), which corresponds to the cell expansion phase (Xie et al. 2013). Of the 93 F 1 progenies, only 49 bore fruits. Fruit length (L) and diameter (D) were measured using a digital vernier caliper (CD-15CP, Mitutoyo, Tokyo, Japan). L and D were measured as maximum height and width, respectively. L/D ratio was calculated. Phenotyping for fruit shape was investigated once with three replications.
Pear 70 K SNP array Genotype data of 'Whangkeumbae' × 'Yali' were produced through Axiom Pear 70 K Genotyping Array (Montanari et al. 2019). Raw data were subjected to several filtering steps with Axiom Analysis Suite v4.0 software (Thermo Fisher Scientific, Inc., Waltham, MA, USA). Best Practice Workflow was conducted to obtain high quality genotype data. Sample quality control (QC) thresholds were as follow: dish quality control (DQC) ≥ 0.82, QC call rate ≥ 97%, percentage of passing samples ≥ 95%, and average call rate for passing samples ≥ 98.5%. SNP QC thresholds followed the default value for diploid species type.

Construction of genetic linkage map and QTL analysis
JoinMap 5.0 was used to construct an integrated 'Whangkeumbae' × 'Yali' map (WY map) using array-SNPs and SSRs with a pseudo-testcross mapping strategy. The SNPs and SSRs genotype data were converted to cross pollinator (CP) code and loaded into JoinMap 5.0. Among the array-SNPs, PHR type had genotype < hk × hk > and NMR type had genotype < lm × ll > and < nn × np > . Markers that showed a segregation distortion rate of more than 10% of missing data were excluded in the subsequent analysis. Grouping was performed with a logarithm of odds (LOD) threshold of 10.0. Genetic distance was calculated by Kosambi's mapping function and regression mapping.
The phenotype data and WY map were loaded onto MapQTL 6.0 to identify the QTL related to the pear fruit shape. Kruskal-Wallis test, interval mapping, and 1000 permutation tests were carried out and the LOD threshold was decided at p < 0.05. After the QTL analysis, MapChart 2.3 was used to represent significant QTLs.
HRM analysis was performed using Roter-Gene6000 real-time cycler (Qiagen, Hilden, Germany) with Type-it HRM PCR Kit (Qiagen, Hilden, Germany). A total of 10 µL PCR reactant comprised 2 × HRM PCR Master Mix, including HotStarTaq Plus DNA polymerase, Type-it HRM PCR buffer with EvaGreen dye, dNTP, 10 µM of forward and reverse primers, 30 ng of gDNA, and RNase-free water. The amplification steps were as follows: predenaturation at 95 ℃ for 5 min, followed by 35 cycles 122 Page 4 of 11 Vol:. (1234567890) of denaturation at 95 ℃ for 10 s, annealing at 60 ℃ for 30 s, and extension at 72 ℃ for 10 s. Subsequently, fluorescence intensity was measured by increasing the temperature by 0.1 ℃ every 2 s in the range of 70-95 ℃. Then, the melting curve was analyzed using Rotor-Gene Q Series Software (Qiagen, Hilden, Germany).

HRM marker validation
The developed HRM markers discriminating fruit shape were applied to pear cultivars including round-(P. pyrifolia cv. Whasan and Wonhwang, P. bretschneideri cv. Dangshansuli) and pyriform-shaped cultivars (P. bretschneideri cv. Cili, Yaguangli, and Qiubaili, and P. communis cv. Bartlett, Abate Fetal, and Williams). Through the above-described DNA extraction process, DNA was extracted from young leaves of the 9 cultivars. After that, HRM analysis was repeated 2 times for each cultivar using Roter-Gene Q Series Software (Qiagen, Hilden, Germany). 'Whangkeumbae' and 'Yali' were used as control cultivars with round and pyriform genotypes, respectively.

Results
Phenotype evaluation for fruit shape Fruit shapes of 'Whangkeumbae' and 'Yali' were round and pyriform, respectively, at 93 DAFB (Fig. 1). The average L was longer in 'Yali' (5.38 ± 0.53 cm) than in 'Whangkeumbae' (4.54 ± 0.40 cm) at 93 DAFB. The average L of the F 1 individuals was 5.79 ± 0.67 cm and the L distribution ranged from 4.55 to 7.31 cm. The average D for 'Whangkeumbae' and 'Yali' was 5.00 ± 0.29 and 4.20 ± 0.36 cm, respectively. The F 1 individuals displayed 3.57-6.95 cm of D. Based on the observed L and D, the L/D ratio of 'Whangkeumbae' and 'Yali' were 0.88 ± 0.03 and 1.30 ± 0.13, respectively. L, D, and L/D ratio of F 1 progenies displayed normal distribution (Fig. 2). For F 1 progenies with a L/D ratio of more than 0.91, the maximum L/D ratio of 'Whangkeumbae', was considered pyriform and the remaining F 1 progenies were considered round. The number of F 1 progenies with a L/D ratio of 0.91 or higher was 15.3-folds more than that of F 1 progenies with a L/D ratio of less than 0.91.

SNP calling and filtering
A total of 71,363 SNPs were detected in 'Whangkeumbae' × 'Yali' after the Axiom Pear 70 K SNP array. Among them, 11,200 NMH (< lm × ll > and < nn × np >) and 8431 PHR type (< hk × hk >) SNPs were selected to construct a genetic linkage map. After filtering the SNPs showing an unexpected segregation ratio, 10,649 SNPs were finally selected for genetic linkage map construction.
Genetic linkage map of 'Whangkeumbae' × 'Yali' The WY map was divided into 17 linkage groups (LGs) corresponding to the basic chromosome number of pear (Fig. S1). The order of LG numbers corresponded with the chromosome number. The genetic linkage map consisted of 1630 loci, including 1570 SNPs and 60 SSRs (Table 1). The WY map had a total genetic distance of 998.2 cM. The number of markers per LG ranged from 58 to 138 and the average marker density was 1.6 cM. Each LG anchored at least one SSR, except for LG8. Based on the physical location of the anchored markers in each LG, the average chromosome coverage of the WY map was 90.5% (Table 1).
Fruit shape-related QTLs in 'Whangkeumbae' × 'Yali' Three candidate QTLs for L/D ratio were identified in the WY map and they were located in LG6, 7, and 12 with LOD thresholds 2.8, 3.0, and 2.9, respectively (Fig. 3). Fruit shape-related QTLs were distributed from 41.2 to 55.4 cM in LG6 and accounted for 17.8% of the explained variance. In the QTL region of LG6, 6 loci were significantly associated with L/D ratio at p ≤ 0.0001. For QTLs in LG7 and 12, the loci ranged from 0.0 to 8.6 cM and 2.0 to 30.8 cM with an explained variance 16.5 and 23.3%, respectively. Six and 10 loci in those QTL regions of LG7 and 12, respectively, were significantly associated with the L/D ratio (p ≤ 0.0001). Although there were no significant QTLs related to fruit L and D, LOD graphs between the fruit L and L/D ratio were similar (Fig. 3).
Development of HRM markers associated with fruit shape HRM primers were designed using the flanking sequences of the 22 SNPs, but 8 primer pairs were not designed because there were no primer sequences matching the criteria. Fourteen HRM markers were analyzed and 8 of 14 HRM markers did not showed association between genotype and phenotype. Thus, six HRM markers associated with fruit shape were developed using 6 SNPs out of 22 SNPs associated with fruit shape QTLs (Table 2). Two SNPs designated by CBp06sn01 and CBp06sn02 in chromosome 6 had < nn × np > and < lm × ll > genotypes, respectively. A SNP designated by CBp07sn01 in chromosome 7 had < lm × ll > genotype. Three SNPs designated by CBp12sn01, CBp12sn02, and CBp12sn03 in chromosome 12 had < nn × np > genotype. These six HRM markers discriminated homozygous and heterozygous genotypes in 'Whangkeumbae' × 'Yali' (Fig. 4). Fruit shapes of 'Whangkeumbae' × 'Yali' were divided into round and pyriform. The F 1 progenies with L/D ratio of more than 0.91 were considered pyriform and the remaining F 1 progenies were considered round. The average selection rate of the 6 HRM markers was 55.1%. Among the 6 HRM markers, CBp12sn03 had the highest selection rate (65.3%) and CBp07sn01 showed the lowest selection rate (28.6%). The selection rate for fruit shape was increased in 'Whangkeumbae' × 'Yali' using HRM marker combinations consisting of two and three markers. The HRM marker combinations composed of two markers increased the selection rate to an average of 76.6%, and the combination of CBp06sn01 with CBp12sn01, CBp06sn02 with CBp12sn03, and CBp07sn01 with CBp12sn03 showed the highest selection rate (79.6%). The HRM marker combinations consisting of three markers were able to discriminate fruit shapes with an average selection rate of 87.1%. In particular, the selection rate of HRM marker combination consisting of CBp06sn02, CBp07sn01, and CBp12sn03 was 89.8% (Table 3).
As a result of validation of the HRM marker combination in pear cultivars, CBp06sn02 discriminated 'Yali' and 'Cili'. In case of CBp07sn01, all 6 cultivars displayed the same genotype as 'Whangkeumbae'. However, the CBp12sn03 discriminated pear cultivars according to the fruit shape (Fig. S3).

Discussion
Pear fruit shape is determined by differences in the longitudinal and horizontal development of fruit. The fruit shapes of 'Whangkeumbae' and 'Yali' were confirmed as round and pyriform, respectively, at 93 DAFB (Fig. 1). Bain (1961) reported that longitudinal growth progresses more rapidly in pyriform fruit than the round fruit during the cell division period. The average length of 'Yali' was longer than the diameter, thus these results were in agreement with report of Bain (1961) and suggest that the pear fruit shape could be evaluated before the cell expansion period. In addition, phenotype distributions of L, D, and L/D ratio in 'Whangkeumbae' × 'Yali' (Fig. 2) suggest that the fruit shape is controlled by polygene (Grandillo et al. 1999;Sadrnia et al. 2007).   Table 2 Primer sequences, annealing temperature (Tm ℃), amplicon size, and SNP types of six HRM markers based on fruit shape related QTLs located in linkage groups (LGs) 6, 7, and 12 a Bi-alleic variants are denoted by K (G or T), R (A or G), and Y (C or T) LG   The WY map consisted of 1,570 SNPs and 60 SSRs, which constitute 15.2% of the 10,730 markers used for linkage analysis. The remaining 84.8% markers were not mapped due to segregation distortion or physical location mismatch with closely linked markers. Nevertheless, the resolution of WY map was increased compared to our previous 'Whangkeumbae' × 'Minibae' (WM) and 'Greensis' × 'Whasan' (GW) maps (Han et al. 2019;Oh et al. 2020) in terms of the number of mapped markers, average marker density, and genome coverage. The WM and GW maps were constructed using SNPs derived from GBS and SSRs. GBS can detect a large number of SNPs in a short time by reducing genome complexity with a restriction enzyme but, SNPs are detected in partial genome sequences (Elshire et al. 2011). The Axiom Pear 70 K Genotyping Array is a powerful fixed genotyping platform for SNP screening and is more effective than GBS (Montanari et al. 2019). Indeed, Montanari et al. (2019) observed numerous polymorphic markers using the Axiom Pear 70 K Genotyping Array rather than GBS in a segregating population. As a result, the genome coverage of WY map was higher than that of the previous maps constructed using GBS-SNPs.
A segregating population of 'Bayuehong' (P. communis × P. bretschneideri) × 'Dangshansuli' (P. bretschneideri) (BD) have been utilized for QTL analysis related to fruit characteristics including fruit shape Wu et al. 2014). Zhang et al. (2013) constructed single parental maps of 'Bayuehong' and 'Dangshansuli' using amplified fragment length polymorphism (AFLP), sequencerelated amplified polymorphism (SRAP), and SSR markers, and detected QTLs related to L/D ratio in LG1, 2, 7, and 8 of 'Bayuehong' map. Wu et al. (2014) used SNPs detected by restriction site-associated DNA sequencing and SSRs derived from pears and apples to construct an integrated BD map and identified QTLs associated with L (LG11 and 17) and D (LG3, 11, and 17). However, three QTLs related to L/D ratio were detected in LG6, 7, and 12 of WY map (Fig. 3). Although Zhang et al. (2013) and Wu et al. (2014) used the same segregating population, the QTL analysis results were different and Wu et al. (2014) thought that the results could be affected by the resolution of genetic linkage maps. As the WY map anchored more informative array-SNPs than the reduced representation library sequencing-derived SNPs, the three QTLs related to L/D ratio in WY map could also control the fruit shape of Asian pears. Moreover, we thought that the different QTL results between the present and previous studies were influenced by the genomic differences between 'Bayuehoung' (P. communis × P. bretschneideri) × 'Dangshansuli' (P. bretschneideri) and 'Whangkeumbae' (P. pyrifolia) × 'Yali' (P. bretschneideri).
Theoretically, the number of F 1 progenies for genetic study should be at least 50 (Collard et al. 2005). Although 93 F 1 progenies were used for genetic mapping, only 49 F 1 progenies were evaluated for fruit shape, as the remaining 44 F 1 progenies did not bear fruits. However, LOD values of significant QTLs (> 3.0) in LG6, 7, and 12 of the WY map supported that the 3 QTLs are associated with pear fruit shape. Vales et al. (2005) suggested that the number of QTLs detected in randomly selected subpopulations was higher than in selective genotype and selective phenotype subpopulations when the subpopulation size was small (n = 50). The 49 F 1 progenies evaluated for fruit shape phenotype were naturally selected from the total of 93 F 1 progenies, thus, it was possible to overcome the relatively small number of F 1 progenies in the QTL analysis and predict the fruit shape of mature fruit as immature fruit.
Six HRM markers associated with pear fruit shape were developed, and their melting curves and normalized melting curves divided the F 1 progenies according to the marker genotypes (Figs. 4 and S2). However, several F 1 progenies showed a discrepancy between genotype and phenotype. These results suggest that the genotype of array-SNPs is correct, and the discrepancy between genotype and phenotype is due to the influence of the environment on fruit shape (Sadrnia et al. 2007).
Among the 6 HRM markers, CBp07sn01 had the lowest selection rate (28.6%), suggesting that the QTL detected in LG7 of WY map is a minor QTL associated with fruit shape. The average selection rate of single markers ranged from 28.6 to 65.3%, whereas the marker combination comprised 3 markers increased the selection rate by 89.8% (Table 3). This means that all the three QTLs could influence fruit shape development in 'Whangkeumbae' × 'Yali'. However, the F 1 population size (49 progenies) is small to prove association between genotype and phenotype. Therefore, the HRM marker combination was applied to pear 122 Page 10 of 11 Vol:. (1234567890) cultivars bearing round or pyriform fruits. Interestingly, CBp12sn03 discriminates fruit shape of pear cultivars regardless of species (Fig. S3). Since the CBp06sn02 and CBp07sn01 were able to discriminate two cultivars ('Yali' and 'Cili') and one ('Yali') cultivar, respectively, it was thought that QTL regions related to fruit shape could be different depending on pear cultivars. Nevertheless, the CBp12sn03 could be a universal marker to discriminate fruit shape in pears regardless of cultivar and/ or species. The marker combination consisting of CBp06sn02, CBp07sn01, and CBp12sn03 could be used for marker-assisted selection in the pear breeding program.
Longitudinal growth of 'Yali' fruit was observed before cell expansion (Fig. 1). Although no markers had an LOD value above the LOD threshold in QTL analysis related to L, the two LOD graphs associated with L/D ratio and L were similar (Fig. 2). Notably, Zhang et al. (2013) found QTL associated with L/D ratio in the genetic linkage map of 'Bayuehong' (pyriform). Therefore, genetic factors related to pear fruit shape could control pyriform shape during the cell division period. Indeed, ovate cloned in tomatoes regulates cell division patterns and suppressor of ovate interacts with ovate resulting in a pyriform fruit (Wu et al. 2018). The three QTLs associated with fruit shape in 'Whangkeumbae' × 'Yali' will contribute to understanding the genetic mechanism of pear fruit shape. The HRM markers will facilitate pear breeding by selecting the desirable fruit shape early in breeding.