Identification of QTLs for Flesh Mealiness in Apple (Malus × domestica Borkh.)

Abstract

During apple (Malus × domestica Borkh.) storage, a loss in fruit firmness can occur. This is frequently associated with mealiness, an undesirable trait. There have been studies, such as phenotypic analyses and transcriptomics, as well as others employing a transgenic approach, focusing on this trait. Certain genetic approaches, such as quantitative trait loci (QTL) approach, however, have not been attempted. In this study, to identify and characterize QTLs controlling flesh mealiness and to investigate their application in apple breeding, we performed classical QTL mapping based on a bi-parental population and a genome-wide association study (GWAS) of mealiness. Phenotypic data for mealiness allowed us to identify two QTLs in the bi-parental family located on linkage group 10. The GWAS discovered significant marker-trait associations on chromosomes 2, 9, and 10. The MdPG1 locus, located on chromosome 10, was identified as the locus with the strongest significance by both QTL mapping and GWAS, suggesting its primary contribution to flesh mealiness. Using a tri-allelic simple sequence repeat marker, Md-PG1_SSR10kd, 10 kb downstream of the MdPG1 coding sequence, we divided apple accessions into six groups based on their genotypes. Among the six groups, the Md-PG1_SSR10kd genotype “2/2” had the least mealy phenotype.

Introduction

Apples (Malus × domestica Borkh.) are cultivated in almost all the temperate regions worldwide. The fruit stores well and is available for sale all year round. During storage, the texture of apple flesh changes, frequently losing its firmness, depending on the cultivar and storage methods. The loss of firmness can be accompanied by mealiness, which is caused by a loss of cellular adhesion. This results in a powdery texture with very little juice (Harker and Hallett, 1992). The changes in flesh texture after harvest directly affects the consumer preferences; crispy and juicy apples are preferred to soft and dry, i.e., mealy, apples (Galmarini et al., 2013; Harker et al., 1997, 2002; Symoneaux et al., 2012). Therefore, development of apple cultivars resistant to mealiness is highly desirable.

The changes in flesh texture have frequently been described using the flesh firmness as criterion. However, flesh firmness is not always directly associated with mealiness. The mealiness evolves after flesh softening reaches a plateau, and softening can occur without mealiness (Iwanami et al., 2005). The flesh firmness and several textural parameters at harvest and after storage have been well studied. In general, flesh softening is considered a result of structural modifications of the primary cell wall and middle lamella polysaccharides (Goulao and Oliveira, 2008). Differences between cell wall structures in rapidly softening ‘Gala’ and firmness-maintaining ‘Scifresh’ have been found at the fruitlet stage and after harvest (Ng et al., 2015). Several enzymes, such as polygalacturonase (PG), pectin methylesterase (PME), β-galactosidase, and α-L-arabinofuranosidase (AF), which affect the cell wall structure, are associated with the softening of apple flesh (Johnston et al., 2002). Their activities, gene expression levels, and interactions with plant hormones, especially ethylene, have been investigated (Ireland et al., 2014; Wakasa et al., 2006; Wei et al., 2010).

Among the many genes studied, MdPG1 is recognized as a main genetic determinant of flesh texture properties after harvest (Longhi et al., 2013b). Simultaneous recordings of mechanical and acoustic parameters allow a comprehensive characterization of flesh texture (Costa et al., 2010). Longhi et al. (2012, 2013a, b) employed this phenotypic assessment with a quantitative trait locus (QTL) mapping approach using an F₁ population, and an association study using genetic resources. They have shown that the QTLs for texture parameters two months after harvest colocalize in the middle of the linkage group (LG) 10, where MdPG1 is located. An unfavorable allele of MdPG1, expected to cause high enzymatic activity, is associated with the allele “3” of a tri-allelic simple sequence repeat (SSR) marker Md-PG1_SSR10kd, 10 kb downstream of the MdPG1 coding region (Longhi et al., 2013b). There are no significant phenotypic differences between the favorable alleles “1” and “2” of Md-PG1_SSR10kd (Longhi et al., 2013b). The significance of Md-PG1_SSR10kd was confirmed using populations with different genetic backgrounds (Longhi et al., 2013a). Based on these results, Longhi et al. (2013a) concluded that Md-PG1_SSR10kd is the most suitable DNA marker to be used in breeding for a superior fruit texture. The acoustic parameters measured in their method allow the evaluation of mealiness indirectly, although, these are associated with the cell wall breaking phenomenon that generates crispness rather than with mealiness (Duizer, 2001; Longhi et al., 2013b).

To date, several other methods have been used available to evaluate the mealiness feature. Because cellular adhesion is fundamentally related to this feature, the methods focus on the strength of the cellular adhesion. Until now, the easiest way to evaluate mealiness has been a sensory analysis. However, this is difficult to transfer to other laboratories and/or panels because there are some problems with robustness, especially in quantitative measurements caused by differences in testers and criteria. Measuring tensile strength is a widely used method (Glenn and Poovaiah, 1990; Harker and Hallett, 1992; Stow, 1989; Tu et al., 2000), but it measures the combined forces associated not only with cell separation, but also with fracturing and rupture (Harker et al., 1997). The method is also time-consuming when applied to many samples. Simple cell separation data can be obtained by measuring the loss of weight after shaking fruit discs in a sucrose solution (Iwanami et al., 2005; Motomura et al., 2000) and by counting cells isolated from the flesh after vortexing (Segonne et al., 2014). The reliability of these two methods has been confirmed by comparisons with sensory analyses (Motomura et al., 2000; Segonne et al., 2014). Both methods have advantages when dealing with hundreds of samples and examining the evolution of mealiness after the softening reaches a plateau.

Recently, some molecular biology studies have deciphered a part of the mechanism responsible for the evolution of mealiness. A global transcriptomic approach has revealed a role of MdPME2 in the development of this feature (Segonne et al., 2014). The expression level of MdPME2 in fruit of mealy genotypes are lower than those of non-mealy genotypes before and at harvest in addition to after 2 months of storage, suggesting that this gene has a role in stiffening the cell wall. In general, the PME enzyme is considered to accelerate mealiness because it solubilizes pectin in the cell walls. Thus, the study presents a new insight into the role of this gene in the development of apple-flesh mealiness. It has been suggested that MdAF3 is involved in mealiness (Nobile et al., 2011). However, Segonne et al. (2014) did not observe consistent MdAF3 expression profiles using three mealy/non-mealy pairs for comparison. The MdPG1 expression level increased in mealy fruit only after four months of cold storage, suggesting that it is involved in the process during the later storage stages (Segonne et al., 2014). The transgenic approach has clarified the involvement of MdPG1 in the evolution of mealiness. After four months of cold storage, a transgenic ‘Gala’ containing downregulated MdPG1 showed a stronger cellular adhesion than the wild type (Atkinson et al., 2012). Moreover, cellar adhesion levels in the leaves of transgenic apple plants containing overexpressed MdPG1 decreased; however, the experiment has not been conducted in fruit (Atkinson et al., 2002). These studies suggest that MdPG1 plays an important role in cell-to-cell adhesion.

As mentioned above, there have been some mealiness-focused studies, such as phenotypic analyses and transcriptomics, as well as studies using transgenic approaches. However, the contribution of MdPG1 to mealiness in terms of cellular adhesion has not been fully clarified using natural genetic resources. Moreover, genetic approaches such as bi-parental QTL analysis employing linkage maps derived from bi-parental populations, and genome-wide association studies (GWAS), have not been used. However, in separate bi-parental QTL mapping and GWAS employments, it is difficult to validate the non-parental allele effect and detect rare-variant QTL alleles and those QTLs with a slight effect. A simultaneous implementation of these approaches is a good method to overcome this limitation (Myles et al., 2009). However, GWAS is not a common approach in apple studies because of the scarcity of sufficient single nucleotide polymorphism (SNP) markers. Recently, a high-density 20K SNP array for the apple has been developed (Bianco et al., 2014), allowing QTL mapping by GWAS.

The aim of our study was to identify and characterize QTLs controlling flesh mealiness employing a bi-parental QTL mapping approach in an F₁ population and a GWAS using modern apple genetic resources. The application of identified QTLs in breeding programs was also considered.

Materials and Methods

Plant material

An F₁ population derived from the cross between ‘Orin’ × ‘Akane’ (OA) comprising 119 individuals, was used for the QTL analysis. The progeny were planted in the experimental orchard of the Apple Research Station, NARO (Morioka, Japan). All trees were grafted onto the dwarfing rootstock JM1. For GWAS, 82 diploid apple accessions were used (Table 1). These accessions mainly comprised modern Japanese apple cultivars and their founders. Accessions grafted onto the dwarfing rootstocks (M.9, JM1, JM7, JM8, or M.26) were planted in the same orchard. Orchard and tree management was regularly conducted in the same way as in commercial orchards in Japan. Genomic DNA was isolated from young leaves using a DNeasy Plant Mini Kit (Qiagen, Hilden, Germany) or an automatic extracting device PI-50α (Kurabo, Osaka, Japan).

Table 1

Genotypes of genetic markers for apple accessions used in the genome-wide association study (GWAS).

Quantitative measurement of mealiness

The assessments were conducted during two successive years, in 2012 and 2013, using OA, and in 2013 and 2014 using the accessions. Fruit samples from each OA individual and accession were collected during one harvest when plants reached maturity (based on the criteria of Kunihisa et al. (2014)). The measurement of mealiness was conducted according to Iwanami et al. (2005), with slight modifications. Namely, the harvested apples were stored at 20°C, 85 ± 5% relative humidity, for 4 weeks before the measurement. The measurement was conducted using five fruits per genotype. If a genotype did not yield sufficient fruit because of biennial bearing and/or storage-related diseases, then all the available fruit of the genotype was used for the analysis. For each fruit, 10 discs, 10 mm in diameter and 5-mm thick, were taken from the equator cortex with a cork borer. The discs were soaked in 12% sucrose solution at reduced pressure using a water aspirator (A-3S; EYELA, Tokyo, Japan) for 45 min and then weighed. The discs were shaken (180 rpm, 3-cm-wide shake) for 7 h in a 30-mm-diameter test tube containing 10 mL of 12% sucrose solution on a continuous shaker (Plus shaker EP-1; Taitec, Tokyo, Japan), and then re-weighed. The degree of mealiness (DM) was calculated as (W_i—W_s)/W_i, where W_i and W_s are the weights of discs before and after shaking, respectively. To obtain a normally distributed dataset, arcsine-transformation (ADM) of DM values was performed. Negative DM values were considered equal to zero. Then, the ADM values were calculated. A negative DM value indicates an increase in the fruit disc weight after being shaken, which may be caused by the sucrose solution filling the remaining air spaces in the fruit disc.

Genetic linkage map and QTL analysis

The integrated parental genetic linkage map (Kunihisa et al., 2014), which comprises 17 LGs, was used for the QTL analysis using the ADM values. We added the SSR marker Md-PG1_SSR10kd (Longhi et al., 2013b) to the map before the QTL analysis. The amplification procedure for Md-PG1_SSR10kd and the primers were as described in the study of Longhi et al. (2013b). Amplified alleles were scored using the allele codes 1, 2, or 3. The QTL analysis was then performed using MapQTL6 software (van Ooijen, 2009). It included an initial interval mapping followed by the automatic cofactor selection. Then, the QTL analysis using the Multiple QTL Mapping algorithm was performed with the suggested cofactors. The LOD threshold to define the significance of the QTLs was computed using 5000 permutations.

Genotyping of apple accessions

Eighty-two apple accessions were genotyped using the 20K SNP array for apples (Bianco et al., 2014) following the standard Illumina protocol (San Diego, CA, USA). Genotyping data were analyzed with a GenCall threshold of 0.15 using the GenomeStudio software. SNPs with no polymorphisms in the accessions were removed. Finally, a dataset of 15844 SNPs was used for the subsequent analyses.

The accessions were also tested for Md-PG1_SSR10kd (Longhi et al., 2013b) and AF057134-SSR (Silfverberg-Dilworth et al., 2006), which were chosen as cofactors in the QTL analysis with MapQTL6 (see Results) and were flanking QTLs, as described in the original papers.

Estimation of linkage disequilibrium (LD) and GWAS

LD estimations and GWAS were performed using data obtained from the 20K SNP array. Based on the published SNP information (Bianco et al., 2014), SNPs with positional data described as “primary assembly” were aligned according to their positions and their missing data were imputed using BEAGLE 3.3.2 software (Browning and Browning, 2009). SNPs with incomplete positional data, assembled only in “scaffold” or without positional information, had their missing data filled using the population mean of that marker subsequently located on a fictive chromosome (chr), chr 18.

LD values between pairs of markers were estimated using the “LD” function in the R package “genetics”. We calculated the r² values for all the SNP pairs within a 5-Mb physical distance. The fictive chr 18 was excluded from the calculations. The relationships between the r² values and physical distances between the corresponding markers were further modeled by fitting local polynomials using the function “locpoly” in the R package “KernSmooth”.

An association analysis was performed using a mixed model (Yu et al., 2006) implemented in the R package “rrBLUP” (Endelman, 2011) as a “GWAS” function. SNPs with minor allele frequencies below 0.05 were removed from the analysis. A kinship matrix was calculated from all the genetic data using the “A.mat” function (Endelman and Jannink, 2012) in the “rrBLUP” package. Subsequently, principal components (PCs), which were used as the covariates in GWAS, were calculated from the kinship matrix. The number of PCs to include in the model was based on the Bayesian Information Content value calculated using the GAPIT software (Lipka et al., 2012). The false discovery rate of 0.05, determined using the “qvalue” package implemented in the “rrBLUP” package of R was applied to a threshold.

Statistical analyses

All the statistical analyses were conducted using R and Microsoft Excel.

For estimations of the variance components of ADM, we applied a balanced dataset. We performed calculations for the OA and accessions separately. A model of two-way analysis of variance (ANOVA) was applied to estimate the genetic variance:   

$$P_{ij}=\mu +G_i\mathrm{ }+Y_j+E_{ij}$$

, where P_ij is the ADM in the ith genotype of the year j, and μ is the grand mean. G_i is the effect of the ith genotype, Y_j is the effect of the jth year, and E_ij is the residual in the ith genotype of the jth year. i = 1–81 and j = 1–2 for OA, and i = 1–72 and j = 1–2 for accessions. The broad-sense heritability (h_B²) for the mean ADM in 2 years in each dataset was estimated using the formula: h_B² = σ_g²/(σ_g² + σ_e²/2), where σ_g² is the variance of G_i and σ_e² is the variance of E_ij.

Results and Discussion

Phenotypic distribution and h_B² values

All 119 OA individuals and all 82 accessions yielded phenotypic data for at least 1 year. We obtained the data for 2 years from 81 OA individuals and 72 accessions and for the rest, only for 1 year because of the biennial bearing or storage-associated disease. The average ADM values for 119 OA individuals and 82 accessions were 41.6 ± 21.5 and 32.5 ± 23.8, respectively (Tables 1 and 2). The ADM values for OAs and accessions ranged from 0 (non-mealy) to 80.2 (quite mealy) and from 0 to 78.0, respectively. The ADM values for OA showed transgressive segregation (Fig. 1).

Table 2

Arcsine-transformed degree of mealiness (ADM) and harvest date for tested apple materials based on Md-PG1_SSR10kd genotype subgroups.

Fig. 1

Distribution of arcsine-transformed degree of mealiness (ADM) in F₁ population ‘Orin’ × ‘Akane’ (119 individuals).

Variance components were calculated using ANOVA (Table 3). To generate a balanced dataset, we chose 81 OA individuals and 72 accessions tested over two successive years. The ratio of the genetic variance to the phenotypic variance of OAs and accessions was 71% and 83%, respectively (Table 3). Yearly variances were quite low in both datasets. h_B² values of OAs and accessions were 0.83 and 0.92, respectively.

Table 3

Variance components of arcsine-transformed degree of mealiness (ADM) of tested apple materials over 2 years.

QTL analysis of the ADM data

For the QTL analysis, we used all 119 OA individuals, for which we obtained phenotypic values for at least 1 year. The QTL analysis of the OA ADMs, using the integrated map, identified two QTLs in LG 10 (Fig. 2). One QTL, located at the Md-PG1_SSR10kd locus, explained 14.4% of the phenotypic variance and the other QTL, located very close to AF057134-SSR, explained 11.6%. The QTL analysis, using both parental maps, identified Md-PG1_SSR10kd as a QTL in both ‘Orin’ and ‘Akane’. AF057134-SSR was identified as a QTL only in ‘Akane’ (data not shown). Although the estimation of genetic variance was performed for a subset of the data, the sum of the genotypic variance explained by the two QTLs was 26.0% of phenotypic variance, which represents 36.7% of the genetic variance. Therefore, the two QTLs were considered major genetic factors inducing fruit mealiness in the progeny. To obtain an accurate genetic contribution of these QTLs, preparation of a balanced dataset with a larger number of individuals and test years will be needed.

Fig. 2

Genomic positions of quantitative trait loci (QTLs) for arcsine-transformed degree of mealiness (ADM) in apple identified on the integrated genetic map of ‘Orin’ × ‘Akane’ (OA). Genetic distance is indicated in centimorgans. OA10, linkage group 10 of the integrated map. Markers indicated in bold are co-factors in the QTL analysis. The dotted line indicates the threshold of P = 0.05 obtained by 5000 permutations.

Estimation of LD

LD structure of genetic resources is essential for calculating the genetic heterozygosity and recombination content of materials (Nordborg and Tavaré, 2002). Because GWAS exploits the LD between markers and causal polymorphisms, a large LD is useful for identifying the associations between markers and traits of interest using a small number of markers. However, the identification of candidate polymorphisms (genes) under such conditions is difficult as the confidence interval may increase. Thus, to examine the LD structure of our materials, we calculated the r² of LD. In 82 accessions, the average r² values dropped below 0.2 at 375 kb (Fig. 3). This suggested that our material had a smaller LD block (higher levels of genetic diversity and recombination) than the materials of Kumar et al. (2013), but a larger LD block than in the study of Leforestier et al. (2015). Thus, in the present GWAS, a large number of SNPs used in the analysis showed a medium level of diversity.

Fig. 3

Plots of average linkage disequilibrium (LD) values (r²) against physical distances in increments of 10 kb. Gray curves show local polynomial fits obtained using kernel smoothing regression.

GWAS for the ADM

The association analysis of the ADM values for 82 accessions detected 39 significant SNPs located in six regions (one each in the chr 2, 9, and 10, and three in the fictive chr 18) (Table 4; Fig. 4A). The observed −log₁₀(P) values fitted well to the expected −log₁₀(P) values up to ~2 (Fig. 4B), suggesting that the population structure and kinship in the applied model were controlled. To compare the physical positions of significant SNPs, the physical locations of their source sequences (Bianco et al., 2014) were obtained from the Apple Genome v. 1.0 (Velasco et al., 2010). The search identified three regions of significant SNPs on chr 2, 9, and 10 (Table 4; Fig. 4C). However, a relatively high rate of errors in the apple genome assembly means that there are some inconsistencies between the positions obtained by genetic mapping and the positional data in the public database (Antanaviciute et al., 2012; Kunihisa et al., 2016). Thus, to confirm the origin of these regions, a r² LD analysis of all the significant SNP pairs was conducted. The inter-regional significant marker pairs showed low r²; the means for chr 2–9, 2–10, and 9–10 were 0.05, 0.07, and 0.04, respectively. However, the intra-regional r² values were 0.85, 0.83, and 0.53 for chr 2, 9, and 10, respectively. This result suggested that these three regions are independent and possibly associated with mealiness. Thus, the association study identified three regions associated with mealiness.

Table 4

Significant SNPs identified by genome-wide association study of the arcsine-transformed degree of mealiness in apples.

Fig. 4

Genome wide association analysis of arcsine-transformed degree of mealiness in the apple. Manhattan plot of the −log₁₀(P) values of all the tested SNPs aligned according to Bianco et al. (2014), (A); quantile-quantile plot of the observed and expected −log₁₀(P) values of all the tested SNPs, (B); Manhattan plot of the −log₁₀(P) values of the significant SNPs located on chromosome 10, aligned according to the Apple Genome v. 1.0 (Velasco et al., 2010), (C). Chromosomes are shown in two colors and are ordered from 1 to 18. Chromosome 18 is a fictive chromosome for placing the SNPs anchored only to the “scaffold”, as described in Bianco et al. (2014). Horizontal lines indicate the genome-wide significance threshold defined by the false discovery rate of 0.05. The names of the most significant and the second most significant SNPs are shown.

The region on which significant SNPs on chr 10 located ranged from 16.35 Mb to 18.15 Mb and contained MdPG1 (MDP0000326734, 18.13 Mb; Fig. 4C). The SNP with the most significant −log₁₀(P) value in the genome was SNP_FB_0028782. It is located ~600 kb from MdPG1 according to the Apple Genome v. 1.0 contig (Velasco et al., 2010). However, we did not find accessions homozygous for the minor allele in SNP_FB_0028782. This suggested that the probe of this SNP may have hybridized to multiple loci in the genome. Homozygotes of SNP_FB_0028782 were completely associated with the null dose of Md-PG1_SSR10kd-3, and heterozygotes were completely associated with the single and double doses of Md-PG1_SSR10kd-3 (data not shown). The second most significant SNPs, SNP_FB_0832810 and SNP_FB_0832819, were located on the contig flanking the MdPG1. MdPG1 weakens the cellular adhesion in fruit (Atkinson et al., 2012) and could be involved in mealiness development during prolonged storage (Segonne et al., 2014). As our storage conditions were sufficient for expression of MdPG1 (20°C, 4 weeks), the MdPG1 may be the causal gene underlying QTL on chr 10, detected both in the QTL analysis and GWAS. The high significance of the MdPG1 region also suggested its major contribution to mealiness. Polymorphisms of SNP_FB_0832810 and SNP_FB_0832819 were consistent with the dose of the Md-PG1_SSR10kd-3 allele (data not shown). These results indicated that for the markers tested, the dose level of Md-PG1_SSR10kd-3 is the most significant genetic factor affecting the variation in the ADM values.

Relationship between harvest time, ADM and genotypes of Md-PG1_SSR10kd

The relationships between harvest time, ADM, and genotypes of Md-PG1_SSR10kd were of interest because QTLs for harvest time and storability traits have been identified in the flanking region of MdPG1 (Chagné et al., 2014; Costa et al., 2010; Kenis et al., 2008; Kunihisa et al., 2014). Weak correlations observed between harvest time and ADM values for the OA (r = −0.39, P < 0.001) and accessions (r = −0.37, P < 0.001) were consistent with the weak negative correlation between harvest time and storability found in a bi-parental population (Chagné et al., 2014) and genetic resources (Nybom et al., 2013). However, our GWAS for harvest time did not identify any significant SNPs (data not shown). This suggested the existence of a more complex genetic architecture involving intermediate and small QTLs for harvest time rather than for mealiness.

The comparisons of harvest time and ADMs for the six Md-PG1_SSR10kd genotypic subgroups (1/1, 1/2, 1/3, 2/2, 2/3, and 3/3) identified an early harvest time and the highest ADM in the Md-PG1_SSR10kd-3/3 subgroup (Table 2). This result suggests that the 3/3 subgroup tends to be mealy and early ripening. However, this tendency of the Md-PG1_SSR10kd-3/3 subgroup is not supported by sufficiently robust evidence; to strengthen our conclusion, an analysis using a larger sample is needed. Less mealy individuals and a wider range of the harvest periods were observed in the Md-PG1_SSR10kd-2/2 subgroup. The 2/2 genotype could be ideal for obtaining hybrids bearing less mealy fruit and not affected by the harvest period (Fig. 5).

Fig. 5

Relationship between arcsine-transformed degree of mealiness (ADM) in apple and harvest time for all the tested materials, subdivided on the basis of the Md-PG1_SSR10kd genotypes. The datasets of ‘Orin’ × ‘Akane’ individuals and accessions are shown as ● or marked with an identification number corresponding to Table 1, respectively. Harvest time is the number of days counted from August 1. *, ***, and NS indicate significance at the level of 5%, 0.1%, and non-significance, respectively.

We also observed differences in the relationships between the ADM and harvest time among Md-PG1_SSR10kd genotypic subgroups. Scatter plots based on subgroups of accessions and OA individuals identified different trends for ADM-harvest time associations in different subgroups (Fig. 5). Significant negative correlation coefficients were observed in the Md-PG1_SSR10kd-1/1, -2/3, and -3/3 (P < 0.05) subgroups, but not in other subgroups (1/2, 1/3, and 2/2). These results suggest that the relationship between ADM and harvest time depends on the Md-PG1_SSR10kd genotype. Despite the small sample size in the specific subgroups, clear phenotypical differences due to alleles 1 and 2 were observed.

The amino acid sequence of the MdPG1 haplotype associated with allele 3 predicts a higher enzymatic activity than the other two haplotypes associated with allele 1 and/or 2 (Longhi et al., 2013b). Longhi et al. (2013b) also reported that Md-PG1_SSR10kd-1 and -2 are not completely associated with particular haplotypes of MdPG1. This could indicate that the nucleotide variants of Md-PG1_SSR10kd are not the causal mutations themselves. Therefore, we need further research on the enzymatic activity and expression mechanism of the MdPG1 haplotype, even in the noncoding region, to identify the causal mutation of MdPG1.

Other QTLs

In OA progeny, only additive effects of MdPG1 and QTL flanking AF057134-SSR were confirmed by ANOVA (data not shown). However, we compared the effects of only four Md-PG1_SSR10kd genotypes and the 207- and 223-bp alleles of AF057134-SSR. To investigate why QTL flanking AF057134-SSR was not detected in GWAS, we tested this SSR marker using the 82 accessions (Table 1). Six alleles were amplified from the 82 accessions. The 223-bp allele associated with severe mealiness in OA was observed in a heterozygous state in four of these accessions. Interestingly, the 223-bp allele was observed only in ‘Akane’ and its F₁ (‘Sansa’) and F₂ (Morioka 64 and Morioka 68), suggesting that ‘Akane’ is the founder of this allele in Japanese modern apple breeding. The frequency of 223-bp allele was 0.02 in the 82 accessions, which was under the threshold for discarding SNP data from GWAS. These results suggested that the low frequency of the causal allele associated with 223 bp of AF057134-SSR, and insufficient sample size, prevented the identification of this association in GWAS. To increase the power of GWAS, the sample size should be increased (Korte and Farlow, 2013).

In the region near AF057134-SSR, QTLs for fruit firmness at harvest (Costa et al., 2010) and after two months of cold storage (Bink et al., 2014), were identified by bi-parental and pedigree-based QTL mapping, respectively. It is not clear whether the causal gene considered in our study and described in the previous reports is the same; all these studies had a lower marker density and larger confidence interval. To improve our understanding of these QTLs, marker enrichment and other segregating populations should be employed. A new strategy for detecting QTLs (e.g., metaQTL analysis) could also be beneficial. Several QTLs for fruit texture and storability traits have been identified on this chromosome (Ben Sadok et al., 2015; Bink et al., 2014; Chagné et al., 2014; Costa et al., 2010; Kenis et al., 2008; King et al., 2000; Kumar et al., 2013; Kunihisa et al., 2014; Longhi et al., 2013b). Our findings confirm the importance of chr 10 for elucidating the regulatory mechanisms of fruit texture and storability.

A haplotype-based QTL survey, using only SNPs heterozygous in ‘Fuji’, has identified one QTL for mealiness in the middle of chr 1 (Kunihisa et al., 2016). The lack of significance of the MdPG1 locus in Kunihisa et al. (2016), possibly due to the Md-PG1_SSR10kd genotype of ‘Fuji’ (1/2), prevented the detection of QTL. We did not detect the chr 1 QTL. This may have been caused by the lack of allelic effect in the OA in the QTL mapping study and the different approaches employed by Kunihisa et al. (2016) in the present study.

We identified two significant SNPs on chr 9, the chromosome harboring the MdPME2 (MDP0000245813), involved in flesh mealiness (Segonne et al., 2014). However, these two loci were physically separate according to the Apple Genome v. 1.0 (Velasco et al., 2010). The SNP was aligned to 1.1–2.0 Mb, and MdPME2 to 17.1 Mb. We did not find any reports of a QTL for a textural trait on chr 2, where seven significant SNPs were identified by our GWAS. Further research will be required to clarify the effect of these QTLs. Their identification suggests a complex regulation of mealiness in terms of cellular adhesion strength.

Use of the Md-PG1_SSR10kd marker for breeding purposes

Despite the different storage conditions, our results and those of the previous study identifying QTLs for texture parameters (Longhi et al., 2013a) indicate that Md-PG1_SSR10kd is a very effective molecular marker of flesh texture. Based on our results, breeders can easily obtain less mealy hybrids from the Md-PG1_SSR10kd-2/2 and -1/2 seedlings. Additionally, 1/1 genotypes are favorable for late-maturing hybrids. Genotypes possessing at least one Md-PG1_SSR10kd-3 (i.e., 1/3, 2/3, and 3/3) are unfavorable because highly mealy fruit hybrids are expected during the entire maturation period. Thus, obtaining less mealy, early-maturing hybrids is more difficult than producing similar middle- and late-maturing hybrids. Only two genotypic classes, 1/2 and 2/2, are available in the early season, but there are three genotypic classes, 1/1, 1/2, and 2/2, in the later fruiting season. Genotypes of Md-PG1_SSR10kd can be employed for parental selection in breeding using other conventional criteria, such as phenotype. For example, Morioka 67 is a valuable breeding material because it bears less mealy fruit in the early season and has a favorable Md-PG1_SSR10kd (2/2) genotype. Thus, the use of Morioka 67 as a crossing parent should result in early-maturing, less-mealy hybrids. Moreover, the transgressive ADM segregation in the OA population suggests that crosses between cultivars with the Md-PG1_SSR10kd-2 allele can be employed to generate less mealy hybrids.

Conclusion

We identified four QTLs for flesh mealiness in apples during storage at 20°C. One was associated with MdPG1 and was identified by both the QTL mapping and GWAS. The other new QTLs were identified using only one approach. Identifying the genes underlying these QTLs is necessary for a better understanding of the mechanism of mealiness. The high significance level of MdPG1 indicates its primary importance in flesh mealiness. To the best of our knowledge, the detailed relationships among Md-PG1_SSR10kd genotypes, mealiness, and harvest time are described here for the first time. Our results strongly support the reliability and advantages of using the Md-PG1_SSR10kd marker in apple breeding programs. The identification of three other QTLs suggests the involvement of many genes. Our GWAS used the largest number of currently available SNPs in the apple. However, the number of SNPs and accessions included in the analysis was insufficient for detecting QTLs with small effects and/or rare variants, suggesting that the number of SNPs and accessions should be increased for a more accurate and meaningful GWAS.

Acknowledgments

We thank Dr. Charles-Eric Durel and Dr. Hélène Muranty (Institute de Recherche en Horticulture et Semences, INRA, France) for fruitful discussions about GWAS. We would also like to express our gratitude to the breeding sections of the following institutes for kindly providing plant materials of domestic cultivars: Apple Research Institute of the Aomori Prefectural Industrial Technology Research Center and Nagano Fruit Tree Experiment Station.

Literature Cited

•  Antanaviciute,  L.,  F.  Fernandez-Fernandez,  J.  Jansen,  E.  Banchi,  K. M.  Evans,  R.  Viola,  R.  Velasco,  J. M.  Dunwell,  M.  Troggio and  D. J.  Sargent. 2012. Development of a dense SNP-based linkage map of an apple rootstock progeny using the Malus Infinium whole genome genotyping array. BMC Genomics 13: 203.
•  Atkinson,  R. G.,  R.  Schröder,  I. C.  Hallett,  D.  Cohen and  E. A.  MacRae. 2002. Overexpression of Polygalacturonase in transgenic apple trees leads to a range of novel phenotypes involving changes in cell adhesion. Plant Physiol. 129: 122–133.
•  Atkinson,  R. G.,  P. W.  Sutherland,  S. L.  Johnston,  K.  Gunaseelan,  I. C.  Hallett,  D.  Mitra,  D. A.  Brummell,  R.  Schroder,  J. W.  Johnston and  R. J.  Schaffer. 2012. Down-regulation of POLYGALACTURONASE1 alters firmness, tensile strength and water loss in apple (Malus × domestica) fruit. BMC Plant Biol. 12: 129–129.
•  Ben Sadok,  I.,  A.  Tiecher,  D.  Galvez-Lopez,  M.  Lahaye,  P.  Lassere-Zuber,  M.  Bruneau,  S.  Hanteville,  R.  Robic,  R.  Cournol and  F.  Laurens. 2015. Apple fruit texure QTLs: year and cold storage effects on sensory and instrumental traits. Tree Genet. Genomes 11: 119.
•  Bianco,  L.,  A.  Cestero,  D. J.  Sargent,  E.  Banchi,  S.  Derdak,  M.  Di Gualdo,  S.  Salvi,  J.  Jansen,  R.  Viola,  I.  Gut,  F.  Laurens,  D.  Chagné,  R.  Velasco,  E.  van de Weg and  M.  Troggio. 2014. Development and validation of a 20K single nucleotide polymorphism (SNP) whole genome genotyping aray for apple (Malus × domestica Borkh). PLoS ONE 9: e110377.
•  Bink,  M. C. A. M.,  J.  Jansen,  M.  Madduri,  R. E.  Voorrips,  C.-E.  Durel,  A. B.  Kouassi,  F.  Laurens,  F.  Mathis,  C.  Gessler,  D.  Gobbin,  F.  Rezzonico,  A.  Patocchi,  M.  Kellerhals,  A.  Boudichevskaia,  F.  dunemann,  A.  Peil,  A.  Nowicka,  B.  Lata,  M.  Stankiewicz-Kosyl,  K.  Jeziorek,  E.  Pitera,  A.  Soska,  K.  Tomala,  K.  Evans,  F. F.  Fernandez,  W.  Guerra,  M.  Korbin,  S.  Keller,  M.  Lewandowski,  W.  Plocharski,  K.  Rutkovski,  E.  Zurawicz,  F.  Costa,  S.  Sansavini,  S.  Tartarini,  M.  Komjanc,  D.  Mott,  A.  Antofie,  M.  Lateur,  A.  Rondia,  L.  Gianfranceschi and  E. W.  Van de Weg. 2014. Bayesian QTL analyses using pedigreed families of an outcrossing species, with application to fruit firmness in apple. Theor. Appl. Genet. 127: 1073–1090.
•  Browning,  B. L. and  S. R.  Browning. 2009. A unified approach to genotype imputation and haplotype phase inference for large data sets of trios and unrelated individuals. Am. J. Hum. Genet. 84: 210–223.
•  Chagné,  D.,  D.  Dayatilake,  R.  Diack,  M.  Oliver,  H.  Ireland,  A.  Watson,  S. E.  Gardiner,  J. W.  Johnston,  R. J.  Schaffer and  S.  Tustin. 2014. Genetic and environmental control of fruit maturation, dry matter and firmness in apple (Malus × domestica Borkh.). Hort. Res. 1: 14046
•  Costa,  F.,  C. P.  Peace,  S.  Stella,  S.  Serra,  S.  Musacchi,  M.  Bazzani,  S.  Sansavini and  W. E.  Van de Weg. 2010. QTL dynamics for fruit firmness and softening around an ethylene-dependent polygalacturonase gene in apple (Malus × domestica Borkh.). J. Exp. Bot. 61: 3029–3039.
•  Duizer,  L. 2001. A review of acoustic research for studying the sensory perception of crisp, crunchy and crackly textures. Trends Food Sci. Technol. 12: 17–24.
•  Endelman,  J. B. 2011. Ridge regression and other kernels for genomic selection with R package rrBLUP. Plant Genome 4: 250–255.
•  Endelman,  J. B. and  J.-L.  Jannink. 2012. Shrinkage estimation of the realized relationship matrix. G3: Genes, Genomes, Genet. 2: 1405–1413.
•  Galmarini,  M. V.,  R.  Symoneaux,  S.  Chollet and  M. C.  Zamora. 2013. Understanding apple consumers’ expectations in terms of likes and dislikes. Use of comment analysis in a cross-cultural study. Appetite 62: 27–36.
•  Glenn,  G. M. and  B. W.  Poovaiah. 1990. Calcium-mediated postharvest changes in texture and cell wall structure composition in ‘Golden Delicious’ apples. J. Amer. Soc. Hort. Sci. 115: 962–968.
•  Goulao,  L. F. and  C. M.  Oliveira. 2008. Cell wall modifications during fruit ripening: when a fruit is not the fruit. Trend. Food Sci. Technol. 19: 4–25.
•  Harker,  F. R. and  I. C.  Hallett. 1992. Physiological changes associated with development of mealiness of apple fruit during cool storage. HortScience 27: 1291–1294.
•  Harker,  F. R.,  J.  Maindonald,  S. H.  Murray,  F. A.  Gunson,  I. C.  Hallett and  S. B.  Walker. 2002. Sensory interpretation of instrumental measurements. I. texture of apple fruit. Postharvest Biol. Technol. 24: 225–239.
•  Harker,  F. R.,  R. J.  Redgwell,  I. C.  Hallett,  S. H.  Murray and  G.  Carter. 1997. Texture of fresh fruit. Hort. Rev. 20: 121–224.
•  Ireland,  S. H.,  G.  Kularajathevan,  R.  Muddumage,  J. E.  Tacken,  J.  Putterill,  J. W.  Johnston and  R. J.  Schaffer. 2014. Ethylene regulates apple (Malus × domestica) fruit softning through a dose × time-dependent, mechanism and through differential sensitivities and dependencies of cell wall-modifying genes. Plant Cell Physiol. 55: 1005–1016.
•  Iwanami,  H.,  S.  Moriya,  N.  Kotoda,  S.  Takahashi and  K.  Abe. 2005. Influence of mealiness on the firmness of apples after harvest. HortScience 40: 2091–2095.
•  Johnston,  J. W.,  E. W.  Hewett and  M. L. A. T. M.  Hertog. 2002. Postharvest softning of apple (Malus domestica) fruit: a review. New Zealand J. Crop Hort. Sci. 30: 145–160.
•  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.
•  King,  G. J.,  C.  Maliepaard,  J. R.  Lynn,  F. H.  Alston,  C.-E.  Durel,  K. M.  Evans,  B.  Griffon,  F.  Laurens,  A. G.  Manganaris,  T.  Schrevens,  S.  Tartarini and  J.  Verhaegh. 2000. Quantitative genetic analysis and comparison of physical and sensory descriptors relating to fruit flesh firmness in apple (Malus pumila Mill.). Theor. Appl. Genet. 100: 1074–1084.
•  Korte,  A. and  A.  Farlow. 2013. The advantage and limitations of trait analysis with GWAS: a review. Plant Methods 9: 29.
•  Kumar,  S.,  D. J.  Garrik,  M. C. A. M.  Bink,  C.  Whitworth,  D.  Chagne and  R. K.  Volz. 2013. Novel genomic approaches unravel genetic architecture of complex traits in apple. BMC Genomics 14: 393.
• Kunihisa, M., S. Moriya, K. Abe, K. Okada, T. Haji, T. Hayashi, Y. Kawahara, R. Itoh, T. Itoh, Y. Katayose, H. Kanamori, T. Matsumoto, S. Mori, H. Sasaki, T. Matsumoto, C. Nishitani, S. Terakami and T. Yamamoto. 2016. Genomic dissection of a ‘Fuji’ apple cultivar: re-sequencing, SNP marker development, definition of haplotypes, and QTL detection. Breed. Sci. 66: 499–515.
•  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.
•  Leforestier,  D.,  E.  Ravon,  H.  Muranty,  A.  Cornille,  C.  Lemaire,  T.  Giraud,  C.-E.  Durel and  A.  Branca. 2015. Genomic basis of the differences between cider and dessert apple varieties. Evol. Appl. 8: 650–651.
•  Lipka,  A.,  F.  Tian,  Q.  Wang,  J.  Peiffer,  M.  Li,  P. J.  Bradbury,  M. A.  Gore,  E. S.  Buckler and  Z.  Zhang. 2012. GAPIT: genome association and prediction integrated tool. Bioinformatics 28: 2397–2399.
•  Longhi,  S.,  L.  Cappellin,  W.  Guerra and  F.  Costa. 2013a. Validation of a functional molecular marker suitable for marker-assisted breeding for fruit texture in apple (Malus × domestica Borkh.). Mol. Breed. 32: 841–852.
•  Longhi,  S.,  M. T.  Hamblin,  L.  Trainotti,  C. P.  Peace,  R.  Velasco and  F.  Costa. 2013b. A candidate gene based approach validates Md-PG1 as the main responsible for a QTL impacting fruit texture in apple (Malus × domestica Borkh). BMC Plant Biol. 13: 37.
•  Longhi,  S.,  M.  Moretto,  R.  Viola,  R.  Velasco and  F.  Costa. 2012. Comprehensive QTL mapping survey dissects the complex fruit texture physiology in apple (Malus × domestica Borkh.). J. Exp. Bot. 63: 1107–1121.
•  Moriya,  S.,  H.  Iwanami,  K.  Okada,  T.  Yamamoto and  K.  Abe. 2011. A practical method for apple cultivar identification and parent-offspring analysis using simple sequence repeat markers. Euphytica 177: 135–150.
•  Motomura,  Y.,  J.  Takahashi and  K.  Nara. 2000. Quantitative measurement of mealiness in apple flesh. Bull. Fac. Agric. Life Sci. Hirosaki Univ. 3: 23–28.
•  Myles,  S.,  J.  Peiffer,  P. J.  Brown,  E. S.  Ersoz,  Z.  Zhang,  D. E.  Costich and  E. S.  Buckler. 2009. Association mapping: critical considerations shift from genotyping to experimental design. Plant Cell 21: 2194–2202.
•  Ng,  J. K. T.,  R.  Schroder,  D. A.  Brummell,  P. W.  Sutherland,  I. C.  Hallett,  B. G.  Smith,  L. D.  Melton and  J. W.  Johnston. 2015. Lower cell wall pectin solubilisation and galactose loss during early fruit development in apple (Malus × domestica) cultivar ‘Scifresh’ are associated with slower softening rate. J. Plant Physiol. 176: 129–137.
•  Nobile,  P. M.,  F.  Wattebled,  V.  Quecini,  C. L.  Girardi,  M.  Lormeau and  F.  Laurens. 2011. Identification of a novel alpha-L-arabinofuranosidase gene associated with mealiness in apple. J. Exp. Bot. 62: 4309–4321.
•  Nordborg,  M. and  S.  Tavaré. 2002. Linkage disequlibrium: what history has to tell us. Trends Genet. 18: 83–90.
•  Nybom,  H.,  M. A.  Afzadi,  J.  Sehic and  M.  Hertog. 2013. DNA marker-assisted evaluation of fruit firmness at harvest and post-harvest fruit softening in a diverse apple germplasm. Tree Genet. Genomes 9: 279–290.
•  Segonne,  S. M.,  M.  Bruneau,  J. M.  Celton,  S. L.  Gall,  M.  Francin-Allami,  M.  Juchaux,  F.  Laurens,  M.  Orsel and  J. P.  Renou. 2014. Multiscale investigation of mealiness in apple: an atypical role for a pectin methylesterase during fruit maturation. BMC Plant Biol. 14: 375.
•  Silfverberg-Dilworth,  E.,  C. L.  Matasci,  W. E.  Van de Weg,  M. P. W.  Van Kaauwen,  M.  Walser,  L. P.  Kodde,  V.  Soglio,  L.  Gianfranceschi,  C. E.  Durel,  F.  Costa,  T.  Yamamoto,  B.  Koller,  C.  Gessler and  A.  Patocchi. 2006. Microsatellite markers spanning the apple (Malus × domestica Borkh.) genome. Tree Genet. Genomes 2: 202–224.
•  Stow,  J. 1989. The involvement of calcium ions in maintenance of apple fruit tissue structure. J. Exp. Bot. 40: 1053–1057.
•  Symoneaux,  R.,  M. V.  Galmarini and  E.  Merinagic. 2012. Comment analysis of consumer’s likes and dislikes as an alternative tool to preference mapping. A case sudy on apples. Food Qual. Preference 24: 59–66.
•  Tu,  K.,  B.  Nicolai and  J.  De Baerdemaeker. 2000. Effects of relative humidity on apple quality under similated shelf temperature storage. Sci. Hortic. 85: 217–229.
• van Ooijen, J. W. 2009. MapQTL 6 Software for the mapping of quantitative trait loci in experimental populations of diploid species. Kyazma B.V., Wageningen, Netherlands.
•  Velasco,  R.,  A.  Zharkikh,  J.  Affourtit,  A.  Dhingra,  A.  Cestaro,  A.  Kalyanaraman,  P.  Fontana,  S. K.  Bhatnagar,  M.  Troggio,  D.  Pruss,  S.  Salvi,  M.  Pindo,  P.  Baldi,  S.  Castelletti,  M.  Cavaiuolo,  G.  Coppola,  F.  Costa,  V.  Cova,  A.  Dal Ri,  V.  Goremykin,  M.  Komjanc,  S.  Longhi,  P.  Magnago,  G.  Malacarne,  M.  Malnoy,  D.  Micheletti,  M.  Moretto,  M.  Perazzolli,  A.  Si-Ammour,  S.  Vezzulli,  E.  Zini,  G.  Eldredge,  L. M.  Fitzgerald,  N.  Gutin,  J.  Lanchbury,  T.  Macalma,  J. T.  Mitchell,  J.  Reid,  B.  Wardell,  C.  Kodira,  Z.  Chen,  B.  Desany,  F.  Niazi,  M.  Palmer,  T.  Koepke,  D.  Jiwan,  S.  Schaeffer,  V.  Krishnan,  C.  Wu,  V. T.  Chu,  S. T.  King,  J.  Vick,  Q.  Tao,  A.  Mraz,  A.  Stormo,  K.  Stormo,  R.  Bogden,  D.  Ederle,  A.  Stella,  A.  Vecchietti,  M. M.  Kater,  S.  Masiero,  P.  Lasserre,  Y.  Lespinasse,  A. C.  Allan,  V.  Bus,  D.  Chagne,  R. N.  Crowhurst,  A. P.  Gleave,  E.  Lavezzo,  J. A.  Fawcett,  S.  Proost,  P.  Rouze,  L.  Sterck,  S.  Toppo,  B.  Lazzari,  R. P.  Hellens,  C. E.  Durel,  A.  Gutin,  R. E.  Bumgarner,  S. E.  Gardiner,  M.  Skolnick,  M.  Egholm,  Y.  Van de Peer,  F.  Salamini and  R.  Viola. 2010. The genome of the domesticated apple (Malus × domestica Borkh.). Nat. Genet. 42: 833–839.
•  Wakasa,  Y.,  H.  Kudo,  R.  Ishikawa,  S.  Akada,  M.  Senda,  M.  Niizeki and  T.  Harada. 2006. Low expression of an endopolygalacturonase gene in apple fruit with long-term storage potential. Postharvest Biol. Technol. 39: 193–198.
•  Wei,  J.,  F.  Ma,  S.  Shi,  X.  Qi,  X.  Zhu and  J.  Yuan. 2010. Changes and postharvest regulation of activity and gene expression of enzymes relaed to cell wall degradation in ripning apple fruit. Postharvest Biol. Technol. 56: 147–154.
•  Yu,  J.,  G.  Pressoir,  W. H.  Briggs,  I.  Vroh Bi,  M.  Yamasaki,  J. F.  Doebley,  M. D.  McMullen,  B. S.  Gaut,  D. M.  Nielsen,  J. B.  Holland,  S.  Kresovich and  E. S.  Buckler. 2006. A unified mixed-model method for association mapping that accounts for multiple levels of relatedness. Nat. Genet. 38: 203–208.