Introduction

Fruit shape is an important external quality trait for fresh market apples and, thus, is a priority breeding objective (Hazbavi 2014). However, little is known about quantitative trait loci (QTL) and candidate genes for fruit shape index (FSI), which limits the efficiency of selecting elite hybrid genotypes with positive fruit quality attributes.

With the development of apple linkage maps, many QTLs have been identified and associated with a wide range of quantitative traits, including disease and pest resistance (Gardiner et al. 2007; Durel et al. 2009; Wöhner et al. 2014), growth and tree habit (Kenis et al. 2008; Stoeckli et al. 2008), and especially fruit quality (Dunemann et al. 2009; Sarah et al. 2014; Kunihisa et al. 2014). Apple fruit shape, including height, diameter, and FSI, is of strong interest in QTL mapping analysis, but there are few reports on QTLs for FSI in apple. Using 250 F1 individuals of a ‘Telamon’ × ‘Braeburn’ segregating population, QTLs for fruit height were mapped on linkage groups (LG) 2, 6, 15, and 17; and QTLs for fruit diameter were mapped on LGs 2, 5, 9, 10, and 17, but the QTLs varied between two sampling years (Kenis et al. 2008). Using another hybrid population (Co-op 17 × Co-op 16), two QTLs for fruit diameter were identified on LGs 5 and 8, and three QTLs for fruit length were detected on LGs 3, 5, and 17 (Sarah et al. 2014). In ‘Jonathan’ × ‘Golden Delicious’, eight QTLs for fruit length and 11 for diameter were identified in 2 years, with most of these QTLs located on LGs 5, 8, 11, and 15 (Chang et al. 2014). These results clearly show inconsistency in QTL localisation due to differences in mapping populations and environmental factors.

Fruit shape in tomato is regulated by several major gene loci. One major gene, ovate, was detected on chromosome 2 of the tomato genome and was cloned and characterised as encoding a new class of nucleus-localised putative regulatory proteins (Liu et al. 2002). Although allelic variation in both ovate and another gene, sun, caused elongated fruit shape, the two loci differed in their genetic and morphological effects in tomato. For example, sun causes uniform elongation in both longitudinal directions, whereas ovate usually causes asymmetric elongation (van der Knaap and Tanksley 2001). Locule number, which has a pleiotropic effect on fruit shape and size, is controlled by the fasciated (fas) and locule-number (lc) loci. Fas encodes a transcription factor, and down-regulation of fas is caused by a large insertion in the first intron, which results in fruits with high locule number (Cong et al. 2002). The molecular nature of lc indicates that two single-nucleotide polymorphisms (SNPs) are located approximately 1200 bp downstream of the stop codon of a gene encoding a WUSCHEL homeodomain protein (Gustavo et al. 2011). In general, sun, and ovate control elongated shape, whereas fas and lc control fruit locule number and flat shape in tomato. In apple, the expression of two genes associated with cell proliferation, MdANT1 (MDP0000175309) and MdANT2 (MDP0000190889), was high from anthesis until 15 days after full bloom, a period coinciding with active cell division and rapid longitudinal fruit growth (Madhumita and Anish 2012). We previously identified five major gene loci involved in the regulation of FSI on four LGs (10, 11, 12, and 13) in a ‘Jonathan’ × ‘Golden Delicious’ mapping population (Sun et al. 2012). Recently, using the same population and simple sequence repeat (SSR) markers, the year-stable QTL for FSI, fsij08.11.2/fsij09.11 at 7.371 cM on chromosome 11 of the female parent, ‘Jonathan’, co-localised with a major gene locus, F11-1 (Chang et al. 2014). The consistency between QTL analysis using SSRs and major gene mapping puts fsij08.11.2/fsij09.11 at the forefront of candidate gene mining. However, because of the large coverage range of these QTLs (5.2 cM for fsij08.11.2; 5.4 cM for fsij09.11, 0.5 Mb/cM density in LG 11), exploration of QTL-based candidate genes cannot take place until fine mapping is performed.

Here, using the mapping population (‘Jonathan’ × ‘Golden Delicious’) reported previously by Chang et al. (2014) and a super-high-density genetic linkage map with SNP markers derived by restriction-associated DNA sequencing (RADseq), QTLs associated with FSI, fruit length, and fruit diameter were identified, and candidate genes were predicted.

Materials and methods

Plant materials

We used a hybrid population of ‘Jonathan’ × ‘Golden Delicious’ that consisted of 1733 seedlings (Sun et al. 2012; Chang et al. 2014). After preliminary selection for breeding in 2011, scions of all seedlings in the population were grafted onto Malus micromalus rootstock with a 30 cm ‘SH40’ dwarfing interstem and replanted at China Agricultural University (Beijing) for further genetic mapping. In total, 318 seedlings were randomly selected for RAD sequencing.

Phenotyping

The phenotypic characteristics of fruit length, fruit diameter, and FSI (ratio between height and diameter at the widest part of the fruit) were evaluated for 5 years. From 2008 to 2011, ripening fruit were harvested from the seedling populations. In 2014, the fruit were sampled from the grafted trees. Using a vernier caliper, fruit length and diameter were measured and the average values of at least five apples per tree, picked from mid-canopy, were calculated for further analysis. Differences in sample size over the 5 years were due to year-to-year variations in fruit bearing. From 2008 to 2011, only 152, 113, 21, and 66 genotypes had sufficient fruit for sampling in the respective years, and only 122 genotypes had sufficient fruit in 2014. No fruit were picked in 2012 and 2013 because the grafted plants had not yet matured.

QTL analysis

QTL detection was carried out using the MapQTL 6.0 software package (Van Ooijen 2009). Interval mapping was performed for each trait. Genome-wide limit of detection (LOD) thresholds for QTL significance were calculated by performing 1000 iterations with MapQTL’s permutation test. QTLs with LOD values higher than the genome-wide threshold (LOD > 3.50) were considered significant at P < 0.05. The linkage maps and QTL positions were drawn using MapChart (Voorrips 2002). However, a 99 % confidence interval (LOD > 4.10) was obtained using a two-LOD support interval (Van Ooijen 2009).

Prediction of candidate genes

Resequencing data for Jonathan and Golden Delicious and their SNPs were developed by our laboratory (unpublished). The apple genome reference scaffolds and predicted transcripts were downloaded from the genome database for Rosaceae (GDR) (http://www.rosaceae.org/). The minimum coverage for paired reads to call SNPs was in the range of 3–100. Only SNPs within QTL intervals with a frequency threshold of >80 % were filtered. According to the classification of SNPs in QTL intervals, annotation of all genes with nonsynonymous SNPs was performed in reference to the apple genome. After that, gene ontology (GO) annotation was performed by using the online WEGO software (Ye et al. 2006).

Gene expression analysis of candidate genes by RNAseq

In our study, the candidate gene expression data were cited, but unpublished RNA-seq results from our laboratory. ‘Greensleeves’, which has a similar fruit development period to ‘Golden Delicious’, were sampled at three stages (about 18, 67, and 132 days after anthesis, corresponding to the middle stage of cell division, cell expansion and the mature fruit stage, respectively.). Total RNA was extracted from the fruit samples and sequenced in the Genomics Facility, Institute of Biotechnology, Cornell University using Illumina HiSeq™ 2000 (San Diego, CA, USA).

Verification of SNPs in candidate genes and their segregation in hybrids

Primers were designed using Primer-blast software (NCBI, Maryland, USA) based on the sequences around SNPs in candidate genes. Then, PCR products comprising SNP loci of interest were amplified in both parents and sequenced using Sanger sequencing (BGI, Beijing, China) to verify SNPs obtained from resequencing data between parents. The causal SNPs were genotyped in 30 seedlings with high FSI and in 30 seedlings with low FSI values. The sequencing results were assessed using analysis of variance (ANOVA) to detect positive relationships between SNPs and phenotype.

Results

Changes in population phenotype over time

To estimate year-to-year variation in fruit shape traits, correlation coefficients for fruit length, fruit diameter, and FSI traits between years were calculated using 3 years of sampling data from 43 ever-bearing seedlings (Table 1).

Table 1 The correlation analysis of fruit shape traits in different years
Full size table

Significant correlations between fruit shape traits and year were found for 2008 and 2009. No correlation existed for fruit length and year between 2008 and 2014, or between 2009 and 2014, which reflected strong environmental variation between 2008 or 2009 and 2014. Fruit diameter in 2008 was correlated with that in 2014, but this relationship was not apparent for 2009 versus 2014. Stronger correlations were also detected for FSI between any two of the three years, indicating better year-to-year stability of FSI phenotypes. From these results, we conclude that fruit length is more sensitive to environmental conditions than fruit diameter or FSI.

QTL analysis

QTL analysis was performed using phenotypic data on fruit length, diameter, and FSI collected in 2008, 2009, 2010, 2011, and 2014 based on a high-density SNP linkage map. Fifteen QTLs were identified at the genome-wide threshold for QTL significance (LOD > 3.50) and explained genotypic variation ranging from 10.2 to 23.0 % (Table 2). The highest values in the LOD profile in 2010 and 2011, respectively, were 1.76 and 1.40 for fruit length, 2.50 and 2.94 for fruit diameter, and 1.71 and 2.50 for FSI. Therefore, no significant QTLs for the fruit shape traits were identified in 2010 and 2011 because there were fewer available individuals (Table 2).

Table 2 Summary data for QTLs of fruit length, diameter, and FSI using an interval mapping model of the ‘Jonathan’ × ‘Golden Delicious’ population
Full size table

Of the 15 identified QTLs, four were associated with fruit length, one with diameter, and 10 with FSI. Surprisingly, all of the QTLs were mapped to ‘Jonathan’ except for one, flg14.1, which was associated with fruit length and derived from ‘Golden Delicious’ in 2014. A year-stable QTL cluster associated with FSI was mapped to LG 11 of the maternal parent, ‘Jonathan’ (Table 2). Among the 10 QTLs for FSI, two overlapping regions were detected in 3 years. The first (designated ‘Locus 1’) was overlapped by fsij08.11.3, fsij09.11.1, and fsij14.11.2 in the region from 9.60 to 17.90 cM in ‘Jonathan’, while the second (designated ‘Locus 2’) spanned the region from 22.21 to 22.46 cM of LG11 and was overlapped by fsij08.11.4, fsij09.11.2, and fsij14.11.3 (Fig. 1). Four QTLs associated with fruit length were distributed on LGs G1, J7, and J11 in 3 years without overlapping regions. Only one QTL associated with fruit diameter was detected, in LG 3 of ‘Jonathan’ in 2008, and it explained a small amount of variance for this trait (12.0 %).

Fig. 1
figure 1

Interval mapping of QTLs for FSI at a whole-genome LOD threshold ≥4.10 using the ‘Jonathan’ linkage map for the upper region of linkage group 11. Only the upper region of linkage group 11, which was strongly associated with FSI trait, is shown on the left of the figure. QTLs for FSI are shown to the right of the linkage group. Marker positions are listed to the left side of the linkage group. The LOD profile plots for simple interval mapping of FSI in 2008 (red line), 2009 (green line), and 2014 (blue line) are shown to the right of the linkage group. (Color figure online)

Full size image

Delineation of FSI QTLs for sequence variation analyses

Locus 1 and Locus 2 were easily anchored to the genome via the pre-anchored SNP markers at 4411077–6723403 nt and 7408490–7474945 nt, respectively, on chromosome 11 of ‘Jonathan’. Then, reads in the database of whole-genome resequencing of the two parents (unpublished data) were aligned to the Malus × domestica reference genome sequence (Whole Genome ver. 1.0p, Velasco et al. 2010) to call SNPs in Locus 1 and Locus 2 between the parents. In ‘Jonathan’, the resequencing depth of Locus 1 and Locus 2 was 27.4× and 39.3×, respectively, and the corresponding values in ‘Golden Delicious’ were 28.5× and 24.2× (Table 3). Because both QTL regions for FSI trait were from the female parent, heterozygous SNPs in ‘Jonathan’ that could be segregated in hybrids were selected for further analysis. A total of 11,028 and 716 SNPs were detected in Locus 1 and Locus 2, respectively. In Locus 1, 2076 of these SNPs were distributed within exon regions and 913 and were nonsynonymous substitutions. In Locus 2, 91 were in exon regions and 61 were nonsynonymous. These nonsynonymous substitutions putatively affected 187 genes within Locus 1 and 12 genes within Locus 2.

Table 3 Female parent-specific SNPs within the two FSI QTL intervals on chromosome 11
Full size table

Based on GO annotations in the apple genome, the 199 genes in Locus 1 and Locus 2 could be divided into nine functional groups (Fig. 2). The detailed results of GO annotation can be seen in Additional file 1 after uploading Additional file 2 to WEGO (http://wego.genomics.org.cn/). Genes whose functions included pigmentation, response to stimulus, localisation, and establishment of localisation, were not considered candidate genes for the FSI trait. Seventy-five genes involved in metabolic process, 58 genes involved in cellular processes, one gene for cellular component organization, one gene for cell death, and 10 involved in biological regulation were prioritised. Genes involved in alcohol or nitrogen compounds, lipids, pigments, organic acids, nucleotides, vitamin metabolic processes, photosynthesis, cellular macromolecule metabolic processes, and oxidation reduction were also excluded from the grouping. Among the remaining genes, MDP0000135244 encoding peptidoglycan-binding LysM domain-containing protein was of particular interest because it involved in cell wall macromolecule metabolism. In addition, one gene without GO annotation, MDP0000896238, captured our attention because of its high similarity with Arabidopsis thaliana ovate family protein 17 (AtOFP17, Wang et al. 2011). AtOFP family members function as transcription repressors, and they regulate multiple aspects of plant growth and development. However, AtOFP17 was only highly expressed in mature roots, but not in flowers, siliques and leaves.

Fig. 2
figure 2

Histogram of gene ontology classifications for apple of all annotated genes in the Locus 1 and Locus 2 QTL intervals of FSI trait. The genes corresponded to three main categories: biological process, cellular component, and molecular function. The right-hand y-axis indicates the number of genes and left-hand y-axis indicates percent of genes from apple in a category

Full size image

In addition, candidate genes were predicted based on involvement in the fruit shape formation pathway in apple and tomato to locate them on the apple genome sequence using BLASTP analysis. The five genes most correlated with ovate (NP_001234221) were located on chromosomes 2 (MDP0000155311 and MDP0000243940), 8 (MDP0000296199), 15 (MDP0000140421), and the unanchored chromosome (MDP0000921224). The genes most correlated with fas (NP_001234390) were on chromosomes 11 (MDP0000331808), 2 (MDP0000255415), 10 (MDP0000252640), 5 (MDP0000366291), and 9 (MDP0000192940). However, the detailed physical position of MDP0000331808 was at the bottom of chromosome 11, which was not consistent with the results in this study. The genes most correlated with sun (NP_001233793) were on chromosomes 2 (MDP0000639465), 14 (MDP0000299240), unanchored (MDP0000226363), and 1 (MDP0000185385 and MDP0000238367). There were no correlated genes for lc (JF284938 and JF284939) in the apple genome. In Arabidopsis, ANT and AINTEGUMENTA-LIKE genes regulate floral growth and ovule development, respectively (Mizukami and Fischer 2000; Krizek 2009). In apple, two ANT genes, MdANT1 and MdANT2, were also reported to be involved in cell division during fruit growth (Madhumita and Anish 2012). One predicted apple protein (MDP0000182395, chromosome 11: 2860500–2863619), which also encoded AP2-like ethylene-responsive transcription factor (similar to the MdANT1 and MdANT2), was screened and positioned at the top of chromosome 11. Unfortunately, this gene was not located in the Locus 1 interval.

Expression of two candidate genes for FSI trait in apple fruit

The results of the expression analysis of two candidate genes, MDP0000135244 and MDP0000896238, by RNAseq are shown in Additional file 3. The expression of MDP0000135244 increased during fruit development, especially at the cell expansion stage, but not during cell division (Janssen et al. 2008). The expression of MDP0000896238 was almost negligible, similar to findings in A. thaliana (Wang et al. 2011) that suggested MDP0000896238 was unrelated to fruit shape development.

Verification of candidate genes in parental cultivars and their hybrids

The nonsynonymous SNP in the gene of interest, MDP0000135244, was validated between the parents using Sanger sequencing. The result showed that the SNP of MDP0000135244 in the female parent ‘Jonathan’ was reliable. This SNP (designated C11.6053728) was polymorphic in ‘Jonathan’ (T/G) and monomorphic (T/T) in ‘Golden Delicious’ (Fig. 3). Domain motif annotation of the full-length genomic DNA sequence of MDP0000135244 revealed that the SNP was located in the first predicted exon. Amino acid translation revealed a semi-conserved substitution from asparagine (codon AAT) to lysine (codon AAG).

Fig. 3
figure 3

SNP in MDP000013524 between ‘Jonathan’ and ‘Golden Delicious’ apples. (Color figure online)

Full size image

Then, C11.6053728 of MDP0000135244 was genotyped in 30 hybrids with high FSI (≥0.9) and in 30 hybrids with low FSI (≤0.8). The allele T/G was present in 23 of 30 high-FSI hybrids, and T/T was present in half of the low-FSI hybrids. The C11.6053728 genotypes were significantly correlated with FSI phenotypes, revealing a close association of C11.60537228 with FSI trait (Table 4).

Table 4 Genotype of nonsynonymous SNPs for MDP0000135244 in 60 hybrids
Full size table

Discussion

Effect of population and environment on detection of stable QTLs

Considering genotype × genotype and genotype × environment interactions, it is difficult to detect stable QTLs using different populations combined with changing environmental conditions among different planting locations and sampling years (Kenis et al. 2008). For example, QTLs for fruit length were located on LGs 2, 6, and 17 of ‘Telamon’ in a ‘Telamon’ × ‘Braeburn’ population in 1 year and on LGs 15 and 17 in the next year (Kenis et al. 2008). Sarah et al. (2014) detected QTLs for fruit length on LG 3 of the ‘Co-op 17’ linkage map in 2008 and on LGs 3 and 5 in 2009. In our study, fruit length was located on LG 7 of the ‘Jonathan’ linkage map in 2008 and on LG 11 in 2009. Fruit shape is a quantitative characteristic that is controlled by several genes (Sun et al. 2012). King et al. (2000) reported that QTL detection in a population is the result of single alleles derived from one parent, which must therefore be heterozygous for the locus underlying this trait. Therefore, differential segregation of alleles controlling fruit shape in parents from different populations is the primary reason stable QTLs were not detected. In addition, despite the identification of QTLs for fruit diameter in this study, fdj08.3 was mapped on LG 3 (5.4 Mb from the top of the chromosome), the same as the LG reported by Sarah et al. (2014). However, the two QTLs were far apart, indicating that two genes were included in those QTLs.

Environmental differences between years will lead to the detection of different QTLs, even in the same population. Here, QTLs for fruit length were mapped at different LGs in 3 years, coinciding with weak correlations for the trait between 2008 or 2009 and 2014 (Table 1). However, FSI has high heritability in apple (Sun et al. 2012) and a strong correlation among sampling years, as Diaz et al. (2014) reported for melon. Huang et al. (1997) suggested that stable QTLs of traits with high heritability are easy to detect under different environmental conditions. Our findings for QTLs of FSI support that conclusion.

Two stable QTLs of FSI with highly explained variance in different environment and population were detected for high efficiency molecular breeding

The variance explained by a QTL is a key factor affecting the value of that marker for practical use. In general, a marker is considered useful when it is obtained from ‘major’ QTLs that explain more than 20 % of population variance (Kenis et al. 2008). In addition, marker value is dependent on population size. Sarah et al. (2014) identified a main-effect QTL for fruit length in 2008 that explained up to 46.4 % of the variance with only 46 seedlings. The variance explained decreased to 20.1 % when the population size increased to 86 in 2009. Here, the QTL fsij09.11.1 explained 23.0 % of variance in 2009 and was detected using 112 seedlings, and fsij14.11.2 was detected in 2014 using 122 seedlings and explained 22.8 % of the variance. These values indicate that these two QTLs are major determinants of FSI.

A marker is highly efficient for marker-assisted selection (MAS) when it is stable across different genetic surroundings and years. For example, Md-ACS1-2 and Md-ACO1-1 are functional genetic markers associated with fruit texture when both are homozygous in a specific variety. These markers indicate that the variety has low ethylene production and high storability (Costa et al. 2005). Md-ACS1-2 and Md-ACO1-1 have been widely applied in apple breeding after the verification of their accuracy in 255 varieties (www.rosbreed.org). In our study, two year-stable loci associated with FSI (Locus 1 and Locus 2) were mapped to 4.4–6.7 Mb and 7.4–7.5 Mb, respectively, on chromosome 11 of ‘Jonathan’. Locus 2 was located near the previously reported QTL region (8.9–10.8 Mb on LG 11) using SSR linkage maps (Chang et al. 2014), and that location was also close to a major gene in a study using bulked segregant analysis with the same population (Sun et al. 2012). Although the present study did not detect significant QTLs of FSI trait in 2010 or 2011, a peak LOD was present at 5.8 cM of LG 11 of the ‘Jonathan’ linkage map (data not shown). Therefore, the QTLs of FSI trait identified here appear to have been stable between years and identification methods.

However, previously reported QTLs for apple shape mostly focused on fruit length and diameter, and no reports about QTLs for FSI trait are available for comparing the locations between our and other populations. We found one study that examined pear using SSR and amplified fragment length polymorphism (AFLP) markers, which reported an anchored SSR marker, NB105a, positioned on a linkage group with QTLs for pear FSI trait (Zhang et al. 2013). Surprisingly, we found that NB105a was located on LG 11 of the apple genome Celton et al. (2009). Because pear and apple chromosomes show high co-linearity, this result indicated that the QTLs identified by Zhang et al. (2013) should be located on LG 11 in pears. Thus, QTLs of FSI trait obtained in our study and confirmed by findings in pears reflect that a stable QTL region exists among different populations.

Combining high-density linkage maps with whole-genome resequencing is helpful for screening candidate genes

High-density linkage maps provide a useful tool for the efficient identification of candidate genes. In this study, Locus 2 for FSI was mapped to a 67-kb genomic region using high-density SNP linkage maps and was available for further candidate gene prediction. Locus 1 was located in a ~2.3 Mb region, an interval that was too long for direct prediction of the candidate gene. However, the nearest marker, euC11.4515793 at 4.5 Mb from the top of chromosome 11, supplied useful information about the physical location for indirect screening. Van der Knaap et al. (2013) also maintained that SNP markers were more efficient than traditional AFLP and SSR markers for locating QTLs.

Fine mapping was formerly an efficient approach for map-based candidate gene mining. However, the use of large numbers of seedlings limits the use of this technology for fruit crops. For example, the apple scab resistance gene Rvi12 was fine-mapped to an interval spanning 958 kb of the ‘Golden Delicious’ genome based on a mapping population of 1285 seedlings (Padmarasu et al. 2014). The construction of mapping populations in fruit is time-consuming and laborious. Therefore, narrowing the interval for Locus 1 in our study by increasing population size was not the best choice. Fortunately, Moriya et al. (2012) described a new method for QTL detection that is more efficient than traditional map-based cloning in fruit trees. Using 1000 F1 seedlings from 31 population crosses, the Co gene, which controls the columnar growth habit of apple, was fine-mapped to a 196 kb region of LG 10 (Huang et al. 2011). Their results illustrated that the use of multiple families for QTL mapping, which considers a high number of allelic variations, showed high resolution in QTL detection (Huang et al. 2011). The same approach has been reported for other fruit crops (Rosyara et al. 2013).

Combined with whole-genome re-sequencing data from parents or hybrids with various phenotypes, candidate genes within major QTLs could be explored without using large fine-mapping populations. Using whole-genome sequencing of four sibling trees from an F2 mapping population, candidate genes for chilling requirement traits were prioritised in peach (Zhebentyayeva et al. 2014). In our study, using whole genome resequencing data from the parental cultivars, nonsynonymous SNPs affecting 187 candidate genes within Locus 1 were predicted over a 2.3-Mb genomic region. This result provides a foundation for future research related to FSI trait. However, a disadvantage of this method is that only nonsynonymous SNPs in exons were considered. Several mutations in the promoter region, which is also involved in regulation of traits, were ignored by this bias. Therefore, the combined use of multiple families, high-density SNP maps, and resequencing information from accessions with various phenotypes will improve the efficiency and effectiveness of causal gene identification.

MDP0000135244 encoding an LysM domain receptor-like kinase protein was a key candidate gene related to FSI in apple

MDP0000135244 possesses a peptidoglycan-binding lysM domain. In previous years, this sequence motif was found in enzymes involved in cell wall degradation in bacteria (Bateman and Bycroft 2000), but whether it is involved in cell wall development in plants is unclear. Proteins including this motif in A. thaliana and Oryza sativa have antibacterial functions (Erbs et al. 2008; Shimizu et al. 2010). The protein with this motif in A. thaliana (AtLYK5) is the primary receptor for chitin and forms a chitin-inducible complex with CERK1 to induce plant immunity (Cao et al. 2014). It is well known that hypersensitive response (resulting in programmed cell death), callose deposition, and enlarged cells are general symptoms during immune responses (Rate and Greenberg 2001). That study may indicate that AtLYK5 is related to cell development, with the next step in that research being to determine how CERK1 and LYK5 are able to activate plant defenses. To identify the relationship between MDP0000135244 and the FSI trait in apple, the genotype of SNP C11.6053728 in the population was analyzed. The heterozygous genotype (T/G) frequency of SNP C11.6053728 in the 30 individuals with high FSI was 23/30, which showed good correlation between the T/G genotype and high-FSI phenotype. However, the T/T genotype frequency in the 30 individuals with low FSI was 15/30, suggesting that the heterozygous and homozygous genotypes should segregate randomly in the low-FSI group. This result was consistent with the expected segregation ratio of 1:1 in the low-FSI group because the trait was dominant for high FSI, and five segregating loci were involved in the variation in FSI (Sun et al. 2012). In general, the gene MDP0000135244 containing SNP C11.6053728 presumably is important in cell wall metabolism. In future work, a diagnostic marker designed based on these SNPs will be used to select elite hybrids for FSI trait in apple.