SNP identification by GBS analysis
The analysis of GBS data in conjunction with a well-established reference genome is a relatively straightforward route for SNP calling and marker ordering along chromosomes . In this work, most of the sequence tags (84.4%) could be uniquely aligned to the tomato reference genome. This was expected because complete genome assemblies of tomato are available.
In the present study, although 23,677 SNPs were discovered by comparison of the S. lycopersicum and S. pimpinellifolium genome sequences, most of the SNPs (80%) were excluded due to a high proportion of missing data. Nevertheless, a sufficient number of validated SNP markers (3,125 SNPs) remained and were found to be useful for QTL mapping. In comparison with previous work, this study discovered fewer validated SNPs than are available in The Solanaceae Genomics Network database (9,226 SNPs) and those reported by Chen et al.  (4,697 SNPs). The current number of SNPs is also much lower than discovered by Causse et al.  (16,000 SNPs) and Kim et al.  (4,680,647); however, these SNPs were not validated. The average frequency of SNPs identified in this study was 1 SNP per 256.4 kb, much higher than reported for the S. lycopersicum × S. pimpinellifolium linkage maps of Salinan et al.  (1 SNP per 8,482 kb), Capel et al.  (1 SNP per 4,077 kb) and Chen et al.  (1 SNP per 1,821 kb). Thus, the present research demonstrated that the GBS approach was efficient in constructing a SNP-based physical map of sufficient resolution for QTL mapping in tomato.
The IBL population and parental genotypes were evaluated for 11 fruit quality traits in order to identify associated QTLs. Sizable variation for all traits except soluble solids content and pH, and normal continuous distribution of all but three traits (external color, locule number and fruit shape) were observed in the IBL population. External color and locule number tended to skew toward more intense red color and higher locule numbers due to the unbalanced nature of the IBL population which favors the recurrent parent genotype. The parents of the IBL population had extreme alleles for fruit weight, wall thickness, stem scar and soluble solids content traits (Table 2). Although the parental alleles for soluble solids content were extreme, low variation was observed in the IBL population for the trait. This finding implies an unbalanced introgression of S. pimpinellifolium alleles for soluble solids content into the S. lycopersicum genome.
The present study demonstrated correlations between fruit quality traits, however, most of the significant correlations were weak. Correlations between fruit weight and all traits except fruit shape, internal color, dry matter weight and pH demonstrated that fruit weight was associated with fruit quality traits such as locule number, wall thickness, firmness and stem scar. Fruit weight had a high positive correlation with locule number. This is expected because increased locule number has a direct effect on fruit size and weight. Negative correlations of fruit weight with external color and soluble solids content indicate that intensity of external color decreases with increased fruit size due to decreased lycopene content and that sucrose content is negatively correlated with fruit volume. This negative correlation was also reported by Chen et al. , Doganlar et al. , Sun et al.  and Fulton et al. . Correlation results between fruit weight and quality traits were consistent with the results of Lippman and Tanksley , Okmen et al.  and Fulton et al. . A direct effect of soluble solids content on dry matter weight was observed in the IBL population. The positive correlation between internal color and external color was expected and consistent with previous reports [38–40]. These correlations can also be attributed to the pleiotropic effects of genes on different fruit quality traits.
Fruit quality parameters are important agronomic traits that increase the market value of both fresh market and processing tomatoes. Thus, there are many reports on QTL identification for fruit quality traits. All previous QTL mapping studies were performed using low density linkage maps constructed with PCR-based markers (SSR and COSII) and RFLP probes. Various parental lines and mapping populations such as BC2F2, IBL and RIL were used in these previous studies. This is the first study in which QTLs for fruit quality traits were identified by constructing a high density SNP-based physical map using a recently developed IBL population that carries introgressions from the S. pimpinellifolium genome. The physical map of SNP markers was useful for QTL mapping as IBLs are unbalanced populations which are not suitable for linkage map construction.
Fruit weight is the focus of many studies because increased fruit weight has direct effects on tomato yield [1–3, 14, 26, 27]. Fruit size is also an important trait that directs consumer preferences. Medium and large tomatoes are usually preferred by consumers . In this work, three QTLs were identified on chromosomes T2, T4 and T6 for fruit weight. Previous studies identified three major and two minor QTLs on chromosomes T1, T2, T3, T7 and T11. Although QTL locations varied among these studies, all studies identified a QTL with major effect on chromosome T2 corresponding to a cloned gene that controls fruit weight (fw2.2) . In the present study, the fruit weight QTL on chromosome T2 explained a variance of 15% for the trait, a value which is relatively low when compared with the same QTL in other studies (PVEs ranged between 15 and 40%). Differences in QTL magnitudes of effect and locations are most likely due to differences in population type used in the studies. The present work is most similar to the work of Doganlar et al.  which also studied an IBL population, which used a processing tomato as the recurrent parent. PVE of the QTL on chromosome T2 was the same as that reported by Doganlar et al.  (15%) due to the similarity of the genetic structures of the populations (IBL) used in the two studies. Identification of previously undetected QTLs on chromosomes T4 and T6 in the present work can be attributed to variation in the genetic backgrounds of the two mapping populations which is due to the use of different recurrent parents.
Because dried tomatoes have a high economic value, fruit dry matter weight can be as important as fruit weight. A previous QTL mapping study performed by Saliba-Colombani et al.  identified QTLs (with PVEs ranging from 9 to 25%) on chromosomes T2, T4 and T9 in a RIL population developed from the cross between a cherry tomato cultivar and S. lycopersicum. In other work, QTLs were identified on chromosomes T8, T10, T11 and T12 using 20 introgression lines carrying S. chmielewski introgressions in a S. lycopersicum genetic background . In the present study, none of the above mentioned QTLs were detected. This result can be due to insufficient variation for dry matter weight between the parents and the moderate coefficient of variation detected for the trait in the mapping population. PVEs of identified QTLs ranged from 14 to 19%, suggesting that in contrast to fruit weight, dry matter weight is not controlled by major effect QTLs.
In the present study, while a total of nine QTLs were identified for internal color, only two loci were identified for external color. The low number of QTLs identified for external color might be due to the unbalanced segregation of the trait in the IBL population. Previous work detected QTLs for external color on chromosomes T1, T3, T4, T7, T8, T9, T11 and T12 [38.43]. Although a QTL was also identified on chromosome T1 in this work, the physical position of the closest marker (C2_At5g13030: 1.1 Mb) to the locus on the same chromosome by Okmen et al.  reveals that the two QTLs are not identical. For internal color, previous studies identified QTLs on chromosomes T1, T3, T4, T7, T8, T9 and T12 with PVEs that ranged between 5 to 30% [7, 38]. In the present work, QTLs for internal color were identified on chromosomes T4, T7 and T8. The physical positions of the markers (65.4 Mb, 55 Mb and 58.1 Mb for At1g47830, T0671 and TG307, respectively) linked to the three QTLs indicated that they do not overlap with the QTLs identified in previous work.
Previous studies showed that locule number is controlled by six QTLs on chromosomes T2, T3, T4, T7, T10 and T12 [38, 43]. In addition, a major gene for locule number was mapped at the 48.1 Mb position on chromosome T2 . The major QTL (ln2.1) containing this single gene (lc) was also identified in the present study (PVE of 30%). In addition to this major QTL, a new QTL with minor effect was identified on chromosome T4.
Wall thickness and firmness are important fruit quality traits that define the shelf life of tomatoes. QTLs with minor effects on wall thickness were reported on chromosomes T6, T8, T11 and T12 , however, these loci do not overlap with those reported in the present work. Previously, QTLs for firmness were identified on chromosomes T1, T2, T3, T4, T5, T8 and T10 [3, 38]. In addition to these previously identified QTLs, four new QTLs were identified for firmness trait in this work.
Fruit shape and stem scar are appearance traits analysed in this study. Globular fruits with small stem scar are favoured in the market. More than 10 QTLs for fruit shape were identified in previous studies [3, 28, 38]. In addition to these QTLs, a total of four new QTLs were identified in this work with minor effects on fruit shape. For stem scar, seven QTLs were previously identified in tomato [2, 3, 38]. One of the two QTLs identified in this study for the stem scar was previously reported at 65.5 Mb position on chromosome T7 with a low PVE of 8% .
Soluble solids content and pH are important traits for fresh market tomatoes as they help define flavor . A total of five QTLs were detected on chromosomes T1, T6, T8 and T9 in previous studies for soluble solids content [1, 3, 14, 26] The present report demonstrated that different QTLs (chromosome T1, T2, T8 and T10) control soluble solids content in fresh market tomatoes. For pH, a total of six QTLs were identified in tomato on chromosomes T1, T2, T4, T5, T9 and T12 in previous studies [1, 14, 26]. While the position of the previously identified QTL on chromosome T1  was at 86 Mb, the major effect QTL (47%) identified on the same chromosome in this work was positioned at 66.8 Mb. Thus, the QTL identified in this study is close to the QTL previously identified by Chen et al. . These two QTLs might actually overlap because the SNP-based map of the present study has much higher resolution than the linkage map of Chen et al. .
Some QTLs colocalized as expected. For example, QTLs for locule number coincided with those for fruit weight and fruit shape because increased locule number results in larger tomatoes. However, colocalization of a QTL for pH with one for external color and colocalization of loci for wall thickness and soluble solids content were unexpected. These unexpected colocalizations might be due to linkage of the genes that control the traits .
This present study confirmed the high breeding potential of S. pimpinellifolium by detecting useful alleles for breeding of fruit quality traits such as fruit weight, external and internal color, firmness, soluble solids content and stem scar. The findings were expected for external color and soluble solids content because S. pimpinellifolium had higher values than cultivated tomato. In contrast, although S. pimpinellifolium had lower values than S. lycopersicum for fruit weight, internal color and stem scar, favorable S. pimpinellifolium alleles were detected for these traits. This result was consistent with the work of Top et al. . In that study, although S. pimpinellifolium had lower values than cultivated tomato for fruit weight and firmness, some individuals from an IBL (BC2F9) population derived from the cross S. lycopersicum and S. pimpinellifolium (LA1589) showed higher values than S. lycopersicum (TA209).