ddRAD-seq derived genome-wide SNPs, high density linkage map and QTLs for fruit quality traits in strawberry (Fragaria x ananassa)

Understanding the genetic determinants are essential for improving the fruit quality traits of strawberry. In this study, we focused on mapping the loci for fruit-length (FL), -diameter (FD), -weight (FW) and -soluble solid content (SSC) using the genome-wide single nucleotide polymorphisms (SNPs) identified via ddRAD-sequencing of the F1 population raised from Maehyang (♀) X Festival (♂). A total of 12,698 high quality SNPs were identified of which 1554 SNPs that showed significant Mendelian segregation (p < 0.05) were mapped to 53 linkage groups (LG) spanning a total of 2937.93 cM with an average marker density of 2.14 cM/locus. Six QTLs for FL and four QTLs for each of FD, FW and SSC were identified that explained 24–35%, 21–42%, 24–54% and 23–50% of overall phenotypic variations, respectively. The genes that lie within these QTL regions were extracted and discussed thoroughly. In addition, a high resolution melting marker (MF154) were designed based on the SNP A1723G of the UDP-glucose 4-epimerase GEPI48-like gene FAN_iscf00021287. The marker detected the high vs low sugar containing F1 plants and commercial cultivars with 81.39% and 86.95% detection accuracy, respectively. These SNPs, linkage map, QTLs and candidate genes will be helpful in understanding and improving the fruit quality traits of strawberry. Electronic supplementary material The online version of this article (10.1007/s13205-020-02291-5) contains supplementary material, which is available to authorized users.


Introduction
The cultivated strawberry (Fragaria × ananassa), an economically important berry crop, is favored across the globe due to its characteristic aesthetic property, flavor, nutritional value and health benefitting anti-oxidative content (Alvarez-Suarez et al. 2014;Henning et al. 2010). The increasing popularity of the fruit is evident from 21.21% and 57.1% increase in global cultivation area and production, respectively, during the last decade (FAOSTAT 2017). The increasing demand for the fruit necessitates the improvement of both agronomic and fruit quality attributes of the crop. Wider understanding of the genetic determinants of these traits are thus very important. However, compared to other economically important fruits, genetic control of key traits of cultivated strawberry is yet to be completely understood.
Fragaria × ananassa has undergone complex evolutionary changes during its evolutionary journey starting from wild diploid progenitors to the current cultivated octoploid (2n = 8 × = 56) species (Potter et al. 2000). Evidences from recent studies indicate three diploid progenitors namely, F. vesca, F. nubicola and F. orientalis; or F. orientalis, F. iinumae and F. vesca or F. mandshurica (Edger et al. 2019;Rousseau-Gueutin and Gaston 2009), which ultimately lead to the extant octoploid species with four relatively similar sub-genomic components (Rousseau-Gueutin and Gaston 2009;Sargent et al. 2016;Tennessen et al. 2014). The genome composition of the current Fragaria × ananassa is proposed to be either AABBBBCC, AAA′A′BBBB, or AAA′A′BBB′B′, with the latter being supported by most of the studies (Isobe et al. 2013;Kunihisa 2011;Sargent et al. 2012;Shaw 1997). The complexity of the genetics of this species thus can be attributed to the various combinations of gene interactions between the multiple alleles, complex meiotic behaviors and highly heterozygous nature of the species (Lerceteau-Köhler et al. 2003;Rousseau-Gueutin et al. 2008;Spigler et al. 2010;Weebadde et al. 2008).
The complex octoploid genomic structure made it difficult to disentangle the species genetically (Sánchez-Sevilla et al. 2015). However, the recent release of Fragaria × ananassa scaffolds in 2014 (Hirakawa et al. 2014) and the release of the complete genome of its diploid progenitor F. vesca (a simper model species) in 2011 (Shulaev et al. 2011) have facilitated the genetic mapping of economically important traits of cultivated strawberry. Genetic mapping of various important traits, particularly prior to the release of complete and annotated genome, can facilitate the identification of the causal loci (and specific genes, in some cases). For polyploid species, genetic mapping is also helpful in identifying the homoeoalleles and their contribution in governing the complex trait, for example, the identification of homoeoalleles for drought resistance in durum wheat (Peleg et al. 2009) and for cane yield in sugarcane (Aitken et al. 2006). Several homoeologous linkage groups have also been identified in the octoploid cultivated strawberry as well (Rousseau-Gueutin et al. 2008;Sargent et al. 2009).
The draft scaffold of Fragaria × ananassa have shown high collinearity with its diploid progenitor, F. vesca (Rousseau-Gueutin and Gaston 2009;Sargent et al. 2012;Hirakawa et al. 2014). This offers the opportunity to develop genome-wide molecular markers and to construct linkage maps with higher marker density. Moreover, the advancement in the cost-effective and readily available reduced genome sequencing technologies such as restriction site-associated DNA sequencing (RAD-Seq) and genotyping-by-sequencing (GBS) made the identification of the genome-wide SNPs, for this complex octoploid species, more convenient these days (Elshire et al. 2011;Baird et al. 2008). These genome-wide SNPs are helpful in constructing high-density linkage map, which can eventually lead to the identification of novel QTLs and putative functional candidate genes for both diploid and polyploid species (Bassil et al. 2015;Shirasawa et al. 2017). Although, few reports of SNP based linkage mapping in strawberry are available (Sánchez-Sevilla et al. 2015;Davik et al. 2015;Hossain et al. 2019;Vining et al. 2017), no SNP based QTL is reported yet, particularly for fruit quality traits.
In this study, we thus aimed at identifying genome-wide SNPs via ddRAD-seq technique, constructing high density linkage map and mapping QTLs for few fruit quality traits using the F 1 population raised by crossing two cultivars having contrasting agronomic and economic traits. The findings will enrich our current genome wide SNP knowledgebase for cultivated strawberry and will add new insight in understanding the genetics of the studied fruit quality traits.

Plant materials, growth conditions and phenotyping
Seedlings of high sugar cultivar, Maehyang (♀) and low sugar cultivar Festival (♂) were collected from Damyanggun Agricultural Technology Center, Damyang, South Korea. The F 1 population were raised in the green house facility of Sunchon National University, South Korea. The F 1 seeds were germinated in growth chambers and then transplanted to large (66 cm × 20 cm × 16 cm) rectangular pots filled with commercial nursery soil mixture (50% cocopit and 50% soil) in green house maintaining 25 ± 2 °C, 16 h day length, 80% relative humidity and 440 μmol m −2 s −1 light intensity. The length (mm) and diameter (mm) of fruits were recorded and the weights (g) were measured using OHAUS balance (model PAG2102, OHAUS Corporation, NJ, USA). Total soluble solid content (SSC) was measured in a refractometer (model PAL-1, ATAGO CO., LTD, Tokyo, Japan) and expressed in °Bx. The field experiment was performed with five replications. Measurements were taken from five uniformly ripe fruits, each (2nd fruit of the 1st fruit cluster) collected from an individual plant. In addition, 23 commercial culitvars (listed in Table S10) were grown and phenotyped for SSC in the green house facility of Damyang-gun Agricultural Technology Center, Damyang, South Korea. A °Bx value of '8' was considered as the cutoff value for high vs low sugar content. These cultivars were used for validating the markers developed for detecting high vs low fruit sugar containing genotypes.

Isolation of DNA
DNA was collected from the newly emerged young leaves of the parents and F 1 plants. Fresh leaves were disrupted in TissueLyser II (Qiagen, CA, USA) and DNA was extracted using the protocol of DNeasy Plant Mini Kit (Qiagen, CA, USA). The DNA quality was evaluated by electrophoresis in agarose gel (1.2%) while the concentration and the purity of extracted DNA samples were estimated using Nanodrop-2000 (Nanodrop Technologies, Wilmington, DE, USA).

Preparation of ddRAD-seq library and sequencing
The ddRAD-seq libraries for all strawberry genotypes were constructed by digesting the genomic DNA (500 ng) with restriction enzymes, PstI and MspI (New England Biolabs, MA, USA). Unique barcodes of eight nucleotide base pairs and Illumina adaptors were ligated to the digested DNA to enable tracing the samples of each genotype (Table S1).
The adaptor-ligated DNA fragments were amplified by 22 cycles of polymerase chain reaction and the amplicons of 300-900 bp length were separated using BluePippin (Sage Science, MA, USA). The extracted DNA fragments were then sequenced by the Illumina HiSeq 2000 platform (Illumina, Inc., CA, USA) using 93 base pair (bp) paired-end (PE) mode following the procedures described in Shirasawa et al. (2017).
The high quality reads were then mapped to the Fragaria × ananassa reference genome sequence (FAN_r1.1, https ://straw berry -garde n.kazus a.or.jp/) using Bowtie 2 v2.1.0 (parameters: -minins 100 -no-mixed) (Langmead and Salzberg 2012). The resulting SAM (sequence alignment/ map) format files were converted into BAM (binary alignment/map) files prior to sorting, indexing and removal of duplicates by SAMtools version 0.1.19 (parameters: -Duf) (Li et al. 2009). The variants from the high-quality alignments were called using mpileup module from SAM tools and BCF tools (parameters: -vcg) view. The resulting variant call format (VCF) files were filtered using VCFtools version 0.1.11 (Danecek et al. 2011) and the missing data were imputed using Beagle4 (Browning and Browning 2007). The SNP associated sequences were retrieved from strawberry reference genome (FAN_r1.1). SnpEff version 4.11 (snpeff. sourceforge.net) was used to predict the effect of SNP annotations on gene functions (Cingolani et al. 2012).

Construction of linkage map
The SNP markers that were heterozygous either in one parent (segregation genotype code lmxll and nnxnp in JoinMap version 4.1) or in both parents (segregation genotype code hkxhk) were used for linkage mapping. Markers that did not fit significantly with the respective segregation ratio of 1:1 or 1:2:1 in the progeny as per the chi-square test of goodness of fit (p < 0.05) were excluded from map construction. Markers showing more than 10% missing data were excluded as well.
Marker order and map distances (centi-Morgans) were calculated using regression mapping algorithm and Kosambi's mapping function (maximum recombination fraction = 0.45, goodness of fit jump threshold = 5 and a ripple value = 1) in JoinMap v4.1 (Ooijen 2006). Based on the minimum logarithm of odds (LOD) score limit of 5.0, linkage groups (LGs) were selected and visualized in MapChart (version 2.32). All LGs were mapped on the F. vesca genome v4.0.a1 and the LGs were numbered '1'-'7' based on the number of F. vesca chromosomes where the LGs were mapped. The suffixes 'A'-'D' were assigned to the LGs based on their sequential physical position along the length of a particular F. vesca chromosome. Two small LGs, if mapped one after another or partially overlapped, to a particular F. vesca chromosome, they were considered as one linkage group.

Identification of QTLs and extracting genes from QTL regions
QTL mapping was performed by integrating the genotypic and phenotypic data of the mapping population through composite interval mapping (CIM) in Windows QTL Cartographer tool version 2.5_011 (parameters: linkage map method = 10, segregation test size = 0.01, linkage test size = 0.35, mapping function = Kosambi, objective function = SAL and step size = 1 cM). Besides, several multilocus mixed linear models such as mrMLM, pLARmEB and ISIS EM-BLASSO methods were also used to validate the QTLs identified by CIM methods Wang et al. 2016;Wen et al. 2018;Zhang et al. 2017). Putative QTLs were declared based on the significant LOD threshold determined by 1000 permutations. The square of the partial correlation coefficient (R 2 ) was used to indicate the proportion of phenotypic variation explained by a QTL. Sequence tags associated with the flanking markers of each QTL were mapped to the Fragaria vesca annotated genome version v4.0.a1 and the genes within these flanking regions were extracted.

Development of High resolution melting (HRM) Marker
Five HRM markers were designed to validate their linkage with high vs low sugar contents. These markers were used to genotype the F 1 population and 23 commercial cultivars in a final reaction volume of 20 μL, containing 50 ng of genomic DNA, 10 μL of 'HS Prime LP Premix' (GeNet Bio, Deajeon, Republic of Korea), 0.6 μL of 2xSYTO9 green fluorescent nucleic acid stain (GeNet Bio, Deajeon, Republic of Korea), 0.2, 1.0 and 1.0 μL of forward, reverse and probe primers, respectively (Table 4), and ultra-pure water for the remainder of the volume. High resolution melting was performed in a LightCycler96 (Roche, Mannheim, Germany) using 96-well plate in a 20 μL/well final reaction mix with the cycling condition of initial denaturation at 95 °C for 5 min followed by 40 cycles of three step amplifications at 95 °C for 10 s, 60 °C for 15 s and 72 °C for 15 s. The HRM program included denaturation at 95 °C for 1 min, re-naturation at 40 °C for 2 min and melting from 60 to 90 °C, with a ramp of 0.3 °C per second and five fluorescent acquisitions per degree centigrade. High resolution melting data were analyzed using LightCycler ® 96 software v1.1.

Statistical analysis
Significance test was done using Analysis of variance (ANOVA) and statistically significant differences between cultivars for the fruit quality traits were determined by Students t test using Minitab ® 18 software package (Minitab Inc., State College, PA, USA).

ddRAD-seq based genotyping and SNP discovery
The genomic DNA of strawberry genotypes were double digested with PstI and MspI restriction enzymes to generate ddRAD-seq representation libraries which were subsequently sequenced using Illumina HiSeq 2000 platform. The paired-end sequencing of individuals yielded a total of 65,915,564 reads (total 10.31 GB sequenced data) with an average of 1,345,212 (~ 1.3 million) reads per accessions. A total of 26,655,685 high quality reads were retained and an average of 80.8% of total reads (which ranged from 77.6 to 82.6%) were aligned against Fragaria × ananassa genome version FAN_r1.1 (Fig. 1a). The detailed statistics of raw-, cleaned-, and mapped-reads, and genotype-wise alignment ratio are summarized in Table S1. Finally, based on strict SNP filtration criteria (see materials and methods section), a total of 12,698 high quality SNPs were retained from the mapped reads.

Distribution patterns of identified SNPs
Variable distribution of different types of SNPs were observed (Fig. 1b). Among the 12,698 high quality SNP genotypes, the highest proportion of The SNP variant was C/T (1941) followed by G/A (1914), A/G (1787) and T/C (1765). The lowest proportion of SNP genotype was G/C (508) followed by A/C and C/A (642 and 678, respectively) ( Fig. 1b). Out of 12,698 SNP genotypes, 3626 (28.5%) and 2790 (21.9%) SNPs were classified as transitions (A/G or C/T) and transversions (G/T, A/C, A/T, or C/G), respectively (Fig. 1c). In general, occurrence of nucleotide transition type SNPs were higher than transversion types (with a transition/ tranversion (TS/TV) ratio of 1.30). This could be due to the interchange between purine and pyrimidine nucleotide bases. SnpEff tool version 4.11 was used to annotate the high quality SNPs based on their functional classes ( Fig. 1d) which revealed the majority of SNPs (6094; 47.99%) are exonic, followed by 3245 (25.56%) intergenic SNPs. A total of 2630 (20.71%) and 2420 (19.06%) SNPs were located in 1 kb downstream and upstream regions, respectively, while 1844 (14.52%) SNPs were located in intronic regions (Fig. 1d).

Construction of high density linkage map
Among 12,698 high quality SNPs, 4638 fitted our selection criteria of polymorphism between the parents i.e., presence of heterozygous loci at least in 1 parent. Out of 4638 markers, 207 markers that had a minimum of 10% missing data in the population and 1165 markers that showed significant distortion from expected ratio of Mendelian segregation were excluded in the subsequent analysis. From the 3266 markers showing significant Mendelian segregation, 1332 pairs (consisting of 1139 unique markers) showed exactly identical patterns of segregation (Table S2). Finally, 53 linkage groups were selected based on the minimum LOD threshold of 5.0 where 1554 SNP markers were mapped that spanned a total genetic distance of 2937.93 cM representing all 28 chromosomes of Fragaria × ananassa (Table 1; Fig. 2). The markers that were not mapped within these 53 LGs were either mapped in smaller groups or not linked to any of the recognized LGs or showed conflicting segregation pattern with other markers of the same linkage group at the selected LOD threshold. The average length of the linkage groups was 56.12 cM that ranged between 19.89 cM (LG7D2) and 120.27 cM (LG6C1). A minimum of eight and a maximum of 90 markers were mapped in linkage group LG7A2    Table 1 and Tables S3;  and the details of the QTLs are shown in Table 2. Underlined markers indicate markers located at peak LOD position (24.62 cM) and LG5B (94.5 cM), respectively; with an average of 29.30 markers per linkage group. The map resolution corresponded to an average marker density of 2.14 cM per locus with the longest marker interval of 18.61 cM being observed in LG1A2. The sequence tags corresponding to the SNP markers were mapped on the F. vesca genome v4.0.a1 (one of the four sub-genomes of the cultivated octoploid Fragaria × ananassa), which revealed a total coverage of 231.27 Mbp (96.36%) of the diploid F. vesca genome (240 Mb). In addition, the map distances of the SNP markers and their corresponding physical positions on the F. vesca chromosomes showed high degree of collinearity, except for few less dense and shorter linkage groups such as LG1A1, LG1B2, LG1C2, LG2A1, LG2A2 and LG7A2 (Fig. S1). The details of the linkage groups, marker density and marker distance along with their corresponding physical positions on F. vesca genome are shown in Table 1 and Table S3. The SNP positions and SNP alleles in all genotypes along with their corresponding gene IDs are shown in Table S4.

Phenotypic assessment
Significant (p < 0.05) differences were observed for the length, diameter, weight and soluble sugar content of the fruits of two parental cultivars, 'Maehyang' and 'Festival' (Fig. 3). The traits FL, FD and FW of 'Festival' were higher (Fig. 3) while the SSC of this cultivar (6.17°Bx) was significantly lower compared to that of 'Maehyang' (10.13°Bx). The crossing population raised from these two parents showed normal distribution of the traits with FL ranging from 33.81 to 55.60 mm, FD ranging from Fig. 3 Fruit phenotypes (a), differences in length, diameter, weight and sugar content of fruits (b) of high-and low-sugar containing cultivars, Maehyang (♀) and Festival (♂), and their frequency distribution in F 1 population (c). Phenotypic data in parents are presented as mean ± SE (n = 5). Asterisks * and ** represent significant differences between the parental cultivars at p < 0.05 and p < 0.01, respectively, by Student's t test 26.36 to 70.14 mm, FW ranging from 10.45 to 33.75 g and SSC ranging from 4.90 to 13.10°Bx (Fig. 3). Among these, extreme phenotypes in F 1 plants, indicating transgressive segregation, is observed for FD, FW and SSC. Correlation analysis of the strawberry fruit quality traits among the F 1 progeny indicates that fruit SSC was significantly negatively correlated with fruit length and fruit weight. However, the correlation between fruit SSC and fruit diameter was not significant (Table S5).
QTLs for FD and FW were co-localized in two linkage groups such as qFW-1B2.2 and qFD-1B2.1 on LG1B and qFW-4D and qFD-4B on LG4B. The former linkage group hosted two more QTLs namely, qFD-1B2.2 for FD and qFW-1B2.1 for FW. Sugar related QTL qSSC-2B were very closely located with FD related QTL qFD-2B on LG2B (Fig. 2). All other QTLs were solitary i.e., they were located on individual linkage group where no other QTLs were located ( Fig. 2; Table 2).

Potential candidate genes within QTL regions
The genes that lie within the flanking regions of the identified QTLs (Table 2) were extracted from the corresponding regions of Fragaria vesca genome v4.0.a1. Altogether, 2348 genes were found within the regions of FL related QTLs (Table S6). The marker MF3023 of QTL qFL-6C1 corresponded to the Fragaria × ananassa gene FAN_ iscf00104879.1 encoding 'Transcription initiation factor TFIID subunit 12' and the flanking marker MF2104 QTL of qFL-3B1 corresponded to gene FAN_iscf00387732.1 encoding 'Serine/arginine repetitive matrix protein 1'. Within FD and FW related QTL regions, a total of 806 and 1629 genes were found, respectively. Both the flanking markers, MF1601 (FAN_iscf00275054.1) and MF994 (FAN_iscf00135334.1) of fruit diameter related QTL qFD-2B encoded for transcription factor genes (Transcription factor TFIIIB component B and Transcriptional activator DEMETER, respectively) and the flanking marker MF2390 of QTL qFD-1B2 encoded a Translation machinery-associated protein (FAN_iscf00346482.1). The corresponding gene (FAN_iscf00345769.1) of the flanking marker MF818 of FW-related QTL qFW-4B encoded a 'NA-binding protein HEXBP' and the corresponding gene (FAN_iscf00266073.1) of the flanking marker MF332 of QTL qFW-6A2 encoded a putative receptor protein kinase ZmPK1. These transcription regulation related activity of the linked marker associated genes may have roles in genetic regulation of the corresponding traits.
Within genes of SSC related QTL regions, fourteen genes were identified to be related with sugar biosynthesis. These fourteen genes, particularly UDP-glucose 4-epimerase GEPI48-like gene (FvH4_5g04740.1), UDP-glucose 6-dehydrogenase family protein (FvH4_5g14230.1), Glucose-6-phosphate/phosphate translocator-related protein (FvH4_5g05430.1), along with the gene FAN_ iscf00049825.1 hosting the flanking marker MF472 of QTL qSSC-2B encoding a UDP-glucosyltransferase gene may have potential roles in sugar biosynthesis in strawberry (Tables S7 and S8). In addition, seven invertase family genes and the transcription factors genes identified within sugar content related QTL regions may also have roles in sugar metabolism.

Developemnt of markers linked with sugar content
Four SNP markers (MF518, MF1321, MF1169, and MF3696) located at the peak LOD positions of the corresponding SSC related QTLs and one SNP (marker MF154) on the gene FAN_iscf00021287, a key candidate gene for sugar biosynthesis (Shanmugam et al. 2017) located within the flanking region of QTL qSSC-5A2.1 were targeted for developing high resolution melting (HRM) marker linked with sugar content in strawberry (Table S9). Among these five HRM markers, reasonable match between the phenotypic and genotypic observations was only found for the marker MF154, designed based on the SNP A1 723 G (for high and low sugar contents, respectively) of the UDP-glucose 4-epimerase GEPI48-like gene FAN_iscf00021287 (Table S9). The prediction accuracy of the marker was 81.39% and 86.95% for the tested 43 F1 individuals and 23 commercial culitvars, respectively ( Fig. 4; Table S10). This indicates it's suitability in marker assisted breeding aiming at improving the trait via molecular breeding. In addition, the allelic influence of this marker on the overall variation in fruit weight in the F 1 population was evaluated using a general linear model with the "lm" function from the "stats" package in R. We found no significant (R 2 = 0.073, p = 0.39) effect of allelic difference on fruit weight (Fig. S2). This indicates that selection for fruit SSC is independent of fruit weight.

Discussion
This study describes construction of high-density linkage map and detection of QTLs for key fruit quality reated traits using genome-wide SNPs derived ddRAD-seq technique from a mapping population raised from 'Maehyang' and 'Festival' (M × F). These two cultivars were chosen based on their contrasting characteristics for a number of agronomic and fruit quality traits and for their diverse geographical origin and adaptation. The variety, Maehyang is a relatively new Korean cultivar originated from 'Tochinomine' × 'Akihime' and is highly valued for its vigorous and erect type growth, weak dormancy, long, conical and red colored fruit, high sugar content, resistance to powdery mildew and relatively high marketable yield (Kim et al. 2004). 'Festival', on the other hand, is a US patented variety developed from hand-pollinated cross of 'Rosa Linda' and 'Oso Grande' and is characterized by profuse runnering habit, large calyces, long fruit pedicel, firm fleshed, flavorful, deep red, large fruits having soluble sugar content comparatively less than that of 'Maehyang' (Chandler 2004).

Transgressive segregation of fruit quality traits
Transgressive segregations were observed for diameter and weight of fruits in F 1 population. Such transgressive segregations have been reported in intra-specific levels such as Oryza sativa, Cucumis melo, Lolium perenne and Pisum sativum etc. (Mao et al. 2011); in inter-specific level such as Oryza, Prunus and Solanum etc. (Gutiérrez et al. 2010;Quilot et al. 2004) and in polyploids such as Triticum durum (Maccaferri et al. 2011) and Saccharum (Ming et al. 2001).
The allelic combinations of the genetically distant parents having complementary gene action could possibly be a reason behind such extreme F 1 phenotypes in our study. Besides, potential epistatic interactions between the multiple alleles of this high-ploidy species may contribute to the dominance or over-dominance effect which results in such transgressive segregants (Coelho et al. 2007;Flagel and Wendel 2009). Nonetheless, such transgressive phenotypic expressions in crossing population of this species could be exploited for developing high performing varieties.

Linkage map of 'Maehyang' × 'Festival' population
The constructed 53 LGs spanned a total genetic distance of 2937.93 cM with an average marker density of 2.14 cM per locus in M × F population. This marker density and physical span is greater than the previously reported linkage maps (Isobe et al. 2013;Lerceteau-Köhler et al. 2003;Fig. 4 Normalized melting peaks (a) and the difference plots (b) of the high resolution melting (HRM) analysis of F1 lines and 23 commerial strawberry lines using the developed HRM marker MF154 Rousseau-Gueutin et al. 2008;Weebadde et al. 2008;Sargent et al. 2009;Zorrilla-Fontanesi et al. 2011;Castro et al. 2015;Dijk et al. 2014). Among the SNP marker based linkage groups, Davik et al. (2015)

QTLs for fruit growth and quality traits
Altogether eighteen QTLs; six for FL and four for FD, FW and SSC each were identified in the M × F population ( Fig. 2; Table 2). Most of the QTLs were also detected by the multilocus mixed linear models such as mrMLM, pLARmEB and ISIS EM- BLASSO (Tamba et al. 2017;Wang et al. 2016;Wen et al. 2018;Zhang et al. 2017 -Köhler et al. 2012;Zorrilla-Fontanesi et al. 2011;Castro and Lewers 2016). In these three studies, QTLs were identified based on AFLP, SSR and SCAR markers. The use of few common markers facilitated the comparison of the constructed linkage groups and identified that a number of QTLs were actually located on homeologous linkage groups, suggesting common genetic control of related traits in multiple backgrounds (Pott et al. 2000). The QTLs of this study were, however, based on genome-wide SNP based markers, thus could not be directly compared with those previous reports. QTLs detected for total SSC in fruits in this study are fewer compared to the cumulative number of previously identified QTLs for individual sugar compounds such as sucrose (2), glucose (5) and fructose (2) by Lerceteau-Köhler et al. (2012). The higher proportion of phenotypic variations (23-50%) explained by the QTLs for SSC, in contrast to the smaller proportions of phenotypic variations explained by the QTLs for individual sugar components, provides a possible explanation for this. Besides SSC, the titratable acidity (TA) and their delicate ratio determines the flavor of strawberry fruits, requiring the selection of fruits that have suitable balance between sweetness and tartness. Several QTLs for TA and SSC/TA has previously been reported to be collocated with the QTLs for SSC (Lerceteau-Köhler et al. 2012;Zorrilla-Fontanesi et al. 2011;Castro and Lewers 2016;Vallarino and Pott 2019). However, the significant but low correlation co-efficient value between these two traits suggested the possibility of independent selection of each of the traits (Castro and Lewers 2016). Overall, identification of several QTLs for a particular trait, each explaining a small individual proportion of the total phenotypic variation indicates complex quantitative inheritance and multi-loci control of these traits (Lerceteau-Köhler et al. 2012;Pott et al. 2000;Paran and Knaap 2007).
Correlation analysis indicated that fruit length is positively correlated with fruit weight, and both these attributes are negatively correlated with fruit SSC (Table S5). This has practical breeding implication in the sense that selection for one trait might negatively affect the selection of another trait. Careful consideration of this point will be necessary for breeding schemes aiming at improving fruit size and fruit sugar contents.

Distribution of QTLs within the linkage map
Among the 53 linkage groups, only 14 hosted all the 18 detected QTLs with LG1B2 hosting maximum four, and LG2B, LG4B and LG5A2 hosting two QTLs each (Fig. 2). Moreover, QTLs for FD and FW co-localized in two linkage groups such as QTLs qFD-1B2.1 and qFW-1B2.2 on LG1B2, and QTLs qFD-4B and qFW-4D on LG4B. The high and significant correlation between these traits (R = 0.512, p value = 0.003) in the F 1 population also indicates potential control by the common genetic regions, possibly by same or tightly linked genes (Table S5). Besides the two co-located QTLs, LG1B2 hosts two more QTLs, qFW-1B2.1 for FW and qFD-1B2.2 for FD (total four QTLs throughout the length of this LG). This indicates that this chromosomal region may be the key genetic determinants of these fruit growth and size related traits. A previous study has reported one-fourth of fruit quality related QTLs to be 1 3 353 Page 14 of 18 clustered in two linkage groups where QTLs for fruit shape, diameter and weight were particularly found to be co-located (Lerceteau-Köhler et al. 2012). In our population, sugar content related QTL, qSSC-2B and FD related QTL, qFD-2B were found to be very closely located on LG2B, even though no significant correction (R = 0.46, p value = 0.13) were observed for these two traits.
Such QTL clusters have also been observed in strawberry for anthocyanin content, antioxidant property (Castro and Lewers 2016), fruit numbers, yield and quality traits (Zorrilla-Fontanesi et al. 2011), and for various fruit quality traits in other crops including pepper, tomato and for fiber development related traits in cotton (Ring et al. 2013;Fulton et al. 1997;Saliba-Colombani et al. 2001). QTL clusters, in genetic terms, suggest the intricately regulated gene networks with potential pleotropic effects; while in breeding terms, indicate the possibility of selection for multiple traits, simultaneously.
The invertases irreversibly hydrolyze sucrose to glucose and fructose and thus play crucial roles in carbohydrate partitioning and plant development (Chen et al. 2015). Three invertase genes namely, FaINVG-1, FaINVA-3 and FacwINV2-1 were recently found to show higher expressions in high sugar-content cultivars (Lee et al. 2018). However, none of these are found within our QTL regions. Instead, the seven invertase family genes, of which three are neutral/alkaline invertases, were found within our QTL regions (Tables S7 and S8). These genes may have roles in sugar metabolism.
In addition, a HRM marker (MF154) was developed based on the SNP A1 723 G of the UDP-glucose 4-epimerase GEPI48-like gene FAN_iscf00021287 for selecting high vs low sugar containing genotpes. The gene was located within the QTL qSSC5A2.1 on linkage group LG5A2 and was found to be a key gene regulating sugar biosynthesis in progressively ripening strawberry fruits (Shanmugam et al. 2017). Evaluation of the influence of allelic difference of the developed HRM marker (MF154) for fruit SSC on fruit weight in the segregating F 1 population via general linear model revealed that there is no significant influence of allelic difference of fruit SSC on the fruit weight. This is understandable, since multiple non-collocating QTLs with minor effects were identified for these two traits indicating that different sets of genes, each with minor effects, may be acting on these two traits. In breeding terms, this indicate that selection for fruit SSC is independent of selection for fruit weight in this population.

Conclusion
This is the first report of QTLs for fruit quality traits using genome-wide SNP based high density linkage map in cultivated strawberry. The detected genome-wide SNPs will supply abundant choice of transferrable markers for future genetic studies. The constructed high density linkage map will serve as reference for precise sequence scaffold anchoring and orientations in this species. Besides the identified QTLs, the linked markers upon validation using larger populations, and in other genetic backgrounds and environments, and the developed HRM marker can be used as mass screening tools in breeding programs. The genes identified within these QTL regions will enhance our understanding of the genetics of these traits and can be targeted for development of varieties with desired fruit quality traits via breeding and biotechnological means.