Clarifying sub-genomic positions of QTLs for flowering habit and fruit quality in U.S. strawberry (Fragaria×ananassa) breeding populations using pedigree-based QTL analysis

The cultivated strawberry (Fragaria×ananassa) is consumed worldwide for its flavor and nutritional benefits. Genetic analysis of commercially important traits in strawberry are important for the development of breeding methods and tools for this species. Although several quantitative trait loci (QTL) have been previously detected for fruit quality and flowering traits using low-density genetic maps, clarity on the sub-genomic locations of these QTLs was missing. Recent discoveries in allo-octoploid strawberry genomics led to the development of the IStraw90 single-nucleotide polymorphism (SNP) array, enabling high-density genetic maps and finer resolution QTL analysis. In this study, breeder-specified traits were evaluated in the Eastern (Michigan) and Western (Oregon) United States for a common set of breeding populations during 2 years. Several QTLs were validated for soluble solids content (SSC), fruit weight (FWT), pH and titratable acidity (TA) using a pedigree-based QTL analysis approach. For fruit quality, a QTL for SSC on linkage group (LG) 6A, a QTL for FWT on LG 2BII, a QTL for pH on LG 4CII and two QTLs for TA on LGs 2A and 5B were detected. In addition, a large-effect QTL for flowering was detected at the distal end of LG 4A, coinciding with the FaPFRU locus. Marker haplotype analysis in the FaPFRU region indicated that the homozygous recessive genotype was highly predictive of seasonal flowering. SNP probes in the FaPFRU region may help facilitate marker-assisted selection for this trait.


INTRODUCTION
The cultivated strawberry (Fragaria × ananassa Duch. ex Rozier) is widely consumed fresh or in processed form. Improved fruit quality is desired by consumers and sought by breeders. 1 Major components of fruit quality attributes include appearance, texture, taste, aroma and nutritional content. 2 The intense flavor of strawberry is characterized by high titratable acidity (TA), high soluble solids content (SSC) and strong aroma. 3 Some progress toward selecting for aroma compounds and flavor has been made over the last few years. 3,4 In addition to good flavor, strawberries contain antioxidant and other nutritional compounds that may help reduce risk of some chronic ailments, [5][6][7] and there has been substantial research aimed at identifying the genes involved in the biosynthetic pathways involved in the accumulation of these compounds. [8][9][10] Cultivated strawberries are allo-octoploid (2n = 8 × = 56) and mapping studies show major disomic inheritance. 1,11 In the last decade, numerous quantitative trait loci (QTLs) for strawberry fruit quality traits have been detected across diverse breeding material. Lerceteau-Köhler et al. 12 pioneered QTL mapping in octoploid strawberry, detecting several QTLs for 19 fruit development and quality traits in a cross between 'Capitola' and the selection CF1116. Most of the QTLs explained 10-17% of the total phenotypic variation for the corresponding trait. Low genetic effects for fruit quality traits are expected as they are strongly influenced by environmental conditions. 13 Zorrilla-Fontanesi et al. 3,14 detected several QTLs for fruit quality traits and volatile compounds in a cross between the Spanish selections 232 and 1392. Castro and Lewers 15 used 177 F 1 individuals from a cross between 'Delmarvel' and 'Selva' to conduct QTL analysis for fruit quality traits and identified QTLs in many of the same regions identified by Lerceteau-Köhler et al. 12 Flowering behavior is another important trait in strawberry plants and is strongly affected by temperature and photoperiod conditions. [16][17][18] Strawberry plants are classified into seasonal flowering (SF also known as 'short-day' or 'June-bearing') or perpetual flowering (PF also known as 'everbearing', 'day-neutral' or 'remontant'). SF plants typically initiate flower buds under photoperiods of o 14 h or at temperatures below 15°C and usually have a shortened harvest period compared with PF plants. 15,17 Conversely, PF strawberries can initiate flowering when temperatures are o 30°C during the day all summer and into the autumn. 19,20 An extended harvest season is an attractive quality for a strawberry, particularly in the central coast of California and other summer production regions. Numerous sources of the PF trait have been identified in octoploid strawberry. 19,20,21 However, only the PF phenotype conferred by the FaPFRU locus from a F. virginiana subsp. glauca accession originating from the Wasatch Mountains in Utah has been studied in great detail. 15,16,18,[22][23][24][25][26][27] This single dominant locus confers the ability to initiate flowers in F. × ananassa under long days at permissive temperatures and is located on the F. vesca chromosome 4 homolog. 15,16,23,25 Recently, Honjo et al. 28 conducted QTL analysis in three populations segregating for PF that were derived from three different PF cultivars. Two of the PF cultivars, 'Summer-berry' and 'Ever-berry', are cultivars classified as 'older everbearers' and are derived from an unknown Japanese PF source while 'Hecker' is a day-neutral cultivar developed at the UC Davis program with the FaPFRU locus. 21,28 A QTL located on chromosome 4 was detected in all three PF cultivars and a subsequent allelism test suggested the PF phenotype in the Japanese cultivars is controlled by a tightly linked gene or the same gene as that present at FaPFRU. 28 Salinas et al. 18 identified a diagnostic allele for the FaPFRU locus in the PF Japanese cultivar Ooishi-shikinari 2, which is derived from the same source as 'Ever-berry', and suggested that the FaPFRU locus may have been introgressed numerous times before its initial discovery by Bringhurst in 1955. There were ambiguities in clear sub-genome assignments of QTLs in the past and this study uses advanced molecular tools to refine and clarify those ambiguities. In each of the previous QTL studies, low-density genetic maps in bi-parental populations were used for QTL detection and sub-genome identification. These lowdensity maps often contained different marker types (amplified fragment length polymorphism and simple sequence repeats (SSRs)) and as a result, the studies are difficult to compare. In addition to that, in the past, QTLs were mostly assigned to a homeologous group (HG) rather than a specific sub-genome. Moreover, bi-parental populations often exhibit limited genetic diversity, resulting in only a few alleles within the germplasm being identified. 29,30 To maximize gain of genetic information, a QTL analysis approach using multiparental pedigree-linked strawberry populations was developed 29,30 and in conjunction with the Axiom IStraw90 SNP array 31 has become a powerful tool for high-resolution QTL mapping. 32,33 Using this pedigree-based QTL analysis (PBA) approach, implemented through FlexQTL, provides advantages over traditional bi-parental population methods, allowing for a QTL to be evaluated in numerous genetic backgrounds while simultaneously increasing the chances for recombination events near QTLs controlling traits of interest. 29,30 The IStraw90 array 31 allows for the high-throughput, repeatable genotyping necessary to produce a high-quality consensus map and clear sub-genome assignments. The genetic map used in this study follows sub-genome assignments from Koehorst-vanc Putten and van de Weg (2017, unpublished data), which clearly defines and separates octoploid strawberry sub-genomes. Individually, the bi-allelic single-nucleotide polymorphism (SNP) markers on the array provide limited information on the segregation of marker alleles via pedigrees because of potential homozygosity at these loci. However, when the markers are analyzed jointly, haplotypes useful for marker-assisted selection (MAS) can be identified. 34 PBA is especially attractive in clonally propagated plants as statistical power can be increased by phenotypically evaluating ancestral germplasm in the same environments as descendants. 30 When using PBA, it is critical to use a dense map as low marker density can add ambiguity in the transfer of alleles through generations. 29 FlexQTL implements a Bayesian statistical approach using pedigree-linked populations to improve statistical power and detect vertically inherited QTL alleles. 29 Confirmation of accurate pedigree information is particularly important when utilizing PBA, and pedigree records from a breeding program must be checked for accuracy because of human error or pollen contamination. 29,30 FlexQTL traces founder alleles to contemporary generation and if pedigree records entered incorrectly, it provides information on which pedigree records are associated with inconsistencies.
The objectives of this study were to further refine and clarify the sub-genomic locations of major fruit quality and flowering habit QTLs, to dissect the genetics controlling fruit quality and flowering in F. × ananassa using the Axiom IStraw90 SNP array for genotyping together with FlexQTL for QTL analysis of multiparental pedigree-linked populations, and to compare the QTL results with previous studies. SNP haplotype analysis was applied to define the locus of a trait controlled by one strong and consistent QTL for PF detected by FlexQTL.

Breeding germplasm
This study focused on breeding populations established in the RosBREED Project to represent the Michigan State University (MSU) and the USDA-ARS Horticultural Crops Research Unit (HCRU-indicated by Oregon U.S. (ORUS)) strawberry breeding programs in Michigan (MI) and Oregon (OR), respectively. 35 The pedigree-connected germplasm consisted of 23 F 1 families representing 11 important breeding parents from the two breeding programs and their 249 progeny, collectively (Supplementary Information S1). The 11 parents included the PF parents 'Tribute', 'Fort Laramie' and 'Seascape' and the SF 'Earliglow', 'Honeoye', 'Puget Reliance', 'Totem' and selections, MSU 49 and MSU 56.

Phenotypic data
Phenotypic data for seven fruit quality traits were measured or scored in 2011 and 2012 in MI and OR according to Mathey et al. 35 and included: SSC (°Brix), fruit weight (FWT) (g per fruit), firmness (Firm; 1 (very soft) − 9 (very firm) scale), TA (g L -1 citric acid), ratio of SSC and TA (SSC:TA), pH, flavor and external color (ExtCol). In addition to the fruit quality traits, the ability of an individual to perpetually flower (PF) was assessed as described by Mathey et al. 35 In this study, flowering was recorded as a binary trait, 0 for SF and 1 for PF, based on the presence of flowers on or after 17 July. 35 Phenotyping at two locations and over 2 years allowed testing of genetic control of traits in different environments and years. Density plots were visualized for each trait. Spearman's rank correlation coefficients (r) and P values were adjusted for multiple comparisons using the Bonferroni correction method for each trait over 2 years using R version 3.2.5. 36 Genotyping and genetic map DNA from seedlings and parents was extracted from actively growing leaf tissue with the E-Z 96 Plant DNA extraction kit (Omega BioTek, Norcross, GA, USA) as previously described. 37 The resulting genomic DNA was quantitated with the Quant-iT Picogreen Assay (Invitrogen, Eugene, OR, USA) according to the manufacturer's recommendations using a Victor 3 V 1420 Multilabel Counter (Perkin Elmer, Downers Grove, IL, USA). DNA concentrations were adjusted to 20-50 ng μL -1 and submitted to Affymetrix, Inc. (Santa Clara, CA, USA) for genotyping with the Axiom IStraw90 SNP array. 31 Genotypic data acquired from Affymetrix, Inc. was recoded with GenomeStudioConverter 34 to convert it into the software input file format needed for Pedimap version 1.2 38 and FlexQTL version 099130. 29 Genetic positions of 10 560 filtered PolyHighResolution and NoMinorHomozygous SNP markers identified in Bassil and Davis et al. 31 were obtained from a high-density SNP genetic map constructed using the 'Holiday' × 'Korona' population (Koehorst-vanc Putten and van de Weg, 2017, unpublished data). Sub-genome assignments and orientation of each linkage group (LG) was based on van Dijk et al. 39 The homozygous regions across parents were mapped using R software to identify genetic regions where lower QTL detection power would be expected. 36 Pedigree confirmation Genotypic data were tested for pedigree relationships and marker inheritance errors using phased marker allele information from FlexQTL version 099130. 29 Extensive checking of parentage records and marker calls were conducted using SNP marker data and a Microsoft Excel-based tool (van de Weg E, 2016, unpublished data). The Excel-based tool was utilized to identify and eliminate off-type individuals and to identify correct parentage. In general, marker genotypes were filtered for consistency of calls and incorrect pedigree records were corrected after validation. When genotypic data were not available for the parents, they were predicted with FlexQTL and the PediHaplotyper R package, 38 and then used to confirm pedigree relationships.
Strawberry flowering and fruit quality QTL S Verma et al.

QTL detection and haplotyping
Haploblock (HB) alleles 29,38 were created using the R package fp_haplotyper (Roeland Voorrips, 2014, unpublished data) and 10 560 SNP markers, in order to save computation time. FlexQTL was used for genome-wide QTL analysis using HB alleles. 29,34 LGs where significant QTLs were identified in at least two out of four location-years were reanalyzed with SNP markers using FlexQTL. 32,[40][41][42][43] Phenotypic data of only unselected material (seedlings) were used for QTL analysis to prevent prediction biases. Parameter settings used for all FlexQTL runs followed Mangandi et al. 33 The Bayes factor parameter (2lnBF) was interpreted as nonsignificant (0-2), positive (2)(3)(4)(5), strong (5-10) or decisive (410) evidence for the presence of QTLs. Genome-wide evidence was collected in support of QTL models ranging from 1 QTL versus 0 QTLs to n versus n-1 QTLs, where n is the number of QTLs detected in the model. MapChart version 2.3 44 and VisualFlexQTL version 0.1.0.42 were used to visualize FlexQTL output on traces of convergence of QTL models to identify the genomic regions associated with significant QTLs. Additional posterior inference analysis, using the postQTL module, was conducted to delineate significant and repeatable QTL regions in order to obtain the best posterior estimates of genotype probabilities, breeding values and weighted additive variances. The phenotypic variation explained by a QTL was estimated as PVE (%)= [(probability×AVt)/Vp]×100, where probability is the cumulative probability of presence of a QTL in the interval, AVt is the additive variance for the trait, and V P is the total phenotypic variance. QTL effects were calculated and reported.

Assessment of EMFv006 marker for DNA informed breeding
The SSR marker EMFv006 was previously found to be associated with an SSC QTL on chromosome 6A. 12 EMFv006 was assessed for potential use in MAS by genotyping these populations for which phenotypic data was also gathered. PCR was conducted using the

Phenotypic data
The phenotypic distribution of many of the fruit quality traits approached normal distributions ( Table S1). Significant Spearman's rank correlation coefficients (α = 0.05) were observed for PF within a location across years and between locations within 2011 and 2012. Most of the correlations were not significant (α = 0.05), however, significant negative correlation coefficients (r) were observed between TA and pH and significant positive correlations were observed for TA between locations ( Supplementary Fig. S2).  Table S2). Polymorphic markers were well distributed among LGs and minor allele frequency ranged from 0 to 0.5 (Supplementary Figure S3). QTLs for only four (SSC, FWT, pH and TA) out of the eight fruit quality traits were significant (  Figure S5). For pH, a major effect QTL was detected on LG 4CII (MI 2012; Table 1; Supplementary Figure S6), which explained 20% of the phenotypic variation. One moderate effect (5 ≤ 2lnBF o 10) TA QTL was detected on LG 2A in MI 2011 and a minor effect QTL for TA was detected on LG 5B in OR 2012 (Table 1; Supplementary Figure S7). The moderate effect TA QTLs on LG 2A explained 8% of the total phenotypic variation and the one on LG 5B explained 22% of the total phenotypic variation.

Pedigree confirmation and correction
A single major effect QTL for PF was detected by genome-wide QTL analysis of haplotype block alleles (Table 1; Supplementary Figure S8) across both locations (MI and OR) and years (2011 and 2012). The PF QTL explained 68% and 33% of the total phenotypic variation in 2011 and 2012, respectively, at both locations and was detected on the distal end of LG 4A between 44.7 and 50.2 cM using SNP markers (Figure 2). Based on the founders used in this study, this QTL likely corresponds to the FaPFRU locus. 18,23,24 To confirm co-location of the PF QTL with FaPFRU, the sequences of 15 SNP markers associated with the QTL, and six additional SNP markers that were not used for the QTL analysis because they were either monomorphic or not high-quality markers within the QTL Strawberry flowering and fruit quality QTL S Verma et al.
region, were BLAST searched against the scaffolds of the F. vesca v1.0 genome assembly. All markers aligned with scaffold 0513158 (Figure 3), which is the same scaffold previously identified as containing the FaPFRU region. 24 To better visualize the FaPFRU region, these 21 SNP markers and the SSRs developed by Perrotte and Gaston et al. 24 were physically mapped to scaffold 0513158. All of the SNP markers mapped proximally to the FaPFRU region defined by Perrotte and Gaston et al. 24 (Figure 3).

Assessment of EMFv006 marker for DNA informed breeding
The SSR marker EMFv006 linked to the 6A SSC QTL 12 was used to screen the RosBREED strawberry population set. The SSR marker amplified five different alleles: 207bp, 209bp, 213bp, 215bp and 217bp. These five marker alleles constituted 12 different genotypes and phenotypic data were available for nine marker genotypes. Phenotypic distribution associated with each of the nine diplotypes were visualized by density plots and no clear separation of density curves were observed (Supplementary Figure S9). At the same time, phenotypic distribution associated with FlexQTL assigned QQ, Qq, qq diplotypes were also visualized and we observed separation in phenotypic distributions associated with QQ and qq with at least 50% overlapping region of the distributions (Supplementary Figure S9). This indicated that there might be significant association at the SNP level at the SSC locus on LG 6A. An analysis of variance model using these three QTL diplotypes (QQ, Qq, qq) explained approximately 20% of the total SSC variation within the strawberry breeding sets evaluated in OR 2012. However, when genetic screening of the RosBREED material was conducted with the SSR marker EMFv006, substantial homozygosity was observed. Further, high-resolution melting curve assay was performed to detect any sequence difference between the SSR marker EMFv006 homozygous alleles; however, no differences were detected. Therefore, EMFv006 was found to not be useful for MAS.
Assessment of FaPFRU locus HBs and haplotypes for MAS Four HBs extending across the 15 SNPs associated with PF were identified within the germplasm using Haploview (Figures 3 and  4).  Fig. S10).
The three SNP markers of HB-2 were used jointly to haplotype the 249 seedlings and 11 parents from the 23 MSU and ORUS pedigree-linked populations. As these three SNP markers were in high LD, haplotype diversity was low. Four haplotypes were segregating in the unselected material (seedlings). Haplotype AAA was present in 29% of the seedlings, whereas haplotype BBB was found in 70% of the seedlings. Two other haplotypes, ABB and AAB, were rare (frequency ≤ 1%). Based on the PediHaplotyper prediction, the AAB haplotype is the result of a recombination between alleles originating within unidentified parents near the FaPFRU locus in families MSU_9-15 and MSU_9-16. One progeny, MSU_9-14-8, carried the rare haplotype ABB, which was inherited from recombination between haplotypes of 'Seascape' (mother) and an unidentified father. Among the most frequent haplotypes within unselected material, only four diplotypes (compound haplotype) were possible: AAA|AAA (4%), AAA|BBB (35%), BBB| AAA (14%) and BBB|BBB (45%).
The homozygous diplotype BBB|BBB was most predictive for SF as demonstrated by higher negative predictive values over positive predictive values in each environment ( Figure 5; Table 2) and was most predictive of SF in OR 2012. Heterozygous and homozygous individuals carrying the AAA haplotype expressed mixed flowering behavior. The cultivars Fort Laramie, Seascape, Tribute and Fort Laramie 2 were heterozygous at the FaPFRU locus and PF, whereas 'Puget Reliance' was heterozygous at the FaPFRU locus and SF. However, 'Earliglow', 'Totem', 'Honeoye', ORUS 2427-1, 'Mary's Peak', MSU 49 and MSU 56 were homozygous BBB|BBB and SF. Environmental variability appears to affect the diagnostic accuracy of HB-2, which ranged from 64 to 80.7% (Table 2). LGs 2lnBF Interval (cM)

DISCUSSION
In this study, numerous QTLs were identified for horticulturally important traits. Many of the QTLs identified using the PBA approach were also identified in other studies using the pseudo test cross strategy 3,11,12,14,15,23,24,26,27 demonstrating this new methodology performs as good as previously established QTL analysis methods for strawberry. The PBA approach allowed for the QTLs to be evaluated in numerous genetic backgrounds, illustrating robust genotype-trait associations. In this study, we were able to clarify and pinpoint sub-genomic locations of QTLs. Moreover, the use of the 90K axiom array 31 in the present study has allowed for the identification of significant QTLs with much higher resolution than in previous studies.
Correct pedigree information is critical for the PBA approach used and extreme care was taken to ensure correct pedigree information before conducting the QTL analysis. The power to detect a QTL using a PBA approach is affected by the frequency of the founder alleles in the population and is assessed via average allele representation (AAR). An AAR of 12.5 results in a 77.0%, 95.7%, 98.5% and 99.6% probability that each alternative allele is present at least 10, 8, 7 and 6 times in the population, respectively. 53 An AAR of 12.5 units is recommended as the minimum for statistical power in representing the alleles of breeding parents. A minimum AAR of 12.5 can be achieved either by 23 progeny (0.5 × 23 = 11.5 AAR) and two parents (0.5 × 2 = 1AAR) or by 25 progeny (0.5 × 25 = 12.5 AAR) with unknown  parents. There are other possibilities because of pedigree connections that can increase AAR because of relatedness as detailed by Peace et al. 53 The AAR for the parents ORUS 2427-1, 'Puget Reliance' and 'Totem' was below the recommended threshold because of low numbers of true seedlings as compared with the rest of the parents. However, the alleles of the PF cultivars Seascape, Tribute and both genotypes of Fort Laramie, as well as the SF cultivar Earliglow, Honeoye, MSU 49 and MSU 56 were well represented with an AAR ranging between 12.5 in MSU 56 and 51.5 in 'Tribute' (Supplementary Information S1). Although only minor (2lnBF o 5) and moderate (5 ≤ 2lnBF o10) QTLs were detected for four fruit quality traits, one major effect (2lnBF ≥ 10) QTL was detected for flowering habit. The moderate effect QTL for SSC detected on LG 6A is supportive of the SSC QTLs on LGs associated with chromosome 6 previously reported in QTL studies conducted in the cultivated strawberry thus far. 3,12,16 However, owing to the low mapping densities and the type of markers used in previous studies, it is not clear that these previously reported QTLs belong to LG 6A or to homoeologous chromosomes. The remaining five fruit quality QTLs identified in this study coincided with QTLs reported by at least one other study for the same trait on the same HG chromosome. The FWT QTL we detected on LG 2BII was also observed on HG 2 by Lerceteau-Köhler et al. 12 The pH QTL detected on LG 4CII in MI 2012 was also previously reported on HG 4, by Lerceteau-Köhler et al. 12 and Zorrilla-Fontanesi et al. 3 As with the SSC QTL on LG 6A, however, it cannot be confirmed that the other fruit quality QTLs reported here are in the same regions found in the earlier studies.
The SSC QTL on LG 6A explained 10% of total SSC variation in the breeding populations evaluated in OR 2012. The SSR marker EMFv006, which was previously reported by Lerceteau-Köhler et al. 12 to be closely associated with the SSC QTL, was also found flanking the current SSC QTL (van de Weg E, unpublished data). The SSR marker EMFv006 alleles were not able to differentiate between QTL alleles assigned by FlexQTL and was rendered not suitable for MAB applications. To determine if unknown polymorphism existed within the SSR marker EMFv006, analysis by high-resolution melting and amplicon sequencing was performed and no polymorphism associated with SSC level was determined, supporting ineffectiveness of EMFv006 marker for MAS.
Interestingly, few QTLs of significance were detected for each of the fruit quality traits and many were not consistent across environments or years. Missing phenotypic data were not  consistent across locations and years and this unbalanced representation of the phenotypic data might have hindered detection of fruit quality QTLs across locations and over years. In addition, many fruit quality traits have been shown to be quantitatively inherited and under large environmental influence. 3,[12][13][14][15] Environmental influences may have prevented the detection of some smaller effect QTLs, although QTLs may not have been detected because of two sources of ascertainment bias that are present in this study. First, regions of low mapping resolution will exist in areas of the 'Holiday' × 'Korona' map where both parents were homozygous. As genetic positions from the 'Holiday' × 'Korona' map were used for our analysis, this source of lower mapping resolution was transferred to this experiment. Second, artificial regions of homozygosity may have been created in the parents because of using only markers that are polymorphic in the 'Holiday' × 'Korona' population but not in the parental germplasm. In all likelihood, unobserved markers exist that are polymorphic between parental lines in these regions that could aid in the detection of QTLs, as well as assist in the phasing of alleles. As many of the cultivars used in previous PF studies share ancestry with those used in this study, the major effect QTL for PF detected at the distal end of LG 4A is likely the FaPFRU locus. 11,15,16,18,23,24,26,27 This is further supported by the mapping of all 21 SNP markers within the QTL region to F. vesca scaffold 0513158. 24 High LD between SNP markers within HB-2 of the PF QTL region indicated co-inheritance of SNP marker haplotypes. HB-2 was approximately 1.5 Mb away from the FaPFRU locus and had significant association with flowering habit, with diplotype BBB|BBB being highly predictive of SF. Despite being highly associated with flowering, test accuracy varied among environments, ranging from 64 to 80.7%.
The genetic system that enables PF appears to be complex and likely constitutes several key genes and proteins that are capable of sensing environmental cues and hormone levels, which affect their function. 54 This was reflected in the lack of clear 1:1 association of the putatively dominant AAA allele with PF in all families (Supplementary Table S3). We propose two possible reasons for this lack of clear association: (1) environmental conditions suppressed the expression of flowering related genes, and (2) the contribution of founder alleles whereby a recombination event or independent mutational events have caused the AAA haplotype to be linked in coupling to SF. OR and MI have very different environments, and fluctuations in temperature were an everyday phenomenon within these areas during the study period. Seasonal cues like temperature, light conditions and day length have a significant role in the floral transition leading to PF in strawberry. 55,56 In addition, age-dependent vernalization also influences flowering. 57 Perrotte and Gaston et al. 24 and Salinas et al. 18 identified SF cultivars, such as Earliglow and Elsanta, which had the diagnostic allele of Bx215 for PF demonstrating that recombination or mutational events have occurred in the region. As the SNP markers from HB-2 are physically more distant than Bx215 from FaPFRU, it is not unexpected for one of these events to have occurred.
In conclusion, by using multiple pedigrees, PBA determines the effect of alleles from multiple founders simultaneously. Ten QTLs combined were detected for fruit quality traits in MI and OR over 2 years and a major effect QTL was detected for flowering habit in all environments. The flowering QTL was detected in the same genomic region as that of FaPFRU. A haplotyping approach applied to the SNPs within the FaPFRU allowed us to achieve greater concordance with the flowering habits of germplasm in different environments. Alleles predictive of SF were identified and can be converted to a breeder-friendly DNA test to distinguish flowering habits in the cultivated strawberry. Future work will include further dissection of the genetic regions identified in order to better understand the molecular mechanisms driving these traits. This work will ultimately help breeders identify cultivars with desired traits and provide information to develop DNA-based markers for MAS.