Introduction

In the southeastern USA, the most limiting factor for raspberry production is the lack of adapted cultivars. Typical seasonal conditions in southeastern states are warm humid summers and winters with intense temperature fluctuations. In order to establish major production in this area, cultivars will require tolerance to both of these conditions, which is very seldom found in available raspberry cultivars (Ballington and Fernandez 2008; Jennings 1988). Domestication of red raspberry has resulted in a reduction of genetic diversity, with most modern cultivars being genetically similar and adapted to climates like that of the Pacific Northwest (Graham and McNicol 1995; Graham et al. 2004; Jennings 1988). After fulfilling the chilling requirement, fluctuations in temperature in the winter cause premature bud break when temperatures increase and expose fruiting laterals to winter damage after late frosts (Jennings 1988; Westwood 1993). This can cause complete loss of the crop when plants are not tolerant. Woody perennial plants can tolerate these fluctuations by requiring high chilling hours to break winter dormancy (Jennings 1988; Westwood 1993). Thus, for regions such as North Carolina, high chilling requirements are desirable for tolerance to winter fluctuations. However, in growing regions such as Mexico, a lower chilling requirement is necessary to break dormancy and produce a healthy crop in the following growing season. Therefore, breeding for region-specific chilling requirements is important for production of Rubus; this is also true for other woody perennial plants (Westwood 1993).

In the North Carolina State University Rubus Breeding Program, efforts are underway to develop cultivars with high chilling requirements, but because of the biennial nature of the plant, breeding for important traits is a time-consuming process. Initial field trials alone take up to 3 years and cultivar development takes 10–15 years. This emphasizes the need for molecular tools to help expedite the breeding process. Molecular research in Rubus has recently made greatly advances, and tools such as a genetic linkage map with important quantitative trait loci (QTL) would further contribute to the tool set for marker-assisted selection (MAS).

Few studies have looked at bud dormancy and chilling requirement at the molecular level, and thus molecular and genetic mechanisms are not well understood (Mazzitelli et al. 2007; Rohde and Bhalerao 2007; Sánchez-Pérez et al. 2007). Perhaps the best characterized system in the Rosaceae family related to chilling requirement is blooming time in Prunus. Because flowering time is an important agronomic trait in Prunus, there have been several genetic and quantitative studies on that trait, as well as the characterization of an evergrowing mutant in peach (Bielenberg et al. 2004; Sánchez-Pérez et al. 2007; Silva et al. 2005; Wang et al. 2002). Blooming time is inherited quantitatively in the majority of fruit species (Anderson and Seeley 1993), and thus a quantitative analysis method of studying chilling requirement is a plausible approach.

Cultivars without cane prickles (or spines) are easier to manage (i.e. easier to prune and harvest) and are therefore highly desirable for breeders and growers. There are several major genes that confer prickles in Rubus idaeus (Jennings 1988; Jennings and Ingram 1983). One of these is gene s, which is dominant for prickles. R. parvifolius can have gene(s) for the prickle-free trait and are of particular interest because its occurrence in F1 hybrids suggested that these are dominant (Jennings and Ingram 1983). However, these genes have not been genetically characterized. There are several major genes that confer the prickle-free trait in addition to gene s: genes Sf (Rosati et al. 1986), S TE (Hall et al. 1986), and Sf L (Rosati et al. 1988). Cane pubescence, fine hairs on the surface of the canes, is determined by gene H (genotype HH or Hh), and in homozygous recessive individuals (hh) hairs are absent (glabrous). Homozygous dominant (HH) individuals are rarely found due to linkage with a lethal recessive gene (Jennings 1988; Jennings 1967; Keep et al. 1977), and gene H also has a pleiotropic effect on prickles, where its presence slightly increases their density (Jennings 1962).

A notable quality of raspberry fruit is the high level of antioxidants, which is largely attributed to anthocyanins and other plant-derived compounds (Cao et al. 1997; Mullen et al. 2002). Raspberries have a variety of colors attributed to different anthocyanins and can range from yellow, to red, to black. The plant material in this study is of a complex nature, and the cross involves three different species; all three bear different fruit colors. The mapping population used in this study had a wide array of different colors, showing possible differences in types of anthocyanins. These color differences suggests that segregation of anthocyanin genes is occurring, and therefore regions that are closely related to different anthocyanins can be mapped.

The mapping population also segregated for flower color, fruit shape, and fruit size. For some of these traits, only some genes have been described, and therefore mapping these traits can contribute to their genetic understanding.

Herein, we present the construction of a genetic linkage map for performing quantitative trait analysis of chilling requirement, fruit color, fruit shape, relative fruit size, and prickle density. This is the first study to analyze the chilling requirement in raspberry for the identification of QTL.

Materials and methods

Plant material

The mapping population used for this study was generated from a modified backcross between (R. parvifolius × Tulameen) (NC497) and Qualicum, and the offspring segregate for different chilling requirements, as well as for other traits (Ballington and Fernandez 2008; Daubeny and Anderson 1991; Daubeny and Kempler 1995). The blackberry cultivar Kiowa (Moore and Clark 1996) was used as a low-chilling control. Four hundred seedlings of the mapping population were planted in June 2006 at the Sandhills Research Station in Jackson Springs, North Carolina, including the parents and Kiowa. This region is located between the coastal plains and the piedmont area, where temperatures fluctuated between −8.0 and 20.0 °C during the months of January and February between the years 2006 and 2008 (State Climate Office of North Carolina, NC State University; CRONOS internet database available at http://www.nc-climate.ncsu.edu/cronos/). Watering was applied as needed by overhead water reels throughout the growing season to equal approximately 2.5 cm per week from May to September. One hundred and ninety-eight individuals from the mapping population were randomly selected for the construction of the genetic linkage map.

Polymerase chain reactions and molecular markers

Purification of total genomic DNA was extracted from young leaves of 198 individuals using a 2 % CTAB method (Graham et al. 2003). For amplified length polymorphism (AFLP) fingerprinting (Vos et al. 1995), 500 ng of template DNA from the mapping population and parents was digested with 12 U of EcoRI and 8 U of MseI and was performed as described by Vos et al. (1995). To maximize the number of polymorphic markers, a preliminary screen of 171 different selective amplification primer combinations was performed on six of the progeny. The selective primer combinations were selected based on the number of polymorphic markers and clarity of bands on a polyacrylamide gel. The 5′ end of the EcoRI selective amplification primer was labeled with IRDye™ 800 or IRDye™ 700 (LI-COR, Lincoln, Nebraska, USA). AFLP products were denatured and resolved in an 8 % polyacrylamide gel that ran for approximately 3 h in a 4300 DNA Analyzer System (LI-COR). Simple sequence repeat (SSR) and SSR-expressed sequence tag (SSR-EST) markers were used for the development and anchoring of the linkage map (Graham et al. 2004; Graham et al. 2002). Polymerase chain reactions (PCR) for SSR and SSR-EST were carried out using a three-primer tailing system, where the IRDye™ label was bound to the 5′ end of the M13 universal primer (Schuelke 2000). Reactions were performed using 20–50 ng of genomic DNA and were carried out in a 10-μl reaction volume with 1 × standard Taq DNA polymerase buffer containing MgCl2, 0.2 mM of each nucleotide, 0.2 pmol forward primer, 1.0 pmol reverse primer, 1.0 pmol M13-universal IRDye™ labeled primer, and 0.5 U Taq DNA polymerase (Roche), and placed in a Gene Amp PCR System 9700 thermal cycler (Applied Biosystems). The thermal profile had an initial denaturing step at 94 °C for 2.5 min, then 15 cycles of 94.0 °C for 30 s, annealing at 59 °C for 30 s, and extension at 72.0 °C for 1 min, followed by 25 cycles of denaturing at 94.0 °C for 30 s, annealing at 50 °C for 30 s, and extension at 72.0 °C for 1 min, and a final extension at 72 °C for 7 min.

Linkage map construction and QTL analysis

The linkage map was constructed by analyzing data from the polymorphic SSR and AFLP fragments in the mapping population using JoinMap 4.0 and setting the population type as CP for cross-pollinated crops (Van Ooijen 2006). The expected segregation ratios of the molecular markers (either 1:1 for aa × ab and ab × aa, or 3:1 for ab × ab) were tested for goodness-of-fit using a Chi squared test. Linkage groups were named following the revisions made by Bushakra et al. (2012). The regression mapping algorithm was used at a logarithm of odds (LOD) score of 6.0. Markers that produced weak linkages or that did not form a linkage group of more than three markers were excluded. Recombination units were converted into genetic distances using the Kosambi function. QTL analysis was conducted using MapQTL® 5 (Van Ooijen 2004). Significant markers associated with the chilling trait were examined by single point analysis and by interval mapping. LOD significance thresholds were determined by running a permutation test with 1,000 iterations. LOD scores, percent variation explained, and associated mean genotypic values were calculated using the interval mapping function and, whenever possible, were further analyzed by MQM analysis (composite interval mapping). Because of their qualitative nature, significant markers for prickle density, berry color, berry size, berry shape, and flower color were analyzed by single-point analysis using the Kruskal–Wallis method. Markers were considered significant at p ≤ 0.001. LOD scores, percent variation explained, and associated mean genotypic values were calculated using the interval mapping function for those traits. Graphical representation of linkage groups and QTL were generated using MapChart 2.2 (Voorrips 2002) and histograms of trait distributions were generated in KaleidaGraph 4.1.3 (Synergy Software).

Chilling hour requirement screening and determination

Chilling requirement in the mapping population was measured using the method described by Warmund and Krumme (2005) with one modification: in the method used herein, chilling hours are allowed to accumulate under refrigeration at 4 °C instead of field conditions. Chilling requirements were measured on the 196 selected genotypes of the mapping population during three non-growing seasons: the winters of 2006/2007, 2007/2008, and 2008/2009. Time of chilling inception and accumulated chilling hours were calculated using standard model 2, where chilling inception occurs at the first incidence of −2.2 °C and one chilling hour is defined as one cumulative hour at temperatures between 0 and 7 °C (Warmund and Krumme 2005). The accumulated chilling hours and the point of chilling inception at Jackson Springs, NC were monitored through the CRONOS internet database. Eight- to ten-node wood sections were collected from the middle portion of the floricanes (reproductive canes) when at least 100 chilling hours had been accumulated under field conditions to ensure all genotypes had entered dormancy. Wood sections were then kept under refrigeration at 4 °C, in hydrated conditions. Three randomly selected wood sections from each genotype were removed from refrigeration and placed in trays with Oasis Wedges (Smithers Oasis Company, Kent, OH, USA) after sections had accumulated approximately 200 chilling hours. Trays were then placed in the greenhouse on Redi-Heat heating mats plugged into a thermostat set at 25 °C (Phytotronics, Inc., Earth City, MO, USA), and kept under misted conditions (10 s every 10 min). The position and direction of the trays were changed every week to normalize any variability in misters and heating mats. Percent bud break was measured every week for 4 weeks. A bud was defined as broken when the first leaf expanded and became visible. All buds were counted, except for buds inside the Oasis Wedge, in case the bud had died due to lack of air. This process was repeated each time buds accumulated 200–400 chilling hours (approximately every 8–16 days) inside the coolers. Chilling requirements were considered met at the first significant increase in bud break as described by Carter et al. (2006) using PROC GLM in SAS (SAS Institute, Cary, NC, USA). In the first screening (2006–2007), only scanty plant material was available, and therefore two wood sections per genotype per data point were used.

Prickle density

Canes were scored visually on a scale of 0–5, where 0 is no prickles and 5 is densely covered in prickles. Plants were scored for density in 2006. Canes were also scored for pubescence to account for potential pleiotropic effects.

Fruit color

Fruit color was categorized using the Royal Horticultural Society Colour Chart (Royal Horticultural Society 1966). Color scores were yellow (11A), pink (37A), orange (23A), bright red (45A), and dark red (59A); varying degrees of each score were used when applicable. Fruit color was scored at two ripening stages and two growing seasons (2007 and 2008). However, there were little to no differences in significant marker regions in seasons or ripening stage after analyses and therefore values for both years were averaged and only scores for one ripening stage were used.

Fruit shape and size

Fruit shape in the progeny was either round, semi-conical, or conical, with few intermediates. The size of the fruits had similar variability, where relative size was small, medium, or large. The size and shape of the fruit was scored on a scale of 1–3, where 1 was round or small, and 3 was conical or large. Because the berries in the progeny tended to be small and round, or conical and large, a correlation matrix of the traits was generated using the PROC CORR function in SAS. Fruit size and shape were scored during two growing seasons (summer 2007 and 2008). Very few differences in significant markers were observed for either trait in both seasons, and thus scores were averaged for the two seasons.

Flower color

This was defined as the color of the petals, i.e. excluding the color of the stamens and carpels. Color was scored visually on a scale of 1–3, where 1 was white, 2 was light pink, and 3 was pink. There were no differences in significant markers between seasons, and only a few individuals differed between the two evaluations, and thus data were combined by averaging both seasons.

Results

Linkage map construction and QTL analysis

Of the prescreened AFLP primers, 12 out of 39 selective combinations were chosen for the construction of the genetic linkage map (Online Resource 1a). A total of 341 AFLP markers were scored, and 161 were used for the construction of the map after linkage analysis. Thirty-nine SSR markers were tested on the mapping population. Ten of these were monomorphic, seven had no or inconsistent amplification, one was heterozygous, and twenty were polymorphic. However, two of the polymorphic markers did not amplify the expected SSR region using the M13-tailing approach, and consequently were not mapped. One of the markers (leaf97) had insufficient linkage information to be placed on the map, but was grouped onto linkage group 1 (LG1). Thus, a total of 17 SSR markers are shown on the map (Fig. 1). Allele sizes varied slightly in most cases; however, several null alleles were present in NC497, probably due to the introduction of R. parivifolius (Table 1). The Rubus map reported here had a total length of 564.64 cM. At least one SSR marker was placed onto each LG, and was compared to that developed by Graham et al. (2006). LGs were of similar size and SSR markers mapped to similar positions to those reported in Graham et al. Initially, parental maps were constructed; however, the majority of the markers segregated in NC497, and therefore a combined map was constructed. LOD scores beyond 6 resulted in the separation of one or a few markers from the group. Overall, the majority of the markers within the groups fragmented beyond a LOD score of 8. One SSR marker, Rub157, has now been mapped to LG3 (Fig. 1).

Fig. 1
figure 1

Genetic linkage map of raspberry (NC497 × Qualicum) constructed using AFLP and SSR molecular markers. QTL were found for the following traits: chilling requirements measured in winters 2006/2007 (Chilling1), 2007/2008 (Chilling2), and 2008/2009 (Chilling3); fruit size (Size), shape (Shape), and color (Color); flower color (Flower); and prickle density (Prickle)

Full size image
Table 1 Microsatellite (SSR) allele sizes and motifs genotyped in the mapping population NC497 × Qualicum
Full size table

A total of 15 QTL were identified (Table 2). Approximately half of the QTL explained a large percentage of the variation of the traits. The QTL for prickles on LG6 explained the least variation (9.0 %), while the QTL for fruit shape on LG6 had the highest explained variation (86 %). Chilling requirement QTL had LOD scores that ranged between 3.0 and 7.0, while qualitative traits, such as flower color, had LOD scores as high as 36.91. All Kruskal–Wallis analyses revealed at least one highly significant marker associated with the traits (p ≤ 0.0001).

Table 2 QTL detected in the in the NC497 × Qualicum population: chilling requirements measured in winters 2006/2007 (Chilling1), 2007/2008 (Chilling2), and 2008/2009 (Chilling 3); fruit size (Size), shape (Shape), and color (Color); flower color (Flower); and prickle density (Prickle)
Full size table

Chilling hour requirements

The general trend was normal for all seasons except for the first, where most genotypes had chilling requirements below 1,000 h (Fig. 2). The correlation between seasons was positive and significant (data not shown). The chilling requirement of the parents was 800 and 400 for Qualicum and NC497, respectively. The range of chilling requirement in the progeny, however, was larger, ranging from 200 to 1,600 h. Therefore, there was transgressive segregation. Kiowa was used as a low-chilling control, and the estimated chilling requirement was generally consistent with previous studies (Warmund and Krumme 2005), but differed in the 2007–2008 screening from low chilling (200 h) to mid chilling (700 h). The QTL in the second non-growing season (Chilling2) had the lowest percent explanation, each accounting for 32.2, 24.3, and 15.3 % of the variation (Table 2). QTL were found on LG1 and LG4 for the second and third season (Chilling2 and Chilling3), and one QTL was found on LG6 in the first and second season (Chilling1 and Chilling2). The QTL Chilling2 and Chilling3 co-localized to the middle portion of LG1, and in close proximity on the upper portion of LG4. Chilling1 and Chilling2 on LG6 co-localized to the lower portion of the group (Fig. 1). Alleles coming from the NC497 parent had a positive effect on chilling requirement for LG4 in the second season. However, in the third season, a heterozygous model was reflected where both parents contributed alleles (Table 2). A heterozygous model where both parents contributed alleles was reflected in the first season on LG5 (Chilling1; Table 2). NC497 also contributed alleles for higher chilling requirement on LG6, but only for the second season (Chilling2; Table 2). When analyzed by single point analysis, three SSR markers were associated with the QTL for chilling on LG6 (Rub103a, Rub228a, Rub259f), and one AFLP marker on LG4 (E38M36_50.4). Interestingly, Rub228s and Rub259f were two of the four most significant markers in the Kruskall–Wallis analysis for all three non-growing seasons (p ≤ 0.0005 for the second and third seasons, and p ≤ 0.01 for the first season), while E38M36_50.4 was significant in the second and third seasons (p ≤ 0.001). In both of these SSR markers, the segregating allele coming from the NC497 parent was associated with a lower mean, while the AFLP marker was associated with a higher mean (data not shown).

Fig. 2
figure 2

Distribution of chilling hour requirement (Chilling Hours) for the mapping population NC497 × Qualicum. Chilling hour requirement was tested in the population in the non-growing seasons in 2006–2007 (Chilling1), 2007–2008 (Chilling2), and 2008–2009 (Chilling3). Chilling hour requirements were estimated by bud break of chilled cuttings

Full size image

Prickle density

All individuals in the mapping population were pubescent, and likely heterozygous for gene H. Both parents in this study exhibited prickly canes, but the population ranged from densely prickly to prickle-free (Fig. 3). Approximately 20 % of the mapping population is prickle-free, but the majority of the progeny (~50 %) were mildly prickly (score of 4 or 3), while a small percentage was sparsely prickly or densely prickly (score of 1, 2, or 5). Two QTL were detected for prickle density, one of which explained a large portion of the variation and was associated with LG4, while the second QTL was associated with LG6 and explained less than 10 % of the variation (Table 2). The means associated with the different genotypes for the QTL on LG6 support a heterozygous condition for both parents, where the “ac” alleles are additive and increase prickle density (Table 2). The QTL with the larger percentage of the explained variation was found on one of the ends of LG4 close to marker Rub116a, and explained approximately 84 % of the variation with a high LOD score (Table 2). When prickle-free scores were removed from the analysis, the same QTL were detected, and LG4 still explained a large percentage of the variation (46.3 %) but the QTL had a lower LOD score than that of LG6 (3.60 vs 5.32) and the percent variation for the QTL on LG6 increased from 9.0 to 16.7 %. The segregation ratio of 3:1 for the trait fits a single gene model, assuming both parents are heterozygous for gene s (χ 2 = 3.29, α = 0.05 and df = 1). The means associated with the genotypes for the QTL on LG4 also support a heterozygous condition for both parents, i.e. the “bd” genotype is homozygous recessive (score of 0) and the “ac” genotype is homozygous dominant for prickles (score of 4 or 5; Table 2). When the trait was scored as a marker (0 as prickle-free, and 1 as prickles present on canes), the marker mapped to the QTL region on LG4, close to marker Rub116a, and was designated gene s (Fig. 1). The marker Rub116a was highly significant (p ≤ 0.0001), and the associated mean in the “ab” genotypes was for dense prickles, i.e. in “ab” individuals, the “b” allele coming from NC497 was associated with a mean score of 3.42, while “aa” individuals had a mean score of 1.93.

Fig. 3
figure 3

Distribution of prickle density on canes of the mapping population NC497 × Qualicum evaluated in 2006. A density score of 0 was prickle-free, while a density score of 5 was densely prickly

Full size image

Fruit color

A broad range of colors segregated in the population, but, in summary, berries were yellow, orange, pink, red, or dark red, and changed color depending on the ripening stage. The distribution of the scores shows two asymmetric peaks, where most scores were dark yellow–orange and red (Online Resource 1b). The QTL explained between 18 and 32 % of the variation on LG5 and LG1. The means associated with the different genotypes show that alleles coming from the parent NC497 contributed to orange berry color (score of 23), and alleles that contribute or allow visualization of red-pink berry colors (scores of 37 and 45; Table 2).

Fruit shape and size

The berry shape and size of the NC497 parent was small and round, while that of the other parent, Qualicum, was large and conical. Most of the berries were small and round, while a few individuals had large and conical fruit (Online Resource 1c). There was a significant positive correlation between berry shape and size, but only within each growing season, where berries increased in size as shape became conical (data not shown). The QTL for shape on LG6 explained most of the variation (86 %), while the QTL for size on LG5 explained roughly a quarter of the variation (Table 2). The associated means for shape show an effect only in the “ac” genotype, and size has a positive or negative effect with alleles coming from the NC497 parent (Table 2).

Flower color

Most of the population had white petals, while very few individuals in the population had light pink or pink petals (Online Resource 1d). Both significant regions on LG6 and LG5 explain a high percentage of the variation. Single-point analyses show two and three highly significant markers on those linkage groups, respectively (Rub25a, E38M36_509.9, E45M35_312.0, E36M35_106.9, and E35M32_248.3; p ≤ 0.0001; Fig. 1). When color scores were entered as marker data, flower color grouped onto LG6; however, the linkage phase could not be determined.

Discussion

Molecular research in raspberry has been steadily growing, providing more potential tools that can be implemented in molecular breeding of Rubus. Currently, there are five genetic linkage maps available in raspberry (Bushakra et al. 2012; Graham et al. 2004, 2006; Pattison et al. 2007; Sargent et al. 2007; Ward et al. 2013). The most elaborate map to date has been developed by Ward et al. (2013), while that developed by Graham et al. offers extensive coverage of various QTL regions for numerous traits. Ward et al. (2013) report an extensive single nucleotide polymorphism (SNP) marker coverage of all seven LGs. The findings in that study represent a great potential for developing markers associated with traits, such as the QTL in this study, and for further application in MAS, especially since SNP markers allow maps to be compared, and can be screened easily and cost-effectively using high-resolution melting. Further studies should focus on discovering the nature of the SNPs, i.e. if these represent protein, structural, or other significant sequences. Here a sixth linkage map of raspberry is reported, and this was the first study to look at chilling requirement in Rubus for QTL analysis. The application of this study for developing cultivars tolerant to temperature fluctuations in winter could have an important impact on production, since growing regions would expand to the southeastern USA and places with similar climatic conditions. Additionally, other important traits had QTL regions associated to the linkage groups, and will help further research, as localization of QTL represents the first step for molecular characterization of genes underlying traits (Collard et al. 2005).

The map herein had a coverage of 2–5 mapping units in dense groups, which offers potential for development of markers for MAS. Most of the markers segregated for NC497. This is likely due to the integration of R. parvifolius that is contributing heterozygous regions in the progeny. Nevertheless, Qualicum has R. occidentalis in its parentage, but to a lesser extent (Daubeny and Kempler 1995). Similarities were found between the map reported herein and the map reported by Graham et al. (2006), but there were some differences in the linkage groups. For example, the original LG1 (OLG1, and LG7 in this study) reported in Graham et al. (2006) has a distance of 125 cM, while in this study the group was 66 cM. Although LG1 (OLG6 in Graham et al.) had limited marker coverage, the genetic distance agrees with the previously published map. LG4 (OLG7) was denser and slightly larger than that in Graham et al. (2006) but had less marker coverage than that reported in Graham et al. (2009).

Chilling requirements

Transgressive segregation occurred in the population. Transgressive segregation for chilling requirements has also been observed in open-pollinated progenies of the red raspberry cultivar Malling Exploit (Barrientos and Rodriguez 1980). This implies that different allelic interactions are likely involved in the chilling requirement trait. One QTL from the first and second seasons (Chilling1 and Chilling2) co-localized to the lower portion of LG6 and supports the involvement of this region in the chilling hour requirement trait (Fig. 1). Although the QTL regions were in close proximity, a QTL for two different seasons also supports the involvement of regions of LG1 and LG4 in chilling hour requirement. Graham et al. (2009) have studied different stages of reproductive development of red raspberry, from blooming to fruiting in the field and under row covers. In that study, QTL were associated with the different developmental and reproductive stages of raspberry; numerous QTL were associated with OLG3 (LG6), were scattered throughout the group, and had a strong effect on ripening. Regions in OLG5 (LG5) and OLG6 (LG1) were also associated with floral development and ripening; the RiMADS1 gene mapped to OLG5 and encodes a short vegetative phase (SVP)-type MADS box transcription factor (Mazzitelli et al. 2007). These same three linkage groups have co-localized QTL or are in close proximity in the the present study. For example, QTL on LG5 (OLG5) were associated with the upper portion in both maps where the RiMADS1 gene was mapped. In a colder season, RiMADS1 was associated with earlier flowering, and has sequence similarity to Prunus dam6 and SVP MADS box genes that negatively control the expression of FLOWERING LOCUS T (Graham et al. 2009). In this study, Chilling1 on LG5 was associated with lower chilling requirements. The findings also suggest a strong association between the two traits, since time of bud break would be closely linked with required chilling hours, and therefore confounded within floral phenology. Several studies have examined the genetics and genomics of blooming time as a quantitative character in Prunus (Ballester et al. 2001; Dirlewanger et al. 1999; Joobeur et al. 1998; Silva et al. 2005). Similar numbers of QTL affecting blooming time in Prubus have been identified; Joobeur et al. (1998) found four QTL affecting blooming time in four LGs (PLG1, PLG4, PLG6, and PLG7). Because blooming time in Prunus and chilling requirement in Rubus are measured by the same method, parallels among loci are expected. Three candidate genes were found to be linked to markers in QTL regions for blooming time: PrpFAR1, a Phytochrome A gene involved in the flowering time through the photoperiod pathway on PLG7; the Lb gene, a major gene involved in late blooming on PLG4; and prdMADS1, which mapped very close in position to the peach Evergrowing gene on PLG1 (Ballester et al. 2001; Silva et al. 2005; Wang et al. 2002). One study has done comparative mapping work between Prunus and Rubus (Bushakra et al. 2012); several Prunus markers found on PLG7 had homology to Rubus LG1, where QTL for both blooming time and chilling requirements were found. Similarly, some marker sequences in Rubus LG3 and LG6 had homology to Prunus sequences found on PLG6, both of which have QTL for floral development and chilling requirement, respectively. Therefore, regions in LG1 and LG6 play a significant role in chilling requirements and flowering traits. With sequencing efforts of the Rubus genome, these genes could be blasted to help better understand flowering in Rubus, and to develop tightly linked markers. Two markers were found associated with alleles for lowering chilling requirement in the NC497 parent. Nevertheless, the associated means for chilling requirement in LG4 and LG6 suggested that NC497 contributed alleles that increase the required chilling hours (Table 1). This is also supported by Ballington and Fernandez (2008), where backcrosses between red raspberry cultivars and R. parvifolius × red raspberry cultivars generated progeny with high tolerance to fluctuating winter temperatures. Further studies should focus on these marker regions as a potential tool in selecting for longer or shorter chilling requirements.

Prickle density

Although the trait has been previously mapped (Graham et al. 2004), completely prickle-free individuals were not present in the population for that particular study. The QTL for density on LG6 (OLG3) on the map reported herein was located at a similar genetic distance as the one reported by Graham et al. (2006), close to marker Rub103 (Fig. 1). The presence of prickles has been observed to be tightly associated with several fungal disease resistances in Rubus (Jennings 1988). Graham et al. (2006) found three QTL that were associated with fungal resistances overlapping the same region as the QTL for prickle density on OLG3 (LG6). Therefore, inferences on these fungal resistances can be made between the maps. However, a second QTL for prickle density reported by Graham et al. (2006) was not detected, and explained a large percentage of the variation (36 %; Graham et al. 2006), probably because this region is associated with gene H, which has a pleiotropic effect on prickle density (Jennings 1988), and the glabrous phenotype does not segregate in the (NC497) × Qualicum progeny. The prickle-free phenotype was attributed to gene s. The phenotypic distribution suggests that each of the parents was heterozygous for gene s as well as the means associated with the genotypes. Both parents are heterozygous for gene s because each grandparent of the parents carries recessive alleles for gene s; i.e. both Glen Prosen and Glen Moy are prickle-free cultivars (Daubeny and Anderson 1991; Daubeny and Kempler 1995; Jennings 1988). Tulameen, the male parent of NC497, is prickly and received the recessive allele from Glen Prosen, while Qualicum inherited the recessive allele from Glen Moy (Online Resource 1e). Because the QTL on LG4 was still significant after the prickle-free individuals were removed from the analysis, gene s had an effect on the prickle density. In addition, there are other genes present in R. parvifolius that are known to suppress the prickly phenotype and convert glands into soft prickles (Jennings 1988; Jennings and Ingram 1983). Although a closely linked marker for gene s is not necessary for selecting prickle-free canes, a molecular marker is an easy way of distinguishing homozygous dominant and heterozygous individuals in parental selections. In this case, marker Rub116a (with a segregation type ab × aa) is a plausible candidate; however, selections would have to be against the marker, since there was an allele associated with higher prickle density coming from the NC497 parent.

Fruit color

The mapping population is of a complex background involving three species, R. idaeus, R. parvifolius, and to a lesser extent R. occidentalis. Qualicum, the male parent, has R. occidentalis in its parentage (Daubeny and Kempler 1995; Jennings 1988), while NC497 is one-half R. parvifolius. Berries from both R. occidentalis and R. idaeus mostly contain cyanidin pigments, while R. parvifolius berries mostly contain pelargonidin as the major pigment (Jennings 1988; McCallum et al. 2010). The QTL on LG5 co-localized to the QTL for flower color, suggesting that traits for pigmentation in petals and in fruits are linked. A series of genes control the expression of anthocyanin concentrations in the fruit, but the most important one is gene T. Two recessive alleles of gene T (tt) give rise to yellow-colored fruit and stems with non-pigmented spines. Although completely yellow fruits were not present, approximately 24 % of the population exhibited yellow fruits with a pinkish blush. The mean association genotypes suggest that alleles from NC497 give rise to orange berry color (“bc” and “bd” genotypes), while also contributing alleles involved in the expression of pink and red berry colors (“ac” and “ad” genotypes), either by allowing or suppressing the expression of cyanidin. Another study has evaluated QTL analyses of fruit color in the same raspberry map developed by Graham et al., and looked at genetic as well as environmental effects of fruit color (McCallum et al. 2010). In that study, berry color was measured in field, commercial, and protected trials and showed QTL regions associated with the middle and lower portion of OLG6 (LG1). OLG6 (LG1) was also associated with color in this study, but the significant region was in the upper portion of the group. Although McCallum et al. (2010) found QTL regions associated to several groups, no QTL were found on OLG5 (LG5). One of the major differences in this study was the color variation; McCallum et al. (2010) had color differences only in the reds, thus measuring the effects of cyanidin but not pelargonidin.

Fruit shape and size

Genes for fruit size in raspberry have been examined in Norfolk Giant, Baumforth’s Seedling B, and a mutant cultivar of Malling Jewel (Jennings 1966). A single mutation in Malling Jewel conferred larger fruiting laterals, flowers, and fruit by “compromising” the first year’s growth (Jennings 1966, 1988). This shows that fruit size is influenced by a single gene or a few genes, and this is reflected in the QTL analysis and the distribution of the trait. In the QTL analysis, alleles for larger fruit size are inherited by NC497, or NC497 has alleles that inhibit larger fruit size. Fruit shape did not change over the two seasons, as observed in another study (Haskell 1960; Jennings 1988). Little is known about the genetics of fruit shape, although a conical shape is the industry standard. Nevertheless, the QTL with a large effect in LG6 explained most of the variation and, as is the case with fruit size, implied the involvement of a single gene or a few genes.

Flower color

The inheritance of a gene for pink petals, gene An, has been studied in R. coreanus (Keep et al. 1977), where gene An is dominant for pink petals. However, in R. parvifolius, one of the parents of NC497, a case for a dominant gene is not likely because a 1:1 ratio of white to pink petals or all pink petals would be expected. Instead, less than 10 % of the population expressed pink petals, and almost 17 % expressed light pink petals. Moreover, petal color in R. parvifolius involves a series of allelic genes that confer different intensities of pink (Jennings 1988), and this was exposed in the QTL analysis, where two different regions were involved (LG6 and LG5).

Conclusion

Appropriate chilling requirement is dependent upon geographical location, and requires breeders to select for each trait to produce cultivars adapted to particular winter conditions. In southeastern states, such as North Carolina, plants will require longer periods of rest completion in the winter to avoid the consequences of severe fluctuations in temperature. To facilitate the breeding of this trait, molecular tools should be considered to expedite the breeding process. This study has established a baseline for the development of these tools. Four QTL associated with chilling requirement have been identified, and in most cases are consistent in different seasons. One QTL associated with prickle density has been corroborated with previous studies, and gene s, which encodes for prickle-free canes, has been mapped to the upper region of LG4. Additionally, significant regions for fruit color, fruit shape, fruit size, and flower color have been identified and show potential for improving fruit quality by MAS. One future strategy that could further strengthen and accelerate the development of these tools is more comparative mapping studies with the well-established Prunus and Malus maps. Thus comparative mapping is encouraging for further development of the QTL regions, significant markers for MAS, and the sequencing of the Rubus genome.