Introduction

The domesticated apple (Malus x domestica Borkh.) has been distributed into diverse climatic conditions worldwide for commercial production of fruit. Apple trees need exposure to cold temperatures, referred to as chill unit (CU) accumulation during winter, in order for budbreak to occur promptly and uniformly after winter (Cook and Jacobs 2000). In warmer production areas, such as the Western Cape region of South Africa, the application of dormancy breaking chemicals, forming part of standard orchard management, enable successful production of high chilling requiring apple cultivars in suboptimal environmental conditions. Failure to apply dormancy-breaking chemicals can result in prolonged dormancy symptoms, which include extended rest, less synchronized breaking of buds, and reduced branching (Labuschagné et al. 2002b). An increasing awareness of both global temperature increase and the negative effects associated with the use of chemical sprays (for both pest and disease resistance and growth regulation) has resulted in the need to breed cultivars better adapted to current and future environmental conditions.

The breeding of new cultivars using conventional breeding methods is a time-consuming process, especially in perennial tree species with a long juvenile phase such as apple. Markers linked to genes involved in apple disease resistance for a variety of pests and pathogens have been identified (Gardiner et al. 2007) and are already in use in breeding programs (Kellerhals et al. 2008; Tartarini and Sansavini 2003; Tartarini et al. 2000), through the implementation of marker-assisted breeding (MAB) and selection (MAS) that enables the selection of favorable genotypes at a very early seedling stage. The genetic determinants of dormancy-related characteristics, such as time of initial vegetative budbreak (IVB), are still poorly understood, and this hampers the genetic improvement of such characters using MAB. Dormancy characteristics can be controlled by factors residing within the bud itself, referred to as endodormancy, by factors in the plant but outside of the bud (paradormancy) and control by environmental factors (ectodormancy; Khan 1997; Lang et al. 1985). Although our study focused on time of IVB, a character related to endodormancy (Bradshaw and Stettler 1995), various other characteristics can be associated with dormancy, such as position and number of budbreak and budbreak duration.

Unravelling of the genetic basis of complex traits such as dormancy can be undertaken through the construction of a genetic linkage map followed by quantitative trait loci (QTL) identification (Falconer and Mackay 1996; Young 1996). A first attempt towards understanding the genetic control of ‘leaf break’ in apples through the identification of QTLs, was performed by Conner et al. (1998) using a population of 172 trees derived from a cross between ‘Wijcik McIntosh’ and NY 75441-58. Eight genomic regions on seven linkage groups (LGs) could be associated with time of budbreak. The genetic linkage map constructed during their investigation, however did not include transferable simple sequence repeat (SSR) markers, resulting in their inability to align this map with the now more commonly used LG numbering for apple genetic linkage maps (Maliepaard et al. 1998). Further investigation resulted in alignment of three LGs from these two maps, including one (LG 3) that was homologous to LG 9 of Maliepaard et al. (1998) and carried a QTL for leaf break (Kenis and Keulemans 2004). More recently Segura et al. (2007) used 123 seedlings derived from a cross between ‘Starkrimson’ and ‘Granny Smith’ to identify two QTLs for time of budbreak. The first on LG 8, corresponded to that identified on the corresponding LG 7 by Conner et al. (1998) (see Kenis and Keulemans 2004). The second QTL for time of budbreak identified by Segura et al. (2007) was on LG 6.

In the present study, genetic linkage maps were constructed for two mapping pedigrees with the low chilling requiring cultivar ‘Anna’ as common male parent. ‘Anna’ is one of only a few cultivars worldwide characterized by a low chilling requirement (CR) and with ‘Dorsett Golden’ was reported as varieties needing less than 300 h of chilling in Southern California (http://ucce.ucdavis.edu/files/filelibrary/5764/33384.pdf) and North and North Central Florida (Andersen and Crocker 2000). Both published SSR markers (Celton et al. 2009; Guilford et al. 1997; Hemmat et al. 2003, 1997, Liebhard et al. 2002; Silfverberg-Dilworth et al. 2006; Yamamoto et al. 2002a, b) and 116 new SSR markers, developed from expressed sequence tags (ESTs), were used for the construction of the genetic linkage maps used to identify a major QTL for time of IVB on LG 9.

Materials and methods

Plant material

Two F1 progenies, derived from crosses between the low chilling ‘Anna’ (common male parent) and the higher chill requiring ‘Golden Delicious’ (population A) and ‘Sharpe’s Early’ (population B), containing 87 and 92 individuals, respectively, were used. Seedlings from population A were kept in seedling bags under shade netting in Groot Drakenstein (Western Cape, South Africa; 33°50′36″ S 18°58′39″ E). Seedlings in this population were cut back and re-grown to single shoots on a seasonal basis and no chemical treatment was used to induce budbreak. Seedlings from population B were planted in an orchard in Vyeboom (Western Cape, South Africa; 34°4′15″ S 19°4′47″ E) characterized by low winter chilling. Resulting trees were in their fifth growing season at the onset of this investigation. Seven clonal replicates from seedlings in population B and the two parental cultivars were grafted onto rootstocks (M793) and planted in seven randomized blocks in an adjacent orchard (34°8′21″ S 19°0′44″ E). Both sites are characterized by warmer winters and fluctuating chilling accumulation between winters. At these sites CU accumulation varies between 500 and 1,000 CU annually. Chill units were calculated according to a modified Utah model found to be more suitable for local chilling conditions where negative CU values are not carried from one day to the next (Linsley-Noakes et al. 1994). Orchard management of adult and juvenile clonal trees from population B were typical of commercial practice, except that no pruning and tree growth manipulations, such as dormancy breaking chemicals, were applied.

Phenotypic assessment

The time of IVB was scored as the day on which the first green leaves emerged from the vegetative buds (day 1 being January 1; Labuschagné et al. 2002a, b). Phenotypic trait assessments were performed four times over a period of 6 years (1999, 2000, 2002, and 2004) on the 87 seedlings from population A. Trait assessment of population B was first performed during a 3-year period, from 1996 to 1998, on 60 adult trees, initially in their fifth growing season, followed by trait assessment on the seven clonally replicated juvenile trees of all 92 siblings from 1998 to 2000. The data obtained from population B has been used in previous studies (Labuschagné et al. 2002a, b) during which broad sense heritability of IVB was estimated between 0.62 and 0.92 in clonal trials on young seedlings and between 0.57 and 0.83 for adult seedling trees. We calculated Pearson’s correlation coefficients to determine the relationship between different years of phenotypic trait assessment.

DNA extraction

Extraction of seedling and parental cultivar DNA were performed using the hexadecyl trimethyl ammonium bromide (CTAB) method described by Doyle and Doyle (1990) with the addition of polyvinylpyrollidone (Kim et al. 1997) in order to bind secondary plant products such as polyphenolics.

SSR marker development and implementation

Unigene sets obtained from the large public EST database (>240,000; http://www.ncbi.nlm.nih.gov; Naik et al. 2006; Newcomb et al. 2006) for Malus, were searched for SSRs using the Tandem Repeats Finder algorithm (Benson 1999). SSRs were selected based on length of the repeat unit, number of repeats (>10 for di-, >7 for tri-, >5 for tetra- and penta-, and >3 for hexanucleotide repeats) and length of sequences flanking SSR regions. Conserved sequences flanking 196 selected SSRs (100 di-, 60 tri-, 25 tetra-, 5 penta-, and 6 hexanucleotide repeats) were used to design primers resulting in amplicons ranging between 100 and 450 bp in length. Newly developed SSR markers were tested on the three parental cultivars, ‘Anna’, ‘Sharpe’s Early’, and ‘Golden Delicious’.

Markers for map construction were selected based on map position as well as heterozygosity observed during previous studies. They included 238 previously published SSR markers (Celton et al. 2009; Guilford et al. 1997; Hemmat et al. 2003, 1997; Liebhard et al. 2002; Silfverberg-Dilworth et al. 2006; Yamamoto et al. 2002a, b), marker AG11 (unpublished data, A. Patocchi (ETH-Zürich, CH)) and marker Md-EXP7 (Costa et al. 2008) and were initially screened for polymorphism over the three parental cultivars, ‘Anna’, ‘Sharpe’s Early’, and ‘Golden Delicious’.

SSR analysis

All SSR markers implemented in mapping populations were fluorescently labeled and up to 16 markers were multiplexed using both size and fluorescent dye (6-FAM™, VIC™, NED™, and PET™) differences. Polymerase chain reaction (PCR) reactions were performed using the Qiagen multiplexing kit (QIAGEN Ltd., West Sussex, RH10 9NQ) according to the manufacturer’s instructions. Resulting PCR products were prepared for capillary electrophoresis by adding 1 μl of a 1:10 diluted PCR product to 10 μl Hi-Di formamide containing 0.15 μl GeneScan™-500 LIZ™ size standard (Applied Biosystems). Genotyping was performed using the ABI Prism 310 and 3130 (16-capillary array system) Genetic Analyzers (Applied Biosystems, Foster City, CA, USA). Data collection and analysis were performed using GeneMapper®4 software (Applied Biosystems, Foster City, CA, USA).

Genetic linkage map construction

For both progenies, parental genetic linkage maps and integrated genetic linkage maps were constructed using JoinMap®4 (Van Ooijen 2006). A logarithm of the odds (LOD score) of 4 was used to define LGs and genetic distances between markers were calculated using the Kosambi mapping function. On the basis of previously mapped SSRs, LGs were numbered in accordance with the 17 LGs obtained by Maliepaard et al. (1998).

QTL analysis

QTL analysis was performed using MapQTL®5 (Van Ooijen 2004) using the average phenotypic value for the 4 years of phenotypic trait assessment performed on population A and the two 3-year periods of trait assessment performed on adult and juvenile trees from population B. Analyses were also performed separately for each year of phenotypic trait assessment and in the case of clonal replicates, the mean value per genotype was used. Regions with potential QTL effects were identified using interval mapping with a step size of 1 cM. QTLs were declared significant if the maximum LOD, obtained after multiple rounds of MQM mapping, exceeded the genome wide (GW) LOD threshold (calculated with an error rate of 0.05 over 1,000 permutations). QTLs were characterized by the maximum LOD score and the percentage of phenotypic variation explained. For each QTL the differences in mean time of IVB associated with the different genotypic classes, ac, ad, bc, and bd, derived for an ab x cd cross, are reported. QTLs were graphically displayed as bars next to the LGs on which they were identified, with bars corresponding to a 95% confidence interval (LOD score drop of 0.5) and dotted lines corresponding to a 90% confidence interval (LOD score drop of 1).

Results

Phenotypic trait assessment

Bi-modal distribution patterns were observed during most years of phenotypic trait assessment (Fig. 1). The distribution patterns indicate budbreak was occurring earlier during consecutive years as trees matured. Significant levels of correlation were found between the different years during which phenotypic trait assessment were conducted (Table 1). High broad sense heritability values (h 2 = 0.69) for IVB were calculated by Labuschagné et al. (2002a).

Fig. 1
figure 1

Histogram showing the distribution of time of initial vegetative budbreak (IVB) observed in 1 ‘Golden Delicious’ x ‘Anna’ seedlings during 4 years of phenotypic trait assessment a 1999, b 2000, c 2002, and d 2004; 2 ‘Sharpe’s Early’ x ‘Anna’ juvenile trees during 3 years of phenotypic trait assessment a 1998, b 1999, and c 2000; and 3 ‘Sharpe’s Early’ x ‘Anna’ adult trees during 3 years of phenotypic trait assessment a 1996, b 1997, and c 1998

Full size image
Table 1 Pearson’s correlation coefficients indicating phenotypic association (P < 0.0001) between different years for time of initial vegetative budbreak (IVB)
Full size table

SSR marker development and implementation

The amplification success of newly developed SSR markers was 86% (168 SSRs from a total of 196). From these, a total of 116 new SSR markers were polymorphic in at least one of the three parental cultivars used and were mapped in one or both mapping populations (Table 2). Of the 240 previously published markers, including 238 SSR markers (Celton et al. 2009; Guilford et al. 1997; Hemmat et al. 2003, 1997; Liebhard et al. 2002; Silfverberg-Dilworth et al. 2006; Yamamoto et al. 2002a, b), marker AG11 (unpublished data, A. Patocchi (ETH-Zürich, CH)) and marker Md-EXP7 (Costa et al. 2008), 232 markers yielded amplification products of which 204 markers were heterozygous in one or more of the three cultivars tested.

Table 2 Summary of 116 new SSR markers, accession number, repeat motif, primers, resulting fragment sizes and genetic linkage map position
Full size table

Designing new SSR markers so that the resulting amplicons vary in size, enabled effective multiplexing of up to 16 markers in one PCR reaction, greatly reducing the cost involved in the screening of mapping populations. Markers used within each multiplex are very flexible when using the QIAGEN multiplexing kit (QIAGEN Ltd., West Sussex, RH10 9NQ) that provides optimal reaction conditions that increases specificity and minimizes the effect of primer-dimers and non-specific artifacts often associated with multiplex PCR reactions. The ease with which different multiplexes could be assembled enabled easy assembly of new multiplexes containing highly informative markers for each specific mapping pedigree.

Genetic linkage map construction

The four parental maps constructed (Fig. 2) enabled the positioning of 286 SSR markers on 17 LGs corresponding to the number of chromosomes in the apple haploid genome. The number of SSR markers per LG range from ten SSR markers on LG 3 to 28 SSR markers on LG 10, with an average of 17 SSR markers per LG. The positioning of the 116 newly developed SSR markers (Table 2) range from two SSR markers on LG 1 to 15 SSR markers on LG 10.

Fig. 2
figure 2figure 2

Parental genetic linkage maps of ‘Golden Delicious’ (GD) and ‘Anna’ from population A and ‘Sharpe’s Early’ (SE) and ‘Anna*’ from population B. Numbering of LGs are according to Maliepaard et al. (1998). Newly developed SSRs are prefixed by SAms and are indicated in bold, italic, and underlined

Full size image

Genetic linkage map construction allowed the positioning of five previously published but unmapped markers (Liebhard et al. 2002). CH01b09b was mapped to LG 4, CH01e09b was mapped to LG 10 and CH02h11b was mapped to LG 12 in both mapping populations. CH01e121 was mapped to LG 8 and CH05c02 was mapped to LG 11 in the ‘Golden Delicious’ x ‘Anna’ mapping population. Three markers were mapped to different LGs when compared to their location on previously published maps: (1) CH03e03 was mapped to LG 5 compared to LG 3 (Liebhard et al. 2002), most likely due to the amplification of a different locus as observed fragment sizes are slightly larger than published (a fragment size of 216 bp was observed in ‘Prima’ compared to the published 186 bp), (2) Hi23g12 was mapped to LG 15 compared to LG 8 (Silfverberg-Dilworth et al. 2006) confirming results obtained by Patocchi et al. (2009); (3) CH05d04 was mapped to LG 5 compared to LG 12 (Liebhard et al. 2002), also most likely due to the amplification of a different locus as observed fragment sizes are slightly smaller than published (fragments of 154 and 175 bp were observed in ‘Prima’ compared to the published 176 and 186 bp. The marker CH05g07 (Liebhard et al. 2002) was found to amplify two loci, both mapping to LG 12. A locus amplified by the marker Hi03a03 (Silfverberg-Dilworth et al. 2006) was confirmed to map onto LG 6 in both mapping populations used while a second locus amplified by the same marker was found to map to LG 14 in the ‘Anna’ x ‘Sharpe’s Early’ mapping population, confirming structural homology between LG 6 and LG14 (Celton et al. 2009).

Population A

Of the 285 SSR markers screened on 87 seedlings from the ‘Golden Delicious’ x ‘Anna’ mapping pedigree, 260 markers were positioned on the integrated F1 genetic linkage map (map coverage: 1,376.7 cM). Genetic linkage maps constructed for the parental cultivars ‘Golden Delicious’ (map coverage: 1,124.5 cM) and ‘Anna’ (map coverage: 1,292.6 cM) consisted of 163 (including 72 new SSRs) and 170 (including 71 new SSRs) markers, respectively. Parental maps were aligned using 92 SSR markers in common (Fig. 2).

Population B

The ‘Sharpe’s Early’ x ‘Anna’ genetic map was constructed using 230 SSRs genotyped over the 92 F1 seedlings. The integrated F1 genetic linkage map (map coverage: 1,242.6 cM) consisted of 207 mapped SSR markers. Genetic linkage maps constructed for the parental cultivars ‘Sharpe’s Early’ (map coverage: 1,012.9 cM) and ‘Anna’ (map coverage: 1,050.6 cM) consisted of 127 (including 41 new SSRs) and 126 (including 45 new SSRs) markers respectively. Parental maps were aligned using 79 SSR markers in common (Fig. 2). The parental map constructed for ‘Anna’ has 94 SSR markers in common with the parental map for ‘Anna’ constructed for population A.

QTL detection and mapping

A single major QTL for time of IVB was detected on LG 9 (Fig. 3). Analyses performed on the average time of IVB for the different populations and developmental stages showed that this QTL exceeded the GW LOD threshold during phenotypic trait assessment performed on adult trees from population B. LOD scores obtained for the analyses performed on averages from population A and juvenile trees from population B were just below the GW LOD thresholds. Separate QTL analysis for the different years of phenotypic trait assessment performed on seedlings from population A resulted in GW LOD thresholds being reached during trait assessment performed on seedlings in their fourth (2002) and sixth (2004) year (Table 3). GW LOD thresholds were exceeded during all 3 years phenotypic trait assessment has been performed on adult trees from population B (Table 3). Separate QTL analysis for the three different years of phenotypic trait assessment performed on juvenile trees from population B resulted in GW LOD thresholds not being reached during the first three juvenile years (Table 3). One-way analysis of variance indicated significant association (P < 0.0001) between specific NZmsCN943946 alleles inherited from the parental cultivar ‘Anna’ and time of (IVB). This association was true during all years of phenotypic trait assessment on ‘Golden Delicious’ x ‘Anna’ (30.22<F > 91.73) and ‘Sharpe’s Early’ x ‘Anna’ adult (34.39 < F > 49.9) and juvenile (30.6 < F > 69.27) trees.

Fig. 3
figure 3

Position of the QTL for time of IVB detected on LG 9 of the consensus ‘Golden Delicious’ x ‘Anna’ (GDxAn) map and the ‘Sharpe’s Early’ x ‘Anna’ (SExAn) map. QTLs are represented by boxes where the length of the box corresponds to a 5% confidence interval and extended lines to a 10% confidence interval. Boxes representing average time of IVB are filled and boxes representing time of IVB for separate years are open. Boxes indicating QTL detected on juvenile trees are green and those indicating QTL detected on adult trees are red

Full size image
Table 3 Parameters associated with the QTL for time of initial vegetative budbreak (IVB) identified on LG 9 of the consensus map used for population A and population B, using multiple QTL mapping (MQM)
Full size table

Differences in time of IVB associated with the four genotypic classes, ac, ad, bc, and bd, derived from an ab x cd cross, indicate that the phenotypic variation can be associated with alleles inherited from the common male parent ‘Anna’. This QTL explains between 4.8% and 40.1% of the phenotypic variation observed in population A and between 11.9% and 44.6% of the phenotypic variation observed in population B.

Discussion

The genetic linkage maps constructed are composed entirely of SSR markers and since a very large proportion of these markers are derived from EST sequences (more than 120) these maps are the most functional maps yet available. The newly developed and mapped SSR markers will enable the expansion of the 15 cM reference map, currently consisting of 86 SSR markers covering 85% of the genome, proposed by Silfverberg-Dilworth et al. (2006) with up to 11 SSR markers. Depending on polymorphic information content determined on a larger number of cultivars, some of the newly developed SSR markers might be used to replace markers with low polymorphism now included in the reference set, due to lack of more polymorphic SSR markers in certain regions (Silfverberg-Dilworth et al. 2006).

The time of IVB showed a wide bi-modal distribution in the seedlings derived from both mapping populations. Although bi-modality could be explained by seedlings having a difference in their rapidity of response to favorable conditions after their CR was satisfied (Labuschagné et al. 2003), the distribution of time of IVB can be explained by the fact that the trait is controlled by a major QTL together with some minor QTLs. High heritability estimates, although specific to the experimental conditions in which they have been calculated, were calculated for time of IVB by both Labuschagné et al. (2002a) (h 2 = 0.69) and Segura et al. (2007) (h 2 = 0.58), indicating that the trait has a strong genetic influence and that it can be selected for using marker-assisted selection. Heritability is not always related to the power of QTL detection (Segura et al. 2007), as the latter is also influenced by population size and the number of QTLs affecting the trait. The small number of individuals included in phenotypic trait assessment (87 from ‘Golden Delicious’ x ‘Anna’ and 60 and 92 for adults and juveniles from the ‘Sharpe’s Early’ x ‘Anna’ mapping pedigrees) and the amount of variation observed among seedlings from the same mapping population, allowed for the detection of only one QTL with large effect. The fact that this QTL explains up to 40.1% and 44.6% of the phenotypic variation observed in populations A and B, respectively, indicates that there are further QTLs affecting time of IVB. These may include several QTLs with smaller effect that are statistically not detectable due the restricted population sizes used and the phenotypic variation observed in the seedlings. During initial interval mapping (van Dyk et al. 2009), the involvement of several minor QTLs were suggested. Implementation of more markers leading to better genome coverage and the ability to perform MQM analysis enabled the identification of a QTL with large effect in the current study.

Genetic linkage maps constructed for both mapping populations enabled the efficient detection of a major QTL affecting the time of IVB on LG 9 (Table 3). This QTL may co-localize with one of eight QTLs involved in leaf break that was identified by Conner et al. (1998). The QTL identified on LG 3 of the genetic linkage map produced by Conner et al. (1998) was, after the implementation of transferable microsatellite markers, shown to be homologous to the LG now known to be LG 9 (Kenis and Keulemans 2004). In the present study, the QTL on LG 9 can be associated with specific allele inheritance from the common parent ‘Anna’. Performing QTL analyses on an integrated parental map when working with an outbreeder, as was done during this study, enables the determination of both the effect of alleles inherited from a single parent and the interaction between alleles inherited from both parents. Results (Table 3) indicated a clear difference in average time of IVB between seedlings that inherited allele “c” from ‘Anna’ (average “ac” and “bc”) compared to seedlings that inherited allele “d” from ‘Anna’ (average “ad” and “bd”). No clear difference could be detected between seedlings that inherited different alleles from the other parental cultivar involved in each mapping pedigree or seedlings with a specific combination of parental alleles.

The power of QTL detection (LOD score) increased during consecutive years of phenotypic trait assessment being performed on seedlings from population A and juvenile trees from population B. This suggests that although the QTL can be associated with time of IVB in young seedlings, the association between the QTL and the trait becomes stronger as the tree matures. The QTL was found to be significant (LOD score exceeding GW LOD thresholds) in all 3 years during which phenotypic trait assessment was performed on adult trees from population B (Table 3). Although significant GW LOD thresholds are not met in juvenile trees from population B, the association between the QTL and time of IVB can be seen from obtained phenotypic means associated with each of the genotypic classes (Table 3). Budbreak occurring earlier as trees mature has not been reported before. Preliminary results suggest no correlation between the earlier time of vegetative budbreak, associated with seedling age in two apple populations studied, and the CU accumulated during different years. These results suggest that the CR, which is the major determinant of time of budbreak (Bradshaw and Stettler 1995), has been met and that the time of vegetative budbreak is also influenced by factors associated with tree age. These results need to be confirmed in future studies, including several years of phenotypic trait assessment performed during different developmental stages and on different populations.

Markers linked to the QTL identified will be used in a validation test on a larger progeny sharing common parentage. The QTL region will be saturated with markers selected for their positioning on the genetic linkage map as a result of selective (bin) mapping on a subset of individuals (van Dyk and Rees 2009). The ideal will be the identification of markers flanking the QTL that can be used for the implementation of MAS in breeding for cultivars that are better adapted to local climatic conditions.