Introduction

The demand for new varieties with improved and novel quality traits means that the fruit breeder is continually on the lookout for methods and strategies to improve the efficiency and to reduce the time in which such varieties can be produced. As a result, interest has increasingly focused on the use of molecular markers in fruit breeding programmes. In principle, such markers can be utilised at an early (seedling) stage to select for individuals with a favourable combination of several (fruit) quality traits of interest. There are a number of examples of the use of these techniques of marker-assisted selection (MAS) and marker-assisted breeding (MAB) in cereal crops (Bouchez et al. 2002; Masood et al. 2004; Xu et al. 2005; Bortiri et al. 2006; Vinod et al. 2006; Wan et al. 2006; Yoon et al. 2006; Song et al. 2007; Stangoulis et al. 2007). However, MAS holds particular promise in tree-fruit breeding, where the long juvenile period of these species and the polygenic nature of fruit quality traits are major bottlenecks in conventional breeding programmes, meaning that the progeny resulting from a cross have to be maintained for a long time before fruit quality traits can even be evaluated (Etienne et al. 2002; Liebhard et al. 2003a, c). In addition, because members of the Rosaceae such as apple and pear are highly heterozygous and obligate cross-pollinators, large numbers of progeny need to be screened to identify promising genotypes. Therefore, any technology which can speed up or focus this selection process can have substantial economic benefits.

For all these reasons, trait-mapping experiments to identify quantitative trait loci (QTLs) co-segregating with important fruit quality traits have become popular as a means to help unravel the mechanisms underlying the genetic control of important characteristics in apple (King et al. 2000, 2001; Liebhard et al. 2003a). Much of the initial excitement over the potential benefits of QTLs however has been tempered by the realisation that the results of individual QTL mapping experiments need to be validated over different years, growth sites, generations and genetic backgrounds, and currently, the results of such studies are still a long way from being implemented as a part of standard breeding practices. By comparison, MAS and MAB for monogenic traits such as apple scab (Venturia inaequalis) resistance are already in use (Xu et al. 2001; Gygax et al. 2004; Vinatzer et al. 2004; Patocchi et al. 2005).

An overview of the apple fruit quality trait QTLs identified to date and their localisation on the integrated Malus (Fiesta × Discovery) ‘reference’ linkage map, is provided by Liebhard et al. (2003a) and includes QTLs for fruit sugar and acidity levels, which are important determinants of fruit taste. The same population was also used by King et al. (2000, 2001) to localise several QTLs for apple fruit flesh texture. Before these results can be developed into practical breeding tools, however, it is necessary to compare QTL positions for these traits in different populations and to use common transferable markers to align the genetic linkage maps of the different parents. For these reasons, genetic linkage maps containing microsatellites (simple sequence repeats (SSRs)) are indispensable in MAS and/or MAB programmes and further developments in this direction will undoubtedly be aided by the recent availability of large public Malus expressed sequence tag (EST) collections (Newcomb et al. 2006).

In this manuscript, we report on the further saturation of the genetic linkage maps of the apple cultivars, ‘Telamon’ (T) and ‘Braeburn’ (B) with microsatellite markers to obtain an alignment (and renumbering) of all the linkage groups (LGs) to correspond with the integrated ‘Fiesta’ (F) and ‘Discovery’ (D) reference map published by Liebhard et al. (2003b). We also present the results from two successive years of QTL analysis for all major physiological fruit quality traits in our T×B population and discuss the importance of the stability of the detected QTLs. Finally, the localisation of these QTLs is compared with results obtained for the same traits in the ‘Fiesta’ × ‘Discovery’ (F×D) and the ‘Prima’ × ‘Fiesta’ (P×F) mapping populations (King et al. 2000; Liebhard et al. 2003b) to try and estimate the influence of genetic backgrounds on the QTL identification and localisation.

Materials and methods

Plant materials

A cross between the apple cultivars ‘Telamon’ (T) and ‘Braeburn’ (B), with ‘Telamon’ as the female parent, was carried out in spring 1998, resulting in a T×B F1 population of around 250 individuals (Kenis and Keulemans 2005). ‘Telamon’ is a descendant of ‘Wijcik’ × ‘Golden Delicious’ and has inherited the dominant monogenic columnar growth habit of ‘Wijcik’. It produces sweet red fruits which are mostly used for processing because post-harvest storage qualities are poor. ‘Braeburn’ is a chance seedling from New Zealand, which by comparison produces juicy, crisp and firm fruits with high vitamin C contents and excellent post-harvest storage properties (Davey and Keulemans 2004; Davey et al. 2007). In 2004, when these QTL experiments were started, the T×B population was 5 years old. Each genotype was represented by a single non-grafted tree maintained in Rillaar, Aarschot (Belgium) and were cultivated using a standardised integrated pest management system as utilised in the Belgian commercial sector. Pruning and thinning was performed only when necessary to maintain a planting distance of 3.5 by 1.3 m.

Harvesting

Fruit from individual genotypes were harvested once a week when considered fully mature. Apples were considered ready to be picked when they came off the tree when gently twisted and when the seeds had a dark brown colour. Fruit were also tasted to estimate starch–sugar conversion. In 2004, the population harvest lasted from between 7th of September and 3rd of November, while in 2005 fruit were collected between 5th of September and 31st October. In general, 10–20 randomly selected fruit per genotype were collected for the fruit quality measurements. Of the original 257 T×B genotypes used to create the ‘Telamon’ and ‘Braeburn’ genetic linkage maps (Kenis and Keulemans 2005), 233 trees were still alive in 2004, and of these 199 produced more than ten fruits and could thus be included in the analyses. By comparison, in 2005, only 165 genotypes produced sufficient fruit.

Fruit quality measurements

Fruit quality measurements were carried out on the same day of harvest using both nondestructive (fruit height, diameter, weight and the stiffness) and destructive (firmness, sugar content, acidity and rate of flesh browning) procedures. In all cases, measurements represent the mean values obtained from the analysis of at least ten randomly chosen fruit per genotype. Fruit height and diameter were measured with digital callipers and stiffness by acoustic resonance with weight corrections as according to Chen and De Baerdemaeker (1993). Flesh firmness was determined on both the green (shaded) and the red (sunny) side of the fruit using an automated penetrometer (TA-XT2i Texture Analyser, Stable Micro Systems) fitted with an 11-mm diameter plug. Simultaneously, the juice released by penetration of the probe was used to measure the sugar content (ºBrix) using a digital refractometer (Pocket PAL-1, Atago, Japan). The rate of flesh browning was determined as the time between the peeling of a small strip of the skin from the fruit (for the firmness measurements) and the moment that the peeled surface became brown in the majority of the ten fruit and as evaluated at regular time points by two independent investigators. For acidity measurements, the juice of ten bulked apples per genotype were collected and combined. Twenty-five milliliter of this sample was diluted with 25 ml of distilled water and the titratable acid content determined with a 719S Titrino autotitrator (Metrohm, Switzerland).

Data analysis

Statistical analysis of data was carried out to determine the variation, to test distributions and to calculate correlation coefficients between the individual traits, as well as to see if the results of the two harvests were significantly different from each other. All statistical analyses were carried out using the commercially available software packages SAS® (release 9.1) and Microsoft Excel 2003.

Genetic linkage maps

The existing genetic linkage maps of ‘Telamon’ and ‘Braeburn’ (Kenis and Keulemans 2005) were further saturated with 36 additional microsatellite loci using sequences obtained from Liebhard et al. (2002; Table 1) and the JoinMap 3.0 software (Van Ooijen et al. 2002) and as previously described (Kenis and Keulemans 2005). However, the LG numbering was adjusted to correspond to the numbering of the LGs of the integrated Malus reference map published by Liebhard et al. (2003b) and which also corresponds to the labelling of the genetic linkage maps of Maliepaard et al. (1998). Linkage maps were drawn with the MapChart 2.1 Software.

Table 1 Results of the SSR analysis of ‘Telamon’ and ‘Braeburn’ with 33 SSR primer pairs
Full size table

QTL analysis

QTL analyses were carried out using all markers of the updated genetic linkage maps of each parental cultivar. The MapQTL v4.0 software (Van Ooijen et al. 2002) was used to perform interval mapping (IM), in combination with restricted multiple QTL model mapping (rMQM), as previously described (Davey et al. 2006; Kenis and Keulemans 2007). A logarithm of odds (LOD) score threshold of 3.0 was used to define QTL significance and an LOD level of >3.5 was set for co-factor selection in subsequent rounds of rMQM analysis. QTLs identified in this way were described by the marker with the highest LOD score in the corresponding QTL region. An estimation of the total variance explained by this marker was given by MapQTL 4.0 software and QTL regions were defined at the 1.0 and 1.5 LOD support interval, which corresponds to a QTL coverage probability of approximately 95%.

Results and discussion

Genetic linkage maps

A total of 257 individuals of an F1 progeny were used to construct genetic linkage maps of ‘Telamon’ and ‘Braeburn’ (Kenis and Keulemans 2005). To improve the alignment with other published maps, an additional 33 microsatellite (SSR) markers were mapped, resulting in 36 additional loci because three primer pairs produced two sets of bands. Twenty-seven of these loci were mapped onto the ‘Telamon’ map and 28 onto the ‘Braeburn’ map. The SSR markers were labelled as according to Liebhard et al. (2002) except for when additional bands were produced, in which case an extension of the original name was used, e.g. CH04c06a and CH05b06b. An overview of marker properties is given in Table 1 and a schematic overview of the final ‘Telamon’ and ‘Braeburn’ linkage maps is provided in Fig. 1. Additional map and marker details are given in Table 2.

Fig. 1
figure 1figure 1

Genetic linkage maps of ‘Telamon’ (TEL) and ‘Braeburn’ (BR). Homologous LGs of both maps are presented side by side with SSR markers in bold type. AFLPs heterozygously present in both cultivars are underlined and anchoring markers used to help align homologous LGs are connected with a line. LG numbers and orientation have been adjusted to those of the apple reference map (Liebhard et al. 2003b)

Full size image
Table 2 Overview of the number, density and type of markers positioned on the genetic linkage maps of ‘Telamon’ and ‘Braeburn’, including common markers and the largest gap between markers in each map
Full size table

The additional microsatellite markers allowed us to improve the alignment of the ‘Telamon’ and ‘Braeburn’ genetic linkage maps, both with each other as well with the homologous LGs of ‘Fiesta’ and ‘Discovery’ (Liebhard et al. 2003b). Correct orientation of the LGs was based on the order of the SSRs as well as the position of shared AFLP markers when available. With the inclusion of 36 additional SSRs, mean map marker densities were increased to one marker every 3.78 and 4.39 cm in ‘Telamon’ and ‘Braeburn’, respectively, which is only slightly lower than the marker densities reported for the ‘Fiesta’ and ‘Discovery’ maps. However, the integrated F×D map has the major advantage that it contains many more (156) microsatellite markers offering greater potential for alignment and data transfer to other populations.

Population data

From the 257 F1 progeny used to construct the T and B linkage maps, 199 genotypes produced a sufficient number of fruit in 2004 to perform fruit quality measurements, while in 2005, only 165 genotypes could be analysed. An overview of the trait analyses carried out is provided in Table 3.

Table 3 Fruit quality traits, abbreviations and units used for fruit traits measured in the T×B F1 mapping population at harvest
Full size table

For all the measured traits, descriptive statistics were used to determine the population variance—see Table 4. These calculations show that there is a sufficient degree of variability within the population for all the traits analysed and that all traits display a statistically normal distribution (data not shown). Unfortunately, data for the parental fruit quality measurements are not available because in 2003 the maternal trees (‘Telamon’) were lost and had to be replaced by new examples which at the time of these experiments were still not fruiting.

Table 4 Mean, maximum, minimum and standard deviation of the traits measured on the fruit of the T×B genotypes at harvest in 2004 and 2005
Full size table

Correlation coefficients were calculated to determine the extent to which traits are correlated with each other (data not shown) because it has been noted that QTLs for correlated traits often map to similar positions (Paterson et al. 1991). Unsurprisingly, high correlation coefficients (>0.6) were found between the firmness (SMS) measurements on the red (sunny) and green (shaded) sides of the fruit, between the °Brix on the red and green side of the fruit and between the weight, diameter and height of the fruit. The Tukey test at the 5% level was used to determine whether the red and green quality traits were significantly different from each other (Table 4) and whether the trait results were significantly different over the 2 years—see Table 5.

Table 5 Mean values of the quality traits measured on fruit of the T×B genotypes harvested in 2004 and 2005
Full size table

As can be seen in Table 4, there are significant differences between the red and green ºBrix values in both 2004 and 2005, while for firmness a significant difference between the red and the green side existed only in 2004. Because of these differences, data from measurements on the red and green sides of fruit were treated separately in all further analyses rather than being combined to generate a mean value for the whole fruit. Significant differences between the harvests of 2004 and 2005 were observed for the traits fruit height, °Brix and acidity, with on average, fruit from 2005 being longer, having a higher °Brix content and a higher titratable acid content, even though there was no significant difference in the population’s mean fruit fresh weight. On the basis of these statistical analyses, therefore, we might have expected to identify either colour (red or green), or year-specific QTLs for these particular traits.

Comparisons with grafted plants

The T×B mapping population was replicated by grafting onto commercial M9 dwarfing rootstock and planted out in the field in 2002. However, at the time of these experiments, there were too few genotypes with a sufficient number of fruit for a full QTL analyses to be carried out. Nonetheless, a statistical analysis in 2005 of fruit from 50 of these M9-grown genotypes showed that there were no significant differences in quality traits relative to fruit derived from the same 50 genotypes grown on their own root systems. This was perhaps surprising because it is known that light interception, canopy size and crop load are important factors in determining final fruit quality (Wunsche et al. 1996, 2005). Therefore, it might have been expected that fruit from genotypes on dwarfing rootstocks would have higher °Brix contents and a smaller size, due to an improved light penetration and less vegetative growth, respectively (Wunsche and Lakso 2000; Wunsche et al. 2005). The fact that these effects were not detected could be due to either the relatively small number of genotypes analysed or the fact that the columnar growth present in almost half of the T×B genotypes analysed overrides the effects of the rootstock on vegetative growth reduction and thus ultimately fruit quality.

Differences in fruit production between years

Because the columnar growth habit of the ‘Telamon’ parent is also often associated with pronounced biennial bearing due to its spur-type flowering shoots (Lauri and Lespinasse 1993), the smaller number of fruiting genotypes present in 2005 compared to 2004 (165 and 199 trees, respectively, with 144 trees providing sufficient fruit in both years) could have been due to biennial bearing, linked to the presence of the Co gene. In fact, the columnar tree form is present in about half of all the F1 progeny (47%), but of the 55 trees that fruited only in 2004, 39 had a normal growth habit (i.e. were lacking the Co gene) and only 16 had a columnar growth habit (29%), so that this cannot explain the differences in fruiting behaviour in these genotypes. Possibly, this effect is related to juvenility effects and these trees flowered and fruited too intensely in 2004 and were thus unable to produce sufficient fruit in the subsequent year.

QTL analysis of the fruit quality traits

QTL analyses were carried out using fruit from the T×B population grown on its own root system. Because all the traits analysed were found to have a statistically normal distribution (data not shown), QTLs were identified using the IM method in combination with rMQM as recommended by Van Ooijen et al. (2002). The results here once again confirm the quantitative nature of all the traits analysed because generally more than one QTL was detected per trait. A summary of these results is provided in Table 6 and Fig. 2.

Fig. 2
figure 2figure 2

Overview of the all fruit quality traits QTLs identified. Common markers on corresponding LGs are connected with a line. The solid part of the bars of the QTL symbols indicates the most likely position of the QTLs, while the thin lines represent the confidence interval at the 95% level. The QTL symbols for the harvest of 2004 are open and for the harvest of 2005 are filled black

Full size image
Table 6 Overview of fruit quality trait QTLs detected in the segregating progeny of the T×B cross, with results listed per LG
Full size table

Apart from the trait ‘rate of flesh browning’ in 2004, QTLs were detected for all traits analysed and in both parents. Trait QTLs that were detected on overlapping regions of homologous LGs were assumed to be the same QTL, and on this basis the number of QTLs detected per trait varied from zero (for rate of flesh browning in 2004) to up to six (for acidity in 2004). Other workers have reported the detection of up to 12 QTLs per trait (e.g. fruit flesh firmness; Liebhard et al. 2003a).

Interestingly, we also found significant clustering of QTLs with a large cluster containing QTLs for all measured traits except fruit height, stiffness and rate of flesh browning localised to LG 10 in both 2004 and 2005 (Fig. 2). This clustering may be due to either tight linkage between the QTLs (clustering of different genes) and/or may indicate the presence of a single QTL with pleiotropic effects. Similar but smaller clusters of fruit quality QTLs have been reported in other apple varieties (Liebhard et al. 2003a) and other fruit species including peach (Etienne et al. 2002; Quilot et al. 2004) and tomato (Causse et al. 2002; Lecomte et al. 2004). QTL analyses of the same T×B fruit after 2-month post-harvest storage with and/or without a further period of 10-day ‘shelf life’ storage also revealed a cluster of QTLs for most of these quality traits localised to this same region on LG 10 (Kenis et al., manuscript in preparation). Previous work from our laboratory identified QTLs for a number of plant architectural characteristics including ‘height increment’, ‘growth rate’, ‘internode length’, the ‘number’ and ‘length’ of branches and ‘growth increment’ in exactly this same region of LG 10 of ‘Telamon’ as the cluster of fruit quality traits (Kenis and Keulemans 2007). In addition, the Co gene, which is responsible for the columnar phenotype of apple trees, also maps to this region (Kenis and Keulemans 2007). Finally, we have also reported that QTLs associated with fruit flesh vitamin C (L-ascorbate) and total vitamin C (L-ascorbate + dehydroascorbate) contents are also found in this same region of LG 10 of ‘Telamon’ and ‘Braeburn’ in 2005 (Davey et al. 2006). Taken together, these results suggest that this area on LG 10 of the apple genome may control aspects of tree growth and development and that these then have knock-on pleiotropic effects which impact on fruit quality traits. In this respect, our results are similar to the results of Schauer et al. working in tomato, who identified 889 fruit metabolite QTLs in tomato and yet found that at least 50% of these loci were associated with QTLs that modify “whole plant yield” (Schauer et al. 2006). In general though, little is known about the location of genes for plant growth regulators on apple linkage maps. For example, while the gene for 1-aminocyclopropane-1-carboxylic acid oxidase (Md-ACO1), which catalyses one of the two critical steps of ethylene biosyntheses, has been mapped to this region of LG 10 in a ‘Prima’ × ‘Fiesta’ population and in a ‘Fuji’ × ‘Modial Gala’ population (Costa et al. 2005), in ‘Braeburn’ it is located on the bottom of LG 10 (Dreesen, unpublished results) and is thus unconnected with the fruit quality, architectural and vitamin C QTLs on LG 10. MdACS-1 maps to a location on LG 15 in both ‘Telamon’ and ‘Braeburn’ (Dreesen, unpublished results).

QTL clusters were also detected on other LGs, and this was largely as expected on the basis of the measured correlation coefficients. For example, the high correlation coefficients between red and green °Brix were supported by the fact that all the QTLs for red °Brix in 2004 and 2005 co-localised with QTLs for green °Brix on LG 2. Because the percentage of phenotypic variance explained by these QTLs is also similar, our results suggest that the statistically significant differences found by analysis of variance between the red and green °Brix values is primarily due to differing environmental conditions or to interactions with other genetic factors, rather than differences in genetic control alone. We did however identify one unique QTL for green firmness in both 2004 as in 2005 on LG 14 of ‘Telamon’, but no co-localising QTLs for red firmness was found.

Fruit acidity in apple is primarily determined by the concentration of only one major organic acid, which is malic acid. Maliepaard et al. (1998) showed that the malic acid content is determined by a single gene and that the Ma gene maps to LG 16 in the cultivars ‘Prima’ and ‘Fiesta’. While we have not mapped Ma in our parents, major year-stable QTLs for acidity were detected on LG 16, which individually accounted for 20–34% of the total observed variance in both years. This is similar to the results of Liebhard et al. who also localised a major acidity QTL on LG 16 of ‘Fiesta’. This QTL was also linked to high LOD scores for crispness, juciness and consumer ‘overall liking’, accounts for 36% of the population acidity variance and is tightly linked to the SSR marker CH05e04z (Liebhard et al. 2003a). In ‘Telamon’ and ‘Braeburn’, we observed similar results, with the major acidity QTLs in both parents being localised between the adjacent SSR markers CH05e04z and CH05c06 on LG 16 and we therefore strongly expect the Ma gene to map here. Although a second major acidity QTL (33%) on LG 8 was also reported by Liebhard et al., in ‘Braeburn’ and ‘Telamon’ we observed only one relatively minor QTL (~8%), and this was only in 2004.

Colour (fruit side)-specific QTLs

Interestingly, a minor ‘green-specific’ QTL for firmness was detected on LG 14 of ‘Telamon’ in both harvest years. In 2004, this QTL was one of only two QTLs detected for green firmness (SMSG), and the percentage of phenotypic variance explained by this QTL (7.2%) was similar to that of the other green firmness QTL on LG 10 (7.8%). However, this second green firmness QTL on LG 2 co-localised with a QTL for red firmness (SMSR) in 2004, so the fact that we observe statistically significant differences between red and green firmness values in 2004 is probably primarily due to the presence of the first QTL. In 2005, however, the green firmness QTL on LG 14 is only one of four green firmness QTLs and the percent of phenotypic variance explained by this QTL (7.2%) is much less than the sum of the percentages explained by the other three QTLs (for ‘Telamon’ a total of 53.4%). This dominating effect of the other three QTLs probably explains why there was no significant difference between red and green firmness detected in 2005, even though ‘green-specific’ firmness QTLs were found. Overall, therefore, the ‘colour-specific’ genetic determinants of ºBrix and firmness are either very small or non-existent in our T×B population, suggesting that a distinction in the measurement position of the fruit is probably not necessary in future experiments and that an average value of the two measurements is sufficient. In other words, red- and green-specific differences are primarily an environmental or a gene × environment effect.

QTL stability

QTL stability has emerged as a major limiting factor in the transference and application of QTL results in breeding programmes. In this work, we considered QTLs to be stable if they were detected in successive years in homologous regions of the same LG. Often, these year-stable QTLs also had the same marker as the most significant (highest LOD score) marker. Such markers are currently the best candidates for further development of these results into breeding tools. For traits such as ‘harvest 2004’ and ‘harvest 2005’ on LG 10 of ‘Telamon’, QTLs were detected on the same LGs in both years, but as the QTL regions did not overlap they were treated as individual loci. Stability results are summarised in Table 6, with year-stable QTLs being shown in bold.

Overall, we can see that a total of 30 year-specific QTLs were identified, with 14 for traits in 2004 and 16 for traits in 2005. An additional 16 year-stable QTLs were detected, with the majority of these being localised within the two large clusters of fruit quality traits found on LGs 10 and 16. In fact, year-stable QTLs were detected for all traits analysed except ‘fruit height’ and ‘rate of flesh browning’, with a maximum of four year-stable QTLs for the trait fruit flesh acidity. Interestingly, we found very strong, highly significant QTLs for ‘rate for flesh browning’ in 2005, which explained 66% and 77% of the population variance for this trait. However, not a single ‘browning’ QTL was detected in the previous year. This contrasts with other major QTLs, where in general if a major QTL (>20%) is present, it is usually a year-stable QTL, even if it is not always a major QTL. The exceptions to this rule are the QTLs for fruit fresh weight and fruit diameter on LG 9 and for fruit height on LG 15 of the ‘Telamon’ parent. This relative instability of QTLs from year to year appears to be fairly typical (e.g. King et al. 2000; Liebhard et al. 2003a) and possibly provides an indication of the importance of environmental conditions on quality traits. Indeed, similar results were reported by Rousseaux et al. (2005), who found that most tomato fruit anti-oxidant QTLs in near isogenic lines were only detected in one out of three seasons and even disappeared when the same lines were tested under greenhouse conditions.

It is known of course that apple fruit quality is closely linked to pre-harvest environmental conditions, with perhaps the most obvious link being the requirement for ultraviolet light and low temperatures to induce fruit anthocyanin synthesis and red skin colouration (Honda et al. 2002; Merzlyak et al. 2002). However, pre-harvest exposure of fruit (and vegetables) to direct sunlight and the associated high tissue temperatures can also result in differences in sugar contents, tissue firmness and mineral content (Ferguson et al. 1999; Woolf and Ferguson 2000) and can lead to changes in the contents of anti-oxidants (Davey et al. 2000; Treutter 2001; Cisneros-Zevallos 2003; Hodges et al. 2004). Therefore, the appearance of highly significant, major QTLs for the rate of flesh browning in 2005 is likely to be related to the seasonal growth conditions in that year. Interestingly, the ‘rate of flesh browning’ QTL on LG 17 also coincided with a major QTL for fruit flesh total vitamin C and a minor QTL for flesh dehydroascorbate contents (Davey et al. 2006) and it has been shown that fruit vitamin C contents are involved in determining the incidence of internal browning disorder, in pear (Franck et al. 2003a, b) and in apple. Indeed, in ‘Braeburn’, internal browning disorder has been linked to lowered activities of the anti-oxidant enzymes superoxide dismutase and catalase, such that ‘Braeburn’ fruit from an orchard located in a colder area had lower catalase activity, lower lipid soluble anti-oxidant levels and a higher risk of browning disorder (Gong et al. 2001). This again suggests that the appearance of the strong ‘rate of browning’ QTLs in 2005 is related to the different environmental conditions in 2005. An overview of meteorological measurements for 2004 and 2005 in our orchards is provided in Fig. 3.

Fig. 3
figure 3

Summary of the climactic conditions for 2004 and 2005, present at the field site where the T×B mapping population is maintained

Full size image

In fact, the climatic conditions in 2004 and 2005 in Rillaar (Belgium), where the T×B population is situated were quite different. Overall, 2004 can be considered to be a relatively warm season with high temperatures and high rainfall during the summer followed by a mild autumn season, although for a period in the beginning of August 2004 the mean daytime temperatures were up to 10°C lower than in 2005. By comparison, the summer of 2005 had ‘normal’ temperatures, while the beginning of the autumn was characterised with unusually high seasonal temperatures and markedly less rainfall. In terms of physiology, these climatic differences appeared to have little influence on the timing of the harvest but did result in a population where the fruit was on average slightly higher (longer) and had higher acidity and Brix contents in 2005 (Table 5). Unfortunately, we do not have information on changes in fruit anti-oxidant contents in these years, but presumably the more extreme climatic conditions (particularly lower rainfalls) lead to more variability in the rate of browning, which facilitated the detection of browning QTLs in 2005.

QTL significance values and population variances

The potential usefulness of QTL results depends not only on their significance (LOD score) but also on the percentage of the population phenotypic variance explained or accounted for by that particular QTL. The ‘percentage of phenotype variance’ values for each QTL are provided in six and range from 5.2% (for ‘acidity’ on LG 15 of ‘Telamon’) to 34.2% (for ‘acidity’ on LG 16 of ‘Braeburn’) in 2004 and in 2005 from 6.2% (for ‘weight’ on LG 10 of ‘Braeburn’) to 76.9% (for ‘rate of flesh browning’ on LG 3 of ‘Braeburn’). In general, a distinction is made between ‘major’ QTLs which account for more than 20% of the explained population variance and ‘minor’ QTLs which explain less than 20% of the observed population variance for that trait. In 2004, only the QTL for ‘acidity’ located on LG 16 of both ‘Telamon’ (30.4%) and ‘Braeburn’ (34.2%) could be considered as major, whereas 12 of the 32 QTLs detected in 2005 were major QTLs. Again, we presume that this difference between the years is related to the more severe environmental conditions in 2005, facilitating QTL detection.

Comparison of QTL results in ‘Telamon’ × ‘Braeburn’ with QTL results in other mapping populations

When evaluating the use of QTLs in breeding, it is important to consider QTL stability across genetic backgrounds, in addition to stability across years and locations–environments. Ideally, traits of interest should be measured simultaneously in different mapping populations and at the same location. In reality, however, the practical difficulties associated with this means that comparisons are generally limited to populations grown on different sites, harvested in different years and measured by different scientists using different methodologies.

To date, only a limited number of apple fruit trait QTL analyses are available. King et al. (2000, 2001) published results from the analysis of the progeny of a cross between ‘Prima’ and ‘Fiesta’ (P×F), but of the traits measured, only fruit fresh weight, stiffness and flesh firmness were common to our work. However, the results for the analyses of other quality traits such as harvest date, fruit weight and firmness, ºBrix and acidity of the fruit flesh have been reported by Liebhard et al. (2003c) working in a population derived from a cross between ‘Fiesta’ and ‘Discovery’ (F×D). A direct comparison with these published results is still difficult though for a number of reasons. Firstly, some of the trait-measurement protocols differed; these groups also did not distinguish between the red and the green side of the fruit for measurements of firmness and ºBrix, and finally accurate QTL positions were not always provided in the published texts, meaning that it is difficult to be sure whether these QTL regions actually co-localise and thus represent the same locus. This last problem is compounded by the lack of markers common to all the various genetic linkage maps. Nonetheless, in Table 7, we present an overview of the LGs where QTLs for traits common to the different mapping populations are located.

Table 7 Location (LG number) of QTLs for fruit quality traits detected on the genetic linkage maps of the ‘Telamon’ × ‘Braeburn’ population (T×B), the Fiesta × Discovery population (F×D; Liebhard et al. 2003a) and the Prima × Fiesta population (P×F; King et al. 2000, 2001)
Full size table

For °Brix and stiffness, there were no QTLs on homologous LGs common between the two mapping populations in which these traits were measured (T×B and F×D). For fruit fresh weight, one QTL was detected on LG 6 in all three populations, although it was absent in the T×B population in 2004 and was only a minor QTL in 2005. However, the T×B and F×D populations did share a minor year-stable fresh weight QTL on LG 10, and the F×D and P×D populations shared a fresh weight QTL on LG 16, which was not detected in the T×B population.

The major acidity QTL which coincides with the location of the Ma gene on LG 16 was detected in all three populations. In addition, the major QTL for acidity found on LG 8 of ‘Fiesta’ was also detected on LG 8 of ‘Telamon’ and ‘Braeburn’ but only for the harvest of 2004, suggesting that environmental conditions masked this effect in 2005, as well being responsible for an overall higher mean titratable acid content (Table 5). The fact that these acidity QTLs were detected in both the T×B and the F×D populations confirms the heterozygous status of the Ma locus in these varieties. Three homologous firmness (SMS) QTLs were identified (on LGs 3, 10 and 14) in two different mapping populations, while finally for harvest date, one QTL on LG 3 was found to be common to both the T×B and F×D mapping populations.

Of course, care should be taken when interpreting the significance of the absence of a QTL in a variety because the fact that a QTL is not detected may also be the result of the statistical methods used. In other words, a QTL can only be detected if there is segregation for that trait within the population and if measurable effects can be attributed to the different parental alleles. For example, King et al. reported that the QTL detected on LG 10 in the ‘Prima’ × ‘Fiesta’ population for firmness is the result of a single allele derived from the Fiesta parent which must therefore be heterozygous for the locus underlying this trait. The fact that this QTL was not detected in the F×D population can thus only be ascribed to environmental effects or to the different genetic background of the populations.

To summarise, therefore, while there were 45 QTLs detected in the T×B population, for the six fruit quality traits that were measured in all populations, only nine QTLs are found to be common to at least two different populations (Table 7). This demonstrates the importance of genetic backgrounds and/or environment in modulating the strength of QTL effects. Of course, ideally, validated, internationally standardised trait analysis protocols should also be applied to help minimise the sources of variation between studies. Even then, however, similar effects were reported by Chaïb et al. (2006) studying the inheritance of organoleptic quality in tomato who showed that QTLs could be lost or new QTLs detected not only in different mapping populations but also in successive backcross generations.

Conclusions

In this manuscript, we present results from a comprehensive study over 2 years of various fruit physiological traits. Comparing our results with previously published studies again emphasises the importance of well-executed experimental designs. In particular, it is clear that it is necessary to confirm the stability of the QTL results in other mapping populations, at different sites and over several years before the co-segregating molecular markers could be considered for progeny screening in breeding programmes. Nonetheless, our results have identified a number of year- and population×location-stable QTLs which hold some promise for development. A more complete interpretation and analysis of these results however will depend on a better integration and alignment of the ‘Telamon’ and ‘Braeburn’ linkage maps with other existing maps and will involve further linkage map saturation using microsatellite markers, such as those recently developed for Malus (Silfverberg-Dilworth et al. 2006).

As a final remark, it is worth pointing out that to date, QTL investigations of fruit quality traits have focused on traits measured at harvest. However, the majority of fruit reaches the consumer only after a period of up to several months in cold storage and post-harvest storage. For these reasons, we have also carried out QTL analyses for the same quality traits after storage and/or 10-day shelf life using fruit of the T×B population. These analyses revealed significant differences both in the position of QTLs after storage but also QTL co-localisation (Kenis et al., manuscript in preparation).