Abstract
A full-sibling F1 population of 144 individuals from a cross of Ruby Seedless x Thompson Seedless (Vitis vinifera L.) was used to generate a consensus genetic map based on 154 markers (105 SSRs, 37 AFLPs, five ISSRs, one RAPD, four SCARs, and two phenotypic markers) distributed across 19 linkage groups covering 1,340 cM. The following traits were evaluated for later quantitative trait loci (QTL) analysis for seedlessness: number of seeds, total fresh and dry weight of seed and seed traces, berry size and weight, and ripening dates. A QTL with a major effect (K up to 45%) for seedlessness subtraits was found on linkage group 18. QTLs for berry size, berry weight, and ripening date were found in the same location, suggesting a pleiotropic effect. Seedless berries were smaller and had an earlier ripening date. Four other minor QTLs for seedless were found in linkage groups 4, 8, 15, and 16. These results are consistent with previous studies and reflect the quantitative and qualitative nature of this complex trait.
The grapevine (Vitis vinifera L.; n = 19) is an interesting model for the study of seed development in fruit crops because two different mechanisms are involved in seedlessness: parthenocarpy and stenospermocarpy. In parthenocarpy, true seedlessness occurs and berries develop from the ovary without any seed formation. This mechanism is rare and confined to a few genotypes such as Black Corinth. In stenospermocarpy, pollination and fertilization occur normally, but seeds abort at an early stage of development (Stout 1936). In this more prevalent mechanism, the pericarp (berry flesh) keeps growing but the embryo and/or endosperm arrests its development, resulting in the presence of seed traces and a reduced berry size at harvest (Doligez et al. 2002, Fanizza et al. 2005). Stenospermocarpic seedlessness can be separated into several quantitative components or subtraits: fresh and dry weight of seeds and seed traces, total number of seeds and seed traces, degree of seed coat hardness, and degree of endosperm development (Striem et al. 1996).
Stenospermocarpy in table grapes can be studied with quantitative genetic approaches. To do this, a large sibling population is needed along with numerous molecular markers for the construction of detailed genetic maps. In addition, for self-pollinated species such as grape, reliable markers are needed to overcome complications derived from the estimation of recombination frequencies and the linkage phases of the markers in both parents. Among the most reliable markers are simple sequence repeats (SSRs), which have been used to map the grapevine genome (Adam-Blondon et al. 2004, Riaz et al. 2004) and to detect quantitative trait loci (QTL) for phenotypic traits (Doligez et al. 2002, Fanizza et al. 2005). A complete understanding of the inheritance of seedlessness would greatly aid breeding programs through the use of marker assisted selection (MAS).
Quantitative trait loci have been identified for seedlessness subtraits and yield components in table grape (Doligez et al. 2002, Fanizza et al. 2005), and the well-known negative correlation between berry size and seedlessness has been confirmed. In addition, the negative correlation between berry size and cluster weight has been described (Fanizza et al. 2005), but this negative correlation was not supported by molecular data. In general, QTLs detected in both papers were the same, except for a QTL for fruit yield, which Fanizza and colleagues found unstable over years and with little phenotypic variance. Using a progeny derived from two partially seedless genotypes, MTP2223–27 (Dattier de Beyrouth x Pirovano 75) and MTP2121–30 (Alphonse Lavallée (Ribier) x Sultanina), a major stable QTL for seedlessness and berry weight was found in linkage group 18 (Doligez et al. 2002). Additionally, three other regions were identified with minor effects on berry weight or seedlessness. A factor that could be influencing the heritability of this trait is the background of parental genotypes. Fanizza and colleagues used a progeny derived from a cross between two contrasting varieties, Italia (seeded) and Big Perlon (seedless), which gives a segregation ratio of 1:1 (seedless:seeded), instead of 1:2:1 (seedless:noticeable seed traces:seeded) obtained when crossing two partially seedless genotypes (Doligez et al. 2002).
Our work was initiated to further understand the genetics of seedlessness and to determine or confirm the chromosomal location of possible QTLs for this trait, berry size and ripening date, using a population derived from the seedless cultivars Ruby Seedless x Thompson Seedless (Sultanina). Ruby Seedless is a University of California, Davis variety derived from a cross of Emperor x Pirovano 75, which in turn is derived from the cross of Muscat of Alexandria x Sultanina. In addition, possible associations between markers and phenotypes were also sought to enable MAS in the Chilean table grape-breeding program.
Materials and Methods
Plant material.
Progeny (n = 144) were obtained via embryo rescue (Cain et al. 1983) from a cross of Ruby Seedless x Thompson Seedless (RxTS) made in spring 1995. Plants were grown on their own roots in a seedling block planted by 1998 at La Platina Experiment Station of the Instituto de Investigaciones Agropecuarias, in Santiago, Chile. Phenotypic data were recorded in 2002, 2003, and 2004. Fifty-five individuals flowered the first season, 71 the second, and 98 the last season. QTL analyses were performed for seasons 2003–2004 and 2004–2005.
Phenotypic evaluations.
The following traits were scored: berry weight (g) (BW); berry equatorial diameter (cm) (BED); total fresh and dry weight (TFW, TDW) of seeds or seed traces; number of fully developed seeds (SED); total number of seeds and/or seed traces (SST); and ripening date (17 Brix) (RDA). BW, BED, TFW, SST, and SED were measured using 25 randomly sampled berries for each genotype. Percentage of seed dry matter (%DM) was calculated to evaluate seed coat hardness.
Molecular markers.
For the genotypic analyses, young immature leaves (not fully expanded) were collected in spring (September–October) 2 weeks after bud-break and kept at −80°C until DNA extraction. DNA was extracted from 100 mg of frozen leaves using a scaled-down (1/5) version of an existing protocol (Lodhi et al. 1994) without any other modification.
A total of 109 SSR primer pairs were selected based on an existing method (Costantini et al. 2007) and according to previous reference maps of Syrah x Grenache (Adam-Blondon et al. 2004) and Riesling x Cabernet Sauvignon (Riaz et al. 2004) to provide at least four to six polymorphic SSR per linkage group. Microsatellites were amplified in a 16-μL reaction mixture containing 0.25 μM of each primer, 0.25 mM of each dNTP, 1.5 mM MgCl2, 0.25 U Taq polymerase, and 10 ng template DNA. A three-step protocol consisting of denaturation (30 sec at 95°C), annealing (30 sec at 56°C), and extension (30 sec at 72°C) for 25 cycles, followed by a fill-in step of 4 min at 72°C was used for PCR amplification. If no amplification was observed, then annealing temperature was lowered to 50°C, the number of cycles was increased to 35, or MgCl2 was adjusted to a final concentration of 2.0 mM.
Amplified fragment length polymorphism (AFLP) analyses were performed according to (Vos et al. 1995). Preselective PCR AFLP were amplified in a 25 μL reaction mixture containing 5 μL diluted (1/10) digested, and ligated template DNA, 0.24 μM of each primer, 0.15 mM of each dNTP, 2.0 mM MgCl2, and 0.5 U Taq polymerase. A three-step protocol was used consisting of denaturation (30 sec at 94°C), annealing (60 sec at 56°C), and extension (60 sec at 72°C) for 20 cycles, followed by a fill-in step of 4 min at 72°C. The PCR product was diluted 50- to 70-fold and selective PCR AFLPs were amplified using primers with three additional selective nucleotides. Conditions were the same as the preselective reaction, except that 0.05 μM of EcoRI primer, 0.25 μM of MseI primer, and 5 μL diluted preamplification product were used as template. A touch-down PCR amplification was programmed, consisting of 38 cycles of denaturation (30 sec at 94°C), annealing (30 sec), and extension (60 sec at 72°C). The annealing temperature was 65°C for the first cycle and was reduced by 0.7°C for the next 13 cycles, and 56°C for the last 23 cycles.
Intersimple sequence repeat (ISSR) markers (British Columbia University, Canada: BC800 serial ISSR primers) were performed according to Moreno et al. 1998. SSR, AFLP, and ISSR amplified products were mixed with a sequencing dye in a 1:1 ratio, denatured at 95°C for 3 min, and loaded onto a 6% polyacrylamide gel. After electrophoresis, gels were silver stained and data was recorded twice by naked eye.
For random amplified polymorphic DNA (RAPD) analyses, 12-mer primers (Wako Pure Chemical Industries, Osaka, Japan) were used in standard 50 μL PCR mixture according to Williams et al. 1990. The amplification products were separated by electrophoresis on a 1.5% agarose gels in 0.5x TBE buffer at 8 V/cm. Banding patterns were visualized under UV light after staining with ethidium bromide and documented with a digital camera.
Putative markers (RAPD or AFLP) related to seedlessness were transformed into SCAR (sequence characterized amplified regions) markers as described by Lahogue et al. (1998) and evaluated on the mapping population as has been described (Mejía and Hinrichsen 2003).
Phenotypic marker.
Seed development inhibitor (SdI) locus was mapped analyzing seedlessness as a qualitative trait, RxTS progeny were categorized in terms of TFW and SST as described elsewhere (Bouquet and Danglot 1996, Lahogue et al. 1998). A minor correction was adopted because of the difference in seedless degree in the progeny compared with previously described progenies. In our case, SdI+ were those seedless individuals with complete absence of developed seeds and with a total fresh weight of seeds or seed traces per 25 berries <0.1 g; SdI- were seeded individuals with one or more well-developed seeds and >1.0 g of total fresh weight of seeds or/and seed traces per 25 berries; SdI± were partially seeded individuals with hard seed traces or noticeable seeds or/and seed traces per 25 berries. ANOVA was performed using SPSS 11.0 (SPSS Inc., Chicago, IL) to compare the different classes. A Duncan test was performed to determine homogeneity of classes. Color was evaluated as a binary trait, colorless (green) or red (ruby) colored.
Mapping.
The double pseudo-testcross strategy (Grattapaglia and Sederoff 1994) and JoinMap 3.0 software (Plant Research International, Wageningen, Netherlands) were used to build genetic maps. Markers with a high distortion or unexpected χ2 test results were discarded, especially if they were dominant markers such as RAPD or AFLP. Linkage groups were determined using the Kosambi function for the translation of recombinatorial units to genetic distances. The software determined the linkage phases automatically. A LOD score threshold for the determination of linkage groups was set at 3.5. The recombination fraction permitted was 0.45. Markers within the resulting groups were ordered relative to each other by automatic multipoint analyses using the default values of JoinMap 3.0 (mapping threshold LOD > 1, recombination frequency threshold <0.4).
Parental maps were constructed as two double-haploid populations instead of two cross-pollinated populations, because the double-haploid strategy removes the dominant heterozygous markers that add distortion to JoinMap 3.0, so the software could identify the proper linkage phase. A consensus map was constructed using the parameters for a cross-pollinated derived population and the integrate map function of JoinMap 3.0. Codominant markers and double heterozygous dominant markers are used to align the genetic parental maps. In this analysis, all configuration patterns were used. Two recombination data sets, derived from the parental meiosis, were analyzed together.
The linkage groups were numbered according to the reference map of Riaz et al. (2004) and the international agreement achieved within the IGGP (International Grape Genome Program; www.vitaceae.org).
QTL analysis.
QTL analyses were performed separately for both parental and consensus maps for each season; reproducibility among seasons was analyzed with ANOVA. For parental maps, framework maps constructed with the double-haploid population strategy were used for the detection and analysis of QTLs using MapQTL 4.0 (Plant Research International). For the consensus map, a cross-pollinated population strategy was used in the same way. Interval mapping (IM) (Lander and Botstein 1989) was performed based on the stringent maps for each parent and consensus in combination with the field data for the analyzed traits. To declare a QTL with confidence, a permutation test (included in MapQTL 4.0) was used to establish the threshold level at which a QTL was declared significant or suggestive (Doerge and Churchill 1996). The threshold level was determined after 1,000 permutations of the data with a genomewise and chromosome-wise type error rate of 0.05. A QTL was considered statistically significant or suggestive if the LOD detected by IM was greater than the LOD threshold determined after 1,000 permutations, for the genomewise or chromosome-wise type error, respectively.
Interval mapping assumes normal distribution of phenotypic data, but in this study seedless subtraits had a bimodal distribution. Because of this distribution, data was tested for normality with both the Kolmogorov–Smirnov test included in SPSS 11.0 software and the Shapiro–Wilk test for data with less than 50 categories. Data was then normalized according to Yn = √X + 0.5 where X is the raw data and Yn is the normalized data. A nonparametric Kruskal–Wallis (KW) rank-sum test was applied to verify the global segregation of each locus using the KW option of MapQTL 4.0 to detect putative QTLs. Detected QTLs were considered significant when p > 0.005.
Results and Discussion
Map construction.
Five types of molecular markers (SSR, AFLP, RAPD, SCAR, and ISSR) were genotyped to construct a linkage map for the cross RxTS. The total number of loci in the final mapping data set was 247 (Table 1⇓). The number of fully informative and codominant markers segregating in both parents that could be used as bridges for the construction of the consensus map was 62, at least three per each of the 19 linkage groups. Mapping data set contained two phenotypic markers, seedlessness (SdI locus) and color, both analyzed and mapped as qualitative traits. SdI and color mapped close to SSR VMC7F2 in LG 18 (Figure 1⇓) and to SSR VMC5G7 in LG 2 (not shown) respectively, as previously reported (Doligez et al. 2002).
With the double pseudo-testcross strategy, markers from each parent were analyzed separately. At LOD 3.5 for grouping, 90 markers from Ruby Seedless and 108 from Thompson Seedless were linked and mapped on the basis of segregation data obtained from 144 individuals. Because of the limited number of markers used in this mapping project, stringency could not be incremented to avoid the appearance of several unlinked markers or more linkage groups than the number of haploid chromosomes. However, the use of a low stringency level was compensated by the use of reference maps (Doligez et al. 2002, Riaz et al. 2004), a SSR set chosen in agreement with members of the MASTER project (Costantini et al. 2007) and the use of codominant markers.
For the map construction, some parental segregating AFLPs were removed (11 in Ruby Seedless and 6 in Thompson Seedless) because of strong segregation distortion. AFLP markers segregating in both parents were not considered because of the strategy chosen for map construction (double-haploid population). Finally, from 116 genotyped AFLP markers only 37 were mapped. In contrast to the relatively low utility of AFLP markers for the construction of the framework maps, 105 of 116 genotyped SSRs markers were mapped. Eighty-one SSRs were mapped in Thompson Seedless and 67 in Ruby Seedless. Nonmapped SSRs were removed because they distorted the final order of more informative markers or simply because they could not be linked under the established parameters.
The effective map coverage was lower for parental maps (<60%) than for the consensus map (75%) (Table 2⇓). Large gaps between markers were present in parental and consensus maps, of more than 35 cM and 45 cM, respectively. Despite the modest genome coverage, the use of reliable markers (SSRs) and reference maps allowed for the generation of a framework map for QTL identification. The linkage groups where QTLs were found for seedless subtraits and overall berry size are shown in Figure 1⇑.
Phenotypic data evaluation.
Not all 144 siblings of the Ruby Seedless x Thompson Seedless cross have reached maturity. The number of individuals producing fruit increased yearly during this study (from 55 individuals the first season to 98 in the final season). However, there is a group of individuals (7–8%) that have never fruited even after 8 or 9 years in the field. Some have never developed flowers; in others, flowers developed, but no fruit set, and in some cases, whole clusters aborted after fruit set. These observations suggest several causes of infertility and, therefore, genetic dissection analysis for seedlessness requires a larger mapping population and evaluation over many years. It is also likely that inbreeding depression plays a role with the use of Ruby Seedless, which has a parentage that is 50% Thompson Seedless.
Trait evaluations for the last two seasons produced very similar results, with similar ANOVA values (data not shown). For most of the traits, transgressive distribution was observed (data not shown). For subtraits related with seedlessness (TFW, TDW, SED, SST), the Shapiro–Wilk test showed a nonnormal distribution and traits remained nonnormal even after data transformation. Instead of a normal distribution, most of these traits showed a bimodal distribution or a mixture of normal distributions with dominance. A bimodal distribution of seed size based on fresh weight was clearly noticeable, and the intercept of the two types of populations was between 625 to 875 mg for 25 berries (Figure 2⇓). Similar results were obtained with a population derived from a seeded x seedless cross (Ledbetter and Burgos 1994).
Analyses of total fresh weight and number of normally developed seeds revealed that individuals characterized as seeded in the RxTS progeny belong to the hard seed traces classs in the progeny Mtp 3140 described elsewhere (Bouquet and Danglot 1996, Lahogue et al. 1998). This difference might be due to the direct inheritance of this trait from Thompson Seedless in the RxTS progeny, instead in Mtp3140 where Thompson Seedless is a grandparent. If seedlessness inheritance is due to polygenic factors, then the use of Thompson Seedless as one of the parents might generate more progeny with a larger degree of seedlessness. The ANOVA between groups (SdI+, SdI±, and SdI-) for total fresh weight of seeds or seed traces revealed an F = 408 for fresh weight (p < 10−12). Once categorized, SdI mapped as a 1:2:1 marker in LG 18 in the same position as the peak of the main QTL for seedlessness (LG 18).
Seedlessness QTL analysis.
For the main seedlessness components (TFW and SED), significant and copositioning QTLs were found in LGs 16 and 18 (Figure 1⇑, Figure 3⇓, Table 3⇓). In LG 18, QTLs for SST, TDW, and %DM were also identified in the same confidence interval as TFW and SED. For SED, SST, and %DM other QTLs were found in LG 4, 8, and 15, respectively. These results suggest that seed development (number, size, and weight) might be under the control of several genomic regions containing genes also responsible for the number and the weight of seeds. However, three additional loci might have an additional influence on seeds or seed trace number (QTLs from LGs 4, 8, and 15).
In this investigation, QTLs found in LG 18 were the most reproducible among the two evaluated seasons (Figure 1⇑, Table 3⇑). The lack of reproducibility for QTLs found in LGs other than 18 could be explained by environmental conditions that affected the phenotypic variance or by a difference in the size of the evaluated population, because a smaller population causes an underestimation of the number of QTLs detected and an overestimation of QTL effects (Melchinger et al. 2004, Vales et al. 2005). For seedless QTLs, the number of detected QTLs increased with the number of fruiting individuals. However, QTLs of large effect (LG 18) were detected even with a small fruiting population, revealing the high stability of this trait independently of the evaluation season.
Phenotypic variation (H) explained by these QTLs (Figure 1⇑, Table 3⇑) is extremely high, with values ranging from 17 to 95% for a single QTL. These H values could be overestimated because of small population size and nonsaturated confidence intervals.
However, QTL effects might also be biased because the interval mapping (IM) approach does not take into account features such as nonnormal distribution of phenotypic data. Skewness and Kurtosis values revealed that distributions were biased toward seedless for most of the seedless subtraits (SED, TFW, and TDW) (Figure 2⇑, Table 3⇑). This lack of normality may be explained by the presence of the major gene SdI (Bouquet and Danglot 1996, Doligez et al. 2002), which creates a mixed nonnormal distribution. Another explanation may be gene x environment interaction (GxE) effects, or it may be the data are intrinsically nonnormal. GxE effects are difficult to avoid in this perennial woody fruit crop; however, the trait under study has high heritability, and biases due to environmental fluctuations would be minimized after analyzing phenotypic data for several seasons in different environments while using larger populations.
The major QTL LOD peak (LG 18) is co-positioned with the SdI locus (Figure 1⇑). Mapping of SdI locus suggests that seedlessness could be a monogenic trait, despite previous segregation studies and analyses of progenies (Bouquet and Danglot 1996) suggesting that seedlessness is a complex trait determined by several genes. The QTL analysis reported here revealed that at least four genomic regions are involved in seedlessness (Figure 1⇑, Table 3⇑), and one of them (LG 18) harbors the main seedless QTL. The co-positioning of a major QTL with the SdI locus suggests that seedlessness is a complex quantitative trait controlled by a major locus (SdI) that causes bias in phenotype distributions and bias in the determination of the QTL effects.
Nonparametric QTL analysis (Kruskal-Wallis; KW) confirmed the major seedless QTL in LG 18 and confirmed QTLs detected as suggested by IM in LGs 8 and 15 (Table 4⇓). Additionally, it revealed a minor QTL in LG 9, which was not detected by IM. KW analysis revealed more reliable proportions of explained phenotypic variance than those detected by IM. For main seedless subtraits (SED, TFW, TDW, %DM, and SST), the closest marker to the QTL peak in LG 18 (VMC7F2) had K values that varied from 26 to 45%, which also reflect the presence of a major QTL in this region.
Berry weight, berry size, and ripening date QTL.
Since berry weight and berry size (diameter) are highly correlated, overall berry size could be analyzed with either trait. To analyze any possible correlation between overall berry size and presence of seed, both berry weight and berry size (diameter) were tested in the QTL analysis. Seeds produce and act as sinks for hormones such as cytokinin and auxin, which induce rapid growth of the developing ovary by increasing cell division and cell expansion (Bohner and Bangerth 1988). If this statement is valid for grape berries, then there might be a correlation between seed number or size and berry weight or size.
For overall berry size (BW and BED), IM revealed significant and co-positioning QTLs in LG 18, in the same confidence interval with seedless QTLs (Figure 1⇑, Table 3⇑). Two other suggestive and co-positioning QTLs were found in LGs 8 and 15. The Pearson test showed a significant positive correlation (r = 0.70) between seed presence and berry size or weight. For berry weight alone, a suggestive QTL was found in LG 1. The high correlation and co-positioning of QTLs for berry size and weight with the major seedless QTL confirms that seedless and berry size are tightly linked, and the presence of several QTLs confirms that overall berry size is also a quantitative trait that may be strongly affected by the major seedlessness locus. However, phenotypic variance explained by these QTLs is also overestimated for the same reasons mentioned above (H values up to 67% for QTLs found in LG 18). Nonparametric KW analysis confirmed overall berry size QTLs detected in LGs 15 and 18 (Table 4⇑). The proportion of explained phenotypic variance was >40% for the VMC7F2 marker located closest to the peak on LG 18 QTL and 13% for the VVIB63 marker found in LG 15 QTL.
Ripening date (RDA) is a trait that is very dependent on environmental conditions, but each variety has a specific range of ripening dates. Thus, despite environmental conditions, there might be a specific genetic control for this trait. Interval mapping indicated two RDA QTLs, which were confirmed by KW analysis. Also found were a significant RDA QTL in LG 18, co-positioning with previous QTLs detected for seedlessness and overall berry size, and a suggestive RDA QTL in LG 17. In KW analysis, the closest markers to RDA QTL peaks (VMC2H3 in LG 17 and VMC7F2 in LG 18) revealed a proportion of explained phenotypic variance of 11% and 21%, respectively. Co-positioning of RDA and seedlessness QTLs suggests a pleiotropic effect of seed weight or number over ripening date, and concomitantly seedless berries are smaller and have earlier ripening dates (data not shown).
Thus far, QTL analyses for seedlessness have revealed the presence of at least five QTLs (LGs 4, 8, 15, 16, and 18). The presence of a major QTL in LG 18 confirmed previous studies (Doligez et al. 2002) and did not exclude an earlier thesis (Bouquet and Danglot 1996), which proposed that the degree of seed development is under control of three complementary recessive genes that are independently inherited and regulated by a major dominant inhibitor gene, SdI. Seedlessness is a complex quantitative trait and the degree of seed development is under control of several genomic regions, with one of them harboring a major gene (SdI) that not only affects seed development but also may affect overall berry size and ripening date. Effects of this major QTL are great enough to consider seed development a multigenic character rather than a monogenic trait.
It has been postulated that gibberellins synthesized by developing seeds are transported to the pericarp where they regulate berry growth controlling cell division and cell expansion (van Huizen et al. 1996). A reduced number of seeds or reduced seed size will produce proportionally less gibberellin, which may have a direct impact on berry growth. Gibberellins might also be involved in the regulation of ripening by acting as an ethylene antagonist, preventing its role in cell expansion and berry softening during grape ripening (Tesniere et al. 2004).
More SSR markers are being added to LGs where QTLs were detected, and as additional members of the mapping population flower or become available they will also be incorporated in future analyses. These additions will increase the mapping precision and reduce QTL confidence intervals, making them more reliable and useful for the identification of markers for assisted selection.
Conclusions
Using a set of microsatellites chosen to facilitate the rapid development of reliable maps, a midresolution genetic map was developed that covered most of the grape genome using a cross between two seedless genotypes. The existence of a major QTL for stenospermocarpic seedlessness was confirmed on LG 18. This QTL was associated with a pleiotropic effect on berry size or weight and on ripening date, and it was not possible to dissociate seedlessness and small berry size. At least four independent minor QTLs for seedlessness were identified in different linkage groups. These results are consistent with the previous hypothesis that seedlessness is controlled by three complementary recessive genes (minor QTLs) independently inherited and regulated by a dominant gene SdI (major QTL). Additional recombinant and phenotypic data are needed to differentiate how much of the phenotypic variation is due to this pleiotropic effect or to environmental influences.
Footnotes
↵3 Present address: Departamento de Ciencias Vegetales, Facultad de Agronomía e Ingeniería Forestal, Pontificia Universidad Catolica de Chile, Vicuña Mackenna 4860, Santiago, Chile.
The authors thank J.A. Cabezas, J.M. Martínez-Zapater, and A. Doligez for their helpful collaboration in training and discussion. We are also grateful to M.H. Castro and H. Prieto for their help in this work.
Acknowledgments: Research supported by FONDECYT-Chile, grant 1990204, and the INCO-DEV contract ICA4-CT-2001-10065: MASTER (Marker Assisted Selection for TablE gRape).
- Received January 2006.
- Revision received July 2006.
- Revision received May 2007.
- Copyright © 2007 by the American Society for Enology and Viticulture