Complex genetics controls natural variation among seed quality phenotypes in a recombinant inbred population of an interspecific cross between Solanum lycopersicum × Solanum pimpinellifolium

Authors

  • RASHID H. KAZMI,

    1. Wageningen Seed Lab, Laboratory of Plant Physiology, Wageningen University, Droevendaalsesteeg 1, NL-6708 PB Wageningen, the Netherlands
    Search for more papers by this author
    • These authors contributed equally to this paper and should both be considered first author.

  • NOORULLAH KHAN,

    1. Wageningen Seed Lab, Laboratory of Plant Physiology, Wageningen University, Droevendaalsesteeg 1, NL-6708 PB Wageningen, the Netherlands
    Search for more papers by this author
    • These authors contributed equally to this paper and should both be considered first author.

  • LEO A. J. WILLEMS,

    1. Wageningen Seed Lab, Laboratory of Plant Physiology, Wageningen University, Droevendaalsesteeg 1, NL-6708 PB Wageningen, the Netherlands
    Search for more papers by this author
  • ADRIAAN W. VAN HEUSDEN,

    1. Wageningen UR Plant Breeding, PO Box 386, NL-6700 AJ Wageningen, the Netherlands
    Search for more papers by this author
  • WILCO LIGTERINK,

    Corresponding author
    1. Wageningen Seed Lab, Laboratory of Plant Physiology, Wageningen University, Droevendaalsesteeg 1, NL-6708 PB Wageningen, the Netherlands
    Search for more papers by this author
  • HENK W. M. HILHORST

    1. Wageningen Seed Lab, Laboratory of Plant Physiology, Wageningen University, Droevendaalsesteeg 1, NL-6708 PB Wageningen, the Netherlands
    Search for more papers by this author

W. Ligterink. Fax: +31 317418094; e-mail: wilco.ligterink@wur.nl

ABSTRACT

Seed quality in tomato is associated with many complex physiological and genetic traits. While plant processes are frequently controlled by the action of small- to large-effect genes that follow classic Mendelian inheritance, our study suggests that seed quality is primarily quantitative and genetically complex. Using a recombinant inbred line population of Solanum lycopersicum × Solanum pimpinellifolium, we identified quantitative trait loci (QTLs) influencing seed quality phenotypes under non-stress, as well as salt, osmotic, cold, high-temperature and oxidative stress conditions. In total, 42 seed quality traits were analysed and 120 QTLs were identified for germination traits under different conditions. Significant phenotypic correlations were observed between germination traits under optimal conditions, as well as under different stress conditions. In conclusion, one or more QTLs were identified for each trait with some of these QTLs co-locating. Co-location of QTLs for different traits can be an indication that a locus has pleiotropic effects on multiple traits due to a common mechanistic basis. However, several QTLs also dissected seed quality in its separate components, suggesting different physiological mechanisms and signalling pathways for different seed quality attributes.

INTRODUCTION

Seed quality is the ability of seeds to germinate under a wide variety of environmental conditions and to develop into healthy seedlings. Seed quality is determined by several factors including genetic and physical purity, mechanical damage and physiological conditions, such as viability, germination, dormancy, vigour and uniformity (Dickson 1980; Hilhorst & Toorop 1997; Hilhorst 2007; Hilhorst et al. 2010). The physiological condition of seeds during development and maturation has a strong effect on ultimate seed quality. It is influenced by several environmental factors such as temperature, humidity, light and nutrients during the seed filling and maturation stages, by seed treatments (harvesting and processing) and by accumulated damage (Ouyang et al. 2002; Spano et al. 2007). Thus, seed quality is a complex trait governed by interactions between the genome and the environment (Koornneef, Bentsink & Hilhorst 2002) and therefore, seed quality can be challenged over the entire seed production chain. These quality-specific interactions are primarily expressed as germination, which is defined as the event that begins with the uptake of water by the seed and ends with the start of elongation by the embryonic axis, usually the protrusion of the radicle (Bewley 1997; Finch-Savage & Leubner-Metzger 2006). In the case of tomato, protrusion of the radicle through the surrounding layers (endosperm and testa) is considered to be the completion of germination. Thus, successful germination is determined by the balance between two opposing forces.

Abiotic stresses, such as extreme temperatures, low water availability, high salt levels, mineral deficiency and toxicity, are frequently encountered by plants in both natural and agricultural systems (Langridge, Paltridge & Fincher 2006; Eswaran et al. 2010). Higher plants have developed strategies to avoid abiotic stresses whereas these strategies are lost in agricultural crops. The most striking effect of abiotic stresses is on the yield of crops, which is estimated to be less than half under abiotic stress, as compared with normal growing conditions. Traditional approaches to improve the abiotic stress tolerance of crop plants by breeding have been of very limited success. This is mainly because of the difficulty of selecting for stress tolerance traits in traditional breeding programs. However, the natural variation among crop species can be used to cross desired traits from wild relatives and, for tomato, extensive abiotic stress tolerance has been identified in screens of land races and related wild species. Nevertheless, there is relatively little known about the molecular basis of abiotic stress tolerance in tomato species and there is still ample scope for improvement.

Substantial genetic variation for abiotic stresses exists within the cultivated tomato (Solanum lycopersicum; Wudiri & Henderson 1985; Moyle & Muir 2010), as well as in its related wild species, such as Solanum habrochaitis, Solanum pimpinellifolium and Solanum pennellii. These wild species offer the genetic resources for cold, temperature and water stress tolerance with respect to seed quality (Foolad & Lin 1998; Foolad et al. 2003a). However, rather limited efforts have been devoted to the physiological and genetic characterization of this variation in tomato to warrant its use for developing drought-tolerant cultivars (Kahn et al. 1993; Martin, Tauer & Lin 1999). This is in contrast with the considerable amount of research that has been conducted on abiotic stress in relation to other crop species, including rice (Oryza sativa L.; Zhang et al. 2001 and lettuce; Johnson et al. 2000). In a recent germplasm evaluation study, several wild tomato cultivars were identified as possessing the ability to germinate rapidly under abiotic stresses, including S. pimpinellifolium Mill. accession LA722 (Foolad et al. 2003a). S. lycopersicum is sensitive to cold, salt and drought stress during seed germination, whereas S. pimpinellifolium germinates rapidly under most conditions, including cold, salt and drought stress. Among the wild species of tomato, S. pimpinellifolium is the most closely related to S. lycopersicum and the only species for which natural introgression with S. lycopersicum has been demonstrated (Rick 1958). Accessions within this species are red fruited and can be readily hybridized with the cultivated tomato. Furthermore, in comparison with other wild tomato species, S. pimpinellifolium possesses fewer undesirable horticultural characteristics and thus has been frequently used as a genetic resource in tomato genetics and breeding programs.

Crop performance is the end result of the action of thousands of genes and their interaction with the environment. Conventional breeding has been very successful in raising the yield potential of crops (Borlaug & Dowswell 2003; Campos et al. 2004; Collins, Tardieu & Tuberosa 2008). Breeders have exploited genetic variability for crop improvement with very limited knowledge of factors governing it. However, this approach may become inadequate as the pressure to provide improvements will mount if global climate change increases the frequency and severity of abiotic constraints. Temperature stress, drought and salinity will be more prevalent in marginal areas with an increased demand for agricultural products and reduced availability of arable land and natural resources, such as water and fertilizers. Consequently, the genetic dissection of the quantitative traits controlling the adaptive response of crops to abiotic stress is a prerequisite to allow cost-effective applications of genomics-based approaches to breeding programs aimed at improving the sustainability and stability of yield under adverse conditions.

Consistent with the proposition that seed quality has a complex genetic basis, quantitative trait locus (QTL) studies of seed quality have generally revealed the influence of numerous QTLs of small to large phenotypic effect. Quantitative trait mapping of seed quality traits in common bean, sunflower, rapeseed, tomato and Arabidopsis has revealed numerous QTLs (Foolad et al. 2003a; Foolad, Zhang & Subbiah 2003b; Clerkx et al. 2004; Asghari 2007; Ebrahimi et al. 2008; Bentsink et al. 2010; Perez-Vega et al. 2010). S. lycopersicum is severely susceptible to environmental stresses (e.g. salt, drought, cold and high temperature) during seed germination and seedling growth, delaying the onset, rate and distribution of the germination events (Foolad, Subbiah & Zhang 2007). To take up the challenges manifested in uncovering the causal polymorphisms for QTLs, genomics tools are now also available for S. lycopersicum and these offer promising opportunities to unravel network mechanisms underlying complex quantitative traits (Collins et al. 2008). To elucidate the molecular mechanisms underlying quantitative traits, we analysed quantitative responses of tomato seed quality phenotypes in a structured recombinant inbred line (RIL) mapping population.

In the present study we used the RIL population generated from S. lycopersicum (cv. Moneymaker) and S. pimpinellifolium (G1.1554) (Voorrips et al. 2000). This population provides a valuable resource for the study of genes affecting complex phenotypes for seed quality as they allow isolation of the effect of a particular QTL from those of the entire genome, thus increasing our statistical power to dissect quantitative seed quality phenotypes, shaping a complex underlying mechanism.

MATERIALS AND METHODS

Plant material

S. lycopersicum cv. Moneymaker, a horticulturally superior, advanced tomato breeding line, was crossed with S. pimpinellifolium G1.1554, a self-compatible inbred accession of the wild species to produce 83 RILs to F8 (Voorrips et al. 2000). This population was genotyped for a total of 865 single nucleotide polymorphism (SNP) markers in F7.

Growth conditions and seed collection

The S. lycopersicum × S. pimpinellifolium RIL population was grown twice under controlled conditions in the greenhouse facilities at Wageningen University, the Netherlands. The day and night temperatures were maintained at 25 and 15 °C, respectively, with 16 h light and 8 h dark (long-day conditions). All the RILs were uniformly supplied with the basic dose of fertilizers and other nutrients. Seeds were extracted from healthy fruits and treated with 1% hydrochloric acid (HCL) to remove the large pieces of the pulp sticking onto the seeds. The solution of tomato seed extract with diluted hydrochloric acid was passed through a fine mesh sieve and washed with water to remove the remaining parts of the pulp and remnants of the hydrochloric acid. The seeds were processed and disinfected by soaking in a solution of trisodium phosphate (Na3PO4.12H2O). Finally, seeds were dried on clean filter paper at room temperature and were brushed to remove impurities with a seed brusher (Seed Processing Holland BV, Enkhuizen, The Netherlands, http://www.seedprocessing.nl). The cleaned seeds were dried for 3 d at 20 °C and were stored in a cool, dry storage room (13 °C and 30% RH) in paper bags.

Linkage analysis

The genetic linkage map consists of 12 individual linkage groups corresponding to the 12 chromosomes of tomato. Sequence information was used to study the segregation of parental alleles in the S. lycopersicum G1.1554 × S. pimpinellifolium cv. Moneymaker RIL population. Custom-made Infinium Bead arrays containing 5529 SNPs were used to genotype the RIL population. In total, 5529 SNP markers were used to genotype S. pimpinellifolium G1.1554 and S. lycopersicum cv. Moneymaker. The identical markers (no recombination between two markers) were removed and left 2251 polymorphic markers out of 5529 SNPs. The loci with identical segregation patterns were removed before calculating the map. The remaining 865 unique markers were used for calculating the maps of all chromosomes. Map construction was done in JoinMap 4 (Van Ooijen & Voorrips 2001) based on recombination frequency and Haldane's mapping function by incorporating the available SNP marker dataset for 83 RILs. The name of each marker on the tomato linkage map corresponds to the position on the tomato genome sequence version SL2.40 (http://solgenomics.net/organism/solanum_lycopersicum/genome).

Seed phenotyping

Germination assay

Germination assays were performed in triplicate with seeds of the parents and the RILs, which were sown under aseptic conditions on germination trays (21 × 15 cm; DBP Plastics NV, Antwerpen, Belgium, http://www.dbp.be) containing 15 mL water (non-stress condition) or NaCl, polyethylene glycol (PEG) or H2O2 (stress conditions) and one layer of white filter paper (20.2 × 14.3 cm white blotter paper; Allpaper BV, Zevenaar, The Netherlands, http://www.allpaper.nl). Each germination tray contained two lines and 45 seeds of each line and was considered one replicate. Germination trays were placed in a completely randomized design with three replications per sample. A maximum of 17 trays were piled up with two empty trays on both the top and the bottom end of the stack, with 15 mL water and two layers of white filter paper, to prevent unequal evaporation. The trays were covered with tightly fitting lids and the whole pile was wrapped in a closed transparent plastic bag and incubated at 4 °C for 3 d for stratification. Subsequently, the bags where placed randomly in an incubator at 25 °C in the dark (type 5042; Seed Processing Holland), except for brief intervals when germination was counted under laboratory (fluorescent) lighting. Germination responses were scored visually as radicle protrusion at eight hourly intervals for 10 consecutive days during the period of most rapid germination, and at longer subsequent intervals, until no additional germination was observed.

Salt, osmotic and oxidative stress

Salt, osmotic and oxidative stress tolerance treatments were applied in germination trays with 15 mL of the corresponding solution on a piece of filter paper. Salt stress was estimated by germinating seeds in different concentrations of NaCl. Osmotic potentials were established through aqueous solutions of polyethylene glycol (PEG 8000; Sigma, St Louis, MO, USA) measured in megapascal (MPa). Specific concentrations of NaCl and PEG 8000 were determined with the Solute Potential and Molar-Molal-g Solute/g Water Interconversion (SPMM) program (Michel & Radcliffe 1995). Tolerance to hydrogen peroxide was estimated by germinating seeds on filter paper saturated with a solution of 300 mm H2O2.

Low- and high-temperature stress

All RIL genotypes were subjected to suboptimal temperature regimes in order to test their response to temperature stress. Germination was monitored during incubation for 10 d at 12 °C in the case of cold stress, and at 35 and 36 °C to test for high-temperature stress response.

Statistical and genetic analyses

Calculation of Gmax, t10−1, t50−1, MGR, U7525−1, AUC and estimation of means

In this study, the curve-fitter module of the Germinator package was used for analysing different parameters of the cumulative germination curves (Joosen et al. 2010). Parental lines and the RIL population were subjected to different germination conditions, and maximum germination (Gmax, %), the onset of germination [t10−1; reciprocal of time to 10% of germination of viable seeds (h−1)], the rate of germination [t50−1; reciprocal of time to 50% of the germination of viable seeds (h−1)], MGR = mean germination rate, which is reciprocal of the mean germination time (MGT−1), uniformity (U7525−1, reciprocal of time interval between 75 and 25% viable seeds to germinate; h−1) and area under the germination curve [AUC; the integration of the fitted curve between t = 0 and a user-defined end point (x)] were determined. A full description of the validity and assessment of calculated parameters is available elsewhere (Thomson & El-Kassaby 1993; Bradford 1995; Hayashi, Aoyama & Still 2008; Alonso-Blanco et al. 2009; Landjeva, Lohwasser & Borner 2010). The t10−1, t50−1 and U7525−1 were calculated only for those treatments where seeds of the majority of RILs (>80%) completed a corresponding fraction (10, 50 and 75% or more) of germination (Hayashi et al. 2008; Galpaz & Reymond 2010). For germination parameters, the means of the three replicates were calculated and these were transformed to a probit regression model using the R module ‘VGAM’ (http://www.r-project.org). Means of transformed data were used for QTL analysis.

Identification of QTLs

QTL analysis was performed on the basis of the established marker linkage map of the RIL population, which contains 865 SNP markers. The mapping software MapQTL®5.0 (Van Ooijen & Maliepaard 2003) was used for identifying QTL positions in the genome for a given trait. A multiple QTL mapping model (MQM) was used to identify potential QTLs (Jansen et al. 1995) as implemented in MapQTL®5.0. In this method, background markers are selected to take over the role of the putative QTL as cofactors to reduce the residual variance. A two-stage MQM analysis was performed. In the first stage, conventional interval mapping was performed at a 2 cM interval; the logarithm-of-odds (LOD) profiles from interval mapping were inspected and the marker closest to each LOD peak was selected as the cofactor to perform further MQM mapping analysis. Several cycles were performed to obtain the potentially maximum number of cofactors for the MQM analysis. These cofactor markers were then subjected to backward elimination, as implemented in MapQTL®5.0, in order to select the best model for the second-stage MQM analysis. Such a backward elimination procedure leaves out one cofactor at a time in order to create a subset of cofactors. The likelihood of each of these subset models is compared with the likelihood of the full model with all cofactors, and the subset model that causes the smallest change in likelihood is chosen as the starting set for a subsequent round of elimination. This process continues until the change in likelihood is significant according to the 0.002 P-value for the test. The set of cofactors then retained was used in the second stage of the MQM analysis. In the final LOD profile, QTLs were affirmed according to the threshold LOD scores ranging from 2.0 to 7.0 (genome-wide false-positive rate 5%), depending on chromosome map length and the number of chromosome pairs (Van Ooijen 1999). To determine whether QTLs among different traits were significantly co-located, firstly, the number of QTLs from different traits that had overlapping confidence intervals is determined. Then, QTL confidence intervals were randomized across the genome 1000 times, and the distribution of the number of overlapping QTLs of different traits determined. If this number of randomized QTLs was less than the original QTL overlap 95% of the time, the co-location was deemed significant.

Analysis of heritability and epistasis

Broad-sense heritability (h2b) was estimated from one-way random effects of analysis of variance [anova, SPSS version 19.0 (SPSS Inc., Chicago, IL, USA)] with the equation:

image

where σ2g is the genetic variance and σ2e is the environmental variance (Keurentjes et al. 2007). Significant differences among all means of the RILs were estimated using one-way anova followed by a least significant difference (LSD) test.

A two-dimensional genome-wide epistatic interactions analysis was performed using the R/qtl software package (Broman et al. 2003) in order to identify epistatic interactions contributing to variation in the seed germination parameters: Gmax, t10−1, t50−1, MGR, U7525−1 and AUC. Each chromosomal region (tomato chromosomes 1–12) was considered jointly with all other chromosomal regions throughout the genome for each seed quality phenotype analysed. The statistical analysis of epistasis as implemented in the R/qtl software package consists of nested linear model fitting for each pair of loci tested for an epistatic interaction, as described previously (Koller et al. 2009). To obtain appropriate genome-wide significance thresholds for the epistasis results and properly account for the large number of tests considered in the genome-by-genome scan, 10 000 permutation tests (Doerge & Churchill 1996) were performed with the Haley–Knott regression method (Broman et al. 2003). In this manner, the LOD significance threshold of the maximum genome-wide interaction was found to be 4.09; for full model (lod.full) and conditional interactive model (lod.fv), LOD significance thresholds were found to be 6.04 and 4.63, respectively. Interacting QTL pairs were only reported if all of these thresholds were exceeded. Specifically, the 42 traits measured of each RIL were randomly reassigned as a group across the 83 RILs resulting in a permuted dataset (Spano et al. 2007). By keeping all phenotypic data together, the underlying phenotypic correlations were preserved. The epistasis analysis was then performed across the whole genome and the resulting maximum LOD scores for linkage for each phenotype were recorded.

RESULTS

Distribution, means and heritability

To investigate the genetic architecture of seed quality traits, we measured phenotypes of the 83 F8 RILs. The population was derived from a cross between S. lycopersicum (cv. Moneymaker) and S. pimpinellifolium (G1.1554). Seeds of the wild accession S. pimpinellifolium G1.1554 germinated significantly more rapidly than seeds of the breeding line S. lycopersicum cv. Moneymaker under non-stress (control) as well as salt, osmotic, cold and temperature stress conditions (Table 1). The germination parameters were calculated only for those traits in which a corresponding fraction (10, 50 and 75% or more) of seeds completed germination. For example, under control and salt (−0.3 MPa NaCl), seeds from the majority of RILs surpassed 80% of germination and all parameters like Gmax, t10−1, t50−1, MGR, U7525−1 and AUC were obtained. On the other hand, if final germination fell below the particular fraction, t10−1 and t50−1, those traits were not calculated, for example, in case of osmotic (−0.3, −0.5 MPa PEG), cold (12 °C), high-temperature (36 °C) and oxidative stress conditions, Gmax, t10−1, MGR and AUC were obtained but t50−1 and U7525−1 were not, as the final germination percentage was too low to calculate meaningful values.

Table 1.  Means of germination traits (± SD) for the parental genotypes and the F7 population of cross between Solanum lycopersicum (Money) and Solanum pimpinellifolium (Pimp) in the control (non-stress), salt, osmotic, cold, temperature and oxidative stress treatments
TreatmentGenotypesGmaxt10−1(×100)t50−1(×100)MGR(×100)U7525−1(×100)AUC
  1. Money, Solanum lycopersicum; Pimp, Solanum pimpinellifolium; Gmax (%), maximum germination; t10−1, t50−1, reciprocal of time to respectively 10 and 50% of viable seeds to germinate (h−1); MGR, mean germination rate (reciprocal of the mean germination time; MGT−1); U7525−1, uniformity (reciprocal of time interval between 75 and 25% viable seeds to germinate; h−1); AUC, area under the germination curve (integration of fitted curve between 0 and 200 h); nd, not determined; RIL, recombinant inbred line.

ControlMoney100.0 ± 0.01.703 ± 0.0321.237 ± 0.0541.198 ± 0.0613.865 ± 0.488115.3 ± 4.9
Pimp100.0 ± 0.03.663 ± 0.1062.910 ± 0.0052.652 ± 0.06117.762 ± 0.290165.0 ± 0.4
RILs92.5 ± 11.32.390 ± 0.6821.811 ± 0.5591.799 ± 0.6077.111 ± 4.168127.0 ± 27.5
Salt I
(−0.3 MPa NaCl)
 Money85.1 ± 1.01.230 ± 0.0480.960 ± 0.0400.954 ± 0.0353.823 ± 0.15677.7 ± 1.8
Pimp99.6 ± 0.42.609 ± 0.2092.016 ± 0.1591.974 ± 0.1557.682 ± 0.476148.0 ± 3.4
RILs86.7 ± 16.11.547 ± 0.4191.180 ± 0.3191.170 ± 0.3014.804 ± 2.40094.5 ± 30.3
Salt II
(−0.5 MPa NaCl)
 Money85.7 ± 0.80.694 ± 0.0300.502 ± 0.0190.498 ± 0.001nd16.7 ± 4.1
Pimp99.6 ± 0.41.659 ± 0.1011.234 ± 0.0021.200 ± 0.013nd115.9 ± 1.0
RILs67.9 ± 29.61.153 ± 0.3920.857 ± 0.2780.840 ± 0.262nd57.1 ± 39.7
Osmotic I
(−0.3 MPa PEG)
 Money46.9 ± 19.40.810 ± 0.096nd0.653 ± 0.051nd14.0 ± 6.4
Pimp95.5 ± 2.91.594 ± 0.256nd1.107 ± 0.162nd102.5 ± 15.8
RILs54.7 ± 28.71.176 ± 0.470nd0.844 ± 0.261nd43.9 ± 15.6
Osmotic II
(−0.5 MPa PEG)
 Money38.3 ± 9.40.629 ± 0.046nd0.563 ± 0.002nd8.31 ± 1.5
Pimp70.8 ± 6.30.872 ± 0.061nd0.698 ± 0.045nd28.9 ± 6.5
RILs57.8 ± 19.50.773 ± 0.202nd0.638 ± 0.099nd20.2 ± 10.4
Cold stress
12 °C
 Money5.2 ± 2.2ndndndndnd
Pimp100.0 ± 0.00.853 ± 0.048nd0.754 ± 0.025nd68.5 ± 3.9
RILs37.2 ± 18.30.568 ± 0.125nd0.508 ± 0.080nd9.5 ± 3.3
High-temperature stress I
35 °C
 Money72.8 ± 8.21.224 ± 0.1300.736 ± 0.1220.751 ± 0.087nd45.5 ± 9.7
Pimp100.0 ± 0.02.803 ± 0.0122.426 ± 0.0092.305 ± 0.003nd158.2 ± 0.1
RILs77.6 ± 28.11.889 ± 0.6951.359 ± 0.5101.325 ± 0.507nd93.2 ± 35.6
High-temperature stress II
36 °C
 Money3.1 ± 1.3ndndndndnd
Pimp93.1 ± 3.62.507 ± 0.226nd1.788 ± 0.139nd134.0 ± 8.9
RILs33.9 ± 15.91.826 ± 0.764nd1.254 ± 0.416nd39.0 ± 14.5
Oxidative stress
(300 mm H2O2)
 Money64.2 ± 2.70.796 ± 0.032nd0.642 ± 0.013nd24.3 ± 4.8
Pimp3.1 ± 0.9ndndndndnd
RILs39.4 ± 19.40.816 ± 0.281nd0.649 ± 0.124nd17.8 ± 9.6

In most cases, seeds of the RIL population germinated intermediately between the two parental lines, indicating the inheritance of rapid germination from G1.1554 to the progeny (Table 1, Fig. 1). However, we also observed transgressive segregation for the seed quality traits (Table 1, Fig. 1). This implies that the different seed phenotypes shown in the S. lycopersicum and S. pimpinellifolium parental lines result from the presence of distinct genetic polymorphisms with antagonistic effects contributed by each parent. Estimates of the broad-sense heritability of different seed quality traits differed considerably among seed phenotypes studied across different treatments (Table 2). Heritability estimates for different germination-related traits indicated that genetic variation exists for seed quality phenotypes under control conditions, as well as salt, osmotic, cold, high-temperature and oxidative stress conditions, and the germination characteristics in the RIL population are highly heritable (Table 2). The RILs showed great phenotypic variation with regard to seed quality traits; Gmax showed a slight negative skew and t10−1, t50−1, MGR, U7525−1 and AUC a stronger positive skew (Fig. 1).

Figure 1.

Figure 1.

Frequency distributions of non-normalized data of all traits in the Solanum lycopersicum and Solanum pimpinellifolium recombinant inbred line (RIL) population. Seed quality traits determined under control conditions, salt stress I (−0.3 MPa NaCl), salt stress II (−0.5 MPa NaCl), osmotic stress I (−0.3 MPa PEG), osmotic stress II (−0.5 MPa PEG), cold stress (12 °C), high-temperature stress I (35 °C), high-temperature stress II (36 °C) and oxidative stress. The average parental value is indicated with a solid arrow for S. lycopersicum and a dashed arrow for S. pimpinellifolium parents. AUC, area under the germination curve; MGT, mean germination time.

Figure 1.

Figure 1.

Frequency distributions of non-normalized data of all traits in the Solanum lycopersicum and Solanum pimpinellifolium recombinant inbred line (RIL) population. Seed quality traits determined under control conditions, salt stress I (−0.3 MPa NaCl), salt stress II (−0.5 MPa NaCl), osmotic stress I (−0.3 MPa PEG), osmotic stress II (−0.5 MPa PEG), cold stress (12 °C), high-temperature stress I (35 °C), high-temperature stress II (36 °C) and oxidative stress. The average parental value is indicated with a solid arrow for S. lycopersicum and a dashed arrow for S. pimpinellifolium parents. AUC, area under the germination curve; MGT, mean germination time.

Table 2.  Chromosomal location of the QTL associated with seed quality traits of tomato Solanum lycopersicum/Solanum pimpinellifolium RIL population under control (non-stress), salt, osmotic, cold, high-temperature and oxidative stress conditions
 Chr.aMarker peakbSupportLODdExplained varianceeTotalEffectsgHeritabilityh
intervalcexplained variancef
(cM)(%)(%)
  • Gmax (%), maximum germination; t10−1, t50−1, reciprocal of time to respectively 10 and 50% of viable seeds to germinate (h−1); MGR, mean germination rate (reciprocal of the mean germination time; MGT−1); U7525−1, uniformity (reciprocal of time interval between 75 and 25% viable seeds to germinate; h−1); AUC, area under the germination curve (integration of fitted curve between 0 and 200 h). QTL, quantitative trait locus; RIL, recombinant inbred line; LOD, logarithm-of-odds.

  • a

    Chromosome number.

  • b

    Name (= physical position) of marker closest to the QTL peak.

  • c

    1-LOD support interval of QTL.

  • d

    LOD score that represents the significance threshold for QTL (P = 0.002) obtained by permutation tests.

  • e

    Percentage of variation explained by individual QTLs.

  • f

    Percentage of the total variance explained by genetic factors for a single trait as estimated by MapQTL.

  • g

    Effect of QTL calculated as µB − µA, where A and B are RILs carrying S. lycopersicum and S. pimpinellifolium alleles at the QTL position, respectively. µB and µA were estimated by MapQTL. Effects are given in percentage (Gmax) and h−1 (t10−1, t50−1, MGR, U7525−1).

  • h

    Broad-sense heritability estimate for each trait, estimated as the proportion of phenotypic variance explained by genotype in a one-way analysis of variance model; calculated as inline image

Control 
Gmax       0.89
 715592910.0–26.02.2811.911.91.39 
t10−1       0.88
 4877728573.1–80.43.3911.859.30.70 
63410082838.9–56.23.429.8 0.64 
643582592102.7–108.35.5920.6 0.96 
966917748100.8–112.72.789.4 −0.66 
124784530841.4–64.02.307.7 −0.57 
t50−1       0.89
 45657052465.1–80.42.297.736.80.29 
643582592101.1–107.33.7413.1 0.37 
85709950472.6–87.82.257.6 −0.29 
124784530849.5–63.02.488.4 −0.30 
MGR       0.90
 4877728565.1–81.92.157.236.60.57 
643582592101.1–107.33.6812.9 0.75 
85709950472.6–87.82.327.9 −0.59 
124784530849.5–63.02.548.6 0.61 
U7525−1       0.92
 35880282471.7–82.63.3412.842.9−0.72 
45967861286.9–108.33.2312.3 0.72 
72807570433.7–56.72.609.6 −0.69 
85709950472.5–86.62.258.2 −0.81 
AUC       0.80
 23491415623.7–34.22.5712.322.60.96 
45647530869.1–81.92.1810.3 0.69 
Salt I
(−0.3 MPa NaCl)
Gmax       0.93
 5671112259.8–66.43.3215.715.7−0.42 
t10−1       0.79
 45647530865.1–87.02.278.139.30.58 
64358259299.5–109.33.2011.7 0.71 
11547248210.7–17.83.0711.2 −0.70 
124498779248.9–54.52.338.3 −0.60 
t50−1       0.94
 45647530868.2–85.02.778.552.10.30 
643582592101.1–109.34.8915.9 0.42 
966917748106.5–112.72.136.4 −0.27 
114700828020.7–36.32.618.0 −0.29 
124498779248.9–54.54.1813.3 −0.38 
MGR       0.89
 1704403051.8–65.72.025.657.7−0.50 
45701360868.2–85.02.949.3 0.62 
643582592102.4–108.35.1517.4 0.90 
966917748106.5–112.72.417.5 −0.59 
1154724829.0–17.82.196.8 −0.54 
124498779248.9–54.53.4511.1 −0.69 
U7525−1       0.94
 417673820.0–18.92.019.022.20.65 
72807570439.2–56.32.8813.2 0.75 
AUC       0.86
 644674784100.5–112.73.1613.726.10.81 
114828325222.7–35.32.8612.4 −0.74 
Salt II
(−0.5 MPa NaCl)
Gmax       0.85
 45817488485.0–93.23.1314.427.10.79 
5753396160.7–67.82.7812.7 −0.37 
t10−1       0.68
 2337523084.6–26.72.147.222.70.59 
45808128473.1–95.12.267.7 0.85 
64376306099.5–112.72.307.8 0.54 
t50−1       0.79
 45808128485.0–93.22.9511.431.00.35 
64376306099.5–112.72.9911.6 0.35 
85709950478.4–84.82.118.0 −0.29 
MGR       0.85
 23375230815.4–26.63.4315.715.71.04 
AUC       0.72
 45817488485.0–93.23.3612.533.90.78 
643582592101.1–109.33.5613.4 0.82 
96691774899.6–112.72.208.0 −0.64 
Osmotic I
(−0.3 MPa PEG)
Gmax       0.91
 45817488474.1–93.22.6211.229.90.70 
5671112255.7–67.82.3810.1 −0.32 
9487740.00–12.12.058.6 0.60 
t10−1       0.85
 2313481247.6–22.43.5716.827.20.94 
4465411441.0–52.82.2810.4 0.65 
MGR       0.89
 2337523087.6–23.73.5014.834.80.96 
4471101541.0–54.12.8311.8 0.72 
12753668339.4–64.02.028.2 −0.61 
AUC       0.85
 45817488464.1–95.13.5312.848.10.74 
643702064102.4–108.34.9418.7 0.94 
966917748106.5–112.72.709.6 −0.68 
12439760735.2–48.32.017.0 −0.56 
Osmotic II
(−0.5 MPa PEG)
Gmax       0.87
 45454139264.1–78.82.2211.911.90.72 
t10−1<       0.83
 2349141569.6–31.32.887.962.30.77 
45967861293.4–100.04.9514.4 0.81 
643023484101.1–107.35.1615.2 0.83 
96626038498.9–105.32.988.2 −0.62 
124797620857.2–62.45.5816.6 −0.86 
MGR       0.53
 23375230815.4–31.32.9413.513.50.91 
AUC       0.88
 23491415626.7–33.33.4511.744.10.94 
45454139261.2–80.42.959.8 0.65 
64304641699.5–108.33.0710.3 0.70 
124797620857.2–63.03.6012.3 −0.75 
Cold stress
12 °C
Gmax       0.88
 16922778454.7–65.62.079.632.4−0.61 
521661319.6–35.02.8914.0 0.72 
644674784105.3–112.03.398.8 0.78 
t10−1       0.65
4493594027.4–54.12.3811.942.10.69 
5251528720.6–38.62.089.1 0.62 
643582592101.1–109.32.2411.2 0.67 
715592917.0–38.72.299.9 −0.64 
MGR       0.74
 7331748424.0–42.22.6312.612.6−0.79 
AUC        
 16922778461.7–65.72.408.937.0−0.610.86
35749939253.0–76.72.017.5 −0.55 
64467478499.5–112.72.5310.7 0.71 
114858606413.3–35.32.339.9 −0.63 
High-temperature stress I
35 °C
Gmax114640836818.7–30.02.8614.714.7−0.370.91
t10−1       0.80
 45507629265.1–72.52.4910.033.00.64 
644674784101.1–112.15.3123.0 1.02 
t50−1       0.79
 45501458065.1–88.32.189.228.70.31 
644674784101.1–112.14.3819.5 0.48 
MGR       0.88
 16922778461.7–65.72.7911.635.0−0.71 
644674784101.1–112.14.6323.4 1.04 
AUC       0.90
 45834063685.0–96.12.017.630.70.58 
643763060101.1–110.13.8215.2 0.85 
114640836818.7–32.32.087.9 −0.58 
High-temperature stress II
36 °C
Gmax       0.89
 64358259297.5–109.32.3912.712.70.70 
t10−1       0.75
 644905196110.1–112.75.3631.642.51.16 
966710096101.5–112.72.0910.9 −0.69 
MGR       0.93
 644905196111.1–112.74.8928.841.71.10 
 966710096103.5–111.42.4112.9 −0.75 
AUC       0.85
 63410082842.9–64.12.2612.112.10.67 
Oxidative stress
(300 mm H202)
Gmax       0.91
 56230740481.6–96.94.4615.240.3−0.88 
64002537674.9–92.62.126.7 −2.20 
81568409653.1–60.55.2518.4 0.81 
t10−1       0.74
 2313481240.0–22.42.307.676.30.11 
45808128473.1–91.03.7815.3 0.82 
64358259297.2–112.74.2117.3 0.63 
76149496483.2–90.63.4013.6 0.58 
81568409652.2–65.63.5914.3 0.90 
105361470.0–9.52.498.2 −0.90 
MGR       0.91
 2313481240.0–15.42.936.569.50.61 
45677342475.1–78.85.4916.3 0.83 
64358259299.5–110.15.4816.3 0.89 
76149496483.8–90.63.409.4 0.67 
81568409654.2–58.66.7721.0 0.98 
AUC       0.90
 56210079679.6–96.94.7217.036.6−0.61 
63901000027.5–110.02.193.9 0.60 
81568409652.1–63.54.3115.7 0.96 

Identification of QTLs for germination potential under different conditions

The map position and characteristics of the QTLs associated with the studied seed phenotypes under non-stress (control) and stress conditions are summarized in Tables 2 and 3. We found that individual QTLs mapped to specific regions of the tomato genome. We used an LOD threshold of 2.0 to investigate putative QTLs where seed quality phenotypes map. Figure 2 displays a heatmap of LOD profiles. In this way, QTLs can be visualized and global ‘hot spots’ and empty regions across the 12 chromosomes can be seen (Fig. 2).

Table 3.  Summary of quantitative trait locus (QTL) of seed quality traits in Solanum lycopersicum/Solanum pimpinellifolium recombinant inbred line population
TreatmentsTraitsaQTL (nr)bRange ofTotal explained variance (%)d
explained variance (%)c
  • a

    Gmax (%), maximum germination; t10−1, t50−1, reciprocal of time to respectively 10 and 50% of viable seeds to germinate (h−1); MGR, mean germination rate (reciprocal of the mean germination time; MGT−1); U7525−1, uniformity (reciprocal of time interval between 75 and 25% viable seeds to germinate; h−1); AUC, area under the germination curve (integration of fitted curve between 0 and 200 h).

  • b

    Number of QTLs detected.

  • c

    Range of explained variance for QTLs.

  • d

    Total explained variance for each trait.

Control
 Gmax111.911.9
t10−1511.8–20.659.3
t50−147.6–13.136.8
MGR47.2–12.936.6
U7525−147.2–12.842.9
AUC210.3–12.322.6
Salt stress I
(−0.3 MPa NaCl)
 Gmax115.715.7
t10−148.1–11.239.3
t50−156.4–13.352.1
MGR65.6–17.457.7
U7525−129.0–13.222.2
AUC212.4–13.726.1
Salt Stress II
(−0.5 MPa NaCl)
 Gmax212.7–14.427.1
t10−137.2–7.822.7
t50−138.0–11.631.0
MGR115.715.7
AUC38.0–13.433.9
Osmotic stress I
(−0.3 MPa PEG)
 Gmax38.6–11.229.9
t10−1210.4–16.827.2
MGR38.2–14.834.8
AUC47.0–18.748.1
Osmotic stress II
(−0.5 MPa PEG)
 Gmax113.5–13.511.9
t10−157.9–16.662.3
MGR113.5–13.513.5
AUC49.8–12.344.1
Cold stress
12 °C
 Gmax38.8–14.032.4
t10−149.1–11.942.1
MGR112.612.6
AUC47.5–10.737.0
Temperature stress I
35 °C
 Gmax114.714.7
t10−1210.0–23.033.0
t50−129.2–19.528.7
MGR211.6–23.435.0
AUC310.030.7
Temperature stress II
36 °C
 Gmax112.712.7
t10−1210.9–31.642.5
MGR212.9–28.841.7
AUC111.612.1
Oxidative stress
(300 mm H2O2)
 Gmax36.7–18.440.3
t10−167.6–17.376.3
MGR56.5–16.369.5
AUC33.9–17.036.6
Figure 2.

Genomic locations of quantitative trait locus (QTL) identified for seed quality traits. Tomato chromosomes are identified by numbers (1–12), with centimorgans ascending from the left to right; chromosomes are separated by white lines. Control indicates germination phenotypes under optimal condition. Coloured cells indicate QTL significant at P = 0.002 in multiple QTL mapping models [1-logarithm-of-odds (LOD)]. The LOD colour scale is indicated, showing blue and light blue when the Solanum pimpinellifolium (Pimp) allele, and yellow and red when the Solanum lycopersicum (Money) allele, at that marker results in an elevated level of seed quality phenotype. QTL positions, LOD scores, effects and hb values are provided in Table 2. AUC, area under the germination curve; MGR, mean germination rate.

QTL for germination under non-stress conditions

To distinguish between loci specific for regulation of germination traits under stress versus non-stress conditions, the latter were determined using the germination traits, that is Gmax, t10−1, t50−1, MGR, U7525 and AUC. The germination phenotypes were calculated only for those traits in which a corresponding fraction (10, 50 and 75% or more) of seeds completed germination. Although we did analyse rate of germination using a number of rate traits (t10−1, t50−1, MGR) as stated in Table 1, in order to avoid repetition and unnecessary complication, we will explicitly discuss t10−1 in the results. One QTL was detected for Gmax on chromosome 7 with an explained variance of 11.9% (Tables 2 & 3, Fig. 2). QTL analysis revealed five loci for t10−1, one each on chromosomes 4, 9, 12 and two on chromosome 6. In total, these loci accounted for 59.3% of explained variance (Tables 2 & 3, Fig. 2). Four QTLs were identified for U7525−1 on chromosomes 3, 4, 7 and 8, which explained 42.9% of the total variance observed. Two QTLs were revealed for AUC, one each on chromosomes 2 and 4 which explained 22.6% of the total variance (Tables 2 & 3, Fig. 2).

QTL for germination under salt stress conditions

Several QTLs were found to be associated with the tested germination traits (Tables 2 & 3) at −0.3 MPa (low) and −0.5 MPa (high) NaCl levels. For Gmax, one QTL was found on chromosome 5 at −0.3 MPa and two QTLs were revealed at −0.5 MPa, one each on chromosomes 4 and 5, which explained 15.7 and 27.1% of the total variance observed, respectively (Tables 2 & 3, Fig. 2). For t10−1, four QTLs were found, one each on chromosomes 4, 6, 11 and 12 under −0.3 MPa which explained variance of 39.3%, whereas three loci were revealed on chromosomes 2, 4 and 6 at −0.5 MPa with a total explained variance of 22.7% (Tables 2 & 3, Fig. 2). Furthermore, for U7525−1 under low salt stress, two QTLs were identified at chromosomes 4 and 7. In total, these loci explained 22.2% of the variance, whereas in the case of high salt level U7525−1 was not calculated as the majority of RILs did not reach a final germination percentage above 75%. For AUC, two QTLs were found on chromosomes 6 and 11 at −0.3 MPa which explained 26.1% of the variance and three QTLs were revealed for −0.5 MPa NaCl on chromosomes 4, 6 and 9 which explained 33.9% of the variance (Tables 2 & 3, Fig. 2). In a majority of cases, the same QTLs were identified in both levels; however, there were few instances where additional QTLs were identified in one of the salt stress levels (Fig. 2).

QTL for germination under osmotic stress conditions

QTL analysis was carried out in the case of osmotic stress for germination-related traits at both low and high (−0.3 and −0.5 MPa PEG) osmotic stress conditions (Tables 2 & 3). Three QTLs were identified for Gmax under low osmotic stress on chromosomes 4, 5 and 9, whereas for high osmotic stress one QTL on chromosome 4 was identified, which explained 29.9 and 11.9% of the total variance, respectively (Tables 2 & 3, Fig. 2). For t10−1, two QTLs where identified for low osmotic stress on chromosomes 2 and 4, which explained 27.2% of the total variance, whereas at high osmotic stress five QTLs were identified, one each on chromosomes 2, 4, 6, 9 and 12 with a total explained variance of 62.3% (Tables 2 & 3, Fig. 2). The U7525−1 was not calculated as the final germination percentage was too low to calculate meaningful values for the corresponding fraction, as previously described. Four QTLs were identified for AUC in case of low osmotic stress on chromosomes 4, 6, 9 and 12, and four QTLs were detected at high osmotic stress conditions, one each on chromosomes 2, 4, 6 and 12 (Tables 2 & 3, Fig. 2), which accounted for 48.1 and 44.1% of the total explained variance, respectively. Similar as described for salt, in a majority of cases the same QTLs were identified in both levels; however, there were few instances where additional QTLs were identified in one of the osmotic stress levels (Fig. 2).

QTL for germination under temperature stress conditions

Cold stress.  Three QTLs were found for Gmax at 12 °C on chromosomes 1, 5 and 6, which accounted for 32.4% of the total explained variance (Tables 2 & 3, Fig. 2). For t10−1, four QTLs were found on chromosomes 4, 5, 6 and 7 with 42.1% of the total explained variance (Tables 2 & 3, Fig. 2), whereas U7525−1 was not obtained as the final germination percentage was too low to calculate meaningful values. Four QTLs were found for AUC at 12 °C on chromosomes 1, 3, 6 and 11 with 37.0% of the total explained variance.

High temperature.  One QTL each on chromosomes 11 and 6 was found for Gmax, at 35 and 36 °C, which explained 14.7 and 12.7% of the variance, respectively (Tables 2 & 3, Fig. 2). One QTL each on chromosomes 4 and 6 for t10−1 was identified at 35 °C whereas two QTLs on chromosomes 6 and 9 at 36 °C were found, which explained 28.7 and 42.5% of the total variance, respectively (Tables 2 & 3, Fig. 2). U7525−1 was not calculated as the majority of RILs did not reach a final germination percentage above 75%. Three QTLs were found, one each on the chromosomes 4, 6 and 11 for AUC at 35 °C and one QTL on chromosome 6 for AUC at 36 °C, which explained 30.7 and 12.1% of the variation, respectively (Tables 2 & 3, Fig. 2).

QTL for germination under oxidative stress conditions

Three QTLs were identified for Gmax on chromosomes 5, 6 and 8 for oxidative stress, which explained 40.3% of the total variance (Tables 2 & 3, Fig. 2). QTL analysis revealed six QTLs for t10−1 on chromosomes 2, 4, 6, 7, 8 and 10 with 76.3% of the total explained variation (Tables 2 & 3, Fig. 2). No estimate for U7525−1 was obtained as the final germination percentage was too low to calculate meaningful values. For AUC, three QTLs were found on chromosomes 5, 6 and 8 accounting for 36.6% of the total explained variance (Tables 2 & 3, Fig. 2).

Shared QTLs among seed phenotypes

Permutation tests conducted onto all −1LOD QTL intervals allowed to compare and estimate the level of overlapping QTLs between phenotypic traits where occurrences of overlapping QTLs between different seed quality traits were considered highly significant with 1 P-value of 0.99 or 1.0. Seven QTL clusters positioned onto chromosomes 1, 2, 4, 6, 8, 9 and 12 were identified as affecting different seed germination traits with an overlapping proportion ranging from 62.5 to 100% at −1LOD (Fig. 2). QTLs positioned onto chromosomes 1, 2, 4, 6, 9 and 12 also revealed at −1LOD a significant overlap (from 91.6 to 100%) between QTL clusters for rate of germination parameters (t10−1, t50−1, MGR). QTLs detected for Gmax, t10−1, t50−1 and MGR co-located significantly onto three chromosomes: chromosomes 6, 9 and 12 (Fig. 2). The overlapping range between QTLs affecting simultaneously t10−1, t50−1 and MGR varied from 90.0 to 100% (Fig. 2). QTLs involving Gmax and AUC traits co-located together onto the chromosomes 4, 6 and 11, whereas AUC and t10−1, t50−1 and MGR QTLs were significantly overlapping (from 79.4 to 100%) onto chromosomes 3, 4, 6, 8, 9, 11 and 12 (Fig. 2).

To investigate associations among characteristics at the phenotypic level, a correlation matrix was generated by performing Pearson correlation analysis for all pairs of measured traits across the whole population. This analysis used average values calculated from all raw determinations for a given trait/RIL pair. Pearson correlation coefficients (Rp) and accompanying false discovery rate (FDR)-corrected P-values (PBH; Benjamini & Yekutieli 2001) are provided in Supporting Information Table S1. Using the Pearson correlation coefficient to calculate relationships among seed quality phenotypes concerned, a number of low to high significant correlations were observed for seed phenotypes under different germination conditions (Fig. 3 and Supporting Information Fig. S1, Supporting Information Table S1). For instance, Gmax in almost all germination conditions was slightly to highly correlated with t10−1, t50−1 and U7525−1 (Rp = 0.49–0.76; PBH = 0.00). In case of AUC, significant correlations were also observed between these traits (up to Rp = 0.87; PBH = 0.00). Significant positive correlations were also observed between the Gmax and AUC under different germination conditions. Furthermore, there was a strong correlation between the t10−1, t50−1, MGR and U7525−1 (PBH < 0.0001) (Fig. 3 and Supporting Information Fig. S1, Supporting Information Table S1). This is most obvious between t10−1 and t50−1. Examples of t10−1-t50−1 correlations include control t10−1 and t50−1 (Rp = +0.95; PBH = 0.00), salt (−0.3 MPa, −0.5 MPa), t10−1 and t50−1 (Rp = +0.95; P = 0.00; Rp = +0.97; PBH = 0.00 respectively) and between t10−1 and t50−1 at high-temperature stress (35 °C) (Rp = +0.97; PBH = 0.00). The trend was similar while comparing MGR with Gmax, t10−1, t50−1 and AUC; a number of low to high significant correlations were observed for seed phenotypes under different germination conditions (Fig. 3 and Supporting Information Fig. S1, Supporting Information Table S1).

Figure 3.

Heatmap of correlations between seed quality phenotypes. Each square represents the Pearson correlation coefficient between the seed phenotypes of the column with that of the row. Seed phenotype order is determined as in hierarchical clustering using the distance function 1-correlation. The dissimilarity index is employed for cluster analysis to arrange different seed phenotypes according to their similarity (Legendre & Legendre 1998). Self-self correlations are identified in black. Individual correlation coefficients can be found in Supporting Information Table S1. Supporting Information Fig. S1 displays the correlation heatmap organized in logical order for calculated seed traits, for example, Gmax, t10−1, t50−1, MGR, U7525−1 and AUC. AUC, area under the germination curve; MGR, mean germination rate.

Epistasis

The results of genome-wide epistasis analysis for each of the seed quality phenotypes are presented in Table 4. These analyses tested all pairwise combinations of the markers closest to each target QTL. The analysis of this interaction among seed quality QTL revealed several instances where epistatic interactions among QTLs may obscure relationships between loci and phenotypes. These epistatic interactions contribute to phenotypic variability, but hinder detection and affect estimation of QTLs examined singly. A survey of epistasis with the R\qtl module detected reasonable instances of epistasis in our experiments, whereby only pairwise interactions involving two loci were tested. This analysis revealed novel loci on several chromosomes interacting to influence seed quality traits.

Table 4.  Interaction LOD scores for phenotypes significant at the genome-wide level (P < 0.05)
PhenotypeChr APosition (cM)Chr BPosition (cM)Lod.fullaLod.fv1bLod.intc
  • Two-way epistatic interactions for Solanum lycopersicum/Solanum pimpinellifolium recombinant inbred line population across all 12 chromosomes. AUC, area under the germination curve; LOD, logarithm-of-odds.

  • a

    Lod.full is the LOD score of the full model with two loci and their interaction compared with the null model with no quantitative trait locus (QTL).

  • b

    Lod.fv1 is the LOD score of the full model compared with the best single QTL model with one locus on either chromosome A or B.

  • c

    Lod.int is the LOD score of the interaction term which is found by comparing the full model with an interaction term to the two QTL models with no interaction term.

Control U7525−14855157.626.004.56
Salt I (−0.3) U7525−141075210.417.435.00
Salt II (−0.5) t10−12254659.486.624.00
Osmotic I (−0.3) t10−122242511.976.624.55
Cold stress (12 °C) AUC35511158.984.704.91

The analysis revealed a locus on chromosomes 4 and 5 interacting to influence U7525−1 under control conditions (Table 4, Fig. 4). Similarly, for salt (−0.3 MPa), strong evidence of interaction was observed for U7525−1 on chromosomes 4 and 7 (LODint = 5.00). This was the highest level of statistical significance obtained in our epistasis screen. A two-way interaction was also revealed for t10−1 on chromosomes 2 and 4 under salt stress conditions (−0.5 MPa), whereas a locus on chromosome 2 also interacts with a locus on chromosome 4 under osmotic stress condition (−0.3 MPa PEG) for the same parameter (Table 4, Fig. 4). An epistatic interaction was also observed for AUC under cold stress (12 °C) between QTLs on chromosomes 3 and 11 (Table 4, Fig. 4).

Figure 4.

Epistatic interaction network. Graphical visualization of the epistatic interactions found between different loci controlling seed quality phenotypes in Solanum lycopersicum and Solanum pimpinellifolium recombinant inbred line population. The 12 chromosomes are represented as different circle segments, and their sizes are proportional to the corresponding genetic sizes measured in centimorgan (cM) units. The colour of the lines indicates the trait for which the epistatic interaction was observed (Arends et al. 2010). AUC, area under the germination curve.

DISCUSSION

This study makes clear that the genetic control of seed quality is complex. We have detected numerous QTLs with moderate to large phenotypic effects that influence tomato seed quality attributes consistently across all studied traits. Contributions to seed quality from both tomato parental genotypes produced transgressive segregation for some traits. We also found significant evidence for pairwise epistatic interactions. Differences in QTL detection among phenotypic traits added new dimensions to the complexity of seed quality. The recognition and assessment of sources of variation of seed quality is essential for developing a realistic understanding of how tomato seed phenotypes interact across different conditions, with the ultimate goal of obtaining durable seed quality in tomato crop plants.

The S. lycopersicum × S. pimpinellifolium RIL population and QTL locations

The power of detecting QTLs depends on several factors, including heritability (h2) of the trait, gene action, the type of mapping population, the number and individual effects of QTLs, marker coverage, and the distance between marker loci and QTL(s) affecting the trait (Mackay 2001; Foolad et al. 2003b; Mackay, Stone & Ayroles 2009). The overall heritability of traits (i.e. heritability in the broad sense) strongly affects the quality of QTL analysis, including the number of QTLs detected and the accuracy of their map positions and effect estimates (Alonso-Blanco & Koornneef 2000). However, heritability in the broad sense can be controlled by several factors, which are experimentally manipulable when scoring the traits (Kobayashi & Koyama 2002). We have utilized homogenous and strictly controlled plant growing and seed phenotype testing conditions and this has contributed to increasing the broad-sense heritability of the seed quality traits in both control and stressed conditions (hb2 > 0.53–0.94; Table 2).

Interpretation of seed germination traits

Several methods and mathematical expressions to measure the germination process have been proposed over the past two decades (Hilhorst & Karssen 1988; Bradford 1990; Bewley & Black 1994). One of the most significant current discussions in seed science concerns the measurement of time, rate, homogeneity and synchrony of germination, as they can provide information about the dynamics of the seed germination process. These characteristics are important for physiologists and seed technologists as it is at the heart of their understanding of germination potential of seedlots. This study is an effort of indexing different aspects of cumulative germination in order to quantify the different seed quality traits under different germination conditions. The final germination of seeds is one of the qualitative attributes of the germination process; it portrays the overall germination potential of crop species based on a binary answer: germinated or non-germinated. There is consensus as to the meaning, methods and calculation of germinability in time or at the end of the observations (Ranal & Santana 2006). Although final germination is an important factor for estimating the expected seedling yield of a seedlot, it can be partly independent of other germination characteristics like rate of germination. The germination characteristics of a seedlot are determined by the species, genetic diversity, as well as germination conditions and seed pretreatments. In fact, it has been shown that germination parameters are under strong genetic control (El-Kassaby 1991) and therefore, analysing different aspects of cumulative germination curves, like the onset of germination and germination rate as important phenotypic attributes of a seedlot, is of unprecedented importance with respect to the consequences of genetic diversity present in the S. lycopersicum × S. pimpinellifolium RIL population. However, it has been emphasized that the onset of germination and germination rate (t10−1 and t50−1, respectively) is useful for comparisons only when samples have a sufficient level of final germination (Goodchild & Walker 1971), and to address this issue, we only measured these parameters for those traits that show at least 10 and 50% germination, respectively, in more than 80% of the RILs. There is a large volume of published studies describing genetic characterization of onset and rate of seed germination (t10−1, t50−1, MGR) and exploitation of the natural variation using different mapping populations, for example, RILs, ILs, etc., for germination rate phenotypes (Quesada et al. 2002; Foolad et al. 2003a,b, 2007; Clerkx et al. 2004; Langridge et al. 2006; Landjeva et al. 2010). In this study, we performed QTL analysis with all these different germination parameters and we found genomic regions where QTLs for different rate measurements were mapped to the same approximate location, indicating that common factors are associated with the rate measurements to different germination conditions. Strong correlations were also evident among the different rate measurements, and Pearson correlation analysis among all rate estimates indicated high correlations among t10−1, t50−1 and MGR (P < 0.0001).

Despite the agronomic importance of the rate and uniformity of germination, these traits have not been specifically targeted by breeders. Longer germination times for tomato seeds have been associated with a greater likelihood of producing an abnormal seedling. In terms of seed vigour, the rate and uniformity of germination is a sensitive indicator of a high-quality seed, and these attributes deteriorate more quickly than final germination and are therefore a key component to seed quality. To simplify quantification of germination responses, both the rate and the percentage of germination were incorporated into AUC. Thus, simultaneous germination responses can be interpreted by the AUC as increases in germination rate and final germination percentage, as well as an earlier onset and uniformity of germination. Seedlots that germinate rapidly and fully will have high AUC values, while those that germinate slowly and lowly will have low values. The analysis of germination can be enriched if, in addition to the final germination, t10−1, t50−1, MGR, U7525−1 and AUC values are communicated, because they measure different aspects of the germination process. T10−1 is predominantly a measure of the onset of germination (lag time) whereas t50−1 and MGR are measures for the germination rate, U7525−1 for uniformity and AUC as the combinatorial parameter. This study demonstrates that the usefulness of these germination parameters for describing the extremes of pattern differences of seed germination and all these germination measurements can be applied to evaluate seed germination.

QTL overlapping among seed quality phenotypes

Because seed quality is attributable to an overall tolerance to various seed stresses, we expected, and found evidence for, the co-location of QTLs for control and all stress conditions. A number of significant occurrences of overlapping QTLs among Gmax, t10−1, t50−1, MGR, U7525−1 and AUC were observed among most of the detected QTL positions across different germination conditions. For instance, on chromosomes 4, 6 and 11 the confidence intervals of Gmax and AUC QTLs overlapped with those detected for t10−1, t50−1 and MGR across different stress conditions (Fig. 2). Another instance of significant co-locations of QTLs was identified for these seed quality traits on linkage groups 1, 2, 9 and 12 (Fig. 2). Such co-locations indicate that the shared QTL clusters may bear pleiotropic effects. The co-locations of QTLs identified for seed quality traits in the present study indicated a variable number of overlapping QTL clusters among them. The co-location of roughly two-thirds of the QTLs affecting the t10−1, t50−1 and MGR across different stresses highlights the positive relationship between seed quality phenotypes and different stress types. The present results indeed corroborate previous QTL mapping studies of germination under salt, drought and cold stresses in tomato where 71% of the detected QTLs affected germination under two stresses or more (Foolad et al. 2007). Although QTLs for the seed quality parameters (Gmax, t10−1, t50−1, MGR, U7525−1 and AUC) in each germination condition often co-located as partly may be explained by the fact that they are all descriptors for the same germination curves, interestingly however in several instances, germination parameters mapped to unique regions, for example, QTLs for Gmax on chromosome 5 at 12 °C, t10−1 under oxidative stress on chromosome 10 and QTL for U7525−1 under control condition on chromosome 3 (Fig. 2). Furthermore, inspection of the QTLs affecting individual parameters across different chromosomes also revealed striking significant hot spots for one parameter but not for others. Examples include on chromosome 5 we had QTLs for Gmax, but not for t10−1, t50−1 or MGR, whereas on chromosome 7 we had co-location for t50−1 and MGR, but no revelation of any QTL for Gmax. Furthermore, overlapping QTLs were found on chromosome 9 for t10−1, t50−1 and AUC, but not for other measured traits. Similarly, on chromosome 12 we had QTL overlaps for t10−1, t50−1, MGR and AUC traits but not for Gmax (Fig. 2). Apparently there are specific loci that affect some germination characteristics and not the others. It is also interesting to note that besides QTLs at the same loci for all salt and osmotic levels, in some instances additional QTLs under certain concentration were revealed (chromosomes 2, 4, 5, 6, 9, 11 and 12). As an example, Gmax QTL on chromosome 5 was detected in both salt stress levels whereas a QTL on chromosome 4 was only detected at −0.5 MPa salt. The magnitude of different stresses is variable in soil and stress tolerance to environmental stresses depends on the stage, length and severity of the stress (Bray 2002). These results indicate that seeds respond to one or more stresses through physiological mechanisms depending on the nature and magnitude of the stress (Capiati, Pais & Tellez-Inon 2006). Similarly, while comparing QTLs for salt and osmotic stress conditions, we found QTLs co-locating for some seed germination parameters for both salt and osmotic stresses, but we were also able to identify novel loci (Fig. 2). These findings further support the idea that the regulation of germination under salt and osmotic stresses involves the action of common as well as independent loci, revealing the existence of loci specifically associated with the toxic component of salt and not just its osmotic effect (Vallejo, Yanovsky & Botto 2010). Furthermore, identification of QTLs for non-stress condition indicates the genetic relationships between germination phenotypes under stress and non-stress conditions, and it has been suggested that germination of tomato is genetically controlled and hence can be increased by selection (Foolad, Lin & Chen 1999). QTLs corresponding to different seed parameters in our study have shown overlaps, and correlations among germination-derived parameters were also high. Thus, establishing the correspondence between QTL co-locations and correlations between phenotypic characters appears possible. Considering together the traits studied herein, significant correlations were observed: up to 0.76 between Gmax and AUC, and up to 0.95 between Gmax and t10−1, t50−1, MGR and U7525−1, and likewise up to 0.87 between AUC and aforementioned parameters. The QTL analysis indicated the presence of genetic relationships between germination under different conditions. These observations suggest that the QTLs detected for Gmax, t10−1, t50−1, MGR, U7525−1 and AUC in tomato seed are overlapping on the same linkage groups and could be related to significant correlations among these traits. Previous quantitative trait genetic studies have reported similar co-locations (Foolad & Chen 1999; Clerkx et al. 2004) and suggest that trait correlations may be attributable to either pleiotropic effects of single genes or to tight linkage of several genes that individually influence specific traits (Pelgas et al. 2011). It should not be too difficult to disentangle these two effects in the near future.

Physiological mechanism of seed quality phenotypes under different conditions

Productive and sustainable crop growth necessitates growing plants in suboptimal environments with less input of precious resources. This study was intended to make a step forward towards better understanding and rapid improvement of abiotic stress tolerance in tomato, and to link physiological and underlying molecular mechanisms of seed quality. Excessive salt lowers the rate of, or completely inhibits, seed germination (Foolad et al. 2003a, 2007). This may be accomplished by lowering the osmotic potential of the germination medium, but a saline germination medium could also lower the rate of seed germination by specific salt stress. However, accumulating evidence suggests that the low water potential of the external medium, rather than ion toxicity effects, is the major limiting factor to germination under salt stress in different crop species, including tomato (Ni & Bradford 1992; Bradford 1995; Foolad et al. 2007). Another possible explanation for some of our results may be the release of reactive oxygen species (ROS) in all of these stress types (Clerkx et al. 2004; Wahid et al. 2007; Collins et al. 2008). Saline conditions are known to generate ROS (Zhu 2002). Prior studies have noted that lowered rates of seed germination under drought stress are due to reduced osmotic potential of the germination medium (Bradford 1995; Hilhorst & Downie 1996) similar to that under salt stress. Therefore, it is expected that seeds that germinate rapidly under salt stress would also germinate rapidly under osmotic stress, and vice versa. This is partly in agreement with the findings of the present study. It is conceivable that similar or identical genes (and physiological mechanisms) control the seed germination process of tomato under salt and drought stresses. This is evident from the correlation between salt and PEG treatments (Fig. 3). There is hardly any information whether genetic and physiological processes that maintain rapid seed germination under salt and/or drought stress are also responsible for rapid seed germination under cold stress. However, low temperature (cold stress) may affect the water status of the cell and, thus, could delay seed germination by causing osmotic stress (Liptay & Schopfer 1983). In the present study, however, the finding that most of QTLs for seed quality traits under cold stress co-localized with QTLs for germination under salt and/or osmotic stress suggests that the same genes (or physiological mechanisms) may contribute to rapid seed germination under these three conditions. This suggestion is consistent with the finding that selection for rapid seed germination under salt or drought stress resulted in progeny with improved germination under cold stress, and vice versa (Foolad & Lin 2000).

In the present study, QTLs were identified affecting germination phenotypes under non-stress (control) and stress conditions (Fig. 2). The QTLs located on chromosomes 4, 6 and 11 affected germination under three or more conditions. Correlation analysis indicated highly significant correlations between the various germination traits at all treatment levels, and this suggests that for response time traits like germination, the earlier traits may be good predictors of crop performance. Genes related to reserve mobilization and endosperm weakening are likely to be involved and these could conceivably affect the rate of germination as metabolic processes and reserves utilized early during germination are different from those required later during the process, but before its completion (Fait et al. 2006; Bethke et al. 2007; Hayashi et al. 2008), and indeed, presence of QTLs for different germination phenotypes, in particular t10−1, t50−1 and MGR, on different chromosomes of the tomato genome possibly corresponds to metabolic or physiological processes that are themselves occurring during different stages of the germination process. A number of QTLs associated with time to 50% of germination (t50) were mapped in tomato (Foolad et al. 1999, 2003b), Arabidopsis (Quesada et al. 2002), and 1 QTL was also mapped in sunflower (Al-Chaarani et al. 2005).

This study clearly illustrates the complexity underlying the genetic basis for seed germination. Identifying QTLs associated with the different parameters of seed germination facilitates elucidation of molecular mechanisms controlling seed germination. As suggested by transgressive trait distributions within the RILs, both parental genotypes S. lycopersicum cv. Moneymaker and S. pimpinellifolium contributed to increased trait means for different germination parameters under control (non-stress) and stressed conditions. This phenomenon has frequently been described for other traits in many crops (Devicente & Tanksley 1993; Foolad 1996), including tomato. The presence of favourable alleles in both parents suggests a strong likelihood for recovering transgressive variants among segregating progeny (Devicente et al. 1993). Given the result that alleles serving to enhance the ability to complete germination under environmental stress are present in both cultivars, improvement of germination traits must be conducted at an individual QTL level (Hayashi et al. 2008).

Detection of QTLs generic to germination traits under control and stressed conditions suggests the presence of genetic relationships between the ability to germinate rapidly under different conditions and the prediction that selection and improvement of seed germination under one condition would lead to progeny with improved germination under other conditions. There was evidence of greater germination variances in the current study under stress conditions, which is partly due to slower germination and, thus, longer time intervals between germination events. Under stress conditions, germination variances increased in the RIL population, and broad-sense heritabilities were larger under stress than non-stress conditions, suggesting the contribution of some genetic factors to the larger variance under the stress treatment. Greater genetic variance in stress environments is rather uncommon, but is one of the more favourable situations for plant breeders (Rosielle & Hamblin 1981). Furthermore, seed germination under different stress conditions was genetically controlled with additivity being the major genetic component. Significantly large genetic correlations between germination responses at different stress levels indicate that similar or identical genes contributed to the germination response under different stress conditions. Thus, selection for rapid germination at one stress level would result in progeny with improved germination at diverse stress levels. Nonetheless, the co-location of QTLs for different seed germination traits supports the genetic dissection of seed quality in order to facilitate a more strategic approach to breed for better seed quality in tomato. Those regions identified across different germination environments are candidates that can be used in marker-assisted selection (MAS) or gene cloning, especially those with moderate to high broad-sense heritabilities (Dudley 1993; Tanksley 1993). However, isolation, characterization and comparison of functional genes, which facilitate rapid seed germination under the various conditions, are necessary in order to determine the exact genetic relationships among these traits.

Identification of epistasis

We have performed a genome-wide epistasis screen in the S. lycopersicum × S. pimpinellifolium cross for seed quality phenotypes and obtained evidence for multiple significant QTL pairs. The identification of significant epistasis controlling seed quality phenotypes both benefits and complicates this analysis. Epistasis may identify genes that function together in distinct genetic networks, potentially providing a valuable insight into function. Our identification of higher-order epistatic networks that control quantitative seed quality phenotypes in S. lycopersicum suggests that these QTLs may be caused by polymorphism in genes that function in a coordinated network. These findings exemplify an advantage of interaction analyses in plant models for complex phenotypes such as seed quality, because by the use of R/qtl analysis we had more than sufficient statistical power to detect two-way epistatic interactions, implicating genomic regions that would otherwise likely have been passed over (Buescher et al. 2010).

Identification of epistatic pairs of loci contribution to seed quality variation in tomato represents a step forward in the delineation of the genetic architecture of these phenotypes in tomato and provides a powerful approach to identify novel gene candidates and chromosomal regions for further pursuit in seed quality studies. Our results, however, also illustrate the degree of complexity of the genetic architecture of these phenotypes. Strong epistasis in the genetic network controlling germination under salt stress was revealed in an Arabidopsis Sha × Col RIL population (Galpaz & Reymond 2010). Validation of this epistatic network hypothesis will require cloning of the full complement of interacting QTLs. Accounting for these seed quality QTL interactions is not only essential for developing strategies to clone seed quality QTLs, but may also allow the useful inclusion of metabolomics and transcriptomics data in the formulation of hypotheses regarding mechanisms of seed quality of the tomato.

In conclusion, this study has identified numerous QTLs contributing to variation in seed quality trait interactions between the tomato accessions S. lycopersicum and S. pimpinellifolium. The QTL approach appears to be valuable not only in elucidating the genetics, but also the physiological background of the seed quality phenotypes. Both stress-specific and non-specific QTLs control the germination process under different conditions in the tomato. This approach offers a way in which simultaneous improvement of these traits and progress towards identifying the underlying genetic mechanisms may be realized. Genome-scale prediction of a large-effect DNA sequence and transcript accumulation polymorphisms differentiating S. lycopersicum and S. pimpinellifolium permit an informed approach to selection and investigation of gene candidates in identified QTL regions (Joosen et al. 2009). The present study is a significant effort in this direction. Robust QTL mapping with SNP-based linkage maps resulted in a much-improved estimation of the genetic architecture of a tomato genome in terms of the magnitude of QTL effects, QTL-environment interactions and putative pleiotropy. Identification of causal polymorphisms for QTLs influencing a majority of S. lycopersicum and S. pimpinellifolium phenotypes will provide potential breeding targets for enhanced seed quality in tomato. Furthermore, fine mapping, validation and further investigation of seed quality-specific QTLs will provide valuable insight into pleiotropic variation as suggested by the co-location of the QTLs

ACKNOWLEDGMENTS

This work was supported by the Technology Foundation STW (R.K., L.W., W.L.) and by the Higher Education Commission, Pakistan (N.K.).

Ancillary