Report

Zooming In on a Quantitative Trait for Tomato Yield Using Interspecific Introgressions

See allHide authors and affiliations

Science  17 Sep 2004:
Vol. 305, Issue 5691, pp. 1786-1789
DOI: 10.1126/science.1101666

Abstract

To explore natural biodiversity we developed and examined introgression lines (ILs) containing chromosome segments of wild species (Solanum pennellii) in the background of the cultivated tomato (S. lycopersicum). We identified Brix9-2-5, which is a S. pennellii quantitative trait locus (QTL) that increases sugar yield of tomatoes and was mapped within a flower- and fruit-specific invertase (LIN5). QTL analysis representing five different tomato species delimited the functional polymorphism of Brix9-2-5 to an amino acid near the catalytic site of the invertase crystal, affecting enzyme kinetics and fruit sink strength. These results underline the power of diverse ILs for high-resolution perspectives on complex phenotypes.

The genetic basis of many natural phenotypes takes the form of a continuous range rather than discrete classes. The complexity of traits showing continuous distribution often results from the segregation of numerous QTL, whose expression is modified both by the environment and by genetic background (1). Genetic resolution of quantitative traits in populations that segregate simultaneously for different QTL scattered throughout the genome [e.g., second filial generation (F2), backcross, and recombinant inbreds] is low compared with QTL analysis in lines that segregate for a single region (i.e., ILs in plants and congenic strains in animals) (2). Multiple segregating QTL at the whole-genome level tend to mask the effects of one another by introducing high variances in statistical analyses. In sharp contrast, ILs are identical for the entire genome except for a single introgressed region, and therefore all the phenotypic variation in these lines is associated with the introduced segment. The use of such targeted population structures increased the identification power for QTL by several times in both plants and animals (3, 4).

QTL cloning projects in plants have focused primarily on variation derived from naturally divergent genetic resources and have generated new biological insights about plant development, whereby some of the effects could be delimited to single-nucleotide alterations (58). Biodiversity research in crop plants has shown that when chromosome segments of wild species are introgressed into cultivated varieties, it is possible to identify genomic regions that substantially improve yield (9). To enhance the progress of tomato breeding, we developed a population of 76 segmental ILs that are composed of marker-defined genomic regions of the wild species S. pennellii (accession code LA716), substituting for the homologous intervals of the cultivated variety M82 [S. lycopersicum; the taxonomic classification of tomato in the genus Solanum is available in (10)]. The ILs that represent the entire genome partition the genetic map into 107 bins, which are defined by singular or overlapping segments (11). Over the past 10 years, the tomato ILs have been assayed for yield-associated traits and the data are presented, in silico, in a search engine that displays the components of the genetic variation (12).

The QTL database creates a firm basis to investigate the genetic control of yield-associated traits. An important yield component in “ketchup tomatoes” is the total soluble content of the fruit measured in refractometer brix units (consisting mainly of the sugars glucose and fructose). The S. pennellii IL data for brix (B) show 25 different genomic regions with significant effects on the trait relative to M82; in the majority of these cases, the S. pennellii alleles increased B (fig. S1). To uncover the molecular control of B, we characterized the QTL Brix9-2-5, the effects of which ranged from an increase of 11 to 25%, depending on genetic background and environmental factors (13, 14). This moderate QTL improves B without reducing total yield, thus increasing the sugar output per unit area, which is a key yield parameter in industrial tomatoes (9). High-resolution genetic mapping delimited Brix9-2-5 to a single-nucleotide polymorphism (SNP)–defined region of 484 base pairs spanning part of the third exon and the third intron of the cell-wall (CW) invertase (LIN5) (14). Another component of the complexity in the molecular analysis of this gene is that LIN5 is a member of a larger gene family including three additional CW invertases (LIN6, LIN7, and LIN8). After investigating the RNA profile of the LIN family, we concluded that the window of exclusivity of LIN5 expression is in the conductive tissue of the flower's ovary, fruit placenta, and pericarp during the early stage of cell division (15) (Fig. 1). The expression of LIN5 in the conducting tissue next to a potential “sugar-unloading site” near the ovaries is consistent with the role of LIN5 as a “sink gene” that regulates the ability of the fruit to import photosynthetic sugars (16). Because of the partially dominant nature of the S. pennellii allele in elevating fruit B (14), we compared transcription rates of the wild and cultivated species alleles in young ovaries of heterozygous plants. Steady-state mRNA levels of the S. pennellii allele relative to the S. lycopersicum allele were statistically similar (fig. S2), indicating that the QTL effect was not due to differential transcription modulation by the third intron (14). To compare the quantities of the LIN5 protein, we extracted the CW-bound enzymes from the ovules of the Brix9-2-5 nearly isogenic lines and analyzed them on protein gel blots with antibodies raised against CW invertase (17). The lines did not differ statistically in the mean quantity of LIN5 peptide (fig. S3). However, CW invertase activity was statistically different (P < 0.01) between the genotypes, with values three to five times as high in lines homozygous for the S. pennellii allele compared with those of the S. lycopersicum lines (Table 1).

Fig. 1.

LIN5 expression in the conductive tissues of the developing tomato fruit. In situ localization of LIN5 transcripts in a longitudinal section of an ovary at anthesis showing expression mainly in the conductive tissues within the placenta that lead to the developing seeds as well as the surrounding pericarp.

Table 1.

IL9-2-5 effects on invertase activity in flowers. Invertase activity (mmol reducing sugars per gram fresh weight per hour) was determined in the insoluble fraction of ovaries before anthesis and during anthesis. Mean values were calculated based on five replicates of 10 ovaries each (±TE).

Genotype/stage Cell wall invertase activity
Preanthesis Anthesis
M82 9.5 ± 4.1 20.4 ± 4.1
IL9-2-5 32.0 ± 4.1 63.2 ± 4.1

The fact that the IL9-2-5 differed from the control for enzyme activity in flowers did not mean that these differences map to the same LIN5 interval as the fruit sugar QTL, which is manifested 50 days after flowering. An example for the complexity of the genetic factors that regulate quantitative traits was revealed for Drosophila alleles of alcohol dehydrogenase of differing activity, which is controlled by at least three different zones of the gene (18). To evaluate whether the differences in CW invertase activity were associated with the 484–base pair B QTL interval, we assayed the high-resolution recombinant families that span the gene (Fig. 2). The mapping analysis demonstrates that mature fruit B and CW invertase activity in flower ovaries are associated with the same LIN5 interval. The data presented so far imply that the three amino acid differences in the third exon are responsible for the QTL effects (Fig. 3A).

Fig. 2.

Map of the effect of Brix9-2-5 on B and on CW invertase activity. Each recombinant is shown by a combination of empty and hatched bars, representing the S. lycopersicum and S. pennellii sequence, respectively. Each SNP is represented by its position in the S. pennellii (top) and S. lycopersicum nucleotides. The SNP2878, which underlines the QTN effects, is boxed. Phenotypic effects of each recombinant family's alleles on the B and the CW invertase activity is shown as the percentage difference from the S. lycopersicum line. Full bars and asterisks denote a significant difference from the control (P < 0.001).

Fig. 3.

Alignment of the Brix9-2-5 peptide sequence (28). (A) Tomato species Brix9-2-5 allelic series. The relevant three amino acid substitutions between S. lycopersicum and S. pennellii are in bold. The phenotypic effect (+, present; –, absent) of the wild species' alleles located along the short arm of chromosome 9 on B is indicated on the right and is based on multiple-site testings of advanced backcross populations (1821). (B) Sequence alignment of the protein region spanning Brix9-2-5 in several plants showing a conserved aspartate (D).

Interspecific advanced backcross breeding populations in tomato involving S. pimpinelifollium, S. habrochaites, S. neorickii, and a different accession of S. pennellii (LA1657) from that used for the IL construction were next assayed for yield-associated traits (including B) in three to seven locations. Brix9-2-5 maps to the middle of the short arm of chromosome 9; none of these populations revealed a B QTL on the short arm of chromosome 9, in any of the sites tested (1922) (Fig. 3A). Sequencing of the QTL region of LIN5 from the above accessions showed that they all share the Asp366 and Val373 residues with S. pennellii (LA716). However, Asp348 was uniquely associated with LA716 (Fig. 3A). Interestingly, multiple alignment of LIN5 with its duplicated family members (LIN6, LIN7, and LIN8) and other plant invertases exposed aspartate as a conserved residue at this position (Fig. 3B), suggesting a nonneutral role for this amino acid.

The candidate quantitative trait nucleotide (QTN) (SNP2878; Fig. 2) responsible for the Asp348 substitution was evaluated in complementation tests in an attempt to study its effects on enzyme activity. We used a yeast invertase–deficient strain that completely lacks the ability to degrade sucrose (23) as a host for plasmids harboring the tomato LIN5 cDNA from S. lycopersicum (LIN5lyc-E, where the letter in the superscript depicts the identity of the amino acids at positions 348), IL9-2-5 (LIN5pen-D), and the mutated S. lycopersicum allele engineered to code for an invertase with the Asp348 residue (LIN5lyc-D). Unlike the empty-vector control transformants, the three invertase-transformed yeast strains exhibited substantial growth on sucrose, with LIN5pen-D and LINlyc-D alleles seemingly complementing the invertase deficiency of the yeast strain more faithfully than did the LIN5lyc-E allele (fig. S4). These differences could not be attributed to changes in the quantity of the enzyme because immunoblot quantification did not reveal significant differences in the total amount between the three different yeast genotypes (fig. S5). When sucrose hydrolysis activities of the expressed invertases from the three strains were plotted as a function of substrate concentration, the double reciprocal plots were linear (Fig. 4) and the Michaelis constant for sucrose (Km[sucrose]) values calculated from these plots are 31.0, 11.6, and 5.1 mM for LIN5lyc-E, LIN5pen-D, and LIN5lyc-D alleles, respectively. The values for the S. pennellii and the mutated S. lycopersicum enzymes were similar to those previously reported for plant invertases, but the nonmutated S. lycopersicum was considerably less efficient than the values documented in the BRENDA database (www.brenda.uni-koeln.de). The complementation data indicate that Asp348 plays a role in enhancing the activity of LIN5 and establish that SNP2878 (Fig. 2) is the QTN responsible for this phenotype as well as for fruit sugar yield. The recent resolution of the three-dimensional structure of Thermotoga maritima invertase revealed a bimodular arrangement of the protein with a catalytic active site that has a pocket topology (24). The S. pennellii Asp348 aligns with the amino acid Tyr266 in T. maritima, which is within 10 Å of the catalytic site that is likely associated with sucrose recognition and binding. The results presented in our QTL study appear to be consistent with the crystal structure analysis, thus providing a clue about how a point mutation can alter tomato invertase kinetics and sugar yield in a quantitative way.

Fig. 4.

Biochemical analysis of invertase-deficient yeast complemented by LIN5 alleles. Kinetic analysis of the LIN5 enzymes from yeast strains overexpressing the proteins from S. lycopersicum (LIN5lyc-E), IL9-2-5 (LIN5pen-D), and mutated S. lycopersicum (LIN5lyc-D). Acid invertase activity was measured in soluble fractions of a 3-day yeast culture with different substrate (sucrose) concentrations. Each point represents a mean of four replicates.

The present study relies on the high-resolution QTL mapping attributes of the multispecies tomato IL resource and presents a zoom-in view of a QTN, from the whole-genome perspective to a single functional amino acid in the crystal structure. In Drosophila, which is a model organism for the dissection of complex traits, most of the polymorphisms affecting quantitative variation are in putative regulatory regions that tend to fractionate into multiple linked QTL upon high-resolution analysis (25). This is in contrast to Brix9-2-5 and other plant QTL, which in most cases maintain their distinct “single-factor” identity in fine mapping (5). It is too early to tell whether the difference in the genetic architecture of quantitative traits between Drosophila and plants reflects different biology or relates to the amount of diversity between the parents that were used in the analyses. The cultivated and wild species that parented the ILs are highly divergent in sequence, but more importantly in phenotype, thus providing abundant segregation for naturally selected variation affecting yield, morphological, and biochemical traits. For ILs, chromosome substitution strains (4), and other permanent populations, resource construction must be followed by the much more daunting task of phenotyping of different traits over varied environments and genetic backgrounds (26, 27). Such QTL databases, housing hundreds of phenotypes measured on a common set of ILs, will lead to association between seemingly unrelated traits and allow us to explore higher level organization of complex phenotypes and the role of pleiotropy.

Supporting Online Material

www.sciencemag.org/cgi/content/full/305/5691/1786/DC1

Materials and Methods

Figs. S1 to S5

References

References and Notes

View Abstract

Navigate This Article