Research Article

fw2.2: A Quantitative Trait Locus Key to the Evolution of Tomato Fruit Size

See allHide authors and affiliations

Science  07 Jul 2000:
Vol. 289, Issue 5476, pp. 85-88
DOI: 10.1126/science.289.5476.85


Domestication of many plants has correlated with dramatic increases in fruit size. In tomato, one quantitative trait locus (QTL), fw2.2, was responsible for a large step in this process. When transformed into large-fruited cultivars, a cosmid derived from the fw2.2 region of a small-fruited wild species reduced fruit size by the predicted amount and had the gene action expected for fw2.2. The cause of the QTL effect is a single gene, ORFX, that is expressed early in floral development, controls carpel cell number, and has a sequence suggesting structural similarity to the human oncogene c-H-ras p21. Alterations in fruit size, imparted byfw2.2 alleles, are most likely due to changes in regulation rather than in the sequence and structure of the encoded protein.

In natural populations, most phenotypic variation is continuous and is effected by alleles at multiple loci. Although this quantitative variation fuels evolutionary change and has been exploited in the domestication and genetic improvement of plants and animals, the identification and isolation of the genes underlying this variation have been difficult.

Conspicuous and important quantitative traits in plant agriculture are associated with domestication (1). Dramatic, relatively rapid evolution of fruit size has accompanied the domestication of virtually all fruit-bearing crop species (2). For example, the progenitor of the domesticated tomato (Lycopersicon esculentum) most likely had fruit less than 1 cm in diameter and only a few grams in weight (3). Such fruit was large enough to contain hundreds of seeds and yet small enough to be dispersed by small rodents or birds. In contrast, modern tomatoes can weigh as much as 1000 grams and can exceed 15 cm in diameter (Fig. 1A). Tomato fruit size is quantitatively controlled [for example, (4)]; however, the molecular basis of this transition has been unknown.

Figure 1

(A) Fruit size extremes in the genusLycopersicon. On the left is a fruit from the wild tomato species L. pimpinellifolium, which like all other wild tomato species, bears very small fruit. On the right is a fruit fromL. esculentum cv Giant Red, bred to produce extremely large tomatoes. (B) Phenotypic effect of thefw2.2 transgene in the cultivar Mogeor. Fruit are from R1 progeny of fw107 segregating for the presence (+) or absence (−) of cos50 containing the small-fruit allele.

Most of the loci involved in the evolution and domestication of tomato from small berries to large fruit have been genetically mapped (5, 6). One of these QTLs,fw2.2, changes fruit weight by up to 30% and appears to have been responsible for a key transition during domestication: All wild Lycopersicon species examined thus far contain small-fruit alleles at this locus, whereas modern cultivars have large-fruit alleles (7). By applying a map-based approach, we have cloned and sequenced a 19-kb segment of DNA containing this QTL and have identified the gene responsible for the QTL effect.

Genetic complementation withfw2.2. A yeast artificial chromosome (YAC) containing fw2.2 was isolated (8) and used to screen a cDNA library (constructed from the small-fruited genotype, L. pennellii LA716). About 100 positive cDNA clones were identified that represent four unique transcripts (cDNA27, cDNA38, cDNA44, and cDNA70) that were derived from genes in the fw2.2 YAC contig. A high-resolution map was created of the four transcripts on 3472 F2 individuals derived from a cross between two nearly isogenic lines (NILs) differing for alleles at fw2.2(Fig. 2A) (8). The four cDNAs were then used to screen a cosmid library of L. pennellii genomic DNA (9). Four positive, nonoverlapping cosmids (cos50, cos62, cos69, and cos84) were identified, one corresponding to each unique transcript. These four cosmid clones were assembled into a physical contig of thefw2.2 region (10) (Fig. 2B) and were used for genetic complementation analysis in transgenic plants.

Figure 2

High-resolution mapping of thefw2.2 QTL. (A) The location offw2.2 on tomato chromosome 2 in a cross betweenL. esculentum and a NIL containing a small introgression (gray area) from L. pennellii[from (8)]. (B) Contig of thefw2.2 candidate region, delimited by recombination events at XO31 and XO33 [from (8)]. Arrows represent the four original candidate cDNAs (70, 27, 38, and 44), and heavy horizontal bars are the four cosmids (cos62, 84, 69, and 50) isolated with these cDNAs as probes. The vertical lines are positions of restriction fragment length polymorphism or cleaved amplified polymorphism (CAPs) markers. (C) Sequence analysis of cos50, including the positions of cDNA44, ORFX, the A-T–rich repeat region, and the “rightmost” recombination event, XO33.

The constructs (11) were transformed into two tomato cultivars, Mogeor (fresh market–type) and TA496 (processing-type) (12). Both tomato lines carry the partially recessive large-fruit allele offw2.2. Because fw2.2 is a QTL and the L. pennellii allele is only partially dominant, the primary transformants (R0), which are hemizygous for the transgene, were self-pollinated to obtain segregating R1 progeny. In plants containing the transgene (13), a statistically significant reduction in fruit weight indicated that the plants were carrying the small-fruit allele of fw2.2 and that complementation had been achieved. This result was only observed in the R1 progeny of primary transformants fw71 and fw107, both of which carried cos50 (Fig. 1B and Table 1) (14). That the two complementing transformation events are independent and in different tomato lines (TA496 and Mogeor) indicates that the cos50 transgene functions similarly in different genetic backgrounds and genomic locations. Thus, the progeny of plants fw71 and fw107 show that fw2.2 is contained within cos50.

Table 1

Average fruit weights and seed numbers (23) for R1 progeny of several primary transformants. Unless otherwise noted, progeny are from independent R0 plants. Numbers in parentheses are the numbers of R1 individuals tested.

View this table:

Most QTL alleles are not fully dominant or recessive (5). The small-fruit L. pennellii allele forfw2.2 is semidominant to the large-fruit L. esculentum allele (7). R2 progeny of fw71 were used to calculate the gene action [d/a = dominance deviation/additivity; calculated as described in (5)] of cos50 in the transgenic plants. The transgene had a d/a of 0.51; in previous work with nearly isogenic lines (NILs),fw2.2 had a d/a of 0.44. This similarity of gene action is consistent with the conclusion that the cos50 transgene carries fw2.2.

fw2.2 corresponds to ORFX and is expressed in pre-anthesis floral organs. Sequence analysis of cos50 (15) revealed two open reading frames (ORFs) (Fig. 2C): one corresponding to cDNA44, which was used to isolate cos50, and another 663-nucleotide (nt) gene, ORFX, for which no corresponding transcript was detected in the initial cDNA library screen. The insert also contains a highly repetitive, AT-rich (80%) region of 1.4 kb (Fig. 2C). Previous mapping of fw2.2had identified a single recombination event that delimited the “rightmost” end of the fw2.2 candidate region [XO33 in (8)]. Comparison of genomic DNA sequence from this recombinant plant with that of the two parental lines indicated that XO33 is within 43 to 80 nt 5′ from the end of ORFX(Fig. 2C). Because genetic mutation(s) causing change in fruit size must be to the left of XO33, cDNA44 cannot be involved, andORFX or an upstream region is the likely cause of thefw2.2 QTL phenotype.

ORFX is transcribed at levels too low to be detected through standard Northern hybridization protocols in all pre-anthesis floral organs (petal, carpels, sepals, and stamen) of both large- and small-fruited NILs; however, semi-quantitative reverse transcriptase–polymerase chain reaction (RT-PCR) analysis indicated that the highest levels were expressed in carpels (16) (Fig. 3A). In addition, comparison of the relative levels of ORFXtranscript in the carpels of the NILs showed significantly higher levels in the small-fruited NIL (TA1144) than in the large-fruited NIL (TA1143) (TA1143/TA1144 carpel transcript levels, mean ratio = 0.51; for the null hypothesis mean = 1, P = 0.02). The observation ofORFX transcription in pre-anthesis carpels suggests thatfw2.2 exerts its effect early in development. To test this hypothesis, we compared the floral organs from the small- and large-fruited NILs. Carpels (which ultimately develop into fruit), styles, and sepals of the large-fruited NIL were already significantly heavier at anthesis (P = 0.0007, 0.001, and 0.001, respectively) than their counterparts in the small-fruited NIL. Stamen and petals showed no significant difference (P = 0.63 and 0.74, respectively). Cell sizes at anthesis were similar (P = 0.98 and P = 0.85) in the NILs (Fig. 3, B to E); hence, carpels of large-fruited genotypes contain more cells. Therefore, we conclude that allelic variation atORFX modulates fruit size at least in part by controlling carpel cell number before anthesis.

Figure 3

Reverse transcriptase and histological analyses of the large- and small-fruited NILs (TA1143 and TA1144, respectively). (A) RT-PCR detection ofORFX transcript in floral organs. Gel showing RT-PCR products for ORFX in various stages and organs. Stage I, 3- to 5-mm floral buds; Stage II, 5 mm to anthesis; Stage III, anthesis; lane 1, sepals; lane 2, petals; lane 3, stamen; lane 4, carpels; L, leaves. (B to E) Transverse thick sections (1 μM) of tomato carpels at anthesis. Top sections (B and C) display cortical cells from carpel septum. Bottom sections (D and E) display pericarp cells from carpel walls. Sections on the left (B and D) are derived from carpels of NIL homozygous for large-fruit allele. Sections on the right (C and E) are derived from carpels of NIL homozygous for small-fruit allele. TA1143 and TA1144 were not significantly different for cell size in either carpel walls (cells per millimeter squared = 17,600 ± 700 versus 17,700 ± 1000; P = 0.98) or carpel septa (cells per millimeter squared = 10,100 ± 500 versus 10,300 ± 900; P = 0.85) (statistical analysis based on 144 cell area counts from 48 sections). Carpels were fixed in 2.5% glutaraldehyde, 2% paraformaldehyde, and 0.1 M Na cacodylate buffer (pH 6.8) and embedded in Spurr plastic. Bar, 20 μM.

ORFX has homologs in other plant species and predicted structural similarity to human oncogene RAS protein. Sequence analysis ofORFX (17) revealed that it contains two introns and encodes a 163–amino acid polypeptide of ∼22 kD (Fig. 4). Comparison of the predicted amino acid sequence of the ORFX cDNA against sequences in the GenBank expressed sequence tag (EST) database found matches only with plant genes. Matches (up to 70% similarity) were found with ESTs in both monocotyledonous and dicotyledonous species. In addition, a weaker match (56.7% similarity) was found with a gymnosperm, Pinus(Fig. 4). In tomato, at least four additional paralogs ofORFX were identified in the EST database. Although only oneArabidopsis EST is represented in the database, eight additional homologs of ORFX appear in Arabidopsisgenomic sequence, often in two or three-gene clusters and having intron-exon arrangements similar to those of ORFX. None of the putative homologs of ORFX has a known function. Thus,ORFX appears to represent a previously uncharacterized plant-specific multigene family.

Figure 4

The results of CLUSTALW alignment of LpORFX (L. pennellii, AF261775) and LeORFX (L. esculentum, AF261774) with 7 representatives of 26 matched from the GenBank EST and nucleotide databases and the contigs assembled from the TIGR (The Institute for Genomic Research) tomato EST database (24). LpORFX and LeORFX residues are shaded black when identical to at least 73% of all the genes included in the analysis. Shading in the other genes represents residues identical (black) or similar (gray) to the black residues in LpORFX, and the dashes are gaps introduced to optimize alignment. Percentages of identical (%ID) or similar (%SIM) amino acid residues over the length of the available sequence are noted (some ESTs may be only partial transcripts). ESTs included in the list are Ph (Petunia hybrida, AF049928), Gm (Glycine max, AI960277), Os (Oryza sativa, AU068795), Zm (Zea mays, AI947908), and Pt (Pinus taeda, AI725028). The L. esculentum EST is contig TC3457 from the TIGR EST database. At represents a predicted protein fromArabidopsis genomic sequence (AB015477.1). The positions of the introns in ORFX are indicated as I1 and I2, and the three residue differences between LpORFX and LeORFX are denoted by asterisks. Abbreviations for the amino acid residues are as follows: A, Ala; C, Cys; D, Asp; E, Glu; F, Phe; G, Gly; H, His; I, Ile; K, Lys; L, Leu; M, Met; N, Asn; P, Pro; Q, Gln; R, Arg; S, Ser; T, Thr; V, Val; W, Trp; and Y, Tyr.

Analysis of the predicted amino acid sequence of ORFX indicates that it is a soluble protein with alpha/beta-type secondary structure. The threading program LOOPP (18) assigns ORFX to the fold of 6q21, domain A, which is human oncogene RAS protein. TheZ scores for global and local alignments of ORFX are high (3.2 and 4, respectively). Such scores were never observed in false positives and suggest an overall shape similar to that of heterotrimeric guanosine triphosphate–binding proteins. The detailed comparison of ORFX sequence with that of the RAX (where X can be S, N, or D) family reveals conserved fingerprints at RAX-binding domains (19). The RAX family includes proteins with wide regulatory functions, including control of cell division (20).

The basis for allelic differences at fw2.2. In order to understand the basis for allelic differences atfw2.2, we compared the L. pennellii andL. esculentum ORFX alleles by amplifying and sequencing an 830-nt fragment containing ORFX [including 55 nt from the 3′ untranslated region (UTR) and 95 nt from the 5′ UTR] from both NILs (Fig. 4). Of the 42 nt differences between the two alleles, 35 fell within the two predicted introns, 4 represent silent mutations, and only 3 cause amino acid changes. All three of the substitutions occurred within the first nine residues of the ORF (asterisks in Fig. 4). Although the start methionine cannot be determined with certainty, if the second methionine in the ORF (M12 inFig. 4) were used, this would place all three potential substitutions in the 5′ UTR. Conservation between the alleles suggests that thefw2.2 phenotype is probably not caused by differences within the coding region of ORFX, but by one or more changes upstream in the promoter region of ORFX. Variation in upstream regulatory regions of the teosinte branched1 gene has also been implicated in the domestication of maize (21). However, differences in fruit size imparted by the differentfw2.2 alleles may be modulated by a combination of sequence changes in the coding and upstream regions of ORFX(22).

A reduction in cell division in carpels of the small-fruited NIL is correlated with overall higher levels of ORFX transcript, suggesting that ORFX may be a negative regulator of cell division. Whether the ORFX and RAX proteins share common properties other than predicted three-dimensional structure and control of cell division awaits future experimentation. An affirmative result may reflect an ancient and common origin in the processes of cell cycle regulation in plants and animals.

  • * These authors contributed equally to this work.

  • Present address: Department of Biological Sciences, Clapp Laboratory, Mount Holyoke College, South Hadley, MA 01075, USA.

  • Present address: Research Institute for Vegetable and Ornamental Plant Breeding, IMOF-CNR, Via Universita 133, 80055 Portici, Italy.


View Abstract

Stay Connected to Science

Navigate This Article