INTRODUCTION

TOP ABSTRACT INTRODUCTION RESULTS AND DISCUSSION CONCLUSIONS MATERIALS AND METHODS LITERATURE CITED

Tomato (Lycopersicon esculentum) is one of the most important and intensely studied model dicot plants. Compared with the pioneering research in disease resistance and fruit maturation, little is know about the microsyntenic relationship of tomato and other plant species as revealed by sequencing long contiguous stretches of genomic DNA. To date, the longest tomato sequence reported in GenBank is a 105-kb bacterial artificial chromosome (BAC) sequence from the chromosome 2 ovate region (Ku et al., 2000). In contrast, the Arabidopsis genome has been sequenced (The Arabidopsis Genome Initiative [AGI], 2000; http://www.Arabidopsis.org/; http://www.kazusa.or.jp/) and provides a good opportunity for comparative genomics. It is time to answer the questions such as: To what extent can the knowledge gained from the Arabidopsis genome sequence be applied to understand the tomato genome, a species that separated from Arabidopsis more than 100 million years ago (MYA; Gandolfo et al., 1998; Yang et al., 1999)? The sequence of the 105-kb BAC from tomato chromosome 2, containing the ovate locus, was compared with the Arabidopsis genome, shedding interesting light on the synteny of the genomes of these two model species (Ku et al., 2000). Considering the genome size of tomato (approximately 970Mb), more comparative sequence analysis will be required before we can develop a comprehensive and confident comparison of the genomes of these two dicot plants.

In our effort to map-base clone the tomato JOINTLESS gene, a gene that controls the development of pedicel abscission zones (Mao et al., 2000a), we sequenced a 118,813-bp BAC clone encompassing the tomato JOINTLESS region on chromosome 11. In addition to the 15 putative open reading frames (ORFs) detected, repetitive sequences, especially small DNA elements such as foldback transposons, were also described in detail here. A comparison of the predicted tomato genes with the Arabidopsis genome showed that segmented gene colinearity between the two species was preserved, but located on different Arabidopsis chromosomes or on the same chromosome but physically separated. Unlike the chromosome 2 ovate locus, the JOINTLESS region seemed to be synteneous to the non-duplicated portion of the Arabidopsis genome.

	RESULTS AND DISCUSSION

TOP ABSTRACT INTRODUCTION RESULTS AND DISCUSSION CONCLUSIONS MATERIALS AND METHODS LITERATURE CITED

Sequence Characteristics of the Tomato JOINTLESS Locus

Tomato BAC clone 240K04 (GenBank accession no. AF292003.), containing the JOINTLESS locus, was isolated from an L. esculentum cv Heinz1706 BAC library (Budiman et al., 2000) by screening with jointless-linked RFLP markers TG523 and TY159L. Previous work showed that the genomic region containing the jointless locus has a low genetic/physical ratio of less than 100 kb per centiMorgan (cM) as demonstrated by yeast artificial chromosome physical mapping (Zhang et al., 1994), in contrast to the whole genome average of 750 kb cM¹ (Tanksley et al., 1992). The tomato genomic sequence on BAC 240K04 was found to be 118,813 bp long and contained TG523 and TY159L, genetically 1 cM apart, but separated by only approximately 50 kb (Fig. 1). In total, 15 putative genes (including two that belong to one retrotransposon) were predicted on 240K04, giving a gene density in this region as one every 8 kb. This ratio was close to a recent estimation of one gene per 6.2 kb at the ovate region of tomato chromosome 2 (Ku et al., 2000) and nearly twice that of Arabidopsis chromosome 2 (one gene per 4.4 kb; Lin et al., 1999). BAC 240K04 had an average GC content of approximately 32% with the 15 putative gene coding regions having a higher GC content of 42%. These data were comparable to a previous study where the whole genome GC content was estimated as 37% and coding regions as 46% (Messeguer et al., 1991). Concerning the repetitive sequences on BAC 240K04, the only known transposable element was a copia-like retrotransposon. Small DNA elements, however, were found frequently when the sequence of 240K04 was searched against GenBank, reminding us of small elements like miniature inverted-repeat elements (MITEs) in rice and maize (Bureau and Wessler, 1994; Bureau et al., 1996). These DNA elements (short DNA repeats [SDRs]) were of great interest because of their frequent association with gene sequences. In particular, some SDRs were directly involved in the mutation of tomato genes such as jointless (Mao et al., 2000a) and yellow flesh (Fray and Grierson, 1993).

View larger version (23K): [in this window] [in a new window]   Figure 1.   A graphical display of the predicted ORFs and the DNA elements on BAC 240K04. Arrows for each ORF indicate the coding orientation. The pointed arrows on 240K04.1 and 240K04.15 indicate that these two ORFs are 3' incomplete and the blunt arrows flanking the retrotransposon represent the two LTRs.

Genes

Fifteen putative genes were predicted by gene prediction programs such as GenScan and their functions were determined by searching the nonredundant protein database in GenBank (Table I, Fig. 1). ORFs 240K04.1 and 240K04.15 were incomplete because they were located at either end of the BAC. Six predicted genes (240K04.2, 3, 5, 6, 8, and 11) had no significant match in GenBank. However, five had significant expressed sequence tag (EST) matches with The Institute for Genomic Research tomato gene index (Lycopersicon Gene Index, http://www.tigr.org/tdb/lgi/), and thus were annotated as unknown proteins. In total, 11 out of 15 putative genes had significant EST matches in the LGI, including ORFs 240K04.12 and 240K04.13, that apparently belong to the same retrotransposon as delimited by the two LTRs as described below. In the tomato ovate region, however, only six out of the 17 ORFs were found to have EST matches (Ku et al., 2000). ORFs 240K04.6 and 240K04.7 may be from the same gene family because a portion of their sequences were significantly similar to each other. Based on the GenBank search results, we were able to assign putative functions to nine ORFs including JOINTLESS, a member of the MADS-box transcription factor gene family (Table I).

Repetitive Sequences

Previous work showed that the tomato genome contains mostly low copy number sequences (Zamir and Tanksley, 1988). Ganal et al. (1988) estimated by hybridization with the major tomato repeated sequences that the total amount of highly repeated sequences might lie between 10% and 15%. An analysis of 1,205 random BAC end sequences also showed that the repetitive sequences in the tomato genome is around 12% (Budiman et al., 2000). Although a DNA database search of the 240K04 sequence did not detect large blocks of repetitive sequences, regions of small DNA segments were found with high similarity to more than 20 gene sequences from tomato or potato Solanum spp., indicating the repetitive nature of these DNA sequences.

Simple Sequence Repeats (SSRs)

Retrotransposons

Foldback Transposon Tomato Anionic Peroxidase Inverted Repeat (TAPIR)

View larger version (47K): [in this window] [in a new window]   Figure 2.   Characterization of foldback transposon TAPIR. a, Strong secondary structure of the TAPIR element. b, A PCR survey of the three TAPIR loci in various tomato lines/species. Primers used are: TAPIR1f, 5'GAT AGT TAA AGA TGC GCC TAA C3'; TAPIR1r, 5'CGT GTG GGT GTA TAT CTA TTC3'; TAPIR2f, 5'GGT AGA TAG GCA AAA GTT TC3'; TAPIR2r, 5'GAA TAG ATA TAC ACC CAC ACG3'; TAPIR3f, 5'CAC TTG TTA TAC ACT TGT GAT GG3'; and TAPIR3r, 5'CCT TGT TGG GTA TTT GCA TGT G3'. c, The sequence of the direct repeats (DRs) flanking the tomato TAPIRs. d, A phylogenetic tree developed using the middle loop sequences of 12 TAPIRs. 1BF09F, 3DC06F, and 3AB02F were three BAC ends. The lowercase r indicates the reverse strand.

SDRs

Gene Colinearity of Tomato BAC 240K04 and the Arabidopsis Genome

Four criteria were used for synteny analysis: homology, physical distance, colinearity, and gene orientation. This means that all the genes considered had significant homology with each other, were physically close, and were in the same order and transcriptional orientation. First, the protein sequences from the fifteen putative tomato genes from BAC 240K4 were searched against the Arabidopsis sequences in GenBank using TBlastN. All but one (240K4.03) of the putative genes had significant similarity with the Arabidopsis genome with an expectation value of E < 10¹⁰ (Table III). Besides 240K04.12 and 240K04.13 that belong to the retrotransposon polyprotein, six more ORFs had more than five Arabidopsis matches with an E value less than 10¹⁰, indicating that they were members of various gene families. An example was 240K04.14, the MADS-box gene whose MADS-box domain is present in numerous MADS-box transcription factors in Arabidopsis. ORF 240K04.9, a putative auxin-independent growth promoter, had more than 25 Arabidopsis matches with E<10¹⁰. The multiple Arabidopsis matches of tomato ORFs were apparently due to the high number of gene families in the Arabidopsis genomemore than 37% of Arabidopsis genes have >five members (AGI, 2000). Gene families may cause problems when to determine the Arabidopsis orthologue for the correspondent tomato gene based on sequence homology. Therefore, in our microsynteny study, homology is only one of the four criteria evaluated.

Figure 3 shows the network of microsynteny of the tomato ORFs on BAC 240K04 with the Arabidopsis genome. Regions in Arabidopsis were not labeled with the names of the ORFs due to the fact that in both tomato and Arabidopsis most of the ORFs were derived from the gene prediction programs and may not be correspondent ORF by ORF. Because the borders of ORFs on tomato BAC 240K04 were distinguishable we believe that the correspondent regions in Arabidopsis genome should represent separate ORFs. Five Arabidopsis DNA segments were found to contain clusters of genes that were physically close to each other and were in the same order as their homologs in tomato (G1-G5). The coding orientations of the corresponding genes were exactly the same in both genomes, meeting the criteria set above. The number of genes in each segment were from two to five. G3 and G5 contained two genes in less than 10 kb that were significantly similar to the tomato genes. The largest segment G4 contained five genes that were colinear with those on the tomato BAC 240K04. The ORF 240K04.6 had nine matches in the Arabidopsis genome. Three of them were located in the three syntenic segments from the Arabidopsis chromosomes II, IV, and V, respectively. Another ORF 240K04.14, the MADS-box gene, was associated with two Arabidopsis syntenic segments, G1 and G3. The presence of the MADS-box genes in the vicinity of other tomato homologs in Arabidopsis may be coincidental due to a large number of such genes and their highly conserved DNA binding domain. However, the identical coding orientation in both Arabidopsis and tomato when compared with the other gene(s) in the same syntenic segment supported that these MADS-box genes did fall within syntenic clusters in Arabidopsis.

View larger version (31K): [in this window] [in a new window]   Figure 3.   A schematic display of the Arabidopsis syntenic segments to tomato BAC 240K04. The number of centiMorgans (cM) indicate the approximate positions of the corresponding Arabidopsis BAC/P1-derived artificial chromosome clones on the Arabidopsis genome. Black arrows indicate ORFs that do not fall in any colinear segments. Arrows on the tomato BAC indicate approximate positions of the predicted ORFs, whereas arrows on Arabidopsis BAC/P1-derived artificial chromosome clones indicate major homologous regions.

Overall, the five Arabidopsis syntenic segments were from three chromosomes. G2 and G3 were from chromosome IV and more than 37 cM apart. G4 and G5 were from chromosome V and more than 120 cM apart. However, no duplicated Arabidopsis segments were observed for these syntenic clusters, compared with the tomato chromosome 2 ovate region where the Arabidopsis syntenic segments were duplicated (Ku et al., 2000). Thus, the Arabidopsis segments syntenic to the tomato JOINTLESS region may fall in the portion of the genome that was not duplicated because the duplicated part in the Arabidopsis genome comprises only 60% (AGI, 2000). It is interesting that none of the genes in the Arabidopsis syntenic segments contained a gene that had a different coding orientation to its tomato counterpart. In the tomato chromosome 2 ovate region, however, three out of 12 syntenic tomato genes were located on the opposite strand in Arabidopsis (Ku et al., 2000). This may indicate that the duplicated portion of the Arabidopsis genome could be subject to less selection pressure because of gene copy redundancy and therefore may accommodate more genetic rearrangements.

Although the duplicated Arabidopsis genome may fit the model of polyploidization and subsequent gene loss (AGI, 2000; Ku et al., 2000), we observed that ORF 240K04.6 was associated with three syntenic segments from three Arabidopsis chromosomes. In each segment, 240K04.6 was in the middle of the syntenic cluster and therefore the possibility of a false correlation was low, although ORF 240K04.6 had nine significant matches with the Arabidopsis sequences. The presence of one ORF in more than two synteneous segments supported the hypothesis that other alternatives such as independent segmental duplication may also be possible in shaping the Arabidopsis genome (AGI, 2000). We also noticed that three Arabidopsis segments (G1, G3, and G4) were spanning the tomato region that contained the retrotransposon (represented by 240K04.12 and 240K04.13). However, none of these segments contained a retrotransposon in their syntenic context, indicating that the transposition event in tomato occurred after the divergence of these two species about 100 MYA (Gandolfo et al., 1998; Yang et al., 1999).

	CONCLUSIONS

TOP ABSTRACT INTRODUCTION RESULTS AND DISCUSSION CONCLUSIONS MATERIALS AND METHODS LITERATURE CITED

The analysis of the 118,813-bp region from the tomato JOINTLESS region revealed that the tomato genome may contain an abundance of SDRs or small transposons. The presence of some of these elements may be very recent events as demonstrated by the PCR experiment to investigate the related loci in the wild tomatoes. Although SDRs were frequently associated with tomato genes, they did not appear to have functions in regulating adjacent genes. Unlike in rice and maize where small DNA elements, such as MITEs, have been found in large numbers and associated with genes too (Bureau and Wessler, 1994; Wessler et al., 1995; Mao et al., 2000b), the Arabidopsis genome has fewer numbers of such elements (Casacuberta et al., 1998; Ade and Belzile, 1999; Le et al., 2000). Therefore, the large number of such elements may contribute significantly in the expansion of the genomes of rice, maize, and tomato.

Comparison of the tomato JOINTELSS locus with the Arabidopsis genome demonstrated that small syntenic segments could be easily found in the genomes of these two dicot plants that have been diverged more than 100 MYA. In particular, the strict conservation of the gene-encoding orientations in these colinear segments indicates that the region may represent an ancestral block of genes. It is interesting that the comparison of this region did not reveal duplicated segments in Arabidopsis in which more than 60% of its genome is duplicated. It is tempting to put forward a hypothesis about the synteny between the tomato and Arabidopsis genomes, but a clearer picture will be visible only when we have additional long stretches of tomato sequence available for investigation.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION RESULTS AND DISCUSSION CONCLUSIONS MATERIALS AND METHODS LITERATURE CITED

Tomato (Lycopersicon esculentum) BAC Library Screening

Hybridization of high-density tomato BAC filters using TG523 and TY159L insert is as described elsewhere (Mao et al., 2000a). DNAs of positive clones for insert check and Southern analysis were isolated using the standard alkali lysis method (Sambrook et al., 1989).

Construction of a Shotgun Library of Clone 240K4 and Sequencing

A maxiprep of tomato BAC clones was performed using alkali lysis and further purified by CsCl gradient centrifugation using a standard protocol (Sambrook et al., 1989). BAC DNA was nebulized under 6.5 pounds per square inch for 4 min and end repaired. Fragments of 2 to 5 kb were cut out from an agarose gel, purified, and ligated to pBluescipt. The ligation products were electroporated into DH10B competent cells (Gibco BRL, Rockville, MD). Clones were picked randomly and stored in 96-well microtiter plates.

DNA sequencing was performed on ABI 377XL automatic sequencers. Templates of shotgun clones were prepared by the AutoGen (Integrated Separation Systems, Portsmouth, NH) from 3-mL overnight cultures. DNA from each preparation was finally dissolved in 80 µL of water or Tris EDTA and 4 µL was used for sequencing reactions using the BigDye Cycle Sequencing Kit following manufacturer's instructions (ABI, Columbia, MD).

Sequence Analysis

Production sequencing of approximately 2,700 shotgun clones was performed with an additional 200 reactions for finishing. Phred (Ewing and Green, 1998; Ewing et al., 1998) was used for base calling from the trace files, and Consed (Gordon et al., 1998) was used to assemble the sequences into contigs. The final consensus sequence was searched against GenBank using BlastN and BlastX algorithms (Altschul et al., 1997) through the Web site of the National Center for Biotechnology Information (www.ncbi.nlm.gov). Gene structures were predicted with Genscan, Genscan+ (Chris Burge, Massachusetts Institute of Technology, Cambridge, http://CCR-81.mit.edu/GENSCAN.html), and GeneMarkHMM (Mark Borodovsky, Georgia Tech, Atlanta, http://genemark.biology.gatech. edu/GeneMark/). Arabidopsis databases was searched through The Arabidopsis Information Resource Web site (http://www.Arabidopsis.org/) and The Kazusa Arabidopsis data opening site (http://www.kazusa.or.jp/kaos/).

The GCG package (University of Wisconsin, Madison, Version 10) command FOLDRNA was used for analyzing secondary structures. The phylogenetic tree was made using PUAPSEARCH and PAUPDISPLAY from multiple sequence alignment generated by PILEUP.