A Solanum lycopersicum × Solanum pimpinellifolium Linkage Map of Tomato Displaying Genomic Locations of R-Genes, RGAs, and Candidate Resistance/Defense-Response ESTs

We have identified an accession (LA2093) within the tomato wild species Solanum pimpinellifolium with many desirable characteristics, including biotic and abiotic stress tolerance and good fruit quality. To utilize the full genetic potential of LA2093 in tomato breeding, we have developed a linkage map based on an F2 population of a cross between LA2093 and a tomato breeding line, using 115 RFLP, 94 EST, and 41 RGA markers. The map spanned 1002.4 cM of the 12 tomato chromosomes with an average marker distance of 4.0 cM. The length of the map and linear order of the markers were in good agreement with the published maps of tomato. The ESTs were chosen based on their sequence similarities with known resistance or defense-response genes, signal-transduction factors, transcriptional regulators, and genes encoding pathogenesis-related proteins. Locations of several ESTs and RGAs coincided with locations of several known tomato resistance genes and quantitative resistance loci (QRLs), suggesting that candidate-gene approach may be effective in identifying and mapping new R genes. This map will be useful for marker-assisted exploitation of desirable traits in LA2093 and other S. pimpinellifolium accessions, and possibly for utilization of genetic variation within S. lycopersicum.


INTRODUCTION
The tomato Solanum species (Solanum subsection Lycopersicon) include the cultivated tomato, S. lycopersicum L. (formerly Lycopersicon esculentum Miller), and more than 10 related wild species (http://www.sgn.cornell.edu/about/ solanum nomenclature.pl). It is estimated that S. lycopersicum accounts for only about 5% of the total genetic variability in the tomato gene pool [1]. Conversely, the tomato wild species bear a wealth of genetic variability for many agriculturally and biologically important characteristics. During the past several decades, tomato wild species have been utilized extensively in traditional breeding programs, however, mainly for improvement of simply inherited traits such as vertical disease resistance. Genetic variation in the wild species for complex traits such as tolerance to environmental stresses, quantitative disease resistance, and fruit yield and quality has remained largely unexploited [2]. This is mainly due to the inadequacy of traditional breeding protocols to identify, select, and successfully transfer genes controlling such complex traits. The identification of genes underlying quantitative characters is often difficult, particularly if their phenotypic effects are unrecognizable from the phenotype [3]. Furthermore, transfer of desirable genes from wild species into elite breeding lines is not without inherent difficulties. Upon interspecific hybridization, a major task becomes eliminating the great bulk of undesirable exotic genes while maintaining 2 International Journal of Plant Genomics and selecting for desirable characteristics. These limitations, however, may no longer be insurmountable with the advent of molecular biology tools such as genetic markers and maps and marker-assisted selection (MAS). Among various advantages, molecular markers and maps can facilitate determination of the number, chromosomal location, and individual and interactive effects of genes (or quantitative trait loci (QTL)) that affect complex traits. Following their identification, desirable genes or QTLs can be introgressed into the cultigen and undesirable characteristics eliminated by foreground and background MAS.
During the past few decades, several molecular linkage maps of tomato have been developed mainly based on interspecific crosses between the cultivated and related wild species of tomato (for a complete list see Foolad 2007). The first molecular linkage map of tomato was published in 1986, which included 18 isozyme and 94 DNA markers [4]. The high-density linkage map of tomato, which originally was developed based on an F 2 population of a S. lycopersicum × S. pennellii cross and 1030 molecular markers [5], currently comprises more than 2000 markers with intermarker spacing of ≤1 cM (http://www.sgn.cornell.edu/cview/map.pl?map id=9). The high level of molecular marker polymorphism between S. lycopersicum and S. pennellii facilitated the development of this high-density map. With this genetic map, it is likely that any gene of interest would be within one to a few centiMorgan (cM). However, many important agricultural traits are not segregating in this population and many of the markers in this map are not polymorphic in other populations of tomato, in particular those derived from intraspecific crosses within the cultigen or between the cultigen and closely related wild species such as S. pimpinellifolium L. (formerly L. pimpinellifolium (L.) Miller) and S. cheesmaniae (L. Riley) Fosberg (formerly L. cheesmaniae L. Riley). For example, it has been determined that only ∼30% of the RFLP markers in the high-density map detect polymorphism in S. lycopersicum × S. pimpinellifolium populations following digestion of genomic DNAs with many restriction enzymes [6,7]. In a more recent study, only less than 15% of the RFLP markers from the high-density map detected polymorphism between a Mexican accession of S. pimpinellifolium and a S. lycopersicum breeding line (MR Foolad et al., unpubl.). Such low levels of marker polymorphism necessitated the development of several species-specific molecular maps, as listed elsewhere [2]. Among the different wild species of tomato, however, genetic maps developed based on crosses between the cultivated tomato and S. pimpinellifolium would be more useful for practical purposes, as described below.
S. pimpinellifolium is the only red-fruited wild species of tomato and the only species from which natural introgression into the cultivated tomato has been detected [8]. In addition, during the past several decades, extensive genetic introgressions from this species into the cultivated tomato have been made through plant breeding [8][9][10]. Accessions within S. pimpinellifolium are highly self-compatible and bidirectionally cross-compatible with the cultivated tomato. Because of the close phylogenetic relationships between the two species, there is little or no difficulty in initial crosses or in subsequent generations of prebreeding and breeding activities. Furthermore, S. pimpinellifolium harbors numerous desirable horticultural and agronomic characteristics, including disease resistance [11][12][13], abiotic stress tolerance [14,15], and good fruit quality [2,16], and much fewer undesirable traits than most other wild species of tomato. However, to utilize the full genetic potential of this species, it is necessary to detect molecular polymorphisms between this species and the cultivated tomato. Detection or development of polymorphic markers, in particular functional markers (see below), and construction of new molecular linkage maps based on desirable S. lycopersicum × S. pimpinellifolium crosses are a step toward genetic exploitation of this species. Furthermore, because of extensive introgressions from S. pimpinellifolium into modern cultivars of tomato, such markers and maps will also be useful when exploiting the available genetic variation within the cultigen.
Most of the previous genetic linkage maps of tomato were constructed based on random genetic markers such as RFLPs, RAPDs, AFLPs, and SSRs. Recently, however, DNA sequences based on expressed sequence tags (ESTs) and resistance gene analogs (RGAs) have become available, which can be used to develop genetic markers and maps or used as candidates to identify functional genes. Development of markers and maps based on informative sequences will be useful for identification and potentially cloning of genes and QTLs of agricultural and biological significance. ESTs are generally derived from cDNA clones and may have applications in gene sighting, genome mapping, and identification of coding regions in genomic sequences. While ESTs can serve the same purposes as random DNA markers, they provide the additional feature of pointing directly to expressed genes and thus can expedite gene discovery and comparative genomics. The growing EST databases in different plant species, including tomato, have provided valuable resources for development of EST-based markers. The association of EST markers with phenotypes can lead to a better understanding of biochemical pathways and mechanisms affecting important traits. Identification and characterization of RGAs has also been proposed as a candidate-gene approach to identify genes potentially related to disease resistance [17][18][19][20][21]. Although not all amplified products may correspond to functional disease resistance genes [21], RGA primers have been shown to amplify the conserved sequences of leucine-rich repeats (LRR), nucleotide-binding sites (NBS), or serine/threonine protein kinases (PtoKin), thereby targeting genes and gene families for disease resistance, defense response, or other important signal transduction processes [22]. Thus, RGAs have been considered useful not only as genetic markers but also as potential that leads to the identification of important genes. During the past decade, RGAs have been used for mapping of QTLs for many important characters, including disease resistance.
Recently, we identified several accessions of S. pimpinellifolium (including LA2093) with desirable horticultural characteristics such as disease resistance, abiotic stress tolerance, and good fruit quality. To facilitate genetic characterization Arun Sharma et al. 3 and exploitation of LA2093, and possibly other accessions, we have developed a genetic linkage map based on an F 2 population of a cross between LA2093 and tomato breeding line NCEBR-1 using 250 DNA markers, including RFLPs, ESTs, and RGAs. Previously, two molecular linkage maps of tomato based on different crosses between S. lycopersicum (denoted as L) and S. pimpinellifolium (denoted as PM) were reported by Grandillo and Tanksley [6] (referred to as L × PM1 map) and Chen and Foolad [7] (referred to as L × PM2 map). The map presented here (referred to as L × PM3) is different but complementary to the previous two L × PM maps, as it contains a large number of ESTs and RGAs along with some new RFLP anchor markers that can facilitate molecular investigation and exploitation of this and other accessions of S. pimpinellifolium. We have compared the L × PM3 map with other molecular linkage maps of tomato and discussed similarities and differences in relation to phylogenetic relationships between parents of the various mapping populations.

Plant materials and mapping population
Inbred sources of NCEBR-1 (S. lycopersicum) and LA2093 (S. pimpinellifolium) were hybridized and F 1 progeny produced. NCEBR-1 (PVP) is a horticulturally superior, multiple disease resistant, advanced tomato breeding line received from RG Gardner, University of North Carolina, Fletcher, NC, USA. A single F 1 hybrid plant was self-fertilized to produce F 2 seed. A total of 900 F 2 individuals were grown under field conditions and screened for various characteristics. Among other traits, the population was segregating for growth habit (determinate versus indeterminate). Indeterminate growth habit is an undesirable characteristic with confounding effects on other characteristics such as disease resistance and fruit quality. To obtain a population suitable for QTL mapping and breeding purposes, indeterminate plants were eliminated. A total of 172 F 2 individuals, hereafter referred to as the L × PM3 F 2 population, were chosen and grown to maturity and used to construct the molecular linkage map.

RFLP analysis
Nuclear DNA was extracted from approximately 10 g of leaf tissue from each of the parental lines and F 2 individuals using standard protocols for tomato [23,24]. Genomic DNAs were treated with RNase and digested with eight restriction enzymes, including DraI, EcoRI, EcoRV, HaeIII, HindIII, RsaI, Sca1, and Xba1 following manufacturers' instructions, and parental polymorphism survey blots were prepared. To identify sufficient number of polymorphic anchor RFLP markers to develop a framework linkage map, parental survey blots were probed with a total of 340 random tomato genomic (TG) or cDNA (CD or CT) clones, originally chosen from the high-density molecular linkage map of tomato [25]. Agarose gel electrophoresis, Southern blotting, hybridizations, and autoradiography were conducted as described elsewhere [26]. Probes were labeled with [ 32 P]dCTP by primer extension [27]. Following identification of polymorphic RFLP markers (see Section 3 for rates of polymorphism), genomic DNAs of the 172 F 2 individuals were digested with the 8 restriction enzymes and multiple sets of Southern blots were prepared. Blots were hybridized with clones detecting polymorphism and a total of 115 RFLP markers were scored in the F 2 population.

EST analysis
A set of unique ESTs was selected from the tomato gene index sources maintained by The Institute for Genomic Research (TIGR; http://www.tigr.org/) (now at the Computational Biology and Functional Genomics Laboratory at Harvard University; http://compbio.dfci.harvard.edu/ tgi/cgi-bin/tgi/est report.pl). Each EST represents a valid (partial or complete) copy of a transcribed functional allele. We selected 140 ESTs from a diverse array of candidate genes and gene families, many of which are known or assumed to play roles in disease-resistance or defense-response mechanisms. Among them we included ESTs with homology to resistance (R) genes, signal transduction genes, transcriptional regulator factors, and genes encoding pathogenesisrelated proteins. We used this targeted strategy to obtain a set of potentially functional markers for marker-assisted selection in our tomato-breeding program. The 140 EST clones, purchased from the Clemson University Genomics Institute (http://www.genome.clemson.edu/), Clemson, SC, USA, were used as RFLP probes to identify polymorphism between the two parents. Among them, 96 provided polymorphic alleles ( Table 1). The polymorphic ESTs were used as RFLP probes to genotype the F 2 individuals, examine their segregation, and map onto the tomato chromosomes.

Selection of primers
Ten pairs of oligonucleotide primers, previously designed based on conserved LRR, NBS, and PtoKin motifs of several resistance genes, were used (Table 2; [28]). Some primers were chosen to be degenerate at the redundant third position (3 end) in the codons to cover a range of possible sequences encoding the motifs, and thus to increase the efficiency of PCR amplification [19,29]. Only one pair of primers was used for each PCR amplification.

PCR amplification
PCR conditions for amplification of RGAs were described elsewhere [11]. Briefly, each amplification was performed in a 25-μL volume consisting of 300 μM each of dATP, dCTP, dGTP, and dTTP, 5 mM of MgCl 2 , 1 unit of Taq DNA polymerase, 2.5 μL of 10X buffer (PCR Core system I; Promega, Madison, Wis, USA), 2 μL of each primer, and 40 ng of genomic DNA that was used as template. For control reactions, the template was substituted with sterile, nucleasefree ddH 2 O. All PCR mixtures were overlaid with mineral oil and carried out in a Perkin Elmer DNA Thermal Cycler  Genomics Laboratory). c The exact or approximate copy number of ESTs in tomato genome was determined based on the number of hybridized bands on Southern blot gels and may be varied in different labs. Where there is a "/" sign, the figures in the left side denote the number of copies in S. lycopersicum parent and those in the right side denote the number of copies in S. pimpinellifolium parent.

6
International Journal of Plant Genomics

Gel electrophoresis and silver staining
Denaturing polyacrylamide gel electrophoresis (PAGE) was used to separate and detect individual RGA bands [30]. Briefly, a denaturing gel (7 M Urea-6% polyacrylamide) was prepared in a sequencing gel apparatus (420 × 330 × 0.4 mm; Fisher Biotech, Springfield, NJ, USA) using Bind-and Repel-Silane (Promega). After polymerization, the gel was prerun in 1X Tris-borate-EDTA (TBE) buffer for 30 minutes at 40 W (∼1400 V) to reach a gel temperature of 50 • C. Twelve μL of loading buffer (10 M Urea-0.08% xylene cyanole) were added to each 25 μL amplified DNA sample and the mixture was denatured at 95 • C for 5 minutes and immediately put on ice. After cleaning the gel-loading surface, a 0.4 mm-thick shark comb (Fisher Biotech, Springfield, NJ, USA) was inserted into the gel. Subsequently, 7 μL of each PCR-amplified sample were loaded. Each gel accommodated 60 DNA samples and three DNA size markers (1 Kb, 100 bp, 50 bp). The gel was run at 35 W (∼1350 V) for 3.5-4 hours. After electrophoresis, the gel, fixed to the Bind-Silane surface of one glass plate, was silver stained following the manufacturer's protocol (Promega). The gel was air dried at room temperature overnight and stored in dark for future scoring and scanning. All amplifications and gel electrophoresis procedures were repeated at least once.

Identification of informative RGA markers
Following gel electrophoresis and staining, polymorphic and monomorphic bands were observed. Polymorphic bands were directly scored as dominant markers and used for genetic mapping. To determine whether monomorphic bands could detect polymorphism if used as RFLP probes, they were excised from the gel (as described in [28,32]), purified with the QIAgene quick Gel Extraction Kit (QIA-GEN, Valencia, Calif, USA), labeled with 32 P-dCTP, and used to hybridize the parental survey blots. Probes which detected polymorphism between the two parents were then used to hybridize Southern blots of the F 2 population, and scored as either dominant or codominant markers. Overall, a total of 43 RGA markers were successfully scored and mapped onto the 12 tomato chromosomes.

Size determination of RGA fragments
PAGE polymorphic and monomorphic fragments were excised from the dried polyacrylamide gel and reamplified, by using a needle scratching and PCR reamplification method [32]. The reamplified products and DNA size markers (1 Kb, 100 bp, and 50 bp) were run on a 1.0% agarose gel, stained with ethidium bromide, and photographed to determine the size.

Statistical and mapping analyses
Segregation of the 250 DNA markers (115 RFLPs, 94 ESTs, and 41 RGAs) in the F 2 population was tested for deviation from the expected Mendelian genotypic ratios of 1 : 2 : 1 (for codominant) or 1 : 1 (for dominant markers) using chi-square (χ 2 ) goodness-of-fit analysis. Multipoint linkage analysis of the genetic markers in the F 2 population was performed using the MapMaker program v. 3.0 [33] and a genetic linkage map was constructed. Briefly, the group command was used to assign markers into linkage groups using a minimum LOD score of 3.0 and a maximum recombination fraction of 0.20. Three-point linkage analysis was performed to determine the maximum likelihood recombination fraction and the associated LOD score for each combination of loci. The "order" and "compare" commands were used to find the best order of loci within each group, followed by using the "ripple" command to verify the order. Markers were included within the framework map only if the LOD value for the ripple was greater than 3.0. Once the linear order of markers along each chromosome was determined, recombination frequencies between markers were estimated with multipoint linkage analyses. The Kosambi mapping function [34] was used to convert recombination frequencies to map distances in cM. The distribution of percentage of the S. lycopersicum genome (L) in the F 2 population was estimated using the computer program Qgene [35].

RFLP polymorphism between S. lycopersicum and S. pimpinellifolium
RFLP clones were chosen from two sources, a previously published S. lycopersicum (NC84173) × S. pimpinellifolium (LA722) linkage map (L × PM2) [7] and the high-density S. lycopersicum (VF36 -Tm2 a ) × S. pennellii (LA716) linkage map of tomato (L × P) [25]. Of the 152 RFLP clones chosen from the L × PM2 map, 82 (54%) were polymorphic between the two parents (NCEBR-1 and LA2093) in the present study. Of the 120 clones that were chosen based on the high-density L × P map, 40 (30%) were polymorphic between NCEBR-1 and LA2093. The latter level of polymorphism was similar to those previously reported by Grandillo and Tanksley [6] and Chen and Foolad [7] for different S. lycopersicum × S. pimpinellifolium crosses. A lower level of DNA polymorphism between S. lycopersicum and S. pimpinellifolium compared to that between S. lycopersicum and S. pennellii was expected as S. pimpinellifolium is phylogenetically much closer to the cultivated tomato [1,36,37]. The high-density map of tomato was constructed based on A S. lycopersicum × S. pennellii cross mainly because of the presence of high level of marker polymorphism between the two species. However, identification of polymorphic markers and development of maps based S. lycopersicum × S. pimpinellifolium crosses are essential to facilitate marker-assisted exploitation of genetic variation present in S. pimpinellifolium. Such information may also be useful for exploitation of intraspecific variation within S. lycopersicum. This is because of frequent introgressions from S. pimpinellifolium into the cultivated tomato,which have occurred both naturally and deliberately via plant breeding [8]. In the present study, a total of 117 polymorphic RFLP clones were used to construct the backbone linkage map.

EST polymorphism between S. lycopersicum and S. pimpinellifolium
From a total of 140 tomato ESTs examined, 91 (65%) were polymorphic between the two parents. Five of 91 EST clones produced more than one polymorphic band, thus resulting in the detection of a total of 96 polymorphic EST loci, including 91 codominant (∼95%) and 5 dominant markers. Of the 96 EST markers, 94 were successfully scored in the F 2 population and mapped onto the 12 tomato chromosomes using the 115 RFLP anchor markers. The number of EST markers per chromosome ranged from 4 (on chr. 12) to 12 (on chr. 10). Observation of a high level of polymorphism in EST markers between S. lycopersicum and S. pimpinellifolium was unexpected, but encouraging. This high level of polymorphism could be due to various reasons including high copy number of EST bands (compared to the often single-copy RFLP markers) and the nature of the genes or gene families from which ESTs were selected. As indicated earlier, most ESTs were chosen based on their sequence similarities with genes or proteins related to disease resistance. It is likely that chromosomal regions containing resistance gene families accumulate a great deal of variation during their evolution, thus increasing the frequency of restriction sites, which are a basis for polymorphism. Because modern breeding lines have received frequent introgressions from different tomato wild species, in particular for disease resistance, presence of such introgressions in NCEBR-1 could have contributed to the high level of observed polymorphism. Further inspections of the chromosomal locations of ESTs support this submission, as discussed below. However, the observation of high level of EST polymorphism is promising as larger number of ESTs are becoming available.

Marker segregation
Of the 250 marker loci scored in the L × PM3 F 2 population, 41 (16.4%) exhibited significant deviation from the expected 8 International Journal of Plant Genomics 1 : 2 : 1 (codominant) or 1 : 1 (dominant) segregation ratios at P ≤ .01. Markers with skewed segregation were located on chromosomes 1, 3, 4, 5, and 6, with those on chromosome 6 exhibiting the highest level of skewness (Table 3). Markers on chromosomes 1, 3, and 4 exhibited distortion in favor of S. pimpinellifolium alleles whereas those on chromosomes 5 and 6 were in favor of S. lycopersicum. Observation of extensive segregation distortion for markers on chromosome 6 was not unexpected and could be attributed to the selection of determinate F 2 plants (as described in Section 2) and the presence of self-pruning (sp) locus on this chromosome (∼3 cM from RFLP marker TG279) [6]. Skewed segregation for markers on this chromosome was previously reported in other interspecific crosses of tomato, where phenotypic selection (PS) or MAS was employed to remove indeterminate plants from mapping populations [28,38,39]. However, in the present study, despite skewed segregation for markers on chromosome 6, no major differences in genetic map distances were observed when they were compared with the high-density map of tomato [39] or the previous S. lycopersicum × S. pimpinellifolium maps [6,7], where no such selections were practiced. Skewed segregation has been reported in many interspecific crosses of tomato, with the extent of skewness being greater in wider crosses compared to crosses between closely related species, and also generally greater in F 2 than in backcross populations [6,[40][41][42][43][44][45]. A survey of recently published results of interspecific crosses of tomato indicated that skewed segregation was 8.3% in the L × PM1 BC 1 population [6], 9.9% in the L × PM2 BC 1 population [7], 51% in a S. lycopersicum × S. cheesmaniae (L × CH) F 2 population [42], 69% in a S. lycopersicum × S. chmielewskii (L × CL) BC 1 population [46], 15% in a S. lycopersicum × S. habrochaites (L × H1) BC 1 population [38], 62% in the L × H2 BC 1 population [28], and 80% in a S. lycopersicum × S. pennellii (L × P) F 2 population [47]. The L × PM populations exhibited less overall skewed segregation than the other interspecific crosses, consistent with the close phylogenetic relationship between S. lycopersicum and S. pimpinellifolium. However, the relatively high level of skewed segregation in the L × CH F 2 population [42] and the low level of skewed segregation in the L × H1 BC 1 populations [38] were unordinary because S. cheesmaniae is a closely related and S. habrochaites is a distantly related wild species of tomato [1,9,10,48,49]. Skewed segregation in interspecific crosses of tomato has been attributed to various causes, including self-incompatibility (SI), unilateral incongruity, and gametophytic, zygotic, and viability selection in segregating populations, as discussed elsewhere [44,[50][51][52].

Genome composition of the F 2 population
The genomic compositions of the 172 F 2 individuals were determined based on the 220 codominant markers using qgene program. On average, the F 2 population was inferred to contain 51.5% of its genome from the S. lycopersicum parent (L alleles), which is very close to   (Figure 1) could be returned to nearly 100% L genome within 2-4 backcrosses, far more rapid than the 4-6 backcrosses routinely needed to eliminate donor genome without MAS. Alternatively, in a pedigree-type breeding program, marker analysis (if economically feasible) can facilitate inbreeding to homozygosity by selecting progeny at each generation which are homozygous over a maximal proportion of the genome.  (Figure 2), and this was mainly because of the low level of polymorphism between the two parents of this mapping population for markers on these chromosomes. This map was compared with several other molecular linkage maps of tomato for marker order, recombination frequencies, and total map length, as described below.
Inspections of the distribution of ESTs on different chromosomes indicated that in some cases they were clustered, for example, ESTs on chromosomes 4, 8, 10, and 11. Further inspections indicated that chromosomal locations of some clustered or individual ESTs were colocalized with approximate locations of some major disease-resistance genes (Rgenes) or quantitative resistance loci (QRLs), as inferred from other published researche (see Figure 2). While such colocalization suggests that these ESTs may be genetically related to resistance genes or QRL, their actual functionality relationships can only be determined by further analyses such as isolation and sequencing of full EST sequences and functional genomic studies.
Currently, there are more than 214 000 ESTs identified in tomato (http://compbio.dfci.harvard.edu/tgi/cgi-bin/tgi/ gimain.pl?gudb=tomato), of which only a small percentage has been mapped onto the tomato chromosomes (http://www.sgn.cornell.edu/cgi-bin/search/direct search.pl? search=EST). The ESTs were derived from more than 23 cDNA libraries [116,117] and their sequences are available on Solanaceae Genome Network (SGN; http://www.sgn.cornell.edu/). All but four (cLET10E15, cLER4F5, cLEC6O2, and cLEG9N2) of the ESTs mapped in the present study were not previously mapped onto tomato chromosomes. Moreover, of the four that were previously mapped, different members of the corresponding contigs were mapped onto the same or different tomato chromosomes as in the present study. For example, cLET10E15 and cLER4F5 have overlap sequences with cLET1A5 and cLET3F16, respectively, and were mapped on the same chromosomes (http://www.sgn.cornell.edu/) as in the present study. cLEC6O2, which was mapped to chromosome 1 in the present study, was mapped to chromosome 8 and named cLPT1J10 (SGN: F 2 population ofa cross S. lycopersicum LA925 × S. pennellii LA716). EST clone cLEG9N2, which was mapped to chromosome 1 in this study, was previously mapped under cLET20B4 but with no known chromosomal position (http://www.sgn.cornell.edu/). Also, as indicated earlier, five of the EST clones resulted in two pairs of polymorphic bands each. For two of these clones, the two polymorphic bands were mapped onto two linked loci, that is, cTOC2J14a and cTOC2J14b on chromosome 5 and cLEC14I18a and cLEC14I18b on chromosome 11. Others were mapped onto different chromosomes; for example, cLEW22D11a was mapped to chromosome 6 whereas cLEW22D11b to chromosome 4, and cTOE7J7a was mapped to chromosome 1 whereas cTOE7J7b to chromosome 5.

Mapping of RGAs
PCR amplification using the 10 pairs of RGA primers ( Table 2) followed by denaturing PAGE resulted in the detection of a few hundred polymorphic and monomorphic bands. As described in Section 2, of the detected bands, 41 were strong and verifiable and thus were scored in the F 2 population. The amplified fragment size of these RGA bands ranged from 150 to 760 bp. Linkage analysis indicated that the 41 RGA markers were located on the 12 tomato chromosomes, ranging in number from 1 (on chrs. 3, 5, and 7) to 9 (on chr. 1) (Figure 2). The results indicated that RGA loci could be used as genetic markers for genome mapping, consistent with previous suggestions [28,30]. In several cases, RGA loci were clustered, similar to that observed for R-genes in various plant species [17,19,29,60,[118][119][120]. For example, on each of chromosomes 1, 2, 9, 10, 11, and 12, three or more RGA loci that were amplified from the same or different primer pairs mapped to the same or nearby positions (Figure 2). This observation indicated that different primers might initiate amplification of closely linked RGA loci that might be members of the same or different gene families.
Map positions of RGA loci were compared with chromosomal positions of known tomato R-genes and major QRL, whose positions were inferred from the previously published maps, as displayed and described in Figure 2. Most positions were inferred based on linkage to reference markers and thus should be considered best approximations. Colocalization of RGA loci with R-genes and QRLs were observed on a few chromosomes, including regions on chromosomes 1,2,7,8,9,10,11, and 12 ( Figure 2). These observations suggest the possibility of the presence of R or DR genes at the locations of RGAs, though this hypothesis could be confirmed only by extensive mapping and functional analysis of RGAs. Specifically, mapping of the associated RGAs in populations segregating for the colocalized Rgenes and cloning and molecular characterization of RGAs are necessary before any functional relationship could be established.
The map positions of the RGA loci in the present map (L × PM3) were compared with those reported in a S. lycopersicum × S. habrochaites (L × H2) map [28]. There were 19 common RGA loci between the two populations and 13 (68%) of which mapped to the same locations in the two maps, suggesting consistent and reproducible positions of RGAs across populations.

Comparison of the map with other molecular linkage maps of tomato
The present map (L × PM3) was compared with two previously developed S. lycopersicum × S. pimpinellifolium maps, including L × PM1 [6] and L × PM2 [7] as well as the high-density S. lycopersicum × S. pennellii (L × P) map of tomato [25]. The present map is different but complementary to L × PM1 and L × PM2 maps in several ways. First, different S. lycopersicum and S. pimpinellifolium parents and pretty much different molecular markers were used in the construction of the three maps. The L × PM1 was constructed based on a cross between a processing tomato cultivar (M82-1-7) and S. pimpinellifolium accession LA1589 using ∼120 RFLP and RAPD markers. The L × PM2 was constructed based on a cross between a fresh market tomato breeding line (NC84173) and S. pimpinellifolium accession LA722 using 151 RFLP markers. The current map (L × PM3) was constructed based on 250 RFLP, EST, and RGA markers using superior parental lines, as described earlier. It is expected that this map will have great utilities,  including exploitation of the genetic potential of LA2093 and other S. pimpinellifolium accessions. The second point of difference is that relatively a small percentage of the markers used in the present study were used in the previous two S. lycopersicum × S. pimpinellifolium linkage maps. Specifically, a new set of RFLP clones that detect polymorphism between S. lycopersicum and S. pimpinellifolium has been identified in the present study,  [7]; L × PM1, S. lycopersicum (M82-1-7) × S. pimpinellifolium (LA1589) map [6]; E × P, S. lycopersicum (VF36-Tm2) × S. pennellii (LA716) map [25].
beyond those that were identified in the construction of the previous two maps. However, an important observation is that markers that are polymorphic in one L × PM cross usually have a greater chance of being polymorphic in other L × PM crosses, compared to markers directly chosen from the high-density L × P map. Nonetheless, the observation that only 54% of the mapped RFLP clones in the L × PM2 population were polymorphic in the L × PM3 population indicates the presence of considerable DNA sequence variation among S. pimpinellifolium accessions. The overall results suggest that while for each S. pimpinellifolium accession new polymorphic markers need to be identified, the most useful sources would be those markers that have already been mapped in other S. lycopersicum × S. pimpinellifolium crosses. Third, unlike in the previous two L × PM maps, in the present map, "functional" markers such as ESTs and RGAs were used. Such markers may be more useful than random genetic markers for identification of candidate genes. The use of a large number of markers and the incorporation of functional markers in the present map extends its practical value in various genetics and breeding studies. However, the availability of three L × PM maps with rather different molecular markers should facilitate markerassisted exploitation of these and other S. pimpinellifolium accessions.
When the current map was compared with L × PM1 [6], L × PM2 [7], and the high-density L × P map [25], it was determined that the linear order of the common markers were generally the same. However, there were differences in interval lengths for several adjacent markers. For example, of 13 common marker intervals between L × PM3 and L × PM2 maps, 6 intervals on chromosomes 2, 3, 5, 6, and 12 differed in length by 2-3 fold, of which 1 interval was expanded in L × PM3 map. The difference between the two maps in marker interval lengths was not unexpected given the use of different type populations (F 2 versus BC 1 ), rather small size populations (172 and 119) and different number of markers (250 versus 151), all of which could have affected the occurrence and detection of recombination in different intervals. When the L × PM3 was compared with the high-density L × P map, which was constructed based on >1 000 genetic markers and 67 F 2 plants, genetic distances differed markedly for a large number of marker intervals. For example, for 36 common marker intervals, genetic distances differed between the two maps by at least twofold; of these, 7 intervals (23%) showed decreased and 13 (36%) showed increased recombination in the L × PM3 map. Greater differences in marker interval lengths between L × PM3 and L × P maps compared to that between L × PM3 and L × PM2 maps was not unusual considering the relatively close phylogenetic relationships between the L × PM3 and L × PM2 mapping populations.
When comparing the L × PM3 map with the highdensity L × P map, the most striking differences in genetic distances were observed in centromeric regions of chromosomes 3, 4, and 9, where substantial expansions in map distances were observed in the L × PM3 map, and in two locations on chromosome 12, where substantial contractions were observed in the L × PM3 map (Table 4 and Figure 1). The decrease in recombination frequencies in the centromeric regions of tomato chromosomes was previously attributed to the centromeric suppression of recombination [5,121,122]. Such suppression was suggested to be more frequent in wider crosses than in intraspecific crosses and crosses between closely related species. Further inspections indicated that the differences in genetic distances between the two maps across the rest of the genome were generally interval specific and not a characteristic of individual chromosomes. For example, for chromosomes 2, 3, 6, 10, 11, and 12, the L × PM3 map exhibited expansion in some intervals and contraction in others ( Table 4). As indicated earlier, such differences were due in part to the detection of chance recombination given the limited population sizes used in these studies.
Comparisons were also made across the four maps (L × PM3, L × PM2, L × PM1, and L × P) in terms of individual chromosome and total map lengths. The total length of the current map (1002 cM) was comparable with that of the L × P (1277 cM), L × PM1 (1275 cM), and the L × PM2 (1186 cM) maps. Furthermore, across the maps 14 International Journal of Plant Genomics the length of each chromosome in the current map was comparable to the corresponding chromosome in the other maps (Table 5).

CONCLUSION
A medium-density molecular linkage map of tomato is developed based on a cross between S. lycopersicum and S. pimpinellifolium, two phylogenetically closely related species. The parents of this map are superior genotypes and are expected to be useful for tomato crop improvement. This map will provide a basis for the identification, characterization, and introgression of useful genes and QTLs present in LA2093 and other S. pimpinellifolium accessions. It will also facilitate studies of gene and genome organization and evolution, dissection of complex traits, and targeted gene cloning. The map includes different types of molecular markers and provides a basis for identifying and adding other markers. The genomic locations of several EST and RGA markers coincided with locations of several known tomato R-genes or QRL, suggesting that candidate gene approach may be an effective means of identifying and mapping new R-genes and defining the genetic content of specific chromosomal regions. Because of the close phylogenetic relationship between the two species and the past frequent introgression of DNA from S. pimpinellifolium into S. lycopersicum, this map is expected to be particularly useful to breeding programs that exploit intraspecific variability within the cultivated tomato. The combined information from this and the two previously published S. lycopersicum × S. pimpinellifolium maps will facilitate further identification and exploitation of genetic variation within S. pimpinellifolium, S. lycopersicum var. cerasiforme, and S. lycopersicum.