An Introgression Line Population of Lycopersicon Pennellii in the Cultivated Tomato Enables the Identification and Fine Mapping of Yield-Associated QTL

Home , Introgression

An Introgression Line Population of Lycopersicon pennellii in the Cultivated Tomato Enables the Identification and Fine Mapping of Yield-Associated QTL

Yuval Eshed and Dani Zamir

Department of Field and Vegetable Crops and The Otto Warburg Center for Biotechnology, Faculty of Agriculture, The Hebrew University of Jerusalem, Rehovot 76100, Israel Manuscript received March 6, 1995 Accepted for publication August 11, 1995

ABSTRACT Methodologies for mapping of genes underlying quantitative traits have advanced considerably but have not been accompaniedby a parallel developmentof new population structures.We present a novel population consisting of 50 introgression lines (ILs) originating from a cross between the green-fruited species Lycopersicon pennellii and the cultivated tomato (cv M82). Each of the lines contains a single homozygous restriction fragment length polymorphism-definedL. pennellii chromosome segment, and together the lines provide complete coverage of the genome and a set of lines nearly isogenic to M82. A field trial of the ILs and their hybrids revealed at least23 quantitative trait loci (QTL) for total soluble solids content and 18 for fruit mass; these estimates are twice as high as previously reported estimates based on traditional mapping populations. For finer mappingof a QTL affecting fruit mass, the introgressed segment was recombined into smallerfragments that allowed the identification of three linked loci. At least 16 QTL for plant weight, 22 for percentage green fruit weight, 11 for total yield and 14 for total soluble solids yield were identified. Gene action for fruit and plant characteristics was mainly additive, while overdominance (or pseudo-overdominance) of wild species introgressions was detected for yield.

OMPLETE linkage maps of DNA markers have fa- major QTL. For example, 33-37% of the phenotypic C cilitated mapping of genes affecting quantitatively variation for seed weight in cowpea and mungbean was inherited traits (PATERSONet al. 1991a;TANKSLEY 1993). explained by a single QTL (FATOKUNet al. 1992). A Suitable mappingpopulations for such studies must QTL for glumehardness explained 42% of the variation possess sufficient polymorphism for markeranalysis and in a maize-teosinte cross (DOEBLEYand STEC1991). A for quantitative traits. For self-pollinated crops, such as major QTL can overshadow the effects of minor inde- rice (CAUSSEet al. 1994), soybean (KEIM et al. 1990) and pendently segregating QTL by increasing the total phe- tomato (MILLERand TANKSLEY1990), little variation notypic variation, and thusgenes withlesser effects between cultivated varieties is detectable by DNA mark- might fall below the threshold of detection. The over- ers. To overcome this problem, studies of quantitative shadowing effect interferes with the correct estimation trait loci (QTL) were performed on wide crosses be- of the number of QTL and with their fine mapping. tween species or races. The population structures most An additional restriction of standard populations may commonly used for QTL mapping in self-pollinated result from the interaction between two unlinked QTL. crops were FY/FBor recombinant inbreds. Interacting loci can reduce the difference between the Conventional mapping populationshave several limi- subgroups of the tested QTL and therefore theloci may tations in the accurate identification and fine mapping escape detection. of QTL. One of the major shortcomings is their resolu- In tomato, which has been a model plant for QTL tion power. In a simulation study on backcross popula- mapping, the low level ofvariation among thecultivated tions, it was demonstrated that forQTL of large effects, varieties is not limited to FWLPs but is also typical of in experiments with large populations and high density agronomically important traits. The narrow genetic ba- linkage maps, theconfidence interval for QTL map sis of the cultivated tomato has emphasized the value location is on the order of 10 cM (DARVASIet al. 1993). of exotic germplasm for improvementof the crop(RICK Another limitation is in the ability to identify QTL with 1982). QTL studies with complete genome coverage in tomato have been conducted on backcross (PATERSON small effects. In many instances it was found that alarge et al. F2/F3 (PATERSONet al. 1991b) and recombi- portion of the phenotypic variation for the measured 1988), nant inbred populations (GOLDMANet al. 1995) involv- traits could be explained by the segregation of a few ing interspecific crosses. In these populations each of the segregating plants possessed a large fraction of the Corresponding author: Dani Zamir, Department of Field and Vegeta- ble Crops, Faculty of Agriculture, The Hebrew University of Jerusa- wild species genome andsome individuals were charac- lem, P.O. Box 12, Rehovot 76100, Israel. terized by partial or complete sterility. A few major

Grnetics 141: 1147-1 162 (November, 199.5) 1148 Y. Eshed and D. Zamir

genes for sterility can exert an overshadowing effect on was not assayed in homozygous condition; this was because QTL that might be of great interest. For this reason it of elimination of male gametes carrying the L. pmnellii allele has not been possible to map QTL for increased yield (ESHEDand ZAMIR, unpublished data). For the analysis described in this study, the 50 11,s were in tomato, and studies have focused instead on fruit both selfed and crossed with M82 and A8 (a different pro- characteristics, abiotic stress tolerance, insect resistance cessing inbred with larger fruit and higher contentof soluble or morphological variation of the young seedlings solids) (Table 1). Thefollowing genotypes were transplanted (NIENHUISet nl. 1987; PATERSONet al. 1988; MARTINet in the field in Akko on March 1993 in a completely randomized design: I,. prnnellii (eight plants),M82 (30 plants), AS (23 nl. 1989; DE VICENTEand TANKSLEY1993). plants), hybrids ofAS with M82 (20 plants), FI-interspecific To overcome the problem of interspecific mapping hybrids (20 plants), selfed progenies of each of the 49 homo- of QTL associated with yield, we modified population zygous ILs (six plants), hybrids of ILs withM82 (six plants structures used in the past by combining them with DNA each) and hybrids of ILs with A8 (six plants each). Hybrids markers. WEHRHAHNand (1965) have demon- heterozygous for the introgression IL8-1 were selected using strated that effects of individual QTL in wheat can be the isozyme marker Ap-2. Seedlings (35 days old) were transplanted into a drip-irrigated field with 50 cm between plants measured by using backcross inbred lines (BILs). BILs and 2 m between rows (1 m2 per plant). are characterized by the low proportion of the donor Two regions of the genome were selected for the fine map parent in each of the population members and therefore ping analysis in 1994. One region was covered by IL2-5 and are ideally suited for mapping interspecific variation. IL2-6 and was associated with small fruit size. The other was In this paper we describe the application of a novel contained in IL1-4 and was associated with high yield of total soluble solids (BY). Large F2s of a cross between these 11,s introgression line (IL) population,which resembles the with M82 were grown, and recombinants in the targeted re- BILs, forthe mapping of interspecific variation for gions were selected following RFLP analysis (BERNATZKYand quantitative traits associated with yield in tomato. The TANKSLEY1986) with the most distal markersdefining the IL population is comprised of 50 L. esculentum lines, introgressions. Recombinant plants were subjected to RFLP analysis withadditional markers available for the introgressed each containing a single homozygous restriction frag- segments to define their genotype. Recombinants of interest ment length polymorphism (RFLP)-defined chromo- were selfed again to obtain lines homozygous for shorter in- some segment, introduced from the green-fruited spe- trogressions. After lines for the fine mapping analysis were cies L. pennellii (ESHEDand ZAMIR 1994b). Among the fixed in a homozygous condition, their progeny were trans- lines there is acomplete representation of the wild planted in the field in Akko on March 1994 (10 plant? from each IL and 30 plants from M82 in a completely randomized species genome. The ILs are nearly isogenic to the re- design). For fruit mass mapping only selfed plants were cipient genotype, and therefore all the genetic variation grown, while for the BY analysis selfed progenies and hybrids that differentiate them can be associated with the intro- with MS2 were tested. gressed segment. Since each line carries only a small Phenotyping: Fruits of all lines were harvested when 95- fraction of the wild species genome, most of the fertility 100% of the tomatoes of MS2 were red (IO5 days after trans- problems can be eliminated and yield-associated traits planting in thefield). The following measurements were taken for each of the plants: weight of the vegetative part can be measured. We demonstrate that the alternative (PW), percentage green fruit weight at harvest time (G), and mapping population structure presentedhere could in- red fruit weight. Total fresh yield per plant (Y) included both crease the ability of geneticists to dissect a quantitative the red and green fruit. The experiment was harvested ac- trait by using saturated FWLP linkage maps. cording to the order of the planting in the rows and not based on maturity level. Under field conditions toward harvest time, everyday -3% of the fruit change their color from MATERIALSAND METHODS green to red. Therefore it is reasonable to expect that even Plant material: The parental lines for the IL population the late maturing genotypes would have reached complete were the processing tomatoinbred variety M82 (L.esculentum) maturity had the plants been harvested a few days later. For and the inbredaccession of L. pennellii (LA716). The develop- this reason our yield parameter for each of the genotypes ment scheme of the ILs, through repeated backcrossing and included both the red and green fruit weight. Total soluble RFLP selection, was described previously by ESHEDand ZAMIR solids concentration - Brix (B) represents mainly the soluble (1994b). The IL population is composed of 50 L. esculentum sugars and acid concentration in the fruit and is a standard (x = 12) lines, each containing a single RFLPdefined chro- quality parameter for the processing tomato industry. B was mosome segment of L. pennellii (Figure I).The lines contain assayed on a sample of 20 red fruits per plant (measured an average of 33 cM from a total genome size of 1200 cM using the digital refractometer RFM-80 BS), and mean fruit (2.75%); overlapping regionsbetween neighboring lines were mass (FM) was calculated from all the red fruits. The product selected to ensure complete representation of the wild species of Y and B provides an estimate of the grams of soluble solids genome. Thetotal length of the overlaps is 480 cM. Determi- produced per plant (BY). nation of the size and identity of introgressed segments was Statistical analysis: Statistical analyses were performed on based on RFLP analysis of 375 markers chosen to cover the the JMP V.3.1 software package for Macintosh (SAS Institute entire tomato genetic map at minimal intervals. Our claim 1994). Mean values of the parameters measured for thetested that the lines contain a single introgression is based on the genotypes were compared to the appropriate control geno- RFLP analysis, which in nocase revealed additional indepen- types using the “Fit Y by X’ function and “Compare with dent introgressions. The purity of the ILs is also supported control” with an alpha level of 0.05 (DUNNET 1955).M82 was by fingerprinting of the lines with multicopy microsatellite the common control for the ILs and their hybrids with M82; probes ( ESHEDand ZAMIR, unpublished data). It should be M82 X AS was the common control for thehybrids of the ILs noted that the introgression of IL8-1 was the only one that with A8. The additive effect (a) was half of the difference Introgression Lines for QTL Mapping 1149

Plant % Green Fruit YieldBrix Brix weight fruit mass (Y)(B) x Yield (P W ) weight (FM) (BY) (FM) weight (PW) (GI

1 10 cM CT233 Pnl E301 0 TG24

c CD15 c CTl22A rl TG334 TGBO Skdh 1 TG196 TG310 TG70 CT91 B:TG224 Y TG5 9 - TU3 150~d -17.1 250~0.5 40 11.8 40 9.6 50 59.3* 75 TG71 p rl CTl65 3- d TG335 T TGI 97 r -1 50 -250 -40 40 -50 -75 CT224A a 57.7 83.4 -1 3.4 10.6 -1-6.2 3.2 d TG161 -5.0 TG237 19.9 7.9 -6.1 27.1 14.9 27.1 -6.1d 7.9 19.9 TG245 TG430 TG85 TG617:TG17 TGS3 f TGZSS;TG267 c TG158 TG258A;TG389 d TG159 TGZ 7 25.9 44.3a 25.9 15.4-1 .5 6.2 9.4 TG259 7.1 6.7 8.6 24.4 34.1 24.4 d 8.67.0 6.7 -1 7.1 FIGURE1.-The chromosomal location, size, identity and phenotypic effects of the 50 L. pennellii ILs. The genetic map was constructed on thebasis of 119 BC1 plants as described by ESHEDet al. (1992). Mapped markersare connectedto the chromosome with a horizontal line, and markers not assayed on the BCl map are placed according to theirapproximate positions according to TANKSLEXet al. (1992). Each line was probed with all the markers, and the ones showing the wild species alleles are marked by the bar to the left of the chromosome. The phenotypic difference (as percentage of control) for each of the ILs and their hybrids is given for the following traits: PW, plant weight; G, percentage green fruit at harvest time; FM, fruit mass; B, total soluble solids - Brix; Y, yield; BY, Brix X yield. For each trait the left bar represents the relative performance of the ILs (p/p), the central bar shows the effects of hybrids with M82 (e/p), and the bar at the right is the relative performance of the hybrid with the tester A8 (e’/p) (the control isM82 X AS). Bars in black indicate significant differences exceeding p < 0.05, and empty bars indicate nonsignificant differences. The following components of genetic variability for each IL X trait are presented as percentage of control (M82). The additive effect (a) is half of the difference between the IL and M82, and its significance was calculated on the basis of the comparison between them. The dominance deviation (d)is the difference between IL X M82 and the midvalue of its parents. Significant d values at the p < 0.05 are marked by *. Significant overdominance at the p < 0.05 level is marked by an underline of the d value. between each IL and M82, and its significance level was deter- 2) was calculated on the basis of the following assumptions: mined by the comparison between the IL and M82. The domi- 1) each IL affecting the trait carries only a single QTL, 2) nance deviation (d) and its significance were calculated by two overlapping introgressionswith a significant effect on the contrasting the IL X M82 (+1) with M82 (-0.5) andthe trait (in the same direction relative to the control) carry the appropriate IL (-0.5). The threshhold level for significant d same QTLand 3) a QTLis counted only if the IL or its hybrid values was 0.001, which provides an experiment-wise error of is significantly different from the corresponding control, re- 0.05 for the 50 compared genotypes. All calculations were gardless of the significance of a and d. The mean degree of performed with the measured values, except for G, where dominance for eachtrait (d/[a])(Table 2) was calculated square root transformation was evaluated to improve normal- from the mean dominance deviation for all ILs divided by ity. Results arepresented as the percent difference (A%) the mean additive effect. from isogenic control. Interaction with genetic background For the 1994 trial, a multiple range test between the lines (hybrids of ILs with M82 and with AS) was determined for evaluated for the finer mappingof the introgression of inter- each introgression by two way analysis of variance with signifi- est was performed by the “Fit Y by X’ function and “Each cance threshhold of 0.001. The coefficient of variation (CY) pair comparison” with an alphalevel of0.05. For comparisons for each trait was calculated by dividing the general (over all of the lines derived from ILl-4, the inbredILs and thehybrids tested genotypes) SD by the general mean. The minimum of ILs X M82 were compared separately. M82 was included number of p < 0.05 significant QTL affecting a trait (Table in both comparisons as a control. 1150 Y. Eshed and D. Zamir

PW G FM B Y BY - 15, 250b 40b40b '5b

ni d -1 50 -250 50 -1 40 40 -50 -75 a -23.8 -38.0 -1 2.9 -6.4 -1 7.0 -21.5 d -9.8 50F -1 7.5 2 150 R45S TG31 TGZll CTlO6A TG276 I -1 50 -50 CT638 a -9.1 0.6 4.3 -0.6 -4.3 -5.6 TMl2 2 TG1 8 c1 CT255 c 406 TG165 4Op TG554 8 rc 0 40 c1 a -15.8 48.9 -1 0.2 6.5 50p-16.5 -1 2.4 c d 19.0 12.2 i'?F ?2;TG353 TG308 Q) TG454 0 TG493 L TG191 Q) CD9C e TU6 TGI 31 B -50 CT75;TG426 a 26.6 90.6 40b-3.3 4Of 0.9 50F1.4 1.3 TG48 a CT9 :7F TG537 TG34 TG91;CD66

ii;O;TG50B TG361

40 -50 40L-28.7 4Of 13.610.3 -1.3 TG141A TG154 -41.0" -34.0

". 40 a 63.6 75.0 -20.2 11.7 11.1 23.6 d -3.4 10.3 8.7 2.4 -1 2.8 10.3 FIGURE1.- -Continued

RESULTS respectively), while plant weight of their hybrid was seven times higher than that of the parents. Phenotype.- of parental species and their F1 hybrid The parental species and their interspecific hy- Another factor that might influence PW is growth brid were characterized for thequantitative traits under habit: M82 is a determinate plant due to homozygosity investigation. Highly significant differences between L. for the recessive mutation sp (self-pruning) , whereas L. esculentum and the F1 hybrid were found for all mea- pennellii and the FI are indeterminate. L. pennellii was sured traits except BY (Table l). The vigor of the F1 also compared to an indeterminate male-sterile L. escu- hybrid was most striking for PW, which was 24 times lentum line (Sp/Sp) for evaluation of the effect of the higher than for M82 and eight times higher than for mutation sp on plant growth. The sterile indeterminate L. pennellii. line was more than twice as large as the sterile M82 or L. pennellii did not set fruitunder thefield conditions L. pennellii. PW (both fresh and dry) of the F1 hybrid where the Y of the hybrid was 50% of that of M82; this between the wild species and the indeterminate Id. escu- difference in fruit set could have affected PW, since lentum line was 3.7 times higher than thatof the indeter- there is competition for assimilates between the vegeta- minate sterile L. esculentum parent. This result indicates tive and the reproductive organs of the plant. To elimi- that only a part of the heterotic effect observed for nate the effect of variation for fruit set from the study growth rate of the vegetative parts can be attributed of vegetative growth, L. pennellii was compared in the to the differences in fruit set ability and growth habit following year to a male-sterile isoline of M82; their between the two species. plant weights were verysimilar (8.8 and 7.1 kg per plant, Large differences in FM, B and G between the two Introgression Lines for QTL Mapping 1151

PW G FM B Y BY

-2.5 8.1 -2.2 4.5 10.0-2.2 101.58.1 -2.5 d -21.9 -72.0 4.8 -1 4.3 -1 1.9

a 6.5 41.8 10.3-1 2.8 -1 1.0 4.1 1509"d -28.1 25Op-56.1 40b-3.9 40~ 1.2 50b-11.6 75b -11.4 TGt88;TGSOC

0 0 1

-1 5 -250 -40 -40 -50 -75 a91.6 55.7 -1 1 .8 8.3 -27.8-30.5 7.5 4.1 26.7 33.1 26.7 4.1 7.5 40

-250 147.9 40 TG244;TG94 25F-41.7-1 8.8 2.7 8.1 d11.9 -39.1 41.6. 8.039.5' -3.7 mP 1501 250~40b 40~ 5Ob75k

d -1 50 -250 -40 -40 -50 -75 a 3.8 68.8 -5.1 7.1 -5.0 5.4 -8.2 7.9 6.7 10.5 d 3.6 6.7 -3.07.9 -8.2 FIGURE1 .- Continued parental species can be inferred from a comparison of both lines carry the S (self-incompatibility) (LIEDLet L. esculentum with the interspecific F, hybrid. The wild al. 1993) locus, which in a homozygous state produces species was responsible for lower FM, higher B and almost sterile plants of very strong vegetative growth. higher G (Table 1). IL6-2 and IL6-3 are both of indeterminate growth habit Phenotypicanalysis of the IL population: The IL due to the allele Sp originating from L. pennellii; how- population did not appearmarkedly different from the ever, PW of IL6-2 was 58% lower than that of the con- cultivated varieties (Table 1).The plants of the ILs were trol, while PW of IL6-3 was 1'73% higher. Since PWs of bigger, with later ripening,lower fruit mass, higher solu- the hybrids of both ILs were much higher than that of ble solids content (Brix) and lower yield than the con- the control, we conclude that the difference between trol. When tested as hybrids, BY was higher than the the two is due to homozygosity in IL6-2 for therecessive controls, while for the other measured traits they were mutation ndw (necrotic dwarf) (WEIDEet al. 1993). generally intermediate between the inbred ILs and the In 22 ILs there was a significant additive effect (a) controls. associated with an increase of PW, whereas onlytwo ILs Plant weight (PW): Of the 49 ILs presented in Figure were associated with a significant decrease in PW. The 1, 24 had a significantly different PW at harvest date two indeterminate ILs (IL6-2 and IL6-3) had a signifi- compared to M82 (Table 2). IL1-1 had the highest PW cantly increased dominance deviation, of which one in the experiment, i.e., 209% higher than M82, while (IL6-2) was overdominant. Overall, PW showed additive IL1-1 X M82 had a 100% increase relative to M82. In inheritance (d/ [ a] = 0.06), and the minimal number the A8 X M82 genetic background the same introgres- of QTL affecting the trait was 16. sion had a very similar effect, contributing to a 91% Percentage green fruit weight (G): Processing tomato va- increase in plant weight (Figure 1). IL1-1 and IL1-2 had rieties were developed for a single harvest that takes very similar effects on plant weight aswell as other traits place when 95-100% of the fruits are mature red.Since discussed below, suggesting that these QTL reside in the ILs were harvested together, G represents the rela- their overlapping segment. It is interesting to note that tive earliness of the different genotypes. G in the ILs 1152 Y. Eshed and D. Zamir

PW G FM B Y BY

cT1 ZZC d -15 -250 27.5 4040t-= -2.3 40 4.2 -50 -7571- 3.0 CT63C I=-I 7.3 25k "I- -0.6 GP180 COS9 d 3.5 -37.4 -3.6 4.6 3.9 0.8 7649

TG41 $8 TG123 d TG146A;TE483 TG1 62 -1l5I" 5 -250 -40.1 40 -4.8 40 0.3 -7571- 5.4 TG1468 a -2.7 "I- 6.3 "2 d -1.0 25b-7.4 2.7 "t- 4.8 1.7 -0.4 TGZOB TG272 Adhl TGZ84B TG6L TG65-CT188 TGs7i TG305 -J -1l5I-5 a 6.0 -2502501" -9.1 40 40 2.3 -50 -0.9 -7571" 0.2 TG155 3.6 "I- 5k TG345 CT5 0 d 7.7 -32.7 -3.2 2.3 10.3 1 1.7 CT133 CT7 3 TG22;CT1266 CT2248;TG37 16464 d -1 5 -250 40 -40 -50 -75 a 31.9 101.4 -3.1 15.5 -9.1 2.5 d -9.8 48.5 12.2 4.1 25.7 33.8' FIGURE1. - Continued showed a bi nomial distribution and therefore a square higher Y. Unlike thLe previous traits, Y of thLe hybrids root transformation was performed. Significantly late could not be predicted from the Y of their inbred IL ripening characterized 33 of the 49 ILs; only IL2-1 and parents. The ILs with the lowest Y values in a homozy- IL42 ripened significantly earlier than the control.Two gous condition (ILl-1, IL1-2,IL3-3, IL3-4, IL6-2 and ILs (IL 6-2 and IL 7-2) showed a significant dominance IL7-2) gave rise to hybrids with Y values equal or supe- deviation but none was overdominant. G was generally rior to those of controls. Seven ILs with significant dom- inherited in an additive mannerand the minimum inance deviation for increased Y were identified. Three number of significant QTL was estimated as 22. of these (ILl-1, IL5-4, and IL7-3) showed significant Fruit mass (FM): In 22 ILs, FM was lower than in the overdominance, indicating that the hybrid outyielded control. However, 1L7-5 and IL12-1 contained genes its highest parent. Hybrids of ILI-4, IL6-3, IL7-5 and originating from L. pennellii, which increased FM, indi- IL9-2 had higher Y values than M82 but not from their cating transgressive segregation. The smallest fruit size homozygous IL parent. The minimum number of QTL was measured in IL6-2 (homozygous for ndw); this was for Y was 11, and the mode of inheritance was largely followed by IL2-5, which was subjected to fine mapping overdominant (d/[a] = 2.16). (discussed later). The L. esculentum alleles, generally Total yield X Brix (BY): In tomatoes there is a negative producing higher FM, showed partial dominance over relationship between total fruit yield and soluble solids the L. pennellii alleles (d/[a] = 0.34). The minimum concentration (STEVENSand RUDICH1978). Therefore, number of QTL estimated for this trait was 18. the parameter BY provides both a biological and an Total soluble solids concentration - Brix (B): In 31 ILs agricultural estimate for the productivity of the plant. the B values were significantly higher than in the con- To obtain a better estimate of QTL affecting the horti- trol, and only IL2-1 had a lower B value than M82. The cultural yield, we analyzed the derived product BY. This highest B values were found for IL6-2 and IL6-3, which product gives an estimate for the g of soluble solids were indeterminate. These two lines differ from M82 produced per plant. IL2-6, IL6-3, IL7-5, IL9-2 and ILl1- in their growth habit (indeterminate 11s. determinate), 2 had significantly higher BY than M82. All of these and this is known to have a major effect on B (EMERY lines were also distinguished from M82 in having larger and MUNGER1970). The minimum number of QTL and later maturing plants so their advantage can be affecting B was 23, with generally partial dominance to explained by the larger source and the longer time of the L. pennellii alleles for higher B values. assimilate production. The same was true for the hy- Total fmit yield (Y): Nine homozygous ILs had signifi- brids of IL1-1, IL1-2, IL5-4, IL6-2 and IL7-3, all of which cantly lower Y values than M82, while only IL7-5 had a showed significant overdominance for this trait. IL2-1 Introgression Lines for QTL Mapping 1153

PW G FM B Y BY 150

d -1 50 -50 5.4 6.9 4.6 5.9 1.6 0.7 1.6 5.9 a 4.6 ,i- dt3" 2.72.2 -1.89 4.2 5b-1.6 -0.9 150

d -1 50 5.6 82.3 1.4 1.1 -14.0 1.1 1.4 82.3a 5.6 -1 4.3 t":Kk=:b-24.8 28.1 150dkml~41p4ci~ 7.8

d -1 5 -50501- -5.2 -3.3 a3.2 1.6 -2.5 58.4 .i" d -13.2 13.2 3.7 8.2 -0.6 -1.9 150

d -1 50 -50 -1 5.1 -0.7 a 75.7 180.2 -3.5 14.8 ::v d 1.6 27.2 4.3 0.2 150 50v

d x -150 -50 10.59.1 9.7 6.8 37.8 -7.0 37.8a 6.8 1.8 s°F d Kf-.4lk=:k- 16.9 12.0 3.4 10.7 3.4 12.0 5.8 :F FIGURE1.- -Continued was the only line for which the heterozygous was infe- respectively). The region that spans TGI 91 to TG426 is rior to the control. At least 14 QTL were detected for shared between these lines and we therefore assigned BY, and the degree of dominance for this trait was the Fm2-1 to this interval, IL2-5-1 and IL2-5-4 contain longer highest (d/[a]= 10.26) (Table 2) of all the measured segments of the L. pennellii chromosome, extending to traits. TG91 and TG167, respectively. These lines exhibited a Fine mapping of QTL: Fine mapping of a FM QTL 50% reduction in FM, which indicated the existence of (chromosome 2): In the 1993 experiment, IL2-5 and IL2- an additional QTL close to Fm2-1. The finer position 6 showed a considerable reduction (40-60%) in FM of this QTL, designated Fm2-2, could be inferred from relative to the control; this effect was attributed to a therecombined progenies ofIL2-6. IL2-6-1 hadthe single QTL that is overlapped by the two introgressions same fruit mass as IL2-6 and IL2-5-5, while IL2-6-2 had (Figure 1). To achieve finer mapping of this QTL in larger fruit. Only TG91 is shared by the first three of 1994, we conducted RFLP analysis in second and third these lines and not by the fourth, so Fm2-2 is tightly generation progenies of IL2-5 and IL2-6 crosses with linked to TG91. The position of the third QTL, Fm2-3, M82; 12 lines with different subsets of introgressed re- was deduced from the regions covered by IL2-6-3 and gion were evaluated for FM (Figure 2). The effect of IL2-6-4 (with 25% reduction in FM) but not by IL2-6-5 the entire introgression of IL2-5pwas similar to that in (with FM similar to the control). Fm2-3 was placed be- the 1993 experiment (65% reduction), while its true tween TGSOB and CT59. isogenic line IL2-5" (plants from F2 of IL2-5 X M82, The mapposition of a QTL was located to an interval selected for the cultivated tomato genotype) was not between flanking markers that were not introgressed to different from the control M82. all the ILs carrying this QTL. According to the map The IL2-5, IL2-6 region of L. pennellii appears to har- (TANKSLEYet al. 1992) Fm2-1 (CD9C-CT75) is defined bor three different QTL for FM (Figure 2). None of by an interval of 3.2 cM, Fm2-2 (cD66- TG204) by 3.7 the recombined progenies of IL2-5retained the magni- cM and Fm2-3 (TG151- TG14la) by 14.1 cM. Recombi- tude of its effect on FM reduction. Two lines, IL2-5-2 nation frequencies detected for these intervals in the and IL2-5-3, had similar effects on FM (28 and 30%, IL crosses weregenerally three to four times lowerthan 1154 Y. Eshed and D. Zamir

PW G FM B Y BY

"b- 6 GP164 TG61 ZB -40 -4.6 -50 -9.0 -75 I TG29RCr216 5.2 50b-5.4 -1 19 TGZS1;Ap.I 4k10.3 -1.2 11.8 9.4 TG97 40 50 75 ;$:&GI 53 TGZ5;ndw TG24eTG406A TG54 CD25A TG446 -40 -40 16.2 -50 70.7' -75 TM4 "P I!!TGZSS d -108.0' 18.3' irp. CTSOB TG292 CTllOA TG27S 3 rfl;TG279;ClZ06 ad 150p;=;kd TG548 75f! -7575f! 26.5 TG99 -1 50 -40 0.6 -40 25.1 -50 2.0 PC5 a 50.383.7 "I" 50P TG422;TG581 40P TG11S d -2.148.0" -0.3 1 1.8. 24.8 43.7* TGl93 TG221 TGZ58B 75 1

-40 1.9 -4040t- 1.4 -50 -75 a 25.412.7 "I- 5.0 5.4 3.5 5.2d 3.5 0.1 4.4 0.3 3.2 FIGURE1.- Continued those presented in the map; reduced recombination trait. Three derived lines, IL1-44, IL1-45 and IL1-46, rates have been previously reported for certain L. pen- retainedtheir significant effectsover control when nellii introgressions (RICK 1969). tested as hybrids. All three share the introgression from Fine mapping of a BY QTL (chromosome 1): In a previous TG267 to TGI58, which was not covered by IL1-47 or study, we reported that a 37cMintrogression of chromo- any of the other recombined ILs. In contrast to IL1-4P, some 1 from L. pennellii (IL14) improved the total solu- where the hybrid had higher BY values than those of ble solids yield oftomato hybrids in plotsby 16% (ESHED the inbred IL and the control, the lines IL1-4-4, IL1-4 and ZAMIR 1994a). This introgression in a heterozygous 5 and IL1-46 had higher BY values than the hybrids condition was effective over a 2-year period and for two and the control M82. This indicates a partially domi- genetic backgrounds. The same effect of this introgres- nant mode of inheritance for the higher BY QTL in sion was observed in the 1993 experiment (Figure 1). the shortened introgressions. The inconsistency in gene The improvement in BY was most apparent in the hy- action between the complete and shortened segments brids of IL1-4; for the fine mapping we used a similar could be explained if an additional, partially dominant strategy to that described above for FM but included the QTL near TG245 reduces BY in a homozygous condi- heterozygous lines in the analysis (Figure 3). tion. Indeed, IL1-41 and IL1-43 had lower BY values Seven lines derived from IL1-4 and their hybrids with than the control (ILl-4'); however, the BY value of IL1- M82 were measured for B and Y. Overall, the pheno- 42 was similar to that of the control. If such a QTL typic effect of the introgressed segments on the mea- exists, it has a minor effect and should be tested in sured quantitative traits was smaller than the effects of more replicates than the 10 examined here. the FM QTL. In addition, the phenotypic variation for The conclusive result of this fine mapping attempt is Y was much higher than for FM. For these reasons, the that the increase in BY is associated with a 12-cM seg- results of the finer mapping are less conclusive. The ment of L. pennellii between TG53 and TG258A. The effect of the entire introgression on BY was lower than advantage of the different lines and their hybrids over the 49% increase found in 1993 (Figure 1);the hybrid M82 with respect to BY was an increase in either B or of IL1-4P X M82 was 31% higher than the control,while Y, but no consistent pattern could be observed (data its true isogenic line, IL1-4', derived from the F2 genera- not presented). tion of IL1-4 but selected against the introgressed segment, was identical withM82. IL1-4 had a BY value DISCUSSION similar to that of the control butlower than that of the QTL identification using the ILs: Of the 50 ILs heterozygous hybrid, indicating overdominance for the tested, eight had no significant effect on any of the Introgression Lines for QTL Mapping 1155

PW G FM B Y BY 150

r 7 -1 50 a -8.0 8.5 -10.6 3.1 -1 1.7 -1 0.1 150~d 16.4 11.7250p 4Ob15.1 4p 4.6 6.9 75k11.3 2 a -1 50 -250 -40 40 -75 -50 a 65.4 130.3 -1 6.5 -21.317.8 -1 0.4

40 ~ 150pd -23.3 p ;p23.5 75p d

-1 50 40 65.7 162.4 4.9 a162.4 65.7 9.3 -1 1 .8 -5.0 15.4 48.4 1.7 150~ 48.4d 15.4 40b250~ 8.1 Alz52Li75k

I? a

-1 50 -250 -40 -40 -50 -75 0.2 42.4a 0.2 5.8 5.5-2.9 -7.3 29.5 42.1d 29.5 -0.5 40 1": WF 75F

-40 3.3 16.7 3.3 19.3 -2.9 23.1 2.4 23.1d -2.9 16.6 3.6 20.1

obc 1501 250~401 40k 75b

I 0 L w d c 50 -40 0 -1 -250 -40 -50 -75 4.3 1.4 -0.3 14.4 -8.4 2.4 -8.4 14.40 d(Hyb-MB2)-0.3 1.4 4.3 + 150 0 c, C= 2Q) -1sa i-_p..14.5 103.1 II--3.4 :I- 3.4 .i.--8.5 :p--4.1 I TG529 Q) e ~ 150d k;mJ+9 -0.9 :p 4oF SJF 75F d

-1 50 068 27.0 108.0 -7.0 9.1 -3.0 4.3 -3.0 9.1 -7.0 108.0a 27.0 3.5 62.2 -7.9 -2.8 2.6 -2.8 -7.9 62.2d 3.5 -0.8 FIGURE1. - Continued traits measured. These ILs cover a total of 203 cM FM and B showed relat ive ly low environmental varia- (17%) of the wild species genome. The 294 (49 X 6) tion (low CV) (Table 1) and can be compared to previ- homozygous IL X trait combinations yielded 135 that ous QTL findings in tomato where a similar experi- were significant at the p < 0.05 level. Of the 300 (50 ment-wise error was used (5%).PW, G, Y and BY showed X 6) heterozygous IL X trait combinations measured, higher environmental variation and were assayed here 67 were significant at thep < 0.05 level in both genetic in tomato for the first time using a complete linkage backgrounds while 36 were significant in one back- map. In spite of the very small number of replicates ground only. (six) used to determine the phenotypic value of each 1156 Y. Eshed and D. Zamir

9 CTZ25A PW G FM B Y BY - TG254 6l TG105B d tg18 TGS PI TG225 TGlO Y 0) cd52a CTZl 5A -3.1 -0.5 d -3.1 0.5 -2.0-0.6 7.6 d CTZ158 rcO? ISO~ 250~ 413~ 40p50p75y 0215C TGlOl;TG5918 TO1 SO ul TGSSO u tg55 e= TGl44;TG186 TG248 -150 -250 -40 40 -50 -75 cT74 a 20.9 65.9 -6.2 12.0 10.2 22.0 CT198 Q)E 20.3 d 5.0 7.1 -27.314.7. 1.6 ctz18 2 76421 d TG8 TG424 CTl 1 ZA TG328;TG591A d -19.67.2 -54.04.6 2.1 11.1 CTZZO

TG230ST113C TG31 S CAB7 r 150~250b 40 40p 5Ok 75b TG395 c z TG3OS;u r I TG596 d CD45 E' CTSlA -150 -250 -40 40 -50 -75 TG43 a 20.5 20.2 -12.5 12.7 -1 1.6 -5.3 tg52 c CTl Z6A E d 4.1 -21.6 -2.0 0.3 25.3 31 .O TGlOS -7w 150~ 250~ 40k 40f 75k TG408 TGZ85:CT26OC CTll 28 0: TG420 CD34B +rd x CP.0 - CD34A = -150 -250 -40 40 -50 -75 p! a 20.0 73.2 5.7 3.4 -1.4 0.5 d TGlA;CTS7 U TG241 k d 1.9 5.943.4 4.4 -0.6 2.1 TG229

CTSL ?z ad CT240 150~250~ 40 4015ob 7s~ z tg65 -1 50 -250 -40 40 -50 -75 c a -1.7 9.2 -3.2 3.1 -9.1 -7.2 TGZ06A 4 d3.4 6.6 -0.9 -1 9.4 16.5 19.4 CD32B TG233

FIGURE1 .- Continued

IL and its hybrids, 18 QTL for FM and 23 QTL for B The advantage of the ILs in identification of QTL were identified. In previous studies of the inheritance cannot be attributed to the phenotype of the parental of FM and B, using a complete linkage map in a BC1 lines used to generate them because the other experi- generation derived from a cross with L. chmielewskii, six ments also used interspecific crosses with similarpheno- QTL were identified for FM and four forB (PATERSON typic differences between parents. One factor contribut- et al. 1988). Mapping studies involving L. cheesmanii in- ing to the high efficiency of QTL identification using terspecific F2 and FB generations revealed 13 QTL for the ILs is the minimal overshadowing effect.All studies FM and seven for B (PATERSONet al. 1991). In a recom- of intraspecific (EMERYand MUNGER1970) and interspe- binant inbred population derived from the same cross, cific crosses involvingL. chmiehskii, L. chesmanii and L. 13 QTL were identified for FM and for B (GOLDMANet pennellii identified a major QTL for B on chromosome al. 1995). 6 linked to sp; the indeterminate plants always had Introgression Lines for QTL Mapping 1157

PW G FM B Y BY 150 71

d -1 50 40*i.. -1 3.2 4040k= 6.5 -50 -2.4 0.6 31.2 86.5a 31.2 ,t- d -10.9 25.4 150 5oF;;ir;

d -1 50 40 -0.5 40 13.3 -50 a 36.9 100.5 ")lb: 40F3.0 16.7 dPlP -10.1 13.0 150

a -1 50 40 -8.4 40 7.3 -50 ak:P 9.5 68.0 -6.3 -1.4 d 7.9 40.0 -1.7 -0.1 15.3 17.3 150 aP'F

=! -150 40 -8.2 4040P- 1 1.4 -50 i-l- 50k4.7 4.4 a 2.7 19.4 40k I- -8.2 35.8d -8.2 4.4 -2.8 -3.0-1 0.4

TG180 Aco 1 i501" i501" -d -1 50a -3.5 -250250t- -2.7 -40 12.5 -40 TG68 5.7 -4.8 -0.7 ' TGZ63A 2.6 9.229.2 2.6 13.4 22.9 CTl 9 d 23.4 "P CT1 ZOB 40 c179 TGZ63B

CnllA TGZBB =!i50t--1 a 5.050 ppr1 -250250t- 10.4 40 CMB 4.0 9.3 -0.4 7. 5 TGSOA -3.3 -3.0 6.3 17.5 25.8 TG111 d 3.3 TG565 40 bP@dhZ ESC

CT276

- TG473;TGbOZ =! CDZ -1 50 -250250t- -6.5 z150ka -15.9 4.9 2.9 -1 0.7 -9.1 d -7.6 -1 .2 4.4 0.2 -6.8 -5.3 FIGURE1 .- Continued higher B values. Such a gene contributes to large pheno- generated on the basis of the independent markers and typic variation when the effect of an unlinked marker a gene-induced elevation of the mean B value in the on B is examined in a segregating population. The over- population. Since variation is usually correlated with the shadowing effect can result from two main factors: an mean of a measured trait, the existance of high popula- unequal distribution of sp in the subgroups of genotypes tion means will reduce the possibility of detecting QTL 1158 Y. Eshed and D. Zamir

TABLE 1 Mean phenotypic values for the tested genotypes

Percentage No. of Plant weight greenFruit fruit Brix mass X yield Genotype replicates (kg) weight (g) Brix Yield (kg) (g) L. pennellii 8 4.4 t 1.7 NA NA NA 0 0 M82 X I,. pennellii 36.920 i- 11.2** 74.9 i- 13.0** 5.4 2 0.8** 10.3 i- 1.3** 4.0 t 0.7** 415 t 91 M82 30 1.5 t 0.5 3.9 t 4.0 58.2 t 7.1 4.3 ? 0.2 8.2 t 1.7 361 t 82 M82 30 1.5 t 0.5 3.9 -+ 4.0 58.2 t 7.1 4.3 t 0.2 8.2 +- 1.7 361 t 82 M82 X A8 20 1.5 ? 0.3 3.1 5 2.5 64.7 i- 8.1"" 4.7 ? 0.3* 8.6 -+ 1.5 396 ? 66 A8 23 1.6 t 0.5 6.8 t 5.0 73.0 -+ 8.0"" 5.0 t 0.4** 7.0 5 1.6" 351 i- 89 49 inbred ILs 6 2.2 t 0.9** 14.4 t- 18.5"" 50.9 t 11.0** 4.9 t 0.5** 7.0 2 2.2** 347 t 115 M82 X 50 Ls 6 1.8 i- 0.5* 8.9 i- 6.5** 56.0 f 6.0 4.7 f 0.4"" 8.9 t 1.2 424 -t 76** A8 X 50 1Ls 6 1.9 t 0.6** 7.4 2 5.3** 60.7 i- 6.8* 4.9 t 0.4* 9.2 f 1.4 461 t 96** C.V." 22.225.3 32.5" 21.2 11.5 6.9 Mean phenotypic values and standarddeviations of three genotypic groups are presented. Thegenotypic groups areas follows: the parental species, L. pennellii, L. esculentum (cv M82) and their interspecific hybrid; the L. esculentum inbred lines M82, A8 and their hybrid; the homozygous introgression lines (ILs) and their hybrids with M82 and A8. All means were compared to M82 except for A8 X 50 ILs that were compared to M82 X AS. Means marked with *, ** differ significantly (t test: p < 0.05, p < 0.01 respectively). NA, data not available because the plants did not set fruit. Coefficient of variation (%) was calculated on the basis of all genotypes. ' Coefficient of variation (%) was calculated after a square root transformation. with milder effects. An additional factor that might ac- account for QTL and may be an important source of count for the large number ofQTL detected in this trangressive variation. study is the exposure of novel variation not detected in The QTL for PW and G had the largest effects of conventional segregating populations. The two examples all the traits measured. These parameters allowed the of such variation are the male gamete eliminator (IL8- identification of 16 and 22 QTL, respectively, despite 1) and the ndw mutation (IL 62).These genes have not the large environmental variation. Yield is among the been previously reported for any of the FTsegregating most difficult traits to manipulate genetically; not only populations of L. esculentum X L. pennellii, although this does every metabolic pathway ultimately affect repro- cross was used extensively in genetic studies. Only when duction, but the environment is also a major determi- introgressed into the cultivated background, without ad- nant of yield. Genetic analysis of yield in interspecific ditional wild species chromosome segments, were these crosses is often affected by some overshadowing QTL genes identified (WEIDEet al. 1993). Such an inheritance associated with partial sterility. In this study we identi- pattern is probably due to other unlinked epistatic genes fied at least four such QTL, in which homozygosity for from L. pennellii. The same mode of inheritance can the wild species allele resulted in almost complete steril-

TABLE 2 The number of significant effects (p< 0.05) of L. pennellii introgressions on the components of genetic variation for the quantitative traits studied

Genetic Plant Genetic % Green Fruit Brix Total Brix X components" weight fruit weight mass (TW yield fruit yield a+ 22 33 2 31 1 5 a- 2 2 20 1 9 7 d+ 2 0 0 3 7 8 d- 0 2 0 0 0 0 od 1 0 0 0 2 5 Mean d/ [a]" 0.06 0.06 0.34 0.45 2.16 10.26 Minimal number of QTL 16 22 18 23 11 14 "The genetic components of the variations are as follows: additive (a), dominance deviation (d), and overdominance (od). Values higher relative to the control genotype are indicated by + the values lower than the control are indicated by -. The degree of dominance (mean d/[a])was calculated on the basis of all ILs, regardless of significance of d and a.

I The minimal number of significant QTL affecting a traitwas calculated on thebasis of the assumptions described inMATERIALS AND METHODS. for QTL MappingIntrogression QTL Lines for 1159

FM in A 70 from line M82

IL2-!je -4.5 a 112-5P -64.9 d II 112-5-1 -49.5 c 112-5-2 -29.7 b lL2-5-3 -27.6 b 112-5-4 -49.3 c IL2-5-5 -53.7 c IL2-6 -51.1 c IL2-6-1 -54.4 c lL2-6-2 -25.0 b 112-6-3 -25.4 b 112-6-4 -23.3 b 112-6-5 -2.6 a M82 (VALUE g) 61.6 a

FIGURE2.-Fine mapping of QTL for fruit mass (FM) on the distal end of the long arm of chromosome 2. The only L. pennellii chromosome segments introgressedinto the listed lines are marked by the dark bars. ILs 2-5P and 2-5’ are truly isogenic and are derived from an FYof IL2-5 crossed to M82 and selected by RFLPs to be homozygous for the L. pennellii alleles (l’) or the cultivated tomato alleles (‘.). For each line, the phenotypic value (presented as percentage difference from M82, A%) based on 10 replications was determined for the homozygous recombined ILs. Multiple range test by “Each pair comparison” with an alpha level of 0.05 was performed between the lines; different lettersdenote significant differences. The smallest chromosome sement that retains a significant effect on FM relative to the isogenic lines is marked by two vertical lines and is the postulated position of the FM QTL”(l;m). ity (on chromosomes 1, 3, 6 and 7). If we exclude IL6 additive inheritance. General additivity was also de- 2 (because it carries ndzu), the remaining three QTL tected for G. However, effects contrary to expectations can account for partial sterility in >50% of the plants based on parental line phenotype were seen for IL2-1, of the F2 generation [l - (3/4)’]. The IL population which had small plants and early fruit set. In contrast revealed at least 11 significant QTL for yield, of which to the other measured traits, Y was strongly associated seven L. pennellii segments wereassociated with in- with overdominance. Significant dominant deviation creased yield. The effect of selected chromosome seg- for yield was detected in seven of the ILs and was always ments from L. pennellii has been previously reported associated with increased yield. A similar pattern of re- (DE VICENTEand TANKSLEY1993; ESHED and ZAMIR sults was obtained in an experiment designed to identify 1994a); as in the present study, introgressed regions QTL for yield in an intraspecific cross between two elite near the centromeres of chromosomes I and 7 and inbreds of maize (STUBERet al. 1992). from the long armsof chromosomes I and 5 were sig- The most striking feature of the interspecific hybrid nificantly associated with increased yield. was its vigor for PW (Table 1). Dissection of the wild Gene action of detected QTL: The gene action of genome into small independent fragments revealedthat detected QTL was determined by comparing thehomo- only the indeterminate lines, ILG2 and IL63, exhibited zygous ILto its hybrid in the same genetic background. positive dominance deviation for PW, these lines can ex- The inheritance modes of the QTL for FM and for B plain only a small fraction of the heterotic effects found are intermediate between additivity and dominance, in for this trait in the interspecific hybrid. On the other agreement with PATEWONet al. (1991b). High devia- hand, at least four QTL withheterotic effects for BY were tions from the general additivity can be found when identified. No heterotic effect for this trait was observed other genes with strong pleiotropic effects are involved, in the interspecific hybrid. We therefore suggest that by such as the poor performanceof IL62 (ndw) leading to elimination of the factors responsiblefor the low produc- the large dominance deviation for all measured traits. tivity ofthe FI hybrid, the vegetative vigorwas transformed In contrast to the strongheterosis for PW of the inter- into a reproductive one. specific FI hybrid (Table l), the QTL for PW showed In 1993 DE VICENTEand TANKSLEXreported on trans- Y. Eshrtl and D. Zamir

Brix x Yield in A Oh from M82 IL IL IL x M82

II I ‘0 Byl-2 ? Byl-1 5 CM FIGL’RE3.-Fine mapping of QTI. for yield of total solulde solids (BY) in the distal end of the long arm of chromosome 1. The only I,. 1~c,717~lliichromosomesegments introgressed into the listed lines are marked by the dark bars (all lines are progenies of 11, 1-4). 1L.s 14’ and 1-4‘ arc truly isogenic and arc derivcd from F! of IL. 1-4 crossed to M82 and selected by RFLPs to be homozygous for the I.. pnncllii allelcs (1’) or the cultivated tomato alleles (I.). For each line, the phenotypic value (presented as percent difference from “82, A%) based on 10 rcplications was dctcrminetl for both thr homozygom 1I.s and the heterozygous IL. X M82. Multiple range test by “Each pair comparison” with an alpha level of 0.0.3 \vas performed Iletween the inbreds and between the hyhrids separately; different letters denote significant differences. The location of the QTI, is determined by the smallest chromosomal segment that retains a significant eff’ect on BY and is marked by the area between the vertical lines, which is the postulatcd location of the QTI. for this trait. gressive segregation in 10 out of 11 quantitative traits the tomato genome can be covered by a single YAC measured on F2 seedlings of the sameinterspecific (SEGXI.r/ nl. 1992). cross. In this study we show that the same is true for The heterotic effect of the 37cM introgression of ILl- quantitative traits measured on maturc plants and for 4 on BY was consistcnt over 3 years and for different traits having economic importance. geneticbackgrounds and growing stands (ESI-IEI)and Sensitivity of QTL togenetic background: Culti- ZASIIR1994a). Finer mapping of this BY QTL suggested vated tomato varieties represent only a small fraction that two loci were involved: a partially dominant gene of the variation presented in Lycopcrsicon (RICK1982; originating from the wild parent was responsible for the MILLERand TmKStm 1990). This leadsto the expecta- increase in BY and a linked recessive gene of I,. pnnellii tion that significant effects of the introgressions found origin reduced BY. It appears that combined action of in one cultivar are likely to be maintained in others. these QTL, which fit5 the pseudo-overdominance model In this study,the effect of the I-. prnnrllii introgressions for heterosis (CROW1952), resulted in the lower BY of was tested using an additional processing inbred line the homozygous IL1-4 compared to it5 hybrids. (AS) with larger FM and higher B. None of the 300 Implications for general genetics and plant breed- possible interactions with genetic background was sig- ing: For each of the traits analyzed in this study, intro- nificant. Theseresults confirm the broad and dramatic gressions with phenotypic effects of different magni- effects of the novel variation introduced into theculti- tudes were identified (Figure 1). These results are in vated crop. agreement with the accumulated data from numerous Fine mapping of QTL: To overcome the problem of QTL studies that establish definitively that polygenes highresolution mapping of QTL in conventional vary widelyin their effects, and in many instances a crosses, PATERSOXrt nl. (1990) used selected overlap- large proportion of the variation can be explained by ping recombinant chromosomes.In this study we dem- the segregation of a few major QTL (TANKSLEY1993). onstratethat a FM QTL can be resolved intothree Although the number of replicates used in the study linked QTL; it is possible that upon finer mappingaddi- was \‘en small (six per genotype), more QTL could be tional loci will be revealed. I:m,2-1 was mapped to an identified by comparison with similar studiesin tomato, interval of 3.2 cM (Figure 2), which in certain areas of where each genotype has been measured from a few Introgression LinesIntrogression Mappingfor QTL 1161 dozen to a few hundred times. The numbers of QTL ply inherited traits such as disease resistances.Although reported in this study are skewed downward for the in a pioneering study RICK (1974) has demonstrated following reasons. First, it is not necessarily correct that that high soluble solids content in tomato could be two overlapping introgressions with a similar effect improved through wild species crosses, the use of these carry the same QTL. Second, an IL can carry more than resources for improvement of complex traits was con- one QTL, as demonstrated here for IL2-5 and IL2-6. sidered largely impractical because of the many undesir- The higher sensitivity of this population is a result of able traits carried by wild species. This study demon- the minimal genetic variation within the lines. strates that genes from a very small green-fi-uited,poor- A statistical advantage for the experimentaldesign is yielding wild species can serve as a source for many the use of a common control for all lines that can be agriculturally important traits. Even traits that were not grown in large numbers and can ensure enough de- apparent in the parental lines eventually segregated grees of freedom. An additional advantage of the ILs is among the ILs. the ease of distinction between additiveand dominance Methodologies for the identification and mapping effects. Moreover, once an inbred IL is produced, its of genes underlying quantitative traits in plants have effect can be determined for various genetic back- advanced considerably in recent years since the advent grounds (only in a heterozygous state) and in different of marker-saturated linkage maps (TANJSSLEY1993). environments. This factor is of major importance for The developments in marker technologies have not identification of QTL with broad effects. however been accompanied by corresponding progress We favor the development of a permanent resource in population structures. In this study we describe a ofILs that provides several advantages over conven- new population capable of facilitating the utilization of tional populations. In addition to the high efficiency wild germplasm by providing the means for identifica- of QTL identification and fine mapping, it could also tion and fine mapping of QTL. contributeto investigations of interaction between All DNA clones were provided by Dr. S. D. TANKSLEY,except for QTL. Reliable estimates of theinteraction between TM clones provided by Dr. E. LIFSCHITZand GP clones provided by QTL are difficult to obtain in segregating populations Dr. C. GEBHARDT.We thank G. GEM from Akko Experiment Station since the frequencies of some of the genotypic combi- for his assistance in conducting the field trialsand T. PLEBAN,H. VAN nations are too low to allow meaningful statistical com- Oss, T. BLOCH,M. EMANUELand A. NATOR for technical assistance. parisons. The availability of nearly isogenic lines con- We thank Dr. S. D. TANKSLEY,Dr. I. PARAN,Dr. H. VOETand N. ORI for stimulating discussions and S. SMITHfor editing. This research taining different QTL would allowexamination of these was supported in part by grant no. US-2427-94 from BARD, The lines in various genotypic constitutions in balanced ex- United States-IsraelBinational Research and Development Fund. periments aimed at the precise analysis of epistatic in- Seeds of the ILs have been distributed through the Tomato Genetics teraction. Resource Center, University of California, Davis. The IL database has An important advantage of the IL approach is its been deposited in the Solanaceae database and is available through the internet: [email protected]. immediate applicability in breeding. If the recipient genome is a leadingvariety, the derived lines may represent asignificant improvement and abasis for the intro- LITERATURECITED duction of a new cultivar. ILs can also be applied in a BERNATZKY,R., and S. D. TANKSLEY,1986 Methods for detection of number of ways other than those described in this arti- single or low copy sequences in tomato on Southern blots. Plant cle (ESHEDand ZAMIR 1994b), such as mapping of new Mol. Biol. Rept. 4: 37-41. DNA clones to the genome, identification of region- CAUSSE,M. A., T. M. FULTON,Y. G. Gu, S. N. AHN,J. CHUNWONGSE et al., 1994 Saturated molecular map of the rice genome based specific DNA markers, and fine mapping of qualitative on an interspecific backcross population. Genetics 138: 1251- genes. 1274. The main disadvantage of IL populations is the long CROW,J. F., 1952 Dominance and overdominance, pp. 282-297 in Heterosis, edited by J. W. GOWEN.Iowa State College Press, Ames, time and the large amount of work required for their IA. development. We estimate that after complete genotyp- DARVASI,A,, A. WEINREB,V. MINKE,J. I. WELLERand M. SOLLER, 1993 ing of a BCl population and selection of the appre Detecting marker-QTL linkage and estimating QTL gene effect and map location using a saturated genetic map. Genetics 134: priate lines providing coverage of the genomewith min- 943-951. imal overlaps, two additional cyclesof backcrosses DE VICENTE,M. C., and S. D. TANKSLEY,1993 QTL analysis of trans- accompanied by genotypic and phenotypic selection gressive segregation in an interspecific tomato cross. Genetics 134: 585-596. are required. In a selfed BC3 population the founders DOEBLEY,J., and A. STEC, 1991 Genetic analysis of the morphologi- of an IL population can be identified. It is however cal differences between maize and teosinte. Genetics 129: 285- expected that in outcrossing species the proportion of 295. DUNNET,C. W., 1955 A muliple comparison procedure for compar- genes associated with deleterious recessive mutations ing several treatments with a control. J. Amer. Stat. Asso. 50: would be higher than reported here. 1096-1121. Wild species represent a widely divergent gene pool EMERY,G. C., and H. M. MUNGER,1970 Effect of inherited differences in growth habit on fruit size and soluble solids in tomato. for the improvement of cultivated plants. Exotic germ- J. Am. SOC.Hort. Sci. 95: 410-412. plasm have been used extensively for breeding for sim- ESHED,Y., and D. ZAMIR,1994a Introgressions from Lycopersiconpen- 1162 Y. Eshed and D. Zamir

nellii can improve the soluble solids yield of tomato hybrids. PATERSON,A. H., S. DAMON,J. D. HEWITT,D. ZAMIR, H. D. RABINO- Theor. Appl. Genet. 88: 891-897. WITCH et al., 1991b Mendelian factors underlying quantitative ESHED,Y., and D. ZAMIR, 1994b A genomic library of 1.yropersicon traits in tomato: comparison across species, generations and envi- panellii in L. esculentum: a tool for fine mapping of genes. Euphy- ronments. Genetics 127: 181-197. tica 79: 175-179. RICK,C. M., 1969 Controlled introgression of chromosomes of Soh- ESHED,Y., M. ABU-ABIED, Y. SARANGAand D. ZAMIR,1992 Lycopwsi- num pennellii into Lycqpersicon esculatum: segregation and recom- con esculentum lines containing small overlapping introgressions bination. Genetics 62 753-768. from L. pennellii. Theor. Appl. Genet. 83: 1027-1034. RICK, C. M., 1974 High soluble-solids contentin large-fruited to- FATOKUNC. A,, D. I. MENANCIO-HAUTEA,D. DANESH and N. D. YOUNG, mato lines derived from a wild green-fruited species. Hilgardia 1992 Evidence for orthologousseed weight genes in cowpea 42: 493-510. and mung bean based on RFLP mapping. Genetics 132: 841- RICK,C. M., 1982 The potential of exotic germplasm fortomato 846. improvement, pp. 1-28 in Plant Improvement and Somatic Cell GP- GOLDMAN, I. L., I. PARANand D. ih~lR,1995 Quantitative trait locus netics, edited by I. K. VASIL,W. R.SCOWCROFT and K. J. FREY. analysis of a recombinant inbred line population derived from Academic Press, New York. a Lycopwsiconesculentum X Lycopwsiconchrrsmanii cross. Theor. SAS INSTITUTElnc., 1994 JMP Statistics and Graphics Guide: version Appl. Genet. 90: 925-932. 3. SAS Institute Inc., Cary, NC. KEIM, P., B. W. DIERS,T. OLSONand R. C. SHOEMAKER,1990 RFLP SEGAI.,G., M. SmxrI, M. A. SCHAFFER,N. OM, D. ZAMIRet al., 1992 mapping in soybean: association between marker loci and varia- Correlation of genetic and physical structure in the region sur- tion in quantitative traits. Genetics 126: 735-742. rounding the 1-2 FuJarium oxy.~porumresistance locus in tomato. LIEDI.,B. E., S. E. Lrc, D. ESPOSITOand M. A.MUTSCHI.ER, 1993 Mol. Gen. Genet. 231: 179-185. Identification and mapping of S in a self compatible F2 popula- STEYENS,M. A., and J. RUDICH,1978 Genetic potential for overcom- tion of L. rsculmtum X I.. pennellii. Rept. Tomato Genet. Coop. ing physiological limitations on adaptability, yield, and quality 43: 33-35. in the tomato. Hort. Sci. 13: 673-678. MII.I.ER,J. C., and S. D. TANKSIXY,1990 FWLP analysis of phyloge- STUBER,C. W., S. E. LINCOLN,D. E. WOLFF,T. HELENTJARISand netic relationships and genetic variation in the genus Lycopwsi- E. S. LANDER,1992 Identification of genetic factors contribut- con. Theor. Appl. Genet. 80: 437-448. ing to heterosis in a hybrid from two elite maize inbred lines NIENHIJIS,J., T. HELENTJARIS,M. SI.OC:UM,B. RUGGEROand A. using molecular markers. Genetics 132: 823-839. TANKSIXY,S. D. 1993Mapping polygenes. Annu. Rev. Genet. 27: SCHAEFER,1987 Restriction fragmentlength polymorphism 205-233. analysis of loci associated with insect resistance in tomato. Crop TANKSLEY,S. D., M.W. CANAL,J. C. PRINCE,M. C. DE VICENTE, Sci. 797-803. 27: M. W. BONIERABALEet al., 1992 High density molecular linkage PATEKSON,A. H., E. S. IANDER,J. D. HEWITT,S. PETERSON,S. E. maps of the tomato and potato genomes: biological inferences I.IN(:OLNet al., 1988 Resolution of quantitative traits into Men- and practical applications. Genetics 132: 1141-1 160 delian factors, using a complete linkage map of restriction frag- WEHRHAHN,C., and W. AI.IARD,1965 The detection and measure- ment length polymorphisms. Nature 335: 721-726. ment of the effects of individual genes involved in the inheri- PATERSON,A. H.,J. W. DE-VERNA,B. LANINI and S. D. TANKSLEY,1990 tance of a quantitative character in wheat. Genetics 51: 109-1 19. Fine mapping of quantitative traitloci using selected overlapping WEIDF.,R., M. F. VANWORDRAGEN, R. KLEIN IANKHORST,R. VERKERK, recombinant chromosomes, in an interspecific cross of tomato. PI al., 1993 Integration of classical and molecular linkage maps Genetics 124: 735-74'2. of tomato chromosome 6. Genetics 135: 1175-1186. PATERSON,A. H., S. D. TANIGI.EYand M.E. SORRFLLS,l99la DNA markers in plant improvement. Adv. Agron. 46: 39-90. Communicating editor: M. R. HANSON