Genetic Analysis of Traits Distinguishing Outcrossing and Self-Pollinating Forms of Currant Tomato, Lycopersicon pimpinellifolium (Jusl.) Mill.

Corresponding author: Elizabeth M. Lord, University of California, Riverside, CA 92521., lord{at}citrus.ucr.edu (E-mail)

Communicating editor: J. B. WALSH

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

The evolution of inbreeding is common throughout the angiosperms, although little is known about the developmental and genetic processes involved. Lycopersicon pimpinellifolium (currant tomato) is a self-compatible species with variation in outcrossing rate correlated with floral morphology. Mature flowers from inbreeding and outcrossing populations differ greatly in characters affecting mating behavior (petal, anther, and style lengths); other flower parts (sepals, ovaries) show minimal differences. Analysis of genetic behavior, including quantitative trait locus (QTL) mapping, was performed on representative selfing and outcrossing plants derived from two contrasting natural populations. Six morphological traits were analyzed: flowers per inflorescence; petal, anther, and style lengths; and lengths of the fertile and sterile portions of anthers. All traits were smaller in the selfing parent and had continuous patterns of segregation in the F₂. Phenotypic correlations among traits were all positive, but varied in strength. Quantitative trait locus mapping was done using 48 RFLP markers. Five QTL total were found involving four of the six traits: total anther length, anther sterile length, style length, and flowers per inflorescence. Each of these four traits had a QTL of major (>25%) effect on phenotypic variance.

EVOLUTIONARY change in mating system from outcrossing (cross-pollination) to inbreeding (self-pollination) has occurred frequently throughout the flowering plants. Indeed, it has been described as the most common evolutionary trend in angiosperm reproduction (STEBBINS 1957 , STEBBINS 1970 ). Although the majority of angiosperms possess hermaphroditic flowers, many are highly outcrossed. This outcrossing can be regulated by a variety of mechanisms, including genetic self-incompatibility and temporal and physical separation of male and female functions in a self-compatible flower. These systems can and do break down in nature, leading to a condition in which a plant is able to pollinate and fertilize itself (WYATT 1983 ). At present, relatively little is known about the genetic and developmental mechanisms involved in this change in mating system. This study is an investigation into the genetics and development associated with the evolution of inbreeding in one species, currant tomato, Lycopersicon pimpinellifolium (Jusl.) Mill.

The evolution of inbreeding in flowering plants typically involves a syndrome of changes affecting several morphological characters (ORNDUFF 1969 ). Many simple cases of plant evolution, in which one or two genes seem to control a given trait, have been described (GOTTLIEB 1984 ); the existence of such genes of major effect in plants has been known at least since the time of Mendel. But many of the characters involved in mating system evolution, such as sizes of floral organs or amount of pollen produced, are quantitative in nature. They do not show the discrete segregation that is typical of monogenic or digenic systems; these are termed "quantitative traits" and are regarded as having a more complex genetic basis.

The genetic control of flower development has been the subject of considerable research in recent years, mainly involving the identification and characterization of major regulatory genes. A number of genes controlling meristem and floral organ identities have recently been discovered and characterized. Several of these genes appear to control the differentiation of floral organs produced during the development of flowers (WEIGEL and MEYEROWITZ 1994 ). These organ identity genes are part of a hierarchy of regulatory genes controlling the organization of flowers; many are MADS-box genes that code for transcription factors. It is not currently known if these regulatory genes are involved in quantitative changes in plant form. Relatively little is known about the control of quantitative characters involved in flower development. The development of quantitative trait locus (QTL) mapping techniques (LANDER and BOTSTEIN 1989 ) has provided a means of identifying regions of a genome affecting phenotypic traits. This technique can provide information on the map location, relative effect, gene action, and dominance properties of each identified locus, as well as provide a basis for eventual isolation of the actual genes. QTL mapping has been used in the study of maize domestication and has aided in the detection of several genes of major effect in maize morphological evolution (DOEBLEY 1992 ; DOEBLEY et al. 1997 ).

Mating system evolution in plants has been the subject of several studies over the past decade (MACNAIR and CUMBES 1989 ; SHORE and BARRETT 1990 ; HILL et al. 1992 ; HOLTSFORD and ELLSTRAND 1992 ; SHERRY 1994 ; FENSTER et al. 1995 ). These studies either dealt exclusively with morphology and development or used biometrical techniques to analyze genetics. More recently, QTL mapping has been employed to gain insight into the genetic basis of mating system evolution. BRADSHAW et al. 1995 used QTL mapping to investigate the differences between two species of Mimulus, one bee pollinated, the other hummingbird pollinated. They found at least one major QTL of at least 25% effect for each phenotypic trait. More recently, LIN and RITLAND 1997 performed QTL mapping on two species of Mimulus, one outcrossing and the other selfing. The QTL found in their study were of smaller effect, with only one exceeding 20%.

Species of Lycopersicon range from self-incompatible outcrossers to self-compatible selfers (RICK et al. 1977 ; MILLER and TANKSLEY 1990 ). Several self-compatible species vary in their rates of outcrossing, making them ideal candidates for the study of mating system evolution. In a study by RICK et al. 1977 , variation among populations of L. pimpinellifolium was documented in terms of outcrossing rate, anther length, and degree of stigma exsertion. These characters were found to be positively correlated and the floral features were found to be inherited in a quantitative fashion. Self-incompatible species of Lycopersicon, which are obligate outcrossers, also differ from self-pollinating types by the same characters (RICK 1982 ). RICK et al. 1976 have also documented a closely related pair of species differing mainly in mating system, L. parviflorum and L. chmielewskii (collectively referred to as the L. "minutum" complex). This species pair is distinguished by the same floral characters that differentiate the selfing and cross-pollinating populations of L. pimpinellifolium and that are typical of many such pairs of closely related taxa throughout the angiosperms. RICK et al. 1977 also noted that the self-pollinating forms are found typically at the periphery of the native range of the species, which is compatible with the hypothesis that they are derived from the more highly outcrossed forms.

Several studies have addressed the genetics of floral form in relation to pollination and fruit characters in tomato, using classical methods of genetic analysis (LEVIN et al. 1994 ) and QTL mapping (GRANDILLO and TANKSLEY 1996 ). However, they used cultivated tomato varieties and examined only those aspects of mature flowers considered important in agriculture, without regard for the developmental basis of floral form or for the natural range of mating system variation.

To establish the best morphological characters to use for genetic study of the evolution of inbreeding in L. pimpinellifolium, an analysis of developmental differences between selfing and outcrossing flower types in this species was undertaken (GEORGIADY and LORD 2002 ). On the basis of this developmental analysis and previous work on the pollination biology of this species (RICK et al. 1977 , RICK et al. 1978 ), six characters were chosen for genetic analysis: number of flowers per inflorescence and petal, anther (three characters), and style lengths. The number of flowers per inflorescence is one of the factors affecting the extent of floral display and, by inference, pollinator attraction. This character has also been examined in other QTL mapping experiments (GRANDILLO and TANKSLEY 1996 ), so it provides a basis for comparison. Petal length is used as a measure of corolla size, also a factor in floral display. The developmental study (GEORGIADY and LORD 2002 ) revealed that anther differentiation was the first indication of divergence in the two flower forms. Anther length is used as a measure of both anther-tube length and anther size; in tomato, anthers add to floral display, produce pollen, and control pollination by their proximity to the stigma. The fertile portion of the anther represents the site of pollen production, and the sterile portion, which is a unique and defining feature of Lycopersicon species (TAL 1967 ), represents a functional part of the anther tube. Style length is important for pollination, as is the relative position of the anthers and stigma (RICK and DEMPSEY 1969 ; RICK et al. 1978 ). Differences in pistil length in this species are due to differences in style length alone (GEORGIADY and LORD 2002 ).

In this study on the evolution of inbreeding flower form in L. pimpinellifolium, the central question is: what is the genetic basis for the differences in flower form in this species? Different populations of a single species, which show divergence in mating system from outcrossed to inbred forms, are examined. The plants used are derived from natural populations and are not cultivars. Environmental variation is controlled by growing plants together, under the same conditions, in a climate-controlled greenhouse. It has been shown that outcrossing rate correlates with floral morphology in tomato (RICK et al. 1978 ), so morphological characters are used as an indirect measurement of outcrossing rate.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Plant material:
Seed for L. pimpinellifolium (Jusl.) Mill. of contrasting mating system was obtained from the Tomato Genetics Research Cooperative (TGRC, Davis, CA). Two accessions were selected to represent selfing and outcrossing types: LA1237, the "selfer," and LA1581, the "outcrosser" (RICK et al. 1977 ). These two lines were originally obtained from wild populations that were estimated to have outcrossing rates of 0 and 37%, respectively, and that represented the extremes for this species in terms of morphology and outcrossing rate. Plants were grown in a greenhouse at Riverside, California, and periodically propagated by cuttings.

Three selfing and four outcrossing individuals were initially surveyed. The pair showing the greatest degree of genetic polymorphism was crossed, producing F₁ hybrids. The choice of parents was largely arbitrary, since the individuals within each accession were very similar in terms of their restriction fragment length polymorphism (RFLP) fragment patterns, and the two accessions were quite distinct (Fig 1). A few probes showed heterozygosity in one parent and were resurveyed in 28 F₁ plants. An F₁ individual having the most informative genotype was selfed and used to generate a mapping population of 147 F₂ plants. This F₂ population was used for both linkage analysis and QTL mapping.

View larger version (8K): [in this window] [in a new window] [Download PPT slide]   Figure 1. UPGMA tree based on RFLP data for the seven L. pimpinellifolium individuals selected as potential parents for the mapping cross. The TGRC accession number for the outcrosser (LA1581) and selfer (LA1237) are shown, along with a lowercase letter suffix indicating an arbitrary designation for each individual within the accession. The tree was generated with the PHYLIP program NEIGHBOR (FELSENSTEIN 1993 ).

Scoring of phenotypic characters:
Measurements of phenotypic traits for all plants were made on fresh material. Flowers were harvested at midday on the day of anthesis. Each flower was dissected with the aid of a stereoscope and length of individual parts (sepals, petals, anthers, filaments, styles, and ovaries) was measured to the nearest 0.1 mm using an ocular reticle. In addition, for anthers, lengths of fertile and sterile regions were estimated by measuring stomium length, which corresponds well to pollen-sac length (Fig 2). Whole anthers were stained in 0.2% aqueous crystal violet for 30 sec and destained in dH₂O to aid in visualizing the stomia. Thirty flowers from parents and F₁ and five from each F₂ plant were measured.

View larger version (21K): [in this window] [in a new window] [Download PPT slide]   Figure 2. Morphological characters used for genetic analysis. (A) Petal length was measured as the length of the petal from its apex to its adaxial junction with the receptacle. (B) Anther length was measured as the entire length of the anther. Fertile length was measured as the length of the stomia, from their basal end to their junction at the midline of the anther. Sterile length was measured as the distance from the junction of the stomia to the apical end of the anther. (C) Style length was measured from the junction of the style and ovary to the tip of the stigma.

The number of flowers produced per inflorescence was counted, 30 inflorescences each for parents and F₁ and 10 each for F₂ individuals. Counts were made using inflorescences that were fully developed, i.e., in which the youngest (most distal) bud was at or near anthesis; often a final bud was initiated but did not develop fully, and these buds were not counted. Branched inflorescences, and those bearing foliage leaves, were not used.

Statistical analysis of phenotypic data:
Means, standard deviations, Pearson correlation coefficients, and covariances were calculated for each phenotypic trait using the general linear model (GLM) procedure of SAS (SAS INSTITUTE 1989). Genetic variances and covariances were calculated by subtracting the phenotypic variances and covariances of the F₁ population from those of the F₂ population (FALCONER 1989 ); this makes the assumption that genotype-environment interactions are not a factor.

Southern blotting and hybridization:
Genomic DNA was extracted from fresh young shoots via a modified cTAB extraction (OKADA et al. 1997 ). Southern transfers were produced by digestion of genomic DNA with one of four restriction endonucleases (EcoRI, EcoRV, HindIII, and DraI), agarose gel electrophoresis, and solvent transfer to nylon membrane (WHITKUS et al. 1992 ).

Tomato genetic probes (TANKSLEY et al. 1992 ) were obtained as plasmid DNA from Steven Tanksley at Cornell University and prepared as described in OKADA et al. 1997 . Briefly, probe DNA for hybridization was generated by bacterial transformation, miniprep extraction, and insert excision via gel isolation. Alternatively, direct amplification of plasmid inserts was used to prepare some probes. Oligolabeling of probe DNA, hybridization, and washing of Southern blots were done as described in OKADA et al. 1997 .

Marker selection, map construction, and QTL mapping:
Of 117 probes surveyed, 54 appeared to be polymorphic between the two parents. Behavior of questionable probes was rechecked in parents and F₁. Forty-seven probes were deemed usable in the F₂ and were used for mapping. Of these, 5 were dominant and 42 were codominant. Segregation ratios were checked using a chi-square test (WEAVER and HEDRICK 1989 ). A linkage map was constructed using MAPMAKER/EXP v3.0 (LANDER et al. 1987 ; LINCOLN et al. 1992 ). QTL Cartographer v1.13 (BASTEN et al. 1999 ) was used for single-marker regression analysis and interval mapping and for permutation tests to determine experiment-wide significance levels for each of the traits.

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Phenotype of parents, F₁, and F₂:
Results of phenotypic measurements for parents, F₁, and F₂ are summarized in Fig 3. The parents possessed large and distinct differences for each character, the outcrosser consistently being the larger of the parents. All characters showed quantitative inheritance in the F₂ generation.

View larger version (49K): [in this window] [in a new window] [Download PPT slide]   Figure 3. Phenotypic data for parents, F1, and F2. At the top of the graphs, open diamonds represent selfing parent means, solid diamonds represent outcrossing parent means, and shaded diamonds represent F1 means. Means for the entire F2 population are shown by open circles. Error bars represent ±1 SD. In the vertical bar graphs, bar lengths indicate relative frequencies of F2 phenotypes in 12 arbitrary classes.

The number of flowers per inflorescence was 4.8 (60%) greater in the outcrosser than in the selfer, although with a substantial amount of variation, particularly in the outcrosser (Fig 3A). Relative to their parents, the F₁ was intermediate in this character, with essentially no dominance toward either parent. F₂ plants exhibited significant transgressive segregation, with the mean of the F₂ population slightly exceeding the larger parent. The smaller parent inflorescence size was not recovered in the F₂ population.

Petals were 4.5 mm (46%) longer in the outcrosser than in the selfer; they were intermediate in the F₁, although with a moderately high degree of dominance toward the larger parent (Fig 3B). The mean petal length in the F₂ population was very close to the mean of the parents. A small portion of the F₂ population had petals shorter than the smaller parent, and the larger parent length was not recovered. Two widely spaced peaks occur in the F₂ distribution, a large peak between 12 and 13 mm and a smaller peak between 10 and 11 mm.

Anther length showed a similar pattern of inheritance to petal length. Anthers, and their fertile and sterile portions, were greater in the outcrosser by 3.0 mm (41%), 1.1 mm (24%), and 1.9 mm (77%), respectively. Total anther length in the F₁ was intermediate with some dominance toward the larger parent, and the F₂ mean was exactly intermediate between the parents (Fig 3C). There are also two peaks in the F₂ distribution, although they are closer together than those in petal length. The dominance in the F₁ appears to be due to the dominance in the fertile portion of the anther (Fig 3D), which is nearly complete. In contrast, the sterile portion of the anther (Fig 3E) shows very little dominance. The F₂ mean for fertile length is well below the mean parental length, and the mean F₂ sterile length is greater than that of the parental mean by an equal amount. The secondary peak in overall anther length is also seen in sterile length, but not in fertile length. All parental lengths were recovered in the F₂, with the exception of sterile tip length for the outcrosser.

Styles in the outcrosser were 4.8 mm (66%) longer than those in the selfer and were of intermediate length in the F₁, with some dominance toward the larger parent (Fig 3F). The F₂ mean was close to the mean of the parents. As with petal length, a small portion of the F₂ population had styles shorter than the smaller parent, and the larger parental length was not recovered.

Means and variances for the F₂ population are given in Table 1, and correlations among traits are given in Table 2. Flowers per inflorescence is poorly correlated with all other traits in terms of phenotype; this lack of correlation is reflected in the poor statistical significance for this trait. The best correlation for flowers per inflorescence is with style length, but only with a coefficient of 0.208. Overall, the best correlations were total anther length vs. length of anther sterile portion (0.870) and petal length vs. style length (0.839). All correlations among the floral (noninflorescence) traits are statistically highly significant, and all are above 0.5, with the exception of fertile vs. sterile portions of the anther (0.329).

Linkage behavior of markers:
Initial processing of the marker data produced 14 linkage groups (Fig 4); 2 of the 12 chromosomes had linkage groups spaced far enough apart that MAPMAKER did not recognize them as being linked, and their position was inferred from TANKSLEY et al. 1992 . In all, 44 of the 47 markers mapped to locations essentially the same as given in TANKSLEY et al. 1992 . The remaining 3 were apparently previously unmapped secondary loci. Using the Haldane map function to convert recombination frequencies to map distances, the total map length was 1296 cM, and the average distance between markers was 37 cM.

View larger version (46K): [in this window] [in a new window] [Download PPT slide]   Figure 4. Linkage map with QTL. Marker designations are from TANKSLEY et al. 1992 ; those in italics are unique to this study. Map distances (Haldane) from data in this study are given to the left of each chromosome; values in italics are comparable distances from TANKSLEY et al. 1992 , and distances for terminal portions of that map not covered by markers in this study are in parentheses. Distances for intervals covering markers unique to this study are in brackets. All distances are in centimorgans. To the right of each chromosome containing a putative QTL detected in this study, a graph of the test statistic from interval mapping is given, with LOD scale. The five significant QTL identified in this study are shown, along with a one-LOD confidence interval. Dots on the graphs aligned with each marker indicate the results of linear regression of the trait on that marker.

Segregation distortion was significant in 4 of the 47 markers, in all cases toward the selfing parent. All 3 markers on chromosome 4 showed this distortion, as well as TG96 on chromosome 5. In general, the use of markers exhibiting segregation distortion is to be avoided; however, particularly in the case of chromosome 4, there would seem to be little alternative here. This should be kept in mind when interpreting the results.

Quantitative trait loci:
Initially, 18 QTL were identified for the six traits (Table 3), using an arbitrary LOD threshold of 2.4 (GRANDILLO and TANKSLEY 1996 ). Subsequent permutation analysis incorporating 1000 resamplings was used to estimate the genome-wide significance of each QTL, and those exceeding the 0.05 level of significance were retained. Five QTL met these criteria; 2 were associated with anther sterile length and 1 each with total anther length, style length, and flowers per inflorescence. No QTL were detected in this way for petal length or anther fertile length. Four of the 5 QTL were of major effect, explaining >25% of phenotypic variance. The QTL were found on three chromosomes (Fig 4), with 3 QTL on chromosome 3 and 1 each on chromosomes 8 and 11.

Results from single-marker analysis (Fig 4) confirm the interval mapping results. As expected, QTL close to markers were identified by linear regression, while those not near a marker were not detected as reliably. In the one case in which a QTL occurred coincident with a marker (ste11, TG400), results for interval mapping and linear regression were identical, as expected.

The QTL for flowers per inflorescence had a large additive component and a large but negative dominance component that correspond well to the small inflorescences of the F₁ and very large inflorescences of the F₂ (Fig 2A). This QTL is in a region of the long arm of chromosome 3 near two other major QTL. One QTL was found for total anther length (ant), also on chromosome 3. Two QTL were found for anther sterile length (ste), on chromosomes 3 and 11. It should be noted that these two anther-related traits each have a QTL on the long arm of chromosome 3. The QTL of largest effect found in this study, explaining 42% of the variance in style length (sty), is located on chromosome 8.

A substantial degree of dominance was associated with two of the QTL, fpi3 and sty8. The dominance effect was of greater magnitude in sty8 than the additive effect. All of the QTL had negative dominance components, although all six traits had some degree of positive dominance in the phenotypic analysis. These negative dominance components indicate dominance of the selfing parent or, alternatively, recessiveness in the outcrossing parent, given that it is presumed to be the ancestral form.

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

The two most striking results of this study are the variation in genetic behavior of the traits examined and the presence of QTL of large effect on phenotypic variance for four of the six traits: total anther length, anther sterile length, style length, and flowers per inflorescence. The differences between the fertile and sterile portions of the anther, and the QTL of large effect on variance in style length, are of particular interest, given their importance in the mechanics of pollination.

Anther length:
The lengths of fertile and sterile regions of anthers are controlled by substantially different genetic mechanisms, as evidenced by low phenotypic correlation, dominance behavior in F₁ phenotype, and transgressive segregation in the fertile region in the F₂. A QTL for total anther length was found on chromosome 3 and explained 35% of phenotypic variation for this trait. The QTL has a small additive component and a small negative dominance component. As the F₁ shows phenotypic dominance toward the larger outcrossing parent, other factors must be involved in the regulation of total anther length. Two QTL were found for anther sterile length, one of major effect (31.0%) and the other of lesser (11.3%) effect on phenotypic variance. The major QTL is located on chromosome 3; given its location and effects, it may be the same as the QTL on that chromosome affecting overall anther length.

It should be noted that GRANDILLO and TANKSLEY 1996 examined anther length in a backcross between L. esculentum and L. pimpinellifolium. They found two QTL affecting this trait, on chromosomes 2 and 7, which accounted for only 24% of the phenotypic variation. Neither of those loci corresponds to any of the three found in this study. This could be due to a monomorphic state for such loci in the plants used, given the species involved in each.

The sterile anther tip has been used as a defining character for the genus Lycopersicon, although L. pennellii (formerly Solanum pennellii) has the typical Solanum-type anther, which lacks this feature. TAL 1967 performed a crossing experiment between L. esculentum and L. pennellii and attempted to characterize the genetic basis for several traits, including the sterile anther tip. Other than documenting the quantitative nature of this trait, however, he shed little light on this interesting aspect of tomato flower morphology. The interspecific cross produced a wide range of intermediates, but quantitation of the intermediates was not done due to their unusual morphology. From purely qualitative observations, he estimated that two major genes controlled this trait.

The use of component characters here, namely lengths of component portions of the anther, allowed for detection of a QTL (ste11) that would not have been found using total anther length alone. This might be expected, since the more complex a character is, the more genes would be involved in its regulation, and the less effect each would have, making detection more difficult. This situation points up the complexity of doing genetic analyses and the importance of judicious character selection.

Style length:
The presence of a QTL of very large effect (42.2%) for style length is somewhat of a surprise. It is well documented that continuous variation in stigma exsertion exists among wild populations of this species (RICK et al. 1977 ). From this, one might expect a larger number of genes of lesser effect to control style length. However, if several modifier genes are also present, with significant diversity in dominance effects, this would not be a limiting factor. Given the large negative dominance component of sty8 and the positive dominance seen in the F₁, other factors are apparently modifying the effect of the QTL on the phenotype to a substantial degree.

It is interesting to note that the outcrossing parental phenotype was not recovered in the F₂. This may be due in large part to chance, since relatively few F₂ individuals would be expected to be homozygous for all outcrossing parent alleles, given that several unlinked genes affect the trait. The low level of environmental variance, as evidenced by the parents and F₁, and dominance effects also may have played a part.

Flowers per inflorescence:
Flowers per inflorescence differed from all the other traits in the dominance of the selfer in the F₁ and the large degree of transgressive segregation of the F₂. It is not surprising that behavior of flowers per inflorescence in the F₁ and F₂ is very different from the other traits. The inflorescence is a product of an indeterminate shoot apical meristem, which produces foliage leaves and then branch floral meristems. The floral meristems are determinate in nature and produce only the floral organs (which comprise the other traits) and are thus very different in character from the inflorescence meristem. Thus, one would expect the genetic basis of these two types of traits to be very different, as was seen here.

The large inflorescences in the F₂ seem to be due to the breakdown of normal regulation rather than to the heterotic increase in vigor. Inflorescences with multiple branches were common in some of the F₂ plants, but are almost never seen in nature. This behavior has, however, been described in cultivated "Multiflor" plants and anantha-3 mutants (STANCHEVA et al. 1997 ). The atypical branching and pronounced transgressive segregation in the F₂ provide evidence that large inflorescences are under significant negative selection pressure in wild populations. Although a large inflorescence provides a larger floral display for pollinators and a greater possibility of fruit production, there are limiting factors that are also involved. These could include the ability of the plant to support large inflorescences, both physiologically and mechanically, and the effect on outcrossing rate.

The QTL detected for flowers per inflorescence (fpi3) corresponds to a major QTL for the same basic character (number of fruits per truss) in a L. esculentum x L. pimpinellifolium backcross (GRANDILLO and TANKSLEY 1996 ). As differentiation of the shoot meristem is one of the earliest events in embryogeny, one would expect genes affecting shoot growth to be evolutionarily conserved. Also, since flowers are always a product of shoot growth, genes affecting flower development always operate downstream of those affecting shoot growth and would be expected to be more prone to evolutionary change. Thus some commonality for this trait with another species (L. esculentum in this case) is not surprising.

Correlations between traits:
Flowers per inflorescence did not show significant correlation with any of the other traits. Again, this was not unexpected, as FPI is different in nature from the other traits. All the remaining traits showed significant positive phenotypic correlation between one another. The low phenotypic correlation between fertile and sterile portions of the anther, as noted above, is particularly striking and provides good evidence that they are distinct components of the anther that warrant separate attention. The high correlation between petal and style is also notable, since they are separated spatially and temporally during flower development by the anthers. The excessively large values derived for the genetic correlation coefficients in 4 of the 10 comparisons indicate that the assumption of no genotype-environment interaction is not well supported in this experiment.

Comparison with other tomato QTL mapping experiments:
QTL mapping has been used by plant breeders to identify loci affecting traits for yield and disease resistance (LEONARDS-SCHIPPERS et al. 1994 ; ALPERT and TANKSLEY 1996 ). Several QTL mapping experiments have been performed on various tomato species; a few of these have included morphological traits affecting mating system (DE VICENTE and TANKSLEY 1993 ; GRANDILLO and TANKSLEY 1996 ; BERNACCHI and TANKSLEY 1997 ; FULTON et al. 1997 ). However, interpreting these results in an evolutionary light presents some difficulty. Relatively wide, interspecific crosses between L. esculentum cultivars and other species were used, which tend to obscure the evolutionary significance of morphological differences. Wide crosses also produce major changes in the background in which genes operate, making analysis of their function more difficult (BERNACCHI et al. 1998 ; ALLARD 1999 ). The use of complex traits (flower size, stigma exsertion) and subjective scoring of traits in these studies also reduces their usefulness. Still, they provide some basis for comparison with our study, and results from these other studies have been included for comparison as appropriate.

Monogenic traits:
There does not seem to be good correspondence between any of the QTL found in this study and genes of known effect on tomato floral morphology (TANKSLEY and MUTSCHLER 1990 ). For example, many genes of the ms (male sterile) series can alter the size of anthers as well as reduce pollen production. One might expect some of them to be involved in the reduction of those traits in the selfer in this study. However, all the ms genes are on different chromosomes from the QTL for anther fertile length found in this study, except for ms-9 on chromosome 3; it is at the extreme opposite end of the chromosome from ant3, so it would seem an unlikely candidate.

Several monogenic mutants that affect inflorescence architecture, including anantha, conjunctiflora, and multifurcata, are also known. These generally have conspicuous effects on vegetative organs also, so it would seem unlikely that the vegetatively normal plants used in this study would have significant variation in these genes. An exception might be sft (single-flower truss), which can reduce the number of flowers per inflorescence. It has been linked to chromosome 3, but its position is not known; this leaves open the possibility that it is the gene responsible for fpi3.

Other effects:
Most previous studies of mating system evolution in plants have dealt exclusively with phenotypic data, which made analysis of dominance more difficult. In two studies of the genus Mimulus, involving different species, MACNAIR and CUMBES 1989 found dominance toward the outcrossing parent, but FENSTER and RITLAND 1994 found dominance toward the selfer. In the only other QTL mapping study done to date, LIN 1996 found a mixture of dominance effects, but many QTL had dominance toward selfing.

LATTA and RITLAND 1993 attempt to model conditions under which a stable mixed mating system can occur. They concluded that mixed mating was more likely if selfing alleles were dominant and when multiple genes are involved in controlling outcrossing rate. Both those conditions would seem to be present in the L. pimpinellifolium system. However, the monotonic decline in fitness from inbreeding that was assumed in their model does not seem to be realistic; populations with little or no outcrossing are found through the native range of L. pimpinellifolium (RICK et al. 1977 ) and seem to be quite successful. Both selfing and mixed mating systems appear to be stable in this species, which indicates that both long- and short-term elements of fitness may be involved.

Based on the additive and dominance values found for the QTL, predictions of F₁ phenotypes are always smaller than those observed in the actual F₁ plants. This could be due to error in the estimates, the action of undetected loci, or more complex genetic interaction than the simple additive-dominance model assumes. The data here do show that F₁ phenotype may not be a reliable indicator of the dominance of individual genes controlling quantitative traits such as these.

Transgressive segregation has been documented in tomato crosses and attributed to complementary gene action and dominance effects (RICK and SMITH 1953 ; DE VICENTE and TANKSLEY 1993 ). Each of the five floral traits in this study shows a small amount of transgressive segregation. This is probably due to normal environmental variation, although the underdominance of several QTL could also be factors. The transgressive segregation seen in flowers per inflorescence is particularly striking. The question here is perhaps not why the transgressive segregation is occurring in this trait, but why the inflorescences are apparently constrained to smaller numbers of flowers in the outcrossing parent and F₁. This is an interesting situation that deserves further investigation.

General observations:
The most interesting result of this study is that a QTL of major effect on phenotypic variance exists for four of the six traits. This is in agreement with several other studies in which major QTL were found for floral traits (PATERSON et al. 1988 , PATERSON et al. 1991 ; DOEBLEY and STEC 1991 ; BRADSHAW et al. 1995 ; GRANDILLO and TANKSLEY 1996 ; FULTON et al. 1997 ) and stands in contrast to the study of LIN and RITLAND 1997 in which primarily QTL of lesser effect were found. This last study is of particular interest as it is the only other QTL mapping study that specifically addresses the phenomenon of the evolution of inbreeding in plants that differ by the same basic syndrome of floral traits as those in this study; the contrasting results may be due in part to the interspecific cross used or to character choice (e.g., stigma-anther separation).

In most of the traits in our study, the genetic basis seems to be essentially that described by GRANT 1975 , with a macromutation and modifiers. This result is particularly interesting, both because there was relatively little existing work to base expectations on and because the debate over the mechanisms of evolution of characters is very much current. In this system at least, DARWIN's "innumerable slight variations" (1859, p. 459) or change by means of many genes of small effect (LANDE 1981 ) would not seem to be the mechanism by which evolutionary change has occurred. It should be noted that a QTL of large effect on phenotypic variance, such as sty8, could be made up of several closely linked genes of lesser effect, although this need not be the case; FRARY et al. 2000 have recently shown that a single gene underlies a QTL of major effect on variation in tomato fruit weight.

Taken as a whole, these data indicate that slightly different genetic mechanisms have resulted in similar changes in each of the characters involved in the evolution of inbreeding in this species. The genetic regulation of development can be very complex; it remains to be seen what the functions of the specific genes involved are and to decipher the regulatory hierarchy of the genes controlling development.

This study and others, such as LIN 1996 , show that many of the simplifying assumptions made in the theoretical modeling of mating system evolution are not supported by empirical data. Several loci can be involved in the control of mating system traits; these loci can be of large and unequal effect, and additive and dominance components can vary widely. There may be species with simple enough reproductive evolution that such models might be adequate. However, tomato seems to be a relatively simple system to deal with, because of its regular flower architecture and its low degree of environmental input into flower development, and yet the genetics appear to be relatively complicated. Without question, more empirical data on different species will be needed before a general understanding of mating system evolution in plants will be reached.

The main disadvantage of using a close cross, as in this study, is that marker polymorphism is difficult to find. That was certainly the case in this system, in which fewer than one in eight probe-enzyme combinations showed polymorphism in the parents. The lack of thorough coverage of the genome by markers leaves open the possibility that some QTL were not discovered or that the estimates of gene action and additivity are not very accurate. A standard caveat in studies of this type is that the effects of detected QTL are often overestimated (LYNCH and WALSH 1998 ). The use of a larger mapping population to increase the power of analysis could also improve the quality of results, particularly in terms of detecting QTL of small effect (BEAVIS 1998 ). One study (PATERSON et al. 1991 ) investigated plants grown in different environments and showed that different QTL were sometimes found; that approach could be used for further investigation also. The use of developmental traits such as growth rates would be laborious, but perhaps interesting, given that the duration of growth is a major factor in distinguishing the selfing and outcrossing flower types in a study of their development (GEORGIADY and LORD 2002 ). An example of this might be to perform floral organ allometry in the F₂, to see if the simple heterochronic changes suggested by our developmental analysis are regulated independently in each floral organ or on a global basis. And finally, the discovery and characterization of the individual genes involved is becoming a realistic prospect at this point.

	FOOTNOTES

¹ Present address: Department of Biology, Sonoma State University, Rohnert Park, CA 94928.

	ACKNOWLEDGMENTS

The authors thank Charles Rick for tomato seed and advice, Shizhong Xu for advice on data analysis and interpretation, Martha Mutschler for helpful advice, and Steven Tanksley for marker clones. Support for this study was received through a University of California, Riverside, Graduate Dean's Dissertation Research Grant and National Science Foundation grants IBN-9801157 to E.M.L. and M.S.G. and GER-9355042 to M.S.G. This work represents partial fulfillment of requirements for the Doctor of Philosophy degree for M.S.G.

Manuscript received November 30, 2001; Accepted for publication February 13, 2002.

	LITERATURE CITED

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

ALLARD, R. W., 1999 Principles of Plant Mating, Ed. 2. John Wiley & Sons, New York.

ALPERT, K. B. and S. D. TANKSLEY, 1996  High-resolution mapping and isolation of a yeast artificial chromosome contig containing fw2.2: a major fruit weight quantitative trait locus in tomato. Proc. Natl. Acad. Sci. USA 93:15503-15507[Abstract/Free Full Text].

BASTEN, C. J., B. S. WEIR and Z-B. ZENG, 1999 QTL Cartographer, Version 1.13. Department of Statistics, North Carolina State University, Raleigh, NC.

BEAVIS, W. D., 1998 QTL analysis: power, precision, and accuracy, pp. 145–162 in Molecular Dissection of Complex Traits, edited by A. H. PATERSON. CRC Press, Boca Raton, FL.

BERNACCHI, D. and S. D. TANKSLEY, 1997  An interspecific backcross of Lycopersicon esculentum x L. hirsutum: linkage analysis and a QTL study of sexual compatibility factors and floral traits. Genetics 147:861-877[Abstract].

BERNACCHI, D., T. BECK-BUNN, Y. ESHED, J. LOPEZ, and J. UHLIG et al., 1998  Advanced backcross QTL analysis in tomato. I. Identification of QTL for traits of agronomic importance from Lycopersicon esculentum. Theor. Appl. Genet. 97:381-397.

BRADSHAW, H. D., JR., S. M. WILBERT, K. G. OTTO, and D. W. SCHEMSKE, 1995  Genetic mapping of floral traits associated with reproductive isolation in monkeyflowers (Mimulus). Nature 376:762-765.

DARWIN, C., 1859 On the Origin of Species by Means of Natural Selection, or The Preservation of Favored Races in the Struggle for Life. John Murray, London.

DE VICENTE, M. C. and S. D. TANKSLEY, 1993  QTL analysis of transgressive segregation in an interspecific tomato cross. Genetics 134:585-596[Abstract].

DOEBLEY, J., 1992  Mapping the genes that made maize. Trends Genet. 8:302-307[Medline].

DOEBLEY, J. and A. STEC, 1991  Genetic analysis of the morphological differences between maize and teosinte. Genetics 129:285-295[Abstract].

DOEBLEY, J., A. STEC, and L. HUBBARD, 1997  The evolution of apical dominance in maize. Nature 386:485-488[Medline].

FALCONER, D. S., 1989 Introduction to Quantitative Genetics, Ed. 3. Longman, New York.

FELSENSTEIN, J., 1993 PHYLIP (Phylogeny Inference Package), version 3.5c. Department of Genetics, University of Washington, Seattle.

FENSTER, C. B. and K. RITLAND, 1994  Quantitative genetics of mating system divergence in the yellow monkeyflower species complex. Heredity 73:422-435.

FENSTER, C. B., P. K. DIGGLE, S. C. H. BARRETT, and K. RITLAND, 1995  The genetics of floral development differentiating two species of Mimulus (Scrophulariaceae). Heredity 74:258-266.

FRARY, A., T. C. NESBITT, A. FRARY, S. GRANDILLO, and E. VAN DER KNAAP et al., 2000  Fw2.2: a quantitative trait locus key to the evolution of tomato fruit size. Science 289:85-88[Abstract/Free Full Text].

FULTON, T. M., T. BECK-BUNN, D. EMMATTY, Y. ESHED, and J. LOPEZ et al., 1997  QTL analysis of an advanced backcross of Lycopersicon peruvianum to the cultivated tomato and comparisons with QTL found in other species. Theor. Appl. Genet. 95:881-894.

GEORGIADY, M. S. and E. M. LORD, 2002  Evolution of the inbred flower form in the currant tomato, Lycopersicon pimpinellifolium. Int. J. Plant Sci. in press.

GOTTLIEB, L. D., 1984  Genetics and morphological evolution in plants. Am. Nat. 123:681-709.

GRANDILLO, S. and S. D. TANKSLEY, 1996  QTL analysis of horticultural traits differentiating the cultivated tomato from the closely related species Lycopersicon pimpinellifolium.. Theor. Appl. Genet. 92:935-951.

GRANT, V., 1975 Genetics of Flowering Plants. Columbia University Press, New York.

HILL, J. P., E. M. LORD, and R. G. SHAW, 1992  Morphological and growth rate differences among outcrossing and self-pollinating races of Arenaria uniflora (Caryophyllaceae). J. Evol. Biol. 5:559-573.

HOLTSFORD, T. P. and N. C. ELLSTRAND, 1992  Genetic and environmental variation in floral traits affecting outcrossing rate in Clarkia tembloriensis (Onagraceae). Evolution 46:216-225.

LANDE, R., 1981  The minimum number of genes contributing to quantitative variation between and within populations. Genetics 99:541-553[Abstract/Free Full Text].

LANDER, E., P. GREEN, J. ABRAHAMSON, A. BARLOW, and M. DALY et al., 1987  MAPMAKER: an interactive computer package for constructing primary genetic linkage maps of experimental and natural populations. Genomics 1:174-181[Medline].

LANDER, E. S. and D. BOTSTEIN, 1989  Mapping Mendelian factors underlying quantitative traits using RFLP linkage maps. Genetics 121:185-199[Abstract/Free Full Text].

LATTA, R. and K. RITLAND, 1993  Models for the evolution of selfing under alternative modes of inheritance. Heredity 71:1-10.

LEONARDS-SCHIPPERS, C., W. GIEFFERS, R. SCHAFER-PREGL, E. RITTER, and S. J. KNAPP et al., 1994  Quantitative resistance to Phytophthora infestans in potato: a case study for QTL mapping in an allogamous plant species. Genetics 137:67-77[Abstract].

LEVIN, I., A. CAHANER, H. D. RABINOWITCH, and Y. ELKIND, 1994  Effects of the ms10 gene, polygenes, and their interaction on pistil and anther-cone lengths in tomato flowers. Heredity 73:72-77.

LIN, J.-Z., 1996 The genetic basis of mating system differences in Mimulus (Scrophulariaceae). Ph.D. Dissertation, University of Toronto, Canada.

LIN, J.-Z. and K. RITLAND, 1997  Quantitative trait loci differentiating the outmating Mimulus guttatus from the inbreeding M. platycalyx.. Genetics 146:1115-1121[Abstract].

LINCOLN, S., M. DALY and E. LANDER, 1992 Constructing genetic maps with MAPMAKER/EXP 3.0, Ed. 3. Whitehead Institute Technical Report, Whitehead Institute, Cambridge, MA.

LYNCH, M., and B. WALSH, 1998 Genetics and Analysis of Quantitative Traits. Sinauer Associates, Sunderland, MA.

MACNAIR, M. R. and Q. J. CUMBES, 1989  The genetic architecture of interspecific variation in Mimulus. Genetics 122:211-222[Abstract/Free Full Text].

MILLER, J. C. and S. D. TANKSLEY, 1990  RFLP analysis of phylogenetic relationships and genetic variation in the genus Lycopersicon.. Theor. Appl. Genet. 80:437-448.

OKADA, M., R. WHITKUS, and T. K. LOWREY, 1997  Genetics of adaptive radiation in Hawaiian and Cook Islands species of Tetramolopium (Asteraceae; Astereae). I. Nuclear RFLP marker diversity. Am. J. Bot. 84:1236-1246[Abstract].

ORNDUFF, R., 1969  Reproductive biology in relation to systematics. Taxon 18:121-133.

PATERSON, A. H., E. S. LANDER, J. D. HEWITT, S. PETERSON, and S. E. LINCOLN et al., 1988  Resolution of quantitative traits into Mendelian factors by using a complete linkage map of restriction fragment length polymorphisms. Nature 335:721-726[Medline].

PATERSON, A. H., S. DAMON, J. D. HEWITT, D. ZAMIR, and H. D. RABINOWITCH et al., 1991  Mendelian factors underlying quantitative traits in tomato: comparison across species, generations, and environments. Genetics 127:181-197[Abstract].

RICK, C. M., 1982  Genetic relationships between self-incompatibility and floral traits in the tomato species. Biol. Zent. Bl. 101:185-198.

RICK, C. M. and W. H. DEMPSEY, 1969  Position of the stigma in relation to fruit setting of the tomato. Bot. Gaz. 130:180-186.

RICK, C. M. and P. G. SMITH, 1953  Novel variation in tomato species hybrids. Am. Nat. 88:359-373.

RICK, C. M., E. KESICKI, J. F. FOBES, and M. HOLLE, 1976  Genetic and biosystematic studies on two new sibling species of Lycopersicon from interandean Peru. Theor. Appl. Genet. 47:55-68.

RICK, C. M., J. F. FOBES, and M. HOLLE, 1977  Genetic variation in Lycopersicon pimpinellifolium: evidence of evolutionary change in mating systems. Plant Syst. Evol. 127:139-170.

RICK, C. M., M. HOLLE, and R. W. THORP, 1978  Rates of cross-pollination in Lycopersicon pimpinellifolium: impact of genetic variation in floral characters. Pl. Syst. Evol. 129:31-44.

SAS INSTITUTE, 1989 SAS/STAT User's Guide, Version 6, Ed. 4. SAS Institute, Cary, NC.

SHERRY, R., 1994 Developmental stability and floral ontogeny in self- and cross-pollinating populations of Clarkia tembloriensis (Onagraceae). Ph.D. Dissertation, University of California, Riverside, CA.

SHORE, J. S. and S. C. H. BARRETT, 1990  Quantitative genetics of floral characters in homostylous Turnera ulmifolia var. angustifolia Willd. (Turneraceae). Heredity 64:105-112.

STANCHEVA, Y., V. RODEVA, and L. STAMOVA, 1997  Multiflor—a tomato line with super large inflorescences. TGC Rep. 47:33.

STEBBINS, G. L., 1957  Self fertilization and population variability in the higher plants. Am. Nat. 91:337-354.

STEBBINS, G. L., 1970  Adaptive radiation of reproductive characteristics in angiosperms. I. Pollination mechanisms. Annu. Rev. Ecol. Syst. 1:307-326.

TAL, M., 1967  Genetic differentiation and stability of some characters that distinguish Lycopersicon esculentum Mill. from Solanum pennellii Corr. Evolution 21:316-333.

TANKSLEY, S. D., and M. A. MUTSCHLER, 1990 Linkage map of the tomato (Lycopersicon esculentum), pp. 6.3–6.15 in Genetic Maps: Locus Maps of Complex Genomes, Book 6, Ed. 5, edited by S. J. O'BRIEN. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY.

TANKSLEY, S. D., M. W. GANAL, J. P. PRINCE, M. C. DE VICENTE, and M. W. BONIERBALE et al., 1992  High-density molecular linkage maps of the tomato and potato genomes. Genetics 132:1141-1160[Abstract].

WEAVER, R. F., and P. W. HEDRICK, 1989 Genetics. William C. Brown, Dubuque, IA.

WEIGEL, D. and E. M. MEYEROWITZ, 1994  The ABCs of floral homeotic genes. Cell 78:203-209[Medline].

WHITKUS, R., J. DOEBLEY, and M. LEE, 1992  Comparative genome mapping of sorghum and maize. Genetics 132:1119-1130[Abstract].

WYATT, R., 1983 Pollinator-plant interactions and the evolution of mating systems, pp. 51–95 in Pollination Biology, edited by L. REAL. Academic Press, Orlando, FL.

This article has been cited by other articles:

				K. Roselius, W. Stephan, and T. Stadler The Relationship of Nucleotide Polymorphism, Recombination Rate and Selection in Wild Tomato Species Genetics, October 1, 2005; 171(2): 753 - 763. [Abstract] [Full Text] [PDF]

				K.-Y. Chen and S. D. Tanksley High-Resolution Mapping and Functional Analysis of se2.1: A Major Stigma Exsertion Quantitative Trait Locus Associated With the Evolution From Allogamy to Autogamy in the Genus Lycopersicon Genetics, November 1, 2004; 168(3): 1563 - 1573. [Abstract] [Full Text] [PDF]

				K. D. Jermstad, D. L. Bassoni, K. S. Jech, G. A. Ritchie, N. C. Wheeler, and D. B. Neale Mapping of Quantitative Trait Loci Controlling Adaptive Traits in Coastal Douglas Fir. III. Quantitative Trait Loci-by-Environment Interactions Genetics, November 1, 2003; 165(3): 1489 - 1506. [Abstract] [Full Text] [PDF]

This Article Abstract Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Georgiady, M. S. Articles by Lord, E. M. Search for Related Content PubMed PubMed Citation Articles by Georgiady, M. S. Articles by Lord, E. M.