Hybrid classification: insights from genetic map-based studies of experimental hybrids.
Studies of the ecological and evolutionary consequences of hybridization have long been hampered by the difficulty of accurately identifying hybrids. Even when hybrids are successfully detected, estimating their genealogy or pedigree may be difficult or impossible. Yet, accurate classification of hybrids is crucial to correct interpretations of hybridization phenomena or the interaction of hybrids with their biotic or abiotic environment. For example, confident documentation of introgressive race formation, or measurements of pest abundance on hybrid and parental genotypic classes, rely on accurate methods of hybrid identification.
Although the classification of hybrids remains problematic, considerable progress has been made over the past 60 years. The first major contributions were by Edgar Anderson, who devised a number of morphometric approaches for detecting and describing hybrids, such as the hybrid index and the scatter diagram (Anderson 1949). Although these methods were useful for describing morphological variation in hybrid populations, it soon became clear that the character correlations and morphological intermediacy detected by these approaches could result from evolutionary phenomena other than hybridization (Dobzhansky 1941, Baker 1947, Barber and Jackson 1957, Heiser 1973). Dobzhansky (1941) recognized that intermediacy could arise from convergent morphological evolution. He also noted that remnants of the ancestral population from which two species differentiated might also exhibit intermediacy, i.e., an early and explicit recognition of symplesiomorphy (shared ancestral characters). Barber and Jackson (1957) recognized that differentiation within a series of populations in continuous contact (primary intergradation) could be difficult to distinguish from zones of hybridization involving secondary contact between previously isolated species (secondary intergradation). As a result, they questioned the assumption that steep clines always result from the merger of previously differentiated populations. Other authors were skeptical of the use of hybrid indices and other biometric tools in the absence of information regarding the genetic basis of the characters being scored (e.g., Baker 1947, Gottlieb 1972, Lamb and Avise 1987, Rieseberg et al. 1988). For example, morphological analysis of genetically characterized treefrog hybrids revealed that [greater than] 40% of individuals with a known hybrid ancestry would have been misclassified as "pure" parental species, based on morphology (Lamb and Avise 1987). Likewise, a review of 46 studies reporting morphological character expression in plant hybrids (Rieseberg and Ellstrand 1993) revealed that only 45% of morphological characters displayed "intermediate" expression in first generation hybrids; the remaining characters were either the same as one parent or the other (45%), or extreme relative to either parent (10%). By contrast, Floate et al. (1994) report surprisingly strong correlations between morphology and genotype of hybrid cottonwoods, suggesting that in some instances morphology can be a reliable indicator of hybrid ancestry. However, the diversity of genotypes included in the Floate et al. analysis is unclear, making it difficult to evaluate their result.
Despite the uncertainties associated with morphological classification of hybrids, many studies continue to rely on this approach, including studies of herbivore resistance (e.g., Boecklen and Spellenberg 1990, Siemens et al. 1994, Whitham et al. 1994). Clearly, the value of morphology as a predictor of genotype will continue to attract study for the foreseeable future.
Modern contributions to the identification of hybrids have been primarily methodological. Most important has been the development of molecular markers, which have proved to be extremely sensitive tools for the detection of hybrids and for the dissection of hybrid genomes (e.g., Levin 1975, Keim et al. 1989, Harrison 1990, Arnold 1992, Rieseberg et al. 1996a, Howard et al. 1997). Advantages of molecular markers often cited include (1) the large number of independent markers available for analysis, (2) simple genetic control, (3) low levels of nonheritable variation, (4) apparent selective neutrality, and (5) the ability to distinguish between maternal and paternal parents (e.g., Palmer et al. 1983, Powell 1983). Several authors also have noted the advantages of employing physically linked markers for distinguishing between morphological patterns resulting from hybridization and those resulting from primary intergradation, convergence, or symplesiomorphy (e.g., Avise and Saunders 1984, Doebley 1989, Rieseberg et al. 1990).
In addition to providing rigorous documentation of both recent (primary) and ancient (secondary) hybrids, multilocus molecular data sets are now being used to discriminate among hybrid categories (e.g., Ellstrand et al. 1987, Paige et al. 1991, Nason et al. 1992, Paige and Capman 1993). One of the difficulties with this approach is that hybrid and parental genotypic classes often differ minimally in terms of expected marker proportions (Nason and Ellstrand 1993, Floate et al. 1994, Boecklen and Howard 1997). For example, when a small number of markers are used, some [F.sub.2]s are expected by chance to have the same multilocus molecular genotype as [F.sub.1]s, and certain [BC.sub.1]s are likely to be identical in molecular constitution to [F.sub.1]s or parental individuals (Rieseberg and Ellstrand 1993). Nason et al. (1992) have developed maximum-likelihood methods that partially solve this problem by providing estimates of the population-wide frequencies of early genealogical classes of hybrids. However, the pedigree of individual hybrids remains uncertain. Another potential limitation of multilocus approaches to genealogical reconstruction concerns assumptions of marker neutrality. Although most molecular markers used in these studies are indeed likely to be neutral, they may be physically linked to genic or chromosomal factors that affect hybrid fitness. Thus, natural selection could bias marker distributions, leading to faulty genealogical assignments.
The purpose of this paper is to examine the distribution of molecular markers in experimentally synthesized hybrids between two wild sunflower species, Helianthus annuus and H. petiolaris. We will demonstrate that selection has led to significant deviations from neutral patterns of inheritance in these sunflower hybrids and will argue that multilocus marker data probably will not allow precise reconstruction of hybrid genealogies beyond the first segregating generation. Evidence also will be presented demonstrating that selection can greatly reduce the diversity of hybrid genotypes in populations and that the surviving hybrids may represent only a subset of the genotypic diversity expected under conditions of random mating. Although the implications of these data for genealogical reconstruction are discouraging, we will argue that it is the actual genetic constitution of hybrids, not their genealogy, that is most predictive of their characteristics or behavior.
MATERIALS AND METHODS
Helianthus annuus and H. petiolaris are self-incompatible annual sunflower species with the same chromosome number (n = 17). Both species are native to North America and are abundant in the western United States. Comparative genetic linkage mapping (Rieseberg et al. 1995b) indicates that seven linkage groups appear to be collinear between the two species, although this does not rule out the possibility of small undetected rearrangements. By contrast, the remaining 10 linkages differ structurally due to a minimum of seven large interchromosomal translocations and three large inversions. These structural changes generate multivalent formations, as well as bridges and fragments, in hybrids (Heiser 1947, Chandler et al. 1986), apparently leading to semisterility; [F.sub.1] pollen viabilities are typically [less than] 10%, and seed set is [less than] 1% (Heiser 1947, Chandler et al. 1986). Nonetheless, fertility is rapidly restored in later generation hybrids and backcrosses (Heiser 1947). In addition to the chromosomal rearrangements, patterns of introgression suggest that genic factors also contribute to reduced hybrid fertility or viability (Rieseberg et al. 1995a, b).
As part of an ongoing study of hybrid speciation, three hybrid lineages were generated between H. annuus and H. petiolaris: lineage I, P-[F.sub.1]-[BC.sub.1]-[BC.sub.2]-[F.sub.2]-[F.sub.3]; lineage II, P-[F.sub.1]-[F.sub.2]-[BC.sub.1]-[BC.sub.2]-[F.sub.3]; and lineage III, P-[F.sub.1][F.sub.2]-[F.sub.3]-[BC.sub.1]-[BC.sub.2] (Rieseberg et al. 1996a). The initial interspecific cross was H. annuus (cmsHA89; female) x H. petiolaris subsp. petiolaris (PET-PET-1741-1; male; Seiler 1991). The [F.sub.1] hybrids were either backcrossed in the direction of H. annuus, the maternal parent, or crossed among themselves (sib-mated). This process was continued sequentially for an additional three generations with the only difference among the three lineages being the order of the sib-cross vs. backcross generations. For example, in hybrid lineage I, [F.sub.1] s were initially backcrossed to H. annuus. The resulting [BC.sub.1] generation was also backcrossed to H. annuus. The [BC.sub.2] generation derived from this cross was allowed to sibmate, as was the [BC.sub.2][F.sub.2] generation. The product of these crosses (the [BC.sub.2][F.sub.3] generation) was used for the analyses we describe. At least 20 plants were used for each generation. Crosses were performed by applying pooled pollen from all plants from a given generation to stigmas of the same individuals. All achenes from each generation were pooled, and [much greater than] 30 achenes were arbitrarily chosen as founders of the next generation.
Marker surveys and graphical genotype construction
DNA was isolated from 56 or 58 plants representing the final generation of each of the hybrid lineages ([BC.sub.2][F.sub.3] plants) and surveyed for 197 mapped random amplified polymorphic DNA (RAPD) markers (Rieseberg et al. 1996a, b). These markers cover [greater than]80% ([approximately]1160 cM) of the sunflower genome currently mapped (Berry et al. 1995, Gentzbittel et al. 1995, Rieseberg et al. 1995b), with an average distance of 6.5 cM between markers based on H. annuus map distances (Rieseberg et al. 1995a, b).
Graphical genotype construction is described in detail in Rieseberg et al. (1995a). Graphical genotypes are graphical representations of the genomic constitution and parental derivation for all points on a genome (Young and Tanksley 1989). Briefly, introgressed markers of the donor parent (H. petiolaris) were plotted onto the genomic map of the recipient species (H. annuus), generating a graphical genotype for each of the 170 progeny. The presence of two adjacent introgressed markers on the graphical genotype of a single individual was taken as evidence that the entire fragment between the markers was derived from H. petiolaris through introgression. However, because the RAPD markers employed here are largely dominant, we often were unable to determine the linkage phase of adjacent markers. As a result, the possibility that small proportions of adjacent introgressed markers are in repulsion phase (i.e., on different homologues) cannot be ruled out, and the size of the introgressed segment(s) in these situations might be considerably smaller. Likewise, because of dominance, we sometimes were unable to determine whether the introgressed markers or fragments were homozygous or heterozygous. Nonetheless, because the hybrid lineages included two backcross generations, and because of the low frequency of many introgressed markers, it is likely that most introgressed markers are heterozygous and coupled. Moreover, because hybrid lineage III ends in a backcross generation, we know that at least in this lineage, all introgressed markers are heterozygous and coupled.
Simulations of neutral introgression
Previous analyses of this data set focused on the frequency of introgression of individual H. petiolaris markers over all progeny (Rieseberg et al. 1996a, b), rather than on the genotypes of individual plants. To determine whether the genotypes of individual plants from the three hybrid lineages differed from expectations under neutral conditions, we performed simulations of unrestricted marker introgression over the entire genome, as well as for the collinear and rearranged portions of the genome separately. This method assumes that markers are neutral and that recombination rates do not vary across hybrid generations. Because of linkage, the number of independently introgressing markers on a chromosome was less than the total number of markers identified (197 in each lineage).
We took two approaches to account for linkage relationships of markers on the same chromosome. In the first and most conservative case, we assumed that entire chromosomes would be inherited as unrecombined linkages, reducing the number of markers that could introgress independently to the number of chromosomes (n = 17 for the entire genome, n = 7 for collinear linkage groups, and n = 10 for rearranged linkage groups). This simulation is especially conservative for the collinear linkage groups that would be expected to recombine freely during the four generations following the [F.sub.1]. Hence, although this simulation provides less power than is available in the data set, detection of significant deviations from it in the observed data indicates very strong support for lack of marker neutrality.
In the second set of simulations, the number of independent markers on a linkage group was determined by the minimum number of independently introgressing regions observed for all three hybrid lineages (Table 1). Adjacent blocks of markers that differ completely in their behavior (i.e., one block introgresses, whereas the other does not) were considered to be effectively independent. Hence, we simply counted the number of introgressing and nonintrogressing blocks for each linkage group. For example, linkage group A [ILLUSTRATION FOR FIGURE 1 OMITTED] has four independent regions. There are two regions into which linked H. petiolaris markers have introgressed that are consistently separated by a marker that always failed to introgress. In addition there is a single marker following the second set of introgressing markers that always failed to introgress.
For each simulation (entire linkage groups or independent blocks within linkage groups), the independent markers in each individual were allowed to introgress at random. Since we knew the average probability that a marker would introgress under unrestricted introgression for each lineage (lineage I: 0.156, lineage II: 0.1875, lineage III: 0.25), a random number between zero and one was generated for each independent marker in an individual. If the random number was less than or equal to the probability of introgression, the marker was scored as introgressed. This process was repeated for the number of individuals in the progeny array, and the entire simulation was replicated 100 times for each hybrid lineage (Table 1). These simulations allowed us to generate null-hypothesis distributions of the numbers of markers that should introgress per individual and to compare our actual results with those of the null hypothesis.
TABLE 1. Number of markers per linkage group, number of independent blocks per linkage group, and number of markers per block. Number of Number of Number of Linkage group markers blocks markers per block Collinear A 6 4 2, 1, 2, 1 B 7 2 6, 1 C 10 5 1, 3, 2, 3, 1 D 8 3 2, 4, 2 E 8 4 2, 1, 4, 1 F 7 3 3, 1, 3 S 12 3 5, 4, 3 Total 58 24 Rearranged L 9 3 6, 2, 1 HK 15 1 15 IJ 14 1 14 NO 26 1 26 M 8 1 8 PWG 32 1 32 U 6 1 6 RQ 12 1 12 T 11 1 11 V 6 3 2, 2, 2 Total 139 14
The number of markers actually scored per linkage group was much greater than the number of independent markers in our unrestricted, neutral simulations of introgression. Hence, to compare our null-model results with the observed patterns of introgression, it was necessary to reduce the number of observed markers available for introgression to either the number of linkage groups or independent blocks. The following procedure was performed on the observed markers for each individual in the progeny arrays of the three hybrid lineages. For each linkage group or independent fragment, a marker was chosen at random, and whether the marker was observed to have introgressed was noted. After the process of choosing markers had been performed on all of the linkage groups/blocks, the total number of markers introgressing per individual was recorded. This process was repeated 300 times for each individual in the progeny arrays. From the results, we calculated the average number of markers introgressing per individual and their distributions (standard deviations). These were compared with the values from the unrestricted introgression simulations. If the observed markers were behaving in accordance with unrestricted, neutral introgression, half of the individuals should have had average levels of introgression above the average null-model levels of introgression. This expectation was tested using [X.sub.2] statistics. Marker neutrality was rejected if the statistic was significant at the [Alpha] = 0.05 level.
Over the entire genome, comparisons of observed and expected genotypes of individual plants from the three hybrid lineages revealed that the majority of individuals had fewer H. petiolaris markers than expected (Table 2, [ILLUSTRATION FOR FIGURE 1 OMITTED]). This was true regardless of hybrid lineage and whether the number of independently introgressing markers was estimated conservatively (the number of chromosomes or linkage groups) or based upon observed numbers of independent blocks. In all cases, the average number of markers observed to introgress was less than the expectation (Table 3), a highly improbable result if markers were introgressing neutrally.
[TABULAR DATA FOR TABLE 2 OMITTED]
At first glance, the lower than expected frequency of H. petiolaris markers in these hybrid individuals appears to result largely from selection against H. petiolaris chromosomes that differ from H. annuus by translocations and inversions (Rieseberg et al. 1995a, b). As a result, numbers of introgressed markers from the rearranged portion of the genome deviated strongly and consistently from expected means for individuals in all three lineages, no matter the nature of the null models used (Tables 2 and 3, [ILLUSTRATION FOR FIGURE 1 OMITTED]). By contrast, numbers of introgressed markers per individual from collinear linkages presented a more complex picture. Under the conservative null model of unrestricted introgression, the collinear linkages of lineage II consistently had higher numbers of markers introgressing than expected, while lineages I and III had numbers of markers that could not be distinguished from the null model (Tables 2 and 3, [ILLUSTRATION FOR FIGURE 1 OMITTED]). When the observed levels of introgression were compared with the less conservative null model, the situation was strongly different. Individuals in lineages I and III consistently had less introgression than expected and individuals in lineage II had levels of introgression indistinguishable from expectation (Tables 2 and 3). If the markers on the collinear linkages were introgressing neutrally, the results from both the chromosomal and independent-fragment analyses should have been consistent with expectation. This is clearly not the case.
TABLE 3. Chi-square tests of deviations from expectation that the average observed numbers of markers introgressing in individuals was distributed equally above and below the expected values. Region Lineage I Lineage II Lineage III Entire genome Linkage groups 58(***) 52.07(***) 56(***) Blocks 58(***) 8.28(***) 56(***) Collinear Linkage groups 0.07 18.28(***) 1.14 Blocks 17.66(***) 1.78 44.64(***) Rearranged Linkage groups 58(***) 56(***) 56(***) Blocks 58(***) 56(***) 56(***) Note: For all tests, df = 1. *** P [less than] 0.001.
In order to understand the behavior of the collinear linkages, closer examination of the distribution of introgressed markers in collinear linkages was undertaken. Previous work (Rieseberg et al. 1996a, b) has shown that the patterns of introgression are much more similar among the three hybrid lineages than would be expected by chance and that many individual markers or linkage blocks were either completely eliminated or introgressed at higher than expected rates in all three lineages [ILLUSTRATION FOR FIGURE 2 OMITTED]. Under neutral conditions, all of the H. petiolaris markers in the collinear linkages should have introgressed and should have done so at nearly the rates predicted by the null hypothesis (Rieseberg et al. 1995b). Selection, or some other mechanism, seems to have eliminated or favored nearly identical sets of markers in all three lineages.
In lineages I and III, it is likely that the chromosomal null model lacked sufficient power to reveal the patterns of lower than expected introgression seen when testing the larger number of independent blocks. The case of lineage II is more difficult to explain, as it went from significantly higher levels of introgression at the chromosomal level to levels insignificantly different from expectation when examined on an independent-fragment basis. Lack of power cannot be invoked here. It is more likely that regions that had higher than expected levels of introgression balanced the effect of regions that had lower than expected levels of introgression. Such an outcome could easily be the natural consequence of determining independent blocks in the manner that we did. Although other explanations could account for the patterns of introgression observed in the collinear linkages, neutral introgression is clearly not one of them.
A predicted benefit of virtually unlimited numbers of molecular markers was the ability to accurately detect and classify hybrids (e.g., Harrison 1990, Rieseberg and Brunsfeld 1992). Although hybrid detection has indeed been greatly enhanced by these markers, accurate genealogical classification has been less successful for several reasons (although see Nason et al. 1992, Nason and Ellstrand 1993). First, hybrid and parental genotypic classes often differ minimally in terms of expected marker numbers. Thus, an extremely large number of markers are required to accurately distinguish between them, and assignments by inspection will provide biased estimates of the frequencies of certain classes of hybrids (Nason and Ellstrand 1993, Boecklin and Howard 1997). Second, the most easily developed molecular markers are often dominant (e.g., RAPDs and amplified fragment length polymorphisms [AFLPs]) and, therefore, lack information about heterozygosity. As a result, hybrid classes that differ largely in heterozygote proportions rather than parental marker frequencies (e.g., [F.sub.2]s vs. [F.sub.3]s) may be difficult to discriminate. Although both of these issues complicate hybrid classification, maximum-likelihood methods have been developed and successfully used to estimate the frequencies of different classes of early generation hybrids, using either co-dominant or dominant molecular markers (Nason et al. 1992, Nason and Ellstrand 1993; J. D. Nason, personal communication).
A third and more fundamental problem, and the focus of this paper, is selection. As we have illustrated with interspecific sunflower hybrids, selection can quickly lead to large deviations from expected marker numbers. All 170 hybrid plants had fewer H. petiolaris markers than the expected means. In fact, with the exception of two individuals, marker numbers were more consistent with pedigrees involving three to seven backcross generations rather than the two generations of backcrossing actually employed.
The deviations in marker proportions observed in these synthetic sunflower hybrids appear to result from fertility selection. First generation hybrids exhibit pollen viabilities of [less than]10% and seed set is [less than]1% (Heiser 1947). Furthermore, intraspecific pollen has been shown to out-compete interspecific pollen. Our crossing strategy employed pollen and seed pooling, which appears to have led to rapid selection for the most fertile hybrids in terms of pollen viability, seed set, and pollen tube growth rates; uniformly high fertility ([greater than]90% pollen viability) was recovered by the fifth generation in all three hybrid lineages. Because all backcrosses were performed in the direction of H. annuus, this strong fertility selection should lead to the loss of H. petiolaris genes or chromosomal rearrangements that reduce hybrid fertility. As predicted, very few H. petiolaris markers from rearranged linkage groups were recovered in hybrids. Twenty-two markers from collinear linkage groups were absent from all 170 hybrids as well. However, as discussed in the Results, higher than expected rates of introgression for other markers from collinear linkage groups appear to have partially compensated for the loss of the 22 markers. Thus, overall deviations in the numbers of H. petiolaris markers found in hybrid plants appear to be almost entirely due to the effects of chromosomal rearrangements. Possibly, fertility selection will be less of a problem for the reconstruction of hybrid genealogies in species that do not differ in terms of chromosome structure or that are simply less differentiated genetically.
This is not the first evidence that particular genomic regions or even entire genomes can be rapidly eliminated in hybrids by selection. For example, Stephens (1949) noted that in backcrosses between Gossypium species, the donor parent genotype is selectively eliminated, regardless of the direction of the backcrosses. Similar observations have been made for species hybrids in Antirrhinum (Mather 1947), Melilotus (Baenziger and Greenshields 1958), Lycopersicon (Rick 1963), Zea (Mangelsdorf 1958), and Phaseolus (Wall 1968). In fact, skewed segregation ratios in hybrids now appear to be the rule rather than the exception. For example, Zamir and Tadmor (1986) report segregation distortion in 54% of loci from interspecific crosses of Lenz, Capiscum and Lycopersicon, compared to only 13% in intraspecific crosses. In Helianthus, segregation distortion has been observed at 7-13% of loci in intraspecific mapping populations (Rieseberg et al. 1993, Berry et al. 1995, Gentzbittel et al. 1995) compared to 23-90% of loci in interspecific crosses (Quillet et al. 1995, Rieseberg et al. 1995b, 1996b). We can think of no examples of interspecific mapping populations in which distorted ratios have not been reported.
Not only are distorted ratios prevalent, they can also be extreme. For example, segregation ratios skewed 12:1 in favor of "wild" alleles have been reported in crosses between cultivated pearl millet (Pennisetum glaucum) and one of its wild relatives (P. violaceum) (Liu et al. 1996). Likewise, map-based studies of introgression often reveal the unidirectional elimination of many genomic regions in backcross lines (e.g., Jena et al. 1992, Williams et al. 1993, McGrath et al. 1994, Garcia et al. 1995, Rieseberg et al. 1995b, 1996b, Wang et al. 1995). An overall result of these skewed segregation ratios is that hybrid progeny receive more alleles from one parent than would be expected under Mendelian rules of segregation and thus resemble that parent more closely than Mendelian rules would predict.
There are many plausible explanations for skewed segregation ratios. In most cases, deviations probably result from gamete lethality and selective fertilization of surviving gametes (Levin 1975). Typically, the most highly recombinant gametes are eliminated (Rieseberg et al. 1996b). Differential zygote survival, meiotic drive, and preferential segregation phenomena also appear to affect genetic transmission (e.g., Palopoli and Wu 1996), but appear to be less important than gametic survivorship and fertilization. Underlying causes of these phenomena include both large and small chromosomal rearrangements (e.g., Stephens 1949, Quillet et al. 1995), sterility genes (e.g., Wan et al. 1996), segregation distorter loci (i.e., selfish genes; Palopoli and Wu 1996), and the breakup of co-adapted gene complexes (Harlan 1936, Dobzhansky 1941, Mayr 1954, Rick 1963, Carson and Templeton 1984).
Although the vast majority of deviating ratios observed in species backcrosses have favored the genes of the recurrent parent, there have been several exceptions to this general rule. For example, the white lint gene of the donor parent was favored over brown lint alleles of the recipient parent in backcrosses from Gossypium barbadense into G. hirsutum (Stephens 1949). Likewise, 5% of H. petiolaris markers introgressed at significantly higher than predicted rates into an H. annuus genetic background (Rieseberg et al. 1996a). This is a relatively small fraction, however, when compared to the 85% that introgressed at significantly lower than expected rates (Rieseberg et al. 1996a). Finally, Wang et al. (1995) noted that the same Gossypium hirsutum chromosome fragments were maintained in independently generated G. barbadense introgression lines. It is not clear whether these loci or chromosomal fragments are selectively favored in the recurrent parent or whether they represent examples of "selfish genes"; i.e., genes that enhance the success of gametes they inhabit, even if they pose a significant fitness cost during the diploid phase of the life cycle (Haldane 1932).
So far this discussion has focused on the effects of endogenous selection (particularly gametic selection) on the genetic constitution of individual hybrid plants. In natural populations, however, other forces such as environmental selection and assortative mating affect the abundance and distribution of hybrid genotypes. Although these forces are unlikely to directly affect hybrid genomic composition, they do appear to reduce overall genotypic diversity in hybrid populations. In Louisiana irises, for example, environmental selection, assortative mating, and viability selection all appear to have contributed to the formation of hybrid genotypes that tend to be similar to the parental species; intermediate individuals have been eliminated (Cruzan and Arnold 1993, 1994). By contrast, mixed populations of Quercus kelloggii and Q. wislizenii var. frutescens are restricted to [F.sub.1] hybrids and parental individuals, apparently due to competition with parental species (Nason et al. 1992).
Endogenous selection also appears to reduce genotypic diversity in hybrid populations. For example, after only five generations of hybridization, individuals from the three sunflower hybrid lineages converged onto essentially identical gene combinations ([ILLUSTRATION FOR FIGURE 2 OMITTED]; Rieseberg et al. 1996a). Similar convergent patterns have also been observed for Gossypium introgression lines (Wang et al. 1995) and chromosomal genotypes of Caledia grasshopper hybrids (Shaw et al. 1982, 1983). Thus, hybrid populations are predicted to maintain only a fraction of the genotypic diversity expected under conditions of random mating.
Although the data presented here suggest that reconstruction of hybrid genealogy may be difficult or impossible, there are several factors that reduce the complexity of this problem. First, segregation distortion will be most pronounced in wide species crosses and will affect the genotypes of later generation hybrids most severely. It may be possible, therefore, to accurately estimate the frequency of first and second generation hybrid classes between closely related species that lack chromosomal rearrangements, and the methods of Nason and Ellstrand (1993) provide an elegant approach for accomplishing this goal. Second, most studies of hybrids do not require a precise pedigree. Often, hybrid classification simply provides the basis for ecological studies of different genotypic classes (e.g., Paige et al. 1990, Aguilar and Boecklen 1992, Fritz et al. 1994, 1996, Christensen et al. 1995). In these studies, the actual genotype of the hybrid plant is more critical than the historical process by which it was derived. For example, if an [F.sub.6] plant has converged onto the same multilocus genotype as a [BC.sub.4] plant, the critical fact for ecological studies is that they are identical genetically, not their different pedigrees. Only the former will be predictive of the hybrid plants' phenotype and behavior.
One solution to this problem would be to classify plants from hybrid populations based on overall genetic relatedness (Lynch 1990, Lynch and Milligan 1994, Weir 1996). Multivariate methods or various clustering algorithms (e.g., Sneath and Sokal 1973, Saitou and Nei 1987) are available for objectively grouping related individuals. This approach is eminently feasible given the large number of molecular markers that can be easily developed for almost any organismal group. Moreover, classifying hybrids based on genetic relatedness would correct the likely problem that individuals with the same pedigree might differ extensively in terms of parental gene combinations. For example, it is theoretically possible (although highly unlikely) for two [F.sub.4] individuals to have completely nonoverlapping distributions of parental-species genes.
Alternatively, it may be useful to classify hybrids based on the overall proportions or admixture of parental-species genes they carry. This approach can implemented easily as well, since maximum-likelihood methods are available for calculating genetic admixtures (Wheeler and Guries 1987) or hybrid indices (Rieseberg et al. 1998), with either codominant or dominant markers (S. Baird, personal communication). As with estimates of relatedness, species-specific markers are not required for these calculations, although they do increase their power (Rieseberg et al. 1998).
In conclusion, attempts to infer the genealogies of later generation hybrids or even early generation hybrids between genetically divergent species will be difficult or impossible due to selection, regardless of the number of molecular markers used. Fortunately, most ecological studies require genotypic, rather than pedigree, information. Thus, we suggest that hybrids be classified either in terms of genetic relatedness or in terms of the admixture of parental-species genes they carry. Both represent simple approaches for generating biologically cohesive categories for ecological or comparative studies.
This research was supported by NSF grant BSR-9419206 and USDA grant 92373007590 to L. H. Rieseberg and NSF grant BIR-9411128 to C. R. Linder.
Aguilar, J. M., and W. J. Boecklen. 1992. Patterns of herbivory in the Quercus grisea x Quercux gambelii species complex. Oikos 64:498-504.
Anderson, E. 1949. Introgressive hybridization. John Wiley, New York, New York, USA.
Arnold, M. L. 1992. Natural hybridization as an evolutionary process. Annual Review of Ecology and Systematics 23: 237-261.
Avise, J. C., and N. C. Saunders. 1984. Hybridization and introgression among species of sunfish (Lepomis): analysis of mitochondrial DNA variability in a hybrid swarm between subspecies of bluegill sunfish (Lepomis macrochirus). Evolution 38:931-941.
Baenziger, H., and J. E. R. Greenshields. 1958. The effect of interspecific hybridization on certain genetic ratios in sweet clover. Canadian Journal of Botany 36:411-420.
Baker, H. G. 1947. Criteria of hybridity. Nature 159:221-223.
Barber, N. H., and W. D. Jackson. 1957. Natural selection in action in Eucalyptus. Nature 179:1267-1269.
Berry, S. T., A. J. Leon, C. C. Hanfrey, P. Challis, A. Burkholz, S. R. Barnes, G. K. Rufener, M. Lee, and P. D. S. Caligari. 1995. Molecular marker analysis of Helianthus annuus L. 2. Construction of a RFLP linkage map for cultivated sunflower. Theoretical and Applied Genetics 91: 195-199.
Boecklen, W. J., and D. J. Howard. 1997. Genetic analysis of hybrid zones: numbers of markers and power of resolution. Ecology 78:2611-2616.
Boecklen, W. J., and R. Spellenberg. 1990. Structure of herbivore communities in two oak (Quercus spp.) hybrid zones. Oecologia 85:92-100.
Carson, H. L., and A. R. Templeton. 1984. Genetic revolutions in relation to speciation phenomena: the founding of new populations. Annual Review of Ecology and Systematics 15:97-131.
Chandler, J. M., C. Jan, and B. H. Beard. 1986. Chromosomal differentiation among the annual Helianthus species. Systematic Botany 11:353-371.
Christensen, K. M., T. G. Whitham, and P. Keim. 1995. Herbivory and tree mortality across a pinyon pine hybrid zone. Oecologia 101:29-36.
Cruzan, M. B., and M. L. Arnold. 1993. Ecological and genetic associations in an Iris hybrid zone. Evolution 47: 1432-1445.
Cruzan, M. B., and M. L. Arnold. 1994. Assortative mating and natural selection in an Iris hybrid zone. Evolution 48: 1946-1958.
Dobzhansky, T. H. 1941. Genetics and the origin of species. Columbia University Press, New York, New York, USA.
Doebley, J. 1989. Molecular evidence for a missing wild relative of maize and the introgression of its chloroplast genome into Zea perennis. Evolution 43:1555-1559.
Ellstrand, N. C., J. M. Lee, J. E. Keeley, and S.C. Keeley. 1987. Ecological isolation and introgression: biochemical confirmation of introgression in an Arctostaphylos (Ericaceae) population. Acta Oecologia/Oecologia Plantarum 8: 299-308.
Floate, K. D., T. G. Whitham, and P. Keim. 1994. Morphological vs. genetic markers in classifying hybrid plants. Evolution 48:929-930.
Fritz, R. S., C. M. Nichols-Orians, and S. J. Brunsfeld. 1994. Interspecific hybridization of plants and resistance to herbivores: hypotheses, genetics, and variable responses in a diverse community. Oecologia 97:106-117.
Fritz, R. S., B. M. Roche, S. J. Brunsfeld, and C. M. Orians. 1996. Interspecific and temporal variation in herbivore responses to hybrid willows. Oecologia 108:121-129.
Garcia, G. M., H. T. Stalker, and G. Kocherr. 1995. Introgression analysis of an interspecific hybrid population in peanuts (Arachis hypogaea L.) using RFLP and RAPD markers. Genome 38:166-176.
Gentzbittel, L., F. Vear, Y.-X. Zhang, A. Berville, and P. Nicolas. 1995. Development of a consensus linkage RFLP map of cultivated sunflower (Helianthus annuus L.). Theoretical and Applied Genetics 90:1079-1086.
Gottlieb, L. D. 1972. Levels of confidence in the analysis of hybridization in plants. Annals of the Missouri Botanical Garden 59:435-446.
Haldane, J. B. S. 1932. The causes of evolution. Princeton University Press, 1990 edition, Princeton, New Jersey, USA.
Harlan, S. C. 1936. The genetical conception of the species. Biological Review 11:83-112.
Harrison, R. G. 1990. Hybrid zones: windows on evolutionary process. Oxford Surveys in Evolutionary Biology 7: 69-128.
Heiser, C. B. 1947. Hybridization between the sunflower species Helianthus annuus and H. petiolaris Evolution 1: 249-262.
-----. 1973. Introgression re-examined. Botanical Review 39:347-366.
Howard, D. J., R, W. Preszler, J. Williams, S. Frenchel, and W. J. Boecklen. 1997. How discrete are oak species? Insights from a hybrid zone between Quercus grisea and Q. gambelii. Evolution 51:747-755.
Jena, K. K., G. S. Khush, and G. Kocherr. 1992. RFLP analysis of rice (Oryza sativa L.) introgression lines. Theoretical and Applied Genetics 84:608-616.
Keim, P., K. N. Paige, T G. Whitham, and K. G. Lark. 1989. Genetic analysis of an interspecific hybrid swarm of Populus: occurrence of unidirectional introgression. Genetics 123:557-565.
Lamb, T., and J. C. Avise. 1987. Morphological variability in genetically defined categories of anuran hybrids. Evolution 41:157-165.
Levin, D. A. 1975. Interspecific hybridization, heterozygosity and gene exchange in Phlox. Evolution 29:37-51.
Liu, C. J., K. M. Devos, J. R. Witcombe, T. S. Pittaway, and M.D. Gale. 1996. The effect of genome and sex on recombination rates in Pennisetum species. Theoretical and Applied Genetics 93:902-908.
Lynch, M. 1990. The similarity index and DNA fingerprinting. Molecular Biology and Evolution 7:478-494.
Lynch, M., and B. G. Milligan. 1994. Analysis of population genetic structure with RAPD markers. Molecular Ecology 3:91-99.
Mangelsdorf, P. C. 1958. The mutagenic affect of hybridizing maize and
teosinte. Cold Spring Harbor Symposium on Quantitative Biology 23:409-421.
Mather, K. 1947. Species crosses in Antirrhinum. I. Genetic isolation of species majus glutinosum and orontium. Heredity 1:175-186.
Mayr, E. 1954. Change of genetic environment and evolution. Pages 157-180 in J. Huxley, A. C. Hardy, and E. B. Ford, editors. Evolution as a process. George Allen and Unwin, London, UK.
McGrath, J. M., S. M. Wielgus, T. G. Uchytil, H. Kim-Lee, G. T. Haberlach, C. E. Williams, and J.P. Helgeson. 1994. Recombination of Solanum brevidens chromosomes in the second backcross generation from a somatic hybrid with S. tuberosum. Theoretical and Applied Genetics 88:917-924.
Nason, J. D., and N. C. Ellstrand. 1993. Estimating the frequencies of genetically distinct classes of individuals in hybridized populations. Journal of Heredity 84:1-12.
Nason, J. D., N. C. Ellstrand, and M. L. Arnold. 1992. Patterns of hybridization and introgression in populations of oaks, manzanitas, and irises. American Journal of Botany 79:101-111.
Paige, K. N., and W. C. Capman. 1993. The effects of host-plant genotype, hybridization, and environment on gall aphid attack and survival in cottonwood: the importance of genetic studies and the utility of RFLPs. Evolution 47: 36-45.
Paige, K. N., W. C. Capman, and P. Jennetten. 1991. Mitochondrial inheritance patterns across a cottonwood hybrid zone: cytonuclear disequilibria and hybrid zone dynamics. Evolution 45:1360-1369.
Paige, K. N., P. Keim, T. G. Whitham, and K. G. Lark. 1990. The use of restriction fragment length polymorphisms to study the ecology and evolutionary biology of aphid plant interactions. Pages 69-87 in R. D. Eikenbary and R. K. Campbell, editors. Mechanisms of aphid-plant genotype interactions. Elsevier, Amsterdam, The Netherlands.
Palmer, J. D., C. R. Shields, D. B. Cohen, and T. J. Orton. 1983. Chloroplast DNA evolution and the origin of amphidiploid Brassica species. Theoretical and Applied Genetics 65:181-189.
Palopoli, M. F., and C.-I. Wu. 1996. Rapid evolution of a co-adapted gene complex: evidence from the segregation distorter (SD) system of meiotic drive in Drosophila melanogaster. Genetics 143:1675-1688.
Powell, J. R. 1983. Interspecific cytoplasmic gene flow in the absence of nuclear gene flow: evidence from Drosophila. Proceedings of the National Academy of Sciences USA 80:492-495.
Quillet, M. C., N. Madjidian, T. Griveau, H. Serieys, M. Tersac, M. Lorieus, and A. Berville. 1995. Mapping genetic factors controlling pollen viability in an interspecific cross in Helianthus section Helianthus. Theoretical and Applied Genetics 91:1195-1202.
Rice, W. R. 1989. Analyzing tables of statistical tests. Evolution 43:223-225.
Rick, C. M. 1963. Differential zygotic lethality in a tomato species hybrid. Genetics 48:1497-1507.
Rieseberg, L. H., D. M. Arias, M. Ungerer, C. R. Linden and B. Sinervo. 1996b. The effects of mating design on introgression between chromosomally divergent sunflower species. Theoretical and Applied Genetics 93:633-644.
Rieseberg, L. H., S. Baird, and A. Desrochers. 1998. Patterns of mating in wild sunflower hybrid zones. Evolution, 52: 713-726.
Rieseberg, L. H., and S. B. Brunsfeld. 1992. Molecular evidence and plant introgression. Pages 151-176 in D. E. Soltis, P. S. Soltis, and J. J. Doyle, editors. Molecular systematics of plants. Chapman and Hall, New York, New York, USA.
Rieseberg, L. H., R. Carter, and S. Zona. 1990. Molecular tests of the hypothesized hybrid origin of two diploid Helianthus species (Asteraceae). Evolution 44:1498-1511.
Rieseberg, L. H., H. Choi, R. Chan, and C. Spore. 1993. Genomic map of a diploid hybrid species. Heredity 70:285-293.
Rieseberg, L. H., and N. C. Ellstrand. 1993. What can morphological and molecular markers tell us about plant hybridization. Critical Reviews in Plant Science 12:213-241.
Rieseberg, L. H., C. R. Linder, and G. Seller. 1995b. Chromosomal and genic barriers to introgression in Helianthus. Genetics 141:1163-1171.
Rieseberg, L. H., B. Sinervo, C. R. Linder, M. C. Ungerer, and D. M. Arias. 1996a. Role of gene interactions in hybrid speciation: evidence from ancient and experimental hybrids. Science 272:741-745.
Rieseberg, L. H., D. E. Soltis, and J. D. Palmer. 1988. A molecular reexamination of introgression between Helianthus annuus and H. bolanderi. Evolution 42:227-238.
Rieseberg, L. H., C. Van Fossen and A. Desrochers. 1995a. Hybrid speciation accompanied by genomic reorganization in wild sunflowers. Nature 375:313-316.
Saitou, N., and M. Nei. 1987. The neighbor-joining method: a new method of reconstructing phylogenetic trees. Molecular Biology and Evolution 4:406-425.
Seiler, G. J. 1991. Registration of 15 interspecific sunflower germplasm lines derived from wild annual species. Crop Science 31:1389-1390.
Shaw, D. D., P. Wilkinson, and D. J. Coates. 1982. The chromosomal component of reproductive isolation in the grasshopper Caledia capriva. II. The relative viabilities of recombinant and nonrecombinant chromosomes during embryogenesis. Chromosoma 86:533-549.
Shaw, D. D., P. Wilkinson, and D. J. Coates. 1983. Increased chromosomal mutation rate after hybridization between two subspecies of grasshoppers. Science 220:1165-1167.
Siemens, D. H., B. E. Ralston, and C. D. Johnson. 1994. Alternative seed defense mechanisms in a palo verde (Fabaceae) hybrid zone: effects on bruchid beetle abundance. Ecological Entomology 19:381-390.
Sneath, P. H. A., and R. R. Sokal. 1973. Numerical taxonomy. W. H. Freeman, San Francisco, California, USA.
Stephens, S. G. 1949. The cytogenetics of speciation in Gossypium. I. Selective elimination of the donor parent genotype in interspecific backcrosses. Genetics 34:627-637.
Wall, J. R. 1968. Leucine aminopeptidase polymorphism in Phaseolus and differential elimination of the donor parent genotype in interspecific backcrosses. Biochemical Genetics 2:109-118.
Wan, J., Y. Yamaguchi, H. Kato, and H. Ikehashi. 1996. Two new loci for hybrid sterility in cultivated rice (Oryza sativa L.). Theoretical and Applied Genetics 92:183-190.
Wang, G.-L., J.-M. Dong, and A. H. Paterson. 1995. The distribution of Gossypium hirsutum chromatin in G. barbadense germplasm: molecular analysis of introgressive hybridization. Theoretical and Applied Genetics 91:1153-1161.
Weir, B. S. 1996. Genetic data analysis. Sinauer Associates, Sunderland, Massachusetts, USA.
Wheeler, N. C., and R. P. Guries. 1987. A quantitative measure of introgression between lodgepole and jack pines. Canadian Journal of Botany 65:1876-1885.
Whitham, T. G., P. A. Morrow, and B. M. Potts. 1994. Plant hybrid zones as centers for biodiversity: the herbivore community of two endemic Tasmanian eucalypts. Oecologia 97:481-490.
Williams, C. E., S. M. Wielgus, G. T. Harberlach, C. Guenther, H. Kim-Lee, and J.P. Helgeson. 1993. RFLP analysis of chromosomal segregation in progeny from an interspecific hexaploid hybrid between Solanum brevidens and Solanum tuberosum. Genetics 135:1167-1173.
Young, N. D., and S. D. Tanksley. 1989. Restriction fragment length polymorphism maps and the concept of graphical genotypes. Genetics 77:95-101.
Zamir, C., and Y. Tadmor. 1986. Unequal segregation of nuclear genes in plants. Botanical Gazette 147:355-358.
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||Hybridization and Resistance to Parasites|
|Author:||Rieseberg, Loren H.; Linder, C. Randall|
|Date:||Mar 1, 1999|
|Previous Article:||Scaling-Up: From Cell to Landscape.|
|Next Article:||Natural hybridization: how low can you go and still be important?|