c annual p.,rthar reviews runner Click here for quick links to Annual Reviews content online, including: •Otherarticles in this volume •Top cited articles •Top down loaded articles • Our comprehensive search The Role of Hybridization in Plant Speciation Pamela S. Soltis1 and Douglas E. Soltis2 The Genetics Institute, Florida Museum of Natural History, and 2 Department of Botany, University of Florida, Gainesville, Florida 32611; email: psoltis@flmnh.ufl.edu; dsoltis@botany.ufl.edu Annu. Rev. Plant Biol. 2009. 60:561-88 The Annual Review of Plant Biology is online at plant.annualreviews.org This article's doi: 10.1146/annurev.arplant. 043008.092039 Copyright © 2009 by Annual Reviews. All rights reserved 1543-5008/09/0602-0561$20.00 Key Words polyploidy, genome duplication, angiosperms Abstract The importance of hybridization in plant speciation and evolution has been debated for decades, with opposing views of hybridization as either a creative evolutionary force or evolutionary noise. Hybrid speciation may occur at either the homoploid (i.e., between two species of the same ploidy) or the polyploid level, each with its attendant genetic and evolutionary consequences. Whereas allopolyploidy (i.e., resulting from hybridization and genome doubling) has long been recognized as an important mode of plant speciation, the implications of genome duplication have typically not been taken into account in most fields of plant biology. Recent developments in genomics are revolutionizing our views of angiosperm genomes, demonstrating that perhaps all angiosperms have likely undergone at least one round of polyploidiza-tion and that hybridization has been an important force in generating angiosperm species diversity. Hybridization and polyploid formation continue to generate species diversity, with several new allopolyploids having originated just within the past century or so. The origins of polyploid species—whether via hybridization between species or between genetically differentiated populations of a single species—and the immediate genetic consequences of polyploid formation are therefore receiving enthusiastic attention. The time is therefore right for a review of the role of hybridization in plant speciation. S6i Contents INTRODUCTION.................. 562 Extreme Reticulation: Ancient and Recent Polyploidy.......... 562 SPECIES CONCEPTS.............. 565 HOMOPLOID HYBRID SPECIATION.................... 567 POLYPLOIDY...................... 569 Types of Polyploids................ 569 Genetic Expectations for an Allopolyploid Species........... 570 Polyploid Evolution: Rapid and Diverse Changes........... 571 Polyploidy and Diversification..... 572 Conditions That Favor Polyploidization Versus Hybidization................... 573 Autopolyploidy.................... 574 Polyploidy at the Population Level: Unanswered Questions......... 575 ISOLATING MECHANISMS....... 578 CONCLUSIONS.................... 580 appears to be higher than traditionally maintained and may be very similar to estimates for animals (134). Furthermore, many plant species appear to be reproductively isolated and therefore meet the same criterion for species recognition as do animals (see below and Reference 160). Nonetheless, despite some underappreciated similarities in speciation in plants and animals, there are also important differences. Most notable is the high frequency of hybridization and its role in speciation. Hybridization— typically considered to represent crossing between species—has been extended to include crossing between genetically divergent populations or races within a species (e.g., 17, 72), and we follow this broad definition here. Hybridization has often been viewed as a creative force in evolution (e.g., 13, 17). To understand plant speciation, the origin of many adaptations, and the maintenance of plant diversity, we therefore need a renewed emphasis on the processes of species formation through hybridization. Here we review the field of hybrid speciation and offer suggestions for fertile research. Clade: a monophyletic group; an ancestor and all of its descendants Hybridization: crossing between species or between genetically differentiated races or populations of the same species Homoploid hybrid speciation: the origin of a new species through hybridization of two species of the same ploidy (e.g., two diploid species) Allopolyploidy: polyploidy involving interspecific hybridization INTRODUCTION Approximately half a million species of green plants—the clade that encompasses green algae and land plants—have been recognized scientifically. Of these, more than half are angiosperms (198), although estimates of the number of an-giosperm species vary from 350,000 to 400,000 (e.g., P. Raven in Reference 80). The causes and nature of plant speciation are therefore important for understanding the origin and maintenance of a large proportion of the world's biodiversity (estimates of total species numbers range from 1 million to 10 million species). Plants exhibit diverse speciation mechanisms and modes of reproductive isolation (66, 97, 98, 159, 192, 193). Whereas earlier investigators (e.g., 66, 193) emphasized the dramatic differences between plants and animals with regard to population dynamics and speciation, recent work has illustrated many noteworthy similarities. For example, gene flow in plants Extreme Reticulation: Ancient and Recent Polyploidy A significant portion of speciation events in plants involves hybridization, in contrast to most other clades, in which speciation is divergent. Such hybridization results in a phyloge-netic net, rather than a classic bifurcating tree. Hybrid speciation can occur either at the same ploidal level (homoploid hybrid speciation) or much more commonly via allopolyploidy (speciation via hybridization and genome doubling) (Figure 1). [Even autopolyploidy—genome doubling within a species—may typically involve hybridization between populations of the same species (e.g., 181)]. Hybridization seems to be a ubiquitous feature of green plant evolution, although it is particularly pronounced in angiosperms and ferns (178, 212). Whereas both homoploid hybridization and allopolyploidy can be potential sources of new species, allopolyploidy appears to be much more common than homoploid hybrid 562 Soltis • Soltis o o x o o o o Homoploid hybrid species ox°° = oooo Allopolyploid o o X o o = oooo Allopolyploid □ Species A D Species B Figure 1 The origins of species via homoploid hybridization and polyploidy, as represented by a single pair of chromosomes in each diploid parental species, A and B. A homoploid hybrid species (here, a diploid) arises through hybridization of species A and B; the resultant diploid hybrid species has one chromosome complement (here, one chromosome) of each parental species. Allopolyploid formation is similar in that species A and B hybridize; however, unlike homoploid hybrid speciation, allopolyploid formation involves chromosome doubling. The resultant allopolyploid species combines the entire nuclear genomes of both parental species. Autopolyploid formation also involves chromosome doubling and may occur via crossing of genetically differentiated diploid individuals from the same or different populations of a single species. As illustrated here, crossing occurs between two individuals of the same species; chromosome doubling yields the autopolyploid. Note that in both allopolyploid and autopolyploid formation, chromosome doubling may occur either before or after the crossing event takes place. speciation for reasons discussed by Stebbins (193) and Grant (66) and summarized most recently by Rieseberg & Willis (159). Homoploid hybrid species may have greatly reduced fitness in early generation hybrids, whereas this may not be the case in early generation allopolyploids (despite a possible sterility bottleneck in polyploids; 97). Furthermore, genome doubling reduces or eliminates the possibility of the new polyploid backcrossing with its parents— such is not the case for a homoploid hybrid. For these reasons, formation of a new species via allopolyploidy is more likely than through homoploid hybridization. Genomic studies indicate that many, if not most, angiosperms are ultimately of ancient polyploid origin. Thus, the intertwined processes of hybridization and genome doubling have been significant in generating species diversity. These processes that have shaped much of angiosperm species diversity continue to the present, with allopolyploid species having arisen during the past century (see below). In the following paragraphs we review both the extent of ancient polyploidy in the angiosperms and several examples of allopolyploids that have arisen within the past 150 years. Ancient polyploidy. Although genomic studies are clear in identifying many examples of ancient polyploidy, distinguishing between ancient allopolyploidy on the one hand and ancient autopolyploidy followed by genomic diploidization on the other hand may be difficult. However, given that allopolyploidy is more common than autopolyploidy, extrapolating into the past suggests that most ancient polyploid events were allopolyploid. Ancient episodes of polyploidy have clearly played major roles in angiosperm evolution as well as in the evolution of ferns and lycopods (66, 194, 212). Inference of ploidy in angiosperms from chromosome numbers and hypothesized breaks between diploid and polyploid base numbers yielded estimates of 30-35% (193), to nearly 50% (40, 62, 66, 135), to as high as 70-80% (62,102). In contrast, Masterson (118) used leaf guard cell size in fossil and extant taxa to estimate that 70% of all angiosperms have experienced polyploidy in their evolutionary history. On the basis of genomic evidence, even these older values underestimate the frequency of ancient polyploidy in the angiosperms. Genomic studies have completely altered our view of the frequency of polyploidy in angiosperms. The question is no longer "What proportion of angiosperms are polyploid?" but Autopolyploidy: polyploidy that arises within a single species, although it may involve crossing between genetically differentiated populations vrww.annualreviews.org • Reticulation and Speciation in Plants Eudicots Fabidae Malvidae Asterids Monocots Magnoliids OJ P OJ 3 11 fr O Kj l/l o U CQ Figure 2 Whole-genome duplication (WGD) events in angiosperm evolution, inferred from complete genome sequences or other genome-level data. Each colored bar represents a separate WGD. Note that the elongated blue bar along the spine of the tree indicates that the position of this WGD is not clear from current data. However, a WGD unites either all eudicots or all core eudicots, or possibly monocots + eudicots. Likewise, the red bar near the base of the tree indicates that either all angiosperms share a WGD or all extant angiosperms except Amborella share a WGD. Additional data are needed for Amborella to resolve the placement of this WGD and determine whether or not the origin of the angiosperms coincided with a WGD. Orange bars indicate more WGD events that characterize smaller groups of species. Greek letters refer to duplications described by Bowers et al. (Reference 24). Modified with permission from Reference 176. "How many episodes of polyploidy characterize any given lineage?" (see References 32 and 176). Complete sequencing of the nuclear genome has revealed evidence of ancient polyploidy throughout angiosperms and in other eukary-otes. All plant nuclear genomes sequenced to date show evidence of ancient genome duplication (176, 199) (Figure 2): Arabidopsis (24, 173, 209), Oryza (143), Populus (206), Vitis (79, 208), and Carica (130). A growing body of evidence suggests that the common ancestor of Vitis, Populus, Arabidopsis, and Carica was an ancient hexaploid that arose after the split between monocots and eudicots. Following this paleopolyploidy event, subsequent genome duplications occurred within Brassicales (leading to Arabidopsis) and in the lineage leading to Populus (Salicaceae). In the absence of complete genome sequences, ESTs (expressed sequence tags) are an important source of genomic data that can be used to infer occurrences of ancient genome duplication (22, 39, 168). ESTs can be analyzed following Lynch & Conery (109) to estimate whether an ancient polyploid event may have occurred as well as its approximate age. Using this approach, ancient polyploidy has been identified in a number of crops, including Zea (maize), Glycine (soybean), and Gossypium (cotton), among others, with multiple whole-genome duplications in Fabaceae, Solanaceae, and Poaceae (22, 168). Likewise, ancient polyploidy is evident in several lineages of basal angiosperms: Nuphar (water lilies; Nymphaeaceae), Persea (avocado; Lauraceae), Liriodendron (tulip tree; Magnoliaceae), and Saruma (Aristolochiaceae) as well as the basal monocot Acorus (sweet flag; Acoraceae) and the basal eudicot Eschscholzia (California poppy; Papaveraceae) (39). Nuphar may in fact exhibit signatures of two ancient duplications, the older of which may be the oldest duplication so far discovered in flowering plants (39). Surprisingly, despite evidence for extensive polyploidy throughout the angiosperms, Amborella, the sister to all other extant angiosperms, does not show evidence of ancient polyploidy. Although Amborella has a high chromosome number (2w = 26), few duplicate gene pairs were detected in an initial survey of 10,000 ESTs (39); even with a greatly expanded data set, there is still no evidence for genome duplication in Amborella (176). However, lack of evidence for polyploidy in analyses of duplicate gene pairs does not rule out the possibility of ancient polyploidy, of which the signature may have aeen erase d. Recent polyploidy. Several plant species have originated via polyploidy within just the past 150 years: Spartina anglica (6, 7, 76, 116, 117) (Figure 3), Senecio cambrensis (Figure 4) and S. eboracensis (1, 163), Cardamine schultzii (207), and Tragopogon mirus and T. miscellus (139, 183) (Figure 5). These systems represent important evolutionary models because they formed 564 Saltis * Saltis very recently, their parentage is known with certainty, and in several, the polyploid formed multiple times, permitting one to ask whether evolution is repetitive. Studies on the genetic consequences of allopolyploidy in these recently formed polyploid species reveal some significant similarities as well as differences. In Spartina anglica, allopolyploidy resulted in few changes in genome structure, but there is evidence for changes in methylation or epigenetic reprogramming (6, 7, 164). In contrast, in recently formed allotetraploids in both Senecio (1, 73, 74) and Tragopogon (119,183,200; R. Buggs, A. Doust, J. Tate, J. Koh, K. Soltis, F. Feltus, A. Paterson, P. Soltis & D. Soltis, unpublished data), evidence exists for major genetic changes, including loss of homeologs and DNA sequence as well as changes in DNA expression, including tissue-specific expression. These recently formed natural polyploids are unique evolutionary models that afford the opportunity to compare the consequences of polyploidization with results for better-studied genetic models, such as Arabidopsis (e.g., 32), Gossypium (cotton; 2, 3), Oryza (rice; e.g., 223), and Brassica (e.g., 60), and with synthetic polyploids (e.g., 60, 84, 105, 189). SPECIES CONCEPTS A discussion of speciation requires that we step—however trepidatiously—into the realm of species concepts. The issue of when two entities should be considered distinct species has been a longstanding controversy, particularly for plants (reviewed in References 66, 97, 216, and 220). In fact, the uncertainty of what constitutes a plant species may well have slowed progress in the study of some aspects of plant speciation (159, 184). It is beyond the scope of this review to cover species concepts in detail; however, without a coherent definition of a species, it is impossible to determine when a new one has arisen. Comprehensive reviews of species concepts are given elsewhere (18, 19, 37, 41, 58, 59, 66, 97, 155, 203, 216). Rather than rely on the adage that "a species is whatever a good taxonomist says it is," we Spartina anglica Ancestral hexaploid [In = 60) S. maritima S. alterniflora S. foliosa H----------------1 ,_ r ^ F1 hybrids S.xtownsendii S.xneyrautii Introgression i Allopolyploid S. anglica England France California Figure 3 Examples of recent allopolyploids: Spartina anglica. Introduction of S. alterniflora outside its native range has resulted in hybridization with local species in England, France, and California. Hybridization has led to hybrid species formation (S. Xtownsendii and S. Xneyrautii, both from S. maritima and S. alterniflora), polyploid formation (S. anglica), and introgression (S. alterniflora X S. foliosa, denoted by curved arrow). Modified from Reference 7; photo contributed by M. Ainouche. review some of the more prominent concepts below and return eventually to the issue of hybridization and species. The morphology-based taxonomie species concept—an assemblage of morphologically similar individuals that differs from other such assemblages (66)—continues to be widely used in plants. Although practical for taxonomie purposes, this system is subjective—the amount of difference that "is worthy of species rank cannot be prescribed objectively" (66). Different taxonomists may have different criteria and emphasize different characters. Adherence to this concept has certainly been an impediment to the recognition of the importance of autopoly-ploid speciation, in which autopolyploids, even those formed via some degree of hybridization, very closely resemble their diploid progenitors (see below and Reference 184). The biological species concept (122), which maintains that a species is a "a group of interbreeding (or potentially interbreeding) populations that are reproductively isolated Homeologs: chromosomes (and the genes that reside on them) in an allopolyploid that are similar due to shared evolutionary descent; for example, chromosome pair 1 of parental species A is homeologous to chromosome pair 1 of parental species B in allopolyploid species AB www.annualreviews.org • Reticulation and Speciation in Plants 565 S. vulgaris Non-ray, SC S.Xbaxteri Sterile 5. cambrensis SC S. squalidus Ray, SI Homoploid speciation in Senecio S. vulgaris S. cambrensis S. squalidus S. squalidus [In = 20) S. aethnensis [In = 20) S. chrysanthemifolius [In = 20) Figure 4 (a) Examples of recent allopolyploids: Senecio cambrensis. (b) Homoploid hybrid speciation occurs in Senecio as well: S\ squalidus. Photos contributed by S.J. Hiscock. SC, self-compatible; SI, self-incompatible. from other such groups," remains the prevailing view of species in animals (e.g., 36, 37) and has long played a major role in views of plant species as well. However, the application of the biological species concept is difficult in plants because of frequent hybridization and asexual reproduction. In part because of frequent hybridization between plant species and coupled with theoretical arguments against the biological species concept, many plant systematists have abandoned the biological species concept (e.g., 20, 42, 48, 52, 82, 125, 132, 136). However, the biological species concept has recently had a modest resurgence in popularity (e.g., 160, 167), due, at least in part, to the interpretation that some hybridization between species should not dictate that only a single species be recognized (e.g., 37). Despite this renewed pragmatism and empirical work linking morphological similarity to reproductive cohesion (and the converse) (160), many systematists continue to oppose the biological species on theoretical grounds (i.e., the feature uniting members of a biological species—intercrossability—is generally a symplesiomorphy, a shared, ancestral trait). The evolutionary species concept (174, 218, 219) recognizes ancestral-descendant sequences of populations that evolve separately from other such lineages and have their own ecological niches, evolutionary tendencies, and historical fates. Limited hybridization can be accommodated by the evolutionary species concept as long as the hybridizing species do not merge (66, 174). Likewise, the origin of a new species via homoploid hybridization or allopolyploidy would yield a new evolutionary lineage with its own evolutionary tendencies and historical fate. Again, as long as the parental lineages remain intact, such hybrid speciation $66 Saltis * Saltis causes no problems for the evolutionary species concept. The widespread acceptance of phylogenetic approaches prompted the development of several phylogenetic species concepts. Following the phylogenetic species concept (sensu Reference 38; the diagnosability species concept, Reference 82), a species "is the smallest diagnosable cluster of individual organisms within which there is a parental pattern of ancestry and descent." Likewise, species are considered "the minimal elements of hierarchic descent systems" (41). Operationally, species are defined as "the smallest aggregation of (sexual) populations or (asexual) lineages diagnosable by a unique combination of character states" (42, 136, 217). Under a second phylogenetic species concept (48, 131, 133; the apo-morphic species concept, Reference 82), species are recognized on the basis of monophyly and are defined as "the least inclusive taxon recognized in a formal phylogenetic classification" (133; but see also References 48 and 131). Variants on these phylogenetic concepts have also been proposed (e.g., 19, 41, 45), but discussion of them is beyond the scope of this review. Application of either the diagnosable or apomorphic (sensu Reference 82) species concepts to hybrid-derived lineages is not straightforward. The diagnosable species concept may be difficult to apply in cases of hybrids and polyploids in which a (new) derivative species cannot be diagnosed as distinct from its progenitors. Recognition of a hybrid/polyploid species based on monophyly is likewise problematic; although the new hybrid/polyploid is a new evolutionary lineage, it may not appear as such in a phylogenetic tree, unless provisions for incorporating hybridization events are in place. Furthermore, recurrent formation of allopolyploid "species" raises questions about the strict monophyly of most allopolyploids: under the apomorphic species concept, should each polyploid lineage of independent origin be considered a separate species, particularly in the absence of evidence on crossing relationships with and among populations of separate origin? Tragopogon T. dubius a^^fe 2,7 = 12 ^ / \ *ilfc / \». ú T. miscellus 2,7 = 24 \f / - T. mirus S? 2n = 24 \ / T. pratensis % 2n=12 *WPr \ T. porrifolius 2/7=12 Figure 5 Examples of recent allopolyploids: Tragopogon mirus and T. miscellus. HOMOPLOID HYBRID SPECIATION The laws of inheritance allow for predictions about the genetic properties of a hybrid species at either the diploid or polyploid level. Here we review the expectations for a diploid hybrid under a range of scenarios for the mode of origin of the hybrid and then consider how empirical data support or deviate from these expectations. Similar discussion for polyploid species follows (see below). A typical expectation of hybridization is ad-ditivity, but what exactly does additivity mean? From the perspective of quantitative genetics underlying polygenic traits, this is the additive genetic variance, and in a hybrid between inbred lines, additivity will result in some sort of intermediate phenotype, barring dominance or epistasis (see Reference 53). From the joint perspective of morphology and classical systematics, additivity in a hybrid has been typically equated with morphological interme-diacy. However, morphological analyses of natural and artificial hybrids clearly indicate that a range of morphological outcomes is possible: from characters that are identical to those of one parent, to those that are truly intermediate, to those that are identical to the second parent, to novel traits, with intergradations among these conditions (124, 151). Likewise, both theory vrww.annualreviews.org • Reticulation and Speciation in Plants 567 Hybrid swarm: results when two (or more) species interbreed extensively; includes parental species, Fi, F2, and later-generation hybrids, and backcrosses to one or both parental species Introgressive hybridization/ introgression: the transfer of genetic material from one species to another through hybridization and backcrossing (63) and analyses of flavonoid chemistry (100, 101) support the hypotheses of both additivity and novelty in hybrids. If we turn to patterns exhibited by molecular markers, we do not, by definition, see interme-diacy, but the expectation of additivity remains. For example, consider the simple case of two parental individuals belonging to species A and B, respectively, and fixed for alternative alleles at all loci. In this case, the Fi hybrid AB would show complete additivity at all loci. But does that mean the hybrid species AB should be expected to exhibit strict additivity at all loci? The answer is: only if the hybrid species reproduced asexually, through either vegetative reproduction or some other method of asexual propagation. This form of hybrid species has been recognized by many systematists, for example in characterizing hybrids in the genus Rubus (Rosaceae; blackberries and raspberries; e.g., 8). If the Fi were fertile and led to an F2, the F2 would not show additive patterns, but instead the genes would follow the laws of segregation and independent assortment. These combined processes could produce an array of genotypes spanning from one parental genotype to the other (the Fi) and various novel genotypes. Under these simplistic conditions, molecular markers and minimal analyses could probably identify with confidence likely cases of hybrid speciation. However, let us consider a more real-life situation: Both parental individuals are heterozygous at some or all loci (although they do not share alleles), so a sample of the parental species compared with the hybrid may show some differences. These differences may or may not be extensive, but they could interfere with the recognition of a hybrid as a true hybrid and with the identification of a species' progenitors. A more complicated scenario is that the parents share alleles at some loci, which is not unlikely if they are closely related species (e.g., 51). The result may be that the hybrid is additive of the parental alleles at some loci, but the pattern may not be clear at other loci because of the parents' shared alleles; thus, the expectation of additivity is a little more difficult to evaluate under this situation. Perhaps the most real-life scenario involves the formation of the hybrid species from a hybrid swarm that was originated by heterozygous parents. With this mode of formation, Fis, F2S, later-generation progeny, and possibly even backcrosses may all have contributed to the formation of a stabilized hybrid derivative, and strict additivity is not to be expected. Under this scenario, a hybrid species should share approximately half of its alleles with each parent; thus, it is additive in a very loose sense, but certainly not in the way in which an Fi formed from two completely genetically differentiated parents is additive. If multiple origins of the species have occurred from genetically differentiated parental populations (typically associated with hybrid species at the polyploid level rather than the diploid level, but possible nonetheless for a diploid hybrid species), the additivity is even looser. Furthermore, depending on the age of the hybrid species, it may or may not have had time for the origin of novel alleles through point mutations (e.g., 63). Thus, several processes may contribute to patterns that are not strictly additive, even in a species that in fact was formed through hybridization. Gallez & Gottlieb (61) described the genetic expectations of a hybrid species from a popu-lational perspective. First, a hybrid species is expected to show additivity at a single locus, with alleles derived from the two parents. In addition, additivity is expected at the popula-tional level when alleles from the different parents are combined across loci (e.g., locus A has alleles from parent 1 and locus B has alleles from parent 2, with perhaps locus C having alleles from both parents). These predictions were formulated for use with allozyme markers, but they remain valid for both sequence-based and microsatellite-based alleles. Homoploid hybrid speciation appears to occur rarely, but is also difficult to detect, with perhaps only 20 good examples in the literature (reviewed in Reference 68). The best-studied examples of diploid hybrid species in plants are found in Helianthus and Iris. Examples from both genera have been studied for decades. The concept of introgressive hybridization j68 Soltis * Soltis (or introgression)—that is, interspecific hybridization followed by backcrossing to one or both parental species, with the ultimate transfer of genetic material between species—was first described on the basis of patterns of hybridization in Iris (161; see also References 11, 12, 14, 16, and 17 and references therein). The stabilized hybrid derivative I. nehonil was later recognized as a three-way hybrid among I. bre-vifolia, I. hexagona, and I. fulva. Patterns of genetic variation in I. nekonii are complex (e.g., 15), but the contributions of the three proposed parental species are all evident. Helianthus annum, the widespread cultivated sunflower, has hybridized with other species of Helianthus all across North America (e.g., 75, 154, 158). At least three species have formed through hybridization between H. annuus and H. petiolaris, which is also fairly widespread, at least in the American Southwest. H. anomalus, H. deserticola, and H. paradoxus, all diploid hybrid derivatives from the same parents, exhibit genetic contributions of the two parents, but not in the most explicit combinations. In addition, all these species have adapted to extreme, but different, habitats: H. deserticola occurs in the Great Basin in Nevada, Utah, and northern Arizona; H. anomalus is found on sand dunes in Utah and northern Arizona; and H. paradoxus is restricted to saline wetlands in western Texas and New Mexico (157). Most spectacularly, all have undergone parallel chromosomal evolution; all share the same prominent rearranged segments of the genome relative to their parents, and these rearrangements have also occurred in synthetic hybrids between H. annum and H. petiolaris. Both the causes of this parallel evolution and its consequences remain under study, but clearly demonstrate that hybridization and hybrid speciation may have far-reaching and unanticipated effects. POLYPLOIDY Types of Polyploids How many types of polyploids should be recognized and how to characterize have has long been debated (e.g., 34, 40, 66, 192, 193) and are reviewed elsewhere (e.g., 177, 201). In brief, two general types of polyploids have historically been recognized: those involving the multiplication of one chromosome set and those resulting from the merger of structurally different chromosome sets. Kihara & Ono (86) used the terms autopolyploidy (auto = same) and allopolyploidy (alio = different), respectively, to distinguish between these two types. This approach was later employed by other early workers (34, 40, 66, 135). However, despite long-term efforts to categorize polyploids as either auto- or allopolyploids, a continuum of polyploid types clearly exists in nature; some polyploids simply cannot be easily placed in either of these two groups (102, 193). Stebbins (192, 193) referred to polyploids that comprise slightly differentiated chromosome sets as segmental allopolyploids. Such polyploids likely resulted through chromosome doubling associated with hybridization between parental individuals that bear complements that differ, for example, in the arm length of one pair of chromosomes. Hence, researchers have debated for more than 70 years about the types of polyploids that should be recognized in nature and the proper definitions of autopolyploidy and allopolyploidy. True genomic allopolyploids are typically derived from hybridization between two (or more) distantly related species and thus combine divergent genomes with chromosome complements that are unable to pair with each other. In contrast, strict autopolyploids result from genome doubling within a single individual or by crossing between different plants or populations within a species, which involves the production and merger of unreduced (diploid) gametes from genetically and chromosomally similar individuals. [Because an autopolyploid may arise via crossing between genetically different individuals (179, 195), the concept of hybridization may extend to autopolyploids as well (17, 72), and we therefore include autopolyploidy in this review.] Thus, an autotetraploid will contain four copies of each chromosome (all four are homologs), whereas vrww.annualreviews.org • Reticulation and Speciation in Plants jóťj Fixed heterozygosity: nonsegregating heterozygosity, usually due to the additivity of divergent parental genomes in an allopolyploid; may also arise following gene duplication and divergence in a diploid an allotetraploid will contain two of each pair of the counterpart chromosomes derived from two different species (homeologous chromosomes). In (most) allopolyploids, inheritance patterns are therefore disomic because multiple copies of the same genetic loci are not present. In contrast, a strict autotetraploid will likely have tetrasomic inheritance. However, although tetrasomic inheritance is a useful tool for recognizing autopolyploids, it is not an absolute marker of autopolyploidy because a given polyploidy may have some loci with disomic inheritance and others with tetrasomic inheritance (184, 196; see below). Polyploidy has long been recognized as a major force in plant evolution (e.g., 34, 40, 107, 118, 135, 192, 193, 194). Following the work of Stebbins (191, 192, 193, 194), Clausen and coworkers (34), and Grant (e.g., 65, 66), polyploidy became a major focus of biosys-tematic research. Botanists have long appreciated that polyploid lineages may show complex relationships with each other and their diploid ancestors (reviewed in References 159 and 184). The past 10 to 15 years have witnessed a dramatic resurgence in the study of polyploidy (e.g., see Reference 91; also reviewed in References 4, 5, 185, 201, and 214), with renewed interest in the mechanisms of polyploid formation and establishment (77, 147, 148); the frequency of recurrent polyploidiza-tion (e.g., 195, 200); the ecological effects of plant polyploidy (e.g., 204, 205); and the genetic, epigenetic, chromosomal, and genomic consequences of polyploidization (e.g., 4, 5, 24, 49, 60, 90, 92, 104, 106, 138, 144, 149, 199). Recent research has resulted in major modifications to many of the traditional tenets of polyploid evolution (e.g., 3, 49, 138, 185). Recent research has also confirmed that polyploidy is not limited to plants; it has also played a major role in the evolution of other eukaryotes (67, 112, 113). Two episodes of polyploidy are hypothesized for the common ancestor of vertebrates (57, 114, 128, 137, 141, 190). Polyploidy has also been important in the subsequent evolution of amphibians (21), as well as salmonids and other fish (10, 44, 89, 127). The genomes of yeast and other Saccharomyces were anciently duplicated (50, 85, 222). Genetic Expectations for an Allopolyploid Species As for a diploid hybrid species, the fundamental genetic expectation for an allotetraploid (and other allopolyploids) is additivity of the parental genotypes. However, unlike a fertile diploid hybrid with segregating parental alleles, an allotetraploid will sequester its parental genetic variation into its component genomes. Thus, some genetic diversity in an allotetraploid will segregate and some will not. That is, genetic variation on homologous chromosomes (i.e., those that pair, contributed from the same parent species) will segregate, whereas genetic variation on homeologous chromosomes (i.e., similar chromosomes derived from the two parents) will not. Let us consider the simplest scenario: that of two homozygous parents, fixed for alternative alleles at all loci. If these two parents gave rise to an allotetraploid via the combined processes of hybridization and chromosome doubling (see References 147 and 148 for reviews of polyploid formation), the allotetraploid would be homozygous at all loci within a parental genome and heterozygous at all homeologous loci contributed by the two parents. In other words, the allotetraploid would exhibit complete additivity of the parental genes and would appear heterozygous at all homeologous loci. Note, however, that there is no segregating variation under this scenario. Because the parental individuals were homozygous at all loci, this homozygosity is maintained in the allotetraploid; even though the homologous chromosomes are segregating, they do not bear different alleles and thus there is no segregating variation. Furthermore, the fixed heterozygosity contributed by the parental individuals does not segregate because it is borne on homeologous, rather than homologous, chromosomes. This heterozygosity results from the divergence of the parental alleles at homeologous loci and appears similar to the 57° Soltis * Soltis heterozygosity of an Fi, except that it does not segregate. Depending on the mode of formation, true heterozygosity may have been introduced into the first allotetraploid individual, thus generating a more complex scenario than that described above. If we continue to assume that the parents have completely different alleles, rather than homozygosity at all loci within the contributed genome of each parent, as above, here segregating allelic variation occurs between ho-mologs for at least some loci. Furthermore, nonsegregating fixed heterozygosity will still be maintained at homeologous loci, combining the genotypes of the parents. Finally, the most complex scenario—and that most likely to reflect patterns of genetic variation in natural populations of allotetraploids—incorporates the role of multiple origins of polyploids in shaping the genetic diversity of polyploid individuals, populations, and species. It is now widely recognized that nearly all polyploid species comprise populations of independent formation from genetically distinct progenitor populations (see, e.g., Reference 201). If each of several constituent populations of an allotetraploid species of multiple origin had the genetic attributes described above (i.e., both segregating variation and fixed heterozygosity), then each allotetraploid population would have its own set of genotypes and all populations would be genetically distinct. However, with even limited gene flow among populations, novel genotypes could result from crossing between genetically different individuals, followed by segregation and independent assortment of genetic variants (182). The result would be a highly diverse allotetraploid species, with more genetic diversity than could possibly be incorporated into it via a single origin. Thus, in contrast to classical interpretations of polyploids as genetically depauperate and potential evolutionary dead ends, polyploids of multiple origin, especially those with interpopulational gene flow, are expected to maintain levels of genetic diversity—and therefore evolutionary potential—comparable to that of their diploid progenitors. Polyploid Evolution: Rapid and Diverse Changes Flowering plants exhibit remarkable genome plasticity (90), yet the merger of two genomes in a common nucleus should be viewed as a major upheaval at the nuclear and cellular levels— that is, as a sudden, violent disruption. However, angiosperm genomes tend to tolerate this merger, but with the effect of "genomic shock" (35). The genomic interactions that occur following interspecific hybridization are among the causes of genomic shock, which is essentially a response to severe stress (123). Newly formed polyploid genomes subsequently undergo movement of transposable elements and rapid changes in genome size, genome structure (e.g., insertions, deletions, translocations), and epigenetic control. Polyploidy may therefore result in a dramatic "restructuring of the tran-scriptome, metabolome and proteome" (90). On the basis of studies of synthetic and natural polyploids, these changes may begin to occur almost immediately postpolyploidization (e.g., 60, 73, 84, 87, 103, 119, 189, 213; see also References 4, 5, 49, 91, 183, 201, and 214). For example, structural changes in synthetic wheat and Brassica occur within the first few generations (60, 84, 189). In natural populations of Tragopogon, major structural changes, including translocations and trisomy/monosomy, are apparent in new synthetic polyploids of Tragopogon and in natural populations that cannot represent more than 30-40 generations since polyploidy formation (103) (Figure 6). The dynamism of polyploid genomes is also manifested by changes in gene silencing, DNA methyla-tion, and tissue-specific gene expression (e.g., 4, 5, 32, 138). Polyploidy clearly plays a major role in modifying global patterns of gene expression (2, 73, 138, 197) and may be a major source of developmental novelty in polyploid systems (49). All these processes are sources of novel genetic variation and as such can play a major role in the evolutionary success of polyploids (49,90). These rapid changes are the fuel of polyploid evolution. As a result of genetic and genomic changes, individuals may arise with a vrww.annualreviews.org • Reticulation and Speciation in Plants 571 2601-7 GISH of D-genome GISH of P-genome 2601-8 ?) H U M II M |f»HH»" It HM**" DAPI GISH of D-genome f É ft|| »A J* I« •• HLH9 H «« GISH of P-genome 1% If I 2602-3-10 ..____. 5Sand45SrDNA j GISH of D-genome J ^^ \ SSand4SSDNA /( )| )\ || N| H GISH of P-genome /I 2603-33 *.____. 5Sand45SrDNA GISH of D-genome ^1 || ' " 5Sand45SrDNA GISH of P-genome 2603-33B 5Sand45SrDNA GISH of D-genome 5Sand45SrDNA GISH of P-genome ft h h n «m« (r H«i>««" j) u u ••«■ *|) u n»«» K H n i"»» \í U H K II K modified phenotype and ecological preferences and hence are able to exploit new niches or to outcompete progenitor species. Polyploidy and Diversification Given the many suggested benefits of polyploidy (e.g., 96, 97) and the often proposed relationship of genome duplication to speciation (110, 111, 215), a major question remains: Are rates of diversification higher in polyploid lineages than in diploid groups (because of either increased rates of speciation, decreased rates of extinction, or both)? Identifying ancient polyploid events in angiosperm phylogeny provides the opportunity to assess the correspondence between inferred genome duplication events and large diversifications. That is, anecdotal data suggest that polyploid lineages are successful, but a statistical association of polyploidy and species richness has not been rigorously tested. To address this question, Soltis and coworkers (176) compared species richness in clades that are ancient polyploids with sister clades that are not. Using this approach, polyploidy certainly appears to have been a major driving force behind the diversification of the angiosperms. Figure 6 Rapid chromosomal change in Tragopogon mirus. Genomic/fluorescence in situ hybridization (GISH/FISH) and 4',6-diamidino-2-phenylindole (DAPI) karyotypes of mitotic chromosomes of naturally occurring plants of tetraploid T. mirus. Fluorochrome colors are as follows: Yellow/green is fluorescein isothiocyanate (FITC), digoxigenin-labeled probes; orange/red is Cy3, biotin-labeled probes; blue/purple is DAPI staining. Each karyotype is shown with DAPI staining, sometimes also simultaneously labeled for 45S rDNA {yellow) or 5S rDNA (red), and after GISH with T. dubius (green) and T. porrifolius (red) total genomic DNA probes. Plant 2601-7-translocation at end of chromosome C is shown (arrows). Also shown are plant 2601-8-trisomy for B in the D genome (contributed from T. dubius) and monosomy for C in the P genome (contributed from T. porrifolius), possible translocation at the end of chromosome C (see arrow), and plant 2602-3-10-monosomy for the D genome for chromosomes B and C. Soltis * Soltis 9 For example, comparisons of diversification rates suggest that genome doubling may have led to a dramatic increase in species richness in several angiosperm lineages (see Reference 176), including Poaceae (22, 115, 143, 168) (Figure 7), Solanaceae (168), Fabaceae (88, 146), Cleomaceae (170), and Brassicaceae (130). The importance of polyploid events in diversification may be even more profound at deeper levels in the angiosperm tree. Polyploidy may be associated with the origin of the eudicots and perhaps even the origin of the angiosperms (29, 43, 176). Conditions That Favor Polyploidization Versus Hybidization What conditions favor the formation of polyploids? What conditions favor the formation of homoploid hybrid species? Plant biologists have long maintained that the divergence between diploid parents impacts the likelihood that those species will successfully produce a polyploid derivative. That is, closely related diploids are less likely to form a polyploid than are more divergent congeneric diploid species. Digby (47) and Clausen & Goodspeed (33) demonstrated that successful allopolyploids could be derived more easily than homoploid hybrids from distantly related parents (reviewed in Reference 28). In fact, a high degree of differentiation between diploid parents was considered much more likely to lead to polyploidization than to stabilization of a homoploid hybrid (34, 40). More recently, Grant (66) considered chromosomal repatterning between diploids to be a particularly important feature that would promote allopolyploid formation. The distance of the relationship between diploid parents and the likelihood of those parents to form polyploid versus homoploid hybrid species has been reconsidered in light of molecular data. On the basis of sequence divergence in the internal transcribed spacers (ITS) of nuclear ribosomal DNA as a proxy for overall genetic (and presumably phylogenetic) divergence, Chapman & Burke (31) found that 658 genera/-10,000 species BEP clade PAC M ADD clade u n ^ ri b v 0) n o .c u O p 0) T3 'o v T3 u o T3 'n b n