Annu. Rev. Ecol. Syst. 1998. 29:467-501 Copyright (C) 1998 by Annual Reviews. All rights reserved PATHWAYS, MECHANISMS, AND RATES OF POLYPLOID FORMATION IN FLOWERING PLANTS Justin Ramsey and Douglas W. Schemske Department of Botany, University of Washington, Seattle, Washington 98195-5325; e-mail: jramsey@u.washington.edu; schem@u.washington.edu key words: polyploidy, allopolyploidy, allopolyploidy, hybridization, speciation Abstract Polyploidy is widely acknowledged as a major mechanism of adaptation and speciation in plants. The stages in polyploid evolution include frequent fertility bottlenecks and infrequent events such as gametic nonreduction and interspecific hybridization, yet little is known about how these and other factors influence overall rates of polyploid formation. Here we review the literature regarding polyploid origins, and quantify parameter values for each of the steps involved in the principal pathways. In contrast to the common claim that triploids are sterile, our results indicate that the triploid bridge pathway can contribute significantly to autopoly-ploid formation regardless of the mating system, and to allopolyploid formation in outcrossing taxa. We estimate that the total rate of autotetraploid formation is of the same order as the genie mutation rate (10-5), and that a high frequency of interspecific hybridization (0.2% for selfing taxa, 2.7% for outcrossing taxa) is required for the rate of tetraploid formation via allopolyploidy to equal that by autopolyploidy. We conclude that the rate of autopolyploid formation may often be higher than the rate of allopolyploid formation. Further progress toward understanding polyploid origins requires studies in natural populations that quantify: (a) the frequency of unreduced gametes, (b) the effectiveness of triploid bridge pathways, and (c) the rates of interspecific hybridization. 0066-4162/98/1120-0467$08.00 467 468 RAMSEY & SCHEMSKE INTRODUCTION Polyploidy, defined as the possession of three or more complete sets of chromosomes, is an important feature of chromosome evolution in many eukaryote taxa. Yeasts, insects, amphibians, reptiles, and fishes are known to contain polyploid forms (100), and recent evidence of extensive gene duplication suggests that the mammalian genome has a polyploid origin (112). In plants, polyploidy represents a major mechanism of adaptation and speciation (24,56,95, 104,120,157,159). It is estimated that between 47% and 70% of angiosperm species are polyploid (56,110). Differences in ploidy have been observed among related congeners and even within populations of taxonomic species (24,34,56,100,156), and there is evidence that individual polyploid taxa may have multiple origins (154). These observations suggest that polyploid evolution is an ongoing process and not a rare, macroevolutionary event. Research in agricultural and natural systems indicates that polyploids often possess novel physiological and life-history characteristics not present in the progenitor cyto-type (95,104). Some of these new attributes may be adaptive, allowing a plant to enter a new ecological niche. Because plants of different ploidies are often reproductively isolated by strong post-zygotic barriers, polyploidy is also one of the major mechanisms by which plants evolve reproductive isolation (34,56). In spite of the prevalence and importance of polyploidy, the factors contributing to polyploid evolution are not well understood (165). Two critical stages of polyploid evolution can be identified: formation and establishment. To understand the process of polyploid formation requires information on the pathways, cytological mechanisms, and rates of polyploid formation. To assess the likelihood that a new polyploid will successfully establish requires information on the viability and fertility of new cytotypes, the extent of assortative mating and reproductive isolation within and between different cytotypes, and the ecological niche of new polyploids. Here we review the literature concerning polyploid formation to answer the following questions: (a) What are the primary pathways and mechanisms of polyploid formation? (b) What are the parameters for each of the steps involved in polyploid formation? (c) What are the numerical values reported for these parameters? and (d) What is the estimated rate of polyploid formation by each pathway? One major motivation for this review is to synthesize the diverse literature on polyploid origins and thereby provide a resource for the development of future empirical and theoretical studies of polyploid evolution. To this end, we have tabulated data from many studies and made this information available on the Annual Reviews web site (http://www.annualreviews.org; see Supplementary Materials). We summarize these data throughout the text and identify the location of each database on the web site. POLYPLOID FORMATION 469 By necessity, many of the plants considered in this review are agricultural or horticultural cultivars and their wild relatives, as well as taxa widely used in classical genetic studies (e.g. Oenothera and Datura). We believe that the studies reviewed here provide insights into the process of polyploid formation in natural populations, but caution that further research in natural populations is needed to test our findings. Our survey draws from a wide range of plant taxa, but because of the limited number of studies, we do not interpret our results in a phylogenetic context. In this chapter, In refers to the somatic chromosome number and n to the gametic chromosome number regardless of the degree of polyploidy, while x is the most probable base number. This gives the following cytological designations: diploids (2m = 2x), triploids (2m = 3x), tetraploids (2m = Ax), etc. In describing crosses within and between cytotypes, the maternal parent is always listed first. MECHANISMS OF POLYPLOID FORMATION Several cytological mechanisms are known to induce polyploidy in plants. Somatic doubling in meristem tissue of juvenile or adult sporophytes has been observed to produce mixoploid chimeras (2,66,82,128,153). For example, Primula kewensis, one of the first described allopolyploids, originated from fertile tetraploid shoots on otherwise sterile diploid FjS of P. floribunda x P. verticellata (127). Similarly, a tetraploid shoot was observed on a diploid F1 hybrid between Mimulus nelsoni and M. lewisii (66), and in wounded ("decapitated") tomato plants (82). Somatic polyploidy is known to be common in many non-meristematic plant tissues (30, 31). For example, normal diploid Viciafaba contains tetraploid and octoploid cells in the cortex and pith of the stem (26). Such polyploid cells occasionally initiate new growth, especially in wounds or tumors, and are a potentially important source of new polyploid shoots (30,31,99). The frequency of endopolyploidy, and the relative likelihood of polyploid formation from different endopolyploid tissues, are not well known. Somatic doubling can also occur in a zygote or young embryo, generating completely polyploid sporophytes. This phenomenon is best described from heat shock experiments in which young embryos are briefly exposed to high temperatures (43,140). Corn plants exposed to 40°C temperatures approximately 24 h after pollination produced 1.8% tetraploid and 0.8% octoploid seedlings (140). Polyploid seedlings are also known to arise from polyem-bryonic ("twin") seeds at a high frequency (122,176), but it is now believed that such polyploids are generally of meiotic rather than somatic origin (29). In general, little is known about the natural frequency of somatic doubling in plants nor of the effects of interspecific hybridization on its occurrence. 470 RAMSEY & SCHEMSKE A second major route of polyploid formation involves gametic "nonreduc-tion," or "meiotic nuclear restitution," during micro- and megasporogenesis. This process generates unreduced gametes, also referred to as "2n gametes," which contain the full somatic chromosome number (see reviews in 19, 63). The union of reduced and unreduced gametes, or of two In gametes, can generate polyploid embryos. As will be described in detail below, In gametes have been identified in many plant taxa. Polyspermy, the fertilization of an egg by more than one sperm nucleus, is known in many plant species (172), and has been observed to induce polyploidy in some orchids (59). However, it is generally regarded as an uncommon mechanism of polyploid formation (56). Distinguishing between somatic doubling and In gametes as mechanisms of polyploid formation requires a system of genetic markers and a detailed knowledge of the cytological mechanism of gametic nonreduction, which are seldom available. There is, however, strong circumstantial evidence that In gametes are often involved in polyploid formation. The parents of spontaneous polyploids have, upon cytological analysis, commonly been found to produce In gametes (18,21,24,46,52,55,83,90,94,101,166,168, 173). Conversely, plants known to produce In gametes can be crossed to produce new polyploids (18,37, 81,136,138). In many cases, spontaneous polyploids have cytotypes that appear to have been formed by the union of reduced and unreduced gametes (21,67,68,84,125) rather than by somatic mutation, which generally only doubles the base chromosome number (e.g. Ax to 8x) (30,66,82,127). For example, Navashin (125) found triploids and pentaploids in the progeny of open-pollinated diploid Crepis capillaris, and these appear to have been produced by the union of reduced («) and unreduced (2m and An) gametes. Similarly, triploids generated by backcrossing diploid hybrid Digitalis ambigua x purpurea are thought to have arisen from unreduced gametes produced by this interspecific hybrid (21). Because nonreduction appears to be the major mechanism of polyploid formation (19,63,165), we focus on the role of In gametes in polyploid origins. It is clear, however, that much research remains to determine the relative roles of the various cytological mechanisms of polyploid formation in natural populations. Auto- and Allopolyploidy: An Evolving Terminology Kihara & Ono (86) first described two distinct types of polyploids: "autopoly-ploids," which arise within populations of individual species, and "allopolyploids," which are the product of interspecific hybridization. Because chromosome pairing behavior was believed to be a reliable indicator of chromosome homology, early workers emphasized the frequency of multivalent formation at synapsis as a criterion for distinguishing auto- and allopolyploidy (32,120). It was subsequently recognized that some polyploids of known hybrid origin POLYPLOID FORMATION 471 exhibit multivalent pairing, while bivalent formation is prevalent in some non-hybrid polyploids (24,156,157). The term "segmental allopolyploid" was thus coined to denote polyploids of hybrid origin that possess chromosome pairing characteristics of autopolyploids, while "amphiploid" was used to indicate all polyploids that combine the chromosome complements of distinct species (24, 157). The term "autopolyploid" was reserved for polyploids that arose within single populations or between ecotypes or races of a single species (24,56). Although this terminology is recognized by many students of polyploidy (56,165), several confusing aspects remain. For example, there is considerable variation in the criteria used to delimit related taxa as "species." Moreover, some authors reserve the term allopolyploidy for hybrid polyploid derivatives of species that are largely reproductively isolated by barriers of hybrid sterility, because such species are more likely to differ in chromosome structure and pairing and to generate polyploids that behave cytogenetically as "true" allo-ploids (24,160). Also, it is to be expected that some interpopulation polyploids may represent a class of polyploidy intermediate between auto- and allopolyploidy. Because of these and other difficulties, several alternate terminologies have been suggested. Jackson (71) proposed that the terms auto- and allopolyploidy be used in their original, cytological meaning (32,120)—that autopolyploids exhibit multivalent pairing while allopolyploids do not—and developed statistical criteria for distinguishing these types of polyploids (72). Lewis (99) used "intraspecific" and "interspecific" polyploidy to distinguish polyploids that are morphologically distinguishable from those that are not, and considered these terms to correspond roughly to "autopolyploidy" and "allopolyploidy," respectively. We believe that the primary criterion for classifying a polyploid is its mode of origin. We use the term "autopolyploid" to denote a polyploid arising from crosses within or between populations of a single species, and "allopolyploid" to indicate polyploids derived from hybrids between species, where species are defined according to their degree of pre- and/or post-zygotic isolation (biological species concept). We consider polyploids arising from hybridization between species with minor aneuploid differences (dysploidy) to be allopolyploids, following Clausen et al (24). Considerable differences in the mechanisms and rates of polyploid formation within "types" of autopolyploid and allopolyploid systems may exist. In particular, the frequency of meiotic irregularity and spontaneous polyploid production may differ between hybrids of recently and anciently diverged taxa. Pathways of Polyploid Formation Several different pathways of both auto- and allopolyploid formation have been described. In this section, we identify the major routes to polyploid formation 472 RAMSEY & SCHEMSKE and highlight examples in which one or more steps in the pathway have been directly observed. autotetraploid, triploid-bridge Triploids are formed within a diploid population, and backcrossing to diploids, or self-fertilization of the triploid, produces tetraploids. For example, 1% tetraploid progeny were obtained by backcrossing a spontaneous triploid clone of Populus tremula to a diploid (15). Similarly, a small number of tetraploid progeny were obtained from triploid apple varieties that had themselves originated as spontaneous polyploids (14). Although all the steps in this pathway have rarely been observed in their entirety, the individual mechanisms are well substantiated. Triploids have often been observed in diploid populations (41,47,69,143,168), and it is generally believed that these are produced by the union of reduced («) and unreduced (2n) gametes. Studies of such spontaneous triploids, as well as triploids produced by crossing diploids and tetraploids, indicate that many of the gametes produced by autotriploids are not functional, because they possess aneuploid, unbalanced chromosome numbers. However, triploids generate small numbers of euploid (x, 2x) gametes (12,40,44,87,91,148,149) and can also produce 3x gametes via nonreduction (12,91,117). Autotriploids can produce tetraploids by self-fertilization or backcrossing to diploids (40,44,74,174,179). autotetraploid, one-step Tetraploids are formed directly in a diploid population by the union of two unreduced (2«) gametes or by somatic doubling. For example, Einset (47) found a small fraction (4.39 x 10-4) of tetraploid seedlings while cytotyping the progenies of open-pollinated diploid apple varieties. Tyagi (168) crossed clones of Costus speciosus that were observed to produced some In pollen, and recovered a small number of tetraploid seedlings. This process has been observed in several other taxa (18,69,81, 85, 89, 122). allotetraploid, triploid-bridge Hybrid triploids are formed by diploids in the Fj or F2 generation of interspecific crosses, and self-fertilization or backcrossing to diploids produces allotetraploids. For example, Miintzing (119) crossed Galeopsis pubescens and Galeopsis speciosa to generate a highly sterile, diploid Fj hybrid. One of the 200 F2 progeny was found to be triploid, and a backcross of this plant to G. pubescens formed a single viable seed, which was tetraploid (119). Similarly, Skalihska (152) found a single triploid plant in the F1 progeny of a cross between Aquilegia chrysantha and Aquile-gia flavellata. Selling this triploid produced a small number of F2 progeny, two thirds of which were tetraploid. Allotriploids have commonly been observed in the F2 generation produced by backcrossing or selling interspecific Fx hybrids (21,24,94,101,118,137), and in the Fx generation by the union of reduced and unreduced gametes from the parent genotypes (27, 60, 67, POLYPLOID FORMATION 473 152). Studies of such spontaneous allotriploids, and of allotriploids obtained by crossing different diploid and tetraploid species, indicate that the production of diploid gametes (8,54,152,169) and nonreduction (45,70,90,162) enable allotriploids to produce allotetraploids by selling or backcrossing (67,80, 118,152,167). allotetraploid, one-step Allotetraploids are formed directly from diploids in the Fj or F2 generation of interspecific crosses. For example, 90% of the F2 progeny of Digitalis ambigua and Digitalis purpurea were tetraploid (21). Half of the F2 progeny of Allium cepa and Allium fistolium were tetraploid or hypo-tetraploid (i.e. Ax — 1) (94), and 2% of the F1 progeny of Manihot epruinosa x glaziovii were tetraploid (60). There are many other examples of tetraploids being produced in one step by F1 interspecific hybrids (1, 21, 24, 46, 55, 66, 83, 128, 137), or in the F1 generation of an interspecific cross (60,79,92,169). Other important pathways involve the evolution of ploidy levels above tetraploidy. higher ploidy, one-step Within a polyploid population, the union of reduced and unreduced gametes generates a new cytotype of higher ploidy. For example, 2% hexaploid cytotypes were recovered from the progeny of open-pollinated autotetraploid Beta vulgaris, apparently from the union of reduced (2x) and unreduced (4x) gametes (68). Similarly, 1 % of the progeny of tetraploid alfalfa were found to be hexaploid (16). There is circumstantial evidence of auto-hexaploid formation in tetraploid populations in several other systems (23,42). New odd-ploidy cytotypes could also be produced by this mechanism. For example, it has been suggested that unreduced gamete production in hexaploid Andropogon gerardii generated a 9x cytotype, which is now widely distributed (130). allopolyploidy, via hybridization of autopolyploids Hybridization between distinct autopolyploids directly produces allopolyploids. For example, crosses between autotetraploid Lycopersicon esculentum and autotetraploid Lycopersicon pimpinellifolium produced a fertile allotetraploid Lycopersicon that was identical to the allotetraploid made by doubling the diploid F1 hybrid (103). An allotetraploid Tradescantia was produced by crossing autotetraploid forms of T. canaliculata and T. subaspera (4). It has repeatedly been found that the post-zygotic barriers that isolate diploid taxa break down in autopolyploids, so that interspecific hybrids are formed easily (23,65). Not surprisingly, there may be extensive intergradation among polyploids, while diploid taxa remain morphologically distinct (65,98). allopolyploidy, via hybridization of different cytotypes Hybridization between different cytotypes (which may be of auto- or allopolyploid origin) 474 RAMSEY & SCHEMSKE generates intermediates of odd-ploidy, which subsequently produce new even-ploidy cytotypes. For example, a high frequency (~81%) of allohexaploids was generated by crossing triploid hybrids of Nicotiana paniculata (2x) and Nicotiana rustica {Ax) (90). A similar process has been observed in many other systems (24,25,90,166,173). Second-Generation Polyploids The production of later-generation polyploids can be achieved through a variety of pathways. For example, a new self-compatible tetraploid can self to produce tetraploid offspring. For outcrossing taxa, second-generation tetraploids can be produced by matings between independently produced tetraploids. Alternatively, backcrossing to the diploid progenitor can produce triploids (18,73,76), which contribute to further tetraploid formation by crossing to either diploids (44,80,152,170,179) or tetraploids (40,139,152,171,174). Later-generation tetraploids can also be produced by backcrosses of tetraploids to diploids that produce unreduced gametes (18,51,69,178). Clearly, the frequency and cytotype composition of later-generation polyploids will depend on factors such as the mating system and the degree of pre- and post-zygotic reproductive isolation between cytotypes (50,144). There is a need for further empirical and theoretical research on these and other issues related to polyploid establishment. UNREDUCED GAMETES AND THE ORIGIN OF NEW POLYPLOIDS Unreduced gametes are believed to be a major mechanism of polyploid formation (19,64). Both 2n pollen and 2m eggs have been observed in hybrid and non-hybrid agricultural cultivars and natural plant species (11, 13,24,36,37,83,101, 116,136,142,147,151,163,164,177). Unreduced pollen grains can often be identified by size, as they typically have a diameter 30^-0% larger than that of reduced pollen (78, 168, but see 101, 105), and the distribution of pollen size in plants known to produce 2m pollen is often bimodal (131,168). Unreduced female gametophytes can sometimes be identified by size (161), but more often the frequency of 2m gametes is indirectly estimated using controlled, inter-ploidy crosses in plants with very strong interploidy crossing barriers. The progeny generated are usually the products of 2m gametes (3, 36; see below). Little correlation has been observed between the production of 2m pollen and 2m eggs (36, 134, 142, 164, but see 147). Meiotic aberrations related to spindle formation, spindle function, and cytokinesis have been implicated as the cause of 2m gamete production in nonhybrid crop cultivars (19). For example, a parallel spindle orientation at anaphase II POLYPLOID FORMATION 475 results in the reconstitution of diploid nuclei in microsporogenesis, and premature cytokinesis that immediately follows the first meiotic division creates diploid nuclei that never undergo a second meiotic division (116). The cyto-logical causes of nonreduction in hybrids are less well studied. Often, poor chromosome pairing in F1 hybrids leads to asynapsis at the first meiotic division, and a single "restitution" nucleus containing the full somatic chromosome number forms in the spore mother cell (55, 83,166,173). Other cytolog-ical mechanisms related to cytokinesis are also known to produce 2« gametes in interspecific hybrids (64). The timing and type of cytological anomaly producing In gametes affects both the level of genie heterozygosity and the yield of polyploid cultivars (114). Unfortunately, there are few data on the cytological origins of 2« gametes in natural systems. Identifying the origins of unreduced gametes is complicated, because different individuals in the same species often produce In gametes by different cytological mechanisms, and more than one mechanism may operate within an individual plant (134, 177). What Is the Frequency of Unreduced Gametes ? The frequency of 2n gametes determines the rate of new polyploid formation, as well as the types of polyploids being produced, and is therefore critical for understanding polyploid formation. We summarized the observed frequencies of 2n pollen in hybrid and nonhybrid systems, excluding those selected for their tendency to produce In gametes. The mean frequency of In gametes found in studies of hybrids (27.52%) was nearly 50-fold greater than that in nonhybrids (0.56%), and this difference was significant (Mann-Whitney U test, P < 0.001) (web Table 1; 8 web Tables are located at www.AnnualReviews.org). This result is consistent with the qualitative impressions of Harlan & de Wet (63). Because interspecific hybrids often experience severe meiotic irregularities involving poor chromosome pairing and non-disjunction, the "reduced" gametes produced by hybrids often possess unbalanced, aneuploid cytotypes, and are thus inviable (22,52). This suggests that the effective frequency of In pollen in hybrid systems may be even higher than estimated here. The few existing data for In eggs suggest that the natural frequency of non-reduction is similar in megasporogenesis and microsporogenesis. The mean frequency of In eggs in a sample of approximately 100 field-collected individuals of Dactylis glomerata was 0.49% (36), while the frequency of In pollen in a similar collection of individuals was found to be 0.98% (105). The mean frequency of In eggs was 0.06% in Trifolium pratense (136) and 0.09% in maize (3). We know of no published reports on the frequency of In eggs in interspecific hybrids. 476 RAMSEY & SCHEMSKE Polyploid Formation Is Facilitated by a Breakdown in Self-Incompatibility The "effective" frequency of In pollen in plants with gametophytic self-incom-patability may be increased by the breakdown of incompatibility in diploid pollen produced by either reduction division in established tetraploids, or non-reduction in diploids. This tendency, which appears to be related to genie interactions in diploid pollen grains (20,96,97), may allow self-pollination by 2m pollen, leading to polyploid formation. For example, Marks (108) obtained some polyploids by selling diploid self-incompatible Solanum. Lewis (96) found only triploids in the selfed progeny of self-incompatible strains of Pyrus producing 2« pollen, and described this phenomenon as the "incompatibility sieve" for polyploid formation. This mechanism could contribute to polyploid formation where abiotic or biotic factors increase the frequency of self pollen deposition. Genetic Factors Influence the Rate of Unreduced Gamete Production Bretagnolle & Thompson (19) provide an exhaustive review of the genetic basis of 2n gamete production in nonhybrid crop species, and we provide only a brief summary here. Plant populations often possess heritable genetic variation for the capacity to produce 2m gametes, as illustrated by a rapid response to selection for 2m gamete production in crop cultivars (135,164). For example, the mean frequency of 2m pollen increased from 0.04% to 47% in three generations of selection on Trifolium pratense, giving a realized heritability of 0.50 (135). In Medicago sativa, selection experiments on 2m pollen and 2m egg production gave realized heritabilities of 0.39 and 0.60, respectively (164). Meiotic analysis of progeny derived from crosses between plants differing in their level of 2m gamete production indicate that this phenotype can be under strong genetic control and is often determined by a single locus (115,142,147). Why Is There Genetic Variation for Unreduced Gamete Production ? Because different cytotypes are typically reproductively isolated, 2m gametes do not contribute to the gene pool of their progenitor cytotype. Thus, we expect strong selection against 2m gamete production, and it is perhaps surprising to sometimes find high heritabilities for this trait. As yet, there is insufficient information to determine if the frequency of genes influencing 2m gamete production is different from that expected by mutation-selection balance. Polyploidy often occurs in perennial taxa capable of vegetative reproduction (58,155). POLYPLOID FORMATION 477 Characters related to sexual reproduction may be under relaxed selection in these systems, resulting in a potentially higher frequency of 2« and nonfunctional gametes. This hypothesis is supported by the observation that many of the taxa in which In gamete production has been documented are perennials with means of vegetative propagation (9,105,131). Another possible mechanism contributing to In gamete production is that the cytological abnormalities leading to non-reduction are the pleiotropic effect of genes with other, perhaps beneficial, effects. Environmental Factors Can Affect the Frequency of Unreduced Gametes Several researchers have found that 2n pollen production is stimulated by environmental factors such as temperature, herbivory, wounding, and water and nutrient stress. Temperature, and especially variation in temperature, have particularly large effects (11, 38,96,111,151). Belling (11) observed a dramatic increase in In pollen production in field and greenhouse cultures of Strizolo-bium sp., Datura stramonium, and Uvularia grandiflora following aberrant cold spells. Potato genotypes selected for the tendency to undergo gametic non-reduction had approximately twice the mean frequency of In pollen in a coastal field as in a greenhouse, an effect attributed to the temperature differences of the two environments (111). Ramsey & Schemske (unpublished data) found that the frequency of 2« pollen in randomly selected Achillea millefolium plants reared in a temperature-cycling growth chamber was approximately six times that in the natural population from which the study plants had been sampled. Plant nutrition, herbivory, and disease may also affect In gamete production. Grant (55) found that the rate of polyploid production per flower in F1 Gilia hybrids grown in low-nutrient conditions was almost 900 times greater than that of plants grown in high-nutrient conditions, a result attributed to poor pairing at meiosis in the former treatment. However, the higher level of polyploid production per flower in the low-nutrient treatment was partially offset by a much lower flower number, such that the number of polyploids produced per plant was only seven-fold greater. Kostoff (88) and Kostoff & Kendall (89) described an effect of gall mites and tobacco mosaic virus on In pollen formation. Many of the environmental factors known to influence In gamete production are experienced by plants in their natural habitats. This suggests that natural environmental variation, as well as large-scale climate change, could substantially alter the dynamics of polyploid evolution. The high incidence of polyploidy at high latitudes, high altitudes, and recently glaciated areas may be related to the tendency of harsh environmental conditions to induce In gametes and polyploid formation (23,151). 478 RAMSEY & SCHEMSKE TRIPLOIDS: FORMATION, MEIOSIS, FERTILITY, AND PROGENY The evolution of tetraploidy may proceed directly from diploids via the union of two 2m gametes, or in two steps via a triploid bridge (19). Because the probability of the union of two 2m gametes is expected to be very low, it has been hypothesized that triploids usually play a role in the evolution of tetraploids (39,63). However, the low fertility of triploids, coupled with the existence of cytological barriers that may prevent or limit triploid formation by diploids, may restrict the role of triploids (19,150). We review components of the triploid bridge pathways, including the likelihood of triploid formation via 2m gametes, triploid fertility and meiotic behavior, and the cytotype composition of the progeny produced by triploid parents. How Effective Are Unreduced Gametes in Triploid Formation? In most flowering plants, fertilization of the egg by a sperm nucleus is accompanied by fusion of the other sperm nucleus with two haploid polar nuclei in the female gametophyte to form the triploid endosperm that functions to nourish the 2x embryo. In polyploids, the process proceeds in an analogous fashion, but the ploidies of all tissues are proportionately increased. Crosses between diploid and tetraploid plants often fail because intercytotype hybrid seed development does not proceed normally, and nonviable seeds are produced. The difficulty of obtaining viable triploid seeds by diploid-tetraploid and tetraploid-diploid crosses has been termed the "triploid block" (108). Barriers to intercytotype hybridization have been observed at higher ploidy levels, but are not well described. Viable seeds produced by crosses between diploids and tetraploids are often tetraploid and result from unreduced gametes produced by the diploid parent. Abnormalities in the growth and structure of the endosperm have often been implicated as the source of triploid block (28, 49, 113, 178; see reviews in 62, 175). The ratios between the embryo, endosperm and/or maternal tissue, as well as the maternahpaternal ploidy ratio in the endosperm, are all altered in 2x x Ax and Ax x 2x crosses, and it has been suggested that normal seed development depends on these ploidy ratios. Miintzing (123) hypothesized that proper function requires an embryo:endosperm:maternal tissue ratio of 2:3:2, while others have suggested that it is the 2:3 ratio of the embryo to endosperm that is critical for proper seed development (175). An alternate explanation is that the maternahpaternal ploidy ratio in the endosperm, irrespective of the ploidy of the embryo or maternal tissue, determines seed viability (129). This has been described as the imprinting hypothesis (61,102), and supporting evidence POLYPLOID FORMATION 479 is provided by several studies showing that a 2:1 ratio of the maternahpaternal genomes in the endosperm is required for normal seed development (77,102). Johnston et al (77) proposed a modification of this hypothesis to account for the anomalous findings that viable seed is produced in some systems where the 2:1 ratio is violated, but not in others where the 2:1 ratio is met. Their endosperm balance number hypothesis suggests that seed development is affected by the effective maternal :paternal ploidy ratio of the endosperm, which may not always reflect the actual ploidy composition (77). Although the genetic mechanisms responsible for interploidy crossing barriers remain an open area of investigation, there is general agreement that the ploidies of an embyro and/or its associated endosperm are the critical factors influencing successful seed development. In diploids, fertilization of a reduced egg by an unreduced sperm nucleus will generate the same embryo:endosperm ploidy ratio (3:4) as a 1x x Ax cross, and the union of an unreduced egg and a reduced sperm nucleus will form the ploidy ratio (3:5) of a Ax x 1x cross [most described mechanisms of unreduced egg formation involve nonreduction in the megaspore mother cell, thus producing gametophytes with unreduced polar and egg nuclei (142,147, 177, but see 29)]. Similarly, seed from a 1x x Ax cross contains endosperm with the same maternal :paternal ratio (2:2) of seeds produced by the union of reduced eggs and unreduced pollen, while seed from a Ax x 1x cross contains endosperm with the same maternal :paternal ratio (4:1) of seeds produced by the union of unreduced eggs and reduced pollen. Note that crosses in both directions violate the normal (2x x 1x) embryo:endosperm ratio of 2:3 and the endosperm maternahpaternal ratio of 2:1. Irrespective of the cytological cause of triploid block, the success of crosses involving diploid reduced gametes from tetraploids and haploid reduced gametes from diploids should parallel that of crosses involving diploid unreduced gametes from diploids and reduced gametes from diploids. Here we use this approach to evaluate the likelihood of triploid formation via In gametes. We surveyed the literature for data on crosses between autotetraploids and their progenitor diploids. We excluded studies of naturally occurring polyploids because genie differences arising after autopolyploid formation could contribute to crossing barriers. Our review focuses on studies of autopolyploids because there are too few data on crossing success in allopolyploids. No viable triploid seed production was observed in 13 of the 19 studies of 1x x Ax crosses, or in 7 of the 17 studies of Ax x 1x crosses (web Table 2 located at www.AnnualReviews.org). These data indicate that complete triploid block is present in many taxa, and that the possibility of triploid formation may differ for In pollen and eggs. In the 11 studies reporting viable triploid production in 1x x Ax and/or Ax x 1x crosses, 10 found that more viable triploid seeds were generated by Ax x 1x crosses, while in only a single study 480 RAMSEY & SCHEMSKE was higher triploid production from 2x x Ax crosses observed. This difference is statistically significant (Paired sign test, P < 0.05). We calculated an index of triploid block via 2n gametes using studies in which data on intracytotype crosses were available. For diploid pollen, this is 1 _ £(2x4).3 1 £(2x2) where £(2x4).3 is the viable triploid seed production from 2x x Ax crosses, and £(2x2) is the viable seed production from 2x x 2x crosses (assumed to include only diploid seed). The corresponding likelihood of triploid formation from 2n eggs is 1 _ £(4x2).3 2 £(2x2) where £(4X2).3 is the viable triploid seed production from Ax x 2x crosses, and £(2x2) is defined as above. For both calculations, values greater than 0 indicate a triploid block, and a value of 1.0 indicates a complete block. In the nine studies involving 2x x Ax crosses, the mean block via diploid pollen was 0.952 (range 0.57 to 1.0), and of the eight studies of Ax x 2x crosses, the mean block via diploid eggs was 0.801 (range 0.34 to 1.0), a difference that is marginally significant (Wilcoxon Sign Rank test, P — 0.07). Together, these data demonstrate there is generally a large barrier to triploid formation via unreduced gametes, but that the barrier is often not complete. A direct estimate of triploid block would compare observed and expected triploid production in crosses between diploids producing 2n gametes. In Dactylis glomerata, this approach gave an overall block of 0.98 (18), which is similar to the mean value we report here. The observed reciprocal differences in intercytotype crossing success suggest that unreduced eggs are more effective in polyploid formation than are unreduced pollen. This is consistent with the finding that spontaneous triploids in well-studied genetic systems arise via nonreduction in female parents (17,27). Reciprocal differences in interploidy crossing success may be a consequence of the embryo:endosperm and endosperm maternahpaternal ploidy ratios, which differ with the direction of the cross (62,175). By cytotyping the embryos and endosperm of Citrus seeds resulting from intracytotype and intercytotype crosses of parents producing both reduced and unreduced gametes, Esen & Soost (49) demonstrated that seeds with embryo:endosperm ratios of 2:3, 3:5, 4:6, and 6:10 were viable, while those with 3:4, the expected ratio resulting from the union of unreduced pollen with reduced eggs, were not. Using a meiotic mutant that generated endosperm of varying ploidy, Lin (102) showed an effect of endosperm genome composition on seed viability in 2x x 2x and 2x x Ax POLYPLOID FORMATION 481 crosses in maize. Another possible cause of reciprocal differences is that some mutations in megasporogenesis create embryo sacs containing nuclei of varying ploidy, and thus triploid embryos produced by these megagametophytes can be accompanied by normal, functional triploid endosperm (29). These mutations are a cause of polyembryony (29) and may be an important route of viable triploid formation. Regardless of its cause, the possibility that unreduced eggs are of primary importance in triploid formation seems significant in light of the fact that nearly all studies of gametic nonreduction have focused on unreduced pollen (24,46,101,168,173). Several taxa with low triploid block have atypical endosperm characteristics (62,178). In Oenothera, only a single polar nucleus is involved in the formation of a diploid endosperm, and viable triploid seeds are easily produced by 2x x Ax crosses (62). In Populus, mature seeds have no endosperm, and seed development is very rapid; viable seed set in 1x x Ax crosses is 80% of the 1x x 1x yield (75). These observations suggest a possible relationship between endosperm characteristics, the strength of triploid block, and polyploid formation. Plant families such as the Asteraceae, Crassulaceae, Onagraceae, Rosaceae, and Salicaceae that lack endosperm in mature seeds have a high incidence of polyploidy. Triploids Generate Some Euploid Gametes, and Are Often Semi-Fertile The reduction division in triploids is expected to generate aneuploid gametes with half the triploid chromosome number, or 3x/2. However, the possession of an unmatched complement of chromosomes leads to the formation of multivalents and univalents during pairing, and subsequent irregularities during disjunction can create varied chromosome assortments (33). We surveyed the literature to examine the cytotype composition of pollen produced by triploids of both hybrid and nonhybrid origin, as indicated by examination of anaphase I, metaphase II, and anaphase II in pollen mother cells, or the first postmeiotic mitosis in maturing pollen. In the 26 studies examined, the most common modal pollen chromosome number was 3x/2 (17 studies), followed by 3x/2 — 1 (5 studies; web Table 3). The tendency to produce aneuploid pollen is similar in auto- and allotriploids, so in the remaining analyses we consider both triploid types together. Figure 1 shows the average frequency of haploid, diploid, triploid, and the most common aneuploid cyto-types in pollen produced by triploids (n — 25 studies; web Table 3 located at www.AnnualReviews.org). Low mean frequencies were found for haploid (3%) and diploid (2%) pollen relative to the frequency of the common aneuploid class 3x/2 (34%). Rare euploid gametes are formed when the separation of multivalents and unpaired chromosomes in a spore mother cell is so unequal 482 RAMSEY & SCHEMSKE T 1x (3x/2)-1 3x/2 (3x/2)+1 2x Pollen cytotype Figure 1 Frequency of euploid (Ix, 2x, 3x) and common aneuploid (3x/2, 3x/2 — 1, 3x/2 + 1) cytotypes in pollen produced by hybrid and nonhybrid triploids, as determined by investigation of pollen mitoses as well as metaphase and anaphase of pollen mother cells. Data from web Table 1. that haploid-diploid chromosome assortments are produced at the first meiotic division (12,44,149,152). Triploid pollen, the result of gametic nonreduction (12,45,70,91), was also observed at a low mean frequency (5.2%; n — 9 studies). Together, these analyses suggest that triploids produce mostly aneuploid classes of pollen. Unfortunately, there are relatively few studies of mega-sporogenesis to complement the data on microsporogenesis. Satina & Blakeslee (148) observed a modal egg chromosome number of 3x/2 — 2 in triploid Datura stramonium, with 7% haploid and 1% diploid complements. These values are similar to those observed for pollen. Triploids are often expected to be sterile because of their meiotic irregularities and high frequency of aneuploid gametes. However, in a survey of the literature, we found a mean pollen fertility of 31.9% (range 0-97; web Table 4 located at www.AnnualReviews.org). The mean fertility of autotriploids (39.2%, n — 23 studies) was greater than that of allotriploids (23.7%, n — 18 studies), but this POLYPLOID FORMATION 483 difference was not significant (Mann-Whitney U test, P = 0.17). For those studies with data on both pollen cytotype and fertility (web Table 4 located at www.AnnualReviews.org), there was a significant positive correlation between the frequency of euploid (x and 2x) pollen and pollen fertility (Spearman Rank Correlation, r — 0.63, P < 0.05, n — 11), suggesting that euploid pollen contribute disproportionately to overall pollen fertility (the frequency of 3x pollen was not included in this analysis because most studies quantifying gametic non-reduction did not provide data on the frequency of Ix and 2x pollen). Further evidence that triploids are often semifertile comes from the few studies that have examined the relative fertility of crosses involving triploids. The available data suggest that some viable progeny are typically obtained from 2x x 3x, 3x x 2x, and 3x x 3x crosses, and that crossing success may vary with the direction of the cross (51,126,152,170). New Polyploids Can Be Generated Through a Triploid Bridge The cytotypes of the progeny derived from triploid crosses are often different from what might be expected from triploid meiotic behavior. Figure 2 illustrates this phenomenon in allotriploid Aquilegia chrysantha x flavellata (3x — 21) and autotriploid Zea mays (3x — 30). In both cases, pollen chromosome numbers had an approximately normal distribution (Figure 2a, b), with modes corresponding to the "expected" aneuploid value of 3x j2. However, the cytotype distribution of the progeny differed significantly from the expected distributions calculated from the chromosome number distribution in microsporogenesis for self (Figure 2c, d) and backcross (not shown) progeny (Kolmogorov-Smirnov One-Sample test, P < 0.01). The selfed progeny in both studies had a bimodal cytotype distribution; in Aquilegia, most of the offspring were fully tetraploid (Figure 2c), while the modes in Zea were aneuploid (Figure 2d). These results suggest that gametes with cytotypes near the modal class of 3x/2 do not function as well as other gametes, especially compared to those with euploid or near-euploid cytotypes. We surveyed the literature to examine the frequency of polyploid cytotypes in the progeny of triploids. Auto- and allotriploids produced similar progeny cytotypes, so we combined them for the following analyses. Figure 3 illustrates the mean frequency of several euploid and aneuploid cytotypes resulting from 2x x 3x, 3x x 2x, 3x x 3x, 3x self, 3x x 4x, and 4x x 3x crosses (see web Table 5 for the complete data set). We first investigated the two-step pathway of tetraploid formation, which proposes that triploids produced in diploid populations generate tetraploids via backcrossing to diploids, triploid selling, or crossing among triploids. Tetraploid (4x) progeny were observed in four of 18 studies of 3x x 2x crosses, with a mean frequency of 9.8% (range 0-85.7), 484 RAMSEY & SCHEMSKE 30- 25- >» 20- o c a> 15- 3