Reproductive isolation between autotetraploids and their diploid progenitors in fireweed, Chamerion angustifolium (Onagraceae)


  • Brian C. Husband,

    Corresponding author
    1. Department of Botany, University of Guelph, Guelph, Ontario, N1G 2W1 Canada
      Author for correspondence: Brian C. Husband Tel: +519 824 4120 × 54790 Fax: +519 767 1991 Email:
    Search for more papers by this author
  • Holly A. Sabara

    1. Department of Botany, University of Guelph, Guelph, Ontario, N1G 2W1 Canada
    Search for more papers by this author

Author for correspondence: Brian C. Husband Tel: +519 824 4120 × 54790 Fax: +519 767 1991 Email:


Polyploidy is viewed as an important mechanism of sympatric speciation, but few studies have documented the reproductive barriers between polyploids and their diploid progenitors or explored the significance of assortative mating for polyploid establishment. Here we synthesize new and existing data on five prezygotic (geographic isolation, flowering asynchrony, pollinator fidelity, self-pollination, gametic selection) and two postzygotic (selection against triploid hybrids, inbreeding depression) reproductive barriers between diploid and autotetraploid individuals of the perennial plant Chamerion angustifolium. We also present estimates of realized rates of between-ploidy mating and examine the impact of assortative mating on polyploid dynamics using computer simulation. Reproductive isolation (measured from 0 to 1) was enforced by each barrier, including: geographic separation (RI = 0.41), flowering asynchrony (0.13), pollinator fidelity (0.85), self-pollination (0.44), gametic selection (0.44) and postzygotic isolation (0.87). Total reproductive isolation was 0.997, with the largest relative contributions by geography (41%) and pollinator fidelity (44%). Prezygotic barriers accounted for 97.6% isolation overall; however, tetraploids were more assortatively mating (98%) than diploids (79%). Realized reproductive isolation between ploidy levels in sympatric populations was 87% and tetraploids produced significantly fewer triploids than did diploids. Simulations indicated that the observed prezygotic isolation will reduce the strength of minority disadvantage acting on tetraploids and increase the importance of differences in viability and fertility between cytotypes in regulating polyploidy establishment.


Biologists view polyploidy as both a prevalent and distinct mechanism of speciation in plants. Its uniqueness is attributed to the fact that it can arise seemingly instantaneously through a single genetic event and can persist, owing to the strong genetic incompatibility expected between diploids and their polyploid descendants. For these reasons, genome multiplication is often identified as one of the few speciation mechanisms that can operate in sympatry (Mayr, 1942; Briggs & Walter, 1997; Futuyma, 1998). This perspective ultimately has its roots in the mutationist views of the early 20th century, in which new species were viewed as the instantaneous product of rare macro-mutations and little importance was ascribed to population processes such as natural selection (Mayr, 1982).

Evolutionary biologists, however, are discovering that the process of polyploid speciation may be more complex. Recent research shows that many genic and genomic changes can occur within the early generations after polyploid formation (Song et al., 1995; Galitski et al., 1999; Soltis & Soltis, 1999; Soltis & Soltis, 2000; Ozkan et al., 2001; Shaked et al., 2001; Adams et al., 2003). This suggests that polyploids are far from stable when they first arise, and that selection may be instrumental in transforming neopolyploids into fertile and viable genotypes. Second, theoretical studies show that, with strong postzygotic reproductive isolation, certain population processes may actually restrict the establishment of polyploids after they have been formed (Levin, 1975; Fowler & Levin, 1984; Felber, 1991). Arguably, the most important barrier is ‘minority cytotype exclusion’ (Levin, 1975), which refers to the theoretical fact that, with random mating between diploids and polyploids, equilibrium frequencies will be governed by positive frequency-dependent selection. In particular, rare polyploids will experience a fitness disadvantage because they most often mate with the majority cytotype (i.e. diploids), a union that is largely ineffectual. Therefore, tetraploids that occur below a threshold frequency (0.5, when diploid fitness [w2x] = tetraploid fitness [w4x]) will be selectively eliminated unless there are counteracting forces such as recurrent tetraploid formation (Felber, 1991; Husband, 2003a), stochastic events (Levin, 1975) or assortative mating (Rodriguez, 1996).

Reproductive isolation, and thus assortative mating, is the hallmark of species formation but little is known of its magnitude, underlying mechanisms or consequences in polyploid systems. Polyploid species are often presumed to be isolated from their diploid progenitors by strong postzygotic barriers (triploid block: Ramsey & Schemske, 1998) yet the magnitude of prezygotic factors and the relative contributions of pre- and postzygotic barriers to within-ploidy mating are unknown. The goal of this paper is to provide an overview of recent and new research on reproductive isolation between diploids and tetraploids in fireweed, Chamerion (formerly Epilobium) angustifolium L. Holub (Onagraceae). For the past 10 yr, Husband and coworkers have been studying polyploid speciation in C. angustifolium (Husband & Schemske, 1998; Husband & Schemske, 2000; Husband, 2000; Burton & Husband, 2001; Husband et al., 2002; Husband, 2003a) by investigating the population processes governing the formation and establishment of polyploids. Here our specific objectives were to: synthesize research on several potential reproductive barriers to gene flow and evaluate their individual and combined effects; estimate rates of reproductive isolation between cytotypes based on the production of triploid hybrids in the field; and use computer simulations to evaluate the impact of prezygotic isolating mechanisms on polyploid establishment in C. angustifolium and in general.

Background on Chamerion angustifolium

Chamerion angustifolium, known as fireweed, is a herbaceous perennial plant of open and disturbed habitats found throughout the northern hemisphere. In North America, C. angustifolium consists of diploids (2n = 2x = 36) and autotetraploids (2n = 4x = 72) (Mosquin, 1967). While they may differ in overall size, these cytotypes are very similar morphologically and are often indistinguishable in the field. The two ploidies are largely allopatric, with diploids occurring at higher latitudes than tetraploids (Mosquin, 1967; Flint, 1980; Husband & Schemske, 1998). However, they do co-occur in a contact zone that runs along the southern border of the boreal forest and in the Rocky Mountains. Research has largely been focused on this zone, as the cytotypes occur in close proximity, and the reproductive and competitive interactions that directly influence their evolutionary dynamics can be studied.

Two important results regarding the evolution of polyploidy have emerged from studies of C. angustifolium from the contact zone. First, using experimental populations that differed in the proportions of diploids and tetraploids, Husband (2000) tested whether minority cytotype exclusion could operate in populations of C. angustifolium. As expected, the relative fitness of tetraploids was low when they were in the minority (33%) and increased linearly with their frequency in the population. Second, from a series of studies on unreduced gamete production (Burton & Husband, 2001; P. Kron & B.C. Husband, unpublished) and the fitness of plants of different ploidy (Burton & Husband, 2000; reviewed in Husband, 2003a), there is evidence that new tetraploids are produced recurrently by diploids at rates of up to 2.4 tetraploids per 1000 zygotes each generation (Husband, 2003a). This process represents a unilateral force favouring the maintenance of polyploids at low frequencies. However based on simulations, this rate is about one third of that required (9 × 10−3/generation) to overcome minority cytotype exclusion and cause the fixation of tetraploids (Husband, 2003a). Together, these studies suggest that other factors must operate to overcome the initial frequency disadvantage that new polyploids inevitably experience. Current research on C. angustifolium is focused on the role of reproductive isolating mechanisms in this process.

Barriers to diploid-tetraploid mating in Chamerion angustifolium

Despite the central role of reproductive isolation in most speciation models, there are no comprehensive estimates of isolation between closely related diploid and tetraploid plant taxa. There are many published estimates of postzygotic barriers (i.e. the fitness of within- vs between-ploidy crosses, known as triploid block); however, most of these are for crop plants (Ramsey & Schemske, 1998) and focus on a limited number of life stages (seed production). Some prezygotic mechanisms have been described for natural populations, particularly geographic isolation and phenology (summarized in Levin, 1983; Petit et al., 1999; Levin, 2002), but these barriers are reported in a piece-meal fashion and their relative contributions to total reproductive isolation are rarely considered. Here we examine reproductive isolation between diploid and tetraploid C. angustifolium owing to five prezygotic (geographic isolation, flowering asynchrony, pollinator fidelity, self-fertilization and gametic selection) and two postzygotic barriers (triploid block, inbreeding depression), as derived from published as well as new research. All measures are based on research within the contact zone, where there is potential for gene flow between diploids and tetraploids. For each reproductive barrier we estimate isolation (RI) as the minimum proportion of matings that are within-ploidy. Reproductive isolation is also calculated for each ploidy separately (RI4x, RI2x) to determine whether there are any consistent asymmetries in the direction of gene flow. We then combine these independent measures into a single cumulative measure of reproductive isolation that can be compared to realized rates of gene flow in natural populations.

Geographic isolation

Geographic isolation between diploids and tetraploids was estimated as the proportion of populations within the contact zone that are composed of a single ploidy. This approach assumes that gene exchange among populations is insignificant relative to that within populations and probably underestimates the opportunities for gene flow. Fifty-one populations from throughout the central Canadian Rocky Mountains were surveyed for ploidy. The frequencies of diploids, triploids and tetraploids were estimated by sampling a minimum of 30 plants (separated by > 2 m) from each population and inferring their ploidy using flow cytometry (Burton & Husband, 2001). Overall, 59% (30) of populations contained two or more ploidies (H.A. Sabara & B.C. Husband, unpublished) and therefore geographic isolationinline image was 0.41(Table 1). Of the 21 monocytotypic populations, 19 were tetraploid and only 2 were diploid. Therefore, 61% of populations that contained tetraploids also had other cytotypes (RIgeog, 4x = 0.39), whereas 94% of populations with diploids were mixed (RIgeog, 2x = 0.06). As in previous studies (Flint, 1980; Husband & Schemske, 1998), geographic isolation was associated with the tendency for diploids to occur at higher elevations than tetraploids.

Table 1.  Measures of reproductive isolation (RI) and the relative contributions (RC) to total reproductive isolation for five prezygotic and two postzygotic reproductive barriers between diploid and tetraploid Chamerion angustifolium. Postzygotic isolation is based on the combined effects of triploid block and inbreeding depression. Values are given for diploids and tetraploids separately and combined
Reproductive barrierOverallDiploidTetraploid
Geographic isolation0.41 41.10.06  6.40.39 38.9
Flowering asynchrony0.13  7.70.13 12.50.00  0.0
Pollinator fidelity0.85 43.80.43 35.90.83 50.9
Self-fertilization0.44  3.40.45 21.40.43  4.5
Gametic selection0.44  1.90.17  4.50.70  4.2
Postzygotic0.87  2.10.89 19.30.85  1.5
Total isolation0.997100.00.977100.00.997100.0

Flowering time

Within mixed-ploidy populations, gene exchange may be further reduced by asynchrony in flowering time. Researchers have repeatedly noted that autopolyploids flower either earlier or later than closely related diploids (Levin, 1983; Levin, 2002). In the case of C. angustifolium, tetraploids begin flowering an average of 10 d later than diploids, based on two glasshouse comparisons of eight diploid and nine tetraploid populations combined (Husband & Schemske, 1997; Husband, 2003b) and a field study of one mixed-ploidy population (Husband & Schemske, 2000). The degree of overlap over the entire flowering period was estimated in a study involving 300 diploid and tetraploid plants grown in a common garden (Husband, 2000). In this case, diploids began flowering in mid June, tetraploids flowered 8 d later and both ploidies stopped flowering in August Overall, diploids and tetraploids were both flowering for 87% of the entire flowering period. Phenological isolation, the converse of overlap inline image, was therefore 0.13(Table 1). Ploidy-specific phenological isolation (RIphen, 4x, RIpheno, 2x) was 0.13 and 0.0 for diploids and tetraploids, respectively, reflecting the fact that tetraploid flowering is completely encompassed within the diploid flowering period. These estimates indicate that flowering is asynchronous although preliminary data suggest the extent may vary widely in the field (H.A. Sabara & B.C. Huunpublished).

Pollinator fidelity

Within mixed-flowering populations, the interactions between pollinators and plants of different ploidy will ultimately determine whether cross-pollination will ever occur (Grant, 1949, 1994). If pollinators restrict their foraging to only a single ploidy, or if they are ineffective at transferring pollen because of mismatches between flower and bee morphology, then reproductive isolation will be high. Unfortunately, such interactions are almost never examined in polyploid systems (although see Taylor & Smith, 1979; Segraves & Thompson, 1999). Husband & Schemske (2000) measured the strength of pollinator fidelity in one mixed population by identifying the ploidy of flowering stems in a small plot and then following the sequence of flights by three bumblebee species within and between stems over a 4 d period. All three types of bees were observed on both cytotypes and had very similar foraging patterns. In general, pollinator flights occurred between tetraploid stems 20%−40% (depending on bee species) more often than expected based on the frequency of tetraploids in the plot. Fifteen percent of flights were between stems of different ploidy. Therefore, reproductive isolation caused by pollinator fidelity inline image was 0.85 (Table 1). Pollinator isolation for each ploidy separately, estimated as the proportion of flights to a specific ploidy that originated from a plant of the same ploidy, was 0.83 for tetraploids and 0.43 for diploids (Table 1). While these results suggest that pollinators visit a disproportionately high number of tetraploids, it is important to point out that this pattern of foraging was accounted for by the fact that cytotypes were clumped to different degrees and bees tended to forage over short distances and between neighbouring stems (Husband & Schemske, 2000). Additional work is being undertaken to determine whether tetraploids generally occur in larger patches than diploids as they did in this particular study.


In plants that are hermaphroditic, pollinators can also promote reproductive isolation by causing self-pollination through the transfer of pollen within or between flowers on the same plant. In a previous study of C. angustifolium, bees visited 2.6 flowers per diploid stem and 3.7 on tetraploids (Husband & Schemske, 2000). With 8–10 open, hermaphroditic flowers per inflorescence, most plants may therefore experience substantial self-pollination. A previous genetic analysis of mating, based on the segregation of allozymes in open-pollinated progeny arrays, confirmed that 45% and 43% of progeny from diploids and tetraploids, respectively, were the product of self-fertilization (Husband & Schemske, 1997, 1995). Reproductive isolation causedby selfing inline imageis represented by the average (RIselfing = 0.44) of these ploidy-specific values (Table 1). Because flowers in this species are protandrous (male organs mature before female), most selfing likely occurs as a result of the transfer of pollen between flowers on a stem rather than within flowers (Routley & Husband, 2003). Irrespective of the mechanism, these studies suggest that self-fertilization is an important cause of assortative mating in C. angustifolium, and that its impact is similar in both cytotypes.

Gametic selection

Pollinator foraging patterns and rates of selfing provide a rough measure of the likelihood of within- vs between-ploidy pollen transfer, assuming that pollen deposited on any given flower is derived largely from the previous flower visited. In fact, Galen & Plowright (1985) found that pollen deposition on C. angustifolium flowers declined rapidly with successive visits, but was detectable even after five flowers. Thus, even with a strong pollinator preference for tetraploid flowers, many flowers will receive some combination of both diploid and tetraploid pollen. This raises the question of whether pollen from diploids and tetraploids have different siring abilities when applied to the same stigma. To investigate this, Husband et al. (2002) conducted single and mixed-ploidy pollinations on both diploid and tetraploid recipients. They then compared paternity of seeds from each mother using flow cytometry, and compared patterns of pollen tube attrition down the style in single donor pollinations. Interestingly, in mixed pollinations, pollen from tetraploids consistently sired more seeds (mean = 77%) than diploids. This pattern was similar on diploid mothers (83% tetraploid paternity) and tetraploids (70%), which suggests that tetraploid plants will experience a higher rate of assortative mating than diploids when they receive mixtures of pollen. In total, gametic isolationinline image was 0.44 forall plants and was 0.17 and 0.70 for diploids and tetraploids separately. These results were echoed in the studies of pollen tube growth in the style, with more tubes from tetraploids occurring at the base of the style and higher rates of attrition in pollen from diploids (Husband et al., 2002).

Postzygotic barriers

The most widely studied barriers to gene exchange between diploids and polyploids are postzygotic and involve mechanisms that lead to hybrid inviability (triploid block) and sterility (Ramsey & Schemske, 1998; Petit et al., 1999). These mechanisms have been of interest not so much for their significance to polyploid speciation as their practical implications for plant breeding and the manipulation of germ plasm. In C. angustifolium, the fitness of triploid hybrids has been compared to diploids and tetraploids in a glasshouse comparison (Burton & Husband, 2000). Mean seed formation in 2x× 4x crosses was only 45% of that in within-ploidy pollinations, which corresponds to triploid block of roughly half that estimated in a recent survey of the plant literature (80%−95%; Ramsey & Schemske, 1998). Furthermore, survival was as high in triploid offspring as it was for diploids and tetraploids; however, pollen production and viability were significantly less (Burton & Husband, 2000). Taken together, the relative fitnesses of diploids, triploids and tetraploids were 1, 0.09, and 0.61, respectively, demonstrating that selection against triploids and, hence, between-ploidy mating is strong but not complete.

A second form of postzygotic selection affecting reproductive isolation is inbreeding depression, or reduced fitness of selfed vs outcrossed offspring. To the extent that within-ploidy matings are the result of self-fertilization, inbreeding depression may reduce the benefits of assortative mating and hence diminish reproductive isolation. Inbreeding depression has been documented in both diploid and tetraploid C. angustifolium by comparing the performance of their offspring derived from self- and cross-pollinations (Husband & Schemske, 1997). On average, the fitness cost of selfing was 30% less in tetraploids than diploids, which translates into inbreeding depression values (inbreeding depression= inline image) of 0.67 and 0.95, respectively (Husband & Schemske, 1997). We then adjusted the overall fitness measures of diploids, triploids and tetra-ploids by multiplying their glasshouse fitnesses (described above) by their respective rates of self-fertilization and relative fitnesses of selfed offspring. Under the assumption that triploids have selfing rates and inbreeding depression rates intermediate to tetraploids, we find the relative fitness of diploids, triploids and tetraploids to be 1, 0.11 and 0.75, respectively. The reduction in fitness of triploids relative to their parents indicates that postzygotic isolationinline image is 0.87, withploidy specific rates of 0.89 (RIpostzygotic, 2x) and 0.85 (RIpostzygotic, 4x) (Table 1).

Total reproductive isolation

We have examined seven possible reproductive barriers (five prezygotic, two postzygotic) in sequence, and their individual effects on reproductive isolation between diploid and tetraploid C. angustifolium. To assess their contribution to total reproductive isolation, we used a multiplicative function of the individual components. We applied a method described by Coyne & Orr (1998) and extended by Ramsey et al. (2003) in which the absolute contribution (ACn) of the nth component of reproductive isolation (RIn) is calculated as:

image(eqn 1)

ACn, then, represents the probability of isolation caused by barrier n, after all previous reproductive barriers in the sequence have been considered. For m isolating barriers, total reproductive isolation (RItotal) is:

image(eqn 2)

and varies from 0 to 1. Following Ramsey et al. (2003), we also estimated the relative contribution of different barriers (RCi for i = 1 to n) to total isolation (RItotal) as:

image(eqn 3)

The relative contribution for any given reproductive barrier will approach the absolute contribution as total reproductive isolation approaches 1 (i.e. isolation is complete).

Considering all five prezygotic and both postzygotic barriers (combined), the overall cumulative measure of reproductive isolation in C. angustifolium was 99.7% (Table 1, Fig. 1). In other words, at least 99% of the ovules in diploid and tetraploids, combined, should be sired through within-ploidy mating. This measure of reproductive isolation is, of course, a crude estimate that neglects some potential barriers (e.g. geographic isolation outside of the contact zone; different pollen transfer efficiencies caused by morphological differences), and relies on limited data for others (e.g. pollinator fidelity). Nevertheless, we believe our value is a good first estimate and is likely a minimum value because additional barriers are most likely to further reduce the chance of between-ploidy mating.

Figure 1.

Cumulative measures of reproductive isolation between diploid and tetraploid Chamerion angustifolium based on five prezygotic and two postzygotic reproductive barriers. Values are given for overall isolation (open columns) as well as diploid-specific (grey columns) and tetraploid-specific (closed columns) measures after the successive addition of each reproductive barrier.

Our analysis provides a number of insights into reproductive isolation that are often not available in polyploid systems. First, it is immediately obvious that reproductive isolation is very strong and a product of many different barriers rather than one dominant one (Table 1; Fig. 1). In fact, all barriers examined contributed to isolation. That being said, the contributions that each barrier makes are far from uniform. For example, prezygotic barriers combined cause 0.98 isolation between cytotypes, which is nearly 99% of total reproductive isolation. This is not to say that postzygotic factors are not strong, just that they act on a small proportion of ovules that are not isolated by prezygotic barriers. In other words, ecological barriers may play a much larger role in reproductive isolation of polyploids than is often realized. Among the prezygotic barriers, geographic isolation and pollinator fidelity make a proportionally large contribution (41% and 44%, respectively) to cumulative fitness, whereas flowering time, a stage that has traditionally received attention, contributes less (7.7%; Table 1; Fig. 1).

Perhaps most surprising, the overall measure of prezygotic isolation (0.98) masks the marked asymmetry in rates of assortative mating between diploids and tetraploids (Fig. 1; Table 1). When the effects of all prezygotic barriers are included, diploids are only 79% isolated compared to 98% in tetraploids. This difference in cumulative isolation is largely the result of three specific barriers, geographic isolation, pollinator fidelity and gametic selection, in which assortative mating is reinforced in tetraploids more than diploids. The net effect of this difference is that diploids will be more likely to experience hybrid matings in sympatric populations and will be influenced more by both the cost of producing triploids and frequency of tetraploids in populations. This difference has important implications for polyploid establishment, as described below.

Realized gene flow between diploid and tetraploid C. angustifolium

The most direct method for evaluating the magnitude of reproductive isolation between diploids and tetraploids is to observe mating patterns in natural populations and from that infer realized rates of gene exchange. We recently did this in C. angustifolium by sampling 30 seed families from each of 10 mixed-ploidy populations in the Canadian Rocky Mountains and inferring the ploidy of 10 offspring per family using flow cytometry. We estimated the degree of reproductive isolationas inline image. Because the likelihood ofhybridization between cytotypes will depend on the frequency of diploids and tetraploids in each population, the expected frequency of triploids was based on random mating and is estimated as two times the product of the diploid and tetraploid frequencies. In this calculation, we assume that triploid production, as a result of unreduced gametes and matings involving triploids, is negligible.

On average, 200 offspring were screened in each population. Of these, an average of eight (4.1%) were triploid, which is significantly less than the 62 (31%) expected under random mating (G = 108.6, P < 0.0001; Fig. 2a; H.A. Sabara & B.C. Husband, unpublished). Based on this retrospective approach, reproductive isolation was 87%, which is less than the 98% value derived from all measured prezygotic reproductive barriers. It should be noted, however, that the realized estimate is based on mixed populations and does not consider the role of geographic isolation within the contact zone. In addition, we have used a highly conservative measure of expected triploid production. In fact, under complete reproductive compatibility, between-ploidy mating could be considerably higher than what would be observed through random mating. These two caveats may explain the 11% discrepancy between the realized and expected rates of reproductive isolation.

Figure 2.

Realized rates of between-ploidy mating based on the frequency of triploids in open-pollinated progeny arrays from 10 mixed populations of Chamerion angustifolium (H. Sabara & B.C. Husband, unpublished). (a) Mean rates of triploid production compared to random expectations, which are based on the frequency of diploids and tetraploids in the population. (b) Ploidy–specific rates of gene exchange as measured by triploid production and expressed as the percent deviation from random.

We were also able to examine the rate of triploid production for diploid and tetraploid mothers separately using data from two populations (the two cytotypes were not sufficiently represented in the other mixed populations). Relative to random expectations, triploid offspring were produced much less often by tetraploid mothers and close to expected values by diploids (Fig. 2b; H.A. Sabara & B.C. Husband, unpublished). These results are consistent with our earlier observation that reproductive isolation is asymmetric and higher in tetraploids than diploids.

Significance of assortative mating for polyploidy establishment

In polyploid systems, reproductive isolation will obviously affect gene exchange; however, its effect on the likelihood and rate of polyploid establishment is less clear. Levin (1975) and Felber (1991) showed that strong postzygotic isolation contributes to minority cytotype exclusion and thus restricts the establishment of rare polyploids within diploid populations. On the other hand, it is often assumed that prezygotic isolation (assortative mating) will counteract this process and promote the evolution of polyploids; however, theoretical studies by Levin (1975) and Rodriguez (1996) on self-fertilization suggest that the effects of assortative mating may not be so straightforward.

We examined the general effects of prezygotic, assortative mating on polyploid speciation in a simulation analysis. Each simulation consisted of a population of 1000 individuals with nonoverlapping generations. Individuals were initially assigned a ploidy (diploid or tetraploid) and the population was then passed through a series of selection and mating phases. During selection, individuals were subject to selective mortality in proportion to their ploidy-specific relative fitnesses (a combination of viability and fertility) and their frequencies then standardized to one. During mating, n = 1000 pairs of individuals were drawn at random from the population to serve as parents, each of a single progeny. Each individual had a probability, a, of mating assortatively (with an individual of the same ploidy), or, 1−a, randomly. The ploidy of the offspring was determined according to the offspring ploidy distributions for different parental combinations. For example, if the parental pair consisted of two diploids, the likelihood of producing a diploid offspring was proportional to the squared frequency of reduced gametes (n = x) produced by diploids; triploid offspring would occur at the rate of two times the frequency of reduced gametes multiplied by the rate of unreduced gametes (n = 2x = 2n), and tetraploid offspring would occur at a rate equivalent to the square of the rate of unreduced gamete production. We assumed that tetraploids produced only n = 2x pollen, triploids produce equal portions of x and 2x gametes, and pentaploid and hexaploid offspring were not viable. Each simulation was run for 1000 generations and was replicated 100 times to allow for stochastic variation.

As Levin (1975) pointed out and as illustrated through our simulations (Fig. 3a), prezygotic, assortative mating affects polyploid dynamics by reducing the proportion of matings that are frequency-dependent (i.e. influenced by minority cytotype exclusion). Thus, it does not affect the rate of tetraploid formation, only their frequency once they have arisen. Our simulations started with an initial tetraploid frequency arbitrarily set at 0.4. In addition, diploids and tetraploids were assigned equal fitness and there was no recurrent tetraploid formation (i.e. unreduced gamete production = 0). Not surprisingly, with complete random mating (assortative = 0), tetraploids were quickly eliminated as a result of minority cytotype exclusion, because they were below the threshold frequency of 0.5. Increasing the rate of assortative mating had no effect on the outcome except to delay it (Fig. 3a). This was true as long as some portion of the fertilizations were frequency-dependent. Once assortative mating was complete (i.e. assortative = 1), the population was free from the effects of minority cytotype exclusion and, at that point, cytotypes either coexisted (in the absence of any deterministic effects) or their frequencies changed as a function of drift or environmentally based selection. Therefore, depending on whether tetraploids are more or less fit than diploids, assortative mating will either enhance or diminish polyploid establishment.

Figure 3.

Temporal changes in autotetraploid frequencies (diploids, not shown) based on individual-based computer simulations. (a) Population has an initial tetraploid frequency of 0.4, with no recurrent tetraploid formation, and the fitnesses of tetraploids and diploids = 1. Results are shown for four rates of assortative mating (0, 0.4, 0.8, 1). (b) Parameter values the same as in (a) except tetraploids are produced recurrently through the union of unreduced gametes produced by diploids (UG = 0.03).

When tetraploids were produced recurrently through the union of unreduced gametes, then assortative mating qualitatively changed the evolutionary outcome (Fig. 3b). Under such conditions, a small increase in assortative mating had the effect seen previously, which is to slow the eventual elimination of tetraploids. However, when assortative mating was sufficiently high (0.8 in this case), the impact was fundamentally different. In this case, tetraploids spread to fixation even though their initial frequency was well below the threshold of 0.5. This is because high assortative mating slows the loss of tetraploids through minority exclusion. With high assortative mating, the rate at which tetraploids were excluded as a result of minority disadvantage was less than the rate at which they were added through the union of unreduced gametes. In this way, assortative mating expands the conditions under which tetraploids become established.

The asymmetry in assortative mating between tetraploids and diploids also affected both the direction and rate of tetraploid evolution (Fig. 4). To illustrate, we present the results from a series of simulations in a population with an initial tetraploid frequency of 0.2 (unreduced gametes production = 0.03 and diploid fitness = tetraploid fitness) and assortative mating ranging from symmetrical (4x = 0.8, 2x = 0.8) to highly asymmetric (4x = 0.8, 2x = 0). The results indicate that the greater the asymmetry is in favour of tetraploids, the more likely and the more rapid tetraploid fixation becomes. This is consistent with the idea that tetraploid advantage increases because it not only suffers less from minority cytotype exclusion but simultaneously reduces the transmission of diploids in the next generation through hybrid matings on diploids.

Figure 4.

Temporal changes in autotetraploid frequency when tetraploids and diploids experience different rates of assortative mating. Each pair of numbers represents assortative mating for tetraploids and diploids, respectively. For each simulation, N = 1000, fitness of diploids and tetraploids were equal (triploid fitness = 0) and unreduced gametes were produced at a rate of 0.03 by diploids individuals.

How, then, does assortative mating affect tetraploid – diploid dynamics in Chamerion angustifolium? To examine this, we incorporated previous estimates of assortative mating, ploidy-specific relative fitness as well as unreduced gamete production into the simulations and examined the fate of tetraploids at three different initial frequencies (0.2, 0.4, 0.8). When the fitnesses of diploids and tetraploids were constrained to 1 (w2x = w4x = 1), assortative mating enhanced tetraploid fixation (Fig. 5a). Specifically, with high assortative mating, tetraploids spread to fixation at initial frequencies as low as 0.2, well below the 0.5 threshold. These results reinforced the earlier observation that showed that assortative mating can increase the range of conditions favouring tetraploids when tetraploid fitness is equal to or greater than that of diploids. In C. angustifolium, however, our only measure of fitness is from a glasshouse study that shows tetraploids to be 39% less fit than diploids (Burton & Husband, 2000). Under this condition, high assortative mating was actually disadvantageous to tetraploids (Fig. 5b), as they were excluded even with initial frequencies as high as 0.8. This likely happens because assortative mating reduces the proportion of matings affected by minority cytotype exclusion. Hence, the dominant force governing the dynamics of mixed-ploidy populations shifts from the relative frequency of cytotypes in the population to their relative fitness in the local environment. As a result, the low fitness of tetraploids determines the outcome over a much larger range of conditions than it would if mating was random. Apparently, assortative mating can minimize not only the disadvantages of minority cytotype exclusion but also its advantages to the spread of tetraploids in diploid populations.

Figure 5.

Temporal changes in autotetraploid frequencies using mating parameters based on data from Chamerion angustifolium. Three different initial frequencies of tetraploids were used. (a). fitness of diploids is equal to tetraploids and triploids are lethal (b). Fitness of diploids, triploids and tetraploids are as in C. angustifolium (1, 0.11, 0.75).


Collectively, our research represents the first comprehensive analysis of reproductive isolation between extant autopolyploids and their diploid progenitors. The analysis of seven reproductive barriers (five prezygotic, two postzygotic) and estimates of realized rates of between-ploidy mating point to the same general conclusions. First, diploids and tetraploids are strongly but not completely reproductively isolated. Second, prezygotic, ecological factors such as plant–pollinator interactions and geography play a potentially large role in achieving this isolation. This pattern is in contrast to the considerable emphasis traditionally placed on postzygotic, genetic factors such as triploid block. While these barriers are indeed present, their impacts on isolation in natural populations are relatively small because they act comparatively late in regulating mating. Thirdly, the magnitude of prezygotic isolation is much higher in tetraploids than diploids. This strong assortative mating has the effect of reducing the strength of minority cytotype exclusion to the point where other factors, such as local adaptation, recurrent tetraploid formation and genetic drift may dominate the evolutionary dynamics of the contact zone.

Our results represent patterns of mating between extant diploids and their tetraploid descendants. What is unclear, however, is which barriers have arisen as a direct consequence of chromosome doubling and which have evolved subsequently as a result of drift and/or selection. These questions are difficult to address and indeed plague most studies of the evolution of reproductive isolation and speciation. The good news is that polyploid systems offer a unique opportunity to address this by synthesizing new polyploids (neopolyploids) from existing diploids and exploring their behaviour in subsequent generations. Polyploids have been synthesized numerous times in crop plants and Ramsey & Schemske (2002) review their genetic and morphological characteristics. More recently, synthesized polyploids have been used to study the genetic and genomic changes that occur after chromosome doubling (Song et al., 1995). No information, however, is available about the magnitude of reproductive isolation between diploids and neopolyploids or the specific reproductive barriers that arise as a result of genome multiplication. Without this information, it will be extremely difficult to determine the rate or pathways by which polyploid speciation occurs.

Our research findings also have important taxonomic implications. Historically, autopolyploids have rarely been given separate species designations from their diploid ancestors, and C. angustifolium is consistent with this observation. This practice is usually a result of the close morphological resemblance between cytotypes, rather than any information about gene exchange. If C. angustifolium is representative at all, we suggest that this nomenclature masks the fact that little or no gene exchange is occurring between these taxa. This in itself may not be a serious problem, but what is more puzzling is that these criteria have not been applied evenly across plant groups. Interestingly, while very little information on reproductive isolation is available for allopolyploids, they have more often been given separate species status from their parents. While this fact has little bearing on the study of polyploidy as a speciation mechanism, the taxonomic importance of autopolyploidy is likely underestimated as a result.

Overall, our results suggest that minority cytotype exclusion is neutralized by strong prezygotic isolation, especially in tetraploids, and that an understanding of the distribution of ploidies within and between populations of C. angustifolium depends on knowledge of the relative fitness of diploids and tetraploids in different environments. Under strong assortative mating, tetraploids are most likely to invade diploid populations when there is ecological differentiation and local environments favour tetraploids. Although morphological and physiological differentiation has been reported between diploids and their tetraploid descendants, we are not aware of any studies such as reciprocal transplants that test for ploidy–environment interactions along ecologically meaningful gradients. Interestingly, in an age when the ecological aspects of speciation are receiving more attention in homoploid organisms (Schluter, 2001), it appears that similar attributes may be the final arbiters of success in tetraploid species.


We thank B. Mable for comments on the manuscript, A. Crawford, D. Duval, B. Kennedy, P. Kron, M. Ross, and J. Sabau for field and lab assistance, and the Natural Sciences and Engineering Research Council of Canada, Premier's Research Excellence Award and Canada Research Chair for financial assistance to BCH.