Evolutionary consequences of autopolyploidy

Immediate structural changes after genome doubling

Studies of structural changes in synthetic autopolyploids have reported contrasting trends (Table 1). Immediate loss of c. 17% of total DNA has been observed in synthetic autopolyploids of Phlox drummondii, with a further reduction of up to 25% upon the third generation (Raina et al., 1994). This decrease in genome size was accompanied by a substantial increase in seed set, suggesting an adaptive diploidization process. In the exclusively bivalent pairing Elymus elongatus, both newly synthesized autopolyploids as well as natural accessions showed a loss of 10% DNA compared with their diploid progenitors (Eilam et al., 2009). Reduction in DNA content immediately follows genome doubling in this species and possibly sustains bivalent pairing. By contrast, Santos et al. (2003) revealed no major chromosomal rearrangements and a preponderance of multivalent pairing in a synthetic autopolyploid of Arabidopsis thaliana, for which Ozkan et al. (2006) reported no loss of DNA. In addition, genetic markers did not reveal any structural changes during the first generations following autopolyploid formation in A. thaliana (Ozkan et al., 2006). A synthetic autotetraploid of Paspalum notatum exhibited 9.5% sequence elimination relative to its diploid progenitor (Martelotto et al., 2007).

In conclusion, rapid elimination of DNA sequences from autopolyploid genomes seems to occur, but the limited data available so far imply that autopolyploids experience less genome restructuring than allopolyploids. To what extent differentiation of homologous chromosomes is required for the proper functioning of nascent autopolyploid genomes remains uncertain.

Long-term structural changes in autopolyploids

Genome restructuring occurring throughout the lifespan of an established polyploid lineage may promote its long-term diversification and persistence. Lim et al. (2007) nicely illustrated that insertion, deletion and homogenization of repetitive sequences in Nicotiana allopolyploids provoked a complete turnover of nongenic portions of the genome in < 5 million yr, leading to structural diploidization within a well-conserved karyotype structure. Comparable data are lacking for autopolyploids, but it is tempting to speculate that genetic redundancy leads to long-term genome instability and that similar sequence turnover may be at work after chromosome doubling to gradually restore a diploid-like behavior of autopolyploid genomes.

Established autopolyploids generally experience a significant decrease in chromosome length (Rivero-Guerra, 2008). For example, the young autotetraploid Biscutella laevigata shows successive down-sizing of its genome along its post-glacial recolonization pathway (König & Müllner, 2005). Genome size variation in established autopolyploids in the Triticeae reveals a variety of patterns (Eilam et al., 2009; Table 1). Whereas the multivalent forming Hordeum bulbosum showed no significant change in genome size compared with its diploid progenitor, the strictly bivalent forming Hordeum murinum subsp. murinum showed a 23% reduction in the amount of DNA. However, the bivalent forming Hordeum marinum subsp. gussoneanum unexpectedly showed the exact additive genome size of its progenitor, indicating that at least in some cases there is no association between genome down-sizing and pairing behavior of chromosomes. Genome arrangements might also involve the relocation of chromosomal segments. Asymmetrical relocation and loss of rDNA loci has been reported for the autopolyploid Hepatica nobilis var. pubescens (Weiss-Schneeweiss et al., 2007). In a synthetic autotetraploid lineage of Arabidopsis thaliana grown for 20–30 generations, translocation of the 45S rDNA from chromosome 4 to chromosome 3 has been identified (Weiss & Maluszynska, 2000). This accession forms > 85% bivalents, which contrasts with multivalent pairing observed in newly synthesized autopolyploids (Santos et al., 2003), suggesting that diploidization proceeded through structural rearrangements within 30 generations.

In contrast to nascent autopolyploids, established autopolyploids show substantial genome reorganization compared with their diploid relatives. Structural changes thus accumulate over time (Table 1). However, additional studies addressing the timing and abundance of restructuring after genome multiplication per se are clearly required.

Short-term functional changes after genome doubling

In synthetic autopolyploids of Isatis indigotica, only 4.2% of the genes analysed by cDNA microarrays changed expression patterns compared with diploids (Lu et al., 2006). Using a similar approach, Stupar et al. (2007) reported only few significant expression changes among 9000 genes surveyed in synthetic autopolyploids of potato (Solanum phureja). Martelotto et al. (2005) reported that 129 out of 9617 genes surveyed (1.34%) changed expression levels in synthetic autopolyploids of P. notatum. Investigating the proteome, Albertin et al. (2005) revealed no significant changes in gene expression in synthetic autopolyploids of Brassica oleracea, while an analogous study on synthetic allopolyploid Brassica napus found up to 38% of nonadditive gene expression compared with the parental species (Albertin et al., 2007).

To what extent autopolyploidy induces a ‘rewiring’ of the regulatory network remains largely unknown, but the limited evidence available suggests that genome doubling per se induces only a little functional reorganization. Accordingly, since structural and functional reorganization seem to be limited after genome multiplication per se, only a few evolutionary novelties or transgressive traits (i.e. exceeding the range of parental values; Rieseberg & Willis, 2007) might be immediately available for nascent autopolyploids to establish via adaptive processes.

Long-term functional changes and selection in autopolyploids

Little is currently known about the functional reorganization of autopolyploid genomes in the long term. A recent analysis of naturally occurring populations of diploid, autotetraploid and autohexaploid Helianthus decapetalus with cDNA-AFLPs revealed only 6.6% variation among established ploidy levels (Church & Spaulding, 2009), suggesting that gene expression is not strongly affected by genome doubling in the long term.

Theoretically, genetic redundancy may, however, imply relaxed selection on the copies of duplicated genes and allow them to follow new evolutionary trajectories. Long-term drift, mutation and selection mostly lead to inactivation of duplicated genes, but a certain fraction of duplicated alleles are expected to fix by adaptive processes, because they either acquire a new function (i.e. neofunctionalization) or evolve towards complementation functions (i.e. subfunctionalization; Lynch et al., 2001). Whether such models apply to autopolyploids is speculative, and we lack compelling data supporting the diversification of redundant alleles into evolutionary innovations in autopolyploids (Flagel & Wendel, 2009).

In this context, a crucial question is whether autopolyploids respond more rapidly to selection than their diploid progenitors (Ramsey & Schemske, 2002). Natural selection is theoretically more difficult to analyse in autopolyploids than in diploids, because of their complex population genetics (Luo et al., 2006). Response to selection is always greater in diploids than in autotetraploids, but is less dependent on the intensity of selection for dominant than for recessive or additive alleles in autopolyploids (Bever & Felber, 1992). Although this suggests that autopolyploids adapt more slowly than diploids, adaptive processes may be more efficient after genome doubling as effective population size rises, rendering selection more efficient, and increases in fitness seem to depend more on the establishment of new mutations or alleles than on the efficiency of selection, at least in moderately sized populations (Otto & Whitton, 2000).

In conclusion, genetic redundancy could potentially facilitate adaptive divergence of duplicated genes, increasing the long-term genome flexibility of autopolyploids and favoring their retention. However, the efficiency of selection and the long-term adaptive potential of autopolyploids remain largely unexplored.

Prevalent formation of autopolyploid lineages under environmental change?

Polyploids generally appear to be particularly common in habitats affected by climatic and edaphic fluctuations (Stebbins, 1971; Favarger, 1984; Brochmann et al., 2004). For example, diploids are typically found in disjunct refugial areas, while polyploids are more frequent in previously glaciated areas (Ehrendorfer, 1980). This pattern has also been confirmed for most autopolyploid complexes investigated to date and suggests that genome doubling per se is tightly associated with environmental change (Table 2).

Stebbins (1984) proposed the ‘secondary contact hypothesis’ to explain the origin of allopolyploids: climate-enforced range shifts promote secondary contact and admixture between formerly allopatric taxa, with the resulting hybrids evolving via chromosome doubling to restore fertility. Brochmann et al. (2004) reviewed polyploid evolution in arctic plants and concluded that differentiation in distinct glacial refugia produced the raw material for allopolyploid formation during postglacial expansion. Stebbins’ secondary contact hypothesis can also explain the origin of autopolyploids in cases where diploid populations of the same species accumulated genome differences in isolated glacial refugia. Once brought into contact by climate-induced range shifts and after having undergone genetic admixture, intraspecific hybrids could be genetically stabilized by chromosome doubling (although still showing polysomic inheritance). Evidence supporting such an origin of autopolyploids is scarce. For example, the intermingled distribution of differentiated lineages of B. laevigata across the Western Alps suggests secondary contact between differentiated populations (Fig. 3; Parisod & Besnard, 2007). The circumpolar genus Draba encompasses 18% of diploid taxa and 82% phylogenetically young polyploid taxa (Jordon-Thaden & Koch, 2008). Based on experimental crosses between populations lacking obvious morphological or ecological differentiation, a multitude of cryptic biological species has been identified within diploid taxa of Draba (Grundt et al., 2006). Chromosomal rearrangements are involved in the reduction of fertility in these intraspecific hybrids (Skrede et al., 2008), opening up the possibility that arctic Draba may have evolved autopolyploid lineages from hybridization between cryptic diploid species in secondary contact zones.

In conclusion, secondary contact might favor the origin of autopolyploid lineages by restoring the fertility of admixed individuals between divergent populations. Such a mechanism could explain why polyploid lineages originate as discrete bursts during periods of intense environmental change (Fawcett et al., 2009). However, whether secondary contact could support the successful establishment of autopolyploid populations through adaptive or neutral processes, probably depends on additional processes, as outlined in the following text.

Autopolyploid origin and genetic diversity

The origin of autopolyploids (Fig. 1) is mostly from the union of unreduced gametes, either through fusion of two unreduced gametes (bilateral polyploidization) or the fusion of reduced and unreduced gametes producing (fertile) triploids that can in turn generate tetraploid progeny through selfing or backcrossing (unilateral polyploidization; Bretagnolle & Thompson, 1995; Ramsey & Schemske, 1998). The production of unreduced gametes is heritable, governed by a few genes, and increases with increasing environmental stress (e.g. frost, wounding, herbivory, water deficit and lack of nutrients), which might further enhance the rate of autopolyploid formation during environmental change. Noticeably, autopolyploid formation allows a large part of the genetic diversity present in diploid progenitors to be incorporated into the autopolyploid gene pool (Husband, 2004). Furthermore, multiple origins of autopolyploid taxa are common (Heuchera micrantha,Soltis & Soltis, 1989; Heuchera grossulariifolia, Segraves et al., 1999; B. laevigata, Parisod & Besnard, 2007) and increase the genetic diversity of autopolyploids, also by adding different maternal lineages (Soltis & Soltis, 1999).

In contrast to synthetic genome doubling experiments, natural autopolyploid lineages potentially include differentially adapted genotypes. To what extent the merging of slightly divergent genomes supports the establishment of autopolyploid populations through adaptive processes is, however, still unknown.

Autopolyploid establishment and ecological divergence

Natural autotetraploids arise within diploid populations and thus face a considerable minority disadvantage (Levin, 1975; Felber, 1991). While restricted seed dispersal, vegetative propagation or self-fertilization might mitigate this minority disadvantage by increasing mating within cytotypes (Baack, 2005), the successful establishment of autopolyploids is thought to require (1) a substantial competitive advantage, (2) ecological divergence and/or (3) stochasticity favoring of autopolyploids in small populations (Petit et al., 1999; Rausch & Morgan, 2005), while the probability of establishment may be enhanced by (4) recurrent polyploidy (Soltis & Soltis, 1999).

Ecological divergence between cytotypes is rather common (Table 2), but autopolyploids do not consistently tolerate harsher environmental conditions and do not always show a broader ecological niche than their diploid progenitors (Levin, 2002). For example, autotetraploids of Santolina pectinata are restricted to disturbed habitats while diploids are not (Rivero-Guerra, 2008). Recent studies have demonstrated that, among other mechanisms, differential interspecific interactions can contribute to ecological divergence among cytotypes and support the establishment of autopolyploid populations (Thompson et al., 2004). For example, established cytotypes of the autopolyploids Chamerion angustifolium (Husband & Sabara, 2004) and H. grossulariifolia (Thompson & Merg, 2008) differ in respective pollinator assemblages, leading to partial reproductive isolation and facilitation of the coexistence of multiple ploidy levels. Cytotypes of H. grossulariifolia also differ with respect to herbivore assemblages (Nuismer & Thompson, 2001). Mathematical models suggest that nascent autopolyploids are more resistant to pathogens under dominance models of resistance alleles, but whether this also is a general advantage for natural and established autopolyploids is unknown (Oswald & Nuismer, 2007).

Only strict sympatric speciation requires immediate niche differentiation among cytotypes or competitive superiority of autopolyploids over diploid progenitors. Autopolyploid lineages may also spread from the periphery of a diploid taxon’s distribution, establishing at sites where progenitors are absent and forming sources for further range expansion (Levin, 2002).

In conclusion, diploids and autopolyploids commonly show ecological differentiation, but neither the mechanisms nor the timing of ecological divergence among cytotypes are easy to identify. In contrast to allopolyploidy, autopolyploidy does not necessarily produce transgressive traits to fuel adaptive ecological divergence. Autopolyploids can maintain more than two alleles per locus, allowing them to produce a variety of allozymes (i.e. enzyme multiplicity owing to heterozygosity), which, in principle, would allow them to achieve high fitness in a variety of environments. However, marked overdominance would be required for allelic diversity to explain ecological divergence and, more generally, the success of autopolyploids.

Polysomic inheritance and inbreeding depression in autopolyploids

Because of tetrasomic inheritance (Fig. 2), autotetraploid populations are characterized by high heterozygosity (i.e. partial and complete heterozygotes) and by nearly doubled effective population size compared with populations showing disomic inheritance (Ronfort et al., 1998; Ronfort, 1999). Polysomic inheritance also implies that autopolyploid populations reach Hardy–Weinberg equilibrium gradually, as opposed to diploid populations, where one generation of random mating is sufficient (Bever & Felber, 1992). In response to inbreeding, heterozygosity accordingly declines at a slower rate in autopolyploid than in diploid populations (Moody et al., 1993). Polysomic inheritance thus has important consequences for microevolutionary processes such as the distribution of genetic variation and the amount of inbreeding depression.

If fitness solely depends on overdominance, autopolyploids should present a higher fitness loss caused by inbreeding depression than their diploid progenitors because genotypic combinations of more than two alleles will rapidly decrease (Husband et al., 2008). This model has, however, received only limited and indirect support from empirical studies (Bingham, 1980; Bever & Felber, 1992; Ramsey & Schemske, 2002). By contrast, under the classical model of recessive deleterious mutations negatively affecting fitness when in a homozygous state, inbreeding depression should be half as high in autopolyploid taxa than in progenitor diploids because of a higher level of heterozygosity (Lande & Schemske, 1985). Taking partial heterozygotes and a range of dominance effects into account, Ronfort (1999) concluded that inbreeding depression might only be slightly lower or even equivalent in autopolyploids compared with diploids. If the degree of dominance was related to the dosage of alleles, inbreeding depression should increase in subsequent generations of selfed offspring. However, most empirical studies indicate an immediate reduction of inbreeding depression by almost 50% in autotetraploid populations compared with populations of related diploid species (Husband & Schemske, 1997; Galloway et al., 2003; Galloway & Etterson, 2007; Husband et al., 2008).

Genome doubling per se can lighten autopolyploids from inbreeding depression compared with diploids and might thus represent an immediate advantage favoring the establishment of autopolyploid populations. Nevertheless, further studies assessing the relationship among allelic diversity, allelic dosage and fitness are necessary to better appreciate inbreeding depression and adaptation in autopolyploids.

Expansion of natural autopolyploid lineages under environmental changes

Range expansion of autopolyploid lineages subjects them to stochasticity and environmental heterogeneity. An understanding of patterns under range expansion is thus crucial to appraise the success of autopolyploids and to understand their putative evolutionary advantages. While Levin (2002) and Soltis et al. (2007) have reviewed the evolutionary ecology of various natural autopolyploid taxa, we will detail selected additional recent studies that focused on the spatio-temporal dynamics of autopolyploid lineages under environmental change.

Studies on the autopolyploid complex Achillea millefolium provided exciting evidence for dramatic adaptive radiation in < 1 million yr (Clausen et al., 1948; Ramsey et al., 2008). The expansion of this autopolyploid lineage across the New World offered many opportunities for populations to colonize new and divergent habitats, forming ecological races with morphological, physiological and life-history adaptations. Interestingly, the proliferation of these distinct races was not linked with strong genetic differentiation at neutral markers, stressing the need to better understand the mechanisms of adaptive genetic differentiation in autopolyploids.

In the B. laevigata complex, diploid populations are genetically depauperate and relictual to glacial refugia, while autotetraploid populations have colonized a large distribution range across the European Alps after the last glacial maximum and exhibit substantial genetic diversity (Manton, 1937; Tremetsberger et al., 2002; C. Parisod, unpublished). Multiple origins of autotetraploid lineages from distinct diploid populations are a likely explanation for the high genetic variation in the recently colonized main distribution area of the complex (Parisod & Besnard, 2007; Fig. 3). Moreover, genetic diversity has probably been further increased during range expansion in places where independently originated autopolyploids came into contact and produced new genotypes through recombination (Soltis & Soltis, 1999; Parisod & Besnard, 2007). In such situations, tetrasomic inheritance in B. laevigata might have maintained especially high levels of heterozygosity in recombinant populations (Soltis & Rieseberg, 1986; Bever & Felber, 1992), thus fostering genomic variation upon which selection could act to produce advantageous genotypes (Soltis & Soltis, 2000). Although not tested yet, admixture between distinct maternal lineages as seen in B. laevigata might also induce new and variable cytoplasmic–nuclear interactions, adding to the diversity of potentially adaptive traits.

Studies on populations of B. laevigata from a single maternal lineage offered additional evidence on the putative advantages of autopolyploidy under climate change. An expanding population at the altitudinal limit of the species’ distribution range demonstrated the ability of autopolyploids to maintain genetic diversity during colonization (Parisod & Bonvin, 2008). This finding fits with the theoretical expectation that autotetraploids may be less susceptible to founder effects than diploids, because tetrasomic inheritance increases effective population sizes and heterozygosity, and thereby reduces genetic drift and inbreeding depression. Polysomic inheritance is thus immediately advantageous under environmental change. Autotetraploid B. laevigata also exhibits large phenotypic variation and wide ecological tolerance, being locally adapted to its environment (Gasser, 1986). Parisod & Besnard (2007) showed that independent autotetraploid lineages are spread along the whole altitudinal gradient in a pattern that is not habitat-specific (Fig. 3), questioning the assumption that a wide ecological niche of autopolyploids results from the spatial juxtaposition of lineages with distinct habitat requirements (Bayer et al., 1996; Soltis & Soltis, 2000). Furthermore, a fine-scale genetic survey within a large continuous population of B. laevigata revealed that different multilocus genotypes were consistently associated with particular ecological conditions, indicating local adaptation at a fine spatial scale (Parisod & Christin, 2008). Therefore, autopolyploidy also allows for intralineage diversification and adaptation to local environmental heterogeneity.

Autopolyploidy as a nearly neutral process

Autopolyploidy may represent a neutral, ratcheting process, as suggested by Meyers & Levin (2006). As is apparent from the literature, the lack of attributes consistently associated with genome doubling per se supports this hypothesis to a certain extent. We suggest that this model should to be taken as a null model in future studies.

Genome multiplication might initiate speciation by promoting post-zygotic reproductive isolation, thereby decreasing maladaptive gene flow from related diploids and facilitating adaptive differentiation via other evolutionary processes (Grant, 1981; Rieseberg & Willis, 2007). Some studies have revealed nearly complete reproductive isolation between established cytotypes, mainly owing to geographical separation and assortative mating (Husband & Sabara, 2004). However, it should be noted that studies in diverse polyploid complexes, including allopolyploids, have reported significant gene flow across ploidy levels through fertile triploids (or pentaploids and heptaploids) and/or unreduced gametes (Brochmann et al., 1992a; Bretagnolle & Thompson, 1995; Ramsey & Schemske, 1998; Petit et al., 1999; Husband, 2004; Slotte et al., 2008). In addition, the recurrent production of interbreeding polyploid lineages contributes to gene exchange between diploid and established polyploids (Brochmann et al., 1992b; Soltis & Soltis, 1993, 1999). Accordingly, genome multiplication per se can be seen as an incomplete speciation mechanism that needs to be complemented by other processes, either adaptive or neutral (Otto & Whitton, 2000).

Autopolyploidy as a mechanism enhancing colonization ability

Recently disturbed landscapes may offer habitats with relaxed competition, where polyploids find opportunities for establishment. ‘Faster and more efficient mobilization’ of polyploid genetic variability in common early-successional communities was invoked early to explain the correlation between polyploid distribution and glaciation (Favarger, 1967; Ehrendorfer, 1980). In extreme or repeatedly glaciated environments, polyploidy may serve as an escape from genetic and ecological niche depauperation caused by drift and selfing at the diploid level (Brochmann, 1992). Accumulating evidence suggests that autopolyploidy may be prevalent under environmental change owing to: increased rate of autopolyploid formation; suitable conditions for establishment because of relaxed competition; polysomic inheritance increasing the likelihood of successful range shifts; the potential of autopolyploids for local adaptation also promoting long-term persistence.

The hypothesis that autopolyploidy is advantageous under environmental change mainly relies on those two traits that are consistently associated with autopolyploidy, namely genic redundancy and polysomic inheritance. In marked contrast to allopolyploids, there is no example of rapid and drastic genome reorganization induced by genome multiplication (Table 1). As neither large genomic structural nor functional changes have been detected after genome doubling per se (although only a few studies are available), we hypothesize that divergence among homologous chromosomes is not strongly selected for, supporting maintenance of polysomic inheritance as an imm-ediate adaptive trait in young autopolyploid lineages; and genome multiplication per se does not generate substantial genetic diversity upon which selection immediately acts to favor the establishment and range expansion of autopolyploids. Although polysomic inheritance might seem disadvantageous as it counteracts efficient selection, it also immediately increases effective population size and should reduce inbreeding depression in the short term, allowing autopolyploids to cope with climate-induced range shifts. Conversely, genic redundancy and polysomic inheritance seem to be only transient evolutionary stages in the autopolyploid’s lifespan. The level of polysomic inheritance is expected to decline in the long-term, restoring disomic inheritance and stabilizing adaptive genetic variation accumulated under genic redundancy. Although we have highlighted the many open questions still remaining to be solved with respect to the evolution of autopolyploid lineages, the recent accumulation of knowledge challenges Stebbins’ statement that ‘chromosome doubling by itself is not a help but a hindrance to the evolutionary success of higher plants’ (Stebbins, 1971: 126).