Answers are grounded in this paper and cite their sources.
Meta-analysis examines polyploidy classifications in plants, highlighting discrepancies in taxonomy and genetics.
Polyploidy has long been recognized as a key process in plant evolution and it is the subject of a plethora of reviews (for just two recent examples, see Wendel (2015) and Soltis et al. (2016)). Estimates of the prevalence of polyploidy can vary dramatically – for example, Ramsey & Schemske (1998) gave a range of 47 to 70%. The question of what percentage of plants are polyploid has been resolved, since it is now known that all seed plants are fundamentally polyploid (Jiao et al., 2011), and most lineages have experienced multiple cycles of whole-genome duplication and diploidization (Soltis et al., 2015). More relevant is the percentage of plant speciation events that involve polyploidy; a recent meta-analysis put that figure at c. 15% for flowering plants and double that in ferns (Wood et al., 2009). ‘Definitions of autopolyploidy and allopolyploidy either emphasize mode of origin (hereafter, MO) or cytological criteria (hereafter, CC) as primary criteria. MO autopolyploids arise within single populations or between ecotypes of a single species, whereas allopolyploids are derived from interspecific hybrids... . CC allopolyploids are expected to display bivalent pairing, lack of allosyndesis and disomic inheritance, while CC autopolyploids will exhibit multivalent configurations, nonpreferential pairing at metaphase, and multisomic inheritance... .’ Doyle & Egan (2010) also distinguished two classes of definitions, ‘genetic’ (based on polysomic vs disomic inheritance) vs ‘taxonomic’ (origin from one species vs two species) and considered genetic criteria more useful and informative. Parisod et al. (2010) included both sets of criteria in their definitions: ‘Autopolyploids are traditionally considered to arise within a single species by doubling of structurally similar, homologous genomes (AAAA), whereas allopolyploids arise via interspecific hybridization and subsequent doubling of nonhomologous (i.e. homoeologous) genomes (AABB)’. However, they considered inheritance type ‘diagnostic’. The same cannot be said for taxonomic definitions. Taxonomy, coupled with chromosome counts, is a much simpler criterion to apply than is genetic data – for any given genus, there exist species names, and although chromosome counts are missing for many of them, a ‘vast’ cytological database exists (Wood et al., 2009), relative to the paucity of natural polyploids for which segregation/inheritance data are available. Lacking the diagnostic trait of inheritance pattern, efforts to characterize and catalogue autopolyploids rely on taxonomy – either formal classifications or interpretations of taxonomic treatments that attempt to mitigate such issues as the undercounting of autopolyploids that result from taxonomic definitions (Soltis et al., 2007). Is this a problem? In a recent Letter published in New Phytologist, Barker et al. (2016) report results of a meta-analysis on the relative abundance of autopolyploids vs allopolyploids in a carefully chosen set of 47 vascular plant genera. They gleaned information on these genera from databases such as the Index to Plant Chromosome Numbers (IPCN), and consulted nearly 300 references to arrive at numbers of autopolyploids and allopolyploids in each genus, including not only named species, but also unnamed taxa they interpreted as warranting recognition, thus attempting to address the problem of under-counting of autopolyploids emphasized by Soltis et al. (2007). Among these 47 genera was the legume, Glycine, which has been the focus of our work for many years, with an even longer history of genetic and taxonomic studies, and whose perennial subgenus (subg. Glycine) includes an extensive polyploid complex (Fig. 1; reviewed in Sherman-Broyles et al., 2014). We therefore did our own tabulation for Glycine, which differed from that of Barker et al. (Table 1; details given in Supporting Information Table S1 and Notes S1). We outline the problems that we feel led to these discrepancies, use these to explore the issue of the use of taxonomic criteria for defining classes of polyploids, and make some general suggestions about the interpretation of meta-analyses of polyploid frequency and kind. The IPCN (http://www.tropicos.org/Project/IPCN) is a wonderful resource, but it is hardly shocking that it is far from infallible. Entries for Glycine include long-defunct names for Glycine species, a species with diploid and (auto)polyploid cytotypes that was transferred to the genus Neonotonia in the 1970s, and the mistaken report of a cytotype unknown in the genus that, if real, would revolutionize our understanding of Glycine genome evolution (Table S1; Notes S1). Being unfamiliar with Glycine, Barker et al. understandably accepted these at face value, accounting for some of the discrepancies in Table 1. A general lesson is that such databases are under-curated, resulting in errors that may be incorporated into meta-analyses. Polyploid complexes – whether autopolyploid or allopolyploid – are, almost by definition, taxonomically difficult. Even though Glycine, unlike most plant genera, has a strong genetic foundation for assessing polyploidy (Sherman-Broyles et al., 2014), genetic information has not yet been fully incorporated in classification and nomenclature. Though progress is being made in naming distinct genetic lineages (see Table S1 and Notes S1), several unambiguous biological species in the complex – both diploids and polyploids – still lack formal names (Fig. 1), as do other distinct genetic lineages in the perennial subgenus (Sherman-Broyles et al., 2014). Specifically, patterns of morphological variation within the polyploid complex are obscured by the reticulate pattern of shared diploid genomes. It is possible to name taxa solely on the basis of molecular variation (Cook et al., 2010) or to assume that diploids and polyploids are biological species (Soltis et al., 2007; Barker et al.), but we are working to provide characters useful for field identification before naming species. The taxonomy of Glycine is thus evolving, and the current nomenclature is a snapshot that captures only imperfectly our understanding of genetic lineages. A simplified example from Glycine subgenus Glycine illustrates the problems this creates for meta-analyses. The 1980s taxonomy treated all of the taxa of the Glycine tomentella polyploid complex (Fig. 1) as G. tomentella. Table 2 compares that treatment with the current taxonomy, and also with the expected future taxonomy of the complex when the major genetically distinct, reproductively isolated lineages are described. Under a strictly taxonomic definition, the proportion of autopolyploids shifts from 100% to a current 67% and a future value of 0% (Fig. 2b–d). Barker et al. did not rely entirely on these strictly taxonomic definitions, but their Glycine counts were strongly influenced by taxonomy (Table S1; Notes S1). This is no doubt true of other genera, and it makes it very difficult for anyone relying on taxonomy to produce an accurate estimate of the number of polyploid species. In perennial Glycine, there are three cases where a species has diploid and polyploid cytotypes: two allopolyploids (G. tomentella, G. tabacina) and the putative autopolyploid (G. hirticaulis). In all three cases, the name belongs to the polyploid, not the diploid (Table S1; Notes S1) but Barker et al. assumed that the name belonged to the diploid in each case. This may be reasonable as an evolutionary expectation, because polyploids at least initially will be a minority cytotype (Levin, 1975). However, taxonomy is a product of where, when, and by whom a species happens to be collected and named first, and may have little connection with evolutionary pattern or process. Indeed, the greater colonizing ability of polyploids (e.g. Pandit et al., 2011, 2014; te Beest et al., 2012) may make it more likely that taxonomists will encounter and name polyploid cytotypes first. All subgenus Glycine species have Australian ranges, but only five taxa – all of them allopolyploids – have populations on islands of the Pacific Ocean (Fig. 1; Sherman-Broyles et al., 2014). Four of these were named from those colonial populations, three from Taiwan and nearby islands. Colonizing populations of an invasive species may be more distinctive morphologically than those in the source range. In Glycine many of the allopolyploids and their progenitors are sympatric or parapatric in Australia, forming a near-continuum of morphological variation. In Taiwan, no diploids and only a subset of the polyploid species are found, limiting the range of morphological variation, which is decreased still further because each polyploid is the product of a genetic bottleneck produced by colonization. Thus it is likely that the Taiwanese populations of G. tabacina and G. tomentella were easier to recognize (and thus name) than would be the case in Australia. Although Barker et al. ‘do not estimate or account for multiple origins of a polyploid species’ they also did not describe their criteria for distinguishing multiple origins from multiple unnamed species. Since independent genetic lineages are often considered species, this decision is necessarily subjective, representing another instance where the use of taxonomic criteria will inevitably lead to disagreements in counting or classifying polyploids. In Glycine, several of the unnamed allopolyploid species remain lumped together as G. tomentella s.l. (Fig. 1), and it appears that Barker et al. assumed that these represent ‘multiple origins’ of G. tomentella; they cite a paper (Doyle et al., 2002) with that phrase in its title. But taxa that share only one of their two homoeologous genomes are clearly different biological species. Thus G. tomentella s.l. T1, T2 (= G. dolichocarpa), T3, T4 (= G. tomentella s.s.), T5, and T6 are all separate species, and not simply different origins of G. tomentella s.s. (Fig. 1). Doyle et al. (2002) limited the term ‘multiple origins’ in Glycine to the incorporation of different genotypes of the same diploid genome into a polyploid species (Fig. 1). Taxonomy is an imperfect surrogate for categorizing the products of phenomena that are fundamentally genetic. Barker et al. ‘... divided polyploids into autopolyploids and allopolyploids based on reports by researchers in our survey’, and because the taxonomic definition is the prevalent one in the literature we assume that it dominated the categorizations used in many of their 47 genera. The age-old problem of different species concepts and recognition criteria employed by different taxonomists for different genera adds to the uncertainty. De Queiroz (2007) discussed how taxa fulfill different species recognition criteria on the way from a single species to two daughter species, and how the fact that different taxa achieve these landmarks in different orders leads to a ‘gray zone’ in which there is inevitable disagreement about whether there is one species or two (Fig. 2e). Because the taxonomic (mode of origin) criterion for distinguishing autopolyploids from allopolyploids depends on whether one vs two diploid progenitors are involved, the existence of this gray zone means that for the same polyploid taxon, one worker will consider it an allopolyploid because two diploid species are involved, but another, lumping these two diploid species into one species under a different criterion, will consider it an autopolyploid. This creates an analogous gray zone of ambiguity (Fig. 2f–h) in a meta-analysis such as that of Barker et al., which ‘divided polyploids into autopolyploids and allopolyploids based on reports by researchers’ whose species recognition criteria inevitably differed from one another. The distinction between autopolyploidy and allopolyploidy is genetically meaningful. The most critical feature of polyploids is their ability to retain genome-wide genetic variability at levels not possible for typical diploids, and what is important about the distinction between autopolyploidy and allopolyploids is how they accomplish this (e.g. Stebbins, 1947; Grant, 1971; Parisod et al., 2010; Washburn & Birchler, 2014; Lloyd & Bomblies, 2016; Soltis et al., 2016). As captured by Parisod et al.'s (2010) definition, a ‘textbook’ genetic allopolyploid is the product of genome merger and doubling in which two homoeologous genomes coexist, making the plant a fixed hybrid. Ramsey & Schemske (2002) noted numerous violations of fixed hybridity in allopolyploids, particularly in the earliest stages of polyploid evolution (neopolyploids), but because they used a mode of origin (taxonomic) definition in their review it is perhaps not surprising that some taxa they classified as allopolyploids do not meet genetic criteria for allopolyploidy. Perhaps there is a class of taxa that meets these genetic criteria from their moment of origin, or perhaps the neopolyploid stage is transient and genetic allopolyploidy is quickly stabilized. In any case, the allopolyploid genome evolves over time through gene loss (Soltis et al., 2015), but with full disomic segregation it can preserve much of the variation from both progenitor genomes for millions of years. A genetically defined autopolyploid can also have higher levels of genetic variation than its diploid progenitor genotype(s) (reviewed by Parisod et al., 2010), most simply by doubling the number of chromosomes present in a population (Meirmans & Van Tienderen, 2013). Because autopolyploid genomes consist of four karyotypically interchangeable chromosomes, allelic variation can be lost by segregation, unlike the situation with homoeologous loci in an allopolyploid. Loss is slowed, however, due to polysomic rather than disomic segregation (Grant, 1971; Lloyd & Bomblies, 2016) and loss of diversity by genetic drift is also slowed (Moody et al., 1993). Recessive alleles are also less exposed to selection than is the case in diploids or in allopolyploids with disomic segregation. Still, the autopolyploid genetic system only retards the loss of diversity, whereas a fixed hybrid allopolyploid potentially can preserve homoeologous diversity indefinitely. How well this binary genetic classification agrees with the binary taxonomic classification is unknown (Fig. 3a,b), mostly because the segregation data that provide the most compelling genetic criterion (Parisod et al., 2010) are lacking for most species. But the situation is further complicated because the genetic classification is not absolute, perhaps particularly in neopolyploids (Ramsey & Schemske, 2002). Moreover it has long been known that autopolyploids and allopolyploids represent endpoints in a genetic continuum (e.g. Stebbins, 1947) in which intermediates, termed segmental allopolyploids, show a mixture of disomic and polysomic segregation patterns (‘mixosomic’: a useful term coined in Soltis et al., 2016), depending on the amount of interchange among different duplicated regions of their constituent genomes. Thus, even if genetic and taxonomic polyploidy are strongly correlated, taxonomic polyploids of either type could show a wide range of segregation patterns (Fig. 3c). Although the existence of stable natural segmental allopolyploids has been questioned (Sybenga, 1996), there is mounting evidence for plants with mixosomic inheritance (e.g. Jannoo et al., 2004; Klie et al., 2014; Leal-Bertioli et al., 2015), and increasing interest in modeling the consequences of complex meiotic behavior on patterns of genetic diversity (e.g. Stift et al., 2008; Arnold et al., 2012; Suda & Herben, 2013; Dufresne et al., 2014). However, as noted by Barker et al., ‘most literature reports do not distinguish these gradations’. The age of a polyploid also adds to uncertainty in classifying polyploids. It is thought that taxa that begin their existence with polysomic chromosome behavior evolve to disomic segregation patterns over time (e.g. Otto & Whitton, 2000), though evolution in the opposite direction is theoretically possible (Sybenga, 1996). Species can also make the transition from homologous to homoeologous chromosome behavior abruptly, by mutations at a single locus, though this has only been characterized in Triticum and Brassica (Grandont et al., 2013). If taxonomic uncertainty is coupled with a full genetic spectrum of segregational variation (Fig. 3d) the meaningfulness of meta-analyses that recognize two classes of polyploid under a variety of genetic and taxonomic definitions seems questionable. The answer for Glycine is ‘yes’: Barker et al.'s major conclusion is that there is ‘near parity’ in the numbers of allopolyploid and autopolyploid species across the 47 genera, and Glycine was a case in point, with three autopolyploids and three allopolyploids in their estimate. By contrast, there is no ‘parity’ in our count, with an 8 : 1 preponderance of allopolyploids over autopolyploids (Table 1). They also concluded that many polyploids – particularly autopolyploids – are disproportionately unnamed; we note that as many diploids as polyploids remain unnamed in Glycine, and the only autopolyploid in the complex is named. Barker et al. used their meta-analysis primarily as a springboard for speculating about interesting aspects of polyploid evolution, and we do not dispute their conclusions. But our point here is that, although Glycine may or may not be representative in its proportion of named or unnamed, autopolyploid or allopolyploid species, the genus is probably very representative when it comes to the difficulty of categorizing and tabulating two kinds of polyploids across diverse studies. Clearly, the actual numbers in a meta-analysis of polyploidy should be interpreted as an estimate within a very large confidence interval. For the question of autopolyploidy vs allopolyploidy, we feel strongly that much more information is needed on the genetic behavior of polyploids across a wide diversity of genera to understand the shape of the polysomic–mixosomic–disomic continuum and how it relates to taxonomy-based categorizations (Fig. 3). Although genetic definitions of polyploidy are certainly not black and white, they are inherently meaningful: they deal with genetic diversity, which is a key, but underappreciated and understudied aspect of the ongoing global biodiversity crisis (Graudal et al., 2014). The required research program is daunting, because genetic criteria for classifying polyploids require considerable work to assess (e.g. Doyle et al., 1986; Arnold et al., 2015) and have their own limitations. But it is essential for achieving a richer understanding of polyploid evolution that subjective taxonomic definitions be replaced by evolutionarily meaningful genetic criteria. Our understanding of the nomenclature of the perennial subgenus of Glycine benefitted from conversations and collaboration with the late James Reveal. The authors thank Kirsten Bomblies, Jonathan Wendel, and Walter De Jong for comments on drafts of all or part of this Letter, and anonymous reviewers, particularly Doug Soltis, for incisive criticisms of the submitted manuscript. Jane Doyle, Jeremy Coate and Adrian Powell have provided comments and years of collaboration and discussion shaping our understanding of evolution in perennial Glycine. The authors acknowledge many years of grant support for research on Glycine from the US National Science Foundation (currently grant no. 1257522). J.J.D. conceived the analysis. J.J.D. and S.S-B. conducted the tabulations. J.J.D. and S.S-B. wrote the manuscript. Please note: Wiley Blackwell are not responsible for the content or functionality of any Supporting Information supplied by the authors. Any queries (other than missing material) should be directed to the New Phytologist Central Office. Table S1 Comparison of Barker et al. (2016) classification with our current understanding of Glycine taxonomy and cytology Notes S1 Detailed discussion of Glycine cytology and taxonomy. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.
No takes yet. Share an insight, caveat, or question.
Doyle et al. (2016) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: