PloiDB: the plant ploidy database

Polyploidy, namely the acquisition of additional, complete sets of chromosomes to the genome, is widely recognized as a key feature of extant organismal diversity, particularly in plants. It is generally accepted that all angiosperm species have experienced at least one polyploidization event in their evolutionary past (Jiao et al., 2011). Therefore, most (if not all) plant species should be considered as paleo-polyploids that have since diploidized to some extent. As such, the distinction between diploids and polyploids should be made with respect to a reference timepoint. In recent decades, polyploid research has experienced a resurgence among plant evolutionary biologists. This is largely due to the use of genomic analyses that have revealed a rich history of genome duplications across multiple plant lineages. Indeed, numerous studies have investigated the impact of polyploidy on morphological and life-history traits, ecology, diversification patterns, and genome evolution (reviewed in Soltis & Soltis, 2000; Otto & Whitton, 2003; Ramsey & Schemske, 2003; Otto, 2007; Ramsey & Ramsey, 2014; Wendel, 2015; Soltis et al., 2016; Van de Peer et al., 2017; Fox et al., 2020). Notably, many of these studies focus on the effect of polyploidy in the context of a specific taxonomic group, which limits our ability to draw conclusions regarding the universal consequences of polyploidy, and to distinguish broad convergent trends from species-specific idiosyncrasies. There is thus a growing need to expand the examination of ploidy estimates across the seed plants clade, to obtain robust and broad information about the effect of polyploidy.

In the last few years, multiple methods for ploidy inferences based on sequenced genomic data have been developed (e.g. Jiao et al., 2011; Rabier et al., 2014; Vanneste et al., 2014; Tiley et al., 2018; Zwaenepoel & Van de Peer, 2020). However, due to the computational complexities and substantial amount of genomic data involved, the applications of such methods are still somewhat limited and are usually applied at phylogenetic scales above the species level, for example, by sampling representatives from several clades of interest. As such, the most comprehensive sequence-based analysis to date, which was conducted by the 1KP initiative and encompassed the transcriptomes of roughly 1100 plant species, has identified 244 whole-genome duplication (WGD) events occurring within Viridiplantae (One Thousand Plant Transcriptomes Initiative, 2019; Li & Barker, 2020).

Ploidy estimation at the species level is still largely based on information derived from chromosome numbers. A simple utility of chromosome number information for determining ploidy employs threshold techniques to classify polyploid species relative to the lowest chromosome number (or some other measure) found in the genus (Stebbins Jr, 1938; Grant, 1963; Goldblatt, 1980; Wood et al., 2009). Other studies have incorporated phylogenetic information and inferred ploidy values based on the maximum parsimony principle (Guggisberg et al., 2006; Ohi-Toma et al., 2006; Timme et al., 2007). A more advanced approach utilizes likelihood models of chromosome-number evolution that accounts for the branch lengths of the phylogeny and allows for different types of transitions in chromosome numbers, whose rates are estimated from the data. Such models have been implemented in the chromEvol probabilistic framework and its extensions (Mayrose et al., 2010; Glick & Mayrose, 2014; Freyman & Höhna, 2018; Zenil-Ferguson et al., 2018; Blackmon et al., 2019). ChromEvol treats the evolution of chromosome numbers along a phylogeny as a continuous time Markov process and includes parameters that represent the transition rate for different types of events. Three types of transitions correspond to polyploidization events: (1) WGD, an exact duplication of the number of chromosomes; (2) demi-polyploidization, a 1.5-fold multiplication of the chromosome number, representing, for example, a triplication event; and (3) base-number transition, the addition to the genome of any multiplication of an inferred base number, which represents the monoploid chromosome number of the focal group (Glick & Mayrose, 2014). In addition to ploidy transitions, chromEvol also considers dysploidization events, which may result in either the addition or the subtraction of a single chromosome number. Such events are the result of rearrangements within chromosomal DNA, and are triggered by double-strand breaks and subsequent misrepair at the breakpoints. A decrease in chromosome number (descending dysploidy) is caused by chromosome fusion, resulting from recombination between at least two non-homologous chromosomes. By contrast, chromosome fission leads to an increase in the chromosome number (ascending dysploidy; Mayrose & Lysak, 2021). Given a phylogeny and the respective chromosome numbers at the tips, a standard application of chromEvol allows the selection of the most appropriate model and provides estimates of the expected number of polyploidy and dysploidy transitions along each branch of the phylogeny, thereby allowing categorization of tip taxa as either diploid or polyploid relative to other taxa in the clade (Glick & Mayrose, 2014).

Under the chromEvol phylogenetic framework, inferences are made with respect to a reference timepoint, represented as the root of the phylogeny examined. Consequently, a lineage that experienced a polyploidization event following divergence from the most recent common ancestor (MRCA), but has since diploidized, is still classified as a polyploid. Thus, the choice of a reference timepoint has a large impact on the inferred ploidy values. To date, the chromEvol framework has been applied to phylogenies of different taxonomic scales, ranging from genus (e.g. Mayrose et al., 2011; Rice et al., 2019), to family (e.g. Chacón & Renner, 2014; Mota et al., 2016; Román-Palacios et al., 2020; Romero-da-Cruz et al., 2022), or even higher (e.g. Clark et al., 2016; Escudero et al., 2018; Carta et al., 2020), but the consequences of the chosen time scale remain unclear.

Information on chromosome numbers across a wide array of plant clades is growing continuously (Rice et al., 2015). In parallel, in the past few years, phylogenetic relationships across the seed plants have been more accurately and extensively investigated (Liu et al., 2022). We present the plant ploidy database (PloiDB) as an online source that is available online at http://ploidb.tau.ac.il/. PloiDB contains inferences across multiple phylogenetic scales, according to either divergence time or taxonomic resolution. This further allows us to provide a more continuous scale of ploidy inferences, rather than a strictly dichotomous outcome.

We present PloiDB, as a community resource of ploidy inferences for tens of thousands of plant taxa – including mostly species and occasionally lower taxonomic ranks (e.g. subspecies or varieties). All inferences were generated using the chromEvol framework, with inputs derived from extensive chromosome count information (Rice et al., 2015) and a broad seed-plant phylogeny (ALLMB; Smith & Brown, 2018). The ALLMB phylogeny consists of 356 305 tips, corresponding to plant taxa at the taxonomic rank of species or below. In our baseline inference scheme, we partitioned the broad ALLMB phylogeny to focal clades according to their generic circumscription. This resulted in 2063 genus-level phylogenies, each with at least five taxa having chromosome-number information, ranging across 210 families belonging to the angiosperms and 8 to the gymnosperms, that together encompass 57 493 taxa. Considering the number of accepted plant taxa in World Flora Online (Borsch et al., 2020), the PloiDB inferences cover 14.4% of angiosperm taxa (56 126 out of 389 376) and 26.9% of gymnosperms (442 out of 1643).

Our results reveal that 31.9% of angiosperm and 5.7% of gymnosperm taxa have experienced one or more polyploidization events since their divergence from the MRCA of their respective genus. Polyploid frequency is similar in monocots and eudicots (32.9% and 31.7%, respectively), but changes dramatically within each group. Polyploidy is common in higher monocots (commelinids, represented by Arecales, Commelinales, Poales, and Zingiberales; 49.3%) but relatively rare in basal monocots (non-commelinid monocots; 21.5%). In eudicots, on the other hand, polyploids are similarly frequent in rosids and asterids (30.5% and 30.8%, respectively) but are more common in basal dicots (37.4%). The distribution of polyploid frequency across genera is left-skewed (Fig. 1a), with 708 genera containing only diploids, while polyploid-rich genera (containing over 50% polyploids) are relatively rare and are only found in 359 angiosperm and zero gymnosperm genera. In addition, our inferences indicate that polyploid frequency is < 20% in roughly half of seed-plant families, while eight families are extremely polyploid rich (i.e. > 80% of the taxa are polyploids). Within the 20 largest angiosperm families, Poaceae, Rosaceae, and Malvaceae are the most polyploid rich (55.1%, 51.3%, and 44%, respectively), while Myrtaceae and Apocynaceae are polyploid poor, containing < 10% polyploids (see Supporting Information Table S1 for polyploid frequency across all families, sorted from the largest to the smallest families).

None declared.

IM and KH conceived the study. KH developed the ploidy inference scheme, assembled the PloiDB, and conducted the data analyses. AS provided the chromEvol implementation and consultation during the development of the ploidy inference scheme. IM supervised the study.