On the road to diploidization? Homoeolog loss in independently formed populations of the allopolyploid Tragopogon miscellus (Asteraceae)

Background Polyploidy (whole-genome duplication) is an important speciation mechanism, particularly in plants. Gene loss, silencing, and the formation of novel gene complexes are some of the consequences that the new polyploid genome may experience. Despite the recurrent nature of polyploidy, little is known about the genomic outcome of independent polyploidization events. Here, we analyze the fate of genes duplicated by polyploidy (homoeologs) in multiple individuals from ten natural populations of Tragopogon miscellus (Asteraceae), all of which formed independently from T. dubius and T. pratensis less than 80 years ago. Results Of the 13 loci analyzed in 84 T. miscellus individuals, 11 showed loss of at least one parental homoeolog in the young allopolyploids. Two loci were retained in duplicate for all polyploid individuals included in this study. Nearly half (48%) of the individuals examined lost a homoeolog of at least one locus, with several individuals showing loss at more than one locus. Patterns of loss were stochastic among individuals from the independently formed populations, except that the T. dubius copy was lost twice as often as T. pratensis. Conclusion This study represents the most extensive survey of the fate of genes duplicated by allopolyploidy in individuals from natural populations. Our results indicate that the road to genome downsizing and ultimate genetic diploidization may occur quickly through homoeolog loss, but with some genes consistently maintained as duplicates. Other genes consistently show evidence of homoeolog loss, suggesting repetitive aspects to polyploid genome evolution.


Background
Allopolyploidy combines the processes of hybridization with genome doubling, and together, these provide a potential avenue for instantaneous speciation [1][2][3].
Following allopolyploidization, several evolutionary outcomes are possible for the genes duplicated by polyploidy (homoeologs). Both copies may be retained in the polyploid and remain functional, one copy may accumulate mutations and either diverge in function or become silenced, or one copy may be physically lost [8,33,34]. The fate of these duplicated gene pairs seems to vary depending on the system under investigation and the loci involved [35][36][37][38][39][40][41]. Over longer evolutionary timeframes, gene loss, genome downsizing, and, ultimately, genetic 'diploidization' appear to be common phenomena [8,[42][43][44][45]. Homoeologous recombination appears to play an important role in the loss of small genomic fragments during the early stages of polyploid formation [46][47][48][49][50], which contributes to gene loss and genome downsizing in allopolyploids [39,43,51,52]. Wolfe (2001) pointed out that within a species, some loci may remain 'tetraploid', while others are diploidized; evidence from wholegenome analyses supports this idea [e.g., [36,40]]. Although polyploidy is clearly a recurrent process on both recent and ancient timescales, we know very little about the evolutionary fate of genes duplicated by polyploidy in independently formed allopolyploid populations. Specifically, are homoeologs consistently retained or lost in a repeated manner among individuals from independently formed polyploid populations?
The allopolyploids Tragopogon mirus and T. miscellus (Asteraceae) are textbook examples of speciation following polyploidy and provide an ideal system to investigate the evolutionary fate of duplicated genes in independently formed populations. These allopolyploids formed recently in the Palouse region of the western United States (eastern Washington and adjacent Idaho) following the introduction of three diploid species (T. dubius, T. porrifolius, and T. pratensis) from Europe in the early 1900s [53]. Tragopogon mirus formed independently several times from T. dubius and T. porrifolius, while T. miscellus formed multiple times from T. dubius and T. pratensis [53][54][55][56][57]. Only T. miscellus has formed reciprocally in nature, and these reciprocally formed individuals can be distinguished morphologically. The 'short-liguled' form has T. pratensis as the maternal progenitor, while the 'longliguled' form has T. dubius as the maternal parent ( Figure  1). Today, only one long-liguled population exists (in Tragopogon populations sampled

Genomic CAPS analyses
Two hundred individuals (83 T. dubius, 33 T. pratensis, and 84 T. miscellus) from 10 populations (Table 1) were screened for 13 markers Table 2). Variation in the restriction digestion patterns of Tragopogon dubius was evident for a single marker (TDF72.3) (Figure 2). No variation was observed in T. pratensis based on the present sampling. Two individuals, each grown from a seed collected from a T. dubius plant in the field (one each from Troy and Albion), apparently were hybrids, as the individuals possessed both T. pratensis and T. dubius fragment patterns in the genomic restriction digests for all markers screened (data not shown). Because T. pratensis does not occur in either locality, these individuals likely represent hybrids between T. miscellus and T. dubius.
Combining the new data generated here with data from Tate et al. (2006), for the 13 loci examined, 11 showed loss of a homoeolog in at least one of the T. miscellus  ? A B C lost a homoeolog at more than one locus, the same parental homoeolog was lost more often than different homoeologs (i.e., 11 individuals lost homoeologs from the same parent, while four individuals lost alternative homoeologs). Of those that lost the same parental homoeolog, nine cases were losses of T. dubius, while two were losses of T. pratensis. Considering all populations, regardless of the parental origin, the loss of one homoeolog was the most common scenario (25 cases), followed by homoeolog losses at two loci in eight individuals, three loci in six individuals, and six losses in one plant (individual 2625-3 from Albion, mentioned above). For these multiple losses, no clear pattern emerged (i.e., when two or more loci were lost from multiple individuals, they were not the same pairs of loci).
At the population level, differences in the number of losses were also evident, but without a clear genomic, genic, or geographical pattern (See Additional file 1). The Albion and Moscow populations showed the greatest number of total homoeolog losses (12 losses in three and seven individuals, respectively), followed by Oakesdale

Homoeolog loss in independently formed populations
Our extended survey of 13 loci for ten populations of Tragopogon miscellus indicates that some genes are maintained in duplicate in all populations, while others show loss among some individuals from the independently formed populations. This result is consistent with our previous finding of loss in two populations (Moscow and Pullman) for ten of these same genes [59]. Although homoeolog loss is not unique to Tragopogon, the present study represents the largest survey of individuals from natural populations conducted thus far. Homoeolog loss appears to be a common phenomenon in polyploids and may occur rapidly following their formation. For example, synthetic polyploids of wheat [47] and Brassica [46,48,61] show loss of homoeologous loci in early generations. In Tragopogon, we have not detected loss in F 1 hybrids or first-generation synthetic polyploids [59,60].
Thus, homoeologous loss does not appear to occur instantaneously upon hybridization or polyploidization in Tragopogon, at least based on the loci examined thus far.
Given that genome downsizing and other processes may ultimately contribute to genetic 'diploidization' in polyploid organisms [8,43], what impact does homoeolog loss have on recently and independently formed polyploid populations? Our data indicate that homoeolog loss in Tragopogon miscellus is stochastic among individuals from polyploid populations that are less than 80 years old (<40 generations as these are biennials The loss or retention of certain classes of genes appears to be a recurrent pattern when ancient whole-genome duplication patterns are examined, although the classes that are retained in duplicate differ depending on the lineage under study [35,36,41,68]. For example, in Asteraceae, Barker et al. (2008) found that genes associated with structural components and cellular organization were retained in duplicate, while genes involved with regulatory (e.g., transcription factors) and developmental functions lack duplicates. In Arabidopsis (Brassicaceae) and rice (Poaceae), however, genes involved with transcription were retained in duplicate [36]. In Tragopogon miscellus, the two genes that were retained in duplicate (TDF46 and TDF85) in all individuals did not fall into the category of significantly enriched (or reduced) when compared to the Barker et al. (2008) study. Similarly, the genes that were lost did not match gene ontology (GO) slim categories that were significantly either underrepresented or enriched. TDF46 is a putative protein phosphatase 2C family protein that functions in the plasma membrane, and TDF85 is a putative β-fructosidase that acts in the vacuole. As additional genomic resources are developed for Tragopogon and these genes are analyzed in the polyploid species, it will be imperative to determine whether certain gene classes are consistently lost or retained following allopolyploidization.

Mechanism for homoeolog loss
Studies of Brassica [46,50,61] allopolyploids have revealed a significant role for homoeologous recombination in DNA loss, although this process does not appear to affect wheat allopolyploids [38]. A recent karyological study using fluorescent and genomic in situ hybridization (FISH, GISH) of natural and synthetic Tragopogon allopolyploids identified extensive chromosomal changes, including monosomy and trisomy, intergenomic translocations, and variation in nrDNA loci [69]. Importantly, the same study [69] showed that some chromosomal changes occurred in the first synthetic generation of T. mirus (synthetics of T. miscellus have not yet been investigated). Ownbey [53] observed multivalent formation in individuals of T. mirus and T. miscellus from natural populations and also noted univalents and a ring of four chromosomes in F 1 hybrids between T. dubius and T. pratensis.
We have also observed frequent multivalent formation in synthetic lineages of T. mirus and T. miscellus [62]. These prevalent meiotic irregularities suggest a mechanism for the homoeolog losses observed here. That is, through homoeologous recombination in early generations following polyploid formation, genome reshuffling and gene loss could act to stabilize the new polyploid genome [63]. Perhaps in Tragopogon a combination of factors acts over time to stabilize the new polyploid genomes. For example, some chromosomal changes could happen immediately following polyploid formation, with homoeolog loss acting gradually over successive generations. The study of additional genes and comparisons with synthetic T. miscellus lineages [62] over several generations will be important for establishing the overall pattern of genome change in this system.

Conclusion
Our survey of 13 homoeologous loci in individuals from ten populations of Tragopogon miscellus represents the most extensive survey of the fate of duplicate genes in allopolyploid genomes from independently formed natural populations. In this species, loss of a parental homoeolog has occurred for several loci in individuals from these populations. Some loci are consistently maintained as duplicates in all individuals from these populations. Other genes consistently show evidence of homoeolog loss across populations of independent origin; significantly, the T. dubius homoeolog is typically lost. Hence, some aspects of genome evolution appear to have been repeated in these new polyploids. In these young (~40 generations) allopolyploids, genomic incompatibilities may be resolved, in part, through loss of a parental homoeolog for some loci. As polyploidy and genome downsizing are recurrent processes in many lineages, other polyploid groups should be investigated to determine if similar patterns emerge for the loss and retention of genes duplicated by polyploidy. In total, we included ten populations of T. miscellus, four populations of T. pratensis, and nine populations of T. dubius (Table 1). Our sampling strategy was intended to survey as many individuals and populations from the Palouse as possible. Similarly, we recognize that sympatric diploid populations may not represent the progenitor genotypes for a particular local polyploid population (although they typically do; Symonds et al. unpublished). Therefore, we wanted to survey as many diploid individuals as possible to screen for potential variation in the loci examined. The number of populations and number of individuals from the diploid populations included in the study differed because of changes in population dynamics since the formation of the Tragopogon polyploids [70]. For example, while once locally common, T. pratensis has become sparse in the Palouse over the last several decades [58] and is not always found in the vicinity of T. miscellus populations (Table 1). Nevertheless, data accumulated from previous studies [23,57] indicate that very little genetic variation exists within and between populations of T. pratensis in the Palouse. On the other hand, T. dubius is more widely distributed [58] and harbors more genetic variation than does T. pratensis [57]. Of the two diploid parents of T. miscellus, T. dubius is more likely to exhibit variation in the genes examined.
[59]. The Moscow and Pullman populations of T. miscellus were the subject of a previous study [59]; those data are combined here with data for three new loci (cry1, TDF27.10, and nrDNA).
To verify that the observed homoeolog losses based on CAPS analysis were not the result of a point mutation at the diagnostic restriction site in T. miscellus post-polyploid formation, PCR products were sequenced for all individuals of Tragopogon miscellus that showed loss of a homoeologous fragment. For a given individual, a homoeolog loss was scored only when the sequence data verified the pattern from the CAPS gel analysis (i.e., no sequence polymorphisms were detected in the chromatogram either at the diagnostic restriction site or at other positions where T. dubius and T. pratensis differ). These same criteria applied for nrDNA loci. However, when the intensity of the digested parental fragments differed in the CAPS gel, the nrDNA patterns were scored as P > D or D > P to reflect differing copy numbers in the allopolyploid individuals [65,66]. For all loci, when a loss was determined, we assumed that both alleles of a parental homoeolog were lost. In cases where one allele of a homoeolog was lost, CAPS analysis might not detect these losses. Furthermore, identical patterns of loss in individuals from the same population may be the result of shared ancestry. Therefore, total losses from a single population were tabulated as both minimum and maximum number of losses.