On both recent and ancient time scales, polyploidy (genome doubling) has been a significant evolutionary force in plants. Here, we examined multiple individuals from reciprocally formed populations of Tragopogon miscellus, an allotetraploid that formed repeatedly within the last 80 years from the diploids T. dubius and T. pratensis. Using cDNA–AFLPs followed by genomic and cDNA cleaved amplified polymorphic sequence (CAPS) analyses, we found differences in the evolution and expression of homeologous loci in T. miscellus. Fragment variation within T. miscellus, possibly attributable to reciprocal formation, comprised 0.6% of the cDNA–AFLP bands. Genomic and cDNA CAPS analyses of 10 candidate genes revealed that only one “transcript-derived fragment” (TDF44) showed differential expression of parental homeologs in T. miscellus; the T. pratensis homeolog was preferentially expressed by most polyploids in both populations. Most of the cDNA–AFLP polymorphisms apparently resulted from loss of parental fragments in the polyploids. Importantly, changes at the genomic level have occurred stochastically among individuals within the independently formed populations. Synthetic F1 hybrids between putative diploid progenitors are additive of their parental genomes, suggesting that polyploidization rather than hybridization induces genomic changes in Tragopogon.
POLYPLOIDY, or genome doubling, has been an important process in many eukaryotic lineages, particularly in flowering plants. Recent genome-level studies have revealed that even the model “diploid” plant Arabidopsis thaliana has undergone several rounds of whole-genome duplication (Vision et al. 2000; Blanc et al. 2003; Bowers et al. 2003; Blanc and Wolfe 2004b). Many polyploid plants are allopolyploids, having arisen through hybridization between genetically distinct entities (usually species) and chromosome doubling. This combination of divergent but compatible genomes in allopolyploids may provide a novel substrate for evolutionary processes (Stebbins 1950; Levin 1983; Grant 2002; Osborn et al. 2003). As all genes in allopolyploids are duplicated, a number of possibilities exist for the evolutionary fate of the homeologs (genes duplicated by polyploidy). Theory predicts three potential outcomes for these duplicated genes: (1) both copies are retained and remain functional, (2) one copy retains the original function while the other copy is lost or silenced, or (3) the two copies diverge such that each copy assumes only a part of the original gene function (subfunctionalization) or one copy acquires a new function (neofunctionalization) (Ohno 1970; Lynch and Conery 2000; Prince and Pickett 2002).
In recent years, studies of genome evolution and gene expression have been in the foreground of polyploidy research (Leitch and Bennett 1997; Soltis and Soltis 1999; Wendel 2000; Liu and Wendel 2003). Much of this work has been conducted on Arabidopsis (Comai et al. 2000; Lee and Chen 2001), Brassica (Lukens et al. 2004, 2006; Pires et al. 2004b), and crop plants, such as wheat (Feldman et al. 1997; Ozkan et al. 2001; Shaked et al. 2001; Kashkush et al. 2002, 2003) and cotton (Zhao et al. 1998; Liu et al. 2001; Adams et al. 2003). Among the significant findings from these studies are that polyploids may experience rapid genomic rearrangements (Song et al. 1995; Ozkan et al. 2001; Lukens et al. 2004, 2006), gene loss (Shaked et al. 2001; Kashkush et al. 2002; Lukens et al. 2006), or gene silencing (particularly due to epigenetic mechanisms, such as DNA methylation) (Comai et al. 2000; Lee and Chen 2001; Lukens et al. 2006).
Despite these important insights, however, little is known about genome evolution and gene expression changes in natural polyploid populations. These data are beginning to emerge for some natural systems, including the Glycine polyploids (Joly et al. 2004) and Spartina hybrids (Salmon et al. 2005). In a study of rDNA evolution and expression in the allotetraploids Glycine tomentella and G. tabacina, Joly et al. (2004) found that most natural allopolyploids possessed only one parental repeat type in their genomes. For polyploids maintaining both copies, preferential expression of one homeolog was detected in most cases. Synthetic polyploids and some artificial diploid hybrids did not exhibit preferential expression of parental rDNA, suggesting that neither polyploidization nor hybridization immediately resulted in nucleolar dominance in Glycine (Joly et al. 2004). In the salt marsh grass Spartina, recent work has shown that methylation differences exist in the independently formed F1 hybrids Spartina × neyrautii and S. × townsendii (the latter of which gave rise to the invasive allopolyploid S. anglica) (Salmon et al. 2005). Although these studies of Glycine and Spartina have yielded important evolutionary insights into natural polyploids, a fundamental question yet to be addressed is whether individuals within polyploid populations show similar or divergent patterns of genome evolution and gene expression. Further, although many natural polyploids have likely formed multiple times (Soltis and Soltis 1993, 1999), little is known about the consequences of recurrent formation on polyploid genome evolution and gene expression.
A unique system for the study of recent and recurrent allopolyploidy in natural populations is provided by the genus Tragopogon (Asteraceae). This system represents a classic example of instantaneous speciation via polyploidy. Following the introduction of three diploid (2n = 12) species (Tragopogon dubius, T. pratensis, and T. porrifolius) from Europe to western North America during the early 1900s, two allotetraploid species (T. mirus and T. miscellus) formed (Ownbey 1950). The ancestries of both allotetraploids were confirmed through flavonoid, isozymic, and DNA studies (Ownbey and McCollum 1953, 1954; Brehm and Ownbey 1965; Roose and Gottlieb 1976; Soltis and Soltis 1989; Soltis et al. 1995; Cook et al. 1998). Molecular data indicate that T. miscellus and T. mirus have formed repeatedly, perhaps as many as 21 and 13 times, respectively, in just the past 80 years (reviewed in Soltis et al. 2004). T. dubius and T. pratensis are the diploid progenitors of T. miscellus, while T. dubius and T. porrifolius are the parents of T. mirus (Figure 1A). T. miscellus is of particular interest because it has formed reciprocally, and the reciprocally formed lineages can be readily distinguished morphologically: when T. pratensis is the maternal parent, T. miscellus has short-ligulate ray flowers, and when T. dubius is the maternal parent, T. miscellus has long-ligulate ray flowers (Ownbey 1950; Ownbey and McCollum 1953). Only one existing population of T. miscellus is long liguled (Pullman, WA); all others (∼40) are short liguled (Novak et al. 1991).
In this study, we examine the evolution and expression of homeologous loci in multiple individuals of T. miscellus from reciprocally formed populations, compared to the diploid progenitors, T. dubius and T. pratensis. Our main objectives were to identify genomic changes and expression differences in T. miscellus relative to its diploid progenitors, to determine the identity of the genes that showed these changes and to assess whether multiple individuals within and between the reciprocally formed populations showed similar patterns of genome evolution and gene expression. To address these issues, we employed a two-step approach. We first used cDNA–amplified fragment length polymorphisms (cDNA–AFLPs) (Bachem et al. 1996) to identify potentially differentially expressed genes. This approach has been used successfully in several polyploid systems, including Arabidopsis (Comai et al. 2000; Lee and Chen 2001), cotton (Adams et al. 2004), and wheat (Kashkush et al. 2002; He et al. 2003), and it has proven to be particularly useful in nonmodel systems that lack developed genomic resources. However, a shortcoming of this method is that fragment differences (i.e., absent fragments) on a cDNA–AFLP gel may result from true expression differences, sequence polymorphisms, or gene or homeolog loss and these sources of variation cannot be differentiated from one another merely by the cDNA–AFLP gel banding patterns (see Wang et al. 2005). Because of this limitation, we also isolated several polymorphic cDNA–AFLP fragments and investigated these candidate genes using both genomic and cDNA cleaved amplified polymorphic sequence (CAPS) analysis (Konieczny and Ausubel 1993). With this follow-up method, changes at the genomic level can be readily distinguished from true expression differences.
MATERIALS AND METHODS
Field-collected seeds were grown in the greenhouse (at Washington State University, Pullman, WA) and allowed to self-fertilize for one generation to reduce individual allelic variation and maternal effects. The plants used in this study were derived from these S1 seed, which were germinated and grown under controlled conditions in a greenhouse at the University of Florida (Gainesville, FL). We analyzed individuals from two populations of T. miscellus of reciprocal origin: the short-liguled form from Moscow, Idaho (Soltis and Soltis collection no. 2604), and the long-liguled form from Pullman, Washington (Soltis and Soltis collection no. 2605). One population of each parental species, T. dubius (Pullman, WA; Soltis and Soltis collection no. 2613) and T. pratensis (Moscow, ID; Soltis and Soltis collection no. 2608), which represent the most likely progenitor genotypes for these T. miscellus populations (Soltis et al. 1995), were also sampled.
F1 hybrids were created by crossing T. pratensis and T. dubius plants grown in the greenhouse. These plants were grown to maturity from seed in the greenhouse at the University of Florida. To induce flowering, mature plants (∼6 months old) were sent from the University of Florida to Washington State University in October 2004 and overwintered in an unheated glasshouse for a period of 3 months. Following this cold treatment, the plants were returned to the University of Florida and were again placed in a greenhouse under standard conditions. After ∼2 weeks, most plants began to bolt, and within 1 month, the first heads were produced. The parental plants crossed were T. pratensis (Moscow, 2608-11) and T. dubius (Pullman, 2613-41) and T. pratensis (Spangle, 2609-28) and T. dubius (Spokane, 2615-22). In both cases, T. pratensis served as the maternal progenitor. These plants were chosen to generate hybrids that would represent the natural T. miscellus genotype from Moscow and an additional independently formed T. miscellus genotype. Heads were bagged with glassine envelopes after emasculation (Fahselt et al. 1976) and after pollination, which occurred 2–3 days following emasculation. Seed from these crosses were germinated following standard protocols. Progeny from the crosses were screened using diagnostic rDNA markers (Kovarík et al. 2005) to determine if individual plants were hybrids or selfed maternal plants.
To investigate the utility of cDNA–AFLPs in Tragopogon, we conducted a preliminary study for which a single individual from each diploid and tetraploid population was included. From this initial screen, several polymorphic fragments were characterized and subsequently analyzed for 10 individuals from each population using both genomic and cDNA–CAPS analyses (see CAPS analyses). An expanded cDNA–AFLP study of these same populations used six individuals from each diploid and tetraploid population (see cDNA–AFLP display and identification of polymorphic fragments), including the same individuals from the preliminary analysis. The individuals examined for comparative expression analyses were germinated at the same time and grown under uniform conditions, and the same tissue type was collected from them concurrently.
cDNA–AFLP display and identification of polymorphic fragments:
Leaf tissue was collected from seedlings 4 weeks after germination and frozen in liquid nitrogen. Total RNA was extracted using Trizol reagent (Invitrogen, San Diego) following the manufacturer's instructions. Messenger RNA isolation, cDNA synthesis, and cDNA–AFLP techniques were performed as previously described (Lee and Chen 2001), except that here we used [33P]dATP to label the EcoRI primer. Thirty-four primer combinations (EcoRI and MseI) were used in the selective amplification reactions for the first screening. The primer combinations used in the preliminary screen and those used in the expanded study are listed in supplemental Table 1 at http://www.genetics.org/supplemental/. The selective amplification reactions were run on polyacrylamide gels, and the resulting cDNA–AFLP bands were scored as monomorphic (present in all individuals) or polymorphic (absent in at least one individual).
To determine the putative identity of selected polymorphic fragments, we excised and sequenced the transcript-derived fragments (TDFs) as previously described (Lee and Chen 2001). Approximately 80 polymorphic bands were cut from the polyacrylamide gels, reamplified using the same set of AFLP primers, and cloned using Promega (Madison, WI) pGEM-T easy vector following manufacturer's instructions. Ten clones from each reaction were sequenced.
The resulting sequences were subjected to BLAST searches against online databases. For most fragments, we used two search strategies. First, the sequences were submitted to a BLAST search against The Arabidopsis Information Resource (TAIR) database (http://www.arabidopsis.org/), which contains annotations for the Arabidopsis genome. Second, the sequences were subjected to a BLAST search against the NCBI EST database (http://www.ncbi.nlm.nih.gov/), which contains ∼110,000 Lactuca and Helianthus expressed sequence tags (ESTs) resulting from the Compositae Genome Project (http://compgenomics.ucdavis.edu/). If putative homologs from Lactuca or Helianthus were returned, then these sequences were submitted to TAIR for another round of annotated searches.
To determine if cDNA–AFLP variation could be detected among individuals from the diploid and polyploid populations, we conducted an expanded cDNA–AFLP analysis, which included six individuals from each diploid and tetraploid population. Again, frozen leaf tissue from young plants was used and RNA extracted as described. Twelve selective primer pairs were used (the most variable as determined from our initial screen) for these individuals, and the fragments were separated on polyacrylamide gels followed by silver staining using standard protocols. The resulting cDNA–AFLP fragments were scored and tabulated as described previously. We further characterized cDNA–AFLP fragment variation among individuals within the diploid and polyploid populations by calculating “within-population fragment variation.” This measure was determined by calculating the percentage of individuals that were polymorphic at each fragment locus averaged over all fragments produced in that population.
To determine if cDNA–AFLP fragment polymorphisms resulted from changes at the genomic or transcriptomic level, we conducted genomic and cDNA CAPS analyses. In CAPS analysis, amplified PCR products are digested with diagnostic restriction enzymes, separated by agarose gel electrophoresis, stained, and visualized. For both genomic and cDNA CAPS analyses, we designed primers for 20 of the 80 sequenced TDFs using the web-based program Primer3 (http://frodo.wi.mit.edu/cgi-bin/primer3/primer3_www.cgi) (Rozen and Skaletsky 1997). These TDFs were selected because they showed the most interesting cDNA–AFLP patterns (e.g., missing fragments or novel fragments in the allopolyploids). In some cases, we used the Lactuca or Helianthus ESTs from our BLAST searches to design primers that would include a larger region of the gene and/or a region that would include introns. Primer sequences are given in supplemental Table 2 at http://www.genetics.org/supplemental/.
Genomic DNA was extracted using a modified CTAB protocol (Doyle and Doyle 1987). For each of the 20 TDFs, genomic fragments were amplified in a 25-μl volume with 50 ng template, 10× buffer, 1.5 mm MgCl2, 0.4 mm dNTPs, 0.2 μm each primer, and 0.5 unit Taq polymerase. Thermal cycling conditions were as follows: 94° for 2 min, followed by 35 cycles of 94° for 30 sec, 52°–54° for 30 sec, 72° for 1 min, and a final 5-min extension at 72°. Products were separated on a 1.5% agarose gel, stained with ethidium bromide, and visualized by UV on a transilluminator. Sephadex G-50 (Fisher Scientific) columns were used to purify the PCR products. Cycle sequencing was performed using Big Dye (ABI Prism) terminator chemistry, and the products were separated on an ABI 377 DNA sequencer at the Institute of Mammalian Genetics, University of Florida. Homologous sequences for T. dubius and T. pratensis were aligned by eye in Sequencher v. 4.1.4 (Gene Codes, Ann Arbor, MI) and inspected for diagnostic restriction sites between the species (using the “cut map” function in Sequencher and selecting all restriction enzymes). Diagnostic restriction sites were identified for 10 of the 20 TDFs. Several additional TDFs contained single nucleotide polymorphisms between T. dubius and T. pratensis, but these did not occur at restriction sites and were therefore not pursued further.
Genomic CAPS analyses:
Genomic DNA was extracted from 10 individuals from each diploid and tetraploid population as described. PCR amplification and digests for all individuals for each gene region were performed twice to determine potential bias in PCR amplification. Likewise, an additional set of primers was designed for each region to verify the results. Mutations that occur at the priming sites (in either the diploids or the polyploids) would result in differing amplification and digestion patterns. Given that the T. miscellus polyploids are <80 years old, we expected that novel mutations in the allotetraploids that might cause mis-priming would be rare. Because T. dubius and T. pratensis have fairly divergent genomes (Mavrodiev et al. 2005), we anticipated that mis-priming in the polyploids resulting from sequence differences between the two parents would be more plausible. In most cases, the first set of primers was designed from only one diploid species (i.e., the individual fragment isolated from the cDNA–AFLP gel), while the second primer set was designed from sequences of both diploid progenitors. PCR and digests were conducted for this second set using the same conditions as the first. Genomic digests were carried out in a 10-μl volume, containing 1× buffer (New England Biolabs, Beverly, MA), 1 μl PCR product, 2–10 units of restriction enzyme (New England Biolabs), and 100 μg/ml bovine serum albumin (when required), and were allowed to incubate at the appropriate temperature for 3 hr. Digested products were separated on either a 2% agarose gel or 3–4% Metaphor (Cambrex) agarose gel, stained with SybrGold (Molecular Probes, Eugene, OR), and visualized on a transilluminator.
To determine if genomic changes might occur as early as the F1 generation, we conducted genomic CAPS analyses for these same genes on 11 F1 hybrid individuals that resulted from the two independent crosses between T. pratensis and T. dubius described earlier.
cDNA CAPS analyses:
For the same set of TDFs for which genomic CAPS analyses were performed, we also conducted cDNA CAPS analyses. Different patterns in the cDNA and genomic CAPS analyses would distinguish expression differences from alterations at the genomic level. Total RNA was extracted from frozen leaf tissue using the RNeasy plant minikit (QIAGEN, Chatsworth, CA) with optional on-column DNase digestion and was quantified using a Nanodrop spectrophotometer (NanoDrop Technologies, Wilmington, DE). First-strand cDNA synthesis was carried out on 200 ng total RNA using Superscript II reverse transcriptase (Invitrogen) and a poly(T) (T17) primer. Using the same primer combinations as the genomic CAPS analyses, RT–PCR was performed using one-twentieth cDNA template from the first-strand synthesis reaction, 10× buffer, 1.5 mm MgCl2, 0.4 mm dNTPs, 0.2 μm each primer, and 0.5 unit Taq polymerase. Thermal cycling conditions were as follows: 94° for 2 min, followed by 30–35 cycles of 94° for 1 min, 52°–54° for 45 sec, 72° for 1 min, and a final 5-min extension at 72°. For each set of reactions, negative (without reverse transcriptase) and positive (glyceraldehyde 3-phosphate dehydrogenase, a housekeeping gene; Harris and Water 1976) controls were included. Amplified RT–PCR products were separated on a 1.5% agarose gel, stained with ethidium bromide, and visualized by UV on a transilluminator. RT–PCR products were digested and analyzed as described for the genomic CAPS analyses.
cDNA–AFLP polymorphism in populations of T. miscellus and identification of putatively differentially expressed genes:
To investigate changes in the evolution and expression of homeologous loci in T. miscellus relative to its parents, we used cDNA–AFLPs as a first step toward identifying candidate genes. From our initial screen using 34 EcoRI/MseI primer pairs on four individuals, 1262 fragments were surveyed, and of these, 556 were monomorphic (44.1%) and 706 were polymorphic (55.9%) among T. miscellus, T. dubius, and T. pratensis. Missing cDNA–AFLP bands in either one or both of the reciprocally formed polyploids accounted for 13.5% (171 bands) of the total fragments, while novel cDNA–AFLP bands in the polyploids composed 4.0% (50 bands) of the total fragments. Shared fragments with a maternal origin in the polyploids were 3.7% (47 fragments) of the total bands produced in the preliminary study, while fragments with a paternal origin composed 4.1% (52 fragments) of the bands.
On the basis of the larger sampling, which included six individuals from each diploid and tetraploid population and for which we used a subset (12) of the original 34 primer pairs, 610 fragments were scored, and of these, 216 were monomorphic (35.4%) and 394 were polymorphic (64.6%) (Table 1, Figure 2). Polymorphisms were detected among individuals within the diploid and polyploid populations. Within-population variation was calculated as the percentage of individuals that lacked a fragment at each fragment locus, averaged over all fragment loci produced by that population. In T. dubius, within-population variation was highest at 7.4%, while T. pratensis was 3.8%. Within the short-liguled (Moscow) population of T. miscellus, fragment variation was 2.5%, while in the long-liguled (Pullman) population of T. miscellus, fragment variation was 5.8% (Table 1). In contrast to the results from the preliminary study, additional sampling revealed that most of the maternal and paternal patterns were not uniform within populations and were therefore not consistent between the reciprocally formed populations. No bands uniformly corresponding to a strict maternal origin were detected in either T. miscellus population, but three fragments (0.6% of fragments produced) in each reciprocally formed population were shared with the respective paternal progenitor. Five novel bands (∼1% of fragments produced) that were not present in either diploid progenitor appeared in individuals from both allotetraploid populations (Figure 2).
Approximately 50% of the cloned fragments from the initial screen showed high sequence identity, in terms of E-value (< 2 × 10−10) and nucleotide equivalence (85–95% identical), to ESTs in the Compositae database (http://compgenomics.ucdavis.edu/). The remaining fragments showed moderate -to-low levels (E-value <1–2) of sequence identity to ESTs from other taxa including Glycine (Fabaceae), Solanum (Solanaceae), and Stevia (Asteraceae). BLAST searches against the TAIR database typically produced matches of higher sequence identities than those against the NCBI database. A subset of the polymorphic fragments identified is shown in Table 2. The TDFs matched several genes of known function that belong to various gene classes, including several transcription factors and kinases. These potentially differentially expressed genes appear to be involved in diverse cellular processes, such as carbohydrate metabolism, signal transduction, protein transport and degradation, and cell division. Interestingly, we found two polymorphic fragments that belong to the same gene class (TDF12 and TDF44; leucine-rich-repeat transmembrane protein kinase), although they corresponded to different loci in the A. thaliana genome. These fragments were identified on different gels using different selective amplification primers. Approximately 15% of the fragments did not match any plant sequence in the available databases and may be too short in length to produce either a match or AFLP artifacts.
Rapid and stochastic changes in homeologous loci in T. miscellus:
The genes analyzed with CAPS analysis, the enzymes used for each gene, and the approximate product sizes for the digested genomic and cDNA amplifications for the diploid progenitors, T. dubius and T. pratensis, are listed in Table 3. The results of the genomic and cDNA CAPS analyses for these 10 genes are shown in Figure 3. For 7 of 10 genes surveyed in this study, at least one allopolyploid individual is missing one parental homeolog or a fragment of this homeolog from its genome (Table 4) . To determine if the missing parental fragments resulted from a mutation at the restriction site, we sequenced the PCR products for the T. miscellus individuals that were not additive of their parental genomic fragments. For all individuals sequenced, there were no nucleotide changes in the polyploid sequences when compared to the parental fragments either at the diagnostic restriction site or along the length of the fragment that was sequenced. Further, the T. miscellus individuals that were missing one parental homeolog did not display any nucleotide polymorphisms in the chromatograms at any of the sites where the parental sequences differ (in contrast to the other “additive” individuals, which showed double peaks at all positions where the parents differ). Only for TDF17.4, TDF46, and TDF85 are both parental copies maintained in all allopolyploid individuals of T. miscellus. For the remaining genes, genomic CAPS analysis revealed much diversity in terms of genomic changes in these loci among individuals from the polyploid populations (Figure 3). For several fragments (TDF7, TDF36.3, TDF44, TDF72.3), a few T. miscellus individuals have apparently lost the T. dubius homeolog, while the T. pratensis copy was eliminated less frequently (TDF44, TDF74). The pattern of genomic change is much more striking for other genes. For TDF90, some individuals from each polyploid population retained both parental copies, and others lost one or the other homeolog (more frequently, the T. dubius homeolog). Across the 10 loci examined, individuals from the short-liguled population of T. miscellus lost parental homeologs more frequently (almost twice as many) than individuals from the long-liguled population (Table 4). Within each population, some individuals retained both parental copies for all genes, while others lost one parental homeolog or the other for as many as three genes (Table 4, e.g., individual 2604-22).
Conflicting digestion patterns in the genomic CAPS analysis (using two different sets of PCR amplification primers) occurred in only one case. For TDF74, when the first primer set (TDF74-F1/TDF74-R1) was used, only the T. pratensis copy was present in each T. miscellus individual from Pullman. However, with the second primer set (TDF74-F2/TDF74-R2), both parental copies of TDF74 were detected in the genomic CAPS analysis for these same individuals, indicating that a mutation exists in these individuals at the priming site of the first primer set. The T. miscellus individuals from Moscow displayed the same pattern for both primer sets.
F1 hybrids are additive of their parental genomes:
Genomic CAPS analyses for the synthetic F1 hybrids show that the individuals resulting from two independent crosses between T. pratensis and T. dubius are additive of their progenitors for the genes analyzed (Figure 4). These results provide a critical framework for the interpretation of the genomic changes observed in the young polyploid populations (i.e., compare Figure 3 to Figure 4).
Genomic change vs. differential expression of homeologous loci:
For 8 of the 10 genes surveyed here, the cDNA CAPS results match those of the genomic CAPS analyses, indicating no qualitative changes in expression for these genes. Expression differences between T. miscellus individuals and the diploid progenitors, as well as among the allopolyploid individuals, were evident for TDF72.3 and TDF44 (Figure 3). For TDF72.3, no expression was detected for the T. dubius individuals analyzed, despite repeated RT–PCR amplifications with different pools of cDNA, nor was TDF72.3 expression detected for one individual of T. miscellus (2604-22) from Moscow, despite the amplification of the fragment in genomic CAPS analysis in all T. dubius individuals and despite the retention of the T. pratensis homeolog in T. miscellus (Figure 3). The other T. miscellus individuals expressed the T. pratensis TDF72.3 homeolog only. For TDF44, the T. miscellus individuals expressed only one parental homeolog or the other. One polyploid individual (2604-22, the same individual that did not express either parental copy for TDF72.3) expressed the T. dubius homeolog, while all other individuals in both allotetraploid populations expressed the T. pratensis copy.
cDNA–AFLP variation within and between the diploid progenitor populations:
cDNA–AFLPs provide a useful tool for detecting potentially differentially expressed genes in polyploid systems, particularly those that lack developed genomic resources. However, because missing fragments on the gel may result from differences at the genomic level (e.g., sequence polymorphism, gene, or homeolog loss), the results must be interpreted cautiously and followed up with additional methods, such as CAPS analysis (see Wang et al. 2005). Despite this potential shortcoming of cDNA–AFLPs, this method provides a relatively fast and inexpensive means to identify candidate genes for further study.
The high level of genetic differentiation detected by cDNA–AFLPs between the diploid progenitor species T. dubius and T. pratensis (38.4% in the preliminary study and 35.2% in the extended study) is consistent with previous Tragopogon studies based on allozymes, RAPDs, and chloroplast DNA restriction site data, which showed wide divergence among the T. dubius, T. pratensis, and T. porrifolius genomes (Roose and Gottlieb 1976; Soltis and Soltis 1989; Soltis et al. 1995; Cook et al. 1998). Recent phylogenetic analyses of Tragopogon using internal and external transcribed spacer sequence data also indicate that T. dubius and T. pratensis are well differentiated, with these two species placed in separate major clades, further attesting to their divergent evolutionary histories (Mavrodiev et al. 2005). The majority of the polymorphic cDNA–AFLP fragments between these two species likely result from differences at the primary sequence level. However, real expression differences must account for at least some of the cDNA–AFLP variation detected, because we were able to verify expression differences between T. pratensis and T. dubius for TDF72.3 in the cDNA CAPS analysis (Figure 3).
Rapid and stochastic genomic changes in T. miscellus:
On the basis of both the cDNA–AFLP and the CAPS data, a number of differences in the evolution and expression of homeologous loci are apparent within and between the reciprocally formed allotetraploid populations. In particular, the genomic CAPS data indicate stochastic small-scale losses of homeologous loci. Previous genomic work on the Tragopogon allopolyploids, using fluorescent in situ hybridization of rDNA and short, tandem centromeric and telomeric repeats, indicated that no large-scale genomic rearrangements or losses had occurred (Pires et al. 2004a). Genome size estimates (Pires et al. 2004a) for the Moscow (4C = 20.30 ± 1.50 pg) and Pullman (4C = 20.99 ± 1.14 pg) populations of T. miscellus, while within the range of the sum of the diploid progenitors T. dubius (4C = 10.83 ± 0.65 pg, 11.76 ± 0.79 pg) and T. pratensis (4C = 12.44 ± 0.87 pg), were slightly less than the estimate for another short-liguled T. miscellus population from Spangle, Washington (4C = 21.76 pg ± 0.87). Genome downsizing is a well-documented phenomenon for other polyploid groups (Leitch and Bennett 2004). For most of the genes surveyed here, the pattern of loss appears to be stochastic, although we currently lack a genomic framework for identifying the relative locations of these loci. Furthermore, the size of the fragments lost is not known. The losses detected could be small regions of the individual genes, entire genes, or short chromosomal fragments.
The parental homeolog that is more often lost in both populations of T. miscellus is the T. dubius copy. The T. pratensis homeolog was lost in only three cases (TDF44, TDF74, and TDF90; Figure 3). This result agrees with previous rDNA studies, which found that the number of T. dubius repeats has been reduced in the genomes of both T. mirus and T. miscellus (Kovarík et al. 2005). Concerted evolution has apparently not acted to completion in these young polyploids, however, as some T. dubius rDNA units are still maintained and expressed. Furthermore, both reciprocally formed T. miscellus populations have lost the T. dubius units, suggesting a lack of a cytoplasmic effect on rDNA parental copy number (Kovarík et al. 2005). Directed silencing of one parental genome has been demonstrated in allopolyploids of A. suecica, which has as its parents A. thaliana and A. arenosa (Wang et al. 2004). In A. suecica, genes from the A. thaliana genome are often silenced by DNA methylation. These genes were reactivated in ddm-1 and met1-RNAi transgenic mutants (Wang et al. 2004). In contrast to this system, however, where the genes of one progenitor are silenced by epigenetic mechanisms, the Tragopogon allopolyploids appear to be eliminating homeologous loci, at least for most of the genes examined here.
Other polyploid systems, most notably synthetic polyploids of Brassica (Song et al. 1995) and allohexaploid wheat (Feldman et al. 1997; Shaked et al. 2001; Kashkush et al. 2002; Levy and Feldman 2004), have also shown rapid genomic changes. Synthetic allopolyploids of Brassica showed considerable changes in parental fragments at the F5 generation (Song et al. 1995) on the basis of RFLPs. Comparing the synthetic Brassica polyploids with the natural Tragopogon polyploids, we find similarities in that frequent and rapid changes in parental fragments have occurred in the allopolyploids formed from distantly related diploid progenitors. However, while T. dubius homeologs are more frequently eliminated than the T. pratensis homeologs, there does not appear to be a significant directional aspect (in terms of the maternal or paternal progenitor) to loss in the reciprocally formed populations of T. miscellus, at least for the genes examined in detail for this study, because individuals from these two populations of separate origin show similar patterns of homeolog loss and expression.
The data for T. miscellus allopolyploids, suggesting rapid loss of genes or gene fragments, seem similar to comparable reports of rapid gene loss in synthetic wheat polyploids (Feldman et al. 1997; Liu et al. 1998; Ozkan et al. 2001; Shaked et al. 2001; Kashkush et al. 2002). In wheat, the same low-copy (most likely noncoding) sequences were eliminated in natural and early generation synthetic allohexaploids, suggesting that genome evolution is reproducible in these wheat allopolyploid species (Feldman et al. 1997; Ozkan et al. 2001; Shaked et al. 2001). Additional studies showed that gene loss could occur as early as the F1 or first amphiploid generation in synthetic wheat tetraploids (Kashkush et al. 2002). The loss of homeologs in Tragopogon has apparently occurred rapidly, as these polyploids were formed <80 years ago. Synthetic F1 hybrids between T. pratensis and T. dubius are additive of their parental fragments for the same loci investigated for the T. miscellus polypoids (Figure 4), suggesting that polyploidization rather than hybridization may be acting as a “genomic shock” (McClintock 1984). In Tragopogon, the loss of homeologous loci seems to be stochastic among individuals within and between polyploid populations (Table 4). Moreover, the homeologs that have been purged from the T. miscellus genomes appear to be functional genes that are involved in essential developmental and physiological processes (Table 2). In A. thaliana, which is known to be an ancient polyploid, certain classes of genes have been retained in duplicate (e.g., genes with functions in transcription and signal transduction), while others (e.g., DNA repair) have been preferentially lost (Blanc and Wolfe 2004a).
The mechanism responsible for the loss of homeologs from the T. miscellus allopolyploid genomes is currently unknown. However, Ownbey (1950), who constructed the first karyotypes for T. mirus and T. miscellus, noted that, while both species primarily form bivalents at metaphase I of meiosis, multivalent formation was quite frequent. Interestingly, in F1 hybrids between T. dubius and T. pratensis, Ownbey (1950) observed an occasional pair of univalents and a ring of four chromosomes. On the basis of Ownbey's accounts, Roose and Gottlieb (1976) and Soltis et al. (1995) proposed that recombination among parental chromosomes was responsible for the nonadditive patterns that they observed in some T. miscellus allozyme profiles. Even with only occasional multivalent formation, sporadic loss of homeologous loci via recombination is plausible. Given the evidence of loss that we detected in the T. miscellus allopolyploids, multivalent formation may be more frequent than previously appreciated (or was at least frequent in early generations) and, at present, is the most likely mechanism responsible for homeolog loss in these polyploids. Determining how early (i.e., at what generation) this loss might occur in synthetic allopolyploids of T. miscellus will be of particular interest.
Genomic changes vs. differential expression in T. miscellus relative to its diploid progenitors:
For most of the genes examined here, genome-level changes appear to be responsible for most of the differences observed in T. miscellus relative to its diploid progenitors on the basis of cDNA–AFLPs. This finding demonstrates both the potential power and the shortcoming of cDNA–AFLPs. For one gene (TDF44), we found differences between the genomic CAPS and cDNA CAPS analyses, which together indicate silencing of one parental homeolog in the allopolyploid individuals. Although most of the “expression” differences between polyploids and their diploid parents result from genomic changes, other genes not studied here may be subject to epigenetic phenomena including DNA, RNA, and chromatin-mediated processes (Chen and Ni 2006).
The Tragopogon allotetraploids occupy an important position in the continuum of polyploid formation because they are natural and established allopolyploids that are still young (<80 years, perhaps closer to 60). Given that these plants appear to be biennials, the time frame involved may be fewer than 30–40 generations. The allotetraploid individuals of T. miscellus show frequent and stochastic elimination of homeologous loci, with recombination of the parental homeologs the most likely mechanism. This loss of genetic material agrees with earlier allozyme studies (Roose and Gottlieb 1976), as well as with more recent genome size estimates for these same populations of T. miscellus (Pires et al. 2004a), which indicate genome downsizing. McClintock (1984) introduced the idea of “genomic shock” in which the genome reorganizes as a response to events such as mutagenesis, transposable elements, or hybridization. On the basis of our data to date, we can rule out hybridization-mediated events as being directly responsible for the differences between polyploids and their diploid parents. Clearly, these successful Tragopogon allopolyploids have been able to contend with the genomic effects of both wide hybridization (between divergent progenitors) and chromosome doubling. In this system, it will be particularly important to determine when genomic changes occur during the evolution of these polyploids and to assess further the genomic consequences of additional polyploidization events that produced multiple lineages of the short-liguled T. miscellus.
We thank C. Cody and B. Pratt for greenhouse care and seed collection from greenhouse-grown plants; O. Hassan, H.-S. Lee, and L. Tian for technical assistance; and V. Symonds, A. Kovarík, and three anonymous reviewers for their comments on this manuscript. This research was funded by National Science Foundation (NSF) grant MCB0346437 to D.S., P.S., J.T., and Z.J.C. and by the University of Florida Research Foundation (D.S. and P.S.). Work in the Chen lab was also supported by the Texas Agricultural Experiment Station and a grant (0077774) from the NSF Plant Genome Research Program.
- Received March 1, 2006.
- Accepted April 20, 2006.
- Copyright © 2006 by the Genetics Society of America