Concerted Evolution of rDNA in Recently Formed Tragopogon Allotetraploids Is Typically Associated With an Inverse Correlation Between Gene Copy Number and Expression
Roman Matyášek, Jennifer A. Tate, Yoong K. Lim, Hana Šrubařová, Jin Koh, Andrew R. Leitch, Douglas E. Soltis, Pamela S. Soltis, Aleš Kovařík


We analyzed nuclear ribosomal DNA (rDNA) transcription and chromatin condensation in individuals from several populations of Tragopogon mirus and T. miscellus, allotetraploids that have formed repeatedly within only the last 80 years from T. dubius and T. porrifolius and T. dubius and T. pratensis, respectively. We identified populations with no (2), partial (2), and complete (4) nucleolar dominance. It is probable that epigenetic regulation following allopolyploidization varies between populations, with a tendency toward nucleolar dominance by one parental homeologue. Dominant rDNA loci are largely decondensed at interphase while silent loci formed condensed heterochromatic regions excluded from nucleoli. Those populations where nucleolar dominance is fixed are epigenetically more stable than those with partial or incomplete dominance. Previous studies indicated that concerted evolution has partially homogenized thousands of parental rDNA units typically reducing the copy numbers of those derived from the T. dubius diploid parent. Paradoxically, despite their low copy number, repeats of T. dubius origin dominate rDNA transcription in most populations studied, i.e., rDNA units that are genetic losers (copy numbers) are epigenetic winners (high expression).

EPIGENETIC silencing of parental genes is an important feature of newly established allopolyploids and is believed to reconcile regulatory incompatibilities between parental subgenomes (Adams and Wendel 2005; Birchler et al., 2005). Nucleolus organizer regions (NORs) contain tandemly arranged highly reiterated ribosomal rRNA genes coding for 18S-5.8S-26S rRNA whose expression is under epigenetic control (Pikaard 2000; for review see Neves et al. 2005). Indeed nucleolar dominance (ND) was one of the first examples of differential gene expression discovered in plant hybrids nearly a century ago (Navashin 1934). Numerous studies in both plants and animals support the view that ND/rRNA gene silencing may be stable or unstable, partial, or reversible (Volkov et al. 2007). In Arabidopsis and Brassica biochemical (Chen and Pikaard 1997a) and genetic (Lawrence et al. 2004; Probst et al. 2004) studies provided evidence that rRNA silencing depends on epigenetic factors, including DNA methylation and histone acetylation (for review see Preuss and Pikaard 2007). Other factors may also play a role in ND, e.g., the position of rRNA genes in the nucleus (Shaw and Jordan 1995), the relative amount of heterochromatin (Viegas et al. 2002), and the activity of unlinked genes (Lewis et al. 2004). Despite considerable progress in our understanding of ND, little is known about the relationship between ND and sequence homogenization, particularly in young populations of newly formed polyploidy species. The goal of this article is to study this relationship and the inheritance patterns of ND and sequence homogenization at individual and population levels in recently formed allotetraploid species.

Besides epigenetic silencing homogenization of repeats seems to be another interesting aspect of rDNA biology. The consequence of homogenization is little sequence variability among the units within and across the arrays in genome. In many allopolyploids' rDNA one particular unit overwrites preexisting units, probably mediated by recombination or copy number expansion/contraction mechanism that leads to the fast evolution of novel families, a process called concerted evolution (for review see Alvarez and Wendel, 2003; Volkov et al. 2007). However, the frequency/occurrence of homogenization does vary between species. Allopolyploids of Arabidopsis and Brassica apparently have no rDNA unit loss or concerted evolution, despite an evolutionary history lasting several thousands of generations (Bennett and Smith 1991; O'kane et al. 1996). In contrast studies in some recently formed allopolyploids (Kovarik et al. 2005), synthetic interspecific hybrids and polyploids (Lin et al. 1985; Cluster et al. 1996; Weiss and Maluszynska, 2000; Skalická et al. 2003; Pontes et al. 2004; Lim et al. 2006) reveal evidence of astonishingly fast genetic change influencing thousands of rDNA units, perhaps in just a few generations following allopolyploidy. The reasons for interspecies differences remain enigmatic, but evolutionary change probably occurs through a range of processes including locus loss/gain, amplification/reduction in repeat copy number, gene conversion, and recombination.

In nature there are several examples of allopolyploids of recent (<150 years) origin, including in the genera Senecio (Abbott and Lowe 2004), Spartina (Ainouche et al. 2004), and Tragopogon (Soltis et al. 2004), enabling direct comparison of expression patterns between parents and polyploids. However, in none of these systems has the question of parental rRNA genes expression been addressed so far. Tragopogon mirus Ownbey and T. miscellus Ownbey are allotetraploids (2n = 24) that formed during the past 80 years following the introduction of three diploids [T. dubius, T. pratensis, T. porrifolius (2n = 12)] from Europe to western North America (Ownbey 1950; Soltis et al. 2004). T. mirus and T. miscellus originated by hybridization between T. dubius and T. porrifolius, and T. dubius and T. pratensis, respectively. Both allotetraploids have formed repeatedly—there may be as many as 21 lineages of separate origin of T. miscellus and 11 of T. mirus in the Palouse region of Washington and Idaho (Soltis and Soltis 1989; Cook et al. 1998). We previously reported little or no deviation from additivity for chromosome numbers, satellite sequences, and genome size (Pires et al. 2004). However, a quantitative study of the inheritance of rRNA genes among several populations of T. mirus and T. miscellus revealed up to 70–95% reduction in the number of T. dubius repeats, likely by deletion or recombination events. The genetic lability of the T. dubius genome component in the allopolyploids was recently demonstrated by Tate et al. (2006), who found stochastic gene loss at several low-copy loci in T. miscellus individuals from two populations.

Here, we addressed the following questions. Is there any association between: (i) epigenetic silencing and sequence elimination, (ii) rDNA copy number and gene expression, and (iii) rDNA chromatin condensation and rDNA copy number? Do populations of T. mirus and T. miscellus have the same patterns of rDNA evolution, and do individuals within a population/species become fixed for epigenetic or genetic characteristics after at most 80 years (40 generations) of evolution? To address these questions we have analyzed rDNA copy numbers, rRNA gene expression, and rDNA chromatin condensation at interphase and metaphase in several populations each of T. mirus and T. miscellus.


Plant material:

Seeds of the three diploid and two allotetraploid species of Tragopogon were collected from natural populations in Idaho and Washington (Table 1). For each population sampled, we collected seeds from 10 to 20 individual plants; the seeds from each plant were kept separate. These seeds were subsequently planted in the greenhouses at the University of Florida; plants were grown to maturity and allowed to self-pollinate. The resultant selfed seeds (S1) were collected from individual plants (Kovarik et al. 2005). Some of these seeds, representing one generation of selfing (S1) in the greenhouse, were later germinated in a greenhouse in the Institute of Biophysics, Brno for use in this investigation. Collected leaf samples were dipped into RNAlater solution (Ambion, Austin, TX), stored at −20°, and analyzed within 12 months after harvest. Root tips for cytogenetic studies and leaf tissue for DNA isolation were harvested from 5- to 8-week-old plants. Plants were sampled in duplicate. In some cases, plants were sampled during both the first and second year of the life span of these biennials. All individuals were genotyped using PCR-CAPS analysis of the PP2C gene (Tate et al. 2006) and their diploid parentage was confirmed.

View this table:

Summary of collections of Tragopogon used

Reverse transcription-cleaved amplified polymorphic sequence (RT-CAPS) assay:

This assay was used to determine the relative abundance and parental origin of pre-rRNA transcripts. Total cellular RNA from fresh leaves was isolated using TRIzol Reagent (Invitrogen, Carlsbad, CA) as recommended by the supplier. RNA quantity was measured using a spectrophotometer, and its quality was checked by agarose-formaldehyde gel electrophoresis. Total RNA was treated with RNase-free DNase I RQ1 (Promega, Madison, WI) or DNA free kit (Ambion) and checked for the presence of genomic DNA using PCR.

To study the transcription of parental rDNA, we analyzed the ITS1 and, in some plants, the ITS2 regions of the rRNA precursors, using two-step RT-PCR. Reverse transcription reactions (20 μl) typically contained 1 μg of total RNA, 2 pmol of either 5.8Srev primer (for the ITS1 region) or 26Srev primer (for the ITS2 region) (for sequences see Kovarik et al. 2005), 10 nmol of each dNTP, and 20 units of reverse transcriptase (Superscript II RNAse H; Invitrogen) using the conditions recommended by the supplier. After reverse transcription, the RNA was digested using RNase A (Sigma, St. Louis).

For PCR (10 μl) amplification we used 0.5 μl of RT mixture as template, 4 pmol of each primer, 2.4 nmol of each dNTP, and 0.4 units of DyNAzyme II DNA polymerase (Finnzymes, Epsoo, Finland). Cycling conditions were as follows: initial denaturation step (94°, 180 sec) and 15–35 cycles (94°, 20 sec; 57°, 30 sec; 72°, 30 sec) with 18S-FOR and 5.8S-REV primers for ITS1 or 35 cycles (94°, 20 sec; 60°, 30 sec; 72°, 30 sec) with 5.8-FOR and 26S-REV primers for ITS2, followed by a final 72° extension for 10 min. To discriminate among ITS variants originating from the diploid progenitors, the PCR products were digested with restriction enzymes that produce locus-specific restriction patterns. Mapping of locus-specific sites was carried out on sequences in the GenBank/NCBI database [AY645813, AY508169, AY508167 (Mavrodiev et al. 2004)] using Wisconsin GCG package software. The ITS1 PCR product was digested with BstNI (Figure 1) or MspI (New England Biolabs, Beverly, MA). The ITS2 region was analyzed by double-digestion with Fnu4HI and Tsp509I, and the resulting fragments were separated on a 7% polyacrylamide gel.

Quantification of expression phenotypes and statistical procedures:

After RT-CAPS analysis, the gel was stained with ethidium bromide and the resulting DNA bands in the gel were visualized using ultraviolet light transillumination (CCD camera was a Discovery model; Ultra-Lum, Claremont, CA). Images were processed by UltraQuant molecular imaging and analysis software (Ultra-Lum) fluorescent signals quantified by a rectangle integration method. In some PCR experiments radioactive incorporation of α-[32P]dCTP (ICN, Irvine, CA) precursor was used to monitor linearity of amplification. Both radioactive and fluorescence quantification methods were in a good agreement. The fluorescence integer of both T. dubius-specific bands was divided by the intensity of all bands and expressed as a percentage of the total expression.

Restriction site analysis and Southern blot hybridization:

Genomic DNA was isolated from fresh young leaves according to the modified cetylammonium bromide (CTAB) method of Saghai-Maroof et al. (1984). For Southern blot hybridization, purified DNAs were digested with restriction enzymes, separated by gel electrophoresis, and blotted onto nylon membranes (GE Healthcare). The Southern blot hybridization was carried out in a modified Church–Gilbert buffer (Lim et al. 2000) using the 26S rDNA labeled with α-[32P]dCTP (ICN). The probe was an ∼220-bp PCR product derived from the 3′ end of the 26S rRNA tobacco gene (Lim et al. 2000). The bands were visualized by PhosphorImaging (GE Healthcare).

Genomic-cleaved amplified polymorphic sequence (genomic-CAPS) assay:

This assay was used to determine the relative abundance and parental origin of rDNA sequences. The protocol followed the RT-CAPS assay except that 0.5 ng of genomic DNA was used as template in the PCR step rather than the products of reverse transcription.

Ribonuclease protection assay:

A riboprobe was prepared from a DNA template covering a polymorphic ITS2 region downstream from the 5.8S gene. Briefly, the 138-bp fragment was amplified using ITS2FORdub 5′-GCA TCGCGTCGCCCCCCCCCAT and ITS2T7REV 5′-TAATACGACTCACTATAGGGAGAACAAACACAACCACC ACTAGCCGTGCGTCAA primers and a cloned ITS2 from T. dubius. The ITS2T7REV primer contained a binding site for T7 polymerase at its 5′ end. The gel-purified DNA template was transcribed with T7 polymerase using a nonradioactive biotinylated precursor (MAXIscript; Ambion). The antisense ∼168-nt band was purified by electrophoresis in a polyacrylamide gel, eluted, and used in downstream procedures.

Homeologous primary rRNA transcripts were identified by hybridization of RNAs with a riboprobe derived from the T. dubius unit using a RPA III kit (Ambion). Briefly, 5–10 μg of total RNA was mixed with ∼1 ng of biotinylated RNA probe, lyophilized, resuspended in hybridization buffer, and denatured for 15 min at 95° in a hybridization oven. The probe was hybridized to RNA samples at 68° overnight followed by incubation at 42° for 3 hr. Resulting RNA:RNA hybrid was treated with RNase A/RNase T1 mix for 30 min at 37° using a 1:100 enzyme dilution. After enzyme inactivation the samples were precipitated, vacuum dried, resuspended in gel loading buffer, denatured for 5 min at 94°, placed on ice, and electrophoresed in 7% (44:1, 48% urea) denaturing polyacrylamide gel at 22V/cm. After electrophoresis, the urea was washed out and RNA was blotted onto nylon membrane (Hybond XL; GE Healthcare) using semi-dry transfer (TRANS-BLOT SD; BIO-RAD, USA). The protected probe fragments were visualized by chemiluminiscence (BrightStar BioDetect Nonisotopic Detection Kit; Ambion) using a luminescent image analyzer (LAS-3000; FUJIFILM, Japan).

Single-strand conformation polymorphism (SSCP):

The species-specific polymorphisms in rDNA units were studied by SSCP (Sambrook and Russel 2001). Briefly, ITS1 sequences generated by PCR were digested with Fnu4HI, isopropanol precipitated, resuspended in 30% dimethyl sulfoxide, heat denatured (10 min, 99°), immediately cooled on ice, and loaded onto a pre-electrophoresed and pre-cooled (5°) 10% polyacrylamide gel. After electrophoresis (Tris-borate buffer, pH 8.0, 5°, 8 V/cm), the gels were silver stained (Sambrook and Russel 2001).

Chromosome preparation and fluorescent in situ hybridization (FISH):

Chromosome preparations and FISH were performed according to Leitch et al. (2001) and Pires et al. (2004). The rDNA probe pTa71 (Gerlach and Bedbrook 1979) was labeled with biotin-16-dUTP (Sigma-Aldrich Company, Poole, Dorset, UK), and hybridization was detected with avidin-Cy3 (GE Healthcare). The 5S probe was an ∼120-bp coding region of the Nicotiana tabacum gene (Fulnecek et al. 1998). Images were taken with a Leica CTRMIC epifluorescent microscope with an Orca ER camera, and images were processed using Openlab software (Improvision). Only those functions that apply equally to all pixels in the image were used. For every set of observations, 5 to 10 metaphases were examined from each of one to three slides.


rRNA gene expression in diploid and tetraploid Tragopogon:

To examine expression of progenitor rDNA homeologues, we isolated RNA and DNA from leaf tissue of 10–30 individuals from each population analyzed of the two allotetraploids. The parental origin and relative abundance of the rRNA genes themselves were previously studied by Southern blot hybridization (Kovarik et al. 2005). Here, we analyzed parental genes by genomic-CAPS assay allowing a direct comparison of copy number and expression profiles. To determine relative abundance of rRNA primary transcripts (pre-rRNA), we used the RT-CAPS assay, taking advantage of a BstNI restriction polymorphism in the ITS1 region (Figure 1). There is a single BstNI site in the ITS1 of T. dubius, but not in the units of T. porrifolius and T. pratensis. Therefore, RT-CAPS and genomic-CAPS in T. porrifolius and T. pratensis generate 700-bp bands, while in T. dubius there are two bands at 500 bp and 200 bp (Figures 2 and 3). Consequently, in T. mirus (Figure 2) and T. miscellus (Figure 3) RT-CAPS and genomic-CAPS were expected to generate bands at 700, 500, and 200 bp. However, in many populations this was not always the case (see below). RT-CAPS data were compared with the following controls: (i) no amplified products resulted when RNA was used alone, indicating an absence of contamination with DNA (not shown), (ii) DNA mixing experiments showed that the primers do not discriminate between the parental unit types (Kovarik et al. 2005), and (iii) selected RNA samples were analyzed by nuclease protection assay.

Figure 1.—

Restriction enzyme maps of the most abundant T. dubius, T. pratensis, and T. porrifolius rDNA units. The position of a conserved BstNI site in the T. dubius units is indicated. Large arrows indicate positions of primers. Distances are approximately to scale.

Figure 2.—

Expression analysis of parental rRNA genes in natural populations of T. mirus. (A) Example of a representative polyacrylamide gel with genomic-CAPS (lanes “DNA”) and RT-CAPS (lanes “RNA”) assays. Primary RNA transcripts were analyzed in the ITS1 region by RT-PCR using BstNI restriction site polymorphisms (Figure 1). The digested PCR products were separated on a 7% polyacrylamide gel and stained with ethidium bromide. The sizes of the diagnostic bands were as follows: ∼200 bp (T. dubius), ∼500 bp (T. dubius), ∼700 bp (T. porrifolius). (B) Quantification of rDNA expression and parental genes ratios. Signals in individual locus-specific bands were counted by fluorescence scanning. Data were scored for several individuals per population (numbers are given in Table 1) and expressed as a percentage of T. dubius genes and transcripts out of total. Each sample was analyzed twice and values averaged. The expression patterns in population 2602 differed significantly from those in other four populations (paired t-test, P < 0.01). Differences between populations 2601, 2603, 2673, and 2690 were statistically insignificant (P > 0.1).

Figure 3.—

Expression analysis of parental rRNA genes in natural populations of T. miscellus. (A) Example of a representative polyacrylamide gel with genomic-CAPS (lanes “DNA”) and RT-CAPS (lanes “RNA”) assays. The methods were same as in Figure 2. (B) Quantification of rDNA expression and parental gene ratios among the populations. The expression patterns among populations differed significantly. Statistical analysis (paired t-test): 2604 vs. 2605, 2606, and 2605 vs. 2606, P < 0.01).

Expression patterns in T. mirus:

At the DNA level, the 700-bp bands inherited from T. porrifolius were stronger (Figure 2, lanes “DNA”) than those inherited from T. dubius (500 and 200 bp), which is consistent with reduced copy numbers of the T. dubius-origin units in most Tragopogon allotetraploids. The quantitative proportions of parental genes determined by genomic-CAPS were in a good agreement with those obtained by the Southern blot analysis (Kovarik et al. 2005). In lanes loaded with RT-CAPS products (lanes “RNA”) from individuals of T. mirus (populations 2601 and 2603), the 700-bp band was absent or weak, suggesting strong silencing of T. porrifolius-origin rDNA expression (Figure 2). Quantitative measurements using densitometry scanning of ethidium bromide gels, or incorporated radioactivity, indicated that ∼95% of transcripts in T. mirus are derived from T. dubius-origin units. This result holds true even in cases where the total number of genes of T. dubius origin is extremely small (individual 33 from population 2603). Similar results were obtained by reverse transcription of the ITS2 subregion (not shown). In contrast to populations 2601 and 2603, T. mirus population 2602 had a more balanced number of genes inherited from each parent and this was usually reflected in transcript abundance (Figure 2). However, two individuals (1, 49) of this population showed only trace amounts of T. dubius-origin rRNA, suggesting reciprocal silencing as compared to individuals of population 2601 and 2603. The mean rRNA expression pattern characteristic for each population is summarized in Figure 2B. The meiotic stability of nucleolar dominance was studied in the progeny of 2601 and 2602 individuals. While all 2601 plants showed stable silencing there was up to fivefold variation in the expression of parental rRNA genes among the 2602 individuals (Figure S1 at

The nuclease protection assay was further carried out to confirm the RT-PCR results. The assay is highly specific and quantitative due to the RNase sensitivity of mismatched base pairs and the use of solution-phase hybridization driven toward completion by excess of probe. The sequence alignment revealed several mismatches and indels between the T. dubius and T. porrifolius ITS2 (Figure 4) forming potential targets for enzyme digestion of RNA–RNA heteroduplexes. First, we carried out a control experiment by mixing of T. dubius and T. porrifolius RNA at different ratios (left) and hybridizing these with a 168-bp probe derived from T. dubius ITS2. It is evident the probe remained fully protected to digestion when mixed with fully homologous T. dubius RNA only. However, increased amounts of T. porrifolius RNA in a hybridization reaction resulted in an enhanced digestion of the probe forming distinct low-molecular-weight bands on the gel. Thus, the parental ITS2 could be distinguished based on the differential sensitivity of RNA–RNA homo- and heteroduplexes. Next we selected RNA samples to map all three expression phenotypes in allotetraploids: the T. dubius expression dominance (2601-7), a codominance (2602-7) and T. porrifolius dominance (2602-1). The T. dubius probe was fully protected from digestion with a nuclease when hybridized to RNA from an individual with a typical T. dubius nucleolar dominance (2601-7). In contrast, the probe was partially or completely digested when hybridized with RNA from individuals with codominant and T. porrifolius phenotypes. The expression phenotypes in allotetraploids were thus confirmed by both RT-CAPS and nuclease protection methods.

Figure 4.—

rDNA transcription in T. mirus determined by RNAse protection assay. (A) Experimental strategy and probe design. Mismatched nucleotides between T. dubius (accession no. AM 493990) and T. porrifolius (AM 493995) sequences are indicated. The probe was designed according the T. dubius sequence. (B) Polyacrylamide gel with probe digestion products. The specificity of a probe was confirmed by hybridization with variable amounts of RNA from T. dubius and T. porrifolius (left). There was a progressive digestion of probe with increasing concentration of T. porrifolius RNA due to mismatched nucleotides in an RNA–RNA heteroduplex. Right: RNAse protection analysis of representative samples of T. mirus RNA showing nucleolar dominance of T. porrifolius (2602-1) and T. dubius (2601-7) unit types and a codominance (2602-7).

Expression patterns in T. miscellus:

We analyzed expression patterns in three populations of T. miscellus (Figure 3). Genomic-CAPS revealed genes of both T. dubius and T. pratensis origin in all populations. RT-CAPS of individuals of populations 2606 and 2604 revealed prominent bands of T. dubius-origin transcripts. In population 2605, the ratio of transcripts of T. dubius and T. pratensis origin was more balanced. The mean rRNA expression pattern characteristic for each population is summarized in Figure 3B.

SSCP analysis:

Compared to RT-CAPS, SSCP exhibits higher sensitivity, and alteration of the nucleotide sequence of the molecule by as little as a single base can reshape the secondary structure, with consequent changes in electrophoretic mobilities through native gels. We carried out SSCP on selected samples (Figure 5) from diploid and tetraploid species. In parental diploids, four or five species-specific bands (conformers) could be distinguished. At the DNA level, all four conformers of T. dubius origin (Du1-4) were significantly reduced in three allotetraploid individuals (T. mirus 2601 and both T. miscellus 2604 and 2606) while they clearly dominated the cDNA (“RNA”) profiles. No parental transcripts of T. porrifolius and T. pratensis origin could be detected in 2601 and 2606 individuals, respectively. The sample from population 2602 was again exceptional in having relatively equal proportion of bands inherited from both parents and no apparent differences between cDNA and DNA profiles. In general, the SSCP analysis was in good agreement with the RT-CAPS data. Interestingly, some conformers were more pronounced at the DNA level while others were more prominent at the RNA level. For example, the D4 conformer of T. dubius origin was barely expressed in T. miscellus 2604, whereas it was highly expressed in other accessions. The Pr5 conformer was transcriptionally enhanced in T. miscellus 2604. One interpretation is that only a subset of genes characterized by distinct ITS1 conformers is expressed within the array.

Figure 5.—

SSCP analysis of ITS1 subregion in genomic DNA and cDNA. The high-resolution gels resolved the profiles in the progenitor species and respective allotetraploids. Diagnostic conformers: D, T. dubius origin; Po, T. porrifolius origin; Pr, T. pratensis origin. U is a unique conformer specific to T. mirus population 2602. The gels show analysis of individuals from populations of T. mirus (pops. 2601-0-10 and 2602-7) and T. miscellus (2604-4 and 2606-15).

Structural polymorphism of rDNA among tetraploid populations:

To determine if there were polymorphisms in IGS, we digested genomic DNA with restriction enzymes and hybridized the blots with rDNA probes. Several enzymes, including combined BstYI/SspI (Figure 6) digestion, revealed IGS polymorphisms among diploid species. We analyzed individuals from three different populations of diploids collected in the Palouse region and found little or no variability. In the allotetraploid species, most parental BstYI/SspI bands were inherited. There were, however, several distinct deviations from band additivity. First, the intensity of the upper doublet band inherited from T. dubius was significantly reduced in the 2601 and 2603 individuals, consistent with the previous observation (Figure 2). Second, the 2602 population of T. mirus differed from the other two populations by having an extra doublet of 3–4 kb that did not occur in either of the diploid progenitors or was present in low amounts beyond the sensitivity limit. Further, the band positions are slightly shifted in the 10-kb region. Within-population variation in the structure of the rDNA units was low, and the hybridization profiles of several T. mirus individuals were identical.

Figure 6.—

Southern blot hybridization showing IGS restriction site polymorphisms. DNAs from several populations of the diploid progenitors, T. dubius and T. porrifolius, and three populations of the allotetraploid T. mirus were digested with BstYI /SspI restriction enzymes and hybridized with the 26S rDNA probe. Asterisks indicate bands unique to population 2602 of T. mirus. Note extremely weak T. dubius-specific bands in populations 2601 and 2603.

Condensation patterns in rDNA chromatin:

We used FISH to compare the distribution of rDNA chromatin in the nucleus and its condensation state in individuals with silenced (displaying nucleolar dominance) and non-silenced (codominant) phenotypes. In T. dubius (Figure 7A) and T. pratensis (Figure 7G) there was a single rDNA locus (NOR) that localizes to chromosome A as described by Pires et al. (2004). In T. porrifolius, one NOR occurs on chromosome A, while a second NOR is localized to chromosome D (Figure 7B). In this species, chromosome A, unlike chromosome D, carries a 5S rDNA locus, allowing discrimination of both NORs after double labeling with the 35S and 5S rDNA probes (Pires et al. 2004). In most cases, only the NOR on chromosome D shows 35S rDNA decondensation (Figure 7B), suggesting its dominance over the NOR on chromosome A.

Figure 7.—

rDNA chromatin condensation patterns in populations of T. mirus with dominant (C, D, population 2601 plant no. 0-10) and codominant (E, F, population 2602 plant no. 7) phenotypes. T. miscellus (H) was an individual from population 2604 (plant no. 4). Diploid T. dubius, T. porrifolius, and T. pratensis progenitors are in A, B, and G, respectively. rDNA was visualized by FISH to root-tip metaphase (A–C, E, G, H) and interphase (D, F) using FITC-labeled 35S rDNA probe (green fluorescence) and biotin-labeled 5S rDNA detected with avidin-Cy3 (red fluorescence). Nuclei were counterstained with DAPI for DNA (blue fluorescence). Note: in C, the rDNA sites that show some decondensation have large, linked 5S rDNA sites (arrows); in F and D, note the condensed extranucleolar rDNA sites (arrows).

An individual from T. mirus population 2601 (0-10), representing a plant with strong silencing of T. porrifolius-origin rDNA (Figure 2A), had two partially decondensed and four highly condensed rDNA signals at metaphase (Figure 7C). The locus on chromosome D of T. porrifolius origin was identified by an absence of a linked 5S rDNA site. These sites were highly condensed at metaphase. The chromosome A homeologue of T. dubius origin, characterized by a strong DAPI band in the centromeric position and relatively strong 5S signal, was partially decondensed at metaphase. At interphase, the condensed loci remained condensed while the rDNA units on chromosome A of T. dubius origin formed a dispersed array of signals within the nucleolus (Figure 7D). These data indicate that only the latter locus is decondensed and active.

The rDNA condensation patterns were also analyzed in a plant of T. mirus population 2602 (7) that had a codominant rDNA expression profile (Figure 2A). At metaphase, four of the six rDNA sites showed secondary constrictions (Figure 7E). The two condensed sites were from chromosome A of T. porrifolius origin. At interphase (Figure 7F), the pattern of rDNA decondensation reflected the secondary constrictions at metaphase. The inactive loci from chromosome A of T. porrifolius origin formed a condensed, extranucleolar cluster of genes. Thus, at least one rDNA locus from T. porrifolius was active in this T. mirus individual.

In T. miscellus we analyzed population 2604 that show nearly complete silencing of T. pratensis units (Figure 7H). At metaphase of individual 4, two of the four rDNA sites showed secondary constriction. Relatively small size of decondensed signals on both homologs suggests that the active NOR in T. miscellus is formed from low copy T. dubius units.


Nucleolar dominance occurs in some populations of Tragopogon allotetraploids:

We analyzed here expression of rRNA genes and rDNA loci in T. mirus and T. miscellus, two recently and recurrently formed allotetraploids that originated naturally within the past 80 years. Among several populations of independent origin, we found evidence for complete dominance (4 populations), partial dominance (2), and codominance (2).

In T. mirus both NORs inherited from T. porrifolius were frequently silenced and their expression negligible. At interphase, units of T. porrifolius origin were often condensed and extranucleolar in location. These data indicate that a single T. dubius-origin locus can efficiently dominate expression over two other loci. The ability of a single master locus to dominate supernumerary loci has been noted before, e.g., a single NOR-bearing chromosome from Aegilops umbellutata can suppress all remaining NORs in hexaploid wheat (Martini et al. 1982), and in Solanum somatic hybrids, a single tomato locus can dominate NORs from potato (Komarova et al. 2004). But in Arabidopsis suecica, six A. arenosa-origin NORs dominated expression over two NORs inherited from the A. thaliana parent (Pontes et al. 2003).

In general, variation among individuals for expression ratios of the progenitor rDNA types was low, whereas differences between populations were significant. Phenotypes were usually stably inherited in progeny (with the exception of one population of T. mirus), suggesting that interspecific hybridization and subsequent polyploidization have generated distinct rDNA epiallelic states, each being typical for a particular population. In synthetic A. suecica, the direction of silencing was concordant with natural A. suecica, suggesting that there might be a predisposition toward silencing of parental genomes. Nevertheless, one strain of A. suecica showed codominance (Pontes et al. 2003), and only certain inter-ecotype diploid Arabidopsis hybrids exhibited NOR silencing (Lewis et al. 2004). These data indicate that ND may depend on the particular genotype or combination of alleles.

Two T. mirus individuals from population 2602 had strongly dominant T. porrifolius rRNA genes, suggesting that the general trend in expression patterns can be reversed. Stochastic losses of low- and single-copy genes have been observed in populations 2605 and 2604 of T. miscellus at ∼5% frequency (Tate et al. 2006). Perhaps there is a stochastic factor influencing which rDNA loci are silenced. Experiments with synthetic allotetraploids of Arabidopsis suggested that rRNA gene silencing is stochastic in the early generations, but becomes fixed in later generations (Chen et al. 1998). Perhaps T. mirus population 2602 remains unstable because it is relatively young (recently formed) with incompletely established epigenetic patterns. However, Ownbey collected plants of T. mirus in Palouse, WA (the collection site of 2602) in 1949 (Ownbey 1950; Novak et al. 1991).

Previous allozymic (Soltis et al. 1995) and AFLP (Tate et al. 2006) analysis indicated little or no inter- and intrapopulation genetic variability among the diploids, T. pratensis and T. porrifolius, while there is some variability among populations of the widespread T. dubius consistent with expectations based on the introduction of a limited number of individuals initially introduced into the Palouse region. Similarly, there were some polymorphisms in the ITS (Kovarik et al. 2005) and IGS (Figure 6) regions among populations of T. dubius and T. mirus suggesting that T. dubius subgenome is a major source of genetic variability in Tragopogon tetraploids. The influence of genetic diversity within or outside of rDNA locus on rRNA silencing remains to be determined. IGS sequence might not, however, play a major role in nucleolar dominance in allotetraploid Tragopogon because rDNA units in populations 2605 and 2606 are identical or near identical (based on ITS sequencing and blot hybridization data with 10 restriction enzymes; data not shown) but show different expression patterns. It is more likely that epigenetic factors are important. Variation in rDNA methylation levels occurs among ecotypes of A. thaliana, and this has been related to NOR activity (Riddle and Richards 2002). In Tragopogon, it is likely that rapid epigenetic changes following polyploidization are the likely cause of expression variability between populations.

Frequent dominance of low-copy rDNA:

In many allopolyploids, concerted evolution has homogenized rDNA repeats to variable degrees and directions, altering the copy number of inherited genes (Alvarez and Wendel 2003; Kovarik et al. 2004; Volkov et al. 2007). In most cases, expression is dominated by a high-copy rRNA gene family (Flavell 1986; Lim et al. 2000; Joly et al. 2004). Nevertheless, certain incongruence exists between copy number and expression. For example, a less abundant unit was expressed in one accession of Glycine (Joly et al. 2004) and a NOR in wheat with fewer repeats was more active than another with more repeats (Flavell and O'Dell 1976).

A previous study of Tragopogon allotetraploids indicated that concerted evolution has resulted in the partial homogenization of rDNA units, typically involving a reduction in the number of T. dubius-origin repeats (Kovarik et al. 2005). Surprisingly, in Tragopogon allopolyploids, the less common T. dubius-origin units were expressed and, in most cases (all except 3 individuals out of >100 tested), dominated overall rDNA expression. It is significant that both tetraploid species, T. mirus and T. miscellus, combine the T. dubius genome with different partner genomes, indicating that the capacity to dominate rDNA expression might be encoded, in part, by the T. dubius-origin subgenome. This hypothesis is supported by higher expression of T. dubius-origin units in synthetic diploid F1 hybrids (R. Matyášek, unpublished data). While dominance was nearly complete in four populations of T. mirus, it was partial in most individuals of T. miscellus, suggesting that the partner genome may modulate the magnitude of silencing. One T. mirus individual (33), belonging to the population with nucleolar dominance (2603), had extremely few T. dubius-origin units, perhaps only ∼200–300 copies per diploid cell, yet these units were highly transcribed, while >3000 units of T. porrifolius origin were completely suppressed. In A. thaliana, which has ∼800 rRNA genes per diploid cell, ∼50% of the units are estimated to be active. Electron microscopy observations indicate that there are ∼300 units active in pea nucleoli (Gonzalez-Melendi et al. 2001). By extension, nearly all units of T. dubius origin might be active in the 2603-33 individual. The situation here is clearly different from Crepis, for example, in which low-copy genes on supernumerary B-chromosomes do not seem to significantly contribute to the total rRNA pool despite their low-level transcription (Leach et al. 2005).

T. mirus population 2602 had a more balanced ratio of parental genes that were equally expressed. This result indicated that expression of T. dubius units themselves does not suppress rDNA units on partner chromosomes. Perhaps elimination of a part of the array might have contributed to elevated transcriptional activity in populations with a strongly silenced phenotype (2601 and 2603). Alternatively, there might have been a recombination event between homeologous chromosomes resulting in formation of recombinant NORs and perhaps novel units. Appearance of unique SSCP conformers (Figure 5) and restriction bands (Figure 6) in population 2602 samples is in accord with this hypothesis. Chromosomal rearrangements are thought to alter expression patterns in wheat (Schubert and Kunzel 1990), and deletions of heterochromatin adjacent to the rDNA locus resulted in the loss of suppression at the underdominant NOR in Drosophila (Durica and Krider 1978). Interestingly, loss of suppression did not prevent rDNA amplification in polytene tissues (termed replicative dominance) indicating that nucleolar dominance and developmentally directed genetic changes might be independent processes (Goodrich-Young and Krider 1989). Nevertheless, rDNA copy number at individual NORs did influence both nucleolar and replicative dominance in Drosophila interspecific hybrids. In yeast, spontaneous deletions of rDNA repeats released silencing factors, including histone deacetylases that acted in trans to silence distal heterochromatic regions (Michel et al. 2005). Interestingly, histone acetylation and DNA methylation seem to be in the center of regulation of nucleolar dominance in Arabidopsis and Brassica allopolyploids (Earley et al. 2006). Perhaps the reduction of rDNA repeats at the T. dubius NOR could have released putative silencing factors (e.g., histone deacetylases or even RNA signals (Mayer et al. 2006) that might have silenced the NOR activity on the partner chromosome.

The question arises as to how rDNA expression and copy number will evolve over longer time frames. Because it is unlikely that the most active, decondensed units will in the long run be lost, we might expect one or all of the following outcomes: (1) over time, the direction of homogenization processes is reversed and the units of T. dubius origin (or their variant) amplify once again; (2) the inactive T. porrifolius-origin genes are eliminated, mutate to become pseudogenes, or are overwritten by T. dubius-origin units; (3) tissue-specific partitioning of unit activity evolves, so that some tissues utilize T. porrifolius-origin units and others retain the use of T. dubius-origin repeats; or (4) when the number of active units is decreased beyond a certain threshold, the transcriptional block of rDNA on partner chromosomes is released. Irreversible inactivation of alleles on partner chromosomes would eventually lead to reduced fitness and population extinction. Each of the above-mentioned evolutionary trajectories has its precedent among relatively ancient allopolyploids. For example, Nicotiana allopolyploids have evolved their own rDNA types which have partially or completely overwritten the parental type and dominate expression (Kovarik et al. 2004). Yet their copy number is <30% of the number expected from Mendelian inheritance, and Lim et al. (2000) proposed that gene conversion and homogenization of parental rDNA could involve preferentially transcriptionally active units. Recent experiments in Nicotiana synthetic hybrids confirmed an inverse correlation between silencing (nucleolar dominance) and tendency toward intergenomic recombination (Dadejová et al. 2007). In insects, there is evidence that ITS within transcribed rDNA loci is more homogeneous than that in inactive loci (Van Vugt et al. 2005). On the other hand, Brassica allotetraploids, which do not seem to homogenize parental rDNA, display stable silencing in leaves; this stability becomes relaxed in floral tissues, suggesting tissue-specific partitioning of transcription activity (Chen and Pikaard 1997b). It will be interesting to see if similar developmental switches occur in Tragopogon allotetraploids that have evolved population-specific rDNA epigenotypes within just the 80 years or less since their origin.


We thank Avi Levy (Weiszman Institute, Israel) for stimulating discussion and L. Jedlickova for excellent technical assistance. This work was supported by the Grant Agency of the Czech Republic (204/05/0687, 521/07/0116), Ministry of Education, Youth and Sports of the Czech Republic (LC06004), and Academy of Sciences of the Czech Republic (AVOZ50040507), Natural Environment Research Council grant to A.R.L. and Y.K.L., and a National Science Foundation grant (MCB0346437) to D.E.S., P.S.S., and J.A.T.


  • Sequence data from this article have been deposited with the EMBL/GenBank Data Libraries under accession nos. AM493990 and AM493995.

  • Communicating editor: S. R. Wessler

  • Received March 1, 2007.
  • Accepted May 23, 2007.


View Abstract