Speciation via interspecific or intergeneric hybridization and polyploidization triggers genomic responses involving genetic and epigenetic alterations. Such modifications may be induced by small RNAs, which affect key cellular processes, including gene expression, chromatin structure, cytosine methylation and transposable element (TE) activity. To date, the role of small RNAs in the context of wide hybridization and polyploidization has received little attention. In this work, we performed high-throughput sequencing of small RNAs of parental, intergeneric hybrid, and allopolyploid plants that mimic the genomic changes occurring during bread wheat speciation. We found that the percentage of small RNAs corresponding to miRNAs increased with ploidy level, while the percentage of siRNAs corresponding to TEs decreased. The abundance of most miRNA species was similar to midparent values in the hybrid, with some deviations, as seen in overrepresentation of miR168, in the allopolyploid. In contrast, the number of siRNAs corresponding to TEs strongly decreased upon allopolyploidization, but not upon hybridization. The reduction in corresponding siRNAs, together with decreased CpG methylation, as shown here for the Veju element, represent hallmarks of TE activation. TE-siRNA downregulation in the allopolyploid may contribute to genome destabilization at the initial stages of speciation. This phenomenon is reminiscent of hybrid dysgenesis in Drosophila.
INTERSPECIFIC or intergeneric hybridization and whole-genome doubling provide a mechanism for “overnight” speciation (Rieseberg and Willis 2007; Leitch and Leitch 2008; Soltis and Soltis 2009; Woodet al. 2009). Wheat, Spartina, Senecio, and Tragopogon are examples of recent speciation via wide hybridization and polyploidization. It has been argued that in nature, there are plenty of opportunities for interspecific hybridization and genome doubling, with an estimated 10% hybridization among animals and 25% among plants (Mallet 2007). Moreover, unreduced gametes, which are occasionally produced in most plant species (Bretagnolle and Thompson 1995), were shown to occur at a frequency of up to 50% in some combinations of intergeneric hybrids of wheat (Kihara and Lilienfeld 1949). Therefore, one might expect new hybrid and polyploid species to be formed more frequently than documented (Van De Peeret al. 2009). This apparent paradox may be resolved by works demonstrating the genome-wide genetic and epigenetic impact of genome merging or doubling, consequently challenging maintenance of genome functionality and stability. While some intergenomic combinations show hybrid incompatibility (Bomblies and Weigel 2007; Ishikawa and Kinoshita 2009; Waliaet al. 2009) or hybrid dysgenesis (Malone and Hannon 2009), or feature reduced fitness, others show improved vigor (Chen 2010). The mechanisms dictating nascent hybrid and polyploid survival are therefore of particular interest when attempting to understand the opportunities and bottlenecks of speciation.
A genetic model for such hybrid incompatibility was proposed by Dobzhansky (1936) and thoroughly reviewed by Landryet al. (2007), whereby divergent loci or alleles demonstrate incompatibility when merged into the same nucleus. More recently, Tiroshet al. (2009) performed a genome-wide analysis of gene expression in yeast interspecific hybrids and analyzed interactions between cis- and trans-acting regulatory factors and their relation to gene expression rewiring in hybrid genomes. Distinct expression patterns in the hybrid were shown to account for both hybrid failure (Landryet al. 2007) and hybrid vigor (Niet al. 2009; Birchleret al. 2010).
Whole-genome doubling of interspecific or intergeneric hybrids gives rise to allopolyploid species, which present the same type of novel intergenomic interactions as those found in interspecific hybrids while also providing fertility rescue. Plant allopolyploidy also offers an additional level of opportunities for the newly formed species, such as mutation buffering, sub- and neofunctionalization of duplicated genes, fixed heterozygosity, and a broader range of dosage responses (Comai 2005; Doyleet al. 2008; Chen 2010). Most angiosperms undergo one or more polyploidization events in their lineage (Van De Peeret al. 2009) and 2–4% of all speciation events seem to occur via polyploidization (Otto and Whitton 2000). Despite these potential advantages, polyploidization events can lead to “genomic shock” due to gene redundancy, imbalanced and antagonistic gene expression, orchestration of DNA replication among the multiple and sometimes different genomes, and pairing between multiple homologous and homeologous chromosomes. Studies in several species have shown a broad range of adverse genetic and epigenetic responses, which occur soon after hybridization and polyploidization, including DNA deletions, rearrangements or cytosine methylation, gene silencing, activation of transposons, and modification of parental imprinting (Comai 2005; Doyleet al. 2008; Chen 2010). Small RNAs have been associated with all of these events (Matzke and Birchler 2005) and therefore, changes in small RNA species in hybrids and polyploids might provide mechanistic insight into the control of genetic and epigenetic changes that occur in response to polyploidization.
Endogenous plant small RNAs can be divided into several classes (Ghildiyal and Zamore 2009), where two of the most prominent classes include the 21-nucleotide (nt)-long species, mostly corresponding to microRNAs (miRNA), and the small interfering RNAs (siRNAs), typically 24 nt in length. Several dsRNA-mediated pathways operating at the level of the nuclear genome have been described in plants and include RNA-directed DNA methylation (Matzke and Birchler 2005) and small RNA-mediated DNA demethylation (Pentermanet al. 2007). Small RNAs have been shown to be involved in a broad range of functions including heterochromatin formation and silencing (Lippman and Martienssen 2004). siRNAs seem to function as guardians against transposable elements (TEs) during plant development (Levy and Walbot 1990; Hsiehet al. 2009; Gehrinet al. 2009; Mosheret al. 1990; Slotkinet al. 2009). However, their role in transposon regulation in the context of hybridization or polyploidization has received far less attention.
In this work, we performed a high-throughput screen of small RNAs in parental, hybrid, and allopolyploid plants mimicking the events that shaped the genome during the speciation of bread wheat. We show that miRNAs and repeat-derived siRNAs respond differently to changes in ploidy level. In addition, the siRNA pools corresponding to transposons were significantly reduced upon allopolyploidization. In parallel, we show that the reduction or total disappearance of siRNAs correlated with decreased CpG methylation of target transposons. This deregulation of siRNAs, and the associated reduction in transposon methylation, may contribute to genome instability and may hinder speciation via hybridization and polyploidization.
MATERIALS AND METHODS
Seeds were germinated on wet 3MM papers in petri dishes. All seeds were imbibed for 24 hr at room temperature under regular light. Seeds of Aegilops tauschii then underwent 3–6 weeks of vernalization, at 4° in the dark. After germination, plantlets were placed in 5-liter pots and grown in a greenhouse. All plants were grown during the winter, at 20° and under short-daylight conditions.
Pollen from the diploid wheat Aegilops tauschii, cultivar TQ113 (genome DD, kindly provided by J. Dvořak), was used to manually pollinate stigmas of the tetraploid wheat Triticum turgidum ssp. durum, cultivar Langdon (TTR16, genome BBAA), which served as the female parent, to generate BAD F1 hybrids. Pollination was performed early in the morning, using freshly collected pollen, 2–3 days after emasculation of the female-parent flowers. The hybrid plants were mostly sterile. F1 spikes were bagged to prevent outcrossing. Occasionally, seeds were obtained from the spikes of F1 plants. Typically, spikes from F1 plants spontaneously gave zero to one seed, corresponding to an average frequency of ∼1%, consistent with reports of frequent occurrence of unreduced gametes, in similar intergeneric hybrids of wheat (Kihara and Lilienfeld 1949). Allopolyploid plants derived from these seeds were then characterized (see results for details).
Preparation of metaphase chromosomes from wheat root tips:
To count chromosomes in F1 hybrids and in newly synthesized amphiploids, root tips of germinated seeds were cut, subjected to 24 hr cold treatment in double-distilled water to arrest mitosis, and then transferred to a 2% acetocarmine staining solution for 2–3 days. Before squashing, root tips were briefly heated in a solution containing 4 drops of 2% acetocarmine and 1 drop of 1 n HCl. Squashing was performed on a preparative slide, in one drop of 2% acetocarmine.
Total RNA was extracted from 2 g of whole 2-month-old plant tiller tissue, using 25 ml TRI reagent (Molecular Research Center, Cincinnati, OH), according to the manufacturer’s protocol, with the exception of isopropanol precipitations and ethanol washes, which were performed overnight, at −20°. The plant material used in this study consisted of six TTR16 plants, six TQ113 plants, three F1 hybrid plants, and six newly synthesized S1 amphiploid plants, whose polyploid nature had been previously confirmed by karyotype analysis. Tissues from all genotypes were at the vegetative phase.
Small RNA library preparation:
RNA quality of all samples was verified using a bio-analyzer. Cloning of small RNAs was performed by Illumina (San Diego, CA), using the DGE small-RNA sample preparation kit protocol v1.0 (Illumina, Hayward, CA). In brief, small RNAs (18–32 nt long) were size selected, purified, and ligated with 3′ and 5′ adaptors. Four cDNA libraries of size-selected RNA from the four pools of plants were prepared by reverse-transcription PCR, followed by PCR.
Bioinformatic analysis of small RNA high-throughput sequencing data:
Eighteen million reads were obtained from high-throughput sequencing of small RNAs from the four libraries. Within each library, the unique sequences were reported as tags containing sequence information and frequency. There were 9.8 million tags in total for all four libraries. A script was developed to trim the 3′ adaptor: in the first step, the first 7 bases from the adaptor and any downstream sequence were removed. Then the process was repeated allowing one mismatch in the adaptor sequence. Sequences that contained more than 3 N’s were also eliminated. Tag frequencies were recalculated after these steps and resulted in 8.9 million tags.
Following quality control, tags with a frequency of at least 30 reads were assembled into short contigs using the Staden sequence analysis program’s (Staden 1996) “normal shotgun assembly” module with a minimal initial match of 20 bases, allowing 5% maximal mismatch, and then allowing additional assembly with 17 bases with no mismatch. Eventually, 21,787 tags were assembled into 6892 contigs. The number of reads and tags in each library is summarized in supporting information, Table S1. The Staden program produces a file, which reveals the relationship between the contigs and the sequences and thus helps determine the contigs composition and library frequencies. Contigs were annotated using Eland version 0.3 and BLAST 2.2.17 (Altschulet al. 1990). Searches were performed against the following databases: NCBI nr/nt (July 2008), the rice genome (ftp://ftp.plantbiology.msu.edu/pub/data/Eukaryotic_Projects/o_sativa/annotation_dbs/pseudomolecules/version_5.0 and version_6.0), wheat repeats (http://www.tigr.org/tdb/e2k1/tae1/wheat_downloads.shtml – wheat repeat database), T. aestivum release 2 (ftp://ftp.tigr.org/pub/data/plantta/Triticum_aestivum), T. turgidum release 1 (ftp://ftp.tigr.org/pub/data/plantta/Triticum_turgidum), and mature miRNA (http://microrna.sanger.ac.uk/sequences/). Specifically, for miRNA detection, Eland_21 (seed of 21 bases) was run against plant miRNAs using a script that slides a window of 21 bases along the contig sequence. Only miRNAs with no mismatches were reported. Using the output of BLASTn table against the TIGR wheat repeat database, the contig hits were divided into several subcategories: retroelements, DNA transposons, repetitive element, and unclassified repeats.
Contig frequency was calculated by summing the frequency of each tag comprising the contig. Interlibrary normalization was performed by dividing the contig frequency count per library, by a factor that was calculated by dividing by the total number of reads from tags with more than 30 reads. Therefore, the calculation represents the frequency per million reads. A rigorous significance test was performed according to Audic and Claverie (1997). For miRNA counting, all contigs with the same seed sequence were summed.
Northern blot analyses:
Northern blot analyses were performed as previously described (Brown and Mackey 2001). Briefly, 15 μg of total RNA was loaded on formamide-agarose gels and later transferred to a Hybond N+ nylon membrane (Amersham) via the standard wet transfer method. The Wis2-1A LTR probe was prepared using the forward primer 5′ TGTTGGAAATATGCCCTAGAG 3′ and the reverse primer 5′ GCACAATCTCATGGTCTAAGG 3′; the Veju LTR probe was prepared using the forward primer 5′ TAGAATAAATCCGAGGCATACC 3′ and the reverse primer 5′ TTAGGTTACAGTTGGACTTGG 3′, for amplification from the T. aestivum var. Chinese Spring genome.
Probes were labeled with [α-32P]dCTP (Amersham) using Klenow DNA polymerase (Fermentas). The membranes were blotted overnight with 20 pmol of the labeled probe and then exposed overnight to a phosphorimager screen, and images were visualized using the Image Gauge program. Membranes were stripped for reuse, in 0.1% SDS and verified to give no signal.
Hybrid and polyploid genomic DNA was extracted from 100 mg of leaf tissue, using the GeneElute Plant genomic DNA miniprep kit (Sigma). Two micrograms of genomic DNA suspended in 0.5 ml double-distilled water was sheared on ice by ultrasound (Microson sonicator, Misonix) set at 3 W output power, four pulses of 10 sec each with 50-sec intervals between pulses. Sheared DNA was concentrated to a final volume of 20 μl, using a speed vacuum set at 60° for 2 hr. The genomic DNA was then subjected to bisulfite conversion and purification using the EpiTect Bisulfite kit (Qiagen), according to the manufacturer’s instructions. Bisulfite ions convert nonmethylated cytosine residues to uracil residues. Fifty nanograms of purified DNA were then subjected to PCR, using degenerate primers (Metabion), designed to amplify the Veju retrotransposon LTR (Veju degenerate forward primer, 5′ AGTGAATGTYAAGTTGTTGGTG 3′; Veju degenerate reverse primer, 5′ TCRAACAACCTARCTCATRATAC 3′). The PCR products were then separated on a 2% agarose gel and purified using a PCR purification kit (RBC). Cleaned PCR products were ligated to a pGem-T easy vector (Promega), which were then used to transform Escherichia coli (top 10 strain); transformed bacteria were then plated on selective plates. Plasmid DNA was extracted from 25 bacterial colonies for each plant and then sequenced. Sequences were analyzed using the Sequencher 4.9 program, and methylation patterns were analyzed using the Kismeth program http://katahdin.mssm.edu/kismeth/revpage.pl (Gruntmanet al. 2008).
Phenotypic analysis of hybrid and polyploid wheat:
To gain insight into the molecular mechanisms leading to the genetic and epigenetic changes occurring upon hybrid and polyploid formation, the small RNA profiles of parental wheat lines were compared to those of the derived hybrid and the first allopolyploid generation. More specifically, the four wheat species analyzed included the parental tetraploid T. turgidum ssp. durum (genome BBAA) and the diploid Ae. tauschii (genome DD), their synthetic triploid hybrid (genome BAD), and their derived hexaploid (genome BBAADD). This synthetic hexaploid is analogous in genome structure to bread wheat T. aestivum ssp. aestivum. Parental lines were self-pollinated for several generations, with spike bagging to prevent cross-pollination, and were thus considered inbred. Hybrid seeds were obtained by crossing T. turgidum and Ae. tauschii. The hybrid F1 seeds were obtained from the female tetraploid plants and were found to be very small and shriveled compared to their parents (Figure 1). Upon germination, F1 seeds gave rise to F1 plants bearing necrotic leaves (Figure 1), a typical hybrid incompatibility phenotype (Bomblies and Weigel 2007). Some F1 plants died shortly after germination. Nevertheless, those that survived were vigorous and featured spikes larger than their parents (Figure 1), but most were sterile. Occasionally, seeds were spontaneously obtained from the spikes of F1 plants (See materials and methods) and were termed “S1.” While these seeds were slightly shriveled, they were larger in size than the seeds of either parent (Figure 1). Plants germinating from S1 seeds has a duplicated genome as confirmed by karyotype analysis (Figure S1). The resulting S1 plants were fertile and bore spikes that yielded S2 seeds, which were less shriveled than, but similar in size to, S1 seeds. While S1 plant leaves were highly necrotic, the plants were vigorous and featured spikes larger than those in the hybrid or in the parents and leaves similar in size to those of the tetraploid parent (Figure 1).
High-throughput sequencing of small RNAs:
Total RNA was isolated from tillers of 2-month-old wheat plants, which primarily included somatic tissues, namely leaves and stems and some meristematic tissues. A small RNA library was prepared for each of the analyzed wheat types and was composed of a pool of RNA collected from three to six plants of each genotype. Eighteen million reads were obtained from small RNA sequencing of the four libraries (Table S1). The unique sequences within each library were reported as tags (containing sequence information and frequency). These tags underwent strict quality checks (Figure S2 and materials and methods), and only tags that had 30 reads or more were included in the analysis, due to lack of a fully sequenced reference genome. As the data from a number of small RNA sequencing experiments proved highly reproducible (with a correlation of Spearman’s ρ = 0.95–0.98) for tags with ≥30 reads (Fahlgrenet al. 2009), this number was set as the threshold for data analysis. Of the 18 million reads obtained, 6 million high-quality reads, corresponding to 21,787 unique sequence tags, were analyzed. The small RNA sizes ranged between 18 and 35 nt and included the two prominent classes of 21 nt- and 24 nt-long small RNAs (Figure 2). The 21-nt class corresponds mainly to miRNA, while the 24-nt class corresponds most likely to siRNAs (Ghildiyal and Zamore 2009). The 24-nt-long small RNAs were most abundant within the two parental and and the hybrid libraries, whereas, the 21-nt-long small RNAs were most prevalent in the allopolyploid library.
The various small RNAs obtained from tags with ≥30 reads were categorized according to sequence similarity (Figure S3). miRNAs were one of the most abundant small RNA species (21–44%, depending on the genotype). Small RNAs that matched repeats, which were largely TEs, represented ∼12% of the total reads for all libraries, aside from the allopolyploid library, where a significant decrease (P(χ2) < 0.0001) to only 6% was observed. Surprisingly, tRNAs comprised ∼9% of total hits in the parental and hybrid libraries and rose to ∼20% (P(χ2) < 0.0001) in the allopolyploid library (Figure S3). Small RNA contigs matching tRNA genes ranged in size between 18 and 36 bases. However, 55% of the small RNA tags corresponding to tRNAs showed distinct peaks at 19, 20, and 21 nt, suggesting that they were not degradation products of larger tRNA molecules. No exceptional RNA degradation was observed in any of the samples, hence degradation could not explain the higher tRNA frequency observed in allopolyploid samples.
Abundance of miRNAs in parent, hybrid and allopolyploid libraries:
Twenty-one known plant miRNA sequences were identified with high certainty, i.e., with ≥30 reads each, and were identical to known miRNAs from other plant species (Table S2). For example, miR168, the most abundant miRNA in rice (Luet al. 2008), was the most abundant in the present data. Interestingly, the amount of miRNA, relative to total small RNA, increased with increasing levels of ploidy, being the lowest (21%) in the diploid Ae. tauschii DD genome, 33% in the triploid hybrid genome (BAD), 38% in the tetraploid T. turgidum BBAA genome, and 44% in the synthetic hexaploid BBAADD genome (Figure 3).
To assess the changes in the abundance and profile of miRNAs expressed as a result of hybridization and polyploidization, we compared the normalized number of hits (in reads per million) found in the hybrid and the polyploid libraries to the average number of hits in the two parental libraries, namely the midparent value (MPV). Under the assumption of additive expression, the number of hits in F1 or S1 should be similar to the midparent value: log2(F1/MPV) or log2(S1/MPV) = 0. Positive log2 values indicate that the number of small RNAs for a given tag is higher in the hybrid or the polyploid than in the parent average; negative values indicate the opposite. When applying a cutoff value in which |log2| < 0.5 is considered similar to MPV, most miRNAs were expressed to similar degrees as the MPV (Figure 4). However, miR390a was underrepresented (log2(F1/MPV) = −4.8), and miR160f was overrepresented (log2(F1/MPV) = 4.86) when comparing hybrid miRNA expression profiles to those of the parent libraries. Interestingly, both miR390 and miR160 play key roles in auxin regulation, via regulation of TAS3, a trans-acting siRNA involved in auxin signaling (Fahlgrenet al. 2006) by the former and via targeting of ARF10 (Liuet al. 2007) by the latter. In the allopolyploid plants, several miRNA expression patterns significantly deviated from those observed in the parental lines. miR157a, miR160f, miR159a, miR396d, and miR168 were significantly overrepresented relative to MPV, while miR156b, miR165a, and miR1135 were significantly underrepresented. miR168 had the highest number of reads in almost all libraries, amounting to 25% of the polyploid library reads and 119,710 and 79,244 reads in T. turgidum and Ae. tauschii libraries, respectively (Table S2). It was overrepresented by 1.27-fold in the hybrid and by 2.46-fold in the synthetic polyploid, relative to the midparent value (MPV = 99,477) and is suggested to have genome-wide effects on small RNA species (see discussion). miR156 was the second-highest tag expressed in the model, while miR1135 was only moderately expressed (Table S2). miR156a was overrepresented by 2.4-fold in the polyploid relative to the MPV. miR156 and its related sequence is of particular interest, as its overexpression has been suggested to prolong the vegetative phase and to delay flowering in Arabidopsis (Schwarzet al. 2008) and may explain the heterotic effects seen here in the allopolyploid plants.
siRNAs that correspond to repeats:
Small 24-nt-long RNAs were abundant in all the libraries, where a large percentage of these siRNAs corresponded to repeats: 1113 Staden contigs out of a total of 6892 correspond to transposons and retrotransposons. An additional 130 contigs corresponded to ribosomal RNA genes and telomeric and centromeric repeats. siRNAs matching known genes often corresponded to repeats as well. Here, we focused on repeats, using the complete wheat repeat database (http://wheat.pw.usda.gov/ITMI/Repeats/flatfile.total) as reference.
Transposons, which make up approximately 80% of the wheat genome (Charleset al. 2008), comprise a major class of repeats. Recently, small RNAs have been mapped to the repeat sequences of the Triticeae repeats database (Cantuet al. 2010). This work demonstrates the changes in TE-related small RNAs following hybridization and polyploidization and confirms that TEs are targets of small RNAs, particularly at their termini. In contrast to miRNAs, the amount of siRNAs corresponding to transposons decreased with increased ploidy (Figure 3). In the hybrid, 42% of siRNAs corresponding to transposons were represented similarly to the MPV (defined here as |log2| < 0.5), while 39% were underrepresented and 19% were overrepresented. In the allopolyploid, a massive reduction and significant shift (P(χ2) < 0.0001) in the relative abundance of siRNAs corresponding to transposons was observed, with 85% of the hits below the MPV (Figure 5). This underrepresentation affected both DNA elements and retroelements (Table 1). Small RNA matching to the sequenced and annotated T. aestivum BAC genome region (GenBank accession no. CT009735), using IGB (Integrative Genome Browser), confirmed the massive reduction of small RNAs corresponding to transposons, in the polyploid compared to the parental lines (Figure 6). Note that in the data used by Cantuet al. (2010), the number of reads corresponding to Veju in natural hexaploid wheat was very low (1/400,000), similar to the data shown here for the synthetic hexaploid, but contrasting with the present data collected from natural tetraploid and diploid wheat.
We then examined whether the reduction in TE-related siRNAs correlated with hallmarks of transposons activation. Table 1 summarizes several examples of retrotransposons and DNA transposons for which small RNAs were underrepresented in the polyploid compared to the parental lines. We then tested whether siRNA underrepresentation correlated with transcriptional activation of the copia-like retrotransposon Wis2-1A and of Veju, a “terminal-repeat retrotransposon in miniature” (TRIM) element (Sanmiguelet al. 2002). In both cases, the allopolyploid expressed higher transcript levels than the hybrid, further substantiating the observed reduction in corresponding siRNAs. However, the tetraploid parent also expressed high transcript levels, thus, no straightforward correlation can be drawn between small RNA and corresponding mRNA levels, detected by Northern hybridization (Figure S4).
Cytosine methylation represents an additional hallmark of transposon activity and has been reported to be reduced in active and hypermethylated in silent transposons (Chandler and Walbot 1986). As a case study, we performed bisulfite sequencing of the LTR of a Veju element, using the available sequence from a transposon-rich region in the T. aestivum genome (GenBank accession no. CT009735, coordinates 38,989–39,363). Seven different small RNAs, corresponding to hundreds of reads, were mapped along the LTR sequence (Figure 7, middle) and shown to derive from both strands. All seven were almost fully suppressed in the allopolyploid line (Figure 7, top). Sequencing of 25 different Veju-LTR clones per plant type, following bisulfite conversion, enabled us to determine the average methylation of Veju elements similar in sequence to the T. aestivum LTR (GenBank accession no. CT009735; see methods). A decrease in CG methylation (Figure 7, bottom) but not in CHG or CHH methylation (data not shown), was observed in the allopolyploid line and correlated with decreased abundance of Veju LTR small RNAs in the polyploid. These results are consistent with and further elucidate data reported by Kraitshteinet al. (2010) with regard to AFLP-based Veju methylation.
Small RNAs are involved in a number of key cellular processes, and perturbations in their steady-state expression levels can lead to genome-wide changes in gene expression or in chromatin and genome structure. Modified small RNA expression levels have been widely reported in the context of developmental regulation of plants, but to a lesser extent in the context of speciation via interspecific hybridization and allopolyploidization (Haet al. 2009). Haet al. (2009) describe an allotetraploid hybrid, analogous to Arabidopsis suecica, formed between a synthetic Arabidopsis autotetraploid line and the natural Arabidopsis tetraploid, A. arenosa, and report significant deviations in hybrid miRNA expression from MPV. In this study, most hybrid miRNA expression profiles did not deviate from their MPVs (Figure 4). However, a number of deviations were found in the allopolyploid (Table S2). In addition, some classes, such as miR156 and miR166, included several sequence variants, each demonstrating differential expression patterns. However, in the absence of the wheat genome sequence, it is difficult to associate changes in miRNA levels to the expression of target genes or to plant phenotypes.
Another difference between the present model and that described in the Arabidopsis allopolyploid analysis (Haet al. 2009) lies in the nature of the genetic material analyzed, where the Arabidopsis work involved plants of the same ploidy level (parents and hybrid were all tetraploids), while this study encompassed lines of different ploidy levels (parents [2× and 4×], hybrid [3×], and a derived allohexaploid [6×] analogous to natural bread wheat). The wheat material offered unique and novel insight into an unexpected response to changes in ploidy levels. Remarkably, and unexpectedly, the relative amount of small RNAs corresponding to miRNAs increased with ploidy (Figure 3). Inversely, the relative amount of 24-nt small RNAs corresponding to transposons decreased with increased ploidy (Figure 3). Similarly, in the work of Cantuet al. (2010), the count per basepair of small RNAs corresponding to transposons was lower in hexaploid than in tetraploid wheat varieties (Cantuet al. 2010). This ploidy dependence seems to be insensitive to genomic composition, but sensitive to dosage. Indeed, while the diploid, tetraploid, and hexaploid wheat had different genomic composition, the triploid (hybrid) and allohexaploid, which both featured the same three wheat genomes (A, B, and D) but at varying doses (3× vs. 6×), expressed divergent small RNAs profiles.
Mechanistically, a correlation can be drawn between specific miRNA overrepresentation and repressed siRNAs formation. More specifically, miR168 overrepresentation in polyploids may account for ARGONAUTE 1 (AGO1) suppression (Mallory and Vaucheret 2009), which in turn reduces the production of siRNAs (Vaucheret 2008). Although, the regulation of AGO1 is quite complex (Mallory and Vaucheret 2009), it is conceivable that affecting regulatory loops responsible for AGO1 activity, e.g., via alterations in miR168, may bear genome-wide effects on siRNAs. In fact, the novel intergenomic interactions and novel dosage effects demonstrated in the allopolyploid lines may account for the disruption of the activity of several genes related to small RNA expression and stability. Previously, we have shown that novel interactions between cis- and trans-acting regulatory elements can lead to overexpression or, alternatively, to suppression of ∼10% of the genes of an interspecific hybrid (Tiroshet al. 2009). Deregulation of central genes involved in the gene silencing machinery could account for some of the alterations in small RNA profiles reported here, which in turn can affect TE activities, as discussed below.
siRNAs play a critical role in heterochromatin maintenance and in transposable element silencing (Zaratieguiet al. 2007). The strong decrease in the percentage of siRNAs corresponding to TEs reported here could thus lead to transposon activation. Transposons, notoriously responsive to “genomic shocks” (Mcclintock 1984), may be responsive to hybridization- and polyploidization-induced “shocks,” linked with downregulation of small RNAs. Indeed, it has been reported that silent transposons and retrotransposons may become transcriptionally and sometime transpositionally active upon hybrid and polyploid formation (Kashkushet al. 2003; Madlunget al. 2005; Chen and Ni 2006; Kraitshteinet al. 2010). The correlation between small RNAs cytosine methylation and transposon silencing (Matzkeet al. 2007) may further explain transposon activation following small RNAs-related demethylation, as reported here for the Veju TE.
A fascinating aspect of TE regulation involves their capacity to transit between inactive and active states. Remarkably, a hypomethylated TE can become remethylated within one generation, as shown here for a specific element (Figure 7) and as shown on a genome-wide scale for the Veju element (Kraitshteinet al. 2010). An attractive model for such kinetics is demonstrated via the disappearance of small RNAs, as seen here in S1 for several elements, including Wis2-1A (Table 1), which then leads to transcription of otherwise silent TEs. As a result, high levels of aberrant and dsRNAs may be generated, as many TEs are nested within each other in opposite orientations (Sanmiguelet al. 1996). In the subsequent generation, small RNAs produced by such transcriptional activation, might reverse the element to its original methylated and silenced state. This proposed mechanism may account for the reported maintenance silencing of repeats by small RNAs produced by RNA polymerases IV and V (Pikaardet al. 2008; Matzkeet al. 2009). In other words, TE activation may catalyze TE silencing via upregulation of small RNA production.
However, the association between small RNAs and TE transcription still remains elusive. In an earlier work (Kashkushet al. 2003), we analyzed an allopolyploid resulting from a cross between two diploid parents and measured S1 transcriptional activation of transposons that were silent in the parents. The present model, using different parental species, is a bit more complex (Figure S4), where S1 transposon transcript levels were higher than those in F1, substantiating an earlier report of transcriptional activation in the S1 generation (Kashkushet al. 2003). However, high levels of both small RNAs and TE transcripts were found in the tetraploid parent (Figure S4), which may be due to transcription of a divergent subfamily of transposons, not targeted by the small RNAs, or to alteration in small RNA mobility, processing, or other unknown phenomenon.
Another intriguing question remaining relates to the changes measured in allopolyploid small RNA species, and to a lesser extent in the hybrid. This observation may be linked to polyploidy per se. Alternatively, this may be a result of the kinetics of small RNA alterations, which begin in the hybrid and peak in the S1 generation. Another interesting possibility is that global demethylation and activation of TEs may occur through disruption of developmentally regulated processes in the germline. Such modifications may occur during meiosis, or gametogenesis, or during the development of an embryo derived from “deregulated” gametes. In such cases, massive changes would be found in the S1 only, and not in the F1.
In summary, the data reported here, together with previous findings, suggest that deregulation of small RNAs may stimulate TE activation in interspecific hybrids and allopolyploids. This phenomenon is reminiscent of hybrid dysgenesis in Drosophila (Malone and Hannon 2009), a mechanism of incompatibility that can hinder speciation via hybridization and polyploidization. The correlation between the different hallmarks of transposon activation (small RNAs, hypomethylation, transcription, and transposition) is complex, as they undergo rapid changes between generations. In addition, it is important to note that this study was performed on plants that survived hybridization and allopolyploidization. The events of seedling death and seed abortion, which were observed in the hybrid and allopolyploid lines, may have occurred as a result of severe transposon activation and were not considered in our analysis.
We thank Naomi Avivi-Ragolski and other members of the Levy laboratory for their help and discussions; Idan Efroni for help in data analysis; Yehudit Posen for editing the manuscript; and members of the bioinformatics unit, especially Jaime Prilusky for writing the quality-control script and Ester Feldmesser for help with statistical analysis. This work was funded by a grant from the Israeli Science Foundation, no. 616/09, to A.A.L.
Supporting information is available online at http://www.genetics.org/cgi/content/full/genetics.111.128348/DC1.
Sequence data from this article have been deposited with the GenBank Data Libraries under accession no. GSE29243.
- Received March 3, 2011.
- Accepted April 4, 2011.
- Copyright © 2011 by the Genetics Society of America