The role played by whole-genome duplication (WGD) in evolution and adaptation is particularly well illustrated in allopolyploids, where WGD is concomitant with interspecific hybridization. This ‘Genome Shock’, usually accompanied by structural and functional modifications, has been associated with the activation of transposable elements (TEs). However, the impact of allopolyploidy on TEs has been studied in only a few polyploid species, and not in Brassica, which has been marked by recurrent polyploidy events.
Here, we developed sequence-specific amplification polymorphism (SSAP) markers for three contrasting TEs, and compared profiles between resynthesized Brassica napus allotetraploids and their diploid Brassica progenitors. To evaluate restructuring at TE insertion sites, we scored changes in SSAP profiles and analysed a large set of differentially amplified SSAP bands.
No massive structural changes associated with the three TEs surveyed were detected. However, several transposition events, specific to the youngest TE originating from the B. oleracea genome, were identified.
Our study supports the hypothesis that TE responses to allopolyploidy are highly specific. The changes observed in SSAP profiles lead us to hypothesize that they may partly result from changes in DNA methylation, questioning the role of epigenetics during the formation of a new allopolyploid genome.
Polyploidy, or whole-genome duplication (WGD), is now recognized as an important evolutionary force in plants, particularly in angiosperms, with all flowering plant lineages having experienced one or more rounds of ancient polyploidy (Cui et al., 2006; Jiao et al., 2011). The recurrent polyploidization events which have marked angiosperm evolutionary history occurred before or simultaneous to major evolutionary transitions and adaptive radiation of species. This striking coincidence suggests that WGD played a predominant role in the origin of adaptive speciation (De Bodt et al., 2005), as illustrated by the key role that appears to have been played by WGD in the survival and proliferation of plant lineages during the Cretaceous–Tertiary mass extinction event (Fawcett et al., 2009). Indeed, polyploidy is one of the main mechanisms (with unequal crossing over and retroposition) that results in genetic redundancy, which is then acted upon by selection, drift and mutation, and represents, with the emergence of new gene functions, an important source of novelty (Ohno, 1970; Zhang, 2003). Polyploids are usually divided into two categories: autopolyploids arising within a single species by chromosome doubling, and allopolyploids resulting from genome duplication, implying interspecific hybridization between two or more distinct species. It is currently estimated that, in natural populations, allopolyploids are more prevalent than autopolyploids, even though the rate of autopolyploid formation appears to be higher than that of allopolyploids (Ramsey & Schemske, 1998). Thus, the evolutionary success of allopolyploids seems to be a result of the interspecific hybridization leading to immediate evolutionary advantages (even though the question of the evolutionary advantages of genome multiplication per se still remains unanswered; Parisod et al., 2010b).
One of the main features of allopolyploids is their ‘doubled interspecific hybrid’ composition which fixes the heterozygosity at the origin of the well-known phenomenon of heterosis, or hybrid vigour, responsible for increased growth and robustness in hybrids relative to their parents (Chen, 2010). In addition, this hybridization step – the merger of differentiated complete genomes in one single cell – may constitute a major ‘Genome Shock’, as first hypothesized by McClintock (1984). This is accompanied by rapid and profound structural and functional modifications (Chen & Ni, 2006; Doyle et al., 2008; Hegarty & Hiscock, 2008; Soltis & Soltis, 2009; Gaeta & Pires, 2010), representing an important source of novelty. Such an immediate response from neo-allopolyploid genomes explains their ability to overcome competition with their parental species and to colonize new habitats (Hegarty & Hiscock, 2008), as well as a large extent of their adaptability and, indeed, their evolutionary success.
Hybridization and allopolyploidy have often been associated with the transposition of transposable elements (TEs). TEs are a major component of plant genomes (e.g. 21% of the smallest plant genome from Arabidopsis thaliana (Ahmed et al., 2011), and over 85% of the reference maize genome (Schnable et al., 2009)). They are divided into two main classes depending on their mode of transposition (Wicker et al., 2007): class I elements, or retrotransposons, move via reverse-transcribed RNA intermediates, and class II elements, or DNA transposons, move via DNA intermediates. As a consequence of their repetitiveness in the genome, TEs have played a major role in shaping the structure of plant genomes through transposition, but also recombination (Vitte & Panaud, 2005). TEs are also involved in unequal and illegitimate recombination events, and can therefore generate a large variety of structural mutations (Devos et al., 2002; Vitte et al., 2011). When TEs are activated, associated with their amplification and mobility across the genome, they can be a major source of diversity for gene expression, acting as continuous mutagenic agents, but can also be responsible for deleterious effects. Thus, TEs are mainly silenced, and their capacity to recombine, as well as their transcriptional/transpositional activity, is usually controlled by heterochromatinization and epigenetic host regulatory machineries (Slotkin & Martienssen, 2007; Lisch, 2009). Two major natural phenomena are assumed to reactivate TEs: biotic and abiotic stresses (Grandbastien, 1998; Kalendar et al., 2000); and interspecific hybridization and allopolyploidy in response to the ‘Genome Shock’ accompanied by extensive demethylation, which could potentially lead to the derepression and activation of previously silenced TEs (Madlung & Comai, 2004). Nevertheless, the effective impact of allopolyploidy on TEs has been studied in only a few specific polyploid systems (reviewed in Parisod et al., 2010a) and, surprisingly, has not yet been surveyed in any polyploid Brassica species.
The evolutionary history of the Brassica genus has been particularly marked by polyploidy. In addition to the α (3R) polyploidy event, which is unique to the Brassicaceae family (Barker et al., 2009), an ancient triplication has been shown to be unique to the Brassicaceae tribe, with strong evidence from various approaches for genome triplication in diploid Brassica species (Schranz et al., 2006). Finally, additional rounds of allopolyploidization occurred recently (< 0.01 million yr ago), with B. napus (AACC, 2n = 4x = 38), in particular, originating from natural interspecific hybridization between the diploid species B. rapa (AA, 2n = 2x = 20) and B. oleracea (CC, 2n = 2x = 18). Consequently, this allotetraploid Brassica species and its progenitors represent a good model for the investigation of the impact of allopolyploidy on TEs, in order to identify potential structural modifications linked to TE derepression. Resynthesized B. napus allotetraploids have been successfully used to characterize the regulation of gene expression following allopolyploidy by comparative proteomics (Albertin et al., 2006) and transcriptomics (Gaeta et al., 2009; Marmagne et al., 2010). The extent of the structural changes occurring during the formation of the allopolyploid genome has also been investigated (Gaeta et al., 2007; Szadkowski et al., 2010, 2011). In the present study, we analysed the same plant material to characterize the impact of allopolyploidy on three contrasting TEs (considering their genomic localization, copy number in the genome and mechanism of transposition). We developed a transposon display approach to identify the extent and nature (transposition vs genomic rearrangement) of the structural modifications which can be triggered by these TEs in the early generations of a newly formed Brassica allopolyploid genome.
Materials and Methods
Plant material and DNA extraction
The resynthesized Brassica napus allotetraploids analysed were produced by A-M. Chèvre et al. and have been used in previous studies (as specified later) (please note that all the resynthesized B. napus lines and the different diploid Brassica genotypes (except Z1) are the property of INRA Le Rheu, France (contact A-M. Chèvre)). Three homozygous doubled haploid lines, ‘HDEM’ B. oleracea var. botrytis italica, ‘RC34’ B. oleracea var. alboglabra, ‘Z1’ B. rapa ssp. oleifera (with Z1 provided by K. C. Falk, Agriculture and Agri-Food, Ottawa, ON, Canada), and two selfed progenies of ‘C1.3′ B. rapa ssp. rapifera and ‘C10’ B. oleracea var. acephala were used as diploid progenitors. Reciprocal crosses between RC34 and C1.3 led to the synthetic allotetraploids BoRCC (with RC34 as the maternal parent) and BrCRC (with C1.3 as the maternal parent) (Szadkowski et al., 2010). Another cross between C10 (maternal parent) and Z1 resulted in a biannual synthetic allotetraploid BoC10Z1 (specifically analysed in the present study). Five independent crosses between HDEM and Z1 led to the synthetic allotetraploids BoEMZ (Albertin et al., 2006; Marmagne et al., 2010; Szadkowski et al., 2010). Three of the five S0 synthetic lines (obtained from colchicine treatment of the first hybrids F1), namely BoEMZ138, BoEMZ238 and BoEMZ338, were each then used as paternal parent in crosses with the corresponding hybrid F1 to produce three other synthetic allotetraploids obtained from female unreduced gametes, designated BoEMZ119,38, BoEMZ219,38 and BoEMZ319,38, respectively (Szadkowski et al., 2011). Therefore, our plant material consisted of 11 independently resynthesized B. napus allotetraploids and their progenies, represented by a total of 109 individual plants comprising the first hybrids F1 (CA or AC), the corresponding doubled hybrids S0 (CCAA or AACC) and the following generations of selfing S1 and S2. All S0, S1 and S2 plants were verified for chromosome numbers 2n = 38 using flow cytometry and classical cytogenetics. Details of the plant material used are given in Table 1. Three cuttings per plant, randomized in three blocks, were grown under controlled conditions (18°C : 21°C, 8 h dark : 16 h light). Samples were collected and bulked from three cuttings from 10:00 h to 12:00 h. Total genomic DNA was extracted from young leaves following a standard cetyltrimethylammonium bromide (CTAB) protocol.
Table 1. Description of the plant material surveyed in the present study: scheme showing the generation of the resynthesized Brassica napus allotetraploids, sample sizes and details of the synthetic progenies
The three TEs surveyed
We chose to survey three contrasting Brassica TEs, considering their genomic localization, copy number in the genome and mechanism of transposition. To represent class I TEs, we chose a 10-kb-long Athila-like retrotransposon (Alix & Heslop-Harrison, 2004). This TE is represented by c. 400 copies in the B. oleracea genome, mainly concentrated in the pericentromeric regions of the chromosomes (Alix et al., 2005). This retrotransposon has been used previously to develop PCR-based markers which preferentially target centromeres (Pouilly et al., 2008). This element should be efficient for the identification of new TE-related genomic rearrangements occurring in the usually hypermethylated pericentromeric regions, in response to an eventual demethylation following allopolyploidy. We also targeted the Brassica MITE BraSto, which is 250 bp long with c. 125 and 310 copies in the B. oleracea and B. napus genomes, respectively. This element has been demonstrated to be located preferentially in the gene space (Sarilar et al., 2011). As we observed signs of recent proliferation in the diploid and allotetraploid Brassica species (Sarilar et al., 2011), it appeared worthwhile to examine whether this transposon is mobilized immediately after the polyploidization event. Finally, the class II CACTA transposon Bot1 was chosen for its Brassica C genome specificity; genome A is nearly devoid of any Bot1 copies (Alix et al., 2008). This 10-kb-long transposon has gone through several rounds of amplification in the C genome, whereas only a few ancestral copies are still present in B. rapa. Bot1 is present at c. 1500 copies in the B. oleracea genome and is highly dispersed along the full length of all B. oleracea chromosomes (as observed using fluorescence in situ hybridization – Alix et al., 2008).
Our SSAP strategy followed the protocol optimized for the Athila-like retrotransposon and described by Pouilly et al. (2008), which was based on previously described protocols with some modifications (Waugh et al., 1997; Ellis et al., 1998). Additional primers were specifically designed for BraSto and Bot1. The endonuclease EcoRI was used to restrict genomic DNA, which was employed, after ligation to adaptors, as the template for a two-step SSAP reaction including pre-amplification before selective amplification. Table 2 lists the sequences and designation of adaptors and primers, as well as the SSAP primer combinations used. The final amplified products were separated by electrophoresis in denaturing 5.5% acrylamide sequencing gels using a LI-COR® DNA analyser (LI-COR Biosciences, Lincoln, NE, USA) (Pouilly et al., 2008). All the SSAP fingerprints were analysed manually, and any ambiguous SSAP bands were removed from the analysis. First, we compared the parental SSAP multiband profiles with each other for each cross, and then compared the SSAP profiles of the resynthesized B. napus allotetraploids with both parental SSAP profiles, again for each cross. SSAP bands were scored as present (amplified) or absent (not amplified) for each line of the resynthesized B. napus allotetraploids surveyed: (1) SSAP bands amplified in synthetic lines and present in at least one of the corresponding diploid parents were considered as additive for these lines (Nadd); (2) extra SSAP bands amplified in synthetic lines only were scored as nonadditive (N+) for these lines; (3) SSAP bands that were missing in synthetic lines compared with diploid parents were scored as nonadditive (N–) for these lines. The proportion of nonadditivity (i.e. global nonadditivity) was estimated for any given cross, and for each primer × generation combination. It corresponds to the ratio calculated as the division between the total number of nonadditive bands and the total number of amplified bands (parental SSAP bands + SSAP bands newly amplified in progenies): Nnad = [(N− + N+)/(n × Ndiploid parents + N+)] × 100, where ‘n’ is the number of lines per generation and Ndiploid parents is the total number of bands amplified in the diploid progenitors (Ndiploid parents = Nspecific to parent 1 + Nspecific to parent 2 + Nparent 1 and parent 2). We classified the global nonadditivity observed into four main classes (strict additivity (Nnad = 0), Nnad < 5%, 5% ≤ Nnad ≤ 10%, Nnad > 10%) taking into account each ‘cross × primer × generation’ combination as a single case.
Table 2. List of primer sequences and primer pairs used for sequence-specific amplification polymorphisms (SSAPs)
We looked for differences in the total number of SSAP bands between the two parental species (B. oleracea vs B. rapa), the five parental genotypes (three B. oleracea genotypes and two B. rapa genotypes) and the different primers. To analyse the nonadditivity of the SSAP profiles among the different synthetic lines, we looked for differences in the number of nonadditive SSAP bands in relation to the crosses, generations and primers. The χ2 tests for contingency tables were all performed under the following H0 hypothesis: ‘the number of nonadditive SSAP bands observed for a specific condition is proportional to the number of individuals tested for this condition’. As an example, for the parental SSAP profiles and the variable ‘number of SSAP bands generated’, the ‘species’ effect was tested against a 3/5–2/5 theoretical distribution, as we studied three different B. oleracea genotypes and two different B. rapa genotypes (each genotype being represented by the same number of individuals). In the Results section, we report the P value associated with each χ2 test performed for the different comparisons made. The Wilcoxon signed-rank test was also applied to test the significance of more nonadditive SSAP bands appearing or missing in one given combination, with W > 0 indicating N+ > N− (with H0: E(N+) = E(N−)).
Molecular characterization of nonadditive SSAP bands
For each TE surveyed, c. 30 nonadditive SSAP bands were cloned and sequenced in order to identify the molecular polymorphism responsible for the nonadditive profile observed. SSAP bands were excised from dried silver-stained polyacrylamide gels (our protocol was modified slightly from that described in the Promega technical manual, with fixation in 10% ethanol for 5 min and oxidation in 1% nitric acid for 3 min), suspended in 40 μl of sterile ultrapure water, submitted to three freezing/defrosting cycles, boiled for 5 min and, finally, re-amplified by PCR with the unlabelled primers used for the selective amplification. PCR products were cloned (pGEM-T vector, Promega) and transformed into Escherichia coli strain DH5α. PCR was used to check that the DNA inserts were of the correct size and at least five clones per amplified SSAP band were sequenced (GenoScreen, Lille, France).
Nucleotide sequences from clones of the same SSAP band were aligned using ClustalW2 (http://www.ebi.ac.uk/Tools/msa/clustalw2/); the predominant consensus sequences were used as queries for similarity searches against nonredundant nucleotide databases at the National Center for Biotechnology Information (NCBI) (http://www.ncbi.nlm.nih.gov/) with the programme BLASTN and against Arabidopsis genes at TAIR (http://www.arabidopsis.org/Blast) using WU-BLAST2. Primers compatible with the corresponding TE-selective SSAP primers were designed against genomic regions flanking putative insertions of the TE surveyed. To identify the molecular nature of the differentially amplified SSAP bands, new primer pairs were used for direct PCR amplification on genomic DNA from individuals showing nonadditive patterns, controls represented by the diploid parents and synthetic lines displaying additive patterns. Amplified products were separated by electrophoresis in 2.5% agarose gels, and visualized after staining with ethidium bromide.
Nucleotide sequences of the SSAP bands and primer sequences, with detailed results, are provided as Supporting Information Tables S1–S3.
SSAP profiling and summary of SSAP bands differentially amplified between B. rapa and B. oleracea diploid parents
Structural changes in copies of the retrotransposon family were detected using Athila primers, which were designed in the nucleotide sequence of reverse transcriptase (Pouilly et al., 2008). Seven primer combinations were used (Table 2). For further statistical analyses, nonambiguous SSAP bands were scored; the number of scored bands varied from 12 to 52, depending on the primer pair and cross considered. When SSAP patterns between the two diploid parental species were compared (B. oleracea and B. rapa represented by three and two different genotypes, respectively), B. oleracea displayed more SSAP bands than B. rapa (P ≤ 0.05). Statistical analysis showed a genotype effect (P ≤ 0.05) with the highest number of Athila-SSAP bands amplifying from HDEM (199 – sum of the bands amplified with the seven SSAP combinations) and the lowest number from C1.3 (153). A primer effect was also detected (P ≤ 0.001) and was caused by a deficit of bands amplified with the Athila40/18-E55 (Eco-CGA) primer pair.
The BraSto-SSAP primer (BrMiselect), designed in the terminal inverted repeat (TIR) of the MITE, was used in combination with three different selective Eco primers (Table 2). It amplified between 20 and 38 SSAP bands. We analysed the SSAP patterns obtained for the five parental genotypes and observed that B. oleracea displayed more BraSto-SSAP bands than did B. rapa (P ≤ 0.001). The primer combination had an effect on the total number of SSAP bands amplified (P ≤ 0.05), with significantly more bands amplified with the selective primer E55 (Eco-CGA). The E80 (TAC)/BrMiselect primer pair amplified less SSAP bands than the other two primer pairs.
The Bot1-SSAP primer (3Botselect) was defined upstream of the CACTA motif of the transposon and was used in combination with three selective Eco primers (Table 2). Consistent with the C genome specificity of Bot1, 32–58 SSAP band levels were amplified, with more bands in B. oleracea than in B. rapa (P ≤ 0.001). The selective primer, E55 (Eco-CGA), amplified more SSAP bands than E33 (AAG), which also amplified more bands than E35 (ACA). Thus, once again, a primer effect was observed (P ≤ 0.01).
The Athila-like SSAP profiles were mainly additive in the resynthesized B. napus allotetraploids compared with their diploid parents
The Athila-like SSAP multiband profiles in B. napus lines were compared with the parental SSAP patterns, enabling the additivity hypothesis to be tested (see 'Materials and Methods'). However, when the entire dataset as a whole was considered, with no distinction between crosses, generations or primers, the SSAP bands were mainly additive, with 52% showing additivity (142 of the 273 cases studied corresponding to the different cross × generation × primer combinations; Table 3). For the remaining cases, 32% showed levels of nonadditivity between 1% and 5%, 13% between 5% and 10%, and only 3% above 10% (with a maximum of 17% for two combinations involving two different F1 hybrids). When the total number of nonadditive SSAP bands for every cross was considered, we detected a cross effect (P ≤ 0.001), with cross BrCRC showing a larger number of nonadditive SSAP bands (+ 190%) than expected (i.e. as estimated under the hypothesis that the number of nonadditive SSAP bands is proportional to the number of individuals tested, equivalent here to the mean number of SSAP bands per line multiplied by the number of lines originating from the cross BrCRC). Interestingly, the reciprocal cross BoRCC did not show any particular deviation from the expected number of nonadditive SSAP bands. The three crosses BoEMZ119,38, BoEMZ338 and BoEMZ538 generated fewer than expected nonadditive SSAP bands (−74%, −73% and −75%, respectively; Table 3). The generation (F1, S0, S1 or S2) did not have an influence on the number of nonadditive bands; nevertheless, the appearance of newly amplified SSAP bands was more frequent than the disappearance of parental SSAP bands (W > 0, N+ > N−) in the generations F1 (P ≤ 0.05), S0 (P ≤ 0.001) and S1 (P ≤ 0.001). In addition, we found a significant cross effect (P ≤ 0.001) at each generation (with F1 and S0 considered together). Interestingly, BoEMZ219,38 was the only cross to display a deficit of nonadditive SSAP bands in S1 and a high excess of nonadditive bands in S2; the general rule was that crosses displayed comparable levels of nonadditivity in both generations. Finally, we estimated the impact of each primer pair on the number of nonadditive SSAP bands amplified. We found that this parameter had a significant effect (P ≤ 0.001): more nonadditive SSAP bands were obtained with the primer pair Athila40/18-E55 (+ 69%) and, to a lesser extent, Athila40/18-E35 and Athila100/18-E84, than theoretically expected, whereas the primer pair Athila100/18-E80 generated less nonadditive SSAP bands (−68%).
Table 3. Characterization of the sequence-specific amplification polymorphism (SSAP) profiling approach developed for the three contrasting transposable elements (TEs) Athila-like, BraSto and Bot1 (the proportion of nonadditive SSAP bands identified and the main parameters affecting this nonadditivity are shown in detail)
SSAP approach: number of SSAP band levels scored:
A single case is defined as a cross × generation × primer combination.
All the percentages were calculated from the results of χ2 contingency tests as follows: deviation = ((Nobs – Ntheor)/Ntheor) × 100. Nobs, number of nonadditive SSAP bands observed; Ntheor, number of nonadditive SSAP bands theoretically expected according to the contingency table.
N+, number of newly amplified SSAP bands not present in the parental SSAP profiles; N−, number of SSAP bands present in at least one of the diploid parents but absent in one synthetic line.
Differences between crosses significant in S1 and S2
Differences between crosses significant in S2
Not significantN+ > N− in F1, S0, S1N+ = N− in S2
Significant with more nonadditive bands in S2N+ > N− in F1 and S0N+ = N− in S1;N+ < N− in S2
Significant with more nonadditive bands in S2N+ > N− in F1, S0, S1, S2
Primer effect (primer indicated if it generated more nonadditive bands)
E33 (in S2 only) +10%
The BraSto-SSAP profiles of the resynthesized B. napus allotetraploids showed the highest levels of nonadditivity relative to their diploid parents
Next, we quantified the global nonadditivity of the BraSto-SSAP profiles across the different B. napus synthetic lines. Most of the SSAP bands were additive, but in a lower proportion than observed with the Athila-like element or Bot1 (see later and Table 3): for 51% of the different cases studied (60 of the 119 cases corresponding to the different cross × generation × primer combinations), < 5% of the SSAP bands (out of the total number of SSAP bands amplified) were nonadditive, with 22% showing strict additivity. For the remaining cases, 20% displayed a level of nonadditivity of > 10% (with one combination reaching 19% in S2). Taking into consideration the different crosses available, we observed significant differences (P ≤ 0.001) between the 11 crosses which generated nonadditive SSAP bands. Of note, more nonadditive SSAP bands were amplified from crosses BrCRC and BoEMZ219,38 than theoretically expected (+ 43% and + 40%, respectively), whereas the cross BoEMZ338 generated less nonadditive SSAP bands than expected (−52%). We also observed that the reciprocal cross BoRCC displayed the expected number of nonadditive SSAP bands, and that the different BoEMZ crosses did not show homogeneous behaviour, even when considering the method used to produce them (colchicine treatment or female unreduced gametes – Table 1). Such differences for nonadditivity between crosses were significant only for the S1 and S2 generations (P ≤ 0.05 and P ≤ 0.001, respectively). We found that the generation (F1, S0, S1 or S2) had an impact on the proportion of nonadditivity (P ≤ 0.05), so that the number of nonadditive SSAP bands was, at first, stable, but then increased significantly in the S2 generation. Interestingly, this nonadditivity in S2 was characterized by a higher number of lost (N−) parental SSAP bands rather than a gain (N+) in novel bands (W < 0, N− > N+). The appearance of newly amplified SSAP bands was relatively more frequent (W ≥ 0, N− < N+) in the F1 hybrids and S0 generation (P ≤ 0.05), whereas it was more balanced in the S1 generation (not significant). We found an effect (P ≤ 0.001) of the selective Eco- primer used for SSAP amplification: the BrMiselect/E55 primer pair amplified + 88% of nonadditive SSAP bands than theoretically expected, whereas discrepancies of −28% and −60% were observed for E36 and E80, respectively.
Bot1-SSAP profiles included significantly more nonadditive bands in the S2 generation of B. napus synthetic lines
When the Bot1-SSAP profiles of the different individuals from all the different B. napus synthetic lines were compared with the profiles of the corresponding diploid parents, again the majority of the SSAP bands surveyed were additive. Indeed, 71% of the different cases studied (86 of the 114 cases corresponding to the different cross × generation × primer combinations) displayed < 5% nonadditivity, with 26% showing additivity (Table 3). The remaining 22% displayed between 5% and 10% nonadditivity, and 7% showed levels of > 10% (with one cross × primer combination showing 17% nonadditive bands in S2). When the nonadditivity of the SSAP profiles for each cross was analysed, we observed differences (P ≤ 0.001) between crosses for the number of nonadditive SSAP bands. The relative proportion of nonadditive SSAP bands for cross BoRCC was consistent with the expected numbers. Its reciprocal cross BrCRC, however, displayed more nonadditive SSAP bands than theoretically expected (+ 51%). Among the eight different BoEMZ synthetic progenies, five showed the expected number of nonadditive SSAP bands. A greater proportion of nonadditivity (+ 39%) was found in cross BoEMZ319,38, whereas fewer nonadditive SSAP bands than expected were observed in the crosses BoEMZ138 (−31%) and BoEMZ338 (−37%). Thus, the method used to generate the BoEMZ crosses (colchicine treatment or female unreduced gametes) did not have an effect. Interestingly, we estimated that the differences observed between crosses for the general nonadditivity of the SSAP profiles analysed were only statistically significant in the S2 generation (P ≤ 0.001). To complete this analysis, we examined the impact of generation. Although a significant effect was found (P ≤ 0.001) by testing all four generations (F1, S0, S1 and S2) for homogeneity in the proportion of nonadditivity, no conclusion could be drawn when the three generations F1, S0 and S1 (P = 0.1) were tested. This indicated that relatively more nonadditivity was observed for the S2 generation. When the number of disappearing (N−) vs appearing (N+) SSAP bands was evaluated, for each generation, most of the nonadditive SSAP bands corresponded to newly amplified SSAP bands (W > 0, N+ < N−;P ≤ 0.001 for each generation). Of note, the nonadditivity estimated here was dependent on the primer used to amplify the SSAP (P ≤ 0.001), with the selective primer E55 producing more nonadditive bands in all four generations (+ 41% in total), and E35 producing fewer nonadditive bands (−39%).
Table 3 provides a summary of the nonadditivity represented by each TE SSAP profile, the parameters influencing nonadditivity and their respective effects.
Some TE-mediated structural changes occurred but, overall, very few transposition events took place in the early generations of resynthesized B. napus allotetraploids
A total of 21 (N+ = 17, N− = 4), 27 (N+ = 21, N− = 6) and 30 (N+ = 18, N− = 12) nonadditive bands, which amplified with the Athila-, BraSto- and Bot1-SSAPs, respectively, were cloned and sequenced. Tables S1A, S2A and S3A provide a description and full molecular characterization of the nonadditive SSAP bands analysed. For 17 SSAP bands, more than one major sequence was obtained and thus analyzed, illustrating the complexity of the detailed analysis of multiband profiles. For the three TE-derived SSAP profiling approaches, not all the SSAP bands could be confidently related to the targeted TE. Only 35% of the Athila-SSAP sequences contained Athila-like or any reverse transcriptase fragments, and only 38% of the BraSto-SSAP sequences contained the BraSto-specific TIRs (Tables S1B, S2B). Nevertheless, more than half of the Bot1-SSAP sequences contained the typical CACTA ending motif (Table S3B).
PCR tests were then developed by designing PCR primers in all the cloned sequences; primers were used in combination with the corresponding TE-specific primer (Athila40/18 or Athila100/18, BrMiselect, 3Botselect) in order to verify the presence/absence of the corresponding nonadditive SSAP band in the two diploid parents and synthetic lines of interest (those showing the nonadditive SSAP profile, and others used as references). Figure 1 provides a summary of the information obtained from the detailed analysis of each SSAP sequence. First, for several SSAP bands (25% of all the analysed SSAP sequences), we could not draw any conclusions about the nature of the polymorphism revealed by SSAP because of invalid PCR test results. This was particularly marked for the Athila-SSAP profiling, with PCR tests of 39% of the Athila-SSAP sequences failing to give conclusive results (multiband amplicons or smears, data not shown). For a total of 61% of all the sequences analysed, we showed that these were not accompanied by any structural modifications, as the resulting PCR test profiles were additive in all the synthetic lines tested compared with the diploid parents. We could identify structural changes associated with the remaining 14% of sequences. These represented 9% and 19% of the sequences analysed for Athila- and BraSto-SSAP, respectively. According to sequence analysis, it appears that partial genomic rearrangements in the vicinity of the insertions of the TEs surveyed have occurred, rather than new transposition events. Interestingly, although 7% of the Bot1-SSAP sequences investigated were of this kind, three Bot1-SSAP sequences could be associated with putative transposition events, which corresponded to one new insertion and two excision events (one excision event is illustrated in Fig. 2). These polymorphic sites were all identified in the S2 generation of the synthetic lines.
Allopolyploidy was not accompanied by massive reshuffling of TEs in early generations of resynthesized B. napus allotetraploids
Until recently, Barbara McClintock's Genome Shock hypothesis (1984), which proposes that a burst of transposition occurs immediately after allopolyploidization, had been rarely investigated in detail. During the last decade, efforts have been made to examine the impact of allopolyploidy on the mobilization of TEs across the genome. The mobilization of various TE families has been studied in different polyploid plant species at different evolutionary time scales (e.g. Gossypium – Zhao et al., 1998; Hu et al., 2010; Triticum – Kashkush et al., 2002; Charles et al., 2008; Arabidopsis – Madlung et al., 2005; Nicotiana – Petit et al., 2007; Spartina – Parisod et al., 2009). This led to the conclusion that ‘the effect of allopolyploidy on TE genome fractions may be more complex than generally assumed’ (Parisod et al., 2010a). In the present study, we analysed early generations of different resynthesized B. napus allotetraploids for changes in the insertion profiles of three contrasting TEs. By developing a SSAP profiling approach specific to each of the three TEs surveyed, we could observe a general trend for additivity of the insertion profiles between the synthetic lines and their corresponding diploid parents. In other words, we could not detect any massive structural changes associated with the three contrasting TEs surveyed, immediately after allopolyploidization. As in previously published studies on the topic, our results reinforce the concept that highly specific TE responses occur during a particular allopolyploidy event. In Arabidopsis polyploids, Madlung et al. (2005) demonstrated that Sunfish transposons were mobilized, whereas no evidence for transposition was reported from a global survey of potential TE re-activation (Beaulieu et al., 2009). In synthetic Nicotiana tabacum allopolyploids, it was demonstrated that only a few specific young populations of the Tnt1 retrotransposon were affected and mobilized by the polyploidization event (Petit et al., 2010). These observations underline the lack of any general rules concerning the global impact of allopolyploidy on TE mobilization.
The very few transposition events identified here were related specifically to the putative activity of the Bot1 CACTA transposon, activity which was already suspected because of the existence of Bot1 transcripts in the Brassica expressed sequence tag (EST) database (Alix et al., 2008). This Brassica C genome-specific transposon appears to be the youngest TE surveyed in the present study. Indeed, the low level of sequence divergence between Bot1 and its C genome specificity suggests that this element was amplified after divergence from the A genome (Alix et al., 2008). For comparison, BraSto is present in both the A and C genomes, whereas the Athila-like element is closely related to the Athila retrotransposons of Arabidopsis. Therefore, it appears that the age of a TE could be indicative of its potential transposition activity, as observed previously for Tnt1 in Nicotiana (Petit et al., 2010). With regard to the suspected transposition activity of Bot1, which is highly supported by our current data, one evolutionary question still remains: why did Bot1 not spread to the A subgenome during the recent formation of the natural allotetraploid B. napus (Alix et al., 2008)? Indeed, this phenomenon was demonstrated for Gossypium allopolyploids with a copia-like retrotransposon specific to one parental genome (Zhao et al., 1998). One possibility is that specific Bot1 repression mechanisms set in immediately after allopolyploidization, avoiding anarchic transposition across the genome and ensuring stable establishment of the newly formed allopolyploid species (even though very few copies may escape this control). In addition, even if the hypothesis that the few transpositions depicted here may reflect the stochastic standard activity of Bot1 in the diploid parental genome (which would thus reinforce our conclusion of the low impact of allopolyploidy on Bot1 activation) cannot be completely ruled out, our data still ask the question of the mechanisms controlling Bot1 activity. The actors involved and the extent of this regulatory mechanism remain to be elucidated, especially considering that TE transcription was also demonstrated as a major target for control by small interfering RNAs (Mirouze et al., 2009).
Meiotic recombination appears to be involved in the TE-mediated structural changes observed
We identified several genomic rearrangements which resulted in part of the nonadditivity depicted (at least 11% of all the nonadditive SSAP bands analysed; however, some of the undetermined nonadditive SSAP bands may also correspond to other genomic rearrangements). It was striking that rearrangements were significantly more numerous at the MITE BraSto insertion sites, and that they were all observed in the S1 generation synthetic lines (Table S2A). This makes sense in that BraSto copies were found to be associated with the Brassica gene space, and were usually in more interstitial positions than the other two TEs surveyed here (Sarilar et al., 2011). These data echo major results reported previously from the analysis of the same resynthesized B. napus allotetraploids: it was demonstrated that structural changes were generated by the first meiosis – which took place in S0 to produce the S1 lines – and were significantly more abundant in interstitial and distal locations along the chromosomes (Szadkowski et al., 2010). It can thus be hypothesized that restructuring at TE insertion sites is strongly influenced by meiotic recombination. This is also supported by the identification of a pronounced generation effect on SSAP nonadditivity for the two TEs BraSto and Bot1, with significantly more nonadditive SSAP bands detected at the second selfed generation S2 (particularly marked with Bot1-SSAP). This effect was not significant in the Athila-SSAP profiles and Athila-like copies are concentrated in pericentromeric regions (Alix et al., 2005), where the recombination rate is low, as a result of hypermethylation of this region. We also found that the Eco primer used had a significant effect on the generation of nonadditive SSAP bands (with the selective Eco primer E55 generating significantly more nonadditive SSAP bands than the others). This finding underlines the importance of the location of the genomic regions targeted by the different primer pairs used on the extent of the changes affecting the TE insertion sites analysed by SSAPs. Here, again, our observations highlight the putative key role of the recombination rate in the generation of nonadditive SSAP profiles. Indeed, reactivation of Tnt1 retrotransposons in resynthesized Nicotiana tabacum allotetraploids was effective only in the S4 generation, after four rounds of meiosis, and was accompanied by major restructuring at the Tnt1 insertion sites (Petit et al., 2010).
Among the additional parameters examined for their involvement in SSAP nonadditivity, we paid attention to the way in which each allotetraploid was formed. Szadkowski et al. (2011) demonstrated that the type of allotetraploid formation pathway involved has a significant effect on the generation of genomic rearrangements. They reported that significantly more rearrangements occurred in synthetic lines originating from unreduced gametes. In the present study, however, we could not validate this effect. Similar nonadditivity proportions were found in the synthetic BoEMZ19,38 allotetraploids produced from unreduced gametes as those of the synthetic BoEMZ38 allotetraploids obtained following colchicine treatment of F1 hybrids. In our study, in all the different generation × primer combinations analysed and for the three different TEs surveyed, the cross effect (i.e. the union at random of two samples of gametes) was the major effect, suggesting a statistically significant proportion of stochastic SSAP nonadditivity.
Towards a hypothesis that some nonadditive SSAP bands may result from changes in DNA methylation
A relatively large number of nonadditive SSAP bands were found to be additive at the genomic structure level, according to the PCR tests performed (corresponding to 61% of all the nonadditive SSAP bands analysed in detail). Nonadditivity results, which could not be associated with transposition events or structural rearrangements (indicated as ‘no structural modification’), could be explained by competition between multiple target sites during PCR amplification, but also polymorphisms at the EcoRI endonuclease restriction sites used to develop the SSAP markers. The EcoRI recognition site is GAATTC, and it has been demonstrated that EcoRI is moderately sensitive to cytosine methylation in eukaryotic cells, as in bacteria, with a reduced rate of cleavage at hemi-methylated GAATTm5C and no cleavage with DNA containing GAATTm5C on both strands (Brennan et al., 1986; Nelson & McClelland, 1991). It is thus tempting to hypothesize that some nonadditive SSAP bands observed may correspond to methylation changes at the restriction sites. Epigenetic responses of the plant genome to allopolyploidy have been documented previously. Remodelling of DNA methylation was frequently observed in various polyploid plant species, corresponding mainly to rapid alterations occurring immediately after the allopolyploidization event (e.g. Arabidopsis – Madlung et al., 2005) or already in the F1 hybrids (e.g. Triticum – Shaked et al., 2001; Spartina – Parisod et al., 2009; Senecio – Hegarty et al., 2011). In Brassica, changes to the extent of DNA methylation were described in newly resynthesized B. napus allotetraploids (Lukens et al., 2006) and also in the same plant material as that surveyed in the work reported here (A. Salmon & A-M. Chèvre, unpublished). In addition, different studies on Arabidopsis hybrids have revealed the influence of the maternal cytoplasm on epigenetic regulation and the control of transposons and gene expression in reciprocal crosses through the production of specific small RNAs (Martienssen, 2010; Lu et al., 2012). With the involvement of small RNAs in the mediation of DNA methylation (Simon & Meyers, 2011), we can thus hypothesize that cytotype (cross direction) would also influence changes in DNA methylation in response to interspecific hybridization. Notably, when testing the cross effect for the reciprocal crosses BoRCC and BrCRC, we identified significantly more nonadditive SSAP bands in the progeny carrying the A cytoplasm, for each of the three TE-SSAP profiles (BrCRC progeny; Table 3); however, we did not observe any deviations from the expected proportions of nonadditive SSAP bands in the corresponding progeny carrying the C cytoplasm (BoRCC). This can be compared with the previously demonstrated significant cytoplasmic effect on the transmission of meiosis-related genomic rearrangements to progenies (Szadkowski et al., 2010). Although a deficit in DNA methylation has recently been demonstrated to boost recombination, at least in euchromatic regions (Melamed-Bessudo & Levy, 2012), the relationship between changes in DNA methylation patterns and the structural rearrangements which occur following interspecific hybridization and allopolyploidization (Hegarty et al., 2011) should also be investigated in newly resynthesized B. napus allotetraploids.
In conclusion, our study of the impact of allopolyploidy on the restructuring of three contrasting TEs in resynthesized B. napus allotetraploids demonstrated the lack of an anarchic burst of transposition, as generally assumed until recently. The nonadditivity of TE-anchored SSAP profiles between synthetic B. napus allotetraploids and their diploid progenitors, accompanied or not by structural rearrangements, also asks the question of some potential concordance with DNA methylation changes. These observations imply that the tight control of TEs at their insertion sites, by heterochromatinization and other hypermethylation processes, was partly relaxed in response to the recent allopolyploidization event. The very few Bot1 transposition events identified may thus be related to a higher activation of TEs, primarily at the level of transcription, but this still needs to be confirmed. Future studies are needed to ascertain whether or not the structural and functional changes induced by allopolyploidy reflect the upstream disruption of the epigenetic regulatory machinery of the neo-polyploid genome.
We thank Jean-Claude Letanneur and graduate students from AgroParisTech (Wenxuan Zhang, Thomas Murarasu, Jordan Peltier, Florent Guinot) for technical assistance. We also particularly thank Clémentine Vitte and Adrienne Ressayre for valuable discussions, and Leigh Gebbie for corrections to the English. We also thank the three anonymous referees for providing us with valuable comments and suggestions. This work was supported by the project ‘Effect of polyploidy on plant genome biodiversity and evolution’ funded by the French Agence Nationale de la Recherche (Biodiversity programme #ANR-05-BDIV-015). V.S. was supported by a PhD fellowship from the French Direction Générale de l'Enseignement et de la Recherche (DGER) via AgroParisTech and the French Centre National de la Recherche Scientifique (CNRS). P.M.P. was supported by a PhD fellowship from the French Ministère de l'Enseignement Supérieur et de la Recherche (MESR).