Centromeres of allopolyploid A. suecica incorporate both CENH3 variants, but more from A. arenosa
To address whether subgenome-specific CENH3 variants undergo subgenome-specific centromere loading in an allopolyploid species, we used the allotetraploid A. suecica as a model. First, the hybrid nature of A. suecica was confirmed by multicolour FISH using centromere-specific DNA probes for A. thaliana (AtCEN) and A. arenosa (AaCEN) centromeres (Fig. 1a). After FISH, chromocenters revealed either A. thaliana- or A. arenosa-specific signals. Next, an A. arenosa CENH3-specific antibody (anti-AaCENH3) was produced and tested for specificity with mitotic and interphase chromosomes of A. arenosa (Fig. 1b, Suppl. Figure 1b). The absence of immunosignals in A. thaliana nuclei confirmed the species specificity of anti-AaCENH3. The species specificity of anti-A. thaliana CENH3 (anti-AtCENH3) was demonstrated as only the centromeres of A. thaliana but not of A. arenosa displayed immunosignals (Fig. 1b, c). Double immunostaining using both antibodies and subsequent FISH with the A. arenosa centromere-specific probe confirmed that both CENH3 variants are incorporated in all centromeres of A. suecica independent of their origin and sequence (Fig. 1c).
Next, the centromeric localization of both subgenome-specific CENH3 variants was tested in synthetic allotetraploid A. suecica (N22665) plants. Double immunostaining revealed that about 60% of analyzed sorted nuclei incorporated equally both subgenome-specific CENH3s, resulting in either strong (AT/AA) or weak (at/aa) signals indicating an unbiased loading of subgenome-specific CENH3s (Fig. 2). In cases with a biased subgenome-specific CENH3 loading (i.e. AT/aa, at/AA, AT/0, at/0, 0/AA or 0/aa), we never found nuclei with strong AtCENH3 signals (i.e. AT/aa and AT/0) (Fig. 2 and Suppl. Table 2). In contrast, about 40% of nuclei, showed strong AaCENH3 signals (i.e. at/AA and 0/AA), indicating a biased loading of A. arenosa subgenome-specific CENH3.
To check whether the same bias exists in natural allopolyploid A. suecica, derived from ancient A. thaliana x A. arenosa hybridization events, we performed double immunostaining on sorted nuclei of A. suecica accession “Sue2”. About 70% of analyzed nuclei, had equally incorporated subgenome-specific CENH3s (Fig. 2). In cases with a biased subgenome-specific CENH3 loading (i.e. AT/aa, at/AA, AT/0, at/0, 0/AA or 0/aa), no nuclei with strong AtCENH3 signals (i.e. AT/aa and AT/0) were observed (Fig. 2 and Suppl. Table 2). However, about 30% of all nuclei, showed strong AaCENH3 signals (i.e. at/AA and 0/AA), demonstrating a similar biased A. arenosa CENH3 incorporation. To exclude that the observed signal differences of both species-specific CENH3 antibodies were not caused by a lower quality of the AtCENH3 antibody, an additional control experiment was performed with tetraploid A. thaliana, the parent genotype we used for wide hybridization with A. arenosa. The same combination of antibodies resulted in nuclei with about 71% strong (AT/0) and 27% weak (at/0) AtCENH3-specific signals, respectively (Fig. 2). Thus, our findings indicate that both subgenome-specific CENH3 variants are incorporated into the centromeres of A. suecica, but with a positive bias towards A. arenosa CENH3.
Preferential expression and loading of AaCENH3 become established immediately after hybrid formation
Crossing of both species was performed to determine whether the preferential A. arenosa-type CENH3 incorporation is already established in the first generation after the wide hybridization. Only pollination of tetraploid A. thaliana with tetraploid A. arenosa resulted in fertile seeds. As aneuploidy and chromosome elimination were described for crosses between A. thaliana and A. arenosa (Comai, Tyagi et al. 2000, Wright, Pires et al. 2009), flow cytometry was employed to identify successful hybridization events. In an initial pre-screening without an internal reference standard, one (plant 6) out of 30 plants was identified as diploid (Suppl. Figure 2a) while all other plants were confirmed to be tetraploid. Immunostaining with both types of CENH3 antibodies on nuclei of plant 6, confirmed the absence of the pollinator genome, as only up to 10 AtCENH3-specific signal clusters were found (Suppl. Figure 2b). Hence, here all A. arenosa-derived chromosomes were eliminated, likely during hybrid embryogenesis. Uniparental elimination of chromosomes is a common phenomenon in hybrids derived from distantly related species (reviewed in (Ishii, Karimi-Ashtiyani et al. 2015)). Furthermore, we noticed in the flow cytometric pre-screen an obvious variation in the peak positions between the measurements of the individual plants (Suppl. Figure 2c). Therefore, we determined the genome size of 24 putative F1 hybrids using Raphanus sativus as an internal reference standard and compared the data to values obtained for the tetraploid parents (Suppl. Figure 3a). For tetraploid A. thaliana and A. arenosa, we estimated genome sizes of 0.683 pg/2C and 0.825 pg/2C, respectively, indicating an expected genome size for the interspecific hybrid of 0.754 pg/2C. Among the analyzed plants, 13 plants revealed a genome size deviating by less than 3% from the expected value of a hybrid plant and were considered as euploid hybrid plants. 11 plants showed a deviation from the expected genome size of more than 3%, presumably as a result of aneuploidy. One plant is most likely the product of an A. thaliana selfing event or spontaneous doubling of haploid progeny (plant No. 25, Suppl. Figure 3a, b).
Based on the determined DNA contents, leaf nuclei of selected 3-month-old F1 hybrid plants were sorted on slides and analyzed by double immunostaining. About 50% of nuclei showed an unbiased loading of subgenome-specific CENH3 variants (AT/AA and at/aa nuclei). Out of the analyzed nuclei, only 3% showed stronger AtCENH3 signals (i.e. AT/aa, AT/0 and at/0 nuclei). In contrast, about 45% of CENH3-loaded nuclei, showed stronger A. arenosa signals (i.e. at/AA, 0/AA and 0/aa nuclei) (Fig. 2, Suppl. Table 2), indicating a biased loading of A. arenosa-specific CENH3. The observed increase in the frequency of nuclei with an equal proportion of parental CENH3s (i.e., AT/AA and at/aa nuclei) from F1 to natural hybrids of A. suecica through generations suggests a gradual step-wise adaptation of CENH3 variant loading in allopolyploid Arabidopsis (Fig. 2, Suppl. Figure 4). Super-resolution microscopy confirmed the mixed composition of both parental CENH3s in F1 hybrid nuclei (Fig. 3, Suppl. Movie 1). However, the ultrastructure of AtCENH3 and AaCENH3 signals differed but intermingled, suggesting that the subgenome-specific CENH3 variants are preferentially loaded into different centromeric nucleosome arrays.
The observed biased CENH3 incorporation prompted us to analyze the transcription of both parental CENH3s. The relative transcript levels of A. thaliana- and A. arenosa-derived CENH3 were quantified and then normalized to ACTIN2 (At3g18780) after qPCR with AaCENH3- and AtCENH3-specific primer pairs (Fig. 4, Suppl. Table 1). In all analyzed tissues (young rosette leaves, flower buds, and siliques) of allopolyploid A. suecica (natural hybrid) and synthetic hybrids (F1, older synthetic hybrid), the relative expression of A. arenosa CENH3 was higher in comparison to A. thaliana.
In summary, we conclude that the majority (above 90%) of centromeres of A. thaliana x A. arenosa F1 hybrid, synthetic and natural A. suecica incorporate CENH3s of both parental genomes, despite a different centromere DNA composition (Fig. 5). However, after the formation of the F1 hybrid, the contribution of A. arenosa-derived CENH3 is higher than that of AtCENH3.