Epigenetic histone H3 phosphorylation marks discriminate between univalent- and bivalent-forming chromosomes during canina asymmetrical meiosis

Abstract Background and Aims Dogroses (Rosa sect. Caninae) are mostly pentaploid, bearing 2n = 5x = 35 chromosomes in somatic cells. They evolved a unique form of asymmetrical meiosis characterized by two types of chromosomes: (1) chromosomes forming bivalents and distributed in the normal sexual way; and (2) chromosomes occurring as univalents and transferred by a female gamete only. In the mature pollen of pentaploid species, seven bivalent-derived chromosomes are transmitted to offspring, and 21 unpaired univalent chromosomes are eliminated during microsporogenesis. To discriminate between bivalent- and univalent-forming chromosomes, we studied histone H3 phosphorylation patterns regulating meiotic chromosome condensation and segregation. Methods We analysed histone modification patterns during male canina meiosis in two representative dogrose species, 5x Rosa canina and 5x Rosa rubiginosa, by immunohistochemical and molecular cytogenetics approaches. Immunostaining of meiotic cells included α-tubulin, histone H3 phosphorylation (H3S10p, H3S28p and H3T3p) and methylation (H3K4me3 and H3K27me3) marks. In addition, fluorescent in situ hybridization was carried out with an 18S rDNA probe. Key Results In the first meiotic division, univalent chromosomes underwent equational division into chromatids, while homologues in bivalents were segregated as regular dyads. In diakinesis, bivalent chromosomes displayed strong H3 phosphorylation signals in proximal regions, spreading to the rest of the chromosome. In contrast, in univalents, the H3 phosphorylation signals were weaker, occurring mostly outside proximal regions largely overlapping with the H3K4me3 signals. Reduced phosphorylation was associated with relative under-condensation of the univalent chromosomes, particularly at early diakinesis. Conclusions We hypothesize that the absence of pairing and/or recombination in univalent chromosomes negatively affects the histone H3 phosphorylation of their chromatin and perhaps the loading of meiotic-specific cohesins. This apparently destabilizes cohesion of sister chromatids, leading to their premature split in the first meiotic division.

Section Caninae of the genus Rosa L. represents a highly successful group of wild roses called dogroses (~60 species in total) distributed in Europe and Western Asia (Wissemann, 2003;Koopman et al., 2008).All dogroses are polyploids originating from multiple hybridization events (Wissemann, 2000;Ritz et al., 2005); most species are pentaploid (2n = 5x = 35), with some tetraploids, hexaploids and heptaploids (2n = 4x, 6x, 7x = 28, 42, 49) (Klášterská, 1969;Wissemann, 2002;Herklotz and Ritz, 2017).Dogroses display a peculiar form of asymmetrical meiosis called 'canina meiosis', which was observed more than a century ago (Täckholm, 1920;Blackburn and Harrison, 1921).Regardless of ploidy level, only 14 chromosomes pair in meiosis (seven bivalents) and are regularly distributed to the male and female gametes, while the remaining chromosomes are transmitted as univalents via the egg cell (Fig. 1).Thus, male and female gametes contribute unequally to the chromosome complement of the zygote.The resulting matroclinal inheritance of genetic material has been confirmed by microsatellite and ribosomal DNA (rDNA) markers (Nybom et al., 2004(Nybom et al., , 2006;;Kovarik et al., 2008;Herklotz and Ritz, 2017).A population-level study involving thousands of European samples revealed considerably fewer copies of identical multi-locus genotypes in dogroses in comparison to species with regular sexual reproduction (Reichel et al., 2023), supporting their considerable genetic diversity.In male meiosis, univalent chromosomes behave as laggards, i.e. display retarded migration compared with that of bivalents in the first meiotic division.The expression activity of genes located on univalent chromosomes has been debated considerably (Darlington, 1965;Khaitová et al., 2010;Ritz et al., 2011;Vogt et al., 2015).Earlier studies suggested these chromosomes to be primarily inactive and heterochromatic (Darlington, 1965), whereas more recent studies revealed transcriptional activity of genes encoded by chromosomes forming univalents (Ritz et al., 2011;Herklotz et al., 2018).However, chromosome studies in roses are methodologically challenging because of the small size of the chromosomes and their sticky behaviour (Klášterská and Natarajan, 1974a, Kirov et al., 2014, 2016;Klásterská and Natarajan, 2014) and generally low number of mitotic and meiotic cells.In addition, a high content of polyphenolic compounds in the cytoplasm complicates the preparation of microscopic samples (Roman et al., 2013).Nevertheless, several classical cytogenetic studies confirmed the original findings of Täckholm (1922), showing highly polarized meiosis in dogrose spermatocytes (Klášterská and Natarajan, 1974a;Roberts, 1975;Lata, 1982;Lim et al., 2005).
The epigenetic control of the cell cycle through posttranslational histone modifications (phosphorylation, acetylation, methylation, etc.) plays a fundamental role directly related to changes in chromatin structure and function.Phosphorylation of a serine residue of the H3 histone at position 10 (H3S10p) is the most common histone modification with a direct link to meiosis and mitosis in both plants and animals (Wei et al., 1998;Houben et al., 1999;Kaszas and Cande, 2000;Manzanero et al., 2000;Paula et al., 2013).In animals, this mark is essential for the condensation of meiotic chromosomes and their segregation, because experiments show that artificially reduced H3S10 phosphorylation levels lead to meiotic problems and sterile phenotypes (Wei et al., 1999).In plants, the mitotic H3S10p chromatin mark is localized in (peri)centromeric regions of chromosomes, and in the first meiotic division the mark tends to occupy the whole chromosome (Manzanero et al., 2000).More variable patterns have been reported in the second meiotic division, when histone phosphorylation levels are generally reduced compared with those in the first meiotic division, and differences are likely to occur in the function of H3S10p between animals and plants.The other phosphorylated amino acids of histone H3 linked with nuclear division are the serine at position 28 (H3S28p) and threonine at position 3 (H3T3p) (Houben et al., 2007;Sawicka and Seiser, 2012).As demonstrated in Secale cereale (rye), the H3S10p and H3S28p sites frequently co-localize and are likely to influence sister chromatid cohesion (Manzanero et al., 2000;Gernand et al., 2003).Interestingly, at meiosis II, no phosphorylation of H3S10 and H3S28 occurs on chromatids derived from prematurely separated univalents (Manzanero et al., 2000).Although notable exceptions to this trend exist (Fernandes et al., 2008), it is likely that loss of phosphorylation is needed for proper chromatid pair segregation.
Detailed studies of asymmetrical meiosis in dogroses are scarce, and virtually no reports exist on epigenetic modification of their chromosomes.To address this lack of studies, we analysed meiotic and mitotic chromosomes of R. canina and R. rubiginosa by immunostaining with antibodies recognizing modified residues of histone H3.In addition, transmission of chromosomes during meiosis was analysed by rDNA markers.We found that univalent and bivalent chromosomes acquire differential phosphorylation statuses in the first meiotic division, underlining the role of epigenetics in their differential segregation.
The fluorescence in situ hybridization (FISH) procedures followed protocols described by Lim et al. (2005) and Herklotz et al. (2018).Briefly, anthers were macerated on a slide, squashed in a drop of 70 % acetic acid and fixed by snap freezing in liquid nitrogen.A cloned 1.7 kb fragment of the Solanum lycopersicum 18S rRNA gene (GenBank X51576.1) was used as a probe.The probe was directly labelled by nick translation using Atto647N (Jena Bioscience, Jena, Germany; pseudo-coloured to red in the figures).The hybridization mixture contained 100 ng labelled probe, 50 % (v/v) formamide, 10 % (w/v) dextran sulphate and 2× saline-sodium citrate (SSC) buffer.Chromosomes were counterstained with DAPI and hybridized against the probe overnight at 37 °C.The posthybridization washing steps were carried out in a water bath (42 °C) in the dark.The slides were washed twice in 2× SSC for 5 min (stringency ~60 %), followed by a more stringent wash in 0.1× SSC (stringency ~82 %) twice for 5 min.The slides were then removed from the water bath (still in a dark box), cooled to room temperature and washed in 2× SSC for 5 min, followed by 4× SSC with 0.1 % Tween 20 for 7 min and a final brief wash in 1× PBS.
All slides were examined under an Olympus Provis AX70 epifluorescence microscope (Metasystems, Germany).The imaging software used was ISIS (MetaSystems), and images were optimized for contrast and brightness with Adobe Photoshop PS2021.

Immunostaining procedures
Histone modification marks were immunolocalized according to several combined protocols (Houben et al., 2003;Demidov et al., 2014).Anthers from young buds were dissected and fixed for 20-25 min in freshly prepared 4 % (w/v) paraformaldehyde with 0.1 % (w/v) Triton X-100 in 1× phosphate-buffered saline (1× PBS, pH 6), followed by additional incubation on ice for 20 min, if necessary.The anthers were washed three times for 15 min in an ice-cold mixture containing 2 % (w/v) PVP-40 and 0.5 % (v/v) Triton X-100 diluted in 1× PBS and three times for 5 min in ice-cold 1× PBS.Tissues were then digested with a mixture of enzymes containing 2.5 % (w/v) hemi-cellulase (Sigma-Aldrich), 2.5 % (w/v) cellulase Onozuka R-10 (Serva), 2.5 % (w/v) macerozyme (Duchefa) and 1.0 % (w/v) pectolyase Y-23 (Sigma-Aldrich) dissolved in 1× PBS or 1× microtubule stabilizing buffer (for microtubule detection) overnight at 4 °C.Enzymes were then removed, and tissue was disrupted by gentle tapping with a metal rod to generate a suspension of cells.The mixture was filtered through a 30 µm mesh filter.To reduce polyphenolic contaminants of microscopic preparations, the anther mash was incubated in an ice-cold solution of 2 % PVP-40 and 0.5 % Triton X-100 in 1× PBS for two rounds of 5 min each in a rotating shaker, followed by washes with 1× PBS for three rounds of 5 min each.For microtubule detection, we used 1× microtubule stabilizing buffer as an incubation buffer.After washing steps, the meiocyte suspension was placed on poly-l-lysine-coated slides (Sigma-Aldrich) and squashed between a glass slide and coverslip.After freezing in liquid nitrogen, the coverslip was removed, and slides were transferred immediately into 1× PBS.For immunostaining, the slides were incubated in blocking buffer [1× PBS with 5 % (w/v) bovine serum albumin and 0.1 % (v/v) Triton X-100] in a prewarmed humid chamber (for immunochemistry) at 37 °C for 60 min.Primary antibodies were added to slides and incubated in a humid chamber at 4 °C overnight, followed by subsequent washes (three times for 5 min in 1× PBS with 0.1 % Tween 20 on ice).The slides were incubated with secondary antibodies for 2 h at 37 °C.Slides were washed three times in 1× PBS for 5 min each and subsequently dehydrated in an ethanol series (70, 90 and 100 %).The slides were air dried in a dark box, mounted in Vectashield containing DAPI and observed under a microscope.All primary and secondary antibodies were diluted in 1× PBS, 1 % bovine serum albumin and 0.1 % Tween 20.Working dilutions of primary antibodies are listed in Table 1.The secondary antibodies included anti-rabbit IgG Cy3 conjugate (Jackson Immunoresearch, USA; catalogue number 111167003) and anti-rabbit IgG fluorescein isothiocyanate (FITC) conjugate (Jackson Immunoresearch, USA; catalogue number 111097003), both diluted 1:300.For simultaneous immunostaining, we used a combination of antimouse IgG FITC conjugate (Sigma-Aldrich; catalogue number F0257; 1:200 dilution) and anti-rat IgG Cy3 conjugate (Thermo Fisher Scientific, Invitrogen, USA; catalogue number A10522; 1:1000 dilution) antibodies.
The numerical pixel values of fluorescence intensity were obtained from manually annotated chromosomes after selection for DAPI (blue), FITC (green) and Cy3 (red) fluorescence filters.The mean grey value was calculated as the sum of the grey values of all the pixels in the selection divided by the total number of pixels.Box plots were constructed in RStudio (BoxPlotR, 2013; R Development Core Team, 2013) Mann-Whitney non-parametric tests were performed using an online calculator (StatisticsKingdom, 2017).A total of 639 meiocytes from several R. canina individuals and one R. rubiginosa individual were examined with the antibodies listed in Table 1.A summary of the samples and analyses is given in Supplementary Data Table S1.

Viability staining of pollen grains
For pollen vitality tests, mature pollen grains were collected from fresh yellow anthers of closed flowers, air dried for 5 days and stored in reaction tubes until use.To estimate the vitality of pollen grains, we followed the protocol of Alexander (1969).Briefly, the staining reagent (1 mL) contained 10 µL of Malachite Green, 50 µL of Acid Fuchsin and 5 µL of Orange G diluted in 95 µL of absolute ethanol (Sigma-Aldrich Chemie GmbH, Taufkirchen, Germany), 250 µL of glycerol, 40 µL of acetic acid and 550 µL of distilled water.Pollen grains were incubated with the reagent in rotating collection tubes for 1 h and counted in a Burker chamber.

The course of canina meiosis as analysed by FISH
Fluorescence in situ hybridization in combination with DAPI staining of chromatin was used to identify chromosomes at different stages of meiosis in R. canina and R. rubiginosa.In diakinesis, all seven homologous chromosome pairs (forming bivalents) resembled thick rods along their entire lengths (Fig. 2A, top preparations, asterisks).In contrast, most of the 21 unpaired chromosomes (univalents) tended to form thin fibres exhibiting relatively weak staining.The 18S rDNA probe hybridized to four sites, two on bivalents and two on univalents.No multivalents were detected in diakinesis.All four 18S rDNA sites were also visible in metaphase I (Fig. 2A, middle panel).In anaphase I, homologous pairs in bivalents separated by moving to the poles as in regular meiosis.In contrast, 21 univalents were separated into 42 chromatids that had delayed migration (Fig. 2A, bottom).A single (corresponding to tightly linked homologues) or a double (corresponding to loosely linked homologues) rDNA signal was visualized in each bivalent sector.In the univalent sector, there were four rDNA signals (arrowheads) derived from a split of chromatid pairs of the two nucleolus organizer regions (NOR)-positive chromosomes.We did not observe cytokinesis between meiosis I and meiosis II.In metaphase II, each of the two daughter nuclei inherited four 18S rDNA sites (two on bivalents and two on univalents) (Fig. 2B, top).In anaphase II, the bivalents separated into chromatids (arrows) and migrated to the poles (Fig. 2B, middle).In contrast, the univalent-derived chromatids (arrowheads) remained between the two meiotic spindles and probably never reached the poles.There were eight 18S rDNA signals indicating separation of chromatids of all NOR chromosomes at this stage.Additionally, at telophase II (Fig. 2B, bottom), eight rDNA signals were visible.The four signals were located in large nuclei with nucleoli probably representing premature microspores, and another four signals were visible in micronuclei.In contrast to Klášterská (1971), we did not observe nucleolus-like bodies in the cytoplasm, and all rDNA loci seemed to associate with chromosomes.Among the univalent chromosomes, one rDNA locus was consistently missing, indicating that one of the subgenomes lacked essential rRNA genes and was no longer able to provide fertile gametes.In other R. canina accessions, all five rDNAs are still present (Lim et al., 2005;Herklotz et al., 2018), although one of them is extremely small, in line with the loss of rDNA copies observed in many allopolyploids (Garcia et al., 2017).In summary, premature separation of univalent chromatid pairs in phase I has been confirmed by rDNA markers in two dogrose lineages of independent origin.Finally, we determined the pollen viability of 5x R. canina, 5x R. rubiginosa and 2x R. rugosa (Supplementary Data Table S2).Pollen was collected from the same population over four consecutive years (2020-2023).The pollen viability of the diploid R. rugosa was high (80-90 %) and relatively stable over the years.In contrast, the viability of pentaploid dogrose pollen ranged from 17 to 50 % in R. canina and from 30 to 45 % in R. rubiginosa.

Histone H3 phosphorylation marks in diakinesis/early metaphase I in canina and non-canina meiosis
Major changes in histone H3 by phosphorylation are expected to occur at the early stages of the first meiotic division (Fuchs et al., 2006).Therefore, we analysed late diakinesis/ early metaphase I nuclei by immunostaining of nuclei with antibodies against phosphorylated H3S10p, H3S28p and H3T3p histones.In R. canina, both bivalents and univalents showed positive staining with all three marks (Fig. 3).The bivalent chromosomes were intensely stained with the H3S10p and H3S28p antibodies along their whole lengths (Fig. 3A, B, arrows).In univalents (arrowheads), most phosphorylation signals were localized on the chromosome arms and in distal regions, while the proximal regions were less intensively stained, particularly with the H3S28p antibody (Fig. 3B).Selected chromosomes are enlarged in the insets in Fig. 3.Staining patterns were reproducible between independent meiotic cells of R. canina (Supplementary Data Fig.S1A, B).
In the diploid R. nitida (2n = 2x = 14), a species with standard meiosis, all 14 chromosomes formed bivalents that were heavily stained with the H3S10p antibody across their whole lengths in metaphase I (Supplementary Data Fig.S3A).
Immunolocalization of α-tubulin and H3S10p at different meiotic stages in the R. canina and R. rubiginosa pentaploids To visualize the relative position of microtubules and chromosomes during canina meiosis, we used immunostaining with antibodies against α-tubulin and H3S10p.Therefore, a more complete set of meiotic stages is shown for R. canina (Fig. 5A-E); meiotic stage II is also shown for R. rubiginosa (Fig. 5F, G).The polar view of R. canina metaphase I chromosomes showed an overlap of H3S10p signals with the α-tubulin signals (Fig. 5A).Likewise, in anaphase I, both bivalent and lagging univalents were associated with α-tubulin (Fig. 5B).In the second meiotic division, the H3S10p signals were clearly visible up to the anaphase II stage (Fig. 5D-F).In contrast to that in anaphase I, the association of chromatids with α-tubulin was not as pronounced in anaphase II, particularly for univalents that remained at the equator (Fig. 5E, F).At telophase II, we observed diffuse H3S10p signals of decondensing chromosomes (Fig. 5G; Supplementary Data Fig.S4A).Fluorescence signals became almost undetectable in polyads (Supplementary Data Fig.S4B).

Histone H3 phosphorylation patterns in somatic chromosomes
To analyse the histone H3 phosphorylation profiles in somatic tissues, we analysed root tip nuclei from R. canina pentaploid and R. nitida diploid plants.The metaphase chromosomes were immunostained with antibodies against H3S10p, H3S28p and H3T3p.In both R. canina (Supplementary Data Fig.S5A, B) and R. nitida (Supplementary Data Fig.S3B), the H3S10p and H3S28p signals were localized preferentially to pericentromeric positions.The H3T3p phosphorylation signals were more dispersed (Supplementary Data Fig.S5C) and appeared to be somewhat stronger at (peri)centromeric regions.

Euchromatic and heterochromatic histone marks on dogrose chromosomes
Next, we wondered whether bivalent and univalent chromosomes differ in euchromatin/heterochromatin landscapes.Therefore, we examined euchromatic H3K4me3 and heterochromatic (facultative) H3K27me3 marks on metaphase I chromosomes (Supplementary Data Fig.S6).In both R. canina (Supplementary Data Fig.S6A) and R. rubiginosa (Supplementary Data Fig.S6B), the H3K4me3 signals were located mostly at interstitial and distal positions, often on both arms.This pattern corresponded to that of H3S28p staining (for comparison, see Fig. 3B).Immunostaining of the H3K27me3 mark produced diffuse patterns with several spot-like signals in distal regions (Supplementary Data Fig.S6C, D).Staining of somatic root tip nuclei with antibodies against the H3K4me3 and H3K27me3 marks resulted mostly in whole chromosome patterns with some local maxima and minima (Supplementary Data Fig.S7).In general, there was no or difference only a small difference in eu-and heterochromatin immunostaining patterns between mitotic and meiotic cells and between univalent and bivalent chromosomes, at least for the two histone H3 marks investigated.

DISCUSSION
In asymmetrical canina meiosis, undivided univalents are transmitted to offspring via the female gamete, but they are eliminated during microsporogenesis and do not occur in mature pollen.The mechanism of univalent chromosome elimination in pollen mother cells has often been overlooked.On the basis of molecular cytogenetics methods, this study provides some insights into the mechanisms of univalent chromosome behaviour during male canina meiosis.

Molecular cytogenetics reveals equational division of univalents in the first meiotic division, followed by their gradual elimination in the second meiotic division
We investigated individual stages of canina meiosis in R. canina and R. rubiginosa (both 2n = 5x = 35).The rDNA FISH marker clearly revealed separation of univalent chromatids in anaphase I (Fig. 2A, bottom).The chromatids migrated as laggers, corroborating the findings of all classical cytological studies (Blackburn and Harrison, 1921;Täckholm, 1922;Klášterská and Natarajan, 1974a;Roberts, 1975).We were unable to detect nondisjunction of any univalent chromosome, indicating that each daughter nucleus receives seven bivalent-derived chromatids and 21 univalent-derived chromatids.Thus, the first meiotic division of canina meiosis includes the reductional division of homologous bivalent chromosomes and equational division of univalent chromosomes.In contrast to Oenothera (Onagraceae), in which some genotypes also exhibit irregular meiosis due to permanent reciprocal translocations (Golczyk et al., 2008), dogroses do not appear to have multivalents in diakinesis, with their occurrence seemingly limited to the earlier prophase stage (Supplementary Data Fig.S8).As in other allopolyploids, these structures might be fixed later or removed by selection (Cifuentes et al., 2010;Soares et al., 2021).In the second meiotic division, former univalents exist in the form of chromatids that cannot be divided further.However, they are still capable of binding microtubules migrating irregularly across the meiotic plate.In anaphase II, we observed irregular counts of prematurely separated chromatids, chromatids located outside the meiotic plate and an uneven distribution of rDNA sites (Figs 2B and 4E,F).These features suggest a near-to-stochastic distribution of univalent chromosomes in meiosis II.We noted variable (~21 ± 5) distributions of chromatids between daughter cells in anaphase II.It is possible that cells with higher chromosome counts falsely lead to the inference of univalent replication between meiosis I and II (Täckholm, 1922).Thus, univalents are eliminated gradually at the terminal stage of male meiosis, probably owing to a cascade of events triggered in its early stages.

Bivalent and univalent chromosomes contain both active and inactive epigenetic marks
In contrast to histone phosphorylation marks, methylation marks have been studied less often in meiotic cells.In our experiments, the H3K4me3 euchromatin marker was abundantly present on chromosomes at diakinesis and early metaphase I (Supplementary Data Fig.S6).Zygotene/pachytene chromosomes were also strongly stained with antibodies against diand trimethylated lysine in tobacco (Mursalimov et al., 2019).These results support the notion that active H3K4me3 marks are not erased during meiotic division and might even be enhanced to some extent.In this context, H3K4me3 chromatin might be associated with meiotic recombination sites, as shown for Saccharomyces cerevisiae (Borde et al., 2009).It remains to be determined whether the observed hypermethylation of H3K4 is correlated with histone acetylation, which has been shown to be enhanced in prophase in some systems (Feitoza et al., 2017), and whether it reflects the transcriptional activity of genes residing in univalent chromosomes.

Reduced histone H3 phosphorylation of univalent chromatin in meiotic phase I
Histone H3 phosphorylation at S10 and S28 in plants is usually found at (peri)centromeric regions, where it functions in mitotic and meiotic sister chromatid cohesion and successful chromosome segregation (Manzanero et al., 2000;Rossetto et al., 2012;Loginova and Silkova, 2017).Strikingly, in the first meiotic division, univalent chromosomes showed only weak phosphorylation levels in these regions (Figs 3 and 4), and most signals were localized to the interstitial and distal parts of chromosomes.This pattern contrasts with that of the bivalents, which showed strong histone phosphorylation signals spreading from (peri)centromeric regions across the entire chromosome length.
How can the unusual staining patterns of univalent chromosomes be explained?Given that phosphorylation is a hallmark of functional (peri)centromeres, it is possible that their centromeres are impaired or degenerated.However, the following observations argue against this hypothesis.First, in root mitotic cells, all chromosomes showed the canonical (peri) centromeric localization of the H3S10p phosphorylation mark (Supplementary Data Fig.S5).Furthermore, there was no evidence for abnormal mitosis (Lim et al., 2005;Herklotz et al., 2018).Second, the (peri)centromeric CANR4 repeat previously found in most Caninae genomes occurs in both univalent and bivalent chromosomes and is more abundant in the former (Lunerová et al., 2020).Hence, the univalent chromosomes seem to harbour functional centromeres, and their abnormal phosphorylation patterns seem to be linked to the first meiotic stage.It will be interesting to analyse H3 phosphorylation in prophase to see whether aberrant phosphorylation is a cause or consequence of pairing failure.
Intriguingly, the reduction in histone H3S10 and H3S28 phosphorylation in (peri)centromeric regions preceded a split of univalent chromatid pairs in the first meiotic division.Centromeres are chromosomal regions that interact with the spindle apparatus during each nuclear division to ensure the disjunction of chromosomes (Talbert and Henikoff, 2020).It has been shown that the reduction in histone H3S10 phosphorylation in these regions is associated with chromosome defects and meiotic abnormalities in animals (Wei et al., 1999).Thus, one of the functions of histone phosphorylation could be to assist sister chromatid cohesion, perhaps through increased chromatin condensation.In comparison to univalents, bivalents were strongly stained with DAPI in diakinesis (Fig. 4) and prophase (Supplementary Data Fig.S8).This finding is consistent with some earlier cytological observations.For example, Erlanson (1933) attributed non-synapsis of univalent chromosomes to their late condensation at prophase I. Later, irregular contraction of chromosomes in the first meiotic division was reported in several Canina species (Klášterská and Natarajan, 1974a, b).It should be mentioned that full decondensation of univalents has never been observed in meiosis, and heterochromatic regions (satellites and rDNA) seem to be condensed normally (Fig. 2; Lunerová et al., 2020).
It might be informative to compare the behaviour of canina chromosomes with that in other systems bearing non-pairing chromosomes in their karyotypes, such as B chromosomes.These supernumerary chromosomes occur in both plants and animals, exhibiting a non-Mendelian mode of transmission, and they do not recombine with autosomes.There are notable differences between both systems.For example, the canina univalents divide equationally in phase I, whereas the B chromosomes, in general, undergo non-disjunction and migrate as dyads to one of the poles (Chiavarino et al., 2000;Chen et al., 2022).Interestingly, canina histone H3S10 phosphorylation is not completely erased in the second meiotic division (Fig. 5E, F), indicating that separation of chromatid pairs does not lead to complete loss of H3S10p chromatin marks.This contrasts with the situation in maize and rye (Manzareno et al., 2000), where the B chromosomes lack H3S10 phosphorylation completely.The immunostaining patterns of canina meiotic chromosomes somewhat resemble those of the grasshopper Eyprepocnemis plorans, whose univalent X-chromosome retains H3 phosphorylation throughout the meiotic cycle.Of note, similar to R. canina, E. plorans exhibits equationally dividing univalents at the first division (Rebollo and Arana, 1995).

A hypothetical model of chromosome behaviour in canina male meiosis
It is well established that the cohesin complex plays a central role in the successful transmission of chromosomes through meiosis (Bolanos-Villegas, 2021).The complex includes meiosis-specific subunits, such as the meiotic α-kleisin REC8/ SYN1 (Castellano-Pozo et al., 2020), which are deposited on chromatin during leptotene.The synaptonemal complex (SC) is then assembled fully between paired homologous chromosomes at pachytene.Centromeric cohesion is protected by the conserved SUGOSHIN (SGO1/2 in Arabidopsis) family of proteins until anaphase II (Cromer et al., 2013).Based on these suppositions, a hypothetical model of chromosome behaviour in canina meiosis is proposed in Fig. 6.In anaphase, there is a balance between opposing forces, a mitotic spindle on the one hand and cohesins holding sister centromeres on the other.However, the loading of cohesins onto the chromosomes could be reduced or even fail in dogrose univalents, which are structurally heterogeneous and lack homologous partners (V.Heklotz (Senckenberg Museum of Natural History, Senckenberg, Germany), J. Lunerová (Czech Academy of Sciences, Czechia), A. Marquez (MPI, Germany), C.M. Ritz (Senckenberg Museum of Natural History, Senckenberg, Germany), unpubl.res.).Thus, in univalents, a repulsion force of the mitotic spindle dominates over the cohesion force, perhaps as a result of aberrant chromatin modifications and cohesin insufficiencies.This apparently leads to their equation division in anaphase I.In contrast, the cohesive force associated with hyperphosphorylated bivalent chromatin keeps their chromatids together.Certainly, in the second meiotic division, bivalent chromosomes gradually lose phosphorylation, degrade residual cohesins and, as in other systems, divide equationally.

Conclusions
We found differences in the epigenetic histone H3 phosphorylation patterns between pairing and non-pairing chromosomes during canina asymmetrical meiosis.We hypothesize that reduced H3 phosphorylation levels of univalent chromosomes, particularly in pericentromeric regions, could destabilize the cohesion of chromatid pairs, leading to their premature separation in the first meiotic division.

Fig. 2 .Fig. 3 .Fig. 4 .
Fig. 2. Fluorescence in situ hybridization of Rosa canina and Rosa rubiginosa meiotic chromosomes.Sites of 18S rDNA probe hybridization are shown in red.Arrows and arrowheads mark positions of rDNA loci on bivalent and univalent chromosomes, respectively.(A) First meiotic division.(B) Second meiotic division.In diakinesis, bivalent chromosomes are marked by asterisks.In metaphase, individual chromosome types could not be resolved.Note the separation of univalentderived chromatids in anaphase I (equatorial position).Note that only a few univalents were retained in R. rubiginosa anaphase II.Note the newly formed nucleoli surrounded by rDNA (arrows) in telophase II; micronuclei containing univalent loci (arrowheads) lacked nucleoli.Scale bar: 10 µm.

Fig. 5 .
Fig. 5. Immunolocalization of histone H3S10p marks (green) and α-tubulin (red) on chromosomes during the first and second meiotic divisions.(A-E) Rosa canina.(F, G) Rosa rubiginosa.Note the association of α-tubulin with both bivalent (leading) and univalent (lagging) chromosomes in anaphase I (B) and to a much lesser extent in anaphase II (E, F).Note the relatively strong H3S10p signals in metaphase chromosomes (A, D).The bivalents, bivalent-derived chromatids and major tetrad cells are depicted by arrows.Scale bars: 10 µm.
Fig. 6.A hypothetical model showing the fate of univalents (A) and bivalents (B) in canina meiosis.A role of histone H3 phosphorylation (S10, S28) in chromatid separation is proposed during the metaphase/anaphase transitions.Dark green circles indicate highly phosphorylated (peri)centromeric chromatin.Light green circles indicate no or weakly phosphorylated (peri)centromeric chromatin.Black arrows indicate cohesive force maintaining the integrity of chromatids.Blue arrows indicate repulsive force of the kinetochore.

Table 1 .
List of antibodies and their targets used in cytogenetic analyses.