When sexual meets apomict: genome size, ploidy level and reproductive mode variation of Sorbus aria s.l. and S. austriaca (Rosaceae) in Bosnia and Herzegovina

Abstract

Background and Aims Allopolyploidy and intraspecific heteroploid crosses are associated, in certain groups, with changes in the mating system. The genus Sorbus represents an appropriate model to study the relationships between ploidy and reproductive mode variations. Diploid S. aria and tetraploid apomictic S. austriaca were screened for ploidy and mating system variations within pure and sympatric populations in order to gain insights into their putative causalities.

Methods Flow cytometry was used to assess genome size and ploidy level among 380 S. aria s.l. and S. austriaca individuals from Bosnia and Herzegovina, with 303 single-seed flow cytometric seed screenings being performed to identify their mating system. Pollen viability and seed set were also determined.

Key Results Flow cytometry confirmed the presence of di-, tri- and tetraploid cytotype mixtures in mixed-ploidy populations of S. aria and S. austriaca. No ploidy variation was detected in single-species populations. Diploid S. aria mother plants always produced sexually originated seeds, whereas tetraploid S. austriaca as well as triploid S. aria were obligate apomicts. Tetraploid S. aria preserved sexuality in a low portion of plants. A tendency towards a balanced 2m : 1p parental genome contribution to the endosperm was shared by diploids and tetraploids, regardless of their sexual or asexual origin. In contrast, most triploids apparently tolerated endosperm imbalance.

Conclusions Coexistence of apomictic tetraploids and sexual diploids drives the production of novel polyploid cytotypes with predominantly apomictic reproductive modes. The data suggest that processes governing cytotype diversity and mating system variation in Sorbus from Bosnia and Herzegovina are probably parallel to those in other diversity hotspots of this genus. The results represent a solid contribution to knowledge of the reproduction of Sorbus and will inform future investigations of the molecular and genetic mechanisms involved in triggering and regulating cytotype diversity and alteration of reproductive modes.

INTRODUCTION

Hybridization coupled with polyploidization (allopolyploidy) represent determinant occurrences shaping plant diversity (Hegarty and Hiscock, 2008; Soltis and Soltis, 2009). Recurrent formation and recombination of allopolyploid genomes may profoundly recompose genome structure, alter gene expression and induce phenotypic changes (Soltis and Soltis, 2009; Ďurkovič et al., 2012). In certain groups, hybridization and polyploidy are associated with changes in the mating system, following a familiar direction from sexuality to asexuality (Ozias-Akins and van Dijk, 2007; Potter et al., 2007; Talent and Dickinson, 2007a, b; Cosendai and Hörandl, 2010; Hojsgaard et al., 2014). Polyploidy is related to a breakdown of self-incompatibility, allowing uniparental reproduction and apomixis (Mable, 2004; Hörandl, 2010; Ludwig et al., 2013).

Apomixis (asexual seed formation), seen as an evolutionary mechanism, provides a way to preserve and maintain particular hybrid and heterozygous genotypes along with genomes having unbalanced numbers of chromosomes, allowing their long-term propagation and persistence (Hörandl, 2006; Ozias-Akins and van Dijk, 2007). Such uniparental reproduction is seen as having advantages in the colonization of disturbed areas and ecologically unfavourable habitats with respect to outcrossing (Hörandl, 2006; Paun et al., 2006). The exact molecular and genetic mechanisms triggering and regulating apomixis are still not fully understood; such research is intriguing from both the developmental and the evolutionary perspective (Hand and Koltunow, 2014).

In the genus Sorbus, apomixis is a powerful factor in population structuring and divergence. European members of the genus Sorbus include sexual diploid species and numerous polyploids, predominantly apomictic taxa that are concentrated in three diversity hotspots, i.e. Scandinavia, Britain and south-eastern Europe (Jankun, 1993; Warburg and Kárpáti, 1993; Rich et al., 2010; Robertson et al., 2010). Prevalence of apomictic polyploids across different parts of the European continent appears to have been initially generated by interspecific hybridization and backcrosses of four diploid sexual species: S. aria, S. aucuparia, S. torminalis and S. chamaemespilus (Liljefors, 1953, 1955; Warburg and Kárpáti, 1993). This hypothesis implies that interspecific hybridization events have been linked, at the early stages of divergence of the genus, with unreduced gametes in diploids that produced heteroploid progeny, which subsequently participated in further hybridization events.

Diploid Sorbus species are outcrossing and self-incompatible, while polyploids are mostly apomictic, thus at least in part reproductively isolated (facultative apomixis) in relation to their parents. These findings have been achieved by embryological (Liljefors, 1953; Jankun and Kovanda, 1986, 1987, 1988), isozymatic (Proctor et al., 1989) and molecular (Nelson-Jones et al., 2002; Chester et al., 2007; Lepší et al., 2008, 2009, 2013; Kamm et al., 2009; Robertson et al., 2010; Ludwig et al., 2013) studies. Early embryological observations provided fundamental information on the type of embryo sac, sexual and asexual pathways of embryo development, pseudogamy, mono- and dispermy of central cell nuclei, as well as disturbed meiosis in hybrids and individuals having odd ploidies. Such investigations have provided an invaluable basis and have been most useful for later interpretation of cytometric data to infer sexual and asexual seed origin along with putative pathways of endosperm formation in different plant groups, including Crataegus, which was among the first rosaceous genera studied using cytometric seed screening (Talent and Dickinson, 2007a, b). Molecular markers, particularly microsatellites, combined with ploidy data have provided plenty of important evolutionary information on Sorbus, such as mating system discrimination, parentage identification, complex relationship reconstruction and genetic diversity (Chester et al., 2007; Lepší et al., 2008, 2009, 2013; Robertson et al., 2010; Ludwig et al., 2013).

Variations in reproduction mode, an extremely important aspect in Sorbus evolution, have not been evaluated by flow cytometry. Flow cytometric seed screening (FCSS) has become a prevailing method of analysing reproduction modes, providing reliability, accuracy and speediness with relatively low costs (Doležel et al., 2007 and references therein). Flow cytometric seed screening has been applied in some rosaceous genera [Amelanchier (Burgess et al., 2014), Crataegus (Talent and Dickinson, 2007a), Potentilla (Dobeš et al., 2013), Rubus (Šarhanová et al., 2012)], shedding new light on their mating systems. Data on embryo and endosperm ploidies allow the discrimination of sexual from asexual seed origins and the inference of ploidies of contributing gametes and pathways of endosperm formation. All such information is of outstanding relevance for Sorbus, in which the dynamics of ploidy fluctuation is one of the best indicators of ongoing diversification, especially by comparing ploidy data of progenitors with data of their derived progeny. In contrast to molecular markers, information on ploidy levels does not facilitate distinguishing auto- from allopolyploidy nor resolving the exact parentage of hybrid derivatives.

Diploid Sorbus are characterized by having an eight-nucleated Polygonum-type embryo sac that follows the most common pattern of sexual reproduction, resulting in a diploid embryo and a triploid endosperm (Jankun and Kovanda, 1986). In Sorbus polyploid apomicts, megagametophytes are unreduced due to the bypassing of meiosis (apomeiosis), meaning that the embryo develops from the unreduced egg (parthenogenesis), while the unreduced central cell requires fertilization (pseudogamy) for normal endosperm development (Liljefors, 1953; Jankun and Kovanda, 1986, 1987; Ludwig et al., 2013). The successful establishment of apomictic polyploids is highly dependent on their adaptive potential to overcome the problem of endosperm imbalance. For normal endosperm development, flowering plants usually require a balanced contribution of maternal and paternal genomes (2 m : 1p) (Scott, 2007) as a necessary precondition for imprinted gene expression (Haig and Westoby, 1991; Vinkenoog et al., 2003). An overdose of either parental genome could result in endosperm failure and finally seed abortion (Scott et al., 1998). Due to the bypassing of meiosis, fusion of unreduced central cell polar nuclei results in high ploidies of the maternal genome that contravene the balanced ratio in the endosperm. However, pseudogamous apomicts develop mechanisms such as changes in megagametophyte structure and modifications of fertilization behaviour to cope with deviations from the 2 m : 1p balance (Scott, 2007), successfully yielding normally developed seeds. One of the solutions for achieving endosperm balance in apomicts lies in the availability of one spare sperm, leading to dispermy or fertilization of the central cell’s nuclei with both sperms (Talent and Dickinson, 2007a; Šarhanová et al., 2012; Ludwig et al., 2013; Burgess et al., 2014). Sorbus polyploids show different mating strategies for endosperm formation as a prerequisite for successful apomictic seed production (Ludwig et al., 2013). Dispermy of the central cell, as a way to overcome the problem of endosperm imbalance, was documented for tetraploid Sorbus species in the embryological study of Jankun and Kovanda (1988). Sorbus tetraploids are fully self-compatible and capable of self-pollination while triploid apomicts are self-incompatible and require pollen from other related Sorbus taxa (Ludwig et al., 2013). Male lineages regularly produce fertile pollen (Rich, 2009); apomicts can therefore exchange genetic material via pollen and consequently transfer apomictic genes to their sexual relatives (Whitton et al., 2008; Cosendai and Hörandl, 2010).

A critical point in Sorbus divergence is the encounter between sexual diploids and apomictic polyploids, usually resulting in heteroploid offspring due to apomicts’ polyploidy. The involvement of newly derived offspring continuously drives, among different related sexual and asexual lineages, a reticulation in future hybridization events that produces novel diversity (Robertson et al., 2010). Hence, apomictic polyploids play a double role in the formation of particularly complex patterns of variation within Sorbus: they preserve particular variation and initiate interploid crosses.

Classical models of polyploid formation require either unreduced gametes in diploid crosses or the exchange of reduced gametes between diploids and tetraploids (Ramsey and Schemske, 1998). Although triploids are considered to be the main link in the formation of tetraploids via the triploid bridge (Nelson-Jones et al., 2002; Robertson et al., 2004a; Talent and Dickinson, 2007a, b), recent studies have identified tetraploids as the key link in the formation of Sorbus triploids (Ludwig et al., 2013).

Owing to extensive recent work on European Sorbus species, genome size, ploidy level and chromosome counts have been determined for most of them (Aldasoro et al., 1998; Bailey et al., 2008; Lepší et al., 2008, 2009, 2013; Siljak-Yakovlev et al., 2010; Pellicer et al., 2012). In general, the most frequent ploidy levels for Sorbus taxa are di-, tri- and tetraploidy, while pentaploids are extremely rare (Bailey et al., 2008; Pellicer et al., 2012).

A recent study on Sorbus in Bosnia and Herzegovina detected a mixture of S. aria-like cytotypes in coexistence with S. austriaca, representing novel diversity in the Balkans (Hajrudinović, 2012), which motivated a more detailed cytometric study of both species. Sorbus aria s.l. (subgenus Aria) is an aggregate of related taxa exhibiting a series of ploidy levels: diploid (2n = 2x = 34), triploid (2n = 3x = 51) and tetraploid (2n = 4x = 68) (Bailey et al., 2008; Rich et al., 2010; Siljak-Yakovlev et al., 2010; Pellicer et al., 2012). It is the most variable and taxonomically intriguing species aggregate of the genus, being involved in most interspecific hybridization events across Europe (Robertson et al., 2010). According to recent taxonomic treatment, S. aria s.s. represents a sexual diploid species while morphologically distinctive polyploids are designated as apomictic species (Rich et al., 2010). On the other hand, S. austriaca (subgenus Soraria) is considered to be a stable apomictic allotetraploid species (2n = 4x = 68) (Kovanda, 1986); it is relatively widespread, growing in the Balkans, Carpathians and eastern Alps (Warburg and Kárpáti, 1993) and contains genomes of S. aucuparia and S. aria s.l. (Nelson-Jones et al., 2002).

The working hypothesis considered was that the presence of apomicts in sexual populations of Sorbus leads to the generation of novel polyploid cytotypes, which deeply impact population structuring via apomixis. In this study we carried out a cytometric survey to assess: (1) the range of genome size and ploidy variation in pure and sympatric populations of the studied species; (2) the increase or decrease in progeny ploidy in relation to the parental generation; (3) the frequency of sexuality in polyploids; and (4) the most frequent endosperm formation pathways among sampled polyploids and their requirements for endosperm balance. The panel comprised seven pure populations of diploid S. aria, one of tetraploid S. austriaca and four of their sympatric populations. We discuss our findings in the context of evolutionary processes governing Sorbus diversification. Also, we advocate that the generation of novel diversity is primarily driven by the interaction of sexual and asexual taxa, an evolutionary model as important for Sorbus in the Balkans as for the other diversity hotspots for this genus in Europe.

MATERIALS AND METHODS

Plant material

Flow cytometry was done on leaves of 380 Sorbus individuals from 12 sites and 303 seeds of 80 Sorbus adult individuals from nine natural sympatric and pure populations encompassing each taxon/cytotype group (Table 1, Fig. 1). The number of sampled individuals ranged from 6 to 105 per population (Table 1, Fig. 1). More individuals were sampled at site 1 (N = 105) and site 3 (N = 52). The reasons for sampling asymmetry were as follows: sites 1 and 3 were initially the research focus as series of ploidy levels in S. aria s.l. had been recorded there for the first time in Bosnia and Herzegovina; our intention was to use certain localities with new Sorbus diversity as a criterion to designate specific areas within the High Conservation Value Forest programme in Bosnia and Herzegovina; the convenient geographical proximity of these sites was conducive to repeated sampling when needed. Sampled sites mostly represent mixed stands of Fagus sylvatica and Abies alba in various stages of degradation (coppices, scrublands) on calcareous or dolomitic rocks. Sites 4 and 8 represent degradation stages of Pinus heldreichii forests on dolomitic soils. Site 2 represents a relic thermophilous community of Quercus petraea and Fraxinus ornus while site 5 belongs to degraded forests of Quercus pubescens. Degraded sites are mostly exposed to the south-east, sunny and dry.

Table 1.

Geographical origin of 12 populations of Sorbus mother trees covering four cytotypes and corresponding sample sizes (N) for genome size, reproductive mode and pollen stainability

Site	Origin of sample				Number of samples (N)
	Locality	Longitude (E)	Latitude (N)	Altitude (m)	S. aria 2x	S. aria 3x	S. aria 4x	S. austriaca	Total N per site
1	Grkarica, Mt Igman	18°17′48″	43°44′21″	1350	8, 35, 7	55, 58, 42	11, 30, 5	31, 42, 26	105, 165, 80
2	Pratača, Mt Igman	18°11′29″	43°45′49″	913	12, 8, 6	–, –, –	–, –, –	–, –, –	12, 8, 6
3	Umoljani, Mt Bjelašnica	18°13′34″	43°39′51″	1300	1, –, 1	9, 1, 7	32, 37, 19	10, 2, 3	52, 40, 30
4	Rujište, Mt Prenj	17°57′05″	43°27′26″	890	6, –, 6	–, –, –	–, –, –	–, –, –	6, –, 6
5	Gradac, Posušje	17°23′33″	43°25′30″	760	25, 6, 6	–, –, –	–, –, –	–, –, –	25, 6, 6
6	Borova glava, Livno	17°07′25″	43°46′52″	1183	39, 13, 15	–, –, –	–, –, –	–, –, –	39, 13, 15
7	Bosiljna, Mt Čvrsnica	17°29′27″	43°30′09″	1319	10, 24, –	14, 12, –	1, 2, –	4, –, –	29, 38, –
8	Blidinje	17°30′26″	43°35′59″	1220	7, –, 1	1, –, –	3, 5, 1	14, –, 3	25, 5, 5
9	Kamenica, Mt Oštrelj	16°24′38″	44°28′30″	1210	12, –, 5	–, –, –	–, –, –	–, –, –	12, –, 5
10	Mt Slovinj	16°57′22″	43°59′10″	1380	41, 13, –	–, –, –	–, –, –	–, –, –	41, 13, –
11	Zavaline, Drvar	16°20′34″	44°20′32″	820	9, –, –	–, –, –	–, –, –	–, –, –	9, –, –
12	Vrba	18°34′06″	43°10′52″	1090	–, –, –	–, –, –	–, –, –	25, 15, 5	25, 15, 5
	Total N per cytotype				170, 99, 47	79, 71, 49	47, 74, 25	84, 59, 37	380, 303, 158

Fig. 1.

Geographical distribution of cytotypes and their frequency in each site (% relative to N = number of Sorbus individuals analysed), for 12 populations in Bosnia and Herzegovina. Site numbers correspond to those in Table 1.

We labelled each individual at each site due to the need for repeated sampling in different vegetation periods. Pollen stainability analysis covered only part of the study population since some trees did not flower during the sampling period (Table 1).

For the purpose of this study we treated S. aria individuals having different ploidy levels as cytotype groups due to their complicated taxonomy. Vouchers of analysed individuals are deposited in the Herbarium of the National Museum of Bosnia and Herzegovina (SARA).

Genome size and ploidy level determination

Genome size determination followed the basic protocol described by Marie and Brown (1993). Parts of Sorbus fresh leaves (∼1 cm²) were chopped simultaneously with leaf material of internal standard [Solanum lycopersicum ‘Montfavet 63-5’, 2C = 1·99 pg (Lepers-Andrzejewski et al., 2011) or Salvia brachyodon, 2C = 0·95 pg (Maksimović et al., 2007)] using a razor blade in Petri dishes in 600 µL of cold Gif nuclear buffer [45 mm MgCl₂, 30 mm sodium citrate, 60 mm 4-morpholinepropane sulphonate pH 7, 0·1 % (w/v) Triton X-100, 1 % polyvinylpyrrolidone (M_r∼10 000; Sigma P6755), 5 mm sodium metabisulphite]. The suspension was filtered through a 50-µm nylon mesh (CellTrics, Partec) and RNAse (Roche) was added to 2·5 U mL^–1. Nuclei were then stained with propidium iodide (Sigma-Aldrich), with a final concentration of 50 µg mL^–1. This preparation was incubated on ice for 15–20 min to obtain full saturation of nuclei with propidium iodide before analysis. The fluorescence of 5000 nuclei was recorded for each sample using a Partec CyFlow SL3 (Partec, Münster, Germany) 532-nm laser cytometer. Fluorescence histograms were analysed using FloMax ver. 2·8 (Partec, Münster, Germany). A linear relationship between the fluorescence signals of the unknown sample and known internal standards was used to calculate the absolute 2C DNA values. Individual DNA ploidy levels (Suda et al., 2006) were inferred following earlier chromosome counts on Sorbus species and compared with obtained 2C DNA values (Siljak-Yakovlev et al., 2010). Each individual was assigned to a particular taxon/cytotype group of di-, tri- and tetraploid S. aria and tetraploid S. austriaca. Monoploid values of individual cytotypes were calculated using the total 2C DNA value/ploidy level. The mean values of monoploid genome size of cytotype groups were tested using analysis of variance (ANOVA) followed by a post hoc Scheffé’s test. Prior to ANOVA, the homogeneity of group variances was checked using Levene’s test and data distribution was checked using the Kolmogorov–Smirnov test. Analyses were carried out in SPSS ver. 20 (IBM, Armonk, NY, USA).

Flow cytometric seed screen

Flow cytometric seed screening (Matzk et al., 2000) was successfully conducted on 303 seeds collected from attached fruit on 65 S. aria s.l. and 15 S. austriaca mother trees, previously cytotyped (Table 1, Supplementary Data Table S1). Only well-formed seeds, dried for 20–30 h at room temperature and kept in paper bags at 4 °C prior to analysis, were used. Each seed was analysed separately. Entire seeds were co-chopped with part of a fresh leaf (0·5 cm²) of internal standard Oryza sativa ssp. japonica ‘Nipponbare’ [2C = 0·9 pg (Uozu et al., 1997)] in cold Gif nuclear buffer. (This internal standard with nuclei below the 2C level of the Sorbus spp. has the advantage of avoiding any confusion with the numerous nuclear DNA levels occurring in some seed samples.) Prior to this main analysis, seeds with seed coats removed were also cytotyped as a trial but no differences were observed in fluorescence peaks compared with entire seeds, except in one case in which the seed coat, which was formed from maternal tissue, showed a trace of diploid nuclei while the embryo was triploid (Supplementary Data Fig. S1). The rest of the procedure followed the steps described for genome size and ploidy level determination, but taking both linear and log histograms after the fluorescence had been divided between two photomultipliers with a 50/50 mirror. (Because of the laser beam point focus, the fluorescence signal can be partly clipped on this 532-nm cytometer, so a correction factor must be used to interpolate values for large nuclei, namely 2–10 % depending upon their size. This factor was assessed experimentally by analysing endoreplicated foliar nuclei from Phalaenopsis spp., shown to be perfectly 2, 4, 8 and 16C on another cytometer.) Endosperm ploidy was calculated using the inferred monoploid DNA amount of the embryo of the same seed. Estimated DNA ploidies of embryo and endosperm were compared to distinguish between sexual and apomictic origin for each analysed seed and to propose fertilization pathways according to Talent and Dickinson (2007a, b), Hörandl et al. (2008), Cosendai and Hörandl (2010), Šarhanová et al. (2012) and Burgess et al. (2014), all dealing with plant groups having the same Polygonum-type of embryo sac as Sorbus (Liljefors, 1953).

Pollen viability

Pollen viability estimation was performed using Alexander’s stain (Alexander, 1969). Pollen walls stain green and the cytoplasm red so potentially viable pollen grains can be easily identified. Anthers with fresh pollen grains from different flowers of each analysed individual were gently pressed into a drop of Alexander’s stain on a microscopic slide and left with a coverslip overnight at room temperature. One hundred pollen grains were assessed for each of 158 individuals analysed (Table 1).

Seed set

Potential fertility of cytotypes was examined by seed counting. The seeds were collected from the same 80 individuals as for FCSS. Well-formed, firm seeds fully filled with white embryo and endosperm were treated as viable while small, thin and empty seeds, or seeds containing a deformed grey embryo were regarded as sterile. Ten ripe fruits were sampled from each tree. Seeds were extracted and cut in half and the well-formed ones were counted for each fruit and scored as the mean value of an individual. These individual values were used to calculate the mean of well-formed seeds for each taxon/cytotype group.

RESULTS

Nuclear DNA content and ploidy variation

Analysis of nuclear DNA content yielded stable histograms and high-resolution peaks, providing accurate estimation with low coefficients of variation (CVs; mean CV = 3·1 %). Measured on different days, mean CV values among individuals of the same S. aria cytotype and S. austriaca ranged from 2·4 to 3·9 %. Nuclear DNA content averages ranged from 1·39 pg in diploid S. aria to 2·72 pg in tetraploid S. austriaca (Table 2). Such DNA content variation was related to the presence of di-, tri- and tetraploid levels in our sample for S. aria and only tetraploids for S. austriaca (Table 2). The ratios between fluorescence peaks of samples and the internal standards were used to deduce different ploidy levels. The values ranged from 1·29 to 1·48 pg for the diploid S. aria group, from 1·79 to 2·20 pg for triploids and from 2·51 to 2·83 pg for tetraploids. Sorbus austriaca had values ranging from 2·52 to 2·94 pg. Mean values of monoploid genome size [1Cx (Greilhuber et al., 2005)] ranged from 0·67 to 0·70 pg (Table 2). Diploid S. aria had the highest 1Cx DNA values while triploid S. aria had the smallest. ANOVA showed significant differences among the groups for monoploid genome size (F_3,376 =40·512, P < 0·001). Scheffé's test revealed differences between diploids and all other groups (P < 0·01), while no differences were found among polyploids (Table 2).

Table 2.

DNA content of S. aria s.l. and S. austriaca leaf nuclei and deduced ploidy

Species	DNA ploidy	N	2C DNA		CV (%)	1Cx DNA
			Mean ± s.d. (pg)	Min–max (pg)		Mean ± s.d. (pg)	Mean (Mbpa)
S. aria	2x	170	1·392 ± 0·034	1·293–1·476	2·4	0·696 ± 0·017**	681
	3x	79	2·010 ± 0·071	1·787–2·200	3·5	0·670 ± 0·024	655
	4x	47	2·696 ± 0·082	2·513–2·833	3·0	0·674 ± 0·020	659
S. austriaca	4x	84	2·722 ± 0·075	2·520–2·944	2·8	0·680 ± 0·019	665
Total N		380

^a1 pg = 978 Mbp (Doležel et al., 2003).

**Significant difference at P ≤ 0·01.

Four out of 12 sampled populations were sympatric (sites 1, 3, 7 and 8) and comprised S. austriaca and all three S. aria cytotypes (Table 1). The distribution of cytotypes was different at each of these sites (Fig. 1). The more extensively sampled sites 1 and 3 clearly had fewer diploid than polyploid S. aria individuals. Triploid S. aria was the prevailing cytotype at sites 1 (52 %) and 7 (48 %), tetraploid S. aria at site 3 (62 %) and S. austriaca at site 8 (56 %) (Fig. 1). The other sampled populations were pure species having either diploid S. aria or S. austriaca.

Embryo and endosperm ploidies and origin of seeds

Here, FCSS was used for detailed analysis of Sorbus ploidy variations in relation to mother plants and reproductive modes. Only clearly interpretable results are included in the paper, i.e. results for seeds yielding only one fluorescence peak or having a second peak doubling the first peak (e.g. nuclei in G2 phase or trace endoreduplications of the embryo) were not considered. More than one-third of seeds had minor supernumerary peaks corresponding to doubled nuclear DNA levels mirroring embryo and/or endosperm nuclei. Again, these are trace populations of G2-differentiated or endoreduplicated nuclei. Different ploidies of embryos and endosperms reflected different pathways of seed origin (Table 3). At least two different FCSS profiles were determined in each taxon/cytotype group (Table 3, Fig. 2). The results are coherent with either a reduced or unreduced binucleate central cell being fertilized with either one or two sperm cells.

Table 3.

Flow cytometric results and hypothesized pathways of seed formation and mode of reproduction in analysed Sorbus taxa/cytotypes

Mother species	Cytometric results						Hypothesized pathways of seed formation
	Maternal ploidy*	Embryo mean ± s.d. (pg)	Endosperm mean ± s.d. (pg)	Embryo ploidy	Endosperm ploidy	Number of seeds	Egg ploidy	Polar nuclei ploidy/number	Sperm ploidy	Number of sperm fecundating the endosperm	Endosperm maternal :paternal genome ratio	Endosperm/embryo fluorescence (mean ratio)	Seed origin
S. aria	2x	1·39 ± 0·06	2·11 ± 0·12	2x	3x	88	1x	1x/2	1x	1	2 : 1	1·5	Sexual
	2x	2·10 ± 0·11	2·82 ± 0·13	3x	4x	11**	1x	1x/2	2x	1	1 : 1	1·3	Sexual
	3x	2·11 ± 0·10	5·81 ± 0·23	3x	8x	25	3x	3x/2	1x or 2x	2 or 1	3 : 1	2·6	Apomictic
	3x	2·11 ± 0·09	6·37 ± 0·40	3x	9x	18	3x	3x/2	3x	1	2 : 1	2·8	Apomictic
	3x	2·12 ± 0·07	7·16 ± 0·19	3x	10x	21	3x	3x/2	2x	2	1·5 : 1	3·1	Apomictic
	3x	2·15 ± 0·17	7·67 ± 0·56	3x	11x	6	3x	3x/3	1x or 2x	2 or 1	4·5 : 1	3·4	Apomictic
	3x	2·93 ± 0·13	5·88 ± 0·14	4x	–	1**	3x	–	1x	–	–	–	Sexual
	4x	2·84 ± 0·05	4·22 ± 0·03	4x	6x	3	2x	2x/2	2x	1	2 : 1	1·5	Sexual
	4x	2·92 ± 0·23	8·12 ± 0·83	4x	11x	6	4x	4x/2	3x	1	2·7 : 1	2·7	Apomictic
	4x	2·80 ± 0·09	8·40 ± 0·26	4x	12x	64	4x	4x/2	2x	2	2 : 1	2·8	Apomictic
	4x	2·89 ± 0·00	10·16 ± 0·00	4x	14x	1	4x	4x/3	2x	2	3 : 1	3·5	Apomictic
S. austriaca	4x	2·92 ± 0·05	7·17 ± 0·16	4x	10x	3	4x	4x/2	1x or 2x	2 or 1	4 : 1	2·5	Apomictic
	4x	2·85 ± 0·13	7·86 ± 0·43	4x	11x	5	4x	4x/2	3x	1	2·7 : 1	2·6	Apomictic
	4x	2·75 ± 0·06	8·31 ± 0·22	4x	12x	51	4x	4x/2	2x	2	2 : 1	2·8	Apomictic
Total seeds						303

*Data from cytometric leaf measurements.

**Embryo ploidy was increased compared with the mother’s ploidy.

Fig. 2.

Flow cytometric seed screen histograms (log abscissa). Histograms of Sorbus seeds originated sexually (A–D) and apomictically (E–L). The first fluorescence peak corresponds to the internal standard (S, Oryza sativa L. ssp. japonica ‘Nipponbare’), the second to the embryo and the third to the endosperm.

Embryo and endosperm ploidies and seed origin in diploids

All seeds (99) from 21 diploid mothers were of sexual origin (Table 3, Supplementary Data Table S1). Most seeds (88) from all mothers had a regular sexual profile with 2x embryos and 3x endosperms (Fig. 2A) formed via fertilization of meiotically reduced female gametophytes with haploid (n = 1x) sperm cells. Increase in embryo ploidy was observed in 11 seeds having 3x embryos and 4x endosperms (Table 3, Fig. 2B). These seeds came from three out of four analysed diploid plants from site 1, representing 31 % of the total number of analysed seeds (35; Supplementary Data Table S1). The 3x embryos and 4x endosperms originated from an interploid cross in which a reduced megagametophyte (n = 1x) was fertilized with 2x sperm cells, either unreduced (2n = 2x) from a diploid male parent or reduced (n = 2x) from a tetraploid one.

Embryo and endosperm ploidies and seed origin in triploids

Triploid mothers (22) produced seeds (70) of apomictic origin (Table 3, Supplementary Data Table S1). Apomeiotic parthenogenetic embryos were always triploid, while endosperm ploidies ranged from 8x to 11x (Fig. 2E–H). Endosperm ploidies of 8x, 9x and 10x had almost equal frequencies within this group (Table 3). These ploidies indicate diverse pathways of endosperm formation. Endosperm ploidies of 8x, 9x and 10x were derived from unreduced binucleate central cells (2n = 3x) fertilized, respectively, by a single 2x sperm cell or two reduced 1x sperm cells; a single unreduced 3x sperm cell or two with unbalanced chromosome numbers; and a single unreduced 4x sperm cell or two reduced 2x sperm cells.

In addition, six seeds had 11x endosperm, implying a rare case of a trinucleate central cell being fertilized with one 2x (n or 2n) sperm cell or two 1x (n) sperm cells (Table 3, Fig. 2H).

Exceptionally, one seed had a sexually derived embryo corresponding to 4x (Table 3, Fig. 2D), probably due to fertilization of the apomeiotic triploid egg cell with 1x sperm. The second peak, corresponding to 8x, could represent either endoreduplicated embryo nuclei or nuclei in G2 phase (Fig. 2D).

Embryo and endosperm ploidies and seed origin in tetraploids

All seeds (133) of tetraploid mother plants (36) had tetraploid embryo and endosperm ploidies that ranged from 6x to 14x (Table 3, Fig. 2C, I–L, Supplementary Data Table S1). Sexually generated seeds of S. aria had 6x endosperms in only three cases (Fig. 2C); they arose via fertilization of meiotically reduced female gametophytes (n = 2x) with 2x sperm cells. The remaining 130 seeds were formed apomictically (Fig. 2I–L). In 115 apomictically formed seeds in both S. aria and S. austriaca, the prevailing endosperm ploidy was 12x, formed through the fusion of two central cell nuclei (2n = 4x) with two 2x sperm cells (Fig. 2K). Eleven seeds had 11x endosperms (Fig. 2J) suggesting fertilization with 3x pollen (a single unreduced sperm cell or two reduced unbalanced sperm cells from triploids). Of the remaining seeds, three had 10x endosperms, involving a single reduced (n = 2x) sperm cell (Fig. 2I); the last one had 14x endosperm, for which the occurrence of an extra central cell nucleus and the fertilization with a reduced (n = 2x) sperm cell is proposed (Fig. 2L).

Endosperm balance

Balanced endosperm with a maternal-to-paternal genome ratio of 2 m : 1p was present in the majority (74 %) of both sexually and asexually derived seeds (Table 3). Seeds with balanced endosperm had sexual embryo : endosperm profiles of 2x : 3x (29 %) and 4x : 6x (1 %) and apomictic profiles of 3x : 9x (6 %) and 4x : 12x (38 %). Unbalanced endosperm with a 1 m : 1p ratio was observed in sexually produced seeds from diploid S. aria via the involvement of 2x male gametes, resulting in a 3x : 4x profile (4 %) (Table 3). The remaining seeds (22 %) with unbalanced endosperms (1·5 m : 1p, 2·7 m : 1p, 3 m : 1p, 4 m : 1p and 4·5 m : 1p) were of apomictic origin and produced mostly by triploid mothers.

Mean endosperm and embryo ratios were 1·3 and 1·5 in sexually derived seeds and ranged from 2·5 to 3·5 in apomictically derived seeds (Table 3).

Pollen stainability

All analysed groups formed viable pollen. Sorbus aria cytotypes showed the highest frequencies of viable pollen (diploids, 83 ± 15 %, range 39–100 %; tetraploids, 86 ± 8 %, range 63–95 %). Mean values for triploid S. aria were the lowest (64 ± 16 %, range 32–95 %). Sorbus austriaca displayed 74 ± 17 % (range 32–97 %) viable pollen.

Seed set

Individual values for well-formed seeds ranged from 0 to 5 seeds per fruit. Compared with both tri- and tetraploid S. aria cytotypes (1·1 ± 0·8; range 0–3), diploid S. aria individuals showed almost twice the number of well-formed seeds (2·1 ± 1·0; range 0–5) while S. austriaca yielded 1·5 ± 0·9 (range 0–4) well-formed seeds per fruit.

DISCUSSION

Genome size and ploidy variation

In this study, leaf nucleus cytometry precisely differentiated cytotypes. Ploidy levels were easily inferred following monoploid values (1Cx) for both mother plants and the respective embryos. Monoploid values were quite similar among S. aria cytotypes and S. austriaca and remained stable across the polyploid series (2x, 3x and 4x). However, the highest difference between mean monoploid values was 0·026 pg, namely between di- and triploid S. aria (Table 2). Although minor, differences between monoploid mean values in cytotypes of S. aria indicate a genome downsizing trend. The causes of a decrease in monoploid genomes following polyploidization (Leitch and Bennett, 2004) are usually related to the loss of repetitive DNA, such as retroelements or retrotransposons (Bennetzen et al., 2005).

Our estimates of cytotype genome size are similar to those obtained by other authors (Lepší et al., 2008, 2009; Siljak-Yakovlev et al., 2010; Pellicer et al., 2012). The slight differences (∼10 %) in relation to other authors for genome size values in diploid S. aria and other cytotypes might be attributed to the use of different internal standards, protocols and sample sizes, but also to aneuploidy, small segmental amplifications, deletions, and retrotransposition of DNA short repeats during polyploidization (Bennetzen et al., 2005).

Nevertheless, to date, values for monoploid genomes of all studied Sorbus taxa have always shown stability, regardless of hybrid origin and parental combinations or possible autopolyploidy of taxa (Czech Republic: Lepší et al., 2008, 2009; Balkans: Siljak-Yakovlev et al., 2010; Britain: Pellicer et al., 2012; and the present study). This fact argues in favour of limited variation or conservatism of the 1Cx monoploid genome within the genus. Despite the high rate of hybridization, the numerous combinations of genome crosses and recombination among different taxa across Europe, monoploid genomes remain fairly stable. This scenario requires an evolutionary constraint regulating the stability of monoploid genomes.

Assignment of ploidy level to endosperm was facilitated by using embryo monoploid values, but the endosperm/embryo ratio was sometimes slightly shifted from the expected value. For example, in seeds of diploid mothers the mean ratio was 1·52 ± 0·04 and ranged from 1·41 to 1·64 (Supplementary Data Table S2). Slight deviations of DNA peak positions (from simple multiples of the monoploid value) can also be expected in higher endosperm ploidies (8x–14x), as the endosperm contains not only maternal genomic DNA but also sperm cells of variable origin, even from different taxa. Thus, the variation might be caused by a series of factors, such as aneuploidy of either central nuclei or sperm cells (variation hardly distinguishable in diploid nuclei but obvious in higher endosperm ploidies) and slight non-linearity of the cytometer (which might induce 2–5 % variation). Nevertheless, the highest endosperm/embryo ratio deviation was detected in a portion of triploid S. aria individuals having 3x : 9x (expected ratio 3, observed ratio 2·84 ± 0·14, CV = 5·0 %) and 3x : 11x profiles (expected ratio 3·67, observed ratio 3·35 ± 0·15, CV = 4 %) and in S. austriaca with a 4x : 11x (expected ratio 2·75, observed ratio 2·63 ± 0·13, CV = 5·0 %) profile (Supplementary Data Table S2). Higher deviations were observed for odd endosperm ploidies, where values were on the boundary between ploidy levels.

Patterns and processes in single- and mixed-ploidy populations

Our survey of single- and mixed-ploidy Sorbus populations reveals the pattern in which ploidy dynamics depend on the interaction of different cytotypes and their mating systems. In single-cytotype populations of either S. aria or S. austriaca, no intrapopulational ploidy variation or changes in the mating system were recorded (Fig. 1; Supplementary Data Table S1). On the other hand, mixed-ploidy populations were characterized by different proportions of a particular cytotype, which may reflect active processes within these populations, i.e. one cytotype slowly replacing the other (Sabara et al., 2013). In mixed cytotype populations we evidenced ploidy variations in the progeny relative to their maternal plants. Embryo ploidy increases in relation to maternal ploidy levels resulted from interploid crossings in mixed-ploidy Sorbus populations. Eleven sexually produced triploid embryos of diploid S. aria mothers were fertilized with 2x pollen, likely from a tetraploid parent (Table 3). These seeds came from the most extensively sampled population (site 1), where the diploid cytotype was in the minority (eight individuals amongst 105 sampled) and the pressure of polyploid pollen was high (site 1, Table 1, Fig. 1). All analysed Sorbus taxa had the same flowering time, and possible 2x gamete donors in open pollinations were either tetraploid S. aria or S. austriaca. Formation of unreduced gametes has been documented in Sorbus (Liljefors, 1953; Jankun and Kovanda, 1986, 1987; Aas et al., 1994); thus, the involvement of unreduced pollen of diploid S. aria cannot be ruled out. However, the presence of unreduced gametes via cases of embryo ploidy increase was not observed in pure populations of diploid S. aria (Fig. 1, Supplementary Data Table S1).

The 11 observed triploid embryos represent 31 % of those analysed from diploid mother plants at site 1 (Table 1). This is a high proportion relative to diploid embryos produced; it is reasonable to assume that the number of triploid progeny would be much higher in larger sample sizes of both seeds and maternal trees. These triploid embryos represent novel genetic lineages formed through independent crossing events. Their success and long-term survival will depend critically on potentials determined by their mating system (Hörandl, 2010). Therefore, the genetic structure of triploid populations is not necessarily clonal due to apomixis, but rather comprises multiple clonal lineages having arisen via independent crossing events (A. Hajrudinović, unpubl. res.). The triploid cytotype was prevalent in two out of four sympatric populations analysed (Fig. 1), which indicates their ecological and reproductive success. All triploid mothers produced seeds by apomixis, having the potential to establish themselves in coexistence with other cytotypes. Moreover, they are generally considered to have higher potential for long-distance dispersal and colonization of new habitats compared with their related sexuals (reviewed in Hörandl, 2006).

Ludwig et al. (2013) found that apomictic triploid species investigated from Avon Gorge were self-incompatible, requiring pollen of other Sorbus taxa for successful endosperm and seed formation. In the majority of cases, triploids used either 1x or 2x pollen (Table 3). They were also capable of using 3x sperm cells, as concluded from the 3x : 9x embryo : endosperm profile (25 % in the sample of triploid seeds) (Table 3). At all four sympatric sites (Fig. 1), possible pollen donors could be diploid S. aria, different genetic lineages of tri- and tetraploid S. aria-like and S. austriaca individuals producing reduced 1x, unbalanced 3x and reduced 2x sperm cells, but also S. aucuparia, which is abundant at these sites (A. Hajrudinović, field observations).

Our findings are concordant with the existing model of apomictic taxon formation, which requires crosses of an apomictic polyploid and sexual diploid progenitor (Robertson et al., 2004a, 2010; Ludwig et al., 2013). We propose that the production of triploids occurs in the direction of pollen exchange from apomictic tetraploids to sexual diploids. This is in favour of triploid formation via the tetraploid bridge (Robertson et al., 2004a, b; Ludwig et al., 2013).

Exceptionally, one single case indicating sexuality was detected in a seed with a 4x embryo (Table 3, Fig. 2D), which could occur through the union of a 3x unreduced female egg cell and a 1x sperm. Fertilization of unreduced triploid eggs is a rather rare case but documented in Sorbus teodori (Liljefors, 1953) and triploid individuals of Sorbus chamaemespilus (Jankun and Kovanda, 1986). The same pathway of origin was established for new tetraploid apomicts in Great Britain: Sorbus pseudofennica on Arran (Robertson et al., 2004b) and Sorbus × houstoniae in Avon Gorge (Robertson et al., 2010). In addition to tetraploidy, these facts also favour a triploid bridge in Sorbus.

The exclusive occurrence of polyploid S. aria individuals in sympatric populations and their predominantly apomictic reproduction mode (Table 3, Fig. 1) suggest close connections with processes operating elsewhere in Europe in Sorbus communities that are ploidy-mixed. Evolutionary processes in Sorbus have been thoroughly studied using molecular markers, a prime example being the situation on the island of Arran involving different combinations of species (Robertson et al, 2004a, b). There, initial crossings of diploid sexuals and tetraploid apomicts produced apomictic triploid offspring. Subsequently, unreduced triploid eggs were fertilized by reduced pollen of the parental diploid taxon to produce new tetraploids. Newly derived tetraploids were facultatively apomictic and participated in further crossings to generate new diversity. Our results suggest that the active nature of processes governing Sorbus diversity in the Balkans is probably parallel to those determined in Arran and other diversity hotspots of the genus.

While S. austriaca produced all seeds via apomixis, several sexually originated seeds were found in S. aria tetraploids (Table 3, Fig. 2C). These were documented in two different mothers from site 3. The same mothers also produced seeds of apomictic origin (Supplementary Data Table S1). It is known that tetraploid Sorbus apomicts can produce meiotically reduced egg cells and undergo regular sexual pathways (Liljefors, 1953; Jankun and Kovanda, 1986). The preserved sexuality of tetraploids provides an additional evolutionary potential for the exchange of genetic material within Sorbus agamic–sexual complexes in Bosnia and Herzegovina. However, ploidy data cannot reveal their exact origins so this question remains open for our S. aria polyploids. We assume two different scenarios. The most parsimonious scenario would initially include crossing of apomictic S. austriaca reduced 2x sperm cells with either reduced 1x or unreduced 2x megagametophytes of sexual S. aria, to produce triploids and tetraploids, respectively. The other involves fertilization of unreduced female gametes of an apomictic triploid to give rise to a tetraploid.

Endosperm formation and balance requirements

The developmental processes and pathways in sexually and asexually formed female gametophytes were thoroughly presented in embryological and cytological studies of Sorbus taxa from Scandinavia and the Czech Republic (Liljefors, 1953; Jankun and Kovanda, 1986, 1987, 1988). In our study, the pathways for seed formation inferred from FCSS data were conclusive and congruent with the body of knowledge derived from embryological studies. FCSS resulted in various embryo : endosperm profiles and pointed to the common pathways of female gametophyte formation and the ploidies of contributing sperm cells (Table 3). Female gametophytes were either reduced or unreduced, with the binucleate central cell fertilized with either one or two sperm cells.

Addition to present knowledge is observed in a tendency towards a balanced 2m : 1p parental genome contribution to the endosperm shared by Sorbus diploids and tetraploids, regardless of their sexual or asexual origin (Table 3). The balanced endosperm in seeds of sexual origin was maintained by the exchange of meiotically reduced homoploid gametes. On the other hand, the vast majority of tetraploid apomicts (86 %) maintained a balanced 12x endosperm by using either two reduced 2x sperm cells or a single unreduced 4x sperm cell for the fertilization of the central cell (Table 3). Dispermy of the central cell has been suggested for tetraploids of several rosaceous genera: Amelanchier (Burgess et al., 2014), Rubus (Šarhanová et al., 2012) and Crataegus (Talent and Dickinson, 2007a). Interestingly, the 10x endosperm, indicating endosperm imbalance, was prevalent in Crataegus and Rubus but quite rare in Amelanchier and Sorbus (Table 3).

The balanced endosperm in triploids’ seeds, found in one-quarter of the sample, was achieved by the contribution of three paternal chromosome complements, either by a single 3x unreduced sperm cell or two meiotically reduced unbalanced sperm cells, resulting in a 9x endosperm as the most balanced state in terms of parental contributions (Table 3). The same uncommon apomictic profile with balanced 9x endosperm was detected in Rubus triploids (7 %; Šarhanová et al., 2012). This pathway greatly depends on the frequency of production of such gametes. Irregular microsporogenesis, resulting in a series of different chromosome numbers (17–25) was reported in triploid Sorbus bohemica that had a low pollen stainability of ∼20 % (Jankun and Kovanda, 1987). In our study, all cytotype groups showed high pollen stainability, including triploids, in which stainability reached 95 % in certain individuals (see the Results section).

Most seeds (75 %) from triploid mothers showed deviations from a balanced endosperm, having ratios of 1·5m : 1p, 3m : 1p and 4·5m : 1p (Table 3), as is also frequent in rosaceous triploid apomicts of Amelanchier (1·5m : 1p and 3m : 1p, Burgess et al., 2014), Rubus (1·5m : 1p and 3m : 1p; Šarhanová et al., 2012) and Crataegus (1·5m : 1p; Talent and Dickinson, 2007a), but also Sorbus (3 m :1p; Jankun and Kovanda, 1987).

However, tolerance of imbalance and viability in these seeds remains questionable. The analysis of seed set showed that triploids yielded the lowest number of seeds relative to diploids and tetraploids. However, triploids apparently tolerate endosperm imbalances, given their numerical dominance in some of the populations investigated here.

Tolerance of endosperm imbalance potentiates the formation of polyploids via interploidal crosses in the genus (Hörandl et al., 2008). This potential particularly refers to unbalanced endosperm with the 1m : 1p ratio observed in sexually produced seeds from diploid S. aria via the involvement of 2x male gametes, resulting in a 3x : 4x profile (Table 3). Crossings between diploids and tetraploids probably do occur recurrently, since this was the only observed pathway of triploid formation in the present dataset. In any case, the endosperm balance requirement must always be violated in seeds originating via diploid–tetraploid crosses, regardless of the direction of the cross. Hence, as has been suggested by Ludwig et al. (2013), we advocate that formation of Sorbus triploids is facilitated by relaxed sensitivity to the endosperm balance requirement.

Concluding remarks

Results on flow cytometry proved valuable in the analysis of ploidy dynamics and the identification of mating systems in Sorbus. Cytotype diversity in mixed-ploidy populations demonstrated a link with the coexistence of sexual S. aria diploids with apomictic S. austriaca tetraploids. Tetraploid-to-diploid gene flow was suggested by combining ploidy data of progenitors and progeny. The consequence of being a Sorbus polyploid is a change in reproduction mode that favours apomixis. Triploids exhibited near obligate apomixis, as well as did tetraploid S. austriaca. Tetraploids of S. aria were characterized by facultative apomixis, having a small portion of retained sexuality. Facultative apomixis provides additional evolutionary potential in further interactions within Bosnian Sorbus agamic–sexual complexes. The ecological and reproductive success of triploids and their putative capability for tolerating endosperm imbalance pose intriguing evolutionary questions from developmental perspectives.

Our data show similarities with processes governing Sorbus diversity in other hotspots of the genus in Europe. Bosnian Sorbus sexual–agamic complexes harbouring mixtures of cytotypes and the interacting sexual and asexual lineages represent potential reservoirs of new biodiversity that merit a status of special concern for conservation. By using cytometric data as a solid baseline, our forthcoming research will use molecular methods to identify precise parentages and the underlying genetic mechanisms.

SUPPLEMENTARY DATA

Supplementary data are available online at www.aob.oxfordjournals and consist of the following. Table S1: flow cytometric seed screening results for 303 Sorbus aria s.l. and S. austriaca seeds from natural populations in Bosnia and Herzegovina. Table S2: group statistics for observed FCSS profiles in Sorbus aria s.l. and S. austriaca. Figure S1: DNA histograms of Sorbus aria diploid mother tree seeds for FCSS, using linear and log axes.