Introduction

Natural hybridization is common and its contributions to speciation have been recognized in higher plants (Stebbins 1959; Ellstrand et al. 1996; Arnold 1997). On the basis of the assumption that mating compatibility weakens with increasing genetic distance (Rieseberg and Carney 1998; Mallet 2005, 2008), data from crossing experiments have been and are still used to evaluate intra- and interspecific genetic relationships and test the accuracy of taxonomic schemes in many different plant groups (for example Andersson 1993; Raimondi et al. 2003; Favero et al. 2006; Pellegrino et al. 2009; Marcussen and Borgen 2011). Novel phylogenetic approaches are being developed to be able to accommodate the reticulate evolutionary patterns resulting from hybridization (Vriesendorp and Bakker 2005).

In the genus Oryza L., one of the first accounts of natural hybridization between wild and cultivated taxa was reported by Roschevics (1931). Nezu et al. (1960) analyzed the crossability among 17 Oryza species and detected interfertility among the Asian members as well as among the African species of series Sativae although some taxonomically dubious species names (e.g., O. perennis Moench, O. stapfii Roshev.) were used in their study. Much later, more extensive crossing experiments on Oryza series Sativae were conducted where Naredo et al. (1997, 1998) reported highly variable crossabilities among the species from Asia (O. nivara and O. rufipogon), Australasia (specifically Australia and New Guinea) (O. meridionalis) and tropical America (O. glumaepatula Steud.) and Juliano et al. (2005) observed reproductive compatibility between populations of O. meridionalis that are from the same geographic province (e.g., Irian Jaya., Northern Territory and Queensland).

The current research focuses on the three wild species of Oryza series Sativae in Asia Pacific. O. rufipogon is perennial, predominantly outcrossing, photoperiod sensitive and thrives in permanently submerged areas of tropical Asia and northern Australia. O. nivara and O. meridionalis are both annual, inbreeding, photoperiod insensitive and can be found in seasonally wet habitats. O. rufipogon overlaps with the former in continental Asia and with the latter in Australasia. O. meridionalis is widely accepted as a taxonomic species (Ng et al. 1981; Duistermaat 1987; Juliano et al. 2005) while O. nivara (considered in this study as a species) is sometimes treated as an ecotype of O. rufipogon (Tateoka 1963; Oka 1988; Vaughan et al. 2003). Recent studies revealed that these two Asian taxa differ morphologically and have genetically distinct populations that overlap across their distribution but are differentiated at the local scale (Banaticla-Hilario 2012). Despite previous experiments (Naredo et al. 1997, 1998), the full extent of reproductive isolation between O. nivara and O. rufipogon remains to be established.

This study examines the strength of post-pollination isolation within and between the three species in relation to other factors that were not accounted for in the earlier crossing studies such as the spatial distance between the parents’ geographic origins, and sympatricity versus non-sympatricity of parents. The specific objectives are to: (a) compare the crossability exhibited by different intra and interspecific combinations of Asia Pacific Oryza series Sativae species; (b) determine the correlation between crossability and spatial distance between parental origins; (c) test whether reproductive success differs between sympatric and non-sympatric crosses; and (d) evaluate the morphology of the obtained F1 hybrids.

Materials and methods

Crossing experiments

A set of 15 accessions from the International Rice Genebank (IRG) in the International Rice Research Institute (IRRI) was selected to represent the three species across their geographic range and include sympatric population pairs of O. nivara and O. rufipogon and of O. meridionalis and O. rufipogon (Fig. 1; Table 1). This material has been previously phenotyped and genotyped with simple sequence repeat (SSR) markers (Banaticla-Hilario 2012). Five plants from each accession were grown at the wild rice nursery screenhouse in IRRI, where all pollination experiments (September–December of 2008 and 2009) and succeeding phenotyping of F1 plants (June 2009–September 2010) were conducted. To achieve self-pollination, five panicles from each plant were covered with glassine bags prior to anthesis and harvested once the grains ripened. The seed set of each selfed plant was then determined by obtaining the average percentage of filled spikelets from the five bagged panicles (Online Resource 1). The seed set of each accession was also derived from the mean seed set of the 25 bagged panicles (Table 1).

Fig. 1
figure 1

Geographic distribution of the 15 parental accessions used in the crossing experiments

Full size image
Table 1 Parental accessions and their mean seed set obtained from 25 self-pollinated panicles (five panicles from each of the five plants per accession)
Full size table

A total of 25,692 spikelets from 803 panicles were artificially pollinated in 114 combinations: 8 within accessions, 37 between accessions of the same species, and 69 between species (Online Resource 2). Emasculation of female parents, artificial pollination, and harvesting of F1 seeds were performed as described by Naredo et al. (1997) while F1 seed germination and F1 plant establishment followed the methods of Naredo et al. (1998).

Evaluation of hybrids

The 145 F1 plants obtained were evaluated using 10 characters that can discriminate between either all or certain pairs of the three parental species (based on Banaticla-Hilario 2012). These were culm length (cm), leaf length (cm), flag leaf width (cm), spikelet length (mm) and width (mm), awn length (mm), anther length (mm), anther length to spikelet length ratio, spikelet fertility (i.e., seed set of selfed panicles) and number of days from seeding to first heading. Measurements were obtained and fresh leaf samples (for DNA extraction) were collected from each F1 plant. Herbarium voucher specimens of each hybrid were collected and deposited at the IRG Herbarium.

DNA samples were extracted and genotyped as described in Fulton et al. (1995) and Banaticla-Hilario (2012), respectively. Based on the SSR profile of the parents, certain combinations of 26 microsatellite markers were used to validate the true F1 nature of the plants resulting from the different crosses (Online Resource 3).

Data analyses

All crossing and phenotype data were analysed using R 2.14 (R Development Core Team 2011). Five crossability estimates were examined: (1) seed set (the number of filled spikelets as a percentage of the total number of pollinated spikelets); (2) germinability (the number of germinated F1 seeds as a percentage of the total number of filled spikelets; (3) seedling survival (the number of established F1 plants as a percentage of the total number of germinated seeds); (4) cross fertility (the number of non-sterile hybrids as a percentage of the total number of established F1 plants); and (5) F1 fertility (the mean seed set of the selfed (bagged) panicles of the resulting mature fertile F1 hybrids).

Box and whisker plots of these estimates were constructed using the boxplot() function. All crossability data (expressed as percentages) were transformed to arcsine values, then subjected to a one-way analysis of variance (ANOVA) with the car package (Fox and Weisberg 2011) to determine whether crossability across all and within certain species combinations varies with different factors (e.g., maternal/paternal species, sympatricity of parents). The correlation of crossability estimates with spatial distance between the geographic origin of parents was also tested using the cor.test() function of the base package.

Parental and F1 phenotypes were compared by conducting a principal components analysis (PCA) of nine species-distinguishing traits (culm length, leaf length, flag leaf width, spikelet length, spikelet width, anther length, awn length, spikelet fertility, and number of days from seeding to heading), using the prcomp() function.

Anther length, anther length to spikelet length ratio, awn length and spikelet width can readily distinguish the three parental species from each other (Banaticla-Hilario 2012). Scatter plots of these highly discriminating characters were constructed to determine whether the interspecific F1s would exhibit intermediate morphology. Measurements from 10 genetically confirmed hybrid accessions identified by Banaticla-Hilario (2012) as intermediate forms between O. nivara and O. rufipogon were incorporated in the plots for comparison and validation.

Results

Intraspecific crossability

Self-compatibility of parental accessions

Fertility of self-pollinated panicles does not differ significantly among the three species, among accessions within O. meridionalis and among accessions within O. rufipogon. However, accessions of O. nivara differ significantly at the p < 0.05 level. All species show a high variability in seed set among plants of the same accession and in a few cases, even among panicles from the same plant (Online Resource 1, Table 1).

Crosses within and between accessions

The crossability results of each cross combination are tabulated in Online Resource 2 while the ANOVA results are listed in Table 2. Comparisons of crossability among the different intra- and interspecific crosses are presented in Fig. 2. Variations in the mean seed set of crossed panicles and of self-pollinated F1 plants resulting from specific parental combinations are displayed in Figs. 3 and 4, respectively.

Table 2 Variations in intra- and interspecific crossability of Asia Pacific Oryza series Sativae
Full size table
Fig. 2
figure 2

Crossability within and between Oryza series Sativae species in Asia Pacific: a seed set; b germinability; c seedling survival; d cross fertility; and e F1 fertility

Full size image
Fig. 3
figure 3

Mean seed set of selfed panicles and of crosses between and within the Asia Pacific species of Oryza series Sativae

Full size image
Fig. 4
figure 4

Mean seed set of selfed F1 hybrids from inter and intra-specific crosses of Asia Pacific Oryza series Sativae compared to the mean seed set of selfed parental accessions

Full size image

Three out of five crosses within O. meridionalis and four out of six within O. nivara had zero seed set (Online Resource 2, Fig. 3). In O. meridionalis, only the combinations M1 × M2 and M1 × M1 produced F1 seeds that however, failed to germinate (Online Resource 2). In O. nivara, the crosses between N18 and N51 produced F1 seeds that germinated but did not survive the seedling stage (Online Resource 2). Intraspecific combinations (crosses within and between accessions) produced fewer seeds than the selfed (i.e., bagged) panicles in both annual species (Fig. 3).

In contrast, out of 34 crosses within O. rufipogon, 11 generated non-sterile F1 hybrids and only four had zero seed set (Online Resource 2; Figs. 3, 4). Intraspecific O. rufipogon combinations exhibited significantly higher seed set than all the other intraspecific and interspecific combinations (at p < 0.0001 level) and were also more successful than crosses between and within the two annual species in terms of survival, cross fertility and F1 fertility (Table 2; Fig. 2).

In O. rufipogon, crosses within the same accession displayed lower seed set (p < 0.05) but higher germinability (p < 0.01) than crosses between different accessions (Table 2). The germinated F1 seeds of these intra-population crosses did not survive the seedling stage (Table 2). Panicles resulting from both intra- and inter-population crosses exhibited significantly lower seed set (p < 0.0001) than the selfed panicles (43.81 %). The seed set of combinations R30 × R2 (32.3 %), R51 × R64 (46.4 %) and R53 × R64 (51.4 %) exceeded the seed set (of self-pollinated panicles) of one or both of their corresponding parental accessions (Fig. 3). The most successful combinations in terms of F1 fertility were R59 × R30 (69.3 %), R63 × R66 (60.1 %) and R53 × R59 (55.1 %) (Fig. 4).

Geographic distance between parents was not significantly correlated with the different estimates of crossability in O. rufipogon. The spatial distance between parents of combinations that produced non-sterile F1s ranged from 1,062 to 3,813 km. Crosses with distance below and beyond the said limit exhibited zero seedling survival (Online Resource 2).

Interspecific crossability

Crosses between O. meridionalis and O. nivara were as unproductive as their corresponding intraspecific crosses (Table 2; Fig. 2). Six out of seven combinations had zero seed set (Online Resource 2; Fig. 3). Only N51 × M2 produced a single seed that did not survive the seedling stage (Online Resource 2; Fig. 3).

In contrast, 12 out of 21 crosses between O. meridionalis and O. rufipogon had seed sets that ranged from 0.37 to 35.75 % (Fig. 3). Yet, the 38 fertile F1 plants obtained were mainly from three combinations: M2 × R66 (20.22 % F1 fertility), M2 × R30 (3.21 %) and R30 × M2 (0.77 %) (Fig. 4). Crosses with maternal O. meridionalis had significantly higher seed set (p < 0.0001), survival (p < 0.0001), proportion of fertile hybrids (p < 0.05) and F1 fertility (p < 0.01) than crosses with maternal O. rufipogon (Fig. 2).

Among the 41 crosses between O. nivara and O. rufipogon, 30 combinations had seed sets that ranged from 0.4 to 40.5 % (Online Resource 2; Fig. 3). Nonetheless, only five combinations generated 14 fertile F1 plants: R51 × N18 (53.2 % mean F1 fertility); N18 × R59 (31.1 %); N2 × R2 (35.2 %); N18 × R64 (30.4 %); and R63 × N2 (13.6 %) (Fig. 4). All crossability estimates were comparable between crosses with maternal O. nivara and crosses with maternal O. rufipogon (Table 2; Fig. 2). The F1s of O. nivara and O. rufipogon exhibited lower survival (p < 0.0001) but higher fertility (p < 0.01) than the F1s of O. meridionalis and O. rufipogon.

In interspecific combinations involving O. rufipogon, all crossability measures were not significantly linearly correlated with geographic distance between parental origins. Significant crossability differences between sympatric and non-sympatric combinations were detected in certain parental combinations. Non-sympatric parents produced higher seedset (significant at p < 0.05 level for combinations with maternal O. rufipogon and paternal O. meridionalis), lower germinability (significant at p < 0.01 level for combinations with maternal O. nivara and paternal O. rufipogon) and generally higher seedling survival (significant at p < 0.05 level for combinations with maternal O. meridionalis and paternal O. rufipogon) than sympatric parents. Non-sympatric crosses were also more successful in terms of F1 fertility since, among the sympatric combinations, only the crosses between maternal O. nivara and paternal O. rufipogon produced fertile F1 plants while the rest produced F1s that did not survive the seedling stage.

Characteristics of F1 plants

F1 nature of progeny

Several of the SSR markers used enabled the validation of the F1 nature of the progeny resulting from the crosses. RM44, RM316 and RM154 discriminate among most of the intra- and interspecific combinations of Asia Pacific Oryza series Sativae. All hybrids of O. nivara and O. rufipogon are distinguished by RM152 while all intraspecific O. rufipogon hybrids (except the offspring of R2 × R51 and R51 × R64) are discriminated by RM237 (Online Resource 3). RM44, RM316, RM154, RM271, RM484 and RM124 differentiate the hybrids of O. meridionalis and O. rufipogon. Among the screened plants, those that appear to be produced by self-pollination (probably due to mechanical errors in crossing) were excluded from the analyses.

Life cycle and phenology

The F1s resulting from crosses within O. rufipogon and between O. meridionalis and O. rufipogon exhibit a perennial life cycle while all those from crosses between O. nivara and O. rufipogon display an annual life cycle (except for the single offspring of R63 × N2).

Intraspecific O. rufipogon F1s flower significantly later (mean of 271 days from seeding to first heading) than the parental accessions (mean number of days: O. rufipogon—128; O. nivara—104; O. meridionalis—98) and other interspecific F1s (mean number of days: F1s of O. meridionalis and O. rufipogon—121; F1s of O. nivara and O. rufipogon—80) at p < 0.0001 level. Unlike most intraspecific O. rufipogon combinations, R63 × R66 and R2 × R51 produced F1 hybrids that flowers relatively early (number of days from seeding to first heading ranged from 109 to 122).

Among the F1s of O. meridionalis and O. rufipogon, those with maternal O. rufipogon take longer to flower (247 days) than those with maternal O. meridionalis (107 days). On the other hand, in crosses between O. nivara and O. rufipogon, the average number of days from seeding to first heading of hybrids with maternal O. nivara (85 days) is higher but does not differ significantly from that of hybrids with maternal O. rufipogon (70 days).

Morphology

The PCA results are shown in Table 3 and Fig. 5. The first two principal components account for 57.6 % of the total variance. Hybrids from the same cross combination generally cluster together. F1 plants from interspecific crosses are more morphologically similar to their annual parental species (Fig. 5).

Table 3 Factor loadings and importance of the first three principal components of the PCA of the nine morphological characters of parental accessions and F1 hybrids
Full size table
Fig. 5
figure 5

PCA plot showing the clustering patterns of F1 hybrids and parental accessions based on nine quantitative morphological characters

Full size image

The first principal component (34.74 %) separated the intraspecific F1s of non-Australasian O. rufipogon (long-culmed, long-anthered, and late flowering) from two groups of short-culmed, short-anthered, and early flowering interspecific F1s (O. meridionalis × Australasian O. rufipogon and O. nivara × O. rufipogon) (Fig. 5). The parental O. rufipogon accessions together with the hybrids of O. meridionalis and non-Australasian O. rufipogon and most of the intraspecific Australasian O. rufipogon hybrids are positioned in between these groups (Fig. 5).

The second principal component (22.84 %) separates the interspecific hybrids of O. nivara and O. rufipogon from the F1s of O. meridionalis and O. rufipogon (Fig. 5; Table 3). The former has shorter and broader spikelets, shorter awns and more fertile panicles than the latter. The third principal component explains 17.93 % of the total variance (with leaf length and flag leaf width exhibiting the highest factor loading) (Table 3) and, along with the rest of the principal components, do not show any distinct clustering pattern.

Figure 6 shows that anther, awn and spikelet measurements of intraspecific O. rufipogon F1s fall within the range of their parental species. An offspring of M2 × R30 also appears to be more similar to O. rufipogon based on the four traits. Anther length and anther length to spikelet length ratio of interspecific hybrids are intermediate between O. rufipogon and the annual species while awn length and spikelet width tend to be more similar to the annual parent. Most F1s between O. nivara and O. rufipogon have shorter awns and broader spikelets than F1s between O. meridionalis and O. rufipogon. Accessions genetically identified as hybrids between O. nivara and O. rufipogon also exhibit intermediate anther length and anther length to spikelet length ratio but some populations had slightly longer awns and narrower spikelets than the F1 hybrids of O. nivara and O. rufipogon obtained in this experiment (Fig. 6).

Fig. 6
figure 6

Scatter plot of the four quantitative characters that can readily discriminate the species and hybrids of Asia Pacific Oryza series Sativae: a awn length versus anther length; b spikelet width versus anther length to spikelet length ratio

Full size image

Discussion

Varying fertilities of self-pollinated populations within species

The reported difference between the seed productivity of O. nivara (high) and O. rufipogon (low) (Vaughan and Morishima 2003) is not evident from our results. This may be due to the limited number of accessions used in the present study. It should also be noted that panicle fertilities may vary between in situ and ex situ-conserved wild rice populations as observed in O. officinalis from Malaysia (Kairudin et al. 1996) and O. rufipogon from Dongxiang, China (Yu et al. 2007). Banaticla-Hilario (2012) examined the morphology of 121 populations (including the parental accessions in this experiment) and found significant differences in the spikelet fertility among the three Asia Pacific Oryza series Sativae species where O. rufipogon displayed the lowest and O. nivara the highest value.

The high variability of seed set observed among accessions of O. nivara (Table 1) and among plants of the same accession (Online Resource 1) reflects the high genetic diversity between and within wild rice populations. The plants in each accession might be segregating both for self- and cross-compatibility. Wei et al. (2010) detected segregation at the S5 locus (one of the major genes that control female sterility) in O. rufipogon accessions.

Incompatibility within and between the selfing species

The high seed production of selfed parental panicles and generally poor seed set of crossed panicles (Fig. 3) demonstrate the highly selfing nature of O. meridionalis and O. nivara.

The absence of seed set in most of the crosses within and between these two annual species suggests a strong pre-zygotic barrier. Concurrently, the inviability of the few generated seeds implies post-zygotic isolation. Although Naredo et al. (1998) obtained several F1 plants from crosses within and between O. nivara and O. meridionalis, the hybrids exhibited very low fertility. Substantial post-zygotic barriers (i.e., low fitness and reduced fertility of intraspecific hybrids) were also observed in other selfing species (Grundt et al. 2006; Bomblies et al. 2007; Widmer et al. 2009).

Nevertheless, Juliano et al. (2005) reported high F1 fertility (>50 % spikelet fertility) in crosses between O. meridionalis populations from the same geographic province indicating that reproductive compatibility is higher within the regions of Australasia (e.g. Irian Jaya, Northern Territory and Queensland). This cannot be confirmed in the current study as only three O. meridionalis accessions from different regions were used allowing only for intra-population and inter-regional comparisons. It would be interesting to know whether interfertility at the regional scale also exists in O. nivara.

The observed incompatibilities within and between O. meridionalis and O. nivara probably reflect the genetic dissimilarities among the geographically isolated parental accessions used in this experiment. Pollen-pistil interactions should be investigated to confirm and provide insights into the pre-zygotic mechanisms involved in post-pollination isolation of these selfing species.

Reproductive coherence and spatial crossability patterns within O. rufipogon

As expected, O. rufipogon shows greater intraspecific compatibility than the two inbreeding species. Outcrossing perennials that can propagate vegetatively (like O. rufipogon) tend to be more predisposed to natural hybridization (Ellstrand et al. 1996). However, it should be noted that this wild rice species is also self-compatible as exemplified by the abundant seed set of selfed panicles (Online Resource 1, Table 1; Fig. 3). Intraspecific O. rufipogon crosses also had significantly higher seed set than all other interspecific combinations (Table 2; Figs. 2, 3) signifying the presence of a pre-zygotic isolation mechanism.

In intraspecific O. rufipogon crosses among the selected accessions, production of viable and non-sterile F1 plants seemed feasible only within a certain range of parental spatial distance (approximately 1,100–3,900 km). Reproductive failure was expected in crosses with spatially remote and consequently genetically diversified parents. Surprisingly though, post-zygotic barriers also seemed to favour combinations with spatially proximate parents. This suggests that even geographically close O. rufipogon populations may be genetically differentiated and reflects the complex diversity patterns within this perennial wild rice species.

Within O. rufipogon, crossability estimates vary considerably in crosses between those from the same region (e.g., South Asia, continental Southeast Asia, maritime Southeast Asia and Australasia) as well as in inter-regional combinations (Online Resource 1). Still, it can be discerned that crosses within and between maritime Southeast Asian and Australasian populations (especially those involving R53, R59 and R64) are generally the most successful in terms of seed set and F1 fertility (Figs. 3, 4).

Direction, extent and spatial patterns of interspecific crosses involving O. rufipogon

Differences in the success of reciprocal crosses

O. meridionalis and O. rufipogon exhibit asymmetrical compatibility where combinations with maternal O. meridionalis have more reproductive success than crosses with maternal O. rufipogon (Table 2; Fig. 2). Reciprocal crosses of O. nivara and O. rufipogon display comparable crossability success. Similar compatibility patterns were detected by Naredo et al. (1997).

Asymmetrical reproductive isolation appears to be widespread in plants (Rieseberg and Carney 1998; Tiffin et al. 2001; Yasumoto and Yahara 2006) and may result from parental differences in style/stigma length, breeding system and fruit abortion and also from nuclear-cytoplasmic incompatibilities (Tiffin et al. 2001). The stigma length factor seems plausible in the case of the three Oryza species since this trait separates O. meridionalis from O. rufipogon (specifically the Australasian populations) but does not differ between O. nivara and the perennial species (Banaticla-Hilario 2012). Molecular studies identified O. meridionalis as the first lineage to diverge from the rest of the series (Zhu and Ge 2005; Kwon et al. 2006; Duan et al. 2007). It was even reported to be genetically more affiliated to the African species than the Asian series Sativae members (Duan et al. 2007), hence some incongruencies in nuclear-cytoplasmic interactions are to be expected in interspecific crosses. Conversely, despite the differences in the breeding systems of O. nivara and O. rufipogon, these recently diverged taxa (Zheng and Ge 2010) seem to still maintain a certain degree of cytoplasmic compatibility.

It is worth mentioning that asymmetric compatibilities have also been detected in several intraspecific crosses of O. rufipogon, in line with the large diversity within this outcrossing species.

Extent of post-mating isolation

Mallet (2008) discussed that the reproductive failure of crosses within species and the relative success of interspecific crosses preclude the use of crossability estimates in confirming the species status of certain plant taxa. This also applies to the two annual Oryza taxa. It is well known that O. meridionalis is a genetically distinct species (Xu et al. 2005; Kwon et al. 2006; Banaticla-Hilario 2012) and that O. nivara and O. rufipogon are closely related taxa (Vaughan et al. 2003; Kwon et al. 2006; Zhu and Ge 2005; Duan et al. 2007; Zheng and Ge 2010). Hence, the reproductive compatibilities of O. meridionalis and O. nivara with O. rufipogon are discussed in the context of their genetic relationships, not their taxonomic statuses.

Theoretically, O. nivara and O. meridionalis would be reproductively isolated from O. rufipogon in the wild by similar pre-pollination barriers (e.g., differences in habitat, phenology and breeding system). Such pre-mating isolation mechanisms are potent yet leaky reproductive barriers between species and post-zygotic mechanisms are needed to counter residual gene flow that penetrates these pre-mating barriers (Widmer et al. 2009). In this experiment, post-pollination pre-zygotic isolation (i.e., seed set) was quite comparable in interspecific combinations of O. meridionalis and O. rufipogon and of O. nivara and O. rufipogon (Table 2). However, since symmetrically compatible crosses are likely to have more chances of successful mating than asymmetrical ones, O. nivara could have a reproductive advantage over O. meridionalis under natural conditions.

Disparities in the strength of several post-zygotic barriers were detected when studying the results of crosses with different annual parental species. As predicted, reduced F1 fertility is more pronounced in crosses between the genetically distant O. meridionalis and O. rufipogon (17.2 % mean F1 fertility) when compared to those between O. nivara and O. rufipogon (36.9 % mean F1 fertility). Yet, the reduced F1 fertility is counterbalanced by the greater abundance of generated established F1 plants (38 F1s, mean of 12.7 F1s per cross vs. 14 F1s, 2.8 F1s per cross in O. nivara and O. rufipogon combinations). Furthermore, unlike the predominantly annual F1s of O. nivara and O. rufipogon, hybrids between O. meridionalis and O. rufipogon are sustained for longer periods by their perennial habit. Thus, in crosses involving O. meridionalis, high production and (to a certain degree) perenniality of hybrids can compensate for the low F1 fertility while high F1 fertility in crosses involving O. nivara may account for the low production and short life span of hybrids.

Despite differences in the intensity of several post-zygotic mechanisms, the extent of post-mating isolation of both O. meridionalis and O. nivara against O. rufipogon seems comparable. This indicates that the magnitude of reproductive isolation may not always correspond to genetic relationships and that gene flow barriers may remain permeable even between distantly related taxa. Mallet (2005, 2008) claimed that reproductive barriers could remain porous for millions of years after species divergence. Furthermore, Widmer et al. (2009) reported the ability of certain recently diverged species to rapidly develop strong post-zygotic barriers. It can be said that within Asia Pacific Oryza series Sativae reproductive isolation is incomplete and a certain amount of gene flow transgresses species boundaries. However, to confirm this conclusion, hybrid breakdown or its absence must be confirmed and the viability and fertility of succeeding hybrid generations should be monitored to ascertain the ultimate reproductive success of interspecific crosses.

Reproductive isolation in plants is a product of the complex synergy between numerous pre- and post-zygotic barriers (Rieseberg and Willis 2007; Widmer et al. 2009). The present study was limited to several easily measurable isolating mechanisms. The contributions of gametic, habitat, temporal and other barriers in the maintenance of reproductive isolation within the series remain unaccounted for and await exploration.

Spatial crossability patterns

Geographical patterns seem obscure in crosses between O. meridionalis and O. rufipogon as well as between O. nivara and O. rufipogon. Crossability estimates are highly variable and both intra- and inter-regional combinations produce viable and non-sterile F1 hybrids (Online Resource 2). The lack of significant correlation between distance of parental origins and crossability suggest at most a weak (if any) spatial influence on the mating compatibility within and between the species of Oryza series Sativae in Asia Pacific. This complements the weak mantel correlation of genetic distance with geographic distance detected within and between the species of the series (Banaticla-Hilario 2012).

Crossability differences between sympatric and non-sympatric crosses are significant only for a few parental combinations but still hint at slightly stronger reproductive isolation under sympatric conditions. This indicates that sympatry generates a selection pressure against hybrids, favouring the evolution of stronger genetic barriers. This also agrees with the genetic differentiation detected in local populations (Banaticla-Hilario 2012) and supports the separation of O. nivara from O. rufipogon. Enhanced genetic barriers under sympatric conditions are a good indication that closely related taxa have evolved into separate species (Wu 2001).

Variations in F1 hybrids

F1 life cycle and phenology also differ between interspecific combinations of O. rufipogon with the two annual parental species. The asymmetry observed in the reproductive compatibility of O. meridionalis and O. rufipogon extends to F1 phenology as hybrids seem to inherit the photoperiod sensitivity or insensitivity of their maternal species. F1s between O. nivara and O. rufipogon flower much earlier than their parental accessions regardless of which species is the maternal or paternal parent. On the other hand, intraspecific O. rufipogon F1s exhibit prolonged periods of vegetative growth and flower much later than their parents.

The number of days from seeding to flowering, culm length and anther length are greater in the majority of the O. rufipogon intraspecific F1s compared to the parental accessions and other hybrids (Table 3; Fig. 5), conforming to the general assumption that hybrids between geographic races tend to be vegetatively superior to their parents (Rieseberg and Carney 1998). However, crosses between Australasian O. rufipogon populations produced F1s that were morphologically comparable to the parental O. rufipogon accessions (Fig. 5). The slightly different morphology displayed by the F1s of Australasian populations is unsurprising as this geographic population group is indeed genetically distinct (Banaticla-Hilario 2012; Waters et al. 2012) and can even be morphologically differentiated from the rest of O. rufipogon by the length of ligules, flag leaves, panicles, spikelets, awns and stigmas (Banaticla-Hilario 2012).

In interspecific combinations with O. meridionalis, Australasian and non-Australasian populations of O. rufipogon produced fractionally different offspring. The F1s with Australasian O. rufipogon as parents appeared more similar to O. meridionalis while the F1s with non-Australasian O. rufipogon as parents appeared more similar to the perennial species (Fig. 5). The chloroplast genomes of Australian O. rufipogon are more akin to O. meridionalis than to Asian O. rufipogon (Waters et al. 2012). The observed morphological disparities imply that many F1 phenotypic traits are cytoplasmically inherited. The variations in F1 morphology of different interspecific crosses can most likely be attributed to maternal or cytoplasmic effects, as seen in the hybrids between the indica and japonica groups of O. sativa (Li et al. 1997). Certain cytoplasmic genes that control hybrid sterility seem to influence the agronomic traits of F1 japonica hybrid rice (Wang et al. 1998). Crosses between O. nivara and O. rufipogon resulted in F1s that were more similar to the former, regardless of the geographic origin of the latter implying that a considerable portion of phenotype-determining genes are inherited predominantly from O. nivara.

Species coherence in terms of morphology is evident from the scatter plots of highly discriminating characters where all parental accessions and intraspecific F1s of O. rufipogon form a distinct cluster separate from the other parental species and interspecific F1s (Fig. 6). Conversely, similar to the observations of Rieseberg and Carney (1998), our interspecific hybrids display some character traits that are similar to their annual parental species and some that are intermediate between the two parents (Fig. 6). The poor morphological resolution of interspecific F1s causes difficulty in distinguishing the hybrid- from the non-hybrid plants of Asia Pacific Oryza series Sativae. Therefore, the use of more reliable molecular techniques such as SSR and single nucleotide polymorphism (SNP) genotyping in evaluating these hybrids is highly recommended.

Conclusions

A clearer perception of how post-pollination barriers operate within and between the Asia Pacific species of Oryza series Sativae has been provided. In the case of the studied taxa, post-zygotic barriers do not necessarily intensify with increasing genetic distance. Hybridization data is not a reliable basis in gauging taxonomic relationships among these wild rice species since gene flow patterns could either transcend or fall short of species limits.

Still, the results pose several implications for the evolution of Oryza series Sativae species in Asia and Australia. The constrasting levels of intraspecific crossability displayed by O. nivara (low) and O. rufipogon (high) correspond to the low intra-population diversity and high inter-population differentiation of the former and high intra-population diversity and low inter-population differentiation of the latter (Banaticla-Hilario 2012). This shows that the extent of intraspecific gene flow can influence population structure and genetic diversity patterns within species.

The modest crossability of O. rufipogon with the two annual species affirms the possibility of interspecific gene flow via secondary contact and can help explain the genetic similarities observed between O. nivara and O. rufipogon and between O. meridionalis and O. rufipogon from Australasia. Nevertheless, sympatric populations of O. nivara and O. rufipogon and of O. meridionalis and O. rufipogon tend to have stronger pre- and postzygotic barriers than non-sympatric populations. This tendency to evolve reinforced reproductive barriers in sympatry is characteristic of closely related but distinguishable species (Wu 2001) and implies species cohesiveness in the three Oryza taxa. Therefore, caution should be taken in re-introducing populations in the wild so that locally existing gene flow barriers are not disrupted. To prevent unwanted hybridization, the crossability patterns recognized in this study should be considered in layout planning for genebank regeneration activities. O. rufipogon should have wider spaces between accessions to maintain the genetic integrity of populations while such wide spacing might not be necessary for accessions of the inbreeding species O. nivara and O. meridionalis.

The compatibilities detected among the three wild Oryza species can be exploited in breeding programs for the improvement of cultivated rice. The highly crossable accessions and reproductively successful combinations identified in this study can be considered in future hybridization experiments. O. sativa L. is reportedly interfertile with O. nivara and O. rufipogon but poorly crossable with O. meridionalis (Lu et al. 2003). The fertile F1 hybrids of O. meridionalis and O. rufipogon can be used in introgressing useful genes from O. meridionalis to O. sativa.