Gamete fertility and ovule number variation in selfed reciprocal F1 hybrid triploid plants are heritable and display epigenetic parent-of-origin effects


  • Dorota Duszynska,

    1. Plant and AgriBiosciences Centre (PABC), School of Natural Sciences, National University of Ireland Galway, Galway, Ireland
    Search for more papers by this author
    • These authors contributed equally to this work.
  • Peter C. McKeown,

    1. Plant and AgriBiosciences Centre (PABC), School of Natural Sciences, National University of Ireland Galway, Galway, Ireland
    Search for more papers by this author
    • These authors contributed equally to this work.
  • Thomas E. Juenger,

    1. Section of Integrative Biology & Institute for Cellular and Molecular Biology, University of Texas, Austin, TX, USA
    Search for more papers by this author
  • Anna Pietraszewska-Bogiel,

    1. Plant and AgriBiosciences Centre (PABC), School of Natural Sciences, National University of Ireland Galway, Galway, Ireland
    2. Molecular Cytology, Swammerdam Institute for Life Sciences (SILS), University of Amsterdam, Amsterdam, the Netherlands
    Search for more papers by this author
  • Danny Geelen,

    1. Department of Plant Production, Faculty of Bioscience Engineering, University of Ghent, Ghent, Belgium
    Search for more papers by this author
  • Charles Spillane

    Corresponding author
    • Plant and AgriBiosciences Centre (PABC), School of Natural Sciences, National University of Ireland Galway, Galway, Ireland
    Search for more papers by this author

Author for correspondence:

Charles Spillane

Tel: +353 91 494148



  • Polyploidy and hybridization play major roles in plant evolution and reproduction. To investigate the reproductive effects of polyploidy and hybridization in Arabidopsis thaliana, we analyzed fertility of reciprocal pairs of F1 hybrid triploids, generated by reciprocally crossing 89 diploid accessions to a tetraploid Ler-0 line.
  • All F1 hybrid triploid genotypes exhibited dramatically reduced ovule fertility, while variation in ovule number per silique was observed across different F1 triploid genotypes. These two reproductive traits were negatively correlated suggesting a trade-off between increased ovule number and ovule fertility. Furthermore, the ovule fertility of the F1 hybrid triploids displayed both hybrid dysgenesis and hybrid advantage (heterosis) effects.
  • Strikingly, both reproductive traits (ovule fertility, ovule number) displayed epigenetic parent-of-origin effects between genetically identical reciprocal F1 hybrid triploid pairs. In some F1 triploid genotypes, the maternal genome excess F1 hybrid triploid was more fertile, whilst for other accessions the paternal genome excess F1 hybrid triploid was more fertile.
  • Male gametogenesis was not significantly disrupted in F1 triploids. Fertility variation in the F1 triploid A. thaliana is mainly the result of disrupted ovule development. Overall, we demonstrate that in F1 triploid plants both ovule fertility and ovule number are subject to parent-of-origin effects that are genome dosage-dependent.


Polyploid organisms frequently display serious reproductive defects, which can vary in severity between species (Henry et al., 2007). Differences in polyploid reproductive phenotypes are of evolutionary importance as polyploids act as bridges between lineages of different ploidy, and can also act as reproductive isolating barriers underlying speciation events (Spillane et al., 2002; Henry et al., 2007; Otto, 2007; Kohler et al., 2010). In this context, triploid plants play particularly important roles in terms of both genetic bridges and reproductive viability.

The effects of triploidy on fitness can vary dramatically across species and taxa. For instance, triploidy in humans is predominantly fatal and is the leading cause of human miscarriages (Sankaranarayanan, 1979; Egozcue et al., 2002; Kohler et al., 2010). Similarly, in some plant species (e.g. maize, poplar), triploidy is also not tolerated. However, in other plant (and some fish) species, triploidy is well tolerated, allowing for reproductively competent individuals (Adams & Wendel, 2005). As a result, some lineages, species, populations or varieties may be predominantly triploid, including many important crop species such as bananas, watermelons and citrus fruits (Stöck et al., 2002). In addition, many asexually reproducing plants (whether by apomixis or vegetative propagation) are triploids where sexual reproduction via meiosis is compromised (Spillane et al., 2004).

Even in species where polyploids can form successfully, reproductive defects may reduce the survival of the polyploidy lineage by affecting the transition to flowering, meiotic efficiency or gametogenesis (Comai et al., 2003; Chen, 2007). The effects of ploidy on reproduction are particularly severe in cases where the ploidy is odd-numbered (triploidy, pentaploidy). Indeed, in banana and other fruit crops, triploid infertility is a key agronomic trait for generating seedless fruits. However, triploidy can also constrain breeding programs as a result of inherent difficulties of crosses and reproduction involving triploid plants (Lora et al., 2011). Classic studies by Darlington, Blakeslee, McClintock and others demonstrated that the reduced fertility of triploids is largely the result of their tendency to generate univalents during meiosis and subsequent disruption of genome dosage in the gametes and offspring (Belling & Blakeslee, 1922; Lesley, 1928; Darlington, 1929; McClintock, 1929; Newton & Darlington, 1929; Satina & Blakeslee, 1937a,b; Avers, 1954). In extreme cases, this tendency to meiotic catastrophe can lead to complete sterility, for example in tomato (Lesley, 1928). However, triploid plants of some plant species are at least partially fertile, despite the problems encountered during meiosis and subsequent gametogenesis. In some instances, this is a result of the evolution of highly specialized pollen or ovule meioses that cause all univalents to migrate to a single gamete, as occurs in dog-rose (Rosa canina ) and bearded heath (Leucopogon juniperus) (Fagerlind, 1940; Smith-White, 1948), allowing the formation of eutriploid zygotes. In other species, including the model plant Arabidopsis thaliana, triploids show intermediate degrees of fertility (Henry et al., 2005, 2009).

Despite the long history of research into the reproduction of triploid plants, a number of key questions remain. A number of studies have shown that within a species different triploid individuals can vary in their fertility. However, systematic analysis of the impact of genetic variation on the fertility of polyploid plants remains limited (Wendel, 2000; Henry et al., 2005; Matsuoka et al., 2007). In particular, the extent of variation within and between species regarding the fertility of selfed triploids is not yet fully elucidated. Similarly, little is known regarding how hybridization between divergent genotypes of a species can differentially affect polyploid fertility (Baack & Rieseberg, 2007). Finally, the underlying mechanisms allowing some triploids to maintain fertility remain unknown. In particular, the mechanisms governing the tolerance of gametes of selfed F1 triploid plants to genome and/or allelic dosage effects are poorly understood (McClintock, 1984; Birchler et al., 2001; Comai et al., 2003; Henry et al., 2007).

The model plant A. thaliana provides an excellent system to investigate the fertility of both isogenic and hybrid F1 triploids, by facilitating dissection of the relative roles of triploidy (genome dosage) vs hybridity (allelic dosage). In addition, the generation of viable F1 triploids from reciprocal crosses of diploid and tetraploid genotypes allows for elucidation of parental genome dosage effects on fertility characteristics. In this study, we use a panel of self-fertilized triploid F1 plants (generated from interploidy crosses of different accessions of A. thaliana ) to determine the impact of natural variation on both ovule number and ovule fertility in the selfed F1 triploids. In addition, we investigate the impact of hybridization on F1 ovule number and ovule fertility. Finally, using reciprocal pairs of maternal and paternal genome excess F1 triploids, we investigated whether ovule number and ovule fertility in selfed F1 triploids are subject to genome-dosage dependent, parent-of-origin effects.

Materials and Methods

Plant material

Diploid accessions of Arabidopsis thaliana L. were obtained from NASC (CS22564–CS22659). 4x seeds originated from: Ler-0 – Ueli Grossniklaus (originally from Cold Spring Harbour Laboratory); Col-0 and C24 – Luca Comai (University of Washington), and Zurich – Ortrun Mittelsten-Scheid (Gregor Mendel Institute, Vienna); all were generated by colchicine doubling (Blakeslee, 1941). Sterilized seeds were sown on MS medium (Murashige & Skoog, 1962) and grown in a Percival Tissue Culture cabinet under a 16 : 8 h light : dark (21 : 18°C) regime (Boyes et al., 2001) and then grown in a randomized design on individual pots of soil (eight parts Westland compost (Dungannon, N. Ireland) : one part perlite : one part vermiculite) in growth chambers under fluorescent lamps at 200 μmol m−2 s−1. Plants were crossed by manual emasculation and cross-pollination before anthesis under a Leica MZ6 dissecting microscope and Dumostar No. 5 tweezers (Leica Microsystems GmbH, Wetzlar, Germany). Siliques were harvested for analysis 7 d postpollination. Euploidy was confirmed using a PA-1 ploidy analyzer (Partec GmbH) and Cystain UV Precise P kits (Partec GmbH, Münster, Germany), according to manufacturer's instructions.

Reproductive phenotypes

Confirmed triploid plants were allowed to self-pollinate and nearly mature siliques dissected. Five siliques were analyzed for each of three individual F1 plants, with both maternal and paternal genome excess F1 triploids generated for each accession tested (89 accessions × two cross directions × three independent plants × five siliques = c. 2670 siliques examined; the total numbers of ovules analyzed varied by genotype and is listed for each F1 triploid in Supporting Information, Table S1). Siliques were dissected and contents scored for percentage of unfertilized ovules (%U ), identified as very small white bodies (Meinke, 1994), and as a proportion of the total number of ovules (T ) whether fertilized or not (Fig. S1). Pollen viability was determined by Alexander staining of mature anthers (Alexander, 1969) and unviable pollen grains scored (n = 100 grains for each of three flowers from three biological replicates from each of 13 study accessions).

Statistical analyses

Percentage unfertilized ovules and T were analyzed using linear mixed models in PROC Mixed in SAS (Littel et al., 1996), with accession, cross-direction, and their interaction as fixed factors and plant as a random effect. In this framework, the contribution of parental genetic variation in the triploid traits (%U, T ) is indicated by the accession term; parent-of-origin effects are indicated by the direction term; and genetic variation in parent-of-origin effects are indicated by the interaction term. This model appropriately accounts for measurements of individual siliques nested within replicate plants as a random effect. The significance of terms was determined using standard type 3 analysis and F-ratios (Littel et al., 1996). To further explore significant interaction terms and to identify accessions with extreme parent-of-origin effects, we used the ‘slice’ option in SAS. Slicing is a statistical approach for testing simple main effects in the presence of interaction in factorial ANOVA: in our case, slicing allowed tests of parent-of-origin effects for specific accessions while still benefiting from the power of our global model.

In addition, we also explored natural genetic variation in the measured triploid reproductive characters by considering accession as a random effect in data split by cross-direction. The proportion of the variation explained by the single-dose contribution of the accessions was calculated as the ratio of the among-accession variance component to the total phenotypic variability in triploid traits. We estimated this proportion (analogous to a broad-sense heritability, H2) and 95% confidence limits using a percentile bootstrap approach and 10 000 bootstrap samples with the software H2boot (Phillips & Arnold, 1999). Single-dose genetic correlations among the triploid reproductive characters (in both cross directions) were estimated by the standard Pearson product-moment correlation of triploid line means. The significance of each genetic correlation and 95% confidence limits were determined using a percentile bootstrap approach and 10 000 bootstrap samples with the software resampling2 (Howell, 2007).


Fertility of F1 isogenic triploids of A. thaliana varies by genotype

Selfed isogenic F1 triploids of A. thaliana in the accession Col-0 are subfertile and form many ovules which remain unfertilized (Henry et al., 2005). To test whether ovule fertility of selfed isogenic F1 triploids varies by genetic background, we generated F1 triploids in four accessions. Given the low rates of heterozygosity in A. thaliana accessions (Horton et al., 2012), these triploids can be considered isogenic, as their parents differ only in their ploidy levels. We demonstrate that unfertilized ovules are also observed in the Ler-0, C24 and Zu genetic backgrounds, with 41–62% of ovules in the siliques remaining unfertilized (Fig. 1; Table S1). To test for parent-of-origin effects on ovule fertility, we compared selfed reciprocal F1 isogenic triploids containing either two maternal or two paternal genomes. No differences in ovule fertility were found between the reciprocal pairs (t-test, P > 0.05), indicating no parent-of-origin effects on %U within accessions (Fig. 1). Hence, in A. thaliana, genetic background affects ovule fertility of isogenic F1 triploids in a parent-of-origin independent manner.

Figure 1.

Ovule fertility of selfed F1 isogenic triploids in Arabidopsis thaliana by cross direction and parental genotype (genetic background). Bars represent mean percentage of unfertilized ovules (%U ) observed in selfed reciprocal F1 isogenic triploids in isogenic Col-0, Zu, C24 and Ler-0 genetic backgrounds. Error bars represent standard error (SE). Mean %U values were compared using a Student's t-test and no significant differences were found in %U between pairs of reciprocal F1 isogenic triploids (P-value < 0.05).

Ovule fertility of F1 hybrid triploids can display hybrid dysgenesis or hybrid advantage (heterosis)

To investigate the effect of hybridity on the fertility of F1 hybrid triploids, a tetraploid Ler-0 line (LLLL) was reciprocally crossed to 89 diploid accessions (AA). Where the tetraploid line was used as the female parent (a 4x × 2x cross) the resulting F1 hybrid offspring had a maternal genome excess (2m:1p). In the reciprocal cross direction, the tetraploid Ler-0 line was used as the male parent (a 2x × 4x cross), producing a paternal genome excess F1 hybrid triploid (1m:2p). The crossing scheme used is illustrated schematically for clarity (Fig. S1). All 166 F1 hybrid triploid lines were validated as eutriploid by flow cytometry (data not shown). Each of the F1 hybrid triploid genotypes were grown in triplicate and allowed to self-fertilize before dissection of siliques to determine relative proportions of unfertilized ovules (%U) in each line. The total numbers of ovules analyzed varied by group and is listed for each F1 hybrid triploid in Table S1.

Similar to the analysis of the selfed F1 isogenic triploids, all selfed F1 hybrid triploids produced many unfertilized ovules within their siliques (Figs 2, S2). However, in selfed F1 hybrid triploids the percentage of unfertilized ovules (%U ) was affected by the interaction of diploid parental genotype with the cross direction (Table 2) (accession, degrees of freedom (df) = 88,325, F-value = 2.30, P < 0.0001; cross direction, df = 1,320, F-value = 0.92, P = 0.3384; accession × cross direction, df = 75,321, P < 0.0004). The impact of interploidy cross direction on %U (from either maternal or paternal excess F1 hybrid triploids) depends on the genotype of the diploid parent. In the most extreme cases of ovule infertility (i.e. %U observed in selfed F1 hybrid triploids generated from the 2x × 4x cross direction), 74.2% of ovules of selfed 1m:2p F1 hybrid triploids generated from diploid Tamm-27 were unfertilized, while only 39.7% of those formed by selfed 1m:2p F1 hybrid triploids generated from diploid Pro-0 were unfertilized (Fig. 2; Table S2). Depending on genotype, the F1 hybrid triploids generated in the two interploidy cross directions displayed %U values that were either higher or lower than the %U values observed in the F1 isogenic triploids generated in the Ler-0 background (Fig. 1). This indicates that hybrid dysgenesis (higher %U) or hybrid advantage (lower %U) effects on ovule fertility can be generated in F1 hybrid triploids depending on the genotype.

Figure 2.

Ovule fertility (%U ) of selfed Arabidopsis thaliana F1 hybrid triploids arranged by parental 2x genotype. Bar-plots of %U are shown for each of 78 pairs of reciprocal F1 hybrid triploids showing the differences between paternal excess (2x × 4x; 1m:2p; white bars) triploids and maternal excess (4x × 2x; 2m:1p; black bars) triploids. Statistically significant differences between reciprocal F1 hybrid triploid pairs are indicated by asterisks as described in the 'Materials and Methods' section (*, P < 0.05; **, P < 0.01; ***, P < 0.001). Error bars represent the standard error.

Ovule fertility of F1 hybrid triploids can display parent-of-origin effects

Parent-of-origin effects can be detected in reciprocal F1 offspring which are genetically identical yet display different phenotypes depending on the parent-of-origin of the genetic material (whether maternally or paternally transmitted). It should be noted that both the F1 isogenic triploids and the F1 hybrid triploids (generated from reciprocal 4x × 2x and 2x × 4x crosses) are genetically identical (LLA vs ALL genotypes), but differ according to whether they have two maternally derived (2m:1p) or two paternally derived (1m:2p) sets of chromosomes. Such a reciprocal crossing design allows for a test of parent-of-origin effects between reciprocal triploids.

In marked contrast to the isogenic triploids, we determined that the percentage of ovule infertility frequently displayed parent-of-origin effects between genetically identical F1 hybrid triploids generated in the two cross directions. Such parent-of-origin effects on the percentage of ovule infertility were statistically significant, at P < 0.05 or less, for 13 out of 83 pairs of reciprocal F1 hybrid triploids (Fig. 2). The parent-of-origin effect seen on ovule fertility in reciprocal F1 hybrids was not biased to either the 2m:1p or 1m:2p F1 hybrids. Six F1 hybrid triploids were significantly more ovule-fertile as maternal excess 2m:1p triploid plants (e.g. CS22941), while seven displayed the opposite trend and were more fertile in the paternal excess 1m:2p triploid (e.g. Fei-0). Not only is the observed range of F1 hybrid triploid fertility greater than that found in isogenic F1 triploid plants, but the ovule fertility of interaccession reciprocal F1 hybrid triploids can also vary significantly via parent-of-origin effects. To the authors' knowledge, this is the first evidence of a parent-of-origin effect on any aspect of fertility in an F1 hybrid triploid plant.

Ovule number per silique shows genotype-dependent effects in F1 triploids

Previous studies in A. thaliana have shown differences in ovule number between diploid accessions Cvi-0 and Ler-0, and their reciprocal F1 hybrid offspring (Alonso-Blanco et al., 1999). To determine whether total ovule number displayed variation across F1 triploids of different genotypes, we calculated the total sum of the fertilized and unfertilized ovules (T ) produced per silique for each isogenic or hybrid triploid (Table S3). In F1 hybrid triploids, the total number of ovules was affected by the interaction of diploid parental genotype and the cross direction (Table 2; Fig. 4) (accession, df = 1323, F-value = 2.44, P < 0.0001; cross direction, df = 1,320, F-value = 0.25, P = 0.6151; accession × cross direction, df = 75,321, P < 0.0002). The impact of cross direction (i.e. maternal or paternal excess F1 hybrids generated from interploidy crosses) on total ovule number depends on the genotype of the diploid parent. The isogenic F1 triploids displayed some variation in total ovule number between accessions (Fig. 3), which indicates that there are accession-specific differences between isogenic triploids in relation to the total number of ovules they produce per silique. However, a wider range of variation for total ovule number was observed across hybrid triploids (Fig. 4), with the 2m:1p Ct-1 F1 hybrid triploid displaying the highest value (79 ovules per silique) and the 2m:1p Sq-1 F1 hybrid triploid displaying the lowest value (33 ovules per silique) (Table S3). We conclude that the genetic background of an F1 hybrid triploid can affect the number of ovules formed in its carpels, and that hybrid dysgenesis or hybrid advantage can affect total ovule number in F1 hybrid triploids depending on their genotype.

Figure 3.

The ovule number (T) per silique of selfed Arabidopsis thaliana F1 isogenic triploids varies by cross direction as well as parental genotype. Bars represent mean ovule number per silique in four reciprocal F1 isogenic triploid lines (Col-0, Zu, C24 and Ler-0). Error bars represent standard error (SE). Mean values were compared using Student's t-test. Significant differences are marked by asterisks (P-value < 0.05).

Figure 4.

Total ovule number (T) per silique of selfed Arabidopsis thaliana F1 hybrid triploids arranged by parental 2x genotype. Bar-plots of T are shown for each of 78 pairs of reciprocal F1 hybrid triploids showing the differences between paternal excess (2x × 4x; 1m:2p; white bars) triploids and maternal excess (4x × 2x; 2m:1p; black bars) triploids. Statistically significant differences between reciprocal F1 hybrid triploid pairs are indicated by asterisks as described in the 'Materials and Methods' section (*, P < 0.05; **, P < 0.01). Error bars represent standard error.

Ovule number per silique displays parent-of-origin effects in F1 triploids

To determine whether total ovule number in selfed reciprocal F1 triploids was subject to any parent-of-origin effects, we compared ovule numbers of each pair of reciprocal triploids for both the isogenic and hybrid genotypes. In the isogenic triploids, significant parent-of-origin effects were seen on the total number of ovules for the Col-0 and Zu isogenic triploids (Fig. 3). For Col-0, a higher ovule number occurs in the selfed paternal excess 1m:2p F1 triploid compared with the 2m:1p F1 triploid, while use of the Zu accession displayed the opposite effect. Hence, in two genetic backgrounds, there is a parent-of-origin effect on total ovule number in selfed reciprocal F1 triploids which are genetically isogenic.

Variation in ovule number was also observed across F1 hybrid triploids generated from crossing different diploid accessions with the Ler-0 tetraploid. As with %U, the total ovule number (T) varied significantly between F1 hybrid triploids produced from parents of different genetic backgrounds, and significant parent-of-origin effects on ovule number were observed in 12 reciprocal crosses (Fig. 4). As with %U, there was little bias of the ovule number data towards F1 hybrids generated from one or the other cross direction. The most extreme parent-of-origin effect occurred for the Sq-1 reciprocal F1 hybrid triploids, where the 2m:1p triploid had a mean of 33.3 ovules per silique, yet the 1m:2p triploid had a mean of 62.1 (Table S3). Overall, our results provide the first evidence for parent-of-origin effects on ovule number in reciprocal triploids of plants, in both isogenic and hybrid F1 genotypes.

Fertility of F1 hybrid triploids is genetically variable and involves a trade-off with ovule number

We used a variance component analysis to estimate the genetic and environmental contributions to variation in the F1 triploid reproductive characters (%U and T ). We observed significant genetic variation in both %U and total ovule number based on significant variance components in mixed models (P < 0.0001) and bootstrap estimates and confidence intervals. Broad-sense heritabilities ranged from 0.17 and 0.35 for %U and 0.14 and 0.33 for total ovule number T (Table 1). Lower heritability for both traits was observed when diploid accessions were the maternal parent (Table 1). The pattern of genetic correlation between the two triploid F1 reproductive characters (%U, T) within and among parental cross directions was also investigated (Fig. 5). We observed significantly positive genetic correlations between %U and T within either maternal or paternal excess cross directions (Fig. 5, Table 2). This indicates that a genetic trade-off exists between the number of ovules (T) and the likelihood that an ovule will be fertilized in selfed triploids (%U). Finally, despite highly similar patterns of genetic correlation between the two reproductive traits within a cross direction, we found generally low genetic correlations between either trait across maternal and paternal excess crossing directions (Table 2). This result concurs with the observation of significant accession × cross direction interaction effects on the two reproductive traits (%U, T) measured.

Table 1. Broad-sense heritabilities, percentage of unfertilized ovules (%U ) and the total number of ovules per silique (T) demonstrating heritable variation in fertility of selfed F1 Arabidopsis thaliana paternal excess (1m:2p) and maternal excess (2m:1p) F1 triploids
Trait2x × 4x4x × 2x
  1. 95% bootstrap confidence limits are shown.

% U 0.17 (0.04–0.30)0.35 (0.22–0.48)
T 0.15 (−0.01 to 0.31)0.33 (0.18–0.47)
Table 2. Genetic correlations (rg) between percentage of unfertilized ovules (%U ) and the total number of ovules (T) and the effect of parental origin in selfed reciprocal F1 hybrid triploid plants of Arabidopsis thaliana
Trait 1Trait 2Genetic correlation (rg)
  1. 95% bootstrap confidence limits are shown.

%U (4n × 2n )T (4n × 2n )0.51 (0.35–0.64)
%U (2n × 4n )T (2n × 4n )0.45 (0.20–0.65)
%U (4n × 2n )%U (2n × 4n )0.12 (−0.14 to 0.36)
T (4n × 2n )T (2n × 4n )0.16 (−0.04 to 0.36)
%U (2n × 4n )T (4n × 2n )0.10 (−0.09 to 0.30)
T (2n × 4n )%U (4n × 2n )0.37 (0.16–0.55)
Figure 5.

Scatterplot of genetic correlations between ovule number (T) and fertility (%U ) within and across reciprocal F1 Arabidopsis thaliana triploids based on accession means. Histograms showing frequency (y-axis) of different mean %U (or mean T ) (x-axis) in maternal (2m:1p; 4x × 2x) and paternal excess (1m:2p; 2x × 4x) cross directions are presented on the diagonal. Significance levels are displayed in Table 2.

Differences in F1 hybrid triploid fertility are not the result of hybrid dysgenesis effects that are manifest in diploids

As differences in total ovule number and ovule fertilization percentages in F1 hybrid triploids could be a consequence of hybrid dysgenesis between Ler-0 and some accessions (Bomblies, 2006; Bomblies & Weigel, 2007), we determined whether F1 hybrid diploids display any evidence of similar effects. Hence, to confirm that the observed parent-of-origin effects in the F1 hybrid triploids is specific to genotypic effects seen only in the polyploid state, as suggested by the significance of accessions × cross direction interaction effects (Table 2), %U was also measured in selfed reciprocal F1 hybrid diploids. These F1 hybrid diploids were generated from reciprocal crosses between diploid Ler-0 and six diploid accessions representing the extremes of the %U fertility defects observed in the F1 hybrid triploids. All five displayed high ovule fertility (low %U ) when selfed, indicating that no hybrid dysgenesis effect on ovule fertility is present in F1 hybrid diploids (Fig. 6). Only minor parent-of-origin effects were seen in the F1 hybrid diploids, with Ga-0 and Wei-0 both showing modest but significant increases in fertility in one cross direction (Mann–Whitney test, P < 0.05). Overall, this indicates that genotypic and parent-of-origin effects on ovule fertility seen in the F1 hybrid triploids result from a genome dosage effect that is principally manifest in F1 triploids and not F1 hybrid diploids.

Figure 6.

Minor effects on ovule fertility (% unfertilized ovules, %U) in F1 hybrid F1 Arabidopsis thaliana diploids. Bars represent mean %U produced by selfed F1 hybrid diploids, generated by crosses between diploid Ler-0 and five different diploid accessions. Error bars represent standard error.

High proportion of unfertilized ovules in F1 triploids is not correlated with pollen viability

Selfed F1 triploids in A. thaliana generate many aneuploid gametes (Henry et al., 2005). To determine the extent to which aberrant female and/or male gametogenesis are linked to altered %U fertility in reciprocal F1 hybrid triploids, the viability of pollen grains produced by selfed F1 triploids ranging from the highest to the lowest %U was determined. Each group of F1 triploids chosen for analysis is indicated by a numbered horizontal line. In all reciprocal F1 hybrid triploids tested, there were no significant reductions in pollen viability. This was the case for the female excess F1 triploids with high %U (Fig. 7a, accession group 1); female excess F1 triploids with low %U (Fig. 7a, accession group 2); male excess F1 triploids with high %U (Fig. 7a, dataset 3); male excess F1 triploids with low %U (Fig. 6a, accession group 4); the F1 triploids displaying the greatest %U differences between reciprocal hybrid F1 triploids (Fig. 7a, datasets 5–6); and F1 triploids displaying the least differences (Fig. 7a, accession group 7). In a few cases, pollen viability was slightly reduced (e.g. Cvi-0). While this indicates some degree of disrupted pollen viability in these F1 hybrid triploids, a poor correlation was found between pollen inviability and %U (r = −0.24 in 1m:2p triploids; r = −0.263 2m:1p triploids, P > 0.05) (Fig. 7b). For example, the triploids generated from Fei-0 have the largest difference in %U of any pair of reciprocal triploids, but both have pollen viability > 97% (Fig. 7a). Furthermore, the proportion of viable pollen always exceeded 50% and hence was in high excess over the number of available ovules. This strongly suggests that there is no causal link between pollen viability and F1 triploid fertility in A. thaliana.

Figure 7.

Pollen viability is not correlated with altered F1 triploid fertility. (a) Pollen viability of 13 reciprocal Arabidopsis thaliana F1 hybrid triploids. Data for male excess triploids are represented by black bars, data for female excess by white bars; numbered horizontal bars indicate different categories of triploids as described in the text. Error bars represent standard error. (b) Pollen viability correlated against %U (% unfertilized ovules) and association found to be nonsignificant (r = −0.24 in the 2x × 4x cross direction and r = −0.262 in the 4x × 2x cross direction, > 0.05). Lines of best fit are shown. (c) Alexander staining indicates extent of unviable pollen, visible as pale, shrunken grains (arrow) in 4x × 2x (A) and 2x × 4x (B) hybrid triploids, and anther from diploid showing only viable pollen (plump, purple grains, C). Scale bars, 100 μm.

To further confirm that parent-of-origin effects are unlikely to be a result of pollen effects, we determined the sizes of pollen grains from F1 triploids with the largest parent-of-origin effects on %U (Fei-0 and Zdr-1) together with Bur-0 and Pro-0 as controls (Fig. 1). We found that pollen size ranges (which mirror ploidy) are greater in all F1 triploids than in a diploid control (with many smaller aneuploid pollen and occasional larger grains). However, all pollen size profiles were similar, and there was no distinct difference between pollen size or counts produced by reciprocal F1 triploids (Fig. S3). This was observed regardless of whether the reciprocal F1 triploids varied in their %U. Overall, these results rule out extensive pollen abortion resulting from widespread meiotic catastrophe in selfed F1 triploids of A. thaliana. We suggest that the high proportion of unfertilized ovules identified in F1 hybrid triploids most likely occurs as a result of aberrant megagametogenesis or ovule development.


In this study, we analyzed the reproduction of F1 triploid plants of different hybrid and isogenic genotypes of A. thaliana. We found that fertility of F1 triploid plants is subject to natural variation which can improve or diminish ovule fertility and number, traits which we demonstrate are heritable and negatively correlated. Unusually for triploid plants, pollen viability and size are relatively unaffected and appear to play a minimal role in determining the fertility of the selfed F1 triploid plants. The fertility of selfed F1 triploid plants is, however, strongly influenced by parental genome dosage from the original cross, revealing a potentially epigenetic parent-of-origin effect on polyploid fertility that has hitherto been unreported.

Significant heritable variation in ovule fertility and total ovule number in selfed F1 triploids of A. thaliana

Selfing of triploid plants can lead to different reproductive outcomes ranging from complete sterility (e.g. banana) to high fertility (e.g. Aster; Avers, 1954). Triploid fertility defects are the result of meiotic pairing problems generating aneuploid gametes which cannot undergo fertilization to generate viable progeny. To date, the extent of natural variation for ovule fertility of selfed F1 triploids has not been explored, in either isogenic or hybrid backgrounds. In this study, we demonstrate that there is considerable genetic variation in ovule fertility and ovule number in selfed F1 triploids (isogenic and hybrid) of A. thaliana across many different genotypes (Figs 1–3).

Hybrid advantage and hybrid dysgenesis effects on ovule fertility and ovule number in selfed F1 hybrid plants

While heterosis and hybrid dysgenesis effects are typically detected in F1 hybrids, the relative role of hybridity vs gene dosage effects in heterosis type-effects in polyploids remains unclear (Birchler et al., 2003, 2006). We demonstrate that, in contrast to isogenic Ler-0 triploids, F1 hybrid triploids (generated from crosses of different accessions to tetraploid Ler-0) often display either hybrid advantage or dysgenesis effects in relation to ovule fertility or number relative to Ler-0 F1 isogenic controls (Figs 1–3, 6). To our knowledge this is the first demonstration of hybrid advantage/dysgenesis effects on ovule traits in selfed F1 hybrid triploids. The presence of transgressive variation in the offspring of interploid crosses may help to explain the persistence of plant populations containing individuals of different ploidy. Such situations have been observed in genera such as Spartina and Tragopogon and may be widespread (Soltis et al., 2004; Salmon et al., 2005).

Parent-of-origin effects on ovule number and fertility in reciprocal F1 hybrid triploids

Parent-of-origin effects on phenotypes (e.g. %U, T ) can be revealed in reciprocal F1 triploid hybrids which are genetically identical apart from having alleles inherited via the male or female gametes (Fig. S1). Our study reveals that many of the isogenic reciprocal F1 hybrid triploid pairs displayed parent-of-origin effects on both ovule fertility and number (Figs 1–3, 6). Such parent-of-origin effects between isogenic reciprocal F1 triploid lines may represent epigenetic effects, wherein mitotically and/or meoitically heritable changes in gene function cannot be explained by changes in DNA sequence (Russo et al., 1996). Such epigenetic effects could involve changes to DNA methylation or chromatin, or alternatively could result from cytoplasmic effects/factors that somehow persist throughout plant development to manifest their effects in the next round of gametogenesis. While parent-of-origin dependent genome-dosage effects can affect the development of F1 triploid seeds obtained from interploidy crosses (Lin, 1984; Johnston & Hanneman, 1995; Scott et al., 1998), it has never previously been shown that such parent-of-origin effects can affect later stages of plant development, including beyond the transition to flowering. In addition, our results demonstrate novel genome-dosage-dependent parent-of-origin effects on ovule number, whereby ovule development genes (CUC1-2, AFF5; Colombo et al., 2008) involved in ovule primordia may be dysregulated. Possible explanations for such late-acting parental effects could include differences in the tolerance to genome dosage imbalance in ovules of the reciprocal triploids, or the fact that the set of triploid chromosomes in the 2m:1p vs 1m:2p plants are not epigenetically equivalent. We conclude that, while previous studies have demonstrated parent-of-origin effects on phenotypes of F1 seeds generated from interploidy crosses (Scott et al., 1998; Berger & Chaudhury, 2009; Erilova et al., 2009), parent-of-origin effects in reciprocal F1 triploid plants are also manifest throughout the life cycle.

Trade-off between ovule formation and fertility in triploids

Many wild plants produce higher ovule numbers in order to take advantage of variable pollen supplies (Burd et al., 2009). Despite the importance of ovule number to both fitness and agriculture, the relationship between ovule number per gynoecium and ovule fertility remains largely unknown. Our study has revealed a novel relationship between ovule number and the likelihood that an ovule will be fertilized in F1 triploids (Fig. 7, Table 2), suggesting that gynoecia with smaller numbers of ovules produce more fertile ovules. Possible explanations could involve reduced vigour of F1 triploid A. thaliana and consequent reduction in available photosynthates. Alternatively, ovules formed by F1 triploids may require higher resource investment during their development and may compete with each other (e.g. Greenway & Harder, 2007).

High and uniform viability of pollen produced by F1 A. thaliana triploids

Our results from the analysis of pollen from selfed F1 triploids strongly suggest that the differences in triploid fertility described are a result of female gametophytic rather than male gametophytic effects, because pollen grains produced by F1 triploids are predominantly viable (Fig. 7a,b). The fertility of F1 triploid A. thaliana plants is therefore differentially affected by defects during the formation of the male and female gametophytes, which confirms a trend observed in Blakeslee's classic work in Datura (Belling & Blakeslee, 1922; Satina & Blakeslee, 1937b). For example, Blakeslee observed lagging chromosomes in over 50% of megaspore meioses in triploid Datura, compared with just 2.5–4.5% following the first and second meiotic divisions in the pollen mother cell (Satina & Blakeslee, 1937a). However, the differences between the male and female gametophytes of A. thaliana are even more striking with respect to the high pollen fertility, which was usually around 90% (Fig. 7) compared to with c. 50% observed in Datura. Datura has pollen grains more tolerant of triploidy than many other species (e.g. tomato or maize), although not as high as that observed in artificial triploid varieties of Aster, which also reached 90% (Avers, 1954). It should be noted that many classic reports used plants grown under field conditions, the pollen of which might have been more prone to temperature stress than the plants grown under controlled conditions that were used in this study. We conclude that A. thaliana may have a high tolerance to triploid pollen meiosis and altered microgamete chromosome dosage, and that this leads to the observed disparity between the effects of ovules and pollen on overall fertility.

Epigenetic effects in speciation and polyploid crops

One consequence of the variation we have described in reproductive success of different F1 triploids is that the effects of polyploidy could affect reproductive success of different populations within a species in different ways (Rick & Baron, 1953). Such variation among genotypes has been observed in crops – for example, triploid tomato strains vary in the proportions of aneuploidy offspring they produce (Rick & Baron, 1953; Rick & Notani, 1961). The presence of widespread parent-of-origin-dependent phenotypes in reciprocal F1 triploids could explain phenomena reported from the plant breeding literature, such as reports that desirable cucumber triploids can only be generated if the excess genome is paternally inherited, and show decreased pollen fertility and seedlessness (Mackiewicz et al., 1998). More generally, our data also provide a useful model for understanding transgressive variation and potentially heterosis/hybrid dysgenesis in triploid plants (Birchler et al., 2003; Lippman & Zamir, 2007), as more profound effects on reproduction were observed in the F1 hybrid than in isogenic triploids.


Polyploidy is a key process in angiosperm evolution (Adams & Wendel, 2005). Many studies of the genetic consequences of autopolyploid have been conducted in plants, but the factors affecting the phenotypes of triploid plants have rarely been assessed. Data derived from pairs of reciprocal F1 triploid A. thaliana demonstrate key roles for natural variation in the balance between ovule number and ovule fertility in determining triploid fertility. Curiously, both of these traits are influenced by genome-dosage-dependent, parent-of-origin-specific effects in both hybrid and isogenic selfed F1 triploid plants. These lend support to models assuming that evolution of polyploids may involve changes to epigenetic regulation as well as genetic dosage effects (Rapp & Wendel, 2005; Rapp et al., 2009). By contrast with many triploids, pollen viability is relatively unaffected in A. thaliana, so the success of female gametogenesis may be the key determinant of F1 triploid fertility in A. thaliana. These novel findings have implications for our understanding of the relative reproductive abilities of maternal vs paternal excess F1 triploids in both natural and agricultural environments.


This work was funded through grant funding to C.S. from Science Foundation Ireland (SFI; grants 02/IN.1/B49 and 08/IN.1/B1931), the Irish Department of Agriculture, Fisheries and Food (grant RSF 07-534), the Irish Research Council for Science, Engineering and Technology (grant R 12345) and the National University of Ireland Thomas Crawford Hayes Research Fund. The support of COST Action FA0903 (HAPRECI) on ‘Harnessing Plant Reproduction for Crop Improvement’ is acknowledged.