For angiosperms, apomixis means asexual reproduction by seed [1]. It is strongly associated with hybridity and polyploidy, and molecular mechanisms responsible for it remain shrouded in complexity [2–4]. Apomixis involves the reprogramming of unreduced (2n) cells of the ovule, which thereafter follow a very different developmental trajectory than had the plant been sexual. Specifically, ovules of apomictic plants produce asexual totipotent cells. These form in the nucellus, chalaza or integuments, and embryos develop from them either directly (adventitious embryony) or after 2n embryo sac (ES) formation (gametophytic apomixis). Apomictic (2n) ES usually resemble sexual ES, but embryony in them occurs parthenogenetically and often precociously. Whether in sexual plants or apomicts, embryony is the result of epigenome modifications that begin as early as floral transition [5, 6].

Gametophytic apomixis is further divided into i) apospory, where the 2n aposporous ES (AES) forms from a cell of the nucellus, chalaza or rarely an integument, and ii) diplospory, where the 2n ES forms from an ameiotic megasporocyte (MMC). The formation of viable seed in apomicts requires the formation of functional endosperm, and this occurs pseudogamously or autonomously, i.e. with or without fertilization of the ES central cell, respectively. In adventitious embryony, a sexual ES with functional endosperm forms from which the developing adventitious embryo derives nutrients. The sexual embryo may survive and compete for nutrients with adventitious embryos [1, 7].

Apomixis in angiosperms occurs in polyploids or polyhaploids and is found in 31 of 63 orders (compiled from [2] using APG III nomenclature [8]). Though widespread, it occurs infrequently, being reported in only 223 genera (of about 14,000), 41 of which belong to the Poaceae. Of these, 24 belong to the Panicoideae, which is a large and ancient subfamily of grasses many members of which, including Sorghum L. (but not Zea L.), have undergone few chromosome rearrangements and no whole genome duplications since a whole genome duplication occurred 65 million years ago that differentiated grasses from other monocots [9–11]. Accordingly, Sorghum is an anciently diploidized paleotetraploid (n = 10). It is divided into five subgenera, Sorghum, Chaetosorghum, Heterosorghum, Parasorghum and Stiposorghum. Subgenus Sorghum includes perennial S. halapense Pers. (2n = 4× = 40), perennial S. propinquum (Kunth) Hitchc. (2n = 2× = 20), and annual S. bicolor (L.) Moench (2n = 2× = 20). The latter is divided into subsp. bicolor (domesticated grain sorghums), subsp. drummondii (stabilized derivatives between grain sorghums and their closest wild relatives), and subsp. verticilliflorum (formerly subsp. arundinaceum, wild progenitors of grain sorghum). Subspecies bicolor is further divided into five races, bicolor, guinea, caudatum, kafir and durra, and 10 intermediate races [12].

Low frequency AES formation occurs in several subsp. bicolor lines [13–17]. However, none of the reports provide convincing molecular or cytological evidence of parthenogenesis, and claims to the contrary have met with skepticism [18, 19]. In this respect, Gustafsson [20] reviewed evidence from several species that the 2n egg in an AES from a plant that rarely produces AES may not be capable of parthenogenesis, an opinion shared by Asker and Jerling [21]. Nevertheless, the interrelatedness of Panicoideae [22] suggests that the AES formation observed in S. bicolor may be symplesiomorphic with that observed in the fully functional aposporous Panicoideae.

In practice sexual and apomictic plants are differentiated by i) cytological analyses of ovule development [23], ii) progeny tests using morphological or molecular markers [24], and iii) flow cytometry of seed nuclei to identify distinguishing embryo to endosperm ploidy level ratios [25]. However, several less-distinct traits also differentiate many apomicts from their related sexuals. For example in diplosporous species of Tripsacum L. [26, 27] and Elymus L. [28], onset of 2n ES formation, relative to stage of ovule development, occurs prior to onset of meiosis in related sexuals. Whether this is a general phenomenon of diplospory has not been investigated. In aposporous apomicts, the potentially competitive sexual germline is usually terminated by apoptosis from the MMC stage to early sexual ES formation. AES formation is detected cytologically as early as the MMC stage to as late as ES maturation. Timing of apospory is not rigid, and much within species and within plant variation occurs [1, 20, 21, 29]. Likewise, parthenogenesis occurs prior to flower opening in many apomicts. This has been observed in Alchemilla L., Aphanes L., Taraxacum Cass., Wikstroemia Endl., Ochna L., Allium L., Chondrilla L., Hieracium L., Crepis L., Potentilla L., Poa L., Elatostema J. R. & G. Forst., Tripsacum, and Parthenium L. [20, 21].

In the present study, we determined onset timing of megasporogenesis (female meiosis) and sexual ES formation relative to stage of ovule development for 569 genotypes from three populations of S. bicolor. We also determined the frequency of AES formation for each genotype. The genotypes were then grouped according to AES frequency, and the groups were compared based on onset timing of megasporogenesis and sexual ES formation. The results suggest that the apospory program in S. bicolor heterochronically accelerates, relative to stage of ovule development, the onset of meiosis and sexual ES formation.

In grasses, a single ovule develops from the ovary placenta. Initially, the ovule primordium (young funiculus) grows inward and perpendicular to the inner ovary wall. As the ovule grows, the nucellus and integuments form and undergo anisotropic curvature downward and away from the developing style (Figure 1A). In the present study, ovule curvature values at specific germline stages (meiosis and early sexual ES formation) were determined and found to be less variable, likely more canalized, than ovule area values. As a result, curvature measurements were superior to area measurements in detecting differences among genotypes in onset timings of germline stages.

Meiosis and sexual ES formation occurred precociously, relative to stage of ovule development, in high AES-producing plants (Figure 5; Additional files 6, 7, 8). This was an unexpected result, and four possible explanations for its occurrence were considered. First, early onset of germline development may trigger apospory, especially in Sorghum, which, being a panicoid grass, may already be prone to apospory (24 Panicoideae genera contain aposporous species). However, many genotypes underwent early germline development but were not aposporous. Hence, while apospory was a good predictor of early germline development, the latter was a poor predictor of the former (Additional files 7, 8: compare ANOVA P and r² values for ovule curvature among F₂ and RIL clustered by apospory with those obtained for frequency of apospory among F₂ and RIL clustered by ovule curvature).

Second, meiotic instabilities due to recent hybridity may trigger apospory and early germline development. As noted above, a disproportionately high percentage of genotypes with >3% AES formation were hybridization-derived breeding lines. However, aposporous activity among the 150 genotypes tested (from 65 accessions) was not correlated with meiocyte abortion, even at P < 0.25. Hence, while hybridity may have increased the frequency of apospory, meiotic instability does not appear to be a factor.

Third, heterozygosity, due to recent hybridity, might trigger apospory and early germline development. If this were correct, we would expect apospory and early germline development to decline substantially during the production of the RIL population. However, apospory was present among the homozygous F₈ RIL at nearly the same frequency (5.0% of RIL had >3.0% AES formation) as in genotypes from the accessions (7.3%) and F₂ (7.7%). Thus, hybridity in S. bicolor may bring together different alleles that interact quantitatively to enhance aposporous activity, but heterozygosity does not appear to be important.

Fourth, the expression of an apomixis program in S. bicolor, though weak, may cause precocious reproduction, whether apomictic or sexual. This possibility best explains our observations. As noted above, apospory in a given genotype, even at the low frequencies observed herein, was a good predictor of early onset of sexual germline development. The implication is that even though the apospory program was too weak to induce consistent AES formation, it was strong enough to more consistently induce early onset of sexual germline development. While precocious aposporous and diplosporous ES formation have been documented in many apomicts [21, 26–29], to our knowledge the present report is the first to document what may be a controlled heterochronic acceleration of sexual germline development by apomixis. Studies using additional sexual plants and closely related facultative apomicts are required to determine if precociousness of sexual reproduction in facultative apomicts is a general phenomenon. For such studies, curvature measurements should be useful in quantifying stages of ovule development.

Phenological traits other than ovule development also differentiate some apomicts from related sexuals. Early flowering is one. In the Netherlands, peak flowering of apomictic Taraxacum occurred 5 and 10 d earlier than that observed for sympatric diploid sexuals on south and north facing slopes, respectively [31]. Early flowering in apomicts was also observed among 52 apomictic and 879 sexual angiospermous species in Sweden. Here, a significantly higher proportion of apomicts (compared to the proportion of sexuals) flowered in the early spring [21]. Early flowering was also observed in natural sympatric populations of sexual and apomictic Antennaria Gaertn., Boechera Á. Löve & D. Löve, and Elymus. For Antennaria, Boechera, Elymus as well as Tripsacum, flowering not only occurred earlier in the apomicts but tended to continue indefinitely when grown continuously in ideal greenhouse conditions. In contrast, more specific environments were required to induce flowering in related sexuals (JGC, field collection and greenhouse notes). These examples coupled with findings presented herein, of a precocious meiosis and sexual ES formation, suggest that sexual dimorphism in plants (systematic molecular, phenological or ontogenetic differences between male, female, sexual or apomict) may be more life-cycle-pervasive than previously recognized. Sexual dimorphism at the transcriptome level (mRNA extracted from young vegetative tissues) was recently reported between male and female Silene L. [32].

The precocity of temporally distinct life-cycle events (floral induction, apomeiosis, ES formation, and parthenogenesis) may have evolved independently in apomicts. However, Asker and Jerling [21] doubted this stating that a fitness-based rationale for such directional selection at different life-cycle stages is lacking. Alternatively, the evidence to date is consistent with the existence of an apomixis program that epigenetically controls, throughout the life cycle, onset timings of temporally divergent reproduction-related events (Figure 6). In cyclically apomictic animals, e.g., certain water fleas, aphids, flatworms, rotifers, gall wasps, gall midges, and beetles, favorable environments induce a greatly accelerated rate of reproduction through apomictic live-birth parthenogenesis. But when these same individuals encounter stress, the apomixis program is suppressed, and sexual reproduction, through the formation of quiescent and stress-tolerant eggs, occurs [33]. Tendencies toward a similar cyclical apomixis in plants have been reported. Where this has been studied, percentage sexual ES formation was highest when plants were grown in suboptimal conditions (as in cyclically apomictic animals). Examples include facultative apomicts of i) Boechera, where sexual ES formation was most frequent in stressed inflorescences [34], ii) Calamagrostis Adans., where sexual ES formation was most frequent in early-forming spikelets [35], iii) Ageratina Spach [36] and Limonium Mill. [37], where sexual ES formation was most frequent in plants exposed to cold stress, iv) Dichanthium Willem. [38–40], where sexual ES formation was most prevalent when these short-day plants were grown in long days, and v) Paspalum L. [41] and Brachiaria (Trin.) Griseb. [42], where frequency of sexual ES formation was highest for plants grown in conditions unfavorable for flowering.

The hypothesis that apomixis evolves repeatedly in eukaryotes by a hybridization or polyploidization induced genetic or epigenetic uncoupling of sexual stages, where some stages are discarded and others are fortuitously retained and re-coupled [2], has received serious consideration [3–5, 43]. However, a reliance on fortuity at the molecular level is a troubling component of this hypothesis, and the hypothesis in general is inconsistent with the observation that apomixis has failed to arise spontaneously (even once) among many tens of thousands of intra and inter-specific hybrids and amphiploids that have been produced artificially during the past 100 years. Herein, we suggest that the apparent uncoupling/recoupling process is not fortuitous but evidence of an ancient sex/apomixis switch (Figure 6) the molecular components of which have been retained, to a greater or lesser extent, in relatively few eukaryote lineages during evolution. Hybridization and polyploidization may occasionally epigenetically trigger the switch (from sex to apomixis or vice versa) but only in lineages that have retained, at the molecular level, a sufficient capacity for each mechanism. If this ancient alternatives hypothesis is correct, apomixis may be more complex than previously envisioned. It may be a life-cycle phenomenon, like sexual reproduction, that includes resetting the epigenetic clock each generation. Accordingly, apomixis in eukaryotes would share a common fundamental theme, i.e., the formation of unreduced and epigenetically reset parthenogenetically active cells from germline cells or closely associated cells (cells normally associated with sexual reproduction).

Similarities in the environmental control of the sex/apomixis switch between cyclically apomictic animals and facultatively apomictic plants that exhibit cyclical apomixis tendencies were recognized in the 1960s [39]. These similarities suggest that the unicellular common ancestor of plants and animals was cyclically apomictic or at least possessed processes by which cyclical apomixis could evolve by parallel evolution. In this respect, the precocious meiosis and sexual ES formation observed in the present study (Figure 5) may be regulated by the same epigenetic network that induces early flowering in apomicts, a reproductive step occurring much earlier in the life cycle, as well as precocious embryogenesis from parthenogenetic eggs [20, 21], a reproductive step occurring much later in the life cycle. Molecular studies are required to evaluate these possibilities.

Much variation was found among S. bicolor accessions in timing of germline development relative to ovary and ovule development. In this respect, ovule curvature appeared to be strongly canalized, and was more consistent than ovule area in predicting onset timing of specific germline events. AES formation was most prevalent in subsp. verticilliflorum and in the breeding lines of subsp. bicolor. It was uncommon in races of subsp. bicolor. Correlations between AES and AI were lower than expected, which suggests that additional factors are required for AES formation. Meiosis and sexual ES formation occurred precociously in genotypes with high AES frequencies. AES formation did not appear to be triggered by early onset of sexual germline development, meiotic instabilities or heterozygosity. Instead, a weakly expressed apomixis program in certain genotypes appeared to accelerate onset of reproduction, whether apomictic or sexual.

The present study adds onset of meiosis and sexual ES formation to onset of the vegetative/floral transition, apomictic ES formation, and parthenogenesis as processes that occur early in apomictic plants. The temporally diverse nature of these events suggests that an epigenetic memory of the apomixis status of the plant exists, which is maintained throughout the life cycle (Figure 6). In some plants, as in cyclically apomictic animals, this memory is degraded by reproductively marginal (stress-related) conditions. The result is an increased frequency of progeny that are produced sexually.

Apomictic plants share developmental and phenological traits characteristic of apomictic organisms from other kingdoms. These include i) a first division apomeiotic restitution (observed in many apomictic plants and animals), ii) parthenogenesis, iii) precocious onset of reproduction, and iv) tendencies toward cyclical apomixis. In cyclically apomictic animals and in plants exhibiting cyclical apomixis tendencies, sex is favored during stress and genetically reduced quiescent eggs are produced. In the same individuals, apomixis drives clonal fecundity during reproductively favorable conditions. The quiescent egg phase is skipped: cyclically apomictic animals, which produce quiescent eggs when reproducing sexually, undergo live birth, and the parthenogenetic eggs of apomictic plants produce embryos precociously. Whether apomicts from diverse kingdoms share molecular components of a conserved apomixis/sex switch is a question that awaits further elucidation. Such a finding would imply that apomixis is more ancient and more complex than previously envisioned.