Apospory and Diplospory in Diploid Boechera (Brassicaceae) May Facilitate Speciation by Recombination-Driven Apomixis-to-Sex Reversals

Apomixis (asexual seed formation) in angiosperms occurs either sporophytically, through adventitious embryony, or gametophytically, where an unreduced female gametophyte (embryo sac) forms and produces an unreduced egg that develops into an embryo parthenogenetically. Multiple types of gametophytic apomixis occur, and these are differentiated based on where and when the unreduced gametophyte forms, a process referred to as apomeiosis. Apomeiotic gametophytes form directly from ameiotic megasporocytes, as in Antennaria-type diplospory, from unreduced spores derived from 1st division meiotic restitutions, as in Taraxacum-type diplospory, or from cells of the ovule wall, as in Hieracium-type apospory. Multiple types of apomeiosis occasionally occur in the same plant, which suggests that the different types occur in response to temporal and/or spatial shifts in termination of sexual processes and onset timing of apomeiosis processes. To better understand the origins and evolutionary implications of apomixis in Boechera (Brassicaceae), we determined apomeiosis type for 64 accessions representing 44 taxonomic units. Plants expressing apospory and diplospory were equally common, and these generally produced reduced and unreduced pollen, respectively. Apospory and diplospory occurred simultaneously in individual plants of seven taxa. In Boechera, apomixis perpetuates otherwise sterile or semisterile interspecific hybrids (allodiploids) through multiple generations. Accordingly, ample time, in these multigenerational clones, is available for rare meioses to produce haploid, intergenomically recombined male and female gametes. The fusion of such gametes could then produce segmentally autoploidized progeny. If sex re-emerges among such progeny, then new and genomically unique sexual species could evolve. Herein, we present evidence that such apomixis-facilitated speciation is occurring in Boechera, and we hypothesize that it might also be occurring in facultatively apomictic allodiploids of other angiospermous taxa.

Apomixis (asexual seed formation) in angiosperms occurs either sporophytically, through adventitious embryony, or gametophytically, where an unreduced female gametophyte (embryo sac) forms and produces an unreduced egg that develops into an embryo parthenogenetically. Multiple types of gametophytic apomixis occur, and these are differentiated based on where and when the unreduced gametophyte forms, a process referred to as apomeiosis. Apomeiotic gametophytes form directly from ameiotic megasporocytes, as in Antennaria-type diplospory, from unreduced spores derived from 1st division meiotic restitutions, as in Taraxacum-type diplospory, or from cells of the ovule wall, as in Hieracium-type apospory. Multiple types of apomeiosis occasionally occur in the same plant, which suggests that the different types occur in response to temporal and/or spatial shifts in termination of sexual processes and onset timing of apomeiosis processes. To better understand the origins and evolutionary implications of apomixis in Boechera (Brassicaceae), we determined apomeiosis type for 64 accessions representing 44 taxonomic units. Plants expressing apospory and diplospory were equally common, and these generally produced reduced and unreduced pollen, respectively. Apospory and diplospory occurred simultaneously in individual plants of seven taxa. In Boechera, apomixis perpetuates otherwise sterile or semisterile interspecific hybrids (allodiploids) through multiple generations. Accordingly, ample time, in these multigenerational clones, is available for rare meioses to produce haploid, intergenomically recombined male and female gametes. The fusion of such gametes could then produce segmentally autoploidized progeny. If sex re-emerges among such

INTRODUCTION
The genus Boechera (Brassicaceae) evolved about 2.5 Myr ago (Mandakova et al., 2015) and is closely related to Arabidopsis (Bailey et al., 2006;Rushworth et al., 2011). It encompasses c. 83 primarily inbreeding sexual diploid taxa (Li et al., 2017), many of which have relatively narrow geographic ranges. Boechera also includes hundreds of genomically distinct diploid, triploid and tetraploid hybrids that are partially to fully sterile sexually. These hybrids produce most of their seeds through apomixis (without meiotic recombination, chromosome reduction or fertilization), but sexually derived seeds are also occasionally produced (Aliyu et al., 2010). This dual capacity, to produce seeds sexually and apomictically, is called facultative apomixis, and it is characteristic of most if not all angiospermous apomicts (Asker and Jerling, 1992).
Most Boechera taxa belong to a well-supported western North American clade (Alexander et al., 2013), the distribution of which extends from northern Mexico to the Arctic with outlying populations (mostly apomictic and polyploid) in Greenland and around the Great Lakes and the St. Lawrence River. Another clade of nine taxa, previously assigned to the genus Borodinia (Alexander et al., 2013), is here included in Boechera due to the recent discovery of inter-clade hybridization (Windham et al., field observations). The latter are distinctive in being sparsely pubescent and restricted to forested regions of eastern North America and the Russian Far East. Boechera s.l. is, by far, the largest genus of tribe Boechereae, a morphologically disparate group that includes seven other genera whose phylogenetic affinity only became apparent through recent chromosomal and molecular analyses. Indeed, the primary defining characteristic of Boechereae consists of a reduction in chromosome base number from n = 8 to n = 7 (Al-Shehbaz, 2003). Evidence suggests that the n = 8 Boechereae ancestor entered North America from Asia about 5 Mya via the Bering land bridge. The chromosome base number reduction likely occurred thereafter by multiple translocations (Mandakova et al., 2015).
Frequent hybridization with or without homoeologous recombination (Kantama et al., 2007) explains the proliferation of apomictic alloploid Boechera (Beck et al., 2012;Windham et al., 2015;Li et al., 2017), but how the sexual diploids originate is less obvious. The traditional view is that they arise by range expansion and speciation along ecological gradients (Alexander et al., 2015;Li et al., 2017). However, the slow pace of such speciation is inconsistent with the large numbers of rare sexual diploids described for this youthful genus. Here we provide a cytological and theoretical framework that addresses this question.
The pioneering study of female meiosis (megasporogenesis) and female gametophyte formation in Boechera (Bocher, 1951) was motivated by observations of 2n pollen formation. This, plus two subsequent studies of 2n pollen forming Boechera (Naumova et al., 2001;Taskin et al., 2004), revealed meiotic first division restitutions that produced dyads of 2n spores in ovules and anthers. On the male side, both spores formed 2n pollen. On the female side, one 2n spore degenerated and the other developed into a 2n gametophyte (Taraxacum-type diplospory). This limited embryological sampling led to an incorrect notion that 2n and 1n pollen in Boechera are diagnostic of apomixis and sex, respectively (Roy, 1995;Windham and Al-Shehbaz, 2006). More thorough sampling in recent years has provided a clearer picture of apomixis development in Boechera (Carman, 2007;Carman et al., 2015). Certain accessions of Boechera microphylla (B. imnahaensis × yellowstonensis) were found to be highly apomictic despite apparently normal male and female meioses (Mateo de Arias, 2015). In these plants, functional pollen grains form from all four 1n microspores. However, on the female side, all meiotically produced spores generally degenerate, and a 2n gametophyte forms adventitiously from a nucellar cell of the ovule wall (Hieracium-type apospory).
To better understand the unusual pervasiveness and origins of multiple apomixis types in Boechera, we expanded our taxonomic sampling to 64 accessions representing 44 operational taxonomic units (OTUs). Our sampling includes sexual and apomictic taxa that span the Boechera phylogeny (Alexander et al., 2013), and it represents a mix of taxa traditionally treated as species as well as recently discovered but as yet unpublished entities. Hereafter, published names are used for the diploid sexual species. However, the apomictic hybrids are identified by genome composition as found in the Boechera Microsatellite Website (BMW) http://sites. biology.duke.edu/windhamlab/ (Li et al., 2017). We show that both apospory (normal male and female meioses with female sexual development failing thereafter) and diplospory (first division restitution male and female meioses) occur frequently in Boechera and are widely dispersed across the genus. Based on these findings, we provide a possible explanation for the origins of rare and allelically poor sexual endemics that are often encountered in habitats otherwise populated by allelically complex apomicts.

Plant Materials
Cytological analyses were performed using floral buds taken from plants growing in native habitats, plants transplanted from native habitats, or plants grown from seeds (Supplementary Table S1). Seeds were placed on moist filter paper, stratified at 4 • C for 21 days, and planted. Potted seedlings or transplants were grown in 600 mL cone-shaped (68 mm diameter × 255 mm tall) pots or 350 mL square (85 mm wide × 95 mm tall) pots that contained Sunshine Mix #1 potting soil (Sun Gro Horticulture Canada Ltd., Vancouver, BC, Canada). Vernalization was accomplished by cold incubation (4 • C) for 10-12 weeks with minimal lighting (8/16 day/night photoperiod). Vernalized plants were transferred to controlledenvironment greenhouses or growth chambers that maintained a 16/8 h day/night photoperiod using supplemental light provided by a combination of cool white florescent bulbs, incandescent bulbs, and high-pressure sodium-vapor lamps. These provided a minimum photosynthetic photon flux of 400 µmol m −2 s −1 at the tops of the canopies. Day/night temperatures were maintained at 22/16 • C, and plants were watered regularly with a dilute solution (250 mg L −1 ) of Peters Professional 20:20:20 fertilizer (Scotts, Maryville, OH, United States).

Allelic Diversity and Geographic Distributions of Sexual Diploids
For sexual diploids, taxonomic names, specimen numbers, locations of origin, and allele lengths for 13 single-locus microsatellite (simple sequence repeat, SSR) loci (A1, BF3, B6, B9,  Frequencies by taxon of ovules exhibiting sexual tetrads, sexual or Taraxacum-type diplosporous dyads, Hieracium-type aposporous gametophytes, or Antennaria-type diplosporous gametophytes (accession numbers correspond to those in Supplementary Table S1). Tetrad, dyad, and Antennaria-type diplosporous gametophyte frequencies per accession sum to 100%. Aposporous gametophyte frequencies are listed separately (red bars). These develop adventitiously while meiotic tetrads form and degenerate. Median (Med.) numbers of SSR alleles observed among homozygous samples of each sexual accession (Supplementary Table S2) are shown (Med.) as are numbers (No.) of correctly staged ovules analyzed per accession. (B) A hypothesis of evolutionary cycling between hybridization induced apomixis and apomixis facilitated reticulate evolution of new sexual species. Blue box: taxa with mostly 2n pollen with some reduced and shrunken pollen, which suggests meiotic anomalies due to recent interspecific hybridization but with a transition from diplospory to apospory occurring in some taxa possibly due to early genome diploidization events associated with infrequent selfing (15-21); red box: taxa with mostly fertile reduced pollen coupled with apospory and sexual tetrad formation and degeneration, which suggests more extensive genome diploidization with a gradual restoration of sexual fertility; green box: sexually fertile anthers and pistils with no cytological evidence of apomeiosis.
B11, BF15, B18, B20, B266, C8, E9, I3, and I14) were downloaded from the BMW. To minimize inclusion of apomicts (mistakenly collected as sexual diploids), specimens were excluded if they were heterozygous for any of the 13 loci, yielded data for less than six of the 13 loci, or represented taxa with less than six homozygous specimens. Taxa meeting these criteria were then ranked based on median and mean numbers of alleles per locus (population level allelic polymorphism). This variability was used to identify putative sexual or apomictic ancestors.

RESULTS
Clearing and mounting of whole pistils using BBDP (Crane and Carman, 1987) was efficient for high throughput analysis of megasporogenesis and gametophyte formation. Each pistil contained, depending on species, from 40 to 200 ovules (Al-Shehbaz et al., 2010). When pistils were mounted horizontally (c. 16 per slide), 20-30% of their ovules were in sagittal orientation, which permitted efficient analyses of MMC origins as well as details of dyad, tetrad and early gametophyte formation.

Diplospory and Apospory Are Common in Boechera
In most sexual angiosperms, the mature female gametophyte is a seven celled (eight nucleate) structure (Polygonum type) that forms from a genetically reduced megaspore of a meiotic tetrad (Johri et al., 1992). Accordingly, the consistent observation of the following four phenomena was taken as strong evidence for near-obligate to obligate sexual reproduction: (i) meiotic tetrad formation, (ii) absence of aposporous gametophytes at the meiotic tetrad stage, (iii) absence of vacuolate enlargement or endomitotic activity in a dyad obtained from a MMC, and (iv) vacuolate enlargement of the surviving megaspore of a meiotic tetrad coupled with endomitotic activity. These criteria were consistently observed in the ovules of 23 of the 64 accessions analyzed (Figure 1A, 42-64). In these accessions, the tetrad to functional megaspore stage lasted c. 2 d, which permitted tetrads to accumulate in rows along the placentae (Figures 2A-C). Additional photomicrographs diagnostic of sexual ovule development (meiotic tetrads and vacuolate 1-to 2-nucleate gametophytes forming from surviving megaspores of meiotic tetrads) are shown in Supplementary Figures S1A-P for eight of the sexual taxa identified herein.
Ovules from 21 of the 64 accessions analyzed (Supplementary Table S1, 16 OTUs) generally underwent Taraxacum-type diplospory ( Figure 1A, 1-21). Here, a first division meiotic restitution occurred, which was followed by a mitotic-like second division to produce a dyad of 2n megaspores (Figures 2D,E). The 2n gametophyte then developed from the chalazal most spore (Figure 2D), and the micropylar-most spore degenerated. The dyad stage terminated Taraxacum type diplosporous megasporogenesis and, as with sexual megasporogenesis (resulting in tetrads), a pause in development preceded gametophyte formation. These pauses allowed diplosporous dyads to accumulate in rows, like sexual tetrads, along the ovary placentae (compare Figure 2A with Supplementary Figure S2A). Additional photomicrographs diagnostic of Taraxacum-type diplosporous ovule development (vacuolate 1to 2-nucleate unreduced gametophytes forming from surviving megaspores of unreduced apomeiotic dyads) are shown in Supplementary Figures S2A-F for six of the Taraxacum-type diplosporous taxa identified herein. Also shown are unreduced microspore dyads (Supplementary Figure S2G), which are commonly produced in the anthers of diplosporous Boechera (Bocher, 1951;Naumova et al., 2001).
Aposporous gametophytes generally replaced all four megaspores of meiotic tetrads in 27 of 64 accessions, which represented 19 OTUs (Figures 1A, 15-41, 2H-K). Photomicrographs diagnostic of apospory (degenerating meiotic tetrads being replaced by vacuolate 1-to 2-nucleate aposporous gametophytes) are shown in Supplementary Figures S2H-O for seven aposporous taxa. Because we scored reproduction from the dyad to 2-nucleate gametophyte stages, our apospory frequencies are probably underestimates. This is because some ovules scored as sexual in the dyad to early tetrad stages would have likely produced aposporous gametophytes had they been fixed at a later date. Apospory and diplospory occurred together in seven of the 27 aposporous accessions (Figures 1A, 2F).
In angiosperms, the nucellus develops by periclinal divisions of subepidermal cells of the funiculus, and the cells of the epidermis divide anticlinally to accommodate this nucellar enlargement. The periclinal divisions produce columns of nucellar cells. The central column extends from the middle of the chalaza to the distil most position at the micropylar epidermis (see narrow white lines, Figures 2D,E,G,H,L and Supplementary Figures S1I,J, S2B,D,K,L,N). The distil cell of the central column enlarges to produce the archesporial cell. In some angiosperms, enlarging archesporial cells divide mitotically to produce a parietal cell that separates the archesporial cell from the nucellar epidermis (Johri et al., 1992). In our sexual and apomictic taxa, parietal cells formed in 10-20% of the ovules, and they were observed from the MMC stage until they degenerated during early gametophyte formation (Figures 2B,E,G,H and Supplementary Figures S1C,D,G,L,O,  S2C,I,K-M). Histochemical evidence suggests that abnormal meioses can also produce parietal-like cells in Boechera (Rojek et al., 2018).

Evidence for Homoeologous-Recombination-Driven Reticulation
To evaluate possibilities of apomixis-to-sex reversions in allodiploid Boechera, we searched the BMW for sexual endemics with limited geographic distributions and limited allelic variable (listed at the bottom of Supplementary Table S2). One of these, B. mitchell-oldsiana, is endemic to a 4 km stretch along the rim of Hells Canyon in northeastern Oregon. This location is within the center of diversity of two prominent sexual taxa, B. retrofracta and B. sparsiflora. Mean and median numbers of alleles per locus in the homozygous BMW samples for B. mitchelloldsiana were 1.4 and 1, respectively (Supplementary Table S2). According to traditional views, B. mitchell-oldsiana, with its low allelic variability, could represent an ancient, nearly extinct sexual species that has experienced a genetic bottleneck followed by a weak comeback. Alternatively, it may have evolved from a single sexual species along an ecological gradient by directional selection. Then again, it may have evolved by reticulate evolution via a recombination-driven apomixis-to-sex reversion (Figure 4). If by directional selection, from a single species, most of its alleles should be found within single plants of its ancestral sexual species. However, if it evolved recently by apomixis-facilitated reticulation, its alleles should be found equally distributed between two ancestral sexual species, and local apomicts formed by hybridizations between these putative parental species should possess all or nearly all of the B. mitchell-oldsiana alleles.
As expected for an apomixis-facilitated reticulation, B. mitchell-oldsiana alleles were nearly evenly distributed between the two putative sexual parents, B. retrofracta and B. sparsiflora, and neither parent alone appeared to be capable of providing all of the needed alleles ( Table 1, allele columns of putative sexual ancestors). In contrast, each of three local apomictic hybrids contained nearly all of the B. mitchell-oldsiana alleles ( Table 1, allele columns for the three apomicts). The microsatellite-genotyped apomictic B. retrofracta × sparsiflora hybrids in the BMW represent only a small fraction of hundreds of such hybrids in this region from which B. mitchell-oldsiana may have evolved.
To reacquire meiotic stability after interspecific hybridization, apomictic Boechera must have undergone genome modifications that enhance chromosome pairing and recombination (diploidization). Since progeny of near-obligate apomicts are usually clonal and genetically identical to their mothers, well established allodiploid Boechera apomicts, which are also facultatively sexual, should have ample time (even hundreds of years) to sooner or later simultaneously produce 1n (genomically recombined) male and female gametes. In contrast, nonapomictic species hybrids are generally sterile, and these usually die without reproducing (Dobzhansky et al., 1977).
When allodiploid apomicts facultatively produce progeny by production and union of genomically recombined 1n gametes, a 50% reduction in homoeologous chromosome regions occurs. This is accompanied by a compensating increase in homologous (and homozygous) chromosome regions (Figure 4). With each additional autogamous generation, a 50% decrease in remaining homoeologous regions occurs. After several generations of selfing, each interspersed with perhaps multiple generations of apomixis, allodiploid apomicts should become sufficiently diploidized (Figure 4) for successful and efficient meioses to occur. Their chromosomes at this point represent chiasmagenerated composites of alternating homozygous sections of the homoeologous genomes of their parents (Sybenga, 1996;Carman, 2007). This process is analogous to recombinant inbred line (RIL) production where multiple generations of selfing produce new chromosomes consisting of alternating segments of the original parental chromosomes. Chromosome painting studies provide evidence that such inter-genomic recombination in apomictic Boechera is extensive (Kantama et al., 2007;Koch, 2015).
If recombinational loss of parental chromosome regions eliminates alleles responsible for one apomixis type over another, or for apomixis in general, then new apomictic or sexual plants with uniquely recombined genomes may evolve FIGURE 4 | Process whereby a facultatively aposporous allodiploid may produce new genomically unique sexual species consisting of alternating sections of homologous chromosomal regions from its homoeologous parental genomes. One homoeologous chromosome pair is represented. Black and gray chromosome regions are homoeologous. On average, remaining homoeology is decreased by 50% with each facultative autogamous generation (recombination driven autoploidization). Regions of strict homology, where chiasma formation is likely, are more common in homologous than in homoeologous regions. Hence, recombination probabilities should increase in subsequent autogamy formed generations. Fortuitous loss or silencing of apomixis-causing alleles is likely in some lineages. For reproduction in these lineages to continue, reversion to sexual reproduction must occur. This may happen gradually or rapidly, with plants regaining complete sexual fertility after several generations of recombination. (Figure 4). Such processes could explain the existence of a B. imnahaensis × yellowstonensis accession that is mostly diplosporous and another accession of the same combination that is mostly aposporous (Figure 1A, 19, 27). It is noteworthy that many aposporous hybrids contain a B. microphylla clade genome (B. thompsonii, B. imnahaensis, or B. yellowstonensis), which suggests that the B. microphylla clade may be predisposed to switch from diplospory to apospory. Our data also suggest that tendencies toward apospory may persist for many sexual generations ( Figure 1A, note high frequencies of tetrad formation with widely varying frequencies of apospory). While unreduced pollen is occasionally observed among apomicts of other angiospermous families, as well as among sexual plants (Asker and Jerling, 1992;Carman, 1997), the correlations between diplospory or apospory and 2n or 1n pollen, respectively, are unique to Boechera, and these correlations add to the uniqueness of the Boechera agamic complex.
The cytogenetic data available for the plants investigated herein (e.g., production of fertile 1n pollen) support the hypothesis of gradual, reticulation-driven shifts from recently evolved (sexually sterile or semisterile) diplosporous apomicts ( Figure 1A, 1-14), to plants that produce 1n and 2n pollen and exhibit diplospory, apospory and sex (Figure 1A, 15-21), to plants that produce mostly 1n pollen and exhibit mostly apospory and sex (Figure 1A, 22-41), and finally to completely sexual plants that produce 1n pollen (Figures 1A, 42-64,B). It should be noted that only a very small percentage of progeny, if any, in each hybrid combination might fortuitously undertake this evolutionary route. In this respect, the vast majority of seeds produced by apomictic hybrids are genetic clones of the mother plant. Hence, while apomixis to sex reversions may on occasion occur for a given hybrid combination, the parental apomictic hybrid remains happily apomictic. The definitive test for verifying this process would be to observe it firsthand. As noted above, both diplosporous B. exilis × thompsonii and aposporous B. imnahaensis × yellowstonensis produce about 4% of their seeds sexually. Accordingly, sexual gametophyte formation frequencies among sexually produced progeny (off types) could be determined. If segmental diploidization (Figure 4) and apomixis-to-sex reversions occur, they should be detectable within 2-4 generations. Another approach would be to genotype rare sexual endemics and their sexual and apomictic neighbors using phylogenetically stable markers. If a rare sexual endemic evolved recently from another sexual plant, most of its genetic markers should be similar to its progenitor. However, if it evolved by a recombination driven apomixis-to-sex reversal, then most or perhaps all of its molecular markers should be found in neighboring apomictic hybrids. In turn, these hybrids should contain near equal numbers of alleles from two distinct sexual parents, as was observed for B. mitchell-oldsiana herein (Table 1). Interestingly, B. mitchell-oldsiana exhibits a low frequency of apospory ( Figure 1A, 41), which is consistent with a putative apomixis-to-sex origin.
The B. mitchell-oldsiana germplasm analyzed here is unique among SSR genotyped diploids. Specifically, it's geographic distribution is restricted to a few flourishing populations, within 4 km of each other, in a single Oregon county. Similarly restricted populations of diploids have been reported, but only a few samples of SSR genotypes are available for them. It will be interesting to conduct analyses similar to that shown in Table 1 as additional rare sexual diploids are more thoroughly SSR genotyped.
Given the documented diversity of apomictic hybrids in Boechera [over 400 unique genomic combinations reported by Li et al. (2017)], it is evident that the association between apomixis and hybridization in this youthful genus is strong and that apomixis arises quickly following the amalgamation of divergent, mostly self-pollinating lineages. Likewise, if reversions from apomixis to sex require only a few successful sexual generations (Figure 4), then the entire process could reasonably occur in nature within a few decades. This would include (i) hybridization of sexual diploids, (ii) an homoeologous hybrid apomixis phase, (iii) a weakly apomictic segmental allodiploid phase, and (iv) a fully diploidized fledgling sexual endemic phase with early interspecific hybridizations of its own (Figure 1B). Variably repetitive patterns of microsatellite markers, as observed in the BMW (Li et al., 2017), could be explained by such a rapid recombination-driven speciation. Corresponding values are presented for its three most allelically similar sexual taxa (putative progenitors), B. retrofracta (retr), B. sparsiflora (spar), and B. puberula (pube), and for three allelically similar triploid apomictic hybrids, the parents of which include one or more of the putative ancestors (MS346, B. retrofracta × sparsiflora × sparsiflora; FW133, B. rectissima × retrofracta × sparsiflora; FW1042, B. cusickii × puberula × retrofracta). Numbers in the heading next to sexual taxa indicate number of homozygous lines in the BMW that were tallied to obtain frequency data. High frequency alleles from the two putative parents are shaded. For the three apomicts, allele presence values are from single plants. This highlights the possibility that sexual B. mitchell-oldsiana may have evolved from a single interspecific apomictic hybrid. Numbers of additional diploids that contain each allele are also listed (a measure of allele rarity). Locations are those from which BMW DNA samples were obtained. CAN, Canada; United States abbreviations as commonly used.

Apomixis Types May Simply Reflect Temporal and Spatial Variations in Termination of Sexual Development and Onset of Gametophyte Formation
Apomixis in plants (Asker and Jerling, 1992), animals (Suomalainen et al., 1987), and protists (Bilinski et al., 1989) involves three single-cell processes: termination of sexual development, production of unreduced spores or eggs, and parthenogenesis where spores or eggs reinitiate the life cycle without syngamy. It has been hypothesized that these seminal events of apomixis are anciently polyphenic with the corresponding seminal events of sexual reproduction, and that eukaryotes in general have more or less retained, during evolution, abilities to switch from one reproductive mode (phenism) to the other (Carman et al., 2011;Hojsgaard et al., 2014;Albertini et al., 2019). Accordingly, onset timings and locations of unreduced gamete formation, which in angiosperms requires gametophyte formation, could be the event that defines apomixis types in angiosperms (Battaglia, 1989;Carman, 1997).
If this hypothesis is correct, then apomixis types are not dependent on apomixis-type-specific mutations per se but on genetically controlled temporal and spatial variations in the induction of unreduced gametophyte formation. Drought and heat stress can switch facultatively diplosporous Boechera from mostly apomeiotic dyad formation to mostly meiotic tetrad formation (Mateo de Arias, 2015). Hence, some of the variability in dyad to tetrad ratios observed among diplosporous accessions (Figure 1A, 1-21), especially those fixed in the field (Supplementary Table S1, Windham collections), may have been caused by variations in the weather prior to field collections.
In certain ovules of the present study, sexual development was terminated prior to meiosis and was immediately replaced by unreduced gametophyte formation. This Antennaria-type diplospory occurred while integuments were still budding (Figures 2L,M). Likewise, unreduced gametophyte formation also followed the termination of sexual development during early meiotic prophase, which defines Taraxacum-type diplospory (Figures 2D,E), and shortly after meiosis in aposporous Boechera FIGURE 5 | Timing of sexual program termination may determine apomixis type in Boechera. In Antennaria-type diplospory, the sexual program aborts early, meiosis fails completely, and the gametophyte (ES) forms directly from the megaspore mother cell (MMC). In Taraxacum-type diplospory, the sexual program aborts during meiotic prophase I. Restitution of the first meiotic division (the reductional division) then occurs, two unreduced spores form, and the ES generally forms from the chalazal most spore. In Hieracium-type apospory, sexual reproduction can be terminated as early as early meiosis or as late as early ES formation, with unreduced ESs forming adventitiously from sporogenous nucellar cells.
( Figures 2H,J). That multiple types of apomixis occur in the same plant is evidence that timing and location of unreduced gametophyte formation dictates apomixis type (Figure 5). Interestingly, high frequency shifts between types of apomixis, as well as between sexual and apomictic development, have been induced in sexual and apomictic Boechera through pharmacological treatments that affect stress response pathways and DNA methylation (Gao, 2018).

Apomixis and Speciation, a Reappraisal
Facilitating the origins of genomically unique sexual species and genera runs counter to long held opinions concerning the involvement of apomixis in evolution. Historically, biologists considered apomixis, as well as wide hybridization and polyploidy, as antitheses of speciation (Darlington, 1939;Stebbins, 1971;Van Dijk and Vijverberg, 2005). Clearly, these processes block the selection-based shifts in allele frequencies thought to be required for gradual speciation along ecological gradients (Mayrose et al., 2011). However, studies now implicate reticulation as a prominent player in speciation (Carman, 1997;Rieseberg, 1997;Martis et al., 2013;Sochor et al., 2015;Payseur and Rieseberg, 2016;Vargas et al., 2017;Hojsgaard, 2018). In this respect, the immortality conferred by apomixis to allodiploid Boechera should provide them with unlimited time for rare recombinations to occur and for sexually fertile species, which possess multi-species-recombinant genomes, to evolve (Figure 4). To date, only a few cases of apomixis to sex reversions have been reported (Chapman et al., 2003;Domes et al., 2007;Horandl and Hojsgaard, 2012;Ortiz et al., 2013;Hojsgaard et al., 2014). However, this could change if the speciation mechanism described herein is found to be of more general occurrence among agamic complexes.
Establishment of apomixis-to-sex founder plants, like the recombinational events required to generate them, are probably rare, and this rarity could explain the low levels of allelic variability observed among some of the sexual diploids of Boechera (Supplementary Table S2). Also, since geographic ranges of apomicts often exceed those of their sexual progenitors (Bayer, 1997;Hojsgaard et al., 2014), newly formed apomixis-tosex founder populations could reasonably be allopatric with their most recent sexual ancestors but sympatric with clones of their immediate apomictic parents.

CONCLUSION
The unique situation in Boechera of self-pollinating, aposporously fertile allodiploids that facultatively produce 1n eggs and sperm may facilitate reversions from apomixis to sex. In fact, multiple genomically unique sexual species could in theory evolve from the same allodiploid, the divergent genomes of which would contain different assortments of homozygous segments of the original parental genomes (Figure 4). In this manner, apomixis may serve as an effective springboard in stabilizing reticulate evolution processes (Carman, 1997). Since allodiploidy increases rates of homoeologous recombination (Dewey, 1984;Wang, 1989;Grandont et al., 2014;Poggio et al., 2016), diploidization possibilities should be enhanced. Accordingly, the aposporous allodiploid Boechera identified herein are well suited for studying this putative speciation mechanism. While occurring at a much slower pace, this process could also occur among polyploid apomicts. Here, the process would originate in allodiploids that form from allotetraploid apomicts by haploid parthenogenesis.

ACKNOWLEDGMENTS
The authors thank Michelle Jamison, Devin Wright, Bryan Cox, John Carman Jr., and George Hampton II for assistance in collecting, growing, and preparing Boechera samples for cytology.