RH17 restricts reproductive fate and represses autonomous seed coat development in sexual Arabidopsis

In higher plants, the life cycle alternates between two distinct generations, with the sporophyte as the dominant generation and the gametophytes being reduced to a few cells only. The transitions between generations are of crucial importance and need to be thoroughly regulated. The formation of the female and male gametophytic lineages takes place in specialized reproductive organs of the flowers, which are the ovules and the anthers, respectively. In sexual plants, a single sporophytic cell typically specifies in the subepidermal ovule tissues to give rise to the functional megaspore (FMS) as the first cell of the female gametophytic lineage (Schmidt et al., 2015). This megaspore mother cell (MMC) is committed to a meiotic fate. Three of the four meiotically derived megaspores undergo apoptosis. The surviving FMS subsequently goes through three rounds of mitosis in a syncytium followed by cellularization. In Arabidopsis, like in the majority of angiosperms, a Polygonum-type mature gametophyte is formed comprising four specialized cell types (Schmid et al., 2015). These are the two female gametes (egg and central cell), two synergids, which are important for pollen tube attraction and double fertilization, and three antipodal cells. Similarly, the male reproductive lineage initiates development with the selection of a single sporophytic cell in the anther, termed the pollen mother cell (PMC) (Borg et al., 2009). Four microspores are derived from meiosis of the PMC. Two subsequent mitotic divisions lead to the formation of two sperm cells engulfed in a vegetative cell. In sexual plants, double fertilization of the female gametes with one sperm cell each initiates the coordinated development of the embryo, endosperm and seed coat. The zygote resulting from the fertilized egg cell first elongates before an apical cell and a basal cell are formed through an asymmetric division (Hove et al., 2015). The apical cell then gives rise to most of the embryo with the exception of its basal end; the basal cell differentiates to form the suspensor. Embryo development thereby proceeds through defined developmental stages, with the formation of a globular embryo, which subsequently traverses through heart and torpedo stages to give rise ultimately to the mature embryo (Hove et al., 2015).

In addition to sexual reproduction, asexual reproduction through seeds (apomixis) is also common in plants. From an evolutionary point of view, the co-existence of sexual reproduction and apomixis is of great interest. Apomixis has classically been regarded as a dead end of evolution. However, alternative hypotheses have recently emerged, including the possibility of a switch of the mode of reproduction in response to environmental stress (Albertini et al., 2019; Carman et al., 2019; Schmidt, 2020). Furthermore, apomixis has great potential for applications in agriculture, as it leads to formation of offspring that is genetically identical to the mother plant (Barcaccia and Albertini, 2013; Conner and Ozias-Akins, 2017). Engineering of apomixis in crops would therefore enable the fixation of superior genotypes. However, a precise understanding of the molecular programs governing plant reproduction will be required to achieve this.

Developmentally, different types of apomixis are recognized. First, sporophytic and gametophytic apomixis are distinguished based on the origin of the embryo (Bicknell and Koltunow, 2004; Koltunow and Grossniklaus, 2003). In gametophytic apomicts, the clonal embryo is derived from the egg cell. During diplospory, an apomictic initial cell forms instead of the MMC and alters meiosis to give rise to the gametophytic cell lineage. In contrast, during apospory an additional sporophytic cell, the aposporous initial cell (AIC), specifies adjacent to the MMC. It adopts a gametophytic fate without undergoing meiosis. In both cases, after gametogenesis the egg cell gives rise to the embryo in the absence of fertilization (parthenogenesis). In contrast, fertilization is required for endosperm formation in the majority of apomicts, but autonomous endosperm development also occurs. Unlike in gametophytic apomixis, in sporophytic apomicts asexual embryos form directly from sporophytic cells of the ovule tissues enclosing the sexual gametophyte. This frequently leads to polyembryony. Furthermore, apospory occasionally leads to polyembryony resulting from the formation of both the sexual and the apomictic reproductive lineages, derived from the MMC and the AIC, respectively, in the same ovule (Schmidt, 2020).

Formation of additional embryos per seed can be observed during sexual reproduction in some taxa. In gymnosperms, as well as in Orchidaceae and Poaceae, they derive from irregular divisions of the developing embryo, typically from the suspensor (Kishore, 2014). Alternatively, cell type mis-specification during gametogenesis can give rise to a second egg cell, mostly instead of one synergid cell, as has been described for French bean (Kishore, 2014). In most sexual species, including Arabidopsis, polyembryony is very rare. Nevertheless, in twin mutants two to three embryos per seed originate from the proliferation of suspensor cells (Vernon and Meinke, 1994), whereas in lines lacking the function of ALTERED MERISTEM PROGRAM 1 supernumerary egg cells are formed (Kong et al., 2015). In addition, formation of two embryos per seed occurs at low frequencies in septuple mutants deficient in all members of the gene family encoding for KIP-related proteins (ICK/KRP) (Cao et al., 2018). KRP genes act redundantly during germline formation and are involved in controlling the entry into meiosis (Zhao et al., 2017). Therefore, the embryos likely originate from egg cells formed in supernumerary meiotically derived gametophytes. In contrast, polyembryony from apospory-like development has thus far not been described for Arabidopsis.

Nevertheless, components of apomixis have been observed in sexual species (Schmidt et al., 2015), involving elements of autonomous seed development. During sexual reproduction, both endosperm and seed coat development typically stay repressed in the absence of fertilization, a process governed by the activity of Polycomb group proteins (Figueiredo and Köhler, 2018). These act in multimeric complexes to repress the activity of target genes by introducing histone methylation marks (Figueiredo and Köhler, 2018; Mozgova et al., 2015). Interestingly, Polycomb repressive complex 2 (PRC2) also represses genes involved in auxin biosynthesis, and auxin is sufficient to induce autonomous development of endosperm and seed coat (parthenocarpy) (Figueiredo et al., 2016; Figueiredo and Köhler, 2018). Moreover, features of apospory or diplospory have been described for certain mutants of core meiotic genes or genes involved in epigenetic regulatory pathways, which control gene expression based on DNA methylation or the activity of small RNAs (Hand and Koltunow, 2014; Lora et al., 2019; Nakajima, 2018; Schmidt et al., 2015). Phenotypes closely resembling apospory have also been reported for lines carrying a mutant allele of the RNA helicase MNEME (MEM) (Schmidt et al., 2011).

RNA helicases are becoming increasingly recognized as crucial for germline development (Schmidt, 2020). They are required for various aspects of RNA metabolism, including RNA stabilization, degradation and translation, and are involved in small RNA pathways (Owttrim, 2013). In addition, RNA helicases, together with other RNA-binding proteins, are crucial components of ribonucleoprotein particles (RNPs) (Martin et al., 2013). Ribosome assembly and activity has classically been regarded as constitutive in cells. Nevertheless, increasing evidence suggests that in germlines regulatory mechanisms involving ribosomes play roles in controlling gene activity and developmental processes (Schmidt, 2020). In Arabidopsis, 58 DEAD-box RNA helicase (RH) proteins are represented (Mingam et al., 2004), 24 of which are putative ribosome biogenesis factors (Liu and Imai, 2018), but only a few of them have been described to date (Chi et al., 2012; Hsu et al., 2014; Lee et al., 2013; Nishimura et al., 2010; Paieri et al., 2018). RH3 and RH22 are active in chloroplasts and are among the 510 EMBRYO DEFECTIVE genes recognized to date (Meinke, 2020). The importance of gametogenesis has only been reported for RH36 (SLOW WALKER 3), which is involved in 18S-preRNA processing (Huang et al., 2010; Liu et al., 2010). Interestingly, homology with the yeast protein Dbp7p and investigations on the rice homolog OsRH17 suggest that Arabidopsis RH17 could potentially act in ribosome biogenesis (Liu and Imai, 2018; Xu et al., 2015). This identifies RH17 as a promising candidate for controlling aspects of reproductive development.

To study the potential roles of RH17 in reproduction, we used independent Arabidopsis lines carrying mutant alleles. Homozygous mutants were not recovered, in line with an embryo lethal and partially penetrant male gametophytic defect. Consistently, defects during seed development were observed, including embryo developmental arrest and over-proliferation of embryos. Interestingly, two embryos per seed formed at low frequencies. These likely derived from a second gametophyte originating from an AIC-like cell specifying adjacent to the MMC in the same ovule. In addition, we identified that in our lines carrying mutant alleles seed coat and likely also endosperm development initiated in the absence of fertilization at certain frequencies. In agreement with this, RNA sequencing (RNA-Seq) on ovules collected 3 days after emasculation (DAE) from a line carrying a mutant allele of RH17 confirmed an enrichment of genes involved in seed coat development in the set of differentially expressed genes (DEGs) compared with wild type, in addition to genes involved in stress response and hormonal pathways. Taken together, our results show that mutant alleles of RH17 lead to phenotypes resembling elements of apomixis, which could link ribosomal pathways, stress response and hormonal pathways to the regulation of reproductive development.

Increasing evidence suggests that certain RHs are involved in ribosome assembly and turnover, thereby contributing to the control of gene activity (Liu and Imai, 2018; Schmidt, 2020). Interestingly, Arabidopsis RHs likely representing ribosome-binding factors are broadly expressed in reproductive tissues (Fig. S1). Based on its expression and localization of the protein to the nucleus (Fig. S1A-G) (Kiefer et al., 2020; Nguyen et al., 2018), we identified RH17 as a candidate for regulation of aspects of reproductive development. RH17 is present in a variety of cell and tissue types, including female reproductive tissues and germline cells, particularly the MMC and the surrounding nucellus, ovules at early and late stages of gametogenesis, and the female gametes (Fig. S1A,B). During male germline development, RH17 is expressed in meiocytes, in uninucleate microspores and bicellular pollen, unlike in mature pollen or sperm (Fig. S1A,B). Following double fertilization, RH17 is present at early stages of seed development in the embryo, endosperm and seed coat, predominantly at the preglobular stage of embryogenesis. In contrast to sexual Arabidopsis, expression of the RH17 homolog was not observed in the apomictic initial cell or in cells of the mature gametophyte in the apomict Boechera gunnisoniana (Fig. S1B).

To gain insights into the functional roles of RH17 in Arabidopsis, two independent lines, rh17-1 and rh17-2, harboring T-DNA insertions in the eighth exon and the fifth intron, respectively, were selected for analysis. A CRISPR/Cas9 line designated rh17-3 was also generated, leading to a frame-shift deletion in the seventh exon. Consistent with crucial roles for reproductive development, plants heterozygous for either of the three alleles showed 37% (n=895), 39% (n=1514) and 33% (n=233) arrested seeds or ovules in rh17-1/RH17, rh17-2/RH17 and rh17-3/RH17, respectively, in contrast to 3% (n=793) observed in wild-type plants. As all lines showed seed abort at similar frequencies, only rh17-1/RH17 and rh17-2/RH17 were investigated in further detail.

In accordance with the high frequencies of seed abort, homozygous plants were not recovered. To describe the defect in more detail at the genetic level, we studied the segregation of the mutant alleles in the offspring of heterozygous plants by scoring sulfadiazine resistance linked to the T-DNA insertion for rh17-1 or by genotyping of rh17-2. For rh17-1/RH17, we observed 62.0% (n=645) resistant seedlings and 63.5% heterozygous plants were identified by genotyping of rh17-2/RH17 (n=158). These ratios of the numbers of plants carrying a mutant allele and wild-type plants differed significantly from the ratios expected for Mendelian inheritance (3:1, χ-squared=57.9 and 11.6 for rh17-1/RH17 and rh17-2/RH17, respectively). Taken together with the frequencies of seed abort, the observed segregation distortion points towards an embryo lethal defect likely combined with partially penetrant gametophytic lethality. To elucidate which parent was affected, analysis of transmission efficiency through male and female gametes was performed after reciprocal crosses of heterozygous lines to wild type (Table 1). The results indicated a slightly reduced transmission efficiency through the male, rather than the female parent. Therefore, we addressed pollen viability in rh17-1/RH17 by Alexander staining (Peterson et al., 2010). The percentage of viable pollen was slightly reduced to 80% (n=531) compared with 90% (n=709) in wild type (Fig. S1H). Together, these data indicate that mutant alleles of RH17 result in embryo lethality combined with a partially penetrant male gametophytic defect.

As high frequencies of seed abort were observed in siliques from heterozygous plants, we used differential inference contrast (DIC) microscopy after clearing with chloral hydrate to identify developmental defects during embryo and endosperm development. In contrast to wild type, which exhibited normal seed development and formation of one embryo per seed (n=91), in siliques harboring maturing seeds, both lines carrying a mutant allele of RH17 showed pleiotropic phenotypes (Fig. 1, Fig. S2). Consistent with the embryo lethal defect, in 60% (rh17-1/RH17) and 62% (rh17-2/RH17) of the seeds normal embryo and endosperm development was observed, whereas 25% of embryos were aborted typically at the globular or heart stage in both rh17-1/RH17 (n=321) and rh17-2/RH17 (n=227) (Fig. 1, Fig. S2). Occasionally, in seeds aborted at the globular stage of embryogenesis no endosperm was formed (Fig. 1B,D). Furthermore, in both lines, over-proliferating embryos were present in 7% of seeds (Fig. 1I, Fig. S2D,E). In addition, enlarged endosperm nuclei were frequently visible close to the embryo (Fig. 1C,E). In the remaining ovules, mostly arrested or unfertilized mature gametophytes were visible, or gametophytes were lacking (Fig. S2B,C).

Interestingly, two embryos in one seed were occasionally observed (Fig. 1F-H, Fig. S3U). However, developmental defects of one embryo or differences in the developmental stage were apparent, suggesting that successful germination of both embryos is unlikely (Fig. 1G,H). To score frequencies, we analyzed siliques with seeds at early stages of embryo development (up to ∼128 cell embryo proper). Two embryos and endosperm were formed in one seed at frequencies of 2.8% in rh17-1/RH17 (n=109) and 1.5% in rh17-2/RH17 (n=268) (Fig. 1, Fig. S3U). In summary, these observations suggest that the mutant alleles cause alterations in reproductive development and defects in the embryo and in endosperm.

Despite low frequencies, formation of more than one embryo per seed is exceptional in Arabidopsis. To gain insights into the origin of the additional embryo, we studied the formation and development of the female germline. We introgressed the pKNU::nlsYFP marker, which labels the MMC (Tucker et al., 2012), into rh17-1/RH17 and rh17-2/RH17 plants. In wild type, typically only a single enlarged subepidermal cell, the MMC, is visible in young ovules. Nevertheless, formation of more than one enlarged cell has been reported at low frequencies of ∼6% (Grossniklaus and Schneitz, 1998; Olmedo-Monfil et al., 2010; Schmidt et al., 2011). Consistent with this, in the pKNU::nlsYFP lines a second enlarged cell next to the MMC or female meiocyte was visible in 5.9% of scored ovules. In the vast majority of ovules only one enlarged subepidermal cell expressing the marker was observed (Fig. 2A,G, Fig. S3A,B). In contrast, in rh17-1/RH17 carrying the marker, a second enlarged non-labeled cell was visible next to the MMC at a frequency of 23.5% (Fig. 2C,D,G, Fig. S3C,D,G), and only occasionally a second marked cell was observed (not shown). In rh17-2/RH17, 27.3% of ovules showed a second enlarged cell next to the MMC not expressing the marker (Fig. 2F,G, Fig. S3E,F). Based on the absence of YFP activity, we can conclude that these cells do not typically acquire MMC identity.

Whereas in wild type the rarely observed additional enlarged cells typically do not give rise to formation of gametophytic lineages (Fig. 2N,Q) (Pinto et al., 2019), the development of additional gametophytes was observed by clearings in 27.8% and 28.2% of ovules in rh17-1/RH17 and rh17-2/RH17, respectively (Fig. 2O-Q, Fig. S3Q). To determine whether FMS-like cells and developing cell lineages are specified as gametophytic, we used the ANTIKEVORKIAN (AKV) reporter for cell identity (Fig. 2H-L, Fig. S3H-P,R-T,V), which is active in female gametophytic nuclei from the FMS stage to cellularization (Pillot et al., 2010; Schmidt et al., 2011). Consistent with our expectations, gametophytic identity could be confirmed based on the activity of the AKV reporter. In addition, atypical positioning of gametophytic nuclei in the ovule likely results from more than one gametophyte per ovule (Fig. S3M-P,R-T,V). In contrast to the pAKV::H2B-YFP marker line serving as control (Fig. S3J-L,V), the occurrence of ectopically formed gametophytes was also occasionally detected in rh17-1/RH17 and rh17-2/RH17 expressing the reporter construct (Fig. S3M-O,R).

Interestingly, different sizes of gametophytic nuclei were sometimes presented in the same ovule, which might indicate different ploidies of the nuclei (Fig. 2J-L). To determine whether cells other than the MMC undergo meiosis, we used the KINGBIRD marker line, which allows visualization of the activity of the core meiotic protein REC8 (Prusicki et al., 2019). However, only the MMC was labeled as meiotic in young ovules of rh17-1/RH17 (n=43) (Fig. S4). Taken together, in the lines carrying mutant RH17 features resembling apospory were displayed, where an additionally enlarged sporophytic cell adjacent to the MMC potentially initiates gametogenesis without undergoing meiosis.

However, in agreement with the fact that two embryos were observed only infrequently, the additionally formed gametophytes rarely reached maturity (Figs S5E and S6I). Most of the additionally formed gametophytes appeared to arrest at early stages of gametogenesis (Fig. S6G,H). Furthermore, no evidence for mis-specification of gametophytic cells was given in mature gametophytes of rh17-1/RH17 carrying a multicolor fluorescent marker line labeling each cell of the mature gametophyte with a different fluorescent protein at 2-3 DAE (Fig. S5). This suggests that the second embryo likely derived from an egg cell of a second gametophyte that formed in the same ovule.

The question remains regarding how two embryos and functional endosperm can form in one seed, as typically during double fertilization only one sperm cell each is delivered to the egg and the central cell. Therefore, we tested for the possibility of autonomous embryo or endosperm development. To this aim, we emasculated floral buds and subjected the ovules to clearing at 5 DAE (Fig. S6). A tendency was identified for an enlargement of ovules and the formation of additional nuclei in lines carrying a mutant rh17 allele. To gain more insight into developmental changes over time, we further analyzed ovules at 6-7 DAE. Unlike in wild type (n=284), we observed an enlargement of ovules and features of autonomous seed coat formation at frequencies of 11.4% and 12.1% in rh17-1/RH17 (n=667) and rh17-2/RH17 (n=331), respectively (Fig. 3). In addition, evidence of autonomous endosperm formation, mostly with low numbers of nuclei, was observed in 7.5% and 9.4% of ovules from rh17-1/RH17 and rh17-2/RH17, respectively (Fig. 3B-D,F). Nevertheless, endosperm identity was not unambiguously confirmed and an origin of nuclei by proliferation of gametophytic nuclei cannot be ruled out. To confirm parthenocarpy, vanillin staining forming a red color upon reaction with proanthocyanidins in the developing seed coat was applied at 7 DAE (Xuan et al., 2014). The staining supported the hypothesis of autonomous seed coat development (Fig. 4).

Unlike these features of autonomous seed development, we did not find evidence for autonomous embryo formation. To test whether offspring were derived by fertilization, we used rh17-1/RH17 (Ecotype Col-0) as a mother and the AKV- or multicolor fluorescent marker lines (Ecotype Landsberg erecta, Ler) as father. We used established mapping primers to identify Ler alleles in 105 seedlings (Zhang et al., 2007). In all plants tested, the presence of Ler genomic information was confirmed (Fig. S7). This suggests that offspring is usually formed by fertilization of the egg cell.

Autonomous endosperm and seed coat development typically involve a derepression or deregulation of genes and pathways repressing such development. To identify genomic pathways involving the activity of RH17, we used RNA-Seq for transcriptional profiling of ovules at 3 DAE. In the wild type, gametophytes are fully mature at this stage, whereas we expect the developmental programs enabling autonomous development to be activated in the lines carrying a mutant allele of RH17. Illumina Next Seq500 was used to sequence three and two biological replicates of RNA isolated from ovules of wild type or rh17-1/RH17, respectively. This resulted in between 41,637,502 and 60,884,568 paired mapped reads per sample after quality control and trimming (Table S1). We found that 19,726 genes were expressed in all samples of rh17-1/RH17 and the wild type, whereas only 374 and 396 genes were active either in both samples of rh17-1/RH17 or all three replicates from the wild type at ≥10 TMM normalized read counts, respectively (Fig. S8A). Compared with the wild type, 1558 genes were identified as differentially regulated in ovules of rh17-1/RH17 (FDR ≤0.05 after Benjamini–Hochberg adjustment) (Robinson et al., 2009) (Tables S2 and S3, Fig. S8B). As only a heterozygous line could be used for this analysis, the tissues of rh17-1/RH17 contained a mixture of heterozygous maternal sporophytic tissues of the ovule harboring on average about 50% gametophytes carrying the rh17-1 allele and 50% wild-type gametophytes. Consequently, the fold changes (FC) observed for differential expression will be reduced (Table S3). For example, a 0.6 log FC was observed for RH17. Interestingly, the DEGs identified with the highest FC included (long) non-coding RNAs (Table S3). For the long non-coding RNA AT1G07887, we independently validated this great change in abundance by real-time PCR (Fig. S9).

To categorize the gene regulatory programs and to identify biological processes that are over-represented in the set of genes, we analyzed Gene Ontology (GO) enrichment. Biological processes that were upregulated (P<0.01) pointed towards the relevance of stress responses including redox-related processes and salt stress, hormonal pathways related to auxin, gibberellin and jasmonic acid, and ribosome-related processes (i.e. GO:0034976 ‘response to endoplasmic reticulum stress’ and GO:0034620 ‘cellular response to unfolded protein’) (Table S4). Interestingly, the DEGs representing the biological process of ‘response to endoplasmic reticulum stress’ included the luminal binding proteins BIP1 and BIP2, which are relevant for fusion of polar nuclei prior to endosperm proliferation (Table S5) (Maruyama et al., 2020). The category related to redox regulation comprised a number of cytochrome P450 genes, including CYP78A5 (also known as KLUH, KLU), previously reported to have roles in regulation of reproduction during germline specification and to control seed size and seed coat development (Adamski et al., 2009; Zhao et al., 2018). Apart from KLU, AGAMOUS-LIKE 11 (AGL11; also known as SEEDSTICK, STK), CYTOKININ OXIDASE/DEHYDROGENASE 1, UBIQUITIN-SPECIFIC PROTEASE 15, ABA DEFICIENT 2, ABSCISIC ACID-INSENSITIVE 5 and DA (LARGE IN CHINESE) 2 were among the DEGs with reported roles in control of seed size (Orozco-Arroyo et al., 2015). Furthermore, YUCCA6, which is an important player in auxin synthesis (Robert et al., 2015), was identified as one of the DEGs more highly expressed in rh17-1/RH17 ovules compared with the wild type. The category of auxin-responsive DEGs comprised 52 genes, including AUXIN RESPONSE FACTOR 5 (ARF5), ARF18, a number of SMALL AUXIN UPREGULATED RNA (SAUR) genes, different INDOLE-3-ACETIC ACID INDUCIBLE (IAA) genes, in addition to PIN-FORMED 7 (PIN7), PIN3 and GH3.6 (Table S5). Importantly, transcriptional regulation was also identified as an upregulated term (GO:0006355, ‘regulation of transcription, DNA templated’) (Table S4). This category comprised 152 DEGs, including a number of AGL MADS box genes, such as SEPELATA2 (SEP2), SEP3, DIANA, STK, AGL20, AGL99, and others, e.g. BASIC HELIX-LOOP-HELIX PROTEIN 100 (BHLH100) (Table S5). In addition to metabolic processes, developmental processes were also affected, including GO:0080155 ‘regulation of double fertilization forming a zygote and endosperm’. In good agreement with our observations, the biological process of ‘seed coat development’ (GO:0010214) was enriched (Fig. 5, Tables S4 and S5). Interestingly, all genes in this category were significantly more highly expressed in our lines carrying a mutant rh17-1 allele compared with the wild-type ovules, with the exception of LEUNIG, which is a regulator of seed coat mucilage extrusion (Bui et al., 2011). Importantly, STK, which is upregulated in our lines carrying a mutant allele, has previously been reported to be a master regulator of seed coat development (Mizzotti et al., 2014). Furthermore, epigenetic regulation, in particular ‘histone lysine demethylation’, was differentially regulated (Tables S4 and S5). Taken together, the data indicate differential regulation of a number of genes and pathways, including, importantly, known regulators of reproductive development.

In natural apomicts, several alterations compared with sexual reproduction are required, including the formation of an unreduced gametophytic lineage, and the coordinated development of the parthenogenetic embryo and of endosperm. Despite several decades of research, the longstanding dream of engineering apomixis into crops has not been accomplished to date. Nevertheless, as a proof of concept, clonal offspring has been obtained in sexual plants by combinations of mutations in three meiotic genes and systems for haploid induction (Marimuthu et al., 2011; Schmidt, 2020). Whereas haploid induction requires egg cell fertilization prior to elimination of the paternal genome, even parthenogenesis can be induced in rice by expression of the transcription factor BABY BOOM 1 in the egg cell (Khanday et al., 2019). The applicability of these genes and systems is thus far limited, as typically clonal offspring is only obtained at low frequencies (Scheben and Hojsgaard, 2020). Furthermore, it would be a great advancement if less complex breeding systems could be established with the requirement to modify fewer genes. In partial contradiction to the longstanding hypothesis of apomixis being derived from deregulation of the underlying gene regulatory program of sexual reproduction (Carman, 1997), it is typically genetically linked to one or a few loci (Barcaccia and Albertini, 2013). This has previously stimulated the idea of a master regulator that might be in control of a complex regulatory program (Eckardt, 2003). Here, we demonstrate that mutant alleles of RH17 lead to alterations in sexual development resembling different features of apomixis. Although a straightforward application is not foreseeable owing to the embryo lethality observed, these findings suggest that it might be possible to induce more than one element of apomixis by manipulation of one or a few genes only.

Interestingly, low frequencies of formation of two mature gametophytes in one ovule and of polyembryony have previously been observed in natural apomicts of the genus Boechera (Carman et al., 2019). This is similar to our observations in Arabidopsis lines carrying a mutant allele of RH17. Here, we cannot exclude the possibility that two embryos and functional endosperm derive from fertilization processes involving the attraction of more than one pollen tube, which occurs at low frequencies in Arabidopsis (Maruyama et al., 2013). Nevertheless, we provide evidence for parthenocarpy and potentially also autonomous endosperm development, suggesting an alternative explanation. Although viable offspring typically derived from fertilization of the egg cell, the embryo lethal defect suggests that gene regulatory programs governing embryogenesis might be imbalanced.

In contrast to the tight control of seed formation, greater developmental flexibility is observed upon germline specification. Increasing evidence suggests that complex gene regulatory systems are required to mediate the important developmental transition from sporophytic to gametophytic fate, the repression of multiple reproductive lineages per ovule, and the commitment of the MMC to meiosis (Lora et al., 2019). Accordingly, several types of mutants in which surplus germlines develop have been described, mainly differing in the fate of the MMC or the sporophytic cells acquiring germline identity (Pinto et al., 2019; Schmidt et al., 2015; Su et al., 2020; Zhao et al., 2018, 2017). Unlike in several other mutants, but similar to lines carrying a mutant allele of RH17, in stk mutants the pKNU::nlsYFP marker for MMC identity is typically not active in additional enlarged subepidermal cells (Mendes et al., 2020). Although STK was deregulated in ovules at 3 DAE, it remains to be determined whether RH17 also affects the expression of this gene at early stages of reproduction. Furthermore, ectopic MMCs in mutant lines of SET DOMAIN GROUP 2, an epigenetic regulator of H3K4 methylation, are not consistently labeled with the marker KNU (She, 2014; Yao et al., 2013). The control of histone modifications, specifically the eviction of repressive H3K27me3 marks established by PRC2, is crucial for activation of KNU (Sun et al., 2014). Therefore, it is possible that the epigenetic setup prevents activation of KNU in ectopic cells adjacent to the MMC. Together with the low, or lack of, expression of the KNU homolog in Boechera nucellus tissues (Zühl et al., 2019), this could be an indication that KNU activity is not generally required upon germline specification. Interestingly, KNU is involved in repression of the gene encoding the homeodomain protein WUSCHEL (WUS), which is an important regulator of stem cell fate (Sun et al., 2014). As repression of WUS is crucial for entry into meiosis (Zhao et al., 2017), in our lines the additional enlarged cells potentially do not express KNU as they likely do not become committed to meiosis, in agreement with the lack of REC8 activity. Therefore, the control of KNU itself might depend on meiotic fate of the MMC rather than on germline identity.

An important question remaining is the molecular mechanism involving RH17 that leads to the restriction of germline fate to one cell lineage per ovule. As RH17 is expressed in the MMC and in surrounding nucellus tissues, both cell-autonomous and non-cell-autonomous mechanisms might be involved. Interestingly, similar apospory-like phenotypes can be obtained by modifications of other genes, including MEM (Schmidt et al., 2015, 2011). Therefore, redundancy is unlikely, as the underlying genetics of the developmental defects differ, with mutations in MEM leading to a female gametophytic defect and mutations in RH17 causing embryo lethality combined with a slightly reduced transmission efficiency through the male parent. The cause of this remains to be elucidated, but it is possible that RH17 acts in certain RNPs that are important during pollen development and pollen tube growth (Honys et al., 2009).

Evidence for the involvement of RH17 in ribosome biogenesis has previously been described for homologs in yeast and rice (Liu and Imai, 2018; Xu et al., 2015). This is in agreement with the deregulation of ribosome-associated processes identified in our transcriptional analysis. Biogenesis and function of ribosomes is highly dynamic and an important level of translational regulation (Dalla Venezia et al., 2019). Previously, for an Arabidopsis line conferring overexpression of RH17, implications for response to salt stress have been demonstrated as well as a decreased accumulation of reactive oxygen species (Nguyen et al., 2018). In contrast, overexpression or knockdown of the rice homolog did not affect sensitivity to different biotic or abiotic stress treatments (Xu et al., 2015). The over-representation of biological processes related to endoplasmic reticulum stress and protein folding, in combination with responses to salt stress and redox regulation, suggests that AtRH17 might function in the regulation of stress response via association with ribosomes. This is consistent with previous reports that mutations in genes involved in ribosomal pathways lead to increased sensitivity to salt stress (Palm et al., 2019). Alternatively, deregulation of processes involved in the stress response might be a consequence of a proteotoxic effect resulting from imbalanced ribosome biosynthesis, as recently demonstrated for Saccharomyces cerevisiae (Albert et al., 2019). Given this, the higher tolerance of overexpression lines to salt stress might reflect a stabilization of ribosomal processes. Moreover, upregulation of processes related to hormonal, epigenetic regulatory, and developmental pathways could serve as an indication that RH17 integrates the control of developmental processes in response to environmental conditions. In agreement with the differential activity of genes involved in auxin synthesis, transport and signaling, auxin has been shown to play important roles in ribosome biosynthesis (Rosado et al., 2010). In turn, ribosomes are involved in regulation of auxin signaling, and, for example, translational control of ARF5, which is an important regulator of embryo development, is mediated by ribosomes (Merchante et al., 2017; Rosado et al., 2012). Interestingly, ARF5 was one of the DEGs identified in our study. Consistent with the importance of auxin signaling for embryo development and patterning control (Möller and Weijers, 2009), the observed over-proliferation of embryos might result at least in part from the deregulation of auxin signaling. This suggests that RH17 acts in ribosomal pathways potentially both controlling and being under the control of auxin signaling.

The upregulation of YUCCA6 indicates that auxin synthesis might be enhanced in ovules carrying a mutant rh17 allele. Importantly, auxin can induce autonomous seed coat and endosperm development (Figueiredo et al., 2016; Figueiredo and Köhler, 2018). This pathway also involves the activity of different AGL MADS box transcription factors, which are generally broadly active and involved in the regulation of gametophyte and seed development (Bemer et al., 2010). In particular, AGL62 mediates the transport of auxin from endosperm to seed coat to induce development (Figueiredo et al., 2016). Furthermore, AGL36 shows increased expression in response to elevated auxin levels (Batista et al., 2019; Shirzadi et al., 2011). Although neither of these were identified as DEGs at 3 DAE, we cannot rule out the possibility that differential regulation might occur at later stages of autonomous seed development. In contrast, DIANA, a gene crucial for central cell development and specification, mutants of which fail to give rise to endosperm formation (Steffen et al., 2008), was identified as DEG. Interestingly, defective or failed endosperm formation was observed in a number of seeds. However, mutant alleles of RH17 also induce features of autonomous endosperm development. Mostly low numbers of nuclei were observed, indicating that endosperm nuclei likely only undergo first divisions, but in cases of two nuclei the possibility of persistent unfused polar nuclei of the central cell cannot be unambiguously excluded. In wild type, autonomous endosperm development and parthenocarpy are typically repressed by PRC2 activity. Interestingly, long non-coding RNAs are involved in recruitment of PRC2 complexes to target sites (Brockdorff, 2013). Whether RH17 plays a role in this pathway remains to be elucidated by future studies.

Increasing evidence suggests that RNA helicases may play a role in gene regulation by association with mechanisms involving non-coding RNAs (Nair et al., 2020; Schmidt, 2020). Our transcriptional analysis identified (long) non-coding RNAs as differentially expressed with high levels of fold changes, for example AT1G07887, which potentially modulates the activity of HOMOLOG OF HUMAN HEI 10, a gene involved in meiotic crossover formation (De Muyt et al., 2014; Ziolkowski et al., 2017). Consistent with previous reports of RH17 being involved in the formation of R-loops, which represent DNA-RNA hybrids and associated single-stranded DNA, functional roles of RH17 during meiosis in the control of genome integrity are likely (Castellano-Pozo et al., 2012; Santos-Pereira and Aguilera, 2015; Xu et al., 2020).

In our lines carrying mutant alleles of RH17, pleiotropic phenotypes were observed, consistent with the deregulation of various genes. It remains to be uncovered which of the DEGs are direct targets and which are indirectly affected in their expression. Interestingly, the mechanism of action of RH17 likely resides in ribosomal-associated pathways, underpinning the notion that specialized ribosomes might play a crucial role in the modulation of gene activity. In conclusion, we uncovered RH17 as a player in the control of plant reproductive development. Strikingly, several alterations of sexual development were observed, including features resembling apomixis. Our findings suggest that hormonal pathways, ribosomal pathways and regulation via non-coding RNAs are involved, providing an excellent basis for future studies investigating the specific molecular mechanisms and machineries underlying underlying the developmental alterations caused by mutant alleles of RH17.

Arabidopsis thaliana (L.) Heynh. Ecotype Col-0 was used as wild type. T-DNA insertion lines were obtained from the Eurasian Arabidopsis Stock Centre (NASC, http://arabidopsis.info/) and GABI-Kat (https://www.gabi-kat.de/). The multicolor fluorescent marker line [in Landsberg erecta (Ler) genomic background] to label the cell types of the mature female gametophyte has been described previously (Lawit et al., 2013). Seeds of the pKNU::nlsYFP (Tucker et al., 2012), the pAKV::H2B-YFP (in Ler genomic background) (Rotman et al., 2005), and the KINGBIRD (Prusicki et al., 2019) marker lines were kindly provided by Matthew Tucker (University of Adelaide, Australia), Weicai Yang (Beijing University, Beijing, China) and Arp Schnittger (University of Hamburg, Germany), respectively. The lines were introgressed into rh17-1/RH17 and rh17-2/RH17, and F1 or F2 offspring of the crosses were analyzed. Seeds were stratified for at least 1 day in darkness before surface sterilization as described previously (Wuest et al., 2010). Seedlings were germinated and grown on Murashige Skoog (MS) plates for about 2 weeks before transfer to soil (ED73, Universalerde, Germany). Subsequently, plants were grown in a growth chamber at 16 h light/8 h darkness at 21°C and 18°C, respectively. Plants were treated with nematodes against fungus gnats.

Two independent T-DNA insertion lines disrupting the eighth exon and the fifth intron of the gene AT2G40700 encoding for RH17 (rh17-1 and rh17-2, which are GABI_814E05 and SALK_076414, respectively) were obtained from GABI-Kat and NASC. Lines rh17-1 and rh17-2 were genotyped with primers 5′-CTCCCTTGTGAGCTGCATATC-3′, 5′-ATGGAATACCATCACAGCTGG-3′ and the left border (LB) primer 5′-GGGCTACACTGAATTGGTAGCTC-3′, or with primers 5′-CGATCTCTACGAGACACTGGC-3′, 5′-AAAAGCAAAGATTTAGAAGATTTAAATC-3′ and the LB primer 5′-ATTTTGCCGATTTCGGAAC-3′, respectively. Positions of T-DNA LBs were confirmed by sequencing to be inserted after base 1769 on genomic DNA counted from the A of the start codon for rh17-1 and after base 938 for rh17-2. Linkage of sulfadiazine resistance to the locus of the rh17-1 insertion was verified by genotyping. For investigation of segregation distortion, offspring of heterozygous plants carrying the rh17-1 or rh17-2 mutant alleles were analyzed either by counting the ratio of plants sensitive or resistant to sulfadiazine or by genotyping, respectively. Resistance was tested on MS medium supplied with 7.5 mg/l sulfadiazine. Analysis of transmission efficiency was performed by reciprocal crosses of rh17-1/RH17 and rh17-2/RH17 to wild type and subsequent analysis of the presence of the mutant allele by scoring sulfadiazine resistance or by genotyping, respectively.

Generation of the CRISPR/Cas9 line was carried out as previously described (Kiefer et al., 2020), except that small guide RNAs 5′-ATTGTGCAGTGCCATGTGGTGCA-3′ and 5′-AAACTGCACCACATGGCACTGCA-3′ were used and genotyping was performed with primers 5′-ACCATCACAGCTGGTACAGAGA-3′ and 5′-CAGGAACTTCTGGCTTGAGATT-3′ and the restriction enzyme ApaLI (New England Biolabs). To generate the line rh17-3/RH17, a heterozygous mother plant was crossed to wild type. The offspring resulting from this cross was germinated on MS plates with hygromycin and sensitive plants were rescued. Genomic DNA was isolated from sensitive plants and tested for the absence of the CRISPR/Cas9 construct with primers 5′-GATTTCATGCTCACACATGCTC-3′ and 5′-GAAGAGCTGGTTGTACGTCTG-3′. A frameshift deletion of GT in the target region after base 1430 (counted on genomic DNA from the A of the start codon) was confirmed by amplification of the region with the primers used for genotyping, cloning of the PCR construct into pJET1.2/blunt (Thermo Fisher Scientific) and sequencing of five clones using the T7 primer.

To test for potential formation of clonal offspring, rh17-1/RH17 plants were emasculated for 2-3 days prior to pollination with either the pAKV::H2B-YFP marker line or the multicolor fluorescent marker line both representing Ler ecotype. Sulfadiazine-resistant F1 offspring was tested for the presence of Ler genomic DNA by using primers for map-based cloning with Col-0 and Ler ecotypes with the primer combination T2E12F and T2E12R (Zhang et al., 2007). PCR products were analyzed on 1% agarose gels stained with GelRed (Biotium).

The AtUBIQUITIN10::RH17-mVENUS construct for transient expression in Nicotiana benthamiana was assembled using the GreenGate cloning system (Lampropoulos et al., 2013). To generate the pGGC-RH17 module, the coding sequence of Arabidopsis RH17 was amplified with Phusion (New England Biolabs) using primers 5′-AACAGGTCTCAGGCTCAACAATGATGAAGAGAGCCCAACAATC-3′ and 5′-AACAGGTCTCTCTGAAGTTTTTTGTGTACTTCTATTGCC-3′. Sanger sequencing to confirm the sequence was performed at Eurofins Genomics (Ebersberg, Germany). pGGD-mVenus was kindly provided by Rainer Waadt [Centre for Organismal Studies (COS), Heidelberg University, Germany]. pGGA006, pGGB003, pGGZ003, pGGE-tHSP18.2 M and pGGF-HygR were described previously (Lampropoulos et al., 2013; Waadt et al., 2017). Assembly of GreenGate modules pGGA006, pGGB003, pGGC-RH17, pGGD-mVENUS, pGGE-tHSP18.2 M and pGGF-HygR into pGGZ003 was performed as previously described (Lampropoulos et al., 2013). The pUBI10::H2B:RFP::term construct (Lucas et al., 2013) used as marker for localization to the nucleus was kindly provided by Alexis Maizel (COS, Heidelberg, Germany). The expression constructs were independently transformed into the Agrobacterium tumefaciens ASE strain harboring the pSOUP plasmid, and bacterial cultures were mixed directly before they were used for leaf infiltration of N. benthamiana as described (Chen et al., 2013).

Quantification of aborted ovule or seed development was performed after opening siliques with dissecting needles. To characterize and quantify germline and seed development, inflorescences, buds or siliques were fixed in an ice-cold mixture of 75% ethanol and 25% acetic acid, infiltrated by vacuum twice for 15 min before incubation in fixative on ice overnight followed by replacing the fixative with 70% (v/v) ethanol. Clearing was accomplished in chloral hydrate/glycerol/water (8:1:2; w/v/v) or with Hoyer's solution (Anderson, 1954). Ovules and seeds were microdissected with dissecting needles prior to microscopy. They were mounted in 80% (w/v) glycerol for epifluorescence microscopy or in 5% (w/v) glycerol with 0.1% (v/v) of the cell wall dye Renaissance 2200 (SR2200) for laser-scanning confocal microscopy (Musielak et al., 2015). Laser-scanning confocal microscopy was performed with a Leica TCS SP8 microscope using a 63× oil immersion objective and Leica LAS X software. Images were acquired sequentially. Excitations/emissions were set to 405/410-519 nm for SR2200, and 514/519-659 nm for YFP. To score frequencies of ovules showing activity of the pKNU::nlsYFP and pAKV::H2B-YFP marker lines, laser-scanning confocal microscopy or epifluorescence microscopy were used. To score wild-type ovules expressing the multifluorescent marker line, wild-type segregants of rh17-2/RH17 carrying the construct were used. Differential inference contrast (DIC) and epifluorescence microscopy were performed using a Zeiss Axio Imager M1 to capture images. For Alexander staining, images were taken with a Zeiss Axioskope HBO50 connected to an Axiocam MrC5 camera. Images were processed in ImageJ2.1 or processed and cropped in Adobe Photoshop CS2 Version 9.0 (Adobe Systems).

Two and three biological replicates of ovules harboring mature gametophytes 3 DAE were isolated with dissection needles from rh17-1/RH17 and wild type, respectively. Per sample, approximately 150-200 ovules were collected and immediately frozen in liquid nitrogen. Ovule tissues were disrupted using glass beads and a mixer mill prior to isolation of total RNA using the PicoPure RNA isolation Kit (Arcturus) following the manufacturer's instructions including DNaseI digest. RNA integrity was confirmed using RNA Pico Chips on an Agilent 2100 Bioanalyzer. Libraries for RNA-Seq were prepared with the NEBNext Ultra II Directional RNA Library Prep Kit for Illumina (New England Biolabs) using NEBNext Multiplex Oligos for Illumina (Index Primers Set 1 and Index Primers Set 2) for sample indexing. Sequencing was performed on an Illumina NextSeq500 instrument by the Deep Sequencing Core Facility of Heidelberg University using the 75 bp paired end protocol.

The quality of raw reads was controlled with fastQC version 0.11.4 (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Cutadapt 1.14 was used for trimming adapters and low-quality bases with the parameters set to hard trimming of the final base, and a 3′ trimming with a cutoff of phred score threshold 20 (Martin, 2011). Reads shorter 30 bp were filtered out. The TAIR10 A. thaliana genome sequence and Araport11 genome annotations were downloaded from Araport (araport.org) and used for mapping with Star version 2.5.3a_modified (Dobin et al., 2013). The number of fragments of correctly paired reads, of which at least 70% of read length mapped to an exon, was determined using featureCounts in Rsubread package (version 1.20.6), based on usage described by Schmid (2017). Subsequent analysis of DEGs was carried out using the Bioconductor library EdgeR implemented in R for filtering of genes with low expression and estimations of dispersions using pairwise comparisons of wild-type and rh17-1/RH17 samples (Robinson et al., 2009). Genes with FDR <0.05 after Benjamini–Hochberg adjustment were considered to be significantly differentially expressed. The Bioconductor package topGO (http://www.bioconductor.org/packages/release/bioc/manuals/topGO/man/topGO.pdf) was used to test for the enrichment of molecular functions and biological processes. To this aim, the ATH_GO_GOSLIM annotations were downloaded from TAIR (arabidopsis.org) and used for custom gene2GO annotations in topGO. To test for overrepresented GO terms, Fisher's exact test was used combined with the function ‘weight’.

For analysis of expression, 24 DEAD-box RNA helicases were selected with homologies to yeast and human ribosomal binding factors (Liu and Imai, 2018). To analyze expression in 74 different Arabidopsis cell and tissue types, data were used as described previously (Kotliński et al., 2017). RH17 was designated as present (P) when present in at least two out of three or three out of four replicates, and otherwise referred to as absent (A). Analysis of presence or absence in each sample was described previously (Schmidt et al., 2011). The heatmap in Fig. S1A was based on log2-transformed mean expression values as described (Schmidt et al., 2011). To generate the heatmap in Fig. S1B, normalized log2-transformed read counts from RNA-Seq on Arabidopsis and Boechera gunnisoniana cell and tissue types were used as described previously (Schmidt et al., 2014). For all heatmaps presented, the heatmap.2 function implemented in the Bioconductor package gplots (https://cran.r-project.org/web/packages/gplots/gplots.pdf) was used to generate the heatmap using agglomerative clustering (complete linkage) and Euclidean distance. The Venn diagram was created with the online tool Venny 2.1 (https://bioinfogp.cnb.csic.es/tools/venny/).

Bar graphs and box and whiskers plots were created in R (www.r-project.org) with the package ggplots2 (Wickham, 2016) or with the function boxplot() using default settings. Two-sided Fisher's exact test for count data was performed in R using the fisher.test() function. Two-sided Student's t-test was performed in R using the function t.test().

For independent validation, three independent replicates of ovules from wild-type plants and rh17-1/RH17 were collected from buds 3 DAE. Ovule collection and RNA isolation was carried out as described above for RNA-Seq. For cDNA synthesis, 500 µg of total RNA were used with oligo (dT) primers 12-18 (Thermo Fisher Scientific) and the Superscript IV Reverse Transcriptase Kit (Thermo Fisher Scientific) following the manufacturer's instructions. qPCR was subsequently performed on cDNA (diluted 1:30) using Luna Universal qPCR Master Mix (New England Biolabs) following the manufacturer's instructions on a Rotor Gene 6000 cycler (Corbett Life Science). Three technical replicates each were included in the analysis. Primer combinations 5′-CAAGTCATTCGACCAAAACA-3′ and 5′-GTGAATTTCGGTTGGGTAGC-3′, and 5′-ACTCTTGAGGTGGAGAGTTCTG-3′ and 5′-GAAGATCAGCCTCTGCTGGTC-3′, were used to amplify the long non-coding RNA AT1G07887 and UBIQUITIN10, respectively. Primer pair efficiencies were tested prior to the analysis. For analysis, the ΔΔCt method was applied. Relative expression levels were inferred as 2^−ΔCt values of biological replicates. The significance (P<0.001) of differences between the wild type and rh17-1/RH17 was tested with two-tailed, unpaired Student's t-test.

We thank Marcus A. Koch [Centre for Organismal Studies (COS), Heidelberg, Germany] for sharing the working time of his technical assistant and laboratory space. We are grateful to the groups at COS (Heidelberg) for generous access to equipment, facilities, and GreenGate modules. We acknowledge Marlena Pozoga and Xu Fei (COS, Heidelberg) for help with the plant work. We thank Markus Kiefer (COS, Heidelberg) for help with the bioinformatics data analysis, and Rainer Waadt (COS, Heidelberg), Alexis Maizel (COS, Heidelberg), Weicai Yang (Chinese Academy of Science, Beijing, China), Matthew Tucker (University of Adelaide, Australia) and Arp Schnittger (University of Hamburg, Germany), for providing vectors and marker lines.