The MAP3K-Coding QUI-GON JINN (QGJ) Gene Is Essential to the Formation of Unreduced Embryo Sacs in Paspalum

Apomixis is a clonal mode of reproduction via seeds, which results from the failure of meiosis and fertilization in the sexual female reproductive pathway. In previous transcriptomic surveys, we identified a mitogen-activated protein kinase kinase kinase (N46) displaying differential representation in florets of sexual and apomictic Paspalum notatum genotypes. Here, we retrieved and characterized the N46 full cDNA sequence from sexual and apomictic floral transcriptomes. Phylogenetic analyses showed that N46 was a member of the YODA family, which was re-named QUI-GON JINN (QGJ). Differential expression in florets of sexual and apomictic plants was confirmed by qPCR. In situ hybridization experiments revealed expression in the nucellus of aposporous plants’ ovules, which was absent in sexual plants. RNAi inhibition of QGJ expression in two apomictic genotypes resulted in significantly reduced rates of aposporous embryo sac formation, with respect to the level detected in wild type aposporous plants and transformation controls. The QGJ locus segregated independently of apospory. However, a probe derived from a related long non-coding RNA sequence (PN_LNC_QGJ) revealed RFLP bands cosegregating with the Paspalum apospory-controlling region (ACR). PN_LNC_QGJ is expressed in florets of apomictic plants only. Our results indicate that the activity of QGJ in the nucellus of apomictic plants is necessary to form non-reduced embryo sacs and that a long non-coding sequence with regulatory potential is similar to sequences located within the ACR.

Apomixis is a clonal mode of reproduction via seeds, which results from the failure of meiosis and fertilization in the sexual female reproductive pathway. In previous transcriptomic surveys, we identified a mitogen-activated protein kinase kinase kinase (N46) displaying differential representation in florets of sexual and apomictic Paspalum notatum genotypes. Here, we retrieved and characterized the N46 full cDNA sequence from sexual and apomictic floral transcriptomes. Phylogenetic analyses showed that N46 was a member of the YODA family, which was re-named QUI-GON JINN (QGJ). Differential expression in florets of sexual and apomictic plants was confirmed by qPCR. In situ hybridization experiments revealed expression in the nucellus of aposporous plants' ovules, which was absent in sexual plants. RNAi inhibition of QGJ expression in two apomictic genotypes resulted in significantly reduced rates of aposporous embryo sac formation, with respect to the level detected in wild type aposporous plants and transformation controls. The QGJ locus segregated independently of apospory. However, a probe derived from a related long non-coding RNA sequence (PN_LNC_QGJ) revealed RFLP bands cosegregating with the Paspalum aposporycontrolling region (ACR). PN_LNC_QGJ is expressed in florets of apomictic plants only.
Our results indicate that the activity of QGJ in the nucellus of apomictic plants is necessary to form non-reduced embryo sacs and that a long non-coding sequence with regulatory potential is similar to sequences located within the ACR.

INTRODUCTION
Asexual reproduction can naturally occur in ovules of several flowering plant taxa through apomixis, an alternative route to sexuality, which allows the formation of maternal embryos within seeds (Nogler, 1984;Carman, 1997). This atypical trait relies on developmental alterations which cause unreduced cells within the ovule to acquire a reproductive fate. Although mechanistically diverse, apomictic pathways are usually classified into two major classes (i.e., sporophytic and gametophytic), depending on the origin of maternal embryos (Hand and Koltunow, 2014). During sporophytic apomixis, embryogenesis occurs spontaneously in somatic cells of the ovule, leading to the formation of seeds that harbor supernumerary maternal embryos. In contrast, gametophytic apomixis involves the differentiation of functional, unreduced embryo sacs (2n-ES) within the ovule, followed by egg cell parthenogenetic development into embryos (Bicknell and Koltunow, 2004). Depending on the origin of 2n-ESs, gametophytic apomixis can be further subcategorized into: (1) diplospory, when the megaspore mother cell (MMC) fails meiosis and enters into gametogenesis; or (2) apospory, when one or several nucellar or integumental cells, which are usually somatic companions of the MMC, acquire a gametic fate. In all gametophytic apomicts the embryo develops autonomously, while the formation of the endosperm can be either autonomous or fertilization-dependent (pseudogamous) (Bicknell and Koltunow, 2004).
Based on these results and considering the essential roles of MAPKs in plant development (Musielak and Bayer, 2014;Xu and Zhang, 2015), we rationalized that the P. notatum At1g53570 homolog N46 might be involved in the switch from sexuality to aposporous apomixis in this species. The central biological question of our work was the following: is the Paspalum At1g53570 ortholog (N46) involved in the developmental molecular cascade controlling apospory, either as trigger or participant? To test this hypothesis, we first mined transcriptomic resources available for P. notatum to complete the characterization of N46 sequences. We also conducted spatio-temporal expression analyses in sexual and apomictic genotypes of P. notatum, and used Brachiaria brizantha, a related aposporous species, as a validation control. Finally, we made functional analyses in P. notatum by producing RNA interference (RNAi) lines. Moreover, we mapped N46 onto the P. notatum genome to explore the occurrence of genetic linkage with apomixis. Finally, we determined that the At1g53570-like transcript previously identified by Polegri et al. (2010) (see above) was not a protein-coding ortholog of At1g53570 and N46, but a lncRNA showing only partial similarity with these genes.

Sequence Analysis
The N46 full-length sequences were retrieved from 454/Roche FLX + floral transcriptome databases generated in prior work (Ortiz et al., 2017) and available at DDBJ/ENA/GenBank under the accessions GFMI00000000 and GFNR00000000, versions GFMI02000000 and GFNR01000000, respectively. Analysis of DNA similarity was done by using the BLASTN and BLASTX packages at the NCBI 1 , the Arabidopsis Information Resource 2 and the Gramene 3 websites, as well as exploring the Oryza Repeats Database 4 . For open reading frame (ORF) detection, the NCBI ORF Finder tool was used 5 . Gene schemes were constructed with the WormWeb Exon-Intron graphic maker 6 . Alignments and phylogenetic analyses were done with ClustalW2 (Larkin et al., 2007) and MEGA6 (Tamura et al., 2013) software packages, respectively. The evolutionary history was inferred using the UPGMA method (Sneath and Sokal, 1973). Evolutionary distances were computed using the Poisson correction method (Zuckerkandl and Pauling, 1965) (units: number of amino acid substitutions per site). The lncRNA similarity survey was done onto the plant lncRNA GreeNC database (Paytuví Gallart et al., 2016).
To evaluate the representation of different splice variants by semiquantitative RT-PCR, total RNA was extracted from leaves and/or spikelets at premeiosis/meiosis and reverse transcribed using Superscript II (INVITROGEN). Amplifications were conducted with: (1) the same primer pair flanking the 76-nt intron described above; or (2) a primer pair located inside the intron: IP upper (5 -AAACAGCATGGTGCAGTCAA-3 , T m : 60.31 • C) and IP lower (5 -TCAGGTGGACAATTGATGAGA-3 , T m : 59.07 • C). Each reaction (final volume 25 µL) included the same components used for genomic amplifications and were run in the same thermocycler, but 20 ng of cDNA were used as template, and the cycling was 5 min at 94 • C, 25 cycles of 30 s at 94 • C, 30 s at 57 • C and 45 s at 72 • C and a final elongation step of 5 min at 72 • C.
Total RNA was extracted with the SV Total RNA Isolation Kit (PROMEGA), which includes a DNAse treatment step. cDNAs were synthetized with Superscript II (INVITROGEN). All qPCR reactions (final volume: 25 µL) included 200 nM gene-specific primers, 1X Real mix qPCR (BIODYNAMICS) and 20 ng of cDNA. Three biological replicates were processed each into three technical replicates. Replicates with templates produced in the absence of Superscript II (INVITROGEN) and without templates were included (negative controls). Amplifications were performed in a Rotor-Gene Q thermocycler (QIAGEN), programmed as follows: 2 min at 94 • C, 45 cycles of 15 s at 94 • C, 30 s at 62 • C and 40 s at 72 • C and a final elongation step of 5 min at 72 • C. QGJ-specific primers were: (1) N46N upper (5 GGCCCTGCATCTCCTACTTCAT3 , T m : 68 • C) and N46N lower (5 'TGCCCAAACGTCCCACTGC3 , T m : 62 • C), which amplified QGJ in all allelic contexts (used for chronological expression analysis); (2) N46S upper (5 AATCGAAGTTGCTTGCCATC3 , T m : 60 • C) and N46S lower (5 GCTCTGTTAGACCGCTGCTT3 , T m : 59 • C), which were located outside the N46 segment cloned into the pBS86-N46 vector (used for analysis of expression in transgenic plants). Nontemplate reactions were included as controls. β-tubulin was used as an internal reference gene, as recommended by Felitti et al. (2011), Ochogavía et al. (2011), and Podio et al. (2014b, who worked in the same plant model. Relative quantitative expression levels were calculated by using REST-RG (Relative Expression Software Tool V 2.0.7 for Rotor Gene, Corbett Life Sciences) considering take-off and amplification efficiency values for each particular reaction.

In situ Hybridization (ISH) Analyses
Spikelets of P. notatum (genotypes Q4117 and C4-4x) and B. brizantha (genotypes B30 and B105) were collected at premeiosis/meiosis. Flowers were dissected, fixed in 4% paraformaldehyde/0.25% glutaraldehyde/0.01 M phosphate buffer pH 7.2, dehydrated in an ethanol series and embedded in paraffin (for P. notatum) or butyl-methyl-methacrylate (BMM) (for B. brizantha). Specimens were cut into sections of 10 µm (Paspalum) or 3.5 µm (Brachiaria) and placed onto slides treated with poly-L-lysine 100 µg/mL. The paraffin or BMM were removed with xylene or acetone series, respectively. Prior to hybridization, control sections were stained with acridine orange and examined under UV light to verify RNA integrity. A plasmid including the original N46 fragment isolated by Laspina et al. (2008) was linearized using restriction enzymes NcoI or SalI (Promega). Sense and anti-sense probes were labeled with the Roche Dig RNA Labeling kit (SP6/T7), following the manufacturers' instructions, and hydrolyzed to 150-200 bp fragments. Prehybridization was carried out in 0.05 M Tris-HCl pH 7.5 buffer containing 1 µg/mL proteinase K in a humid chamber at 37 • C for 10 min. Hybridization was carried out overnight in a humid chamber at 42 • C, in 10 mM Tris-HCl pH 7.5 buffer containing 300 mM NaCl, 50% formamide (deionized), 1 mM EDTA pH 8, 1 X Denhardt's solution, 10% dextran sulfate, 600 ng/mL tRNA and 600 ng/mL of probe. Detection was performed following the instructions of the Roche Dig Detection kit, using anti DIG AP and NBT/BCIP as substrates. Sections were mounted in glycerol 50% and observed under Leica DMRX (Paspalum experiments) or Zeiss-Axiophot (Brachiaria experiments) light microscopes.

Cytoembryological Observations and Pollen Viability Tests
Spikelets at anthesis were fixed in FAA (70% ethanol:formaldehyde:acetic acid 18:1:1) for 24-48 h. Ovaries were dissected and placed in 70% ethanol for at least 24 h, treated with 3% H 2 O 2 during 2 h and dehydrated in an ethanol series (50%, 70%, 95% and twice 100%; 30 min each step). Next, they were cleared using a series of methyl salicylate/ethanol (v:v) solutions (1:1, 3:1, 5.6:1; 30 min for each step). Finally, ovaries were incubated in methyl salicylate for at least 12 h and examined using a Leica DM2500 microscope equipped with DIC optics. Pollen viability was estimated by staining with Alexander's reagent (Alexander, 1980). Purple-stained grains were considered to be viable whereas lack of staining (i.e., pale-green/non-colored grains) indicated sterility. Observations were carried out in a Nikon Eclipse E200 microscope.

Statistical Methods
The average number of aposporous and meiotic embryo sacs per ovule was compared among four independent transgenic events and the control genotype Q4117. A modified Shapiro-Wilk test was used to test the normal distribution of the variables (Shapiro and Wilk, 1965). Due to the non-normal distribution detected, the variables were compared using the non-parametric tests of Wilcoxon (Wilcoxon, 1945) and Kruskal-Wallis (Kruskal and Wallis, 1952). Confidence intervals for observed proportions were calculated following the method described by Newcombe (1998), derived from a procedure outlined by Wilson (1927) with a correction for continuity 7 . Chi 2 tests for homogeneity were calculated with the R software 8 .

Linkage Analyses
An F 1 population of 55 individuals, derived from a cross between sexual Q4188 as pistillate parent and apomictic Q4117 as male progenitor and characterized for reproductive modes (29 sex: 26 apo) (Stein et al., 2007) was used for linkage analyses. For N46 bulked segregant analysis (BSA), 30 µg of genomic DNA from the two parental lines and two equitable bulks of 10 sexual and 10 apomictic F 1 hybrid progenies were digested with EcoRI, HindIII, and PstI. Samples were loaded in 1% agarose gel (1xTAE), electrophoresed at 40 mA and blotted onto nylon membranes (Hybond N, Amersham) using 10x SSC buffer. DNA was fixed at 80 • C for 2 h. DIG probe labeling (the same N46 fragment used for ISH analyses), hybridization and detection were performed as described by Ortiz et al. (1997). For A-148-3 linkage analysis, 5 µg of genomic DNA from the two parental lines and the 55 F 1 s were digested using EcoRI, electrophoresed and finally blotted onto Nylon membranes. A-148-3 was converted into an RFLP probe according to Polegri et al. (2010). Probe 32 P labeling, blot hybridization and exposition to X-ray films was performed according to Pupilli et al. (2001).

LncRNA Expression Analysis
PCR amplifications were conducted from 50 ng of cDNA produced from total RNA extracted from leaves or flowers, with upper primer LNCU: 5 -AATTGTGCGAAATCCAATCA-3 and lower primer LNCL: 5 -TTCACCATTACTGCCCACAA-3 . The cycling program included 1 cycle of 1 min at 94 • C, 30 cycles of 1 min at 94 • C, 2 min at 57 • C and 2 min at 72 • C and a final elongation cycle of 5 min at 72 • C. Laspina et al. (2008) reported similarity between a mRNA fragment differentially expressed in florets of sexual and apomictic P. notatum plants (N46) and a full-length cDNA transcribed from the maize gene GRMZM6G513881 (see footnote 3; NCBI Reference Sequence NM_001137220.1), which encodes a MAP3K protein. Here, we took advantage of Paspalum 454/Roche FLX + floral transcriptomes recently developed in our laboratory to recover the N46 full cDNA sequences from apomictic (Q4117) and sexual (C4-4x) genotypes and carry out molecular phylogenetic analysis. In the apomictic floral transcriptome library, we detected one isogroup (apoisogroup 00379) represented by four homologous isotigs, namely apoisotig 03083 (GFMI02003139.1), apoisotig 03084 (GFMI02003140.1), apoisotig 03085 (GFMI02003141.1) and apoisotig 03086 (GFMI02003142.1). In the sexual floral transcriptome library, we also found one isogroup (sexisogroup 02509), but it contained a single isotig (sexisotig 08547; GFNR01008571.1). ClustalW nucleotide (nt) sequence alignments revealed that apoisotigs 03085 and 03086 and sexisotig 08547 were highly similar, differing only by a few polymorphisms (SNPs and INDELs) (Supplementary Figure S1). Apoisotigs 03083 and 03084, were respectively, identical to apoisotigs 03085 and 03086, except for a 76-nt insertion with the canonical donor-receptor sites of GU-AG-type introns, which corresponds to a partially conserved intron in maize GRMZM6G513881. These results suggest that P. notatum N46-like floral sequences are genetic and splice variants of a single locus with at least two different alleles in the apomictic genotype (apoisotigs 03085/03083 and 03086/03084) and a third allele detected in the sexual genotype (sexisotig 08547). The intron-like insertion (located between positions 1857-1934 and 1833-1908 in apoisotigs 03083 and 03084, respectively) modifies the reading frame to produce a protein with a variable C-terminal end (Figure 1 and Supplementary Figure S2). Genomic amplification with flanking primers showed that the intron-like region is present in both apomictic and sexual plants (Supplementary Figure S3). Although the non-processed form had been sequenced only from the apomictic samples in the 454/Roche FLX + transcriptome libraries, semi-quantitative RT-PCR experiments showed that both variants (processed and non-processed) were represented in flowers of apomictic and sexual plants (Supplementary Figure S3). Moreover, qPCR of cDNAs originated from a mix of flowers at different developmental stages (from premeiosis to anthesis) with primers located inside the intron revealed no significant differential representation between reproductive modes (not shown). BLASTX searches using N46 full-length sequences as queries identified homology to MAP3K genes belonging to the YODA family (best annotated match: Oryza sativa mitogen-activated protein kinase kinase kinase YODA isoform X1 XP_015617106.1; 79% identity; E-val: 0.0; query coverage: 74%; alignment length: 1,688). A phylogenetic tree inferred using 22 homologous protein sequences from different species showed that all Paspalum sequences grouped into a single cluster within the Poaceae clade, supporting the conclusion that they are allelic isoforms or, alternatively, expressed from gene copies that diverged recently (Supplementary Figure S4). Finally, the whole QGJ sequence was amplified by using primers located at the borders (see "Materials and Methods"), to confirm its existence without the need of computational assembly. Based on its identity as a member of the YODA family, we named the N46 locus QUI-GON JINN (QGJ), after another character of the Star Wars saga.

QGJ Quantitative Expression in Reproductive Organs
The QGJ expression was quantified in spikelets of sexual and apomictic P. notatum genotypes at different developmental stages (1: premeiosis; 2: late premeiosis/meiosis; 3: postmeiosis) by using real-time PCR. The primer pair used for amplification was complementary to all known QGJ variants (see "Materials and Methods, " qPCR experiments). At stage 1 (premeiosis), QGJ transcripts were equally represented in both P. notatum reproductive types (Figure 2A). Later, during late premeiosis/meiosis, the expression in the sexual plant was significantly higher (Figure 2A). In contrast, an opposite pattern was observed at post-meiosis (Figure 2A). Besides, we took advantage of the EMBRAPA (Brasilia, Brazil) collection of Brachiaria brizantha plants, another wellcharacterized aposporous pseudogamous system (Pagliarini et al., 2012), to validate the results. Brachiaria brizantha (syn. Urochloa brizantha) is, like P. notatum, a rhizomatous perennial grass (Poaceae), which reproduces through aposporous pseudogamous apomixis. As P. notatum plants, aposporous Brachiaria genotypes form supernumerary non-reduced embryo sacs lacking antipodals from nucellar somatic cells surrounding the MMC. The B. brizantha genome also includes a single ACR lacking recombination, which may be evolutionary related to the Paspalum one, since it is located in a chromosomal background displaying partial synteny to rice chromosome 2 (Pessino et al., , 1998. In Brachiaria spikelets a similar expression profile was detected, but at premeiosis overexpression was detected in the apomictic genotype ( Figure 2B). However, a more detailed quantification of QGJ in RNA samples extracted in Brachiaria isolated ovaries revealed overexpression in apomictic plants at late premeiosis/meiosis (Figure 2C), which suggest the occurrence of contrasting representation patterns in different tissues.

In situ Expression Pattern
The site of expression was then examined through in situ hybridization in developing ovaries and anthers of P. notatum, using the original N46 clone to produce the sense and antisense probes (Figure 3). N46 is complementary to a region conserved among all QGJ variants, so the experiment has no potential to differentiate them. In premeiotic flowers of sexual plants, the antisense probe showed weak signal in the ovule nucellus and integuments and moderate to strong signal in the MMC, the anther tapetum and pollen mother cells ( Figure 3A). Meanwhile, in apomictic plants the same probe revealed strong signal in the ovule nucellus and MMCs, and a diminished signal in anthers ( Figure 3B). The sense probe showed weal signal in ovules of both sexual and apomictic plants (Figures 3C,D). We validated the observed in situ differential expression of QGJ genes in B. brizantha aposporous ovaries, by using the same N46 probe (Figure 4). In the Brachiaria experiments, thinner microtome slice cuts, together with a microscope with a higher resolving power were used (see "Materials and Methods"), allowing a more accurate detection of the hybridization pattern. In premeiotic ovules of sexual plants, a weak to moderate signal was detected in the ovule nucellus, while a moderate to strong signal appeared in the MMCs (Figure 4A). After meiosis I, the signal became mainly restricted to the micropylar cell of female dyads, yet some signal could also be observed in the nucellus (Figure 4B and Supplementary Figure S5). In tetrads, a strong signal was detected in the non-functional (micropylar) megaspores, while the functional one (located close to the chalazal end of the ovule) showed low signal ( Figure 4C). In premeiotic ovules of apomictic plants, a weak to moderate signal was detected in the ovule nucellus, while a moderate to strong signal appeared in the MMCs ( Figure 4D). During aposporous initials (AI) differentiation, moderate to strong signal was detected in the ovule nucellus, except for the cell layer surrounding the MMC (the AI onset site) (Figures 4E,F). Note that at this stage the MMC has enlarged and formed a meiocyte instead of entering meiosis I (meiosis frequently fails in obligate aposporous plants). Apospory initials originate from this proximal cell layer lacking signal ( Figure 4F). While strong signal was detected in pollen mother cells of the sexual plant ( Figure 4G), a moderate to weak signal was observed in pollen mother cells of the aposporous genotype ( Figure 4H). Finally, hybridizations using sense probes detected weak signals in the nucellus of both plant types (Figures 4I,J). Our results indicate that, in sexual plants, QGJ is weakly expressed in nucellar tissues during meiosis. At this stage, its expression is restricted to the non-functional (micropylar) megaspores, which are adjacent to the functional (chalazal) megaspore. In contrast, and in agreement with our previous results from Paspalum, a strong expression of QGJ was observed in nucellar cells of apomictic plants. However, the proximal layer of cells originating the AI lacked signal, suggesting that QGJ is (C) B. brizantha ovaries. White bars: apomictic plant (genotype Q4117). Black bars: sexual plant (genotype C4-4x). 1: premeiosis; 2: late premeiosis/meiosis; 3: postmeiosis. Control: sexual plant, postmeiosis stage (relative expression = 1). Samples displaying statistically significant differential expression between apomictic and sexual plants were marked with an asterisk at the top. expressed in cells adjacent to the functional germ cells, i.e., nonfunctional reduced megaspores or nucellar cells located aside the non-reduced megaspores in sexual and apomictic plants, respectively.

A Decrease of QUI-GON JINN Expression Impairs the Formation of Aposporous Embryo Sacs (AES)
Next, we decided to investigate if a diminished expression of QGJ in an apomictic background gives rise to altered reproductive phenotypes. Firstly, a plant transformation vector including an N46 hairpin (pBS86-N46) was constructed by cloning the complete N46 fragment in sense and antisense orientation within plasmid pBS86 (see "Materials and Methods"). Then, QGJ RNAi lines were obtained by Q4117 biolistic co-transformation with plasmids pact1-gfbsd2 (which expresses an enhanced green fluorescent protein gene eGFP under the rice ACT1 promoter) and pBS86-N46 (see "Materials and Methods"). From 41 positive transgenic events (for pact1-gfbsd2, pBS86-N46 or both), two groups of lines were selected, which had, respectively, been transformed with: (1) the reporter plasmid pact1-gfbsd2; and (2) the RNAi plasmid pBS86-N46 and the reporter plasmid pact1-gfbsd2 (Figures 5A-H,K,L). Plants belonging to the first group were classified as transformation control lines (since they allow evaluation of reproductive phenotypes in plants subjected to in vitro culture and transformation procedures), while those corresponding to the second group were labeled as RNAi lines. These plants were classified as T0, since they were regenerated from bombarded calli. Prior to the molecular and cytoembryological analysis, two rounds of small rhizomes subculture were conducted (P. notatum is perennial and reproduces by forming rhizomes). Only four plants flowered in the isolated GMO chamber under controlled conditions (transformation control lines #TC1/#TC2 and RNAi lines #RNAi1/#RNAi2), together with a wild type control. Fluorescence analysis in transgenic lines carrying the pact1-gfbsd2 vector (both RNAi and TC) confirmed that the rice Act1 promoter drives expression in male and female reproductive tissues of P. notatum (Figures 5I,J).
Quantitative RT-PCR analyses revealed that the QGJ expression was significantly attenuated in floral tissues of both RNAi lines compared to the apomictic wild type ecotype (relative expression levels ranging from 0.478 to 0.656; Supplementary Figure S6A). Lower expression levels were detected in leaves in comparison to flowers of Q4117 and no significant reduction in expression was detected in leaves of RNAi lines (Supplementary Figure S6B). Rates of viable pollen were slightly reduced in the two RNAi lines (#RNAi1 and #RNAi2) with respect to both the control lines (#TC1 and #TC2) and the wild type (Q4117) ( Table 1). However, even if statistically significant, this minor alteration might not imply physiological consequences. Female reproductive development was examined by observing cleared ovules at anthesis and determining the type and number of ES per ovule. Relatively high proportions of aborted ovaries, i.e., containing no ES, were detected for lines #RNAi1 and #RNAi2 compared to both wild type and control plants (20-30% vs. <10%; Table 2). Defects in both initiation and completion of AES formation affected female reproductive development in RNAi lines (Table 2 and Figure 6). A lower number of AES per ovule was detected for both #RNAi1 and #RNAi2 ( Table 2) and most of them exhibited weak/abortive phenotypes such as small size, ragged borders, and no detectable polar nuclei (Figure 6). Conversely, in control lines (#TC1 and #TC2), the average number of AES per ovule was similar to that of Q4117 (Table 2). Finally, the proportion of ovules containing meiotic ES (MES) showed no statistical difference among Q4117, control lines and RNAi lines ( Table 2). We concluded that the significant reduction of QGJ transcripts after introducing a QGJ hairpin References: an, anther; mmc, megaspore mother cell; ov, ovule; tp, tapetum; pmc, pollen mother cells.
construction in the apomictic genotype Q4117 impaired the formation of AESs. Our data suggest that expression of QGJ in nucellar cells is necessary for aposporous development in P. notatum.

Genetic Linkage Analysis Between the QGJ Locus and the ACR
Possible co-segregation of the QGJ locus with apospory was examined in P. notatum by using bulked-segregant analysis (Figure 7). The complete N46 original fragment was hybridized onto genomic DNA samples originated from two parental plants (Q4188, sexual female parent and Q4117, apomictic pollen donor) and genomic DNA bulks made from 10 sexual and 10 apomictic F 1 plants derived from the Q4188 × Q4117 cross. Segregation in F 1 is expected due to ACR hemizygosity and the heterozygous nature of a high number of parental genomic loci. Although genomic DNA digestion with three different restriction enzymes (EcoRI, HindIII, and PstI) produced several polymorphic bands between parental plants, none of them resulted polymorphic between F 1 bulks. The bulked results show the sexual polymorphisms are not specific to sexual F 1 s. However, the apomictic band present in the parental PstI digest disappeared in the apomictic F 1 s. Since PstI is sensitive to certain contexts (CpNpG sites), this observation might suggest a methylation change occurring during hybridization. In silico mapping onto the Gramene website revealed that the putative ortholog to QGJ is located in chromosome 11 (Os11g0207200, E-value 0.0), in a genomic region showing no synteny with the ACR of P. notatum. However, other high score hits (Os02g0555900, Os02t0666300-01, Os12g0577700, E-values 5.7E-86, 3E-58, and 7.6E-15) are protein kinase genes located in a rice genome region (chromosome 2 long arm, positions: 21,002,439-21,008,209, 27,048,920-27,061,955 and chromosome 12 long arm, positions: 23,885,845-23,888,835, respectively) syntenic to the P. notatum ACR. Our results showed no evidence of a genetic link between QUI-GON JINN and the ACR, but suggested that sequences showing significant similarity to this gene might be located within the genomic region controlling apospory.
The Paspalum ACR Transcribes a Long Non-coding Sequence Showing Partial Similarity to QGJ In a previous work, Polegri et al. (2010) reported full linkage between the Paspalum simplex transcriptome fragment A-148-3 and apospory, and determined that this candidate was also homologous to At1g53570. A search for the complete sequence of A-148-3 in the Paspalum Roche 454 floral libraries revealed that it was not a PN_QGJ ortholog, since it showed the best hit of similarity with apoisotig 04689 (GFMI02004742.1) (query cover 100%; E-value: 6 e −143 , identities 90%; position on apoisotig 04689: 5588-5955). Interestingly, apoisotig 04689 is a 6835-nt sequence with no coding potential and specific to the apomictic P. notatum floral transcriptome libraries (reads in the apomictic library: 133; reads in the sexual library: 0). A search in the GreeNC database revealed similarity with six predicted plant lncRNAs at E-values ≤ 1 e −10 (best match: Zmays_GRMZM2G024551_T01; E-value: 1.00411e −27 ; alignment length: 226 nt, involving the segment flanked by positions 4304-4530 within apoisotig 04689; positive: 177). Given its partial similarity with QGJ, its lack of protein-coding potential, its similarity to sequences included in the GreeNC database and its detected expression in the apomictic floral transcriptome, we inferred that isotig 04689 is a long non-coding RNA (lncRNA) related to QGJ, and was renamed accordingly as PN_LNC_QGJ (after P. notatum long non-coding QGJ). The complete QGJ functional sequence (apoisotig 03085, 2377 nt) has similarity to the PN_LNC_QGJ transcript in positions ranging 1033-1326/1352-1475 (matching positions 5493-5800/6255-6387 in the PN_LNC_QGJ sequence) (Figure 1). Meanwhile, the original N46 fragment spanned positions 733-1174 in the QGJ sequence (apoisotig 03085) (Figure 1).
Mapping of the A-148-3 original transcript in a P. notatum population (55 F 1 plants) revealed one polymorphic band strictly cosegregating with apomixis (Figure 8). Another band showed partial linkage, confirming association of the sequence with proximal regions. Besides, the two additional monomorphic/non-segregating bands were detected, which could be related with the same locus or, alternatively, with other genomic regions located elsewhere (Figure 8). Furthermore, reverse-transcribed PCR experiments with PN_LNC_QGJ specific primers conducted in several apomictic and sexual P. notatum individuals showed that PN_LNC_QGJ is expressed only in apomictic plants (Supplementary Figure S7).
Based on these observations, we concluded that original transcript fragment A-148-3 is part of a long non-coding RNA (namely PN_LNC_QGJ) differentially expressed in apomictic and sexual plants. Besides, a sequence similar to the A-148-3 probe is located in the P. notatum ACR. However, further experiments should be conducted to determine if the copy of the gene located at the ACR is producing the differentially expressed lncRNA (considering the existence of monomorphic bands, which could represent copies located in other parts of the genome). Moreover, the existence of a functional link between

DISCUSSION
The extracellular signal-regulated kinase 1/2 (ERK1/2) cascade is a central signaling pathway that modulates a wide variety of cellular processes, including proliferation, differentiation, survival, apoptosis, and stress response (Wortzel and Seger, 2011). The intracellular communication between membrane receptors and their nuclear or cytoplasmic targets upon stimulation is mediated by a limited number of signaling pathways, including a group of mitogen-activated protein kinase (MAPK) cascades (Wortzel and Seger, 2011). MAPK signal transduction cascades consist of three sequentially activated kinases. Upstream signals activate MAPK kinase kinases (MAPKKKs), which in turn phosphorylate MAPK kinases (MKKs); subsequently, MKKs activate specific MAPKs. The downstream targets of MAPKs can be either transcription factors or cytoskeletal proteins. Phosphorylation and activation of a MAPK can lead to changes in its subcellular localization and its activity on transcriptional effectors, thereby reprogramming gene expression (Fiil et al., 2009). In particular, the Arabidopsis genome encodes 60 putative MAP3Ks (including the MEKK, RAF, and ZIK subfamilies), 10 MAPKKs, and 20 MAPKs, involved in a plethora of different responses to specific ligands (Ichimura et al., 2002). Here, we characterized the  (Newcombe, 1998), with a correction for continuity (http://vassarstats.net/prop1.html), as described in the section "Materials and Methods." expression and function of QUI-GON JINN (QGJ), the putative ortholog to Arabidopsis At1g53570/AtMEKK3, in reproductive organs of sexual and apomictic P. notatum plants. AtMEKK3 belongs to the MEKK subfamily related to budding yeast Ste11p (Lukowitz et al., 2004). It comprises 12 members (Ichimura et al., 2002), including critical regulators of plant cell division and differentiation during reproduction, i.e., YODA/AtMAPKKK4 (early embryogenesis) (Lukowitz et al., 2004), AtMEKK20 (male gamete differentiation) (Borg et al., 2011) and ScFRK/AtMEKK19/20 (female gametophyte development) (Daigle and Matton, 2015). Our results suggests that QGJ plays a role in promoting the acquisition of a gametophytic cell fate by AIs, a critical step in the establishment of the aposporous pathway, or alternatively affects the development of the embryo sacs. The fact that QGJ is not expressed in the cell layer originating AIs, but in the adjacent ones, suggests that a non-cell-autonomous signaling mechanism might be operating. Such mechanisms   , it is difficult to evaluate the QGJ post-transcriptional attenuation consequences on the sexual developmental pathway when using the genetic background transformed here. Such analyses should be conducted in a mutant/transformant line derived from a sexual plant, an experiment that we plan to complete in the near future. Moreover, the slight decrease in pollen viability detected in the Paspalum transgenic lines might reflect a mild response to the introduction of the hairpin in a tissue where the gene is naturally down-regulated with respect to sexual plants, as revealed by in situ hybridization. Podio et al. (2012) had analyzed anaphase I configurations and pollen viability in aposporous and sexual tetraploid cytotypes of P. notatum and found reduced pollen viability in the aposporous genotypes, including Q4117. A reduced activity of QGJ in pollen mother cells of aposporous plants (Figures 3, 4) could also explain the diminished expression detected in apo plants in comparative qPCR experiments conducted on florets (Figure 2), since anthers represent a high proportion of the floret tissue at this stage. Based on all these observations, we hypothesized that, similarly to YODA (Lukowitz et al., 2004), QGJ mediates a signaling pathway acting as a key regulator to define cell lineage during plant reproduction. While YODA is required for normal development of the zygote and the cells of the basal lineage originating the suspensor, QGJ might play a role during the sporophytic-togametophytic transition phase. However, although our results indicate that QGJ activity is essential to non-reduced embryo sacs formation, overexpression experiments under ovule specific promoters will be necessary to assess if expression in the nucellus of sexual plants is fully responsible for apomictic development.
An alternative hypothesis is that this candidate is involved later, after the fate decision has been made, in either sexual or aposporous embryo formation. Polegri et al. (2010) reported the isolation of A-148-3, a P. simplex transcript homologous to the predicted Arabidopsis QGJ ortholog (At1g53570), showing constitutive expression in apomictic genotypes during all reproductive developmental stages and linkage to the P. simplex ACR. However, full sequence analysis revealed that A-148-3 is not a QGJ ortholog, but an lncRNA with partial similarity to QGJ. Genetic linkage analyses in a Q4188 (sexual) × Q4117 (apomictic) F 1 population confirmed that a sequence showing similarity with A-148-3 is located within the P. notatum ACR. On the contrary, we found no evidence that N46 and the ACR are genetically linked. Note that the A-148.3 probe has some potential to hybridize QGJ (the original 354 bp-long A-148-3 sequence includes a 147-nt insertion with 78 % similarity to P. notatum QGJ). Moreover, from the 451 nt covered by the original fragment N46, a 112-nucleotide segment keeps partial similarity (72%) with PN_LNC_QGJ. Although these similarities are limited, involve only a portion of the probes (40% of A-148.3 and 25% of N46), and the experimental conditions were strict enough to ensure specific detection, the possibility of some cross hybridization cannot be fully discarded, especially for A-148.3. However, the detection of contrasting genomic hybridization patterns when FIGURE 7 | Bulked segregant analysis of the QGJ locus on the P. notatum genome. Three different restriction enzymes (EcoRI, HindIII, and PstI) were used to digest the genomic DNA samples. The complete original N46 fragment was used as a probe. Left panel: Polymorphisms (marked with arrows) were detected between Q4188 (sexual genotype, mapping population female parent) and Q4117 (apomictic genotype, mapping population male parent). Right panel: Linkage analysis was performed by bulking genomic DNA extracted from 10 sexual (SB) and 10 apomictic (AB) F 1 individuals obtained from crossing Q4188 (female) and Q4117 (male).
using N46 and A-148.3 as alternative probes (Figures 7, 8), suggests that each of them has capacity for specific detection. The ACR is a genomic region specific to apomictic genotypes, which is highly heterochromatic and harbors almost intact exonic sequences interlaced within highly repetitive sequences (Podio et al., 2014a). PN_LNC_QGJ is expressed in floral tissues of aposporous plants only, and it includes large, non-coding stretches similar to retrotransposons and two short exonic QGJ regions (439 nt in total out of 6835). Long non-coding RNAs have recently emerged as critical regulators of gene expression in many eucaryotes, including plants (Ariel et al., 2015;Chekanova, 2015;Liu et al., 2015). Therefore, considering the sequence relationship between QGJ and PN_LNC_QGJ, it is tempting to speculate that PN_LNC-QGJ could mediate QGJ modulation in reproductive tissues of Paspalum apomicts. Among many putative mechanisms, our results point out at least two of them: (1) a change in splicing leading to the formation of variants; and (2) the induction of nucellar expression via miRNA hijacking. However, no claim of a functional relationship between QGJ and PN_LNC_QGJ can be currently made, since it is not supported by functional analysis data. Moreover, expression of PN_LNC_QGJ from the ACR should be further confirmed. Shortly, we will focus in determining if the miss-expression of QGJ is caused by a transcriptional regulatory event or is alternatively influenced by PN_LNC-QGJ, and to determine if PN_LNC-QGJ is expressed from the ACR. Interestingly, non-coding transcripts carrying exonic sequences were proposed to regulate PN_SERK and PS_ORC3, two genes putatively involved in apomixis in P. notatum and P. simplex, respectively (Podio et al., 2014b;Siena et al., 2016). FIGURE 8 | A-148-3 mapping onto the P. notatum genome by RFLP analysis. Hybridizing banding pattern of genomic DNA EcoRI digests of a P. notatum segregating population, composed of a female sexual parent (Q4188,♀), a male aposporous parent (Q4177, ♂) together with 29 sexual and 26 aposporous F 1 plants. The arrow indicates a band inherited from the aposporous parent, which strictly co-segregates with apospory.
In the past decade, numerous candidate genes for apomixis were identified (Hand and Koltunow, 2014;Ronceret and Vielle-Calzada, 2015) but how the underlying networks integrate into sexual reproduction and alter expression patterns remain largely unknown. Our work posits that a MAP3K signaling pathway of an ERK1/2 cascade is pivotal to aposporous embryo sac differentiation. However, the rest of the members of the ERK cascade and their interactions with this kinase remain unknown. Interestingly, besides N46 (QGJ), Laspina et al. (2008) reported the differential expression of several other genes involved in ERK cascades in comparisons between sexual and apomictic plants: an LRR family protein (N79), a GPI anchored protein (N20), phosphatidylinositol 4K (N23), a Ser/Thr phosphatase (N102), the PRIP-interacting protein (N69) and a kinesin (N114). From them, only N20 and N69 have been further characterized (Felitti et al., 2011;Siena et al., 2014). N20 (later renamed N20GAP-1) is ortholog to genes At4g26466 (LORELEI, encoding a GPI-anchored protein) and/or At5g56170 (LORELEI-like), shows partial cosegregation with apospory and is increasingly overexpressed in apomictic plants from premeiosis to antesis (Felitti et al., 2011). N69 is ortholog to gene AT1G45231 (TGS1, encoding a trimethylguanosine synthase which has a dual role in splicing and transcription), and, contrarily, is increasingly overexpressed in sexual plants from premeiosis to antesis . Moreover, evidence was shown that TGS1 Ser 298 phosphorylation is promoted by an ERK cascade to activate transcriptional activity at some promoters (Kapadia et al., 2013). The availability of Paspalum RNAi lines for N20 (LORELEI), N69 (TGS1) and N46 (QGJ MAP3K, reported here) would allow to investigate in detail a possible biological link among these molecules. Though functional approaches are challenging in polyploid, highly heterozygous apomictic species like P. notatum, the development of reference genomes, transformation protocols and advanced microscopy tools will likely accelerate the discovery of the central mechanisms underlying the switch from sexuality to apomixis.