Spotting the Targets of the Apospory Controller TGS1 in Paspalum notatum

Sexuality and apomixis are interconnected plant reproductive routes possibly behaving as polyphenic traits under the influence of the environment. In the subtropical grass Paspalum notatum, one of the controllers of apospory, a main component of gametophytic apomixis reproduction, is TRIMETHYLGUANOSINE SYNTHASE 1 (TGS1), a multifunctional gene previously associated with RNA cleavage regulation (including mRNA splicing as well as rRNA and miRNA processing), transcriptional modulation and the establishment of heterochromatin. In particular, the downregulation of TGS1 induces a sexuality decline and the emergence of aposporous-like embryo sacs. The present work was aimed at identifying TGS1 target RNAs expressed during reproductive development of Paspalum notatum. First, we mined available RNA databases originated from spikelets of sexual and apomictic plants, which naturally display a contrasting TGS1 representation, to identify differentially expressed mRNA splice variants and miRNAs. Then, the role of TGS1 in the generation of these particular molecules was investigated in antisense tgs1 sexual lines. We found that CHLOROPHYLL A-B BINDING PROTEIN 1B-21 (LHC Ib-21, a component of the chloroplast light harvesting complex), QUI-GON JINN (QGJ, encoding a MAP3K previously associated with apomixis) and miR2275 (a meiotic 24-nt phasi-RNAs producer) are directly or indirectly targeted by TGS1. Our results point to a coordinated control exercised by signal transduction and siRNA machineries to induce the transition from sexuality to apomixis.


Introduction
Apomixis is a revolutionary trait in terms of increasing food production on a sustainable basis [1]. This peculiar mode of plant reproduction, described in at least 293 angiosperm genera [2], consists of the spontaneous formation of maternal seeds in absence of meiotic recombination and fertilization [3]. The joint use of sexuality and apomixis in plant breeding programs allows for a permanent fixation of heterotic genotypes, thus accelerating work outlines and reducing hybrid seed costs [4]. Apomixis-based programs aimed at developing new adapted varieties basically consist of crossing sexual mother plants and apomictic pollen donors, followed by selecting superior F 1 hybrids with asexual reproduction capacity. In theory, a single cross involving any genotype, even a highly heterozygous one, might suffice to produce improved varieties capable of cloning themselves by seeds [4]. Such relatively simple breeding schemes are currently operative in apomictic forage grasses of the Paspalum and Brachiaria genera, but still remain unfeasible in sexual major crops [5][6][7]. However, widely cultivated species such as sorghum, sunflower or cassava show reproductive anomalies mimicking particular apomixis steps [8][9][10][11]; this allows for raising the hypothesis that, at least for some species/genotypes, the induction Plants 2022, 11, 1929 3 of 19 plants through antisense technology caused a decrease in the proportion of ovules carrying meiotic embryo sacs and a concurrent emergence of supernumerary megagametophytes resembling aposporous embryo sacs [38]. No signal of parthenogenesis was detected in tgs1 antisense lines and seeds showed a decrease in the germination capacity, which suggests a possible additional role during endosperm development and/or embryogenesis [38].
In other eukaryote systems (yeasts, mammals, drosophila), TGS1 was initially cloned and characterized as an RNA binding protein with a methyltransferase domain, which enhances the PRIP (PPAR-interacting protein) nuclear receptor coactivator function [39]. It can act as a transcriptional coactivator bridge that, in presence of a ligand-bound PPAR, connects the CBP/p300-anchored coactivator complex (with histone acetyltransferase activity) with the PBP-anchored TRAP/DRIP/ARC core, facilitating the recruitment of general transcription factors (GTFs) and RNA polymerase II holoenzymes, to initiate transcription of specific target genes [40,41]. Moreover, at least in yeasts and animals, the TGS1 methyltransferase domain is involved in trimethylation of the capping of different RNA (snRNAs, snoRNAs, telomerase RNA, pri-pre miRNAs), with direct impact in RNA processing [42][43][44]. This gene is essential to meiosis in Saccharomyces [45] as well as to development in Drosophila [46,47] and mice [48]. Recently, TGS1 was shown to control SWI6/HP1-independent siRNA production and the establishment of heterochromatin in fission yeast [49].
To initiate the identification of molecular pathways modulated by TGS1 in P. notatum ovules, we decided to evaluate a group of RNA molecules showing differential processing in plants with attenuated TGS1 expression. First, we compared sexual and apomictic floral transcriptomes in search for differentially represented mRNA splice variants and miRNAs, taking advantage of the contrasting TGS1 activity previously described for these reproductive biotypes [37]. Then, we carried out intron-specific (for mRNAs) or stem-loop (for miRNAs) qPCR analysis in sexual tgs1 antisense lines as well as in apomictic and sexual controls, to confirm a causal association between the TGS1 knock-down and the emergence of particular splice variants/miRNAs. Using this procedure, we identified three transcripts whose processing or expression is altered in tgs1 antisense lines, i.e., LHC Ib-21 (encoding CHLOROPHYLL A-B BINDING PROTEIN 1B-21, a protein with reported RNA binding activity) [50], QGJ (encoding a MAP3K previously associated with aposporous development) [51] and miR2275 (shown to produce PHAS siRNAs essential to meiosis in rice) [52]. Our results provide information on TGS1 functional targets, open the way to further analysis of apomixis candidates, and contribute to assemble the molecular puzzle redirecting plant seed reproduction from sexuality to asexuality.

The Emergence of a Particular LHC Ib-21 Splice Variant Is Influenced by TGS1
Considering that TGS1 is required for specific RNA processing/cleavage in several organisms [42,43], we started by comparing the representation of particular mRNA splice variants in florets of apomictic and sexual plants, in which TGS1 was shown to be contrastingly expressed (i.e., overexpressed in ovules and anther tapetum of sexual plants) [37]. We first surveyed a list of DETs expressed in flowers of apomictic and sexual plants, which had been identified via RNAseq [34]. In particular, we focused on the 316 DETs with the lowest false discovery rates (FDRs < 6.74 × 10 −10 ) from the 3732 ones differentially represented between apomictic and sexual libraries at p-values ≤ 0.01 and logFC ≥ |3| [34]. The full mRNA sequences of these 316 transcripts, which had been assembled from reads originated from both sexual and apomictic libraries (Global Assembly), were used to recover all homologs from the apomictic and the sexual floral transcriptomes. Alignments of sexual and apomictic isoforms and BLAST surveys on the NCBI green plants nucleotide databases revealed that 20 DETs represented possible splice variants (Supplementary Table S1). Six of them were chosen for subsequent analysis, based on the following criteria: (1) putative differential processing occurring within internal regions of the transcript (to discard incomplete sequencing); (2) presumed alternative splicing sectors representing characterized Plants 2022, 11, 1929 4 of 19 introns in other species and (3) presence of the intron typical context GU-AG. A graphic displaying the structure of the six selected transcripts' putative splice variants from apomictic and sexual libraries are provided in Figure 1. To construct Figure 1, the variants carrying introns were used as query, to reveal the absence of this part of the sequence in the rest of the variants. Annotations and relevant statistical parameters are listed in Table 1.  Table S1). Six of them were chosen for subsequent analysis, based on the following criteria: (1) putative differential processing occurring within internal regions of the transcript (to discard incomplete sequencing); (2) presumed alternative splicing sectors representing characterized introns in other species and (3) presence of the intron typical context GU-AG. A graphic displaying the structure of the six selected transcripts' putative splice variants from apomictic and sexual libraries are provided in Figure 1. To construct Figure 1, the variants carrying introns were used as query, to reveal the absence of this part of the sequence in the rest of the variants. Annotations and relevant statistical parameters are listed in Table 1.

Figure 1.
Representation of particular splice variants in florets of sexual and apomictic P. notatum plants. Structural scheme of DETs i10779, i23387, i11548, i22343, i24572 and i22630 variants expressed in apomictic and sexual 454/Roche FLX+ libraries [34]. We used the NCBI BLASTN alignment tool to compare all homologs. The query applied to produce the alignments was denoted at the top as a thicker line and corresponded always to an unprocessed variant (with intron). Arrows inside the query mark the position of the primers designed for the qPCR assays. Since one of the primers was always located within the putative intron, the primer pair was capable of amplifying only the unprocessed variant. Sexual isotigs are represented in black and apomictic ones in gray. The variant detected as differential in the 454/Roche FLX+ libraries was marked with an asterisk (*).

Figure 1.
Representation of particular splice variants in florets of sexual and apomictic P. notatum plants. Structural scheme of DETs i10779, i23387, i11548, i22343, i24572 and i22630 variants expressed in apomictic and sexual 454/Roche FLX+ libraries [34]. We used the NCBI BLASTN alignment tool to compare all homologs. The query applied to produce the alignments was denoted at the top as a thicker line and corresponded always to an unprocessed variant (with intron). Arrows inside the query mark the position of the primers designed for the qPCR assays. Since one of the primers was always located within the putative intron, the primer pair was capable of amplifying only the unprocessed variant. Sexual isotigs are represented in black and apomictic ones in gray. The variant detected as differential in the 454/Roche FLX+ libraries was marked with an asterisk (*).
Next, we designed PCR primer pairs to amplify internal regions of those splice variants that carried introns. One primer was invariably located inside the target intron and the other one on the adjacent exon, to produce PCR amplicons specific for the non-processed variant (+intron) (Figure 1, Supplementary Table S2 and Data S1. Note that the processed variants (−intron) cannot be amplified without coamplifying the non-processed ones (+intron) (i.e., primers located on exons surrounding the missing intron amplify products of different size from both variants). Therefore, successful qPCR quantification can be achieved only for the non-processed variant (+intron). For i10779, three primer pairs were designed, corresponding to different exon-introns boundaries. Initially, non-quantitative PCR reactions were used to check the number of bands produced by each primer combination in cDNA from florets of an obligate apomictic (Q4117) and a full sexual (C4−4x) genotype ( Figure 2A). Successful fragment amplification in both plants was obtained for i10779, i23387, i11548, i22343, and i24572, while no band neither from the apomictic nor the sexual genotypes were obtained for i22630, even after applying different cycling conditions ( Figure 2A). While i10779 and i23387 homologs amplified single bands, those corresponding to i11548, i22343 and i24572 amplified more than one isoform, possibly due to the occurrence of overlapped splice variants or unspecific annealing. Therefore, they were discarded for further qPCR analysis, which requires the generation of a single product.
Real-time qPCR assays were used to amplify i10779 and i23387 on floral cDNA samples from three obligate apomictic and three fully sexual genotypes, using the same primer pairs tested in the non-quantitative PCR analyses ( Figure 2B, Supplementary Data S2). While i10779 amplification levels did not significantly differ between apomictic and sexual samples (not shown), i23387 displayed higher relative expression levels in sexual genotypes in comparison with apomictic ones ( Figure 2B, Supplementary Data S2).
Since the i23387 processed (−intron) variant cannot not be specifically amplified in qPCR experiments, we decided to examine its expression in recently delivered Illumina RNAseq libraries originated from florets of sexual and apomictic genotypes at different developmental steps (premeiosis, meiosis, postmeiosis, anthesis) [36]. Blast analysis revealed it corresponded to transcript TRpn_180064 [36] (ID: 100%; E-value: 0.0) and was overexpressed in apomictic libraries respect to sexual ones at anthesis (logFC: 17516; p-adjusted value: 0.0091) [36] ( Figure 3A). Moreover, the number of reads detected for this processed (−intron) variant was well above of that of the non-processed (+intron) one (transcript TRpn_99393) (ID: 99.85%; E-value: 0.0), which was represented at low levels in the libraries and for which no significant differential expression could be detected, due to low count ( Figure 3A). In summary, our results indicate in flowers of apomictic P. notatum plants there is a drastic upregulation of the processed (−intron) form of i23387, while the non-processed (+intron) form remains at low levels in both genotypes but, at least according to the more sensitive qPCR results, it is overexpressed in sexual ones.
To investigate a possible causal association between the differential TGS1 representation detected in apomictic and sexual plants [37] and the occurrence of particular splice variants of the i23387 homologs, we examined two independent tgs1 lines (tgs1-1: E2.9 and tgs1-2: E2.13) previously generated by transformation of a sexual P. notatum genotype with a construction carrying an antisense copy of TGS1 under the rice act1 promoter [38]. These antisense lines display a constitutively attenuated TGS1 expression, form putative apospory initials, supernumerary gametophytes with a morphology resembling AESs as well as abundant leaf trichomes [38]. qPCR analyses revealed a significantly higher expression of the non-processed (+intron) i23387 isoform in florets of the sexual control with respect to the two tgs1 antisense lines ( Figure 2C, Supplementary Data S2). These results suggest that TGS1 participates of the i23387 transcript processing. Moreover, they Plants 2022, 11, 1929 6 of 19 are in coincidence with the higher relative expression levels of non-processed i23387 form detected in sexual flowers (where TGS1 is highly expressed) in comparison with apomictic flowers (where TGS1 is naturally downregulated). In agreement with the ontology analysis made by Ortiz et al. (2017) [34], our NCBI BLASTX surveys confirmed significant homology (E-value: 3 × 10 −121 ; % ID: 84.21%) between i23387 and CHLOROPHYLL A-B BINDING PROTEIN 1B-21 (LHC Ib-21) of Setaria italica (XP_004965344) (alternative names: LHCI TYPE I CAB-1B-21; LHCI-730 CHLOROPHYLL A/B BINDING PROTEIN; LIGHT-HARVESTING COMPLEX I 21 KDA PROTEIN). Even when the detection of a photosynthesis related transcript was not expected, a possible relation between LHC Ib-21 and development is commented in the discussion. (transcript TRpn_99393) (ID: 99.85%; E-value: 0.0), which was represented at low levels in the libraries and for which no significant differential expression could be detected, due to low count ( Figure 3A). In summary, our results indicate in flowers of apomictic P. notatum plants there is a drastic upregulation of the processed (−intron) form of i23387, while the non-processed (+intron) form remains at low levels in both genotypes but, at least according to the more sensitive qPCR results, it is overexpressed in sexual ones. To investigate a possible causal association between the differential TGS1 representation detected in apomictic and sexual plants [37] and the occurrence of particular splice variants of the i23387 homologs, we examined two independent tgs1 lines (tgs1-1: E2.9 and tgs1-2: E2.13) previously generated by transformation of a sexual P. notatum genotype with a construction carrying an antisense copy of TGS1 under the rice act1 promoter [38]. These antisense lines display a constitutively attenuated TGS1 expression, form putative apospory initials, supernumerary gametophytes with a morphology resembling AESs as well as abundant leaf trichomes [38]. qPCR analyses revealed a significantly higher expression of the non-processed (+intron) i23387 isoform in florets of the sexual control with respect to the two tgs1 antisense lines ( Figure 2C, Data S2). These results suggest that TGS1 participates of the i23387 transcript processing. Moreover, they are in coincidence with Note that the processed splice variant is expressed at higher levels than the non-processed and it is differentially represented in apomictic and sexual plants at anthesis. No significant differential expression could be detected for the non-processed variant in RNAseq experiments, although in qPCR analysis (known to be more sensitive than sequencing approaches) the non-processed variant is consistently detected overexpressed in sexual plants at anthesis. (B) Normalized read counts corresponding to the processed (TRpn_64321) and non-processed (TRpn_120075) QGJ splice variants. The processed splice variant is upregulated in sexual plants at premeiosis and remains with higher expression along development (yet not at statistically significant levels). The non-processed variant is expressed at low levels. Premei, Mei, Postmei and Anth indicate ovules at the developmental stages premeiosis, meiosis, postmeiosis and anthesis, respectively. The asterisk denotes the differential expression significance (* p < 0.05).

The Expression of the MAP3K QUI-GON JINN (QGJ) Is Influenced by TGS1
As an independent strategy to identify additional TGS1 targets, we focused on the mitogen activated protein kinase kinase kinase (MAP3K) QUI-GON JINN (QGJ), which is necessary for the formation of AESs in Paspalum [51]. Since attenuation of TGS1 leads to the development of AES-like gametophytes in the same species [38], we decided to investigate a possible functional link between both genes. QGJ is represented by two splice variants in apomictic and sexual plants [51]. We used non-specific primers (Supplementary Table S2) to measure: (i) the relative TGS1 expression levels in two constitutive qgj RNAi lines (qgj-1, qgj-2) generated by transforming an apomictic genotype with a hairpin QGJ construction [51]; and (ii) the relative QGJ expression levels in one of the constitutive tgs1 (tgs1-1) antisense line used in the former section [38]. No differential TGS1 expression was detected in the qgj defective lines with respect to an apomictic control at anthesis (when expression of TGS1 peaks [37]; the apomictic control used here was the same genotype originating the qgj defective line by biolistic transformation, i.e., Q4117) ( Figure 4A, Supplementary Data S2). Contrarily, a statistically significant downregulation of QGJ expression was detected in both the tgs1 defective line and the apomictic control with respect to the sexual control at meiosis (when the highest QGJ expression is detected [51]; the sexual control was C4-4x; the apomictic control was Q4117) ( Figure 4B, Supplementary Data S2). The latest result indicates that a downregulation of TGS1 have the same impact in the total floral concentration of QGJ at meiosis as observed in apomictic plants (a decrease, possibly due to downregulation in pollen mother cells, according to Mancini et al. 2018) [51].
(TRpn_120075) (ID 96.96%; E-value: 0.0) equally expressed in sexual and apomictic samples at all developmental stages, while the processed (−intron) variant (TRpn_64321) (ID 99.93%; E-value: 0.0) resulted significantly upregulated in sexual plants at premeiosis, and remained upregulated at other developmental stages but at statistically non-significant levels ( Figure  3B). Altogether, these data indicate that, at premeiosis/meiosis, a downregulation of TGS1 causes an alteration in the expression level of the processed (−intron) QGJ isoform but not in the non-processed (+intron) one.  Next, we investigated the previously-reported occurrence of two QGJ splice variants. First, we used a primer pair that specifically identifies the QGJ non-processed (+intron) splice variant in qPCR experiments (Supplementary Table S2). No significant variation was observed for this particular isoform ( Figure 4C,D, Supplementary Data S2), which is in agreement with results presented by Mancini et al. (2018) [51], who detected no differences in levels of the non-processed variant in sexual and apomictic plants. Moreover, we checked the Illumina libraries published by Podio et al. (2021) [36] and found the non-processed (+intron) splice variant (TRpn_120075) (ID 96.96%; E-value: 0.0) equally expressed in sexual and apomictic samples at all developmental stages, while the processed (−intron) variant (TRpn_64321) (ID 99.93%; E-value: 0.0) resulted significantly upregulated in sexual plants at premeiosis, and remained upregulated at other developmental stages but at statistically non-significant levels ( Figure 3B). Altogether, these data indicate that, at premeiosis/meiosis, a downregulation of TGS1 causes an alteration in the expression level of the processed (−intron) QGJ isoform but not in the non-processed (+intron) one.

In Situ Analysis of QGJ Expression in tgs1 Antisense Lines
As we detected a change in the quantitative expression level of QGJ in florets of antisense tgs1 lines, in situ experiments were designed to test if there were modifications in the spatial distribution pattern of QGJ in the ovule, as it was observed in aposporous plants with respect to sexual ones ( Figure 5). At premeiosis, in both tgs1 antisense lines and wild type sexual plants, the QGJ signal was detected across the whole nucellus and within the MMC, but while in sexual genotypes the signal was generally robust, in tgs1 lines was rather fainter ( Figure 5A,D). During meiosis, in the control sexual plants the QGJ expression shifted to the micropylar region and was also observed in the funiculus, but had lower levels in the chalaza ( Figure 5B); the degenerating megaspores showed moderate to intense signal, but the functional one (at a more chalazal position) had no signal. On the contrary, during meiosis, in tgs1 antisense lines the signal moved to the chalaza instead of the micropyle, and was also observed in the funiculus as well as in all meiotic products ( Figure 5E). Regarding male development, in control sexual genotypes a consistent signal was observed within pollen mother cells (PMCs) but not in the tapetum ( Figure 5C). Nevertheless, in anthers of tgs1 genotypes there was moderate signal in the tapetum, while PMCs showed heterogeneous staining (i.e., some of them displayed no signal, while others did) ( Figure 5F). The sense probe showed undetectable hybridization signals ( Figure 5G-J). In general, results obtained here for the tgs1 antisense lines are analogous to those presented by Mancini et al. (2018) [51] for florets of apomictic P. notatum and Brachiaria brizantha plants. Additional images of the in situ experiments were provided in Supplementary Figure S1. This evidence strongly suggests that the silencing of TGS1 causes a shift of QGJ expression from the micropyle to the chalaza in premeiotic/meiotic ovules, and from pollen mother cells to the tapetum in premeiotic/meiotic anthers, which means that QGJ expression moves from a typical sexual expression pattern to an apomictic-like pattern.

In Situ Analysis of QGJ Expression in tgs1 Antisense Lines
As we detected a change in the quantitative expression level of QGJ in florets of antisense tgs1 lines, in situ experiments were designed to test if there were modifications in the spatial distribution pattern of QGJ in the ovule, as it was observed in aposporous plants with respect to sexual ones ( Figure 5). At premeiosis, in both tgs1 antisense lines and wild type sexual plants, the QGJ signal was detected across the whole nucellus and within the MMC, but while in sexual genotypes the signal was generally robust, in tgs1 lines was rather fainter ( Figure 5A,D). During meiosis, in the control sexual plants the QGJ expression shifted to the micropylar region and was also observed in the funiculus, but had lower levels in the chalaza ( Figure 5B); the degenerating megaspores showed moderate to intense signal, but the functional one (at a more chalazal position) had no signal. On the contrary, during meiosis, in tgs1 antisense lines the signal moved to the chalaza instead of the micropyle, and was also observed in the funiculus as well as in all meiotic products ( Figure 5E). Regarding male development, in control sexual genotypes a consistent signal was observed within pollen mother cells (PMCs) but not in the tapetum ( Figure 5C). Nevertheless, in anthers of tgs1 genotypes there was moderate signal in the tapetum, while PMCs showed heterogeneous staining (i.e., some of them displayed no signal, while others did) ( Figure 5F). The sense probe showed undetectable hybridization signals ( Figure 5G Figure S1. This evidence strongly suggests that the silencing of TGS1 causes a shift of QGJ expression from the micropyle to the chalaza in premeiotic/meiotic ovules, and from pollen mother cells to the tapetum in premeiotic/meiotic anthers, which means that QGJ expression moves from a typical sexual expression pattern to an apomictic-like pattern.  the signal was transferred to the micropylar region and the funiculus; the functional megaspore has no signal; the degenerating megaspores are with signal; (C) Anthers of Q4188: no signal is observed in the tapetum; almost all pollen mother cells display signal (D) Premeiotic ovule of tgs1.1: moderate to low signal is observed in the whole ovule, including the MMC (E) Postmeiotic ovule of tgs1.1: the signal was transferred to the chalaza and the funiculus and is detected inside all meiotic products, which are surrounded by a cell layer with lower signal; an enlarged cell reminiscent to an apospory initial is located aside the meiotic products, within the less-hybridized area; (F) Anthers of tgs1.1: moderate signal is observed in the tapetum; only some of the pollen mother cells display strong signal, the rest remains unstained. (G-H) meiotic ovules of sexual and tgs1 lines, respectively, hybridized with QGJ sense probe (10×). (I-J) premeiotic ovules of sexual and tgs1 lines, respectively, hybridized with QGJ sense probe (40×). ai: a putative apospory initial cell; an: anthers; cha: chalaza; dm: degenerated megaspores; fm: functional megaspore; fu: funiculus; mi: micropyle; mmc: megaspore mother cell; pmc: pollen mother cells; tp: tapetum. Bars: 20 µm.

Identification of miRNA Variants Associated with TGS1 Activity
In human fibroblasts, TGS1 mediates the cleavage of a group of pri-pre-miRNA precursors in order to produce mature miRNAs associated with EXPORTIN 1 [44]. We could not identify the orthologs to such group of miRNAs in Paspalum, possibly due to low conservation. However, since TGS1 is naturally upregulated in spikelets of sexual plants, we decided to search miRNAs which were overrepresented in floral transcriptome libraries of sexual plants with respect to apomictic ones. To do so, we exploited the P. notatum floral small RNA libraries of apomictic and sexual individuals generated by Ortiz et al. (2019) [35]. Based on the quantitative analysis of miRNA representation, only one miRNA (miR2275) was detected upregulated in spikelets of sexual plants in comparison with apomictic ones. miR2275 showed a normalized counting of 23 reads in the apo libraries and 48 in the sexual ones [35]. Meanwhile, in long-read Roche 454/FLX + libraries involving the same floral samples, the miR2275 precursor (Cluster_47025, sex_isotig38116) was detected twice in the sexual sample but was absent from the apo one [34].
In order to assess the miR2275 in vivo representation within in the P. notatum floral system, a stem-loop PCR assay was designed to amplify the mature miRNA in sexual control plants and a tgs1 defective line (Supplementary Figure S2 and Table S2). At meiosis, miR2275 resulted significantly downregulated in the tgs1 line, reaching levels closer to those detected in apomictic plants ( Figure 6A, Supplementary Data S2), yet still significantly higher. As miR2275 was reported as a potential AGO1 repressor in Paspalum [35], we decided to check the levels of miR168 (a characterized AGO1 repressor) in ovules of sexual/apomictic controls and tgs1 defective lines. Interestingly, miRNA168 was also found differentially expressed in sexual and aposporous plants, with overexpression in apomictic ones [35]. However, our stem-loop PCR analysis showed similar levels of miR168 representation in tgs1 defective and sexual lines, which were, as expected, lower that those observed in apomictic lines ( Figure 6B, Supplementary Data S2).

Discussion
In the last few years, an unprecedented amount of data originated from genome an These results show that two potential AGO1 controllers (miR2275 and miR168) are differentially expressed in apomictic and sexual P. notatum ovules at premeiosis (upregulated in sexual and apomictic plants, respectively), but only one of them (miR2275) seems to be modulated by TGS1. Then we checked the AGO1 expression in the available P. notatum floral transcriptome libraries of sexual and apomictic plants [36]. We found 21 detectable AGO1-like transcripts, from which 8 are expressed at considerable levels. The most expressed transcript (TRpn_87312) is significantly upregulated in sexual plants at anthesis [36]. The rest of the transcripts are significantly upregulated in apomictic plants at all stages (TRpn_171351, TRpn_104454, TRpn_153031, TRpn_176544), at premeiosis, meiosis and anthesis (TRpn_86127), at postmeiosis and anthesis (TRpn_109434) or only at meiosis (TRpn_103957) [36].

Discussion
In the last few years, an unprecedented amount of data originated from genome and transcriptome sequencing projects have flooded the apomixis field. Hundreds of candidate genes showing differential expression in sexual and apomictic plants were identified at different reproductive developmental stages [53]. However, little is known on the functional role of these candidates, the operative interactions among them or the identity of the molecules controlling the coexistence of a subtle balance between both reproductive types in natural populations. The identification of molecular markers co-segregating with apomixis, the generation of artificial sexual polyploids after colchicine duplication, the construction of transcriptome and genomic databases and the establishment of biolistic transformation platforms represent good perspectives for selecting traits of interest in natural apomictic species such as P. notatum [21]. In fact, heterosis for forage yield and cold tolerance has been repeatedly reported in this species, and an upright and fast-growing apomictic P. notatum hybrid was recently released as a forage cultivar [21]. Instead, the harnessing of apomixis in major crops such as rice and maize requires a much more detailed knowledge of the molecular mechanisms controlling the balance between apomixis and sexuality. The establishment of such programs requires sophisticated molecular tools and might bring new ecological challenges related with spreading the trait via pollen and seeds (e.g., uniparental reproduction, unidirectional gene transfer) [17]. Therefore, before any attempt of using the trait, we should expand our information on the molecular, functional and organizational mechanisms operative in natural apomictic plant populations, to improve our capacity to evaluate and avoid any damage.
Natural apomicts display a wide range of developmental approaches to balance their competence for both genetic variation and cloning. Both features can coexist in the same species (i.e., confined at different ploidy levels), within the same plant (i.e., in facultative apomicts) and even the same ovule (i.e., occurrence of polyembryony of sexual and apomictic origin). Moreover, the proportion of offspring formed by each reproductive mode can be influenced by the environment [17]. In any case, there seem to be regulators operating to favor apomixis or sexual reproduction in different contexts (variable ploidy, diverse genetic backgrounds, particular environmental conditions), which could be harnessed to abolish the ecological impact of the trait ensuring a safe use in agriculture. Understanding the functional interactive dynamics of both reproductive modes at the molecular level will be crucial to predict how an apomictic crop may behave in natural fields and visualize potential ecological threats.
In connection with this, only a few sexuality/agamospermy switch regulators have thus far been reported in natural apomictic species [53]. Among them, TGS1 is both a repressor of the formation of supernumerary AESs-like gametophytes from the nucellus of sexual individuals and a promoter of the meiosis capacity [38]. Once described as a gene with a physiological function (i.e., chilling tolerance) [54], it was also found differentially expressed in reproductive organs of sexual and apomictic Paspalum plants [37] and its attenuation in sexual individuals caused a reduction in the meiosis capacity and the emer-gence of AES-like gametophytes, pointing to an parallel function as a major developmental modulator of the apomixis-sexuality switch [38].
The availability of sexual P. notatum tetraploid plants with impaired TGS1 activity [38] enabled studies aimed at identifying its target transcripts in Paspalum notatum ovules, in order to investigate the possible molecular pathways controlling the apomictic-sexuality connection. The first candidate we identified (CHLOROPHYLL A-B BINDING PROTEIN 1B-21) shows upregulation of a particular splice variant in sexual plants with respect to apomictic ones and tgs1 lines, which might be related with the TGS1 function as a major splicing controller as reported in other eukaryotic systems [42,43]. The modulation of a member of the CHLOROPHYLL A-B BINDING PROTEIN 1B-21 in heterotrophic reproductive organs was an unanticipated result, since we were expecting candidates with a female reproductive function. However, light-induced phenotypes had already been described in antisense tgs1 plants (i.e., the development of abundant leaf trichomes under a light regime) [38]. Besides, recently several components of the photosynthesis machinery were included within a group of RNA binding proteins (RBPs) with a key role in RNA metabolism [50]. Moonlighting functions have been hypothesized for these proteins, even when further analysis is required to understand the crosstalk between photosynthesis and RNA metabolism [50]. About 8% of the leaf RBPs described by Bach-Pages et al. (2020) [50] have annotations related to photosynthesis or photosystems, and many photosynthesis-related domains are enriched in the leaf RBPome, including the CHLOROPHYLL A-B BINDING domain. Moreover, 33 photosynthesis-related leaf RBPs have been independently shown to associate with RNA in other Arabidopsis tissues [55,56] and some photosynthesis components, such as the large subunit of rubisco (LSU) and cytochrome f, are known to bind RNA [57][58][59][60]. In future work, the link among the TGS1 capacity to repress the formation of trichomes [38], the TGS1-mediated regulation of LHC Ib-21 splice variants shown here and the capacity of LHC Ib-21 to bind RNA [50] should be investigated in the context of the previously reported influence of the chloroplast on alternative splicing [61] and the well-known relation between trichome formation and light incidence.
We also investigated the functional link between TGS1 and the transcript mitogenactivated 3-kinase QGJ, which was previously described as differentially represented in reproductive organs of sexual and aposporous plants and to be essential for AESs formation [51]. Natural apomictic plants show a downregulation of TGS1 in the ovule chalaza and the anther tapetum [37], concurrent with a spatial deregulation of the QGJ representation (upregulation in chalaza, proximal nucellus and anther tapetum; downregulation in pollen mother cells and the cell layer surrounding the MMC) [51]. In qPCR analysis involving spikelets (somatic tissues, ovules and anthers), apomictic plants display lower quantitative QGJ levels with respect to sexual plants, since anthers represent most of the floret tissue [51]. Two alternative QGJ splice variants are expressed in Paspalum florets, one of them including an intron that introduces several additional aminoacids and shifts the open reading frame. However, no differential apo/sex representation was detected for the non-processed variant, the only one that can be evaluated in qPCR assays [34,51]. In our experiments, antisense tgs1 lines showed a QGJ expression profile analogous to that of apomictic plants, suggesting that the attenuation of TGS1 might be causing the deregulation of the QGJ expression detected during aposporous apomixis. While constitutive repression of TGS1 in sexual plants induces the formation of AES-like gametophytes from somatic cells of the nucellus and/or the chalaza [38], down-regulation of QGJ in the ovule of aposporous plants impairs the development of AES [51], suggesting that both genes belong to antagonist pathways. Thus, the modulation exercised on QGJ could partially imply the activity of a non-cell autonomous mechanism, that, as we will discuss in the next paragraph, might be mediated by epigenetic mechanisms.
In human fibroblasts, TGS1 was shown to control the processing of a group of miRNAs associated with the Exportin 1 function [44]. Given that the low degree of conservation among plants and human miRNAs prevents direct identification of homologs via sequence similarity analysis, and considering that sexual P. notatum plants have upregulated TGS1 activity, we decided to examine the influence of TGS1 expression on miRNAs overexpressed in sexual plants. Using previously available sRNA databases, we detected a single miRNA significantly upregulated in spikelets of sexual plants: miR2275 [35]. Stem-loop qPCR analysis revealed that this molecule was downregulated in tgs1 lines at meiosis, reaching similar levels to those detected in apomictic plants. Interestingly, in rice and maize, miR2275 was shown to trigger a 24-nt phasiRNA pathway mediated by a specific DCL protein (DCL 5), which is essential to meiosis [52,62]. Recently, Xia et al. (2019) [63] reported that the miR2275/24-nt phasiRNA pathway is widely present in eudicots plants, but absent in wellcharacterized species of the Fabaceae and Solanaceae, as well as in the model plant Arabidopsis. In rice anthers, miR2275 is formed in the tapetum cell layer and trigger a 24-nt phasiRNAs pathway necessary for meiosis in pollen mother cells [52]. Migration of the 24-nt phasiRNAs from the site of origin (the tapetum) to pollen mother cells was proposed as the only possible regulatory mechanism to promote meiosis [62]. As commented above, in Paspalum, TGS1 is expressed in the tapetum of sexual plants (a location coincident with miR2275 synthesis site in rice), as well as in the ovule nucellus/chalaza [37]. A decrease in TGS1 expression is detected in apomictic plants at both locations [37], concurrently with a general decrease in the miRNA2275 levels [35], the formation of extra megagametophytes [38] and a reduction in pollen viability [27,64]. Moreover, tgs1 antisense lines show lower miRNA2275 levels (results reported here) and emergence of AES-like structures [38]. An interesting point is that the P. notatum predicted a target of miR2275 is AGO1 [35], a transcript previously implicated in meiotic development in yeasts [65]. AGO1 was also found overexpressed in egg cells [66] and differentially expressed in male and hermaphrodite gametophytes of the homosporous fern Ceratopteris richardii [67], which suggest a particular role during female gametophyte development.
Based on the information presented here as well as other reported in the articles commented in this Discussion [37,38,51,52,62] we propose the hypothetical model for TGS1 action shown in Figure 7, which could be useful to plan future experiments regarding the elucidation of the apomixis molecular control. According to our model, in sexual plants, TGS1 is expressed in the ovule chalaza and the anther tapetum, where it promotes the generation of mature miR2275 molecules. In turn, miR2275 produces 24nt-meiotic phasiRNAs, which are redirected to the MMC and the PMCs to promote the entrance into meiosis, which occurs concurrently with the acquisition of a gametophytic fate (i.e., the commitment to eventually form a gametophyte) via induction of QGJ. Besides, TGS1 repress the expression of the processed QGJ variant within its own expression domain by a still unknown mechanism. In apomictic plants and antisense tgs1 lines, the expression of TGS1 is attenuated, thus mature miR2275 are not effectively formed and the delivery of the meiotic 24nt-phasiRNAs becomes compromised. Therefore, meiosis in both the MMC and PMCs is impaired. Moreover, the absence of TGS1 in the chalaza allows an ectopic expression of QGJ at this location and the concomitant induction of a gametophytic fate in nucellar somatic cells, with the consequent emergence of extra non-reduced gametophytes (Figure 7).
After considering the growing amount of evidence reported in the last few years, TGS1 shapes up as a major promoter of sexuality and inhibitor of aposporous apomixis [38] and emerges as a possible candidate for a reversible switch going from one mode of reproduction to the other. Since TGS1 is a well-known controller of the splicing and transcription machinery in other organisms, the identification of the molecular routes underlying its function in the plant ovule will certainly lead to a general conception of the mechanisms modulating the reproductive developmental transitions, a requirement for thoughtful ecological evaluation and future harnessing of clonal reproduction into the breeding of major crops. apomictic plants and antisense tgs1 lines, the expression of TGS1 is attenuated, thus mature miR2275 are not effectively formed and the delivery of the meiotic 24nt-phasiRNAs becomes compromised. Therefore, meiosis in both the MMC and PMCs is impaired. Moreover, the absence of TGS1 in the chalaza allows an ectopic expression of QGJ at this location and the concomitant induction of a gametophytic fate in nucellar somatic cells, with the consequent emergence of extra non-reduced gametophytes (Figure 7). TGS1 could be repressing the expression of QGJ within its own activity domain, but inducing it in the germinal line via a non-cell autonomous miR2275/24-phasiRNA-mediated mechanism. In absence of TGS1 (i.e., apomictic genotypes and tgs1 antisense lines) QGJ would overexpress ectopically across the canonical TGS1 activity domain (the anther tapetum and the chalaza). Around the MMC, a limited number of cell layers show low QGJ activity, allowing the differentiation of apospory initials (i.e., cells with functional megaspore identity, but non-reduced). Besides, the absence of TGS1 is also associated with a decrease in miR2275 and the loss of the miR2275/24-phasiRNA pathway. The latter condition compromises the induction of QGJ within MMCs and pollen mother cells, impairing proper differentiation of the meiotic germline. mmc: megaspore mother cell.
After considering the growing amount of evidence reported in the last few years, TGS1 shapes up as a major promoter of sexuality and inhibitor of aposporous apomixis [38] and emerges as a possible candidate for a reversible switch going from one mode of reproduction to the other. Since TGS1 is a well-known controller of the splicing and transcription machinery in other organisms, the identification of the molecular routes underlying its function in the plant ovule will certainly lead to a general conception of the mechanisms modulating the reproductive developmental transitions, a requirement for thoughtful ecological evaluation and future harnessing of clonal reproduction into the breeding of major crops.

Identification of Splice Variants in Floral Transcriptomes
Splice variants were initially identified from a long-read 454/Roche (Branford, CT, USA) FLX+ (454 Life Sciences Corporation, Branford, Connecticut, USA) reference floral transcriptome of sexual and apomictic P. notatum, in which raw data originated from floral samples had been assembled de novo into sexual, apomictic and global (sexual + apomictic) transcriptomes (NCBI Bioproject: PRJNA330955) [34]. DETs identified using the global transcriptome as reference were ordered according to the False Discovery Rate (FDR) value [34]. The top 316 genes with the lowest FDRs were selected to investigate the presence of alternative splice variants in the apomictic (NCBI SRX1971037) and the sexual (NCBI SRX1971038) assemblies. The apomictic and sexual transcripts were aligned with Clustal Omega (http://www.ebi.ac.uk/Tools/msa/clustalo/) (accessed on 1 January 2020).
Besides, alignments were made with BLASTN/BLASTX against the NCBI non-redundant databases. The differential expression of particular splice variants was also tested in Illumina HiSeq floral transcriptome libraries [36].

Qualitative PCR
Qualitative PCR amplifications were performed on cDNA samples generated from floral total RNA. Primers pairs used are shown in Supplementary Table S2 and were designed with Primer3 v.0.4.0 (http://bioinfo.ut.ee/primer3-0.4.0/) (accessed on 1 January 2020). To validate splice variants, one of the primers was located on internal sequences of the putative intron and the other one on the adjacent exon. Total RNA was extracted from spikelets at stage VII (i.e., with highest TGS1 expression) according to the Paspalum reproductive calendar [33] and using the SV Total RNA Isolation Kit (PROMEGA, Madison, WI USA). cDNAs were synthetized with Superscript II (INVITROGEN, Carlsbad, CA, USA) following the manufacturers' recommendations. PCR reactions were carried out in a final volume of 25 µL containing: 50 ng cDNA, 0.5 µM of each specific primer (Table 1)

qPCR Experiments
For qPCR, total RNA was extracted from spikelets at the required developmental stage [33] by using the SV Total RNA Isolation Kit (PROMEGA, Madison, WI, USA). Primers pairs used are shown in Table S2 and were designed with Primer3 v.0.4.0 (http: //bioinfo.ut.ee/primer3-0.4.0/) (accessed on 1 January 2020). cDNAs were synthetized with Superscript II (INVITROGEN, Carlsbad, CA, USA) following the manufacturers' recommendations. Quantitative PCR reactions (final volume: 20 µL), were performed in a Rotor-Gene Q thermocycler (QIAGEN, Hilden, Germany) and included: 0.5 µM genespecific primers (Table S1), 1× Real Mix qPCR (BIODYNAMICS, Buenos Aires, Argentina) and 20 ng of cDNA. In each experiment, two biological replicates were processed, in which each determination was run in three technical replicates. The constitutive gene β-TUBULIN was used as housekeeping reference, since in previous work this gene had been selected as the best reference for comparisons in this particular floral apomictic-sexual Paspalum system [70][71][72]. Negative controls were processed without template DNA. Amplifications were performed with the following program: 2 min at 94 • C, 45 cycles of 94 • C for 15 min, 57 • C for 30 s, and 72 • C for 17 s, and a final elongation step of 5 min at 72 • C. Relative quantitative expression levels were assessed using the REST-RG 2009 software (QIAGEN, Hilden, Germany).

Conclusions
The RNA methyltransferase TGS1 participates in the alternative splicing of a LHC Ib-21, a protein of the chloroplastic light harvesting antenna with a predicted uncharacterized moonlighting function related with RNA regulation. However, TGS1 is responsible for the expression level of a processed transcript encoding the mitogen activated protein kinase kinase kinase (MAP3K) QGJ in florets of sexual plants. Moreover, in sexual Paspalum flowers, TGS1 induces the expression of miR2275, a miRNA responsible for the synthesis of essential meiotic 24-nt phasi-RNA previously characterized in maize and rice. In the context of previous results, the evidence shown here suggests that the TGS1 action on QGJ and miR2275 might: (i) trigger a molecular route leading to the acquisition of gametophyte commitment and the induction of meiosis in the MMC and pollen mother cells, and (ii) repress gametophyte commitment in cells surrounding the MMC, so in absence of TGS1, the spatial QGJ distribution is altered, inducing the formation of unreduced supernumerary gametophytes.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/ 10.3390/plants11151929/s1, Figure S1: Additional images of QGJ in situ hybridization analysis in control and tgs1 lines, Figure S2: Stem-loop RT-PCR scheme for miRNA225a and miRNA168, Data S1: Primer design for specific amplification of the splice variants carrying introns (non-processed forms), Data S2: qPCR data analysis, Table S1: List of the 20 selected transcripts representing putative differentially-expressed splice variants in libraries of apomictic and sexual P. notatum plants; Table S2: Primers used in qPCR and stem-loop qPCR analysis.

Data Availability Statement:
The RNAseq sequence datasets used in this work are openly available at the NCBI repository under the accession numbers SRX1971037 y SRX1971038 (454/Roche FLX+), SRP099144 (small RNAs) and PRJNA511813 (Illumina sequencing). All plant materials used here are maintained at the IICAR greenhouses or controlled chambers and are available after official request for academic purposes.