Overexpressing lncRNA LAIR increases grain yield and regulates neighbouring gene cluster expression in rice

Long noncoding RNAs (lncRNAs) are essential regulators of gene expression in eukaryotes. Despite increasing knowledge on the function of lncRNAs, little is known about their effects on crop yield. Here, we identify a lncRNA transcribed from the antisense strand of neighbouring gene LRK (leucine-rich repeat receptor kinase) cluster named LAIR (LRK Antisense Intergenic RNA). LAIR overexpression increases rice grain yield and upregulates the expression of several LRK genes. Additionally, chromatin immunoprecipitation assay results indicate H3K4me3 and H4K16ac are significantly enriched at the activated LRK1 genomic region. LAIR binds histone modification proteins OsMOF and OsWDR5 in rice cells, which are enriched in LRK1 gene region. Moreover, LAIR is demonstrated to bind 5′ and 3′ untranslated regions of LRK1 gene. Overall, this study reveals the role of lncRNA LAIR in regulating rice grain yield and lncRNAs may be useful targets for crop breeding.

R ice (Oryza sativa L.) is a staple food crop for more than 3 billion people worldwide 1 . It is also considered as one of the most important model monocotyledonous plants for genome research due to its small genome size and considerable synteny with other cereal genomes 2,3 . Yield is a crucial agronomic trait for cereal crops' genetic improvement. Its improvement relies on the manipulation of a range of component traits such as number of panicles, grains per plant, and 1000-grain weight, which often exhibit continuous phenotypic variations. It has been shown that these traits are regulated by quantitative trait loci (QTLs), whose effects are usually mediated by at least two genes and environmental conditions.
The traditional definition of a gene has recently been challenged by the discovery of noncoding RNAs (ncRNAs), which include functional transcripts encoded by a specific genome region. A large-scale sequencing analysis reveals that most of the eukaryotic genome is transcribed 4 , and large repertoires of RNAs are generated, including short (22-33 nucleotides) and long (>200 nucleotides) ncRNAs. Long noncoding RNAs (lncRNAs), which lack any obvious open reading frames (ORFs) [5][6][7] , function as key regulators of diverse mechanisms in a range of biological processes 6,8,9 . They may regulate the expression of neighbouring genes with cis-acting (close to the transcription site) or transacting (multiple loci throughout the genome) mechanisms 10 . Approximately one-third of lncRNAs can interact with multiple protein partners 11 . The concept of lncRNAs serving as molecular scaffolds that help assemble and target the chromatin-modifying complex to specific genomic loci for the regulation of gene expression is likely part of a globally emerging mechanistic theme 6,12 . One major theme involves the regulatory role of lncRNAs regarding the expression of nearby protein-coding genes or gene clusters. For example, HOX transcript antisense RNA (HOTAIR), one of the first identified trans-acting lncRNAs, binds to multiple genomic loci of HOXD genes and serves as a molecular scaffold that enables the recruitment of chromatinmodifying complexes to repress the expression of HOXD genes 13,14 . In Drosophila melanogaster, the dosage compensation complex with lncRNAs [RNA on X1 (roX1) and RNA on X2 (roX2)] that combine with proteins [male-specific lethal-1 (MSL-1), MSL-2, MSL-3, Males-absent-on-the-first (MOF), and Maleless (MLE)] acetylates H4K16 so that chromatin can assume a more open conformation in the hyperactive cells of male fruit flies [15][16][17] . Genomic occupancy maps of roX RNAs have also been done to show the targeting of X chromosomes 18 .
In plants, it is known that lncRNAs can regulate gene silencing, flowering time, reproduction, stress responses, organogenesis in roots, and photomorphogenesis in seedlings [19][20][21][22][23][24] . Because the overwhelming majority of well-characterized plant lncRNAs with established functions were studied in Arabidopsis thaliana, our understanding of the mechanisms regulating lncRNAs in crop species remains limited. Thus, elucidating the mechanisms of action for the rice lncRNAs may clarify the functions of ncRNAs in different species and provide useful information for the breeding of food crops. We previously detected and cloned a leucine-rich repeat receptor kinase (LRK) gene cluster from a QTL associated with increased grain yield in rice 25 . In this study, we identify a lncRNA transcript (LRK Antisense Intergenic RNA, LAIR) from the antisense strand of the LRK gene cluster. Transgenic lines, in which LAIR is silenced, exhibit inhibited growth and low expression levels for all LRK genes. In contrast, considerable increases in total grain yield and upregulated expression of some members of the LRK gene cluster are observed in lines overexpressing LAIR. Moreover, luciferase reporter assays reveal that LAIR is capable of differentially activating promoters of the LRKs gene. Meanwhile, an analysis of chromatin modifications related to LRK genes indicates H3K4me3 and H4K16ac are enriched in transcriptionally active genes. Next, LAIR binds histone modification proteins OsMOF and OsWDR5 (WD repeat domain 5) in rice cells, which are detected enrichment in the LRK1 gene region. Additionally, ChIRP assays reveal that LAIR binds 5′ and 3′ UTR genomic regions of the LRK1 gene. These results imply that LAIR overexpression increases rice yield and regulates the expression of neighbouring LRK gene cluster.

Results
Identification of the antisense lncRNA LAIR. We previously identified a QTL for improved rice yield consisting of several tandemly arranged intronless LRK gene copies in a 51.3-kb region of chromosome 2 25 . The LRK gene cluster in the O. sativa ssp. indica MH63 genome was subsequently sequenced. Genomic and transcriptomic analyses revealed the LRK gene cluster in MH63 contained eight highly similar LRK genes (LRK1-8), which were transcribed normally, with the exception of LRK5 (Fig. 1a).
An antisense transcript encoded by the 5′ terminal region of the LRK gene cluster was named LAIR. The LRK1 gene was located in the first intron of LAIR, and there was no overlap between the two antisense transcripts (Fig. 1a). The 5′ and 3′ ends of LAIR were determined by rapid amplification of cDNA ends (RACE) ( Supplementary Fig. 3). We observed that LAIR contains a 5′ cap structure and a polyadenylated 3′ end, which is typical of mRNAs. Thus, it is likely that RNA polymerase II (RNAPII) is responsible for the transcription of LAIR. The spatial and temporal expression of LAIR was analyzed, LAIR showed relatively high levels in 3-leaf-stage shoot, flowering-stage node, pistil and caryopsis ( Supplementary Fig. 1). Additionally, up to 10 alternative spliced isoforms of LAIR were identified and sequenced (JX512719-JX512728). All splicing events were detected in the second exon of LAIR, and consisted of the shuffling of specific motifs rather than random events. The most enriched LAIR RNA (JX512726) was 1797 bp long, while the shortest LAIR RNA (JX512719) was 1585 bp long. The corresponding genomic and transcript sequences were up to 7.3 kb and approximately 1.8 kb long, respectively (Supplementary Fig. 2 and Fig. 1a).
We evaluated the coding potential of LAIR. The ORF Finder online tool (https://www.ncbi.nlm.nih.gov/orffinder/) was used to predict ORFs based on LAIR isoform JX512726. As indicated in Fig. 2a, the termination codon was distributed in all three frames, and only a few sequences encoding short peptides were detected. The two longest peptides comprised 133 and 104 amino acids. Furthermore, there were no functional domain matches for any of the peptides according to homology searches of large databases of protein families and domains (Pfam 26 and SMART 27 ). The possibility that these peptides function as small proteins was assessed by the following genetic investigations. An analysis of the protein-coding potential of transcripts 28 indicated that all 10 alternative spliced isoforms of LAIR represented ncRNAs (Fig. 2b). Additionally, the DNA sequences of the different alternatively spliced LAIR isoforms had diverse coding frames. Therefore, we hypothesized that there is no functional constraint associated with the LAIR coding regions.
To summarize, we identified an antisense lncRNA (LAIR) from the 5′ terminal region of the LRK gene cluster; further analyses indicated its association with rice grain yield.
LAIR affects rice yield traits. To test whether LAIR is functional in individual plants, we conducted genetic transformation experiments. Full-length cDNA sequences of sense or antisense LAIR (JX512726) were cloned and transformed into rice MH63 under the control of the 35S Cauliflower Mosaic Virus (CaMV) promoter (35S::LAIR and 35S::anti-LAIR). Compared with the wild-type control plants, the 35S::LAIR plants had larger primary panicles and more panicles per plant, leading to a considerable increase in total grain yield per plant (Fig. 1b-e). In contrast, the 35S::anti-LAIR transgenic plants exhibited inhibited growth and decreased grain yield. Similar rice growth traits were reported for LRK1-overexpressing transgenic lines 29 . These results demonstrate that LAIR expression affects rice plant growth and grain yield.
The LAIR transcript is long (approximately 1.8 kb) and was predicted to lack coding potential according to the ORF and functional domain analyses. To investigate whether LAIR functions as a lncRNA, we mutated the sequences encoding the two longest peptides (133 and 104 amino acids according to the JX512726 sequence). Two LAIR mutant vectors were constructed, with a single nucleotide mutation introducing a premature stop codon in the LAIR-MU1 mutant, and a single deleted nucleotide after the start codon (ATG) in the LAIR-MU2 mutant (Fig. 2c). We analyzed transgenic rice MH63 plants carrying the 35S:: LAIR-MU1 and 35S::LAIR-MU2 constructs and observed that the morphological characteristics and grain yield of these mutants LAIR lines were similar to those of the wild-type LAIRoverexpressing (35S::LAIR) transgenic lines ( Fig. 2d-f). These results implied that LAIR functions were not dependent on the encoded products. Thus, LAIR likely serves as a lncRNA that regulates rice plant growth and increases grain yield.
LAIR influences the transcription of LRK gene cluster. An earlier study that confirmed non-coding loci can affect the expression of neighbouring protein-coding genes was critical for the functional characterization of ncRNAs 30 . Additionally, the 35S::LAIR transgenic rice lines exhibited improved growth traits that were consistent with the growth characteristics of the LRK1overexpressing transgenic rice lines. Thus, a qRT-PCR analysis was used to investigate the expression of LRK genes in LAIRassociated transgenic rice lines. In the LAIR-overexpressing 35S:: LAIR lines, the LRK genes were differentially expressed. The LRK1 and LRK4 transcript levels were high, while the expression levels for the other LRK genes remained unchanged or decreased slightly (Fig. 3a). Furthermore, LRK genes expression in the LAIR mutant lines with the 35S::LAIR-MU1 or 35S::LAIR-MU2 constructs was similar to that in the 35S::LAIR lines (Fig. 3b). Accordingly, LAIR may have a key role influencing the transcriptional regulation of LRK genes. All LRK genes were expressed at relatively low levels in the 35S::anti-LAIR lines, suggesting that LAIR was crucial for the expression of the LRK gene cluster (Fig. 3d).
To clarify the regulatory effects of LAIR on the expression of LRK genes, rice protoplasts were transformed with sense or antisense LAIR expression vectors along with the vector harboring the LRK promoter-driven luciferase reporter gene. The LUC/REN signal ratios suggested that wild-type LAIR (LAIR) could significantly promote the promoter activations of LRK1 and LRK4. Simultaneously, mutant LAIR (LAIR-MU1) showed similar activity ability on LRKs (Fig. 3c). On the contrary, antisense LAIR (anti-LAIR) suppressed the promoter activations of all the LRKs (Fig. 3e). The different regulatory effects between sense LAIR and antisense LAIR may be caused by the fact that sense LAIR transformation only contained one isoform (JX512726) of alternative spliced LAIR, and antisense LAIR could  interfere all isoforms. Together, our results implied that LAIR was capable of activating the LRK promoter in vivo, with different regulatory effects among different LRK loci.
LAIR overexpression changes histone modification marks. Recent advances revealed the broad functional repertoire of lncRNAs, including their effects on chromatin modifications and epigenetic changes to specific genomic loci that regulate the expression of neighbouring protein-coding genes 30 . To assess whether altered LAIR expression results in changes of histone modifications, we completed a chromatin immunoprecipitation (ChIP) assay to investigate the modification status of LRK loci. We analyzed the histone modifications associated with transcriptional activation, including trimethylated histone H3 lysine 4 (H3K4me3), acetylated H4K5 (H4K5ac), H3K9ac, H4K16ac, and H3K27ac [31][32][33][34] . We mainly focused on LRK1 because its transcription was increased in the 35S::LAIR lines, while LRK3 was used as a control because its transcription was unaffected by LAIR expression (Fig. 3a-c).
Our ChIP-qRT-PCR results indicated that H3K4me3 and H4K16ac were unaffected at the LRK3 locus, but were obviously enriched at the promoter, 3′ untranslated region, and LRK1 gene body in the LAIR-overexpressing line relative to the levels in the wild-type plants (Fig. 4b). These observations suggested that the overexpression of LAIR increased the H3K4me3 and H4K16ac levels at the LRK1 chromatin region, which was consistent with the upregulated transcription of LRK1 in the 35S::LAIR lines. In contrast, when enrichments of H3K4me3 and H4K16ac were analyzed in the 35S::anti-LAIR lines, reduction were observed both at the LRK1 and LRK3 locus (Fig. 4c), which were consistent with the suppression of anti-LAIR to LRKs promoter (Fig. 3d, e). Additionally, analysis of histone modifications on LRK4 which was also activated by LAIR (Fig. 3a-c) was completed. The H3K4me3 and H4K16ac levels were increased in the 35S::LAIR lines, and reduced in the 35S::anti-LAIR lines ( Supplementary  Fig. 5). The results could exclude the possibility that the changes of H3K4me3 and H4K16ac levels at LRK1 locus might be caused by the transcription of LAIR itself. Taken together, the results suggested that LAIR overexpression resulted in altered expression of specific LRK genes and correlated with the changes in histone modifications.
LAIR binds OsMOF and OsWDR5 that target LRK1 genomic region. A considerable proportion of lncRNAs was recently confirmed to associate with chromatin-modifying complexes and modulate gene transcription 11 . Because of the histone modification changes of H3K4me3 and H4K16ac in our ChIP assay (Fig. 4b, c), we tested the interaction between LAIR and chromatin-modifying complex proteins. The MOF enzyme is the primary histone H4K16 acetyltransferase in mammalian cells 35 , and is reportedly a component of MSL and the nonspecific lethal (NSL) complex 36 Fig. 2). Coding potential scores were generated using the CPC program. Transcripts with scores beyond −1 and 1 are marked as non-coding or coding in this CPC classification 28  promoter regions of NSL-targeted genes 38 . Additionally, MLE is a critical protein that associates with ncRNAs to form the MSL complex 39,40 . An NSL component, WDR5, has recently been reported to interact with hundreds of lncRNAs 33,41 .
Homologs of human MOF (NM_032188), D. melanogaster MLE (NM_057293), and human WDR5 (NM_017588) in rice (OsMOF XM_015789227, OsMLE XM_015756470, and OsWDR5 XM_015772971, respectively) were cloned and fused in-frame with the FLAG tag sequence. Rice protoplasts were co-infiltrated with these genes and the 35S::LAIR vector or 35S::anti-LAIR for a RNA immunoprecipitation (RIP) assay. Our RIP-qRT-PCR results revealed the significant binding of OsMOF and OsWDR5, but not OsMLE, to the 5′ and 3′ regions of LAIR (Fig. 4d). These results were further confirmed by RIP assay using the OsMOF-FLAG and OsWDR5-FLAG transgenic rice lines ( Supplementary  Fig. 4). Thus, LAIR appears to be able to interact with two subunits of the chromatin-modifying complex in rice cells.
Furthermore, to investigate the binding ability of OsMOF and OsWDR5 to LRKs genomic loci, we performed ChIP assay in the OsMOF-FLAG and OsWDR5-FLAG transgenic rice lines. A notable enrichment was detected in the LRK1 region, while the binding levels at the LRK3 locus were similar to those of the controls (Fig. 4e). Our findings confirmed that OsMOF and OsWDR5 can target LRK1 genomic region.
LAIR binds to LRK genomic loci. To investigate the association between genomic DNA region of LRKs and LAIR, we performed chromatin isolation by RNA purification (ChIRP) assay. Successful enrichment of LAIR RNA using biotinylated LAIRantisense DNA (LAIR-asDNA) probes was confirmed by LAIR 5′-primer qRT-PCR. The result showed that LAIR-asDNA probes retrieved~65% of LAIR RNA and undetectable OsActin (Fig. 4f,  left panel). Identification of DNA fraction extracted from ChIRP samples showed that LAIR significantly occupied at the 5′ and 3′ untranslated regions of LRK1 compared with LRK3 (Fig. 4f, right  panel). These data suggested that LAIR RNA could physically locate to LRKs gene genomic DNA region, with different effects among the LRK loci.

Discussion
There has recently been considerable interest in the noncoding transcripts in diverse eukaryotes. Accordingly, there has been a rapid increase in the number of well-studied lncRNAs in mammals, while related plant studies have lagged behind. In this study, a lncRNA, LAIR, was identified in rice. Most of the tens of thousands of ncRNA sequences encoded in plant genomes are transcribed by RNAPII, and the majority of the identified plant lncRNAs are polyadenylated 42 . Similarly, LAIR contains a 5′ cap structure and a polyadenylated 3′ end, implying RNAPII is responsible for the transcription of LAIR as well as the nearby protein-coding genes. Moreover, it is likely that most of the uncharacterized lncRNAs help regulate the expression of neighbouring genes 43 . The LAIR sequence is encoded in the antisense strand of the 5′ terminal region of the LRK gene cluster. Additionally, LRK1 is located in the first intron of LAIR, with no overlap between the two antisense transcripts (Fig. 1a). Thus, LAIR functions may not involve a direct base-paired interaction with the transcript of the protein-coding LRK genes. Furthermore, we discovered that LAIR overexpression can result in highly upregulated expression of several LRK genes as well as an obvious increase in rice grain yield. Meanwhile, the increased transcription of antisense LAIR downregulates the expression of all LRK genes and inhibits plant growth (Figs. 1b-e and 3a, d). Considering exogenous LAIR was randomly incorporated into the rice genome and not near the LRK gene cluster, we concluded that LAIR is a crucial trans-acting regulator of LRK transcription. We also detected an intriguing gene expression pattern in LAIR-overexpressing rice lines. Specifically, there were differences in the expression level changes among the LRK genes (Fig. 3a). These differences were also observed in the luciferase reporter assays (Fig. 3c). Therefore, LAIR regulates LRKs expression with diverse effects among genes.
The mechanisms underlying the lncRNA-associated regulation of gene expression are very diverse. Recent reports described a crucial role for lncRNAs involving epigenetic modifications to specific genomic loci, which then regulate the expression of neighbouring protein-coding genes 6,30,44 . Given that LAIR activates the expression of LRK genes, we analyzed the histone     OsActin was used as a reference control and for data normalization. d RIP assays of epigenetic modificationassociated proteins (OsMOF, OsMLE, and OsWDR5) confirmed the interaction with LAIR. The proteins were fused to the FLAG tag. CK: empty vector control with only the FLAG tag. The RIP was completed in rice protoplasts that were co-infiltrated with fusion proteins vector or the control, and the vector with the 35S::LAIR or 35S::anti-LAIR. e ChIP assays of modification-associated proteins on LRK loci. The analyses were completed for LRK1 and LRK3 using an anti-FLAG antibody in rice wild-type (WT) as well as OsWDR5-FLAG (OsWDR5-1 and OsWDR5-2) and OsMOF-FLAG (OsMOF-1 and OsMOF-2) transgenic lines. OsActin was used as a reference control and for data normalization. f ChIRP assays of LAIR on LRK loci. ChIRP enriched~65% of LAIR RNA in both odd and even probes pool, with undetectable OsActin (Actin) (left panel). Immunoprecipitated DNA genomic region analyzed by ChIRP-qPCR showed the remarkable interaction of LAIR with the 5′ and 3′ untranslated regions of LRK1 compare with LRK3 (right panel). Probes that targeted LacZ mRNA were served as negative controls. b-f Student's t-test: *P < 0.05. Data are presented as the mean ± standard deviation (n = 3). g A proposed model of LAIR during the regulation of LRKs. LAIR interacts with epigenetic regulation proteins OsMOF and OsWDR5; both lncRNA and epigenetic modification proteins locate to LRK1 genomic region; and LAIR overexpression results in changes in histone marks (H3K4me3 and H4K16ac) and activating transcription of LRK1. The process is not executed on LRK3 genomic region. This difference may contribute to the observed differential regulation of LRK genes modification status of LRK loci. We observed that H3K4me3 and H4K16ac were enriched at LRK1 in LAIR overexpression lines, implying that epigenetic changes to specific LRK loci were associated with altered LAIR expression (Fig. 4b). Hundreds of lncRNAs have been confirmed to interact with multiple protein partners, suggesting lncRNAs influence gene expression by targeting chromatin remodelers to specific genomic regions as part of a molecular scaffold 6,11 . RNA-binding proteins OsMOF and OsWDR5, which have shown to associate with the H3K4me3 and H4K16ac histone modification complex, were detected significantly binding with LAIR ( Fig. 4d and Supplementary Fig. 4). Additionally, OsMOF and OsWDR5 were enriched at the LRK1 locus (Fig. 4e). Finally, we found evidence on LAIR RNA binding with LRK1 genomic locus (Fig. 4f). Thus, LAIR could interact with epigenetic regulation proteins OsMOF and OsWDR5; both lncRNA and epigenetic modification proteins could locate to specific LRK genes; and LAIR overexpression resulted in altered LRK genes expression and changes in histone marks (Fig. 4g). However, current evidence do not support that LAIR acts through LRKs to regulate grain yield and it is not clear whether the changes in histone marks are the cause of LRKs expression or occurring as a consequence. While the lncRNA activities continue to be clarified, the mechanisms regulating plant lncRNAs remain relatively uncharacterized. As a key histone acetyltransferase in mammalian cells, MOF has been well studied in human and fruit fly, but the research of MOF in rice is almost zero. There are currently no published reports describing lncRNA-regulated gene expression associated with the histone modification protein OsMOF in rice. The NSL complex is a universal and sex-independent major regulator of housekeeping genes 37,45,46 . Because two of its crucial components (OsMOF and OsWDR5) interact with LAIR, it is likely that LAIR functions are associated with the evolutionarily conserved NSL complex. However, additional research is required to elucidate this potential relationship.
The biological features of lncRNAs have attracted widespread interest, and important advances have been made in nearly all cellular systems or species that have been studied at the genome level.
Regarding applied research, current efforts are mostly related to the potential utility of lncRNAs in cancer therapeutics, regenerative medicine involving stem cell technology, and biomarker development for molecular interventions 42,[47][48][49][50] . However, studies on the application of lncRNAs in plants, especially for crop breeding, remain limited. To satisfy future food and bio-fuel demands, innovative molecular methods to improve crop yields must be developed. The rice LAIR described herein may be useful for enhancing plant growth and grain yield. Relative to the wild-type plants, the LAIR-overexpressing transgenic rice lines produced larger primary panicles and had more panicles per plant, which likely contributed to the considerable increase in total grain yield ( Fig. 1b-e). As a new genetic element identified at a QTL, LAIR may represent a new option for introducing stable and heritable increases of grain yield in rice. Rice is considered a model cereal crop because of its well-established genetic background and molecular breeding system. Thus, it may be relevant for characterizing lncRNAs in terms of the associated molecular mechanisms and individual gene expression levels, as well as for the application of lncRNAs in crop breeding. Future studies should aim to characterize the molecular mechanism underlying LAIR-regulated gene expression with diverse effects, and explore the potential application of LAIR for improving crop yield.

Methods
Plant materials and gene transformation. O. sativa ssp. indica rice cultivar MingHui63 (MH63) was used in this study. Additionally, the binary vector pCAMBIA1304 (Center for the Application of Molecular Biology to International Agriculture), which carries the kanamycin-and hygromycin-resistance genes for screening transformed bacteria and plants, respectively, was assembled. The LAIR isoform of JX512726 was cloned and used to construct 35S:LAIR transgenic line. To induce rice embryogenic calli, dehusked MH63 rice seeds were surface-sterilized by soaking in 70% (v/v) ethanol for 5-10 min and then 0.1% (w/v) HgCl 2 for 20-30 min. Seeds were then washed 5-7 times with sterile distilled water and sown on agar-solidified Murashige and Skoog medium. About 20 days later, the calli were bombarded using a gene gun and selected after a 2-week exposure to 50 mg l −1 hygromycin. The hygromycin-resistant calli were regenerated and grown (T 0 lines). A total of 10-20 independent T 0 lines were tested to confirm they were successfully transformed. The hygromycin gene was amplified by a polymerase chain reaction (PCR) with genomic DNA used as the template.
All rice materials were grown in a greenhouse at 28°C under a 14 h-light/10 hdark photoperiod or in test fields in Shanghai (31°11′N, 121°29′E)  Rice protoplasts generation and gene expression analysis. A 1.5-kb sequence upstream of the ATG start codon (−1500 to −1), which corresponded to the LRK promoter region, was amplified with the LRKn promoter primers (Supplementary Data 1). The amplified fragment was cloned into the pGreenII 0800-LUC 51 dual luciferase (LUC) assay system transient expression vector so that it was followed by the LUC reporter gene. Leaves from rice plants at the three-leaf stage were treated with cellulase and macerozyme to generate protoplasts, which were transiently transformed with the prepared vectors using a 40% polyethylene glycol solution 52 . After a 12-16 h incubation at 28°C in darkness, the protoplast lysate was collected and tested using the Dual-Luciferase ® Reporter Assay System (Promega, Madison, WI, USA) and the Synergy™ 2 Multi-Mode Reader (BioTek, Winooski, VT, USA). The expression levels of the LUC reporter gene under the control of the target promoter and the Renilla luciferase (REN) reporter gene under the control of the CaMV 35S promoter were analyzed. The LUC/REN signal ratio can be used to determine the relative ability of ncRNAs to induce the expression of genes under the control of different promoter sequences.
Chromatin immunoprecipitation assay. Two-week-old rice seedlings were harvested and immersed in a fixing buffer [0.4 M sucrose, 10 mM Tris-HCl, pH 8.0, 1 mM EDTA, pH 8.0, 1% formaldehyde, and a protease inhibitor cocktail tablet (Roche, Indianapolis, IN, USA)] under vacuum conditions for 15-20 min. The seedlings were incubated for an additional 5 min after 0.1 M glycine was added. Seedlings were rinsed 5 times with distilled water and then frozen with liquid nitrogen. Fixed seedlings were ground to a powder and resuspended in 30 ml lysis buffer (50 mM HEPES, pH 7.5, 1 mM EDTA, pH 8.0, 150 mM NaCl, 1% Triton X-100, 10% glycerol, 5 mM β-mercaptoethanol, and a protease inhibitor cocktail tablet). Solutions were filtered through a 100-µm nylon mesh and centrifuged at 1500 × g for 20 min. The pelleted chromatin was resuspended and sheared by sonication to produce approximately 200-bp fragments. After a centrifugation at 13,000 × g for 10 min, the supernatants were transferred to new tubes and the appropriate antibodies were added [2 μl  Beads were washed with low salt, high salt, LiCl, and TE buffers (1 ml each). Immunocomplexes were eluted from the beads with 500 μl elution buffer (1% SDS and 0.1 M NaHCO 3 , with 20 μl 5 M NaCl in each tube). Additionally, crosslinks were reversed by incubating samples at 65°C for at least 5-6 h. Residual proteins were degraded by the addition of a solution consisting of 10 mM EDTA, pH 8.0, 40 mM Tris-HCl, pH 8.0, and 2 μl Proteinase K followed by an incubation at 45°C for 1 h. Moreover, DNA was purified and resuspended in 100 μl sterile distilled water.
A qRT-PCR was conducted to assess the enrichment of immunoprecipitated DNA during the ChIP experiments. Details regarding the gene-specific primers are listed in Supplementary Data 1. OsActin was used as an internal standard for normalizing data.
RNA immunoprecipitation assay. The full-length coding sequences for genes associated with epigenetic modifications (OsMOF, OsMLE, and OsWDR5) were cloned from rice MH63 cDNA and fused in-frame with the gene encoding the FLAG tag under the control of the 35S CaMV promoter in the pCAMBIA1306 vector. Rice protoplasts were co-infiltrated with this vector or the control (i.e., empty vector encoding the FLAG tag) and the vector with the 35S::LAIR or 35S:: anti-LAIR vector. After a 12-16-h incubation at 28°C in darkness, protoplasts were harvested for RIP assays.
The RIP assays were conducted using the EZ-Magna RIP™ Kit (Millipore, Bedford, MA, USA). Monoclonal anti-FLAG M2 (Cat. No. F1804, Sigma-Aldrich, St. Louis, MO, USA) was added (5 μl) to each reaction. Purified RNA was used as the template to prepare first-strand cDNA with the PrimeScript™ RT Reagent Kit. A qRT-PCR was completed using gene-specific primers (Supplementary Data 1) to assess the enrichment of the immunoprecipitated RNA, with OsActin serving as the reference control. The RIP efficiency of the protoplasts transformed with the control vector was used to normalize data.
Chromatin isolation by RNA purification. Rice protoplasts were generated from MH63 and harvested for ChIRP assays. The ChIRP assays were conducted using the EZ-Magna ChIRP™ RNA Interactome Kit (Millipore, Bedford, MA, USA). Sonication was performed in ice-cold water bath at setting with high-power, 30 s ON, 30 s OFF pulse intervals for 2 h (total process time). Anti-sense biotinylated DNA tiling probes for selective retrieval of RNA target were designed by online probe design program (www.singlemoleculefish.com), and details are provided in Supplementary Table 1. For LAIR, probes were separated into two groups (even and odd), which served as internal control for each other. Probes that targeted LacZ mRNA, which absent from rice cell, were used as negative control probes 14 .
After ChIRP and RNA/DNA isolation, qRT-PCR was completed using genespecific primers (Supplementary Data 1) to assess the enrichment of the retrieved RNA and immunoprecipitated DNA, with OsActin serving as a negative control gene.