Oryza sativa LIGULELESS 2s determine lamina joint positioning and differentiation by inhibiting auxin signaling

The lamina joint (LJ) is a grass-specific organ that connects the leaf blade to the sheath and causes the leaf blade to bend away from the vertical axis. The angle between the leaf blade and the vertical axis is termed the leaf angle. The leaf angle determined by the LJ in cereal crops affects light capture, particularly under conditions of high-density planting, and therefore is an important trait for high yields (Sakamoto et al., 2006; Liu et al., 2019; Tian et al., 2019). The LJ is a highly plastic organ that determines leaf angle dynamics and optimizes light capture in response to developmental stages, internal cues, and the changing environments (Asahina et al., 2014; Sun et al., 2015; Qiao et al., 2017; Zhou et al., 2017; Ruan et al., 2018). The structure of LJ varies among different grass species. In rice (Oryza sativa) and wheat (Triticum aestivum), the LJ consists of the collar (mechanical hinge), the ligule (a thin and tongue-like white membrane), and a pair of auricles (hairy and sickle-shaped tissue) (Hoshikawa, 1989; Liu et al., 2019); by contrast, the maize (Zea mays) LJ consists of the ligule, the midrib, and a pair of auricles (thickened tissues that connect blade and sheath) (Kong et al., 2017). The organogenesis of LJ has been mainly studied in maize using the liguleless mutants (Becraft et al., 1990; Sylvester et al., 1990; Harper & Freeling, 1996; Moreno et al., 1997; Walsh et al., 1997). The LIGULELESS 1 (LG1) gene, which encodes a nuclear-localized protein that contains a SBP (SQUAMOSA promoter-binding protein) domain (Moreno et al., 1997), was first identified in maize through analysis of the recessive mutant lg1-R. This mutant shows LJ defects, and the ligule and auricles are absent from about the first 10 leaves, but ligule vestiges without auricles are produced on the upper five to 10 leaves (Becraft et al., 1990; Sylvester et al., 1990). The LIGULELESS 2 (LG2) gene, which encodes a basic leucine zipper (bZIP) protein (Walsh et al., 1997), has only been identified in maize. The recessive lg2-R mutant shows age-dependent LJ phenotypes: the ligule and auricle of the first leaf are absent, the asymmetric auricles and associated ligules are observed on the margins of the fourth leaf, and subsequent leaves generate increasingly more ligules and auricles until the phenotype is similar to that of wild type (Harper & Freeling, 1996; Walsh et al., 1997). The lg1-R;lg2-R double mutant displays a novel phenotype, with the absence of ligules on all leaves (Harper & Freeling, 1996), suggesting that LG1 and LG2 might function redundantly in the same biological process in maize. By contrast, mutations in OsLG1 and TaSPL8, the orthologs of LG1 in rice and wheat, respectively, result in the absence of ligules, auricles and collar on all leaves (Lee et al., 2007; Liu et al., 2019), whereas the function of LG2 has not been reported in other grass species except maize. Together, the mechanism that underlies LJ organogenesis remains to be elucidated. In this study, we first identified the functions of LIGULELESS 2 genes in rice and then investigated the mechanism how LG2 and LG1 regulate LJ organogenesis in rice. To define the cytological processes that underlie LJ organogenesis in rice, we monitored the second leaf that is the first complete leaf along a developmental trajectory after seed germination (Fig. 1a; Supporting Information Fig. S1). The morphological and cytological features of the LJ suggested that the boundary between the blade and pre-sheath formed at 3 days after germination (DAG), indicating that the proximal–distal axis of the leaf has been determined before this point; the location of LJ was observed at 4 DAG, which is indicated by the first round of anticlinal and periclinal divisions of the adaxial epidermal cells in the preligule region; then the LJ differentiated to form the collar, ligule and auricles after 4 DAG (Fig. 1a). This suggests that two successive processes are involved in LJ organogenesis, including the determination of LJ position and the differentiation of the collar, ligule, and auricles. Moreover, the determination of LJ position might contribute to form a clear boundary between blade and sheath. Based on the amino acid sequence of ZmLG2 (GRMZM2G060216 in maize), we identified Os01g0859500 and Os05g0443900 encoding the closest orthologs to ZmLG2 in rice, and named them OsLG2 and OsLG2L, respectively (Fig. S2). β-Glucuronidase (GUS) staining assays showed that both OsLG2 and OsLG2L highly expressed in the LJs (Fig. S3), indicating their potential function in LJ. Using CRISPR/CAS9 technology, we obtained the knock-out lines OsLG2-cri OsLG2L-cri and OsLG2-cri;OsLG2L-cri, as CAS9-free homozygous mutations (Fig. S4). The LJ morphology in OsLG2-cri was similar to that in wild-type Nipponbare (Ni) (Figs 1b, S5), whereas in OsLG2L-cri, LJ organogenesis was defective in early leaves, and the phenotype is gradually rescued to wild type in later-initiated leaves (Figs 1b, S5). This indicated that OsLG2 and OsLG2L might function redundantly in LJ organogenesis in rice. Consistent with this, the collar, ligules, and auricles, were absent on all leaves of the OsLG2-cri;OsLG2L-cri line (Figs 1b, S5). Notably, in OsLG2-cri;OsLG2L-cri, the sheath irregularly protruded into the blade, resulting in a perturbed and disordered boundary between blade and sheath (Figs 1b, S5), indicating that the position of LJ was indeterminate. Meanwhile, we used the amino acid sequence of ZmLG1 (GRMZM2G036297 in maize; Moreno et al., 1997) as a query, and identified that OsLG1 (Os04g0656500) was the closest ortholog to ZmLG1 in rice, which has been previously reported (Lee et al., 2007). Using CRISPR/CAS9 technology, we generated the knock-out line OsLG1-cri as CAS9-free homozygous mutations (Fig. S4), and found that although no collar, ligule or auricles were presented on any leaf in OsLG1-cri, the sheath connected the blade in a line, and a smooth boundary between blade and sheath was formed, which was obviously different from that in OsLG2-cri;OsLG2L-cri, and similar to that in wild-type Ni (Figs 1b, S5). To further investigate the function of OsLG2/2L and OsLG1 in LJ positioning and differentiation at the cytological level, we first performed longitudinal sections of the LJ regions in the second leaves of these CRISPR lines. We found that the ligules were completely absent in LJ regions of OsLG2-cri;OsLG2L-cri and OsLG1-cri lines (Figs 1c, S1). Then we used scanning electron microscopy to observe the epidermis at the adaxial and the abaxial sides of LJ regions. Because the epidermis of the blade cannot be distinguished from that of the sheath on the abaxial side (Fig. 1d), we turned to the phenotype on another side. On the adaxial side, the epidermis of the sheath, consisting of rectangular cells with stomata, was easily distinguishable from the blade epidermis, which was covered by wax crystals, papillae, stomata, and hairs (Fig. 1d). Notably, the epidermis of the collars in Ni, OsLG2-cri, and OsLG2L-cri consisted of wrinkled cells, which can be distinguished from the epidermis of blade and sheath; by contrast, in the OsLG2-cri;OsLG2L-cri and OsLG1-cri lines, no collar-specific cells were presented in the LJ region, but a boundary between blade epidermis cells and sheath epidermis cells can be observed. Importantly, the boundary in OsLG2-cri;OsLG2L-cri was distinct from that in OsLG1-cri: in OsLG2-cri;OsLG2L-cri, the sheath cells protruded into the blade to form a discrete and aberrant boundary; by contrast, in OsLG1-cri, blade cells connected sheath cells in a line, which leads to a smooth boundary between blade and sheath. Therefore, the morphological and cytological results demonstrate that OsLG2 and OsLG2L determine LJ position, and that OsLG1 initiates LJ differentiation. Notably, the regulation of LJ differentiation by OsLG1 is dependent on the LJ position determination by OsLG2 and OsLG2L, because the LJ differentiation was absent in OsLG2-cri;OsLG2L-cri. To further investigate the functional relationship between OsLG2/2L and OsLG1 during LJ organogenesis, we analyzed the spatiotemporal expression pattern of these genes using promoter-driven GUS reporter lines and quantitative reverse transcription polymerase chain reaction (qRT-PCR) (Fig. 1e,f). Although the boundary between the blade and pre-sheath formed at 3 DAG (Fig. 1a), the expression levels of OsLG2, OsLG2L and OsLG1 were lower in the LJ region than that at 4 DAG (Fig. 1e,f). The expression of OsLG2 and OsLG2L was higher than OsLG1 at 3 DAG and 4 DAG, and the expression of OsLG1 was significantly increased from 5 DAG (Fig. 1e,f). Notably, OsLG1 expression was slightly higher at 4 DAG than that at 3 DAG (Fig. 1e,f), which might affect ligule initiation after determination of the LJ position by OsLG2/2L (Fig. 1a). Therefore, together with the phenotypes of OsLG2/2L and OsLG1 knock-out lines, the spatiotemporal expression patterns indicated that OsLG2 and OsLG2L might function earlier than OsLG1 in these two consecutive processes. Furthermore, we generated the homozygous triple mutant OsLG2-cri;OsLG2L-cri;OsLG1-cri (Fig. S4) and found that the triple mutant displayed indeterminate LJ positioning that was similar to that in OsLG2-cri;OsLG2L-cri (Figs 1g, S5). This suggests that determination of the LJ position by OsLG2/2L is independent of OsLG1. However, the expression level of OsLG1 in OsLG2-cri;OsLG2L-cri was largely inhibited in LJs at 3 and 4 DAGs (Fig. 1h) and 6 DAG (Fig. S6a), which is consistent with the absent differentiation of LJ in OsLG2-cri;OsLG2L-cri (Fig. 1b,c) and suggests that OsLG1-induced LJ differentiation may depend on OsLG2/2L-determined LJ position. In addition, we also found the slightly decreased expression levels of OsLG2/OsLG2L in LJ of OsLG1-cri (Fig. S6b,c), which might be feedback regulation. The LJ organogenesis can determine the leaf angle in cereal crops (Moreno et al., 1997; Walsh et al., 1997; Lee et al., 2007; Zhou et al., 2017; Liu et al., 2019). We also found that the CRISPR knock out lines, OsLG1-cri, OsLG2-cri, OsLG2L-cri and OsLG2-cri;OsLG2L-cri, showed the reduced leaf angle (Fig. S7a), which was accompanied with the nonexpanded cells in the LJ adaxial side (Fig. 1c). In addition, to further uncover the function of OsLG1 and OsLG2/2L in LJ organogenesis, we generated knockdown lines using RNA interference (RNAi), and found that OsLG1-RNAi and OsLG2/2L-RNAi showed the intact LJs (Fig. S8a–d), but the reduced leaf angle (Fig. S8a,e), which results from the decreased length in the LJ adaxial side (Fig. S8f). To further detect whether the expression levels of these genes is involved in the leaf angle formation, we analyzed the leaf angles in several independent RNAi lines and overexpression lines with these genes (Fig. S7b,d), and found that leaf angle was decreased in RNAi lines and increased in overexpression lines compared to that in wild-type Ni, respectively (Fig. S7c,e). These results indicate that OsLG1 and OsLG2/2L play roles in LJ organogenesis in a dose-dependent manner, which can affect the leaf angle formation. To investigate the mechanism by which OsLG2/2L and OsLG1 determine LJ position and initiate LJ differentiation, we quantified the transcriptional profiles in the LJ regions of liguleless-related plant materials at 3 DAG and 4 DAG, including OsLG1-cri, OsLG2-cri;OsLG2L-cri and wild-type Ni (Fig. S9). We identified 3595 differentially expressed genes (DEGs) from the comparison between OsLG1-cri and Ni, including 1388 upregulated and 2207 downregulated genes (Table S1); and 1095 DEGs from the comparison between OsLG2-cri;OsLG2L-cri and Ni, including 798 upregulated and 297 downregulated genes (Table S1). Notably, 62.1% (680) of LG2/2L-regulated genes were co-regulated by OsLG1 (Fig. 2a), consistent with the conclusion that the initiation of LJ differentiation by OsLG1 is dependent on the determination of LJ position by OsLG2/2L. We then performed Mapman functional enrichment analysis of the DEGs in OsLG1-cri and OsLG2-cri;OsLG2L-cri (Fig. 2b). Some functional clusters related to DNA synthesis and cell division were mainly enriched in the downregulated gene clusters in OsLG1-cri (Fig. 2b,c; Table S2), which is consistent with the active cell division involved in OsLG1-dependent LJ differentiation (Fig. 1a). Auxin signaling and spatiotemporal distribution play key roles in organogenesis and patterning processes during plant development (Smet et al., 2010). The core auxin signaling pathway includes three components: Transport Inhibitor Resistant1/Auxin signaling F-Box proteins (TIR1/AFB) auxin receptors, AUXIN (Aux)/INDOLE-3-ACETIC ACID (IAA) transcriptional repressors, and AUXIN-RESPONSIVE FACTOR (ARF) transcription factors. Auxin promotes the interaction between TIR1/AFB and Aux/IAA proteins, which results in the degradation of Aux/IAAs and the release of ARFs from repression, and triggers auxin responses (Lavy & Estelle, 2016), including expression of the early auxin-responsive SMALL AUXIN UP RNA (SAUR) genes (Leyser, 2018). In addition, feedback control can regulate auxin signal output and is important to prevent excessive auxin response, which includes upregulation of Aux/IAA and GRETCHEN HAGEN 3 (GH3) genes (key genes involved in auxin metabolism) (Fig. S10) (Benjamins & Scheres, 2008). Notably, various pathways for controlling auxin signal output were enriched by OsLG1 and OsLG2/2L (Fig. 2b). First, the gene clusters of SAURs were enriched by the upregulated genes in the OsLG1-cri and OsLG2-cri;OsLG2L-cri (Fig. 2d,e; Table S2), indicating that auxin signal output might be enhanced in LJs of liguleless mutants compared to Ni. Second, the Aux/IAA family was enriched by the upregulated genes in the OsLG2-cri;OsLG2L-cri lines, and was also upregulated in the OsLG1-cri lines (Fig. 2f; Table S2), which indicates the potential feedback regulation of enhanced auxin signal output, and further suggests that auxin signaling may be enhanced in LJs of the liguleless mutants. Third, the ARF gene cluster was specifically enriched among the downregulated genes in OsLG1-cri (Fig. 2g; Table S2), which is consistent with the previous report that TaSPL8, the ortholog of LG1 in wheat, directly induces the expression of TaARF6 (Liu et al., 2019). Because ARFs have been reported to induce GH3 gene expression to decrease the content of free IAA (Zhang et al., 2015), the downregulation of ARF family expression suggests that the content of free IAA in OsLG1-cri might be elevated. In addition, we further confirmed the expression patterns of these auxin related clusters in OsLG1-cri and OsLG2-cri;OsLG2L-cri by qRT-PCR assays (Fig. S11). Taken together, these results suggest that OsLG2/2L and OsLG1 may, through inhibiting auxin signaling in LJs, determine the LJ position and trigger the LJ differentiation. In conclusion, the LJ is a grass-specific organ, but its composition and structure slightly differ among grass species. In rice, OsLG1 is required for LJ development (Lee et al., 2007), but little is known about rice LJ organogenesis. Here, our comprehensive cytological, genetic and transcriptional network analysis identified the function of OsLG2/2L in LJ organogenesis, and revealed the spatiotemporal relationship between OsLG2/2L and OsLG1 during LJ organogenesis and development, as well as suggesting a potential mechanism for their role in LJ organogenesis (Fig. 2h). OsLG2 and OsLG2L are essential for determining the LJ position; OsLG1 is necessary to initiate LJ differentiation, which depends on the LJ positioning determined by OsLG2/2L. Furthermore, the early processes in LJ organogenesis that are regulated by OsLG2/2L and OsLG1 require the decreased auxin signaling output in the LJ (Figs 2d–g, S11). Notably, although OsLG2/2L and OsLG1 are conserved bZIP and SPL transcriptional factors, respectively (Figs S12, S13), phylogenetic analysis of their orthologs from different species in the plant kingdom identified a grass-specific subfamily (Fig. S14a,b), indicating that the grass-specific LG2 and LG1 subfamilies might be responsible for the unique organ LJ in grass. In addition, we identified another close homolog of ZmLG1 and ZmLG2 in maize by the phylogenetic analysis (Figs S2, S14b), which might function redundantly with ZmLG1 and ZmLG2, respectively, in maize LJ organogenesis. Finally, the results demonstrate that OsLG2/2L and OsLG1 show a strong dose-dependent function in LJ organogenesis, which determine the leaf angle formation. Therefore, LG2 and LG1 may be excellent targets for improving the architecture of cereal crops. The rice wild type was the japonica (O. sativa) cultivar Ni. All rice transgenic lines were in the Ni background. The seeds were soaked for 1 d at room temperature, and then imbibed in an incubator at 37°C for 2 d. Seedlings were grown in water in the 96-well plates with the bottoms cut away under 16 h light at 28°C and 8 h dark at 25°C. The leaf angle and LJ length were measured in wild-type Ni and transgenic rice at 9 DAG using ImageJ software. The protein sequence of ZmLG2 was used as a Blast query to identify homologs in rice and maize via Phytozome (https://phytozome.jgi.doe.gov/pz/portal.html) using a cut-off score greater than 316. The protein sequences of ZmLG1, ZmLG2 and their homologs in other plant species were obtained from the Phytozome database. To illustrate the specificity of LG1 to grasses, we included the Aco031754 sequence in Fig. S14(b), and although it had a lower score than 316, it represents the ortholog of ZmLG1 in Ananas comosus, which is the most similar protein to that of the grasses shown in Fig. S14(b). The protein sequences were aligned by ClustalX and GeneDoc software. The phylogenetic neighbor-joining tree was generated using Mega6.0 with a bootstrap setting of 500 (Tamura et al., 2013). For Fig. S14(a), because there were too many homologs of ZmLG2 in plant, we selected the most homologous one except for rice in the phylogenetic trees of LG2. The vector pTCK303 (Wang et al., 2004) was used to clone constructs to generate RNAi transgenic lines. To create OsLG1-RNAi, OsLG2/2L-RNAi plants, the fragments from nucleotides 267–506 and 59–299 within the coding regions of OsLG1 and OsLG2, respectively, were amplified by specific primers, digested with Kpn I and BamH I and then Spe I and Sac I sites, and inserted into pTCK303 driven by the 35S promoter. The vector pCXUN-CAS9 (He et al., 2017) was used to generate CRISPR/CAS9 knock-out mutants. The target sequences were selected from Crispr-p 2.0 (http://crispr.hzau.edu.cn/CRISPR2/). To genotype homozygous mutants generated by CRISPR/CAS9, DNA fragments c. 800-bp long that contained the CRISPR target sequences were amplified and sequenced. The primers are listed in Table S3. The RNAi and pCXUN-CAS9 vectors were transformed into the Agrobacterium strain EHA105, which was used to transfect callus of Ni. For scanning electron microscopy, the second leaf of CRISPR/CAS9 mutants and wild-type Ni at 9 DAG were cut longitudinally and fixed in 3% glutaraldehyde. Samples were dehydrated in an ethanol concentration gradient (25%, 50%, 75%, 95%, 100%) and were then dried and coated with gold. The morphology of the leaf blade, LJ, and sheath of each sample was observed with a scanning electron microscope (JSM-6390LV; Jeol, Tokyo, Japan). Samples were harvested from plants at 3, 4, 5, 6 and 9 DAG and were frozen in liquid nitrogen. Total RNA was extracted using the Trizol method (Invitrogen, Beijing, China). First-strand cDNA was synthesized using the PrimeScript RT reagent kit (TaKaRa). SYBR master mix for PCR (Invitrogen) was used for qRT-PCR. The expression data were normalized against expression of OsUBQ (Os03g0234200). Three biological replicates were performed. The primers for qRT-PCR are listed in Table S3. Tissues at different stages were harvested and immersed in GUS staining solution containing 100 mM sodium phosphate (pH 7.0, Macklin, Shanghai, China), 10 mM Na2EDTA (Macklin), 1 mM K3[Fe(CN)6] (Macklin), 1 mM K4[Fe(CN)6] (Macklin), 0.5% Triton X-100 (Sigma Aldrich, St Louis, MO, USA), 20% methanol (Sinopharm Chemical Reagent Co. Ltd, Shanghai, China), 1 mg ml−1 X-gluc (Goldbio, St Louis, MO, USA), and were stained in the dark at 37°C. Samples were then washed and conserved in 70% ethanol (Sinopharm Chemical Reagent Co. Ltd, Shanghai, China), and imaged using a stereomicroscope (Zeiss Discovery V20, Germany) or camera (Nikon, Japan). Total RNA was extracted by the Trizol method (Invitrogen). At 3 and 4 DAG, LJs from OsLG1-cri, OsLG2-cri;OsLG2L-cri and wild-type Ni were harvested, and equal amounts of RNA from both sampling time points were pooled for each genotype to create a single sample. Two replicates were performed for each sample. Libraries were constructed and sequenced on an Illumina Hiseq 2000 platform to generate 150-bp paired-end reads. Low-quality reads were removed by Trimmomatic-0.36 and the remaining reads were mapped to the rice reference genome IRGSP-1.0 via TopHat v.2.1.1 (Kim et al., 2013). The expression levels of genes were estimated by Cufflinks v.2.2.1 (Trapnell et al., 2012) by calculating the FPKM (fragments per kilobase of transcript per million reads) using the uniquely mapped reads. The DEGs between mutants and wild type were defined to meet the two criteria of an FDR (false discovery rate) < 0.05 and a fold change > 2. Functional enrichment analysis of DEGs was performed by Mapman gene annotation (P-value.adjust < 0.05) (Thimm et al., 2004). The clean sequencing data for this study have been deposited in the European Nucleotide Archive (ENA) at EMBL-EBI under accession no. PRJEB40447 (https://www.ebi.ac.uk/ena/browser/view/PRJEB40447). The authors thank Yun-de Zhao for providing the CRISPR/CAS9 vector. This work was supported by funding from the Ministry of Science and Technology of China (grant 2016YFD0100403), the National Natural Science Foundation of China (grants 31540080, 31671265 and 2015CFB487). SS designed the research. RW performed the experiments. ZC helped to detect the transgenic plants. CL helped to analyze the transcriptomic data. SS and RW analyzed the results and wrote the manuscript. XW gave suggestions for the manuscript. All authors approved the manuscript. Fig. S1 Three-dimensional schematic diagram of the longitudinal sections of rice lamina joints. Fig. S2 A Mega6 neighbor-joining tree inferred from the amino acid sequences of ZmLG2 homologs in rice and maize. Fig. S3 The spatiotemporal expression patterns of OsLG2 and OsLG2L. Fig. S4 Sequencing of the CRISPR/CAS9 knockout lines. Fig. S5 Phenotypes of lamina joints (LJs) in the second leaf of wild-type Nipponbare (Ni), OsLG2-cri-#2, OsLG2L-cri-#2, OsLG2-cri;OsLG2L-cri-#2, OsLG1-cri-#2 and OsLG1-cri;OsLG2-cri;OsLG2L-cri-#2. Fig. S6 The spatial expression pattern of OsLG1 and OsLG2/2L. Fig. S7 The degree of lamina inclination of the second leaves in transgenic rice. Fig. S8 The phenotype of OsLG1-RNAi and OsLG2/2L-RNAi plants. Fig. S9 Schematic diagram for RNA-seq of the lamina joint regions. Fig. S10 Schematic diagram of feedback mechanisms in the auxin signaling pathway. Fig. S11 The expression patterns of auxin related genes by qRT-PCR. Fig. S12 Alignment of the amino-acid sequences of LG2 in grass. The purple lines highlight the functional domains. Fig. S13 Alignment of the amino-acid sequences of LG1 in grass. The purple lines highlight the functional domains. Fig. S14 Phylogenetic analysis of LG2 and LG1 in plant kingdom. Table S1 Differentially expressed genes (DEGs) in lamina joints (LJs) of OsLG1-cri and OsLG2-cri;OsLG2L-cri, compared to wild-type Nipponbare (Ni). Table S2 The FPKM (fragments per kilobase of transcript per million reads) values of genes involved in cell division and auxin-related functional categories. Table S3 Primers used in this study. Please note: Wiley Blackwell are not responsible for the content or functionality of any Supporting Information supplied by the authors. Any queries (other than missing material) should be directed to the New Phytologist Central Office. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.