Research ArticlePLANT SCIENCES

RALF1-FERONIA complex affects splicing dynamics to modulate stress responses and growth in plants

See allHide authors and affiliations

Science Advances  20 May 2020:
Vol. 6, no. 21, eaaz1622
DOI: 10.1126/sciadv.aaz1622


The environmentally responsive signaling pathways that link global transcriptomic changes through alternative splicing (AS) to plant fitness remain unclear. Here, we found that the interaction of the extracellular rapid alkalinization FACTOR 1 (RALF1) peptide with its receptor FERONIA (FER) triggered a rapid and massive RNA AS response by interacting with and phosphorylating glycine-rich RNA binding protein7 (GRP7) to elevate GRP7 nuclear accumulation in Arabidopsis thaliana. FER-dependent GRP7 phosphorylation enhanced its mRNA binding ability and its association with the spliceosome component U1-70K to enable splice site selection, modulating dynamic AS. Genetic reversal of a RALF1-FER–dependent splicing target partly rescued mutants deficient in GRP7. AS of GRP7 itself induced nonsense-mediated decay feedback to the RALF1-FER-GRP7 module, fine-tuning stress responses, and cell growth. The RALF1-FER-GRP7 module provides a paradigm for regulatory mechanisms of RNA splicing in response to external stimuli.


Alternative splicing (AS) involves the production of multiple transcript isoforms from a single gene, which increases the coding capacity and the regulatory potential of the whole genome (1). In higher plants, AS is particularly prevalent during development and in response to environmental changes (2) and has tissue-specific characteristics (3), suggesting that it plays an important role in plant performance. Individual RNA binding proteins can interact with distinct regulatory motifs in RNA molecules to determine which splice sites are to be used by the spliceosome to shape cellular identity and adjust the transcriptome to different conditions (4, 5). However, our understanding of the complete signaling pathways in plants that dynamically regulate AS in response to internal or external signals, thus fine-tuning cellular adaptation, is rather limited. FERONIA (FER) is a receptor-like kinase (RLK) that acts as a receptor for the rapid alkalinization factor (RALF) peptides (e.g., RALF1 and RALF23) (68). FER has recently emerged as a potential target for crop improvement and protection because of its versatile, fundamental, and tissue-specific roles in plant growth, yield control, and multiple stress responses (9). Thus far, our mechanistic understanding of this central RALF-FER pathway remains incomplete. For instance, it is poorly understood which molecular mechanisms FER uses to regulate such diverse cellular activities as cell growth and stress responses via a single FER gene, i.e., the molecular basis of the versatile roles of FER (9). Furthermore, the mechanism how FER regulates cell type– or stimulus-specific signaling output remains unknown (10), i.e., LRE (LORELEI) and LLG1 [LRE like glycosylphosphatidylinositol-anchored proteins (GPI-APs)], which act as chaperones/coreceptors for FER, work together with FER to perform its context-specific roles after LLG1/LER-FER perceives different RALF peptides (8, 11). Here, we found that the RALF1-FER pathway triggers rapid and global changes in AS programs. This process entails phosphorylation of the splicing factor glycine-rich RNA binding protein7 (GRP7), which changes its RNA binding ability and elicits changes in AS patterns. Thus, the RALF1-FER module integrates cell-intrinsic factors with environmental signals to shape the transcriptome.


Binding of the RALF1 peptide to FER triggers inhibition of root growth (6). To determine how RALF1 affects the transcriptome in roots, we performed RNA sequencing (RNA-seq) of the roots of wild-type (WT; Col-0) Arabidopsis plants after 1 hour of RALF1 treatment. Principal component analysis (PCA) was carried out to determine the role of RALF1 in the transcriptome (i.e., in gene expression and AS). The effect of RALF1 on mRNA AS could be observed in PC1, which explained 46.1% of all splicing variance, much higher than the within-group variance (PC2) (fig. S1A). However, the effect of RALF1 on gene expression was minor (fig. S1A) and may have been a secondary effect caused by RALF1 treatment. This hypothesis is consistent with the small number of differentially expressed genes ( logFC > 1; P < 0.05) and the large number of splicing defects detected in the RALF1-treated plant RNA-seq data (table S1). We detected 3401 significantly changed AS events in RALF1-treated plants compared with mock controls (P < 0.05, exon inclusion level change > 0.05, n = 3; see Materials and Methods for details) (Fig. 1A and table S1). Gene ontology (GO) analysis indicated that among these AS events, functions related to DNA repair and cellular amino acid biosynthesis were prevalent, followed by functions related to stress responses [e.g., abscisic acid (ABA) and abiotic stresses] and root cell differentiation (fig. S1B and table S1). This result indicates that many stress-related genes and cellular amino acid biosynthetic genes rapidly respond to RALF1 through pre-mRNA processing. Then, we randomly selected 20 AS events with a wide range of inclusion level changes according to the RNA-seq data for independent validation (see Materials and Methods for details). Sixteen of them were revealed to have significant differences using semiquantitative polymerase chain reaction (semi-qPCR), similar to the findings obtained with the RNA-seq data (fig. S2, A to D). The remaining four AS events that were not confirmed had relatively high P values (i.e., relatively low reliability), as indicated by the RNA-seq data (fig. S2, A to D). Compared with those in mock-treated plants, 23 genes displayed transcript levels that were at least twofold higher in RALF1-treated plants, while 220 genes showed at least twofold lower transcript levels (P < 0.05) (fig. S3A and table S1). The differentially expressed genes ( logFC > 1; P < 0.05) in RALF1-treated plants encoded proteins with diverse functions, such as signal transduction and defense responses, but not the root growth-related pathway (fig. S3B and table S1). Furthermore, we assumed that the observed splicing changes were mediated by the spliceosome, suggesting that a disruption of spliceosome components and key splicing factors would alter the RALF1 response. We used a RALF1-mediated primary root growth inhibition assay as a readout to test this hypothesis (6). We observed the RALF1 responses of mutants in core spliceosome components (lsm8-1) and certain splicing factors (sr45-1 and sua-4) and found that the sr45-1 and lsm8-1 mutants, but not the sua-4 mutant, showed reduced sensitivity to RALF1. The fer-4 mutant, which is deficient in the cognate receptor kinase for RALF1, was less sensitive to lower concentration of RALF1 (1 μM) than WT plants in the root growth inhibition assays (Fig. 1B and fig. S3, C and D) as described in a previous work (6). Notably, fer-4 was still sensitive to higher concentrations of RALF1 (3 to 5 μM; fig. S3, E and F), which has been previously reported (6). This suggested that there might be other RALF1 receptors under different growth conditions. In addition, we also tested whether other RALF peptides are involved in AS regulation. RALF23 (1 μM), but not 1 μM RALF17, induced less root growth inhibition in sr45-1 and lsm8-1 compared to WT, suggesting that RALF23 was also involved in AS regulation (Fig. 1B and fig. S3, C and D). Together, these data indicated that splicing contributes to the responsiveness to RALF1 and RALF23. Because LSM8 has been implicated in the decay of nuclear RNA, it remains possible that RNA degradation also contributes to the effect of RALF1 on root growth in lsm8-1 (12).

Fig. 1 RALF1 affects global RNA splicing, and its receptor FER interacts with the RNA binding protein GRP7.

(A) Different types of AS events are significantly changed in RALF1-treated roots of WT (Col-0) plants relative to mock-treated roots (3401 events in total). (B) RALFs-mediated root growth inhibition. Different genotypes were grown in mock-treated medium and medium with 1 μM RALFs (n = 20 for each group); each bar indicates the means ± SD of three biological replicates. One-way ANOVA with Tukey’s test were used to determinate the statistical difference, **P < 0.01. (C) β-Galactosidase assay in the Y2H system. Negative controls (AD and BD) are shown. Data are shown as the means ± SD of three biological replicates. One-way ANOVA with Tukey’s test, **P < 0.01. (D) GST pull-down assay. The GST protein and His-FER-KD proteins were detected by Western blot (WB) analysis. The GST input protein was visualized with Coomassie brilliant blue (CBB). The GRP7 protein was divided into two parts: the N-terminal region, named GRP71–87, contained an RRM domain, and the C-terminal region, named GRP787–176, contained a GC-rich domain. (E) BiFC assay in Arabidopsis protoplasts. The cell membrane was visualized by FM4-64 staining. Negative controls (GRP6-cCFP+FER-nVenus and GRP7-cCFP+CVY1-nVenus) are also shown. DIC, differential interference contrast. (F) Co-IP assays. The immunoprecipitated FER and coimmunoprecipitated GRP7 were recognized using anti-FLAG and anti-GRP7 antibodies, respectively. The input lanes are indicated, β-actin was used as loading control. Three independent experiments were conducted with similar results to those shown in (D) to (F).

Given that FER acts as a receptor of RALF1 (6), we considered that RALF1 might regulate AS via FER. We assessed several RALF1-triggered AS events in Col-0 and the FER knockout (fer-4) mutant and found that the RALF1-triggered splicing changes (illustrated by the genes AT4G32060, AT1G48175, AT1G54110, AT5G49230, and AT3G27340) were impaired in the fer-4 mutant (fig. S3G). Next, we focused on potential downstream factors of RALF1-FER. Screening of Arabidopsis complementary DNA (cDNA) libraries with the FER kinase domain (FER-KD) as a bait identified a truncated version of GRP7 (60 to 176 amino acids). GRP7 contains an N-terminal RNA recognition motif (RRM) and a C-terminal glycine-rich domain and has been reported to function in RNA processing (13, 14), flowering (15), the immune response (16), and the circadian system (17). We first confirmed that the full-length GRP7 protein interacted with the FER-KD in yeast (Fig. 1C and fig. S4A) (18). We then found that the glycine-rich domain (87 to 176 amino acids, C terminus), but not the RRM domain (1 to 86 amino acids, N terminus), of GRP7 was responsible for the interaction (Fig. 1C and fig. S4A). We further noticed that GRP8, a paralog of GRP7, also interacted with FER. In addition to FER, GRP7 interacts with certain FER gene family members, such as HERK2 (fig. S4, B and C). Consistent with the yeast two-hybrid (Y2H) results, pull-down experiments with recombinant GRP7 variants suggested that GRP7-GST, GRP787–176-GST, and GRP8-GST, but not GRP71–86-GST, interacted with His-FER-KD (18) in vitro (Fig. 1D and fig. S4, D and E). Next, we confirmed the interaction between GRP7 and FER by bimolecular fluorescence complementation (BiFC) in Arabidopsis protoplasts. We confirmed the expression of GRP7–C-terminal cyan fluorescent protein (cCFP) [or GRP6-cCFP (AT1G18630) as a negative control] and FER-nVenus (or CVY1-nVenus as a negative control; CVY1 encodes a member of the CrRLK1L family similar to FER) (fig. S4F). The results showed that GRP7, but not the negative control GRP6, interacted with FER at the plasma membrane (indicated by the styryl dye FM4-64); GRP7 did not interact with CVY1 in the BiFC assay (Fig. 1E). Furthermore, coimmunoprecipitation (Co-IP) assays using Ubi:FER-FLAG transgenic plants and Col-0/Mock control confirmed the interaction between FER and GRP7 in vivo (Fig. 1F). Collectively, these data indicate that FER physically interacts with GRP7.

To investigate whether FER can phosphorylate GRP7, we performed in vitro phosphorylation assays (18) with purified GRP7-GST, His-FER-KD, and His-FERK565R-KD kinase-dead (18) negative control proteins. As shown in Fig. 2A, an antibody against phosphorylated serine residues (anti-pSer) recognized the GRP7 protein when GRP7 was coincubated with His-FER-KD (lane 2) but not when GRP7 was coincubated with the negative control His-FERK565R-KD (lane 3). His-FER-KD was detected because of its autophosphorylation ability (18, 19), and alkaline phosphatase [calf intestinal phosphatase (CIP)] dephosphorylated His-FER-KD and GRP7 (lane 4). Next, we used an ABA-induced in vitro phosphorylation system to identify the GRP7 phosphorylation site(s) (19). Briefly, we coexpressed FER-KD (or FERK565R-KD kinase-dead as a negative control), the ABA receptor pyrabactin resistance like 1 (PYL1), the abscisic acid insensitive 1 (ABI1) protein phosphatases type 2C that dephosphorylates FER and inhibits FER kinase activity (20), and GRP7 in one Escherichia coli strain. ABA triggered FER-KD but not FERK565R-KD kinase-dead phosphorylation levels. Six phosphorylated residues (Tyr111, Ser112, Ser132, Tyr138, Ser139, and Ser140) from GRP7 were identified in the ABA-incubated PYL1/ABI1/FER-KD/GRP7 system and not in the ABA-incubated PYL1/ABI1/FER-KDK565R/GRP7 system (fig. S5, A to F). Tyr111 and Tyr138 were conserved in 13 plant species (fig. S6, A and B). Furthermore, we mutated these identified six phosphorylation sites (fig. S5, A to F) to alanine residues (GRP7mut6A) to constitutively inactivate these phosphorylation sites and tested whether mutated GRP7mut6A-GST was phosphorylated by His-FER-KD in vitro. The results showed that GRP7mut6A-GST was not phosphorylated by His-FER-KD (lane 5; Fig. 2A), suggesting that certain residues may be phosphorylated by FER. To determine whether the phosphorylation of GRP7 by FER is regulated by RALF1, we obtained GRP7-GFP lines expressing GRP7-GFP (green fluorescent protein) under the control of the constitutive Cauliflower Mosaic Virus 35S promoter in the WT background and GRP7-GFP/fer-4 lines (fig. S6, C to F). Immunoprecipitation (IP) of GRP7-GFP using GFP-trap beads showed that the phosphorylation levels of both Ser and Tyr were enhanced after RALF1 treatment in GRP7-GFP plants (lanes 1 and 2; Fig. 2B) but were lower, with or without RALF1 treatment, in GRP7-GFP/fer-4 plants than in GRP7-GFP plants (lanes 3 and 4; Fig. 2B). Furthermore, we analyzed the phosphorylated residues of GRP7 in GRP7-GFP plants with mock treatment or RALF1 treatment using label-free quantitative mass spectrometry (MS). Phosphorylation of Ser132 and Ser139 was significantly higher after RALF1 treatment than after mock treatment (Fig. 2, C and D). We did not detect phosphorylation of the other four residues that were identified in the ABA-incubated PYL1/ABI1/FER-KD/GRP7 system (fig. S5, A to F). This result may have been due to noise and detection limits in MS and/or to some other unknown mechanisms (e.g., reversible phosphorylation).

Fig. 2 FER phosphorylates GRP7 in a RALF1-dependent manner.

(A) FER phosphorylates GRP7. The figure shown here is representative of three independent experiments with similar results. (B) FER phosphorylates GRP7 in response to RALF1 (1 μM, 30 min). The ratio of anti-pSer/anti-GFP or anti-pTyr/anti-GFP is displayed below the blots. The data shown are representatives of three independent experiments. “Control” refers to the WT (Col-0) plants. (C) Examples of phosphopeptides of GRP7. The maximum probability shown for each phosphorylation site was calculated by Proteome Discoverer. (D) The relative abundances of two phosphopeptides detected in GRP7-GFP purified from RALF1-treated and untreated seedlings. The intensity of the indicated phosphopeptide containing the respective phosphorylation site was determined by the label-free MS1 peak area of each phosphopeptide as estimated by MaxQuant. The data represent three biological replicates. The label-free quantitation (LFQ) intensity of the phosphopeptide derived from MaxQuant was normalized against the LFQ intensity of the GRP7-GFP input and then converted to relative abundance by dividing by the mean (n = 3) of the most abundant phosphopeptide (S132) detected, which was set to 100%. Data are presented as the means ± SD, Student’s t test. **P < 0.01. ATP, adenosine triphosphate.

To test whether GRP7 is responsible for AS changes observed upon RALF1 treatment (Fig. 1A), we performed RNA-seq. For this, we used a grp7-1 T-DNA mutant that additionally carries an RNA interference (RNAi) construct against GRP8, designated grp7-1 8i (15). GRP7 negatively regulates its paralog GRP8 (14), and the RNAi construct serves to counteract the up-regulation of GRP8 that occurs in the grp7-1 mutant due to relief of repression (14). Considering that GRP7 is a circadian clock-regulated protein (17), we carried out RNA-seq for Col-0, fer-4, and grp7-1 8i seedling roots after 12 hours of light, when GRP7 shows the highest mRNA expression (17). We identified 3509 significant AS events in the fer-4 mutant compared with the WT control (P < 0.05, exon inclusion level change > 0.05, n = 3) (Fig. 3A and table S1). In addition, we detected 3822 AS events (P < 0.05, inclusion level change > 0.05, n = 3) in the grp7-1 8i mutant compared with the WT control (Fig. 3A and table S1). Notably, all of the AS events described in our work represent the relative abundance of AS events in each group compared with its WT mock. There was a significant overlap between events detected in the two comparisons (P = 3.2 × 10−342, hypergeometric test) (Fig. 3A). It is possible that FER regulates AS also via other RNA binding proteins, e.g., GRP8, and our grp7-1 8i mutant is the only mutant of GRP7 with normal GRP8 expression (15). Thus, the overlap in AS between the fer-4 and grp7-1 8i mutants was approximately 30%. Most of the splicing changes were in the same direction, and the values were highly consistent (r = 0.936, P < 0.0001) (Fig. 3B and table S1). Furthermore, clustering analysis indicated that most splicing changes in fer-4 and grp7-1 8i were correlated, changing in the same direction (Fig. 3C). The genes with splicing defects in both fer-4 and grp7-1 8i plants compared with WT plants encoded proteins associated with distinct biological processes, including stress responses and development (e.g., ABA signaling and root development) (fig. S7, A and B). From the RNA-seq data, we chose 12 AS events detected in both fer-4 and grp7-1 8i plants. We successfully validated 11 of the 12 selected AS events using semi-qPCR (fig. S7, C to E). In parallel, we performed RNA-seq on Col-0, fer-4, grp7-1 8i, and RALF1-treated Col-0 plants. The splicing changes in FER or GRP7 knockout mutants relative to RALF1-treated Col-0 plants were highly correlated (fig. S8, A and B, and table S1), and the splicing changes in common targets (P < 0.05 for both of RALF1-triggered and fer-4 or grp7-1 8i) were also highly consistent (r = 0.797, P = 3.54 × 10−161 for RALF1 versus fer-4 and r = 0.816, P = 1.51 × 10−176 for RALF1 versus grp7-1 8i) (fig. S8, C and D, and table S1), suggesting that RALF1-triggered splicing changes were partially mediated by FER and GRP7.

Fig. 3 GRP7 functions downstream of RALF1-FER to regulate AS and plant fitness.

(A) Venn diagram depicting the overlap of significant splicing changes (P < 0.05, exon inclusion level difference > 0.05, n = 3) between fer-4 and grp7-1 8i (compared with Col-0). P = 3.2 × 10−342; the hypergeometric test was used in the overlap analysis. (B) Density figure showing the comparison between splicing changes upon FER knockout and GRP7 knockout (n = 3). The x and y axes represent changes in inclusion level; the red area indicates high density. The Spearman correlation was used for the correlation analysis. (C) Heat map of changes in differential AS events in fer-4 (n = 3) and grp7-1 8i (n = 3) compared with WT (Col-0). Sig. indicates P < 0.05 for either mutant and splicing difference > 0.05. (D) Statistical data of root length with or without RALF1 treatment (n = 20 roots per condition). Different lines are indicated as “-number.” Data are presented as the means ± SD. (E) Bar graph of the ROS inhibition ratios of different genotypes without or with 1 μM RALF1 treatment (n = 2). The total photon count after flg22 triggering was defined as I, the total photon count after RALF1, and flg22 triggering was defined as J, the ROS inhibition ratio = (I-J)/I × 100%, and the detailed total photon counts are shown in fig. S8E. Bar indicates the means ± SD of two independent experiments. (F) Root lengths of different genotypes treated with 0 or 5 mM ABA. The photograph was taken 4 days after transfer of the seedlings. Scale bars, 1 cm. (G) Statistical analysis of root length in response to ABA. The lengths of 324 roots (n = 18 roots per group) were measured. Data are shown as the means ± SD. All assessments were independently repeated four times with similar results. One-way ANOVA with Tukey’s test, *P < 0.05, **P < 0.01. n.s., not significant.

To elucidate whether GRP7 is involved in RALF1-FER signal transduction, we assayed the inhibition of primary root growth by RALF1 (6). The fer-4 and grp7-1 8i mutants were less sensitive to RALF1 than WT plants, while GRP7-GFP, which overexpressed GRP7, was more sensitive (Fig. 3D). Moreover, the overexpression of GRP7 in the grp7-1 8i mutant complemented its RALF1-tolerant phenotype, indicating that loss of function of GRP7 was responsible for this phenotype (Fig. 3D). In addition, to address whether the phosphorylation sites of GRP7 were involved in the RALF1 response, we mutated the six GRP7 phosphorylation sites to Ala (GRP7mut6A) or Asp (GRP7mut6D) to mimic disrupted or constitutive GRP7 phosphorylation, respectively. GRP7mut6D-GFP/grp7-1 8i plants were more sensitive to RALF1, while GRP7mut6A-GFP/grp7-1 8i plants were less sensitive to RALF1 than GRP7-GFP/grp7-1 8i plant (Fig. 3D and fig. S6E), indicating that FER-mediated phosphorylation of GRP7 is important for the GRP7-mediated RALF1 response. These results suggest that GRP7 may be involved in the RALF1-FER pathway.

Reactive oxygen species (ROS) production triggered by bacterial flagellin22 (flg22) (21) and RALF1 as well as related RALF peptides lead to inhibition of flg22-induced ROS production (7, 19), thus RALF1-mediated, flg22-induced ROS production inhibition is a hallmark of the RALF1 responses. We monitored the ROS production triggered by flg22 and found that fer-4 and grp7-1 8i plants had weaker ROS bursts than Col-0, while GRP7-GFP overexpression lines had stronger ROS bursts (fig. S8E). These data are consistent with previous studies, showing that FER and GRP7 play roles in immune responses (7, 16). When treated with flg22 and RALF1 combined, fer-4 and grp7-1 8i were less sensitive than the WT to RALF1 treatment, as reflected by reduced inhibition of the ROS bursts (52.5% for Col-0, 19.1% for fer-4, and 24.9% for grp7-1 8i) (Fig. 3E and fig. S8E). In contrast, GRP7-GFP#2 mutants were more sensitive to RALF1 than WT plants (66.9% inhibition rate) (Fig. 3E and fig. S8E). When expressed in grp7-1 8i, GRP7-GFP can complement the impaired ROS production of grp7-1 8i (Fig. 3E and fig. S8E), indicating that the loss of GRP7 was indeed the cause of this phenotype. Together, these results further indicated that FER and GRP7 are involved in RALF1 responses.

In addition to plant immunity, the RALF1-FER pathway regulates ABA responses, thus modulating responses to abiotic stresses (10). In the absence of ABA, the primary roots of the different genotypes analyzed were nearly identical (Fig. 3, F and G). However, in the presence of 5 μM ABA, the grp7-1 8i and fer-4 mutants showed stronger root growth inhibition than the WT (Fig. 3, F and G). The ABA-sensitive phenotypes of grp7-1 8i with regard to root growth were rescued by overexpression of GRP7. When comparing the phosphorylation sites mutations of GRP7, we found that GRP7mut6A-GFP/grp7-1 8i showed hypersensitivity to ABA, while GRP7mut6D-GFP/grp7-1 8i showed higher ABA tolerance than GRP7-GFP/grp7-1 8i plants (Fig. 3, F and G). We next assessed the cotyledon greening responses of these different genotypes with or without ABA and reached conclusions similar to those for the ABA-induced primary root growth inhibition (fig. S8, F and G). These results indicate that GRP7, together with FER, plays a negative role in ABA responses and that the phosphorylation of GRP7 is crucial to its role in this process.

We next wondered how the RALF1-FER pathway mechanistically affects RNA splicing via the FER-mediated phosphorylation of GRP7. Given that pre-mRNA splicing occurs in the nucleus and that GRP7 is a nucleocytoplasmic shuttling protein (22), we tested the impact of RALF1 on the subcellular localization of GRP7. We treated GRP7::GRP7-GFP seedlings (grp7-1 plants expressing GRP7-GFP under the control of its native promoter, thus mimicking the endogenous expression level) (23) with RALF1. Confocal microscopy revealed that GFP was present in the nuclei of more cells after RALF1 treatment than mock treatment (Fig. 4, A and B). Similarly, immunofluorescence assays showed that RALF1 triggered the accumulation of GRP7 in the nucleus, and fewer cells showed nuclear localization of the GRP7 protein in the fer-4 background than in the WT background after RALF1 treatment (fig. S9, A and B). Nuclear and cytoplasmic fractionation assays showed that GRP7 accumulated in the nucleus in WT plants after RALF1 treatment, but this process was impeded in fer-4 mutants (fig. S9C). With or without RALF1 treatment, fewer cells showed GFP in the nucleus in the GRP7-GFP/fer-4 and GRP7mut6A-GFP lines than in the GRP7-GFP line (fig. S9, D and E). In addition, more cells showed GFP in the nucleus in the GRP7mut6D-GFP/grp7-1 8i compared to GRP7-GFP/grp7-1 8i (fig. S9, F and G). On the basis of these observations, we suggest that the nuclear accumulation of GRP7 is promoted by the RALF1-FER pathway and that this promotion is partially dependent on FER-mediated phosphorylation.

Fig. 4 RALF1-FER regulates RNA splicing via FER-mediated phosphorylation of GRP7.

(A) Representative images of GRP7::GRP7-GFP roots in 5-day-old seedlings with or without RALF1. (B) The chosen layer [yellow lines inside (A) show the areas used for line scan measurements] was analyzed for intensity using ImageJ, and the resulting plot profiles are shown. (C) Y2H analysis of the interaction between GRP7 and U1-70K or U2AF35A. (D) BiFC assay in Arabidopsis protoplasts. The interaction between GRP7 and U1-70K, which produced GFP, occurred in the nucleus. (E) Bar graph of the numbers of protoplasts with nuclear GFP in (D) with or without RALF1 peptide [1 μM, 3 hours; n (protoplast number) >114 in each experiment]. The values represent the means ± SD of three replicates. Student’s t test, **P < 0.01. (F) Ability of GRP7 to bind ABF1 sense RNA probes. The quantities of 1× GRP7 and 1× pGRP7 were the same. (G) RIP assays. ABF1, PUB9, and LRK10L1.1 pre-mRNA fragments containing putative GRP7 binding sites are indicated in fig. S8A. The levels in the GFP-trap precipitate are presented relative to the levels in the input. Values are means ± SD (n = 3 technical replicates). ANOVA with Tukey’s test; **P < 0.01. (H) Semi-qPCR analysis of GRP7 transcripts in different backgrounds. All assessments were independently repeated four times with similar results.

To begin to understand how GRP7 influences the activity of the spliceosome, we tested whether GRP7 interacted with the U1 and U2 small nuclear ribonucleoproteins (snRNPs), which play roles in defining 5′ and 3′ splice sites, respectively (24). We first assessed the interaction of GRP7 with U1-70K or U2AF35A using Y2H and found that GRP7 interacted with U1-70K but not U2AF35A (Fig. 4C). We then used BiFC to confirm that GRP7-cCFP interacted with U1-70K-nVenus in the nucleus, but not with the GATA-type transcription factor ZML1 negative control (Fig. 4D). The percentage of cells that showed a GFP fluorescence signal in the nucleus increased when RALF1 was added to Arabidopsis protoplasts (Fig. 4E), indicating that RALF1 treatment increased the interaction between GRP7-cCFP and U1-70K-nVenus in the nucleus.

In previous work, 452 GRP7 binding targets have been identified by both individual nucleotide resolution cross-linking and IP (iCLIP) and RNA IP (RIP) sequencing (23). We performed GO analysis and noticed that these targets were involved in photosynthesis, the cold response, and the ABA response (fig. S10A). Taking ABA-related genes as examples, we analyzed the genomic DNA sequences of ABA-related genes with defective splicing in grp7-1 8i and found that several of them, including ABF1 (25), PUB9 (26), and LRK10L1.1 (27), contained one of the reported core GRP7 RNA binding consensus motifs (UU/GCUGG) (22) (fig. S10B). An RNA electrophoretic mobility shift assay (EMSA) was used and confirmed that GRP7-GST, but not the GST control protein, bound the PUB9, ABF1, and LRK10L1.1 mRNAs (fig. S10, C to E). We also tested whether FER-mediated phosphorylation affected the RNA binding activity of GRP7. We used an ABA incubation system to control the in vitro phosphorylation of GRP7-GST by FER-KD (19) and found that the RNA binding ability of phosphorylated GRP7 (pGRP7-GST) was higher than that of unphosphorylated GRP7 (GRP7-GST) toward ABF1 mRNA (Fig. 4F). As RALF1 enhanced the phosphorylation level of GRP7, we speculated that the RALF1-FER module may enhance the RNA binding ability of GRP7 in vivo. Thus, we compared the RNA binding ability of GRP7-GFP in the WT background (GRP7-GFP) and the FER knockout background (GRP7-GFP/fer-4). RIP assays showed that GRP7 bound to the ABF1, PUB9, and LRK10L1.1 mRNAs (Fig. 4G), whereas the negative control BAK1 transcript was not enriched in our RIP assays (16). Next, we tested whether RALF1 affects the mRNA binding ability of GRP7. The results showed that RALF1 enhanced the ability of GRP7 to bind ABF1, PUB9, and LRK10L1.1 in the WT background but not in the fer-4 background (Fig. 4G), further demonstrating that FER mediates the RALF1-triggered regulation of the RNA binding ability of GRP7. These data suggest that RALF1 enhances the FER-mediated phosphorylation of GRP7 to modulate the ability of GRP7 to bind its target mRNAs.

We further investigated the functional consequences of the RALF1-FER-GRP7 module-driven splicing changes. Given that ABA response element binding factors (ABFs) play crucial roles in ABA signaling (28), we decided to characterize the AS of the ABA general transcription factor ABF1 in detail. ABF1 is a bZIP (basic region/leucine zipper motif) transcription factor for which our RNA-seq data detected two mRNA isoforms. The full-length transcript ABF1.1 encoded a protein with 392 amino acids (29), while the truncated variant ABF1.2 encoded a protein with 403 amino acids due to a frameshift (fig. S11A). According to our RNA-seq and semi-qPCR results, in the fer-4 and grp7-1 8i mutants, ABF1.1 displayed increased expression levels, while ABF1.2 exhibited reduced expression (fig. S11B); this phenomenon may have contributed to the ABA response defects of the fer-4 and grp7-1 8i mutants. To confirm this, we overexpressed ABF1.1 and ABF1.2 in WT and found that plants overexpressing ABF1.2 were resistant to ABA in seed germination assays, while plants overexpressing ABF1.1 were sensitive compared with WT (fig. S11, C and D). ABF1.2 overexpression partially rescued the ABA response defects of grp7-1 8i (fig. S11, E to H). Reversal of a single splicing event (e.g., enhancement of ABF1.2 expression) after GRP7 depletion only modestly rescued plant viability. This result was not unexpected, given that multiple splicing events are affected upon perturbation of the RALF1-FER pathway. These data suggest that the RALF1-FER-GRP7 module regulates specific target genes (e.g., ABF1) to modulate specific cellular activities and thus affects ABA responses and immune responses in plants.

Regulated unproductive splicing and translation (RUST) is a mechanism that links AS with nonsense-mediated mRNA decay (NMD) to regulate protein abundance by fine-tuning the generation of unproductive isoforms that are further degraded by the NMD pathway (30). We noticed that GRP7 could be regulated via RUST. An alternatively spliced variant (as_GRP7) retaining part of the intron with a premature termination codon can be degraded via NMD, which reduces the GRP7 RNA level and ultimately affects the protein abundance (23, 31). We found that the level of as_GRP7 in GRP7mut6D-GFP and GRP7-GFP was higher than that in WT, while its level was lower in GRP7mut6A-GFP (Fig. 4H), indicating that increases in GRP7 activity (i.e., phosphorylation and/or overexpression of GRP7) will trigger the production of as_GRP7 and suggesting that as_GRP7 up-regulation may act as a negative feedback mechanism. We took the confirmed FER-GRP7 roles in ABA response as an assay to check this hypothesis because the FER-GRP7 pathway inhibits ABA responses. We found that the GRP7 protein level was sharply reduced upon ABA treatment (fig. S12A). Addition of the proteasome inhibitor MG132 (N-carbobenzyloxy-l-leucyl-l-leucyl-l-leucinal) did not block this process, indicating that the reduction in GRP7 protein may not have occurred via the ubiquitination pathway (fig. S12A). We then evaluated whether ABA regulates GRP7 protein levels by regulating RNA levels. However, the total GRP7 level increased after ABA treatment (fig. S12B). With regard to AS, we found that the levels of as_GRP7, which harbors a premature stop codon, increased after ABA treatment (fig. S12C), while those of fully spliced GRP7 RNA (fs_GRP7) decreased significantly (fig. S12C). We further confirmed this result by mutants defective in ABA responses and found that the as_GRP7 levels were increased in three ABA hypersensitive mutants [abi1-2/abi2-2/hab1-1 (20), hab1-1/abi1-2/pp2ca (20), and fer-4], whereas as_GRP7 levels were decreased in an ABA response–reduced mutant [ABA receptor hextuple, 112458 (20)] (fig. S12D). These data indicate that ABA signaling can lead to the NMD-mediated regulation of GRP7. In summary, these data are consistent with the hypothesis that regulation of GRP7 via RUST can fine-tune the signaling strength of the FER-GRP7 pathway.


The splicing factor GRP7 is involved in endogenous processes, including circadian timekeeping, flowering time control, and responses to a plethora of abiotic and biotic factors (1517, 32). Furthermore, RALFs-FER acts as a central regulator of the stress responses, via posttranslational modification and/or transcription (10). This study reports a link between RALFs-FER and global AS changes.

Environmental cues (e.g., abiotic stress) may regulate ABA and RALF1 levels in plants (10, 20), and we found that RALF1 triggers massive AS changes dependent on its interaction with FER. Mechanistically, we found that FER phosphorylates GRP7 at six serine/tyrosine residues, enhancing GRP7 accumulation in the nucleus, GRP7 RNA binding, and GRP7 interaction with U1-70K. This provides a basis for the observed AS changes, affecting genes involved in stress responses, such as ABA signaling and salt stress (fig. S12E). Despite the fact that RALF1 triggers GRP7 phosphorylation, expression of GRP7mut6D in the grp7-1 8i mutant did not exert a constitutive inhibition of root growth. Therefore, in addition to GRP7 phosphorylation, other factors may contribute to the AS changes in response to RALF1 (fig. S12E). Overall, this work reveals an entire major pathway in which extracellular peptide signals regulate global splicing dynamics to allow organisms to anticipate and prepare for environmental changes. Different peptides may recruit FER and/or its relatives (e.g., HERK2), LRE-related GPI-APs [e.g., LLG1 (11)], and leucine-rich repeat extensin [LRX (33)] proteins to regulate various downstream RNA binding proteins (e.g., GRP7, GRP8, and others); however, whether these proteins are specifically recruited by a given peptide is still not clear and requires future studies. In addition, FER and distinct FER relatives might also recruit different downstream factors, such as guanine nucleotide exchange factor–ROP, RPM1-induced protein kinase (RIPK); ErbB3-binding protein 1 (EBP1); ABA insensitive 2 (ABI2); and MYC2, to mediate different signaling pathways (8, 10, 11, 19, 20, 33, 34), representing the molecular mechanisms of the observed versatile and context-specific roles of the RALF/FER module (9). Notably, we observed that AS of GRP7 itself and accumulation of an isoform with a premature termination codon at the expense of the mature mRNA in response to ABA led to reduced GRP7 protein levels, providing a feedback control mechanism for RALF1-FER signaling, as previously shown for circadian oscillations of GRP7 (17). In the future, more studies are needed to elucidate how RLKs, such as FER, expand their regulatory capabilities by controlling RNA-processing events in response to environmental cues. Understanding the mechanisms involved in the RALFs-FER-GRP7 pathway will provide a notable model for interpreting the phenotypic consequences of the complexity and heterogeneity of gene expression across different tissues. In addition, using CRISPR-Cas gene editing to (re)program RNA processing (such as splicing) may provide entry points for modulating stress responses and growth in plants.


RNA-seq sample preparation

WT (Col-0), fer-4, and grp7-1 8i seedlings were grown on ½ MS medium for 7 days under long-day conditions (16-hour light/8-hour dark). To avoid an effect of the manipulation of the seedlings during the transfer from solid to liquid ½ MS medium for RALF1 treatment, we first removed the seedlings from the solid ½ MS medium. Then, we transferred the seedling roots into liquid ½ MS medium and preincubated them for 1 hour before treatment. Then, Col-0, fer-4, and grp7-1 8i seedlings were soaked with 1 μM RALF1 (included in the ½ MS liquid medium) or with mock control (½ MS medium containing protein elution buffer used for RALF1 purification) for 1 hour. The root samples were collected at ZT (zeitgeber time) 12 for RNA extraction. Total RNA was extracted with a mirVana microRNA Isolation kit (Ambion, AM1561) following the manufacturer’s protocol. RNA integrity was evaluated using the Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA). For all samples, the RNA integrity number was ≥9, indicating excellent RNA quality. The cDNA libraries were constructed by using the TruSeq Stranded mRNA LT Sample Prep Kit (Illumina, San Diego, CA, USA) according to the manufacturer’s instructions. The library was prepared for Illumina sequencing by Shanghai OE Biotech. Co. Ltd. and was sequenced on the Illumina sequencing platform (Illumina HiSeq X Ten).

Bioinformatics analysis

Read quality was evaluated, and low-quality reads were removed using Trimmomatic v0.36 ( (35) with the parameters “CROP:150 ILLUMINACLIP:TruSeq3-PE-2.fa:2:30:10:8:true LEADING:3 TRAILING:3 SLIDINGWINDOW:4:15 MINLEN:50”. High-quality reads were mapped to the Arabidopsis TAIR10 genome using HISAT2 v2.1.1 ( (36) with the parameter “-rna-strandness rf-fr”. Sam files were converted to sorted bam files using SAMtools v1.4 (37). The FPKM value of each gene was calculated using cufflinks (38), and the read counts of each gene were obtained by htseq-count (39). Differentially expressed genes were identified using the DESeq (40) R package functions estimateSizeFactors and nbinomTest. A P value of <0.05 and a fold change of >2 or <0.5 were set as the threshold for significant differential expression. Uniquely mapped reads were used for the identification of differential AS events using rMATS v4.0.1 ( (41) with parameters “-t paired –len 150 –c 0.0001 –analysis U –libType fr-firststrand –novelSS 1.” Events with P < 0.05 and exon inclusion level change of >0.05, calculated on the basis of junction reads by rMATs, were identified as significant differential AS events. Inclusion level (ψ) was a quantitative measurement of AS (41). With exon skipping as an example, inclusion level is estimated by the proportion of exon-exon junction counts supporting the exon-inclusion isoform. Inclusion level can be applied to all AS categories (exon skipping, intron retention, alternative 5′/3′ splice site, mutually exclusive exons), with minor changes made when defining “inclusion.” The GO enrichment analysis was conducted using DAVID (42) v6.8 ( PCA was carried out on the basis of 1000 gene expression levels or inclusion levels (splicing) with the largest variance across the six samples (three mock controls and three RALF treatment samples) using the prcomp function in R (version 3.5.1). The RNA-seq results of splicing changes are summarized in table S1.

RNA extraction

Samples under different treatments, as indicated, were powdered in liquid nitrogen for the RNA extraction. Total RNA was extracted by TRIzol reagent (Ambion, 15596-026) and digested by deoxyribonuclease I (DNase I; Takara) to remove genomic DNA. First-strand cDNA was synthesized by using a cDNA synthesis kit (Fermentas, K1622), according to the manufacturer’s instructions.

Gene expression analyses

qPCR was performed using the CFX96 Touch Real-Time PCR Detection System (Bio-Rad) with SYBR Premix ExTaqII (Takara). Semi-qPCR mixes were prepared using 2× Master Mix (TsingKe, TSE004) according to the manufacturer’s instructions with a final primer concentration of 200 nM. Reactions were performed using a PCR System (C1000, Bio-Rad) with initial incubation at 95°C for 10 min, followed by 26 cycles of 15 s at 95°C, 30 s at 55°C, and 1 min/kb at 72°C. ACTIN2 was used as a reference in qPCR analysis. PP2A was used as an internal reference in semi-qPCR analysis (43). Data are shown as the mean expression ± SD. When indicated, the fold change is shown for the treated sample in comparison to the control sample.

The following genes with significant AS changes in the RNA-seq data were chosen for independent validation by semi-qPCR. We used ImageJ to detect the ratio of two splice isoform bands. We obtained three results from three independent repeats and used one-way analysis of variance (ANOVA) to calculate statistical significance.

AT1G13450 is a homeodomain-like superfamily protein, which is assumed to act as a molecular switch modulated through Ca2+-dependent phosphorylation/dephosphorylation in response to light signals. AT1G78070 encodes a transducin/WD40 repeat-like superfamily protein. AT5G49230 is a drought-responsive family protein. AT1G28560 is a small nuclear RNA-activating complex family protein that encodes a protein similar to human SNAP50. AT3G27340 encodes a Myb domain protein. AT2G25000 encodes the WRKY60 transcription factor. AT1G13460 encodes the protein phosphatase 2A regulatory B subunit family protein. AT5G42030 encodes an ABL interactor-like protein 4. AT1G19025 encodes a DNA repair metallo-beta-lactamase family protein. AT3G12640 is an RNA binding (RRM/RBD/RNP motif) family protein. AT1G69400 encodes a transducin/WD40 repeat-like superfamily protein. AT1G79790 is a haloacid dehalogenase-like hydrolase superfamily protein that encodes a chloroplast-localized flavin mononucleotide (FMN) hydrolase. AT3G44740 is a class II aaRS and biotin synthetase superfamily protein. AT3G59810 encodes a snRNP family protein. AT1G29030 encodes an apoptosis inhibitory protein. AT1G44180 is a peptidase M20/M25/M40 family protein. AT1G54110 encodes a membrane fusion protein. AT1G76040 encodes the calcium-dependent protein kinase 29. AT5G15460 encodes a membrane-anchored ubiquitin-fold protein. AT1G48175 encodes a cytidine/deoxycytidylate deaminase family protein. AT1G13460 encodes a protein phosphatase 2A regulatory B subunit family protein. AT4G32060 encodes a calcium-binding EF hand family protein.

Because GRP7 and GRP8 have highly similar cDNA sequences, we used GRP7 primers that could distinguish GRP7 from GRP8 as previously described (13). The primers used for qPCR and semi-qPCR are listed in table S1.

Seeds and plant growth conditions

Arabidopsis thaliana ecotype Col-0 was used as the WT for all experiments. grp7-1 8i and GRP7::GRP7-GFP, fer-4, and Ubi:FER-FLAG were described in our previous works (15, 18, 23). sua-4 (44), sr45-1 (45), and lsm8-1 (12) were provided by Y. Zhang (University of British Columbia), P. Duque (Instituto Gulbenkian de Ciência), and J. Kufel (University of Warsaw), respectively. The pCAMBIA2300 plasmid carrying GRP7-GFP, GRP7mut6A-GFP, or GRP7mut6D-GFP and the pCAMBIA1300 plasmid carrying ABF1.1-OE or ABF1.2-OE were transformed into Agrobacterium tumefaciens strain Ag10 before transformation of the Arabidopsis ecotype Col-0. Expression of the transgenes was confirmed via qPCR. We also detected the GRP7 protein levels in these transgenic plants using the anti-GRP7 antibody (fig. S6, C and D). The GRP7-GFP/grp7-1 8i-3, -4, GRP7mut6A-GFP/grp7-1 8i-1, -5, GRP7mut6D-GFP/grp7-1 8i-3, -4, ABF1.1-OE/grp7-1 8i-1, -3, and ABF1.2-OE/grp7-1 8i-5, -6 lines were obtained by transformation of plasmids into the grp7-1 8i plants (fig. S6E). Transgenic plants were selected on kanamycin (40 μg/ml) (pCAMBIA2300)/hygromycin (pCAMBIA1300). GRP7-GFP-2/fer-4 lines were obtained by crossing GRP7-GFP-2 with the fer-4 mutant (fig. S6F). For Arabidopsis growth, seeds were surface-sterilized with 75% ethanol for 3 min, washed three times with sterile water, sterilized with 15% bleach for 3 min, washed three times with sterile water, and then grown on ½ MS medium containing 0.8% (w/v) sucrose and 0.8% (w/v) agar (Sigma-Aldrich). The plates were stratified in darkness for 2 days at 4°C and then transferred to a chamber set at a light intensity of 80 μmol m−2 s−1 and a constant temperature of 22°C under 16-hour light/8-hour dark for 6 days. Then, the seedlings were potted in soil and placed in a growth room.

RALF1 and ABA response assays

For RALF1 root growth inhibition assays, 4-day-old seedlings with comparable growth vigor were transferred to liquid ½ MS medium containing 1 μM RALF1 for 2 days and then analyzed as previously described (18). For the ABA germination assay, plants of different genotypes were grown side by side, and seeds were harvested at the same time. After harvest, the fresh seeds (~40) of each genotype were surface-sterilized and sown on ½ MS plates supplemented with different ABA concentrations, as indicated in the figure. Seedlings with expanded green cotyledons were considered germinated and were used for statistical analysis at the indicated time points. For root elongation assays, seeds were stratified for 2 days in the dark at 4°C, sown, and grown on vertically oriented ½ MS plates for 4 days. Seedlings with similar primary root lengths were transferred to new ½ MS plates supplemented with the indicated concentrations of ABA. The plates were scanned on the fourth day after transfer, and primary root length was measured using ImageJ (

Y2H assays

The kinase domain coding region of FER (469 to 896 amino acids, FER-KD) and the kinase domains of its relatives (At5g24010, CVY1, At4g39110, At5g39020, At5g38990, HERK2, and THE1) were cloned into the pGBKT7 (BD) vector (18), and FER-KD-BD was applied as bait in a Y2H screen of the Arabidopsis cDNA library for positive protein-protein interactions (20). Briefly, the FER kinase domain cloned into the pGBKT7 vector and the Arabidopsis cDNA library cloned into the prey vector pGADT7 were transformed into AH109 yeast cells containing FER-KD-BD. Then, the transformed cells were plated on synthetic dropout selection medium that lacked Trp, Leu, and His supplemented with 20 mM 3-AT to inhibit self-activation. For further confirmation of the interaction between GRP7/GRP8 and FER-KD, the full-length, N-terminal RRM, a C-terminal region comprising the glycine-rich region of the GRP7 coding sequence, or the full-length coding sequence of GRP8 was cloned into the pGADT7 (AD) vector and cotransformed with FER-KD into yeast strain AH109 to test their interaction. To analyze the interaction between GRP7 and U1-70K or U2AF35A, the full-length GRP7 coding sequence was subcloned into the pGBKT7 vector, and the full-length U1-70K (AT3G50670) or U2AF35A (AT1G27650) coding sequence was cloned into pGADT7and cotransformed into AH109 to test for an interaction (20). The primers used for cloning are listed in table S1.

BiFC and confocal microscopy

The vectors for BiFC were pSAT1-cCFP (pE3449) and pSAT1-nVenus (pE3308) (18). When cCFP and nVenus are brought together via the proteins that are fused to them, GFP fluorescence will be detected (18, 20). The coding regions of GRP6 and GRP7 were cloned into pSAT1-cCFP to generate GRP6-cCFP and GRP7-cCFP, while the coding regions of FER and CVY1 were cloned into pSAT1-nVenus to generate FER-nVenus and CVY1-nVenus. The full-length coding sequence of U1-70K was cloned into pSAT1-nVenus. The primers used for the PCR are listed in table S1. The constructs were cotransfected into Arabidopsis protoplasts using polyethylene glycol according to our previous work (20). Briefly, mesophyll protoplasts were isolated from 4-week-old seedlings and transformed with 2 μg of purified plasmid DNA per construction. The transformed protoplasts were incubated in the dark at 22°C for 14 hours. After incubation, the transformed protoplasts were divided into two parts for RALF1 peptide treatments. For RALF1 treatment, 1 μM RALF1 (or the elution buffer used for RALF1 purification as mock) was added to the protoplasts and incubated for 3 hours before microscopic analysis. Fluorescence signals were detected by using a Nikon confocal laser scanning microscope with a 20× objective lens. The following wavelengths were used for fluorescence detection: excitation at 488 nm and emission at 490 to 530 nm for GFP, excitation at 543 nm and emission at 560 to 620 nm for FM4-64 and red fluorescent protein, and excitation at 350 nm and emission at 430 to 480 nm for 4′,6-diamidino-2-phenylindole (DAPI). At least three independent experiments were performed for all confocal imaging data.

Protein purification and GRP7 antibody preparation

The RALF1 protein was purified as described in the manual of His Purification System (Invitrogen, R901-15) as described previously (18). The GST-tagged GRP7 were purified as described according to the manual of Pierce Glutathione Agarose (Thermo Fisher Scientific, 16100). For GRP7 antibody production, purified GRP7-GST protein was used as the antigen, and a 30-day ICR (Institute of Cancer Research) mouse (Shanghai Laboratory Animal Center laboratory animal) was injected with 50 μg of GRP7-GST protein emulsified with Complete Freund’s adjuvant (Sigma-Aldrich, F5881). The injection was repeated after 2 and 3 weeks with Incomplete Freund’s adjuvant (Sigma-Aldrich, F5506). The serum of the immunized mouse was obtained as a GRP7 antibody, and we tested it by WB detection using protein extracts from grp7-1 mutant and plants that were transformed with distinct GRP7 mutant forms tagged with GFP (fig. S6D). We found that the GRP7 antibody can strongly detect the GRP7 and GRP7-GFP protein, but the GRP7 antibody still detects a dim band in the grp7-1 mutant background. We assume that this band was caused by elevated levels of the highly similar homolog GRP8 in the grp7-1 mutant background, as reported previously (15). Because our GRP7 antibody detected only a dim band when GRP8 was strongly expressed in combination with a relatively high loading amount in WB, and FER can also interact with GRP8, we suggest that our GRP7 antibody was suitable to perform the RALF1-FER pathway–related biochemical assays in this project.

GST pull-down analyses

The construction and expression of His-FER-KD were described previously (18). GRP7 or GRP8 was cloned into the pGEX4T-1 vector and transformed into E. coli BL21 star (DE3) cells that produced the GST fusion protein upon induction by 0.5 mM isopropyl-β-d-thiogalactopyranoside. The His-FER-KD and GRP7-GST/GRP71–86-GST/GRP787–176-GST/GST proteins were coincubated with 30 μl of glutathione agarose beads (Thermo Fisher Scientific, 16100) in binding buffer [50 mM tris-HCl (pH 8.0), 150 mM NaCl, and 10 mM MgCl2] at 4°C for 6 hours. The beads were washed with washing buffer I [50 mM tris-HCl (pH 8.0), 300 mM NaCl, and 0.5% Triton X-100] for 10 min and then two times with washing buffer II [50 mM tris-HCl (pH 8.0), 150 mM NaCl, and 10 mM MgCl2], each for 20 min. The proteins on the beads were eluted by boiling in SDS loading buffer, separated by 1× SDS–polyacrylamide gel electrophoresis (PAGE), and detected by immunoblotting using a GST (SC-80998, Santa Cruz Biotechnology) or His antibody (M20001, Abmart). The GST input protein was visualized by Coomassie brilliant blue (CBB) staining.

Co-IP analyses

One-week-old Col-0 and FER-FLAG transgenic seedlings were ground to powder with liquid nitrogen. The total proteins were extracted with NEB buffer [20 mM Hepes-KOH (pH 7.5), 40 mM KCl, 1 mM EDTA, 1 mM phenylmethylsulfonyl fluoride (PMSF), and protease inhibitor (Thermo Fisher Scientific, 78420)] and NEB-T buffer (NEB buffer containing 1% Triton X-100) for 1 hour at 4°C. After centrifugation at 14,000g for 20 min, the supernatant was incubated with 30 μl of anti-FLAG affinity gel (Sigma, A2220) at 4°C for 4 hours. The agarose beads were washed three times with washing buffer [20 mM Hepes-KOH (pH 7.5), 40 mM KCl, and 0.1% Triton X-100]. The immunoprecipitates associated with the agarose gel were boiled in 1× SDS loading buffer for 10 min, separated by 12% SDS-PAGE, and detected with anti-FLAG (Abmart, M20008; 1:4000) and anti-GRP7 antibodies (1:3000). Notably, to analyze the total protein level of FER and compare the interaction impacts of the RALF peptide, we shortened the SDS-PAGE running time to prevent separation of the phosphorylated and dephosphorylated forms of FER (18, 20).

Phosphorylation assays and identification of phosphorylation sites

Recombinant His-FER-KD and its kinase-dead variant (His-FERK565R-KD) were generated as previously described (18). The in vitro phosphorylation assay was performed as described by Du et al. (18). Briefly, 1 μg of His-FER-KD or His-FERK565R-KD and 1 μg of GRP7-GST or GRP7mut6A-GST were added to kinase assay solution [25 mM tris-HCl (pH 7.5), 10 mM MgCl2, 1 mM CaCl2, and 1 mM dithiothreitol (DTT)]. The assay was initiated by adding 1 mM adenosine triphosphate and incubated for 30 min at 30°C. For CIP assays, alkaline phosphatase (Thermo Fisher Scientific, EF0651) was added to the solution and incubated at 30°C for 10 min. The reaction was stopped by the addition of 4× SDS loading buffer and subsequent incubation at 95°C for 10 min. The proteins were then separated by 12% (w/v) SDS-PAGE and analyzed by WB using anti-pSer (1:3000; Abcam, ab9332) and anti-pTyr (1:3000; Abcam, ab17302) antibodies. For phosphorylation site identification, GRP7 was coexpressed with the FER-KD or FERK565R-KD via our coexpression system that can identify the substrate phosphorylation residue(s) of FER kinase (19) and then purified and subjected to alkylation/tryptic digestion followed by MS (19). The His-FERK565R-KD and GRP7 combination was used as a negative control.

For phosphorylation assays in vivo, 2 g of GRP7-GFP, GRP7-GFP/fer-4, or Col-0 seedlings were mock-treated or treated with 1 μM RALF for 30 min, and then, total protein was extracted by using 0.5 ml of NEB buffer [20 mM Hepes-KOH (pH 7.5), 40 mM KCl, 1 mM EDTA, 1 mM PMSF, 1% protease inhibitor mixture (Thermo Fisher Scientific, 78420), and 1% phosphatase inhibitors (Bimake, B15001)] and 0.5 ml of NEBT buffer [20 mM Hepes-KOH (pH 7.5), 40 mM KCl, 1 mM EDTA, 1% Triton X-100, 1 mM PMSF, 1% protease inhibitor mixture (Thermo Fisher Scientific, 78420), and 1% phosphatase inhibitors (Bimake, B15001)]. The homogenized sample was centrifuged twice at 12,000g for 10 min each at 4°C. Then, 20 μl of GFP-trap bead (Chromotek, gta-100) for each sample was used to immunoprecipitate GRP7-GFP protein complexes at 4°C for 5 hours. The beads were washed three times in washing buffer [20 mM Hepes-KOH (pH 7.5), 40 mM KCl, and 0.1% Triton X-100] at 4°C. Last, the beads were resuspended in 40 μl of 2× SDS loading buffer, boiled for 10 min, separated by 12% (w/v) SDS-PAGE, and probed with anti-pSer (1:3000; Abcam, ab9332), anti-pTyr (1:3000; Abcam, ab17302), and anti-GFP antibodies (1:5000; CMC, at0028). For phosphorylation site identification in vivo, 8 g of GRP7-GFP was mock-treated or treated with 1 μM RALF1 for 30 min and extracted with 1 ml of NEB and 1 NEBT buffer for 1 hour at 4°C. Eighty microliters of GFP-trap beads (Chromotek, gta-100) was used for each sample. After washing three times in washing buffer, the beads were added 120 μl of 2× SDS loading buffer and boiled for 10 min. Then, 12% (m/v) SDS/PAGE was used for protein separation. The GRP7-GFP bands were excised and cut into small pieces after CBB staining. Then, the pieces were dehydrated, and the proteins were reduced, alkylated, digested, and analyzed by MS as described previously (19, 46).

MS data analysis

For label-free quantification, raw files were processed using MaxQuant (47) (version and Proteome Discoverer (Thermo Fisher Scientific, version 1.4). GRP7 phosphopeptides were identified by searching all tandem MS spectra against version of the Araport11_pep_201606 sequence database and filtering using a false discovery rate of <0.01 at the peptide level and of <0.05 at the protein level by MaxQuant. We selected carbamidomethylation of cysteines as the fixed modification and used oxidation (M) and phosphorylation (STY) as variable modifications. In addition, we used Proteome Discoverer to assess phosphorylation localization. After identification, we performed GRP7 protein label-free quantitation of phosphopeptides with the MaxLFQ algorithm integrated in the MaxQuant software suite according to Liu et al. (46).

Phylogenetic tree construction

Protein sequences of different species of GRP7 were blasted and downloaded from the National Center for Biotechnology Information (NCBI) on the basis of the protein sequence of AtGRP7. Sequences of GRP7-like protein from different species were aligned using DNAMAN and displayed with iTOL (48). To analyze the phosphorylation sites, we randomly selected 12 species in the same branch and aligned the sequences using DNAMAN and BioEdit (version 7.0.4) with default settings.

Flagellin22-induced ROS production

Leaves of 4-week-old plants were cut into 4 mm by 4 mm pieces. Leaf pieces were transferred to 96-well plates containing 100 μl of sterile water (pH 5.7) and allowed to recover in the dark for 16 hours. The next day, the water was replaced by 75 μl of 2 mM MES-KOH (pH 5.8) to mimic the apoplastic pH. Leaf pieces were incubated further for 5 hours before adding 20 μM luminol L-012 (Sigma-Aldrich), horseradish peroxidase (10 μg/ml; Sigma-Aldrich), 100 nM flg22, and 200 nM RALF1 peptide solution. Luminometric output was measured with a FLUOROSKAN ASCENT FL (Thermo Fisher).

Immunofluorescence assay

Immunofluorescence was performed as previously described (49). Briefly, 5-day-old Col-0 and fer-4 seedlings were mock-treated or treated with 1 μM RALF1 for 3 hours. Then, the seedlings were incubated in fixative solution [1× phosphate-buffered saline (PBS) (pH 7.2) containing 2% paraformaldehyde and 0.1% Triton X-100] for 5 min and subsequently vacuum-infiltrated for 10 min in a vacuum pump (0.05 Mpa) at room temperature (24°C). Then, we removed the fixative solution, washed the seedling three times for 10 min each with 1× PBS, and digested the cell wall with cell wall digestion solution [2% Driselase (Sigma, D8037) in 1× PBS] for 18 min at 37°C. We washed the seedling with 1× PBS five times and then incubated the seedling in tissue permeation solution (3% IGEPAL CA-630 and 10% dimethyl sulfoxide in 1× PBS) for 20 min. We washed the seedling three times with 1× PBS and then incubated it with blocking solution [3% BSA (Ameresco, 0332) in 1 × PBS] for 1 to 2 hours. Then, we incubated it with anti-GRP7 or serum control antibody for 10 hours at 4°C in the dark and washed it with 1× PBS three times before incubating it with goat anti-mouse secondary antibody (Sungene Biotech, GM200G-37C) in blocking solution at 37°C for 5 hours in the dark. Last, we washed the sample five times with 1× PBS, incubated the seedling with DAPI for 30 min, washed the sample three times with 1× PBS, stored it at 4°C, and prepared it for observation within 3 days.

Subcellular fractionation

Nuclear and cytoplasmic protein fractionation was performed as previously described (19, 22). Briefly, 1-week-old Col-0 and fer-4 seedlings were treated with 1 μM RALF1 for 3 hours and then ground in liquid nitrogen. Samples were extracted with 100 μl of fractionation buffer [20 mM tris-HCl (pH 7.0), 250 mM sucrose, 25% glycerol, 20 mM KCl, 2 mM EDTA, 2.5 mM MgCl2, 30 mM beta-mercaptoethanol, 1× protease inhibitor cocktail, and 0.7% Triton X-100] and then centrifuged at 5000g for 5 min. The supernatant was considered the cytoplasmic fraction and was stored on ice until use. The pellet was further washed with resuspension buffer [20 mM Tris-HCl (pH 7.0), 25% glycerol, 2.5 mM MgCl2, and 30 mM beta-mercaptoethanol] three times to obtain the nuclear fraction. We used glyceraldehyde-3-phosphate dehydrogenase [(GAPC) (1:1000; AS152894, Agrisera)] and histone H3 antibodies (1:5000; ab1791, Abcam) to confirm the fractionation of the nucleus and cytoplasm by WB, respectively.

GRP7 binding site analysis

We downloaded the genomic sequences of 2225 genes that had splicing defects in the grp7-1 8i mutant from TAIR ( FIMO (within the MEME suite) (50) was used to find the UU/GCUGG motif in the genomic sequences. We chose P < 0.001 as the filter and scanned the given strand only.

RNA-EMSA assays

Fluorescein isothiocyanate (FITC)–labeled RNA probes were generated by TsingKe Biological Technology Co. Ltd. Probes were also amplified with the primers indicated by the arrowheads in fig. S10B. The RNA-protein binding reaction was performed by incubating 0.2 pM FITC-labeled probe with 30 to 50 pM purified GRP7-GST and GST protein. To perform the competition experiments, 50 pM purified GRP7-GST and GST protein and 100× unlabeled competitor A or 100× nonspecific competitor B were incubated in binding buffer for 15 min, followed by the addition of 0.2 pM FITC-labeled probe for 20 min. To avoid RNA degradation, we added 1 μl of ribonuclease (RNase) inhibitor (Thermo Fisher, N8080119) to each reaction. Next, the binding reaction mixture was loaded onto a native 4% polyacrylamide and 0.5× tris-borate EDTA gel, which was run for 40 min and then exposed to a fluorescence imager plate. Phosphorylated GRP7 and unphosphorylated GRP7 were generated using the ABA-induced FER substrate phosphorylation system as previously described (19).

RIP and qPCR analyses

Three grams of 7-day-old 35S::GFP, GRP7-GFP, and GRP7-GFP/fer-4 seedlings were harvested at ZT 12 and mock-treated or treated with 1 μM RALF1 peptide (in ½ MS liquid medium) for 2 hours and then placed under vacuum for 15 min (0.1 Mpa). The working concentration of 0.125 M glycine was added to quench cross-linking, and the seedlings were subjected to an additional 6 min of vacuum treatment. The plants were rinsed twice with water, then ground to powder with liquid nitrogen, and solubilized with 300 μl of extraction buffer [50 mM tris-HCl (pH 8.0), 150 mM NaCl, 4 mM MgCl2, 0.1% Igepal, 5 mM DTT, 0.1% SDS, 10 mM ribonucleoside vanadyl complex (NEB, S1402S), RNase inhibitor (80 U/ml; Thermo Fisher, N8080119), 1 mM PMSF, and protease inhibitor tablets]. After incubating for 1 hour at 4°C, the extracted samples were centrifuged at 12,000g for 15 min, and the supernatant was transferred to a fresh tube. One hundred microliters of supernatant was used as input and stored at −80°C, and 100 μl of supernatant was added to 20 μl of GFP-trap beads that had been prewashed three times in binding/washing buffer [50 mM tris-HCl (pH 8.0), 150 mM NaCl, 2 mM EDTA, 1% Triton X-100, and 0.1% SDS]. Then, the beads were incubated at 4°C overnight, washed with washing buffer three times, and centrifuged at 1000g for 2 min. The beads were incubated for 10 min at 55°C in RIP elution buffer [100 mM tris-HCl (pH 8.0), 100 mM EDTA, 1% SDS, and RNase inhibitor (80 U/m)]. The protein was degraded by proteinase K, and RNAs were isolated by adding 1 ml TRIzol of RNA extraction reagent (Takara). The RNA pellet was washed with 70% (v/v) ethanol, air-dried, and dissolved in RNase-free water. The RNA sample was incubated with DNase I and reverse-transcribed using a cDNA synthesis kit (Fermentas, K1622) for qPCR. In parallel, input samples were used for quantification.


Any significant differences in data were analyzed by two-tailed Student’s t test or by multivariate comparison (one-way ANOVA) using SPSS (version 17.0) software. All statistical tests are clearly described in the figure legends and/or in the Materials and Methods. Correlations between splicing changes were represented with the Spearman correlation coefficient r; |r| ≥ 0.7 indicates strong correlation. Bar graphs were generated by GraphPad Prism 6 and show the means ± SD. Density dot plots and heat maps were drawn using R software.


Supplementary material for this article is available at

This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial license, which permits use, distribution, and reproduction in any medium, so long as the resultant use is not for commercial advantage and provided the original work is properly cited.


Acknowledgments: We thank J. Alfano, Y. Zhang, P. Duque, and J. Kufel for providing the seeds. We thank H. Xue and D. Caddell for editing the manuscript. Funding: This work was supported by grants from the National Natural Science Foundation of China (NSFC-31871396, 31571444, and 31400232), Young Elite Scientist Sponsorship Program by CAST (YESS20160001), and China Postdoctoral Science Foundation funded project (2019 M662763). Author contributions: F.Y. conceived the project and designed research. L.W., S.Z., C.L., Y.M., J.T., H.L., X.L., and J.X. performed research. D.S., Q.L., and T.Y. contributed new reagents/analytic tools. B.W. and Z.H. performed bioinformatics analyses. L.W. and F.Y. analyzed data. LW., F.Y., and D.S. wrote the manuscript. All authors reviewed and approved the manuscript for publication. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. The sequencing data reported in this paper are summarized in table S1 and are deposited in the NCBI SRA database (SRR8362598).

Stay Connected to Science Advances

Navigate This Article