PRP4KA phosphorylates SERRATE for degradation via 20S proteasome to fine-tune miRNA production in Arabidopsis

Abstract

Phosphorylation can quickly switch on/off protein functions. Here, we reported pre-mRNA processing 4 kinase A (PRP4KA), and its paralogs interact with Serrate (SE), a key factor in RNA processing. PRP4KA phosphorylates at least five residues of SE in vitro and in vivo. Hypophosphorylated, but not hyperphosphorylated, SE variants could readily rescue se phenotypes in vivo. Moreover, hypophosphorylated SE variants had stronger binding affinity to microprocessor component HYL1 and were more resistant to degradation by 20S proteasome than hyperphosphorylated counterparts. Knockdown of the kinases enhanced the accumulation of hypophosphorylated SE. However, the excessive SE interfered with the assembly and function of SE-scaffolded macromolecule complexes, causing the se-like defects in the mutant and wild-type backgrounds. Thus, phosphorylation of SE via PRP4KA can quickly clear accumulated SE to secure its proper amount. This study provides new insight into how protein phosphorylation regulates miRNA metabolism through controlling homeostasis of SE accumulation in plants.

PRP4KA-mediated phosphorylation tags Serrate/Ars2 for degradation versus 20S proteasome to control miRNA processing in plants.

INTRODUCTION

Posttranslational modifications exemplified by protein phosphorylation determine protein folding, subcellular localization, catalytic activity, or stability. Sporadic reports have implicated protein phosphorylation in regulating biogenesis of microRNAs (miRNAs), a large family of small noncoding regulatory RNAs (1, 2). In animals, miRNAs are initially processed from long primary substrates (pri-miRNAs) by a microprocessor that constitutes of ribonuclease III (RNase III)–like enzyme Drosha and DiGeorge syndrome critical region 8 (DGCR8) (3, 4). Phosphorylation of Drosha and DGCR8 stimulates pri-miRNA processing and elevates miRNA accumulation (1). Plant microprocessor/dicing complex minimally includes Dicer-like 1 (DCL1), a double-stranded RNA binding protein, and hyponastic leaves 1 (HYL1) (5). HYL1 can be phosphorylated by MPK3 (mitogen-activated protein kinase 3) (6) and SnRK2 [sucrose nonfermenting (SNF1)-related protein kinase of Group 2] (7) in vitro, but loss-of-function mutations of two kinases have an opposite effect on HYL1 accumulation and miRNA level in vivo (2), inferring context-dependent regulatory roles of HYL1 phosphorylation in miRNA production.

Serrate (SE), an ortholog of the mammalian arsenic resistance protein 2 (Ars2) (8, 9), is a multifunctional protein. SE has been best known to partner with DCL1 and HYL1 to produce miRNAs (10, 11). Whereas earlier reports argue for a direct role for SE in promoting the enzymatic activity and accuracy of DCL1 (12, 13), recent studies propose that SE might act as a scaffold or mediate liquid-liquid phase separation to recruit the core processing machinery including DCL1/HYL1 to the proper RNA substrates, or vice versa, to generate miRNAs in vivo (5, 14–16). On the other hand, SE recruits switch/sucrose non-fermentable complex (SWI2/SNF2) adenosine triphosphatase subunit chromatin remodeling factor 2 (CHR2) to remodel the secondary structure of pri-miRNAs to control miRNA production (17). SE also interacts with the nuclear exosome targeting (NEXT) complex to degrade pri-miRNAs to fine-tune miRNA biogenesis in Arabidopsis (18). Similarly, Ars2 is engaged in miRNA- and small interfering RNA (siRNA)–dependent silencing in mammals, suggesting the conserved function of SE/Ars2 in RNA silencing throughout eukaryotes (8, 9). SE/Ars2 also contributes to other aspects of RNA metabolism, for instance, splicing of pre-mRNA, biogenesis of noncoding RNAs, RNA transport, and RNA stability (19–26). Some of these functions are fulfilled presumably through the interaction with nuclear cap-binding complex (CBC), which consists of two subunits [Cap-binding protein 20 (CBP20) and CBP80] and binds to 7-methylguanylate (m⁷G) caps at the 5′ ends of polymerase II (pol II)–produced transcripts (19, 22, 23). Besides being a key component in RNA metabolism, SE/Ars2 protein acts as a transcriptional factor, regulating expression of transposons (27) and protein-coding genes (28–31).

Given the critical roles of SE/Ars2 in RNA metabolism, regulation of the proteins themselves becomes pivotal to secure transcript accumulation and processing. Recent studies show that SE and possibly Ars2 are intrinsically disordered proteins (IDPs) and are typically sheltered in various macromolecular complexes to perform their functions (32). Unpacked or excessive SE/Ars2 are scavenged by 20S proteasome alpha subunit G1 (PAG1) or possibly 20S proteasome subunit alpha type-3 (PSMA3), the mammalian ortholog of PAG1 (33), key components of 20S core proteasome, for degradation via an ubiquitin-independent pathway. However, the degradome signal that prompts deconstruction of SE/Ars2 has not been yet identified.

Here, we reported that pre-mRNA processing 4 kinase A (PRP4KA), its paralog PRP4KB, and possibly PRP4KC are new bona fide partners of SE. In vitro phosphorylation assays followed by liquid chromatography–tandem mass spectrometry (LC-MS/MS) analysis pinpointed 17 confident phosphorylation residues of SE by PRP4KA in vitro, 8 of which can also be recovered from in vivo phosphorylation-proteomics analysis. Hypophosphorylated but not hyperphosphorylated SE variants could readily rescue se-1 phenotypes in vivo. Moreover, hypophosphorylated SE variants displayed increased binding affinity to HYL1 and were relatively less vulnerable to the activity of 20S proteasome than hyperphosphorylated SE mutants. Unexpectedly, the knockdown mutants of the kinases via artificial miRNAs enhanced the accumulation of hypophosphorylated SE but the improperly accumulated functional forms of SE in the kinase knockdown mutants and wild type interfered with the assembly of SE-scaffolded macromolecule complexes, correspondingly, compromising SE functions and causing the molecular and morphological defects reminiscent of se loss-of-function mutants. Therefore, we concluded that the phosphorylation of SE via PRP4KA, PRP4KB, and PRP4KC (PRP4KA-C) represents a regulatory mechanism to rapidly clear excessive SE and to maintain homeostasis of SE accumulation in vivo to secure its proper functions. Thus, the study revealed a new regulatory layer of miRNA metabolism at a posttranslational level in plants.

RESULTS

Identification of PRP4 kinases as new SE-interacting proteins

We identified PRP4KA (AT3G25840) as a new partner of SE through MS analysis of SE immunoprecipitate (fig. S1A) (17, 27, 32). The native PRP4KA protein is predominantly localized in the nucleus in stable transgenic plants expressing PRP4KA-eYFP under its own promoter, inferring its main function in the nucleus (fig. S1B). To examine whether PRP4KA is a bona fide interactor of SE, we first carried out a split luciferase complementation (LCI) assay (Fig. 1A). In our LCI assays, PRP4KA displayed luciferase (LUC) complementation with SE, as did the positive control of Argonaute 1 (AGO1) with cucumber mosaic virus–encoded 2b (CMV2b) (34), implying that PRP4KA and SE could interact with each other in vivo. We next conducted coimmunoprecipitation (Co-IP) assays with Nicotiana benthamiana that was co-infiltrated with 35S-SE-3HA, together with 35S-Flag-4Myc (FM)–PRP4KA (Fig. 1B). We could readily detect SE in the immunoprecipitants of PRP4KA but not in the control protein flowering locus VE (FVE). This result implies that SE can associate with PRP4KA in planta. We also performed bimolecular fluorescence complementation (BiFC) assays (Fig. 1C). Cotransfected N-terminal fragment of YFP (nYFP)-tagged PRP4KA with C-terminal fragment of YFP (cYFP)-tagged SE into Arabidopsis protoplast displayed clear and punctate foci in the nucleus, reminiscent of D-bodies (35) or splicing bodies (36). Coexpression of nYFP-PRP4KA with cYFP-HYL1 also showed dispersed fluorescence complementation. These results further validated the interaction of PRP4KA and SE, and the process might take place adjacent to HYL1 in vivo. Last, we validated the PRP4KA-SE interaction by yeast two-hybrid (Y2H) assays (Fig. 1, D and E). Further Y2H assays showed that PRP4KA interacted with the C-terminal part of SE (469 to 720 amino acids) (Fig. 1E). However, further truncations of SE abolished its interaction with PRP4KA, inferring that the integral conformation of SE (469 to 720 amino acids) is essential for their association. On the other hand, SE physically interacted with the N-terminal domain of PRP4KA (1 to 149 amino acids) (Fig. 1D).

Fig. 1. Experimental validation of PRP4KA and PRP4KB as new partners of SE by LCI.

(A), Co-IP (B), BiFC (C), and Y2H assays (D and E). (A) The infiltration scheme of leaves shows different combinations of constructs fused to either the N-terminal (nLUC) or C-terminal (cLUC) regions of luciferase (left). AGO1 and CMV2b serve as a positive control (34). CPS, luciferase activity, counts per second. (B) The construct of 35S-SE-3HA was co-infiltrated with 35S-FM-PRP4KA-C in N. benthamiana. IP was conducted by an anti-Myc antibody. Western blot analysis was done using anti-Myc, hemagglutinin (HA), or actin antibodies to detect the indicated proteins in input and IP. FVE and actin serve as negative controls. (C) The 35S-nYFP-PRP4KA construct was cotransfected with 35S-cYFP-SE or 35S-cYFP-HYL1 into Col-0 protoplasts, and the YFP signal indicated the interaction of PRP4KA with SE or HYL1. Scale bars, 10 μm. (D) Y2H shows that SE interacts with the N terminus (1 to 149 amino acids) of PRP4KA. Top: Schematic illustration of PRP4KA variants. U2AF, U2 small nuclear ribonucleoprotein auxiliary factor interaction domain; STKc, catalytic domain of the serine/threonine kinase (bottom). (E) Y2H shows that PRP4KA interacts with the integrative C-terminal region (469 to 720 amino acids) (bottom). Schematic illustration of SE variants. ZnF, zinc finger domain; GAPE, a conserved region enriching Gly, Ala, Pro, and Glu residues (top); LT, Leu and Trp; LTHA, Leu, Trp, His, and Ade.

PRP4KA has two paralogs, PRP4KB and PRP4KC (AT1G13350 and AT3G53640), with the amino acid identities of 54 and 46%, respectively (fig. S1C). We repeated Co-IP, LCI, and Y2H assays and observed that PRP4KB could interact with SE (Fig. 1B and fig. S1, D and E). PRP4KC displayed its interaction with SE in the LCI assay, but not in the Co-IP and Y2H assays (Fig. 1B and fig. S1E). In line with these results, PRP4KC contains an N-terminal domain that is less conserved compared with the ones of PRP4KA and PRP4KB, which harbor the SE interaction interface. Together, we concluded that PRP4KA, PRP4KB, and possibly PRP4KC are the new bona fide partners of SE protein in vivo.

PRP4KA phosphorylates SE in vitro

We next examined whether PRP4KA could phosphorylate SE. An in vitro kinase assay showed that PRP4KA was an active kinase and could conduct autophosphorylation (Fig. 2A). PRP4KA could also phosphorylate SE, but not control protein HYL1 (Fig. 2A). Next, the SE protein, either mock-treated or PRP4KA-phosphorylated, was recovered from SDS–polyacrylamide gel electrophoresis (SDS-PAGE) and subjected to trypsin digestion and LC-MS/MS analysis (Fig. 2B and fig. S2). The mock-treated SE appeared to contain five phosphorylation sites (S76, S185, S187, S295, and S299), presumably resulting from the activity of a kinase(s) from Escherichia coli (Fig. 2D and fig. S2B). However, LC-MS/MS analysis identified 12 confident PRP4KA phosphorylation sites on SE, including S22, S23, S24, S103, Y133, Y181, S291, T294, S324, S571, S669, and S678 (Fig. 2, B and D, and fig. S2A).

Fig. 2. PRP4KA phosphorylates SE in vitro and in vivo.

(A) In vitro kinase assays showed that PRP4KA could directly phosphorylate SE but not HYL1 protein. Coomassie Brilliant Blue staining (CBB) of SDS-PAGE of PRP4KA, SUMO-SE, and HYL1 proteins purified from an insect/baculovirus or E. coli system (left). Self-phosphorylated PRP4KA and phosphorylated SE are indicated with red arrows in autoradiography. (B) Selected phosphorylation peptides identified by LC-MS analysis. m/z, mass/charge ratio. (C) Hypophosphorylation of endogenous SE was detected in prp4k3 compared to Col-0. SE was detected by SDS-PAGE and phosphor-tag gel in parallel using an anti-SE antibody. Actin was a loading control. (D) Eight phosphorylation residues (S22, S23, S24, S76, S291, T294, S295, and S299; marked in yellow) were recovered from both in vitro PRP4K kinase assay and public phosphorproteomics datasets from in vivo. Note that three residues recovered from both in vitro and in vivo (written in blue; S76, S295, and S299) and two in vitro–specific residues (S185 and S187) could be also phosphorylated in E. coli. (E) Schematic diagram of locations of the identified phosphorylation sites in SE protein.

To investigate whether PRP4KA-C are the genuine kinases for SE in vivo, we generated mutants of these genes. Transferred DNA insertion mutants of prp4ka, prp4kb, and prp4kc did not exhibit clear morphological defects except that prp4ka displayed a late flowering phenotype as previously reported (36). We applied CRISPR-cas9 technology to knock out PRP4KA-C genes concurrently. We were unable to recover the homozygotes for all three mutants, inferring that simultaneous loss of all three genes would cause lethal phenotype. We next generated knockdown transgenic lines of PRP4KA-C by expressing two different artificial miRNA constructs concurrently targeting PRP4KA-C (amiR-PRP4KA-C-1) or targeting PRP4KB and PRP4KC in the prp4ka background (amiR-PRP4KB-C-2). Because transformants of the two constructs displayed similar knockdown efficiency and morphological patterns, we selected four amiR-PRP4KA-C-1 lines for further analysis and renamed it as prp4k3 hereinafter. Notably, Western blot assays of SDS-PAGE showed that SE was accumulated in prp4k3 compared to wild type (Fig. 2C, bottom). Furthermore, the SE protein migrated faster in prp4k3 than that in Columbia (Col-0) in a phos-tag gel, implying a decreased phosphorylation level of SE in prp4k3 (Fig. 2C, top). These results suggested that PRP4KA-C could phosphorylate SE protein in vivo.

Earlier, high-throughput phosphoproteomics also revealed that SE could be phosphorylated in plants (2, 37). We mined the public phosphoproteomics datasets (37) and pooled all endogenous phosphorylation sites of SE in vivo detected up to date. Among all 17 possible sites, 8 were recovered from the in vitro PRP4KA assay and thus the focus in the next physiological studies. These residues included S22, S23, S24, S291, and T294, in addition to S76, S295, and S299 that underwent default phosphorylation in E. coli (Fig. 2D). Notably, these phosphorylation sites are not located in the interaction interface between SE and PRP4KA, indicating that PRP4KA targets N-terminal residues while interacting with the C terminus of SE (Fig. 2E).

Hypophosphorylated but not hyperphosphorylated SE variants rescue the se phenotype

To investigate how phosphorylation affects SE function in vivo, we generated a series of hypo-and hyperphosphorylation variants of SE. These variants included S22A, S23A, S24A; S22D, S23D, and S24D; S291A, T294A, S295A, and S299A; and S291D, T294D, S295D, and S299D, which were hereafter renamed as 3A SE^22–24, 3D SE^22–24, 4A SE^291–299, and 4D SE^291–299, respectively. S76A and S76D mutants were also included. These FM-tagged SE variants were driven by its native promoter and were introduced into the se-1 background (Fig. 3, A and B). Numerous complementation lines of the SE variants with representative protein accumulation were screened and maintained for further analyses (Fig. 3C). Both S76A and S76D transformants rescued se-1 efficiently, inferring that the mutations of Ser⁷⁶ might not affect the SE function in vivo (Fig. 3A). By contrast, the hypophosphorylation variants of 4A SE^291–299, but not its phosphor-mimic form (4D SE^291–299), could fully rescue the se-1 developmental defect (Fig. 3A and fig. S3A). Similarly, the 3A SE^22–24 could rescue se developmental defect more efficiently than 3D SE^22–24, but to a lesser extent compared to the 4A SE^291–299 variant above (Fig. 3B and fig. S3A).

Fig. 3. Hypophosphorylated SE variants rescue se phenotypes more efficiently than hyperphosphorylated SE counterparts.

(A and B) Phenotypes of T1 transformants expressing hypo- and hyperphosphorylation variants of SE in the se-1 background. Four-week-old plants were photographed. Note that SE S76A and SE S76D appeared to have similar effect on the complementation of the se developmental phenotype. (C) Western blot analysis using an anti-SE antibody shows accumulation of hypo- and hyperphosphorylation forms of SE in the complementation lines. Actin was a loading control. (D) Quantitative reverse transcription polymerase chain reaction (qRT-PCR) analysis shows that the accumulation of selected miRNA (bottom) and their targets (top) was better restored in the transformants expressing hypophosphorylated SE variants compared to the ones expressing hyperphosphorylated SE mutants. The expression of the miRNAs and target transcripts in the indicated mutants and complementation lines were first normalized to the internal control UBQ10 and then to Col-0 where the value was arbitrarily assigned as 1. The data are presented as means ± SD from three biologically independent replicates (n = 3). Note that the effect of SE S76A and SE S76D on miRNAs and the target accumulation varied among the independent lines.

We next conducted quantitative reverse transcription polymerase chain reaction (qRT-PCR) assays for the four founding miRNAs such as miR166, miR156, miR164, and miR159. Whereas the accumulation of the tested miRNAs was similar in se-1;P_SE-FM-SE S76A and se-1;P_SE-FM-SE S76D lines, the amount was higher in se-1;P_SE-FM-4A SE^291–299 transformants than that in se-1;P_SE-FM-4D SE^291–299 lines (Fig. 3D, bottom). The restoration of miRNA production was also observed in se-1;P_SE-FM-3A SE^22–24 lines and sometimes in se-1;P_SE-FM-3D SE^22–24 transgenic lines as well (Fig. 3D, bottom). We also examined the expression levels of PHB, SPL10, NAC1, and MYB33 transcripts, which are targeted by the tested miRNAs, respectively (Fig. 3D, top). Typically, there was inverse correlation in expression levels of miRNAs and their targeted transcripts. For instance, the steady-state levels of the tested transcripts are comparable in se-1;P_SE-FM-SE S76A and se-1;P_SE-FM-SE S76D, despite variations between the individual lines. By contrast, the levels of the selected miRNA targets in se-1;P_SE-FM-4A SE^291–299 transformants, but not in se-1;P_SE-FM-4D SE^291–299 transgenic lines, were restored to the ones of wild-type plants. This pattern was generally observed between se-1;P_SE-FM-3A SE^22–24 and se-1;P_SE-FM-3D SE^22–24 lines with exceptions of NAC1 and MYB33 (Fig. 3D, top). These results were well in line with the complementation efficiency of se phenotypic defects in the transformants, indicating that the hypophosphorylated, but not hyperphosphorylated SE variants are the functional forms in vivo.

We visited the phenotypes of the hypo- and hyperphosphorylation variant transformants in T2 or T3 higher generations. The phenotype of se-1;P_SE-FM-3A SE^22–24 and se-1;P_SE-FM-4A SE^291–299 complementation lines was stably inherited (Fig. 3B and fig. S3A). The leaf morphology of se-1;P_SE-FM-3D SE^22–24 and se-1;P_SE-FM-4D SE^291–299 transformants became less severe than the shapes of T1 generation, despite that the transgenic plants remained dwarf (Fig. 3, A and B). These results inferred that the transformants of SE variants might have self-compensatory mechanism to progressively revert to its wild-type function in the higher generations.

Phosphorylation affects SE association with HYL1 protein, but not with pri-miRNAs

We assessed whether phosphorylation altered SE binding to pri-miRNAs and to microprocessor components. Electrophoretic mobility shift assays (EMSAs) showed that SE and its hypo- or hyperphosphorylation variants (3A, 3D, S76A, S76D, 4A, and 4D) were bound to pri-miRNAs with comparable dissociation constants (K_d = approximately 30 to 50 nM) (fig. S4, A to E). The results inferred that the phosphorylation status of the residues might not affect the association of SE with pri-miRNAs.

We next performed Y2H assays to examine the interaction of SE variants with HYL1 (Fig. 4, A to C). Considering that there might be potential difference in plasmid transfection efficiency and protein expression levels in individual colonies, we randomly picked up at least 16 colonies for each combination and examined their growth curves, which served as a proxy of the HYL1-SE interaction affinity. Western blot analysis revealed that all hemagglutinin (HA)–tagged hypo- and hyperphosphorylated SE variants were expressed in yeast cells, and so was MYC-tagged HYL1 (Fig. 4C). The colonies of DNA binding domain (BD)–HYL1 and activation domain (AD)–3A SE^22–24 or AD-4A SE^291–299 grew faster than the ones with BD-HYL1 and AD-3D SE^22–24 or AD-4D SE^291–299, respectively, despite the occasional observation of large colonies with the combinations of BD-HYL1 and hypophosphorylated SE variants (Fig. 4A). Further Y2H assays with a series of yeast titers also showed the same result (Fig. 4B). These results imply that the hypophosphorylated SE variants might have stronger interaction with HYL1 than their hyperphosphorylated SE counterparts.

Fig. 4. Hypophosphorylated SE variants displayed stronger binding affinities to HYL1 than hyperphosphorylated SE mutants.

(A) Y2H assays showed that yeast colonies cotransfected with BD-HYL1 and AD-hypophosphorylated SE variants grew bigger than those with BD-HYL1 and AD-hyperphosphorylated SE counterparts. Sixteen colonies that were randomly selected were shown. (B) Yeast growth patterns of representative colonies cotransfected with BD-HYL1 and AD-hypophosphorylated SE variants versus the ones with BD-HYL1 and AD-hyperphosphorylated SE counterparts. (C) Western blot analysis validated the expression of HYL1 and SE proteins in the cotransfected yeast cells. Western blot assays were performed with anti-HA and anti-Myc antibodies to detect HA-tagged SE and Myc-tagged HYL1, respectively. Ponceau staining of Rubisco serves as a loading control. (D) In vitro pull-down assay showed that increased amount of hypophosphorylated SE variants was recovered from HYL1 immunoprecipitant compared to hyperphosphorylated SE mutants. One microgram of HYL1 and His-SUMO–tagged SE was mixed and incubated in room temperature, and HYL1 was immunoprecipitated with an anti-HYL1 antibody. Western blot assays were performed with specific anti-HYL1 and anti-His antibodies to detect the bait HYL1 and the prey SE in input and the IP, respectively. His-SUMO serves as a negative control.

We further conducted in vitro pull-down assays to examine the HYL1-SE interaction. Because SE could not stably bind to the microprocessor components under a physiological salt condition such as 150 mM NaCl (5, 17), we lowered the salt stringency to 75 mM NaCl, which is also maximally allowed for microprocessor activity assay in vitro (5, 17). Notably, SE does bind to HYL1 in vitro under this condition (Fig. 4D and fig. S4F). Furthermore, the amount of the hypophosphorylated SE variants recovered from the HYL1 IP was increased compared to that of the hyperphosphorated SE mutants (Fig. 4D and fig. S4F). Thus, we concluded that the hypophosphorylation of the tested residues of SE increased its binding affinity with microprocessor relative to the hyperphosphorylation of the counterparts.

Phosphorylation destabilizes SE in vivo and in vitro

Differential affinities of SE variants to microprocessor might affect their stabilities and accumulation in vivo. To test this, we adopted a cell-free degradation system using in the complementation lines (32). We prepared cell lysates from 10-day-old seedlings and treated the extracts with cycloheximide (CHX) to block protein synthesis. Without the new protein synthesis and adenosine triphosphate (ATP) supply, the hyperphosphorylated SE (4D SE^291–299) had a half-life of less than 20 min, whereas the hypophosphorylated SE (4A SE^291–299) extended half-life to more than 20 min (Fig. 5A and fig. S5A, left). The addition of carbobenzoxy-leu-leu-leucinal (MG132), a potent peptide aldehyde proteasome inhibitor, to the reaction mixture could largely slow down the degradation progress of 4D SE^291–299 from the complementation lines. MG132 treatment could also inhibit the destruction of 4A SE^291–299 in the complementation lines, but to a much lesser extent compared to 4D SE^291–299 (Fig. 5A and fig. S5A, right). Similar scenarios were observed with SE^22–24 variants (Fig. 5B and fig. S5B), but less clearly with S76 mutants (Fig. 5C and fig. S5C). Thus, we concluded that phosphorylation of SE^291–299 and SE^22–24 triggers SE degradation in vivo.

Fig. 5. Hypophosphorylated SE variants are more resistant to 20S proteasome activity than hyperphosphorylated counterparts.

(A to C) Western blot analysis revealed that hypophosphorylated SE variants had extended half-life than hyperphosphorylated SE forms in cell-free extracts. Total cell lysis prepared from the various SE complementation lines was treated with CHX with or without MG132 at the indicated time points before Western blot assays using an anti-Flag antibody. Actin was a loading control. (D) Western blot analysis showed that SE stability is enhanced in prp4k3 compared to Col-0. Total protein extracts were prepared from the 10-day-old seedlings that were pretreated with CHX, and Western blots were probed with antibodies specifically against indicated proteins. Actin is a loading control. (E) Silver staining of 20S proteasome isolated via immunoprecipitation of FM-tagged PAG1 complex (32). (F to H) In vitro reconstitution assays of protein degradation by 20S proteasome. Recombinant 6xHis-SUMO-SE variants were incubated with purified 20S proteasome, and the reactions were stopped at the indicated time intervals. The numbers below the gels indicate the relative mean signals of SE variants at different time points that were normalized to those of the proteins at time 0, where the value was arbitrarily assigned a value of 1. The data are presented as means ± SD from three biologically independent replicates (n = 3).

We also examined the protein stability of SE in prp4k3. We treated 10-day-old seedlings with CHX to block protein synthesis in planta. Wild-type SE protein had a half-life of approximately 2 hours in Col-0, indicating that SE was readily destroyed in the absence of protein synthesis in planta. However, this degradation was largely inhibited in prp4k3 (Fig. 5D and fig. S5D). These results indicated that hypophosphorylated SE in prp4k3 is less vulnerable to protease activity compared to wild-type SE in Col-0.

SE is an IDP and easily subjected to 20S proteosome–mediated degradation in vitro and in vivo (32). We purified recombinant His-small ubiquitin-like modifier (SUMO)–tagged hypo- and hyperphosphorylated SE variants and conducted in vitro 20S proteasome reconstitution assays according to our recently established protocol (32). The silver staining of immunoprecipitates of FM-tagged PAG1, a subunit of 20S proteasome, showed that Arabidopsis 20S proteasome was recovered successfully, and the patterns of proteasome subunits were identical to those described previously (Fig. 5E) (32). Hyperphosphorylation forms of SE including 3D SE^22–24 and 4D SE^291–299, similar to recombinant SE protein that harbors five E. coli–phosphorylated residues, were quickly degraded, whereas the hypophosphorylation variants of SE such as 3A SE^22–24 and 4A SE^291–299 were relatively more resistant to the 20S proteasome activity (Fig. 5, F to H, and fig. S5E). Notably, similar patterns were also observed when the His-SUMO tag was removed from the recombinant SE proteins, although they were degraded much faster than the His-SUMO–tagged counterparts (fig. S5F). These results further indicated that hypophosphorylated SE is more stable than phosphor-mimic SE in vitro as well. These results also suggested that differential stabilities of hypo- and hyperphosphorylated SE variants should account for contrasted efficiencies of rescuing se-1 in their complementation lines in vivo (Fig. 3 and fig. S3).

Knockdown mutants of PRP4KA-C display pleiotropic developmental defect

Given that the above-mentioned hypophosphorylated SE variants are stable and functional, PRP4KA-C would be expected to be a negative regulator of SE function in vivo. To test this, we carefully reaccessed the developmental and molecular phenotypes of prp4k3 mutants (Fig. 6A and fig. S6, A and B). The prp4k3 lines exhibited developmental abnormalities with varying severities. One group of transgenic lines (type I; ~32 of 100 transformants) displayed the severest growth retardation and frequently reddish leaves, and these plants did not survive through the emergence of a few pairs of true leaves (Fig. 6A). Lines with moderately severe phenotype (e.g., type II plants; ~35% of transformants) displayed spoon-like cotyledons and serrated true leaves. Notably, these lines superficially phenocopied hypomorphic se and ago1 mutants, inferring that the miRNA pathway might be repressed. Lines with mild developmental defects that are represented by type III (24%) had rhomboid cotyledons and slow-growing leaves. The adult mutants of these lines showed wrinkled rosette leaves with delayed flowering time (fig. S6B, top). The rest of transformants (type IV; ~9%) displayed seemly normal growth and development. The phenotypic severities of prp4k3 were inversely correlated to the accumulation of PRP4KA and PRP4KB transcripts, implying that amiR-PRP4KA-C should account for the developmental defects (Fig. 6A and fig. S6B). amiR-PRP4KA-C does not have off-targets (fig. S6H), and thus, the prp4k3 lines are specific knockdown mutants of PRP4KA-C. Types II and III plants survived and were propagated to higher generations for further experiments.

Fig. 6. PRP4KA-C positively regulate SE-mediated RNA metabolism at a genetic level.

(A) Morphological defects of 2-week-old prp4k3 knockdown mutants. (B) Gene Ontology (GO) enrichment analysis of PRP4KA-C and SE-regulated differentially expressed genes (DEGs). The numbers in or adjacent to the pies represent the ratios of genes in each category over the total DEGs. ncRNA, noncoding RNA. (C) Overlapping of up-regulated and down-regulated genes between prp4k3 and se-2 mutants. (D) IGV snapshots of splicing defects of the selected transcripts. The dashed rectangles mark introns with higher retention in se-2 and prp4k3. (E) RT-PCR assay validated the alternative splicing of the selected genes in Col-0 and the indicated mutants. EF-1α serves as an internal control. The red and blue arrows indicate the unspliced and spliced forms, respectively. gDNA, genomic DNA. (F) Small RNA (sRNA) sequencing analysis of miRNA profiling in Col-0 and prp4k3 mutants. miRNAs with expression of at least 1.5-fold higher (prp4k3/Col-0 ≥ 1.5), lower (Col-0/prp4k3 ≥ 1.5), or within <1.5-fold are indicated by red, blue, and gray dots, respectively. (G) Overlapping of up- and down-regulated miRNAs in prp4k3 with SE-dependent miRNAs. (H) sRNA blot analyses of the selected miRNAs in the indicated mutants. U6 is a loading control. (I) Western blot analysis shows increased accumulation of key components of the miRNA pathway in prp4k3 using antibodies specifically against the indicated proteins.

prp4k3 and se mutants display similar transcriptome defect

We next performed RNA sequencing (RNA-seq) analysis of 3-week-old Col-0, se-2, and prp4k3 mutants (Fig. 6, B to D). Expression levels of PRP4KA and PRP4KB transcripts were reduced to about 62 ~ 75% of Col-0 in prp4k3, whereas PRP4KC mRNA was barely detectable (fig. S6E). These data indicated the reliable quality of our RNA-seq data (fig. S6C). RNA-seq analysis showed that the se-2 mutation caused 5602 differentially expressed genes (DEGs) (32). Among these DEGs, 3485 were up-regulated, whereas 2117 were down-regulated. By contrast, approximately 6815 or 5765 genes were either significantly increased or decreased in prp4k3 mutants, respectively (fig. S6C). These high numbers of deregulated loci in prp4k3 mutants likely accounted for their pleotropic developmental defects. Gene Ontology (GO) analysis placed the DEGs observed in prp4k3 into numerous functional categories. The most affected genes (5082 of 12,580; 40.4%) are classified into generic metabolism that includes metabolic processes (26.4%), RNA metabolism (5.9%), and phosphorylation (8.1%) (Fig. 6B). These results underscored the critical roles of PRP4K proteins in regulating RNA molecules and a cascade signaling of phosphorylation. The next two most affected groups are involved in plant responses to stimuli (26.8%) and developmental processes (13.2%). This transcriptome profiling is well in line with physiological unfitness and morphological abnormality of prp4k3. An additional significantly affected group belongs to protein transport and localization, inferring that PRP4KA-C–mediated phosphorylation might directly or indirectly control trafficking and compartmentation of cellular factors (Fig. 6B). GO analysis of DEGs in se-2 revealed that SE-affected genes also belonged to metabolism, RNA processing, posttranslational modifications, and responses to stimulus among others (Fig. 6B).

Further comparative analysis of transcriptome in prp4k3 and se-2 mutants revealed that, among 6815 genes significantly up-regulated in prp4k3, 2364 (34.7%) were also up-regulated in se-2 (Fig. 6C). Conversely, among 3485 genes increased in se-2, 67.8% was also enhanced in prp4k3. The overlap of up-regulated genes between prp4k3 and se-2 mutants is statistically significant [log(P) = −1802.3; hypergeometric test]. Similarly, among the 5765 down-regulated genes in prp4k3, 1647 (28.6%) were also repressed in se-2. In parallel, among 2117 decreased in se-2, 77.8% was also reduced in prp4k3. Down-regulated genes also represent a significant overlap between prp4k3 and se-2 mutants [log(P) = −1747.9; hypergeometric test]. The significantly overlapped DEGs displayed an orchestrated pattern in prp4k3 and se-2 mutants. In line with this observation, only a few DEGs exhibited opposite expression patterns in prp4k3 and se-2 mutants (Fig. 6C). Together, prp4k3 and se-2 mutants displayed comparable transcriptome profiling. These results were astonishing as PRP4KA-C appeared to be a positive regulator, rather than a presumed negative regulator, of SE protein at the genetic level.

As PRP4KA regulates alternative splicing (36), we also assessed whether PRP4KA-C affected SE-mediated pre-mRNA splicing. Previous studies identified 14 splicing retention and 52 alternative splicing events in se. We were unable to recover all these splicing defects in se-2 because of the low-sequencing depth. However, we carefully examined the Integrative Genomics Viewer (IGV) files and manually pinpointed numerous splicing defective transcripts in se-2 (Fig. 6D). These abnormal splicing events were detected in RNA-seq of prp4k3 mutants. Furthermore, this defect could be easily validated by RT-PCR assays, suggesting that PRP4KA-C might control SE accumulation to regulate pre-mRNA splicing (Fig. 6E).

PRP4KA-C affect SE-mediated miRNA production

We next compared small RNA (sRNA) profiles in Col-0, se-2, and prp4k3 mutants. As previously observed, sRNAs have numerous categories including miRNAs, trans-acting siRNAs (ta-siRNAs), and transposable element–associated siRNAs, depending on their originalities and biogenesis pathways (fig. S6D). Comparative analysis did not reveal any obvious change in the sRNA profiling between prp4k3 and Col-0 except marginal but consistent reduction of miRNAs and ta-siRNAs. Of the 366 annotated miRNA species, 134 exhibited at least 1.5-fold decrease, whereas only 16 showed increase in prp4k3 relative to Col-0 (Fig. 6F and table S5). Notably, the down-regulated miRNAs in prp4k3 predominantly overlapped with the ones that depend on SE (Fig. 6G). The sRNA-seq results were readily validated by sRNA blot assays, as the expression of miR159, miR164, and miR166 was generally reduced in the independent transgenic lines of prp4k3, although, to a less extent, relative to se-2 (Fig. 6H). Consistently, the accumulation of the targeted transcripts of the reduced miRNAs was increased in RNA-seq analysis (fig. S6F). Thus, these results suggested that SE-mediated miRNA biogenesis was compromised in the loss-of-function prp4k3 mutants.

We next accessed the protein accumulation of the selected microprocessor and RNA-induced silencing complex (RISC) components. Western blot analysis showed that the amount of AGO1, DCL1, HYL1, and SE was all elevated in prp4k3 compared to that of Col-0 (Fig. 6I). One plausible explanation is that AGO1 and DCL1 are all feedback-regulated by miR168 and miR162, respectively, and the compromised miRNA production could increase the steady-state levels of their targeted transcripts (fig. S6F). Another reason is that a mutation in miRNA pathway often causes deregulation of other components in the pathway (38).

We also examined the expression patterns of MIR159a and MIR164b loci in prp4k3. To the end, we crossed two native promoter-driven GUS reporter lines with prp4k3 and conducted Western blot analysis with F2 segregation lines based on plant statues. Accumulation of a silenced β-glucuronidase (GUS) protein from the pooled samples in prp4k3 was reduced relative to wild-type background for both MIR159a and MIR164b loci. These results indicated that the compromise of miRNA production in prp4k3 mutants, in part, resulted from the reduced transcription of pri-miRNAs, with further suggestion that PRP4KA-C may regulate endogenous gene expression, directly or indirectly at a transcriptional level (fig. S6G).

Accumulation of hypophosphorylated SE variants interferes with its native function

Whereas hypophosphorylated SE is a stable and functional form, the prp4k3 mutant contains accumulated hypophosphorylated SE but displays se-like phenotype. The inconsistency of biochemical and genetic data is reminiscent of the pag1 mutant in which excess amount of SE protein interrupts with its native function and leads to the phenotype of se loss-of-function mutant (32). To examine whether this was the case, we compared SE profiling in prp4k3 and Col-0 (Fig. 7A). Similar to the previous observation (32), size exclusion chromatography (SEC) showed that SE is predominantly distributed from peaks #7 to #8 in Col-0, representing the macromolecular ribonucleoprotein complexes of SE (Fig. 7A, top). By contrast, SE was distributed in a broader range of fractions with the peaks at the fractions of #8 and #9 in prp4k3, which represented mixtures of assembled macromolecular complexes, incomplete or intermediate assemblies (Fig. 7A, bottom). These results indicated that the hypophosphorated variants of SE were quite stable in vivo, but overaccumulated hypophosphorated SE interferes with the assembly of functional macromolecule complexes.

Fig. 7. Excess accumulation of hypophosphorylated SE variants interferes with the native SE function.

(A) SEC revealed different distribution spectra of SE protein in Col-0 and prp4k3 plants. The bands framed in the red dotted boxes are full-length SE proteins. (B) BiFC assays showed that the assembly of SE/DCL1-contained microprocessors was compromised in prp4k3 compared with Col-0. (C) Statistical analysis of numbers of D-body–like foci in Col-0 and prp4k3. The data are presented as means ± SD from n = 10 biologically independent samples (***P < 0.001; unpaired two-tailed Student’s t test). (D) P_SE-FM-SE hypophosphorylation variants more easily caused se-like morphology in Col-0 background versus the SE hyperphosphorylation counterparts. Two-week-old T2 plants are shown. Scale bar, 1 cm. (E) Western blot analysis of endogenous and transgenic SE proteins using an anti-SE antibody. Single and double asterisks indicate endogenous SE and FM-SE variants, respectively. Actin is the loading control. (F) qRT-PCR analysis shows that the introduction of hypophosphorylated SE variants into Col-0 readily caused deregulation of the accumulation of selected miRNA targets versus the hyperphosphorylated SE mutants.

We next accessed whether the overaccumulated hypophosphorylated SE interrupted microprocessor formation. To this end, we cotransfected cYFP-SE and nYFP-DCL1 into the protoplasts of Col-0 and prp4k3 and examined the SE-DCL1 interaction patterns. The complementation of cYFP-SE and nYFP-DCL1 formed numerous fluorescence-brightened foci in the nucleus, reminiscent of previously reported dicing bodies in Col-0 (35). However, the number and fluorescence intensity of dicing body–like foci were substantially reduced in prp4k3 (Fig. 7, B and C). This result indicated that the accumulated hypophosphorylated SE protein did affect the formation of microprocessors, compromising miRNA production. This result also explained the comparable molecular and developmental defects between se and prp4k3 with further suggestion that homeostasis of SE accumulation in vivo is critical for its proper functions, as previously observed in the pag1 mutant (32).

To further examine the critical role of homeostasis of SE accumulation in development, we introduced the hypophosphorylated and hyperphosphorylated SE variants expressed under the SE native promoter into Col-0 background (Fig. 7, D and E, and fig. S7). Approximately 57 and 58% of transgenic lines expressing Col-0;P_SE-FM-3A SE^22–24 or Col-0;P_SE-FM-4A SE^291–299 exhibited developmental defects somewhat mimicking the se loss-of-function and prp4k3 mutants. This result again indicates that an extra dosage of hypophosphorylated SE protein interfered with the assembly of functional SE-engaged complexes, leading to abnormal developmental defect. By contrast, introducing hyperphosphorylated SE variants (3D SE^22–24 and 4D SE^291–299) into Col-0 only caused subtle abnormal phenotypes, compared to the hypophosphorylated counterparts. In line with the morphological defects, the expression levels of a few tested miRNA targets were increased in Col-0;P_SE-FM-3A SE^22–24 and -4A SE^291–299 compared to Col-0;P_SE-FM-3D SE^22–24 and -4D SE^291–299 (Fig. 7F). Notably, the hyperphosphor-mimic forms of SE were detectable but accumulated relatively to a lesser extent compared to the hypophosphomimic SE variants in the transgenic lines (Fig. 7E). These results further indicated that the hypophosphorylated SE mutants were relatively more stable in vivo, and the accumulated endogenous wild-type SE and SE variants in the Col-0 background led to the se-like phenotype.

DISCUSSION

PRP4KA-C phosphorylate SE and repress its function

Here, we reported PRP4KA-C as new partners of SE. These kinases could directly phosphorylate SE protein and mechanistically repress its functions in RNA metabolism. Several lines of evidence support our model (Fig. 8). First, PRP4KA-C have the physical association with SE in multiple independent experiments (Fig. 1 and fig. S1). Second, PRP4KA can phosphorylate 17 residues of SE with confidence in vitro (Fig. 2 and fig. S2). Moreover, at least five of these residues could be retrieved from the high-throughput phosphorproteomics in vivo. Third, the hypophosphorylated SE variants, but not phosphor-mimic forms, could rescue the se phenotype in the complementation assays. Fourth, the hypophosphorylation variants of SE are relatively more stable and readily accumulated compared to the hyperphosphorylation forms in vivo. Last, the hypophosphorylation variants of SE are less vulnerable than the hyperphosphorylation forms to activity of 20S proteosome in vitro. Together, phosphorylation acts as degradome signaling to trigger SE deconstruction in vivo, and, thus, PRP4KA-C negatively regulate SE function at a biochemical level.

Fig. 8. A proposed model for PRP4KA-C that phosphorylate SE protein for degradation to secure its homeostasis and functions in vivo.

SE serves as scaffolds to assemble multiple ribonucleoprotein complexes that participate in transcription and posttranscriptional processes of different species of RNA. Once accumulated, excessive SE can be phosphorylated (shown in red circles) by PRP4KA-C for destruction via 20S proteasome so that the SE function is properly maintained in plants. Note that SE might undergo additional phosphorylation by yet unidentified kinases (shown in green circles), and the related functions remain to be defined.

How does phosphorylation affect SE stability? It has been reported that phosphorylation of the disordered domains can affect local folding and aggregation of IDPs, as well as protein-protein interactions (39, 40). Here, the identified phosphorylation sites are in the intrinsically disordered regions of SE (32); hence, it is possible that the phosphorylation of SE might promote the IDP formation, whereas dephosphorylation promotes the folding of the protein and then its interaction with other cellular components. In line with this note, the hypophosphorylated forms of SE have a stronger binding affinity to HYL1 than the hyperphosphorylated SE variants. Consequently, the hypophosphorylated SE would advance the assembly of ribonucleoprotein complexes such as the microprocessor, securing the functions of the complexes. Under this scenario, hypophosphorylated forms will be more resistant to degradation by the 20S proteasome, whereas hyperphosphorylated SE is easily vulnerable (Fig. 5). We have no reason to exclude other possibilities: For instance, PAG1 or other 20S proteosome subunits might have increased affinities with hyperphosphorylated SE than the hypophosphorylated forms (40). In addition, several splicing factors interact with SE, and it is plausible that phosphorylation might affect SE interaction with the splicing factors, leading to splicing defect (41).

Homeostasis of SE protein is critical for fine-tuning its function in vivo

Given that hypophosphorylated SE is a functional form, the knockdown mutants of kinases harbored enhanced functional SE protein, but, unexpectedly, prp4k3 phenocopied se regarding molecular and developmental defects. The inconsistency is reminiscent of pag1 scenario in which PAG1 guides SE degradation biochemically, whereas the pag1 loss-of-function mutant displays se-like molecular and developmental defects (32). The consensus from the two independent cases is that both excess and lack of SE proteins could interrupt the assembly of SE-scaffolded macromolecular ribonucleoprotein complexes, leading to the se phenotype. These results have been underscored in the SEC and BiFC assays (Fig. 7, A and B). Because of these facts, excessive accumulation of functional SE in prp4k3 or introduction of extra copies of functional SE into Col-0 has similar consequence; but this scenario would not take place with the phosphor-mimic forms of SE because they are easily degraded. These results indicate that the homeostasis of SE needs to be strictly maintained for its proper function in vivo. SE is an IDP and acts in various ribonucleoprotein complexes. It has been noticed that the expression homeostasis is critical for functions of IDPs that form scaffolds or parts of numerous complexes, and any spare or deficiency of IDPs could impair the integrity of the macromolecular complexes and could interfere with their biological functions (42). This scenario also applies to the multifunctional SE protein in Arabidopsis. It has been reported that production of numerous miRNAs is reprogrammed in response to environmental stresses (43), and this process likely entails proper adjustment of protein accumulation of microprocessor components including SE. Under these scenarios, PRP4KA-C could quickly phosphorate excessive SE and channel it to proteasome for clearance. On the other hand, lowered expression of PRP4KA-C would promote SE accumulation. The expression of PRP4KA-C displays various fluctuations under different biotic and abiotic stresses, inferring their surveillance roles on expression homeostasis of SE and other targets. Thus, PRP4KA-C–mediated phosphorylation serves as a regulatory mechanism to fine-tune the SE amount and to maintain its homeostasis in responses to physiological and environmental variations.

Perspective

Up to date, phosphorylation of several components of microprocessor and RISC has been reported in animal systems. In many cases, phosphorylation increases the protein stability of Drosha (44, 45), DGCR8 (45, 46), and transactivation response RNA-binding protein (TRBP) (47, 48) and promote miRNA production. In different contexts, phosphorylation could also cause relocalization and resultant inactivation of Drosha (49) and Dicer (50). Thus, the fact that PRP4KA phosphorylates SE and triggers its degradation via 20S proteasome represents a distinct mechanism in miRNA pathway. Phosphoregulation of SE is reminiscent of its microprocessor partner, HYL1, in Arabidopsis as hypophosphorylated HYL1 is a functional form, whereas phosphorylation leads to the protein destabilization (51, 52). This notwithstanding, several outstanding questions remains unclear. First, it did not escape from our attention that wild-type SE in Col-0 migrates more slowly than the one in prp4k3, and this fact infers that SE might already have some phosphorylation modifications under normal physiological conditions (Fig. 2C). Because wild-type SE harbors 17 phosphorylation residues in vivo, it is likely that different phosphorylation sites have different impacts on SE functions. Among the 17 phosphorylation residues, only 5 have been confidently accredited to PRP4KA; the identities of the bona fide kinases that account for phosphorylation of the rest residues and how the phosphorylation of these residues alters SE function await future investigation. Second, because phosphorylation of SE is crucial to fine-tune its level in vivo, it would be expected that this process should be reversible for precise control of its function in vivo. The phosphatase(s) that dephosphorylates SE and antagonizes PRP4KA-C remains to be identified. Third, it seems that SE, different from its partners, HYL1 and DCL1, which are degraded by cellular proteases (53, 54), is subjective to the regulation through 20S proteasome pathway, and whether SE is controlled by additional protein quality control pathways such as autophagy and some cellular proteases remains unclear. Fourth, Ars2 has been reported to be hypophosphorylated in human cell lines (55); how the phosphorylation status affects the functions of Ars2 awaits clarification. Last, several alleles of prp4k3 have been initially recovered from the genetic screening of splicing defect mutants in splicing. In yeast, PRP4KA has been proposed to participate in splicing regulation. Because SE has been known to be engaged in splicing, it is plausible that PRP4KA regulates splicing through SE. However, it is very likely that PRP4KA might target additional splicing factors, especially that multiple phosphorylation peptides of several splicing factors are recovered from phosphorylation proteomics. This would be another interesting topic.

MATERIALS AND METHODS

Plant materials and growth conditions

Arabidopsis thaliana ecotype Col-0, se-1 (CS3257), se-2 (SAIL_44_G12), hyl1-2 (SALK_064863), dcl1-9 (CS3828), P_MIR159a-FM-GUS (Col-0 background), P_MIR164b-FM-GUS (Col-0 background), and pBA002a-P_PAG1-gPAG1-FM(PAG1-FM) were used for this study as described previously (17, 32). The pBA002a-P_PRP4KA-gPRP4KA-eYFP was transformed into the dcl1-9^+/− background. The pBA-35S-amiR-PRP4K3 transformants (prp4k3) were screened by sRNA blot assay to detect amiR-PRP4K3 and qRT-PCR and by RNA-seq to detect the expression of the target transcripts. A series of hypo- or hyperphosphorylation mutants of SE [pBA002a-P_SE-FM-SE (3A^22–24, 3D^22–24, 4A^291–299, 4D^291–299, S76A, or S76D)] were transformed into Col-0 or se-1 and screened by Western blot analysis. The prp4k3; P_MIR159a-FM-GUS and prp4k3; P_MIR164b-FM-GUS lines were generated from genetic crossing between prp4k3 and P_MIR159a-FM-GUS or P_MIR164b-FM-GUS and screened by Western blot analysis. All plants were grown on soil (Jorry Gardener/LP5) or MS plates in 12-hour light/12-hour dark at 22° ± 1°C.

Vector construction

Most cloned coding sequences (CDSs) and genomic DNA sequences were cloned into the pENTR vector, confirmed by sequencing, and then transferred into the final gateway vectors by atL-attR (LR) recombination reaction. All constructs and related construction primers are listed in tables S1 and S2, respectively.

For plant transformation, the binary vector of pBA002a-P_SE-FM-DC was first generated through the ligation of EcoR V/Xba I–digested SE promotor fragment and pBA002a-FM-DC. Then, various CDSs of SE variants were transferred to the binary vector through LR reactions. The pBA002a-P_SE-FM-SE was described previously (17).

For LCI assays, the tested CDSs were cloned to the pCAMBIA1300-35S-cLUC-DC or pCAMBIA1300-35S-DC-nLUC vector. For the BiFC assays, the tested CDSs were cloned into the pBA-35S-cYFP-DC or pBA-35S-nYFP-DC vector. For the Co-IP assay, the tested CDSs were cloned into the pBA-35S-DC-HA or pBA-35S-FM-DC vector. For Y2H assays, the tested CDSs such as PRP4KA-C and SE mutants in the pENTR vectors were transferred into the pGADT7-DC or pGBKT7-DC vector as described (17).

For protein purification, PRP4KA CDS was digested by Xho I/Sma I and ligated to the pAcGHLT-C vector that was treated by the same enzymes before being transfected into insect cells. The Bam HI/Xho I–digested CDSs of SE variants were ligated to the pET-28a-6xHis-SUMO vector before being transformed into BL21. The construction of wild-type SE and HYL1 expression vectors was described previously (17).

SEC assays

SEC was performed as previously described (17, 56). Ten-day-old Col-0 and prp4k3 mutant seedlings were harvested and then grounded to a fine powder in liquid nitrogen mixed with two volumes of extraction buffer [20 mM tris-HCl (pH 7.5), 300 mM NaCl, 4 mM MgCl₂, 200 μM ZnCl₂, 0.1% Triton X-100, 1% glycerol, 4× EDTA-free protease inhibitor (Roche), 2 mM phenylmethyl sulfonyl fluoride (PMSF), and 15 μM MG132]. The total protein extracts were centrifuged twice at 15,000 rpm for 15 min at 4°C. The supernatant was then filtered through a 0.2-μm filter. Next, the total protein extracts for each sample were loaded onto a Superdex 200 10/300 GL column (GE Healthcare) that was prewashed with the balance buffer [20 mM tris-HCl (pH 7.5), 300 mM NaCl, 4 mM MgCl₂, 200 μM ZnCl₂, 0.1% Triton X-100, 1% glycerol, 1/3× EDTA-free protease inhibitor, 0.5 mM PMSF, and 15 μM MG132]. The running buffer contained 20 mM tris-HCl (pH 7.5), 300 mM NaCl, 4 mM MgCl₂, 200 μM ZnCl₂, 0.1% Triton X-100, 1% glycerol, 1× EDTA-free protease inhibitor, 2 mM PMSF, and 15 μM MG132. Fractions were collected for Western blot analysis using an anti-SE antibody for SE. The Superdex 200 column was also calibrated by the gel filtration standard (Bio-Rad).

LCI assay

The LCI assays were performed as previously described (57). In Fig. 1A and fig. S1D, at least three independent leaves for each combination were agroinfiltrated, and all LCI assays showed similar results. The LCI signal was recorded as counts of luciferase activities per second.

Co-IP assay

For the Co-IP experiments with the N. benthamiana system, all tested constructs were transformed into the Agrobacterium strain GV3101 and then co-infiltrated into 4-week-old N. benthamiana leaves. After two days, leaf samples were collected and were grounded to fine powder in liquid nitrogen. Total protein extracts were prepared by mixing 0.4 g of the ground powder with 1.2 ml of IP buffer [40 mM tris-HCl (pH 8.0), 300 mM KCl, 5 mM MgCl₂, 0.2 mM EDTA (pH 8.0), 5 mM dithiothreitol, 1 mM PMSF, 0.2% Triton X-100, 2% glycerol, and one pellet per 25 ml of complete EDTA-free protease inhibitor, with or without RNase A (0.05 mg/ml) during incubation). The total protein extracts were then centrifuged twice for 15 min at 15,000 rpm at 4°C. The final supernatants were immunoprecipitated with 20 μl of anti-Myc antibody agarose beads at 4°C for 2 hours. Then, the unspecific-bound proteins were removed by three-time wash with the IP buffer. The beads were boiled with 2× SDS loading buffer for Western blot analysis using an anti-Myc or anti-SE antibody.

BiFC assays

Isolation and transfection of Arabidopsis leaf protoplasts from 4-week-old Col-0 and prp4k3 were performed as previously described (38). The constructs of nYFP-PRP4KA-C (fused with N-terminal YFP) were coexpressed with either cYFP-SE or cYFP-HYL1 (SE and HYL1 fused with C-terminal YFP) in protoplasts. Fluorescence signals in the protoplasts were visualized at 12 hours after transfection by Leica SP8 confocal microscopy (YFP fluorescence signal excited at 514 nm and chlorophyll fluorescence signal excited at 633 nm). At least 10 independent protoplasts for each interaction were examined and showed similar results in Figs. 1C and 7B.

Y2H assays

The Y2H assays were performed as previously described (17). All complementary DNAs (cDNAs) were cloned into the pGADT7-DC and pGBKT7-DC by LR reaction and subsequently transfected into yeast for examination of the possible protein-protein interaction. AD and BD refer to galactose-responsive transcription factor activation domain and DNA binding domain, respectively, in Figs. 1 (D and E) and 4 (A and B). In these Y2H assays, a combination of AD-HYL1 and BD-SE was typically used as a positive control, whereas combinations of AD/BD vectors served as negative controls. The yeast colonies were photographed 2 or 3 days after transfection. For the Y2H assays to examine the interaction of HYL1 and various SE mutants in Fig. 4 (A and B), at least 16 independent colonies for each combination were tested, and all showed similar results.

Confocal microscopy

YFP signal was detected from the root tips of 7-day-old T1 dcl1–9/+; P_PRP4KA-gPRP4KA-eYFP transgenic plants. PRP4KA localization was imaged on a Nikon D-ECLIPSE C1si confocal laser scanning microscope.

In vitro phosphorylation assay and LC-MS/MS

GST-6xHis-PRP4KA protein was purified from a baculovirus/insect expression system. The His-SUMO-SE and HYL1 (without His-SUMO tag) protein were purified from E. coli BL21.

In vitro phosphorylation assay was carried out by diluting and mixing 1 μg of GST-6xHis-PRP4KA and 3 μg of His-SUMO-SE or HYL1 with H₂O into a volume of 15 μl. The mixed proteins were then added into 15 μl of 2× phosphorylation buffer [20 mM tris-HCl (pH 7.5), 10 mM MgCl₂, 100 mM NaCl, 5 mM EDTA (pH 8.0), 1 mM dithiothreitol (DTT), 100 μM ATP, and 0.5 to 2 μl of [γ-32 P] ATP of each sample]. The reactions were incubated for 4 hours at room temperature before the termination with the addition of 7.5 μl of 5× SDS loading buffer. The samples were denatured at 95°C for 5 min and resolved by SDS-PAGE gel, followed by the detection by autoradiography.

For the MS assay, the in vitro phosphorylation assays were performed in parallel but with cold ATP. The reaction was loaded onto SDS gels. After separation, the gel was stained with Bio-Safe Coomassie.

The gel pieces were trypsin-digested overnight at 37°C after reduction by DTT at 55°C for 45-min shaking and alkylation by CAA (chloroacetamide) in the dark for 1 hour. In-gel trypsin digestion was conducted at 37°C overnight. The digested peptides were extracted with acetonitrile (ACN) and SpeedVac to dryness. Phosphopeptides were enriched using a glygen NuTip TiO₂ + ZrO₂ (part no. NT2TIZR). After enrichment, 5 μl of formic acid was added to preserve the phosphopeptides, and the samples were speed vac again. The phospho-enriched peptides were resuspended in 0.1% formic acid in water for LC-MS/MS analysis as previously described (58). Briefly, the peptide samples were loaded onto an Acclaim PepMap 100 C18 precolumn (20 mm by 75 μm; 3 μm) and separated on a PepMap RSLC C18 analytical column (250 mm by 75 μm; 2 μm) at a flow rate at 300 nl/min through a linear gradient from solvent A [0.1% formic acid (v/v)] to 35% solvent B (0.1% formic acid and 99.9% ACN) in 110 min and to 98% solvent B for additional 10 min. The mass spectrometer used was an Orbitrap Fusiong Tribrid system with a collision-induced dissociation (CID) and electron transfer dissociation (ETD) decision tree method. The Orbitrap MS1 scan range was 350 to 1800 mass/charge ratio, the automatic gain control target was set to 400,000, and the maximum inject time was set to 50 ms. The MS/MS spectra were acquired in the linear ion trap after CID for two to four charges and/or ETD for three to eight charges. The threshold for precursor ion selection was 5000 counts, and selected parent ions were isolated using a mass window of 1.3.

The raw LC-MS/MS data were processed to search against the Arabidopsis TAIR10 database (32,785 entries; www.arabidopsis.org/) by using the Mascot software (version 2.4; Matrix Science Inc.). The parameters are as follows: precursor mass tolerance at 10 parts per million, fragment mass tolerance at 0.8 Da, and trypsin as the enzyme, allowing two missed cleavages, oxidation (M) and phosphorylation (S, T, and Y), as dynamic modifications. Confident database matching spectra were manually inspected for evidence of phosphorylation.

Phos-tag analysis

Total protein extracts were prepared from the 10-day-old plants of Col-0 and prp4k3. Briefly, 0.1 g of tissue was grounded into 1 ml of 10% (w/v) trichloroacetic acid (TCA)/acetone to precipitation for 30 min at 4°C. Then, the precipitated protein was boiled with 2× SDS loading buffer for the phos-tag (Nard Institute Ltd., AAL-107M) analysis following the Phos-tag SDS-PAGE Guidebook (https://labchem-wako-pages.fujifilm.com/US-Phostag-Catalog-Download.html).

Western blot assays

Western blot analysis was typically performed with 10-day-old plants that were grown on an MS plate or 3-week-old plants that were grown on soil. The extraction and experimental procedures were performed as described (59). The primary antibodies were used against Myc (Sigma-Aldrich, C 3956), HA (Sigma-Aldrich, H9658), actin (Sigma-Aldrich, A0480), His (Sigma-Aldrich, H1029), SE (Agrisera, AS09 532A), DCL1 (Agrisera, AS12 2102), AGO1 (Agrisera, AS09 527), and HYL1 [from S. W. Yang’s laboratory (53)]. Secondary antibodies used were goat-developed anti-mouse immunoglobulin G (GE Healthcare, NA931) and anti-rabbit (GE Healthcare, NA934).

For Western blot analysis of yeast extracts, an amount of 10⁵ yeast cells was harvested, resuspended in 200 μl of lysis buffer [0.1 M NaOH, 0.05 M EDTA (pH 8.0), 2% SDS, and 2% β-mercaptoethanol], and boiled in 95°C for 10 min. The cell resuspension was supplied with 5 μl of 4 M acetic acid into and boiled in 95°C for 10 min. Then, the cell resuspension was mixed with 50 μl of loading buffer [0.25 M tris-HCl (pH 6.8), 50% glycerol, and 0.05% bromophenolblue).

sRNA blot assays

The sRNA blot assays from 10-day-old seedlings were performed as previously described (59). The sequences of oligo probes used for sRNA blot assays are listed in table S2.

RT-PCR and qRT-PCR

Total RNA was extracted by TRIzol (Sigma-Aldrich) from 10-day-old seedlings and then treated with deoxyribonuclease I (DNase I; Sigma-Aldrich, AMPD1). The cleaned RNA was reverse-transcribed with SuperScript III reverse transcriptase (Invitrogen, catalog no. 18080093) with an oligo(dT) primer according to the manufacturer’s instructions. Through experiments, EF-1α or UBQ10 genes were used as internal controls. For Figs. 3D and 7F and fig. S6B, the data are presented as means ± SD from three biologically independent replicates (n = 3). For the regular RT-PCR analysis, the PCR products were fractioned on agarose gels. qRT-PCR was performed using the CFX384 Touch Real-Time PCR Detection System (Bio-Rad). The primers used for the assays are listed in table S2.

RNA-seq and sRNA-seq

Total RNA was extracted with the TRI Reagent (Sigma-Aldrich, T9424) from 3-week-old soil-grown plants. The library preparation for Illumina sequencing and bioinformatic analysis were performed as previously described (27, 60).

GUS staining

F2 seedings of P_MIR159a-FM-GUS and P_MIR164b-FM-GUS in Col-0 and prp4k3 mutant background were used in this experiment. Fourteen-day-old seedings of F2 were harvested and used for GUS staining as previously described (17).

Expression and purification of recombinant proteins

The GST-6xHis-PRP4KA protein was expressed in a baculovirus (BD Biosciences, catalog no. 554740)/insect cell (BD Biosciences, catalog no. 554738) expression system. For expression and purification of PRP4KA, pAcGHLT-C-GST-6xHis-PRP4KA and BaculoGold baculovirus DNA were cotransfected into sf9 insect cells to generate the recombinant baculovirus. The recombinant viruses were amplified for two rounds and then the P3 virus was added to 2.5 × 10⁶ sf9 insect cells/ml for propagation and cultured for 60 hours before the collection for large-scale protein expression. The cell pellet was resuspended in lysis buffer [40 mM tris-HCl (pH 7.5), 500 mM NaCl, 5% glycerol, 1 mM β-mercaptoethanol, 1 mM PMSF, 0.1% Triton X-100, and one pellet per 50 ml of EDTA-free protease inhibitor) and disrupted with a high-pressure homogenizer (AVESTIN, catalog no. EF-C3). The disrupted cell pellet was centrifugated at 18,000 rpm for 15 min at 4°C and then was filtrated with 0.4 μm of membrane. The cleared lysate was supplemented and loaded on a HisTrap HP column (GE Healthcare, catalog no. 17-5248-02). The recombinant GST-6xHis-PRP4KA protein were dialyzed in a dialysis buffer [40 mM tris-HCl (pH 8.0), 100 mM NaCl, 10 mM MgCl₂, 2 mM β-mercaptoethanol, 2 mM EDTA (pH 8.0), and 50% glycerol) at 4°C overnight, and the final purified protein was aliquoted for storage at −80°C.

For expression of recombinant proteins in E. coli, the recombinant SE and its variants or HYL1 were transformed into BL21 DE3 cells, grown in Luria-Bertani (LB) liquid medium at 37°C until OD₆₀₀ (optical density at 600 nm) = 0.6 and then supplied with 0.5 mM isopropyl-β-d-thiogalactopyranoside. The protein expression was induced at 16°C overnight before the pellet collection. For the protein purification, the cell pellet was resuspended in lysis buffers [SE and its variants: 20 mM tris-HCl buffer (pH 8.5), 500 mM KCl, 1 mM β-mercaptoethanol, 2 mM PMSF, 2% glycerol, 0.1% Triton X-100, and one pellet per 50 ml of EDTA-free protease inhibitor; HYL1: 40 mM tris-HCl buffer (pH 8.0), 300 mM KCl, 1 mM β-mercaptoethanol, 1 mM PMSF, 2% glycerol, 1% Triton X-100, and one pellet per 50 ml of EDTA-free protease inhibitor]. The resuspended culture was homogenized with a high-pressure homogenizer. The disrupted cell pellet was centrifugated at 18,000 rpm for 15 min at 4°C and then was filtrated with 0.4 μm of membrane. The cleared lysate was supplemented and loaded on a HisTrap HP column. The recombinant His-SUMO-SE/SE variant protein was dialyzed in dialysis buffer [20 mM tris-HCl (pH 8.5), 150 mM KCl, 5 mM β-mercaptoethanol, and 50% glycerol] at 4°C overnight, and the final purified protein was aliquoted for storage at −80°C. The recombinant His-SUMO-HYL1 protein fractions from the HisTrap column were pooled and treated with SUMO protease at 4°C overnight for the removal of the His-SUMO tag. The fractions were concentrated by 50-kDa molecular weight cut-off centricon (Millipore) before the gel filtration assays with HiLoad 16/600 Superdex 200-pg column (GE Healthcare). The recombinant HYL1 protein was finally supplemented in 50% glycerol and stored at −80°C, frozen by liquid nitrogen.

Electrophoretic mobility shift assay

EMSA was performed as described with modifications (17). Recombinant proteins were mixed in the EMSA buffer [2 mM tris-HCl (pH 7.5), 2 mM MgCl₂, 2 mM DTT, 0.3% NP-40, and the SUPERase-In RNase Inhibitor (1 U/μl; Thermo Fisher Scientific)]. The mixture was incubated on ice for 30 min before γ-32 P–labeled RNA was added. The bound ribonucleoprotein complexes were resolved on a native agarose gel. The gel was incubated in the fixation buffer (40% ethanol, 10% acetic acid, and 5% glycerol) for 15 min and subsequently dried at 80°C for 2.5 hours before visualization via radiography. The K_d was calculated using the Prism GraphPad 8 software fit with a Hill slope model.

In vitro 20S proteasome decay assay

The 20S proteasome purification and in vitro 20S proteasome decay assays were performed as previously described (32). The activity of the purified proteasome was first tested with the substrate succinyl-Leu-Leu-Val-Tyr-7-amido-4-methylcoumarin (Suc-LLVY-AMC) (Sigma-Aldrich, S6510) as previously described (32).

In vivo CHX decay assay and chemical treatments

For the CHX decay assay, Col-0 and prp4k3 plants were germinated and grown on solid MS medium for 10 days before being transferred to liquid MS medium supplemented with the indicated concentrations of CHX (Sigma-Aldrich, C1988). The samples were treated for 15 min under vacuum and then incubated at room temperature for the indicated times (0, 1, 2, 4, and 6 hours) before Western blot analysis.

In vitro cell-free decay assay

The in vitro cell-free decay assay was carried out as previously described with modifications (32). Ten-day-old seedlings of Col-0 and prp4k3 plants were ground to a fine powder in liquid nitrogen, mixed with two volumes of lysis buffer [25 mM tris-HCl (pH 7.5), 10 mM NaCl, 10 mM MgCl₂, and 10% glycerol]. The total protein extracts were centrifuged twice at 4°C for 10 min at 15,000 rpm and then adjusted to equal concentrations with the lysis buffer. The final supernatant was supplemented with 0.5 mM CHX, and the mixtures were then divided into two parts. One aliquot was added with 50 μM MG132 (Calbiochem, 474787), whereas the other was treated with 2% dimethyl sulfoxide as a control. The mixtures were then incubated at 22°C for the indicated times (0, 10, 20, 30, 45, 60, and 75 min) described in each experiment before Western blot analysis.

In vitro pull-down assay

One microgram of recombinant proteins (HYL1, His-SUMO or His-SUMO-tagged SE, or SE variants) was supplied with 1 ml of binding buffer [40 mM tris-HCl (pH 7.4), 0.1% NP-40, 75 mM NaCl, and one pellet per 50 ml of complete EDTA-free protease inhibitor]. The mixtures were incubated by gentle rotation for 1.5 hours at room temperature before the addition of 3 μl of anti-HYL1 and 30 μl of agarose A beads. The reactions were then rotated for 1 hour at room temperature. After the incubation, the agarose beads were washed three times with the binding buffer before the addition of 40 μl of 2× SDS buffer for Western blot analysis.

5′ Rapid amplification of cDNA ends

The RNA was isolated from a 10-day-old T3 prp4k3 mutant. The amount of 5 μg of DNase-treated total RNA was ligated to 100 pmol of 5′ adapter (rGrUrUrCrArGrArGrUrUrCrUrArCrArGrUrCrCrGrArCrGrArUrC). After the ligation, first-strand cDNAs were synthesized using the Superscript III Reverse Transcriptase (Invitrogen, CA) and a random primer. The cDNA was further amplified using the primers listed in table S2.

Model diagram

The model diagram was created with BioRender.com.

Supplementary Materials

This PDF file includes:

Figs. S1 to S7

Other Supplementary Material for this manuscript includes the following:

Tables S1 to S5

View/request a protocol for this paper from Bio-protocol.