m⁶A modification plays an integral role in mRNA stability and translation during pattern-triggered immunity

To determine whether the m⁶A modification could affect plant defense against pathogens, we characterized the fip37-4 mutant, a developmentally viable partial knockdown mutant of m⁶A modification (5), in response to pathogen challenge. We first examined the basal resistance of fip37-4 plants to a natural oomycete pathogen of Arabidopsis, Hyaloperonospora arabidopsidis (Hpa) Noco2 (3-5 × 10⁴ spores/mL), and a bacterial pathogen, Pseudomonas syringae pv maculicola (Psm) ES4326 (OD_600nm = 0.0001). As a negative control for the bacterial infection experiment, a mutant with enhanced disease susceptibility, nonexpresser of pathogenesis-related genes 1 (npr1), was included. These infection assays revealed significantly increased growth of Hpa Noco2 and Psm ES4326 in the fip37-4 mutant compared to wild-type (WT) plants, and this enhanced susceptibility was rescued in the gFIP37-GFP/fip37-4 transgenic complementation plants (5) (Fig. 1 A and B). Since the m⁶A modification is present in thousands of mRNAs even in the absence of pathogen challenge, and the transcriptome-wide m⁶A abundance is reduced in fip37-4 (5), it is likely that the mutant has distinctly different basal transcriptome from that of the WT due to a high level of genetic interference both during development and at homeostasis. In order to more discretely observe the effects of m⁶A modification on plant immunity, we generated dexamethasone (DEX)-inducible silencing lines against the MTA (DEX:siMTA), the catalytic component of the m⁶A writer complex (12), to inhibit m⁶A modification only after the plants have developed normally and reached maturity. We found that without DEX treatment, the two independent DEX:siMTA lines showed bacterial growth at levels similar to that of the DEX:YFP control plants. However, after DEX treatment for 24 h, which reduced MTA mRNA levels to 50%, bacteria grew to higher levels, similar to fip37-4 and npr1 (Fig. 1 C and D). These results confirm the positive role of FIP37 and MTA in plant defense and suggest a general requirement for the m⁶A machinery in basal immunity.

To further distinguish a direct involvement of m⁶A modification in plant defense from a pleiotropic effect, we examined the responsiveness of the fip37-4 mutant plants to MAMP by first infiltrating plants with 1 µM of elf18, an epitope of bacterial EF-Tu, followed by inoculation with Psm ES4326 (OD_600nm = 0.001 which is 10-fold higher than that used for detecting a defect in basal resistance), with the elf18 receptor mutant, efr-1, as the negative control. We found that the protection against Psm ES4326 conferred by the elf18-pretreatment was significantly compromised in the fip37-4 mutant. Complementation of the mutant phenotype was achieved in the gFIP37-GFP/fip37-4 line (Fig. 1E). Consistently, the two independent DEX:siMTA lines also showed a significant loss of elf18-mediated protection against Psm ES4326 infection after the DEX treatment (Fig. 1F), demonstrating that deposition of the m⁶A modification is needed for establishing PTI.

Based on liquid chromatography–tandem mass spectrometry (LC-MS/MS) analysis, we found that in response to elf18 treatment, no discernable alteration in the total m⁶A abundance was detected in mRNA (SI Appendix, Fig. S1A). To further investigate the dynamics of m⁶A upon PTI induction, we performed global m⁶A-sequencing assays of mock- and elf18-treated WT plants. While m⁶A modifications were detected along the coding sequences (CDS) of mRNAs, there was a preference for the stop codon, consistent with previous reports (5, 12, 24), in both mock- and elf18-treated samples (SI Appendix, Fig. S1 B and C). The canonical m⁶A consensus, RRACH, was significantly overrepresented in the modified regions when compared to a randomized background (SI Appendix, Fig. S1D).

Despite the lack of an overall change in the total m⁶A abundance in response to elf18, we found 623 modifications in 518 mRNAs that were specific to the mock-treated plants, while 614 modifications in 459 mRNAs that were specific to the elf18-treated treatment, in addition to the 5008 shared modifications in 4098 transcripts (SI Appendix, Fig. S1E and Dataset S1). Among the 459 elf18-specific genes, 126 were found to be transcriptionally induced upon elf18 treatment, and the GO terms for the elf18-specific transcripts were enriched with “response to biotic stimulus,” such as immune regulators EDS5, XLG3, and WRKY27, suggesting that m⁶A may have a role in the newly synthesized defense-related transcripts during PTI (SI Appendix, Fig. S1F).

To examine whether specific cis-elements are involved in regulating the deposition of m⁶A in response to elf18, we searched for enriched sequences in both the m⁶A modified regions and the 150 nucleotides preceding and following the modification sites. Interestingly, clear consensus sequences were detected in the flanking regions, with a degree of overlap between mock and elf18-treated samples such as “GAAGAAGA” (SI Appendix, Fig. S1G). Therefore, condition-specific m⁶A modifications detected in our sequencing experiments likely result from a combination of transcript level changes and some specific cis-elements.

After establishing the dynamics of m⁶A modification in the transcriptome upon immune induction, we next tested our hypothesis that binding of the m⁶A reader proteins to modified- mRNAs plays a role in the dynamic posttranscriptional regulation of gene expression during PTI. We characterized the response of mutants of the putative m⁶A readers, which are proteins containing a known m⁶A-binding YT521-B homology (YTH) domain. In Arabidopsis, 11 proteins have such YTH domains in the evolutionarily conserved C terminus (ECT1-11) (7). When we tested ect single mutants, we did not detect any significant immune phenotypes (SI Appendix, Fig. S2), potentially due to the functional redundancy of these ECTs. Indeed, when we subsequently generated high-order mutants based on their homology (7) through traditional crossing or CRISPR/Cas9, we found that both the ect2/ect3 (ect2/3) double and the ect2/ect3/ect4 (ect2/3/4) triple mutants from the YTHDFA clade showed a clear defect in PTI (Fig. 2A). Since these readers are known to be cytosolic localized (14, 25), their immune-deficient phenotype indicates that active recognition and processing of m⁶A mRNA by these proteins is necessary for PTI.

ECT2, ECT3, and ECT4 have been reported to function redundantly during plant development (8), and targets of ECT2 and ECT3 have significant overlap (14). To capture the ECT2-bound transcripts during PTI, we utilized the transgenic complementation plants gECT2-YFP/ect2 to perform formaldehyde fixation and cross-linking immunoprecipitation-sequencing (FA-CLIP-Seq) with 35S:GFP as a control under mock or elf18-treatment (Dataset S2). Consistent with previous findings (9), we found that ECT2 was mainly bound to translation stop site and 3′ UTRs, and elf18-treatment did not affect the overall pattern of the ECT2 binding (Fig. 2 B and C). Among the enriched peaks, UGUA is the most representative consensus sequence in the center of the peaks (Fig. 2D), as reported previously (9). Meanwhile, we compared the target genes identified under both mock and elf18-treatment. Only 131 transcripts were mock-specific, and 278 transcripts were elf18-specific, with the defense-related GO terms enriched in the pool of elf18-specific mRNAs (Fig. 2 E and F). However, most of the targets were bound by ECT2 under both conditions, consistent with the observations of the global m⁶A-seq.

To further investigate how reader proteins in the YTHDFA clade, ECT1-4, regulate the fate of mRNAs, we utilized a MS2-tethering system (26–28) to directly test the effects of ECT protein binding to target mRNAs (Fig. 2G). Protein of interest (POI), i.e., the ECT protein or the YFP control was fused to the C-terminal of the MS2 coat protein (MCP), whereas tandem MS2-binding sequences (MBS) were inserted into the 3′ UTR of the Firefly Luciferase (FLUC) mRNA where m⁶A modification and ECT2 binding were preferentially detected (Fig. 2 B and C). As an internal control, Renilla Luciferase (RLUC) was constitutively coexpressed. The resulting effectors and reporters were transiently coexpressed in Nicotiana benthamiana for 2 d. We found that ECT1 promoted the degradation of the FLUC mRNA with a similar level of reduction in the FLUC protein produced (Fig. 2 H and I), which is consistent with the recent finding on the ect1 mutant in Arabidopsis (16). In contrast, tethering ECT3 to the FLUC mRNA enhanced the accumulation of the transcript with the corresponding increase in protein production (Fig. 2 H and I), indicating that ECT3-binding has the opposite effect on mRNA stability as ECT1. Distinct from ECT3, ECT2-binding increased the FLUC protein levels without significantly impacting the mRNA stability. These results indicate that in the ect2/3 mutant, both the stability of the m⁶A-modified mRNAs and their translation were negatively affected. In this assay, we did not capture any detectable activity of ECT4. However, ect2/3/4 has even a stronger development phenotype than ect2/3 (8, 29) and a deficiency in PTI as ect2/3 (Fig. 2A) suggesting ECT4 may have similar functions as ECT2 and ECT3. Therefore, we used the ect2/3/4 mutant for further analyses.

To further elucidate the function of m⁶A during PTI, we performed global measurements of mRNA decay in WT, fip37-4, and ect2/3/4 seedlings in response to mock or elf18 induction, followed by treatment of a cocktail of transcriptional inhibitors (cordycepin 150 µg/mL, actinomycin 10 µM) for 0.5 h prior to sample collection. The abundance of mRNA at Time 0 and at subsequent intervals of 1, 2, and 4 h was measured by QuantSeq (Fig. 3A). The decay rates of individual transcripts across different genotypes and conditions were determined using a previously reported mathematical modeling approach (30) (Dataset S3).

For comparison, we selected transcripts modified by m⁶A and bound by ECT2 (“m⁶A&ECT2 targets”) as the highly confident targets, and those without m⁶A or ECT2 binding (“w/o m⁶A or ECT2 binding”) as the control based on the m⁶A-IP-seq and FA-CLIP-seq data. Under mock condition, we observed that m⁶A-modified ECT2-targets were relatively less stable compared with nontargets in both fip37-4 and ect2/3/4 mutants, indicating m⁶A modification and ECT2/3 binding stabilize mRNA (Fig. 3 B and C). This observation aligns with previous findings suggesting that ECT2/3 can promote RNA stability (14, 15). Interestingly, under elf18-induced conditions, a similar result was only observed in ect2/3/4, not in the fip37-4 mutant (Fig. 3 D and E), suggesting that during PTI, the m⁶A functions may not be fully determined by ECT2/3/4. Other reader proteins, such as ECT1 (16), which destabilizes mRNA (Fig. 2H) may also be involved.

Indeed, when we focused on transcripts that are transcriptionally induced by elf18 (“elf18-inducible”), we found, surprisingly, that they intrinsically had relatively shorter overall half-life compared to the noninducible ones under mock conditions, and the difference between the two groups became more pronounced upon elf18 treatment (Fig. 3F). This accelerated degradation observed in the elf18-inducible transcripts suggests the need for rapid turnover of these defense-related mRNAs, perhaps to avoid unnecessary growth inhibition (31). The role that m⁶A plays in this accelerated turnover of elf18-inducible mRNAs was further demonstrated by their enhanced relative stability in the writer complex mutant fip37-4 compared to that in the WT background (Fig. 3G). Interestingly, this phenomenon was not observed in the ect2/3/4 mutant (Fig. 3H), indicating that these three readers are not involved in the turnover of these elf18-inducible mRNAs. It is conceivable that other readers, such as ECT1, facilitate the RNA degradation during immune responses.

The longer half-life of immune-inducible mRNAs in fip37-4 (Fig. 3G) contradicts the PTI-deficient phenotype observed in this m⁶A writer mutant. The discrepancy suggests that this immune response must have additional layers of regulation by m⁶A modification. Besides mRNA stability, which does not seem to correlate with the mutant phenotype, we next considered the effect of m⁶A modification on translation because our MS2-tethering assays showed that ECT2 could enhance the translation efficiency of the target mRNA (Fig. 2 H and I) and a possible role for m⁶A modification in mediating translational activation during plant immune induction has not been previously examined. To fill this knowledge gap, we conducted polysome profiling on WT, fip37-4, and ect2/3/4 seedlings exposed to either mock- or elf18-treatment for 1 h. Subsequently, we assessed translation efficiency (TE) by determining the ratio of mRNAs associated with polysomes to their total mRNA levels through QuantSeq (Dataset S4).

Calculations of TE revealed that, under mock treatment, WT plants and the fip37-4 or ect2/3/4 mutants had no detectable overall TE difference (SI Appendix, Fig. S3A), whereas elf18 treatment led to a slight reduction in the overall TE in the mutants (SI Appendix, Fig. S3B). This reduction suggests a potential role for m⁶A and reader proteins in the efficient translation of defense-related transcripts during this immune response. For further analysis, we once again focused on comparisons between transcripts with m⁶A modification and ECT2-binding and transcripts with neither m⁶A modification nor ECT2-binding as the control. The cumulative curves, representing changes in TE between the mutants and WT, showed that m⁶A-modified ECT2-targets had lower TE values in the ect2/3/4 mutant compared with the control transcripts under mock conditions, and a dramatic decrease in TE was observed in both fip37-4 and ect2/3/4 mutants upon elf18 treatment (Fig. 4 A–D). The only shared GO term enriched in the top 10% transcripts with the most impaired TE between fip37-4 and ect2/3/4 was “response to external stimulus” (SI Appendix, Fig. S4 A and B). This suggests that the translation of defense-related transcripts is defective without m⁶A modification or recognition. Indeed, while elf18-induced transcriptomes for WT and the mutants showed high levels of correlation (SI Appendix, Fig. S5), elf18-inducible transcripts were translated less efficiently in both fip37-4 and ect2/3/4 mutants upon elf18-treatment (Fig. 4 E–G). This indicates that the m⁶A writing and reading are crucial for rapidly reshaping of the transcriptome and translatome to facilitate the transient switch from growth to defense. The compromised translational response may underlie the observed deficiency in basal defense and PTI in the m⁶A writer and reader mutants.

To more clearly show how m⁶A-mediated mRNA stability and translation collectively shape the plant immune response, we made scatterplots to illustrate changes in RNA stability and TE between WT and the fip37-4 mutant for the elf18-inducible transcripts. Comparing the elf18 treatment with mock, it is evident that the elf18-inducible transcripts exhibited extended half-lives in the m⁶A-deficient fip37-4 (x axes in Fig. 5 A and B), yet showed lower TEs, compared to the WT (y axes in Fig. 5 A and B). These plots indicate that the dual function of m⁶A in PTI, destabilizing immune-induced mRNAs while promoting their TE, results in a pulsatile production of immune-associated proteins and enhanced disease resistance (Fig. 5C).

Arabidopsis thaliana and Nicotiana benthamiana were grown in soil under 12/12-h light/dark cycles at 22 °C with 55% relative humidity. All Arabidopsis lines used in this study are of the Columbia-0 accession. The fip37-4 mutant (SALK_018636) and the ect single mutants were obtained from the Arabidopsis Biological Resource Center and listed in Dataset S5. The gFIP37-GFP/fip37-4 transgenic line was provided by Hao Yu from the National University of Singapore (5). Previously published lines are npr1-2 (42), efr-1 (43), and DEX:YFP (44). The ect2/3 mutant was generated by crossing ect2 (SALK_002225C) with ect3 (SALKseq_077502) and the ect2/3/4 triple mutant was generated by CRISPR/Cas9 knock-out using a previous described method (45) with the guide RNA (gRNA) sequences shown in Dataset S5.

To generate DEX:siMTA, primers P1/P2 (sequences for all primers used are listed in Dataset S6) were used to amplify the 332 bp fragment corresponding to the CDS of MTA to generate the PCR product 1 (LIC1-MTA-LIC2) with additional flanking sequences LIC1 and LIC2. PCR product 2 (LIC4-LIC1-MTA-LIC2-LIC3) was amplified by primers P3/P4 using LIC1-MTA-LIC2 as a template to add another two flanking sequences LIC3 and LIC4. Both products containing the same sequence were inserted into pRNAi-LIC in the opposite orientation as described previously to generate the hairpin RNA (hpRNA) cassette (46). The hpRNA cassette was amplified using primers J1/J2 and ligated into pBAV154 via XboI/SpeI to build the construct pDEX:siMTA. The floral dip method was used to generate transgenic plants (47).

For generating the ECT CRISPR mutants, egg cell-specific CRISPR/Cas9 vector pHEE401E was used to obtain the ect2/3/4 mutant (45). The website http://skl.scau.edu.cn/targetdesign/ was employed for designing gRNA targets and choosing the restriction enzymes for genotyping (Dataset S5). The CRISPR/Cas9 construct, pHEE401E-ect234, was transformed into WT plants using the floral dipping method. PCR and restriction enzyme digestion were applied to identify knock-out plants. The ect2/3/4 mutant plants identified in the T3 generation were confirmed by Sanger-sequencing of gRNA-targeted regions. In the T2 and T3 generations, Cas9-F and Cas9-R primers were used to identify Cas9-free plants, and Cas9-free T4 plants were also confirmed through viable growth screening on MS plates supplemented with 20 mg/L hygromycin B.

For complementation of the mutation in ECT2, the ECT2 coding sequence was inserted into the plasmid pLIC-YFP by ligation-independent cloning (LIC) (48) to obtain 35S:ECT2-YFP. To construct gECT2-YFP, the 35S promoter was replaced by the ECT2 native promoter through restriction cloning EcoRI. The construct was transformed to the ect2 plants for complementation. For the MS2 assay, the pMCP-LIC vector was generated by replacing YFP sequence in pYFP-LIC with the HA-MCP sequence. CDS sequences of YFP, ECT1, ECT2, ECT3, or ECT4 were amplified and inserted into the pMCP-LIC vector by ligation-independent cloning. Six tandem MS2 binding sites were amplified from pEGAD-M6 (27) and inserted into the pFLUC-RLUC vector (22) after the stop codon of FLUC to generate the pFLUC-M6-RLUC reporter.

Hpa Noco2 infection assay was performed as previously described (49). For Hpa Noco2 infection, 12-d-old plants were grown under 12/12-h light/dark cycles. The plants were sprayed with a suspension of 3-5 × 10⁴ spores/mL in H₂O and covered with a dome to achieve 100% saturated humidity for 1 d. The plants were then exposed to ambient humidity for 3 d before being covered again. 7 d after the initial infections, spores were collected by suspending infected plants in 5 mL of water. Spores were then counted in a hemacytometer under a microscope.

The basal resistance assay was performed as previously described (49). Leaves from 3.5-wk-old plants were infiltrated with Psm ES4326 (OD_600nm = 0.0001) suspended in 10 mM MgCl₂. Bacterial growth was scored on Day 0 and Day 3 by serial dilutions in 10 mM MgCl₂. For the DEX:siMTA silencing lines, 50 µM dexamethasone (DEX) or mock (H₂O) was sprayed on leaves 1 d prior to infiltration with Psm ES4326 (OD_600nm = 0.0001). The elf18-induced protection assay was performed as previously described (22). Leaves from 3.5-wk-old plants were infiltrated with 1 µM elf18 in H₂O or mock (H₂O) or pre-treated for 1 d with 50 µM DEX or mock (H₂O) prior to 1 µM elf18 treatment. After 1 d, the same leaves were then infiltrated with Psm ES4326 (OD_600nm = 0.001) and bacterial growth was scored 2 d later.

Arabidopsis plants were grown on MS plates (1/2 MS basal salts, 1% sucrose, and 0.8% agar) for 6 d. Plants were then transferred to 6-well plates containing liquid MS media (1/2 MS basal salts, 1% sucrose) and grown for four additional days. For elf18 treatment, the growth media were replaced with fresh liquid MS media with or without 1 µM elf18 and incubated for 1 h. The sample was collected, immediately frozen in liquid nitrogen, ground using the Genogrinder (SPEX SamplePrep), and subjected to total RNA isolation using the Direct-zol RNA Miniprep Plus Kit (Zymo). mRNA was enriched twice through poly(A) selection using Oligo d(T)₂₅ Magnetic Beads (NEB). Around 200 ng purified mRNA was digested in a two-step manner with nuclease P1 (1 μL, Sigma-Aldrich) in 20 μL reaction buffer containing 10 mM of NH₄OAc (pH= 5.3) at 42 °C for 2 h. Then, 1 μL of shrimp alkaline phosphatase (rSAP, NEB) was added along with 2.5 μL of 10× CutSmart buffer (NEB) and incubated at 37 °C for 2 h. After the incubation, the sample was diluted with additional 35 μL water and filtered with 0.22 μm filters (4 mm diameter, Millipore) and 6 to 8 μL of the entire solution was injected as one replicate into C18 reverse phase column coupled to Agilent 6460 LC-MS/MS spectrometer in positive electrospray ionization mode. The nucleosides were quantified by using retention time and the nucleoside to base ion mass transitions (282-to-150). For all the quantification, a mock control with only digestion buffers and enzymes was included each time and was later used for the subtraction of baseline signals. Quantification was performed in comparison with the standard curve, obtained from pure nucleoside standards running with the same batch of samples. The m⁶A level was calculated as the ratio of m⁶A to A.

3.5-wk-old WT plants were infiltrated with 10 µM elf18 or mock (H₂O). Whole leaf tissue (~1.5 g per replicate, three replicates per treatment) was collected and flash frozen in liquid nitrogen 1 h after infiltration. Total RNA was then extracted using TRIZOL (Ambion) according to the manufacturer’s instructions. mRNA was isolated from the total RNA using the Dynabeads mRNA DIRECT purification kit (ThermoFisher). mRNA concentration was adjusted to 15 ng/µL in 100 µL and fragmented using a Bioruptor ultrasonicator (Diagenode) with 30 cycles of 30 s on/off. m⁶A-immunoprecipitation (m⁶A-IP) and library preparation were performed according to the published protocol (50). Specifically, the total input mRNA used was reduced to 1 µg. Correspondingly, 2.5 µg of anti-m⁶A antibody was used. Input and m⁶A-IP eluted RNA libraries were constructed using the TruSeq Stranded mRNA kit (Illumina) following the standard protocol. Sequencing was carried out using Illumina HiSeq 4000 according to the manufacturer’s instructions.

Three replicates of m⁶A input and IP libraries were aligned to the Arabidopsis genome (TAIR10) using Bowtie2.0 (51). Reads were assigned to genes by alignment to the exons of the longest isoform of each gene. For m⁶A peak calling, a sliding window of 100 nt in length with a step of 50 nt was used to determine read coverage of the longest isoform of each gene in each library. Read coverage in each window was pseudocount transformed and normalized by the median window read coverage. In m⁶A IP libraries, windows with a greater than 3 peak-over-median (POM) score were condensed. These regions were normalized by the read coverage in the corresponding input library to determine a peak-over-input (POI) score. A region of greater than 3 POI in at least two of the three replicates was determined to be a putative m⁶A methylation site. Putative peaks are considered specific to a condition if none of the windows in the peak is found with a significant POI score in the other condition.

FA-CLIP-seq was modified from the previous study (9). gECT2-YFP and 35S:YFP plants were grown on MS plates for 6 d. Plants were then transferred to 6-well plates containing liquid MS media and grown for four additional days. For elf18 treatment, the growth media were replaced with fresh liquid MS media with or without 1 µM elf18 and incubated for 1 h. For input, plant tissues were collected right after treatment. For immunoprecipitation samples, plants were fixed and crosslinked by 1% formaldehyde under a vacuum to enhance infiltration. Two replicates were collected for the experiment. The lysate from 2 g tissues was partially digested with RNase T1 and immunoprecipitated by 25 µL GFP-trap magnetic beads to pull down ECT2-bound RNA fragments. The other steps were same as described in the previous study (9). The final libraries were sequenced using the Illumina NovaSeq 6000 platform. The method of peak calling was similar to that used for m⁶A peak calling. ECT2 binding sites were determined by a sliding window analysis of read density between IP and input samples for each genotype and condition. Briefly, for each transcript model reads were assigned to windows along the length of the transcript model with a sliding window step size of 10 nt. Read density was compared for each window in the IP and input sample to calculate the POI score (IP/input). Sensitivity of the POI was determined by analyzing the YFP samples and set to the 90th percentile of POI scores across YFP libraries, with all windows below this POI score removed from further analysis. For each transcript model, significance of the POI score was determined by Z-score analysis. Windows with a Z-score below 2.576 were removed from future analysis. The false discovery rate (FDR) was calculated by calling peaks with input and IP samples inverted (input/IP). Windows with a POI Z-score of >2.576 and an FDR < 0.05 were considered putative ECT2 binding sites.

Agrobacteria were cultured overnight, spun down, and resuspended in the infiltration buffer (10 mM MgCl₂, 10 mM MES pH = 5.6, and 200 µM acetosyringone) for 2 h and diluted to OD_600nm = 0.8, and strain harboring pFLUC-M6-RLUC was mixed with that of pMCP-YFP or pMCP-ECT in a 1:3 ratio and infiltrated into N. benthamiana leaves. Samples were collected for RNA and protein extraction after 2-d expression. Protein samples were subject to dual-luciferase assay as previously described (17), whereas RNA samples were extracted by TRIZOL according to the manufacturer’s instructions and followed by cDNA synthesis and qPCR. All qPCR primers used are described in Dataset S6.

WT, fip37-4, and ect2/3/4 plants were grown on MS plates for 6 d. Plants (6 to 10 seedlings) were then transferred to 12-well plates containing liquid MS media and grown for four additional days. For elf18 treatment, the growth media were replaced with fresh liquid MS media with or without 1 µM elf18 and incubated for 1 h. At half an hour after adding inhibitors of mRNA transcription (0.6 mM cordycepin, 10 µM actinomycin), plants were collected (time 0) and flash frozen in liquid nitrogen. The following timepoints were subsequently collected 1, 2, and 4 h later. All frozen tissue samples were processed with TRIZOL to isolate total RNA according to the manufacturer’s instruction. n = 3.

WT, fip37-4, and ect2/3/4 plants were grown on MS plates for 6 d. Plants were then transferred to 6-well plates containing liquid MS media and grown for four additional days. For elf18 treatment, the growth media were replaced with fresh liquid MS media with or without 1 µM elf18 and incubated for 1 h. Polysome profiling was performed as previously described (22). Briefly, 0.3 g tissues for each sample were ground in liquid nitrogen and extracted with 1.2 mL polysome extraction buffer [200 mM Tris pH 9.0, 200 mM KCl, 35 mM MgCl₂, 25 mM EGTA, 5 mM DTT, 1 mM phenylmethanesulfonyl fluoride, 50 µg mL⁻¹ cyclohexamide, 50 µg mL⁻¹ chloramphenicol, 1% (v/v) Brij-35, 1% (v/v) Igepal CA630, 1% (v/v) Triton X-100, 1% sodium deoxycholate, and 1% (v/v) polyoxyethylene 10 tridecyl ether]. The resulting lysate (0.8 mL) was loaded on a sucrose gradient (15 to 60%) and centrifuged at 4 °C for 10 h (35,000 rpm; Ti-SW41 rotor), and 0.1 mL of lysate was saved as input. Polysome profiles and fractions were collected using a fractionator and 254 nm UV monitor. Fractions corresponding to polysomes were pooled, as determined by their sedimentation patterns. RNAs were isolated from input and polysome pooled fractions by TRIZOL-LS according to the manufacturer’s instructions. n = 3.

Total RNA (500 ng) was used for library preparation via the QuantSeq 3′ mRNA Seq Library Prep FWD Kit from Illumina (Lexogen). Sequencing was performed at the Duke Center for Genomic and Computational Biology using NovaSeq 6000 or NovaSeq X for 50 bp single-end reads. Raw reads were trimmed to remove adaptors and poly(A) sequences by using Trim Galore (52), and then mapped to the Arabidopsis genome TAIR10 by using the STAR RNA sequencing aligner (53). The raw counts for each gene were calculated by HTSeq and normalized by DESeq2 for further analyses (54).

To calculate the RNA decay rate, the Bioconductor RNAdecay package was utilized to normalize the data, model mRNA decay, and compare genotype effects. With limitation of the program, different genotypes with the same treatment were fitted to the decay models together. For each transcript, the best decay model was selected by the lowest Akaike information criterion (AICc), and the initial decay rate α and the decay of decay rate β of the transcript were estimated by that model. Half-life of mRNA was calculated as t_1/2 = ln(2)/α. For the downstream analysis, the transcripts with the variance σ² of the model less than 0.0625 and half-life less than 720 min across all genotypes under the same condition were selected.

To measure the translation efficiency (TE), TE for each transcript was calculated in individual replicates as mRNA(polysome)/mRNA(input), and the average TE from all three replicates was used for global analysis. DESeq2 was used for differential expression analysis, and the inducible transcripts of WT were detected by fold-change of mRNA(input) >1.5 and an adjusted P-value < 0.05.

T.C., G.H.G., J.M., and X.D. designed research; T.C., G.H.G., J.M., M.M., G.-Z.L., G.X., S.K., Y.X., and C.L. performed research; C.H. contributed new reagents/analytic tools; T.C., G.H.G., J.M., and M.M. analyzed data; and T.C., G.H.G., J.M., and X.D. wrote the paper.

X.D. is a founder of Upstream Biotechnology Inc. and a member of its scientific advisory board, as well as a scientific advisory board member of Inari Agriculture Inc. and Aferna Bio, Inc. G.H.G. is a founder of Upstream Biotechnology Inc. C.H. is a scientific founder, a member of the scientific advisory board and equity holder of Aferna Bio, Inc. and Ellis Bio Inc., a scientific cofounder and equity holder of Accent Therapeutics, Inc., and a member of the scientific advisory board of Rona Therapeutics and Element Biosciences.