The m⁶A reader SlYTH2 negatively regulates tomato fruit aroma by impeding the translation process

Significance

It is generally accepted that many modern commercially produced fruits with high firmness have reduced flavor quality, particular in volatiles that appeal to consumers. Our work generates tomato fruit with increased aroma-associated volatiles and consumer preferences by genome editing an N⁶-methyladenosine (m⁶A) reader SlYTH2, without sacrificing fruit ripening and firmness. Thus, SlYTH2 is distinct from m⁶A regulators such as erasers that influence many ripening processes. Enriched aroma is caused by increased translation of SlYTH2 target genes responsible for volatile synthesis. Our study provides insight into the mechanism of tomato fruit flavor regulation during ripening.

Abstract

N⁶-methyladenosine (m⁶A) is a fundamentally important RNA modification for gene regulation, whose function is achieved through m⁶A readers. However, whether and how m⁶A readers play regulatory roles during fruit ripening and quality formation remains unclear. Here, we characterized SlYTH2 as a tomato m⁶A reader protein and profiled the binding sites of SlYTH2 at the transcriptome-wide level. SlYTH2 undergoes liquid–liquid phase separation and promotes RNA–protein condensate formation. The target mRNAs of SlYTH2, namely m⁶A-modified SlHPL and SlCCD1B associated with volatile synthesis, are enriched in SlYTH2-induced condensates. Through polysome profiling assays and proteomic analysis, we demonstrate that knockout of SlYTH2 expedites the translation process of SlHPL and SlCCD1B, resulting in augmented production of aroma-associated volatiles. This aroma enrichment significantly increased consumer preferences for CRISPR-edited fruit over wild type. These findings shed light on the underlying mechanisms of m⁶A in plant RNA metabolism and provided a promising strategy to generate fruits that are more attractive to consumers.

Among hundreds of distinct chemical modifications in eukaryotic messenger RNAs (mRNAs), N⁶-methyladenosine (m⁶A) is the most abundant and prevalent modification type (1). In plants, dynamic and reversible changes of m⁶A levels are regulated by enzymes (namely m⁶A writers and erasers) (2). The m⁶A marks in target RNAs are recognized and decoded (1–3) by m⁶A-binding proteins, so-called m⁶A readers, containing YT521-B homology (YTH) domain. m⁶A readers recognize m⁶A modifications of RNAs and widely regulate RNA-related processes, including alternative splicing, nuclear export, mRNA stability, and translation efficiency (1–3). In plants, biological functions of m⁶A readers have been well studied in Arabidopsis, where they are involved in plant development, plant morphology, flora transition, and stress responses (4–10).

Increasing evidence indicates that m⁶A readers determined target RNAs’ fate via their liquid–liquid phase separation (LLPS) ability (11). LLPS is often triggered by proteins with prion-like domains (PrLDs) or intrinsically disordered regions (IDRs) that induce the aggregation of complex proteins or RNA–protein granules. In aggregated liquid-like condensates, mRNA decay and translation inhibition were reported (12–14). In Arabidopsis, m⁶A reader evolutionarily conserved C-terminal region 1 (ECT1), ECT2, and CPSF30-L were shown to undergo phase separation (4, 6, 8). For example, m⁶A-modified RNAs become sequestered into ECT1-induced cytosolic condensates under salicylic acid (SA)-mediated stress, leading to mRNA decay in response to phytohormone stress (6).

In addition to model plant Arabidopsis, m⁶A readers have also been identified in multiple fruit crops, including tomato (15, 16), strawberry (17), apple (18–20), pear (21), and lichi (22). However, whether these candidate m⁶A readers could recognize m⁶A modification and how they play regulatory roles during fruit ripening and quality formation is poorly understood.

Tomatoes are the most valuable fruit and vegetable crop produced throughout the world, with an annual worldwide production of ~257 million tons (2021, FAOSTAT). For consumers, flavor preferences, together with health benefits, are key contributors to consumption of fruits. Flavor is determined by the specific chemical composition of the fruit. Although sugars and acids influence our perception of flavor, it is the aroma volatiles that determine the unique flavor of a ripe fruit (23, 24). For instance, cold storage alters the contents of specific volatiles without altering sugars and acids, resulting in reduced flavor and consumer enjoyment of tomato fruit (25). Among ~400 volatiles that are detectable in tomato fruit, only 20 to 30 chemicals correlate with flavor (26). Although the content of most flavor-impacting volatiles tends to increase during fruit ripening, the regulatory mechanism of their synthesis is not well understood.

Over the last decade, great progress has been achieved in identifying genes and pathways for the synthesis of aroma-associated volatiles in fruits (23). Transcription factors play a vital role in coordinating metabolic activities. CRISPR-Cas9-mediated genome editing of the ripening regulator MADS-RIN and NAC-NOR delayed fruit ripening, resulting in higher firmness but significant reductions in aroma volatiles (27, 28). In addition to transcriptional regulation, epigenetic modifications such as DNA methylation are involved in aroma formation. Reduced production of aroma volatiles and delayed fruit ripening were associated with high levels of DNA methylation caused by knockout of DNA demethylase SlDML2 (25, 27). However, whether RNA methylation such as m⁶A modification has any relationship with fruit aroma, and more fundamentally, how m⁶A readers affect the processes of fruit ripening and aroma formation, remain unexplored.

Here, we characterized the tomato (Solanum lycopersicum) YTH-domain family protein SlYTH2 as an m⁶A reader. Integrating transcriptome, proteome, polysome profiling, and LLPS assays, we determined that SlYTH2 represses translation of target m⁶A-modified mRNAs via its phase separation ability, thereby regulating the formation of fruit aroma. Knockout of SlYTH2 led to the enrichment of tomato fruit aroma without compromising ripening, pigmentation, and softening. These findings show that modulation of an m⁶A reader is a promising approach to generate appealing fruits for consumers and growers.

Results

SlYTH2 Loss-of-Function Increases Aroma-Associated Volatiles Production in Fruit.

The tomato genome contains nine putative YTH-domain containing proteins, designated SlYTH1-SlYTH9 (16). Within the YTH family, the function of SlYTH1 was initially selected for analysis due to its highest expression level in tomato. Knockout of SlYTH1 resulted in low seed germination rate and flattened fruits with reduced number of locules (29). Since these phenotypes are unacceptable for an agronomic crop, we turned our attention to other m⁶A reader proteins. SlYTH2 (Solyc01g028860) is highly homologous to Arabidopsis ECT2, ECT3, and ECT4 (SI Appendix, Fig. S1), and exhibits the second-highest expression level, behind SlYTH1, during fruit ripening in two different cultivars (Ailsa Craig, SI Appendix, Fig. S2; M82, SI Appendix, Fig. S3), suggesting potential roles of SlYTH2 in multiple biological processes. We generated slyth2 mutants through CRISPR/Cas9 genome editing system in Ailsa Craig tomato and identified two homozygous mutants, slyth2 #1 and slyth2 #2, harboring a 5-bp deletion and a 65-bp deletion in two designed target sites, respectively (SI Appendix, Fig. S4A). In both lines, protein translation is predicted to be terminated prior to the YTH domain, producing truncated proteins of 49 and 144 amino acids, respectively (SI Appendix, Fig. S4B). No off-target editing events were observed based on the results generated by CRISPR-P (http://crispr.hzau.edu.cn/CRISPR2/). Protein blot assays revealed full-length protein of SlYTH2 in wild type (WT), but not in slyth2 mutants (SI Appendix, Fig. S4C). Therefore, slyth2 #1 and slyth2 #2 mutants were used for further studies.

No discernible differences in plant growth, flowering, or fruit set were observed in slyth2 mutants (SI Appendix, Figs. S5 and S6). Given the substantial upregulation of SlYTH2 in fruit, we conducted a detailed analysis of fruit-associated phenotypes between slyth2 mutants and WT. There were no significant differences observed in pigmentation or firmness during fruit ripening from Breaker (Br) to the full red ripe stage (7 d after Br, B + 7) (Fig. 1 A and B and SI Appendix, Fig. S7). Remarkably, a substantial difference in fruit aroma between slyth2 mutants and WT fruits were detected using electronic nose analysis on B + 7 fruits. The first two principal component (PC) factors explained 96.8% of the difference in volatile profiles between slyth2 mutants and WT fruit (Fig. 1C). Next, we performed GC–MS analysis to quantify differences in individual volatile content caused by mutation of SlYTH2. The slyth2 mutants produced higher levels of aroma-associated volatiles than WT during fruit ripening (Fig. 1D and SI Appendix, Table S1). Total volatile content in slyth2 #1 mutant at B + 7 was approximately 1.5-fold higher than WT fruit (Fig. 1D). Loss-of-function of SlYTH2 led to significant increases in volatiles derived from fatty acids, carotenoids, nonaromatic amino acids, and phenylpropanoids (Fig. 1E). These results suggested that knockout of SlYTH2 enhances production of aroma-associated volatiles without obvious negative effects on fruit pigmentation or firmness during ripening.

Fig. 1.

SlYTH2 is essential for aroma-associated volatiles synthesis in tomato fruits. (A) Tomato fruit ripening process of WT and slyth2 mutants. Br, breaker; B + n, n days after breaker. (Scale bar, 1 cm.) (B) Fruit firmness of WT and slyth2 mutants during fruit ripening. Data are presented as mean ± SD of nine independent biological replicates, each indicated by one dot. Different lowercase letters indicate significant differences (Tukey’s multiple range test, P < 0.05). (C) E-nose with principal component analysis of WT, slyth2 #1, and slyth2 #2 fruits at B + 7 stage. Each cluster consists of six replicates (three biological replicates × two technical replicates), each replicate represented by one dot. (D) Content of total flavor-related volatiles during WT and slyth2 mutant fruit ripening. (E) Major classes of fruit volatiles in WT and slyth2 mutant fruits at B + 7 stage. Data in (D) and (E) are indicated as mean ± SD of three biological replicates, each dot represents one replicate. Significant differences (Tukey’s multiple range test, P < 0.05) are denoted with different lowercases. (F) Aroma evaluations by untrained panelists (n = 28) and preference tests (n = 28) of WT and slyth2 mutant fruits at B + 7 stage.

The noticeable increase in aroma-associated volatile contents hinted at potential impacts on the fruits’ sensory attributes. To assess flavor perception, we employed an untrained consumer panel. Over 90% of the panelists discerned differences in aroma between WT and slyth2 fruits. Within the panel, 82% reported a more noticeable aroma in slyth2 mutant fruits, while 11% perceived a stronger scent in WT fruits, with only 7% detecting no difference (Fig. 1F). Furthermore, 71% of the panelists preferred the aroma of slyth2 mutant fruits, 14% preferred the WT tomatoes (Fig. 1F). Collectively, our results show that SlYTH2 loss-of-function enhances tomato fruit aroma quality.

SlYTH2 Functions As an m⁶A Reader.

To investigate whether SlYTH2 could recognize m⁶A modification, we first performed bioinformatic analysis. Multiple sequence alignment using ESPript 3 revealed highly conserved functional sites for YTHDF subfamily proteins (30), including the aromatic cage structure, m⁶A, and RNA contact sites (SI Appendix, Fig. S8A), indicating that SlYTH2 might have a similar m⁶A reading mechanism to that of human HsYTHDF and Arabidopsis ECT proteins (4, 31–33). The homology-modeling server SWISS-MODEL (34) generated a model for the 3D structure of SlYTH2 using the most homologous protein HsYTHDF2 (PDB: 4RDN) as a template (SI Appendix, Fig. S8B). Three functional tryptophan residues that form the aromatic cage structure for m⁶A binding are conserved between HsYTHDF2 and SlYTH2: Trp-432, Trp-486, and Trp-491 in HsYTHDF2 (31), Trp-468, Trp-525, and Trp-530 in SlYTH2. Three amino acids involved in hydrogen bond interactions with m⁶A are also highly conserved: Tyr-418, Asp-422, and Cys-433 in HsYTHDF2 (31) and Tyr-454, Asp-458, and Ala-469 in SlYTH2. Collectively, the conserved patterns for multiple alignment and homology modeling suggest that SlYTH2 functions as a tomato m⁶A reader.

We next isolated poly(A) RNA and purified glutathione S-transferase (GST) tagged SlYTH2 (GST-SlYTH2) recombinant protein to perform in vitro RNA immunoprecipitation (RIP) followed by liquid chromatography–tandem mass spectrometry (LC–MS/MS). Significant enrichment of m⁶A was observed in the SlYTH2-bound sample (SI Appendix, Fig. S8C), supporting the m⁶A-binding ability of SlYTH2. As a control, SlYTH2 lacking the putative binding structure (GST-SlYTH2m) by mutation of Trp-468 to Ala was produced. Biotin-labeled RNA probes containing either A or m⁶A were designed, followed by electrophoretic mobility shift assay (EMSA) to detect the m⁶A-binding ability of GST-SlYTH2 and GST-SlYTH2m (SI Appendix, Figs. S8D and S9). SlYTH2 bound the m⁶A-modified probe, but not the unmethylated one (SI Appendix, Fig. S8D). Meanwhile, the mutated protein of SlYTH2 lost its m⁶A-binding ability (SI Appendix, Fig. S8D). Additionally, to further test the m⁶A-binding ability of SlYTH2 in vivo, we generated UBQ:SlYTH2-FLAG transgenic plants overexpressing SlYTH2 with a 3 × Flag tag and performed in vivo formaldehyde (FA) cross-linking-RIP-LC–MS/MS analysis. After immunoprecipitation (IP), we found a significant accumulation of m⁶A-modified RNAs in SlYTH2-Flag-IP sample compared to control immunoglobulin G (IgG)-IP and input samples (Fig. 2A). Taken together, our results strongly indicate that SlYTH2 functions as an m⁶A reader in tomato.

Fig. 2.

The m⁶A binding of SlYTH2 to volatile synthesis-related genes SlHPL and SlCCD1B is associated with the formation of fruit aroma. (A) in vivo FA-RIP-LC–MS/MS showing that m⁶A is enriched in SlYTH2-Flag-IP sample compared with input and IgG-IP samples. (B) Venn diagrams showing the overlap of identified SlYTH2-binding peaks and m⁶A peaks corresponding to 3,850 transcripts (termed SlYTH2 & m⁶A targets). (C) Distribution of SlYTH2 & m⁶A targeted sites across transcript with three nonoverlapping segments (5′ UTR, CDS, and 3′ UTR). UTR, untranslated region; CDS, coding sequence. (D) Sequence motif identified within SlYTH2 & m⁶A targeted sites by HOMER software. (E) Integrative Genomics Viewer (IGV) showing the distributions of SlYTH2-targeted peaks and m⁶A peaks in transcripts related to volatile synthesis: SlHPL and SlCCD1B. The light-yellow rectangle indicates the position of SlYTH2 & m⁶A targeted sites. The input and IgG-IP reads are presented in the foreground. (F) m⁶A-IP-qPCR showing the enrichment of SlHPL and SlCCD1B in m⁶A-IP sample compared to Input. (G) RIP-qPCR showing the binding affinity of SlYTH2 to SlHPL and SlCCD1B in vivo. Data in (A), (F), and (G) are shown as mean ± SD (n = 3, each biological replicate is marked with one dot). ** indicates significant difference of P < 0.01 with unpaired t test.

SlYTH2 Recognizes Volatile Synthesis-Related Genes SlHPL and SlCCD1B via m⁶A Modification.

Having established that SlYTH2 is an m⁶A reader and affects fruit aroma-associated volatile synthesis, we next investigated whether genes related to volatile synthesis have m⁶A modifications and how SlYTH2 acts on them. We first performed m⁶A-seq with three biological replicates using WT fruit at the Br stage. After peak calling with the R package exomePeak2 (35), m⁶A peaks with a false discovery rate (FDR) <0.01 were identified as high-confidence m⁶A peaks. A total of 11,376 high-confidence m⁶A peaks detected in all three biological replicates were selected for further analysis (SI Appendix, Fig. S10A and Dataset S1). The distribution of m⁶A peaks throughout the transcriptome showed that they were highly enriched in 3′ untranslated regions (3′ UTR) and around stop codons (SI Appendix, Fig. S10 B and C), in agreement with previous studies in tomato fruit (36, 37). Moreover, motif analysis of m⁶A peaks revealed a UGUAY (Y = U > A) sequence motif (SI Appendix, Fig. S10D), which is consistent with observations in Arabidopsis and tomato (10, 37).

Meanwhile, we employed the formaldehyde cross-linking and immunoprecipitation (FA-CLIP) method, conducting two biological replicates using UBQ:SlYTH2-FLAG transgenic Br stage fruits to enable the identification of transcriptome-wide binding sites of SlYTH2. Based on IP enrichment criteria (SlYTH2/Control ≥ 2, P < 0.01), we identified 8,638 peaks as high-confidence SlYTH2-binding sites and found that 56.9% (4,911 out of 8,638) of SlYTH2-binding sites overlapped with m⁶A peaks (termed SlYTH2 & m⁶A targeted sites) corresponding to 3,850 transcripts (termed SlYTH2 & m⁶A targets; Fig.r 2B and Dataset S2), further confirming the m⁶A-binding capability of SlYTH2 in tomato. We subjected the 4,911 SlYTH2 and m⁶A targeted sites for further analysis and found a heightened enrichment of these sites in the 3′ UTR and around the stop codon (Fig. 2C and SI Appendix, Fig. S11). Motif analysis revealed that the binding motif of SlYTH2 was highly overlapping with the m⁶A occurrence motif: UGUAY (Fig. 2D). Moreover, Gene Ontology (GO) enrichment analysis revealed a high enrichment of SlYTH2 & m⁶A targets in cellular metabolic process, organic compound metabolic process, and peptide metabolic process, which prompted us to investigate the intrinsic link between fruit aroma synthesis and m⁶A recognition (SI Appendix, Fig. S12).

Fatty acid-derived volatiles in tomato fruit stand as the most abundant aroma-associated chemicals (23). These volatiles are synthesized from polyunsaturated fatty acids, which are liberated from glycerolipid by lipase, including SlLIP8. Subsequently, six-carbon (C6) aldehydes such as E-2-hexenal and hexanal are catalyzed by lipoxygenase (LOX) such as SlLOXC and hydroperoxide lyase (HPL) such as SlHPL. These C6 chemicals could be further converted to their corresponding alcohols through alcohol dehydrogenase (ADH) activity (38–41) (SI Appendix, Fig. S13A). Volatiles derived from carotenoids are also important contributors to tomato fruit aroma. Carotenoid cleavage dioxygenase 1 (CCD1) cleaves multiple carotenoids to release apocarotenoid volatiles, including 6-methyl-5-hepten-2-one (MHO), geranylacetone, and β-ionone (42, 43) (SI Appendix, Fig. S13B).

We focused on these genes associated with volatile synthesis and found that the transcripts of SlHPL and SlCCD1B were indeed SlYTH2 & m⁶A targets, with the recognition sites in the 3′ UTR (Fig. 2E). Subsequently, we designed biotin-labeled RNA probes for SlHPL and SlCCD1B within the 3′ UTR based on m⁶A-seq and FA-CLIP results. EMSA results showed that m⁶A-modified probes of both SlHPL and SlCCD1B could be recognized by GST-SlYTH2 protein. In addition, the binding signals were enhanced with increasing protein content in the reaction system (SI Appendix, Fig. S13 C and D). When m⁶A in the probes were replaced with A, GST-SlYTH2 failed to recognize RNA probes (SI Appendix, Fig. S13 C and D), indicating that the protein–RNA combination occurred through m⁶A. To investigate whether there are SlYTH2–RNA interactions in vivo, we performed m⁶A-IP-qPCR and RIP-qPCR in tomato fruits. m⁶A-IP-qPCR confirmed that mRNAs of SlHPL and SlCCD1B displayed significantly higher accumulations in m⁶A-IP samples compared to input (Fig. 2F). A polyclonal antibody that specifically recognized SlYTH2 was used to immunoprecipitate SlYTH2 and SlYTH2-bound RNAs. The RIP-qPCR assays indicated that transcripts of SlHPL and SlCCD1B were directly bound by SlYTH2 in vivo (Fig. 2G). Taken together, we confirmed that SlYTH2 could recognize m⁶A sites of SlHPL and SlCCD1B transcripts in tomato fruit.

Interestingly, we observed that slyth2 mutants produced significant higher content of fatty acid-derived volatiles, including E-2-hexenal and hexanal, whose synthesis is specifically catalyzed by SlHPL, the target of SlYTH2 (SI Appendix, Fig. S13E). Similar results were also noticed in carotenoid-derived volatiles: MHO, geranylacetone, and β-ionone were highly enriched in slyth2 mutants, which are released by SlCCD1B (SI Appendix, Fig. S13F). Collectively, these results give us the hint: Increased contents of volatiles derived from fatty acids and carotenoids in slyth2 mutant fruits correlate with the m⁶A-binding ability of SlYTH2.

SlYTH2 Affects Protein Accumulation, but Not Transcript Abundance.

It has been demonstrated that m⁶A readers precisely regulate the fate and metabolism of m⁶A-modified RNA, such as mRNA stability and translation efficiency (3). To elucidate how SlYTH2 affects the fate of target RNAs, we first analyzed the genome-wide gene expression patterns of slyth2 #1 and WT fruits by RNA-Seq. The results showed a total of 577 differentially expressed genes (DEGs) (|fold change| > 2, P < 0.05), of which 149 were down-regulated and 428 were up-regulated in slyth2 #1 compared to WT (SI Appendix, Fig. S14A and Dataset S3). GO enrichment analysis of DEGs indicated that genes related to response to light stimulus and to radiation were highly enriched (SI Appendix, Fig. S14B).

Notably, only 17 out of the 577 DEGs overlapped with 3,850 SlYTH2 & m⁶A targets (SI Appendix, Fig. S14C), suggesting that variations in gene expression might not directly align with SlYTH2-mediated mRNA regulation. Additionally, none of the known fruit volatile synthesis-related genes were observed in the DEGs list (Dataset S3), including SlLIP8, SlLOXC, SlHPL, SlADH2, SlCCD1B, SlBCAT1, SlTNH1, and SlPAR2 (SI Appendix, Fig. S13 A and B). As for the gene expression pattern of SlYTH2 & m⁶A targets SlHPL and SlCCD1B, qRT-PCR assays confirmed that there were no significant differences between slyth2 mutants and WT fruits (SI Appendix, Fig. S14D). The mRNA stability assays showed that the mRNAs of SlHPL and SlCCD1B were degraded after treatment with the transcription inhibitor actinomycin D, but the degradation rate was not significantly changed between the slyth2 mutant and WT (SI Appendix, Fig. S14E). These results suggested that the accumulation of volatiles in slyth2 mutant fruits may be associated with translational regulation.

To gain further understanding of SlYTH2–mRNA regulatory mechanism, we performed polysome profiling assays using slyth2 #1 and WT fruits at Br stage. We found that knockout of SlYTH2 did not change the levels of the 40S or 60S ribosomal subunits but coincided with considerable decreases in the level of 80S monosomes (Fig. 3A). Notably, starting from fraction 8, large increases in disomes, trisomes, and polysomes were observed in the slyth2 mutant compared to WT (Fig. 3A). Collectively, increased polysome peaks with concomitant decreases in 80S monosomes indicated increased translation efficiency in the slyth2 mutant fruits compared to WT. To detect the translation efficiency of SlYTH2-targeted SlHPL and SlCCD1B, qRT-PCR assays were performed using RNAs recovered from total (Input) and collected fractions. We found that in slyth2 mutant fruits, the mRNA levels of SlHPL and SlCCD1B in fraction 8, 10, and 11 were more than twofold higher than in WT fruits (Fig. 3B). In addition, we selected the ripening regulator NAC-NOR, the fatty acid-derived volatile synthesis gene SlADH2, and a housekeeping gene ubiquitin as negative controls, none of which are SlYTH2 targets. There were no significant increases in translation efficiency in factions 8, 10, and 11 of these mRNAs in slyth2 mutant compared to WT (SI Appendix, Fig. S15). These results demonstrate that SlYTH2-induced translational repression of SlHPL and SlCCD1B is achieved by direct targeting of mRNAs.

Fig. 3.

SlYTH2 inhibits translation efficiency of SlHPL and SlCCD1B, thereby suppressing protein accumulation. (A) Quantitative sucrose density gradient analysis showing the polysome profiles of WT and slyth2 #1 fruits. (B) Translation efficiency of SlHPL and SlCCD1B in WT and slyth2 #1 fruits. The translation efficiency of each fraction was measured by the abundance ratio of RNAs in corresponding recovered fractions versus the total RNAs. Data are shown as mean ± SD (n = 3, each biological replicate is marked with one dot). NS, not significant, **P < 0.01 with Sidak’s multiple comparisons test. (C) Venn diagrams showing the overlap of the 4,137 proteins detected in WT and slyth2 #1 proteomes with SlYTH2 & m⁶A targets. (D) Violin plot showing relative protein levels of the 1,649 overlapped proteins obtained in (C) in WT and slyth2 #1 fruits. **P < 0.01, unpaired t test. (E) Relative protein levels of SlHPL and SlCCD1B in WT and slyth2 #1 fruits. Data are shown as mean ± SD (n = 3, each biological replicate is marked with one dot). *P < 0.05, unpaired t test. (F) Schematic diagram of the pCAMBIA1300 vector containing CDS and 3′UTR fragment of SlHPL and SlCCD1B genes. (G) Protein blot assay showing the accumulation of SlHPL-Flag and SlCCD1B-Flag is inhibited by SlYTH2-Myc protein in tobacco leaves. Actin was used as a loading control. Two lanes indicate two replicates.

Having found that SlYTH2 negatively regulated ribosome occupancy of its target mRNAs, the next question to ask was whether protein accumulation of target RNAs would be disturbed by SlYTH2. To address this question, we performed proteomic analysis using WT and slyth2 #1 fruits. Of the 4,137 detected proteins, a total of 1,649 overlapped with SlYTH2 & m⁶A targets (Fig. 3C and Dataset S4). There was a significant accumulation of these 1,649 proteins in the slyth2 mutant compared with WT (Fig. 3D). Through differential expression analysis, a total of 1,601 differentially expressed proteins (DEPs) (|fold change| > 1.3, P < 0.05) were identified in slyth2 #1 compared to WT. Among these DEPs, 585 up-regulated DEPs overlapped with SlYTH2 & m⁶A targets (SI Appendix, Fig. S16), as well as SlHPL and SlCCD1B (Fig. 3E), which is consistent with the increased volatile contents in slyth2 mutants (SI Appendix, Fig. S13 E and F). As an additional confirmation, we constructed expression vectors for SlHPL and SlCCD1B, facilitating transcription of both the coding sequence (CDS) and the 3′ UTR. However, translation was only observed in full-length proteins fused with a 3× Flag tag, referred to as SlHPL-Flag and SlCCD1B-Flag, respectively (Fig. 3F). By transiently coexpressing SlYTH2 protein fused with a 4× Myc tag (designated as SlYTH2-Myc) or SlYTH2m protein (designated as SlYTH2m-Myc), which lacks the previously confirmed m⁶A-binding ability (SI Appendix, Fig. S8D), in Nicotiana benthamiana leaves, we observed that the protein content of SlHPL and SlCCD1B was inhibited by SlYTH2-Myc compared to SlYTH2m-Myc (Fig. 3G). Furthermore, we investigated the impact of m⁶A site mutagenesis on the protein content of SlHPL and SlCCD1B. To achieve this, we generated mutant constructs for both proteins, namely SlHPLm-Flag and SlCCD1Bm-Flag, where the m⁶A sites identified through m⁶A-seq and FA-CLIP analysis were mutated (SI Appendix, Fig. S17 A and B). The protein blot assay revealed that SlYTH2-Myc exerts inhibitory effects on the protein accumulation of SlHPL-Flag and SlCCD1B-Flag, while having no significant impact on their mutated counterparts (SI Appendix, Fig. S17C). Collectively, these findings lead us to conclude that SlYTH2 represses the translation efficiency of its targets, namely SlHPL and SlCCD1B, thereby diminishing the levels of corresponding proteins in an m⁶A-dependent manner.

SlYTH2 Undergoes Phase Separation with Target m⁶A-Modified RNAs.

It is believed that the human m⁶A reader HsYTHDF1 recognizes m⁶A-modified mRNAs and interacts with translation initiation factors (eIFs) to promote ribosome occupancy and facilitate the translation process (44). Additionally, m⁶A reader could repress mRNA translation via its ability to promote LLPS (45). While how m⁶A readers influence translation in plants is not fully understood. Several studies in plants have also reported that m⁶A reader regulates the fate of target RNAs by forming liquid‐like condensates, such as Arabidopsis CPSF30-L undergoes LLPS in nuclear bodies to regulate poly(A) site choice of its target mRNAs (8), and ECT1 forms cytosolic condensates to modulate SA-mediated plant stress responses (6). These studies prompted us to explore whether SlYTH2 represses mRNA translation by enhancing the formation of LLPS.

Through motif prediction, we found that SlYTH2 contains IDRs and PrLDs which are essential for LLPS generation (Fig. 4A). Using GST-SlYTH2 recombinant protein, we observed that the sample solutions became turbid when treated with PEG8000 (Fig. 4B). For visual observation, we purified the GFP-tagged SlYTH2 recombinant protein (GFP-SlYTH2) (SI Appendix, Fig. S18), which could form spherical droplets in the presence of 10% PEG8000, unlike the GFP protein (Fig. 4C). GFP-SlYTH2 droplets could fuse into a larger condensate within a few seconds under time-lapse microscopy (Fig. 4D). Moreover, fluorescence recovery after photobleaching (FRAP) assay showed that the fluorescent signal of photo–bleached GFP-SlYTH2 droplet recovered from 40% to nearly 90% within 4 min (Fig. 4E). Collectively, these results strongly indicate that SlYTH2 undergoes LLPS to form condensates.

Fig. 4.

SlYTH2 undergoes phase separation and sequesters target m⁶A–RNAs into condensates. (A) PrLD and IDR domains in SlYTH2 were predicted using “Prion-Like Amino Acid Composition” (PLAAC; http://plaac.wi.mit.edu/) and “Predictor of Natural Disordered Regions” (PONDR; http://pondr.com/), respectively. (B) Visualization of turbid solution of GST-SlYTH2 proteins under PEG8000 treatment. GST and Bovine serum albumin (BSA) proteins serve as controls. (C) Fluorescence image showing in vitro LLPS of recombinant GFP-tagged SlYTH2 proteins (GFP-SlYTH2). (Scale bar, 20 μm.) GFP and GFP-SlYTH2 proteins in droplet-promoting buffer containing 10% PEG8000. The inner box indicates the enlarged area (Scale bar, 5 μm). (D) Time-lapse microscopy showing dynamic fusion of GFP-SlYTH2 proteins. (Scale bar, 2 μm.) (E) FRAP assay and quantification data showing the dynamic property of GFP-SlYTH2 droplets. (Scale bar, 2 μm.) Data are presented for three independent FRAP events. (F) Fluorescence image showing condensate formation of GFP-SlYTH2 with Cy3-labeled m⁶A-modified and Cy5-lebeled non-m⁶A-modified RNA probes of SlHPL and SlCCD1B, respectively. Green, GFP-SlYTH2; red, Cy3-labeled probe; blue, Cy5-labeled probe; yellow, merged images. (Scale bar, 5 μm.) (G) Colocalization intensity of GFP-SlYTH2 with Cy3-labeled m⁶A-modified RNA probes or Cy5-labeled non-m⁶A-modified RNA probes of SlHPL and SlCCD1B, respectively. The colocalization intensity was quantified using 10 different confocal images. Data are presented as mean (n = 10, each dot represents one replicate). **P < 0.01 with unpaired t test.

We next evaluated whether condensates induced by SlYTH2 could selectively concentrate m⁶A-modified target mRNAs. The Cy3-labeled m⁶A-modified RNA probes, including SlHPL and SlCCD1B, were individually incubated with GFP-SlYTH2 protein. We found that the droplet signals formed by RNA–protein interactions became stronger with increasing probe concentration (SI Appendix, Fig. S19). Notably, when we introduced Cy5-labeled RNA probes without m⁶A modification, the formation of RNA–protein condensates had a clear bias toward m⁶A-modified ones (Fig. 4 F and G). Taken together, these results demonstrated that SlYTH2 undergoes LLPS and recognizes its target m⁶A-modified mRNAs to form RNA–protein condensates.

SlYTH2 Forms Cytosolic Condensates with Translation Regulators.

The mechanism by which m⁶A readers guide and decide RNA fate is determined by their interacting proteins. Thus, to investigate functional proteins interacting with SlYTH2 in sequestered condensates, we performed formaldehyde cross-linking immunoprecipitation combined with mass spectroscopy (FA-IP/MS) analysis using UBQ:SlYTH2-FLAG transgenic tomato fruits. We identified a total of 126 candidate protein partners (FC > 4, P < 0.01) (Fig. 5A and Dataset S5) that were associated with RNA binding, peptide metabolic process and translation initiation based on GO enrichment analysis (Fig. 5B). We found that SleIF3C, SleIF4B, and SlPABP interacted with SlYTH2 (Fig. 5A). As translation initiation components, eIFs and polyA-binding proteins (PABPs) could act as components of LLPS-induced RNA–protein condensates to stall the mRNA translation process (13). We hypothesize that SleIF3C, SleIF4B, and SlPABP interact with SlYTH2 in an LLPS-induced manner. To confirm this hypothesis, we conducted a split-yellow fluorescent protein (YFP)–based system for bimolecular fluorescence complementation (BiFC) in N. benthamiana leaves. We observed that SlYTH2 interacted with these translation regulators and formed visible cytosolic condensates (Fig. 5C). These results showed that tomato SlYTH2 interacts with proteins associated with mRNA translation in cytosolic condensates, potentially synergistically impeding polysome assembly and mRNA translation.

Fig. 5.

SlYTH2 forms cytosolic condensates with translational regulators. (A) Scatterplot showing SlYTH2-interacting proteins in tomato fruits. The dashed line indicates the threshold of significantly enriched proteins (FC > 4 and P < 0.01). (B) GO enrichment analysis showing protein partners of SlYTH2 in (A) are significantly enriched in RNA binding, peptide metabolic and translation related pathways. (C) BiFC assays showing that SlYTH2 interacts with translational regulators and forms cytosolic condensates, respectively. N. benthamiana leaves were used for transient coexpression and confocal observation. (Scale bar, 20 μm.) (D) Model for the m⁶A reader SlYTH2 in regulation of tomato fruit aroma. SlYTH2, an m⁶A reader, selectively binds to m⁶A-modified transcripts associated with fruit volatile synthesis and promotes LLPS into liquid-like condensates with translational regulators. This interaction impedes ribosome occupancy and translation of target mRNAs, thereby regulating the synthesis of fruit aroma.

Taken together, our results suggest a working model for SlYTH2-modulated fruit aroma formation in tomatoes. The SlYTH2 targets m⁶A-modified mRNAs associated with fruit aroma volatile synthesis and promotes the formation of RNA–protein condensates. Within these condensates, SlYTH2 recruits translation regulators to inhibit ribosome occupancy and mRNA translation, thereby impacting the formation of fruit aroma-related volatiles (Fig. 5D).

Discussion

Epigenetic modification of m⁶A is critical for plant growth, crop yield, and fruit ripening (37, 46, 47). YTH domain-containing proteins are m⁶A readers that function on RNA metabolisms. Thus, investigating the characterization and the mechanistic insights of m⁶A readers is an essential prerequisite for fully understanding the roles of m⁶A in plants. Here, we report tomato SlYTH2 as an m⁶A reader, which inhibits the translation of its target mRNAs in an LLPS-dependent manner, thereby impacting the content of aroma-associated volatiles.

The mRNAs translation involves the formation of mRNA–ribonucleoprotein (mRNP) complexes, namely polysomes, as well as membraneless RNA granules formed via LLPS (48). A recent study revealed that HeLa cell m⁶A reader IGF2BP3 represses mRNA translation by facilitating mRNA trafficking from polysomes to processing bodies (P-bodies) (45). Here, we show that tomato m⁶A reader SlYTH2 recognizes and enriches target transcripts SlHPL and SlCCD1B in RNA–protein condensates (Fig. 4), leading to a reduced ribosome occupancy of these two genes (Fig. 3). Additionally, our study showed that translational factors, including SleIF3C, SleIF4B, and SlPABP, could interact with SlYTH2 to form cytosolic condensates (Fig. 5). Being sequestered in the condensates, these protein complexes may lose their abilities to facilitate ribosome assembly and recruit mRNAs to maintain the translation process. Consequently, this could potentially contribute as an additional factor in SlYTH2-induced translation repression. Taken together, the role for SlYTH2 in repressing mRNA translation indicates a unique role for plant m⁶A readers and enriches our understanding of m⁶A regulatory mechanism in eukaryotes.

The present results show that SlYTH2 represses protein translation. Interestingly, human HsYTHDF1 facilitates protein translation (44). Therefore, these observations show that SlYTH2 modulates translation in a distinctly different manner than human HsYTHDF1. A previous study in Arabidopsissuggested that cytoplasmic ECT2 promotes m⁶A-modified mRNA stability (10), in contrast to the human HsYTHDF2 that induces mRNA degradation (49). These different roles highlight the diversity and complexity of m⁶A functions among different species.

Increasing evidence indicates that m⁶A is critical for LLPS, which plays pivotal roles in plant stress responses (11). For example, Arabidopsis m⁶A reader ECT1 derives RNA–protein condensates to dampen SA-induced stress (6). Setaria italica SiYTH1 confers plant drought tolerance by forming cytosol condensates (50). Our study shows that tomato SlYTH2 promotes LLPS and enhances condensate formation. We speculate that tomato SlYTH2 may possess a regulatory role in responding to environmental factors using LLPS as a mechanism, a topic for further investigation.

CRISPR/Cas9-mediated mutation of SlYTH2 caused an increase in contents of aroma-associated volatiles synthesized from multiple precursors. For instance, the fatty acid-derived C6 aldehyde such as hexanal, and alcohol such as hexanol, were increased by up to twofold and sixfold in slyth2 #1 mutant fruits, respectively (SI Appendix, Fig. S13E and Table S1). Meanwhile, carotenoid-derived MHO and geranylacetone increased by ~twofold and ~threefold after SlYTH2 mutation, respectively (SI Appendix, Fig. S13F and Table S1). These data clearly indicate that in addition to ripening-associated transcription factors and DNA methylation (27, 28), the formation of fruit aroma-associated volatiles is posttranscriptionally regulated by SlYTH2. The interactions between different epigenetic modifications and their links to transcriptional regulation remain elusive, which will be interesting for future research. Notably, aroma-associated volatiles production can be influenced by transcription factors MADS-RIN and NAC-NOR, as well as DNA methylation-related SlDML2; however, all of these regulators have negative impacts on fruit ripening (27, 28). Interestingly, knockout of SlYTH2 enhances the accumulation of volatiles contributing to fruit aroma, without obvious changes in fruit color or firmness during tomato ripening.

In summary, we produced tomato fruit with enriched aroma volatiles by genome editing m⁶A reader SlYTH2. The major effect of SlYTH2 on its target mRNAs is to repress the translation process through its LLPS ability, indicating a unique role for m⁶A readers in plants. Our results reveal molecular and biochemical mechanisms underlying how SlYTH2 inhibits translation and affects fruit quality during fruit ripening. Moreover, our data indicate that improving fruit aroma without negatively affecting fruit firmness is feasible, providing a different route to produce agricultural and horticultural crops with improved flavor.

Materials and Methods

Plant Materials.

Wild-type tomato (S. lycopersicum cv Ailsa Craig, AC) and slyth2 mutants were grown in a condition-controlled glasshouse (16 h of light at 25 °C, 8 h of dark at 20 °C, alternation of light and dark). Fruits were harvested at different ripening stages: Br, B + 3 (3 d after Br), B + 7 (7 d after Br). Three independent biological replicates were collected, each consisting of at least six fruits.

Fruit Quality and Molecular Analysis.

Details for fruit quality (aroma-associated volatiles, consumer panel, electronic nose, firmness, and pigment analysis), molecular (CRISPR/Cas9 gene editing, protein blot, EMSA, in vitro RIP-LC–MS/MS, in vivo FA-RIP-LC–MS/MS, m⁶A-seq, m⁶A-IP–qPCR, FA-CLIP, RIP-qPCR, RNA-sequencing and qRT-PCR, mRNA stability assay, polysome profiling assay, proteome analysis, LLPS assay, FA-IP/MS, and BiFC assay) analyses are provided in SI Appendix. All primers used in this work are listed in SI Appendix, Tables S2–S7.

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Dataset S02 (XLSX)

Dataset S03 (XLSX)

Dataset S04 (XLSX)

Dataset S05 (XLSX)

Data, Materials, and Software Availability

The raw sequencing data of m⁶A-seq, FA-CLIP, and RNA-seq reported in this paper have been deposited in the National Genomics Data Center (https://ngdc.cncb.ac.cn/) under project number PRJCA024042 (51). All other data are included in the manuscript and/or supporting information.