SlERF.D6 positively influences fruit ripening
Much molecular and physiological ripening research has focused largely or exclusively on the pericarp but recent analyses indicate locule tissue undergoes similar molecular changes prior to pericarp maturation (2, 32, 33, 36, 37). As examples, RIN-MADS, NOR-NAC, GRAS38 and SlLOB1 all influence ripening traits and all display their first elevated expression in tomato locule tissue preceding their induction in pericarp (Fig. S1a-d), suggesting that ripening regulators often first respond in locule tissues.
We searched the Tomato Expression Atlas (https://tea.solgenomics.net/) (32) for uncharacterized TFs with similar expression patterns to early locule expression ripening TFs and identified 12 for which we attempted to generate CRISPR/Cas9 edited mutations in tomato cultivar Ailsa Craig (Dataset S2). The edited mutants showed diverse phenotypes: CRISPR editing of SlWOX14 (encoding a WUSCHEL-related homeobox protein), SlHsfA6b (encoding a heat shock factor), and SlERF.D6 (encoding an ethylene response factor) delayed fruit ripening (Fig. 1; Fig. S1e); editing of SlZHD23 (encoding a zinc finger family protein) delayed plant development and altered morphology (Fig. S1f); editing of SlARR11 (encoding a two-component response regulator) displayed reduced pigments in flowers and fruits (Fig. S2), confirming prior transgenic over-expression results (38).
We focused on SlERF.D6 (Solyc04g071770), given the stronger ripening delay in edited lines. SlERF.D6 was additionally repressed in both the ripening deficient rin and nor mutants (Fig. S3a) and was upregulated by exogenous ethylene supplied to MG fruit and downregulated by ethylene inhibitor 1-MCP, indicating that this gene is positively ethylene responsive (Fig. S3b). Examination of reported ChIP-Seq data (2, 39) indicates that the upstream sequences of SlERF.D6 bind TFs RIN-MADS, NOR-NAC and TAGL1 (Fig. S3c). It is noteworthy that all three TFs result in ripening inhibition when mutated (2), and RIN-MADS and TAGL1 have been shown to physically interact (40). In addition to the AP2/ERF DNA binding domain, analysis of SlERF.D6 protein structure using MobiDB (41) and PLAAC (42) revealed four intrinsically disordered regions (IDRs), two of which were predicted to have Prion-like domains (PrLD) (Fig. S4). Proteins containing nucleic acid binding motifs, IDRs, and PrLDs are required for the formation of biomolecular condensates through liquid-liquid phase separation (43, 44). It is noteworthy that condensate formation has been previously associated with the regulation of auxin (44) and ethylene signaling and response in Arabidopsis (45). Together these observations led us to hypothesize that SlERF.D6 is an important contributor to tomato fruit maturation and ripening control in part through interaction with previously described central ripening TFs.
Delayed fruit ripening in SlERF.D6 gene-edited mutants
To assess SlERF.D6 function, three independent CRISPR edited (Cr) lines (Cr3, Cr8, Cr7) harboring distinct mutations in the gene were selected for further characterization (Fig. 2a). All three lines showed similar delayed fruit ripening initiation of approximately 5 days, uneven ripening and slower attainment of full color (Fig. 2b, 2c). We selected the Cr3 and Cr8 lines for detailed analysis. Consistent with delayed pigment accumulation, peak ethylene evolution in the mutants was reduced by 70% as compared to WT controls, occurring 3 days later than WT (Fig. 2d). Fruit texture analysis showed softening of fruits was delayed significantly during early ripening but achieved similar firmness to WT controls in very mature fruit at 50 days post anthesis (DPA) (Fig. 2e). Though substantially delayed in ripening onset and color development, the mutant and WT fruit shared similar post-harvest water loss phenotypes and cuticle integrity as assessed by Toluidine blue staining (Fig. S5).
We further examined carotenoid accumulation in SlERF.D6 mutant lines. Pericarp tissues of all Cr mutants initially accumulated less lycopene (red) and β-carotene (orange) when compared to WT fruits (Fig. 2f, 2g), though by 50 DPA carotenoid levels in very mature mutant and WT fruit became similar. In addition to changes in pericarp coloration and carotenoid profiles, SlERF.D6 CRISPR mutants also displayed altered locule pigmentation (Fig. S6). Mutant lines presented lighter green locules than WT at both 35 DPA and 39 DPA (Fig. S6a) resulting from reduced chlorophyll a/b contents (Fig. S6b, c). The predominant carotenoids in the locules, lycopene, beta-carotene and lutein, varied with lutein notably lower in early ripening mutant fruit with changes largely overcome by late ripening (50 DPA) (Fig. S6d-g).
Transcriptome profiling of SlERF.D6 mutants
To identify transcriptome changes resulting from SlERF.D6 mutation, RNA-Seq analysis was performed on fruit pericarp and locular tissues from WT and two independent mutant lines at the 29 DPA, 33 DPA, 35 DPA, 39 DPA, and 42 DPA stages (Fig. S7a-b). In WT fruit 33 DPA and 35 DPA correspond to the Mature Green (MG) and Breaker (BR) stages, respectively. To align fruit developmentally for comparison of similar appearing fruit of different ages, additional RNA-Seq was conducted on 46 DPA (~ BR + 7) and 54 DPA (~ BR + 15) mutant and 50 DPA (~ BR + 15) WT (Fig. S7a-b) fruit. Principal component analysis (PCA) showed that 29 DPA and 33 DPA belonged to the pre-ripening cluster in pericarp tissue among all lines. PCA divergence began at 35 DPA when WT initiated ripening (BR) while mutants remained more similar to the unripe MG stage (Fig. S7a). In locule tissue, the divergence began at 33 DPA (Fig. S7b) as some ripening factors altered expression (Fig S1a-d). All genotypes clustered at the very end of measured development, consistent with their similar appearance and coloration (Fig. 2b).
Analysis of differentially expressed genes (DEGs; fold change ≥ 2 and adjusted P < 0.05) in all tissues revealed a substantial perturbation of the transcriptome by the SlERF.D6 mutation (Dataset S3). We focused on differences in 35 DPA and 33 DPA pericarp and locule tissues, respectively, as this is when ripening changes became manifest in WT fruit but were still not evident in the mutants. A total of 874 upregulated and 683 downregulated genes were identified in 35 DPA pericarp tissues, as well as 714 and 1114, respectively, in 33 DPA locular tissues, in both Cr3 and Cr8 lines compared to the WT controls (Figure S6c-d).
Gene Ontology (GO) term enrichment analysis was conducted for the identified DEGs in 35 DPA pericarp and 33 DPA locule. A total of 19 and 59 GO terms enriched in upregulated and downregulated genes, respectively, were identified in 35 DPA pericarp tissues, as well as 26 and 41, respectively, in 33 DPA locular tissues (Dataset S4). Cell wall related and membrane integrity related GO terms were enriched in upregulated genes in both tissues (Fig. S7e-f). Pigment biosynthetic process, fruit ripening, carotenoid metabolism, flavonoid biosynthetic and ethylene biosynthetic processes were enriched in downregulated genes in 35 DPA mutant pericarp tissue as compared to WT (Fig. S7g). For 33 DPA locule tissue, photosystem associated pathways were the predominant GO terms enriched for downregulated genes (Fig. S7h). The photosystem and carotenoid gene expression changes were consistent with the coloration changes observed in the locule tissue (Fig. S6). Ethylene activated signaling pathway was also downregulated in this tissue (Fig. S7h).
Altered cell wall and ripening-associated genes in SlERF.D6 mutant fruit
Upon further mining of our GO analysis results, we found that despite delayed ripening initiation, the mutants displayed some upregulation of cell wall genes prior to ripening. Investigation of cell wall associated DEGs revealed six and two of the most differently expressed genes in the pericarp and locule tissue of the mutants versus WT, respectively (Fig. S8a, b). These genes encode cellulose synthases (Fig. S8a, h), COBRA-like protein (Fig. S8d), expansin-associated proteins (Fig. S8e, i), BURP domain-containing protein (Fig. S8f), pectin acetylesterase like protein (Fig. S8g), and endotransglucosylase/XTH8 (Fig. S8j). All have reported or suggested roles in cell wall structure and/or remodeling (46–50), were upregulated at earlier ripening stages compared to WT and achieved WT expression levels during later development and ripening in the mutants. These changes alone obviously were not sufficient to alter fruit texture and some, for example increased expression of cellulose synthase, would be consistent with enhanced cell wall integrity more typical of pre-ripening fruit.
Investigation of additional cell wall regulators which were reported recently to be important for textural changes of ripening tomatoes (36), showed that LOB1, EXP1, E6, CEL2, PL1-27, and GASA were all suppressed in the mutants at early stages of ripening in pericarp (Fig. S9a-c). In locule, PL1-27 and GASA were notable for being suppressed even earlier from 29 DPA to 35 DPA, consistent with the earlier expression of SlERF.D6 in this tissue, with continued suppression into later development (Fig. S9d, e).
PG2a and PL, classified under the GO term “fruit ripening”, were also inhibited in early ripening (Fig. S9f, g). Another four ripening- and ethylene-associated genes, ACS2, ACS4, E8 and E8-Homolog were repressed at the same stages (Fig. S9h-k). The decreased expression levels of ACS2 and ACS4 likely contributed to the reduced production of ethylene observed in the mutants. Additionally, RIN-MADS, NOR-NAC, CNR-SPL, DML2 and FUL1-MADS were notable for being strongly suppressed in 35 DPA pericarp and 33 DPA locule tissues (Fig. S9l, m), consistent with the ripening delay observed in mutant fruit.
Altered ethylene signaling and photosystem associated pathways in the locular tissues of the SlERF.D6 mutants
Based on GO term analysis of downregulated genes in 33 DPA locule (Fig. S7h), sixty DEGs in the ethylene-activated signaling and photosystem associated pathways (Fig. S10a, b) were identified and those most significantly altered were shown (Fig. S10c-e). Ethylene receptor ETR4 was consistently repressed from 33 DPA to 39 DPA (Fig. S10c), while receptor ETR3 was most highly suppressed at 33 DPA (Fig. S10d). Both receptor genes are involved in ripening (51–53). At 33 DPA, EIL4 and ERF.E1, primary and secondary ethylene responsive factors, along with one AP2/ERF transcription factor SlDEAR2 and phosphate transporter were also inhibited (Fig. S10d).
Genes encoding photosystem proteins including chlorophyll a-b binding proteins (CABs), PsaD subunit of photosystem I, photosystem II 10kDa polypeptide (PsbR) were suppressed at early ripening, consistent with the lower chlorophyll content in locule tissues of the mutants (Fig. S10e; Fig.S6a, b). Two other photosystem associated genes, ERD15 (encoding an early response to dehydration 15) and HSMT (encoding a homocysteine S-methyltransferase) were also repressed in early ripening (Fig. S10e).
Hindered pigment biosynthetic and metabolic processes in SlERF.D6 mutants
Consistent with the clear carotenoid phenotypes observed in the mutants, twenty DEGs related to pigment biosynthetic, carotenoid metabolic and flavonoid biosynthetic processes were identified in 35 DPA pericarp (Fig. S11a, b). Five genes, 1-D-deoxyxylulose 5-phosphate synthase (DXS), pheophytinase (PPH), phytoene synthase 1 (PSY1), alcohol acyl transferase (AAT1), and green flesh (GF) were among the strongest differentially expressed (Fig. S11b-f). GF encodes a stay-green protein associated with chlorophyll retention (54) and whose repression is consistent with the delayed color change (Fig. 2) of SlERF.D6 mutant fruit.
Given the carotenoid phenotypes of fruit locules, we analyzed the expression of carotenoid biosynthetic genes in that tissue as well (Fig. S12). Several carotenoid pathway genes were altered in expression in SlERF.D6 mutant lines from prior to committed carotenoid synthesis (DXS) to key rate limiting steps (PSY1, ZISO) through downstream pathway activities (CrtR-B2) (Fig. S12a-d), all consistent with the altered coloration phenotypes of delayed and modified carotenoid accumulation in the pericarp and locule, respectively.
Accelerated ripening initiation and reduced fruit size in SlERF.D6 overexpression lines
To confirm the positive role of SlERF.D6 in ripening, we generated transgenic SlERF.D6 overexpression (OE) plants and three independent lines were selected for further analysis (Fig. 3a). The OE lines displayed accelerated ripening initiation (Fig. 3b, c) along with earlier ethylene induction and enhanced fruit softening (Fig. 3d, e). OE lines also presented smaller fruit size and reduced fruit weight (Fig. 3f, g). OE lines additionally showed altered plant vegetative phenotypes including epinasty (Fig. S13a), constitutive triple response seedlings suggestive of elevated ethylene and/or response (Fig. S13b), elongated and curled leaf morphology (Fig. S13c), and larger flower size with longer sepals (Fig. S13d, e). All fruit phenotypes were opposite of those observed in gene-edited mutant lines and vegetative phenotypes were consistent with elevated ethylene response phenotypes.
Fruit RNA-Seq data showed numerous ripening associated genes were upregulated in a representative overexpression line, OE1 (Dataset S5, Fig. S14). Ripening regulatory transcription factors RIN-MADS, NOR-NAC, CNR-SPL and DML2 were upregulated at one or more stages of fruit ripening (Fig. S14a-d) reiterating the positioning of SlERF.D6 in context of previously described key regulators and consistent with the gene-editing loss-of-function results. Carotenoid biosynthesis pathway gene PSY1 was also upregulated in OE1 at 33 DPA and 35 DPA (Fig. S14e). Meanwhile, LCYB1 was downregulated in the same tissues (Fig. S14f). As PSY1 and LCYB1 are regulated positively and negatively by ethylene, respectively (Alba et al., 2005), to promote carotenoid synthesis and block lycopene catabolism, these observations are consistent with a role of SlERF.D6 in mediating ethylene regulation which in turn controls fruit lycopene accumulation. Fruit cell wall metabolism genes PG2a and PL were also upregulated at the beginning of maturation in OE1, consistent with enhanced and early softening (Fig. 3e; S14g, h). Ethylene biosynthesis and responsive genes ACS2, ACS4, E4, and E8 were upregulated early in OE1 fruit as well (Fig. S14i-l).
Primary target genes of SlERF.D6
DNA affinity purification sequencing (DAP-Seq) was carried out to further investigate the potential targets of SlERF.D6. A total of 11,772 highly reliable binding sites associated with 3,317 target genes in the tomato genome were identified (Supplementary Data 5). The most significant motif (E-value = 8.7 × 10− 1037) found by the programs MEME, DREME, and CentriMo (55) in the DAP-Seq data was the GCC box (Fig. S15a), which is known as the conserved binding site of ERFs (56, 57).
Among the targets identified by DAP-Seq, multiple contribute to or are associated with fruit ripening (Dataset S6): AP2/ERF family members (ERF.F7, Solyc03g006320; SlERF3-19, Solyc03g119800; SlDEAR2, Solyc04g078640; ERF.H9, Solyc07g042230; SlERF12-4, Solyc12g009490), tetratricopeptide repeat protein (TPR5, Solyc05g008420), ABA receptors (SlRCAR12, Solyc03g007310; SlRCAR15, Solyc05g052420), auxin associated genes (ARF5, Solyc04g081240; ARF7, Solyc07g042260; SAUR65, Solyc11g011730), bHLH transcription factors (bHLH3, Solyc01g081090; bHLH31, Solyc04g007430; bHLH61, Solyc09g089870), zinc finger transcription factors (ZFP45, Solyc06g069440; ZFP69, Solyc11g069340), aldehyde dehydrogenase (SlALDH2B1, Solyc02g086970), and UDP-glycosyltransferase (SlUDPGT19, Solyc03g078780).
Comparing with the transcriptome data of loss-of-function mutants, we identified 428 DEGs that had SlERF.D6 consensus binding sites on their promoters while 1197 DEGs that had binding sites in their gene bodies (Fig. S15b, c; Dataset S7). By investigating genes with binding sites in their bodies, we found twelve previously and mostly functionally described genes: SlTCP12, SlARF2a, SlGGPPS2, SlZDS, CrtR-B1, SlPL, Expansin1, SlE4, SlE8, SlERF.A2, SlERF.A3, and ERF.F12 (Fig. S16). Gene sequences themselves are known to serve as sites of regulatory information as has been widely reported in human cells, maize, rice, wheat, and Arabidopsis (58–63).
Among genes with promoter binding sites included three target genes of SlERF.D6, SlDEAR2, SlRCAR12, and SlUDPGT19 (Fig. S15d-f) that were significantly downregulated in edited lines (Fig. S15g-i). UDP-glycosyltransferase has been reported to play a role in fruit ripening (64, 65). Co-silencing SlRCARs including SlRCAR12 delayed fruit ripening in tomato (66). AP2/ERF transcription factor SlDEAR2 was previously linked with fruit ripening (67) as a possible regulator of the fruit ripening associated TF SlTCP12 (Solyc11g020670) based on co-expression and yeast one hybrid analyses (68).
Impaired fruit ripening in CRISPR edited mutants of SlERDF.D6 target genes
Because the TF SlDEAR2 is a target of SlERF.D6 and has been implicated in ripening but not functionally studied to date, we generated CRISPR/Cas9 editing mutations at the SlDEAR2 locus in tomato cultivar Ailsa Craig. Three independent CRISPR edited lines (CR1, CR10, CR13) harboring distinct mutations in the gene were selected for further characterization (Fig. S17a). For CRISPR edited line CR1, 1 bp base was replaced and 8 bp bases were deleted resulting in a truncated protein (11 amino acids total predicted length). The CR10 and CR13 mutant lines have a 1 bp deletion and 4 bp deletion resulting in truncated proteins of 21 and 20 aa, respectively. All three lines showed similar delayed fruit ripening initiation of approximately 4 days (Fig. 4a; Fig. S17b) resembling SlERF.D6 edited mutants (Fig. 2). Consistent with the delayed and slower pigment accumulation phenotype (Fig. S17c), ethylene emission and softening of fruits in the mutants were impaired (Fig. S17d, e). Unlike SlERF.D6 edited mutants, SlDEAR2 mutants’ locular tissues showed the same color as the WT lines (Fig. S18a). SlDEAR2 was expressed both in pericarp and locular tissues (Fig. S18b), indicating that the locular color changed in SlERF.D6 edited mutants may be caused by other factors and the effects of this gene downstream of SlERF.D6 are more prominent in the pericarp.
To gain further insight into the role of SlERF.D6 in regulating the ripening cascade, we also edited the TF SlTCP12, a downstream target of SlDEAR2. Its expression pattern during fruit development was similar to that of both SlERF.D6 and SlDEAR2 (Fig. S19a). Resulting mutants showed similar delayed ripening initiation (Fig. 4b; Fig. S19b, c) and ethylene emission (Fig. S19d) as mutants in genes encoding its upstream regulators, strongly indicating that this TF is in a direct regulatory chain led by SlERF.D6. In summary, we provide novel information describing three TFs working in a sequential ripening regulatory cascade where each is necessary as defined by genetic mutation in normal ripening manifestation. Furthermore, we can place this regulatory module in the context of additional central and essential ripening regulators and the ripening hormone ethylene.