Role of the tomato MARS1/ROUGH gene encoding a LYSINE‐SPECIFIC HISTONE DEMETHYLASE 1 in adventitious root and fruit skin formation

ABSTRACT

In contrast to animals, plants possess a remarkable regenerative capacity and are capable of forming new organs or even entire organisms from a limited number of cells present in adult tissues, in response to injury or environmental changes. In this study, we describe the isolation and characterization of the more adventitious roots1‐1 (mars1‐1) mutant, which exhibits enhanced regenerative potential upon wounding in tomato hypocotyl explants. Additionally, the mars1‐1 fruits exhibited a rough surface because of ectopic subepidermal cell proliferation, resulting in the formation of callus‐like structures on their cuticle. The MARS1/ROUGH gene encodes a conserved lysine‐specific histone demethylase, SlLDL1, which regulates processes such as cell proliferation, stem cell pluripotency, and embryogenesis in metazoans. Two CRISPR/Cas9 null alleles, mars1‐2 and mars1‐3, were generated, and their pleiotropic phenotypes were characterized. ChIP‐seq analysis revealed elevated levels of methylated lysine 4 in histone 3 within specific regions of the mars1/rough genome. To determine the impact of altered epigenetic marks on gene expression regulation in mars1/rough mutants, we analyzed the transcriptome of tomato hypocotyls during adventitious root formation. Using specific bioinformatic workflows and the resolution of the directional RNA‐seq data, we identified several dozen distinct genomic regions with de novo expression in the mars1/rough mutants. One such region includes a novel B‐type cyclin gene that is upregulated in mars1/rough mutants and may account for adventitious root and fruit skin phenotypes. Our findings indicate that SlLDL1 plays a role in maintaining silencing in specific genomic regions that are essential for tissue‐specific reprogramming.

The tomato mars1/rough mutant displays enhanced root regeneration and rough fruits due to ectopic cell proliferation. The causal gene encodes a lysine‐specific histone demethylase that normally maintains gene silencing. Its loss alters histone methylation, upregulating several genes, including those B‐type cyclins involved in tissue‐specific cellular reprogramming.

INTRODUCTION

Plants display an impressive capacity for regeneration, enabling them to repair damaged tissues and even regenerate lost organs (Xu and Huang, 2014; Ikeuchi et al., 2016, 2019). In the context of environmental stress, the regeneration of damaged plant parts has been demonstrated to be a consequence of hormonal crosstalk, thereby restoring normal tissue function. Following injury, the regeneration of missing organs can occur in various tissues, resulting in the formation of adventitious roots (ARs) or shoots. Studies conducted on the model plant Arabidopsis thaliana have identified several transcription factors and genes involved in hormone crosstalk as key players in wound‐induced organ regeneration (Mathew and Prasad, 2021).

The tomato cultivar Micro‐Tom (MT) serves as an excellent model for research in Solanaceae because of its similarities with Arabidopsis, including its small size, short life cycle, and simplicity of transformation (Shikata and Ezura, 2016). The implementation of an optimized CRISPR‐Cas9 editing system enables the utilization of reverse genetics methodologies (Dahan‐Meir et al., 2018). In previous studies (Alaguero‐Cordovilla et al., 2021; Larriba et al., 2021a, 2021b, 2022), we established a new experimental system to study regeneration in tomato, using hypocotyl explants of MT to simultaneously study wound‐induced de novo shoot formation and AR development without exogenous application of hormones. In this experimental system, the cellular reprogramming of the founder cells of the new organs requires precise spatial and temporal regulation of auxin and cytokinin (CK) gradients in the most apical and basal regions of the explants, coupled with extensive metabolic and dynamic homeostasis of reactive oxygen species (ROS). We also identified several conserved transcription factor modules, such as those homologs of WIND1, ESR1/2, WOX13 (WOUND‐INDUCED DEDIFFERENTIATION 1, ENHANCER OF SHOOT REGENERATION 1/2, WUSCHEL‐RELATED HOMEOBOX 13), and SLR/IAA14, ARF7/ARF19, LBD (SOLITARY ROOT/INDOLE‐3‐ACETIC ACID 14, AUXIN RESPONSE FACTOR 7/19, LATERAL ORGAN BOUNDARIES‐DOMAIN), involved in the specification of callus and root identity during de novo shoot and AR formation, respectively. We also reported significant variation in the expression of long noncoding RNAs (lncRNAs), depending on both the region studied and time (Larriba et al., 2021a), suggesting a regulatory role of these noncoding sequences in wound‐induced organ formation in tomato.

Recent work in tomato hypocotyl explants (Yang et al., 2024) characterized loss‐of‐function mutants of REGENERATION FACTOR 1 (REF1; Solyc04g072310), which exhibited reduced callus formation and shoot regeneration upon exogenous hormone treatment. In contrast, plants overexpressing REF1 presented significantly increased regeneration capacity. REF1 encodes a small peptide that induces the expression of SlWIND1, the tomato homolog of WIND1 in Arabidopsis that promotes wound‐induced callus formation in many species (Iwase et al., 2017). Furthermore, a key role for GROWTH‐REGULATING FACTOR 1 (SlGRF1; Solyc04g077510) as a central transcriptional activator of de novo shoot formation from hypocotyl explants has been established through the combination of single‐cell RNA sequencing, transcriptomic analysis over time, and genetic validation (Xu et al., 2026). REF1 and SlGRF1 do not appear to influence wound‐induced AR formation, suggesting that distinct genetic pathways regulate shoot and root formation in tomato hypocotyl explants.

To elucidate the genetic mechanisms that modulate wound‐induced AR formation in tomato explants, we screened a highly mutagenized ethyl methanesulfonate (EMS) mutant population generated from the miniature cultivar MT at INRA Bordeaux (Just et al., 2013). Most of the AR mutants identified in our screen displayed some defects in AR initiation, which caused a complete absence or a reduced number of ARs. In this work, we performed molecular characterization of mutants with an increased number of ARs, which might be defective in some of the negative regulators of wound‐induced AR formation, and we termed them more adventitious roots (mars) mutants. One of these mutants, mars1/rough, was also independently identified in a genetic screen for tomato mutants affected by fruit cuticle cracking. The present study reports the results of a genetic and molecular analysis of these mutants. Our findings indicate that MARS1/ROUGH encodes a protein that is homologous to LYSINE‐SPECIFIC HISTONE DEMETHYLASE 1 (LSD1), also known as LYSINE (K)‐SPECIFIC DEMETHYLASE 1 A (KDM1A) (Shi et al., 2004), and plays a key role in a novel genomic surveillance mechanism to broadly downregulate silenced regions within euchromatin.

RESULTS

Identification of mutants affected in wound‐induced adventitious root formation and fruit skin formation

A highly mutagenized collection of lines generated by EMS in S. lycopersicum cv. MT was evaluated for early seedling and root system architecture mutant phenotypes. A small number of the lines presented seedlings with an increased number of ARs following the excision of the entire root system (Alaguero‐Cordovilla et al., 2020). These findings may have been the consequence of mutations that affect the expression of negative regulators of wound‐induced AR formation. Compared with their MT siblings, the P19E6 M₃ line presented a segregation pattern for mutant seedlings, characterized by a significant increase in wound‐induced ARs (Figure 1A, B). The observed phenotypic proportions were consistent with the recessive inheritance of the mutant phenotype. The primary root (PR) length was significantly shorter in the P19E6 mutants than in the MT background (Figure 1C), mainly because of the light delay in germination observed in the mutants (Figure S1A). The root hairs in the P19E6 mutants were closer to each other (Figure 1D), although there were no significant differences in length compared with those in the MT siblings (Figure S1B). The number of lateral roots (LRs) was subsequently quantified following excision of the PR tip, as previously described (Alaguero‐Cordovilla et al., 2020). No significant differences were observed between the P19E6 mutants and MT in terms of the number or density of LRs (Figure S1C–E).

Figure 1.

Morphological phenotypes of the P19E6 mutant

(A) Representative images of rooted shoot explants at 14 d after excision (dae). The number of ARs is indicated. (B, C) Box plots of AR number (B) and PR length (C). The numbers indicate the means ± standard deviations (SD). The P values are indicated (Student's t‐test; P < 0.05; n > 18). (D) Representative images of the differentiated region of the PRs, which exhibited fully developed root hairs. (E–G) Observations of the fruit surface in MT and the P19E6 mutant at the red ripe stage. (H) Toluidine blue staining of the outer pericarp section of the fruit of MT and P19E6 fruits at the mature green stage. E1, outer epidermal layer of the exocarp; E2 and E3 layers of the exocarp, and mesocarp layers. The asterisks indicate the presence of meristematic cells in the region of the outer pericarp where the knobs are found in the P19E6 mutant. (I) Representative images of the vegetative phenotypes of adult plants after 7 weeks of growth. Scale bars: 5 cm (I), 10 mm (A, E), 500 μm (F, G), and 150 μm (D, H).

The P19E6 mutation was also independently identified in a genetic screen for tomato mutants affected by fruit cuticle cracking. In the red ripe stage, compared with the MT fruits, this mutant exhibited cuticular defects on the fruit surface, including skin microcracks (Figure 1E). Detailed observation of the surface of P19E6 mutant fruits further revealed that lighter‐colored patches of tissue protruded outward from the fruit skin (Figure 1F), resulting in a rough surface in these areas (Figure 1G). A histological study of cross‐sections of fruits from P19E6 mutants confirmed the presence of supernumerary cells in the exocarp, leading to outgrowth of the fruit skin (Figure 1H). Uncontrolled outward cell proliferation was initiated as early as 5 d post‐anthesis (DPA) (Figure S1F). Analysis of the fruit cuticle of the P19E6 mutants revealed an almost threefold increase in the cutin polymer load (3.29 ± 0.23 mg/cm² in P19E6 and 1.23 ± 0.61 mg/cm² in MT fruits; P‐value = 0.005), whose composition was reminiscent of that of suberin, indicating an increased abundance of very long‐chain fatty acids such as 20–OH–C20:0 (Figure S1G). The wax load increased by almost twofold, but its composition remained unaffected. Compared with the MT plants, the P19E6 mutants also presented an altered floral phenotype, characterized by more branching and more flower trusses (Figure 1I). On the basis of the AR and fruit skin phenotypes, the P19E6 mutants are referred to herein as the more adventitious roots1‐1 (mars1‐1) or rough‐1.

Molecular identification of the MARS1/ROUGH gene encoding a LYSINE‐SPECIFIC HISTONE DEMETHYLASE 1‐LIKE (LDL) protein

Segregation analysis of the rough fruit phenotype in the BC₁F₂ population issued from a P19E6 × MT cross confirmed the presence of a single recessive mutation. To identify the mutation responsible for this phenotype, we used an established mapping‐by‐sequencing strategy (Garcia et al., 2016; Petit et al., 2016). Whole‐genome sequencing of a bulk of 44 BC₁F₂ plants displaying a rough fruit phenotype, followed by analysis of the allelic frequency distribution of EMS‐induced mutations, revealed a dramatic increase in mutation frequency on chromosome 10 (Figure 2A). Subsequent fine mapping using 768 BC₁F₂ recombinant plants allowed us to unequivocally identify the causal mutation of the rough fruit phenotype, which is a G‐to‐A transition in the coding region of Solyc10g047350 (Figure 2B). This gene encodes a protein that is homologous to the LSD1/KDM1A histone demethylase (Shi et al., 2004; Figures 2D, S2A). The tomato genome contains six genes that encode LSD1‐like (LDL) proteins (Li et al., 2020; this work). These have been renamed on the basis of their sequence homology to LDL proteins in A. thaliana (Figure S2B). The P19E6 mutation changes a glycine (G) to an arginine (R) residue at position 227 of the SlLDL1 protein, within the SWIRM (Swi3p, Rsc8p, and Moira) domain, which is found in many chromosomal proteins involved in chromatin modification (Da et al., 2006). This mutation affects a highly conserved glycine residue in all tomato and Arabidopsis LDL proteins (Figures 2E, F, S2C; Gao et al., 2012). LSD1/KDM1A also contains an amine oxidase domain (Figure S2B, C), which is responsible for catalyzing the oxidation of methylated lysines in several proteins, including the histone H3 tail (Shi et al., 2004).

Figure 2.

Identification of the gene responsible for the pleiotropic mars1/rough phenotype

(A) Mapping‐by‐sequencing of the P19E6 mutant. The pattern of mutation allelic frequencies along tomato chromosomes obtained in the mutant pool is represented by a green line. Chromosome 10 shows a dramatic increase in mutation allelic frequency. (B) Fine mapping of the causal mutation using the BC₁F₂ population. Recombinant scoring of 768 individual BC₁F₂ plants identified the causal mutation at position 40,585,035 on chromosome 10 (SL2.50, ITAG2.4). The P19E6 EMS mutant contains a G–to–A transversion at position 679 of the Solyc10g047350 gene, encoding a protein homologous to LYSINE‐SPECIFIC HISTONE DEMETHYLASE 1. (C) CRISPR/Cas9‐induced mutations in the first exon of Solyc10g047350 obtained from three independent homozygous lines. Bases highlighted in gray indicate the two guide RNA target sequences. Insertions and deletions are indicated in red. CRISPR/Cas9‐induced frameshift insertions and deletions lead to the formation of truncated proteins. (D) Phylogeny of full‐length LSD1‐like (LDL) homologs in representative plant species. A. thaliana and S. lycopersicum sequences are shown in green and blue, respectively. (E) Partial alignment of the SWIRM domains of the four LDL family proteins from A. thaliana (LDL1, LDL2, LDL3, and FLD) and the six LDL proteins from S. lycopersicum. The mars1‐1 mutation in the P19E6 line resulted in the replacement of a highly conserved glycine [G] with arginine [R] in SlLDL1. (F) 3D structure of the protein region surrounding the mutation found in the P19E6 line, according to the models predicted by AlphaFold (Jumper et al., 2021).

We next generated several gene‐edited CRISPR/Cas9 lines in MT and selected three of these lines with different mutations (Figure 2C; Table S1). Line mars1‐2 has a one‐base insertion between positions 287 and 288 and a one‐base deletion at position 315 from the ATG, which generates a frameshift and a truncated protein of 104 amino acids because of the appearance of a premature termination codon (Figure 2C). Line mars1‐3 presents the insertion of a base pair between positions 287 and 288 and the deletion of six base pairs at positions 310−315 from the ATG, resulting in the production of a truncated protein of 104 amino acids because of the appearance of a premature termination codon after the frameshift (Figure 2C). Gene editing of line mars1‐4 resulted in a four‐base pair deletion at position 316−319 from the ATG, which generated a truncated protein of 118 amino acids because of the appearance of a premature termination codon after the frameshift (Figure 2C). Our research revealed that the MARS1/ROUGH gene is expressed in different tomato tissues, including stems and young fruits (Figure S3). For further studies, we chose the gene‐edited lines mars1‐2 and mars1‐3, which exhibited distinct mutations but produced analogous mutant proteins. These proteins were composed solely of a short amino‐terminal region (104 amino acids), which was nonfunctional because of the absence of the SWIRM and amino oxidase domains.

Compared with the MT background, the mars1‐2 and mars1‐3 mutants exhibited increased AR production in shoot explants following whole root excision (Figure 3A, B), with more pronounced phenotypes than those observed in the mars1‐1 mutant (Figure 1). In addition, the rough fruit phenotype of mars1‐1 was more pronounced in both mars1‐2 and mars1‐3 than in mars1‐3, where it was concomitant with increased branching and flower production (Figures 3F, S4A–C). A detailed microscopic examination of the developing fruit exocarp revealed that mars1‐2 and mars1‐3 mutants exhibited aberrant periclinal division and subsequent cell proliferation in the fruit epidermis during the cell division phase (5−8 DPA). This resulted in the formation of disorganized epidermal cell masses on the fruit surface (Figure 3F, G). During the cell expansion phase (20 DPA), the loss of cell adhesion in these callus‐like structures was followed by the rupture of the fruit skin in protruding areas (Figure 3G). During the phenotypic characterization of the CRISPR/Cas9 mutants of SlLDL1, a significant proportion of the seeds of the mutants did not germinate or exhibited early lethality (Figure 3H, I). These results are consistent with the more severe molecular damage observed in the two CRISPR/Cas9 mutants than in the EMS‐induced mars1‐1 mutants. These findings suggest that mars1‐2 and mars1‐3 mutants might carry null mutations in the SlLDL1 gene, whereas mars1‐1 could be a hypomorphic mutation resulting in a mild loss of function in SlLDL1.

Figure 3.

AR formation and fruit phenotypes in the mars1/rough mutants

(A) Representative images of rooted shoot explants at 14 d after excision (dae). The number of ARs is indicated. Wild‐type images corresponding to the MT background are shown in Figure 1A. (B) Rooting capacity of shoot explants at 14 dae. (C, D) Rooting capacity of hypocotyl explants at 14 dae. (E) AR emergence in the explants. Diagrams (B–E) indicate the hypocotyl fragments used in the different experiments. Different letters in (B–E) indicate significant differences between conditions (Fisher's LSD; P < 0.05; n > 15). (F) Macroscopic observations of the fruit surface in mars1/rough mutants at the red ripe stage. Wild‐type images corresponding to the MT background are shown in Figure 1E. (G) Calcofluor staining and confocal scanning microscopy images of pericarp tissue cells of MT and mars1‐3 fruits; DPA, day post‐anthesis. The yellow asterisks indicate the presence of meristematic cells in the region of the outer pericarp where the knobs are found in the mars1/rough mutants. (H) Phenotypes of early seedlings. Emb: seedlings resembling mutants defective in embryo development; NG: non‐germinated seeds. Different letters indicate significant differences (Fisher's LSD; P < 0.05; n > 225) between genotypes. (I) Representative images of Emb phenotypes in mars1/rough mutants at 10 d after sowing. Scale bars: 1 mm (I), 10 mm (A, F), and 20 μm (G).

mars1/rough mutants affect hormone regulation during de novo organ formation

Previously, we characterized wound‐induced AR formation in young shoot and hypocotyl explants in several tomato cultivars, including MT (Alaguero‐Cordovilla et al., 2021). Our findings indicated that basipetal auxin transport from the shoot and local auxin biosynthesis triggered by wounding play pivotal roles in this process (Alaguero‐Cordovilla et al., 2021). We found that mars1/rough mutations also increased the rooting capacity of whole hypocotyl explants without the shoot (Figure 3C), as well as that of smaller explants obtained from sectioning hypocotyls in halves, corresponding to the apical or basal region of the hypocotyl (Figure 3D). The emergence of ARs was significantly delayed in the apical‐derived explants compared with the basal‐derived explants, regardless of genotype (Figure 3E). Transverse sections of the basal region of the hypocotyl revealed that the AR primordia arise from the proliferation of interfascicular cambial cells that are on the side facing the phloem poles and adjacent to the vascular bundle in both the MT background (Figure 4A, A′) and mars1/rough mutants (Figure 4B, B′). The observed increase in the number of ARs in the mars1/rough mutants cannot be attributed to the longitudinal expansion of the AR formative region at the base of the hypocotyl. Rather, it appears to be the result of an increase in the radial density of ARs, as evidenced by cross sections (Figure 4B). These results are consistent with the hypothesis that mars1/rough mutations could alter the response to the dynamic hormonal balance between auxins and CKs, which is necessary for the specification of new AR primordia from the cambium (Larriba et al., 2021b). The dual‐color transcriptional reporter DR5:mScarletI‐NLS;TCSn:mNeonGreen‐NLS was employed to study endogenous auxin and CK responses (Omary et al., 2022). At the time of wounding (0 h after excision, hae), the expression of the CK response marker TCSn:mNeonGreen‐NLS was observed in numerous cells of the vascular cylinder in the MT hypocotyls. In contrast, the expression of the auxin response marker DR5:mScarletI‐NLS was significantly lower in these cells, with some overlap between the two markers expressed in specific cells, as evidenced by green (CK response) and red (auxin response) fluorescence (Figure 4C). Similar results were observed in the mars1‐1 mutants (Figures 4D, S5). At 40 and 75 hae, the expression of the CK reporter (TCSn) in the vascular stele significantly decreased in both the MT and mars1‐1 mutants (Figures 4E, S5). The most conspicuous differences between MT and mars1‐1 were due to the higher number of auxin peaks after 40 hae in the latter genotype (Figures 4E, S5). These results indicate that compared with the MT background, the mars1‐1 mutant exhibited increased auxin responsiveness after wounding.

Figure 4.

Morphological, anatomical, and physiological changes in the basal regions of hypocotyl explants in mars1/rough mutants during AR formation

(A, B) Representative cross sections of the basal regions of the hypocotyls during AR formation in MT (A, A′) and mars1‐3 (B, B′). The AR primordia arise from the proliferation of the interfascicular cambium (ifc) on the side facing the phloem (ph) poles and adjacent to the vascular bundle; c, cambium; xl, xylem. Double asterisks indicate the quiescent center of the newest AR primordia. (C, D) Fluorescence microscopy and confocal microscopy images of MT (C, C′) and mars1‐1 (D, D′) hypocotyl sections expressing DR5:mScarletI‐NLS (pink) and TCSn:mNeonGreen‐NLS (green) markers (white ovals: auxin maximum defining the AR formative regions) at the indicated time after excision; (C′, D′) show AR primordia protruding from the hypocotyl. (E) Quantification of the phytohormonal responses in hypocotyl cross‐sections at different times during auxin‐induced AR formation. (F) Regeneration percentage and (G) rooting capacity of cotyledon explants incubated with 0.15 μM NAA for 21 d. (H) Callus fresh weight in cotyledon explants incubated with 2.5 μM 6‐BAP for 21 d. (I) Shoot regeneration stages of hypocotyl explants at 17 dae as described in Larriba et al. (2021a). (J) Representative images of de novo organ formation in the apical region of hypocotyl explants at 17 dae. Scale bars: 200 μm (A–D), and 1 mm (C′, D′, J). Different letters in (E–I) indicate significant differences between samples (Fisher's LSD; P < 0.05; n > 16).

Cotyledon explants of the mars1/rough mutants exhibit a heightened response to the exogenous addition of naphthaleneacetic acid (NAA), a synthetic auxin, as evidenced by their enhanced regeneration rate (Figure 4F) and rooting capacity (Figure 4G). Conversely, when cotyledon explants were incubated with the synthetic CK 6‐benzylaminopurine (6‐BAP), compared with their MT genetic background, the mars1/rough mutants exhibited a diminished capacity for callus formation (Figure 4H). These findings indicate that there has been a disruption in the endogenous signaling of the hormonal balance of auxins and CKs in the mutant tissues, as already suggested by the results obtained with the fluorescent markers TCSn:mNeonGreen‐NLS and DR5:mScarletI‐NLS (see above). The extent of de novo shoot formation in the apical region of hypocotyl explants was subsequently evaluated in accordance with previously described methods (Larriba et al., 2021a). While a substantial proportion of MT hypocotyl explants exhibited the ability to produce functional shoot apical meristems with fully developed leaves after 3 weeks (Figure 4I, J), the mars1/rough mutations significantly reduced the shoot regeneration capacity (Figure 4I, J). The mars1‐3 mutant exhibited a pronounced delay in de novo shoot regeneration, with most explants remaining in the stage of callus‐like tissue formation (i.e., stage 2) for a longer time (Figure 4I, J).

Gene expression profile at the base of the hypocotyl explants of mars1/rough mutants during the early stages of adventitious root formation

Two independent RNA‐seq experiments (Figure S6) yielded expression data for 22,578 protein‐coding genes (PCGs) and 1,206 lncRNAs (Figures S7, S8; Table S2). Principal component analysis and k‐means clustering of these datasets revealed substantial differences in transcript profiling over time and condition in both experiments, with less variation due to genotype (Figures S7A–D, S8A–D). A total of 1,027 PCGs and 67 lncRNAs were differentially expressed in the first experiment (Figure S7F; Table S3A), and 741 PCGs were differentially expressed in the second experiment (Figure S8F; Table S4A) in the mars1/rough mutant hypocotyl explants compared with the MT background at the different time points and conditions (Figure S6). Given the consistency of our results with an enhanced auxin‐to‐CK response in mars1/rough mutants, our subsequent focus was on the 122 auxin‐related and 88 CK‐related genes expressed in our RNA‐seq data (Table S5A, B). Most of the variation observed in the expression of auxin‐ and CK‐related genes (Figure S9A, B) was attributed to the time course and not to the genotype. However, certain genes related to endogenous auxin levels were dysregulated in the hypocotyls of mars1/rough mutants shortly after wounding. SlTAR3 (Solyc05g031600) exhibited slightly higher basal expression in mars1/rough mutants than in MT, and this expression increased significantly after wounding at all studied times (Figure S9C). On the other hand, SlARF24 (Solyc05g056040) was downregulated in mars1/rough mutants at 8 hae. Of note are the following genes associated with CKs that were significantly dysregulated in mars1/rough mutants at 8 hae: Solyc10g079930 (SlZOG2) and Solyc12g042600 (SlUGT73C1) (Figure S9D). These genes are hypothesized to encode zeatin O‐glucosyltransferase‐like enzymes that may play a role in negatively regulating CK homeostasis. These results suggest that locally altered auxin and CK homeostasis affects a subset of auxin responses in these mutants shortly after wounding, which is in line with the DR5:mScarletI‐NLS and TCSn:mNeonGreen‐NLS marker results obtained above.

Given that the most evident phenotype of mars1/rough mutants is increased cell proliferation, the focus was now placed on the 96 cell cycle‐related genes expressed in the RNA seq data. These genes were classified into 12 distinct functional groups (Table S5C). We subsequently sought to determine whether there were substantial discrepancies in their dysregulation between 0 hae and 8 hae, with particular attention given to the MT and mars1/rough mutants. The two categories, “DNA replication” and “Cyclin A”, were significantly different between the mars1/rough mutants and MT (Figure S9E). A comparative analysis of the expression patterns of genes involved in DNA replication revealed that the expression of SlMCM6 (Solyc02g082180) and SlPCNA (Solyc02g076660) was more highly induced in mars1/rough mutants than in MT at 8 hae (Figure S9F). These results suggest that compared with MT, mars1/rough mutants exhibit a greater increase in proliferative capacity following wounding.

A Venn diagram of all DEGs identified in both RNA‐seq experiments revealed that only 66 genes were significantly deregulated in all comparisons, with most of them being upregulated (n = 57) in the mars1/rough mutants (Figure 5A, B). No significant enrichment of GO terms was observed among these DEGs. Additionally, our observations indicated that a significant proportion of these deregulated DEGs in the mars1/rough mutants were in genomic regions that exhibited low or residual expression levels in the MT background but were flanked by genes whose expression levels were relatively high (Figure 5C). The fold changes in the expression of the PCGs and lncRNAs contained in 50‐Kb windows centered on these selected DEGs were determined and compared with the levels of deregulation in similar regions randomly selected within the genome (see Materials and Methods). The results unequivocally demonstrated that there was a substantial increase in the gene expression levels of the mars1/rough mutants around the selected DEGs (Figure 5D). These regions were enriched in the presence of intergenic lncRNAs compared with the scattered distribution of lncRNAs across the genome (Figure 5E). Our results indicate that the DEGs in mars1/rough mutants are mostly associated with genomic regions containing intergenic lncRNAs.

Figure 5.

Identification of commonly deregulated genes in mars1/rough mutants during AR formation

(A) Venn diagram of deregulated genes in the mars1/rough mutants compared with MT, as indicated in the main text and Figure S6A. The numbers indicate the number of DEGs found in each comparison (P < 0.01). Selected DEGs that were studied further are shown in white. (B, C) K‐means clustering of selected DEGs. Upregulated and downregulated clusters are shown in red (cluster UP) and blue (cluster DOWN), respectively. The colored peaks indicate the coverage of reads from multiple replicates at each time point, genotype, or condition. The MT samples are depicted in green, the mars1‐3 samples in red, and the mars1‐1 samples in orange. PCGs and lncRNAs are displayed at the top of each panel in blue and purple, respectively, with arrowheads indicating the direction of transcription. The red asterisks indicate genes within this region that were significantly upregulated in the mars1/rough mutants compared with MT. (D) Deregulation of neighboring PCGs in a 50‐Kb interval of the commonly deregulated DEGs in the mars1/rough mutants compared with MT. (E) The ratio of differentially expressed lncRNAs in the mars1/rough mutants compared to those in MT. Hierarchical ranking of the transcripts annotated as lncRNAs was performed on the basis of their positions in relation to these PCGs.

Thus, our RNA‐seq results suggest that the SlLDL1 protein encoded by the Solyc10g047350 gene may play a pivotal role in the silencing of specific genomic regions within euchromatin. To determine whether the mars1/rough mutations impact other non‐expressed regions of the genome, we conducted de novo transcript annotation (see Materials and Methods; Table S6). The transcripts were grouped on the basis of their expression levels, from the lowest (first quartile, Q1) to the highest (fourth quartile, Q4). This allowed us to determine that the most deregulated transcripts in the mars1/rough mutants were included in Q1 and, to a lesser extent, in Q2 (Figure 6A, B). Upon closer examination of the deregulated transcripts in Q1 in the mutants, it became evident that these transcripts exhibited a broad distribution along the entire chromosome (Figure 6C). Conversely, deregulated genes in the mutants assigned to Q2 and Q3 were predominantly located within euchromatin regions of the chromosome arms (Figure 6C). Notably, a significant proportion of the deregulated transcripts in Q1 were in intergenic regions where no transcripts had been previously assigned (Figure 6D).

Figure 6.

Transcriptome landscape of the mars1/rough mutants during AR initiation

(A) Clustering of expressed transcripts at the base of the hypocotyl during AR initiation was performed based on their expression levels, from the lowest expression (first quartile, Q1) to the highest expression (fourth quartile, Q4). The rows were clustered according to the fold change in expression between the mutants (mars1‐1 and mars1‐3) and the MT datasets. The columns were clustered using the Euclidean distance between the different libraries. The relative expression levels in each row are presented on a scale of 0 (blue) to 1 (yellow). The fold change in expression ranged from −8.85 (blue) to 11.04 (red). (B) Box plot of the fold change in expression of the expressed transcripts in the studied quartiles. The red lines connect the maximum and the minimum values in each quartile. The P‐values are indicated, with those in italics indicating statistical significance (Student's t‐test; P < 0.05; n = 6,697). (C) Fold change in expression of the expressed transcripts along the chromosome (in bp). (D) Functional classification of transcriptional regions based on whether they contain at least one PCG or lncRNA (green), intergenic transcriptional regions that do not contain a PCG or lncRNA (purple), or regions with transcriptional units not included in the above categories (orange).

SlLDL1 is involved in the maintenance of methylated histone 3 lysine 4

To verify that the pleiotropic phenotype of the mars1/rough mutants is indeed a result of the inactivation of the Solyc10g047350 gene, which encodes the SlLDL1 protein, seedlings that express the SlLDL1 coding region under the control of the Solyc09g010800 constitutive promoter (proPMtb) were obtained (see Materials and Methods). The mars1‐3 mutants transformed with the empty vector produced rough fruits characterized by epidermal knobs associated with cell proliferation in the exocarp (Figure 7A). In contrast, the mars1‐3 plants transformed with the proPMtb:SlLDL1:T35S construct displayed smooth fruits with minimal protuberances (Figure 7B). These T₁ fruits exhibited a morphology consistent with that of the MT genetic background (Figure 1E), confirming that SlLDL1 upregulation can rescue the wild‐type fruit phenotype in mars1‐3 mutants. Furthermore, we confirmed that, in T₂, the constitutive expression of SlLDL1 in mars1‐3 restored the wild‐type AR formation phenotype in shoot explants after wounding (Figure 7C, D). These findings serve to substantiate the hypothesis that Solyc10g047350 is indeed MARS1/ROUGH.

Figure 7.

MARS1/ROUGH encodes SlLDL1, which is required for histone H3 lysine 4 demethylation

(A–C) Genetic complementation of mars1‐3 mutants with the SlLDL1 genomic sequence under the control of the ubiquitous proPMtb promoter from the Solyc09g010800 gene. Representative images of T₁ plants and mature fruits of mars1‐3 mutants transformed with an empty vector (A) or with the proPMtb:SlLDL1:T35S construct (B). (C) Representative images of rooted T₂ shoot explants 10 d after excision (dae). (D) Boxplot showing the number of ARs for each genotype (n = 6). The rescue of the wild‐type phenotypes in (B, C) confirms that SlLDL1 (Solyc10g047350) is MARS1/ROUGH. Asterisks indicate highly significant differences with respect to the MT background (Student's t‐test; P < 0.01). Scale bars: 10 mm.

In metazoans, LSD1/KDM1A demethylates monomethylated or dimethylated histone H3 lysine 4 (H3K4; Shi et al., 2004) and H3K9 (Metzger et al., 2005), which are associated with both gene repression and gene activation, respectively. To determine which specific methylation was significantly altered by mars1/rough mutations, we conducted immunoblot analysis with several antibodies to analyze the abundance of modified H3 proteins (see Materials and methods). The levels of H3K4me1 proteins were significantly greater in the mars1‐3 mutants than in the MT background (Figure S10A, B). The levels of H3K4me2 proteins also appeared to be higher in mars1‐3 mutants than in MT plants (Figure S10A, B). In contrast, no significant differences in the levels of H3K4me3 proteins were observed between the MT and mars1‐3 mutants (Figure S10A, B). The levels of two additional histone H3 marks associated with transcriptional activation, H3K36me3 and acetylated H3K9 (H3K9Ac) (Berr et al., 2011), were quantified, and no statistically significant differences were detected between the MT and mars1‐3 mutants (Figure S10C). These findings support the hypothesis that the tomato SlLDL1 protein regulates the demethylation of H3K4me2 and H3K4me1.

H3K4 methylation marks are altered in interspersed heterochromatin regions in mars1/rough mutants

Given the potential of immunoblotting to obscure gene‐specific effects, a ChIP‐seq experiment was conducted for H3K4me1, H3K4me2, and H3K4me3. Genome‐wide enrichment profiles of H3K4 methylation revealed a preference for H3K4me3 near the transcription start site (TSS) of genes (Figure S10D, E), in line with the established role of H3K4me3 as a marker of transcriptional activation (Guo et al., 2010). In contrast, H3K4me1 exhibited a uniform distribution across the gene bodies (Figure S10D, E). No significant differences were observed between the marks studied on the MT and mars1‐3 when their whole genome profiles were considered (Figure 10D, E). However, comparative analysis of peak gains in mars1‐3 mutants with respect to MT revealed specific genomic regions with significant fold enrichment (Figure S11). We subsequently investigated the potential impact of the observed alterations in H3K4 methylation profiles between mars1‐3 and MT on gene expression. An emphasis was placed on the DEGs of the Cg1 and Cg2 contrasts, thereby highlighting basal differences in gene expression between mars1‐3 and MT (Figures S6A, S12A–C). This enabled the identification of a clear correlation between increased H3K4 methylation levels and elevated gene expression in mars1‐3 compared with those in MT. In all these cases, H3K4me1 levels increase throughout the gene body. This increase is associated with higher levels of H3K4me2 and H3K4me3 near the TSS region and the activation of the expression of these genes in mars1/rough mutant tissues, which are otherwise silenced in MT. Concurrently, reduced levels of H3K4 methylation were observed in the genes whose expression was downregulated in the mars1‐3 mutants (Figure S12D–F).

Figure 10.

Downregulation of Solyc04g081650 and Solyc04g081660 using virus‐induced gene silencing (VIGS)

(A) Representative images of mature fruits of mars1‐3 mutants 14 d after being injected with Agrobacterium tumefaciens strains containing pTRV1 or pTRV2‐GG, either empty (Empty vector) or with Solyc04g081650 (VIGS‐1F1R), Solyc04g081660 (VIGS‐2F2R), and Solyc03g123760 (VIGS‐3F3R, PHYTOENE DESATURASE [PDS]) probes. The white arrowheads in (A) indicate the injection point(s). The dashed lines indicate the extent of PDS downregulation with the VIGS‐3F3R probe. The numbers in the lower left corner represent the number of fruits that exhibited the observed phenotype relative to the total number of assayed fruits. (B) Details of the fruit surface displaying tissue outgrowths in these fruits. White asterisks indicate the position of the epidermal knobs. Note the absence of these structures in the epidermis of mature mars1‐3 fruits inoculated with VIGS‐1F1R or VIGS‐1F1R and VIGS‐2F2R [VIGS‐1F1R + 2F2R] but not with VIGS‐2F2R. (C) Representative images of rooted shoot explants of mars1‐3 at 6 d after excision (dae). (D) Boxplot showing the number of ARs for each treatment (n = 7). Asterisks indicate highly significant differences between treatments (Student's t‐test; P < 0.01). Scale bars: 10 mm (B, C), and 1 mm (A).

The focus was subsequently directed toward the 66 genes that demonstrated significant deregulation across the all comparisons (Figure 8A, B). A thorough investigation into the distribution of H3K4 methylation signals within these genes revealed a substantial increase in H3K4me1 levels within the gene body of mars1‐3 mutants, in contrast to the reduced levels observed in MT. Furthermore, an increase in H3K4me2 and H3K4me3 levels was observed near the TSS in these genes in the mars1‐3 mutants relative to those in the MT (Figures 8C top panels, S13). In contrast, genes located at some distance from these regions, which were not found to be deregulated in the mars1‐3 mutants, did not show significant differences in H3K4 methylation levels between mars1‐3 and MT (Figures 8C bottom panels, S13). Eight genes were selected on the basis of the following criteria: Minimal expression in MT and a significant increase in expression in mars1/rough mutants (Figure S12G). Among these genes are Solyc06g035940, which encodes a member of the class IV homeodomain‐Leu zipper (HD‐ZIP IV) family, and Solyc12g044440, which encodes a zinc finger protein of unknown function. These observations underscore the local impact of mutations in mars1/rough on the demethylation status of H3K4 in silenced heterochromatin, which is adjacent to euchromatin domains.

Figure 8.

Comparative landscape of H3K4 methylation marks in MT and mars1/rough plants

(A, B) Average profile plots (A) and corresponding heatmaps (B) illustrating the distribution of ChIP‐seq signals across gene bodies from the transcription start site (TSS) to the transcription end site (TES) in the 66 DEGs in all comparisons between the MT and mars1‐3 genotypes. In (B), the differential occupancy plot (H3K4me_diff) highlights regions of increased (red) and decreased (blue) histone methylation in mars1‐3. (C) Metaplots showing the average ChIP‐seq signal intensity across a subset of genes (n = 57) identified as upregulated in all comparisons (top panels) or non‐regulated and randomly selected genes (bottom panels). The plot shows the normalized read density from 1.0 Kb upstream to 1.0 Kb downstream of the TSS and the TES. Red and blue lines represent the mean signal for each genotype, highlighting a significant increase in H3K4 methylation marks across gene bodies in the mars1‐3 background.

Functional validation of the candidate gene cluster responsible for the mars1/rough phenotype

Among the regions whose expression was upregulated in the mars1/rough mutants and whose H3K4 demethylation status was affected, a ∼100 Kb region on chromosome 4 (SL4.0ch04: 63,531,701–63,631,100) was identified as a potential candidate region for further investigation (Figures 9A, S13). This region contains an antisense lncRNA, Solyc04r023410, which was found to be upregulated in the mars1/rough mutants, as well as several other RNAs spanning the coding sequences of three contiguous genes: Solyc04g081650, Solyc04g081660, and Solyc04g081670 (Figure 9B). Our directional RNA‐seq allowed us to reconstruct seven different transcripts, including these three genes (Figure 9A). Solyc04g081660 and Solyc04g081650 have been shown to putatively encode proteins with partial homology to the N‐ and C‐terminal domains of mitotic B‐type cyclins, respectively. Both genes are expressed at very low levels in different tissues and across developmental stages in the M82 background, except in generative and sperm cells during pollen development (Figure S14). We confirmed by RT‐PCR that several transcripts in this region were upregulated in both young shoot explants and fruit skin in mars1‐1 and mars1‐3 mutants, with no detectable expression in their corresponding MT background tissues (Figure 9C). We cloned and sequenced five of the transcripts encompassing this genomic region, which were differentially expressed in shoot explants and fruit skin tissue in mars1/rough mutants (Figure 9D). Owing to the absence of expression of these transcripts in the MT background in our RNA‐seq data (Figure 9B), we hypothesize that their increased expression in the mars1/rough mutants may contribute to increased AR formation in the hypocotyl after wounding and uncontrolled cell proliferation in the fruit skin. In silico translation of the transcripts encompassing the Solyc04g081650 and Solyc04g081660 genes within this cluster revealed two peptides that partially encode a B‐type cyclin closely related to the G2/mitotic‐specific cyclin B2;4 encoded by Solyc04g082430 (Figure 9E). The peptide encoded by Solyc04g081650 encompassed the canonical C‐terminal domain of the cyclin protein (PF02984), with characteristic alpha‐helixes fully overlapping with the C‐terminal domain of cyclin B2;4 (Figure 9F). The peptide encoded by Solyc04g081660 showed some similarity to the N‐terminal region of the B‐type cyclins but lacked most of the internal amino acids present in the canonical N‐terminal domain sequence (PF00134) (Figure 9E). B‐type cyclins are negatively regulated by APC/C‐mediated ubiquitination through their DEAD‐box (Davey and Morgan, 2016), and we found a sequence matching the consensus DEAD‐box in the N‐terminal peptide encoded by Solyc04g081660 (Figure 9E).

Figure 9.

Functional validation of the gene cluster responsible for the mars1/rough phenotype

(A) A 20‐Kb region corresponding to the gene cluster that is highly deregulated in mars1/rough mutants. PCGs and lncRNAs are shown in dark blue and green, respectively, and the transcripts identified by RNA‐seq are shown in red. Light blue rectangles indicate products (1, 2) obtained from RT‐PCR. (B) Expression data (in raw counts) of the six genes and the lncRNA included in this region obtained by RNA‐seq (experiment 1). Different letters indicate highly significant differences between genotypes for a given gene (Student's t‐test; P < 0.01). (C) RT‐PCR of hypocotyl samples at time 0 (T0). RT‐PCR 1 and 2 correspond to partial amplification of the annotated genes Solyc04g081660 and Solyc04g081650, respectively. RT‐PCR of Solyc04g082430 (cyclin B2;4) was used as a control for cDNA quality. Water was used as a negative control (neg.). The numbers indicate the expected size of the amplified bands from the cDNA. (D) Novel transcripts expressed in mars1/rough mutants. The percentage was estimated on the basis of the number of clones identified for each transcript. (E) Phylogenetic tree of the closest relatives to cyclin B peptides encoded by Solyc04g081660 (N‐ter) and Solyc04g081650 (C‐ter). The protein sequences were aligned using Clustal Omega, and a phylogenetic tree was constructed with IQ‐TREE2, employing the JTT‐G4 model and iToL representation. (F) Alignment of the 3D structures of peptides encoded by Solyc04g082430 (cyclin B2;4, orange), Solyc04g081660 (N‐ter, purple), and Solyc04g081650 (C‐ter, green). The putative DEAD‐box domain is indicated in light green. The predicted template modeling (pTM) score of the 3D structure is indicated.

To test the hypothesis that the upregulation of the Solyc04g081660 and/or Solyc04g081650 genes is responsible for the mars1/rough phenotype, a virus‐induced gene silencing (VIGS) strategy (Fu et al., 2005) was employed to suppress the expression of these genes in mars1/rough mutants (see Materials and Methods). Agroinjection of immature fruits (8−10 DPA) of the mars1/rough mutants was performed, and the region surrounding the VIGS‐1F1R transformation, which targeted Solyc04g081650, did not exhibit the rough phenotype characteristic of the mature fruit epidermis of mars1‐3 mutants. However, the VIGS‐2F2R‐mediated downregulation of the Solyc04g081660 gene, encoding the N‐terminal cyclin B region, did not rescue the wild‐type epidermis phenotype in mars1‐3 mutants (Figure 10A, B). The use of both VIGS‐1F1R and VIGS‐2F1R constructs in a single transformation event led to the observation of mars1‐3 fruits that lacked callus‐like structures within their epidermis (Figure 10A, B). This phenomenon can be attributed to the absence of ectopic proliferation after VIGS induction when the VIGS‐1F1R probe was used; however, this response was not observed when the VIGS‐2F2R probe was used (Figure 10A, B). Furthermore, the use of both the VIGS‐1F1R and VIGS‐2F1R constructs in mars1‐3 shoot explants led to a substantial decrease in the number of ARs following wounding (Figure 10C, D). The results of our studies unequivocally demonstrate that several transcripts within the genomic region encompassing Solyc04g081650 and Solyc04g081660 are upregulated in mars1/rough mutants. We suggest that those transcripts encoding an N‐terminally truncated mitotic cyclin B protein have the potential to interfere with normal cell cycle progression.

DISCUSSION

Two independent mutant screenings revealed that the altered function of Solyc10g047350 results in mild developmental phenotypes. This gene encodes a putative demethylase of the histone marks H3K4me1/2, which is homologous to the LSD1/KDM1A protein (Shi et al., 2004) and henceforth designated SlLDL1 (this work). The altered function of SlLDL1 resulted in a pleiotropic phenotype throughout the tomato life cycle. Mutations in this gene affect different traits, including early seedling growth, AR formation, inflorescence architecture, fruit exocarp development, and embryonic development. On the basis of the AR and fruit pericarp phenotypes, we defined these mutants as mars1/rough.

The size and shape of tomato fruits are determined by the relative contributions of cell division and cell expansion, which are specific at both the tissue level and the temporal level (Renaudin et al., 2017). The pericarp tissues of the fruit are derived from the carpel walls and are composed of three layers of exocarp (E1 to E3), the central layers of the mesocarp, and two inner layers of the endocarp (I1 and I2). During early fruit development, the exocarp is the primary site of cell division, whereas cell expansion occurs mainly in the mesocarp and is closely associated with the progression of endoreduplication (Tourdot et al., 2024). Auxin and gibberellins play important roles in the coordination of cell division and expansion during the early stages of fruit development (Pattison and Catalá, 2012; Renaudin et al., 2023), and endogenous CKs regulate pericarp cell division (Gan et al., 2022). The majority of the divisions are anticlinal in the epidermal layer, whereas periclinal divisions are observed in the underlying subepidermal cells, leading to the formation of additional mesocarp layers (Renaudin et al., 2017). In the fruit epidermis of mars1/rough mutants, abnormal cell division patterns, which are absent in the ovary at anthesis, occur in the exocarp as soon as 5 DPA. The fruit epidermis, which normally consists of a single layer of cells with anticlinal divisions, undergoes periclinal divisions in mars1/rough mutants, which are normally largely limited to the E2 and E3 subepidermal cell layers. As a result, supernumerary cell layers appear in the fruit epidermis, eventually leading to the appearance of protuberances on the fruit surface. Later, during the cell expansion and ripening stages, uncontrolled cell division in the epidermis leads to the disruption of the fruit skin in the protruding areas. The observed fruit phenotype in the mars1/rough mutants is consistent with the hypothesis that a temporal dysregulation of cell division and/or cell division plane determination occurs in some cells of the fruit epidermis. B‐type cyclins are among the primary regulators of the G2/M transition, which is critical for the regulation of cell division and, consequently, fruit development. The results of this study suggest that peptides with incomplete cyclin B features, potentially encoded within the Solyc04g081660 and Solyc04g081650 genes, were upregulated in mars1/rough fruits. These peptides may play a role in the development of the rough fruit phenotype. The constitutive overexpression of a nondegradable cyclin B1 protein in transgenic tobacco seedlings resulted in strong developmental abnormalities due to impaired cell cytokinesis, which led to the formation of polynucleated cells (Weingartner et al., 2004). Conversely, the ectopic overexpression of APC10, one of the key subunits of the anaphase‐promoting complex/cyclosome (APC/C) that functions to degrade mitotic cyclins at the G2/M transition and exit from mitosis (Vodermaier, 2004), was found to increase the rates of cell division during the early stages of leaf development without affecting plant morphology (Eloy et al., 2011). However, the cellular phenotype of mars1/rough cell clusters in fruits resembles the ectopic constitutive expression of CYCD3;1 in Arabidopsis seedlings, which resulted in a shortened G1 phase and inhibited cell cycle exit, which in turn led to smaller cells and ectopic divisions in many different tissues through regulation of the RETINOBLASTOMA RELATED‐EF2 canonical pathway (Dewitte et al., 2003). We hypothesize that the C‐terminal B‐type cyclin peptide encoded by Solyc04g081650, which is upregulated in mars1/rough fruits, increases the mitotic rate of some epidermal cells in a cell‐autonomous manner, independent of its regulation by APC/C. These findings align with the short‐range activation of irregular divisions observed in the epidermis of mars1/rough fruits.

In tomato fruits, the cell expansion phase, after 10−12 DPA, is associated with a considerable increase in the thickness of the fruit cuticle, which is formed by epidermal cells (Reynoud et al., 2023). One of the upregulated genes in the mars1/rough mutants was CUTIN DEFICIENT 2 (CD2), Solyc06g035940, which encodes a member of the class IV homeodomain‐Leu zipper (HD‐ZIP IV) family (Nadakuduti et al., 2012). Mutations in CD2 have been shown to cause a significant reduction in the biosynthesis of cutin, the main cuticle polymer in tomato, and alter wax deposition (Isaacson et al., 2009; Nadakuduti et al., 2012). Accordingly, while wax load and composition were not altered in mars1/rough mutants, the cutin load was considerably increased, suggesting the coordinated upregulation of genes involved in cuticle polymer formation, as observed in other tomato cutin‐enriched mutants (Petit et al., 2021). Moreover, the composition of mars1/rough cuticle lipid polyesters revealed the signature of suberin, which normally covers underground plant organs and seeds. Suberin is also synthesized in aerial organs for wound healing, as shown in tomato fruit cuticle mutants (Lashbrooke et al., 2016), suggesting that suberin is synthesized in response to the disruption of the fruit epidermis of mars1/rough fruits, resulting in their russet color. Consequently, the complex and pleiotropic phenotype observed in the mars1/rough mutant fruits might be a consequence of altered expression of not one but several genes that are normally silenced through SlLDL1‐mediated H3K4 demethylation. A number of studies have demonstrated a direct correlation between the demethylation of repressor histone marks, facilitated by demethylases belonging to the Jumonji domain‐containing (JmjC) protein family, and the regulation of genes implicated in different stages of fruit development (Li et al., 2020, 2024; Zeng et al., 2025). SlJMJ6 and SlJMJ3 expressions promote fruit ripening through the removal of H3K27me3/me2 from the promoters of key ripening‐related genes (Li et al., 2020, 2024). More recently, the H3R2me2‐specific demethylase SlJMJ10 has been shown to regulate the transcription of brassinosteroid‐regulated genes involved in cell expansion during fruit growth (Zeng et al., 2025). To our knowledge, this is the first report of the involvement of a histone lysine demethylase of the LDL family in the regulation of cell proliferation in the tomato fruit epidermis.

In previous studies (Alaguero‐Cordovilla et al., 2021; Larriba et al., 2021a, 2021b, 2022), we characterized wound‐induced AR formation in tomato using explants of the MT cultivar. The prevailing mechanistic model incorporates dynamic spatial and temporal regulation of auxin and CK gradients in the apical and basal regions of the hypocotyl subsequent to wounding (Figure 11A), thereby inducing dedifferentiation (i.e., cell reprogramming) and cell cycle reactivation of certain vascular cambium cells, ultimately resulting in the formation of new AR primordia in the basal region of the hypocotyl. Further evidence of the causal role of auxins in this process is provided by the characterization of the aerial roots (aer) mutant, which produced excessive amounts of ARs in the stem because of a greater accumulation of auxins and the high induction of the SHOOTBORNE ROOTLESS (SBRL) gene (Kevei et al., 2024), which encodes the tomato ortholog of the Arabidopsis LOB DOMAIN‐CONTAINING PROTEIN 29 (LBD29) protein involved in lateral root initiation in Arabidopsis (Omary et al., 2022). ARABIDOPSIS TRITHORAX‐RELATED 2 (ATXR2) is a histone methyltransferase that facilitates the accumulation of H3K36me3 at the promoters of the LBD16 and LBD29 genes, which are activated during cellular dedifferentiation in response to elevated levels of endogenous auxin during hormone‐triggered root regeneration. ATXR2 is recruited to the LBD18 and LBD29 promoters through its interaction with the AUXIN RESPONSE FACTOR 7 (ARF7) and ARF19 transcription factors (Lee et al., 2017). Another methyltransferase responsible for H3K36me2/3‐mediated transcriptional activation is SET DOMAIN GROUP 8 (SDG8). Its deficiency results in a range of pleiotropic developmental abnormalities, including a reduction in wound‐induced AR formation in leaf explants (Zhang et al., 2019). SDG8 function is required to adequately transduce the wound signal to upregulate ANTHRANYLATE SYNTHASE α1 (ASA1), a tryptophan biosynthesis gene in the auxin production pathway, which is required to trigger formative cell division in the cambium (Zhang et al., 2019). Moreover, the formation of callus in response to hormones is contingent upon the stimulatory effect of auxin on the upregulation of SDG8, which in turn affects the status of H3K36me3 on some cell cycle‐related genes, including CYCLIN B1;1 (Ma et al., 2022). These results indicate that auxins act both upstream and downstream of epigenetic regulators such as SDG8 and ATRX2, which may complicate the interpretation of the pleiotropic phenotypes observed in most epigenetic mutants. Compared with MT plants, mars1/rough mutants respond more strongly to the endogenous auxin‐to‐CK ratio dynamically established at the base of the hypocotyl after wounding (Larriba et al., 2021b). This phenomenon could be attributed to a local increase in auxin biosynthesis and the inactivation of CKs after wounding in the mars1/rough mutants, which is consistent with the increased levels observed for SlTAR2 and SlZOG2 in our RNA‐seq data.

Figure 11.

Working model for wound‐induced organ formation in tomato explants and the involvement of SlLDL1 in cell cycle regulation

(A) Micro‐Tom (MT) shoot and hypocotyl explants can be used to study de novo adventitious root (AR) and shoot formation in response to wounding without the need for exogenous hormone application (Alaguero‐Cordovilla et al., 2020; Larriba et al., 2021a, 2021b, 2022). Cellular reprogramming of the founder cells of new organs requires precise spatial and temporal regulation of auxin and cytokinin (CK) responses in the apical and basal regions of explants. This regulation is altered in mars1/rough mutants, leading to increased AR formation and delayed shoot regeneration. mars1/rough mutants exhibit ectopic subepidermal cell proliferation, resulting in the formation of callus‐like structures on their fruit cuticles. (B) The MARS1/ROUGH gene encodes SlLDL1, whose active chromatin mark (methylated H3K4) is removed and whose expression is repressed at specific genomic regions. mars1/rough mutants cannot demethylate H3K4me1/2, resulting in the constitutive upregulation of several dozen genes, including the candidate gene cluster encoding proteins with partial homology to the N‐ and C‐terminal domains of mitotic B‐type cyclins. (C) Cyclin B is targeted for degradation by APC/C in a cell cycle‐dependent manner through a conserved N‐terminal D‐box sequence. Ectopic cyclin B peptides are expressed in mars1/rough mutants, and the retention of the C‐terminal domain results in their enhanced proliferative potential, indicated by the red banner.

According to the proposed working model depicted in Figure 11B, SlLDL1 predominantly functions as a histone demethylase, specifically by targeting H3K4me1/me2 in various regions of constitutive heterochromatin situated within euchromatin‐rich chromosome arms. In the SlLDL1‐deficient mars1/rough mutants, increased gene expression is observed in these regions, leading to the accumulation of novel transcripts, some of which have coding potential. One region with significant deregulation in mars1/rough mutants contains two contiguous genes, namely Solyc04g081660 and Solyc04g081650, encoding two peptides that are homologous to the N‐ and C‐terminal domains of the protein encoded by Solyc04g082430, respectively. Solyc04g082430 encodes G2/M‐specific SlCycB2;4, which is expressed mainly during gametogenesis. In the reprogramming system leading to callus formation from protoplasts obtained from the leaf mesophyll in Arabidopsis, the biosynthesis of endogenous auxin induces the transcriptional reactivation of cell cycle genes from the G2/M phase, including CYCLIN‐DEPENDENT KINASES (CDKs) and cyclin B genes, among others (Sakamoto et al., 2022). In mars1/rough mutants, the upregulation of genes involved in DNA synthesis at the base of the hypocotyl occurs as early as 8 h after excision. These findings are consistent with the enhanced response to auxins, leading to accelerated formation of AR primordia in mars1/rough mutants. Additionally, the genes encoding the CDKB1 and CDKB2 proteins in tomato are predominantly expressed in the pericarp during the early developmental stages, when cell division takes place (Czerednik et al., 2012). CDKA1 is expressed in the pericarp throughout development but is strongly upregulated in the outer pericarp cell layers at the end of the growth period, when CDKB gene expression has ceased. The overexpression of CDKB genes at later stages of development and the downregulation of CDKA1 result in a very similar fruit phenotype, with a reduction in endoreduplication and cuticle thickness (Czerednik et al., 2012). Cyclin B and CDK interactions positively regulate CDK kinase activity, thereby enhancing mitotic progression in these tissues. Given the structural similarity between the two peptides that are constitutively expressed in mars1/rough mutants and SlCycB2;4, it is hypothesized that the N‐ and C‐terminal cyclin B peptides expressed in mars1/rough mutants spontaneously associate because of the affinity between their structural domains, resulting in the formation of an active cyclin B protein that could interact with CDKs (Figure 11C). However, the DEAD‐box, which is necessary for cyclin B regulation via APC/C‐mediated ubiquitination, is present only in the N‐terminal peptide encoded by the upregulated Solyc04g081660 gene in the mars1/rough mutants. Consequently, the prolonged binding of the C‐terminal peptide encoded by Solyc04g081650 to CDKs in the mars1/rough mutants could maintain their kinase activity for a longer period and induce the increased proliferation observed in the fruit exocarp, as well as in the cambial hypocotyl during wound‐induced AR formation (Figure 11C). Although the possibility cannot be ruled out that an analogous mechanism may still function in other mars1/rough tissues where SlLSD1 is normally expressed, established mitotic regulators might counteract the surplus of cyclin B/CDK activity in these other tissues. These regulators include INHIBITOR OF CDK/KIP‐RELATED PROTEIN (ICK/KRP) (Nafati et al., 2010) and WEE1 protein kinase (Gonzalez et al., 2007). Alternatively, the expression of other tomato LDL orthologs could compensate for the loss of SlLDL1 function in these organs. Consequently, increasing the levels of the C‐terminal cyclin B peptide in mars1/rough mutant hypocotyls following wounding increases the cell cycle capacity within the AR formative region. This activity acts in a synergistic manner with the enhanced auxin effect in response to wounding observed in the mutants.

The precise upregulation of mitotic cyclin B peptides in mars1/rough mutants is responsible for the observed AR formation and fruit epidermis phenotypes, as demonstrated by our VIGS experiments. Furthermore, transcriptomic analyses of tomato hypocotyl explants during wound‐induced AR formation revealed that mars1/rough mutations increased the basal expression of several dozen silenced genomic regions. Our findings suggest that SlLDL1 functions through local silencing of specific regions within euchromatin in the chromosome arms, as well as through global silencing of regions of pericentromeric heterochromatin, as evidenced by our RNA‐seq data.

The loss‐of‐function mutations of two partially redundant members of the LDL family in Arabidopsis, LDL1 and LDL2, result in a pleiotropic phenotype, including increased seed dormancy, late‐flowering, and altered expression of defense‐related genes (Jiang et al., 2007; Zhao et al., 2015). A number of studies have reported the genome‐wide occupancy profile of LDL1 and LDL2 demethylases, which are known to interact with the histone deacetylase HDA6 to synergistically downregulate gene expression by removing H3Ac and H3K4me1/2 from their target genes (Hung et al., 2018, 2020). Notably, the findings indicated the potential involvement of the HDA6‐LDL1/2 complex in the regulation of a specific subset of lncRNAs, which exhibited low or undetectable expression in wild‐type Arabidopsis seedlings but was upregulated in the mutants. Moreover, the expression levels of their neighboring genes were found to be positively correlated with the expression of these lncRNAs (Hung et al., 2020). It can thus be postulated that the Arabidopsis HDA6‐LDL1/2 complex may play a pivotal role in plant development by repressing some lncRNAs that may be induced under specific environmental conditions. We observed that the genomic regions that were upregulated in the mars1/rough mutants were enriched for intergenic lncRNAs. Additional studies, including RNA immunoprecipitation sequencing (RIP‐seq) and cross‐linking immunoprecipitation followed by high‐throughput sequencing (CLIP‐seq), are needed to determine whether SlLDL1 participates in a similar complex in tomato.

The most deregulated transcripts in the mars1/rough mutants exhibited a broad distribution along the entire chromosome, extending beyond the euchromatic arms to encompass the centromeric and pericentromeric chromatin. These regions are distinguished by the presence of intergenic lncRNAs and are abundant in transposable elements (TEs) and pseudogenes (Naish and Henderson, 2024). In Arabidopsis, several H3K9 methyltransferase genes, including SU(VAR)3–9‐RELATED 5 (SUVR5) and KRYPTONITE (KYP)/SUVH4, SUVH5, and SUVH6, are known to be involved in H3K9me2 deposition in TEs in a DNA methylation‐independent and DNA methylation‐dependent manner, respectively (Caro et al., 2012; Du et al., 2014). The loss of H3K9me2 in TEs, as observed in suvh456 triple mutants, induces a gain of H3K4me1, which in turn leads to their transcriptional activation (Inagaki et al., 2017). Among the four LDL genes in Arabidopsis, LDL2 has been proposed to function downstream of H3K9me2, most likely by decreasing H3K4me1 levels (Inagaki et al., 2017). Furthermore, SUVR5 physically and genetically interacts with LDL1, another H3K4 demethylase that is partially redundant with LDL2 (Caro et al., 2012). The loss of KYP histone methyltransferase in tomato revealed that the levels of H3K9me2 in kyp mutants diminished in discrete pericentromeric regions, which in turn resulted in a sharp deposition of H3K9 acetylation and the emergence of novel topologically associating domain (TAD)‐like structures that could contribute to 3D chromatin conformation and changes in KYP‐controlled gene expression (An et al., 2024). Therefore, we cannot rule out the possibility that SlLDL1 is involved in the epigenetic regulation of intercalated heterochromatin through a pathway similar to that proposed for LDL2 in Arabidopsis (Inagaki et al., 2017). The use of tagged SlLDL1 proteins in conjunction with ChIP‐seq experiments utilizing H3K4/K9 marks will facilitate the identification of the specific genomic regions that are directly regulated by SlLDL1 and its interacting protein partners.

MATERIALS AND METHODS

Plant materials and growth conditions

The seeds of Solanum lycopersicum L. cv. Micro‐Tom (MT) and the P19E6 line, obtained from a highly mutagenized EMS mutant collection (Just et al., 2013), were utilized in all the experiments. A total of 30 M₂/M₃ mutant families were selected through the screening of our phenotypic mutant database (Just et al., 2013) for the presence of fruit cracking mutants. The reporter line DR5:mScarletI‐NLS;TCSn:mNeonGreen‐NLS has been previously described (Omary et al., 2022). The seeds were surface‐sterilized and germinated in vitro in accordance with the methodology described by Alaguero‐Cordovilla et al. (2020). Afterward, 1‐month‐old seedlings were transplanted and cultivated in the greenhouse under standard conditions, as described by Rothan et al. (2016).

Morphological analysis of mars1/rough phenotypes

The early seedling phenotype of mars1/rough mutants was studied as described by Alaguero‐Cordovilla et al. (2020). The formation of ARs was induced by removing the entire root system 2–3 mm above the hypocotyl−root junction of in vitro grown tomato seedlings at the 100–101 growth stage, after which the shoot explants were subsequently transferred to glass jars containing standard growth medium (Alaguero‐Cordovilla et al., 2021). Each jar contained six explants of the same genotype and/or treatment. All experiments were conducted in triplicate. Additionally, hypocotyl explants were obtained by sectioning the shoot apex just below the cotyledons. The emergence of ARs from the basal region of the hypocotyl explants was monitored and recorded at regular intervals. The emergence of ARs was estimated on the basis of the day before the first AR was observed. The regeneration of shoot tissue in hypocotyl explants was characterized according to the methodology described by Larriba et al. (2021a).

To assess the impact of exogenous hormones on the process of tissue regeneration, cotyledon explants were harvested from 7‐d‐old seedlings and divided into two halves. The explants were subsequently incubated in 90‐mm diameter Petri dishes containing 1/2 Murashige and Skoog (MS) basal salt media supplemented with 2% sucrose, and 0.8% agar with or without 0.15 μM NAA or 2.5 μM 6‐BAP. After 21 d of incubation, the number of ARs and the fresh weight of the calluses were quantified. The experiments were conducted in triplicate with a minimum of six explants used in each trial.

Mapping‐by‐sequencing and recombinant analysis

Mapping‐by‐sequencing and recombination analysis were performed as previously described (Garcia et al., 2016; Petit et al., 2016). A mapping population (BC₁F₂) was created by crossing the P19E6 mutant with an MT parental line and by selfing a single BC₁F₁ hybrid. The mutant bulk was then constituted by pooling 44 plants with a rough fruit phenotype, and the extracted DNA was used to prepare a library that was sequenced using a HiSeq. 2500 sequencer (Illumina, 100‐bp paired‐end run mode) on the INRAGeT‐PlaGe‐GENOTOUL platform. Sequence analyses were performed as previously described by Garcia et al. (2016), using version SL2.50 of the reference tomato genome for read mapping (https://solgenomics.net/ftp//genomes/Solanum_lycopersicum/Heinz1706/assembly/build_2.50/). EMS variants with read depths between 10 and 100 were considered for allelic frequency analysis. Recombinant BC₁F₂ individuals were detected using the EMS‐induced SNPs flanking the putative mutation as markers in a Kompetitive allele‐specific PCR (KASP) assay (Smith and Maughan, 2015). Subsequent genotyping of the causal mutation was performed through Sanger sequencing of the PCR products. All primer sequences are listed in Table S1.

Generation of CRISPR/Cas9 lines

CRISPR/Cas9 mutagenesis of SlLDS1 and tomato cotyledon transformation were performed as described by Bollier et al. (2018), except that the level 1 construct pICH47742:2x35S‐5′UTR‐hCas9(STOP)‐NOST was replaced by pICH47742:2x35S‐5′UTR‐pcoCas9(STOP)‐NOST (Addgene plasmid #112079), and the level 1 construct was assembled into the level 2 pICSL4723 (Addgene plasmid #86172). The sgRNA target sequence was designed using CRISPR‐P 2.0 software (Lei et al., 2014). The T₀ plants resulting from the regeneration of the cotyledons were genotyped, and their fruits were phenotyped. T₁ seeds selected from T₀ plants with rough fruits were subsequently sown for further characterization. CRISPR/Cas9‐positive lines were further genotyped for indel mutations using primers flanking the target sequence. The resulting mutations were detected by PCR and sequencing in the T₀ and T₁ lines. A list of the sgRNAs and primers used is given in Table S1.

Vector construction and tomato genetic transformation

The coding sequence (CDS) of Solyc10g047350 was domesticated using the GoldenBraid (GB) domestication tool (https://goldenbraidpro.com/). The domesticated sequence, flanked by B3‐B5 overhangs, was synthesized, including a 3× HA tag (Integrated DNA Technologies, Iowa, USA), and subsequently cloned into the pUPD2 vector. This process resulted in the creation of a Level 0 (L0) module pUPD2‐SlLDL1. The L1 expression cassette was assembled through a multipartite BsaI‐mediated reaction, employing the L0 parts proPMtb (Addgene plasmid #GB0080), pUPD2‐SlLDL1 (this work), pT35S (Addgene plasmid #GB0036), and pDGB1_alpha2R as the backbone. The final transformation construct was generated through a BsmBI‐mediated restriction‐ligation L2 assembly that incorporated proPMtb:SlLDL1:T35S, GB0235, and pDGB3_omega1. The L1 and L2 assemblies were constructed in accordance with the established GB protocols (Sarrion‐Perdigones et al., 2014; Vazquez‐Vilar et al., 2017). The L2 vector was then transformed into chemically competent Escherichia coli DH5α, and the vector sequence was further verified by next‐generation sequencing (GENEWIZ New Jersey, USA).

The L2 plasmid was used to transform the A. tumefaciens GV3101 strain through electroporation. After antibiotic selection, a single colony was inoculated into the LB liquid medium and incubated until an OD₆₀₀ of 0.8 was reached. The culture was harvested, and the pellet was washed twice with 10 mM MgCl₂. The final OD₆₀₀ was adjusted to 0.5. Then, the seedlings were co‐cultivated for 24 h with A. tumefaciens culture in buffer supplemented with 0.005% Silwet L‐77 and 100 μM acetosyringone (Li et al., 2009). After 24 h, the seedlings were washed with a 1/4 MS liquid solution containing timentin and cefotaxime. Then, the co‐inoculated seedlings were washed with sterile water and placed on an agar plate containing antibiotics for 1−2 weeks. The proPMtb:SlLDL1:T35S transformed seedlings were subsequently transferred to soil and genotyped by PCR. Because proPMtb:SlLDL1:T35S complemented the rough mutant phenotype of mars1‐3, smooth fruits were chosen from among the PCR‐positive plants for seed collection. The seeds were sown, and F₂ seedlings were obtained and used for subsequent experiments.

Histological analyses

Five‐millimeter basal sections from shoot explants were fixed in a paraformaldehyde/Triton solution, dehydrated, and embedded in Technovit 7,100 resin as previously described (Alaguero‐Cordovilla et al., 2021). Sections were stained with 0.05% weight/volume (w/v) toluidine blue (Sigma‐Aldrich) and observed using a bright‐field Motic BA210 microscope (Motic Spain, Spain). To visualize the auxin and CK response, DR5:mScarletI‐NLS;TCSn:mNeonGreen‐NLS reporters were introduced by crossing to MT and mars1‐1 and selected in F₃ by Sanger sequencing. Hypocotyls from seedlings at the 100–101 growth stage were cut into 5 mm sections and transferred to 1/2MS basal salt media supplemented with 2% sucrose, 0.8% agar, and 0.30 μM NAA. The expression of the reporter proteins mScarletI‐NLS and mNeonGreen‐NLS was observed at 0, 40, and 75 hae under a ZEISS Axio Observer microscope equipped with an Apotome 2 module (Carl‐Zeiss‐Stiftung, Oberkochen, Germany). Excitation of the mScarletI‐NLS was achieved at 450−490 nm wavelengths, while the fluorescence emission was measured between 500 and 550 nm. Excitation of mNeonGreen‐NLS was achieved at 538−562 nm wavelengths, and the fluorescence emission was measured between 570 and 640 nm. Images of the vascular bundles were captured at varying Z‐positions of the sample and processed by using ImageJ (Carotenuto et al., 2019).

Histological analyses of pericarp tissue in the fruits were performed as previously described (Musseau et al., 2017, 2020). Transverse fresh sections of the fruit exocarp were obtained from three independent 20−DPA−stage fruits from MT and mars1/rough plants. Transverse sections were stained with calcofluor white stain (Sigma‐Aldrich) for cell walls. The sections were submerged for 30 s in 0.05% calcofluor white stain and washed for 5 min with phosphate‐buffered saline two times. The stained sections were mounted on a slide in fluorescence mounting medium (CitiFluor AF1, England) and observed by confocal microscopy (Zeiss LSM880, Germany). Images were acquired with Zen 2011 software. All these experiments were performed at the Bordeaux Imaging Center (http://www.bic.u-bordeaux.fr/). Observations were made on MT, mars1/rough homozygous mutant, and wild‐type individuals from the mapping BC₁F₂ population detected by KASP assay and three CRISPR/Cas9 T₂ homozygous lines. For each genotype, a minimum of three pericarp sections from three fruits from different plants grown side‐by‐side were observed.

RNA isolation, next‐generation sequencing, and RNA‐seq analysis

In each sample, 3–4 mm of the basal region of the hypocotyl was collected at the indicated times: 0, 8, 20, or 48 hae (Figure 6A). Three biological replicates, each consisting of 8–10 basal fragments, were harvested and immediately frozen in liquid nitrogen. RNA was extracted as previously described (Larriba et al., 2021a). Total RNA was extracted from ~100 mg of powdered tissue using the Spectrum Plant Total RNA Kit (Merck, Burlington, MA, USA) and treated with DNAse I (Thermo Fisher Scientific, Waltham, MA, USA); the RNA was then stored at −80°C. Sequencing libraries and next‐generation sequencing were carried out using the BGISEQ‐500 pipeline (BGI‐Tech, Shenzhen, China). In the first experiment, three libraries were sequenced for the MT and mars1‐3 mutant, and two libraries were sequenced for the P19E6 mutant using the stranded DNBSeq protocol (BGI) in Pair End (PE) mode with 100 cycles of sequencing; in the second experiment, three libraries were sequenced for samples at 8 hae, and two libraries were sequenced for samples at 20 and 48 hae using the non‐stranded DNBSeq protocol (BGI) in PE with 100 cycles of sequencing (BioProject: PRJNA1289037).

The bioinformatics workflow utilized for RNA‐seq analysis is shown in Figure 6B. Briefly, the sequencing libraries from both experiments were aligned against the tomato SL4.0 genome assembly using the STAR aligner. The assignment of reads to gene and lncRNA models (ITAG4.0 annotation) was conducted with the featureCounts tool from the Subread package. The number of reads was determined using the featureCounts tool with the –s 2 parameters in experiment 1 and the default setting in experiment 2. The reads were normalized using counts per million (CPM), and differential expression analysis was performed using DESeq. 2, from iDEP 2.0 (Ge et al., 2018) to identify DEGs with a false discovery rate (FDR) = 0.01 and a fold change (FC) = | 1.5 | . The expressed transcripts (X_LOC) were obtained from the collapsed BAM files of each time/condition and genotype of experiment 1 using Stringtie and GFFCompare. The number of reads associated with the identified X_LOC in the collapsed BAMs of both experiments, differentiated by time and genotype, was quantified using the featureCounts tool from the GFF. To calculate the FC of each X_LOC, we utilized the count values per genotype and calculated the difference between the CPM values of the mutant and control genotypes. The chromosomal positions of the lncRNAs were annotated in relation to the genes annotated in ITAG4.0 using GFFCompare.

Protein extraction and Western blot analysis

Nucleus isolation was performed as described by Tian et al. (2020). Briefly, 1.5 g of frozen leaf tissue from 4‐week‐old MT and mars1‐3 plants was homogenized in liquid nitrogen. The tissue was incubated in 15 mL of nuclear isolation buffer (NIB) for 30 min and inverted every 5 min. The mixture was filtered sequentially through 70‐ and 40‐μm filters. The filtrate was subsequently centrifuged at 900 × g for 10 min at 4°C. The resulting pellet containing the nuclei was washed with 1 mL of NIB and then centrifuged again at 900 × g for 10 min. Histones were isolated from the nuclei using the acid extraction protocol described by Shechter et al. (2007). A total of 1 mL of 4 NH₂SO₄ was added to the pellet, which was subsequently incubated for 1 h with rotation at 4°C and then centrifuged at 21,000 × g for 10 min at 4°C to obtain the histone‐containing supernatant. Histone precipitation was carried out with trichloroacetic acid at a final concentration of 33% for 1 h at 4°C. The histones were then precipitated by centrifugation, washed twice with acetone, dried, and resuspended in nuclease‐free water. The proteins were separated on a 15% SDS–polyacrylamide gel and subsequently blotted onto PVDF blotting membranes (GE HealthCare Technologies, Chicago, IL, USA) as indicated elsewhere (Kenesi et al., 2023). The PVDF membrane was immersed in the blocking solution overnight. The blocking solution used was TBS‐T (150 mM NaCl, 50 mM Tris‐HCl, 0.2% Tween; pH 7.5) with 5% nonfat milk (Kenesi et al., 2023). The membranes were subsequently washed with TBS‐T at least three times. After the membranes were washed, they were incubated with histone‐specific primary antibodies (anti‐H3 [Abcam, ab1791], anti‐H3K4me1 [Abcam, ab8895], anti‐H3K4me2 [Abcam, ab7766], anti‐H3K4me3 [Agrisera, AS16 3190], anti‐H3K36me3 [Abcam, ab9050], or anti‐H3K9Ac [Sigma Aldrich 07‐352],) for 4 h at room temperature. The membranes were then washed with TBS‐T and incubated for 1 h with secondary antibodies (goat‐anti‐rabbit‐HRP: 1:10,000 [Agrisera, AS09602], or goat‐anti‐mouse‐HRP: 1:5,000 [GE HealthCare, NA 931VS]) at room temperature. Immunoreactive signals were subsequently developed with the use of WesternSure PREMIUM chemiluminescent substrate and detected by using a ChemiDoc XRS+ imaging system (Bio‐Rad, Hercules, CA, USA).

ChIP‐seq assays

ChIP‐seq assays were performed on the fourth leaves of 4‐week‐old tomato plants using the following antibodies: anti‐H3K4me3 (Millipore 07‐473), anti‐H3K4me2 (Active Motif. 39142), and anti‐H3K4me1 (Active Motif. 39300), according to the procedure described in (Ramirez‐Prado et al., 2021). ChIP‐seq libraries were prepared from 10 ng of DNA using the NEBNext Ultra II DNA Library Prep Kit for Illumina (NEB) according to the manufacturer's instructions. The quality of the DNA libraries was assessed with an Agilent 2100 Bioanalyzer. Single‐end 75 bp reads were produced by the Illumina NextSeq 500 platform. The cleaned reads were then aligned to the SL4.0 tomato genome using STAR software with the·‐‐alignIntronMax 1 parameter. MAC3 software (https://macs3-project.github.io/MACS/; Zhang et al., 2008) was used with default parameters and a q‐value threshold of 0.05, except for H3K4me1, for which broad peak calling was employed. BigWig files were normalized using S3norm (Xiang et al., 2020). Differential hyper‐ and hypomethylated peaks were identified using MAnorm2 (Tu et al., 2021) with a P‐value cut‐off of 0.05 and subsequently annotated to genes using the BEDTools intersect function (Quinlan, 2014). The plots and visualizations of the genomic peaks were produced using deepTools (https://deeptools.ie-freiburg.mpg.de/; Ramírez et al., 2016) and IGV (https://igv.org/), respectively. The ITAG4.0 version of the tomato genome was used for the structural annotation of the genes (https://solgenomics.net/ftp/tomato_genome/annotation/ITAG4.0_release/).

Reverse transcription, PCR, D‐Topo cloning, and sequencing

Total RNA was extracted from approximately 80−100 mg of roots, shoot explants, leaves, stems, and ripe green and red fruits using TRIzol reagent (Invitrogen, Carlsbad, CA, USA) according to the manufacturer's protocols. The obtained RNA was subjected to reverse transcription (RT) using the RevertAid kit (Thermo Fisher Scientific, Waltham, MA, USA) according to the manufacturer's recommendations. The resulting cDNA was used as a template for PCR amplification using several primer pairs listed as RT−PCR in Table S1. RT‐quantitative PCR (RT−qPCR) was performed as described by Justamante et al. (2025), using the primers listed in Table S1 for qPCR. PCR products amplified with the Vacuolar2F_DT and CYC‐3R primers (Table S1) were cloned using the pENTR/D‐TOPO kit (Thermo Fisher Scientific), and unique clones, as determined by the size of the PCR products obtained using the M13F and M13R primers, were Sanger sequenced. The genomic region encompassing the Solyc04g081650 to Solyc04g081670 genes was sequenced. To this end, the PCR products were amplified using the primers listed in Table S1 for DNA sequencing.

Vector construction and agroinjection for virus‐induced gene silencing (VIGS)

The SGN VIGS Tool (Fernandez‐Pozo et al., 2015; https://vigs.solgenomics.net/) was used to design specific probes against Solyc04g081650 and Solyc04g081660 within the candidate gene cluster, and against Solyc03g123760, which encodes the phytoene desaturase (PDS) that was used as a positive control. The primers listed as VIGS in Table S1 were used for PCR amplification. PCR products were cloned into the pTRV2‐GG vector (Addgene plasmid #105349) using Eco31I, and positive clones were verified by Sanger sequencing. The pTRV1 and pTRV2‐GG vectors were subsequently electroporated into Agrobacterium tumefaciens GV3101 strains, which were subsequently selected on the basis of antibiotic resistance and PCR. A. tumefaciens strains containing pTRV1 or pTRV2‐GG (either empty or with Solyc04g081650, Solyc04g081660, or Solyc03g123760 probes) were grown in YEB medium supplemented with antibiotics for 2 d at 28°C, after which the cells were harvested by centrifugation and resuspended in infiltration buffer (10 mM MgCl₂, 10 mM MES pH 5.6, 150 mM acetosyringone) to a final OD₆₀₀ of 0.5 at a 1:1 ratio. This infiltration solution was injected through the carpopodium of young tomato fruits (5−8 DPA) attached to the plant using a 1 mL syringe. For each pTRV2‐GG strain, three replicates of the agroinjection protocol were performed, each with a minimum of 3−5 tomatoes. The injected fruits were observed every week for 1 month. Furthermore, shoot explants from 2‐week‐old mars1‐3 and MT seedlings were obtained, and immersed in co‐inoculation media containing either pTRV1 or both pTRV1 and pTRV2‐GG (see above) for 1−2 h. Subsequently, the seedlings were placed on 1/2MS medium overnight without being washed off the bacteria. After 24 h, the seedlings were washed with distilled water and transferred to 1/2MS medium supplemented with the antibiotics timentin and cefotaxime. AR numbers were scored after a period of 4 d.

Statistical analysis

The descriptive statistics were calculated using StatGraphics Centurion 16.1.03 (StatPoint Technologies, Inc., Warrenton, VA, USA) and GraphPad Prism version 8.3.1 for Windows (GraphPad Software, La Jolla, California, USA) software. We performed multiple testing analyses using the ANOVA F−test or Fisher's least significant difference (Fisher's LSD) method (P < 0.01, unless otherwise indicated). Nonparametric tests were used when necessary.

CONFLICTS OF INTEREST

The authors declare no conflicts of interest.

AUTHOR CONTRIBUTIONS

E.L., C.B., A.A.‐C., R.R., Q.W., J.P., J.‐P.M., L.C., D.L., B.B., R.B.‐C., and D.E.‐B. designed and performed all the experiments. The first draft of the manuscript was written by C.R. and J.M.P.‐P. J.M.P.‐P. wrote the final manuscript. All the authors read and approved the final manuscript.

Supporting information

Additional Supporting Information may be found online in the supporting information tab for this article: http://onlinelibrary.wiley.com/doi/10.1111/jipb.70287/suppinfo

Figure S1. Root phenotype in the P19E6 mutant

Figure S2. Phylogenetic analysis of the LSD1‐like gene family in tomato

Figure S3. Expression profile of MARS1/ROUGH in different tissues and across developmental stages

Figure S4. Pleiotropic phenotype in mars1/rough mutants

Figure S5. Auxin and cytokinin response in the vascular cylinder of mars1/rough mutants during AR formation

Figure S6. Design of the RNA‐seq experiments

Figure S7. Read mapping and analysis of gene expression in experiment 1

Figure S8. Read mapping and analysis of gene expression in experiment 2

Figure S9. Expression of functionally annotated genes in selected pathways

Figure S10. MARS1/ROUGH encodes the SlLDL1, which is required for histone H3 lysine 4 demethylation

Figure S11. Quantitative ChIP‐seq signaling in mars1‐3 mutants

Figure S12. H3K4 methylation profile of dysregulated genes identified by RNA‐seq

Figure S13. IGV visualization of gene expression and H3K4 methylation in the genomic regions flanking upregulated DEGs in mars1/rough mutants

Figure S14. Expression profiles of Solyc04g081650, Solyc04g081660, and Solyc04g081670 in different tissues and across developmental stages

Table S1. List of primer sequences

Table S2. Raw gene expression data

Table S3. DEG analysis in the first experiment

Table S4. DEG analysis in the second experiment

Table S5. Expression of auxin, cytokinin, and cell cycle genes

Table S6. Raw transcript expression data

Table S7. ChIP‐seq results

Biographies

Larriba, E. , Bres, C. , Alaguero‐Cordovilla, A. , Riyazuddin, R. , Wang, Q. , Petit, J. , Mauxion, J.‐P. , Caballero, L. , Latrasse, D. , Bakan, B. , et al. (2026). Role of the tomato MARS1/ROUGH gene encoding a LYSINE‐SPECIFIC HISTONE DEMETHYLASE 1 in adventitious root and fruit skin formation. J. Integr. Plant Biol. 68: 3644–3669.

Data availability statement

Raw sequence reads data in this study have been deposited in the National Center for Biotechnology Information Sequence Read Archive database under accession PRJNA1289037. All other data generated or analyzed during this study are provided in this article and its supplementary data files or will be provided upon reasonable request.