Autophagy restricts tomato fruit ripening via a general role in ethylene repression

Disclaimer

The New Phytologist Foundation remains neutral with regard to jurisdictional claims in maps and in any institutional affiliations.

Introduction

Ripening involves complex biochemical and molecular reprogramming, resulting in color, texture, aroma, and flavor changes to attract humans and other animals (Giovannoni et al., ²⁰¹⁷). In climacteric fruits, this process is controlled by a myriad of phytohormones, predominantly ethylene (Li et al., ²⁰²¹; Huang et al., ²⁰²²). To allow these changes, fruits constantly reshape their cellular proteome by fine‐tuning protein degradation and synthesis (Szymanski et al., ²⁰¹⁷). While the ubiquitin–proteasome system was shown to be critical in ethylene signaling and ripening (Fenn & Giovannoni, 2021; Jia et al., ²⁰²³), knowledge of the impact of autophagy, another central degradation system, is rather limited.

Autophagy delivers cytosolic components to the vacuole for degradation and recycling. Double‐membrane vesicles, termed autophagosomes, are generated around the cellular cargo destined for degradation. The autophagosome then fuses with the tonoplast to release a single membrane structure, termed autophagic body. Inside of the vacuole, the autophagic body degrades along with its cargo, and its constituents are recycled to replenish cellular energy. Autophagy is executed via the function of > 30 autophagy‐related (ATG) proteins (Ding et al., ²⁰¹⁸; Marshall & Vierstra, ²⁰¹⁸). Notably, ATG proteins annotated with different numbers are not related and have distinct roles in autophagy. Among these, the ubiquitin‐like ATG8 proteins are important for autophagosome biogenesis, fusion with the vacuole, and selective recognition of the cargo to be degraded. They are found conjugated either to phosphatidylethanolamine (PE) lipids (lipidated) or in a nonactive (nonlipidated) free form (Kellner et al., ²⁰¹⁷). The ATG8 family consists of nine members in Arabidopsis thaliana (Arabidopsis; AtATG8a to AtATG8i). When conjugated to a fluorescent protein, ATG8 family proteins are considered optimal markers for autophagy activity (or flux) assessment, as the stability of the fluorescent protein moiety allows for the estimation of the amount of autophagic material that was delivered to the vacuole. While assessing the ATG8 lipidation status (ATG8‐PE : ATG8 ratio) provides a measure of autophagic membrane levels in the cytosol, it does not necessarily reflect autophagy activity (Qi et al., ²⁰²³). Due to some level of redundancy (Del Chiaro et al., ²⁰²⁴), ATG8 family members are not used for functional analysis. Alternatively, the downregulation of other essential (and usually single‐copy) ATG genes, such as ATG2, ATG5, or ATG7, each independently, can serve for functional analysis (Marshall & Vierstra, 2018). Each of these genes has a critical role in autophagy, rendering the respective knockout mutant autophagy‐deficient. ATG2, together with ATG9 and ATG18 family proteins, functions in the lipid delivery system to the developing phagophore. Recently, the ATG2‐dependent recruitment of ATG18a onto the phagophore to promote its expansion and closure was revealed (Luo et al., ²⁰²³). ATG5 and ATG7 are responsible for different yet essential steps of the ubiquitin‐like conjugation of ATG8 proteins to autophagic membranes. ATG4, a protease allowing the lipidation or delipidation of ATG8 members, is represented by a single gene in Solanum lycopersicum (tomato) and two genes in Arabidopsis (AtATG4a and AtATG4b) and is also used for functional analysis (Seo et al., ²⁰¹⁶).

Initially, it was assumed that ethylene and autophagy are not linked based on the lack of leaf senescence recovery phenotype in Arabidopsis plants deficient in autophagy and Ethylene‐Insensitive 2 genes (Yoshimoto et al., ²⁰⁰⁹). However, several studies later suggested that such an interaction exists (Liao et al., ²⁰²²). First, the transcript abundance of several ethylene biosynthesis and ethylene signaling genes was higher in atg5 and atg9 mutants than in wild‐type (WT) (Col‐0) Arabidopsis plants (Masclaux‐Daubresse et al., ²⁰¹⁴). Moreover, tomato plants treated with the ethylene precursor, 1‐aminocyclopropane‐1‐carboxylate (ACC), exhibited increased autophagy activity and expression of SlATG8d and SlATG18h genes during drought. The authors suggested the direct binding of Ethylene Response Factor 5 to the promoters of these genes (Zhu et al., ²⁰¹⁸). 1‐Aminocyclopropane‐1‐carboxylate was also reported to induce autophagy in Arabidopsis (Rodriguez et al., ²⁰²⁰). Finally, pollination‐induced petal leaf senescence, accompanied by enhanced ethylene emission, was further accompanied by increased expression of petunia PhATG8 isoforms and generation of autophagosomes. Notably, the ethylene antagonist, 1‐methylcyclopropene (1‐MCP), delayed the induction of PhATG8 proteins following pollination and the appearance of the subcellular structures presumed to be autophagosomes (Shibuya et al., ²⁰¹³). Therefore, ethylene is repeatedly suggested to regulate autophagy. Nonetheless, it is yet unclear whether autophagy has any regulatory impact on ethylene and whether it is involved in climacteric fruit ripening.

Materials and Methods

Plant growth and transformation

Tomato (Solanum lycopersicum L.) Micro‐Tom, Moneymaker and Ailsa Craig cultivars, and Arabidopsis thaliana (L.) Heynh. Col‐0 ecotype were used in this study. Tomatoes were grown in pots at 23°C in a climate‐controlled glasshouse under a 12‐h, light : dark cycle. Arabidopsis was grown in growth chambers at 16 h : 8 h, 22°C, light : dark cycles. Transgenic Del/Ros1 Micro‐Tom (Orzaez et al., ²⁰⁰⁹) seeds were obtained from Prof. Asaph Aharoni's laboratory (Weizmann Institute). GFP‐SlATG8‐2.2 Moneymaker lines were generated by transforming the 35S::GFP‐StATG8‐2.2 construct (Potato StATG8‐2.2 is identical to the tomato counterpart), which was kindly provided by Dr Yasin Dagdas (Zess et al., ²⁰¹⁹). E8::SlATG4‐RNAi lines were generated in a Micro‐Tom background.

Tomato transformation with the E8::ATG4‐RNAi construct was performed using Agrobacterium strain GV3101 harboring the appropriate plasmids as described in Sun et al. (2015) with minor modifications. Tomato seeds were surface‐sterilized and germinated on a ½‐strength Murashige & Skoog (MS) medium (+30 g l⁻¹ sucrose). After 10 d, fully expanded cotyledons were dissected, with only their central sections used. Explants were immersed in Agrobacterium suspensions (OD₆₀₀ = 0.6) for 20 min. Explants were then gently dried on Whatman paper and transferred to a cocultivation medium (MS medium supplemented with 3% (w/v) sucrose, 0.8% agar, and 2 mg l⁻¹ 6‐Benzylaminopurine) for 2 d in the dark. Then, explants were transitioned to a shoot induction medium (MS medium supplemented with 1 mg l⁻¹ zeatin, along with 0.8% (w/v) agar, 3% (w/v) sucrose, 50 mg l⁻¹ kanamycin, and 500 mg l⁻¹ cefotaxime). Incubation took place at 26°C with a 16‐h photoperiod. After 30 d, plantlets were separated from the original explants and transferred to a fresh induction medium. Shoots reaching lengths of 2–3 cm were excised and transplanted to a root induction medium (MS medium containing 1 mg l⁻¹ indole‐3‐acetic acid (IAA), 0.3% (w/v) agar, 3% (w/v) sucrose, and 50 mg l⁻¹ kanamycin). All substances used were from Duchefa Biochemie (Haarlem, the Netherlands). Eventually, 20 lines were obtained, of which five lines (L2, L9, L15, L18, and L20) were examined postharvest, all showing a clear delayed ripening phenotype at their T2 generation. Progeny of some lines, such as L9 (T3 and T4 generations), showed a reduction in phenotype severity, potentially due to trans‐generational silencing of the transgene.

GFP‐SlATG8‐2.2 Moneymaker line was generated similarly with a few differences. The seeds were sown on a germination medium (4.3 g l⁻¹ MS including vitamins, 30 g l⁻¹ sucrose, 100 mg l⁻¹ myo‐inositol, 0.8% phytoagar, pH 5.8). Two microliters of the Agrobacterium suspension (10 mM MgSO₄, 200 μM acetosyringone; OD₆₀₀ = 1.0) was applied per cotyledon, and they were cocultivated on the germination medium supplemented with 1 mg l⁻¹ 6‐Benzylaminopurine and 1 mg l⁻¹ NAA for 2 d in the dark at 22°C. In 2 d, the cotyledons were placed abaxial surface down on the germination medium supplemented with 35 mg l⁻¹ kanamycin, 1 mg l⁻¹ trans‐Zeatin, and 250 mg l⁻¹ ticarcillin disodium/clavulanate potassium. The cotyledons were placed in 14 h : 10 h, 23°C, light : dark, 50% humidity, and were transferred to a fresh selection medium every 7 d. In the second and fourth weeks, kanamycin concentration was increased to 50 and 100 mg l⁻¹, respectively. Regenerating shoots were cut at the base and transferred to the germination medium supplemented with 20 mg l⁻¹ kanamycin, 0.1 mg l⁻¹ IAA, and vancomycin (500 mg l⁻¹).

Plasmid construction

We employed the ClonExpress II One Step Cloning (Vazyme Biotech, Jiangsu, China) and the Gateway Cloning system (Thermo Fisher Scientific, Waltham, MA, USA) for virus‐induced gene silencing (VIGS)‐related cloning. The ‘VIGS tool’ at the Sol Genomics Network website (vigs.solgenomics.net) was used to select appropriate 300 base‐pair sequences. The primers used are listed in Supporting Information Table S1. Approximately 300‐bp fragments of SlATG2 (Solyc01g108160) and SlATG7 (Solyc11g068930) were PCR‐amplified and cloned into pENTER‐Gus (Thermo Fisher Scientific; modified to have spectinomycin instead of kanamycin resistance). Then, the Gateway LR reaction was performed to allow the transfer of the gene fragment into the pTRV2‐Del/Ros1 vector. For ripening‐specific SlATG4 (Solyc01g006230) silencing, the E8::SlATG4‐RNAi vector was generated by switching the promoter in the 35S::SlATG4‐RNAi expression cassette (Alseekh et al., ²⁰²²). The original plasmid was amplified (Phusion polymerase) as a linear fragment without the 35S part. Then, ClonExpress II was employed to introduce a PCR‐amplified E8 promoter with compatible flanking regions of the linearized plasmid. All cloned plasmids were transformed into the Mix & Go DH5α competent cells (Zymo Research, Irvine, CA, USA) and selected on Luria Broth (LB) agar plates containing the relevant antibiotics. Colony PCR confirmed positive clones, and plasmids were then purified using the Promega plasmid purification kit.

Immunoblotting

Liquid nitrogen‐frozen tomato fruit pericarp samples were manually ground to a fine powder using a mortar and pestle. One gram of samples was mixed with 500 μl of a 1× Laemmli buffer. The mixture was then heated to 95°C for 5 min and centrifuged at 16 000  g (4°C) for 10 min. The resulting supernatant was collected for immunoblotting. To allow the separation of ATG8 and ATG8‐PE fragments, equal amounts of protein were loaded onto custom‐made 15% SDS‐PAGE gels supplemented with 6 M urea in the resolving gel. Gels were blotted onto polyvinylidene difluoride membranes, which were incubated overnight at 4°C with 1 : 1000 anti‐ATG8 antibodies (Abcam, Cambridge, UK) and subsequently with goat‐anti‐rabbit‐HRP (1 : 10 000; Jackson ImmunoResearch, West Grove, PA, USA) before visualization in the Fusion Plus Imaging System (Vilber, Marne‐la‐Vallée, France). Anti‐H3 antibodies (EMD Millipore Corp., Burlington, MA, USA) were used for loading visualization (1 : 10 000). For the GFP‐release assay, total protein extracts were loaded onto commercial precast gels (SurePage; GenScript, Piscataway, NJ, USA) and immunoblotted with anti‐GFP antibodies (Ab290; 1 : 2000; Abcam). Arabidopsis samples were processed as previously described (Michaeli et al., ²⁰¹⁹).

RNA isolation and qRT‐PCR

Total RNA was extracted from 100 mg of young leaves (three biological replicates) or pink‐stage tomato fruits (three biological replicates, each comprised of three pooled fruits) using the Plant Spectrum RNA isolation kit (Sigma). RNA concentration and purity were determined, and cDNA was synthesized from 1 μg of total RNA using the verso cDNA synthesis kit (Thermo Fisher Scientific). For quantitative polymerase chain reaction (qPCR), the cDNA of each sample was diluted 1 : 10 with nuclease‐free water before mixing with the Luna^® Universal qRT‐PCR Master Mix (New England Biolabs, Ipswich, MA, USA). Then, samples were processed using the StepOnePlus Real‐Time PCR System (Thermo Fisher Scientific). The comparative C _t (ΔΔC _t) method was employed for data analysis, normalizing target gene expression to reference genes (TIP41, GAPDH, and Actin) and comparing the relative expression levels between samples.

Virus‐induced gene silencing

pTRV1 and pTRV2‐Del/Ros1 harboring the SlATG7‐ or SlATG2‐derived fragments were introduced into Agrobacterium strain GV3101 using electroporation (Bio‐Rad). A 5‐ml culture was grown overnight at 28°C in a medium containing kanamycin. The next day, the culture was transferred to a 50‐ml LB medium containing antibiotics, 10 mm 2‐(N‐morpholino)ethanesulfonic acid (MES), and 20 μM acetosyringone and grown overnight in a 28°C shaker. Agrobacterium cells were harvested, resuspended in an infiltration medium (10 mm MgCl₂, 10 mm MES, and 200 μM acetosyringone), left at room temperature for 3 h, and adjusted to an optical density (OD₆₀₀) of 0.1. Agrobacterium cells were then infiltrated, using a 1‐ml syringe, through the peduncle of mature‐green (MG)‐stage fruits while attached to the plant.

Fluorescence imaging and quantification of transgenic fruits using the in vivo imaging system

GFP‐SlATG8‐2.2 tomato fruits were harvested at different ripening stages (MG, turning, and red‐ripe (RR)) and kept in the dark for 12 h before evaluation. GFP fluorescence (465 nm excitation and 520 nm detection) was acquired and analyzed using the in vivo imaging system (IVIS) Lumina II equipped with an XFOV‐24 lens and the Living Image 4.3.1 software (PerkinElmer, Waltham, MA, USA). Fluorescence is presented as radiance intensity (photons s⁻¹ cm⁻² sr⁻¹).

Confocal microscopy, ConA treatment, and quantification

Confocal imaging was performed using the Leica SP8 (Leica Microsystems, Wetzlar, Germany) confocal laser‐scanning microscopy (CLSM) system at the Volcani Institute microscopy core facility. Thin sections of tomato fruit pericarp (using razor blades) or intact Arabidopsis seedlings were positioned between a microscope slide and a coverslip containing a ½MS liquid medium. Excitation/detection‐range parameters for GFP were 488 nm/500–550 nm, and emissions were collected using the system's hybrid (Hyd) detectors. ×20 dry (NA 0.75) or ×63 water immersion (NA 1.3) objectives were used. Scanning was routinely performed in ‘line’ mode. The images were processed and analyzed using Fiji (ImageJ).

For concanamycin A (ConA) treatment of fruit cells, thin fruit sections were incubated in liquid ½MS supplemented with 2 or 10 μM ConA, or the corresponding amount of dimethyl sulfoxide (DMSO) (mock), for 7 h before imaging or 15 h for immunoblots. For Arabidopsis analysis, GFP‐ATG8E seeds were germinated in liquid ½MS (Duchefa Biochemie) medium containing 1% sucrose. Four‐day‐old seedlings were transferred into the same medium containing either 0.4% DMSO (Sigma‐Aldrich; as mock) or 20 μM ACC or 1 μM ConA (Santa Cruz Biotechnology, Dallas, TX, USA) and 0.4% DMSO or 1 μM ConA and 20 μM ACC. Seedlings were incubated for 2 h in the dark and transferred to slides in the corresponding medium. Labeled autophagosomes were counted manually from single confocal‐slice images. Forty‐five seedlings and 677 cells were analyzed in three independent experiments.

Triple‐response assay

The triple‐response assay was made as described previously (Merchante & Stepanova, 2017) with minor modifications. Sterilized seeds were placed onto plates containing a ½MS medium (Duchefa Biochemie) and 1% sucrose, with or without 20 μM ACC. The plates were then moved to 4°C in darkness for 48 h before exposure to light for 2 h. Then, the plates were relocated to the growth room (22°C) and kept in darkness for 72 h.

Phylogenetic tree construction

The amino‐acid sequences of tomato and Arabidopsis ATG8 proteins were retrieved from Sol Genomics Network (solgenomics.net) and TAIR (www.arabidopsis.org) databases. Sequences were aligned using ClustalW in the EMBL‐EBI web server (https://www.ebi.ac.uk/Tools/msa/clustalw2/) using the default settings. The resulting file was uploaded to IQ‐Tree v.1.6 (http://www.iqtree.org/) for phylogenetic analysis, which included ModelFinder, tree reconstruction, and ultrafast bootstrap with 1000 replicates. ModelFinder was used to determine the best‐fit model, with the Bayesian information criterion guiding the model selection, ultimately choosing LG + G4. This model, incorporating a gamma distribution with four rate categories, was used for managing rate heterogeneity among sites. Branch support values were calculated using the SH‐aLRT test and ultrafast bootstrap approximation with 1000 replicates. Finally, the tree was visualized in the Newick format, using the Interactive Tree of Life (iTOL) web‐based tool (https://itol.embl.de/). The tree image was exported and refined using Inkscape to enhance clarity and presentation.

Fruit color (hue) and firmness measurements

Color values a* and b* and the hue parameter were measured using the Konica Minolta Chroma Meter CR‐400 Series v.1.11 (Konica Minolta, Tokyo, Japan). Fruits were considered at the MG stage when −0.59 < a*/b* < −0.47 (Batu, 2004). Then, fruits were harvested, cleaned, and measured for hue and firmness on different days postharvest, as indicated in related figures. Firmness was measured using the TA.XT Plus Texture Analyzer (Stable Microsystems, Surrey, UK) using a 3‐mm‐diameter probe with 5% penetration at a speed of 1 mm s⁻¹, recording the maximal endpoint force in Newtons (N).

Ethylene production measurements and 1‐MCP treatment

Harvested fruits were placed in 200‐ml sealed containers for 2 h. Ethylene was measured by injecting 10 ml of head‐space air into a gas chromatograph (GC; Varian 3300, Walnut Creek, CA, USA) with a flame ionization detector and an alumina column. For Arabidopsis, seedlings of the different genotypes were grown on a ½MS + 1% sucrose solid medium. 1‐Aminocyclopropane‐1‐carboxylate and mock treatments were applied on 10‐d‐old seedlings by adding 20 μM ACC (dissolved in a liquid ½MS medium) or mock (a ACC‐free ½MS medium) to the plates. Following 96 h of incubation, seedlings were removed, weighed, and quickly sealed in 20‐ml syringes containing a 1‐ml ½MS liquid medium, with or without ACC, for 2 h. Each syringe contained 30 seedlings. Head‐space was then collected with a fresh 10‐ml syringe (by penetrating the needle into the 20‐ml syringe) and injected into the GC apparatus. For 1‐MCP treatment, fruits from WT and the E8::ATG4‐RNAi lines were placed in a sealed flask containing 600 ppb of 1‐MCP (RIMI, Petah Tikva, Israel) for 20 h. Then, fruits were released and placed in a ‘shelf‐life’ chamber (22°C) for the indicated duration for hue measurements.

Dark‐induced leaf senescence and electrolyte leakage measurements

The third leaf from the apical meristem of 5‐wk‐old E8::ATG4‐RNAi lines (L9, L15, and L18) and WT plants was detached. Leaves were then washed and kept in Petri dishes with moist Whatman paper. Petri dishes were then wrapped with aluminum foil, incubated in a growth chamber, and photographed at 5, 7, 9, 12, and 14 d. Each experiment included four replicates. Similar leaves were used for electrolyte leakage measurements. From each leaf, 1.2–cm‐diameter leaf disks were incubated in 10 ml of de‐ionized water in the dark with solution measurements taken at 0, 3, 6, 7, 10, 11, 13, 17, 19, and 21 d using a conductivity meter. The experiment was repeated four times.

Statistics

Statistical analysis and data visualization were conducted using Python 3.9.6 with the following libraries:

1. NumPy 1.25.2 for numerical operations and data handling;

2. Pandas 2.0.3 for data importation and data frame handling;

3. Matplotlib 3.7.2 for basic plotting; and

4. Seaborn 0.12.2 for advanced data visualization.

Calculations were performed using the subpackage scipy.stats from the open‐source SciPy Python‐based library. The Shapiro–Wilk test (scipy.stats.shapiro – SciPy v.1.11.2) was used for normality determination, where P‐value > 0.05 suggests that the data are normally distributed. The Levene's test (scipy.stats.levene) was used for equal variance determination, where P‐value > 0.05 suggests equal variance. If two datasets showed equal variance and normal distribution, then an independent two‐sample t‐test (scipy.stats.ttest_ind) was used. Otherwise, a nonparametric Mann–Whitney U test (scipy.stats.mannwhitneyu) was used.

Results

Autophagy activity climaxes at mid‐ripening

To decipher the role of autophagy in tomato fruit ripening, we first followed its activity by monitoring SlATG8 dynamics (Qi et al., ²⁰²³). There are seven tomato protein orthologs of Arabidopsis AtATG8s clustered into four subgroups (Fig. 1a; Kellner et al., ²⁰¹⁷; Zess et al., ²⁰¹⁹). Data from the Sol Genomics Network – Tomato Expression Atlas (SGN‐TEA; tea.solgenomics.net; Shinozaki et al., ²⁰¹⁸) revealed that five of the seven members showed increasing expression along ripening progression. Two of these, SlATG8‐1.1 (Solyc08g078820) and SlATG8‐2.2 (Solyc07g064680), exhibited a relatively higher transcript abundance (Fig. 1b). To verify that SlATG8‐2.2, as a representative, encodes an autophagy‐associated protein in fruit cells, we generated transgenic tomato plants expressing GFP‐SlATG8‐2.2. CLSM of fruit endocarp cells at the breaker stage highlighted small spherical bodies (1–2 μm in diameter) reminiscent of autophagosomes (Fig. 1c, images C1–C3). These structures were motile (Video S1) and occasionally seemed to be ring‐like structures. Shown here is a presumably mature phagophore at an advanced stage before closure (Fig. 1c, image C4, edges defined by arrowheads), later appearing as a fully closed and complete autophagosome (Fig. 1c, image C5).

Fig. 1.

Autophagy activity climaxes at mid‐ripening. (a) A phylogenetic tree showing the four Solanum lycopersicum (tomato) ATG8 (SlATG8) family subgroups with the related Arabidopsis thaliana AtATG8 members. The scale bar represents branch length in substitutions per site. (b) Modified representation of SlATG8 family members' relative transcript expression in reads per million. Data obtained from SGN‐TEA. (c) Confocal laser‐scanning microscopy (CLSM) imaging of fruit endocarp cells expressing GFP‐SlATG8‐2.2 (green). The magenta signal represents Chl autofluorescence. Yellow arrowheads point toward autophagosomes. Image C2 also includes the bright‐field channel. Image C3 is an enlargement of the area defined with a dashed rectangle at C1. Images C4 and C5 show a developing autophagosome imaged 1 min apart. White arrowheads in image C4 denote the estimated location of phagophore edges. The white arrowhead in image C5 points toward the estimated autophagosome closure site. (d) Total pericarp proteins from six ripening stages were separated in resolving gel supplemented with 6 M urea and immunoblotted with @ATG8 antibodies (Abcam). Histone 3 (@H3; Millipore) serves as a loading control. A representative blot of three biological repeats is shown. (e) Fluorescence emission from GFP‐SlATG8‐2.2 fruits at three ripening stages: mature‐green (MG), orange, and red‐ripe (RR), imaged using in vivo imaging system. None of the 22 RR‐stage fruits examined exhibited a detectable fluorescence signal. (f) Total pericarp proteins of GFP‐SlATG8‐2.2 fruits from three ripening stages and wild‐type (WT) MG‐stage as control were immunoblotted with @GFP. Histone 3 (H3) serves as a loading control. Orange‐stage (Or) fruits were also treated with the indicated concentrations of concanamycin A (ConA) or mock (DMSO). Representative blots of three biological repeats are shown. (g) CLSM imaging of ConA‐treated endocarp cells of MG‐ and orange‐stage GFP‐SlATG8‐2.2 fruits. The magenta signal represents Chl autofluorescence. Yellow arrowheads point toward autophagosomes. White arrows denote the location of cytosolic aggregates, potentially culminating due to their inability to degrade. Representative images of 10 replicates (representing > 50 cells) per ripening stage are shown. No autophagic bodies were detected in any of the MG‐stage cells. Vac, vacuole lumen.

To gain insight into autophagy activity, we immunoblotted total pericarp protein extracts of tomato fruits at different ripening stages with an anti‐ATG8 antibody. We suspect this polyclonal antibody recognizes all expressed ATG8 isoforms. Our analysis showed a sharp increase in ATG8‐PE levels, from undetectable levels at the MG stage to prominent expression at mid‐ripening, and maintaining these high levels until the final RR stage (Fig. 1d). The increase in ATG8‐PE may reflect either an elevation in autophagy activity or a blockage in autophagy flux to the vacuole. To determine which of these options is factual, GFP‐SlATG8‐2.2 expressing fruits were evaluated to follow autophagy flux. Using an IVIS (Perkin Elmer), we detected sufficient GFP fluorescence in fruits of MG to orange stages. However, the signal was not detected in any of the RR fruits examined (n = 22; Fig. 1e). We then employed the GFP‐release assay, which allows assessment of autophagy flux from the cytoplasm to the vacuole according to the ratio of the free GFP degradation product relative to the full‐length GFP‐ATG8 fusion protein. Immunoblotting fruit protein extracts of three ripening stages, MG, orange, and RR, with anti‐GFP antibodies revealed a sharp increase in free GFP level (i.e. increasing flux) from the MG stage to the orange stage (Fig. 1f). Consistent with our IVIS analysis (Fig. 1e), the level of GFP detection at the RR stage was significantly lower than at earlier stages. Nonetheless, it was sufficient to observe a shift back toward balanced levels of GFP‐ATG8‐2.2 and the free GFP, indicating a reduction in autophagy flux (Fig. 1f; RR sample). We then examined the impact of ConA, a drug that inhibits vacuolar degradation (Qi et al., ²⁰²³). When applied at the orange stage, both mild (2 μM) and high (10 μM) ConA concentrations reduced the level of the free GFP degradation product (Fig. 1f, right blot), suggesting that most of it was produced via vacuolar degradation. ConA further allows visualization of fluorescent proteins within the vacuole lumen. CLSM imaging of orange‐stage vacuoles following ConA treatment showed punctate structures corresponding to autophagic bodies (Fig. 1g), exhibiting the typical random movement within the vacuole lumen (Video S2). Notably, no such autophagic bodies were detected in any of the MG‐stage vacuoles following analysis of > 50 cells in 10 different fruits (Fig. 1g). This suggests autophagy flux occurs at the mid‐ripening, whereas it is blocked at the MG stage. Altogether, increasing ATG8‐PE levels along ripening renders increased autophagy activity, which peaks at the mid (orange) stage.

Autophagy deficiency accelerates fruit ripening and results in increased ethylene production in both tomato fruits and Arabidopsis seedlings

Autophagy is involved in many processes, including energy homeostasis and nutrient remobilization (Avin‐Wittenberg et al., ²⁰¹⁵; Li et al., ²⁰¹⁵; Masclaux‐Daubresse et al., ²⁰¹⁷). Therefore, knockout or constitutive knockdown of any key ATG gene would result in pleiotropic effects, preventing the understanding of ripening‐specific roles. To overcome this, we applied two approaches. First, we used VIGS directly in the MG‐stage Del/Ros1 tomato fruits, harboring the Snapdragon transgenes Delila (DEL) and Rosea 1 (ROS1) driven by the E8 ripening‐induced promoter. Therefore, silencing Del/Ros1, which prevents the ripening‐associated appearance of purple fruits (Orzaez et al., ²⁰⁰⁹), is a visual indicator for silencing (Fig. 2a; D/R vs NT fruits). Silencing of the core autophagy genes, SlATG2 or SlATG7, each with DEL/ROS1 (D/R/ATG2‐VIGS or D/R/ATG7‐VIGS, respectively), resulted in SlATG8 accumulation, presumably due to its decreased turnover (Fig. S1a). This is consistent with the accumulation of AtATG8 in Arabidopsis atg2 and atg7 mutants (Munch et al., ²⁰¹⁴; Kang et al., ²⁰¹⁸; Luo et al., ²⁰²³). Phenotypically, silencing SlATG2 resulted in accelerated color transition (Fig. 2a). A quantitative examination of fruit populations' green‐to‐red transition, manifested by declining hue angle values, showed an accelerated color transition of ATG2‐silenced fruits (D/R/ATG2‐VIGS) relative to reference fruits (D/R‐VIGS; Fig. 2b). A similar outcome was revealed following silencing of SlATG7 (D/R/ATG7‐VIGS), albeit with differences observed at an earlier time point (Fig. S1b). Moreover, D/R/ATG2‐VIGS fruits exhibited accelerated fruit softening as indicated by firmness measurements (Fig. 2c).

Fig. 2.

Silencing SlATG2 or SlATG4 in mature fruits results in accelerated ripening and ethylene production. (a–c) D/R‐VIGS, silencing of Del/Ros1. D/R/ATG2‐VIGS, silencing of Del/Ros1 and ATG2. D/R/ATG7‐VIGS, silencing of Del/Ros1 and ATG7. (a) Representative Del/Ros1 tomato fruits at 10 d post injection of Agrobacterium tumefaciens (Agrobacterium) strain GV3101 cells harboring the indicated virus‐induced gene silencing (VIGS) constructs. NT, nontreated. (b) Quantification of the green‐to‐red color transition (hue) of VIGS‐treated fruits at the indicated days post Agrobacterium injection. Decreasing hue values indicate color transition. Values are the mean ± SD (n = 30; each value represents the average of three technical measurements for each fruit). (c) Firmness measurements of populations of VIGS‐treated fruits. Firmness is displayed as the force in Newton (N) required to achieve a similar probe penetration (see the Materials and Methods section) on different days post Agrobacterium injection. Values are the mean ± SD (n > 30). (d) Relative expression of SlATG4 in pink‐stage fruits from the wild‐type (WT) and two E8::ATG4‐RNAi lines was measured by quantitative reverse transcription polymerase chain reaction (RT‐PCR). Values are the mean ± SD after normalization to three reference genes (TIP4, Actin, and GAPDH). Significance was tested relative to WT using one‐tailed and paired Student's t‐test n = 9. *, 0.01 < P < 0.05; ***, P < 0.001. (e) Total pericarp proteins of pink‐stage fruits from the WT and E8::ATG4‐RNAi line L18 were immunoblotted with @NBR1 antibodies. For each genotype, three biological replicates are presented (noted as 1–3). Histone 3 (@H3; Millipore) serves as a loading control. Neighbor of BRCA 1 band intensities (quantified using ImageJ) were normalized to the H3 loading control bands and are presented as the ratio relative to the far‐left band (WT1). (f) Relative expression of SlNBR1a (Solyc03g112230) and SlNBR1b (Solyc06g071770) in the pericarp of pink‐stage fruits from WT and L18 was measured by quantitative RT‐PCR. Values are the mean ± SD after normalization relative to two reference genes (Actin and GAPDH) with similar results. Three biological (each representing three fruits) and technical replicates were performed. Significance was tested relative to the WT using one‐tailed and paired Student's t‐tests. *, P < 0.01. (g) WT, L9, and L18 tomato fruits across 25 d post harvest. The first fruit to initiate ripening out of the sampled population of each genotype is shown here. (h) Quantification of the green‐to‐red color transition (hue) of WT, L9, and L18 fruits at the indicated days post harvest. Values are the mean ± SD (n = 30). (i) Quantification of green‐to‐red color transition (hue) of WT, L9, and L18 fruits, treated or not with 1‐methylcyclopropene. Values at the indicated days post treatment are the mean ± SD (n = 10). (j) Representative WT and L18 fruits from the experiment described in (i). (k) Ethylene production in WT, L9, and L18 tomato fruits over 31 d post harvest. The numbers next to the key represent total ethylene production, on average, for each genotype during the entire measurement duration. Values are mean ± SE (n = 30; each measurement represents an average of three fruits). (l) Ethylene production from Del/ROS1 fruits following Del/Ros1 VIGS‐mediated silencing (D/R‐VIGS as control) or fruits additionally silenced in ATG2 (D/R/ATG2‐VIGS). Values at the indicated days post injection of Agrobacterium cells are the mean ± SD (n = 5; each measurement recorded from three pooled fruits). Significance was tested relative to controls (D/R‐VIGS in (b, c, l), or WT in (h, k)) for each time point. Either unpaired and two‐tailed Student's t‐tests or the Mann–Whitney U test was performed based on population distribution and variance (see the Materials and Methods section). No asterisk, not significant; *, 0.01 < P < 0.05; **, 0.001 < P < 0.01; ***, P < 0.001.

For the second approach, we targeted SlATG4. Similar to most ATG8 members, its expression also increases along ripening progression, suggesting they may be synchronized to allow proper lipidation or delipidation of ATG8 proteins (Fig. S2). For ripening‐specific silencing, we generated transgenic lines harboring a previously confirmed ATG4‐RNAi construct (Alseekh et al., ²⁰²²), albeit under the regulation of a ripening‐induced promoter (E8::SlATG4‐RNAi). We focused on line no. 18 (L18), which exhibited significant silencing, and L9, which was revealed as a relatively weak line (Fig. 2d). Since ATG4 may act in both lipidation and delipidation of ATG8 (Zou et al., ²⁰²⁵), we reasoned that autophagy activity assessment via ATG8‐PE levels would not fit here. Therefore, we turned to examine the levels of neighbor of BRCA 1 (NBR1), a selective‐autophagy cargo receptor known to accumulate under autophagy deficiency (Fig. S3a) and to reflect the autophagy flux status (Bassham, 2015; Yoshii & Mizushima, ²⁰¹⁷; Zhao et al., ²⁰²²). There are two predicted NBR1 genes in tomatoes: SlNBR1a and SlNBR1b (Zhou et al., ²⁰¹⁴). Constitutive ATG4 silencing (in 35S::ATG4‐RNAi plants (Alseekh et al., ²⁰²²)) resulted in the marked accumulation of SlNBR1a (Fig. S3b).

Consistently, SlNBR1a accumulated in the E8::ATG4‐RNAi L18 pink‐stage fruits relative to WT pink‐stage fruits (Fig. 2e), without a concomitant increase in SlNBR1a nor in SlNBR1b transcript levels (Fig. 2f), demonstrating a reduced autophagic breakdown in L18 fruits. Fruits from WT and the transgenic lines were harvested at the MG stage and monitored postharvest for color transition (Fig. 2g). On average, L18 fruits initiated earlier ripening and reached full ripeness 7 d before WT and L9 fruits (Fig. 2h). Notably, accelerated ripening was also observed in nondetached fruits of E8::SlATG4‐RNAi lines L15 and L18 (Fig. S4). Altogether, these results demonstrate that autophagy is involved in ripening restriction.

Previously, constitutive silencing of SlATG4 (35S::ATG4‐RNAi) was reported to result in early tomato leaf senescence (Alseekh et al., ²⁰²²), a canonical phenotype of autophagy deficiency, which has the potential to affect fruit development and ripening. To verify that our ripening‐context (E8::ATG4‐RNAi) lines phenotype is restricted to fruits, we visually monitored leaves from L9, L15, and L18, as well as their electrolyte leakage measurements. These tests showed no significant differences in leaf senescence of the silencing lines relative to WT leaves (Fig. S5a–c). To assess the tissue‐dependent activity of the E8 promoter, we quantified the expression of the cloned ATG4 fragment (Fig. S5d) in pink‐stage fruits and leaves of WT and L18 plants. This showed that its expression in L18 fruits is, on average, almost 10 times that of the expression in WT fruits (the latter representing the expression of the native WT ATG4 gene). On the contrary, there were no significant differences in ATG4 expression in the leaves of these genotypes (Fig. S5e). These data suggest that E8 promoter activity, driving ATG4 silencing, is indeed restricted to fruits.

To examine whether the impact of autophagy on ripening is linked to ethylene, we treated WT, L18, and L9 fruits with the ethylene antagonist, 1‐MCP, before hue measurements were taken to assess color transition. Fruits were harvested at a relatively advanced color stage (hue value of c. 105) to minimize the ripening initiation gap. 1‐Methylcyclopropene abolished the ripening advantage of L18, which exhibited similar delayed ripening dynamics as the WT fruits (Fig. 2i,j), suggesting that autophagy might regulate ripening via interaction with ethylene. Examination of ethylene emission showed earlier climacteric phase onset and higher ethylene production in L18 than in WT and L9 fruits. On average, L18 initiated ethylene production 7 d before WT and 3 d before L9 fruits. However, differences at 7–13 d post harvest were found to be statistically insignificant (using the Mann–Whitney U test) due to the high variance usually found within fruit populations around the ripening initiation time. Notably, L18 fruits reached ethylene production climax 5 d before WT and 3 d before L9 (Fig. 2k). Moreover, pink‐stage L18 fruits did not show significant alteration in the expression of tomato ACC‐Oxidase 1, ACC‐Oxidase 4, ACC‐Synthase 2, ACC‐Synthase 4, Polygalacturonase, or Expansin 1 relative to their levels in WT pink‐stage fruits (Fig. S6). These genes are involved in climacteric ethylene production and fruit softening (Li et al., ²⁰²¹), suggesting that autophagy's regulation of ripening is not at the level of transcription. We further examined ethylene emission from the Del/Ros1/ATG2‐VIGS fruits compared with the Del/Ros1‐VIGS reference. Experiments showed elevated ethylene production in the ATG2‐silenced fruits throughout the experiment, albeit significantly higher at 14 d post injection (Fig. 2l). A similar experiment testing ATG7‐VIGS revealed a similar outcome, with a significantly elevated ethylene emission at 7‐ and 8 d post injection (Fig. S1c). These results, together with the 1‐MCP treatment data, suggest autophagy affects ripening via its impact on ethylene production. That said, we currently cannot exclude the possible involvement of additional hormones known to interact with ethylene and to affect ripening, such as abscisic acid and auxin (Fenn & Giovannoni, 2021).

To gain further insight into the autophagy‐ripening and autophagy‐ethylene crosstalk, we examined its activity in fruits of the ripening mutants, never‐ripe (nr) and colorless nonripening (cnr). NR encodes an ethylene receptor (Tieman et al., ²⁰⁰⁰), while CNR encodes a SQUAMOSA promoter‐binding protein‐like transcription factor (Wang et al., ²⁰²⁰). In WT, SlNBR1a was detected at the MG and turning stages (40 and 47 d post anthesis (DPA), respectively) and significantly reduced at RR (54 DPA), potentially reflecting an increased flux of NBR1 toward the end of the ripening process. Notably, SlNBR1a levels were significantly higher in nr than in WT fruits at all three stages examined. On the contrary, SlNBR1a was undetectable at any of the cnr fruits examined (Fig. S7a). Quantitative PCR showed no significant difference in SlNBR1a transcript in nr relative to WT fruits, suggesting a decreased autophagy activity in nr fruits (Fig. S7b). However, we estimate that the c. 20% reduction in NBR1a expression in cnr fruits (Fig. S7b) does not fully explain the lack of SlNBR1 protein detection, suggesting strong autophagy flux in cnr fruits. Although we currently cannot provide a clear explanation, we do note that cnr, which is impaired in ethylene production, appears to show hyperautophagic activity, whereas nr, which is impaired downstream of ethylene production, shows the opposite. In other words, the negative correlation between ethylene production and autophagy activity is yet again apparent in these fruit‐ripening mutants.

To see whether autophagy impacts ethylene beyond ripening, we examined Arabidopsis atg2‐1 knockout seedlings under the triple‐response assay, a known test for detecting ethylene‐sensitive or ethylene‐insensitive mutants (Merchante & Stepanova, 2017). Overall, the atg2 mutants did not show altered root growth relative to Col‐0 seedlings in the dark. However, differences were significant following ACC application (Fig. 3a,b). A similar examination of atg5‐1 and atg7‐2 autophagy‐deficient mutants and ebf2‐3 as a known ethylene hypersensitive mutant (Potuschak et al., ²⁰⁰³) revealed a similar outcome, except for atg5 seedlings exhibiting a shorter root already under dark treatment (Fig. S8a,b). Notably, atg2, atg5, and atg7 seedlings exhibited between 1.75‐ and threefold increase in ethylene emission relative to Col‐0 following ACC treatment, while atg2 and atg7 produced significantly more ethylene even under mock treatment (Figs 3c, S8c). These results suggest that the root phenotype of atg mutants following ACC application may be due to increased ethylene production, highlighting a potentially general role for autophagy in ethylene repression.

Fig. 3.

Autophagy‐deficient atg2 Arabidopsis seedlings display increased ethylene emission and 1‐aminocyclopropane‐1‐carboxylate (ACC) sensitivity, ACC induces autophagy, and the current working model. (a) Representative Arabidopsis seedlings of the indicated genotypes were assessed for ACC‐induced triple‐response assay (dark‐grown, etiolated seedlings, with or without ACC treatment; see the Materials and Methods section). Bars, 0.25 cm. (b) Root length measurements of seedlings subjected to the triple‐response assay. Values represent means ± SD (n > 19). (c) Ethylene production from 10‐d‐old Arabidopsis seedlings. Values represent mean ± SD (n = 5, each measurement recorded from 30 pooled seedlings). (d) Total protein extracts of Arabidopsis seedlings, grown with or without ACC‐containing medium for the indicated time points (in hours), were immunoblotted with @ATG8. The upper lane shows standard separation, while the lower lane shows separation in a 6 M urea‐supplemented gel. Coomassie staining serves as loading control. (e) Representative confocal images of Arabidopsis root cells expressing GFP‐ATG8e following treatment with ACC, mock, concanamycin A (ConA) + mock, or ConA + ACC. Bars, 20 μm. (f) Quantification of GFP‐labeled puncta in image slices of plants treated as described in (e). Numbers were normalized to image area (1000 μm²). n > 35. (g) Total proteins of transgenic GFP‐ATG8e Arabidopsis seedling roots that were treated with mock, ACC (20 μM), or ACC and ConA (1 μM) for 72 h before immunoblotting with @GFP. The asterisk denotes a nonspecific protein band, as can be seen in Col‐0 samples (Supporting Information Fig. S9a). Histone 3 (H3) serves as a loading control. Representative blots of two biological repeats are shown. (h) Representation of ethylene and autophagy dynamics along wild‐type (WT) or ATG knockdown (KD) fruit ripening. In WT, autophagy and ethylene share similar dynamics, and autophagy participates in buffering ethylene levels. In ATG KD (atg2‐KD, atg7‐KD, or atg4‐KD) fruits, the milder buildup of autophagy results in earlier induction of climacteric ethylene production and, hence, earlier ripening. Below is a representation of ethylene production in Col‐0 or ATG‐deficient Arabidopsis seedlings and its impact on root elongation. Following ACC application, atg5, atg2, or atg7 mutants produce more ethylene, which may lead to their reduced root length under the ‘triple‐response’ assay conditions (ACC + darkness) compared with Col‐0. Significance was tested relative to controls (Col‐0 in (b, c) or mock (f)). Either unpaired and two‐tailed Student's t‐tests or the Mann–Whitney U test was performed based on population distribution and variance (see the Materials and Methods section). No asterisk, not significant; *, 0.01 < P < 0.05; **, 0.001 < P < 0.01; ***, P < 0.001.

ACC induces autophagy activity

It was previously reported that ACC induces autophagy based on GFP‐ATG8a foci numbers (without ConA), GFP‐release assay, and expression of the selective‐autophagy cargo receptor, NBR1 (Rodriguez et al., ²⁰²⁰). Considering our observations, we wanted to reaffirm this with complementary approaches. First, we tested ATG8 levels in Arabidopsis Col‐0 seedlings subjected to ACC or mock. Results showed that both total ATG8 (Fig. 3d, upper blot) and ATG8‐PE (Fig. 3d, lower blot) levels were higher in ACC‐treated seedlings at 72 and 96 h post treatment. Then, we followed GFP‐ATG8e in Arabidopsis roots subjected to combinations of ACC, ConA, or mock (Fig. 3e). Quantification revealed an increased number of puncta whether ACC was applied solely or combined with ConA, demonstrating that ACC increases autophagosomes production and their flux to the vacuole (Fig. 3f). Examination of GFP‐ATG8e using the GFP‐release assay supported the microscopy results, showing increased flux following 72 h of ACC treatment, which was reduced with the addition of ConA (Fig. 3g). To examine whether ACC reflects ethylene's function rather than ACC‐specific function, we further assessed GFP‐ATG8e flux following ethylene application. Ethylene (2 ppm) induced autophagy flux, yet at an earlier time point (48 h) than 20 μM ACC and quite transiently (Fig. S9a). We speculated that the time difference in reaching autophagy flux between ACC and ethylene might be related to the concentrations applied. Indeed, when 2 μM ACC was applied (instead of the 20 μM in Fig. 3g), an earlier flux at 48 h was apparent (Fig. S9b). Together with the report by Rodriguez et al. (2020), these results confirm the ability of ACC and ethylene to induce autophagy.

Discussion

It is well‐accepted that there is a physiological and molecular resemblance between ripening and senescence, as seen in dry fruits such as Arabidopsis siliques (Seymour et al., ²⁰¹³; Gómez et al., ²⁰¹⁴). Indeed, large‐scale data analysis suggested that three types of transcriptional circuits controlling ethylene‐dependent ripening have evolved from senescence or floral organ identity pathways (Lü et al., ²⁰¹⁸). Autophagy is an antisenescence and antiaging mechanism in plants and animals (Liu & Bassham, 2012; Minina et al., ²⁰¹⁸; Aman et al., ²⁰²¹), reconciling with a repressive effect on ripening. We propose that the milder buildup of autophagic capacity in the SlATG‐silenced fruits permits earlier and elevated ethylene production, leading to accelerated ripening (Fig. 3h, upper panel). Intriguingly, the participation of autophagy in nonclimacteric fruit ripening of pepper and strawberry was recently reported (López‐Vidal et al., ²⁰²⁰; Sánchez‐Sevilla et al., ²⁰²¹). Contrary to our results, it was concluded that autophagy acts in strawberry ripening promotion. If so, we cannot rule out the possibility of autophagy having contradicting roles in climacteric and nonclimacteric fruits. An intriguing possibility is that this reflects the different impacts of ethylene on both fruit types (Perotti et al., ²⁰²³). The increased sensitivity of Arabidopsis atg5, atg2, and atg7 to ACC, coupled with their elevated ethylene production (Figs 3a–c, S8), suggests that the repressive effect of autophagy is potentially widespread and may extend to other roles of ethylene beyond ripening, for example during root elongation (Fig. 3h, lower panel). However, we note that fruit ripening and root elongation are independent processes in which identical molecular mechanisms are not necessarily expected to exhibit similar functions. Therefore, the fact that autophagy restricts both of them does not suggest they are linked but rather the outcome of the repressive impact of autophagy on ethylene in both systems. Although the transcript level of genes involved in ethylene production was reportedly elevated in atg5 and atg9 Arabidopsis mutants (Masclaux‐Daubresse et al., ²⁰¹⁴), we have not detected a similar trend in our ATG4‐RNAi L18 fruits (Fig. S6), potentially highlighting the different manner of regulation between both systems.

As it might be difficult to interpret the buildup of autophagy as a repressive mechanism while ripening is ongoing, we wish to clarify this point. Let us assume that ripening progression is determined by the sum of mechanisms that promote and repress it, with the pace determined by their cumulative effects. In other words, such a highly regulated process would require functional breaks from its initiation to control process onset and their full performance at full speed (in our case, at the turning stage) in order to prevent loss of control (i.e. to mediate the transition from climacteric to postclimacteric ethylene production). In a similar manner, autophagy activity increases with leaf senescence progression, although it delays senescence, as is clearly seen in various atg mutants (Yoshimoto et al., ²⁰⁰⁹). That said, it is not yet fully understood how ripening is initiated. Changes to histone marks and DNA methylation seem to be necessary and are associated with ripening gene activation (Lü et al., ²⁰¹⁸; Li et al., ²⁰²¹). Here, we propose autophagy as an additional layer of regulation.

Several questions emerge from this study. First, how does autophagy limit ethylene? This may happen via the selective degradation of ethylene or ACC production components, such as ACC‐Synthase or ACC‐Oxidase enzymes (Houben & Van de Poel, 2019; Park et al., ²⁰²¹). Elimination of even further upstream precursors or other regulatory elements of ethylene production is another possibility. Finally, it cannot be excluded that autophagy may also regulate ethylene signaling components (Binder, 2020). Notably, the ability of ACC to induce autophagy highlights a potential feedback loop for ethylene regulation. Unlike ethylene, ACC levels were reported to continuously increase along with ripening (Van de Poel et al., ²⁰¹²). Additional questions are as follows what is the weight of ethylene in the well‐known early senescence phenotype of autophagy mutants? Especially since, so far, it has mostly been associated with salicylic acid (Yoshimoto et al., ²⁰⁰⁹). Finally, does autophagy regulate ripening solely via its impact on ethylene? Although the autophagy‐ethylene crosstalk seems significant for ripening progression, it is reasonable to assume that several components, such as protein complexes and organelles, not necessarily related to ethylene, would be recycled via selective‐autophagy during ripening (Clavel & Dagdas, 2021; Eckardt et al., ²⁰²⁴). Further studies are required to settle these questions.

Competing interests

None declared.

Author contributions

GK, PKP, EQ, S Mursalimov, JD, SA‐T, EL and S Michaeli designed experiments and conducted experiments. GK performed all virus‐induced gene silencing‐related experiments, data and statistical analysis, phylogenetic tree construction and autophagy activity assays in tomatoes. PKP generated the E8:ATG4‐RNAi tomato lines and performed most Arabidopsis experiments and SlNBR1 immunoblots. EQ assisted in several aspects of the study and conducted hue, firmness, ethylene measurements and tomato fruit GFP‐release assay blots. S Mursalimov performed confocal microscopy and quantification in Arabidopsis. JD examined the E8:ATG4‐RNAi lines and performed qPCRs. SA‐T assisted in hue, firmness, and ethylene measurements and performed the 1‐MCP experiments. EL performed qPCR. JXL and KS generated the GFP‐ATG8‐2.2 tomato lines. SÜ supervised students and contributed to the draft. All authors discussed the results and contributed to editing the manuscript. S Michaeli conceived the study, supervised the work, and wrote the manuscript. GK and PKP contributed equally to this work.

Supporting information

Fig. S1 Impact of ATG2 or ATG7 virus‐induced gene silencing on ATG8 abundance and of the latter on hue angle readings and ethylene production.

Fig. S2 SlATG4 expression increases along tomato fruit ripening progression.

Fig. S3 Constitutive silencing of SlATG4 results in NBR1a accumulation.

Fig. S4 E8::ATG4‐RNAi fruits display earlier ripening when attached to plants.

Fig. S5 Leaves of E8::ATG4‐RNAi lines do not show early senescence nor altered expression of the ATG4 fragment used to drive silencing in the fruits.

Fig. S6 ATG4 silencing does not result in transcript‐level alteration of key ripening‐related genes involved in ethylene biosynthesis or softening.

Fig. S7 Fruits of never‐ripe and colorless nonripening mutants exhibit markedly altered levels of SlNBR1 protein, while SlNBR1 transcript levels are largely not affected.

Fig. S8 Autophagy‐deficient atg5 and atg7 Arabidopsis seedlings display increased ethylene emission and 1‐aminocyclopropane‐1‐carboxylate sensitivity.

Fig. S9 Ethylene induces ATG8e flux.

Table S1 Primers used in this study.

Video S1 GFP‐ATG8‐2.2 labeled autophagosomes movement within fruit endocarp cells.

Video S2 GFP‐ATG8‐2.2 labeled autophagic bodies movement within concanamycin A‐treated endocarp cell vacuoles of an orange‐stage fruit.

Please note: Wiley is not responsible for the content or functionality of any Supporting Information supplied by the authors. Any queries (other than missing material) should be directed to the New Phytologist Central Office.

Data availability

Accession nos. associated with this work: Solyc01g108160 (ATG2), Solyc11g068930 (ATG7), Solyc01g006230 (ATG4), Solyc08g078820 (SlATG8‐1.1), Solyc07g064680 (SlATG8‐2.2), Solyc03g112230 (SlNBR1a), and Solyc06g071770 (SlNBR1b). All research data are contained in the article (Figs 1, 2, 3) and the Supporting Information (Figs S1–S9; Table S1; Videos S1, S2).