A transcriptomic, metabolomic and cellular approach to the physiological adaptation of tomato fruit to high temperature

Funding information H2020 Food, Grant/Award Number: H2020 programme TomGEM / 679766 Abstract High temperatures can negatively influence plant growth and development. Besides yield, the effects of heat stress on fruit quality traits remain poorly characterised. In tomato, insights into how fruits regulate cellular metabolism in response to heat stress could contribute to the development of heat-tolerant varieties, without detrimental effects on quality. In the present study, the changes occurring in wild type tomato fruits after exposure to transient heat stress have been elucidated at the transcriptome, cellular and metabolite level. An impact on fruit quality was evident as nutritional attributes changed in response to heat stress. Fruit carotenogenesis was affected, predominantly at the stage of phytoene formation, although altered desaturation/isomerisation arose during the transient exposure to high temperatures. Plastidial isoprenoid compounds showed subtle alterations in their distribution within chromoplast sub-compartments. Metabolite profiling suggests limited effects on primary/intermediary metabolism but lipid remodelling was evident. The heat-induced molecular signatures included the accumulation of sucrose and triacylglycerols, and a decrease in the degree of membrane lipid unsaturation, which influenced the volatile profile. Collectively, these data provide valuable insights into the underlying biochemical and molecular adaptation of fruit to heat stress and will impact on our ability to develop future climate resilient tomato varieties.


| INTRODUCTION
Plant growth and development are vulnerable to abiotic stresses such as higher temperatures, leading to detrimental effects on agricultural yield (Battisti & Naylor, 2009;Bita & Gerats, 2013). Increased global mean temperatures and extreme climate-related events are now occurring with increased frequency and intensity (Horton et al., 2015;IPCC, 2013). The consequences arising from the warmer temperatures have already been observed and hold potential to destabilise food systems and to threaten local to global food security (Lesk et al., 2016;Zhao et al., 2017). Higher temperatures can trigger developmental, physiological, cellular stress responses in plants which is highly dependent on duration and severity of stress as well as sensitivity of plant cell type and developmental stage Ohama et al., 2017). When heat stress is moderate, the changes occurring to crops may be rapidly reversible, but severe episodes of elevated temperatures are irreversible and can lead to crop failure (Zhang et al., 2010). Tomato (Solanum lycopersicum) is among the crops for which yield losses have been well documented when different high-temperature regimes occur during reproductive phase. Yield (e.g., fruit number and weight) is adversely affected by daily mean temperatures above 29 C, ranging from a few days-when pollen development or fruit set is disturbed-to a whole developmental period (Peet et al., 1998;Pressman et al., 2002;Sato et al., 2000Sato et al., , 2006. Notably, the tomato vegetative development is less sensitive to episodic temperature increases as structural damages of photosystem II were not detected for temperatures reaching 38 C (Lu et al., 2017;Spicher et al., 2017). Nevertheless, the influence of heat stress on tomato nutritional composition and quality has received less attention. This is surprising considering tomato is one of the most widely consumed fruits globally (Bergougnoux, 2014), it is grown worldwide and is an important source of vitamins and bioactives in the human diet (Viuda-Martos et al., 2014).
Metabolite composition of tomato fruit is affected by adverse environmental conditions (Quinet et al., 2019). While some abiotic stresses as water deficit lead to an increase in sugars, organic acids, vitamin C and carotenoids (Albert et al., 2016), high-temperature conditions seem to have a diverse impact on fruit quality. Most information relies on studies performed with differing post-harvest conditions. Early studies have demonstrated that the long-known failure to achieve normal pigmentation of excised tomatoes ripening at high temperatures (Tomes, 1963) is associated with changes in ethylene production, fruit softening and colour development, a phenotype that could be reversed when fruits were transferred to optimal temperature (Lurie et al., 1996;Picton & Grierson, 1988). More recent evaluations corroborated these findings highlighting the heatsensitivity of antioxidant accumulation as carotenoids and vitamin C (Gautier et al., 2008;Massot et al., 2013). For vine-attached fruits, the detrimental effects of high ambient temperature on bioactive compounds have also been observed (Hernández et al., 2015;Mulholland et al., 2003). Vitamin C (ascorbate) and carotenoids (provitamin A) are significantly lower when a heat-stress treatment is imposed during the advanced stages of fruit development. However, when the temperature is raised at earlier stages, the lack of effects indicates differential thermo-sensitivity of fruit developmental stages (Hernández et al., 2015). While these changes in tomato metabolism have been reported, a comprehensive evaluation exploring the cellular and molecular modifications associated with heat response awaits elucidation.
Alteration in metabolite composition can be direct consequence of several molecular mechanisms underlying the heat stress responses. A typical signature response is a wide-scale transient reprogramming of gene expression, including the expression of heat shock proteins (HSPs, Kotak et al., 2007;Ohama et al., 2017). The majority of HSPs function as molecular chaperones which act not only in protection against stress damage but also in folding, intracellular distribution and degradation of proteins (Mishra et al., 2002). Interestingly, HSPs seem to be important for tomato fruit ripening (Fragkostefanakis et al., 2015;Neta-Sharir et al., 2005). To maintain membrane stability and deal with oxidative stress generated in response to heat, plants induce synthesis of hormones, and other protective molecules including osmoprotectants and antioxidants (Gray & Brady, 2016;Wahid et al., 2007). The antioxidant network, in part, is based on the action of several molecules including carotenoids, tocopherols (vitamin E), ascorbate and phenolic compounds, all with potential to contribute to fruit nutritional proprieties (Li et al., 2018b).
Furthermore, as heat can lead to membrane damage caused by lipid hyper-fluidity and lipid peroxidation, modifying membrane lipid composition, particularly the acyl moieties of glycerolipids, is another critical aspect of plant thermotolerance (Falcone et al., 2004;Murakami et al., 2000;Higashi & Saito, 2019) that is linked to fruit quality.
Changes in 18-carbon (C18) polyunsaturated fatty acids can affect the enzymatic/nonenzymatic formation of oxylipins derived thereof (Feussner & Wasternack, 2002). Lipoxygenase (LOX) pathway produces hydroperoxide intermediates for the synthesis of different compounds, including jasmonic acid and volatiles, the latter an essential aspect of fruit quality (Tieman et al., 2017). Lipid-derived signaling molecules as oxidised derivatives constitute important components of heat stress response, acting, for example, in the control of gene expression related to protective responses (Balogh et al., 2013;Farmer & Mueller, 2013;Hou et al., 2016).
In this present study, transcriptomic, metabolomic and cellular analysis has been applied to tomato fruits at different ripening transitions, following exposure to transient high-temperature treatment (40 C day/ 30 C night). These conditions may replicate events of heat stress experienced during commercial production. Collectively, the data provide new insights into the metabolic plasticity of tomato fruit to heat stress episodes and may contribute generically to the development of climatic resilient crops.

| Isoprenoid determination and quantification
Isoprenoids (carotenoids, tocochromanols and chlorophylls) were extracted from lyophilised tissue powder (15 mg) as described by Enfissi et al. (2010). Compounds were analysed by reverse-phase chromatography using an ultra-performance liquid chromatography (UPLC) system (Acquity, Waters) equipped with a Photo Diode Array (PDA) detector (Acquity, Waters). A UPLC BEH-C18 column (100 mm × 2.1 mm; 1.7 μm, Acquity, Waters) was used for separation as described by Nogueira et al. (2013). Peak identification was achieved by comparison of characteristic UV/Vis spectrum with authentic standards, reference spectra and retention times (Fraser et al., 2007). Quantification was performed using dose-response curves obtained from authentic standards.

| Metabolite profiling by gas chromatography (GC)-MS
Polar extracts were prepared from freeze-dried fruit powder (10 mg), extracted with 1 ml of solution containing methanol and water acidified with 0.1% formic acid [80:29.9:0.1, (v/v/v)] and agitated for 1 hr. After centrifugation, the polar extract was spiked with ribitol (1 mg/ml in MeOH; 10 μg final concentration) as the internal standard. For nonpolar extracts, alkaline hydrolysis with KOH was performed with a fruit powder aliquot (10 mg) during 1 hr at 40 C followed by extraction as described for isoprenoids. Nonpolar extracts were spiked with deuterated myristic acid-d27 as the internal standard. The dried residues were derivatised in methoxyamine hydrochloride (in pyridine) followed by silylation with N-methyl trimethylsilyl trifluoroacetamide. The GC-MS analysis was achieved on Agilent 7890A GC system interfaced with a 5975C mass-selective detector as described in Uluisik et al. (2016).
For lipid and fatty acid compositional analysis, extraction was performed as described for isoprenoids and resolved on highperformance thin-layer chromatography (HTLC) silica gel 60 F254 plates (Merck) developed in a solvent mixture of acetone, toluene, and water [91:30:7, (v/v/v)]. Regions containing the lipid classes were identified based on the comparison with authentic standards visualised with iodine vapour and scraped from the HTLC plate. Elution and conversion to fatty acid methyl esters (FAMEs) by acid-catalysed transmethylation, followed by quantification using GC-MS were performed as previously described by Nogueira et al. (2013). FAMEs were quantified using myristic-d27 acid as an internal standard.
Components were identified using a mass spectral library built from in-house standards and NIST11 database. Each analytical batch was validated with quality control samples.

| Profiling of volatile compounds by GC-MS
Frozen fruit samples were ground in liquid N 2 and aliquots (0.5 g) used for the analysis of volatile compounds. Homogenates were weighed out into screw-top headspace amber glass vials (20 ml) and spiked in with deuterated acetophenone-d3 as internal standard (20 ppb). Capped vials were incubated at 40 C and shaken for 30 min. Volatile compounds were then adsorbed onto a SMPE fibre (Car/DVB/PDSM) for 20 min, followed by desorption into the injection port for 5 min. Chromatographic separation was conducted in a DB-5MS 30 m × 250 μm × 0.25 μm column (J&W Scientific, Folsom, CA), equipped with a 10 m guard column and using a step-temperature gradient from 40 to 300 C at 5 C/min. The linear temperature gradient included a 2 min hold-temperature and then steps at 40, 120, 250 C and 5 min at 300 C. Helium was employed as the carrier gas and the flow rate was 1 ml/min. The inlet and the mass spectrometer transfer line were heated to 250 C. A 7890B-5977B GC-MS system (Agilent Technologies, Palo Alto, CA) was used in splitless mode, and data processing and analysis proceeded using AMDIS (version 2.73) software.

| Subchromoplast fractionation
Chromoplasts were isolated from fruits (90 g) at B3 to B4 stage, and sub-compartments were fractionated using a discontinuous gradient of sucrose, according to Nogueira et al. (2013).

| Transmission electron microscopy
Pericarp fruit segments were fixed at room temperature in solution [3% (v/v) glutaraldehyde, 4% (v/v) formaldehyde buffered with 0.1 M PIPES buffer pH 7.2] and then stored at 4 C for at least 24 hr until processing. Samples were post-fixed in buffered 1% (w/v) osmium tetroxide and uranyl acetate, washed, dehydrated in a graded series of acetone, and embedded in resin. Ultrathin sections were stained with Reynolds lead citrate and imaged on a Tecnai T12 Transmission Electron Microscope (Field Electron and Ion Company).

| qPCR expression analyses
Total RNA was extracted from frozen leaves and fruit pericarps using the RNeasy kit (Qiagen) according to manufacturer's instructions. RNA from at least four biological replicates was prepared from each tissue and ripening stage. RNA quality was assessed by agarose gel electrophoresis. Total RNA (1 μg) was treated with DNase and converted into cDNA using the QuantiTect Reverse Transcription kit (Qiagen), according to the manufacturer's protocols. Real-time quantitative PCR (qPCR) assays were performed in technical duplicates using RotorGene SYBR green PCR kit (Qiagen) on Rotor-Gene Q, with approximately 10 ng of reverse-transcribed RNA. Primer sequences are listed in Table S1. Relative expression was calculated as described by Quadrana et al. (2013). For reference gene selection, expression stability of five known reference genes (CAC, EXP, GAGA, ACT1 and ACT2) (Cheng et al., 2017;Exposito-Rodriguez et al., 2008) was evaluated on control and heat-stressed samples using GeNorm (Vandesompele et al., 2002).

| RNA sequencing
Total RNA from three biological replicates samples was isolated using Trizol RNA Purification kit (Thermo Fisher Scientific). cDNA libraries were prepared and sequenced by IGA Technology Services facility (Udine, Italy). Single-end sequence reads (75 nt) at a read depth of 31.3 million reads on the average per sample (24.1 to 39.9 M reads) were obtained from the NextSeq500 platform (Illumina). Raw reads were processed using ERNE (Del Fabbro et al., 2013) and Cutadapt (Martin, 2011) software. The reads were mapped onto tomato genome (S. lycopersicum, cv. Heinz) reference SL3.0, with gene models ITAG3.10, using STAR (Dobin et al., 2013) applying default parameters. Assembling and quantification of full-length transcripts were accomplished by Stringtie (Pertea et al., 2015). The counting was achieved by HTseqcount (Anders et al., 2015). Gene ontology (GO) term annotation was performed using Blast2GO Pro (version 5.2.5) (Conesa et al., 2005).
All raw RNA sequencing data are available on NCBI, under the Bioproject accession number PRJNA603594.

| Data analyses
Significant differences between the control and heat-stressed conditions were determined by Student's t test or ANOVA followed by a Dunnett's multiple comparison with the level of significance set to 0.05, using GraphPad Prism software. For pair-wise differential expression analysis of transcriptome data, statistical analyses were performed by DeSeq2 (Love et al., 2014). Differentially expressed genes were determined using false discovery rate (FDR) ≤ 0.01 (adjusted p-value) and jfold-changej ≥ 1.5 (or jlog 2 FCj ≥ 0.58). GO enrichment analysis with Fisher's Exact Test was conducted using Blast2GO.

| Heat stress at advanced ripening stages negatively affects carotenoid accumulation
Plants grown under control conditions (C) were exposed to 48 hr high-temperature treatment (H). The H was imposed when plants possessed fruits undergoing specific ripening transitions, from mature green to breaker (H B ) and yellow to light-red transition (H B3 ) ( Figure 1). For comparison, an additional H treatment at early to late mature green (H MG ) was included. After H, plants were returned to control conditions (recovery, R) until fruits ripened (B7 stage).
The levels of plastidial isoprenoids were determined by UPLC-PDA. H negatively influenced the carotenoid levels in tomato fruit, and changes from the heat stress were highly dependent on the fruit stage used (Table 1). Remarkably, phytoene, the first carotene product of the pathway, was reduced both in H B (10-fold) and H B3 (4.6-fold) compared to their corresponding non-stressed fruits. The other carotenes in the pathway, phytofluene and ζ-carotene, when detected, responded similarly to phytoene when H was applied.
Lycopene, the predominant carotenoid found in red ripe tomato, was significantly lower, below detection in H B fruits and reduced by 50% in H B3 fruits, compared to corresponding non-stressed controls.
Interestingly, lycopene levels only partially recovered in H B3 R4 fruits, to about 60% of the content found in ripe C B7 fruit, despite the levels of phytoene and phytofluene precursors being fully restored at this stage ( Table 1).
The total fruit carotenoid content was consistent with lycopene levels at later stages of ripening. While lycopene decreased in H B3 and H B3 R4 fruits, total carotenoids remained unchanged both in H B fruits and in those fruits allowed to recover from heat stress (H B R7). By contrast, H MG carotenoid levels were mostly similar to the control, except for a modest increase in β-carotene levels. The same response was also observed in heat-stressed leaves (H L ). Additionally, lutein levels as well as total carotenoid showed an increase in H L compared to control conditions. F I G U R E 1 Outline of heat stress experiment. Tomato plants were kept at 25 C day/20 C night (control, C) or exposed to transient hightemperature treatment (H) at 40 C/30 C (day/night) for 48 hr. After H, fruits at the following stages were harvested: mature green (H MG ), breaker (H B ) and 3 days post-breaker (H B3 ). Fruits allowed to recover (R) under normal conditions were harvested at ripe stage (H B R7 and H B3 R4) T A B L E 1 Transient changes in isoprenoid profile of tomato fruits and leaves exposed to high-temperature treatment 3.2 | Metabolite profiling of tomato fruit exposed to transient heat stress Metabolite profiling using GC-MS was carried out on fruits exposed to heat stress and concurrently samples were collected for transcriptome analysis. Principal component analysis (PCA) was used to compare primary metabolism of fruits under C, H and R conditions.
The score plot obtained from polar and nonpolar extracts could not discriminate non-stressed and heat-stressed conditions among the fruit stages evaluated, though a clear separation between early-and late-ripening stages was achieved ( Figure. 2a, b). Overall primary metabolism remained unchanged after H and followed by R (Table S2).
Our high-temperature treatment did not appreciably affect the levels of known osmoregulators such as proline and GABA, and only a few metabolites responded significantly to H mostly in a fruit stagedependent manner (Figure 2c, d). Sucrose was very responsive to heat, consistently accumulating in fruits (H MG and H B ) at early ripe stages ( Figure 2c). Some amino acids changed in content, particularly threonine was increased in H B and H B3 . The tricarboxylic acid (TCA) F I G U R E 2 Effect of heat stress on tomato fruit metabolite profiling. Scores plot obtained by Principal Component Analysis (PCA) for metabolite levels measured in polar (a) and nonpolar (b) extracts. Metabolites at mature green, MG (c), breaker, B (d), breaker+3, B3 (e), red ripe, B7 (f) stage of fruits kept at C, exposed to H or followed R. Quantification was determined relative to the internal standard and values are presented as mean ± SD from five biological replicates. Nonpolar compounds are shown as left insets in the graphs. Only significant changes compared to respective control are shown (pair-wise t test corrected for multiple comparison using Holm-Sidak's post-test; * Adjusted p < .05, ** p < .01, *** p < .001). Full data set available in Table S2 (Table S2).

| Fruit lipid metabolism is highly responsive to heat
Lipid remodelling while-decreasing the level of lipid unsaturation is a crucial aspect of plant thermotolerance under suboptimal temperature conditions (Falcone et al., 2004). Nevertheless, the analysis of the total lipid fraction by GC-MS showed no significant differences between C and H in fatty acid composition. To increase the sensitivity and address the potential changes in complex lipid moieties, lipid spe-  (Table S3). In contrast, H B fatty acid composition of the TAG fraction showed an opposite response with higher levels of the trienoic acid C18:3 proportion compared to control C B (Figure 3).
Overall, tomato fruit response to high-temperature conditions includes lipid remodelling, which leads to TAG accumulation, and decreasing the level of membrane lipid unsaturation, particularly trienoic acid composition. The volatile organic compounds (VOC) analysis revealed C18 fatty acid-derived volatiles were altered in heat-stressed fruits ( unaffected except for the lycopene derived 6-methyl-5-hepten-2-one; which was found to be lower in H B3 compared to C B3 . These data complement the lower levels of lycopene found in these samples following H. Finally, the accumulation of different VOCs belonging to terpenoid parent molecules, such as α-pinene, δ-4-carene, cymene, β-phellandrene was detected in at least one ripening stage either immediately after stress and/or after recovering. A similar trend was found for the phenolic derived molecule o-guaiacol (Table S4).
3.4 | Chromoplast structure of fruits exposed to high temperatures Heat stress-induced changes in carotenoids, tocopherols and neutral lipid levels may be associated with perturbations not only in the metabolic pathways but also in compound sequestration (Spicher et al., 2017;Zhang et al., 2010). To ascertain whether high temperature alters the distribution of liposoluble antioxidants into plastidial sub-compartments, analysis of fractionated chromoplast from heatstressed and non-stressed fruits was carried out (Figure 4).
First, the total amounts obtained from the sum of all fractions (i.e., plastoglobuli, envelope membranes, stroma, thylakoids) corroborated the lower levels of phytoene, phytofluene and lycopene in H B3 fruits (Table S5) Figure 5a).

| Transcriptome analysis
Considering the most specific GO terms, up-regulated genes were significantly enriched for only a few biological process GOs F I G U R E 5 Gene expression changes associated with heat stress in tomato fruit. (a) Venn-diagrams of the up-regulated and down-regulated differentially expressed (DE) genes following H at B and B3 fruit stages. (b) GO terms enriched in the common set of DE genes observed at B and B3 stages according to Fisher's exact test (FDR < 0.05). Only the most specific GO terms for biological process category were shown. (c) Relative expression of HsfA2 and PSY1 by qPCR. Abbreviations and colour codes for fruit treatments are the same as in Figure 1. Nonstressed leaves (C L ) and heat-stressed (H L , brown bars) leaf samples were included for comparison. Values are expression levels normalized to CAC and ACT2 reference genes (mean ± SEM of at least four biological replicates) from samples kept at C, exposed to H or followed R. Significant differences (Student's t test, * p < .05, ** p < .01, *** p < .001) between heat and control conditions at corresponding organ/developmental stage are shown [Colour figure can be viewed at wileyonlinelibrary.com] (10) associated with general terms as RNA processing, mitotic cell cycle and chromatin remodelling (Figure 5b). When a more relaxed significance threshold (p-value <.01) was applied, GO terms as "mRNA splicing, via spliceosome" (GO:0000398; p-value 2.15 E −03 ), and molecular function "SWI/SNF superfamily-type complex" (GO:0070603, p-value 4.68 E −04 ), the latter acting in chromatin remodelling processes (Table S7) (Table S6).
The epigenetic mechanisms also featured when the heat-induced genes were queried for each comparison separately; "histone modification" and "RNA processing" processes were overrepresented among up-regulated H B3 genes (Table S7).
By contrast, genes down-regulated by H were enriched for GOs mainly associated with defence response to biotic stress, hormone synthesis and signaling pathway, metabolic processes related to lipids, carbohydrate, amino acids, and redox-related compound glutathione ( Figure 5b). GO terms found overrepresented such as "carotenoid biosynthetic process," "carbohydrate metabolic process," "alpha-amino acid biosynthetic process" are closely related to the metabolic reprogramming triggered by heat in fruits.
Extreme temperature is known to induce the expression of HEAT SHOCK TRANSCRIPTION FACTORS (Hsfs). In tomato, HsfA1, which is constitutively expressed and post-translationally regulated, is responsible for the initial heat stress response controlling the HS-induced expression of HsfA2 and HsfA3 (Fragkostefanakis et al., 2015;von Koskull-Döring et al., 2007). Both HsfA2 and HsfA3 were found upregulated under H in RNA-seq dataset compared to control conditions (Table S6). qPCR assays confirmed higher HsfA2 transcript levels not only for H B and H B3 but also in heat-stressed leaves (H L ), though significant differences for H MG were not detected. Importantly, higher HsfA2 transcripts were not sustained after R (Figure 5c).

| Ripening regulators and targeted pathways
Given that processes related to fruit ripening were significantly enriched among the genes repressed by heat (Figure 5b (Table S8).
The sucrose accumulation in heat-stressed fruit may be related to differences in sucrose turnover enzyme abundance (Qin et al., 2016).
Although sucrose accumulated only in H B fruits (Figure 2c), perturbations in the expression of genes related to sugar metabolism were detected at both B and B3 stages, with notable repression of the transcripts of vascular invertase (VI), cell wall invertase (LIN5) and sucrose synthase (Susy6) (Figure 6a). Sucrose turnover enzymes coding genes were higher repressed in H B than H B3 (e.g., for VI, fold change ratio was about four-fold and two-fold in C B vs H B and C B3 vs H B3 , respectively), though the transcripts encoding a vacuolar invertase inhibitor (VIF) were found slightly up-regulated in H B3 , which suggests further capping to invertase activity.

| Lipid metabolism
Upon heat stress, lipid metabolism-related transcripts were found to be overrepresented among downregulated genes ( Figure 5(b)).
From a manually curated list derived from tomato loci showing homology to Arabidopsis genes associated with acyl-lipid metabolism (Higashi et al., 2015), a subset of genes putatively involved in the plastidial de novo fatty acid biosynthesis was suppressed under H in both B and B3 stages (Figure 6b, Table S9) For TAG biosynthesis, higher levels of expression were associated with biosynthetic pathway genes. Different types of enzymes can synthesise TAG from DAG, including acyl-CoA dependent enzymes, acyl-CoA:DAG acyltransferases (DGATs), and diacylglycerol acyltransferase (PDAT) which uses PL as acyl donor (Fan et al., 2017). Genes encoding DGAT and PDAT were upregulated under heat, following the higher levels of TAG observed in H B and H B3 (Figure 3). A similar trend was found for the transcripts encoding proteins associated with TAG hydrolysis, which were predominantly upregulated at both stages analysed (Figure 6b).
Fatty acid β-oxidation pathway was overrepresented among the genes down-regulated by heat (Figure 5b). In accordance, the genes encoding enzymes were found actively repressed under heat, such as acyl-CoA oxidase (ACX1a, Li et al., 2005) and members of the multifunctional protein (MFP), as well as the peroxisomal isoform of long-chain acyl-CoA synthetase (LACS), which activates free fatty acids to acyl-CoA thioesters to generate acyl-CoA derivatives. Indeed, genes associated with oxylipin biosynthesis were repressed after H, including allene oxide cyclase (AOC3), allene oxide synthase (AOS2) and acyl-hydrolase patatin-like, involved in the production of jasmonate from polyunsaturated fatty acids.
As the levels of β-sitosterol increased after heat stress (Figure 2e), the transcripts encoding sterol 22-desaturase (CYP710A11), involved in the conversion of β-sitosterol to stigmasterol, were checked, revealing a heat-sensitive expression pattern.

| DISCUSSION
Our study has undertaken an integrative approach to address the metabolic, cellular and molecular changes associated with transient heat stress imposed on tomato fruits, elucidating several key features that impact on fruit quality traits.

| Nutritional attributes were altered by heat stress
Our findings revealed that transient heat stress can alter carotenoid accumulation in tomato fruits, with sensitivity to temperature increasing as ripening advanced. The negative impact of heat on carotenoid levels was associated with changes in the initial steps in carotene formation.
Our study showed that ripe fruits have tremendous plasticity to restore carotenoid levels following a heat wave, suggesting no permanent damage was achieved. The temperature of 40 C caused moderate stress in tomato as reported earlier (Spicher et al., 2017).
Reversible effects of heat treatment have been observed on vine detached fruits previously (Lurie et al., 1996). Nevertheless, the lower carotenoids levels in ripe fruits experiencing heat at B3 transition implies that the length of the recovery period may be critical. Boosting in carotenoid synthesis during fruit ripening is achieved predominantly by the up-regulation of genes encoding key biosynthetic enzymes (Hirschberg, 2001;Enfissi et al., 2017). Thus, the heat-induced transcriptional misregulation at advanced ripening stages (Figure 6a) may explain the decrease in fruit carotenoid levels. Firstly, both phytoene formation and subsequent isomerisation are potentially compromised as the expression of fruit-specific PSY1 (Figure 5), encoding the major flux-controlling enzyme of carotenogenic pathway (Fraser et al., 2002(Fraser et al., , 2007, and CrtISO (Isaacson et al., 2002) were repressed by heat. Secondly, efficient carotenoid desaturation conducted by phytoene desaturase (PDS) and ζ-carotene desaturase (ZDS) depends on the redox status of plastoquinone/plastoquinol pool dependent on the activity of PTOX (Shahbazi et al., 2007) whose transcripts were found heatsensitive. Finally, methyl-erythritol phosphate (MEP)-derived precursors for carotenogenesis (Almeida et al., 2015;Nogueira et al., 2018) might be altered as the expression of ISPE, GGPPS3 and the ripening-  , 2018). Therefore, it is expected some contribution of post-translational mechanisms curbing the activity of carotenoidrelated enzymes under heat stress. It is worth noting that this response is likely fruit-specific as heat-stressed leaves showed an opposite effect on carotenoids; indeed, the increased levels of lutein observed in H L were consistent with a previous report using tomato (Spicher et al., 2017).
Higher vitamin E levels found in heat-stressed fruits and their recovered counterparts may serve as a molecular signature of fruits which have experienced stress previously irrespective of their developmental stage. The production of tocopherols has been linked to high-temperature response in tomato leaves supported by transcriptional regulation (Spicher et al. 2016(Spicher et al. , 2017. In our study, the lack of correlation between fruit tocopherol content (Table 1) and the expression of genes involved in tocopherol biosynthesis (Figure 6a), could be associated with the redirection of isoprenoid precursors from carotenoid formation into tocopherol formation instead (Almeida et al., 2015;Fraser et al., 2007). Importantly, carotenoid and tocopherol heat-responsive genes typically exhibit ripeningassociated expression pattern (Quadrana et al., 2013;Sato et al., 2012;) and the changes observed may be due to inhibition of fruit ripening.
4.2 | Ripening related processes are misregulated in heat-stressed fruits In tomato, fruit ripening encompasses highly coordinated processes orchestrated by a network of interacting genes and signaling pathways, which involves differentiation of chloroplasts into chromoplasts Seymour et al., 2013). A peak in ethylene production and burst in cellular respiration are associated with profound metabolic transitions, leading to alterations not only in pigmentation but also in sugar accumulation, tissue softening and volatile production (Klee & Giovannoni, 2011;Rambla et al., 2015).
Besides, repressors of carotenogenesis PIF1a, DET1 and COP1, which transduce phytochrome-sensed changes in the environmental light, hence affecting carotenoid biosynthesis and plastid development (Enfissi et al., 2010;Liu et al., 2004;Llorente et al., 2015), were found to be up-regulated under heat conditions. As PIF1a binds to PSY1 promoter region and represses PSY1 transcription in tomato fruits (Llorente et al., 2015), the higher PIF1a expression may cause temperature-induced PSY1 repression.  Figure S1). Moreover, consistent with previous studies in tomato (Mishra et al., 2002), fruit heat transcriptional response seems to be mediated by Hsfs ( Figure 5). Finally, our transcriptome analysis suggested a role of epigenetic mechanisms mediating heat-induced transcriptional changes as chromatin remodelling and histone methylation were enriched among up-regulated genes in response to heat ( Figure 5, Table S7). Epigenetic mechanisms as DNA methylation add another layer of regulation for the tomato ripening program. DNA methylation relies, in part, on the RNA directed DNA methylation (RdDM) pathway, dependent on small RNAs and the activity of DRM and DRD1 (Gallusci et al., 2016). The up-regulation of tomato homologs DRM1 and DRD1 (Table S6) might contribute to rearrangements of epigenome landscape under high-temperature treatment and raises the possibility that plant adaptive responses to heat mediated by epigenetic mechanisms (Li et al., 2018a;Quadrana et al., 2019;Ohama et al., 2017) also operate in tomato fruit.

| Fruit primary metabolism changes in response to heat stress
The absence of signatures commonly shared through all stressed samples ( Figure 2, Table S2) further supports the idea the thermoresponsive is highly dependent on fruit stage. Among the known osmoregulators, higher threonine levels at advanced ripening stages are in line with its conserved biomarker for abiotic stress, accumulating in Arabidopsis leaves under heat stress (Obata & Fernie, 2012).
Notably, citric and malic acids that contribute most to the typical acidity of tomato fruit (Baldwin et al., 2008) remained unaffected.
Sucrose accumulation in heat-stressed fruit at early ripening stages ( Figure 2) correlated with the sensitivity of sucrose metabolism to high temperatures previously reported in tomato male reproductive system and in fruits after pollination (Li et al., 2012;Liu et al., 2016;Sato et al., 2006). Indeed, enhanced LIN and VI activity in tomato has been associated with fruit thermotolerance at early developmental stages (Li et al., 2012;Liu et al., 2016). It is known that, at late-ripening stages, sucrose accumulation is limited since invertase activities intensify as ripening progresses, with VI controlling sucrose/hexose ratio (Biais et al., 2014;Klann et al., 1996;Qin et al., 2016;Yelle et al., 1991). In H B fruits, lower VI transcripts may explain why sucrose increased. However, control sucrose levels found at H B3 did not correlate with down-regulation of VI, LIN5 and Susy6. In this case, sugar metabolic fluxes at B3 stage may prevent sucrose from accumulating under heat. Lack of correlation between invertase activity and sucrose/hexose levels has been reported in tomato fruits of lines with increased LIN activity (Liu et al., 2016).
Effects on fruits seem to be minor compared to the vegetative system or pollen development, where altered carbon metabolism upon exposure to high temperatures can promote yield losses (Ruan et al., 2010;Rieu et al., 2017). Moreover, fluctuations in sugar levels are important, as cell signals, since they can act in crosstalk with hormones (e.g., auxin) and reactive oxygen species (ROS) signaling pathways during stress responses; sugars can also contribute to stress alleviation by facilitating production of HSP even in reproductive systems, property that correlates to high fruit sensitivity under stress (Liu et al., 2013). Molecule signaling and protective roles are, therefore, possible to intersect under heat stress in fruits.

| Lipid remodelling triggered by heat stress in fruits
Lipid remodelling was a pronounced feature of heat-stressed fruits, leading to TAG accumulation, and decreasing the level of membrane lipid unsaturation (Figure 3, Table S3). Polyunsaturated acyl chains contribute to membrane fluidity and stability and, in response to higher temperatures, the degree of unsaturation is decreased to maintain optimal fluidity and integrity of membranes (Nishida & Murata, 1996;Murakami et al., 2000;Falcone et al., 2004;Zheng et al., 2011). Moreover, the observed changes in sterols might also be linked to membrane stability. Heat-induced β-sitosterol accumulation may contribute to control membrane permeability and membrane protein activity (Guo et al., 2019).
TAG accumulation associated with heat stress-induced lipid remodelling has been reported in photosynthetic organisms (Légeret et al. 2016;Mueller et al., 2015;Narayanan et al., 2016 substantial heat-induced decrease in C18:3/C18:2 ratio of plastid membrane lipids in fruits is similar to previous reports in heat-stressed tomato leaves (Spicher et al., 2016). Besides the prevention of physical-chemical damages of membranes, it has been proposed that selective decline of trienoic acid acyl moieties might confer survival advantage imposed by cellular oxidative stress associated to excessive ROS generated under heat Légeret et al., 2016). As galactolipids are highly enriched in polyunsaturated fatty acids, and thus easily prone to lipid peroxidation, photosynthetic organisms may transfer trienoic fatty acids from membrane lipids to TAG sequestered in lipid droplets as a strategy to control the extension of lipid oxidation (Du et al., 2018;Légeret et al., 2016).
Our transcriptome data provided insights into molecular mechanisms supporting lipid remodelling (Figure 6b, Table S9). The heatinduced decrease in the level of plastid lipid unsaturation coincided with the down-regulation of FADs, mainly the FAD7/FAD8 encoding the plastidial ω3 desaturase. Transcripts of the ER-counterpart desaturase (FAD3) were only significantly repressed in H B3 , following the temperature-induced decrease of extraplastidial 18:3-acyl-containing lipids specific to this stage ( Figure 3). Lipid remodelling may also be supported for the up-regulation of (a) SFR2/GGGT, whose corresponding enzyme contributes to diminish the MGDG/ oligogalactolipids ratio and to release DAG further used for TAG biosynthesis (Higashi & Saito, 2019;Moellering & Benning, 2011), and (b) DGAT and PDAT, which encode TAG biosynthetic-related enzymes (Fan et al., 2017). Together, these data suggest that the higher amounts of 18:3-acyl-containing TAGs upon heat, at B stage, may have been derived from C18:3 released from membrane lipids than from de novo synthesised fatty acids (Légeret et al., 2016), therefore reflecting the changes in the proportion of membrane glycerolipid composition (Higashi & Saito, 2019). By contrast, at B3 stage, 18:3 acyl moieties are likely redirected to lipid oxidation pathways (Schilmiller et al., 2007), for example, volatile production. In tomato, levels of C18:2 and C18:3 positively correlate to volatile derivatives hexanal and hexenal, respectively (Domínguez et al., 2010;Ties & Barringer, 2012 A decrease in HPL activity by heat might enhance the hydroperoxide pool, which can be redirected towards the C5 branch of LOXpathway as proposed previously based on tomato HPL-deficient lines (Shen et al., 2014). Notably, the tomato fruit volatile profile is highly sensitive to heat exhibiting alterations even when the stress ceased.
In conclusion, our findings illustrated the impact of brief exposure to high-temperature events on tomato fruit quality and revealed potential molecular mechanisms associated with heat response (Figure 7). Depending on the ripening stage, heat may have underestimated yet significant effects on nutritionally value and other quality-related attributes in tomato, with sensitivity to high temperature increasing in more advanced ripening stages. Several heat-stress responsive genes, including fruit ripening regulators, have been identified from transcriptome analysis correlating with the metabolite changes. Collectively the data acquired provides a significant advancement to our understanding of fruit metabolic reprogramming associated with heat stress. It is now clear that cold storage is not the only stress affecting fruit quality but perturbations in heat will also alter quality attributes. These data provide an exploitable resource for the development of climate resilient crop varieties.

ACKNOWLEDGMENTS
This work is supported by the H2020 programme No. 679766.
TomGEM; A holistic multi-actor approach towards the design of new tomato varieties and management practices to improve yield and quality in the face of climate change. The authors thank Mr Chris Gerrish for the technical assistant with fractionation experiments. IGA Technology Services facility for assistance in the utilisation of RNA-seq data and Dr Genny Enfissi for advice and input in the experimental approach.

CONFLICT OF INTEREST
The authors declare no conflict of interest.