Analyses of tomato fruit brightness mutants uncover both cutin-deficient and cutin-abundant mutants and a new hypomorphic allele of GDSL lipase

Cuticle is a protective layer synthesized by epidermal cells of the plants and consisting of cutin covered and filled by waxes. In tomato fruit, the thick cuticle embedding epidermal cells has crucial roles in the control of pathogens, water loss, cracking, postharvest shelf-life and brightness. To identify tomato mutants with modified cuticle composition and architecture and to further decipher the relationships between fruit brightness and cuticle in tomato, we screened an EMS (ethyl methanesulfonate) mutant collection in the miniature tomato cultivar Micro-Tom for mutants with altered fruit brightness. Our screen resulted in the isolation of 16 glossy and 8 dull mutants displaying changes in the amount and/or composition of wax and cutin, cuticle thickness and surface aspect of the fruit as characterized by optical and Environmental Scanning Electron microscopy. Main conclusions on the relationships between fruit brightness and cuticle features are that ( i ) screening for fruit brightness is an effective way to identify tomato cuticle mutants, ( ii ) fruit brightness is independent from wax load variations, ( iii ) glossy mutants show either reduced or increased cutin load, ( iv ) dull mutants display alterations in epidermal cell number and shape. Cuticle composition analyses further allowed the identification of groups of mutant displaying remarkable cuticle changes e.g. , mutants with increased dicarboxylic acids in cutin. Using genetic mapping of a strong cutin-deficient mutation, we next discovered a novel hypomorphic allele of GDSL lipase carrying a splice junction mutation, thus highlighting the potential of tomato brightness mutants for advancing our understanding of cuticle formation in plants.


INTRODUCTION
The epidermis of all aerial plant organs is covered with an extracellular layer, the cuticle, which is synthesized by the epidermal cells. Cuticle is localized on the outer face of primary cell walls and is largely composed of cutin embedded with polysaccharides, filled with intracuticular waxes and covered with a thin layer of epicuticular waxes (Nawrath, 2006).
Cutin is a polyester of glycerol, hydroxy and epoxy fatty acids; in most species, the main cutin monomers are C16 and C18 -hydroxy fatty acids (Pollard et al., 2008). Besides cutin, another lipid polyester named suberin typically contains in addition ,-dicarboxylic acids, hydroxycinnamic acids and fatty alcohols. Suberin forms a hydrophobic layer in cell walls of specific plant organs (e.g., roots and seeds) or is synthesized in response to stress (Pollard et al., 2008). Waxes are a mixture of very long chain fatty acids (VLCFA, C24-C34), and their derivatives such as alkanes,aldehydes, primary and secondary alcohols, ketones or esters and include occasionally of triterpenoids and phenylpropanoids (Kunst and Samuels, 2009).
In the recent years, availability of Arabidopsis genome sequence, high throughput gene expression analysis tools and mutant collections enabled deciphering the biosynthetic pathways and transport networks involved in cutin, suberin and wax biosynthesis (Pollard et al., 2008;Li-Beisson et al., 2009;Yeats and Rose, 2013). The synthesis of cutin monomer starts with the synthesis of long chain fatty acids in the plastids. Fatty acids are then transported to the cytoplasm where they undergo a series of modifications including the activation to coenzyme A thioesters by long-chain acyl-CoA synthetases (LACS), oxidation by cytochrome P450-dependent fatty acid oxidases (CYP) and esterification to glycerolbased acceptor by glycerol-3-phosphate acyl transferases (GPAT) to produce acyl-glycerols (Li-Beisson et al., 2009;Pollard et al., 2008). Though the sequential order of the reactions remains to be determined, the implication in cutin biosynthesis of several LACS, CYP86A and CYP77A and GPAT has been confirmed in Arabidopsis. Mechanisms of transport of the cutin monomers and their assembly into the cuticle remain largely unknown. The plasma membrane ATP-binding cassette (ABC) transporters have been implicated in the transport of both wax and cutin to the apoplast while lipid transfer proteins (LTP) very likely contribute to the transport of cutin monomers through the cell wall to the cutin layer (Yeats and Rose, analyses revealed that mutant fruit with altered brightness displayed wide variations in cutin load and composition and in epidermal patterning. Only few changes in wax composition and increases in wax load were observed. Fruit brightness modifications could not be attributed to a single cause but rather to combinations of various and sometimes opposite alterations of fruit cuticle. We further identified the mutation underlying a glossy cutin-deficient mutant by genetic mapping of the corresponding cuticle-associated traits and discovered a novel allele of the tomato GDSL lipase involved in cutin polymerization (Girard et al., 2012;Yeats et al., 2012). Thus our study highlights how the exploitation of artificially-induced genetic diversity and of genomic tools currently available in tomato can efficiently contribute to the study of cuticle biosynthesis and properties in plants.

Screening Micro-Tom EMS mutant collection for plants with altered fruit brightness
Analysis of cuticle formation and properties using reverse genetic approaches may be hampered by other modifications of plant and fruit physiology often induced in strong cuticle mutants. In this study, we chose to focus on mild tomato fruit cuticle mutants displaying no obvious phenotypic alterations such as dwarfism, wilting or organ fusion in plant or large cracks, multiple micro-cracks (russeting) and strong water loss in fruit. Towards this end, the criterion we used for selecting tomato mutants was fruit brightness, the variation of which results from milder modification of fruit surface. In addition, thin cuticle and glossy fruit appearance associated with adequate postharvest shelf-life and absence of fruit defects are among the desirable breeding traits for several types of fresh-market tomatoes.
The tomato mutant collection used was an EMS (ethyl methanesulfonate) mutant population generated in the miniature cultivar Micro-Tom by our group (Causse and Rothan 2007;Just et al., 2013). The collection comprises ca 3500 highly-mutagenized mutant families thoroughly phenotyped (12 plants per M2 families) for ca 150 plant and fruit phenotypic criteria ranged in categories and sub-categories. All the phenotypic data are compiled and stored into a dedicated web-searchable database called MMDB (MicroTom Mutant DataBase). Selection of fruit brightness mutants was done using requests centered on the fruit (category: fruit; sub-categories: color, epidermis, brightness i.e. fruit glossier or duller than the wild-type control). Strong fruit cuticle/color mutants such as the slcyp86a9 cuticle synthesis mutant recently identified from the same Micro-Tom mutant collection (Shi et al., 2013), were excluded from the analysis. As shown in Figure 1, the first query using these criteria resulted in the identification of 274 mutant families. Taking into account additional criteria such as the presence/absence of other gross plant and fruit phenotypic alterations and the previous confirmation (or not) of the observed fruit brightness phenotypes in independent cultures of the retained M2 families allowed the final selection of 40 mutant families.
For each selected family, 24 plants were grown in greenhouse. Visual evaluation of fruit brightness was done for each plant of each family on fruits at Red Ripe (RR) stage, in comparison with wild-type (WT) fruits (Supplemental Table S1). In the M2 families studied, which segregates for the EMS-induced mutations, a plant was considered as a fruit brightness mutant when all the fruits on that plant displayed the glossy or dull trait. Previous fruit brightness phenotypic annotation (dull or glossy fruit) was confirmed for 20 families (i.e., at least one plant in the family showing uniformly dull or glossy fruits) whereas 7 families displayed fruit brightness phenotypes opposite to those previously observed. In addition, two displayed both dull/glossy fruits on the same plants and 11 did not show any visual difference with the WT fruits. We finally selected the 20 confirmed mutant families plus 4 families displaying fruit brightness phenotypes opposite to those previously observed (i.e. one plant with dull or glossy fruits for each of the selected families) Among these 24 fruit brightness mutant plants, 16 were glossy mutants and 8 dull (Supplemental Fig. S1).

Cuticle is altered in fruit brightness mutants which display remarkable changes in cutin load and composition
In order to investigate whether variations in fruit brightness could be associated with quantitative and/or qualitative differences of cuticle, we analyzed wax and cutin load and composition of RR fruits (~45 days post-anthesis or DPA) collected on the 24 selected mutant lines.
Both the glossy and the dull mutants exhibited a wide range of variation in wax (identified compounds) load ( Fig. 2 and Table I). For up to 50% of the mutants, no significant differences in cuticular wax load were observed relative to wild type (Fig. 2). Remarkably, the glossy and dull mutants displaying wax load variation all showed an increase in wax load in comparison with WT. However, while the glossy mutants showed a gradual variation in wax load from ~6 µg/cm² (i.e., close to WT values) to 13,5 µg/cm², most of the dull mutants had similar wax loads (7,6-9 µg/cm²) with the exception of one (P6D6 mutant; 13,1 µg/cm²).
When taking into account the unidentified wax compounds, which represent 17% of the total wax in the wild type and up to 40% in the P4E2 mutant, similar variations were observed (data not shown). triterpenoids (  and  type amyrin; ~18%), isoalkanes (7-8%) and alcohols (~3-4%). While changes in alkane and in wax load showed the same trend in most mutants, several mutants showed remarkable modifications of wax composition. Amyrins increased to ~29 to 33% of total wax load in the P17F12 and P23F12 glossy mutants and to almost ~30% in the P6D6 dull mutant. Alcohols increased to ~7 to 12% in the P11H2 and P4E2 glossy mutants. In conclusion, most fruit brightness mutants displayed no or slight variations in wax load and glossy mutants could not be distinguished from dull mutants based only on single wax analysis.
While changes in cutin load were usually paralleled by similar changes in the various classes of cutin monomers, several glossy mutants displayed remarkable and specific patterns of accumulation of cutin monomers. Indeed, further exploration of the cutin composition data by Principal Component Analysis (PCA) and Hierarchical Clustering Analysis (HCA) revealed groups of mutants displaying similar cutin composition changes (Fig. 4). The first principal component (PC1) explaining 53% of the total variability, clearly separated three clusters on the positive side and one cluster on the negative side. The second principal component (PC2) explaining 23% of the total variability separated two clusters on the positive side and two clusters on the negative side. A bulk of 8 mutants, either dull or glossy, was not separated by the principal components PC1 and PC2 and were aggregated within the WT cluster.
Examination of PC1 and PC2 loadings (Supplemental Fig. S3) allowed the identification of cutin monomers responsible for cutin modifications in the various mutants. On the positive side of PC1 and PC2, the primary alcohols and 2-hydroxy fatty acids and the dicarboxylic acids discriminated the P17F12 glossy mutant and the P5E1/P26E8 cluster of glossy mutants. The P17F12 mutant, in which the total cutin load is increased by more than twofold, exhibited a 30-fold enhancement of 2-hydroxy fatty acids, which increased from trace amounts in the WT (6.6 µg/cm²) to ~16% of the total cutin monomers in the mutant (~201 µg/cm²). The dicarboxylic acid contents were strikingly similar between the P5E1 and P26E8 mutants and increased from ~32 µg/cm² in the WT to 134-139 µg/cm² in the mutants.
Another remarkable cluster groups the strong cutin-deficient mutants which are similarly affected for all cutin monomers (e.g. the P15C12 and the P23F12 mutants clustered on the negative side of PC1). On the positive side of PC1 and negative side of PC2, the poly hydroxy fatty acids, the -hydroxy fatty acids, the fatty acids and the cinnamic acid compounds were discriminating the P30A12/P18H8 cluster of glossy mutants. These mutants accumulate large amount of cutin monomers (X 1.6), like P17F12, but show a different composition, especially with respect to the 2-hydroxy fatty acids which amount is close to that of the WT (Table II). Last, a cluster discriminated on negative side of PC2 groups several dull and glossy mutants with moderate to large changes in cutin load.
Thus, though cutin is altered in most dull and glossy mutants, no obvious link could be made between fruit cutin load and/or composition and fruit brightness, as already observed for waxes. Moreover, glossy mutants can show either strong deficiencies (e.g., P15C12) or increases (e.g., P17F12) in cutin load (Fig. 3). To further explore the possible effect of wax and cutin co-variations on fruit brightness, we combined the biochemical (wax and cutin composition and load) and phenotypic data and analyzed them by HCA (Supplemental Fig.   S4 and S5). However, as for cutin alone, no obvious relationships could be found between cutin and wax variations and the dull or glossy aspect of the fruit.

Epidermal and Cuticle architecture affect fruit brightness
In order to get more insights into the relationships between fruit brightness and cuticle, we selected three groups of mutants according to cutin monomer and/or wax load of the fruits and further characterized them by optical and scanning electron microscopy. Among these were glossy mutants with high cutin load (P18H8 and P17F12) and low cutin load (P23F12 and P15C12) and dull mutants with either high cutin load (P16H5) or high wax load (P6D6) (Fig. 5A). As expected, glossy mutants with low cutin load exhibited very thin cuticles when compared to WT, with cuticle thickness reduced by ~49% for P23F12 and ~67% for P15C12 ( Fig. 5D and E). Surprisingly, cuticle thickness was very similar between WT fruits and glossy mutants with high cutin loads, despite the strong difference in cutin load between these mutants and WT (60 to 120% more cutin in P18H8 and P17F12 respectively). Examination of exocarp sections indicated that neither tissue structure nor epidermal cell size and shape were affected in these glossy mutants (data not shown). Dull mutants did not show any significant difference in cuticle thickness relative to WT but both high cutin and high wax dull mutants displayed distinctive morphological alterations of the cuticle-encased epidermal cells, which appeared less elongated and more conical-shaped than in WT and other mutants.
When examined under environmental scanning electron microscopy (ESEM), native fruit surface from glossy mutants with low cutin load mutants looked much smoother than that of WT fruits, which showed more irregular surface with small domes (Fig. 5B). At the opposite, fruit surface from dull mutants was rough and entirely covered with circular-shaped dome-like structures, in agreement with optical microscopy observations (Fig. 5D). These differences appeared even more clearly when epicuticular waxes were removed from cuticle by treatment with chloroform (Fig. 5C). Cutinized epidermal cell walls were clearly visible in the WT while the surface of high cutin load glossy mutants remained remarkably smooth with no (P18H8) or very few (P17F12) surface irregularities. In contrast, de-waxing low cutin load glossy mutants revealed very thin-walled epidermal cells underneath the cuticle proper, with no cutin deposit on top of these cells unlike WT. De-waxing did not change the surface aspect of the dull fruits, though the dome-like structures which correspond to epidermal cells ( Fig. 5D) were even more apparent. Their quantification clearly indicates that fruit epidermal cell number is increased by 66% to 122% in the dull mutants (Supplemental Table S3).
Together these observations suggest that both the cuticle and the development of the epidermis are altered in dull fruit mutants. In contrast, only cuticle is affected in glossy fruit mutants and increased fruit brightness can be provoked by both deficiency and increased accumulation of cutin.

The cutin-composition mutants P5E1 and P26E8 show constitutive alteration of suberin biosynthesis in the fruit
To investigate the mutations underlying the variations in fruit brightness, we focused on the glossy mutants since dull mutations are likely pleiotropic and affect epidermis. We further focused on cutin mutants because cutin load is possibly the major factor controlling fruit brightness in tomato (Girard et al., 2012;Nadakuduti et al., 2012;Yeats et al., 2012), and the possible effect of wax changes on glossiness have already been well described in Arabidopsis (Chen et al., 2003;Aharoni et al., 2004;Bourdenx et al., 2011). One way to identify the possible origin of the mutation is to combine the wealth of information now available on cuticle biosynthesis and regulation with the data on fruit surface chemistry. Two of the cutin-accumulation mutants that clustered together in PCA-HCA (P5E1 and P26E8 components of the P5E1 and P26E8 mutants (Supplemental Table S2) indicates that both display a 2.5 fold increase in C16:0 dicarboxylic acid and a 5 fold increase in C18:1 dicarboxylic acid when compared to WT. In most species, except Arabidopsis and other Brassicacea, this composition is considered as indicative of suberin (Pollard et al., 2008).
Remarkably, no russetting or any other visible mark of suberin accumulation had been observed in the fruit sample analyzed, which presented a uniform glossy appearance.
Following cuticle composition analysis, we therefore planted one of these mutants (P26E8) in order to observe its cuticle phenotype. Close examination of the fruits revealed the obvious accumulation of suberin-like material at the distal end of the fruit (Supplemental Fig. S6), a trait not observed in the fruit sample previously analyzed. This result reinforces the hypothesis that the biosynthesis of both cutin and suberin are altered in these mutants. Since growth period and position of the fruit on the plant apparently influence the extent of accumulation of this suberin-like material, this is likely under the control of the environmental conditions.

The P15C12 cuticle mutant displays constitutive cutin-deficiency in the fruit
Two other obvious targets are the P15C12 and P17F12 cutin mutants, which are respectively strong cutin-deficient (cutin load: 84.1 µg/cm²) and strong cutin-accumulating (cutin load: 1175.5 µg/cm²) mutants. Fruits from the P17F12 mutant display obligatory parthenocarpy and plants carrying the cuticle mutation never set seeded fruits, despite attempts of pollination with WT pollen. This mutant was therefore not studied further. In contrast, the P15C12 mutant, in which plant and fruit were not affected by the mutation except for cuticle alteration, was further characterized along fruit development for cutin load and water loss Alterations of cutin biosynthesis in the P15C12 mutant were also accompanied by large modifications of the cuticle permeability to water, though mutant wax load was 40% higher than that WT. The water loss kinetics of representative fruits from WT and P15C12 mutant ( Fig. 6B) clearly show the impaired ability of P15C12 detached fruits to avoid water loss, the mutant fruit retaining only 32.2% of its original weight at the end of the experiment versus 72.5% in the WT. Consistently, the tests of cuticle permeability performed on Mature Green (MG) fruit demonstrated that mutant fruit was very permeable to the toluidine blue (TB) dye, unlike WT fruit (Fig. 6C). Thus, besides fruit brightness, the mutation affecting fruit cutin load in the P15C12 mutant has dramatic consequences on the integrity and properties of its cuticle. Since these various phenotypic traits (fruit brightness, cutin width, water loss, permeability to TB) describe various aspects of the same mutation affecting cutin load, they were therefore used for characterizing tomato genotypes carrying the cuticle-deficiency allele found in P15C12.

Mapping of the P15C12 locus and identification of the gdsl2-b mutation through candidate gene approach
A F2 mapping population segregating for the cuticle-deficiency mutation found in P15C12 was generated through crossing the 'Micro-Tom' P15C12 homozygous mutant with a M82 dwarf mutant, previously selected amongst EMS mutants generated in the widely used Nevertheless, the major locus controlling each trait was located in a 4.84 Mb region on chromosome 11, between markers 11289_715 and 10722_814 (Fig. 7A). LOD score, R² percentage and effect of QTL are presented in Table III. We next screened the chromosomal region of interest for candidate genes with known implication in cutin synthesis or regulation. Among these were genes identified as specifically expressed in the outer epidermis of tomato by laser microdissection of various tomato pericarp tissues followed by RNAseq analysis (Matas et al., 2011). Two of them (Solyc11g007540 and Solyc11g006250; SGN: http://solgenomics.net/) were localized between the markers of interest on chromosome 11. The first presents a strong homology with an Arabidopsis gene (AT5GO4660.1) encoding a CYP77A4 cytochrome P450 oxidase catalyzing the epoxidation of free fatty acids (Sauveplane et al., 2009) Sequencing the SlGDSL2 (Solyc11g006250) gene from the WT and P15C12 plants revealed an A => T mutation disrupting the 3' splice site of intron 4 (Fig. 7 B). Since a mutated allele of SlGDSL2 was already described in tomato (Yeats et al., 2012), the P15C12 mutant is thereafter named gdsl2-b. As in other introns belonging to the major class of introns processed by the U2 spliceosome, the splice site sequences of intron 4 from SlGDSL2 gene fit the canonical GT-AG consensus borders, where the nearly invariant GT dinucleotide is at the 5' end and the AG dinucleotide at the 3' end of the intron. Point mutations in this 3' AG dinucleotide, which is essential to the definition of the 3' splice junction, may lead to the production of mRNA with unspliced intron 4. Actually, missplicing of intron 4 of SlGDSL2 leads to the accumulation of several species of SlGDSL2 mRNA in both the P15C12 mutant and its F1 hybrid (Fig. 7D). The larger mRNA effectively corresponds to unspliced mRNA, as confirmed by sequencing. In-frame reading of intron 4 produces 13 additional incorrect residues after exon 4 (Fig. 7C), thus leading to the production of a truncated protein in which the 64 residue C-terminal region is missing. Interestingly, a second mRNA is produced, which size is consistent with that of correctly spliced SlGDSL2 mRNA. However, close examination of the mutated sequence reveals the presence of a cryptic 3' splicing site in exon 5, 17 nucleotides downstream of the canonical 3' consensus splice sequence of intron

DISCUSSION
Considerable progresses in the knowledge of cuticle synthesis and regulation have been made in the recent years thanks to the availability of Arabidopsis mutants and genomic tools.
Main pathways for the synthesis of waxes and cuticle polymers have being deciphered (Pollard et al., 2008;Bernard and Joubès, 2013;) and increasing evidence on the mechanisms of transport of cuticular components and on the regulation of cuticle biosynthesis are now available (Yeats and Rose, 2013). Currently, one of the main challenges is to understand how cuticle properties are linked with cuticle composition and structure, how cuticle components interact with cell wall polymers, and how this will affect plant characteristics e.g. plant growth or resistance to biotic and abiotic stresses (Dominguez et al., 2011).
Though Arabidopsis remains the model of choice for plant functional genomics, this species is not well adapted for studying cuticle properties due to its very thin cuticle and some specificities (e.g. the high level of dicarboxylic acids in Arabidopsis cutin, unlike most other plants). In contrast, tomato fruit has a thick and easy-to-study cuticle synthesized along early fruit development (Yeats and Rose, 2013). Tomato has therefore long been used for studying cuticle biomechanics and permeability (Dominguez et al., 2011;Schreiber, 2010) and has recently emerged as a new model for functional genomics of cuticle formation in plants.
Because tomato is both a major crop species and a model for fleshy fruits, a wealth of information and genomic tools are now available for this species (Tomato Genome Consortium, 2012). In addition, several major agronomical traits in tomato and in other fleshy fruit species e.g., fruit growth, visual aspect, cracking, water loss, resistance to pathogens and postharvest shelf-life are highly dependent on fruit cuticle (Bargel and Neinhuis, 2005;Saladie et al., 2007;Matas et al., 2009;Dominguez et al., 2011;Parsons et al., 2012). An increasing number of studies highlight the possibilities offered by tomato for analyzing cuticle architecture, mechanical properties and permeability (Lopez-Casado et al., 2007;Saladie et al., 2007;Mintz-Oron et al., 2008;Buda et al., 2009;Isaacson et al., 2009;Wang et al., 2011) and for discovering genes contributing to cuticle synthesis and regulation (Hovav et al., 2007;Mintz-Oron et al., 2008;Nadakuduti et al., 2012;Girard et al., 2012;Yeats et al., 2012;Shi et al., 2013). Nevertheless, to further our understanding of the relationships between cuticle

Tomato EMS mutants for studying cuticle composition and properties
Collections of artificially-induced genetic diversity resulting from fast-neutron or Ethylmethane Sulfonate ( fruit surface defects such as micro-cracks (russetting), cracks, strong shriveling, peel browning or less severe alterations such as increased water loss and altered fruit color or brightness (data not shown). Because strong cuticle alterations are detrimental to fruit quality, we preferred to focus in a first step on the phenotypic alterations only responsible for variations in fruit brightness (glossy or dull fruits), which likely arise from more subtle changes in cuticle properties. Pleiotropic mutations were also excluded since in that case the origin of the cuticular defect is not always easy to trace e.g., in fruit developmental mutants (Czerednik et al., 2012). The disadvantage of using fruit brightness as a screen for detecting cuticle mutations is that this trait can be very sensitive to environmental conditions e.g. the growing season (data not shown). However, out of the 40 mutants examined (from the 274 originally found in the mutant database), more than 20 were further confirmed as being fruit brightness mutants. Among them, 8 displayed glossy/dull trait for both fruit and leaf (Supplemental table S1) indicating that fruit can be used as an attractive model for studying alterations of leaf cuticle.
Our study clearly shows that, though all selected brightness mutants displayed cuticle alterations, the glossy/dull fruit trait is not due to one single alteration. Recent studies of tomato mutants and transgenic lines established that increased fruit glossiness was associated with cutin deficiency (Isaacson et al., 2009;Girard et al., 2012;Nadakuduti et al., 2012;Shi et al., 2013). We indeed confirmed this relationship for several mutants (Fig. 4) but further showed that more complex cuticle architecture can be responsible for this trait. Most of the glossy mutants showed significant variations in either total cutin load or cutin composition (Table II) Buschaus and Jetter, 2011). Due to these characteristics, WT fruit is moderately glossy. In high cutin load mutant, a thick cutin layer covers all cell surface and thus likely levels most irregularities resulting in a smooth surface, which can be seen in both native and de-waxed fruit (Fig. 4). This likely increases the specular reflection, hence giving a glossy aspect to the fruit. In contrast, the de-waxed surface of low cutin load mutants displays a very irregular aspect due to the presence of thin to very thin cutin layer. The epicuticular wax film covering these small surface irregularities likely "polishes" the surface of the fruit, which therefore displays increased glossiness. At the opposite, dull mutants present highly irregular surface due to an increased number of small epidermal cells of different morphological aspect, as shown by ESEM (Fig. 4). The rugged aspect of fruit surface, seen for both native and dewaxed fruit, is probably responsible for increased diffuse light reflection hence giving to the fruit a matt aspect (Fig. 8). Higher wax load of some mutants may also contribute to increase light scattering.
No glossy tomato fruit mutants with reduced total wax load were observed in the present study. In Arabidopsis, stem glossiness is generally indicative of altered epicuticular wax crystallization and is either due to general wax load reduction or to alteration of specific wax compounds (Jenks et al., 2002). Similar observations have been made in maize and in rice (Islam et al., 2009;Jenks et al., 2002). In addition, no organ fusion was observed among all the glossy/dull fruit mutants studied, in contrast to the glossy mutants described in Arabidopsis or to the SlCER6 loss-of-function wax tomato mutant (Smirnova et al. 2013).
Though the participation of wax in the glossy aspect of the fruit cannot be excluded (several glossy mutants display significant increases in wax load), our results strengthen the conclusion that the glossy/dull fruit appearance in our tomato mutants is not due to wax deficiency. The most striking result is however that dull fruit aspect of tomato cuticle mutants

Cuticle composition analysis of EMS mutants as a tool for discovering genes and pathways
Insights into cuticle composition may give some hints on the regulations or biochemical pathways affected in the various cuticle mutants. Before undertaking allelism tests by crossing the various mutants, as done for the wax glaucous mutants in Sorghum (Peters et al., 2009), we first analyzed mutants for their cutin and wax composition (Table 2 and 3).
Indeed, several groups of mutants displayed remarkably similar cuticle characteristics suggesting the involvement of the mutated genes in related biochemical pathways (e.g. in the cutin-deficient mutants P15C12 and P23F12) or in their regulation (e.g. in the wax-and cutin-rich mutants P18H8 and P30A12). These groups may first help discovering new genes and functions related to poorly known aspects of cuticle formation. In addition, as discussed below for the gdsl lipase gene, these mutants may help uncovering new alleles of known candidate genes such as those involved in cutin formation or in the coordinated regulation of wax and cutin biosynthesis (Yeats and Rose, 2013). Finally, these groups may also unravel genes involved in suberin biosynthesis. In the P5E1 and P26E8 mutants, the accumulation in cutin fraction of dicarboxylic acids, which are considered as markers for suberin, and the

Fruit brightness assessment
Fruit brightness of the mutant lines was visually estimated at the Mature Green (MG) and

Water loss and cuticle permeability measurements
For water loss measurements, one RR fruit was harvested from each plant of the F 2 mapping population, including the parents. Fruit sealing wax was then applied on stalks and fruits were stored at room temperature. Fruit fresh weight was recorded at T0 and each week, until 6 weeks. Water loss was calculated as a percentage of weight loss. For measurements of cuticle permeability to stain, one MG fruit was collected from each plant of the F 2 mapping population and from the parents, and dropped in 1% Toluidine Blue solution during 6-h as described in Tanaka et al., (2004). For the cutin monomer analysis, a 1 cm diameter disk was cut off from a RR fruit epidermis, carefully scratched with a scalpel blade in order to remove exocarp cells and incubated 30 min in isopropanol at 85°C. The disk was then delipidated by successive deeps in chloroform / methanol (C/M), at different ratios: C/M (2:1) for 24h, C/M (1:1) for 24h, C/M (1:2) for 24h and 100% methanol for 24h. The delipidated epidermis disk was dried 48h under pulsed air and 48h in a desiccator. Cutin was then depolymerized, analyzed and quantified as already described (Domergue et al., 2010).

Light microscopy
For cuticle thickness measurements, fruit exocarp (including cuticle) was obtained from 2 independent RR fruits of WT and mutant lines. Samples were fixed and embedded in paraffin as previously described (Mounet et al., 2009). Eight micrometer slides of exocarp were stained using saturated and filtered Sudan Red solution in ethanol. Mean cutin thickness was assessed from 60 measurements.
For analyzing the thickness of the cutin layer between two adjacent fruit epidermal cells (thereafter called cutin width) in the mapping population, a 10 mm² square tissue sample was peeled off from one RR fruit harvested from each plant of the F 2 mapping population, including the parents. As much epidermal cells as possible were then removed from the peel by scratching the internal surface with a scalpel blade. The resulting cuticle-enriched fragment was immerged in water, placed flatten on a glass slide and observed under optical microscope at x20 magnification. Cutin width was then determined by measuring thickness of the cutinized cell wall between two adjacent epidermal cells (Supplemental Fig. 6). Mean cutin width was assessed from 12 measurements.

Environmental Scanning Electron Microscopy (ESEM)
For the observation of wax on the surface of the fruit, 5 mm side cubes of exocarp including cuticle were excised from the equatorial part of RR fruits and placed directly into the observation chamber of a scanning electronic microscope FEI Quanta 200, in environmental mode. For cutin observation, cubes were beforehand submerged 30 s into chloroform, with gentle agitation. Observations were performed with a plate at 4°C, a 6 Tor pressure in the chamber and a 4.7 V voltage applied to the filament. Pictures were captured with x600 and x3000 magnifications.

PCR and Quantitative Reverse Transcription PCR analysis of SlGDSL2
Genomic DNA was extracted from 100 mg fresh weight (