Use of metabolomics data analysis to identify fruit quality markers enhanced by the application of an aminopolysaccharide

Chitosan is a biostimulator that has a great effect either on plant physiology, productivity, or fruit quality. However, the metabolic mechanism regulated by chitosan still remains unknown. Untargeted metabolomics analysis, using LC-MS/MS mass spectrometry, was used to investigate fruit quality markers. Thus, this study was focused on the identification of untargeted metabolites of tomato fruits produced under the application of five doses of chitosan at different concentrations (0, 0.25, 0.50, 0.75, and 1 mg ml−1) that was extracted from Parapenaeus longirostris shrimp shells. The identification was carried out using two ion modes (ESI−/ESI+), a web application “Metfamily” to analyze signals, and reference libraries. The analysis of data using partial least squares discriminant analysis (PLS-DA) and hierarchical cluster analysis (HCA) showed that chitosan application, especially 0.75 mg ml−1, had a clear and remarkable effect regarding the number of metabolite families identified in both ion modes. This treatment has increased the relative abundance of many metabolites that belong to anthocyanins decorated with sugars, terpenoids, phenylpropanoids, acylsugars, glucosinolates, folates, galactolipids, fatty acids, and phospholipids. Thus, these results showed that chitosan application increased the quality of tomato fruits due to its involvement in the regulation of many metabolic pathways that might be responsible for enhancing the nutritional characteristics as well as the defense of fruits.


Introduction
Agrosystems can change over time due to the increase in drought, salinity, heat, metal toxicity, and mineral deciency or by infection with pathogens. 1,2 The use of synthetic molecules to overcome some of these issues is not always the best solution because of its feature to appear as a double edged-sword. As an example, pesticides from the 1940s are considered the best phytochemicals due to their ability to enhance either the quantity and the quality of important crops or the protection of plants against diseases, pests, fungus, and weeds. 3 Despite their benets, however, their excessive use, especially by exceeding the dose line to have more productivity, has developed into another issue as the deterioration of human health and the environment.
Recently, many researchers have focused on the valorization of shellsh waste to produce a natural fertilizer, such as chitosan, which therefore reduces the pollution generated by this waste, replaces pesticides, and increases plant growth and productivity. 4 Chitosan, an aminopolysacharide, is a deacetylated form of chitin which is the second major polysaccharide in nature aer cellulose. 5 The application of chitosan in different elds has been raised by the employment of its diverse biological properties, such as non-toxicity and biodegradability. In agriculture, this polymer has been demonstrated to be a good elicitor due to its ability to stimulate the defense mechanisms and to protect the plant against pathogens. 6 Among the mechanisms underlying its effect, there is a faster accumulation of Ca 2+ in plant cells which then stimulates the stomata closure and therefore stops and limits the invasion of microorganisms inside the plant. 7 It could also prime the plant by either the activation of defense hormones, such as jasmonic acid, or the induction of pathogenesis related-genes, such as chitinase and b-glucanase. 8 Besides strengthening plant defense, the stimulation of defense metabolites as well as the limitation of the assimilation of CO 2 because of stomata closure will adversely affect productivity. In contrast, it has been demonstrated that chitosan is a good biostimulator for plant growth. It could increase yield and improve the quality of fruits and vegetables of many species by enhancing the production of secondary metabolites, such as phenolic and terpenoid compounds, vitamin C, and lycopene. 8 Therefore, it seems that chitosan prepares plants to be ready and adapt in any steady and urgent circumstance with a positive effect on plant growth and fruits quality.
The effect of foliar application of chitosan during seedling and fruit development on the production of secondary metabolites in fruits has not been yet studied. Therefore, a deep analysis of fruit metabolome is required to understand chitosan effects and to know, how to enhance either the esthetic or the taste quality of fruits.
Nowadays, many analytical techniques, such as HPLC/MS and NMR, were used to describe the chemical composition in varieties, to detect adulteration, and to discover new metabolites. 9 Untargeted metabolomics, using liquid chromatography with mass spectrometry (LC MS/MS), is among the efficient analytical techniques, which may help to detect unforeseen changes in the food metabolome. 10 This technique has been employed to compare volatile metabolites of 94 ripe tomato genotypes and to investigate the mechanisms and metabolic regulation that inuence fruit quality. 11,12 Therefore, we used LC-MS in combination with bioinformatics tools as an approach to perform non-targeted metabolite analyses of fruits from plants treated with different concentrations of chitosan. Resulting differences in the relative abundance of several metabolites showed the effects of chitosan application on tomato fruit metabolome.

Growing conditions and chitosan application
Tomato seeds (Solanum lycopersicum cv. Campbell33) were germinated in a commercial peat substrate at 28 C. Aer 10 days, the seedlings were transplanted into plastic pots (4 kg) that contained a mixture of sand and peat (2 : 1) and then placed in a greenhouse with 24 C temperature, 330 mmol m À2 s À1 photosynthetic photon ux density, and 69% relative humidity. At the four-leaf stage, uniform plants were chosen and divided into ve groups which represent chitosan treatments at different concentrations (0, 0.25, 0.50, 0.75, and 1 mg ml À1 ) Then, plants were sprayed once every two weeks for three months either with chitosan or distilled water. Three replicates of each treatment were applied. Fully ripened fruits were harvested, cut into pieces, immediately frozen, and ground in liquid nitrogen followed by lyophilization.

Metabolite extraction
Extraction of metabolites was performed using 25 mg of freezedried fruit material. In cryotube (1.6 ml; Precellys Steel Kit 2.8 mm, Peqlab Biotechnologie GmbH, Erlangen, Germany), material was mixed in a bead mill (5.0 m s À1 ; 20 s; FastPrep24, MP Biomedicals LLC, Santa Ana, California) with 900 ml of a cold mixture of dichloromethane:ethanol (2 : 1, À80 C) and 200 ml of trifuoroacetic acid (TFA, pH 1) for three times. The homogenized samples were centrifuged at 12 623 g for 3 min at 4 C. The upper phase of each sample (z200 ml, aqueous phase) was removed, while the rest was mixed again with 50 ml of TFA. Aer elimination of the aqueous phase of the second extraction, 600 ml of the lower phase (organic phase) was collected and transferred to a new tube. Then, 500 ml of tetrahydrofuran was added to the rest. Aer a fast mixture and centrifugation, the supernatant was collected and combined with the previous organic extract and dried in a nitrogen stream.

Analysis of untargeted metabolites using LC MS/MS
Dried extracts were resuspended in 180 ml 80% MeOH and centrifuged. Per sample 5 ml supernatant was injected for UPLC analysis. The separation of extracted metabolites was carried out using a Waters ACQUITY UHPLC system, equipped with a Nucleoshell RP18 column (2.1 Â 150 mm, particle size 2.1 mm, Macherey and Nagel, GmbH, Düren, Germany), a binary solvent manager, and an ACQUITY sample manager (Waters GmbH, Eschborn, Germany). The mobile phase A consisted of 0.3 mmol l À1 of ammonium formate (NH 4 HCO 2 , pH 3.5 with formic acid), while the mobile phase B was acetonitrile. A gradient elution was as follow: 2 min, 5% B; 2 to 19 min, 95% B; 19 to 21 min, 95% B; 21.01 to 24 min, 5% B. The ow rate was 400 ml min À1 and the temperature was 40 C. The metabolites were ionized by electrospray ionization in positive and negative mode. Mass spectrometric analysis of small molecules was performed by MS-TOF-SWATH-MS/MS (TripleToF 5600, both AB Sciex GmbH, Darmstadt, Germany) operating in negative or positive ion mode and controlled by Analyst 1.6 TF soware (AB Sciex GmbH, Darmstadt, Germany) (Fig. 1). The source operation parameters were as the following: ion spray voltage, À4500 V/+5500 V; nebulizing gas, 60 psi; source temperature, 600 C (600 TripleToF); drying gas, 70 psi; curtain gas, 35 psi. TripleToF instrument tuning and internal mass calibration were performed every 5 samples with the calibrant delivery system applying APCI negative or positive tuning solution, respectively (AB Sciex GmbH, Darmstadt, Germany).
TripleToF data acquisition was performed in MS1-ToF mode and MS2-SWATH mode. For MS1 measurements, ToF masses were scanned between 65 and 1250 Dalton with an accumulation time of 50 ms and a collision energy of 10 V (À10 V). MS2-SWATH-experiments were divided into 26 Dalton segments of 20 ms accumulation time each. Together the SWATH experiments covered the entire mass range from 65 to 1250 Dalton in 48 separate CID scan experiments, which allowed a cycle time of 1.1 s. For all MS/MS scans a declustering potential of 35 (or À35 V) was applied. Collision energies for all SWATH-MS/MS were set to 35 V (À35 V) and a collision energy spread of AE25 V, maximum sensitivity scanning, and elsewise default settings.

Annotation of metabolites
Prior metabolites identication and statistical analysis, raw data pre-processing were done using MS-Dial (version 4.0) (Fig. 1). Then, 2139 precursor ions detected in negative and positive ionization modes were ltered using two parameters: MS1 abundance threshold of 5000 counts and log2-fold change (LFC) of 0.5.
The identication and the annotation of metabolites were performed as described in Fig. 1 and 2. Briey, to annotate metabolites, the monoisotopic mass of the precursor ions (MS1; [M + H] + , [M + NH 4 ] + , [M À H] À adducts) was ltered from the publicly available database (i.e., Metacyc). If this mass matched a known metabolite, then a putative identication would be possible. Then, the identication of MS1 ions was conrmed if the mass of fragment ions detected by TOF MS was similar to those found either in reference spectra (MassBank and Pub-Chem) or by ChemDraw soware. Metabolite annotation was performed using a "MetFamily" web application, which is available free of charge at http://msbi.ipb-halle.de/MetFamily/. This website provides a dynamic link between MS1 data (m/z, retention time, and relative abundance) and its corresponding MS/MS spectra.
Moreover, in this study, 232 metabolites were identied (Tables S1 and S2 †) and 16 of them were selected because of their importance in human diet as well as their high level in fruits from non treated plants or treated plants with chitosan, especially 0.75 mg ml À1 .

Statistical analysis
Graphical representations, such as partial least squares regression in the discriminant analysis (PLS-DA) and hierarchical cluster analysis (HCA), were created to visualize the correlation between treatments and metabolites extracted. PLS-DA and HCA were performed using MetFamily web application (Fig. 1). Moreover, the analysis of a signicant difference in metabolite abundance was done by CoStat 6.400 (CoHort soware), using Duncan's test at p #0.05.

Foliar spray with chitosan changes metabolite prole of tomato fruits
Tomato fruits from plants treated with chitosan treatments at different concentrations (0, 0.25, 0.50, 0.75, and 1 mg ml À1 ) were used for extraction of metabolites. The extracts were analyzed by electrospray ionization mass spectrometry (ESI-MS/ MS) in negative and positive mode and revealed 713 and 1426 metabolite features per sample, respectively.
3.1.1. Identication of metabolite families. To identify fruit metabolites with altered levels due to treatment with chitosan as well as metabolic pathways that were altered by the application of chitosan, the metabolomics data set for both ionization modes (positive and negative) were analyzed. Due to a large number of data, we used multidimensional statistical analysis to simplify the information obtained. In addition, graphic representations using Partial Least Squares Discriminant Analysis (PLS-DA) and Hierarchical Cluster Analysis (HCA), were generated to show similarities and dissimilarities between fruit groups (Ch0, Ch0.25, Ch0.50, Ch0.75, and Ch1).
Regarding negative ion mode data, PLS-DA score plot showed that samples of each group were clustered together and separated from others (Fig. 3A). The total variation between those groups was equal to 42.8% (PLS factor 1 (t 1 ): 24.3% + PLS factor 2 (t 2 ): 18.5%). Additionally, fruits treated with Ch0.75 (group 4) were so strongly different from those of control (group 1). However, fruits treated with other concentrations of chitosan were clustered closely to non-treated fruits.
To clarify the reason for discrimination between samples, the loadings plot was created (Fig. 3B). This plot revealed that the metabolites were frequently localized on the right side which therefore conrms the positive correlation between these metabolites and samples treated with Ch0.75.
Concerning positive ion mode data, the score plot (Fig. 3C) showed that groups 4 and 5 (1 mg ml À1 chitosan) were closed to each other on both axis (PLS1 (21.3%) and PLS2 (17%)), which therefore indicates a high correlation between them. However, samples of non-treated fruits (group 1) were clustered far from  those treated with chitosan (group 2, 3, 4, and 5), indicating a strong negative correlation between treated and non-treated fruits. Loadings plot showed also that metabolites detected in positive ion mode are less scattered, in comparison to those detected in negative ion mode ( Fig. 3B and D).
To investigate the difference in the metabolome of treated (0.75 mg ml À1 chitosan) and non-treated fruits, HCA plots were created ( Fig. 4 and 5). A rst ler setting (average MS1 abundance is equal to 5000) was performed to reduce the number of ions analyzed to 232 and 560, in negative and positive ion mode, respectively. The annotation of negative ions using HCA spectra, showed that samples were composed of several metabolites belonging to eleven main classes (i.e., benzenesulfonates, benzoates, desulfoglucosinolates, fatty acids, avonoids, oligosaccharides, organosulfates, organosulfur compounds, phenylpropanoids, phospholipids, and terpenoids) (data not shown). Then, a second lter setting (fold change is equal to 1.41) was used to dene metabolites that their abundances were decreased ( Fig. 4A and 5A) or increased ( Fig. 4B and 5B) in fruits treated with Ch0.75 in comparison to non-treated fruits. Concerning non-treated fruits, the number of metabolites was decreased to achieve 49 and two metabolite families (desulfoglucosinolates and oligosaccharides) disappeared (Fig. 4A). However, Fig. 4B revealed the abundance of Moreover, a few parent classes (8 classes) were identied in positive ion mode ( Fig. 5A and B). Five of these classes were most common in both ionization modes; however, the remaining three classes (i.e., acylsugars, galactolipids, and pteroates) were new and abundant in group 4 than group 1.
3.1.2. Identication of feature. The annotation of ions presented in Tables S1 and S2 † was done based on the similarity between the monoisotopic mass of precursor ion detected by mass spectrometry and those found in the database (theoretical mass) as well as their fragment ions. Fig. 6 lists some product ions used for metabolomics data analysis.
For instance, some avonoids that have a positive charge were detected in negative ion mode, as anthocyanidins linked sugar (peak 22 and 23), pelargonidin (or luteolindin) (peak 113), cyanidin (peak 114), and peonidin (peak 115). The masse fragment ion found at m/z 271.0600 (C 15  ) were attributed to cyanidin and peonidin ions, respectively. These ions were generated aer the deprotonation of a hydroxyl group which then switch one carbon double bond to a single bond (Fig. 6).
Additionally, different molecules that have a negative change were identied, as rutin (peak 25). The annotation of this peak was done based on the similarity between the monoisotopic mass of precursor ion detected (m/z 610.1572) and of theoretical mass (m/z 610.1534; C 27 H 30 O 16 ) as well as the presence of deprotonated fragment ion of quercetin (m/z 301.0327; C 15 H 9 O 7 À ) (Fig. 6).

Action mode of chitosan
To evaluate the effect of foliar application of chitosan on tomato fruit quality, sixteen metabolites were selected based on their major abundance either in fruits from treated or non-treated plants (Fig. 7).

Discussion
In this study, we addressed the question, whether foliar spray with chitosan affects the quality of tomato fruits. Performing a non-targeted metabolite proling approach on mature tomato fruits from plants treated or not with different concentrations of chitosan led to the detection of many signals by LC-MS. Bioinformatics analyses revealed differences in the metabolome of fruits from chitosan-treated plants in comparison to those from non-treated plants being related to some specic metabolites, which might change the aesthetic of fruits (color and avor), increase their defense, and improve their nutritional values.

Chitosan impacts on colors and avors of tomato
Anthocyanins are classied as important avonoids, derived from phenylpropanoid pathway. The majority of anthocyanins produced by plants provide the dark color of purple, blue, and red. 14 In our study, chitosan application to tomato fruits increased the level of cyanidin-3-O-b-D-glucoside (CG) and pelargonidin-3-O-b-D-glucoside (PG) and decreased the level of phlorizin, pelargonidin, cyanidin, and peonidin (Tables S1 and S2 †). The increase in the content of anthocyanidins decorated with sugars could be due to the activation of UDP-glucose:-avonoid-3-O-glycosyltransferase enzyme (UFGT). 14,15 It is known that accumulation of anthocyanins can be increased in response to abiotic stress, but also treatment of strawberry fruits with chitosan led to enhanced levels of two anthocyanins. 16 Usually, anthocyanins are less available in the skin of some fruits and vegetables (i.e., grape, eggplant, blueberry, potato, purple potato, and red cherry), and not present in other edible products that are consumed widely, as tomato. 17 Due to their great role to overcome a range of chronic diseases, several attempts have been done to raise anthocyanins levels in fruits. Despite the applying of transgenic approaches to improve the anthocyanin composition in tomatoes, 18 the biochemical and genetic complexity of the fruits made gene selection for improving anthocyanins levels extremely difficult. Therefore, treatment of plants with chitosan might provide a simple way to improve anthocyanins levels in tomato.
In addition, metabolomics data analysis revealed that chitosan application to tomato fruits increased the level of trans-3,4-didehydrolycopene and b-apo-8 0 -carotenal. These two compounds not only give orange pigments, but also can reduce the risk of cancer and heart diseases. 19 Recent studies reported that chitosan elicited lycopene and carotenoid biosynthesis. 8,20 To our knowledge, the accumulation of carotenoid derivatives (trans-3,4-didehydrolycopene and b-apo-8 0 -carotenal) appears for the rst time in tomato. Although chitosan mechanisms, involved in the biosynthesis of these metabolites, are unclear, we made a hypothetical biochemical pathway in Fig. 8.
Besides the colors, avors are among the characteristics that have an impact on the quality of fruits. To date, over 400 avors have been discovered in mature tomato fruits, 21 but with relatively low levels. 22 The reasons of this minority can be attributed to either a genetic aspect or biochemical changes that taking place during post-harvest and handling practices. 22 However, metabolomics analysis of our data showed the presence of a molecule that is very close to 1-(3,4-dihydroxyphenyl)-5hydroxy-3-decanone, a particular precursor of ginger avor (Fig. 8). Chitosan effect on tomato fruits avor has not yet been documented. But, the enhancement of the level of this avor could be explained by the fact that chitosan application to fruits could make a barrier allowing to keep this aroma longer. 23

Chitosan induces the biosynthesis of defense metabolites
Chitosan is known to be an elicitor that stimulates the natural defense mechanisms of plants. 6 The analysis of metabolomics proles showed that the relative abundance of sucrose 2isovaleroyl-3-isodecanoyl-4-isobutanoate, a class of acylsugar, was increased in all fruits of chitosan-treated plants. Acylsugars are sticky molecules synthesized and stored in the glandular trichomes type IV of tomato leaves and fruits to enhance plant resistance against herbivores and micro-organisms infection. 24 In Artemisia annua L., Kjaer et al. (2012) 25 showed that foliar application of chitosan increased the leaf trichome density. Therefore, we suppose that its application to leaves could stimulate the biosynthesis of acylsucrose even in fruits. Besides acylsugar, other defense compounds identied were gallotanins. The defense mechanisms of these compounds are related to their ability to form complexes with proteins and lipids. 26 However, as it shows in Fig. 7, chitosan application had a negative impact on the level of 3-O-digalloyl-1,2,4,6-tetra-O-b-D-galloylglucose. The deactivation of gallotanins biosynthesis might be due to the involvement of sugars and fatty acids in biosynthesis pathways of avonoids, acylsugars, and lipids (Fig. 8).

Improvement of the nutritional quality of tomato under chitosan application through the increase in the level of bioactive molecules decient in fruits
Our study revealed that chitosan application increased the level of sulfur-containing glucosides: glucosinolates (GLs) (i.e., 4sinapoyloxybutylglucosinolate). Gls are one of the most abundant groups of phytochemicals in cruciferous plants, as broccoli (Brassica oleracea var. italica), cabbage (B. oleracea var. capitata f. alba), cauliower (B. oleracea var. botrytis), rapeseed (Brassica napus), mustard (Brassica nigra), and horse-radish (Armoracia rusticana); however, rare groups of metabolites in tomato. 27 Gls can break down into different products (i.e., isothiocyanates, thiocyanates, oxazolidinethione, and epithionitriles) that are responsible for plant taste and odor. 28 Meanwhile, other study has shown that the conversion of GLs to isothiocyanates in the small intestine and colon may help the body to detoxify the dangerous product and to treat cancer. 29 Changes in environmental conditions (i.e., temperature, heavy metal stress, and herbivore infections) were found to increase the contents of Gls. 30 Our results were, however, incongruent with the fact that the application of 100 mg of elicitor (chitin or chitosan) to brassica seedlings did not change the content of glucosinolates. 31 Additionally, folates are essential vitamins for human health due to their involvement in many biochemical processes, such as the carbon metabolism and the DNA biosynthesis. 32 Although tomatoes are rich in bioactive molecules, they contain small amounts of folate. The increase in the consumption of food decient in this molecule may cause several human disorders as cancer. 33 Our results showed that there is an increase in the relative abundance of (6S)-5-formyltetrahydrofolate tri-L-glutamate in fruits from tomato plants treated with medium (0.50 mg ml À1 ) and high (0.75 mg ml À1 and 1 mg ml À1 ) doses of chitosan. Such an accumulation of folate under chitosan application could be attributed to the stimulation of the biosynthesis of chorismate, purine, and glutamate 34 (Fig. 8). Therefore, the increase in native folate in tomato fruits could be considered as an alternative approach to avoid the fortication of fruits with synthetic folate (i.e., pteroylmonoglutamyl folate). 35

Chitosan could delay the deterioration of fruits
The metabolomics data analysis revealed the presence of two classes of galactolipids (monogalactosyldiglyceride (MGDG) and digalactosyldiglyceride (DGDG)). These compounds are the major lipid component in photosynthetic tissues (chloroplasts) of higher plants. 36 Moreover, Whitaker (1986) showed that the mature-green tomato contained almost 70% of galactolipids, in comparison to leaf chloroplasts. 37 Other studies have reported that ripening of fruit did not alter the content of simple compounds (i.e., fatty acids and triacylglycerols); in contrast, it affected the level of large molecules (i.e., phospholipids and galactolipids) that were dropped. 38,39 Thus, our results might suggest that chitosan application could delay the fruit soening process which is usually caused by the degradation of phospholipids and triacylglcerols, and therefore could prevent fruit spoilage. 40

Conclusions
The analysis of metabolomics data is a successful strategy that helped us to specify which features of tomato fruits were changed by the application of chitosan to leaves. The aesthetic, the nutritional, and the defense properties of tomatoes were promoted by foliar application of chitosan to tomato plants. With these data, our understanding of the effects of this polymer on tomato plants has been increased. However, additional analyses as the quantication of metabolites related to those properties will be useful to prove these ndings. On the other hand, analyzing metabolomics data enabled us to identify some specic ion markers that we can used as libraries to examine food ingredients.

Author contributions
F. E. A carried out the experiment, worked in data curation, and analyzed data, and wrote the original dra. All authors contributed to the conception and design of the study, the analysis and discussion of the results, and the revision of the manuscript. All authors have read and agreed to the published version of the manuscript.