Placenta, Pericarp, and Seeds of Tabasco Chili Pepper Fruits Show a Contrasting Diversity of Bioactive Metabolites

Chili pepper (Capsicum spp.) is one of the most important horticultural crops worldwide, and its unique organoleptic properties and health benefits have been established for centuries. However, there is little knowledge about how metabolites are distributed throughout fruit parts. This work focuses on the use of liquid chromatography coupled with high resolution mass spectrometry (UHPLC-ESI-HRMS) to estimate the global metabolite profiles of the pericarp, placenta, and seeds of Tabasco pepper fruits (Capsicum frutescens L.) at the red mature stage of ripening. Our main results putatively identified 60 differential compounds between these tissues and seeds. Firstly, we found that pericarp has a higher content of glycosides, showing on average a fold change of 5 and a fold change of 14 for terpenoids when compared with other parts of the fruit. While placenta was the richest tissue in capsaicinoid-related compounds, alkaloids, and tocopherols, with a 35, 3, and 7 fold change, respectively. However, the seeds were richer in fatty acids and saponins with fold changes of 86 and 224, respectively. Therefore, our study demonstrates that a non-targeted metabolomic approach may help to improve our understanding of unexplored areas of plant metabolism and also may be the starting point for a detailed analysis in complex plant parts, such as fruits.


Introduction
Chili pepper (Capsicum spp.) is one of the most important crops worldwide. It is used as a main ingredient for many dishes in different cultures, such as Asian, Latin-American, and Mediterranean cultures, due to its organoleptic properties [1]. There are 40 accepted chili species but only five are considered domesticated: C. annuum, C. chinense Jacq, C. frutescens, C. baccatum, and C. pubescens [2]. In 2017, 36 million tons of chili peppers were produced globally, with Mexico being the second largest producer [3]. Wild pepper populations of Tabasco pepper (C. frutescens L.) and C. annuum can be found in some states of Mexico, increasing the relevance for collecting and characterizing these species as resources for the breeding of cultivated peppers [4,5]. Previous studies have extensively described metabolite diversity in C. annuum [6,7] but very little is known on C. chinense and C. frutescens. Most of these studies have undertaken a targeted approach, where the main focus has been to quantitate for specific metabolites, such as capsiate, dihydrocapsiate, capsaicin, dihydrocapsaicin, carotenoids, fatty acids, and amino acids [8,9]. To date, there are no studies with a comprehensive global profiling of tissue specific in C. frutescens and C. chinense. that contains 60 putative identified metabolites, which include alkaloids and terpenoids that are unreported in Capsicum, with significant differences and relative abundances between three sections (pericarp, placenta, and seed) of Tabasco pepper fruit at the mature red stage, using an UHPLC-QTOF-HRMS platform, combined with the use of Progenesis QI for small molecules, as a tool for the pre-identification of unknown metabolites. Our results underline the global metabolic differences in complexity, mainly based on the secondary metabolism of these fruit parts.

Non-Targeted Metabolomic Analysis
Our data for two tissues and the seed of Tabasco chili fruit comprised a total of 87 files or chromatograms per ionization mode (positive and negative polarities) as shown in Figure 1 and were uploaded into Progenesis QI. The dataset was first aligned (retention time): Each chromatogram was aligned against each other and automatically compared to a reference profile selected by the software that contained the highest number of features (potential compounds). Then, peak picking was performed by default parameters. A total of 1980 features were detected in the aqueous phase and 1481 were detected in the diethyl ether extract, for both ionization modes. Figure 1 shows a typical chromatographic profile in positive and negative ionization mode of placenta tissue, displaying some representative compounds of chili pepper fruit. (pericarp, placenta, and seed) of Tabasco pepper fruit at the mature red stage, using an UHPLC-QTOF-HRMS platform, combined with the use of Progenesis QI for small molecules, as a tool for the pre-identification of unknown metabolites. Our results underline the global metabolic differences in complexity, mainly based on the secondary metabolism of these fruit parts.

Non-Targeted Metabolomic Analysis
Our data for two tissues and the seed of Tabasco chili fruit comprised a total of 87 files or chromatograms per ionization mode (positive and negative polarities) as shown in Figure 1 and were uploaded into Progenesis QI. The dataset was first aligned (retention time): Each chromatogram was aligned against each other and automatically compared to a reference profile selected by the software that contained the highest number of features (potential compounds). Then, peak picking was performed by default parameters. A total of 1980 features were detected in the aqueous phase and 1481 were detected in the diethyl ether extract, for both ionization modes. Figure 1 shows a typical chromatographic profile in positive and negative ionization mode of placenta tissue, displaying some representative compounds of chili pepper fruit.  (a) Base peak intensity chromatographic profile of placenta tissue from chili pepper fruit on a Charged Surface Hybrid (CSH) C18 column obtained with Electrospray Ionization (ESI) positive mode on a mass range from 100 to 1500. 1. 2-[(1H-Indol-3-ylacetyl) amino]-4-methylpentanoate; 2.
Distribution of the features between parts was compared using principal component analysis (PCA) of loading and score plots describing a significant grouping by part for both extraction phases ( Figure 2). In addition, quality controls (QC samples) were also considered, as shown in Figure 2. The QC samples cluster tightly in comparison to the total variance in the projection, suggesting a dataset deemed to be of high quality. Tables 1 and 2 describe loadings that mostly contribute to principal components for each extraction. Putative organic compounds catalogued as saponins (SPNS), such as Tuberoside J, Matesaponin 5, and Asparagoside B significantly contributed to component 1 in the aqueous phase, while other important features for component 2 belong to flavonoid class (FLV) and SPNS compounds. Furthermore, both components in the organic phase were mainly composed of glycerolipids (GL) and terpenoids (TER), as well as a putative carotenoid, (5cis,5 cis,9cis,11 cis)-1,2,7,7 ,8,8 -Hexahydro-1,2-epoxy-ψ, ψ-carotene were important for contribution of component 1 in organic phase. Distribution of the features between parts was compared using principal component analysis (PCA) of loading and score plots describing a significant grouping by part for both extraction phases ( Figure 2). In addition, quality controls (QC samples) were also considered, as shown in Figure 2. The QC samples cluster tightly in comparison to the total variance in the projection, suggesting a dataset deemed to be of high quality. Tables 1 and 2 describe loadings that mostly contribute to principal components for each extraction. Putative organic compounds catalogued as saponins (SPNS), such as Tuberoside J, Matesaponin 5, and Asparagoside B significantly contributed to component 1 in the aqueous phase, while other important features for component 2 belong to flavonoid class (FLV) and SPNS compounds. Furthermore, both components in the organic phase were mainly composed of glycerolipids (GL) and terpenoids (TER), as well as a putative carotenoid, (5cis,5'cis,9cis,11'cis)-

Level 1 and 2 Metabolomic Identification Analysis
We putatively identified approximately 270 compounds and classified them in 52 compound classes (Table S1). Putative identifications were taken into consideration with a high match score (>90%). Terpenoids, fatty acids, and glycosylated compounds were the most abundant groups. As was predicted, different capsaicinoid compounds were also detected with a high match score. Alkaloids, carotenoids, saponins, and phenolic compounds were also detected in our study.
Capsaicin and dihydrocapsaicin were validated by matching their retention times and MS/MS spectra with those of the analytical standard (Level 1 identification). Furthermore, commonly reported compounds, such as carotenoids and capsaicin related compounds, were detected and putatively identified in the same manner in our samples (

Different Metabolomic Profiles in Capsicum Sections
Based on results from volcano plot comparisons, we found a total of 60 putative compounds with significant differences between the parts of the Tabasco pepper fruit. Figure 4 shows the distribution of features between placenta and pericarp tissues. As was expected [28], capsaicinoids were more abundant in placenta than pericarp. In contrast, pericarp was richer in glycosylated compounds and terpenoids such as acalyphin, capsiate, and capsidiol. Some significant ions remain unknown that still need to be identified.

Different Metabolomic Profiles in Capsicum Sections
Based on results from volcano plot comparisons, we found a total of 60 putative compounds with significant differences between the parts of the Tabasco pepper fruit. Figure 4 shows the distribution of features between placenta and pericarp tissues. As was expected [28], capsaicinoids were more abundant in placenta than pericarp. In contrast, pericarp was richer in glycosylated compounds and terpenoids such as acalyphin, capsiate, and capsidiol. Some significant ions remain unknown that still need to be identified. After feature screening and putative identification, Venn diagrams were generated ( Figure 5) to show similarities and differences between fruit parts, according to the fold change values obtained in Table S1. Noticeably, around 30% of the complete dataset of identification was shared by all three fruit parts in almost all the extraction solvents, the exception being for the organic phase in the negative ionization mode, which only share 9.32% similarities. As can be seen in the Venn diagrams, a greater number of putative metabolites were identified in the positive ionization mode. Those easily detected in the positive ionization mode were the compound classes alkaloids, carotenoids, fatty acids, glycosylated compounds, terpenoids, and saponins; while in the negative ionization mode amino acid-derivate compounds, sphingolipids, and phospholipids were detected. After feature screening and putative identification, Venn diagrams were generated ( Figure 5) to show similarities and differences between fruit parts, according to the fold change values obtained in Table S1. Noticeably, around 30% of the complete dataset of identification was shared by all three fruit parts in almost all the extraction solvents, the exception being for the organic phase in the negative ionization mode, which only share 9.32% similarities. As can be seen in the Venn diagrams, a greater number of putative metabolites were identified in the positive ionization mode. Those easily detected in the positive ionization mode were the compound classes alkaloids, carotenoids, fatty acids, glycosylated compounds, terpenoids, and saponins; while in the negative ionization mode amino acid-derivate compounds, sphingolipids, and phospholipids were detected. Listed in Table 3 is the putative identification that presents a differential abundance between fruit parts, including the fragmentation pattern (MS 2 ) and adducts for putative annotation. Fold change values are shown in Table S1.  Table 3 is the putative identification that presents a differential abundance between fruit parts, including the fragmentation pattern (MS 2 ) and adducts for putative annotation. Fold change values are shown in Table S1.

Discussion
The global metabolic comparison between the tissues and seeds of C. frutescens showed several feature differences between the pericarp, placenta, and seed. The Level 1 and 2 confidence metabolite annotations allowed us to assign a putative identification to these ions. Around 30% of metabolites were shared between all three parts. Compounds related to the primary metabolism showed few significant differences, they included amino acid related compounds, fatty acids, and phospholipids. As shown in the Venn diagram ( Figure 5) and Table S1, placenta and pericarp have the biggest compound class diversity. Significantly, the seeds presented a higher number of putative identifications, and these were primarily saponins, terpenes, and fatty acids.
Pericarp compound classes were mainly composed of glycosylated compounds and terpenoids. Complementary to these findings, Materska demonstrated that chili pepper pericarp is abundant in glycosylated compounds [30]. Likewise, terpenoids were distributed in the whole fruit but pericarp showed a slightly higher proportion of them. These compounds are highly abundant in spices and herbs and give a wide range on the aroma and flavor spectrum [1]. Similarly, terpenoids have been described as showing antibiotic properties [41] and have been used in fragrances [42].
Placental tissue showed a large number of previously reported compounds with bioactivity, mainly capsaicin-and capsinoids-related compounds. In addition, alkaloids and tocopherols were present, a fact that is in agreement with current literature [28,30]. Found to be abundant in this fruit compartment were 6-O-acetylaustroinulin (terpenoid) and Myrciacitrin V (flavonoid) and they have not been reported in the Capsicum fruit.
The compounds found in chili seeds were predominantly fatty acids (3,16-Dihydroxypalmitate, Sterculynic acid) and saponins (Capsicoside A, Eleutheroside L and Tragopogonsaponin F) where they function as reserve nutrients for embryo development and propagation [43]. Moreover, seeds showed the presence of terpenes (Ursolic acid 3-[glucosyl-(1->4)-xyloside]) which are known to function as a natural promoter of predation and, as a consequence, a seed disperser [44]. Ritota et al. (2010) reported an abundance of fatty acids in sweet pepper species by nuclear magnetic resonance spectroscopy, in which polyunsaturated fatty acids were easily detected and pre-identified [6].
Capsaicin, Dihydrocapsaicin, and capsaicinoids derivatives mainly accumulate in the placenta, as previously reported [27,49]. This class of compounds represents the most described and abundant metabolite in this genus and are predominantly known as being responsible for the pungency. Different bioactivity assays have been developed, demonstrating properties of capsaicinoids over different cell lines and metabolism, including as an analgesic and for weight-loss [22,50]. Large abundance of capsaicinoids in the chili fruit placenta was proposed by Tewksbury and Nabhan (2001), who suggest that capsaicin selectively discourages vertebrate predators (capsaicin has been found to repel or poison mammals) without deterring more effective and important seed dispersers, such as birds. [51].
A variety of new compounds in Capsicum genus, also reported in different species, were detected in pericarp, including isopimaric acid [41], which is a terpenoid with bioactive properties. Additionally, other compounds such as Pedalitin [52], Xi-8-acetonyldihydrosanguinarine [53], Pratenol B [54], Uralenneoside [55] were detected in pericarp and have been previously reported as bioactive compounds. Quercetin 3-(6"-malonyl-glucoside) is an anthocyanin-related compound that has not been reported in pepper fruit, but this compound class is well known to be localized mainly in pericarp [56,57], due to its involvement as a protection system against solar damage in plants and to attract potential pollinators [29].
The existence of terpenes in seeds may result in different aromas that have been shown to firstly attract birds to mature fruits during the day [63] and secondly, to promote the dispersal of seeds. This function supports the ecological relationship between birds and chili pepper fruits, attracting the most beneficial vertebrate predators [51].
In summary, the non-targeted LC-MS metabolomics method that was developed in this study is shown to be a powerful tool for the putative identification of tissue-specific secondary metabolites at the red mature stage of chili pepper fruit. The use of databases available online gave rise to a faster comprehensive elucidation of global characteristics of a complex matrix than more traditional phytochemical studies. Nutraceutical, aroma, flavor, and new compounds that have not been reported before were putatively identified and related to pericarp, placenta or seeds of C. frutescens. As presented here, some of these compounds have been reported with bioactivity properties, supporting empirical properties of pepper fruit that have been known for centuries. The procedure developed here will be utilized for further studies in our laboratory, including to enable the exploration of comparisons between wild cultivars of chili pepper fruit with their cultivated counterparts and for the further understanding of secondary metabolism in this crop. We recommend that complementary analysis should be carried out to confirm structural elucidation. In addition, compound isolation and bioactivity properties should be considered in future studies.

Plant Material and Dissection of Tissues and Seed
Seeds of Tabasco pepper (C. frutescens L.) were treated with 3% hypochlorite solution. Plants were grown in optimum conditions (30-32 • C), at greenhouse facilities between June and September of 2016. Fruits from different plants were collected at 60 DPA (red ripe stage), washed with deionized water and immediately frozen with liquid nitrogen and stored at −80 • C until dissection and analysis.
Five biological replicates (plants) were considered for the experiment and three fruits per plant were collected. Each fruit was first placed into dry ice to facilitate hand dissection into pericarp, placenta, and seed using a sterile scalpel. All fruit parts were ground using a ball mill (Retsch MM301) under cold conditions and applying liquid nitrogen.

Chemicals, Reagents, and Standards
All chemicals and reagents were purchased from AccesoLab S.A. de C.V. (Mexico, Mexico). Capsaicin and dihydrocapsaicin analytical standards, formic acid, methanol, acetonitrile were HPLC grade and purchased from Sigma-Aldrich (Mexico, Mexico).

Sample Extraction and UHPLC-MS Analysis
For metabolite extraction, the method employed was adapted from Matyash [64] as follows: Methanol, 1.5 mL, was added to 100 mg of sample in a test tube and vortexed for 1 min, then, 5 mL of diethyl ether was added. The mixture was incubated with gentle stirring for one hour at room temperature. Next, 1.5 mL of ultra-pure water (18 Ω, milli-Q system) was added and mixed vigorously for a further minute then kept at room temperature for 10 min to allow phase separation. After that, the sample was centrifuged at 1000 × g for 10 min. Aqueous and organic layers were recovered separately and vacuum dried (miVac®, Genevac) at 30 • C for 30 min and finally kept at −80 • C until further analysis.
Three quality control (QC) samples were prepared to account for instrument drift and system calibration during analysis in UHPLC-QTOF-HRMS; each QC sample was prepared by mixing homogenously all sample extracts into a new single vial, in both separated phases containing polar and non-polar compounds. QC samples were distributed at the beginning, middle, and end of the injection run list. Analytical standards of capsaicin and dihydrocapsaicin were injected under the same conditions as samples. Extraction blanks were also considered during the experiment.
For LC-MS analysis, all samples (including QC, analytical standards and blank extraction) were resuspended in 1 mL of acetonitrile/ultra-pure water 50:50 (v/v) and filtered through a membrane of 0.2 µm (PTFE, Agilent Technologies, Santa Clara, USA). Samples were injected according to a randomized list order on an UPLC®(Acquity class I, Waters, Milford, CA, USA) coupled with an orthogonal QTOF (SYNAPT G1 HDMS, Waters, Milford, CA, USA) mass spectrometer. Chromatographic separation was achieved on a reversed phase CSH C18 column (2.1 mm × 150 mm, 1.7 µm, Waters, Milford, USA) maintained at 30 • C during chromatographic separation. Auto-sampling of 10 µL per sample was injected. Compounds were eluted using ultra-pure water with 0.1% (v/v) formic acid (solvent A) and acetonitrile with 0.1% (v/v) formic acid (solvent B) with a flow rate of 0.3 mL/min with the following gradient program: From 0.5 to 30 min, 1-75% B; 30 to 31 min, 75% B; 31 to 31.5 min, 75-100% B; 31.5 to 34.5, 100% B; 34.5 to 34.6, 100-1% B; 34.6 to 36 min, 1% B. The mass spectrometer mass range was set from 50 to 1500 Da. Both ionization modes were injected separated. For negative electrospray ionization (ESI) mode, the conditions were set as follows: Capillary voltage 2 kV; cone voltage 40 V; source temperature 150 • C; cone gas flow 20 L/h; desolvation temperature 350 • C; desolvation gas flow 600 L/h. For the positive ESI mode: Capillary voltage 3 kV; cone voltage 40 V; source temperature 130 • C; desolvation temperature 350 • C; desolvation gas flow 700 L/h. Leucine-Enkephalin (2 ng/mL) was infused as LockSpray reference internal mass calibrant at a flow rate of 5 µL/min and its signal was monitored every 10 s. The data format was collected in a continuum mode with a MS scan time of 1.5 s. In both the positive and negative ionization mode, data were acquired in MS E experiments; using Argon as the collision gas with a collision energy in the trap region of 6 eV (Function 1, low energy) and ranged from 20-40 eV (Function 2, high voltage).

Data Analysis
Raw data was imported to Progenesis QI for small molecules software (Non-Linear Dynamics, Waters, Milford, MA, USA) for automatic alignment, normalization, deconvolution, and compound pre-identification over all samples separating the aqueous and organic phases. The RT range was limited from 0.5 to 35 min for pre-identification method. Pre-identification was performed using Chemspider Databases (PlantCyc, Plant Metabolic Network, KEGG, HMDB and ChEBI) and with an in-house database with a minimum match of 90% for precursor ions, MS/MS data and isotope distribution was included for increasing match score values. Statistics and graphics were performed using EZinfo 3.0 (Waters, Milford, MA, USA) and R (3.3.3v, Vienna, Austria) [65] software. Compounds were grouped according to their compound classes. The resulting data was first mean centered and scaled to Pareto and then submitted to a principal component analysis (PCA) using the first three components. Results were analyzed using one-way ANOVA and q-values were established using the false discovery rate (FDR < 0.01) to correct multiple comparisons by the Benjamini-Hochberg procedure [66].