Impaginato 305 Adv. Hort. Sci., 2021 35(3): 305­318 DOI: 10.36253/ahsc­10630 (*) Corresponding author: cosimo.taiti@unifi.it Citation: COMPARINI D., TAITI C., LANZA M., VITA F., PAN­ DOLFI C., LUTI S., SPINELLI F., PAZZAGLI L., MAN­ CUSO S., 2021 ­ Comparison of wild and domesti‐ cated hot peppers fruit: volatile emissions, pun‐ gency and protein profiles. ­ Adv. Hort. Sci., 35(3): 305­318. Copyright: © 2021 Comparini D., Taiti C., Lanza M., Vita F., Pandolfi C., Luti S., Spinelli F., Pazzagli L., Mancuso S. This is an open access, peer reviewed article published by Firenze University Press (http://www.fupress.net/index.php/ahs/) and distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Data Availability Statement: All relevant data are within the paper and its Supporting Information files. Competing Interests: The authors declare no competing interests. Received for publication 9 March 2021 Accepted for publication 3 August 2021 AHS Advances in Horticultural Science Comparison of wild and domesticated hot peppers fruit: volatile emissions, pungency and protein profiles D. Comparini 1, Taiti C. 1 (*), Lanza M. 2, Vita F. 3, Pandolfi C.1, Luti S. 4, Spinelli F. 1, Pazzagli L. 4, Mancuso S. 1 1 Department of Agrifood Production and Environmental Sciences, University of Florence, Viale delle Idee, 30, 50019 Sesto Fiorentino (FI), Italy. 2 IONICON Analytik GmbH, Eduard Bodem Gasse 3, 6020 Innsbruck, Austria. 3 Department of Agriculture, Food and Environment, University of Pisa, Via Mariscoglio, 34, 56124 Pisa, Italy. 4 Department of Biomedical Experimental and Clinical Sciences, University of Florence, Viale Morgagni, 50, 50134 Florence, Italy. Key words: Artificial Neural Network, Capsicum spp. classification, capsaicin, SDS­ PAGE, VOCs. Abstract: Capsicum plant species are globally cultivated in warm and temperate regions, being important for agro­economic, biological and cultural aspects. While their worldwide spread and their ability of cross­pollination to easily hybridize play an important role in the formation of numerous species and vari­ eties but also create confusion for their classification. For this reason, the cate­ gorization of species and varieties is complex and several methods have been used to evaluate pepper plant origin and evolution. Therefore, the objectives of this study were to compare a wild pepper (Capsicum chacoense) with other two domesticated cultivars belonging to different species such as Capsicum annuum and C. baccatum and draw conclusions about their origins using different approaches. For this purpose three methodologies have been used and com­ pared: the comparison of their fruits volatile organic compounds (VOCs) emis­ sions , their capsaicin and dihydrocapsaicin content and the leaves proteomic profiles. The VOCs analysis has been conducted by a time­of­flight mass spec­ trometry (ToF­MS) with an innovative approach to better identify all the com­ pounds detected, in particular using two different ionization agents (H3O + and NO+) to better identify all the compounds detected. The VOCs and pungency analyses were then used to build back propagation neural networks (BPNN) and a Random Tree classifier to conduct a multivariate analysis and evaluate the most species­specific volatiles. The outcomes appeared to be a most accu­ rate approach with respect to the traditional varieties descriptors used for pep­ pers discrimination. The BPNN led to the identification of several putative volatiles as good candidates for the recognition of these species or significant nodes in a decision learning tool. Finally, protein profiles have been obtained by SDS­PAGE analysis on the leaves to perform a fast proteomic comparison among the species. The protein profiles showed the C. baccatum and C. cha‐ coense were more similar to the domesticated pepper C. annuum. https://doi.org/10.36253/ahsc-10630 http://www.fupress.net/index.php/ahs/ http://creativecommons.org/licenses/by/4.0/ http://creativecommons.org/licenses/by/4.0/ http://creativecommons.org/licenses/by/4.0/ Adv. Hort. Sci., 2021 35(3): 305­318 306 1. Introduction Origin and classification of Capsicum The origin of chili peppers has been located in sev­ eral locations of Latin America as testified by archeo­ logical records and pepper is usually classified as one of the first new world domesticated plants (Pickersgill, 1969; Long­Solis, 1986; Perry et al., 2007). Peppers represent one of the most ancient plants cultivated in America together with the Phaseolus L., maize and other plants of the Cucurbitaceae family, fundamental ingredients of natives’ diet. About 36 different species belong to the genus Capsicum and even today there are many wild species to be defined from the taxonomic point of view, consequently, we cannot exclude the existence of new individuals currently unknown (Davenport, 2004). The identification and maintenance of the genetic diversity in Capsicum are important to avoid genetic erosion. Several species exist and Capsicum species have recently been described from Bolivia (Nee et al., 2006). Among these, only five species respectively C. annuum (variety annuum), C. chinense, C. frutescens, C. baccatum (variety pendolum and umbelicatum) and C. pubescens have been domesticated in the past by the American natives and later, in the post­ Colombian period, they have been widely imported and cultivated in temperate and tropical regions for the characteristics of their fruits (McLeod et al., 1979; McLeod et al., 1983). The broad geographical distribution of this genus, usually used as feeds not only by humans but also by birds that don’t have receptors for capsaicin, associated with the antiquity of the origin and the high frequency of hybridization, created a broad genetic variability during the evolu­ tion and resulted in many morpho­qualitative differ­ ences among cultivars of homologous species (Pozzobon et al., 2006). In fact peppers plants have been used for several reasons, starting from their high nutritional value, good content in vitamins but also as medicine or mystic rituals. This determined a rapid diffusion of these plants in the old continent, stimulated also by their spicy flavour similar to the black pepper, a very valued spice, whilst peppers pre­ sented higher production and more flexibility as food. Furthermore the mechanism of cross­pollina­ tion and the ability of peppers to easy hybridize played another role in the formation of numerous varieties with specific features. During the last cen­ turies, botanists have been active to cross­pollinate creating confusion for the classification of the vari­ eties and the identification of indigenes species. For these reasons the zone where each species originat­ ed is still subjected to debate. Brazil is considered as the center of origin of the genus capsicum and cur­ rently represents the most important source of genetic diversity (Buso et al., 2001). For example, a recent survey of chili cultivars from the state of Roraima in northwestern Brazil noted 60 distinct lan­ draces of peppers from four different species: Capsicum annuum; C. frutescens, C. baccatum and C. chinense (Barbosa et al., 2006). Therefore Brazil and Bolivia possess the highest number of wild species (Pickersgill, 1984), however, a broad and complete study about diversity has not been done yet for the native species. It is anticipated that continuing, plant exploration in southern Peru, Bolivia and Brazil, will yield additional new narrow endemic pepper species (Russo, 2012). As consequence, a taxonomic classification of peppers and the determination of all relations among species are challenging and created many debates about the origin and evolution of the Capsicum genus. Most current peppers are derived from domesticated species. The difficulties in identifying them lie in dialectal names linked with the local tradi­ tion, making challenging their classification. For example, a study among the Mexican peppers popu­ lation underlined that among about 200 common names used to refer to different peppers, only approximately 15 types were different commercial pods (Russo, 2012). This situation contributes in the last filthy years several authors to focus on the taxon­ omy and origin of the species of this genus without coming out with a complete and flawless analysis (Heiser and Smith, 1953; Pickersgill, 1988; Hunziker 2001; Barboza and Bianchetti, 2005). It has been proposed that the C. chacoense in Bolivia is one of the most ancestral nuclear centers for the origin of Capsicum (2n=24) and it is considered as the basal for the evolution of the species (McLeod et al., 1982). Furthermore, a study based on karyotype and other considerations of different Capsicum by Moscone et al. (2007) concluded the existence of a triple origin of domesticated Capsicum species. The whole genus has been hypothesized to be originated in the tropical regions of the Americas and the ances­ tral species should be born in the Chaco zone which extends into several parts in the country of Argentina, Bolivia, Brasil and Paraguay and propose that the genus “chacoense” as one of the most primitive origi­ Comparini et al. ‐ Wild vs domesticated peppers 307 nated pepper taxon. All the lines that we know today derived from this first line of evolution and have devel­ oped and differentiated according to the characteris­ tics of the new habitat: in the north of the Amazon to the ancient forms of C. annuum, on the north coast of Brazil and Venezuela in the forms of C. chinense and C. frutescens; and in a subsequent period from the same ancestral form were differentiated to C. rhomboideum (Colombia, Ecuador) and C. galapagoense (Galapagos Islands) (Moscone et al., 2007). Starting from a second evolutionary line instead, they are originated different species: in subtropical areas of Brazil and later the C. baccatum and C. exim‐ ium; in the arid regions of Peru the C. cardenasii, C. tovarii and C. pubescens, while in Paraguay C. flexuo‐ sum and C. praetermissum. Finally, it was hypothe­ sized that the migration of ancestral forms of C. flex‐ uosum and C. praetermissum gave rise, in some areas of Brazil, the largest center of diversification of the genus represented by the large group of wild species in 26 chromosomes (Moscone et al. , 2007). Currently, the species that best reflects the ancestral morphologically and physiologically, is the C. cha‐ coense (Hunziker, 2001). C. chacoense has also been used in breeding C. annuum programs, focusing on tobacco mosaic virus resistance (Boukema, 1982). The wild species with 24 and 26 chromosomes pre­ sent different morphological traits and geographical distribution (Pozzobon et al., 2006). The differences that arise are probably related to the different agents of seed dispersal, birds for the first whilst bats and other small mammals for species of 26 chromosomes that present hanging fruit, inconspicuous, and a little spicy. In a second time, the domestication and subse­ quent human selection caused a selective pressure in favor of large hanging fruits, less attractive for birds with small exception (e.g. C. frutescens “Tabasco” varieties), (AISPES, 2010). The classification of chilies, like that of any multifarious group of cultivars, is con­ fusing. For example, a very large amount of pod types exist in C. annuum, C. chinense and C. bacca‐ tum (Bosland and Votava, 2012). Usually, the classification of the genus Capsicum and the varieties, belonging to each species, is car­ ried out based on morphological descriptors that define the shape of flowers and fruits (Pickersgill, 1971; Moscone et al., 2007; Ince et al., 2009; Sudré et al., 2010) and the current system for classifications involved genus, species, variety, pod type and culti­ vars (Bosland and Votava, 2012). However, other descriptors are considered essential for an accurate germplasm characterization, such as those indicated by IPGRI (International Plant Genetic Resources Institute). Furthermore, the characterization and evaluation of the species belonging to the genus Capsicum are particularly interesting for breeders, gene banks, because of the large genetic variability available (Guzmán et al., 2005; Sudré et al., 2006; Ince et al., 2009). They can furthermore be identified from the different flavors of the fruits and two main factors that contribute to the aroma perception are pungency and aroma, and these are associated with the fruit volatile compounds (Taiti et al., 2015). With the advancement of computer technology, multivariate methods have become an important tool for taxonomic classification (Ortiz et al., 2008). However, the procedures of statistical classification require a data set based on a large number of vari­ ables. Thus, the paper has aimed to analyze and char­ acterize the wild C. chacoense and to compare its pro­ file of volatile emissions, pungency, and proteins with two domesticated species i.e. C. annuum and C. bac‐ catum to elucidate the evolution of the Capsicum genus. For this reason, the volatiles compounds emis­ sions profile of each species has been used to build an Artificial Neural network and differentiate and identify the species. Finally, an additional proteomic analysis of the leaves of the peppers has been used to evalu­ ate the relationships among the species and compare the results obtained from the volatiles emission, pun­ gency and protein expression profiles of each species. 2. Materials and Methods Fresh pepper material Ripe pepper fruits belonging to 3 different species were used in this study: Capsicum annuum var. Ciliegino, C. baccatum var. Brasileiro and C. cha‐ coense (wild accession). The fruits were collected from ten plants, for each species, grown into green­ house (Florence, Italy). Fully matured uniform­sized fruits were collected within 24h after the 100% color surface was reached. All plants were obtained from seeds, germination and growing phases were made following the same system used in the previous work of Taiti et al., 2015. All plants belonging to each species were grown in the greenhouse in three differ­ ent rooms to avoid the effects of cross­pollination. Capsaicin and dihydrocapsaicin quantification Capsaicinoids are a group of alkaloids produced as Adv. Hort. Sci., 2021 35(3): 305­318 308 secondary metabolites by chili peppers responsible for the pungency. Among several structural analogs capsaicinoids, capsaicin and dihydrocapsaicin are the two most potent and abundant compounds account­ ing for more than 90% of total capsaicinoids in chili pepper (Ziino et al., 2009). Capsaicin and dihydrocap­ saicin have been extracted from whole frozen fruits and calculated as the average of five extraction for each species (n=5, SD). About 2 g of frozen fruits have been weighted and then pulverized in 10 mL of cold acetone at 4°C and kept overnight. Then 50 µl of the sample have been collected by using a 0.22 µm filter syringe and used for the quantification. All data have been calculated as µg of capsaicin or dihydro­ capsaicin content per gram of fresh weight after nor­ malization based on the exact weight of the initial fresh sample. RP­HPLC quantification of capsaicin and dihydro­ capsaicin was performed by using a C18 column, 3 µm, 15x4.6 cm (Supelco, Bellefonte, Pennsylvania, USA). The stock solution containing capsaicin and dihydrocapsaicin (cod, 360376 Sigma­Aldrich, St Louis, MO, USA) was prepared in 20% Acetonitrile at concentrations of 7.5 mg/mL. For calibration curves construction 30 µl, 60 µl, 90 µl, 120 µl, 150 µl, 300 µl and 500µl of stock solution were analyzed obtaining a linear curve for both capsaicin and dihydrocapsaicin (R2=0.9912 and R2=0.9923, respectively). Elution gra­ dient was performed at a flow rate of 0.8 ml min­1 with the following solvent system: 10mM trifluo­ roacetic acid (TFA) in acetonitrile (solvent A); 10mM TFA in water (solvent B). The gradient used was 20% A for 2 min, from 20% to 100% A in 15 min, holding at 100% A for 10 min, from 100% A to 20% A in 2 min, and detection was based on UV absorbance at 280 nm. Under these conditions, the capsaicin peak appeared at a retention time (Rt) of 10.6 min and dihydrocapsaicin at Rt of 11.3 min. Quantification was calculated using the Chromeleon software. Volatiles organic compounds (VOCs) analysis SRI‐MS ToF protocols. For headspace analysis, pepper fruits have been selected among ten plants and the uniform­sized fruits were collected at the optimal ripening stage (100% of coloration). For each species ten samples were analyzed, each constituted of 10 grams of fresh pepper fruits. Using a commer­ cial PTR­TOF 8000 (Ionicon Analytik GmbH, Innsbruck, Austria) instrument with SRI­MS (Switchable Reagent Ions ­ Mass Spectrum) upgrade, the analysis of the samples was carried out following a similar procedure described in Taiti et al. (2015). In short, 10 g of freshly cut chili pepper (including the seeds) without any pre­treatment were placed in a glass jar (750 mL at 25°C, with a dynamic headspace flushing flow rate of 200 mL min­1) equipped with two Teflon inlet and outlet tubes on the opposite side, which were respectively connected to a zero­air generator (Peak Scientific) and the PTR­TOF MS instrument. Moreover, for the first time, the Switchable Reagent Ion System (H3O + and NO+) has been used to produce different ionization agents for capsicum study. Using the additional precursor (reagent) ions as NO+, besides the usual H3O+, improved the analytical possibilities of this technique (Wang et al., 2004; Mochalski et al., 2014). In particu­ lar, the SRI system allows: (1) the analysis of VOCs which are not detectable with the reference ion H3O + (e.g. alkanes); (2) the separation of isobaric com­ pounds as in the case of aldehydes and ketones using NO+ (Jordan et al., 2009; Del Pulgar et al., 2013). For a detailed explanation of the system see Blake et al. (2006) and Blake et al. (2009). All samples were ana­ lyzed using the same procedure and the VOCs were assessed with H3O + and NO+ as reagent ions; the drift conditions for each primary ion used are reported in Table 1. The sampling time for each channel of TOF acquisition was 0.2 ns, for a mass spectrum com­ prised between m/z 20­210. The duration of a single sample measurement was 120 seconds, which corre­ sponds to 60 mass spectra. All the samples were ana­ lyzed in an air­conditioned room, with a constant temperature of 25±1°C (Mancuso et al., 2015). The SRI­MS upgrade consists of an additional mass flow controllers for the respective reagent gases (water vapor for H3O +, charcoal filtered air for NO+); for a detailed explanation see Jordan et al. (2009). In short, the use of NO+ and the H3O + as a reagent ion Table 1 ­ Instrumental condition throughout the experiment Primary Ion Drift voltage (V) Pressure (mB) Temperature (°C) Us (V) Uso (V) Ihc (mA) Udx (V) E/N (Td) Mass calibration Mass calibration Mass calibration H3O + 594 2.25 110 110 85 4.0 35 140 21.022 29.997 59.041 NO+ 600 2.30 110 25 80 5.0 36 137 21.022 32.002 47.997 Comparini et al. ‐ Wild vs domesticated peppers 309 Table 2 ­ List of the average m/z­signals that can be specifically assigned using H3O + as reagent ion: Volatile Organic Compounds headspace intensity expressed in ncps (n=10; ±SD); chemical formulae and tentative identifications for each signals detected; the compounds identification was linked to the PTR­ToF­MS pattern fragmentation references (a) or previously reported in Capsicum species (b) Measured mass (m/z) Capsicum Annuum Ciliegino Capsicum Baccatum Brasileiro Capsicum Chacoense Wild pepper Sum formula Tentative identification PTR Pattern fragmentation a Capsicum Literature b 1 27.022 799.86±234.50 200.93±109.28 496.95±140.33 C2H3 + Acetylene 2 31.018* 615.25±205.40 184.12±62.55 81.22±21.30 CH3O + Formaldehyde [2] 3 33.033 1712.59±220.12 415.39±167.27 1067.70±200.55 CH5O + Methanol [1] 4 41.038 2612.818±638.34 983.12±320.11 378.78±100.33 C3H5 + Alkyl fragment (Alcohols and esters) [1] 5 43.054 578.30±190.22 133.23±33.90 101.94±25.90 C3H7 + Alkyl fragment (Alcohols) [1] 6 45.033 3049.50±982.28 529.04±67.40 5638.53±1230.45 C2H5O + Acetaldheyde [2] 7 47.010 353.83±111.89 224.89±51.58 636.43±160.55 CH3O2 + Formic Acid/Formates [2] 8 47.049 4.80±1.20 3.50±0.80 2.95±1.00 C2H7O + Ethanol [1] Rodriguez­Burruezzo et al., 2010 9 53.030 102.26±35.40 50.90±22.33 17.53±5.00 C4H5 + Cyclobutadiene 10 55.050 6.82±1.90 1.99±0.50 3.32±1.30 C4H7 + C4 aldehydes fragment [1] 11 57.033* 2910.50±1109.12 366.28±78.54 179.55±49.50 C3H5O + C3 aldehydes and ketones fragments [6] 12 57.069 274.43±85.40 45.91±17.22 33.02±9.25 C4H9 + Alcohol fragments (1­Butanol, 1­ Pentanol, 1­Hexanol, 2­Methyl­1­ propanol, Pentanol, 1­Heptanol, Octanol, Nonanol) [1] 13 59.049* 553.06±133.09 196.04±65.90 459.80±180.44 C3H7O + Propanal, Acetone [2] Ziino et al., 2009 14 61.028 303.46±44.65 120.22±39.10 355.05±44.44 C2H5O2 + Acetates [6] Ziino et al., 2009 15 63.027 4.45±1.30 4.64±1.20 5.05±0.80 C2H7S + Dimethylsulfide Taiti et al., 2015 * The signals that mostly contributed to the BPNN classification have been marked. to be continued... improves the analytical performance of the tool, par­ ticularly for the separation of isobaric compounds and for the detection of compounds with proton affinities lower than that of water. Furthermore, as reported by Edtbauer et al. (2014) when the PTR­MS instrument works in NO+ mode can improve the selectivity of compounds detection. Mass data statistical analysis. Since the external calibration provided by the tool gave a poor mass accuracy, it has been performed off­line thus ensur­ ing, a high mass accuracy generally than 0.001Th, which in most cases allowed the formula identifica­ tion (Taiti et al., 2017). The raw data of each peak spectrum (calculated as number of counts per sec­ ond, cps) were acquired with the software TOFDaq (TOFwerk AG, Switzerland) by setting a dead time of 20 ns for the Poisson correction, instead, for peak quantification, the resulting data were corrected according to the duty cycle and the signals were nor­ malized (ncps “normalized count per second”) as described by Herbig et al. (2009). Moreover, the Poisson correction has been applied to correct all spectra for any count losses. Finally, VOCs putative identification was based on a high instrumental mass resolution and the fragmentation patterns of pure standards available in the bibliography (Buhr et al., 2002; Lee et al., 2006; Maleknia et al., 2007; Kim et al., 2009; Tani, 2013; Aprea et al., 2015; Taiti et al., 2019) and integrated with previously detected VOCs emitted from Capsicum fruit available in literature (Table 2 and 3). Data processing and classification methods Back propagation neural network. In this study, the capsaicinoid contents and the 52 volatile signals detected by the PTR­ToF­MS by using H3O + as reagent ion were used as input layers, and the 3 pepper species represented the output (30 single pepper fruits analysis, ten for each species). The BPNN was built using a data mining software (Weka 3.6.14) and the Multilayer Perceptron classifier was used for the classification. Two BPNN were made, one with only the VOCs emission profiles, and another with VOCs and capsai­ cinoids. The number of hidden neurons and the num­ ber of iterations was adjusted to optimize the neural network activity. Many factors, such as learning schemes, numbers of nodes, and connections between them, play an important role in determing of the best configuration of the hidden layers (Zurada 310 Adv. Hort. Sci., 2021 35(3): 305­318 Table 2 ­ List of the average m/z­signals that can be specifically assigned using H3O + as reagent ion: Volatile Organic Compounds headspace intensity expressed in ncps (n=10; ±SD); chemical formulae and tentative identifications for each signals detected; the compounds identification was linked to the PTR­ToF­MS pattern fragmentation references (a) or previously reported in Capsicum species (b) Measured mass (m/z) Capsicum Annuum Ciliegino Capsicum Baccatum Brasileiro Capsicum Chacoense Wild pepper Sum Formula Tentative identification PTR Pattern fragmentation a Capsicum Literature b 16 65.038 4.87±3.25 2.49±0.80 3.45±1.11 C5H5 + Alkyl fragment/ Terpenes fragment [1/2] 17 67.050 61.35±16.66 13.86±6.40 6.77±1.40 C5H7 + Terpenes fragment [2] 18 69.033 80.90±21.98 17.82±4.50 11.92±4.10 C4H5O + Furan 19 69.069 395.06±133.90 139.11±80.33 92.50±18.44 C5H9 + Isoprene/Alkyl fragment (e.g. 2­ methylbutanal, 1­octen­3­ol) [2/6] 20 71.049* 21.29±7.10 6.52±2.12 8.70±2.50 C4H7O + 2­Butenal Taiti et al., 2015 21 71.086 15.92±5.90 4.71±190 7.65±1.90 C5H11 + Alchol (3­methyl­1­butanol, Pentanol, Iso­pentanol, 2­ethyl­1­hexanol, Eggink et al., 2012 a 22 73.060* 43.02±12.35 6.62±2.10 15.45±4.90 C4H9O + Isobutanal/Butanone/Butanal [1] Ziino et al., 2009 23 75.044 57.76±22.25 14.33±4.50 12.16±6.00 C3H7O2 + Butanol/Methyl acetate/Propanoates [1] 24 77.038 7.09±2.79 7.77±2.20 14.53±3.50 C6H5 + Alkyl fragment 25 79.054 134.86±48.60 22.84±7.98 15.7±4.10 C6H7 + Benzene/Alkyl and terpenes fragment [2] 26 81.068 949.61±301.39 178.02±68.78 152.06±50.30 C6H9 + Terpenes fragment/Aldehydes frag­ ment (trans­ 2­hexenal) [4] Taiti et al., 2015 27 83.049 76.90±19.75 32.36±9.94 13.69±2.90 C5H7O + 2­Methylfuran Rodriguez­Burruezzo et al., 2010 28 83.086 175.59±50.10 92.78±29.50 16.29±4.90 C6H11 + C6 compounds/ Hexenol fragment [6] 29 85.064 46.91±15.40 9.90±3.10 18.20±4.30 C5H9O + Methyl­butenal/1­penten­3­one Ziino et al., 2009 30 85.101* 32.25±12.30 2.08±0.94 7.41±2.00 C6H13 + Alcohol (1­Hexanol/Nonanol) Ziino et al., 2009 31 87.045* 25.21±5.40 6.92±3.20 12.65±4.10 C4H7O2 + Diacetyl /2,3­butanedione [1] Ziino et al., 2009 32 87.080 10.23±1.21 6.44±0.95 2.01±0.55 C5H11O + 2,3­Methylbutanal /(Z)­2­penten­1­ ol/3­Pentanone [1] Rodriguez­Burruezzo et al., 2010 33 91.075 22.38±5.40 17.68±4.40 6.35±1.10 C4H11O2 + 2,3­Butanediol/Monoterpene ketone [5] Ziino et al., 2009 34 93.069 7.07±1.95 12.12±5.40 7.59±1.50 C7H9 + Terpene fragments (e.g. cymene, limonene) [3] Rodriguez­Burruezzo et al., 2010 35 95.086 27.45±6.99 4.18±2.20 14.97±1.10 C7H11 + 1­Methyl­1,4­cyclohexadiene Eggink et al., 2012 b 36 97.064 50.45±10.50 14.90±4.80 10.18±2.50 C6H9O + 2­Ethylfuran Rodriguez­Burruezzo et al., 2010 37 99.080* 60.65±20.30 8.62±2.75 15.98±6.30 C6H11O + cis­3­Hexenal/ (E)­2­Hexenal Rodriguez­Burruezzo et al., 2010 38 101.096 33.66±5.70 9.62±3.55 17.02±5.50 C6H12O + Hexanal/ (E)­2­Hexenol [1] Ziino et al., 2009 39 103.075 54.22±20.44 18.20±6.50 9.28±2.80 C5H11O2 + 3­Methylbutanoic acid [6] Zimmermann and Schieberle, 2000, Azcarate et al., 2010 40 105.069 4.98±0.65 5.23±0.78 5.96±1.12 C8H9 + Styrene/Styrol/Phenylethanol [6] Rodriguez­Burruezzo et al., 2010 41 107.085 9.50±5.44 3.65±2.30 12.85±7.20 C8H11 + p­Xilene [6] Eggink et al., 2012 a 42 109.101 13.85±4.50 3.15±0.90 6.06±2.20 C8H13 + Terpenes fragments [3] 43 115.111* 17.27±4.30 6.04±2.50 1.36±0.50 C7H14O + Heptanal [6] Ziino et al., 2009 44 117.091 6.35±1.60 14.05±3.00 3.96±1.30 C6H13O2 + Hexanoic acid/Hexanoates [6] Eggink et al., 2012 b 45 119.085 7.75±3.70 3.26±1.35 4.89±2.40 C9H11 + Terpenes fragment [3] Ziino et al., 2009 46 121.101 6.40±2.50 15.08±3.50 6.10±2.00 C9H13 + Terpenes fragment [3] 47 123.120 8.40±2.80 1.65±0.60 1.72±0.85 C9H15 + Sesquiterpene fragments [4] 48 135.117* 7.15±1.83 2.69±1.40 6.62±1.30 C10H15 + p­Cymene/Monoterpene ketone frag­ ment [6/5] Ziino et al., 2009 49 137.132* 17.12±4.05 9.75±1.10 23.88±4.35 C10H17 + Monoterpenes (e.g. (Z)­b­ocimene) [5] Eggink et al., 2012 a 50 149.132* 7.13±4.40 3.91±1.05 9.65±3.20 C11H17 + Sesquiterpenes fragments (e.g. Ectocarpene) [4] Taiti et al., 2015 51 205.195 25.44±7.10 11.90±3.33 9.59±2.65 C15H25+ Sesquiterpenes [4] Eggink et al., 2012 b TOTAL VOCs EMISSION (average ncps) 22116 4850 13214 * The signals that mostly contributed to the BPNN classification have been marked. Comparini et al. ‐ Wild vs domesticated peppers 311 Table 3 ­ SIFT­MS signals obtained using NO+ as reagent ions, in the range between to m/z 20­200. The analysis showed only the signals with intensity expressed in ncps higher than 1 (n=10; ±SD) Number of compounds Measured mass (m/z) Capsicum Annuum Ciliegino Capsicum Baccatum Brasileiro Capsicum Chacoense Wild pepper Sum Formula Tentative identification 1 41 10.91±3.10 45.53±16.60 19.47±8.85 C3H5 + Alkyl fragment (Alcohols and esters) 2 45 18.72±2.10 25.67±8.40 21.61±6.80 C2H5O + C6 fragment (e.g. (E)­2­Hexenol) 3 57 7.41±2.98 48.61±18.07 9.68±1.55 C3H5O + C3 Aldehyde 4 58 50.01±8.86 39.10±3.25 25.63±9.52 C3H6O + C3 ketones 5 69 13.07±2.33 45.73±18.80 20.95±8.55 C4H4O + Furan 6 83 15.11±2.23 79.05±20.44 13.85±0.60 C6H11 + C6 aldehydes (e.g (E)­2­Hexenol) 7 84 4.17±0.88 3.64±0.55 2.94±0.90 C5H8O + C5 8 85 7.62±3.20 27.85±7.22 10.69±3.33 C5H11O + Valeraldehyde 9 88 59.06±24.44 44.29±27.70 61.12±20.32 C3H6O NO+ Acetone 10 99 11.12±2.33 15.24±3.33 10.76±2.22 C6H13O + Hexanal/ C6 aldehydes (e.g (E)­2­ Hexenol) 11 106 3.67±0.78 2.87±0.30 2.780±6.52 C8H11 + Xylene 12 113 1.05±0.33 6.65±.55 1.34±0.55 C7H13O + Heptenal 13 114 2.60±0.44 10.15±1.1 2.16±0.30 C5H8 NO2 + Cluster C5 unsaturated ketones 14 116 9.47±2.63 26.33±8.33 6.39±1.20 C5H10 NO2 Cluster C5 ketones/Pentanone 15 128 3.34±0.81 2.04±1.01 1.88±0.72 C8H16O + 6­Methyl­5­hepten­2­ol 16 136 17.85±1.66 7.78±0.95 26.57±6.77 C10H16 + Monoterpene compounds 17 144 0.80±0.62 3.21±0.81 0.70±0.55 C6H10 NO3 + Cluster hexanedione/heptanone 18 166 0.73±±55 2.38±0.35 0.80±0.68 C10H16 NO+ Monoterpenes fragment Total VOCs emission average (ncps) 236.71 436.09 239.29 1992; Zurada and Malinowski, 1994). In our case, the minimum error was reached with a network com­ posed of 29 hidden neurons for both BPNN, posi­ tioned on one level, with the hidden layer activated by a logistic sigmoid activation function: f(x) = 1/(1+e­x) (1) These sigmoid functions fix the output signal limit between 0 and 1. The resulting function works as an output logic­gate that can be opened (1) or closed (0). Also, as part of a continuous function it can hap­ pen that a gate is partially opened (i.e., its value results between 0 and 1). Ideally, only a group of out­ puts, which represents an accession, would express a value of 1 (meaning correct identification) while the remaining groups would show a value of 0 (incorrect identification). In reality, this take place rarely, for this reason it is usually considered as “incorrect” a value closer to zero (wrong identification), while “correct” when the resulted value is close to 1 (cor­ rect identification) (Pandolfi et al., 2009). A 10­folds cross­validation was applied to test the performance of the model. The original dataset was essentially randomly segregated into 10 equal­sized groups. Each set is divided into two groups: 90% of data are used for training the network and 10% of data are used for the validation test. The cross­validation process is then repeated 10 times (the folds), in which every time a subsamples is validated. The results deriving from all folds are finally averaged to result in a single evaluation of the network’s performance. The values of the identification for each species were highlighted using a misidentification matrix. All identification processes executed by the network were averaged and the results represented in Table 4. The rows refer to the species in the test set, the columns report the species to which the test plants are referred by the neural network. An “Attribute selection filter” provided by Weka was also applied to the two sets of data, to determine the more dis­ criminant parameters. Random tree. The 51 volatile signals detected by the PTR­ToF­MS using H3O + (30 single pepper fruits analysis, ten for each species) were also analyzed using the “Random tree” algorithm, a decision tree Adv. Hort. Sci., 2021 35(3): 305­318 312 learning tool provided by the software (Weka 3.6.14). In classification trees, each internal node is labeled with an input feature, which can be informative to detect similarities or differences among the pepper species. Similarly to the BPNN, a 10­folds cross validation was applied to test the performance of the model, and the results from the folds are then averaged to produce a single estimation of the performance of the algorithm. SDS‐page protein analysis Protein extraction and quantification. Soluble pro­ teins were extracted from leaves of three Chili pep­ per plants of each species (Capsicum Chacoense, C. annuum and C. Baccatum) according to Vita et al. (2013), with some modifications. In short, for each analysis 100 mg of fresh leaves were grounded in liq­ uid nitrogen and homogenized with 1 mL of extrac­ tion buffer (5 M urea, 2 M thiourea, 40 mM Tris­HCl, 2% CHAPS, 50 mM DTT). The homogenates were cen­ trifuged for 15 min at 15,000 rpm. Supernatants were precipitated using TCA (15%, v/v) containing 0.007% β­mercaptoethanol in acetone at ­20°C for 2 h and successively at 4°C for a minimum of 2 h. Samples were centrifuged at 4°C for 15 min at 14,000 rpm, supernatants were discarded and pellets were washed twice with ice­cold acetone containing 0.007% β­mercaptoethanol. Pellets were dissolved in a rehydration buffer (5 M urea, 2 M thiourea, 4% CHAPS, 40 mM DTT). Protein quantification was per­ formed using a Bradford­based assay kit assay (Bio­ Rad Hercules, CA), using bovine serum albumin as a standard. SDS‐PAGE. The protein separation was carried out by the established technique of SDS­PAGE. In detail, the gels used (size 20 * 20 cm) had a thickness of 1 mm and were constituted by a stacking gel (4.8% T, 1.3% C, pH 6.8) and a running gel (15%, 1.3% C, pH 8.5). SDS­PAGE analyses were performed 4 times (n=4). Electrophoresis runs were carried out using the Protean XI cells (Bio­Rad Laboratories, Inc, Hercules CA) with specific parameters (for each gel 25 mA, 8h running time, temperature 15°C). Precision Plus Protein™ Unstained was the molecular marker used for the essay (Bio­Rad Laboratories Inc., Hercules, CA). The protein samples were analyzed by SDS­PAGE on gels stained with the Brilliant Blue G­ Colloidal Concentrate Coomassie (Sigma­Aldrich) according to the manufacturer’s instructions. The images of each gel were acquired using a Bio­ Rad densitometer GS­800 ™ in greyscale colors, with a definition of 300 dpi. The images were analyzed using the software Quantity One 1­D Analysis ™ soft­ ware (Bio­Rad Laboratories, Inc, Hercules CA). Dendrogram based on signal quantities was created using correlation­based distances and Ward’s method of agglomeration was used in the present analysis (Ward, 1963). 3. Results and Discussion Capsaicinoid In all the species analyzed both the capsaicin and the dihydrocapsaicin (DHC) contributed to the pun­ gency of the fruits. As expected (Stoica et al., 2016), the capsaicin content was higher than the DHC in all species and C. chacoense resulted in the higher con­ tent of about 300 µg/g of fresh weight (FW) com­ pared to C. baccatum whilst C. annum have been the most variable samples with an high standard devia­ tion that made it not statistical different from the other two species (Fig. 1). Interestingly the C. cha‐ coense resulted in the higher content of DHC of 243±36 µg/g FW, followed by the C. annum and C. baccatum that did not resulted statistical different Fig. 1­ Capsaicin and dihydrocapsaicin (DHC) content obtained from ripe fresh pepper fruits for each species analyzed, using HPLC detection (n=5; SD). Different letters repre­ sent statistical significance (ANOVA, p<0.05). Table 4 ­ Confusion Matrix derived from the Random tree analy­ sis from the aromatic profiles of the pepper species obtained using a decision tree learning tool C. baccatum C. chacoense C. annuum C. baccatum 9 1 0 C. chacoense 0 10 0 C. annuum 0 0 10 Comparini et al. ‐ Wild vs domesticated peppers 313 from each other with respectively 73±11 µg/g FW and 40±11 µg/g FW. The capsaicin content influences more the spiciness followed by the DHC, therefore C. baccatum have showed the lowest level of spiciness. PTR‐ToF‐MS volatiles compounds analysis The volatiles compounds analysis by SRI­ToF­MS revealed the volatile profile of each pepper species. The extraction of every single peak permitted the detection and the quantification of specific signals resultant of the protonation of numerous VOCs (m/z range = 20­210). From the volatile fraction composi­ tion of different pepper species, were detected many volatile compounds specifically alcohols, aldehydes, esters, ketones, hydrocarbons and terpenes com­ pounds. For each peak identified using H3O + and NO+ as ion reagents, all of the m/z detected have been assigned to the mass formulas reported respectively in Table 2 and 3 and expressed in ncps higher than 1 (n=10; ±SD). In particular, by trusting the high accura­ cy and resolution of this tool, the chemical com­ pounds have been tentatively proposed and matched with the existing documentation of VOCs in literature and by the acknowledged VOCs emitted by peppers. Thus, the identification of the compounds has been further improved by the use of two different reagents ions. All the peaks obtained have been fil­ tered and 51 mass spectral peaks have been detect­ ed when H3O + was the reagent ion and 18 using NO+ as reagent ion. Remarkably even if the pepper analyzed derived from different species (in particular Capsicum Chacoense), all the peaks identified using H3O + or NO+ as reagent ions were always present in the three species of chili pepper included in this study. Moreover, since the aroma is linked to the species and varieties (Taiti et al., 2019) the differences of VOCs emission among these three hot peppers were expected (Table 2 and 3). Above all, by using H3O +, C. annuum showed the highest total VOCs emission (22,116 ncps) followed by C. chacoense (13,214 ncps) and C. baccatum (4,850 ncps) as reported in Table 2. In particular, C. annuum (“Ciliegino” var.) seems to have a richest volatile profile compared to the other ones and showed the highest VOCs intensity for the compounds linked to the herbaceous notes. This trend could be confirmed by the high intensity of compounds detected at m/z 81.069, 83.086, 85.101, 99.080 and 101.096 all identified as C6 compounds (Table 2). Notably, peaks that confirmed this trend were detected at C4H9 + (measured at m/z=57.069) probably derived from Alkyl fragment (Hexanol/valeric acid), C5H7 + (measured m/z=67.055) and C6H9 + (measured m/z=81.069) likely correspond to Alkyl fragment (isoprene and terpenes or aldehy­ des fragments respectively), C6H13 + (measured m/z=85.101) probably refers to fragments of 1­ Hexanol and/or Nonanol, C5H11O2 + (measured m/z=103.075) refers as 3­Methylbutanoic acid and C15H25 + (measured m/z=205.195) attributed to Sesquiterpene compounds. On the contrary, the VOCs emission observed in the wild accession (C. chacoense) is characterized by the high emission of signals detected at m/z 33.033, 45.033 and 137.132, with the first two compounds which are linked to the ripening process whilst the last compound belongs to the terpene class. Moreover, it is interesting to note that the signal intensity of monoterpenes (m/z 137.132) is higher in C. chacoense than that observed for sesquiterpenes (m/z 205.101), in contrast to the emissions of C. annum and C. baccatum (Fig. 2, Table 2). Regarding the wide chemical classes of terpenes, common compounds which strongly contributes to the aroma of fruits and vegetables (Rodríguez­ Burruezo et al., 2010), in this study were identified several peaks belonging to monoterpenes (C­10) and sesquiterpenes (C­15) while no one was detected as oxygenated terpenes. This result is also confirmed by what was already reported (Rodríguez­Burruezo et al., 2010), where the oxygenated terpenes were found at only traces and mainly in C. chinense. Finally, the aroma of “Brasileiro” variety (C. bacca‐ tum) was characterized by a lower signals intensity, except for the compounds detected at m/z 117.091 (TI: Hexanoic acid/Hexanoates) and m/z 121.101 (TI: Terpenes fragment). Additional representative chem­ ical classes of the volatile fraction were aliphatic Fig. 2 ­ Example of schematic chart of mass peaks detected with H3O + ion as reagent for each capsicum species used in this paper. Signal intensities are given in normalize count per second (ncps) log­1 and higher than 1. Adv. Hort. Sci., 2021 35(3): 305­318 314 aldehydes, alcohols, and branched hydrocarbons. In particular, Hexanal, Hexanol, cis­2­Hexanal and cis­2­ Hexenol were the main compounds that contribute to the flavour note described as the odour of freshly cut grass or ground leaves, which are typically pro­ duced in fresh Capsicum fruits later on the tissue destruction (Ziino et al., 2009). On the contrary by using NO+ as donor ion, the differences of VOCs emis­ sion among all varieties studied were smaller as well as the number of peaks detected (Table 3). Sometimes these spikes were different, while some­ times were identical to those obtained using H3O + as a reagent ion and this behavior is usual and has been already reported elsewhere (Jordan et al., 2009). Indeed, even if we use NO+ as reagent ion in a com­ plex matrix, H3O + could occur in the ionization of the compounds present in the sample headspace (Jordan et al., 2009, Del Pulgar et al., 2013). Interesting, in contrast to what was observed with H3O+ , when using NO+ the Capsicum species C. baccatum showed for the majority of signals the highest intensity detected (Table 3). Moreover, Table 3 shows signifi­ cant differences in the concentration of many peaks between the three pepper species, especially as far as they are concerned with protonated aldehydes and ketones. Artificial neural networks Two neural networks have been built: the first one uses only VOCs emission, the second one com­ bines the pungency with capsaicin and DHC analysis as inputs layers. Both the ANN (Artificial Neural Network) were able to discriminate among the acces­ sion with 100% accuracy. Thus, the confusion matrix assigns all the aromatic profiles correctly to the relat­ ed pepper species (Table 4). According to the attribute selection filter applied, 12 VOCs profile and DHC content were the most discriminant parameters. In particular, the m/z signal of the VOCs and the ten­ tative identification are reported here below and are marked in Table 2 with the asterisk symbol: m/z 31.018 (Formaldehyde), m/z 57.033 (C3 aldehydes and ketones), m/z 59.049 (Propanal, Acetone), m/z 71.049 (2­Butenal), m/z 73.060 (Isobutanal/ Butanone), m/z 85.101 (1­Hexanol/Nonanol), m/z 99.080 (cis­3­Hexenal/(E)­2­Hexenal), m/z 115.111 (Heptanal), m/z 135.117 (p­Cymene), m/z 137.132 (Monoterpenes) and m/z 149.132 (Sesquiterpenes fragments). Random tree The data coming from the aromatic profiles of the pepper species were also analyzed using a decision tree learning tool. The identification was successful in 96.6% of the cases, and, as shown in the confusion matrix, Table 4, only one instance belonging to C. baccatum was incorrectly attributed to C. chacoense. The decision tree showing the two significant nodes is reported in figure 3. According to the tree, only two VOCs are fundamental to discriminate among the three species: m/z 81.068 and m/z 85.064. Protein profiles SDS­PAGE analyses performed on Capsicum species (Fig. 4) identified specific profiles linked to each sample. Gel images were then analyzed to gen­ erate data like a phylogenetic tree based on similarity comparison (Fig. 4) to graphically display relation­ ships among samples. Dendrograms results showed as two samples, C. chacoense and C. baccatum, clus­ terized independently from the third sample C. annu‐ um. Differences detected in the protein profiles could be associated with quantitative differences in the band densities as soon as some quantitative differ­ Fig. 3 ­ On the top the VOCs signal of the two most significant masses that help the identification of the species using a decision tree learning tool (n=10; SD). On the bottom, the two significant nodes have been reported. Comparini et al. ‐ Wild vs domesticated peppers 315 Fig. 4 ­ (A) SDS­PAGE analysis of proteins getting from Capsicum samples, where 1= C. chacoense; 2= C. annuum; 3= C. baccatum; M= molecular marker. (B) Phylogenetic tree resulting from the analysis of protein by SDS­PAGE analy­ sis. Image data were processed with Quantity One software (Bio­Rad Laboratories, Inc, Hercules CA) using Ward's method for clustering (Ward, 1963). ences. C. chacoense and C. baccatum showed a band (a) with a molecular weight slightly higher 20 kDa that were not detected in the C. annuum sample (Fig. 4). 4. Conclusions Our study underlined the high potential of the using the PTR­ SRI­MS to obtain a species­specific fin­ gerprinting of the volatile compounds emitted by the pepper fruits. This technology can help to highlight particular VOCs signals that are specific of species or specific growing conditions of chili pepper fruits with a rapid analysis without any pre­treatment of the samples. Indeed, the switching reagent ion system in PTR­MS instrumentation was applied for the first time to analyze pepper fruits by using not only H3O+ but also NO+ as precursor ions. This tool has permitted to find the VOCS able to discriminate among the species and by using two ionization agents a more accurate identification of the volatile compounds has been possible. In particular, the PTR­ ToF­MS analysis with NO+ as reagent ion has allowed (i) the detection of aldehydes and ketones in separat­ ed peaks, (ii) to detect some molecules not found using H3O +, (iii) to confirm the results obtained using H3O+ as reagent ion for VOC analysis. VOCs results were thus confirmed by protein analysis according to qualitative and quantitative differences, which turn out to be able to differentiate samples within the capsicum genus. Moreover, the multivariate statisti­ cal approach revealed that some of these com­ pounds can be successfully used for the species recognition in our artificial neural networks. The BPNN classifier utilized in the work had always 100% of success, both for H3O + and NO+ (data not shown) and this was probably due to the marked differences in the volatiles emissions of the three species. Further studies will aim to use the same method for a higher number of species, to challenge the analysis. Our investigation permitted the identification of 12 promising VOCs as more discriminant for each species and, among them, the masses m/z 81.068 and m/z 85.064 have been recognized as the most promising volatile markers of these species (Fig. 3). In addition to 12 VOCs, regarding the capsaicinoids, the DHC content was a more effective parameter to dis­ tinguish among domesticated and wild species of the genus Capsicum with respect to the capsaicin. The results from the random tree have been according with the protein analysis, which showed that the C. chacoense and C. baccatum have more similarities with respect to the C. annuum, although this was a preliminary analysis and additional analysis need to be done to support our hypothesis. According to our findings, both domesticated species, i.e. C. baccatum and C. annuum differ from the wild species C. cha‐ coense, in particular, the C. annuum resulted in being the most dissimilar from the wild species. This was probably due to a more strict genetic selection that the C. annuum faced during years of domestication and supports the hypothesis that C. chacoense is one of the most ancient species of the genus and that the C. baccatum and C. annuum evolved separately from this common predecessor. References AISPES, 2010 ­ The evolution of Capsicum genus. ­ I° Congress AISPES, International Association for the Study of Peppers and Solanaceae http://www. aispes.org. APREA E., ROMANO A., BETTA E., BIASIOLI F., CAPPELLIN L., FANTI M., GASPERI F., 2015 ­ Volatile compound changes during shelf life of dried Boletus edulis: com‐ parison between SPME‐GC‐MS and PTR‐ToF‐MS analy‐ sis. ­ J. Mass. Spectrom., 50(1): 56­64. AZCARATE C., BARRINGER S.A., 2010 ­ Effect of enzyme activity and frozen storage on Jalapeño pepper volatiles by selected ion flow tube‐mass spectrometry. ­ J. Food Sci., 75(9): 710­721. BARBOSA R., LUZ F., FILHO H., MADURO C., 2006 ­ Adv. Hort. Sci., 2021 35(3): 305­318 316 Pimentas de Roraima (Catálogo de Referência). ­ Manaus, INPA/EDUA, pp. 93. BARBOZA G.E., BIANCHETTI L.B. , 2005 ­ The new species of capsicum (Solanaceae) and a key to wild species from Brazil. ­ Syst. Bot., 30: 863­871. BLAKE R.S., MONKS P.S., ELLIS A.M., 2009 ­ Proton‐transfer reaction mass spectrometry. ­ Chem. Rev. 109(3): 861­ 896. BLAKE R.S., WYCHE K.P., ELLIS A.M., MONKS P.S., 2006 ­ Chemical ionization reaction time‐of‐flight mass spec‐ trometry: Multi‐reagent analysis for determination of trace gas composition. ­Int. J. Mass Spectrom., 254(1): 85­93. BOSLAND P.W., VOTAVA E.J., 2012 ­ Peppers: vegetable and spice capsicums. ­ Cabi Publishing, London, UK, pp. 248. BOUKEMA I.W., 1982 ­ Resistance to TMV in Capsicum chacoense in Capsicum L. ­ Euphytica, 29: 433­439. BUHR K., VAN RUTH S., DELAHUNTY C., 2002 ­ Analysis of volatile flavour compounds by Proton Transfer Reaction‐Mass Spectrometry: fragmentation patterns and discrimination between isobaric and isomeric com‐ pounds. ­ Int. J. Mass Spectrom., 221(1): 1­7. BUSO G.S.C., LOURENÇO R.T., BIANCHETTI L.B., DE LINS T.C.L., POZZOBON M.T., DE AMARAL Z.P.S., FERREIRA M.E., 2001 ­ Espécies silvestres do gênero Capsicum coletadas na Mata Atlântica Brasileira e sua relação genética com espécies cultivadas de pimenta: uma primeira abordagem genética utilizzando marcadores moleculares. ­ Embrapa Recursos Genéticos e Biotecnologia. Boletim de Pesquisa e Desenvolvimento, Brasília, Brazil, pp. 78. DAVENPORT L.J., 2004 ­ Genera Solanacearum: The genera of Solanaceae illustrated, arranged according to a new system by Armando T. Hunziker. ­ American Society of Plant Taxonomists, 29(1): 221­222. DEL PULGAR J.S., SOUKOULIS C., CARRAPISO A.I., CAPPELLIN L., GRANITTO P., APREA E., ROMANO A., GASPERI F., BIASIOLI F., 2013 ­ Effect of the pig rearing system on the final volatile profile of Iberian dry‐cured ham as detected by PTR‐ToF‐MS. ‐ Meat Sci., 93(3): 420­428. EDTBAUER A., HARTUNGEN E., JORDAN A., HANEL G., HERBIG J., JÜRSCHIK S., LANZA M., BREIEV K., MARK L., SULZER P., 2014 ­ Theory and practical examples of the quantification of CH4, CO, O2, and CO2 with an advanced proton‐transfer‐reaction/selective‐reagent‐ ionization instrument (PTR/SRI‐MS). ­ Int. J. Mass Spectrom., 365: 10­14. EGGINK P.M., MALIEPAARD C., TIKUNOV Y., HAANSTRA J.P.W., BOVY A.G., VISSER R.G.F., 2012 a ­ A taste of sweet pepper: Volatile and non‐volatile chemical com‐ position of fresh sweet pepper (Capsicum annuum) in relation to sensory evaluation of taste. ‐ Food Chem., 132(1): 301­310. EGGINK P.M., MALIEPAARD C., TIKUNOV Y., HAANSTRA J.P.W., POHU­FLAMENT L.M.M., DE WIT­MALJAARS S.C., WILLEBOORDSE­VOS F., BOS S., DE BENNING­ WAARD C., VAN GRAUW­LEEUWEN P.J., FREYMARK G., BOVY A.G., VISSER R.G.F., 2012 b ­ Prediction of sweet pepper (Capsicum annuum) flavor over different har‐ vests. ‐ Euphytica, 187(1): 117­131. GUZMÁN F.A., AZURDIA H.A.C., DUQUE M.C., VICENTE C.M., 2005 ­ AFLP assessment of genetic diversity of Capsicum genetic resources in Guatemala: home gar‐ dens as an option for conservation. ­ Crop. Sci., 45: 363­ 370. HEISER C.B., SMITH P.G., 1953 ­ The cultivated Capsicum spp. ­ Econ. Bot., 7: 214­227. HERBIG J., MÜLLER M., SCHALLHART S., TITZMANN T., GRAUS M., HANSEL A., 2009 ­ On‐line breath analysis with PTR‐TOF. ­ J. Breath Res., 3(2): 027004. HUNZIKER A.T., 2001 ­ Genera Solanacearum: the genera of Solanaceae illustrated, arranged according to a new system. ­ Gantner Verlag, Ruggell, Liechtenstein, pp. 516. INCE A.G., KARACA M., ONUS A.N., 2009 ­ Genetic relation‐ ships within and between Capsicum species. ­ Biochem. Genet., 48: 83­95. JORDAN A., HAIDACHER S., HANEL G., HARTUNGEN E., HERBIG J., MÄRK L., SCHOTTKOWSKY R., SEEHAUSER H., MÄRK T.D., 2009 ­ An online ultra‐high sensitivity proton‐transfer‐reaction mass‐spectrometer combined with switchable reagent ion capability (PTR + SRI−MS). ­ Int. J. Mass Spectrom., 286: 32­38. KIM S., KARL T., HELMIG D., DALY R., RASMUSSEN R., GUENTHER A., 2009 ­ Measurement of atmospheric sesquiterpenes by proton transfer reaction‐mass spec‐ trometry (PTR‐MS). ­ Atmos Meas. Tech., 2(1): 99­112. LEE A., GOLDSTEIN A.H., KROLL J.H., NGA N.L., VARUTBANGKUL V., FLAGAN R.C., SEINFELD J.H., 2006 ­ Gas‐phase products and secondary aerosol yields from the photooxidation of 16 different terpenes. ­ J. Geophys. Res. Atm., 111(D17). LONG­SOLIS, 1986 ­ Capsicum y cultura: La Historia del chilli. ­ Fondo de Cultural Economica, Mexico. MALEKNIA S.D., BELL T.L., ADAMS M.A., 2007 ­ PTR‐MS analysis of reference and plant‐emitted volatile organic compounds. ­ Int. J. Mass Spectrom., 262(3): 203­210. MANCUSO S., TAITI C., BAZIHIZINA N., COSTA C., MENESATTI P., GIAGNONI L., ARENELLA M., NANNIPIERI P., RENELLA G., 2015 ­ Soil volatile analysis by proton transfer reaction‐time of flight mass spec‐ trometry (PTR‐TOF‐MS). ­ Appl. Soil Ecol., 86: 182­191. MCLEOD M.J., GUTTMAN S.I., ESHBAUGH W.H., 1979 ­ A preliminary biochemical systematic study of the genus Capsicum­Solanaceae, pp. 701­714. ­ In: HAWKES J.G., R.N. LESTER, and A.D. SKELDING (ed.) The biology and taxonomy of the Solanaceae. Academic press, London, UK, pp. 738. MCLEOD M.J., GUTTMAN S.I., ESHBAUGH W.H., 1982 ­ Early evolution of chili peppers (capsicum). ­ Econ. Bot., Comparini et al. ‐ Wild vs domesticated peppers 317 36: 361­368. MCLEOD M.J., GUTTMAN S.I., ESHBAUGH W.H., RAYLE R.E., 1983 ­ An electrophoretic study of evolution in Capsicum Solanaceae. ­ Evolution, 37: 562­574. MOCHALSKI P., UNTERKOFLER K., ŠPANĚL P., SMITH D., AMANN A., 2014 ­ Product ion distributions for the reactions of NO+ with some physiologically significant aldehydes obtained using a SRI‐TOF‐MS instrument. ­ Int. J. Mass Spectrom., 363: 23­31. MOSCONE E.A., SCALDAFERRO M.A., GABRIELE M., CECCHINI N.M., SANCHEZ­GARCIA Y., JARRET R., DUCASSE D.A., BARBOZA G.E., EHRENDOFER F., 2007 ­ The evolution of Chili peppers (Capsicum ­ Solanaceae): a cytogenic perspective. ­ Acta Horticulturae, 745: 137­ 169. NEE M., BOHS L., KNAPP S., 2006 ­ New species of Solanum and Capsicum (Solanaceae) from Bolivia with clarifica‐ tion of nomenclature in some Bolivian Solanum. ­ Brittonia, 58: 322­356. ORTIZ R., CROSSA J., FRANCO J., SEVILLA R., BARGUENO J., 2008 ­ Classification of Peruvian highland maize races using plant traits. ­ Genet. Res. Crop Evol., 55: 151­162. PANDOLFI C., MUGNAI S., AZZARELLO E., BERGAMASCO S., MASI E., MANCUSO S., 2009 ­ Artificial neural networks as a tool for plant identification: a case study on Vietnamese tea accessions. ­ Euphytica, 166(3): 411­ 421. PERRY L., DICKAU R., ZARRILLO S., HOLST I., PEARSALL D., PIPERNO D., BERMAN M.J., COOKE R., RADEMAKER K., RANERE A., RAYMOND S., SANDWEISS D., SCARAMELLI F., TARBLE K., ZEIDLER J., 2007 ­ Starch fossils and the domestication and dispersal of Chili peppers (Capsicum spp. L.) in the Americas. ­ Science, 135: 966­998. PICKERSGILL B., 1969 ­ The archaeological record of chili peppers (Capsicum spp.) and the sequence of plant domestication in Peru. ­ Am. Antiqu., 23: 54­61. PICKERSGILL B., 1971 ­ Relationships between weedy and cultivated form in some species of chili peppers (genus Capsicum). ­ Evolution, 25: 683­691. PICKERSGILL B., 1984 ­ Migration of Chili Peppers, Capsicum spp., in the Americas, pp. 105­123. ­ In: STORE P. (ed.) Pré‐Columbian Plant Migration. Harvard University Press, Cambridge, UK, pp. 183. PICKERSGILL B., 1988 ­ The genus Capsicum: a multidiscipli‐ nary approach to the taxonomy of the cultivated and wild plants. ­ Biol. Zent., 107: 381­389. POZZOBON M.T., SCHIFINO­WITTMAN M.T., BIANCHETTI L.B., 2006 ­ Chromosome numbers in wild and semido‐ mesticated Brazilian Capsicum L. (Solanaceae) species: do x = 12 and x = 13 represent two evolutionary lines?. ­ Bot. J. Linn. Soc., 151: 259­269. RODRÍGUEZ­BURRUEZO A., KOLLMANNSBERGER H., GONZÁLEZ­MAS M.C., NITZ S., FERNANDO N., 2010 ­ HS‐SPME comparative analysis of genotypic diversity in the volatile fraction and aroma‐contributing com‐ pounds of Capsicum fruits from the Annuum, Chinense, Frutescens complex. ­ J. Agric. Food Chem., 58(7): 4388­ 4400. RUSSO V.M., 2012 ­ Peppers ‐ botany, production and uses. ­ Cabi publishing, Wallingford, UK, pp. 280. STOICA R.M., MOSCOVICI M., TOMULESCU C., BĂBEANU N., 2016 ­ Extraction and analytical methods of capsai‐ cinoids‐a review. ‐ Scientific Bulletin. Series F. Biotechnologies, 20: 93­98. SUDRÉ C.P., CRUZ C.D., RODRIGUES R., RIVA E.M., AMARAL Jr. A.T., SILVA D.J.H., PEREIRA T.N.S., 2006 ­ Variáveis multicategóricas na determinação da divergência gené‐ tica entre acessos de pimenta e pimentão. ­ Hortic. Bras., 24: 88­93. SUDRÉ C.P., GONÇALVES L.S.A, RODRIGUES R., AMARAL Jr. A.T., RIVA­SOUZA R.M., BENTO C.S., 2010 ­ Genetic variability in domesticated Capsicum spp. as assessed by morphological and agronomic data in mixed statisti‐ cal analysis. ­ Genet. Mol. Res., 9(1): 283­294. TAITI C., COSTA C., MENESATTI P., COMPARINI D., BAZIHIZINA N., AZZARELLO E., MASI E., MANCUSO S., 2015 ­ Class‐modeling approach to PTR‐TOF‐MS data: a peppers case study. ­ J. Sci. Food Agric., 95(8): 1757­ 1763. TAITI C., COSTA C., MIGLIORI C.A., COMPARINI D., FIGORILLI S., MANCUSO S., 2019 ­ Correlation between volatile compounds and spiciness in domesticated and wild fresh chili peppers. ­ Food and Bioprocess Technology, 12(8): 1366­1380. TAITI C., COSTA C., NISSIM W.G., BIBBIANI S., AZZARELLO E., MASI E., PANDOLFI C., PALLOTTINO F., MENESATTI P., MANCUSO S., 2017 ­ Assessing VOC emission by dif‐ ferent wood cores using the PTR‐ToF‐MS technology. ­ Wood Sci. Technol., 51(2): 273­295. TANI A., 2013 ­ Fragmentation and reaction rate constants of terpenoids determined by proton transfer reaction‐ mass spectrometry. ­ Environ. Control Biol., 51(1): 23­ 29. VITA F., LUCAROTTI V., ALPI E., BALESTRINI R., MELLO A., BACHI A., ALESSIO M., ALPI A., 2013 ­ Proteins from Tuber magnatum Pico fruiting bodies naturally grown in different areas of Italy. ­ Proteome Sci., 11(1): 7. WANG T., ŠPANĚL P., SMITH D., 2004 ‐ A selected ion flow tube study of the reactions of H3O +, NO+ and O2 +• with some phenols, phenyl alcohols and cyclic carbonyl com‐ pounds in support of SIFT‐MS and PTR‐MS. ­ Int. J. Mass Spectrom., 239(2): 139­146. WARD Jr. J.H., 1963 ­ Hierarchical grouping to optimize an objective function. ­ J. Am. Stat. Assoc., 58: 236­244. ZIINO M., CONDURSO C., ROMEO V., TRIPODI G., VERZERA A., 2009 ­ Volatile compounds and capsaicinoid content of fresh hot peppers (Capsicum annuum L.) of different Calabrian varieties. ­ J. Sci. Food Agric., 89(5): 774­780. ZIMMERMANN M., SCHIEBERLE P., 2000 ­ Important odor‐ ants of sweet bell pepper powder (Capsicum annuum cv. annuum): differences between samples of Hungarian and Morrocan origin. ­ Eur. Food Res. Adv. Hort. Sci., 2021 35(3): 305­318 318 Technol., 211(3): 175­180. ZURADA J.M., 1992 ­ Introduction to artificial neural sys‐ tems. ­ West publishing company, Boston, USA, pp. 790. ZURADA J.M., MALINOWSKI A., 1994 ­ Multilayer percep‐ tron networks: selected aspects of training optimiza‐ tion. ­ Appl. Math. Comp. Sci., 4: 281­307.