Volatile Compounds in Citrus Essential Oils: A Comprehensive Review

The essential oil fraction obtained from the rind of Citrus spp. is rich in chemical compounds of interest for the food and perfume industries, and therefore has been extensively studied during the last decades. In this manuscript, we provide a comprehensive review of the volatile composition of this oil fraction and rind extracts for the 10 most studied Citrus species: C. sinensis (sweet orange), C. reticulata (mandarin), C. paradisi (grapefruit), C. grandis (pummelo), C. limon (lemon), C. medica (citron), C. aurantifolia (lime), C. aurantium (bitter orange), C. bergamia (bergamot orange), and C. junos (yuzu). Forty-nine volatile organic compounds have been reported in all 10 species, most of them terpenoid (90%), although about half of the volatile compounds identified in Citrus peel are non-terpenoid. Over 400 volatiles of different chemical nature have been exclusively described in only one of these species and some of them could be useful as species biomarkers. A hierarchical cluster analysis based on volatile composition arranges these Citrus species in three clusters which essentially mirrors those obtained with genetic information. The first cluster is comprised by C. reticulata, C. grandis, C. sinensis, C. paradisi and C. aurantium, and is mainly characterized by the presence of a larger abundance of non-terpenoid ester and aldehyde compounds than in the other species reviewed. The second cluster is comprised by C. junos, C. medica, C. aurantifolia, and C. bergamia, and is characterized by the prevalence of mono- and sesquiterpene hydrocarbons. Finally, C. limon shows a particular volatile profile with some sulfur monoterpenoids and non-terpenoid esters and aldehydes as part of its main differential peculiarities. A systematic description of the rind volatile composition in each of the species is provided together with a general comparison with those in leaves and blossoms. Additionally, the most widely used techniques for the extraction and analysis of volatile Citrus compounds are also described.


INTRODUCTION
Citrus essential oils, widely used to the procurance of natural fruity perfumes and as flavoring ingredients in food, pharmaceutical and cosmetic products (Tranchida et al., 2012;Palazzolo et al., 2013;Othman et al., 2017), are obtained mainly from the fruit rind (flavedo), although flowers or leaves have also been used. Volatile and semi-volatile compounds represent 85-99% of the entire oil fraction Tranchida et al., 2012;Sarrou et al., 2013), which is represented typically by over 200 compounds. Hydrocarbon and derivate mono-and sesquiterpenes are the compounds most frequently reported, followed by aliphatic and olefinic C 6 -C 12 non-terpene aldehydes, alcohols, ketones, esters and acids, along with several aromatic compounds. The non-volatile residue is mainly composed of flavonoids, coumarins, diterpenoids, sterols, and fatty acids.
Additionally, several other extraction methods not involving the use of water or organic solvents have been employed. This is the case of water-free microwave extraction (Ferhat et al., 2007;Uysal et al., 2011); or supercritical CO 2 fluid extraction (Akakabe et al., 2010;Dong et al., 2014;Sun et al., 2014a), a technique which extracts more efficiently higher MW compounds, such as squalene (Benelli et al., 2010) or linoleic acid (Dong et al., 2014). Headspace techniques have also been often used to extract volatile compounds, such as headspace extraction using a gas-washing bottle and a Porapak-Q sorbent tube (Elmaci and Onogur, 2012), and most frequently HS-SPME. HS-SPME has been used to extract volatiles and semi-volatile compounds of flowers and leaves, predominantly (Flamini et al., 2007;Jabalpurwala et al., 2009;Lin et al., 2010;Cheong et al., 2011b). When using this technique, the plant material is introduced in a septum-capped vial and after an equilibration period, a pre-conditioned fiber is exposed to the headspace of the vial for several minutes either at room temperature or at higher temperatures. There are currently several types of fibers suitable for adsorbing Citrus VOCs. A 100 µm PDMS-coated fiber was used to analyze the aroma of C. unshiu Marcov. fresh and dry blossoms (Choi, 2003a), C. limon flower organs and pollen (Flamini et al., 2007; and Taiwan Citrus leaves (Lin et al., 2010). In other studies, a 75 µm/85 µm CAR/PDMS-coated fiber was chosen to extract the VOCs from Citrus flowers Cheong et al., 2011b;Chung, 2012). Also, an additional two fibers have been used for Citrus peel oil. One fiber was a 65 µm DVB/PDMS coated fiber, which was used to analyze C. clementine EOs (González-Mas et al., 2015). For this analysis the oil was diluted 1:100 with dichloromethane and 10 µL were introduced in the septum-cap vial for HS-SPME with 990 µL of milli-Q water. Choi and Min (2004) also used this last fiber to analyze the VOCs in fresh flavedo of Hallabong Citrus. The other fiber used for Citrus oil and fresh flavedo was a 50 µm/30 µm DVB/CAR/PDMS fiber (Qiao et al., 2008;Mitropoulou et al., 2017), which is more polar than the other named fibers, and often used to extract volatile and semi-volatile compounds from the Citrus juices (González-Mas et al., 2011;Benjamin et al., 2013;Rambla et al., 2014;Qiu and Wang, 2015). Although recently it has also been used to analyze VOCs in Citrus flowers . According to the results reported in our literature review, the SPME-fibers used most often to analyze Citrus flower and leaf volatiles seems be the minor polar 100 µm PDMS or 75 µm/85 µm CAR/PDMS fibers, while the 65 µm DVB/PDMS fiber is more used for Citrus essential oil.

TECHNIQUES USED TO ANALYZE Citrus VOLATILE COMPOUNDS
The technique most widely used for the analysis of volatile and semi-volatile compounds in Citrus species currently is gas chromatography coupled to mass spectrometry (GC-MS). The principal option used to introduce the oil in the chromatograph is by direct injection of a dilution in organic solvent: most often pentane (Verzera et al., 2003;Boussaada and Chemli, 2006;Jazet Dongmo et al., 2009;Hosni et al., 2010) or hexane (Alonzo et al., 2000;Ahmed et al., 2001;Mondello et al., 2003;Dugo et al., 2005;Hamdan et al., 2010;Wang and Liu, 2014), but also dichloromethane (Blanco Tirado et al., 1995;Frizzo et al., 2004), diethyl ether (Merle et al., 2004) or even acetone (Viuda-Martos et al., 2009). Sun et al. (2014b) diluted the C. maxima EO in two steps: by 5,000-fold dilution in ethanol for analysis of the most abundant volatiles and 100-fold dilution for the minor volatiles. However, in other studies, the oil is not diluted (Mitiku et al., 2000;Song et al., 2000b;Choi et al., 2001;Minh Tu et al., 2002b, 2003bChoi, 2005;Njoroge et al., 2005aNjoroge et al., ,b, 2008Sawamura et al., 2005;Akakabe et al., 2010;Cheong et al., 2011b;Omori et al., 2011). When volatile and semi-volatile compounds are extracted with an organic solvent, the samples are concentrated before direct injection using a rotary evaporator and further reduced under a stream of N 2 (Hamdan et al., 2010;Miyazato et al., 2012). However, during these concentration steps, highly volatile compounds tend to be lost (Kerry et al., 2002). When VOCs are extracted with a fiber (HS-SPME), they are transferred directly to the injection port of the GC-MS system (Flamini et al., 2007;Jabalpurwala et al., 2009).
Different GC capillary columns have been used to analyze the volatile and semi-volatile Citrus compounds. Supplementary Table S1 in the Supporting Information shows the most often GC columns reported since 2010. Additionally, multidimensional gas chromatography consisting on the use of two columns in tandem with different separation characteristics (typically one polar and the other non-polar) has also been often used in the last decade, in order to achieve a higher chromatographic resolution of the diverse compounds comprising complex matrices such as EOs (Tranchida et al., 2012;Cannon et al., 2015).
There also exist more complex techniques based on carbon isotope ratios which allow the authentication of Citrus VOCs for the control of adulteration of Citrus essential oils. Thus, the employment of GC hyphenated to carbon isotope ratio mass spectrometry (GC-C-IRMS) can detect compounds of different botanical origin such as, for example, the presence of citral extracted from lemon grass (Cymbopogon citratus) (Schipilliti et al., 2018) in lemon EO. Moreover, this technique provides the authenticity assessment to differentiate natural and synthetic compounds in different Citrus essential oils, to discriminate Citrus peel oils of different geographical origin. Also, GC-C-IRMS is useful to eliminate the environmental influences and to differentiate genuine Citrus essential oils belonging to different species. However, this technique presents limitations related to chromatographic resolution due to instrumentation engineering. It will be necessary the optimization of the multidimensional GC separations to more effectively establish the ranges of authenticity of Citrus essential oils.

VOLATILE COMPOUNDS IDENTIFIED IN ESSENTIAL OILS FROM PEEL, LEAVES, AND FLOWERS OF Citrus SPECIES
We have listed in this review over one thousand volatile and semivolatile compounds that have been reported during the last two decades in rinds, leaves or flowers (EOs and solvent extracts) of Citrus species cultivated worldwide, although some early studies have also been included (Supplementary Table S2). In this review we have focused on the composition of C. reticulata, C. grandis (also named C. maxima), C. sinensis, C. paradisi, C. aurantium, C. junos, C. medica, C. aurantifolia, C. bergamia, and C. limon. This study also allowed us to establish which species are more similar or distant to each other according to the rind composition of their volatile and semi-volatile fractions as described in the literature. Finally, the volatile pattern of Citrus rinds has been compared to those of flowers and leaves.
Our literature review on VOCs in Citrus species corroborates some commonalities for the 10 most widely used and analyzed Citrus species. Thus, the EOs obtained from Citrus rinds always show limonene, a hydrocarbon monoterpene, as the most abundant compound, its concentration generally representing about 60-95% of the oil (Jing et al., 2014). However, limonene can show lower levels, as in C. bergamia, in which it can decrease down to 30% (Jing et al., 2014), or in C. limon, where it can decrease down to 48% (Lota et al., 2002). The following compounds in abundance are also monoterpenes, usually representing less than 15% (Supplementary Table S2), although γ-terpinene (Fanciullino et al., 2006) and linalyl acetate (Verzera et al., 2003) can reach an abundance of 23% and 36%, respectively. Non-terpenoid compounds very rarely represent more than 1%, although this does not necessarily mean that these compounds do not have an impact on the EO aroma. On the other hand, Citrus flowers and leaves show a different volatile profile. In these organs, limonene is not always the major volatile compound. Depending on the species, other monoterpenes such as linalool, β-myrcene or β-citronellol, in the case of flowers (Boussaada and Chemli, 2006;Jabalpurwala et al., 2009;Sarrou et al., 2013), or sabinene, in the case of leaves (Tomi et al., 2008), may be the most prominent compounds.
Attending to the number of compounds identified, sesquiterpene hydrocarbons are the most diverse group in most of the species. Monoterpene hydrocarbons and oxygenated monoterpene alcohols also tend to be among the most copious groups (Figure 1). Nevertheless, in some species such as C. reticulata or C. limon, the most numerous group corresponds to non-terpenoid aldehydes (Figure 2).
It became obvious when performing this review that the nomenclature of Citrus VOCs compounds, especially for monoand sesquiterpene derivates needs standardization, as in its present form it can lead to misinterpretation of the results. Compounds are named according to IUPAC in some papers, but this nomenclature is very complex and results in very long names (Yu et al., 2009). Most of the papers use shorter non-systematic names, but this results in many synonyms for the same compound. An example of this is p-cymenene (Lota et al., 2002;Fanciullino et al., 2006;Pino et al., 2006), also named α-p-dimethylstyrene (Huang et al., 2000;Jabalpurwala et al., 2009;Cannon et al., 2015) and dehydro-4-cymene (Tomiyama et al., 2012). Similarly, (Z)-limonene oxide (Minh Tu et al., 2003b;Sun et al., 2014b) is also named (Z)epoxylimonene (Blanco Tirado et al., 1995), (Z)-limonene,1,2epoxide (Lota et al., 1999;Omori et al., 2011) or 1,2-epoxylimonene (Casilli et al., 2014). Monoterpene compounds are named in many occasions according to the structure of p-menthane (Supplementary Figure S1A). This is the case of pseudolimonene (Lan Phi et al., 2009;Sun et al., 2014b), α-terpineol (Lota et al., 2002;Cheong et al., 2011b;Liu et al., 2012;Sun et al., 2014a), α-terpinyl acetate (Lota et al., 2002) (also α-terpenyl acetate) (Verzera et al., 2004), 4-terpinenol Liu et al., 2012) and dihydrocarveol (Njoroge et al., 2005a) (also dehydrocarveol) (Shen et al., 2002), which have also been named as p-mentha-1(7),8-diene FIGURE 1 | Relative frequency of each group of compounds present in the 10 most common Citrus species. Left, all the compounds described; right outer circle, compounds up to date detected in only one species (one-species only); right inner circle, compounds detected in all 10 species (common). Orange bars indicate separation between mono-, sesqui-, and other terpenoids, and non-terpenoid compounds. The qualitative data used is available in Supplementary Table S4. FIGURE 2 | Heatmap showing the relative frequency of each group of compounds is presented for each of the most common Citrus species. Data represent the number of compounds reported in each category divided by the total number of compounds reported in each particular species, according to the scale below: red, high frequency; black, intermediate frequency; green, low frequency. The qualitative data used is available in Supplementary Table S5. (Tomiyama et al., 2012), p-menth-1-en-8-ol (Chen et al., 2014), p-menth-1-en-8-yl acetate (Kirbaslar et al., 2001), 1-p-menthen-4-ol (Delort and Jaquier, 2009), and p-menth-8-en-2-ol (Delort et al., 2015), respectively. It seems necessary to reach a scientific agreement to name these compounds with a single criterion or, at the least, to indicate the several common synonyms for each compound in related articles. Along this review we have used one of the most common names for each compound, but all the different terms used in the scientific literature have been kept in Supplementary Table S2.

Volatile Composition of Peel Extracts and Peel Essential Oil
In this study, we have particularly focussed on the presence of volatile compounds in the peel of the 10 Citrus species most commonly analyzed, for which a significant amount of information is available in the scientific literature. Depending on each particular study, compounds have been identified with a different degree of confidence. We have systematically classified the reliability of identification for each compound in each of the revised articles in a scale from A+++ (highest confidence) to H (lowest confidence), as detailed in Supplementary Table S2. We considered it a key feature in this review, as our own experience in Citrus juice volatile profiling is that misidentification of volatile compounds is far to be rare in the scientific literature, particularly for those tentatively identified based on retention indices (González-Mas et al., 2011). Therefore, we have used a conservative approach all along this review for the description of the volatile profiles, as well as for the generation of figures and statistical analyses. We have selected only the 763 volatile and semi-volatile compounds which have been identified by, at least, both retention index and mass spectrum (reliability classified from A+++ to B in Supplementary Table S2), while discarding those identified with a lower degree of confidence. We used this approach to perform a qualitative description (described/never described) of the volatile profiles of these species, summarized in Figures 1, 2, and to elaborate the HCA shown in Figure 3. Applying this criterion, we observed that slightly more than half of these 763 reliably identified compounds are terpenoids, whilst the other half corresponds to a wide range of non-terpenoid compounds. Oxygenated monoterpenes is the most numerous group, comprising almost a quarter of all the compounds identified. Forty-nine of these volatile compounds have been described in all 10 species (Table 1 and Supplementary  Figures S1B-H), almost 90% of which are terpenoid. Therefore, in some way they could be considered as those defining the characteristic Citrus volatile profile. Interestingly, most of these shared compounds are monoterpenoids (15 monoterpene hydrocarbons; 17 oxygenated monoterpenes) or sesquiterpenoids (9 sesquiterpene hydrocarbons; 2 oxygenated sesquiterpenes). The only non-terpenoid common volatiles are 1-octanol and the five C 8 to C 12 linear aliphatic aldehydes from octanal to dodecanal. On the other hand, 410 compounds of ample chemical diversity have been described up to date in only one of these species, most often in only one study (Figure 1 and Supplementary Table S2). It cannot be discarded that in some cases this might be due to misidentification, particularly when compound identification was not confirmed with a pure standard. Nevertheless, when unambiguously identified, they could be considered as biomarkers for each particular species.
As previously described, several different -and somewhat complementary-analytical techniques have been used for the identification of the VOCs in the EOs along the scientific literature. In the case of these 10 species we considered that the number of studies already published is sufficient to rely on the total list of compounds identified as quite representative of their actual volatile profile. Therefore, we addressed the classification of these species by HCA based on their respective volatile profiles as described in the literature. This approach allowed us to classify these species into three clusters based on their volatile profile as described in the literature (Figure 3) . The first cluster is comprised by C. reticulata, TABLE 1 | Common volatile compounds identified in C. sinensis, C. reticulata, C. paradisi, C. grandis, C. aurantium, C. limon, C. medica, C. aurantifolia, C. bergamia, and C. junos.
However, a recent study about the origin and evolution of Citrus based on the total genome shows higher similarities with the volatile profiles here summarized. Our HCA clustered C. sinensis, C. paradisi and C. aurantium with C. reticulata and C. grandis (C. maxima). In fact, Wu et al. (2018) showed that the first three species are actually genetical hybrids and admixtures of the latter two. This genomic study also revealed that C. aurantifolia is a hybrid of C. medica. In accordance with FIGURE 3 | Hierarchical clustering of the 10 most common Citrus species according to their volatile profile as described in the scientific literature. For each species, data were normalized by dividing the number of times a particular compound was reported by the total number of reports of all the compounds in that species. The statistical parameters used were as described in Rambla et al. (2015). The qualitative data used to create HCA is available in Supplementary Table S4. this, our HCA based on the volatile profile of the fruit peel also clustered them together.
In the following sections, the volatile and semi-volatile profiles of the peel are presented independently, describing the peculiarities of each species. Species are presented following the results of the HCA, which provides information about the similarities and differences between them. The compounds up to date described in only one species have been highlighted (Supplementary Table S3), as well as those that allowed to establish correlations between species. In addition, the compounds quantitatively most abundant in each species have been indicated. Finally, the compounds frequently isolated in most Citrus but never reported in a particular species have also been mentioned here.
The volatile profile most apart from C. grandis is that of C. limon (Figure 3). However, provided that C. limon has been widely studied by Cannon et al. (2015), several compounds have been exclusively isolated in C. limon and C. grandis peels, specially some non-terpenoid aliphatic ester compounds such as (Z)-3-hexenyl hexanoate (Sun et al., 2014a;Cannon et al., 2015) and methyl octanoate, methyl nonanoate and methyl decanoate Cannon et al., 2015). Moreover, for this same reason, there are compounds only isolated in C. grandis, and C. limon and also C. reticulata or C. sinensis or C. paradisi such as (E)-piperitol [C. grandis (Sun et al., 2014a;Liu et al., 2015), C. reticulata , and C. limon (Cannon et al., 2015)].
One of the most relevant features of C. grandis volatile profile is the absence of some compounds that are typical of other common species of Citrus, mainly some sesquiterpene compounds. This is the case of the sesquiterpene (E)α-bergamotene, which has never been reported in C. grandis peel but frequently identified in all the other common Citrus species studied (Mondello et al., 2003;Lan Phi et al., 2009;Aliberti et al., 2016;Petretto et al., 2016;Zhang et al., 2017). As far as we know, neither the compound thymol has been reported in C. grandis nor C. paradisi nor C. bergamia peel and neither (E)-2-decenal had been reported in C. grandis nor C. aurantifolia, but in the rest of the species reviewed.
Finally, it is important to indicate that when Citrus essential oils are exposed to sunlight, UV radiation or the air, the volatile profile can be modified due to chemical reactions. It has been reported in C. grandis oil that the concentration of several compounds, especially aldehydes such as decanal, dodecanal, neral or geranial, drops dramatically when exposed to these conditions. Instead, other compounds increase their concentration or appear novelly such as β-citronellal, α-humulene, linalool oxides, limonene oxides, (E)-carveol, (Z)carveol, perilla alcohol, carvone, α-pinene oxide, γ-elemene or caryophyllene oxide (Supplementary Table S2) (Sun et al., 2014b;Li et al., 2016). This aldehyde transformation in C. grandis promoted an increase of strong oily notes and a decrease of fresh notes. Recently, sensory evaluation, GC-MS and GC-O demonstrated the pummelo rind oil co-treated by both oxygen and heating (70 • C O 2 ) showed significant differences in both volatile composition and aromatic profile from those of the corresponding fresh, nitrogen-protected (25 • C N 2 ), heated (70 • C N 2 ) and oxygen-exposed (25 • C O 2 ) oils (Sun et al., 2018). This study indicates that oxygen and heating co-treatment increases the content of limonene oxides and carvone in pummelo EO. The first compounds enhanced the sweet and floral notes, while the latter promoted the intensity of the minty note. These results indicate the importance of a correct production and storage of Citrus essential oil to preserve its original freshness quality.

C. sinensis
The volatile composition of C. sinensis peel EO is the most studied of all Citrus species along with those of C. reticulata and C. limon (Supplementary Table S2). This species seems to be richer in the diversity of sesquiterpene hydrocarbons than other species as C. reticulata or C. limon. An example of these compounds is aromadendrene (Njoroge et al., 2005a;Hosni et al., 2010) or sesquiphellandrene (Ruberto and Rapisarda, 2002;Sawamura et al., 2005), although it should not be forgotten that many of these sesquiterpenes are identified ambiguously, because there are no commercial standards or are unattainable for their price for many scientific studies. Also, in this species have been reported a relative large group of monoterpenoid compounds, being some of them almost exclusive compounds as limonene diepoxide (Högnadóttir and Rouseff, 2003;Njoroge et al., 2005a;Sawamura et al., 2005;Esquivel-Ferriño et al., 2014).
However, these distinctive VOCs of C. sinensis are minor compounds with a concentration usually lower than 0.1%. The major C. sinensis VOCs are very similar to that of C. reticulata, with comparable proportions. Limonene is usually reported between 90 and 97% in C. sinensis, although this percentage decreased down to 64% in some studies (Chen et al., 2014). Quantitatively, the most important difference with C. reticulata is for α-sinensal, since in C. reticulata this compound can reach a percentage above 0.7% (Lota et al., 2001b), while in C. sinensis its percentage does not rise of 0.05% (Njoroge et al., 2005a;Sawamura et al., 2005).

C. paradisi
The volatile profile from this species is very close to those of C. aurantium and C. sinensis, and a little less so to C. grandis and C. reticulata (Figure 3), and further apart from the other species studied. As in C. sinensis, numerous sesquiterpene hydrocarbons in have been reported in C. paradisi (Figure 2 and Supplementary Table S2), such as the sesquiterpene α-cedrene, isolated in C. paradisi (Njoroge et al., 2005b), C. grandis (Njoroge et al., 2005b), C. sinensis (Njoroge et al., 2005a;Sawamura et al., 2005), C. aurantifolia (Minh Tu et al., 2002a), and C. medica (Wu et al., 2013).
The major compounds in C. paradisi are very similar to those in C. grandis, often with lower levels of γ-terpinene and linalool to those with C. sinensis and C. reticulata (Njoroge et al., 2005b;Petretto et al., 2016), and with higher levels of nootkatone, which can reach 0.2% in C. paradisi (Njoroge et al., 2005b) and 177.02-5.05 µg/g in C. grandis (Zhang et al., 2017), while in the other two varieties they usually appear in lower percentages (Supplementary Table S2) (Ruberto et al., 1997;Zhang et al., 2017).
Some VOCS have been exclusively described in Citrus aurantium, principally monoterpene compounds, hydrocarbons such as α-ocimene (Baik et al., 2008) and oxygenated compounds such as α-terpinen-4-ol acetate (Pino and Rosado, 2000). The set of species most different to C. aurantium include C. junos, C. bergamia, C. medica, C. aurantifolia and especially C. limon, although some of them have several compounds in common with C. aurantium and some species in other clusters. This is the case of thymol, reported in C. aurantium, C. reticulata, C. sinensis, C. aurantifolia, C. medica, C. junos and C. limon (Supplementary Table S2) (Miyazawa et al., 2009;Aliberti et al., 2016;Zhang et al., 2017).
Most of the C. junos characteristic compounds previously mentioned are compounds that have been reported at trace levels, with percentages lower than 0.02%. In this species the major compounds are also very similar to those described in C. sinensis or C. reticulata. Thus, the most abundant compound remains limonene, although in concentrations between 63 and 77%, lower than usually those of the species closest to C. reticulata (Song et al., 2000b;Lan Phi et al., 2009). This decrease is associated to the increase of some monoterpene and sesquiterpene hydrocarbons, such as α-phellandrene, β-phellandrene, terpinolene or bicyclogermacrene (Song et al., 2000b;Lan Phi et al., 2009;Tomiyama et al., 2012).
It is important to emphasize that some VOCs frequently isolated in the other common Citrus species were not found in C. junos peel, such as geranial, isolated in the peel of all the other nine Citrus species, terpenyl acetate, neryl acetate, eucalyptol (also missing in C. medica peel), tetradecanal, β-bisabolene or decyl acetate (the latter two compounds found in the juice but not in the peel of C. junos) (Tomiyama et al., 2012).

C. medica
According to our review, the volatile profile most similar to that of C. medica are those of C. aurantifolia and C. bergamia, whilst C. junos also clusters together in terms of volatile profile similarity (Figure 3). As in C. bergamia, C. aurantifolia and C. junos peel, and also in C. sinensis and C. paradisi, many sesquiterpene hydrocarbons have been described in C. medica (Figure 2). The recent study realized by Aliberti et al. (2016) has reported some compounds in C. medica not described, or very rarely identified, in other Citrus species so far as like the monoterpenes cuminyl alcohol, dehydrosabina ketone or cis-4-caranone, and specially sesquiterpene like longifolene, 9-epicaryophyllene, α-cuprenene, γ-cuprenene, italicene, nootkatol, or β-oplopenone, among many others. However, it is important to note that most of these compounds were tentatively identified, because a standard was not used for their identification, with some exception like the monoterpene trans-4-caranone and the sesquiterpenes longifolene or 9-epi-caryophyllene.
As in all the revised species, the most abundant compound in C. medica is limonene, but its percentage can decrease down to 51% (Verzera et al., 2005), while other monoterpene compounds such as γ-terpinene, β-pinene or camphene are present at higher concentration, in comparison with species of C. reticulata cluster where these compounds are usually described in percentages below 1%. Thus, in C. medica oil, γ-terpinene, β-pinene or camphene can reach percentage of 31%, 9.7%, and 10%, respectively (Lota et al., 1999;Aliberti et al., 2016;Petretto et al., 2016). Also C. medica presents higher concentrations of some sesquiterpenes, as is the case of (E)-α-bergamotene (Aliberti et al., 2016) or germacrene D , although their abundance is usually lower than 0.5%.
Finally, C. medica peel lacks some compounds especially non-terpenoid aldehyde that many times are also absent in C. aurantifolia and/or C. bergamia peel, but which have been detected in the rind of all the other species analyzed. Thus, (E,E)-2,4-decadienal has never been reported in C. medica and C. bergamia, and (Z)-3-hexenal, (E)-2-dodecenal and the sesquiterpenoid α-sinensal have not been reported in C. medica, C. aurantifolia and C. bergamia. As far as we know, other compounds which have not been reported in C. medica peel are: (E)-2-hexenal, neither reported in C. aurantium nor in C. bergamia peel, or hexanal, neither present in C. bergamia, C. aurantium nor in C. junos peels. Also, the aromatic monoterpenoid carvacrol has not been described in C. medica, C. bergamia, C. aurantifolia or C. paradisi.

C. aurantifolia
Many mono-and sesquiterpene hydrocarbons and oxygenated monoterpenes have been reported in C. aurantifolia peel (Figure 2). In fact, C. aurantifolia has some exclusive terpenes such as the sesquiterpene santal-10-en-2-ol (Lota et al., 2002) (Supplementary Tables S2, S3). According to our cluster, the species most similar to C. aurantifolia is C. bergamia and further are C. medica and C. junos (Figure 3).
Despite C. aurantifolia peel being rich in total terpenes, it is striking that we do not find some terpene compounds found in other Citrus species, as indicated in the previous section about C. medica. Among the terpenes absent in C. aurantifolia, that were not mentioned above, we highlight the sesquiterpenoid nootkatone, also not reported in C. junos.
The major compound in C. bergamia is limonene, but its percentage can decrease down to 33% (Poiana et al., 2003). Instead, C. bergamia shows higher percentage of other monoterpene compounds, most notably linalyl acetate, which is sometimes almost as abundant as limonene, which does not happen for any other species of the reviewed, as far as we know (Poiana et al., 2003;Verzera et al., 2003). Linalool is also more abundant in C. bergamia EO than in any other of these species; it can reach up to 13.9% in SC-CO 2 extracts (Poiana et al., 2003). Moreover, other terpenes which are abundant in C. medica and C. aurantifolia are also increased in C. bergamia, when compared to the species of the C. reticulata cluster, such as α-pinene, β-pinene, neryl acetate, geranyl acetate, β-bisabolene, (E)-α-bergamotene, β-caryophyllene (Mondello et al., 2003;Poiana et al., 2003;Verzera et al., 2003;Sawamura et al., 2006;Belsito et al., 2007). Finally, compounds such as octyl acetate are more abundant in C. bergamia than in other species, while compounds such as octanal seems to decrease, with values lower than 0.1% (Poiana et al., 2003;Verzera et al., 2003).
On the other hand, several coumarin and psoralen (furocoumarin) compounds have been isolated in C. bergamia EOs. These compounds are more polar and less volatile than most of the compounds in Citrus EOs. Not surprisingly, they has been mainly reported in oils obtained by cold press (Costa et al., 2010;Russo et al., 2015), organic solvent (polar extracts) (Chisholm et al., 2003b;Craske et al., 2005), or more recently by ultrasonic (Zhang et al., 2017) extractions, and are usually identified by means of HPLC techniques (Costa et al., 2010;Russo et al., 2015). In some studies, such compounds are considered as part of the non-volatile peel fraction (Costa et al., 2010). Following this criterion, we have not included them as part of the volatile profile, although some of them have been listed in Supplementary Table  S2. Although this type of compounds are usually not included in Citrus peel oil studies, when reported they have been principally described in C. bergamia, C. aurantifolia and C. aurantium EOs, and also in other oils such as those from C. limon (Bonaccorsi et al., 1999;Cannon et al., 2015;Russo et al., 2015). These compounds have not been found in C. sinensis or C. reticulata peel, which instead are the richest in polymethoxyflavones such as tangeretin (Bonaccorsi et al., 1999;Russo et al., 2015). Among them, the furocoumarins bergaptene (Poiana et al., 2003;Belsito et al., 2007;Costa et al., 2010;Russo et al., 2015) and bergamottin (Costa et al., 2010;Russo et al., 2015) are characteristic of the rind of C. bergamia, but they are not exclusive for this species as they have also been described in other species as C. aurantifolia (Craske et al., 2005;Russo et al., 2015), C. limon (only bergamottin) (Bonaccorsi et al., 1999;Russo et al., 2015), C. paradisi (Russo et al., 2015), and C. aurantium (Dugo et al., 1996Pino and Rosado, 2000;Russo et al., 2015). Other examples are citropten [C. aurantifolia (Craske et al., 2005;Russo et al., 2015;Zhang et al., 2017), C. bergamia, and C. limon] or 5-geranyloxy-7-methoxy-coumarin [C. aurantifolia (Russo et al., 2015), C. bergamia and C. limon].
Quantitatively, the major compound of C. limon EO is limonene, at levels usually ranging between 70 and 48%, similar to those in C. medica and C. aurantifolia, and lower than those in the species comprising the first cluster. The following more abundant compounds are terpenoids similar to those higher in C. medica and C. aurantifolia. Among these compounds which have not been named so far, we can highlight geranial and neral, that in C. limon (Lota et al., 2002;Loizzo et al., 2016) can reach abundances of 2.9% and 1.5%, respectively, compared to percentages between 0.3%-traces in the species in the C. reticulata cluster (Minh Tu et al., 2002a;Ruberto and Rapisarda, 2002;Mondello et al., 2003;Tao et al., 2014).

Volatile Composition of Citrus Blossoms
There are fewer studies about the VOCs in the flowers of Citrus most common species in comparation with Citrus rind. These analyses have been mostly performed on C. grandis, C. aurantium, C. limon, C. reticulata, C. sinensis, C. aurantifolia, and C. paradisi blossoms. According to our review, many of the volatile compounds present in the Citrus peel have also been reported in Citrus blossom and neroli Boussaada and Chemli, 2006;Jabalpurwala et al., 2009;Cheong et al., 2011b;Sarrou et al., 2013;Družić, 2016). This is the case of many monoterpene hydrocarbons such as limonene, some monoterpene alcohol such as linalool and geraniol, some sesquiterpene hydrocarbons such as (E)β-farnesene and β-caryophyllene, some alcohol sesquiterpene such as (E)-nerolidol and (Z,E)-farnesol, the norcarotenoid 6-methyl-5-hepten-2-one, and a few non-terpenoid aliphatic compounds such as hexanal. The common VOCs identified in the peel of all the 10 species described ( Table 1) are not always reported in their corresponding blossoms. This is the case of the terpene compounds perillaldehyde, carvone, valencene or nootkatone (Supplementary Table S2). Interestingly, the non-terpenoid C 8 -C 12 aliphatic aldehyde compounds, all of them common in the peel of all the revised species, have been sporadically described in flowers of some species as C. paradisi  or C. grandis .
Unlike in the peel EO, the most abundant compound in neroli is not always the monoterpene limonene, as it is sometimes replaced by linalool (Boussaada and Chemli, 2006;Sarrou et al., 2013). Moreover, according to  and Huang et al. (2017), C. limon and C. grandis neroli composition are different to those of whole flowers, because these neroli are richer in monoterpenes such as limonene, while their corresponding whole flowers are richer in alcohol monoterpenoids such as linalool. Moreover, some studies show that VOCs are produced in distinctive amounts by the different flower organs (Flamini et al., 2007;Huang et al., 2017).
Recently a volatilome of Citrus leaves about differentially accumulated volatiles has been developed (Lamine et al., 2018). This study was performed in the leaves of four Citrus species (C. limon, C. reticulata, C. sinensis and C. aurantium) and these four species were clearly discriminated, especially C. limon. The number of taxonomic classifiers was narrow to eight (linalyl acetate, geranyl acetate, neryl acetate, β-cisocimene, δ-3-carene, β-myrcene, sabinene and α-terpineol). However, according to our review, it is very risky to establish taxonomic volatile metabolites with a single study, given the great variability of the volatile profile for the same species.

CONCLUSION
This review about the volatiles in the Citrus essential oil summarized critically the volatile compounds identified up to date in the 10 most frequently studied Citrus species. It allowed their clustering based on the qualitative profile of the volatiles more reliably identified, which was in concordance with their genomic similitude. It also provides the possibility of developing species biomarkers based on this type of compounds, and points toward those which could be useful for the industry to monitor quality and consumer safety (Salgueiro et al., 2010). In addition, traceability, quality control, and adulterations of foods and drugs could be better addressed and optimized by using the Citrus volatile chemical database presented here. This review also provides important information to help in the selection of the most adequate species and the best source for specialized volatile chemicals for pharmaceutical, food, beverages, fragrances, and cosmetic industries.

AUTHOR CONTRIBUTIONS
MG-M reviewed the literature and elaborated the tables. JR performed statistical analysis and figures. ML-G elaborated Supplementary Figure S1. MG-M and JR drafted the manuscript. All authors listed have made a substantial, direct and intellectual contribution to the work, and approved it for publication.

FUNDING
This work was supported in part by the European Commission Horizon 2020 program TRADITOM grant 634561 and TomGEM grant 679796 to JR and AG.