Systematic Detection and Identification of Bioactive Ingredients from Citrus aurantium L. var. amara Using HPLC-Q-TOF-MS Combined with a Screening Method

Bitter orange, Citrus aurantium L. var. amara (CAVA), is an important crop and its flowers and fruits are widely used in China as a food spice, as well as in traditional Chinese medicine, due to its health-promoting properties. The secondary metabolites that are present in plant-derived foods or medicines are, in part, responsible for the health benefits and desirable flavor profiles. Nevertheless, detailed information about the bioactive ingredients in CAVA is scarce. Therefore, this study was aimed at exploring the phytochemicals of CAVA by high performance liquid chromatography/quadrupole time-of-flight mass spectrometry (HPLC-Q-TOF-MS). Here, a systematic screening method combined with HPLC-Q-TOF-MS was presented. This technique was used to systematically screen metabolites, primarily from the complex matrix of CAVA, and to identify these compounds by their exact masses, characteristic fragment ions, and fragmentation behaviors. A total of 295 metabolites were screened by the screening method and 89 phytochemicals were identified in the flowers, fruits, roots, leaves, and branches of CAVA. For the first time, 69 phytochemicals (flavonoids, alkaloids, terpenoids, etc.) were reported from CAVA. The results highlight the importance of CAVA as a source of secondary metabolites in the food, medicine, and nutraceutical industries.


Establishment of the Screening Method
LC-MS is a fast and sensitive tool for the detection and identification of metabolites in plant medicines and foods. In previous studies, many components, especially trace compounds, have been missed due to poor screening methods. In a specific plant, analogues with the same skeleton but different substituent groups are synthesized synchronously in different amounts through specific biosynthetic pathways. Abundant compounds, or ones exhibiting a high-quality MS response, are easy to detect, while trace analogs, or compounds that exhibit a poor MS response, are always overwhelmed by complex matrices, and are difficult to discover [23,24]. In light of this situation, a method for detecting analogues in CAVA using HPLC-Q-TOF-MS combined with a screening strategy was established (Figure 1). Three approaches, non-, accurate-, and extensive-target were used for systematic screening metabolites from the TICs of different CAVA samples. A non-target method was used to screen compounds that were abundant or compounds that exhibited a high-quality MS response which can present distinct peaks in TIC and, then, fragment ions were obtained by tandem mass spectrometry (MS/MS). An accurate-target method was performed by first developing a list of all reported compounds in previous studies from the genus including their structure, molecular formula, accurate mass, and identification method. Then, the measured exact masses of candidates were obtained using extracted ion chromatogram (EIC) of the calculated precise mass of reported compounds on the TICs. Finally, the characteristic fragment ions of candidates were produced by target-MS/MS. In this study, 142 previously reported compounds were summarized; 106 components were detected in CAVA using the accurate-target method and 44 of them were identified. An extensive-target method combines known basic molecular units with different sugars to obtain a series of theoretical calculated masses and, then, EIC of the formed theoretical exact masses on the TICs of the samples were performed. If the measured MS data match the theoretical calculated mass, those combined molecules are considered to be present in the sample. Finally, the fragments of each candidates were obtained by target-MS/MS. In this study, 272 theoretical exact masses were formed by combining the eight basic units (hesperitin, naringenin, apigenin, eriodictyol, diosmetin, acacetin, luteolin, and cirsimaritin) with four common sugars (glucose, rhamnose, arabinose, and glucuronic acid). According to the TICs of flowers, fruits, leaves, roots, and branches of CAVA, the measured exact masses of 67 candidates were obtained by using an EIC method, and the target-MS/MS analysis was conducted for each candidate. Finally, the most likely structures of 38 metabolites were inferred by the fragmentation pathway of references. The above three methods have been used individually for screening of metabolites in specific plants, however, comprehensive and systematic detection of bioactive ingredients by combining the three methods has been rarely reported ( Figure 1). In this study, 295 compounds were screened from CAVA by combining all three methods and 89 of them were identified.

Screening and Identification of Flavonols and Flavonol Glycosides
A series of similar compounds with the same framework but different substituent groups are distributed throughout CAVA. Since these analogues typically display similar MS fragmentation behaviors, investigation of the fragmentation pathways of well-known references is a valid approach for identifying the unknown analogues. The fragmentation behaviors and characteristic diagnostic ions of ten reference samples were investigated in detail and used for identifying the flavonols and flavonol glycosides in CAVA.
In the MS/MS spectra of neoeriocitrin (36), poncirin (39), eriocitrin (40), naringin (42), naringenin (43), narirutin (44), neohesperidin (51), hesperidin (56), apigenin (59), and hesperitin (61) ( Figure S1), four fragmentation behaviors dominated. The first fragmentation pathway was the successive neutral loss of sugars from the protonated flavonol glycoside, and formation of the basic unit.  Figure 2). The second fragmentation pattern was the cleavage of the C-ring and formation of a series of relatively low m/z fragment ions (   [20,21]. The dehydrated glucose (Glc, 162.0528), rhamnose (Rha, 146.0579), arabinose (Ara, 132.0423), and glucuronic acid (Glc A, 176.0321) were the primary substituent groups for those components. By adding no more than three sugar molecules to the skeleton, a total of 272 different theoretical exact masses were obtained. Sixty-seven candidates were produced using EIC based on the TIC of flowers, fruits, roots, leaves, and branches of CAVA and their MS/MS spectra were produced by the target-MS/MS model. The structures of 38 candidates were tentatively determined by the fragmentation pathways of flavonols and flavonol glycosides. In addition, 142 potential compounds were obtained by the non-target and accurate-target methods and 47 compounds were tentatively identified by their characteristic fragmentation behaviors. Finally, 209 flavonols and flavonol glycosides were screened by the non-, accurate-, and extensive-target methods and 58 components, including 19 flavones, 27 flavanones, and 12 polymethoxyflavonoids were identified and 45 of them were reported for the first time from this plant.
Compound 59 was screened by the three screening methods simultaneously and its MS/MS data was obtained by target-MS/MS. Compound 59 was identified as apigenin unambiguously by comparison of the retention time, MS, and MS/MS data with that of the standard (Table 1). It was difficult to find compound 52 (TR = 14.22 min, Figure 3) using the non-target method because of the low content or poor response and the lack of distinct peaks in the TICs. However, this compound was easily detected by the accurate-and extensive-target methods using EIC on the TIC of different parts of CAVA. In the MS/MS spectrum of compound 52 (     Compounds 36, 39, 40, 42, 43, 44, 51, 56, and 61 were unambiguously identified as neoeriocitrin, poncirin, eriocitrin, naringin, naringenin, narirutin, neohesperidin, hesperidin, and hesperitin, respectively, by comparison of the retention time, MS, and MS/MS data with the corresponding standards ( Table 1). The protonated ion of compound 57 was submerged in high abundance ions or complex biological matrices, making it difficult to detect by the non-target method (Figure 3). In addition, this secondary metabolite has not been reported in this genus previously. Therefore, not surprisingly, it was difficult to detect compound 57 by the accurate-target mean, however, the compound was detected with the extensive-target method by screening the theoretical exact mass on the TICs. The extensive-target method indicated that the basic skeleton of this compound was naringenin and the substituent group was Glc. In the MS/MS spectrum of compound 57 (Figure 4), the basic skeleton ion at m/z 273.0742 was formed by the loss of a Glc residue from the protonated ion occurring at m/z 435.1295. The fragments at m/z 153.0175 and 273.0742 indicated that the basic skeleton was naringenin. Thus, compound 57 was preliminarily identified as naringenin-O-glucoside (Figure 4). Using the same method, the remaining flavanone-type compounds (33, 34, 35, 37, 41, 46, 47, 48, 49, 53,  60, 64, 68, 69, 74, 75, and 77) were tentatively identified (Table 1) and the relevant MS/MS spectra are provided in the Supplementary Materials ( Figure S2).
Compounds 85 and 87 were unambiguously identified as nobiletin and tangeretin, respectively, by comparing the retention time, MS, and MS/MS data with the references (Table 1). Compound 86 presented a distinct peak in the TIC of CAVA roots ( Figure 3) and has previously been reported in this genus. Therefore, compound 86 was easily detected with the non-target and accurate-target methods. The fragmentation pathways of polymethoxyflavonoid-type compounds were investigated in detail using nobiletin (85) and tangeretin (87) as references before identifying their structures of compound 86 and other compounds. The MS behaviors of polymethoxyflavonoid-type compounds were different from other types of flavonoids. First, this type of compound only responded well in positive mode of ESI. Second, the main fragmentation route was the successive losses of small groups, such as H 2 O moiety and CH 3 radical from the basic skeleton. In the MS/MS spectra of references 85 and 87 ( Figure S1), fragment ions observed at m/z 388.1133 and 358.1032 were generated by loss of CH 3 radical from the protonated ions at m/z 403.1368 and 373.1264, respectively (Figure 5a). Fragments at m/z 355.0790 and 325.0689 were formed by neutral loss of H 2 O moiety from the ions at m/z 373.0895 and 343.0793, respectively. The third fragmentation pattern was cleavage of the C-ring and formation of relatively low m/z fragment ions. The characteristic fragment ions at m/z 211.0220 and 211.0223 for compounds 85 and 87 were formed by RDA reaction (cleavage the C-ring, Figure 5). In the MS/MS spectrum of compound 86, the fragmentation behavior had a high similarity with polymethoxyflavonoid-type compounds. The difference in m/z values of compounds 86 and 87 was 30.0079 Da, which indicated that the structure of compound 86 has an OCH 3 moiety fewer than that of compound 87. According to a previous report [27], compound 86 was preliminarily identified as 4 ,5,6,7-pentamethoxyflavone by comparison with characteristic ions. Using the similar method, the remaining polymethoxyflavonoid-type compounds (26, 54, 72, 81, 82, 83, 84, 88, and 89) were tentatively identified (Table 1) and the relevant MS/MS spectra are provided in the Supplementary Materials ( Figure S2).

Screening and Identification of Coumarin
In the MS/MS spectra of the three references (xanthotoxol (71), scopoletin (76), and auraptene (80) (Figure S1), it is difficult to cleave the skeleton of coumarin, therefore, the primary characteristic fragmentation pathway was the loss of small molecular groups, such as CO, CH 3 Figure 5b.
Compounds 71, 76, and 80 were clearly identified as xanthotoxol, scopoletin, and auraptene (Table 1), respectively, by comparing the retention time, MS, and MS/MS data with those of the standards. Compound 78 presented a distinct peak in the TICs (Figure 3). In addition, this compound has been previously reported in this genus. Therefore, compound 80 could be detected easily by the non-and accurate-target methods. In the MS/MS spectrum of compound 78, the fragmentation behavior was highly consistent with coumarin-type compounds. The difference in m/z values between compounds 78 and 71 was 14.0146 Da, which indicated that compound 78 has a CH 3 moiety more than compound 71. In the MS/MS spectrum of compound 78, the high abundance fragment ion at m/z 202.0250 was generated by the loss of a CH 3 radical from the protonated ion at m/z 217.0483, which indicated that a CH 3 moiety was included in the structure of metabolite 78. According to a previous report [28], compound 78 was preliminarily identified as bergapten (Figure 4).

Screening and Identification of Alkaloids and Triterpenoid
In the MS/MS spectrum of synephrine (15) (Figure S1), the primary fragmentation route was the loss of small molecular groups, such as CH 3 and H 2 O, from the basic skeleton. The fragment observed at m/z 150.0917 was formed by the neutral loss of H 2 O from the protonated ion at m/z 168.1014. Successive loss of CH 3 and CHNH moieties were observed, resulting in the ions at m/z 135.0670 and 107.0500. The proposed fragmentation pathways are shown in Figure 5c.
By comparing the retention time, MS, and MS/MS data with the standard substance, the structure of compound 15 was clearly determined (Table S1). Compound 21 presented a distinct peak in the TICs ( Figure 3) and has been reported previously in this genus. Therefore, compound 21 was detected easily by the non-and accurate-target methods. In the MS/MS spectrum of compound 21, the difference in m/z value between compounds 21 and 15 was 15.9908 Da, which indicates that the structure of compound 21 results from the loss of an OH moiety in compound 15. Moreover, the MS/MS fragmentation behaviors of both compounds are highly similar. The ion observed at m/z 121.0639 was generated corresponds to the loss of -NH 2 CH 3 from the protonated ion at m/z 152.1065. The subsequent loss of an H 2 O moiety and formation of a peak at m/z 103.0528 was also observed. The MS/MS data indicates that -NHCH 3 and -OH groups are present in the structure of compound 21. According to the previous report [29], it was preliminarily identified as N-acetylnorsynephrine. Using the same method, the remaining alkaloids (2, 7, 10, 11, 14, 16, 17, 18, 19, 20, 24, 25, and 27) were tentatively identified (Table 1) and the relevant MS/MS spectra are provided in the Supplementary Materials ( Figure S2).
Limonin compounds have been reported previously as the most common triterpenoids in CAVA [20]. Compound 79 was screened by non-target and accurate-target methods and its structure was unambiguously identified as limonin by comparing the retention time, MS and MS/MS data with the reference (Table 1).

Distribution of Metabolites in CAVA
The distribution of all identified compounds in roots, fruits, flowers, leaves, and branches of CAVA were determined using EIC, based on the TICs. More than 90% ingredients were detected and identified from the flowers and fruits, however, the number of identified metabolites from other parts were relatively small. This is the reason why the flowers and fruits were used as main medicinal parts in traditional Chinese medicine. Flavonoids, alkaloids, and coumarins were the main active ingredients of the flowers and fruits. Thirty-two characteristic compounds, such as limonin (79) and auraptene (80), were detected only from both parts. Flavonoids and flavonoid glycosides were the main metabolites of flowers, however, the polymethoxyflavonoid-type compounds (such as 78, 81, 85, and 87) were only in fruits, which demonstrated that the flowers and fruits have different functions as herbal medicine or food additives. The polymethoxyflavonoid-type compounds were detected only from the family of Citrus reticulata Blanco in previous studies and have a wide range of biological activities, such as antioxidant, anti-inflammatory, antitumor, and antifungal activity [3,30]. Those types of flavonoids were mainly distributed in the fruits and roots of CAVA. Nevertheless, the species and amounts of polymethoxyflavonoid-type compounds have a huge difference between the two parts. Some high content polymethoxyflavonoids (such as compounds 82, 83, 88, and 89, comparing the relative peak area) were detected only from the roots of CAVA. Although compounds 84 and 86 were found in fruits and roots, the level of both compounds in roots was far more than that in fruits. Interestingly, some high content polymethoxyflavonoids (such as compounds 78, 79, 81, and 85) in fruits were difficult to detection in the roots (Figure 3). The results revealed that the roots of CAVA, usually discarded in the previous disposal process, were an important source of polymethoxyflavonoid-type metabolites for the food, medicine, and nutraceutical industries.

Conclusions
In this study, we have shown that the combination of HPLC-Q-TOF-MS with a systematic screening method constitutes a powerful analytical tool for the detection and identification of bioactive ingredients in all parts of CAVA. A total of 295 secondary metabolites were primarily found from the flowers, fruits, roots, leaves, and branches of CAVA with a systematic screening method, which is comprised of non-, accurate-, and extensive-target approaches. Eighty-nine compounds, including 19 flavones, 27 flavanones, 12 polymethoxyflavonoids, 4 coumarins, 15 alkaloids, 1 limonoids, and 11 other phytochemicals were identified by their exact masses, fragment ions, and characteristic fragmentation patterns. Sixty-nine of the compounds are reported for the first time from CAVA. To the best of our knowledge, this work marks the first comprehensive study of secondary metabolites from different parts of CAVA. In addition, the established screening method can also be applied to other plant-derived foods and medicines for systematic detection of the bioactive compounds from the complex biological matrices.