Comparative Metabolite Profiling and Fingerprinting of Genus Passiflora Leaves Using a Multiplex Approach of UPLC-MS and NMR Analyzed by Chemometric Tools

Home , Passiflora ambigua, Passiflora biflora, Passiflora capsularis, Passiflora coriacea, Passiflora laurifolia, Passiflora lutea

Anal Bioanal Chem DOI 10.1007/s00216-016-9376-4

RESEARCH PAPER

Comparative metabolite profiling and fingerprinting of genus Passiflora leaves using a multiplex approach of UPLC-MS and NMR analyzed by chemometric tools

Mohamed A. Farag1 & Asmaa Otify 1 & Andrea Porzel2 & Camilia George Michel1 & Aly Elsayed1 & Ludger A. Wessjohann2

Received: 9 December 2015 /Revised: 19 January 2016 /Accepted: 28 January 2016 # Springer-Verlag Berlin Heidelberg 2016

Abstract Passiflora incarnata as well as some other those of C-flavonoids including isovitexin-2″-O-xyloside, Passiflora species are reported to possess anxiolytic and sed- luteolin-C-deoxyhexoside-O-hexoside, schaftoside, ative activity and to treat various CNS disorders. The medic- isovitexin, and isoorientin. P. inc arn ata was found most inal use of only a few Passiflora species has been scientifical- enriched in C-flavonoids, justifying its use as an official drug ly verified. There are over 400 species in the Passiflora genus within that genus. Compared to NMR, LC-MS was found worldwide, most of which have been little characterized in more effective in sample classification based on genetic and/ terms of phytochemical or pharmacological properties. or geographical origin as revealed from derived multivariate Herein, large-scale multi-targeted metabolic profiling and fin- data analyses. Novel insight on metabolite candidates to me- gerprinting techniques were utilized to help gain a broader diate for Passiflora CNS sedative effects is also presented. insight into Passiflora species leaves’ chemical composition. Nuclear magnetic resonance spectroscopy (NMR) and mass Keywords Passiflora . C-glycosides . Chemotaxonomy . spectrometry (MS) spectra of extracted components derived Biomarkers . Flavonoids . Multivariate data analysis . from 17 Passiflora accessions and from different geographical UPLC-MS . NMR origins were analyzed using multivariate data analyses. A total of 78 metabolites were tentatively identified, that is, 20 C- flavonoids, 8 O-flavonoids, 21 C, O-flavonoids, 2 cyanogenic Introduction glycosides, and 23 fatty acid conjugates, of which several flavonoid conjugates are for the first time to be reported in Passifloraceae is a family of flowering plants, including over Passiflora spp. To the best of our knowledge, this study pro- 700 species classified in around 25 genera [1]. They are most- vides the most complete map for secondary metabolite distri- ly found in tropical and temperate regions including ornamen- bution within that genus. Major signals in 1H-NMR and MS tal plants such as Passiflora incarnata L., Passiflora foetida spectra contributing to species discrimination were assigned to L. as well as commercialized species as Passiflora edulis Sims and Passiflora laurifolia L., both widely cultivated for their edible fruits. The genus Passiflora from the Passifloraceae Electronic supplementary material The online version of this article plant family comprises about 400 dicotyledonous species, (doi:10.1007/s00216-016-9376-4) contains supplementary material, and the genus is considered the largest and namesake of the which is available to authorized users. Passifloraceae family [2]. Several species have a long history * Mohamed A. Farag in traditional herbal medicine, though the medicinal use has [email protected]; [email protected] been scientifically verified only for very few Passiflora species to varied extents. Aerial parts of Passiflora species have been traditionally used to treat anxiety, insomnia, and ner- 1 Pharmacognosy Department, College of Pharmacy, Cairo University, vousness [3]. In particular, P. incarnata L. (passion flower) Kasr el Aini St., Cairo P.B. 11562, Egypt is regarded as the most common species used in contemporary 2 Department of Bioorganic Chemistry, Leibniz Institute of Plant Western phytotherapy for the treatment of anxiety disorders Biochemistry, Weinberg 3, 06120 Halle (Saale), Germany [4]. Other reported biological activities include antioxidant, M.A. Farag et al. antihypertensive, anti-inflammatory, antitumor, and antimi- Plant material crobial effects [5]. P.incarnata L. is indeed listed as an official plant drug in the 1970s and 1990s in several pharmacopoeias Leaves from 17 different species within the genus Passiflora [6–8]. Passion extracts are worldwide distributed in health (a total of 20 samples) were collected from different geograph- stores and often formulated with other herbal drugs as cham- ical origins (Table 1). omile, hops, kava, skullcap, and valerian in nutraceuticals used as sedatives. Several Passiflora species have been ana- Chemicals and reagents lyzed with respect to their flavonoid composition, being a rich source of the otherwise rare C-glycosyl flavones [9, 10]. Methanol-d4 (99.80 % D), acetone-d6 (99.80 % D), and Nevertheless, several metabolites, i.e., maltol [11], alkaloids hexamethyldisiloxane (HMDS) were provided from Deutero [12], and essential oil [13, 14], have also been proposed to GmbH (Kastellaun, Germany). Acetonitrile and formic acid mediate for Passiflora pharmacological effects. The active (LC-MS grade) were purchased from J. T. Baker principles have not yet been fully identified which warrants (The Netherlands); Milli-Q water was used for LC analysis. for a more detailed, initially untargeted type of analysis to Vitexin, isovitexin, harmine, orientin, and isoorientin were provide better insight on the global chemical composition of purchased from Chromadex (Wesel, Germany). All other that genus. Metabolomics is defined as the comprehensive chemicals and standards were provided by Sigma-Aldrich analysis of thousands of metabolites (metabolome), providing (St. Louis, MO, USA). a way for accurate analysis and or standardization of herbal products. The metabolomic technique enables large numbers Extraction procedure and sample preparation for NMR of samples to be processed quickly and simultaneously “glob- and MS analyses al approach”; in an untargeted manner, such approach often utilizes hyphenated analytical techniques most commonly liq- A one pot extraction protocol developed in Farag et al. [20] uid chromatography (LC) coupled to mass spectrometry (MS) was used for extraction of Passiflora specimens. Dried and or direct analysis using nuclear magnetic resonance spectros- deep frozen Passiflora leaves were ground with a pestle in a copy (NMR). Such technologies have great potential for ac- mortar using liquid nitrogen. The powder (120 mg) was ho- curate analyses or standardization of herbal products. mogenized with 5 ml 100 % MeOH containing 10 μg/ml Successful examples for the use of metabolomics in quality umbelliferone (an internal standard for relative quantification control of herbal drugs involve analyses of licorice [15], ephe- using UPLC-MS) using a Turrax mixer (11,000 RPM) for five dra [16], hops [17, 18], or St. John’sWort[19]. Combination 20-s periods. To prevent heating, a period of 1 min separated of the complementary advantages of both platforms, i.e., pow- each mixing period. Extracts were then vortexed vigorously erful NMR structural elucidation accuracy and MS high sen- and centrifuged at 3000×g for 30 min to remove plant debris. sitivity and peak resolution, was employed to profile metabo- For NMR analysis, 3 ml was aliquoted using a syringe and the lites in 17 different Passiflora species. Our aim is to explore solvent was evaporated under a stream of nitrogen to dryness. variation in secondary metabolites from a wide range of Dried extracts were resuspended with 800 μl100%methanol- Passiflora species in terms of genotypes and geographical d4 containing HMDS. After centrifugation (13,000×g for origins to provide a detailed map for metabolite pattern- 1 min), the supernatant was transferred to a 5-mm NMR tube. based taxonomy, quality control analysis, and to help identify All 1H-NMR spectra for multivariate data analysis were ac- alternatives for the official drug P. incarnata. To achieve our quired consecutively within a 48-h time interval with samples goal, metabolite profiling and fingerprinting using ultra per- prepared immediately before data acquisition. Repeated con- formance liquid chromatography (UPLC)-MS in parallel to trol experiments after 48 h showed no additional variation. For NMR techniques were employed for the analysis of the offi- UPLC-MS analyses, 500 μl was aliquoted and placed on a cial P. in carn ata leaves from three geographical origins. (500 mg) C18 cartridge preconditioned with methanol and Additionally, metabolic profiles of 16 other different water. Samples were then eluted using 6 ml methanol, the Passiflora leaf extracts (Table 1) were performed to help iden- eluent was evaporated under a nitrogen stream, and the ob- tify a substitute candidate and for drug quality control analysis tained dry residue was resuspended in 1 ml methanol. Three and or adulteration detection. microliters was used for UPLC-MS analysis. For each speci- men, three biological replicates were provided and extracted in parallel under identical conditions.

Experimental High resolution UPLC-PDA-MS and MS/MS analysis

Materials and methodology used in this study were applied Chromatographic separations were performed on an Acquity following the protocol of Farag et al. [20]. UPLC system (Waters) equipped with a HSS T3 column Comparative metabolite profiling and fingerprinting of Passiflora leaves

Table 1 Origin of Passiflora leaf samples used in this study

Species Country Herbarium voucher

Passiflora ambigua Hemsl. Germany, Botanische Gärten der Universität Bonn xx-0-BONN-26956 Passiflora caerulea L. Germany, Botanische Gärten der Universität Bonn xx-0-BONN-10173 Passiflora capsularis L. Germany, Botanische Gärten der Universität Bonn xx-0-BAS-369/2004GP Passiflora costaricensis Killip Germany, Botanische Gärten der Universität Bonn xx-0-BONN-25338 Passiflora edulis Sims Germany, Botanische Gärten der Universität Bonn xx-0-BONN-28265 Passiflora helleri Peyr. Germany, Botanische Gärten der Universität Bonn xx-0-BONN-1019 Passiflora holosericea L. Germany, Botanische Gärten der Universität Bonn xx-0-BONN-8748 Passiflora incarnata L. Germany, Botanische Gärten der Universität Bonn xx-0-BONN-26958 Passiflora perfoliata L. Germany, Botanische Gärten der Universität Bonn xx-0-BONN-9626 Passiflora quadrangularis L. Germany, Botanische Gärten der Universität Bonn xx-0-BONN-1020 Passiflora racemosa Brot. Germany, Botanische Gärten der Universität Bonn xx-0-BONN-9620 Passiflora serratifolia L. Germany, Botanische Gärten der Universität Bonn xx-0-BONN-9648 Passiflora subpeltata Ortega Germany, Botanische Gärten der Universität Bonn xx-0-BONN-10623 Passiflora incarnata L. Germany, Botanischer Garten Göttingen GT-4509-401 Passiflora incarnata L. Germany, Pharmazie Kiel A6509-112 Passiflora lutea L. USA, North Carolina, Mecklenburg Co US-0-BONN-32505 Passiflora microstipula L.E. Gilbert and J.M. MacDougal Mexico, sine loc, A.Lau MX-0-BONN-10620 Passiflora coriacea Juss. Guatemala, O-Guatemala, Yucatan Tical GT-0-BONN-9617 Passiflora × kewensis Hort. Guatemala, BG Osnabrück GT-0-OSN-95-0025 Passiflora edulis Sims Egypt, Giza, El-Orman Botanical Garden 26-12-2012

(100 × 1.0 mm, particle size 1.8 μm; Waters) applying the (ThermoElectron, San Jose, USA) equipped with a ESI source following elution binary gradient at a flow rate of (electrospray voltage, 4.0 kV; sheath gas, nitrogen; capillary 150 μlmin−1: 0 to 1 min, isocratic 95 % A (water/formic acid, temperature, 275 °C) in positive and negative ionization 99.9/0.1 [v/v]), 5 % B (acetonitrile/formic acid, 99.9/0.1 [v/v]); modes. The Ion Trap MS system is coupled with the exact 1to16min,linearfrom5to95%B;16to18min,isocratic Waters UPLC setup, using same elution gradient. The MSn 95 % B; and 18 to 20 min, isocratic 5 % B. The injection spectra were recorded during the UPLC run by using the fol- volume was 3.1 μl (full loop injection). C18-bonded phase lowing conditions: MS/MS analysis with a starting collision- used for the HSS T3 sorbents is compatible with 100 % aque- induced dissociation energy of 20 eVand an isolation width of ous mobile phase and provides ultra-low MS bleed while pro- +2 amu in data-dependent positive and negative ionization moting superior polar compound retention which has been modes [20]. Metabolites were characterized by their UV–vis successfully used for the profiling of similar plant extracts spectra (220–600 nm), retention times relative to external [17, 19]. Samples were randomly allocated in the sequence standards, mass spectra and comparison to our in-house data- run to avoid for any bias in measurements. Eluted compounds base, MassBank, phytochemical dictionary of natural product were detected from m/z 100 to 1000 using a MicroTOF-Q database, and reference literature. hybrid quadrupole time-of-flight mass spectrometer (Bruker Daltonics) equipped with an Apollo-II electrospray ion source MS data processing for multivariate data analyses (PCA, in positive and negative ion modes using the following instru- HCA, and OPLS) ment settings: nebulizer gas, nitrogen,1.6 bar; dry gas, nitro- − gen, 6 l min 1, 190 °C; capillary, −5500 V (+4000 V); end Relative quantification and comparison of Passiflora metabol- plate offset, −500 V; funnel 1 resolve filter (RF), 200 Vpp; ic profiles after UPLC-MS were performed using XCMS data funnel 2 RF, 200 Vpp; in-source CID energy, 0 V; hexapole analysis software under R 2.9.2 environment, which can be RF, 100 Vpp; quadrupole ion energy, 5 eV; collision gas, downloaded for free as an R package from the Metlin argon; collision energy, 10 eV; collision RF 200/400 Vpp Metabolite Database [21]. This software approach employs (timing 50/50); transfer time, 70 μs; prepulse storage, 5 μs; peak alignment, matching, and comparison to produce a peak pulser frequency, 10 kHz; and spectra rate, 3 Hz. Electrospray list. The resulting peak list was processed using the Microsoft ionization-tandem mass spectrometry (ESI-MSn) spectra were Excel software (Microsoft, Redmond, WA), where the ion obtained from a UPLCQ Deca XP MAX system features were normalized to the total integrated area (1000) M.A. Farag et al. per sample and imported into the R 2.9.2 software package for multivariate analyses. PCA and hierarchical clustering analy- principal component analysis (PCA) and orthogonal projec- sis (HCA) were performed with R package (2.9.2) using tion to latent structures-discriminant analysis (OPLS-DA). custom-written procedures after scaling to HMDS signal and Absolute peak area values were auto-scaled (the mean area exclusion of solvent regions as described elsewhere [15]. value of each feature throughout all samples was subtracted from each individual feature area and the result divided by the standard deviation) prior to principal component analysis. Results and discussion This provides similar weights for all the variables. PCA was then performed on the MS-scaled data to visualize general To allow for a comparative analysis of the metabolite data clustering, trends, and outliers among the samples on the score derived from different metabolomic platforms that is found plot. OPLS-DA was performed with the program SIMCA-P compatible with both NMR and UPLC-MS metabolomics version 13.0 (Umetrics, Umeå, Sweden). Biomarkers for spe- [15, 18, 20], a one pot extraction method was developed cies were subsequently identified by analyzing the S-plot, (see “Experimental”). Chemometric tools were further used which was declared with covariance (p) and correlation (pcor). to classify between samples and to ensure good analytical Distance to the model (DModX) test was used to verify the rigorousness and define both similarities and differences presence of outliers and to evaluate whether a submitted sam- among samples. ple fell within the model applicability domain. UPLC-MS identification of Passiflora leaf metabolites NMR analysis Chemical constituents of the 17 examined Passiflora species All spectra were recorded on an Agilent VNMRS 600 NMR leaves (Table 2 and Electronic Supplementary Material (ESM) spectrometer operating at a proton NMR frequency of Table S1) were analyzed via reversed-phase UPLC/PDA/ESI- 599.83 MHz using a 5-mm inverse detection cryoprobe. 1H- qTOF-MS, using a gradient mobile phase that allowed for a NMR spectra were recorded with the following parameters: comprehensive elution of plant analytes within ca. 900 s. digital resolution 0.367 Hz/point (32 K complex data points), Extracts were analyzed in both positive and negative ioniza- pulse width (pw) = 3 μs (45°), relaxation delay = 23.7 s, acqui- tion modes as Passiflora is known to contain indole alkaloids sition time = 2.7 s, and number of transients = 160. Zero filling [34] in addition to flavonoids, of which the latter preferentially up to 128 K and an exponential window function with lb = 0.4 ionize under negative ionization. Changes in ESI polarity can were used prior to Fourier transformation. 2D NMR spectra alter competitive ionization and suppression effects revealing were recorded using standard CHEMPACK 4.1 pulse se- new metabolites [35]. The positive ion ESI mass spectra were quences (gDQCOSY,gHSQCAD, gHMBCAD) implemented characterized by cations corresponding to pseudo-molecular in Varian VNMRJ 2.2C spectrometer software. The species (i.e., (M + H)+ and several lower m/z fragment ions heteronuclear single quantum coherence spectroscopy attributed to the sequential loss of sugar moieties that typically 1 (HSQC) experiment was optimized for JCH = 146 Hz with allow for the determination of molecular weight and aglycone DEPT-like editing and 13C-decoupling during acquisition molecular weight. In general, and compared with positive ESI, time. The heteronuclear multiple bond correlation (HMBC) negative ionization provided better sensitivity and revealed experiment was optimized for a long-range coupling of more observable peaks particularly of fatty acids due to the 1 8 Hz; a two-step JCH filter was used (130–165 Hz). For presence of ionized carboxylate moieties. A total of 97 chro- NMR analysis, samples were randomly allocated in the se- matographic peaks belonging to various metabolite classes quence run. were obtained including mostly flavonoids, alkaloids, cyanogenic glycosides, and fatty acids (Table 2). A representative NMR data processing and multivariate data analysis UPLC-MS base peak chromatogram of extracts of P. incarnata L., P. edulis Sims, and Passiflora caerulea L. is The 1H-NMR spectra were automatically Fourier transformed shown in Fig. 1a–c, and the structures of metabolites com- to [.esp] files using ACD/NMR Manager lab version 10.0 monly found in Passiflora genus and discussed throughout software (Toronto, Canada). The spectra were referenced to manuscript are shown in Fig. 2. Flavone glycoside conjugates internal HMDS at 0.062 ppm for 1H-NMR and to internal were shown to be the major class in Passiflora spp. as evi- 13 CD3OD signals at 49 ppm for C-NMR, respectively [20]. denced from photodiode array detection which revealed two Spectral intensities were reduced to integrated regions, re- distinct UV maxima, i.e., 270 and 325–350 nm. MS/MS was ferred to as buckets, of equal width (0.04 ppm) within the employed for metabolite structural elucidation, where there is region of δ = 11.4 to −0.4 ppm. The regions between δ =5.0– a striking difference in the fragmentation pathways of the of 4.7 ppm and δ =3.4–3.25 ppm corresponding to residual water O-glucosyl and C-glucosyl flavonoids that allow them to be and methanol signals, respectively, were removed prior to readily differentiated. For all of the O-glucosyl flavonoids, the oprtv eaoiepoiigadfnepitn of fingerprinting and profiling metabolite Comparative Table 2 Metabolites identified in Passiflora species leaf methanol extracts via UPLC-PDA-MS in positive (P) and negative (N) ionization modes

+/− n Peak no. Mode M Rt (s) UV Molecular Error MS Identification Species formula

L1 P 213.2600 139 271 C13H13N2O +8.4 171, 199 Harmine P. incarn ata − L2 N 456.1464 176 271, 340 C22H26NO11 +1.8 258, 293, 325, 431 Mandelonitrile-O-di-glucoside (amygdalin) P. edulis + L3 P 282.1709 178 264 C15H24NO4 −3.6 173, 237, 246, 254, 261, 277 Unknown P. ambigua − L4 N 609.1558 178 269, 346 C27H29O16 +0.8 369, 399, 447, 489, 519, 591 Luteolin-6,8-C-di-glucoside (lucenin-2) [22] P. caerulea, P. edulis, P. microstipula + L5 P 583.1645 182 292 C26H31O15 +2.3 320, 463, 565 Unknown flavonoid P. microstipula − L6 N 593.1651 188 270, 335 C27H29O15 +1.2 353, 383, 473, 503, 575 Apigenin-6,8-C-di-glucoside (vicenin-2) [9] P. microstipula, P. incarn ata, P. edulis + L7 P 274.0689 191 311 C14H12NO5 +7.4 184, 218, 228, 234, 252 Unknown P. coriacea − L8 N 579.1495 191 271, 333 C26H27O15 +0.9 459, 489, 399, 369, 519 Luteolin-6-C-hexoside-8-C-pentoside P. lutea − L9 N 609.1632 196 270, 347 C27H29O16 −4.4 447, 327, 357, 429 Isoorientin-4′-O-glucoside [23] P. incarn ata, P. perfo lia ta − L10 N 440.1518 195 270, 338 C20H26NO10 +1.4 270, 294, 318, 337, 388, 415 Mandelonitrile-O-rutinoside P. edulis − L11 N 563.1512 201 270, 349 C26H27O14 +1.3 443, 473, 383, 503, 487, 547 Apigenin-6-C-glucoside-8-C-arabinoside P. caerulea, P. coriacea, P. microstipula, (schaftoside) [24] P. incarn ata, P. edulis Passiflora + L12 P 595.1651 201 270, 350 C27H31O15 −3.5 303, 329, 359, 395, 431, 449 Luteolin-C-hexoside-O-deoxyhexoside (isomer) P. subpeltata, P. perfo lia ta − L13 N 447.1030 203 269, 349 C21H19O11 −0.2 327, 357, 369, 285, 545 Luteolin-6-C-glucoside (isoorientin) [25] P. caerulea, P. capsularis, P. lutea,

P. incarn ata, P. edulis leaves + L14 P 593.1872 204 269, 348 C28H33O14 −1.2 329, 359, 431, 449, 463, 561 Luteolin-C-hexoside-O-(dideoxy)hexoside-methyl ether P. perfoliata − L15 N 447.1030 207 269, 348 C21H19O11 −0.8 327, 357, 285, 545 Luteolin-8-C-glucoside (orientin) [25] P. capsularis, P. lutea, P. incarnata, P. edulis + L16 P 463.1243 207 270, 347 C22H23O11 −1.8 269, 271, 280, 343, 445, 449 Orientin-4′-methyl ether [26] P. perfoliata + L17 P 741.2260 207 270, 339 C33H41O19 −4.2 595, 433, 415, 313, 272 Apigenin-C-hexoside-O-hexoside-O-deoxyhexoside P. x-kewensis ¯ L18 N 577.1556 208 270, 343 C27H29O14 +3.4 457, 487, 559, 311, 473 Chrysin-6,8-C-di-hexoside P. edulis + L19 P 771.2360 208 270, 347 C34H43O20 −1.9 295, 303, 405, 463, 591, 625 Luteolin-di-hexoside-deoxyhexoside-methyl ether P. x-kewensis − L20 N 593.1656 210 268, 335 C27H29O15 +0.2 413, 327, 431, 327, 357, 285 Luteolin-C-deoxyhexoside-2″-O-hexoside P. ambigua, P. serratifolia, P. quadrangularis, P. incarn ata, P. perfo lia ta + L21 P 535.1449 213 249, 346 C25H27O13 +1.3 287, 303, 317, 369, 377, 449 Unknown luteolin conjugate P. coriacea, P. microstipula − L22 N 563.1512 215 268, 335 C26H27O14 −1.6 433, 415, 313, 271 Isovitexin-2″-O-xyloside [27] P. serratifolia, P. quadrangularis − L23 P 579.1565 215 269, 337 C27H31O14 −1.5 413, 457, 353, 383, 457, 559 Luteolin-C-deoxyhexoside-O-deoxyhexoside P. subpeltata, P. helleri, P. caerulea, P. perfoliata, P. edulis + L24 P 625.2000 215 270, 337 C28H33O16 +0.9 286, 449, 463 Luteolin-6,8-C-di-glucoside-methyl ether [26] P. helleri + L25 P 581.1507 217 324 C26H29O15 −1.4 287, 340, 331, 399, 449, 535 Luteolin-6-C-pentoside-8-C-glucoside [26] P. coriacea + L26 P 419.0973 218 340 C20H19O10 0.0 287, 295, 343, 369, 409 Unknown flavonoid P. lutea − L27 N 431.097 219 270, 333 C21H19O10 −1.0 311, 341, 353, 383, 413 Apigenin-6-C-glucoside (isovitexin) [25] P. caerulea, P. capsularis, P. lutea, P. incarn ata, P. edulis + L28 P 609.1809 222 nd C28H33O15 −0.9 595, 449, 519, 287 Orientin-2′′-O-rhamnoside-3′-methyl ether P. helleri, P. x-kewensis (scoparin-2″-O-rhamnoside) [28] − L29 N 445.1299 224 327 C22H21O10 −2.9 325, 355, 413, 311, 341, 427 Apigenin-C-hexoside-methyl ether P. perfoliata, P. incarnata + L30 P 449.1086 225 325 C21H21O11 −1.6 287, 291, 313, 409, 419 Luteolin-O-hexoside P. lutea + L31 P 593.1723 226 324 C28H33O14 −2.1 346, 406, 447, 478 Isovitexin-4′-O-rhamnoside-7-methyl ether P. holosericea (Swertisin-4′-O-rhamnoside) [29] − L32 N 563.1512 228 271, 327 C26H27O14 +3.1 459, 473, 545, 369, 443, 503 Luteolin-6-C-deoxyhexoside-8-C-pentoside P. edulis Table 2 (continued)

+/− n Peak no. Mode M Rt (s) UV Molecular Error MS Identification Species formula

+ L33 P 317.0556 230 nd C9H17 O10S −5.7 147, 295 Unknown P. helleri + L34 P 609.1659 230 274, 325 C29H37O14 +1.8 324, 477, 499 Unknown luteolin conjugate P. perfoliata + L35 P 537.2704 231 313 C28H41O10 −2.5 163, 219, 227, 237 Unknown P. costaricensis − L36 N 593.1651 231 270, 338 C27H29O15 −2.4 447, 429, 357, 327 Luteolin-C-hexoside-O-deoxyhexoside (isomer) P. subpeltata + L37 P 725.2310 231 270, 312 C33H41O18 −3.2 303, 374, 382, 417, 433, 579 Luteolin-3′-O-di-rhamnoside-7-O-rhamnoside [30] P. caerulea − L38 N 577.1557 233 nd C27H29O14 +2.6 311, 269, 397, 415, 473 Apigenin-C-deoxyhexoside-O-hexoside [31] P. edulis + L39 P 319.2015 233 270, 324 C18H27N2O3 −0.2 220, 275, 299 Unknown P. microstipula − L40 N 575.1551 234 270, 340 C27H27O14 −2.2 411, 429, 357, 473, 285 Luteolin-8-C-6″deoxy-3″-hexuloside-2″-O-rhamnoside [32] P. subpeltata − L41 N 431.1018 238 270, 347 C21H19O10 +2.7 285, 327, 357, 413, 311, 341 Luteolin-6-C-quinovoside/fucoside [33] P. incarn ata, P. edulis − L42 P 417.1039 239 270, 315 C21H21O9 +2.8 297, 399 Chrysin-6-C-hexoside P. edulis + L43 P 577.1551 241 270, 340 C27H29O14 −2.2 329, 347, 358, 413, 431, 559 Luteolin-deoxyhexoside-6-C-deoxyhexuloside P. subpeltata + L44 P 595.1647 241 nd C27H31O15 +0.4 431, 449, 329, 359, 287 Isoorientin-2″-O-rhamnoside [29] P. coriacea + L45 P 481.1678 243 273 C23H29O11 +5.7 249, 258, 265, 295, 303 Unknown P. ambigua + L46 P 535.1081 243 269, 334 C24H23O14 +0.1 287, 317, 358, 369, 393, 449 Luteolin-malonylhexoside P. perfoliata + L47 P 449.1081 244 340 C21H21O11 −0.7 274, 286, 371, 357, 431, 433 Unknown flavonoid P. lutea + L48 P 493.1337 247 nd C23H25O12 +0.2 163, 287, 357, 430, 449, 463 Methoxyluteolin-O-hexoside-methyl ether P. coriacea + L49 P 549.1609 248 271, 317 C26H29O13 −1.3 211, 312, 324, 399, 417 Apigenin-6-C-rhamnoside-8-C-arabinoside [30] P. caerulea + L50 P 621.1822 248 nd C29H33O15 −2.4 475, 355, 272, 603 Apigenin-C-hexoside-O-deoxyhexoside acetate P. x-kewensis + L51 P 609.1809 249 270, 345 C28H33O15 +0.5 353, 431, 449, 577, 579, 595 Isoorientin-4′-O-rhamnoside-methyl ether P. subpeltata (Swertiajaponin 4′-O-rhamnoside) [29] + L52 P 741.2034 251 271, 314 C36H37O17 −0.1 303, 343, 399, 414, 433, 595 Apigenin-C-hexosyl-O-p-coumaroylhexoside P. ambigua + L53 P 773.1527 252 327 C36H37O19 −1.0 303, 407, 415, 471, 743 Unknown P. coriacea − L54 N 561.1717 253 270, 325 C27H29O13 −2.4 295, 457, 543, 441 Chrysin-C-hexosyl-6″-O-deoxyhexoside P. caerulea, P. edulis, P. holosericea + L55 P 431.0981 254 350 C21H19O10 −2.0 413, 327, 357, 287, 329 Luteolin-6-C-deoxyhexuloside P. lutea + L56 P 577.1560 254 270, 340 C27H29O14 −1.6 431, 413, 287, 329 Luteolin-deoxyhexuloside-O-deoxyhexoside (isomer) P. subpeltata + L57 P 457.2074 258 Nd C22H33O10 −0.7 209, 248, 287, 303, 417 Unknown P. x-kewensis, P. incarnata + L58 P 459.2227 260 270, 241 C22H35O10 −0.5 211, 249, 258 Unknown P. subpeltata, P. incarnata − L59 N 431.1128 262 269, 347 C21H19O10 +1.0 285, 327, 357, 413 Luteolin-8-C-deoxyhexoside P. edulis + L60 P 757.1980 262 nd C36H37O18 −0.6 287, 309, 392, 407, 449, 727 Luteolin-O-hexoside-O-p-coumaroylhexoside P. coriacea + L61 P 559.1442 267 270, 339 C27H27O13 +0.8 413, 287, 515, 473 Luteolin-O-deoxyhexoside triacetate P. subpeltata − L62 N 561.1717 267 269, 337 C27H29O13 +1.4 457, 399, 295, 325, 253, 381 Chrysin-C-deoxyhexoside-O-hexoside P. subpeltata, P. edulis + L63 P 561.1609 272 270, 339 C27H29O13 +1.6 413, 415, 499, 543, 397, 353 Apigenin-O-deoxyhexoside-O-deoxyhexuloside P. subpeltata − L64 N 415.104 273 269, 330 C21H19O9 +1.6 269, 294, 311, 341, 397 Apigenin-8-C-deoxyhexoside P. edulis al. et Farag M.A. + L65 P 595.1439 279 269, 317 C30H27O13 +0.8 287, 309, 347, 355, 411, 419 Luteolin-O-p-coumaroylhexoside P. microstipula + L66 P 547.1802 281 270, 336 C27H31O12 +2.3 529, 383, 417, 473, 443, 271 Apigenin-C-deoxyhexoside-O-(dideoxy)hexoside P. subpeltata + L67 P 287.0551 284 347 C15H11 O6 −3.6 201, 205, 210, 213 Luteolin P. caerulea, P. lutea + L68 P 575.1395 285 269, 338 C27H27O14 +0.1 429, 411, 339, 311, 273 Unknown flavonoid P. subpeltata oprtv eaoiepoiigadfnepitn of fingerprinting and profiling metabolite Comparative Table 2 (continued)

+/− n Peak no. Mode M Rt (s) UV Molecular Error MS Identification Species formula

+ L69 P 579.1486 294 269, 317 C30H27O12 +2.5 271, 309, 339, 425, 517, 567 Apigenin-O-p-coumaroylhexoside P. microstipula − L70 N 327.2167 311 nd C18H31O5 +2.6 113, 215, 257, 269, 301, 318 Trihydroxy-octadecadienoic acid P. incarn ata, P. helleri, P. quadrangularis, P. perfoliata, P. edulis + L71 P 213.1456 317 nd C12H21O3 +8.2 144, 148, 162, 176, 187 Unknown fatty acid P. incarn ata, P. costaricensis, P. coriacea, P. lutea, P. perfoliata, P. edu lis + L72 P 485.1068 321 nd C24H21O11 +1.9 213, 315, 333, 353, 413, 431 Unknown flavonoid P. subpeltata + L73 P 251.1257 324 nd C14H19O4 +6.4 173, 194, 203, 213, 229 Unknown P. serratifolia, P. helleri, P. lutea, P. perfoliata − L74 N 329.2320 327 nd C18H33O5 +1.2 113, 183, 227, 249, 259, 309 Trihydroxy-octadecaenoic acid P. incarn ata, P. edulis + L75 P 311.2194 338 nd C18H31O4 +8.1 253, 271 Dihydroxy-octadecatrienoic acid P. capsularis, P. holosericea, P. edulis − L76 N 287.2215 342 nd C16H31O4 +4.6 175, 241, 248, 251, 256 Dihydroxy-hexadecanoic acid All except P. caerulea, P. racemosa, P. subpeltata, P. quadrangularis

+ Passiflora L77 P 239.1627 413 nd C14H23O3 +6.3 111, 129, 199, 217 Hydroxy-tetradecatrienoic acid All except P. caerulea, P. serratifolia, P. helleri, P. edu lis + L78 P 239.1643 425 nd C14H23O3 +7.8 111, 199, 217 Hydroxy-tetradecatrienoic acid All except P. caerulea, P. serratifolia,

P. helleri leaves − L79 N 293.2120 452 nd C18H29O3 +0.5 113, 171, 235, 249, 265, 275 Hydroxy-octadecatrienoic acid All − L80 N 293.2120 457 nd C18H29O3 +0.5 113, 197, 242, 249, 267, 275 Hydroxy-octadecatrienoic acid All − L81 N 295.2276 481 nd C18H31O3 +0.3 201, 209, 221, 227, 277 Hydroxy-octadecadienoic acid All except P. caerulea, P. capsularis, P. edu lis + L82 P 295.2271 494 nd C18H31O3 −0.8 213, 277 Hydroxy-octadecatrienoic acid P. ambigua, P. coriacea, P. racemosa, P. serratifolia, P. helleri P. quadrangularis, P. edu lis + L83 P 295.2267 501 nd C18H31O3 −0.4 213, 277 Hydroxy-octadecatrienoic acid P. ambigua, P. helleri, P. edulis + L84 P 279.2325 516 nd C18H31O2 −2.8 177, 257, 261 Octadecadienoic acid All + L85 P 510.3904 545 nd C33H52NO3 +7.1 213, 283, 293, 301, 397, 398 Unknown P. edulis + L86 P 280.2633 556 nd C18H34NO −2.5 194, 199 Octadecadienoic acid amide P. ambigua, P. costaricensis, P. incarn ata, P. perfoliata + L87 P 309.2035 558 nd C18H29O4 +7.5 199, 213, 280 Dioxo-octadecadienoic acid P. edulis, P. racemosa, P. incarnata − L88 N 277.2172 572 nd C18H29O2 −0.5 211, 251 Linolenic acid All + L89 P 256.2638 581 nd C16H34NO −0.4 185, 191, 213, 217, 224, 243 Hexadecanoic acid amide (palmitamide) All except P. ambigua + L90 P 282.2792 596 nd C18H36NO −0.5 194, 213, 265 Octadecenoic acid amide (oleamide) All − L91 N 279.2331 604 nd C18H31O2 −2.9 171, 211, 239, 245 Linoleic acid All + L92 P 255.2311 611 nd C16H31O2 +1.4 213, 219, 237 Palmitoleic acid P. ambigua, P. helleri, P. lu tea − L93 N 255.2311 632 nd C16H31O2 +0.7 165, 183, 223, 239, 240, 241 Palmitic acid All + L94 P 284.2932 639 nd C18H38NO +4.3 195, 213, 257, 265 Stearic acid amide (stearamide) All − L95 N 281.2484 641 nd C18H33O2 +0.6 113, 165, 171, 239 Oleic acid P. edulis, P. helleri − L96 N 283.2484 680 nd C18H35O2 +5.7 113, 165, 183, 223, 239 Stearic acid P. edulis + L97 P 338.4322 695 nd C22H44NO −0.8 213, 228, 250 Docosenoic acid amide (docosenamide) All M.A. Farag et al.

Fig. 1 UPLC-ESI-MS total ion current chromatograms of the methanol extracts of P. incarnata leaf (A), P. edu lis (B), and P. caerulea (C). Chromatographic conditions are described under “Experimental” section. The identities, Rt values, and basic UV and MS data of all peaks are listed in Table 2 most intense fragment results from the loss of the sugar unit, L48, L51, L55, L59, L61, and L65), and chrysin containing i.e., [M-162]+/− (hexoses), [M-132]+/− (pentose), [M-146]+/− peaks (L18, L42, L54, and L62). Of these the annotated peaks (deoxyhexose), or [M-130]+/− ((dideoxy)hexose), or from mo- (L8, L12, L17, L19, L29, L30, L32, L40, L43, L46, L48, L50, lecular ions [36]. Nevertheless, the dominant fragmentation L52, L54-56, L59-63, L65, L66, L69, L70, L74, L75, L77– pathway of C-glucosyl flavonoids includes loss of water 83, L86, L87, L89, L90, L94, and L97), the respective sug- [M-18]+/− and cross-ring cleavages [(O-C1 and C2-C3) or gested compounds are for the first time reported in Passiflora [(O-C1 and C3-C4)] of the sugar units, namely, [M-120/ species. To the best of our knowledge, this is the first compre- 90]+/− for C-hexosides, [M-90/60]+/− for C-pentosides, and hensive profile of Passiflora secondary metabolites that help [M-104/74]+/− for C-deoxyhexosides [37, 38]. Additionally, provide chemical-based evidence for its biological effects. fragment ions [Agl + 41/71]− in mono-C- and [Agl + 83/ The following paragraphs show the details used for metabo- 113]− for di-C-glycosides which represent the aglycone plus lites structural elucidation. the residues of the sugars that remained linked to it identified − − the type of aglycone, i.e., [311/341] and [353/383] for Identification of luteolin C-flavonoid conjugates apigenin, [327/357]− and [369/399]− for luteolin, and [295/ − − 325] and [337/367] for chrysin mono-C and di-C-glyco- TheMS/MSspectraofpeaks(L4)[m/z 609.1558 − − sides, respectively [39]. Such ions were identified in apigenin (C27H29O16) ], (L8) [m/z 579.1495 (C26H27O15) ], and − containing peaks (L6, L11, L17, L22, L27, L29, L31, L38, (L32) [m/z 563.1391 (C26H27O14) ] all exhibit fragmentation L49, L50, L52, L63, L64, L66, and L69) and in luteolin con- patterns of luteolin-di-C-glycoside owing to the presence of jugates in peaks (L4, L8, L9, L12, L13, L15, L16, L20, L21, ions [369]− and [399]−. Peak (L4) showed fragment ions at m/ L23, L24, L25, L28, L32, L34, L36, L37, L40, L41, L44, z 489 [M-120-H]−,andm/z 519 [M-90-H]− was assigned as Comparative metabolite profiling and fingerprinting of Passiflora leaves

Fig. 2 Structure of the major secondary metabolites detected in Passiflora species methanol extracts: (A) cyanogenic glycosides, (B) harmine alkaloid, (C) apigenin derivatives, (D) luteolin derivatives. Metabolite numbers follow those listed in Table 2 for metabolite identification using UPLC-PDA-MS luteolin 6,8-C-di-glucoside (lucenin-2) reported to occur in Such ion showed less intensity in orientin (L15, 8-C- − P. edulis [22, 40]. Peak (L32) [m/z 563.15122 (C26H27O14) ] glucusyl luteolin). Both isomers are widely distributed in ge- exhibiting a base peak at m/z 459 [M-104-H]− characteristic nus Passiflora [25]. Derivatives of orientin and isoorientin for 6-C-linked deoxyhexose, and a product ion at m/z 503 were also identified in peaks (L16) [m/z 463.1243 − + + [M-60-H] indicative of C-pentosides was annotated as (C22H23O11) ] and (L48) [m/z 493.1337 (C23H25O12) ]. luteolin-6-C-deoxyhexoside-8-C-pentoside (ESM Fig. S1). Peak (L16) was characterized as orientin-4′-methyl ether from − + Peak (L8) [m/z 579.1495 (C26H27O15) ] showing fragments its peak ions at m/z 343 [M-120 + H] and m/z 449 [M-14 + at m/z 459 [M-120-H]−, 489 [M-90-H]−, and 519 [M-60-H]− H]+, whereby the latter is indicative for the loss of a methyl was annotated as luteolin-6-C-hexoside-8-C-pentoside. group [26]. In contrast, mass fragments in (L48) at m/z 463 Similarly, other identified C-glycosides included luteolin-6- [M-30 + H]+ and m/z 449 [M-30-14 + H]+ indicating loss of C-glucoside (isoorientin L13) and luteolin-8-C-glucoside methoxyl plus methyl groups led to its assignment as − − (orientin L15), with [M-H] of 447.1030 (C21H19O11) and a methoxyluteolin-C-hexoside-methyl ether. Peaks (L41) [m/z − − base peak at m/z 327 [M-120-H] characteristic for C- 431.1018 (C21H19O10) ]and(L59)[m/z 431.1128 − hexosides. Differentiation between the two isomers depends (C21H19O10) ] exhibit similar molecular formulae and an in- on the high abundance of product ion at m/z 429 [M-18-H]− in tense peak at m/z 327 [M-104-H]− indicative for a C- (L13) indicating flavone-6-C-hexoside (isoorientin), due to deoxyhexose residue in these flavonoids. Such isomers could the formation of an additional hydrogen bond between the be distinguished by an intense water loss signal (−18 amu) for 2″-hydroxyl group of the sugar and the aglycone 5- or 7- 6-C-glycoside in peak (L41), albeit the same peak is also hydroxy group which confers additional rigidity [38, 41]. detected but at much lower intensity in (L59), indicative of M.A. Farag et al.

+ + flavone-8-C-glycoside. Eventually, (L41) was assigned as (C30H27O12) ] both show a fragment ion [M-308 + H] for luteolin-6-C-quinovoside or luteolin-6-C-fucoside, previously the loss of coumaroyl hexose, viz., 287 [M-308 + H]+ and 271 reported in P. edulis [33, 40], whereas peak (L59) was identi- [M-308 + H]+, respectively. Accordingly, these two peaks fied as luteolin-8-C-deoxyhexoside. Similarly, (L55) [m/z were assigned as luteolin-O-(4-coumaroyl) hexoside and + 431.0981 (C21H19O10) ] shows a similar fragmentation pat- apigenin-O-(4-coumaroyl) hexoside. tern as (L59) with 2 amu mass difference relative to the daugh- ter ion, i.e., m/z 329 [M-102 + H]+ revealing for a 3″-keto Identification of C/O-flavonoid conjugates group in its sugar moiety (deoxyhexulose) and was assigned as luteolin-6-C-deoxyhexuloside. A total of 21 mixed C, O-glycosylated flavonoids have been identified in this study. The relative abundance of the major Identification of apigenin C-flavonoid conjugates mass fragments in these conjugates allowed for distinguishing O-linked sugar positions to be either bound to a phenolic − The MS spectrum of peak (L6) [m/z 593.1651 (C27H29O15) ] aglycone group or to a C-linked sugar. For example, the oc- exhibits product ions at m/z 503 [M-90-H]− and m/z 473 currence of an abundant ion [(M-132/-146/-162]+/−,(mono-O- [M-120-H]− (ESM Fig. S2) and annotated as apigenin-6,8-C- pentosyl/deoxyhexosyl/hexosyl-C-glycosyl derivatives)] di-glucoside (vicenin-2) previously isolated from several characterizes the O-glycosylation on an aglycones phenolic − Passiflora spp. [9]. Peak (L11) [m/z 563.1512 (C26H27O14) ] group. The preferential fragmentation leading to the shows typical fragmentation patterns of flavone-C-hexosides as [M-(132/-146/-162)-18]+/− fragment is characteristic of 2″- evident from fragments m/z 473 [M-90-H]− and m/z 443 O-glycosyl-C-glycosyl derivatives, whereas 6″-O-glycosyl- [M-120-H]− (ESM Fig. S3); additionally, the appearance of a C-glycosyl conjugates are additionally characterized by the product ion at m/z 503 [M-60-H]− characteristic for C-pentosides loss of O-sugar moiety by cross-ring cleavages of the C-gly- and the higher abundance of C-hexose fragment relative to cosyl residue, e.g., [M-162-120]+/− [39]. These MS data char- C-pentose indicate hexose attachment at the C-6 position in acterize peaks (L9, L12, L17, L22, L31, L36, L50, L62, and (L11) and was thus annotated as apigenin-6-C-glucoside-8-C- L66) as C-glycosyl flavones with additional O-glycosylation arabinoside (schaftoside) previously isolated from P. in carn ata at the aglycone moiety, peaks (L20, L40, L44, and L51) en- [24]. Isovitexin (apigenin-6-C-glucoside) was identified in (L27) compass 2″-O-linked sugar moieties, whereas peaks (L38 and − [m/z 431.0971 (C21H19O10) ] from its fragmentation pattern, L54) contain 6″O-glycosyl-C-glycosyl derivatives. typical of mono-C-hexosyl flavones, viz., m/z 341 [M-90-H]− and 311 [M-120-H]−[25]. Peak (L29) [m/z 445.1299 Identification of luteolin-C/O-flavonoid conjugates − (C22H21O10) ] exhibits similar fragmentation patterns with a n − − n 14-amu difference in its MS ions, i.e., m/z 355 [M-90-H] ,a Peak (L9) [m/z 609.1632 (C27H29O16) ]showingMS ions at base peak at m/z 325 [M-120-H]−, and an additional ion at m/z m/z 519 [M-120-H]−, 489 [M-90-H]−,and447[M-162-H]− 430 [M-14-H]− assigned as apigenin-C-hexoside methyl ether. was assigned to isoorientin-4′-O-glucoside, previously reported in Passiflora spp. [23]. Similarly peaks (L36) [m/z − Identification of chrysin C-flavonoid conjugates 593.1651 (C27H 29O 15 )], (L12) [m/z 595.1651 + + (C27H31O15) ], and (L44) [m/z 595.1647 (C27H31O15) ]ex- The ESI-MS spectrum of peak (L18) [m/z 577.1556 hibit typical fragmentation patterns of C, O-glycoside, yet − − (C27H29O14) ] exhibits product ions at m/z 487 [M-90-H] with different O-sugar units (deoxyhexose) accounting for and m/z 457 [M-120-H]− assigned as chrysin-6,8-C-di- the appearance of ion at m/z 447 [M-146-H]− in (L12 and + + hexoside. Peak (L42) [m/z 417.1039 (C21H21O9) ] exhibits L36) and m/z 449 [M-146 + H] in (L44). Peak (L44) shows fragment ions typical of mono-C-glycosides, namely, m/z an intense ion at m/z [M-146-18] characteristic for 2″-O-gly- 297 [M-120 + H]− and an intense ion at m/z 399 [M-18 + cosides and was identified as isoorientin-2″-O-rhamnoside H]+ identified as chrysin-6-C-hexoside. previously reported in Passiflora biflora [29], whereas peaks (L12 and L36) were characterized as two positional isomers of Identification of luteolin and apigenin O-flavonoid luteolin-C-hexoside-O-deoxyhexoside. Peak (L51) [m/z + conjugates 609.1809 (C28H33O15) ] exhibits a fragmentation pattern similar to that observed in peak (L44), though with additional ions + Peak (L63) [m/z 561.1609 (C27H29O13) ] exhibits a base peak resulting from the loss of a methyl group (−14 amu) at m/z 595 at m/z 413 [M-144 + H]+ and m/z 415 [M-146 + H]+ due to the [M-14 + H]+ and was identified as isoorientin-4′-O- respective loss of hexulose and deoxyhexosyl moieties, indic- rhamnoside-methyl ether [29]. Peak (L20) [m/z 593.1656 − ative of a di-O-glycoside and identified as apigenin-O- (C27H29O15) ] was identified as luteolin-C-deoxyhexoside- deoxyhexoside-O-deoxyhexuloside. Peaks (L65) [m/z 2″-O-hexoside from its MSn fragment ions at m/z 431 + − 595.1439 (C30H27O13) ] and (L69) [m/z 579.1486 [M-162-18-H] for 2″-O-hexoside, and m/z 327 [M-162- Comparative metabolite profiling and fingerprinting of Passiflora leaves

104-H]− and m/z 357 [M-162-74-H]− indicative of a C- ions at m/z 295 [M-162-104-H]− and m/z 457 [M-104-H]− deoxyhexosyl flavone (ESM Fig. S4). Other identified C/ O- characteristic for flavone-C-deoxyhexosides, and eventually glycosides include luteolin-8-C-6″-deoxy-3″-hexuloside-2″- led to the assignment of (L62) as a chrysin-C-deoxyhexosyl- O-rhamnoside (L40) [32]. O-hexoside.

Identification of apigenin-C/O-flavonoid conjugates Identification of cyanogenic glycosides

+ Peaks (L17) [m/z 741.2260 (C33H41O19 )] and (L22) [m/z Two cyanogenic (mandelonitrile) di-O-glycosides were de- − 563.1512 (C26H27O14 )] show fragment ions at m/z tected, namely, (L2 and L10) in P. edulis.Identificationwas [M-162-146-120]+/− and [M-132-120]+/−, characteristic based on fragmentation pathways which include the loss of a of a C-sugar moiety, i.e., m/z 313 and m/z 311, respec- cyanide group (−26 amu) in their MS/MS spectra. In details, − tively, and an extra ion due to loss of O-linked sugar peak (L2) [456.1499, C20H26NO11 )] with a fragment ion at residue (L17) exhibited fragment ions at m/z 595 m/z 431 for the loss of a cyano group [M-26-H]−, m/z 294 for a [M-146 + H]+ and 579 [M-162 + H]+ due to the respec- loss of glucose [M-162-H]−, and a further ion at m/z 269 due tive losses of O-deoxyhexosyl and O-hexosyl units and a to loss of both units [M-26-162-H]− was assigned as coming base peak at m/z 433 [M-146-162 + H]+ (due to loss of from mandelonitrile-O-di-glucoside (2-hydroxy-2-phenyl- both sugar units); the compound was annotated as acetonitrile-O-di-glucoside), reported as a major constituent apigenin-C-hexoside-O-hexoside-O-deoxyhexoside. In in P. edulis [9, 40]. Likewise, (L10) [440.1550 − contrast, peak (L22) (ESM Fig. S5) shows an intense (C20H26NO10) ] was characterized as mandelonitrile-O- MS peak at m/z 431 [M-132-18-H]− indicative for an rutinoside, showing the same pathway, that is, ion signals at interglycosidic 2″-O-pentose and was assigned as m/z 415 [M-26-H]− and m/z 294 [M-146-H]− from terminal isovitexin-2″-O-xyloside reported for Passiflora rhamnose loss [40, 42]. serratifolia L. [27]. Isovitexin-4′-O-rhamnoside-7-methyl + ether (L31) [m/z 593.187 (C28H33O14 )] previously iso- Identification of fatty acid conjugates lated from P. biflora [29] was identified from its frag- + ment ions at m/z 473 [M-120 + H] , m/z 431 [M-162 + In the second half of the chromatographic run (Rt 400–800 s), H]+,andm/z 447 [M-146 + H]+ for C-hexose, O-hexose, the ESI-MS spectra revealed many unsaturated fatty acids, and O-deoxyhexose units, respectively. Peak (L38) [m/z namely, linolenic acid (L88), linoleic (L91), palmitoleic acid − n 577.1557 (C27H29O14) ] exhibits MS fragments at m/z (L92), and oleic (L95), as evident from their high resolution 415 [M-162-H]− for the loss of O-hexose moiety and a masses at m/z 281.2484, m/z 279.2331, m/z 277.2172, and m/z − − base peak at m/z 311 [M-162-104-H] indicative of C- 255.2311 with predicted molecular formulae of (C18H33O2) , − − + deoxyhexoside assigned as apigenin-C-deoxyhexoside-O- (C18H31O2) ,(C18H29O2) , and (C16H31O2) , respectively. hexoside [31, 40]. Similarly, peak (L66) [m/z 547.1802 MS signals were also assigned to the saturated palmitic + (C27H31O12) ] exhibits a similar fragmentation pattern, (L93) and stearic acid (L96), based on their molecular ions though with a different O-linked sugar moiety, a at m/z 255.2311 and 283.2484 with corresponding molecular − − (dideoxy)hexose as is evident from the loss of 130 amu formulae of (C16H31O2) and (C18H35O2) , respectively. at m/z 417[M-130+H]+ and was assigned to apigenin- Additionally, some new hydroxylated fatty acids could also C-deoxyhexoside-O-(dideoxy)hexoside. Another apigenin be tentatively identified (L74, L75, L77, L78, L79, L80, L81, + conjugate (L50) [m/z 621.1822 (C29H33O15 )] shows, in L82, and L83). In detail, (L79, L80, L82, and L83 isomer addition to sugar losses of 146 and 120 amu, a loss of peaks) [m/z 293.2120] and peak (L81) [m/z 295.2276] show − 42 amu indicative of an acetyl group and consequently predicted molecular formulae of (C18H29O3) and − led to its identification as apigenin-C-hexoside-O- (C18H31O3) , respectively. A mass difference of 2 amu be- deoxyhexoside acetate. tween the two sets of compounds revealed an extra double bond and led to their identification as hydroxy- Identification of chrysin-C,O-flavonoid conjugates octadecatrienoic and hydroxy-octadecadienoic acids. Several fatty acid amides were also identified from their even mass − Peak (L54) [m/z 561.1717 (C27H29O13) ]givesm/z 415 weights indicating the presence of a nitrogen atom in [M-146-H]− and m/z 325 [M-146-90-H]−, typical of flavone- octadecadienoic acid amide: (L86) [m/z 280.2633 − + C, O-glycoside, a base peak at m/z 295 [M-146-120-H] ,in- (C18H34NO) ], hexadecanoic acid amide (palmitamide); + dicative of 6″-O-linked sugar, identified (L54) as chrysin-C- (L89) [m/z 256.2638 (C16H34NO) ], octadecenoic acid amide + hexoside-6″-O-deoxyhexoside. (L62) [m/z 561.1717 (oleamide); (L90) [m/z 282.2792 (C18H36NO) ], stearic acid − − + (C27H29O13) ] exhibits a distinct ion at m/z 399 [M-162-H] amide (stearamide); (L94) [m/z 284.2932 (C18H38NO) ]and indicating a phenolic linked O-hexose, in addition to product docosenoic acid amide (docosenamide); and (L97) [m/z M.A. Farag et al.

+ 338.4322 (C22H44NO) ]. This is the first report for the pres- P. edulis ence of oxygenated fatty acids and fatty acid amides in P. edulis Passiflora species and suggests that UPLC-MS represents a P. x-kewensis useful technology for fatty acid profiling in Passiflora spp. P. helleri The detailed analysis and full structural elucidation to determine oxygenation position in oxygenated fatty acids are still P. perfoliata underway. There is an increasing interest in hydroxylated fatty P. ambigua acids, due to their anti-inflammatory, antimicrobial, and cyto- P. subpeltata toxic activities [43, 44]. Fatty acid amides have emerged as an P. microstipula intriguing family of diverse, mammalian neuroactive lipids. P. costaricensis The anxiolytic-like effect of oleamide was studied in several experimental models of anxiety in group-housed and socially P. coriacea isolated mice and was found to have an anxiolytic-like effect 1b P. racemosa suggesting that fatty acid amides might be involved in the P. holosericea regulation of anxiety-related behavior in mice [45, 46]. P. quadrangularis Whether fatty acid amides also contribute to Passiflora anxiolytic effect in animals has yet to be reported. P. serratifolia P. incarnata Identification of alkaloids P. incarnata P. incarnata Harman alkaloids (Fig. 2b) are reported in Passiflora spp. [25] 1a P. caerulea though at much lower levels compared to flavonoids. Alkaloids exhibit a higher response in positive versus negative P. lutea ionization which warrants for employing MS analysis in pos- P. capsularis itive mode. P. incarnata analyzed in positive mode showed Fig. 3 Hierarchical cluster analysis (HCA) of Passiflora species based trace levels of harmine alkaloid (peak L1, Table 1)detected on group average cluster analysis of mass spectrometric biochemical only in its chloroform fraction from which it was concluded profiles (n =3) that the anti-anxiety effect of P. incarnata may be due to constituents other than the harman alkaloids. To confirm that All other Passiflora spp. were clustered in group 1b. It should elution method used herein did not preclude for harman alka- be noted that within a species, separation based on geograph- loid detection, harmine standard was analyzed under same ical region or growing habitat could not be observed from conditions and was detected at a level up to 100 ng/μl[40]. HCA, i.e., P. incarnata samples derived from different geographical sites (Table 1) were all clustered together. Regarding Unsupervised multivariate data analysis of Passiflora geographical origin, the same observation was made in metabolite profiles via UPLC-MS P. edulis specimens suggesting that the profiling method applied describes signatures inherent to taxonomic divisions and Although different metabolite patterns could be observed by that these can help to reveal genetic relationships among spec- visual inspection of UPLC-MS traces from different species imens. The tight clustering in P. incarnata and P. caerulea (Fig. 1a–c), PCA and HCA were attempted as more holistic individual specimens suggests that they share comparable sec- approaches to test for possible heterogeneities among acces- ondary metabolite profiles. As a note of caution, it should be sions and for the chemical composition-based classification of mentioned that the overall chemical similarity does not nec- Passiflora species. PCA and HCA are unsupervised clustering essarily define similar bioactivity profiles, but single constit- methods, requiring no knowledge of the data set and act to uents responsible for the activity may not (significantly) con- reduce the dimensionality of multivariate data while preserv- tribute to one of the principal components, even if that is ing most of the variance within; such methods are increasingly statistically unlikely. applied for the analysis of herbal drugs [47]. HCA applied to PCAwas further applied to the matrix of Passiflora UPLC- the total ion current chromatogram (TIC) data from the qTOF-MS peak intensities from TICs. The metabolome clus- UPLC-qTOF-MS analysis retrieved a dendogram with two ters were located at different points in the 2D space described distinct clusters, of 4 (group 1a) and 13 (group 1b) genotypes by two vectors, that is, principal component 1 (PC1), account- (Fig. 3). Examination of 1a revealed that P. caerulea is the ing for 33 % of variance, and PC2 which explained 14 % of species chemically most closely related to official drug source the variance (Fig. 4a). Generally, most biological replicates P. incarnata, followed by Passiflora lutea and Passiflora from each sample clustered together and were separated from capsularis clustering together in a subdivision of group 1a. other genotypes, confirming the method’s consistency and Comparative metabolite profiling and fingerprinting of Passiflora leaves

Fig. 4 UPLC-qTOF-MS (m/z a 100–1000) principal component analyses of different Passiflora samples analyzed by UPLC- qTOF-MS (n = 3). The metabolome clusters are located at the distinct positions described by two vectors of principal component 1 (PC1 = 33 %) and principal component 2 (PC2 = 14 %). (A) Score plot of PC1 versus PC2 scores. (B) Loading plot for PC1 (33 %) and PC2 (14 %) with contributing mass peaks and their assignments, with each metabolite denoted by b its mass/Rt (s) pair. It should be noted that ellipses do not denote statistical significance but are rather for better visibility of clusters as discussed

repeatability of the data used in this study. In PC analyses of will eventually allow to determine whether the P. incarnata the MS data, most accessions sorted to the negative (left) side flavonoid fingerprints are sufficiently unique to be markers oftheverticallinerepresentingPC1,whereasP. incarnata, that allow to distinguish it from its allied most similar species. P. caerulea, P. capsularis,andP. lutea samples were placed Such information is of value for future quality control analysis on the right side. Along PC2, samples for P. serratifolia and and for adulteration detection in the important drugs for which Passiflora quadrangularis were the most distant with positive its extracts are used. An OPLS-DA model was constructed score values (Fig. 4a). Examination of the loading plot reveals with P. incarnata (ESM Fig. S6) against all other coral that MS variables refer mostly to signals of C-flavonoids Passiflora species present in its class group. The (Fig. 4b). In detail, schaftoside was more enriched in P. incarnata model shows one orthogonal component with P. incarnata and P. lutea, isovitexin in P. incarnata and R2 = 0.83 (and explains 83 % of the total variance), prediction P. caerulea, and isoorientin in P. capsularis and goodness parameter, Q2 = 0.79, and a cutoff value of P <0.05 P. incarnata. The distant clustering of P. serratifolia and (ESM Fig. S6A).Aparticularlyusefultoolthatcomparesthe P. quadrangularis was mostly attributed to an abundance of variable magnitude against its reliability is the S-plot obtained isovitexin-2″-O-xyloside in these species. It should be noted by the OPLS-DA model and represented in ESM Fig. S6B, that MS signals for fatty acids and nitrogen containing fatty where axes plotted from the predictive component are the acidsdetectedbyMS(Table2) did not contribute for segre- covariance p[1] against the correlation p(cor)[1]. For the indi- gation in PCA loading plots along PC1 in the UPLC-MS cation of plots with retention time m/z values, a cutoff value of dataset, suggestive that they are present at comparable levels P < 0.05 was used. Compared with all other Passiflora spe- in all Passiflora or at least their extracts. cies, P. incarnata is particularly enriched in C-flavonoids, i.e., isovitexin and isoorientin, though it should be noted that these Supervised OPLS-DA of Passiflora metabolite profiles via constituents were found in all samples and therefore cannot UPLC-MS serve as a chemical marker for P. in ca rn at a. To distinguish more precisely between P. incarnata and P. caerulea, which In spite of the clear separation observed in PCA analyses of cluster together in PCA and HCA analysis (Figs. 3 and 4), Passiflora samples based on genotype, supervised OPLS-DA supervised OPLS-DAwas used to build a classification model was attempted to build an even better classification model that to distinguish between them with the derived score plot. This M.A. Farag et al. shows a clear separation between both samples (ESM spectrum from P. incarnata.The1H-NMR spectrum of Fig. S7). The OPLS-DA score plot explains 78 % of the total P.incarnata L. leaf samples shows two main regions. The first variance (R2 = 0.78) with the prediction goodness parameter region (δ1H 0.0 to 6.0 ppm) showed the most intense signals Q2 = 0.69. The S-plot derived from OPLS-DA reveals that mainly attributed to primary metabolites. Nevertheless, the P. caerulea samples, compared with P. incarnata, contain downfield region (δ1H 6.0 to 9.0 ppm) was characterized with more luteolin-C-deoxyhexoside-O-deoxyhexoside, apigenin- much less intense resonances related to C-flavonoids and in C-deoxyhexoside-O-pentoside, and chrysin-C-hexoside-6″- accordance with the LC-MS analysis (Fig. 1). It should be O-deoxyhexoside. noted that only very few peaks in the P. incarnata 1H-NMR spectra could be readily assigned, e.g., fatty acids (N1) by the Metabolite fingerprinting of Passiflora species via 1D two signals at δ1H 1.31–1.33 and δ1H 0.88–0.91 ppm for and 2D NMR methylenes and the terminal methyl groups, respectively. However, the identification of most of the compounds was The use of NMR fingerprinting has not been used for the mainly based on the analysis of 2D NMR spectra. analysis of Passiflora.WechoseP. incarnata to demonstrate Structures of metabolites identified in P. incarnata by the NMR spectroscopic characterization of its metabolites. NMR are shown in ESM Fig. S8. The 1H-NMR spectrum Peaks were assigned on the basis of the comparison with the revealed for two triplets at δ1H2.81(J =6 Hz) and δ1H chemical shifts of standard compounds and 2D NMR using 0.97 ppm (J = 7.4 Hz), which upon examination of the 1H-1H-correlation spectroscopy (COSY), 1H-13C HSQC, and HMBC spectrum were readily identified as the allylic methy- HMBC and in comparison with spectra of reference standards lene and terminal methyl of ω-3 fatty acids, viz., linolenic [15, 20]. Chemical shifts of metabolites that were identified acid (N2) (Fig. 5a and ESM Figs. S9 and S10). Additionally, are listed in Table 2. Figure 5 shows a representative 1H-NMR ω-6 fatty acid, that is, linoleic acid (N3), was also

Fig. 5 1H-NMR spectrum of P. incarnata leaf methanol extract showing are labeled as follows: fatty acids (N1), ω-3-fatty acids (N2), ω-6-fatty characteristic signals for primary and secondary metabolites: (A) δ 0.5– acids (N3), sucrose (N4), GABA (N6), vitexin (N7), and isovitexin (N8). 5.5 ppm, (B) δ 6.0–8.35 ppm (expanded). Peaks assigned in the spectrum The assignments to NMR peaks are listed in Table 3 oprtv eaoiepoiigadfnepitn of fingerprinting and profiling metabolite Comparative

Table 3 Resonance assignments with chemical shifts of constituents identified in 600 MHz 1H-NMR, HSQC, and HMBC spectra of P. incarnata leaf extract (methanol-d4)

Metabolite Assignment δ1Hinppm δ13C in ppm (HSQC) HMBC correlations δ13C in ppm

Fatty acids (N1) C-2 2.30 (br. m) 33.3 24.4 (C-3), 28.6 (CH2)n, 173.4 (C-1) C-3 1.60 (br. m) 24.4 28.6 (CH2)n, 33.4 (C-2), 173.4 (C-1) (CH2)n 1.31–1.33 br. s 28.6–28.8 24.5 (C-3), 28.7 (CH2)n, 33.5 (C-2) (t-CH3)0.88–0.91 (br. m) 12.6 21.8, 31.4 ω-3 fatty acids (N2) and Olefinic Hs 5.30–5.36 127–129 24.7 (C-11/C-14) ω-6 fatty acids (N3) (overlapped) ω-3 fatty acid (N2) C-11/C-14 (ω-3) 2.81 t (J = 6 Hz) 24.7 126.5, 127.5, 129.5, 131.2 (C-9, C-10, C-11, C-13)

C-18 (t-CH3)0.97t(J = 7.4 Hz) 12.8 19.9 (C-17), 131.2 (=CH–) Passiflora ω-6 fatty acid (N3) C-11 (ω-6) 2.77 t (J = 6 Hz) 24.8 127.4 (C-9/C-13), 129.3 (C-10/C-12) C-18 (t-CH3) 0.88–0.91 (br. m) 12.6 21.8 (C-17), 31.4 (C-16) Sucrose (N4) C-1 5.39 (J = 3.6 Hz) 91.9 71.6(C-2), 72.3 (C-3), 103.7 (C-1′) leaves C-4′ 3.75 82.0 69.4, (C-6′), 74.0 (C-4′) Glucose (N5) C-1 4.86 98.7 69.8, 70.9 GABA (N6) C-3 1.86 23.1 33.2 (C-2), 39.2 (C-4), 178.7 (C-1) C-2 2.33 33.2 23.1 (C-3), 39.2 (C-4), 178.7 (C-1) C-4 2.95 39.2 23.1 (C-3), 33.2 (C-2) Vitexin (N7) and isovitexin (N8) C-3′/C-5′ 6.94 d (J = 8.7 Hz) 115.4 121.4 (C-1′), 161.4 (C-4′) C-1′′ 4.90 d (J = 10.2) 73.4 81.0 (C-5′′), 78.5 (C-2′′), 107.6 (C-6/8), 160.5 (C-5), 163.4 (C-7) Vitexin (N7) C-6 6.51 s 99.5 107.6 (C-8), 163.4 (C-7), 157.2 (C-9) C-3 6.68 s 102.2 104.4 (C-10), 164.9 (C-2), 121.4 (C-1′), 182.6 (C-4) C-2′/C-6′ 7.90 d (J = 8.7 Hz) 127.9 161.4, (C-4′), 164.9 (C-2) Isovitexin (N8) C-8 6.52 s 93.5 107.6 (C-6), 103.7 (C-10), 163.4 (C-7), 157.2 (C-9) C-3 6.62 s 102.1 103.7 (C-10), 182.6 (C-4), 164.7 (C-2), 121.5 (C-1′) C-2′/C-6′ 7.85 d (J = 8.7 Hz) 127.7 115.4 (C-3′,C-5′), 161.4 (C-4′), 164.7 (C-2), 102.1 (C-3) p-Coumaric acid derivative (N9) (Zform) C-2 5.84 d (J =12.9)a 115.4a – C-3 6.81d(J =12.9)a 142.8a 132.1 (C-1′)a p-Coumaric acid derivative (N9) C-2 6.37 d (J =16.1)a 114.0a 125.6 (C-2′/6′), 158.5 (C-1)a (E-form) C-3 7.64 d (J =16.1)a 144.6a 129.6 (C-1′), 158.5 (C-1)a a Signals were detected in P. coriacea extract M.A. Farag et al. characterized in a similar manner through the appearance of of (E) configuration, and confirmed from their respective two other triplets at δ1H2.77(J = 6 Hz), for the allylic meth- HSQC correlations (Table 2). ylene and δ1H0.89ppm(J = 7.5 Hz) for terminal methyl [20] (Fig. 5a and ESM Figs. S9 and S10). 1H-NMR also confirmed Multivariate data analysis of Passiflora species using the presence of unsaturation in fatty acids evident from proton 1H-NMR multiplets appearing at chemical shift range of olefinic protons at δ1H5.30–5.36 ppm (Fig. 5a), further confirmed by HSQC The performance of multivariate data analyses derived from correlation cross-peaks with carbon signals at δ13C127– both MS and 1H-NMR spectra of Passiflora extracts was 129 ppm. Among other primary metabolites detected in compared in order to better evaluate the classification poten- P. incarnata extract were sucrose (N4) and glucose (N5) and tial of both technologies for this series. HCA was performed γ-amino-butyric acid (N6), summarized with other assign- for all Passiflora samples within the 1H-NMR region of δ 9.0 ments in Table 2.Althoughγ-amino-butyric acid (GABA) is to −0.4 ppm. When compared to the HCA plot obtained from a recognized CNS inhibitory neurotransmitter, it is unlikely to UPLC-MS data (Fig. 3), it is apparent that the NMR result mediate for Passiflora leaf biological effect as it cannot pass (ESM Fig. S16A) was not in agreement, with even specimens through the blood–brain barrier [48]. GABA evaded detection for P. in ca rn at a derived from different origins failing to clus- using UPLC-MS (Table 2) which highlights for the potential ter together as observed in UPLC-MS dataset. Similar results of our comparative MS and NMR approach to profile plant were found for P. edulis which question the validity of the extracts. In P. incarnata, flavonoids occur mostly as flavone NMR-based model to reveal for chemotaxonomic relatedness conjugates, namely, of luteolin and apigenin [9]. Several fla- within that genus or with the selected signal regions. Such vonoids including vitexin and isovitexin were identified in the variability was mainly ascribed to compositional differences 1H-NMR spectrum downfield region via their characteristic in fatty acids and sugars among species, as evidenced from the ring A and B proton signals (Table 2). In detail, the 1H- 1H-NMR loading plot (data not shown), i.e., primary metab- NMR spectrum indicates the presence of an apigenin gluco- olite, the accumulation of which usually is more dependent on side as evident from the glucose anomeric proton at δ1H growth conditions (soil, water, light, temperature) than on 4.90 ppm (d, J = 10.2 Hz) appearing upfield in the chemical specific (genetic) differences. Considering that our goal is a shift range of C-glycosides. This was confirmed from HSQC related to the distribution of Passiflora natural product classes, cross-peak correlation with a carbon signal at δ13C 73.4 ppm i.e., flavonoids, HCA was performed for all samples in a sec- [49]. 4′-Mono-substitution in the flavone B-ring was con- ond step, with the fatty acids and sugar region of 1H-NMR firmed from two doublets at δ1H 6.94 ppm (d, J = 8.7 Hz) spectra excluded from analysis (ESM Fig. S16B)[20]. and δ 7.90 ppm (d, J = 8.7 Hz,), assigned to the ortho-coupled Though clustering was successful in the past in the context protons H-3′/H-5′ and H-2′/H-6′, respectively. In contrast, the of extracts more enriched in secondary metabolites [17], HCA presence of a singlet aromatic proton for H-6 at δ1H6.51ppm results derived from aromatic proton region in Passiflora also and the HMBC correlations from the anomeric proton H-1′′ at did not retrieve a signal that could correlate to the UPLC-MS- δ1H4.90ppm(J = 10.2 Hz) to carbons at δ13C 107.6 (C-8), derived results. In contrast, only from auto-scaled NMR aro- 160.5 (C-5), and 163.4 (C-7) ppm confirmed the tri- matic region could we retrieve a dendrogram (ESM substituted A-ring structure and glucose attachment to the Fig. S16C) in which P.incarnata specimens clustered together C-6 position. Altogether, these signals of (N7) led to its iden- similar to UPLC-MS data-derived HCA plot (Fig. 3) but much tification as apigenin-8-C-glucoside, vitexin (Fig. 5b and less comprehensively. ESM Figs. S11–S13). Metabolite (N8), apigenin-6-C-gluco- Absolute integral area values were auto-scaled (the mean side, or isovitexin exhibits a similar set of signals, albeit with area value of each bin throughout all samples was subtracted those of B-ring having a slight upfield shift, i.e., at δ1H from each individual bin area and the result divided by the 6.94 ppm (d, J =8.7Hz)and δ 7.85 ppm (d, J = 8.7 Hz,) for standard deviation) prior to multivariate data analysis. This the ortho-coupled protons H-3′/H-5′ and H-2′/H-6′, respec- provides similar weights for all the variables, as described tively (Fig. 5b and ESM Figs. S11–S13). elsewhere [50]. These results point to the advantage of the UPLC-MS analysis revealed unexpected p-coumaroyl de- utilized comparative metabolomic approach to reveal for spe- rivatives in P. coriacea (Table 2), also visible in the NMR cies relatedness from different technology platforms. PCAwas spectra of its extract. Two sets of p-coumaric acid conjugates further performed on the aromatic NMR-scaled data to visu- (N9) were identified from the characteristic doublets of the alize general clustering and trends among all samples on the olefinic protons H-2 and H-3 exerting distinct cross-peaks in scores plot, where PCA results are now in agreement with the a 1H–1H COSY experiment (ESM Figs. S14 and S15). The previous MS-based dataset, regarding the compositional dif- first set, appearing at δ 5.84 and 6.81 ppm (d, J =12.9Hz),is ferences in flavonoids among samples being responsible for indicative of (Z)-oriented olefinic protons, whereas a second species discrimination. The NMR-derived PCA score plot set appears at δ 6.37 and 7.64 ppm (d, J = 16.1 Hz), indicative (ESM Fig. S17A) exhibits two major clear clusters Comparative metabolite profiling and fingerprinting of Passiflora leaves representing the 17 different species examined. Passiflora in any of the examined species (except for trace levels of perfoliata and Passiflora subpeltata are positioned right to harmine in P.incarnata) suggesting that the anxiolytic activity PC1 side, whereas P. coriacea is placed on the far left side. well-known for genus Passiflora is mostly associated with Examination of the loading plot along PC1 (ESM Fig. S17B) compounds other than the alkaloids. In contrast, the enrich- revealed for proton signals of H-3, H-5′, and H-8 in luteolin- ment of fatty acid amides in Passiflora extracts suggests that it C-hexoside-based conjugates appearing at δ 6.56, 7.37, and is these compounds that might contribute to its well-reported 6.51 ppm, respectively, and a distinct signal at δ 5.52 ppm for CNS inhibitory effects. Nevertheless, it should be noted that an anomeric proton of an O-linked sugar, all having positive these are just a hint and will have to be followed by more PC1 values. Other NMR signal with a negative effect on PC1 detailed studies of Passiflora extracts in vivo or ideally by belonged to the doublet signals at δ 5.84 and 6.81 ppm of studies on single, isolated compounds to be conclusive. olefinic protons (Z form) in a p-coumaric acid flavonoid conjugate. Proton resonances at δ 6.37and7.64ppmwere Acknowledgments This work was supported by the Alexander von assigned to the olefinic protons (E) of the same flavonoid. Humboldt-Foundation, Germany to Dr. Mohamed Farag and Prof. Dr. Along PC2, samples for P. incarnata, P. caerulea,and Ludger Wessjohann. Dr. Mohamed Ali Farag acknowledges the funding received by Science and Technology Development fund STDF, Egypt P. edulis were the most distant ones due to enrichment in (grant number 12594). Prof. Dr. Ludger Wessjohann and Dr. Andrea isovitexin, as revealed from PC2 loading plot signals of the Porzel acknowledge partial support from Leibniz-Wettbewerb SAW- aromatic protons H-2′/H-6′ appearing at δ 7.85 ppm. Visual 2015-IPB-2. inspection of 1H and 2D-NMR spectra further showed the presence of aromatic proton signals of luteolin-C-hexoside- Compliance with ethical standards O-deoxyhexoside in P. perfoliata and P. subpeltata samples, Conflict of interest The authors declare that they have no competing olefinic proton signals of luteolin-O-hexoside-O-p-coumaroyl interests. hexoside in P. coriacea, and aromatic signals for isovitexin in P. incarnata, P. caerulea,andP. edulis all in agreement with UPLC/MS-PCA results (Fig. 4).

References Conclusion 1. Simpson MG. Plant systematics. 2nd ed. Burlington, MA: This work represents the first attempt to explore variation in Academic; 2010. secondary metabolites from a wide range of Passiflora species 2. Mabberley DJ. The plant-book: a portable dictionary of the vascular in context to the genotypes and different geographical origins. plants. 2nd ed. New York: Cambridge University Press; 1997. 3. Brasseur T, Angenot L. The pharmacognosy of the passion flower. J It also provides the most detailed map of flavonoid composi- Pharm Belg. 1984;39(1):15–22. tion within that genus. A total of 97 metabolites were detected 4. Bergner P. Passionflower. Med Herbalism. 1995;7:13–4. in 17 species, of which 78 were identified including C/O-fla- 5. Ingale AG, Hivrale AU. Pharmacological studies of Passiflora sp. vonoids, mandelonitrile glycosides, and fatty acid conjugates and their bioactive compounds. African J Plant Sci. 2010;4:417–26. using NMR and UPLC-MS. The abundance of C-flavonoids 6. Lakhan SE, Vieira KF. Nutritional and herbal supplements for anxiety and anxiety-related disorders: systematic review. Nutr J. (41 peaks) in all examined species suggests activation of high 2010;9(42):1–14. doi:10.1186/1475-2891-9-42. abundance of C-glycosyl transferases or of liberal glycosyl 7. Fiebich BL, Knoerle R, Appel K, Kammler T, Weiss G. transferases [10]inPassiflora. The functional characterization Pharmacological studies in an herbal drug combination of St. of these glycosyl transferases (GTs) has yet to be reported for John’sWort(Hypericum perforatum) and passion flower that genus. Coupling these differential metabolite profile data (Passiflora incarnata): in vitro and in vivo evidence of synergy between Hypericum and Passiflora in antidepressant pharmacolog- with gene transcript levels may assist in probing involved ical models. Fitoterapia. 2011;82(3):474–80. doi:10.1016/j.fitote. genes, biosynthetic pathway analyses, and eventually breed- 2010.12.006. ing [51]. By using the comparative metabolomic approach, 8. Sarris J, Panossian A, Schweitzer I, Stough C, Scholey A. Herbal chemotaxonomic relatedness among 17 Passiflora species medicine for depression, anxiety and insomnia: a review of psycho- pharmacology and clinical evidence. Eur Neuropsychopharmacol. could be established with C-flavonoids which are most rele- 2011;21(12):841–60. doi:10.1016/j.euroneuro.2011.04.002. vant for providing the metabolite signature for species classi- 9. Dhawan K, Dhawan S, Sharma A. Passiflora: a review update. J fication. Among the examined Passiflora species, P. caerulea Ethnopharmacol. 2004;94(1):1–23. doi:10.1016/j.jep.2004.02.023. appeared to be the closest to P. incarnata in terms of overall 10. Wessjohann LA, Keim J, Weigel B, Dippe M. Alkylating enzymes. – chemical composition, suggesting it as the best candidate to Curr Opin Chem Biol. 2013;17(2):229 35. doi:10.1016/j.cbpa. 2013.02.016. substitute of the official drug. However, without exactly 11. Aoyagi N, Kimura R, Murata T. Studies on Passiflora incarnata knowing the activity-relevant constituents, this will require dry extract. I. Isolation of maltol and pharmacological action of additional confirmation. Harman alkaloids were not detected maltol and ethyl maltol. Chem Pharm Bull. 1974;22(5):1008–13. M.A. Farag et al.

12. Lutomski J, Malek B. Pharmacochemical investigations on raw 28. Escobar LK, Liu Y-L, Mabry TJ. C-glycosylflavonoids from materials genus Passiflora. 3. Phytochemical investigations on Passiflora coactilis. Phytochemistry. 1983;22(3):796–7. doi:10. raw materials of Passiflora edulis forma flavicarpa 1016/S0031-9422(00)86995-2. (Pharmakochemische Untersuchungen von Drogen der Gattung 29. McCormick S, Mabry TJ. O-andC-glycosylflavones from Passiflora. 3. Mitteilung: Phytochemische Forschung der Drogen Passiflora biflora. Phytochemistry. 1983;22(3):798–9. 10.1016/ aus Passiflora edulis forma flavicarpa). Planta Med. 1975;27(3): S0031-9422(00)86996-4. 222–5. 30. Sakalem ME, Negri G, Tabach R. Chemical composition of 13. Congora C, Proliac A, Raynaud J. Isolation and identification of hydroethanolic extracts from five species of the Passiflora genus. two mono-C-glucosyl luteolins and of the di-C-substituted 6,8- Revista Brasileira De Farmacognosia (Braz J Pharmacogn). diglucosyl luteolin from the leafy stalks of P. incarnanta L. Helv 2012;22(6):1219–32. doi:10.1590/s0102-695x2012005000108. Chim Acta. 1986;69:251–3. 31. Ferreres F, Sousa C, Valentao P, Andrade PB, Seabra RM, Gil- 14. Wolfman C, Viola H, Paladini A, Dajas F, Medina JH. Possible Izquierdo A. New C-deoxyhexosyl flavones and antioxidant prop- anxiolytic effects of chrysin, a central benzodiazepine receptor li- erties of Passiflora edulis leaf extract. J Agric Food Chem. gand isolated from Passiflora caerulea. Pharmacol Biochem 2007;55(25):10187–93. doi:10.1021/jf072119y. Behav. 1994;47(1):1–4. 32. Zhao J, Pawar RS, Ali Z, Khan IA. Phytochemical investigation of 15. Farag MA, Porzel A, Wessjohann LA. Comparative metabolite Turnera diffusa. J Nat Prod. 2007;70(2):289–92. doi:10.1021/ profiling and fingerprinting of medicinal licorice roots using a mul- np060253r. tiplex approach of GC-MS, LC-MS and 1D NMR techniques. 33. Mareck U, Herrmann K, Galensa R, Wray V. The 6-C-chinovoside Phytochemistry. 2012;76:60–72. doi:10.1016/j.phytochem.2011. and 6-C-fucoside of luteolin from Passiflora edulis. 12.010. Phytochemistry. 1991;30(10):3486–7. doi:10.1016/0031-9422(91) 16. Okada T, Nakamura Y, Kanaya S, Takano A, Malla KJ, Nakane T, 83241-c. et al. Metabolome analysis of Ephedra plants with different contents 34. Avula B, Wang YH, Rumalla CS, Smillie TJ, Khan IA. of ephedrine alkaloids by using UPLC-Q-TOF-MS. Planta Med. Simultaneous determination of alkaloids and flavonoids from aerial 2009;75(12):1356–62. doi:10.1055/s-0029-1185577. parts of Passiflora species and dietary supplements using UPLC- 17. Farag M, Mahrous E, Lübken T, Porzel A, Wessjohann L. UV-MS and HPTLC. Nat Prod Commun. 2012;7(9):1177–80. Classification of commercial cultivars of Humulus lupulus L. 35. de Rijke E, Out P, Niessen WMA, Ariese F, Gooijer C, Brinkman (hop) by chemometric pixel analysis of two dimensional nuclear UAT. Analytical separation and detection methods for flavonoids. J magnetic resonance spectra. Metabolomics. 2013;10(1):21–32. Chromatogr A. 2006;1112(1–2):31–63. doi:10.1016/j.chroma. doi:10.1007/s11306-013-0547-4. 2006.01.019. 18. Farag MA, Porzel A, Schmidt J, Wessjohann LA. Metabolite pro- 36. Simirgiotis MJ, Cuevas H, Tapia W, Borquez J. Edible Passiflora filing and fingerprinting of commercial cultivars of L. (hop): a (banana passion) fruits: a source of bioactive C-glycoside flavo- comparison of MS and NMR methods in metabolomics. noids obtained by HSCCC and HPLC-DAD-ESI/MS/MS. Planta Metabolomics. 2012;8(3):492–507. doi:10.1007/s11306-011- Med. 2012;78(11):1242–3. 0335-y. 37. Davis BD, Brodbelt JS. Determination of the glycosylation site of 19. Farag MA, Wessjohann LA. Metabolome classification of commer- flavonoid monoglucosides by metal complexation and tandem mass cial Hypericum perforatum (St. John’s Wort) preparations via spectrometry. J Am Soc Mass Spectrom. 2004;15(9):1287–99. doi: UPLC-qTOF-MS and chemometrics. Planta Med. 2012;78(5): 10.1016/j.jasms.2004.06.003. 488–96. doi:10.1055/s-0031-1298170. 38. Figueirinha A, Paranhos A, Pérez-Alonso JJ, Santos-Buelga C, 20. Farag MA, Porzel A, Mahrous EA, El-Massry MM, Wessjohann Batista MT. Cymbopogon citratus leaves: characterization of flavo- LA. Integrated comparative metabolite profiling via MS and NMR noids by HPLC-PDA-ESI/MS/MS and an approach to their poten- techniques for Senna drug quality control analysis. Anal Bioanal tial as a source of bioactive polyphenols. Food Chem. 2008;110(3): Chem. 2015;407(7):1937–49. doi:10.1007/s00216-014-8432-1. 718–28. doi:10.1016/j.foodchem.2008.02.045. 21. Smith CA, Want EJ, O’Maille G, Abagyan R, Siuzdak G. XCMS: 39. Ferreres F, Gil-Izquierdo A, Andrade PB, Valentao P, Tomas- processing mass spectrometry data for metabolite profiling using Barberan FA. Characterization of C-glycosyl flavones O-glycosyl- nonlinear peak alignment, matching, and identification. Anal ated by liquid chromatography-tandem mass spectrometry. J Chem. 2006;78(3):779–87. doi:10.1021/ac051437y. Chromatogr A. 2007;1161(1–2):214–23. doi:10.1016/j.chroma. 22. Li H, Zhou P, Yang Q, Shen Y, Deng J, Li L, et al. Comparative 2007.05.103. studies on anxiolytic activities and flavonoid compositions of 40. Otify A, George C, Elsayed A, Farag MA. Mechanistic evidence of Passiflora edulis ‘edulis’ and Passiflora edulis ‘flavicarpa’.J Passiflora edulis (Passifloraceae) anxiolytic activity in relation to Ethnopharmacol. 2011;133(3):1085–90. doi:10.1016/j.jep.2010. its metabolite fingerprint as revealed via LC-MS and chemometrics. 11.039. Food Funct. 2015;6(12):3807–17. doi:10.1039/C5FO00875A. 23. McCormick S, Mabry TJ. Flavonoids of Passiflora pavonis. J Nat 41. Abad-Garcia B, Garmon-Lobato S, Berrueta LA, Gallo B, Vicente Prod. 1981;44(5):623–4. doi:10.1021/np50017a025. F. New features on the fragmentation and differentiation of C-gly- 24. Geiger H, Markham KR. The C-glycosylflavone pattern of cosidic flavone isomers by positive electrospray ionization and tri- Passiflora incarnata. Z Naturforsch C. 1986;41:949–50. ple quadrupole mass spectrometry. Rapid Commun Mass 25. Patel SS, Soni H, Mishra K, Singhai AK. Recent updates on the Spectrom. 2008;22(12):1834–42. doi:10.1002/rcm.3560. genus Passiflora: a review. Int J Res Phytochem Pharmacol. 42. Chassagne D, Crouzet J. A cyanogenic glycoside from Passiflora 2011;1(1):1–16. edulis fruits. Phytochemistry. 1998;49(3):757–9. doi:10.1016/ 26. Simirgiotis MJ, Schmeda-Hirschmann G, Borquez J, Kennelly EJ. s0031-9422(98)00130-7. The Passiflora tripartita (banana passion) fruit: a source of bioac- 43. Li SS, Jiang ZM, Thamm L, Zhou GT. 10-Hydroxy-2-decenoic tive flavonoid C-glycosides isolated by HSCCC and characterized acid as an antimicrobial agent in draft keg-conditioned wheat beer. by HPLC-DAD-ESI/MS/MS. Molecules. 2013;18(2):1672–92. J Am Soc Mass Brew Chem. 2010;68:114–8. doi:10.3390/molecules18021672. 44. Martin-Arjol I, Bassas-Galia M, Bermudo E, Garcia F, Manresa A. 27. Ulubelen A, Mabry TJ. C-Glycosylflavonoids of Passiflora Identification of oxylipins with antifungal activity by LC-MS/MS serratifolia. J Nat Prod. 1980;43(1):162–3. doi:10.1021/ from the supernatant of Pseudomonas 42A2. Chem Phys Lipids. np50007a017. 2010;163(4–5):341–6. doi:10.1016/j.chemphyslip.2010.02.003. Comparative metabolite profiling and fingerprinting of Passiflora leaves

45. Waluk DP, Battistini MR, Dempsey DR, Farrell EK, Jeffries KA, Neuropharmacology. 1971;10(1):103–8. doi:10.1016/0028- Mitchell P, Hernandez LW, McBride JC, Merkler DJ, Hunt MC 3908(71)90013-X. (2014) Omega-3 fatty acids in brain and neurological health: chap- 49. Tai Z, Cai L, Dai L, Dong L, Wang M, Yang Y, et al. Antioxidant ter 9—mammalianfattyacidamidesofthebrainandCNS. activity and chemical constituents of edible flower of Sophora Academic Press, Boston. doi:http://dx.doi.org/10.1016/B978-0- viciifolia. Food Chem. 2011;126(4):1648–54. doi:10.1016/j. 12-410527-0.00009-0 foodchem.2010.12.048. 46. Wei XY, Yang JY, Dong YX, Wu CF. Anxiolytic-like effects of 50. van den Berg RA, Hoefsloot HC, Westerhuis JA, Smilde AK, van oleamide in group-housed and socially isolated mice. Prog der Werf MJ. Centering, scaling, and transformations: improving Neuropsychopharmacol. 2007;31(6):1189–95. doi:10.1016/j. the biological information content of metabolomics data. BMC pnpbp.2007.04.008. Genomics. 2006;7:142. doi:10.1186/1471-2164-7-142. 47. Gad HA, El-Ahmady SH, Abou-Shoer MI, Al-Azizi MM. 51. Farag MA, Deavours BE, de Fatima A, Naoumkina M, Dixon RA, Application of chemometrics in authentication of herbal medicines: Sumner LW. Integrated metabolite and transcript profiling identify a – a review. Phytochem Anal. 2013;24(1):1 24. doi:10.1002/pca. biosynthetic mechanism for hispidol in Medicago truncatula cell 2378. cultures. Plant Physiol. 2009;151(3):1096–113. doi:10.1104/pp. – γ 48. Kuriyama K, Sze PY. Blood brain barrier to H3- -aminobutyric 109.141481. acid in normal and amino oxyacetic acid-treated animals.