A targeted prenylation analysis by a combination of IT-MS and HR-MS: Identification of prenyl number, configuration, and position in different subclasses of (iso)flavonoids

MS fragmentation-based guideline identifies prenylated (iso)flavonoids


Introduction
(Iso)flavonoid metabolite profiling in complex crude plant extracts is classically performed with a combination of liquid chromatography (LC) with mass spectrometry (MS), and relies on comparisons with standard compounds and a search in databases and literature. This involves comparing obtained MS information (e.g. parent ions, neutral losses (NLs)) with those of previously isolated and analyzed compounds [1e4]. However, annotations are often tentative, as well-curated databases are not widely available and standards are often not commercially available [5]. Therefore, annotation of these secondary metabolites in complex matrices (such as plant extracts) with LC-MS is still a bottle-neck. In recent years, advanced state-of-the-art metabolite profiling and data analyses methods based on e.g. integrating both LC-MS and nuclear magnetic resonance (NMR) spectroscopy profiling have emerged, enabling unambiguous full characterization of secondary metabolites, including (iso)flavonoids in complex matrices [5,6]. Additionally, LC with high resolution MS profiling provides more indepth metabolite identification, as it enables rapid assignment of elemental formulas to the molecular ions and all fragment ions derived thereof [7]. A major drawback of some of these state-ofthe-art metabolite profiling methods is their high cost and limited availability in most (research) laboratories; therefore, techniques with broader applicability in phytochemistry research are needed. Moreover, these state-of-the-art studies have focused on i.a. analysis of flavonoid glycoconjugates [8] and flavonoid aglycones [7,9], and not specifically on prenylated (iso)flavonoids.
Commonly, separation and identification of prenylated (iso)flavonoids in complex plant extracts is performed with a combination of LC with UVeVis spectroscopy and electrospray ionization (ESI) MS. MS identification of (iso)flavonoids is possible in positive (PI) and negative ionization (NI) modes, based on (i) their specific fragmentation pattern and (ii) NL screening [1e4,20]. Fragmentation of (iso) flavonoids is characterized by the retro-Diels-Alder (RDA) reaction in which the C-ring of the (iso)flavonoid is cleaved, resulting in characteristic A-and B-ring fragment ions that provide information on the number and type of substituents of these rings (Fig. 1D). RDA fragments are also useful for identification of prenylation, as it has previously been observed that fragments i,j A þ Àprenyl moiety and i,j B þ Àprenyl moiety are indicative for A-or B-ring prenylation, respectively [1,21]. It is favorable to perform fragmentation in PI mode, as cleavage of the C-ring of isoflavones is often limited or not observed in NI mode [22]. From previous research, a few specific spectral features in relation to prenylation are known; in PI mode NLs of 56 and 68 u are associated with cleavage of a 3,3-DMA prenyl substituent upon fragmentation, whereas NLs of 15, 42, and 54 u are associated with 2,2-DMP prenylation [20,23e25]. Additionally, Simons et al. established a screening method for identification of prenyl configuration (3,3-DMA chain vs. 2,2-DMP ring) based on the ions corresponding to NLs of 42 and 56 u; a main loss of 56 u indicates chain prenylation and a main loss of 42 u indicates ring prenylation [20]. Moreover, recently it was cautiously proposed that the position of 3,3-DMA prenylation in isoflavones could be determined with NLs of 84 and 98 u, where a major NL of 84 u indicated prenylation on C6, and a major NL of 98 u indicated prenylation on C8 [21]. It was also proposed that the position of 2,2-DMP prenylation in isoflavones can be putatively assigned based on characteristic NLs; a loss of 54 u with >20% relative abundance was characteristic for prenylation on C6, whereas a loss of 54 u with <5% relative abundance was characteristic for prenylation on C8 [21]. However, aforementioned NLs related to prenylation have been identified with various MS ion analyzers with only a limited selection of standard compounds; Fang and co-workers [23] and Xu et al. [24] used quadrupole time of flight MS (Q-TOF-MS) and included 15 and 12 standards, respectively, whereas Zhang and co-workers [25], Simons et al. [20], and Aisyah et al. [21] used ion trap MS (IT-MS) with 10, 1, and 2 standard compounds, respectively. An overview of prior research is shown in Table 1 and Table S1. From Table 1, it is observed that fragmentation in IT-MS and Q-TOF-MS lead to different NLs as a result of differences in the fragmentation mechanisms of Q-CID and IT-CID; IT-MS 2 mainly gives NLs 42 and 56 u, whereas Q-TOF-MS 2 yields additional NLs of 15, 54, and 68 u. Even though this literature provides a starting point about characteristic spectral information on identification of prenyl configuration in (iso) flavonoids and prenyl position in isoflavones by MS, little is known about (i) if NLs of 42 and 56 u can be used to detect other prenyl configurations (besides 3,3-DMA and 2,2-DMP prenylation), (ii) rapid identification of single or double prenylation in (iso)flavonoids, and (iii) identification of prenyl position on other (iso)flavonoid backbones, besides the isoflavone subclass.
In this work, we elucidated characteristic spectral features related to prenylation and based on this, we developed a targeted prenylation guideline that annotates prenyl configuration, prenyl number, and prenyl position on different (iso)flavonoids backbones. For this, we systematically approached the mass spectrometric analysis of 23 prenylated (iso)flavonoid standards (Fig. S1) from different (iso)flavonoid subclasses with different number and prenyl configurations (i.a. 3,3-DMA, 2,2-DMP, 1,1-DMA 2 00 -IPF; the last two being analyzed for the first time in this study). We specifically aimed te develop a widely applicable guideline that can be used with the widely available technique IT-MS, to rapidly annotate prenylation in complex plant extracts. For this purpose, a combination of LC with UVeVis spectroscopy, coupled to ESI-IT-MS with CID fragmentation was used. Additionally, we employed highresolution Orbitrap-MS (ESI-FT-MS) with higher collisional dissociation (HCD) for confirmation of elemental formulas of molecular ions and fragments. As a proof of concept, compositional analysis of prenylated (iso)flavonoids in an extract of Glycyrrhiza glabra roots was performed. We hypothesized that prenylated (iso)flavonoids can be identified in complex crude plant extracts based on common spectral features; (i) prenyl number can be defined by the m/z of the parent compound, (ii) prenyl configuration can be defined by the ratio of relative abundances of NLs 42 and 56 u in MS n , (iii) performing FT-MS 2 analyses to confirm elemental formulas of prenylrelated fragments can be used to ascertain prenyl configuration, and (iv) prenyl position (A-ring or B-ring prenylation) can be defined by analysis of RDA fragments upon fragmentation.

Sample preparation and extraction
Licorice roots were freeze dried prior to milling. Roots were milled in a Retsch SM 2000 (Retsch, Haan, Germany) and subsequently sieved (pore size 250 mm, Retsch) to yield root powder. The resulting powder was extracted via ultrasound-assisted extraction with EtOAc (1:25 [w:w]) in 3 consecutive cycles of 15 min/cycle at 35 C and the supernatants (after centrifugation for 20 min at 4,696 g at RT) of all three cycles were combined to yield the root extract. EtOAc in the root extract was removed under reduced pressure and dried extract was resolubilized in tert-butanol and freeze-dried. Prior to UHPLC-MS analysis, the dried root extract was resolubilized in MeOH to a final concentration of 0.1 mg mL À1 for ESI-FT-MS n or 1 mg mL À1 for ESI-IT-MS n and centrifuged (15,000 g, 5 min, RT) prior to further analysis. Standards were dissolved in MeOH and injected (after centrifugation for 5 min, 15,000 g, RT) at 100 mg mL À1 for ESI-IT-MS n analysis and 3 mg mL À1 for ESI-FT-MS n analysis.

Reversed phase liquid chromatography (RP-UHPLC-PDA)
Samples were separated on a Thermo Vanquish UHPLC system (Thermo Scientific, San Jose, CA, USA) equipped with a pump, degasser, autosampler and photodiode array (PDA) detector. The flow rate was 400 mL min À1 at a column temperature of 45 C. Injection volume was 1 mL. Eluents used were water acidified with

Electrospray ionization ion trap mass spectrometry (ESI-IT-MS n )
Mass spectrometric data were acquired using a LTQ Velos Pro linear ion trap mass spectrometer (Thermo Scientific) equipped with a heated ESI probe coupled in-line to the Vanquish UHPLC system. Nitrogen was used as sheath gas (48 arbitrary units), auxiliary gas (11 arbitrary units), and sweep gas (2 arbitrary units). Data were collected in negative ionization (NI) and positive ionization (PI) mode between m/z 200e1000. Based on experience with fragmentation of prenylated (iso)flavonoids in our laboratory [20,21,27,28], data dependent MS n analyses were performed by collision-induced dissociation with a normalized collision energy of 35%. MS n fragmentation was performed on the most intense product ion in the MS nÀ1 spectrum. Dynamic exclusion, with a repeat count of 3, repeat duration of 5.0 s and an exclusion duration of 5.0 s was used to obtain MS 2 spectra of multiple different ions present in full MS at the same time. Ion transfer tube temperature was 254 C, source heater temperature 408 C, and the source voltage was 3.5 (PI) and 2.5 (NI) kV. Data were processed using Xcalibur 4.1 (Thermo Scientific).

Electrospray ionization hybrid quadrupole Orbitrap mass spectrometry (ESI-FT-MS)
Accurate mass data were acquired using a Thermo Q Exactive Focus hybrid quadrupole-Orbitrap Fourier transform mass spectrometer (FT-MS) (Thermo Scientific) equipped with a heated ESI probe. Samples were separated on a Vanquish UHPLC system (Thermo Scientific), as described above. Prior to analysis, the mass spectrometer was calibrated in PI and NI mode using Tune 2.11 (Thermo Scientific) by injection of Pierce negative and positive ion calibration solutions (Thermo Scientific). Used gas flows and source conditions were the same as for ESI-IT-MS. Full MS and higher energy C-trap dissociation (HCD) fragmentation data were recorded at 70,000 and 35,000 resolution, respectively. Normalized collision energy was 35%. MS n fragmentation was performed on the most intense product ion in the MS nÀ1 spectrum. Data were processed using Xcalibur 4.1 (Thermo Scientific). Table 1 Characteristic neutral losses related to prenyl configuration and position, reported for different MS ion analyzers in positive ionization mode.

Quantification of prenylated (iso)flavonoids in licorice extract
Quantification of (iso)flavonoids was based on UV absorbance at 280 nm. For this, a six-point (0.2e100 mg mL À1 ) calibration curve based on an external standard of glabridin (R 2 ¼ 0.999) was used. Subsequently, content of each compound was corrected for the differences in molar extinction coefficients between the standard and the compounds of interest, using a derivative of Lambert-Beer's law (Eq. (1)).
In which ε (AU/M•cm at 280 nm) is the molar extinction coefficient, C is the molar concentration, Glab is glabridin, and X is the (iso)flavonoid to be quantified. Concentrations of compounds were recalculated to mg per g of dry weight (DW) of the licorice root powder (mg g À1 DW). See Table S2 for an overview of the molar extinction coefficients used for these calculations [26,27,29,30].

Spectral properties of prenylated (iso)flavonoids
A diverse set of 24 standards was selected (Fig. S1), which included different (iso)flavonoid subclasses (isoflavan, isoflavone, isoflav-3-ene, 3-arylcoumarin, pterocarpan, flavanone, chalcone), and various prenyl configurations (chain, furan, and pyran) and positions (C3', C6', C6, and C8). In the paragraphs below, all standards are discussed with respect to their UV absorbance and parent m/z, their characteristic NLs associated with the prenyl moiety, and their RDA fragments and specific NLs associated with prenyl position. An overview of all obtained spectrometric data (UV, NI and PI mode with IT-MS n and FT-MS n ) of standards is listed in Table S5.

(Iso)flavonoid subclass, presence, and number of prenylation
UV absorbance spectra of all standards complied with the (iso) flavonoid subclass specific UV absorbances known from literature; for example, glabridin (isoflavan) showed a maximum UV absorbance at 278 nm, which is in agreement with the typical UV absorbance of isoflavans, which was reported to be between 270 and 285 nm [31], whereas licochalcone A (chalcone) showed a maximum UV absorbance at 378 nm corresponding to the typical band I absorption of chalcones between 340 and 390 nm [32]. Based on our results (Table S3) and previously reported spectral data [33,34], we propose to use a UV absorbance cut-off of at least 240 nm as a first criterion to identify potential (prenylated) (iso)flavonoids. Next, m/z ratios of the parent ions of all standards were compiled; m/z ratios in PI ranged from 323 for glabrene (single prenylated isoflav-3-ene) to 453 for 6 0prenylpiscidone (double prenylated isoflavone) ( Table S5). The smallest natural prenylated flavonoid is 7,8-(2,2-dimethylchromeno) flavone, with a molecular weight of 304 g mol À1 (C 20 H 16 O 3 ) [14,35,36]. Thus, a second criterion was defined based on parent m/z of the compound; m/z ratio of the parent molecule in PI mode <305 indicates a non-prenylated (iso)flavonoid, m/z of 305e373 indicates a single prenylated (iso)flavonoid, and m/z >373 (305þ68) indicates a double prenylated (iso)flavonoid.

Prenyl configuration
All standards were screened for the presence of ions in IT-MS n that corresponded to NLs of 42 and 56 u. An overview of these NLs, with their relative abundances, is shown in  (Fig. 2A2). Molecular formulas of fragments were confirmed by FT-MS 2 (Fig. 2A3). FT-MS 2 fragmentation showed a similar fragmentation pattern as IT-MS 2 with different relative intensities of fragments. Differences in relative intensities of fragments are due to the use of CID in our IT-MS n analyses and HCD in our FT-MS 2 analyses. Additionally, HCD improves fragmentation in the low mass region [37], which can provide useful ions for confirmation and screening purposes to detect prenylated compounds based on this MS 2 fragment. It should be noted that correct identification of prenyl configuration by the ratio of the relative abundances of the ions corresponding to NLs 42 and 56 u was instrument dependent, as it was correctly identified by IT-MS n and not always by HR-MS 2 .
An interesting result was that the ratio of the relative abundances of the ions corresponding to NLs 42 and 56 u did not correctly identify 1,1-DMA prenylation (licochalcone A) and furan ring prenylation (pterocarpans glyceollin V and glyceofuran). For these compounds, high resolution MS 2 was required to identify molecular formulas of fragments. IT-MS 2 fragmentation of licochalcone A showed a major fragment at m/z 297 (NL of 42 u) and a minor fragment at m/z 283 (NL of 56 u) (Fig. 3B1). The ratio 56:42 <1 suggested 2,2-DMP prenylation, however, FT-MS 2 (Fig. 3B3)  . In our analyses, fragment m/z 69 was detected in FT-MS 2 with HCD, but it should be noted that detection of fragment m/z 69 can also be achieved with CID by lowering the activation Q or using pulsed Q CID (PQD), thereby circumventing the "1/3rd rule" [40]. Next, fragmentation behavior of furan ring prenylated pterocarpans glyceollin III (6,7-(2 00 -isopropenyldihydrofuran)), glycollin V (6,7-(2 00 -isoproprenylfuran)), and glyceofuran (6,7-(2 00 -(2-hydroxy-) isopropenylfuran)) was studied to elucidate whether five-membered (furan) prenyl rings fragment similarly to six-membered (pyran) prenyl rings, as suggested by Simons and co-workers [41]. Only for glyceollin III (Fig. S2F1) were not associated with the prenyl (Fig. S2G and S2 Thus, when a complex plant extract is analyzed which potentially contains novel prenylated compounds, or if it is known from literature that furan prenylated compounds may be present, it is valuable to use high resolution MS; we recommend verification of the molecular formulas of the fragment ions with high resolution MS to prevent false positive identification of 3,3-DMA prenylation. Moreover, molecular formulas obtained with high resolution MS can give insights in new compounds in complex extracts. Peak labels show the corresponding fragmentation pathway, fragment (with its molecular formula for FT-MS), and m/z (with error in ppm in parentheses). Fragments were labelled based on a cut-off value at 10% relative abundance; unless they were specific fragments used for screening (including RDA fragments, fragments associated with the prenyl). Structural formulas of compounds are shown in Table S3 (Supplementary information). (Iso)flavonoid subclass, molecular formula, and exact mass of the precursor ions are shown in A.1 and B.1. In B.2 and B.3, specific fragments used for screening are shown in amplified regions of the spectra, displayed in dashed boxes.

Prenyl position
For all tested standards, prenyl position on the (iso)flavonoid backbone (A-or B-ring) and subsequent C6 or C8 prenylation (for Aring prenylated compounds) was identified based on RDA fragments  formulas. An overview of characteristic RDA fragments and NLs is listed in Table S3.
From above results, we conclude that analysis of RDA fragments in IT-MS n , in combination with molecular formula confirmation by FT-MS 2 , is a useful tool to identify A-or B-ring prenylation in different classes of (iso)flavonoids. Fragments i,j A þ Àprenyl and i,j B þ Àprenyl are diagnostic ions, i.e. fragment ion i,j A þ ÀC 4 H 8 or i,j B þ ÀC 4 H 8 confirms 3,3-DMA prenylation on the corresponding ring and fragment ion i,j A þ ÀC 3 H 6 or i,j B þ ÀC 3 H 6 serves the same purpose for 2,2-DMP prenylation. Additionally, we propose that A-or B-ring prenylation can easily be distinguished in IT-MS 2 by the presence of fragment [MþHeC 5 H 8 ] þ in combination with absence of RDA fragments in the case of B-ring 3,3-DMA prenylated isoflavones.
3.1.3.2. C6 or C8 prenylation. In a study by Aisyah and co-workers, it was proposed that NLs 84 and 98 u in IT-MS 3 can be used to identify C6 or C8 3,3-DMA prenylation, respectively, whereas the relative abundance of the ion corresponding to a NL of 54 u was used to distinguish C6 (>20% relative abundance) or C8 (<5% relative abundance) 2,2-DMP prenylation [21]. We compared these proposed NL-related guidelines to the mass spectra of all our A-ring prenylated standards. Analysis of the relative abundance of the ion corresponding to a NL of 54 u ([MþHeC 4 H 6 ] þ ) in IT-MS n correctly identified C8 2,2-DMP prenylation in the C8 2,2-DMP prenylated isoflavans glabridin, hispaglabridin A, and hispaglabridin B; fragments with a NL of 54 u were absent in IT-MS n (Table S3). A limitation in our study was the absence of C6 2,2-DMP prenylated standards; we suggest to verify a NL of 54 u in IT-MS n with authentic standards of e.g. parvisoflavone A (C8 2,2-DMP prenylated isoflavone) and parvisoflavone B (C6 2,2-DMP prenylated isoflavone), which were not commercially available. This could determine with more certainty if a NL of 54 u can distinguish C6 and C8 2,2-DMP prenylation. For isoflavone isomers wighteone (C6 3,3-DMA prenylated) and lupiwighteone (C8 3,3-DMA prenylated), IT-MS 3 was useful to elucidate prenyl position (Table S3); lupiwighteone yielded a major ion at m/z 241 (NL of 98 u) (Fig. 2A2), indicating the loss of C 2 H 2 O from [MþHeC 4 H 8 ] þ . Wighteone did not give m/z 241 in MS 3 , instead an ion at m/z 255 was seen (NL of 84 u), indicating the loss of CO from [MþHeC 4 H 8 ] þ . Both fragments were confirmed with FT-MS 2 (Fig. 2A3). The mechanism behind this difference in fragmentation between C6 and C8 prenylated isomers remains to be elucidated. However, the relative abundances of the ions corresponding to NLs of 84 and 98 u in IT-MS 3 did not always correctly identify C6 and C8 3,3-DMA prenylation (Table S3). Incorrect identification did not seem to be subclass specific, as this was observed for both isoflavone and flavanone standards; the IT-MS 3 spectra of luteone (C6 3,3-DMA prenylated isoflavone) and 6-prenylnaringenin (C6 3,3-DMA prenylated flavanone) showed minor NLs (with <1% relative abundance) of 84 and 98 u with similar intensities (Fig. S3A2 and Fig. 2B2), making it impossible to confirm C6 prenylation.
In order to correctly identify C6 or C8 3,3-DMA prenylation, we tested fragmentation of the standard compounds with lower normalized collision energies (NCE 15, 20, 25, and 35) (Fig. 4). It was shown previously that 8-prenylated flavanones were less likely to lose C 4 H 8 during fragmentation than their 6-prenylated isomers [43,44]. Indeed, when we lowered the NCE from 35 to 20, the relative abundance of fragment [MþHeC 4 H 8 ] þ of the 6-and 8prenylated isomers changed differentially for all tested standards, making them easily distinguishable in IT-MS 2 . For example, 6prenylnaringenin (Fig. 4A) readily lost its C 4 H 8 moiety and yielded fragment [MþHeC 4 H 8 ] þ at m/z 285 with 100% relative abundance, whereas 8-prenylnaringenin showed parent ion [MþH] þ at m/z 341 with 100% relative abundance and fragment [MþHeC 4 H 8 ] þ with 62% relative abundance. Isoflavones isowighteone, lupiwighteone, luteone, and 2,3-dehydrokievitone showed similar fragmentation behavior (Fig. 4B and C). Based on these results, we formulated a new rule of differentiating between C6 and C8 3,3-DMA prenylation; if the relative abundance of [MþH] þ is lower than the relative abundance of [MþHeC 4 H 8 ] þ (at low NCE) then the molecule is C6 prenylated, whereas if the relative abundance of [MþH] þ is higher than [MþHeC 4 H 8 ] þ then the molecule is C8 prenylated. The underlying reason for this difference in relative abundance of fragment [MþHeC 4 H 8 ] þ in IT-MS 2 between C6 and C8 prenylated isomers is still unknown. It could be hypothesized that the hydroxyl groups adjacent to the prenyl moiety at C6 make the prenyl moiety a better leaving group; the C6 chain prenylated compounds (wighteone, luteone, and 6-prenylnaringenin) have two ortho phenolic hydroxyl groups present adjacent to the prenyl, whereas the C8 chain prenylated compounds (lupiwighteone, 2,3dehydrokievitone and 8-prenylnaringenin) have only one. Moreover, when applying this new rule, we recommend optimization of the NCE and the required reduction in NCE for each individual MS system, as the actual CID energy will differ between instruments and may even change over time.

Decision guideline for identification of prenylation configuration, number, and position in (iso)flavonoids
Based on the characteristic spectral properties of prenylated (iso)flavonoid standards (Section 3.1.), a decision guideline was developed in order to quickly identify prenyl number, configuration (Fig. 5A), and position ( Fig. 5B and C) in (iso)flavonoids. Our guideline can be used as follows; when starting at the upper arrow (Fig. 5A), the first criterion is based on UV absorbance (Fig. 5A, "UV absorbance); if the compound has a UV absorbance >240 nm, the second criterion is faced, based on the m/z ratio of the parent molecule (Fig. 5A, "full MS"). Our decision guideline does not take into account O-or C-glycosylated prenylated (iso)flavonoids. For guidelines on identification of O-or C-glycosides, we refer to work by Vukics and co-workers [45] and Pilon et al. [8] that discuss identification of flavonoid glycosides by mass spectrometry. However, we do not expect that glycosylation affects our guideline (besides changing the m/z of parent ion), as sugar moieties are easily cleaved in IT-MS 2 and identified with characteristic neutral losses associated with the glycoside; Jin and co-workers showed that ESI-FT-MS 2 fragmentation (with a CID of 35%) in PI mode of prenylated and glycosylated flavonols yielded fragments [MþHÀsugar] þ , but also [MþHeC 4 H 8 ] þ [46]. Prenylation (the third criterion in our decision guideline) is confirmed based on detection of NLs of 42 and 56 u in PI mode with IT-MS 2 and MS 3 (Fig. 5A, "MS 2 ", "MS 3 "); a ratio 56:42 >1 indicates 3,3-DMA prenylation and a ratio 56:42 <1 2,2-DMP prenylation (Fig. 5A, " Our decision guideline also facilitates identification of prenyl position for single prenylated compounds with 2,2-DMP prenylation and compounds with 3,3-DMA prenylation ( Fig. 5B and C). The golden standard for identifying A-or B-ring prenylation based on MS is analysis of the RDA fragments in IT-MS n ; fragment i,j A þ ÀC 3 H 6 or i,j B þ ÀC 3 H 6 confirms 2,2-DMP ring prenylation and fragment i,j A þ ÀC 4 H 8 or i,j B þ ÀC 4 H 8 confirms 3,3-DMA chain prenylation (Fig. 5B). Additionally, for 3,3-DMA prenylated compounds, a NL of 68 u ([MþHeC 5 H 8 ] þ ) in combination with absence of RDA fragments in IT-MS 2 indicates Bring prenylation (Fig. 5C). With respect to 2,2-DMP ring prenylation on the A-ring (i.e. C6 or C8), the ion that corresponds to a NL of 54 u in IT-MS n with a relative abundance >20% indicates C6 prenylation, whereas a relative abundance <5% indicates C8 2,2-DMP prenylation (Fig. 5C). A relative abundance between 5 and 20% results in an inconclusive outcome. As for the position of 3,3-DMA prenylation on the A-ring, the ratio of relative abundances of ions  (Fig. 5C). It should be noted that identification of C6 or C8 prenylation based on MS fragmentation is only possible for single prenylated compounds. We validated our decision guideline with reported spectral data of standards from literature, which is shown in Table S4 (validation of decision guideline).
To summarize, the entirety of our proposed decision guideline can be used with IT-CID-MS, the majority can be used with FT-HCD-MS (an exception is identification of prenyl configuration, which was found to be instrument dependent, section 3.1.2), and at least part of it is applicable to Q-TOF-MS (validation of decision guideline, Table S4). The full range of the guideline's applicability should be further evaluated in the future by analyses on a wider variety of mass spectrometers.

Annotation and quantification of prenylated (iso)flavonoids in extracts of G. glabra roots
The chromatographic UV profile of EtOAc extract G. glabra roots is shown in Fig. 6A. In total, 33 peaks were selected, based on (1) a cut-off that the peak should account for at least 1% of the total UV area at 280 nm and (corresponding to 69% of total UV) (2) compounds that were tentatively identified previously in G. glabra extracts at our laboratory [20,28]. Using the established fragmentation behavior of different classes of (iso)flavonoids and the prenylation decision guideline (Fig. 5), the 33 chromatographic UV peaks (representing 36 compounds due to co-elution of compounds) were annotated. Separation on RP-UHPLC was used in combination with IT-MS n and FT-MS for high resolution mass  (Table S5, Supplementary Information). (B) Distribution of (iso)flavonoid subclasses in EtOAc extract of G. glabra root, and (C) prenyl configuration (chain ¼ 3,3-dimethylallyl, ring ¼ 2,2-dimethylpyran) of prenylated (iso)flavonoids identified in EtOAc extract of G. glabra roots. determination. UVeVis absorbance was used to determine (iso) flavonoid subclass. Annotations are listed in Table S5. Proposed fragmentation schemes in PI mode (with FT-MS n ) are shown in Figs. S6eS10. The total content of annotated compounds in the EtOAc extract of G. glabra roots was 5.8±0.4 mg g À1 DW, of which 5.0±0.4 mg g À1 DW were prenylated (iso)flavonoids. Content of each individual compound is given in Table S6. To date, no literature is available on prenylated phenolic content in EtOAc extract of G. glabra roots. Nevertheless, in a MeOH extract of G. glabra roots, Cheel and co-workers [47] reported total phenolic and total flavonoid content between 72 and 108 mg g À1 DW and 18e44 mg g À1 DW, respectively [18,47,48]. This higher content is explained by the higher polarity of MeOH compared to EtOAc, leading to extraction of a wider range of non-flavonoid phenolics and non-prenylated flavonoids.

Screening for prenylation in G. glabra root extract
To apply the decision guideline for prenylation (Fig. 5), the EtOAc extract of G. glabra roots was screened for the presence of prenylated (iso)flavonoids in addition to the tentatively annotated compounds (Table S5). This screening was based on screening of the IT-MS 2 and MS 3 chromatograms for presence of NLs of 56 and 42 u. Additionally, we performed NL triggered IT-MS 3 on the ions that corresponded to [MþHÀ42] þ and [MþHÀ56] þ in order to identify double prenylation. In total, 209 peaks were selected based on this NL screening (Table S7), thus revealing a wide variety of possible prenylated compounds that were present <1% abundance. All peaks showed an UV absorbance >240 nm. By following our decision guideline, 13 of these peaks were assigned as nonprenylated (m/z < 305 u).
Of the resulting 196 peaks, 75 peaks were assigned as single prenylated and 104 peaks were assigned as double prenylated. The remaining 17 peaks were most likely double prenylated, of which 4 peaks were at least single chain prenylated, 7 peaks at least single ring prenylated, and for 6 peaks no definite prenyl configuration could be determined (Table S7). Regarding single prenylated compounds, 50 peaks were assigned as chain prenylated and 25 peaks as ring prenylated. For the double prenylated compounds, 69 peaks were assigned as double chain prenylated, 32 peaks as ring and chain prenylated, two peaks as double ring prenylated, and one peak was at least single ring prenylated (Tables S7 and S146); the ratio of relative abundances of NLs 56 and 42 u in IT-MS 3 was 1, leading to an inconclusive outcome.
The distribution of prenyl configuration in all 209 compounds annotated based on this screening is visualized in Fig. S5. We compared this distribution with the distribution observed in the main peaks observed in the extract, which were annotated and quantified (section 3.3, Fig. 6C). Based on this, we concluded that the distribution of prenyl configurations found via these two analytical approaches was generally quite comparable; by using our prenylation guideline, fast screening of prenyl configuration without laborious annotation of compounds gives a reasonably accurate representation of prenyl configuration.
To conclude, by applying our decision guideline for prenyl configuration (Fig. 5A), we rapidly identified 196 prenylated compounds in crude EtOAc extract of G. glabra roots. This indicates that many more unique prenylated compounds are present in G. glabra roots than the~75 that have been annotated so far [18,51,52]. To develop a full picture that also includes identification of prenyl position, we propose that further work is required regarding automation of our decision guideline; this will facilitate fast identification of prenyl number, configuration, and position in complex plant extracts.

Conclusion
In this study, we developed a widely applicable decision guideline that enables identification and characterization of prenylated (iso)flavonoids from different subclasses in complex crude plant extracts. We systematically analyzed fragmentation of prenylated (iso)flavonoids using a combination of IT-MS n with CID fragmentation and high resolution FT-MS with HCD fragmentation in PI mode. Based on this systematic analysis, we elucidated fragmentation pathways and characteristic spectral features of different subclasses of prenylated (iso)flavonoids, as well as fragmentation patterns and corresponding NLs that resulted in development of an annotation guideline. Whereas previous annotation guidelines only facilitated identification of prenyl configuration, our new guideline allows identification of (i) the presence of an (iso)flavonoid backbone, (ii) prenyl number, (iii) prenyl configuration, and (iv) prenyl position, in complex plant extracts by ESI-IT-MS. Moreover, high resolution MS with HCD fragmentation was used to confirm molecular formulas of fragments and led to the new insights, which uncovered inconsistencies in previously proposed annotation guidelines. Structural characteristics were annotated based on: (i) UV absorbance; (ii) the m/z ratio of the parent compound; (iii) the ratio of relative abundances between NLs 42 and 56 u in MS n ; and (iv) RDA fragments, NLs of 54 and 68 u, and the ratio [MþHeC 4 H 8 ] þ /[MþH] þ in MS n . With this guideline, we tentatively identified 196 prenylated (iso)flavonoids in G. glabra root extract. Prenylated (iso)flavonoid content in the EtOAc extract of G. glabra roots was calculated at 5.0±0.4 mg g À1 DW G. glabra roots.
CRediT authorship contribution statement

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.