Feature-based molecular networking and building blocks identification advance novel natural products characterization: Sesquiterpene-chromone hybrids in agarwood as an application

Graphical abstract

Highlights

• •Mass defect filtering and feature-based molecular networking accelerated capturing sesquiterpene-chromone hybrids (SCHs).
• •FEER-MS³ matching strategy facilitated fragment ion identification for SCHs.
• •Twenty-five SCHs were captured and structurally annotated by separately deciphering units A and B.
• •Seven new SCHs, including one bearing C–C linkage between substructures were purified for structural justification.
• •Two new SCHs showed cell-protective, and six exhibited anti-inflammatory activities.

As important sources of new drugs, natural products (NPs) are conceptually biosynthesized from simple structural pioneers (i.e. building blocks). The traditional non-targeted purification strategy extensively suffers from the time-consuming and laborious bottlenecks. Fortunately, liquid chromatography–mass spectrometry/mass spectrometry (LC–MS/MS)-guided separations widely succeed in recent decades. However, it is still challenging for LC–MS/MS to precisely capture new NPs. Efficiently extracting information from the chaotic chemical composition and confident structural annotation are two primary technical barriers for pursuing the interesting structures, particularly those exhibiting trace distributions and high-level structural complexity. Here, to provide accurate guidance for the follow-up phytochemical purification, molecular defect filtering (MDF) and feature-based molecular networking (FBMN) [1] were incorporated to explore NPs and thereafter, bottom-up structural analysis was undertaken through identifying building blocks with full exciting energy ramp (FEER)-MS³ matching. Sesquiterpene-chromone hybrids (SCHs) structurally configured by two building blocks such as units A (chromone) and B (sesquiterpene) [2] in agarwood were characterized as a proof-of-concept. Twenty-five SCHs were captured and identified. Thereof, seven new SCHs were purified with a LC–MS/MS-guided manner and annotated using nuclear magnetic resonance (NMR) spectroscopy to justify the proposed structures. Moreover, their cell-protective and anti-inflammatory activities were evaluated. Together, the incorporation of post-acquisition data processing strategies and FEER-MS³ spectrum matching-assisted building blocks identification facilitated novel NPs exploration and purification.

The step-wise workflow was conducted as follows (Fig. 1A): 1) recording MS/MS spectra to summarize mass fragmentation rules using authentic compounds; 2) filtering SCHs through applying MDF and FBMN to MS¹-MS² dataset; 3) constructing full collision energy ramp (FCER)-MS² spectra [3] to characterize substructures linkage manners; and 4) configuring FEER-MS³ spectra [4] for key 1st-generation fragment ions correlating to units A and B to annotate fragment ions. Thereafter, targeted separation was undertaken to obtain SCHs.

Fig. 1.

Overview of the integrated mass spectrometry-based strategy for the discovery and structural characterization of sesquiterpene-chromone hybrids (SCHs) from agarwood. (A) The workflow to discovery, screening, and identification of SCHs in agarwood. (B) Full collision energy ramp (FCER)-mass spectrometry (MS)², full exciting energy ramp (FEER)-MS³, and structures of aquilacrassnins B and C. (a) FCER-MS² (Left) and structure (Right) of aquilacrassnin B by involving all shaped breakdown graphs of 1st-generation fragment ions. (b) FCER-MS² spectra (Left) and structure (Right) of aquilacrassnin C by involving all shaped breakdown graphs of 1st-generation fragment ions. (c, d) FEER-MS³ of a featured 1st-generation fragment ion at m/z 217 for aquilacrassnin B (c) and aquilacrassnin C (d) by involving all 2nd-generation fragment ions, along with m/z 217 residue and its structure (Right). (C) Molecular defect filtering (MDF) and feature-based molecular networking (FBMN) processing results of SCHs. (a) The two dimensional (2D) scattering plot formed by the integer and decimal values of [M+H]⁺ for SCHs in agarwood. (b) FBMN processing results of SCHs. (c) The overlaid extracted ion current chromatogram of SCHs (∗ indicated the justification with authentic compounds). (D) FCER-MS², FEER-MS³, and the proposed structures of C2–C5. FCER-MS² of m/z 549.25, [M+H]⁺ of C2 (a1), C4 (b1), C5 (c1), and C3 (d1); FEER-MS³ of unit A (m/z 283 for a2, b2, c2, and d2) and unit B (m/z 231 for a3, b3, c3, and d3) for C2, C4, C5, and C3, respectively; and the proposed structures of C2, C4, and C5 (e) and C3 (f). (E) Structures of C2 (aquisinenin L), C3 (aquisinenin M), C15 (4′-demethoxy-3′-dehydroxyl-aquisinenin J), C21 (aquisinenin K), C23 (3′-dehydroxyl-aquisinenin J), C24 (aquisinenin J) and N1 (aquisinenin N). EE: exciting energy; CE: collision energy; t_R: retention time.

Five authentic SCHs were assayed to complete step 1 (Fig. S1). At MS¹ spectrum, SCH typically exhibited [M+H]⁺ exceeding m/z 460 and displayed unique mass defect feature within 0.2–0.5 Da, resulting in a MDF opportunity to efficiently capture SCHs from complicated MS¹ dataset (Fig. S2). H₂O neutral loss, the linkage fission between units A and B, and the dissociation of either unit (Figs. S3 and S4) served as the primary fragmentation routes. Featured fragment ion corresponding to unit B was lower than m/z 300 with a defect > 0.15 Da and the one for unit A was greater than m/z 240 with a defect < 0.12 Da. Their sum was 1.01 or 35.01 Da lower than [M+H]⁺. For FCER-MS², in addition to the previously defined correlations for optimal collision energy (OCE), CE level corresponding to 50% survival yield of precursor ion (CE₅₀) and the maximal relative ion intensity (RII_max) against structural properties [3], the intramolecular hydrogen bonds between hydroxyl groups of two units decreased RII_max of fragment ions belonging to unit B (Fig. S5). When 1st-generation ion at m/z 121.06 showed great RII_max, methoxy moiety occurred at C-ring of unit A (Figs. 1B(a) and 1B(b)). For FEER-MS³, the same substructure existed for identical FEER-MS³ spectra from different SCHs, and moreover, the relationships for optimal exciting energy (OEE), EE level corresponding to 50% survival yield of fragment ion (EE₅₀), and the relative ion intensity at OEE (RII_OEE) against substructure were similar to the correlations between FCER-MS² and structure (Figs. 1B(c) and 1B(d)).

For step 2, after pre-processing with MZmine software, data filtering parameters including mass range as 460–700 Da, mass defect range as 0.2–0.5 Da and minimum intensity threshold as 5000 cps (Fig. 1C(a)) were set to capture SCHs by constructing fit-for-purpose rectangle frame. The top-500 signals were imported into GNPS platform for FBMN processing, leading to a distinct SCHs cluster containing 24 nodes (C1–C24) (Figs. 1C(b) and 1C(c)), embracing one at m/z 535.27, four ones at m/z 549.24, eight ones at m/z 551.26, six ones at m/z 553.28, one at m/z 565.28, two ones at m/z 579.26, one at m/z 583.29, and the rest one at m/z 599.29. Further, another compound (N1) showing similar MS/MS spectra and mass defect features also appeared for m/z 517.26. By matching with authentic compounds, C1, C20, C9, C10, and C19 were assigned as aquilacrassnins B and C, together with aquisinenins D, F, and C, sequentially.

Steps 3 and 4 were operated to interpret the other compounds (Figs. S6–S12). Representatively, C2–C5 shared m/z 549.25 ([M+H]⁺, C₃₂H₃₇O₈) and 1st-generation fragment ions including m/z 531.24, 513.23, 283.10, 255.10, 231.14, 227.11, 192.04, 185.13, 164.05, 143.09, 105.07, and 91.05. Thereof, m/z 283.10 and 231.14 correlated to units A and B, respectively, because their sum was 35.01 Da lower than m/z 549.25. After matching EE₅₀, OEE, and RII_OEE features (Figs. 1D(a1–a3, b1–b3, c1–c3) and Table S1), structural congruence occurred for units A amongst C2, C3, and C5. Although C4 also owned 1st-generation fragment ion at m/z 283.10, different abundance ranking style existed for 2nd-generation fragment ions, i.e. m/z 255 > m/z 227 > m/z 164 > m/z 192 > m/z 91 (Figs. 1D(d1–d3). There were at least two hydroxyl groups linking to A-ring of units A for C2–C5. Regarding m/z 231.14, C2–C5 showed quite similar FEER-MS³, except RII_OEE features of m/z 143 (C2: 60.23%; C4: 55.20%; C5: 53.01%; and C3: 75.73%), attributing to the subtle structural differences in units B. After referring to the literature [5], selina-4,11-diene-12,15-dial as unit B was shared by C2, C4 and C5, and 12,15-dioxo-selina-4,11-dine belonged to C3. C2–C5 produced totally identical FCER-MS², suggesting the same linkage manner between two units. Consequently, C2, C4 and C5 were tentatively deciphered as stereoisomers of structure shown in Fig. 1D(e), and C3 was structurally shown in Fig. 1D(f). Notably, N1 didn't show 1st-generation fragment ion belonging to unit B, and nonetheless, produced fragment ions at m/z 295.10 and 283.10 (Table S2) [5], indicating a rare SCH bearing C–C bond linkage between two units.

After targeted purification and NMR spectroscopic assays (Figs. S13–S90 and Tables S1–S4), C2, C3, C15, C21, C23, C24, and N1 were obtained and confirmatively identified as new SCHs, namely as aquisinenins L and M, 4′-demethoxy-3′-dehydroxyl-aquisinenin J, aquisinenin K, 3′-dehydroxyl-aquisinenin J, and aquisinenins J and N, sequentially (Fig. 1E). As desired, all configured structures agreed well with the structures constructed above (Fig. 1E and Table S1), and exactly, units A and B of aquisinenin N were linked with C–C bond. Their bioactivity features, including the viability enhancement of TCA-damaged GES-1 cells and inhibitory effects against LPS-induced NO production in RAW264.7 macrophages, were further evaluated. Only C2 and N1 significantly protected GES-1 cells against TCA-induced apoptosis (Fig. S91 and Table S5). Except C23, six SCHs exhibited significant anti-inflammatory activities and NO production inhibition rates ranged from 20% to 50%, and notably, the values of C2 and C3 reached 86.35% and 91.17% at 20 μM, respectively. (Fig. S92 and Table S6).

Summarily, to achieve precisely targeted acquisition, an integrated strategy was proposed for capturing and structural annotation of concerned SCHs from agarwood. Twenty-five SCHs were captured by MDF and FBMN. The correlations between chemical structures and MS/MS behaviors were constructed by assaying five authentic SCHs. Thereafter, units A and B of each SCH were identified by matching FEER-MS³ with known fragment ions, corresponding to substructures from authentic SCHs. Seven new SCHs were purified and identified, agreeing well with the proposed structures. Although this approach remains unable to characterize entirely novel substructures that lack reference spectra or literature analogues, the integration of FBMN, MDF, and FEER-MS³ matching constitutes an effective strategy for the exploration and subsequent purification of novel NPs in a precisely LC–MS/MS-guided manner.

CRediT authorship contribution statement

Qian Wang: Writing – original draft, Methodology. Han Li: Writing – original draft, Methodology. Huiting Liu: Validation, Data curation. Yujie Pei: Validation, Data curation. Pengfei Tu: Supervision. Huixia Huo: Writing – review & editing, Funding acquisition. Jun Li: Supervision, Funding acquisition. Yuelin Song: Writing – review & editing, Conceptualization.

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.

Appendix A. Supplementary data

The following is the Supplementary data to this article: