Integrative Approaches for the Identification and Localization of Specialized Metabolites in Tripterygium Roots^,

An analysis of the thunder god vine (genus Tripterygium) offers innovative analytical approaches for the identificationand localization of specialized metabolites.

Abstract

Members of the genus Tripterygium are known to contain an astonishing diversity of specialized metabolites. The lack of authentic standards has been an impediment to the rapid identification of such metabolites in extracts. We employed an approach that involves the searching of multiple, complementary chromatographic and spectroscopic data sets against the Spektraris database to speed up the metabolite identification process. Mass spectrometry-based imaging indicated a differential localization of triterpenoids to the periderm and sesquiterpene alkaloids to the cortex layer of Tripterygium roots. We further provide evidence that triterpenoids are accumulated to high levels in cells that contain suberized cell walls, which might indicate a mechanism for storage. To our knowledge, our data provide first insights into the cell type specificity of metabolite accumulation in Tripterygium and set the stage for furthering our understanding of the biological implications of specialized metabolites in this genus.

Extracts from plants of the genus Tripterygium, most prominently Tripterygium wilfordii (Celastraceae; also known as léi gōng téng in Mandarin, which can be translated to thunder god vine), have a long history in traditional Chinese medicine as a remedy for diverse ailments ranging from sores to fever and inflammation (first being mentioned in herbal compendia of the 15th century; Helmstädter, 2013). Root extracts were later evaluated by allopathic medicine, but several randomized controlled clinical trials (conducted during the 1980s to 2000s) that evaluated the efficacy for treating rheumatoid arthritis reached inconsistent conclusions (Liu et al., 2013). Challenges for standardization remain because a plethora of metabolites have been isolated and structurally characterized from Tripterygium extracts (many of which are bioactive), including sesquiterpenoids, diterpenoids, triterpenoids, and sesquiterpene pyridine alkaloids (Brinker et al., 2007). The most promising clinical candidate is a derivative of the diterpene triepoxide, triptolide, termed F60008. A phase I study with this compound in France was only partially successful due to the toxicity of the semisynthetic derivative (Kitzen et al., 2009). There is still considerable enthusiasm for minnelide (Arora et al., 2015), a prodrug derivative of triptolide, which is currently undergoing phase I clinical trials as a treatment for advanced gastrointestinal tumors in the United States (clinicaltrials.gov identifier NCT01927965). Other metabolites with encouraging preclinical performance are the appetite-reducing triterpenoid quinone celastrol (Liu et al., 2015) and the immunosuppressive sesquiterpene pyridine alkaloid wilfordine (and naturally occurring structural analogs; Duan et al., 1999).

While the contents of bioactive specialized metabolites in commercial Tripterygium root extracts vary significantly between growing areas in China, triptolide is always present in very low concentrations (less than 0.008% of dry biomass), sesquiterpene pyridine alkaloids range between 0.0001% and 0.1% of dry biomass, and celastrol can accumulate up to about 1% of dry biomass (Zeng et al., 2013; Zhuo et al., 2013; Guo et al., 2014). To improve the yields of active principles extractable from Tripterygium roots, an understanding of the factors that control their accumulation is highly desirable. Specialized metabolites often are found only in certain plant lineages and are accumulated in specialized anatomical structures and cell types (Lange, 2015). The spatially restricted occurrence of these metabolites presents challenges for analytical chemistry. When bulk organs or tissues are processed to obtain sufficient material for metabolite quantitation, information about the localization of specialized metabolites is lost.

Newer technologies using mass spectrometry-based imaging (MSI) have been developed to allow the detection and mapping of metabolites at cellular resolution (Bjarnholt et al., 2014). One approach involves laser desorption ionization with or without prior application of a chemical matrix to aid with ionization (commonly used acronyms are MALDI [for matrix-assisted laser-desorption ionization] and LDI, respectively). The sample is introduced into a loading chamber and is rastered with respect to a laser as stationary ionization source, while mass spectra are recorded at each position across the sample. The signal intensity for an ion can then be plotted versus the XY position on the specimen, thereby enabling the generation of an ion map or mass spectral image. While the chemical diversity and the occurrence of several cell types in close proximity have complicated the development of robust LDI- and MALDI-MSI methods with plant sections, exciting recent successes include the imaging of various lipid species in cotton (Gossypium hirsutum) seed embryos (Horn et al., 2012), the localization of defense phenalenone-type phytoalexins in lesions caused by nematode infection of banana (Musa spp.) roots (Hölscher et al., 2014), and the differential accumulation of secoisolariciresinol diglucoside (lignan) and cyanogenic glycosides in flax (Linum usitatissimum) seeds (Dalisay et al., 2015).

The high-confidence identification of metabolites in highly complex plant matrices is still a challenge. While several online mass spectral databases now allow spectral comparisons for thousands of metabolites, including those occurring uniquely in plants, the coverage of open-source search tools for NMR spectra is much narrower (for review, see Lange, 2016). We developed Spektraris as a one-stop source for the integrative analysis of both accurate mass and NMR data (Cuthbertson et al., 2013; Fischedick et al., 2015). As part of this study, we updated the Spektraris database with multiple spectral data types for metabolites occurring in the genus Tripterygium. We then fractionated root extracts, obtained chromatographic and spectroscopic data sets for these fractions, and identified the major constituents by searching against Spektraris, thereby demonstrating the utility of the comprehensive online spectral resource. MALDI-MSI indicated a differential localization of quinone methide triterpenoids to the periderm and of sesquiterpene pyridine alkaloids to the cortex layer of Tripterygium roots, which provides important cellular context for follow-up studies to assess the biological functions of these classes of structurally complex metabolites.

RESULTS

Quantitation of Specialized Metabolites in Tripterygium Root Extracts

Before attempting to identify and assess the localization of specialized metabolites in roots of Tripterygium regelii, a robust method for the quantitation of known active principles had to be validated. Homogenates from root samples were extracted with acetone, and metabolites were separated and detected, based on previously published protocols, using HPLC-quadrupole time of flight (QTOF)-mass spectrometry (MS; Cuthbertson et al., 2013; Liu et al., 2013; Fischedick et al., 2015). Authentic standards of triptolide (Fig. 1A) and celastrol (Fig. 1B) were readily available and therefore served to determine the reproducibility, sensitivity, ion suppression, and extraction efficiency of the protocol. When the electrospray ionization source was operated in positive mode, the highest triptolide peak intensities were obtained for [M+H]⁺, with [M+Na]⁺, and [M+K]⁺ ions occurring at significantly lower abundances (Fig. 1, C and D). In addition to relative retention time and molecular ion (accurate mass time [AMT] tag; Cuthbertson et al., 2013), tandem mass spectrometry (MS/MS) fragmentation patterns were used as a criterion for peak identification (Fig. 1, E and F). The quantitation of triptolide was performed by recording peak intensity in extracted ion chromatograms for [M+H]⁺. Celastrol also was identified based on comparisons of AMT tag and MS/MS data (Fig. 1, G, H, K, and L), which matched the records deposited in the Spektraris and MassBank databases (Horai et al., 2010; Cuthbertson et al., 2013). The diode array detector (DAD) trace at 424 nm was used for the quantitation of celastrol (Fig. 1, I and J). Extraction efficiency (recovery), reproducibility of the extraction, linearity of detection, limits of detection and quantitation, and matrix effects were evaluated and a highly robust method was validated (for details, see “Materials and Methods” and Supplemental Table S1).

Figure 1.

A and B, Identification of triptolide (A) and celastrol (B) in Tripterygium root extracts. C and D, HPLC-QTOF-MS detection (at m/z 361.1648 in positive mode) of a triptolide standard (C) and the corresponding root metabolite (D). E and F, MS/MS fragmentation obtained at 30 eV with a triptolide standard (E) and the corresponding root metabolite (F). G and H, HPLC-QTOF-MS detection (at m/z 451.2844 in positive mode) of a celastrol standard (G) and the corresponding root metabolite (H). I and J, HPLC-DAD detection at 424 nm of a celastrol standard (I) and the corresponding root metabolite (J). K and L, MS/MS fragmentation obtained at 30 eV with a celastrol standard (K) and the corresponding root metabolite (L).

Using this method, the contents of triptolide and celastrol in T. regelii roots were determined as 11.8 μg g⁻¹ and 12.4 mg g⁻¹, respectively (Fig. 2A). These concentrations are in the same range as those reported by others (toward the lower end for triptolide and among the highest for celastrol; Zeng et al., 2013; Zhuo et al., 2013; Guo et al., 2014). Triptolide also was detected in extracts of stems (5.2 μg g⁻¹) and leaves (6.4 μg g⁻¹), whereas only trace amounts (below the limit of quantitation [LOQ]) of celastrol were detectable in these organs (Fig. 2A). Visual inspection of T. regelii roots revealed three main morphological types: white lateral roots (youngest root tissue), orange roots (medium age root tissue), and woody roots (oldest root tissue; Fig. 2B). White, orange, and woody root material was collected separately, and metabolites were extracted and analyzed by HPLC-QTOF-MS. Triptolide was most abundant in woody roots (16.5 μg g⁻¹), with lesser concentrations in orange roots (9.8 μg g⁻¹), and the lowest abundance in lateral roots (0.6 μg g⁻¹; Fig. 2A). Celastrol abundance was highest in orange roots 17.8 mg g⁻¹), with lower quantities in woody roots (8.6 mg g⁻¹) and lateral roots (6.6 mg g⁻¹; Fig. 2A). To our knowledge, this is the first study to report on metabolite concentrations in different root types of Tripterygium plants; therefore, we do not have a reference for comparison.

Figure 2.

A, Quantitation of triptolide (white bars) and celastrol (gray bars) in extracts obtained from different Tripterygium organs. B, Different root types.

Development of Spectral Libraries to Expedite the Identification of Specialized Metabolites in Tripterygium Extracts

Very few standards for specialized metabolites occurring in Tripterygium extracts are commercially available, which is a significant impediment to the annotation of peaks in HPLC-QTOF-MS runs at high confidence. As a first step to remedy this situation we assembled the currently available information on the physicochemical properties of Tripterygium metabolites and integrated these data sets into the Spektraris online database (Cuthbertson et al., 2013; Fischedick et al., 2015). Based on our literature searches (as of June 2016), 415 specialized metabolites had been isolated and characterized from Tripterygium. Of these metabolites, 105 are abietane diterpenes, 22 are kaurane diterpenes, 86 are dehydroagarofuran sesquiterpenes, 64 are sesquiterpene alkaloids, three are spermidine alkaloids, 26 are friedelane triterpenes, 38 are friedooleanane triterpenes, 48 are oleanane triterpenes, 15 are ursane triterpenes, and eight additional metabolites belong to other structural classes (Supplemental Table S2). Spectral data for the most commonly identified metabolites from root extracts (a total of 115) were integrated into Spektraris (Fig. 3). Following the recent integration of selected data sets from the NAPROC-13 database (López-Pérez et al., 2007), the Spektraris database now contains NMR records for over 20,000 plant natural products (http://langelabtools.wsu.edu/nmr).

Figure 3.

Updating the Spektraris resource with NMR spectral records of Tripterygium root metabolites.

Acetone extracts of T. regelii were then analyzed by HPLC-QTOF-MS and spectral data (exact mass and calculated elemental formula) compared with the updated Spektraris database. However, the annotation of peaks was still not satisfactory due to the frequent occurrence of isobars (metabolites with the same monoisotopic mass) in Tripterygium; therefore, we proceeded with additional characterizations. NMR spectroscopy can provide unambiguous data for the identification of isomers, but, for complex structures, highly purified fractions containing larger quantities are generally required (even with modern cryoprobes). We wanted to investigate if the spectral search approach enabled by the Spektraris database would yield peak annotations with only partially purified fractions. Since our greenhouse-grown materials were limited, we obtained T. wilfordii roots from a commercial supplier in China and fractionated ethanolic extracts by silica gel gravity column and HPLC (for details, see Supplemental Methods and Data File S1). HPLC-QTOF-MS, MS/MS, and NMR data were acquired for partially purified metabolites (Supplemental Fig. S1) and compared with the Spektraris database.

The earliest eluting metabolite (13.3 min) was detected with a mass-to-charge ratio (m/z) value of 874.2764, which corresponded to a formula of C₄₁H₄₇NO₂₀ for the [M+H]⁺ ion (Table I). Previous studies that used reverse-phase chromatography with Tripterygium root extracts consistently listed wilfortrine as the earliest eluting abundant peak associated with this formula (Zeng et al., 2013; Cai et al., 2015; Su et al., 2015; Luo et al., 2016). The MS/MS fragmentation patterns we obtained at a collision energy of 30 eV were almost identical to those reported for wilfortrine when recorded at 25 eV (major product ion at m/z 846; Cai et al., 2015). A search with our ¹H-NMR data (isolated metabolite was about 88% pure according to the mass spectrometric estimate) against the Spektraris database returned wilfortrine (peak 1 in Fig. 4A) as the top hit with an identity score of 96% (combined MS and ¹H-NMR data; Table I). However, we did not have sufficient material to acquire ¹³C-NMR spectra; therefore, our peak assignment is of high confidence but not unambiguous.

Table I. Properties of metabolites detected by HPLC-QTOF-MS in ethanolic Tripterygium root extracts.

MS/MS spectra were acquired with a fragmentation voltage of 30 eV. n.a., Not available.

HPLC Peak	Retention Time	Formula (Hill Notation)	[M+H]+		Error	MS/MS Signal Patterns	Annotation	Spektraris Score
			Calculated	Observed				1H-NMR	13C-NMR	MS/1H/13C
	min				Δppm	% relative abundance		% identity
1	12.8	C41H47NO20	874.2764	874.2772	0.82	874 (2.1), 856 (18.6), 846 (100), 828 (18.6), 786 (4.5), 674 (12.5), 194 (13.4), 176 (17.0)	Wilfortrine	93	n.a.	96
2	13.9	C41H47NO20	874.2764	874.2770	0.56	874 (100), 856 (6.1), 846 (46.0), 828 (7.2), 786 (3.6), 674 (12.8), 194 (5.9), 176 (18.5)	Unknown	n.a.	n.a.	n.a.
3	15.3	C41H47NO20	874.2764	874.2773	1.06	874 (100), 856 (2.3), 152 (2.5), 134 (5.8)	Wilfordinine H	65	n.a.	83
							Hypoglaunine B	63	n.a.	81
4	15.8	C38H47NO18	806.2866	806.2882	1.94	806 (79.8), 788 (52.4), 746 (96.1), 704 (33.9), 686 (33.2), 644 (22.2), 259 (17.0), 206 (100)	Peritassine A	n.a.	n.a.	n.a.
							Wilformine	n.a.	n.a.	n.a.
5	16.3	C41H47NO19	858.2815	858.2825	1.17	858 (100), 840 (48.0), 798 (22.3), 780 (11.6), 746 (20.0), 686 (40.4), 206 (72.7), 178 (14.1)	Wilforgine	84	73	86
							Wilfordinine D	75	73	83
							Hypoglaunine D	68	71	80
6	17.7	C41H47NO19	858.2815	858.2828	1.46	858 (73.5), 840 (38.3), 798 (63.8), 738 (26.4), 686 (30.8), 644 (28.9), 206 (100), 178 (29.7)	Hypoglaunine D	64	39	68
							Wilforgine	58	37	65
							Wilfordinine D	57	34	64
7	18.1	C43H49NO18	868.3022	868.3036	1.54	868 (100), 850 (50.0), 808 (22.2), 746 (26.5), 686 (44.6), 206 (60.2), 105 (19.1)	Wilforine	n.a.	n.a.	n.a.
8	19.0	C41H47NO19	858.2815	858.2830	1.73	858 (72.7), 840 (100), 798 (69.3), 756 (19.7), 738 (24.6), 259 (27.1), 213 (30.9), 206 (98.9)	Wilfordinine D	n.a.	n.a.	n.a.
							Hypoglaunine D	n.a.	n.a.	n.a.
							Wilforgine	n.a.	n.a.	n.a.
9	19.7	C29H36O6	481.2585	481.2587	0.48	245 (10.3), 231 (100), 203 (23.4), 109 (11.1)	Demethylzeylasteral	100	100	100
10	22.7	C29H38O4	451.2843	451.2844	0.31	215 (5.0), 201 (100), 163 (3.5), 158 (3.8), 123 (2.5), 109 (3.1)	Celastrol	96	90	95

Figure 4.

Metabolites detected in and purified from ethanolic extracts of Tripterygium roots. A, HPLC-QTOF-MS chromatogram (total ion current) with numbered peaks. B, Structures of identified (or in some cases tentatively identified) metabolites of Tripterygium roots. The structure of celastrol (peak 10) is shown in Figure 1.

A second, less abundant, peak associated with a predicted formula of C₄₁H₄₇NO₂₀ eluted at 13.9 min (peak 2 in Fig. 4A). The MS/MS spectrum of this peak indicated less fragmentation ([M+H]⁺ was the peak with highest intensity) compared with the data obtained with wilfortrine (Table I). We did not acquire NMR data for the metabolite associated with this peak; therefore, a high-confidence annotation was not possible.

A third peak corresponding to a predicted formula of C₄₁H₄₇NO₂₀ was detected at 15.3 min (peak 3 in Fig. 4A). Once again, very little fragmentation was observed at 30 eV (Table I). When the calculated elemental formula and ¹H-NMR data (generated with a fraction of 59% purity according to mass spectrometric estimate) were searched against the Spektraris database, several sesquiterpene alkaloid records were returned with high identity scores (wilfordinine H [83%], hypoglaunine B [81%], and wilfortrine [81%; assigned to peak 1]). Wilfordinine H and hypoglaunine B are structurally complex isomers differing only in the size of the macrolide ring (9′ methyl group with a contracted ring in hypoglaunine B) and the stereochemistry on the chiral carbon (C9′ in wilfordinine H and C8′ in hypoglaunine B; Fig. 4B). Based on the available data, we could not distinguish these two possibilities.

A peak eluting at 15.8 min (peak 4 in Fig. 4A) was predicted, based on accurate mass data (m/z value of 806.2882), to be consistent with a formula of C₃₈H₄₇NO₁₈ (Table I). The fragmentation patterns (abundant pseudomolecular ion and base peaks at m/z 746 and 206) were similar to those reported for two previously characterized metabolites, wilformine and peritassine A. The elution order of Tripterygium root constituents in reverse-phase chromatography was in accordance with the properties reported for peritassine A (Luo et al., 2016; Fig. 4B). However, we did not obtain NMR data for this metabolite; therefore, an unambiguous annotation was not possible.

A formula of C₄₁H₄₇NO₁₉ was predicted for a peak eluting at 16.3 min (peak 5 in Fig. 4A), a second peak at 17.7 min (peak 6 in Fig. 4A), and a third peak at 19 min (peak 8 in Fig. 4A; Table I). The MS/MS fragmentation pattern of peak 5 (pseudomolecular ion as base peak, with an abundant fragment at m/z 206) was closest to those reported for wilforgine (Luo et al., 2016). When the exact mass (857.274 g mol⁻¹) and ¹H-NMR spectra for peaks 5 and 6 were searched against Spektraris (no data acquired for peak 8), three records were returned with almost equally high scores: wilforgine, hypoglaunine D, and wilfordinine D (Table I). Because of the complexity and high similarity of the structures of these metabolites, an unambiguous annotation of peaks 5, 6, and 8 was not possible with the available data.

A peak eluting at 18.1 min (peak 7 in Fig. 4A) was associated with a predicted formula of C₄₃H₄₉NO₁₈ (Table I). The only abundant metabolite commonly reported with this formula is wilforine (Fig. 4B). This annotation also was consistent with the elution order and fragmentation patterns of sesquiterpene alkaloids reported previously for reverse-phase HPLC-MS and MS/MS (molecular ion as base peak, with abundant fragments at m/z 206 and 856; Zeng et al., 2013; Cai et al., 2015; Su et al., 2015; Luo et al., 2016). No NMR data were acquired in this study.

The HPLC-MS properties of a peak eluting at 19.7 min (peak 9 in Fig. 4A; retention time and exact mass [480.251 g mol⁻¹] and MS/MS fragmentation patterns [base peak at m/z 231 and additional lower abundance fragments at m/z 245, 203, and 109]) were consistent with those reported for demethylzeylasteral (Table I; Fig. 4B; Zeng et al., 2013). The same metabolite record was retrieved with a high score in searches of the exact mass (480.251 g mol⁻¹) and ¹H-NMR spectral data against Spektraris. However, the published ¹H-NMR data (Gamlath et al., 1987) were incomplete, and no ¹³C-NMR data had been reported. Therefore, we acquired these spectra and updated the Spektraris database accordingly.

The properties of the peak eluting at 22.7 min (peak 10 in Fig. 4A; absorption maximum at 424 nm; 450.610 g mol⁻¹; predicted formula C₂₉H₃₈O₄; MS/MS with m/z 201 as base peak and smaller fragment peaks at m/z 215, 163, 158, 123, and 109) and ¹H-NMR spectra corresponded to those of an authentic standard of celastrol (Table I; Fig. 1).

Cell Type-Specific Localization of Specialized Metabolites in Tripterygium Roots Using MSI Technology

The cryosectioning of small Tripterygium woody roots (diameter 2 mm or less) presented challenges because the stele is lined by densely packed pericycle and endodermis layers, which are surrounded by large thin-walled cortex cells, and the cortex cells are lined by a compact periderm. Various trials indicated that consistent results could be obtained with a thickness of 30 µm (with partial rupturing around the stele; Fig. 5A). MALDI-MS was performed with authentic standard solutions and highly enriched metabolite fractions spotted onto an imaging target plate, which allowed us to optimize the settings for detection in MALDI-MSI with tissue samples. For all metabolites (triterpenoids and sesquiterpene alkaloids), the pseudomolecular ion ([M+H]⁺) and the K⁺ adduct were detectable at high intensity, while signals for Na⁺ adducts were generally of lower abundance (Table II). Ions also were passed through the drift tube of the mass spectrometer, thereby providing an additional dimension of separation by drift time (Table II). Finally, MS/MS experiments were performed and fragmentation patterns of signals in tissue samples were compared with those obtained with standards (Supplemental Fig. S2). In MALDI-MSI, the K⁺ adduct of celastrol was detected exclusively in the periderm of Tripterygium woody roots (Fig. 5B). The signal for demethylzeylasteral ([M+H]⁺) was considerably lower and the MALDI-MSI localization was not as clear cut as that of celastrol, but we still observed a preferential localization to the periderm (Fig. 5C). In contrast, the K⁺ adduct signals for sesquiterpene pyridine alkaloids, representing the elemental formulas C₃₈H₄₇NO₁₈ (wilforine), C₄₁H₄₇NO₁₉ (wilforgine, hypoglaunine D, and wilfordinine D), and C₄₁H₄₇NO₂₀ (wilfordinine H and hypoglaunine B), had only background signal in the stele and periderm, while intense signals were detected throughout the root cortex (Fig. 5, D–F).

Figure 5.

MALDI-MSI of Tripterygium roots. A, Root cross section prepared for MALDI-MSI. B, [M+K]⁺ ion map for celastrol. C, [M+H]⁺ ion map of demethylzeylasteral. D, [M+K]⁺ ion map of wilforine. E, [M+K]⁺ ion map of C₄₁H₄₇NO₁₉ (wilforgine, hypoglaunine D, or wilfordinine D). F, [M+K]⁺ ion map of C₄₁H₄₇NO₂₀ (wilfortrine, wilfordinine H, or hypoglaunine B).

Table II. MALDI-MSI detection of quinone methide triterpenoids and sesquiterpene pyridine alkaloids in Tripterygium roots.

n.a., Not applicable; n.d. not detectable. n = 8.

Metabolite	[M+H]+	Δppm	Drift Time	[M+Na]+	Δppm	Drift Time	[M+K]+	Δppm	Drift Time
			ms			ms			ms
Celastrol	451.2840 ± 0.0010	1.74 ± 1.34	89.21 ± 0.40	473.2660 ± 0.0011	1.98 ± 1.12	106.85 ± 0.33	489.2403 ± 0.0010	1.73 ± 0.98	106.08 ± 0.38
Demethylzeylastral	481.2582 ± 0.0007	1.26 ± 0.76	94.83 ± 0.20	n.d.	n.a.	n.a.	519.2151 ± 0.0013	1.95 ± 1.92	109.19 ± 0.35
C38H47NO18	806.2862 ± 0.0012	1.31 ± 0.59	141.53 ± 0.50	828.2677 ± 0024	2.56 ± 1.46	147.89 ± 0.38	844.2423 ± 0.0018	1.59 ± 1.37	149.75 ± 0.30
C41H47NO19	858.2819 ± 0.0015	1.26 ± 1.15	154.90 ± 3.26	n.d.	n.a.	n.a.	896.2373 ± 0.0026	2.30 ± 1.41	161.51 ± 1.22
C41H47NO20	874.2759 ± 0.0029	2.45 ± 2.19	159.61 ± 1.04	n.d.	n.a.	n.a.	912.2329 ± 0.0014	1.00 ± 1.33	161.07 ± 0.62

To obtain independent data on metabolite localizations (albeit at lower resolution), the outer layers of Tripterygium woody roots (mostly periderm) were removed surgically, and the outer and inner root layers were extracted separately with ethanol and then subjected to HPLC-QTOF-MS. In accordance with MALDI-MSI data, triterpenoid quinone methides were detected at high concentrations in the outer layer (but only as small peaks in the inner layers), whereas sesquiterpene pyridine alkaloids were present at significantly higher concentrations in the inner layers compared with the outer layers of the roots (Supplemental Fig. S3).

Cross sections of orange-colored roots, obtained using a vibratome, indicated that the intense color (with considerable absorption at 390–430 nm) was restricted to the periderm (cork cells, cork cambium, and possibly endodermis; Fig. 6, A and B). Interestingly, these spectral features were consistent with those reported for celastrol (λ_max at 256 and 424 nm) and demethylzeylasteral (λ_max at 247, 304, and 388 nm; Harada et al., 1962; Tamaki et al., 1996). We did not observe oil bodies or similar subcellular structures for the storage of these triterpenoids (as described previously for diterpenoids in Coleus forskohlii; Pateraki et al., 2014). But how can triterpenoids be accumulated to several percent of the periderm layer biomass?

Figure 6.

Accumulation of triterpenoids in Tripterygium root periderm cells with suberized cell walls. A, Cross section of a Tripterygium woody root with characteristic orange coloration of periderm. B, Higher magnification of the periderm layer. C, Quantitation of membrane sterols in the outer and inner layers of the root. D, Yellow fluorescence (560–600 nm) from Nile Red localizes to the cell walls of periderm cells. E, Bright-field image of the Tripterygium root employed for CLSM in D (control). F, Overlay of images D and E. G, Sudan IV staining of a clarified root cross section. H, A clarified unstained root cross section (control).

One possibility would be an incorporation into membranes by displacing (partially or entirely) phytosterols. The expectation in this case would be comparatively low concentrations of phytosterols in the outer layers of the root. To assess this hypothesis, the outer layers of Tripterygium roots (mostly periderm) were peeled off surgically and transferred to glass tubes. The inner root layers were transferred to separate glass tubes. Following saponification, membrane sterols (primarily β-sitosterol) were quantified by gas chromatography-MS. The outer cell layers contained β-sitosterol at 1 µmol g⁻¹ fresh weight, while the inner layers accumulated β-sitosterol to 0.47 µmol g⁻¹ fresh weight (Fig. 6C), which is evidence against a displacement of membrane phytosterols by triterpenoids.

It also would be plausible that triterpenoids are accumulated outside the periderm cells in suberized cell walls (but not cross-linked with these polymeric materials). Nile Red staining was used to visualize lipophilic regions in root cross sections using confocal laser-scanning microscopy (CLSM). A scan of emission wavelengths revealed strong fluorescence at both the yellow and red wavelength ranges for stained root sections (Supplemental Fig. S4), indicative of the presence of lipidic materials in the semipolar to neutral range (Fowler and Greenspan, 1985). The maximum fluorescence intensity was emitted from peridermal cell layers in the yellow wavelength range (Supplemental Fig. S4). Red fluorescence (630–670 nm) was emitted by cells of the stele and periderm, while yellow fluorescence (560–600 nm) was localized exclusively to the periderm (Fig. 6, D–F; Supplemental Fig. S5). Sudan IV treatment resulted in staining of the same cell layers (Fig. 6, G and H), which provided independent evidence that the periderm cells of Tripterygium roots contain highly suberized walls. The orange color in bright-field microscopy (due to celastrol and demethylzeylasteral accumulation) and yellow fluorescence in CLSM (due to suberin deposition) affected the same periderm cells (Fig. 6, A, D, and G).

DISCUSSION

Development of Combined HPLC-MS and NMR Spectral Libraries Facilitates the Identification of Metabolites in Complex Plant Matrices

Specialized metabolites purified and characterized from members of the genus Tripterygium (family Celastraceae) represent a remarkable structural diversity (Supplemental Table S2). For example, according to our literature searches, 105 metabolites have been classified as abietane-type and an additional 22 as kaurane-type diterpenoids, the vast majority of which were isolated from roots. Dehydro-β-agarofuran sesquiterpenoids are widely distributed in the Celastraceae (more than 500 structures) and very common in Tripterygium (our searches returned 150 unique structures). While sesquiterpene polyesters have been isolated primarily from fruit, sesquiterpene pyridine alkaloid macrolides are present in several organs, and their concentrations are often fairly high in roots (Gao et al., 2007). Friedelane triterpenoids are a third prominent class of specialized metabolites in the Celastraceae (more than 50 structures; Shan et al., 2013), and our searches indicated that 26 different friedelanes have been described in Tripterygium.

Only a few studies have attempted to quantify specialized metabolites in Tripterygium, and the concentrations reported therein vary widely between cultivars and growing areas (Zeng et al., 2013; Zhuo et al., 2013; Guo et al., 2014; note that reports about metabolite contents in Tripterygium tablets were not considered, because these are nonstandardized extracts of poorly characterized or even mixed provenance). There is, however, agreement that diterpenoids occur ubiquitously at low levels, with the highest concentration in roots (generally below 20 µg g⁻¹), which is consistent with the data presented here for the diterpene epoxide triptolide. Triterpenoids tend to be much more abundant, and the concentration of celastrol, the signature metabolite of this class, can exceed 1 mg g⁻¹ in roots of mature Tripterygium plants (Zhuo et al., 2013). We detected similar levels in commercial root samples but even higher concentrations (up to 17 mg g⁻¹) in roots of young plants (less than 4 years since germination). Sesquiterpene alkaloids also are prominent specialized metabolites that can reach concentrations in the milligram per gram of root material range (Zeng et al., 2013; Guo et al., 2014).

Although triterpenoids and sesquiterpene alkaloids are quite abundant in Tripterygium, the structural diversity and lack of access to authentic standards have been impediments for more research with this pharmaceutically important genus. Spectral databases can be invaluable tools to aid in metabolite identification, but their coverage of plant chemical diversity is currently very limited (Johnson and Lange, 2015; Lange, 2016). We have been working toward developing database resources that allow orthogonal searches with HPLC-MS, MS/MS, and NMR data sets. For example, our Spektraris database enabled the rapid identification of the major taxanes (diverse class of diterpenoids) in Taxus × media var Hicksii cell cultures (Cuthbertson et al., 2013; Fischedick et al., 2015). We have now expanded these efforts to include other highly diverse classes of specialized metabolites (those occurring in Tripterygium spp.) that had not been incorporated into any existing spectral databases. The advantage of our approach is that the discovery process is significantly accelerated. AMT tag information and MS/MS spectra can be acquired with crude preparations and, for this project, provided tentative identifications of several metabolite peaks obtained with Tripterygium root extracts. Our database records now also include information about the occurrence of metabolites across species, which can be used to further solidify tentative structural assignments.

However, unraveling the identities of isobars (metabolites with the same exact mass), which are common in Tripterygium, requires additional spectral information that could only be provided by NMR. One would generally attempt to further process extracts to fairly high purity to allow unambiguous interpretations of NMR spectra. We demonstrate that the availability of an NMR spectral database for all major metabolite classes of Tripterygium allows high-confidence structural assignments even when only partially purified fractions are obtained in a simple work flow (silica gel gravity column chromatography followed by one-step HPLC fractionation; Supplemental Methods and Data File S1). While this approach does not always allow an unambiguous identification of a metabolite (due to very small structural variations among isobars that would be differentiable only with extensive NMR investigations outside the scope of this study), it is very powerful in supporting dereplication efforts (identifying known metabolites and excluding these from further consideration in phytochemical screening efforts). Adding to the utility of Spektraris, our search algorithm does not require in-depth expertise in NMR spectrum interpretation. Users can upload their own spectral records through a simple Web interface (http://langelabtools.wsu.edu/nmr/submit), thereby contributing to our efforts to continuously broaden the scope of Spektraris. Finally, in addition to providing access to our spectral data via our Spektraris online resource, we share our records with both MassBank (Horai et al., 2010) and NMRShiftDB (Steinbeck and Kuhn, 2004) and, as a result, contribute to the expansion of the phytochemical repertoire of complementary open-source databases.

MS Imaging Provides Evidence for Cell Type-Specific Differences in the Localization of Specialized Metabolites in Tripterygium Roots

Many of the metabolites that have been isolated from Tripterygium, in particular those accumulating in roots, have been demonstrated to exert considerable toxicity in cell-based assays and animal models (Brinker et al., 2007). Therefore, one would expect that a spatial sequestration is required to prevent similar effects in Tripterygium cells. MSI-based approaches allow the evaluation of metabolite localization, but plant tissues can present notable experimental challenges because of their anatomical and chemical complexity. The finding that, based on MSI data, celastrol and demethylzeylasteral (the major quinone methide triterpenoids) accumulate primarily in the periderm was not a surprise, as these cell layers have a bright orange color in young Tripterygium roots, which is consistent with the absorption characteristics of the above-mentioned triterpenoids. The high specificity of this localization, however, with only background noise being detected for the signature ions in other cell types, was not necessarily expected. Given their location in the outermost cell layers of the root, one may speculate that triterpenoids play important roles in the first line of defense against bacterial and/or fungal soil pathogens attempting to penetrate roots. Thus far, celastrol has been tested, and found highly active, only against gram-positive bacteria with clinical pertinence (Moujir et al., 1990), but the relevance for plant-bacteria interactions has remained unexplored. Fairly broad inhibitory effects against several phytopathogenic fungi were reported for both celastrol and its naturally occurring methyl ester (Luo et al., 2016). However, these are preliminary studies, and investigations assessing the ecological significance of Tripterygium triterpenoids, by taking advantage of the natural variability in their accumulation levels in comparative ecological studies, would certainly be desirable.

The concentration of triptolide was too low for consistent detection above background by MSI. However, triptophenolide, a diterpenoid that accumulates to significantly higher levels in Tripterygium woody roots (150 µg g⁻¹), was detectable by MSI. There was a slight enrichment of ions related to this metabolite in the periderm; however, the background was still fairly high, so this localization needs to be regarded as tentative (Supplemental Fig. S6). The measured concentration of triptophenolide (0.015% of dry biomass averaged for the entire root) can thus be regarded as an estimate of the limit of detection by MSI. Compared with triterpenoids, sesquiterpene alkaloids appeared to be more evenly distributed; therefore, the local concentrations were lower (but ion intensities were significantly higher than those of diterpenoids). The high-confidence localization of sesquiterpene alkaloids by MSI, which required a relatively large number of replicate experiments (n = 8), indicated a primary occurrence in the root cortex (with only very low levels in the stele) in this study. MSI was recently employed to unravel the complex localization of intermediates and end products of terpenoid indole alkaloid biosynthesis in several cell types (phloem-associated parenchyma, epidermis, idioblast, and laticifers) of Catharanthus roseus stems (Yamamoto et al., 2016). Magnoflorine, an aporphine alkaloid occurring in Podophyllum spp., was shown, using MSI, to accumulate mainly in the pith and epidermis of the rhizome and throughout emerging roots (Marques et al., 2014). These reports are in general agreement with our finding that alkaloids occur in multiple cell types in roots. However, MSI studies are just beginning to shed light on alkaloid localization, and it is too early to generalize such conclusions. It would now be instructive to evaluate the localization of transcripts related to sesquiterpene alkaloid biosynthesis in Tripterygium, but the genes involved in this process have not been identified yet. Furthermore, it would also be enlightening to determine the ecological roles of sesquiterpene alkaloids, which have intriguing macrolide structures. The data presented here provide the metabolite localization context for these follow-up investigations.

How Are Triterpenoids Sequestered within Periderm Cells?

The high concentration of celastrol in Tripterygium roots begs the question of how this highly bioactive metabolite (Salminen et al., 2010) might be segregated from cellular metabolism to avoid cytotoxic effects. The localization of celastrol and demethylzeylasteral to the periderm of Tripterygium roots is reminiscent of the occurrence of the diterpenoid forskolin in organellar oil bodies of C. forskohlii (Lamiaceae; Pateraki et al., 2014), which are visible in cross sections by light microscopy and appear to occur only in cork cells. We did not find microscopic evidence for the existence of oil bodies in cork cells of Tripterygium roots.

A second plausible hypothesis is that triterpenoids in Tripterygium roots are incorporated into membranes. Assuming that lipid bilayers constitute about 10% of the cell mass (Alberts, 2015), the measured concentration of 18 mg g⁻¹ celastrol in orange Tripterygium roots would translate into a concentration of 180 mg g⁻¹ in membranes. Considering the fact that triterpenoids accumulate only in the periderm (which constitutes about 20% of the volume of orange roots), the concentration in this layer would be 900 mg g⁻¹ (corresponding to 90 mass %), which is considerably more than the highest concentrations estimated for the plant plasma membrane (50 mass %; Dufourc, 2008). Therefore, one would have to postulate that celastrol replaces other sterols in the plasma membrane of Tripterygium periderm cells. However, our measurements indicated that the concentration of β-sitosterol, the signature sterol of Tripterygium roots, was in fact higher in the outer layers compared with the inner layers, which we interpret as evidence against the membrane incorporation hypothesis. An explanation for higher rather than lower concentrations of β-sitosterol in the outer root layers might be that the inner layers of the root contain many metabolically inactive cell types.

Where else could triterpenoids be stored? Both celastrol and demethylzeylastral are readily extracted in lipophilic organic solvents, indicating that they are not glycosylated or in some other way modified (such as triterpenoid saponins) to facilitate storage in the vacuole (Thimmappa et al., 2014). Triterpenoids in Tripterygium roots are soluble, as indicated by the rapid loss of orange color when roots are immersed in organic solvents. An extracellular storage also is conceivable, but cellulose (polysaccharide polymer) and lignin (phenolic polymer) might be too polar to serve as the matrix for triterpenoid deposition. Suberin is a more lipophilic macromolecule of secondary cell walls; therefore, we assessed its distribution in Tripterygium roots. Based on microscopic investigations using both histochemical stains and CLSM, suberized cell walls were found primarily in the periderm, which also was the location of triterpenoid accumulation. Providing direct evidence for a colocalization of soluble lipophilic metabolites and a lipophilic cell wall polymer is exceedingly difficult, as protocols for visualizing suberin result in the removal of soluble constituents. It would take significant efforts beyond the scope of this study to develop protocols aimed at testing the potential absorption of triterpenoids in a suberized cell wall matrix, but it remains the most plausible hypothesis consistent with our experimental data.

MATERIALS AND METHODS

Chemicals and Solvents

Triptolide and celastrol were purchased from Cayman Chemical, and stock solutions were prepared in ethanol (stored at −20°C until use). Acetone was of HPLC grade (Fisher Scientific), ethyl acetate and ethanol were obtained as OmniSolv high-purity reagents (EMD Chemicals), and acetonitrile, methanol and, water were liquid chromatography-MS grade (Sigma-Aldrich). CDCl₃ was obtained from Cambridge Isotope Laboratories. Red phosphorus, α-cyano-4-hydroxycinnamic acid, and 9-anthracenecarboxylic acid were purchased from Sigma-Aldrich. 2,5-Dihydroxybenzoic acid was sourced from TCI America, and Leu encephalin was from Waters.

Plant Materials

Tripterygium regelii plants were obtained from Woodlanders and maintained under greenhouse conditions (illumination, 16-h day, 8-h night (250–500 µE); temperature, 24°C day, 20°C night; relative humidity, 45%–55%). A voucher specimen was deposited, after careful morphological and phytochemical evaluation, with the herbarium of the Field Museum of Natural History in Chicago (collector, D.D. Soejarto). Whole roots of Tripterygium wilfordii were purchased online from Chemleader Biomedical. The identity of Tripterygium metabolites was established by isolation and purification from whole roots, with subsequent high-resolution MS and NMR analyses (Supplemental Fig. S1).

Tissue Extraction

Root, leaf, or stem material (50 ± 0.5 mg) was harvested into liquid nitrogen. Frozen tissue samples were homogenized under liquid nitrogen in a Mixer Mill (MM01; Retsch) using a stainless steel ball in a Teflon receptacle with a shake rate of 20 s⁻¹ for 45 s. Finely powdered root homogenate was transferred to 10-mL glass tubes, and metabolites were extracted with 5 mL of acetone for 30 min in an ultrasound bath (FS30H; Fisher Scientific). Tubes were then centrifuged at 3,000g for 5 min, and the supernatant was transferred to another 10-mL glass tube. The solvent was removed in vacuo, and the remainder was dissolved in 5 mL of acetone for a second extraction following the same protocol as listed above. Tissue was extracted a total of four times, and the finally obtained residue (after solvent removal) was dissolved in 1 mL of acetonitrile:water (80:20, v/v) containing 10 μg mL⁻¹ internal standard (anthracene 9-carboxylic acid). Extracts were passed through a 20-μm filter (polytetrafluoroethylene) and stored at −20°C until further analysis. All samples were analyzed within 5 d of preparation.

HPLC-QTOF-MS, MS/MS, and NMR Data Acquisition

The HPLC system (Agilent Technologies) consisted of a G1379B degasser, G1312B binary pump, G1330B thermostat, G1367C autosampler, G1316B column heater, G1315C DAD, and a G1310A isocratic pump (for infusion of mass references). High-resolution mass spectrometric analyses were performed on a 6520 QTOF device (Agilent Technologies) with an electrospray ionization source. Separation of Tripterygium metabolites was achieved on a Zorbax Eclipse XDB-C18 Rapid Resolution HT 4.6-mm × 50-mm × 1.8-μm column connected to a Zorbax SB-C8 Rapid Resolution guard cartridge (2.1 mm × 30 mm × 3.5 μm; Agilent Technologies). The initial conditions were 70% solvent A (water with 0.1% [v/v] formic acid) and 30% solvent B (acetonitrile with 0.1% [v/v] formic acid). A linear gradient (flow rate of 0.6 mL min⁻¹) was used to increase solvent B to 80% over 35 min, followed by a more rapid gradient to 95% solvent B at 40 min. The DAD was set to record at 219, 254, and 424 nm, with spectra being recorded from 200 to 500 nm. To enable a continuous mass correction, a reference mass solution containing 300 nm purine (exact mass, 120.043596 g mol⁻¹) and 250 nm hexakis-(¹H,¹H,³H-tetrafluoropropoxy)-phosphazine (exact mass, 921.002522 g mol⁻¹) in acetonitrile:water (95:5, v/v) was infused into the ion source at a flow rate of 0.2 mL min⁻¹. The electrospray source was operated in positive polarity mode at a gas temperature of 325°C, a gas flow of 10 L min⁻¹, and a nebulizer pressure 2.41 bar. The QTOF m/z scan range was set to 50 to 1,200 amu, the capillary entrance was maintained at 3,500 V, and the MS/MS fragmentor was set to 175 V. MS/MS data were acquired by employing a collision energy of 30 eV, and electronic files were submitted to the MassBank database (Horai et al., 2010). Data analysis was performed using the MassHunter Workstation software version B.03.01 (Agilent Technologies). A second HPLC-QTOF-MS method was used to acquire data to be included in the Spektraris-AMT database as described previously (Cuthbertson et al., 2013). NMR spectra were acquired in CDCl₃. A detailed listing of acquisition parameters is given in Supplemental Table S3. Images of spectra are provided in Supplemental Figure S1.

Analyte Extraction and HPLC-QTOF-MS Method Validation

The recovery (R) of target analytes from Tripterygium roots (50 ± 0.5 mg homogenate) was determined by (set 1) spiking with 1 µg of triptolide and 50 µg of celastrol before extraction, (set 2) spiking with 1 µg of triptolide and 50 µg of celastrol after extraction, and (set 3) processing the tissue without spiking, which was then taken into account to determine absolute quantities of analytes using the following equation:

The recovery of triptolide and celastrol from T. regelii root extracts was almost complete (Supplemental Table S1). The same analyses were used to determine the reproducibility of the extraction protocol (by comparing replicate data sets) and evaluate matrix effects (by directly comparing detector responses obtained with extracts from sets 2 and 3). The reproducibility of the extraction, determined as relative sd (RSD) of three replicate experiments, was 3.7% and 4.7% for triptolide and celastrol, respectively (Supplemental Table S1). The linear range of detection was determined for triptolide and celastrol. The linearity of detection was tested for triptolide by injecting 0.01, 0.05, 0.1, 0.5, 1, 5, 10, and 50 ng on column (r² = 0.99, RSD ≤ 10%) and for the more abundant celastrol by injecting 1, 50, 100, 500, and 2,500 ng on column (r² = 0.99, RSD ≤ 6%). The limit of detection (signal:noise ratio of 3:1) was determined as 0.01 and 1 ng on column for triptolide and celastrol, respectively (Supplemental Table S1). The LOQ (signal:noise ratio of 10:1) was recorded as 0.1 and 10 ng on column for triptolide and celastrol, respectively (Supplemental Table S1). The LOQ for celastrol determined with DAD was significantly higher than that for MS (0.2 ng), but, because of the high abundance of the metabolite in T. regelii root extracts, sensitivity was not a concern. The intraday variation of the quantitation was 0.9% and 0.6% for triptolide and celastrol, respectively (Supplemental Table S1). The RSD for interday precision, determined by analyte quantitations on three different days, was 6.5% and 0.5% for triptolide and celastrol, respectively (Supplemental Table S1). When a separation using HPLC is combined with MS detection, matrix effects can lead to ion suppression or enhancement, thus leading to an incorrect quantitation of target analytes. To evaluate the relevance of such matrix effects for the detection of triptolide (celastrol was quantified by DAD, where matrix effects are insignificant), peak areas were compared for unspiked and spiked extracts from T. regelii roots. Triptolide was present at 0.8 µg per 50 mg of root material, while a root extract spiked with 1 µg of triptolide contained 1.6 µg per 50 mg, indicating that matrix effects were negligible.

Sample Preparation for MSI

Stock solutions of analytes (0.5 mg mL⁻¹ in ethanol) were mixed 50:50 (v/v) with matrix 1 solution (1 mg mL⁻¹ 2,5-dihydroxybenzoic acid in methanol:water [50:50, v/v]) and spotted into standard wells of an imaging target plate (55 × 41 mm; Waters). A stock solution of Leu enkephalin (1.6 mg mL⁻¹ in water) was mixed 50:50 (v/v) with matrix 2 solution (1 mg mL⁻¹ α-cyano-4-hydroxycinnamic acid in acetonitrile:water [70:30, v/v]) and also deposited on the imaging target plate to be used for lock mass correction (for details, see below).

T. regelii plants were removed from pots, and roots were washed gently with tap water to remove soil. Roots were cut with scissors, placed in 15-mL screw-cap plastic tubes, immediately submerged in liquid nitrogen, and subsequently embedded in 3% agarose. These samples were stored at −80°C for up to 3 weeks. Directly before analysis, the chamber of a CM 1950 cryostat (Leica Biosystems) was set to −20°C, and embedded root samples were sectioned to 30 µm thickness. Sections were transferred to an imaging target plate, and matrix 3 solution (40 mg mL⁻¹ 2,5-dihydroxybenzoic acid in methanol:water [50:50, v/v]) was applied at a flow rate of 0.05 mL min⁻¹ with an HTX Imaging Sprayer (HTX Technologies) connected to an 1100 Series HPLC Binary Pump (Agilent Technologies). The settings for the sprayer were as follows: nozzle temperature at 80°C; spraying velocity at 1,250 mm min⁻¹; 20 passes; drying time between passes, 0.3 min; and track spacing of 1 mm.

MSI

A MALDI Synapt G2 mass spectrometer (Waters) was operated in resolution mode with positive polarity and with a mass range setting of m/z 100 to 1,000 amu. Data acquisition was performed with MassLynx software version 4.1 (Waters). A saturated suspension of red phosphorus in acetone was applied to the standard wells of an imaging target plate and used for the mass calibration according to the manufacturer’s instructions. The imaging target plate was introduced into the MALDI sample chamber, and the laser was operated with the following settings: 1,000-Hz firing rate; laser energy of 350 (arbitrary units); and a step size of 50 µm. Lock mass correction with Leu enkephalin was repeated every 400 s for 5 s. Ion mobility separations were performed with the following settings: helium gas flow of 90 mL min⁻¹; trap wave velocity of 311 m s⁻¹; trap wave height of 4 V; ion mobility wave velocity of 650 m s⁻¹; ion mobility wave height of 40 V; transfer wave velocity of 191 m s⁻¹; transfer wave height of 0.1 V; and ion mobility wave delay of 450 µs. MS/MS experiments were performed by selection of a precursor mass and a collision energy of 30 eV in the transfer cell. MSI data were processed using the High Definition Imaging software version 1.2 (Waters) with lock mass correction.

Microscopy

Roots (approximately 1 mm in diameter) were harvested from greenhouse-grown plants and embedded in 7% (w/v) low-melting-point agarose. Sectioning (30 µm thickness) was performed using a VT 1000 S vibratome (Leica Biosystems). Fresh sections were immediately visualized using a BH2 microscope (Olympus). Sections for staining were transferred from water to 100% ethanol using a gradient and incubated for 1 h to clarify the tissue.

Nile Red Staining

Sections were stained in 70% (v/v) ethanol containing 0.01% (w/v) Nile Red (9-diethylamino-5H-benzo[α]phenoxazine-5-one; Sigma-Aldrich) for 1 h, washed with 50% (v/v) ethanol, and transferred to water for CLSM.

Sudan IV Staining

Sections were stained in 70% (v/v) ethanol containing 0.07% (w/v) Sudan IV [1-(2-methyl-4-(2-methylphenyldiazenyl) phenyl)azonapthalen-2-ol; Sigma-Aldrich] for 3 h, washed with 50% (v/v) ethanol, and transferred to water for bright-field microscopic observation.

CLSM

Imaging was performed using an SP5 confocal laser-scanning microscope (Leica Biosystems). Tissue sections were illuminated with an argon laser (488 nm, laser intensity setting of 15%). The XYλ scan function was employed to determine emission signatures and maxima across the 510- to 775-nm waveband in 10-nm steps. Yellow and red fluorescence emission from Nile Red-stained tissue was detected at 560 to 600 nm and 630 to 670 nm, respectively. Clarified root sections without staining were used as negative controls.

Supplemental Data

The following supplemental materials are available.

• Supplemental Figure S1. Spectral data for metabolites isolated from Tripterygium roots.

• Supplemental Figure S2. MALDI-MS/MS spectra for metabolites in Tripterygium roots.

• Supplemental Figure S3. HPLC-QTOF-MS chromatograms of extracts of the outer layers and inner layers of Tripterygium roots.

• Supplemental Figure S4. Fluorescence emission from a Tripterygium root cross section scanned across the wave band 510 to 775 nm.

• Supplemental Figure S5. Emission of yellow fluorescence and red fluorescence from Tripterygium roots, with bright-field image and overlay of bright-field and fluorescence images for comparison.

• Supplemental Figure S6. MS-based imaging of triptophenolide in Tripterygium roots.

• Supplemental Table S1. Validation of HPLC-based methods for the quantitation of triptolide and celastrol.

• Supplemental Table S2. Metabolites isolated and characterized from members of the genus Tripterygium (literature search).

• Supplemental Table S3. NMR acquisition parameters.

• Supplemental Methods and Data File S1. Isolation and characterization of metabolites from Tripterygium roots.

Glossary

MSI

mass spectrometry-based imaging

MALDI

matrix-assisted laser-desorption ionization

LDI

laser-desorption ionization

QTOF

quadrupole time of flight

AMT

accurate mass time

MS/MS

tandem mass spectrometry

DAD

diode array detector

LOQ

limit of quantitation

MS

mass spectrometry

CLSM

confocal laser-scanning microscopy

RSD

relative sd