Chlorination Diversifies Cimicifuga racemosa Triterpene Glycosides

Abstract

Extracts from roots and rhizomes of black cohosh (Cimicifuga racemosa) are widely used as dietary supplements to alleviate menopausal symptoms. State-of-the-art QC measures involve phytochemical fingerprinting of the triterpene glycosides for species identification and chemical standardization by HPLC. In the course of developing materials and methods for standardization procedures, the major C. racemosa triterpene glycoside (1) was isolated ans initially thought to be cimicifugoside (2). Detailed HR-LC-MS and 1/2D NMR analysis of 1 and 2 unambiguously revealed that 1 is its chlorine-containing derivative of 2, namely 25-chlorodeoxycimigenol-3-O-β-D-xyloside. Accordingly, HPLC profiles of black cohosh preparations require revision of the assignments of the chlorinated (1) and non-chlorinated (2) pair. Besides explaining the substantial shift in polarity (Δt_R[RP-18] ca 20 min), 25-deoxychlorination opens a new pathway of structural diversification in triterpene glycosides chemistry. As chemical conversion of 2 into 1 could be demonstrated, deoxychlorination may be interpreted as artifact formation. Simultaneously, however, it is a potentially significant pathway for the gastric in vivo conversion (“nature's pro drug”) of the relatively polar triterpene glycosides into significantly less polar chlorinated derivatives with altered pharmacological properties.

Cimicifuga racemosa (L.) Nutt., (synonym Actaea racemosa L.) (Ranunculaceae), commonly referred to as black cohosh, is native to North America and has had a history of use by Native American Indians.¹ The roots/rhizomes have traditionally been used to treat a variety of medical conditions including colds, kidney ailments, malaise, rheumatism and woman conditions such as uterine disorders and menstrual complaints.² From the results of a recent clinical trial,³ a water-ethanol extract was demonstrated to relieve climacteric symptoms of menopause, such as hot flashes. No toxicity was observed even at high doses (127 mg/day, recommended daily dose: 40∼80 mg/day).⁴ Further exemplary cases of recent studies regarding clinical efficacy and safety of black cohosh are documented in the literature.^5-10 While virtually none of the C. racemosa preparations used in any reported clinical study has been chemically and biologically standardized, the dual standardization using both phytochemical and biological parameters has been a core aspect of the recent botanical research in the UIC/NIH Botanical Center.

For many years, investigations of C. racemosa have led to the isolation of phytochemicals that constitute, or are part of, the underlying bioactive principles of the plant.^11-13 In a quantitative context, however, and as typically observed with botanical dietary supplements, the constituents characterized thus far do not account for all of the biological activity observed and relative to the biological endpoints of similar clinical trials. Accordingly, there is continued demand on the identification of new marker compounds for chemical standardization of black cohosh.

In addition, because C. racemosa has been associated with more than 20 common names, this botanical has often been confused with other Cimicifuga species, e.g., the Asian varieties such as Rhizoma Cimicifugae from C. foetida L., or C. dahurica Maxim, which is typically marketed in China, as well as rhizomes and roots of C. simplex Warm., which is typically marketed in Japan. There is no recent literature indicating that Asian Cimicifuga rhizomes (Sheng Ma) have been used to treat women's disorders during menopause, although there have been earlier reports relative to the treatment of uterine disorders and menorrhagia.¹⁴ In addition to a significant number of case reports from Europe, Australia, Canada, and the United States, which often lack verification of botanical identity of C. racemosa, two adverse medical events that recently cited the use of black cohosh as inducing hepatitis or liver damage were reported from both Australia¹⁵ and the United States (black cohosh tablets, unknown brand or dose).¹⁶ Although there is ongoing discussion as to whether or not black cohosh was the common causative agent noted in these two reports, there was no analytical documentation that could unequivocally identify the consumed supplement as black cohosh. HPLC analysis of black cohosh exhibits a distinct fingerprint, characteristic of its chemical constituency with the major constituents being the 9,19-cyclolanostane triterpene glycosides present in all Cimicifuga species. The triterpene glycoside HPLC fingerprint is useful for differentiating Cimicifuga racemosa (black cohosh) from the Asian Cimicifuga species and other related plants possessing the same common name.¹⁷

As representative of one of the plants presently under investigation in the UIC/NIH Botanical Center, the phytochemical study of C. racemosa has previously led to the identification of a number of new triterpene glycosides with the total assignment of 26 triterpene glycoside structures.^18,19 In the process of standardizing the root extracts, using a previously established HPLC method,²⁰ a high abundance peak was observed (R_t = 42 min) (Figure 1), which had not been previously identified. Isolation of the compound underlying this chromatographic peak and preliminary survey spectroscopic analysis (NMR, 1-D ¹H/¹³C) of the isolate suggested that it possessed nearly identical spectroscopic data to that obtained for cimicifugoside, i.e., cimigenol-3-O-β-D-xyloside (2). The chromatographic peak originally assigned to 2, however, was reported to have a considerably longer retention time (R_t = 62 min, S1).²⁰ As a result of this observation, it was clear that an ambiguity existed with respect to the assignment of the respective chromatographic peaks, which prompted a more detailed NMR and LC-MS investigation (see Supplementary Data for the HMBC spectra of 1 [S2] and 2 [S3]). The results revealed that the unassigned HPLC peak (R_t = 42 min) is in fact the known triterpene glycoside 2, and that the later eluting peak (R_t = 62 min), which had originally been assigned structure 2, is in fact the chlorinated triterpene glycoside, chlorodeoxycimigenol-3-O-β-D-xyloside, a new chemical entity possessing structure 1. While 1 and 2 are structurally almost indistinguishable, they have distinctly different HPLC retention times with a ΔR_t of 20 min. Consistent with replacement of the C-25 hydroxyl substituent with a chlorine atom, the ¹H and ¹³C data for the two triterpene glycosides are almost indistinguishable, even for ¹H data recorded at 900 MHz (Supplementary Data S4 and S5). Herein, we wish to describe the details of the isolation and structure elucidation of 1, report on the chemical conversion of triterpenes into chlorinated analogues, and discuss the broader implications of this exchange with regard to chemical diversification and biological impact.

Figure 1.

HPLC profile of a hydroalcoholic C. racemosa extract with ELSD detection using a previously reported method (briefly: RP-18 [YMC ODS-AQ, 5 um, 120 Å, 4.6mm × 250 mm] with 0.05% aqueous. TFA – MeCN gradient).²⁰ Of the several unassigned chromatographic peaks, the arrow indicates a major constituent (“?”) that was originally isolated as being new, but was later identified as the known 2. The latter had previously been assigned to a peak eluting at 62 min (see HPLC of reference standards in supplementary data, S1), which as a result of this study corresponds to the new chlorinated analogue 1.

Results and Discussion

Pure 25-chlorodeoxycimigenol-3-O-β-D-xyloside (1) was obtained from the MeOH extract of roots/rhizomes of C. racemosa as a pale yellow powder after repeated chromatography on silica gel followed by reverse phase silica gel LPLC (RP-18, Lobar®). During positive ion electrospray, 1 exhibited a protonated molecule of m/z 639.3666 and the typical isotope pattern of chlorinated compounds, consistent with the molecular formula of C₃₅H₅₅O₈Cl, (calc. 639.3658; 1.3 ppm error). Compound 1 was clearly recognized as a highly oxygenated 9,19-cycloartane triterpene monoglycoside from its proton NMR data (Table 1). The characteristic high-field methylene signals of 1 appearing at δ 0.297 and 0.536 (each 1H, d, J = 4.2 Hz) were diagnostic for the presence of the cyclopropane ring present in the 9,19-cyclolanostane triterpene ring skeleton. The presence of seven methyl groups were indicated from the ¹H NMR data (pyridine-d₅) appearing at δ_H 0.867 (d, J = 6.3 Hz, Me-21), 1.077 (s), 1.141 (s), 1.189 (s), 1.336 (s), 1.710 (s), 1.720 (s). A single β-anomeric glycoside proton resonance was observed at δ_H 4.889 (d, J = 7.6 Hz) and confirmed the presence of a single carbohydrate moiety. The ¹³C and DEPT-135 NMR data (Table 2) for 1 showed signals that were ascribed to four oxygen-bearing methine carbons appearing at δ_C 89.44 (C-24), 88.28 (C-3), 80.25 (C-15), and 72.48 (C-23), and quaternary carbon resonances appearing at δ_C 113.16 (C-16) and 71.56 (C-25) corresponding to the spiro-ketal carbon and chlorine-bearing carbons, respectively. The spectra for 1 also exhibited signals consistent with the presence of five oxygenated carbon signals associated with the glycoside residue and the ¹³C chemical shifts were found to be identical to those reported for the β-D-xylopyranose moiety [δ_C 107.67 (C-1′), 75.62 (C-2′), 78.68 (C-3′), 71.26 (C-4′), 67.17 (C-5′)] found in 26-deoxyactein and 23-epi-26-deoxyactein; their structures have been previously confirmed by single-crystal X-ray analysis.¹⁸

Table 1.

¹H NMR data of compounds 1 and 2.

Proton	1a	1b	2a	2b
1	1.24 m	1.248 (3.5, 4.3, 13.3)	1.24	1.250 (3.5, 4.3, 13.3)
	1.61 m	1.613 (1.3, 4.0, 12.9, 13.3)	1.62	1.616 (4.0, 12.9, 13.3)
2	1.981 (4.2, 11.8, 12.0, 13.0)	1.975 (4.3, 12.9, 11.7, 12.9)	1.983 (4.0, 12.5, 16.5)	1.977 (4.3, 12.9, 11.7, 12.9)
	2.392 (3.0, 3.5, 4.5, 13.0)	2.382 (3.5, 4.0, 4.5, 12.9)	2.395 (4.0, 7.8, 12.5)	2.385 (3.5, 4.0, 4.5, 12.9)
3	3.536 (4.5, 11.8)	3.530 (4.5, 11.7)	3.539 (4.3, 11.8)	3.530 (4.5, 11.7)
5	1.341 (4.5, 11.2)	1.348 (4.5, 12.4)	1.347 (4.3, 13.2)	1.352 (4.5, 12.4)
6	0.731 (1.7, 12.6, 12.3, 12.3)	0.735 (2.5, 12.4, 12.7, 12.7)	0.729 (2.0, 13.0, 14.2)	0.729 (2.0, 13.0, 14.2)
	1.58 m	1.546 (2.5, 2.5, 4.5, 12.7)	1.50	1.537 m
7	1.17 m,	1.177 (2.5, 12.5, 12.5, 12.7)	1.12	1.179 (2.5, 12.5, 12.5, 12.7)
	2.14 m	2.125 (2.5, 4.5, 4.5, 12.7)	2.09	2.090 overlapped
8	1.681 (4.2, 11.8)	1.698 (4.5, 12.5)	1.687 (4.4, 12.5)	1.707 (4.5, 12.5)
11	1.07 m	1.085 (3.9, 10.7, 13.8)	1.10	1.088 (3.9, 10.7, 13.8)
	2.12 m	2.090 (5.5, 10.7, 15.2)	2.13	2.090 overlapped
12	1.67 m,	1.554 (5.5, 11.0, 13.8)	1.52, 1.70	1.562 m
	1.52 m	1.675 (0.8, 3.9, 11.0, 15.2)		1.675 overlapped
15	4.263 (9.4)	4.257 (9.4)	4.281 (8.9)	4.276 (9.0)
OH-15	4.755 (9.4)	4.703 (9.4)	4.488 (8.9)	4.460 (9.0)
17	1.464 (11.0),	1.466 (11.3)	1.513 (10.8)	1.518 (10.9)
18	1.141 s	1.141 (0.8)	1.161 s	1.163 (0.8)
19	0.297 (4.2)	0.299 (4.2)	0.294 (4.2)	0.299 (4.2)
	0.536 (4.2)	0.536 (1.8, 4.2)	0.538 (4.2)	0.536 (1.8, 4.2)
20	1.65 m	1.642 (6.7, 6.9, 11.1, 11.3)	1.65 m	1.675 overlapped
21	0.867 (6.3)	0.869 (6.7)	0.863 (6.5)	0.866 (6.5)
22	1.011 (1.7, 11.2,13.2)	1.010 (2.0, 11.3, 13.3)	1.039 (1.8, 11.4, 13.2)	1.035 (2.1, 11.9, 13.3)
	2.268 (6.9, 9.5, 13.2)	2.268 (6.9, 9.4, 13.3)	2.283 (6.7, 9.7, 13.2)	2.281 (6.9, 9.4, 13.3)
23	4.699 (0.9, 1.8, 9.5)	4.693 (0.8, 2.0, 9.4)	4.780 (brd 8.5)	4.727 (1.1, 1.7, 9.5)
24	3.885 (0.9)	3.880 (0.9, 0.9)	3.788 brs	4.785 (1.0, 1.3)
26	1.710 s	1.697 s	1.488 s	1.480 s
27	1.720 s	1.714 s	1.509 s	1.504 s
28	1.189 s	1.188 s	1.197 s	1.197 s
29	1.336 s	1.330 s	1.335 s	1.329 s
30	1.077 s	1.073 s	1.077 s	1.074 s
1′	4.889 (7.6)	4.881 (7.5)	4.891 (7.6)	4.881 (7.5)
2′	4.061 (7.6, 8.6)	4.054 (7.5, 8.9)	4.061 (7.6, 8.5)	4.052 (7.5, 8.9)
3′	4.187 (8.6, 8.9)	4.248 (8.7, 8.9)	4.185 (8.5, 8.7)	4.175 (8.7, 8.9)
4′	4.260 (5.0, 8.9, 10)	4.248 (5.2, 8.7, 10.0)	4.260 (5.1, 8.7, 10.2)	4.246 (5.2, 8.7, 10.0)
5′	3.758 (10.0, 11.2, b)	3.751 (10.0, 11.5, b)	3.760 (10.2, 11.3, b)	3.752 (10.0, 11.5, b)
	4.384 (5.0, 11.2, a)	4.375 (5.2, 11.5, a)	4.386 (5.0, 11.3, a)	4.376 (5.2, 11.5, a)

^aRecorded at 360 MHz in pyridine-d₅.

^bRecorded at 900 MHz in pyridine-d₅.

Table 2.

¹³C NMR data^a of compounds 1 and 2, and calculated ¹³C chemical shift differences indicating the substituent chemical shift effect of deoxychlorination

Carbon	1	2	Δδ[1−2]
1	32.46 t	32.44 t	Ø
2	30.02 t	30.20 t	Ø
3	88.28 d	88.58 d	Ø
4	41.39 s	41.38 s	Ø
5	47.58 d	47.61 d	Ø
6	21.08 t	21.07 t	Ø
7	26.33 t	26.37 t	Ø
8	48.67 d	48.65 d	Ø
9	20.00 s	20.03 s	Ø
10	26.66 s	26.67 s	Ø
11	26.44 t	26.47 t	Ø
12	34.00 t	34.08 t	Ø
13	41.72 s	41.86 s	Ø
14	47.16 s	47.30 s	Ø
15	80.25 d	80.24 d	Ø
16	113.16 s	112.00 s	+1.16
17	59.48 d	59.58 d	Ø
18	19.52 q	19.53 q	Ø
19	30.92 t	30.90 t	Ø
20	23.96 d	24.12 d	Ø
21	19.48 q	19.60 q	Ø
22	37.49 t	38.15 t	−0.66
23	72.48 d	71.83 d	+0.65
24	89.44 d	90.20 d	−0.76
25	71.56 s	70.97 s	+0.59
26	30.20 q	27.25 q	+2.95
27	26.36 q	25.41 q	+0.95
28	11.81 q	11.84 q	Ø
29	15.47 q	15.47 q	Ø
30	25.74 q	25.75 q	Ø
1′	107.67 d	107.67 d	Ø
2′	75.62 d	75.62 d	Ø
3′	78.68 d	78.68 d	Ø
4′	71.26 d	71.27 d	Ø
5′	67.17 t	67.18 t	Ø

^aAt 90 MHz in pyridine-d₅.

^b Δδ=δ 1 − δ 2 in ppm; Ø indicates |Δδ|<0.30 ppm.

Detailed analysis of the ¹H, ¹³C, HMQC, HMBC and COSY spectra for 1 established the structural elements for rings A to D. The partial structure, −CHCH(CH₃)CH₂CH- (for H-17, H-20 to H-23), present in both 1 and 2 displayed a clearly observable spin pattern network in the COSY spectrum. The HMBC spectrum, in addition to revealing the long range correlation information associated with rings A to D, showed ³J correlations between the methine carbon signal at δ_C 59.48 (C-17) with the protons of the two methyl signals at δ_H 0.867 (Me-21) and 1.141 (Me-18). The pair of methyl groups at δ_H 1.710 and δ_H 1.720 (Me-26, and Me-27) showed ²J correlations to the quaternary chlorine-bearing carbon C-25 (δ_C 71.56) and a ³J correlation to the C-24 methine carbon (δ_C 89.44), which is consistent with the acyclic partial structure present in both 1 and 2. Finally, both rings E and F were established on the basis of HMBC correlations observed between the quaternary ketal carbon C-16 (δ_C 113.16) and H-23 (δ_H 4.699), as well as with H-24 (δ_H 3.885). This confirmed a prior suspicion that compound 1 must be structurally very similar to the known 2.²¹

The only significant differences that were observed in the proton NMR spectra of 1 and 2 were associated with the chemical shifts of H-24 (δ_H 3.788 in 2), which appeared to be shifted downfield (Δδ = 0.107 ppm) as compared to 1; H-23 (δ_H 4.780 in 2), which appeared shifted upfield (Δδ = 0.081 ppm) in 1 (Figure 2); and the two methyl signals for Me-26 (δ_H 1.509 in 2) and Me-27 (δ_H 1.488 in 2), which appeared shifted downfield to δ_H 1.720 and 1.710 (Δδ = 0.211 and 0.222 ppm) in 1 (Figure 3); respectively. Comparative analysis of the ¹³C NMR spectra of 1 and 2 demonstrated the fact that most of the carbon resonances for rings A – E were also essentially identical. In the ¹³C domain, the only signals exhibiting a significant but still small deviation in chemical shift values were C-16, and C-22 to C-27 (Table 2). The strongest substituent chemical shifts (SCSs) of ¹³C NMR signals were observed for the ketal carbon C-16 [δ_C 112.00 (s)] in 2, which is subject to a slight downfield shift (Δδ = 1.16 ppm) to δ_C 113.16 (s) as compared to 1. At the same time, the signals due to C-26 and C-27 [δ_C 27.25 q and 25.41 q] in 2 were also shifted downfield as compared to 1 (Δδ = 2.95 and 0.95 ppm, respectively). The proton signals for Me-26 and Me-27 were shown to correlate (HMBC) with the carbon signals for C-26 and C-27 in 2. A strong ³J HMBC correlation was observed between H-24 and C-16 in 1, and parallels a similar observation present in the HMBC spectrum of 2. Evidence for homonuclear correlation between H-24 and Me-27 in 1 was observed in the COSY data as well. Proton H-24 in 1 showed spin-spin coupling with only one of the methyls, i.e., Me-27, through four single bonds (W or zigzag effect).²²

Figure 2.

Comparison of the downfield signal pattern (3.45∼4.95 ppm) of the 900 MHz ¹H NMR spectra of the new chlorodeoxycimigenol-3-O-β-D-xyloside (1, above/red), and cimicifugoside (2, below/blue, both recorded in pyridine-d₅). In contrast to the minor chemical shift differences of the spiroketal protons of H-23, H-24, only the exchangeable OH-15 proton showed a prominent change of chemical shift, most likely due to a change in intramolecular hydrogen bonding. Accordingly, extreme care needs to be taken when comparing ¹H NMR spectra of triterpene glycosides, especially when analyzing closely related structures and “isomers”, and when working in protic solvents, because the signals of the significant exchangeable protons are lost under these conditions. The fact that solvent induced shifts and concentration effects may even exceed the ¹H induced substituent chemical shift (SCS) differences observed between chlorodeoxy and hydroxy compounds and/or potential other (stereo)isomers, further complicates the unambiguous structural identification of triterpene glycosides.

Figure 3.

Close similarities are also observed in the upper field signal pattern (0.20∼2.60 ppm) of the 900 MHz ¹H NMR spectra of chlorodeoxycimigenol-3-O-β-D-xyloside (1) (above/red) and cimicifugoside (2) (below/blue; both in pyrindine-d₅). Interestingly, the methylene and methine protons of the rings A, B, C, remain largely unaffected by the alteration of the iso-propyl substituent attached to C-24 of the five-membered ketal (ring F). On the other hand, there are small but defined changes in the chemical shift of the methyl groups (Me-26 and Me-27). This feature exemplifies the importance of standardized NMR conditions, in particular solvent choice, when identifying/elucidating the structure of Cimicifuga triterpene glycosides.

Two-dimensional NOESY experiments, carried out at 500 MHz, provided information regarding the relative stereochemistry of 1 vs 2 (Table 3). In addition to the NOEs expected for and associated with rings A to D of the 9,19-cyclolanostane basic skeleton, NOEs were observed between H-24 and protons H-22α and Me-26 in 1, which is consistent with the assignment of H-24 to an α-orientation. Another key NOE was observed between H-23 and Me-27 in 1, which served to confirm the stereochemistry of H-23 as possessing a β-orientation. In 2, NOEs were observed between H-23 with H-22β and Me-26, and H-24 with H-22α and Me-26, thus confirming the partial structure C-25 to C-27 as not being constrained in a ring, but rather reflecting an acyclic sub-structural element in both 1 and 2. The observed NOEs for the side-chain protons in 1 and 2 are essentially identical. The structural relationship of protons H-23 and H-24 places the dihedral angle at or near 90 degrees and, thus, coupling between these two protons is expected to be zero. No coupling between H-23 and H-24 is observed in either 1 or 2. The “extra” NOE between H-23 and H-22α in 2, but not observed in the ¹H NMR spectrum of 1, may reflect a subtle difference in the averaged side-chain conformational populations of 1 vs 2. All spectroscopic evidence for compound 1 is consistent with the structure of 25-chloro-15β-hydroxyl-16β,23β:16α,25-di-epoxy-24β-hydroxyl-9,19-cyclolanostan-3-O-β-D-xylopyranoside, which represents the first chlorine-containing triterpene glycoside isolated from C racemosa. In analogy to the nomenclature of halogenated sugars, the aglycone of 1 is named 25-chlorodeoxy-cimigenol, assigning 1 to 25-chlorodeoxy-cimigenol-3-O-β-D-xylopyranoside.

Table 3.

A summary of major NOEs ^a observed for 1 and 2.

	NOE effects
Proton	1	2
H-2β	Me-30 (1.077)	Me-30 (1.077)
15	Me-18 (1.141)	H-8 (1.687), Me-18 (1.161)
17	Me-21 (0.867), Me-28 (1.189)	no NOE was observed
18	H-8β (1.861), H-12β (1.52)	H-8β (1.687)
21	H-22α (1.011)	H-20 (1.65)
23	Me-27 (1.720), H-22β (2.268), H-22α (1.011 w)	H-22β (2.268), Me-26 (1.710)
24	H-22α (1.011), Me-26 (1.710)	H-22α (1.011), Me-26 (1.710),
28	H-11α (1.07), H-7α (2.14), H-17α (1.464)	H-11α (1.10)
30	H-19b (0.297), Me-29 (1.336)	H-19b (0.294)

^aObtained from 2D NOESY at 500 MHz in pyridine-d₅, mixing time 250 ms; δ of NOE enhanced signals in ppm.

The ¹H and ¹³C NMR data for 1 and 2 were practically identical at the field strengths that are typically used for routine analysis of natural products (300−500 MHz ¹H, data not shown) and are often employed for preliminary examination of isolates. Only the ¹H NMR data subsequently recorded at 900 MHz NMR allowed for the resolution of important details and led to the unambiguous assignment of all ¹H resonances. The proton NMR data obtained at both 360 and 900 are summarized in Table 1, and a detailed comparison of various regions of the 900 MHz ¹H NMR spectra of 1 and 2 are shown in Figure 2 and Figure 3 to illustrate the subtleties of the differences and clarify the signal assignments. Although much attention must be paid to minute detail, even when performing ¹H NMR analysis at high magnetic field strengths, the distinction between 1 and 2 can in principle be made by means of NMR. However, the contribution of MS in resolving what otherwise represented a structural ambiguity needs to be emphasized, with the recognition of the chlorine isotope pattern being a key aspect. At the same time, it must be recognized that, especially for polyhydroxylated natural products (glycosides), and depending on ionization conditions and other MS parameters, frequently observed dehydrated species ([M* − n·18] ions) may provide a challenge to spectrum interpretation. Considering the 17 amu mass difference between a chlorodeoxy derivative and its parent compound, the recognition of a chlorine substitution can escape routine examination, especially in view of the very close similarity of their NMR spectra.

The occurrence of the chlorinated triterpene 1 not only expands the structural diversity of 9,19-cycloartane-type triterpenes, but also provides valuable information pertaining to the relationship of chromatographic trends and the structure of chemically related triterpene glycosides. During the analysis of triterpene glycosides present in black cohosh by HPLC-ELSD and LC-MS, it was observed that, for a pair of epimers only involving one stereocenter, the difference in the retention times was 7.3 minutes.¹⁸ In contrast, the HPLC profile of the extract of black cohosh shown in Figure 1 displays a considerably larger difference in the retention times between compounds 1 and 2 (ΔR_t ∼ 20 minutes). However, in this instance, the compounds are not isomeric but differ because of the replacement of the 25-OH found in 2 with a chlorine atom in 1. Clearly, introducing significant changes in the sub-structural elements by atom replacement serve to alter the inter- and intramolecular hydrogen bonding networks associated with 1 vs 2 (also refer to discussion of SCSs of exchangeable proton signals in the caption of Figure 2). The chromatographic trends associated with 1 and 2, as well as their congeners, may serve as useful indicators to assist in subsequent structural characterizations.

A question immediately arises as to whether 1 occurs naturally or whether its presence is an artifact of the isolation protocol. During the isolation procedure (see Experimental), the initial methanol extract was partitioned using CHCl₃ and n-BuOH. Chloroform, depending on purity, age of the solvent, and length of time exposed to light, is known to break down to phosgene and HCl. The HCl could react with the hydroxyl function at C-25 (a quaternary carbon center) and replace the 25-OH with 25-Cl. Experiments were performed to answer this question. First LC-MS was run on a crude botanical extract of black cohosh, prior to the CHCl₃/n-BuOH partition, to look for 1. None could be detected. Secondly, a series of experiments was conducted in which a dilute chloroform solution of 2 was allowed to come in contact with HCl vapor for a period of time. The result was that 2 chemically converted into 1, suggesting that the origin of 1 may indeed be, at least in part, an isolation artifact. Since naturally occurring chlorinated compounds derived from plants are rare, in this instance, experimental verification of its possible source was clearly justified. In an effort to simulate the interaction of 2 with chloroform, which is known to contain varying levels of HCl from decomposition, chemical conversion studies were carried out with more concentrated CHCl₃ solutions containing HCl gas. The reactions were monitored by LC-MS. As a result, it was shown that 2 was readily but not quantitatively, converted into 1.

The occurrence of organohalogen natural products in terrestrial plants, marine organisms and animals is typically underrated.^23,24 In addition to organochlorines, organobromines have been detected across all phyla, including plants.^25,26 Specific biosynthetic pathways exist for enzymatic halogenation through haloperoxidases (e.g., haloperoxidases [CPO] and bromoperoxidases [BPO]),^27,28 FADH₂-dependant halogenases, S-adenosylmethionine fluorinating enzymes, and non-heme Fe²⁺, α-ketoglutarate, and O₂ dependent halogenases.^29-32 More recently, biogenetic conversion by diatomaceous microorganisms has been shown to generate halogenated medium chain hydrocarbons from C₂₀ fatty acids.³³ These specific pathways add to the possibly non-specific or spontaneous formation of halohydrins from epoxide precursors. It should be noted that numerous Cimicifuga triterpenes contain the epoxide motif, and that the halide/peroxidase/hydrogen peroxide chemical system is well established in mammals.³⁴

Regardless of whether it represents an acid-catalyzed “artifact” formation or an enzyme-controlled reaction, the demonstration of the likelihood of a straightforward chemical exchange of a tertiary OH for a Cl atom opens a new pathway of structural diversification in triterpene glycoside chemistry. Moreover, deoxychlorination converts a relatively polar triterpene alcohol (2) into a much more lipophilic chlorohydrin derivative (1). It might be reasonable to expect that deoxychlorination is a universal reaction, and can produce new chemical entities that are equally complex assortments as their precursor triterpenes. Given the reactivity of tertiary alcohols, the numerous Cimicifuga triterpenes with an oxirane partial structure, e.g., the whole acetein series, are even more likely to undergo reactions that eventually yield chlorohydrin derivatives. Although different in basic skeleton, the very rare finding of the chlorinated cycloartane-type glycoside 1 shares an interesting relationship with the chlorinate oleanane-type triterpene aglycone known from Mentha villosa.³⁵

As conversion of 2 into 1 was experimentally demonstrated to occur in vitro, deoxychlorination may at first sight be interpreted as artifact formation. At the same time, however, it represents a potentially significant pathway for the gastric (pH 2 HCl) in vivo conversion of the triterpene glycosides to afford pharmacologically enhanced agents. Another plausible spontaneous pathway leading to the formation of organochlorine compounds has been reported for withanolides, in which the ubiquitous NaCl serves as the halogen source.³⁶ Under all of these considerations, when orally administered, the relatively polar glycosides would actually represent “nature's prodrugs” that are bioconverted into significantly more lipophilic chlorinated derivatives. These bioconversion products can be expected to possess altered pharmacodynamic and, almost certainly, improved pharmacokinetic properties, particularly with regard to their otherwise poor oral absorption. Therefore, the question of whether chlorohydrin (ketal derivative) artifact formation in C. racemosa and related botanical preparations is desirable, or undesirable, is only one facet of the interpretation of this finding: The concept of the potential gastric in situ bioconversion (bioactivation?) of a large variety of the Cimicifuga triterpenes may in fact decrease the need for stringent control of the “genuine” triterpene spectrum.

An aspect with potentially much broader impact is the new chemical space opened by deoxychlorination of the structural class of triterpenes and related terpenes. Exploration of this space is in need of further study and a prerequisite for determining the biological impact of deoxychlorination in vitro and in vivo. The fact that (bioconverted) chlorohydrin derivatives can potentially be (part of the) active principle(s), underlines the need for parallel chemical and biological standardization of dietary supplements and evaluation of other natural products, respectively. At the same time, considering this form of bioconversion, it must be borne in mind that correlations between phytochemical profiles and in vitro or clinical biological endpoints might be less straightforward than they appear. While this report establishes basic evidence for this relationship in black cohosh preparations, it is reasonable to predict that similar mechanisms apply to other natural products containing a similar (tri)terpene profile. Finally, it is perceivable that, besides chlorination of hydroxylated triterpenes, highly functionalized and hydroxylated natural products present in almost all dietary supplements have a similar potential for in vivo bioconversion when taken orally.

While it remains to be demonstrated, to what extend chlorodeoxygenation might even be a universal mechanism that equally affects biologically active and (genuinely) inactive natural products, the phenomenon of chlorodeoxygenation deserves added attention from all relevant biological perspectives (in vitro, QSAR, in vivo, clinical). Regarding C. racemosa, we are currently pursuing the possibility of identifying other chlorinated derivatives of the cimigenol and perhaps other (epoxy) series of triterpene glycosides, and of further characterizing the conditions of such bioconversion processes.

Experimental Section

General Experimental Procedures

Melting points were determined on a Fisher-Johns melting point apparatus. Optical rotations were obtained with a Perkin-Elmer 241 polarimeter (Perkin-Elmer, Inc., MA). ¹H NMR spectra were measured at 360, 500, and 900 MHz and ¹³C NMR spectra were measured at 90 MHz on Bruker Avance spectrometers (Bruker, Zurich, Switzerland) in pyridine-d₅ as the solvent. The chemical shifts were referenced to tetramethylsilane (TMS) using the signals from the solvent (δ_H 8.737, δ_C 149.911 relative to TMS) as internal standard. NMR data were processed with NUTS (Pro version for Microsoft Windows, Acorn NMR Co., USA) and MestRe-C (Version 3.6.3 for Microsoft Windows, MestRe-C Co., Spain). ¹H NMR spectra were acquired with 64 K data points and zero-filled to 256 K data points after Fourier transformation. HR-ESI data were recorded on a Micromass (Manchester, UK) Q-TOF-2 system. Infrared spectra were recorded on an ATI Mattson Genesis Series FTIR (ATI Mattson Instruments Inc., WI) by using liquid layer on NaCl. Thin-layer chromatography was performed on precoated TLC plates (250 μm thickness, K6F Si gel 60 and K6F RP-18 Si gel 60, EM Science, Germany) with compounds visualized by spraying the dried plates with 5% H₂SO₄ in EtOH and followed by heating until dryness. Semi-preparative HPLC was carried out on a Waters Delta 600 system with a Waters 996 photodiode array detector, Waters 717 plus autosampler, and Millennium³² Chromatography Manager (Waters Co, MA) on a GROM-Sil 120 ODS-4 HE (Watrex-International, Inc., NY) semi-preparative column (5 μm, 300 × 20 mm) with a flow rate of 6 mL/min. Reversed-phase column chromatography was carried out on Merck Lobar LiChroprep RP-18 columns (EM Science, Germany) with a Fluid pump (FMI, Fluid Metering, Inc., NY) and fraction collector (Spectrum, Spectrum Chromatography, TX). Silica gel (230−400 mesh, Davisil, W.R. Grace & Co., MD) was used for column chromatography.

Plant Material

Cimicifuga racemosa (L.) Nutt. roots/rhizomes were collected in Rockbridge County, Virginia (June 1999), GPS coordinates 37 48.27 N × 79 18.67 W, and identified by Dr. G. Ramsey, Department of Biology, Lynchburg College, Lynchburg, VA. Voucher specimens have been deposited at the Ramsey-Freer Herbarium at Lynchburg College, Lynchburg, VA, and at the Field Museum of Natural History Herbarium, Chicago, IL.

Extraction and Isolation

The dried, milled roots/rhizomes of C. racemosa (6 kg) were extracted with MeOH and fractionated by successive partitions with CHCl₃ and n-BuOH. The CHCl₃-soluble fraction (160 g) was chromatographed on a silica gel (1.8 kg) column and eluted with CHCl₃/MeOH (20/1, 500 mL), CHCl₃/MeOH (10/1, 8 × 500 mL), CHCl₃/MeOH (2/1, 3 × 500 mL), CHCl₃/MeOH (1/1, 6 × 1500 mL), and MeOH (3 × 1500 mL) to give a total of 22 fractions. Fractions 21−22 (34.19 g) were further fractionated by normal-phase silica gel column chromatography (330 g) eluting with a gradient CHCl₃-MeOH (10:1 in increasing polarity to 0:1) solvent system to afford six sub-fractions (SF-I to SF-VI). The fraction SF-I (3.47 g) was subjected to a RP-18 open column eluted with MeOH/H₂O (2/1, 2 × 250 mL), MeOH/H₂O (4/1, 3 ×250 mL) and MeOH to yield the fractions SF-I-A to SF-I-F. After combining fractions SF-II and SF-I-B (2.40 g), the material was further fractionated on a LiChroprep RP-18 column (Lobar®, 310 × 25 mm, E. Merck), and finally purified by semi-preparative HPLC column (gradient, from 50% to 80% MeCN in water in 30 minutes) to yield pure 1 (5.4 mg).

For the isolation and characterization of cimicifugoside (2), see previous report¹⁹ and S5 for ¹H and S3 for HMBC NMR spectra.

25-Chlorodeoxy-cimigenol-3-O-β-D-xylopyranoside (1)

pale yellow powder, mp 241−243°C, [α]²⁵_D +4.8 (c 0.003, CHCl₃); ¹H and ¹³C NMR see Table 1 and Table 2, respectively, as well as S4 for ¹H and S2 for HMBC NMR spectra; IR max (cm⁻¹) 3425, 2935, 2869, 1376, 1233, 1042, 760; HRESIMS m/z 639.3666 (calc 639.3658 for C₃₅H₅₆O₈Cl; [M+H]⁺)

Chlorination of 2

The following experiments were conducted in order to attempt to “simulate” approximate isolation conditions under which the formation of 1 might have occurred. Two 10 μL samples of 2 were removed from a pyridine-d₅ NMR solution (conc. 3.3 mg/mL) and transferred into two 1 mL vials, respectively. After evaporating the pyridine under vacuum, to one vial was added distilled 0.5 mL CHCl₃, and to the other vial was added 0.35 mL CHCl₃, 0.15 mL HPLC grade acetone and 0.1 g Si gel. These two samples were stored on a lab bench at ambient temperature for 3 weeks. The solvent and Si gel were removed and the samples were analyzed by LC-MS. The chlorodeoxy compound 1 could not be detected in either of the two samples.

Two 60 μL samples of 2, obtained from a pyridine-d₅ NMR solution (see above), were placed in two 1 mL vials, respectively. After removing the pyridine under vacuum, to one vial was added distilled 0.5 mL CHCl₃, and to the other vial was added 0.40 mL CHCl₃, 0.10 mL HPLC grade acetone and 0.1 g Si gel. The CHCl₃ was pre-treated with the gaseous vapor from 37% hydrochloric acid by removing 5 × 1 mL HCl gas vapor from the surface of the HCl solution with a glass disposable Pasteur pipette, bubbling the gas into 2 mL freshly distilled CHCl₃. Both vials were kept on a lab bench at ambient temperature for 1 week. After removal of the solvent and the Si gel, the samples were analyzed by LC-MS. The chlorodeoxy compound 1 was detected in both samples. This clearly indicates that small amounts of HCl present in the chloroform solution can lead to 25-deoxychlorination 2 to form 1.

Repetition of the Isolation Procedure

50 g of C. racemosa (L.) Nutt. roots/rhizomes (identical raw material previously used) were grounded and exhaustively extracted with MeOH (4000 mL total). The MeOH extract was divided into two portions, and each evaporated to afford 4.38 g and 4.53 g portions of crude extract. The extract portions were dissolved in 90% MeOH and partitioned with EtOAc (4×100 mL). The EtOAc-soluble fractions yielded 1.48 g and 1.53 g after removal of the EtOAc. The 1.53 g EtOAc portion was re-dissolved in MeOH, mixed with Si gel (100−230 mesh, 2 g), and loaded on a vacuum column of Si gel (46 g). Elution with 600 mL of CHCl₃, CHCl₃-acetone (10%), CHCl₃-acetone (20%), CHCl₃-acetone (30%), CHCl₃-MeOH (10%), CHCl₃-MeOH (50%), and MeOH, provided 9 fractions based on TLC monitoring: F1 (CHCl₃), F2 (CHCl₃ + 10% acetone), F3 (400 mL CHCl₃ + 20% acetone), F4 (200 mL CHCl₃ + 20% acetone and 300 mL CHCl₃ + 30% acetone), F5 (300 mL CHCl₃ + 30% acetone and 100 mL 10% MeOH), F6 (150 mL 10% MeOH), F7 (400 mL 10% MeOH), F8 (600 mL 50% MeOH), and F9 (600 mL MeOH). Fractions F2 to F6, the EtOAc-soluble fraction, and the crude MeOH extract were analyzed by LC-MS. The chlorodeoxy compound 1 could not be detected in the MeOH extract nor in the EtOAc fraction, while a small amount (near the LOD) of 1 was detected in fraction F3 by LC-MS.

Supplementary Material

si20070417_020

S1. HPLC profile of reference compounds from Cimicifuga racemosa with ELSD detection using a previously reported method (briefly: RP-18 [YMC ODS-AQ, 5 um, 120 Å, 4.6mm × 250 mm] with 0.05% aqueous. TFA – MeCN gradient).²⁰ The assignment of peak 18 was revised by elucidating its structure as the new chlorinated triterpene glycoside 1. Correspondingly, peak 17 ismost likely the arabinose epimer of 1, i.e., chlorodeoxycimigenol-3-O-β-D-arabinoside, which had previously been assigned as cimigenol-3-O-α-L-arabinoside (see ²⁰ and S1 for further peak numbers and assignments).

1. Caffeic acid, 2. Ferulic acid, 3. Isoferulic acid, 4. Cimicifugoside H-2, 5. Kaempferol, 6. Cimiracemoside A, 7. Formononetin, 8. Cimicifugoside H-1, 9. (26R)-Actein, 10. 26-Deoxycimicifugoside, 11. (26S)-Actein, 12. 23-epi-26-Deoxyactein, 13. 23-OAc-shengmanol-3-O-β-D-xyloside, 14. 26-Dexoyactein, 15. 25-OAc-cimigenol-3-O-α-L-arabinoside (24S), 16. 25-OAc-cimigenol-3-O-β-D-xyloside (24S), 17. (designated) Chlorodeoxycimigenol-3-O-β-D-arabinoside (previously assigned as cimigenol-3-O-α-L-arabinoside (24S), 18. Chlorodeoxycimigenol-3-O-β-D-xyloside (1, previously identified as cimigenol-3-O-β-D-xyloside, syn. cimicifugoside)

S2. HMBC (360 MHz) of chlorodeoxycimigenol-3-O-β-D-xyloside (1)

S3. HMBC (360 MHz) of cimifugoside (2)

S4. 900 MHz ¹H NMR spectrum of chlorodeoxycimigenol-3-O-β-D-xyloside (1)

S5. 900 MHz ¹H NMR spectrum of cimifugoside (2)