Structural Elucidation of an Atropisomeric Entcassiflavan-(4β→8)-Epicatechin Isolated from Dalbergia monetaria L.f. Based on NMR and ECD Calculations in Comparison to Experimental Data.

A rare dihydoxyflavan-epicatechin proanthocyanidin, entcassiflavan-(4β→8)-epicatechin, was isolated from Dalbergia monetaria, a plant widely used by traditional people from the Amazon to treat urinary tract infections. The constitution and relative configuration of the compound were elucidated by HR-MS and detailed 1D- and 2D-NMR measurements. By comparing the experimental electronic circular dichroism (ECD) spectrum with the calculated ECD spectra of all 16 possible isomers, the absolute configuration, the interflavan linkage, and the atropisomers could be determined.

1. Introduction

Dalbergia monetaria L.f., popularly known as “veronica”, is a medicinal plant widely used in the Amazon region by traditional people to treat some infectious diseases such as urinary tract infections (UTIs) [1]. Regarding the chemical composition of D. monetaria, phenolic compounds, such as isoflavonoids and proanthocyanidins (PAs), were isolated from the bark [2,3]. PAs possess a range of biological effects; among them, antibacterial activity is widely reported, especially against Gram-negative and -positive strains [4,5]. However, condensed tannins have not been widely reported from the Dalbergia genus so far. Proanthocyanidin A-2 was identified in the leaves and heartwood crude extracts of D. boehmii [6]. Additionally, (2R,3R,4R)-3,3′,4′,7-tetrahydroxyflavan-(4β→8)-epicatechin and (2R,3R,4R)-3,4′,7-trihydroxyflavan-(4β→8)-epicatechin, built from flavan-3-ol units, were isolated from the stembark of D. monetaria [3]. Flavan-3-ols are the best-known building blocks of proanthocyanidins, mainly catechin and epicatechin. In contrast to flavanols, flavans are very rare as monomeric units of PAs [7]. Thus, dimeric dihydoxyflavan-(epi)catechin proanthocyanidins have only rarely been described, mainly in plants of the genus Senna [7,8,9,10,11]. Recently, our research group characterized proanthocyanidins, built with flavan and flavan-3-ol monomeric units, from D. monetaria using high-resolution mass spectrometry [12].

Nuclear magnetic resonance (NMR) is widely used for the identification of proanthocyanidins. However, it is difficult to determine the positions of interflavan bonds, such as 4→8 or 4→6 [13]. Recently, new studies demonstrated a successful approach to identify and determine both the relative and the absolute configuration of such compounds on the basis of NMR chemical shifts and experimental electronic circular dichroism (ECD) spectra [14,15,16]. In the work presented here, we compared the calculated and experimental ECD spectra in addition to extensive NMR studies in order to determine the complete structure of a proanthocyanidin.

2. Results and Discussion

2.1. Compound Identification

Compound 1 was isolated as a reddish-brown amorphous powder. The molecular formula C30H26O9 was determined by HR-ESI-MS as m/z 529.1504 ([M − H]− calculated for C30H25O9− 529.1504). HR-ESI-CID-MS2 resulted in fragments m/z 511.1402, 419.1140, 409.0932, and 289.0721 (Scheme S1, Supporting Information) indicating compound 1 to be a dimeric proanthocyanidin (Scheme S2, Supporting Information). The constitution and relative configuration of 1 were determined by detailed 1D- and 2D-NMR measurements (Figures S1–S7, Supporting Information). The 13C-NMR spectrum (Figure S1) revealed two signal sets with 30 signals each. Despite partial signal overlap in the 1H-NMR spectrum (Figure S2), the proton signals could also be assigned to the two signal sets using the HSQC spectrum (Figure S3). Since all 1H-NMR signals of the two sets had the same correlations in the COSY, TOCSY, ROESY, and HMBC 2D-NMR spectra (Figures S4–S7), and the 1H−1H coupling constants were almost the same for both sets, two atropisomers were present. Rotational isomerism in proanthocyanidins often results in broad NMR signals [13,14,16]; however, for compound 1, both atropisomers were apparently stable at room temperature in the solvent used (CD3OD). From the quantitative evaluation of the 1H-NMR spectrum, a molar ratio of 1.6:1 was obtained for the major and minor atropisomers. From the COSY and TOCSY spectra, as well as the analysis of 1H signal multiplicities and coupling constants, four aromatic spin systems could be assigned (for the numbering scheme, see Scheme 1): an AA’BB’ system (H-B2′/6′, H-B3′/5′), two tri-substituted aromatic rings (H-A5, H-A6, H-A8 and H-E2′, H-E5′, H-E6′), and one isolated proton (H-D6). Furthermore, two heterocyclic four-spin systems could be identified: H-C2, H-C3A, H-C3B, H-4 andH-F2, H-F3, H-F4A, H-F4B, with oxygen substitution at C-C2, C-F2, and C-F3, as shown by the corresponding 1H and 13C chemical shifts (Table 1 and Table S1). The assignment of the six-spin systems to rings A, B, C, D, E, and F was possible on the basis of the HMBC and NOE correlations (see Table 1; Scheme 1). H-C2 showed HMBC correlations to C-B1′ and C-B2′/6′; both protons attached to C3 to C-A10, while H-C4 showed correlations to C-A5, C-A9, C-A10. On the other hand, HMBC correlations of H-F2 to C-E1′ and C-E2′/6′, of H-F3 to C-F10, and of both protons at F4 to C-D5, C-D9, and C-D10 were identified. The linkage of the two molecular parts followed from the HMBC correlation of H-C4 to C-D7, C-D8, C-D9 on the one hand and the correlation of H-F2 to C-D9 on the other hand. Due to the large vicinal coupling constants, both protons, H-C2 and H-C4, were axially oriented (3JH-C2/H-C3A = 11.6 Hz; 3JH-C4/H-C3A = 12.0 Hz). H-F2 was also axially aligned, resulting from the NOE to one of the protons at F4 (H-F4 at 2.846 ppm). Since the coupling constant between the axial proton H-F4A and H-F3 was small (3JH-F3/H-F4A = 4.6 Hz), the latter was equatorially aligned.

Scheme 1

Structure of compound 1 with axial position of the protons at C2, C4, and F2, as well as the OH group at F3, and the key HMBC correlations.

Table 1

NMR data (600/150 MHz, CD3OD, +25 °C) of the major rotamer of entcassiaflavan-(4β→8)-epicatechin (1).

Pos. a	δ 13C (ppm)	δ 1H (ppm) m (J [Hz])	Key HMBC (H→C)	Key NOE
C2	80.43	4.974 dd (11.6/2.0) ax	C4, B1′, B2′/6′	C3 eq, C4, B2′/6′
C3	36.71	2.640 ddd (13.2/12.0/11.6) ax; 1.836 ddd (13.2/5.8/2.0) eq	C2, C4, B1′, A10, D8; B4, A10	B2′/6′; C2, C4
C4	33.05	4.771 ddd (12.0/5.8/1.1) ax	B2, B3, A5, A9, A10, D7, D8, D9	C2, C3 eq, A5
A5	129.84	6.709 dd (8.4/1.2)	A7, A8, A9	A6
A6	108.99	6.262 dd (8.4/2.5)	A7, A8, A9, A10	A5
A7	156.75	-
A8	103.97	6.311 d (2.5)	A6, A7, A9, A10
A9	157.26	-
A10	120.12	-
B1′	134.77	-
B2′/6′	128.62	7.056 d-like (8.6)	B6′/2′, B4′	B3′/5′
B3′/5′	115.95	6.660 d-like (8.6)	B1′, B5′/3′, B4′	B2′/6′
B4′	157.99	-
F2	79.30	4.731 br s-like ax	F3, F4,E1′, E2′, E6′	F3, F4ax, E2′, E6′
F3	67.44	4.067 ddd (4.6/2.3/1.2) eq	F2, D10	F2, F4ax, F4eq, E2′, E6′ (w) b
F4	29.35	2.846 dd (16.8/4.6) ax; 2.757 ddd (16.8/2.3/0.9) eq	F2, F3, D5, D9, D10; F2, F3, D5, D9, D10	F2 ax, F3; F3
D5	156.03	-
D6	96.09	6.081 s	D5, D7, D8, D10
D7	155.74	-
D8	110.05	-
D9	155.50	-
D10	100.78	-
E1′	131.89	-
E2′	114.14	6.540 dd (2.1/0.6)	E3′, E4′, E6′
E3′	145.72	-
E4′	145.43	-
E5′	115.95	6.638 d (8.2)	E1′, E3′, E4′	E6′
E6′	119.76	6.132 ddd (8.2/2.1/0.6)	E4′	E5′

a For numbering scheme, see Scheme 1 (for the sake of clarity, the position number is preceded by the designation A, B, C, D, E, or F of the corresponding ring.); b w: weak correlation.

As result, the analysis of the 2D-NMR spectra allowed the assignment of all 1H and 13C chemical shifts for both atropisomers (see Table S1) and the determination of the relative configuration at the four stereo centers, but it could not be determined whether the major isomer had a P or M configured biaryl axis.

2.2. Determination of the Absolute Configuration and the Biaryl Position by Means of ECD Calculation

The NMR results showed the protons at C2, C4, and F2, as well as the OH group at F3, to be axially oriented (see Scheme 1); thus, the relative configuration at the four stereogenic centers was determined. However, according to NMR measurements on the nonderivatized compound, neither the absolute configurations nor the atropisomers could be deduced, due to missing NOE interactions between the upper and lower parts of the proanthocyanidin. Therefore, combined calculations of electronic circular dichroism (ECD) spectra and experimental values were used to determine both the absolute configuration and the atropisomers. Additionally, the proposed position of the biaryl linkage (C4–D8 instead of C4–D6; see Scheme 2) could be proven.

Scheme 2

Alternative structures of PA 1. (A) with C4–D8 and (B) with C4–D6 connectivity of the two ring systems.

Since the rotation about the C4–D8 bond is hindered by an energy barrier of 22.9 kcal/mol, the ECD spectra for each of the theoretically possible four stereoisomers C2R/C4R/F2R/F3R, C2R/C4R/F2S/F3S, C2S/C4S/F2R/F3R, and C2S/C4S/F2S/F3S (for numbering scheme see Scheme 2), referred to in this paper as RRRR, RRSS, SSRR, and SSSS, respectively, were calculated with two alternative atropisomers each. Thus, altogether, 16 calculated ECD spectra (eight each for C4–D8 and C4–D6 connections) were compared with the experimental spectrum (see Table 2 for an overview).

Table 2

Results of ECD calculations for the 16 isomers under consideration.

Biaryl Bond	Energy 1(kcal/mol)	Config. 2-Atropoisomer	S 3	Shift	Figure	Config. 2-Atropisomer	S 3	Shift	Figure
4–8	0.0	RRRR-P	0.9674	+19	Figure 1A	SSSS-M	0.6071	−30	Figure S8
4–8	1.7	RRRR-M	0.9642	+30	Figure 1C	SSSS-P	0.8631	−3	Figure S15
4–8	0.0	SSRR-P	0.7511	−26	Figure S9	RRSS-M	0.6300	+16	Figure S10
4–8	2.4	SSRR-M	0.7720	+20	Figure S16	RRSS-P	0.7557	−21	Figure S17
4–6	0.0	RRRR-M	0.6709	−5	Figure S11	SSSS-P	0.7564	−26	Figure S18
4–6	2.3	RRRR-P	0.9278	+27	Figure 2	SSSS-M	0.8379	−1	Figure S12
4–6	0.0	SSRR-P	0.7730	−30	Figure S13	RRSS-M	0.8427	+29	Figure S20
4–6	1.3	SSRR-M	0.8580	−2	Figure S19	RRSS-P	0.7863	−27	Figure S14

1 Energy difference is the energy relative to the lowest energy found by the DFT calculations. 2 Order of the stereogenic C atoms: C-2/C-4/F-2/F-3 (for numbering, see Scheme 1 and Scheme 2). 3 Relative similarity of the calculated with the experimental ECD spectrum.

Below, we discuss in detail only the ECD spectra for the nine isomers with the lowest conformational energy (Figure 1 and Figure 2 and Figures S8–S14). The ECD spectra and structures of the related energetically disfavored atropisomers are shown in the Supplementary Materials (Figures S15–S20).

Figure 1

(A) Calculated ECD spectrum (red curve) for the most stable conformation of the C4–D8 atropisomer with a high similarity of 0.9674 and a shift of 19 nm to the experimental ECD-spectrum (black curve); the spectrum of the C4–D8 SSSS M enantiomer is shown by the blue curve, which does not fit. (B) Related structure with the dihedral angle (A10–C4–D8–D9) = 135.9°. (C) Calculated ECD spectrum (red curve) for the C4–D8 atropisomer with a relative energy of 1.7 kcal/mol and a high similarity of 0.9642, but a high shift of 30 nm to the experimental ECD spectrum (black curve); the spectrum of the C4–D8 SSSS P enantiomer is shown by the blue curve, which does not fit. (D) Related structure with the dihedral angle (A10–C4–D8–D9) = −60.6°. Combining the Boltzmann weighted spectra of the most stable structures ( and atropisomer) gives a ratio of P/M = 19:1 and a slightly increased similarity of 0.9682. However, the superimposed spectrum looks almost identical to that shown in Figure 1A.

Figure 2

(left) Calculated ECD spectrum (red curve) for the P atropisomer with a relative energy of 2.3 kcal/mol of the C4–D6 isomer (M isomer: see Figure S11) with a similarity of 0.9278 and a shift of 27 nm to the experimental ECD spectrum (black curve); (right) related structure with the dihedral angle (A10–C4–D6–D5) = −61.4°.

In general, the analyses carried out showed that the calculated and experimental ECD spectra of isomers with a C4–D8 connection showed subtle differences when compared with those with a C4–D6 connection. This could be a result of many different factors, such as solution state effects and level of theory, as reported by Stephens et al. [17]. Nevertheless, the results presented below show that the method used in this work was successful in assigning the structure (mostly likely) of compound 1.

The calculated ECD spectrum that fit best with the experimental one resulted in a structure with C4–D8 connectivity with the P atropisomer RRRR-configuration (Figure 1A). For comparison, the figure includes the mirrored spectrum for the enantiomer (blue curve for SSSS-M) to visualize the difference between both. The superposition (with low similarity 0.6071) of the SSSS-M spectrum with the experimental one is displayed in the Supplementary Materials (Figure S12). The very high similarity of 0.9674 with a low shift of 19 nm between the calculated and experimental ECD spectra is a strong indication that this configuration represents the true experimental structure, particularly when compared with the much lower similarity of the other calculated spectra for C4–D8 connectivity (Figures S8–S10 and S15–S17). However, for the M atropisomer, the calculated and experimental ECD spectra also fit quite nicely (similarity 0.9641) with the C4–D8 connection and RRRR configuration (Figure 1C), but with a significantly higher shift of 30 nm. The NMR spectra indicated the occurrence of both atropisomers. The M atropisomer had a relative energy of 1.7 kcal/mol, resulting in a Boltzmann weighted distribution of P/M = 19:1. 1H-qNMR data showed a relative ratio of 1.6:1 for the two atropisomers. From our ECD calculations, taking into consideration the slightly higher conformational energy and the higher shift of the calculated spectrum when superimposed with the experimental one for the M atropisomer, it seems that the P atropisomer was the preferred one, even though the molar ratio differed from that determined by qNMR. The reason could be missing entropy contributions for the free-energy calculations or the choice of insufficiently high basis sets for the DFT energy calculations. The Boltzmann weighted sum of both spectra resulted in a slightly enhanced similarity to the experimental one of 0.9682, but the fitted curves were almost identical to those shown in Figure 1A. The comparison of the calculated ECD spectra with the experimental one for the structures with C4–D6 connectivity (Figure 2 and Figures S11–S14 and S18–S20) showed a high similarity of 0.9278 only for the energetically disfavored (2.3 kcal/mol) P atropisomer with RRRR configuration (Figure 2), but a shift of 27 nm. The unfavorable energy and the high shift make it very unlikely that this structure represents the true one. All other ECD spectra for structures with C4–D6 connectivity did not fit with the experimental one.

Thus, in conclusion, on the basis of the comparison of all 16 calculated ECD spectra with the experimental ones, including the consideration of relative conformational energies, the most likely structures shown in Figure 1 were characterized by C4–D8 connectivity and RRRR configuration. Accordingly, compound 1 is entcassiaflavan-(4β→8)-epicatechin. This compound was described by Coetzee et al. [7], and it was isolated from Senna petersiana as a 4′,7-di-O-methyl-ent-cassiflavan-(4β→8)-3′,4′,5,7-tetra-O-methyl-3-O-acetylepicatechin derivative.

3. Materials and Methods

3.1. General Experimental Procedures

The 1D (1H, 13C) and 2D (1H, 13C gHSQCAD, 1H, 13C gHMBCAD, 1H, 13C gH2BCAD, 1H, 1H DQFCOSY, 1H, 1H zTOCSY, 1H, 1H ROESYAD) NMR spectra were measured with an Agilent VNMRS 600 instrument at 599.83 MHz (1H) and 149.84 MHz (13C) using standard CHEMPACK 8.1 pulse sequences implemented in the VNMRJ 4.2A spectrometer software. All spectra were obtained with CD3OD as solvent at +25 °C. 1H and 13C chemical shifts were referenced to internal hexamethyl disiloxane (δH 0.062 ppm; δC 1.96 ppm), with the following parameters: TOCSY mixing time = 80 ms; ROESY mixing time = 300 ms; HSQC optimized for 1JCH = 146 Hz; HMBC optimized for nJCH = 8 Hz.

High-resolution ESI-MSn (m/z 100 to 2000) analyses were performed by direct injection into an Orbitrap Elite Mass spectrometer (Thermofisher Scientific, Waltham, MA USA), and HR-ESI-CID-MS2 analyses were performed using a collision energy dissociation (CID) of 25 eV, in a negative mode of ionization. Data were acquired and processed using the Xcalibur® 2.2 software.

ECD data were acquired on a JASCO J-815 CD spectrometer (solvent: methanol).

3.2. Extraction and Isolation of Compound

The subfraction SL6–5 (27 mg) was obtained from the ethyl acetate fraction of Dalbergia monetaria L. leaves from our previous study [12]. The subfraction was purified using preparative high-performance liquid chromatography (HPLC, Knauer system coupled with a WellChrom K-1001 pump and WellChrom K-2501 UV detector) using an ODS-A column (5 μm, 120 Å, 150 × 10 mm ID, YMC, USA). The mobile phase was H2O (A) and MeOH (B; Fluka Analytical, HPLC-MS grade Chromasolv®) (acidified 0.1% formic acid, v/v). The flux rate was 16 mL/min. The gradient was as follows: 2–30% B, 0 to 15 min; 30–70% B, 15 to 25 min. An isocratic condition was established (70% B) for 2 min before the reverse gradient of 2% B for 3 min. Compound 1 (14 mg) eluted at 13.49 min.

3.3. Computational Methods

All structures were constructed using the Molecular Operating Environment (MOE) software [18]. LowMode molecular dynamics simulations were applied for conformational search using the MMFF94 molecular mechanics force field [19]. Except for two alternative conformations of each phenolic hydroxyl group, only two low-energy conformations (atropisomers) resulted. The force field minimum-energy structures were subsequently optimized by applying density functional theory (DFT) using the BP86 functional with the def2-TZVPP basis set [20,21,22,23,24] implemented in the ab initio ORCA 3.0.3 program package [25]. The influence of methanol solvent was included in the DFT calculations using the COSMO model [26]. For the estimation of the rotational barrier for C4–D8 connectivity (Scheme 1), a conformational scan around the C4–D8 bond in steps of 5° was performed. The quantum chemical simulations of the UV and ECD spectra were also carried out using ORCA. For this purpose, the first 50 excited states of each enantiomer and conformation were calculated by applying the long-range corrected hybrid functional TD CAM-B3LYP with the def2-TZVP(-f) and def2-TZVP/J basis sets [22,23,24]. The ECD curves were visualized with the help of the software SpecDis 1.64 [27,28] from the calculated rotatory strength values using a Gaussian distribution function at a half-bandwidth of σ = 0.3 eV. In all cases, the ECD spectra were superimposed with the experimental one to reach maximal similarity using SpecDis. A maximum shift of ±30 nm was allowed. Alternative superpositions based on UV spectra gave no reasonable results because of the rather flat curve of the experimental one (Figure S21, Supplementary Materials). Since enantiomers show mirrored spectra, the calculated spectra of the opposite configurations listed in Table 2 were mirrored and additionally superposed with the experimental ones. Since alternative hydroxyl group conformations influenced the calculated ECD spectra only minimally, and the energies between atropisomers were mostly more than 2 kcal/mol different from each other, the energetically unfavored conformations contributed less than 5% to Boltzmann statistical weights. Therefore, only single spectra were compared with the experimental spectrum except when discussing the relevant structures related to Figure 1.

4. Conclusions

For the first time, entcassiflavan-(4β→8)-epicatechin (1) was isolated from a plant of the genus Dalbergia, from Dalbergia monetaria L.f. This is also the first isolation of the underivatized compound. The constitution and relative configuration were determined by detailed MS and NMR studies. All 1H and 13C signals for the two atropisomers could be assigned.

By comparison of the experimental electronic circular dichroism (ECD) spectra with the calculated ECD spectra, for the 16 reminiscent possible stereoisomers (post NMR), the absolute configurations concerning the four stereocenters could be determined. In addition, this method confirmed the position of the interflavan bond and assigned the main compound (most likely) to be the P atropisomer. Thus, we could demonstrate that, without considerable effort, e.g., by derivatization, breakdown, or synthesis of a parent compound, the absolute configuration, including atropisomerism, could be determined on the basis of NMR data and ECD spectral calculations in comparison with the experimental spectrum.