Anti-proliferative activity of a novel tricyclic triterpenoid acid from Commiphora africana resin against four human cancer cell lines

Myrrh, a resin derived from the damaged bark of Commiphora genus, has traditionally been used for treatment of various human diseases, such as amenorrhea, ache, tumors, fever, and stomach pains. In spite of this widespread use of the myrrh in Ethiopia, the pharmacological activity and chemical composition have not been studied in detail. A new tricyclic triterpene acid (3S,4S,14S,7E,17E,21Z)-3,30-dihydroxypodioda-7,17,21-trien-4-carboxylic acid (commafric A) has been isolated from a crude methanolic extract of Commiphora africana (A. Rich.) Engl. resin along with the known pentacyclic triterpene α-amyrin. The structure of commafric A was characterized using different spectroscopic techniques such as 1D and 2D NMR, IR, and VCD combined with computations. The anti-proliferative activity of both isolated compounds was evaluated using SRB based colorimetric cellular assay against four human cancer cell lines. Etoposide was used as a positive control. Commafric A showed significant anti-proliferative effects against non-small cell lung cancer (A549) with IC50 values of 4.52 μg/ml. The pentacyclic triterpene α-amyrin showed a weak anti-proliferative activity against A2780 (ovarian cancer), MIA-PaCa-2 (pancreatic cancer), and SNU638 (stomach cancer) cell lines tested with IC50 values ranging 9.28 to 28.22 μg/ml. Commafric A possessed anti-proliferative activity against non-small cell lung cancer (A549), which suggests that commafric A has potential to be further optimized being a lead compound in the search for new drugs against cancer diseases.

In spite of the use of Commiphora africana in traditional medicine, to the best of our knowledge, there has been no study aimed at isolation and characterization of anti-proliferative constituents of this plant. As part of the exploration of the pharmacological potential of the resins from Commiphora genus, we herein report the isolation, structural elucidation, and anti-proliferative activity screening of a new oxygenated podioda-7,17,21-triene [21,22] and α-amyrin obtained from C. africana.

Plant material
Resins and botanical specimens of C. africana were collected from district of Yabello, Borena zone, Ethiopia in December 2016 and identified by a botanist Shambel Alemu at the Biology Department, Addis Ababa University. A voucher specimen has been deposited at the National Herbarium (number: 072801) of Addis Ababa University, Addis Ababa, Ethiopia. This tree is known by its vernacular name hammesa Dihro (Afaan Oromo). There is no permission required to collect samples from wild to investigate their chemical composition and evaluate their anti-proliferative effects.

General experimental procedure
Melting points are uncorrected. TLC was performed on precoated plates (Silica gel 60 F 254 , Merck) using n-hexane/EtOAc/MeOH (10:10:0.1) as a developing solvent and with vanillin-H 2 SO 4 as the detecting reagent. CC was performed on silica gel. IR spectra were measured on a Perkin Elmer 1600 instrument using KBr tablets. Optical rotation was measured on a Perkin-Elmer 241 polarimeter. UV spectra were recorded on a Shimadzu UV-VIS spectrophotometer.  1 H and 13 C NMR data for commafric A (1) was acquired on an Avance III HD spectrometer (Bruker-BioSpin GmbH, Germany) equipped with an inverse detected TCI cryoprobe with a cryogenic enhancement for 1 H, 2 H, and 13 C, operating at 600 MHz for 1 H. All spectra were recorded using TopSpin 3.5pl7 at 298 K in dichloromethane-d 2 and chloroform-d 1 , using gradient selected and adiabatic inversion versions of pulse sequences where applicable. Analysis of NOE buildup for compound 1 was performed by complete relaxation matrix analysis using Mspin 2.3.2-694 (MestreLab Research S. L., Spain). The results are presented as Boltzmann averaged NOE enhancement across the conformational ensemble generated for analysis of IR and VCD section. The sample for anisotropic NMR measurements (10 mg) was dissolved in CDCl 3 with 0.03% TMS to which increasing amounts of poly-γ-benzyl-l-glutamate (PBLG, 150,000-300,000 Da, Sigma-Aldrich CAS 25014-27-1) was added to yield weight-to-volume ratios of 7.9%, 12.1%, and 16.2%. For compound 1 1D carbon, CLIP-HSQC, IPAP-HSQC, and J-res carbon was acquired to extract the 1 J CH coupling constants and anisotropic carbon shifts. The data treatment is thoroughly described in additional file.

Experimental vibrational circular dichroism (VCD) spectra
VCD and absorption spectra of 1 were measured in CDCl 3 solutions (~ 1 mg/100 μl) using a BaF 2 cell of an optical path of 50 μm and a BioTools ChiralIR-2X  instrument. Spectra of pure CDCl 3 solvent were subtracted as a baseline; the accumulation time was ~ 12 h using blocks of 1200 scans and 4 cm −1 resolution.

Computations
Using the Gaussian suite of programs [23] a systematic conformer search of compound 1 was performed considering the three torsional angles in the vicinity of the tricyclic system. Other C-C-C-C dihedral angles were all-trans and all-extended (180°) at the beginning of the optimizations and left to optimize without constraints. The B3LYP [24] PCM [25] (chloroform)/6-31 ++G** level was used for all quantum chemistry. For stable conformers thus obtained the magnetic field perturbation theory [26,27] was used to simulate IR and VCD intensities; final spectra were obtained as a Boltzmann average, using a convolution with Lorentzian lines (FWHM = 10 cm −1 ). Since the consideration of the full molecular flexibility was not possible with our computational means, atomic axial tensors of the linear side chain (from carbon number 16) were deleted for VCD generation. The chain itself is not chiral and thus supposedly the error introduced by this approximation is small, whereas the dominant signal from the chiral more rigid molecular part may help to assign the absolute configuration. Alternated simulations of the spectra based on molecular dynamics were attempted as well, but did not provide results significantly different from the limited conformer model. Boltzmann-averaged isotropic shielding and spin-spin coupling constants were computed at the same level as for VCD.

Tumor cell lines
The cell lines used in this study were the human nonsmall cell lung cancer cell lines (A549), ovarian cancer cell line (A2780), pancreatic cancer cell line (MIA-Paca-2), and stomach cancer cell line (SNU-638), and they were maintained using RPMI1640 cell growth medium (Gibco, Carlsbad, CA), supplemented with 5% fetal bovine serum (FBS) (Gibco), and grown at 37 °C in a humidified atmosphere containing 5% CO 2 [28]. All the cancer cells used were adapted for 6 months at least to the RPMI1640 medium at Korea Research Institute of Chemical Technology, then used in the anti-proliferative assays. Cells were exposed to the isolated compounds at concentrations ranging from 0.1 to 30 μg/ml.

Anti-proliferative assessment
All experiments were conducted by the NCI's protocol [28]. Experimental cultures were placed in 96-well microtiter plates (Corning) containing 0.15 ml of growth medium per well with a cell density of 5 × 10 3 (A549, A2780, and MIA-Paca-2) and 1 × 10 4 (SNU-638). The culture was incubated at 37 °C and humidified by 5% CO 2 for 1 day. Then, media were aspirated off and added the test material in triplicate which was dissolved in media at varying concentrations. In case of necessity, the test material was dissolved in small amount of dimethyl sufoxide (DMSO), but the final concentration of DMSO in the medium did not exceed 0.5%. The culture was incubated for additional 3 days. After 3 days of continuous drug exposure, the medium was removed by flicking plates over a sink. The cells attached to the plastic substratum were fixed by gently layering with 0.1 ml of cold 10% trichloroacetic acid (TCA). Incubation at 4 °C for 1 h was followed by wash with tap water five times to remove TCA solution. After being washed and dried at RT overnight, TCA fixed cells were stained with 0.1 ml of 0.4% sulforhodamine B (SRB) in 1% acetic acid per well for 30 min. At the end of the staining period, SRB supernatant was removed and the remaining cells were rinsed with 1% acetic acid five times. It was dried until no standing moisture was visible, and then the bound dye in cells was extracted with 0.1 ml of l0 mM unbuffered Tris base (pH 10.5) per well by stirring on a gyratory shaker for 5-10 min, followed by measured the optical density at 520 nm by MR700 microplate reader (Dynatech Laboratories). Antitumor activity of the test material at varying concentrations was estimated as the net growth % of cells compared with that of control (without test material, net growth = 100%). The dose-response curves of test material were constructed and the IC 50 value was calculated as the concentration of the test material that caused 50% inhibition of cell growth against the cancer lines A549, A2780, MIA-PaCa-2, and SNU638. Etoposide, a standard drug was used as a positive control. IC 50 values of etoposide were compared with those for crude extract, n-hexane fraction, and isolated compounds.

Isolation of compounds
Flash column chromatography of the n-hexane fraction using n-hexane/EtOAc (7:3) as an eluent furnished two pure compounds: commafric A (1) the most abundant isolated molecule and the known compound α-amyrin (2), were obtained as white solids. The structures of the isolated compounds were elucidated employing 13 Fig. S1).
The UV spectrum of 1 (EtOH) exhibited an absorption band at 209 nm indicating the absence of conjugated double bonds. The experimental IR-spectrum agrees with the simulated spectrum of commafric A (Fig. 2); the most prominent features comprise free (3607 cm −1 ) and hydrogen-bonded (3400 cm −1 ) OH stretching bands, split (2956/2872 cm −1 ) CH stretching signal indicating the presence of both sp 2 and sp 3 carbons, and the C=O stretching band (1695 cm −1 ). 1 H NMR resonances (Table 1)  The 13 C NMR (Table 1 and Additional file 1: Fig. S3) and multiplicity edited HSQC/DEPT confirmed the presence of thirty carbons, of which seven were quaternary, ten methylene, seven methine, and six methyl groups. A set of 2D NMR experiments, HSQC, HMBC, H2BC, HSQC-TOCSY, DQF-COSY, and 1,1-ADEQUATE (Additional file 1: Fig. S4-S9), was used to establish the molecular framework of commafric A as a tricyclic triterpenoid with structural similarity to podioda-7,17,21trienol [21]. All carbon-carbon connections were unambiguously confirmed by 1,1-ADEQUEATE, except for the two quaternary to quaternary carbon bonds (C4 to C23 and C8 to C13), which did not correlate in this pulse sequence because of the absence of protons on both sides of the C-C bond (Fig. 3b). The C23 carboxyl group could be placed by HMBC correlations to H3, H5, and H24 from C23, while the C8 to C13 bond was confirmed by HMBC correlations between H26 and H12, and C8. The location of the hydroxymethylene group as well as the position of the 21, 22 double bond was confirmed by 2D HMBC (Fig. 3a), which showed three bond couplings between the olefinic proton H21 (δ H 5.30) and the carbons at the C29 methyl group (δ21.4) and the C30 hydroxymethylene carbon (δ C 61.5). Furthermore, H2BC two-bond couplings appeared between C22 and Me29 and hydroxymethylene protons H30′ / ″ (Fig. 3a).

Determination of the relative configuration
Based on 3 J CH couplings measured by selective CLIP-HSQMBC [32], as well as NOESY/ROESY analysis, the C17/C18 configuration was determined as E ( 3 J C28H17 = 8.0 Hz and 3 J C19H17 = 6.0 Hz) and the C21/C22 configuration was determined as Z ( 3 J C29H21 = 6.0 Hz and 3 J C30H21 = 9.3 Hz, Additional file 1: Figs. S10-S12). These conclusions are based on the fact that the trans-3 J CH is expected to be stronger than the corresponding cis-3 J CH . This was also consistent with the direct observations of the presence of the NOE H21H29 and ROE H21H29 correlations, and the absence of the NOE H21H29 and ROE H21H29 correlations. The Z-configuration of the C21 to C22-double bond is also consistent with the 13 C NMR chemical shifts of C21, C22, C29, and C30 in previously reported compounds with similar side chain configurations [33,34].
The relative configuration of the tricyclic system was established by NOESY and ROESY correlations (Additional file 1: Fig. S12) supplemented with some 3 J HH and 3 J CH couplings. Most notably, an N/ROE correlation trace between H3-H5-H9-H11″-H12″ place these protons below the ring plane while the trace between H24-H25-H11′-H12′-H26 place these above the ring plane. The key correlations are displayed in Fig. 3c. The configuration at the C14 stereocenter could not be unambiguously determined from the N/ROE correlations without a full conformational analysis because of relatively free rotation about the C13-C14 bond.

Determination of the C14 configuration Vibrational circular dichroism spectroscopy
In order to relate the measured VCD spectra to the structure, theoretical VCD spectra were simulated computationally. The theoretical modelling focused on the tricyclic system supposedly dominating in VCD spectrum, because a full conformer analysis (> 10 6 possible conformers) was not possible with available computational means. A reduced ensemble of 27 conformers was generated by 120° rotations around the C13-C14, C4-C23, and C3-O3 bonds only, and relative conformer energies were obtained for both the R and S C14 isomers. The molecular tail (C14-C30) was kept in the extended conformation and its contribution to VCD was not considered. The equilibrium geometries and spectral properties were calculated using the Gaussian software [24] and the B3LYP/6-311 ++G**/PCM (CHCl 3 ) level of theory, relative enthalpies were used for property averaging. The simulation reproduced some spectral features observed experimentally (Fig. 4), although the theoretical R and S VCD sign patterns of 1 are similar. The 1150-1000 cm −1 spectral region is however predicted to be significantly

Isotropic NMR parameters
For a more detailed NMR data analysis of distances and coupling constants localized near the C14, only the three Boltzmann averaged rotamers around the C13-C14 bond were used. The H14 proton is preferentially pointing towards the edge of the ring system (and H-7) in both the R-and the S C14 configurations (Table 2). This was problematic for using the otherwise very informative 3 J CH couplings originating from the only proton (H14) directly attached to the rotating C13-C14 bond since the dihedral angles formed between the H14-C14 vector and the C13-C8, C12, C26 vectors will be very similar for R-and S. The predicted differences between the two configurations were smaller than 0.5 Hz. The experimental couplings (Table 3) are in agreement with the calculated rotamer populations of both configurations. These couplings were very challenging to measure accurately experimentally as the signal of H14 overlaps with that of H15s. That made it necessary to use heavily chemical shift filtered selective methods and there is unavoidable phase modulation from J-coupling. Coupling sums were utilized because of significant second order/phase contributions. For all these reasons, the experimental error is estimated to at least 1 Hz. Furthermore, the possibility to determine the relative configuration from NOE buildup rates involving the C27 methyl group were investigated. The C27 methyl group is predicted to be predominantly anti to the C26 methyl in the S-configuration and gauche to the C26 methyl in the R-configuration. NOE buildups are intrinsically difficult to quantify for flexible molecules due to the fact that they depend on the distance as r −6 . A scarcely populated rotamer can contribute significantly to the observed NOE if the two protons are positioned very close to each other. The measured NOEs around and across the C13-C14 bond (Additional file 1: Fig.  S13) are not conclusive alone as both theoretical configurations result in r 2 values in the order of 0.7. However, the most important NOE between H26 and H27 that is expected to be the most sensitive to the C14 configuration indicates the S-configuration. The weighted experimental NOE was determined to η = 0.0054 (corresponding to r = 2.9 Å), i.e., closer to the value predicted for the S-model (η = 0.0032, r = 3.2 Å), than that predicted for the R-model (η = 0.0144, r = 2.5 Å).

Confirmation of the C14 configuration by anisotropic NMR parameters
Anisotropic NMR parameters have emerged as powerful tools in structural elucidation [35,36]. Residual dipolar coupling (RDC) depend on the relative orientation of the 13 C-1 H bond vectors, while residual chemical shift anisotropy (RCSA) depend on the relative orientations of the carbon chemical shielding tensors. When combined, they can provide the configuration of stereogenic centers that are difficult to establish by traditional methods. A method described by Liu et al. [35] utilizing poly(γbenzyl-l-glutamate) (PBLG) to form a liquid crystal that induces anisotropy in chloroform-d 1 was employed in the present study. 13 C residual chemical shifts were referenced to TMS, and combined with RDC data from HSQC-IPAP spectra (Additional file 1: Figs. S14-S16, Tables S1 and S2). In the comparison of the experimental and theoretical anisotropic parameters for the commafric A the S-model provided lower quality factors (Q = 0. 19) than the R-model (Q = 0.33, Fig. 5). Based on the anisotropic parameters, together with the VCD and NOE data, we conclude that commafric A's stereocenter C14 has the (S)-configuration.

In vitro anti-proliferative effect of the MeOH extract, n-hexane fraction, and isolated compounds
The anti-proliferative effects of the crude MeOH extract, n-hexane fraction, and isolated compounds commafric A and α-amyrin were tested for four cancer cell lines, A549 (non-small cell lung cancer), A2780 (ovarian cancer), MIA-PaCa-2 (pancreatic cancer), and SNU638 (stomach cancer). The crude methanol extract showed a strong anti-proliferative activity against A549 (IC 50 = 3.55 μg/ ml) ( Table 4, Additional file 1: Fig. S17). The crude extract exhibited weaker activity against A2780, SNU-638, and MIA-Paca-2 compared with A549 cell lines. Among the cell lines tested the methanol crude extract showed strong effect on A549 cell lines (3.55 μg/ml). The n-hexane fraction exhibited significant inhibition of cell proliferation on the all four cell lines with dose dependent relationship in vitro. However, it was less sensitive and had a weaker net growth effect on A549 cell lines compared with the crude methanol extract (Table 4, Additional file 1: Fig. S17).

Discussion
A number of bioactive compounds were previously reported from different Commiphora species. In this study, the n-hexane fraction of C. africana resin yielded a new compound commafric A. Cancer, cardiovascular diseases, chronic respiratory disorders and diabetes are the primary cause of deaths worldwide. The knowledge of herbal medications and anticancer activity tests assist in the development of important anticancer drugs [37].
In our study for in vitro anti-proliferative activity, we found that the crude extract and related n-hexane fraction is lower than 30 μg/ml, which is within the limit of criteria set by the American National Cancer Institute for further purification [38]. For compound 1, a moderate anti-proliferative activity was observed for the A549 cancer cell lines, with IC 50 value of 4.52 μg/ml. This is comparable with the activity of the crude extract and more than twice better than for α-amyrin. On the other hand, compound 1 showed weak anti-proliferative effects > 9 μg/ml for the other cell lines. The compound exhibited about twice higher activity on MIA-PaCa-2 compared with the methanol extract. The presence of carboxylic acid or terminal hydroxyl group in the compound might be the reason for its bioactivity. α-Amyrin (2) having an ursane skeleton showed weaker activity on A549, A2780, and SNU638 cell lines compared with both compound 1 and the crude methanol extract. Compounds with similar skeleton to α-amyrin such as ursolic acid and its derivatives showed strong anti-proliferative activity against ovarian carcinoma, pancreatic carcinoma, prostate cancer, cervical carcinoma, hepatic cancer, breast cancer, colorectal cancer, leukemia, neuroma, and colon adenocarcinoma [39]. The weak activity of α-amyrin might be due to lack of carboxylic acid group. According to the literature on α-amyrin pentacyclic triterpenes stimulate proliferation of human keratinocytes but do not protect them against UVB damage [40][41][42]. A review on antitumor effect of triterpene acid compounds revealed that triterpene acid type compounds have many effects including antiinflammatory, regulating blood sugar level, antiviral, and antitumor activity. More important, triterpene acid type compounds have become one of the most popular topics recently because of its selective toxic effects on cancer cells and harmless to normal cells [43].
The tricyclic triterpene acid (3S,4S,14S,7E,17E,21Z)-3,30-dihydroxypodioda-7,17,21-trien-4-carboxylic acid (commafric A) was isolated from the n-hexane fraction of the resin of C. africana and its structure was determined. It showed the variable IC 50 values against the four cancer cell lines studied. The anti-proliferative effect against non-small cell lung cancer (A549) cells with IC 50 values of 4.52 μg/ml was the highest among the cancer cells tested. The anti-proliferative effect suggests that commafric A has a potential in further investigations as a drug against different cancer lines. Further work is required to evaluate the mechanism of action of the commafric A as an antitumor agent against non-small cell lung cancer and evaluate its anti-proliferative on a number of cancer cell lines. Further isolation of compounds from the resin and other parts of this plant will be pursued which might yield some novel bioactive compounds.