Article pubs.acs.org/jnp
Quantitative Purity−Activity Relationships of Natural Products: The Case of Anti-Tuberculosis Active Triterpenes from Oplopanax horridus⊥ Feng Qiu,† Geping Cai,†,‡ Birgit U. Jaki,†,‡ David C. Lankin,† Scott G. Franzblau,†,‡ and Guido F. Pauli*,†,‡ †
Department of Medicinal Chemistry and Pharmacognosy, College of Pharmacy, University of Illinois at Chicago, Chicago, Illinois 60612, United States ‡ Institute for Tuberculosis Research, College of Pharmacy, University of Illinois at Chicago, Chicago, Illinois 60612, United States S Supporting Information *
ABSTRACT: The present study provides an extension of the previously developed concept of purity−activity relationships (PARs) and enables the quantitative evaluation of the effects of multiple minor components on the bioactivity of residually complex natural products. The anti-tuberculosis active triterpenes from the Alaskan ethnobotanical Oplopanax horridus were selected as a case for the development of the quantitative PAR (QPAR) concept. The residual complexity of the purified triterpenes was initially evaluated by 1D- and 2DNMR and identified as a combination of structurally related and unrelated impurities. Using a biochemometric approach, the qHNMR purity and anti-TB activity of successive chromatographic fractions of O. horridus triterpenes were correlated by linear regression analysis to generate a mathematical QPAR model. The results demonstrate that impurities, such as widely occurring monoglycerides, can have a profound impact on the observed antimycobacterial activity of triterpene-enriched fractions. The QPAR concept is shown to be capable of providing a quantitative assessment in situations where residually complex constitution contributes toward the biological activity of natural products.
W
patterns, which were not congruent with the results of the quantitative analysis. The present study explores quantitative purity−activity relationship (QPAR) as an extension of the PAR concept by enabling the quantitative evaluation of the effects of multiple minor components on the bioactivity of a purified natural product. The term QPAR is analogous to the more widely used terms quantitative structure−activity relationship (QSAR), which summarizes a mathematical relationship between related chemical structures and bioactivity in a regression or classification model. Recently, we have reported a biochemometric approach6 by which the chemical and biological profiles of successive chromatographic fractions of crude plant extracts can be correlated statistically for the identification of bioactive and potentially synergistic components, without the need for isolation. The present study integrates this methodology and develops an experimental procedure for the establishment of QPARs of natural products, as shown in Figure 1. Instead of using samples of pure natural products of different origin, the present study uses successive chromatographic fractions to perform impurity profiling and biological assessment. In these
henever natural products are purified from natural source material, they carry an individually characteristic signature in the form of a profile of minor constituents (impurities), called residual complexity (RC, see S1, Supporting Information), which originates from the metabolomic complexity of the biogenetic source.1−3 As the majority of natural product purification efforts focus on the identification of “pure” compounds, co-eluting minor components can easily be overlooked, particularly if they do not interfere with the spectroscopic analysis of the main component. However, there have been a few cases where the synthetic and natural compounds were found to be chemically but not biologically identical.4,5 These findings suggested the presence of trace but highly potent constituents in the natural product samples. Therefore, it is considered absolutely necessary to evaluate the RC of natural products. Our previous study has introduced the concept of purity− activity relationships (PARs) in natural product research.1 The PAR profiles of natural products were generated by correlating qHNMR purity assessment with antituberculosis (TB) biological data of the same “pure” natural product from different sources. However, as these samples originated from various research and commercial sources, the different impurity profiles and degree of RC of each sample resulted in unique PAR
Special Issue: Special Issue in Honor of Lester A. Mitscher
⊥
Residual Complexity and Bioactivity, Part 17 (see S1, Supporting Information). © XXXX American Chemical Society and American Society of Pharmacognosy
Received: November 7, 2012
A
dx.doi.org/10.1021/np3007809 | J. Nat. Prod. XXXX, XXX, XXX−XXX
Journal of Natural Products
Article
authentic reference materials for creation of the standard curve required for quantification. In contrast, NMR spectroscopy can serve as a more universal and linear detector for biologically active samples and, thus, is better suited for chemical analyses of the underlying RC. Not only does NMR typically provide structural information, but it allows the simultaneous quantification of each detected analyte in a complex matrix, frequently without the need for further separation and, more importantly, without identical calibrants.3,7,8 Despite these advantages, the remaining challenges in NMR result from the complexity of the 1H NMR spectra, such as peak multiplicity and signal overlap. The present study applied a combination of 1D- and 2D-NMR spectroscopy to increase the spectral resolution and obtain additional information about the RC of purified anti-TB active triterpenes of Oplopanax horridus, an Alaskan ethnobotanical previously studied in our laboratory.9,10 For quantification of all 1H NMRdetectable constituents of the anti-TB active fractions, quantitative 1H NMR (qHNMR) was used, applying the modified 100% method scaled to the residual solvent signal as internal reference.
Figure 1. The four steps (S1−4) of the experimental design for the establishment of quantitative purity−activity relationships (QPARs) of natural products. S1: The target natural products are purified chromatographically, yielding a series of fractions represented by gray bars; S2: The fractions are subject to spectroscopic analysis for qualitative and quantitative identification of major and minor components; S3: In parallel to S2, the fractions are evaluated by a high-throughput-capable bioassay for the target bioactivity; S4: Statistical analysis of the chemical and biological data obtained in S1 and S2 generates a quantitative model that correlates purity and activity of the target natural products.
fractions, the various components are present at different concentrations, following the elution profiles of the last chromatographic purification step. The hypothesis was that it is possible to statistically correlate the concentrations of individual components with the bioactivity of each fraction. The study outcome led to a quantitative model, enabling a comprehensive analysis of bioactivity in residually complex samples as it relates to individual components and their possible interactions. The major challenge for the establishment of QPARs is the resolution of the RC of purified natural products. This involves both the chemical and biological complexities of multiple minor components. While LC-UV and LC-MS are used frequently in impurity profiling, these methods provide limited information for structural identification. The challenge increases further when the target analytes exhibit weak UV absorption and/or poor MS ionization, such as the triterpenes investigated here. Moreover, LC-based methods rely on the availability of
■
RESULTS AND DISCUSSION Purification and Identification. Oplopanax horridus (Devil’s Club) has been used by Native Americans to treat colds, arthritis, diabetes, and a variety of forms of cancer.11 Recent studies have shown that the crude extract of the inner stem bark of O. horridus inhibits the growth of Mycobacterium tuberculosis.9 Bioassay-guided fractionation led to the isolation of anti-TB active polyynes and sesquiterpenes.10,12 In a continuation of efforts to identify further anti-TB active constituents from O. horridus, an exhaustive fractionation procedure using NP-VLC was carried out for a combination of petroleum ether and CH2Cl2 sequential extracts of the dried fruits. Elution with stepwise gradients of n-hexane−EtOAc
Figure 2. Offline qHNMR analysis (methanol-d4, 600 MHz) of the preparative HPLC purification of triterpenes from the anti-TB active fractions of O. horridus. The stacked 1H NMR spectra showed that a known triterpene, 1, eluted in Fr-19 to Fr-24, while co-eluting with its isomer, 2, in Fr-23 and Fr-24. At the same time, all six fractions contained ca. 0.5−8.0 mol % of glyceride-type impurities, as implicated by their characteristic signals in the range δH 4.50−3.50. B
dx.doi.org/10.1021/np3007809 | J. Nat. Prod. XXXX, XXX, XXX−XXX
Journal of Natural Products
Article
Table 1. 13C and 1H NMR Spectroscopic Data of 1a
followed by EtOAc−MeOH gave a triterpene-enriched fraction (see the Experimental Section). A 100 mg sample of this fraction was subjected to RP-HPLC purification using isocratic elution with 90% MeOH. Upon TLC analysis of the HPLC fractions, Fr-19 to Fr-24 all showed a single pinkish spot with an apparently identical Rf value. However, 1H NMR analysis indicated that Fr-19 to Fr-22 contained mainly the known triterpene 3α-hydroxylup-20(29)-ene-23,28-dioic acid (1), whereas Fr-23 and Fr-24 were found to be a mixture of 1 and its isomer, 3α-hydroxyolean-12-ene-23,28-dioic acid (2). The ratio of the two isomeric triterpenes in Fr-23 and Fr-24 was measured as 7:3 and 3:7, respectively, by using their compound-specific marker signals: a doublet of triplets (dt) at δH 3.029 with J = 4.7 and 10.9 Hz (H-19) for 1 and a doublet of doublets (dd) at δH 2.853 with J = 4.2 and 13.8 Hz (H-18) for 2 (Figure 2). While the first isolation of 1 dates back 30 years,13 its 1H NMR assignment and elucidation of the relative configuration of its pentacyclic core, particularly of the methylene protons in the overcrowded aliphatic region (2.00−1.00 ppm), are still incomplete in the current literature. In the present study, based on the previously assigned 13C spectrum of 1,14 the COSY and HSQC spectra were used to guide the detailed assignment of the 1H NMR resonances of 1 (Table 1) for use in future structural identification and dereplication.
Resolution of Residual Complexity. An expanded view of the 4.50−3.50 ppm range of the 1H NMR spectra revealed an additional group of minor signals assigned to glyceride-type impurities based on their chemical shifts and splitting patterns. A stack of the six spectra (Figure 2) shows that these impurities were present in all six fractions, with a content of ca. 0.5−8.0 mol %. Additional minor signals were observed in the 5.70− 5.30 ppm range, indicating further RC of the purified triterpenes. As shown by this example, offline qHNMR analysis expands the visualization of elution profiles in multiple dimensions, where both major and minor constituents of a sample can be described qualitatively and quantitatively. While 1 H NMR is often insufficient to fully characterize all the components, 2D-NMR techniques are available to advance the analysis of complex mixtures, and thereby evaluate the RC. Classical COSY spectroscopy remains among the most popular forms of 2D-NMR spectroscopy and is performed routinely as a highly useful method for identifying different molecules through their unique cross-peak patterns.15 In order to verify the presence of minor glycerides in purified triterpenes, Fr-21 was selected and COSY was used to determine the presence of the typical 1H−1H correlations
positionb
δC [ppm]
δH [ppm]
integral
multiplicity
1a 1b 2a 2b 3 4 5 6a 6b 7ax 7eq 8 9 10 11ax 11eq 12ax 12eq 13 14 15ax 15eq 16ax 16eq 17 18 19 20 21a 21b 22a 22b 23 24 25 26 27 28 29a 29b 30
33.57
1.371 1.351 1.926 1.550 3.695
1H 1H 1H 1H 1H
ddd/mc ddd/mc dddd/mc dddd/mc t/dd
1.908 1.499 1.469 1.630 1.320
1H 1H 1H 1H 1H
dd dddd/mc dddd/mc dt/ddd ddd/mc
12.0, 8.2
1.577
1H
dd
13.0, 3.2
1.272 1.478 1.098 1.734 2.315
1H 1H 1H 1H 1H
dq/dddd dddd/mc ddt/dddd br d/dddd ddd
13.0, 4.2 13.3, 13.1, 4.2 13.3 13.2, 11.4, 3.7
1.544 1.185 1.438 2.243
1H 1H 1H 1H
dt/ddd dt/ddd dt/ddd dt/ddd
13.5, 13.5, 12.8, 12.9,
1.644 3.038
1H 1H
t/dd td/ddd
11.4 11.4, 4.8
1.946 1.395 1.928 1.470
1H 1H 1H 1H
m m m m
1.144 0.911 0.987 1.069
3H 3H 3H 3H
s s s s
4.723 4.605 1.711
1H 1H 3H
d dd d
26.11 73.70 52.33 45.72 22.30 35.17 42.45 51.86 38.05 21.86 26.87 39.67 43.78 30.81 33.36 57.51 50.45 48.50 152.00 31.71 38.15 180.07 17.86 16.98 16.84 15.16 177.50 110.16 19.54
J [Hz]
3.0
12.8, 3.7
3.9 3.3 3.4 3.4
2.5 2.5, 1.4 1.2
Solvent: methanol-d4 (500 μL); 1H: 600 MHz, 13C: 100 MHz. bThe axial (ax) and equatorial positions (eq) of some methylene protons were determined on the basis of multiplicity patterns and J values. c Even at 600 MHz, many of the methylene and methine proton resonances remain strongly overlapped and form high-order spin systems, which limits the identification of distinct multiplicity patterns. The multiplicity and J values provided are based on first-order assumptions. a
that can be expected for glycerides, i.e., characteristic signals in the range 4.50−3.50 ppm. Using a 400 MHz NMR spectrometer, the time-domain (TD) data size was set to 2048k/256k (F2/F1) and the number of scans (NS) was increased to 64, for an improved limit of detection (LOD) for these minor impurities. The acquired COSY spectrum clearly showed four 1H−1H correlations, which revealed the spin system of H-1 to H-3 in a 1-monoglyceride (imp1, Figure 3A). In the 1H NMR spectrum, a triplet signal of two protons at 2.372 ppm indicated the methylene group adjacent to the carbonyl group in the monoglyceride. Excluding the cross-peaks C
dx.doi.org/10.1021/np3007809 | J. Nat. Prod. XXXX, XXX, XXX−XXX
Journal of Natural Products
Article
Figure 3. Impurity profiling of a representative O. horridus triterpene sample (Fr-21) by COSY analysis (methanol-d4, 400 MHz). A 1-monoglyceride was implicated by its characteristic 1H−1H correlation pattern shown in region A. The observed correlations in region B indicate two additional minor impurities, both of which possess an allylic hydroxy group.
a prerequisite for their biological assessment. The example of the O. horridus triterpenes and the triterpene-enriched fraction also demonstrates that 1D- and 2D-NMR methods are powerful tools for qualitative and quantitative evaluation of RC, even if definitive assignments cannot be made due to limitations in structural information. Quantitative Purity−Activity Relationships. The 100% qHNMR method7,8 was applied in the quantification of individual components. Because the triterpene-enriched fractions resulted from repeated purification and were all carefully dried under identical conditions, the use of the absolute qHNMR method for the detection of residual solvent and moisture was considered unnecessary, especially as these minor components would not affect the biological activity in the assay. Thus, the molar content of individual components in each fraction was calculated as listed in Table 2. In addition to providing the results of purity evaluation, Table 2 also summarizes the anti-TB activity of the six samples, determined by the microplate Alamar blue assay (MABA). In the present study, the MIC values (μg/mL) are defined as the lowest drug concentrations causing 90% growth inhibition of M. tuberculosis.
belonging to the main component 1 in the HSQC spectrum, additional signals at δH ∼1.32 and δC ∼30.2 were assigned to the methylenes of the aliphatic chain of the monoglyceride. As these signals severely overlap with those of the methylenes of the triterpenes in the already crowded region, 2.00−0.80 ppm, it was impossible to determine the length of the fatty acid chain from the 1H NMR spectrum. However, it appears that the aliphatic chain is free of double bonds, as no olefinic signals were observed in the region 6.00−4.50 ppm. Finally, using the signal at δH 3.814 as the purest qHNMR signal, the content of the monoglyceride was determined to be 6.1 mol %. The COSY spectrum revealed two other impurities (imp2a/ b), with both having an allylic hydroxy group in the molecule (Figure 3B). These results were deduced from a correlation of two double-bond protons in the 5.70−5.30 ppm region. One of these two protons also correlates with a proton at 4.50−3.50 ppm, consistent with a −O−C−H partial structure, while the other proton correlates with two protons at ∼2.00 ppm, overall suggesting an allylic hydroxy group. Literature data12 imply that these partial structures belong to two polyyne-type compounds of the falcarindiol class, previously identified in a lipophilic extract of O. horridus. The total content of these two impurities in Fr-21 was determined to be 5.9 mol % by using their signals at δH 3.948 (imp2a) and 4.365 (imp2b) as quantifiers. In addition to imp1 and imp2, a minor amount of 2 (2.2 mol %) was identified in Fr-21 on the basis of its marker signal at δH 2.852. These findings exemplify that the RC of purified natural products can result from both structurally related and unrelated impurities. In our experience, despite repeated chromatographic purification, reference materials of biosynthetically diverse natural products such as triterpenes often exhibit surprisingly high degrees of RC.16 Regarding their bioactivity, the observed co-occurrence of considerably different chemical species is of crucial importance. Therefore, the evaluation of the degree and pattern of RC of purified natural products should be considered
Table 2. qHNMR Impurity Profiles and Anti-TB Activity of the O. horridus Triterpene Fractions against M. tuberculosis Strain H37Rv HPLC fractions
a
D
component
Fr-19
Fr-20
Fr-21
Fr-22
Fr-23
Fr-24
1 (mol %) 2 (mol %) imp 1 (mol %) imp 2 (mol %) MIC (μg/mL)a
94.9 0.0 5.1 0.0 220
96.3 1.2 2.5 0.0 118
85.8 2.2 6.1 5.9 121
79.9 4.7 7.7 7.7 125
70.8 28.0 0.6 0.6 124
29.3 69.4 1.3 0.0 226
The highest test concentration was 256 μg/mL. dx.doi.org/10.1021/np3007809 | J. Nat. Prod. XXXX, XXX, XXX−XXX
Journal of Natural Products
Article
Figure 4. Graphic representation of the QPARs of 1 in the O. horridus triterpene-enriched HPLC fractions. Panel A shows the relationship between the molar contents of the two triterpenes and the anti-TB activity of the fractions. Panel B illustrates the correlation between the impurities and the observed anti-TB activity.
exhibited a stronger anti-TB activity than 1 or that it was a potentiator of the activity of 1. In order to interpret inhibition of antimycobacterial activity, it is important to note that glycerol is the main carbon source usually employed in the culture of mycobacteria. A recent study has identified monoglyceride lipase Rv0183 from M. tuberculosis, suggesting that this lipolytic enzyme may be involved in the degradation of host cell lipids. Thus, monoglycerides can be hydrolyzed by Rv0183, releasing free glycerol and absorbable fatty acid, which are essential elements for the growth of M. tuberculosis.17,18 This could explain the observed antagonistic effect of monoglyceride impurities such as imp1 on the anti-TB activity of impure triterpenes. On the other hand, several reports have confirmed polyynes as relatively potent anti-TB compounds with MIC values of 20−60 μg/mL.12,19 Therefore, even minor polyyne impurities may exert significant enhancement of the anti-TB activity, which offers an explanation for the potential impact of imp1 and imp2 on the observed overall activity. Finally, it is noteworthy that the triterpenes 1 and 2 exist as an isomeric pair derived from betulinic acid (3) and oleanolic acid (4), respectively, by epimerization at C-3 and oxidation of Me-23 to a carboxylic acid group. The two precursors, 3 and 4, have previously been reported to be active against M. tuberculosis with MIC values of 62 and 29 μg/mL, respectively.20 However, the present QPAR results indicate that the anti-TB activity of 2 is actually weaker than that of 1, resulting in a higher MIC value for the combination of the two compounds relative to that of pure 1. Furthermore, considering our previous study,1 a direct comparison of reported MIC values with the present bioactive compounds would require PAR and/or purity information for 3 and 4. Taking into account our previous1 and present results, it is likely that previously observed anti-TB activity levels of 3 and 4 will require correction once they are studied in a QPAR model.
These data led to the establishment of QPARs for the anti-TB active fractions as visually illustrated in Figure 4. Clearly, curve shape and progression indicate that the activity of the fractions (MIC) and purity of the triterpene 1 were not proportional, nor were they correlated in a linear or logarithmic fashion. Therefore, the activity must be attributed to both the major components and the impurities, or solely to the impurities. Visual inspection of the graph in Figure 4 also indicated that the anti-TB activity of 1 appears to be correlated with the presence of monoglycerides. However, this inhibition was counteracted by the co-occurrence of polyyne analogues, representing additional minor impurities in the same triterpene fractions. In order to quantitatively evaluate the effects of the two impurities, imp1 and imp2, on the observed activity of fractions dominated by 1, the relationship between sample activity and purity was correlated in a mathematical model. For this study, two conditions were initially hypothesized: (1) the concentration and activity of any single chemical entity (SCE) component follows a linear relationship; (2) the combination of activities of all components in the sample is only additive. On the basis of these two conditions, linear regression analysis becomes an appropriate method for the establishment of QPARs. As a result, the anti-TB activity of 1 (MIC) can be expressed as a linear function of the molar content (M) of each of the three minor components (2, imp1, and imp2) as follows: MIC = 53.9 + 1.95 × M 2 + 30.9 × M imp1 − 22.2 × M imp2 (R2 = 0.964)
When extrapolating this function toward 100% purity, i.e., M2 = Mimp1 = Mimp2 = 0, 1 exhibits an MIC of 54 μg/mL. This means that the impurities have a negative impact on the observed activity of 1, leading to higher than expected MIC values for the six samples. Notably, the effect of each impurity on the overall activity can be assessed on the basis of the sign and value of coefficients of the corresponding independent variable in the model. Therefore, in the given example of fractions dominated by 1, the positive coefficients for 2 and imp1 indicate that these two components showed adverse effects on the overall activity. Conversely, imp2 has a negative coefficient, meaning that it
■
CONCLUSION Normally, when a substance is evaluated biologically, the assumption is made that the sample represents an SCE or a mixture with known composition. However, when bioactive materials require isolation from complex matrices, they are E
dx.doi.org/10.1021/np3007809 | J. Nat. Prod. XXXX, XXX, XXX−XXX
Journal of Natural Products
Article
Thus, the dt signal at δH 3.029 of the main component 1 was selected as internal reference with an arbitrary value of 100. The proportionality was then calculated on the basis of all detected impurities, which were normalized to 100% of the total sample. All NMR data processing and analysis was performed using the MestReNova 8.00-10524 software (Mestrelab Research, Santiago de Compostela, Spain). Chemical shifts (δ in ppm) were referenced to the residual methanol-d3 signal at δH at 3.310 and δC at 49.00. Plant Material. The dried fruits of O. horridus were collected from wild specimens of the plant in the vicinity of Anchorage, Alaska, in the winter of 2007, and authenticated by David C. Smith at Alaska Green Gold (Anchorage, AK, USA). Voucher specimens (BC #390) are deposited in the College of Pharmacy at UIC, Chicago. Extraction and Fractionation. The dried fruits of O. horridus (7 kg) were pulverized and percolated sequentially with petroleum ether and CH2Cl2 at room temperature. The crude organic extracts were concentrated in vacuo (
