2.2. Qualitative Structure-Activity Relationship Study
The chemical structures of all 19 cyclopeptide alkaloids are shown in Figure 1
. Based on these structures and the results obtained for the antiplasmodial activity, a preliminary structure-activity relationship study was performed. From the 19 cyclopeptide alkaloids tested, eight had an IC50
value below 10 μM. Out of these eight, five belonged to the 13-membered and three to the 14-membered cyclopeptide alkaloids. Interestingly, the five most promising results were found for cyclopeptides bearing a 13-membered ring, with IC50
values ranging between 2.1 and 7.1 μM. From these results, it could be concluded that a meta-cyclophane structure (13-membered ring) is more likely to be associated with a high antiplasmodial activity compared to a para-cyclophane structure (14-membered ring). Moreover, out of the six tested 13-membered cyclopeptide alkaloids, only one showed an IC50
value above 10 μM, namely jubanine-F (6
value of 12.8 μM), but this result still indicates moderate activity. Thus, all tested meta-cyclophane compounds were found to be antiplasmodially active. As for the 14-membered cyclopeptide alkaloids, only three of the 13 tested compounds were found to have an IC50
value below 10 μM, while nummularine-E (4
) could not inhibit P. falciparum
in concentrations up to 64.0 μM. These findings support our hypothesis that a 13-membered ring is favorable regarding the high antiplasmodial activity.
When looking further into the compounds containing a 13-membered macrocyclic ring, two contain a styrylamine moiety, hydroxylated in position 2 of the aromatic ring, while the other four possess a 2-methoxy styrylamine moiety. With respect to this structural feature, Suksamrarn et al. [4
], who assessed the antiplasmodial activity of four 13-membered cyclopeptide alkaloids, proposed that the methoxy group would be crucial for displaying antiplasmodial activity. Our results do not support this, since the tested hydroxylated meta-cyclophane compounds were also found to be active, but interestingly, compound 15
value of 3.2 μM) seemed to be more active than compound 16
value of 7.1 μM), with their only difference being the presence of a hydroxy or a methoxy group, respectively. Based on this result, it might be concluded that cyclopeptide alkaloids containing either a methoxy group or a hydroxy group can display antiplasmodial activity, but that methoxylated compounds might be associated with higher activity. In view of this, it would be interesting to determine, for example, the antiplasmodial activity of paliurine-C, which is the methoxylated analogue of spinanine-B. Moreover, it would be interesting to test the 13-membered cyclopeptide alkaloids without a methoxy or hydroxy group, in order to know how this affects their antiplasmodial activity. However, to the best of our knowledge, such cyclopeptide alkaloids have not been reported up to now [8
], although they could possibly be obtained by semi-synthesis.
One common feature of all tested meta-cyclophane compounds is the presence of a β-hydroxy proline amino acid moiety. Panseeta et al. [5
] suggested that this feature could be linked to a relatively high antiplasmodial activity. They found promising results for the 13-membered cyclopeptide alkaloids mauritine-M and nummularine-H, and for hemsine-A (17
), the latter belonging to the para-cyclophane compounds. In our case, all para-cyclophane compounds bearing a β-hydroxylated proline showed only moderate or weak activity against P. falciparum
, except for amphibine-D (5
value of 8.9 μM). Moreover, the 14-membered cyclopeptide alkaloids comprised of a β-hydroxy phenylalanine (oxyphylline-F) (19
) or a β-hydroxy leucine (adouetine-X) (9
) showed promising antiplasmodial activities. These results indicate that a β-hydroxy proline moiety is not crucial for displaying antiplasmodial activity. In order to confirm whether this finding is applicable to the 13-membered cyclopeptide alkaloids as well, it would be valuable to know to what extent meta-cyclophane cyclopeptide alkaloids with different β-hydroxy amino acid moieties are capable of inhibiting the growth of the Plasmodium
parasite. However, to the best of our knowledge, all reported 13-membered cyclopeptide alkaloids contain such a β-hydroxy proline moiety; thus, for the moment, it is not possible to determine the activity of other types of meta-cyclophane cyclopeptide alkaloids.
Another interesting finding is the difference in activity between adouetine-X (9
value of 7.5 μM) and frangulanine (10
value of 14.9 μM). These two compounds show many structural similarities, and differ only in two aspects: adouetine-X contains an isoleucine moiety as a ring-bound amino acid, while frangulanine contains a leucine moiety in this position. The opposite is applicable for the side chain, with a leucine unit found in adouetine-X and an isoleucine moiety found in frangulanine. Leucine and isoleucine are constitutional isomers, hence, a small difference in chemical structure can affect the antiplasmodial activity. Suksamrarn et al. [4
] concluded that substitution of a leucine unit with a valine unit, both aliphatic amino acid units as well, resulted in similar biological activities. In their case, the substitution took place in the side chain (ziziphine-N vs. ziziphine-Q). The cyclopeptide alkaloids jubanines-F (6
) and -G (7
) also differ only in one amino acid unit in the side chain, with jubanine-F containing a valine moiety and jubanine-G an isoleucine moiety. However, the antiplasmodial activity found for jubanine-G (IC50
value of 4.7 μM) was significantly higher than the activity for jubanine-F (IC50
value of 12.8 μM). Thus, the statement made by Suksamrarn et al. [4
] is not supported by our results, and it must be concluded that the substitution of one aliphatic amino acid with another does affect the resulting antiplasmodial activity.
2.3. Quantitative Structure-Activity Relationship Study
In order to have a quantitative understanding of the structure-activity relationships underlying some of the compounds’ selectivity with respect to antiplasmodial activity, a QSAR (quantitative structure-activity relationship) study was carried out. PLS (partial least squares regression) and MLR (multiple linear regression) models were built, linking a set of molecular descriptors with the antiplasmodial activity. Different data pretreatments, i.e., autoscaling, direct orthogonal signal correction (DOSC) [9
], and their combinations were first applied to the descriptor data. DOSC was found to result in the best PLS model. The MLR models were built by selecting the descriptors based on their importance in the best DOSC-PLS model. Three models, with three, six, or eight descriptors were built, and revealed which molecular descriptors were of main importance for the antiplasmodial activity. The model with eight descriptors performed similarly well as the best PLS model.
The RMSECV (root mean squared error of cross validation) and RMSEC (root mean squared error of calibration) values of each model were calculated, in order to select the best model. For the best DOSC-PLS model, the RMSECV was 0.0134 (and the RMSEC was 0.0051). The more descriptors included in the MLR models, the lower the RMSECV value found. For the model built with three descriptors, the RMSECV was 0.4306 (RMSEC was 0.2150); for that with six descriptors it was 0.0725 (0.0505); and when eight descriptors were included it was 0.0257 (0.0095), respectively. However, the four models predicted the activity of the calibration samples well, even when they were considered as unknown compounds in the cross validation. For instance, the following MLR equation (eight-descriptor model) can be applied in order to have an estimation of the IC50
value obtained in the antiplasmodial activity assay:
The eight molecular descriptors that were identified as most influential (in sequence added in the three MLR models) were: the Platt index, a path-based topological descriptor which is defined as the sum of the edge degrees of a molecular graph, and two geometrical descriptors, MMFF94 (Merck molecular force field 94) energy and the Dreiding energy (both given in kcal/mol), which are measures for the internal energy of a given conformer. In addition, two other geometrical descriptors, i.e., the maximal and minimal projection area (given in Å2); furthermore, the Wiener polarity, which represents the number of three bond length distances in the molecule (topological descriptor); the percentage of C-atoms in the molecule, and finally the aliphatic bond count. For all descriptors in the model, a higher value is associated with a higher IC50 value, and thus lower antiplasmodial activity. Unfortunately, the theoretical descriptors cannot always easily be linked to the physicochemical properties of the molecule.
Additionally, the models were used to predict the activity of a number of potentially interesting structures, as indicated by the qualitative SAR study. We stated earlier that it would be interesting to determine the antiplasmodial activity of paliurine-C, the methoxylated analogue of spinanine-B, as well as of the 13-membered cyclopeptide alkaloids without a methoxy or hydroxy group. The models built predict that both paliurine-C and 13-membered cyclopeptide alkaloids without a methoxy group would have rather high activities. Moreover, the importance of the β-hydroxy proline moiety in 13-membered cyclopeptide alkaloids was questioned, and it would be valuable to know to what extent meta-cyclophane cyclopeptide alkaloids with different β-hydroxy amino acid moieties are capable of inhibiting P. falciparum. Thus, a prediction of the activity of 13-membered cyclopeptide alkaloids bearing either a β-hydroxy leucine, β-hydroxy valine, or a β-hydroxy phenylalanine moiety was made, although such compounds have not been identified from natural sources yet. The theoretical descriptors derived from their structures allowed the prediction of a high activity for these compounds. In addition, the models seem to confirm the high activity for the 13-membered-ring compounds. When interpreting the predictions, it has to be taken into account that the calibration set was small (19 compounds), and that the structures for which the activities were predicted may be sub-optimally represented in it. Moreover, the data set mainly consisted of active and intermediately active compounds, while inactive compounds are underrepresented. Therefore, in our opinion, the interpretation of the activity at this moment, should or could not go further than the indication of whether or not a given compound is expected to be rather active or inactive. Better predictive models for these types of compounds will require a larger, more representative calibration set.
The chemical structures of promising cyclopeptide alkaloids defined in the qualitative discussion of this paper allowed the prediction of high activities using any of the QSAR models. The QSAR approach thus did not contradict the expectations gained from the qualitative study. In the future, promising compounds could be isolated from natural sources or constructed (semi-) synthetically, after which testing of their antiplasmodial activity can be performed.