QSARs on human carbonic anhydrase VA and VB inhibitors of some new not yet synthesized, substituted aromatic/heterocyclic sulphonamides as anti-obesity agent

Abstract

This paper presents result of quantitative structure–activity relationships (QSAR) study realized with the PRECLAV, omega, brood and MOPAC software. The dependent property is the inhibitory activity against human carbonic anhydrase mitochondrial isoforms VA and VB. The calibration set includes 17 aromatic/heterocyclic sulphonamides incorporating phenacetyl, pyridylacetyl and thienylacetyl tails with three clinically used CA inhibitors namely AZA, TPM and ZNS molecules. The prediction set contains 24 others not yet synthesized substituted aromatic/heterocyclic sulphonamides having unknown observed values of activity. In the presence of prediction set, the predictive quality of QSAR of hCA VA (r² = 0.9789, F = 418.115, r²_CV = 0.9689) and hCA VB (r² = 0.9768; F = 379.717; r²_CV = 0.9637) is large. The obtained models suggest a slightly different inhibition mechanism for the two isoforms. Large percentage, in weight, of CONH molecular fragments seems to be favourable to inhibitory activity of both VA and VB.

Keywords::

• QSAR
• omega
• brood
• PRECLAV
• carbonic anhydrase VA
• VB

Introduction

There are 16 α-carbonic anhydrase (CA, EC 4.2.1.1) isoforms expressed in mammals, CA I–CA XV, two of which, CA VA and CA VB, being present in mitochondria^1–5. These two isozymes are involved in several biosynthetic processes, such as ureagenesis, gluconeogenesis and lipogenesis^1,2,6–8. As hCA VA/VB are involved in several biosynthetic processes catalysed by pyruvate carboxylase, acetyl CoA carboxylase and carbamoyl phosphate synthetases I and II, providing the bicarbonate substrate to these carboxylating enzymes involved in fatty acid biosynthesis, their inhibition may lead to the development of anti-obesity agents possessing a new mechanism of action⁸. Inhibition data for classical sulphonamide CA inhibitors (CAIs) used clinically, such as AZA (acetazolamide), TPM (topiramate), ZNS (zonisamide) were obtained, together with inhibition data for aromatic/heterocyclic sulphonamides possessing varied structures, incorporating phenacetyl, pyridylacetyl and thienylacetyl tails act as potent inhibitors of human mitochondrial isoforms VA and VB. Such compounds may be useful for the development of novel anti-obesity therapies⁹.

The prediction set includes molecules having unknown observed values of dependent property. The quantitative structure–activity relationships (QSAR) studies can be made in absence or in presence of certain prediction set. In the absence of the prediction set, the purpose of QSAR studies is the identification of the molecular features with the highest impact (favourable or unfavourable) on the biochemical activity. In the presence of the prediction set, the purpose is to identify the prediction set molecules having the largest computed activity.

The search for new human mitochondrial isoforms V (VA and VB) inhibitors is important for medicinal chemistry. Therefore, the structures of the prediction set molecules were selected mainly by their possibility to be synthesized in laboratory conditions and taking into account the commercial availability of the raw materials.

The calibration set and the prediction set

Recently one of our groups⁸ reported for the first inhibition study against mitochondrial isoform hCA VA and VB with aromatic/heterocyclic sulphonamides incorporating phenyl(alkyl), halogenosubstituted-phenyl or 1,3,4-thiadiazole-sulphonamide moieties and thienylacetamido; phenacetamido and pyridinylacetamido tails. The aromatic/heterocyclic sulphonamides incorporating phenacetyl, pyridylacetyl and thienylacetyl tails with clinically used CA inhibitors such as AZA, TPM and ZNS (Table 1) were included in the calibration set. The inhibitory activity (as K_I values, in the nanomolar to micromolar range for isozymes) was expressed by means of the equation A = −log K_I.

Table 1.  Structural details of hCA VA and VB inhibitor and their observed and estimated inhibitory activities (−log KI).

Com. no.	See structure above	n	Y	X1	X2	Obs. VA	Est. VA	Obs. VB	Est. VB
1	A	0	H	–	–	−0.85733	−0.896	−0.8451	−0.932
2	A	0	F	–	–	−0.83885	−0.925	−0.89763	−0.889
3	A	0	Cl	–	–	−0.88649	−0.905	−0.9345	−0.894
4	A	0	Br	–	–	−0.87506	−0.86	−0.93952	−0.928
5	A	1	H	–	–	−0.95904	−0.93	−0.85733	−0.958
6	A	2	H	–	–	−1.0086	−0.961	−0.90309	−0.931
7	B	–		–	–	−0.88081	−0.954	−0.86923	−1.151
8	C	0	H	CH	CH	−0.82607	−0.851	−0.89763	−0.802
9	C	1	H	CH	CH	−0.9345	−0.884	−0.91908	−0.889
10	C	2	H	CH	CH	−0.90849	−0.915	−0.91381	−0.852
11	C	0	F	CH	CH	−0.91908	−0.879	−0.90849	−0.787
12	C	0	Cl	CH	CH	−0.91381	−0.86	−0.90849	−0.806
13	C	0	Br	CH	CH	−0.85126	−0.829	−0.77085	−0.765
14	C	0	H	N	CH	−0.83251	−0.855	−0.83251	−0.885
15	C	0	H	CH	N	−0.86332	−0.855	−0.95424	−0.897
16	C	2	H	CH	N	−0.88081	−0.918	−0.96848	−0.922
17	D	–	–	CH	CH	−0.92428	−0.936	−0.78533	−0.964
18	AZA	–	–	–	–	−1.79934	−1.823	−1.73239	−1.609
19	ZNS	–	–	–	–	−1.30103	−1.23	−3.78053	−3.761
20	TPM	–	–	–	–	−1.79934	−1.793	−1.47712	−1.5

The prediction set contains 24 other not yet synthesized substituted aromatic/heterocyclic sulphonamides generated by Brood¹⁰ software (version 2.0.0, open eye science software, Santa Fe, USA), having unknown observed values of activity (Table 2). Brood uses the shape and attachment geometry of the query fragment to identify a family of similar fragments.

Table 2.  The chemical structure of prediction set molecules not yet synthesized having unknown observed values of activity.

Methods and formulas

The minimum energy geometry, for each molecule in the calibration and prediction set, was obtained by the conformational search ability of the Omega v.2.4.3^11–13 (OpenEye Science Software, Santa Fe, USA) program. Isomeric SMILES notation was used as program input in order to avoid any influences on conformational model generation by presenting 3D seed structures. Omega employs a rule-based algorithm^12,13 in combination with variants of the Merck molecular force field 94. For the generation of conformers, following parameters were used: a maximum of 200 conformers per compound, an energy cut-off of 10 kcal/mol relative to global minimum identified from the search. The force field used was the 94s variant of the MMMF_NoEstat (Merck molecular force field^11–13) that includes all MMFF terms except coulomb interactions. The RMSD fit value 2.0 Å was used to avoid redundant conformers.

The conformations of minimum energy obtained by molecular mechanics calculations were further minimized by quantum chemical calculations. The semi-empirical PM6 method¹⁴ included in the MOPAC 2009 software¹⁵ (Stewart Computational Chemistry, Colorado Springs, CO) optimized the geometry more rigorously. In MOPAC analysis, we used the following sequence of keywords: “PM6 Pulay gnorm = 0.01 shift = 50 geo-ok campking mmok bonds vectors”.

In the next step, the MOPAC and PRECLAV software (Center of Organic Chemistry, Bucharest^16,17) produced more than 500 “whole molecule” (global) descriptors including the value of some weighted functions and virtual fragmentation descriptors for each molecule. Set of descriptors includes parabolic functions of whole molecule descriptors, calculated by PRECLAV. The statistical calculations used for obtaining the QSAR equations were done with PRECLAV as reported earlier^17–25.

Using only the “significant” descriptors¹⁸ PRECLAV computes thousands of QSAR equations, i.e. multi-linear formulas of the dependent property.

1

Here “A” represents a dependent property (here the inhibitory activity defined above) and “k” is the number of descriptors in the set. Ordinary Least Square Method computes weighting factors C_k of predictors D_k. The PRECLAV program does not compute errors related to regression coefficients. The “quality” of each QSAR was computed using usual statistical formulas that are a measure of agreement of observed/computed values of activity: standard error of estimation Se, Pearson square correlation r², Fisher function F and cross-validated Pearson square correlation r²_CV.The concordance between the calculated/observed values has been calculated using the quality function Q^16,19 which possesses values in the interval {−1, 1}.

2

where r²_CV is cross-validated (Leave one out method) Pearson square linear correlation between computed/observed values and N is the number of molecules in the calibration set (here N = 20). By increasing the number of descriptors k, the quality Q of the equations increases, reaching a maximum, and then decreases. For predictions, the equation with the highest quality was used, the descriptors present in this equation being called “predictors”.

The relative Utility (U) of a certain predictor on dependent property values was computed by the Specific procedure^16,20. The predictors which present a high value for U, within the range [0, 1000], may be considered very useful in estimating the activity, because they correlate very well with activity and do not correlate with other predictors. Each “useful” predictor offers ample information about the variation in activity from molecule to molecule.

After computing the A_calc values of the inhibitory activity for the prediction set molecules, PRECLAV arranged these molecules according to the estimated values. It computed the average value A_calc^m for the estimated values and standard deviation σ of the estimated values. The program considers “high values” as the values fulfilling the criterion (3) and “low values” as the values fulfilling the criterion (4). Here, the molecules having “high” computed value of inhibitory activity have been taken as “recommended for synthesis²¹”.

3 4

PRECLAV divides the analyzed molecules into virtual fragments, using an algorithm reported earlier^26,27. The virtual fragments identified by PRECLAV do not always coincide with the classical functional groups. The presence of a significant fragment in the molecule greatly influences (in a positive of negative way) the inhibitory activity of the molecule.

Results and discussions

QSAR#1 (human mitochondrial isomers of hCA VA inhibitors)

In absence of prediction set, the number of “significant” descriptors is 232 and we obtained type (1) QSAR equation where

C₀ = −1.6194;

C₁ = 0.0163;

D₁ is (psa) percentage of single conjugated and aromatic bonds¹⁶ (U = 1000);

C₂ = −0.0164; and

D₂ is (war) weight percentage of largest molecular fragment¹⁶ (U = 994).

Whereas the quality of correlation is described by the statistical indices:

Se = 0.0422, r² = 0.9789, F = 418.1153, r²_CV = 0.9689, Q = 0.8721.

The quality of above QSAR is high. There are no outliers in calibration set. The minimum correlation of predictors with inhibitory activity is computed for D₂ (r² = 0.3535). The inter-correlation between predictor is very low (r² = 0.0296). Therefore, D₁ predictor offers a different kind of information from the D₂ predictor. In this study, molecules of analyzed database include 33 virtual fragments but only seven virtual flagments are considered significant. The percentages, in weight, of molecular fragments are well correlated (directly or inversely) with the values of inhibitory activity: CONH (r = 0.6851), C₂HN₃S₂ (r = −0.645), CH (r = −0.645), C atom (r = −0.645), O atom (r = −0.645) and NH₂ (r = −0.5298).Because of the structure of the computed QSAR and the result of virtual fragmentation, we think:

• the presence of substituted CONH groups (compound no. 1–6 and 8–16) is favourable to activity;

• the presence of C₂HN₃S₂ (compound no. 18), CH (compound no.20), C atom (compound no. 20) and O atom (compound no. 20) fragments is unfavourable to the activity;

• the presence of aromatic and conjugated bonds is favourable to inhibitory activity; and

• the large molecular fragments are not significant to inhibitory activity. This conforms to the molecular fragment analysis of calibration set compounds.

In the presence of the prediction set, the number of “significant” descriptors is only 188. We obtained the same QSAR equation and results (r² = 0.9789, F = 418.1153, r²_cv = 0.9689) in presence of the prediction set molecules not yet synthesized substituted aromatic/heterocyclic sulphonamides. Therefore, the calibration set is quite “representative sample” in calibration set + prediction set group. Using this equation the maximum activity computed for calibration set molecules is −0.829, the average activity computed for calibration set molecules is −1.003 ± 0.288 and the average activity computed for prediction set molecules is −0.825 ± 0.067.

In Table 3, the calculated values identified by the program as “high” have been marked in bold letters, while the values identified as “low” have been underlined. According to criterion (3), this equation identified eight molecules in prediction set having “high values” of activity. The large number of molecules identified as “high active” in prediction set is, probably, the statistical effect of small gap between maximum and minimum observed value of activity in calibration set. However, in the Table 3 “most active” 75% molecules include chlorine atoms and CH₃ group and 25% include OH group. So the favourable effect of chlorine atoms and CH₃ group on activity is obvious.

Table 3.  Calculated values of hCA VA and VB inhibitory activity for the molecules in the prediction set.

S. no.	Comp. no. (VA)	Calculated VA	Comp. no. (VB)	Calculated VB
21.	40.	−0.704	25.	−0.752
22.	36.	−0.725	26.	−0.757
23.	34.	−0.727	30.	−0.777
24.	42.	−0.728	37.	−0.783
25.	41.	−0.73	32.	−0.784
26.	44.	−0.776	31.	−0.795
27.	32.	−0.779	35.	−0.797
28.	21.	−0.785	34.	−0.812
29.	38.	−0.803	36.	−0.835
30.	31.	−0.808	40.	−0.845
31.	35.	−0.811	33	−0.855
32.	22.	−0.819	39	−0.873
33.	43.	−0.841	38	0.876
34.	37.	−0.842	22	−0.914
35.	33.	−0.842	27	−0.918
36.	23.	−0.867	23	−0.93
37.	25.	−0.867	41	−0.931
38.	24.	−0.88	24	−1.037
39.	28.	−0.89	44	−1.066
40.	29.	0.893	21	−1.069
41.	27.	−0.9	28	−1.105
42.	26.	−0.9	29	−1.181
43.	39.	−0.919	43	−1.255
44.	30.	−0.922	42	−1.264

QSAR# 2 (human mitochondrial isomers of hCA VB inhibitors)

In absence of prediction set, the number of “significant” descriptors is 222 and we obtained type (1) QSAR equation where

C₀ = −4.3766;

C₁ = 6.7015;

D₁ is (XSC) maximum net charge of C atoms at parabolic region¹⁶ (U = 1000);

C₂ = −10.0927; and

D₂ is (avc) average free valence of C¹⁶ (U = 896).

Whereas the quality of correlation is described by the statistical indices:

Se = 0.102, r² = 0.9768, F = 379.717, r²cv = 0.9637, Q = 0.8673.

The quality of above QSAR is also high. There are no outliers in calibration set. The minimum correlation of predictors with inhibitory activity is computed for D₂ (r² = 0.056). The inter-correlation between descriptors is very low (r² = 0.0303). In this study, molecules of analysed database include 33 virtual fragments but only three virtual fragments CONH (r = 0.5455), C₇H₄NO (r = −0.9396) and NH₂ (r = −0.788) are considered significant. Because of the structure of the computed QSAR#2 and the result of virtual fragmentation, we think:

• C₇H₄NO (compound no. 19) and NH₂ (compound nos. 1–20) fragments are the unfavourable to the activity;

• the presence of substituted CONH (compound nos. 1–6 and 8–16) groups is favourable to activity;

• the avc is inversely proportional to the activity, if the free valance carbon is substituted by radical like Cl and OH increases the inhibitory activity; and

• the maximum net charge of C atoms is favourable to the inhibitory activity.

In the presence of the prediction set, the number of “significant” descriptors is only 184. We obtained the same QSAR equation and results (r² = 0.9768 0; F = 379.717; r²_CV = 0.9637) in the presence of the prediction set molecules not yet synthesized substituted aromatic/heterocyclic sulphonamides. Therefore, in this case also the calibration set is quite “representative sample” in calibration set + prediction set group. Using the above QSAR equation, the maximum activity computed for calibration set molecules is −0.765, the average activity computed for calibration set molecules is −1.105 ± 0.662 and the average activity computed for prediction set molecules is −0.926 ± 0.154.

According to criterion (3), this equation identified 10 molecules in prediction set having “high values” of inhibitory activity for hCA VB. The Table 3 “most active” 70% molecules include halogen atoms and 30% molecule includes CH₃ and OH group. Favourable effect of halogen atoms in activity is obvious in human mitochondrial isomers of hCA VB inhibitors.

Conclusions

• Large percentage, in weight, of CONH molecular fragments seems to be favourable for the inhibitory activity of hCA VA and hCA VB.

• Large percentage, in weight, of CH₃, C₂HN₃S₂, CH, C atom, and O atom molecular fragments is unfavourable to hCA VA inhibitory activity and C₇H₄NO and NH₂ molecular fragments are unfavourable to hCA VB inhibitory activity.

• Aromatic and single conjugated bond are favourable for hCA VA inhibitory activity.

• Maximum free valance of carbon atom is favourable to the hCA VB inhibitory activity.

• The positive correlation of maximum net charge of C atoms plays dominating role of the modelling of hCA VB inhibitory activity.

• The “representative sample” feature of calibration set-in calibration set + prediction set group has a large influence on predictive power of computed QSAR. The most active ten molecules in VB and eight molecules in VA prediction set include halogen (chlorine atoms) group.

Acknowledgement

The authors are thankful to OpenEye Scientific Software, Santa Fe, USA and Center of Organic Chemistry-Romanian Academy, Bucharest (Dr. Tarko L.) for providing software for the QSAR study.

Declaration of interest

The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.