Anti-Human Immunodeficiency Virus Activities of Nucleosides and Nucleotides: Correlation with Molecular Electrostatic Potential Data

Abstract

Examination of the anti-human immunodeficiency virus (HIV) data of some normal and isomeric dideoxynucleosides (ddNs and isoddNs), their three-dimensional (3-D) electron density patterns, their electrostatic potential surfaces (EPS), and their conformational maps reveals some interesting correlations. For example, the EPS of (S,S)-isoddA shows regions of high and low electrostatic potential remarkably similar to those of β-d-3′-azido-3′-deoxythymidine (β-d-AZT), (−)-oxetanocin A, and (−)-carbovir. Such correlations involving EPS data and anti-HIV activity were also found with many other active nucleosides. Conversely, inactive compounds had EPS different from those of compounds in the same series that were active. For example, apio-ddNs, which are inactive against HIV, exhibit clear differences in electrostatic potential and 3-D electron density shape from isoddNs that are active against HIV. Additionally, the inactivity of (S,S)-isoddC and (S,S)-isoddT can be correlated convincingly with a combination of their EPS data and their conformational energy maps. The electrostatic potential distributions of active nucleoside triphosphates show remarkable correlations. For example, (S,S)-isoddATP, AZT triphosphate (AZTTP), and oxetanocin A TP have similar 3-D electron density surface patterns and similar high and low regions of electrostatic potential, which may suggest that these compounds proceed through related mechanisms in their interactions with, and inhibition of, HIV reverse transcriptase (RT). Docking of AZTTP, (S,S)-isoddATP, and other active triphosphates into the active site of HIV RT and calculation of the EPS of both the nucleotide and the active site show that there is excellent matching between inhibitor and enzyme binding site EPS data. The structure-activity profile discovered has contributed to the development of a first predictive quantitative structure-activity relationship analysis in the area.

Correlation of anti-human immunodeficiency virus (HIV) activity with specific structural characteristics of nucleoside inhibitors is of considerable significance in contributing to the understanding of the bioactivation and mechanism of action of these compounds. Additionally, such investigations may provide predictive information on the structural characteristics most likely to elicit activity. In the case of nucleosides active against HIV, quantitative structure-activity relationship (QSAR) correlations have not been proposed. An approach with limited success is the correlation of activity with conformation (i.e., glycosidic torsional angle, orientation of the 5′-OH, and sugar ring puckering) (1, 4, 25, 38). For anti-HIV-active nucleosides targeted at HIV reverse transcriptase (RT), phosphorylation to the triphosphate is a requirement for inhibition. However, molecular recognition studies with the phosphorylating kinases are complicated by the fact that the crystal structures of some of these enzymes have not been determined. However, examination of electrostatic potential surfaces (EPS) has suggested that if molecules have similar electrostatic potentials, along with some other key properties such as conformation, hydrophobicity, and hydrogen-bonding sites, then it is more probable that their molecular recognition by enzymes and receptors would be similar (2, 8, 28, 33). Thus, by using electrostatic potential, fundamental structural properties can be examined for qualitative and quantitative correlations. Areas of high, neutral, and low electrostatic potential can be determined for active nucleosides and nucleotides. The binding sites within relevant enzymes would then be expected to have opposite areas of electrostatic potential.

We have recently investigated the synthesis and anti-HIV activity of isomeric dideoxynucleosides (isoddNs), i.e., dideoxynucleosides (ddNs) with transposed base moieties or transposed -CH₂OH groups (30). One of these isomeric nucleosides, (S,S)-isodideoxyadenosine [(S,S)-isoddA], has potent activity against HIV type 1 (HIV-1) and HIV-2 (31). Its triphosphate is one of the most powerful inhibitors of HIV RT known (K_i = 16 nM) (31). The enantiomer of this compound also shows some activity against HIV (14, 15). In comparison, none of the other isomeric nucleosides of the (S,S), (R,R), or apio families were found to have significant anti-HIV activity (30). The antiviral data for these compounds provided a unique opportunity to investigate the usefulness of electrostatic potential and conformational analysis for the correlation of the anti-HIV activities of normal ddNs and isoddNs.

To facilitate this study, a method for the examination of these active nucleoside derivatives was investigated. This method involved small-molecule manipulation with respect to electron densities and electrostatic potentials. The analysis led to some interesting qualitative relationships between the nucleoside derivatives. One such relationship is the existence of similar electrostatic potential distributions for 3′-azido-3′-deoxythymidine (AZT), (S,S)-isoddA, (−)-carbovir, and oxetanocin A, all of which are active against HIV (27). This relationship was strengthened further by the examination of related but inactive compounds that were found to have different EPS patterns. Extensive development of this initial qualitative correlation led to more relationships among the active nucleoside derivatives and also to the development of the first predictive QSAR analysis in this area. In addition, electrostatic potential data were utilized to produce a model of the active site of HIV-1 RT docked with a variety of competitive triphosphate inhibitors. This paper describes the first comprehensive application of EPS data in the anti-HIV field.

(A preliminary account of some aspects of this work, has appeared previously [27]).

MATERIALS AND METHODS

Molecular modeling of the nucleoside derivatives was carried out on a Silicon Graphics IRIS 4000/5000 system using the SYBYL program (version 6.4; Tripos Associates, St. Louis, Mo.). The structures were first fully minimized to their lowest energy by the Powell method, and charges were considered by using the Gasteiger-Huckel charge calculation. The Gasteiger-Huckel charge calculation can give more accurate results in nucleosides where hydrogen bonding between 5′-OH and O-2 can take place. To construct the EPS, the partial charges were calculated using the Gasteiger-Marsili method. Then the electron density was calculated by using a resolution of 7, followed by the mapping of electrostatic potential on the surface of the electron density. Surfaces were then compared by using their fully minimized structures. For the conformational analysis, a total of 1,369 conformations (37²; 0° and 360° were confirmed by recalculation) were generated and their energies were fully minimized by using GRID search. Both γ and χ were defined and searched through a full 360° range, with 10° increments. Each conformation gave an energy value, and these values were graphed in terms of γ and χ by using the TABLE and GRAPH options in SYBYL. For QSAR results, the torsion angles γ and χ of each molecule in each group were set to those of the representative molecule for each set, e.g., β-d-ddC for the 2′, 3′-dideoxycytidine (ddC) group and (−)-2′,3′-dideoxy-3′-thiacytidine [(−)-3TC] for the oxythiolane group. Then, by using the ALIGN DATABASE command, each group was aligned to give the best overlap of carbon atoms. Realignments and recalculations were necessary in some instances due to poor structural overlap. With a good overlap, a molecular spread sheet (MSS) was created and the comparative molecular field analysis (CoMFA; Tripos Inc.) for each molecule was calculated. Then a partial least-squares (PLS) analysis was performed using the log 1/IC₅₀ (where IC₅₀ is the 50% inhibitory concentration) and the CoMFA data. Bioactivities were calculated by using the VIEW CoMFA command in the QSAR dialog box.

To calculate the electrostatic potential for the active site of HIV RT, the X-ray crystal structure for HIV RT was modified from the structure originally obtained (13, 21) to include Mg²⁺ ions and the appropriate ddN triphosphate. Also, the active site was extracted from the X-ray structure to minimize unnecessary calculations. Charges were then calculated by using the Gasteiger-Marsili method. Connolly surfaces were generated in the Molcad module of SYBYL with a probe radius of 1.4 Å and a dot density of 5.0 points/area to reduce the calculation time. Docking was performed by using the dock feature, and the energy was minimized.

RESULTS AND DISCUSSION

Qualitative structure-activity relationship: EPS maps of nucleosides.

While the pathway for the mechanism of action of anti-HIV nucleosides is complex, we have developed a method to “group” active and inactive molecules by their EPS. This method involves examining the pattern of high (positive, red-orange), neutral, and low (negative, purple-blue) electrostatic potentials located on the surfaces of the electron densities for a correlation. Molecules with similar patterns of EPS can be grouped together and compared with structurally related molecules that are active or inactive. All of the molecules examined and their anti-HIV activities are shown in Table 1 and Fig. 1.

TABLE 1.

Compound numbers, structures,^a and anti-HIV activities of compounds investigated

No.	Templatea	R1	R2	R3	R4	X	Y	Base	Nameb	Activityc	Cell line	Source or reference
1	A							A	(S,S)-IsoddA	0.67	PBM	31
2								A	(R,R)-IsoddA	43	MT-4	14
3	F	Me-						T	β-d-AZT	0.004	PBM	11
4								A	(R,S)-Apio-ddA	>200	MT-4	30
5								A	(S,R)-Apio-ddA	>200	MT-4	30
6	A							G	(S,S)-IsoddG	>200	MT-4	30
7	A							C	(S,S)-IsoddC	>200	MT-4	30
8	A							T	(S,S)-IsoddT	>200	MT-4	30
9								G	(−)-Carbovir	0.19	MT-2	39
10								A	(−)-Oxetanocin A	4	MT-4	36
11	B	H-	H-	H-	H-			C	β-d-ddC	0.046	MT-4	29
12	B	H-	H-	F-	H-			C	5-F-β-d-ddC	0.2	MT-4	16
13	B	H-	H-	Et-	H-			C	5-Et-β-d-ddC	0.49	PBM	NIAID workshopd
14	B	H-	H-	H-	F-			C	3′-F-β-d-ddC	8	ATH8	12
15	D	H-				O	S	C	(−)-3TC	0.24	CEM	9
16	C	H-				S	O	C	(+)-dOTC	0.9	MT-4	24
17	C	F-				S	O	5-FC	(+)-dOTFC	3	MT-4	24
18	D	H-				S	O	C	(−)-dOTC	2.8	MT-4	24
19	D	F-				S	O	5-FC	(−)-dOTFC	3.2	MT-4	24
20	D	F-				O	S	5-FC	(−)-FTC	0.3	MT-2	35
21								G	(−)-DXG	0.03	PBM	17
22	E							A	β-l-d4A	0.38	PBM	5
23	E							5-FC	β-l-d4FC	0.09	MT-2	23
24	E							Hypox	β-l-d4I	5.5	PBM	5
25	E							G	β-l-d4G	14.1	PBM	5
26								5-FC	β-l-FddC	2.3	MT-4	10
27	B	H-	H-	F-	N3-			5-FC	3′-N3-5-F-β-d-ddC	1	PBMC	22
28	B	H-	H-	H-	N3-			C	3′-N3-β-d-ddC	7.6	MT-4	3
29	B	H-	Me-	Me-	N3-			diMeC	3′-N3-N4,5-diMe-β-d-ddC	17.3	MT-4	11
30	B	H-	H-	Cl-	F-			5-CIC	3′-F-5-Cl-β-d-ddC	26	MT-4	37
31								C	α-l-Dioxolane C	0.06	PBM	18
32	F	H-						U	β-d-AZU	0.36	MT-4	6

^aSee Fig. 1. 

^bCompounds are referred to by their most commonly used names. 

^cExpressed as the IC₅₀, in micromolar concentrations. 

^dBaker et al., Workshop on Nucleosides in HIV Chemotherapy, 1988. 

FIG. 1.

Structures of the compounds investigated.

Examination of the electrostatic potential of (S,S)-isoddA (compound 1) and various other ddNs shows some remarkable correlations. For example, comparison of (S,S)-isoddA (compound 1) and (R,S)- or β-d-AZT (compound 3) shows that both molecules, in addition to having similar overall three-dimensional (3-D) electron density shapes in their most stable conformations, also exhibit similar regions of high (red-orange, positive) and low (purple-blue, negative) electrostatic potential (Fig. 2). For purposes of comparison and correlation, we have examined only the key regions on the surface map and have allowed for small deviations resulting from rotational freedom about the glycosidic bond.

This initial observation that two nucleosides that are active against HIV may show similar EPS prompted us to analyze some active and inactive compounds. As expected, (S,S)-isoddA (compound 1) and its less active enantiomer, (R,R)-isoddA (compound 2), exhibited different 3-D electron density shapes and different electrostatic potentials (Fig. 3). Neither the electrostatic potential distribution nor the conformational map of (R,R)-isoddA correlated with that of AZT. Comparisons between (S,S)-isoddA and the isoddNs that were inactive against HIV were also of significance. For example, the anti-HIV-inactive d- and l-related apio-series of isoddA (30) exhibited different electron density shapes and different electrostatic potentials from (S,S)-isoddA (Fig. 4).

Differences in EPS were also apparent within the (S,S)-isoddN group, which was consistent with the anti-HIV data that showed that none of the compounds with bases other than adenine had activity (30). Interestingly, a detailed comparison of (S,S)-isoddA (compound 1) with (S,S)-isoddC (compound 7) and (S,S)-isoddT (compound 8) showed that they all had similar 3-D shapes and similar electrostatic potentials with the exception of the base region. Conformational data (see below) aided further in providing a possible explanation for the inactivity of (S,S)-isoddC and (S,S)-isoddT. Interestingly, the anti-HIV-active carbocyclic nucleoside (−)-carbovir (compound 9) (39), exhibited an electrostatic potential distribution similar to that of (S,S)-isoddA (Fig. 5). The anti-HIV-active (−)-oxetanocin A (compound 10) (36) also had an electrostatic potential map similar to those of (S,S)-isoddA, AZT, and (−)-carbovir (Fig. 6).

Other active molecules have also been examined. Several different patterns have emerged which show remarkable correlations within their respective groupings. One such group includes the anti-HIV-active molecule β-d-ddC (compound 11) and several of its structurally related derivatives, including 5-fluoro-β-d-ddC (5-F-β-d-ddC) (compound 12) and 5-ethyl-β-d-ddC (compound 13) (7, 16, 29; D. Baker, C. K. Chu, and K. Agrawal, presented at the Workshop on Nucleosides in HIV Chemotherapy, Developmental Therapeutics Branch AIDS Program, National Institute of Allergy and Infectious Diseases, 1988). All of these molecules show the same pattern of high and low electrostatic potentials (Fig. 7). Also, other active, structurally related molecules such as 3′αF-β-d-ddC (compound 14) (3, 11, 12, 22, 37) show similar patterns with only minor differences. However, inactive derivatives, such as 3′βF-β-d-ddC and 2′αF-β-d-ddC (26), exhibit completely different patterns.

Another group of active molecules that was investigated was related to (−)-3TC (compound 15) and the active oxathiolane compounds (9, 24, 35). The molecules (+)-2′-deoxy-3′-oxo-4′-thiocytidine [(+)-dOTC] (compound 16) and (+)-dOTFC (compound 17) had closely related electrostatic patterns along with their mirror images (−)-dOTC (compound 18) and (−)-dOTFC (compound 19) (Fig. 8). Also, (−)-2′,3′-dideoxy-5-fluoro-3′-thiacytidine [(−)-FTC] (compound 20) and (−)-3TC (compound 15) showed the same pattern of electrostatic potential within their set and had the same general pattern as their counterparts, the oxathiolane compounds. Additionally, the active compound (−)-β-dioxolane G [(−)-DXG] (compound 21) (17) showed EPS data similar to those of (−)-FTC (compound 20) and (−)-3TC (compound 15). Differences were found between the electrostatic potentials of the potently anti-HIV active compound (−)-3TC and its less active enantiomer, (+)-3TC (Fig. 9).

The third series of active molecules examined comprised the l-isomers of the d4N series (compounds 22 to 25) and the β-l-isomer of 5FddC (compound 26) (5, 10, 23). Remarkably, all molecules investigated, regardless of their base moiety, had the same pattern of high and low electrostatic potentials (Fig. 10).

Correlations with conformational maps of nucleosides.

In cases where compelling correlations, particularly for inactivity, were not available from the EPS data of nucleosides, we resorted to supporting conformational explanations. Examination of the conformational maps of the anti-HIV-inactive (S,S)-isoddC (compound 7) and the anti-HIV-active β-d-ddC (compound 11) demonstrated that they are clearly different (Fig. 11 and 12). Among the major differences are the relatively larger number of less favorable conformations in (S,S)-isoddC versus β-d-ddC and their corresponding locations. Also, (S,S)-isoddC shows, within a local minimum region, several conformations with relatively small distances from 5′-OH to O-2, which would suggest the presence of hydrogen bonding (34) (Fig. 13; Table 2). Finally, the anti-HIV-inactive (R,S)- and (S,R)-apio-isoddAs (compounds 4 and 5) demonstrated little correlation with (S,S)-isoddA (compound 1) with respect to conformational energy maps, in contrast to the similarity of these maps for (S,S)-isoddA and other anti-HIV-active ddNs such as β-d-ddC (compound 11).

TABLE 2.

Hydrogen bonding stabilization data

Compound	Energy before (kcal/mol)	Energy after (kcal/mol)	ΔE (kcal/mol)	Angle of 5′-OH, and O-2 (°)	H-bond distance (Å)
(S,S)-IsoddC	11.000	7.429	3.571	148.56	1.787
β-d-ddC	15.282	13.809	1.473	150.86	1.773

Correlations of EPS data of nucleoside triphosphate inhibitors and their binding to the active site of HIV RT.

It should be carefully pointed out that the correlations described for nucleosides do not give any information on the efficiency with which these compounds are phosphorylated by kinases to their triphosphates, the actual cellular inhibitors of HIV RT. Thus, a compound may have excellent EPS correlations with other anti-HIV nucleosides but may not have significant anti-HIV activity because of poor phosphorylation by kinases. For this reason, we also examined the electron density and EPS data of both active and inactive nucleoside triphosphates. The correlations were remarkable. We discovered that AZTTP and (S,S)-isoddATP have similar 3-D electron density surface patterns and similar high and low regions of electrostatic potential distribution (Fig. 14). The triphosphate of the anti-HIV-active carbocyclic nucleoside, (−)-carbovir (compound 9), also exhibited an electrostatic potential distribution similar to that of (S,S)-isoddATP. Figure 14 shows the similarity of the EPS of the triphosphates of oxetanocin A and (S,S)-isoddA. Correlations were also found with a number of other active triphosphates such as β-d-ddCTP and 5F-β-d-ddCTP. However, there was a clear difference between the EPS data for the anti-HIV-active compound β-d-ddATP and the data for its inactive 2-trifluoromethyl derivative (Fig. 14).

To further examine the binding relationship between the active site of HIV-1 RT (32) and triphosphate inhibitors, an EPS of the active site was then created. For Fig. 15, AZTTP was docked into the active site to show the interactions that occur with Mg²⁺ ions, the amino acid residues Tyr115, Asp110, Asp185, and Asp186, and the growing viral DNA chain. As can be seen, each residue of importance in the active site has an electrostatic potential that matches (is opposite in electrostatic potential) with the corresponding region in the triphosphate. For Fig. 16, (S,S)-isoddATP was docked into the active site of HIV RT and the electrostatic potential was calculated. The EPS matching between inhibitor and active site is clearly apparent. Other anti-HIV-active nucleoside triphosphates were also investigated, with similar results.

QSARs.

To further strengthen the nucleoside groupings that have been highlighted thus far, QSARs were prepared for two representative sets, d-ddC and 3TC. In this case QSAR analysis was attempted in order to further illustrate the relationship between electrostatic potential and anti-HIV activity. A QSAR set with all examined molecules was attempted and found to have no defined relationship. With this in mind, the groupings were developed to perform the analysis. To accomplish this task, molecules were structurally aligned, an MSS was created, and the CoMFA of each molecule was calculated. CoMFA is a quantitative calculation of the electrostatic potential and steric interactions of each molecule. The use of CoMFA in QSAR is well documented (19, 20).

In these analyses, only CoMFA and the log 1/IC₅₀ values were subjected to PLS analysis to ascertain the relationship between each active nucleoside derivative and its corresponding electronic structure. Due to the methods used for testing active nucleosides, more than one cell line and one set of published data were used to construct the QSAR analysis. This allowed confirmation of a statistical relationship between structure and activity but not the determination of specific activities of unknown sets. However, the analysis leads to a qualitative prediction of possible modifications that enhance activity. The groupings of molecules used and the data collected (cross-validated q², r², and F values) are summarized in Table 3, and the residual values for each molecule are shown in Table 4. The QSAR for the ddC series demonstrates the complexity of generating an accurate QSAR with predictive abilities. First, the cross-validated q² value is slightly lower than most q² values for other systems, but the r² and the F values indicate that there is a clear relationship being produced. Graphical representations of each QSAR are shown in Fig. 17. In the oxathiolane series of molecules, the q² was slightly lower than in the ddC series, but the other data suggest a more defined relationship.

TABLE 3.

Data collected from CoMFA

Series	Optimum no. of comparisons	q2	r2	F
d-ddC	2	0.413	0.874	17.367
3TC	5	0.403	0.993	29.426

TABLE 4.

Residual^a values

Series	Compound no.	Residual
d-ddC	11	0.254
	12	−0.548
	13	−0.085
	14	0.573
	27	0.128
	28	0.088
	29	−0.032
	30	−0.378
3TC	15	0.001
	16	0.07
	17	−0.068
	18	−0.081
	19	0.085
	20	−0.004
	21	−0.003

^aA residual is the negative logarithm of the difference between the actual and the predicted value. 

The data collected for the relative contributions of steric interactions (0.601 for d-ddC and 0.395 for 3TC) and electrostatic potential (0.399 for d-ddC and 0.605 for 3TC) also show that the qualitative relationships that were examined previously are adequate for examining electrostatic potential except perhaps in the case of the d-ddC series. In this case, an additional analysis of the respective volumes may be needed for a more accurate qualitative indicator of activity. The other series show that electrostatic potential is a slightly more important factor in activity than steric interactions with regard to a PLS analysis of CoMFA data for each series.

With the ability to generate a correlation among the active nucleosides, the next step was to use this information to generate a predictive profile for the enhancement of antiviral activity. One method for accomplishing this is to use the QSAR data to generate a bioactivity representation of the electron density and EPS. Each series could then be examined for possible modifications, and a log 1/IC₅₀ could be predicted.

In this study, two representative series were examined, and the predictions of changes to enhance their anti-HIV activities are shown in Fig. 18. Each qualitative representation has different colors to show the possible modifications that could improve activity. Steric bulk is represented by yellow and green. Yellow is used to show a region where a decrease in steric bulk would improve activity, while green represents a region where an increase in steric bulk would lead to improved activity. Electrostatic potential is represented by red and blue. Red shows a region where a modification in the electronic structure to a more negative species would improve activity, while blue shows a region where a more positive charge would improve activity. Ideally, modifications in a number of areas would be required to lead to a highly active compound for each series. A summary of the results of the QSAR studies is shown in Table 5.

TABLE 5.

Results of QSAR studies

Compound no.	Name	Activitya	Qualitative grouping	Confirmed with QSARb
1	(S,S)-IsoddA	0.67	(S,S)-IsoddA	Yes
2	(R,R)-IsoddA	43	None	NA
3	β-d-AZT	0.004	(S,S)-IsoddA	No
4	(R,S)-Apio-ddA	>200	None	NA
5	(S,R)-Apio-ddA	>200	None	NA
6	(S,S)-IsoddG	>200	None	NA
7	(S,S)-IsoddC	>200	None	NA
8	(S,S)-IsoddT	>200	None	NA
9	(−)-Carbovir	0.19	(S,S)-IsoddA	No
10	(−)-Oxetanocin A	4	(S,S)-IsoddA	Yes
11	β-d-ddC	0.046	β-d-ddC	Yes
12	5-F-β-d-ddC	0.2	β-d-ddC	Yes
13	5-Et-β-d-ddC	0.49	β-d-ddC	Yes
14	3′-F-β-d-ddC	8	β-d-ddC	Yes
15	(−)-3TC	0.24	(−)-3TC	Yes
16	(+)-dOTC	0.9	(−)-3TC	Yes
17	(+)-dOTFC	3	(−)-3TC	Yes
18	(−)-dOTC	2.8	(−)-3TC	Yes
19	(−)-dOTFC	3.2	(−)-3TC	Yes
20	(−)-FTC	0.3	(−)-3TC	Yes
21	(−)-DXG	0.03	(−)-3TC	Yes
22	β-l-d4A	0.38	l-5FddC	Yes
23	β-l-d4FC	0.09	l-5FddC	Yes
24	β-d-L-d4I	5.5	l-5FddC	Yes
25	β-l-d4G	14.1	l-5FddC	Yes
26	β-l-FddC	2.3	l-5FddC	Yes
27	3′-N3-5-F-β-d-ddC	1	β-d-ddC	Yes
28	3′-N3-β-d-ddC	7.6	β-d-ddC	Yes
29	3′-N3-N4-5-diMe-β-d-ddC	17.3	β-d-ddC	Yes
30	3′-F-5-Cl-β-d-ddC	26	β-d-ddC	Yes
31	αl-dioxolane C	0.06	(S,S)-IsoddA	Yes
32	β-d-AZU	0.36	(S,S)-IsoddA	Yes

^aExpressed as IC₅₀, in micromolar concentrations. 

^bNA, not applicable; inactive compounds which were not included in QSAR studies. 

In conclusion, examination of the anti-HIV data of some normal ddNs and isoddNs, their 3-D electron density patterns, their EPS, and their conformational data reveals some interesting correlations. The EPS of (S,S)-isoddA shows regions of high and low electrostatic potential remarkably similar to those of β-d-AZT, (−)-oxetanocin A, and (−)-carbovir. Correlations involving EPS data and anti-HIV activity were also found with many other active nucleosides such as those exhibiting EPS similarities with β-d-ddC, (−)-3TC, or l-d4A. Conversely, inactive compounds had EPS different from those of compounds in the same series that were active. For example, the anti-HIV-inactive apio-ddNs exhibit clear differences in electrostatic potential, 3-D electron density shape, and conformational maps from to the anti-HIV-active isoddNs. The electrostatic potential distributions of active nucleoside triphosphates show remarkable correlations. For example, (S,S)-isoddATP, AZTTP, and oxetanocin A triphosphate have similar 3-D electron density surface patterns and similar high and low regions of electrostatic potential, which may suggest that these compounds proceed through related mechanisms in their interactions with, and inhibition of, HIV RT. Docking of AZTTP, (S,S)-isoddATP, and other nucleoside triphosphates into the active site of HIV RT and calculation of the EPS of both the nucleotide and the active site show that there is excellent matching between the EPS data for the inhibitor and that for the enzyme binding site. The structure-activity profile discovered has significant ramifications and has contributed to the development of a first predictive QSAR analysis in this area. This is also the first comprehensive application of EPS data in the anti-HIV field.

FIG. 2-10.

FIG.2. EPS of (S,S)-isoddA (top) and AZT (bottom). FIG. 3. EPS of (S,S)-isoddA (top) and (R,R)-isoddA (bottom). FIG. 4. EPS of (S,S)-isoddA (top) and (R,S)-apio-isoddA (bottom). FIG. 5. EPS of (S,S)-isoddA (top) and (−)-carbovir (bottom). FIG. 6. EPS of (S,S)-isoddA (top) and oxetanocin A (bottom). FIG. 7. Electrostatic potentials of ddC (top, both panels), 5-F-ddC (bottom, left), and 5-Et-ddC (bottom, right). FIG. 8. Electrostatic potentials of (+)-dOTC (top) and (−)-dOTC (bottom). FIG. 9. Electrostatic potentials of (−)-3TC (top) and (+)-3TC (bottom). FIG. 10. Electrostatic potentials of l-d4A (top left), l-d4FC (top right), l-d4G (bottom left), and l-FddC (bottom right).

FIG. 11.

Fig. 11. Conformational maps of β-d-ddC (left) and (S,S)-isoddC (right). Fig. 12. Right-hand views of conformational maps for β-d-ddC (left) and (S,S)-isoddC (right). Energies range from 4 to 22 kcal/mol for the β-d-ddC graph and from 6 to 18 kcal/mol for the (S,S)-isoddC graph. Fig. 13. Minimized structure of (S,S)-isoddC showing hydrogen bonding between the 5′-OH and O-2 of the cytosine moiety. Fig. 14. EPS of (S,S)-isoddATP and AZTTP (far left), (S,S)-isoddATP and oxetanocin triphosphate (middle left), ddCTP and 5FddCTP (middle right), and ddATP and 2CF₃ddATP (far right).

FIG. 15.

Fig. 15. Active-site model of HIV RT showing AZTTP binding within the active site. Electrostatic potentials are shown for both the ligand and the active site. Magnesium ions are behind the triphosphate group. Fig. 16. Model of the active site of HIV RT with (S,S)-isoddATP docked inside. Fig. 17. QSAR graphs showing the relationship between the actual and the predicted biological activity (log 1/IC₅₀). The ddC series (left) and the 3TC series (right) are shown. Fig. 18. Bioactivities calculated from the QSAR analysis of each series' CoMFA. Each series is shown with a molecule docked inside the bioactivity measurement. In the d-ddC series (right), β-d-ddC is shown; in the 3TC series (left), (−)-3TC is shown.