BIOCHEMICAL AND BIOPHYSICAL RESEARCH COMMUNICATIONS ARTICLE NO.
227, 484–488 (1996)
1533
Triterpenes as Potential Dimerization Inhibitors of HIV-1 Protease Luc Que´re´, Traudl Wenger, and Hans J. Schramm1 Max-Planck-Institut fu¨r Biochemie, D-82152 Martinsried, Germany Received August 28, 1996 HIV-1 protease is a homodimeric enzyme. A b-sheet consisting of the four terminal segments provides the main driving force for dimerization of the per se inactive protomers. Several short peptides with sequences related to the terminal sequences of the protease are able to inhibit dimerization by blocking the ‘interface’ part of the monomers. From the structures of such inhibitory peptides a ‘pharmacophore’ could be derived. By using a prominent distance from this pharmacophore scaffold for a library search (Cambridge Structural Database), non-peptide inhibitors of HIV-1 protease with polycyclic triterpene structure could be found. The IC50 constants of these compounds are near 1 mM. One of the triterpenes, the ursolic acid (Ki Å 3.4 mM), was further kinetically analysed (according to Zhang). The shape of the graph confirms the expected mechanism of dimerization inhibition. q 1996 Academic Press, Inc.
The active protease of HIV-1 (PR) consists of two identical half-molecules which contribute both to the formation of only one active site. This structure allows the unusual way of enzyme inhibition by preventing dimer formation (1-3) by compounds which stabilize the per se inactive monomers. The X-ray structure of PR shows that this ‘interface’ part consists of the four N- and C-terminal segments of the two monomers which form an anti-parallel b-structure (4). Peptides with sequences similar to those of the natural termini of PR are able to block the interface part of a monomer, and in this way prevent dimerization. The computer modelled peptide-PR complexes show some prominent features (3): • The OH-group of the deeply inserted side chain of the ‘anchor’ residue Tyr97= (of the peptide) is hydrogen-bonded to the peptide backbone of the active site segment Asp25-ThrGly (of the monomer). Additional hydrogen bonds are possible. For instance, Tyr97= may be replaced by dihydroxyphenylalanine or similar residues. Elongations of the side chain 97* or OH-esterifications might allow additional binding interactions to Asp25 or Arg8. • The side chains of the charged amino acid 98* and the free terminal carboxy-group 99* of the inhibitor contribute to the binding energy by interaction with His69 and Pro1 of the monomer. • The hydrophobic side chain of amino acid 99 * points into a hydrophobic pocket nearly normal to the b-sheet axis and below the sheet. This extended pocket in the monomer offers space for larger residues in the case of modified peptides, e.g. those with biphenylalanine in 99*.
The structural consensus features of the docked peptides could be defined as a ‘pharmacophore’ structure (3) which allowed now the design of non-peptide inhibitors. This was done by library search using one of the pharmacophore distances and a bank of crystal structures (Cambridge Structural Database). By computer docking, it was possible to identify several triterpenes which fit well into the hydrophobic interface site of the relaxed monomer. 1
To whom correspondence should be addressed. Abbreviation: PR, HIV-1 protease. 484
0006-291X/96 $18.00 Copyright q 1996 by Academic Press, Inc. All rights of reproduction in any form reserved.
AID
BBRC 5517
/
690d$$$541
09-26-96 09:13:13
bbrcas
AP: BBRC
Vol. 227, No. 2, 1996
BIOCHEMICAL AND BIOPHYSICAL RESEARCH COMMUNICATIONS
MATERIALS AND METHODS Materials. All organic compounds were purchased from Aldrich, D-89555 Steinheim. PR was obtained by bacterial expression in E. coli (3) following the procedure of Billich (5). Enzyme inhibition assay. HIV-1 PR inhibition was measured by an assay using recombinant HIV-1 PR and the nonapeptide H-Lys-Ala-Arg-Val-Nle* (p-nitro-Phe)-Glu-Ala-Nle-NH2 as substrate (6). The spectrophotometric measurement was carried out as follows: the hydrolysis of the chromogenic substrate in 1 ml of assay buffer (pH 4.7) was monitored at room temperature in presence of various concentrations of inhibitors using a Perkin-Elmer spectrophotometer LAMBDA 17 at 295 nM for 300s (1 cm lightpath, quartz cuvette). The determination of Ki values followed the description of Zhang (2). Data search. For the pharmacophore query (for a recent example see ref. 7), the current version (5.11, release April 1996) of the Cambridge Structural Database CSD (8,9) was screened using the program QUEST (10). This databank consists of more than 152000 three-dimensional structures determined by X-ray diffraction. The computer search generated a list of best-fitting solutions. Among these compounds, we identified those which (a) were predicted to fit into the dimerization site and thereby act as dimerization inhibitors, (b) were commercially available for testing and (c) had a rigid nucleus to reduce the tendency for hydrophobic collapse and loss of entropy. Several triterpenes and steroids were selected by this screening mode and subject to the enzyme test. Modelling. The molecular mechanics minimization in this study was performed with the DISCOVER program (Biosym, San Diego, INSIGHT II package, version 95.0) using the CVFF forcefield, on MSI Silicon Graphics Indigo 2 workstations. The initial PR structure was the X-ray crystal form of HIV-1 PR solved by Erickson et al. (11) and deposited in the Brookhaven Data Bank as 9hvp. For further studies to analyse the docking fit of the inhibitors, the LEAPFROG program of the SYBYL package (Tripos, Inc., St. Louis, USA, version 6.2) was used. The INSIGHT II minimization protocol included an initial relaxation of the monomer form (500 steps of steepest descent, 500 steps of conjugate gradient and 500 steps of VA09A minimization). Selected triterpenes were subsequently docked into the putative dimerization site, and the complexes minimized in the same way.
RESULTS AND DISCUSSION
Using the distance range of 1.2 to 1.4 nm—for OH 97* to COOH 99* —as a first query length, the Cambridge Structural Database (CSD) was searched (9) using the program QUEST. As a starting ‘‘site’’, the relaxed PR monomer was used (12). As a first hit, the structure of cholic acid emerged with a satisfactory score. This compound showed a fair inhibition constant IC50 of about 350 mM. For such determinations, one has to avoid inhibitor concentrations above the critical micelle concentration. This was also necessary with the other triterpenes. Further database search concentrated now on the group of triterpenes from which many compounds are commercially available. The identified substances were purchased and evaluated in the PR test (Fig. 1). So far, all commercially available compounds of this type show inhibition in the test, although the pentacyclic triterpenes indicate a tighter binding to the PR than the more hydrophilic tetracyclic analogues of cholic acid. However, the cheaper members of the cholic acid group (e.g. lithocholic acid with only one -OH) may become attractive starting materials for modifications. The particularly low IC50 value of lithocholic acid—which lacks two hydroxy groups present in cholic acid—reveals the high hydrophobicity required for good ligands. Note, that uvaol which has a hydroxy group instead of the carboxylic group (at position 28) is nearly as potent as ursolic acid. This suggests that the formation of a salt bridge in this part of the inhibitor is not crucial. In this region, the main contribution to binding seems to come from hydrogen bonds. However, the hydophobic triterpene scaffold is obviously sufficient for binding in the low micromolar range. Modifications of the polycyclic hydrocarbon part using the program LEAPFROG confirmed this. For instance, the replacement of the CH3groups by the more lipophilic CF3-groups improved the calculated binding energies considerably, as did the fusion of one or two additional rings to the triterpene scaffold without further polar substitutions (data not shown). The addition of a carbohydrate moiety at OH 3 (ahederin) does not strongly interfer with activity (Fig 1). The PR inhibition by ursolic acid was further analysed (data not shown) using Zhang kinetics (2). This method includes the influence of the dimer-monomer equilibrium and yields 485
AID
BBRC 5517
/
690d$$$542
09-26-96 09:13:13
bbrcas
AP: BBRC
Vol. 227, No. 2, 1996
BIOCHEMICAL AND BIOPHYSICAL RESEARCH COMMUNICATIONS
FIG. 1. Chemical structures and HIV-1 protease inhibitory constants of several triterpenes.
completely different graphs for competitive (convergent lines) and dissociative inhibitors (parallel lines) of PR. The method is especially valuable since the fast renaturation of PR does not easily allow the measurement of retardation by peptides after denaturation and dilution (results not given). The validity and applicability of this kinetic method could be confirmed by the demonstration of synergism between inhibitory interface peptides and active site directed inhibitors (3). As assumed, a preliminary measurement shows that ursolic acid belongs to the group of dimerization inhibitors with a Kdim of 3.4 mM, and a slight competitive inhibitory component i Kicom of about 130 mM. Computer modelling shows (Fig.2) that this binding can be further improved not only by modification of the scaffold, but also by rather simple chemical modifications, e.g. the esterification of the OH group. By using dicarboxylic acids, amino acids (as Asp, Lys, Ala, b-Ala) and similar compounds as reactants, simple derivatives of the basic triterpenes can be synthesized in order to reach nanomolar inhibition. 486
AID
BBRC 5517
/
690d$$$542
09-26-96 09:13:13
bbrcas
AP: BBRC
Vol. 227, No. 2, 1996
BIOCHEMICAL AND BIOPHYSICAL RESEARCH COMMUNICATIONS
FIG. 2. Stereoscopic view of the complex of a relaxed HIV-1 protease monomer with ursolic acid (dashed lines). Hydrogen bonds are shown as dotted lines. The lower part of the relaxed structure represents the interface region with the two terminal segments. The main differences to the ‘native’ monomer structure are in this region.
More generally, the scaffold of the triterpene compounds which matches in its volume that of the backbone of a cyclic hexapeptide can be used efficiently as a cheap building unit in the design of peptidomimetics for several reasons (13): • • loss •
Many steroids are used as drugs and have good oral bioavailability, The rigid polycyclic nucleus should reduce the tendency for hydrophobic collapse and of entropy generally associated with peptide-protein interaction, Triterpenes offer multiple possibilities for side chain modifications.
In these experiments, the relaxed monomer structure of PR—which is still similar to the ‘native’ monomer, i.e. the half of the dimer—is used. It is assumed that it reflects the monomer structure in solution. However, it is not clear whether this notion holds. It is possible that the unrelaxed conformer arising directly from the PR dimers in the equilibrium may also be an attractive target structure. The redimerization of the monomer obviously needs this conformer. It is also not clear how stable the inhibitor-protease complex is: it may subsequently unfold or, because of the large hydrophobic patches, aggregate. The binding of active site directed peptidic inhibitors to PR depends on the sum of the single contributions of several side chains interactions. This multi-site mode of active site binding (8 subsites) allows the prediction of cleavage positions in peptides, but also complicates the design of potent mimetic inhibitors, especially since two side chains in neighbouring subsites (e.g. S1 and S3) interact to some extend and weaken the overall binding. In the case 487
AID
BBRC 5517
/
690d$$$542
09-26-96 09:13:13
bbrcas
AP: BBRC
Vol. 227, No. 2, 1996
BIOCHEMICAL AND BIOPHYSICAL RESEARCH COMMUNICATIONS
of the interface inhibition mode, a more simple site structure is available which should facilitate the computer aided development of inhibitors. A literature search showed that some of the triterpenes are already known as anti-HIV agents (14), and betulinylglycine (15) is known as a ‘weak’ PR inhibitor. This confirms our results and draws further attention to this interesting class of natural compounds with very low toxicity. In addition, betulinic acid derivatives are assumed to interact with the post-binding virus-cell fusion process (15). Therefore, it may be possible to design triterpene derivatives which concurrently target the protease and the fusion complex. Triterpene compounds abound in nature in amount and diversity (16-18). For example, the waxy skin of one single apple may contain more than 50 mg of ursolic acid (16). These riches of nature may be utilized for the isolation of cheap but potent agents (or of starting materials for simple chemical modifications), for the use in anti-HIV ‘cocktails’. ACKNOWLEDGMENTS We thank A. Boetzel and J. Bu¨ttner for valuable discussions.
REFERENCES 1. Schramm, H. J., Nakashima, H., Schramm, W., Wakayama, H., and Yamamoto, N. (1991) Biochem. Biophys. Res. Comm. 179, 847–851. 2. Zhang, Z.-Y., Poorman, R. A., Maggiora, L. L., Heinrikson, R. L., and Kezdy, F. J. (1991) J. Biol. Chem. 266, 15591–15594. 3. Schramm, H. J., Boetzel, J., Bu¨ttner, J., Fritsche, E., Go¨hring, W., Jaeger, E., Ko¨nig, S., Thumfart, O., Wenger, T., Nagel, N. E., and Schramm, W. (1996) Antiviral Res. 30, 155–170. 4. Wlodawer, A., and Erickson, J. (1993) Ann. Rev. Biochem. 62, 543–585. 5. Billich, A., Hammerschmidt, F., and Winkler, G. (1990) Biol. Chem. Hoppe-Seyler 371, 265–271. 6. Richards, A. D., Phylip, L. H., Farmerie, W. G., Scarborough, P. E., Alvarez, A., Dunn, B. M., Hirel, P.-H., Konvalinka, J., Strop, P., Pavlickova, L., Kostka, V., and Kay, J. (1990) J. Biol. Chem. 265, 7733–7736. 7. Wang, S., Milne, G. W. A., Yan, X., Posey, I. J., Nicklaus, M. C., Graham, L., and Rice, W. G. (1996) J. Med. Chem. 39, 2047–2054. 8. Kennard, O., Watson, D. G., Allen, F. H., Motherwell, W. D. S., Town, W. G., and Rodgers, J. (1975) Chem. Ber. 11, 213–216. 9. Allen, F. M., and Kennard, O. (1993) Chem. Design and Automat. News 8, 31–37. 10. CSD System Design, Programming and Documentation (October 1992). 11. Erickson, J., Neidhart, D. J., Van Drie, J., Kempf, D. J., Wang, X. C., Norbeck, D. W., Plattner, J. J., Rittenhouse, J. W., Turon, M., Widebrug, N., Kohlbrenner, W. E., Simmer, R., Helfrick, R., Paul, D. A., and Knigge, M. (1990) Science 249, 527–532. 12. Venable, R. M., Brooks, B. R., and Carson, F. W. (1993) Proteins: Struct., Funct. and Genet. 15, 374–384. 13. Hirschmann, R., Sprengler, P. A., Kawasaki, T., Leaky, J. W., Shakespeare, W. C., and Smith, A. B., III (1992) J. Am. Chem. Soc. 114, 9699–9701. 14. Okano, M., Fukamiya, N., Tagahara, K., Cosentino, M., Lee, M. T.-Y., Morris-Natschke, S., and Lee, K.-H. (1996) Bioorg. & Med. Chem. Lett. 6, 701–706. 15. Evers, M., Poujade, C., Soler, F., Ribeill, Y., Lelie`vre, Y., Gueguen, J.-C., Reisdorf, D., Morize, I., Pauwels, R., De Clercq, E., He´nin, Y., Bousseau, A., Mayaux, J.-F., Le Pecq, J.-B., and Dereu, N. (1996) J. Med. Chem. 39, 1056–1068. 16. Ying, Q.-L., Rinehart, A. R., Simon, S. R., and Cheronis, J. C. (1991) Biochem. J. 277, 521–526. 17. Karrer, W. (1976) Konstitution und Vorkommen der organischen Pflanzenstoffe, Birkha¨user Verlag, Basel und Stuttgart. 18. Devon, T. K., and Scott, A. I. (1975) Handbook of Naturally Occurring Compounds, Vol. II, Academic Press, New York/San Francisco/London.
488
AID
BBRC 5517
/
690d$$$542
09-26-96 09:13:13
bbrcas
AP: BBRC