J. Med. Chem. 2002, 45, 1563-1566
1563
β-C-Mannosides as Selectin Inhibitors Neelu Kaila,* Lihren Chen, Bert E. Thomas IV, Desiree Tsao, Steve Tam, Patricia W. Bedard, Raymond T. Camphausen, Juan C. Alvarez, and Giliyar Ullas Departments of Chemical Sciences, Biological Chemistry, and Immunology & Hemostasis, Wyeth Research, 200 Cambridge Park Drive, Cambridge, Massachusetts 02140 Received August 14, 2001 Abstract: Potential E- and P-selectin inhibitors were synthesized to explore a hydrophobic area on the E-selectin surface and the PSGL-1 protein binding site on the P-selectin surface that was recently defined by crystallography. Three series of mannose-based compounds (libraries A, B, and C) were synthesized using solution phase parallel synthesis. Biological evaluation of these compounds was done using two ELISA-based assays and transferred NOE (trNOE) experiments. Some of the compounds showed better activity than sLex in the P-selectin assay.
The tetrasaccharide sialyl Lewisx (sLex, 1) has been identified as the carbohydrate epitope recognized by E-, P-, and L-selectins on the leukocyte surface. During an inflammatory response, the initial rolling of leukocytes is mediated by the carbohydrate-protein recognition between sLex and E- or P-selectin. The excess recruitment of leukocytes can contribute to several acute diseases, such as stroke and reperfusion injury, and chronic diseases, such as psoriasis and rheumatoid arthritis. In recent times the development of a small molecule inhibitor that blocks the interaction of the selectins with sLex has become an attractive therapeutic approach.1-9 Previously, we described our success in accessing a hydrophobic patch on the E-selectin surface7 with phenyl substituents at the C-6 position of β-C-mannosides (2). As a result of our continuing efforts to develop sLex mimetics, we were fortunate to obtain the crystal structure of PSGL-1 (P-selectin glycoprotein ligand 1) bound to P-selectin.10 PSGL-1, the natural ligand for P-selectin, is a complex homodimeric glycoprotein containing the carbohydrate subunit, sLex. It was apparent that the binding site for the protein portion of PSGL-1 is on the opposite side of the sLex binding site relative to the proposed binding site for the C-6 phenyl substituents described previously (Figure 1). We observed that the PSGL-1 protein-binding site was proximal to the proposed location of the glutamic acid side chain in our selectin inhibitors (2). This observation prompted the exploration of this new binding site. This was accomplished by derivatizing the C-1 of the β-mannose core (library A) and C-R′ of the β-mannosyl glutamate (library B) using solution phase parallel synthesis. Reagents for each library were chosen using molecular modeling, as described below. * Corresponding author. Address: Wyeth Research, Department of Chemical Sciences, 200 Cambridge Park Drive, Cambridge, MA 02140. Tel: (617) 665-5626. Fax: (617) 665-5682. E-mail: nkaila@ war.wyeth.com.
Starting with our model of mannose bound to Eselectin, a model for mannose bound to P-selectin was generated in a manner described previously.7 Using these two models, virtual libraries were generated in the presence of E- and P-selectin using QXP.11 Each initial structure was minimized, followed by a Monte Carlo simulation of up to 1000 steps in which the ligand was allowed to rotate, translate, and flex torsions while the protein was held rigid. The 3- and 4- hydroxyls of the mannose core of each construct were constrained to be proximal to the selectin-bound calcium through the use of zero-order bonds. The low energy structure of each construct was retained and ranked using the QXP scoring function. The top 100 structures from the E-selectin and P-selectin simulations were visually screened, and the best 50 structures from the combined lists were retained for synthesis in each library. These amines were used in libraries A and B, respectively. To further explore the hydrophobic area proposed in our last paper,7 a virtual library that corresponds to derivatization of C-6 of β-mannosyl glutamate with hydrophobic groups was generated in the presence of Eand P-selectin using QXP. A similar process to the one described above was used to pick 50 acids for library C. Accordingly, several C-1 substituted β-mannosides (library A) and C-R′ and C-6 substituted β-mannosyl glutamate derivatives (libraries B and C) were prepared. Library A was synthesized by the route shown in Scheme 1. We decided to protect the four hydroxy groups as diacetonide so that we could deprotect the final products in a high throughput manner.12,13 Methyl2-(2,3,4,6-tetra-O-benzyl-β-D-mannopyranosyl) acetate 3 was prepared from 2,3,4,6-tetra-O-benzyl-D-mannopyranose in two steps via a known procedure.14 The benzyl groups in glycoside 3 were removed using Pd/C. The tetraol was then protected as a diacetonide (4) using dimethoxy propane. Hydrolysis of the methyl ester yielded 2-(2,3,4,6-diisopropylidene-β-D-manno-pyranosyl) acetic acid 5, which on coupling with methyl 4-(4aminophenyl)butyrate and further deprotection gave the final acid 6 in 60% yield.15 The optimal conditions for this step were used to couple acid 5 with a variety of amines. The resulting amides were deprotected to give library A (40 compounds).13 The purity of the final products was checked by LC/MS. In cases where the products were less than 95% pure they were purified using the PREP HPLC. The synthetic approach used to prepare C-R′ amide library B is outlined in Scheme 1. Coupling of acid 5 with H-L-Glu(OBut)-OMe followed by deprotection of the
10.1021/jm010390f CCC: $22.00 © 2002 American Chemical Society Published on Web 03/14/2002
1564
Journal of Medicinal Chemistry, 2002, Vol. 45, No. 8
Letters
Scheme 1
Figure 1. Crystal structure of PSGL-1 bound to P-selectin is shown. The modeled conformation of 2 bound to P-selectin is overlaid onto the crystal structure. The surface of P-selectin is represented by a Connelly surface that is colored by lipophilicity. The location of the bound calcium is indicated. The sLex-modified O-glycan residues of PSGL-1 are colored yellow, and the peptide portion is colored magenta. Compound 2 is colored in red.
R′-ester using potassium carbonate yielded acid 7. Acid 7 was coupled with L-aspargine tert-butyl ester and deprotected to yield amide 8 in 50% yield.15 The optimal conditions for this step were used to prepare library B (42 compounds).13 The purity of the final products was checked by LC/MS. In cases where the products were less than 95% pure, they were purified using PREP HPLC. The C-6 substituted derivatives of β-mannosyl glutamate (library C) were prepared by the procedure shown in Scheme 2. The C-4, C-6-acetonide in mannoside 4 was selectively removed by treatment with 70% acetic acid at room temperature.16 The primary alcohol was then converted to the corresponding azide 9. Hydrolysis of the methyl ester in 9 and coupling with H-LGlu(OBut)-OBut yielded amide 10. The azide group was then reduced to give amine 11, which on further coupling with N-(o-tolyl)succinamic acid and subsequent treatment with TFA at -20 °C gave the amide 12 in 70% yield.15 The common intermediate 11 was prepared in large quantities for use in the synthesis of library C (25 compounds).13,17 The purity of the final products was checked by LC/MS. In cases where the products were less than 90% pure they were purified using PREP HPLC. All of the compounds were evaluated in two ELISAbased assays. The E-selectin assay measures the binding of E-selectin to human recombinant AGP (R-1 acid glycoprotein) containing five sLex per molecule. Compounds showing inhibition of 30% or greater in the E-selectin assay are shown in Table 1. These compounds have IC50s around 5 mM. All of the active compounds belong to library C. Compounds from libraries A and B were inactive in the E-selectin ELISA-based assay. The P-selectin assay evaluates the binding of a PSGL-1 fusion protein to immobilized soluble P-selectin, which is preincubated with a test compound. The PSGL-1 fusion protein consists of the functional Nterminal region of PSGL-1 fused to the Fc region of IgG1. Compounds showing inhibition of 30% or greater
Scheme 2
in the P-selectin assay are shown in Table 2. The most potent compound, 13, was 3-fold more active than sLex. All but one of the active molecules came from library C. The other active compound came from library B. Compounds from library A were inactive in the Pselectin assay. The SAR from the three libraries is not very apparent. Compounds that lack a carboxylic acid group (some compounds in library A) were inactive in both the Eand P-selectin assays at 5 mM. Most compounds containing a hydrophobic group at C-6 showed some activity at 5 mM in both the assays. Those with greater than 30% inhibition at 5 mM are shown in Tables 1 and 2. The C-6 hydrophobic groups are believed to bind to a shallow hydrophobic pocket formed by the alkyl portion
Letters
Journal of Medicinal Chemistry, 2002, Vol. 45, No. 8 1565
Table 1. β-C-Mannosides in the E-Selectin Assaya
P-selectin and E-selectin. The trNOE experiment is a well-established technique to study protein-ligand interactions, including the determination of bound ligand structures.18 In the trNOE experiment, the protein-ligand system must be in fast exchange, and the bound small molecule is easily detectable by the observation of strong negative trNOE’s. Those are differentiated from the compounds that do not bind to the protein, which display a weak positive NOE. Table 2 summarizes the results of the different β-C-mannosides binding to P-selectin. With our standard ligand sLex, which shows an IC50 of 8 mM, we observe strong trNOE signals, corroborating previously observed data.10,18a Three compounds, 19, 20, and 21, which had moderate activity in the ELISA assay (∼30-37% inhibition at 5 mM), also showed trNOE cross-peaks, confirming that they bind to P-selectin. The NOESY spectra of these compounds in the absence of protein did not show negative NOE signals. Interestingly, compounds 13 and 18, which displayed higher or comparable inhibition in the ELISA, did not show binding by NMR, as no trNOE signals in the presence of P-selectin were observed. One possibility is that these compounds bind so weakly to P-selectin that they are below the threshold of the trNOE experiment, but the ELISA assay is still able to detect them. Alternatively, the consequence of the natural ligand binding weakly to the receptor necessitates the use of a multimeric presentation of the receptor or the ligand in the assay. The multimeric presentation is variable from one assay to another, which results in larger errors than typical biological assays. Because of the variability of the assay it would not be unexpected in some instances that the ELISA assay gives false positives. The results of these experiments demonstrate the difficulty in designing compounds that bind to a relatively flat surface. The crystal structure of PSGL-1 bound to P-selectin shows that the sLex binding site is flat and generally hydrophilic,10 particularly when compared to a typical drug target. While it is encouraging that a few compounds presented in this paper have a greater affinity for P-selectin than does sLex, they are still several orders of magnitude weaker than PSGL-1. The trNOE experiments provided a valuable tool to confirm that compounds, which show activity in the biological assay, actually bind to their intended target in the activity range indicated by a highly variable assay. Future plans are being directed toward exploring new scaffolds and the binding site of the peptide portion of PSGL-1.
a
These compounds did not show binding in the NMR assay.
Table 2. β-C-Mannosides in the P-Selectin Assays
of the side chains of Ala77, Glu80, Asp100, and Asn105, along with the phenyl group of Tyr94. While hydrophilic groups surround the top of the pocket, the pocket is made up of adjacent hydrophobic entities. The compounds in libraries A, B, and C were designed to explore this hydrophobic pocket as well as the PSGL-1 proteinbinding site. The fact that we did not find very active compounds may have resulted from the promiscuous nature of the selectin surface. In addition to the ELISA-based assays, we also performed transferred NOE (trNOE) experiments by NMR to confirm the binding of several compounds to
Acknowledgment. We thank Daniel Ricca and Dennis Lee at PPD Discovery for experimental assistance. Supporting Information Available: Flowchart for preparation of library A, experimental details for E-and P-selectin ELISAs, and details about transferred NOE (trNOE) experiments by NMR. This material is available free of charge via the Internet at http://pubs.acs.org.
References (1) Ebnet, K.; Vestweber, D. Molecular mechanisms that control leukocyte extravasation: the selectins and the chemokines. Histochem. Cell Biol. 1999, 112, 1-23. (2) Sears, P.; Wong, C.-H. Carbohydrate Mimetics: A New Strategy for Tackling the Problem of Carbohydrate-Mediated Biological Recognition. Angew. Chem., Int. Ed. 1999, 38, 2300-2324.
1566
Journal of Medicinal Chemistry, 2002, Vol. 45, No. 8
(3) Lasky, L. A. Selectin-Carbohydrate Interactions and the Initiation of the Inflammatory Response. Annu. Rev. Biochem. 1995, 64, 113-139. (4) Bertozzi, C. R. Cracking the carbohydrate code for selectin recognition. Chem. Biol. 1995, 2, 703-708. (5) Giannis, A. The sialyl Lewis-X group and its analogues as ligands for selectins: chemo-enzymic syntheses and biological functions. Angew. Chem., Int. Ed. Engl. 1994, 33, 178-180. (6) Springer, T. A. Traffic Signals for Lymphocyte Recirculation and Leukocyte Emigration: The Mutistep Paradigm. Cell 1994, 76, 301-304. (7) For recent examples of small molecules as selectin inhibitors, see refs 7-9. Kaila, N.; Thomas, B. E., IV; Thakker, P.; Alvarez, J. A.; Camphausen, T. R.; Crommie, D. Design and Synthesis of Sialyl Lewis x Mimics as E-Selectin Inhibitors. Bioorg. Med. Chem Lett. 2001, 11, 151-155 and references therein. (8) Simanek, E. E.; McGarvey, G. J.; Jablonowski, J. A.; Wong, C.H. Selectin-Carbohydrate Interactions: From Natural Ligands to Designed Mimics. Chem. Rev. 1998, 98, 833-862 and references therein. (9) Kaila, N.; Thomas IV, B. E. Design and Synthesis of Sialyl Lewis x Mimics as E- and P-Selectin Inhibitors. Med. Res. Rev. Manuscript submitted. (10) Somers, W. S.; Tang, J.; Shaw, G. D.; Camphausen, R. T. Insights into the Molecular Basis of Leukocyte Tethering and Rolling Revealed by Structures of P- and E-Selectin Bound to sLex and PSGL-1. Cell 2000, 103, 467-479. (11) McMartin, C.; Bohacek, R. S. QXP: powerful, rapid computer algorithms for structure-based drug design. J. Comput.-Aided Mol. Des. 1997, 11, 333-344.
Letters (12) Removal of the benzyl groups is labor intensive and requires one by one workup of each reaction. The diacetonide group can be removed in a combinatorial fashion by addition of acetic acid/ water, stirring, and savant. (13) As an example, the flowchart used for the preparation of library A is included in the Supporting Information. (14) Allevi, P.; Ciuffreda, P.; Colombo, D. The Wittig-Horner reaction on 2,3,4,6-tetra-O-benzyl-D-mannopyranose and 2,3,4,6-tetra-Obenzyl-D-glucopyranose. J. Chem. Soc., Perkin. Trans. 1. 1989, 1281-1283. (15) Compounds 6, 8, and 12 gave spectroscopic data consistent with the proposed structures. Satisfactory HRMS data (EI) was obtained for these compounds. (16) Evans, M. E.; Parrish, F. W. Acetal exchange reactions. Part 3. Monomolar acetalations of methyl R-D-mannosides-synthesis of methyl R-D-talopyranoside. Carbohydr. Res. 1977, 54, 105-114. (17) The use of low temperature for TFA deprotection was essential as reaction at room temperature led to γ-lactonization with hydroxyl at position 2. (18) (a) Poppe, L.; Brown, G. S.; Philo, J. S.; Nikrad, P. V.; Shah, B. H. Conformation of sLex Tetrasaccharide, Free in Solution and Bound to E-, P-, and L-Selectin. J. Am. Chem. Soc. 1997, 119, 1727-1736. (b) Ni, F. Recent developments in transferred NOE methods. Prog. Nucl. Magn. Reson. Spectrosc. 1994, 26, 517606. (c) Meyer, B.; Thomas, W.; Peters, T. Screening mixtures for biological activity by NMR. Eur J. Biochem. 1997, 246, 705-709.
JM010390F
