Article No. jmbi.1999.2621 available online at http://www.idealibrary.com on
J. Mol. Biol. (1999) 287, 853±858
COMMUNICATION
The Discovery of Steroids and Other Novel FKBP
Inhibitors Using a Molecular Docking Program
P. Burkhard, U. Hommel, M. Sanner and M. D. Walkinshaw*
Structural Biochemistry Unit
Edinburgh University
Michael Swann Building
Kings Buildings, Edinburgh
EH9 3JR, UK
The molecular docking computer program SANDOCK was used to
screen small molecule three-dimensional databases in the hunt for novel
FKBP inhibitors. Spectroscopic measurements con®rmed binding of over
20 compounds to the target protein, some with dissociation constants in
the low micromolar range. The discovery that FK506 binding protein is a
steroid binding protein may be of wider biological signi®cance. Twodimensional NMR was used to determine the steroid binding mode and
con®rmed the interactions predicted by the docking program.
# 1999 Academic Press
*Corresponding author
The immunophilin families of FK506 binding
proteins (FKBPs) and cyclophilins, are widely distributed in almost all tissues, and closely homologous forms are found in prokaryotes and
eukaryotes (Galat & Metcalfe, 1995). All members
of the family have a peptidyl prolyl isomerase
(PPIase) activity which has been shown to enhance
the rate of protein folding (Schmid et al., 1993).
Immunophilins are also involved in immunosuppression by the immunosuppressive drugs cyclosporin A (CsA) and FK506. These drugs promote a
ternary complex with the serine/threonine phosphatase calcineurin and block the signal transduction pathway in T cells (Schreiber et al., 1993).
Larger two-domain immunophilins, cyclophilin 40
and FKBP59, have been found to bind competitively to a chaperone heat shock protein Hsp90
(Ratajczak & Carrello, 1996) that is intimately
associated with steroid receptor control. It is therefore intriguing that the docking studies described
here show that the FKBP active site provides a
good binding pocket for a wide range of steroid
molecules.
The new computer screening program SANDOCK (Burkhard et al., 1998) was used to search
the Available Chemicals Database (Aldrich ChemiPresent addresses: P. Burkhard, Biozentrum, Basel
University, Klingelbergstrasse, Basel, Switzerland;
U. Hommel, Novartis Pharma AG, CH4002, Basel,
Switzerland; M. Sanner, The Scripps Research Institute,
Department of Molecular Biology, 10550 North Torrey
Pines Road, La Jolla, CA 92037, USA.
Abbreviations used: FKBP, FK506 binding protein.
E-mail address of the corresponding author:
[email protected]
0022-2836/99/150853±06 $30.00/0
cal Company) and the Cambridge Crystallographic
Database (Allen et al., 1991) for potential inhibitors
of FKBP. The approach is similar to that of the
database searching program DOCK (Desjarlais
et al., 1988; Knegtel et al., 1997) but with signi®cant
differences in the description of the binding pocket
and the evaluation of complementarity of ®t
between putative ligand and protein. The method
of using empirical interaction energies in a grid
was ®rst developed in the program GRID
(Goodford, 1985) and used in the fragment ®tting
program LUDI (Bohm, 1992, 1996). There are still
only a few examples of completely new chemical
entities discovered using these database mining
approaches (Bohm & Klebe, 1996; Ajay & Murcko,
1995; Somoza et al., 1998).
The X-ray structure of uncomplexed FKBP was
used to provide a template with the binding pocket
represented as a cluster of 183 site-points (Figure 1)
on the protein surface (Burkhard et al., 1998). The
site-points are assigned chemical properties
(hydrophobic or hydrogen bond donor or hydrogen bond acceptor) and represent idealised positions for complementary ligand atoms. The
scoring function used to grade the quality of ®t of
protein to potential ligand is composed of three
weighted terms comprising geometrical ®t, hydrophobic ®t and hydrogen bonding ®t. The SANDOCK distance-matching algorithm used for this
work required at least nine pairs of site-point distances to match with nine atom pair distances. The
tolerance of the matching distances was varied
Ê and 1.6 A
Ê . The tolerance value
between 0.6 A
depends on the geometric ®t and also on the
match between the atom type and the hydrogen
bonding or hydrophobicity of site-points. Best hits
# 1999 Academic Press
854
Docking of Steroids and Other FKBP Inhibitors
Figure 1. Site-points in FKBP. The dots which were used during the SANDOCK runs are colour-coded according
to their chemical properties: hydrophobic site-points (grey), hydrogen bond acceptor site-points (red), hydrogen bond
donor site points (blue), hydrogen bond donor/acceptor site points (magenta). The molecular surface of the protein
was calculated by the program MSMS (Sanner et al., 1996) and visualised with the graphics program DINO
(Philippsen, 1998; http://www.bioz.unibas.ch/ xray/dino).
from SANDOCK showing the putative ligands
docked to the protein were screened visually using
molecular graphics and ligands were selected for
binding tests.
Binding of the selected ligands was con®rmed
by ¯uorescence quenching which made use of the
fact that FKBP has only one tryptophan residue
which forms the ¯oor of the binding pocket.
Table 1A lists the substances found by SANDOCK
which bind to FKBP as con®rmed by ¯uorescence
quenching tests. Seven compounds were found to
bind in the mM range to FKBP, the best being Nbenzyloxycarbonyl-L-proline-L-proline with a binding constant of 0.8 mM. Similarity searches based
on compounds found by SANDOCK were also
successful in suggesting additional ligand molecules (Table 1B). Such an approach was used to
identify a series of steroids that were also shown
to complex with FKBP. All except 5b-Androstan3a-ol bind to FKBP in the micromolar range, but
the two steroids proposed by SANDOCK were
found to bind strongest (Table 1).
One-dimensional NMR analysis of the ligandFKBP complexes were also carried out on compounds 1, 3, 6 and 7 (Table 1A). The four protein
methyl resonances with the largest up-®eld shifts
are from residues Val55, Ile56 and Ile91 which all
lie in the ascomycin binding pocket. These chemical shifts provide a sensitive probe for speci®c
ligand binding. The one-dimensional NMR spectra
shown in Figure 2(a) indicate that the chemical
environment of these resonances has changed
upon addition of the ligands and clearly demonstrate ligand-binding in a site speci®c manner.
The structures of the FKBP-ascomycin (Petros
et al., 1991) and FKBP-FK506 (Lepre et al., 1992;
Vanduyne et al., 1993) complexes have been studied in solution by NMR, and the chemical shift
assignment of FKBP and its complex with bound
ascomycin have been reported (Rosen et al., 1991;
Xu et al., 1993). Here, two FKBP ligand complexes
(compounds 3 and 6) have been studied by heteronuclear multidimensional NMR techniques using
15
N-labelled FKBP. 15N-HSQC spectra were used to
identify the residues involved in ligand binding.
Figure 2(b) summarises the changes in chemical shift
observed upon complexation of FKBP with compounds 3 and 6. It is apparent from this Figure that
both ligands affect the same site. This is further illustrated in Figure 3(a) where the protein surface has
been colour coded according to chemical shift
changes shown in Figure 2(b). The affected residues
855
Docking of Steroids and Other FKBP Inhibitors
Table 1.
A. FKBP binding ligands selected by SANDOCK
Compounds 1 to 7 are a selection of hits from the SANDOCK program. The X-ray structure of FKBP
(Vanduyne et al., 1993) was used as a three-dimensional template and compounds were selected from the Available Chemical Directory. Fluorescence measurements were carried out using a Perkin-Elmer MPF-66 ¯uorescence
spectrophotometer with excitation wavelength 280 nm and emission wavelength of 340 nm and slit widths of
5 nm. The buffer solution was 50 mM Tris (pH 7.6) and 0.1 M NaCl and the protein concentration was 2 mM.
B. FKBP ligands selected on the basis of similarity to compounds selected in Table 1A
Nr
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
Name
Kd [mM]
N-Benzyloxycarbonyl-L-Pipecolidine
Tert-butyl Ester (similar to No 3)
CHC 100038352 (similar to No 1)
5b Pregnan 3b ol 20 on
5b Pregnan 3a,20a diol
5b Pregnan 3a, 11b diol 20 on
5b Pregnan 3a, 11b, 17a triol 20 on
5b Pregnan 3a, 11b, 20b triol
5b Pregnan 3a, 11a, 20b triol
5b Pregnan 11a, ol 3, 20 dion
5b Pregnan 3, 11, 20 trion
5b Pregnan 3a ol 20 on sulfate, triethyl ammonium salt
5b Pregnan 3a, 12a diol 20 on diacetat
5b Pregnan 20a ol 3 on acetat
5b Pregnan
5b Androstan 3a ol
5b Estran 3a, 17b diol
25
20
130
40
40
300
80
18
15
12
65
15
12
50
4200
18
Compounds 8 and 9 were obtained using similarity searches with MACCS data-base searching software
(Guner et al., 1991). Compound 8 is very similar to compound 3 with a pipecolidine ring like FK506 instead of
the proline ring found by SANDOCK, and also binds to FKBP with similar af®nity. The same holds for compound 9 which was selected because of its similarity to compound 1 and binds even better to FKBP than the
one selected by SANDOCK. Compounds 10 to 23 are steroid-based FKBP binding ligands extracted from the
Available Chemicals Database using a MACCS similarity search based on compound 6.
cluster around the binding site of the docked ligand
and involve a number of active site residues. The
only residues with large shift changes that are not
clustered around the ligand belong to the a-helix.
This may indicate that this helix as a whole is affected
by the complex formation. The effect of ligand binding on the NMR signals of helices has also been
observed by Shuker et al. (1996) and Hajduk et al.
(1997). Taken together, the one and two-dimensional
NMR results show that the binding site of these small
856
Docking of Steroids and Other FKBP Inhibitors
Figure 2. (a) 1H NMR shifts occurring on binding of compounds 1 (cyan), 3 (blue), 6 (red) and 7 (magenta) compared to the spectrum of the unliganded protein (black). The spectra were recorded on a Bruker DMX500 spectrometer. Protein concentrations were 0.2 mM in 100 mM NaCl, 20 mM Hepes (pH 7.4), 10 % 2H2O at 294 K.
(b) Chemical shift changes occurring upon ligand binding on FKBP-12. The weighted chemical shift changes
(w ˆ 0.2 (15N) ‡ (HN); ˆ jdbound ÿ dunboundj) are shown as coloured bars for substances 3 and 6 in red and
blue, respectively. Resonance assignments were based on previously published data (Rosen et al., 1991; Xu et al.,
1993) and were con®rmed using three-dimensional 15N-edited NOESY and DIPSI spectra. Spectra were recorded on a
Bruker AM500 spectrometer at 294 K and protein concentrations were 2 mM in H2O, 10 % 2H2O at pH 6.5.
compounds is the same as that of ascomycin and
FK506.
A conserved feature of the ligand binding is that
the hydrophobic moiety of the ligand is bound in
the active site cavity on top of the indole group of
Trp59. The predicted hydrogen bonding pattern of
the ligands to the protein is, however, variable and
involves the peptide nitrogen atom of Ile56, the
hydroxyl group of Tyr82, or the carboxyl group of
Asp37. Use was made of these hydrogen bonding
patterns in the SANDOCK program by imposing
distance constraints on the docked ligands such
that one of these hydrogen bonds had to be ful®lled. The space-®lling CPK representation of the
docked compound 1 within the surface of the protein molecule as depicted in Figure 3(b) illustrates
well the shape complementarity of the docked
ligand to the protein active site cavity. The ada-
Docking of Steroids and Other FKBP Inhibitors
857
Figure 3. (a) Stereo picture of the structure of FKBP when complexed with compound 7 as determined by the program SANDOCK. The surface of the protein is colour-coded according to the magnitude of chemical shift changes
introduced upon ligand binding. Residues with weighted shifts greater than 0.3 are coloured red, residues with
weighted shifts between 0.2 and 0.3 are coloured orange and residues with weighted shifts between 0.15 and 0.2 are
coloured yellow. The picture shows the largest chemical shift differences to occur in the pocket used for the generation of the cluster of site-points. This con®rms that ligand binding occurs at the predicted protein site (compare with
Figure 1). The best rigid body ®t selected by SANDOCK (shown here) gives 31 non-bonded contacts of less than
Ê between compound 7 and amino acid residues F36, D37, F46, V55, I56, Y82, H87, I90 and F99. Two possible
3.7 A
Ê
FKBP . . . ligand hydrogen bonds are shown as thin lines; Phe36(O) . . . 3aOH ˆ 2.95 A
and Asp37
Ê . There is one unacceptably short O . . . C contact of 2.7 A
Ê which involves the hydroxyl group of
(Od) . . . 3aOH ˆ 3.29 A
Tyr82 and this can be easily relieved to an acceptable distance by an 8 rotation about w1. (b) CPK-representation of
compound 1 in the active site cavity of FKBP illustrating the excellent lock and key (rigid body) ®t between ligand
and protein as predicted by SANDOCK. The adamantyl moiety of the ligand is bound in the hydrophobic pocket
above the residue Trp59. Two potential hydrogen-bonding interactions are shown as thin red lines. There are 35 nonÊ between compound 1 and FKBP. The three unacceptable contacts between 2.8 and
bonded contacts of less than 3.7 A
Ê can be relieved by a w1 twist for Asp37 which also improves the geometry of the N(ligand) . . . Asp (Od) hydrogen
3A
Ê (as shown in the Figure) to 2.8 A
Ê.
bond from a distance of 3.47 A
858
Docking of Steroids and Other FKBP Inhibitors
mantyl moiety sits in the hydrophobic pocket of
the protein on top of Trp59. The phenyl ring is
bound in another hydrophobic compartment of the
active site cavity beneath His87, while the nitrogen
and sulfur atoms of the ligand form hydrogen
bonds to the side-chains of Asp37 and Tyr82,
respectively.
Results presented here show that SANDOCK
has an ef®cient docking algorithm in which the
geometrical and chemical parameters used in the
scoring function seem to provide good discrimination between docked ligands. The discovery of
these novel ligands for FKBP opens the way to the
design of new families of immunophilin inhibitors
and also begs the question of whether steroidFKBP interactions are biologically relevant in the
cell.
Acknowledgements
We thank Novartis AG for supporting this work and
Dr Claudio Dalvit for helpful discussion.
References
Ajay, & Murcko, M. A. (1995). Computational methods
to predict binding free-energy in ligandreceptor complexes. J. Med. Chem. 38, 4953-4967.
Allen, F. H., Davies, J. E., Galloy, J. J., Johnson, O.,
Kennard, O., Macrae, C. R., Mitchell, E. M.,
Mitchell, G. R., Smith, J. M. & Watson, D. G. (1991).
The development of version-3 and version-4 of the
Cambridge Structural Database system. J. Chem.
Info. Comp. Sci. 31, 187-204.
Bohm, H. J. (1992). Ludi-rule-based automatic design of
new substituents for enzyme- inhibitor leads.
J. Comp. Aided Mol. Design, 6, 593-606.
Bohm, H. J. (1996). Towards the automatic design of
synthetically accessible protein ligands - peptides,
amides and peptidomimetics. J. Comp. Aided Mol.
Design, 10, 265-272.
Bohm, H. J. & Klebe, G. (1996). What can we learn from
molecular recognition in protein-ligand complexes
for the design of new drugs?. Ange. Chemie Int. Edit.
35, 2589-2614.
Burkhard, P., Taylor, P. & Walkinshaw, M. D. (1998).
An example of a protein ligand found by database
mining: description of the docking method and its
Ê X-ray structure of a thromveri®cation by a 2.3 A
bin-ligand complex. J. Mol. Biol. 277, 449-466.
Desjarlais, R. L., Sheridan, R. R., Seibel, G. L., Dixon,
J. S., Kuntz, I. D. & Venkataraghavan, R. (1988).
Using shape complementarity as an initial screen in
designing ligands for a receptor-binding site of
known 3-dimensional structure. J. Med. Chem. 31,
722-729.
Galat, A. & Metcalfe, S. M. (1995). Peptidylproline cis/
trans isomerases. Prog. Biophys. Mol. Biol. 63, 67-118.
Goodford, P. J. (1985). A computational-procedure for
determining energetically favorable binding-sites on
biologically important macromolecules. J. Med.
Chem. 28, 849-857.
Guner, O. F., Hughes, D. W. & Dumont, L. M. (1991).
An integrated approach to 3-dimensional information management with maccs-3d. J. Chem. Info.
Comp. Sci. 31, 408-414.
Hajduk, P. J., Meadows, R. P. & Fesik, S. W. (1997).
Drug design - discovering high-af®nity ligands for
proteins. Science, 278, 497.
Knegtel, R. M. A., Kuntz, I. D. & Oshiro, C. M. (1997).
Molecular docking to ensembles of protein structures. J. Mol. Biol. 266, 424-440.
Lepre, C. A., Thomson, J. A. & Moore, J. M. (1992). Solution structure of fk506 bound to fkbp-12. FEBS
Letters, 302, 89-96.
Petros, A. M., Gampe, R. T., Gemmecker, G., Neri, P.,
Holzman, T. F., Edalji, R., Hochlowski, J., Jackson,
M., Mcalpine, J., Luly, J. R., Pilotmatias, T., Pratt, S.
& Fesik, S. W. (1991). NMR-studies of an fk-506
analog, [u-c-13]ascomycin, bound to fkbp - conformation and regions of ascomycin involved in binding. J. Med. Chem. 34, 2925-2928.
Philippsen, A. (1998). DINO. A Visualisation System for
Structural Data, University of Basel, Switzerland.
Ratajczak, T. & Carrello, A. (1996). Cyclophilin-40 (cyp40), mapping of its hsp90 binding domain and evidence that fkbp52 competes with cyp-40 for hsp90
binding. J. Biol. Chem. 271, 2961-2965.
Rosen, M. K., Michnick, S. W., Karplus, M. & Schreiber,
S. L. (1991). Proton and nitrogen sequential assignments and secondary structure determination of the
human fk506 and rapamycin binding-protein. Biochemistry, 30, 4774-4789.
Sanner, M. F., Olson, A. J. & Spehner, J. C. (1996).
Reduced surface - an ef®cient way to compute molecular-surfaces. Biopolymers, 38, 305-320.
Schmid, F. X., Mayr, L. M., Mucke, M. & Schonbrunner,
E. R. (1993). Prolyl isomerases - role in protein-folding. Advan. Protein Chem. 44, 25-66.
Schreiber, S. L., Albers, M. W. & Brown, E. J. (1993).
The cell-cycle, signal-transduction, and immunophilin ligand complexes. Acc. Chem. Res. 26, 412-420.
Shuker, S. B., Hajduk, P. J., Meadows, R. P. & Fesik,
S. W. (1996). Discovering high-af®nity ligands for
proteins - sar by NMR. Science, 274, 1531-1534.
Somoza, J. R., Skillman, A. G., Munagala, N. R., Oshiro,
C. M., Knegtel, R. M. A., Mpoke, S., Fletterick, R. J.,
Kuntz, I. D. & Wang, C. C. (1998). Rational design
of novel antimicrobials: blocking purine salvage in
a parasitic protozoan. Biochemistry, 37, 5344-5348.
Vanduyne, G. D., Standaert, R. F., Karplus, P. A.,
Schreiber, S. L. & Clardy, J. (1993). Atomic structures of the human immunophilin fkbp-12 complexes with fk506 and rapamycin. J. Mol. Biol. 229,
105-124.
Xu, R. X., Nettesheim, D., Olejniczak, E. T., Meadows,
R., Gemmecker, G. & Fesik, S. W. (1993). H-1, C-13,
and N-15 assignments and secondary structure of
the W 06 binding-protein when bound to ascomycin. Biopolymers, 33, 535-550.
Edited by F. E. Cohen
(Received 25 September 1998; received in revised form 29 January 1999; accepted 30 January 1999)