Tài liệu Báo cáo khoa học: Role of hydroxyl group and R/S configuration of isostere in binding properties of HIV-1 protease inhibitors docx

Role of hydroxyl group and R / S configuration of isostere in binding properties of HIV-1 protease inhibitors Hana Petrokova´1, Jarmila Dusˇkova´1, Jan Dohna´lek1, Tereza Ska´lova´1, Eva

Role of hydroxyl group and R / S configuration of isostere in binding properties of HIV-1 protease inhibitors

Hana Petrokova´1, Jarmila Dusˇkova´1, Jan Dohna´lek1, Tereza Ska´lova´1, Eva Vondra´cˇkova´-Buchtelova´1, Milan Soucˇek2, Jan Konvalinka2, Jirˇı´ Brynda3, Milan Fa´bry3, Juraj Sedla´cˇek3and Jindrˇich Hasˇek1

1

Institute of Macromolecular Chemistry,2Institute of Organic Chemistry and Biochemistry and3Institute of Molecular Genetics, Academy of Sciences of the Czech Republic, Praha, Czech Republic

The crystal structure of the complex between human

immunodeficiency virus type 1 (HIV-1) protease and a

peptidomimetic inhibitor of ethyleneamine type has been

refined to R factor of 0.178 with diffraction limit 2.5 A˚ The

peptidomimetic inhibitor Boc-Phe-Y[CH2CH2

NH]-Phe-Glu-Phe-NH2(denoted here as OE) contains the

ethylene-amine replacement of the scissile peptide bond The inhibitor

lacks the hydroxyl group which is believed to mimic

tetra-hedral transition state of proteolytic reaction and thus is

suspected to be necessary for good properties of

peptido-mimetic HIV-1 protease inhibitors Despite the missing

hydroxyl group the inhibition constant of OE is 1.53 nM

and it remains in the nanomolar range also towards several

available mutants of HIV-1 protease The inhibitor was

found in the active site of protease in an extended

conformation with a unique hydrogen bond pattern different from hydroxyethylene and hydroxyethylamine inhibitors The isostere nitrogen forms a hydrogen bond to one catalytic aspartate only The other aspartate forms two weak hydro-gen bridges to the ethylene group of the isostere A com-parison with other inhibitors of this series containing isostere hydroxyl group in R or S configuration shows different ways

of accommodation of inhibitor in the active site Special attention is devoted to intermolecular contacts between neighbouring dimers responsible for mutual protein adhe-sion and for a special conformation of Met46 and Phe53 side chains not expected for free protein in water solution Keywords: ethyleneamine inhibitor; HIV-1 protease; pepti-domimetic inhibitor; X-ray structure

The HIV-1 protease, the aspartic protease that cleaves

specific peptide bonds in precursor gag-pol proteins to form

the mature proteins, is essential for production of infectious

HIV particles Its inhibition is an efficient method of

treatment of the acquired immunodeficiency syndrome

(AIDS) and related diseases [1,2] No matter what is the

inhibitor type (symmetrical or unsymmetrical), the HIV

protease covers any inhibitor under its flaps (Fig 1) forming

thus a characteristic long binding tunnel with the cleavage

site in the middle One structural feature present in most

tight-binding aspartic protease inhibitors is a critical

hydroxyl group that replaces the catalytical water molecule

in the active site This hydroxyl group forms hydrogen

bonds to the catalytically active aspartates [3] in HIV

protease complexes with all hydroxyethylamine inhibitors

being by far the most frequently studied structures in the Protein Data Bank [4] and the HIV Protease Database (http://srdata.nist.gov/hivdb) and therefore it has been supposed necessary for tight-binding of aspartic protease inhibitors

This paper presents the structure of native HIV-1 protease in a complex with inhibitor Boc-Phe-Y[CH2CH2NH]-Phe-Glu-Phe-NH2(denoted here as OE) The inhibition constant of OE remains low (1.53 nM) in spite of the fact that the critical hydroxyl group is completely missing in this inhibitor and it remains in the nanomolar range also for several available mutants of

HIV-1 protease (e.g 4.HIV-1 nM for the A71V/V82T/I84V mutant arising after Indinavir treatment) [5,6]

The structure of OE complex has been solved in the frame

of a systematic structure study of a group of inhibitors with very similar chemistry They differ only in the presence of the hydroxyl group in the isostere and in its configuration (R or S) which is considered crucial for tight binding of hydroxyethylamine inhibitors The fact that the inhibition constant of the respective inhibitors (denoted here as OE,

RE and SE) does not differ much deserves closer attention The inhibitor OE (without OH group) has only a slightly lower inhibition efficiency than similar inhibitors possessing hydroxyl group in S or R configuration (Ki,OE¼ 1.5 nM,

Ki,RE¼ 0.12 nM, Ki,SE¼ 0.15 nM) [5,7–9]

Our recent studies of the hydroxyethylamine inhibitor complexes [7–9] revealed that the binding tunnel of protease, abundant in hydrogen bond donors and acceptors, can bind the inhibitors in several possible ways Most of the

Correspondence to H Petrokova´, Institute of Macromolecular

Chemistry, Academy of Sciences of the Czech Republic, Heyrovske´ho

na´m 2, 162 06 Praha 6 Fax: +420 296809 410,

Tel.: +420 296809 205, E-mail: petrokova@imc.cas.cz

Abbreviations: OE, Boc-Phe-Y[CH 2 CH 2 NH]-Phe-Glu-Phe-NH 2 ; RE,

Boc-Phe-Y[(R)-CH(OH)CH 2 NH]-Phe-Glu-Phe-NH 2 ; SE,

Boc-Phe-Y[(S)-CH(OH)CH 2 NH]-Phe-Glu-Phe-NH 2

Enzyme: retropepsin (EC 3.4.23.16).

Note: The crystallographic data of the complex HIV-1 protease with

OE inhibitor have been deposited with the Protein Data Bank and are

available under access code 1m0b.

(Received 21 June 2004, revised 9 September 2004,

accepted 29 September 2004)

hydroxyethylene inhibitors are bound to the HIV)1

protease by interaction of the hydroxyl group of isostere

with both aspartates Asp25, Asp125 A slight shift (about

0.5 A˚) of the isostere group in the case of

hydroxyethyl-amine inhibitors SE or SQ [7,8] causes that the catalytic

aspartates bind mainly to the isostere NH group leaving

only one contact to the isostere hydroxyl group The NH

group thus partly substitutes the role of the hydroxyl group

in hydroxyethylene inhibitors In this paper, we present

another unusual binding mode of the ethyleneamine

inhibitor OE where only its isostere NH group makes a

contact to one catalytic aspartate We suggest that the

binding mode where the role of the hydroxyl group is

completely overtaken by the isostere NH group can be

generalized for the whole class of ethyleneamine inhibitors

Materials and methods

Crystallization and crystal parameters

A solution of HIV protease at concentration 3 mgÆmL)1

in 50 mMsodium acetate buffer (pH 5.8) containing 0.5%

(v/v) 2-mercaptoethanol was mixed with the inhibitor [12]

dissolved in dimethyl sulfoxide at 11 mMin the volume ratio

20 : 1 and left at 4C for at least 30 min prior to

crystallization; this gave the final fourfold molar excess of

inhibitor [11] Co-crystallization by hanging drop diffusion

technique against 1 mL reservoir of 0.2–0.6M NaCl in

0.1MNa citrate buffer, pH 4.5–5.5 followed Rod-shaped

hexagonal crystals appeared overnight and continued to

grow over the next 7 days Before flash-freezing, the crystals

were soaked for 30 s in the mother liquor containing 20%

glycerol

Data collection and processing

X-ray diffraction data were collected at the European

Synchrotron Radiation Facility (ESRF) in Grenoble at

BM29 beamline equipped with a MAR 345 detector Data were collected from a single crystal (0.06· 0.06· 0.7 mm) at a temperature of 100 K The oscilla-tion range was 1.5 and each frame was exposed for

30 s The distance of the crystal to detector plate was

100 mm The diffraction data extended to 2.0 A˚ Inten-sities were integrated, scaled and merged using the HKL

software [12] Data were reduced in the P61 space group The unit cell dimensions were a¼ b ¼ 62.7 A˚, c ¼ 82.2 A˚ The details of X-ray diffraction data collection are described in Table 1

Refinement Refinement was carried out using the CNS program package [13] Parameters for nonstandard parts of inhibitor were set in agreement with several structures found in the CCDC database [14] The rigid body refinement was performed with the starting model of the protease dimer taken from the Protein Data Bank [4] (PDB code 1aaq) Several cycles of CNS refinement (positional and individual B factor optimization) and rebuilding using the graphics programO[15] were carried out The noncrystallographic symmetry was applied during the refinement to both the protease and the inhibitor at the initial stages of refinement with the weight of 300 kcalÆmol)1ÆA˚)2 Later, it was partially

Fig 1 Front view of the structure of native HIV-1 protease complexed

with OE inhibitor The inhibitor (stick model) sits over the catalytic

aspartates (ball-and-stick model) and is completely covered by

prote-ase flaps belonging to two monomers of proteprote-ase related by an

approximate two-fold symmetry axis The exact C 2 symmetry is

per-turbed by asymmetry of the inhibitor and also by contacts between

neighbouring protease subunits.

Table 1 Statistics of diffracted intensity measurement Complex of native HIV-1 protease with inhibitor OE R sym ¼ S|I ) <I>|/ S<I>.

All reflections The highest shell

Table 2 Parameters describing the quality of refined model of the native HIV-1 protease complexed with inhibitor OE.

Parameter

% of cases in the most favored regions of Ramachandran plot

92.4

% of cases in disallowed regions

of Ramachandran plot

0

liberated to 50 kcalÆmol)1ÆA˚)2 Cross validation was

performed with the use of the Rfree factor calculated

from 5% of the reflections

The inhibitor was uniquely identified and modeled in the

2Fo-Fc map in its expected position in two opposite

orientations at R¼ 0.256 and Rfree¼ 0.294 After the

structure refinement to R¼ 0.235 and Rfree¼ 0.272, a total

of 159 water molecules were gradually included The criteria

for accepting waters in refinement were as follows: the

presence of the Fo-Fcelectron density peak at 3 r level, the

2Fo-Fcpeak at 1 r level, and at least one hydrogen-bonding

partner within the distance 2.2–3.6 A˚

The resulting structure (Fig 1) has R¼ 0.178 and

Rfree¼ 0.242 Parameters describing the quality of the final

structure were checked by programPROCHECK[16] and are

given in Table 2 The symmetry of the complex and rigidity

of different parts of the protease can be seen in Figs 2,3 The

figures were produced using programsMOLSCRIPT[17] and

RASTER3D [18] or WLVIEWERPRO (http://molsim.vei.co.uk/

weblab)

Results

Structure of the complex of native HIV-1 protease with OE inhibitor

The structure of the HIV-1 protease in the complex with inhibitor OE (Fig 1) has been finally refined to R¼ 0.178 and Rfree¼ 0.242 The protein is a homodimeric molecule, made up of two 99-residue polypeptide chains: chain A, Pro1-Phe99 and chain B, Pro101-Phe199 The inhibitor is bound as an extended chain in a tunnel running under flaps across the dimer interface The flaps consist of the amino-acid residues 46–55 and 146–155 from both polypeptide chains and completely close the inhibitor in the active site channel The tips of flaps bind together by hydrogen bond (NH Gly51…C ¼ O Ile150 or NH Gly151…C ¼ O Ile50) The nitrogen atoms of Ile50 and Ile150 in flaps interact with the inhibitor via hydrogen bonding through a single water molecule (W401) typically present in structures of HIV-1 protease with peptidomimetic inhibitors

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4

1

Residue number

Fig 2 Root-mean-square distances between

the corresponding atoms of individual residues

after the least-squares alignment of C a atoms of

two subunits A and B of the complex of HIV-1

protease with OE inhibitor.

0 20 40 60 80

0

20

40

60

80

21 11

Residue number

chain A

chain B

B factor

B factor

Fig 3 The complex of HIV-1 protease with

OE inhibitor B factors averaged over each

residue Top, chain A; bottom, chain B.

Inhibitor binding.The inhibitor OE binds into the protease

binding tunnel in an extended conformation in two opposite

directions (Fig 4) The flaps are locked over the inhibitor by

water molecule W401 forming hydrogen bonds to NH

groups of Ile50, Ile150 and to carbonyls of the inhibitor in

positions P1¢ and P2(Fig 5) The side chains of the

tert-butyloxycarbonyl (P2), phenylalanine (P1), phenylalanine

(P1¢), glutamine (P2¢) and phenylalaninamide (P3¢) bind in

the respective S2, S1, S1¢, S2¢ and S3¢ pockets of the HIV-1

protease (Fig 6)

The inhibitor was modelled into a well defined Fo-Fc

electron density after the protein was refined to R¼ 0.25

and Rfree¼ 0.29 The disordered inhibitor was modelled in

the active site of protease in two opposite orientations

(Fig 4), here referred as I and Y chains (residue numbers in

PDB file 301A-306A and 301B-306B, respectively) The

average rmsd of inhibitors in opposite orientations (I and Y)

was 0.15 A˚

The scheme of hydrogen bonds (Fig 5) shows that all

proton donors and acceptors of the inhibitor are involved in

hydrogen bonding Five hydrogen bonds of the total 18

Fig 4 Stereoview of two alternative positions of the inhibitor OE in the electron density 2F o -F c at 1 r level Both orientations were refined with occupation factors 0.5 In ball-and-stick model the active site Asp25–Gly27 and Asp125–Gly127 are highlighted.

N O

N O

N O

O O

N O

O O

N

OH O N

N O N

O

N

O

O N

N O

N

O

O O

N O

O OH 3.5

Asp125

Gly127

Asp25

Gly27

Gly48

3.6

2.7

2.5 3.2

W401 3.3

W402

3.1

W464

2.4

3.5

Asp29

Asp30

W424 W507

3.3 3.1

2.8 3.0 3.3

2.6

3.0

3.3 2.9

Fig 5 Network of hydrogen bonds formed by the inhibitor OE in the native HIV-1 protease Distances are given in A˚ Statistics: one intramolecular hydrogen bond CO…NH, seven hydrogen bonds to protease main-chain

NH, three to side-chain carboxyls, three to main-chain carbonyls and seven to water molecules.

D129G148 G149 F153 R8

N

N O

N

O OH O

N

O O

D30 A28 V32 I150I47 G48 I84

G149 I150 P81V82

I84 D25T80

G127

A128 D130 V132 I147 D129 G148

G49

V182 L123 P181 G27

G48

I50

D25

S2

S1

S1'

S2' S3'

Fig 6 Review of hydrophobic interactions (short C-C contacts up to 4.1 A˚) of the inhibitor OE side chains with binding pockets of HIV-1 protease.

hydrogen bonds connect directly the inhibitor main chain

with the main chain of protease (NH…CO to Gly23,

Gly123, Gly48, CO…NH to Gly48, Asp29) Two hydrogen

bonds connect the side chain of the inhibitor (Glu) with the

protein main chain (CO…NH to Asp29, Asp30) Three

hydrogen bonds connect inhibitor with protease side chains

(NH…Asp125, COOH…Asp30, NH2…Asp30) Three

structurally important water molecules W401, W402 and

W464 bridge directly the inhibitor with protease and form

five hydrogen bonds to inhibitor P2 (CO…W402), P1

(NH…W402 and CO…W401), P01 (CO…W401), P02

(CO…W464) The whole inhibitor is almost completely

buried in the binding tunnel of protease Only the inhibitor

ends seem to be exposed to solvent Two water molecules

W424 and W507 were found in the difference map forming

two hydrogen bridges to the N terminal group of inhibitor

Unusual conformation of isostere in the inhibitor OE

enables also a weak intramolecular hydrogen bond

con-necting the Boc carbonyl with the NH group at P1¢ position

(O…N at 3.6 A˚) The inhibitor refined in the opposite

orientation binds in a similar way

The inhibitor Glu at P2¢ is totally buried in the protease

S2¢ site and makes six hydrogen bonds to Asp30 and Asp29

The importance of this residue and the strength of its

binding to protease are supported also by the fact that it has

the lowest B-factors of the whole inhibitor

The water molecule W401 that hydrogen bonds to the

main chain NH group of both flaps as well as to the

inhibitor carbonyls is observed in most HIV protease

complexes with peptidomimetic inhibitors It was clearly

seen at the difference map though it has a considerably high

B factor (64 A˚2) High displacement factor is a result of

probable disorder of W401 caused by two orientations of

inhibitor and asymmetrical binding of water with respect to

the noncrystallographic C2 axis The positions of two

inhibitor carbonyls (one from Boc group at P2position and

another from Phe at P1¢) that are bridged by W401 are not

symmetrical with respect to the noncrystallographic C2axis

and thus the alternative positions of the bridging water

W401 are different However, only the average position of

W401 was refined in our structure model

In spite of numerous hydrogen bonds, hydrophobic forces

seem to be a dominating interaction between protease and

inhibitor As an indicator of these hydrophobic interactions,

distances between carbon atoms of inhibitor and the protease

were calculated using cutoffs of 3.6 A˚ and 4.1 A˚ [19] All

hydrophobic contacts and hydrogen bonds that are involved

in the inhibitor OE binding are summarized in Table 3

Conformation of isostere.Two different conformations of

the isostere of inhibitor fitting well the electron density were

modeled (Fig 7) In the first conformation, the NH group

of Phe in position P1¢ makes only one hydrogen bond to one

of the catalytic aspartates and also forms a weak

intra-molecular hydrogen bond to the carbonyl group of the Boc

residue In the second conformation, the isostere NH group

binds almost symmetrically between the Asp25 and Asp125

gaining thus additional hydrogen bond to the protease

However, theCNSrefinement run with both conformations

resulted in the same result very similar to the first

conformation Therefore, the inhibitor was refined in form

that corresponds to the first conformation with only one

hydrogen bond to catalytic aspartates not observed with other inhibitors However, it seems that the isostere conformation is probably not fixed and that we have to admit possible concerted conformational changes at this site Because both conformations of the isostere group fit well into the 2Fo-Fcelectron density map, we assume that the inhibitor can change its conformation inside the cavity and that this positively contributes to the inhibitor binding Two fold noncrystallographic symmetry of HIV protease The two HIV protease subunits are related by an approxi-mate two-fold noncrystallographic axis Figure 2 shows the deviations between the corresponding atoms of chain A and chain B (averaged for each residue) after the least-squares

Table 3 Summary of all contacts and hydrogen bonds for inhibitor OE complexed in the native HIV-1 protease Three water molecules involved in protein inhibitor interaction are included The upper table (six rows) concerns the I orientation of inhibitor, the lower table (six rows) concerns the Y orientation of inhibitor Contacts to water molecules (W) are given after the + sign Short contacts in the C-C column are supposed to be repulsive, the short contacts (hydrogen bonds) in columns 3–5 contribute to good binding ability of inhibitor

to HIV protease.

Hydrophobic contacts

Hydrogen bonds C-C

up to 4.1 A˚

C-O/N (+W)

up to 3.6 A˚

O/N-O/N

up to 3.6 A˚

O/N-W

up to 3.6 A˚

Fig 7 Two conformations of isostere in the inhibitor OE with very similar energy can be placed in the 2F o -F c electron density of the refined structure of the complex of HIV-1 protease with OE inhibitor Torsion angles of the preferred conformation of the OE isostere are listed in Table 4 Only one orientation of the disordered inhibitor OE is shown

in this figure.

alignment of Caatoms According to expectation, the main

differences were found in the tips of flaps where the two-fold

noncrystallographic symmetry is not possible because the

hydrogen bonds connecting tips of flaps cannot be present

with its two-fold image in the same molecule at once

Therefore, the alternative positions were modeled at the

main chains for Ile50-Gly51 and Ile150-Gly151 The

least-squares fit of all Cain superimposed monomers of protease

gives the best overlap by rotation of 178 with the rmsd

0.07 A˚ for all 758 atom pairs and 0.06 A˚ when the residues

49–53 and 149–153 from the tips of flaps are excluded

B-factors.The highest B factors (above 50 A˚2) indicate high

conformational instability in loops Leu38-Lys44 and

Leu138-Lys144, symmetrically in hinges of both flaps,

whereas the flap ends seem to be well stabilized by

interactions with inhibitor, namely those mediated by water

molecule locked over the inhibitor carbonyls (Fig 3) The

map of electron density shows the highest orientational

disorder in Arg41, Arg141, Lys43 and Lys143 side chains

(zero occupation factors in the PDB file)

Adhesion between proteins and protease activity in the

crystal form Adhesion between protein molecules plays

an important role in their function in biological systems

[8] Structure changes of the protein surface when exposed

to solvent or when involved in adhesion with

neighbour-ing protein molecule have undoubtedly important

bio-logical implications Here, the HIV-1 protease molecules

stick together by middle parts of flaps and form a special

helical arrangement with the 61symmetry and with active

sites directed into the wide solvent tunnels passing

through the whole crystal enabling thus an easy exchange

of solvent even in the active sites of proteases This

explains experimentally verified exchange of inhibitors in

the protease single crystal without an extensive destruction

of the crystal [20] In our structure, one of the preferred

interactions between two neighbouring HIV-1 protease

dimers is localized at the top of flaps The Phe53 from

one protease dimer and the Phe153 from the other

symmetrically related dimer form a convenient parallel stacking of phenyl rings joining thus these two dimers together (Fig 8A) These p–p interactions appear on both sides of each protease dimer and thus form an infinite helical arrangement of the protease complexes in the P61 space group This molecular arrangement is supported by Met46 (Met146) which forms S…H-C short contacts to Phe53 (Phe153) from the same molecule (Fig 8B) The time-averaged view of the protease complex shows that all residues involved in contact – Phe53, Phe153, Met46, Met146 were found in two distinct conformations with an occupation factor of 0.5 (confirmed by refinement) These alternative conformations form two favorable parallel stackings leading to two quite different water channels – called here closed and open solvent channel

The inner virtual diameter of the closed solvent channel is 8.7 A˚ The inner surface of the channel is formed by 12 Cc

atoms of six phenyl pairs (Phe53 and Phe153) per one helix turn (Fig 9A) The phenyl pairs are held together by p–p interactions and the Sdatoms of Met46 and Met146 form close interactions with these phenyls Six of these residue quartets form a steep spiral ridge inside the solvent channel The virtual diameter of the open solvent channel is 12.4 A˚ The surface of the channel is formed by 12 sulfur Sd atoms of Met46 and Met146 per one turn of helix (Fig 9B)

In the open solvent channel the methionines are turned into the solvent and do not form significant contacts to protein Thus, the solvent tunnel cross-section is not rigid because different conformations of Phe53, Phe153, Met46 and Met146 lead to different tunnel diameters and also to different hydrophobicity of the solvent tunnel surface The fact that inter–protein interaction can influence the inhibi-tion process may be important for interpretainhibi-tion of the protease function

Alternative conformations The HIV protease complexes are not rigid The crystal structure of HIV-1 PR with OE described here is a mixture of many conformation states In addition to residues Leu38-Lys44 and Leu138-Lys144 localized in flexible flap hinges (see the chapter on

Fig 8 Dimers of the complex of HIV-1 protease with OE inhibitor are linked together by p–p interactions of phenyl rings of Phe53 and Phe153 of neighbouring molecules to form a helix along the crystallographic c axis This p–p interactions are supported by CH…S hydrogen bonds between the Phe53A … Met46A and Phe153A … Met146A (A) Stacking of neighbour molecules and positions of inhibitors in subsequent protease dimers forming the helix (B) The detail of interacting residues Phenylalanines are disordered 1 : 1 in two conformations A and B leading to parallel stacking of phenyl rings in each conformation.

B-factors), other eight amino-acid residues have been found

each in two distinct alternative conformations: Met46,

Met146, Ile50, Ile150, Gly51, Ile151, Phe53 and Phe153 All

of them have clear interpretation Two conformations of

Met46 are associated with corresponding alternative

con-formations of Phe53 so that either Met46(A) and Phe53(A)

or Met46(B) and Phe53(B) can be present simultaneously in

the structure and analogously in the second chain of the

protease dimer The residues Met46 and Phe53 placed on

opposite sides of one flap seem to be of principal importance

for intermolecular arrangement of molecules in crystal

Figure 8 shows that the Phe53 rings are in both orientations

stacked by p–p interaction with Phe153 rings of the

neighbouring protein dimer in crystal The alternative

positions of phenylalanine rings are formed in concert with

the alternative positions of Sdand Ccof Met46 and Met146

(from neighbouring protease dimer)

Residues Ile50 and Ile150, Gly51 and Gly151 are found

at the tips of the flaps making a hydrogen bond to each

other These hydrogen bonds are mediated by NH groups

of one chain with carbonyl groups of the other chain

Because the main chains of the two loops have the same

orientation of C ends at the place of contact, the

noncrystallographic two-fold symmetry cannot be realized

in this location Therefore, two possible orientations were

modelled in this structure, which differ in the flipped

peptide bond between residues Ile50-Ile51 The side chains

of neighbouring Ile50 and Ile150 already fitted well the

electron density in a single position

Comparison of inhibitors OE, SE and RE complexing

the native HIV-1 protease

The availability of three experimental structure

determina-tions of very similar inhibitors OE, RE and SE complexed

with the native protease gives a detailed insight into the nature of interaction inhibitor – protease The chemical structure of OE, RE, SE inhibitors can be seen from the scheme in Table 4 The inhibitors RE, SE differ in absolute configuration of the CHOH group of the isostere and their inhibition constants are surprisingly the same Ki,SE¼ 0.15 nM, Ki,RE¼ 0.12 nM[1] The isostere carboxyl group replaced in OE by CH2 makes the inhibition constant somewhat lower Ki,OE¼ 1.53 nM [1]; however, the resist-ance of OE to protease mutations seems to be better The experimental structure of HIV-1 protease with inhibitor SE was determined by Dohna´lek [8] R¼ 0.18, diffraction limit 3.1 A˚, PDB code 1fqx The structure with the RE inhibitor was determined by Dusˇkova´ (unpublished results) with R¼ 0.173, diffraction limit 2.0 A˚

The overall layout of the inhibitor in the protease binding site is very similar for all the compared inhibitors The side chains of all the inhibitors are placed similarly in their pockets, though, not identically (Fig 10)

Conformation of inhibitor backbones The compared inhibitors differ mainly in their isostere areas The OE inhibitor possesses the nonscissile isostere without any hydroxyl group, whereas the hydroxyethylamine inhibitors

SE and RE have the hydroxyl group of isostere in the S or R configuration, respectively The isostere of the OE inhibitor has quite different conformation compared with the RE and

SE inhibitors The inhibitor OE lacks the hydroxyl group which fixes a unique conformation by strong hydrogen bonds to Asp25 and/or Gly27 allowing thus in principle more possible conformations of the isostere group Possible variations in torsion angles in the inhibitor are cooperative This means that any change in one torsion angle should be compensated by change of other torsion angles in the opposite direction to keep all the side chains in the whole

Fig 9 Solvent tunnels passing through the crystal of HIV-1 protease complexed with OE inhibitor along the crystallographic axis c The residues Met46, Phe53, Met146 and Phe153 participating in the solvent tunnel formation have two different conformations: A, narrow tunnel (A) and B, wide tunnel (B) The inhibitors OE of six complexes forming one turn of the helix are also shown in one orientation in thick stick representation and coloured according to atom types In the case of
narrow tunnel, the diameter of the effective view through the solvent channel is 8.7 A˚ (determined

by distance of projections of two opposite C c atoms of Phe53 and Phe153) In the case of
wide tunnel, the diameter of the effective view through the solvent channel is 12.4 A˚ (measured as a distance of projections of opposite S d atoms of Met46 and Met146).

inhibitor approximately in the same position Thus, in spite

of the fact that the individual torsion angles in the isostere

differ largely (Table 5), the
overall dihedral angles (i.e the

dihedral angles determining the mutual orientation of the

side chains in S1 and S1¢ sites) are almost identical (141, 148

and 148 for inhibitors OE, SE and RE, respectively) The

NH group of Phe in P1¢ position of the OE inhibitor is

rotated with respect to SE, RE preserving one hydrogen

bond to the catalytic Asp25 only The intramolecular

hydrogen bond NH…O between ester oxygen of Boc and

NH of Phe in P1¢ position probably stabilizes this unusual conformation of the OE inhibitor in the active site The aspartate that does not bind to the isostere NH seems to make weak hydrogen bonds to both CH2 groups of the isostere The changes of torsion angles in the isostere also resulted in shifts of Caof Phe in P1¢ position of the OE inhibitor by 0.58 and 0.46 A˚ in comparison with SE and RE complexes, respectively, in both cases in the direction away

Fig 10 Stereoview of the position of inhibitor OE (yellow carbons), RE (green carbons) and SE (violet carbons) in the binding tunnel of native HIV protease Conformation of different inhibitors is not identical Note a different orientation of OH groups in isosteres of RE and SE The

S configuration forces the OH groups to the inconvenient orientation for COH.O hydrogen bonds with aspartates from one side whereas the

R configuration makes the same from the opposite side The conformation of isostere in the case of OE seems to be more conformationally flexible The preferred conformation of OE inhibitor is supported by the intramolecular hydrogen bond CO…HN (P 2 – P 1 ¢) However, more conformation states of OE isostere fit the same map of electron density (Fig 9) Rotations of benzyl groups in the protein pockets P 1 , P 1 ¢ and P 3 ¢ compensate the stress imposed by different geometries of the H-bond network.

Table 4 Torsion and dihedral angles (degrees) describing conformation of isostere in inhibitors OE, RE and SE in complex with HIV-1 protease Diagram shows schematically the structure of inhbitors and measured torsion angles X ¼ H; (R) OH and (S) OH for OE, RE and SE inhibitors, respectively.

Inhibitor

1.

N 1 -CA 1 -C 1 -C 2

2.

CA 1 -C 1 -C 2 -N 2

3.

C 1 -C 2 -N 2 –CA 2

4.

C 2 -N 2 –CA 2 -C 3

N 1 -CA 1 -C 1

CA 2 -C 3 -N 2

from the catalytic aspartates at the bottom of the active site

tunnel The shift of inhibitor OE main chain upwards

propagates to P2¢ Glu However, the Oe1 and Oe2 of

inhibitors RE and OE form already similar hydrogen bond

pattern to the Asp29 and Asp30 in both cases Thus, even if

the backbones of these inhibitors do not follow the same

line, the hydrogen bond pattern remains similar, showing

high flexibility and adaptation of not only the inhibitor to

the protease but also of the protease to the inhibitor

Comparison of inhibitors side chains

Boc groups at P2of SE and RE inhibitors lie in the same

position Only in the case of inhibitor OE, the Boc carbonyl

is significantly rotated to form an intramolecular hydrogen

bond (with O…N distance 3.6 A˚) to NH group in P1¢

position This results also in significant shifts of the

tert-butyl group by 0.3 and 0.6 A˚ when compared with SE

and RE inhibitors

Phenylalanine residues in position P1.The orientation of

benzyl group in P1position of the SE inhibitor differs from

those in OE and RE Both, Fig 10 and Table 6 show that

the rotation of SE benzyl group leads to a lower number of

contacts to protease – compare 12 contacts in SE with 17

and 16 contacts in OE and RE, respectively (contact is

defined here as an interatomic distance lower than 4.1 A˚)

Phenylalanine residues in position P1¢ The Ca of OE is

shifted higher to the flaps than Caof SE and RE inhibitors

(about 0.5 A˚ in both cases) Although phenyl rings of OE

and RE have very similar positions and orientation and

have similar contact patterns (26 and 31 contacts,

respect-ively), the P1¢ phenyl ring of SE is turned down to make less

contacts (19) to protease, namely to the flap residues in

comparison with OE and RE inhibitors

Glutamate residues in position P2¢ In comparison with RE

and SE, the Caat P02in OE is shifted upwards to the flaps

and also closer towards the symmetry axis of protease

However, the Glu Oe1and Oe2in OE and RE overlap and

make the same network of hydrogen bonds to the Asp29

and Asp30 with bond-length differences up to 0.3 A˚ The v3 torsion angle (C-C-C-OH) of Glu is for all inhibitors significantly different, namely in the case of RE (165,)54 and)179 for OE, RE and SE, respectively) This results in different hydrogen bonds between the Glu at P2¢ and protein binding site, although their numbers (6, 7, 7, see Table 6) remain unchanged

Phenylalanine residues in position P3¢ Benzyl groups of Phe at P3¢ are partially exposed to the solvent in all cases However, their positions and orientations are significantly different for the compared inhibitors After the Ca super-position, the phenyl rings of different inhibitors were found rotated to each other The angles between plains of P3¢ phenyl rings are 98 for OE and SE inhibitors, and 38 for

OE and RE inhibitors The S3¢ groups of OE and RE inhibitors having a similar orientation phenyl rings form a similar number of close contacts to protease (22 and 20 short C-C contacts for OE, RE, respectively), whereas in the case of SE inhibitor, there are 13 short C-C contacts to the protease only (Table 6)

Discussion

The experimentally determined structure of a complex of HIV-1 protease with OE, RE and SE inhibitors allowed us

to answer the following general questions: (a) why the presence of hydroxyl group in the isostere of an inhibitor (often referred to as necessary replacement of the catalytic water molecule in reaction intermediate) is not necessary for good inhibition properties of inhibitor (b) why the R or S configuration at the carboxyl of the isosteric group has no influence on the inhibition constant, and (c) what is the influence of small chemical changes in the inhibitor molecule

on its conformation in the binding tunnel of HIV protease

In answer to the first question, it was shown, that the hydroxyl binding to catalytic aspartates considered originally

as the main condition for good inhibition properties of substrate-mimicking inhibitors can be easily replaced by that

of the isostere NH group (if not present as this is the case of OE) Some energy loss caused by a less dense hydrogen bond network in place of the missing hydroxyl is probably

Table 5 Review of interactions of native HIV-1 protease with inhibitors OE, RE, SE analysed residue by residue Comparison of protease residues which are in contact with individual side chains of inhibitors OE, RE and SE All contacts up to 0.41 nm are accounted for Number of contacts to certain residue is given.

compensated by increased flexibility of the central part of the

OE inhibitor leading to a favorable entropy contribution

In answer to the second question, the negligible difference

between inhibition constants of RE and SE can be explained

by the fact that the strain imposed on inhibitor during

docking into the binding site of protease does not allow

correct orientation of the hydroxyl group to catalytic

aspartates either in S or in the R configuration The

resulting orientation of the hydroxyl group deviates from

the most convenient direction similarly, but in opposite

directions In other words, inhibitor remains halfway as far

as correct orientation in both cases is concerned

In answer to the final question, the X-ray structures

discussed in this paper showed that some even small changes

in inhibitor chemistry can have a large influence on

conformation of inhibitor main chain Thus, different

inhibitors differ not only in the binding affinity, but also

in the degree of freedom of the inhibitor inside the binding

tunnel Some special inhibitors (such as the OE discussed

here) can find more conformations in the binding tunnel of

the HIV protease with comparable interaction energies

This can form a significant contribution to the stability of

the complex through the entropy term in the average energy

of the system

Structure determination by X-ray crystallography shows

why theoretical predictions of inhibitor properties have been

relatively unreliable so far [21] Even small changes in the

inhibitor chemistry, no matter whether they have a high or

small effect on inhibition constant, result often in similarly small differences in the overall geometry of the inhibitor inside the binding tunnel Relatively small shifts of Caatoms (about 0.5 A˚) and rotations of side chains can significantly re-form a network of hydrogen bonds, which is, of course, compensated by a change of the inner torsion energy of chains destabilizing the complex Several water molecules always taking part in the formation of hydrogen bond network increase the number of possible configurations and conformation states of the complex Calculation of energy differences between the states containing different number

of atoms and their configurations is a difficult task because this requires an exact mutual scaling of all the energy contributions including the entropy and hydrophobic effects A good check for selection of a good inhibitor conformation among many others theoretically possible is

to verify which of them fit at least approximately the experimental map of electron density that can be nowadays easily calculated for any structure deposited in the PDB database with its experimentally measured intensities

Acknowledgements

The work was supported by the Grant Agency of the Czech Republic (projects 203/98/K023, 204/00/P091, 203/00/D117) and the Grant Agency of the Academy of Sciences of the Czech Republic (projects KJB4050312, A4050811, AVOZ4050913) The authors thank the beamline BM29 staff at ESRF in Grenoble for providing beam time.

Table 6 Comparison of hydroden bonds (2.4–3.6 A˚) of inhibitors OE, RE and SE towards their HIV-1 proteases Stated numbers are distances between inhibitor and protease atoms in A˚.