The molecular basis for high resistance to clinical inhibitors of HIV-1 protease (PR) was examined for the variant designated PRP51 that was selected for resistance to darunavir (DRV). High resolution crystal structures of PRP51 with the active site D25N mutation revealed a ligand-free form and an inhibitor-bound form showing a unique binding site and orientation for DRV. This inactivating mutation is known to increase the dimer dissociation constant and decrease DRV affinity of PR. The PRP51-D25N dimers were in the open conformation with widely separated flaps, as reported for other highly resistant variants. PRP51-D25N dimer bound two DRV molecules and showed larger separation of 8.7 Å between the closest atoms of the two flaps compared with 4.4 Å for the ligand-free structure of this mutant. The ligand-free structure, however, lacked van der Waals contacts between Ile50 and Pro81' from the other subunit in the dimer, unlike the majority of PR structures. DRV is bound inside the active site cavity; however, the inhibitor is oriented almost perpendicular to its typical position and exhibits only 2 direct hydrogen bond and two water-mediated interactions with atoms of PRP51-D25N compared with 11 hydrogen bond interactions seen for DRV bound in the typical position in wild-type enzyme. The atypical location of DRV may provide opportunities for design of novel inhibitors targeting the open conformation of PR drug-resistant mutants.
The molecular basis for high resistance to clinical inhibitors of HIV-1 protease (PR) was examined for the variant designated PRP51 that was selected for resistance to darunavir (DRV). High resolution crystal structures of PRP51 with the active site D25N mutation revealed a ligand-free form and an inhibitor-bound form showing a unique binding site and orientation for DRV. This inactivating mutation is known to increase the dimer dissociation constant and decrease DRV affinity of PR. The PRP51-D25N dimers were in the open conformation with widely separated flaps, as reported for other highly resistant variants. PRP51-D25N dimer bound two DRV molecules and showed larger separation of 8.7 Å between the closest atoms of the two flaps compared with 4.4 Å for the ligand-free structure of this mutant. The ligand-free structure, however, lacked van der Waals contacts between Ile50 and Pro81' from the other subunit in the dimer, unlike the majority of PR structures. DRV is bound inside the active site cavity; however, the inhibitor is oriented almost perpendicular to its typical position and exhibits only 2 direct hydrogen bond and two water-mediated interactions with atoms of PRP51-D25N compared with 11 hydrogen bond interactions seen for DRV bound in the typical position in wild-type enzyme. The atypical location of DRV may provide opportunities for design of novel inhibitors targeting the open conformation of PR drug-resistant mutants.
HIV-1 protease (PR) has been a successful target
in AIDS therapy
due to its critical role in viral maturation by hydrolyzing the Gag
and Gag-Pol precursor polyproteins into mature structural and functional
proteins.[1,2] A series of clinical HIV-1 protease inhibitors
(PIs) has improved the survival of AIDSpatients. One such inhibitor,
darunavir (DRV), which was designed to target drug resistance by introducing
strong polar interactions with the main chain atoms of the PR,[3−5] has been widely used for the treatment of drug-naïve patients
and those infected with multidrug-resistant HIV-1.[6] DRV effectively inhibits PR enzymatic activity with picomolar
binding affinity assessed by isothermal titration calorimetry (ITC).[7] However, HIV evolves resistance to DRV by selecting
a combination of mutations.[6]Highly
DRV-resistant HIV-1 variants were selected in the laboratory
to elucidate the mechanism for resistance.[8] A mixture of 8 highly DRV-susceptible HIV-1 clinical isolates (HIV-1MIX) containing 9–14 PI-resistant mutations was propagated
in the presence of DRV. The viral population at passage 51 (HIV-1MIXP51) replicated well at the concentration of
5 μM DRV, and sequencing revealed 14 amino acid substitutions
in the PR gene (Figure 1).[8] The viral strain HIV-1MIXP51 was
highly resistant, with half maximal effective concentration (EC50) for inhibition of viral replication increased to more than
1 μM for DRV and most other PIs, and showed moderate resistance
to saquinavir (SQV) (0.3 μM EC50).[8]
Figure 1
PRP51 mutations. (A) Sites of the 14 drug resistant
mutations mapped onto the PRP51 dimer (cyan cartoon representation).
The mutations located in the active site cavity are shown as red spheres,
while the flap mutations are shown as blue spheres, and the mutations
distal from the active site are indicated as green spheres. (B) Amino
acid sequence of HIV-1 PR (upper line) and PRP51 (lower
line). The amino acids are colored as in panel A. Note that the wild-type
PR sequence used for structural comparison includes mutations Q7K,
L33I, and L63I to prevent autoproteolysis, and both proteins include
C67A and C95A to eliminate potential cysteine-thiol oxidation.
PRP51 mutations. (A) Sites of the 14 drug resistant
mutations mapped onto the PRP51 dimer (cyan cartoon representation).
The mutations located in the active site cavity are shown as red spheres,
while the flap mutations are shown as blue spheres, and the mutations
distal from the active site are indicated as green spheres. (B) Amino
acid sequence of HIV-1 PR (upper line) and PRP51 (lower
line). The amino acids are colored as in panel A. Note that the wild-type
PR sequence used for structural comparison includes mutations Q7K,
L33I, and L63I to prevent autoproteolysis, and both proteins include
C67A and C95A to eliminate potential cysteine-thiol oxidation.We have investigated the physical
and biochemical properties of
several resistant variants, including the HIV-1MIXP51 protease (PRP51).[9] PRP51 and another highly resistant variant with 20 mutations
(PR20) showed several extreme properties contributing to resistance.
The affinity of DRV and SQV for PRP51 as measured by isothermal
titration calorimetry (ITC) gave KL values
of 37 and 54 nM, respectively, or about 7400-fold and 135-fold weaker
than the corresponding values for wild-type PR. Autocatalytic cleavage
(autoprocessing) of precursor Gag-Pol polyprotein is essential to
produce mature and fully active PR.[10] Autoprocessing
of the precursor comprising the 56-amino-acid transframe region (TFR)
fused to PR (TFR-PR) was inhibited the best by DRV and SQV as compared
to other clinical inhibitors, although in the low micromolar range.However, autoprocessing of the TFR-PRP51 precursor was
uninhibited by DRV and marginally inhibited by SQV, at 150 μM
PI concentration.[9] These properties of
PRP51 are consistent with the high antiviral resistance
to DRV measured for virus bearing this variant (>300-fold increased
EC50) relative to wild-type.[8]We have reported recently the structural analysis of another
extremely
resistant variant, PR20, which showed no inhibition of precursor autoprocessing,
and KL values for mature PR20 increased
by more than 8000- and 2000-fold, respectively, for DRV and SQV.[9,11] Crystallographic analysis of PR20 showed fewer interactions with
inhibitors and widely separated flaps in the absence of inhibitors
with no contact of the flap with residues from the other subunit,
unlike interactions in the corresponding structures of wild-type PR.[12,13] An increased separation of the flaps in the dimer may be typical
of many resistant variants, as seen also for variant MDR769[14] and in solution studies using double electron–electron
resonance (DEER) spectroscopy.[15]In order to further investigate the molecular mechanisms of high-level
resistance, we have determined the crystal structures of PRP51 bearing the inactivating mutation D25N (PRP51-D25N) to abolish self-degradation (autoproteolysis) for sample handing
during crystallization. When the D25N mutation was introduced into
wild-type PR, the affinity for DRV was decreased by about 106-fold, while no substantial changes were observed in the crystal
structures.[16] Two structures were obtained
for PRP51-D25N: a DRV-bound structure (PRP51-D25N/DRV) and a ligand-free structure (PRP51-D25N).
Structural analysis revealed an unusual binding site for DRV and the
widely separated flaps that characterize the ligand-free structures
of many highly resistant variants.[11,14,15]
Results and Discussion
Crystal Structures of Ligand-Free
and DRV-Bound PRP51-D25N
Repeated attempts
with active PRP51 did not
yield crystals, likely due to its enhanced autoproteolysis. Consistent
with this observation even storage of the protein in 12 mM HCl prior
to folding showed degradation products unlike the optimized wild-type
PR. The optimized wild-type PR bears the mutations L33I and L63I shown
to significantly restrict autoproteolysis of wild-type PR in addition
to Q7K, which exists in PRP51.[10,17] Also, as DRV interactions with PRD25N are nearly identical
to those in wild-type PR/DRV complex,[16] we resorted to using PRP51 with mutation D25N for our
studies.Crystals were grown of mutant PRP51-D25N in the presence of DRV, SQV, tipranavir (TPV), and amprenavir (APV)
in order to identify the structural changes associated with high level
resistance. Electron density for inhibitor DRV was observed within
the ligand binding cavity only in the structure of PRP51-D25N/DRV. No inhibitor was visible in the PRP51-D25N structure obtained from crystals grown in the presence of SQV, APV,
or TPV, consistent with the high level resistance of this mutant.
The two crystal structures designated PRP51-D25N/DRV and PRP51-D25N were refined with X-ray data
at resolutions of 1.66 and 1.50 Å and R-factors of 18.9% and
15.9%, respectively. The crystallographic statistics are listed in Supplementary Table S1. PRP51-D25N/DRV was refined in space group P41212 with a monomer of residues numbered 1–99 in the asymmetric
unit, while PRP51-D25N was refined in space group P41 and contained one dimer of residues numbered
1–99 and 1′–99′ in the asymmetric unit.
Alternate conformations were refined for 3 residues in PRP51-D25N/DRV and 7 residues in PRP51-D25N structures. Residues
34–36 from each monomer of the two structures showed similar
alternate conformations with 0.5 relative occupancy. The crystallographic
dimer of PRP51-D25N/DRV was generated for structural
analysis.
PRP51-D25N/DRV Exhibits Unusual Binding Conformation
of DRV
DRV placed at the typical binding site (Figure 2A)[5,18] did not fit the electron density
visible in the ligand binding cavity of PRP51-D25N/DRV, and thus different locations were evaluated for the inhibitor.
After manual adjustment, the final monomer structure was refined with
one full occupancy molecule of DRV oriented approximately perpendicular
to DRV in the typical inhibitor-binding site of wild-type PR or most
mutants (Figure 2B), as indicated by the clear
electron density map (Figure 3A). The two molecules
of DRV are related by 180° rotation and interact with each other
in the binding cavity of the PRP51-D25N dimer (Figure 3B). The polar interactions between the two DRV molecules
include a water-mediated hydrogen bond connecting a sulfonyl oxygen
and hydroxyl group with the amino group on the aniline of the other
DRV and C–H···water interactions with the phenyl
group and the bis-THF of both DRVs. The bound DRV has relatively few
hydrogen bonds and many hydrophobic contacts with the protein (Figure 3C and D).
Figure 2
DRV binds to different sites in PR and PRP51-D25N. (A) The structure of wild-type PR (gray color)
with one conformation
of DRV (red sticks) (PDB ID 2IEN) bound perpendicular to the dimer interface. (B) A
different position and conformation was seen for the two molecules
of DRV (red sticks and pink sticks) bound symmetrically in the PRP51-D25N dimer (green color). On the right side, the
chemical structure of DRV is indicated in approximately the same orientation
for the two structures.
Figure 3
Unique binding site for DRV in PRP51-D25N. (A) Fo – Fc omit
map contoured at 2.0σ showing one molecule of DRV (yellow sticks)
in the crystal structure of the monomer of PRP51-D25N/DRV. (B) Interaction between two DRV molecules in the PRP51-D25N dimer (yellow sticks and cyan sticks). Water molecules are shown
as red spheres. The dotted lines represent the hydrogen bonds, and
the dashed lines represent potential CH···O interactions
with the interatomic distances indicated in Å. (C,D) Interactions
of DRV with PRP51-D25N (green sticks). The interactions
of DRV are separated into panels C and D for clarity. The interacting
residues are labeled, and mutations are indicated by underlining.
The dotted lines show the hydrogen bond interactions, and the double-sided
arrows represent the van der Waals contacts. The number of van der
Waals contacts is indicated.
DRV binds to different sites in PR and PRP51-D25N. (A) The structure of wild-type PR (gray color)
with one conformation
of DRV (red sticks) (PDB ID 2IEN) bound perpendicular to the dimer interface. (B) A
different position and conformation was seen for the two molecules
of DRV (red sticks and pink sticks) bound symmetrically in the PRP51-D25N dimer (green color). On the right side, the
chemical structure of DRV is indicated in approximately the same orientation
for the two structures.Unique binding site for DRV in PRP51-D25N. (A) Fo – Fc omit
map contoured at 2.0σ showing one molecule of DRV (yellow sticks)
in the crystal structure of the monomer of PRP51-D25N/DRV. (B) Interaction between two DRV molecules in the PRP51-D25N dimer (yellow sticks and cyan sticks). Water molecules are shown
as red spheres. The dotted lines represent the hydrogen bonds, and
the dashed lines represent potential CH···O interactions
with the interatomic distances indicated in Å. (C,D) Interactions
of DRV with PRP51-D25N (green sticks). The interactions
of DRV are separated into panels C and D for clarity. The interacting
residues are labeled, and mutations are indicated by underlining.
The dotted lines show the hydrogen bond interactions, and the double-sided
arrows represent the van der Waals contacts. The number of van der
Waals contacts is indicated.This DRV molecule has a different conformation and interactions
from those for the regular binding mode of DRV, which has 11 direct
hydrogen bond interactions and 4 water-mediated ones with PR in the
wild-type PR/DRV complex (PDB ID 2IEN) (Figure 4A and 4B).[5] Many van der Waals
contacts were observed between PRP51-D25N and DRV
with distances ranging from 3.8 to 4.2 Å. In the crystal structure
of PRP51-D25N/DRV, DRV forms two direct hydrogen
bonds with the main chain amides of Gly49 and Gly50, and three water
molecules mediate hydrogen bond interactions with Gly27, Asp30, and
Gly27′ from the crystallographic dimer. van der Waals interactions
occurred between DRV and PRP51-D25N residues Asn25,
Ala28, Ile47, Gly48, Gly49, Ile50, Gly52, Phe53, Met54, Thr80, Pro81′,
Ile82, and Val84. These interacting residues include mutations of
V32I, I54M, V82I, and I84V from the selected P51 isolate and the D25N
mutation. It is probable that these mutations facilitate binding of
DRV to this atypical site. Introducing the single mutation D25N in
the wild-type enzyme, however, does not alter the binding of DRV observed
in the crystal structure.[16] In addition,
DRV forms interactions with residues from other symmetry-related molecules
of PRP51-D25N as shown in Figure 4B and Supplementary Figure S1.
The DRV molecule interacts with Ile72 and Gln61 from two different
symmetry-related monomers of PRP51-D25N (Supplementary Figure S1). The interactions include
two direct hydrogen bonds between the two oxygens of the bis-THF and
the amino group on the side chain of Gln61 and van der Waals contacts
with Ile72 and Gln61. The existence of DRV interactions with the symmetry-related
PRP51-D25N raises the possibility that crystal lattice
contacts influence the atypical binding of the inhibitor.
Figure 4
DRV has distinct
hydrogen bond interactions with wild-type PR (PDB
ID 2IEN) (A)
and PRP51-D25N (B). DRV is represented in gray sticks
and yellow sticks in PR and PRP51-D25N, respectively.
The dotted lines indicate hydrogen bond interactions. Water molecules
are represented by W. NH is the main chain amide group, and C=O
is the main chain carbonyl group. Glu61 from a symmetry-related dimer
is shown in blue within parentheses.
DRV has distinct
hydrogen bond interactions with wild-type PR (PDB
ID 2IEN) (A)
and PRP51-D25N (B). DRV is represented in gray sticks
and yellow sticks in PR and PRP51-D25N, respectively.
The dotted lines indicate hydrogen bond interactions. Water molecules
are represented by W. NH is the main chain amide group, and C=O
is the main chain carbonyl group. Glu61 from a symmetry-related dimer
is shown in blue within parentheses.The question of whether DRV might bind at this atypical site
in
wild-type enzyme can be addressed by structural comparison. The PRP51-D25N/DRV monomer shares an almost identical wide
open conformation (RMSD of 0.33 Å on equivalent Cα atoms)
with the wild-type PR crystallized with Mg2+ coordinated
at the active site (PDB ID 2PC0).[13] Superposition of the
two monomers reveals that three side chains of the wild-type enzyme
lie close to DRV in this atypical site (Figure 5). The side chain of Ile54 in wild-type PR would hinder binding of
the second THF of the bis-THF group of DRV due to short interatomic
distances of 2.0–3.1 Å (Figure 5). Also, the side chains of Val82 and Ile84 in wild-type PR extend
closer to the aniline group of DRV compared to the good hydrophobic
contacts formed by Ile82 and Val84 in PRP51-D25N/DRV. However, rotated conformers of the side chains of Ile54, Val82,
and Ile84 could allow the possible binding of DRV in the atypical
site.
Figure 5
Superposition of the monomers of PRP51-D25N/DRV
(green) and wild-type PR (PDB ID 2PC0) (grey). The double-sided arrows represent
short interatomic contacts of 2.0–3.1 Å between DRV (yellow)
and side chains of Ile54, Val82, and Ile84 in wild-type PR.
Superposition of the monomers of PRP51-D25N/DRV
(green) and wild-type PR (PDB ID 2PC0) (grey). The double-sided arrows represent
short interatomic contacts of 2.0–3.1 Å between DRV (yellow)
and side chains of Ile54, Val82, and Ile84 in wild-type PR.The affinity of DRV for binding
in the typical location is expected
to be decreased drastically by the mutations. The PRP51-D25N variant showed 7,400-fold lower affinity for DRV,[9] while introducing the single D25N mutation in the wild-type
enzyme produced about 106-fold decreased affinity for DRV
as measured by ITC with no change in the stoichiometry of binding.[16] Therefore, the combination of D25N plus the
14 mutations in PRP51 is expected to compromise the affinity
to low micromolar levels.[16] Consequently,
DRV may favor the weaker atypical binding location observed in PRP51-D25N.It is not the first time that DRV has
been observed to bind at
an unusual site in HIV-1 protease variants. A second binding site
was observed in an ultrahigh resolution (0.84 Å) structure of
PRV32I/DRV in addition to the typical active site binding.
The second DRV bound in a groove on one flap surface where the residues
Glu35′, Lys45′, Lys55′, Val56′, and Arg57′
participated in the major interactions.[18] A similar DRV binding site occurred in the crystal structures of
PRM46L/DRV and PR20/DRV.[11,18] Also for SQV,
a second molecule was found in a location adjacent to the usual active
site location in PR20/SQV and PRV32I/I47V/V82I/SQV structures.[11,19] Another unique binding mode was found for GB-18, specifically [3-cobalt
bis(1,2-dicarbollide)]-ion, which belongs to a novel class of inorganic
cobaltacarborane inhibitors, in the active site cavity of the wild-type
PR (PDB ID 1ZTZ).[20] This binding mode involves two molecules
of GB-18 positioned asymmetrically inside the pseudosymmetric active
site cavity.[20] These binding pockets for
GB-18 are formed by the residues Pro81, Ile84, and Val82 and covered
by the flap residues Ile47, Gly48, and Ile54 forming a semiopen conformation
of the PR dimer.[20] The cobaltacarborane
complex was used to design novel inhibitors with linkers between two
metallocarboranes to permit flexible binding to drug-resistant mutants.
A similar strategy can be evaluated for inhibiting the open conformation
dimer by chemically linking two DRV molecules as seen in the PRP51-D25N/DRV complex.
Flaps of PRP51-D25N/DRV and PRP51-D25N Display Different Intersubunit
Interactions
Both DRV-bound
and ligand-free PRP51-D25N dimers have flaps separated
by a large distance between their tips. The flaps of PRP51-D25N/DRV dimer had a larger separation of 8.7 Å between the closest
atoms, while the flaps were separated by 4.4 Å for PRP51-D25N. Superposition of the monomer of PRP51-D25N/DRV
with each subunit of the ligand-free dimer gave the overall RMSD value
of 0.97 and 0.41 Å on Cα atoms, respectively, with large
differences in the conformation of the two flaps and 80′s loops
(residues 79/79′ to 83/83′) (Figure 6A). One of the two flaps (residues 47/47′ to 54/54′)
of the PRP51-D25N/DRV dimer is further away from
the catalytic site than seen for the equivalent flap of PRP51-D25N as indicated by the distance of 5.6 Å between the equivalent
Cα atoms of Ile50 in these two structures, while the other flap
conformation is more similar in the two structures with 1.5 Å
distance between the Cα atoms of Ile50′ residues (Figure 6B).
Figure 6
Comparison of two structures of PRP51-D25N. (A)
Superposition of the overall structures of PRP51-D25N (pink) and PRP51-D25N/DRV (green). (B) The interactions
of flap residues 47–54 are shown in the blue box; the numbers
beside the black arrows show the distance in Å between corresponding
Cα atoms of Ile50/50′ in the two structures.
The two PRP51-D25N dimer
structures were compared with open conformation structures of PR and
PR20 as well as their DRV-bound complexes. The open conformations
of ligand-free PR20 have two unusual features relative to other reported
dimer structures: widely separated flaps and no intersubunit van der
Waals contacts between the flap tip and residues from the other subunit.[11] Therefore, the conformations were assessed by
measuring the closest distance between atoms at the tip of the two
flaps within one dimer structure and the closest intersubunit contact
of Ile50/50′ at the flap tip with Pro81′/81. The shortest
interatomic distances between the flap tips were 4.0 Å in ligand-free
PRP51-D25N, compared to 3.0 Å in a typical
open conformation of wild-type PR (PDB ID 1HHP), 7.7 Å in another wild-type PR
with Mg2+ coordinated at the active site (PDB ID 2PC0), and 6.0 Å
in ligand-free PR20 (PDB ID 3UF3).[11−13]Comparison of two structures of PRP51-D25N. (A)
Superposition of the overall structures of PRP51-D25N (pink) and PRP51-D25N/DRV (green). (B) The interactions
of flap residues 47–54 are shown in the blue box; the numbers
beside the black arrows show the distance in Å between corresponding
Cα atoms of Ile50/50′ in the two structures.The various flap conformations in wild-type PR
(PDB ID 1HHP), ligand-free PRP51-D25N, and ligand-free PR20
are compared in Supplementary Figure S2A. For the DRV complexes,
the shortest interflap distance was 8.7 Å in PRP51-D25N/DRV compared with about 3.3 Å for typical closed conformation
dimers of PR/DRV (PDB ID 2IEN) and PR20/DRV (PDB ID 3UCB) (Supplementary Figure
S2B).[5,11] Ile50/50′ at the flap
tips has intersubunit van der Waals contacts of about 4.0 Å with
Pro81′/81 in PRP51-D25N/DRV as observed for
closed conformation inhibitor-bound dimers. The majority of dimers
show intersubunit van der Waals contacts between Ile50/50′
and Pro81′/81, except for the ligand-free PR20 where these
two side chains are separated by about 7 Å. In contrast, this
intersubunit contact was lost in ligand-free PRP51-D25N since the closest atoms of Ile50 and Pro81′ were separated
by 8.5 Å. The asymmetric flap conformations of ligand-free PRP51-D25N resemble those of another open conformation
dimer of PR20 (PDB ID 3UHL), which had van der Waals contact between the flap
tips and asymmetric flaps with 12.2 and 5.4 Å intersubunit separation
between Ile50/50′ and Pro81′/81 (Figure 7). Therefore, the large separation (∼7–12 Å)
between side chain atoms of Ile50 and Pro81′ from the other
subunit is conserved in the ligand-free dimers of the two highly resistant
mutants PR20 and PRP51-D25N.
Figure 7
Comparison of dimers
of PRP51-D25N and PR20/p2-NCopen (PDB
ID 3UF3). (A)
Superposition of the overall structures of PRP51-D25N (pink) and PR20/p2-NCopen (PDB ID 3UHL, blue). (B) The
flap residues 47–53 and Pro81 of PRP51-D25N and PR20/p2-NCopen are shown below. The numbers beside
the black arrows show the distance in Å between Ile50 and Pro81′.
Comparison of dimers
of PRP51-D25N and PR20/p2-NCopen (PDB
ID 3UF3). (A)
Superposition of the overall structures of PRP51-D25N (pink) and PR20/p2-NCopen (PDB ID 3UHL, blue). (B) The
flap residues 47–53 and Pro81 of PRP51-D25N and PR20/p2-NCopen are shown below. The numbers beside
the black arrows show the distance in Å between Ile50 and Pro81′.
Multiple Mutations Contribute
to the Structural Changes in PRP51-D25N
PRP51-D25N bears
14 mutations (L10I, I15V, K20R, L24I, V32I, L33F, M36I, M46L, I54M,
L63P, K70Q, V82I, I84V, and L89M) as well as D25N relative to the
standard wild-type PR sequence. Four mutations associated with drug
resistance, V32I, M46L, V82I, and I84V, alter residues in the active
site cavity where substrates and inhibitors usually bind. Additionally,
“second shell” mutations L10I, L24I, L33F, and I54M
alter residues that form direct interactions with residues in the
active site cavity. Our crystallographic and biochemical analysis
has demonstrated the changes due to the individual single mutations
of L24I, V32I, M46L, I54M, and I84V, as reviewed in Weber and Agniswamy.[2] Several of the other substitutions in PRP51-D25N are shared by the highly resistant multiple
mutant PR20, and their coordinated effects have been described previously.[11]The mutations of residues V32I, V82I,
and I84V in the active site cavity are assumed to contribute to the
poor affinity of the PRP51 mutant for inhibitors and the
observed unusual binding site for DRV in PRP51-D25N (Figure 3). The side chains of these residues
form hydrophobic interactions with each other and with Val47 in all
structures. Mutants PRV32I and PRI47V have altered
interactions with inhibitors DRV and SQV;[2,11] however,
V82I in the triple mutant PRV32I/I47V/V82I bearing the
active site residues of HIV-2 PR does not significantly alter direct
contacts with inhibitor.[19] Further comparison
of inhibitor interactions is limited since the crystal structures
of the PRP51-D25N mutant are in the open conformation
without inhibitor bound in the standard active site location.Mutations M46L and I54M alter residues in the flaps and are proposed
to have small indirect effects on inhibitor binding and may alter
the flap dynamics.[18,21,22] In PRP51-D25N/DRV, the I54M mutation introduced
new van der Waals interactions with DRV (Figures 3C and 5), while M46L had no contacts
with the ligand. The changes in interactions of mutated residue L24I
are similar in PRP51-D25N and the single mutant
PRL24I.[23] Ile24 gains two good
van der Waals contacts with Leu90 and loses or elongates contacts
(4.7 and 4.2 Å in the two subunits compared to 3.8 Å distance
in wild-type PR) with Phe99′ from the other subunit. Unlike
the mutated side chain of L10F in PR20/DRV, in which a new hydrophobic
contact was formed between the side chains of Phe10 and mutated Ile82,[11] mutated residue L10I in PRP51-D25N yields no new interactions with nearby residues. Mutation L33F introduces
the large bulky Phe side chain, which maintains hydrophobic contacts
of the wild-type enzyme, including contacts with mutated residues
I15V, M36I substituting shorter side chains as reported for L33F in
PR20.[11] Also, the flap hinge region comprising
residues 34 to 43 shares a very similar conformation in PRP51-D25N and in the ligand free conformation of PR20, which is likely due
to the presence of mutations M36I and I33F in both highly drug-resistant
variants. The mutations in the flap hinge and flaps are likely to
contribute to the extended flap conformations observed in PRP51-D25N and in inhibitor-free PR20 structures.Mutation L89M has not
been analyzed previously in structures. The
side chain of Leu89 in wild-type PR forms hydrophobic contacts with
the side chains of Ile64, Ile66, Ala71, Gly73, Ile85, and Asn88, as
well as van der Waals and C–H···O interactions
with the side chain of Thr31 (Figure 8). The
mutated Met89 forms very similar contacts in PRP51-D25N, except for an additional van der Waals contact with the side chain
of Val75. Mutation K20R alters a residue near the protein surface
showing varied interactions with other surface side chains. Mutations
L63P and K70Q also alter surface side chains that form a hydrophobic
contact in the wild-type PR, which is eliminated in the mutant PRP51-D25N.
Figure 8
Interactions of Leu89 in PR/DRV (gray) (PDB
ID 2IEN) and
Met89 in PRP51-D25N/DRV (green) with neighboring
residues. The
double-sided arrows and the dashed lines represent the van der Waals
contacts and putative C–H···O interactions,
respectively, and are colored gray and green to match the structures
when differences occur, while black lines indicate identical numbers
of contacts in both structures. The number of van der Waals contacts
is indicated by the double-sided arrows.
Interactions of Leu89 in PR/DRV (gray) (PDB
ID 2IEN) and
Met89 in PRP51-D25N/DRV (green) with neighboring
residues. The
double-sided arrows and the dashed lines represent the van der Waals
contacts and putative C–H···O interactions,
respectively, and are colored gray and green to match the structures
when differences occur, while black lines indicate identical numbers
of contacts in both structures. The number of van der Waals contacts
is indicated by the double-sided arrows.
Conclusions
Two crystal structures were analyzed for
the PRP51-D25N variant that was selected for high
levels of resistance to DRV. These structures confirm the increased
separation of the two flaps in the dimer and/or loss of intersubunit
contacts between the flap tip and Pro81 in comparison to the open
conformations seen for the ligand-free wild-type PR, as previously
described for the MDR769[14] and PR20[11] highly resistant variants. Other highly resistant
PR mutants have shown greater flap mobility in studies employing various
techniques[11,14,15,24] and decreased interactions with inhibitors
or substrate analogues in crystal structures.[11,25,26] In PRP51-D25N, however,
DRV showed a unique mode of binding within the open conformation flaps
and lying almost perpendicular to the typical active site position.
Importantly, this new binding site for DRV may hint at designs for
novel types of antiviral inhibitors that capture the open, inactive
conformation of the protease.
Methods
Construction,
Expression, and Purification
The HIV-1
PR from Group M (Genbank HIVHXB2CG) is designated as PR. The PRP51 construct contains 14 mutations (L10I, I15V, K20R, L24I,
V32I, L33F, M36I, M46L, I54M, L63P, K70Q, V82I, I84V, and L89M) plus
three other mutations Q7K to minimize autoproteolysis and C67A and
C95A to prevent cysteine-induced aggregation.[9,10,27] The mutant DNA was constructed by oligonucleotide
synthesis and cloned into the pET11a vector between Nde1 and BamH1
restriction sites. To eliminate autoproteolysis, the inactivating
mutation of D25N was introduced using the QuikChange II Site-Directed
Mutagenesis Kit and confirmed by DNA sequencing. The protein was expressed
in Escherichia coli BL21(DE3), purified and folded
using the protocol described previously.[28,29]
Crystallization and Data Collection
Crystals of PRP51 (including the D25N mutation) complexed with clinical inhibitors
DRV and SQV were obtained by the hanging-drop vapor-diffusion method
at RT using 24 well VDX plates (Hampton Research, Aliso Viejo, CA,
USA). PRP51 with a monomer concentration of 1.29 mg mL–1 was mixed with the inhibitors at 5–10-fold
molar excess. Screening Kit I solutions (Hampton Research, Aliso Viejo,
CA, USA) gave good crystals of PRP51 complexed with DRV
(0.1 M HEPESsodium pH 7.5, 0.8 M potassium sodium tartrate tetrahydrate)
and crystals of PRP51-D25N grown in the presence
of SQV (0.1 M imidazole pH 6.5, 1.0 M sodium acetate trihydrate).
The crystals were frozen in liquid nitrogen using 25% (v/v) glycerol
as a cryoprotectant. X-ray diffraction data were collected at 100
K by remote access on the beamline BM-22 of the Southeast Regional
Collaborative Access Team (SER-CAT), the Advanced Photon Source, Argonne
National Laboratory, Chicago.
Data Processing and Structure
Determination
The X-ray
data were indexed, integrated, and scaled with HKL2000.[30] The structures were solved by molecular replacement
with MOLREP in the CPP4i suite of programs[31] using the PR20 complex with Yb+ (PDB ID 3UF3) as the starting
model.[11] The structures were refined by
REFMAC 5.2 in the CCP4 program suite 6.1.13[32] and refitted using COOT 0.6.1.[33] Alternate
conformations were modeled for PR residues, inhibitors, and solvent
molecules based on the observed electron density maps. Anisotropic
B factor refinement was applied for the higher resolution structure,
and TLS restrained refinement was used for the lower resolution structure.
Structural figures were prepared with PyMol.[34]
Protein Data Bank Accession Codes
The structure coordinates
and factors have been deposited in the Protein Data Bank with access
codes 4NPT for
PRP51-D25N-DRV and 4NPU for PRP51-D25N.
Authors: Fengling Liu; Peter I Boross; Yuan-Fang Wang; Jozsef Tozser; John M Louis; Robert W Harrison; Irene T Weber Journal: J Mol Biol Date: 2005-10-21 Impact factor: 5.469
Authors: Yunfeng Tie; Yuan-Fang Wang; Peter I Boross; Ting-Yi Chiu; Arun K Ghosh; Jozsef Tozser; John M Louis; Robert W Harrison; Irene T Weber Journal: Protein Sci Date: 2012-01-24 Impact factor: 6.725
Authors: A M Mildner; D J Rothrock; J W Leone; C A Bannow; J M Lull; I M Reardon; J L Sarcich; W J Howe; C S Tomich; C W Smith Journal: Biochemistry Date: 1994-08-16 Impact factor: 3.162
Authors: Chen-Hsiang Shen; Yu-Chung Chang; Johnson Agniswamy; Robert W Harrison; Irene T Weber Journal: J Mol Graph Model Date: 2015-09-08 Impact factor: 2.518
Authors: Florian Leidner; Nese Kurt Yilmaz; Janet Paulsen; Yves A Muller; Celia A Schiffer Journal: J Chem Theory Comput Date: 2018-04-18 Impact factor: 6.006
Authors: Zhanglong Liu; Xi Huang; Lingna Hu; Linh Pham; Katye M Poole; Yan Tang; Brian P Mahon; Wenxing Tang; Kunhua Li; Nathan E Goldfarb; Ben M Dunn; Robert McKenna; Gail E Fanucci Journal: J Biol Chem Date: 2016-08-30 Impact factor: 5.157
Authors: Ian W Windsor; Michael J Palte; John C Lukesh; Brian Gold; Katrina T Forest; Ronald T Raines Journal: J Am Chem Soc Date: 2018-10-22 Impact factor: 15.419