Download Molecular modeling of HIV-1 reverse

Protein Engineering vol.10 no.12 pp.1379–1383, 1997
Molecular modeling of HIV-1 reverse transcriptase drug-resistant
mutant strains: implications for the mechanism of polymerase
action
Marilyn B.Kroeger Smith1,*, Christopher J.Michejda1,
Stephen H.Hughes1, Paul L.Boyer1, Paul A.J.Janssen2,
Koen Andries2, Robert W.Buckheit, Jr3 and
Richard H.Smith, Jr1,4
A computer model of human immunodeficiency virus type
1 (HIV-1) reverse transcriptase (RT) either alone, or complexed with a non-nucleoside inhibitor (NNI), was constructed using crystal coordinate data from a subset of the
protein surrounding the binding pocket region. Molecular
mechanics calculations were carried out on solvated wildtype RT and RT that contained modifications corresponding
to resistance-engendering mutations. Results from the calculations revealed that the r.m.s. difference between 12
modified proteins and that of wild-type RT could be
qualitatively correlated with the measured polymerase
activity of the enzyme in the presence of these mutations.
In addition, the level of activity was related to the measured
distance between the primer grip and dNTP binding regions
of the protein. These data suggest a direct correlation
between RT structure and function. Complexes of RT–8C1 TIBO and RT–α-APA were also minimized in models
containing modifications corresponding to key drug-resistant mutants. The variant complexes all showed weaker
binding than wild-type RT, while giving rise to similar, but
critical changes in the protein. Therefore, the design of
new inhibitors should center on obtaining stronger binding
drugs to key drug-resistant RT variants.
Keywords: AIDS/computer/inhibitor/resistance/reverse transcriptase
nucleoside inhibitors (NNIRTs) results in the development of
resistance, although the specific mutants that are responsible
differ for the most part between the two classes of drugs
(Boyer et al., 1994a,b). In order to combat this problem,
combination therapy, which combines inhibitors from both
classes, is now being used as a means to enhance virus
suppression (Larder and Kemp, 1990; St Clair et al., 1991;
Larder, 1992; Tisdale et al., 1993). One such combination
involved administration of zidovudine (AZY) along with a
second nucleoside analog such as ddI (Iversen et al., 1996).
Alternative treatment strategies involve concomitant treatment
with either one or more nucleoside and/or non-nucleoside RT
inhibitors, two non-nucleoside drugs or RT inhibitors and
HIV-1 protease inhibitors. Selection of compounds for these
treatments should involve inhibitors that have different resistance spectra with the goal of obtaining escape mutants that
replicate poorly.
As an aid to understanding how mutations affect RT structure
and function, in both the presence and absence of nonnucleoside inhibitors, molecular modeling studies were undertaken on RT modified at specific residues. Assays to determine
polymerase activity of the mutant strains of RT showed that
while some of these strains had levels of activity comparable
to that of the wild-type enzyme, other mutant strains had much
lower levels of enzymatic activity. Correlation of structure
with enzymatic function in RT without inhibitor present may
lead to an explanation for the experimental findings. It has
been suggested that in the presence of NNIRTs, a deformation
of the polymerase active site occurs, locking it into an inactive
conformation (Ren et al., 1995). Presumably, mutation may
either restore the active site conformation through structural
realignment or could destabilize inhibitor binding. Therefore,
assessment of the structural adjustment resulting from these
local changes and the accompanying energetic consequences
may provide insight into the design of more effective RT
inhibitors.
Introduction
Reverse transcriptase (RT) of human virus type 1 (HIV-1) is
an important enzyme for drug therapy because it plays a
critical role in the viral replication cycle. RT, which is virally
encoded, converts the single-stranded viral RNA genome into
a double-stranded linear DNA intermediate that is subsequently
integrated into the host cell DNA (Goff, 1990; Mitsuya
et al., 1990).
The success of treatment of HIV-1 with inhibitors that target
RT has been limited by the development of resistance of the
virus to the various drugs (Larder and Kemp, 1990; St Clair
et al., 1991; Richman et al., 1994; Zhang et al., 1994). The
reasons behind the eventual failure of the inhibitors include
the large numbers of virions in the patient and their rapid
turnover, in addition to continual mutation in the virus (Coffin,
1995). Treatment of patients with either nucleoside or non-
Materials and methods
Computational methods
All molecular modeling of the enzyme complexes was carried
out using Insight II software (MSI), with all calculations
performed using the C force field (cff91) within the Discover
module mounted on a Cray YMP-8 supercomputer at NCIFCRDC. The modeling calculations were based on the X-ray
structure coordinates of the complexes of RT–DNA and
RT–8-chloro-tetrahydro-imidazo(4,5,1-jk)(1,4)-benzodiazepin2(1H)-thione (8-C1 TIBO) (R86183) (Jacobo-Molina et al.,
1993; Ding et al., 1995). For the model of the enzyme without
inhibitor, only the protein coordinates were used. Calculations
with α-anilino-2,6-dibromophenylacetamide (α-APA) were
carried out by superimposing the inhibitor (in its non-minimized cognate complex) over 8-C1 TIBO (in its minimized
complex) and reminimizing the resulting RT–α-APA site. At
present, since there are no crystallographic data available for
1ABL-Basic
Research Program, NCI-Frederick Cancer Research and
Development Center, P.O. Box B, Frederick, MD 21702, USA 2Center for
Molecular Design, Janssen Research Foundation, Antwerpsesteenweg 37, B2350 Vosselaar, Belgium, 3Virology Research Group, Southern Research
Institute, Frederick, MD 21701 and 4Department of Chemistry, Western
Maryland College, Westminster, MD 21157, USA
*To whom correspondence should be addressed.
© Oxford University Press
1379
M.B. Kroeger-Smith et al.
mutant forms of RT either alone or in a complex with nonnucleoside inhibitors, computations involving RT variants
were constructed by extrapolation from wild-type HIV-1 RT
crystal data.
The construction of a suitable computer model from the
respective coordinate data has been documented previously
(Kroeger Smith et al., 1995). The model site consisted of p51
subdomain residues 132–142 and residues 89–116, 156–211,
213–243, 265–271, 313–323, 346—351 and 380–384 from the
p66 subdomain. The C-terminal ends of these strands were
capped with a methylamino group and the N-terminal ends
were capped with an acetyl group. The pH of the site was set
to 7.5, resulting in a total charge of zero, with the charged
residues including Lys1, Arg1, Asp– and Glu–. To simulate
mutations in the protein, amino acid residues were changed
using the Biopolymer module within Insight II. To account for
solvation effects, all RT–inhibitor complexes were surrounded
with a ~6 Å layer of water prior to minimization (the thickness
of the water layer was adjusted in order to obtain 1283 water
molecules surrounding each complex), using the higher cvff
partial charges for water (O, 20.820; H, 0.410). Calculations
were carried out in three stages, with each stage consisting of
steepest descent minimization followed by a conjugate gradient
minimization until the r.m.s. deviation was ø0.001 Å. The
protein was divided into two layers: a primary layer (~10 Å
from the inhibitor), which was constrained during initial solvent
minimization and then progressively freed (side chain first,
followed by backbone) during subsequent stages of the minimization, and a secondary layer (~10–20 Å from the inhibitor)
that was held fully constrained throughout the calculations (for
the specific residues in these two layers Kroeger Smith et al.,
1995). The solvent molecules changed position between ~0.75
and 3.0 Å during the course of the minimization.
The resultant binding energy for each RT–inhibitor complex
obtained from the minimizations was taken to be the nonbonded interaction energy resulting from the formation of
the enzyme–inhibitor complex from the initial inhibitor and
enzyme reactant species. The magnitude of the drug–protein
interaction energy, in addition to its van der Waals and
electrostatic components, was measured using the Docking
module within the Insight II package.
Biochemical procedures
Construction of the expression vector for RT (Hughes et al.,
1990; Jacobo-Molina et al., 1993) and the preparation of clones
expressing individual drug-resistance encoding mutations in
the p66 subunit and a wild-type p51 subunit (Boyer et al.,
1994a,b) were described previously. RNA-dependent DNA
polymerase activity was assayed in Escherichia coli extracts
as described previously (Boyer et al., 1992). The cytoprotective
activity of 8-C1 TIBO or α-APA against either wild-type virus
or strains containing defined mutations in RT was evaluated
in a screening assay described earlier (Weislow et al., 1989).
Results and discussion
Modeling studies on HIV-1 RT containing amino acid
substitutions
The structure of wild-type HIV-1 RT is represented by a model
containing 155 amino acid residues surrounding the nonnucleoside binding pocket. Minimizations on a subset of the
enzyme were undertaken owing to memory limitations. It
should be recognized that calculations performed on only part
1380
Table I. Results from calculations on mutant variants of RT-DNA
RT strain
Polymerase
activity
(%)a
Distance
(primer grip
to dNTP) (Å)b
r.m.s.,
difference
(Å)c
Wild-type
L100I
K101E
K103N
V106A
D110E
V179D
Y181C
D185E
D186E
Y188H
G190E
E138K
100
40
140
100
80
,5
100
115
,5
,5
75
20
150
11.7
10.7
11.3
11.6
11.5
10.8
11.7
11.1
10.2
10.5
11.0
10.9
11.2
–
0.6
0.4
0.4
0.5
0.5
0.5
0.4
0.6
0.7
0.5
0.6
0.4
aData for p66 homodimer reported as % of wild-type RT
bThe measured distance is from G231:CA to D186:CA.
activity.
cThe
difference is a comparison for the mutant and wild-type strains of RT,
as measured by backbone superposition.
of the enzyme are not necessarily representative of the response
of the entire protein and, as such, are subject to errors which
may permeate throughout the study in a differential manner.
However, since the aim of the study was to see if computations
would produce quantative trends that might shed light on the
mechanistic effects of mutations on polymerase activity and
resistance to NNIRTs, it was felt that the method was valid
within these limits. This model was modified separately at
each of 12 key amino acid residues (see Table I) and the
resultant sites were subjected to molecular mechanics energy
minimization. Nine of the amino acid residues that were
modified in the model are representative of changes observed
in drug-resistant strains that emerge either in vitro (cell assays)
(Kleim et al., 1996) or in vivo (patients) (Moermans et al.,
1995) following treatment with non-nucleoside inhibitors. The
other three amino acid mutations that were analyzed, D110,
D185 and D186, compose the triad of aspartyl residues in the
polymerase active site and are crucial for polymerase function.
These were included in the calculations since they show that
removal of one of aspartic acid residues disrupts the site.
The results from the calculations on these modified RT sites
are shown in Table I, along with the measured polymerase
activity of wild-type RT and the various drug-resistant strains.
The r.m.s. differences (measured by backbone superposition
of all 155 residues) of the modified RTs, as compared with
the wild-type starting structure, are also shown in Table I.
These overall measures of difference were found, in general,
to correlate with the measured polymerase activities of the
mutant enzymes. RT drug-resistant strains that showed substantially reduced polymerase activity had r.m.s. values .0.5 Å,
while those variants that maintained near wild-type activity
had r.m.s. values ,0.5 Å. Two notable exceptions are the
V179D and D110E changes, which both showed r.m.s. values
of 0.5 Å. Clearly, measurement of r.m.s. differences is only a
crude measure of possible disruption of the geometry at the
polymerase active site.
A more detailed analysis is necessary to elucidate the
mechanism that underlies these differences. The mutation that
caused the greatest local structural change was the G190E
substitution. In this case, structural differences were especially
evident in the β12–β14 sheet which contains the amino acid
Molecular modeling of HIV-1 reverse transcriptase
Fig. 1. Chart of the distance in Å between an amino acid residue in the
dNTP binding site region (D186) and one in the primer grip region (G231)
vs the polymerase activity for various mutant RTs.
residue W229. This residue appears to play an essential role
in the polymerase process since no viable mutations at that
position have been found to date (Jaques et al., 1994; P.L.Boyer,
H.Q.Gao and S.H.Hughes, in press). The displacement
observed for the CB atom of this residue (1.52 Å) was found
to be respresentative of the alteration of the entire backbone
chain. The sheet was also displaced when each of the other
amino acid substitutions were made, although to a lesser
extent. Other notably large residue adjustments were as follows:
for the V179D mutation, the CB atom of F227 moved 1.8 Å
compared with its position in wild-type; for the Y188H change,
the backbone residues for K101 to V106 shifted ~1.0 Å; and
for the K101E modification, the CB atom of Y183 moved 0.9 Å.
A more direct measure of this geometry alteration is the
distance between the D186 residue (CA atom) near the dNTP
binding site and a reference residue in the primer-grip region
(Jacobo-Molina et al., 1993) of RT (G231:CA). Reduction of
this distance to ,11.0 Å (wild-type 11.7 Å) appears to lead
to polymerase activities less than 50% of wild-type (Table I,
Figure 1), although slightly less reduction in some cases leads
inexplicably to an increase in polymerase activity. In order to
test that the default placement of the mutant residues by the
software program is not artificially influencing the results of
the calculations, the position of the isoleucine side chain of
amino acid 100, for example, was adjusted to an alternative
initial orientation. The measured distance of 10.5 Å following
minimization was comparable to that found from the first
calculation (10.7 Å), ruling out system artifacts. The major
effect of the mutations appears to be a local distortion in the
position of the catalytic triad (as opposed to major changes in
the backbone atoms of the β12–β14 sheet), with the CB atom
of D186, for example, moving 0.3–1.2 Å. This adjustment is
largest (1.2 Å) in the case of D186E substitution, followed by
the G190E and K101E modifications (~1.0 Å). For mutations
that retain a level of polymerase activity comparable to wildtype (with the exception of K101E noted above), the D186:CB
displacement distance is smaller (,0.8 Å).
Correlation of RT structure and function
The locations of the residues in RT that were modified prior
to minimization, together with the corresponding level of
polymerase activity for the mutant RTs, are shown in Figure
2. From the data in Table I, it is apparent that the importance
Fig. 2. Representation of the location and polymerase activity of key drugresistant variants in wild-type RT–DNA. Amino acid residues shown in red
maintain 100% of the wild-type activity upon mutation, those in blue 50–
100% and those in green ,50%.
of these residues with respect to polymerase activity falls into
three classes.
Of utmost importance are the aspartic acid triad residues,
D110, D185 and D186 (shown in green), which play a direct
role in the polymerase activity of the enzyme. Mutation in
any of these residues results in a drop in polymerase activity
to ,5% of wild-type. In addition, the D110E, D185E and
D186E mutations reduce the dNTP to primer grip distance by
0.9–1.3 Å, more than that for any other modification we have
tested. Among the key ramifications of these aspartic acid
mutations is that lengthening the carboxylate-bearing side
chain will result in substantial movement of the Mg21 ion to
which it is complexed (Steitz and Steitz, 1993; Patel et al.,
1995). Hence the location of the dNTP that in turn complexes
with this Mg21 ion through its α-phosphate would be altered
from that found in the wild-type enzyme. The end result is
that the E186 variant would be less able to process the
incoming dNTP.
Mutation of amino acid residues L100 and G190 (also
shown in green) produces enzymes that retain only 20–40%
of wild-type polymerase activity. In these cases, the dNTPprimer grip distance is significantly shortened (see Table I).
This change could interfere with the ability of the incoming
dNTP to be appropriately positioned to permit sugar–phosphate
bond formation. Although no key function has yet been
postulated for L100 in either unliganded RT or RT complexed
with DNA, examination of the structures shows that in its
location on the β5–β6 connecting loop, L100 is situated
between Y181 and Y188, making important van der Waals
contacts with both residues. Hence, in the absence of inhibitor
binding, which causes the reorientation of the side chains of
Y181 and Y188 as well as F227 and W229 during the
formation of the inhibitor binding pocket (Hsiou et al., 1966),
L100 is clearly crucial to the integrity of the folded protein.
Replacement of leucine with isoleucine could account for
substantial structural rearrangement which is reflected in the
1381
M.B. Kroeger-Smith et al.
distortion of the polymerase active site geometry and results
in a marked reduction in polymerase activity. Interestingly,
we have also found that the polymerase activity of an RT
carrying both the L100I and the K103N mutations is 85% of
that observed for wild-type enzyme. Modeling of this variant
showed that only a small decrease in the distance between the
primer grip and dNTP binding site occurred compared with
wild-type (0.1 Å). Turning to the G190 residue, it can be seen
from the model that the substitution of glutamic acid for
glycine leads to significant steric hinderance with Y188, with
the resultant structural reorganization again compromising the
critical geometry of the polymerase site. This explanation is
in keeping with the suggestion by Chao et al. (1995) that the
G190E mutation impairs protein folding of the recombinant
enzyme, as indicated by a reduction in solubility. In our hands,
however, a recombinant HIV-1 RT containing the G190E
mutation was readily soluble.
The V106 and Y188 mutations, shown in blue, retain at
least 75% of the original activity upon mutation. These residues
show slightly reduced dNTP–primer grip separation distances,
in keeping with their relatively high levels of polymerase
activity. V106 makes van der Waals contact with Y188,
suggesting that changes in these residues affect the integrity
of the folded protein, but the distortion of the geometry of the
dNTP binding site is less severe than with other mutations.
Some mutations (shown in red) do not lead to a measurable
reduction in polymerase activity in our assay. K103N and
V179D have activities identical with that of wild-type RT.
Consistent with these data, alterations to those two residues
do not markedly change the geometry of the polymerase active
site, as measured by overall r.m.s. deviations and the dNTP–
primer grip distance. Other changes such as K101E, Y181C
and E138K actually lead to an increase in polymerase activity,
although this should not be taken to mean that they are more
efficient enzymes than wild-type. Other processes, such as
fidelity, processivity, and/or strand transfer, may have defects
that are not readily measurable by a straightforward polymerase
assay. For example, the E138K mutant enzyme shows a
significantly lower Vmax and a lower Km for poly (rC)·oligo
(dG) or dGTP (P.L.Boyer, H.Q.Gao and S.H.Hughes, in press),
enabling it to incorporate more dNTP than the wild-type
enzyme. The effect of these latter mutations is a slight
compression of the dNTP–primer grip distance, 11.2 6 0.1 Å,
possibly suggesting that polymerization efficiency under the
specific conditions used in the in vitro assay may be enhanced
by this type of compression.
The polymerase activity data shown in Figure 2 thus provide
an informative picture of the correlation between structure and
function in RT. When mutations arise in important residues in
the enzyme, a displacement of catalytic site residues occurs
which reduces the ability of the enzyme to incorporate the
incoming nucleotide triphosphate.
Modeling of RT–inhibitor complexes containing amino acid
modifications
Computer modeling studies were also undertaken of RT–
inhibitor complexes that contained one of the amino acid
substitutions (Y181C, K101E, or K103N) seen to emerge
following treatment of patients with 8-C1 TIBO (Moermans
et al., 1995). Following minimization of the complexes containing variant amino acids, the backbone r.m.s. values of
these structures as compared with the wild-type were ~0.4 Å,
with the majority of the deviations being local changes directly
surrounding the site of mutation.
1382
Table II. Interaction energy and EC50 values for NNIs against drug-resistant
strains of RT
RT strain
8-CI TIBO–RTa
(EC50b)
α-APA–RTa
(EC50b)
Wild type
Y181C
K101E
L100I
K103N
–54.0
–53.2
–50.8
–49.5
–49.5
263.5 (0.4)
–52.3 (1.0)
–58.3 (0.2)
–61.4 (0.08)
–55.7 (0.7)
(0.3)
(4.2)
(17.4)
(17.4)
(17.4)
aEnergies
bValues
in kcal mole–1.
for EC50 in µM (in parentheses).
The drug–protein interaction energy for 8-C1 TIBO or αAPA in the 8-C1 TIBO site (wild-type and variant) was
measured following the minimizations. This value is useful in
the evaluation of changes in the strength of inhibitor–protein
binding among different complexes. The use of the same
starting RT–inhibitor complex for all the minimizations eliminated any inherent error associated with structural anomalies
between different crystal structures. In the present study, this
energy value was seen to be more positive (hence less
favorable) for all the variant RTs as compared with wild-type,
whether RT was complexed with α-APA or 8-C1 TIBO (Table
II). In addition, the order of the energy values corresponds not
only to the measured EC50 values for 8-C1 TIBO and α-APA
against wild-type RT, but also against the four drug-resistant
strains (see Table II).
Measurement of the dNTP–primer grip distance following
modification of the complexed protein showed only small
changes from those determined for inhibitor complexes with
wild-type enzyme (15.93 Å), with distances generally 0.12–
0.22 Å shorter than in wild-type. L100I was the only mutant
for which the changes seen upon drug binding were greater
than in wild-type, with increases of 0.16 Å for 8-C1 TIBO
and 0.57 Å for α-APA.
Implications for the design of non-nucleoside inhibitors with
a broader spectrum of activity
No crystallographic data were available for complexes between
non-nucleoside inhibitors with mutant forms of RT. In the
absence of such structural data, calculations were based on
wild-type RT–inhibitor complexes. In our previous studies
where these complexes were used as models (Kroeger Smith
et al., 1995; Smith et al., 1998), indirect evidence was obtained
that binding of various non-nucleoside inhibitors interferes
with polymerase activity in a straightforward manner. Specifically, the dNTP binding site to primer grip distance (as
represented by residues D110, D185, D186 and G231) was
correlated with the potency of the inhibitor against wild-type
RT. It was found that the greater the separation of the two
sites, the better was the EC50 value of the drug, with all of
the distances for the complexed enzyme being substantially
greater than that for RT alone. The resulting spatial rearrangements in the protein could reasonably be expected to impede
the formation of the sugar phosphate bond. Hence we proposed
a more detailed explanation of the allosteric mechanism of
inhibition by the non-nucleoside compounds (Kroeger Smith
et al., 1995).
From data in the current study, complexes of inhibitors with
variant forms of RT show distortions in the dNTP–primer grip
separation similar to wild-type (data not shown). In previous
work (Smith et al., 1998), where α-APA, 8-C1 TIBO and
Molecular modeling of HIV-1 reverse transcriptase
nevirapine were modeled in their cognate sites containing the
Y181C mutation, this same distance was slightly shortened
(0.1–0.5 Å as compared with the wild-type enzyme). In concert,
it has recently been shown experimentally that none of the
inhibitors examined here inhibit variant RT nearly as well as
they arrest the wild-type enzyme (data not shown). In support
of this idea, recent kinetic studies have shown (Spence et al.,
1996) that the Kd of the non-nucleoside inhibitor nevirapine
for RT carrying the Y181C mutation increased 500-fold
compared with the wild-type RT complex. This study suggests
that a similar mechanism (i.e. reduced binding to the various
mutant forms of RT) might also be operative in the RT–TIBO
complex aside from active site distortion. Contributing to the
decreased binding affinity could be loss of aromatic stacking
interactions, van der Waals non-bonded interactions, and/or
coulombic (electrostatic) attraction. Depending on the specific
amino acid change, the subsequent interaction of these modified
residues with inhibitors of varying structure would result in a
differing balance in the reduction of these types of interactions.
Hence, in order for a substance to be an effective inhibitor of
the polymerase activity of either wild-type or mutant HIV-1
RT, it must both bind tightly to the enzyme and, upon binding,
give rise to structural changes in the protein that result in low
polymerase activity.
Two clinically observed mutational changes that appear to
have important implications for RT inhibition are G190E and
L100I. While the exact role of these two residues is unknown,
they are clearly important for RT function (as seen from
polymerase activity data) and in addition, play an important
role in complexation with non-nucleoside inhibitors. Aside
from inhibitor binding, it appears from their location that
mutation of these residues might also interfere sterically (and
electrostatically in the case of G190E) with key contacts along
an entrance route that the inhibitors are believed to travel
during the process of binding. Additional support for a key
role of position of residue 190 comes from in vitro studies
with the non-nucleoside inhibitor HBY 079, where five out of
six resistant strains displayed a G190E mutation following
treatment with the drug (Kleim et al., 1996). In the same paper
the authors noted that the double mutant L100I/K103N may
result in the same phenotype as that seen with the G190E
mutant.
Resistance, which develops following treatment with nonnucleoside inhibitors of RT, is the key factor undermining the
ability of these drugs to lower viral load. While a drug that
effectively inhibits all forms of RT (wild-type and variant)
would be an ideal goal, a highly satisfactory alternative would
be one that selects for weak mutant variants that replicate less
well than the wild-type. With respect to either objective, an
understanding of the structural consequences of mutations in
RT in its unliganded and complexed states is vitally important.
With this information at hand, one strategy to increase the
activity of known inhibitors against variant forms of the virus
would be to modify current inhibitors chemically with the aim
of enhancing the hydrophobic character of the π-electronrich components of the molecule. The addition of a bulky,
hydrophobic group to the inhibitor would fill the void left by
the loss of an aromatic residue, such as Y181. A second tactic
would be to design a chemically modified inhibitor that would
covalently bind to an essential amino acid residue, such as
W229. Hopefully, modifications such as these will lead to
inhibitors that have a broad-based spectrum of action.
Acknowledgements
The authors thank Dr Edward Arnold for critical reading of the manuscript
and Valerie Fliakas-Boltz for performing the anti-HIV-1 assays. In addition,
they acknowledge the Frederick Biomedical Supercomputing Center of the
Frederick Cancer Research and Development Center for access to the CrayYMP computer. Research sponsored input by the National Cancer Institute,
DHHS, under contract with ABL. The contents of this publication do not
necessarily reflect the views or policies of the Department of Health and
Human Services, nor does mention of trade names, commercial products, or
organizations imply endorsement by the US Government.
References
Boyer,P.L., Ferris, A.L. and Hughes, S.H. (1992) J. Virol., 66, 1031–1039.
Boyer,P.L., Ding,J., Arnold,E. and Hughes, S.H. (1994a) Antimicrob. Agents
Chemother., 38, 1909–1914.
Boyer,P.L., Tantillo, C., Jacobo-Molina, A., Nanni, R.G., Ding, J., Arnold, E.,
and Hughes,S.H. (1994b) Proc. Natl Acad. Sci USA, 91, 4882–4886,
Chao,S.-F., Loong Chan,V., Juranka,P., Kaplan,A.H., Swanstrom,R and
Hutchinson, C.A. (1995) Nucleic Acids Res., 23, 803–810.
Coffin,J. (1995) Science, 267, 483–489,
Ding,J., Das,K., Moereels,H., Koymans,L., Andries,K., Janssen,P.A.J.,
Hughes,S.H. and Arnold,E. (1995) Nature Struct. Biol., 2, 407–415.
Goff,S.P. (1990) J. Acquired Immune Defic. Syndr., 3, 817–831,
Hsiou,Y., Ding,J., Das,K., Clark,A.D., Jr, Hughes,S.H. and Arnold,E. (1996)
Structure, 4, 853–860.
Hughes,S.H., Ferris,A. and Hizi,A. (1990) In Laver,W.G. and Air,G.M. (eds)
Use of X-Ray Crystallography in the Design of Antiviral Agents. Academic
Press, New York, pp. 297–307.
Iversen,A.K.N., Shafer,R.W., Wehrly,K., Winters,M.A., Mullins,J.L.,
Chesebro, B. and Merigan,T.C. (1996) J. Virol., 70, 1086–1990.
Jacobo-Molina, A. et al., (1993) Proc. Natl Acad. Sci. USA, 90, 6320–6324.
Jacques,P.S., Wohrl,B.M., Ottmann,M., Darlix,J.-L. and Le Grice,S.F. (1994)
J. Biol. Chem., 269, 26472-26478.
Kleim,J.-P., Rosner,M., Winkler,I., Paessens,A., Kirsch,R., Hsiou,Y., Arnold,E.
and Riess,G. (1996) Proc. Natl Acad. Sci. USA, 93, 34–38.
Kroeger Smith,M.B. et al. (1995) Protein Sci., 4, 2203-2222.
Larder,B.A. (1992) Antimicrob Agents Chemother., 36, 2664–2669.
Larder,B.A. and Kemp,S.D. (1990) Science, 246, 1155–1158.
Mitsuya,H., Yarchoan,R, and Broder,S. (1990) Science, 249, 1533–1544.
Moermans,M. et al., (1995) Abstracts of the Fourth International Workshop
on HIV-1 Drug Resistance, Sardinia, Italy, July 6–9.
Patel,P.A., Jacobo-Molina,A., Ding,J., Tantillo,C., Clark,A.D., Jr, Raag, R.,
Nanni,R.G., Hughes,S.H. and Arnold,E. (1995) Biochemistry, 34, 5351–
5356,
Ren,J., Esnouf,R., Hopkins,A., Ross,C., Jones,Y., Stammers,D. and Stuart,D.
(1995) Structure, 3, 915–926.
Richman,D.D. et al., (1994) J. Virol., 68, 1660–1666.
Smith,R.H., Jr, Michejda,C.J., Hughes,S.H., Arnold,E., Janssen,P.A.J. and
Kroeger Smith,M.B. (1998) J. Struct.Biol., (Theochem), in press.
Spence,R.A., Anderson,K.S. and Johnson,K.A. (1996) Biochemistry, 35,
1054–1063.
St Clair,M.H. et al. (1991) Science, 253, 1557–1559.
Steitz,T.A. and Steitz,J.A. (1993) Proc. Natl Acad. Sci. USA, 90, 6498–6502.
Tisdale,M., Kemp,S.D., Parry,N.R. and Larder,B.A. (1993) Proc. Natl Acad.
Sci. USA, 90, 5653–5656.
Weislow,O.S., Kiser,R., Fine,D.L., Bader,J., Shoemaker,R.H. and Boyd,M.R.
(1989) J. Natl Cancer Inst., 81, 577–586.
Zhang,D., Caliendo,A.M., Eron,J.J., DeVore,K.M., Kaplan,J.C., Hirsch,M.S.
and D’Auuila,R.T. (1994) Antimicrob. Agents Chemother., 38, 282–287.
Received April 29, 1997; revised July 25, 1997; accepted September 5, 1997
1383