Full text of DOI:10.1110/ps.9.11.2232

Protein Science ~2000!, 9:2232–2245. Cambridge University Press. Printed in the USA.
Copyright © 2000 The Protein Society
Trp42 rotamers report reduced flexibility when the
inhibitor acetyl-pepstatin is bound to HIV-1 protease
1
1
1
2
BEÁTA ULLRICH, MONIQUE LABERGE, FERENC TÖLGYESI, ZOLTÁN SZELTNER,
2
1
LÁSZLÓ POLGÁR, and JUDIT FIDY
1Institute of Biophysics and Radiation Biology, Semmelweis University, P.O.B. 263 H-1444 Budapest, Hungary
2
Institute of Enzymology, Biological Research Centre, Hungarian Academy of Sciences, P.O.B. 7, H-1518 Budapest, Hungary
~Received January 25, 2000; Final Revision August 15, 2000; Accepted August 24, 2000!
Abstract
The Q7K0L33I0L63I HIV-1 protease mutant was expressed in Escherichia coli and the effect of binding a substrate-
analog inhibitor, acetyl-pepstatin, was investigated by fluorescence spectroscopy and molecular dynamics. The dimeric
enzyme has four intrinsic tryptophans, located at positions 6 and 42 in each monomer. Fluorescence spectra and
acrylamide quenching experiments show two differently accessible Trp populations in the apoenzyme with k ϭ 6.85 ϫ
q1
9
Ϫ1 Ϫ1
9
Ϫ1 Ϫ1
9
Ϫ1 Ϫ1
1
0 M
s
and kq2 ϭ 1.88 ϫ 10 M , that merge into one in the complex with k
s
q
ϭ 1.78 ϫ 10 M
s
.
5
00 ps trajectory analysis of Trp x
1
2
0x rotameric interconversions suggest a model to account for the observed Trp
fluorescence. In the simulations, Trp60Trp6B rotameric interconversions do not occur on this timescale for both HIV
forms. In the apoenzyme simulations, however, both Trp42s and Trp42Bs are flipping between x 0x states; in the
1
2
complexed form, no such interconverions occur. A detailed investigation of the local Trp environments sampled during
the molecular dynamics simulation suggests that one of the apoenzyme Trp42B rotameric interconversions would allow
indole-quencher contact, such as with nearby Tyr59. This could account for the short lifetime component. The model
thus interprets the experimental data on the basis of the conformational fluctuations of Trp42s alone. It suggests that the
rotameric interconversions of these Trps, located relatively far from the active site and at the very start of the flap region,
becomes restrained when the apoenzyme binds the inhibitor. The model is thus consistent with associating components
of the fluorescence decay in HIV-1 protease to ground state conformational heterogeneity.
Keywords: 500 ps molecular dynamics; acrylamide quenching; dihedral trajectories; dynamics influenced by inhibitor;
ground state heterogeneity; Trp fluorescence; Trp rotamers
HIV-1 protease the homodimeric aspartyl-protease encoded in the
viral genome of the HIV-1 virus is essential for the proper matu-
ration of the virus. It cleaves the polyprotein products of the gag
and pol viral genes ~Kohl et al., 1988!, making the release of
enzyme proteins including the protease itself possible ~Seelmeier
et al., 1988!. Inactivation of the enzyme by mutation or chemical
inhibition causes the production of immature, noninfectious viral
particles ~Seelmeier et al., 1988!. Use of the enzyme was proposed
as a target in anti-AIDS drug design, and thus the effect of binding
various inhibitors on protease structure and activity is currently the
focus of intensive research ~Fitzgerald et al., 1990; Wlodawer &
Erickson, 1993; Pargellis et al., 1994; Polgár et al., 1994; Yamazaki
et al., 1994; Nicholson et al., 1995; Wlodawer & Vondrasek, 1998;
Trylska et al., 1999!. Four such inhibitors, i.e., saquinavir, indi-
navir, ritonavir, and nelfinavir, were recently introduced to clinical
applications ~Flexner, 1998!, as a result of structure assisted drug
design ~Wlodawer & Vondrasek, 1998!. While a great amount of
effort has been invested toward practical applications ~e.g., deter-
mination of inhibitor binding constants, description of the active
site triad!, little has been contributed to the detailed understanding
of the structural consequences of inhibitor binding at the level of
the whole protein molecule. To obtain information about the struc-
ture of the enzyme in solution, NMR spectroscopy has been used
for several years, but no results have been published about the
apoenzyme itself, the studies being limited to the protease complex
with various inhibitors ~Yamazaki et al., 1994; Wang et al., 1996a,
1996b!. The lack of comparative studies is probably due to the fact
that the apoenzyme is known to undergo autoproteolysis ~Rosé
et al., 1993; Mildner et al., 1994!. In the present study, we focused
our interest on complex formation with acetyl-pepstatin and used
the Q7K0L33I0L63I HIV-1 mutant, known to be resistant to auto-
proteolysis ~Mildner et al., 1994!. Highly sensitive fluorescence
spectroscopic methods allowed us to perform the studies in solu-
tions of low concentration.
Reprint requests to: Judit Fidy, Institute of Biophysics and Radiation
Biology, Semmelweis University, P.O.B. 263 H-1444 Budapest, Hungary;
e-mail: judit@puskin.sote.hu.
There are three available crystal structures for HIV-1 protease,
determined at various resolutions: the a-carbon backbone structure
2
232

Tryptophan monitoring of HIV-1 protease dynamics, inhibitor binding
2233
of the NY5 isolate at 3.0 Å ~Navia et al., 1989!, the synthetic
structure replacing two cysteines by ABAs at 2.8 Å ~Wlodawer
et al., 1989!, and the 2.7 Å structure of the native protease ~Lapatto
et al., 1989!. The coordinates of several protease complexes with
peptide or nonpeptide inhibitors have also been deposited at the
Protein Data Bank ~PDB!, including the 2.0 Å resolution structure
of HIV-1 complexed to acetyl-pepstatin ~Fitzgerald et al., 1990!, as
well as that of the Q7K0L33I0L63I mutant complexed to the
Glu-Asp-Leu tripeptide ~2 Å, Louis et al., 1998!.
HIV-1 protease is a dimer ~Fig. 1A!, containing two identical
subunits and two Trp residues per monomer, at positions 6 and 42.
The greater part of the protease is ordered in b-strands except for
residues 86–94, which are ordered in a-helical form. Asp25, Thr26,
and Gly27 form the “active site triad” at the dimer interface where
the inhibitor binds. The binding site can accommodate six to eight
amino acid residues. Residues 42–58 form a highly flexible region
in each subunit of the protease, referred to as the “flap region.” It
is also part of the dimer interface and it is evident from the X-ray
structures that it is significantly involved in the conformational
change occurring upon inhibitor binding. A 4.6 Å motion of the
flap region toward the active site takes place upon binding, as seen
in the upper part of the structures shown in Figure 1. The mono-
mers are slightly asymmetric, as illustrated in Table 1, which lists
some inequivalent distances measured from Trp residues in the two
subunits of the complex. The two Trp residues ~6 and 42, respec-
tively! are also not equivalent, both in terms of geometry and in the
respective environments in which they are found: Trp6—located
close to the N-terminal believed to play a role in the H-bond
Table 1. Inequivalent atomic distances between residues in the
two subunits of the HIV-1 complex with acetyl pepstatina
Distance
Distance
Difference
Trp
Residue
~Å!
Trp
Residue
~Å!
~Å!
6
6
6
6
42
Asn37B
Lys7
Lys7B
Gln92B
Glu65
31.99
7.41
6B
6B
6B
6B
42B
42B
42B
Asn37
Lys7B
Lys7
Gln92
Glu65B
Glu35
Asn37B
30.01
9.76
27.34
6.80
26.42
36.23
11.64
1.98
2.35
1.99
3.23
1.97
2.67
2.52
29.33
10.03
24.45
38.90
9.12
4
4
2
2
Glu35B
Asn37
aThe atomic distances were measured between the indole N of Trp and
the g-C in case of Asn and d-C atoms in cases of Glu and Gln.
network connecting the two monomers ~York et al., 1993!—
stretches out in solvent ~cf. Fig. 2!, while Trp42, located in the flap
region, has its aromatic ring flipped back unto itself and in closer
contact with the protein matrix. The above differences between
Trp6 and Trp42 are maintained in both the complexed and uncom-
plexed forms of HIV-1.
The presence of tryptophan intrinsic probes makes the protease
a good candidate for fluorescence studies. Conformational changes
caused by increased pH and Mg~II! ion concentration were suc-
cessfully monitored using the emission spectrum of the intrinsic
Trp residues ~Tyagi et al., 1994!. In a recent work, the fluorescence
intensity and anisotropy decay of several apoenzyme–inhibitor com-
plexes were studied using a ps laser excitation technique, and the
lifetime components extracted from the fluorescence decay mea-
Fig. 1. Structure of ~A! HIV-1 protease and ~B! its complex with acetyl
pepstatin. The Trp residues and the inhibitor are rendered as balls and
sticks.
Fig. 2. Side view of ~A! the structures of HIV-1 protease and ~B! its
complex with acetyl pepstatin.

2234
B. Ullrich et al.
surements were identified based on simulations of the dynamic
behavior of the two types of Trp residues present in the structure
and that of the complex overlaid. This shows a significant differ-
ence in the blue side of the excitation maximum where the con-
1
~
Kungl et al., 1998; Ringhofer et al., 1999!. The complex with the
tribution of the La R
A transition is known to predominate ~Callis,
inhibitor acetyl-pepstatin has not been studied previously.
1997; Valeur & Weber, 1977!.
In this work, we compare the structure of the Q7K0L33I0L63I
mutant HIV1-protease and its complex with acetyl-pepstatin. We
present models for the structure of the mutant apoenzyme obtained
from the three-dimensional ~3D! structural data available for its
complex with a small tripeptide inhibitor. We describe the dynamic
behavior of the Trp residues during the course of comparative
In Figure 4A, the fluorescence decay of the apoenzyme is shown
at room temperature, as registered by the ns-flashlamp system,
with excitation at 295 nm and emission at 340 nm. The decay
curves could not be well fitted with one exponential, but the sum
of two exponentials resulted in acceptable chi-squared values
~cf. weighted residuals in the figure!. In the case of the complex
~Fig. 4B!, the two exponential fit did not improve the quality of the
fit significantly. The data ~average values of several independent
experimental series! are shown in Table 2. For the apoenzyme,
500 ps molecular dynamics simulations ~MDS! and propose a
model to interpret the ns-experimental fluorescence data on the
basis of Trp42B rotamer states exploring different local environ-
ments. We compare their respective accessibility to potential
quencher groups and discuss the different local environments of
Trp6s and Trp42s.
typical values from the two-component fit were: t ϭ 2.73 ns
1
~19.2%! and t ϭ 5.7 ns ~80.8%!, ^t& ϭ 5.13 ns and for the
2
complex: t ϭ 3.1 ns ~21.7%! and t ϭ 5.9 ns ~78.3%!, ^t& ϭ
1
2
5.29 ns.
Results
Fluorescence quenching studies
Effect on the spectra
Fluorescence spectroscopy and excited state lifetime
The corrected and normalized fluorescence excitation and emis-
sion spectra of the apoenzyme at room temperature are shown in
Figure 3A. Complex formation does not influence the fluorescence
emission spectra significantly, thus it is not shown in the figure.
Figure 3B shows the fluorescence excitation spectra of the enzyme
Fluorescence quenching studies using acrylamide as a quencher
were performed to test if the two discrete lifetime components of
the fluorescence decay can be identified as the contribution of two
Fig. 4. Fluorescence decay of ~A! the apoenzyme and ~B! the complex.
Decays were collected at 295 nm of excitation wavelength and 340 nm of
emission wavelength. Results of two-exponential fits for the presented
decays were: t1 ϭ 2.7 ns, f1 ϭ 19.7%, t2 ϭ 5.6 ns, f2 ϭ 80.3%, ~apoen-
zyme!, t1 ϭ 3.36 ns, f1 ϭ 19.9%, t2 ϭ 5.84 ns, f2 ϭ 80.1%.
Fig. 3. A: Fluorescence excitation ~lem ϭ 340 nm! and emission ~excita-
tion lex ϭ 295 nm! spectra of the apoenzyme. B: Superposed excitation
spectra of apoenzyme and complex ~dashed line!.

Tryptophan monitoring of HIV-1 protease dynamics, inhibitor binding
2235
Table 2. Van der Waal contact with protein quencher groups
by the different Trp rotamers sampled during 500 ps
simulation time
accessible and one inaccessible Trp, using the linearized form of
the Stern–Volmer equation ~Lehrer, 1971!:
I0
1
1
ApoHIV-1
Complex
vdW
ϭ
ϩ
~1!
⌬
I
f K @Q# fa
a D
vdW
contact
contact
a
where f is the fraction of the initial fluorescence that is accessible
gϪ
Ϫ908
gϪ
Ϫ908
to the quencher and DI is the difference between the intensity of
Trp6
x1
x2
N
N
N
N
fluorescence in the absence of quencher ~I ! and the intensity at
0
0
various quencher concentrations. The linear fit to I 0DI plotted as
Trp6B
Trp42
Trp42B
x1
x2
gϪ
Ϫ908
N
N
gϪ
Ϫ908
N
N
a function of @Q# ~not shown! yielded 86% for the contribution of
the accessible Trps in the apoenzyme and around 100% for the
complex ~i.e., the y-intercept was close to 1!.
x1
x2
gϩ
900Ϫ908
N
N
gϩ
908
N
N
The downward curvature of the Stern–Volmer plot in the case of
the apoenzyme agrees well with the multicomponent feature of the
lifetime data. It indicates that there are Trps in the protein with
distinct accessibility for quenching ~Eftink & Ghiron, 1981!. The
inset of Figure 5 shows that quenching leads to a blue shift in the
emission spectrum of the apoenzyme. This result is not only in-
dicative of the presence of distinct Trps, but also shows that the
Trps with red shifted spectra are more effectively quenched.
In the case of the complex, no significant spectral change could
be observed upon quenching ~cf. Fig. 5, inset!. Thus, the Trps in
this conformation are not significantly different in their accessi-
bility for quenching. The upward curvature observed in the inten-
sity plot of the quenching experiment is probably indicative of the
occurrence of both dynamic and static processes, but a similar
effect arising from a significant contribution of the transient term
of the Smoluchowski equation ~Eftink & Ghiron, 1981! during the
quenching process cannot be excluded either.
ϩ
Ϫ
gϩ
908
x1
x2
g 0g
900Ϫ908
Y
Y
N
N
differently accessible Trps of the protein. We selected acrylamide
as quencher because it only has a very slight effect on the enzyme
activity of HIV-1 protease ~3–5% decrease at a 0.3 M acrylamide
concentration!. Similar concentrations of iodide as quencher dras-
tically lowered the enzyme activity.
The fluorescence intensity at a constant emission wavelength
~
lex ϭ 295 nm, lem ϭ 340 nm! as well as emission spectra ~lex
ϭ
295 nm! were measured as described in Materials and methods.
The intensity plot, shown in Figure 5, shows a downward curva-
ture in the case of the apoenzyme. The same experiment in the case
of the complex leads to an upward-curving intensity plot ~cf. Fig. 5!.
A Lehrer plot analysis ~not shown! was also performed, based on
the most simple model, i.e., that the emission occurs from one
Effect on the fluorescence decay
Figure 6A shows a Stern–Volmer plot based on an evaluation
using two lifetime components for the case of the apoenzyme. It is
seen that both lifetime components can be fitted to linear plots,
yielding significantly distinct k bimolecular quenching constants.
q
The values are listed in Table 3. The conclusion from the quench-
ing studies for the apoenzyme is that the data can be well inter-
preted by assuming the existence of two types of Trp populations.
The population with shorter lifetime is more effectively quenched
by a collisional mechanism, and this population has emission spec-
tra red shifted relative to that of longer lifetime.
In the case of the complex, the decay curves could not be so well
fitted by two lifetimes ~as mentioned above!. The Stern–Volmer
plots based on two lifetime components did not lead to straight
lines ~not shown!, but a downward curvature could be observed.
As the intensity plot and the spectral data shown in Figure 5 were
indicative of similar Trp behavior in the quenching experiment, we
evaluated the lifetime data by using the mean lifetime in the Stern–
Volmer plot as shown in Figure 6B. The data points are from three
independent experiments; they can be well fitted by a straight line.
The bimolecular quenching constant is given in Table 3; it is
somewhat smaller than that of the Trps with longer lifetime com-
ponents in the apoenzyme.
Trp dynamics by computer simulations
Fig. 5. Intensity Stern–Volmer plots of the apoenzyme ~Ⅲ! and the com-
plex ~⅙!. Inset: Fluorescence emission spectra of ~A! the apoenzyme and
Figures 7 and 8 plot the variation of the x and x Trp dihedral
angles for the 500 ps time course of the distance-dependent di-
electric simulations acquired after a 100 ps equilibration. All Trp
1
2
~
B! the complex at 0 M ~solid line! and 0.35 M ~dashed line! acrylamide
concentrations, excitation wavelength ϭ 295 nm.

2236
B. Ullrich et al.
tions for the length of available and comparable trajectory time
~cf. Materials and methods!. In the motion of Trp6 and Trp6B, the
trajectories did not show significant differences, both in the apo-
enzyme and in the complex. These residues kept fluctuating about
the same average geometry, i.e., that of the gϩ for the x , and 908
1
2
for the x dihedral case.
To illustrate how the rotamer conformations sampled by the Trp
residues during the time course of the trajectories report on the
local protein environment, we display graphically ~Figs. 9, 10! the
proximity of the Trp indoles to a protein matrix quencher group,
i.e., the phenol group of Tyr59 ~Chen & Barkley, 1998!. For clari-
ty’s sake, we only display the van der Waals surfaces of the stron-
gest quencher present in the respective environments of the Trp
residues. Specifically, Figure 9 compares—for both apoenzyme
and complex—the quenching environment of Trp42 and Trp42B in
the x rotamer conformations extracted from the MDS and shows
1
that, in its “flipping” motion between two rotamers, Trp42BgϪ in
the apoenzyme is the only species which comes into effective van
der Waals contact with the quencher. Figure 10 compares the x2
interconversions. In the case of the apoenzyme Trp42, the 908 and
Ϫ908 rotamer interconversion does not bring the indole closer to
the quencher ~Fig. 10A,B!. However, in the case of Trp42B
~
Fig. 10C,D!, interconversion again results in van der Waals over-
lap of the Ϫ908 rotamer with the quencher group ~Fig. 10D!.
Thus, the dynamics are suggestive of two different Trp42 pop-
ulations: one that would make van der Waals contact with a quench-
ing group, consisting of the Ϫ908 and gϪ rotamers of Trp42B
exposed to quencher in the apoenzyme, and another population,
not exposed to Tyr59, that would consist of Trp42 in both apoen-
zyme and complex and of the 908 and gϩ rotamers of Trp42B in
the complex.
Fig. 6. Stern–Volmer plots for t1 ~Ⅲ! and t2 ~᭹! discrete lifetime compo-
nents of ~A! the apoenzyme and ~B! for ^t& data of the complex. Excitation
wavelength ϭ 295 nm; emission wavelength ϭ 340 nm.
residues are seen to adopt the x
cally favored staggered rotamers, i.e., g , g , or 908 or Ϫ908
cf. Materials and methods!. The distribution of rotamer popula-
1
0x
2
geometries of stereochemi-
The same analysis was performed for all other known quencher
groups of indole present within probable van der Waals overlap
distance, i.e., the sulfhydryl group of Cys ~no contact!; the amide
groups of Asn37, Asn37B, Asn88 ~no contact!, and of Gln58, 92B
~no contact!; the E-amino group of Lys43, Lys55, and Lys55B ~no
contact!. Special attention was paid to Lys7, which replaces Gln7
in the mutant protease. Its E-amino group, also reported as an
indole fluorescence quencher in this pH-range ~Tran & Beddard,
1985!, was observed to be dynamically quite active during the time
course of the simulations, in some instances approaching Trp6 as
close as 2.8 Å with van der Waals overlap of the residues. In the
simulations, no other Lys residue was seen to approach a Trp
indole so closely ~e.g., Lys55, Lys55B, Lys43, Lys43B!. The quench-
ing mechanism of lysine in proteins is not yet known, but studies
performed on model peptides ~Chen & Barkley, 1998! have pro-
posed an excited-state proton transfer mechanism. If this is the
ϩ
Ϫ
~
tions is also listed in Table 2. We note here that Trp does not adopt
the third possible x rotamer, i.e., t ϭ 1808; this is not surprising
1
as this rotamer has been shown to be indicative of a-helical struc-
ture ~Willis et al., 1994!, which does not occur in this region of the
protease.
The most dynamically active Trp is Trp42B, which switches
between two possible x 0x rotamers, but only in the apoenzyme
1
2
~
cf. bottom left graphs in Figs. 7, 8!. In the complex, it adopts only
one x 0x conformation ~cf. bottom right graphs in Figs. 7, 8!. As
1
2
for Trp42, it adopts the gϩ geometry in both the apoenzyme and
the complex, and both 9080Ϫ908 conformations in the apoenzyme.
We note here that this type of dynamic behavior was observed in
both explicitly solvated and distance-dependent dielectric simula-
Table 3. Fluorescence decay parameters of apoenzyme and complex obtained
a
from a double exponential fit
Apoenzyme
Complex
ti
Contribution
kq
ti
~ns!
Contribution
^t&
~ns!
kq
~ns!
~%!
~MϪ1 sϪ1
!
~%!
~MϪ1 sϪ1
!
2
5
.7 6 0.16
.7 6 0.05
19.2
80.8
6.85 ϫ 109
1.88 ϫ 109
3.1 6 0.36
5.9 6 0.14
21.7
78.3
5.29
1.78 ϫ 109
aThe decays were collected at lex ϭ 295 nm and lem ϭ 340 nm.

Tryptophan monitoring of HIV-1 protease dynamics, inhibitor binding
2237
Fig. 7. Variation of the x1 dihedral angles of all four Trp residues ~6, 6B, 42, and 42B! present in HIV-1 protease ~left side! and in
the complex ~right side! during a 500 ps simulation. A dihedral of 608 corresponds to the gϩ rotamer and one of Ϫ608 to the gϪ rotamer.
x1 is defined as the N-CA-CB-CG dihedral angle. AP symbol refers to the results obtained for the complex with acetyl-pepstatin
inhibitor.
case, in addition to proximity, very stringent geometry orientations
would also be required, so as to have the lysine make contact with
either C2, C4, or C7 on the indole ring. No such overlap could be
extracted from the MDS runs.
Possible peptide bond quenching was also investigated. In model
peptide studies, effective quenching requires at least two amides
~Chen et al., 1996!, and such close overlap was not observed with
backbone peptide bond segments in the immediate vicinity of the

2238
B. Ullrich et al.
Fig. 8. Variation of the x2 dihedral angles of all four Trp residues ~6, 6B, 42, and 42B! present in HIV-1 protease ~left side! and in
the complex ~right side! during a 500 ps simulation. x1 is defined as the CA-CB-CG-CD2 dihedral angle. AP symbol refers to the
results obtained for the complex with acetyl-pepstatin inhibitor.
Trps. However, recent work by Engelborgh’s group shows that, in
proteins, very effective quenching can occur by electron transfer
from Trp’s indole CE3 atom to Trp’s own peptide bond—carbonyl
carbon. Quenching can occur via a “through bond” mechanism;
that is, the electron can travel along the covalent bond network
linking the carbonyl C to the indole CE3 or via a “through space”
pathway ~Sillen et al., 2000!. These quenching mechanisms, how-
ever, require specific orientations for the interacting orbitals. We

Tryptophan monitoring of HIV-1 protease dynamics, inhibitor binding
2239
Fig. 9. Van der Waals surface renderings ~mesh! of the x1 Trp rotamers and of the phenol of Tyr59 sampled during 500 ps and showing
no quenching contact for a, b, c, and d ~respectively, Trp42g and Trp42Bgϩ in the complex, Trp42gϩ and B42gϩ in the apoprotease!,
and quenching overlap for e ~Trp42BgϪ in apo!.
ϩ
analyzed the trajectories to monitor the distance of the Trp indole
CE3 atoms to the nearest peptide bond carbonyl C-atoms. This
yielded distances that came within efficient quenching distance for
the Trp42s but not for the Trp6s ~;4.5 Å, cf. Sillen et al., 2000!.
However, even if the analysis resulted in showing that the distance
requirement could be met, we could not investigate the orbital ori-
entation requirement and could therefore not rigorously elaborate
the details of a possible peptide bond additional quenching possibility.

2240
B. Ullrich et al.
Fig. 10. Van der Waals surface renderings ~mesh! of the x2 Trp rotamers and of the phenol of Tyr59 sampled during 500 ps and
showing no quenching contact for the 908 and Ϫ908 rotamers of Trp42 and the 908 rotamer of Trp42B in the apoenzyme ~A, B, and
C! and quenching contact for the Ϫ908 rotamer of Trp42B in the apoenzyme ~D!.
Trp solvent accessible surfaces
Table 4. Solvent accessible surface areas of Trp residues
during 500 ps
The orientation of the protease structures ~cf. Fig. 2! clearly shows
that both Trp6 and Trp6B residues are solvated in the X-ray struc-
tures. It was of interest to know if this degree of solvation under-
goes significant fluctuation extrema in time. The average SASA
Trp
RHIV-1 ~Rmin0Rmx!
RComplex ~Rmin0Rmx!
6
77.2 ~75.4078.2!
76.7 ~75.1079.0!
47.2 ~44.4049.5!
45.0 ~42.7046.3!
47.2 ~43.4048.2!
76.7 ~72.6078.2!
74.8 ~73.2077.1!
75.9 ~72.3076.7!
38.4 ~36.9043.1!
44.3 ~41.2047.8!
6
4
4
4
4
B
~
cf. Materials and methods! calculated for all Trp residues in both
2
complexed and uncomplexed forms during the 500 ps simulation
are listed in Table 4 as well as the extreme maximum0minimum
values. The Trp6 and Trp6B residues are seen to remain clearly
solvated with R values consistently greater than 72.3, i.e., above
the limiting value of 70 for solvent-exposed residues. The R values
2B ~gϩ!
2B ~gϪ!
2B ~tr!

Tryptophan monitoring of HIV-1 protease dynamics, inhibitor binding
2241
of the Trp42 and of the Trp42B rotamers are also indicative of
solvent exposure, but to a much lesser extent than the Trp6s with
R values Ϸ40 ~cf. R value for buried Trp is Ͻ20!. They can thus
be considered as partially buried or with one plane of the indole
ring exposed to the protein matrix and the other to solvent. It
should be noted that, in view of the simulations, as Trp42B flips
from the g to the gϪ conformations, it also becomes fully solvent-
exposed ~42B“tr” in Table 4!.
on the mean lifetime ~i.e., 5.29 ns!, however, shows a linear de-
pendence. There is also no spectral change due to the presence of
the quencher. All these facts lead us to conclude that the proba-
bility of collisional quenching is very similar for all the Trps in the
dimeric protein and that all have about the same fluorescence
decay time. Thus, the binding of the inhibitor somewhat decreases
the long lifetime component of the apoenzyme and significantly
increases the short component.
ϩ
The MDS trajectories suggest a model in which inhibitor bind-
ing would only affect the motion of Trp42B. In the trajectories,
this Trp becomes very similar to that of Trp42 in the complex.
Since Trp42 and Trp6 are located in very different environments
Discussion
Identification of the Trp fluorescence lifetime components
~cf. Table 4!, it is hard to imagine that they would form one
In the ns range, the fluorescence decay of the apoenzyme could be
well fitted with two exponentials, namely 2.7 and 5.7 ns, with a
much higher contribution from the longer component ~cf. Table 2!.
The lifetime values are comparable to those reported by Kungl
et al. ~1998! for native HIV-1 protease. Those values were 2.06 and
population in the complex with respect to collisional quenching
and fluorescence lifetime. Thus, a possible interpretation of our
data would be that the contribution of Trp6 and Trp6B to the
overall fluorescence decay is negligible on the ns timescale. In this
interpretation, the effect of inhibitor binding would be based solely
on the Trp42 rotamers with corresponding assignments that the 2.7
and 5.7 ns lifetimes characterize different rotameric states of Trp42
in the apoenzyme. In one of the monomers, this Trp ~i.e., Trp42B!
would be switching between two x and x rotamer states that are
4
.46 ns, respectively. Besides the significant difference in the long
lifetime, their respective contribution was also different, leading to
an average lifetime of 2.23 ns. The cited study, however, is very
different from ours, both in experimental methods and in studied
samples ~e.g., different pH, buffer, and wild-type enzyme!. The
authors used a frequency-doubled laser excitation source, and an
analysis based on the maximum entropy method ~MEM!. The
pulse shape shown in Figure 4 of their paper has a width of about
1
2
not equally perturbed by the presence of nearby Tyr59. In the other
monomer, no such flipping between rotamer states would occur,
i.e., Trp42 would never be brought into contact with Tyr59. Upon
complex formation, the previously interconverting Trp42B would
stop its flipping, thus becoming similar to the other one. Experi-
mentally, this interpretation is supported by the small magnitude of
the red-shift in the emission spectra of the acrylamide quenching
experiments ~4 nm! resulting in a 340 nm maximum for the red
population, difficult to associate with fully solvated Trp6s, for
which we would expect an emission maximum around 348 nm
~Lakowicz, 1983!. The two populations assigned to the Trp42B
rotamers in the apoenzyme would then consist of one rotamer
population ~the blue one! more exposed to the hydrophobic resi-
dues of the protein matrix ~Try59B, Leu38B! and the other one, the
rotamer population, less exposed to these after the flip ~the red
one!.
1
ns. It is evident that their method leads to a better signal0noise
ratio and thus it allows for MEM analysis even for lifetime distri-
butions centered around values in the ps range. The intrinsic time
resolution of the experiment, however, does not improve much.
The fact is that it would be hard to perform a precise comparative
analysis of the two sets of results, even if the sample conditions
had been identical.
In our quenching experiments, the validity of a heterogeneous
source of emission is supported by the downward curvature of the
Stern–Volmer intensity plot ~Fig. 5!, indicative of two populations
not equally accessible to acrylamide. The dynamic collisional
quenching constants k could reliably be determined for the two
q
q
lifetime components, and they were very different ~k values for
9
Ϫ1 Ϫ1
9
Ϫ1 Ϫ1
the apoenzyme: 6.85 ϫ 10 M
s
for t
1
and 1.88 ϫ 10 M
s
Local Trp environment features related to dynamics that
support the assignment of Trp fluorescence lifetimes
for t !. The short component is quenched with much higher effi-
2
ciency by collision, and the spectra ~cf. Fig. 5, inset! show that the
accessible Trp residues in the system have moderately red-shifted
emission ~from 336 to 340 nm!. From this, one has to conclude that
the Trps of longer lifetime are in a relatively hydrophobic envi-
ronment. As for the two quenching probabilities, one needs to
Known quenchers of Trp fluorescence in model compounds and
proteins include the side chains of Lys, Gln, Asn, protonated and
unprotonated His, protonated Glu and Asp, Tyr, Cys, and cystine,
as well as the peptide bond itself ~Vos & Engelborghs, 1994; Chen
et al., 1996; Hennecke et al., 1997; Chen & Barkley, 1998; Sillen
et al., 2000!. We now discuss the possible reasons as to why the
lifetime of Trp42B could be shorter than that of Trp42, and how to
account for Trp6 and Trp6B.
consider especially the higher k value that is approaching the
q
limiting value given for indole acrylamide quenching in water as
9
Ϫ1 Ϫ1
7
ϫ 10 M
s
~Eftink & Ghiron, 1976!. In the protein litera-
ture, however, one can find similar data: Bismuto and Irace ~1994!
9
report bimolecular quenching constants of even 8.9 ϫ 10 and
9
Ϫ1 Ϫ1
2
.3 ϫ 10 M
s
for the acrylamide quenching of apomyoglobin
The Trp42s
9
Ϫ1 Ϫ1
Trps. More recently, a value of 6.4 ϫ 10 M
s
was reported
for the long component associated with a Trp nearly fully exposed
to solvent in the smooth muscle myosin motor domain with the
essential light chain ~Yengo et al., 1999!. Thus, we conclude that
our values agree well with other reported data.
In the complex form, the two component fit of the fluorescence
decay curves is somewhat questionable; the Stern–Volmer analysis
of the quenching results also suggests that the two lifetime com-
ponents have no real physical meaning. The same analysis, based
The MDS trajectories support the interpretation of two Trp42 pop-
ulations, in that they identify two groups of Trp42 in terms of their
different accessibility to quenching by Tyr59 ~cf. Figs. 7–10!, which
we now discuss. The first Trp42 population consists of Trps that do
not come into close contact with this tyrosine during the time
course of the simulations, and it consists in the apoenzyme of the
ϩ
g
and 908 rotamers of Trp42 and can thus be identified as the
long-term component of the decay; the second population consist-

2242
B. Ullrich et al.
ing of Trp42B that flips between rotamers thus approaching Tyr59
within effective quenching distance. This type of Trp can be iden-
tified as the source of the shorter lifetime component. The contri-
bution of each component to the total decay, i.e., 80% for the long
and 20% for the short lifetime components, agrees well with this
concept. Out of the four Trp42 rotamers in the apoenzyme, three
would be in a conformation that never comes into close contact
with Tyr59 while the fourth would be spending time between two
states, one that overlaps with Tyr59 and one that does not, with the
transition between conformers being fully solvent-accessible. The
number of Trps accessible for quenching ~i.e., of shorter lifetimes!
should be less than ;25% of the total. This agrees with our results.
In the complex, Trp42B rotamers would no longer come within
quenching distance of Tyr59, and they would become similar to the
first population.
of the MDS, the Trp6s do not sample different conformations and
do not come within contact distance of protein quencher groups.
Moreover, their CE3 atoms do not come within efficient peptide
carbonyl C quenching distance, i.e., dCE3-C Ͻ 4.5 Å. It could be
still reasonable to assume that they are even more active than the
Trp42s, simply because there is no apparent steric hindrance to the
mobility of their indoles. As such, their interconversion could then
very well occur on a faster timescale. Inspection of our trajectories,
however, could not yield support for this interpretation. The sim-
ulations show that wat423 ~retained from the initial X-ray struc-
ture! simultaneously H-bonds to the NH of adjacent Lys7 and to
NH of Trp6. Trp42s do not H-bond with solvent. This tethering of
Trp6 to the protein surface would be sufficient to hinder the rota-
tion of Trp6 indoles, by providing a rotational barrier of the order
Ϫ1
of ;5 kcal mol . A similar effect has been modeled for the Trp of
A note of functional interest here is that the timescale of rotamer
interconversion occurs in our trajectories within the timescale of
HIV-protease flap opening, as simulated by activated MDS to be
possible within ;140 ps ~Collins et al., 1995!, which leads us to
speculate as to whether there could be a possible correlation be-
tween Trp42B rotamer interconversion and flap opening. This is
supported by recent preliminary anisotropy data ~J. Fidy, B. Ulrich,
M. Laberge, F. Tolgyesi, L. Polgar, Z. Szeltner, J. Gallay, & M.
Vincent, unpubl. data! obtained by synchrotron radiation ~LURE,
Orsay! and evaluated by MEM analysis. These results showed the
presence of sub-ns rotational correlation time constants besides a
phospholipase A to account for experimental anisotropy decays
indicative of hindered rotation and it could be seen in the trajec-
tories only with explicit -H simulations, which is also the approach
we used. This confers a slightly dipolar character to the indole and
allows realistic simulation of H-bonding patterns ~Axelsen et al.,
1991!.
These considerations allow us to possibly account for Trp6s in
the model suggested by the simulations, i.e., they would be unable
to sample rotameric states with lifetimes distinguishable by their
different local environments. In this interpretation, they could be
assigned a sub-ns lifetime, with possible quenching either by sol-
vent ~Chen & Barkley, 1998! or as a result of efficient nonradiative
pathways such as energy transfer or electron transfer. For the latter
assignment, we need to postulate the presence of possible electron
acceptors in their immediate vicinity. If we compare the respective
environments of Trp6s and Trp42s within a 7 Å sphere, we note the
following arrangements: for the Trp42s, of eight residues present
in the sphere, three are basic ~Lys41B, Lys43B, Arg57B! and two
are hydrophobic ~Tyr59B, Leu38B!. It is of interest to note also the
presence of three so-called “conformationally important residues,”
i.e., Pro44B, Gly40B, and Pro39B, highly flexible and undoubt-
edly required for the flap opening mechanism. In contrast, the five
residues facing the Trp6s are: one basic ~Lys7! and four polar
residues ~Thr4B, Gln92, Asn88, and Thr91!. This availability of
polar groups in the vicinity of the two Trp6s would meet the
requirement for the presence of electron acceptors. Reorientational
relaxation of any of these residues as a result of the much higher
dipole moment of indole in the excited state ~Callis, 1997! would
yield a sub-ns lifetime for the Trp6s ~cf. Hudson et al., 1999 for a
discussion of Trp quenching by Gln in bacteriophage T4 lyso-
syme!. However, for such a quenching mechanism to be operative,
there is a contact requirement, such as, for example, an H-bond
between Trp6 and one of the neighbor Gln or Thr that we do not
observe in the 500 ps trajectories. However, we should point out
that the length of our simulations is not adequate to fully describe
the concerted motions involved in the observed Trp fluorescence
processes. The limitations of MDS as applied to systems of the size
of proteins are essentially twofold: crude approximations have to
be made concerning the interaction potentials and only narrow
time windows can be covered. As such, the statistically probable
energy fluctuations of the system cannot be adequately sampled
~Axelsen et al., 1991!.
;15 ns component attributed to rotation of the whole protein
~
Eftink, 1983!. Similar results were recently published on a slightly
different HIV-1 protease system with better resolution in the sub-ns
range ~Ringhofer et al., 1999!. In our preliminary anisotropy ex-
periments, a 10–100 ps broad unresolved envelope of u-values was
seen in the sub-ns range and we inspected the trajectories for
dynamic evidence other than the Trp42 rotamer interconversion.
We inspected backbone “wobble” rotational motion, defined by
changes along the CA-CB-CD axis and associated with the Rama-
chandran c and f angles of the peptide segment near the Trps.
Significant “wobble” motion could only be extracted for the back-
bone region that includes the Trp42s. In the literature, Trp-associated
backbone motion has been assigned to the 200 ps or ns rotational
correlation time regime, as for example in the case of the Tet
repressor, also a homodimer of some ;207 residues ~Vergani et al.,
2000!.
At this point, we speculate that the unresolved ps-timescale
anisotropy decay, which our preliminary results yield, encompass
the Trp42 rotamer interconversion at the ;50ps “end” of the broad
range as well as the above-mentioned backbone “wobble,” ob-
served only in the peptide region including the Trp42s. These
observations, i.e., evidence for backbone “wobble” in the Trp42
region, its timescale, the timescale of the x 0x rotameric inter-
1
2
conversions of these Trps would seem to support such a Trp420
flap opening correlation, and consequently the assignment of the
fluorescence lifetime components.
The Trp6s
Unlike the Trp42s, located in the flap region of the protease, which
opens by as much as ;7 Å during the inhibitor binding event
~
Wlodawer & Erickson, 1993!, the Trp6s in both apoenzyme and
The other possible Trp6 quenching mechanism could be Förster-
type energy transfer. Unfortunately, our present data does not al-
low the accurate calculation of the proposed energy transfer
efficiency because we are lacking the distinct Trp6 overlap inte-
complex are not located in a dynamically active region of the
protease. In our model, based on the simulations, the backbone
“
wobble” motion is insignificant for this region: during the course

Tryptophan monitoring of HIV-1 protease dynamics, inhibitor binding
2243
grals. However, inspection of the static X-ray structure shows that
the Trp6s are separated by a distance of 25 Å with face-to-face
orientation of the indoles ~cf. Fig. 2!. Using this distance and
literature data for a three-Trp protein ~De Beuckeleer et al., 1999!,
we can roughly estimate the transfer efficiency. Supposing an R0
value of 11–12 Å, the efficiency at a distance of 25 Å would
amount to some ;20%, which does not account for a strong quench-
ing effect.
The concentration of the enzyme was determined from the optical
Ϫ1
Ϫ1
density ~E280 ϭ 25,500 M cm ! with a Cary 4E ~Varian, Aus-
tralia! UV-VIS spectrophotometer. Aliquots of the concentrated
enzyme were stored at Ϫ70 8C. The enzyme activity was controlled
before the experiments by fluorescence spectroscopy making use
of the fluorogenic substrate 2-aminobenzoyl-Thr-Ile-Nle-Phe~NO
Gln-Arg ~Szeltner & Polgár, 1996!.
!-
2
Acetyl-pepstatin @Ac-Val-Val-Sta-Ala-Sta#, obtained from Sigma
Chemicals ~St. Louis, Missouri! and dissolved in DMSO, was used
as inhibitor. Aliquots of 0.5 or 1.0 mM stock solutions were added
to the protein solution; the protein:inhibitor molar ratio was about
Conclusions
1:2; the resultant DMSO content was 4%. The measurements started
The effect of binding the inhibitor
after 40 min of incubation at room temperature.
The present study proposes a model in which the partially solvent-
exposed Trp42 in HIV-1 protease, located far from the active site
for substrate binding, and also far from the tip of the flap region to
feel drastic—and immediately local—structural changes, would
still able to report on the conformational effects caused by the
binding of the inhibitor acetyl pepstatin. The model suggests that
the conformational fluctuations of both Trp42s become restricted
upon binding the inhibitor. It is worth mentioning the special ef-
forts required to use the apoenzyme as a reference system for
which no reliable structural data were available previously because
of the protease autoproteolysis problem. This difficulty was care-
fully handled in the present work. Also, this work supports the
validity of investigating the role of Trp rotamer states in a protease
system using a combined molecular dynamics simulations and
experimental approach. As such, the study supports models of
tryptophan fluorescence based on ground state conformational het-
erogeneity ~Engh et al., 1986; Hutnik & Szabo, 1989; Ross et al.,
Fluorescence measurements
The experiments were carried out using an Edinburgh Analytical
Instruments CD900 luminometer, fitted with a Hamamatsu R955
PMT as detector. The instrument is equipped with a 75 W xenon
lamp for fluorescence spectroscopy; in fluorescence lifetime stud-
ies, a nF900 type flashlamp ~40 kHz, 1.5 ns pulse width! was used
with a high purity H gas ~10 ppm impurity! filling. The individual
2
intensity decay curves were determined using a time-correlated
single photon counting method; data were collected over a time
period of 10 lifetimes. The excitation wavelength was 295 nm, and
emission was typically measured at 340 nm. Both excitation and
emission wavelengths were selected by grating monochromators.
The decay curves were analysed by iterative reconvolution using
discrete exponentials.
Evaluations based on fluorescence intensity data were corrected
for the absorption of acrylamide and for protein self-absorption in
the 1 mL cuvettes used according to the following formula:
1
992!. Furthermore, our interpretation of the lack of Trp6 contri-
bution to the 1–10 ns-timescale fluorescence suggests that the study
of quenching processes in some cases definitely requires the use of
femtosecond techniques. Finally, in spite of the arduous process
involved in trying to account for the proposed non-ns lifetimes of
the Trp6s, we believe that we have presented a plausible model for
the occurrence of Trp42 rotameric states. The rotational barrier
seen in the MDS for the Trp6s is also perhaps suggestive of the
attention required to understand quenching mechanisms that are
not collisional, i.e., those operating through space ~Silva & Pren-
dergast, 1996!.
~
OD 05ϩOD340!
95
2
Icorr ϭ Iobs10
~2!
where Icorr is the corrected fluorescence intensity, Iobs is the ob-
served fluorescence intensity, OD295 and OD340 are the optical
densities of the sample at 295 nm and at 340 nm, respectively, the
volume of cuvettes was 1 mL.
Quenching of fluorescence
Trp accessibility to small solutes was determined using acrilamide,
an efficient quencher of Trp fluorescence and selective for surface
residues ~Eftink & Ghiron, 1981!. Five or ten microliter aliquots of
Materials and methods
1
, 2.5, and 5 M stock solutions of electrophoretic grade acrylamide
Sample preparation
~Aldrich Chemical Co., Milwaukee, Wisconsin! in 50 mM phos-
HIV-1 protease was expressed in Escherichia coli 1458, and pu-
rified from inclusion bodies as described previously ~Polgár et al.,
phate buffer ~pH: 7.5! were used as quencher up to a concentration
of 0.35 M.
1994!. The enzyme contains three mutation sites at positions 7
Bimolecular collisional quenching constants ~k
! were deter-
q
~
Gln r Lys!, 33 ~Leu r Ile!, and 63 ~Leu r Ile!. This mutant
mined from the equation based on the t fluorescence lifetime
~Equation 3!.
form of the enzyme is resistant against autodegradation; its kinetic
properties and inhibition profiles by a variety of inhibitors, how-
ever, were very similar, if not identical with the wild-type enzyme
␶
0
Ϫ1 ϭ k ␶ @Q#
~3!
q
0
~
Mildner et al., 1994!. The high similarity of the kinetic properties
␶
of the wild-type enzyme and that of the mutant form suggests that
the structure of the protein is not changed very much by mutation.
The final concentration of the enzyme was 10.6–12.5 mM in 50 mM
phosphate buffer containing 50 mM disodium-hydrogen-phosphate,
where k t ϭ K , the dynamic quenching constant.
Mean lifetime ~^t&! values were obtained from the results of
two-component exponential fits according to the formula
q
0
D
1
0
mM EDTA, 2 mM DTE, 0.1 M NaCl, 10 v0v% glycerol, and
.1% PEG 2000, pH ϭ 7.5, and was used without further dilution.
͗␶͘ ϭ f ␶ ϩ f ␶
2 2
~4!
1
1

2244
B. Ullrich et al.
where f
1
and f
2
are the fractional contributions of t
1
and t
2
lifetime
NVT simulations were carried out at a constant temperature of
300 K using a leapfrog algorithm with a time step of 1 fs. 200 ps
trajectories were acquired, and the initial equilibration 100 ps were
disregarded in the analysis. Nonbond interactions were taken into
account using the cell multipole method ~Schmidt & Lee, 1991! as
previously described ~Laberge et al., 1998!.
As explicitly solvated, dynamics are quite CPU-time consum-
ing; longer 600 ps trajectories were then acquired using the distance-
dependent dielectric approximation ~Mehler & Solmajer, 1991!.
As the trajectories of comparable timescales ~i.e., 100 ps! yielded
the same results as the explicitly solvated runs with respect to Trp
dihedral angles, we acquired the longer time simulations with the
dd-E approach, thus considered satisfactory.
components to the total emission, respectively.
Computational methods
The starting coordinates for HIV-1 protease complexed to acetyl-
pepstatin ~pdb5hvp.ent, 2.0 Å, Fitzgerald et al., 1990! were re-
trieved from the Brookhaven Protein Data Bank ~Bernstein et al.,
1977!. Residues were replaced as required so as to generate the
Q7K0L33I0L63I mutant and the structure was subjected to energy
minimization using the Discover-3 module of the InsightII soft-
ware package ~MSI, San Diego California! on a SGI R10000
workstation, with the ESFF force field. Missing hydrogens were
added subject to van der Waals constraints and consistent with the
ionization state of charged R groups at the experimental pH, i.e.,
The Trp indole motion was analyzed by following the variations
of the Trp x and x dihedral angles. These angles are defined as
1
2
7.5. To remove artifacts due to the addition of explicit hydrogens,
1 2
follows: x ϭ N-CA-CB-CG and x ϭ CA-CB-CG-CD2 ~IUPAC-
energy minimization was performed using a conjugate gradient
IUB Commission on Biochemical Nomenclature, 1970!, and we
use them as indicators of indole rotation about the CA-CB and
CB-CG bonds, respectively ~see also Figs. 7, 8!. The stereochem-
algorithm until the average ~root-mean-square! energy derivative
Ϫ1 Ϫ1
reached 0.1 kcal mol
A . The HIV-1 protease structure without
inhibitor was modeled in two different ways, the first using the
.7 Å structure ~pdb3phv.ent, Lapatto et al., 1989! and the second
using the available coordinates of the Q7K0L33I0L63I mutant
pdb1a30.ent, 2.0 Å, Louis et al., 1998!. The mutant is bound to
ically favored rotamers of the x
1
conformation are the staggered
2
forms, possible when x ϭ 60, Ϫ60, and 1808; these rotamers are
1
ϩ
Ϫ
called the g , g , and t rotamers, respectively. Similarly, the fa-
~
vored x rotamers are the 908 and Ϫ908 forms ~Szabo & Rayner,
2
the Glu-Asp-Leu tripeptide, which was removed prior to generat-
ing the fully opened flap structure of the unbound protease. Flap
opening was achieved by template forcing the atomic coordinates
of the mutant structure of residues 1 to 25 and 65 to 99 in each
monomer to the coordinates of the “fully flap-opened” 3phv struc-
ture and minimizing as described above, following essentially the
approach described by Collins et al. ~1995! in their simulation of
the protease’s flap opening. The stereochemical quality of all struc-
tures was verified and corrected using the program PROCHECK
1980; Willis et al., 1994!.
Acknowledgments
The authors thank A. Demchenko ~Palladin Institute of Biochemistry, Kiev!
for valuable suggestions and for reading the manuscript and also W. Braun
~UTMB Galveston, Texas! for a copy of his FANTOM program and R.A.
Laskowski ~UC London, GB! for PROCHECK. The authors also thank J.
Gallay and M. Vincent for use of the fluorescence decay facilities at LURE,
Orsay ~France!. M.L. gratefully acknowledges support from Hungarian
postdoctoral grant MKM FKFP 119101997 ~J.F.!. B.U. thanks for support
from the PhD program no. 11 of the Semmelweis University, Budapest.
The technical assistance of Mrs. R. Markács is greatly appreciated.
3.4 ~Laskowski et al., 1993; Rullman, 1996!, which compares a
given structure to well-refined structures at the same resolution,
thus providing an indication of residue reliability in terms of ideal
bond lengths and angles, backbone f and w torsion angles, and Ca
chirality. It was felt that special care had to be paid in generating
optimal apoenzyme structures, since our objective was to run dy-
namics on both complexed and uncomplexed proteases while en-
suring that any observed differences in Trp dynamics between
them would not be due to structural artefacts and0or poor contacts
in the apoenzyme. The apo-HIV protease retained for the simula-
tions were the higher resolution mutated proteases, as PROCHECK
consistently returned better stereochemical parameters for these
structures. The solvent accessible surface areas ~SASA! of the
residues—defined as the surface generated by a sphere of the
radius of a water molecule of 1.4 Å rolled around the residue ~Lee
References
Axelsen PH, Gratton E, Prendergast FG. 1991. Experimentally verufying mo-
lecular dynamics simulations through fluorescence anisotropy measure-
ments. Biochemistry 30:1173–1179.
Bernstein FC, Koetzle TF, Williams GJB, Meyer EF Jr, Brice MD, Rodgers JR,
Kennard O, Shimanouchi T, Tasumi M. 1977. The Protein Data Bank: A
computer-based archival file for macromolecular structures. J Mol Biol
112:535–549.
Bismuto E, Irace G. 1994. Unfolding pathway of apomyoglobin. Simultaneous
characterization of acidic conformational states by frequency domain fluo-
rometry. J Mol Biol 241:103–109.
Callis PR. 1997. L
experimental observations to fluorescence of proteins. Methods Enzymol
78:113–150.
a b
and L transitions of tryptophan: Applications of theory and
&
Richards, 1971!—was calculated using FANTOM ~Fracz-
2
kiewicz & Braun, 1998!. The areas are expressed as ratios of the
indole surface area to the random coil value of Trp, defined as the
Trp SASA in the tripeptide Gly-Trp-Gly in an ensemble of 30
random conformations. Trp is considered solvent-exposed if R Ͼ
Chen Y, Barkley MD. 1998. Toward understanding tryptophan fluorescence in
proteins. Biochemistry 37:9976–9982.
Chen Y, Liu B, Yu HT, Barkley MD. 1996. The peptide bond quenches indole
fluorescence. J Am Chem Soc 118:9271–9278.
Collins JR, Burt SK, Erickson JW. 1995. Flap opening in HIV-1 protease sim-
ulated by “activated” molecular dynamics. Struct Biol 2:334–338.
De Beuckeleer K, Volckaert G, Engelborghs Y. 1999. Time resolved fluores-
cence and phosphorescence properties of the individual tryptophan residues
of Barnase: Evidence for protein-protein interactions. Proteins Struct Funct
Genet 36:42–53.
Eftink M. 1983. Quenching-resolved emisson anisotropy studies with single and
multitryptophan-containing proteins. Biophys J 43:323–334.
Eftink MR, Ghiron CA. 1976. Fluorescence quenching of indole and model
micelle systems. J Phys Chem 80:486–493.
50% and buried if R Ͻ 20%.
The molecular dynamics simulations were first run with solv-
ated structures. To generate bulk solvent, the structures were
explicitly solvated in PBC cell of water molecules ~6888 wa-
ters!, also retaining the water molecules of the X-ray structure.
A dielectric constant of 1 was used throughout the simulations.
First the 3D solvent cell was equilibrated, holding the protein
coordinates fixed, then the entire solvated system was subjected
to energy minimization.
Eftink MR, Ghiron CA. 1981. Fluorescence quenching studies with proteins.
Anal Biochem 114:199–227.

Tryptophan monitoring of HIV-1 protease dynamics, inhibitor binding
2245
Engh RA, Chen LXQ, Fleming GR. 1986. Conformational dynamics of tryp-
tophan: A proposal for the origin of the non-exponential fluorescence decay.
Chem Phys Lett 126:365–372.
Fitzgerald PMD, McKeever BM, VanMiddlesworth JF, Springer JP, Heimbach
JC, Leu ChT, Herber WK, Dixon RAF, Darke PL. 1990. Crystallographic
analysis of a complex between human immunodeficiency virus type 1 pro-
tease and acetyl-pepstatin at 2.0 Å resolution. J Biol Chem 265:14209–14219.
Flexner C. 1998. HIV-protease inhibitors. New Eng J Med 338:1281–92.
Fraczkiewicz R, Braun W. 1998. A new efficient algorithm for calculating
solvent accessible surface areas of macromolecules. J Comp Chem 19:319–
Ross JBA, Wyssbrod HR, Porter RA, Schwartz GP, Michaels CA, Laws WR.
1992. Correlation of tryptophan fluorescence intensity decay parameters
with 1H NMR-determined rotamer conformations: @Tryptophan #Oxytoxin.
Biochemistry 31:1585–1594.
Rullman JAC. 1996. AQUA, Computer program. Department of NMR Spec-
troscopy, Bijvoet Center for Biomolecular Research, Utrecht University,
The Netherlands.
Schmidt KE, Lee MA. 1991. Implementing the fast multipole method in three
dimensions. J Stat Phys 63:1223–1237.
Seelmeier S, Schmidt H, Turk V, von der Helm K. 1988. Human immunodefi-
ciency virus has an aspartic-type protease that can be inhibited by pepstatin
A. Proc Natl Acad Sci USA 85:6612–6616.
Sillen A, Diaz JF, Engelborghs Y. 2000. A step toward the prediction of the
fluorescence lifetimes of tryptophan residues in proteins based on structural
and spectral data. Protein Sci 9:158–169.
Silva ND, Prendergast FG. 1996. Tryptophan dynamics of the FK506 binding
protein: Time-resolved fluorescence and simulations. Biophys J 70:1122–
1137.
Szabo AG, Rayner DM. 1980. Fluorescence decay of tryptophan conformers in
aqueous solution. J Am Chem Soc 102:554–563.
Szeltner Z, Polgár L. 1996. Conformational stability and catalytic activity of
HIV-1 protease are both enhanced at high salt concentration. J Biol Chem
271:5458–5463.
2
3
33.
Hennecke J, Sillen A, Huber-Wunderlich M, Engelborghs Y, Glockshuber R.
997. Quenching of tryptophan fluorescence by the active-site disulfide
1
bridge in the DsbA protein from E. coli. Biochemistry 36:6391–6400.
Hutnik CM, Szabo AG. 1989. Confirmation that multiexponential fluorescence
decay behavior of holoazurin originates from conformational heterogeneity.
Biochemistry 28:3923–3934.
Kohl NE, Emini EA, Schleif WA, Davis LJ, Heimbach JC, Dixon RAF, Scolnick
EM, Sigal IS. 1988. Active human immunodeficiency virus protease is
required for viral infectivity. Proc Natl Acad Sci USA 85:4686–4690.
Kungl AJ, Visser NV, van Hoek A, Visser AJWG, Billich A, Schilk A, Gstach
H, Auer M. 1998. Time-resolved fluorescence anisotropy of HIV-1 protease
inhibitor complexes correlates with inhibitory activity. Biochemistry 37:
2
778–2786.
Laberge M, Vreugdenhill AJ, Vanderkooi J, Butler IS. 1998. Microperoxidase-
1: Molecular dynamics and q-band excited resonance raman of the oxi-
Tran CD, Beddard GS. 1985. Studies of the fluorescence from tryptophan in
melittin. Eur Biophys J 13:59–64.
Trylska J, Antosiewicz J, Geller M, Hodge CN, Klabe RM, Head MS, Gilson
MK. 1999. Thermodynamic linkage between the binding of protons and
inhibitors to HIV-1 protease. Protein Sci 8:180–195.
Tyagi SC, Simon SR, Carter CA. 1994. Effect of pH and non physiological salt
concentrations on human immunodeficiency virus HIV-1 protease dimer-
ization. Biochem Cell Biol 72:175–181.
Valeur B, Weber G. 1977. Resolution of the fluorescence excitation spectrum of
indole into the 1La and 1Lb excitation bands. Photochem Photobiol 25:
441–444.
Vergani B, Kintrup M, Hillen W, Lami H, Piemont E, Bombarda E, Alberti P,
Doglia SM, Chabbert M. 2000. Backbone dynamics of Tet repressor a80a9
loop. Biochemistry 39:2759–2768.
Vos R, Engelborghs Y. 1994. A fluorescence study of tryptophan-histidine in-
teraction in the peptide anantin and in solution. Photochem Photobiol
60:24–32.
Wang YX, Freedberg DI, Grzesiek S, Torchia DA, Wingfield PT, Kaufman JD,
Stahl SJ, Chang ChH, Hodge NC. 1996a. Mapping hydration water mol-
ecules in the HIV-1 protease0DMP323 complex in solution by NMR-
spectroscopy. Biochemistry 35:12694–12704.
1
dized, reduced and carbonyl forms. J Biomol Struct Funct 15:1039–1050.
Lakowicz JR. 1983. Principles of fluorescence spectroscopy. New York: Ple-
num. p 264.
Lapatto R, Blundell T, Hemmings A, Overington J, Wilderspin A, Wood S,
Merson JR, Whittle PJ, Danley DE, Geohegan KF, et al. 1989. X-ray analy-
sis of HIV-1 proteinase at 2.7-Å resolution confirms homology among ret-
roviral enzymes. Nature 342:299–302.
Laskowski RA, MacArthur MW, Moss DS, Thornthon JM. 1993. PROCHECK:
A program to check the stereochemical quality of protein structures. J Appl
Crystallogr 26:283–291.
Lee B, Richards FM. 1971. The interpretation of protein structures: Estimation
of static accessibility. J Mol Biol 55:379–400.
Lehrer SS. 1971. Solute perturbation of protein fluorescence. The quenching of
the tryptophyl fluorescence of model compounds and of lysozyme by iodide
ion. Biochemistry 10:3254–3263.
Louis JM, Dyda F, Nashed NT, Kimmel AR, Davies DR. 1998. Hydrophilic
peptides derived from the transframe region of gag-pol inhibit the HIV-1
protease. Biochemistry 37:2105–2113.
Mehler EL, Solmajer T. 1991. Electrostatic effects in proteins: comparison of
dielectric and charge models. Protein Eng 4:903–910.
Mildner AM, Rothrock DJ, Leone JW, Bannow CA, Lull JM, Reardon IM,
Sarcich JL, Howe WJ, Tomich C-SC, Smith CW, et al. 1994. The HIV-1
protease as enzyme and substrate: Mutagenesis of autolysis sites and gen-
eration of a stable mutant with retained kinetic properties. Biochemistry
Wang YX, Freedberg DI, Yamazaki,T, Wingfield PT, Stahl SJ, Kaufman JD,
Kiso Y, Torchia DA. 1996b. Solution NMR evidence that the HIV-1 protease
catalytic aspartyl groups have different ionization states in the complex
formed with the asymmetric drug KNI-272. Biochemistry 35:9945–
9950.
Willis KJ, Neugebauer W, Sikorska M, Szabo AG. 1994. Probing a-helical
secondary structure at a specific site in model peptides via restriction of
tryptophan side-chain rotamer conformation. Biophys J 66:1623–1630.
Wlodawer A, Erickson JW. 1993. Structure-based inhibitors of HIV-1 protease.
Annu Rev Biochem 62:543–585.
3
3:9405–9413.
Navia MA, Fitzgerald PMD, McKeever BM, Leu ChT, Heimbach JC, Herber
WK, Sigal IS, Darke PL, Springer JP. 1989. Three-dimensional structure of
aspartyl-protease from human immunodeficiency virus HIV-1. Nature
3
37:615–620.
Wlodawer A, Miller M, Jaskolski M, Sathyanarayana KB, Baldwin E, Weber TI,
Selk ML, Clawson L, Schneider J, Kent HBS. 1989. Conserved folding in
retroviral proteases: Crystal structure of a synthetic HIV-1 protease. Science
245:616–621.
Nicholson LK, Yamazaki T, Torchia DA, Grzesiek S, Bax A, Stahl SJ, Kaufman
JD, Wingfield PT, Lam PYS, Jadhav PK, et al. 1995. Flexibility and func-
tion in HIV-1 protease. Struct Biol 2:274–280.
Pargellis ChA, Morelock MM, Graham ET, Kinkade P, Pav S, Lubbe K, Lama-
rre D, Anderson PC. 1994. Determination of kinetic rate constants for the
binding of inhibitors to HIV-1 protease and for the association and disso-
ciation of active homodimer. Biochemistry 33:12527–12534.
Polgár L, Szeltner Z, Boros I. 1994. Substrate-dependent mechanisms in the
catalysis of human immunodeficiency virus protease. Biochemistry 33:9351–
Wlodawer A, Vondrasek J. 1998. Inhibitors of HIV-1 protease: A major success
of structure-assisted drug design. Annu Rev Biophys Biomol Struct 27:249–
284.
Yamazaki T, Nicholson LK, Torchia DA, Stahl SJ, Kaufman JD, Wingfield PT,
Domaille PJ, Campbell-Burk S. 1994. Secondary structure and signal as-
signments of human-immunodeficiency-virus-1 protease complexed to a
novel, structure-based inhibitor. Eur J Biochem 219:707–712.
Yengo CM, Chrin L, Rovner AS, Berger CL. 1999. Intrinsic tryptophan
fluorescence identifies specific conformational changes at the actomyosin
interface upon action binding and ADP release. Biochemistry 38:14515–
14523.
9
357.
Ringhofer S, Kallen J, Dutzler R, Billich A, Visser AJWG, Scholz D, Stein-
hauser O, Schreiber H, Auer M, Kungl AJ. 1999. X-ray structure and con-
formational dynamics of the HIV-I protease in complex with the inhibitor
SDZ283-910: Agreement of time-resolved spectroscopy and molecular dy-
namics simulations. J Mol Biol 286:1147–1159.
York DM, Darden TA, Pedersen LG, Anderson MW. 1993. Molecular-dynamic
simulation of HIV-1 protease in a crystalline environment and in solution.
Biochemistry 32:1443–1453.
Rosé JK, Salto R, Craik CS. 1993. Regulation of autoproteolysis of the HIV-1
and HIV-2 proteases with engineered amino acid substitutions. J Biol Chem
2
68:11939–11945.