Characterization of the human cytomegalovirus protease as an induced- fit serine protease and the implications to the design of mechanism-based inhibitors

Article abstract of DOI:10.1021/ja983904h

The conformational properties of the N-tert-butylacetyl-L-tert- butylglycyl-L-N(δ),N(δ)-dimethylasparagyl-L-alanyl methyl ketone (MK) 1 and its terminal N-isopropylacetyl analogue 2 were investigated. Whereas these compounds are weak (mM IC₅₀ range) inhibitors of the human cytomegalovirus (HCMV) protease, their activated carbonyl analogues are > 1000-fold more potent (e.g., trifluoromethyl ketone 3, IC₅₀ = 1.1 μM). A combination of NMR techniques demonstrated that MK 2 exists in solution as a relatively rigid and extended peptide structure and that the bulky side chains, notably the P3 tert-butyl group, greatly contribute to maintaining this solution conformation. Furthermore, transferred nuclear Overhauser effect (TRNOE) studies provided an enzyme-bound conformation of MK 2 that was found to be similar to its free solution structure and compares very well to the X-ray crystallographic structure of a related peptidyl inhibitor complexed to the enzyme. The fact that ligands such as MK 2 exist in solution in the bioactive conformation accounts, in part, for the observed inhibitory activity of activated ketone inhibitors bearing comparable peptidyl sequences. Comparison of the X-ray structures of HCMV protease apoenzyme and that of its complex with a related peptidyl α-ketoamide inhibitor allowed for a detailed analysis of the previously reported conformational change of the enzyme upon complexation of inhibitors such as 1 and 3. The above observations indicate that HCMV protease is a novel example of a serine protease that operates by an induced-fit mechanism for which complexation of peptidyl ligands results in structural changes which bring the enzyme to a catalytically active (or optimized) form. Kinetic and fluorescence studies are also consistent with an induced-fit mechanism in which a considerable proportion of the intrinsic ligand-binding energy is used to carry out the conformational reorganization of the protease. Issues related to the rational design of both mechanism- and nonmechanism-based inhibitors of HCMV protease, notably in light of the peptidyl ligand-induced optimization of its catalytic functioning, are discussed.

Full text of DOI:10.1021/ja983904h

2974
J. Am. Chem. Soc. 1999, 121, 2974-2986
Characterization of the Human Cytomegalovirus Protease As an
Induced-Fit Serine Protease and the Implications to the Design of
Mechanism-Based Inhibitors
Steven R. LaPlante, Pierre R. Bonneau, Norman Aubry, Dale R. Cameron, Robert De´ziel,
Chantal Grand-Maˆıtre, Ce´line Plouffe, Liang Tong,† and Stephen H. Kawai*
Contribution from the Departments of Chemistry and Biological Sciences, Boehringer Ingelheim (Canada)
Ltd., Bio-Me´ga Research DiVision, LaVal, Quebec H7S 2G5, Canada
ReceiVed NoVember 10, 1998
Abstract: The conformational properties of the N-tert-butylacetyl-L-tert-butylglycyl-L-Nδ,Nδ-dimethylasparagyl-
L-alanyl methyl ketone (MK) 1 and its terminal N-isopropylacetyl analogue 2 were investigated. Whereas
these compounds are weak (mM IC50 range) inhibitors of the human cytomegalovirus (HCMV) protease, their
activated carbonyl analogues are >1000-fold more potent (e.g., trifluoromethyl ketone 3, IC50 ) 1.1 µM). A
combination of NMR techniques demonstrated that MK 2 exists in solution as a relatively rigid and extended
peptide structure and that the bulky side chains, notably the P3 tert-butyl group, greatly contribute to maintaining
this solution conformation. Furthermore, transferred nuclear Overhauser effect (TRNOE) studies provided an
enzyme-bound conformation of MK 2 that was found to be similar to its free solution structure and compares
very well to the X-ray crystallographic structure of a related peptidyl inhibitor complexed to the enzyme. The
fact that ligands such as MK 2 exist in solution in the bioactive conformation accounts, in part, for the observed
inhibitory activity of activated ketone inhibitors bearing comparable peptidyl sequences. Comparison of the
X-ray structures of HCMV protease apoenzyme and that of its complex with a related peptidyl R-ketoamide
inhibitor allowed for a detailed analysis of the previously reported conformational change of the enzyme upon
complexation of inhibitors such as 1 and 3. The above observations indicate that HCMV protease is a novel
example of a serine protease that operates by an induced-fit mechanism for which complexation of peptidyl
ligands results in structural changes which bring the enzyme to a catalytically active (or optimized) form.
Kinetic and fluorescence studies are also consistent with an induced-fit mechanism in which a considerable
proportion of the intrinsic ligand-binding energy is used to carry out the conformational reorganization of the
protease. Issues related to the rational design of both mechanism- and nonmechanism-based inhibitors of HCMV
protease, notably in light of the peptidyl ligand-induced optimization of its catalytic functioning, are discussed.
Introduction
paid to these enzymes as therapeutic targets for the development
of novel antiherpetic agents.5,6 This is notably the case of the
highly prevalent human cytomegalovirus (HCMV),7 for which
existing antiviral treatments suffer from toxicity and limited
effectiveness.8
The quest for new inhibitors of HCMV protease has been
paralleled by much activity focusing on the structural biology
of the target, which has underlined the highly novel nature of
this enzyme. X-ray crystallographic studies9,10 have revealed
HCMV protease to possess a protein fold previously unseen
among serine proteases as well as a unique catalytic triad in
(5) Howerda, B. C. AntiViral Res. 1997, 35, 1-21.
(6) Gibson, W.; Welch, A. R.; Hall, M. R. T. Perspect. Drug DiscoV.
Des. 1994, 2, 413-426.
(7) Mocarski, E. S., Jr. In Virology; Fields, B. N., Ed.; Lippencott-
Raven: Philadelphia, PA, 1995; Vol. 2, pp 2447-2492. Britt, W. J.; Alford,
C. A. In Virology; Fields, B. N., Ed.; Lippencott-Raven: Philadelphia, PA,
1995; Vol. 2, pp 2493-2525.
(8) Goodgame, R. W. Ann. Intern. Med. 1993, 119, 924-935.
(9) Tong, L.; Qian, C.; Massariol, M.-J.; Bonneau, P.; Cordingley, M.;
Lagace´, L. Nature 1996, 383, 272-275.
(10) Qiu, X.; Culp, J. S.; DiLella, A. G.; Hellmig, B.; Hoog, S. S.; Janson,
C. A.; Smith, W. W.; Abdel-Meguid, S. S. Nature 1996, 383, 275-279.
Shieh, H.-S.; Kurumball, R. G.; Stevens, A. M.; Stegeman, R. A.; Sturman,
E. J.; Pak, J. Y.; Wittwer, A. J.; Palmier, M. O.; Wiegand, R. C.; Holwerda,
B. C.; Stallings, W. C. Nature 1996, 383, 279-282. Chen. P.; Tsuge, H.;
Almassy, R. J.; Gribskov, C. L.; Katoh, S.; Vanderpool, D. L.; Margosiak,
S. A.; Pinko, C.; Kan, C.-C. Cell 1996, 86, 835-843.
Serine proteases are perhaps the most well-understood family
of enzymes to date1,2 whose study has played a very large role
in furthering our comprehension of the functioning of enzymes
in general. Much effort is presently being directed toward a
particular group of serine proteases, the herpesvirus proteases
(or assemblins), whose activity is essential for proper assembly
of the capsid3 and, as a consequence, viral replication.4 It is
therefore not surprising that a great deal of attention is being
* To whom correspondence should be addressed. Phone: (450) 682-
4640. Fax: (450) 682-8434.
† Present address: Department of Biological Sciences, Columbia Uni-
versity, New York, New York 10027.
(1) Polgar, L., Ed. Mechanisms of Protease Action; CRC Press: Roca
Baton, FL, 1989.
(2) Fersht, A. Enzyme Structure and Mechanism, 2nd ed.; W. H.
Freeman: New York, NY, 1985.
(3) Trus, B. L.; Booy, F. P.; Newcomb, W. W.; Brown, J. C.; Homa, F.
L.; Thomsen, D. R.; Steven, A. C. J. Mol. Biol. 1996, 263, 447-462.
Newcomb, W. W.; Homa, F. L.; Thomsen, D. R.; Booy, F. P.; Trus, B. L.;
Steven, A. C.; Spencer, J. V.; Brown, J. C. J. Mol. Biol. 1996, 263, 447-
462.
(4) Gao, M.; Matusick-Kumar, L.; Hurlburt, W.; DiTusa, S. F.; New-
comb, W. W.; Brown, J. C.; McCann, P. J., III; Deckman, I.; Colonno, R.
J. J. Virol. 1994, 68, 3702-3712. Matusick-Kumar, L.; McCann, P. J., III;
Robertson, B. J.; Newcomb, W. W.; Brown, J. C.; Gao, M. J. Virol. 1995,
69, 7113-7121.
10.1021/ja983904h CCC: $18.00 © 1999 American Chemical Society
Published on Web 03/19/1999

Characterization of the Human CytomegaloVirus Protease
J. Am. Chem. Soc., Vol. 121, No. 13, 1999 2975
the conformation of the former in the complexed state. We
recently described such studies involving both M- and R-site
sequence N-terminal cleavage products of HCMV protease.17
In the present work, we provide details of our elucidation of an
NMR structure of methyl ketone 2, a probe that was designed
to conform to the kinetic constraints of the TRNOE method. A
combination of NMR methods were also employed to determine
the conformational properties of MK 2 and related compounds
in solution. These studies revealed that the predominant solution
conformation of ketone 2 is very similar to that when complexed
to HCMV protease, a property that is undoubtedly key to the
observed binding of the optimized peptidyl sequence.
Figure 1. Nomenclature used in this study.
Table 1. Peptidyl Methyl and Activated Ketone Inhibitors of
HCMV Protease
The recent disclosure18 of the X-ray crystallographic structure
of the related R-ketoamide inhibitor 18 complexed to HCMV
protease allowed for comparison of enzyme-bound inhibitor
structures derived by NMR and X-ray diffraction. Perhaps more
importantly, it also permitted us to compare the structures of
the free and inhibited protease and define the ligand-induced
conformational changes. These analyses of both E19 and I led
us to conclude that HCMV protease is a unique example of a
serine protease that clearly operates by an induced-fit mecha-
nism. Consistent with this, kinetic studies indicate that factors
(such as salt) which stabilize forms of the enzyme which
approach the catalytically activated species reduce the proportion
of the intrinsic binding energy used to effect the conformational
change, resulting in stronger complexation of peptidyl ligands.
The implications of these factors in the design of inhibitors,
both mechanism- and non-mechanism-based, are addressed.
a Compounds 3, 4, and 5 are epimerized at the P1 R-center.
which the third member is a histidine residue. Confirming
biophysical studies, the enzyme has also revealed itself to exist
as a dimer that is believed to be the sole active form.11 We
have recently provided spectroscopic evidence that the protease
is an induced-fit enzyme, undergoing a conformational change
upon complexation of substrate-based peptidyl inhibitors.12
In our efforts to develop potent inhibitors of HCMV protease,
a large number of substrate-based, activated peptidyl ketones
were synthesized.13 Among the most active were those bearing
Results and Discussion
Design and Synthesis of Peptidyl Methyl Ketones. (a)
Design of Noncovalent Peptidyl Ketone Inhibitors. Peptidyl
activated carbonyl compounds20 represent a very large class of
substrate-based serine protease inhibitors comprised of a peptidic
chain and an electrophilic keto-functionality replacing the
carbonyl group of the scissile peptide bond. Examples include
trifluoromethyl ketones21-23 (TFMKs), pentafluoroethyl ke-
tones24 (PFEKs), and R-ketoamides,25 and among the many
enzymes inhibited by this class of molecules, their action against
chymotrypsin,22,26 elastase,27 and thrombin28 has been most
studied. The generally accepted mechanism of competitive
inhibition is depicted in Figure 2 with the example of a peptidyl
a
tert-butylacetyl-L-tert-butylglycyl-L-dimethylasparagyl-L-
alanyl sequence (which mimics the P4-P114 Val-Val-Asn-Ala
sequence of the natural Maturational (M-) cleavage site) flanking
the activated carbonyl center (Figure 1). The IC50 values, in
the high nanomolar range, of these molecules are provided in
Table 1. As part of our efforts to obtain a detailed understanding
of the workings of HCMV protease and the structural basis of
its inhibition by the above-mentioned compounds, we looked
to NMR15 and, in particular, the transferred nuclear Overhauser
effect (TRNOE) experiment as a tool of choice.16 This powerful
NMR technique utilizes the rapid reversible binding of a ligand
in solution to a macromolecule to obtain information concerning
(17) LaPlante, S. R.; Aubry, N.; Bonneau, P. R.; Cameron, D. R.; Lagace´,
L.; Massariol, M.-J.; Montpetit, H.; Plouffe, C.; Kawai, S. H.; Fulton, B.
D.; Chen, Z. G.; Ni, F. Biochemistry 1998, 37, 9793-9801.
(18) Tong, L.; Qian, C.; Massariol, M.-J.; De´ziel, R.; Yoakim, C.; Lagace´,
L. Nature Struct. Biol. 1998, 5, 819-826.
(19) The following abbreviations are used throughout the text: E,
enzyme; I, inhibitor; S, substrate; ES, Michaelis complex; EP, enzyme-
product complex; E:I, enzyme-inhibitor complex.
(20) Mehdi, S. Bioorg. Chem. 1993, 21, 249-259.
(11) Margosiak, S. A.; Vanderpool, D. L.; Sisson, W.; Pinko, C.; Kan,
C.-C. Biochemistry 1996, 35, 5300-5307. Darke, P. L.; Cole, J. L.;
Waxman, L.; Hall, D. L.; Sardana, M. K.; Kuo, L. C. J. Biol. Chem. 1996,
271, 7445-7449.
(21) Imperiali, B.; Abeles, R. H. Biochemistry 1986, 25, 3760-3767.
Peet, N. P.; Burkhart, J. P.; Angelastro, M. R.; Giroux, E. L.; Mehdi, S.;
Bey, P.; Kolb, M.; Neises, B.; Schirlin, D. J. Med. Chem. 1990, 33, 394-
407.
(22) Brady, K.; Abeles, R. H. Biochemistry 1990, 29, 7608-7617.
(23) Xe¨ı, C. V. Biochemistry 1997, 36, 1308-1313.
(24) Angelastro, M. R.; Baugh, L. E.; Bey, P.; Burkhart, J. P.; Chen,
T.-M.; Durham, S. L.; Hare, C. M.; Huber, E. W.; Janusz, M. J.; Koehl, J.
R.; Marquart, A. L.; Mehdi, S.; Peet, N. P. J. Med. Chem. 1994, 37, 4538-
4554.
(25) Hu, L.-Y.; Abeles, R. H. Arch. Biochem. Biophys. 1990, 281, 271-
274. Brady, S. F.; Sisko, J. T.; Stauffer, K. J.; Colton, C. D.; Qiu, H.; Lewis,
S. D.; Ng, A. S.; Shafer, J. A.; Bogusky, M. J.; Veber, D. F.; Nutt, R. F.
Bioorg. Med. Chem. 1995, 3, 1063-1078.
(26) Liang, T.-C.; Abeles, R. H. Biochemistry 1987, 26, 7603-7608.
(27) Edwards, P. D.; Bernstein, P. R. Medicinal Res. ReV. 1994, 14, 127-
194.
(12) Bonneau, P. R.; Grand-Maˆıtre, C.; Greenwood, D. J.; Lagace´, L.;
LaPlante, S. R.; Massariol, M.-J.; Ogilvie, W. W.; O’Meara, J. A.; Kawai,
S. H. Biochemistry 1997, 36, 12644-12652.
(13) Ogilvie, W.; Bailey, M.; Poupart, M.-A.; Abraham, A.; Bhavsar,
A.; Bonneau, P.; Bordeleau, J.; Chabot, C.; Duceppe, J.-S.; Fazal, G.; Goulet,
S.; Grand-Maitre, C.; Guse, I.; Halmos, T.; Lavalle´e, P.; Leach, M.;
Malenfant, E.; O’Meara, J.; Plante, R.; Plouffe, C.; Poirier, M.; Soucy, F.;
Yoakim, C.; De´ziel, R. J. Med. Chem. 1997, 40, 4113-4135.
(14) The peptide/peptidyl inhibitor and subsite nomenclature used is that
of: Schechter, I.; Berger, A. Biochem. Biophys. Res. Commun. 1967, 27,
157-162.
(15) Wu¨thrich, K. NMR of Proteins and Nucleic Acids; John Wiley &
Sons: New York, NY, 1986. Lian, L. Y.; Barsukov, I. L.; Sutcliffe, M. J.;
Sze, K. H.; Roberts, G. C. K. Methods Enzymol. 1994, 239, 657-700.
(16) Ni, F. Prog. Nucl. Magn. Reson. 1994, 26, 517-590.
(28) Jones, D. M.; Atrash, B.; Ryder, H.; Teger-Nilsson, A.-C.; Gyzander,
E.; Szelke, M. J. Enzyme Inhib. 1995, 9, 43-60.

2976 J. Am. Chem. Soc., Vol. 121, No. 13, 1999
LaPlante et al.
Scheme 1. Synthesis of Peptidyl Methyl Ketones 1 and 2
Figure 2. Inhibition of a serine protease by a peptidyl trifluoromethyl
ketone.
feature of the present methyl ketones is the presence of the
ketonic CH3 group that gives rise to an additional NMR
resonance which could potentially provide further distance
information to aid definition of the conformation in the P1-P1′
region.
TFMK. Following the formation of an initial encounter complex
(analogous to the Michaelis complex), attack of the active site
serine hydroxyl on the reactive carbonyl center results in a
covalent hemiketal adduct which, by virtue of the electron-
withdrawing CF3 substituent and its resemblance to the transition
state, has an appreciable lifetime.
(b) Synthesis of Methyl Ketone Inhibitors 1 and 2.
Although a number of methods have been reported for the
transformation of R-amino acids and peptides to their equivalent
methyl ketones,30 the problems of epimerization at the R-carbon
and/or unsatisfactory yields prompted us to investigate Weinreb
methodogy31 (Scheme 1). Indeed, the alanyl ketone 9 could be
obtained from N-Boc-alanine via reaction of its Weinreb amide
8 with excess methyllithium in 94% yield, and which proceeded
without any detectable racemization of the R-center. Ketone 9
was then reduced with borohydride to a mixture of diastereo-
meric alcohols 10 which were subsequently protected by
benzylation to give 11. Attempted peptide coupling of either
ketone 9 or amino alcohols 10 following N-deprotection proved
unsatisfactory.
Standard peptide coupling employing 1-hydroxybenzotriazole
(HOBT) and 1-(3-dimethylaminopropyl)-3-ethylcarbodiimide
hydrochloride (EDAC) was then used to elongate the peptide
chain from compounds 11, each step proceeding in very good
yield (86-94%). The coupling of the P2 dimethylasparagine
unit was carried out at 0 °C to reduce the epimerization we
observed in related work. Quantitative removal of the benzyl
groups of 14 and 16 by hydrogenation over palladium, followed
by Dess-Martin oxidation of the resultant alcohols, afforded
the target peptidyl methyl ketones 1 and 2, respectively.
NMR Studies. (a) Conformation of Free Peptidyl Ligands.
In contrast to proteins, small peptides are generally viewed as
While the transferred nuclear Overhauser effect (TRNOE)
method is clearly an expedient route to obtaining the bound
conformation of a protein-complexed ligand, a key constraint
of the technique is that the reversible exchange between the
free and bound inhibitor be rapid enough to allow for cross
relaxation. While activated carbonyl compounds such as 3, 4,
and 5 are reversible inhibitors, the exchange rates between their
free and enzyme-bound states are very slow due to their
mechanism of inactivation of the enzyme (covalent modifica-
tion), confering low koff rates (typically 7 × 10-5 s-1 for a
peptidyl TFMK inhibitor of the present enzyme). Moreover, in
the case of trifluoromethyl and pentafluoroethyl ketones, the
high degree of hydration in the free state results in “slow-
binding” kinetics.29 As a consequence, TRNOE methods cannot
be employed to determine the bound conformation of these
inhibitors.
To circumvent the kinetic constraints of the TRNOE method,
we sought to design a ligand that as closely as possible
resembled our more potent inhibitors while not exhibiting its
slow exchange properties. The peptidyl methyl ketones 1 and
2 clearly fulfill these criteria. Although close structural analogues
of TFMK 3, they are not hydrated in aqueous solution and do
not form hemiketal adducts with the active site serine of HCMV
protease. This latter conclusion is based on our NMR studies
involving 13C carbonyl-labeled 1 which clearly demonstrated
its unreactivity toward attack by the active site serine.12 Another
(30) Thompson, R. C. Biochemistry 1974, 13, 5495-5501. Mansour, T.
S.; Evans, C. A. Syn. Commun. 1989, 19, 667-672. Klix, R. C.; Chamberlin,
S. A.; Bhatia, A. V.; Davis, D. A.; Hayes, T. K.; Rojas, F. G.; Koops, R.
W. Tetrahedron Lett. 1995, 36, 1791-1794. McMurray, J. S.; Dyckes, D.
F. J. Org. Chem. 1985, 50, 1112-1115. Myers, A. G.; Yoon, T. Tetrahedron
Lett. 1995, 36, 9429-9432.
(29) Stein, R. L.; Strimpler, A. M.; Edwards, P. D.; Lewis, J. J.; Mauger,
R. C.; Schwartz, J. A.; Stein, M. M.; Trainor, D. A.; Wildonger, R. A.;
Zottola, M. A. Biochemistry 1987, 26, 2682-2689. Brady, K.; Abeles, R.
H. Biochemistry 1990, 29, 7608-7617.
(31) Nahm, S.; Weinreb, S. M. Tetrahedron Lett. 1981, 22, 3815-1818.

Characterization of the Human CytomegaloVirus Protease
J. Am. Chem. Soc., Vol. 121, No. 13, 1999 2977
Table 2. Substitutions of the P3 Side Chain of Pentafluoroethyl
Ketone Inhibitors
Figure 3. (A) Effect of HCMV protease on the P1-âCH3 doublet of
MK 1 (400 MHz): (blue) 2 mM MK 2, (red) after the addition of
protease (2:protease ) 7:1), and (black) after the subsequent addition
of PFEK 4 (2 equiv with respect to protease). (B) Effect of HCMV
protease on the P4-γCH3 doublet of MK 2 (400 MHz): (blue) 3.3 mM
2, (red) after the addition of protease (2:protease ) 10:1), (black) after
subsequent addition of â-lactam 23 (2 equiv with respect to protease;
see Table 5 for structure). Black traces have been displaced for the
sake of clarity.
Table 3. 1H Chemical Shift and Coupling Constant Data for
Inhibitors 1, 2, 4, 6, and 7 in H2O/DMSOa
chemical shift (ppm)
proton
P1 NH
MK (1) MK (2) PFEK (4) PFEK (6) PFEK (7)
7.96
2.03
4.11
1.13
8.17
4.56
7.95
2.03
4.11
1.13
8.19
4.56
7.42
7.43
7.58
COCH3
R-H
â-CH3
4.13
1.04
8.21
4.52
4.14
1.05
7.99
4.52
4.15
1.07
7.98
4.55
P2 NH
R-H
â-CH2
ꢀ-CH3
2.65/2.72 2.65/2.72 2.63/2.70 2.62/2.71 2.61/2.71
2.76/2.92 2.76/2.92 2.75/2.90 2.77/2.91 2.71/2.91
P3 NH
7.61
4.04
7.74
4.06
7.63
4.07
8.02
4.14
1.16
8.15
3.62/3.66
R-H
â-CH3
γ-CH3
P4 R-HA
R-HB
0.88
2.00
2.15
0.87
2.01
2.10
1.91
0.83
0.86
1.98
2.14
1.96
1.99
1.98
2.01
â-CH
γ-CH3
0.91
0.90
0.91
0.91
3J(H)-(H) Coupling Constant (Hz)
P1 NH-R-H 6.2
6.7
7.2
7.7
6.1
6.1
8.2
9.3
6.8
7.6
7
9.3
6.7
7.9
6.5
5.6
7.0
9.3
6.8
7.9
6
R-H-â-H 7.2
P2 NH-R-H 7.7
R-H-â-HA 6.1
R-H-â-HB 6.1
6.1
8.6
6
P3 NH-R-H 8.2
5.8/5.8
a Values reported for the hydrated forms of inhibitors 4, 6, and 7.
MK 2 exists predominantly in an extended peptide conformation.
In addition, a weak but detectable correlation between the P2
side chain and N-terminal isopropyl methyl groups (data not
shown) is consistent with a certain population existing such that
the P4 group is oriented toward the same side of the peptide
backbone as the P2 side chain.
Figure 4. Comparison of 2D spectra of free versus bound MK 2 (all
experiments carried out in aqueous buffer/750 MHz/283 K/3.1 mM
2): (A) NH-aliphatic region of the transferred NOESY spectrum in
the presence of HCMV protease (2:protease ) 7:1/150 ms mixing time),
(B) ROESY spectrum in the absence of protease (125 ms spin-lock
mixing time).
Selected proton NMR parameters for methyl ketones 1 and
2 are listed in Table 3 and, as one would expect, consist of two
very similar sets of values. Of particular interest are the
3
magnitudes of JNH-RH. Consistent with the conformational
being devoid of defined structure in aqueous solution, their NMR
data representing an average of all of the solution conformations
present. Nonetheless, it has been demonstrated through a
combination of NMR techniques that certain short peptide chains
do exist predominantly in one particular conformation.32 The
solution structure of peptidyl ketone 2 was investigated through
rotating-frame NOE (ROESY) methods. The NH-aliphatic
region of the 2D spectrum is shown in Figure 4B and indicates
that the molecule is indeed structured. A clear feature of the
ROESY data is a characteristic pattern of weak intramolecular
[CRH(i)-NH(i)] correlations alternating with intense interresidue
[CRH(i)-NH(i + 1)] signals. This, together with the absence of
any cross-peaks between adjacent NH protons, indicates that
picture derived from the ROESY data, the coupling constants
for P2 and P3 indicate a predominantly extended peptide
structure. The values also compare well to those obtained for
the PFEK analogue of methyl ketone 1 (compound 4). The
difference in the case of P1 is no doubt related to the fact that
the activated pentafluoroethyl ketones are sterically congested
due to hydration whereas the methyl ketones 1 and 2 are not.
We ascribe the smaller coupling constants to substantial
contributions by nonextended conformers.
The fact that compounds bearing the peptidyl sequence of 1
exist predominantly as extended structures is underlined by
comparing their conformational properties with analogues
lacking the bulky side chain at the P3 position for which one
would predict a much greater degree of flexibility in solution.
A series of NMR experiments, performed in 15% H2O in
DMSO-d6 for reasons of solubility, were carried out for the
(32) Dyson, H. J.; Cross, K. J.; Houghten, R. A.; Wilson, I. A.; Wright,
P. E.; Lerner, R. A. Nature 1985, 318, 480-483. Dyson, H. J.; Rance, M.;
Houghten, R. A.; Lerner, R. A.; Wright, P. E. J. Mol. Biol. 1988, 201,
161-200.

2978 J. Am. Chem. Soc., Vol. 121, No. 13, 1999
LaPlante et al.
Table 4. 13C Chemical Shift and NT1 Values of PFEK Inhibitors 4, 6, and 7 in H2O/DMSOa
PFEK (4)
PFEK (6)
PFEK (7)
carbon
R-CH
δ (ppm)
NT1 (s)
δ (ppm)
NT1 (s)
0.3b
1.74
0.32
0.46
1.73
2.12/3.00
1.20
0.37
2.46
δ (ppm)
NT1 (s)
0.3b
1.62
0.30
n.d.
1.85
2.74/3.08
1.04
P1
P2
48.78
15.37
50.27
34.48
170.30
35.51/37.32
170.92
60.96
34.24
27.09
171.29
48.78
31.15
0.30
1.74
0.29
0.38
1.65
2.34/2.76
1.04
0.33
2.69
2.61
0.94
0.58
4.39
3.69
48.78
15.01
49.82
34.87
170.13
35.30/37.15
170.88
48.64
48.64
14.82
49.50
â-CH3
R-CH
â-CH2
γ-CdO
ꢀ-CH3
CdO
R-CH
â-C
γ-CH3
CdO
R-CH2
â-C
γ-CH3
CdO
34.81
169.77
34.98/36.80
170.74
42.29
P3
P4
0.50
17.96
172.83
48.78
30.88
29.96
172.33
1.02
0.7b
4.90
4.50
1.31
169.20
48.64
30.51
29.64
172.28
1.22
0.9b
5.07
4.77
1.47
30.16
172.54
1.02
a N ) number of attached hydrogens, T1 ) longitudinal relaxation time. Values reported for the hydrated forms of the inhibitors. In the case of
carbonyl and quaternary carbons, T1 values are given. b Estimated values (see Supporting Information). The direct measurement of T1 was precluded
by the overlap of the P1 and P4 signals.
series of pentafluoroethyl ketones 4, 6, and 7 (Table 2). The
ROESY spectrum of compound 4 (data not shown) was, not
surprisingly, similar to that of MK 2 in aqueous solution, being
consistent with an extended backbone conformation. Compari-
son of the ROESY spectra of inhibitors 6 and 7 revealed marked
differences upon the reduction of the steric bulk at P3. Most
notable was a sharp decrease in the relative intensities of
particular cross-peaks, notably the volumes of the interresidue
NH-CRH correlations. In the cases of the P1-NH to P2-RH,
P2-NH to P3-RH, and P3-NH to P4-RCH2 cross-peaks, the
relative signal volumes measured for inhibitors 6 and 7 with
respect to those of compound 4 were 74 and 70%, 47 and 56%
(two protons), and 57 and 48%, respectively. Globally, the
comparative analysis of the ROESY spectra argue strongly for
the view that the peptidyl chains of 1, 2, and 4 are fairly rigid
extended structures and that these conformational properties are
maintained to a considerable degree by the P3 tert-butyl
substituent.
to P3 with a higher degree of mobility for the P4 tert-butylacetyl
group. Removal of steric bulk at the P3 position results in an
overall increase in the NT1 values for all of the carbon atoms
from the P3-residue onward and, notably, for the N-terminal
acyl moeity. As one would expect, this rise in conformational
freedom is most pronounced for the P3-glycyl-substituted
compound 7.
(b) Protease-Bound Conformation of MK 2.35 Our initial
studies toward the determination of the bound conformation of
the present peptidyl methyl ketones focused on clearly establish-
ing that they indeed bind to the active site of HCMV protease
in a specific fashion. Upon addition of the enzyme to aqueous
samples of MK 1, uniformly broadened proton resonances of
the ketone were observed. These changes in peak form were
dependent on the protein/ligand ratio, indicative of rapid
reversible complexation to the protease. The specific nature of
the binding was verified by the addition of the potent and slow-
exchanging PFEK inhibitor 4, which resulted in elimination of
line-broadening in the MK signals (Figure 3A). In fact, the
disappearance of the enzyme-induced line-broadening of MK
1 resonances, due to its displacement from the protease by a
more potent competitive inhibitor, was later used to investigate
the action of nonpeptidic inhibitors of HCMV protease. Figure
3B presents the results of such a study involving the â-lactam
23, and demonstrates that this inhibitor acts by blocking access
to the active site. Analogous 2D competition experiments were
also carried out (data provided in the Supporting Information).
Numerous negative correlations in the NOESY spectrum of MK
1 in the presence of HCMV protease were found to disappear
upon the addition of PFEK 4, providing unequivocal evidence
that the binding of the methyl ketone is specific in nature.
While many of the initial NMR experiments were carried
out for MK 1, it was evident that the overlap of the two tert-
butyl singlets in its proton spectrum would limit the information
that could be provided. It was in considering this fact that
TRNOE studies were carried out for the P4 isopropylacetyl
inhibitor 2 for which this unfortuitous signal overlap is not
present. Since the intensities (volumes) of TRNOE cross-peaks
are proportional to interproton distances ( r-6), they provide
valuable distance information concerning the conformation of
Selected coupling constants and chemical shifts for the PFEKs
are also listed in Table 3. These values correspond to the
hydrated forms of the inhibitors which predominate in solution.
3
The JNH-RH value for P3 is also consistent with an extended
conformation for 4 (8.6 Hz). We interpret the value of 7.0 Hz
for inhibitor 6 as being due to a certain contribution by
nonextended populations. In contrast, the values of 5.8 for the
P3-Gly-substituted inhibitor 7 point to a freely rotating φ bond.
The chemical shift data, in particular the R-CH2 signals of the
N-terminal acyl group, also reveal differences within the series
of PFEKs. Whereas the shifts of the diastereotopic protons are
very similar (∆δ ) 0.03 ppm) for molecules 6 and 7, they are
separated by 0.16 ppm in the case of inhibitor 4, indicating that
the two exist in distinct magnetic environments and suggesting
that rotation about the ψ bond is much more restricted in the
latter, the terminal acyl group having a preferred orientation.
Finally, carbon-13 T1 values of the three compounds were
measured to further evaluate the dynamic properties of the
compounds (Table 4). NT1 values (the product of the number
of attached protons and the longitudinal relaxation time) provide
information concerning the mobility of a molecule33 in solution
and have been used to investigate the conformational dynamics
of peptidic compounds.34 Examination of the NT1 values of the
R-carbons of PFEK 4 reveals little segmental motion along P1
(34) Kessler, H.; Bats, J. W.; Griesinger, C.; Koll, S.; Will, M.; Wagner,
K. J. Am. Chem. Soc. 1988, 110, 1033-1049.
(35) Preliminary communication: LaPlante, S. R.; Cameron, D. R.; Aubry,
N.; Bonneau, P. R.; De´ziel, R.; Grand-Maˆıtre, C.; Ogilvie, W. W.; Kawai,
S. H. Angew. Chem., Int. Ed. Engl. 1998, 37, 2729-2732.
(33) Lyerla, J. R., Jr.; Levy, G. C. Top. Carbon-13 NMR Spectrosc. 1974,
1, 79-148.

Characterization of the Human CytomegaloVirus Protease
J. Am. Chem. Soc., Vol. 121, No. 13, 1999 2979
structure of the R-site product 20 was determined to be an
extended peptide. As one would anticipate, comparison with
the present structures generated for MK 2 (Figure 6) reveals
that the two families of conformations overlap very well,
confirming that substrate and inhibitor bind in the same mode.
Figure 7 also includes the structure of protease-complexed
R-ketoamide 18 taken from the X-ray structure of the inhibited
enzyme.18 Again, the extended peptide structures overlap well
from the P4 to P2 residues. The differences at P1 likely stem
from the fact that ketoamide 18 exists as the covalent adduct
with the active site serine whereas MK 2 is not a covalent
inhibitor and may interact differently with the enzyme in the
region of the catalytic serine and oxyanion hole.
It is clear, therefore, that the present activated peptidyl
inhibitors undergo little change in conformation between the
free and covalently linked states. This point is of importance
since it strongly implies that very little adjustment of the bound
inhibitor occurs over the course of the reorganization of the
protease upon inhibitor binding (see below). Otherwise, one
could question whether the NMR-derived structures presented
in Figure 5 represent the pre- or post-conformational change
structure of MK 2, or a virtual structure. TRNOE studies
involving ligands which themselves undergo conformational
exchange concurrent with complexation37 can be problematic
and necessitate the application of particular methods for the
proper handling of transferred NOESY data.38 This, however,
does not appear to be the case for the present system.
Mechanism of Inhibition. (a) Ligand-Induced Structural
Changes in HCMV Protease. Our previous fluorescence and
circular dichroism studies demonstrated that HCMV protease
undergoes a conformational change upon complexation of
inhibitors 1, 3, 4, and 5.12 A key observation to come out of
these investigations was that formation of a covalent adduct
between E and I is not required to induce the structural
alterations which are no doubt elicited by the formation of
contacts between the peptidyl portion of the inhibitors and
binding regions in the active site. Comparison of the X-ray
structures of HCMV protease apoenzyme9 and its covalent
complex with ketoamide 1818 allows for a structural analysis
of this conformational transition.
Many enzymes bear flexible surface loops which have been
observed in either of two conformations depending on whether
a ligand is present.39,40 These structures are believed to play a
role in forming the binding-pocket for the substrate and/or be
indirectly implicated in catalysis. In many cases, these loops
are “lid-hinges”40 of which a large portion is a rigid structure
that moves about a flexible joint. The conformational change
in HCMV protease is a much more complex process, involving
the movement and rigidification of a relatively large portion of
the active site region as shown in Figures 7 and 8.
Figure 5. Family of NMR-derived bound structures of MK 2. Twenty-
nine overlapping conformations are shown in stereoview.
a macromolecule-bound ligand. In the presence of HCMV
protease, the NOESY spectrum of a 12-fold excess of MK 2
presented numerous negative correlations. The volumes of the
signals were scaled and converted to interproton distance
restraints36 with use of the P1-RH to P1-âCH3 cross-peak to
correlate volumes with distances. Twenty-nine low-energy,
TRNOE-derived structures were generated by restrained simu-
lated annealing and are shown in Figure 5.
(c) Comparison of Free and Bound Structures of MK 2.
The investigations of the solution structure of MK 2 led to the
conclusion that the peptide is relatively structured, existing
predominantly in an extended conformation with the somewhat
mobile N-terminal acyl group oriented, at least part of the time,
toward the same side of the backbone as the P2 side chain.
This conformation corresponds very well to the NMR-derived
enzyme-bound structure of the ligand as presented in Figure 5.
The similarity between the protease-bound conformations of MK
2 and its solution structure is also clearly evidenced by
comparison of the ROESY spectrum of the free ligand and its
transferred NOESY spectrum in the presence of HCMV
protease. The corresponding NH-aliphatic regions of the two
traces are presented in Figure 4 to underline the similarity of
cross-relaxation within the free and protease-bound systems. A
more quantitative comparison of the normalized cross-peak
volumes (Supporting Information) emphasizes the very similar
backbone conformations as well as the fact that the N-terminal
product 20 is also bound in the same fashion.
Of equal interest is the appearance or, more importantly, the
increase in the relative intensities of certain NH to side-chain
signals in the TRNOE spectrum relative to the ROESY trace.
Globally, this is a result of the immobilization of certain groups
which exhibit a certain mobility in solution (e.g., P2 side chain
and ketonic methyl groups) and whose ROESY signals are not
observed. Other peaks are weaker due to the fact that only a
certain proportion of the molecules are present in the bioactive
conformation (not excluding the possibility of contributions from
other orientations). The fact that certain signals, such as the
P3-NH to P4-γCH3 and P3-NH to P4-âCH2 correlations, are
present in the ROESY spectrum is consistent with a nonneg-
ligible population of MK 2 existing in solution with the same
side-chain orientation as when protease bound.
On the basis of these comparisons, we conclude that MK 2
exists predominantly in the bioactive conformation when free
in solution, notably with respect to its backbone geometry, and
that complexation by HCMV protease occurs with very little
conformational adjustment within the ligand. This is certainly
the case as well for the much more potent activated carbonyl
analogues of MK 2 which bear essentially identical peptidyl
chains (i.e., inhibitors 3, 4, and 5).
The X-ray structure of HCMV protease inhibited by ketoa-
mide 18 revealed that the peptidyl portion of the ligand is bound
primarily through formation of an antiparallel â-sheet with the
â5-strand of the enzyme, a key interaction being between P3
and Ser-135. In the apoenzyme, this strand, which bears the
catalytic Ser-132, is more or less disordered from Ser-134
(37) Nicotra, M.; Paci, M.; Sette, M.; Oakley, A. J.; Parker, M. W.; Lo
Bello, M.; Caccuri, A. M.; Federici, G.; Ricci, G. Biochemistry 1998, 37,
3020-3027.
(38) Moseley, H. N. B.; Curto, E. V.; Krishna, N. R. J. Magn. Res. B
1995, 108, 243-261.
(39) Derewenda, U.; Brzozowski, A. M.; Lawson, D. M.; Derewenda,
Z. S. Biochemistry 1992, 31, 1532-1541. Jia, Z.; Barford, D.; Flint, A. J.;
Tonks, N. K. Science 1995, 268, 1754-1758. Wlodawer, A.; Erickson, J.
W. Annu. ReV. Biochem. 1993, 62, 543-585.
(40) Joseph, D.; Petsko, G. A.; Karplus, M. Science 1990, 249, 1425-
1428.
(d) Comparison of NMR and X-ray Structures. We
recently reported TRNOE studies involving HCMV protease
N-terminal cleavage products.17 The NMR-derived bound
(36) Baleja, J. D.; Moult, J.; Sykes, B. D. J. Magn. Res. B 1990, 87,
375-384.

2980 J. Am. Chem. Soc., Vol. 121, No. 13, 1999
LaPlante et al.
Figure 6. Comparison (stereoview) of NMR-derived bound conformations of MK 2 (green; 29 structures), R-peptide 20 (gray; 32 structures), and
the X-ray crystallographic structure of bound ketoamide 18 (yellow; 4 conformations).
Figure 7. Peptidyl ligand-induced conformational change of HCMV protease. Stereoviews of the active site region of the apoenzyme (above, in
gray) and of the inhibited enzyme (below, in green) covalently modified by ketoamide 18 (see Table 5 for chemical structure) shown in yellow.
Spheres indicate the points at which the L9-loop was no longer resolved. For the sake of clarity, single representative structures of the apoenzyme
(out of two) and of the complex (out of four) are presented. This applies as well to Figure 8.
through the L9 loop which is not resolved until Lys-156. After
superimposing the apoenzyme and inhibited protease structures,
the extent of the changes which occur upon ligand binding was
revealed and may be summarized as follows: (a) movement
and ordering of the â5-strand; (b) ordering of the L9-loop and
a large change in the position of the L2-loop such that a cleft
in which the P1 and P3 side chains find themselves is formed
(this involves formation of a putative double-salt bridge between
Arg-137, Glu-31, and Arg-165); and (c) repositioning of Arg-
165 and Arg-166. These residues constitute the oxyanion hole
whose structure is also altered upon ligand binding. Thus,
HCMV protease undergoes a substantial reorganization and,
notably, rigidification upon peptidyl ligand binding. It should
also be noted that the inhibitor-induced changes in the L9-loop
result in burying and immobilization of Trp-42, which in the
free enzyme is, more or less, solvent exposed and clearly mobile.
This change in microenvironment is responsible for the reported
shifts in fluorescence emission upon peptidyl ligand binding.12
A number of other structural alterations occur which have been
described18 and are of lesser importance in the present context.
While X-ray crystallography may provide “snapshots” of
discrete states (i.e., apoenzyme and enzyme-ligand complex)
in a crystal lattice which allows for the identification of ligand-
induced alterations in protein structure, the exact manner by
which these changes occur is not revealed.41 It must be
emphasized that viewing the conformational change observed
for HCMV protease as consisting of a well-defined transition
between discreet structures is a gross oversimplification. It is
clear that the free enzyme in solution exists as an ensemble of
a great many, rapidly interconverting structures, and that the
complexation of a rigid peptidyl ligand results, above all, in a
restriction of the flexibility of the protein. In this respect, the
process bears elements of protein conformer selection.42 How-
ever, we believe that this partial immobilization43 of the active
site region can only occur once a ligand is bound. In other
(41) For a review covering the conformational adaptability of enzymes,
see: Citri, N. AdV. Enzymol. 1973, 37, 397-649.
(42) Cao, Y.; Musah, R. A.; Wilcox, S. K.; Goodin, D. B.; McRee, D.
E. Protein Sci. 1998, 7, 72-78.
(43) Davis, J. H.; Agard, D. A. Biochemistry 1998, 37, 7696-7707.

Characterization of the Human CytomegaloVirus Protease
J. Am. Chem. Soc., Vol. 121, No. 13, 1999 2981
Figure 8. Peptidyl ligand-induced conformational change of HCMV
protease showing the alterations in the â5-strand and the oxyanion hole.
The Arg-Glu-Arg double salt bridge is also shown (same color scheme
as for Figure 7; iodine is shown in pink).
words, the peptide guides the reorganization of the protein
around it after initial association.
(b) Inhibition of HCMV Protease by Peptidyl Activated
Carbonyl Compounds. Having analyzed the conformational
properties of both free and enzyme-bound inhibitors bearing
the peptide sequence of MKs 1 and 2, and after examination of
the changes in protein structure upon ligand binding by the
protease, we believe that the inhibition of HCMV by analogous
activated carbonyl compounds proceeds as follows: (a) An
encounter complex forms between the flexible apo-form(s) of
E and I, the latter existing as a relatively fixed and extended
peptide structure; (b) formation of the H-bonds between I and
the â5-strand of E in the active site, notably those involving
the P3 residue, results in the above-described structural reor-
ganization and rigidification of the protease, which brings E to
its active or optimal catalytic form; and (c) after the confor-
mational change, the active site is properly organized for
formation of the hemiketal adduct, which can strongly interact
with the active site owing to its structural similarity to the
tetrahedral intermediate (or, more precisely, the transition state
of its formation or breakdown) of the catalytic reaction.44
(c) HCMV Protease As an Induced-Fit45 Enzyme. While
the X-ray crystallographic comparisons and fluorescence studies
provide very convincing evidence as to the induced-fit character
of HCMV protease, truly unequivocal proof was provided by
observation of the conformational change during substrate
hydrolysis. By fluorescence, the peptidyl ligand-induced struc-
tural change manifests itself as a blue shift of 5-7 nm in the
emission maximum of the enzyme upon specific tryptophan
excitation.12 (See Figure 9A for the spectral change due to the
binding of PFEK 4.) Figure 9B compares the fluorescence
spectra of the enzyme alone and in the presence of the
nonfluorogenic substrate 21 at an initial concentration far above
its KM and at approximately 50% conversion to product (i.e.,
enzyme saturation). A blue-shift similar to that observed in the
presence of peptidic inhibitors can clearly be seen.
Figure 9. Effect of ligand binding and Na2SO4 on the fluorescence
emission spectrum of HCMV protease. All spectra were acquired after
tryptophan excitation at 280 nm. (A) The effect of inhibitor binding:
HCMV protease alone (-) and in the presence of saturating concentra-
tions of PFEK 4 (- - -) and â-lactam 24 (---). The trace for 24 was
scaled to facilitate comparison with that of the free enzyme. The
decrease in fluorescence upon binding of 24 is ascribed to absorption
of excitation energy by the aromatic groups of the inhibitor. (B) The
effect of substrate binding: HCMV protease alone (-) and in the
presence of a 30-fold excess of the nonfluorogenic substrate 21 after
50% conversion (- - -). (C) The effect of sodium sulfate: HCMV
protease alone in buffer (-) and after the addition of Na2SO4 to a
concentration of 0.5 M (---). The emission spectrum of the enzyme
saturated with PFEK
(- - -).
4 in the presence of 0.5 M Na2SO4
correlated to a blue-shift in the emission spectrum of the enzyme.
While similar salt activation has been observed for HCMV
The induced-fit mechanism also accounts for other biophysi-
cal properties of the enzyme. The marked activating effect of
salts on the catalytic efficiency of herpes proteases is well
established46,47 and, in the case of HSV-1 protease, has been
(45) Koshland, D. E., Jr. Proc. Natl. Acad. Sci. U.S.A. 1958, 44, 98-
104. Jencks, W. P. AdV. Enzymol. 1975, 43, 219-410.
(46) Hall, D. L.; Darke, P. L. J. Biol. Chem. 1995, 270, 22697-22700.
(47) Yamanaka, G.; DiIanni, C. L.; O’Boyle, D. R., II; Stevens, J.;
Weinheimer, S. P.; Deckman, I. C.; Matusick-Kumar, L.; Colonno, R. J. J.
Biol. Chem. 1995, 270, 30168-30172.
(44) Mader, M. M.; Bartlett, P. A. Chem. ReV. 1997, 97, 1281-1301.

2982 J. Am. Chem. Soc., Vol. 121, No. 13, 1999
LaPlante et al.
Table 5. Structures of Other Substrates/Inhibitors Referred to in This Work
a Abbreviations of unnatural amino acids: Tbg ) tert-butylglycine, Asn(Me2) ) Nδ,Nδ-dimethylasparagine.
Table 6. Kinetic Parameters for the Hydrolysis of Peptide
Substrate (22) Under Varying Salt Conditionsa
Scheme 2. Mechanism of Induced-Fit Catalysis
Incorporating the Partially Activated Enzyme Species E′
no salt
0.5 M Na2SO4
1.5 M NaCl
kcat, s-1
KM, µM
0.052
132
390
0.19
6.5
30000
0.17
31
5400
k
cat/KM, M-1 s-1
a Values listed represent the averages from two experiments. Both
salt conditions are of the same ionic strength (1.5 M).
protease,47,48 no detailed studies have been reported. Figure 9C
presents the effect of Na2SO4 on the fluorescence emission of
the present enzyme. A blue-shift in the emission maximum upon
the addition of sulfate is evident. Kinetic studies were subse-
quently performed under identical buffer conditions to determine
the effect of Na2SO4 on the kinetic parameters for the hydrolysis
of the optimized substrate 22 (Table 6). The results clearly
indicate that the 75-fold increase in catalytic efficiency between
0 and 0.5 M sodium sulfate is due, in very large part, to a
marked drop (20-fold) in KM.
In other words, the strength of binding of the substrate is reduced
by a factor (K) that reflects the energy cost of converting the
enzyme (conformational change) to its activated form E*. The
presence of sulfate, however, should not be looked upon as
shifting the equilibrium K toward E* but, rather, as introducing
an intermediate species E′ that approaches the activated enzyme
as depicted in Scheme 2. Thus, K is broken into two equilibrium
constants (K ) K1K2) such that part of the free energy cost of
(49) For the derivation of this equation based on the scheme below, see
ref 2 (pp 332-333). KM* (treated as a simple dissociation constant) may
be viewed as the KM for a lock-and-key enzyme that always exists in the
activated form E*. Although the induced-fit reaction proceeds via Eunact:S,
the value KM(obs) also represents the global dissociation constant for E*:S
a Eunact + S proceeding via E*.
These results are clearly consistent with the induced-fit model
for HCMV protease. According to Fersht,49 induced fit mediates
against catalysis by increasing KM for all substrates according
to the equation:
KM(obs) ) KM*/K
where K ) [E*]/[Eunact] in the absence of S, and E* is the
activated enzyme form following the conformational change.50
(48) Burke, P. J.; Berg, D. H.; Luk, T. P.; Sassmannshausen, L. M.;
Wakulchik, M.; Smith, D. P.; Hsiung, H. M.; Becker, G. W.; Gibson, W.;
Villarreal, E. C. J. Virol. 1994, 68, 2937-2946.
(50) K is the equilibrium constant for the hypothetical conformational
equilibrium between unactivated (Eunact) and activated (E*) forms of the
free enzyme in solution. In reality, E* only exists when substrate is bound.

Characterization of the Human CytomegaloVirus Protease
J. Am. Chem. Soc., Vol. 121, No. 13, 1999 2983
Table 7. Truncation of Trifluoromethyl Ketone 3
converting Eunact:S to E*:S, which is normally paid for by the
free energy of ligand binding in the absence of salt, is provided
for by the presence of sulfate such that KM(obs)salt ) KM*/K2.
From the values provided in Table 6, one can calculate51 K1
to be 0.05, which corresponds to 1.8 kcal/mol of free energy
available for substrate binding in the presence of 0.5 M sulfate
which would otherwise have been spent in effecting the change
in enzyme conformation in the absence of the salt. Thus, the
sequential blue-shifts in the fluorescence of HCMV protease
as shown in Figure 9C are due to conversion of Eunact to E′ (or
more precisely a mixture of Eunact and the various species
constituting E′) by sulfate, and then to E*-I upon the binding
of the peptidyl ligand. It is likely that unstructured regions of
the protease which are critical for ligand binding (â5-strand,
L9-loop) acquire a certain preorganization in the presence of
kosmotropic agents such as sulfate which would decrease the
entropic loss when they are rigidified upon ligand complexation.
Table 6 also lists the kinetic parameters in the presence of
sodium chloride, again indicating an increase, albeit smaller,
in the specificity constant and consistent with the observed salt
activation of HSV-1 protease.46,47
The induced-fit nature of HCMV protease is another of the
many characteristics setting it apart from classical serine
proteases which are generally acknowledged to be “lock-and-
key”52 enzymes based on high-resolution X-ray crystallographic
studies. The description of the functioning of HCMV protease
outlined above bears some similarity to the workings of certain
aspartic proteases53 for which P3-S3 interaction alters the active
site such that kcat is increased without changing KM. In the case
of elastase, a large increase in kcat/KM (due, principally, to an
increase in kcat) upon the addition of a P3-residue to shorter
substrates has led to the proposal54 that it is an induced-fit
enzyme for which P3-S3 interaction engages the catalytic triad.
In the present case, the inherently low catalytic activity of
HCMV protease has precluded this type of kinetic investigation.
However, truncation studies of TFMK inhibitors (Table 7)
showed that sequential removal of groups from the C-terminal
end of TFMK 3 results in substantial drops in inhibitory activity.
We attribute these decreases in potency, at least in part, to the
removal of key features of the inhibitors which are required by
the protease to optimally form the covalent adduct.
a The dimethylasparagine analogue of 28 is also expected to have
an IC50 value of well over 300 µM based on the modest increase in
potency observed for the change in other series.
protein scaffold, it is perhaps logical that evolution select a
protease possessing a certain flexibility to accommodate the
binding of such immobilized recognition sequences. As dis-
cussed above, the low turnover rate for the enzyme compared
to digestive serine proteases is consistent with the induced-fit
mechanism that mediates against catalysis. The observed kinetic
properties of the protease may be linked to its function in the
infected cell, being tuned to the relatively slow capsid assembly
process.55 Furthermore, the kinetic activation by salts, which is
again related to induced fit, may be a reflection of the behavior
of the protease when it finds itself in the particular micro-
environment of the nucleus or within the capsid.46,47
Inhibitor Design. (a) Peptidyl Inhibitors. The obvious
question arises as to how one could possibly increase the potency
of the present peptidyl activated carbonyl inhibitors. The
dissociation constant of the complex formed between an ideal
transition-state analogue (KTX) and enzyme target is related to
the rate acceleration observed for the enzymatic reaction with
respect to the uncatalyzed rate:56
Examination of the structure of HCMV protease inhibited
by 18 reveals that the molecule lies on the surface of the protein
and that the side chains which are not solvent exposed (P1 and
P3) are buried upon changes in enzyme structure which are
triggered by ligand binding. There are no highly defined pockets
which one would expect to confer a high degree of substrate
specificity due to steric complementarity. The description of
ligand binding we outline above fits well in the context of the
natural enzyme reaction. The induced-fit recognition of a
particular backbone conformation may be, in part, the mecha-
nism by which HCMV protease achieves a certain selectivity.
Considering the polyvalent function of HCMV protease and the
fact that the target sequence within the assembly protein
precursor is likely held in a rigid manner on a supramolecular
KTX ) kuncat/(kcat/KM)
For dipeptidyl TFMK inhibitors of chymotrypsin, Ki values of
approximately 3 orders of magnitude more than the calculated
KTX value of 0.5 pM were measured, leading to the conclusion
that about 65% of the potential transition-state binding is made
use of.22 In the case of HCMV protease, the KTX value based
on the kinetic parameters for the hydrolysis of substrates bearing
the optimized P4-P1 sequence present in 1 is in the 10-15 to
10-14 M range.57 Thus, inhibitors 3 to 5 remain far from
exploiting all of the binding available for transition-state
stabilization.
(51) KM(obs)salt/KM(obs)no salt ) (KM*/K2)/(KM*/K1 K2) ) K1.
(52) Fischer, E. Chem. Ber. 1894, 27, 2985-2993.
(53) Sali, A.; Veerapandian, B.; Cooper, J. B.; Moss, D. S.; Hofmann,
T.; Blundell, T. L. Proteins: Struct. Funct. Genet. 1992, 12, 158-170.
(54) Stein, R. L.; Strimpler, A. M.; Edwards, P. D.; Lewis, J. J.; Mauger,
R. C.; Schwartz, J. A.; Stein, M. M.; Trainer, D. A.; Wildonger, R. A.;
Zottola, M. A. Biochemistry 1997, 26, 22682-2689. Stein, R. L.; Strimpler,
A. M.; Hori, H.; Powers, J. C. Biochemistry 1987, 26, 1301-1305. Stein,
R. L.; Strimpler, A. M.; Hori, H.; Powers, J. C. Biochemistry 1987, 26,
1305-1314.
(55) Babe´, L. M.; Craik, C. S. Cell 1997, 91, 427-430.
(56) Wolfenden, R. Acc. Chem. Res. 1972, 5, 10-18. Wolfenden, R.
Annu. ReV. Biophys. 1976, 5, 271-306.
(57) Based on a kuncat for amide hydrolysis of 10-11 s-1 (Yamana, T.;
Mizukami, Y.; Tsuji, A.; Yasuda, Y.; Masuda, K. Chem. Pharm Bull. 1972,
20, 881-891).

2984 J. Am. Chem. Soc., Vol. 121, No. 13, 1999
LaPlante et al.
The factors contributing to the inhibitory activity of TFMK
3 and related analogues can be, therefore, dissected as follows:
(i) The rigidified elongated peptidyl chain bearing amide groups
properly disposed to H-bond with the â5-strand. The NMR
studies described above clearly demonstrate that inhibitors such
as 2 and 4 possess a fair degree of structure in solution that
corresponds to the conformation when bound to HCMV
protease, effectively reducing the loss of conformational entropy
upon complexation. It appears that the presence of the bulky
side chains stabilizes the bioactive backbone conformation, the
P3 tert-butyl substituent playing a key role. The measured φ
angles for the peptidyl chain of protease-bound 18 in the X-ray
structure were used to calculate58 the theoretical magnitudes of
3JNH-RH for the bioactive conformation. The values for P1 to
P3 were determined to be 9.5, 9.2, and 8.2 Hz, respectively.
Comparison of these values to those in Table 3 is consistent
with our analysis that, for MK 2, P3 is locked in the optimal
conformation and that other nonextended conformers contribute
to the observed J values for P2 and P1.
It is clear that the binding of the peptidyl portion of the present
ketone inhibitors is not very strong as evidenced by the poor
inhibition by MKs 1 and 2 which are not transition-state
analogues. Moreover, we have observed that the 1D proton
NMR resonances of covalently enzyme-bound TFMK inhibitors
remain surprisingly sharp (data not shown). This would indicate
that, even as hemiketal adducts, the present peptidyl inhibitors
retain a high degree of mobility with respect to the protein and
that the E:I interactions, notably those involving the inhibitor
side chains, are probably quite weak.
inducing the structural reorganization as observed by fluores-
cence emission (Figure 9A). The blue-shift normally observed
upon the binding of peptidyl inhibitors or substrate (as reported
above) was not produced upon saturation of the enzyme with
â-lactam inhibitors such as compound 24, the spectrum remain-
ing essentially identical with that of the free enzyme.
Detailed mechanistic studies have shown these molecules to
be very slowly cleaved competitive substrates of HCMV
protease for which an acylated form of a nonactivated enzyme
may accumulate.63 Comparison of the kinetic parameters for
the hydrolysis of the substrate 19 and those of the â-lactam
probe 25 is telling. In the case of peptide cleavage, respective
kcat and KM values of 0.2 s-1 and 760 µM were measured,64
figures which agree with those reported in the literature for
comparable substrates.5 In the case of â-lactam hydrolysis, the
respective kcat and KM values were found to be 0.00043 s-1 and
0.6 µM.63 Such differences, namely increased binding coupled
to lowered catalytic efficiency, are exactly what one would
anticipate for an inhibitor that does not elicit the protease-
activating conformational change. The degree to which the form
of the enzyme acting on a â-lactam substrate is far from one
optimized for catalysis is further underlined in considering the
slow hydrolysis of the acyl-enzyme (kdeacylation being the rate-
determining step at 0.00043 s-1), being 500-fold slower than
the kcat for peptide hydrolysis. Such a large discrepancy is what
one would anticipate for the induced-fit mechanism. Therefore,
the inhibitory activity of â-lactams such as 25 is owed, in part,
to the fact that their sub-optimal processing by the protease
occurs at such a slow rate.
(ii) The electrophilic carbonyl which allows for formation of
a stabilized hemiketal that mimics the transition state. Formation
of the covalent adduct that, owing to its structural analogy with
the catalytic transition-state, may interact strongly with the
enzyme is the key factor contributing to inhibitor potency. This
is clearly demonstrated by the >1000-fold increase in potency
in proceeding from MK 1 to TFMK 3. One might, therefore,
expect the strongest effects on activity being observed upon
changes to the activated carbonyl moiety.
Conclusion
Through the use of a number of NMR methods, we have
demonstrated that inhibitors such as 1-5 of HCMV protease
exist as fairly rigid, extended structures in solution and that the
bulky P3 tert-butyl side chain is strongly implicated in
maintaining this conformation. Moreover, the structure of the
free ligand, as exemplified by MK 2, is very similar to that
when complexed to the enzyme as determined by TRNOE and
X-ray crystallographic methods. This fact undoubtedly contrib-
utes to a significant degree to the observed inhibitory activity.
Comparison of the X-ray structures of the free and inhibited
protease revealed the changes in protein structure which occur
upon ligand binding. Combined, the information allows us to
propose a mechanism of induced-fit inhibition (and catalysis)
for this novel serine protease.
From an enzymological viewpoint, the characterization of
HCMV protease as an induced-fit enzyme represents a signifi-
cant observation. This is especially so by virtue of the fact that
it is a serine protease for which many members of the family
are among the best characterized enzymes to date. Since classical
serine proteases such as chymotrypsin operate through a lock-
Thus, the inhibition of HCMV protease by TFMK 3 and
related activated carbonyl compounds involves the synergistic
effect of a peptidyl portion required to induce the transition to
a catalytically active or activated form of the enzyme, and an
electrophilic keto moiety that may then react to yield a stabilized
covalent complex mimicking the transition-state. This accounts
for the relatively weak binding of the peptidyl chain and is
consistent with the view that, in induced-fit catalysis, binding
energy is expended to compensate for the energy required to
convert the enzyme to a thermodynamically less favorable state.
In view of the magnitude of the observed alterations in protein
structure, notably the apparent ordering of otherwise random
regions, one would anticipate a very large and unfavorable
entropic factor associated with E*:I complex formation.
(b) â-Lactam Inhibitors. It has been suggested that mol-
ecules which complex to, but do not induce the conformational
change associated with an induced-fit enzyme would be better
suited as inhibitors since free energy of binding is not utilized
to effect the structural change in the macromolecule.53 Such
molecules would be expected to exhibit binding constants
reflecting, more directly, the structural interactions between
ligand and protein. In fact, medicinal chemistry efforts later
focused on a class of inhibitors, the â-lactams,59-62 which
inhibited HCMV protease by binding to the active site without
(59) Yoakim, C.; Ogilvie, W. W.; Cameron, D. R.; Chabot, C.; Guse,
I.; Hache´, B.; Naud, J.; O’Meara, J. A.; Plante, R.; De´ziel, R. J. Med. Chem.
1998, 41, 2882-2891.
(60) De´ziel, R.; Malenfant, E. Bioorg. Med. Chem. Lett. 1998, 8, 1437-
1442.
(61) Yoakim, C.; Ogilvie, W. W.; Cameron, D. R.; Chabot, C.; Grand-
Maˆıtre, C.; Guse, I.; Hache´, B.; Kawai, S. H.; Naud, J.; O’Meara, J. A.;
Plante, R.; De´ziel, R. AntiViral Chem. Chemother. 1998, 9, 379-387.
(62) Borthwick, A. D.; Weingarten, G.; Haley, T. M.; Tomaszewski, M.;
Wand, W.; Hu, Z.; Be´dard, J.; Jin, H.; Yuen, L.; Mansour, T. S. Bioorg.
Med. Chem. Lett. 1998, 8, 365-370.
(63) Bonneau, P. R.; Hasani, F.; Plouffe, C.; Malenfant, E.; LaPlante,
S. R.; Guse, I.; Ogilvie, W. W.; Plante, R.; Davidson, W. C.; Hopkins, J.
L.; Morelock, M. G.; Cordingley, M. G.; De´ziel, R. 1999, 121, 2965-2973.
(64) Bonneau, P. R.; Plouffe, C.; Pelletier, A.; Wernic, D.; Poupart, M.-
A. Anal. Biochem. 1998, 255, 59-65.
(58) Pardi, A.; Billeter, M.; Wu¨thrich, K. J. Mol. Biol. 1984, 180, 741-
751.

Characterization of the Human CytomegaloVirus Protease
J. Am. Chem. Soc., Vol. 121, No. 13, 1999 2985
(3S)-3-[(tert-Butoxycarbonyl)amino]butan-(2R,S)-2-ol Benzyl Ether
(11). Carbonyl diimidazole (23.0 g, 0.142 mol) was added to a stirred
solution of N-Boc-alanine (20.0 g, 0.106 mol) in anhydrous dichlo-
romethane (350 mL) cooled to 0 °C and the reaction was stirred under
a nitrogen atmosphere. After 0.5 h, triethylamine (19.8 mL, 0.142 mol)
and N,O-dimethylhydroxylamine hydrochloride (13.85 g, 0.142 mol)
were added and the stirring at 0 °C was continued for 0.5 h after which
time the reaction was allowed to warm to ambient temperature. After
14 h, the solution was poured into ethyl ether (800 mL) and the organic
phase was washed with hydrochloric acid (1 N, 3 × 250 mL), saturated
sodium bicarbonate (250 mL), and brine (250 mL). The ether phase
was dried (MgSO4), filtered, and evaporated in vacuo to yield Weinreb
amide 8 (23.1 g, 94%) as a white solid that was used without further
purification. 1H NMR (400 MHz, CDCl3): δ 1.29 (d, 3H, J ) 7.0 Hz,
Ala-â-Me), 1.42 (s, 9H, t-Bu), 3.19 (s, 3H, NMe), 3.75 (s, 3H, OMe),
4.65 (br m, 1H, Ala-R-H), 5.23 (br d, 1H, J ) 7.3 Hz, NH). MS: m/z
(rel intensity) 233 (MH+, 92), 177 ([MH+ - C4H8], 100), 133 ([MH+
- Boc], 86), 91 (26), 57 (70).
Methyllithium (1 M in ethyl ether, 60.0 mL, 60.0 mmol) was added
slowly to a stirred solution of Weinreb amide 8 (5.07 g, 21.8 mmol) in
anhydrous tetrahydrofuran (230 mL) at -78 °C under a nitrogen
atmosphere and the resulting solution was stirred for 1.5 h. The reaction
was then quenched with saturated ammonium chloride (100 mL) and
extracted with ethyl acetate (100 + 400 mL). The combined organic
phases were washed with water (500 mL) and brine (500 mL) and then
dried (MgSO4), filtered, and evaporated in vacuo to yield methyl ketone
9 as a pale yellow oil (4.45 g, quantitative) that partially solidified
upon standing and was used without further purification. 1H NMR (400
MHz, CDCl3): δ 1.34 (d, 3H, J ) 7.3 Hz, Ala-â-Me), 1.44 (s, 9H,
t-Bu), 2.20 (s, 3H, C(O)Me), 4.30 (m, 1H, Ala-R-H), 5.24 (br s, 1H,
NH). MS: m/z (rel intensity) 188 (MH+, 37%), 132 ([MH+ - C4H8],
100), 91 (41), 88 ([MH+ - Boc], 61), 57 (66).
and-key mechanism, comparison with HCMV protease, which
effects peptide cleavage through the same chemical steps, should
further our understanding of the differences between lock-and-
key and induced-fit models of catalysis. Detailed kinetic studies
characterizing the present enzyme are clearly wanting and will
no doubt aid in the development of new inhibitors.
From the perspective of drug design, our medicinal efforts
constitute an interesting example of how two different classes
of competitive inhibitors of comparable potency exercise their
activity in two very different ways. In the case of the activated
peptidyl ketones, inhibition relies on optimal functioning of the
catalytic machinery. They are mechanism-based inhibitors which
depend on complexation of their peptidic chains to bring the
enzyme to an activated state, even though this interaction does
not contribute to a great degree to the overall binding.
In the case of the â-lactam inhibitors, the intrinsic binding
energy appears to be a much more important factor in endowing
potency. Having made the comparison between the KM values
for peptide and â-lactam hydrolysis above, it is perhaps more
important to consider the differences in the dissociation constant,
KS (i.e., for E:I a E + I). In the case of peptidyl inhibitor MK
1, KS ) KI ) 2.1 mM, whereas for the â-lactam 25, the KS was
determined to be 64 µM.63 Thus, despite the much higher degree
of noncovalent interaction between E and I for peptidyl
inhibitors, the much smaller â-lactam binds greater than 30-
times more tightly to the protease.
For these latter molecules, nonoptimized functioning of the
catalytic machinery (i.e., the absence of induced-fit activation)
is a key factor since it prevents rapid processing of the inhibitors
which are also substrates for the enzyme. Interestingly, there is
some evidence in the case of â-lactams lacking a heteroatom
substituent at the C-4 position that the acylation rate may
approach that for the deacylation of the protease. If this is so,
the noncovalent E:I complex also accumulates and must be
considered as well to be an inhibitory species. Therefore, the
most advantageous strategy in inhibitor design may, perhaps,
be a molecule that binds in the same mode as a â-lactam (i.e.,
no activating conformational change) but which is not at all
prone to processing by the suboptimal form of the enzyme.
Sodium borohydride (907 mg, 24.0 mmol) was added to a stirred
solution of ketone 9 (4.09 g, 21.8 mmol) in tetrahydrofuran (42 mL)
containing methanol (10 mL) at 0 °C and the reaction was stirred for
2 h. The solution was then diluted with ethyl ether (170 mL) and
quenched by the slow addition of aqueous citric acid (10% w/v, 85
mL). The aqueous phase was extracted with ether (3 × 85 mL) and
the combined organic phases were washed with saturated sodium
bicarbonate (170 mL) and brine (170 mL), dried (MgSO4), filtered,
and evaporated in vacuo to yield the alcohols 10 (4.15 g, quantitative,
mixture of diastereomers) as a white solid that was used without further
1
purification. H NMR (400 MHz, CDCl3): major isomer, δ 1.09 (d,
3H, J ) 7.0 Hz, Ala-â-Me), 1.14 (d, 3H, J ) 6.4 Hz, CH(OH)Me),
1.45 (s, 9H. t-Bu), 3.60 (br m, 1H, Ala-R-H), 3.85 (dq4, J ) 3.2, 6.4
Hz, CH(OH)), 4.62* (br s, 1H, NH); minor isomer, δ 1.16 (d, 3H, J )
7.0 Hz, Ala-â-Me), 1.19 (d, 3H, J ) 6.4 Hz, CH(OH)Me), 1.45 (s, 9H.
t-Bu), 3.56 (br m, 1H, Ala-R-H), 3.69 (dq4, J ) 4.7, 6.4 Hz, CH(OH)),
4.62* (br s, 1H, NH). MS: m/z (rel intensity) 280 (MH+, 52%), 134
([MH+ - C4H8], 100), 90 ([MH+ - Boc], 87), 57 (48).
Experimental Section
Synthesis: General Methods. NMR spectra were recorded on a
Bruker AMX400 spectrometer and are referenced to internal tetram-
ethylsilane. Peak multiplicities are denoted by the following abbrevia-
tions: app, apparent; br, broad; s, singlet; d, doublet; t, triplet; q4,
quartet; q5, quintet; and m, multiplet (asterisk indicates overlapping
signals). A reference value of 77.00 ppm was used for the 13C solvent
signal of chloroform. The following abbreviations are used: TBG (tert-
butylglycine), TBA (tert-butylacetyl), and IPA (isopropylacetyl). IR
spectra were recorded on a Mattson Research Series spectrophotometer.
FAB mass spectra (thioglycerol) were recorded on either Autospec VG
or Kratos MS50 instruments (at the Department of Chemistry, Uni-
versite´ de Montre´al). Optical rotation measurements were carried out
in a Perkin-Elmer 241 polarimeter. UV spectra were obtained with a
Perkin-Elmer Lambda 7 UV/VIS instrument. Analytical HPLC em-
ployed either of the following systems: (A) Vydac C18 10 mm
analytical column (24 × 4.6 mm), mobile phase: acetonitrile/0.06%
trifluoroacetic acid (TFA) in water/0.06% TFA; (B) Symmetry shield
C8 10 mm analytical column (15 × 3.9 mm), mobile phase: acetonitrile
in 20 mM Na2HPO4, pH 9.0. Flash chromatography was performed on
Merck silica gel 60 (0.040-0.063 mm) with nitrogen pressure.
Analytical thin-layer chromatography (TLC) was carried out on
precoated (0.25 mm) Merck silica gel F-254 plates. Anhydrous grade
(Aldrich) dichloromethane and N,N-dimethylformamide were employed
and tetrahydrofuran was distilled from lithium aluminum hydride
immediately prior to use.
The mixture of alcohols 10 (995 mg, 5.26 mmol) was added to a
stirred suspension of sodium hydride (211 mg, 60% oil disp., 5.28
mmol) in dry N,N-dimethylformamide (14 mL) cooled to 0 °C followed
by tetra-n-butylammonium iodide (97 mg, 0.26 mmol). After 0.5 h,
benzyl bromide (0.75 mL, 6.31 mmol) was added and the reaction was
allowed to warm to ambient temperature under a nitrogen atmosphere.
After 1 day, the mixture was diluted with ethyl ether/ethyl acetate (1:1
v/v, 100 mL) and washed with water (50 mL), saturated sodium
bicarbonate (50 mL), and brine (50 mL). The organic phase was then
dried (MgSO4), filtered, and evaporated in vacuo yielding an orange
oil that was purified by flash chromatography (SiO2, hexane/ethyl
acetate ) 6/1 to 4/1) to yield the benzyl ethers 11 (644 mg, 44%,
mixture of diastereomers) as a colorless oil. 1H NMR (400 MHz,
CDCl3): major isomer, δ 1.12 (d, 3H, J ) 7.0 Hz, Ala-â-Me), 1.16 (d,
3H, J ) 6.4 Hz, CH(OR)Me), 1.43 (s, 9H, t-Bu), 3.56-3.64* (br m,
1H, CH(OR)), 3.66-3.77 (br m, 1H, Ala-R-H), 4.44 and 4.62 (ABq4,
2H, J ) 11.8 Hz, CH2Ph), 4.62* (br s, 1H, NH), 7.23-7.37* (m, 5H,
Ph); minor isomer, δ 1.17 (d, 3H, J ) 6.7 Hz, Ala-â-Me), 1.18 (d, 3H,
J ) 6.4 Hz, CH(OH)Me), 1.44 (s, 9H, t-Bu), 3.50 (dq4, 1H, J ) 2.7,
6.4 Hz, CH(OR)), 3.56-3.64* (br m, 1H, Ala-R-H), 4.43 and 4.62
(ABq4, 2H, J ) 11.8 Hz, CH2Ph), 4.62* (br s, 1H, NH), 7.23-7.37*

2986 J. Am. Chem. Soc., Vol. 121, No. 13, 1999
LaPlante et al.
(m, 5H, Ph). MS: m/z (rel intensity) 280 (MH+, 37%), 224 ([MH+
-
Enzyme Assay and Kinetics. All studies employed the HCMV
protease mutant Ala143Gln to overcome the problem of autoproteoly-
sis.65 Full details of the preparation and purification of the recombinant
enzyme (>99% purity) have been published.12 IC50 values were
determined by using a fluorogenic assay that has also been described.13
The kinetic parameters kcat and KM for the hydrolysis of 22 under
varying salt conditions were determined by initial rate kinetics
(according to ref 64) at 30 °C.
Fluorescence Studies. Fluorescence measurements were performed
in 1 mL quartz cuvettes with a Perkin-Elmer LS-50B spectrofluorom-
eter. Emission spectra of HCMV protease were recorded upon excitation
at 280 nm. Further experimental details have been described.12
NMR Studies: (a) Free Inhibitors in H2O/DMSO. Spectra of
samples containing 30 mM MKs 1 and 2 and PFEKs 4, 6, and 7 in
15% H2O in DMSO-d6 were acquired on a Bruker DMX 600
instrument. Further experimental details are provided in the Supporting
Information.
C4H8], 82), 180 ([MH+ - Boc], 46), 91 ([C7H7+], 100), 57 (27). HRMS
(C16H26O3N): calcd 280.1913; found 280.1922.
(3S)-3-[N-tert-Butylacetyl-L-(r-tert-butylglycyl)-L-(Nδ,Nδ-dimeth-
ylasparagyl)amino]butan-2-one (1). Alcohols 15 (200 mg, 0.452
mmol) were dissolved in anhydrous dichloromethane (10 mL) contain-
ing N,N-dimethylformamide (0.5 mL) and Dess-Martin periodinane
(383 mg, 0.904 mmol) was added. After the mixture was stirred at
ambient temperature for 2.5 h under a nitrogen atmosphere, the reaction
was diluted with dichloromethane (10 mL) and a mixture of 10%
sodium thiosulfate/saturated sodium bicarbonate (1:1 v/v, 50 mL) was
added and the phases vigorously stirred. After 0.5 h, the clear phases
were extracted with ethyl acetate (50 + 30 mL) and washed with
saturated sodium bicarbonate (4 × 50 mL) and brine (50 mL). The
combined organic phases were then dried (Na2SO4), filtered, and
evaporated in vacuo to a colorless foam. The product was dissolved in
boiling ethyl acetate and the solution was allowed to cool, resulting in
the precipitation of a white solid. The material was filtered, washed
with cold solvent, and dried under vacuum, affording methyl ketone 1
(b) ROESY Experiments. Homonuclear ROESY data from the
States-TPPI method66 were acquired at 600 MHz, cw spinlock, 300 t1
data points from the addition of 32 transients, and a relaxation delay
of 1.2 s. Three spectra were collected with spinlock times of 300 ms,
800 ms, and 1.5 s. ROESY data were also collected for 2 mM MKs 1
and 2 with a spinlock time of 300 ms.
1
(112 mg, 56%) as a white powder. H NMR (400 MHz, CDCl3): δ
1.04 (s, 18H, 2 × t-Bu), 1.34 (d, 3H, J ) 7.3 Hz, Ala-â-Me), 2.11 and
2.14 (ABq4, 2H, J ) 13.0 Hz, TBA-R-H) 2.17 (s, 3H, C(O)Me), 2.50
(dd, 1H, J ) 7.0, 16.5 Hz, Asn(Me2)-â-HA), 2.94 and 3.00 (two s, 2 ×
3H, NMe2), 3.13 (dd, 1H, J ) 3.2, 16.5 Hz, Asn(Me2)-â-HB) 4.20 (d,
1H, J ) 8.6 Hz, TBG-R-H), 4.46 (app q5, 1H, Japp ) 7.3 Hz, Ala-R-
H), 4.84 (app dt, 1H, Japp ) 3.2, 7.3 Hz, Asn(Me2)-R-H), 6.03 (d, 1H,
J ) 8.3 Hz, TBG-NH), 7.43 (d, 1H, J ) 7.6 Hz, Asn(Me2)-NH), 7.68
(d, 1H, J ) 6.7 Hz, Ala-NH). 13C NMR (100.6 MHz, CDCl3): δ 16.72,
26.29, 26.85, 29.86, 31.00, 34.30, 34.77, 35.53, 37.34, 49.38, 50.76,
54.99, 61.36, 170.45, 170.54, 171.04, 171.89, 206.59. MS: m/z (rel
intensity) 441 (MH+, 100%), 354 ([MH+ - AlaMe], 11), 230 ([Asn-
(c) Carbon-13 NMR. One-dimensional 13C spectra for T1 relaxation
measurements for PFEKs 4, 6, and 7 were acquired at 150 MHz at 27
°C. Inversion recovery experiments were run with power-gated proton
decoupling during acquisition. Ten spectra were acquired corresponding
to the τ delays (0.01, 0.1, 0.25, 0.37, 0.5, 1.0, 1.5, 2.0, 3.0, and 7.0 s).
Each spectrum was acquired by adding 1600 transients and using a
relaxation delay of 6.9 s. The method used to calculate the T1 values
for the P1 and P4 R-centers of 6 and 7 is provided in the Supporting
Information.
(d) Transferred NOE Experiments. Transferred NOESY experi-
ments involving MK 2 were carried out on a Varian UNITY 750
instrument with mixing times of 50, 100, 150, and 250 ms. Sample
preparation and NMR conditions were identical with those described
in ref 17, as were the computational methods used to model MK 2
with Discover 95.0 and the CFF95 force field.
(Me2)AlaMe + H+ ], 21), 212 (19), 184 (11), 115 (74). IR: 1713 cm-1
.
[R]22D +47.3° (c 0.516, CHCl3). UV: λmax ) 282 nm (ꢀ ) 46, MeOH).
HRMS (C22H41O5N4): calcd 441.3077; found 441.3112. HPLC: system
A >99%, system B >99%.
(3S)-3-[N-Isopropylacetyl-L-(r-tert-butylglycyl)-L-(Nδ,Nδ-di-
methylasparagyl)amino]butan-2-one (2). Alcohols 17 (161 mg) were
oxidized as described above for the preparation of 1 above. The crude
product was purified by flash chromatography (SiO2, ethyl acetate/
ethanol ) 10/1), which afforded methyl ketone 2 (112 mg, 70%) as a
Acknowledgment. We are grateful to Dr. Feng Ni and his
research group (Biomolecular NMR Laboratory, Biotechnology
Research Institute, NRC) for helpful discussions and technical
support. We also thank Dr. L. Lagace´ and M.-J. Massariol for
providing the enzyme. A. Abraham, C. Chabot, J.-S. Duceppe,
G. Fazal, S. Goulet, T. Halmos, E. Malenfant, J. A. O’Meara,
Dr. M.-A. Poupart, and D. Wernic are gratefully acknowledged
for the synthesis of compounds used in the present work. We
gratefully acknowledge Dr. W. W. Ogilvie for his important
intellectual contribution. Finally, we thank Drs. P. C. Anderson
and M. G. Cordingley for their support.
white powder. 1H NMR (400 MHz, CDCl3): δ 0.96 (app d, 6H, Japp
)
6.4 Hz, CHMe2), 1.04 (s, 9H, t-Bu), 1.34 (d, 3H, J ) 7.3 Hz, Ala-â-
Me), 2.08-2.16 (m, 3H, IPA-R,â-H), 2.17 (s, 3H, C(O)Me), 2.50 (dd,
1H, J ) 7.0, 16.5 Hz, Asn(Me2)-â-HA), 2.94 and 3.00 (two s, 2 × 3H,
NMe2), 3.13 (dd, 1H, J ) 3.2, 16.5 Hz, Asn(Me2)-â-HB), 4.23 (d, 1H,
J ) 8.6 Hz, TBG-R-H), 4.46 (app q5, 1H, Japp ) 7.3 Hz, Ala-R-H),
4.85 (app dt, 1H, Japp ) 3.2, 7.3 Hz, Asn(Me2)-R-H), 6.10 (d, 1H, J )
8.6 Hz, TBG-NH), 7.47 (d, 1H, J ) 8.0 Hz, Asn(Me2)-NH), 7.65 (d,
1H, J ) 7.0 Hz, Ala-NH). 13C NMR (100.6 MHz, CDCl3): 16.73,
22.46, 22.51, 26.16, 26.28, 26.82, 34.43, 34.69, 35.53, 37.35, 46.16,
49.41, 54.97, 61.21, 170.44, 170.48, 171.06, 172.62, 206.55. MS: m/z
(rel intensity) 427 (MH+, 100%), 340 ([MH+ - AlaMe], 10), 230 ([Asn-
(Me2)AlaMe + H+], 21), 198 (11),170 (18), 115 (67). IR: 1719 cm-1
.
Supporting Information Available: Detailed procedures
and characterization for compounds 12 to 17, as well as
characterization (NMR, MS, HPLC) of inhibitors 6 and 7; details
of NMR experiments; supporting NMR data (NOESY competi-
tion studies for MK 1 and quantitative comparison of cross-
peak volumes for MK 2 (NOESY and ROESY) and 20
(NOESY)) (PDF). This material is available free of charge via
the Internet at http://pubs.acs.org.
[R]22D +49.5° (c 0.513, CHCl3). UV: λmax ) 282 nm (ꢀ ) 48, MeOH).
HRMS (C21H39O5N4) calcd 427.2920; found 427.2909. HPLC: system
A >99%, system B >99%.
Other Inhibitors, Substrates, and Intermediates. The syntheses
of activated carbonyl inhibitors 3-5 and 26-28 have been described.13
Inhibitors 6 and 7 were prepared by the same published method as
described for 4; their physical constants are provided in the Supporting
Information as are the full experimental procedures and characterization
for intermediates 12-17. The peptides 20 and 21 were prepared by
standard solid-phase methods (respective purities of >96% and >97%
by HPLC). The synthesis of â-lactam inhibitors 23-25 and the
substrates 18, 19, and 22 have been described (see Table 5 for the
respective references).
JA983904H
(65) Pinko, C.; Margosiak, S. A.; Vanderpool, D.; Gutowski, J. C.;
Condon, B.; Kan, C.-C. J. Biol. Chem. 1995, 270, 23634-23640.
(66) Bax, A.; Davis, D. G. J. Magn. Reson. 1985, 63, 207-213.