Conformational selection of inhibitors and substrates by proteolytic enzymes: Implications for drug design and polypeptide processing

Article abstract of DOI:10.1021/jm990315t

Inhibitors of proteolytic enzymes (proteases) are emerging as prospective treatments for diseases such as AIDS and vital infections, cancers, inflammatory disorders, and Alzheimer's disease. Generic approaches to the design of protease inhibitors are limi

Full text of DOI:10.1021/jm990315t

J . Med. Chem. 2000, 43, 1271-1281
1271
Con for m a tion a l Selection of In h ibitor s a n d Su bstr a tes by P r oteolytic En zym es:
Im p lica tion s for Dr u g Design a n d P olyp ep tid e P r ocessin g
,
†
†
†
†
†
David P. Fairlie,* J oel D. A. Tyndall, Robert C. Reid, Allan K. Wong, Giovanni Abbenante,
†
†
†
‡
‡
Martin J . Scanlon, Darren R. March, Douglas A. Bergman, Christina L. L. Chai, and Brendan A. Burkett
Centre for Drug Design and Development, University of Queensland, Brisbane, Queensland 4072, Australia, and
Research School of Chemistry, The Australian National University, Canberra, Australian Capital Territory 2600, Australia
Received J une 21, 1999
Inhibitors of proteolytic enzymes (proteases) are emerging as prospective treatments for diseases
such as AIDS and viral infections, cancers, inflammatory disorders, and Alzheimer’s disease.
Generic approaches to the design of protease inhibitors are limited by the unpredictability of
interactions between, and structural changes to, inhibitor and protease during binding. A
computer analysis of superimposed crystal structures for 266 small molecule inhibitors bound
to 48 proteases (16 aspartic, 17 serine, 8 cysteine, and 7 metallo) provides the first conclusive
proof that inhibitors, including substrate analogues, commonly bind in an extended â-strand
conformation at the active sites of all these proteases. Representative superimposed structures
are shown for (a) multiple inhibitors bound to a protease of each class, (b) single inhibitors
each bound to multiple proteases, and (c) conformationally constrained inhibitors bound to
proteases. Thus inhibitor/substrate conformation, rather than sequence/composition alone,
influences protease recognition, and this has profound implications for inhibitor design. This
conclusion is supported by NMR, CD, and binding studies for HIV-1 protease inhibitors/
substrates which, when preorganized in an extended conformation, have significantly higher
protease affinity. Recognition is dependent upon conformational equilibria since helical and
turn peptide conformations are not processed by proteases. Conformational selection explains
the resistance of folded/structured regions of proteins to proteolytic degradation, the susceptibil-
ity of denatured proteins to processing, and the higher affinity of conformationally constrained
‘
extended’ inhibitors/substrates for proteases. Other approaches to extended inhibitor conforma-
tions should similarly lead to high-affinity binding to a protease.
In tr od u ction
the lack of structural information until recently for
protease-inhibitor complexes. Even now there is no
structural information available for most known pro-
teases, let alone their inhibitor-bound complexes. Most
approaches to protease inhibitors have begun with some
knowledge of which peptide substrates are processed by
the target protease. Proteases tend to recognize short
Proteolytic enzymes of the four major classes (aspar-
tic, cysteine, metallo, serine) catalyze both intracellular
and extracellular peptide cleavages during numerous
physiological processes, including digestion, fertilization,
growth, differentiation, cell signaling and migration,
immunological defense, wound healing, apoptosis, pro-
tein turnover, and general ‘housekeeping’ functions.
Proteases are also vital in the propagation of most
(<10 residue) peptide sequences of their polypeptide
substrates, the amino acid side chains being crucial for
determining selectivity and susceptibility to cleavage.1
However when isolated, these short peptides are most
commonly found in rapidly interconverting random
conformations, whereas native polypeptide substrates
can fold via different intramolecular hydrogen bonds
into well-defined fixed conformations such as R-helices,
â- or γ- turns, and â-sheets. Very little is presently
known about whether substrate conformation influences
protease recognition or cleavage. Also, the most abun-
dant substrate/inhibitor conformation in solution may
not necessarily be the one recognized by a protease, and
conformational equilibria may significantly influence
rates of peptide processing.
1
disease processes. Thus selective inhibition of proteases
is increasingly being targeted for the treatment of
2
cancers; parasitic, fungal, and viral infections (e.g.
3
4
5
6
schistosomiasis, malaria, C. albicans, HIV, hepatitis
7
8
9
C, herpes, common cold ); inflammatory, immunologi-
cal, and respiratory conditions; and cardiovascular
and degenerative disorders including Alzheimer’s dis-
ease. There are now many designed potent and selec-
tive protease inhibitors that slow or halt disease pro-
gression, HIV-1 protease inhibitors being notable for
1
0
11
1
2
1
3
6
their speedy emergence from clinical trials.
Progress toward the design of small molecule inhibi-
tors of proteases is significantly handicapped by the
unpredictable conformational changes that can occur
upon substrate/inhibitor binding to proteases, and by
We now pool together some compelling structural
evidence that aspartic, serine, metallo, and cysteine
proteases all share a common conformational require-
ment for recognition, namely an extended (â-strand)
conformation for their active-site-directed inhibitors and
substrate analogues. It has previously been noted for
individual crystal structures of protease-inhibitor com-
*
To whom correspondence should be addressed. Tel: +61 7 336
5
1271. Fax: +61 7 336 51990. E-mail: d.fairlie@mailbox.uq.edu.au.
†
University of Queensland.
The Australian National University.
‡
1
0.1021/jm990315t CCC: $19.00 © 2000 American Chemical Society
Published on Web 03/11/2000

1
272 J ournal of Medicinal Chemistry, 2000, Vol. 43, No. 7
Fairlie et al.
F igu r e 1. Superimposed protease-inhibitor X-ray crystal structures (PDB codes), proteases omitted: (a) endothiapepsin-binding
conformations of 21 inhibitors (1eed, 1ent, 1epl, 1epm, 1epn, 1epo, 1epp, 1epq, 1epr, 1er8, 2er0, 2er6, 2er7, 2er9, 3er3, 3er5, 4er1,
4
1
2
8
er2, 4er4, 5er1, 5er2); (b) porcine pancreatic elastase-binding conformations of 16 inhibitors (1eal, 1eas, 1eat, 1eau, 1eld, 1ele,
elf, 1esb, 1bma, 1fle, 1inc, 1jim, 1nes, 2est, 5est, 7est); (c) thermolysin-binding conformations of 13 inhibitors (1thl, 1tlp, 1tmn,
tmn, 3tmn, 4tln, 4tmn, 5tln, 5tmn, 6tmn, 7tln, 7tmn, 8tln), metal shown as white sphere; (d) papain-binding conformations of
inhibitors (1pad, 1pe6, 1pip, 1pop, 1ppp, 4pad, 5pad, 6pad), cysteine in green with sulfur shown as yellow sphere. Rmsd values
were 0.30 ( 0.19 Å (a), 0.30 ( 0.06 Å (b), 0.17 ( 0.04 Å (c), and 0.37 ( 0.30 Å (d).
plexes (e.g. separate reports for each PDB-filed structure
in Table 1, Supporting Information) that the extended
including human immunodeficiency virus protease,
HIV-1PR (50 inhibitors) and related viral proteases
(HIV-2 (13 inhibitors), SIV (5), FIV (1)), cathepsin D (1),
renin (3), rennin/chymosin (1), penicillopepsin (7), se-
creted aspartic protease (2), pepsin (2), mucoropepsin
(1), retropepsin (1), saccharopepsin (1), rhizopuspepsin
(4), and plasmepsin II (1). All 114 structures (Table 1)
show a common extended (â-strand) conformation for
inhibitors bound to aspartic proteases. Since many of
these inhibitors are substrate analogues, containing
identical peptide sequences to substrates but with the
cleavable amide bond replaced by uncleavable transi-
tion-state analogues, substrates are also likely to be
recognized in their strand conformations.
1
4
strand conformation is important for recognition, but
recent publication of large numbers of three-dimensional
structures for protease-inhibitor complexes now per-
mits this more extensive analysis of inhibitor confor-
mational preferences for all four classes of proteases.
A strategy that fixes the extended conformation of an
inhibitor is also examined below, for several proteases
from different classes, and found to enhance inhibitor
affinity for a protease. For one protease (HIV-1 pro-
tease), additional structural (CD, NMR), binding, and
limited mechanistic observations for substrates and
substrate analogues support the hypothesis that they
too are recognized and processed in only an extended
conformation. Proteases appear most likely to select for,
rather than structurally reorganize inhibitors/sub-
strates to, a single (extended) conformation. This con-
formational selection phenomenon appears to be com-
mon for proteases and has profound implications for
protein folding and stability, as well as for design of
protease inhibitors as potential drugs.
1
6
Serine proteases are classified by their substrate
specificity as either trypsin-like (positively charged
residues (Lys/Arg) preferred at P1), chymotrypsin-like
(large hydrophobic resiudes (Phe/Tyr/Leu) at P1), or
elastase-like (small hydrophobic residues (Ala, Val) at
P1). All three types were compared. Figure 1b shows
superimposed conformations of 20 inhibitors complexed
to porcine pancreatic elastase. Other PDB-filed crystal
structures examined for inhibitor-protease complexes
of serine proteases (Table 1) include trypsin (10 inhibi-
tors), R-chymotrypsin (1), γ-chymotrypsin (19), human
neutrophil elastase (3), R-lytic protease (9), thrombin
(10), subtilisin (7), proteinase A (4), achromobacter (1),
human cathepsin G (1), glutamic acid-specific protease
(1), carboxypeptidase (3), blood coagulation factor VIIa
(1), porcine factor IXA (1), mesentericopeptidase (1), and
thermitase (1). All 95 structures (Table 1) show a
common extended inhibitor conformation for these 17
serine proteases.
Resu lts
P r otea ses w ith Mu ltip le In h ibitor s. Inhibitors of
the aspartic1 protease, endothiapepsin,
5a
15b
adopt an
extended strand conformation when bound to the pro-
tease as exemplified in Figure 1a for 21 superimposed
PDB-filed crystal structures. All proteases were super-
imposed, but only protease-bound conformations of
inhibitors are displayed. We also examined 15 other
aspartic proteases for which there are PDB coordinates
for cystal structures with bound inhibitors (Table 1)

Conformational Selection by Proteolytic Enzymes
J ournal of Medicinal Chemistry, 2000, Vol. 43, No. 7 1273
These results demonstrate that all four classes of
proteases bind inhibitors in a common extended â-strand
conformation forming enzyme-inhibitor-enzyme anti-
parallel â-sheet arrangements.
In h ibitor s w ith Mu ltip le P r otea ses. Figure 2
shows conformations of single inhibitors each bound to
multiple proteases, an arbitrary structure being chosen
for superimposition of the others. The serine protease-
binding domain of Eglin C, a proteinacious inhibitor
1
9
from a leech, adopts an extended conformation be-
tween P4-P1 when bound to five different serine
proteases (Figure 2a). The classic aspartyl protease
inhibitor, pepstatin (or acetylpepstatin), is a hexapep-
tide inhibitor which adopts the same extended confor-
mation between P3-P3′ when bound to at least eight
different aspartic proteases (Figure 2b, rmsd 0.1-0.5
Å versus PDB: 4er2). The natural product cyclotheona-
mide binds in an extended conformation between P3-
P1 (Figure 2c) to both the digestive enzyme trypsin and
the trypsin-like thrombin involved in clotting (rmsd 0.43
Å for superimposed backbone atoms). These examples
strengthen the notion of conformational homogeneity in
protease recognition.
Cyclic P r otea se In h ibitor s. To test the proposition
that proteases preferentially recognize an extended
conformation, we and others have devised methods to
stabilize the â-strand inhibitor conformation.2
0,21
Our
strategy was to fix inhibitors in protease-binding con-
formations by linking together side chains from i,i+2
amino acids to form macrocycles (e.g. 1-3). We have
F igu r e 2. Superimposed protease-inhibitor X-ray crystal
structures (PDB codes), proteases omitted: (a) binding domain
of Eglin C complexed with the serine proteases R-chymotrypsin
(dark blue, 1acb), subtilisin carlsberg (red, 1cse), mesenteri-
copeptidase (yellow, 1mee), subtilisin novo (green, 1sbn), and
thermitase (light blue, 1tec); (b) binding domain of pepstatin
complexed with the aspartic proteases cathepsin D (pink,
1
lyb), pepsin A (red, 1pso), plasmepsin II (yellow, 1sme),
mucoropepsin (green, 2rmp), endothiapepsin (light blue, 4er2),
rhizopuspepsin (dark blue, 6apr), HIV-1 (acetylpepstatin,
orange, 5hvp), and HIV-2 (acetylpepstatin, white, 1phv); (c)
binding domain of cyclotheonamide A (4) complexed with the
serine proteases thrombin (green, 1tmb) and trypsin (yellow,
1
tyn).
Metalloproteases17 for which inhibitor-protease crys-
tal structures could be accessed (Table 1) were ther-
molysin (13 inhibitors), matrilysin (3), neutrophil col-
lagenase (6), interstitial collagenase (2), atrolysin C (1),
stromelysin (4), and carboxypeptidase A (9). These 38
structures establish that all 7 metalloproteases bind to
their inhibitors in extended strand conformations as
depicted in Figure 1c for thermolysin-bound inhibitors.
Cysteine proteases18 exist in three structurally different
classes typified as papain-like (e.g. cathepsins), picorna-
viral proteases (similar to serine proteases with cysteine
replacing serine), and ICE-like proteases (caspases).
Crystal structures were examined for inhibitors bound
to papain (8 inhibitors; Figure 1d), and for inhibitors
complexed with cathepsin B (2), ICE (2), apopain or
CPP32 (2), actinidin (1), glycyl endopeptidase (1), cruzain
earlier reported the syntheses of 1-3 which are potent
2
2
inhibitors of HIV-1 protease. We now show (Figure 3a)
that the structures of 1 and 2 bound to HIV-1 protease
2
3
(PDB codes: 1mtr, 1cpi) superimpose extremely well
upon the structures of 20 other acyclic inhibitors of
HIV-1 protease, not all of which are peptides (Figure
3a). This structural mimicry is achieved through con-
formational restrictions jointly imposed by cyclization
and constraints in the cycle (two planar amides and
aromatic ring). The hydroxyethylamine isostere at-
tached to the cycle adopts slightly different positions
in the enzyme, due to the influence of the charged
secondary amine of 2 which also hydrogen bonds to one
of the catalytic aspartate residues in the enzyme.2
This also has an effect on the location of the benzyl
substituent at P1, and these changes reflect some
2b,23
(2), and calpain (1). In all 19 structures (Table 1) the
protease-bound inhibitors adopt extended strand con-
formations (e.g. Figure 1d).

1
274 J ournal of Medicinal Chemistry, 2000, Vol. 43, No. 7
Fairlie et al.
flexibility in the capacity of the enzyme to bind coop-
eratively to the inhibitor.
6
Unlike most known inhibitors of HIV-1 protease, 1-3
are restrained in an extended conformation (Figure 3a)
2
0,22,23
that is essentially preorganized for receptor binding.
Due to reduced conformational entropy for protease-
binding, these cyclic inhibitors of HIV-1 protease are
1
0-100 times more potent than acyclic analogues of
2
0c,24
comparable size.
Cyclization also protects amide
bonds from proteolytic degradation, a major problem
with bioavailability of acyclic peptides. This mimetic
strategy can be extended to other proteases since, for
example, we find that protease-bound cyclic natural
products 4 (Figure 2c, PDB: 1tyn) and 5 (Figure 3B,
PDB: 1tps) and the synthetic inhibitor 6 (PDB: 1mmp)
also exist in an extended conformation. These com-
pounds are known to potently inhibit trypsin, thrombin,
and matrilysin, respectively.
F igu r e 3. Superimposed protease-inhibitor X-ray crystal
structures (PDB codes), proteases omitted: (a) HIV-1 protease-
binding conformations of cyclic inhibitors 1 (yellow, 1cpi) and
2
1
4
(blue, 1mtr) and 20 acyclic inhibitors (red, 1aaq, 1gnm, 1gnn,
gno, 1hbv, 1hef, 1heg, 1hih, 1hiv, 1hps, 1hsg, 1hvi, 1hxb, 1sbg,
hvp, 4phv, 5hvp, 7hvp, 8hvp, 9hvp); (b) trypsin-binding
conformation of cyclic inhibitor 5 (yellow, 1tps) and four acyclic
inhibitors (green, 1brb, 1brc, 1ppe, 1smf). Rmsd values were
0
.57 ( 0.15 Å (a) and 0.43 ( 0.07 Å (b).
Cyclic peptides in general are usually associated with
2
0a,b
mimicking turn conformations of peptides.
It ap-
pears though that if the macrocycle is small enough,
some of its components present their peptide backbone
F igu r e 4. Ramachandran plot of 30 angle pairs (Φ,Ψ° or
pseudo Φ,Ψ°) for 10 cyclic inhibitors of HIV-1 protease (3),
rhizopuspepsin (1), penicillopepsin (3), trypsin (2), and matril-
or equivalent residues in an extended â-strand confor-
mation.20b For example, Figure 4 shows a Ramachan-
dran plot of Φ,Ψ or pseudo Φ,Ψ angle pairs taken from
available crystal structures for the peptide backbone of
22a
22b
23
25
26
ysin (1). PDB codes: 1cpi,
1mtr,
1b6p, 5apr, 1tps,
27
28
29
30
1
tyn, 1mmp, 1bxo, 2wea, 2wed.
1
0 cyclic inhibitors bound to the proteases: HIV-1
2
2,23
25
26,27
protease (3),
rhizopuspepsin (1), trypsin (2),
Cyclic P r otea se Su bstr a tes. Since cyclization pre-
matrilysin (1),28 and penicillopepsin (3).
29,30
Most of the
organizes inhibitors for binding to a protease (e.g. 1-6),
we decided to examine this approach for optimizing
protease-binding of substrates. Scheme 1 summarizes
the synthesis of a novel bicyclic substrate 7, an analogue
of 3 but with the N- and C-terminal cyclic peptides
linked together by an amide bond instead of a hydroxy-
ethylamine transition-state isostere. The two chiral
centers in each cycle of 7 are simply derived from
L-amino acids. The N-terminal cyclic acid was prepared
(Scheme 1) by bromoacylating the N-terminus of the
dipeptide, H-Val-Tyr-OBz, followed by ring closure via
the p-hydroxy substituent of Tyr.
coordinates lie in the upper left quadrant of the plot (Φ
3
1
-
56.9° to -169.5°, Ψ 35.2° to 161.7°) and typify the
extended â-strand peptide conformation. The two excep-
tions in the upper right quadrant are for compounds 2
and 3 but are due to the transition-state isostere
replacement of an amide bond. The two other exceptions
in the lower left quadrant result from the piperidine ring
of the trypsin inhibitor 5 (lower point) and the turn
region of the rhizopuspepsin cyclic inhibitor (upper
point). These data establish that the 10 cyclic inhibitors
are effective structural mimics of the extended â-strand
peptide conformation. Other cyclic protease inhibitors
and all adopt an extended peptide
or pseudo-peptide conformation.
Each cycle in 7 is constrained by the presence of only
15 or 16 ring atoms, by two trans amide bonds, and by
a para-substituted aromatic ring. Evidence for strain
are also known,2
0,32

Conformational Selection by Proteolytic Enzymes
J ournal of Medicinal Chemistry, 2000, Vol. 43, No. 7 1275
1
in each cycle is demonstrated in the H NMR data for 7
(
information about the binding affinity of 7 versus 8 for
a protease is to consider these compounds as competitive
inhibitors, since:
see Experimental Section) by the appearance of four
separate aromatic proton resonances per cycle, the
inequivalence of the aromatic protons being indicative
of restricted rotation of each aromatic ring.
k1
E + I y z EI
k-1
We therefore chose to measure substrates 7 versus 8
as competitive inhibitors of the processing of another
substrate, the fluorogenic substrate 2-Abz-Thr-Ile-Nle-
Phe(p-NO2)-Gln-Arg-NH2 (Abz-NF*-6; Km ) 26 µM), by
HIV-1 protease at pH 6.5 (I ) 0.1 M, 37 °C). Competition
for the protease by 7 and 8, which are not directly
monitored, is detectable only as reduced proteolytic
processing of Abz-NF*-6. We could thus directly obtain
protease affinities for substrates by determining Ki ()k1/
k-1) for 7 (0.8 µM) and 8 (60 µM) in these experiments.
Ki is the equilibrium constant which defines the affinity
of the inhibitor (I) for the enzyme (E) and is directly
related to the free energy of binding (∆G ) -RT ln K).
Thus bicycle 7 was found to have about 80-fold higher
affinity for HIV-1 protease than linear 8, corresponding
to about 11 kJ /mol difference in binding energy.
We also directly measured the processing of sub-
strates 7 and 8 using HPLC to monitor peptide cleavage
products. Interestingly, 7 was a poorer substrate (Km
)
110 µM) than 8 (Km ) 20 µM) as defined by the
Michaelis-Menton equilibrium constant, yet 7 is more
constrained to an extended conformation and has the
higher affinity for the protease as determined above
using competition experiments with a fluorogenic sub-
strate. This difference can be accounted for if the
enzyme-substrate complex (ES) decomposes more slowly
for 7 than for 8, not a surprising result since a good
substrate should not bind so tightly to enzyme as to
impede turnover, but rather there needs to be a com-
promise between enzyme affinity and product formation
and dissociation. Substrate 7 consists of the same two
cyclic components as in 1 and 2, which each constrain
the peptide backbone to the preferred extended confor-
mation (Figure 3a), and its N-terminal cycle is predicted
to align in the protease active site in a similar location
as does the cycle in 1.
We were interested in comparing the relative affini-
ties for HIV-1 protease of bicycle 7 and its more flexible
linear hexapeptide analogue 8. Traditionally substrates
such as 7 and 8 are examined for their turnover kinetics,
measuring Km and kcat. The latter parameter was
difficult to obtain in water because these small mol-
ecules are not very water soluble and because they are
only processed slowly. Also while Km ()(k-1 + k2)/k1)
does reflect the affinity of the substrate for the enzyme,
it is complicated by also including the rate of decompo-
sition of the enzyme-substrate complex:
k1
k2
E + S y z ES
98 E + P
k-1
Acyclic P r otea se Su bstr a tes. Conformational se-
lection of substrates has not previously been demon-
A much more effective method of obtaining meaningful
Sch em e 1

1
276 J ournal of Medicinal Chemistry, 2000, Vol. 43, No. 7
Fairlie et al.
Figur e 5. CDspectra ofpeptides13(Ac-KARVLVAEAMSQVTNP-
NH , - - -), 14 (EQADKLAKEAAR KVLVF LNEAAAKAA,
s), and 15 (Ac-PSYKGRP GNF VLQS RP-NH , - - -) in water
2
2
at 22 °C. Arrow indicates cleavage site for HIV-1 protease.
Protease-binding domain is shown in bold.
strated for any protease, although many of the inhibitors
crystallized with proteases are substrate analogues and
do adopt an extended conformation. If fixing the ex-
tended conformation of a substrate leads to higher
affinity for a protease (e.g. 7 vs 8), then the reverse
action of stabilizing substrates in alternative conforma-
tions (e.g. R-helices or â-turns) should reduce their
affinity for proteases. To test this proposition, we
synthesized three peptide substrates for HIV-1 protease
F igu r e 6. 1H NMR data for sequence assignment and iden-
tification of secondary structure in 14 (top) and 15 (bottom).
Filled bars indicate sequential NOE connectivities, bar height
indicating their strength; d(i,i+1) connectivities not detected
due to overlap are asterisked. Lightly shaded bars correspond
to sequential connectivities to Pro Hδ’s. Values of J HNHR are
indicated by V if <5 Hz, v if >8 Hz.
(
(
13-15), each containing seven sequential amino acids
P4-P3′) from polypeptide substrates that are known
to bind in the active site of the protease and are known
to be cleaved by HIV-1 protease (Figure 4, bold). For
example, peptide 13 contains the p24 cleavage site, and
3
1
4 has the RT/IN cleavage site, while 15 contains the
p7/p6 cleavage junction. The solution conformations of
the experimental difficulties and different substrate se-
quences make it difficult for direct quantitative com-
parisons of protease affinities and substrate-processing
rates to be made between 13-15, it is clear from their
Km values at least that 14 and 15 do have low affinities
for the protease. These results offer some qualitative
support for the idea that helical and turn conformations
are not preferred by the protease and that the protease
binds only extended conformations of substrates (Fig-
ures 1-3). The low rates of substrate processing may
reflect the unfavorable equilibrium between the recog-
nized extended conformation and the unrecognized
helical/turn conformation (Figure 7). Further work is
progressing toward stabilizing a single peptide sequence
in three different conformations in order to less ambigu-
ously confirm conformational recognition of substrates
by a protease.
Con for m a tion a l Selection . The requirement of
extended conformations for protease substrates is logical
for several reasons. First, energy that is expended in
folding peptides into helices, turns, or sheets will be
conserved if these structures are resistant to proteolytic
degradation but would be wasted if peptide cleavage was
conformationally indiscriminant. Second, it is intuitively
obvious that the stretching of the substrate amide bond
toward the transition state, which is largely dictated
by interactions between enzyme and ‘pocket-filling’
inhibitor/substrate substituents, is most efficient when
these forces are directionally opposed as enforced by the
these peptides are predicted6
e,33
to exist in different
solution conformations (13, random coil; 14, R-helix; 15,
turn) when incorporated into longer peptides (Figure 5),
and these qualitative structure predictions are con-
firmed by their CD spectra in Figure 5.
Peptide 14 has the known protease-binding heptapep-
tide component of polypeptide substrates incorporated
into a conformationally biased unnatural sequence
designed to further stabilize the R-helix. This compound
1
appears to be helical in water by CD (Figure 5), and H
NMR spectroscopy confirmed that 14 and 15 retain
significant structure in aqueous solution (Figure 6).
Sequential NOE connectivities, coupling constants, and
chemical shift indices (Figure 6) indicate that 14 is
helical along its entire length, while 15 has a small
number of medium range NOEs consistent with turns
around residues PSYK and RPGN.
Using rpHPLC to monitor substrate cleavages over
time, we found that 13-15 are all processed by HIV-1
protease but only very slowly, far too slowly to measure
Ki in the presence of the fluorogenic substrate Abz-NF*-
6
. We therefore resorted to measuring Km values for
these substrates. The random coil substrate 13 (Km )
40 µM), the helical substrate 14 (Km ) 300 µM), and
1
6
the turn-containing substrate 15 (Km ) 10 µM) are very
poor substrates for HIV-1 protease, requiring days to
completely turn over thus preventing us from obtaining
reliable values for kcat for these poor substrates. While

Conformational Selection by Proteolytic Enzymes
J ournal of Medicinal Chemistry, 2000, Vol. 43, No. 7 1277
It is conceivable that molecules that are more closely
related to the ultimate protease-binding shape, the
extended conformation, still bind to the protease via a
3
5a
multistep templating process
in which part of the
extended ligand surface draws together the protease
surfaces. Whether proteases in general initially select
for an extended conformation remains to be confirmed.
If it is true in general, then the rate of substrate
processing should be influenced by conformational
equilibria and disfavored by high populations of con-
formations that need to rearrange prior to protease
binding. This was not possible to fully test here with
the limited range of (poor) substrates available for the
one protease examined but is the subject of a more
detailed kinetic study for HIV-1 protease using a single
substrate sequence locked into different highly stable
conformations that do not readily interconvert (unpub-
lished data).
Im p lica tion s for Dr u g Design . Biological proper-
ties of peptides and proteins are frequently interpreted
on the basis of their observed solution or solid-state
structures. However the most abundant solution struc-
ture of a molecule is not necessarily the one recognized
by an enzyme. The above results clearly demonstrate
that inhibitors and substrate analogues at least end up
bound in an extended conformation to the active site of
a protease. There are profound consequences for drug
design of binding to proteases by their substrates,
substrate analogues, and inhibitors in extended â-strand
conformations. First, it suggests the underutilized
strategy of conformationally fixing protease inhibitors
in an extended â-strand in order to optimize their
potency and selectivity, an approach highlighted here
by conformationally constrained macrocycles, but there
are also other ways of stabilizing extended â-strands.
Second, the design of protease inhibitors should be
greatly simplified since new proteases, for which there
is no or limited structural information, can still be
effectively targeted based only upon knowledge of the
class of protease and its substrate specificity. Third,
proteolytic degradation of other bioactive peptides (ago-
nists, antagonists, etc.) might be minimized by the
reverse strategy of stabilizing their turn or helix con-
formations so that the extended strand conformation is
less accessible. Conformational selection should now be
examined for other proteases, and attempts should be
made to see if there are correlations between confor-
mational equilibria and processing rates of polypeptide
substrates and conformationally biased analogues.
F igu r e 7. Conformational equilibria for substrates and
inhibitors and the proposed conformational selection by pro-
teases.
extended form due to the trans amide bond and the
L-conformation of each flanking amino acid. Third,
linear extension of a peptide exposes the main chain
amide atoms, otherwise shielded by side chains in intra-
molecularly hydrogen-bonded secondary structures, and
promotes intermolecular hydrogen bonding with an
enzyme. Fourth, proteins that are substrates for HIV-1
protease are processed between elements of secondary
structure, rather than within helical, turn, or sheet
regions, and rates of processing tend to be greater after
denaturation of protein substrates.34 Fifth, and impor-
tantly in the context of the substrates 13-15, the diam-
eter of the inhibitor-bound conformation of the active
site of HIV-1 protease (∼6 Å) is far too small to accom-
modate substrates in either helical or turn conforma-
tions so the latter forms would not be expected to be
recognized prior to rearrangement to an extended form.
3
5
This latter observation is not consistent with a model
in which the peptide is recognized in its most stable
solution conformation (e.g. helix, turn) and then rear-
ranged into the extended strand on protease binding.
These factors all support a conformational selection
hypothesis, where different conformers interconvert
(Figure 7) but only the extended conformer is recognized
and processed by proteases. This could be construed as
a controversial hypothesis. For example, it has been
suggested elsewhere35 that entropic considerations do
not support the idea that protein-ligand interactions
in general proceed via conformational selection, but
rather are more likely to proceed via a two/multistep
‘zipper’ process. We agree that such an induced fit model
of protein-ligand interaction is more attractive than a
rigid lock-and-key model for protease-ligand interac-
tion, but the truth may well be somewhere in between.
Conformational restrictions that preorganize the ligand
structure close to its final protease-bound conformation
can be expected to reduce the entropic barrier en route
to a protease-ligand complex. We have only presented
data in this paper that suggest that proteases ultimately
bind their ligands in an extended ligand conformation
and cannot definitively conclude from the crystal struc-
tural data alone that the initial interaction between
protease and ligand requires the ligand to be in an
extended conformation. However it does appear from
our preliminary modeling on the size of the active site,
and the structural and binding data for just one pro-
tease, HIV-1 protease, that there is support for the idea
that helical and turn conformations are not recognized
by this particular protease. For other protease active
sites which are larger or more open and accessible to
helices, turns, and sheets, these may well be accom-
modated in the active site though such examples are
not presently known.
Con clu sion s
This work provides the most comprehensive analysis
to date for the importance of inhibitor conformation in
inhibitor-protease interactions. On the basis of super-
positions of over 250 protease-inhibitor crystal struc-
tures, the first strong evidence has been pooled together
to show that aspartic, serine, cysteine, and metallo
proteases commonly bind to the extended â-strand
conformation of their inhibitors, including substrate
analogues and nonpeptidic inhibitors. Cyclic inhibitors
and a cyclic substrate have been used to show that the
basic strategy of conformationally constraining inhibi-
tors/substrates to the protease-binding structure (pre-
organization) is an effective way of increasing affinity

1
278 J ournal of Medicinal Chemistry, 2000, Vol. 43, No. 7
Fairlie et al.
for a protease. This suggests that general methods
which can successfully preorganize molecules in the
protease-binding conformation will similarly increase
protease affinity, and some of these methods are already
being described for specific proteases.20b,21 Furthermore,
for proteases where the structure is unknown, but
substrate specificity and protease type are known,
approaches to fixing an extended conformation for
inhibitors may be useful for producing lead molecules
that bind the protease tightly. Finally, some preliminary
data is presented which is consistent with the idea that
peptide conformations other than the extended strand
may be more resistant to protease cleavage, suggesting
that protein folding protects peptide amide bonds and
thus that equilibria between peptide conformations
could be important determinants of substrate activity.
We are currently examining peptide substrates in more
detail for kinetic and structural evidence of a confor-
mational preference in substrate processing by pro-
teases.
using HBTU activation and DIPEA in situ neutralization. This
involved washing the resin in DMF, cleaving the N-terminal
Boc protecting group by shaking in 100% TFA (2 × 1 min)
and flow washing with DMF (1 min) to remove the TFA. Each
coupling involved activating the amino acid (4 equiv) by mixing
with 0.5 M HBTU (4 equiv) and DIPEA (5 equiv), and then
adding this mixture to the resin (1 equiv) and shaking 10 min.
Couplings were monitored by the quantitative ninhydrin test
for remaining free amine. The resin was then flow washed for
1
min and the procedure of deprotection and coupling was
repeated for the other amino acids. After the final amino acid
was added the resin was flow washed with DMF and DCM
and then dried under a flow of nitrogen gas. Peptide 14 was
synthesized on MBHA resin (Peptide Institute) with substitu-
tion value of 0.57 mmol/g. Peptide 8 was synthesized synthe-
sized on a Boc-Val-Pam resin (Peptide Institute) with substi-
tution value of 0.70 mmol/g and then methylated (HCl/MeOH).
Side-chain-protecting groups were N-tosyl for Arg, xanthyl
(Xan, temporary protection) for Asn and Gln, benzyloxy (BzlO)
for Asp and Glu, and N-2-chlorobenzyloxycarbonyl (ClZ) for
Lys. Prior to cleavage of the peptide from the resin the
N-terminal Boc protecting group was removed, the resin was
flow washed with DMF and DCM and then dried under a flow
of nitrogen. Cleavage of ∼300-400 mg resin was performed
in an all Teflon apparatus specially designed for handling HF.
Treatment with liquid HF (10 mL) and p-cresol (1 mL) at -5
Exp er im en ta l Section
All materials were reagent grade; amino acids were pur-
chased from Novabiochem. Gradient HPLC was performed on
Waters C-18 analytical (15 µm, 8 mm × 100 mm) and
semipreparative (15 µm, 25 mm × 100 mm) columns. Analyti-
cal runs were 100% A for 2 min, then 0-60% B gradient over
°
C for 1-2 h was followed by treatment with deoxygenated
ether to precipitate the cleaved peptide, which was filtered and
dissolved in 50% acetonitrile/water and then freeze-dried.
Purification by rpHPLC used a gradient of 100% A (2 min),
1
00% A/0% B to 20% A/80% B (40 min), 20% A/80% B to 100%
B (10 min), 100% B (5 min). Ac-LVFFIV-OMe (8): t ) 56.0
min; ES-MS calcd 793.5, found 793.2. Suc-EQADKLAKEAA-
RKVLFLNEAAAKAA-NH (14): t ) 24.5 min; ES-MS calcd
755.5, found 2755.3.
P ep tid e 15. This was synthesized by Fmoc chemistry on
6
0 min at 2 mL/min, where buffer A is 0.1% TFA in H
2
O and
R
buffer B is 0.1% TFA in 90% CH CN/H O. Mass spectra were
3
2
obtained on a triple quadrupole mass spectrometer (PE SCIEX
API III) equipped with an ionspray (pneumatically assisted
electrospray) atmospheric pressure ionization source (ISMS).
Solutions of compounds in 9:1 acetonitrile/0.1% aqueous tri-
fluoroacetic acid were injected by syringe into the spectrom-
2
R
2
Ramage resin (Novabiochem) with a substitution value of 0.55
mmol/g. Amino acids were assembled by manual stepwise
solid-phase peptide synthesis using HBTU/DIPEA activation.
The procedure involved washing the resin in DMF, cleavage
of the N-terminal Fmoc protecting group by shaking in 50%
piperidine/DMF (2 × 1 min) and flow washing with DMF (1
min). Each coupling involved activating the amino acid (4
equiv) by mixing with 0.5 M HBTU (4 equiv) and DIPEA (4
equiv), and then adding this mixture to the resin (1 equiv) and
shaking 10 min. Couplings were monitored by the quantitative
ninhydrin test. The resin was then flow washed for 1 min and
the procedure of deprotection and coupling was repeated for
the other amino acids. After the final amino acid was added
the resin was flow washed with DMF and DCM and then dried
under a flow of nitrogen gas. Side-chain-protecting groups were
n+
eter. Molecular ions, {[M + nH] }/n, were generated by the
ion evaporation process and focused into the analyzer of the
mass spectrometer through a 100-mm sampling orifice. Full
scan data was acquired by scanning quadrupole-1 from m/z
1
00-900 with a scan step of 0.1 Da and a dwell time of 2 ms.
Mass spectra were processed using MacSpec version 3.2 (PE-
Sciex Instruments). Accurate mass determinations were per-
formed on a KRATOS MS25 mass spectrometer using electron
impact ionization.
Com p u ter Su p er im p osition s. Protease-inhibitor crystal
structures were identified in the Protein Data Bank (http://
www.rcsb.org/pdb) and the PDB codes used are listed in Table
1
(Supporting Information). These represented the majority
of structures available at the time of analysis. The protease-
inhibitor structures were displayed using the program In-
sightII 95.0 (Molecular Simulations Inc., San Diego, CA) and
the R-carbon atoms of a selection of proteases were superim-
posed using the homology module and superimposition func-
tion. Only the inhibitor/substrate conformations are displayed
in Figures 1-3. The reported rmsd values for Figures 1 and 3
only reflect structural deviations among the protein conforma-
tions. In the cases of multiple proteases bound to a common
inhibitor (e.g. pepstatin A), only the inhibitor conformations
were superimposed (e.g. Figure 2) and the rmsd values indicate
the extent of conformational match for the same inhibitor
bound to different proteases.
2
,2,5,7,8-pentamethylchroman-6-sulfonyl (Pmc) for Arg, trityl
(Trt) for Asn and Gln, tert-butyloxycarbonyl (Boc) for Lys, and
tert-butyl (tBu) for Ser and Tyr. The peptide was cleaved from
resin (200-300 mg) by stirring with 3 mL of a solution of TFA:
TIPS:H O (95:2.5:2.5) for 2-3 h for Arg(Pmc)-protected pep-
2
tides and for 5-6 h for Arg(Mtr)-protected peptides. The TFA
was removed on a rotary evaporator at 25 °C, ether was used
to precipitate the cleaved peptide which was filtered and
dissolved in 50% acetonitrile/water and then freeze-dried.
Purification by rpHPLC used a gradient of 100% A (2 min),
100% A/0% B to 20% A/80% B (40 min). Ac-PSYKGRPGN-
FLQSRP-NH
744.8.
5-Br om op en ta n oyl-L-va lin yl-L-tyr osin e Ben zyl Ester
2 R
(15): t ) 18.54 min; ES-MS calcd 1744.9, found
1
Syn th esis. The synthesis of inhibitors 1-3 has been
2
2
described previously. The bicyclic substrate 7 was made by
some modifications to this method involving solution-phase
coupling of the N-terminal cyclic carboxylate 11 with the
C-terminal cyclic amine 12 using BOP/DIPEA. It was char-
(10). A solution of L-valinyl-L-tyrosine benzyl ester p-toluene-
sulfonate (9; 10 mmol) in THF (30 mL) and water (30 mL)
was stirred at room temperature while solutions of K CO (2.83
2 3
g) in water (10 mL) and 5-bromopentanoyl chloride in THF (5
mL) were added over 10 min. After a further 30 min the
solution was diluted with EtOAc and washed with 2 M HCl,
1
13
acterized by NMR ( H, C) and MS data. Peptides 8 and 13-
1
5 were synthesized by standard solid-phase synthesis pro-
tocols using Boc or Fmoc chemistry, purified by rpHPLC and
characterized by MS and NMR spectra.
P ep tid es 8 a n d 14. Amino acids were Boc-protected and
assembed by manual stepwise solid-phase peptide synthesis
5% NaHCO
solvent gave a white powder 10 (5.06 g, 95%): R
EtOAc/hexane); 13C NMR (CDCl
) δ 172.9, 171.1, 171.0, 155.4,
3
and brine and dried over MgSO
4
. Removal of
f
0.55 (75%
3

Conformational Selection by Proteolytic Enzymes
J ournal of Medicinal Chemistry, 2000, Vol. 43, No. 7 1279
1
3
35.0, 130.4, 128.6, 126.7, 115.6, 67.4, 58.3, 53.3, 36.9, 35.4,
3.1, 32.0, 31.2, 24.2, 19.1, 18.2.
Bicyclic Su bstr a te 7. A solution of the macrocyclic acid
11 (155 mg, 0.428 mmol), the macrocyclic amine 12 (109 mg,
0
.244 mmol) and BOP reagent (190 mg, 0.43 mmol) was stirred
(
9S,12S)-12-Ca r boben zyloxy-7,10-d ioxo-9-isop r op yl-2-
with THF (2 mL) at 0 °C, then diisopropylethylamine (160 µL)
was added. The mixture was stirred at 0 °C for 60 min then
concentrated in vacuo. The residue was dissloved in the
minimum volume of warm DMF then diluted with about 2
oxa-8,11-diazabicyclo[12.2.2]octadeca-14,16,17-tr ien e (11).
5
1
-Bromopentanoyl-L-valinyl-L-tyrosine benzyl ester (935 mg,
.75 mmol) was converted to the corresponding iodide by
refluxing with NaI (395 mg) in acetone (20 mL) for 2 h, then
the precipitated NaCl was filtered off and the solvent was
evaporated. The residual iodide was dissolved in dry DMF (50
mL) and anhydrous K CO (500 mg) was added. The mixture
2 3
was stirred at room temperature overnight then evaporated
in vacuo. The residue was partitioned between water and 50%
volumes of 50% MeCN/H
reverse-phase HPLC before the product aggregated from
solution (linear gradient 30% MeCN/70% H O/0.1% TFA to
O/0.1% TFA over 20 min, retention time
7 min). Lyophilization gave a white powder 7 (91 mg, 55%):
2
O and immediately purified by
2
9
0% MeCN/10% H
2
1
1
H NMR (500 MHz, DMSO-d ) δ 8.09 (d, J ) 9.8 Hz, 1H, Phe-
6
EtOAc/CHCl
solid polymer, then washed with dilute sodium thiosulfate
solution, 2 M HCl, and NaHCO , and dried over MgSO
Removal of solvent gave a solid which was purified by flash
chromatography (CHCl -50% EtOAc/CHCl ) giving the benzyl
ester of 11 as a white powder (290 mg, 37%): R 0.28 (75%
) δ 7.45-7.35 (m,
H, ArH), 7.21 (dd, J ) 8.4, 2.2 Hz, 1H, ArH), 6.92-6.84 (m,
H, ArH), 6.78 (dd, J ) 8.3, 2.7 Hz, 1H, ArH), 5.83 (d, J ) 9.9
3
and the organic layer was filtered to remove
NH), 7.90 (d, J ) 7.4 Hz, 1H, Phe-NH), 7.33 (d, J ) 9.3 Hz,
H, Val-NH), 7.28 (m, 1H, CH NH), 7.18-7.13 (m, 2H, Ar-
1
2
3
4
.
H), 7.01 (m, 2H, Ile-NH and Ar-H), 6.84 (m, 2H, Ar-H), 6.81-
6
4
.74 (m, 2H, Ar-H), 6.67 (dd, J ) 8.2, 2.6 Hz, 1H, Ar-H),
.58 (m, 1H, Phe-RH), 4.53 (m, 1H, Phe-RH), 4.30 (m, 1H, H-3′),
3
3
f
1
4.11 (m, 1H, H-3′), 4.09 (m, 1H, H-3), 4.04 (m, 1H, H-3), 3.99
EtOAc/hexane); H NMR (300 MHz, CDCl
3
(
t, J ) 8.5 Hz, 1H, Val-RCH), 3.49 (t, J ) 7.9 Hz, 1H, Ile-
RCH), 3.30 (m, 1H, H-5), 3.10 (m, 1H, Phe-âCH ), 3.08 (m, 1H,
Phe-âCH ), 2.66 (m, 1H, H-5′), 2.48 (m, 1H, Phe-âCH ), 2.43
m, 1H, Phe-âCH ), 2.13 (m, 1H, H-6), 1.98 (m, 1H, H-4′), 1.83
5
2
2
Hz, 1H, tyr NH), 5.69 (d, J ) 8.8 Hz, 1H, val NH), AB system:
2
2
(
(
(
(
(
(
2
δ
1
3
A
5.26, δ
4.9, 10.0, 4.8 Hz, 1H, tyr RH), 4.29-4.12 (m, 2H, CH
.95 (dd, J ) 8.8, 7.1 Hz, 1H, val RH), 3.43 (dd, J ) 13.6, 4.8
B
5.20 J AB ) 13.1 Hz, 2H, OCH
2
Ph), 5.06 (ddd, J )
m, 1H, H-6), 1.68 (m, 1H, H-4′),1.66 (m, 1H, Val-âCH), 1.54
m, 1H, H-5), 1.46 (m, 1H, H-4), 1.41 (m, 1H, Ile-âCH), 1.30
m, 1H, H-5), 1.28 (m, 1H, Val-γCH
m, 1H, Ile-γCH ), 0.75 (d, J ) 6.8 Hz, Ile-γCH
m, 9H, Val-γCH and Ile-γCH ), 0.63 (d, J ) 6.8 Hz, 3H, Ile-
2
OAr),
), 1.17 (m, 1H, H-4), 0.83
Hz, 1H, tyr âH), 2.53 (dd, J ) 13.6, 12.1 Hz, 1H, tyr âH), 2.26-
2
), 0.76-0.70
2
2
.17 (m, 1H) and 2.09-1.98 (m, 1H, CH
2
CO), 1.98-1.70 (m,
), 0.84
) δ 172.07, 171.25,
70.02, 155.71, 135.04, 131.12, 129.99, 128.72, 128.69, 128.46,
2
2
H, CH ), 1.83 (m, 1H, val âH), 1.55-1.31 (m, 2H, CH
2
2
3
3
1
3
1
3
δCH ); C NMR (DMSO-d ) δ 171.3, 170.7, 170.3, 169.8, 169.8,
(d, J ) 6.8 Hz, 6H, (CH
3
)
2
); C NMR (CDCl
3
3
6
1
1
57.6, 155.1, 131.4, 131.1, 130.5, 129.6, 129.5, 129.3, 118.5,
18.4, 118.0, 117.5, 68.2, 68.1, 57.2, 57.1, 55.2, 54.1, 38.3, 38.2,
1
1
3
28.34, 118.34, 116.52, 67.81, 67.49, 58.16, 52.48, 38.46, 36.12,
+
37.7, 35.7, 35.0, 31.4, 26.7, 25.9, 24.6, 21.7, 19.3, 18.8, 15.0,
1.78, 25.98, 21.66, 18.77, 18.40; HRMS m/e 452.2312 (M ),
+
1
7
51 5 7
1.5; HRMS m/z 700.3666 (MNa ), calcd for C37H N O Na
00.3686.
calcd for C26
32 2 5
H N O 452.2311.
A solution of this benzyl ester of 11 (3.0 g, 6.6 mmol) in
MeOH (75 mL) was hydrogenated over 10% Pd-C, 2 atm, room
temperature for 3 h. The catalyst was filtered off and the
solvent was removed in vacuo giving the carboxylic acid 11 as
En zym e In h ibition a n d P r ocessin g. HIV-1 protease
assays36 were carried out using synthetic [Aba 67,95,167,195]-
HIV-1 protease (SF2 isolate) with Cys residues replaced by
1
R-aminobutyric acid (Aba). K values were calculated from IC
a white powder (2.4 g, 100%): H NMR (300 MHz, CD
.35 (d, J ) 9.9 Hz, 1H tyr NH), 7.47 (d, J ) 9.3 Hz, 1H,
Val-NH) 7.19 (dd, J ) 8.5, 2.2 Hz, 1H, ArH), 7.00 (dd, J ) 8.2,
.2 Hz, 1H, ArH), 6.85 (dd, J ) 8.4, 2.6 Hz, 1H, ArH), 6.76
dd, J ) 8.2, 2.6 Hz, 1H, ArH), 4.87-4.76 (m, 1H, tyr RH),
.29-4.19 (m, 1H, CH OAr), 4.14-4.02 (m, 1H, CH OAr), 4.00
d, J ) 8.4 Hz, 1H, val RH (NH exchanged)), 3.35 (dd, J )
3.5, 4.3 Hz, 1H, tyr âH), 2.62 (dd; J ) 13.5, 12.7 Hz, 1H, tyr
âH), 2.20-2.07 (m, 2H, CH CO), 1.88-1.80 (m, 2H, val âH
and CH ), 1.60-1.25 (m, 3H, CH ), 0.90 (d, J ) 6.8 Hz, 3H,
CH ), 0.84 (d, J ) 6.7 Hz, 3H, CH
3
OD) δ
i
50
values determined for substrates 7, 8, and 13-15 or inhibitors
8
1
2
-3 incubated at pH 6.5, I ) 0.1 M, 37 °C, 50 mM substrate
-Abz-Thr-Ile-Nle-Phe(p-NO )-Gln-Arg-NH (Abz-NF*-6) using
2
2
2
(
3
7
a continuous fluorimetric assay on a Perkin-Elmer LS50B
luminescence spectrometer and assuming competitive inhibi-
4
2
2
tion. K values were determined at pH 5.5, I ) 0.1 M and 37
m
(
1
°
C using fixed point assays for substrates incubated over 6-24
h and the products were identified and analyzed using rpHPLC
and electrospray mass spectrometry.
2
2
2
1
3
3
3
); C NMR (CD
3
OD) δ
CD Sp ectr a . CD data were recorded on a J ASCO 710
spectropolarimeter. Peptides were dissolved in 10 mM phos-
phate buffer (pH 7) and measurements were made at 20 °C in
a 3-mL cuvette with a path length of 1 cm, scanning from 250
to 190 nm every 0.10 nm, with bandwidth of 1 nm. Wavelength
scans were corrected for buffer scans taken at 20 °C. Peptide
concentrations were determined by amino acid analysis.
Percent helicity was calculated38 based on a predicted mean
residue ellipticity ([ø]Hn) using [ø]Hn ) [ø]H∞ (1-2.5/n), where
1
1
1
3
74.93, 174.35, 172.48, 156.66, 132.56, 131.46, 130.56, 119.63,
19.40, 69.36, 59.54, 54.18, 38.50, 36.46, 32.89, 27.24, 22.84,
9.54, 19.04; HRMS m/e 362.1841, calcd for C19
62.1842.
26 2 5
H N O
(
8S ,11S )-11-Am in o-7,10-d ioxo-8-(2-b u t yl)-2-oxa -6,9-
d ia za bicyclo[11.2.2]h ep t a d eca -13,15,16-t r ien e (12). The
2
2b
N-Boc-protected form
of macrocycle 12 (1.00 g, 2.31 mmol)
was dissolved in TFA (10 mL), then after a further 5 min, the
TFA was evaporated in vacuo. The residue was dissolved in
water and purified by preparative reverse-phase HPLC (linear
gradient, 100% water increasing MeCN at 1% per min + 0.1%
TFA), room temperature 15 min. Lyophilization gave the
macrocyclic amine 12 as a white powder (525 mg): 1H NMR
2
-1
[
ø]H∞ ) -37400 deg‚cm ‚dmol at a wavelength of 222 nm
for a helix of infinite length, n is the number of residues in
the helix, and the value of 2.5 is a wavelength-dependent
constant for 222 nm.
NMR Sp ectr oscop y. Proton NMR spectra were recorded
2
(
)
6
(
3
300 MHz, CD
7.8 Hz, 1H, Ile-NH), 7.23 (dd, J ) 8.4, 2.1 Hz, 1H, ArH),
.97-6.79 (m, 3H, ArH), 4.41-4.31 (m, 1H, H-3), 4.28-4.16
m, 1H, H-3), 4.08 (dd, J ) 10.7, 7.0 Hz, 1H, Tyr-RCH), 3.60-
.45 (m, 1H, H-5), 3.47 (t, J ) 7.6 Hz, 1H, Ile-RCH), 3.28 (dd,
J ) 12.4, 7.0 Hz, 1H, Tyr-âCH), 2.86-2.75 (m, 1H, H-5), 2.70
dd, J ) 12.4, 10.7 Hz, 1H, Tyr-âCH), 2.30-2.09 (m, 1H, H-4),
.83-1.69 (m, 1H, H-4), 1.60-1.39 (m, 2H, Ile-âCH and Ile-
γCH ), 1.01-0.88 (m, 1H, Ile-γCH ), 0.83 (t, J ) 7.2 Hz, 3H,
Ile-δCH ), 0.75 (d, J ) 6.8 Hz, 3H, Ile-γCH
OH) δ 171.5, 168.6, 160.0, 132.2, 130.4, 128.2, 118.8, 118.7,
8.6, 59.8, 56.0, 40.0, 38.0, 37.5, 27.6, 26.2, 14.9, 11.7; IS-MS
3
OH) δ 7.79 (broad m, 1H, NHCH
2
), 7.22 (d, J
for peptides (1-3 mM; 90% H3
2
O/10% H
2
O) on a Bruker ARX-
500 at 280 K as described, 9 processed using XWIN NMR
(Bruker), and analyzed using AURELIA (Bruker). Resonances
were assigned using standard sequential procedures. The
pattern of NOEs was used to identify secondary structure.
Sequence specific assignments for 14 were based on sequential
NOE connectivities (Figure 6a). Despite overlap in the spectra,
medium range NOEs (dRN(i,i+2), dRN(i,i+3), dRN(i,i+4), dRâ-
(i,i+3)) were observed for a number of residues, consistent with
a helical conformation. For the majority of residues where
these NOEs have not been identified, the presence of other
overlapping cross-peaks in the spectra prevents their observa-
tion. Also consistent with a helical conformation for the
(
1
2
2
1
3
3
3 3
); C NMR (CD -
6
+
m/z 334 (MH ).

1
280 J ournal of Medicinal Chemistry, 2000, Vol. 43, No. 7
Fairlie et al.
peptide, a large number of residues have coupling constants
C virus NS3 protease domain complexed with a synthetic NS4A
3
cofactor peptide. Cell 1996, 87, 343-55; erratum: Cell 1997, 8,
J
HNHR < 5 Hz and the HR chemical shifts of the majority of
1
59. (b) Love, R. A.; Parge, H. E.; Wickersham, J . A.; Hostomsky,
the residues are shifted significantly upfield compared with
their random coil values. The pattern of NOE connectivity
observed, J HNHR coupling constants, and chemical shift indices
Z.; Habuka, N.; Moomaw, E. W.; Adachi, T.; Hostomska, Z. The
crystal structure of hepatitis C virus NS3 proteinase reveals a
trypsin-like fold and a structural zinc binding site. Cell 1996,
87, 331-42.
3
(Figure 6a) all strongly suggest that the peptide is helical along
its entire length. For peptide 15, sequence-specific assignments
were made from sequential NOE connectivities (Figure 6b).
Strong sequential dRN(i,i+1) connectivities and coupling
(8) (a) Gibson, W.; Hall, M. R. Assemblin, an essential herpesvirus
proteinase. Drug Des. Discov. 1997, 15, 39-47. (b) Shieh, H. S.;
Kurumbail, 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. Three-dimensional structure
of human cytomegalovirus protease. Nature 1996, 383, 279-
82; erratum: Nature 1996, 384, 288.
3
constants J HNHR ) 5-8 Hz suggest that most residues are in
an extended conformation, but a small number of medium
range NOEs are observed consistent with turn conformations
at residues P
1
SYK
4
and R
6
PGN
9
.
(9) Boycott, R.; Klenk, H. D.; Ohuchi, M. Cell tropism of influenza
virus mediated by hemagglutinin activation at the stage of virus
entry. Virology 1994, 203, 313-9.
Ack n ow led gm en t. We thank the Australian Na-
tional Health and Medical Research Council and the
Australian Research Council for some financial support.
(
10) (a) Bernstein, P. R.; Edwards, P. D.; Williams, J . C. Inhibitors
of human leukocyte elastase. Prog. Med. Chem. 1994, 31, 59-
120. (b) Hugli, T. E. Protease inhibitors: novel therapeutic
application and development. Trends Biotechnol. 1996, 14, 409-
12. (c) Fath, M. A.; Wu, X.; Hileman, R. E.; Linhardt, R. J .;
Kashem, M. A.; Nelson, R. M.; Wright, C. D.; Abraham, W. M.
Interaction of secretory leukocyte protease inhibitor with heparin
inhibits proteases involved in asthma. J . Biol. Chem. 1998, 273,
Su p p or tin g In for m a tion Ava ila ble: Table 1 containing
PDB codes for 266 protease-inhibitor structures, as mentioned
in the text. This material is available free of charge via the
Internet at http://pubs.acs.org.
1
3563-9. (d) Tanaka, R. D.; Clark, J . M.; Warne, R. L.; Abraham,
W. M.; Moore, W. R. Mast cell tryptase: a new target for
Refer en ces
therapeutic intervention in asthma. Int. Arch Allergy Immunol.
1
995, 107, 408-9.
(
1) (a) Leung, D.; Abbenante, G.; Fairlie, D. P. Protease Inhibitors:
(
11) Stubbs, M. T.; Bode, W. A player of many parts: the spotlight
Current Status and Future Prospects. J . Med. Chem. 2000, 43,
falls on thrombin’s structure. Thromb Res. 1993, 69, 1-58.
3
05-41. (b) Babine, R. E.; Bender, S. L. Molecular Recognition
of Protein-Ligand Complexes: Applications to Drug Design.
Chem. Rev. 1997, 97, 1359-472. (c) Craik, C. S.; Debouck, C.
Proteases as Therapeutic Targets. In Perspectives in Drug
Discovery and Design; McKerrow, J . H., J ames, M. N. G., Eds.;
ESCOM: Leiden, 1995; Vol. 2, pp 1-125. (d) Van Noorden, J .
F. IBC’s 2nd International Conference on Protease Inhibitors:
Novel Therapeutic Applications and Development, Washington
D.C., Feb 24-26, 1997; Acta Histochem. 1997, 99, 249-55.
2) (a) Beckett, R. P.; Davidson, A. H.; Drummond, A. H.; Huxley,
P.; Whittaker, M. Recent Advances in matrix metalloprotease
inhibitor research. DDT 1996, 1, 16. (b) J ohnson, L. L.; Dyer,
R.; Hupe, D. J . Matrix metalloproteinases. Curr. Opin. Chem.
Biol. 1998, 2, 466-71. (c) Yan, S.; Sameni, M.; Sloane, B. F.
Cathepsin B and human tumor progression. J . Biol Chem. 1998,
(12) (a) Selkoe, D. J . Alzheimer’s disease: genotypes, phenotypes,
and treatments. Science 1997, 275, 630-1. Vassar, R.; Bennett,
B. D.; Babu-Khan, S.; Khan, S.; et al. Beta-secretase cleavage
of Alzheimer’s amyloid precursor protein by the transmembrane
aspartic protease BACE. Science 1999, 286, 735-741.
(13) Seife, C. Blunting nature’s Swiss army knife. Science 1997, 277,
1602-3.
(14) (a) See references to all PDB-filed structures in Table 1. (b) Rich,
D. H. Peptidase Inhibitors. In Comprehensive Medicinal Chem-
istry; Samuels, P. J ., Ed.; Pergamon Press: Oxford, 1990; Vol.
2, pp 391-441 and references therein.
(15) (a) J ames, M. N. G.; Sielecki, A. R. Aspartic proteases and their
Catalytic Pathway inhibitors. In Biological Macromolecules and
Assemblies: Active sites of enzymes; J urnak, F. A., McPherson,
A., Eds.; J ohn Wiley & Sons: New York, 1987; p 413. (b) Bailey,
D.; Cooper, J . B. A structural comparison of 21 inhibitor
complexes of the aspartic proteinase from Endothia parasitica.
Protein Sci. 1994, 3, 2129-43.
(
(
3
79, 113-23.
3) (a) Becker, M. M.; Harrop, S. A.; Dalton, J . P.; Kalinna, B. H.;
McManus, D. P.; Brindley, P. J . Cloning and characterization
of the Schistosoma japonicum aspartic proteinase involved in
hemoglobin degradation. J . Biol. Chem. 1995, 270, 24496-501;
erratum: J . Biol. Chem. 1997, 272, 17246. (b) Brindley, P. J .;
Kalinna, B. H.; Dalton, J . P.; Day, S. R.; Wong, J . W.; Smythe,
M. L.; McManus, D. P. Proteolytic degradation of host hemo-
globin by schistosomes. Mol. Biochem. Parasitol. 1997, 89, 1-9.
4) (a) Silva, A. M.; Lee, A. Y.; Gulnik, S. V.; Maier, P.; Collins, J .;
Bhat, T. N.; Collins, P. J .; Cachau, R. E.; Luker, K. E.; Gluzman,
I. Y.; Francis, S. E.; Oksman, A.; Goldberg, D. E.; Erickson, J .
W. Structure and inhibition of plasmepsin II, a hemoglobin-
degrading enzyme from Plasmodium falciparum. Proc. Natl.
Acad. Sci. U.S.A. 1996, 93, 10034-9. (b) Li, Z.; Chen, X.;
Davidson, E.; Zwang, O.; Mendis, C.; Ring, C. S.; Roush, W. R.;
Fegley, G.; Li, R.; Rosenthal, P. J .; et al. Anti-malarial drug
development using models of enzyme structure. Chem. Biol.
(
(
(
16) Powers, J . C.; Harper, J . W. Inhibitors of serine proteases. In
Protease Inhibitors; Barret, A. J . S., Salverson, G., Eds.; Elsevi-
er: Amsterdam, 1986; pp 55-152.
17) Powers, J . C.; Harper, J . W. Inhibitors of metalloproteases. In
Protease Inhibitors; Barret, A. J . S., Salverson, G., Eds.; Elsevi-
er: Amsterdam, 1986; pp 219-98.
18) (a) Hans-Hartwig, O.; Schirmeister, T. Cysteine proteases and
their inhibitors. Chem. Rev. 1997, 97, 133-71. (b) Rasnick, D.
Small synthetic inhibitors of cysteine protease. In Perspectives
in Drug Discovery and Design; McKerrow, J . H., J ames, M. N.
G., Eds.; ESCOM: Leiden, 1996; pp 47-63.
(
(
(
(
(
19) Dauter, Z.; Betzel, C.; Genov, N.; Pipon, N.; Wilson, K. S.
Complex between the subtilisin from a mesophilic bacterium and
the leech inhibitor eglin-C. Acta Crystallogr. B 1991, 47, 707-
1
994, 1, 31-7.
3
0.
5) (a) Abad-Zapatero, C.; Goldman, R. W.; Muchmore, S. W.;
Hutchins, C.; Stewart, K.; Navaza, J .; Payne, C. D.; Ray, T. L.
Structure of a secreted aspartic protease from C. Albicans
complexed with a potent inhibitor: implications for the design
of antifungal agents. Protein Sci. 1996, 5, 640-52. (b) Abad-
Zapatero, C.; Goldman, R.; Muchmore, S. W.; Hutchins, C.; Oie,
T.; Stewart, K.; Cutfield, S. M.; Cutfield, J . F.; Foundling, S. I.;
Ray, T. L. Structure of secreted aspartic proteinases from
Candida. Implications for the design of antifungal agents. Adv.
Exp. Med. Biol. 1998, 436, 297-313.
(
20) (a) Fairlie, D. P.; Abbenante, G.; March, D. Macrocyclic Pepti-
domimetics: Forcing Peptides into Bioactive Conformations.
Curr. Med. Chem. 1995, 2, 672-705. (b) Fairlie, D. P.; West, M.
L.; Wong, A. K. Towards protein surface mimetics. Curr. Med.
Chem. 1998, 5, 29-62. (c) Reid, R. C.; Fairlie, D. P. Mimicking
extended conformations of protease substrates: designing cyclic
peptidomimetics to inhibit HIV-1 Protease. Advances in Amino
Acid Mimetics and peptidomimetics; J AI Press Inc.: Greenwich,
CT, 1997; pp 77-107.
(
21) (a) Smith III, A. B.; Keenan, T. P.; Holcomb, R. C.; Sprengeler,
P. A.; Guzman, M. C.; Wood, J . L.; Carol, P. J .; Hirschman, R.
Design, synthesis and crystal structure of a pyrrolinone-based
peptidomimetic possessing the conformation of a â-strand:
Potential application to the design of novel inhibitors of pro-
teolytic enzymes. J . Am. Chem. Soc. 1992, 114, 10672-74. (b)
Smith III, A. B.; Hirschmann, R.; Pasternack, A.; Akaishi, R.;
Guzman, M. C.; J ones, D. R.; Keenan, T. P.; Sprengeler, P. A.;
Darke, P. L.; Emini, E. A.; Holloway, M. K.; Schleif, W. A. Design
and synthesis of peptidomimetic inhibitors of HIV-1 proteases
and renin. Evidence for improved transport. J . Med. Chem. 1994,
37, 215-8. (c) Szewczuk, Z.; Rebholz, K. L.; Rich, D. H. Synthesis
and biological activity of new conformationally restricted ana-
logues of pepstatin. Int. J . Pept. Protein Res. 1992, 40, 233-42.
6) (a) Wlodawer, A.; Erickson, J . W. Structure-based inhibitors of
HIV-1 protease. Annu. Rev. Biochem. 1993 62, 543-85. (b)
Darke, P. L.; Huff, J . R. HIV protease as an inhibitor target for
the treatment of AIDS. Adv. Pharmacol. 1994, 25, 399-454. (c)
West, M. L.; Fairlie, D. P. Targeting HIV-1 protease: a test of
drug-design methodologies. Trends Pharmacol. Sci. 1995, 16,
6
7-75. (d) Kempf, D. J .; Sham, H. L. HIV protease inhibitors.
Curr. Pharm. Des. 1996, 2, 225-46. (e) March, D. R.; Fairlie,
D. P. Designing new antiviral drugs for AIDS: HIV-1 Protease
and its inhibitors; R. G. Landes Co.: Austin, TX, 1996.
7) (a) Kim, J . L.; Morgenstern, K. A.; Lin, C.; Fox, T.; Dwyer, M.
D.; Landro, J . A.; Chambers, S. P.; Markland, W.; Lepre, C. A.;
O’Malley, E. T.; Harbeson, S. L.; Rice, C. M.; Murcko, M. A.;
Caron, P. R.; Thomson, J . A. Crystal structure of the hepatitis

Conformational Selection by Proteolytic Enzymes
J ournal of Medicinal Chemistry, 2000, Vol. 43, No. 7 1281
(
22) (a) Abbenante, G.; March, D.; Bergman, D.; Hunt, P. A.;
Garnham, B.; Dancer, R. J .; Martin, J . L.; Fairlie, D. P.
Regioselective Structural and Functional Mimicry Of Peptides:
Design Of Hydrolytically Stable Cyclic Peptidomimetic Inhibitors
Of Hiv 1 Protease. J . Am. Chem. Soc. 1995, 117, 10220-6. (b)
March, D.; Abbenante, G.; Bergman, D.; Brinkworth, R. I.;
Wickramasinghe, W.; Begun, J .; Martin, J . L.; Fairlie, D. P.
Substrate Based Cyclic Peptidomimetics Of Phe Ile Val That
Inhibit Hiv 1 Protease Using a Novel Enzyme Binding Mode. J .
Am. Chem. Soc. 1996, 118, 3375-9. (c) Reid, R. C.; March, D.
R.; Dooley, M. J .; Bergman, D. A.; Abbenante, G.; Fairlie, D. P.
A novel bicyclic enzyme inhibitor as a consensus peptidomimetic
for the receptor-bound conformations of 12 peptidic inhibitors
of HIV-1 protease. J . Am. Chem. Soc. 1996, 118, 8511-7.
23) Martin, J . L.; Begun, J .; Schindeler, A.; Wickramasinghe, W.
A.; Alewood, D.; Alewood, P. F.; Bergman, D. A.; Brinkworth,
R. I.; Abbenante, G.; March, D.; Reid, R. C.; Hunt, P. A.; Fairlie,
D. P. Molecular Recognition of Macrocyclic Peptidomimetic
inhibitors by HIV-1 Protease. Biochemistry 1999, 38, 7978-88.
24) Abbenante, G.; Bergman, D. A.; March, D. R.; Reid, R. C.; Hunt,
P. A.; J ames, I. W.; Dancer, R. J .; Garnham, B.; Stoermer, M.
L.; Fairlie, D. P. Structure-activity relationships for macrocyclic
peptidomimetic inhibitors of HIV-1 protease. Bioorg. Med. Chem.
Lett. 1996, 6, 2531-6.
(32) (a) Ksander, G. M.; De J esus, R.; Yuan, A.; Ghai, R. D.;
McMartin, C.; Bohacek, R. Meta-substituted benzofused mac-
rocyclic lactams as zinc metalloprotease inhibitors. J . Med.
Chem. 1997, 40, 506-14. (b) J anetka, J . W.; Raman, P.;
Satyshur, K.; Flentka, G.; Rich, D. H. Novel cyclic biphenyl ether
â-strand mimetics and HIVProtease Inhibitors. J . Am. Chem.
Soc. 1997, 119, 441-2. (c) Xue, C.-B.; He, X.; Roderick, J .;
DeGrado, W. F.; Cherney, R. J .; Hardman, K. D.; Nelson, D. J .;
Copeland, R. A.; J affe, B. D.; Decicco, C. P. Design and sunthesis
of cyclic inhibitors of matrix metalloproteinases and TNF-R
production. J . Med. Chem. 1998, 41, 1745-8. (d) Cherney, R.
J .; Wang, L.; Meyer, D. T.; Xue, C.-B.; Wasserman, R.; Hardman,
K. D.; Welch, P. K.; Covington, M. B.; Copeland, R. A.; Arner,
E. C.; DeGrado, W. F.; Decicco, C. P. Macrocyclic amino car-
boxylates as selective MMP-8 inhibitors. J . Med. Chem. 1998,
(
(
(
4
1, 1749-51.
(33) Ptitsyn, O. B.; Finkelstein, A. V. Theory of protein secondary
structure and algorithm of its prediction. Biopolymers 1983, 22,
15-25.
(34) (a) Poorman, R. A.; Tomasselli, A. G.; Heinrikson, R. L.; Kezdy,
F. J . A cumulative specificity model for proteases from human
immunodeficiency virus types 1 and 2, inferred from statistical
analysis of an extended substrate database. J . Biol. Chem. 1991,
25) Suguna, K.; Padlan, E. A.; Bott, R.; Boger, J .; Parris, K. D.;
Davies, D. R. Structures of complexes of rhizopuspepsin with
pepstatin and other statine-containing inhibitors. Proteins 1992,
2
66, 14554-61. (b) Tomasseli, A. G.; Heinrikson, R. L. Specificity
of retroviral proteases: an analysis of viral and nonviral protein
substrates. Methods Enzymol. 1994, 241, 279-301.
1
3, 195-205.
(
26) Lee, A. Y.; Smitka, T. A.; Bonjouklian, R.; Clardy, J . Atomic
structure of the trypsin-a90720a complex: a unified approach
to structure and function. Chem. Biol. 1994, 1, 113.
(35) (a) Burgen, A. S. V.; Roberts, G. C. K.; Feeney, J . Binding of
flexible molecules to macromolecules. Nature 1975, 253, 753-
(
27) Lee, A. Y.; Hagihara, M.; Karmacharya, R.; W.albers, M. W.;
Schreiber, S. L.; Clardy, J . Atomic structure of the Trypsin-
Cyclotheonamide a complex: lessons for the design of serine
protease inhibitors. J . Am. Chem. Soc. 1993, 115, 12619.
28) Browner, M. F.; Smith, W. W.; Castelhano, A. L. Matrilysin-
inhibitor complexes: common themes among metalloproteases.
Biochemistry 1995, 34, 6602-10.
5
. (b) J encks, W. P. Catal. Chem. Enzymol. 1987, 717-9.
(
36) Bergman, D. A.; Alewood, D.; Alewood, P. F.; Andrews, J . L.;
Brinkworth, R. I.; Englebretsen, D. R.; K. S. B. H. Kinetic-
properties of HIV-1 protease produced by total chemical syn-
thesis with cysteine residues replaced by isosteric l-alpha-amino-
N-butyric acid. Lett. Pept. Sci. 1995, 2, 99-107.
(
(
29) Khan, A. R.; Parrish, J . C.; Fraser, M. E.; Smith, W. W.; Bartlett,
P. A.; J ames, M. N. Lowering the entropic barrier for binding
conformationally flexible inhibitors to enzymes. Biochemistry
(37) Toth, M. V.; Marshall, G. R. A simple, continuous fluorometric
assay for HIV protease. Int. J . Pept. Protein Res. 1990, 36, 544-
50.
1
998, 37, 16839-45.
(
38) Chen, Y. H.; Yang, J . T.; Chau, K. H. Determination of the helix
and beta form of proteins in aqueous solution by circular
dichroism. Biochemistry 1974, 13, 3350-9.
(
30) Ding, J . H.; Fraser, M. E.; Meyer, J . H.; Bartlett, P. A.; J ames,
M. N. G. Macrocyclic inhibitors of penicillopepsin. 2. X-ray
crystallographic analyses of penicillopepsin complexed with a
P3-P1 macrocyclic peptidyl inhibitor and with its two acyclic
analogues. J . Am. Chem. Soc. 1998, 120, 4610-21.
(
39) Scanlon, M. J .; Fairlie, D. P.; Craik, D. J .; Englebretsen, D. R.;
West, M. L. NMR solution structure of the RNA-binding peptide
from human immunodeficiency virus (type 1) Rev. Biochemistry
1995, 34, 8242-9.
(
31) (a) Kabsch, W.; Sander, C. Dictionary of protein secondary
structure: pattern recognition of hydrogen-bonded and geo-
metrical features. Biopolymers 1983, 22, 2577-637. (b) Rich-
ardson, J . S. The anatomy and taxonomy of protein structure.
Adv. Protein Chem. 1981, 34, 167-339.
J M990315T