A substrate-based difluoro ketone selectively inhibits Alzheimer's γ-secretase activity

Full text of DOI:10.1021/jm970621b

6
J . Med. Chem. 1998, 41, 6-9
A Su bstr a te-Ba sed Diflu or o Keton e
Selectively In h ibits Alzh eim er ’s
γ-Secr eta se Activity
Michael S. Wolfe,*,† Martin Citron,‡,§
Thekla S. Diehl,‡ Weiming Xia,‡
Isaac O. Donkor,† and Dennis J . Selkoe‡
Department of Pharmaceutical Sciences, University of
Tennessee, Memphis, Tennessee 38163, and Center for
Neurologic Diseases, Harvard Medical School, Brigham and
Women’s Hospital, Boston, Massachusetts 02115
F igu r e 1. Sites of APP processing by R-, â-, and γ-secretases
and the resulting proteolytic fragments. Heterogeneity of Aâ
and p3 at the C-terminus is due to differential processing of
the 12 and 10 kDa C-terminal fragments (CTFs) by γ-secre-
tases.
Received September 15, 1997
Amyloid plaques, invariantly associated with Alz-
heimer’s disease (AD), contain amyloid â-protein (Aâ)
as the primary protein component.1 The 39-43 amino
acid Aâ is derived from the â-amyloid precursor protein
(APP), an integral membrane protein of unknown func-
tion.2 Recent findings particularly implicate the more
hydrophobic and highly insoluble 42 amino acid variant
(Aâ42) in amyloid plaque formation and in the patho-
genesis of AD.3-5 Although Aâ42 accounts for only about
10% of total Aâ secreted from cells (roughly 90% is the
40 amino acid variant Aâ40), Aâ42 is the major protein
component of the diffuse, largely nonfibrillar plaques
which precede the development of the dense, fibrillar
neuritic plaques characteristic of AD.6,7 Genetic evi-
dence strongly implicates Aâ in general and Aâ42 in par-
ticular in the etiology of AD: all mutations linked to
familial early-onset AD (FAD) examined to date result
in increased production of Aâ42 or of both Aâ40 and
Aâ42.3-5
Aâ is formed from APP through two protease activi-
ties.8 First, â-secretase cleaves APP at the Aâ N-
terminus, resulting in a soluble, secreted APP derivative
(â-APPs) and a 12 kDa membrane-retained C-terminal
fragment (CTF) (Figure 1). The latter is further pro-
cessed to Aâ by γ-secretases, which cleave within the
single transmembrane region. Other APP molecules
can be cleaved by R-secretase within the Aâ region, thus
precluding Aâ formation. This R-secretase cleavage,
which occurs more commonly than â-secretase cleavage,
also leads to release of a soluble derivative of APP (R-
APPs) and retention of a 10 kDa CTF. The CTF
resulting from R-secretase cleavage can also serve as a
substrate for the γ-secretases to yield a 3 kDa protein
(p3), an N-terminally truncated version of Aâ. As yet,
none of these APP-processing secretases have been iso-
lated, nor have they been assigned to a specific protease
class (although R-secretase is likely a metalloprotease9).
Recent studies suggest that different protease activi-
ties, dubbed γ(40)- and γ(42)-secretases, are responsible
for the C-terminal heterogeneity of Aâ.10,11 Four peptide
aldehydes, all originally designed to inhibit the cysteine
protease calpain, are the only γ-secretase inhibitors
reported to date,10-12 and no compounds selective for
inhibiting Aâ42 production more than that of Aâ40 have
been described. At the time of this writing, there have
been no reports describing the design of γ-secretase
inhibitors. Thus, while γ-secretases are considered
important new targets for inhibitor design,13,14 very
little information on the structural requirements for
inhibiting these enzymes is available. In this report,
we describe a substrate-based difluoro ketone that
inhibits γ-secretase activity in APP-transfected cells.
The design of γ-secretase inhibitors at this time is
challenging, because the proteases are unidentified.
While it has been suggested that γ-secretases are
aspartyl proteases,15-17 the fact that the reported
γ-secretase inhibitors are all calpain inhibitors suggests
that the γ-secretases may be cysteine proteases. We
initially designed peptidomimetic 1 (Scheme 1) based
on the APP γ(42)-secretase cleavage site (Figure 1). This
compound contains an (S)-4-amino-3-oxo-2,2-difluoro-
pentanoyl moiety as a stable pseudodipeptide replace-
ment for the scissile Ala-Thr bond. Although target
compound 1 does not contain the P1′ side chain, related
difluoro ketone peptide analogues without P1′ side
chains potently inhibit aspartyl proteases (renin18 and
HIV protease19) and serine proteases (chymotrypsin20
and porcine pancreatic elastase21). The difluoro ketone
moiety is readily hydrated in aqueous media, and this
hydrate is thought to mimic the tetrahedral intermedi-
ate formed in aspartyl protease catalysis. For serine
proteases, the active site serine attacks the keto form
of the analogue, resulting in a stable, covalent enzyme-
inhibitor complex. A similar mechanism can be envi-
sioned for cysteine protease inhibition, although there
are no reports confirming this. There is at least one
report, however, of fluoromethyl ketone peptide ana-
logues that potently inhibit a cysteine protease.22 Hence,
without clear knowledge of the class of protease into
which γ-secretases fall, difluoro ketone-type pseudo-
dipeptides incorporated into the context of the appropri-
ate APP sequence would appear to have a reasonable
chance of inhibiting these enzymes. Peptidomimetics
of this type would be expected to have important
advantages over the few previously reported γ-secretase
inhibitors: they are specifically designed toward γ-secre-
tases and not toward calpain, and they allow exploration
* To whom correspondence should be addressed. Tel: (901) 448-
7533. Fax: (901) 448-6828. E-mail: mwolfe@utmem1.utmem.edu.
† University of Tennessee.
‡ Harvard Medical School.
§ Present address: Amgen, Inc., Thousand Oaks, California 91320-
1789.
S0022-2623(97)00621-3 CCC: $15.00 © 1998 American Chemical Society
Published on Web 01/01/1998

Communications to the Editor
J ournal of Medicinal Chemistry, 1998, Vol. 41, No. 1
7
Sch em e 1a
a
(a) Boc2O, KHCO3; (b) DMSO, oxalyl chloride, Et3N; (c) BrCF2CO2Et, Zn, reflux; (d) 1 equiv of NaOH, then H+; (e) H2N-Val-Ile-OMe,
DIPC, HOBT; (f) TFA; (g) Boc-Val-Ile-OH, DIPC, HOBT; (h) Dess-Martin oxidation.
of binding sites downstream from the cleavage site (i.e.,
P2′, P3′, etc.).
were retained in the design of 1 to enhance permeability
and localization to membranes in the cell-based assay
employed.
Peptide analogue 1 was designed based on the Ala713
-
1
Thr714 APP cleavage site leading to Aâ42 (i.e., it was
specifically designed as a γ(42)-secretase inhibitor).
Flanking residues were included to increase binding
potency as well as specificity. The synthetic procedure
for realizing this compound is shown in Scheme 1.
Commercially available (S)-2-amino-1-propanol (L-alani-
nol, 2) was protected as a tert-butoxycarbamate (Boc)
and oxidized under Swern conditions to produce alde-
hyde 3 in 75% yield for the two steps. Although many
amino aldehydes are easily racemized or otherwise
unstable, aldehyde 3 was remarkably stable and ob-
tained in enantiopure form (>95% ee) via recrystalli-
zation from hexane and ether.23 This aldehyde was then
converted to difluoro alcohol 4 in 60% yield via a
Reformatsky reaction with ethyl bromodifluoroacetate
under refluxing conditions. A similar transformation
has been reported with L-leucinal in which the more
thermodynamically stable 3R,4S diastereomer was
formed,18 presumably via chelation control. When this
transformation is carried out with L-alaninal, the smaller
methyl substituent does not provide high stereoselec-
tivity, and the difluoro alcohol 4 is formed as a 4:1
mixture of diastereomers. The major diastereomer was
initially assigned as 3R,4S by analogy with the result
from L-leucinal and proven correct through crystal-
lographic analysis of a later synthetic intermediate (vide
infra).
Compound 4 was hydrolyzed with a slight excess of
hydroxide and then coupled to H2N-Val-Ile-OMe using
diisopropylcarbodiimide (DIPC) and 1-hydroxybenzo-
triazole (HOBT) to form 5 in 66% yield after chroma-
tography. Peptide analogue 5 was then isolated as a
single diastereomer after recrystallization from ether/
pentane, the stereochemical assignment confirmed via
X-ray crystallography. Compound 5 was N-deprotected
with trifluoroacetic acid and coupled to BocHN-Val-Ile-
OH to provide the penultimate product 6 in 58% yield.
Alcohol 6 was oxidized with Dess-Martin periodinane
reagent24,25 in virtually quantitative yield to afford final
compound 1. The N- and C-terminal protecting groups
The H NMR spectra of difluoro ketone 1 illustrated
the tendency of this compound to form a hydrate:
addition of D2O to a DMSO-d6 solution of 1 resulted in
dramatic decreases of the peaks corresponding to the
ketone and increases in the peaks corresponding to the
hydrate. In most cases, overlap between ketone and
hydrate peaks was considerable, but the methyl peak
of the pseudopeptide portion of the molecule proved di-
agnostic: this peak is a doublet (J ) 7.0 Hz) at 1.25
ppm in the ketone form of 1 and a doublet (J ) 6.8 Hz)
at 1.08 ppm in the hydrate form.
Compound 1, alcohol precursors 5 and 6, as well as
the oxidized counterpart of 5 (difluoro ketone 7) were
tested for their ability to selectively inhibit the produc-
tion of Aâ40 and Aâ42 according to the protocol described
fully in ref 10. Human embryonic kidney 293 cells were
stably transfected with a plasmid construct carrying a
double mutant form of human APP. This double muta-
tion, K595N/M596L, occurs immediately prior to the
â-secretase cleavage site and is found in certain Swedish
families that are prone to FAD.8 The double mutation
26
leads to increased levels of both Aâ40 and Aâ42 and
thus allows for easier detection of these APP metabo-
lites. The transfected cells were pulse-labeled with [35S]-
methionine for 2 h and then chased for 2 h in media
with or without compound.10 Secreted APP cleavage
products in the conditioned media were then immuno-
precipitated with specific antibodies, and the immuno-
precipitates were analyzed by PAGE and fluorography.
At 200 µM, compounds 5, 6, and 7 did not decrease
either Aâ or p3 production. In contrast, 1 completely
blocked Aâ40 and Aâ42 as well as p340 and p342 formation
at this concentration (Figure 2A). At lower concentra-
tions (25-100 µM), 1 was surprisingly more effective
against Aâ40 than Aâ42 production (Figure 2B). The 10
and 12 kDa C-terminal APP fragments (i.e., γ-secretase
substrates) were dramatically increased in the presence
of 1 (Figure 3A). However, this compound did not
decrease production of R-APPs or â-APPs (Figure 3B),
indicating that R- and â-secretase activities were not

8
J ournal of Medicinal Chemistry, 1998, Vol. 41, No. 1
Communications to the Editor
F igu r e 2. Fluorographic SDS-PAGE analysis of Aâ secretion from transfected cells. Panel A: Effect of 1 and 5-7 on Aâ and p3
production. Immunoprecipitation of media performed with antibody 21F12 (specific for Aâ42 and p342) or with 2G3 (specific for
Aâ40 and p340) either in the absence (lane C) or presence of 200 µM compound. Panel B: Effect of 1 on Aâ and p3 production at
lower concentrations. Antibodies and their specificities are described in ref 10 and citations therein.
F igu r e 3. Fluorographic SDS-PAGE analysis of APP R- and â-secretase cleavage products from transfected cells. Panel A:
Effect of 1 on APP 10 and 12 kDa C-terminal fragments. Immunoprecipitation performed with antibody C7 (specific for full-
length APP and its C-terminal fragments) (upper panel) or with antibody 1282 (specific for the 12 kD C-terminal fragments)
(lower panel). Panel B: Effect of 1 on APPs. Immunoprecipitation performed with antibody 1736 (specific for R-APPs) or with
192sw (specific for â-APPs). Antibodies and their specificities are described in ref 10 and citations therein.
inhibited. Thus, this difluoro ketone peptidomimetic
appears to selectively inhibit Aâ synthesis at the
γ-secretase level. Furthermore, the differential effect
of 1 on Aâ40 versus Aâ42 further supports the hypothesis
that these Aâ variants result from different proteases.
That the ketone functionality is critical for this activity
is indicated by the lack of activity of the alcohol
counterpart 6 (Figure 2A). Also, at least part of the
N-terminal dipeptide is essential, because the truncated
difluoro ketone 7 is likewise devoid of activity. Analysis
of the effects of 1 on Aâ production 2 h after its removal
from the medium indicated that this compound is a
reversible inhibitor of γ-secretase activity: Aâ and p3
levels were higher than control, apparently due to the
buildup of the precursor 10 kDa and 12 kDa C-terminal
APP fragments prior to compound removal (data not
shown).
was captured by monoclonal antibody 266 (specific for
Aâ peptide residues 13-28) and detected via Aâ-specific
biotinylated monoclonal antibody 3D6 (recognizing Aâ
residues 1-5) followed by reaction with an alkaline
phosphatase-avidin conjugate and incubation with
fluorogenic substrate, 4-methylumbelliferyl phosphate.
Using this assay, the IC50 of 1 for total Aâ inhibition
was 13 ( 5 µM (average of three independent experi-
ments). The potency of 1 is thus comparable to the best
reported peptide aldehyde (IC50 ∼5 µM).11 Because the
reported γ-secretase inhibitors were all originally de-
signed as calpain inhibitors, we tested 1 as an inhibitor
of this cysteine protease. Difluoro ketone 1 was quite
ineffective against calpain II (Sigma) in a purified
enzyme assay,28 with an IC50 of about 100 µM. By this
criterion, 1 is a selective γ-secretase inhibitor in com-
parison with the reported peptide aldehyde inhibitors.
Compound 1 was further tested for inhibition of Aâ
production in APP-transfected Chinese hamster ovary
cells using a rapid and quantitative double antibody
(“sandwich”) ELISA.27 Total Aâ secreted from media
The γ-secretases, which catalyze the final step in Aâ
biosynthesis from APP, are thought to be important new
therapeutic targets.13,14 Only four compounds have
been previously reported as inhibitors of these pro-

Communications to the Editor
J ournal of Medicinal Chemistry, 1998, Vol. 41, No. 1
9
(11) Klafki, H.-W.; Abramowski, D.; Swoboda, R.; Paganetti, P. A.;
Staufenfiel, M. The carboxyl termini of â-amyloid peptides 1-40
and 1-42 are generated by distinct γ-secretase activities. J . Biol.
Chem. 1996, 271, 28655-28659.
(12) Higaki, J .; Quon, D.; Zhong, Z.; Cordell, B. Inhibition of â-amyl-
oid formation identifies proteolytic precursors and subcellular
site of catabolism. Neuron 1995, 14, 651-659.
(13) Schenk, D. B.; Rydel, R. E.; May, P.; Little, S.; Panetta, J .;
Lieberburg, I.; Sinha, S. Therapeutic approaches related to
amyloid-â peptide and Alzheimer’s disease. J . Med. Chem. 1995,
38, 4141-4154.
(14) Brennan, M. Bringing back memories: Expanding understand-
ing of Alzheimer’s disease drives development of new drugs.
Chem. Eng. News 1997, J an. 20, 29-35.
(15) Evin, G.; Cappai, R.; Li, Q.-X.; Culvenor, J . G.; Small, D. H.;
Beyreuther, K.; Masters, C. L. Candidate γ-secretases in the
generation of the carboxyl terminus of the Alzheimer’s disease
âA4 amyloid: Possible involvement of cathepsin D. Biochemistry
1995, 34, 14185-14192.
(16) Ladror, U. S.; Snyder, S. W.; Wang, G. T.; Holzman, T. F.; Krafft,
G. A. Cleavage at the amino and carboxyl termini of Alzheimer’s
amyloid-â by cathepsin D. J . Biol. Chem. 1994, 269, 18422-
18428.
(17) Saftig, P.; Peters, C.; von Figura, K.; Craessaerts, K.; Van
Leuven, F.; De Strooper, B. Amyloidogenic processing of human
amyloid precursor protein in hippocampal neurons devoid of
cathepsin D. J . Biol. Chem. 1996, 271, 27241-27244.
(18) Thaisrivongs, S.; Pals, D. T.; Kati, W. M.; Turner, S. R.;
Thomasco, L. M.; Watt, W. Design and synthesis of potent and
specific renin inhibitors containing difluorostatine, difluorosta-
tone, and related analogues. J . Med. Chem. 1986, 29, 2080-
2087.
(19) Dreyer, G. B.; Metcalf, B. W.; Tomaszek, T. A., J r.; Carr, T. J .;
Chandler, A. C., III; Hyland, L.; Fakhoury, S. A.; Magaard, V.
W.; Moore, M. L.; Strickler, J . E.; Debouck, C.; Meek, T. D.
Inhibition of human immunodefiency virus 1 protease in vitro:
Rational design of substrate analogue inhibitors. Proc. Natl.
Acad. Sci. U.S.A. 1989, 86, 9752-9756.
(20) Parisi, M. F.; Abeles, R. H. Inhibition of chymotrypsin by
fluorinated R-keto acid derivatives. Biochemistry 1992, 31,
9429-9435.
(21) Takahashi, L. H.; Radhakrishnan, R.; Rosenfield, R. E., J r.;
Meyer, E. F., J r.; Trainor, D. A. Crystal structure of the covalent
complex formed by a peptidyl R,R-difluoro-â-keto amide with
porcine pancreatic elastase at 1.78-Å resolution. J . Am. Chem.
Soc. 1989, 111, 3368-3374.
(22) Angliker, H.; Anagli, J .; Shaw, E. Inactivation of calpain by
peptidyl fluoromethyl ketones. J . Med. Chem. 1992, 35, 216-
220.
(23) Hamada, Y.; Shioiri, T. New methods and reagents in organic
synthesis. 29. A practical method for the preparation of optically
active N-protected R-amino aldehydes and peptide aldehydes.
Chem. Pharm. Bull. 1982, 30, 1921-1924.
(24) Ireland, R. E.; Liu, L. An improved procedure for the preparation
of the Dess-Martin periodinane. J . Org. Chem. 1993, 58, 2899.
(25) Linderman, R. J .; Graves, D. M. An efficient procedure for the
oxidation of fluorinated carbinols. Tetrahedron Lett. 1987, 28,
4259-4262.
(26) Citron, M.; Oltersdorf, T.; Haass, C.; McConlogue, K.; Hung, A.
Y.; Suebert, P.; Vigo- Pelfrey, C.; Lieberburg, I.; Selkoe, D. J .
Mutation of the â-amyloid precursor protein in familial Alzhe-
imer’s disease increases â-protein production. Nature 1992, 360,
672-674.
teases: all are peptide aldehydes, and all are calpain
inhibitors.10-12 Identifying other types of inhibitors is
critical for creating molecular probes for studying these
proteases and for generating drug leads. We have
demonstrated that a substrate-based difluoro ketone is
a specific inhibitor of Aâ biosynthesis in APP-transfected
cells. While this inhibition occurs at the γ-secretase
level, it is unclear whether this compound and the
reported peptide aldehydes interact directly with γ-secre-
tases. Clarification of this issue will have to await the
development of a purified enzyme assay. These results
again support the existence of pharmacologically dis-
tinct γ-secretases cleaving at amino acids 40 and 42;
thus far, selective inhibitors of γ(40)-secretase activity
have been easier to identify. Currently, we are modify-
ing prototype 1 in various ways to determine structural
requirements for activity, to reverse the selectivity in
favor of Aâ42 reduction, and to create new molecular
tools for AD research.
Ack n ow led gm en t. We thank Drs. Steven W. White
and Charles R. Ross, II, at St. J ude Childrens’ Research
Hospital for the crystallographic analysis of 5. This work
was supported by a Faculty Development Grant from
the University of Tennessee College of Pharmacy
(M.S.W.) and by NIH Grant AG 12749 (D.J .S.).
Su p p or tin g In for m a tion Ava ila ble: Synthetic proce-
dures and characterization for all compounds as well as
crystallographic data on 5 (11 pages). Ordering information
is given on any current masthead page.
Refer en ces
(1) Glenner, G. G.; Wong, C. W. Alzheimer’s disease: Initial report
of the purification and characterization of a novel cerebrovas-
cular amyloid protein. Biochem. Biophys. Res. Commun. 1984,
120, 885-890.
(2) Kang, J .; Lemaire, H.; Unterbeck, A.; Salbaum, J . M.; Masters,
C. L.; Grzeschik, K. H.; Multhaup, G.; Beyreuther, K.; Muller-
Hill, B. The precursor of Alzheimer’s disease amyloid A4 protein
resembles a cell-surface receptor. Nature 1987, 325, 733-736.
(3) Hardy, J . The Alzheimer family of diseases: Many etiologies,
one pathogenesis? Proc. Natl. Acad. Sci. U.S.A. 1997, 94, 2095-
2097.
(4) Selkoe, D. J . Alzheimer’s disease: Genotypes, phenotype, and
treatments. Science 1997, 275, 630-631.
(5) Lamb, B. T. Presenilins, amyloid-â and Alzheimer’s disease.
Nature Med. 1997, 3, 28-29.
(6) Roher, A. E.; Lowenson, J . D.; Clarke, S.; Wolkow, C.; Wang,
R.; Cotter, R. J .; Reardon, I.; Zurcher-Neely, H. A.; Heinrikson,
R. L.; Ball, M. J .; Greenberg, B. D. Structural alterations in the
peptide backbone of â-amyloid core protein may account for its
deposition and stability in Alzheimer’s disease. J . Biol. Chem.
1993, 268, 3072-3083.
(7) Iwatsubo, T.; Odaka, A.; Suzuki, N.; Mizusawa, H.; Nukina, N.;
Ihara, Y. Visualization of Aâ42(43) and Aâ40 in senile plaques
with end-specific monoclonals: Evidence that an initially de-
posited species is Aâ42(43). Neuron 1994, 13, 45-52.
(8) Selkoe, D. J . Cell biology of the amyloid â-protein precursor and
the mechanism of Alzheimer’s disease. Annu. Rev. Cell Biol.
1994, 10, 373-403.
(9) Roberts, S. B.; Ripellino, J . A.; Ingalis, K. M.; Robakis, N. K.;
Felsenstein, K. M. Non-amyloidogenic cleavage of the â-amyloid
precursor protein by an integral membrane metalloendopepti-
dase. J . Biol. Chem. 1994, 269, 3111-3116.
(10) Citron, M.; Diehl, T.; Gordon, G.; Biere, A. L.; Seubert, P.; Selkoe,
D. J . Evidence that the 42- and 40-amino acid forms of amyloid
â protein are generated from the â-amyloid precursor protein
by different protease activities. Proc. Natl. Acad. Sci. U.S.A.
1996, 93, 13170-13175.
(27) J ohnson-Wood, K.; Lee, M.; Motter, R.; Hu, K.; Gordon, G.;
Barbour, R.; Khan, K.; Gordon, M.; Tan, H.; Games, D.; Lieber-
burg, I.; Schenk, D.; Seubert, P.; McConlogue, L. Amyloid
42 deposition in a transgenic
mouse model Alzheimer’s disease. Proc. Natl. Acad. Sci. U.S.A.
1997, 94, 1550-1555.
precursor protein processing and Aâ
(28) Buroker-Kilgore, M.; Wang, K. K. W. A Coomassie Brilliant Blue
G-250-based colorometric assay for measuring activity of calpain
and other proteases. Anal. Biochem. 1993, 208, 387-392.
(29) Hamada, Y.; Shibata, M.; Sugiura, T.; Kato, S.; Shioiri, T. New
methods and reagents in organic synthesis. 67. A general
synthesis of derivatives of optically pure 2-(1-aminoalkyl)-
thiazole-4-carboxylic acids. J . Org. Chem. 1987, 52, 1252-1255.
J M970621B