Identification of small-molecule inhibitors of the Aβ-ABAD interaction

Article abstract of DOI:10.1016/j.bmcl.2006.05.099

The interaction of amyloid beta peptide (Aβ) and Aβ-binding alcohol dehydrogenase (ABAD) was recently implicated in the pathogenesis of Alzheimer's disease (AD). Using an ELISA-based screening assay, we identified frentizole, an FDA-approved immunosuppressive drug, as a novel inhibitor of the Aβ-ABAD interaction. Analysis of the frentizole structure-activity relationship led to identification of a novel benzothiazole urea with a 30-fold improvement in potency.

Full text of DOI:10.1016/j.bmcl.2006.05.099

Bioorganic & Medicinal Chemistry Letters 16 (2006) 4657–4660
Identification of small-molecule inhibitors of the
Ab–ABAD interaction
Yuli Xie, Shixian Deng, Zhenzhang Chen, Shidu Yan and Donald W. Landry*
Department of Medicine, Columbia University, 630 West 168th Street, New York, NY 10032, USA
Received 14 April 2006; revised 30 May 2006; accepted 31 May 2006
Available online 14 June 2006
Abstract—The interaction of amyloid beta peptide (Ab) and Ab-binding alcohol dehydrogenase (ABAD) was recently implicated in
the pathogenesis of Alzheimer’s disease (AD). Using an ELISA-based screening assay, we identified frentizole, an FDA-approved
immunosuppressive drug, as a novel inhibitor of the Ab–ABAD interaction. Analysis of the frentizole structure–activity relationship
led to identification of a novel benzothiazole urea with a 30-fold improvement in potency.
ꢀ 2006 Elsevier Ltd. All rights reserved.
Alzheimer’s disease (AD) is the most common form of
senile dementia, affecting 15 million worldwide.1 AD is
characterized by progressive memory loss, decline in
language skills and other signs and symptoms of cogni-
tive impairment. Currently, no effective cure exists for
this chronic debilitating illness. Current therapeutics
modulate global acetylcholine-based neurotransmission
and clinical benefits are modest, transient, and sporadic.
A specific treatment for AD based on pathogenetic
mechanisms is needed. Although the etiology of AD is
not understood completely, amyloid beta-peptide (Ab),
the major component of senile plaques found in AD
brains, is implicated in the pathogenesis.2 The pathway
by which Ab causes neuronal dysfunction is the subject
of intense interest and evidence supports a role for a
mitochondrial alcohol dehydrogenase, alternately desig-
nated ERAD or HSD-10, found to bind Ab in the yeast
two-hybrid system and thus also known as amyloid-
binding alcohol dehydrogenase (ABAD).3 Recently, we
found that Ab can enter neuronal mitochondria to asso-
ciate with ABAD, and initiate a cascade culminating in
apoptosis. ABAD is upregulated in AD patients and our
mice overexpressing ABAD showed memory impair-
ment and an AD-like phenotype. These results suggest
that ABAD provides a direct molecular link from Ab
to neurotoxicity. Specific interruption of the binding
of Ab to ABAD suppressed Ab-induced neuronal
dysfunction, and thus the Ab–ABAD interaction could
be a useful target for drug development.4,5
We sought to test this hypothesis by identifying small
molecules that block the Ab–ABAD interaction. To this
end, we first developed a high-throughput screening as-
say for the binding of Ab to ABAD. In a standard sand-
wich ELISA,6 we found that biotin-labeled ABAD binds
to fixed Ab in a dose-dependent manner (Fig. 1). An
ABAD–biotin concentration of 2.5 lM was selected
for screening and, as a control, non-labeled ABAD
was found to compete with a maximum inhibition of
77.4% at 25 lM (Fig. 2, ABAD aggregation interferes
with the assay at higher concentration).
Figure 1. The ELISA-based assay of the binding of Ab–ABAD. Dose-
dependent binding of biotinated ABAD to Ab. Ninety-six-well plates
loaded with 5 lg of Ab were incubated with biotin-labeled ABAD at
varying concentrations with detection via ExtrAvidin-peroxidase
absorbance at 490 nm. The binding was corrected for background in
the absence of Ab. The assay was run in triplicate and plated as mean
ISD.
Keywords: Ab–ABAD interaction; Alzheimer’s disease; ELISA; Ben-
zothiazole urea.
*
Corresponding author. Tel.: +1 212 3051152; e-mail: yz2104@
columbia.edu
0960-894X/$ - see front matter ꢀ 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmcl.2006.05.099

4658
Y. Xie et al. / Bioorg. Med. Chem. Lett. 16 (2006) 4657–4660
cially available or prepared from the corresponding phe-
nylamines, were treated with acid chlorides in the
presence of excess triethylamine in THF at room tem-
perature to yield the corresponding amides (2). In paral-
lel, the 2-amino-benzothiazoles were reacted with a
series of imidazole carbamates or thiocarbamates, and
then with aromatic amines in DMF at 100 ꢁC to yield
the corresponding ureas (3, 5, and 6) and thioureas (4)
(Scheme 1).
A total of 45 compounds were synthesized,7 and their
capacity to inhibit Ab–ABAD binding was measured
by ELISA as described above, with IC50s calculated by
SigmaPlot from a five-point dose–effect curve run in
duplicate (Table 1). The preliminary structure–activity
relationship (SAR) study indicated that the urea moiety
was required for inhibitory activity. Amide compounds
(2a–2k) showed less than 50% inhibition at high concen-
tration (1 mM), while the ureas completely blocked
Ab–ABAD binding at 1 mM (data not shown). We
speculated that hydrogen bond donation at the urea
participated Ab or ABAD binding and consistent with
this hypothesis, the corresponding thioureas retained
potency. Replacement of the phenylurea ring with a
Figure 2. Inhibition of the binding with non-labeled ABAD. The
binding of biotinated ABAD (2.5 lM ) to fixed Ab was determined in
the presence of varying concentrations of free ABAD. The inhibition
(%) = [binding without free ABAD—binding with free ABAD]/binding
without free ABAD] · 100%.
With a convenient assay in hand, we initially screened a
focused in-house chemical library of 50 commercially
available compounds that (1) interact with Ab and/or
ABAD such as amyloid-binding dyes, AD imaging
agents, cyclodextrins, NAD+ (ABAD enzyme cofactor),
and nucleotides, or (2) are considered neuroprotective
(for the possibility that this neuroprotection represented
AB or ABAD binding activity). The initial screen iden-
tified three inhibitors: Congo red, thioflavine T, and res-
veratrol with inhibition at 100 lM of 100%, 21.0%, and
27.9%, respectively. The Ab binder Congo red was elim-
inated due to high toxicity and poor cell permeability,
and analogs of the neuroprotective agent resveratrol
had no activity in our assay. Thus, our attention turned
to thioflavine T, a nontoxic amyloid-binding dye. A
secondary screen of thioflavine T analogs identified
frentizole, a benzothiazole urea, as the most promising
hit (Fig. 3).
heterocyclic (Series
3
compounds) or polycyclic
structure (Series 6 compounds) resulted generally in low-
er inhibition. Substitutions on the benzothiazole and
phenylurea rings dramatically affected potency. Small
electron-withdrawing groups were preferred at the ben-
zothiazole ring with Cl and F particularly favored. Also,
compounds with a hydroxyl group at the para position
of the phenylurea were noticeably more potent.
Combining these features resulted in our two most po-
tent inhibitors, 5h and 5l, with IC50s of <10 lM each.
In summary, we have successfully identified a class of
benzothiazole ureas as micromolar inhibitors of the
Ab–ABAD interaction. The compounds 5h and 5l are
presently the most potent inhibitors discovered in this
study. As low-molecular-weight compounds, with
octanol–water partition coefficients (logP) of 1.34 and
1.15,8 respectively, they are likely to cross the blood–
brain barrier adequately. In the future, improved
potency will be sought through an expanded library of
benzothiazole ureas. If adequate CNS penetration can
be demonstrated, as a proof-of-concept, the inhibitors
will be tested in our AD animal models. Further studies
to characterize CNS penetration of 5h and 5l are in
progress.
Frentizole, a nontoxic antiviral and immunosuppressive
agent used clinically in rheumatoid arthritis and system-
ic lupus erythematosus, displayed a slightly improved
activity (IC50 = 200 lM) compared to thioflavine T
(IC50 = 230 lM). More importantly, frentizole allowed
facile construction of a library for SAR analysis and
we synthesized a series of analogs with variations at
the aromatic rings and their linking group. As shown
in Scheme 1, benzothiazole amines (1), either commer-
Figure 3. The chemical structures of the hits.

Y. Xie et al. / Bioorg. Med. Chem. Lett. 16 (2006) 4657–4660
4659
Scheme 1. Synthesis of frentizole analogs. Reagents and conditions: (a) i—NH4SCN, HCO2H/HAc (1:4), 0 ꢁC, 0.5 h; ii—Br2, 0 ꢁC, 5 h; (b)
substituted benzoyl chloride, THF, rt; (c) i—1,10-carbonyldiimidazole, CH3CN, rt; ii—amines, DMF, 100 ꢁC; (d) i—1,10-thiocarbonyldiimidazole,
CH3CN, rt; ii—amines, DMF, 100 ꢁC.
Acknowledgments
Table 1. Inhibition of the Ab–ABAD interaction determined by
ELISAa
These studies were supported by PO1AG17490,
P50AG08702, and the Alzheimer’s Association.
Compound
IC50 (lM)
3a
3b
3c
3d
4a
4b
5a
5b
5c
5d
5e
5f
256.7
243.8
247.3
282.5
190.5
15.6
References and notes
1. Pietrzik, C.; Behl, C. Int. J. Exp. Pathol. 2005, 86, 173.
2. Tanzi, R. E.; Bertram, L. Cell 2005, 120, 545.
3. Yan, S. D.; Fu, J.; Soto, C.; Chen, X.; Zhu, H.;
AlMohanna, F., et al. Nature 1997, 389, 689.
4. Lustbader, J. W.; Cirilli, M.; Lin, C.; Xu, H. W.; Takuma,
K.; Wang, N., et al. Science 2004, 304, 448.
5. Takuma, K.; Yao, J.; Huang, J.; Xu, H.; Chen, X.; Luddy,
J., et al. FASEB J. 2005, 19, 597.
6. ELISA protocol: 96-well plates were coated with
commercially available Ab1–42 (5 lg per well) and then
blocked with 2% of BSA (150 lL per well). After
washing, different amount of biotin-labeled recombinant
ABAD was added followed by incubation at 37 ꢁC for
2.5–3 h. The amount of ABAD bound to the plates was
detected using ExtrAvidin-Peroxidase (Sigma) according
to the protocol provided by the company. For mea-
suring the inhibitory activity, the testing agents were
added to the plates before addition of biotin-labeled
ABAD.
29.5
26.3
26.6
250.0
154.1
162.9
316.2
6.46
5g
5h
5i
23.0
267.3
22.7
5j
5k
5l
6.56
5m
5n
5o
5p
5q
5r
5s
5t
260.6
286.4
207.0
244.3
184.9
47.6
7. Analytical data for representative compounds: 2d, 1H
NMR (CDCl3) ppm 3.91 (s, 3H, OCH3); 3.93 (s, 3H,
OCH3); 4.10 (s, 3H, OCH3); 7.0–7.31 (m, 2H, Ph-H); 7.80
(dd, 1H, Ph-H); ESI-MS (M++1): 360.4. Compound 3d,
1H NMR (DMSO-d) ppm: 7.15–7.30 (d, 1H, Ph-H);
7.31–7.40 (m, 1H, Py-H); 7.58–7.69 (m, 1H, Py-H); 7.75–
7.85 (m, 1H, Ph-H); 7.95–8.00 (d, 1H, Ph-H); 8.20–8.30
(d, 1H, Py-H); 8.53 (s, 1H, Py-H); ESI-MS (M++1):
289.0. Compound 5h, 1H NMR (DMSO-d) ppm: 3.85 (s,
3H, –COOCH3); 6.90–6.98 (d, 1H, Ph-H); 7.35–7.42 (d,
1H, Ph-H); 7.51–7.62 (m, 2H, Ph-H); 8.03 (br, 2H, Ph-
H); 9.10 (s, 1H, OH); ESI-MS (M++1): 377.9. Compound
6a, 1H NMR (DMSO-d) ppm: 7.10 (s, 1H, Naph-H);
53.8
20.6
6a
6b
6c
6d
6e
6f
227.0
280.5
212.3
18.2
27.9
29.0
6g
6h
267.3
36.3
a IC50 is defined as concentration of compounds to obtain 50% of
maximum inhibition (100%).

4660
Y. Xie et al. / Bioorg. Med. Chem. Lett. 16 (2006) 4657–4660
7.35–7.40 (d, 1H, Naph-H); 7.48 (s, 1H, Naph-H); 7.55–
8. The octanol–water partition coefficient logP of 5h and 5l
was experimentally determined as described in the litera-
ture: Rothwell, J. A.; Day, A. J.; Morgan, M. R. J. Agric.
Food Chem. 2005, 53, 4355.
7.62 (d, 1H, Ph-H); 7.70–7.79 (d, 1H, Naph-H); 7.95 (s,
1H, Naph-H); 8.01–8.05 (d, 1H, Ph-H); 8.06 (s, 1H, Ph-
H); 9.35 (s, 1H, OH); ESI-MS (Mꢀꢀ1): 498.0.