A catch-and-release strategy for the combinatorial synthesis of 4-acylamino-1,3-thiazoles as potential CDK5 inhibitors

Article abstract of DOI:10.1016/S0960-894X(03)00726-1

Two-dimensional libraries of 4-acylamino-1,3-thiazoles 9 were prepared via Curtius rearrangement of 1,3-thiazole-4-carbonyl azides 6, trapping of the intermediate isocyanates with oxime resin, and thermal regeneration of the isocyanates from the washed resin in the presence of nucleophiles. Several compounds proved to be selective inhibitors of CDK5/p25 versus the closely homologous CDK2/cyclin A enzyme, with the best analogue (43) possessing over 100-fold selectivity.

Full text of DOI:10.1016/S0960-894X(03)00726-1

Bioorganic & Medicinal Chemistry Letters 13 (2003) 3491–3495
A Catch-and-Release Strategy for the Combinatorial Synthesis of
4-Acylamino-1,3-thiazoles as Potential CDK5 Inhibitors
Scott D. Larsen,a,* Carl F. Stachew,a Paula M. Clare,b Jerry W. Cubbagec
and Karen L. Leachb
aMedicinal Chemistry Research, Pharmacia Corporation, 333 Portage St., Kalamazoo, MI 49007, USA
bCellular & Molecular Biochemistry, Pharmacia Corporation, 333 Portage St., Kalamazoo, MI 49007, USA
cStructural & Computational Chemistry, Pharmacia Corporation, 333 Portage St., Kalamazoo, MI 49007, USA
Received 24 April 2003; revised 3 July 2003; accepted 12 July 2003
Abstract—Two-dimensional libraries of 4-acylamino-1,3-thiazoles 9 were prepared via Curtius rearrangement of 1,3-thiazole-4-
carbonyl azides 6, trapping of the intermediate isocyanates with oxime resin, and thermal regeneration of the isocyanates from the
washed resin in the presence of nucleophiles. Several compounds proved to be selective inhibitors of CDK5/p25 versus the closely
homologous CDK2/cyclin A enzyme, with the best analogue (43) possessing over 100-fold selectivity.
# 2003 Elsevier Ltd. All rights reserved.
One of the pathological hallmarks of Alzheimer’s Dis-
ease is cytoplasmic filamentous material known as neu-
rofibrillary tangles (NFT) that accumulates in the
neuronal cell soma.1,2 The main component of NFT is
an aggregate of the microtubule-stabilizing protein tau
in a hyperphosphorylated state. NFT are associated
with microtubule destabilization, synapse dysfunction
and dementia. Cyclin-dependent kinase 5 (CDK5) is a
serine/threonine kinase that is required for normal neu-
ronal development.3 Association of CDK5 with its reg-
ulatory subunit p35 is necessary for enzyme activation,
and this complex phosphorylates a variety of substrates.
CDK5/p35 (also known at Tau Protein Kinase II or
TPKII) is believed to be one of the major kinases that
phosphorylates tau in AD brains.4 CDK5 is con-
stitutively activated in brain tissue of Alzheimer’s
patients due to the displacement of the normal reg-
ulatory subunit p35 by a truncated form p25.5 In addi-
tion, the enzyme complex has been shown to co-localize
with NFT, suggesting that inhibitors of CDK5 may have
therapeutic efficacy for the treatment and/or prevention
of Alzheimer’s disease.6
degree of specificity it exhibited for CDK5 versus the
closely homologous cell cycle regulatory kinase complex
CDK2/cyclin A (Ki=10–20 mM). For obvious reasons,
specific inhibition of a particular target kinase relative
to the plethora of other crucial regulatory kinases is of
paramount importance.
Preliminary medicinal chemistry work (not shown) on 4
quickly established that simple urea or hydrazide ana-
logues were not active, indicating that the aminourea
moiety was important. The 1,3-thiazole nucleus of 4
suggested a straightforward synthesis of two-dimen-
sional combinatorial libraries (varying the 2-substituent
and the 4-aminourea) commencing with the known
preparation of thiazole-4-carboxylates from thioamides
(Scheme 1, steps a,b).7 This route was designed to avoid
the intermediacy of 4-amino-1,3-thiazoles, which have
been reported8a to be unstable, by directly generating
the 4-amino substituent in an acylated form (isocyanate)
ready for nucleophile introduction.8b Although 4 could
be successfully prepared from thioamide 1 by this route,
the final product was contaminated by numerous impu-
rities, from which the desired product had to be chro-
matographically isolated in poor yield. The major
byproducts were the symmetrical urea derived from the
aminothiazole and compounds resulting from insertion
of the intermediate nitrene into solvent.
Broad screening of the Pharmacia compound collection
for inhibitors of TPK II led to the identification of 4
(Ki=0.5–2.0 mM). Of particular interest was the modest
To facilitate the production of libraries free of these
impurities, we developed a catch-and-release approach,9
*Corresponding author. Tel.: +1-269-833-7713; fax: +1-269-833-
2516; e-mail: scott.d.larson@pfizer.com
0960-894X/$ - see front matter # 2003 Elsevier Ltd. All rights reserved.
doi:10.1016/S0960-894X(03)00726-1

3492
S. D. Larsen et al. / Bioorg. Med. Chem. Lett. 13 (2003) 3491–3495
NaOH, Bu4NHSO4; (ii) Lawesson’s reagent, toluene,
65 ꢀC). A combination of 22 thioamides, 56 hydrazines
and nine alkoxyamines were employed in the course of
preparing several two-dimensional libraries (50–100
compounds each- not all possible combinations were
prepared), from which products could be obtained in
satisfactory purity (>80% by HPLC/MS) from 20
thioamides, 41 hydrazines and all nine alkxoyamines.
Yields for the various libraries averaged about 70%
from the precursor acyl azides. Representative examples
are included in Table 1.
Scheme 1. (a) BrCH2COCO2Et, 100 ꢀC (30%); (b) aq KOH, THF
(93%); (c) oxalyl chloride, CHCl3; NaN3, acetone (66%); (d) toluene,
80 ꢀC; 4-aminomorpholine (34%).
All new compounds were assayed for their ability to
inhibit CDK5/p25 (TPK II). Selectivity was determined
by evaluation in an assay for inhibition of CDK2/cyclin A.
Results for selected analogues are compiled in Table 1.12
that is outlined in Scheme 2. Thioamides 5 were con-
verted to acyl azides 6 via a protocol similar to that
described in Scheme 1 (76% avg yield). Addition of
oxime resin [polystyrene-C(4–NO2–Ph)=NOH] to the
intermediate isocyanates 7, generated from thermolysis
of azides 6, afforded resins 8, which could be rinsed free
of impurities by repeated washing with CH2Cl2 and
MeOH. The presence of the resin-bound carbamate was
confirmed by FTIR (1751–1759 cmÀ1).10 Heating the
washed resins 8 in a mixture of the desired nucleophile
(NuH), triethylamine and either toluene or 1,2-dichloro-
ethane afforded the desired acylamino thiazoles 9. The
addition of triethylamine base was found to greatly facil-
itate the elimination of the isocyanate from the oxime
resin, giving improved yields and purities. Purification of
the crude products was accomplished by treating the
reaction mixtures with isocyanate resin (to remove
excess NuH) prior to filtration, followed by SLE to
remove excess triethylamine. To illustrate the super-
iority of this catch and release protocol to the solution
phase route described in Scheme 1, lead compound 4
could be reproducibly obtained from the corresponding
azide 3 in nearly quantitative yield with a purity of 93–
94% using the method in Scheme 2.11
The first SAR facet to emerge from this work was the
absolute requirement that the hydrazines be 1,1-di-
substituted. This is strikingly apparent when comparing
the compounds 15 and 16 as well as the compounds 17,
19 and 20. In each closely related set, only the terminal
disubstituted isomers are active. In fact, of the numer-
ous monosubstituted and 1,2-disubstituted hydrazine
diversity elements incorporated into these libraries, not
a single one afforded an active analogue (most data not
shown in Table 1). Several 1,1-dialkylhydrazines and
related cyclic analogues possessed significant activity,
although none were superior to the lead 4. Evidently
some steric constraints are operative, as substituting the
pyrrolidine ring of the active analogue 10 abolished
activity (11), and the dibenzyl analogue 14 was virtually
inactive.
Modification of the thiazole moiety was obviously more
labor intensive than altering the hydrazine component,
thereby limiting diversity input. Nevertheless, 20 thio-
amides were successfully incorporated into thiazole
analogues, 16 of which are represented in Table 1. The
2,6-dichlorobenzyl moiety unfortunately proved to be
optimum for both activity and selectivity versus CDK2/
cyclin A. Removal of a single chlorine atom (21) or both
(22) from the benzyl moiety of the lead attenuated
activity over 20-fold, suggesting that the conformation
of the sidechain is important. Interestingly, when the
phenyl ring was attached directly to the thiazole ring,
activity could be maintained without aromatic substitu-
tion (30), but kinase inhibition was non-selective. A
comparison of 2-versus 4-chloro and 3-versus 4-methyl
phenyl analogues (33 vs 34 and 31 vs 32) indicates that
4-substitution is not tolerated, suggesting a steric inter-
action at that point that might be avoidable with the
more flexible benzyl analogues. In fact, the even larger
sidechains present in 24, 26 and 27 were tolerated, pre-
sumably due to flexibility. A requirement for an aro-
matic ring is suggested by the inferior activity of
methylsulfonyl analogue 28 versus arylsulfonyl analo-
gue 27, as well as the inactive ethyl analogue 29.
The route depicted in Scheme 2 proved useful for the
production
of
both
aminourea
analogues
(Nu=R2R3NNR4) and alkoxyurea analogues (Nu=R-
ONH), using commercially available hydrazines and
alkoxyamines. Thioamides 5 were either commercially
available, prepared from the corresponding carbox-
amides (Lawesson’s reagent, toluene, 65 ꢀC) or prepared
from the corresponding nitriles (i) 30% H2O2, 3M
Of particular interest were the results obtained by
replacing the hydrazine moiety of 4 with alkoxyamines.
Despite the fact that the oxygen can only be mono-
substituted, these compounds exhibited remarkably
Scheme 2. (a) BrCH2COCO2Et, EtOH, 75 ꢀC; (b) 12 M KOH, 50 ꢀC;
5 M HCl; (c) Oxalyl chloride, cat. DMF, CH2Cl2, rt; NaN3, THF,
40 ꢀC; (d) toluene, 80 ꢀC; (e) oxime resin, CH2Cl2, rt; washes; (f) tolu-
ene or 1,2-dichloroethane, triethylamine, NuH, 80 ꢀC.

S. D. Larsen et al. / Bioorg. Med. Chem. Lett. 13 (2003) 3491–3495
Table 1. CDK5 versus CDK2 inhibition for selected analogues 9
3493
Compd
R1
Nu
CDK5/p25a IC50 (C.I.)c
CDK2/Aa IC50 (C.I.)c
4
2,6-diCl–PhCH2
1
10
10
11
12
13
2,6-diCl–PhCH2
2,6-diCl–PhCH2
2,6-diCl–PhCH2
2,6-diCl–PhCH2
3 (1.5–5)
>100
20 (18–26)
6 (4–9)
30 (20–42)
10 (8–19)
70 (42–117)
14
15
16
17
18
19
20
2,6-diCl–PhCH2
2,6-diCl–PhCH2
2,6-diCl–PhCH2
2,6-diCl–PhCH2
2,6-diCl–PhCH2
2,6-diCl–PhCH2
2,6-diCl–PhCH2
(PhCH2)2NNH–
PhNHNH–
PhN(Me)NH–
MeNHNH–
NH2NH–
>100
>100
10 (5–29)
>100
>100
60 (45–77)
10 (5–28)
Me2NNH–
MeNHNHMe
4 (3.5–5)
>100
21
22
23
24
25
26
27
28
29
30
31
32
2-Cl–PhCH2
30%b
55 (36–78)
26%b
PhCH2
>100
4-Br-PhCH2
PhSCH2
30 (20–35)
80 (59–102)
30 (10–77)
10
30 (13–86)
>100
PhOCH2
4-MeO-PhOCH2
2-pyr-SO2CH2
MeSO2CH2
CH3CH2
20 (6–31)
5
35%b
>100
Ph
9
6
3-Me-Ph
30
20
4-Me-Ph
>100
(continued)

3494
S. D. Larsen et al. / Bioorg. Med. Chem. Lett. 13 (2003) 3491–3495
Table 1 (continued)
Compd
R1
Nu
CDK5/p25a IC50 (C.I.)c
CDK2/Aa IC50 (C.I.)c
33
34
35
2-Cl–Ph
30
10
4-Cl–Ph
>100
Thiophen-2-yl
30
10
36
37
38
39
40
41
2,6-diCl–PhCH2
2,6-diCl–PhCH2
2,6-diCl–PhCH2
2,6-diCl–PhCH2
2,6-diCl–PhCH2
2,6-diCl–PhCH2
MeONH
EtONH
Allyl-ONH
tBuONH
PhCH2ONH
C6F5CH2ONH
9 (6–17)
5 (3–7)
6 (4–10)
1 (0.5–3)
40%b
>100
20 (13–35)
40 (32–60)
15 (11–20)
>100
42
2,6-diCl–PhCH2
3 (2–6)
90 (55–130)
43
44
45
46
47
2,6-diCl–PhCH2
3-MePh
MeSO2CH2
Ph
PhCH2
PhONH
PhONH
PhONH
PhONH
PhONH
0.9 (0.5–1.8)
16 (4–70)
6 (5–11)
>100
>100
90 (50–150)
35 (25–54)
>100
8 (7–15)
20 (9–43)
aRef 12.
b% Inhibition at 20 mM.
cIC50 and 95% confidence intervals (C.I.) expressed in mM.
good activity, in striking contrast to the corresponding
terminal monosubstituted hydrazine analogues. This is
dramatically illustrated by comparing methyl analogues
36 versus 17 and phenyl analogues 43 versus 15. In fact,
phenoxyamine proved to be superior to aminomorpholine
with regard to both activity and selectivity in several
examples (4 vs 43, 31 vs 44, 28 vs 45 and 30 vs 46). Even
the non-aromatic THP analogue 42 possessed selectivity
(>30-fold) surpassing that of the lead 4. The lack of
activity of the benzyloxyamino analogues 40 and 41 is
puzzling in light of the otherwise forgiving SAR in this
series. One can only speculate that the benzyl groups are
too large to be accommodated by the binding site. The
unique ability of the 2,6-dichlorobenzyl substituent on
the thiazole to confer both potency and selectivity was also
observed in the alkoxyamine series, as illustrated by a
comparison of compounds 43–47. The optimum com-
pound from this series proved to be 43, exhibiting sub-
micromolar activity for CDK5/p25 and over 100-fold
selectivity versus CDK2/cyclin A.
only one low energy conformation each, with the next
higher ones being 4.8 and 6.7 kcal greater in energy,
respectively. Common to all three is a co-planar
arrangement of the thiazole ring and the urea moiety,
with the thiazole nitrogen situated anti to the carbonyl
(Fig. 1, only 19 shown). Also common is the observa-
tion that the urea NH bond distal to the thiazole is slightly
out of plane with the thiazole/urea system (Fig. 2). Of
particular interest is the observation that the two active
analogues (19, 36) each have a pair of electrons on the
terminal heteroatom projecting at an acute dihedral
angle to the carbonyl. The inactive analogue 17, on the
other hand, is steered into a conformation that prefers
to have the hydrogen of the terminal nitrogen projecting
toward the carbonyl in a syn-periplanar orientation (dihe-
dral angle for CNNH system is À32.1ꢀ), thereby forcing
the nitrogen’s lone pair to point in a different direction
than in 19 and 36. This conformational difference makes
In an attempt to rationalize the fairly consistent activity
of the N,N-dialkylhydrazine and alkoxyamine analo-
gues versus the uniformly inactive N-monosubstituted
hydrazine analogues, molecular modeling13 was under-
taken on the three closely related analogues 17, 19 and
36. It was found that 19 has two low-energy conforma-
tions available, differing only by inversion of the term-
inal sp3 nitrogen (1.3 kcal). Compounds 17 and 36 have
Figure 1. One of the low-energy conformations of 19.

S. D. Larsen et al. / Bioorg. Med. Chem. Lett. 13 (2003) 3491–3495
3495
4. Dhavan, R.; Tsai, L.-H. Nat. Rev Mol. Biol. 2001, 2, 749.
5. Patrick, G. N.; Zukerberg, L.; Nikolic, M.; de la Monte, S.;
Dikkes, P.; Tsai, L.-H. Nature 1999, 402, 615.
6. Lau, L.-F.; Seymour, P. A.; Sanner, M. A.; Schachter, J. B.
J. Mol. Neurosci. 2002, 19, 267.
7. (a) Wolfe, S.; Ducep, J.-B.; Greenhorn, J. D. Can. J. Chem.
1975, 53, 3435. (b) Holzapfel, C. W.; Pettit, G. R. J. Org.
Chem. 1985, 50, 2323.
8. (a) Beattie, D.; Crossley, R. H.; Hill, D. G.; Shepherd, R. G.
Eur. J. Med. Chem. Chimi. Thererocycl. 1979, 14, 105. (b) A
Curtius rearrangement to generate 5-isocyanato-1,3-thiazoles
has been reported: South, M. S. J. Het. Chem. 1991, 28, 1003.
9. For a review of catch-and-release protocols see: Thompson,
L. A. Curr. Opin. Chem. Biol. 2000, 4, 324.
Figure 2. End view of low-energy conformations of 19, 36, and 17,
respectively. Lone pairs are shown for illustrative purposes only.
10. For another report of trapping isocyanates with oxime
resin, see: Scialdone, M. A. Tetrahedron Lett. 1996, 37, 8141.
11. Representative H NMR spectra (300 MHz, DMSO-d6) d:
Compound 2: 4.62 (s, 2H), 7.41 (t, J=6 Hz, 1H), 7.56 (d, J=6
Hz, 2H), 8.31 (s, 1H), 12.9 (s, 1H). Compound 3: 4.75 (s, 2H),
7.25 (t, J=6 Hz, 1H), 7.40 (d, J=6 Hz, 2H), 8.15 (s, 1H).
Compound 4: 2.74 (s, broad, 4H), 3.65 (s, broad, 4H), 4.52 (s,
2H), 7.12 (s, 1H), 7.40 (t, J=6 Hz, 1H), 7.55 (d, J=6 Hz, 2H),
7.93 (s, 1H), 8.86 (s, 1H).
the electrostatics about the terminal nitrogen for 19 and
36 quite different than 17, which could conceivably
adversely affect the electrostatic complementarity of the
enzyme binding site with 17 and other terminal NH
analogues.
1
In summary, a catch-and-release protocol was devel-
oped for the synthesis of 2-dimensional libraries of 2-
substituted-4-acylamino-1,3-thiazoles that entailed
trapping of intermediate thiazole isocyanates onto
oxime resin and subsequent release under basic catalysis
following washing. The protocol was successful at
avoiding contamination of the final analogues with
byproducts arising from the intermediate Curtius rear-
rangement. Assay of the resulting libraries for selective
inhibition of CDK5/p25 versus the closely homologous
enzyme CDK2/cyclin A revealed that only analogues
derived from 1,1-dialkylhydrazines and alkoxyamines
retained activity. Molecular modeling indicated that the
only apparent conformational difference between the
active analogues and the inactive ones devrived from
monoalkyl hydrazines was the syn-periplanar orientation
of the terminal NH to the urea carbonyl, which changes
the position of the nitrogen lone pair, altering the elec-
trostatic contribution about the terminal heteroatom.
12. Kinase assays. CDK2/cyclin A and CDK5/p25 kinase
assays were performed using the scintillation proximity assay
(SPA) format with the following conditions. The assays contained
50 mM HEPES, pH 7.5, 20 mM Na3VO4, 15 mM MgCl2, 1 mM
DTT, 0.1 mg/mL BSA, 0.2 mg/mL BGG, 0.01% Triton X-100,
1 mM biotinylated peptide (PKTPKKAKKL), and 0.2 mCi
[g33P]-ATP. CDK5/GST-p25 was used at 0.5 nM and CDK2/
GST-cyclin A was used at 4 nM. Reactions were incubated for
30 min at 37 ꢀC and then terminated with the addition of 500
mg streptavidin SPA beads (Amersham) in 50 mM ATP, 5 mM
EDTA, and 0.1% (v/v) Triton X-100 in PBS without calcium
and magnesium. CPM for each well were determined using a
Packard TopCount scintillation counter. Percent inhibition
compared to 100% control was calculated using the formula
100Â(1-(unknownÀbkgd)/(controlÀbkgd)). IC50 values were
calculated using the sigmoidal dose-response equation in
Prism (Graphpad). When tested more than once, the % inhib-
ition or IC50 values were averaged.
13. Several starting geometries were chosen for each com-
pound and then subsequently each geometry was optimized
using ab initio quantum mechanics (HF/6-31G*) in the
GAMESS14 suite of programs. Each compound led to two or
three low energy conformations, and Hessians were obtained
for the two lowest energy conformations for each compound
to confirm the nature of the stationary points–that is minima
on the potential energy surface. Results were visualized with
MacMolPlt15 and conformations were overlaid using
Mosaic2.16
Acknowledgements
The authors gratefully acknowledge the contributions of
Michael Dupuis and Gary Petzold in enzyme preparation.
References and Notes
14. Schmidt, M. W.; Baldridge, K. K.; Boatz, J. A.; Elbert,
S. T.; Gordon, M. S.; Jensen, J. H.; Koseki, S.; Matsunaga,
N.; Nguyen, N.; Su, S. J.; Windus, T. L.; Dupuis, M.; Mon-
tgomery, J. A. J. Comput. Chem. 1993, 14, 1347.
1. Mandelkow, E.-M.; Mandelkow, E. Trends Cell Biol. 1998,
8, 425.
2. Spillantini, M. G.; Goedert, M. Trends Neurosci. 1998, 21,
428.
15. Bode, B. M.; Gordon, M. S. J. Mol. Graphics Mod. 1998,
16, 133.
3. Nikolic, M.; Dudek, H.; Kwon, Y. T.; Ramos, Y. F. M.;
Tsai, L.-H. Genes Dev. 1996, 10, 816.
16. Mosaic2 is a Pharmacia-developed modeling suite of pro-
grams.