202
Y. Deng et al. / Bioorg. Med. Chem. Lett. 24 (2014) 199–203
shows when CDK2/cyclin A is incubated at 40 °C without 2, fol-
lowed by centrifugation that both CDK2 and cyclin A remain in
the supernatant fraction because the heterodimer complex is sta-
ble at 40 °C. However, a similar incubation with increasing concen-
trations of 2 results in the loss of cyclin A (which is insoluble as a
monomer at 40 °C) from the supernatant fraction as evident by the
decrease in the corresponding Coomassie band for the supernatant
fractions and an increasing amount of cyclin A in the pellet frac-
tion. Thus, CDK2-bound 2 inhibits cyclin A binding.
The synthesis of 2 and its derivatives with variations at the 2, 4
and 6-position is summarized in Scheme 1. 5-Bromoisatin 3, was
converted to the dicarboxylic acid 4 using pyruvic acid and potas-
sium hydroxide.19 The aryl group in 5 was introduced using stan-
dard coupling conditions with the corresponding boronic acid.
The dicarboxylic acid 5 was then activated by converting to di-pen-
tafluorophenol (PFP) ester 6 in the presence of DCC. It was found
that the PFP ester at the 2-position of the quinoline was more reac-
tive than at the 4-position and reacting di-PFP ester 6 with one
equivalent of tyrosine methyl ester, followed by hydrolysis, gener-
ated 2 with ꢀ60% yield. A similar approach yielded 7–30.
The SAR for the analogs of 2 proved to be very consistent with
these structural observations. Substitutions at the 6-position (R1)
of the quinoline showed that a meta substituent on the phenyl ring
was preferred and that limited space around the ortho and para-
positions was available due to steric hindrance from Phe146 and
Leu 58 (7, 8). The un-substituted phenyl group reduces the potency
drastically (10). The R1 group binds in a hydrophobic environment,
and introduction of polar groups such as morpholine resulted in
less potent compounds (11). Hydrophobic substitutions at R1 (12,
3-CF3-phenyl) or extended hydrophobic substitutions (13 and 14,
4-Cl-phenylethyl and 4-Cl-styryl, respectively) were favorable
and resulted in single digit nM binding as determined by TdCD
(Kd = 5 nM for 14). These substitutions are keys for differentiating
the binding of the quinoline compounds to CDK2 and not CDK2/cy-
clin A. In the CDK2/cyclin A binary complex, no binding pocket ex-
ists to accommodate the quinoline scaffold and substitutions at the
6-position (R1). In the structure of CDK2 bound with 2 the C-Helix
is pushed 11.7 Å outward which provides room for the compound
binding but sterically inhibits the binding of cyclin A to CDK2.
The interaction of the quinoline 4-position carboxylic acid with
the backbone of the protein proved to be very important. Reduc-
tion of the carboxylic acid to the hydroxymethyl group (16), con-
version to an amide (17) or to an acylsulfonamide (18)
completely abolished binding. In 2, the hydroxyl group of the phe-
nol (AR substitution) functions as both a hydrogen bond donor and
an acceptor and shares two key hydrogen bond interactions with
the hinge amide backbone (Glu81, Leu83). Replacement of the hy-
droxyl group with a fluorine group was 100 fold less potent (19).
Methylation of the same hydroxyl group or relocating to a meta-
position on the tyrosine ring resulted in complete loss of binding
(20 and 21). The methyl ester at R2 was found to be essential for
binding since 22 (R2 = H) and 23 (R2 = COOH) bound poorly to
CDK2. In anticipation of potential metabolic liabilities with the
methyl ester (R2), we decided to investigate its replacement with
more stable functional groups. The crystal structure of 2 with
CDK2 revealed a hydrophobic interaction between the methyl es-
ter moiety and the side chain of Leu 134. This position tolerated
a wide range of substitutions including alkyl (25), amide (26 and
27), heterocyclic (28) and methoxy–methyl groups (29 and 30).
The chemical and structural SAR described above clearly defines
the observed affinity SAR and why the quinoline series bind to the
CDK2 monomer and not the CDK2/cyclin A heterodimer. Despite
the improvement of binding affinity, none of the compounds de-
scribed above show significant activity in a cell based proliferation
assay or inhibition of pRb phosphorylation while CDK2 control
compounds were active in these assays. Several reasons may exist
Figure 4. CDK2/cyclin A complex dissociation by 2 at 40 °C. CDK2/cyclin A at 1.5 lM
was incubated with increasing concentrations of 2 at rt for 30 min and then heated at
40 °C for 30 min. The samples were centrifuged and the supernatant fractions run on
an SDS–PAGE and stained with Coomassie stain (Fig. 4-top). The loss of cyclin A
fractions at the higher concentrations of 2 is an indication of disruption of the CDK2/
cyclin
A complex. Figure 4-bottom is the centrifugation pellet fraction and
demonstrates that cyclin A becomes dissociated in the presence of increasing 2
and is retained in the pellet fraction due to poor thermal stability at 40 °C. Sample C is
a control sample with only 1% DMSO run at 40 °C under the same conditions as the
compound treated samples. The final DMSO in all samples was 1%.
Compound 2 has been shown to bind to CDK2 by direct compe-
tition of the ATP-competitive fluorescently labeled compound (B-
Alexa-Fluor647), binding in the TdF assay where bound 2 increases
the thermal stability of CDK2, and binds in the isothermal titration
calorimetry assay with a Kd = 0.3 lM. The unique binding mode of
2 with CDK2 suggests that 2 and its analogues may be selective
against other kinases. Indeed counter-screening 2 against a panel
of 310 kinases at 1 lM resulted in only one kinase with greater
than 20% inhibition (PDK1, 30% inhibition) while control kinase
inhibitors demonstrated the expected inhibition profile. No inhibi-
tion was observed for the active state of CDK1, CDK2, CDK5, CDK7,
CDK8, and CDK9. We cannot rule out that compound 2 binds to the
inactive (noncyclin bound) states of the CDK family. Binding stud-
ies would be needed to assess if compound 2 could bind to the
inactive states of these CDKs. The bound conformation of 2 to
CDK2 is similar to that observed for Lapatinib bound to EGFR (Type
I½) where these compounds bind into the core of the protein to-
ward the C-Helix but without inducing a DFG-out conformation.17
It is interesting to note that Betzi et al.18 have recently observed
that two 8-anilino-1-naphthalene sulfonate (ANS) hydrophobic
fluorescent probes (Kd was 37 lM) can also bind into the same re-
gion as the quinoline core and the R3 group and induces a similar
C-Helix conformational change. Selected compounds were also
evaluated in a cell based proliferation assay but none showed cell
proliferation inhibition possibly due to permeability issues due to
the R3 carboxylic acid group. Substitution of the ionic R3 carboxylic
acid group with bioisosteres that retain enzyme activity may im-
prove permeability and cellular activity.
Evidence that 2-induced conformational change prevents cyclin
binding to CDK2 was obtained by incubating the compound with
CDK2/cyclin A and measuring the amount of soluble cyclin A after
incubating at 40 °C. Figure 4 is a Coomassie stained SDS gel that