A triple exon-skipping luciferase reporter assay identifies a new CLK inhibitor pharmacophore

Article abstract of DOI:10.1016/j.bmcl.2016.12.056

The splicing of pre-mRNA is a critical process in normal cells and is deregulated in cancer. Compounds that modulate this process have recently been shown to target a specific vulnerability in tumors. We have developed a novel cell-based assay that specifically activates luciferase in cells exposed to SF3B1 targeted compounds, such as sudemycin D6. This assay was used to screen a combined collection of approved drugs and bioactive compounds. This screening approach identified several active hits, the most potent of which were CGP-74514A and aminopurvalanol A, both have been reported to be cyclin-dependent kinases (CDKs) inhibitors. We found that these compounds, and their analogs, show significant cdc2-like kinase (CLK) inhibition and clear structure-activity relationships (SAR) at CLKs. We prepared a set of analogs and were able to ‘dial out’ the CDK activity and simultaneously developed CLK inhibitors with low nanomolar activity. Thus, we have demonstrated the utility of our exon-skipping assay and identified new molecules that exhibit potency and selectivity for CLK, as well as some structurally related dual CLK/CDK inhibitors.

Full text of DOI:10.1016/j.bmcl.2016.12.056

Bioorganic & Medicinal Chemistry Letters xxx (2016) xxx–xxx
Contents lists available at ScienceDirect
Bioorganic & Medicinal Chemistry Letters
journal homepage: www.elsevier.com/locate/bmcl
A triple exon-skipping luciferase reporter assay identifies a new CLK
inhibitor pharmacophore
⇑
Yihui Shi, Jaehyeon Park, Chandraiah Lagisetti, Wei Zhou, Lidia C. Sambucetti, Thomas R. Webb
Division of Biosciences, SRI International, 333 Ravenswood Ave., Menlo Park, CA 94025, USA
a r t i c l e i n f o
a b s t r a c t
Article history:
The splicing of pre-mRNA is a critical process in normal cells and is deregulated in cancer. Compounds
that modulate this process have recently been shown to target a specific vulnerability in tumors. We have
developed a novel cell-based assay that specifically activates luciferase in cells exposed to SF3B1 targeted
compounds, such as sudemycin D6. This assay was used to screen a combined collection of approved
drugs and bioactive compounds. This screening approach identified several active hits, the most potent
of which were CGP-74514A and aminopurvalanol A, both have been reported to be cyclin-dependent
kinases (CDKs) inhibitors. We found that these compounds, and their analogs, show significant cdc2-like
kinase (CLK) inhibition and clear structure-activity relationships (SAR) at CLKs. We prepared a set of ana-
logs and were able to ‘dial out’ the CDK activity and simultaneously developed CLK inhibitors with low
nanomolar activity. Thus, we have demonstrated the utility of our exon-skipping assay and identified
new molecules that exhibit potency and selectivity for CLK, as well as some structurally related dual
CLK/CDK inhibitors.
Received 15 November 2016
Revised 20 December 2016
Accepted 21 December 2016
Available online xxxx
Keywords:
Cdc-like kinase (CLK) inhibitors
Modulators of pre-mRNA splicing
High throughput screening
Anti-tumor agents
Ó 2016 Elsevier Ltd. All rights reserved.
Introduction
in turn regulated by cis-acting regulatory sites on pre-mRNA sub-
strates.1 Since pre-mRNAs for a given gene may contain many dif-
The processing of pre-mRNA to mature mRNA in metazoans is a
critical process for the development and normal functioning of
cells. The pre-mRNA splicing process involves the removal of inter-
vening sequences from pre-mRNA followed by the ligation of
exons to form mature mRNA. This splicing process is catalyzed
and regulated by a highly complex macromolecular protein-RNA
complex called the spliceosome. The spliceosome is composed of
five small nuclear ribonucleoproteins (snRNPs) (U1, U2, U4, U5
and U6) and over 150 associated proteins.1,2 The pre-mRNA matu-
ration process includes alternative splicing (AS), which is the
mechanism that allows for different forms of mature mRNAs to
be generated from the same pre-mRNA. Commonly, alternative
splicing patterns determine the inclusion or exclusion of portions
of the coding sequence in the mRNA, giving rise to protein isoforms
that differ in their peptide sequence. Alternative splicing is regu-
lated by numerous spliceosomal trans-acting proteins, which are
ferent exon and intron combinations, there are often a very large
number of possible mRNAs that can lead to a correspondingly large
set of proteins with different, even opposing, biological functions
within the cell. The complexity of the spliceosome and the current
scarcity of molecular-resolution X-ray structures complicates a
rapid advancement in the understanding of many of the important
functional mechanisms that are critical to the normal functioning
of the cells of higher organisms.3 Because of the importance of
splicing in normal organismal development, the spliceosome is
increasingly being recognized as a major frontier for molecular
biology and is now accepted as a valid oncology target.4,5
Interest in the spliceosome was dramatically bolstered when
two independent groups reported that a pair of structurally diver-
gent bacterial natural products, FR901464 and pladienolide, both
target a similar site on the SF3B subunit of the spliceosome.6,7 Sub-
sequent to those initial discoveries the list of compounds that are
known to target the SF3B subunit has grown to include additional
bacterial natural products such as herboxidiene (GEX1A)8 (isolated
from Streptomyces sp. A7847) and the thailanstatins (isolated from
Burkholderia thailandensis).9 These bacterial fermentation products
show cytotoxic IC50s in the low nanomolar range in tumor cell lines
and were reported to have a similar distinctive effect on the cell
cycle in mammalian cell lines, which includes cell cycle arrest in
Abbreviations: CLK, cdc2-like kinase; MDM2, mouse double minute 2 homolog;
Luc, luciferase; TEA, triethylamine; DIPEA, diisopropylethylamine; EtOAc, ethyl
acetate.
⇑
Corresponding author.
E-mail address: thomas.webb@sri.com (T.R. Webb).
http://dx.doi.org/10.1016/j.bmcl.2016.12.056
0960-894X/Ó 2016 Elsevier Ltd. All rights reserved.
Please cite this article in press as: Shi Y., et al. Bioorg. Med. Chem. Lett. (2016), http://dx.doi.org/10.1016/j.bmcl.2016.12.056

2
Y. Shi et al. / Bioorganic & Medicinal Chemistry Letters xxx (2016) xxx–xxx
the G1 and G2/M phases.10 Several of these natural products have
also been reported to show potent antitumor activity in vitro and
in vivo.11,12 Work in this area led to the development of the
semisynthetic pladienolide analog E7107 that entered Phase I clin-
ical studies,7,13,14 without the benefit of many subsequent recent
discoveries relevant to mechanism of action, tumor selectivity,
and patient stratification.4,5
Results and discussion
We have previously demonstrated that the SF3B1 targeted
splicing modulator drugs potently induce exon skipping in
MDM2 pre-mRNA.22,24 More specifically, splicing of MDM2 is
altered in response to treatment with the sudemycins, and other
SF3B1 active agents, such that exons 4–11 are skipped.22 Based
on these observations, we designed an exon-skipping MDM2-Luc
reporter (see Fig. 2).38 The principal strategy in the design of this
reporter is that, when expressed, cellular luminescence is depen-
dent on the production of a correctly spliced full-length luciferase
transcript due to the drug induced skipping of MDM2 cassette
exons 4, 10, and 11, which interrupt the luciferase gene in this con-
struct in the absence of a splicing modulator drug (see Fig. 2). We
have previously reported the use of this construct as a reporter for
the real time pharmacodynamics of SF3B1 targeted agents in vivo,
via a stable reporter cell line.38 Using a stable cell line expressing
this reporter, luminescent signals are generated in a dose-depen-
dent fashion in the presence of splicing modulators (such as the
sudemycins, herboxidiene and pladienolide B).38 We have used
this assay to screen a small molecule library (composed of 830
known drugs and 4359 bioactive compounds)39 and identified sev-
eral active ‘hit’ compounds. These hits were then subjected to
dose–response assays in the MDM2-Luc exon-skipping assay and
those actives were then confirmed in an orthogonal PCR-based
MDM2 alternative splicing assay in Rh-18 cells (see Supporting
Information).22 Two structurally related ‘hits’ demonstrated activ-
ity in both of these screens that is similar (but less potent) to that
observed for SF3B1 targeted agents such as SD6 (see Supporting
Information). The most potent of these confirmed hits are shown
in Fig. 3 below.
As part of our effort to develop a class of drug-like synthetic
spliceosome modulators our group has previously reported the
design and synthesis of FR analogs that contain only 3 chiral cen-
ters (the sudemycins),15,20–22 pladienolide analogs,23 and several
herboxidiene analogs24 all of which are active compounds that
effectively modulate alternative splicing.22 Also, new details
regarding the mechanism of action of SF3B1 targeted agents have
been elucidated.25 Additionally results of the genome-wide array
analysis of sudemycin treated tumor cells, which shows that sude-
mycins cause a rapid wide-ranging change in alternative pre-
mRNA splicing and that a biotin-labeled sudemycin probe directly
interacts with the SF3B1 protein have been reported.26 This collab-
orative project ultimately led to sudemycin D6 (SD6),15 which is
currently in preclinical development as an anticancer agent (see
Fig. 1). Several groups have independently discovered new diverse
small molecule structural classes that effect pre-mRNA splicing,
using a range of screening strategies (see Fig. 1). These compounds
include TG-003,16 KH-CB19,17 Araki Cpd-2,18 and Madrasin19 (see
Fig. 1). Compounds KH-CB19 and the Araki Cpd-2 have been
reported to be highly selective inhibitors of the of cdc2-like kinase
(CLK) family,17,18 while the molecular target of Madrasin has not
been reported.19
In parallel to the work described above, strong evidence has
continued to mount that aberrant splicing of pre-mRNA is a driver
of tumorigenesis27 and that the spliceosome is a valid target for
cancer therapy.4,5 Recent groundbreaking discoveries have identi-
fied recurrent mutations in SF3B1 (and/or other splicing factors)
in multiple forms of cancer including: myelodysplastic syndromes
(MDS),14,28 chronic lymphocytic leukemia (CLL),29 acute myeloid
leukemia (AML),30,31 breast cancer,32,33 lung adenosarcoma,34 and
uveal melanoma.35 These genetic studies have also fueled comple-
mentary research in the therapeutic significance of spliceosome
recurrent mutations. Very recently the selective sensitivity of
tumors to agents that target SF3B1 have also been linked to over-
expression of MYC.36,37 Thus the collective progress in natural pro-
duct screening, target identification, spliceosome related medicinal
chemistry, and high-throughput transcriptome sequencing has led
to a remarkable convergence of independent research areas, which
have simultaneously identified new oncology drug targets and new
small-molecule therapeutic leads.
These two confirmed hits, CGP-74514A (1)40 and aminopur-
valanol A (2)41 are known to inhibit cyclin-dependent kinases
(CDKs), and in particular CDK1.41 Compound 1 was reportedly
optimized for selective CDK1 activity through a significant analo-
ging effort.40 However, structurally similar compounds have been
reported to have a very broad spectrum of biological activities,41
which indicates that this inhibitor class may also inhibit other
important protein kinases. In order to narrow down the possible
identity of the target class that led to exon-skipping in our assay
we then screened additional commercially available selective
kinase inhibitor tool compounds that have shown cell-based activ-
ity.42 The compounds that we evaluated included the CDK selective
clinical compound dinaciclib (a nanomolar inhibitor of CDK1,
CDK2, CDK5, and CDK9),43 SRPIN340 (selective serine arginine pro-
tein kinase (SRPK) 1 inhibitor (Ki = 0.89
bits SRPK1 and SRPK2 but does not significantly inhibit other
lM; this compound inhi-
O
Sudemycin D1: R=N(CH3)2
Sudemycin D6: R=NHCH3
TG-003
Cl
KH-CB19
Cl
H2N
O
CN
O
O
R
S
N
O
O
O
H
N
N
O
Cl
Madrasin
O
Araki Cpd-2
O
N
O
N
N
HN
HN
N
N
N
N
N
NH
H
Fig. 1. Some recently reported synthetic alternative splicing modulators.15–19
Please cite this article in press as: Shi Y., et al. Bioorg. Med. Chem. Lett. (2016), http://dx.doi.org/10.1016/j.bmcl.2016.12.056

Y. Shi et al. / Bioorganic & Medicinal Chemistry Letters xxx (2016) xxx–xxx
3
SRPKs such as CLK1 and CLK4, or other classes of SR kinases).44 We
found all of these compounds to be inactive (at 10 M) in the
l
MDM2-Luc assay, which eliminated SRPK1, SRPK2, CDK1, CDK2,
CDK5 and CDK9 as the likely targets that are responsible for the
observed activity in our MDM2-Luc cell based assay.
Fig. 2. The MDM2-Luc construct that contains parts of the MDM2 intronic and
exonic sequences in the luciferase gene.22
In contrast to the lack of activity of the tool compounds dis-
cussed above, the CLK1/2/4 cell-active selective kinase inhibitor
KH-CB19 showed modest activity in the MDM2-Luc assay (see Sup-
porting Information).17 These results, taken together with the
recent observations of pronounced modulation of splicing (through
inhibition of the CLK mediated phosphorylation of SR proteins17
)
by compounds such as Araki compound-2 (see Fig. 1),18 strongly
suggested that CLK inhibition was responsible for the alternative
splicing effects seen with compounds 1 and 2; though splicing
modulation has not been recognized as an activity of 1 or 2, to
our knowledge. In order to explore the CLKs as possible targets,
and to better understand the structure-activity relationships
(SAR), additional new structurally related analogs along with the
associated CLK biochemical inhibition data were clearly needed
so decided to evaluate the most potent of these compounds (com-
pound 1) for CLK activity. Remarkably, as shown below (see
Fig. 3. The most active PCR validated hits from a run of the MDM2-Luc exon-
skipping screen.
Table 1
Analogs of 1 and 2 and their activity in the MDM2-Luc reporter assay, prepared as shown in Scheme 1.a
Compound
Ar
X
R1
Et
R2-NH
MDM2/Luc
Active
Compound
Ar
X
R1
R2-NH
MDM2/Luc
Active
CGP-74514A (1)
3-Cl-phenyl
N
10
3-Cl-phenyl
N
i-Pr
5
6
3-Cl-phenyl
3-Cl-phenyl
N
N
Et
Et
Active
Active
11
12
3-Cl-phenyl
3-Cl-phenyl
CH
N
i-Pr
NA
NA
Et
7
3-Cl-phenyl
N
i-Pr
Active
13
3-Cl-phenyl
N
i-Pr
NA
8
9
3-Cl-phenyl
3-Cl-phenyl
N
N
i-Pr
i-Pr
Active
Active
14
15
3-Cl-5-NH2-phenyl
3-Cl-5-AcNH-phenyl
N
N
i-Pr
i-Pr
Active
Active
a
Active means a significant MDM2-Luc reporter dose response (greater than approximately 3-fold over DMSO control, where DMSO is set as Luc-Fold = 1.0) in luciferase
activation based on the average of three experiments. Sudemycin D6 serves a positive control. Abbreviations: NA, not active. See Experimental Section and Supporting
Information for more details.
Please cite this article in press as: Shi Y., et al. Bioorg. Med. Chem. Lett. (2016), http://dx.doi.org/10.1016/j.bmcl.2016.12.056

4
Y. Shi et al. / Bioorganic & Medicinal Chemistry Letters xxx (2016) xxx–xxx
Table 2
The biochemical inhibition results for the active compounds from Table 1.a
Compound
CLK1 IC50 (nM)
CLK2 IC50 (nM)
CLK3 IC50 (nM)
CLK4 IC50 (nM)
CDK1 IC50 (nM)
CDK4 IC50 (nM)
CDK6 IC50 (nM)
CGP-74514A (1)
5
6
7
SRI-29329 (8)
9
10
14
15
148
868
371
400
78
38
74
18
19
111
498
190
120
16
11
26
16
13
>1000
>1000
>1000
>1000
>1000
>1000
>1000
>1000
>1000
104
>1000
498
634
86
50
70
54
64
382
546
573
>1000
>1000
>1000
>1000
>1000
>1000
>1000
>1000
>1000
>1000
>1000
>1000
>1000
541
>1000
>1000
930
>1000
>1000
>1000
674
>1000
973
>1000
a
Enzyme inhibition of CLK subtypes 1–4 and CDK 1, 4, and 6.
Tables 1and 2) we found that compound 1 is a more potent
inhibitor of CLK1, CLK2 and CLK 4 than it is as at CDK1, based on
biochemical assays.
known that CLK inhibition can lead to modulation of pre-mRNA
splicing,18 and since we observed that the weakly cell-active, but
selective, CLK inhibitor KH-CB19 shows activity in the MDM2-
Luc reporter assay, we decided to explore the activity of 1 (and
our active analogs) in regard to their biochemical inhibition of
CLK 1–4 (see Table 2). The results summarized in Table 2 clearly
show that CGP-74514A (1), which has been considered a selective
CDK inhibitor46 actually also shows significant inhibition of CLK 1,
2, and 4 and so can be considered a dual CLK/CDK inhibitor. Addi-
tionally some of our analogs of 1 and 2 were more active for three
of the four CLK subtypes and the potency trend of CLK inhibition
across the subtypes is similar to the trend that was observed in
the MDM2-Luc reporter assay, in that more potent CLK inhibitors
(compounds 9, 10, 14, 15) showed a strong luciferase response
(see Supporting Information). Interestingly compound 8 shows
some modest (ꢀ5-fold) selectivity for CLK2 over CLK1 and CLK4
without significant CDK1,4,6 activity and compounds 9, 10, 14
and 15 that showed the strongest Luc activation also show potent
CLK inhibition for the 1, 2 and 4 subtypes. We also screened com-
pounds that were inactive in the triple-exon skipping assay, which
were the closest analogs to the active compounds (11, 12 and 13)
for CLK1 activity (as described in the Experimental Section below)
We therefore prepared a set of analogs starting from the versa-
tile starting material 3 of the general structure 4 containing some
of the pharmacophore features of 1 and/or 2 (see Scheme 1 and
Table 1) using well-precedented chemistry that has been used to
make similar analogs,40,41,45 in order to explore the nature of the
pharmacophore that led to the observed activation of the triple-
exon skipping in our assay.38 As can be seen by inspection of
Table 1, replacement of the N7 nitrogen with CH completely abro-
gated all activity at 10 lM (compound 11), as did acetylation of the
basic primary amine (compound 12), and dramatically reduced
activity by replacement of the valanol group of 2 with an unsubsti-
tuted methyl ether (compound 13). The absolute and relative
stereochemistry of the cyclohexane diamine also affected the
activity. By combining potential interaction features from both 1
and 2 into compound 14 and into the acetanilide derivative 15,
we observed strong MDM2-Luc activity (see Supporting Informa-
tion). Thus, with this set of compounds we have identified some
of the critical geometric and interaction pharmacophore features
that lead to activation in the MDM2-Luc assay.
Since the pharmacophores present in 1 and 2 do not match the
previously reported consensus pharmacophore for agents that tar-
get the SF3B1 splicing protein,15,20,24 it appeared unlikely that 1 or
2 target the sudemycin D6 binding site on SF3B1. However, since it
Analogs 11 and 12 have IC50s for CLK1 ꢁ 1
lM (data not shown)
and thus serve as negative controls that further strengthen the
connection between the CLK activity in our compounds and the
activity observed in the triple-exon skipping screen. Compound
13 is active in the CLK assay, but has very poor solubility (data
not shown), which may explain its inactivity in the reporter assay.
The weakly cell active tool compounds Madrasin19 and TG-00316
were also screened for activity in the MDM2-Luc assay at a concen-
tration of 10 lM, but neither compound showed significant activ-
ity in our hands in this screen (data not shown). This is in line
with the weak activity reported for Madrasin with an EC50 of
ꢀ20
CLK inhibitor tool compound that showed activity (KH-CB19, also
reported to be weakly active in cells at 10
M)17 barely passes
l
M.19 In regard to TG-003 it should be noted that the other
l
the minimal hit criteria in our MDM2-Luc screen. This suggests
that TG-003 may be even less ‘drug like’ in terms of its cell
penetration. Given all of this information we selected the active
Scheme 1. Synthesis of substituted Diaminopurine and Diaminopyrrolo[2,3-d]
pyrimidine Analogs.
Fig. 4. Active CLK inhibitors that were profiled against 22 oncogenic kinases (see Supporting Information).
Please cite this article in press as: Shi Y., et al. Bioorg. Med. Chem. Lett. (2016), http://dx.doi.org/10.1016/j.bmcl.2016.12.056

Y. Shi et al. / Bioorganic & Medicinal Chemistry Letters xxx (2016) xxx–xxx
5
compounds 8, 9, and 14 for kinase profiling against 25 kinases42 at
a concentration of 1 M in duplicate for each analog (see Fig. 4
the compounds shown in Tables 1 and 2. This data is presented
in Table 4. Many of these analogs showed potent activity at
CLK1, CLK2 and CLK4 but exhibited some consistent variation
depending on the stereochemistry of the diamine group and
the nature of the diamine constraint. Compound 16 (with the
1R,2S configuration of the cyclohexyldiamine) showed the most
potent inhibition of CLK2 (6 nM) of any compound evaluated.
Compound 17, the enantiomer of 16, showed similar activity at
CLK1 and CLK4 but was significantly less active at CLK2
(26 nM). A similar trend is seen with the enantiomers 18 and
19. An even more pronounced stereochemistry driven
structure-activity difference is seen between the 1S,2S
cyclohexyldiamine analog 20 and its enantiomer 21, with a
potency difference of >10-fold for CLK2.
l
below and the Supporting Information). This kinase profiling
screen showed <50% inhibition against the entire 22 diverse kinase
panel for compound 8, with some significant off-target activity for
9 only against Flt3, Src(1–530) and off-target activity for 14 against
Flt3, Plk1, and Src(1–530) (see Supporting Information for com-
plete screening results).
Given these promising results we decided to prepare a follow-
up set of optically pure analogs 16–25 (see Table 3) that holds
the R1 group constant. This set of analogs includes different ring
constraints and derivatives with a single stereocenter (com-
pounds 22–25). This set of compounds was then subjected to
the same kinase inhibition assays that had been performed on
Table 3
Second-generation analogs of 1 and 2, prepared as shown in Scheme 1.
Compound
Ar
R2-NH
Compound
Ar
R2-NH
16
3-Cl-5-NH2-phenyl
21
3-Cl-5-NH2-phenyl
17
18
3-Cl-5-NH2-phenyl
3-Cl-5-AcNH-phenyl
22
23
3-Cl-phenyl
3-Cl-phenyl
19
20
3-Cl-5-AcNH-phenyl
3-Cl-5-NH2-phenyl
24
25
3-Cl-phenyl
3-Cl-phenyl
Table 4
The biochemical inhibition results for the active compounds from Table 3.a
Compound
CLK1 IC50 (nM)
CLK2 IC50 (nM)
CLK3 IC50 (nM)
CLK4 IC50 (nM)
CDK1 IC50 (nM)
CDK4 IC50 (nM)
CDK6 IC50 (nM)
16
17
18
19
20
21
22
23
24
25
17
21
19
26
56
274
31
66
71
118
6
26
7
17
14
163
26
34
28
92
>1000
>1000
>1000
>1000
>1000
>1000
>1000
>1000
>1000
>1000
25
28
30
31
124
648
61
102
77
532
>1000
>1000
611
>1000
>1000
>1000
>1000
>1000
>1000
>1000
>1000
>1000
>1000
>1000
>1000
>1000
>1000
>1000
>1000
>1000
>1000
>1000
>1000
>1000
>1000
>1000
>1000
>1000
>1000
299
a
Enzyme inhibition of CLK subtypes 1–4 and CDK 1, 4, and 6.
Please cite this article in press as: Shi Y., et al. Bioorg. Med. Chem. Lett. (2016), http://dx.doi.org/10.1016/j.bmcl.2016.12.056

6
Y. Shi et al. / Bioorganic & Medicinal Chemistry Letters xxx (2016) xxx–xxx
Fig. 5. Left Panel: A representation of the structure of debromohymenialdisine (K0010) bound to CLK1 using the published coordinates,47 highlighting the hydrogen bonds to
the ligand (as blue lines). This figure was prepared using UCSF Chimera.48 This perspective is similar to that shown in Fig. 5. Right Panel: A hypothetical 2D overlay of the
hydrogen bond pharmacophore features from K0010 (blue) over the structure of CLK inhibitor 14 (shown in grey) from Table 1 above. The overlay is intentionally offset so
that similar features are visible in both structures. Abbreviations: HB-Don = hydrogen-bond donor: HB-Acc = hydrogen-bond acceptor. The notation ‘Asp-325?’ refers to a
hypothetical interaction with this residue in CLK1, assuming the overlay is correct.
We also considered the published co-crystal structures CLK1
with the CLK selective inhibitor KH-CB19,17 and the promiscuous
debromohymenialdisine,47 which may be helpful in developing a
pharmacophore hypothesis (see Fig. 5) as part of future work. As
shown in Fig. 5 a potential a hypothetical binding mode can be pro-
posed as a conceptual starting point. This possible pharmacophore
is based on our initial SAR data, and the published CLK1/debromo-
hymenialdisine structure together suggest a potentially important
charge-charge hydrogen-bond interaction between CLK1 Asp-325
and the basic primary amine in the active selective compounds
such as 8, 9, 14, and 15. Future work can test this hypothetical
pharmacophore.
in full length MDM2 pre-mRNA in a different cell background.22
Using this screening approach we have identified compounds that
are dual CLK2/CDK1 inhibitors (such as 1) that we optimized to
yield more specific CLK inhibitors (such as SRI-29329, 8). This is
a notable result since CLK250 and CDK151 have been independently
identified as potential targets for breast cancer. Therefore these
active analogs, SAR results and our MDM2-Luc assay may facilitate
the development of new probe and lead molecules with potential
application in breast cancer, and other diseases. Additionally the
MDM2-Luc assay has proven its utility in the identification of com-
pounds with splicing modulatory activity at targets other than
SF3B1. The utility of the triple-exon skipping splicing modulatory
screen is remarkably selective. None of the 830 FDA approved
drugs showed activity at 10 lM in our assay, indicating that this
Conclusion
assay may be useful in selectively identifying compounds that
act through other unique mechanisms involving modulation of
pre-mRNA splicing that may be relevant to innovative drug discov-
ery. Additionally our observations of overlapping pharmacophores
for the CLK and CDK (and potentially structurally related CDK inhi-
bitors) should be considered when interpreting biological results
that are based on the cellular or in vivo responses to compounds
such as CGP-74514A (1).
We have used our triple-exon skipping assay to screen a small
molecule library (composed of 830 known drugs and 4359 bioac-
tive compounds39, see Supporting Information) and have identified
initial hits. Using directed chemistry, which included a range of
optically pure analogs, we also identified new more potent and
selective CLK inhibitors and associated structure-activity relation-
ships. The fact that our screen identified CDK inhibitors that also
show off-target effects against the highly homologous CLK subfam-
ily is not surprising, given the close structural relationship
between these two branches of the kinase family tree and the
reported partial selectivity of many clinical kinase inhibitors.49
We used dose response in our MDM2-Luc exon-skipping assay
and the orthogonal PCR MDM2 alternative splicing assay in Rh-
18 cells (see Supporting Information) to validate the active hits
and analogs and as a guide in the development of SAR. Some of
the new analogs showed significant activity in the MDM2-Luc
assay and are potent biochemical inhibitors of CLK subtypes 1, 2
and 4. Several of these compounds show reasonable overall kinase
selectivity, based on data from a total of 29 kinases (including
CLK1-4; CDK 1, 4, and 6; and a screen against a panel of 22 onco-
genic kinases, see SI). It is important to note that the well-studied
tool compound CGP-74514A (1) shows significantly more bio-
chemical activity against CLK 1, 2, and 4 than either CDK 1 or
CDK 2, under the assay conditions that we used (see Experimental
section below). Additionally, we found that CLK inhibition gener-
ally correlated with MDM2-Luc activity in this series of com-
pounds. The lack of a close correlation between the MDM2 triple
exon skipping and CLK inhibition and biochemical activity may
be due to the fact that inhibitors of CLKs do not efficiently inhibit
this unusual type of alternative splicing modulation, which has
only been previously reported for SF3B1 modulators.22,24,38 The
use of the RT PCR assay as a confirmatory counterscreen for
MDM2 exon skipping in Rh-18 cells validates this splicing effect
Author contributions
All authors have given approval to the final version of the
manuscript.
Notes
The authors declare no competing financial interest.
Acknowledgments
We would like to thank Dr. Amanda Joyner for the design of the
MDM2/Luc reporter construct, and for her work on the initial
screens along with Dr. Gustavo Palacios and Dr. William Shadrick,
whom we also thank. We would also like to thank Dr. Philip M. Pot-
ter for important insights and helpful discussions. This work was
supported in part by National Institutes of Health Grant
CA140474 (TRW) and a Stanford Cancer Institute Translational
2015–2016 Award (TRW).
A. Supplementary material
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.bmcl.2016.12.056.
Please cite this article in press as: Shi Y., et al. Bioorg. Med. Chem. Lett. (2016), http://dx.doi.org/10.1016/j.bmcl.2016.12.056

Y. Shi et al. / Bioorganic & Medicinal Chemistry Letters xxx (2016) xxx–xxx
26. Convertini P, Shen M, Potter PM, et al. Nucl Acids Res. 2014;42:4947–4961.
7
References
27. David CJ, Manley JL. Genes Dev. 2010;24:2343–2364.
28. Ogawa S. Int J Hematol. 2012;96:438–442.
29. Damm F, Nguyen-Khac F, Fontenay M, Bernard OA. Leukemia.
2012;26:2027–2031.
30. Murati A, Brecqueville M, Devillier R, Mozziconacci MJ, Gelsi-Boyer V,
Birnbaum D. BMC Cancer. 2012;12:304.
31. Cancer Genome Atlas Research N. New England J Med. 2013;368:2059–2074.
32. Cancer Genome Atlas N. Nature. 2012;490:61–70.
33. Maguire SL, Leonidou A, Wai P, et al. J Pathol. 2015;235:571–580.
34. Imielinski M, Berger AH, Hammerman PS, et al. Cell. 2012;150:1107–1120.
35. Harbour JW. Genomic, prognostic, and cell-signaling advances in uveal
melanoma. American Society of Clinical Oncology educational book/ASCO
American Society of Clinical Oncology Meeting; 2013: 388–391.
36. Hubert CG, Bradley RK, Ding Y, et al. Genes Dev. 2013;27:1032–1045.
37. Hsu TY, Simon LM, Neill NJ, et al. Nature. 2015.
38. Shi Y, Joyner AS, Shadrick W, et al. Pharmacol Res Perspect. 2015;3:e00158.
39. Walters BJ, Lin W, Diao S, et al. PloS One. 2014;9:e91173.
40. Imbach P, Capraro HG, Furet P, Mett H, Meyer T, Zimmermann J. Bioorganic Med
Chem Lett. 1999;9:91–96.
41. Chang YT, Gray NS, Rosania GR, et al. Chem Biol. 1999;6:361–375.
42. Uitdehaag JC, Verkaar F, Alwan H, de Man J, Buijsman RC, Zaman GJ. Br J
Pharmacol. 2012;166:858–876.
43. Parry D, Guzi T, Shanahan F, et al. Mol Cancer Ther. 2010;9:2344–2353.
44. Fukuhara T, Hosoya T, Shimizu S, et al. Proc Natl Acad Sci USA.
2006;103:11329–11333.
1. Kramer A. Ann Rev Biochem. 1996;65:367–409.
2. Wahl MC, Will CL, Luhrmann R. Cell. 2009;136:701–718.
3. Hang J, Wan R, Yan C, Shi Y. Science. 2015.
4. Webb TR, Joyner AS, Potter PM. Drug Discovery Today. 2013;18:43–49.
5. Bonnal S, Vigevani L, Valcarcel J. Nat Rev Drug Discov. 2012;11:847–859.
6. Kaida D, Motoyoshi H, Tashiro E, et al. Nat Chem Biol. 2007;3:576–583.
7. Kotake Y, Sagane K, Owa T, et al. Nat Chem Biol. 2007;3:570–575.
8. Hasegawa M, Miura T, Kuzuya K, et al. ACS Chem Biol. 2011;6:229–233.
9. Liu X, Biswas S, Berg MG, et al. J Nat Products. 2013;76:685–693.
10. Nakajima H, Sato B, Fujita T, Takase S, Terano H, Okuhara M. J Antibiotics.
1996;49:1196–1203.
11. Nakajima H, Hori Y, Terano H, et al. J Antibiotics. 1996;49:1204–1211.
12. Mizui Y, Sakai T, Iwata M, et al. J Antibiotics. 2004;57:188–196.
13. Eskens FA, Ramos FJ, Burger H, et al. Clin Cancer Res. 2013;19:6296–6304.
14. Yoshida K, Sanada M, Shiraishi Y, et al. Nature. 2011;478:64–69.
15. Lagisetti C, Palacios G, Goronga T, Freeman B, Caufield W, Webb TR. J Med Chem.
2013;56:10033–10044.
16. Muraki M, Ohkawara B, Hosoya T, et al. J Biol Chem. 2004;279:24246–24254.
17. Fedorov O, Huber K, Eisenreich A, et al. Chem Biol. 2011;18:67–76.
18. Araki S, Dairiki R, Nakayama Y, et al. PloS One. 2015;10:e0116929.
19. Pawellek A, McElroy S, Samatov T, et al. J Biol Chem. 2014;289:34683–34698.
20. Lagisetti C, Pourpak A, Jiang Q, et al. J Med Chem. 2008;51:6220–6224.
21. Lagisetti C, Pourpak A, Goronga T, et al. J Med Chem. 2009;52:6979–6990.
22. Fan L, Lagisetti C, Edwards CC, Webb TR, Potter PM. ACS Chem Biol.
2011;6:582–589.
45. Oumata N, Bettayeb K, Ferandin Y, et al. J Med Chem. 2008;51:5229–5242.
46. Dai Y, Dent P, Grant S. Cell Cycle. 2002;1:143–152.
23. Gundluru MK, Pourpak A, Cui X, Morris SW, Webb TR. Medchemcomm.
2011;2:904–908.
24. Lagisetti C, Yermolina MV, Sharma LK, Palacios G, Prigaro BJ, Webb TR. ACS
Chem Biol. 2014;9:643–648.
25. Effenberger KA, Urabe VK, Prichard BE, Ghosh AK, Jurica MS. Rna.
2016;22:350–359.
47. Bullock AN, Das S, Debreczeni JE, et al. Structure. 2009;17:352–362.
48. Pettersen EF, Goddard TD, Huang CC, et al. J Comput Chem. 2004;25:1605–1612.
49. Fabian MA, Biggs 3rd WH, Treiber DK, et al. Nat Biotechnol. 2005;23:329–336.
50. Yoshida T, Kim JH, Carver K, et al. Cancer Res. 2015;75:1516–1526.
51. Kang J, Sergio CM, Sutherland RL, Musgrove EA. BMC Cancer. 2014;14:32.
Please cite this article in press as: Shi Y., et al. Bioorg. Med. Chem. Lett. (2016), http://dx.doi.org/10.1016/j.bmcl.2016.12.056