Optimization of small molecule agonists of the thrombopoietin (Tpo) receptor derived from a benzo[a]carbazole hit scaffold

Article abstract of DOI:10.1016/j.bmcl.2008.08.077

The lead optimization of a novel series of benzo[a]carbazole-based small molecule agonists of the thrombopoietin (Tpo) receptor is reported. The chemical instability of the dihydro-benzo[a]carbazole lead 2 was successfully addressed in the design and eval

Full text of DOI:10.1016/j.bmcl.2008.08.077

Bioorganic & Medicinal Chemistry Letters 18 (2008) 5259–5262
Contents lists available at ScienceDirect
Bioorganic & Medicinal Chemistry Letters
journal homepage: www.elsevier.com/locate/bmcl
Optimization of small molecule agonists of the thrombopoietin (Tpo)
receptor derived from a benzo[a]carbazole hit scaffold
*
Thomas H. Marsilje , Phil B. Alper, Wenshuo Lu, Daniel Mutnick, Pierre-Yves Michellys, Yun He,
Donald S. Karanewsky, Donald Chow, Andrea Gerken, Jianmin Lao, Min-Ju Kim, H. Martin Seidel,
Shin-Shay Tian
Genomics Institute of the Novartis Research Foundation (GNF), San Diego, CA 92121, USA
a r t i c l e i n f o
a b s t r a c t
Article history:
The lead optimization of a novel series of benzo[a]carbazole-based small molecule agonists of the throm-
bopoietin (Tpo) receptor is reported. The chemical instability of the dihydro-benzo[a]carbazole lead 2
was successfully addressed in the design and evaluation of compounds which also demonstrated
improved potency compared to 2. Members of the scaffold have been identified which are full agonists
that demonstrate cellular functional potency <50 nM. Analog 21 demonstrates equivalent efficacy in
Received 22 June 2008
Revised 15 August 2008
Accepted 18 August 2008
Available online 26 August 2008
the human megakaryocyte differentiation (CFU-mega) assay compared to EltrombopagTM
.
Keywords:
Tpo
Ó 2008 Elsevier Ltd. All rights reserved.
Thrombopoietin
TpoR
cMpl
Hematopoietic growth factor
Agonist
The preceding publication in this journal describes our initial
identification of a structurally novel series of human Tpo receptor
agonists based upon the benzo[a]carbazole ring system.1 Figure 1
shows the known2–5 TpoR agonist EltrombopagTM (1) as well as a
representative lead structure from our initial SAR exploration (2).
As previously discussed,1 the dihydro-benzo[a]carbazole scaffold
suffers from chemical instability via spontaneous oxidation of its
ethylene bridge yielding the fully aromatic benzo[a]carbazole ring
system. We wanted to avoid the fully aromatic system for two rea-
sons: (a) the possibility of oxidation of the 5,6-double bond in vivo
potentially generating a reactive species and (b) the flat fully aro-
matic nature of this molecule making intercalation into DNA an
undesired possibility. Additionally, the chemical instability issue
of the ethylene bridge would preclude the pharmaceutical devel-
opment of this TpoR agonist scaffold. Thus, a focus of our optimiza-
tion of this lead scaffold was to identify strategies to address the
chemical instability of the ethylene bridge while retaining or
improving functional potency in the agonism of the TpoR. Simulta-
neously, we were also interested in identifying structural modifica-
tions to reduce the high lipophilicity of the lead scaffold ideally
through the same bridge modifications used to address the lead’s
chemical instability. A series of ethylene bridge modifications are
shown in Table 1.
Contracting the ethylene bridge of 2 to one carbon (3) results in
a compound with only slightly reduced potency (160 nM) which
retains almost full efficacy (91%). We were very encouraged by
these results since compound 3 was significantly more chemically
stable than compound 2. For example, when heated in DMSO solu-
tion, 3 appeared to be completely stable by LC–MS analysis. In con-
trast,
2 degrades to oxidation by-products under the same
conditions. Replacement of the carboxylic acid at R4 with a primary
carboxamide slightly improves both potency (86 nM) and efficacy
(118%). The non-essential requirement of a negatively charged
group at the R4 position for both potency and efficacy mirrors
the SAR seen in our previous study.1 Since the lead molecule’s eth-
ylene bridge had been contracted to a shorter methylene bridge,
* Corresponding author. Tel.: +1 858 332 4604; fax: +1 858 812 1648.
E-mail address: TMarsilje@gnf.org (T.H. Marsilje).
Figure 1. EltrombopagTM (1) and lead structure (2).
0960-894X/$ - see front matter Ó 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmcl.2008.08.077

5260
T. H. Marsilje et al. / Bioorg. Med. Chem. Lett. 18 (2008) 5259–5262
Table 1
In vitro human TpoR functional activity of alternate bridgesa
Compounda
R1
R2
R4
R5
R6
EC50
(l
M)b
% Efficacyc
2
3
4
5
6
7
8
40-n-Butylphenyl
H
H
H
F
F
F
CO2H
CO2H
CONH2
CO2H
CO2H
CONH2
CO2H
CO2H
CO2H
CONH2
CONH2
CH2
CH2
CH2
CH2
CH2
CH2
CH2
C(CH3)2
CH2
CH2
CH2
CO2
0.111
0.160
0.086
0.037
0.026
0.277
0.111
0.332
0.449
1.43
141
91
40-n-Butylphenyl
Bond
Bond
Bond
Bond
Bond
Bond
CH2–CH2
SO2
40-n-Butylphenyl
40-n-Butylphenyl
118
151
144
105
118
71
115
83
0
30-Methyl-40-fluorophenyl
30-Methyl-40-fluorophenyl
40-n-Butylphenyl
F
9d
10
11
12
40-n-Butylphenyl
H
H
F
40-n-Butylphenyl
30-Methyl-40-fluorophenyl
30-Methyl-40-fluorophenyl
O
CH2
F
>30
EltrombopagTM (1)
0.038
127
a
All compounds demonstrated satisfactory LC–MS and 1H NMR characterization.
Ba/F3-hTpoR RGA.6,7
b
c
Relative to 30 ng/mL TPO.
d
Assay conducted with 1 mM Zn.8
the correct placement of the primary carboxamide for interaction
with the TpoR was confirmed by synthesizing the other possible
carboxamide regioisomers. This confirmed that the most potent
and efficacious carboxamide regioisomer is analog 3 (other data
not shown). The addition of a fluorine onto 3 at R2 (5) further in-
creases potency to 37 nM and increases efficacy to 151% relative
to Tpo. Replacing the R1 40-n-butylphenyl group of 5 with 40-flu-
oro-30-methylphenyl (6) results in a very potent (26 nM) and effi-
cacious (144%) compound with reduced lipophilicity (cLogP = 6.9
for 6 vs. 8.3 for 5), albeit the cLogP is still higher than desired in
a drug-like molecule. Surprisingly, the carboxamide version of this
compound (7) is significantly less potent albeit efficacious, indicat-
ing that the scaffold is less tolerant of changes to the R4 portion of
the molecule when a less lipophilic R1 group is employed. Com-
pound 8 incorporates a gem-dimethyl group in the methylene
bridge which would fully stabilize the bis-benzylic bridge position
as well as introduce a tetra-substituted center in the molecule. We
envisioned that this tetra-substituted center would further prevent
the possibility of undesired DNA intercalation of our very flat,
to the lipophilic substituents shown for R1 in Table 1, only similar
highly hydrophobic groups showed high potency. A small number
of less lipophilic aromatic heterocyclic replacements for the benze-
noid rings in 2 were attempted based upon overlapping the struc-
tures of
1 and 2. These attempts generated compounds of
significantly reduced activity. Ether and aniline substituents on
the distal benzene ring (R1 in Table 1) reduced cLogP slightly
and were partially tolerated albeit with reduced activity (data
not shown).
In addition to modifying the unstable ethylene bridge linker of
2, we also investigated analogs in which the bridge is completely
removed, i.e., 2-phenyl indoles and 2-phenyl benzimidazoles
(Table 2). When the bridge is removed from 5 to generate the 2-
phenyl indole 13, this causes only a 3ꢀ loss in potency with reten-
tion of full efficacy. The 2-phenyl indole scaffold’s potency is in-
creased to ca. 50 nM by the reintroduction of a phenolic oxygen
in the R3 position (15). Adding substituents to the R6 position (
16–18), causes a drop in activity correlated to an increase of the
steric bulk of R6. Two possible explanations for this SAR trend
are: (1) the TpoR prefers a coplanar orientation of the indole and
phenyl ring systems or (2) there is a steric clash with the TpoR in
the analogs with large R6 substituents. The former explanation par-
allels a recent SAR analysis of pyrimidine benzamide-based TpoR
agonsts.9 The benzimidazole analog 19 has a potency reduced to
the micromolar range. Interestingly, the benzimidazole salicylate
20 regains ca. 7ꢀ potency compared to 19, indicating a more pro-
nounced dependence on the salicylate moiety for potency in the
benzimidazole sub-series than in the 2-phenyl indole sub-series.
Since it had proven difficult to decrease the scaffold’s lipophil-
icity by introducing hydrophilic functional groups onto the indole
6-position (R1 in Table 1, data not shown), in addition to bridge
modifications, we also attempted to decrease the cLogP of this ser-
ies by introducing hydrophilic substituents in place of the carbox-
ylic acid in analog 5 (R4 position, Table 3). Replacing the carboxylic
acid in 5 with an acetamido group (21) resulted in an equipotent
compound which also retained full efficacy. Previous SAR had indi-
cated that a negatively charged group was not required for recep-
tor activation but it is interesting that the TpoR also tolerates
displacement of the carbonyl away from the aromatic ring. Replac-
ing the acetamido group with methanesulfonamido (22), ureido
(23), and hydroxyacetamido (24) groups is partially tolerated but
mostly aromatic scaffold as well as discourage
p-stacking based
aggregation, improving physiochemical properties of the scaffold.
We were pleased that the gem-dimethyl analog 8 retains full effi-
cacy although its potency (111 nM) drops by 3ꢀ compared to 5.
Expanding the bridge to three carbons yielded (9) which also re-
tains significant potency albeit with reduced efficacy (71%) in com-
parison to the equivalent shorter bridge linker analogs (2, 3).8
Taken together, these data indicate the tolerance of the TpoR for
different bridge linker lengths in this scaffold in terms of potency
and partially in terms of efficacy. Analogs 10–12 introduce polar
functionality into the linker bridge in an attempt to lower the lipo-
philicity of the scaffold. Introducing a sulfone for R5 (10) loses four-
fold in potency in comparison to 2. Introducing an oxygen for R5
(11) results in a drop in potency to the low micromolar range with
83% efficacy. Introducing an ester for R6 (12) results in a complete
loss of activity. In addition to the SAR described in Table 1, addi-
tional SAR exploration on the 6-position of the indole ring system
(R1 in Table 1) was performed in an attempt to decrease the scaf-
fold’s high lipophilicity. The R1 position was chosen for this SAR fo-
cus since preliminary SAR1 indicated that this portion of the
scaffold had the most steric room for modification. In general, po-
tency correlated very well with the lipophilicity of R1. In addition

T. H. Marsilje et al. / Bioorg. Med. Chem. Lett. 18 (2008) 5259–5262
5261
Table 2
In vitro human TpoR functional activity of unbridged analogsa
b
Compounda
R2
R3
R4
R5
R6
EC50
(l
M)
% Efficacyc
13
14
15
16
17
18
19
20
F
F
F
F
F
F
H
H
H
H
CO2H
CONH2
CO2H
CO2H
CO2H
CO2H
CO2H
CO2H
C
C
C
C
C
C
N
N
H
H
H
CH3
CH2CH3
CH(CH3)2
—
0.117
0.160
0.051
0.116
0.228
>1
1.26
0.179
112
81
141
119
97
70
83
121
OH
OH
OH
OH
H
OH
—
EltrombopagTM (1)
0.038
127
a
All compounds demonstrated satisfactory LC–MS and 1H NMR characterization.
Ba/F3-hTpoR RGA.6,7
Relative to 30 ng/mL TPO.
b
c
chemistry to generate the gem-dimethyl analog 8 required the syn-
thesis of the requisite gem-dimethyl ketone 31 which is shown in
Scheme 1. Commercially available 6-methoxy-indanone (28) was
reacted with titanium tetrachloride and dimethyl zinc to convert
the ketone functionality into the corresponding gem-dimethyl,
generating intermediate 29. This intermediate then underwent
benzylic oxidation with chromium trioxide, generating the gem-di-
methyl ketone intermediate 30. Attempts to convert the methyl
ether of 30 to the corresponding phenol using BBr3 were unsuc-
cessful. Instead, the phenol was produced via high temperature
microwave irradiation of ether 30 in the presence of benzenethiol.
The phenol was then converted into the penultimate nitrile ketone
intermediate 31 via conversion first to a triflate intermediate fol-
lowed by reaction with zinc cyanide under palladium catalyzed
conditions.
Table 3
In vitro human TpoR activity of alternate head groupsa
Compounda
R3
R4
TpoR EC50
(
l
M)b
% Efficacyc
5
21
H
H
H
H
H
H
NH2
NHAc
CO2H
NHAc
NHMs
NHCONH2
NHCOCH2OH
NHCOCO2H
CO2H
0.037
0.025
0.166
0.124
0.087
0.021
0.341
0.376
151
117
95
128
98
110
70
90
22d
23d
24d
25
26d
27
CO2H
The chemistry used to generate analog 12 is shown in Scheme 2.
A base promoted condensation between 5-cyanophtalide (32) and
diethyl malonate provided 33. Intermediate 33 then reacted
smoothly with 2-fluoro-3-(40-fluoro-30-methylphenyl)hydrazine
hydrochloride in ethanol without added catalyst followed by
hydrolysis with sodium hydroxide to afford 34. Finally, treatment
with HCl in dioxane at 100 °C for 5 min closed the lactone ring to
provide the final product 12.
In the case of analogs 13–18 (Table 2), the Fisher reaction was
again employed to form the indole ring system.11 The 4-n-butyl-
phenyl moiety could either be introduced as a preexisting moiety
on the hydrazine coupling partner in the Fisher reaction or alterna-
a
All compounds demonstrated satisfactory LC–MS and 1H NMR characterization.
Ba/F3-hTpoR RGA.6,7
b
c
Relative to 30 ng/mL TPO. NA, not applicable.
d
Assay conducted with 1 mM Zn.8
causes a potency loss of 3–7ꢀ. In contrast, the oxalyl amide (25) is
as active as the acetamido group (21) and it successfully reduces
the cLogP from 7.6 (21) to 6.3 (25), albeit the cLogP is still higher
than desired in a drug-like molecule. Finally, the anthranilic acid
(26) along with its acetylated derivative (27) are both ca. 10ꢀ less
potent than 5. A representative sample of scaffold analogs were
tested in the human CFU-mega assay.7 Compound 21 was the most
potent analog identified and was found to have full efficacy relative
to Tpo with an EC50 of ca. 300 nM, equivalent to the activity of the
clinically efficacious EltrombopagTM.
With the exception of 12, the chemistry used to assemble the
compounds in Table 1 is a Fisher indole synthesis reaction between
the appropriate ketone and hydrazine in acetic acid at 100 °C over-
night promoted by ZnCl2.10 In the cases where R4 is a carboxylic
acid, the corresponding ester was used for the Fisher indole forma-
tion and subsequently hydrolyzed with LiOH to generate the final
carboxylic acid product. Likewise, in the cases where R4 is an
amide, the corresponding nitrile was used for the Fisher indole for-
mation and subsequently hydrolyzed with LiOH to an amide in a
final step. With the exception of the gem-dimethyl analog 8, all
of the required ketone and hydrazine synthetic fragments were
either known in the literature or synthesized via standard syn-
thetic manipulation of known literature intermediates.10 The
Scheme 1. Reagents and conditions: (a) titanium tetrachloride (2 equiv), dimethyl
zinc (6 equiv), anhyd CH2Cl2/toluene, ꢁ40 °C to rt, 3 h; (b) chromium trioxide (2.3
equiv), 2:1 AcOH/H2O, 0 °C to rt, 3 h; (c) i—benzenethiol (1 equiv), K2CO3
(0.1 equiv), NMP, 220 °C (microwave), 30 min, ii—triethylamine (3 equiv), N-
phenyltrifluoromethanesulfonimide (1.1 equiv), anhyd CH2Cl2, ꢁ78 °C to rt, 12 h,
iii—ZnCN2 (2 equiv), palladium bis(tri-tert-butyl phosphine) (5 mol%), 2:1 THF/
NMP, 120 °C, 3 h.

5262
T. H. Marsilje et al. / Bioorg. Med. Chem. Lett. 18 (2008) 5259–5262
Scheme 4. Reagents and conditions: (a) i—4-butylphenylboronic acid (2 equiv), CsF
(3 equiv), ((t-Bu)3P)2Pd (5%), dioxane, 120 °C, 15 min; ii—H2, 10% Pd/C, 2:1 MeOH/
EtOAc; (b) i—aromatic aldehyde (1 equiv), NaHSO3 (1.5 equiv), DMA, 140 °C, 1 h;
ii—LiOH (5 equiv), EtOH, THF, 165 °C, 5 min.
Scheme 2. Reagents and conditions: (a) diethyl malonate (1.5 equiv), NaOMe
(3 equiv), MeOH, reflux, 5.5 h; (b) (i) (2,40-difluoro-30-methylbiphenyl-3-yl)-
hydrazine hydrochloride (0.8 equiv), EtOH, 160 °C, 10 min, (ii) excess NaOH,
80 °C, 15 min; (c) EtOH/4 M HCl in dioxane (1:1), 100 °C, 5 min.
tively it could be installed after the Fisher reaction via a Suzuki
reaction on the corresponding 6-halo synthetic intermediate. All
of the precursor ketones were known in the literature or commer-
cially available except the ketones required for 16–18 that were
made as shown in Scheme 3. The known12 aldehyde 35 was treated
with 2 equivalents of the requisite Grignard reagent in THF and
then oxidized to the ketone 36 using PDC.
Scheme 5. Reagents and conditions: (a) i—dibutylvinylboronic ester (1.5 equiv),
Pd(PPh3)2Cl2 (4%), Na2CO3 (7 equiv), 4:1 THF/H2O, 80 °C, overnight; (b) RuCl3.H2O
(5%), NaIO4 (4 equiv), CCl4, 3:2 H2O/MeCN, 50 °C, 2 h.
The chemistry to generate analog 20 is shown in Scheme 4. The
bromide 37 was coupled with n-butylphenyl boronic acid under
Suzuki conditions, followed by hydrogenation to afford the dia-
mine 38. This diamine was then oxidatively coupled with the
appropriate aldehyde ester followed by ester saponification to af-
ford the carboxylic acid analogs 19 and 20.
tify molecules with good in vitro activity and chemical stability,
the high cLogP demanded by this scaffold’s TpoR binding pocket
interactions generated compounds with physicochemical proper-
ties (e.g., extremely low aqueous solubility) which precluded this
series of analogs from further pharmaceutical development.
In the case of analogs 26 and 27, the requisite ketone for the
Fisher reaction was prepared as shown in Scheme 5. The known13
ketone 39 was elaborated to the styrene 40 using Suzuki chemistry
and then oxidized to carboxylic acid 41 with RuO4.
References and notes
1. Alper, P. B.; Marsilje, T. H.; Mutnick, D.; Lu, W.; Chatterjee, A.; Roberts, M. J.; He,
Y.; Karanewsky, D. S.; Chow, D.; Lao, J.; Gerken, A.; Tuntland, T.; Liu, B.; Chang,
J.; Gordon, P.; Seidel, H. M.; Tian, S. S. Bioorg. Med. Chem. Lett. 2008, 18, 5255.
2. Revill, P.; Serradell, N.; Bolos, J. Drugs Future 2006, 31, 767.
3. Bussel, J. B.; Cheng, G.; Saleh, M. N.; Psaila, B.; Kovaleva, L.; Meddeb, B.; Kloczko,
J.; Hassani, H.; Mayer, B.; Stone, N. L.; Arning, M.; Provan, D.; Jenkins, J. M. N.
Engl. J. Med. 2007, 357, 2237.
In conclusion, we designed and evaluated a series of TpoR ago-
nists derived from
a lead dihydro-benzo[a]carbazole scaffold
which had been deemed unsuitable for pharmaceutical develop-
ment due to inherent chemical instability.1 The instability of the
lead 2 was successfully addressed by compounds which also dem-
onstrated improved potency compared to 2. Members of the scaf-
fold are full agonists of the human TpoR that demonstrate
functional cellular potency <50 nM with analog 21 having equiva-
lent activity in the primary human CFU-mega assay compared to
the clinically efficacious EltrombopagTM. An examination of repre-
sentative lead compounds in a cross-selectivity screen against
4. Jenkins, J. M.; Williams, D.; Deng, Y.; Uhl, J.; Kitchen, V.; Collins, D.; Erickson-
Miller, C. L. Blood 2007, 109, 4739.
5. McHutchison, J. G.; Dusheiko, G.; Shiffman, M. L.; Rodriguez-Torres, M.; Sigal,
S.; Bourliere, M.; Berg, T.; Gordon, S. C.; Campbell, F. M.; Theodore, D.;
Blackman, N.; Jenkins, J.; Afdhal, N. H.; Bronowicki, J. P.; DeLedinghen, V.;
Dumortier, J.; Encke, J.; Germanidis, G.; Lawitz, E.; Marcellin, P.; Mills, P.;
Poupon, R.; Rustgi, V.; Teuber, G.; Tran, A.; Zarski, J. P. N. Engl. J. Med. 2007, 357,
2227.
6. Erickson-Miller, C. L.; DeLorme, E.; Tian, S.-S.; Hopson, C. B.; Stark, K.; Giampa,
L.; Valoret, E. I.; Duffy, K. J.; Luengo, J. L.; Rosen, J.; Miller, S. G.; Dillon, S. B.;
Lamb, P. Exp. Hematol. (New York) 2005, 33, 85.
7. Duffy, K. J.; Darcy, M. G.; Delorme, E.; Dillon, S. B.; Eppley, D. F.; Erickson-Miller,
C.; Giampa, L.; Hopson, C. B.; Huang, Y.; Keenan, R. M.; Lamb, P.; Leong, L.; Liu,
N.; Miller, S. G.; Price, A. T.; Rosen, J.; Shah, R.; Shaw, T. N.; Smith, H.; Stark, K.
C.; Tian, S. S.; Tyree, C.; Wiggall, K. J.; Zhang, L.; Luengo, J. I. J. Med. Chem. 2001,
44, 3730.
8. Compounds 9, 22–24, and 26 were assayed in the presence of 1 mM Zn rather
than the 25 mM Zn levels present for the other compounds discussed in this
communication. We have noted that there is usually no more than a two-fold
difference between data collected under the two different conditions (data not
shown).
>50 other receptors showed IC50 > 1 lM for all of the receptors
evaluated, indicating that the observed TpoR activity is not simply
due to non-specific hydrophobic binding or aggregation which is a
concern for highly lipophilic lead structures. Clear SAR trends also
support specific ligand interactions with a binding pocket in the
human Tpo receptor. Although we were able to successfully iden-
9. Reiter, L. A.; Jones, C. S.; Brissette, W. H.; McCurdy, S. P.; Abramov, Y. A.;
Bordner, J.; DiCapua, F. M.; Munchhof, M. J.; Rescek, D. M.; Samardjiev, I. J.;
Withka, J. M. Bioorg. Med. Chem. Lett. 2008, 18, 3000.
10. Alper, P.; Marsilje, T.; Chatterjee, A.; Lu, W.; Mutnick, D.; Roberts, M. J.; He, Y.
WO2007009120, 2007.
11. Marsilje, T.; Lu, W.; Alper, P.; Mutnick, D.; He, Y. WO2007022269, 2007.
12. Dyke, S. F.; White, A. W. C.; Hartley, D. Tetrahedron 1973, 29, 857.
13. Fouad, F. S.; Crasto, C. F.; Lin, Y.; Jones, G. B. Tetrahedron Lett. 2004, 45, 7753.
Scheme 3. Reagents and conditions: (a) i—RMgCl (2 equiv), THF, ꢁ78 °C, 1 h, 25 °C,
1 h; ii—CH2Cl2, PDC (1.2 equiv).