A hit deconstruction approach for the discovery of fetal hemoglobin inducers

Article abstract of DOI:10.1016/j.bmcl.2018.10.032

Beta-hemoglobinopathies such as sickle cell disease represent a major global unmet medical need. De-repression of fetal hemoglobin in erythrocytes is a clinically validated approach for the management of sickle cell disease, but the only FDA-approved medicine for this purpose has limitations to its use. We conducted a phenotypic screen in human erythroid progenitor cells to identify molecules with the ability to de-repress fetal hemoglobin, which resulted in the identification of the benzoxaborole-containing hit compound 1. This compound was found to have modest cellular potency and lead-like pharmacokinetics, but no identifiable SAR to enable optimization. Systematic deconstruction of a closely related analog of 1 revealed the fragment-like carboxylic acid 12, which was then optimized to provide tetrazole 31, which had approximately 100-fold improved cellular potency compared to 1, high levels of oral exposure in rats, and excellent solubility.

Full text of DOI:10.1016/j.bmcl.2018.10.032

Bioorganic&MedicinalChemistryLettersxxx(xxxx)xxx–xxx
Contents lists available at ScienceDirect
Bioorganic & Medicinal Chemistry Letters
journal homepage: www.elsevier.com/locate/bmcl
A hit deconstruction approach for the discovery of fetal hemoglobin
inducers
⁎
Andrew B. Benowitza, , H. Christian Eberlb, Connie L. Erickson-Millera, Aidan G. Gilmartina,
Elizabeth R. Gorea, Monica N. Montoutea, Zining Wua
a
b
GlaxoSmithKline, 1250 South Collegeville Rd., Collegeville, PA 19426, USA
Cellzome, a GSK Company, Meyerhofstrasse 1, 69117 Heidelberg, Germany
A R T I C L E I N F O
A B S T R A C T
Keywords:
Beta-hemoglobinopathies such as sickle cell disease represent a major global unmet medical need. De-repression
of fetal hemoglobin in erythrocytes is a clinically validated approach for the management of sickle cell disease,
but the only FDA-approved medicine for this purpose has limitations to its use. We conducted a phenotypic
screen in human erythroid progenitor cells to identify molecules with the ability to de-repress fetal hemoglobin,
which resulted in the identification of the benzoxaborole-containing hit compound 1. This compound was found
to have modest cellular potency and lead-like pharmacokinetics, but no identifiable SAR to enable optimization.
Systematic deconstruction of a closely related analog of 1 revealed the fragment-like carboxylic acid 12, which
was then optimized to provide tetrazole 31, which had approximately 100-fold improved cellular potency
compared to 1, high levels of oral exposure in rats, and excellent solubility.
Sickle cell anemia
Beta-thalassemia
Fetal hemoglobin
Benzoxaborole
CPSF
Hemoglobin is the oxygen-carrying component of red blood cells. In
adults, it is a structural complex normally composed of two pairs of
alpha- and beta-globin proteins plus four molecules of heme. Beta-he-
moglobinopathies, such as sickle cell anemia and beta-thalassemia, are
rare diseases in the United States and in European countries,1 but they
are the most common Mendelian diseases in the world. Epidemiological
data indicates that approximately 300,000 infants are born each year
with a beta-hemoglobinopathy worldwide,2 and the prevalence of in-
fants born with a beta-hemoglobinopathy is predicted to increase to
more than 400,000 per year by 2050.3
levels of fetal hemoglobin (HbF), which is normally repressed after
birth,4 is phenotypically protective against the sickle cell homozygous
genotype.5 The ribonucleotide reductase inhibitor hydroxyurea was the
first FDA approved medicine for the specific treatment of sickle cell
anemia, and it has been shown to raise HbF levels in erythrocytes and to
reduce the frequency of sickle crisis events.6 In contrast, there is less
clinical experience with hydroxyurea in intermediate and major beta-
thalassemia patients, but the disease biology of beta-thalassemia sug-
gests that elevating HbF levels sufficiently may also be beneficial in
these patients.7
Beta-hemoglobinopathies are characterized by disorders of the beta-
globin component of hemoglobin. In sickle cell anemia, the gene en-
coding for beta-globin contains an E6V mutation which results in ab-
normal hemoglobin structure and causes deoxygenated red blood cells
to adopt a characteristic sickle shape. This abnormal shape combined
with increased adherence to the vascular endothelium leads to reduced
red cell plasticity, longer capillary transit times, and frequent vaso-oc-
clusive events that damage tissues. In contrast, beta-thalassemia is
characterized by inadequate beta-globin production to combine with
normally produced alpha-globin. The resulting accumulation of tetra-
meric alpha-globin is toxic to red cell precursors, and results in red cell
hemolysis and ineffective erythropoiesis.
Despite the importance of hydroxyurea for the treatment of sickle
cell patients, there are important limitations to its use. Significantly,
there is large variability in patient response to hydroxyurea, and many
adult patients have inadequate HbF increases upon treatment to pro-
vide effective clinical disease management.8
In an effort to identify small molecule inducers of fetal hemoglobin
that could be useful for the treatment of beta-hemoglobinopathy pa-
tients, we conducted a phenotypic screen in human erythroid pro-
genitor cells (EPCs) derived from bone marrow CD34+ hematopoietic
stem cells.9,10 Due to the inherent cost and logistical challenges asso-
ciated with conducting a dose-response screen with human primary
cells, we were limited to screening a curated library of approximately
8000 compounds, from which we focused on one confirmed hit
Evidence from human genetics indicates that elevated erythrocyte
⁎
Corresponding author.
E-mail address: andrew.b.benowitz@gsk.com (A.B. Benowitz).
https://doi.org/10.1016/j.bmcl.2018.10.032
Received 30 August 2018; Received in revised form 17 October 2018; Accepted 20 October 2018
0960-894X/©2018ElsevierLtd.Allrightsreserved.
Pleasecitethisarticleas:Benowitz,A.B.,Bioorganic&MedicinalChemistryLetters,https://doi.org/10.1016/j.bmcl.2018.10.032

A.B. Benowitz et al.
Bioorganic&MedicinalChemistryLettersxxx(xxxx)xxx–xxx
Table 1
Table 3
Activity and in vivo rat pharmacokineticsa of 1.
Effect of fluorine substitution on reversed amide benzoxaborole potency.
Cmax (IV/PO)b
Clc
Vdss
1.7
T1/2 (IV/PO)e
6.4/4.5
DNAUC (IV/PO)f
4210/1120
%Fg
d
Compound
R1
R2
R3
pEC50
FaSSIF solubilitya
792/1420
a
4.1
27%
n = 2 (IV; 1 mg/kg), n = 2 (PO; 5 mg/kg), Sprague-Dawley rats, non-
crossover design. Reported values are the average of two animals.12
4
5
6
7
H
F
H
H
F
H
H
H
F
5.99
7.04
< 4.5
5.67
3
4
b
c
d
H
H
Not done
8
Maximum observed concentration (ng/mL).
H
Blood clearance (mL/min/kg).
Volume of distribution at steady state (L/kg).
a
Four hour fasted state simulated intestinal fluid solubility (μg/mL).
e
f
Half-life (h).
Area under the blood concentration vs. time curve, normalized for dose
similar potency as 4 (i.e., 7).
(h*kg*ng/mL/mg).
g
Since both compounds 1 and 4 appeared to be largely intolerant of
substitutions and modifications, we sought to determine the minimum
pharmacophore required for activity in our HbF induction cellular assay
to identify a scaffold more amenable to structural modification as well
as to improve the solubility. To this end, we embarked upon a sys-
Bioavailability.
(compound 1, Table 1). This compound was found to reproducibly in-
duce the production of HbF in an ELISA assay9 employing human EPCs
11
with a pEC50 of 5.57, and was found to have dose-normalized IV
tematic deconstruction of the Ames-negative compound
4 while
exposure in rats of 4210 h*kg*ng/mL/mg, clearance of 4.1 mL/min/kg,
an IV half-life of 6.4 h, and 27% oral bioavailability.12
maintaining the benzoxaborole moiety (Scheme 1), since we reasoned
that this moiety seemed to be essential for HbF induction based on the
results observed with 2 and 3.
Although we did not know the target or mechanism by which
compound 1 induced HbF, we speculated that the HbF induction ac-
tivity in EPCs and lead-like pharmacokinetics in rats of 1 indicated that
this compound would be an attractive starting point for optimization.
However, in practice we found that very few substitutions and struc-
tural modifications of this compound were tolerated without causing a
dramatic loss of HbF induction in human EPCs. Particularly noteworthy
was the observation that the benzoxaborole13 moiety was essential for
activity, as the boronic acid compounds 2 and 3 (Table 2) were not
found to induce HbF at all concentrations up to 33 μM, which was the
highest concentration tested in our assay.
Although systematic reduction of the complexity of 4 mostly re-
sulted in compounds that did not induce HbF when tested in a dose-
response manner up to the highest concentration of our cellular assay,
the fragment-like carboxylic acid 12 was an exception. Compound 12
was found to be highly soluble, and to reproducibly induce HbF with an
average pEC50 of 5.03 (Table 4), which is less than a 10-fold reduction
in cellular HbF induction potency when compared with the sub-
stantially larger and more complex compound 4. It is noteworthy that
the corresponding amide 11 was not found to induce HbF, suggesting
that the carboxylic acid present in 12 is important for activity. How-
ever, the carboxylic acid-containing compound 10 was also not ob-
served to induce HbF, suggesting that the spacing and conformation
between the carboxylic acid and the aromatic benzoxaborole system
may also be important.
Although virtually all substitutions and modifications to 1 that we
attempted were found to result in compounds that did not induce HbF
up to the highest concentration tested, we observed that the reversed
amide 4 (Table 3) was approximately equipotent with 1 in our HbF
cellular induction assay. This finding was significant since 4 lacks the
benzoxaborole aniline substructure present in 1, which carries a mu-
tagenicity risk, and we were pleased to discover that 4 was negative in
an Ames reverse bacterial mutation panel assessment both in the pre-
sence and absence of metabolic activation.14 However, 4 was found to
have very poor fasted state simulated intestinal fluid solubility of 3 μg/
mL after four hours, and attempted optimization of this compound was
largely unsuccessful in that with the exception of the more potent 7-
fluoro compound 5, most analogs prepared with even conservative
changes were inactive (i.e., 6) with very few compounds found to have
To test this hypothesis, we examined a series of benzoxaborole
carboxylic acids to determine the effects both of spacing between these
moieties, as well as of conformational restraint (Table 4). Increasing the
linker length between the carboxylic acid and the benzoxaborole
moiety via insertion of either a methylene (13) or an ethylene (14)
spacer resulted in an approximately 10-fold improvement in cellular
Table 2
Effect of boronic acid modifications on potency.
Compound
R
pEC50
2
3
H
< 4.5
< 4.5
Scheme 1. Minimum pharmacophore identification through deconstruction of
compound 4.
CH3
2

A.B. Benowitz et al.
Bioorganic&MedicinalChemistryLettersxxx(xxxx)xxx–xxx
Table 4
Table 6
Effects of linker length and conformational restraint on HbF induction potency.
Effects of carboxylic acid isosteres on HbF induction potency.
Compound
Structure
pEC50
7.19
FaSSIF
solubilitya
23
900
Compound
X
pEC50
FaSSIF solubilitya
24
6.08
> 1000
12
13
14
15
bond
5.03
5.97
5.95
4.94
950
–CH2–
> 1000
> 1000
> 1000
–CH2CH2–
25
26
7.74
7.82
700
880
16
6.50
510
a
Four hour fasted state simulated intestinal fluid solubility (μg/mL).
potency, although there was essentially no difference in potency be-
tween these lengths. Insertion of a 2-pyrrole spacer (15) did not in-
crease the HbF induction potency, but insertion of a 3-pyrrole spacer
(16) resulted in an approximately 30-fold increase in HbF cellular in-
duction potency compared to compound 12.
(rac)-27
7.57
6.96
> 1000
980
28
The result observed for 16 provided additional evidence that both
the linker length and conformation between the carboxylic acid and the
benzoxaborole moiety are important for HbF induction potency.
However, despite the encouraging cellular potency observed with 16,
this compound was found to have poor oral exposure in Sprague-
Dawley rats (PO DNAUC 186 h*kg*ng/mL/mg, F = 26%; n = 2, non-
crossover design).12
29
30
6.89
6.14
> 1000
> 1000
To further evaluate the influence of the linker on HbF induction
potency, we next prepared a series of compounds in which the linker
was substituted with a small selection of alkyl groups (Table 5).
We observed that disubstitution of the linker with two methyl
groups (compound 17) afforded an approximately 10-fold improvement
in HbF cellular potency compared to 13, but that the addition of a
single methyl substituent to the linker to provide (rac)-18 resulted in an
even greater increase in HbF induction potency. Addition of an ethyl
group to give (rac)-21 or an isopropyl group to give (rac)-22 did not
further increase the potency over (rac)-18. Isolation and evaluation of
each enantiomer of (rac)-18 revealed that the (R)-enantiomer 19 was
more than 10-fold more potent than the (S)-enantiomer 20, with an
HbF induction EC50 of 40 nM in human erythroid progenitor cells.
We next explored whether we could obtain further HbF induction
31
32
33
7.58
8.04
6.56
930
> 1000
> 1000
Table 5
Effects of linker substitution on HbF potency.
a
Four hour fasted state simulated intestinal fluid solubility (μg/mL).
potency improvements by replacing the carboxylic acid moiety with a
suitable isostere. Carboxylic acid bioisosteres have been reviewed
elsewhere,15,16 and we observed that both acylsulfonamide (com-
pounds 23–28) and tetrazole (compounds 29–33) carboxylic acid re-
placements provided compounds with high HbF cellular induction po-
tency and high solubility (Table 6).
Compound
R1
R2
pEC50
FaSSIF solubilitya
13
H
H
5.97
6.94
7.32
7.40
6.06
7.30
6.19
> 1000
> 1000
> 1000
> 1000
950
We found that the simple aryl sulfonamide 23 as well as the cy-
cloalkyl sulfonamides 25 and 26 had higher cellular HbF induction
potency compared to the corresponding carboxylic acid 13. This was
also observed to be the case for the unsubstituted tetrazole compound
32, which was discovered to induce HbF with an EC50 of less than
10 nM in human erythroid progenitor cells. In contrast to the potency
17
CH3
CH3
(rac)-18
19
CH3/H
CH3
CH3/H
H
20
H
CH3
(rac)-21
(rac)-22
CH2CH3/H
i-Pr/H
CH2CH3/H
i-Pr/H
850
> 1000
a
Four hour fasted state simulated intestinal fluid solubility (μg/mL).
3

A.B. Benowitz et al.
Bioorganic&MedicinalChemistryLettersxxx(xxxx)xxx–xxx
Table 7
displaced by an inactive soluble compound 36 (pEC50 < 4.5). By this
method the highest affinity interacting proteins were determined to be
INTS9, INTS11, and CPSF3, which are members of the related In-
tegrator and CPSF (cleavage and polyadenylation specific factor)
complexes. Both INTS11 and CPSF3 have conserved metallo-beta-lac-
tamase (MBL) domains that serve as RNA endonucleases, functioning in
the 3′-end cleavage and maturation of snRNA and mRNAs respec-
tively.17 Notably, multiple MBL containing proteins in prokaryotes and
fungi have previously been identified as the targets of some benzox-
aborole containing compounds.18–20
In vivo PO rat pharmacokineticsa of selected compounds.
Cmaxc
T1/2
DNAUCe
%Ff
d
Compound
PO dose
19
5 mg/kg
10 mg/kg
5 mg/kg
100 mg/kg
5 mg/kg
28 mg/kg
5 mg/kg
10 mg/kg
5 mg/kg
5 mg/kg
100 mg/kg
1863
2710
1727
47,250
1287
52,967
1177
3880
283
16.6
3.03
1.14
4.53
7.29
3.88
1.05
1.53
3.67
1.30
3.54
429
283
451
683
560
2120
612
1049
103
38
83%
19b
20
55%
88%
20
> 100%
28%
23
23b
31
> 100%
68%
31b
32
> 100%
24%
Although we lack direct biochemical evidence for the inhibition of
these complexes by the described compounds, it is notable that muta-
tions in the CPSF binding sites for the alpha- or beta-globin genes have
been observed to significantly change their RNA stability and to result
in thalassaemias.21,22 These mutations affect the stability and transla-
tion efficiency of the globin mRNAs, and suggest a potential role for the
CPSF complex in impacting globin protein expression. Notably, in the
course of writing this manuscript, related compounds have been iden-
tified as inhibitors of CPSF3 from T. gondii.23 Work is ongoing to un-
derstand how these effects on the CPSF and Integrator complexes might
ultimately result in the observed cellular increases in gamma globin
protein as well as the observed in vivo toxicity in rats.
33
71
13%
33b
4903
184
62%
a
n = 3, Sprague-Dawley rats, non-crossover design (IV not shown).
Reported values are the average of three animals.12
b
c
d
e
In vivo toxicity observed.
Maximum observed concentration (ng/mL).
Half-life (h).
Area under the blood concentration vs. time curve, normalized for dose
(h*kg*ng/mL/mg).
f
Bioavailability.
Acknowledgments
The authors gratefully acknowledge the assistance of M. S. Jiang
and W. Xie for cellular assay screening support, S. Lehmann for pro-
teomics assay support, M. Boesche, T. Rudi, M. Kloes-Hudak, and K.
Kammerer for assistance with mass spectrometry, and W. Zou (BioDuro,
LLC) for synthetic chemistry support.
Appendix A. Supplementary data
Fig. 1. Compounds used for chemoproteomic analysis.
Full synthetic routes, experimental procedures, and proteomics data
are available in the Supplementary Data, which can be found online at
https://doi.org/10.1016/j.bmcl.2018.10.032.
improvements observed via linker substitution of carboxylic acid 13 to
give both 17 and (rac)-18, linker substitution of acylsulfonamide 25
and tetrazole 32 to provide (rac)-27 and 31 respectively did not result
in any increase in HbF induction potency.
References
In vivo pharmacokinetic profiling of representative exemplars re-
vealed variable oral exposure in rats (Table 7). For example, carboxylic
acid compounds 19 and 20, and tetrazole compound 31 had moderate
to high oral bioavailability and dose-normalized exposure in rats when
dosed orally at 5 mg/kg. In contrast, acylsulfonamide compound 23,
and tetrazole compounds 32 and 33 all had poor oral exposure when
dosed orally at 5 mg/kg. However, dose escalation revealed substantial
non-linear PK as well as serious in vivo toxicity upon dosing many of
these compounds. Although these compounds were all tolerated upon
5 mg/kg dosing, we observed serious adverse events upon dosing 19 at
10 mg/kg, 23 at 28 mg/kg, 31 at 10 mg/kg, and 33 at 100 mg/kg. In
contrast, 20 was well-tolerated up to and including 100 mg/kg which
was the highest dose tested. Note that tetrazole 32 was only dosed at
5 mg/kg.
1. Hassell KL. Population estimates of sickle cell disease in the U.S. Am J Prev Med.
2010;38(4 Suppl.):S512–S521.
2. Weatherall DJ. The inherited diseases of hemoglobin are an emerging global health
burden. Blood. 2010;115:4331–4336.
3. Piel FB, Hay SI, Gupta S, Weatherall DJ, Williams TN. Global burden of sickle cell
anaemia in children under five, 2010–2050: modelling based on demographics, ex-
cess mortality, and interventions. PLoS Med. 2013;10:e1001484.
4. Wood WG. Haemoglobin synthesis during human fetal development. Br Med Bull.
1976;32:282–287.
5. Weatherall DJ, Clegg JB. Hereditary persistence of fetal haemoglobin. Br J Haematol.
1975;29:191–198.
6. Charache S, Terrin ML, Moore RD, et al. Effect of hydroxyurea on the frequency of
painful crises in Sickle cell anemia. N Engl J Med. 1995;332:1317–1322.
7. Musallam KM, Taher AT, Cappellini MD, Sankaran VG. Clinical experience with fetal
hemoglobin induction therapy in patients with β-thalassemia. Blood.
2013;121:2199–2212.
8. Steinberg MH, Lu ZH, Barton FB, Terrin ML, Charache S, Dover GJ. Fetal hemoglobin
in sickle cell anemia: determinants of response to hydroxyurea. Blood.
1997;89:1078–1088.
Since these compounds had all been identified through phenotypic
cellular testing, we speculated that identification of the putative target
or targets of these compounds might allow us to better understand the
potential for separating the desired HbF induction from the observed
undesired in vivo toxicity.
9. Li H, Xie W, Gore ER, et al. Development of phenotypic screening assays for γ-globin
induction using primary human bone marrow day 7 erythroid progenitor cells. J
Biomol Screen. 2013;18:1212–1222.
10. Human biological samples were sourced ethically and their research use was in ac-
cord with the terms of the informed consents.
11. Defined as the negative log of the molar concentration at which 50% maximal in-
duction of HbF was observed for a given compound.
To that end, chemoproteomic analysis was conducted in EPC lysates
using compound 34 (Fig. 1), which was immobilized onto activated
sepharose beads via an amide bond through the primary amine moiety.
We tested the synthetic intermediate carbamate compound 35 (pEC50
6.04) to confirm that the bead-linked amide compound would likely
still be active in our EPC HbF cellular induction assay. Proteins binding
specifically to the chemoproteomic ligand were competed off with an
active soluble compound 24 (pEC50 6.08), but were minimally
12. All animal studies were conducted in accordance with the GSK Policy on the Care,
Welfare and Treatment of Laboratory Animals and were reviewed by the Institutional
Animal Care and Use Committee either at GSK or by the ethical review process at the
institution where the work was performed.
13. Liu CT, Tomsho JW, Benkovic SJ. The unique chemistry of benzoxaboroles: current
and emerging applications in biotechnology and therapeutic treatments. Bioorg Med
Chem. 2014;22:4462–4473.
14. Escobar PA, Kemper RA, Tarca J, et al. Bacterial mutagenicity screening in the
pharmaceutical industry. Mutation Res – Rev Mutation Res. 2013;752:99–118.
4

A.B. Benowitz et al.
Bioorganic&MedicinalChemistryLettersxxx(xxxx)xxx–xxx
15. Ballatore C, Huryn DM, Smith III AB. Carboxylic acid (bio)isosteres in drug design.
ChemMedChem. 2013;8:385–395.
20. Sonoiki E, Ng CL, Lee MCS, et al. A potent antimalarial benzoxaborole targets a
Plasmodium falciparum cleavage and polyadenylation specificity factor homologue.
Nat Commun. 2017:8.
16. Lassalas P, Gay B, Lasfargeas C, et al. Structure property relationships of carboxylic
acid isosteres. J Med Chem. 2016;59:3183–3203.
21. Higgs DR, Goodbourn SEY, Lamb J. α-Thalassaemia caused by a polyadenylation
signal mutation. Nature. 1983;306:398–400.
17. Matera AG, Wang Z. A day in the life of the spliceosome. Nat Rev Mol Cell Biol.
2014;15:108–121.
22. Orkin SH, Cheng TC, Antonarakis SE, Kazazian Jr HH. Thalassemia due to a mutation
in the cleavage-polyadenylation signal of the human beta-globin gene. EMBO J.
1985;4:453–456.
18. Hernandez V, Crépin T, Palencia A, et al. Discovery of a novel class of boron-based
antibacterials with activity against gram-negative bacteria. Antimicrob Agents
Chemother. 2013;57:1394–1403.
23. Palencia A, Bougdour A, Brenier-Pinchart MP, et al. Targeting Toxoplasma gondii
CPSF3 as a new approach to control toxoplasmosis. EMBO Mol Med. 2017;9:385–394.
19. Rock FL, Mao W, Yaremchuk A, et al. An antifungal agent inhibits an aminoacyl-
tRNA synthetase by trapping tRNA in the editing site. Science. 2007;316:1759–1761.
5