9H-Carbazole-1-carboxamides as potent and selective JAK2 inhibitors

Article abstract of DOI:10.1016/j.bmcl.2015.04.101

The discovery, synthesis, and characterization of 9H-carbazole-1-carboxamides as potent and selective ATP-competitive inhibitors of Janus kinase 2 (JAK2) are discussed. Optimization for JAK family selectivity led to compounds 14 and 21, with greater than 45-fold selectivity for JAK2 over all other members of the JAK kinase family.

Full text of DOI:10.1016/j.bmcl.2015.04.101

Bioorganic & Medicinal Chemistry Letters xxx (2015) xxx–xxx
Contents lists available at ScienceDirect
Bioorganic & Medicinal Chemistry Letters
journal homepage: www.elsevier.com/locate/bmcl
9H-Carbazole-1-carboxamides as potent and selective JAK2 inhibitors
a
a
a
a
Kurt Zimmermann a, , Xiaopeng Sang , Harold A. Mastalerz , Walter L. Johnson , Guifen Zhang ,
Qingjie Liu b, Douglas Batt b, Louis J. Lombardo b, Dinesh Vyas a, George L. Trainor b, John S. Tokarski b,
Matthew V. Lorenzi b, Dan You b, Marco M. Gottardis b, Jonathan Lippy b, Javed Khan b, John S. Sack b,
Ashok V. Purandare b
⇑
a Bristol-Myers Squibb Co., 5 Research Parkway, Wallingford, CT 06492-1951, USA
b Bristol-Myers Squibb Co., Province Line Rd., Princeton, NJ 08540-4000, USA
a r t i c l e i n f o
a b s t r a c t
Article history:
The discovery, synthesis, and characterization of 9H-carbazole-1-carboxamides as potent and selective
ATP-competitive inhibitors of Janus kinase 2 (JAK2) are discussed. Optimization for JAK family selectivity
led to compounds 14 and 21, with greater than 45-fold selectivity for JAK2 over all other members of the
JAK kinase family.
Received 20 March 2015
Revised 28 April 2015
Accepted 30 April 2015
Available online xxxx
Ó 2015 Published by Elsevier Ltd.
Keywords:
JAK2
JAK
JAK family selectivity
JAK1
JAK3
TYK2
Myeloproliferative disorders
Myeloproliferative neoplasms
JAK2 is a member of the Janus kinase family of non-receptor
tyrosine kinases.1 The other members in the family are JAK1,
JAK3 and TYK2. Recently a gain-of-function point mutation in the
pseudokinase domain (JH2) of JAK2 was discovered, where valine
617 is replaced with phenylalanine (JAK2V617F) in myeloprolifera-
tive neoplasm (MPN) patients. This mutation was found in a
majority of patients with polycythemia vera (PV) and
approximately 50% of the patients suffering from essential
thrombocythemia (ET) and myelofibrosis (MF).2,3 The mechanism
by which this and other JAK-STAT pathway mutations cause
growth-factor independent activation is still under investigation.
This discovery generated considerable interest in targeting JAK2
kinase as a potential treatment for MPNs. Prior to the introduction
of the first JAK2 inhibitor, ruxolitinib, only palliative treatment
options existed for MPNs. Ruxolitinib and the more recently
reported momelotinib have shown modest activity in relieving
constitutional symptoms and can temporarily improve the quality
of life for patients with myelofibrosis. However, the side effects of
these drugs result in high discontinuation rates.4 JAK2 inhibitors
also have the potential to treat other malignancies and
inflammatory diseases that are caused by the aberrant activation
of the JAK2/STAT signaling pathway.5–8 A safe, orally bioavailable,
small molecule, JAK2 inhibitor could significantly change the
course of these diseases.6 Since other members of the JAK family
are involved in regulation of immune functions, their inhibition
would likely cause undesired immunosuppression and would need
to be avoided.7 We present 9H-carbazole-1-carboxamides as novel,
ATP-competitive JAK2 inhibitors with selectivity against other JAK
family members.9
Screening of our internal compound collection resulted in the
identification of 1 as a hit with good JAK2 potency but modest
selectivity against other JAK family kinases and poor aqueous
solubility.10
The X-ray crystal structure of 1 bound to the JAK2 kinase
domain showed the inhibitor bound to the hinge region of the pro-
tein via a donor–acceptor–donor triad of hydrogen bonds, namely
from the ring-NH to the carbonyl of Leu932, from the amide NH of
Leu932 to the primary amide carbonyl, and from the amide NH of 1
to the carbonyl of Glu930 (Fig. 1).11 The other NH of the primary
amide is involved in a water-mediated hydrogen bond to the
amide carbonyl of Gly993, an interaction that may contribute to
the selectivity over JAK3 which has an Ala in this position. The
Ala-Asp amide in JAK3 structures is flipped 180° compared to the
⇑
Corresponding author. Tel.: +1 203 6777236; fax: +1 203 6777702.
E-mail address: Kurt.Zimmermann@bms.com (K. Zimmermann).
http://dx.doi.org/10.1016/j.bmcl.2015.04.101
0960-894X/Ó 2015 Published by Elsevier Ltd.
Please cite this article in press as: Zimmermann, K.; et al. Bioorg. Med. Chem. Lett. (2015), http://dx.doi.org/10.1016/j.bmcl.2015.04.101

2
K. Zimmermann et al. / Bioorg. Med. Chem. Lett. xxx (2015) xxx–xxx
hinge region (Fig. 1). In the ribose pocket these are Gly993 (Ala
in JAK3) and Val911 (Ile in TYK2). The different amino acids in
the solvent-exposed extended hinge region are Gln853 (Arg in
JAK1 and TYK2 and Ser in JAK3), Met865 (Leu in the rest of the
JAK family), Tyr931 (Phe in JAK1), Ser936 (Cys in JAK3), and
Asp939 (Glu in JAK1). We first set out to study carbazole substitu-
tions at the C-3 position that project towards the amino acid differ-
ences in the ribose binding pocket to improve JAK family
selectivity (Fig. 2).
The ribose binding pocket in JAK2 is slightly larger than in TYK2
(Val911 v. Ile). Not unexpectedly, reducing the size of the sub-
stituents on the C-ring (e.g., des-chloro-compound 2) resulted in
the erosion of selectivity, but larger substituents (e.g., 3-propoxy
compound 5) reduced JAK2 potency. Ortho-substitution of the phe-
nyl ring (e.g., 4) leads to loss of JAK2 selectivity, likely due to fur-
ther increase in the dihedral angle between the phenyl ring and
the carbazole. Replacement of chloride substituents with methyl
groups retained the activity (data not shown). Overall, one or
two small substituents, in the meta or para position on the phenyl
ring, are optimal for potency and selectivity.
We then turned our attention towards the C-6 and C-7 positions
of the carbazole, to capitalize on the amino acid differences in the
extended hinge region vs. other JAK family members, while at the
same time attempting to improve ADME properties. We hypothe-
sized that the introduction of a tertiary amine, protonated under
physiological conditions, into the C-6 amide portion of the mole-
cule (e.g., compounds 6–8) would position a positive charge near
Gln853 in JAK2 and less favorably near arginine in JAK1 and
TYK2. Contrary to our expectations, these changes resulted in
reduced selectivity for JAK2 versus JAK1 and TYK2.
Introducing a methylene spacer between the carbazole and the
phenyl C-ring (compound 9) reduced JAK2 potency, likely because
of disturbed coplanarity between the 2 ring systems. Increasing the
sp3 content of the molecule by saturation of the A-ring (e.g., 10)
resulted in ꢀ30-fold loss of JAK2 potency, suggesting the necessity
of a flat aromatic ring in this position, to fit into the narrow gap
between Leu855 and Gly935 (see Fig. 1).
Migrating the tertiary amide from C6 to C7 resulted in a small
but consistent (ꢀ2–3-fold) loss of JAK2 potency (compounds
11–13, Table 2). This is consistent with the fact that an amide
carbonyl on C6, but not C7, can form the water-mediated H-bond
with Asp939 that was observed in the X-ray crystal structure of
compound 1 (Fig. 1). In contrast to the C6-amide series, the
introduction of an additional basic nitrogen into the C-7 amide
series (14) was tolerated and even beneficial for kinase
selectivity, cell potency, and solubility. In particular, the
Figure 1. X-ray structure of compound 1 in the ATP binding site of the kinase
domain of JAK2. Carbon atoms of 1 are colored in magenta. Hydrogen bonds shown
as dotted lines.
O
O
NH2
O
NH
H
N
N
NH2
B
3
O
A
C
R
6
9
O
NH2
O
R'
N
H
N
1-8
R"
Cl
Cl
N
O
O
10
Figure 2. Structures of compounds 1–10.
Gly993-Asp994 amide in JAK2, rendering it unable to form the
above mentioned hydrogen bond.12 The dichlorophenyl ring of 1
is tucked under the P-loop near the ribose-phosphate pocket at
the start of the DFG motif of the activation loop.
The C-3-phenyl ring is slightly twisted relative to the carbazole
core of the molecule, with a dihedral angle of ꢀ30°. The-6-amido-
morpholino group extends towards the solvent-exposed extended-
hinge region, and participates in water-mediated hydrogen bonds
with Gln853 and Asp939 (see Fig. 1 and Table 1).
The JAK family selectivity of 1 is likely due to nearby residue
differences in the ribose pocket and solvent exposed extended
Table 1
Kinase and cell potency data for C-6 amides 1–10
Compd
R0R00N
R
JAK2 (nM)
JAK1 (nM)
JAK3 (nM)
TYK2 (nM)
SET-2a (nM)
A549b (nM)
O
O
O
O
O
N
N
N
N
N
1
2
3
4
5
6
3,4-di-Cl
H
2.4
2.6
2.8
6.8
8.0
1.2
52
18
29
31
62
5.7
139
48
31
200
120
16
2140
2400
2160
3376
5820
1750
16
3-OMe
2,4-di-Cl
3-O-nC3H7
4-Cl
66
32
119
130
35
40
216
350
71
129
2.8
N
N
N
N
7
3,4-di-Cl
3.1
41
59
45
140
990
N
N
8
3-OMe
3.5
45
66
39
280
3730
9
10
See Figure 2
See Figure 2
17
86
169
1899
123
2021
179
608
210
n.d.
810
n.d.
a
SET-2 cell proliferation IC50 (JAK2 dependent cell line).
A549 cell proliferation IC50 (JAK2 independent cell line, used as counter-screen for non-selective cytotoxic compounds).
b
Please cite this article in press as: Zimmermann, K.; et al. Bioorg. Med. Chem. Lett. (2015), http://dx.doi.org/10.1016/j.bmcl.2015.04.101

K. Zimmermann et al. / Bioorg. Med. Chem. Lett. xxx (2015) xxx–xxx
3
Table 2
versus bulkier and more polar, positively charged Arg in JAK1
and TYK2).
Kinase and cell potency data for C-7 amides 11–14
Compd R0R00N
R
JAK2 JAK1 JAK3 TYK2 SET-2a A549b
(nM) (nM) (nM) (nM) (nM) (nM)
Introduction of a pyridine ring as replacement for the phenyl C-
ring gave compounds with low selectivity versus JAK1 (e.g., com-
pound 20). However, the N-methyl-indazole C-ring replacement
was well tolerated when combined with the preferred C-7 sub-
stituent. Compound 21 has excellent kinase selectivity and good
antiproliferative activity against JAK2-dependent SET-2 cells.
Metabolic stability of 21 is comparable to 14. (75% and 59%
remaining compound after 10 min incubation with human and
rat liver microsomes resp.). While the solubility of 21 at acidic
O
O
O
N
N
N
11
12
13
H
7.4
5.7
31
49
67
246
193
311
70
171
201
170
210
110
2640
3-OMe
3080
3-F-4-OMe 5.5
3-OMe 3.5
10,000
N
14
158
225
243
80
3840
N
a
SET-2 cell proliferation IC50 (JAK2 dependent cell line).
A549 cell proliferation IC50 (JAK2 independent cell line, used as counter-screen
b
pH was improved compared to the initial screening hit 1
for non-selective cytotoxic compounds).
(0.044 mg/ml at pH 1.0), its solubility at neutral pH and permeabil-
ity were too low to achieve sufficient oral exposure and were infe-
rior to 14 (<0.001 mg/ml at pH 6.5, PAMPA = 37 nm/s).
Compounds 14 and 21 were further tested in the AMBIT
KinomeScan panel of 380 kinases to assess kinase selectivity.
Compounds 14 and 21 bind to <6% and <3% of kinases in the panel,
respectively.14
O
NH2
H
N
3
7
O
R
N
R"
R'
11-14
In summary, we herein report hit optimization of carbazole 1
leading to the identification of novel potent JAK2 inhibitors (14
and 21), with greater than 45-fold selectivity over other JAK family
members (JAK1, JAK3, TYK2). These analogs also displayed excel-
lent cellular potency against JAK2-dependent SET-2 cells (80 and
127 nM; respectively). In addition to having the desired kinase
selectivity profile, compound 14 also provided improved pharma-
ceutics properties over 21. Further studies towards improving the
ADME properties of this series of JAK2 inhibitors will be published
in due course (Fig. 4).
The syntheses of compounds 1–9 was carried out as per
Scheme 1.9 Para-amino-benzoic amides 22a and 22b undergo
SNAr reaction with 5-bromo-2-fluoro-benzonitrile 23 to give
biphenylamines 24a,b. Standard Suzuki conditions were used to
install the ring C 25. Cyclization to the carbazole and hydrolysis
of the CN group to a primary amide was achieved upon heating
25 with palladium acetate in acetic acid. These harsh conditions
also resulted in partial cleavage of the amide, leading to isolation
of acid 26 as a byproduct, which was utilized to access amide 7
Figure 3. Structures of compounds 11–14.
(S)-dimethylamino-pyrrolidine amide, 14, demonstrated greater
than 45-fold selectivity against other members of the JAK kinase
family. The compound also displayed good cell potency (SET-2
IC50 = 80 nM) and cellular selectivity.13 Compound 14 was further
evaluated for its in-vitro ADME properties. The solubility of 14 is
significantly higher than that of compound 1. (>1.9 mg/mL at pH
1.0; 0.41 mg/mL at pH 6.5; compared to <0.001 mg/mL across pH
range for 1). The permeability of 14 is modest (PAMPA = 292
compared to >600 for e.g., 2, 4, 11, 13), as is metabolic stability
(79% and 46% remaining compound after 10 min incubation with
human and rat liver microsomes resp.) (Fig. 3).
We further explored 7-amino-carbazoles (Table 3). The ‘reverse
amide’ 15 displayed poor JAK family selectivity. Several other
amides of the 7-amino-carbazole (not shown) displayed a similar
trend.
Tertiary amines (e.g., 16) exhibited a significant improvement
in the selectivity versus JAK3, however showed modest selectivity
versus JAK1 and TYK2 and also reduced cellular potency. The intro-
duction of a basic amine into the C7-substituent (e.g., 17–19) fur-
ther improved selectivity over JAK1 and TYK2. The effects on JAK
family selectivity observed for various substituents at C7 are con-
sistent with residue differences in its vicinity (Gln853 in JAK2
O
NH2
R
H
N
R"
N
R'
15-21
Figure 4. Structures of compounds 15–21.
Table 3
Kinase and cell potency data for C-7 amines 15–21
Compd
R0R00N
R
JAK2 (nM)
10.2
JAK1 (nM)
77
JAK3 (nM)
59
TYK2 (nM)
87
SET-2a (nM)
100
A549b (nM)
1310
H
O
N
15
O
O
O
N
16
17
18
7.4
2.1
2.9
218
141
195
990
264
197
158
71
500
370
470
10,000
3510
N
N
Cl
Cl
N
N
N
N
160
2730
O
19
20
5.2
6.1
432
100
242
442
197
117
260
120
930
N
N
O
N
2000
N
N
N
21
5.5
339
324
247
130
1410
N
a
SET-2 cell proliferation IC50 (JAK2 dependent cell line).
A549 cell proliferation IC50 (JAK2 independent cell line, used as counter-screen for non-selective cytotoxic compounds).
b
Please cite this article in press as: Zimmermann, K.; et al. Bioorg. Med. Chem. Lett. (2015), http://dx.doi.org/10.1016/j.bmcl.2015.04.101

4
K. Zimmermann et al. / Bioorg. Med. Chem. Lett. xxx (2015) xxx–xxx
NH2
F
N
cyclohexanone 28. Bromoanthranilic acid 29 was converted to
the hydrazine 30 using standard conditions, then condensed with
28 to give 31. Amide formation, followed by DDQ oxidation/dehy-
drogenation and ester hydrolysis gave 33, which was converted to
products 11–14 via amide formation followed by Suzuki coupling
using standard methods.
The partially saturated analog 10 was made following the strat-
egy shown in Scheme 2, using 4-oxocyclohexanecarboxylic acid
instead of 28 and skipping the DDQ oxidation step. See Ref. 9a
for details.
N
H
N
O
N
X
O
Br
a
Br
N
X
22a, b
23
24a,b
a: X=O
b: X=NMe
N
H
OH
B
N
HO
O
R
R
N
c
b
25a,b
X
The C-7-anilines required for the synthesis of compounds 15–
21 were prepared via Curtius rearrangement of 32 with
diphenylphosphoryl azide in t-butanol, followed by BOC cleavage
(Scheme 3). A Suzuki reaction installed the substituted aryl C ring.
Dihydrofuran and N-substituted dihydopyrrole 35 were ozonized
and the resulting di-aldehyde solutions used without isolation in
reductive amination reactions to form the morpholine or piperi-
dine rings of 16–21.
H
N
H
N
CONH2
CONH2
R
+
R
O
O
N
OH
26
7
1-5,9 X=O
6,8
d
X
X=NMe
Scheme 1. Synthesis of compounds 1–8. Reagents and conditions: (a) KOtBu,
DMSO, 1 h, 0 °C to rt, 70%.; (b) Pd(PPh3)4, Na2CO3, Toluene/MeOH 2:1, 105 °C, 40–
80%; (c) 2.5 equiv. Pd(OAc)2, HOAc, 2 h, 130 °C, 30–50%; (d) amine, HATU, DIPEA,
DMAP, DMF 60%.
Acknowledgements
The authors thank the lead profiling department at Bristol-
Myers Squibb for generating the data presented in Tables 1–3.
EtOOC
O
EtOOC
OH
a,b
27
28
Supplementary data
O
OH
O
OH
H
N
H2N
H2N
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.bmcl.2015.04.
101.
c
Br
Br
OH
29
30
O
H
N
28
+
30
EtO
O
d
Br
References and notes
31
O
NH2
1. (a) Leonard, W. J.; O’Shea, J. J. Annu. Rev. Immunol. 1998, 16, 293; (b) Wilks, A. F.
Semin. Cell Dev. Biol. 2008, 19, 319.
2. Palandri, F.; Polverelli, N.; Catani, L.; Cavo, M.; Vianelli, N. Int. J. Hematol. Oncol.
2013, 2, 251.
H
N
11-14
HO
O
e,f,g
h,i
Br
32
3. Quintas-Cardama, A.; Verstovsek, S. Clin. Cancer Res. 1933, 2013, 19.
4. (a) Rosenthal, A.; Mesa, R. A. Pharmacotherapy 2014, 15, 1265; (b)
Abdelrahman, R. A.; Begna, K. H.; Al-Kali, A.; Hogan, W. J.; Litzow, M. R.;
Tefferi, A. Leukemia 2015, 29, 498.
5. Derenzini, E.; Younes, A. Expert Opin. Investig. Drugs 2013, 22, 775.
6. Tam, C. S.; Verstovsek, S. Expert Opin. Investig. Drugs 2013, 22, 687.
7. Verstovsek, S.; Kantarjian, H.; Mesa, R. A.; Pardanani, A. D.; Cortes-Franco, J.;
Thomas, D. A.; Estrov, Z.; Fridman, J. S.; Bradley, E. C.; Erickson-Viitanen, S.;
Vaddi, K.; Levy, R.; Tefferi, A. N. Engl. J. Med. 2010, 363, 12. 1117 claims that
both, JAK1 and JAK2 inhibition, are significant contributor to the clinical
efficacy of ruxolitinib observed in Myelofibrosis patients. But the same
publication also describes significant side effects that appear consistent with
being caused by immunosupression via JAK1 inhibition.
8. (a) Hanan, E. J.; van Abbema, A.; Barrett, K.; Blair, W. S.; Blaney, J.; Chang, C.;
Eigenbrot, C.; Flynn, S.; Gibbons, P.; Hurley, C. A.; Kenny, J. R.; Kulagowski, J.;
Lee, L.; Magnuson, S. R.; Morris, C.; Murray, J.; Pastor, R. M.; Rawson, T.; Siu, M.;
Ultsch, M.; Zhou, A.; Sampath, D.; Lyssikatos, J. P. J. Med. Chem. 2012, 55, 10090;
(b) Hitoshi, Y.; Lin, N.; Payan, D. G.; Markovtsov, V. Int. J. Hematol. 2010, 91,
189; (c) Wang, T.; Duffy, J. P.; Wang, J.; Halas, S.; Salituro, F. G.; Pierce, A. C.;
Zuccola, H. J.; Black, J. R.; Hogan, J. K.; Jepson, S.; Shlyakter, D.; Mahajan, S.; Gu,
Y.; Hoock, T.; Wood, M.; Furey, B. F.; Frantz, J. D.; Dauffenbach, L. M.; Germann,
U. A.; Fan, B.; Namchuk, M.; Bennani, Y. L.; Ledeboer, M. W. J. Med. Chem. 2009,
52, 7938; (d) Ioannidis, S.; Lamb, M. L.; Ameida, L.; Guan, H.; Peng, B.;
Bebernitz, G.; Bell, K.; Alimzhanov, M.; Zinda, M. Bioorg. Med. Chem. Lett. 2010,
20, 1669.
Scheme 2. Synthesis of compounds 11–14. Reagents and conditions: (a) H2 (4 bar),
Rh-Al2O3, EtOH, 98%; (b) CrO3, H2SO4, acetone, 95%; (c) NaNO2, HCl, then SnCl2 89%;
(d) AcOH, reflux, 70%; (e) EDC, HOBT, NH4OH, THF + DCM, 94%; (f) DDQ, Tol, reflux,
93%; (g) NaOH, THF + MeOH + water, quant.; (h) EDC, HOBT, amine, 90%; (i)
arylboronic acid, Pd(PPh3)4, aq Na2CO3, toluene, MeOH, 50–60%.
O
H
N
NH2
33
32
H2N
NH2
a,b
c
Br
O
H
N
f
H2N
15
R
34
16-21
e
X
[(CHO-CH2)2X]
d
X=O, NMe, NBoc
35
36
Scheme 3. Synthesis of compounds 15–21. Reagents and conditions: (a) DPPA,
Et3N, tBuOH, 86%; (b) TFA, DCM, 96%; (c) arylboronic acid, PdCl2(dppf), aq Na2CO3,
DME, 50–70%; (d) Ozone, then PPh3; (e) NaBH3CN, HC(OCH3)3, HOAc, 20–50%; (f)
tetrahydropyran-4-carboxylic acid, HOBT, EDC, DIPEA, DMF, 47%.
9. (a) Purandare, A. V.; Batt, D. G.; Liu, Q.; Johnson, W. L.; Mastalerz, H.; Zhang, G.;
Zimmermann, K. PCT Int. Appl. (2010), WO 2010080474 A1 Jul 15, 2010.; (b)
Liu, Q.; Batt, D. G.; Delucca, G. V.; Shi, Q.; Tebben, A. J.; US Patent Appl. Publ.
(2010), US 20100160303 A1 20100624.
10. Aqueous solubility <0.001 mg/ml.
11. The atomic coordinates of compound 1 have been deposited in the RCSB
Protein Data Bank (code 4ZIM). Experimental procedures for the JAK2 protein
purification and crystallization are available in the Supplemental Material
section.
12. Thoma, G.; Nuninger, F.; Falchetto, R.; Hermes, E.; Tavares, G. A.;
Vangrevelinghe, E.; Zerwes, H.-G. J. Med. Chem. 2011, 54, 284.
via standard amide coupling conditions. The synthesis of benzyl-
analog 9 followed the same route, using 2-benzyl-4,4,5,5-tetram-
ethyl-1,3,2-dioxaborolane instead of the substituted phenyl-
boronic acids.
The C7-substituted amides 11–14 were prepared as shown in
Scheme 2. Meta-hydroxy benzoic ester 27 was hydrogenated over
rhodium on alumina, followed by Jones oxidation to give
13. Cellular selectivity was defined as the ratio of IC50’s for non-JAK2 dependent
A549 cell line versus JAK2-dependent SET-2 cell line.
14. Less than 10% control activity at 1 micromolar concentration.
Please cite this article in press as: Zimmermann, K.; et al. Bioorg. Med. Chem. Lett. (2015), http://dx.doi.org/10.1016/j.bmcl.2015.04.101