Design, synthesis, and biological activity of substituted 2-amino-5-oxo-5H-chromeno[2,3-b]pyridine-3-carboxylic acid derivatives as inhibitors of the inflammatory kinases TBK1 and IKKε for the treatment of obesity

Article abstract of DOI:10.1016/j.bmc.2018.09.020

The non-canonical IκB kinases TANK-binding kinase 1 (TBK1) and inhibitor of nuclear factor kappa-B kinase ε (IKKε) play a key role in insulin-independent pathways that promote energy storage and block adaptive energy expenditure during obesity. Utilizing docking calculations and the x-ray structure of TBK1 bound to amlexanox, an inhibitor of these kinases with modest potency, a series of analogues was synthesized to develop a structure activity relationship (SAR) around the A- and C-rings of the core scaffold. A strategy was developed wherein R₇ and R₈ A-ring substituents were incorporated late in the synthetic sequence by utilizing palladium-catalyzed cross-coupling reactions on appropriate bromo precursors. Analogues display IC₅₀ values as low as 210 nM and reveal A-ring substituents that enhance selectivity toward either kinase. In cell assays, selected analogues display enhanced phosphorylation of p38 or TBK1 and elicited IL-6 secretion in 3T3-L1 adipocytes better than amlexanox. An analogue bearing a R₇ cyclohexyl modification demonstrated robust IL-6 production in 3T3-L1 cells as well as a phosphorylation marker of efficacy and was tested in obese mice where it promoted serum IL-6 response, weight loss, and insulin sensitizing effects comparable to amlexanox. These studies provide impetus to expand the SAR around the amlexanox core toward uncovering analogues with development potential.

Full text of DOI:10.1016/j.bmc.2018.09.020

Accepted Manuscript
Design, synthesis, and biological activity of substituted 2-amino-5-oxo-5H-
chromeno[2,3-b]pyridine-3-carboxylic acid derivatives as inhibitors of the in-
flammatory kinases TBK1 and IKKε for the treatment of obesity
Tyler S. Beyett, Xinmin Gan, Shannon M. Reilly, Andrew V. Gomez, Louise
Chang, John J.G. Tesmer, Alan R. Saltiel, Hollis D. Showalter
PII:
S0968-0896(18)31462-7
https://doi.org/10.1016/j.bmc.2018.09.020
BMC 14545
DOI:
Reference:
Bioorganic & Medicinal Chemistry
To appear in:
Received Date:
Revised Date:
Accepted Date:
14 August 2018
12 September 2018
18 September 2018
Please cite this article as: Beyett, T.S., Gan, X., Reilly, S.M., Gomez, A.V., Chang, L., Tesmer, J.J.G., Saltiel, A.R.,
Showalter, H.D., Design, synthesis, and biological activity of substituted 2-amino-5-oxo-5H-chromeno[2,3-
b]pyridine-3-carboxylic acid derivatives as inhibitors of the inflammatory kinases TBK1 and IKKε for the treatment
of obesity, Bioorganic & Medicinal Chemistry (2018), doi: https://doi.org/10.1016/j.bmc.2018.09.020
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Design, synthesis, and biological activity of substituted 2-amino-5-oxo-5H-
chromeno[2,3-b]pyridine-3-carboxylic acid derivatives as inhibitors of the
inflammatory kinases TBK1 and IKK for the treatment of obesity
Tyler S. Beyetta,b†, Xinmin Ganc,d†, Shannon M. Reillye, Andrew V. Gomeze , Louise Changb,
John J. G. Tesmerb, Alan R. Saltiele and Hollis D. Showalterc,d *
aProgram in Chemical Biology, bLife Sciences Institute, Departments of Pharmacology and
Biological Chemistry, cDepartment of Medicinal Chemistry, dVahlteich Medicinal Chemistry
Core, University of Michigan, Ann Arbor, MI 48109, United States
eDepartments of Medicine and Pharmacology, Institute for Diabetes and Metabolic Health,
University of California, San Diego, La Jolla, CA 92093-0912, United States
†Authors contributed equally to this work.
*Corresponding author
Abstract: The non-canonical IB kinases TANK-binding kinase 1 (TBK1) and inhibitor of
nuclear factor kappa-B kinase  (IKK) play a key role in insulin-independent pathways that
promote energy storage and block adaptive energy expenditure during obesity. Utilizing docking
calculations and the x-ray structure of TBK1 bound to amlexanox, an inhibitor of these kinases
with modest potency, a series of analogues was synthesized to develop a structure activity
relationship (SAR) around the A- and C-rings of the core scaffold. A strategy was developed
wherein R7 and R8 A-ring substituents were incorporated late in the synthetic sequence by
utilizing palladium-catalyzed cross-coupling reactions on appropriate bromo precursors.
Analogues display IC50 values as low as 210 nM and reveal A-ring substituents that enhance
selectivity toward either kinase. In cell assays, selected analogues display enhanced
phosphorylation of p38 or TBK1 and elicited IL-6 secretion in 3T3-L1 adipocytes better than
amlexanox. An analogue bearing a R7 cyclohexyl modification demonstrated robust IL-6
production in 3T3-L1 cells as well as a phosphorylation marker of efficacy and was tested in
obese mice where it promoted serum IL-6 response, weight loss, and insulin sensitizing effects
comparable to amlexanox. These studies provide impetus to expand the SAR around the
amlexanox core toward uncovering analogues with development potential.
Key words: obesity; amlexanox; TANK-binding kinase 1 (TBK1); inhibitor of nuclear factor
kappa-B kinase subunit epsilon (IKK); TBK1·amlexanox co-crystal; SAR

1.
Introduction
O
O
NH2
O
The current worldwide epidemic of obesity is exacting
profound human health costs in both industrialized and
developing nations, where it is associated with an
increased risk of developing a host of medical conditions
such as type 2 diabetes, dyslipidemia, non-alcoholic fatty
liver disease, cardiovascular disease, and cancer.[1] As
of 2015, one-third of adults in the United States were
obese and an additional one-third overweight.
OSO2NH2
topiramate
O
O
phentermine
1
( )
2
( )
O
O
CH3(CH2)5
NH
(CH2)10CH3
O
O
Cl
CH3
lorcaserin
NHCHO
orlistat
The mechanisms by which obesity can lead to type 2
diabetes and other diseases remain unresolved, and little
progress has been made in devising new therapeutic
strategies that attack the core defects underlying either
disorder.[2] The few pharmacological therapies available
for the treatment of obesity act principally through the
central nervous system, but provide only modest
3
( )
4
( )
O
O
N
NH2
8
3
7
CO2H
6
amlexanox
5
( )
Figure 1. Structures of anti-obesity drugs
suppression of appetite at therapeutically viable doses.[3] In some treatments, a combination of
medications that intervene by more than one biochemical pathway is utilized.[4] Exemplary
small molecule drugs are (i) a combination of phentermine (1), a sympathomimetic amine that
acts as an appetite suppressant and stimulant, and topiramate (2) an anticonvulsant[1]; (ii)
lorcaserin (3), a selective 5-HT2C receptor agonist[5]; and (iii) orlistat (4), which blocks fat
absorption in the intestine by acting as a lipase inhibitor [6] (Figure 1). Beyond that, there is
little new in development. Indeed, none of the currently available and widely used treatments for
obesity addresses the underlying energy imbalance that is the root cause.[7]
One hallmark of obesity is a state of chronic, low-grade inflammation in liver and fat tissue.
Prior studies on the inflammatory links between obesity and type 2 diabetes led to the elucidation
of an important role for the nuclear factor kappa-light-chain-enhancer of activated B cells (NF-
B) pathway, revealing the role of two non-canonical inhibitors of kappa B (IB) kinases,
TANK-binding kinase 1 (TBK1) and inhibitor of nuclear factor kappa-B kinase subunit 
(IKK in insulin-independent pathways that promote energy storage and block adaptive energy
expenditure during obesity.[8] The expression and activity of these kinases are increased in mice
fed a high fat diet due to the generation of inflammation in response to obesity. Deletion of the
IKK gene attenuated some of the deleterious metabolic effects of high fat feeding including
reduced weight gain, lowered insulin resistance, and less chronic low-grade inflammation. These
data led to a new hypothesis about the role of inflammation in the generation of insulin resistance
and type 2 diabetes during obesity,[9] prompting a high-throughput screen for small molecule
inhibitors of the TBK1 and IKK kinases. One hit from the screen was amlexanox (5; Figure 1),
an approved drug for the treatment of canker sores, asthma, and allergic rhinitis. The compound

proved to be a modest inhibitor of TBK1 and IKK and reproduced the effects of the IKK
knockout phenotype, which include weight loss in obese mice, improved insulin sensitivity, and
reduced hepatic steatosis and inflammation. Amlexanox also increased the level of the second
messenger molecule cAMP and triggered the release of the hormone interleukin-6 (IL-6) from
fat cells thereby reducing the production of glucose. Two recent clinical trials indicated that
amlexanox treatment effectively lowered glycated hemoglobin (HbA1c) in a subset of patients
with obesity and type 2 diabetes.[10]
Amlexanox is thus an attractive starting point for analogue synthesis. We previously reported the
TBK1·amlexanox structure which revealed the amino pyridine moiety on amlexanox binding the
kinase hinge in a canonical type 1, DFG-in mode.[11] The C-3 carboxylic acid interacts with
Thr156, and its replacement with acid bioisosteres improved the potency of amlexanox but
without improvement in efficacy. In this paper, we detail the synthesis of core modifications of
amlexanox and develop a more detailed structure-activity relationship (SAR) that addresses
some of the liabilities of amlexanox, including its modest potency and low solubility. We also
report the X-ray co-crystal structures of select amlexanox analogues and TBK1, which will guide
our future studies toward generating analogues with enhanced potency and selectivity.
2.
Analogue Design
We first focused on modifications around the A ring of amlexanox and designed a set of
analogues bearing conservative modifications at the C-7 position alongside a small set of C-6
and C-8 derivatives that may stabilize the phosphate-binding loop (P-loop) or interact with the
kinase hinge, respectively. Next, a large series of cyclic modifications were examined at the C-7
position due to their synthetic accessibility and in consideration of the crystallographic data
showing space to accommodate bulkier modifications. Finally, previously reported bioisostere
derivatives of the C-3 carboxylic acid[11] were combined with C-7 modifications to assess the
effect of dual modification of the pharmacophore.
3.
Chemistry
Synthetic methodologies to make amlexanox and analogues were reported several decades ago
by Takeda investigators.[12], [13], [14] In almost all cases, the A-ring substituent had been
incorporated early onto a precursor 4-oxo-4H-l-benzopyran-3- carbonitrile. With a goal to
develop a more efficient route for higher output of analogues, we mapped out a strategy of
introducing A-ring substituents at a later stage of the synthetic sequence. Our overall strategy is
outlined in Scheme 1, wherein suitable tricyclic A-ring bromo-substituted precursors (6 – 10;
Scheme 1) were first synthesized and then subjected to a range of palladium-catalyzed cross-
coupling reactions (carbon-carbon, carbon-nitrogen) to introduce a broad range of R6-R8
functionality. The resultant ethyl ester intermediates (15-24, 32-58) with olefinic/alkynic side
chains would then be reduced via catalytic hydrogenation to products (25 – 31, 59 – 84)
embellished with a wide variety of side chains. Subsequent base hydrolysis of these and selected

R8
R7
O
O
N
NH2
R8
R7
O
O
N
NHR2
R8
R7
O
O
N
NHR2
(b)
Z
(a)
Z
Z
CO2Et
CO2Et
CO2Et
6
7
8
9
: Z = C; R2 = R7 = H; R8 = Br
: Z = C; R2 = H; R7 = Br; R8 = F
: Z-R8 = N; R2 = H; R7 = Br
: Z = C; R7 = Br; R2 = R8 = H
25-31
59-84
(Scheme SM1)
(Scheme SM2)
15-24
32-58
(Scheme SM1)
(Scheme SM2)
R
7 = H, alkyl, cycloalkyl, cycloalkylethyl
ZR8 = N; when Z = C, R8 = H, F, alkyl,
cycloalkyl, cycloalkylethyl
R
R
2 = H, Boc
10
: Z = C; R7 = Br; R2 = Boc; R8 = H
7 = H, c-Pr, c-Bu, i-Bu, alkenyl, cycloalkenyl,
cycloalkylvinyl, alkynyl, Ph, N(alkyl)2
ZR8 = N; when Z = C, R8 = H, F, alkenyl,
cycloalkenyl, cycloalkylvinyl
O
N
NH2
(c)
CO2H
6
(c)
O
C-6 amlexanox
108
(Scheme 3, Table 1)
R8
R7
O
O
N
NH2
R
7 = H, alkyl, alkenyl, alkynyl, cycloalkyl,
cycloalkenyl, Ph, cycloalkylethyl,
cycloalkylvinyl, N(alkyl)2
Z
O
N
NH2
CO2H
N
N
ZR8 = N; when Z = C, R8 = H, F, alkyl,
alkenyl, cycloalkyl, cycloalkenyl,
R7
N
NH
O
105-107, 109-121
(Scheme SM1, Table 1)
(Scheme SM2, Table 2)
cycloalkylethyl, cycloalkylvinyl
122-167
168-171
(Scheme 4, Table 3)
Scheme 1. Synthetic strategy to make linear and cyclic amlexanox analogues. (a) Palladium-catalyzed cross
coupling reactions. (b) Catalytic hydrogenation. (c) Base hydrolysis. For full details, see Schemes 2-4;
unsaturated precursors, along with a few amine congeners generated from Buchwald coupling,
would then provide a broad range of analogues (Tables 1 and 2) for initial evaluation in our
enzyme assays. These were supplemented with a smaller set of congeners with a tetrazole
bioisostere installed in the C-3 position of the C-ring (Table 3).

The synthesis of novel tricyclic
precursors 6-8 followed a fairly
classical sequence of reactions
starting with commercially available
bromo- substituted (2-
hydroxyaryl)ethan-1-ones (11a – 11c;
Scheme 2). Hence, treatment of 11a
or 11b with a large excess of
Vilsmeier reagent resulted in a two-
stage homologation to give 7- or 6-
bromo-4-oxo-4H-chromene-3-
carbaldehydes 12a and 12b,
respectively, in 73 – 88% yield.
Homologation of 2-bromo-5-
hydroxypyridin-4-yl)ethan-1-one[15]
(11c) could not be achieved under
these conditions, thus an alternate
approach was taken. Treatment of 11c
with Brederick’s reagent proceeded
smoothly to generate intermediate
(E)-1-(2-bromo-5-hydroxypyridin-4-
yl)-3-(dimethylamino)prop-2-en-1-
one in 60% yield following
O
O
O
O
O
O
O
O
(c)
CH=NOH
(b)
CHO
(d)
CN
CN
Br
Br
Br
88
87
86
89
(e)
N
(a)
O
O
NH2
OH
O
OH
O
O
O
(h)
(a)
(d)
CO2Et
CHO
R
Br
85
90
91
: R = prop-1-en-2-yl
: R = isopropyl
92
(f)
93
(f)
(h)
(g)
OH
O
O
N
NH2
O
(a)
CHO
CO2H
6
O
O
94
108
(Table 1)
95
Scheme 3. Synthesis of C-6 amlexanox 108. (a) Method I: DMF,
POCl3, 25 °C, 17 h, then H2O, 67 - 94%. (b) Method J: NH2OH·HCl,
EtOH, 60 °C, 1 h, 77%. (c) Method K: Ac2O, 110 °C, 18 h, 99%. (d)
Method A: for 88 to 89: tri-n-butyl(prop-1-en-2-yl)stannane,
Pd(PPh3)2Cl2, p-dioxane, 80-100 °C, 16.5 h, 81%; for 85 to 92:
potassium trifluoro(prop-1-en-2-yl)borate, Pd(PPh3)2Cl2, PPh3, K2CO3,
THF, 100 °C, 18 h, 98%. (e) Method L: CNCH2CO2Et, piperidine,
EtOH, reflux, 2 h, 73%. (f) Method D: H2, 10% Pd/C, 2:1 v/v
EtOH:DCM, 25 °C, 18-45 h, 78 - 95%. (g) Method E: 1N aq NaOH,
EtOH, 50 °C, 1.5 h, 86%. (h) Oximation (Method J) results in
destruction of starting compound.
chromatography. Further homologation using Vilsmeier reagent derived from DMF and triflic
anhydride then provided 6-bromo-4-oxo-4H-pyrano[2,3-c]pyridine-3-carbaldehyde (12c) in 88%
yield. The use of Vilsmeier reagent with the weakly nucleophilic triflate counterion was critical
to the success of this reaction. Further elaboration of carbaldehydes 12a – 12c to corresponding
oximes 13a – 13c and then nitriles 14a – 14c employed conditions utilized for the synthesis of
related 4-oxo-4H-chromenes.
R8
R7
OH
Me
R8
O
R8
O
Z
(a)
(b)
Z
Z
Annulation of nitriles 14a – 14c
with ethyl cyanoacetate to
tricyclic starting materials 6 – 8
of Scheme 1 was achieved in 77
– 93% yield utilizing literature
conditions. Compound 9 was
synthesized as previously
described[16] and converted to
N-Boc derivative 10 in 83%
yield.
R7
CH=NOH
R7
CHO
O
O
O
11a-c
13a-c
12a-c
R8
O
O
(c)
Z
(d)
a:
b:
c:
Z = C; R7 = H; R8 = Br
Z = C; R7 = Br; R8 = F
Z-R8 = N; R7 = Br
6-8
(Scheme 1)
R7
CN
14a-c
Scheme 2. Synthesis of bromo tricyclic precursors 6-8. (a) Method I: for
11a,b to 12a,b: DMF, POCl3, 25 °C, 17 h, then H2O, 73-88%; for 11c to 12c:
t-BuOCH(NMe2)2, 25 °C, 17 h, 60% yield; DMF/(CF3SO2)2O, 25 °C, 3 h,
then H2O, 88%. (b) Method J: NH2OH·HCl, EtOH, 60 °C, 1-2 h, 61-80%. (c)
Method K: Ac2O, 110 °C, 18 h, 94-99%. (d) Method L: CNCH2CO2Et,
piperidine, EtOH, reflux, 2 h, 77-93%.

The synthesis of targeted 7- or 8-(substituted)-5-oxo-5H-chromeno[2,3-b]pyridine-3-carboxylic

acids and 8-aza congeners along with
unsaturated precursors, following the
strategy outlined in Scheme 1, is shown
in detail in Schemes SM1 and SM2
(Supplementary Materials). After
considerable experimentation, we
determined that N-2 unprotected amines
were optimal for Suzuki[17] or
A.
OH
O
O
O
OH
O
(b)
(a)
Br
R7
CHO
R7
96
97 a,b
98 a,b
O
O
N
NH2
CO2R3
O
(d)
(c)
R7
R7
CN
O
100 a,b:
R
R
3 = Et
99 a,b
(e)
113
151
3 = H (
, Table 1;
, Table 2)
(g)
Sonogashira[18] couplings. Thus
a:
b:
R
R
7 = NMe2
7 = morpholinyl
treatment of 6 – 9 with a wide range of
alk/cycloalk-1-en-1-ylboronic acids or 2-
(alk/cycloalk-1-en-1-yl)-4,4,5,5-
B.
O
N
NH2
CN
O
N
NH2
R3
(a)
Br
O
O
tetramethyl-1,3,2-dioxaborolanes under
standard Suzuki conditions provided
alkene products (15 – 18, 24, Scheme
SM1; 36 – 52, 55 – 58, Scheme SM2) in
61-95% yield. Similarly, Sonogashira
reaction of 9 with selected substituted
alk-1-ynes provided alkyne adducts (19 –
21, Scheme SM1; 53, 54, Scheme SM2)
in 67-94% yield. However, Suzuki
coupling to derive cyclopropyl adduct 32
necessitated starting with N-Boc-
102:
R3 = CN
101
(f)
N
169:
R3
=
(Table 3)
N
NH
N
O
N
NH2
CN
O
O
NH2
N
(f)
N
NH
O
N
103
168
(Table 3)
O
O
N
NH2
R3
(d)
(h)
151
99 a,b
a:
b:
R7
R
R
7 = NMe2
7 = morpholinyl
104a,b:
R
=
=
3 = CN
172:
R
7 = morpholinyl
(f)
H
170:
171:
R3
R3
; R7 = NMe2
N
N
O
N
N
NH ; R7 = morpholinyl
N
R3
=
N
NH
N
(Table 3)
protected 10, which was also required for
Negishi cross-coupling[19] of isobutyl-
and c-butylzinc bromides to give adducts
22 and 34, respectively. The N-Boc
protecting group of each was cleaved
with TFA to give the respective products
33, 23, and 35. Hydrogenation of
unsaturated esters (15 – 17, 19 – 21, 24,
Scheme SM1; 36 – 39, 41 – 52, 54 - 58,
Scheme SM2) over 10% palladium on
carbon proceeded smoothly to give
corresponding saturated R7/R8 derivatives
in modest to excellent yields. Ester
hydrolysis of these (and 62, 71- 73, 76
(Table 3)
Scheme 4. Synthesis of R7 amino and R3 tetrazole amlexanox
analogues (a) Method A: for 96 to 97a,b; LiHMDS, NHMe2 or
morpholine; Pd2(dba)3, DavePhos, THF, 75 - 80 °C, 1-2 h, 81-
100%; for 101 to 102; 2-(cyclohex-1-en-1-yl)-4,4,5,5-tetramethyl-
1,3,2-dioxaborolane, PPh3, Pd(OAc)2, K3PO4, aq. p-dioxane, 70 °C,
1.5 h, 78%. (b) Method I: DMF, POCl3, 25 °C, 17 h, then H2O,
68%. (c) Method K: NH2OH·HCl, NaI, CH3CN, reflux, 2-3 h, 34-
58% yield. (d) Method L: for 99a,b to 100a,b: CNCH2CO2Et,
piperidine, EtOH, reflux, 2.5-4 h, 9-58%; for 99a,b to 104a,b:
CNCH2CN, piperidine, EtOH, reflux, 3 h, 65-71%. (e) Method E:
1N aq NaOH, EtOH, 50 °C, 2 h, 85-91% yield. (f) Method M:
NaN3, NH4Cl, DMF, 90 °C, 16-18 h, 27-84%. (g) Method D: H2,
10% Pd/C, 47:53 v/v EtOH:DCM, 25 °C, 22 h, 75%. (h) Method N:
SOCl2 / cat. DMF, DCM, reflux, 3 h, and then 5-aminotetrazole,
pyridine, NEt3, RT, 4h, 38%.
derived from further transformations), along with selected precursor R7 and R8 alkenes/alkynes,
under standard base conditions provided the corresponding linear (105-107, 109-112, 114-121,
Scheme SM1: Table 1) and cyclic (122-150, 152-167, Scheme SM2; Table 2) amlexanox

analogues in 48 – 95% yield.
The synthesis of the C-6 isomer of amlexanox (108; Scheme 3) could not be achieved via the
pathway outlined above for R7/R8 congeners. Standard Vilsmeier homologation of 85 provided
chromone carbaldehyde 86 in 94% yield with further transformations to oxime 87 (77% yield)
and nitrile 88 (99% yield) proceeding uneventfully. However, attempted annulation of 88 with
ethyl cyanoacetate toward securing tricyclic C-6 bromide under conditions providing 6-9 could
not be achieved. Therefore, a different ordering of steps to install the C-6 isopropyl side chain
was investigated. The application of standard Suzuki conditions to 88 resulted in complete
destruction, due to its exquisite sensitivity to base, whereas a milder Stille cross-coupling[20]
reaction in the absence of base gave 89 in 81% yield. This was then parlayed into targeted C-6
amlexanox (108, Table 1) in three standard steps in 49% overall yield. Scheme 3 also outlines
alternative reaction manifolds toward securing 108 via Suzuki-derived intermediate 92.
Subsequent reactions to intermediate aldehydes 93 and 95 proceeded smoothly, but attempted
oximation of each surprisingly resulted in compound destruction.
The synthesis of selected R7 amino analogues of amlexanox is delineated in Scheme 4A. The R7
amine moiety was efficiently installed in the first step from 96 using Buchwald palladium
catalyzed cross-coupling conditions[21] to give 97a,b. Applying similar conditions [22] to
tricyclic bromide 10 did not provide product. Application of a sequence of standard steps
(Vilsmeier homologation, nitrile formation, and annulation) to 97a,b then gave ester
intermediates (100 a,b), which were hydrolyzed under basic conditions to target carboxylic acids
(113, Table 1; 151, Table 2).
A final set of analogues was synthesized incorporating tetrazole into the R3 position as a
carboxylic acid bioisostere (Scheme 4B). Toward this end, tricyclic bromo nitrile 101 was
elaborated via Suzuki chemisty to R7 cyclohexene derivative 102, which was then saturated to
103 by catalytic hydrogenation. Each of these was elaborated into its respective R3 tetrazole
under standard conditions (sodium azide, ammonium chloride, DMF) to provide 169 and 168,
respectively (Table 3). Similarly, bicyclic nitriles 99a,b were annulated into tricyclic R3 nitriles
104a,b, which were then transformed to R7 (dialkylamino)-substituted R3- tetrazoles 170 and
171, respectively (Table 3). A final target was made via Schotten-Bauman acylation of
carboxylic acid 151 to give R7-morpholinyl, R3-tetrazole amide analogue 172 (Table 3).
All compounds were rigorously purified by flash chromatography or crystallization, and their
structural assignments are supported by diagnostic peaks in the 1H NMR spectra and by mass
spectrometry. Digital copies of NMR spectra can be found in Supplementary Materials. The
reagents utilized in reactions to our target compounds are summarized in the legends of Schemes
2-4 and Schemes SM1 and SM2 (Supplementary Materials). All reactions are categorized
according to generalized Methods A – N, and experimental procedures for each are organized
accordingly in the Experimental Section. Representative experimental procedures are given for
the Suzuki reaction (Method A), catalytic hydrogenation (Method D), and base hydrolysis

(Method E). Additional procedures for compounds derived from these three methods are
provided in Supplemental Materials.
4.
Kinase inhibition
All the compounds in Tables 1-3 were investigated for inhibition of recombinant human TBK1
and IKK. The IC50 values shown were determined by dose-response radiometric kinase assays
with myelin basic protein as substrate, and are technically apparent IC50 values. Each is an
average of three separate determinations.
Selected analogues in Tables 1-3 bearing more extensive A-ring modifications, which may
dramatically affect potency, were first evaluated at a single concentration of 1 or 5 µM. This
approximately corresponds to the IC50 of amlexanox (5) toward TBK1 and IKK, respectively.
Based on their % inhibition relative to amlexanox, full dose-response assays were conducted.
Most of these compounds are shown in Tables 2 and 3.
Table 1 lists the structures and kinase inhibitory properties of a series of linear alkyl, alkenyl, and
alkynyl A-ring derivatives of amlexanox (5), which is used for comparison. Analogues bearing
a 2-propenyl substituent at R7 (105, 110, 112) or R8 (107) generally show reduced potency
toward TBK1 and improved potency toward IKK when located at the R7 position. The addition
of a methylene spacer(s) (116, 117) had no significant effect on potency. The introduction of an
alkyne spacer (118) markedly reduced potency with the data likely compromised due to poor
solubility in the assay medium. To improve the solubility of 5, the isopropyl group was replaced
with a dimethylamino moiety in 113 which improved potency 2- and 3-fold toward TBK1 and
IKKε, respectively.
Moving the isopropyl group to the C-6 position (108) in an attempt to stabilize the P-loop by
packing against Leu15 and Gly16, reduced potency toward TBK1 without affecting potency
toward IKK. The introduction of fluorine at the R8 position (109) enhanced potency toward
TBK1 relative to amlexanox (5), whereas the 2-propenyl congener (110) was essentially
equivalent for TBK1 but enhanced for , similar to non-fluorinated congener (105).
Installation of a nitrogen in place of carbon at the 8-position of amlexanox to give 111 enhanced
potency toward both kinases. In contrast, making the same replacement for the R7 2-propenyl
congener 105 to give 112 negated the potency boost of the 2-propenyl substituent toward IKK.
Table I. In Vitro Activity of R6 – R8 Linear 5-Oxo-5H-[1]benzopyrano[2,3-b]pyridines and A-
ring Aza Congeners.

R8
R7
O
B
N
C
NH2
O
O
N
NH2
O
N
NH2
N
A
CO2H
CO2H
R7
CO2H
6
108
O
O 111 112
,
No. R7
R8
Single Concentration, % Inhibition)a
(IC50, µM)b
TBK1
Fold
Fold
IKK
Change
Change
e
e
IC50
IC50
5
i-Pr
H
H
50 ± 1.3
29 ± 7.4
(0.85 ± 0.14)
(1.3 ± 0.1c)
(0.30 ± 0.03d)
(2.3 ± 0.1d)
(1.1 ± 0.02 d)
(0.40 ± 0.08)
(0.79 ± 0.07)
(0.57 ± 0.02 d)
(5.1 ± 1.6)
(1.3 ± 0.5d)
(4.7 ± 2.3)
(7.2 ± 2.0)
(6.0 ± 0.5)
(5.4 ± 0.7)
(1.7 ± 0.3c)
(3.1 ± 1.1)
105 CH2=C(CH3)-
0.7
2.8
0.4
0.8
2.1
1.1
1.5
3.9
1.1
0.7
0.9
0.9
3.0
1.6
106
107
108
H
H
-
i-Pr
CH2=C(CH3)-
-
F
F
-
109 i-Pr
110 CH2=C(CH3)-
111 i-Pr
112 CH2=C(CH3)-
113 Me2N
-
(2.0 ± 0.3c)
0.4
(5.3 ± 0.5)
1.0
H
53 ± 3.4
35± 3.8
(1.6 ± 0.5)
(0.40 ± 0.09) c
2.1
3.2
114 Me2N(CH2)3-
H
H
21 ± 1.8
17 ± 2.0
35 ± 2.2
115
Me2NCH2C C-
20 ± 3.4
116 i-Bu
117 i-pentyl
H
H
H
1.5
1.0
0.1
1.0
1.5
0.3
(0.55 ± 0.08)
(0.83 ± 0.08)
(4.9 ± 2.0)
(3.4 ± 0.4)
118
(CH3)2CHC C-
(12 ± 1c)
32 ± 0.7
14 ± 4.8
8.7 ± 4.2
(20 ± 9)
119 i-pentyl
120 t-BuCH2CH2
121 t-BuCH=CH-
F
H
H
-5.7 ± 4.7
0.6 ± 11
-3.1 ± 3.2
a'TBK1, (1 µM); IKK (5 µM). b Concentration for 50% inhibition in radiometric kinase assays using purified protein. Reported
as IC50 (µM) ± SEM (n=3). Significance relative to amlexanox determined by a two-tailed Student’s T-test (cp<0.05, dp<0.01.
eFold change over amlexanoc (5) calculated from the mean IC50 values. See Experimental Section.
The vast majority of synthesized analogues, incorporating R7 and R8 cycloalkyl, heterocyclic
and alkcycloalkyl substituents displayed decreased inhibition of both kinases relative to
amlexanox (Table 2). Analogues displaying equivalent or improved inhibition were further tested
in dose-response assays. Introduction of conformational restriction in amlexanox (5) to give
cyclopropyl analogue 122 resulted in reduced potency toward both kinases. A general decrease
in inhibition was observed as the ring size at the R7 position was increased from cyclobutyl to
cycloheptyl (123 - 125, and 127, respectively) with TBK1 inhibition the most adversely affected.
Notably, unsaturation of the cyclohexyl ring (126) markedly improved inhibition of IKK
relative to 125, consistent with the prior observation for the unsaturated 2-propenyl substituent
preferentially improving IKK inhibition. Further modification of the 125, 126 pair to
incorporate R8 fluorination (152, 153, respectively) or 8-aza (154, 155, respectively) generally

decreased inhibition toward both kinases, especially IKK. For a series of R7 (156 – 161, 164-
167) and R8 (162, 163) analogues bearing saturated rings joined to the core through short linkers
(ethane, ethene, ethyne), all were poor inhibitors of either kinase, likely due to limited solubility
under the assay conditions. Installation of cyclohexyl (128) or cyclohexenyl (129) at R8
diminished inhibition likely due to the introduction of steric clashes with the kinase hinge.
Table 2. In Vitro Activity of R7 and R8 Cycloalkyl, Heterocyclic and Alkcycloalkyl 5-Oxo-5H-
[1]benzopyrano[2,3-b]pyridines and A-ring Aza Congeners.
R8
R7
O
B
N
C
NH2
O
N
NH2
N
A
CO2H
R7
CO2H
154 155
,
O
O
No.
5
R7
R8
Single Concentration % Inhibitiona
(IC50, µM)b
TBK1
Fold
Fold
IKK
Changee
Changee
i-Pr
H
50 ± 1.3
29 ± 7.4
(0.85 ± 0.14)
(5.1 ± 1.6)
122 c-Pr
H
H
(2.3 ± 0.1d)
44 ± 3.7
0.4
0.2
(10.6 ± 0.4d)
25 ± 3.4
0.5
1.6
123
124
125
126
c-Bu
(4.6 ± 3.6)
(3.1± 0.7c)
9.2 ± 1.9
(6.3 ± 1.0)
9.0 ± 3.1
(>100)
41 ± 4.9
(5.4 ± 1.5)
67 ± 4.1
(1.8 ± 0.3c)
34 ± 1.6
(1.9 ± 0.6)
44 ± 0.2
(1.7 ± 0.5)
c-Pentyl
c-Hexyl
H
H
H
0.5
0.5
0.5
0.8
0.9
127
128
32 ± 8.5
11 ± 9.6
9.2 ± 4.1
8.2 ± 6.3
c-Heptyl
H
H
c-Hexyl
129
130
131
15 ± 3.7
34 ± 2.9
14 ± 3.8
15 ± 4.7
33 ± 6.4
11 ± 2.1
H
H
H
F
F
132
133
27 ± 3.8
19 ± 5.6
19 ± 5.6
3.2 ± 2.9
H
H
F
F

134
135
136
33 ± 8.0
4.0 ± 4.0
15 ± 6.5
11 ± 2.3
0.4 ± 0.4
-0.7 ± 7.5
H
H
H
F3C
F3C
137
138
38 ± 4.5
22 ± 2.4
H
H
37 ± 2.8
27 ± 2.8
12 ± 2.2
26 ± 3.6
HO
HO
139
140
H
H
52 ± 6.2
(0.45 ± 0.07c)
27 ± 4.9
(1.5 ± 1.3)
1.9
1.9
3.4
O
O
141
142
143
H
H
H
63 ± 1.1
30 ± 8.1
(0.44 ± 0.05c)
(3.2 ± 0.03d) 1.6
O
54 ± 2.4
(0.75 ± 0.06)
63 ± 4.6
1.1
(2.1 ± 0.8c) 2.4
O
O
12 ± 8.7
22 ± 3.4
7.9 ± 8.6
23 ± 2.7
HN
. HCl
144
145
H
H
HN
O
68 ± 1.1
34 ± 8.4
(5.2 ± 2.9) 1.0
(0.49 ± 0.05c)
1.7
3.1
N
N
H
146
66 ± 1.3
(0.27 ± 0.03d)
H
26 ± 7.5
(1.9 ± 2.6) 2.7
O
CH3
147
148
149
150
151
46 ± 0.4
(0.75 ± 0.22)
H
H
H
H
H
32 ± 13
(11 ± 5.5)
1.1
1.0
2.9
1.0
0.5
O
O
S
46 ± 4.0
(0.86 ± 0.56)
50 ± 2.3
(2.0 ± 0.6c) 2.6
46 ± 1.8
(0.29 ± 0.10c)
18 ± 2.5
(1.2 ± 0.4d) 4.3
57 ± 4.1
(0.83 ± 0.12)
23 ± 2.3
(11 ± 6.7)
0.5
O2S
O
32 ± 5.9
3.5 ± 10
N

152
153
26 ± 8.4
10 ± 6.8
c-Hexyl
c-Hexyl
F
F
8.2 ± 3.0
-0.5 ± 1.3
154
155
14 ± 4.9
-1.2 ± 6.6
_
_
7.0 ± 1.1
3.0 ± 1.4
156
157
31 ± 6.0
29 ± 4.8
H
H
7.7 ± 1.9
10 ± 6.8
(CH2)2-
158
159
160
161
11 ± 3.2
26 ± 4.9
3.2 ± 10
44 ± 1.0
-6 ± 8.0
8.7 ± 6.8
4.8 ± 5.5
13 ± 4.6
H
H
H
H
(CH2)2
(CH2)2-
162
163
4.7 ± 7.1
8.8 ± 2.5
H
H
5.2 ± 2.0
6.5 ± 2.3
(CH2)2-
164
165
-4.4 ± 7.2
2.9 ± 3.2
F
F
-7.1 ± 4.0
-0.2 ± 2.4
(CH2)2-
(CH2)2
166
167
-1.9 ± 2.3
13 ± 1.3
H
H
-4.1 ± 3.7
4.3 ± 0.7
a'TBK1, (1 µM); IKK (5 µM). b Concentration for 50% inhibition in radiometric kinase assays using purified protein. Reported
as IC50 (µM) ± SEM (n=3). Significance relative to amlexanox determined by a two-tailed Student’s T-test (cp<0.05, dp<0.01.
eFold change over amlexanoc (5) calculated from the mean IC50 values. See Experimental Section.
Based on preliminary cellular and in vivo results with 125 (vide infra), we extended our SAR to
elaborated R7 cyclohexyl and, in selected cases, precursor olefin analogues. All modifications
were made on the distal position of the 6-membered ring substituent. Thus, gem-difluoro (131,
132), methyl (133), trifluoromethyl (134, 135), and gem-dimethyl (136, 137) modifications
decreased compound solubility that resulted in reduced inhibition. Hydroxylation (138, 139) also
reduced potency. Surprisingly, installation of distal ketone (140) and spiroketal (141) functions
improved potency toward both kinases whereas the unsaturated congener of spiroketal (142) only
significantly improved potency toward IKK.
Next, a series of analogues containing R7 nitrogen, oxygen, and sulfur heterocycles was
evaluated. The installation of a piperidinyl group at (143, 144) reduced inhibition toward both
kinases whereas formylation (145) or acetylation (146) of the heterocyclic nitrogen of 143

modestly improved potency toward both kinases. The installation of a tetrahydropyranyl group
(147) had no significant effect on potency but its olefinic congener (148) improved potency
toward IKK. In contrast, replacement of oxygen in 147 with sulfur (149) improved potency
toward both kinases, although no improvement was observed for the sulfone congener 150. The
morpholinyl analogue 151 displayed a decrease in inhibition toward TBK1 and was essentially
inactive toward IKK.
Table 3. In Vitro Activity of R3, R7 Disubstituted 5-Oxo-5H-[1]benzopyrano[2,3-b]pyridines
O
B
N
C
NH2
R3
A
R7
O
No. R3
R7
Single Concentration % Inhibitiona
(IC50, µM)b
TBK1
Fold
Fold
IKK
Changef
Changef
5
CO2H
i-Pr
50 ± 1.3
29 ± 7.4
(0.85 ±
0.14)
(5.1 ± 1.6)
168
169
c-Hexyl
64 ± 1.5
30 ± 2.2
N
(1.9 ± 0.2c)
0.4
1.7
1.6
1.6
1.1
(1.0 ± 0.3c)
5.1
24
NH
NH
N
N
N
52 ± 3.3
(0.49 ±
0.05c)
33 ± 0.6
(0.53 ±
0.16)
34 ± 3.6
(0.54 ±
0.10)
49 ± 0.5
(0.76 ±
0.11)
79 ± 2.1
N
(0.21 ± 0.06e)
N
170
171
172
Me2N
65 ± 1.3
N
(0.58 ± 0.2e)
8.8
16
NH
N
N
N
70 ± 1.1
N
N
O
(0.31 ± 0.14e)
NH
N
H
57 ± 1.7
N
N
N
O
(1.2 ± 0.4d)
4.3
NH
O
N
N
a'TBK1, (1 µM); IKK (5 µM). b Concentration for 50% inhibition in radiometric kinase assays using purified protein. Reported
as IC50 (µM) ± SEM (n=3). Significance relative to amlexanox determined by a two-tailed Student’s T-test (cp<0.05, dp<0.01,
ep<0.001). fFold change over amlexanoc (5) calculated from the mean IC50 values. See Experimental Section.
Analogues incorporating the carboxylic acid tetrazole bioisostere at the R3 position are shown in
Table 3. The combination of R3 tetrazole and R7 cyclohexyl moieties in 168 displayed the same
potency toward TBK1 as its R3 carboxylic acid congener 125 (Table 2), whereas its potency
toward IKK significantly increased. The olefin congener (169) exhibited a >20-fold increase in

potency toward IKK to 210 nM, consistent with prior observations that IKK favors unsaturated
substituents at R7. Significant increases in potency were also observed for the dimethylamine
(170) and morpholinyl (171, 172) analogues bearing R3 tetrazole or amide-linked tetrazole
modifications. Interestingly, 171 is a potent (310 nM) inhibitor of IKK despite its carboxylic
acid congener 151 (Table 2) showing no inhibitory effect, suggesting that the tetrazole moiety is
the primary driver of potency in this subclass of analogues.
5.
Crystallographic studies with select analogues
To verify the binding mode of analogues, select inhibitors were co-crystallized with TBK1
(Figure 2, Table SM1). To represent the series of analogues with extended non-cyclic R7
modifications, such as 117 and 120, the more soluble analogue 114 was co-crystallized with
TBK1 and the structure determined to 3.19 Å spacings (PDB entry 6CQ0, Figure 2A). The core
of the inhibitor binds identically to amlexanox along the kinase hinge. Weak density is observed
for the dimethylamino tail positioned near Pro90 and Cys91. The placement of the R7 tail in
solvent may explain why no significant changes in potency are observed, however these
analogues are generally less soluble which may also contribute to a reduction in potency. The
placement of the tail near Cys91, which is likely in the thiolate form due to its proximity to
Arg143, suggests that incorporation of an acrylamide in place of the dimethylamine in 114 may
serve as a strategy for covalent inhibitor design targeting Cys91. To represent cyclic R7
analogues, we determined the structure of TBK1 in complex with 131 at 3.2 Å and 150 at 3.35 Å
(PDB entries 6CQ4 and 6CQ5, respectively, Figure 2B-C). Both compounds bind along the
hinge similar to amlexanox and the R7 rings extend into solvent between the kinase and
dimerization domains of TBK1. The lack of interaction between the gem-difluorocyclohexyl
modification in 131 and TBK1, and decreased solubility, explains why this analogue displays
decreased inhibition of both kinases. The sulfone-containing ring on 150 also does not interact
with the kinase domain but is within range of Lys567 and Lys568 on the dimerization domain,
which would form favorable electrostatic interactions with the sulfone group.

Figure 2: TBK1 co-crystal structures with analogues (A) 114, (B) 131, and (C) 150. Mesh represents
positive |Fo| - |Fc| omit density contoured at 3 σ. The amino pyridine portion of the inhibitor core interacts
with the backbone of Glu87 and Cys89 in the hinge of the kinase and the carboxylate with Thr156 adjacent
to the DFG motif. The structures have been deposited as PDB entries 6CQ0, 6CQ4, and 6CQ5, respectively.
6.
Cell studies in 3T3-L1 adipocytes
To determine the efficacy of amlexanox analogues in cellular assays, we monitored
phosphorylation of p38 or TBK1 as well as secretion of IL-6 in 3T3-L1 adipocytes, which are
biomarkers of TBK1 and IKK inhibition.[23] TBK1 phosphorylation is increased as the result
of the blockade of a negative feedback loop involving AMPK and its target ULK1, which
directly phosphorylates TBK1[24]. Phosphorylation and activation of p38 occur as a result of de-
repression of cAMP levels through reduced activation of phosphodiesterase 3B (PDE-3B) by
TBK1 and IKK.[23] IL-6 secretion is not only a marker of the efficacy of TBK1/IKK
inhibition and p38 activation in adipocytes, but is also part of the biological response that confers
metabolic benefit in vivo.[25] An example of the western blots used to determine
phosphorylation levels is provided in Figure SM1. Enhanced phosphorylation of p38 relative to
amlexanox (5) is observed for the C-6 isomer 108 and R7 isobutyl homolog 116, whereas the R8
isomer 106 and 8-fluoro congener 109 show essentially equivalent activity to amlexanox (5).
Surprisingly, the 8-N congener of amlexanox (111) is almost completely non-stimulatory,
perhaps due to altered cell permeability. For TBK1 phosphorylation, the SAR is more mixed
with optimal enhancement by the R7 isopentyl homologue 117 and cyclohexyl analogue 125.
Additional enhancement was elicited by the R7 2-propenyl compound 105, C-8 isomer 106, R8
fluoro congener 109, unsaturated pyran (148) and thiopyran (149) analogues. In 3T3-L1
adipocytes, a number of analogues (105, 108, 109, 110, 116, 117, 125, 148, and 149) were
found to elicit IL-6 secretion as good as, or better than, amlexanox. For these, only 108 and 116

also increased p38 phosphorylation whereas 105, 108, 109, 117, and 148 additionally exhibited
increased TBK1 phosphorylation. Analogues 125 and 149 elicited an improved response in all
three readouts.
Table 4. Cellular Activity of Synthesized 5-Oxo-5H-[1]benzopyrano[2,3-b]pyridines and A-ring
Aza Congeners.
O
O
N
NH2
O
N
NH2
R8
R7
O
B
N
C
NH2
R3
N
A
CO2H
R7
CO2H
6
O
O
108
111, 112
No.
R7
R8
R3
p38 T180 / pTBK1
Y182a S172a
IL-6
secretionb
5
i-Pr
H
CO2H
CO2H
CO2H
CO2H
CO2H
CO2H
CO2H
CO2H
CO2H
CO2H
CO2H
CO2H
CO2H
1.00
1.00
1.00
1.11
0.96
0.20
2.33
1.36
1.20
0.19
0.22
0.48
2.56
2.46
0.42
105
106
107
108
109
110
111
112
113
116
117
118
CH2=C(CH3)-
H
0.48
0.96
0.79
1.80
0.93
0.41
0.03
0.03
0.47
1.4
1.42
1.23
0.52
1.20
1.30
0.61
0.04
0.04
0.23
0.46
3.40
0.20
H
i-Pr
H
-
CH2=C(CH3)-
-
i-Pr
F
F
-
CH2=C(CH3)-
i-Pr
CH2=C(CH3)-
Me2N
-
H
H
H
H
i-Bu
i-Pentyl
(CH3)2CHC C-
0.7
0.8
125
140
c-Hexyl
H
H
CO2H
CO2H
1.33
0.00
3.44
n.d.
3.58
0.07
O

146
H
CO2H
0.60
0.29
0.33
N
O
CH3
148
149
168
H
H
H
CO2H
CO2H
0.59
1.19
0.49
1.27
1.54
n.d.
1.53
1.55
0.17
O
S
c-Hexyl
N
NH
NH
NH
NH
N
N
N
N
N
169
170
171
H
H
H
0.10
0.00
0.70
n.d.
n.d.
0.39
0.00
0.03
0.54
N
N
Me2N
N
N
N
N
O
N
aPhosphorylation of p38 or TBK1 in 3T3-L1 adipocytes treated with 100 µM compound normalized to the
response to amlexanox (5). Values are determined from a single experiment with technical duplicates. bIL-6
secretion (pg/mL) from 3T3-L1 adipocytes treated with 100 µM compound normalized to the response to
amlexanox (5). Values are determined from a single experiment with technical triplicates. cn.d: not determined.
7.
Computed properties and aqueous solubility
A balance between lipophilicity and aqueous solubility are key parameters that influence success
in a standard oral, low-dose drug discovery program. Toward assessing drug-like properties
relating to oral absorption for our series of inhibitors, lipophilic ligand efficiency (LLE) and
solubility forecast indicator (SFI) calculators were utilized to derive computed values for
compounds in Table 4. Using cLogP and IC50 data, calculations of LLE can be calculated which
provide a measure of a compound’s lipophilicity associated with potency.[26],[27] That, along
with SFI calculations[28] derived from cLogD7.4 information, provide valuable information
regarding a compound’s potential to be a drug. The data are shown in Table SM2, with equations
for LLE and SFI calculations given in the footnotes. Computed LLE values range from 4.77 –
1.30 vs TBK1 and 3.92 – 0 vs IKK, imbuing several compounds with favorable properties
relative to amlexanox (5) (LLE of 2.76 and 1.98, respectively). Utilizing cLogD7.4 in place of
clogP in the equation[27] markedly enhances LLE for almost all compounds (calculations not
shown). With regard to aqueous solubility, clogD7.4 values together with the number of aromatic

rings in a given molecule provide a simple, useful guide of potential solubility of particular
molecules with values ≤ 6 predicting for good aqueous solubility. Utilizing the SFI calculator, all
compounds in Table SM2 are predicted to have acceptable aqueous solubility (SFI of 5.83 –
1.80) with many displaying lower SFI than amlexanox (3.32). These calculations correlate
loosely with measured thermodynamic aqueous solubility values for selected compounds of
Table SM2 (916 - <1.1 μM vs 172 μM for amlexanox). Measured solubility is highest for
compounds with 2-propenyl (259 μM for 105; 373 μM for 112) and tetrazole (916 μM for 171)
substituents.
8.
In vivo studies
Figure 3: In vivo efficacy of 125 relative to amlexanox (5) and vehicle control in a 4-week
trial. Animals were treated by daily oral gavage with 25 mg/kg, (n = 7 per treatment group).
This sample size was calculated to be sufficient to detect a significant IL-6 response relative
to vehicle (if efficacy is at least 50% of amlexanox), and a 50% change in IL-6 response
relative to amlexanox at a power level of 0.80. A) Serum IL-6 response after 6 h of
treatment. B) Body weight over the 4-week treatment period. C) Glucose tolerance test after
3 weeks of treatment. D) Insulin tolerance test after 4 weeks of treatment. Statisitcal
significance calculated using a one-way ANOVA and two-way ANOVA (time being the
second factor), with a Holm-Sidak calculation of signifigance adjusted for multiple
comparisons. * indicates a p-value < 0.05 versus vehicle (red for 5 and green for 125).
Amlexanox (5) has been shown to reverse insulin resistance, inflammation, and hepatic steatosis
in mice on a high fat diet.[29] Increased serum IL-6 levels are a robust biomarker of drug
efficacy, and have been shown to be responsible for reducing fasting blood glucose levels in

these mice and humans.[10],[25] Analogue 125, which produced the most robust IL-6 production
in 3T3-L1 cells, as well as a phosphorylation marker of efficacy, was investigated in obese mice
and exhibited a serum IL-6 response similar to amlexanox (Figure 3A). Both amlexanonx and
125 treatment resulted in a significant increase in serum IL-6 levels, but there was no significant
difference between amlexanonx and 125. Further investigation of 125 showed a comparable
weight loss effect over the course of 4 weeks (Figure 3B). After 4 weeks the body weight of
amlexanox and 125 treated animals were both significantly lower than the vehicle controls,
without any significant difference between the two treated groups. Intermediate glucose response
(Figure 3C) and insulin sensitizing effect (Figure 3D) were observed in animals treated for 3-4
weeks. Analysis by one-way ANOVA revealed that all three groups were significantly different
from each other in both the insulin and glucose tolerance tests. Overall, the effect of treatment
with 125 closely resembles that of amlexanox, although with slightly altered kinetics.
9.
Conclusions
As part of a research campaign to explore a new hypothesis on the role of inflammation in the
generation of insulin resistance and type 2 diabetes, we have been focusing efforts on the design
of small molecule inhibitors of TBK1 and IKK kinases, which are implicated as key mediators
in insulin-independent pathways that promote energy storage and block adaptive energy
expenditure during obesity. Toward that end, we are utilizing the clinically approved drug
amlexanox (5), a modest inhibitor of TBK1 and IKK, as a starting point for analogue synthesis
in order to develop an SAR that addresses some of its liabilities, principally modest potency and
non-optimal aqueous solubility.
We previously reported the TBK1·amlexanox crystal structure, which revealed canonical ATP-
competitive binding along the kinase hinge and demonstrated the importance of the carboxylic
acid function for potent inhibition.[11] One of the goals of our work has been to improve on
literature synthetic methodologies to make amlexanox analogues wherein A-ring substituents
almost always had been incorporated early into a precursor 4-oxo-4H-l-benzopyran-3-
carbonitrile. To introduce such substituents at a later stage and thus enhance overall route
efficiency, we developed a novel palladium-catalyzed cross-coupling strategy on bromo
precursors 6 – 10 to incorporate a range of R7 and R8 substituents. Due to the base sensitivity of
key precursors using this strategy, it was necessary to devise a modified route to the C-6 isomer
(108) of amlexanox.
Analogues with an olefinic substituent favor binding to IKK relative to saturated congeners,
possibly due to the planarity of the sp2 carbons (compare 105 vs 5, 110 vs 109, and 126 vs 125).
The presence of a serine residue at position 90 in the hinge of IKK in contrast to a proline in
TBK1, may affect the flexibility of the hinge region and, in conjunction with a planar olefinic
moiety, allow for more kinase domain closure and improved binding. This trend is in line with

prior SAR for the R3 position of amlexanox,[11] but does not apply to 8-aza analogues (112 vs
111) wherein the isopropyl congener 111 shows a greater fold change in binding relative to
amlexanox. For binding to 8-aza analogues, the hinge region may assume a conformation that
allows for a hydrogen bond to form with the kinase backbone. This possibility is supported by
the TBK1·amlexanox crystal structure wherein the hinge would need to change conformation in
order for the 8-nitrogen to interact with the backbone.
Based on prior structural studies with amlexanox, cyclic R7 modifications were predicted to
extend through solvent and contact the neighboring dimerization domain. Docking experiments
indicated the potential for global rearrangement of the inhibitor binding mode in such molecules.
However, two independent co-crystal structures show that the inhibitor core of these analogues
binds similarly to amlexanox along the hinge. R7 cyclic substituents containing heteroatoms at
the distal position (e.g., 149 and 150) may interact with Tyr564, Lys570, or Lys576. More
extended ring modifications such as acetylated piperidine (146) and spiroketal (141, 142) may
form contacts with the backbone of the dimerization domain.
The R7 position tolerates ring sizes up to cyclohexyl after which potency markedly decreases,
likely as a result of poor solubility and excessive bulk near the kinase hinge and P-loop. In line
with the prior observation that inhibition of IKKε is accentuated with olefinic substituents
attached to the R7 position, analogue 126 retains inhibition of IKKε while TBK1 inhibition is
decreased. Appending saturated rings via short aliphatic linkers was detrimental to inhibition of
both kinases due to poor solubility.
Some of the most striking results from our SAR campaign are shown by selected R7 derivatives
of Tables 1 and 2 that were further modified with the R3 tetrazole, a carboxylic acid bioisostere.
An analogue of amlexanox (5) bearing a tetrazole at the R3 position was previously described to
display a potency of 200-300 nM toward both TBK1 and IKKε.[11] When combined with a 6-
membered hydrocarbon ring (168, 169), dimethyamino (170), or morpholinyl (171) substituents,
there were marginal changes in TBK1 potency relative to amlexanox. Unexpectedly, potency
toward IKKε was improved by nearly 25-fold (to 210 nM) compared to amlexanox for
cyclohexenyl analogue 169, suggesting that potency for IKKε is optimized with a planar R7
substituent near the hinge along with a R3 tetrazole function. This potency is comparable to the
R3 tetrazole derivative of amlexanox suggesting that interactions with the gatekeeper Met86 and
increased buried surface area drive potency even when a less favorable modification, such as a
R7 morpholine (171), is present. Thus, selectivity toward IKKε for dual modified analogues may
stem from increased hinge flexibility, allowing for the accommodation of two additional cyclic
functions off the amlexanox pharmacophore. Conversely, the installation of an amide-linked
tetrazole at the C-3 position was previously shown to decrease potency toward TBK1.[11]
However, when combined with a R7 morpholinyl (172), potency is comparable to that of
amlexanox.
In 3T3-L1 adipocytes, amlexanox (5) treatment promotes phosphorylation of TBK1 and p38, the

latter of which induces beneficial IL-6 secretion. Movement of the isopropyl group to the C-6
position in 108 enhances p38 phosphorylation and downstream IL-6 secretion. Because the C-6
position of the pharmacophore projects toward the unoccupied P-loop, further SAR around this
position may be more productive, despite a slight reduction in enzyme potency for 108 relative to
amlexanox. Elongation of the isopropyl function to R7 isobutyl (116) and isopentyl (117) greatly
enhances IL-6 secretion and also enhances p38 or TBK1 phosphorylation. This result is
surprising as both analogues display enzyme potencies comparable to amlexanox, even with a
computed loss of solubility and lipophilic ligand efficiency (Table SM2). However, both
inhibitors are predicted to have improved cellular permeability and may more readily enter
cultured adipocytes. Analogue 148 bearing a R7 unsaturated tetrahydropyran moiety elicited a
similar IL-6 response despite the compound having a lower cLogP value. However, it is
significantly more potent toward IKK, which may compensate for a lower permeability.
Most profound are the cellular effects by the R7 cyclohexyl analogue 125, which produced the
greatest level of TBK1 phosphorylation and IL-6 secretion. Along with 117, 125 is predicted to
be highly cell permeable, which likely contributes to its efficacy. Interestingly, 125 is a poor
enzyme inhibitor of IKK, displaying only minimal inhibition at 100 µM. Thus, its cellular
effects likely stem from inhibition of TBK1. When combined with a R3 tetrazole moiety in 168,
the cellular response is abolished. In fact, all analogues bearing dual R3 and R7 modifications fail
to elicit a cellular response. This observation is in line with a previous report that the
introduction of an R3 tetrazole without modification of the R7 position eliminates cellular
activity.[11] The lack of cellular response may stem from complex permeability effects that
abrogate the beneficial effects of R7 cyclohexyl incorporation.
In vivo experiments with 125 showed comparable efficacy to amlexanox (5) in obese mice. In
contrast to robust secretion in cultured adipocytes, IL-6 production by 125 was only comparable
in vivo to that of amlexanox (5). Similar effects on hepatic insulin response and glucose uptake
were observed, although the 125 response lagged relative to amlexanox at some time points
suggesting that the presence of a R7 cyclohexyl moiety alters the kinetics of drug absorption and
distribution. However, the endpoints for 125 were the same as amlexanox suggesting that the
observed efficacy for these two inhibitors may be approaching a ceiling with respect to inhibition
of TBK1 and IKK within the R7 subclass of analogues
In summary, future exploration of the amlexanox pharmacophore will be greatly aided by the
synthetic methodology, co-crystal structure determinations, and SAR reported herein. Whereas
several biomarkers in adipocytes of TBK1 and IKK are described, it remains to be seen whether
the efficacy of amlexanox or 125 represents the maximum response that can be achieved through
inhibition of these targets. As the C-6 isomer of amlexanox displayed improved cellular
response, the installation of additional modifications at this position presents an attractive
direction toward enhancing enzyme potency through burying surface area under the P-loop. That
along with careful consideration of physicochemical properties, such as permeability and

solubility, will aid in the identification of inhibitors that might produce a more efficacious
response in vivo. Further SAR exploration of the amlexanox pharmacophore, as well as truncated
core congeners, is ongoing and will reveal whether these efforts lead to clinical candidates for
the treatment of type 2 diabetes and/or obesity.
10.
Experimental
Chemistry. Materials and Methods. All starting monomers were obtained from commercial
suppliers and were used without further purification. Amlexanox, >99% pure by HPLC, was
purchased from Ontario Chemicals, Inc. 2-Amino-7-bromo-5-oxo-5H-[1]benzopyrano[2,3-
b]pyridine-3-carbonitrile was purchased from Sigma-Aldrich. Routine 1H NMR spectra were
recorded at 400 or 500 MHz on a Varian 400 or 500 instrument, respectively, with CDCl3 or
DMSO-d6 as solvent. Chemical shift values are recorded in  units (ppm). Mass spectra were
recorded on a Micromass TofSpec-2E Matrix-Assisted, Laser-Desorption, Time-of-Flight Mass
Spectrometer in a positive ESI mode (TOFES+) unless otherwise noted. High resolution mass
spectrometry (HRMS) analysis was performed on an Agilent Q-TOF system. Analytical HPLC
was performed on an Agilent 1100 series instrument with an Agilent Zorbax Eclipse Plus C18
(4.6 mm × 75 mm, 3.5 μm particle size) column with the gradient 10% acetonitrile/water (1
min), 10−90% acetonitrile/water (6 min), and 90% acetonitrile/water (2 min) flow = 1 mL/min.
Thin-layer chromatography (TLC) was performed on silica gel GHLF plates (250 µm) purchased
from Analtech. Column chromatography was carried out in the flash mode utilizing silica gel
(220−240 mesh) purchased from Silicycle. Extraction solutions were dried over anhydrous
sodium sulfate prior to concentration.
Method A: Palladium-catalyzed cross coupling reactions
(a)
Suzuki couplings[17]
1.1
1-(2-Hydroxy-6-(prop-1-en-2-yl)phenyl)ethan-1-one (92; Scheme 3). A solution of
nitrogen-degassed potassium carbonate (964 mg, 6.98 mmol) in water (1.2 mL) was added to a
nitrogen-degassed mixture of compound 85 (500 mg, 2.33 mmol), triphenylphosphine (18.3 mg,
0.070 mmol), potassium trifluoro(prop-1-en-2-yl)borate, (475 mg, 3.21 mmol), and PdCl2(PPh3)2
(16.3 mg, 0.023 mmol) in tetrahydrofuran (3.6 mL). The reaction vial was sealed and heated at
100°C for 18 h. The mixture was cooled to room temperature and diluted with saturated aqueous
ammonium chloride and then extracted with dichloromethane (3x) and. The combined organic
phases were washed by brine, dried, and concentrated to give 92 (0.40 g, 98%) as an orange oil.
1H NMR (500 MHz, chloroform-d): δ 11.48 (s, 1H), 7.32 (ddd, J = 8.2, 7.5, 0.6 Hz, 1H), 6.89
(ddd, J = 8.4, 1.3, 0.6 Hz, 1H), 6.76 (ddd, J = 7.5, 1.2, 0.6 Hz, 1H), 5.24 (dt, J = 2.0, 1.0 Hz, 1H),
4.97 (dt, J = 1.7, 0.9 Hz, 1H), 2.57 (d, J = 0.6 Hz, 3H), 2.17 (dt, J = 1.5, 0.7 Hz, 3H). MS: m/z

177.0908 (M+H)+.
1.2
2-Amino-7-(cyclohex-1-en-1-yl)-5-oxo-5H-chromeno[2,3-b]pyridine-3-carbonitrile
(102; Scheme 4). A solution of tripotassium phosphate (701 mg, 3.3 mmol) in water (1.5 mL)
was added to a nitrogen-degassed suspension of 2-amino-7-bromo-5-oxo-5H-chromeno[2,3-
b]pyridine-3-carbonitrile (101; 174 mg, 0.55 mmol), 2-(cyclohex-1-en-1-yl)-4,4,5,5-tetramethyl-
1,3,2-dioxaborolane (252 mg, 1.21 mmol), triphenylphosphine (72.2 mg, 0.28 mmol), and
Pd(OAc)2 (12.4 mg, 0.06 mmol) in p-dioxane (13 mL). The reaction vessel was sealed with a
screw cap and heated at 100 -110 °C for 30 min and then stirred at 70 oC for 1.5 h. After cooling,
the mixture was concentrated and the residue was diluted with a mixture of water and
dichloromethane. Precipitated solids were collected and the organic phase of the biphasic filtrate
was dried and concentrated to leave additional solids. The combined solids were triturated in
dichloromethane and dried to give 102 (136 mg, 78%) as an orange solid. 1H NMR (400 MHz,
DMSO-d6): δ 8.61 (s, 1H), 8.13 (s, 2H), 7.97 (d, J = 2.3 Hz, 1H), 7.89 (dd, J = 8.8, 2.4 Hz, 1H),
7.52 (d, J = 8.7 Hz, 1H), 6.29 (s, 1H), 2.41 (s, 2H), 2.21 (s, 2H), 1.82 – 1.70 (m, 2H), 1.62 (t, J =
5.8 Hz, 2H). MS: m/z 318.0 (M+H)+.
See Supplementary Material for additional experimental procedures: 15-18, 24 (Scheme SM1);
32, 36-52, 55-58 (Scheme SM2)
(b)
Sonogashira coupling[18]
1.1
Ethyl 2-amino-7-(3-methylbut-1-yn-1-yl)-5-oxo-5H-chromeno[2,3-b]pyridine-3-
carboxylate (19; Scheme SM1). A suspension of compound 9 (300 mg, 0.83 mmol), 3-
methylbut-1-yne (338 mg, 4.96 mmol), triphenylphosphine (21.7 mg, 0.08 mmol), N,N-
dimethylformamide (8 mL), tetrahydrofuran (8 mL), and triethylamine (7.5 mL) in a 100 mL
sealed tube was degassed with nitrogen. Then Pd(PPh3)4 (47.7 mg, 0.04 mmol) and CuI (15.7
mg, 0.08 mmol) were added under an atmosphere of nitrogen and the suspension was heated at
85 °C for 18 h. The mixture was diluted with water and extracted with dichloromethane (2x).
The organic phases were dried and concentrated to leave a solid that was triturated in ethanol and
then diethyl ether, and dried to give 19 (246 mg, 85%) as a solid. 1H NMR (400 MHz,
chloroform-d) δ 9.14 (s, 1H), 8.38 (s, 1H), 8.28 (d, J = 2.1 Hz, 1H), 7.69 (dd, J = 8.6, 2.1 Hz,
1H), 7.41 (d, J = 8.6 Hz, 1H), 5.92 (s, 1H), 4.40 (q, J = 7.1 Hz, 2H), 2.80 (p, J = 6.9 Hz, 1H),
1.43 (dd, J = 7.6, 6.7 Hz, 3H), 1.29 (dd, J = 7.0, 1.0 Hz, 6H). MS m/z 351.1341 (M+1)+.
1.2
Ethyl 2-amino-8-fluoro-7-(3-methylbut-1-yn-1-yl)-5-oxo-5H-chromeno[2,3-
b]pyridine-3-carboxylate (20; Scheme SM1). Similar reaction of 7 (40 mg, 0.11 mmol), 3-
methylbut-1-yne (43 mg, 0.63 mmol), triphenylphosphine (2.8 mg, 10.5 umol), Pd(PPh3)4 (6.1
mg, 5.3 μmol), CuI (2.0 mg, 10.5 μmol), triethylamine (2 mL), N,N-dimethylformamide (2 mL),
and tetrahydrofuran (2 mL) gave 20 (26 mg, 67%) as a solid. 1H NMR (500 MHz, chloroform-d)

δ 9.12 (s, 1H), 8.40 (s, 1H), 8.32 (d, J = 7.8 Hz, 1H), 7.17 (d, J = 9.2 Hz, 1H), 5.89 (s, 1H), 4.40
(q, J = 7.1 Hz, 2H), 2.83 (q, J = 6.9 Hz, 1H), 1.43 (t, J = 7.2 Hz, 3H), 1.35 - 1.26 (m, 6H). MS
m/z 369.12 (M+1)+.
1.3
Ethyl 2-amino-7-(3-(dimethylamino)prop-1-yn-1-yl)-5-oxo-5H-chromeno[2,3-
b]pyridine-3-carboxylate (21; Scheme SM1). Similar reaction of 9 (400 mg, 1.10 mmol), 3-
dimethylamino-1-propyne (275 mg, 3.3 mmol), CuI (42 mg, 0.22 mmol), (Ph3P)2PdCl2 (77 mg,
0.11 mmol), triethylamine (20 mL), and N,N-dimethylformamide (6 mL) at 80 °C overnight gave
a residue that was purified by flash silica gel chromatography, eluting with 3% methanol in
dichloromethane, to provide 21 (320 mg, 80%) as light orange solid following sonication in
methanol. 1H NMR (400 MHz, chloroform-d) δ 9.15 (s, 1H), 8.40 (s, 1H), 8.33 (d, J = 2.0 Hz,
1H), 7.82 - 7.68 (m, 1H), 7.44 (d, J = 8.6 Hz, 1H), 5.88 (s, 1H), 4.49 - 4.33 (m, 2H), 3.50 (s, 2H),
2.39 (s, 6H), 1.44 (td, J = 7.1, 1.1 Hz, 3H). MS m/z 366.0 (M+H)+.
1.1
Ethyl 2-amino-7-(cyclopropylethynyl)-5-oxo-5H-chromeno[2,3-b]pyridine-3-
carboxylate (53; Scheme SM2). Similar reaction of 9 (500 mg, 1.38 mmol),
ethynylcyclopropane (364 mg, 5.51 mmol), triphenylphosphine (36.1 mg, 0.14 mmol), N,N-
dimethylformamide (10 mL), tetrahydrofuran (10 mL), triethylamine (10 mL), Pd(PPh3)4 (80
mg, 0.07 mmol) and CuI (26.2 mg, 0.14 mmol) gave 53 (452 mg, 94%) as solid. 1H NMR (400
MHz, chloroform-d) δ 9.14 (s, 1H), 8.39 (s, 1H), 8.26 (s, 1H), 7.67 (dd, J = 8.5, 2.1 Hz, 1H),
7.40 (d, J = 8.6 Hz, 1H), 5.91 (s, 1H), 4.40 (q, J = 7.1 Hz, 2H), 1.52 - 1.46 (m, 1H), 1.43 (t, J =
7.2 Hz, 3H), 0.90 (d, J = 8.1 Hz, 2H), 0.88 - 0.80 (m, 2H). MS m/z 349.12 (M+1)+.
2.
Ethyl 2-amino-7-(cyclobutylethynyl)-5-oxo-5H-chromeno[2,3-b]pyridine-3-
carboxylate (54; Scheme SM2). Similar reaction of 9 (200 mg, 0.55 mmol), triphenylphosphine
(14.4 mg, 0.06 mmol), N,N-dimethylformamide (4 mL), tetrahydrofuran (4 mL), triethylamine (4
mL), ethynylcyclopropane[30] (220 mg, 2.75 mmol), Pd(PPh3)4 (31.7 mg, 0.03 mmol) and CuI
(10.5 mg, 0.06 mmol) gave 54 (173 mg, 87%) as a solid. The compound was used directly in the
next step.
(c)
Negishi coupling[19]
2.1
Ethyl 2-((tert-butoxycarbonyl)amino)-7-isobutyl-5-oxo-5H-chromeno[2,3-b]pyridine-
3-carboxylate (22; Scheme SM1). A mixture of isobutylzinc(II) bromide solution (7.5 ml, 0.5
M in tetrahydrofuran, 3.75 mmol), compound 10 (300 mg, 0.65 mmol) and PdCl2(dppf) · CH2Cl2
(53 mg, 0.07 mmol) was degassed with nitrogen and the reaction vial capped. The mixture was
heated at 67 °C for 16 h, cooled, diluted with water, and extracted with dichloromethane (2x).

The combined organic phases were washed with saturated aq NaHCO3 and then brine, dried, and
concentrated to leave a solid that was purified by silica gel preparative plate chromatography
(elution with 0.2% methanol in dichloromethane) to afford 22 (72 mg, 25%). 1H NMR (400
MHz, chloroform-d) δ 10.89 (s, 1H), 9.29 (d, J = 0.8 Hz, 1H), 8.04 (dd, J = 1.8, 0.9 Hz, 1H),
7.54 (t, J = 1.3 Hz, 2H), 4.46 (q, J = 7.1 Hz, 2H), 2.61 (d, J = 7.2 Hz, 2H), 1.94 (dt, J = 13.6, 6.8
Hz, 1H), 1.58 (s, 9H), 1.47 (t, J = 7.2 Hz, 3H), 0.94 (d, J = 6.6 Hz, 6H). MS m/z 441.2032
(M+H)+.
2.2
Ethyl 2-((tert-butoxycarbonyl)amino)-7-cyclobutyl-5-oxo-5H-chromeno[2,3-
b]pyridine-3-carboxylate (34; Scheme SM2). Similar reaction of 10 (200 mg, 0.43 mmol),
cyclobutylzinc(II) bromide (5.0 ml of 0.5 M solution in tetrahydrofuran, 2.5 mmol),
Pd(dppf)Cl2·CH2Cl2 (35.3 mg, 0.04 mmol) gave 34 (50 mg, 26%). 1H NMR (400 MHz,
chloroform-d): δ 10.89 (s, 1H), 9.30 (d, J = 0.9 Hz, 1H), 8.10 (d, J = 2.3 Hz, 1H), 7.68 – 7.57 (m,
1H), 7.54 (d, J = 8.6 Hz, 1H), 4.47 (q, J = 7.1 Hz, 2H), 3.66 (t, J = 8.9 Hz, 1H), 2.42 (dtd, J =
10.4, 7.9, 2.4 Hz, 2H), 2.27 – 2.14 (m, 2H), 2.14 – 1.98 (m, 1H), 1.95 – 1.84 (m, 1H), 1.58 (s,
9H), 1.47 (t, J = 7.1 Hz, 3H). MS TOFES+: m/z 439.1 (M+H)+.
(d)
Stille coupling[20]
2.3
4-Oxo-5-(prop-1-en-2-yl)-4H-chromene-3-carbonitrile (89; Scheme 3). A mixture of
compound 88 (1.0 g, 4.0 mmol), tri-n-butyl(prop-1-en-2-yl)stannane (1.46 g, 4.4 mmol), and
Pd(PPh3)2Cl2 (281 mg, 0.4 mmol) in a vial was flushed with nitrogen. Dry p-dioxane (20 mL)
was added and the mixture was degassed and flushed with nitrogen. This vial was sealed and
heated at 100 °C for 20 min and then at 80 °C for 16 h after which the mixture was diluted with
water and extracted with dichloromethane. The organic phase was dried and concentrated to
leave a residue that was purified by preparative plate silica gel chromatography (25-50% ethyl
acetate in hexanes) to give 89 (680 mg, 81%) as a solid. 1H NMR (400 MHz, chloroform-d): δ
8.36 (s, 1H), 7.69 (dd, J = 8.5, 7.4 Hz, 1H), 7.47 (dd, J = 8.5, 1.2 Hz, 1H), 7.24 (dd, J = 7.4, 1.2
Hz, 1H), 5.18 (t, J = 1.6 Hz, 1H), 4.89 - 4.78 (m, 1H), 2.09 (t, J = 1.1 Hz, 3H). MS: m/z
212.0702 (M+H)+.
(e)
Buchwald coupling[21]
2.4
1-(5-(Dimethylamino)-2-hydroxyphenyl)ethan-1-one (97a; Scheme 4). To a nitrogen-
degassed solution of 96 (3.0 g, 13.95 mmol) was added dimethylamine (27.9 ml, 2M in
tetrahydrofuran, 55.8 mmol), lithium hexamethyldisilazide (48.8 mL, 1M in tetrahydrofuran,
48.8 mmol), Pd2(dba)3 (204 mg, 0.223 mmol), and 2-dicyclohexylphosphino-2’-N,N-
(dimethylamino)biphenyl (DavePhos; 176 mg, 0.446 mmol). The mixture was degassed again

with nitrogen, sealed in a capped vial and stirred at 80 °C for 2 h. After cooling, the mixture was
concentrated to a residue that was distributed between dichloromethane and saturated aqueous
ammonium chloride solution. The organic phase was washed with brine and concentrated to give
crude 97a (2.5 g, 100%) as a solid. 1H NMR (400 MHz, chloroform-d) δ 11.74 (s, 1H), 7.08 (dd,
J = 9.0, 3.1 Hz, 1H), 7.01 (d, J = 3.0 Hz, 1H), 6.91 (d, J = 9.0 Hz, 1H), 2.89 (d, J = 1.1 Hz, 6H),
2.62 (d, J = 1.1 Hz, 3H). MS m/z 180.1 (M+H)+.
2.5
1-(2-Hydroxy-5-morpholinophenyl)ethan-1-one (97b; Scheme 4). Reaction of 96 (2.0
g, 9.3 mmol) in tetrahydrofuran (15 mL), morpholine (1.3 g, 14.9 mmol), lithium
hexamethyldisilazide (32.6 mL, 1 M in tetrahydrofuran, 32.6 mmol), Pd2(dba)3 (136 mg, 0.15
mmol), and 2-dicyclohexylphosphino-2’-N,N-(dimethylamino)biphenyl (DavePhos; 117 mg,
0.30 mmol) at 75 °C for 1 h, as described above for the synthesis of 97a, gave an oil that was
purified by flash silica gel chromatography, eluting with 0.5:1:100 trimethylamine solution (25%
w/w solution in methanol) : methanol : dichloromethane to give 97b (1.67 g, 81%) as a solid. 1H
NMR (400 MHz, chloroform-d): δ 11.92 (s, 1H), 7.22 – 7.18 (m, 2H), 6.97 – 6.92 (m, 1H), 3.93
– 3.82 (m, 4H), 3.10 – 3.01 (m, 4H), 2.63 (s, 3H). MS TOFES+: m/z 222.1 (M+H)+.
Method B: Formation of Boc derivatives of ethyl 2-amino-5-oxo-5H-chromeno[2,3-b]pyridine-
3-carboxylates[31]
2.6
Ethyl 7-bromo-2-((tert-butoxycarbonyl)amino)-5-oxo-5H-chromeno[2,3-b]pyridine-
3-carboxylate (10; Scheme SM1). To a suspension of compound 9 (50 mg, 0.14 mmol, and 4-
dimethylaminopyridine (1 mg, 8.2 umol) in dichloromethane (5 mL) was added di-tert-butyl
dicarbonate (33 mg, 0.15 mmol) portion-wise. The mixture was stirred at room temperature for
16 h. The solid was collected by filtration, washed with dichloromethane (2 mL) and dried to
afford 10 (53 mg, 83%) as a white solid. 1H NMR (400 MHz, chloroform-d): δ 10.91 (s, 1H),
9.28 (d, J = 0.9 Hz, 1H), 8.41 (d, J = 2.4 Hz, 1H), 7.83 (dd, J = 8.9, 2.5 Hz, 1H), 7.51 (d, J = 8.9
Hz, 1H), 4.46 (q, J = 7.1 Hz, 2H), 1.58 (d, J = 0.9 Hz, 9H), 1.47 (t, J = 7.1 Hz, 3H). MS TOFES-:
m/z 460.9, 462.9 (M-H)-.
Method C: TFA deprotection of ethyl 2-((tert-butoxycarbonyl)amino)-5-oxo-5H-chromeno[2,3-
b]pyridine-3-carboxylates[32]
2.7
Ethyl 2-amino-7-isobutyl-5-oxo-5H-chromeno[2,3-b]pyridine-3-carboxylate (23;
Scheme SM1). To a solution of compound 22 (71 mg, 0.16 mmol) in dichloromethane (10 mL)
was added trifluoroacetic acid (0.25 mL) and the mixture was stirred for 15 min at room
temperature. The mixture was washed with saturated aq NaHCO3, dried, and concentrated to
give 23 (53 mg, 97%) as a solid. 1H NMR (400 MHz, chloroform-d) δ 9.15 (s, 1H), 8.36 (s, 1H),

8.03 (d, J = 2.1 Hz, 1H), 7.54 - 7.49 (m, 1H), 7.42 (d, J = 8.5 Hz, 1H), 5.93 (s, 1H), 4.41 (q, J =
7.1 Hz, 2H), 2.60 (d, J = 7.2 Hz, 2H), 1.93 (dt, J = 13.5, 6.8 Hz, 1H), 1.44 (t, J = 7.1 Hz, 3H),
0.93 (d, J = 6.6 Hz, 6H). MS m/z 341.1522 (M+H)+.
2.8
Ethyl 2-amino-7-cyclopropyl-5-oxo-5H-chromeno[2,3-b]pyridine-3-carboxylate (33;
Scheme SM2). Similar reaction of 32 (59 mg, 0.14 mmol), trifluoroacetic acid (0.5 mL), and
dichloromethane (20 mL) gave 33 (45 mg, 100%) as a white solid. 1H NMR (400 MHz, DMSO-
d6): δ 8.80 (s, 1H), 8.34 (s, 1H), 8.09 (s, 1H), 7.77 (d, J = 2.2 Hz, 1H), 7.58 – 7.42 (m, 2H), 4.35
(q, J = 7.1 Hz, 2H), 2.11 (dt, J = 8.5, 3.7 Hz, 1H), 1.36 (t, J = 7.1 Hz, 3H), 1.07 – 0.92 (m, 2H),
0.79 – 0.65 (m, 2H). MS TOFES+: m/z 325.0 (M+H)+.
2.9
Ethyl 2-amino-7-cyclobutyl-5-oxo-5H-chromeno[2,3-b]pyridine-3-carboxylate (35;
Scheme SM2). Similar reaction of 34 (50 mg, 0.11 mmol), trifluoroacetic acid (0.5 mL), and
dichloromethane (20 mL) for 1 h gave 35 (39 mg, 100%) as a white solid. 1H NMR (400 MHz,
chloroform-d): δ 9.15 (s, 1H), 8.36 (s, 1H), 8.09 (d, J = 2.2 Hz, 1H), 7.56 (ddd, J = 8.5, 2.3, 0.7
Hz, 1H), 7.43 (d, J = 8.5 Hz, 1H), 5.91 (s, 1H), 4.41 (q, J = 7.1 Hz, 2H), 3.65 (p, J = 8.8 Hz, 1H),
2.41 (dtd, J = 10.5, 8.0, 2.4 Hz, 2H), 2.21 (pd, J = 9.6, 2.0 Hz, 2H), 2.12 – 2.01 (m, 1H), 1.95 –
1.82 (m, 1H), 1.44 (t, J = 7.1 Hz, 3H). MS TOFES+: m/z 339.1 (M+H)+.
2.10
Ethyl 2-amino-7-cycloheptyl-5-oxo-5H-chromeno[2,3-b]pyridine-3-carboxylate (62;
Scheme SM2). Similar reaction of 61 (105 mg, 0.23 mmol), trifluoroacetic acid (0.5 mL) and
dichloromethane (20 mL) for 30 min gave 62 (82 mg, 94%) as a solid. 1H NMR (400 MHz,
chloroform-d): δ 9.15 (s, 1H), 8.35 (s, 1H), 8.06 (d, J = 2.3 Hz, 1H), 7.55 (dd, J = 8.6, 2.3 Hz,
1H), 7.41 (d, J = 8.6 Hz, 1H), 5.88 (s, 1H), 4.40 (q, J = 7.2 Hz, 2H), 2.87 – 2.74 (m, 1H), 1.97 –
1.88 (m, 2H), 1.82 (dt, J = 9.5, 3.3 Hz, 2H), 1.72 (td, J = 10.3, 3.4 Hz, 4H), 1.61 (d, J = 6.5 Hz,
4H), 1.44 (t, J = 7.1 Hz, 3H). MS TOFES+: m/z 381.1 (M+H)+.
2.11 Ethyl 2-amino-5-oxo-7-(piperidin-4-yl)-5H-chromeno[2,3-b]pyridine-3-carboxylate,
trifluoroacetic acid salt (71; Scheme SM2). Similar reaction of 70 (100 mg, 0.21 mmol),
trifluoroacetic acid (0.25 mL), and dichloromethane (10 mL) for 5 h at 40 oC, but without the
extractive workup, gave crude 71 (115 mg, 100%) as an off-white solid. It was used directly in
the next step. 1H NMR (400 MHz, chloroform-d): δ 9.17 (s, 1H), 8.70 (s, 1H), 8.62 (s, 1H), 8.20
(s, 1H), 8.13 (d, J = 2.2 Hz, 1H), 7.63 (d, J = 2.3 Hz, 1H), 7.55 (s, 1H), 4.42 (q, J = 7.1 Hz, 2H),
3.70 (d, J = 12.7 Hz, 2H), 3.18 (d, J = 12.0 Hz, 2H), 3.00 (d, J = 10.9 Hz, 1H), 2.15 (d, J = 22.9
Hz, 4H), 1.44 (t, J = 7.1 Hz, 3H). MS TOFES+: m/z 368.1 (M+H)+.
Method D: Hydrogenation of olefins/alkynes to alkyl derivatives[33]
2.12 Ethyl 2-amino-6-isopropyl-5-oxo-5H-chromeno[2,3-b]pyridine-3-carboxylate (91;

Scheme 3). A flask with a solution of compound 90 (500 mg, 1.54 mmol) in dichloromethane
(50 mL) and ethanol (100 mL) was purged with nitrogen and 10% palladium on carbon (400 mg)
was added. The mixture was purged with nitrogen and hydrogen, and hydrogenated at 45 psi
using a Parr shaker at room temperature for 45 h. The mixture was filtered over Celite® and
concentrated to a solid that triturated in ethanol and dried to give 91 (392 mg, 78%) as a white
solid. 1H NMR (400 MHz, chloroform-d): δ 9.12 (s, 1H), 8.32 (s, 1H), 7.61 (t, J = 8.0 Hz, 1H),
7.36 (ddd, J = 15.2, 8.0, 1.2 Hz, 2H), 6.00 (s, 1H), 4.75 (p, J = 6.8 Hz, 1H), 4.38 (q, J = 7.1 Hz,
2H), 1.42 (t, J = 7.1 Hz, 3H), 1.32 (d, J = 6.8 Hz, 6H). MS: m/z 327.1341 (M+H)+.
2.13 1-(2-Hydroxy-6-isopropylphenyl)ethan-1-one (94; Scheme 3). Made similarly as 91
from compound 92 (370 mg, 2.1 mmol) in dichloromethane (10 mL) and ethanol (20 mL) over
10% palladium on carbon (220 mg) at atmospheric pressure of hydrogen for 18 h. Workup left
94 (357 mg, 95%) as a gum. 1H NMR (400 MHz, chloroform-d) δ 7.21 (t, J = 8.0 Hz, 1H), 6.86
(dd, J = 7.8, 1.0 Hz, 1H), 6.72 (dd, J = 8.1, 1.1 Hz, 1H), 3.71 (q, J = 7.0 Hz, 1H), 3.08 (p, J = 6.8
Hz, 1H), 2.59 (d, J = 0.8 Hz, 3H), 1.24 (d, J = 6.8 Hz, 6H). MS: m/z 179.1061 (M+H)+.
2.14 2-Amino-7-cyclohexyl-5-oxo-5H-chromeno[2,3-b]pyridine-3-carbonitrile (103;
Scheme 4). Made similarly as 91 from compound 102 (58 mg, 0.18 mmol) in dichloromethane
(90 mL) and ethanol (80 mL) over 10% palladium on carbon (80 mg) at 42 psi of hydrogen for
22 h. Workup gave 103 (44 mg, 75%) as a white solid. 1H NMR (500 MHz, chloroform-d): δ
8.75 (s, 1H), 8.08 (d, J = 2.3 Hz, 1H), 7.60 (dd, J = 8.6, 2.3 Hz, 1H), 7.44 (d, J = 8.6 Hz, 1H),
5.76 (s, 2H), 2.63 (d, J = 3.4 Hz, 1H), 1.90 (dd, J = 19.2, 12.5 Hz, 4H), 1.78 (d, J = 13.2 Hz, 1H),
1.52 – 1.36 (m, 4H), 1.29 (m, 1H). MS TOFES+: m/z 320.0 (M+H)+.
See Supplementary Material for additional experimental procedures: 25-31 (Scheme SM1); 59-
61, 63-70, 74, 75, 77-84 (Scheme SM2)
Method E: Base hydrolysis of ethyl esters to carboxylic acids[34]
1.1
2-Amino-6-isopropyl-5-oxo-5H-chromeno[2,3-b]pyridine-3-carboxylic acid (108;
Scheme 3). To a suspension of 91 (380 mg, 1.2 mmol) in ethanol (50 mL) and water (6.25 mL)
was added 1 N aq NaOH (6.25 mL). The mixture was stirred for 1.5 h at 50 oC and filtered. The
filtrate was concentrated to a solid that was suspended in water (50 mL) and acidified with 1N
aq. HCl to pH 3-4. The precipitated solids were stirred at room temperature for 24 h, collected,
washed with water, and dried to give 108 (298 mg, 86%). HPLC: rt 6.6 min (98% purity) (98%).
1H NMR (400 MHz, DMSO-d6): δ 13.40 (s, 1H), 8.78 (s, 1H), 8.21 (s, 2H), 7.72 (t, J = 8.0 Hz,
1H), 7.48 - 7.37 (m, 2H), 4.71 (p, J = 6.8 Hz, 1H), 1.24 (d, J = 6.8 Hz, 6H). MS: m/z 299.1026
(M+H)+.