Mechanism of Action of an EPAC1-Selective Competitive Partial Agonist

Article abstract of DOI:10.1021/acs.jmedchem.9b02151

The exchange protein activated by cAMP (EPAC) is a promising drug target for a wide disease range, from neurodegeneration and infections to cancer and cardiovascular conditions. A novel partial agonist of the EPAC isoform 1 (EPAC1), I942, was recently discovered, but its mechanism of action remains poorly understood. Here, we utilize NMR spectroscopy to map the I942-EPAC1 interactions at atomic resolution and propose a mechanism for I942 partial agonism. We found that I942 interacts with the phosphate binding cassette (PBC) and base binding region (BBR) of EPAC1, similar to cyclic adenosine monophosphate (cAMP). These results not only reveal the molecular basis for the I942 vs cAMP mimicry and competition, but also suggest that the partial agonism of I942 arises from its ability to stabilize an inhibition-incompetent activation intermediate distinct from both active and inactive EPAC1 states. The mechanism of action of I942 may facilitate drug design for EPAC-related diseases.

Full text of DOI:10.1021/acs.jmedchem.9b02151

pubs.acs.org/jmc
Article
Mechanism of Action of an EPAC1-Selective Competitive Partial
Agonist
Hongzhao Shao, Hebatallah Mohamed, Stephen Boulton, Jinfeng Huang, Pingyuan Wang, Haiying Chen,
́
Jia Zhou, Urszula Luchowska-Stanska, Nicholas G. Jentsch, Alison L. Armstrong, Jakob Magolan,
Stephen Yarwood, and Giuseppe Melacini*
Cite This: https://dx.doi.org/10.1021/acs.jmedchem.9b02151
Read Online
sı
*
Metrics & More
Article Recommendations
Supporting Information
ACCESS
ABSTRACT: The exchange protein activated by cAMP (EPAC) is a
promising drug target for a wide disease range, from neurodegeneration and
infections to cancer and cardiovascular conditions. A novel partial agonist of the
EPAC isoform 1 (EPAC1), I942, was recently discovered, but its mechanism of
action remains poorly understood. Here, we utilize NMR spectroscopy to map
the I942−EPAC1 interactions at atomic resolution and propose a mechanism
for I942 partial agonism. We found that I942 interacts with the phosphate binding cassette (PBC) and base binding region (BBR) of
EPAC1, similar to cyclic adenosine monophosphate (cAMP). These results not only reveal the molecular basis for the I942 vs cAMP
mimicry and competition, but also suggest that the partial agonism of I942 arises from its ability to stabilize an inhibition-
incompetent activation intermediate distinct from both active and inactive EPAC1 states. The mechanism of action of I942 may
facilitate drug design for EPAC-related diseases.
the CNBD to its active form. The cAMP binding also
allosterically weakens the ionic latch at the N-terminal region
of EPAC1-CNBD, causing the release of the catalytic domain
from the regulatory domain.30−34 Upon the release of the
catalytic domain from the autoinhibitory RR, it becomes
accessible to binding by the Rap1 and Rap2 GTPase substrates,
resulting in guanosine diphosphate (GDP) to guanosine
triphosphate (GTP) exchange1,2 (Figure 1A). Hence, the
CNBD of EPAC1 serves as its central controlling unit.
The CNBD consists of α and β subdomains. The α-
subdomain starts with an N-terminal α-helical bundle
(NTHB) that includes helices α1−α4 as well as the ionic
latch residues.30−35 The α-subdomain also comprises helices α5
and α6, which are distinct from the NTHB, with α5 embedded
within the β-barrel and α6 located C-terminally to the β-
barrel.30,34,35 The α5 helix is part of the phosphate binding
cassette (PBC). The PBC, along with the base binding region
(BBR) in the β-subdomain, is crucial for cAMP docking within
the CNBD.30,34,35 The α6 helix is responsible for the hinge
rotation of the RR relative to the CR and ultimately for the
release of the CR upon binding to cAMP.30,34,35
INTRODUCTION
■
The exchange proteins activated by cAMPs (EPACs) are a
family of guanine nucleotide exchange factors (GEFs) for the
GTPases, Rap1, and Rap2.1,2 EPACs are receptors for the cyclic
adenosine monophosphate (cAMP) secondary messenger and
contribute to the control of cAMP signaling.3 EPAC proteins are
promising therapeutic targets, as they are involved in multiple
physiological effects4,5 in the nervous,6,7 cardiovascular,8−11
endocrine,12−19 digestive,20,21 and immune22−24 systems.
Regulation of EPAC activity has been hypothesized to be a
viable avenue for developing treatment strategies for Alzheimer’s
disease,25 bacterial22 and viral23 infections, breast,18 ovar-
ian,12,13,17 and pancreatic cancer,20,21 cardiomyopathy,8,9
chronic pain,26,27 diabetes,14,16 drug dependence,28 and
obesity.14,16,19
There are two major isoforms of EPAC, 1 and 2, that share
similar structures, but differ in tissue distributions and
physiological functions.3 The EPAC1 isoform is composed of
a regulatory region (RR) and a catalytic region (CR). The RR
includes the Disheveled Egl-10 Plectstrin (DEP) domain and
cyclic nucleotide binding domain (CNBD), while the CR spans
the RAS-exchange motif (REM), the RAS association (RA), and
CDC25 homology domains (CDC25HD)4,5,29 (Figure 1A). In
its apo form, EPAC1 proteins adopt a primarily closed topology,
which autoinhibits GEF activity, as the active site in the catalytic
domain is occluded by the RR, which forms multiple salt bridges
with the CR, collectively referred to as “ionic latch” (Figure
1A).30−34
The EPAC1-CNBD undergoes a distinct allosteric conforma-
tional change upon binding to cAMP. The simplest model of the
Received: January 2, 2020
cAMP modulates EPAC1 allosterically by binding to the
CNBD and changing the conformation of the hinge region on
https://dx.doi.org/10.1021/acs.jmedchem.9b02151
© XXXX American Chemical Society
J. Med. Chem. XXXX, XXX, XXX−XXX
A

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
Figure 1. Activation mechanism of EPAC1 by cAMP and the covalent structure of the EPAC1 competitive inhibitor I942. (A) Domain organization of
EPAC1 and the scheme of EPAC1 activation. The full-length EPAC1 is organized in two major regions, the regulatory and catalytic regions, starting
from the regulatory Disheveled Egl-10 Plectstrin (DEP) domain and cyclic nucleotide binding domain (CNBD) and continuing to the catalytic RAS-
exchange motif (REM), RAS association (RA), and CDC25 homology domains (CDC25HD). The construct involved in this study is colored blue,
with the starting (149) and ending (318) residue numbers stated above. EPAC1 is initially in its autoinhibition state, where an ionic latch (IL) forms
between the regulatory and catalytic regions, effectively autoinhibiting EPAC. Upon binding to cAMP in the cyclic nucleotide binding domain, EPAC1
is released from the autoinhibitory state and Rap1 can access the CDC25 homology domain. (B) Thermodynamic cycle coupling the cAMP binding
and the EPAC1-CNBD active/inactive equilibria. The EPAC1-CNBD is illustrated in the inactive (green) and active (red) states, where the hinge
region and the phosphate binding cassette (PBC) are both in their “out” or “in” conformation, respectively (blue rectangles). The hinge and PBC
orientations in the holo-inactive and apo-active states are shown in gray to emphasize that they are only hypothetical. (C) Illustration of the EPAC2-
CNBD apo/holo structural differences. The dashed boxes indicate the base binding region (BBR), the phosphate binding cassette (PBC), and the
hinge region. The PBC and hinge regions show the out/in conformation shifts from the apo/inactive (green, PDB 2BYV30) to the holo/active (red,
PDB 3CF634) states. The holo/active state is bound to a known EPAC2 activator, Sp-cAMPS.34 (D) Structure highlighting Sp-cAMPS bound to
EPAC2-CNBD.34 Hydrogen bonds between Sp-cAMPS and EPAC2 are marked with dashed lines. The interaction between L408 (L273 in EPAC1) in
the α5 helix (yellow, PBC) and F435 (F300 in EPAC1) in the α6 helix (cyan, hinge) is highlighted with red side chains. (E) Covalent structure of the
EPAC1 partial agonist I942.
cAMP-dependent allostery in EPAC1 is a four-state thermody-
namic cycle arising from the coupling of cAMP binding and
conformational equilibria of the CNBD36−42 (Figure 1B). In the
absence of cAMP, the EPAC1-CNBD samples primarily the
apo-inactive state, but as the cAMP levels increase the holo-
active state is progressively stabilized (Figure 1B). Upon
docking in the PBC and BBR, cAMP allosterically affects the
conformation of the β2−β3 loop and α6 hinge helix.31,33,43 The
interaction between cAMP and the PBC is a key driver of the
allosteric transition. The cAMP interacts with the PBC primarily
through hydrogen bonding and a salt bridge34,35 (Figure 1D)
and changes the conformation of the α5 helix in the PBC from
an “out” to an “in” orientation, subsequently shifting the
orientation of the α6 helix (i.e., the hinge region) from “out” to
“in”30,34,35 (Figure 1B,C). The coupling between the PBC and
hinge regions relies on the L273−F300 interaction (Figure 1D)
and leads to partial unfolding of the α6 C-terminal region,
closure of the lid, rotation of the regulatory region away from the
catalytic region, and ultimately activation of EPAC1.30,31,34,35
Given the potential of EPAC as a therapeutic target, over the
years, several EPAC selective small-molecule inhibitors were
identified through high-throughput screening (HTS).20,44−53
Recently, a new EPAC1-selective sulfonyl acetamide partial
agonist, called I942 (Figure 1E), was discovered through
competitive fluorescence. I942 is a promising lead to develop
EPAC1-specific partial agonists.54 Other EPAC agonists have
been discovered, but these are cAMP analogs.55−60 Although
some of the cAMP analogs are resistant to phosphodiesterase
(PDE) hydrolysis,59,60 being cyclic nucleotides generally makes
most of the cAMP analogs vulnerable to PDE hydrolysis, as
cAMP signaling through EPAC1 is subject to PDE4D3
termination.61,62 Unlike cyclic nucleotides, I942 is less likely
to be hydrolyzed by PDEs. Furthermore, cyclic AMP analogues
typically exhibit limited membrane permeability and may
generate off-target effects. In addition, I942 is significantly
more soluble than ESI-09 and CE3F4R, which minimizes
nonspecific interactions between I942 and EPAC1.
Using EPAC1 GEF assays, I942 was proven to be able to
induce about 10−20% activity compared to cAMP-bound
EPAC1.54 I942 also inhibits the EPAC1 activity induced by
cAMP and is therefore classified as a partial agonist.54
Furthermore, I942 affects multiple EPAC1-dependent cellular
functions, including the induction of the suppressor of cytokine
signaling 3 (SOCS-3), which in turn suppresses the activation of
interleukin 6 (IL-6) signaling, as well as the downstream
induction of vascular cell adhesion molecule 1 (VCAM-1) and
monocyte adhesion to human umbilical vascular endothelial
cells (HUVEC).63 This could be utilized for the suppression of
inflammatory effects and provides an alternative therapeutic
approach for the treatment of cardiovascular diseases, including
pulmonary artery hypertension and atherosclerosis.63,64
To build upon and further improve the I942 lead compound,
it is essential to determine its mechanism of action.
Unfortunately, only sparse structural information about the
complex of EPAC with I942 or cAMP competitive inhibitors is
currently available.64 Hence, here we map, at atomic resolution,
the interactions between I942 and the EPAC1-CNBD using
NMR spectroscopy. To this end, we have utilized various
protein NMR techniques, including chemical shift mapping,
saturation transfer difference (STD), nuclear overhauser effect
spectroscopy (NOESY), and measurement of intermolecular
Nuclear Overhauser Effects (NOEs) through isotope-filtered
15N-NOESY-HSQC, as well as non-NMR techniques, such as
B
https://dx.doi.org/10.1021/acs.jmedchem.9b02151
J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
Figure 2. I942-binding isotherms for the EPAC1-CNBD and I942 conformations either free or bound to the EPAC1-CNBD and heteronuclear single
quantum coherence (HSQC) spectra comparison of the cAMP- and I942-bound EPAC1-CNB domain. (A) Binding isotherm of wild-type EPAC1-
CNBD and I942 through 8-NBD-cAMP competition monitored by fluorescence losses. (B) Similar to (A) but for the L273W mutant of the EPAC1-
CNBD, which stabilizes the hinge helix in the inactive conformation. In both panels (A) and (B), ⟨v⟩ represents the bound percentage of 8-NBD-
cAMP. 2.5 μM of EPAC1-CNBD and 0.5 μM of 8-NBD-cAMP were utilized. (C, D) Zoomed-in regions of HSQC spectra of the apo, cAMP-, and
I942-bound EPAC1-CNB domain, shown in red, blue, and green, respectively. These regions highlight representative residues from the phosphate
binding cassette (PBC). The amides of A280 and A272 donate hydrogen bonds to the cAMP phosphate in the cAMP:EPAC complex.34 (E)
Compounded chemical shift (CCS) differences between the cAMP- and I942-bound EPAC1-CNB domain. Note that any CCS differences over 1 ppm
is off-scale. The secondary structure of EPAC1 is marked on the plot along with residue numbers. The dashed line indicates the average ppm value over
all residues. (F) 3D map of chemical shifts greater than the average (red surface).
Figure 3. CHEemical Shift Projection Analysis (CHESPA) of I942 binding to the EPAC1-CNBD. (A) Vectors included in the CHESPA analysis. (B)
Compounded chemical shift (CCS) difference upon binding of cAMP (red) or I942 (black) to the EPAC1-CNBD. (C) cos θ to measure the extent of
linearity of chemical shift changes. (D) Fractional activation (X) of I942 relative to cAMP, with dashed lines indicating overall or local averages of the
PBC and hinge regions. Values greater than 1 or lower than −1 are off-scale. The secondary structure is plotted along with the residue number in panels
(B−D). (E) Map of residues with X values lower than average. (F) Similar to (E), but in a different orientation to display the β2−β3 loop. (G) Protein
energetic conformational analysis from NMR (PECAN) chemical shift prediction of the secondary structure in the I942-bound EPAC1-CNBD. The
secondary structure of apo EPAC1 is plotted along with the residue number. Red bars indicate the probability of residues forming an α-helix, while blue
bars indicate the probability of residues forming a β-sheet. Residues without any assignment are marked with black bars and secondary structure
probabilities in these regions may be affected by missing data.
C
https://dx.doi.org/10.1021/acs.jmedchem.9b02151
J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
fluorescence. Based on these data, we propose a mechanism of
action for the partial agonism exhibited by I942 toward EPAC1.
hydrogen bonding to the cAMP phosphate causes a major
down-field 1H shift for the amides of A280 and A272 in the PBC,
as expected based on the H-bonds donated by these backbone
amides to the cAMP phosphate, I942 causes only a marginal
RESULTS
■
1
1
down-field H shift (A280; Figure 2C) or an up-field H shift
(A272; Figure 2D), pointing to a significant weakening of the
intermolecular hydrogen bonds donated by the corresponding
amides. Considering that the hydrogen bond with A272 is
unique to the active conformation of the EPAC1-CNBD,35 these
differences are fully consistent with the partial agonism
previously reported for I942. Overall, the comparative chemical
shift analyses of Figure 2C−F reveal that, while I942 targets the
PBC and BBR similar to cAMP, the nature of the short-range
interactions with the EPAC1-CNBD is markedly different from
that of the endogenous effector cAMP. Furthermore, the I942 vs
cAMP chemical shift differences at the allosteric sites (Figure
2E,F) point to variations in long-range effects as well.
To further examine the conformational changes induced by
I942 binding to the EPAC1-CNBD, CHEmical Shift Projection
Analysis (CHESPA) was performed according to the vector
definition illustrated in Figure 3A. We found that consistently
negative cos θ and fractional activation values were observed at
the PBC and hinge regions (Figure 3C,D), which suggests that
I942 induces an inhibitory shift relative to cAMP at these
allosteric sites. The shift of the PBC to the inactive state upon
replacing cAMP with I942 is also supported by the overall
negative CHESPA fractional activation values measured for the
β2−β3 loop region, which is adjacent to the PBC (Figure 3F).
Although not interacting directly with cAMP, the β2−β3 loop is
sensitive to the change in PBC conformation occurring when the
EPAC1-CNBD binds to cAMP.31 Furthermore, the hinge
region, which is not a direct cAMP-binding site but is
allosterically coupled to the PBC through the L273−F300
side-chain interaction,31,34,35 is also partially inhibited. The
inhibited hinge and PBC regions indicate that, upon displace-
ment of cAMP by I942, the EPAC1-CNBD partially shifts back
to the autoinhibited state. Comparing the average of partial
inhibition of the PBC and hinge regions, it is clear that I942
inhibits the hinge region more than the PBC (∼50% for PBC vs
∼70% for hinge; Figure 3D). Other sites within the EPAC1-
CNBD are subject to similar inhibitory shifts upon replacing
cAMP with I942 (Figure 3E,F).
I942 Binds the Inactive and Active States of EPAC1
with Comparable Affinities. As a first step toward under-
standing how I942 interacts with EPAC1, we measured the
affinity of I942 for the EPAC1-CNBD, i.e., human EPAC1
(149−318), under the same experimental conditions utilized for
NMR. For this purpose, the 8-NBD-cAMP analog is a
convenient tool65 since its displacement by I942 causes a loss
of 8-NBD-cAMP fluorescence intensity. Hence, monitoring the
fluorescence decrease during a I942 titration provides an
effective means to measure the affinity of I942 for the EPAC1-
CNBD. The effective Kd between I942 and EPAC1-CNBD
under our experimental conditions is 6.6 0.5 μM (Figure 2A),
which is comparable to the previously published Kd between
cAMP and EPAC1-CNBD of 4.5 0.1 μM,66 indicating that
I942 is an effective competitive inhibitor of the cAMP-bound
EPAC1-CNBD. Furthermore, the Kd value observed for I942
and our EPAC1-CNBD construct does not exceed the IC50 and
AC50 values reported for I942 in the context of a longer EPAC1
construct (i.e., EPAC1 149−881),54 suggesting that our NMR-
amenable construct adequately recapitulates the main determi-
nants of I942 binding to longer EPAC1 constructs.
We also measured the affinity of I942 for the L273W EPAC1
mutant, which is known to prevent activation even in the
presence of cAMP.35 The L273W mutation perturbs the
communication between L273 and F300, which is critical for
controlling the hinge conformational shift upon cAMP binding.
The bulky side chain on the L273W mutant effectively prevents
the hinge region from adopting the “in” (or active)
conformation even when cAMP binds to the PBC.35,43 By
relying again on 8-NBD-cAMP competitive fluorescence
binding experiments, we found that I942 binds to EPAC1-
CNBD with a Kd value of 4.9 1.5 μM (Figure 2B). This result
indicates that the silencing of the allosteric network between the
PBC and hinge regions through the L273W mutation affects the
I942 affinity to the EPAC1-CNBD only marginally. Hence, I942
does not preserve the active vs inactive selectivity of cAMP,
which is known to bind the wild-type (WT) EPAC1-CNBD
with five-fold higher affinity relative to L273W.35 These I942 vs
cAMP differences provide an initial explanation as to why I942
cannot activate EPAC to an extent similar to cAMP.
To gain further insight into how I942 affects the structure of
EPAC1, we assessed the secondary structure of the
I942:EPAC1-CNBD complex using the Protein Energetic
Conformational Analysis from NMR chemical shifts
Binding of I942 to the EPAC1-CNBD Maintains Critical
1
Allosteric Sites in the Inactive State. The H,15N-HSQC
(PECAN).68 The chemical shifts of H and 15N as well as
1
spectrum of the I942-bound EPAC1-CNBD was assigned
through comparison with the apo and cAMP-bound spectra as
well as through triple-resonance spectra. The overlay of the apo,
cAMP- and I942-bound EPAC1-CNBD HSQCs (Figure 2C,D)
reveals that I942 causes major perturbations relative to both apo
and cAMP-bound states. Marked I942 vs apo chemical shift
changes (Figure 3B) are detected for the cAMP-binding sites,
i.e., the phosphate binding cassette (PBC) and the base binding
region (BBR), in agreement with the cAMP competitive nature
of the I942 ligand. Interestingly, significant I942 vs apo chemical
shift variations are also observed beyond the PBC and BBR, i.e.,
at allosteric sites, such as the hinge region and the β2−β3 loop
(Figure 3B).
13Cα and 13Cβ from 1N,15H-HSQC, HNCACB, and HN(CO)-
CACB experiments were utilized in the prediction. The
prediction results (Figure 3G) indicated that the majority of
the secondary structures of cAMP-bound EPAC1-CNBD are
preserved upon replacement of cAMP with I942. Outliers
include the N-terminal α1 and the C-terminal region after α6,
where most of the assignments are missing due to flexibility and
lack of sufficient resolution. In addition, the α6 helix in the hinge
region was partially unfolded based on this prediction similar to
the cAMP-bound EPAC.34 Another important aspect of the
prediction is that the α5 helicity in the PBC is retained by I942
binding.
To gain further insight into the nature of the I942-induced
perturbations, we also examined the I942 vs cAMP chemical
shifts (Figure 2E,F), which reveal major differences at both the
cAMP-binding and allosteric sites.31,67 For example, while
Conformational Change of I942 upon EPAC1 Binding.
The conformational change of the ligand in the protein−ligand
binding process is as important as the conformational shift of the
protein receptor. To probe the conformational shift of I942
D
https://dx.doi.org/10.1021/acs.jmedchem.9b02151
J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
Figure 4. Intra-ligand NOEs for the free and the EPAC1-CNB domain-bound I942. The NOESY spectra acquired in the absence and presence of the
EPAC1-CNBD construct are shown in panels (B) and (D), respectively. Panels (A) and (C) as well as the dashed lines in spectra (B) and (D) indicate
NOE cross-peaks between proton pairs of I942. The black dotted connections in panels (A) and (C) indicate that the chemical shifts of the
corresponding protons are too close to result in resolvable cross-peaks. The green lines in C indicate NOEs between the two aromatic moieties. In the
presence of EPAC1 the transfer NOE effect prevails, generating cross-peaks of the same sign as the diagonal, whereas in the absence of EPAC1 opposite
signs are observed due to the fast tumbling of free I942.
Figure 5. Mapping the EPAC1-CNBD and I942 interface. (A−E) Intermolecular NOE peaks from the 13C, 15N-filtered NOESY-HSQC spectrum. (F)
3D map of residues identified in the NOESY-HSQC spectrum. Red surfaces indicate the presence of an NOE to I942 originating from the labeled
residue.
upon its binding to EPAC1-CNBD, NOESY spectra were
acquired for I942 in the absence and presence of the EPAC1-
CNBD. In the NOESY spectrum recorded for free I942, several
NOEs were observed within each of the aromatic rings, but no
E
https://dx.doi.org/10.1021/acs.jmedchem.9b02151
J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
appreciable NOE was detected between them (Figure 4A,B),
pointing to an “open” conformation for I942 before binding to
EPAC1. However, in the NOESY spectrum acquired in the
presence of substoichiometric amounts of EPAC1-CNBD,
several transfer NOEs were observed between the phenyl and
naphthalene moieties of I942 (Figures 4C,D and S1), suggesting
a less open conformation for I942 bound to the EPAC1-CNBD.
Intermolecular NOEs, STD-HSQC Data, and I942
Analogs Reveal the Molecular Basis of the I942 vs
cAMP Mimicry. The measurement of intermolecular NOEs
between I942 and EPAC1-CNBD is pivotal in mapping the
EPAC1:I942 interface. The intermolecular NOEs were acquired
with a modified 15N-NOESY-HSQC pulse program, including
two 13C isotope filters with adiabatic pulses and a 15N isotope
filter. The 13C, 15N-filtered, and 15N-edited experiment
distinguishes protons bound to 13C/15N vs 12C/14N atoms and
filters out protein−protein intramolecular NOEs within a
15N/13C isotopically labeled protein.69,70 The experiment
resulted in several NOE peaks originating from the backbone
of selected PBC and BBR residues (Figure 5A−E). These
intermolecular NOEs indicate that the naphthalene and phenyl
rings of I942 interact with two different binding regions on
EPAC1-CNBD, i.e., the PBC and BBR, respectively (Figure 5F).
More specifically, the 2′,4′-dimethylphenyl moiety mimics the
ribose ring of cAMP and docks within the PBC, while the
naphthalene ring mimics the adenine base of cAMP and docks at
the BBR. These results are consistent with STD-HSQC data for
the I942:EPAC1-CNBD complex (Figure S2, Supporting
Information).
The conclusions drawn based on the intermolecular NOEs
were independently verified by comparing the HSQC spectra of
the I942- and I942 analog-bound EPAC1-CNBD. We modified
the two aromatic moieties and monitored the resulting chemical
shift changes in the EPAC1-CNBD.71 By comparing the
chemical shift differences between I942 and analogs with
modifications in the 2′,4′-dimethylphenyl ring, we observed that
the PBC region exhibits the largest chemical shift differences
(Figure 6A−E). The interaction between the naphthalene ring
of I942 and the BBR region was probed through a comparison of
I942 with an analog of I942 consisting of a naphthalene
substitution (Figure 6F). The chemical shift difference analysis
showed that in this case it is the BBR region that displays the
largest chemical shift differences (Figure 6F). Thus, we confirm
that the 2′,4′-dimethylphenyl and naphthalene rings are, in fact,
interacting with the PBC and BBR, respectively, in agreement
with our intermolecular NOE data.
Figure 6. Compounded chemical shift differences between I942- and
I942 analog-bound EPAC1-CNBDs. (A, B) Representative super-
imposition of HSQC spectra of EPAC1-CNBD bound to I942 or the
I942 analog from panel (C) for residues A280 (A) and A281 (B). (C−
F) Compounded chemical shift differences between I942 and I942
analog-bound EPAC1-CNBDs. All CCS differences are plotted along
with the residue numbers and the secondary structure. The dashed lines
in all plots represent the average plus two standard deviations. The
structures of analogs are provided within the plots, with the
modifications of I942 colored red or blue to highlight whether they
perturb the binding with PBC (red) or BBR (blue).
The engagement of both the 2′,4′-dimethylphenyl and the
naphthalene moieties of I942 in contacts with the EPAC1-
CNBD was independently confirmed through saturation
transfer difference (STD) NMR. STD NMR identifies which
groups within I942 are in close contact with the EPAC1-CNBD
receptor. STD measures the saturation transfer from protein to
ligand, and the STD vs saturation transfer reference (STR)
intensity ratio of each proton on I942 reflects the proximity of
such protons to the surface of EPAC172 (Figure 7A). Figure 7A
indicates that the majority of protons exhibit similar STD/STR
intensity ratios, confirming their proximity to the EPAC1-
CNBD.
While the intermolecular NOEs together with the I942
analogs and the STD data are extremely useful to position I942
within the EPAC1-CNBD, they provide only a limited number
of constraints for the sulfonyl group, which is devoid of NMR
detectable protons. To further probe to what extent the sulfonyl
moiety mimics the cyclic phosphate of cAMP, we comple-
mented the intermolecular NOE measurements with mutations
of the PBC arginine residue known to form salt bridges with the
cAMP phosphate, i.e., R279.34
Sulfonyl Group on I942 Mimics the Cyclic Phosphate
of cAMP and Interacts with the R279 Guanidinium in
EPAC1. I942 contains a sulfonyl group in place of the cyclic
phosphate group of cAMP. A viable hypothesis is that the
sulfonyl group forms hydrogen bonds and/or salt bridges with
the R279 side-chain guanidinium group34 similar to cAMP. To
test this hypothesis, we acquired guanidinium HSQC spectra
and selectively assigned the R279 guanidinium group through
the R279A mutant. If cAMP stabilizes the guanidinium of R279
by forming hydrogen bonds or salt bridges and by reducing the
rate of rotation of its NH2 moieties to fall within the slow
exchange limit, all N−H signals arising from the guanidinium’s
F
https://dx.doi.org/10.1021/acs.jmedchem.9b02151
J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
from the corresponding spectrum measured in the presence of
the reverse-agonist Rp-cAMPS (Figure 7B), consistent with the
ability of I942 to partially activate EPAC1. To further probe how
I942 affects the allosteric network of EPAC1, we relied on the
CHEmical Shift Covariance Analysis (CHESCA).
CHESCA Reveals How I942 Perturbs the Allosteric
Network of the EPAC1-CNBD. The allosteric network within
the EPAC1-CNBD is critical to release EPAC1 from auto-
inhibition. Previously, the allosteric network of the EPAC1-
CNBD was identified through an NMR chemical shift analysis
method called CHEmical Shift Covariance Analysis (CHES-
CA).67 The previously determined correlation matrix reveals a
strong correlation of the hinge region with both the PBC and the
α4 helix.67 The allosteric network between the hinge region and
the PBC is critical for EPAC1 activation. The residue pair
L273−F300 between the PBC and the hinge region controls the
position of the autoinhibitory equilibrium of EPAC1,34 whereas
the allosteric network between the hinge region and the adjacent
α4 helix is sensitive to the orientation of the hinge region, as the
α4 helix preserves a similar conformation in the active and
inactive form.30,34,35
To examine how I942 perturbs the EPAC1-CNBD allosteric
network, the cAMP CHESCA perturbation was replaced with
I942 and the CHESCA computed (Figure 8B). A control
CHESCA with cAMP-bound EPAC1-CNB including an I942-
matching amount of dimethyl sulfoxide (DMSO-d6) was also
performed to minimize the potential effects arising from the
addition of DMSO-d6 (Figure 8A). Upon replacing cAMP with
I942, we observed a significant loss of correlations relative to the
control CHESCA (Figure 8). Despite such loss, the I942 matrix
still retained multiple correlations of the hinge region with the
α4 helix as well as the PBC (Figure 8B). Selected correlations
between the hinge and the α4 helices were either weakened or
lost, i.e., most of the correlations between residues Q298, D299,
E308, and the α4 helix were retained (Figure S3), whereas most
of the correlations between F300, N301, V307, and α4 helix
were lost (Figure S3). These correlations between the hinge
region and its adjacent region, the α4 helix, are an indicator of
whether the hinge region adopts a third conformation other than
the common “in” or “out” orientation.
Another critical aspect of the allosteric network regulating
EPAC1 activation is the coupling between the hinge region and
the PBC, which controls the orientation of the hinge (α6)
helix.30,34,35 The L273−F300 pair-wise correlation exhibits an
absolute Pearson correlation coefficient of 0.97 in the control
CHESCA (Figure 8C) and 0.98 in the I942-CHESCA (Figure
8D), with I942 partitioned between the inactive apo state and
the agonist states, i.e., I942 falls in the partial activator category
(Figure 8D). A residue close to L273, i.e., N275, is another
interesting case. Its correlations with F300, I303, and V307
(Figure S4A, S4B) were lost with the I942-CHESCA, while its
correlation with E308 was retained and I942 acts as an activator
(Figure S4C,D). Overall, while some key pair-wise residue
correlations remain after I942 binding, e.g., L273−F300, several
other correlations between the two helices are lost, suggesting
that in the presence of I942, the α6 helix samples a
conformational ensemble that is distinct from the traditional
structures of apo- and cAMP-bound EPAC1 involved in the two-
state model previously proposed to explain EPAC1 activation. In
the simplest case, these observations point to a three-state partial
agonism model for I942.
Figure 7. Interactions between I942 and EPAC1-CNBD mapped by
saturation transfer difference and guanidinium HSQC. (A) Saturation
transfer difference and group epitope mapping for I942 binding to the
EPAC1-CNB domain. The 1D STR (red) and STD (blue) spectra are
overlayed. The STD spectrum is rescaled so that the intensities of
proton j in both spectra match. Proton j was selected as it exhibits the
maximum STD/STR ratio. The assignments of the spectra are
indicated through the labels in the molecular structure, whereas the
radii of the red circles indicate the relative normalized STD/STR ratios.
(B) Guanidinium HSQC spectra. The spectra overlay includes the
R279A mutant and the WT of the EPAC1-CNB domain in the apo form
or saturated with different ligands, as per the legend in the spectra. The
circled peaks are assigned to the -NH2 and Nε−Hε guanidinium
moieties of Arginine 279.
two NH2 groups and Nε−Hε bond are expected to be visible in
the HSQC spectrum of the cAMP-bound WT EPAC1-CNBD
domain. These signals should be also expected to disappear and/
or move upon removal of either cAMP (i.e., in the WT apo
sample) or the R279 guanidinium (i.e., in the R279A cAMP-
bound sample). Figure 7B shows the peaks in the Arg-HSQC
spectrum of cAMP-bound EPAC1-CNBD that meet these
criteria and were therefore assigned to the guanidinium of R279.
Interestingly, the Arg-HSQC spectrum of I942-bound WT
EPAC1-CNBD preserves the signals from only one of the two
NH2 groups of R279 observed for the cAMP-bound WT
EPAC1-CNBD. To gauge the allosteric role of the two R279
NH2 groups, we then acquired Arg-HSQC spectra for Sp-
cAMPS-bound WT EPAC1-CNBD. Sp-cAMPS is a well-known
superagonist of EPAC in which the axial exocyclic oxygen of the
cyclic phosphate group is replaced by a sulfur atom. Sulfur is
bulkier and more polarizable than oxygen and therefore perturbs
hydrogen bonds and salt bridges involving the cyclic phosphate.
Comparative analysis of the Arg-HSQC spectra of the Sp-
cAMPS- and I942-bound samples shows that Sp-cAMPS and
I942 preserve clearly different arginine NH2 signals, indicating
that, unlike Sp-cAMPS, I942 is unable to engage the R279 NH2
group necessary for activation, while still forming hydrogen
bonds and/or salt bridges with the other R279 NH2 moiety.
These observations provide a further rationale to explain the
molecular basis of the I942 partial rather than full agonism. On
the other hand, the R279 guanidinium spectrum of I942 differs
G
https://dx.doi.org/10.1021/acs.jmedchem.9b02151
J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
Figure 8. CHESCA and the allosteric networks perturbed by I942 binding. (A, B) CHEmical shift correlation matrices from CHESCA. All secondary
structures are plotted along with the residue number. Each point represents a residue pair with a Pearson correlation coefficient ≥ 0.98 and residues
identified to belong to the major cluster from agglomerative clustering are connected with blue lines. Each matrix includes the apo and the 2′-OMe-
cAMP, Rp-cAMPS, Sp-cAMPS-bound EPAC1-CNBDs, and the DMSO-d6 matched cAMP- (A) or I942- (B) bound EPAC1-CNBDs. (C, D) Pair-
wise residue vs residue chemical shift correlation between L273 and F300 with cAMP (C) or I942 (D).
DISCUSSION AND CONCLUSIONS
allosterically critical guanidinium of R279 (Figures 7B
and 9B). The guanidinium HSQC spectra indicate that
only one NH2 group of Arginine 279 is involved in the
formation of a salt bridge with the sulfonyl group in I942
and that this NH2 group is different from the one engaged
by the superagonist Sp-cAMPS (Figure 7B).
■
Based on the available data, here we propose a viable mechanism
of action for I942 as a partial agonist for EPAC1 (Figure 9). First,
our NMR analyses reveal the molecular basis of the I942 vs
cAMP mimicry. The 2,4-dimethylphenyl group, the sulfonyl
group, and the naphthalene ring of I942 mimic the ribose ring,
the cyclic phosphate, and the adenine base of cAMP,
respectively (Figure 9A). This conclusion is independently
confirmed by our intermolecular NOE and chemical shift data.
In particular, multiple intermolecular NOEs were detected
between the PBC and the 2′,4′-dimethylphenyl moiety as well as
between the BBR and the naphthalene ring, confirming that the
2′,4′-dimethylphenyl ring and the naphthalene moieties bind to
the cAMP ribose and base docking regions, respectively, in full
agreement with the I942 analog chemical shift data (Figure 6).
In addition, considering that, when bound to EPAC1, cAMP
adopts a syn conformation with the adenine base pointing
toward the ribose,73 the phenyl-ribose/naphthalene-adenine
I942 vs cAMP mimicry proposed here suggests proximity
between the phenyl and naphthalene rings for EPAC1-bound
I942. This prediction is fully consistent with the transition to a
less open conformation observed for I942 upon EPAC1 binding.
The replacement of cAMP in the EPAC1-CNBD by I942
results in three main effects on EPAC1:
(b) The I942 vs cAMP exchange also perturbs the allosteric
network between the PBC and the hinge helix, as
indicated by comparative CHESCA analyses (Figures 8
and S3 and S4).
(c) The I942 vs cAMP replacement induces a partial
inhibitory shift of the PBC and hinge regions from the
“in” to the “out” orientation. While the partial in-to-out
shift affects both the PBC and hinge regions, the PBC on
average favors the “in” orientation more than the hinge
region (Figure 3D). The differential in-to-out transition of
the PBC vs hinge sites is consistent with the compromised
allosteric couplings observed by CHESCA for the I942-
bound EPAC1-CNB domain and is explained through the
partial stabilization by I942 of a third “mixed”
intermediate with a PBC-in and hinge-out topology
(Figure 9C). The population of active and inactive
populations was estimated through the fractional
activation values from CHESPA (Figure 3D). The active
population was estimated through the fractional activa-
tion values from the hinge region. The hinge region was
∼70% inhibited, so the active population could be
estimated at ∼30%. Similarly, the inactive population
was estimated through the PBC inhibition at ∼50%.
Hence, the inactive population was estimated to be ∼50%.
There was ∼20% population that was unaccounted for,
which was assigned to the mixed intermediate state.
(a) It compromises selected intermolecular hydrogen bonds
with the PBC, which are necessary to fully engage the
PBC in an active conformation (Figure 9B). For example,
comparative chemical shift analyses clearly indicate that
the hydrogen bonds with the amides of A272 and A280 in
the PBC are significantly weaker in the I942:EPAC1 vs
cAMP:EPAC1 complex (Figures 2C,D and 9B) and that
I942 significantly perturbs the salt bridge with the
H
https://dx.doi.org/10.1021/acs.jmedchem.9b02151
J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
EXPERIMENTAL SECTION
■
Protein Expression and Purification. The human EPAC1 (149−
318) construct, referred here as EPAC1-CNBD, as wild-type, R279A,
or L273W mutants, was expressed and purified based on previously
described protocols.31,52,67,73 In summary, a GST (Glutathione S-
transferase)-tagged fusion protein clone was transformed into the E. coli
BL21 (DE3) strain on ampicillin-infused LB (Lysogeny Broth) agar. A
well-isolated colony was picked and grown in LB broth for 6 h and then
inoculated in 15N- or 15N/13C-labeled M9 minimal media depending on
the requirement of nuclei for the NMR experiments. The inoculated
culture was incubated at 37 °C until the optical density at 600 nm
reached the 0.6−0.8 range. The grown culture was then induced with
0.5 mM of isopropyl β-D-1-thiogalactopyranoside (IPTG) and
incubated at 18−20 °C for 16−18 h. Cells were harvested and lysed
with either a cell disruptor or a sonicator in a lysis buffer composed of
phosphate-buffered saline, 10% glycerol, 10 mM of ethylenediaminete-
traacetate (EDTA), 1 mM of dithiothreitol (DTT), 12 μg of Nα-p-tosyl-
L-lysine-chloromethyl ketone (TLCK) and 12 μg of N-p-tosyl-
phenylalanyl-chloromethyl ketone (TPCK), and 0.24 mM of 4-(2-
aminoethyl)benzenesulfonyl fluoride (AEBSF), pH 7.4. The cell lysate
was purified with a Glutathione Sepharose 4B resin (GE Healthcare).
First, the resin was washed with 50 mM Tris, 500 mM NaCl, and 1 mM
DTT, pH 8.0, and later washed and suspended in 50 mM Tris, 50 mM
NaCl, and 1 mM DTT, pH 7.6. The GST-tag was removed by treating
with biotinylated thrombin (EMD-Millipore) for 12−16 h, and the
biotinylated thrombin was later removed using a Streptavidin
Sepharose High Performance resin (GE Healthcare). The crude
product was further purified with a HiTrap Q HP anion exchange
chromatography column (GE Healthcare) and dialyzed in 50 mM Tris,
50 mM NaCl, and 1 mM DTT, pH 7.6, overnight.
NMR Sample Preparation and NMR Spectroscopy. The final
dialysis buffer mentioned in the previous paragraph was also utilized as
buffer for the NMR experiments and the 8-NBD-cAMP competitive
binding titration, except when a buffer devoid of detectable protons was
required, as specified below. The latter is composed of 20 mM sodium
phosphate buffer and 50 mM NaCl in D2O, pH 7.6. The I942 ligand
(Life Chemicals; purity >99%), i.e., N-((2,4-dimethylphenyl)sulfonyl)-
2-(naphthalen-2-yloxy)acetamide, was dissolved in deuterated DMSO-
d6 as a stock with 50 mM concentration and diluted with the
appropriate experiment-specific buffer, as described below, when
needed. All NMR experiments were carried out on a Bruker Avance
700 MHz spectrometer equipped with a 5 mm TCI cryo-probe, unless
otherwise specified. Experiments were performed at a temperature of
306 K except for ligand-based experiments, such as saturation transfer
difference (STD) and 2D transfer NOESY, which were conducted at
298 K for more efficient cross-relaxation. All samples for NMR
spectroscopy prepared with DMSO-d6 had a DMSO-d6 content ≤ 2%
v/v to minimize potential chemical shift perturbations from the
addition of DMSO-d6,45 and all samples in nondeuterated buffer were
supplied with 5% v/v of D2O for locking purposes. A matching amount
of DMSO-d6 was added to apo and cAMP-bound EPAC as a control for
DMSO-d6-induced chemical shift changes. The 1D experiments were
processed and analyzed with TopSpin (Bruker), and the 2D/3D
experiments were processed with Topspin and/or NMRPipe74 and
analyzed with Sparky.75 The assignment of the I942 1D spectrum was
determined based on COSY and TOCSY experiments (Figure S5). The
I942-bound EPAC1-CNBD NMR spectra were assigned through
comparisons with apo and cAMP-bound spectra as well as 3D triple-
resonance experiments. Details of NMR experimental acquisitions are
described below.
Figure 9. Proposed mechanism of action of I942 as a partial agonist for
EPAC1. (A) Molecular basis for the I942 vs cAMP mimicry. Circles and
dashed lines indicate how I942 mimics cAMP when bound to EPAC1.
(B) Illustration of the interactions preserved or perturbed by I942
compared to cAMP. Interactions disrupted by I942 binding are
depicted by black dashed lines and interactions largely preserved by
I942 binding are shown as red dashed lines. In the case of the R279
guanidinium group, both types of lines are present because of only
partial perturbation by I942. (C) Proposed mechanism of action of
I942 as a partial agonist for EPAC1. The scheme includes the activation
of EPAC1 by cAMP binding for comparison purposes and a three-state
equilibrium for the I942:EPAC1-CNBD complex. The mixed
intermediate includes elements from both active and inactive states
of the EPAC1-CNBD.
The presence of the mixed PBC-in/hinge-out intermediate in
the conformational ensemble sampled by the I942:EPAC1-
CNBD complex is also supported by the comparable WT vs
L273W affinities of I942 (Figure 2A,B). L273W is known to
stabilize such mixed intermediates through the interposition of
the bulky indole side chain between the PBC and the hinge
helices. Hence, the partial agonism of I942 does not arise from a
simple shift of the original two-state active vs inactive
conformational equilibrium but is in agreement with the
presence of a mixed PBC-in/hinge-out intermediate state
(∼20%).30,33−35 Considering that the hinge-in orientation is
necessary for full activation, such an intermediate is expected to
be inhibitory, thus rationalizing the partial agonism observed for
I942 in GEF assays.54 Overall, the model of Figure 9C suggests
that the disengagement of the CNBD hinge helix is the primary
factor limiting I942 agonism. The proposed mechanism of I942
partial agonism provides a basis for the design of the next
generation of I942 analogs with improved potency and efficacy.
Saturation Transfer Difference (STD) and Group Epitope
Mapping. The binding between the EPAC1-CNBD and I942 was
probed by STD experiments. A 50 μM EPAC1-CNB domain sample
was buffer-exchanged into 20 mM sodium phosphate buffer supplied
with 50 mM of NaCl, pH 7.6, in 99.9% D2O with Zeba Spin Desalting
Columns of 7K MWCO (Thermo Scientific). I942 was added to the
EPAC1-CNBD at 300 μM. The samples were incubated at room
temperature for at least 30 min to ensure that protein−ligand binding
reached its equilibrium prior to NMR acquisition. Saturation transfer
I
https://dx.doi.org/10.1021/acs.jmedchem.9b02151
J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
reference (STR) experiments were acquired with an off-resonance
saturation at 30 ppm, whereas STD experiments were acquired with an
on-resonance saturation at 0.85 ppm, where the methyl region of
EPAC1 appears and no I942 1D signal is detected. Hence, the STD
experiment monitors saturation transfer from EPAC1 to I942. The
spectral width was set at 11.9807 ppm with a transmitter frequency of
4.717 ppm for both STD and STR spectra. The STD experiment was
performed with 512 scans and the STR experiment with 256 scans,
along with 8 dummy scans for both. The STD/STR signal intensity
ratios for different I942 peaks were computed and normalized to the
largest STD/STR ratio to probe the binding group epitope mapping.72
2D-NOESY. The conformational changes between free and EPAC1-
CNB-bound I942 were probed by NOESY and transfer NOESY
experiments. EPAC1-CNBD (50 μM) was buffer-exchanged into the
same buffer as the STD experiments with Zeba Spin Desalting Columns
of 7K MWCO (Thermo Scientific) and 300 μM of I942 was added to
the sample. Additional NOESY spectra of 100 μM of free I942 and 50
μM of free EPAC1-CNBD were also acquired as the free ligand and free
protein controls, respectively. All NOESY spectra were acquired with
2048 and 512 or 256 complex points in the direct and indirect
dimensions, respectively, 64 dummy scans, 64 scans, and spectral
widths of 11.9807 ppm with a transmitter frequency of 4.696 ppm. The
NOESY mixing time was 250 ms in the presence of the EPAC1-CNBD
and 750 ms for free I942.
less sensitive compared to STR-HSQC. In addition, the STD-HSQC
experiment was acquired with an irradiation of I942 naphthalene signals
to monitor saturation transfer from I942 to EPAC1.
8-NBD-cAMP Competitive Binding. I942 competes with 8-(2-[7-
nitro-4-benzofurazanyl] aminoethylthio) adenosine-3′,5′-cyclic mono-
phosphate (8-NBD-cAMP) for binding to the EPAC1-CNB domain,
and the Kd of the EPAC1-CNBD:I942 complex was determined
through the loss of fluorescence intensity of 8-NBD-cAMP bound to
EPAC1-CNBD.65 A series of I942 samples with concentrations ranging
from 0 to 300 μM was added to mixtures of 2.5 μM of EPAC1-CNB and
0.5 μM of 8-NBD-cAMP in the NMR buffer. The samples were
incubated at room temperature for at least 30 min to promote protein−
ligand binding equilibrium and loaded onto Corning 96-well half area
plates afterward. The plate was scanned with a Cytation 5 plate reader
(BioTek) using an excitation wavelength of 485 nm and an emission
wavelength of 535 nm. The competitive binding was fitted as previously
described.77 The bound 8-NBD-cAMP percentage (⟨v⟩) is plotted
against the total concentration of I942. The percentage of bound 8-
NBD-cAMP can be described as
θ
−a + 2 a2 − 3b cos
(3 )
⟨v⟩8−NBD−cAMP
where
=
θ
3Kd,8−NBD−cAMP − a + 2 a2 − 3b cos
(3 )
Measurement of Intermolecular NOEs. Intermolecular NOEs
were measured through isotopically filtered NOESY-HSQC experi-
ments. The intermolecular NOEs between the 15N/13C-labeled
EPAC1-CNBD and I942 were measured with samples in a non-
deuterated buffer because the protein HN backbone would otherwise
exchange with D2O and eliminate any NOEs with exchangeable NH
amides. 15N- and 13C-isotopic filters with 13C adiabatic pulses were
implemented in the 15N-NOESY-HSQC pulse sequence to measure
intermolecular NOEs.69,70 The experiment was acquired at 306 K with
128 dummy scans, 8 scans, and complex points of 1024, 256, and 128 in
the 1H direct, 1H indirect, and 15N dimensions, respectively, and
spectral widths of 13.9, 14.0, and 32.0 ppm, respectively. The
a = Kd,8−NBD−cAMP + Kd,I942 + [8 − NBD−cAMP]T + [I942]T
− [EPAC]T
b = Kd,8−NBD−cAMPKd,I942 + Kd,I942([8 − NBD − cAMP]T
− [EPAC]T ) + Kd,8−NBD−cAMP([I942]T − [EPAC]T )
c = −Kd,8−NBD−cAMPKd,I942[EPAC]T
−2a3 + 9ab − 27c
θ = arc cos
2 (a2 − 3b)3
1
transmitter frequency was set at 4.7 and 119.0 for the H and 15N
channels, respectively. The 15N/13C-filtered NOESY-HSQC was
acquired with a mixing time of 150 ms. We also recorded a 15N-edited
NOESY-HSQC spectrum on the 15N,13C-labeled EPAC1 (149−318)
with a mixing time of 250 ms and without J13C‑1H refocusing or
decoupling pulses to confirm the intermolecular NOEs.
The errors on the Kd and ⟨v⟩ values were computed as the standard
deviations of replicate measurements. The reported Kd values are the
average of replica experiments.
Chemical Shift Analysis. The chemical shift projection analysis
(CHESPA) was performed similarly to what previously described43,78
with both apo and cAMP-bound EPAC1-CNBD samples supplied with
2% DMSO-d6 to match the DMSO-d6 content in I942-bound EPAC1-
CNBD sample. The compounded chemical shift difference (ΔCCS),
cos θ, and fractional activation (X) (Figure 3A) were computed as
Guanidinium HSQC. The side-chain guanidinium group 15N
signals resonate generally at around ∼85 ppm,76 which is usually not
within the spectral width of routine HSQC experiments and far from
the 15N carrier frequency, resulting in offset effects. Here, we changed
the 15N carrier frequency to 80 ppm and expanded the HSQC 15N
spectral width to monitor the guanidinium groups in arginine side
chains. To evaluate the involvement of the EPAC1 Arginine 279 side-
chain guanidinium NH and NH2 moieties in I942 binding, guanidinium
HSQC spectra were acquired for wild-type apo, cAMP-, and I942-
bound states as well as the R279A mutant in the cAMP-bound state to
assign the N−H peaks corresponding to the R279 guanidinium. To
improve the signal-to-noise ratio, the number of scans was increased to
64. A guanidine HSQC of the Sp-cAMPS-bound wild-type EPAC1-
CNBD was also acquired to confirm which NH2 group of R279 is
involved in I942 binding, since the crystal structure of Sp-cAMPS-
bound EPAC2-CNB is known.34
ΔCCS = (ΔδH)2 + (0.2×ΔδN)2
(1)
A·B
cos θ =
|A||B|
|A|
|B|
X =
cos θ
The cos θ and fractional activation (X) were computed with a minimum
cutoff of 0.05 ppm for the ΔCCS of vectors A (perturbation) or B
(reference). The CHEmical Shift Covariance Analysis (CHESCA) was
also performed with five different states: apo, cAMP or I942, 2′-OMe-
cAMP, Rp-cAMPS, and Sp-cAMPS bound as previously described67,78
to probe the allosteric perturbations caused by I942. First, the
compounded chemical shifts (CCS) were calculated for each residue in
every spectrum.
STD-HSQC. STD-13C-HSQC spectra were recorded to monitor the
saturation transfer from I942 to the EPAC1-CNB domain. The STD/
STR-13C-HSQC spectra were acquired in D2O to minimize residual
water artifacts in the spectrum. Two samples of 250 μM in the EPAC1-
CNB domain with and without 1 mM of I942 were prepared. The apo
sample was prepared for saturation transfer leak through control
purposes. Both the STR-HSQC and STD-HSQC spectra were acquired
at 306 K with 64 dummy scans, 1024 (1H) and 256 (13C) complex
points, spectral widths of 13.0301 ppm (1H) and 90.0 ppm (13C), and
transmitter frequencies at 1.0 ppm (1H) and 39.0 ppm (13C). The STR-
HSQC spectrum was acquired with 8 scans, while the STD-HSQC
spectrum was acquired with 64 scans as the STD-HSQC experiment is
δNH = 0.2 × δN + δH
Pair-wise residue linear correlations were then identified by computing
the correlation matrix of the transpose of the residue × state CCS
matrix. An absolute value Pearson correlation coefficient cutoff of 0.98
was used. The cAMP, 2′-OMe-cAMP, and Sp-cAMPS samples served
as references for the active states of the EPAC1-CNBD, while the apo
J
https://dx.doi.org/10.1021/acs.jmedchem.9b02151
J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
and Rp-cAMPS samples as references for the inactive states in the
CHESCA analysis. Then, the cAMP-bound state was replaced with the
I942-bound state. Except for I942-bound EPAC1-CNBD, all spectra for
the CHESCA analysis were acquired previously, unless otherwise
specified.67 After the assembly of the two matrices with cAMP or I942
along with the other CHESCA perturbation states, the allosteric
network differences occurring upon replacing cAMP with I942 were
identified through the comparative analysis of the two chemical shift
correlation matrices. A control CHESCA with cAMP replaced by
cAMP in an I942-matching amount of DMSO-d6 was also performed to
minimize any effect caused by DMSO-d6. The errors for the I942-
bound EPAC1-CNBD-compounded chemical shifts were estimated
through a technical triplicate, and the errors of cAMP-bound EPAC1-
CNBD with matching DMSO-d6 were calculated through previously
published cAMP-bound spectra.67 The chemical shift correlation
matrices served as the input for agglomerative clustering using Cluster
3.079 to find EPAC1-CNBD residue clusters that represent functional
allosteric networks. The clustering was performed through the
complete linkage method, and all clusters were extracted with a
correlation coefficient cutoff value of 0.95.
Validation of the I942:EPAC1-CNBD Interactions through
the Chemical Shift Differences Induced by I942 Analogs. HSQC
spectra of several I942 analog-bound EPAC1-CNBD samples were
acquired to validate the I942−EPAC1-CNBD interactions. The analogs
contained modifications in the 2′,4′-dimethylphenyl80 and naphthalene
moieties. Purity was confirmed by LC−MS to be ≥ 95% for all
compounds used. The compounded chemical shift differences between
the I942 and its analog-bound states were computed using eq 1 to probe
the location of the protein residues perturbed by the ligand
modifications.
Estimation of State Populations in the Three-State Equili-
brium. The populations of the inactive, active, and intermediate states
were estimated through the fractional activation values obtained from
CHESPA, as previously described.78,81,82 For instance, the active state is
the only state in which the hinge adopts the “in” orientation as opposed
to the “out” orientation in the inactive and intermediate states. On the
other hand, the inactive state is the only state in which the PBC adopts
the “out” orientation as opposed to the “in” orientation in the active and
intermediate states. Hence, both active and inactive populations were
calculated through the local average fraction activation of the hinge and
PBC regions, respectively. The intermediate state accounts for the
remaining population percentage other than the active and inactive
states.
Stephen Boulton − Department of Biochemistry and Biomedical
Sciences, McMaster University, Hamilton, Ontario L8S 4L8,
Canada; orcid.org/0000-0002-2521-2240
Jinfeng Huang − Department of Chemistry and Chemical Biology,
McMaster University, Hamilton, Ontario L8S 4L8, Canada
Pingyuan Wang − Chemical Biology Program, Department of
Pharmacology and Toxicology, University of Texas Medical
Branch, Galveston, Texas 77555, United States
Haiying Chen − Chemical Biology Program, Department of
Pharmacology and Toxicology, University of Texas Medical
Branch, Galveston, Texas 77555, United States
Jia Zhou − Chemical Biology Program, Department of
Pharmacology and Toxicology, University of Texas Medical
Branch, Galveston, Texas 77555, United States; orcid.org/
0000-0002-2811-1090
́
Urszula Luchowska-Stanska − Institute of Biological Chemistry,
Biophysics and Bioengineering, School of Engineering and
Physical Sciences, Heriot-Watt University, Edinburgh EH14 4AS,
United Kingdom
Nicholas G. Jentsch − Department of Biochemistry and
Biomedical Sciences, McMaster University, Hamilton, Ontario
L8S 4L8, Canada
Alison L. Armstrong − Department of Chemistry and Chemical
Biology and Department of Biochemistry and Biomedical
Sciences, McMaster University, Hamilton, Ontario L8S 4L8,
Canada
Jakob Magolan − Department of Biochemistry and Biomedical
Sciences, McMaster University, Hamilton, Ontario L8S 4L8,
Canada; orcid.org/0000-0002-2947-8580
Stephen Yarwood − Institute of Biological Chemistry, Biophysics
and Bioengineering, School of Engineering and Physical Sciences,
Heriot-Watt University, Edinburgh EH14 4AS, United Kingdom
Complete contact information is available at:
https://pubs.acs.org/10.1021/acs.jmedchem.9b02151
Author Contributions
The manuscript was written by H.S. and G.M. through the
contributions of all authors. All authors have given approval to
the final version of the manuscript.
ASSOCIATED CONTENT
* Supporting Information
The Supporting Information is available free of charge at
https://pubs.acs.org/doi/10.1021/acs.jmedchem.9b02151.
■
Notes
sı
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
Supporting Figures S1−S5; additional results; synthesis
experimental section; compound purity section (PDF)
Molecular formula strings (CSV)
We would like to express our appreciation to Dr. X. Cheng
(UTHealth), Dr. M. Akimoto, R. Ahmed, and J. Huang for their
helpful discussions. We thank Dr. K. Green, Dr. H. Jenkins, and
Dr. B. Berno for technical assistance. This work was supported
by Canadian Institutes of Health Research Grant 389522 (to
G.M.) and Natural Sciences and Engineering Research Council
of Canada Grant RGPIN-2019-05990 (to G.M.).
AUTHOR INFORMATION
Corresponding Author
■
Giuseppe Melacini − Department of Chemistry and Chemical
Biology and Department of Biochemistry and Biomedical
Sciences, McMaster University, Hamilton, Ontario L8S 4L8,
Canada; orcid.org/0000-0003-1164-2853;
Email: melacin@mcmaster.ca
ABBREVIATIONS USED
■
BBR, base binding region; cAMP, cyclic adenosine mono-
phosphate; CCS, compounded chemical shift; CHESCA,
chemical shift covariance analysis; CHESPA, chemical shift
projection analysis; CNBD, cyclic nucleotide-binding domain;
EPAC, exchange protein directly activated by cAMP; GEF,
guanine nucleotide exchange factor; GDP, guanosine diphos-
phate; GTP, guanosine triphosphate; HSQC, heteronuclear
single quantum coherence; NMR, nuclear magnetic resonance;
NOE, nuclear overhauser effect; PBC, phosphate binding
Authors
Hongzhao Shao − Department of Chemistry and Chemical
Biology, McMaster University, Hamilton, Ontario L8S 4L8,
Canada
Hebatallah Mohamed − Department of Chemistry and Chemical
Biology, McMaster University, Hamilton, Ontario L8S 4L8,
Canada
K
https://dx.doi.org/10.1021/acs.jmedchem.9b02151
J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
regulation of leptin expression and secretion in white adipose tissue.
Mol. Cell. Biol. 2016, 36, 2440−2450.
cassette; STD, saturation transfer difference; STR, saturation
transfer reference; WT, wild-type
(17) Gao, M.; Ma, Y.; Bast, R. C.; Li, Y.; Wan, L.; Liu, Y.; Sun, Y.; Fang,
Z.; Zhang, L.; Wang, X.; Wei, Z. Epac1 knockdown inhibits the
proliferation of ovarian cancer cells by inactivating AKT/Cyclin D1/
CDK4 pathway in vitro and in vivo. Med. Oncol. 2016, 33, No. 73.
(18) Kumar, N.; Gupta, S.; Dabral, S.; Singh, S.; Sehrawat, S. Role of
exchange protein directly activated by cAMP (EPAC1) in breast cancer
cell migration and apoptosis. Mol. Cell. Biochem. 2017, 430, 115−125.
(19) Hwang, M.; Go, Y.; Park, J.-H.; Shin, S.-K.; Song, S. E.; Oh, B.-C.;
Im, S.-S.; Hwang, I.; Jeon, Y. H.; Lee, I.-K.; Seino, S.; Song, D.-K.
Epac2a-null mice exhibit obesity-prone nature more susceptible to
leptin resistance. Int. J. Obes. 2017, 41, 279−288.
(20) Almahariq, M.; Tsalkova, T.; Mei, F. C.; Chen, H.; Zhou, J.;
Sastry, S. K.; Schwede, F.; Cheng, X. A novel EPAC-specific inhibitor
suppresses pancreatic cancer cell migration and invasion. Mol.
Pharmacol. 2013, 83, 122−128.
(21) Wang, X.; Luo, C.; Cheng, X.; Lu, M. Lithium and an EPAC-
specific inhibitor ESI-09 synergistically suppress pancreatic cancer cell
proliferation and survival. Acta Biochim. Biophys. Sin. 2017, 49, 573−
580.
(22) Gong, B.; Shelite, T.; Mei, F. C.; Ha, T.; Hu, Y.; Xu, G.; Chang,
Q.; Wakamiya, M.; Ksiazek, T. G.; Boor, P. J.; Bouyer, D. H.; Popov, V.
L.; Chen, J.; Walker, D. H.; Cheng, X. Exchange protein directly
activated by cAMP plays a critical role in bacterial invasion during fatal
rickettsioses. Proc. Natl. Acad. Sci. U.S.A. 2013, 110, 19615−19620.
(23) Tao, X.; Mei, F.; Agrawal, A.; Peters, C. J.; Ksiazek, T. G.; Cheng,
X.; Tseng, C.-T. K. Blocking of exchange proteins directly activated by
cAMP leads to reduced replication of Middle East respiratory syndrome
coronavirus. J. Virol. 2014, 88, 3902−3910.
(24) Dawn, A.; Singh, S.; More, K. R.; Siddiqui, F. A.; Pachikara, N.;
Ramdani, G.; Langsley, G.; Chitnis, C. E. The central role of cAMP in
regulating Plasmodium falciparum merozoite invasion of human
erythrocytes. PLoS Pathog. 2014, 10, No. e1004520.
(25) McPhee, I.; Gibson, L. C. D.; Kewney, J.; Darroch, C.; Stevens, P.
A.; Spinks, D.; Cooreman, A.; MacKenzie, S. J. Cyclic nucleotide
signalling: a molecular approach to drug discovery for Alzheimer’s
disease. Biochem. Soc. Trans. 2005, 33, 1330−1332.
(26) Singhmar, P.; Huo, X.; Eijkelkamp, N.; Berciano, S. R.; Baameur,
F.; Mei, F. C.; Zhu, Y.; Cheng, X.; Hawke, D.; Mayor, F.; Murga, C.;
Heijnen, C. J.; Kavelaars, A. Critical role for Epac1 in inflammatory pain
controlled by GRK2-mediated phosphorylation of Epac1. Proc. Natl.
Acad. Sci. U.S.A. 2016, 113, 3036−3041.
(27) Wang, H.; Heijnen, C. J.; van Velthoven, C. T. J.; Willemen, H. L.
D. M.; Ishikawa, Y.; Zhang, X.; Sood, A. K.; Vroon, A.; Eijkelkamp, N.;
Kavelaars, A. Balancing GRK2 and EPAC1 levels prevents and relieves
chronic pain. J. Clin. Invest. 2013, 123, 5023−5034.
(28) Wan, X.; Torregrossa, M. M.; Sanchez, H.; Nairn, A. C.; Taylor, J.
R. Activation of exchange protein activated by cAMP in the rat
basolateral amygdala impairs reconsolidation of a memory associated
with self-administered cocaine. PLoS ONE 2014, 9, No. e107359.
(29) Bos, J. L. Epac proteins: multi-purpose cAMP targets. Trends
Biochem. Sci. 2006, 31, 680−686.
(30) Rehmann, H.; Das, J.; Knipscheer, P.; Wittinghofer, A.; Bos, J. L.
Structure of the cyclic-AMP-responsive exchange factor Epac2 in its
auto-inhibited state. Nature 2006, 439, 625−628.
(31) Mazhab-Jafari, M. T.; Das, R.; Fotheringham, S. A.; SilDas, S.;
Chowdhury, S.; Melacini, G. Understanding cAMP-dependent
allostery by NMR spectroscopy: comparative analysis of the EPAC1
cAMP-binding domain in its apo and cAMP-bound states. J. Am. Chem.
Soc. 2007, 129, 14482−14492.
(32) VanSchouwen, B.; Selvaratnam, R.; Fogolari, F.; Melacini, G.
Role of dynamics in the autoinhibition and activation of the exchange
protein directly activated by cyclic AMP (EPAC). J. Biol. Chem. 2011,
286, 42655−42669.
(33) Das, R.; Mazhab-Jafari, M. T.; Chowdhury, S.; SilDas, S.;
Selvaratnam, R.; Melacini, G. Entropy-driven cAMP-dependent
allosteric control of inhibitory interactions in exchange proteins directly
activated by cAMP. J. Biol. Chem. 2008, 283, 19691−19703.
REFERENCES
■
(1) Kawasaki, H.; Springett, G. M.; Mochizuki, N.; Toki, S.; Nakaya,
M.; Matsuda, M.; Housman, D. E.; Graybiel, A. M. A family of cAMP-
binding proteins that directly activate Rap1. Science 1998, 282, 2275−
2279.
(2) de Rooij, J.; Zwartkruis, F. J.; Verheijen, M. H.; Cool, R. H.;
Nijman, S. M.; Wittinghofer, A.; Bos, J. L. Epac is a Rap1 guanine-
nucleotide-exchange factor directly activated by cyclic AMP. Nature
1998, 396, 474−477.
(3) de Rooij, J.; Rehmann, H.; van Triest, M.; Cool, R. H.;
Wittinghofer, A.; Bos, J. L. Mechanism of regulation of the Epac family
of cAMP-dependent RapGEFs. J. Biol. Chem. 2000, 275, 20829−20836.
(4) Schmidt, M.; Dekker, F. J.; Maarsingh, H. Exchange protein
directly activated by cAMP (epac): a multidomain cAMP mediator in
the regulation of diverse biological functions. Pharmacol. Rev. 2013, 65,
670−709.
(5) Robichaux, W. G.; Cheng, X. Intracellular cAMP sensor EPAC:
physiology, pathophysiology, and therapeutics development. Physiol.
Rev. 2018, 98, 919−1053.
(6) Kaneko, M.; Takahashi, T. Presynaptic mechanism underlying
cAMP-dependent synaptic potentiation. J. Neurosci. 2004, 24, 5202−
5208.
(7) Yang, Y.; Shu, X.; Liu, D.; Shang, Y.; Wu, Y.; Pei, L.; Xu, X.; Tian,
Q.; Zhang, J.; Qian, K.; Wang, Y.-X.; Petralia, R. S.; Tu, W.; Zhu, L.-Q.;
Wang, J.-Z.; Lu, Y. EPAC null mutation impairs learning and social
interactions via aberrant regulation of miR-124 and Zif268 translation.
Neuron 2012, 73, 774−788.
(8) Morel, E.; Marcantoni, A.; Gastineau, M.; Birkedal, R.; Rochais, F.;
́
Garnier, A.; Lompre, A.-M.; Vandecasteele, G.; Lezoualc’h, F. cAMP-
binding protein Epac induces cardiomyocyte hypertrophy. Circ. Res.
2005, 97, 1296−1304.
́
(9) Metrich, M.; Lucas, A.; Gastineau, M.; Samuel, J.-L.; Heymes, C.;
Morel, E.; Lezoualc’h, F. Epac mediates beta-adrenergic receptor-
induced cardiomyocyte hypertrophy. Circ. Res. 2008, 102, 959−965.
(10) Laurent, A.-C.; Bisserier, M.; Lucas, A.; Tortosa, F.; Roumieux,
́
M.; Regibus, A. de.; Swiader, A.; Sainte-Marie, Y.; Heymes, C.; Vindis,
C.; Lezoualc’h, F. Exchange protein directly activated by cAMP 1
promotes autophagy during cardiomyocyte hypertrophy. Cardiovasc.
Res. 2015, 105, 55−64.
(11) Fazal, L.; Laudette, M.; Paula-Gomes, S.; Pons, S.; Conte, C.;
Tortosa, F.; Sicard, P.; Sainte-Marie, Y.; Bisserier, M.; Lairez, O.; Lucas,
A.; Roy, J.; Ghaleh, B.; Fauconnier, J.; Mialet-Perez, J.; Lezoualc’h, F.
Multifunctional mitochondrial Epac1 controls myocardial cell death.
Circ. Res. 2017, 120, 645−657.
(12) Rangarajan, S.; Enserink, J. M.; Kuiperij, H. B.; de Rooij, J.; Price,
L. S.; Schwede, F.; Bos, J. L. Cyclic AMP induces integrin-mediated cell
adhesion through Epac and Rap1 upon stimulation of the beta 2-
adrenergic receptor. J. Cell Biol. 2003, 160, 487−493.
(13) Bastian, P.; Balcarek, A.; Altanis, C.; Strell, C.; Niggemann, B.;
Zaenker, K. S.; Entschladen, F. The inhibitory effect of norepinephrine
on the migration of ES-2 ovarian carcinoma cells involves a Rap1-
dependent pathway. Cancer Lett. 2009, 274, 218−224.
(14) Yan, J.; Mei, F. C.; Cheng, H.; Lao, D. H.; Hu, Y.; Wei, J.;
Patrikeev, I.; Hao, D.; Stutz, S. J.; Dineley, K. T.; Motamedi, M.;
Hommel, J. D.; Cunningham, K. A.; Chen, J.; Cheng, X. Enhanced
leptin sensitivity, reduced adiposity, and improved glucose homeostasis
in mice lacking exchange protein directly activated by cyclic AMP
isoform 1. Mol. Cell. Biol. 2013, 33, 918−926.
(15) Almahariq, M.; Chao, C.; Mei, F. C.; Hellmich, M. R.; Patrikeev,
I.; Motamedi, M.; Cheng, X. Pharmacological inhibition and genetic
knockdown of exchange protein directly activated by cAMP 1 reduce
pancreatic cancer metastasis in vivo. Mol. Pharmacol. 2015, 87, 142−
149.
(16) Hu, Y.; Robichaux, W. G.; Mei, F. C.; Kim, E. R.; Wang, H.;
Tong, Q.; Jin, J.; Xu, M.; Chen, J.; Cheng, X. Role of exchange protein
directly activated by cyclic AMP isoform 1 in energy homeostasis:
L
https://dx.doi.org/10.1021/acs.jmedchem.9b02151
J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
(34) Rehmann, H.; Arias-Palomo, E.; Hadders, M. A.; Schwede, F.;
Llorca, O.; Bos, J. L. Structure of Epac2 in complex with a cyclic AMP
analogue and RAP1B. Nature 2008, 455, 124−127.
tetrahydroquinoline analogs as EPAC inhibitors. ACS Med. Chem. Lett.
2017, 8, 1183−1187.
(52) Boulton, S.; Selvaratnam, R.; Blondeau, J.-P.; Lezoualc’h, F.;
Melacini, G. Mechanism of selective enzyme inhibition through
uncompetitive regulation of an allosteric agonist. J. Am. Chem. Soc.
2018, 140, 9624−9637.
(53) Ahmed, A.; Boulton, S.; Shao, H.; Akimoto, M.; Natarajan, A.;
Cheng, X.; Melacini, G. Recent advances in EPAC-targeted therapies: a
biophysical perspective. Cells 2019, 8, 1462.
(54) Parnell, E.; McElroy, S. P.; Wiejak, J.; Baillie, G. L.; Porter, A.;
Adams, D. R.; Rehmann, H.; Smith, B. O.; Yarwood, S. J. Identification
of a novel, small molecule partial agonist for the cyclic AMP sensor,
EPAC1. Sci. Rep. 2017, 7, No. 294.
(55) Schwede, F.; Maronde, E.; Genieser, H.-G.; Jastorff, B. Cyclic
nucleotide analogs as biochemical tools and prospective drugs.
Pharmacol. Ther. 2000, 87, 199−226.
(35) Rehmann, H.; Prakash, B.; Wolf, E.; Rueppel, A.; de Rooij, J.; Bos,
J. L.; Wittinghofer, A. Structure and regulation of the cAMP-binding
domains of Epac2. Nat. Struct. Biol. 2003, 10, 26−32.
(36) Kraemer, A.; Rehmann, H. R.; Cool, R. H.; Theiss, C.; de Rooij,
J.; Bos, J. L.; Wittinghofer, A. Dynamic interaction of cAMP with the
Rap guanine-nucleotide exchange factor Epac1. J. Mol. Biol. 2001, 306,
1167−1177.
(37) Kukic, P.; Alvin Leung, H. T.; Bemporad, F.; Aprile, F. A.;
Kumita, J. R.; Simone, A. de.; Camilloni, C.; Vendruscolo, M. Structure
and dynamics of the integrin LFA-1 I-domain in the inactive state
underlie its inside-out/outside-in signaling and allosteric mechanisms.
Structure 2015, 23, 745−753.
́
(38) Gagne, D.; Narayanan, C.; Doucet, N. Network of long-range
(56) Enserink, J. M.; Christensen, A. E.; de Rooij, J.; van Triest, M.;
Schwede, F.; Genieser, H. G.; Døskeland, S. O.; Blank, J. L.; Bos, J. L. A
novel Epac-specific cAMP analogue demonstrates independent
regulation of Rap1 and ERK. Nat. Cell Biol. 2002, 4, 901−906.
(57) Christensen, A. E.; Selheim, F.; de Rooij, J.; Dremier, S.;
Schwede, F.; Dao, K. K.; Martinez, A.; Maenhaut, C.; Bos, J. L.;
Genieser, H.-G.; Døskeland, S. O. cAMP analog mapping of Epac1 and
cAMP kinase. Discriminating analogs demonstrate that Epac and cAMP
kinase act synergistically to promote PC-12 cell neurite extension. J.
Biol. Chem. 2003, 278, 35394−35402.
(58) Rehmann, H.; Schwede, F.; Døskeland, S. O.; Wittinghofer, A.;
Bos, J. L. Ligand-mediated activation of the cAMP-responsive guanine
nucleotide exchange factor Epac. J. Biol. Chem. 2003, 278, 38548−
38556.
(59) Poppe, H.; Rybalkin, S. D.; Rehmann, H.; Hinds, T. R.; Tang, X.-
B.; Christensen, A. E.; Schwede, F.; Genieser, H.-G.; Bos, J. L.;
Doskeland, S. O.; Beavo, J. A.; Butt, E. Cyclic nucleotide analogs as
probes of signaling pathways. Nat. Methods 2008, 5, 277−278.
(60) Vliem, M. J.; Ponsioen, B.; Schwede, F.; Pannekoek, W.-J.; Riedl,
J.; Kooistra, M. R. H.; Jalink, K.; Genieser, H.-G.; Bos, J. L.; Rehmann,
H. 8-pCPT-2′-O-Me-cAMP-AM: an improved Epac-selective cAMP
analogue. ChemBioChem 2008, 9, 2052−2054.
concerted chemical shift displacements upon ligand binding to human
angiogenin. Protein Sci. 2015, 24, 525−533.
(39) Cembran, A.; Kim, J.; Gao, J.; Veglia, G. NMR mapping of
protein conformational landscapes using coordinated behavior of
chemical shifts upon ligand binding. Phys. Chem. Chem. Phys. 2014, 16,
6508−6518.
(40) De, S.; Greenwood, A. I.; Rogals, M. J.; Kovrigin, E. L.; Lu, K. P.;
Nicholson, L. K. Complete thermodynamic and kinetic characterization
of the isomer-specific interaction between Pin1-WW domain and the
amyloid precursor protein cytoplasmic tail phosphorylated at Thr668.
Biochemistry 2012, 51, 8583−8596.
(41) Boehr, D. D.; Nussinov, R.; Wright, P. E. The role of dynamic
conformational ensembles in biomolecular recognition. Nat. Chem.
Biol. 2009, 5, 789−796.
(42) de Simone, A.; Aprile, F. A.; Dhulesia, A.; Dobson, C. M.;
Vendruscolo, M. Structure of a low-population intermediate state in the
release of an enzyme product. eLife 2015, 4, No. e02777.
(43) Selvaratnam, R.; VanSchouwen, B.; Fogolari, F.; Mazhab-Jafari,
M. T.; Das, R.; Melacini, G. The projection analysis of NMR chemical
shifts reveals extended EPAC autoinhibition determinants. Biophys. J.
2012, 102, 630−639.
(44) Tsalkova, T.; Mei, F. C.; Li, S.; Chepurny, O. G.; Leech, C. A.;
Liu, T.; Holz, G. G.; Woods, V. L.; Cheng, X. Isoform-specific
antagonists of exchange proteins directly activated by cAMP. Proc. Natl.
Acad. Sci. U.S.A. 2012, 109, 18613−18618.
(45) Zhu, Y.; Chen, H.; Boulton, S.; Mei, F.; Ye, N.; Melacini, G.;
Zhou, J.; Cheng, X. Biochemical and pharmacological characterizations
of ESI-09 based EPAC inhibitors: defining the ESI-09 “therapeutic
window”. Sci. Rep. 2015, 5, No. 9344.
(46) Brown, L. M.; Rogers, K. E.; McCammon, J. A.; Insel, P. A.
Identification and validation of modulators of exchange protein
activated by cAMP (Epac) activity: structure-function implications
for Epac activation and inhibition. J. Biol. Chem. 2014, 289, 8217−8230.
(47) Brown, L. M.; Rogers, K. E.; Aroonsakool, N.; McCammon, J. A.;
Insel, P. A. Allosteric inhibition of Epac: computational modeling and
experimental validation to identify allosteric sites and inhibitors. J. Biol.
Chem. 2014, 289, 29148−29157.
(61) Dodge-Kafka, K. L.; Soughayer, J.; Pare, G. C.; Carlisle Michel, J.
J.; Langeberg, L. K.; Kapiloff, M. S.; Scott, J. D. The protein kinase A
anchoring protein mAKAP coordinates two integrated cAMP effector
pathways. Nature 2005, 437, 574−578.
(62) Conti, M.; Beavo, J. Biochemistry and physiology of cyclic
nucleotide phosphodiesterases: essential components in cyclic
nucleotide signaling. Annu. Rev. Biochem. 2007, 76, 481−511.
́
(63) Wiejak, J.; van Basten, B.; Luchowska-Stanska, U.; Hamilton, G.;
Yarwood, S. J. The novel exchange protein activated by cyclic AMP 1
(EPAC1) agonist, I942, regulates inflammatory gene expression in
human umbilical vascular endothelial cells (HUVECs). Biochim.
Biophys. Acta Mol. Cell Res. 2019, 1866, 264−276.
(64) Barker, G.; Parnell, E.; van Basten, B.; Buist, H.; Adams, D. R.;
Yarwood, S. J. The potential of a novel class of EPAC-selective agonists
to combat cardiovascular Inflammation. J. Cardiovasc. Dev. Dis. 2017, 4,
22.
(48) Laudette, M.; Coluccia, A.; Sainte-Marie, Y.; Solari, A.; Fazal, L.;
Sicard, P.; Silvestri, R.; Mialet-Perez, J.; Pons, S.; Ghaleh, B.; Blondeau,
J.-P.; Lezoualc’h, F. Identification of a pharmacological inhibitor of
Epac1 that protects the heart against acute and chronic models of
cardiac stress. Cardiovasc. Res. 2019, 115, 1766−1777.
(49) Courilleau, D.; Bisserier, M.; Jullian, J.-C.; Lucas, A.; Bouyssou,
P.; Fischmeister, R.; Blondeau, J.-P.; Lezoualc’h, F. Identification of a
tetrahydroquinoline analog as a pharmacological inhibitor of the
cAMP-binding protein Epac. J. Biol. Chem. 2012, 287, 44192−44202.
(50) Courilleau, D.; Bouyssou, P.; Fischmeister, R.; Lezoualc’h, F.;
Blondeau, J.-P. The R-enantiomer of CE3F4 is a preferential inhibitor
of human exchange protein directly activated by cyclic AMP isoform 1
(Epac1). Biochem. Biophys. Res. Commun. 2013, 440, 443−448.
(51) Sonawane, Y. A.; Zhu, Y.; Garrison, J. C.; Ezell, E. L.; Zahid, M.;
Cheng, X.; Natarajan, A. Structure-activity relationship studies with
(65) Tsalkova, T.; Mei, F. C.; Cheng, X. A fluorescence-based high-
throughput assay for the discovery of exchange protein directly
activated by cyclic AMP (EPAC) antagonists. PLoS ONE 2012, 7,
No. e30441.
(66) Boulton, S.; Selvaratnam, R.; Ahmed, R.; Van, K.; Cheng, X.;
Melacini, G. Mechanisms of specific versus nonspecific interactions of
aggregation-prone inhibitors and attenuators. J. Med. Chem. 2019, 62,
5063−5079.
(67) Selvaratnam, R.; Chowdhury, S.; VanSchouwen, B.; Melacini, G.
Mapping allostery through the covariance analysis of NMR chemical
shifts. Proc. Natl. Acad. Sci. U.S.A. 2011, 108, 6133−6138.
(68) Eghbalnia, H. R.; Wang, L.; Bahrami, A.; Assadi, A.; Markley, J. L.
Protein energetic conformational analysis from NMR chemical shifts
(PECAN) and its use in determining secondary structural elements. J.
Biomol. NMR 2005, 32, 71−81.
M
https://dx.doi.org/10.1021/acs.jmedchem.9b02151
J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
pubs.acs.org/jmc
Article
(69) Zwahlen, C.; Legault, P.; Vincent, S. J. F.; Greenblatt, J.; Konrat,
R.; Kay, L. E. Methods for measurement of intermolecular NOEs by
multinuclear NMR spectroscopy: application to a bacteriophage λ N-
peptide/ boxB RNA complex. J. Am. Chem. Soc. 1997, 119, 6711−6721.
(70) Melacini, G. Separation of intra- and intermolecular NOEs
through simultaneous editing and J -compensated filtering: A 4D
quadrature-free constant-time J -resolved approach. J. Am. Chem. Soc.
2000, 122, 9735−9738.
(71) Medek, A.; Hajduk, P. J.; Mack, J.; Fesik, S. W. The use of
differential chemical shifts for determining the binding site location and
orientation of protein-bound ligands. J. Am. Chem. Soc. 2000, 122,
1241−1242.
(72) Mayer, M.; Meyer, B. Group epitope mapping by saturation
transfer difference NMR to identify segments of a ligand in direct
contact with a protein receptor. J. Am. Chem. Soc. 2001, 123, 6108−
6117.
(73) Das, R.; Chowdhury, S.; Mazhab-Jafari, M. T.; SilDas, S.;
Selvaratnam, R.; Melacini, G. Dynamically driven ligand selectivity in
cyclic nucleotide binding domains. J. Biol. Chem. 2009, 284, 23682−
23696.
(74) Delaglio, F.; Grzesiek, S.; Vuister, G. W.; Zhu, G.; Pfeifer, J.; Bax,
A. NMRPipe: a multidimensional spectral processing system based on
UNIX pipes. J. Biomol. NMR 1995, 6, 277−293.
(75) Goddard, T. D.; Kneller, D. G.Sparky 3; University of California:
San Francisco, 2008.
(76) Yamazaki, T.; Pascal, S. M.; Singer, A. U.; Forman-Kay, J. D.; Kay,
L. E. NMR pulse schemes for the sequence-specific assignment of
arginine guanidino 15N and 1H chemical shifts in proteins. J. Am.
Chem. Soc. 1995, 117, 3556−3564.
(77) Wang, Z.-X. An exact mathematical expression for describing
competitive binding of two different ligands to a protein molecule.
FEBS Lett. 1995, 360, 111−114.
(78) Byun, J. A.; Melacini, G. NMR methods to dissect the molecular
mechanisms of disease-related mutations (DRMs): Understanding how
DRMs remodel functional free energy landscapes. Methods 2018, 148,
19−27.
(79) de Hoon, M. J. L.; Imoto, S.; Nolan, J.; Miyano, S. Open source
clustering software. Bioinformatics 2004, 20, 1453−1454.
́
(80) Wang, P.; Luchowska-Stanska, U.; Chen, H.; Liu, Z.; Wiejak, J.;
Whelan, P.; Morgan, D.; Lochhead, E.; Barker, G.; Rehmann, H.;
Yarwood, S. J.; Zhou, J. Synthesis and biochemical evaluation of non-
cyclic nucleotide exchange protein directly activated by cAMP (EPAC)
partial agonists. Unpublished results.
(81) VanSchouwen, B.; Selvaratnam, R.; Giri, R.; Lorenz, R.; Herberg,
F. W.; Kim, C.; Melacini, G. Mechanism of cAMP partial agonism in
protein kinase G (PKG). J. Biol. Chem. 2015, 290, 28631−28641.
(82) Boulton, S.; Melacini, G. Advances in NMR methods to map
allosteric sites: from models to translation. Chem. Rev. 2016, 116,
6267−6304.
N
https://dx.doi.org/10.1021/acs.jmedchem.9b02151
J. Med. Chem. XXXX, XXX, XXX−XXX