Molecular Modeling Evaluation of the Enantiomers of a Novel Adenylyl Cyclase 2 Inhibitor

Article abstract of DOI:10.1021/acs.jcim.6b00454

Adenylyl cyclase 2 (AC2) is one of nine membrane-bound isoforms of adenylyl cyclase that converts ATP into cyclic AMP (cAMP), an important second messenger molecule. Upregulation of AC2 is linked to cancers like pancreatic and small intestinal neuroendocrine tumors (NETs). The structures of the various isoforms of adenylyl cyclases are highly homologous, posing a significant challenge to drug discovery efforts for an effective, isoform-selective modulator of AC2. In a previous study, a screen identified a potential isoform-selective and noncompetitive inhibitor of AC2, SKF83566. In the present study, molecular modeling is used to explore the mode of inhibition of AC2 by SKF83566 and to investigate the active enantiomer of SKF83566. Homology models of hAC2 were built based on canine AC5-C1a and rat AC2-C2a templates. With these models, a combination of flexible docking, molecular dynamics simulations, and free energy calculations using the MM/GBSA methodology suggested an allosteric mechanism in which (S)-SKF83566 binds to an allosteric site near ATP and alters the protein conformation of the ATP binding site, potentially preventing the adenosine moiety of ATP from forming an archlike shape to form cAMP. The predicted binding preference for the (S)-SKF83566 enantiomer and the predicted free energy are consistent with the experimental data.

Full text of DOI:10.1021/acs.jcim.6b00454

Subscriber access provided by Olson Library | Northern Michigan University
Article
Molecular Modeling Evaluation of the Enantiomers
of a Novel Adenylyl Cyclase 2 Inhibitor
Neha Rana, Jason M. Conley, Monica Soto-Velasquez, Francisco
Leon, Stephen J. Cutler, Val J Watts, and Markus A. Lill
J. Chem. Inf. Model., Just Accepted Manuscript • DOI: 10.1021/acs.jcim.6b00454 • Publication Date (Web): 09 Jan 2017
Downloaded from http://pubs.acs.org on January 10, 2017
Just Accepted
“Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted
online prior to technical editing, formatting for publication and author proofing. The American Chemical
Society provides “Just Accepted” as a free service to the research community to expedite the
dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts
appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been
fully peer reviewed, but should not be considered the official version of record. They are accessible to all
readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered
to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published
in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just
Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor
changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers
and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors
or consequences arising from the use of information contained in these “Just Accepted” manuscripts.
Journal of Chemical Information and Modeling is published by the American Chemical
Society. 1155 Sixteenth Street N.W., Washington, DC 20036
Published by American Chemical Society. Copyright © American Chemical Society.
However, no copyright claim is made to original U.S. Government works, or works
produced by employees of any Commonwealth realm Crown government in the course
of their duties.

Page 1 of 41
Journal of Chemical Information and Modeling
1
2
3
4
5
6
7
8
9
Molecular Modeling Evaluation of the Enantiomers of a Novel
Adenylyl Cyclase 2 Inhibitor
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
1
1
1
2
2
Neha Rana , Jason M. Conley , Monica Soto-Velasquez , Francisco León , Stephen J. Cutler , Val
1
1*
J. Watts , Markus A. Lill
1
Department of Medicinal Chemistry & Molecular Pharmacology, College of Pharmacy, Purdue
University, West Lafayette, IN 47907
2
Department of BioMolecular Sciences, School of Pharmacy, The University of Mississippi, Oxford, MS
38677
ACS Paragon Plus Environment

Journal of Chemical Information and Modeling
Page 2 of 41
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
Abstract
Adenylyl cyclase 2 (AC2) is one of nine membraneꢀbound isoforms of adenylyl cyclase that converts
ATP into cyclic AMP (cAMP), an important second messenger molecule. Up–regulation of AC2 is linked
to cancers like pancreatic and small intestinal neuroendocrine tumors (NETs). The structures of the
various isoforms of adenylyl cyclases are highly homologous, posing a significant challenge to drug
discovery efforts for an effective, isoform–selective modulator of AC2. In a previous study, a screen
identified a potential isoform–selective and non–competitive inhibitor of AC2, SKF83566. In the present
study, molecular modeling is used to explore the mode of inhibition of AC2 by SKF83566 and to
investigate the active enantiomer of SKF83566. Homology models of hAC2 were built based on canine
AC5–C1a and rat AC2–C2a templates. Using these models, a combination of flexible docking, molecular
dynamics (MD) simulations, and free energy calculations using the MM/GBSA methodology suggested
an allosteric mechanism where (S)ꢀSKF83566 binds to an allosteric site near ATP, alters the protein
conformation of the ATP binding site, potentially preventing the adenosine moiety of ATP from forming
an arch like shape to form cAMP. The predicted binding preference for the (S)ꢀSKF83566 enantiomer and
the predicted free energy is consistent with the experimental data.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
ACS Paragon Plus Environment

Page 3 of 41
Journal of Chemical Information and Modeling
1
2
3
4
5
6
7
8
9
Introduction
The transmembrane adenylyl cyclases (AC) are enzymes that upon activation convert intracellular
1
ATP into cAMP, an important secondary messenger involved in numerous physiological responses.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Adenylyl cyclases are important downstream effectors of G proteinꢀcoupled receptor signaling pathways
2
+
and are consequently modulated by the actions of G proteins, protein kinases, and Ca . These regulatory
properties, as well as their sequence homology, form the basis for the classification of the nine
1
mammalian membraneꢀbound isoforms of AC into four families. Group I adenylyl cyclases AC1, AC3,
2+
and AC8 are often referred to as the Ca ꢀstimulated family. The Group II adenylyl cyclase isoforms
(
AC2, AC4, and AC7) are characterized by their ability to be conditionally activated by Gβγ subunits.
2+
Group III adenylyl cyclases include AC5 and AC6 and share the feature of being inhibited by free Ca .
AC9 makes up Group IV, the only adenylyl cyclase isoform that is not robustly activated by forskolin.
Structurally, mammalian adenylyl cyclases (AC) are a single polypeptide chain containing an
intracellular Nꢀterminus, followed by two membrane spanning domains alternating with two cytoplasmic
domains that can be further divided into a and b regions (Figure 1). The two cytoplasmic “a” domains
(
C1a and C2a) form a heterodimer that is sufficient for catalytic activity.
Figure 1: Scheme of structural domains of membrane bound adenylyl cyclase
ACS Paragon Plus Environment

Journal of Chemical Information and Modeling
Page 4 of 41
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
The C1a and C2a monomers contain a region of ~200 amino acids and are highly homologous. These
domains were used to solve the AC crystal structure in the presence of its activator, Gα , forming the
S
ternary complex of Gα ꢀAC5.C1/AC2.C2 (Figure 2). The structure contained the C1a subunit of type 5
S
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
AC from canine (VC1) and C2a subunit of type 2 AC from rat (IIC2). The structure was also bound with
a forskolin analog revealing a C1a/C2a dimer face that forms a long, shallow diagonallyꢀrunning trough,
containing both the forskolin and substrate binding sites (Figure 2).
Figure 2: (Left) Proposed structure of hAC2 (C1a subdomain in yellow, C2a subdomain in orange). The
green structure is GαS subunit crystallized with VC1.IIC2 complex (PDB 3C16), obtained by
superimposing crystal structure onto hAC2 model. (Right) The ventral face is rotated about the horizontal
axis to show forskolin (cyan), ATP (pink) and proposed SKF83556 (green) binding sites.
The extensive contacts between C1a and C2a restrict the entry of the substrates from the dorsal
surface, which is also believed to be interacting with the membrane because the Nꢀtermini of both
subdomains (C1a, C2a) project away from the dorsal surface. The ventral surface facing the cytoplasm
allows access of forskolin and substrates from the cell interior to the binding site. The inverted C1a and
C2a cytoplasmic domains enclose only one catalytic site at their interface. The other site, homologous and
ACS Paragon Plus Environment

Page 5 of 41
Journal of Chemical Information and Modeling
1
2
3
4
5
6
7
8
9
symmetric to the catalytic site, is nonꢀactive due to lack of catalytic residues such as ASPꢀ295 and ASPꢀ
339 (hAC2 numbering) and is therefore, called pseudoꢀsymmetric site. The pseudoꢀsymmetric site is
exploited by the nonꢀphysiological diterpenic activator, forskolin, which results in an increase in the level
2
2+
2+
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
of cAMP production. The conversion of ATP to cAMP is facilitated by two cations (Mg /Mg or
2
+
2+
Mg /Mn ) in the catalytic site which in turn are stabilized by two ASP residues, ASPꢀ295 (βꢀstrand 1)
and ASPꢀ339 (βꢀstrand 3) (hAC2 numbering). Concurrently, the adenosine moiety of ATP is predicted to
2
interact with LEUꢀ337 (hAC2 numbering). The primary difference between the forskolin and ATP
binding sites is that THRꢀ411 and SERꢀ941 (hAC2 numbering) in the forskolin binding site are
substituted by ASNꢀ1026 and ASPꢀ339 (hAC2 numbering), respectively, in the ATP binding site.
2
Forskolin thus cannot bind in the ATP binding site as the latter residues would result in steric conflict.
Similarly, ATP cannot be catalytically converted into cAMP in the forskolin binding site because of the
2
lack of catalytic residues, ASPꢀ295 and ASPꢀ339 in the forskolin binding site.
Notably, the C1b has a molecular weight of nearly ~15kDa. Its structure varies among different
isozymes and its primary function is predicted to be isoform specific regulation of enzyme activity by G
1
protein subunits and other proteins. The C2b domain is either very small or nonꢀexistent in some
isozymes and its function still has to be characterized. All Xꢀray structures of hAC are devoid of atomic
data on these regions. Thus, we have excluded these sections from our modeling study.
There is a growing interest in pursuing adenylyl cyclase (AC) isoforms as drug targets in a variety of
3
disease states. This interest is based on the relationship of AC isoform expression patterns and the
1
contribution of specific AC isoforms to distinct physiological processes. More direct evidence
supporting adenylyl cyclases as therapeutic targets is derived from genetic mouse models. Over the past
1
5 years, a variety of knockout and transgenic mouse model studies have identified that AC isoforms
uniquely contribute to diverse physiological processes that include aging, learning and memory, pain and
1
analgesia, cardiac function, ethanol responses, and obesity.
The AC isoform expression patterns,
ACS Paragon Plus Environment

Journal of Chemical Information and Modeling
Page 6 of 41
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
together with the genetic mouse model studies, suggest that small molecule modulation of AC isoforms
1
may yield therapeutics with novel efficacy and/or specificity.
In an effort to identify inhibitors of AC2, we previously completed a successful small molecule
screening effort (ca. 720 compounds) where we identified the phenylbenzazepine derivative, SKF83566
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
4
as the most robust and potent inhibitor of AC2 known to date. The detailed atomistic mechanism of
inhibition of AC2 by SKF83566, however, is unknown. Furthermore, our studies used the commercially
available racemic mixture of SKF83566, which in addition to inhibiting AC2, is also a potent D1
5
dopamine receptor antagonist. We hypothesized that molecular modeling and docking studies would
allow us to identify the active enantiomer and to understand further the mechanism of inhibition by
SKF83566.
Materials and Methods
Molecular Modeling
The overall design of the molecular modeling process is shown as a flowchart in Figure 3.
Homology modeling: The target hAC2 sequence was downloaded from the UNIPROT KB database (Id:
Q08462) and curated by removing the transmembrane regions from each monomer, 1–207 and 602–821,
yielding C1 (monomer 1: Chain A: 208–601) and C2 (monomer 2: chain B: 822–1091) cytoplasmic
6
domains. Both chains C1 and C2 have further two domains: a and b, of which b domains are not
considered in this study because they are neither conserved nor close to the active site. The template
2
7
structures were taken from 1AZS and 3C16 , both of which used the C1a domain from canine AC5 and
the C2a domain from rat AC2. In 1AZS, only the activator forskolin is bound to the pseudoꢀsymmetric
site and no ATP and ions are present in the active site. In 3C16, both forskolin in the pseudoꢀsymmetric
2+
site, and ATP/Ca in the active site are coꢀcrystallized with AC. The hAC2 sequence was aligned to the
8
1
AZS and 3C16 sequences independently using BLASTp , and these alignments were used to generate
ACS Paragon Plus Environment

Page 7 of 41
Journal of Chemical Information and Modeling
1
2
3
4
5
6
7
8
9
two sets of ten homology models of the hAC2 heterodimer. The homology models were generated using
9
Modeller 9.10. DOPE scores were used to estimate the quality of the models.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 3: Flowchart of molecular modeling study.
The models, derived from 1AZS, did not contain cations whereas the models, derived from 3C16,
automatically have two cations in the binding site of the protein. All homology models were
10
superimposed using PyMol , and the coordinates of the cations in 3C16 were added to the models
derived from 1AZS. ATP coordinates were obtained from the aligned template AC2 structure, 3C16.
Ligand library: SKF83556 was identified as most active inhibitor of AC2 in the previous screening
4
assay. Its three dimensional structure was built using PyMol. The structure was corrected using the
ACS Paragon Plus Environment

Journal of Chemical Information and Modeling
Page 8 of 41
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
sculpting feature of PyMol. Plausible protonation states of SKF83556 were generated using the H++ 3.0
11
server , rendering positively charged SKF83566 as alternative protonation states at physiological pH.
C
C
Furthermore, the stereoisomers of neutral SKF83566 (R ꢀ and S ꢀ) were generated along with the
N
N
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
stereoisomers of the protonated Nꢀatom on the phenylbenzazepine moiety (R ꢀ and S ꢀ isoforms). The
complete library of different enantiomers and protonation states is shown in Figure 4.
Preliminary binding site identification: All SKF83566 stereoisomers were docked into the extended
1
2
heterodimer interface region of the homology models using AutoDock4.2. The protein was kept rigid
through the initial docking study. The dimensions of the search volume (i.e. docking box) were chosen
such that the ligandꢀaccessible pockets along the dimer interface, including pseudoꢀsymmetric and active
site, were enclosed within the box. Based on the size of the ligand and number of degrees of torsional
freedom, each docking run was performed using the Lamarckian Genetic algorithm (LGA) search method
for 250,000 steps of energy evaluations, and 27,000 generations, whichever is reached first. 100 final
poses were obtained, which were clustered based on their rootꢀmean square deviation (rmsd = 2 Å). The
clusters were ranked based on the score of the topꢀscored pose within each cluster.
ACS Paragon Plus Environment

Page 9 of 41
Journal of Chemical Information and Modeling
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 4: Chemical structures of ligands used in the flexible docking study.
Protein flexibility and Limoc ensemble generation: To include protein flexibility into docking, MD
simulations using Limoc were performed on all selected homology models as starting structures. Limoc
probes were distributed to the binding site, which was defined by the energetically lowest docking poses
of SKF83655 obtained in the previous docking runs. The Limoc ensembles were clustered separately for
the 3C16ꢀbased and 1AZSꢀbased homology models using QT clustering with a cluster radius of 1 Å, thus
generating two ensembles for subsequent docking. A detailed description of the Limoc MD approach can
13 14
be found in our previous publications.
Flexible docking in Limoc ensemble: First, the ensemble of Limoc structures based on the 3C16 template
was used for docking all stereoisomers of SKF83566. The reason for selecting this ensemble was the
larger confidence in providing favorable ligandꢀengaging interactions due to 3C16’s prior ATP bound
conformation. As Limoc might underꢀsample the rotameric states of flexible side chains, the following
nine residues of the ATP binding site, which were found within a distance of 5 Å of ATP, were
ACS Paragon Plus Environment

Journal of Chemical Information and Modeling
Page 10 of 41
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
additionally treated as flexible entities during docking: LEUꢀ337, ASPꢀ339, LYSꢀ939, ILEꢀ941, METꢀ
946, VALꢀ1025, ASNꢀ1026, SERꢀ1029 and ARGꢀ1030. Due to the significant increase in number of
torsions, the number of energy evaluations in LGA was increased to 25,000,000 steps. Analyzing the side
chain flexibility throughout this docking run, the observed side chain variability for five out of the nine
residues was well represented by the Limoc ensembles. Thus, the final docking procedure on complete
Limoc ensemble treated the side chain flexibility explicitly only for remaining four residues: LYSꢀ939,
METꢀ946, ASNꢀ1026 and ARGꢀ1030. All poses of the final docking run were clustered based on 2Å
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
rmsd criteria for each ligand using AutoDockTools.
Charge fitting for modeling of potential halogen bond: SKF83566 contains a bromine atom, a halogen,
which displays unique electronic distribution and is poorly modeled by current atomꢀcentered partial
charges force fields. Halogen atoms such as bromine or iodine display partial negative electrostatic
potential perpendicular to the axis of the carbonꢀhalogen bond and partial positive charge in the center,
the soꢀcalled sigma hole, which can potentially act as a donor to form a halogen bond with a negatively
15
charged acceptor as described in Figure 5. To model these electrostatic properties more accurately
compared to standard force fields, we computed the electrostatic potential of the initial ligand
conformations obtained from the two topꢀranked docking poses of SKF83566 isomers using quantum
mechanics and redistributed the partial charges on ligand atoms and a newly introduced virtual point on
1
6
the halogen surface along the carbonꢀhalogen axis, using a previously established procedure. In detail,
17
DFT calculations using Gaussian 09 were performed with mixed basis set of 6ꢀ311+G** for C, H, N,
and O atoms, and an augmented correlationꢀconsistent valence doubleꢀzeta basis set with a polarization
1
6
function, augꢀccꢀpVDZꢀPP, for the bromine atom. A geometry optimization and minimization was
followed by electrostatic field calculation. An additional dummy atom was created at the center of the
periphery of the bromine atom along the CꢀBr axis at a distance of 2.22 Å from the bromine atom center
and charges were fitted to the new distribution of atoms (including dummy atom) using the RESP charge
18
fitting method in Antechamber.
ACS Paragon Plus Environment

Page 11 of 41
Journal of Chemical Information and Modeling
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
1
6
Figure 5: Sigma hole of halogen atom: X is the halogen, A is an acceptor, HꢀD is a donor moiety.
Molecular dynamics simulations: MD simulations were carried out for selected binding poses of
SKF83566. The ligand, enzyme and complex topology files were prepared using tleap and the
1
9
antechamber program of AmberTools13 using the Amber99SB force field. The parameter file of the
halogenated ligand was modified by adding the parameters for the dummy atom (cf. Table 1).
The proteinꢀligand complex was placed in a rectangular box with a minimum distance between face and
nearest atom of 8 Å. The box was solvated using the TIP3P water model and the system was neutralized
by adding two and three Na+ ions for protonated SKF83566 and neutral SKF83566, respectively.
Table 1: Molecular mechanics parameters of extra atom representing the sigma hole.
Parameters of extra atom
Mass
VdW radius
VdW well depth (ε)
Br↔EP bond length
Values
0.00 amu
1.00 Å
0.00 Å
2.22 Å
2
600.0
Br↔EP bond force constant (kcal/mol/Å )
C↔Br↔EP angle phase
180.0°
150.0
2
C↔Br↔EP angle force constant (kcal/mol/deg )
C↔C↔X↔EP dihedral phase
C↔C↔X↔EP dihedral periodicity
180.0°
2
ACS Paragon Plus Environment

Journal of Chemical Information and Modeling
Page 12 of 41
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
After a minimization of 1000 (500 steepest descent and 500 conjugate gradient) steps, the system was
slowly heated up for 50 ps from 0 K to 300 K. After heating up, the system was equilibrated under NPT
conditions for 550 ps using the isotropicꢀposition scaling in periodic boundary condition method to keep
the pressure constant at 1 atm. Final production MD simulations were performed for 50 ns under NPT
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
2
0 21
conditions using the GPU supported pmemd MD module.
All the bonds involving hydrogen atoms
were constrained using the SHAKE algorithm. ParticleꢀMeshꢀEwald methodology was used to calculate
all longꢀrange electrostatic interactions with a distance larger than 8.0 Å. For temperature scaling,
Langevin dynamics was used with collision frequency of 2 ps.
Binding free energy calculations: The binding free energies of the different ligand poses were estimated
2
2
using the MM/GBSA (Molecular Mechanics/Generalized Born Surface Area) method. The trajectory of
each proteinꢀligand complex simulation was used to extract 5000 snapshots for complex, enzyme and
ligand. The free energy without ligand configurational entropy was estimated by the average difference
between complex and individual protein plus ligand interactions, where solvation effects were treated
using an implicit GB solvation model. For estimating the vibrational entropy of the ligand, the normal
mode analysis method of the Amber suite was used. 48 snapshots were used to estimate the components
of entropy: translational, rotational and vibrational entropy. The difference between the binding free
energy of poses was statistically tested using student’s tꢀtest.
Experimental Methods
Materials
Dopamine, 3ꢀisobutylꢀ1ꢀmethylxanthine (IBMX), G418, and LꢀAscorbic acid were purchased from
SigmaꢀAldrich (St. Louis, MO). Phorbol 12ꢀmyristate 13ꢀacetate (PMA), SQ22536 and (±)ꢀSKF83566
were purchased from Tocris Bioscience (Ellisville, MO). OptiꢀMEM, Dulbecco’s modified Eagle’s
medium (DMEM), and antibioticꢀantimycotic 100x solution were purchased from Life technologies
ACS Paragon Plus Environment

Page 13 of 41
Journal of Chemical Information and Modeling
1
2
3
4
5
6
7
8
9
(
Grand Island, NY). Lipofectamine 2000 was purchased from Thermo Fisher Scientific. FetalClone I
serum, and bovine calf serum were purchased from Hyclone (Logan, UT). The HTRF cAMP kits were
purchased from Cisbio Bioassays (Bedford, MA).
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Chemistry
1
H NMR spectra were recorded on a Bruker model AMX 400 NMR spectrometer with standard pulse
1
sequences, operating at 400 MHz in H NMR; chemical shifts are reported in ppm with the solvent
reference relative to tetramethylsilane (TMS) employed as the internal standard (CD OD, δ = 3.31 ppm).
3
13
C NMR were recorder on a 400 MHz spectrometer operating at 100 MHz, with complete proton
decoupling, chemical shifts are reported in ppm. High performance liquid chromatography (HPLC) runs
were conducted on a Waters LC Module I autosampler equipped with a UV detector 486ꢀwaters.
Experimental data were acquired and processed by Millenium 32 Chromatography Manager software
(Waters). Solvents used for chiral chromatography were HPLC grade and supplied by Fisher Scientific
(Hanover Parks, USA). All HPLC analyses were performed at room temperature. Optical rotation values
were measured on a Rudolph Research Analytical Autopol V Polarimeter (Hackettstown, USA) with a
1cm cell at the sodium D line (l = 589 nm); sample concentration values (c) are given in 1 mg/mL.
Circular dichroism (CD) spectra were recorded on an Aviv 202SF spectrometer (Lakewood, USA).
Chiral chromatography
The enantiomers of SKF83566 were completely resolved by a semiꢀpreparative process using a
Quiral Lux Celluloseꢀ1 column (Ø = 1 cm, l = 2 cm, 5 ꢁm) (Phenomenex), eluting with
CH OH/iProp/formic acid 90:10:0.1 e.c. at a flow rate of 0.8 mL/min. The eluate was properly partitioned
3
according to the UV profile. Sample volumes of 10 ꢁL were injected onto the column. The 10 mg
SKF83566 sample was dissolved in 2 mL of methanol. The fractions obtained containing the enantiomers
were evaporated at reduced pressure.
ACS Paragon Plus Environment

Journal of Chemical Information and Modeling
Page 14 of 41
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
(
-)-(S)-8-bromo-2,3,4,5-tetrahydro-3-methyl-5-phenyl-1H-3-Benzazepin-7-ol
[(-)-(S)-SKF83566]:
25
1
solid colorless; HPLC t = 13.72; [α] : ꢀ4.0 (c 0.01, MeOH); H NMR (400 MHz, CD OD): δ = 7.47 (t,
R
D
3
J = 7.6 Hz, 2H), 7.40 (s, 1H), 7.39 (m, 2H), 7.27 (d, J = 7.6 Hz, 2H), 6.28 (s br, 1H), 4.61 (d, J = 10.4 Hz,
1H), 3.89ꢀ3.73 (m, 2H), 3.71ꢀ3.51 (m, 1H), 3.31ꢀ3.27 (m, 1H), 3.23ꢀ3.13 (m, 1H), 3.04 (dd, J = 7.2; 16.0
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
13
Hz, 1H), 2.97 (s, 3H); C NMR (100 MHz, CD OD): δ = 155.3, 142.1, 139.4, 133.8, 129.9, 129.0, 128.0,
3
1
27.4, 116.4, 107.6, 60.1, 56.1, 45.4, 44.5, 30.3 ppm; CD (MeOH, c 0.14): [θ]279 ꢀ 7535 (max), [θ]257 ꢀ 157
(
max).
(
+)-(R)-8-bromo-2,3,4,5-tetrahydro-3-methyl-5-phenyl-1H-3-Benzazepin-7-ol [(+)-(R)-SKF83566]:
25
solid colorless; HPLC t = 14.70; [α] D: + 8.0 (c 0.01, MeOH); CD (MeOH, c 0.14): [θ]272 – 3800 (max),
R
[
θ]257 ꢀ 5573 (max), [θ]233 ꢀ 701 (max).
Cell culture and pharmacology
HEK293 cells were cultured in DMEM supplemented with 5% bovine calf serum, 5% fetal clone I,
and 1% antibioticꢀantimycotic 100x solution and maintained in a humidified incubator at 37°C and 5%
4
CO . HEKꢀAC2 cells were constructed as previously described and HEKꢀhD1 cells were previously
2
23
described.
Cyclic AMP assays: The cellular cAMP levels were measured using the Cisbio HTRF cAMP
dynamic 2 assay kit. For transient transfections, HEK cells were plated in a 12ꢀwell plate at a cell density
5
of 3 x 10 cells per well. After 24 hours, cells were transiently transfected for 48 hours with Gα , and rat
s
AC2 (rAC2) or the forskolin insensitive rAC2 mutant, S942P in a 1:10 [DNA] ratio, respectively. Cells
were harvested in OptiꢀMEM, plated in a low volume 384ꢀwell plate and incubated for 1 hour in a
humidified incubator at 37°C and 5% CO . The cells were then treated with 30 µM (±)ꢀSKF83566 or 100
2
µM SQ22536 for 30 minutes at room temperature, followed by the addition of 100 nM PMA and 500 µM
IBMX to stimulate cAMP production. Assays were terminated after 1 h stimulation as described below.
For assays in stable cell lines, cryopreserved cells were rapidly thawed at 37°C and resuspended in Optiꢀ
ACS Paragon Plus Environment

Page 15 of 41
Journal of Chemical Information and Modeling
1
2
3
4
5
6
7
8
9
MEM. Cells were centrifuged at 500 × g and the supernatant was aspirated. Cells were washed by
resuspending in OptiꢀMEM and centrifuged at 500 × g. The supernatant was aspirated and cells were
seeded into a 384ꢀwell plate and allowed to incubate at 37°C and 5% CO for 1 h. Cells were then treated
2
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
as indicated with the SKF83566 enantiomers for 30 min, followed by the addition of PMA (HEKꢀAC2)
diluted in stimulation buffer (OptiꢀMEM, 500µM IBMX) or dopamine (HEKꢀD1) diluted in stimulation
buffer plus 0.02% ascorbic acid. The cells were incubated for 1 h at room temperature and the stimulation
was terminated by sequential addition of equal parts of cAMPꢀd2 and antiꢀcAMP cryptate conjugate (5
µLꢀ10 µL), each diluted (1:39) in lysis buffer. Following a 1 h incubation at room temperature, the timeꢀ
resolved fluorescence energy transfer (TRꢀFRET) was measured with a lag time of 100 ꢁs and integration
time of 300 ꢁs using a Synergy4 (BioTek) fluorescence plate reader (excitation filter: 330/80 nm and
emission filters: 620/10 nm and 665/8 nm). The resulting cAMP concentrations were calculated in
GraphPad Prism by applying the 620/665 nm fluorescence ratio values to a standard curve of known
cAMP concentrations.
Results and Discussion
Homology modeling
ACS Paragon Plus Environment

Journal of Chemical Information and Modeling
Page 16 of 41
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 6: Ensemble of 10 homology models based on each, 1AZS (apo–) and 3C16 (holo–) templates.
Different homology models have different side chain conformations, providing a pre–generated
ensemble of structures, thus implicitly including a fraction of the flexibility of the protein.
2+
Due to the presence of ATP and Ca in the template structure 3C16, there is a slight conformational
difference between the 1AZS and 3C16 Xꢀray crystal structures as shown in Supplementary Figure S1.
Both structures were employed in our study as they allowed consideration of both conformations of the
template structure, i.e. in the absence (apo–), as well as presence (holo–) of the substrate. The alignment
using BLASTp showed high sequence similarity between target and template sequences as illustrated in
Supplementary Figure S2 (overall sequence identity, 79%; 63% identity between respective chains
A (C1a domain) and 93% between respective chains B (C2a domain)), which in turn resulted into highly
homologous structures of hAC2. The following functionally and structurally important residues, as
ACS Paragon Plus Environment

Page 17 of 41
Journal of Chemical Information and Modeling
1
2
3
4
5
6
7
8
9
annotated by PROSITE, are conserved between template and target sequences: ASPꢀ295, ASPꢀ339, LEUꢀ
337, ASNꢀ1026 and ARGꢀ1030.
As is noted, the rat AC2ꢀC2a template is missing residues near the beginning of chain B in the
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
template structure from PROꢀ955 to GLUꢀ964.
Different homology models showed slight variations in the side chains of residues, including ATPꢀ
binding site residues ASPꢀ295, ASPꢀ339, LEUꢀ337, ASNꢀ1026, ARGꢀ1030, as shown in Figure 6. An
overlay of the perꢀresidue DOPE score of template and homology models is shown in Supplementary
Figure S3, which demonstrates the structural integrity of the model.
Preliminary binding site identification
Docking of SKF83566 into all 20 homology models of hAC2 identified two preferred sites for
binding, the active site where ATP binds and the pseudoꢀsymmetric site where forskolin binds (Figure 7).
After clustering, we obtained 53 different ligand configurations in the two preferred binding sites.
ACS Paragon Plus Environment

Journal of Chemical Information and Modeling
Page 18 of 41
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 7: All docking poses in ensemble of homology models are shown for different stereoisomers of
SKF83566. They are located in two different binding sites, the ATP and forskolin (FSK)ꢀbinding site in
all homology models. Similar poses in different homology models are shown in same colors: pose A
(pink), pose B (green), pose C (yellow), pose D (brown), pose E (cyan), pose F (purple), pose G (light
C
N
pink), pose H (dark green) and pose I (dark blue). (Top left) The poses of (S ,S )ꢀSKF83566 in ensemble
ACS Paragon Plus Environment

Page 19 of 41
Journal of Chemical Information and Modeling
1
2
3
4
5
6
7
8
9
of hAC2 homology models shown in two different sites in five different conformations (A,B,C,D,E).
C
N
(
Top right) The poses of (S ,R )ꢀSKF83566 shown in two binding sites in six conformations (A,B,C,D,E,
C
N
F). (Bottom left) The poses of (R ,S )ꢀSKF83566 bind in two binding sites in six conformations
C
N
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
(A,B,D,G,H,I). (Bottom right) The poses of (R ,R )ꢀSKF83566 bind in two binding sites in four
conformations (A,B,D,G).
One of the conformations, located in the active site (cf. Figure 7, marked in brown), was the most
frequently observed conformation across all stereoisomers of SKF83566 and all protein structures (30%
of 53 conformations, lowest Autodock energy = ꢀ6.72 kcal/mol). It was followed by two conformations,
located in the pseudoꢀsymmetric site (both 18%, ꢀ6.46 kcal/mol, marked in pink cf. Figure 7 and ꢀ6.68
kcal/mol, marked in green cf. Figure 7, respectively).
Protein flexibility in docking: Limoc ensemble generation plus flexible side chains
Limoc generated two ensembles of ten protein structures (EPS = ensemble of protein structures),
representing potential binding site conformations by using top docked proteinꢀligand conformation
(
colored brown) in each set of homology models. The significant differences between the conformations
of the side chains in the binding site of EPS and previous homology models are highlighted in
Supplementary Figure S4. The major differences occur in the binding site in loop regions between βꢀ
strands 2 and 3, and βꢀstrands 10 and 12.
Flexible docking of SKF83566 to the EPS based on the 3C16 templateꢀhomology model with nine
flexible residues revealed that out of the nine residues, the conformational space of the following five
residues, LEUꢀ337, ASPꢀ339, ILEꢀ941, VALꢀ1025 and SERꢀ1029, observed during flexible docking, was
covered in the complete Limoc ensemble derived from 1AZS and 3C16 templates. The remaining four
residues, LYSꢀ939, METꢀ946, ASNꢀ1026 and ARGꢀ1030, showed significantly larger conformational
variations in the flexible sideꢀchain docking protocol compared to the Limoc ensemble (Figure 8).
ACS Paragon Plus Environment

Journal of Chemical Information and Modeling
Page 20 of 41
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
Consequently, all subsequent docking was performed with these four residues kept flexible, whereas for
the other five residues, only the flexibility inherent to the Limoc ensemble was utilized.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 8: Comparison of Limoc ensemble (light brown and light green; cf. Supplementary Figure S4) and
residue conformations sampled during flexible docking (pink). The flexible docking results demonstrate
that the flexibility of the side chains of four residues (LYSꢀ939, ILEꢀ941, METꢀ946 and ARGꢀ1030) need
to be considered explicitly during docking, as these residues occupy a larger conformational space than
sampled by Limoc.
In the final docking run in complete Limoc ensemble, the representative conformations of the top five
ranked clusters were analyzed. Combining the poses of all protonation states and stereoisomers of
SKF83566, ten distinct binding poses were identified, four in the ATPꢀbinding site and six in the pseudoꢀ
symmetric site. In an effort to reduce the number of potential binding poses for future analyses we
experimentally examined the ability of SKF83566 to inhibit a forskolinꢀinsensitive mutant of AC2, S942P
(
rAC2 numbering). Substitution of the proline is predicted to distort the forskolin binding site and has
24
previously been shown to reduce forskolinꢀstimulated AC2 activity.
Our hypothesis was that if
ACS Paragon Plus Environment

Page 21 of 41
Journal of Chemical Information and Modeling
1
2
3
4
5
6
7
8
9
SKF83566 was acting via the pseudo symmetric forskolinꢀbinding site, disrupting that site would
decrease its inhibitory activity. Our experiments showed that SKF83566 inhibited the S942P mutant to
the same extent as wild type rat AC2 (Figure 9). Similar results were also obtained with the Pꢀsite
inhibitor SQ22536 (Figure 9).
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
These results suggest that the binding site(s) for SKF83566 and SQ22536 are distinct from the
forskolinꢀbinding site. Thus, only those poses of SKF83566 that were located in the ATP binding site
were further investigated.
Figure 9: Evaluation of SKF83566 and SQ22536 inhibitory effects on PMAꢀstimulated cAMP production
in HEK cells transiently transfected with rat AC2 or the forskolin insensitive AC2 mutant, S942P. Data
shown represent the average and S.E.M. of three independent experiments.
The two highest ranked poses for each SKF83566 stereoisomer were selected for stability analysis
and free energy calculations using MD simulations, shown in green and orange color in the ATP binding
site in Figure 10. Both poses were observed in all stereoisomers of SKF83566 and are partially
overlapping with the ATP binding site of the Xꢀray structure of AC.
ACS Paragon Plus Environment

Journal of Chemical Information and Modeling
Page 22 of 41
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
In the pose shown in green color, the hydroxyl group of SKF83566 enantiomers is at a distance of 4
Å from the carboxyl group of ASPꢀ1019 at an approximate angle of 150°. The phenyl group is partiallyꢀ
stacked with the side chain of PHEꢀ299, with an angle of approximately 45° between their planes and a
distance of 3.7 Å between their nearest heavy atoms, representing potential π–π interactions. The same
phenyl ring is also stacked under the aromatic side chain of TRPꢀ1021, making an angle of 22° between
their planes and a distance of 4.3 Å between their closest heavy atoms. Due to these two aromatic residues
and other surrounding hydrophobic residues, the phenyl moiety rests in a hydrophobic pocket.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
ACS Paragon Plus Environment

Page 23 of 41
Journal of Chemical Information and Modeling
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
C
Figure 10: Docking poses of SKF83566 stereoisomers as a result of flexible docking: (S )ꢀSKF83566
C
N
C
N
C
(
(
topꢀleft), (S S )ꢀSKF83566 (topꢀmiddle), (S R )ꢀSKF83566 (topꢀright), (R )ꢀSKF83566 (bottomꢀleft),
C
N
C N
R S )ꢀSKF83566 (bottomꢀmiddle) and (R R )ꢀSKF83566 (bottomꢀright). Similar conformations are
ACS Paragon Plus Environment

Journal of Chemical Information and Modeling
Page 24 of 41
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
shown in the same color.
In the other conformation shown in orange color, there is a potential hydrogen bond between the
NH group of the ligand and the carboxylic acid group of ASPꢀ1019. The planes of the phenyl ring and
–
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
PHEꢀ299 are nearly parallel at an angle of approximately 30° and with the distance of 4.4 Å between their
closest heavy atoms. The TRPꢀ1021 aromatic ring is perpendicular to the phenyl group. The hydroxyl
group of the ligand is not involved in any particular proteinꢀligand interaction, while the bromine atom is
pointing in the direction of –NH group of ARGꢀ1030. Like the conformations shown in green color, the
phenyl group was located in the same hydrophobic pocket.
Table 2: Free energy analysis using MMGBSA for different forms of SKF83566 and their selected
poses.
Ligand
Pose (cf. Figure 10)
∆H (kcal/mol) T∆S (kcal/mol) ∆G (kcal/mol)
(±std err)
(±std err)
(±std err)
C
Green
Orange
Green
Orange
Orange
ꢀ25.84 ± 0.05
ꢀ4.27 ± 0.05
ꢀ19.58 ± 0.05
ꢀ31.08 ± 0.04
ꢀ21.47 ± 0.04
ꢀ16.69 ± 0.96
ꢀ16.39 ± 1.18
ꢀ19.23 ± 1.25
ꢀ18.89 ± 1.08
ꢀ17.97 ± 2.54
ꢀ9.15 ± 0.96
12.12 ± 1.18
ꢀ0.35 ± 1.25
ꢀ12.20 ± 1.08
ꢀ3.50 ± 2.54
S –SKF83566
C
R –SKF83566
C
N
S S –SKF83566
C
N
Green
Orange
Orange
Green
ꢀ17.87 ± 0.05
ꢀ21.71 ± 0.09
ꢀ13.53 ± 0.04
ꢀ18.31 ± 0.05
ꢀ11.23 ± 0.04
ꢀ17.13 ± 2.19
ꢀ19.99 ± 2.84
ꢀ17.55 ± 0.97
ꢀ17.99 ± 1.21
ꢀ14.90 ± 0.98
ꢀ0.74 ± 2.19
ꢀ1.72 ± 2.84
4.03 ± 0.97
ꢀ0.32 ± 1.21
3.67 ± 0.98
S R –SKF83566
C
N
R S –SKF83566
C
N
Green
R R –SKF83566
MM/GBSA energy calculation
To determine which stereoisomer and protonation state of SKF83566 has the highest binding affinity,
and which pose from the flexible docking simulations is the most likely pose, we performed MD
simulations and evaluated the free energy of binding for different ligand poses using the MM/GBSA
method. We present the free energy calculations of all stereoisomers and conformations first, and then
describe the results of the MD simulation of only the most energetically favorable pose in the subsequent
section. The enthalpic and entropic contribution, and predicted free energies of binding are tabulated in
ACS Paragon Plus Environment

Page 25 of 41
Journal of Chemical Information and Modeling
1
2
3
4
5
6
7
8
9
C
C
Table 2. Among the set of stereoisomers, the neutral forms of (S /R )ꢀSKF3566 are more energetically
C
C
N
N
favorable than the simulated poses of protonated (S /R ,S /R )ꢀSKF83566.
C
N
N
Protonated, (S ,S /R )ꢀSKF83566, poses generally are energetically more favorable than protonated
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
C
N
N
C
(
R ,S /R )ꢀSKF83566. The energetically favorable neutral form of (S )ꢀSKF83566 has a predicted free
C
energy of binding of ꢀ9.15 kcal/mol and that of (R )ꢀSKF83566 has a predicted binding free energy of ꢀ
12.2 kcal/mol.
C
Figure 11: Conformational transition of (S )ꢀSKF83566 during MD simulation from blue to red. The
C
starting pose of (S )ꢀSKF83566 (in Figure10, green) undergoes conformational changes and finally
stabilizes in the pose shown in thick red sticks.
The two enantiomers reside in different pockets. Interestingly, each enantiomer’s less favorable pose
C
coincides with the other’s favorite binding pose, e.g. the less favorable pose of (S )ꢀSKF83566 occupies
C
the same region as the energetically favorable form of (R )ꢀSKF83566 and vice versa. In the next section,
C
C
we will discuss the binding conformation of the most favorable poses of (S )ꢀSKF83566 and (R )ꢀ
C
SKF83566 and lesser favorable pose of (R )ꢀSKF83566.
MD simulations
ACS Paragon Plus Environment

Journal of Chemical Information and Modeling
Page 26 of 41
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
In general, we observed that the overall structural stability of the SKF83566 stereoisomers is
correlated to their MM/GBSA free energy values. In this section, the structural details of the MD
C
C
simulation results of (S )ꢀSKF83566 and (R )ꢀSKF83566 are described.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
C
(
S )-SKF83566: For MD, the starting conformation of the energetically most favorable pose corresponds
to the flexible docking pose shown in green color in the ATP binding site in Figure 10 (top-left). During
the MD simulation, no significant change in the conformation of the starting pose was observed for the
first 20ns.
In the subsequent 10ns, the configuration of the ligand in the binding site transitioned into another, stable
conformation. This final conformation differs from the initial docking pose by an RMSD of about 4Å
(
Figure 11).
In the final pose in Figure 12, the methylamino group extended into a small hydrophobic pocket formed
by residues PHEꢀ299, LEUꢀ315, ILEꢀ296 and PHEꢀ319.
ACS Paragon Plus Environment

Page 27 of 41
Journal of Chemical Information and Modeling
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
C
Figure 12: Most stable pose of (S )ꢀSKF83566 (shown in red). The positions of ATP and forskolin (FSK),
derived from the Xꢀ ray structure of the template, are shown as gray sticks. The final positions of cations
are shown as spheres against the surface representation of AC2. Residues that directly interact with the
ligand are shown as gray sticks (PHEꢀ299, LEUꢀ315, TRPꢀ1021, GLYꢀ1022 and ASNꢀ1026).
The phenyl group of the ligand rests in an open hydrophobic pocket that is fairly exposed to the solvent.
The hydroxyl group of the benzazepine ring forms a hydrogen bond with the backbone –NH of GLYꢀ
1
022. The phenyl group of the ligand is positioned perpendicular to the benzazepine moiety of the ligand
and the aromatic side chains of PHEꢀ299 and TRPꢀ1021 in Tꢀshape like aromatic interactions.
ACS Paragon Plus Environment

Journal of Chemical Information and Modeling
Page 28 of 41
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
The Br has a dual electrostatic nature and is involved in strong electrostatic interactions with the different
electrostatic components of the –CONH side chain of ASNꢀ1026. This pose, no longer overlaps with the
7
conformation of ATP in the Xꢀray structure of the homologous AC and binds in a site adjacent to ATP.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
C
Figure 13: Most stable pose of (R )ꢀSKF83566 (shown in green). The positions of ATP and forskolin
C
(FSK), derived from the Xꢀ ray structure of the template, are shown as gray sticks. As seen, (R )ꢀ
SKF83566 is directly overlapping with the ATP. The final positions of cations are shown as spheres
against the surface representation of AC2. Residues that directly interact with the ligand (ILEꢀ938, ASPꢀ
1019,ILEꢀ1020, TRPꢀ1021, GLYꢀ1022 and ASNꢀ1026
C
(
R )-SKF83566: For MD, the starting conformation of the energetically most favorable pose corresponds
to the flexible docking pose shown in orange color in the ATP binding site in Figure 10 (bottom-left).
ACS Paragon Plus Environment

Page 29 of 41
Journal of Chemical Information and Modeling
1
2
3
4
5
6
7
8
9
C
During the MD simulation, the (R )ꢀSKF83566 marginally deviates from the starting position. Finally, it
forms two hydrogen bonds through the hydroxyl group: one with the carboxyl group of ASPꢀ1019 and
one with the backbone –NH of ILEꢀ1020. The closest aromatic amino acid residue facing the phenyl
moiety of SKF83566 is TRPꢀ1021 in a displaced Tꢀshape orientation. The methylamino group is solventꢀ
exposed and the Br atom is buried in the interior of hAC2 as shown in Figure 13. This pose significantly
overlaps with ATP in the Xꢀray structure, which may suggest binding in a competitive manner.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
C
C
(
R )-SKF83566: This second pose of (R )ꢀSKF83566, which is barely energetically favorable (ꢀ0.35
C
kcal/mol), is close to the (S )ꢀSKF83566 site at the end of the simulation and does not overlap with the
ATP conformation. The starting conformation of this simulation is displayed in green in Figure 10
(
bottom-left). The final pose indicates that the bromine group is pointing towards ASPꢀ295 and ASPꢀ
339, and the hydroxyl group forms hydrogen bonds with the solvent. Hydrophobic and aromatic residues,
PHEꢀ299, LEUꢀ315 and TRPꢀ1021, surround the phenyl moiety. The amine group is exposed to the
solvent to form a hydrogen bond between –NH and water.
Pharmacological assessment of the resolved SKF83566 enantiomers
We previously used a limited AC selectivity profile to identify racemic SKF83566 as the most AC2ꢀ
4
selective compound in the cellꢀbased assays when compared to AC1 and AC5. A cellꢀfree, reconstituted
system was also used to assess the direct modulation of AC isoforms (i.e. AC1, AC2, and AC5) revealing
that SKF83566 inhibited forskolinꢀstimulated AC2 activity by more than 40%, consistent with a direct
mode of inhibition.
The mechanism for SKF83566 inhibition of AC2 was also explored by determining inhibition under
varying concentrations of ATP and SKF83566. These studies revealed that SKF83566 was behaving as a
4
noncompetitive AC2 inhibitor .
ACS Paragon Plus Environment