Optimization of a Series of Mu Opioid Receptor (MOR) Agonists with High G Protein Signaling Bias

Article abstract of DOI:10.1021/acs.jmedchem.8b01136

While mu opioid receptor (MOR) agonists are especially effective as broad-spectrum pain relievers, it has been exceptionally difficult to achieve a clear separation of analgesia from many problematic side effects. Recently, many groups have sought MOR agonists that induce minimal βarrestin-mediated signaling because MOR agonist-treated βarrestin2 knockout mice were found to display enhanced antinociceptive effects with significantly less respiratory depression and tachyphylaxis. Substantial data now exists to support the premise that G protein signaling biased MOR agonists can be effective analgesic agents. We recently showed that, within a chemical series, the degree of bias correlates linearly with the magnitude of the respiratory safety index. Herein we describe the synthesis and optimization of piperidine benzimidazolone MOR agonists that together display a wide range of bias (G/βarr2). We identify structural features affecting potency and maximizing bias and show that many compounds have desirable properties, such as long half-lives and high brain penetration.

Full text of DOI:10.1021/acs.jmedchem.8b01136

Subscriber access provided by University of South Dakota
Article
Optimization of a Series of Mu Opioid Receptor
(MOR) Agonists with High G Protein Signaling Bias
Nicole M. Kennedy, Cullen L. Schmid, Nicolette C. Ross, Kimberley M. Lovell, Zhizhou
Yue, Yen Ting Chen, Michael D Cameron, Laura M. Bohn, and Thomas D. Bannister
J. Med. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.jmedchem.8b01136 • Publication Date (Web): 10 Sep 2018
Downloaded from http://pubs.acs.org on September 11, 2018
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 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 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.
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 35
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Optimization of a Series of Mu Opioid Receptor
(MOR) Agonists with High G Protein Signaling Bias
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
Nicole M. Kennedy,1,2,‡ Cullen L. Schmid,2,‡ Nicolette C. Ross,1,2 Kimberly M. Lovell,1,2 Zhizhou
Yue,1,2 Yen Ting Chen,1,2 Michael D. Cameron,2 Laura M. Bohn,2* Thomas D. Bannister1,2*
1 Department of Chemistry, The Scripps Research Institute, Jupiter, FL 33458, USA
2 Department of Molecular Medicine, The Scripps Research Institute, FL 33458, USA
KEYWORDS: Benzimidazolone, Bias, Mu Opioid Receptor, G Protein Coupled Receptor,
βarrestin2.
ABSTRACT: While mu opioid receptor (MOR) agonists are especially effective as broadꢀ
spectrum pain relievers, it has been exceptionally difficult to achieve a clear separation of
analgesia from many problematic side effects. Recently, many groups have sought MOR
agonists that induce minimal βarrestinꢀmediated signaling, since MOR agonistꢀtreated βarrestin2
knockout mice were found to display enhanced antinociceptive effects with significantly less
respiratory depression and tachyphylaxis. Substantial data now exists to support the premise that
G protein signaling biased MOR agonists can be effective analgesic agents. We recently showed
ACS Paragon Plus Environment
1

Journal of Medicinal Chemistry
Page 2 of 35
1
2
3
4
5
6
7
8
9
that, within a chemical series, the degree of bias correlates linearly with the magnitude of the
respiratory safety index. Herein we describe the synthesis and optimization of piperidine
benzimidazolone MOR agonists that together display a wide range of bias (G/βarr2). We identify
structural features affecting potency and maximizing bias and show that many have desirable
properties, such as long halfꢀlives and high brain penetration.
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
INTRODUCTION
Agonists of the mu opioid receptor (MOR), such as morphine and fentanyl, are extensively
used for the treatment of moderate to severe pain due to their high efficacy;1 however, the dose
required to achieve adequate pain relief often elicits multiple unwanted side effects, including
respiratory suppression, constipation, and tolerance. The respiratory suppressive effect of opioids
is especially concerning, as it is the root cause of death by opioid overdose, which claimed more
than 40,000 victims in the USA in 2016.2 The identification and development of safer analgesic
agents may play an important role in combatting the opioid epidemic.
The most effective opioid pain relievers are agonists of the MOR.3 As with all G protein
coupled receptors (GPCRs), agonist binding to MOR initiates the dissociation of heterotrimeric
G protein subunits and the activation of subsequent downstream signaling. The MOR also
interacts with βarrestins, scaffolding and regulatory proteins with multifaceted roles, including
receptor desensitization of the G protein cascades and the facilitation of receptor internalization
and signaling that can be distinct from G protein mediated responses.4 Studies using βarrestin2
knockout (βarr2ꢀKO) mice strongly suggest that the interaction between MOR and βarrestin2
produces many of morphine’s undesirable effects in vivo.5,6,7,8,9,10 In the βarr2ꢀKO mice,
morphine retained its analgesic properties, yet, in comparison to wildꢀtype littermates,
ACS Paragon Plus Environment
2

Page 3 of 35
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
constipation, tolerance, and respiratory suppression were largely attenuated.5,6,7,8,10 Based upon
this work, it was hypothesized that a MOR agonist capable of activating G protein signaling
without prompting the engagement of βarrestins might separate analgesia from many of the
adverse effects that arise downstream of MOR activation.
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
The ability of a compound to stimulate one signaling pathway over another upon engaging the
receptor is referred to “functional selectivity” or “biased agonism.”4,11,12,13,14 Compounds that
preferentially activate G protein signaling over βarrestin2 recruitment in the MOR have been
developed by a number of groups.15,16,17,18,19 Two reported clinical trials with TRV130 (aka
Oliceridine) demonstrate that one such G protein biased MOR agonist has analgesic efficacy
with a modest improvement in respiratory suppression compared to morphine.20,21 These clinical
results suggest that imparting G coupling bias in MOR agonists may lead to the discovery and
eventual availability of safer opioid analgesic agents, especially if the magnitude of signaling
bias is augmented so that more than a modest improvement in therapeutic index may be
achieved.
We recently disclosed members of a family of substituted piperidine benzimidazolone MOR
agonists with high affinity for the MOR (0.3ꢀ14 nM), high G protein signaling bias (up to 100
fold bias relative to DAMGO in some assays), and high selectivity for MOR over the other
opioid receptors.22 The modular design of the core scaffold permitted extensive structural
diversification, an important advantage since we found that only a small subset of potential
analogs in the series display significant G protein signaling bias. We have optimized these MOR
agonists for potency, G protein signaling bias, and drugꢀlike properties. Our objective has been
to identify compounds with morphineꢀlike efficacy in vivo (or better) that would not induce
respiratory suppression at, or even well above, an efficacious dose. Within a set of related
ACS Paragon Plus Environment
3

Journal of Medicinal Chemistry
Page 4 of 35
1
2
3
4
5
6
7
8
9
compounds, we achieved this goal and demonstrated that the degree of signaling bias within the
series linearly correlated to the magnitude of protection from respiratory suppression.22 Given
that slight structural changes can greatly impact signaling bias, understanding the structural
features favoring alternative MOR signaling pathways could greatly enhance efforts to identify
new safe and efficacious analgesic agents.
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
Herein, we report the first comprehensive biasꢀfocused structureꢀactivity relationship (SAR)
study for this class of MOR agonists. We systematically varied the R1ꢀR8 substituents of the
generalized Nꢀbenzyl cycloamino benzimidazolone scaffold as well as varied the ring size in the
central saturated ring, which in most cases is a piperidine (Figure 1). Many structural changes
were found to markedly impact the properties of MOR agonists. The iterative design and
evaluation of analogs drove the optimization of G protein signaling potency, the deselection of
βarrestin interactions, and the optimization of desirable drugꢀlike properties, such as a suitable
halfꢀlife, lack of cytochrome P450 (CYP) inhibition, and high bloodꢀbrain barrier (BBB)
permeability. Signaling bias was characterized using separate cellꢀbased assays designed to
measure G protein coupling and βarrestin2 recruitment. Bias was quantitated by applying the
operational model,23 which measures the affinity, potency, and efficacy of the compound in each
assay relative to the reference full agonist [DꢀAla2, NMeꢀPhe4, Glyꢀol5]ꢀenkephalin (DAMGO)
and then allows for comparison across assays.11,23 For optimized biased analogs, we measured
various in vitro parameters, such as liver microsome stability and CYP inhibition, then followed
up with in vivo studies for metabolic stability and BBB permeability. Results from these in vitro
and in vivo studies advance our understanding of the chemical properties that underlie MOR
signaling bias and may aid the development of additional promising leads separating analgesic
activity from unwanted, and potentially deadly, respiratory suppression.
ACS Paragon Plus Environment
4

Page 5 of 35
Journal of Medicinal Chemistry
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 1. General structure of Nꢀbenzyl piperidine 4ꢀbenzimidazolones. The sites for structure diversification R1ꢀR8
and n are highlighted by the colored circles.
RESULTS AND DISCUSSION
Synthesis. As shown in Scheme 1, a general fiveꢀstep synthetic route was used to synthesize
the majority of the benzimidazolone derivatives24(compounds with a central piperidine ring are
shown; see Scheme S1, Supporting Information for an alternate route). In the general procedure,
nucleophilic aromatic substitution of a fluoronitrobenzene starting material with NꢀBocꢀ4ꢀ
aminopiperidine (or in some cases with a homolog having a 5ꢀ or 7ꢀmembered ring) was
followed by nitro group reduction of I, cyclic urea formation of II, Boc deprotection of III, and
finally, either direct alkylation or reductive amination of IV produced the desired analogs.
Reductive amination used an aldehyde (R5 = H) or ketone (R5 = Me or Et). The products were
isolated and characterized in free base form, unless indicated, and then were evaluated in all
biological and pharmacological assays as either the free base or as the mono mesylate salt.
ACS Paragon Plus Environment
5

Journal of Medicinal Chemistry
Page 6 of 35
1
2
3
4
Scheme 1. General FiveꢀStep Synthetic Route for the Preparation of Substituted Piperidine 4ꢀ
5
6
Benzimidazolones.
7
8
9
R1
R1
NH2
R1
R2
R3
NO2
NH
R2
R3
NH2
NH
H
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
R2
R3
R1
O
Raney Ni
EtOH
N
N
R2
R3
NO2
F
CDI
Boc
R4
R4
R4
THF
N2H4 H2O
K2CO3
DMSO
N
N
N
R4
Boc
Boc
Boc
I
II
III
R1
H
R2
R3
Br
N
O
R5
+ Base (alkylation)
R1
N
H
N
R6
O
R8
R2
R3
R4
- or -
TFA
R7
O
N
N
CH2Cl2
R4
R5
R8
R5
+ NaBH(OAc)3 or
NaBH3CN
R6
N
R7
R6
R8
H
(reductive amination)
R7
1, R1-R8 = H
IV
Structure Activity Relationships. In these SAR studies, substituents R1ꢀR8 were varied, as
was the size of the central ring (Figure 1). Each analog was screened for its ability to activate G
protein coupling at MOR using the standard 35SꢀGTPγS binding assay in membranes from CHOꢀ
K1 cells expressing the human MOR. βarrestin2 recruitment profiles to the human MOR were
determined using a cellꢀbased commercially available enzymeꢀfragment complementation assay
(EFC). In both assays, full dose response curves for the test compound were run in parallel to the
reference full agonist, DAMGO, as previously described.22 To quantitatively compare the
differences observed between the two signaling assays, the operational model was used to
calculate ꢁꢁlog(τ/KA) values, with confidence intervals and bias factors (the antiꢀlog of
ACS Paragon Plus Environment
6

Page 7 of 35
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
ꢁꢁlog(τ/KA)).22,23 A bias factor of 1 (ꢁꢁlog(τ/KA) value of 0) indicates that the compound is
unbiased relative to DAMGO. A bias factor less than 1 (ꢁꢁlog(τ/KA) value less than 0) indicates
bias towards βarrestin2 recruitment over GTPγS binding, while a bias factor greater than 1
(ꢁꢁlog(τ/KA) value greater than 0) indicates bias towards GTPγS binding over βarrestin2
recruitment. Additionally, we used morphine as a clinically relevant opioid analgesic for
comparison purposes. In these assays, we found morphine, with a bias factor of 1.8, to show a
slight bias towards G protein coupling over βarrestin2 recruitment compared to DAMGO. Safer
analgesic agents would be expected to have significantly higher levels of G protein coupling
bias.22
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 shown in Table 1, the parent Nꢀbenzyl piperidine 4ꢀbenzimidazolone (1), with all
substituents R1ꢀR8 = H, had modest potency in the GTPγS assay (EC50 = 2.2 µM) but was a full
agonist (EMAX = 78% of DAMGO). Compound 1 performed similar to DAMGO in both assays
(GTPγS/βarr2 bias factor = 0.9). After exploring the impact of various R6ꢀR8 substituents on the
Nꢀbenzyl ring, we found that the addition of an orthoꢀCl (2) or paraꢀCl substituent (4)22
improved potency. The orthoꢀCl substituent also imparted a modest G protein signaling bias
(GTPγS/βarr2 bias factor = 2.5).
Many paraꢀsubstituted analogs (4ꢀ7) were similar in potency and efficacy in the GTPγS assay
(EC50 = 367ꢀ591 nM, EMAX = 68ꢀ88%) and were similarly unbiased (GTPγS/βarr2 bias factors =
0.9ꢀ1.2). Paraꢀsubstituted analogs (8ꢀ10) were less potent and/or efficacious in the GTPγS assay
compared to 4ꢀ7, with 10 showing no significant activity until the 10 µM concentration. The
metaꢀsubstituted compounds 3, 11, and 12 were also less potent and less efficacious in the
GTPγS assay in comparison to their paraꢀsubstituted counterparts, with compound 11 acting as a
partial GTPγS agonist (EMAX = 33% of DAMGO). When a second Cl substituent was added to
ACS Paragon Plus Environment
7

Journal of Medicinal Chemistry
Page 8 of 35
1
2
3
4
5
6
7
8
9
the para position (13), the GTPγS potency and efficacy was reduced further and the potency at
βarrestin2 was outside the experimental range of the assay. Thus, ortho and para substituted Nꢀ
benzyl analogs were of greatest interest in this series from the perspective of obtaining biased
agonists of useful potency. The potential for combining the ortho and para effects was
demonstrated with compounds 14 and 15, in which 14 is a potent full agonist showing some bias
for G protein signaling (EC50 = 152 nM, EMAX = 93%, GTPγS/βarr2 bias factor = 3.8).
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
Table 1. GTPγS Binding and βarrestin2 Recruitment Activity of NꢀBenzyl Piperidine 4ꢀ
Benzimidazolone Derivatives with NꢀBenzyl Substituents R6ꢀR8.a
βarrestin2 Recruitmentb
35SꢀGTPγS Bindingb
Bias (GTPγS/βarr2)
EMAX
ꢁꢁlog(τ/KA)
Bias
Factor
Agonist
R6
R7
R8
EC50 (nM)
ꢁlog(τ/KA) EC50 (nM)c EMAX (%)c ꢁlog(τ/KA)
(%)c
(95% Cl)d
DAMGO
Morphine
33 ± 1
64 ± 3
100
220 ± 8.3
100
0
1.0
1.8
83 ± 1 ꢀ0.3 ± 0.03 379 ± 17
24 ± 1 ꢀ0.6 ± 0.03 0.25 (0.17 to 0.33)e
1
78 ± 3 ꢀ1.8 ± 0.06 8255 ± 1676 40 ± 2 ꢀ1.8 ± 0.07 ꢀ0.04 (ꢀ0.25 to 0.16)
0.9
2183 ± 290
200 ± 19
2
3
Cl
92 ± 2 ꢀ0.8 ± 0.07 4487 ± 606
1354 ± 195 71 ± 1 ꢀ2 ± 0.06
>10,000 c
96 ± 9 ꢀ1.2 ± 0.05 0.4 (0.21 to 0.59)f
10 ± 0.4c ꢀ2.0 ± 0.09 0.22 (ꢀ0.08 to 0.52)
2.5
1.6
1.2
1.0
1.1
Cl
4h
5
Cl
Me
551 ± 58
433 ± 47
367 ± 37
591 ± 49
68 ± 4 ꢀ1.4 ± 0.09 7656 ± 1469 67 ± 10 ꢀ1.5 ± 0.08 0.07 (ꢀ0.25 to 0.39)
86 ± 2 ꢀ1.0 ± 0.06 2400 ± 497 63 ± 3 ꢀ1.1 ± 0.08 0.01 (ꢀ0.27 to 0.29)
ꢀ1.2 ± 0.1 6611 ± 1076 77 ± 6 ꢀ1.3 ± 0.06 0.05 (ꢀ0.22 to 0.31)
6
Br
76 ± 4
7
OMe
88 ± 3 ꢀ1.2 ± 0.05 2254 ± 303
41 ± 3 ꢀ1.2 ± 0.01 ꢀ0.04 (ꢀ0.45 to 0.36) 0.91
ACS Paragon Plus Environment
8

Page 9 of 35
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
8
NHC(O)Me 760 ± 75
89 ± 6 ꢀ1.4 ± 0.04 3435 ± 889
29 ± 3
1 ± 0.5c
0.03 ± 0.1c
8 ± 1c
ꢀ1.4 ± 0.1 0.06 (ꢀ0.36 to 0.47)
1.1
/
9
CN
3424 ± 693 48 ± 2
NC
7.7 ± 1c
ꢀ2.2 ± 0.1
NC c
NC c
/
/
/
/
10
11
12
13
14
15
C(O)NH2
/
/
OMe
OCF3
Cl
1158 ± 160 33 ± 1 ꢀ1.9 ± 0.02 >10,000 c
ꢀ2.2 ± 0.2 0.30 (ꢀ0.03 to 0.62)
2.0
1.1
/
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
1621 ± 183 64 ± 4 ꢀ1.8 ± 0.05 9326 ± 1738 29 ± 4 ꢀ1.8 ± 0.09 0.03 (ꢀ0.19 to 0.24)
Cl
Cl
Br
6423 ± 565 27 ± 2
ꢀ2.8 ± 0.2
NC c
4 ± 3c
75 ± 5
/
/
Me
F
152 ± 10
110 ± 16
93 ± 1 ꢀ0.6 ± 0.03 2880 ± 411
96 ± 0.1 ꢀ0.7 ± 0.04 3357 ± 60
ꢀ1.2 ± 0.1 0.58 (0.30 to 0.87)g
3.8
2.5
94 ± 6 ꢀ1.1 ± 0.04 0.39 (0.22 to 0.56)g
aIf blank or not shown, Rx = H. “/” = not determined. Within each assay, potency (EC50) and efficacy (EMAX
values were calculated by nonꢀlinear regression and ꢁlog(τ/KA) values were determined by the operational model,
relative to DAMGO. Data are presented as mean ± SEM of three or more assays run in duplicate or triplicate. In
instances where the data did not converge to a nonlinear regression curve fitting due to not reaching plateau at the
highest concentrations tested (10ꢀ4.5M in the βarrestin assasy, 10ꢀ5M in the G protein assay) the EC50 is noted as
>10,000 nM (if some stimulation could be observed) or NC (if the maximum concentration produced no effect) and
the percentage of maximum stimulation at 10,000 nM is provided rather than an EMAX value where indicated. For
ꢁꢁlog(τ/KA) values, 95% confidence intervals (95% CI) are provided. Significant bias compared to DAMGO was
)
b
c
d
e
f
g
h
determined by an unpaired, twoꢀtailed t test: p<0.0001; p<0.001;; p<0.01, otherwise p>0.05. Compound 4 was
22
previously reported as SRꢀ20382
As shown in Table 2, adding an Nꢀbenzylic methyl group (R5 = Me) substantially increased
potency in the parent racemic compound (16, EC50 = 102 nM)22 as compared to 2, but had little
effect on GTPγS/βarr2 bias. Analogous to the Table 1 series, the addition of an orthoꢀ or paraꢀCl
substituent further augmented potency (EC50 = 31 and 16 nM, respectively). In this case, the
paraꢀCl compound 1822 was more biased than was the orthoꢀCl compound 17. The boost in
GTPγS potency and efficacy eroded when an ethyl rather than a methyl group was present (19).
Replacement of the paraꢀCl substituent with either paraꢀF substituent (20) or paraꢀOMe group
(22) gave potent compounds but with reduced bias; however, replacement with either a paraꢀBr
substituent (21) or paraꢀOCF3 group (26), increased GTPγS/βarr2 bias to 4.1 and 8.3,
respectively. Increasing the size of the alkoxy group to ethoxy (23), isopropoxy (24), or ethylene
ACS Paragon Plus Environment
9

Journal of Medicinal Chemistry
Page 10 of 35
1
2
3
4
5
6
7
8
9
dioxy (25)22 eroded potency in the GTPγS assay; however, the GTPγS/βarr2 bias factor of 24
was modestly increased to 3.6. Compounds 20 and 25, interestingly, promoted bias towards
βarrestin2 recruitment, opposite of what is expected to provide analgesia with reduced side
effects (GTPγS/βarr2 bias factors = 0.56 and 0.47, respectively). We previously observed
βarrestin2 bias for fentanyl vs. DAMGO at the human MOR when comparing GTPγS binding
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
and the βarrestin2 recruitment using the enzyme fragment complementation assays (as used in
this current study) (GTPγS/βarr2 bias factor = 0.18, measured relative to DAMGO),22 and drew
correlations with a reduced therapeutic index comparing mouse hot plate responses to changes in
arterial oxygen saturation. Compound 25, aka SRꢀ11501, shared similar profiles with fentanyl in
both the in vivo and in vitro assays22; and thus, it is attractive to speculate that preference for
βarrestin2 recruitment over G protein coupling in vitro will consistently indicate a very narrow
respiratory safety index in vivo. Compounds, such as 20 and additional compounds across a
range of bias factors, will allow further testing of this hypothesis. Finally, compounds 27 and 28
showed that the addition of an orthoꢀF substituent had little effect on the potency, efficacy, and
bias of compounds having paraꢀCl (18) and paraꢀBr (21) substituents.
Table 2. GTPγS Binding and βarrestin2 Recruitment Activity of NꢀBenzyl Piperidine 4ꢀ
Benzimidazolone Derivatives with NꢀBenzyl Substituents R5ꢀ R8.a
βarrestin2 Recruitmentb
35SꢀGTPγS Bindingb
Bias (GTPγS/βarr2)
Agonist
R5
R6
R7
R8
EC50
EMAX
ꢁlog(τ/KA)
EC50
EMAX
ꢁlog(τ/KA)
ꢁꢁlog(τ/KA)
Bias
ACS Paragon Plus Environment
10

Page 11 of 35
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
(nM)
33 ± 1
(%)
(nM)
(%)
100
(95% Cl)c
0
Factor
1.0
DAMGO
Morphine
16e
100
220 ± 8.3
64 ± 3
83 ± 1 ꢀ0.3 ± 0.03 379 ± 17
89 ± 7 ꢀ0.5 ± 0.04 447 ± 52
96 ± 1 0.05 ± 0.05 199 ± 19
24 ± 1
67 ± 4
ꢀ0.6 ± 0.03 0.25 (0.17 to 0.33)d
ꢀ0.3 ± 0.07 ꢀ0.18 (ꢀ0.45 to 0.08)
1.8
Me
Me
Me
102 ± 10
31 ± 0.21
16 ± 1.2
0.7
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
17
Cl
79 ± 15 0.1 ± 0.06 ꢀ0.01 (ꢀ0.2 to 0.18)
95 ± 4 0.02 ± 0.06 0.32 (0.09 to 0.56)f
1.0
18e
Cl
97 ± 7
0.3 ± 0.07 184 ± 24
6050 ±
2.1
19
Et
Cl 1593 ± 194 58 ± 1 ꢀ1.9 ± 0.08
32 ± 4
ꢀ2 ± 0.08 ꢀ0.09 (ꢀ0.36 to 0.17) 0.81
ꢀ0.2 ± 0.1 ꢀ0.26 (ꢀ0.53 to 0.02) 0.56
1233
20
21
22
23
24
25e
26
27
28
Me
Me
Me
Me
Me
Me
Me
Me
Me
F
75 ± 8.3
93 ± 2 ꢀ0.4 ± 0.05 377 ± 76
78 ± 5
92 ± 5
75 ± 8
Br
4.8 ± 0.41
91 ± 2
95 ± 2
0.8 ± 0.02 182 ± 42
0.4 ± 0.03 87 ± 18
0.2 ± 0.1 0.61 (0.11 to 1.11)f
4.1
OMe 13 ± 0.64
OEt 189 ± 25
0.4 ± 0.1 ꢀ0.06 (ꢀ0.50 to 0.37) 0.86
ꢀ0.6 ± 0.1 ꢀ0.04 (ꢀ0.33 to 0.24) 0.90
94 ± 2 ꢀ0.7 ± 0.06 1074 ± 165 50 ± 3
83 ± 2 ꢀ1.1 ± 0.06 5413 ± 373 33 ± 1
OiPr 399 ± 31
ꢀ1.6 ± 0.1 0.56 (0.22 to 0.90)f
3.6
59 ± 2.2 ꢀ0.2 ± 0.06 ꢀ0.32 (ꢀ0.45 to ꢀ0.2)d 0.47
ꢀOCH2CH2Oꢀ 98 ± 8.5
OCF3 31 ± 4.7
77 ± 3.9 ꢀ0.5 ± 0.02 407 ± 50
96 ± 1
93 ± 2
98 ± 3
0.1 ± 0.08 1678 ± 328 100 ± 3 ꢀ0.8 ± 0.06 0.92 (0.64 to 1.20)d
8.3
3.3
2.9
F
F
Br
Cl
18 ± 2.4
22 ± 2.1
0.2 ± 0.04 523 ± 67
0.2 ± 0.04 484 ± 94
82 ± 7
82 ± 6
ꢀ0.4 ± 0.05 0.52 (0.39 to 0.66)d
ꢀ0.3 ± 0.09 0.46 (0.10 to 0.82)f
aIf blank or not shown, Rx = H. All compounds were made and tested as racemates. Within each assay, potency
(EC50) and efficacy (EMAX) values were calculated by nonꢀlinear regression and ꢁlog(τ/KA) values were determined
by the operational model, relative to DAMGO. Data are presented as mean ± SEM of three or more assays run in
duplicate or triplicate. For ꢁꢁlog(τ/KA) values, 95% confidence intervals (95% CI) are provided. Significant bias
compared to DAMGO was determined by an unpaired, twoꢀtailed t test: p<0.0001; Compounds 16, 18, and 25
b
c
d
e
were previously reported as (±)SRꢀ8595, (±)SRꢀ11065, and (±)SRꢀ11501, respectively.22
We then investigated the effect of varying benzimidazolone substituents R1ꢀR4 (see Table 3).
In this study, for consistency in interpretation, we held the distal Nꢀbenzyl substituent constant as
an orthoꢀCl (R6 = Cl). Compounds with a Cl at R1 (29), R2 (30), or R3 (31) were full agonists
with increased GTPγS potency, GTPγS efficacy, and GTPγS/βarr2 bias relative to the desꢀCl
ACS Paragon Plus Environment
11

Journal of Medicinal Chemistry
Page 12 of 35
1
2
3
4
5
6
7
8
9
analog 2. Compound 32, with R4 = Cl, however, was far less active. The nearꢀcomplete lack of
potency for compounds 30 and 31 in the βarrestin2 recruitment assay (EC50 > 10 µM) was
encouraging and led us to explore whether groups other than Cl might also deꢀselect βarrestin2
recruitment. Thus, we studied compounds with various R2 substituents (33ꢀ39) and found that the
groups shown erode GTPγ potency and/or GTPγS/βarr2 bias, except for compound 33, R2 = Me,
which was similar in potency and efficacy in the GTPγ assay but was slightly less biased due to
the higher potency in the βarrestin2 recruitment assay (EC50 = 6.8 µM). We concluded that a
chloro substituent in the benzimidazolone ring system (R1, R2, or R3 = Cl) favors higher potency
and efficacy in the GTPγS assay while giving substantial bias (i.e., minimal βarrestin2
recruitment).
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
Table 3. GTPγS Binding and βarrestin2 Recruitment Activity of NꢀBenzyl Piperidine 4ꢀ
Benzimidazolone Derivatives with Benzimidazolone Substituents R1ꢀR4.a
35SꢀGTPγS Bindingb
βarrestin2 Recruitmentb
Bias (GTPγS/βarr2)
ꢁꢁlog(τ/KA)
(95% Cl)d
Bias
Factor
Agonist R1
R2
R3 R4 EC50 (nM) EMAX (%)c ꢁlog(τ/KA) EC50 (nM) EMAX (%)c ꢁlog(τ/KA)
DAMGO
Morphine
2
33 ± 1
64 ± 3
100
220 ± 8.3
ꢀ0.3 ± 0.03 379 ± 17
ꢀ0.8 ± 0.07 4487 ± 606
100
0
1.0
1.8
2.5
5.8
7.1
6.9
83 ± 1
92 ± 2
99 ± 1
93 ± 2
99 ± 2
24 ± 1
96 ± 9
ꢀ0.6 ± 0.03 0.25 (0.17 to 0.33)
ꢀ1.2 ± 0.05 0.4 (0.21 to 0.59)
200 ± 19
48 ± 6.7
184 ± 23
148 ± 16
29
30
31
Cl
ꢀ0.1 ± 0.09 1962 ± 326 69 ± 10 ꢀ0.9 ± 0.08 0.76 (0.51 to 1.01)
Cl
ꢀ0.7 ± 0.09 >10,000
ꢀ0.6 ± 0.08 >10,000
47 ± 4c
40 ± 4c
ꢀ1.6 ± 0.06 0.85 (0.60 to 1.11)
ꢀ1.5 ± 0.2 0.84 (0.46 to 1.22)
Cl
ACS Paragon Plus Environment
12

Page 13 of 35
Journal of Medicinal Chemistry
1
2
3
4
5
32
33
Cl 3109 ± 747 66 ± 16 ꢀ2.2 ± 0.07
NC
1 ± 1c
79 ± 4
/
/
/
6769 ±
1337
Me
224 ± 9.6
99 ± 2
ꢀ0.8 ± 0.03
ꢀ1.5 ± 0.05 0.65 (0.49 to 0.81)
4.5
6
7
8
9
34
35
36
37
38
39
Br
F
57 ± 5
97 ± 2
96 ± 1
ꢀ0.4 ± 0.04 843 ± 31
ꢀ0.7 ± 0.1 1644 ± 315
80 ± 3
78 ± 6
ꢀ0.5 ± 0.03 0.12 (ꢀ0.03 to 0.27) 1.3
ꢀ0.8 ± 0.09 0.11 (ꢀ0.26 to 0.47) 1.3
ꢀ2.2 ± 0.1 0.28 (ꢀ0.17 to 0.73) 1.9
111 ± 16
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
OMe
OCF3
SO2Me
CN
2123 ± 76
1236 ± 111
7618 ± 737
>10,000
103 ± 3 ꢀ1.9 ± 0.01 >10,000
18 ± 2c
3 ± 1c
57 ± 1
75 ± 7
17 ± 2c
ꢀ1.9 ± 0.03
NC
/
/
/
ꢀ2.5 ± 0.07 >10,000
6 ± 1c
ꢀ2.5 ± 0.04 0.01 (ꢀ0.15 to 0.17) 1.0
ꢀ2.8 ± 0.06
NC
ꢀ0.2 ± 0.2c
/
/
/
aIf blank or not shown, Rx = H. “/” = not determined. Within each assay, potency (EC50) and efficacy (EMAX
values were calculated by nonꢀlinear regression and ꢁlog(τ/KA) values were determined by the operational model,
relative to DAMGO. Data are presented as mean ± SEM of three or more assays run in duplicate or triplicate. In
instances where the data did not converge (NC) and therefore no potency value could be calculated, or produced a
value outside the experimental range (EC50 >10,000 nM), the percentage of maximum stimulation at 10,000 nM is
)
b
c
d
provided. For ꢁꢁlog(τ/KA) values, 95% confidence intervals (95% CI) are provided. Significant bias compared to
DAMGO was determined by an unpaired, twoꢀtailed t test: ep<0.0001; fp<0.001; gp<0.01, otherwise p>0.05.
To determine if the benzimidazolone chloro effect was compatible with substituents other than
orthoꢀCl on the distal Nꢀbenzyl ring, as well as to evaluate compatibility with a benzylic methyl
at R5 (which was expected to increase potency, as was shown in Table 2), another diverse series
of analogs was investigated (Table 4). This set of compounds contained, in the Nꢀbenzyl ring,
orthoꢀF and paraꢀBr substituents at R6 and R8, respectively, since this combination had given
advantages in potency and bias (see Table 2, compare 17 and 27).
Compared to compound 15, with R1ꢀR5 = H, the presence of Cl at R1 (40), R2 (41), or R3 (42)
had little effect on potency in the GTPγS assay, but bias was more pronounced, especially in
compound 41 (GTPγ/βarr2 bias factor = 23, Table 4). Potency in the GTPγS assay was greatly
augmented, as noted earlier (Table 2), with the addition of a benzylic methyl group at R5: see
ACS Paragon Plus Environment
13

Journal of Medicinal Chemistry
Page 14 of 35
1
2
3
4
5
6
7
8
9
compounds 15 and 27, while bias was relatively unchanged. Di and trichlorinated analogs were
also prepared. A Cl present at both R1 and R3 (43) had no advantage over the reference
compound 15 with regard to potency in G protein signaling or in bias disfavoring βarrestin2
recruitment. On the other hand, a Cl at both R2 and R3 (44), while similar in potency to
compound 15 in the GTPγS signaling assay (EC50 = 91 nM), was essentially devoid of all
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
βarrestin2 recruitment activity (EC50 > 10 µM, EMAX = 12% of DAMGO, GTPγ/βarr2 bias factor
= 56). As expected based upon earlier data for compound 32 (Table 3), a Cl at R4 is not tolerated;
thus, the trichloro compound 45 lost potency in the GTPγS assay.
Table 4. GTPγS Binding and βarrestin2 Recruitment Activity of NꢀBenzyl Piperidine 4ꢀ
Benzimidazolone Derivatives with Benzimidazolone Substituents R1ꢀR5.a
βarrestin2 Recruitmentc
35SꢀGTPγS Bindingc
Bias (GTPγS/βarr2)
ꢁꢁlog(τ/KA)
(95% Cl)e
Bias
Factor
Agonist R1 R2 R3 R4 R5 EC50 (nM) EMAX (%)d ꢁlog(τ/KA) EC50 (nM) EMAX (%)d ꢁlog(τ/KA)
DAMGO
33 ± 1
64 ± 3
100
220 ± 8.3
100
0
1.0
Morphine
83 ± 1 ꢀ0.3 ± 0.03 379 ± 17
24 ± 1 ꢀ0.6 ± 0.03 0.25 (0.17 to 0.33)f 1.8
94 ± 6 ꢀ1.1 ± 0.04 0.39 (0.22 to 0.56)g 2.5
82 ± 7 ꢀ0.4 ± 0.05 0.52 (0.39 to 0.66)f 3.3
15
27
40
41
42
43
110 ± 16 96 ± 0.06 ꢀ0.7 ± 0.04 3357 ± 60
Meb 18 ± 2.4
55 ± 8.1
93 ± 2
0.2 ± 0.04 523 ± 67
ꢀ0.5 ± 0.1 2912 ± 189 71 ± 6
14 ± 4d
88 ± 7 ꢀ0.6 ± 0.07 3459 ± 634 49 ± 9
91 ± 2 ꢀ0.9 ± 0.1 >10,000
66 ± 10d ꢀ1.3 ± 0.02 0.40 (0.10 to 0.69)h 2.5
Cl
Cl
93 ± 2
ꢀ1.1 ± 0.1 0.59 (0.21 to 0.98)h 3.9
ꢀ2.5 ± 0.2 1.36 (1.08 to 1.64)f
23
ꢀ1.3 ± 0.1 0.62 (0.31 to 0.94)g 4.2
Cl
355 ± 26
89 ± 2 ꢀ1.1 ± 0.04 >10,000
Cl
Cl
127 ± 11
153 ± 32
ACS Paragon Plus Environment
14

Page 15 of 35
Journal of Medicinal Chemistry
1
2
3
4
5
44
45
Cl Cl
Cl Cl
91 ± 5.5
74 ± 4 ꢀ0.6 ± 0.06 >10,000
12 ± 5d
ꢀ2.4 ± 0.1 1.75 (1.36 to 2.14)f
56
Cl
1764 ± 384 99 ± 3
ꢀ1.8 ± 0.1 >10,000
5.9 ± 3d ꢀ2.5 ± 0.06 0.71 (0.22 to 1.21)h 5.2
6
7
8
9
aIf blank or not shown, Rx = H. Compounds were made and tested as racemates. Within each assay, potency
(EC50) and efficacy (EMAX) values were calculated by nonꢀlinear regression and ꢁlog(τ/KA) values were determined
by the operational model, relative to DAMGO. Data are presented as mean ± SEM of three or more assays run in
b
c
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
d
duplicate or triplicate. In instances where the data produced a value outside the experimental range (EC50 >10,000
nM), the percentage of maximum stimulation at 10,000 nM is provided. For ꢁꢁlog(τ/KA) values, 95% confidence
intervals (95% CI) are provided. Significant bias compared to DAMGO was determined by an unpaired, twoꢀtailed t
e
test: fp<0.0001; gp<0.01; hp<0.05.
The potency and bias effects for the compound series in Table 4 are made even more apparent
when comparing full doseꢀresponse curves, as shown in Figure 2. Moving clockwise within the
Figure, panel A shows curves for the GTPγS binding assay (circles) and the βarrestin2
recruitment assay (squares) for the modestly biased reference compound 15 (green) compared
with DAMGO (white). Panel B shows the potency enhancement seen for 27, which has a
benzylic methyl group: note the leftꢀshifted curve (red circles). Panel C shows data for the monoꢀ
chloro substituted compound 41, which is less potent since it lacks a benzylic methyl group but
is still highly biased. Panel D shows data for compound 43, the less biased and less potent
dichloro benzimidazolone analog. Panel E shows the enhanced signaling bias of compound 44:
note the flattened βarrestin2 curve (blue squares). The rightꢀshifted curve (dark blue circles) in
panel F depicts the significantly less potent compound 45. It is worth noting that at the highest
concentrations tested, in some cases we see an increase in response; however, we are unable to
determine if the response would plateau, as solubility limitations preclude higher concentrations
in the assay. Therefore, in the tables, % maximum stimulation at 10,000 nM is provided when
potency cannot be calculated. Finally, panel G graphically compares the analogs and depicts the
ACS Paragon Plus Environment
15

Journal of Medicinal Chemistry
Page 16 of 35
1
2
3
4
substantial signaling bias seen for compounds 41 and 44 (GTPγ/βarr2 bias factors = 23 and 56,
5
6
respectively).
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 2. SAR increasing activation of G protein binding with differential βarrestin 2 signaling profiles. Four
analogs were synthesized to determine how the addition of chlorine substituents to the benzimidazolone ring affects
bias (41, 43ꢀ45). Analogs were compared to compound 15, which has an unsubstituted benzimidazolone ring.
ACS Paragon Plus Environment
16

Page 17 of 35
Journal of Medicinal Chemistry
1
2
3
4
5
6
Compound 27 was made to determine the effect that the addition of a benzylic methyl group had on bias compared
to 15. Concentration response curves in the GTPγS binding (circles) and βarrestin2ꢀEFC assays (squares) are shown
7
8
9
for the test compounds (solid symbols) versus DAMGO (open symbols) and are presented as mean ± S.E.M.; curves
are the result of 3 parameter nonlinear regression analysis. ꢁꢁLog(τ/KA) values are plotted to demonstrate relative
bias, with 95% confidence intervals. Significant bias compared to DAMGO was determined by an unpaired, twoꢀ
tailed t test: ****p<0.0001; **p<0.01; *p<0.05.
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
With the finding that compound 44, with R2 = R3 = Cl, has the greatest bias among Table 4
analogs, we then held this portion of the structure constant and also held the substitution pattern
in the distal Nꢀbenzyl ring constant, with a paraꢀBr substituent, which we had shown to confer
βarrestin2 bias (see compound 15 in Table 1 and compound 21 in Table 2). The additional orthoꢀ
F substituent in Table 4 analogs had only a small effect: compare compound 44 (Table 4) to
compound 46 (Table 5). With both ends of the molecule held constant, we probed the effect of
changing the size of the central ring, as well as altering where the benzimidazolone is attached
(Table 5). The 7ꢀmembered (47) and 5ꢀ membered ring analogs (48 and 49) were less potent in
the GTPγS assay and were less biased relative to the corresponding 4ꢀsubstituted piperidine (46,
EC50 = 149 nM and GTPγS/βarr2 bias factor = 45).22 Furthermore, the 3ꢀsubstituted piperidines
(50 and 51) were essentially inactive as MOR agonists, with no significant efficacy until tested at
the 10 µM concentration.
Table 5. GTPγS Binding and βarrestin2 Recruitment Activity of NꢀBenzyl Benzimidazolone
Derivatives.
ACS Paragon Plus Environment
17

Journal of Medicinal Chemistry
Page 18 of 35
1
2
3
4
5
6
βarrestin2 Recruitmenta
35SꢀGTPγS Bindinga
Bias (GTPγS/βarr2)
7
8
9
ꢁꢁlog(τ/KA)
(95% Cl)c
Bias
Factor
Agonist
X
EC50 (nM) EMAX (%)b ꢁlog(τ/KA) EC50 (nM) EMAX (%)b ꢁlog(τ/KA)
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
DAMGO
Morphine
33 ± 1
64 ± 3
100
220 ± 8.3
100
0
1.0
83 ± 1
ꢀ0.3 ± 0.03 379 ± 17
24 ± 1
ꢀ0.6 ± 0.03 0.25 (0.17 to 0.33)d 1.8
46e
47
149 ± 9.1
83 ± 4
90 ± 9
ꢀ0.8 ± 0.04 >10,000
ꢀ1.7 ± 0.05 >10,000
3.6 ± 1b ꢀ2.4 ± 0.04 1.66 (1.53 to 1.79)d
45
1403 ± 225
6.7 ± 4b
ꢀ2.6 ± 0.2 0.93 (0.31 to 1.55)f 8.5
48
49
2975 ± 400
373 ± 31
87 ± 6
90 ± 4
ꢀ1.9 ± 0.07 >10,000
ꢀ1.1 ± 0.08 >10,000
7.5 ± 3b ꢀ2.7 ± 0.09 0.79 (0.51 to 1.07)d 6.2
29 ± 6b
ꢀ1.8 ± 0.2 0.73 (0.34 to 1.11)g 5.3
50
51
>10,000
>10,000
24 ± 6b
ꢀ2.7 ± 0.4
NC
NC
ꢀ0.3 ± 0.1b
/
/
/
/
/
/
21 ± 10b ꢀ2.8 ± 0.08
ꢀ0.3 ± 0.2
aWithin each assay, potency (EC50) and efficacy (EMAX) values were calculated by nonꢀlinear regression and
ꢁlog(τ/KA) values were determined by the operational model, relative to DAMGO. Data are presented as mean ±
SEM of three or more assays run in duplicate or triplicate. In instances where the data did not converge (NC) or
b
produced a value outside the experimental range (EC50 >10,000 nM), the percentage of maximum stimulation at
c
10,000 nM is provided. For ꢁꢁlog(τ/KA) values, 95% confidence intervals (95% CI) are provided. Significant bias
compared to DAMGO was determined by an unpaired, twoꢀtailed t test: p<0.0001; eCompound 46 was previously
d
reported as SRꢀ15099.22 fp<0.05; gp<0.01. “/” = not determined.
Drug Metabolism and Pharmacokinetics (DMPK) Properties. In order for a compound to
advance into further development, it must be metabolically stable and must have suitable tissue
distribution (including, in this case, high brain exposure). Furthermore, compounds must be nonꢀ
toxic at efficacious doses and must not prompt drugꢀdrug interactions. As a preliminary gauge of
ACS Paragon Plus Environment
18

Page 19 of 35
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
compound metabolic stability, we assessed the stability of promising compounds in the presence
of mouse and human liver microsomes. To determine the potential for drugꢀdrug interactions, we
studied their ability to inhibit four of the major metabolizing cytochrome P450 (CYP) isoforms.
This in vitro DMPK data was used, along with the GTPγS potency and GTPγS/βarr2 bias data, to
select compounds for further PK evaluation in vivo.
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
Table 6 shows selected in vitro and in vivo properties of many compounds in the series that
were chosen for followꢀup due to their significant GTPγS potency (EC50 < 400 nM) and high
bias (GTPγS/βarr2 bias factor > 5). An exception is compound 25, which was used as a
comparator. This compound, like fentanyl,22 is βarrestion2 biased (relative to DAMGO) and
thus, is expected to have enhanced respiratory suppressive effects that could limit its usefulness.
In this series, the inhibition of CYP isoforms was not a major concern: among the compounds
tested, only compound 29 gave > 50% inhibition of any CYP isoform at 10 µM. The liver
microsome stability of these compounds was more widely variable. Fortunately, human liver
microsome stability exceeding 1 hour was seen in some of the compounds (26, 41, 44, and 46).
Though these compounds were generally less stable to mouse liver microsomes compared to the
human, stability was sufficient for in vivo PK evaluations in mice. Of note, we previously
reported that the G protein and βarrestin2 signaling profiles for two compounds in this series (25
and 46) are consistent between the mouse and human MOR.22
To determine if our MOR agonists could penetrate the BBB, which is required in an opioid
analgesic agent, compound 25 and the compounds most highly biased towards GTPγS over βarr2
(41, 44, and 46) were administered intraperitoneally (i.p.) at 6 mg/kg22 and brain levels were
determined after 1 hour. As shown in Table 6, these compounds were present in the brain 1 hour
following systemic injection at levels exceeding that seen with a systemic injection of morphine
ACS Paragon Plus Environment
19

Journal of Medicinal Chemistry
Page 20 of 35
1
2
3
4
5
6
7
8
at the same dose. High brain levels at this time point are consistent with enhanced compound
stability as indicated by liver microsome stability data, suggesting the value of using in vitro
DMPK assessments to drive the selection of compounds for in vivo evaluation.
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
Table 6. In Vitro and In Vivo Evaluation of Select NꢀBenzyl Piperidine 4ꢀBenzimidazolone
Derivatives.
CP450 Inhibition % Microsome Stability
35Sꢀ
Bias
Factor
Brain Level
(µM)c
GTPγS
EC50
Inhibition at 10 µM
T1/2 life (minutes)
Agonist
Morphine
25d
Structurea
(nM)b
1A2 2C9 2D6 3A4 Human Mouse
64
98
31
1.8
0.47
8.3
0.39 ± 0.07
0.70 ± 0.16
6
ꢀ30 ꢀ3
3
16
1
26
ꢀ40
40
1
ꢀ34 ꢀ46
120
14
29
30
48
5.8
7.1
55
50
38
13
ꢀ2
8
6
3
3
184
ꢀ23 20
31
148
6.9
12
34
41
0
17
4
ACS Paragon Plus Environment
20

Page 21 of 35
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
41
44
355
91
23
56
4
41
18
17
84
73
17
31
17 ± 0.35
4.6 ± 0.28
Br
F
Cl
Cl
ꢀ10 ꢀ1 ꢀ16 ꢀ16
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
N
N
HN
O
46d
149
45
5
14
ꢀ1
ꢀ3
233
35
11 ± 2.2
Microsomal stability assays were conducted in vitro. Brain levels of compound were determined in mice 1 hour
following a 6 mg/kg, i.p. injection of the compound indicated. aCompounds 25 and 26 were made and tested as
racemates. bData are presented as mean ± SEM of three or more assays run in duplicate or triplicate. dCompounds 25
and 46 were previously reported as (±)SRꢀ11501 and SRꢀ15099, respectively.22
CONCLUSIONS
In analyzing data for the set of 51 related compounds shown in Tables 1ꢀ5, as well as DMPK
and efficacy properties (Table 6 and Figure 2), we have established a pharmacophore for G
protein biased MOR agonism. Certain hydrophobic substituents, especially halogen atoms in the
distal benzimidazolone and the Nꢀbenzyl ring, are essential for extreme G protein coupling bias.
The presence of a central piperidine ring is also preferred. In many cases, a benzylic methyl
group augments potency and has a small but typically beneficial impact upon bias. The structural
basis for these effects is currently under investigation. Additional ongoing studies include: more
extensive target selectivity profiling of top compounds, safety studies, antinociceptive studies in
multiple species, and the separation and/or the stereoselective asymmetric synthesis of individual
isomers of analogs reported here as racemates (e.g., Table 2 compounds). Because the degree of
ACS Paragon Plus Environment
21

Journal of Medicinal Chemistry
Page 22 of 35
1
2
3
4
5
6
7
8
9
respiratory safety correlates linearly with the magnitude of the GTPγS/βarr2 bias factor as
calculated herein,22 we hope that the trends noted in this SAR study may guide the design of
substantially safer opioid analgesic agents.
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
EXPERIMENTAL SECTION
General Procedures. Materials were purchased from commercial vendors and used without
purification. All moistureꢀsensitive reactions were performed under argon. Experiments were
monitored by LCMS or TLC and visualized using an ultraviolet lamp (254 nm) or staining with
KMnO4. Flash column chromatography was performed using a Teledyne ISCO Combiflash®
Rf+ and Luknova silica gel cartridges. All NMR data was collected at room temperature on a
1
Brüker Ultrashield 400 MHz nuclear magnetic resonance spectrometer. Chemical shifts for H
NMR spectra are reported in parts per million (ppm) relative to residual solvent signal as an
internal standard: DMSO (δ 2.50), CHCl3 (δ 7.26), or MeOH (δ 3.31). Multiplicities are given as:
s (singlet), d (doublet), t (triplet), q (quartet), or m (multiplet). Coupling constants are reported as
a J value in Hertz (Hz). Mass spectra were recorded on a Thermo Scientific 3000 LCQ Fleet
system (ESI) using a Discovery® HS C18 HPLC column (10 cm x 2.1 mm, 5 µm) at 35 oC with
UV detection at 210, 254, and 280 nm. Flow rate was 0.7 mL/min using a solvent gradient of 5ꢀ
95% B over 4 min (total run time = 6 min), where A = 0.1% formic acid in water and B = 0.1%
formic acid in acetonitrile. Analytical HPLC was performed on an Agilent 1100 series using an
Agilent Eclipse XDBꢀC18 HPLC column (4.6 x 150 mm, 5 µm) with UV detection at 254 nm.
Flow rate was 1.75 mL/min using a solvent gradient of 10ꢀ90% B over 8 min (total run time = 10
min), where A = 0.1% TFA and 1% acetonitrile in water and B = acetonitrile. The purity of all
ACS Paragon Plus Environment
22

Page 23 of 35
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
compounds used in the bioassays was determined to be ≥ 95% by analytical HPLC. For the in
vitro studies, DAMGO (Tocris) and morphine sulfate pentahydrate (NIDA Drug Supply
Program) were dissolved in water as 10 mM stocks. All other compounds were dissolved in
100% DMSO as 10 mM stocks. For all assays, the final DMSO concentration was 1%.
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
tert-Butyl 4ꢀ((2ꢀnitrophenyl)amino)piperidineꢀ1ꢀcarboxylate, (I)22
A mixture of 1ꢀfluroꢀ2ꢀnitrobenzene (0.5 mL, 5.0 mmol), Nꢀbocꢀ4ꢀaminopiperidine (1.0 g, 5.0
mmol), K2CO3 (0.8 g, 5.9 mmol), and DMSO (5 mL) was stirred at room temperature overnight
under argon. The reaction mixture was quenched with water and the aqueous layer was extracted
with CH2Cl2; the combined organic layers were dried over Na2SO4 and concentrated to dryness.
Purification afforded pure product I (1.5 g, 93% yield). 1H NMR (400 MHz, CDCl3) δ 8.17 (dd,
J = 8.6, 1.8 Hz, 1H), 8.08 (d, J = 1.6 Hz, 1H), 7.42 (td, J = 7.8, 1.6 Hz, 1H), 6.86 (d, J = 8.4 Hz,
1H), 6.64 (td, J = 7.8, 1.2 Hz, 1H), 4.01 (dt, J = 13.6, 3.8 Hz, 1H), 3.70ꢀ3.64 (m, 1H), 3.04 (td, J
= 12.2, 3.0 Hz, 2H).
tert-Butyl 4ꢀ((2ꢀaminophenyl)amino)piperidineꢀ1ꢀcarboxylate, (II)22
A 50% aqueous suspension of Raney nickel (11.9 mL) was added to a mixture of I (1.5 g, 4.7
mmol) in absolute EtOH (95 mL). Hydrazine hydrate (2.3 mL, 47 mmol) was then added
dropwise. The mixture was heated to 45 oC, stirred for 10 min, and then filtered through a pad of
Celite®. The pad was washed with MeOH and the filtrate was concentrated to dryness.
Purification afforded pure product II (1.0 g, 74% yield). 1H NMR (400 MHz, CDCl3) δ 6.79 (td,
J =7.2, 2.2 Hz, 1H), 6.72ꢀ6.66 (m, 3H), 4.14ꢀ4.02 (m, 1H), 3.41ꢀ3.34 (m, 4H), 2.93 (t, J = 11.8
Hz, 2H), 2.00 (dd, J = 13.0, 3.0 Hz, 2H), 1.48 (s, 9H), 1.36 (qd, J = 12.2, 4.0 Hz, 2H); MS(m/z):
[M+H] calc’d for C16H26N3O2 is 292.19, found 291.76.
ACS Paragon Plus Environment
23

Journal of Medicinal Chemistry
Page 24 of 35
1
2
3
4
tert-Butyl 4ꢀ(2ꢀoxoꢀ2,3ꢀdihydroꢀ1H-benzo[d]imidazolꢀ1ꢀyl)piperidineꢀ1ꢀcarboxylate, (III)22
5
6
7
8
9
CDI (780 mg, 4.8 mmol) was slowly added to a solution of II (1.0 g, 3.4 mmol) in THF (30
mL). The reaction mixture was stirred overnight at room temperature under argon and then
quenched with 10% HCl. The aqueous layer was extracted with EtOAc. The combined organic
layers were dried over Na2SO4 and concentrated to dryness. Purification afforded pure product
III (0.6 g, 55% yield). 1H NMR (400 MHz, CDCl3) δ 9.30 (s, 1H), 7.16ꢀ7.06 (m, 4H), 4.48 (tt, J
= 12.6, 4.0 Hz, 1H), 4.33 (dd, J = 11.6, 2.0 Hz, 2H), 2.88 (td, J = 13.0, 2.4 Hz, 2H), 2.34 (qd, J
= 12.8, 4.4 Hz, 2H), 1.84 (dd, J = 12.0, 2.4 Hz, 2H), 1.51 (s, 9H); MS(m/z): [M+H] calc’d for
C17H24N3O3 is 318.17, found 317.71.
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
1ꢀ(Piperidinꢀ4ꢀyl)ꢀ1Hꢀbenzo[d]imidazolꢀ2(3H)ꢀone TFA salt, (IV)22
Compound III (0.6 g, 1.9 mmol) was dissolved in a 33% solution of TFA in CH2Cl2 (6 mL).
The reaction mixture was stirred at room temperature until completion and then was
concentrated to dryness. The solid was dissolved in 1:1 water/acetonitrile. The solution was
frozen and then subjected to lyophilization overnight giving IV in the form of a TFA salt. This
1
material was used without further purification. H NMR (400 MHz, CD3OD) δ 7.29ꢀ7.26 (m,
1H), 7.11ꢀ7.08 (m, 3H), 4.55 (tt, J = 12.2, 4.2 Hz, 1H), 3.58 (dt, J = 12.8, 2.2 Hz, 2H), 3.22 (td,
J = 13.2, 2.4 Hz, 2H), 2.74 (qd, J = 12.6, 4.4 Hz, 2H), 2.06 (dd, J = 12.8, 1.6 Hz, 2H); MS(m/z):
[M+H] calc’d for C12H16N3O is 218.12, found 217.92.
1ꢀ(1ꢀ(Benzyl)piperidinꢀ4ꢀyl)ꢀ1Hꢀbenzo[d]imidazolꢀ2(3H)ꢀone, (1)25
A mixture of IV (free base, 110 mg, 0.5 mmol), benzyl bromide (87 mg, 0.5 mmol), DIPEA
(132 µL, 0.8 mmol), and DMF (1 mL) was stirred at room temperature overnight under argon.
Upon completion, the solvent was removed under reduced pressure and the residue was
ACS Paragon Plus Environment
24

Page 25 of 35
Journal of Medicinal Chemistry
1
2
3
4
5
6
dissolved in EtOAc (20 mL). This reaction mixture was washed with saturated NaHCO3
followed by brine. The organic layer was dried over Na2SO4 and concentrated to dryness.
7
8
1
Purification afforded pure product 1 (139 mg, 89% yield). H NMR (400 MHz, (CD3)2SO) δ
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
10.82 (s, 1H), 7.35ꢀ7.34 (m, 4H), 7.28ꢀ7.21 (m, 1H), 7.01ꢀ6.95 (m, 3H), 4.14 (tt, J = 12.4, 4.0
Hz, 1H), 3.53 (s, 2H), 2.94 (d, J = 11.6 Hz, 2H), 2.35 (qd, J =12.4, 4.0 Hz, 2H), 2.09 (t, J = 11.0
Hz, 2H), 1.64 (d, J =9.2 Hz, 2H); MS(m/z): [M+H] calc’d for C19H22N3O is 308.17, found
308.06; HPLC tR = 3.39 min.
GTPγS Binding
MORꢀstimulated GTPγS binding was determined in membranes prepared from CHOꢀhMOR
cells as described.22 CHOꢀhMOR cells were serumꢀstarved for 30 minutes and collected via
gentle scraping in EDTA buffer. Membranes were then prepared via dounce homogenization and
centrifugation at 20,000 x g for 30 minutes at 4°C. GTPγS binding reactions were performed in
200 ꢂL volumes containing 10 ꢂg CHOꢀhMOR membranes, 50 ꢂM guanosineꢀ5”ꢀdiphosphate
35
(GDP, Sigma Aldrich G7127), 0.1 nM SꢀGTPγS (PerkinElmer NEG030H) and concentrations
of the compounds ranging from 0.1 nM to 10 ꢂM were incubated for 1 hour at room temperature.
Reactions were terminated by rapid filtration through GF/B glass fiber filter plates (PerkinElmer)
and radioactivity was counted with a TopCount NXT Scintillation Counter (PerkinElmer).
βarrestin2 Enzyme Fragment Complementation Assay
USOSꢀβarrestinꢀhMORꢀPathHunter cells (DiscoveRx 93ꢀ0213C3) were used to determine
βarrestin2 interactions with the MOR by an enzyme fragment complementation assay.22 The
cells (4,000 cells/well) were incubated with 0.1 nM to 10 or 31 ꢂM (depending on compound
solubility) concentrations of the test compounds at 37 °C for 90 minutes and βarrestin2
ACS Paragon Plus Environment
25

Journal of Medicinal Chemistry
Page 26 of 35
1
2
3
4
5
6
7
8
9
translocation was determined according to the manufacturer’s instructions (DiscoveRx).
Luminescence was measured using a SpectraMax M5e Microplate Reader (Molecular Devices)
with 1 second integration times.
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
Pharmacokinetic Parameters
To determine stability in hepatic microsomes, compound (1 µM) was incubated with 1 mg/mL
human or mouse hepatic micosomes at 37°C with continuous shaking.26 At 0, 5, 10, 20, 40 and
60 minute timeꢀpoints, aliquots were removed and acetonitrile was added to quench the reactions
and precipitate the proteins. Samples were then centrifuged though 0.45 micron filter plates and
halfꢀlives were determined by LCꢀMS/MS.
To determine cytochrome P450 inhibition, 1 µM compound was incubated with human liver
microsomes and selective marker substrates (1A2, phenacetin demethylation to acetaminophen;
2C9, tolbutamide hydroxylation to hydroxytolbutamide; 2D6, bufuralol hydroxylation to 4′ꢀ
hydroxybufuralol; 3A4, midazolam hydroxylation to 1′ꢀhydroxymidazolam). After a 10 minute
incubation, the reaction was terminated and the percent inhibition was determined.27
Pharmacokinetics in mice: Male C57BL6/J mice were purchased from Jackson Laboratory
(Bar Harbor, ME) and assessed at 10ꢀ12 weeks per age. All mice were used in accordance with
the National Institutes of Health Guidelines for the Care and Use of Laboratory Animals with
approval by the Scripps Research Institutional Animal Care and Use Committee (IACUC). To
determine brain penetrance, mice were injected with drug (i.p. at mg/kg at 10uL/g volumes
mouse body weight using a vehicle of 1:1:8 DMSO, Tween80, and purified water). Mice were
sacrificed by cervical dislocation 1 hour following drug treatment and isolated brains were flash
ACS Paragon Plus Environment
26

Page 27 of 35
Journal of Medicinal Chemistry
1
2
3
4
5
6
frozen in liquid nitrogen. Drug levels were determined using a LCꢀMS operated in positiveꢀion
mode, 1 hour after treatment.
7
8
9
Data and Statistical Analysis
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
GraphPad Prism (v. 7.0) was used for data and statistical analysis. All data are presented as
mean ± S.E.M or with 95% confidence intervals, as indicated in the figure and table legends. For
the in vitro studies, the assays were run in duplicate, with at least 3 independent replicates and
DAMGO was run as the reference compound in every experiment for normalization. EC50 and
EMAX values were calculated by nonlinear (three parameter) regression analysis. Due to the limit
of solubility for some compounds, EMAX values are reported at 10 µM concentrations in instances
where the data did not converge or where potency values were outside the linear experimental
range of the assay.
Analysis of bias: Bias factors were calculated according to the operational model using
DAMGO as the reference agonist. As in the prior manuscript, the conservative constraint was
applied to the operational model to fit the KA value to fall between 0 and 10ꢀ15M to allow for
curve fitting in the absence of reaching a maximum plateau in one of the responses; the
ꢀlog(τ/KA) values, relative to DAMGO were constrained to be less than 10.22,27
ASSOCIATED CONTENT
The Supporting Information is available free of charge on the ACS Publications website at
DOI:
ACS Paragon Plus Environment
27

Journal of Medicinal Chemistry
Page 28 of 35
1
2
3
4
Synthetic procedures and characterization details of additional compounds, plus molecular
5
6
strings, plus 1 HNMR spectra of key compounds that were used for in vivo studies (PDF).
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
AUTHOR INFORMATION
Corresponding Authors
*Thomas D. Bannister, for medicinal chemistry issues: email, tbannister@scripps.edu; phone,
561ꢀ228ꢀ2206
* Laura M. Bohn, for pharmacology studies: email, lbohn@scripps.edu; phone, 561ꢀ228ꢀ2227
Author Contributions
The manuscript was written through contributions of all authors. All authors have given approval
to the final version of the manuscript. ‡N.M.K and C.L.S have contributed equally.
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENT
This project is supported by NIH grant R01DA033073 (to L.M.B. and T.D.B.) and
R01DA038694 (to L.M.B.). The authors have filed for patent protection for the compounds
described herein.
ABBREVIATIONS
ACS Paragon Plus Environment
28

Page 29 of 35
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
BBB, bloodꢀbrain barrier; Boc, tertꢀButyloxycarbonyl, CDI, 1,1ꢀcarbonyldiimidazole; CH2Cl2,
methylene chloride; CYP, cytochrome P450; DAMGO, [DꢀAla2, NMeꢀPhe4, Glyꢀol5]ꢀ
enkephalin; DIPEA, N,Nꢀdiisopropylethylamine; DMPK, drug metabolism and
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
pharmacokinetics; DMSO, dimethylsulfoxide; EC50, half maximal effective concentration; EMAX
,
maximal efficacy; EFC, enzyme fragment complementation; EtOH, ethanol; GPCR, G proteinꢀ
coupled receptor; HPLC; high pressure liquid chromatography; i.p., intraperitoneally; H2NNH2,
hydrazine; K2CO3, potassium carbonate; kg, kilogram; KMnO4, potassium permanganate; KO,
knockout; mg, milligram; LC/MS, liquid chromatography mass spectrum; MOR, mu opioid
receptor; NaBH3CN, sodium cyanoborohydride; NaBH(OAc)3, sodium triacetoxyborohydride;
Ni, nickel; nm; nanometer; nM, nanomolar; NMR, nuclear magnetic resonance; PK,
pharmacokinetics; 35SꢀGTPγS, G proteinꢀcoupling assay; SAR, structure activity relationship;
TFA, trifluoroacetic acid; TLC, thin layer chromatography; THF, tetrahydrofuran; µM,
micromolar; µg, microgram.
REFERENCES
1
(a) Morgan, M. M.; Christie, M. J. Analysis of Opioid Efficacy, Tolerance, Addiction and
Dependence from Cell Culture to Human. Br. J. Pharmacol. 2011, 164, 1322ꢀ1334. (b)
Melnikova, I. Pain Market. Nat. Rev. Drug Discov. 2010, 9, 589ꢀ590.
2 (a) Seth, P.; Scholl, L.; Rudd, R. A.; Bacon, S. Overdose Deaths Involving Opioids, Cocaine,
and PsychostimulantsꢀUnited States, 2015ꢀ2016. MMWR Morb. Mortal Wkly. Rep. 2018, 67,
ACS Paragon Plus Environment
29