Free-wilson and structural approaches to Co-optimizing human and rodent isoform potency for 11β-hydroxysteroid dehydrogenase type 1 (11β-HSD1) inhibitors

Article abstract of DOI:10.1021/jm3013163

11β-Hydroxysteroid dehydrogenase 1 (11β-HSD1) has been a target of intensive research efforts across the pharmaceutical industry, due to its potential for the treatment of type II diabetes and other elements of the metabolic syndrome. To demonstrate the value of 11β-HSD1 in preclinical models, we required inhibitors with good potency against both human and rodent isoforms. Herein, we describe our efforts to understand how to co-optimize human and murine potency within the (5-hydroxy-2-adamantyl)-pyrimidine-5-carboxamide series. Two approaches are described-a data-driven (Free-Wilson) analysis and a structure-based design approach. The conclusions from these approaches were used to inform an efficient campaign to design compounds with consistently good human/murine potency within a logD_7.4 range of 1-3. Compounds 20 and 26 demonstrated good rodent PK, which allowed us to demonstrate a PK/PD relationship in rat and mouse. We then evaluated 26 against glycemic and body weight end points in murine disease models, where it demonstrated glucose and body weight efficacy at 300 mg/kg/day but only body weight efficacy at 50 mg/kg/day, despite providing >90% target engagement in the liver.

Full text of DOI:10.1021/jm3013163

Article
pubs.acs.org/jmc
Free-Wilson and Structural Approaches to Co-optimizing Human and
Rodent Isoform Potency for 11β-Hydroxysteroid Dehydrogenase
Type 1 (11β-HSD1) Inhibitors
Frederick W. Goldberg,* Andrew G. Leach, James S. Scott, Wendy L. Snelson, Sam D. Groombridge,
Craig S. Donald, Stuart N. L. Bennett, Cristian Bodin, Pablo Morentin Gutierrez, and Amy C. Gyte
AstraZeneca, Mereside, Alderley Park, Macclesfield, SK10 4TG, United Kingdom
S
* Supporting Information
ABSTRACT: 11β-Hydroxysteroid dehydrogenase 1 (11β-
HSD1) has been a target of intensive research efforts across the
pharmaceutical industry, due to its potential for the treatment of
type II diabetes and other elements of the metabolic syndrome.
To demonstrate the value of 11β-HSD1 in preclinical models, we
required inhibitors with good potency against both human and
rodent isoforms. Herein, we describe our efforts to understand
how to co-optimize human and murine potency within the (5-
hydroxy-2-adamantyl)-pyrimidine-5-carboxamide series. Two ap-
proaches are describeda data-driven (Free-Wilson) analysis and a structure-based design approach. The conclusions from these
approaches were used to inform an efficient campaign to design compounds with consistently good human/murine potency
within a logD7.4 range of 1−3. Compounds 20 and 26 demonstrated good rodent PK, which allowed us to demonstrate a PK/PD
relationship in rat and mouse. We then evaluated 26 against glycemic and body weight end points in murine disease models,
where it demonstrated glucose and body weight efficacy at 300 mg/kg/day but only body weight efficacy at 50 mg/kg/day,
despite providing >90% target engagement in the liver.
optimizing the potency against both isoforms was challenging
as the human/murine SAR was orthogonal in some cases. In
this paper, we describe our efforts to demonstrate preclinical
target engagement and identify compounds for murine efficacy
models, by co-optimizing human and rodent potency of neutral
compounds within the (5-hydroxy-2-adamantyl)-pyrimidine-5-
carboxamide series. Two approaches are describeda data
driven (Free-Wilson) analysis and a structural approach.
INTRODUCTION
■
11β-Hydroxysteroid dehydrogenase 1 (11β-HSD1) has been a
target of intensive research efforts across the pharmaceutical
industry,1,2 due to its potential for the treatment of type II
diabetes and other elements of the metabolic syndrome.3,4
Demonstrating preclinical in vivo efficacy has been a challenge,
however, because of the nature of the in vivo models and the
difficulty in identifying inhibitors with good potency against
both human and rodent isoforms, as we found in our discovery
of 2-[(3S)-1-[5-(cyclohexylcarbamoyl)-6-propylsulfanylpyridin-
2-yl]-3-piperidyl]acetic acid (AZD4017).5 In a previous
communication,6 we described our efforts to identify 11β-
HSD1 inhibitors with improved PK; key to that success was
switching the core from pyridine to pyrimidine. Pyrimidines
such as 1 (Figure 1) demonstrated good potency against both
human and murine isoforms, but it was noticed that further co-
RESULTS AND DISCUSSION
■
A Free-Wilson approach7 assumes that the SAR between
substructures is additive and thus potency and properties of
new compounds, comprised of known substituents, may be
predicted using the observed contributions of the substituents
on a specific core. It can be a powerful method in the context of
a mature research program, where data exist on a wide variety
of substructures and the goal is to efficiently and rapidly
optimize parameters such as potency and lipophilicity. We felt
that the (5-hydroxy-2-adamantyl)-pyrimidine-5-carboxamide
scaffold 2 (Figure 1) was suitable for this approach as the
core had given good PK properties (1 gave 100% oral
bioavailability in rat and dog)6 and the R1 and R2 groups are
assumed to be conformationally independent. We therefore
analyzed the data we had to predict effects on murine and
Received: September 12, 2012
Published: November 16, 2012
Figure 1.
© 2012 American Chemical Society
10652
dx.doi.org/10.1021/jm3013163 | J. Med. Chem. 2012, 55, 10652−10661

Journal of Medicinal Chemistry
Article
Figure 2. Free-Wilson model contributions for groups at R1 and R2 toward human and murine in vitro pIC50 vs equivalent group contribution
toward logD7.4. The 1:1 line is shown (black), and groups above this line contribute to improvements in LLE. The blue line (R1) indicates the effect
of increasing size of cycloalkyl groups.
human potency of R1 and R2 for this scaffold. This analysis was
then used to analyze global trends (e.g., with size and polarity)
and to evaluate new combinations of R1 and R2, as well as new
isosteric groups, that might give high potency against both
isoforms when applied to the (5-hydroxy-2-adamantyl)-
pyrimidine-5-carboxamide scaffold. We aimed to do this
without increasing logD, that is, to improve ligand lipophilicity
efficiency (LLE, defined in this article as pIC50-logD7.4).8
To demonstrate that SAR is additive in a Free-Wilson type
analysis, Oyarzabal et al.9 have recommended studying the R2
values for predictions made on a test set after model building
on a training set and recommend that additive models should
produce Rtest2 values greater than 0.6. In our case, most models
for both human and murine potency (only using in-range
values) achieve this and so are shown to be additive. The mean
as estimated by a Free-Wilson model for that property. For R1
in human, the majority of substituents sit on or above a 1:1 line
for pIC50:logD; that is, potency is driven by lipophilicity. The
groups that sit significantly below the 1:1 line are structural
features that are apparently not well tolerated, for example, in
t
this plot, tBu, CH2 Bu, and CH2-cyclopentyl (c-Pn) for human,
suggesting that these groups are too big. Murine potency shows
a similar trend for lipophilicity driving potency, and again, large
lipophilic groups such as cyclohexyl sit off the 1:1 murine
pIC50:logD line. Our conclusion from this analysis is that R1 is
a difficult position to improve LLE as there is an excellent
correlation between potency and logD for the preferred groups,
so limited effort should be expanded on optimizing the R1
position. We choose a narrow range of lipophilic ether/alkyl R1
groups, controlling the size to ensure a good balance between
human and murine potency but focused elsewhere on the
scaffold to use polar groups to improve LLE and
physicochemical properties.
For R2, the correlations between potency and lipophilicity
are less clear, reflecting more subtle SAR where specific polar
interactions can be picked up. This data set largely reflects
amine SAR since most compounds tested were 2-amino-
pyrimidines (see the Supporting Information for the data set
used and resulting Free-Wilson contributions). Within that
restriction, it can be seen that having a donor is apparently
important for murine potency but not essential for human
potency. This general trend is exemplified by the distribution of
points for secondary (NHR) and tertiary (NR2) nitrogen
substituents (Figure 2, R2 mouse), and so we concluded that
secondary groups at R2 should allow us to achieve good levels
of human and murine potency. There is also a slight trend
2
value of Rtest for human potency is 0.72 and for murine
potency is 0.79 with standard deviations of 0.10 and 0.12,
respectively. The standard error in each mean is 0.03,
demonstrating that at 95% confidence, the underlying
population mean is greater than 0.6, thus supporting the
conclusion that group contributions from R1 and R2 are
additive. More detailed parameters are provided in the
Supporting Information.
With confidence that the SAR was additive, we built Free-
Wilson models for each species covering the full set of
compounds to assess the contribution that the different groups
at R1 and R2 make to human/murine in vitro potency and
logD7.4. The coefficients computed for the various groups
within the Free-Wilson model are plotted in Figure 2, where
the contributions of R1 and R2 groups toward human and
murine pIC50 are plotted against contributions toward logD7.4
10653
dx.doi.org/10.1021/jm3013163 | J. Med. Chem. 2012, 55, 10652−10661

Journal of Medicinal Chemistry
Article
Figure 3. Binding site of the complex of 3 with the human isoform of 11β-HSD (4bb6), in which the side chains of all residues within 4 Å of the
ligand are shown. The region that the R1 and R2 groups on the pyrimidine are expected to occupy is indicated in maroon and light blue ovals,
respectively. NADP is shown in the background with green-colored carbon atoms, and the hydrogen bonds between Ser170 and Tyr183 are
indicated with blue dashed lines.
toward bigger, more lipophilic groups increasing both human
and murine potency, although the trend is less pronounced
than for R1. The R2 group apparently offers a wider scope for
different size and lipophilicity and the opportunity to
incorporate polar groups to control logD and properties.
In tandem with the Free-Wilson approach detailed above, the
wealth of structural information available both internally and
externally was examined. At that time, there was no available
structure of a pyrimidine amide bound to 11β-HSD1, but a
structure of a related pyridine had been obtained internally
(PDB code: 4bb6). Along with many of the publicly available
structures of complexes with 11β-HSD1, this suggested that the
binding mode is dominated by correct positioning of the
adamantyl amide (or equivalent donor) group.10 We made the
assumption that although the specific SAR may differ between
the series, the broad positions in space of the groups around the
pyrimidine core would be similar to those around the pyridine
core of compound 3 shown in Figure 3.
in an analogous fashion to the pyridine of 3. This provides
some interesting rationalization of the SAR that was suggested
by the Free-Wilson analysis. The R1 group is in a lipophilic
pocket lined by the Leu126, Tyr177, Pro178, and Val180
residues. The region occupied by R2 is bounded by the
lipophilic regions of Leu171, Leu217, and Met233 but also by
the side chain of Tyr177, which has its hydroxyl group oriented
toward R2. This is consistent with the SAR above in which R1
tends to be lipophilic to enhance human potency and R2 has
more varied SAR in which hydrogen-bonding groups may be
tolerated. Presumably, groups that can interact favorably with
either the hydroxyl or the electron-rich aromatic side chain of
Tyr177 might enhance potency while reducing lipophilicity.
The contrast between the structures of human and murine
11β-HSD1 is illustrated in Figure 4. As a murine structure was
The enzyme consists to a great extent of fairly rigid helices
and sheets, but the site immediately surrounding the binding
site is dominated by flexible loops. This suggests that the
binding site is likely to be mobile. It should be noted that there
are two molecules in the asymmetric repeat, and the
conformation of the ligand and receptor is similar in both,
and so only one is shown here. It is a subject of debate whether
the dimer is relevant to the in vivo situation or just an artifact of
the crystallographic experiment and in vitro assays.11−14 In
Figure 3, the binding site is shown together with the NADP
cofactor. The principal interaction is likely related to the
mechanism of action of the enzyme and consists of two
hydrogen bonds to the amide carbonyl formed by Ser170 and
Tyr183. This carbonyl group is also positioned in close
proximity to the C4 position of the nicotinamide segment of
NADP from where hydride delivery takes place during the
reduction of cortisone to cortisol. The adamantyl group is
buried in the deeper part of the pocket, and it is likely that this
region is occupied by a hydrophobic section of the steroid
during the normal enzymatic process. This is consistent with
the adamantyl amide being a mimic of the substrate and/or
transition state and therefore making a significant contribution
to the affinity of the ligand for the enzyme.
Figure 4. Complex of 3 with murine 11β-HSD1 generated by overlay
of the complex shown in Figure 3 and the complex of murine 11β-
HSD1 with n-octane (1y5m). In pink are the murine residues that
were within 4 Å of the ligand in the human complex and are different
in humans; residue labels are shown in gray for humans and pink for
murine.
not available to us internally, the murine structure is taken from
the complex with n-octane reported by Zhang et al. (PDB code:
1y5m).15 The murine structure was overlaid on that of the
complex of the human form with 3 using the protein
superposition function in Maestro.16
The structure shown in Figure 3 demonstrates the regions
into which the R1 (maroon oval) and R2 (light blue oval)
would be expected to bind if the pyrimidine series were to bind
It is interesting to note that the lipophilic pocket into which
R1 protrudes has a key residue in human (Ala180) that is larger
in murine (Ile180). This reduces the size of the pocket slightly
10654
dx.doi.org/10.1021/jm3013163 | J. Med. Chem. 2012, 55, 10652−10661

Journal of Medicinal Chemistry
Article
a
Scheme 1. Synthetic Route for R1 = Alkyl
a
Reagents and conditions: (i) 2-Methyl-2-thiopseudourea sulfate, NaOAc, DMF, 44−80%. (ii) Oxone, MeCN, H2O or mCPBA, DCM, 82−94%.
(iii) R2NH2, i-Pr2NEt, THF, 23−100%. (iv) 2 M aqueous NaOH, MeOH, 64−95%. (v) HATU, trans-4-aminoadamantan-1-ol·HCl, i-PrNEt, DMF,
58−88%.
a
Scheme 2. Synthetic Route for R1 = Ether
a
Reagents and conditions: (i) R1ONa, THF or R1OH, 51−91%. (ii) R2NH2, i-Pr2Net, n-PrCN or THF, 23−39%.
and also enables smaller groups [such as cyclopropyl (c-Pr), see
Figure 2] to make hydrophobic contacts in the murine isoform
form that they are not able to make in human; hence, small
groups at R1 are better tolerated in murine. Similarly, whereas
c-Pn is a good group for human potency and presumably makes
good contacts with Ala180, in murine, the group may clash with
the larger Ile180. The Free-Wilson and protein−ligand complex
both suggest that large lipophilic groups at R1 are preferred for
human potency, but smaller lipophilic groups are preferred for
murine potency.
By contrast, the region that R2 protrudes into is dominated
by the change from Tyr177 (human) to Gln177 (murine). This
large change in the size and physical properties of the residue is
likely a key cause of the spread about the line seen in the Free-
Wilson analysis shown in Figure 2. It is striking that the flexible
Gln side chain that can interact with donors and acceptors is
consistent with a favoring of NH-alkyl for murine activity in the
group contributions identified in the Free-Wilson analysis. It
could also be argued that the Gln carbonyl is a stronger
acceptor than the Tyr phenol (so stronger preference for a
donor on the ligand for murine potency), although rationalizing
that a priori would have been difficult without the Free-Wilson
analysis. The groups shown in Figure 2 that are able to
contribute to both human and murine potency are presumably
able to form good interactions with both Gln and Tyr in this
position. The position is relatively open to solvent in both
isoforms and so larger groups ought to be tolerated in both
species.
control logD and explore optimal placement of donors and
acceptors. We selected compounds with a predicted logD7.4 of
1−3 to increase the chances of obtaining favorable PK
properties.17,18 The R1 groups chosen were c-Pr, cyclobutyl
(c-Bu), c-Pn, cyclobutyloxy, and isopropoxy, which were all on
the pIC50/logD leading edge of the Free-Wilson plot (i.e.,
predicted to give high LLE).
Synthetic routes were developed that enabled us to readily
vary the R1 and R2 groups. Thus, for targets with an alkyl R1
group, enamine 4 (synthesized from corresponding ketoester)
was converted to 2-methylthiopyrimidine 5 with 2-methyl-2-
thiopseudourea sulfate (Scheme 1). The thioether was then
oxidized to the sulfone with oxone and displaced with an amine
to give 6. Intermediate 6 was then converted to the target
compound 7 via saponification and subsequent 2-(3H-[1,2,3]-
triazolo[4,5-b]pyridin-3-yl)-1,1,3,3-tetramethylisouronium
hexafluorophosphate(V) (HATU)-mediated amide coupling
with trans-4-aminoadamantan-1-ol. This route was generally
reliable across a range of functionality and reproducible on
scale-up. However, an alternative route was also used to convert
5 to the corresponding amide 8 before addition of the R2
amine. Although this route was more capricious, it allowed us
to vary R2 as the last step.
For targets with an ether R1 group, the synthesis started with
addition of trans-4-aminoadamantan-1-ol to 2,4-dichloropyr-
imidine-5-carbonyl chloride to give amide 9 (Scheme 2). The
addition of alkoxides to 9 occurred selectively in the C4
position to give 10, and subsequent addition of the relevant R2
primary amine provided us with the desired product 11. Again,
this route allowed us to vary the R2 group at the last step,
which suited the optimization strategy of fixing R1 as a small
number of lipophilic ethers but varying the R2 amine more
extensively to pick up polar interactions.
Using the conclusions from the Free-Wilson and structural
analyses, our strategy relied upon using a limited set of five
cycloalkyl/ether R1 groups (lipophilic groups without addi-
tional polarity) that were well tolerated for both isoforms
according to the Free-Wilson analysis and a more expansive set
of 20 secondary amine R2 groups with additional polarity to
10655
dx.doi.org/10.1021/jm3013163 | J. Med. Chem. 2012, 55, 10652−10661

Journal of Medicinal Chemistry
Article
Table 1. Effect of R1 and R2 Groups on Human/Murine Potency and logD7.4
a
IC50 (nM)
b
ex.
12
R1
R2
human
murine
logD7.4
rat F (%)
c-Pr
A
B
C
G
R
T
B
C
G
I
74
30
36
53
94
108
23
31
28
54
31
46
9
38
32
30
22
20
31
37
42
27
53
168
47
66
25
32
36
55
50
38
84
64
133
25
42
85
179
73
42
67
32
163
48
53
21
35
33
16
22
19
31
31
20
1.1
1.6
1.7
1.8
1.8
1.5
2
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
c-Pr
c-Pr
73
53
c-Pr
c-Pr
c-Pr
c-Bu
c-Bu
c-Bu
c-Bu
c-Bu
c-Bu
c-Pn
c-Pn
c-Pn
c-Pn
c-Pn
c-Pn
c-Pn
c-Pn
c-Pn
c-Pn
c-Pn
c-Pn
c-Pn
c-Pn
c-Pn
c-Pn
c-Pn
c-Pn
c-Pn
c-Pn
Oi-Pr
Oi-Pr
Oi-Pr
Oc-Bu
Oc-Bu
Oc-Bu
Oc-Bu
Oc-Bu
Oc-Bu
Oc-Bu
99
82
1.8
2.1
1.1
1.3
1.8
2.1
2.4
2.4
2.7
0.9
1.8
2.6
1.9
1.6
1.6
2.7
2.8
1.4
1.8
2.4
2
100
N
T
A
B
C
D
E
86
21
21
29
94
25
23
38
16
27
27
22
34
12
15
22
26
30
45
19
13
11
20
12
12
20
30
48
14
30
63
61
17
F
18
46
G
H
I
J
K
L
M
N
O
P
9
Q
R
S
2
68
2.5
1.4
2.3
2.2
2.7
2.5
2.6
2.9
2.8
3.2
2.1
2.5
3.1
T
A
B
C
A
B
C
G
I
33
57
62
83
19
99
99
83
N
R
51
161
a
Human and murine potency as measured by the geometric mean of (at least) two independent IC50 measurements in a homogeneous time-resolved
fluorescence (HTRF) assay; see the Supporting Information for assay details. The standard error of the mean (SEM) is not shown as both human
b
and murine HTRF assays were found to be highly reproducible on repeat testing (pIC50 SEM typically <0.1). Oral bioavailability observed in male
Han Wistar rats at low dose (po 7.4−14.7 μmol/kg suspension, iv 2.3−5.6 μmol/kg solution); drug concentrations measured in whole blood.
10656
dx.doi.org/10.1021/jm3013163 | J. Med. Chem. 2012, 55, 10652−10661

Journal of Medicinal Chemistry
Article
Figure 5. Effect on potency of switching R2 from tertiary to secondary amines.
Rather than synthesizing the whole R1xR2 matrix, we chose
to enumerate all 20 R2 amines (Table 1) with R1 = c-Pn and
select a smaller selection of key R2 amines for the other R1
groups to give the 42 compounds given in Table 1. c-Pn was
chosen for the full enumeration as it gives higher human
potency than c-Pr and c-Bu (see Figure 2), albeit at the cost of
reduced murine potency, and at the time, we had less data on
the PK properties of O-linked R1 groups so that these groups
were perceived as being higher risk. The design of the R2
amines in part depended on exact groups identified by the Free-
Wilson analysis as being good for human and murine potency.
Examples of this approach include 1,1-dioxothian-4-amine (21,
32, and 52, Table 1) and oxetan-3-amine (12, 24, 44, and 47).
However, we also used the conclusions from the Free-Wilson
analysis, in conjunction with the structural analysis, to design
new (i.e., not in the Free-Wilson data set) isosteric groups at
R2 with suitable placement of acceptors to explore SAR and
give favorable predicted properties. Examples of these isosteric
groups are tetrahydrofuran-3-amine (13/14, 18/19, 25/26, 45/
46, and 48/49 for R/S enantiomers) and 2-aminocyclopentanol
(27). The data in Table 1 demonstrate that we succeeded in
synthesizing compounds with consistently high human and
murine potency. Remarkably, all 42 compounds synthesized in
this campaign had an IC50 < 180 nM in both human and
murine assays, and typical IC50 values are in the 10−100 nM
range. The SAR is relatively flat, as commented upon in the
Free-Wilson analysis. An interesting example of this flat SAR is
the similar potencies given by R/S enantiomers of tetrahy-
drofuran-3-amine, where all human and murine potencies are
within 2-fold between the enantiomers. We concluded that
human and murine potencies within this series are apparently
not particularly sensitive to the exact placement of the acceptor
at R2. Measured logD7.4 was generally in the desired 1−3 range,
with measured logD values being very accurately predicted
( 0.2) by our AstraZeneca calculated logD model AZlogD7.4.18
This model is built on measured logD7.4 data on compounds in
the AstraZeneca corporate collection. The notable exceptions
were all compounds where R1 = Oc-Bu (compounds 47−53),
where the measured logD was ∼0.5 units higher than predicted;
for example, 50 and 53 (the most lipophilic compounds in this
set) both gave a predicted AZlogD7.4 of 2.6, but measured
logD7.4 values were higher at 3.2 and 3.1, respectively
(publically available model ACDlogD7.4 v12 also underpredicts
50 and 53 as being 2.2 and 2.1, respectively). Rat bioavailability
of these compounds was generally good as shown in Table 1
with 18/23 compounds profiled showing rat F > 40% (for more
detailed in vivo PK parameters on selected compounds, see
Table 3). This campaign thus very efficiently generated a large
number of suitable compounds to profile in vivo. In particular,
compounds 13, 20, 26, 46, 48, and 49 all have excellent human
and murine potency with good oral exposure in Han Wistar
rats. We selected compounds 20 and 26 to progress into more
advanced in vivo studies due to them having a shorter half-life
than 13, 46, 48, and 49. This was a desirable feature
preclinically as it allowed us to examine the effects of
continuous cover versus episodic dosing to target the cortisol
diurnal peak,20 to address our concerns that target engagement
may reduce on chronic dosing with continuous cover. Data to
support this hypothesis will be published shortly.
Key to the success was the switch to secondary amines at R2.
As can be seen in Figure 5, switching the morpholine at R2 in
compound 54 (disclosed in a previous communication6) to
tetrahydropyran-4-amine increases murine potency and thus
gives compound 20, which combines good human and murine
potency at moderate logD. Also shown in Figure 5 is isomeric
26, which also demonstrated good human/murine potency at
moderate logD.
Compounds 20 and 26 were profiled further in vitro to
establish that they are potent inhibitors of human 11β-HSD1
(Table 2) and selective against related targets. Thus, 20 and 26
Table 2. Human in Vitro Potency of 20 and 26 against 11β-
HSD1 and Related Targets
11β-HSD1
11β-HSD1
11β-HSD2
17β-HSD1 17β-HSD3
IC50
(HTRF)
(nm)
IC50
(HPLC)
(nm)
IC50
(HTRF)
(μM)
IC50
(HPLC)
inh
(HPLC)
ex.
(μM)
20
28
14
>30
>35
4% at
10 μM
26
21
1.6
>30
>100
7% at
10 μM
gave HTRF potencies of 21 and 28 nM, respectively, and
HPLC potencies of 2 and 14 nM, respectively (see the
Supporting Information for assay details). The HPLC assay is a
low-throughput assay that accounts for tight binding and thus
should give a more accurate in vitro potency than the HTRF
assay. Both compounds were inactive against 11β-HSD2, which
is important as 11β-HSD2 is believed to be responsible for the
reverse reaction in vivo (11β-HSD1 primarily converts
cortisone to cortisol, whereas 11β-HSD2 primarily converts
cortisol to cortisone), and has been associated with increased
risk of hypertension.19 Both compounds were also inactive
against two 17β-hydroxysteroid dehydrogenases tested (17β-
HSD1 and 17β-HSD3).
We were also able to obtain a crystal structure of 26 (Figure
6). The structure of 26 in complex with 11β-HSD1 (PDB code:
4bb5) crystallizes such that there are four protein−ligand
complexes in the asymmetric unit. The protein conformation is
broadly the same in all four complexes including in the vicinity
of the binding site. The exception is the solvent-exposed side
chain of Met233, which is solved with four different
conformations; as might be expected for what seems to be a
dynamic side chain, the electron density does not strongly
10657
dx.doi.org/10.1021/jm3013163 | J. Med. Chem. 2012, 55, 10652−10661

Journal of Medicinal Chemistry
Article
Figure 7 shows the PK/PD relationships between the ex vivo
activity of the 11β-HSD1 enzyme in epididymal adipose tissue
3
3
(measured as the conversion of H-cortisone to H-cortisol)
and drug concentrations of 20 and 26. A direct response PK/
PD model (Emax sigmoidal model) was used to successfully fit
the PK/PD data showing, in each case, a clear correlation
between free compound levels in plasma and inhibition in the
activity of 11β-HSD1 in adipose tissue (see the Supporting
Information for ex vivo assay details and model-fitting
parameters).
In each case, the in vivo IC50 in the ex vivo adipose assay is in
very good agreement with the in vitro potency (values shown in
Table 4), suggesting that systemic exposure, defined as free
compound concentration in plasma, is a good surrogate of drug
concentration at the target in adipose tissue. The absolute in
vitro and in vivo potencies are within 3-fold of each other in
each case, suggesting only 1-fold free cover of the in vitro IC50
is required to demonstrate target engagement in vivo.
Figure 6. Crystal structure of 26 bound to the human isoform of 11β-
We then evaluated 26 against glycemic and body weight
efficacy end points in a murine disease model. Although 26 did
demonstrate a significant lowering in fasting plasma glucose
and body weight efficacy (Figure 8), this was achieved at a high
dose (300 mg/kg/day) where unbound plasma concentrations
of compound 26 were approximately 80 times the in vitro
(HTRF) IC50. A separate study (unpublished results), in
collaboration with Anne White's group at the University of
Manchester, has shown that compound 26 gave body weight
and glucose efficacy at high dose (300 mg/kg/day). However,
lowering the dose to 50 mg/kg/day resulted only in body
weight efficacy without a significant lowering of fasting glucose
despite providing >90% target engagement in liver and ∼30
fold free cover over the in vitro IC50 in plasma.
HSD1 (4bb5).
support any particular conformation. As can be seen by
comparing Figure 6 to Figure 3, the core of the ligand and the
binding site side chains are essentially in the same positions,
supporting our previous approach of using the pyridine
structure for compound 3 to design changes to the pyrimidine
series.
PK Properties and PK/PD Relationship of 20 and 26.
Compounds 20 and 26 were profiled in vivo and demonstrated
suitable PK properties in mouse and rat at low dose (Table 3)
to proceed to PK/PD studies. Good oral exposure was seen in
mouse and rat. The clearance in both species for both
compounds was relatively low considering the high plasma
protein binding (PPB) free levels seen in vitro (i.e., low
unbound clearance). The relatively short half-lives observed
were advantageous as commented previously, to investigate the
effects of continuous cover versus episodic dosing. We also
profiled both compounds in dog to determine the suitability of
20 and 26 as potential clinical candidates. Human PPB was low
for both compounds (52% free for 20 and 38% free for 26) and
both compounds were stable to human hepatocytes (CLint < 2
μL/min/106).
CONCLUSIONS
■
Free-Wilson and structure-based design approaches were taken
to co-optimizing potency against human and murine isoforms
of 11β-HSD1, within the (5-hydroxy-2-adamantyl)-pyrimidine-
5-carboxamide series. The conclusions from both approaches
helped to inform an efficient design based on picking a limited
number of lipophilic, low molecular weight R1 groups with a
more expansive exploration of R2 secondary amines. This work
identified a number of neutral compounds that were potent
against both isoforms in the desired logD7.4 range (1−3) and
led to the identification of compounds 20 and 26, which have
good PK in mouse and rat, and successfully demonstrated
murine/rat target engagement in ex vivo assays. We then
evaluated 26 against body weight and glycemic end points in
Compound 20 was dosed orally to C57 male mice and Han
Wistar male rats to study its capacity to inhibit 11β-HSD1
activity in adipose tissue, and a similar study was performed
with 26 in C57 male mice. Dose−response experimental
designs were carried out with doses of 20 up to 5 (mouse) and
60 mg/kg (rat) and with doses of 26 up to 35 mg/kg (mouse).
Table 3. PK Properties of 20 and 26
d
ex.
species
iv/po dose (μmol/kg) iv CL (mL/min/kg) iv t1/2 (h) iv Vss (L/kg) po AUC (μM h) F (%) PPB % free Hep CLint (μL/min/106)
a
20
mouse
4.7/11.7
4.7/11.7
1.0/1.2
1.0/1.2
4.7/11.7
4.7/7.0
0.6/1.2
0.6/1.2
42
32
19
29
57
9
0.4
0.6
0.8
0.7
0.7
2.4
0.6
0.6
1.1
0.9
1.1
1.6
2.4
1.0
0.7
0.6
2.7
57
100
34
>50
55
3.6
<2
b
rat
10.4
c
dog m
0.38
>66
>66
30
<2
<2
c
dog f
0.16
1.5
23
a
26
mouse
34
9.1
<2
6.7
6.7
b
rat
5.6
61
35
c
dog m
17
14
0.19
0.42
15
38
c
dog f
30
38
a
b
Dosed (suspension for po and solution for iv) to male mice; drug concentrations measured in whole blood. Dosed (suspension for po and solution
c
for iv) to male Han Wistar rats; drug concentrations measured in whole blood. Dosed as solution to Beagles (m = male, and f = female); drug
concentrations measured in plasma. In vitro median CLint observed with hepatocytes from the appropriate species.
d
10658
dx.doi.org/10.1021/jm3013163 | J. Med. Chem. 2012, 55, 10652−10661

Journal of Medicinal Chemistry
Article
Figure 7. PK/PD relationship between the ex vivo activity of the 11β-HSD1 enzyme in epididymal adipose tissue and the drug concentrations of (a)
20 in male C57 mice, (b) 20 in male Han Wistar rats, and (c) 26 in male C57 mice.
5715) were used for TLC analysis. Preparative HPLC was performed
on C18 reversed-phase silica on a Waters or Phenomenex preparative
reversed-phase column using decreasingly polar mixtures of water
(containing 1% formic acid or 1% aqueous NH4OH) and MeCN. All
reactions were performed under nitrogen unless otherwise stated.
Methyl 2-(Cyclobutanecarbonyl)-3-(dimethylamino)acrylate
(4b). DMF dimethyl acetal (5.62 mL, 42.3 mmol) was added in one
portion to methyl 3-cyclobutyl-3-oxopropanoate (5.5 g, 35.2 mmol) in
1,4-dioxane (50 mL) at rt. The resulting solution was stirred at 100 °C
for 4 h. The reaction mixture was concentrated, and the resulting
crude product was purified by flash column chromatography (FCC),
eluting with 50−80% EtOAc/isohexane, to afford the title compound
Table 4. Comparison of in Vitro Potency to Ex Vivo Potency
from the in Vivo PK/PD Study
a
b
ex.
species
in vitro IC50 (μM)
in vivo IC50 (μM)
20
mouse
rat
0.027
0.271
0.032
0.030
0.087
0.050
26
mouse
a
Measured in HTRF assay (see the Supporting Information for
b
details). Measured ex vivo, using epididymal adipose tissue samples
3
from the in vivo PK/PD study as the conversion of H-cortisone to
3H-cortisol.
1
(4.60 g, 62%). H NMR (CDCl3): 1.72−1.82 (1H, m), 1.85−1.97
(1H, m), 2.06−2.13 (2H, m), 2.18−2.29 (2H, m), 3.02 (6H, s), 3.68−
3.75 (1H, m), 3.73 (3H, s), 7.62 (1H, s). m/z [M + Na]+ = 234.
HPLC tR = 1.42 min.
murine disease models, where it demonstrated glucose and
body weight efficacy at 300 mg/kg/day but only body weight
efficacy at 50 mg/kg/day, despite providing >90% target
engagement in the liver at both doses.
Methyl 4-Cyclobutyl-2-(methylthio)pyrimidine-5-carboxy-
late (5b). 2-Methyl-2-thiopseudourea sulfate (1.93 g, 13.9 mmol)
was added to methyl 2-(cyclobutanecarbonyl)-3-(dimethylamino)-
acrylate 4b (2.6 g, 12.3 mmol) and sodium acetate (4.24 g, 51.7
mmol) in DMF (10 mL) at rt. The resulting solution was stirred at 80
°C for 2 h. Water was added to the cooled solution. The reaction
mixture was diluted with EtOAc and washed with water. The organic
layer was dried over MgSO4, filtered, and evaporated to afford crude
product. The crude product was purified by FCC, elution gradient 5−
30% EtOAc in isohexane to afford the title compound (1.30 g, 44%) as
a white solid. 1H NMR (CDCl3): 1.86−1.94 (1H, m), 2.00−2.10 (1H,
m), 2.26−2.35 (2H, m), 2.41−2.51 (2H, m), 2.65 (3H, s), 3.90 (3H,
s), 4.35 (1H, qn), 8.86 (1H, s). m/z MH+ = 239. HPLC tR = 2.75 min.
4-Cyclobutyl-2-(methylthio)pyrimidine-5-carboxylic Acid. A
solution of LiOH.H2O (0.458 g, 10.9 mmol) in water (8 mL) was
added to a stirred solution of methyl 4-cyclobutyl-2-(methylthio)-
EXPERIMENTAL SECTION
■
All solvents and chemicals used were reagent grade. Flash column
chromatography was carried out using prepacked silica cartridges from
Redisep, Biotage, or Crawford and eluted using an Isco Companion
system. Purity and characterization of compounds were established by
a combination of LC-MS and NMR analytical techniques and was
>95% for all compounds. 1H NMR were recorded on a Bruker Avance
DPX400 (400 MHz) and were determined in DMSO-d6, unless stated
otherwise. 13C NMR spectra were recorded at 101 or 175 MHz.
Chemical shifts are reported in ppm relative to tetramethylsilane
(TMS) (0 ppm) or solvent peaks as the internal reference. All MS data
were obtained using electrospray. Merck precoated thin-layer
chromatography (TLC) plates (silica gel 60 F254, 0.25 mm, art.
Figure 8. Fasting plasma glucose and body weight changes of lean mice, obese mice, and obese mice treated with compound 26.
10659
dx.doi.org/10.1021/jm3013163 | J. Med. Chem. 2012, 55, 10652−10661

Journal of Medicinal Chemistry
Article
pyrimidine-5-carboxylate 5b (1.3 g, 5.46 mmol) in THF (16 mL) at rt.
The resulting mixture was stirred at rt for 24 h. The THF was
evaporated, and the resulting aqueous layer was washed with EtOAc,
then was acidified with 1 M aqueous citric acid, and extracted with
EtOAc. The organic layer was washed with saturated brine, dried over
MgSO4, filtered, and evaporated to afford the title compound (1.10 g,
(dimethylamino)acrylate 4c (114 g, 506 mmol) and NaOAc (166 g,
2024 mmol) in DMF (500 mL). The reaction mixture was stirred at
80 °C for 2 h, then at rt for 14 h, and then was partially evaporated.
The reaction mixture was diluted with EtOAc and washed with water.
The organic layer was dried over MgSO4, filtered, and evaporated to
afford crude product. The crude product was purified by FCC, eluting
with 0−20% EtOAc in isohexane to afford the title compound (76 g,
60%) as a cream solid. 1H NMR (CDCl3): 1.67−1.72 (2H, m), 1.79−
1.92 (4H, m), 1.99−2.05 (2H, m), 2.58 (3H, s), 3.91 (3H, s), 3.99−
4.09 (1H, m), 8.85 (1H, s). m/z MH+ = 253. HPLC tR = 3.04 min.
Methyl 4-Cyclopentyl-2-(methylsulfonyl)pyrimidine-5-car-
boxylate. Oxone (48.7 g, 79.3 mmol) was added to methyl 4-
cyclopentyl-2-(methylthio)pyrimidine-5-carboxylate 5c (10 g, 39.6
mmol) in acetonitrile (75 mL) and water (75 mL). The reaction
mixture was stirred at rt for 16 h and then was concentrated, saturated
aqueous NaHCO3 was added dropwise, and the mixture was extracted
with DCM. The organic layers were combined, dried over MgSO4,
filtered, and concentrated to afford the title compound (9.70 g, 86%).
1H NMR: 1.67 (2H, m), 1.81 (4H, m), 2.02 (2H, m), 3.43 (3H, s),
1
90%) as a white solid. H NMR (CDCl3): 1.87−1.96 (1H, m), 2.02−
2.13 (1H, m), 2.31−2.39 (2H, m), 2.44−2.54 (2H, m), 2.67 (3H, s),
4.42 (1H, qn), 9.00 (1H, s), no CO2H peak observed. m/z MH+ =
225. HPLC tR = 0.82 min.
4-Cyclobutyl-N-((2s,5r)-5-hydroxy-2-adamantyl)-2-methyl-
sulfanylpyrimidine-5-carboxamide (8b). i-Pr2NEt (3.39 mL, 19.6
mmol) was added to 4-cyclobutyl-2-(methylthio)pyrimidine-5-carbox-
ylic acid (1.1 g, 4.90 mmol), trans-4-aminoadamantan-1-ol hydro-
chloride (1.10 g, 5.40 mmol), and HATU (2.24 g, 5.89 mmol) in DMF
(20 mL) at rt. The reaction mixture was stirred at rt for 24 h, then was
evaporated to dryness, taken up in EtOAc. and washed sequentially
with water and saturated brine. The organic layer was dried over
MgSO4, filtered, and evaporated to afford crude product. The crude
product was purified by FCC, eluting with 0−6% MeOH in DCM, to
3.92 (1H, m), 3.94 (3H, s), 9.24 (1H, s). m/z MH+ = 285. HPLC tR =
2.08 min.
1
afford the title compound (1.50 g, 82%) as a white solid. H NMR
(CDCl3): 1.55−1.62 (2H, m), 1.66−1.71 (2H, m), 1.78−1.85 (5H,
m), 1.91−1.97 (3H, m), 2.00−2.08 (1H, m), 2.15−2.19 (1H, m),
2.23−2.32 (4H, m), 2.43−2.52 (2H, m), 2.62 (3H, s), 3.92−4.00 (1H,
m), 4.17−4.22 (1H, m), 5.90 (1H, d), 8.41 (1H, s). m/z MH+ = 374.
HPLC tR = 2.00 min.
(S)-Methyl 4-cyclopentyl-2-(tetrahydrofuran-3-ylamino)-
pyrimidine-5-carboxylate. i-Pr2NEt (4.91 mL, 28.1 mmol) was
added to methyl 4-cyclopentyl-2-(methylsulfonyl)pyrimidine-5-car-
boxylate (4 g, 14.1 mmol) and (S)-tetrahydrofuran-3-amine hydro-
chloride (2.61 g, 21.1 mmol) in THF (50 mL). The resulting solution
was stirred at rt for 16 h. The reaction mixture was concentrated, taken
up in DCM, and washed sequentially with water and saturated aqueous
NaHCO3. The organic layer was dried over MgSO4, filtered, and
evaporated to afford the title compound (2.55 g, 62%). 1H NMR: 1.67
(2H, m), 1.82 (4H, m), 1.97 (3H, m), 2.20 (1H, m), 3.61 (1H, m),
3.77 (1H, q), 3.82 (3H, s), 3.97 (3H, m), 4.46 (1H, m), 8.08 (1H, d),
8.72 (1H, s). m/z MH+ = 292. HPLC tR = 2.30 min.
(S)-4-Cyclopentyl-2-(tetrahydrofuran-3-ylamino)pyrimidine-
5-carboxylic Acid. Two molar aqueous NaOH (8.72 mL, 17.4
mmol) was added to a solution of (S)-methyl 4-cyclopentyl-2-
(tetrahydrofuran-3-ylamino)pyrimidine-5-carboxylate (2.54 g, 8.72
mmol) in methanol (40 mL) at rt. The resulting mixture was stirred
at 40 °C for 16 h, and then, the reaction was stirred at 55 °C for 4 h.
The reaction mixture was concentrated, taken up in water, and washed
with EtOAc. The aqueous layer was acidified with 1 M aqueous citric
acid, and the resulting solid was washed with cold water and dried
under vacuum to afford the title compound (2.30 g, 95%) as a white
solid. 1H NMR: 1.60 (2H, m), 1.75 (4H, m), 1.88 (3H, m), 2.14 (1H,
m), 3.54 (1H, m), 3.71 (1H, q), 3.85 (2H, m), 4.06 (1H, m), 4.40
(1H, m), 7.92 (1H, d), 8.64 (1H, s), 12.42 (1H, s). m/z MH+ = 278.
HPLC tR = 1.75 min.
4-Cyclobutyl-N-((2s,5r)-5-hydroxy-2-adamantyl)-2-methyl-
sulfonylpyrimidine-5-carboxamide. 3-Chloroperoxybenzoic acid
(70%) (0.937 g, 3.80 mmol) was added in one portion to 4-cyclobutyl-
N-(5-hydroxy-2-adamantyl)-2-methylsulfanylpyrimidine-5-carboxa-
mide 8b (0.71 g, 1.90 mmol) in DCM (35 mL) at 0 °C. The reaction
mixture was allowed to warm to rt, was stirred at rt for 24 h, and then
was taken up in DCM. The organic layer was washed sequentially with
saturated aqueous NaHCO3, 2 M aqueous NaOH, and saturated brine,
dried over MgSO4, filtered, and evaporated to afford crude product.
The crude product was purified by FCC, eluting with 0−6% MeOH in
DCM, to afford the title compound (0.560 g, 73%) as a white solid. 1H
NMR (CDCl3): 1.44 (1H, s), 1.58−1.65 (2H, m), 1.74−1.87 (6H, m),
1.93−1.98 (3H, m), 2.05−2.15 (1H, m), 2.18−2.30 (3H, m), 2.32−
2.39 (2H, m), 2.43−2.55 (2H, m), 3.34 (3H, s), 4.00−4.09 (1H, m),
4.21−4.28 (1H, m), 6.42 (1H, d), 8.71 (1H, s). m/z MH+ = 406.
HPLC tR = 1.59 min.
4-Cyclobutyl-N-[(2r,5s)-5-hydroxyadamantan-2-yl]-2-(tetra-
hydro-2H-pyran-4-ylamino)pyrimidine-5-carboxamide (20). 4-
Cyclobutyl-N-((2s,5r)-5-hydroxy-2-adamantyl)-2-methylsulfonylpyri-
midine-5-carboxamide (270 mg, 0.67 mmol) and tetrahydro-2H-
pyran-4-amine (269 mg, 2.66 mmol) were dissolved in THF (4 mL)
and sealed into a microwave tube. The reaction was heated to 150 °C
for 1 h and cooled to rt. The reaction mixture was diluted with DCM
and stirred with saturated aqueous NaHCO3 before passing through a
phase separation cartridge. The organic layer was evaporated to afford
crude product, which was purified by preparative HPLC to afford the
4-Cyclopentyl-N-(5-hydroxy-2-adamantyl)-2-[[(3S)-oxolan-3-
yl]amino]pyrimidine-5-carboxamide (26). HATU (45.1 g, 118.6
mmol) was added to (S)-4-cyclopentyl-2-(tetrahydrofuran-3-ylamino)-
pyrimidine-5-carboxylic acid (32.9 g, 118.6 mmol) and i-Pr2NEt (41.4
mL, 237.3 mmol) in DMF (300 mL). The resulting solution was
stirred at rt for 15 min. trans-4-Aminoadamantan-1-ol hydrochloride
(29.0 g, 142.4 mmol) was added, and the reaction mixture was stirred
at rt overnight, then evaporated, taken up in EtOAc, and washed with
water and saturated brine. The organic layer was dried over MgSO4,
filtered, and evaporated to afford the crude product, which was purified
by FCC, eluting with 0−10% MeOH in EtOAc, then recrystallized
from IPA. The resulting solid was suspended in water and heated at
reflux for 24 h (to remove residual IPA in crystal lattice), then filtered,
and dried under vacuum to afford the title compound (25.8 g, 51%) as
1
title compound (169 mg, 60%) as a white solid. H NMR: 1.31 (2H,
d), 1.48−1.58 (2H, m), 1.61 (4H, m), 1.68−2.01 (11H, m), 2.10 (2H,
q), 2.23−2.32 (2H, m), 3.35−3.41 (2H, m), 3.86 (4H, m), 3.98 (1H,
m), 4.37 (1H, s), 7.31 (1H, d), 7.92 (1H, d), 8.13 (1H, s). m/z MH+ =
427. HPLC tR = 1.67 min.
Methyl 2-(Cyclopentanecarbonyl)-3-(dimethylamino)-
acrylate (4c). DMF dimethyl acetal (3.28 mL, 24.7 mmol) was
added in one portion to methyl 3-cyclopentyl-3-oxopropanoate (3.50
g, 20.6 mmol) in 1,4-dioxane (40 mL) at rt. The resulting solution was
stirred at 100 °C for 4 h. The reaction mixture was evaporated to
afford crude product. The crude product was purified by FCC, eluting
with 50−80% EtOAc in isohexane, to afford the title compound (4.50
g, 97%) as a yellow oil. 1H NMR: 1.45−1.73 (8H, m), 2.81−2.86 (1H,
m), 2.95 (6H, s), 3.62 (3H, s), 7.57 (1H, s). m/z MH+ = 226. HPLC
tR= 1.66 min.
1
a white solid. H NMR: 1.29−1.38 (2H, m), 1.54−1.64 (6H, m),
1.65−2.05 (14H, m), 2.09−2.17 (1H, m), 3.38−3.48 (1H, m), 3.50−
3.53 (1H, m), 3.67−3.73 (1H, m), 3.81−3.85 (1H, m), 3.86−3.91
(2H, m), 4.30−4.42 (1H, m), 4.38 (1H, s), 7.42−7.54 (1H, m), 8.00−
8.07 (1H, m), 8.14 (1H, s). m/z MH+ = 427. HPLC tR = 1.69 min.
See the Supporting Information for the preparation of other
representative examples and characterization for all additional
examples.
Methyl 4-Cyclopentyl-2-(methylthio)pyrimidine-5-carboxy-
late (5c). 2-Methyl-2-thiopseudourea hemisulfate (84 g, 607 mmol)
was added in one portion to methyl 2-(cyclopentanecarbonyl)-3-
10660
dx.doi.org/10.1021/jm3013163 | J. Med. Chem. 2012, 55, 10652−10661

Journal of Medicinal Chemistry
Article
Whittamore, P. R. O. Optimisation of Pharmacokinetic Properties in
a Neutral Series of 11β-HSD1 Inhibitors. Bioorg. Med. Chem. Lett.
2012, 22, 6756−6761.
ASSOCIATED CONTENT
* Supporting Information
■
S
Preparation and additional characterization for final com-
pounds, enzymatic assay procedures, procedures and parame-
ters for X-ray crystallography, data set used for Free-Wilson
analysis and model coefficients for R1/R2 groups, ex vivo assay
and in vivo protocols for PD/efficacy studies, PK/PD
parameters for model fitting, and ADMET references. This
material is available free of charge via the Internet at http://
pubs.acs.org.
(7) Free, S. M., Jr.; Wilson, J. W. A mathematical contribution to
structure-activity studies. J. Med. Chem. 1964, 7, 395−399.
(8) Freeman-Cook, K. D.; Amor, P.; Bader, S.; Buzon, L. M.; Coffey,
S. B.; Corbett, J. W.; Dirico, K. J.; Doran, S. D.; Elliott, R. L.; Esler, W.;
Guzman-Perez, A.; Henegar, K. E.; Houser, J. A.; Jones, C. S.;
Limberakis, C.; Loomis, K.; McPherson, K.; Murdande, S.; Nelson, K.
L.; Phillion, D.; Pierce, B. S.; Song, W.; Sugarman, E.; Tapley, S.; Tu,
M.; Zhao, Z. Maximizing lipophilic efficiency: The use of free-wilson
analysis in the design of inhibitors of acetyl-CoA carboxylase. J. Med.
Chem. 2012, 55, 935−942.
(9) Patel, Y.; Gillet, V. J.; Howe, T.; Pastor, J.; Oyarzabal, J.; Willett,
P. Assessment of Additive/Nonadditive Effects in Structure-Activity
Relationships: Implications for Iterative Drug Design. J. Med. Chem.
2008, 51, 7552−7562.
(10) Thomas, M. P.; Potter, B. V. L. Crystal structures of 11β-
hydroxysteroid dehydrogenase type 1 and their use in drug discovery.
Future Med. Chem. 2011, 3, 367−390.
(11) Walker, E. A.; Clark, A. M.; Hewison, M.; Ride, J. P.; Stewart, P.
M. Functional Expression, Characterization, and Purification of the
Catalytic Domain of Human 11-beta-Hydroxysteroid Dehydrogenase
Type 1. J. Biol. Chem. 2001, 276, 21343−21350.
(12) Maser, E.; Voelker, B.; Friebertshaeuser, J. 11β-Hydroxysteroid
Dehydrogenase Type 1 from Human Liver: Dimerization and Enzyme
Cooperativity Support Its Postulated Role as Glucocorticoid
Reductase. Biochemistry 2002, 41, 2459−2465.
(13) Hosfield, D. J.; Wu, Y.; Skene, R. J.; Hilgers, M.; Jennings, A.;
Snell, G. P.; Aertgeerts, K. Conformational Flexibility in Crystal
Structures of Human 11{beta}-Hydroxysteroid Dehydrogenase Type I
Provide Insights into Glucocorticoid Interconversion and Enzyme
Regulation. J. Biol. Chem. 2005, 280, 4639−4648.
(14) Favia, A. D.; Masetti, M.; Recanatini, M.; Cavalli, A. Substrate
Binding Process and Mechanistic Functioning of Type 1 11β-
Hydroxysteroid Dehydrogenase from Enhanced Sampling Methods.
PLoS ONE 2011, 6, e25375.
Accession Codes
The crystal structure of human 11β-HSD in complex with
compounds 3 and 26 has been deposited in the Protein Data
Bank (PDB accession codes 4bb6 and 4bb5).
AUTHOR INFORMATION
Corresponding Author
*Tel: +44(0)1625 519064. E-mail: frederick.goldberg@
astrazeneca.com.
Notes
■
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
We thank Joanne deSchoolmeester for generating high-quality
enzyme data, Mark Denn for generation of high-quality PK
data, Brendan Leighton for his advice and leadership of the
bioscience strategy, Patrik Johansson and Stefan Gerhardt for
generation of crystal structures, and Adrian Gill and Martin
Wild for their support and leadership of the project.
ABBREVIATIONS
■
c-Pr, cyclopropyl; c-Bu, cyclobutyl; c-Pn, cyclopentyl; FCC,
flash column chromatography; HSD, hydroxysteroid dehydro-
genase; HATU, 2-(3H-[1,2,3]triazolo[4,5-b]pyridin-3-yl)-
1,1,3,3-tetramethylisouronium hexafluorophosphate(V);
HMDS, hexamethyldisilazane; LLE, ligand lipophilicity effi-
ciency (pIC50-logD7.4)
(15) Zhang, J.; Osslund, T. D.; Plant, M. H.; Clogston, C. L.; Nybo,
R. E.; Xiong, F.; Delaney, J. M.; Jordan, S. R. Crystal Structure of
Murine 11β-Hydroxysteroid Dehydrogenase 1: An Important
Therapeutic Target for Diabetes‡. Biochemistry (N. Y.) 2005, 44,
6948−6957.
(16) Maestro; Schrodinger, LLC: New York, NY, 2011; 9.2.
̈
(17) Waring, M. J. Lipophilicity in drug discovery. Expert Opin. Drug
REFERENCES
■
Discovery 2010, 5, 235−248.
(1) Ge, R.; Huang, Y.; Liang, G.; Li, X. 11β-Hydroxysteroid
dehydrogenase type 1 inhibitors as promising therapeutic drugs for
diabetes: status and development. Curr. Med. Chem. 2010, 17, 412−
422.
(2) Morgan, S. A.; Tomlinson, J. W. 11β-Hydroxysteroid
dehydrogenase type 1 inhibitors for the treatment of type 2 diabetes.
Expert Opin. Invest. Drugs 2010, 19, 1067−1076.
(3) Masuzaki, H.; Yasue, S.; Nakao, K. Molecular target of the
metabolic syndrome. 11β-HSD1 an adipocyte function. Card. Pract.
2005, 16, 179−186.
(18) Waring, M. J. Defining optimum lipophilicity and molecular
weight ranges for drug candidates-Molecular weight dependent lower
logD limits based on permeability. Bioorg. Med. Chem. Lett. 2009, 19,
2844−2851.
(19) Kotelevtsev, Y.; Brown, R. W.; Fleming, S.; Kenyon, C.;
Edwards, C. R. W.; Seckl, J. R.; Mullins, J. J. Hypertension in mice
lacking 11β-hydroxysteroid dehydrogenase type 2. J. Clin. Invest. 1999,
103, 683−689.
(20) Veniant, M. M.; Hale, C.; Komorowski, R.; Chen, M. M.; St., J.,
D. J.; Fotsch, C.; Wang, M. Time of the day for 11β-HSD1 inhibition
plays a role in improving glucose homeostasis in DIO mice. Diabetes,
Obes. Metab. 2009, 11, 109−117.
(4) Anagnostis, P.; Athyros, V. G.; Tziomalos, K.; Karagiannis, A.;
Mikhailidis, D. P. The pathogenetic role of cortisol in the metabolic
syndrome: a hypothesis. J. Clin. Endocrinol. Metab. 2009, 94, 2692−
2701.
(5) Scott, J. S.; Bowker, S. S.; deSchoolmeester, J.; Gerhardt, S.;
Hargreaves, D.; Kilgour, E.; Lloyd, A.; Mayers, R. M.; McCoull, W.;
Newcombe, N. J.; Ogg, D.; Packer, M. J.; Rees, A.; Revill, J.; Schofield,
P.; Selmi, N.; Swales, J. G.; Whittamore, P. R. O. Discovery of a
Potent, Selective, and Orally Bioavailable Acidic 11β-Hydroxysteroid
Dehydrogenase Type 1 (11β-HSD1) Inhibitor: Discovery of 2-[(3S)-
1-[5-(Cyclohexylcarbamoyl)-6-propylsulfanylpyridin-2-yl]-3-
piperidyl]acetic Acid (AZD4017). J. Med. Chem. 2012, 55, 5951−
5964.
(6) Scott, J. S.; Gill, A. L.; Godfrey, L.; Groombridge, S. D.; Rees, A.;
Revill, J.; Schofield, P.; Sorme, P.; Stocker, A.; Swales, J. G.;
̈
10661
dx.doi.org/10.1021/jm3013163 | J. Med. Chem. 2012, 55, 10652−10661