Synthesis, biological evaluation, and 3D QSAR study of 2-methyl-4-oxo-3-oxetanylcarbamic acid esters as N -acylethanolamine acid amidase (NAAA) inhibitors

Article abstract of DOI:10.1021/jm501455s

N-(2-Oxo-3-oxetanyl)carbamic acid esters have recently been reported to be noncompetitive inhibitors of the N-acylethanolamine acid amidase (NAAA) potentially useful for the treatment of pain and inflammation. In the present study, we further explored the

Full text of DOI:10.1021/jm501455s

Subscriber access provided by Princeton University Library
Article
Synthesis, Biological Evaluation, and 3D QSAR Study
of 2-Methyl-4-oxo-3-oxetanylcarbamic Acid Esters as
N-Acylethanolamine Acid Amidase (NAAA) Inhibitors
Stefano Ponzano, Anna Berteotti, Rita Petracca, Romina Vitale, Luisa Mengatto, Tiziano
Bandiera, Giovanni Bottegoni, Fabio Bertozzi, Daniele Piomelli, and Andrea Cavalli
J. Med. Chem., Just Accepted Manuscript • DOI: 10.1021/jm501455s • Publication Date (Web): 07 Nov 2014
Downloaded from http://pubs.acs.org on November 14, 2014
Just Accepted
“Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted
online prior to technical editing, formatting for publication and author proofing. The American Chemical
Society provides “Just Accepted” as a free service to the research community to expedite the
dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts
appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been
fully peer reviewed, but should not be considered the official version of record. They are accessible to all
readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered
to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published
in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just
Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor
changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers
and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors
or consequences arising from the use of information contained in these “Just Accepted” manuscripts.
Journal of Medicinal Chemistry 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 55
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 2. A: The overlay of the training set molecules superimposed to compound 2; B: Correlation plot
between pIC50exp and pIC50pred; C: In licorice the 3D structure of 2 is reported. The transparent blobs are
the equipotential contours of the lipophilic 3D QSAR APF component. The grey region represents the ideal
lipophilic shape for NAAA inhibition. D: In green is reported the 3D structure of 20 superimposed to
compound 2, in orange. The grey blobs represent the equipotential contours of the Sp2 hybridization 3D
QSAR APF component. The occupancy of such regions is detrimental for the potency. E: In yellow is reported
1
6 superimposed to compound 2, in orange. The grey blobs represent the equipotential contours of the size
D QSAR APF component. The occupancy of such regions is detrimental for the potency.
77x123mm (300 x 300 DPI)
3
1
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 2 of 55
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 3: The molecules selected for the validation of the 3D QSAR model are shown in yellow, as licorice,
superimposed to 2 in pink. A: molecule 9r, B: molecule 9s, C: molecule 9t, D: molecule 9v, E: molecule 9u.
177x123mm (300 x 300 DPI)
ACS Paragon Plus Environment

Page 3 of 55
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
Correlation plot between the experimental pIC50 and the leave-one-out cross-validated pIC50.
1
77x108mm (300 x 300 DPI)
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 4 of 55
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
Synthesis, Biological Evaluation, and 3D QSAR Study
of 2ꢀMethylꢀ4ꢀoxoꢀ3ꢀoxetanylcarbamic Acid Esters as
NꢀAcylethanolamine Acid Amidase (NAAA) Inhibitors
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
†,#
†,#
†
†
†
Stefano Ponzano, Anna Berteotti, Rita Petracca, Romina Vitale, Luisa Mengatto, Tiziano
†
†,
†,
†,
§
†,
‡
Bandiera, Giovanni Bottegoni, * Fabio Bertozzi, * Daniele Piomelli, Andrea Cavalli
†
Drug Discovery and Development, Istituto Italiano di Tecnologia, Via Morego 30, Iꢀ16163 Genova,
Italy
§
Departments of Anatomy and Neurobiology, Pharmacology and Biological Chemistry, University of
California, Irvine, California 92697ꢀ4621, United States
‡
Department of Pharmacy and Biotechnology, University of Bologna, Via Belmeloro 6, Iꢀ40126
Bologna, Italy
1
ACS Paragon Plus Environment

Page 5 of 55
Journal of Medicinal Chemistry
1
2
3
Abstract
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
Nꢀ(2ꢀoxoꢀ3ꢀoxetanyl)carbamic acid esters have recently been reported to be noncompetitive inhibitors
of the Nꢀacylethanolamine acid amidase (NAAA) potentially useful for the treatment of pain and
inflammation. In the present study, we further explored the structure activity relationships of the
carbamic acid ester side chain of 2ꢀmethylꢀ4ꢀoxoꢀ3ꢀoxetanylcarbamic acid ester derivatives. Additional
favourable features in the design of potent NAAA inhibitors have been found together with the
identification of a single digit nanomolar inhibitor. In addition, we devised a 3D QSAR using the
Atomic Property Field method. The model turned out to be able to account for the structural variability,
and was prospectively validated by designing, synthesizing, and testing novel inhibitors. The fairly
good agreement between predictions and experimental potency values points to this 3D QSAR model
as the first example of quantitative structure activity relationships in the field of NAAA inhibitors.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
2
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 6 of 55
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
Introduction
The fattyꢀacid ethanolamides (FAEs) are a class of multiꢀfunctional lipid mediators that have attracted
significant attention due to their potential role in physiological and pathological processes including
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
ꢀ3
4
5
6, 7
pain, innate immunity, reward regulation and feeding.
In particular, palmitoylethanolamide
(PEA), the endogenous amide of palmitic acid and ethanolamine, has been shown to inhibit peripheral
8
, 9
inflammation and mast cell degranulation and to exhibit antinociceptive properties in rat and mouse
10ꢀ12
models of acute and chronic pain.
Furthermore PEA has been indicated to attenuate skin
1
3, 14
15
inflammation
and neuropathic pain in humans. These effects are mainly attributed to the ability of
PEA to activate peroxisome proliferatorꢀactivated receptorꢀα (PPARꢀα), a member of the
1
, 2, 7, 16
steroid/nuclear receptor superfamily.
PEA is deactivated into free palmitic acid and ethanolamine by two intracellular lipid amidases: fatty
1
7
acid amide hydrolase (FAAH), which prefers anandamide over PEA and OEA, and Nꢀ
1
8ꢀ20
acylethanolamine acid amidase (NAAA), the main responsible for PEA cleavage.
In this light, selective NAAA inhibitors could represent a viable strategy to raise local levels of PEA,
1
, 2, 7
potentially leading to antiꢀinflammatory and analgesic effects via PPARꢀα signaling.
Only a
restricted number of NAAA inhibitors, belonging to a narrow range of chemical classes, has been
2
1, 22
reported so far.
Generally most of the first discovered compounds have been shown to block the
23ꢀ27
enzyme hydrolytic activity with micromolar potencies.
Rational drug discovery efforts aimed at
either identifying novel scaffolds or at improving already known ones have been severely hampered by
the limited amount of detailed structural information on this target.
3
ACS Paragon Plus Environment

Page 7 of 55
Journal of Medicinal Chemistry
1
2
3
20,
NAAA is a cysteine hydrolase that belongs to the Nꢀterminal nucleophile (Ntn) family of enzymes.
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
2
8, 29
Like other Ntn enzymes, NAAA is produced as an inactive proenzyme and is activated at acidic
3
0
pH by autocatalytic cleavage at a specific site of the peptide chain.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Recently, a comparative model of NAAA active site was generated starting from the coordinates of
3
1
conjugated bile acid hydrolase (CBAH). The insights gained from this model eventually led to the
discovery of the serineꢀderived 2ꢀoxoꢀ3ꢀoxetanylamide, (S)ꢀOOPP, as a first subꢀmicromolar NAAA
inhibitor. The model turned out to be consistent with both mutagenesis data, explaining the role of
several residues in the catalytic process, as well as with the SAR exploration performed around the
newly identified potent NAAA inhibitors, featuring an αꢀaminoꢀβꢀlactone ring as a reactive
3
2
electrophilic warhead.
Structureꢀactivity relationship (SAR) studies of αꢀaminoꢀβꢀlactone derivatives, as carbamic acid esters,
investigated the effects on NAAA inhibition of side chain modifications and the stereoꢀchemical
33ꢀ35
requirements of the introduction of a βꢀsubstitution.
These works led to the identification of
compounds that were highly potent at inhibiting both rat and human NAAA, such as βꢀlactones 1
3
3, 34, 36
34
(ARN077)
and 2 (Figure 1). The βꢀlactone ring of these compounds was proved to be
responsible for the inhibition mechanism, interacting covalently with the Nꢀterminal catalytic cysteine
36
(Cys126), through the formation of a thioester bond.
In the present study, we further extended the exploration of the SAR for this class of inhibitors focusing
on the carbamic acid ester side chain. Besides generating new compounds, a more extensive
investigation of the structural determinants for NAAA activity could also contribute to shed light on
some threeꢀdimensional features of NAAA binding pocket, thus facilitating future development of new
molecular scaffolds.
4
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 8 of 55
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
34
To this aim, combining previously reported results together with data on 17 new compounds
purposely synthesized to enhance the structural diversity of the set, we devised a threeꢀdimensional
Quantitative StructureꢀActivity Relationship (3D QSAR) model. Our model was built adopting the
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
37
recently reported Atomic Property Field (APF) method. APF can be described as a gridꢀbased model
in which the discrete features of a classic pharmacophore are replaced by continuous, regularly spaced
threeꢀdimensional lattices. Each molecule from our training set contributed to this composite set of
grids proportionally to its reported activity. With respect to more traditional approaches, the main
advantage of this technique is that the strength of the overlap between a new putative compound and
the APF signature of the template molecules can be directly translated into a prediction of activity.
Finally, our model was prospectively validated. Based on the predictions generated on a list of virtual
compounds, we selected 5 highꢀscoring derivatives to be actually synthesized and tested. Notably, the
activities of the prospectively tested compounds turned out to be in line with the predictions.
5
ACS Paragon Plus Environment

Page 9 of 55
Journal of Medicinal Chemistry
1
2
3
Chemistry
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
Novel 2ꢀmethylꢀ4ꢀoxoꢀ3ꢀoxetanylcarbamic acid esters 9a–v (Table 1 and Table 4) were synthesized
3
4, 35
using previously reported synthetic protocols (Scheme 1).
imidazole 1ꢀcarboxylates 4 or as a mixture of 2ꢀpyridyl carbonates 5 and 2ꢀoxopyridineꢀ1ꢀcarboxylates
. The desired final 2ꢀmethylꢀ4ꢀoxoꢀ3ꢀoxetanylcarbamic acid esters 9a–v were obtained either by
First, alcohols 3 were activated either as
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
6
reaction of (2S,3R)ꢀ2ꢀmethylꢀ4ꢀoxoꢀ3ꢀoxetanylammonium tolueneꢀ4ꢀsulfonate (7) with compounds 4–6
or by cyclization of intermediate αꢀsubstitutedꢀβꢀhydroxycarboxylic acids 8, synthesized from
commercially available Dꢀthreonine and the corresponding activated alcohols 4–6.
Most of the alcohols used in the above mentioned methodologies were commercially available or
obtained via reduction of the corresponding commercially available carboxylic acids. Alcohols 3c–e,p
were prepared as reported in Scheme 2. While the oxetane derivative 3c was synthesized in a
straightforward manner from oxetanꢀ3ꢀone (10) by reaction with butyl lithium, compound 3d was
obtained in a one pot procedure starting from 4ꢀbromobutylbenzene (11), which was first converted
into its lithium analogue and subsequently reacted with 10. The syntheses of alcohols 3e and 3p were
accomplished according to literature procedures. The commercially available 2ꢀmethylpropanoic acid
(12) was reacted with (2ꢀbromoethyl)ꢀbenzene using sodium hydride and in situ prepared lithium
3
8
diisopropylamine to afford the carboxylic acid 13. Lewis acid addition (boron trifluoride diethyl
39
etherate) to a mixture of ethyl 4ꢀhydroxybenzoate (14) and cyclohexene afforded the ester 15. Both
intermediates 13 and 15 were submitted to lithium aluminium hydride reduction to furnish the desired
alcohols 3e and 3p.
6
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 10 of 55
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
Results and Discussion
The main objectives of the present study were to i) perform an extended structural exploration
specifically tailored on the side chain of 2ꢀmethylꢀ4ꢀoxoꢀ3ꢀoxetanylcarbamic acid ester derivatives, ii)
exploit this structural variability to devise a 3D QSAR model, and iii) prospectively validate the model
by synthesizing novel active compounds in line with the predictions.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
SAR study
3
3, 34, 36
Starting from previously identified compound 1,
extensive SAR studies led to the discovery of
3
4
the first singleꢀdigit nanomolar inhibitor of intracellular rat (and human) NAAA activity 2. Here, a
new series of 2ꢀmethylꢀ4ꢀoxoꢀ3ꢀoxetanylcarbamic acid esters 9a–q (Table 1) were synthesized to verify
additional features, which could help confirm or better define the structural determinants for NAAA
activity. We therefore explored the effect on NAAA inhibition of different sideꢀchains taking into
account structural modifications, such as the introduction of heteroatoms, the length of the sideꢀchain,
and the diꢀsubstitution in α and β position with respect to the carbamic acid function. Finally, to further
define the contribution of conformationally constrained moieties, more rigid structures based on the
tridimensional shape of 2 were taken into consideration.
The potency of the new synthesized 2ꢀmethylꢀ4ꢀoxoꢀ3ꢀoxetanyl carbamic acid esters 9a–v was
evaluated by their ability to inhibit the hydrolysis of 10ꢀcisꢀheptadecenoylethanolamide (an unnatural
FAE) by native NAAA prepared from rat lungs. Median inhibitory concentration (IC ) values obtained
5
0
using the rat enzyme (rꢀNAAA) are reported in Table 1 and Table 4.
3
4
Previously described analogues (1,2 and 16–29) and their corresponding r-NAAA inhibitory activity
are reported in Table S1 (see Supporting Information).
7
ACS Paragon Plus Environment

Page 11 of 55
Journal of Medicinal Chemistry
1
2
3
The analysis of the IC50 values reported in Table 1 together with those of the previously reported
analogues (Table S1) allows some general considerations to be made. The introduction of a fluorine
atom in the para position of the phenyl ring (9a, IC = 0.24 ꢀM) was detrimental for the activity,
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
5
0
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
leading to a ca. 5ꢀfold drop in potency compared to 1 (Table S1). Taking into consideration possible
modifications on compound 22 (IC = 0.76 ꢀM, Table S1), a substitution of a methyl group with a
5
0
more lipophilic chlorine atom (9b) resulted in an increased inhibitory potency (IC = 0.28 ꢀM), while
5
0
the introduction of an oxetane moiety in αꢀposition with respect to the carbamate (9c) turned out to be
detrimental for activity (IC = 3.76 ꢀM).
50
The same negative effect of a diꢀsubstitution in the αꢀposition to the carbamate was observed in
analogue 9d (IC = 0.52 ꢀM) compared to 1, therefore indicating a limited space in the region of the
5
0
enzyme proximal to the carbamic function. A similar loss in activity was observed with the insertion of
a gemꢀdiꢀmethyl substitution in the βꢀposition as for compound 9e (IC = 3.43 ꢀM), which resulted 9ꢀ
5
0
fold less active than the corresponding unsubstituted analogue 20 (IC = 0.39 ꢀM, Table S1).
5
0
Whereas the presence of a terminal cyclohexyl moiety in a four carbon atoms aliphatic linker (9f)
retained approximately the same potency as for the previously reported 5 carbon atom aliphatic linker
analogue 28 (IC50 = 0.013 ꢀM, Table S1), a ca. 300ꢀfold loss in activity was observed when the
cyclohexyl moiety was directly connected to the carbamic acid function (9g, IC = 5.42 ꢀM).
5
0
The addition of a methylene unit to the aliphatic linker of 1, leading to a 6 carbon atom alkyl chain, was
reported to be detrimental for activity (21, IC = 1.17 ꢀM, Table S1), whereas the insertion of an
5
0
additional carbon atom (9h) surprisingly restored the original activity (IC50 = 0.039 ꢀM). A short
alkylic chain bearing a terminal p-methylphenyl group (9i, IC50 = 1.22 ꢀM) confirmed a low ꢀM
activity as previously observed for the phenethyl derivative (18, IC = 1.22 ꢀM, Table S1).
50
8
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 12 of 55
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
To further investigate and possibly confirm the importance of the rigidity of the side chain for potent
enzyme inhibition, new βꢀlactones structurally related to 2 were designed and tested against rꢀNAAA.
The substitution of the proximal phenyl ring with an acetylene functionality (9j), which retains the
overall shape of the side chain of 2, led to a 3ꢀfold drop in potency. The same variation in activity was
observed when substituting the terminal phenyl ring with a thiophene moiety (9k, IC = 0.022 ꢀM),
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
5
0
whereas the substitution with a cyclohexyl moiety (9l, IC50 = 0.007 ꢀM) resulted in a compound
equally active to 2 (Table S1). To the best of our knowledge, compounds 2 and 9l are the most active,
singleꢀdigit nanomolar NAAA inhibitors so far reported in literature.
We then evaluated the effect of small substituents at para position of the biphenyl moiety. No
significant variation in potency was observed for the pꢀfluoro derivative 9m (IC = 0.011 ꢀM), while a
5
0
methyl or trifluoromethyl group in the same position led to slightly less potent compounds (9n and 9o,
IC = 0.029 ꢀM and IC = 0.036 ꢀM, respectively). Interestingly, increasing the flexibility of the side
5
0
50
chain of compound 9l by insertion of an oxygen atom between the phenyl and cyclohexyl ring (9p) led
to only a 3ꢀfold decrease in activity (IC = 0.023 ꢀM). Finally, replacing the pꢀbiphenyl ring with a
50
more polar, fused bicyclic piperonyl moiety (9q) led to a 25ꢀfold drop in potency.
3
D QSAR model
We assembled a training set encompassing 33 compounds: 17 new, purposely synthesized 2ꢀmethylꢀ4ꢀ
oxoꢀ3ꢀoxetanylcarbamic acid esters derivatives (Table 1) and 16 analogues taken from a previous SAR
34
study (Table S1). The 3D QSAR model was built using the APF approach, namely a continuous 3D
3
7
pharmacophoric potential implemented on a set of regularly spaced grids. This multiꢀcomponent 3D
potential was generated solely by the training set compounds, each individual contribution reflected
preferences for various atomic properties at each point in space, and it was rescaled according to the
9
ACS Paragon Plus Environment

Page 13 of 55
Journal of Medicinal Chemistry
1
2
3
compound’s experimental potency. Multiple ligands consistently displaying atoms with similar
properties in the same location generated a strong pharmacophoric signal for these common features. In
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
3
7
line with the original APF 3DꢀQSAR implementation proposed by Totrov, we selected seven
properties to be analyzed and assigned: charge, hydrogen bond donor propensity, hydrogen bond
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
2
acceptor propensity, sp carbon atom hybridization, and lipophilicity, to encompass the classic
pharmacophoric features, plus two more unconventional properties, namely size and
electropositivity/negativity, to take into account more subtle differences among substituents that would
have been very difficult to capture otherwise.
First, the training set compounds were superimposed using, as template, the transition state model of
compound 2, as previously obtained by quantum chemical calculations (see Experimental Section for
3
4
details). In Figure 2A, the structural alignment of the training set compounds is reported. As
expected, the βꢀlactone ring perfectly matched. Conversely, the conformational arrangement of the side
chains captured the diversity of the training set and represented the truly informative core of our model.
For each molecule, a sevenꢀcomponent property vector was evaluated and used to generate the 3D
QSAR model. The optimal number of latent vectors to be used in the partial least square (PLS) analysis
was selected in order to optimize the overall squared correlation coefficient between the predicted and
2
2
the observed activities (R ) as well as the leaveꢀoneꢀout crossꢀvalidation results (Q ) on the training set.
2
2
In Table 2, R and Q values as a function of the number of latent vectors are reported. Eventually, we
decided to use 6 latent vectors meeting the generally accepted threshold of 1 latent vector for every 5
4
0
2
compounds employed in the training set. In this way, our model showed a good R coefficient
2
between the experimental pIC (pIC
) and the predicted pIC (pIC
), (R = 0.93, see Figure
5
0
50exp
50
50pred
2
2
B), and a good predictivity with a Q = 0.69 (see Figure S1).
1
0
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 14 of 55
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
In Table 3, the experimental, predicted, and leaveꢀoneꢀout crossꢀvalidated pIC values for the training
5
0
set are reported.
The structural implication of the generated 3D QSAR model can be visualized in terms of isopotential
contour maps of the various APF components. For example, the lipophilicity field of the model was
useful to pinpoint regions that react to the presence of a lipophilic moiety, enhancing or reducing the
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
inhibitory potency. As already reported, the size and the shape of the lipophilic tail are key
32ꢀ34
determinants for the inhibitory potency of the βꢀlactones compounds.
In Figure 2C, only the
regions in which a lipophilic moiety improved NAAA inhibition are reported. Interestingly, the
structural implications of the proposed model are in agreement with those proposed by Solorzano and
colleagues by means of mutagenesis and homology modeling studies (see also the covalent docking
3
1
studies and Figure S2 in Supporting Information). The shape of this lipophilic region suggests a
2
preference for linear moieties in the carbamic acid esters for optimal enzyme recognition. The sp
hybridization atomic property field generated by the training set is reported in Figure 2D. In particular,
2
the regions in which the presence of sp hybridized atoms was detrimental for the inhibitory potency
are shown. For example, a terminal cyclohexyl moiety in a four units aliphatic linker was preferred to a
terminal phenyl ring, as confirmed by the drop in potency of 20 with respect to compound 9f. In Figure
2E, the regions in which large substituents led to a drop in the inhibitory potency are reported. This
seems to suggest that the access to the lipophilic subpocket is quite tight and, as such, illꢀsuited to lodge
bulky substituents like the tertꢀbutyl group of compound 16.
Model validation
The 3D QSAR model was then challenged by designing, synthesizing, and testing novel compounds.
An in silico library of 200 βꢀlactones derivatives was assembled starting from a list of commercially
available alcohols. Their activity data were predicted through the 3D QSAR model (pIC50pred) and the
1
1
ACS Paragon Plus Environment

Page 15 of 55
Journal of Medicinal Chemistry
1
2
3
compounds were then ranked accordingly. The compounds with a pIC50pred value greater or equal to
.45 were selected for visual inspection. The structures of the resulting 23 compounds along with
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
7
pIC50pred values are reported in Table S2 (see Supporting Information). Eventually, we selected 5 new
derivatives for chemical synthesis and biological evaluation. The selection was based on the following
criteria: i) good predicted activity (rank), ii) chemical feasibility, iii) actual availability of the alcohol
substituent, iv) introduction of new chemical features with respect to already known βꢀlactones NAAA
inhibitors. The compounds (9r–v) were synthetized according to Scheme 1, and their rank and activity
data are reported in Table 4. For comparison, in Table S3 (see Supporting Information) we also report
the structures of the five least active compounds as predicted by the model.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
It clearly emerges the quite good agreement between predicted and observed pIC50 values, further
highlighting the robustness of the present 3D QSAR model. Indeed, our model not only retrospectively
explained the SAR in the lipophilic region of the NAAA binding pocket, but could also be used
prospectively to design novel NAAA inhibitors.
In Figure 3, the novel compounds are reported. The new molecules showed a good superimposition
with respect to the biphenyl moiety of 2. In particular, the APF lipophilic contribution of these
molecules matched well with the ideal lipophilic shape and size of the side chain obtained by our
model. These compounds were not profoundly different from those used in the training set to devise the
model. In this regard, it should be pointed out that a hopping exercise aimed at identifying completely
novel chemical scaffolds was outside the scope of the present 3D QSAR study. Furthermore, given the
3
6
covalent nature of the interaction established by these compounds with the enzyme, fair activity
predictions based on structure can only be limited to compounds sharing the same mechanism of
covalent inhibition. However, each new compound displayed a new structural feature, which
incrementally expanded our knowledge of the shape of the pocket, particularly relevant also in light of
1
2
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 16 of 55
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
the lack of a NAAA crystal structure to be utilized in structureꢀbased studies. Compound 9r (Figure
2
0
A; IC = 0.015 ꢀM, Table 4) displays an activity only marginally lower with respect to 2 (IC =
50
50
.007 ꢀM, Table S1) and 9l (IC = 0.007 ꢀM, Table 1), confirming that in this region an aliphatic
5
0
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
group is well tolerated and suggesting that more flexible nonꢀring substituents can be explored. In
Figure 3B, compound 9s (IC = 0.022 ꢀM, Table 4) is reported. Although the molecule resembles the
5
0
lipophilic signature of 2, the presence of a tertꢀbutyl substituent in para to the phenyl ring was not
detrimental for potency, indicating that, in this position, a bulky substituent is well tolerated. This was
also confirmed by the potency displayed by compound 9t (Figure 3C, IC = 0.028 ꢀM, Table 4) with
5
0
the bromoꢀsubstituted thiophene ring matching the lipophilic map of the present model while, at the
same time, exploring an additional volume by means of the bromine substituent. On the same line,
compound 9v was characterized by a reasonable activity (IC = 0.077 ꢀM, Table 4) combining a flat,
5
0
1
sp group in the region proximal to the carbamate, as already seen in 9j, with a long aliphatic chain
Figure 3D). Compound 9u represented an interesting exercise by which we explored the possibility to
(
introduce a heteroaromatic ring in close proximity to the carbamate functionality (Figure 3E).
Interestingly, this compound turned out to be very active (IC = 0.013 ꢀM, Table 4). When combined,
5
0
these data seem to suggest that not only this lipophilic region of the binding pocket is separated from
the catalytic machinery by a tight opening that can only accommodate linear and/or flat substituents,
but also that, immediately after this channel, the subpocket expands and can efficiently lodge bulkier
substituents. Putative bound conformations of 9r–v are reported in Figure S3 in the Supporting
Information.
Discovery of NAAA modulators is a quite recent medicinal chemistry field, as few inhibitors of this
enzyme have so far been reported in the literature. Here, we aimed at extending the SAR around the
series of potent βꢀlactone NAAA inhibitors. To this objective, we also generated the first 3D QSAR
1
3
ACS Paragon Plus Environment

Page 17 of 55
Journal of Medicinal Chemistry
1
2
3
model in the field, which turned out to be quite robust in terms of both statistical parameters and
predictivity, as shown by the 3D QSARꢀdriven design of novel analogues. We are confident that the
present 3D QSAR exploration can also, in a qualitative way, help designing novel structurally unrelated
NAAA inhibitors further expanding the chemical diversity in this field.
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
4
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 18 of 55
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
Conclusions
In this study, we investigated a series of NAAA inhibitors in terms of threeꢀdimensional quantitative
structureꢀactivity relationship (3D QSAR), and to the best of our knowledge, this represents the first 3D
QSAR model in this field. To this objective, we first expanded the SAR around the 2ꢀmethylꢀ4ꢀoxoꢀ3ꢀ
oxetanylcarbamic acid esters as NAAA inhibitors, hence investigating the role of previously
unexplored structural features on the inhibition of this enzyme. In particular, we further characterized
the size and shape of the side chain linked to the carbamic acid function, features that were extensively
investigated here for the first time. This eventually led to the discovery of a quite potent NAAA
inhibitor, compound 9l, endowed with singleꢀdigit nanomolar activity against this enzyme. In addition,
this allowed us to generate a consistent series of analogues that was then utilized to build the 3D QSAR
model. The statistical model was obtained using the innovative APF methodology, where discrete
features of classic pharmacophores are replaced by continuous, regularly spaced threeꢀdimensional
lattices. The 3D QSAR model showed a quite robust statistics and good consistency with docking
experiments carried out using a homology built model of NAAA. These promising results prompted us
to generate a small set of novel NAAA inhibitors, which were purposely designed and predicted using
the 3D QSAR model. The fairly good agreement between theoretical and experimental IC values for
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
5
0
the new 2ꢀmethylꢀ4ꢀoxoꢀ3ꢀoxetanylcarbamic acid esters highlights the good descriptive and predictive
power of the model. From a more general standpoint in the field of NAAA inhibitors, the present study
shows that an extended and lipophilic moiety could be the ideal pharmacophoric function for
substituting the side chain of covalent inhibitors of this enzyme. In conclusion, the present SAR
exploration and the 3D QSAR model can help the future design of novel NAAA inhibitors, which can
find therapeutic applications in the treatment of pain and inflammation.
1
5
ACS Paragon Plus Environment

Page 19 of 55
Journal of Medicinal Chemistry
1
2
3
Experimental section
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
a. Chemicals, Materials, and Methods. All the commercial available reagents and solvents
were used as purchased from vendors without further purification. Dry solvents (THF, Et O, CH Cl ,
2
2
2
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
DMF, DMSO, MeOH) were purchased from SigmaꢀAldrich. Optical rotations were measured on a
Rudolf Research Analytical Autopol II Automatic polarimeter using a sodium lamp (589 nm) as the
light source; concentrations are expressed in g/100 mL using CHCl as a solvent and a 1 dm cell.
3
Automated column chromatography purifications were done using a Teledyne ISCO apparatus
®
(CombiFlash R ) with preꢀpacked silica gel columns of different sizes (from 4 g up to 120 g).
f
Mixtures of increasing polarity of cyclohexane and ethyl acetate (EtOAc) or cyclohexane and methyl
tertꢀbutyl ether (MTBE) were used as eluents. NMR experiments were run on a Bruker Avance III 400
1
13
system (400.13 MHz for H, and 100.62 MHz for C), equipped with a BBI probe and Zꢀgradients.
Spectra were acquired at 300 K, using deuterated dimethylsulfoxide (DMSOꢀd ) or deuterated
6
chloroform (CDCl ) as solvents. UPLC–MS analyses were run on a Waters ACQUITY UPLC–MS
3
system consisting of a SQD (Single Quadrupole Detector) Mass Spectrometer equipped with an
Electrospray Ionization interface and a Photodiode Array Detector. PDA range was 210ꢀ400 nm.
Analyses were performed on an ACQUITY UPLC HSS T3 C column (50x2.1 mm ID, particle size
18
1
.8 µm) with a VanGuard HSS T3 C preꢀcolumn (5x2.1mm ID, particle size 1.8 µm). Mobile phase
18
was either 10 mM NH OAc in H O at pH 5 adjusted with AcOH (A) and 10 mM NH OAc in MeCNꢀ
4
2
4
H O (95:5) at pH 5. Electrospray ionization in positive and negative mode was applied. Purifications
2
by preparative HPLC–MS were run on a Waters Autopurification system consisting of a 3100 Single
Quadrupole Mass Spectrometer equipped with an Electrospray Ionization interface and a 2998
Photodiode Array Detector. HPLC system included a 2747 Sample Manager, 2545 Binary Gradient
Module, System Fluidic Organizer and 515 HPLC Pump. PDA range was 210ꢀ400 nm. Purifications
TM
were performed on a XBridge Prep C OBD column (100x19 mm ID, particle size 5 µm) with a
18
1
6
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 20 of 55
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
TM
XBridge Prep C (10x19 mm ID, particle size 5 µm) Guard Cartridge. Mobile phase was 10 mM
1
8
NH OAc in MeCNꢀH O (95:5) at pH 5. Electrospray ionization in positive and negative mode was
4
2
used. All tested compounds (9a–v) showed ≥ 95% purity by NMR and UPLC–MS analysis.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
General Procedure for the Synthesis of Carbamates 9r–v (Scheme 1)
Preparation of Activated Alcohols as Alkyl-2-pyridyl Carbonates 5r–v and Alkyl-2-oxopyridine-
1-carboxylates 6r–v (Step 1). To a stirred mixture of the suitable alcohol 3r–v (1.0 eq.) in dry CH Cl ,
2
2
and under nitrogen atmosphere, DMAP (0.1 eq.) and DPC (1.2 eq.) were added. The reaction mixture
was left to react at room temperature for 15 h, then diluted with CH Cl , washed with a saturated
2
2
NH Cl solution and, subsequently, with a saturated NaHCO solution (3x). The combined organic
4
3
layers were dried over Na SO , filtered, and concentrated to dryness to give a mixture (ratio 1.6:1) of
2
4
alkylꢀ2ꢀpyridylꢀcarbonate 5r–v and alkylꢀ2ꢀoxopyridineꢀ1ꢀcarboxylate 6r–v. The mixture of isomers
was not separated and used in the next step without any further purification.
Preparation of β-Hydroxycarboxylic Acids 8r–v (Step 2). To a stirred mixture of
Dꢀthreonine (1.0
eq.) and NaHCO (1.5 eq.) in H O (3.5 mL), the isomeric mixture containing the pyridyl carbonate 5r–
3
2
v and the 2ꢀoxopyridineꢀ1ꢀcarboxylate 6r–v (1.5 eq.) in THF (3.5 mL) was added. After 15 h at room
temperature, the crude was evaporated and subsequently extracted with Et O (3x5.0 mL). The aqueous
2
layer was acidified with 2.0 M HCl solution to pH 2ꢀ3 and subsequently extracted with EtOAc (3x10
mL). The combined organic layers were dried over Na SO , filtered, and concentrated to dryness to
2
4
give the threonine derivatives 8r–v, which were used in the next step without further purification.
Preparation of carbamates 9r–v (Step 3). To a stirred mixture of threonine derivatives 8r–v (1.0 eq.)
in dry CH Cl , at 0 °C, and under nitrogen atmosphere, Et N (3.0 eq.) and, subsequently, TBTU (1.2
2
2
3
eq.) were added. The mixture was stirred at 0 °C for 1 h and at room temperature for 15 h, then
1
7
ACS Paragon Plus Environment

Page 21 of 55
Journal of Medicinal Chemistry
1
2
3
concentrated, and the crude purified by column chromatography, eluting with cyclohexane:EtOAc
from 100:0 to 0:100) to afford pure 9r–v.
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
(
(4-Butylphenyl)methyl-N-[(2S,3R)-2-methyl-4-oxo-oxetan-3-yl]carbamate (9r). The reaction was
carried out following the general procedure (Step 3) employing (2R,3S)ꢀ2ꢀ[(4ꢀ
butylphenyl)methoxycarbonylamino]ꢀ3ꢀhydroxyꢀbutanoic acid (8r) (0.29 g, 0.92 mmol), dry CH Cl
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
2
2
(28 mL), Et N (0.39 mL, 2.77 mmol), and TBTU (0.36 g, 1.10 mmol) to give 9r. White solid; yield 8%
3
25
ꢀ
+ 1
(
0.02 g). [α] –17.61 (c 0.1, CHCl ). MS (ESI) m/z: 290 [M–H] , 292 [M–H] . H NMR (400 MHz,
D
3
DMSOꢀd ): δ 0.88 (t, J = 7.34 Hz, 3H), 1.22–1.32 (m, 2H), 1.34 (d, J = 6.28 Hz, 3H), 1.48–1.59 (m,
6
2
H), 2.57 (t, J = 7.65 Hz, 2H), 4.86 (dq, J = 6.08, 6.28 Hz, 1H), 5.01 (d, J = 12.38 Hz, 1H), 5.06 (d, J =
2.10 Hz, 1H), 5.44 (dd, J = 6.08, 9.42 Hz, 1H), 7.19 (d, J = 7.69 Hz, 2H), 7.27 (d, J = 7.77 Hz, 2H),
1
1
3
8.32 (d, J = 9.42 Hz, 1H). C NMR (101 MHz, DMSOꢀd ): δ 14.21, 14.98, 22.15, 33.56, 34.97, 60.41,
6
6
6.63, 75.11, 128.58, 128.79, 134.14, 142.76, 156.11, 170.27.
(4-tert-Butylphenyl)methyl-N-[(2S,3R)-2-methyl-4-oxo-oxetan-3-yl]carbamate (9s). The reaction
was carried out following the general procedure (Step 3) employing (2R,3S)ꢀ2ꢀ[(4ꢀtertꢀ
butylphenyl)methoxycarbonylamino]ꢀ3ꢀhydroxyꢀbutanoic acid (8s) (0.37 g, 1.21 mmol), dry CH Cl
2
2
(
37 mL), Et N (0.51 mL, 3.64 mmol), and TBTU (0.47 g, 1.45 mmol) to give 9s. White solid; yield:
3
2
5
ꢀ
+ 1
2
0% (0.07 g). [α] –17.38 (c 0.1, CHCl ). MS (ESI) m/z: 290 [M–H] , 292 [M–H] . H NMR (400
D
3
MHz, DMSOꢀd ): δ 1.27 (s, 9H), 1.34 (d, J = 6.35 Hz, 3H), 4.86 (dq, J = 6.08, 6.35 Hz, 1H), 5.01 (d, J
6
= 12.15 Hz, 1H), 5.06 (d, J = 12.15 Hz, 1H), 5.44 (dd, J = 6.08, 9.42 Hz, 1H), 7.26–7.31 (m, 2H),
1
3
7
5
.37–7.41 (m, 2H), 8.32 (d, J = 9.42 Hz, 1H). C NMR (101 MHz, DMSOꢀd ): δ 14.55, 31.12, 34.32,
6
9.96, 66.07, 74.68, 125.20, 127.91, 133.53, 150.64, 155.67, 169.85.
[4-(3-Bromo-2-thienyl)phenyl]methyl-N-[(2S,3R)-2-methyl-4-oxo-oxetan-3-yl]carbamate (9t). The
reaction was carried out following the general procedure (Step 3) employing (2R,3S)ꢀ2ꢀ[[4ꢀ(3ꢀbromoꢀ2ꢀ
1
8
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 22 of 55
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
thienyl)phenyl]methoxycarbonylamino]ꢀ3ꢀhydroxyꢀbutanoic acid (8t) (0.39 g, 0.95 mmol), dry CH Cl
2
2
(
40 mL), Et N (0.40 mL, 2.85 mmol), and TBTU (0.37 g, 1.14 mmol) to give 9t. White solid; yield:
3
2
5
ꢀ
+ 1
1
3% (0.051 g). [α] –63.21 (c 0.1, CHCl ). MS (ESI) m/z: 395 [M–H] , 397 [M–H] . H NMR (400
D
3
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
MHz, DMSOꢀd ): δ 1.36 (d, J = 6.32 Hz, 3H), 4.87 (dq, J = 6.28 Hz, 1H), 5.11 (d, J = 12.89 Hz, 1H),
6
5.16 (d, J = 12.56 Hz, 1H), 5.46 (dd, J = 6.06, 9.42 Hz, 1H), 7.19–7.22 (m, 1H), 7.48 (d, J = 7.98 Hz,
1
3
2
H), 7.63 (d, J = 7.98 Hz, 2H), 7.71–7.74 (m, 1H), 8.42 (d, J = 9.42 Hz, 1H). C NMR (101 MHz,
DMSOꢀd ): 14.54, 59.96, 65.77, 74.65, 107.25, 127.07, 128.28, 128.73, 131.57, 131.91, 136.85,
6
137.08, 155.56, 169.75.
(5-Methylbenzothiophen-2-yl)methyl-N-[(2S,3R)-2-methyl-4-oxo-oxetan-3-yl]carbamate (9u). The
reaction was carried out following the general procedure (Step 3) employing (2R,3S)ꢀ3ꢀhydroxyꢀ2ꢀ[(5ꢀ
methylbenzothiophenꢀ2ꢀyl)methoxycarbonylamino]butanoic acid (8u) (0.37 g, 1.15 mmol), dry CH Cl
2
2
(
37 mL), Et N (0.48 mL, 3.48 mmol), and TBTU (0.45 g, 1.39 mmol) to give 9u. White solid; yield:
3
2
5
ꢀ
+ 1
2
7% (0.095 g). [α] –20.99 (c 0.1, CHCl ). MS (ESI) m/z: 304 [M–H] , 306 [M–H] . H NMR (400
D
3
MHz, DMSOꢀd ): δ 1.35 (d, J = 6.30 Hz, 3H), 2.40 (s, 3H), 4.87 (dq, J = 6.08, 6.30 Hz, 1H), 5.32 (d, J
6
=
13.20 Hz, 1H), 5.34 (d, J = 13.20 Hz, 1H), 5.45 (dd, J = 6.07, 9.34 Hz, 1H), 7.19 (dd, J = 1.67, 8.23
1
3
Hz, 1H), 7.38 (s, 1H), 7.60–7.64 (m, 1H), 7.81 (d, J = 8.22 Hz, 1H), 8.44 (d, J = 9.34 Hz, 1H). C
NMR (101 MHz, DMSOꢀd ): 14.54, 20.96, 59.99, 61.53, 74.66, 122.22, 123.67, 124.13, 126.42,
6
1
33.83, 136.93, 139.19, 139.65, 155.39, 169.64.
Dec-2-ynyl-N-[(2S,3R)-2-methyl-4-oxo-oxetan-3-yl]carbamate (9v). The reaction was carried out
following the general procedure (Step 3) employing (2R,3S)ꢀ2ꢀ(decꢀ2ꢀynoxycarbonylamino)ꢀ3ꢀ
hydroxyꢀbutanoic acid (8v) (0.35 g, 1.18 mmol), dry CH Cl (35 mL), Et N (0.49 mL, 3.56 mmol), and
2
2
3
2
5
TBTU (0.46 g, 1.42 mmol) to give 9v. White solid; yield: 26% (0.090 g). [α] –15.83 (c 0.1, CHCl3).
D
ꢀ
+ 1
MS (ESI) m/z: 280 [M–H] , 282 [M–H] . H NMR (400 MHz, DMSOꢀd ): δ 0.83–0.88 (m, 3H), 1.20–
6
1
9
ACS Paragon Plus Environment

Page 23 of 55
Journal of Medicinal Chemistry
1
2
3
1
1
.37 (m, 12H), 1.38–1.48 (m, 2H), 2.17–2.24 (m, 1H), 4.65 (d, J = 2.62 Hz, 1H), 4.67 (d, J = 2.57 Hz,
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
1
3
H), 4.86 (dq, J = 6.30 Hz, 1H), 5.43 (dd, J = 6.07, 9.41 Hz, 1H), 8.38 (d, J = 9.41 Hz, 1H). C NMR
(101 MHz, DMSOꢀd ): δ 13.92, 14.48, 17.87, 22.03, 27.88, 28.13, 31.14, 52.98, 59.92, 74.57, 75.11,
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
8
7.12, 99.50, 154.98, 169.62.
b. Pharmacology
3
4
Rat NAAA in vitro assay. For a detailed description of inꢀvitro rat-NAAA assay, see Supporting
Information.
c. Computational Methods
Compounds preparation. For each compound, we generated a twoꢀdimensional representation in
which bond orders, tautomeric forms, stereochemistry, hydrogen atoms, and protonation states were
manually assigned. Conversion from a twoꢀ to a threeꢀdimensional representation was automatically
4
1
performed by the ICM converting procedure (ICM3.7, Molsoft LLC, San Diego, CA). Cartesian
coordinates were translated into internal coordinates and each molecule was assigned the MMFF force
field atom types and charges.
Three-dimensional alignment of the training set molecules. We generated our initial structural
alignment starting from a rigid template, namely the structure of the transition state model of αꢀamino
βꢀlactone 2. Compound 2 transition state was modeled as follows: the 2ꢀmethylꢀ4ꢀoxoꢀ3ꢀ
oxetanylcarbamic acid ester region was assigned the optimized coordinates that we previously
3
4
identified by means of quantum chemical calculations. In order to identify a plausible starting
conformation of the molecule, we performed an energy optimization on the three torsional angles
describing the orientation of the biphenyl moiety by means of the biased probability Monte Carlo
4
2, 43
(BPMC) stochastic optimizer as implemented in ICM.
The vicinity threshold, namely the rootꢀ
meanꢀsquare deviation threshold for the rotatable torsions, was set equal to 15 degrees. In ICM, the
2
0
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 24 of 55
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
number of Monte Carlo steps is automatically set by an adaptive procedure. After performing Cartesian
minimization, we selected the lowest energy conformer appearing compatible with the pocket
3
2
orientation that was described by Solorzano and coworkers. . Compounds 9a–q were structurally
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
aligned to this rigid template according to the Atomic Property Fields (APF), procedure recently
3
7
reported and thoroughly described by Totrov. First, we calculated seven continuous threeꢀ
dimensional grid potentials representing the pharmacophoric features of our template. The grids
covered the template plus a 5 Å margin. These APF potentials accounted for: hydrogen bond donor and
acceptor propensity, lipophilicity, size, charge, hybridization, and electronegativity. Each molecule to
be superimposed to the template was globally optimized within the preꢀcalculated APF. The
intramolecular forceꢀfield energy was also taken into account in order to avoid the generation of
unrealistically strained conformations. An initial superposition was generated matching the inertia
ellipsoid of the molecule in a random conformation with that of the template. Then, an optimization
was undertaken using the BPMC method. Torsional variables describing the conformation of flexible
rings were sampled explicitly. Thoroughness, namely the parameter that controls the length of the
optimization process, was set equal to 1, a value that has been validated to provide a simulation of
suitable length.
3
D QSAR model. The 3D QSAR model was built starting from the design of a chemical table
containing the superimposed compounds and the NAAA inhibitor activity data (Table 1 and Table S1).
4
1
The APF 3D QSAR methodology as implemented in ICM has been used. For each molecule the 7ꢀ
component APF were calculated and added together. In order to make the APF model that could
provide quantitative prediction of activity for new compounds the proper weight for the contributions
of each compound to each APF component has to be defined. Using the activity data for the training set
compounds the weights for the contributions of each compound were fitted to reproduce experimental
44
data. The partial least squares (PLS) methodology was used to derive the optimal weights. Once the
2
1
ACS Paragon Plus Environment

Page 25 of 55
Journal of Medicinal Chemistry
1
2
3
weights were derived, partial perꢀcompound potentials were combined into the total weighted QSAR
APF and an estimate of activity for a new compound can be performed by a single calculation of the
pseudoꢀenergy. The optimal number of latent vectors for PLS was established by leaveꢀoneꢀout crossꢀ
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
2
validation on the training set. In Table 2 the squared correlation coefficients (R ) and leaveꢀoneꢀout
2
squared correlation coefficients (Q ) according to the number of latent vectors are reported.
3
D QSAR model validation. We generated a virtual library of βꢀlactones molecules through the
Markush Combinatorial Library tool of ICM using a library of commercially available alcohols
4
1
reacting with the βꢀlactone amino group. In this way we automatically created a list of 200 new
compounds: their activity data were then derived through the 3D QSAR model. The molecules were
then ranked according to their predicted pIC . We then selected all the compounds with a predicted
5
0
IC greater than a preꢀspecified threshold. Having used 2 as rigid template for the alignment, we set
5
0
our lower threshold for activity equal to 7.45, corresponding to the potency of the less active biphenylꢀ
bearing derivative (9o). Within the selected 23 molecules, only five compounds were then chosen to be
synthetized for the subsequent model validation. The choice of these compounds was based on the
following criteria: chemical feasibility, good predicted activity data, interesting new chemical features
that are not kept by the available known βꢀlactones molecules.
2
2
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 26 of 55
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
Figures, Schemes/Structures, and Charts
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
ARN077 (1)
2
Figure 1.
2
3
ACS Paragon Plus Environment

Page 27 of 55
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 2.
2
4
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 28 of 55
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 3
2
5
ACS Paragon Plus Environment

Page 29 of 55
Journal of Medicinal Chemistry
1
2
3
Figure legends
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
Figure 1. αꢀAmino βꢀlactone potent NAAA inhibitors.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 2. A: The overlay of the training set molecules superimposed to compound 2; B: Correlation
plot between pIC50exp and pIC50pred; C: In licorice the 3D structure of 2 is reported. The transparent
blobs are the equipotential contours of the lipophilic 3D QSAR APF component. The grey region
represents the ideal lipophilic shape for NAAA inhibition. D: In green is reported the 3D structure of
2
0 superimposed to compound 2, in orange. The grey blobs represent the equipotential contours of the
Sp2 hybridization 3D QSAR APF component. The occupancy of such regions is detrimental for the
potency. E: In yellow is reported 16 superimposed to compound 2, in orange. The grey blobs represent
the equipotential contours of the size 3D QSAR APF component. The occupancy of such regions is
detrimental for the potency.
Figure 3: The molecules selected for the validation of the 3D QSAR model are shown in yellow, as
licorice, superimposed to 2 in pink. A: molecule 9r, B: molecule 9s, C: molecule 9t, D: molecule 9v, E:
molecule 9u.
2
6
ACS Paragon Plus Environment