From Experiments to a Fast Easy-to-Use Computational Methodology to Predict Human Aldehyde Oxidase Selectivity and Metabolic Reactions

Article abstract of DOI:10.1021/acs.jmedchem.7b01552

Aldehyde oxidase (AOX) is a molibdo-flavoenzyme that has raised great interest in recent years, since its contribution in xenobiotic metabolism has not always been identified before clinical trials, with consequent negative effects on the fate of new potential drugs. The fundamental role of AOX in metabolizing xenobiotics is also due to the attempt of medicinal chemists to stabilize candidates toward cytochrome P450 activity, which increases the risk for new compounds to be susceptible to AOX nucleophile attack. Therefore, novel strategies to predict the potential liability of new entities toward the AOX enzyme are urgently needed to increase effectiveness, reduce costs, and prioritize experimental studies. In the present work, we present the most up-to-date computational method to predict liability toward human AOX (hAOX), for applications in drug design and pharmacokinetic optimization. The method was developed using a large data set of homogeneous experimental data, which is also disclosed as Supporting Information.

Full text of DOI:10.1021/acs.jmedchem.7b01552

Subscriber access provided by READING UNIV
Article
From experiments to a fast easy-to-use computational methodology to
predict human aldehyde oxidase selectivity and metabolic reactions
Gabriele Cruciani, Nicolò Milani, Paolo Benedetti, Susan Lepri, Lucia Cesarini, Massimo
Baroni, Francesca Spyrakis, Sara Tortorella, Edoardo Mosconi, and Laura Goracci
J. Med. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.jmedchem.7b01552 • Publication Date (Web): 14 Dec 2017
Downloaded from http://pubs.acs.org on December 15, 2017
Just Accepted
“Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted
online prior to technical editing, formatting for publication and author proofing. The American Chemical
Society provides “Just Accepted” as a free service to the research community to expedite the
dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts
appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been
fully peer reviewed, but should not be considered the official version of record. They are accessible to all
readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered
to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published
in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just
Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor
changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers
and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors
or consequences arising from the use of information contained in these “Just Accepted” manuscripts.
Journal of 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 36
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
From experiments to a fast easy-to-use
computational methodology to predict human
aldehyde oxidase selectivity and metabolic
reactions
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
Gabriele Cruciani,1,2* Nicolò Milani,1 Paolo Benedetti,1,2 Susan Lepri,1 Lucia Cesarini,1 Massimo
Baroni,3 Francesca Spyrakis,4 Sara Tortorella,2,5 Edoardo Mosconi2,6 and Laura Goracci1,2
1Department of Chemistry, Biology and Biotechnology, University of Perugia, via Elce di Sotto 8,
06123 Perugia, Italy.
2Consortium for Computational Molecular and Materials Sciences (CMS)2, via Elce di Sotto 8,
06123 Perugia, Italy.
3Molecular Discovery Ltd, Centennial Park, Borehamwood, Hertfordshire, United Kingdom.
4Department of Drug Science and Technology, University of Turin, via P. Giuria 9, 10125 Turin,
Italy.
5Molecular Horizon srl, via Montelino 32, 06084 Bettona, Italy.
6Computational Laboratory for Hybrid/Organic Photovoltaics, National Research Council–Institute
of Molecular Science and Technologies, Via Elce di Sotto 8, Iꢀ06123 Perugia, Italy.
KEIWORDS: Aldehyde oxidase, metabolism, oxidation, hydrolysis, molecular modelling,
molecular dynamics, qualitative structureꢀmetabolism relationships, Parr’s index.
1
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 2 of 36
1
2
3
ABSTRACT
4
5
6
7
8
Aldehyde oxidase (AOX) is a molibdoꢀflavoenzyme that raised great interest in recent years, since
its contribution in xenobiotics metabolism has not always been identified before clinical trials, with
consequent negative effects on the fate of new potential drugs. The fundamental role of AOX in
metabolizing xenobiotics is also due to the attempt of medicinal chemists to stabilize candidates
toward cytochrome P450 activity, which increases the risk for new compounds to be susceptible to
AOX nucleophile attack. Therefore, novel strategies to predict the potential liability of new entities
towards AOX enzyme are urgently needed to increase effectiveness, reduce costs, and prioritize
experimental studies. In the present work, we present the most upꢀtoꢀdate computational method to
predict liability towards human AOX (hAOX), for applications in drug design and pharmacokinetic
optimization. The method was developed using a large dataset of homogeneous experimental data,
which is also disclosed as supplementary material.
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
INTRODUCTION
Evaluation of the metabolism of a new drug or drug candidate in pharmaceutical companies is
essential in measuring both safety and efficacy. The development of reliable screening systems has
produced a 10% decrease in the rate of pharmaceutical drug candidate clinical trials interruption.1
At the same time, the development of in silico methods to predict the metabolism of xenobiotics has
made the chemical discovery phase much more effective.2ꢀ5 Nowadays, medicinal chemists have
various working strategies available to overcome the issue of unstable candidates with respect to the
P450 cytochrome (CYP) family. Once the site of metabolism (SoM) in a molecule has been
identified using fast experimental (MetID via MassꢀMetaSite)6 or in silico methods,2ꢀ4 the
metabolism of the molecule can be modified by directly changing the SoM or different molecular
regions. For instance, the addition of fluorine, halogens, or deuterium to the SoM, when possible,
2
ACS Paragon Plus Environment

Page 3 of 36
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
can significantly modify the pharmacokinetic profile of the compound. The substitution of a known
labile group with a more resistant one towards oxidation has a significant benefit. Secondly, the
medicinal chemist must initially understand which region of the molecule largely contributes to the
exposure of the molecular site where the reaction takes place so that it can be modified. In both
cases, aryl groups or carbocyclic aromatics possibly present in the scaffold can be substituted by
heteroaromatic rings that, in addition to a positive impact on the clearance, can also impact on
solubility and physicalꢀchemical properties such as LogD.
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
Nevertheless, if understanding the mechanisms responsible for molecules liability has allowed
medical chemists to become more effective in planning compounds stable to CYP enzymes, it has
also abundantly increased the number of compounds that can be potentially metabolized by nonꢀ
CYP enzymes. This means that the contribution of nonꢀCYP metabolism is now very significant in
modern pharmaceutical research and is steadily increasing in importance.7 In particular, studies of
the metabolism mediated by aldehyde oxidase (AOX) have assumed a prominent role as
demonstrated by the numerous papers on this subject (about 400 papers published in the last few
years),8 and by the various experimental strategies developed to determine the propensity of a
compound to be an AOX substrate.9 AOX is a cytosolic enzyme that catalyzes oxidation by a
nucleophile mechanism using water as a source of oxygen that is incorporated in the metabolite.
AOX also shows significant ability to hydrolyze the amide bonds that are generally metabolically
stable and hardly modified by CYP enzymes.
But why should medicinal chemists matter about AOX? This effort is actually encouraged by the
numerous recent interruptions of clinical trials due to AOX metabolism. In particular, the clinical
trials of candidates such as BIBX1382, FK3453, ROꢀ1, and carbazeran have been suspended
because of poor bioavailability and rapid elimination in humans,10ꢀ13 while SGXꢀ523 was
interrupted due to renal insufficiency following aldehyde oxidase metabolism in humans.14 In reply
to the growing role assumed by AOX, the scientific community in pharmaceutical companies has
3
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 4 of 36
1
2
3
4
5
6
7
8
9
activated appropriate screening programs to evaluate the metabolic instability caused by new
chemical entities in different biomatrices. However, a more global strategy is needed to increase
efficacy, reduce costs, and to prioritize the experimental study of new candidates. In this context, in
silico methods can offer a valuable strategy to model and predict the role of AOX towards potential
new compounds.
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
In the present work, we present a novel computational method to predict liability towards human
AOX. This new model combines the knowledge acquired from both experimental and in silico
sources. Indeed, more than 600 compounds (acquired or synthesized to meet the needs of assessing
structureꢀactivity relationships) were experimentally tested in human liver cytosol (HLC) to
generate the largest database of homogeneous experimental data publically available so far. In
addition, our experience in modelling metabolism by phase I metabolic enzymes (CYP and Flavinꢀ
containing monooxygenase 3, or FMO3) in terms of chemical reactivity and exposure to the
enzyme represented the background to develop a new model to predict human aldehyde oxidase
(hAOX) metabolism. To better describe why we decided to address our efforts towards the hAOX
modelling, examples of ineffective optimization recently occurred are analyzed. Subsequently,
correlation of hAOX liability with lipophilicity (LogP and LogD) and electronic parameters (Parr’s
index) is discussed. Finally, our proposed model for hAOX liability prediction and its validation are
described.
RESULTS AND DISCUSSION
From a CYP substrate to a hAOX substrate: examples of ineffective optimization
Here we report two real cases of metabolic shift from CYP to hAOX recently occurred in our
laboratories. First, when compound 1 (Figure 1ꢀa), planned to interfere with several nuclear
receptors, was incubated in human liver microsomes (HLM), it demonstrated notable metabolic
instability. After 60 minutes the compound was totally degraded into three metabolites, with the
4
ACS Paragon Plus Environment

Page 5 of 36
Journal of Medicinal Chemistry
1
2
3
major one being reported in Figure 1ꢀa as 1M. Incubation of 1 with recombinant CYP isoforms
4
5
demonstrated the preeminent role of the enzyme CYPꢀ3A4 (data not shown).
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 1. Metabolic stability assessment of compounds 1 and 2. a) Metabolic stability assessment
of compound 1 in HLM. The compound was rapidly metabolized in the ethylendiamine portion to
give compound 1M. b) Metabolic stability assessment of compound 2 in HLM (no metabolites
observed) and HLC (rapid formation of metabolite 2M). In the kinetic profiles, substrates and
metabolites are colored in black and green, respectively.
An analysis of the reactive pose of compound 1 in CYPꢀ3A4 performed with MetaSite3,15,16
demonstrated the fundamental role of the naphthyl group in molecular exposure. Indeed, one of the
two aromatic rings, interacting with Ile301, Phe304 and Ala305, stabilizes the docking pose and
promotes the oxidation and Nꢀdemethylation reactions in the ethylendiamine part of the substrate
(Figure 2).
5
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 6 of 36
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. Docking pose of compound 1 in CYPꢀ3A4 cavity, according to MetaSite prediction.
To make the compound metabolically more resistant, the naphthyl group was substituted with a
quinoxalin group as in compound 2 (Figure 1ꢀb) that notably stabilized the substrate in HLM
biological matrix. However, an analysis of the newly synthesized compound 2 in HLC
demonstrated the rapid degradation of the compound to give the 2M metabolite (Figure 1ꢀb).
Moreover, attempting to make the compound more stable to the biotransformation induced by the
cytochrome favored the attack by the hAOX enzyme. Therefore, the problem has not been solved
but only shifted and in some ways, dangerously hidden.
A similar situation might easily occur with another fundamental organic group to the medicinal
chemist, that is, the amide group. Amide groups have typically been inserted in molecules of
interest to improve chemical stability, rigidity, and bond directionality, and to modulate polarity and
bioavailability. However, it is now clear that amide groups can be rapidly cleaved in vivo by hAOX
enzyme.17,18
6
ACS Paragon Plus Environment

Page 7 of 36
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
In this context, the second case discussed here concerns our attempts to metabolically stabilize
compound 3 (Figure 3), active in signal transduction and metabolism cascades exerting a putative
antiꢀcancer effect (data not shown). To this aim, we decided to substitute the labile 1Hꢀindolꢀ2ylꢀ
methyl group with the 1Hꢀindoleꢀ2 carboxamide group.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Figure 3. Metabolic stability assessment of compound 3 and 4. a) Compound 3 was rapidly
metabolized to give metabolite 3M. b) Compound 4 was stable in HLM, nevertheless it was
hydrolyzed by hAOX in HLC to give, again. metabolite 3M.
Again, although our strategy reduced the CYPꢀ3A4ꢀmediated metabolism, it favored the hydrolysis
reaction promoted by hAOX. It should be noted that this example (phenotypic shift) is even more
relevant than the previous one, as the amide bond is notoriously difficult to break and is always
used by medical chemists to stabilize pharmaceutical drug candidates.
These examples emphasize the fundamental importance of fully understanding the mechanisms of
metabolic clearance, and the relevant differences of species when pharmaceutical drug candidates
are clinically evaluated.
7
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 8 of 36
1
2
3
hAOX liability prediction: which correlations are possible?
4
5
6
7
8
A certain number of in silico approaches were used in the past to predict the liability of a molecule
to hAOX and to predict its potential metabolic site. While the physicalꢀchemical properties of
molecules such as LogP or LogD are significant determinants of the metabolic clearance mediated
by CYP,19ꢀ22 the substrates of hAOX did not demonstrated any correlations with these parameters.23ꢀ
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
25
To confirm this point, we calculated LogP and LogD7.5 values for each chemical structure in our
internal database, which contains more than 600 compounds experimentally tested to assess hAOX
liability (the entire database is provided in the Supporting Information). As clearly reported in
Figure 4, molecular lipophilicity has not proved to be discriminating for hAOX selectivity in either
reaction mechanism.
Figure 4. LogP values distribution for experimentally determined hAOX substrates (green) and
nonꢀsubstrates (red), for the oxidation (left) and hydrolysis reaction (right). LogP and LogD7.5
values were obtained from VolSurf+.26,27 For chemical structures and statistical data see Supporting
Information.
8
ACS Paragon Plus Environment

Page 9 of 36
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Since the hAOX catalyzed oxidation of a heterocyclic ring initially leads to a nucleophile attack by
the hydroxyl group, susceptibility to hAOX can reasonably depend on the electronic deficiency of
the reaction site. Consequently, descriptors of reactivity or parameters that determine
electrophilicity at the metabolic site have been tested to predict susceptibility to hAOX. The energy
of the lowest unoccupied molecular orbital (ELUMO) and the electrostatic potential of the carbon
adjacent to the nitrogen atom in the heteroaromatic ring are two descriptors able to properly predict
the occurrence of a nucleophile attack.28 Ghafourian and Rashidi used these parameters to explore
the quantitative structure ꢀ activity relationships (QSAR) of a group of compounds that contain a
phthalazine and a quinazoline ring.29 Recently, Dalvie et al. compared the ELUMO of zoniporide, a
molecule with great affinity as a substrate for hAOX, and of various analogues to test the capability
of these parameters in predicting hAOX ꢀsubstrate properties.30 Torres et al. demonstrated how
density functional theory (DFT) methodologies can be used to predict AOX metabolism sites in
heteroaromatic rings.31 Pryde et al. and Dalvie et al. used this strategy to predict the hAOX lability
of newly designed compounds.25,30 More recently, the Parr’s index was used to estimate molecule
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
electrophilicity and establish
a
possible relationship with the observed extent of
biotransformation.32,33 However, as demonstrated in Figure 5, the Parr’s electrophilicity index
alone, calculated for each compound in our database, did not allow a good discrimination in
substrate hAOX selectivity for both oxidation and hydrolysis reactions (see section S2 in
Supporting Information for computational details).
9
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 10 of 36
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 5. Distribution plot for hAOX substrates (top) and nonꢀsubstrates (bottom), according to
Parr’s electrophilicity index (ω) values (for molecular structures, experimental details and statistical
data see Supporting Information).
Additionally, we noticed that in some cases hAOX liability cannot be rationalized solely by
analyzing the initial nucleophile attack by the hydroxyl group bound to molibdenum, and thus the
electrophilicity of the ligand sp2 carbon, but also considering the subsequent hydride displacement
step. For instance, for a number of phthalazines derivatives an opposite trend between the hAOX
susceptibility and the electronic effect of the substituting groups was recently observed.17 Indeed, as
reported in Figure 6, 5ꢀnitroꢀphthalazine (5) turned out to be significantly more stable than
phthalazine itself (6) and than 5ꢀaminoꢀphthalazine (7), although the first compound possesses a
10
ACS Paragon Plus Environment

Page 11 of 36
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
strong electron withdrawing group, which is supposed to decrease the electron availability of the
ring. DFT analyses performed to investigate the thermodynamics of the hydride displacement34 step
and the nature of the transition state rationalized the observed trend. In particular, the calculations
(Table S3ꢀ1, Section S3 in Supporting Information) showed that hydride displacement is more
thermodynamically favored (ꢁGsolv) and has a lower activation energy (ꢁG≠solv) for 5ꢀaminoꢀ
phthalazine than for 5ꢀnitroꢀphthalazine, mainly due to the possible intramolecular hydrogen bond
in 5ꢀaminoꢀphthalazine (Figure S3ꢀ1, in Supporting Information). Computational details are
provided in section S3 of the Supporting Information.
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 6. Metabolic stability of 5ꢀnitroꢀphthalazine (5), phthalazine (6) and 5ꢀaminoꢀphthalazine
(7) in HLC.
Obviously, DFT calculations, accounting for the energetics and the nature of intermediates and
transition states of the overall process are not feasible (in term of time and computational cost) in a
drug design approach, where hundreds of candidates are screened at the same time. However, such
results demonstrate that the assumption that electronꢀwithdrawing group (EWG) will increase
electrophilicity, and therefore hAOX liability, might sometimes fail. At the same time, a general
11
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 12 of 36
1
2
3
4
5
6
7
8
9
assumption that a nitroꢀ substituent can lead to a more stable compound cannot be made. Indeed, 5ꢀ
nitroquinoline was recently found to be a substrate of hAOX, and we also confirmed this data with
the experimental conditions used to generate our library.35,36 Thus, it is evident that other effects
like exposure to the catalytic site must be taken into account for a more accurate prediction.
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
From an experimental point of view, with the intention of finding the susceptibility of heteroarenes
to AOX metabolism, O’Hara et al.37 recently reported a “Litmus test” for the early identification of
AOX drug substrates. This method uses zinc difluoromethanesulfinate (DFMS) as a source of CF2H
radical to simulate AOX metabolic reaction. A binary qualitative answer (reactive or not reactive)
was the desired readout of this chemical test (here called AOXꢀLitmus test). Searching for reliable
correlations with hAOX liability, we also tested nine compounds, selected among the list of
experimentally proved nonꢀsubstrates to cover different scaffolds (Table S4ꢀ1, Supporting
Information), by the AOXꢀLitmus test. Interestingly, we found that for six compounds the test
failed (see Figure 7 for compounds containing different heteroarene moieties (imidazo[1,2ꢀ
a]pyrimidine, isoquinoline, and quinoxaline)). When compounds 8, 10 and 12 were submitted to the
AOXꢀLitmus test, all compounds resulted “reactive”, and thus AOX susceptible substrates.
However, when tested in cytosolic hAOX, no one of the compounds was affected by hAOX and all
three compounds were totally stable. Furthermore, AOXꢀLitmus test on compounds 10 and 12
produced different fluorinated regioisomers (Figure S4ꢀ1 in Section 4 of the Supporting
Information). This fact demonstrates that chemical reactivity alone and, consequently, all those
methods taking into account only chemistry cannot be used to predict the susceptibility of chemical
compounds to hAOX. Again, chemical reactivity is only a component of a mechanism that is
biologically much more complex.
12
ACS Paragon Plus Environment

Page 13 of 36
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 7. Three hAOX nonꢀsubstrates (8, 10, and 12) from our database (see Supporting
Information) were tested with AOXꢀLitmus test conditions. Experimental conditions: a) DFMS,
DMSO, TFA, TBHP.
Exposure component extimation from hAOX flexibility and dynamics studies
At this stage, it appeared evident that the rationalization of experimental results could be supported
by structural analyses taking into account also the ligand exposure in the hAOX catalytic site.
Indeed, molecular docking approaches are often used to clarify significant interactions between the
substrates and the residual of the active site of enzymes that metabolize the pharmaceutical drugs.
The threeꢀdimensional model of hAOX38 has been used to model new compounds by using the
algorithm present in the MetaSite software [version 6.0],39 to which appropriate modifications have
been made and are reported below.
The AOX enzyme is composed of two identical subunits of 150 kDa and is prevalently expressed
in the liver but also in other human tissues.
The structure of the hAOX enzyme has recently been solved by xꢀray crystallography (code
4UHX).38 This protein structure is missing two important polar amino acids (Asp881, Glu882)
13
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 14 of 36
1
2
3
4
5
6
7
8
9
situated in the loop between the Leu880 and Ser883 residues that are only at 4.5 Å distance from
the substrate phthalazine. It is clear that they carry out a significant role in molecular interaction
with potential substrates. Furthermore, also residues Phe655, Phe656, Thr657, Glu658, Ala659 and
Glu660, located in the lower part of the binding site, are missing. These residues, mainly
hydrophobic, belong to the gate 1 region, while the previous two are part of the gate 2. Both gate 1
and gate 2 are dynamic loops involved in regulating the ligands entrance into the pocket.38 To avoid
false results from these structural gaps, and from other loops not resolved in the original xꢀray
structure, we built a model of the entire protein using the 4UHX structure as template and adding
the missing loops (Figure 8). The resulting hAOX catalytic cavity includes 30 amino acids, which,
notwithstanding their different physicalꢀchemical natures and different polarity, have generated a
large pathway of hydrophobic interaction with fewer areas of Hꢀbonding interaction with Hꢀbond
interaction donors being almost absent. Calculations made using waterꢀFlap40,41 have shown the
presence of a water molecule in proximity to the molybdenum cofactor. This water is mobile but
localized in a wellꢀdefined region, where the enzymeꢀsubstrate nucleophile attack can take place,
and the oxygen of the water molecule is transferred to the hAOX substrate.
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
It is clear that the entire protein and, in particular, the binding site displays a significant degree of
flexibility, likely fundamental for the protein to perform its action. To better investigate the protein
intrinsic dynamic we performed plain Molecular Dynamics (MD) studies starting from the protein
model previously mentioned, which was minimized, equilibrated and submitted to 200 ns long MD
simulations (see Experimental Section). The obtained trajectory was clustered according to the
variation of the Molecular Interaction Fields (MIFs) of the pocket,42 as defined by the first
mentioned thirty residues. The analyses confirmed a high flexibility of the 655ꢀ663 and of the 881ꢀ
885 segments, supporting their role in regulating the pocket accessibility (Figure 8). This intrinsic
dynamic triggers a significant variation of the pocket MIFs and volume, and of the chemical and
energetic properties of the catalytic site, but also a dramatic change in the pocket dimension and
14
ACS Paragon Plus Environment

Page 15 of 36
Journal of Medicinal Chemistry
1
2
3
4
5
accessibility. These features could reasonably be responsible of the enzyme capability of accepting
and metabolizing a number of different and variable chemical entities.
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 8. Flexibility of gate 1 and 2. The different conformations of the gates have been taken
from the ten medoids obtained by the trajectory clustering. Each color corresponds to a different
conformation. The protein binding site cavity was calculated with FLAPsite module in FLAP
software (Molecular Discovery, UK)43ꢀ45 and is represented as grey surface. The MoCo cofactor is
shown in sphere.
Modelling hAOX -catalyzed reactions
We reasoned that, likewise with cytochromes P450 and FMO enzymes, both oxidation and
hydrolysis reactions are carried out by means of a combination of substrate reactivity (described
here by the electrophilicity) and the substrate interactions with the hAOX enzyme.
15
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 16 of 36
1
2
3
4
5
6
7
8
In previous studies,3,15,16 we reported that, once the component of accessibility to the enzymatic site
and electrophilic reactivity has been calculated, the metabolic site is described by a probability
function for the metabolic site (PSoM):
9
PSoMi = (1 + we)Ei(1 + wr)Ri
(1)
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
This function is considered to be an approximation of the free energy of the process including
substrateꢀenzyme interaction,3,15,16 where exposure (E) and reactivity (R) are opportunely weighted
coefficients (we and wr, respectively). These coefficients are not fixed values, being tuned based on
the enzyme cavity and structural features of substrates, representing the soꢀcalled “reactivityꢀ
equalization”.15 The two factors that contribute to the PSoM function make contributions that depend
on macromolecule utilization (CYP, FMO3, AOX) and on the substrate molecule. In the case of
hAOX, our analysis of the experimental data collected for more than 600 compounds indicated that
the function that describes the substrate exposure plays an important role practically equivalent than
that described by the electrophilic reactivity of the substrate. Whatever the case, it is the product of
the two components that determines whether a compound is a substrate of the enzyme hAOX or
not, and in which atom a reaction can occur (if several reactive atoms are present). Figure 9
illustrates an example of reactivity and exposure contributions for experimental results obtained on
structurally similar molecules but which have shown totally different experimental behaviors.
16
ACS Paragon Plus Environment

Page 17 of 36
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 9. Examples of reactivity and exposure effects. a) Compounds 14 and 15, having an amide
group with similar electrophilicity (due to tertiary amines as anilide substituents) but different shape
showed opposite behavior in hAOX ꢀmediated hydrolysis reaction; b) compounds 16 and 17,
similar in shape but not in electrophilic nature, were hydrolyzed by hAOX with a different extent;
c) and d) molecules with the same azaꢀaromatic electrophilicity (as in 18 and 19) or molecular
shape (12 and 20) showed opposite behavior towards oxidation reaction performed by hAOX.
In Figure 9ꢀa, the carbonyl carbon of the compounds 14 and 15 has the same electrophilicity;
nevertheless, only compound 15 resulted to be a hAOX substrate. This observation suggests that in
17
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 18 of 36
1
2
3
4
5
this case the portion of exposure to the reactive enzyme site will be the dominant part of the
equation (1).
6
7
8
9
Differently, when compounds 16 and 17 are compared (Figure 9ꢀb), the electron withdrawing effect
of chlorine affecting the carbonyl electrophilicity seems to be responsible for the different hAOX
liability. Therefore, even small structural changes can make a significant difference in the metabolic
biochemical response. Similar considerations can be made for the cases reported in Figure 9ꢀc and
9ꢀd related to the oxidation reaction.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
The weights of the exposure and reactivity effects in equation 1 were accurately computed taking
into consideration our dataset of more than 600 molecules with different scaffolds, sizes, and
shapes; different substituents have been designed, synthesized, or acquired, and lastly
experimentally tested, to significantly enlarge the chemical diversity of hAOX substrates.17
Once the models of exposure and reactivity had been introduced in MetaSite, this procedure was
used to predict the substrate selectivity and the site of metabolism for a series of xenobiotics, for
which experimental data were produced in house using the LCꢀMS/MS technique. Consequently, it
is worth recalling that MetaSite is not a dependent trainingꢀset method as the two functions used in
PSoM are obtained entirely independently from the experimental results and only the weights of the
functions were optimized to improve the recalculation of the experimental data.
In summary, the metabolism of azaheterocycle compounds is a complex function of several factors
that range from the atomic charge at the most positive CꢀH, presence and effects of substituents
(EDG or EWG) in the azaheterocycle ring, and protonation of the azaheterocycle for compounds
with basic centers having a pKa greater than 7.0. Furthermore, hindered and hydrophobic
substituents are often able to dramatically change the substrate exposure, thus enhancing the
metabolic stability. This is often caused by interfering with the water network inside or in the
proximity of the catalytic center. Regarding amides, the same effects are present but they are more
relevant in the amine counterpart, which is more sensitive to hAOX susceptibility being established.
18
ACS Paragon Plus Environment

Page 19 of 36
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
Validation
The methodology was validated by comparing the experimental results obtained from more than
600 molecules with the predictions obtained from MetaSite
9
(www.moldiscovery.com/software/metasite/).39 Approximately 200 molecules experimentally
susceptible to oxidation by hAOX and approximately 45 molecules experimentally susceptible to
hydrolysis by the amide group were used together with approximately 430 molecules that, even if
having the structural requirements to be considered potential hAOX substrates, were experimentally
nonꢀ hAOX substrates. Two items of information were obtained from the in silico methodology.
Firstly, the character of the substrate or nonꢀsubstrate for a compound, and secondly the site of
metabolism (or the sites of metabolism if experimentally more than one). The prediction was
evaluated classifying compounds as “correct substrate”, “correctꢀnon substrate”, “wrong substrate”
and “wrong nonꢀsubstrate” (Table 1). For substrates, prediction was counted as “correct” when the
compound was an experimentally confirmed hAOX substrate and the site (or sites) of metabolism
were also correctly predicted; contrarily, a predicted substrate was counted as “wrong” if the
compound was experimentally nonꢀsubstrate or if it was a substrate but our algorithm failed in
assigning the site of metabolism. For nonꢀsubstrates, “correct” or “wrong” predictions were counted
when the compound was a real nonꢀsubstrate or a real substrate, respectively.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Table 1 summarizes the results obtained from the comparison of the experimental tests with those
obtained in silico. Among the tested compounds, 194 molecules presented the oxidation reaction in
one single site while six presented it in at least two sites of metabolism. In addition, 37 molecules
presented the hydrolysis reaction in one single site while seven presented both hydrolysis and
oxidation. Furthermore, the molecules studied showed great structural diversity including both
rigid and very flexible compounds with more than 10 flexible bonds not to mention a wide range of
molecular weights and great variability in lipophilicity.
19
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 20 of 36
1
2
3
Table 1. Validation results for the proposed model.
4
5
6
7
8
9
AOX metabolism
Oxidation1 Hydrolysis2
SUBSTRATE
207
305
23
44
126
3
NOT SUBSTRATE
UNCERTAIN
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
CORRECT SUBSTRATE3
CORRECT NONꢀSUBSTRATE3
WRONG SUBSTRATE3
WRONG NONꢀSUBSTRATE3
% GOOD PREDICTION
ONE SoM
192
261
15
44
122
0
44
4
88
98
37
0
194
6
TWO SoMs
TWO SOMs OX/HYDRO
1 Database of 535 compounds. 2 Database of 173. 3According to MetaSite predictions.39
7
7
Sources of experimental variability
A significant variation of hydroxylase activity analyzed in liver cytosol from human donors46 in
relation to the used substrates has been reported.47 This is probably due to a different quantity of
enzyme present in cytosol in addition to the effect of polymorphism by the human AOX1 gene.48
In order to limit the variability and standardize the experimental results obtained as much as
possible, the experimental study presented in this paper used a method for quantifying the hAOX
enzyme present in the cytosol (see Experimental Section) that facilitated the quantification of the
enzyme in both the various batches of human acquired cytosol and in human hepatocytes and the
normalization of the obtained results. Furthermore, all the compounds found to be potential AOX
substrates were also tested with a hAOX inhibitor at different concentrations, confirming the
isoform selectivity. In the opinion of the authors the database obtained is extremely homogeneous
and at the same time structurally different, and is now available to the scientific community to in
silico predict the liability of new compounds to the hAOX enzyme.
20
ACS Paragon Plus Environment

Page 21 of 36
Journal of Medicinal Chemistry
1
2
3
Chemistry
4
5
6
All synthesized compounds were obtained according to the procedures reported in Schemes 1ꢀ3.
7
8
9
According to Scheme 1, diamine 21 was obtained by nucleophilic substitution of ethylendiamine on
2ꢀchloropyrimidine. Finally, target compound 1 was obtained by Pd2(dba)2ꢀmediated coupling
between diamine 21 and triflate 22.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Scheme 1. Synthesis of compound 1.a
aReagents and conditions: a) reflux, 2 h (62%); b) Triflic anhydride, Pyridine, DCM, 0 °C, 2 h
(99%); c) Pd2(dba)3, tBuXPhos, Cs2CO3, dioxane, reflux overnight (7%).
Amine 4 was obtained in quantitative yield by LiAlH4 reduction of amide 3, in turn prepared by
HATUꢀpromoted coupling between 1Hꢀindoleꢀ2ꢀcarboxylic acid and 2ꢀ(pyrrolidinꢀ1ꢀyl)aniline.
(Scheme 2).
Scheme 2. Synthesis of amine 3.a
aReagents and conditions: a) HATU, DIPEA, DMF rt (48%); b) LiAlH4, THF, 0 °C (99%).
21
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 22 of 36
1
2
3
Nꢀmethyl amides 16 and 17 were obtained by methylation of amides 23a-b (previously
4
5
synthesized),17 respectively (Scheme 3).
6
7
8
Scheme 3. Synthesis of amides 16 and 17.
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
CONCLUSIONS
hAOX is a cytosolic enzyme able to metabolize pharmaceutical drugs and has currently become one
of the main themes of interest in Medicinal Chemistry, Drug Metabolism and Pharmacokinetics.
Numerous original research papers focusing on the role of hAOX in the metabolism of xenobiotics
have been, in fact, published in the last years. This high level of interest has been caused by the
numerous unfortunate clinical outcomes, mainly given to unacceptable pharmacokinetic properties
and drug safety problems, and often related to an increased liability to hAOX. Interestingly, the
highest sensitivity to hAOX seems to be related to the modifications drug candidates generally
undergo to escape CYP metabolism. In this situation, a practical and fast in silico strategy to predict
hAOX substrate selectivity and to identify the site of metabolism, for both the wellꢀknown
oxidation reactions and for the less known hydrolysis, is urgently needed. The numerosity and
diversity of the used dataset, together with the homogeneity of the produced experimental
information should guarantee the reliability of the presented methodology (included in the MetaSite
procedure) for all typical drug discovery and new chemical entities optimization projects. Applied
in an early phase, the method should facilitate a good understanding of the passivity of hAOX ꢀ
catalyzed metabolism and suggest potential modifications to increase the stability of a xenobiote to
hAOX ꢀcatalyzed reactions.
22
ACS Paragon Plus Environment

Page 23 of 36
Journal of Medicinal Chemistry
1
2
3
4
5
EXPERIMENTAL SECTION
Chemistry
6
7
8
9
All commercial products were acquired from Sigma Aldrich and used without further purification
(≥ 97% pure). Anhydrous solvents were acquired from Acros Organics. Compounds 23a-b were
obtained as previously reported,17 from HATUꢀmediated coupling between picolinic acid and 2ꢀ or
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
1
3ꢀchloroaniline, respectively. H and 13C NMR nuclear magnetic resonance (NMR) spectra were
recorded at 400 and 100.6 MHz, respectively, on Bruker Avance III HD 400 MHz spectrometer at
room temperature. Chemical shifts (δ) are given in parts per million (ppm) relative to the internal
standard tetramethylsilane. Peak multiplicities are reported as s (singlet), d (doublet), dd (double
doublet), t (triplet), dt (double triplet), q (quartet), m (multiplet), or br s (broad singlet). Coupling
constants (J) are given in Hz. HRMS spectra were registered on Agilent Technologies 6540 UHD
Accurate Mass QꢀTOF LC/MS system. Purity for all final compound was determined by LCꢀ
MS/MS and resulted >98%.
General Procedure A for the synthesis of amides.
HATU
(1ꢀ[Bis(dimethylamino)methylene]ꢀ1Hꢀ1,2,3ꢀtriazolo[4,5ꢀb]pyridinium
3ꢀoxid
hexafluorophosphate, 1.25 mmol) was added at rt to a stirred solution of suitable carboxylic acid
(1.10 mmol) and DIPEA (N,Nꢀdiisopropylethylamine, 2.25 mmol) in anhydrous N,Nꢀ
dimethylformamide (1.00 mL). After 6 min, the amine was added (1.00 mmol) and stirring was
continued overnight. The mixture was dropped in water (10 mL), the resulting solid filtered off,
washed with water and dried. The solid was purified by column chromatography to give the desired
amide.
N1-(naphthalen-2-yl)-N2-(pyrimidin-2-yl)ethane-1,2-diamine (1). Pd2(dba)3 (18.0 mg, 0.02
mmol) was added to a degased solution of amine 21 (0.150 g, 1.1 mmol), trifluoromethanesulfonate
22 (0.276 g, 1.0 mmol), Cs2CO3 (0.487 g, 1.5 mmol), and tBuXPhos (13.0 mg, 0.03 mmol) in
23
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 24 of 36
1
2
3
4
5
6
7
8
9
anhydrous dioxane (10 mL). The reaction was stirred under reflux for 20 h, cooled to rt, diluted
with dichloromethane and filtered over a celite pad. The solution was concentrated and the residue
purified by flash chromatography on SiO2 (eluent: dichloromethane, followed by
dichloromethane/methanol, 9:1). The resulting oil was further purified by flash chromatography
(eluent petroleum ether/ethyl acetate, from 6:4 to 4:6). The titled product was obtained as a yellow
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
1
solid (19.1 mg, 7% yield). H NMR (400 MHz, CDCl3) δ 8.25 (d, J = 4.8 Hz, 2H), 7.64 – 7.50 (m,
3H), 7.34 – 7.24 (m, 1H), 7.14 (t, J = 7.5 Hz, 1H), 6.82 (dd, J = 2.3, 11.0 Hz, 2H), 6.57 – 6.47 (m,
1H), 5.85 (s, 1H), 3.71 (q, J = 6.0 Hz, 2H), 3.60 (bs, 1H), 3.45 (t, J = 5.9 Hz, 2H); 13C NMR (101
MHz, CDCl3) δ 162.6, 158.0, 145.8, 135.2, 128.8, 127.5, 127.4, 126.2, 125.8, 121.8, 118.0, 110.8,
104.1, 43.9, 40.6, 23.9; HRMS: calcd. for C16H16N4 265.1148 (M+H+), found 265.1148 (M+H+).
N-((1H-indol-2-yl)methyl)-2-(pyrrolidin-1-yl)aniline (3). LiAlH4 was added under argon at 0 °C
to a stirred solution of amide 4 in anhydrous tetrahydrofuran. After stirring at rt overnight, the
reaction was carefully quenched under nitrogen with ethyl acetate, followed by water. The aqueous
phase was extracted with ethyl acetate (x3), and reunited organic phases were evaporated to
dryness. Flash column chromatography on SiO2 (eluent dichloromethane /methanol 95:5) allowed
1
to recover the titled product as white solid, which was characterized as follows. H NMR (400
MHz, CDCl3) δ 8.36 (s, 1H), 7.63 (d, J = 7.7 Hz, 1H), 7.35 (d, J = 8.1 Hz, 1H), 7.26 – 7.09 (m, 3H),
7.00 (td, J = 1.2, 7.6 Hz, 1H), 6.87 – 6.67 (m, 2H), 6.57 – 6.45 (m, 1H), 5.03 (t, J = 5.3 Hz, 1H),
13
4.56 (d, J = 5.5 Hz, 2H), 3.17 – 3.03 (m, 4H), 2.13 – 1.89 (m, 4H); C NMR (101 MHz, CDCl3) δ
143.5, 137.7, 137.6, 135.9, 128.8, 124.3, 121.5, 120.2, 119.8, 118.8, 118.0, 110.9, 110.8, 99.6, 51.5
(2C), 42.7, 24.2 (2C); HRMS: calcd. for C19H21N3 292.1808 (M+H+), found 292.1802 (M+H+).
N-(2-(Pyrrolidin-1-yl)phenyl)-1H-indole-2-carboxamide (4). The titled compound was obtained
according to the General Procedure A from 1Hꢀindoleꢀ2ꢀcarboxylic acid and 2ꢀ(pyrrolidinꢀ1ꢀ
yl)aniline. After column chromatography (eluent dichloromethane), the product was obtained as a
1
turquoise solid (48% yield). H NMR (400 MHz, CDCl3) δ 9.37 (s, 1H), 9.11 (s, 1H), 8.38 (d, J =
7.5 Hz, 1H), 7.74 (d, J = 8.0 Hz, 1H), 7.50 (d, J = 8.3 Hz, 1H), 7.35 (t, J = 7.6 Hz, 1H), 7.26 – 7.08
24
ACS Paragon Plus Environment

Page 25 of 36
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
(m, 4H), 6.94 (s, 1H), 3.13 (s, 4H), 2.07 (s, 4H); 13C NMR (101 MHz, CDCl3) δ 162.6, 142.9,
139.8, 138.5, 131.3, 125.6, 125.2, 121.7, 120.8, 120.7, 119.8, 118.4, 114.9, 111.1, 110.7, 54.4 (2C),
25.5 (2C); HRMS: calcd. for C19H19N3O 306.1601 (M+H+), found 306.1608 (M+H+).
9
N-(4-(dimethylamino)phenyl)-4,5,6,7-tetrahydrobenzo[b]thiophene-2-carboxamide (14). The
titled compound was obtained according to the General Procedure A from 4,5,6,7ꢀ
tetrahydrobenzo[b]thiopheneꢀ2ꢀcarboxylic acid and N1,N1ꢀdimethylbenzeneꢀ1,4ꢀdiamine. After
column chromatography (eluent dichloromethane /ethyl acetate, 9:1), the product was obtained as a
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
1
yellow solid (27% yield). H NMR (400 MHz, Chloroformꢀd) δ 7.60 – 7.40 (m, 3H), 7.29 (d, J =
3.8 Hz, 1H), 6.75 (d, J = 8.9 Hz, 2H), 2.95 (s, 6H), 2.82 (t, J = 5.8 Hz, 2H), 2.64 (t, J = 5.8 Hz, 2H),
2.05 – 1.68 (m, 4H); 13C NMR (101 MHz, Chloroformꢀd) δ 160.09, 148.02, 141.51, 136.31, 134.86,
129.24, 127.68, 121.95 (2C), 113.12 (2C), 40.96 (2C), 25.47, 25.34, 23.28, 22.62; HRMS: calcd.
for C17H20N2OS 301.1369 (M+H+), found 301.1379 (M+H+).
N-(2-(Pyrrolidin-1-yl)phenyl)-4,5,6,7-tetrahydrobenzo[b]thiophene-2-carboxamide (15). The
titled compound was obtained according to the General Procedure A from 4,5,6,7ꢀ
tetrahydrobenzo[b]thiopheneꢀ2ꢀcarboxylic acid and 2ꢀ(pyrrolidinꢀ1ꢀyl)aniline. After column
chromatography (eluent dichloromethane), the product was obtained as a yellow solid (48% yield).
1H NMR (400 MHz, Chloroformꢀd) δ 8.84 (s, 1H), 8.35 (d, J = 7.8 Hz, 1H), 7.32 (s, 1H), 7.18 (d, J
= 7.7 Hz, 1H), 7.24 – 6.87 (m, 2H), 3.08 (t, J = 5.9 Hz, 4H), 2.83 (t, J = 5.8 Hz, 2H), 2.67 (t, J = 5.8
Hz, 2H), 2.02 (p, J = 3.1 Hz, 4H), 1.98 – 1.70 (m, 4H); 13C NMR (101 MHz, Chloroformꢀd) δ
159.84, 141.69, 140.10, 136.56, 135.16, 133.14, 129.68, 124.24, 123.91, 120.29, 119.35, 52.48
(2C), 25.53, 25.40, 24.45 (2C), 23.28, 22.62; HRMS: calcd. for C19H22N2OS 327.1526 (M+H+),
found 327.1522 (M+H+).
N-(2-chlorophenyl)-N-methylpicolinamide (16). To a solution of 23a in DMF (1 mL) was added
sodium hydride (26 mg, 60% in mineral oil, 0.64 mmol), and the mixture was stirred at room
temperature for 30 min. Methyl iodide (55 ꢂL, 0.64 mmol) was then added at 0 °C, and stirring was
continued at rt for 24 h. The reaction was quenched with water, and extracted 3 times with ethyl
25
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 26 of 36
1
2
3
4
5
6
7
8
9
acetate. The reunited organic phases were washed with water and brine, and dried over Na2SO4.
Filtration, concentration of the filtrate under vacuum, and purification of the residue by silica gel
flash column chromatography (ethyl acetate/petroleum ether, 4:6) gave the titled product as a
yellow oil (56.0 mg, 54% yield), which was characterized as follows. Major atropoisomer (89%, in
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
1
13C NMR spectra indicated with carbons C). H NMR (400 MHz, CDCl3) δ 8.25 (d, J = 4.7 Hz,
1H), 7.68 – 7.54 (m, 2H), 7.32 – 7.24 (m, 1H), 7.24 – 7.18 (m, 1H), 7.16 – 7.10 (m, 2H), 7.10 –
13
7.07 (m, 1H), 3.52 – 3.32 (m, 3H). Minor atropoisomer (11 %, in C NMR spectra indicated with
1
carbons C’). H NMR (400 MHz, CDCl3) δ 8.68 (d, J = 3.9 Hz, 1H), 7.84 (t, J = 7.3 Hz, 1H), 7.78
(d, J = 7.5 Hz, 1H), 7.52 (d, J = 7.8 Hz, 1H), 7.47 – 7.32 (m, 4H), 3.41 (s, 3H). 13C NMR (101
MHz, CDCl3) δ 168.7 (C), 153.6 (C), 148.3 (C’), 148.0 (C), 141.8 (C), 137.1 (C’), 136.1 (C), 132.2
(C), 130.5 (C’), 130.4 (C), 130.0 (C), 129.3 (C’), 129.1 (C’), 128.7 (C), 128.1 (C’), 127.4 (C), 124.8
(C’), 124.1 (C), 124.1 (C’), 123.2 (C), 39.8 (C’), 36.9 (C). Three carbons of the minor atropoisomer
are not detected. HRMS: calcd. for C13H1135ClN2O 247. 0638 (M+H+), found 247.0639 (M+H+).
N-(3-chlorophenyl)-N-methylpicolinamide (17). The titled product was obtained according to the
procedure described for 16, but starting from 23b. 1H NMR (400 MHz, CDCl3) δ 8.38 (s, 1H), 7.69
(t, J = 7.4 Hz, 1H), 7.64 – 7.49 (m, 1H), 7.25 – 7.18 (m, 1H), 7.14 (s, 3H), 6.96 (s, 1H), 3.53 (s,
13
3H); C NMR (101 MHz, CDCl3) δ 168.6, 153.7, 148.4, 145.6, 136.5, 134.4, 129.8, 126.7, 126.7,
124.9, 124.3, 123.9, 38.2. HRMS: calcd. for C13H1135ClN2O 247. 0638 (M+H+), found 247.0637
(M+H+).
N1-(pyrimidin-2-yl)ethane-1,2-diamine (21). A mixture of 2ꢀchloropyrimidine (1.02 g, 8.9 mmol)
and ethylendiamine (10 mL, 667 mmol) was refluxed for 2 h. The mixture was diluted with brine
and extracted with dichloromethane. The reunited organic phases were evaporated to dryness to
afford the titled product as a yellow oil (0.75 g, yield 62%), which was used without further
purification in the following step. 1H NMR (400 MHz, CDCl3) δ 8.28 (d, J = 4.8 Hz, 2H), 6.53 (t, J
= 4.8 Hz, 1H), 5.74 – 4.93 (m, 2H), 3.50 (q, J = 6.0 Hz, 2H), 3.06 – 2.78 (m, 2H).
26
ACS Paragon Plus Environment

Page 27 of 36
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Naphthalen-2-yl trifluoromethanesulfonate (22). Trifluoromethanesulfonic anhydride (4.00 mL,
24.0 mmol) was added dropꢀwise at 0 °C to a stirred solution of naphthalenꢀ2ꢀol (2.88 g, 20.0
mmol) and pyridine (3.23 mL, 40.0 mmol) in anhydrous dichloromethane (40 mL). The mixture
was warmed at rt and left under stirring for additional 2 h. HCl 10% aq. was added and the mixture
extracted by dichloromethane. The reunited organic phases were washed with NaHCO3 sat., dried
over Na2SO4, and concentrated under vacuum. After flash chromatography on SiO2 (eluent,
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
1
petroleum ether) a yellow solid was obtained (5.48g, 99% yield). H NMR (400 MHz, CDCl3) δ
8.11 – 7.86 (m, 3H), 7.78 (d, J = 2.5 Hz, 1H), 7.70 – 7.55 (m, 2H), 7.41 (dd, J = 2.5, 9.0 Hz, 1H).
Metabolism Assay on HLC.
Solvents, reagents, inhibitors and pooled mixedꢀgender cryopreserved HLC (1 mg/mL) were
obtained from SigmaꢀAldrich. HLM UltraPool™ HLM 150 were purchased from BD Gentest
(Woburm, MA, USA). Tested compounds for in vitro assays were acquired from SPECS,
ENAMINE, VITASꢀM and MEDCHEMEXPRESS. All tested compounds were > 98% pure as
detected by UHPLꢀLCꢀMS/MS analysis through 6540 UHD accurateꢀmass quadrupole timeꢀofꢀ
flight (QTOF) system. For AOX quantification, TRIS/HCl, urea, dithiothreitol (DTT), trypsin
(sequencing grade), iodoacetamide (IAA), HLC (Sigma Aldrich), H2O (MS grade), formic acid
(MS grade) were acquired from Sigma Aldrich and used without further purification. Standard
peptide YIQDIVASTLK was obtained provided by the Institute of Biostructure and Bioimaging
(IBB) of the Italian National Research Council, (Naples, Italy).
General Procedure for metabolism assay
Metabolism was evaluated upon incubation with HLC according to a modified Dalvie et al.30
procedure, recently described.17 Briefly, each compound (10 ꢂM) was incubated at pH 7.4 and 37
°C in the presence of magnesium chloride hexahydrate (MgCl2·6H2O, 1 mM) and HLC. The
27
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 28 of 36
1
2
3
4
5
6
7
8
9
reaction was stopped after 0 and 60 min by addition of 250 ꢂL of iceꢀcold acetonitrile (containing
0.6 ꢂM labetalol as internal standard). Corresponding blank was prepared following an identical
procedure, in the absence of tested compound. Similarly, in the inhibition study, test compound (10
ꢂM) was incubated in the presence (50 ꢂM) or absence of selective inhibitor DCPIP. Proteins were
precipitated by centrifugation at 5,000 rpm for 10 min (Eppendorf, Italy; centrifuge 5810 R; rotor Fꢀ
45ꢀ30ꢀ11) at room temperature, and an aliquot of supernatant (1 ꢂL) was analyzed by LCꢀMS/MS
on an Agilent 1200 series HPLC coupled to an Agilent 6540 UHD accurateꢀmass QTOF with a dual
Jet Stream electron spray ionization source. Depending on the nature of the substrates three
different analytical methods (1, 2, 3) were used in LC analysis. The mobile phase was a mixture of
water (solvent A) and acetonitrile (solvent B), both containing formic acid at 0.1%. Method 1: Aeris
Widepore 3.6ꢀꢂm (C4, 100 × 4.6ꢀmm) column at 30 °C using a flow rate of 0.850 mL/min in a 10ꢀ
min gradient elution. Gradient elution: 100:0 (A/B) to 70:30 (A/B) over 9 min, 5:95 (A/B) for 1
min, and then reversion back to 100:0 (A/B) over 0.1 min. Method 2: Acquity UPLC BEH C18 1.7ꢀ
ꢂm (C18,150 ×2.1ꢀmm) column at 40°C using a flow rate of 0.650mL/minina10ꢀmin gradient
elution. Gradient elution: 99.5:0.5 (A/B) to 5:95 (A/B) over 8 min, 5:95 (A/B) for 2 min, and then
reversion back to 99.5:0.5 (A/B) over 0.1 min. Method 3: Luna® Omega 1.6 um Polar (C18, 100 x
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
2.1 mm) column at 40 ⁰C using a flow rate of 0.650 mL/min in a 10 minutes gradient elution.
Gradient elution: 99.5:0.5 (A/B) to 50.0:50.0 (A/B) over 8 minutes, 50.0:50.0 (A/B) for 2 minutes,
and then reversion back to 99.5:0.5 (A/B) over 0.1 min. The MS/MS data were processed using
MassꢀMetaSite (version 3.3.2) (MolecularDiscovery, Ltd., UK).6 LCꢀMS chromatograms in Figure
S4ꢀ1 were extracted by MassHunter (Mass Hunter Workstation Software B.06.00 Qualitative
Analysis; Agilent).
Oxypurinol was also detected in the HLC containing samples, indicating that the contribution of
XO to observed metabolism is negligible, as described by Barr and coworkers.49
28
ACS Paragon Plus Environment

Page 29 of 36
Journal of Medicinal Chemistry
1
2
3
hAOX quantification
4
5
6
7
8
50
HLC was digested following Barr et al. method with modifications. Briefly, 20 mg/ml of HLC
(25 µL) were mixed with 8 M urea, 50 mM DTT and incubated for 1 h at 56 °C. 200 mM
iodoacetamide was added to the mixture and incubated in the dark at 37 °C for 30 min.
Subsequently the mixture was diluted with 0.1 M TRIS/HCl buffer containing sequence grade
trypsin (0.25 µg/µL) and incubated overnight at 37 °C. Digestion was terminated by adding a
solution of 50% v/v H2O/HCO2H (final pH <2). Aliquots of supernatant (5 µL) were analysed by
LCꢀMS/MS. Quantification of digested proteins was performed by MassHunter Quantitative
Analysis B.06.00 (Agilent Technologies, CA) using the standard peptide YIQDIVASTLK provided
by IBB (Italian National Research Council, Naples, Italy).
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
Modelling and Molecular Dynamics
The missing loops in the original xꢀray structure were built with the homology modelling program
MODELLER (http://salilab.org/modeller/). The xꢀray human structure of AOX (pdb code 4UHX)
was used as template upon removal of the coꢀcrystallized ligands and of the malonate. The standard
protonation state at physiological pH was assigned to ionizable residues.
The parameters for the FeS group and for the MoCo cofactor were kindly provided by Cerqueira et
al.51 The FAD cofactor was parameterized within the software BiKi52 used to setup the MD
simulation.
The protein was parametrized by Amber99SBꢀILDN force Field53 and Gromacs 5.1.4 was used to
run MD Simulations.54 The water model employed was TIP3P. The solvated system was
preliminary minimized by 5000 steps of steepest descent. The Verlet cutoff scheme, the
Bussi−Parrinello thermostat LINCS for the constraints (all bonds), and the particle mesh Ewald for
electrostatics, with a shortꢀrange cutoff of 11 Å, were applied. The system was equilibrated in seven
subsequent steps: 500 ps in NVT ensemble at 50 K, 500 ps in NVT ensemble at 100 K, 500 ps in
29
ACS Paragon Plus Environment