Species-Selective Pyrimidineamine Inhibitors of Trypanosoma brucei S-Adenosylmethionine Decarboxylase

Article abstract of DOI:10.1021/acs.jmedchem.7b01654

New therapeutic options are needed for treatment of human African trypanosomiasis (HAT) caused by protozoan parasite Trypanosoma brucei. S-Adenosylmethionine decarboxylase (AdoMetDC) is an essential enzyme in the polyamine pathway of T. brucei. Previous attempts to target this enzyme were thwarted by the lack of brain penetration of the most advanced series. Herein, we describe a T. brucei AdoMetDC inhibitor series based on a pyrimidineamine pharmacophore that we identified by target-based high-throughput screening. The pyrimidineamines showed selectivity for T. brucei AdoMetDC over the human enzyme, inhibited parasite growth in whole-cell assay, and had good predicted blood-brain barrier penetration. The medicinal chemistry program elucidated structure-activity relationships within the series. Features of the series that were required for binding were revealed by determining the X-ray crystal structure of TbAdoMetDC bound to one analog. The pyrimidineamine series provides a novel starting point for an anti-HAT lead optimization.

Full text of DOI:10.1021/acs.jmedchem.7b01654

Subscriber access provided by READING UNIV
Article
Species-Selective Pyrimidineamine Inhibitors of
Trypanosoma brucei S-Adenosylmethionine Decarboxylase
Oleg A. Volkov, Anthony J. Brockway, Stephen A Wring, Michael
Peel, Zhe Chen, Margaret A. Phillips, and Jef K. De Brabander
J. Med. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.jmedchem.7b01654 • Publication Date (Web): 22 Dec 2017
Downloaded from http://pubs.acs.org on December 23, 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 80
Journal of Medicinal Chemistry
1
2
3
4
Species-Selective Pyrimidineamine Inhibitors of Trypanosoma brucei S-
5
6
Adenosylmethionine Decarboxylase.
7
8
9
Oleg A. Volkov,1,& Anthony J. Brockway,2,& Stephen A. Wring,3 Michael Peel,3 Zhe Chen,4
Margaret A. Phillips,1,2,* and Jef K. De Brabander2,*
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
Departments of 1Pharmacology, 2Biochemistry, and 4Biophysics, University of Texas
Southwestern Medical Center, 5323 Harry Hines Blvd., Dallas, Texas 75390ꢀ9038, United States
3Scynexis, Inc. (now Avista Pharma Solutions), 3501 Tricenter Boulevard, Suite C, Durham,
North Carolina 27713, United States
&These authors contributed equally to this work
*Coꢀcorresponding Authors: J.K.D.B. and M.A.P.
Emails: Jef.DeBrabander@UTSouthwestern.edu and Margaret.Phillips@UTSouthwestern.edu
Phones: (214) 648ꢀ7808 and (214) 645ꢀ6164
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 2 of 80
1
2
3
4
Abstract
5
6
7
8
New therapeutic options are needed for treatment of human African trypanosomiasis (HAT)
caused by protozoan parasite Trypanosoma brucei. SꢀAdenosylmethionine decarboxylase
(AdoMetDC) is an essential enzyme in the polyamine pathway of T. brucei. Previous attempts to
target this enzyme were thwarted by the lack of brain penetration of the most advanced series.
Herein, we describe a T. brucei AdoMetDC inhibitor series based on a pyrimidineamine
pharmacophore that we identified by targetꢀbased highꢀthroughput screening. The
pyrimidineamines showed selectivity for T. brucei AdoMetDC over the human enzyme,
inhibited parasite growth in wholeꢀcell assay, and had good predicted blood–brain barrier
penetration. The medicinal chemistry program elucidated structure–activity relationships within
the series. Features of the series that were required for binding were revealed by determining the
Xꢀray crystal structure of TbAdoMetDC bound to one analog. The pyrimidineamine series
provides a novel starting point for an antiꢀHAT lead optimization.
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
Keywords: human African trypanosomiasis, Trypanosoma brucei, AdoMetDC,
pyrimidineamine, coꢀcrystallization.
ACS Paragon Plus Environment

Page 3 of 80
Journal of Medicinal Chemistry
1
2
3
4
Introduction
5
6
7
8
Human African trypanosomiasis (HAT), also known as sleeping sickness, is caused by the
single celled protozoal parasite Trypanosoma brucei and is classified by the WHO as a
Neglected Tropical Disease (NTD).1 HAT is transmitted by the tsetse fly to humans and animals
where it replicates extracellularly in the blood and lymph causing fever and influenzaꢀlike
symptoms. Later in infection, the parasite crosses the blood–brain barrier (BBB) and invades the
central nervous system (CNS) leading to a variety of neurological symptoms, including
disruption of the sleep–wake cycle, coma, and eventual death in most patients.2 The number of
reported cases has declined steadily since peaking in 1998 at nearly 40,000 cases a year.3, 4
Control and surveillance programs, which include both vector control and identification and
treatment of infected individuals, have been credited with this decline.3 The WHO has targeted
HAT for sustainable elimination by 2030;5 however, the recent discovery that some infected
individuals are asymptomatic carriers threatens the elimination program.6 Additionally, it is now
recognized that skin and fat serves as a significant reservoir for the parasites and thus screening
of blood underestimates the magnitude of the parasite burden in the endemic communities.7ꢀ9
Currently approved HAT treatments have significant liabilities hindering efforts to control
the disease.3 Particularly problematic are treatments for the CNS stage of the disease.
Nifurtimox–eflornithine combination therapy (NECT) has replaced the poorly tolerated
melarsoprol as a frontꢀline therapy against Gambian form of the infection.10 However, both the
high cost of this therapy 10 and its limited efficacy against Rhodesian form of HAT has led to
continued use of melarsoprol.11 Thus, it is recognized that a safer and less expensive therapy is
needed that would cure both earlyꢀ and CNSꢀstage infections caused by either T. brucei
subspecies, eliminating the need to stage the disease. There are currently only two drug
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
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 4 of 80
1
2
3
4
5
6
7
8
candidates in clinical development: oxaborole SCYXꢀ715812 and nitroheterocyclic
fexinidazole.13 Both have the potential to meet the outlined treatment goals but it is too early to
know if either will make it to registration.
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
Of the approved drugs only eflornithine has a well understood mechanism of action. It
targets ornithine decarboxylase, a key enzyme in polyamine biosynthesis, suggesting that other
enzymes in the pathway would also provide potential targets for drug discovery.14, 15 A second
key enzyme in the pathway, Sꢀadenosylmethionine decarboxylase (AdoMetDC), has been shown
to be essential by genetic studies,16 and smallꢀmolecule inhibitors of AdoMetDC have been
reported that show in vivo activity against T. brucei in mouse models.14, 17ꢀ19 However, T. brucei
AdoMetDC inhibitors, such as CGP 40215, mechanismꢀbased MDL 73811, and its derivative
Genzꢀ644131 (1), were deemed unsuitable for antiꢀHAT development as their physicochemical
properties were not consistent with good blood–brain barrier (BBB) penetration and thus they
were not effective for CNS stage of the disease.14, 17, 18 AdoMetDCs from trypanosomatids have a
novel subunit configuration that differentiates them from the human enzyme: they are
allosterically activated by heterodimerization with an inactive paralogous pseudoenzyme, we
termed prozyme.20 Structural analysis by Xꢀray crystallography demonstrated that
heterodimerization leads to displacement of an autoinhibitory sequence and to a coupled
structural reorganization that stabilizes the active conformation through insertion of the Nꢀ
terminus into the heterodimer interface.21 These structural differences suggested that selective
inhibition of the parasite enzyme over the host enzyme was plausible.
We recently described mass spectrometryꢀbased highꢀthroughput screening (HTS) campaign
that identified 13 classes of novel smallꢀmolecule inhibitors of T. brucei AdoMetDC that also
inhibited parasite growth.22 Several of the identified series demonstrated high propensity to cross
ACS Paragon Plus Environment

Page 5 of 80
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
the BBB and were at least partially onꢀtarget in the parasite cells.22 Herein, we characterize a
pyrimidineamine chemotype that was identified as a hit in the HTS. Initially, the hit failed to
demonstrate substantial AdoMetDC inhibition in validation studies and thus was not previously
reported. We later discovered that the potency of the initial hit compound was pHꢀdependent,
triggering reꢀprioritization of the pyrimidineamines for a medicinal chemistry effort to establish
the structure–activity relationships (SAR) within the series. A number of pyrimidineamine
analogs with lowꢀmicromolar activity on both the enzyme and the parasites were identified.
Compounds in the series also showed good selectivity versus the human enzyme and were
predicted to cross the blood–brain barrier. Finally, a crystal structure of TbAdoMetDC bound to
one analog was determined, providing insight into the structural basis for the pH dependence of
inhibitor binding to T. brucei AdoMetDC, and paving the way for future lead development. This
work has led to the validation of the pyrimidineamines as a hit series that could be taken forward
into a hitꢀtoꢀlead optimization program for the treatment of HAT.
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
Results
Synthesis. Compound 7 was purchased from Maybridge (part of Thermo Fisher Scientific),
and compounds 56-59, 74-88, and 90 were purchased from ChemBridge (San Diego, CA, USA).
The synthesis of all other analogs is shown in Scheme 1.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 6 of 80
1
2
3
4
Results
5
6
7
8
Synthesis. Compound 7 was purchased from Maybridge (part of Thermo Fisher Scientific),
and compounds 56-59, 74-88, and 90 were purchased from ChemBridge (San Diego, CA, USA).
The synthesis of all other analogs is shown in Scheme 1.
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
Scheme 1. Synthesis of pyrimidineaminesa
H
N
H
R1
R1
R3
H2N
N
Cl
H2N
N
Cl
Y
Cl
Y
N
a, or b, or c,
or c-d, or c-e
b, or c,
18-27
N
N
X
X
R1
R1
or f, or g
R2
R2
R2
R2
Cl
Cl
2
8-10, 13-16
3
17-27
a; R1 = R2 = H
R1 = H, Me; R2 = H, Me, Et, iPr,
Cl, CF3; R1-R2 = -(CH2)3-
a; X = N, Y = CH, R1 = NH2, R2 = Me
b; X = N, Y = CH, R1 = NH2, R2 = H
c; X = N, Y = CNH2, R1 = R2 = H
d; X = Y = N, R1 = H, R2 = Me
e; X = Y = N, R1 = NHMe, R2 = Me
f; X = Y = N, R1 = NMe2, R2 = Me
g; X = Y = N, R1 = R2 = NH2
h; X = Y = N, R1 = Me, R2 = NH2
i; X = CH, Y = N, R1 = H, R2 = NH2
j; X = CH, Y = N, R1 = Me, R2 = NH2
X = N, CH; Y = N, CH, CNH2;
R1 = NH2, H, NHMe, NMe2, Me;
R2 = H, Me, NH2; R3 = Cl, Br
b; R1 = H, R2 = Et
c; R1 = H, R2 = iPr
d; R1 = H, R2 = Cl
e; R1 = H, R2 = CF3
f; R1-R2 = -(CH2)3-
g; R1 = R2 = Me
11-12 (from 13)
R1 = H; R2 = Ph, N-morpholino
17
h
AcHN
N
Br
4
H
32-34, 36-48, 52-55, 60-61
R2 = Cl, Br, Me, Et, tBu, CF3,
CN, OMe, OEt, OBu, OCH2CF3,
OCH2CHCH2, OCH2iPr, SF5,
SO2Me, CO2H;
R4 = F, Cl, Br, Me, tBu, CF3,
OMe, SO2Me, CO2H
H2N
N
N
H2N
N
X
Cl
H2N
N
Cl
R2
c, or i, or j
b, or c
N
N
N
6 (R1 = R3 = H)
5
R
Cl
Me
Cl
Me
R4
Me
a; R = NHMe
b; R = S
28-31
2h
c; R = -CH2NH2
d; R = -(CH2)2NH2
X = NMe-, S, NHCH2,
NH(CH2)2
Cl
R1
d, or m
d
32
R
35, 91-117
R2
R3
H2N
2 = Cl,
R2 = Cl; R4 = Ph,
Aryl, heteroaryl
b, or
c
R4 = Br
6
38
118
R4
R2 = Et,
R4 = Br
R
2 = Et; R4 = 4-F-Ph
62-73, 89
k, or l
R1
1 = H, OH, F;
H
48
49, 51
R
R2
R3
H2N
N
N
R2 = R4
= CO2H
R2 = R4 = CO2Me,
C(O)NH(3,5-Cl2Ph
R2 = H, Cl, Br;
R3 = H, F, Cl, Me, OH,
OPh, OCHF2, NMe2,
N
d
7
50
R4
R2 = R4
= Cl
R2 = R4 = Ph
Me
N-morpholino, N-piperidino, 3,5-Cl2Ph;
R4 = H, Cl, Br
a Reagents and conditions: (a) 2a, 3,5ꢀCl PhNH , DMSO, 100 °C, 18 h (8); (b) 2a-c,f-h or
2
2
3g, aniline (3,5ꢀCl2PhNH2 or 3ꢀBrꢀ5ꢀClPhNH2 or 6), conc. HCl, EtOH, microwave (140 °C), 1.5
h (9, 10, 15, 16, 26, 32-34, 36, 37, 44, 45, 53, 54, 61-68, 72, 73, 89); (c) 2a,d,e,h or 3d-f,h,
aniline (3,5ꢀCl PhNH or 3ꢀBrꢀ5ꢀClPhNH2 or 5a or 6) conc. HCl, EtOH, reflux, 18 h (13, 14,
2
2
23-25, 27, 28, 38-43, 46-48, 52, 55, 60, 69-71); (d) 7 or 13 or 32 or 38, ArylB(OH) , Pd(OAc) ,
2
2
PPh , Na CO , H O/DME, 90 °C, 18 h (11, 35, 50, 91-109, 116-118); (e) 13, DIPEA,
3
2
3
2
ACS Paragon Plus Environment

Page 7 of 80
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
morpholine, CH3CN, microwave (140 °C), 2 h (12); (f) 3,5ꢀCl2PhNH2, 3a,c,i,j, conc. HCl,
BuOH, microwave (185 °C), 4 h (18-20, 22); (g) 3,5ꢀCl2PhNH2, 3b, conc. HCl, MeOH/H2O
(1/1), sealed tube, 100 °C, 18 h (21); (h) 4, 3,5ꢀCl2PhNH2, cat. Pd2(dba)3, BINAP, NaOtBu,
toluene, 70 °C, 18 h; then 3N aq. NaOH, EtOH, reflux, 18 h (17); (i) 2h, 5b, Et3N, DMF, rt, 24 h
(29); (j) 2a, 5c,d, DIPEA, CH3CN, microwave (140 °C), 2 h (30, 31); (k) 48, conc. H2SO4,
MeOH, 4Å mol. sieves, reflux, 18 h (49); (l) 48, 3,5ꢀCl2PhNH2, EDI, py/DCM/DMF (0.4/1/1), rt,
18 h (51); (m) 32, heteroaryl boronic acid MIDAꢀester, cat. Pd(OAc)2, PPh3, sodium
triphosphate, dioxane/H2O) (4/1), 90 °C, 18 h (110-115).
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
Compounds 9ꢀ10, 13ꢀ16, 18ꢀ28, 32ꢀ34, 36ꢀ48, 52ꢀ55, 60ꢀ73, and 89 were readily available
via reaction of the appropriate chloropyrimidine (2a-h, 3d-h) or chloropyridine (3a-c, 3i-j) with
a variety of substituted anilines in the presence of acidic ethanol or butanol under conventional23
or microwave heating.24 Compound 8 was obtained similarly but under neutral conditions via
heating of chloropyrimidine 2a with 3,5ꢀCl2PhNH2 in DMSO at 100 °C, whereas amineꢀlinker
analogs 30ꢀ31 resulted from microwave heating of chloropyrimidine 2h with amines 5c-d in
acetonitrile in the presence of iPr2NEt.25 Stirring thiol 5b with chloropyrimidine 2h and
triethylamine in DMF at room temperature yielded thioetherꢀlinked analog 29.26 Analog 12 was
obtained from chlorideꢀdisplacement in 13 with morpholine in the presence of iPr2NEt under
microwave heating in acetonitrile.27 The bisꢀester and amide analogs 49 and 50 resulted from
Fisher esterification and EDIꢀmediated amide coupling of 48, respectively. Pyridine analog 17
was obtained via palladiumꢀcatalyzed amination of pyridinylbromide 4 with 3,5ꢀCl2PhNH2
followed by baseꢀmediated hydrolysis of the acetamide.28 Finally, all biarylꢀcontaining
compounds (11, 35, 50, 91ꢀ118) were prepared via Suzuki crossꢀcoupling of the corresponding
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 8 of 80
1
2
3
4
5
arylchloride (7, 13) or arylbromide (32, 38) analogs with arylboronic acids29 or the
corresponding MIDA boronates.30
6
7
8
9
Discovery of pyrimidineamines as T. brucei AdoMetDC inhibitors. The pyrimidineamine
7 was identified as a singleton in our previously reported HTS campaign to identify inhibitors of
TbAdoMetDC (Figure 1A).22 Compound 7 did not comprise substructural motifs characteristic of
panꢀassay interference compounds (PAINS),31 according to our analysis performed as described
in ref 22. The compound was repurchased and tested for inhibition of T. brucei cell growth in
vitro demonstrating a halfꢀmaximal effective growthꢀinhibitory concentration (EC50) of 5.6 µM
after 48 h treatment. However, retesting of the TbAdoMetDC enzyme activity yielded a halfꢀ
maximal inhibitory concentration (IC50) of 30 µM (assay pH 7.7). Thus, the compound was not
presented in the original HTS publication, where the IC50 threshold for an active hit was set at 25
µM.
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
Later, characterization of the pH dependence of the TbAdoMetDC inhibition profile showed
that there was a 60ꢀfold increase in inhibitory activity of 7 at pH 6.8 compared with pH 7.7
(Figure 1B; IC50 = 30, 3.7, and 0.50 µM at pH 7.7, 7.2, and 6.8, respectively). The pKa value of
the conjugated acid of the aromatic nitrogen in position 1 (further denoted as pKa(N1)) was
estimated using MarvinSketch (version 16.3.28.0, ChemAxon, Budapest, Hungary) to be 7.3
(Figure 1A). Thus, we hypothesized that protonation of N1 in 7 was important for binding to the
T. brucei AdoMetDC. Pyrimidineamine analogs (described below) with predicted pKa(N1) of 7.3
had similar slopes of IC50–pH dependence, while analogs with pKa(N1)) > 9 had much shallower
slopes reflecting more comparable percentages of protonated species among the three pH points
(Figure 2B). IC50 values for the structurally unrelated 1 tested as a control were not appreciably
ACS Paragon Plus Environment

Page 9 of 80
Journal of Medicinal Chemistry
1
2
3
4
5
6
pHꢀdependent, as only a 2ꢀfold difference was observed between potencies at pH 6.8 and pH 7.7
(Figure 2B).
7
8
9
Encouragingly, compound 7 was predicted to have a high likelihood for the ability to cross
the BBB as assessed using an MDCKIIꢀhMDR1 monolayer cell assay in the presence or absence
of Pꢀglycoprotein 1 (Pgp) inhibitor Nꢀ(4ꢀ(2ꢀ(6,7ꢀdimethoxyꢀ3,4ꢀdihydroisoquinolinꢀ2(1H)ꢀ
yl)ethyl)phenyl)ꢀ5ꢀmethoxyꢀ9ꢀoxoꢀ9,10ꢀdihydroacridineꢀ4ꢀcarboxamide (GF120918). The assay
was performed as previously described.22 The apparent permeability of 7 was 135 nm/s ±
inhibitor, suggesting that 7 is not a substrate of Pgpꢀmediated efflux. This permeability rate is
close to an ideal threshold value of 150 nm/s cited as an indicator of a good likelihood that the
compound will cross into the CNS.32
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
Mechanism of inhibition studies. Reversibility of inhibition by 7 was tested in a rapid
dilution assay.33 TbAdoMetDC was equilibrated with 7 (36 µM ~ IC90 at pH 7.2) for 30 min and
then diluted 100ꢀfold to a concentration that was no longer inhibitory (<IC10) into reaction buffer
containing ~Km concertation (50 µM) of substrate, SꢀadenosylꢀLꢀmethionine (AdoMet), to
initiate the reaction. Aliquots were taken out at 5 min intervals, enzyme was quenched, and
produced decarboxylated AdoMet (dcAdoMet) was quantified using RapidFire–mass
spectrometry (MS) (Figure 2A). The data were compared with those obtained for the vehicle
control. Enzyme activity was fully recovered after diluting out the inhibitor following the 30 min
treatment, which is consistent with rapidly reversible inhibition.
To determine the modality of the reversible inhibition, we followed dcAdoMet produced at
varied concentrations of AdoMet and 7. The doubleꢀreciprocal Lineweaver–Burk plot of the
substrate titrations at four inhibitor concentrations is consistent with the competitive inhibition
(Ki = 2.6 (2.2–3.0) µM; 95% confidence interval in parentheses) (Figure 2B).
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 10 of 80
1
2
3
4
5
6
7
8
9
Hit expansion and development of structure–activity relationships (SAR). On the basis
of a tractable chemotype and anticipated BBB permeability, 7 was prioritized for a hitꢀtoꢀlead
expansion. The atomic structure of TbAdoMetDC was not available at the start of this project.
Thus, the medicinal chemistry effort to elucidate the SAR relied on iterative structure alterations
informed by enzyme and T. brucei cellꢀbased assays. AdoMetDC inhibitory activity was
determined at three pH values (6.8, 7.2, and 7.7) to assess the importance of N1 protonation for
binding; only pH 7.2 data are reported in the main text figures and tables. Data collected at the
alternative pH values are provided in Supplemental Information (Table S1). Additionally, the
inhibition of HsAdoMetDC at pH 7.2 was measured to assess parasite–host enzyme selectivity.
SAR studies: pyrimidine ring substituents. We began the SAR studies by modifying the
pyrimidine ring (Table 1). Replacement of the methylꢀsubstituent at C6 with ethyl (9) did not
disrupt activity in the enzyme or cellꢀbased assays, while increasing bulk at C6 by addition of
isopropyl (10) led to a ~3ꢀfold increase in the IC50 for AdoMetDC inhibition, as did C5–C6
cyclization (15) or removal of the C6ꢀMe (8). Introduction of aromatic (11), heterocyclic (12), or
electronegative (13, 14) functionality at C6 caused a sharp reduction in inhibitor potency. It is
unclear whether this drop is due to steric hindrance, adverse effect on the N1 protonation at the
assay pH, or a combination of both. Notably, there was a lack of a corresponding drop in the cell
growth assay, suggesting offꢀtarget activity. C5 could be methylated (16) without loss of enzyme
inhibitory activity.
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
ACS Paragon Plus Environment

Page 11 of 80
Journal of Medicinal Chemistry
1
2
3
4
Table 1. SAR of pyrimidine C5 and C6 positions
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
Compound R1
R2
pKa(N1)a Tb IC50,b µM Hs IC50,c µM Tb427 EC50,d µM
H
H
H
H
H
Me
H
7.3
6.6
7.1
7.0
6.1
3.7 (3.3–4.1) 27% at 180
13 (11–15) >180
3.0 (2.7–3.3) 36% at 180
5.6 (4.5–7.1)
9.9 (7.7–13)
4.7 (4.1–5.4)
4.7 (3.9–5.8)
7.2 (6–8.5)
13 (10–17)
8.9 (7.3–11)
10 (7.5–14)
19 (15–24)
17 (14–20)
7
8
9
10
11
12
13
14
15
16
Et
iPr
Ph
12 (10–14)
24% at 21
>180
>180
>180
12 (11–14)
>180
>21
>180
>180
>180
>21
H Nꢀmorpholino
H
H
6.0
Cl
CF3
ꢀ(CH2)3ꢀ
Me
–2.2
–3.3
6.8
Me
7.0
4.8 (4.5–5.2) 25% at 180
a The pKa value of the conjugated acid of the pyrimidine nitrogen N1 estimated using
MarvinSketch. b Mean IC50 for T. brucei AdoMetDC/prozyme at pH 7.2 determined using the
RapidFire–MSꢀbased enzyme activity assay, in triplicate, with 95% confidence interval (CI)
shown in parentheses. c Mean IC50 for human AdoMetDC at pH 7.2 determined in the
RapidFire–MSꢀbased enzyme activity assay, in triplicate, with 95% CI shown in parentheses. d
Mean EC50 as measured in the ATP–bioluminescenceꢀbased bloodstream form T. brucei Lister
427 cell viability assay, in triplicate, with 95% CI shown in parentheses. Percent inhibition at the
maximal tested concentration (in µM) is shown when maximum mean inhibition was <50%.
SAR studies: pyrimidine ring nitrogens and amino group. Next, compounds were
prepared to probe the function of the pyrimidine ring nitrogen atoms and C2ꢀamino group (Table
2). Substitution of N1 with CH (17 vs. 8) led to complete loss of enzyme inhibitory activity
consistent with the hypothesized importance of protonated NH+ base for binding. This activity
loss of 1ꢀdeaza analogs could not be rescued by moving the C2ꢀamino group to the C6 position
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 12 of 80
1
2
3
4
5
6
7
8
9
(18 and 19). Conversely, replacement of N3 with CH led to only a 2–4ꢀfold decrease in
inhibitory activity (20 and 21 compared with the respective parent compounds 7 and 8).
Interestingly, the pKa(N1) for compounds with CH in place of N3 was predicted to increase by
2–3 units compared with the parent compounds. Thus the concentration of the N1ꢀprotonated
species was expected to be >99% at pH 7.2, compared with only 55% or 20%, respectively, for
the parent compounds, which was predicted to led to an increase in inhibitory activity. These
data suggest that the positive gains of increased N1 protonation were offset by loss of other
interactions with N3.
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 2. SAR of pyrimidine ring nitrogens
Compound X
Y
R1 R2 R3 pKa(N1)a TbIC50,b µM Hs IC50,c µM Tb427 EC50,d µM
N
N
CH
CH
CH
N
N
N
N
N
N
NH2 Me Cl
NH2 H Cl
NH2 H Cl
7.3
6.6
NA
NA
NA
9.6
9.4
8.4
5.2
7.1
7.0
6.4
6.4
3.7 (3.3–4.1) 27% at 180
5.6 (4.5–7.1)
9.9 (7.7–13)
96 (66–140)
38 (31–47)
13 (8.2–22)
2.2 (1.7–2.9)
11 (9.5–12)
18 (14–24)
27 (22–33)
2.7 (2.4–3.1)
14 (8.1–24)
14 (12–17)
20 (17–24)
7
8
13 (11–15)
>180
>180
>180
>180
>180
>180
>180
>180
>180
>63
17
18
19
20
21
22
23
24
25
26
27
H
NH2 Cl
>180
Me NH2 Cl
24% at 180
15 (13–18)
29 (27–31)
21% at 180
>180
CH NH2 Me Cl
CH NH2 H Cl
N
N CNH2
H
H
H Cl
Me Cl
N
N
N
N
N
N
N NHMe Me Cl
>63
N
N
N
NMe2 Me Cl
NH2 NH2 Cl
Me NH2 Br
>180
48 (43–54)
24% at 180
>180
>180
>180
a, b, c ,d See footnotes for Table 1; NA, not applicable.
Deletion (23) or methylation (24 and 25) of the C2ꢀamino group led to the loss of the
enzyme inhibition. While the former decreased the pKa(N1) to 5.2, the latter had little effect on
the predicted pKa for N1 protonation but introduced steric interactions preventing optimal
ACS Paragon Plus Environment

Page 13 of 80
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
engagement of Glu266 (vide infra, Fig. 3B). Additionally, moving the C2ꢀamino group to C3 in
the background of a Me → H substitution at C6 led to further loss of inhibitory activity (22 vs.
21). Taken together, these data suggest the importance of the primary C2ꢀamine and N1 for
binding interactions, which was later supported by crystallographic data (vide infra) showing Hꢀ
bonding and saltꢀbridge of a protonated pyrimidineamine to Glu266 (Fig. 3B).
Substituting an amino group for methyl at C6 (26) or swapping the C2ꢀamino and C6ꢀ
methyl groups (27; compare with the parent compound 44 in Table 4) resulted in lower pKa(N1)
and a marked drop in the enzymatic inhibition. This is reflected in a better activity for 26 at pH
6.8 (5.8 ꢀM against Tb AdoMetDC, see Table S1). Poorer activity of 27 can be rationalized by
the loss of the C2ꢀamino group essential for interaction with Glu266 (vide infra, Fig. 3B).
Notably, most compounds that showed poor enzyme inhibitory activity, with the exception
of 17, retained antiꢀtrypanosomal activity, suggesting the presence of at least a partial offꢀtarget
component in the mechanism of the cell growth inhibition.
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
SAR studies: linker variation. Having established the pyrimidine ring traits required for
activity, we shifted our focus to understanding the role of the linker (Table 3).
Table 3. SAR of the linker
Compound
ꢀXꢀ
pKa(N1)a Tb IC50,b µM Hs IC50,c µM
Tb427 EC50,d µM
ꢀNHꢀ
ꢀNMeꢀ
ꢀSꢀ
7.3
7.2
4.3
8.1
8.2
3.7 (3.3–4.1)
81 (74–89)
>180
139 (130–150)
28 (25–32)
27% at 180
>180
>180
>180
29% at 180
5.6 (4.5–7.1)
38 (34–43)
>25
11 (9.9–13)
5.7 (1.6–21)
7
28
29
30
31
ꢀNH(CH2)ꢀ
ꢀNH(CH2)2ꢀ
a, b, c, d See footnotes for Table 1.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 14 of 80
1
2
3
4
5
6
7
8
9
Methylation of the linking amine (28) resulted in a sharp decrease in inhibitory activity in
both enzyme and cell growth assays. The replacement of the amine with a thiol linker (29)
eliminated enzyme inhibitory activity completely, presumably due to the effect on N1 acidity
(pKa(N1) = 4.3). Extension of the linker decreased potency to varying degrees. Addition of a
single methylene (30) led to a 40ꢀfold decrease in enzyme inhibition, while extension by two
methylene units (31) decreased enzyme inhibition only by 8ꢀfold.
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
SAR studies: aniline ring substitutions. We next focused our efforts on elucidating the
role of the substituted aniline ring. Both the size and electronic nature of the substituents in the
meta positions were probed first (Table 4). Replacement of one of the two chlorines in 7 with
moderately sized or large functional groups (Br, CN, CF3, Ph in 32–35) was well tolerated and
these analogs had TbAdoMetDC IC50 values similar to the parent dichloro compound 7. In
contrast, pentafluorosulfanyl substitution led to 4ꢀfold increase in the IC50 for TbAdoMetDC
inhibition (36 and 37). Since Br was a good substituent for Cl, we used 3ꢀbromo analogs to probe
the second meta positions of the aniline ring with alkyl and Oꢀalkyl functionalities of varying
length, including branched and unsaturated chains (39–43). All of these analogs showed T.
brucei AdoMetDC inhibitory activity. The trifluoroethoxy of 40 conferred the highest potency of
the tested substitutions, showing 3.8ꢀfold lower IC50 than the similarly sized ethoxy of 39. This
underlines the importance of electronics in the terminal position of the substituent. Linear
allyloxy and butoxy of 41 and 42, respectively, performed 2.4ꢀfold better than isobutoxy of 43
suggesting a preference for linear over branched substituents. Also, ethyl in 38 conferred activity
similar to that of ethoxy in 39, suggesting that the phenolic ether was dispensable.
ACS Paragon Plus Environment

Page 15 of 80
Journal of Medicinal Chemistry
1
2
3
4
Table 4. SAR of meta aniline ring substitutions
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
Compound
R2
R4
Tb IC50,b µM Hs IC50,c µM Tb427 EC50,d µM
Cl
Br
CN
CF3
Ph
SF5
SF5
Et
OEt
Cl
Cl
Cl
Cl
Cl
Cl
Br
Br
Br
Br
Br
Br
Br
3.7 (3.3–4.1)
2.2 (2.0–2.4)
3.9 (3.2–4.8)
3.9 (3.4–4.4)
3.3 (2.8–3.9)
16 (14–18)
8.7 (7.4–10)
7.6 (5.7–10)
6.0 (5.2–6.9)
1.6 (1.1–2.4)
4.5 (4.1–4.9)
4.5 (4.0–5.1)
11 (9.5–12)
1.9 (1.7–2.1)
1.5 (1.3–1.7)
11 (9.8–12)
17 (14–20)
>180
27% at 180
33% at 180
29% at 180
39% at 180
30% at 63
>180
>180
>180
36% at 180
44
28% at 180
>180
5.6 (4.5–7.1)
4.7 (3.9–5.8)
12 (10–14)
13 (11–15)
4.5 (2.1–9.4)
13 (11–15)
0.45 (0.32–0.62)
5.5 (4.6–6.6)
4.4 (4.2–4.7)
4.2 (3.6–4.8)
4.4 (3.7–5.3)
1.9 (1.7–2.1)
4.4 (3.5–5.5)
4.8 (4.1–5.6)
>100
7
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
OCH2CF3
OCH2CHCH2
OBu
i
OCH2 Pr
>180
Br
SO2Me
CF3
OMe
CO2H
CO2Me
Ph
Br
31% at 180
39% at 180
>180
21% at 180
>180
SO2Me
CF3
OMe
CO2H
CO2Me
Ph
14 (11–17)
41 (33–51)
>100
40% at 180
>180
>180
>180
42 (22–78)
1.5 (1.3–1.6)
Nꢀ(3,5ꢀCl2Ph)ꢀ Nꢀ(3,5ꢀCl2Ph)ꢀ
>180
>180
2.0 (1.2–3.7)
51
acetamide
acetamide
Me
tBu
SF5
Cl
Br
F
SMe
H
F
Me
tBu
H
H
H
H
H
H
F
141 (119–167)
>63
23% at 180
35% at 180
41% at 180
>180
>180
>63
>180
>180
>180
>180
19 (17–22)
4.1 (4.1–4.5)
14 (11–17)
42 (34–52)
22 (19–26)
>25
52
53
54
55
56
57
58
59
60
61
149 (122–181) 23% at 180
>25
>180
180 (160–201)
32 (13–82)
>180
>180
>180
>25
70 (61–80)
17 (15–20)
F
Cl
All compounds have pKa(N1)a of 7.3. a, b, c, d See footnotes for Table 1.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 16 of 80
1
2
3
4
5
6
7
8
9
When both chlorines were simultaneously replaced with either bromines (44) or
methylsulfonyls (45), inhibitor potency on the enzyme was improved by 2ꢀfold (although 45 had
no antitrypanosomal activity, presumably due to impaired cell uptake). Conversely, other
moderately sized nonpolar and polar (Me, tꢀBu, CF3, OMe, CO2H, CO2Me) or larger aryl (Ph
and Nꢀ(3,5ꢀ(diꢀCl)ꢀPh)ꢀacetamide) substituents in both meta positions led to reduced (46 and 47)
or nearly absent (48–53) enzyme inhibitory activity. Removal of one or both metaꢀsubstituents
(54ꢀ59) or replacement with a smaller fluorineꢀsubstituent (60, 61) substantially reduced (54ꢀ56,
58, 60, 61) or eliminated (57, 59) enzyme inhibitory activity.
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 correlation between enzyme and cell growth inhibition was dependent on the anilineꢀ
ring substitution pattern. Both enzyme and cellꢀbased activity correlated reasonably for analogs
32ꢀ44, 46ꢀ49, 52, 57ꢀ61, whereas enzymeꢀinactive compounds 50, 51, 53 retained substantial
antitrypanosomal activity indicating offꢀtarget activity. The very weak AdoMeDC inhibitors 54ꢀ
56 did retain some, albeit reduced cellular activity indicating a partial offꢀtarget activity or
increased cellular uptake/accumulation.
Having established the effects of metaꢀsubstitution, we looked at SAR related to substitution
patterns that include ortho and para positions (Table 5).
ACS Paragon Plus Environment

Page 17 of 80
Journal of Medicinal Chemistry
1
2
3
4
Table 5. SAR of ortho- and para-aniline ring substitutions
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
Compound R1 R2
R3
R4 Tb IC50,b µM Hs IC50,c µM Tb427 EC50,d µM
H
H
H
H
H
H
H
H
H
H
H
Cl
Cl
Br
Cl
Cl
Cl
Cl
Cl
Cl
H
F
F
Cl
Me
OPh
OH
OCHF2
NMe2
Cl 3.7 (3.3–4.1)
Cl 1.5 (1.4–1.8) 160 (135–191)
Br 1.6 (1.3–1.8)
27% at 180
5.6 (4.5–7.1)
4.8 (3.3–7)
4.6 (3.6–5.9)
4.5 (3.4–5.9)
4.4 (4.2–4.8)
1.1 (0.97–1.3)
13 (10–16)
4.5 (3.7–5.4)
5.8 (4.2–7.9)
18 (14–24)
2.1 (1.3–3.2)
26% at 100
10 (8.5–12)
23% at 25
27% at 25
28% at 25
>25
>25
>25
29% at 25
28% at 25
>25
25% at 25
28% at 25
28% at 25
19 (17–22)
11 (8.5–15)
15 (13–17)
1.0 (0.88–1.2)
25% at 25
7
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
43% at 180
>63
Cl
Cl
Cl
Cl
Cl
Cl
11 (8.8–13)
20 (17–24)
38% at 180
33% at 180
>63
>63
>180
>180
>63
>180
>180
>180
>180
>63
>180
>180
>180
>180
>180
>180
>180
>180
>180
>180
>180
>180
>180
>180
>180
>180
>180
>180
Cl Nꢀmorpholino Cl
Cl Nꢀpiperidino Cl
>180
>180
OH Cl
F
Cl
Cl Cl
Cl
Me H
SMe H
Br
Me Cl
Me H
Me H
F
Cl
H
H
H
H
H
H
H
F
H
H
Cl
H
H
H
H
H
Cl
Br
H
Cl
Et
iPr
CF3
Cl
37 (29–46)
Cl
H
Cl 2.2 (2.0–2.6)
Cl 115 (89–150)
H
H
H
H
H
H
Cl
H
H
137 (95–198)
>180
H
>180
>180
>180
>180
>180
>180
>180
H
H
H
H
H
H
H
H
F
CF3 199 (164–240)
H
H
H
H
H
H
>180
>180
>180
>180
>180
3,5ꢀCl2Ph
F
33% at 180
All compounds have pKa(N1)a of 7.3 except 67 (pKa(N1)= 7.0); 72 (pKa(N1)= 7.1); 73, 74,
and 83 (pKa(N1)= 7.2). a, b, c, d See footnotes for Table 1.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 18 of 80
1
2
3
4
5
6
7
8
9
Introduction of a fluorine into the para position led to up to 2ꢀfold enhancement in potency
towards TbAdoMetDC (62, 63 vs. 7, 44), whereas chlorine and methylꢀsubstitution (64, 65)
reduced activity 3 to 6ꢀfold vs. comparator 7. All other paraꢀsubstituents that were explored,
irrespective of size or electronic nature, led to a complete loss of inhibitory activity towards the
enzyme. Retrospectively, this SAR is in agreement with the coꢀcrystal structure data, which
indicates a sizeꢀrestriction in the AdoMetDC binding pocket along the paraꢀposition vector (vide
infra, Fig. 3B). The ortho position tolerated addition of a fluoro (73), but substitution with OH
(72) resulted in tenꢀfold reduced inhibitory activity vs. 7. Addition of a Cl, Br, Me, or SMe
substituent in the orthoꢀposition of the inactive unsubstituted phenyl analog 59 or very weakly
active mono metaꢀchloro or bromo analogs 55 or 56 was ineffective at introducing any activity
for compounds 74ꢀ84 (some additional inactive compounds are shown in Table S2).
In this series, there was a reasonable correlation between the activity on the enzyme and
cellular activity for compounds without paraꢀsubstituents or with paraꢀhalogens (62ꢀ64, 72ꢀ85,
90). However, this correlation breaks down for every other paraꢀsubstituted compounds tested
(65ꢀ71, 86ꢀ89). Notably, enzyme inactive paraꢀsubstituted analogs 66ꢀ71 and 86ꢀ89 retained
substantial antitrypanosomal activity, with EC50 values between 1 and 20 µM.
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 data so far suggested that metaꢀpositioning of substituents on the aniline ring was
crucial for activity and minimizing offꢀtarget antitrypanosomal activity. Minimally, one of the
two metaꢀpositions requires substitution with a chlorine or bromine, whereas the other metaꢀ
position needs to be substituted but has more flexibility such that halogens, cyano, CF3, Et,
alkylethers, and Ph can be accommodated. However, while these modifications were tolerated,
none resulted in a compound with substantially increased potency over the parent compound.
Since we discovered that phenyl in one of the two metaꢀpositions yielded an active compound
ACS Paragon Plus Environment

Page 19 of 80
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
(35, see Table 4), we prepared a number of substituted phenyl, heteroaromatic, and fusedꢀring
analogs, aiming to further tune the inhibitory activity and offꢀtarget antitrypanosomal activity
(Table 6).
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
Compared to the parent metaꢀphenyl substituted biaryl analog 35, 4ꢀFPh and 4ꢀClPh
substituted biaryl analogs 91 and 92 were about 2 to 3ꢀfold more potent in the enzyme and cellꢀ
based assays, whereas 4ꢀOMe, 4ꢀMe, and 4ꢀCF3 substitution delivered roughly equipotent
compounds in both assays (93–95). A 2ꢀfold loss of enzyme inhibition was observed for the 4ꢀ
CNPh analog (96), but this was accompanied by an increase in cellular activity (~5ꢀfold).
Moving to 4ꢀAc (97), 4ꢀOH (98), and 4ꢀNMe2 (100) substituents, enzyme inhibition dropped off
markedly but cellꢀbased inhibition remained at singleꢀdigit micromolar EC50 levels, indicating
offꢀtarget activity. Similarly, moving chlorine to the 2ꢀ or 3ꢀposition (101, 102), adding an extra
Cl or CF3 to the 3ꢀposition (103–105), or 3,5ꢀ(CF3)2ꢀsubstitution (106) also abolished enzyme
inhibition while retaining the offꢀtarget cellꢀbased activity.
Biaryls (4ꢀPhPh, 99), polycyclic aryls (naphthyl 107, phenanthrenyl 108), benzofused
heteroaryls (110, 112, 114), or 5ꢀmembered heteroaromatics (5ꢀthiazolyl 115, NꢀMeꢀ4ꢀpyrazolyl
116) were largely incompatible with enzyme inhibition but retained potent offꢀtarget cell
activity. Enzyme and cell activity tracked better for benzodioxolyl (109), furanyl (111),
thiophenyl (113), and pyrimidinyl (117) substituents, with the former two (109, 111) active in
the low micromolar range, and the latter two (113, 117) largely devoid of activity in both assays.
Finally, we prepared and tested compound 118 that combined 4ꢀfluorophenyl and ethyl metaꢀ
substitutions. Compared to the corresponding 3ꢀ(4ꢀFPh)ꢀ5ꢀCl analog 91, the similar activity of
118 indicates that metaꢀCl can be exchanged for metaꢀEt without affecting enzyme or cellꢀbased
potency.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 20 of 80
1
2
3
4
Table 6. SAR of meta-aryl aniline ring substitutions
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
118
Compound
R4
Tb IC50,b µM Hs IC50,c µM
Tb427 EC50,d µM
Ph
4ꢀFPh
4ꢀClPh
4ꢀMeOPh
4ꢀMePh
4ꢀCF3Ph
4ꢀCNPh
4ꢀAcPh
4ꢀHOPh
4ꢀPhPh
3.3 (2.8–3.9)
1.1 (0.95–1.2)
1.7 (1.6–2.0)
2.3 (2.1–2.6)
4.6 (4.0–5.3)
4.7 (3.9–5.7)
7.6 (3.2–18)
28% at 21
30% at 63
>21
4.5 (2.1–9.4)
1.6 (0.94–2.8)
2.2 (1.8–2.7)
4.5 (3.1–6.6)
4.3 (3.3–5.7)
2.0 (1.4–3)
1.0 (0.96–1.1)
6.7 (4.1–11)
9.7 (7.5–13)
0.91 (0.82–1)
5.0 (4.2–6)
8.5 (7.6–9.5)
1.9 (1.6–2.3)
2.9 (1.3–3.6)
3.2 (2.8–3.7)
1.5 (1–2.2)
35
91
92
93
94
95
96
97
98
>63
>180
>63
>21
>7
>21
143 (115–178)
167 (150–190)
34% at 180
26 (22–33)
81 (57–114)
102 (82–125)
157 (95–260)
53 (46–61)
>180
>180
>180
>180
>180
>180
>180
>180
>180
>180
>180
>21
99
4ꢀNMe2Ph
2ꢀClPh
3ꢀClPh
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
3,4ꢀ(Cl)2Ph
3ꢀClꢀ4ꢀFPh
3ꢀCF3ꢀ4ꢀClPh
3,5ꢀ(CF3)2Ph
2ꢀnaphthyl
9ꢀphenanthrenyl
28 (25–31)
57 (51–64)
0.53 (0.47–0.59)
1.7 (1.4–2)
>180
1.8 (1.5–2.1)
2.3 (1.6–3.2)
0.57 (0.13–2.5)
7.1 (5.7–8.8)
1.4 (1.4–1.5)
29 (23–35)
1.6 (1.5–1.8)
3.6 (3.1–4.1)
6.7 (4.7–9.6)
58 (33–100)
1.3 (1.1–1.5)
5ꢀbenzoꢀ1,3ꢀdioxolyl 6.2 (4.6–8.3)
5ꢀbenzofurazanyl
3ꢀfuranyl
2ꢀbenzofuranyl
3ꢀthiopheneyl
2ꢀbenzothiopheneyl
5ꢀthiazolyl
27 (21–35)
7.4 (6.2–8.7)
23% at 63
>180
30% at 180
>63
26 (22–32)
41 (36–46)
42% at 21
30% at 180
>180
>21
NꢀMeꢀ4ꢀpyrazolyl 164 (132–203)
>180
>180
>180
5ꢀpyrimidinyl
>180
2.7 (2.2–3.4)
All compounds have pKa(N1)a of 7.3. a, b, c, d See footnotes for Table 1.
ACS Paragon Plus Environment

Page 21 of 80
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Crystal structure of TbAdoMetDC/prozyme in complex with a pyrimidineamine analog
reveals binding interactions. To assist in rational design of pyrimidineamine analogs with
improved binding properties, we solved the crystal structure of the TbAdoMetDC/prozyme
complex coꢀcrystallized with 44. Compound 44 was selected for coꢀcrystallization on the basis of
its potency and solubility in aqueous solutions. Additionally, the presence of two bromines
facilitated location of the bound compound in the complex, and the anomalous diffraction data
collected at the Br Kꢀedge were instrumental in determining the correct orientation of this quasiꢀ
symmetrical molecule in the binding pocket (See Experimental methods). The structure was
solved by molecular replacement using our previously determined TbAdoMetDC/prozyme
heterodimeric structure (Protein Data Bank (PDB) identifier 5TVM)21 as a search model, and
refined to 3.0 Å resolution with Rwork/Rfree = 20.0/25.3%. Overall, the structure was similar to the
unliganded active heterodimer (RMSD of 0.28 Å between current structure and 5TVM as
determined by TMꢀalign34 for 636 equivalent Cα atoms).
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
A single molecule of 44 was found in the catalytic site of the TbAdoMetDC in the
heterodimer of TbAdoMetDC/prozyme, positioned 3.9 Å from the catalytic pyruvate (Figure
3A). This binding mode is consistent with biochemically demonstrated competitive mode of
inhibition (see Figure 2B). The binding interactions observed between 44 and the enzyme also
explain much of the above described SAR for the series. The requirement for N1 to be
protonated and for C2 to have a primary amine is explained by the finding that both groups are
within hydrogenꢀbonding distance of the sideꢀchain carboxyl oxygens of Glu266 in the binding
pocket. It is thus reasonable to assume that a protonated pyrimidineamine participates in
favorable bidentate Hꢀbonding and saltꢀbridge interactions (Figure 3B).
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 22 of 80
1
2
3
4
5
6
7
8
9
Additional binding energy is also likely contributed by π–π stacking interactions between
the pyrimidine ring and Phe28, and between aniline ring and Tyr243. The proximity of the
catalytic Cys100 to the aniline ring of 44 (3.2 Å between SH and C4' of the ring) is suggestive of
the aromatic–thiol interaction (Figure 3B), though the angle of the interaction is not ideal as
determined by previously described interactions of this type.35, 36 Finally, an Hꢀbonding
interaction37 between the partially negative lateral end of one of the aniline ring metaꢀbromines
and Ser249 hydroxyl group is also potentially present, and may contribute to the binding energy
(Figure 3B).
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
Importantly, the structure of TbAdoMetDC coꢀcrystallized with 44 revealed that the 3'ꢀ and
5'ꢀsubstituents on the aniline ring projected into two distinct pockets designated a and b (Figure
3B). Building on the structural data and the observations reported in Tables 4–6, we hypothesize
that of the two pockets, only pocket a could accommodate a phenylꢀsized substitution at the
metaꢀposition of the aniline ring. Conversely, the shape and depth of pocket b suggested that it
would be accessible to linear, flexible substituents (for example, alkyls and alkyl ethers 38ꢀ43 in
Table 4). The fact that 118, functionalized with pꢀfluorophenyl at one metaꢀposition and ethyl at
the other, inhibited the enzyme with good potency (IC50 = 2.7 µM, see Table 6) suggests that
both pockets can be accessed simultaneously in the context of a single compound. Additionally,
our structure suggested that paraꢀsubstituents larger than F, Cl, or Me would be inaccessible as
confirmed by the enzyme inhibition data in Table 5 (66ꢀ71).
Structural basis for parasite–host selectivity. HsAdoMetDC (PDB identifier 3DZ7)38 was
aligned to the TbAdoMetDC bound to 44 to provide insight into the structural basis for
selectivity of this and other pyrimidineamines for T. brucei AdoMetDC compared with the
human enzyme (Figure 4). The catalytic sites are mostly conserved between the two species, and
ACS Paragon Plus Environment

Page 23 of 80
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
their geometries are similar. One residue at the lid of the pocket, however, differs between them.
This residue (His5 in human AdoMetDC and Arg26 in T. brucei AdoMetDC) is positioned
differently in the two structures. His5 in human AdoMetDC is within 3.2 Å of 44 in the overlaid
enzyme, and may function to restrict the ligandꢀbinding site compared with the T. brucei
structure (Figure 4). The analogous position in T. brucei AdoMetDC (Arg26) was not well
resolved in the electron density maps of the structure reported herein; however, it was observed
in previous structures (PDB depositions 5TVF and 5TVM) where it was positioned away from
the opening of the binding pocket (Figure 4).
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
Additionally, the substitution of Cys49 in the human structure for Ala67 in T. brucei
AdoMetDC restricts the size of the human pocket a relative to T. brucei (Figure 4). This size
restriction would be expected to provide additional selectivity for analogs that have larger
substituents in pocket a, such as the pꢀfluorophenyl of 91 and 118.
On-target mechanism-of-action studies. As was noted above, the correlation between the
enzyme and cellꢀgrowth inhibition was weak. This suggested that trypanocidal effect was at least
in part due to offꢀtarget mechanism of action. Thus, we sought to identify whether the
pyrimidineamines were inhibiting AdoMetDC in T. brucei cells. We previously reported that the
levels of prozyme protein markedly increased when AdoMetDC activity in T. brucei cells was
reduced by either genetic knockdown of AdoMetDC or by chemical inhibition.16, 39 We exploited
this feedback loop to evaluate whether hit compounds identified in the HTS were affecting the
parasite cell growth through intracellular AdoMetDC inhibition.22 Here, we compared prozyme
protein levels in cells treated for 24 h with representative pyrimidineamine analogs that 1)
inhibited both T. brucei cell growth and AdoMetDC (7 and 8), 2) inhibited only T. brucei growth
(23 and 13), or 3) were inactive in both assays (29). Only compounds that were active on both
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 24 of 80
1
2
3
4
5
6
7
8
9
the enzyme and cells induced prozyme above the vehicleꢀonly control levels (Figure 5). These
results show that the pyrimidineamines with TbAdoMetDC inhibitory activity on the purified
enzyme also block AdoMetDC activity in T. brucei cells. However, these results also confirm
that there was a substantial offꢀtarget component to pyrimidineamineꢀinduced cell growth
inhibition, as 23 and 13, which did not inhibit TbAdoMetDC — either in the purifiedꢀenzyme
assay or in the cell (as evident from vehicleꢀonly levels of the prozyme protein) — inhibited cell
growth with lowꢀmicromolar potencies (Figure 5).
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
Discussion and Conclusions
The number of cases of HAT has decreased 10ꢀfold since the beginning of the 21st century
owing to proactive surveillance and control measures set in place in endemic countries.3 Yet the
disease has a history of resurgence following years of a downward trend.3 Sustainability of HAT
elimination relies in part on the availability of patientꢀfriendly treatment options.3 Neither
melarsoprol nor eflornithine, alone or as a part of NECT, are ideal due to high toxicity of the
former and high price and requirement for multiple intravenous infusions of the latter.2, 10 The
potential for eflornithine resistance to emerge is another negative factor.40 Thus, new safe antiꢀ
trypanosomal drugs with CNSꢀstage efficacy are needed.
We identified and characterized a pyrimidineamine series as a new class of reversible,
competitive inhibitors of T. brucei AdoMetDC. This series has a number of positive attributes,
including speciesꢀselective inhibition of the T. brucei AdoMetDC compared with the human
enzyme and antitrypanosomal activity against T. brucei in wholeꢀcell assays. Importantly, the
series is predicted to have good brain penetration on the basis of the monolayer permeability
assay performed on the parent compound. Thus, the pyrimidineamines have the potential to be
used for treatment of both earlyꢀstage and lateꢀstage HAT. Herein, we have established the
ACS Paragon Plus Environment

Page 25 of 80
Journal of Medicinal Chemistry
1
2
3
4
5
6
preliminary SAR within the series and have solved the crystal structure of one analog bound to
T. brucei AdoMetDC, providing insight for future lead optimization of the series.
7
8
9
While the pyrimidineamines look to be a promising series for further optimization, three key
issues were identified that would need to be addressed in future chemistry programs. Firstly,
while the pyrimidineamines did show activity against both the enzyme and the parasite, the
correlation between these activities was poor. Onꢀtarget activity was present as evident from the
finding that prozyme was upregulated in cells treated with compounds that inhibited AdoMetDC.
In addition, good correlation between enzyme inhibition and antiꢀparasite activity was observed
for most active analogs with two metaꢀsubstituents in the aniline ring, with at least one
substituent being Cl or Br. If the second metaꢀposition is a phenylꢀring substituted in the 4ꢀ
position with F, Cl, OMe, Me, or CF3, good activity in both assays was observed. However, there
was also clear evidence for an offꢀtarget activity, as a substantial fraction of pyrimidineamines
that had lost TbAdoMetDC inhibitory activity elicited cell growth inhibition in the lowꢀ
micromolar range. Offꢀtarget activity was especially noticeable for paraꢀsubstituted aniline rings
(Table 5) and bisꢀmetaꢀsubstituted aniline rings with on metaꢀposition occupied by 2ꢀ, 3ꢀ, 2,3ꢀ or
other specifically substituted aryls, biaryls, polycyclic aryls and heteroaryls (Table 6). Secondly,
despite synthesis of a number of analogs of the original parent compound, we were unable to
substantially improve compound potency below the lowꢀmicromolar level. This issue may also
relate to the offꢀtarget activity, and thus identification of more potent analogs may lead to
reduced offꢀtarget activity as the onꢀtarget component improves. As the analogs described herein
had been synthesized prior to obtaining the crystal structure with an inhibitor bound, there
remains room for a structureꢀbased lead optimization program to facilitate the identification of
more potent, onꢀtarget compounds. Lastly, 7 showed poor metabolic stability; however, this
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
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 26 of 80
1
2
3
4
5
6
liability was not the focus of the current chemistry program and remains to be addressed during
further optimization of the series.
7
8
9
A notable feature of pyrimidineamines is the finding that inhibitor binding to TbAdoMetDC
is pHꢀdependent in the relevant physiological range (pH 6.8–7.7). This dependence has its basis
in the pKa of N1 in the pyrimidine ring, which is predicted to be 7.3 in the parent compound 7.
Within the series, modifying the ring substituents to lower the pKa(N1) (<6) led to reduced
inhibitory activity at physiological pH, while optimizing the ring electronics to yield a higher
pKa(N1) (>9) reduced the pH dependence of binding. The crystal structure of TbAdoMetDC
bound to 44 revealed that N1 forms a salt bridge with Glu266, explaining the requirement for
protonation of N1 for inhibitor binding. While we were able to identify compounds with higher
pKa, this was achieved through removal of N3 (20 and 21), which led to a loss of the possible
interaction between N3 and mainꢀchain carbonyl of Cys246 (3.5 Å away, see Figure 3B). Thus
we did not accomplish the goal of improved binding affinity by this strategy. One potential
direction for future optimization of this series will be to identify compounds that retain full
charge at N1 at physiological pH, without compromising other binding interactions.
In addition to the protonation state of N1, we identified a number of other features of the
pharmacophore that were required for TbAdoMetDC inhibition. A primary amine is required at
C2, and it is involved in an Hꢀbond with the Glu266 carboxyl group. The amine linker could not
be modified as there was a loss of potency when the linker was methylated (28), extended by one
or two methylene links (30 and 31), or replaced with sulfur (29). The amine linker does not make
any clear interactions with the protein, so its importance may relate to geometry and steric
considerations. For sulfurꢀlinked analog 29, the decreased pKa at N1 (4.3) could explain the loss
of enzyme inhibition. Finally, both pyrimidine and aniline rings are required, and these form
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
ACS Paragon Plus Environment

Page 27 of 80
Journal of Medicinal Chemistry
1
2
3
4
5
6
stabilizing π–π stacking interactions with the two aromatic residues, Phe28 and Tyr243, on
opposite sides of the catalytic site.
7
8
9
The aniline ring is the most buried part of the ligand. SAR and the crystal structure indicate
that substituents in both of the metaꢀpositions are required for good binding. We identified both
steric and electronic requirements to the metaꢀsubstitutions. At least one of the metaꢀpositions
can include an alkyl, alkyloxy, halide, or 4ꢀsubstituted phenyl ring; whereas the second position
is best occupied by Cl, or Br. Notably, both an alkyl (ethyl) and a 4ꢀsubstituted phenyl could be
accommodated at the same time as indicated by antiꢀTbAdoMetDC activity of 118. A hydrogen
bond between Ser249 γꢀhydroxyl and a partially negative acceptor in a meta position, such as
halide or alkoxy may contribute to the binding energy and explain the finding.37 The next steps,
informed by the coꢀcrystal structure data, will involve a much more in depth combinatorial
exploration of both metaꢀpositions in such a manner as to ideally occupy both a and b pockets as
well as maximize π–π stacking interactions.
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
All active analogs were strongly selective for T. brucei AdoMetDC over the human enzyme.
The selectivity ratio (HsIC50/TbIC50) was at least 27 (40) and in many cases exceeded 100 (e.g.,
44 and 62). Yet, the structural basis for species selectivity was not immediately obvious given
very similar geometries and residue composition of the catalytic sites in the two species.
Specifically, the human site has Phe7 and Phe223 in the positions analogous to T. brucei Phe28
and Tyr243, which could provide π–π stacking interactions with a compound (see Figure 4).
Also, Glu247 in the human enzyme closely aligns with T. brucei Glu266, and could potentially
also form Hꢀbonding and saltꢀbridge interactions with pyrimidineamine part of an inhibitor. One
possible hypothesis suggested by our coꢀcrystal structure is that His5 in the human AdoMetDC
sterically limits access to the binding site, excluding certain classes of inhibitors and thus may be
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 28 of 80
1
2
3
4
5
6
7
8
9
responsible for the observed species selectivity of the pyrimidineamines. In T. brucei, the
equivalent position is occupied by an Arg residue with a more labile side chain (see Figure 4).
Notably, all 13 classes of validated active hits from our HTS campaign were also selective
inhibitors of the T. brucei enzyme,22 suggesting this may be a general principle of parasite–host
selectivity. In contrast, a number of compounds that have greater binding affinity for the human
enzyme have also been described, such as methylglyoxal bis(guanylhydrazone), or MGBG,41 and
CGP 40215,42 so evidently this mechanism does not limit general access to the active site. An
additional possible origin for species selectivity can be inferred from the allosteric mechanism of
T. brucei AdoMetDC activation: its Nꢀterminal residues form an autoinhibitory lid blocking the
active site in the monomer that is displaced from the active site upon binding to prozyme.21
Conversely, the Nꢀterminus of the human enzyme is 21 residues shorter than that of the T. brucei
enzyme, and it has not been observed to bind in the active site in any of the published Xꢀray
crystal structures.38, 43ꢀ45 However, these data do not rule out the possibility that these residues
undergo dynamic transitions during catalysis or ligand binding. Thus, potentially selectivity
could arise if the Nꢀterminal residues of the human enzyme compete for ligand binding. In this
scenario, T. brucei selective inhibitors would be unable to effectively compete with the Nꢀ
terminal residue lid, whereas human selective inhibitors and substrate would be effective
competitors. Finally, we note that pocket a in human AdoMetDC is smaller than in the T. brucei
enzyme, and this difference will enhance selectivity for compounds that fully occupy this pocket
(e.g., 91), though this difference is unlikely to explain the selectivity of compounds with smaller
substitutions in pocket a (e.g. 44).
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 work described herein has led to the discovery of pyrimidineamines with lowꢀ
micromolar activity against both T. brucei AdoMetDC and the parasite. Through a combination
ACS Paragon Plus Environment

Page 29 of 80
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
of medicinal chemistry and protein crystallography, we have defined the key components of the
pharmacophore that are required for TbAdoMetDC inhibitory activity. If this series is to advance
to lead optimization, future work will need to focus on identifying more potent compounds. Two
pockets were discovered in the crystal structure that were not fully exploited in the current work.
Pocket a is larger of the two and is lined with both polar (Glu85 and Pvl86) and nonpolar (Ile65,
Ala67, Leu83, and Ile103) side chains (see Figure 3B). The pꢀsubstituted phenyl analogs such as
91 provided for good size complementarity with the available pocket but the crystal structure and
activity data suggest that more polar substituents that could engage in Hꢀbonding interactions
with Glu85 and backbone carbonyls in the pocket might provide better binding energy (see Table
6 and Figure 3B). Alternatively, a nucleophilic group (e.g, aminooxy) might allow formation of a
Schiff base with the prosthetic pyruvoyl group, transforming pyrimidineamines into mechanismꢀ
based inhibitors, as described for substrate analogs.46 Pocket b also has substantial polar
character with several opportunities for Hꢀbonding interactions presented by side chains of
Glu32, Ser273, Ser249, and His262 (see Figure 3B). Current ring substituents did not fully
exploit these opportunities and thus further series development should focus on optimizing these
potential polar interactions at the metaꢀpositions of the aniline ring. The search of the suitable
substituents can be aided by computational protein–ligand docking through virtual screening of
metaꢀposition analogs. Iterative coꢀcrystallization with new analogs will aid development.
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
ACS Paragon Plus Environment