Integrated Synthetic, Biophysical, and Computational Investigations of Covalent Inhibitors of Prolyl Oligopeptidase and Fibroblast Activation Protein α

Article abstract of DOI:10.1021/acs.jmedchem.9b00642

Over the past decade, there has been increasing interest in covalent inhibition as a drug design strategy. Our own interest in the development of prolyl oligopeptidase (POP) and fibroblast activation protein α (FAP) covalent inhibitors has led us to question whether these two serine proteases were equal in terms of their reactivity toward electrophilic warheads. To streamline such investigations, we exploited both computational and experimental methods to investigate the influence of different reactive groups on both potency and binding kinetics using both our own series of POP inhibitors and others' discovered hits. A direct correlation between inhibitor reactivity and residence time was demonstrated through quantum mechanics methods and further supported by experimental studies. This computational method was also successfully applied to FAP, as an overview of known FAP inhibitors confirmed our computational predictions that more reactive warheads (e.g., boronic acids) must be employed to inhibit FAP than for POP.

Full text of DOI:10.1021/acs.jmedchem.9b00642

Subscriber access provided by Nottingham Trent University
Article
Integrated Synthetic, Biophysical and Computational
Investigations of Covalent Inhibitors of Prolyl
Oligopeptidase and Fibroblast Activation Protein #
Jessica Plescia, Stephane De Cesco, Mihai Burai Patrascu, Jerry Kurian, Justin M. Di Trani, Caroline
Dufresne, Alexander Wahba, Naela Janmamode, Anthony K. Mittermaier, and Nicolas Moitessier
J. Med. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.jmedchem.9b00642 • Publication Date (Web): 08 Aug 2019
Downloaded from pubs.acs.org on August 9, 2019
Just Accepted
“Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted
online prior to technical editing, formatting for publication and author proofing. The American Chemical
Society provides “Just Accepted” as a service to the research community to expedite the dissemination
of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in
full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully
peer reviewed, but should not be considered the official version of record. They are citable by the
Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore,
the “Just Accepted” Web site may not include all articles that will be published in the journal. After
a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web
site and published as an ASAP article. Note that technical editing may introduce minor changes
to the manuscript text and/or graphics which could affect content, and all legal disclaimers and
ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or
consequences arising from the use of information contained in these “Just Accepted” manuscripts.
is published by the American Chemical Society. 1155 Sixteenth Street N.W.,
Washington, DC 20036
Published by American Chemical Society. Copyright © American Chemical Society.
However, no copyright claim is made to original U.S. Government works, or works
produced by employees of any Commonwealth realm Crown government in the course
of their duties.

Page 1 of 39
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Integrated Synthetic, Biophysical and Computational
Investigations of Covalent Inhibitors of Prolyl
Oligopeptidase and Fibroblast Activation Protein α
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
Jessica Plescia,‡,1 Stéphane De Cesco,‡,†,1 Mihai Burai Patrascu,‡,1 Jerry Kurian,1 Justin Di
Trani,1 Caroline Dufresne,1 Alexander S. Wahba,1 Naëla Janmamode,1 Anthony K.
Mittermaier,1 and Nicolas Moitessier*
1Department of Chemistry, McGill University, 801 Sherbrooke Street West, Montreal, QC,
Canada H3A 0B8
KEYWORDS covalent druggability • serine proteases • quantum cluster chemical approach •
covalent inhibitors
ABSTRACT. Over the past decade, there has been an increasing interest in covalent inhibition
as a drug design strategy. Our own interest in the development of prolyl oligopeptidase
(POP) and fibroblast activation protein α (FAP) covalent inhibitors has led us to question
whether these two serine proteases were equal in terms of their reactivity towards
electrophilic warheads. To streamline such investigations, we exploited both computational
and experimental methods to investigate the influence of different reactive groups on both
potency and binding kinetics, using both our own series of POP inhibitors and others’
ACS Paragon Plus Environment
1

Journal of Medicinal Chemistry
Page 2 of 39
1
2
3
4
5
6
7
8
9
discovered hits. A direct correlation between inhibitor reactivity and residence time was
demonstrated through quantum mechanics (QM) methods and further supported by
experimental studies. This computational method was also successfully applied to FAP, as an
overview of known FAP inhibitors confirmed our computational predictions that more
reactive warheads (e.g., boronic acids) must be employed to inhibit FAP than for POP.
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
Following the resurgence of covalent inhibition in the last decade,1-3 kinases and serine
proteases have been targeted with numerous covalent inhibitors,4 and covalent inhibitors
have reached the market (Figure 1). Among these targets are prolyl oligopeptidase (POP),
initially associated with neurodegenerative diseases,5-8 and fibroblast activation protein α
(FAP), a promising target for anti-cancer therapies.9-12 In the past, our group and others have
reported a number of potent covalent POP inhibitors,8 including Cbz-Pro-Prolinal (1), JTP-
4819 (2), KYP-2047 (3), and bicyclic derivative 5,13 as well as FAP covalent inhibitors such
as compounds 6, 8, and Talabostat (Figure 2). Although these two druggable targets have
been inhibited by many covalent inhibitors, the major differences lie in the chemical nature
of the warheads, or electrophilic functional groups that form covalent bonds with protein
residues.
O
OH
B
N
H
N
N
N
N
H
N
OH
HO
O
H
O
N
Vildagliptin
Bortezomib
Figure 1. Selected marketed covalent inhibitors.
ACS Paragon Plus Environment
2

Page 3 of 39
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Non-covalent inhibitors often bind and dissociate very quickly and exhibit short residence
times and are therefore often largely under thermodynamic control. In contrast, covalent
inhibitors are believed to often bind in a two-step process (Figure 3a): a fast, non-covalent
initial binding followed by a slower covalent bond formation. As a result, kinetic factors and
residence time cannot be ignored and could indeed be critical for inhibitor efficacy.3 Recently
the reactivity of the warheads used in covalent inhibition has been increasingly investigated
either experimentally14,15 or computationally16 as reactivity often dictates whether the
inhibitor will bind reversibly or irreversibly.17 Similarly, the reactivity of the protein’s
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
catalytic residues in covalent inhibition has been investigated with focus on cysteines.18-22
A
very interesting method based on free energy perturbation (FEP) has recently been reported
which computes binding free energies including both covalent and non-covalent
contribution to binding. 23,24
However, although these methods proved effective, the focus has been primarily on
cysteine residues and acrylamides as warheads, as kinase cysteines have likely been the
major targets for covalent inhibitors (often acrylamides) in the recent years25. A number of
research groups are nowadays searching for alternative warheads and residues to bind to
(marketed covalent drugs are binding to serine) and computational methods must be
assessed.26
Prior to designing covalent inhibitors, the biological target must be first identified as
covalently druggable (i.e., can be targeted with covalent inhibitors). Unfortunately, there are
very few tools currently available, experimental or computational, to accomplish this. We
report herein a developed computational protocol using POP and FAP that could eventually
be used to (1) predict whether an enzyme is covalently druggable and (2) to identify
ACS Paragon Plus Environment
3

Journal of Medicinal Chemistry
Page 4 of 39
1
2
3
4
5
6
7
8
9
potentially potent warheads. To illustrate the potential application of such a protocol: it took
our team months to optimally express and purify POP and to optimize the in vitro activity
assays, which was followed by months of synthesis until we found potent inhibitors. With
our current computational protocol in hand, requiring only 2-3 weeks of calculations, we
would have been able to make a more informed decision on whether to initiate our hit-
discovery endeavor. Similarly, until this protocol was available, our efforts focused on the
unsuccessful development of nitrile-containing FAP inhibitors. Running reactivity
predictions before synthesis would have allowed us to opt for the appropriate warheads
much earlier.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
N
N
N
O
N
NH
O
O
O
O
O
H
OH
O
1, Cbz-Pro-Prolinal
2, JTP-4819
POP: Ki = 0.35 nM
POP: Ki = 0.06 nM
N
N
N
R
O
O
O
N
O
N
3, KYP-2047
4
, R=H, Ki = 160 nM (hPOP)
POP: Ki = 0.02 nM
5
, R=CN, Ki = 25 nM (hPOP)
FAP: IC50 = > 100 nM
R2
R1
O
NH
N
N
N
O
O
B
NH2
O
HO
OH
F
N
HN
O
N
B
HO
6,
R1= Me, R2=H,
OH
O
POP: IC50 >10,000 nM
FAP: IC50 > 36 nM
9, Talabostat
N
FAP: IC50 = 66 nM
POP: IC50 = 980 nM
DPP-IV: IC50 = 22 nM
7
, R1= H, R2=iPr,
8
POP: IC50 = 1.3 nM
FAP: IC50 > 100,000 nM
FAP: IC50 = 20 nM
Figure 2. Selected POP and FAP inhibitors.
ACS Paragon Plus Environment
4

Page 5 of 39
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
As demonstrated via co-crystallization,27,28 POP inhibition can be achieved through
covalent bond formation between the reactive group of a ligand and the catalytic serine in
the active site (Ser554). The reaction of Ser554 with aldehyde 1 leads to the formation of a
hemiacetal, which favorably mimics the tetrahedral intermediate of the endogenous catalytic
reaction, stabilizing its presence in the active site (Figures 3b, 3c). In contrast, reaction with
a nitrile group leads to a trigonal planar iminoether, an intermediate which less favorably
mimics the amide group of the peptide substrates (Figure 3d). While maintaining favorable
non-covalent interactions (e.g. via the scaffold) is essential for both potency and selectivity,
modification of the covalent warhead is also expected to have a significant impact on the
binding affinity and kinetics via its influence on the second step of the binding event (Figure
3a).
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
k1
k2
...

I
a)
b)
E
I
E
E + I
k-1
k-2
POP
R
O
O
POP-OH
O
O
R
R
H2N
OH
R
N
H
R
R
N
H
c)
POP
O
POP-OH
POP-OH
O
O
R
R
H
N
R
H
d)
POP
O
R
NH
Figure 3. a) Two-step process for covalent inhibition. E: enzyme; I: inhibitor; E∙∙∙I: non-
covalent complex; k1: association rate constant; k-1: dissociation rate constant. b) cleavage of
a substrate; b) aldehyde inhibitor covalent binding; c) nitrile inhibitor covalent binding.
ACS Paragon Plus Environment
5

Journal of Medicinal Chemistry
Page 6 of 39
1
2
3
4
5
6
7
8
9
Herein, we describe our collaborative approach, including computational predictions and
experimental evaluations, to the investigation of the relative reactivities of FAP and POP and
the nature of the covalent warheads that are more likely to lead to potent inhibitors.
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
Strategy. We sought to develop a computational protocol which would first be tested
against experimental data collected on POP and then validated on a homologous enzyme,
FAP (Figure 4).
Figure 4. Computational model and experimental data collection. The residence times of
various inhibitors depends on kinetics factors (tR = 1/koff), which are measured using
biophysical methods. These kinetic parameters are related to the energy required for the
inhibitors to break the covalent bond and leave the enzyme (Eoff). Advantageously, Eoff can
be computed, ultimately demonstrating that computations can substitute complex, time-
ACS Paragon Plus Environment
6

Page 7 of 39
Journal of Medicinal Chemistry
1
2
3
4
5
6
consuming, and expensive experiments for initial assessments as to whether a newly
discovered target is covalently druggable.
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
In order to probe the impact of the intrinsic reactivity of the warhead on the overall
binding process, we designed a series of inhibitors 5a, 10a-17a which complement
previously reported inhibitors 5c, 11c, 13c-15c29 (Figure 5). Two strategies were envisioned:
(1) substitution of the warhead – the nitrile, aldehyde, and boronic acid were selected, as
these are known to form covalent bonds with nucleophilic protein residues;30,31 and (2)
modification of the electronic environment of a given warhead – electron withdrawing
fluorine atoms could be introduced on the nitrile analogue, a strategy exploited to prepare
FAP inhibitors (Figure 2).
ACS Paragon Plus Environment
7

Journal of Medicinal Chemistry
Page 8 of 39
1
2
3
4
5
6
7
8
9
Figure 5. Selected inhibitor structures. i) different reactive groups; ii) modulated nitrile
reactivity; iii) effects of fluorine atoms alone. b) Docked binding mode of boronic acid 11a
into the POP active site. c) Snapshot of the POP active site and with 10b bound to Ser554 and
the residues kept for the Quantum Chemical Cluster Approach (QCCA) study.
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
Chemistry. The synthesis of bicyclic scaffold a, previously published by our group, has been
optimized and is presented in Scheme S1 in the Supporting Information.13 Bicyclic precursor
18 (Scheme 1), previously discovered through a virtual screening/virtual optimization
strategy, was selected because of its straightforward and efficient synthesis. More
specifically, this core was readily available in only three synthetic steps and an overall yield
of 74% with no flash chromatography purification, offering the unprotected carboxylic acid
18 as a diversity point.
The individual proline analogs were then coupled to the scaffold 18 to afford the desired
inhibitors. Because the proline analogues were either commercially or readily available,
expedient and efficient synthesis of potential POP inhibitors was achieved in only 1-2 steps.
The synthesis of these selected analogues (4a, 5a and 10a-17a) is outlined in Schemes 1 and
2.
The boron-containing analogue 12a was obtained by coupling scaffold 18 with the
commercially available proline analog 19 to yield product 12a in moderate yield (Scheme 1).
Attempts at hydrolysis of boronic ester 12a to obtain boronic acid 11a were unsuccessful,
exhibiting solubility issues and yielding complex mixtures, so 12a was utilized instead. Our
own liquid chromatography-mass spectrometry experiment revealed that the boronic ester
12a is nearly quantitatively hydrolyzed to boronic acid 11a in the assay buffer (ca. 90% after
ACS Paragon Plus Environment
8

Page 9 of 39
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
10 minutes, Figure 6), and can therefore be tested as a pro-drug cleaved in the assay.
Furthermore, our own in vitro assay controls show that the cleaved pinanediol exhibits no
inhibitory activity against POP (data not shown).
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
Stability of 12a in buffer
100
90
80
70
60
50
40
30
20
10
0
0
10
20
30
40
50
60
time (min)
Figure 6. Liquid chromatography–mass spectroscopy study of boronic ester cleavage in the
POP buffer. The relative abundance of the boronic ester was recorded over one hour,
plateauing at approximately 5% abundance.
Synthesis of the potential inhibitor 10a bearing an aldehyde as the warhead started with
the coupling of commercially available L-prolinol with scaffold 18, to afford the primary
alcohol 20 in excellent yield. Further oxidation under Swern conditions led to the desired
aldehyde 10a. Synthesis of the non-covalent analog 4a was accomplished by coupling
scaffold 18 to pyrrolidine. The nitrile analog 5a was obtained by coupling 18 with readily
available (S)-pyrrolidine-2-carbonitrile (Scheme 1).
ACS Paragon Plus Environment
9

Journal of Medicinal Chemistry
Page 10 of 39
1
2
3
4
Scheme 1. Synthesis of diversely functionalized inhibitors.a
5
6
7
8
9
O
O
B
N
H2
Cl
H
19
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)
a)
OH
N
N
O
N
O
B
N
O
N
O
O
O
O
O
H
18
b)
12a
4a
a)
N
N
O
N
N
c)
O
5a
N
O
N
OH
O
N
O
O
O
20
10a
aReagents: a) i. PivCl, Et3N, DCM, 0°C; ii. amine (see Experimental Section), rt, 18h (4a, 40%,
5a, 92%, 12a, 43%); b) i. PivCl, Et3N, DCM, 0 °C; ii. Prolinol, 18h, rt (87%); c) DMSO,
Oxalylchloride, DCM, −78°C (40%).
Synthesis of the selected fluorocyanopyrrolidine analogues began from readily available
starting materials.32 Coupling of 18 with nitrile 21 led to the corresponding inhibitor 13a.
Reaction of readily available free amines 22 and 23 with the bicycle core 18 under standard
coupling conditions afforded the intended inhibitors 14a and 15a, respectively. The non-
covalent inhibitors 16a and 17a were prepared through coupling of carboxylic acid 18 with
24 and 25, respectively.
ACS Paragon Plus Environment
10

Page 11 of 39
Journal of Medicinal Chemistry
1
2
3
4
Scheme 2.a Synthesis of the fluorine-containing compounds.
5
6
7
X1
F
X2
F
8
9
N
N
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
CN
CN
O
N
O
N
O
O
14a
15a
, X1=H, X2=F
, X1=F, X2=H
13a
X1
F
X2
F
a)
a)
HN
HN
CN
CN
22
23
, X1=F, X2=H
, X1=H, X2=F
21
OH
O
N
O
18
X1
X2
HN
a)
24
25
, X1=H, X2=F
, X1=F, X2=H
X1
X2
N
O
N
O
16a
17a
, X1=H, X2=F
, X1=F, X2=H
aReagents: a) i. PivCl, Et3N, DCM, 0°C; ii. amine (see Experimental Section), rt, 18h (13a,
35%; 14a, 30%; 15a, 30% 16a, 90%; 17a, 90%)
Biophysical characterization. Next, these selected molecules were first evaluated for their
inhibitory potency against recombinant human POP (Table 1). Figure 7 shows the dose-
response curves for covalent inhibitors 10a and 12a. As expected, while non-functionalized
pyrrolidine derivative 4a exhibited a potency of 160 nM, the measured Ki values for the
nitrile (5a), aldehyde (10a) and boronic ester (12a) derivatives were significantly lower. The
high reactivity of aldehydes as electrophiles has often been a major issue for developing safe
ACS Paragon Plus Environment
11

Journal of Medicinal Chemistry
Page 12 of 39
1
2
3
4
5
6
7
8
9
drugs, even in the discovery of Bortezomib (boronic acid proteasome inhibitor, Figure 1).33,34
In contrast, the lower reactivity of nitriles allowed medicinal chemists to use them as
warhead in drugs such as Vildagliptin (nitrile-containing covalent DPP-IV inhibitor, Figure
1). The reactivity of these warheads is further discussed below.
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. Dose-response curves of 10a (left) and 12a (right) against human recombinant POP
after 30-minute pre-incubation periods. Boronic ester 12a was additionally tested with a
two-hour pre-incubation period to allow for in situ hydrolysis of the (+)-pinanediol
protecting group.
As an observed effect of kinetic factors, the Ki decreases over time until equilibrium is
reached. While the nitrile and aldehyde derivatives reached equilibrium after 30 minutes of
pre-incubation, the boronic ester 12a required a longer incubation period. This was most
likely due to the rate of hydrolysis of the boronic ester to the boronic acid 11a, which is
required for enzyme binding. Residence time is largely controlled by binding kinetics of the
covalent ligand, illustrating the importance of this property.2,35,36 Introduction of fluorine
atoms onto the pyrrolidine ring of our lead compound 5a led to complete loss (15a) or a
ACS Paragon Plus Environment
12

Page 13 of 39
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
decrease (13a, 14a) in potency. A similar decrease in affinity was also observed for the non-
covalent inhibitors bearing a fluorine atom on the pyrrolidine ring (4a vs. 16a and 17a),
suggesting that factors other than the nitrile reactivity modulate potency.
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
Additional biophysical characterization experiments were performed for select inhibitors
10a, 12a, and control 1. To extract kinetic parameters, progress curve experiments were
conducted at various inhibitor concentrations. Data from these curves were then used to
extract the respective kobs values, which were further plotted against inhibitor concentration;
the resultant data was subsequently fitted to the corresponding equations (see Supporting
Information) in order to retrieve the inhibitors’ kinetic parameters. Finally, rapid dilution
experiments were performed to obtain residence time tR. (See Supporting Information for
the raw data curves of the above-mentioned experiments.) Kinetic parameters of each
compound are provided in Table 1. Unfortunately, any attempts to obtain kinetic parameters
for the non-covalent, the prolinonitrile, and the fluorinated prolinonitrile derivatives proved
unsuccessful, as the off rates were too quick to measure experimentally.
Table 1. Summary of the kinetic parameters measured experimentally.*
kon
(105 M1 s-1) (10-4 s-1)
0.4 ± 0.02 7.8 ± 2.2 3.9 ± 0.5 42 ± 5
< 1
< 1
8.2 ± 0.2 20 ± 0.8 80% (11%)
koff
Cpd Ki (nM)a Ki (nM)b Ki*(nM)c
tR (min) FAPd
1
1 ± 0.1
160 ± 40
25 ± 4
8 ± 1
< 5%37
<5%
20%
4a
5a
-
-
-
-
-
25 ± 4
3.5 ± 0.2
-
-
10a 4.0 ± 0.4 20 ± 9
1.86 ± 0.7
110
40e
±
12a
60 ± 10 29 ± 2
0.04 ± 0.01 2.3 ± 0.3 73 ± 10 <5%
ACS Paragon Plus Environment
13

Journal of Medicinal Chemistry
Page 14 of 39
1
2
3
4
22 ± 5f
5
6
7
8
3,300 ±
780
13a
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
<5%
<5%
<5%
14a 170 ± 40
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
>
100
15a
µM
16a 290 ± 50
17a 620 ± 90
-
-
-
-
-
-
-
-
-
-
<5%
<5%
aAffinity constant, measured by absorbance assay. bAffinity constant of the first step of the
c
binding event, measured by dilution experiments. Affinity constant of the second step of
inhibition. dinhibition at 100 µM (at 1 µM). e30 minute E—I pre-incubation time. f2 hour E—I
pre-incubation time. *missing parameters (-) indicate that the kinetics of the reaction were
too quick to measure experimentally
Computational study. To study these kinetic parameters and provide insight into the
development of effective prediction methods for covalent inhibitors, the quantum chemical
cluster approach (QCCA)38 was employed. Starting geometries were taken from crystal
structures. The ligands were truncated to focus on energetics of covalent bond
formation/breakage while maintaining the electronics of the electrophile (e.g., 10b as a
model for 10a).
ACS Paragon Plus Environment
14

Page 15 of 39
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 8. Data collected for boronic acid (a) and nitrile (b) as warhead. In blue is the energy
of the enzyme-catalyzed reaction and in green the same reaction with no enzyme. Non-polar
hydrogens are omitted for clarity. Similar data has been collected for all the probes and both
proteins (Figure S15).
ACS Paragon Plus Environment
15

Journal of Medicinal Chemistry
Page 16 of 39
1
2
3
4
5
6
7
8
9
As per the QCCA protocol, the binding site was restricted to the catalytic triad residues
along with other key residues, such as the backbone of residues contributing to the oxy-anion
hole (Figure 5c). The second step of the binding process was then simulated to acquire
several thermodynamic and kinetic parameters, such as binding energies and activation
energies for binding and unbinding, which together with enzyme mobility and non-bonded
interactions contribute to kon and koff (Table 2, Figure 8, SI). In order to develop a protocol
that would minimize calculation time, we decided to compute the energy at evenly
distributed distances only. As a result, the “ideal” distance or transition state distances were
not necessarily assessed, only close-to-minima structures. Although the search for the
energy at the optimal distances is expected to improve accuracy, it is also expected to
significantly increase computational time and hence decrease efficiency, as locating a
transition state is not a simple task. Rather, our method is expected to provide us with trends
which would be accurate enough to make informed decisions on whether a target is
covalently druggable. Considering this approximation, we also considered different levels of
theory, from semiempirical (AM1) to higher levels (PBE0/def2-TZVP/D3BJ).
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. Summary of the parameters obtained computationally (all values are in kcal/mol).
Eoff
Eon
ΔG (cov. – non-cov.)
Binding Energy
Cpd.
5b
POP
11.9
18.8
34.1
14.2
FAP
6.2
POP
6.2
1.8
6.0
8.4
FAP
7.5
POP
-5.7
FAP
1.3
POP
-36.0
-49.9
-63.5
-37.3
FAP
-27.0
-42.9
-55.8
-29.1
10b
11b
13b
16.9
26.0
12.6
3.9
-17.0
-28.1
-5.8
-13.0
-26.0
-6.7
<1.0 a
5.9
ACS Paragon Plus Environment
16

Page 17 of 39
Journal of Medicinal Chemistry
1
2
3
4
5
6
14b
9.8
5.7
8.5
6.6
10.6
10.2
-1.3
-7.1
4.9
-34.4
-39.2
-25.4
-32.8
15bb
13.7
10.9
-0.7
7
8
9
a
b
The compound forms a covalent bond without an energy barrier. computed with the
pseudo axial conformation (see text).
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
The computational data suggests that the warhead does in fact have a direct influence on
the kinetics and thus activity of the second step in ligand binding. According to this data, the
aldehyde and boronic acid are predicted to have longer residence times than any of the
nitrile derivatives, represented by significantly larger Eoff values. This is consistent with the
in-depth intrinsic electrophilicity analysis at the HF/def2-TZVP level of theory we have
conducted on compounds 5b, 10b and 11b (Tables S5-7) using our own QM package QUEMIST.
This analysis is based on computing the LUMO energies of the three compounds, alongside
intrinsic atomic orbital (IAO) atomic charges39 for the reactive atoms (boron, carbonyl
carbon and nitrile carbon) and total atomic nucleophilic superdelocalizabilities (TANS) for
the reactive atoms.40 The TANS is a measure of interaction with a nucleophile by an
electrophile, and can be reliably used to compare two or more sites for this specific
interaction. A high value of TANS is representative of a larger capacity of an atom to interact
with a nucleophile through their LUMO orbital. According to this data (Tables S5-7), the
boronic acid 11b has the lowest LUMO energy (7.4 kcal/mol lower than the aldehyde and 5.3
kcal/mol lower than the nitrile) along with a higher IAO atomic charge for the reactive atom
(0.74 vs. 0.33 for 10b and 0.15 for 5b) and a higher TANS for the reactive atom (9.84 vs. 1.24
for 10b and 1.06 for 5b). This analysis is also consistent with the binding energies presented
ACS Paragon Plus Environment
17

Journal of Medicinal Chemistry
Page 18 of 39
1
2
3
4
5
6
in Table 2, which suggests the boronic acid is the strongest binder/more intrinsically
reactive.
7
8
9
The data also suggested that the nitriles had a slightly higher activation energy (Eon) in FAP
than in POP. More surprisingly, the conversion of the aldehyde to the hemi-acetal and the
boronic acid to the boronate in the active site appear to possess a very low energy barrier.
This would imply that covalent bond formation is rapid, only limited by diffusion of the
ligand into the active site and reorientation of the electrophilic warhead to allow covalent
bond formation. This low energy barrier is attributed to two observations: (1) the ligands
appear to be pre-activated by Tyr473 as they approach the nucleophilic serine, and (2) the
transition states of these reactions resemble the transition state adopted by the natural
substrates and are highly stabilized through hydrogen bonding. Interestingly, in the case of
nitriles, the proximal Tyr473 hydroxyl group is properly positioned to transfer a proton to
the forming imine despite the sp2 character of this intermediate. While ligand-enzyme
kinetics were the focus of these computations, we also observed that in the case of the nitrile,
energy of the covalently bound state was not necessarily significantly lower than that of the
non-covalent complexes, keeping in agreement with low residence time. We next
investigated whether computing the full binding process was required. First, we computed
the correlation between the binding energies and Eoff for POP (Table 2; Figure S16),
according to the Bell-Evans-Polanyi principle.41 This principle establishes a linear
relationship between the activation energy and the enthalpy of reaction in the same family
of reactions. The R2 coefficient of 0.96 confirms a significant correlation between the binding
energies and Eoff, and suggest that future investigations should simply focus on ground
states, thus streamlining the process. We next computed theoretical half-lives t1/2 of both the
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
18

Page 19 of 39
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
initial bond formation as well as the bond breakage (Tables S3 and S4) which are
significantly lower than the residence times (Table 2). The bond formation/breaking was
significantly faster than the experimental residence time, suggesting that the inhibition
process is under thermodynamic control. Thus, the binding energies should also correlate
with the experimentally determined Ki values. We plotted the relevant graph (Figure S16)
and we observed no correlation between the binding energies and Ki values (R2 = 0.17).
Several factors are responsible for this apparent lack of correlation: first, we computed the
binding energies of truncated fragments, while the Ki values were determined using the full
molecules. While we tried to obtain Ki values for the fragments alone, they showed no
activity in POP or FAP (data not shown), revealing the critical contribution of the rest of the
molecules to the binding affinity. Secondly, the Ki values were computed in solution, while
our computations were performed exclusively in the gas phase. We believe that for accurate
solvent effects an explicit solvent should be used,42 which can be reliably done using
molecular dynamics simulations. However, this is beyond the scope of our computations,
which intend to offer a qualitative analysis and trends with respect to reactive warheads. In
addition, the use of small fragments assumes that the warheads are properly positioned to
form a reversible covalent bond. Adding groups to these fragments certainly modulates these
optimal alignments hence the Ki. Finally, we observed that association rate of 10a showed a
significant temperature dependence, which we attribute to the large conformational
rearrangement of POP that accompanies ligand binding,43 a motion not considered when
computationally binding fragments.
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
19

Journal of Medicinal Chemistry
Page 20 of 39
1
2
3
4
5
6
7
8
9
It should be noted that the objective of this research is to define whether a covalent bond
is possible (covalent druggability) and which warhead would be optimal but not whether a
given molecule is to be a strong inhibitor.
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
Although the computational trends match the experimental trends, the computed low
energy barriers for the aldehyde and boronic acid contrast with the commonly reported slow
covalent binding step (Figure 3a). One such disagreement between computations and
experiment exists for compound 15a/b. A closer look at the models revealed that the
preferred conformation (pseudo-axial fluorine) cannot geometrically form the covalent
bond and must rearrange into the less energetically favored pseudo-equatorial
conformation. We and others have previously found that fluorine atoms have major control
on five-membered ring conformations.42 This phenomenon was also observed in FAP by
Jansen et al., with inhibitors bearing the cis-fluorine as in 14c exhibiting potency two orders
of magnitude greater than those bearing the trans-fluorine as in 15c.29 Computations of
binding energies (from dissociated complexes to covalently bound complexes) suggests that
POP binds these fluorinated proline mimics more tightly than FAP. As discussed above, this
apparent discrepancy between experimental and computational results stems from the use
of ideally positioned fragments used in the computational investigations of the covalent
bond formation which experimentally do not inhibit the enzymes vs. the larger molecules
used in the in vitro assays.
By means of progress curve analyses and rapid dilution experiments, relevant parameters
such as the Ki (affinity constant of the non-covalent component of binding), the Ki* (affinity
of the second step of inhibition), kon (association rate) and koff (dissociation rate) were
experimentally determined (Table 1, see SI for details). As a control experiment, Cbz-Pro-
ACS Paragon Plus Environment
20

Page 21 of 39
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
prolinal (1) was first investigated, and the data obtained was in agreement with a previous
report.31 The results for our designed inhibitors confirmed that the intrinsic reactivity of the
warhead of the covalent inhibitor greatly influenced the on- and off-rates of ligand binding.
The boronic ester inhibitor (12a) displayed the slowest on-rate of inhibition (kon). As
discussed above, hydrolysis of the boronic ester was an additional factor in the on-rate of
inhibition.44,45 Similarly, aldehyde 10a, although displaying significantly faster on-rates than
boronic ester 12a, interacted relatively slowly compared to nitrile 5a, where no slow-binding
was detected. The relatively slow binding of the aldehyde-containing inhibitor was
attributed to the presence of a pre-existing equilibrium in aqueous solution of an active
aldehyde form and an inactive hydrate form of the ligand, while no such equilibrium exists
with nitriles. While fast binding was predicted when covalent bond formation was computed
(k2 in Figure 3a), kon measures the entire two-step process.
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 agreement with the computations, aldehydes and boronic acids have the longest
residence times. Experimentally, our aldehyde compound 10a has a residence time 20
minutes shorter than that of Cbz-Pro-prolinal (1). In contrast, boronic ester 12a displayed
nearly a four-fold longer residence time, rationalized by the additional stabilizing
interactions of the resultant boronic acid hydroxyl groups with residues in the active site of
POP, as proposed by docking and QM studies (Figures 5b, 5c). Nitrile inhibitor 5a displayed
a very short residence time, despite bearing the same scaffold as 12a and 10a, correlating
with the low Eoff computed in POP. These short residence times may also result from 5a
binding non-covalently as suggested by the small difference in energy between covalently
bound and non-covalently bound complexes (Table 2). From the experimental data, we can
conclude that the nitrile does not provide a strong enough covalent adduct needed to
ACS Paragon Plus Environment
21

Journal of Medicinal Chemistry
Page 22 of 39
1
2
3
4
5
6
7
8
9
maintain a longer-bound time in POP, which is in agreement with the computations. As
discussed above, covalent ligand-protein complexes of both 12a and 10a with POP exhibit a
tetrahedral geometry, resembling that of the transition state geometry formed by the
enzyme while carrying out its peptidase activity. In contrast, the covalent complex resulting
from reaction with nitrile 5a exhibits a trigonal planar geometry (Figure 3d).
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 computations suggested that the pre-orientation of the ligand by the tyrosine
facilitates rapid bond formation and, coupled to the small activation barrier, that the covalent
binding step may be very fast (Figure 8). This prediction is supported by kinetic results we
recently obtained for a related scaffold in which addition of a nitrile warhead actually led to
a moderate increase in the binding rate compared to the equivalent non-covalent inhibitor.46
In contrast, the association rate for the covalent inhibitor would be lower than that of the
non-covalent analogue if bond formation were indeed rate-limiting.
The correlation between computations and experiments provides validation for the QCCA
method’s ability to predict the overall trends in binding kinetics in POP for covalent ligands
and gave us confidence in the data for FAP, an enzyme which is extremely difficult to express
and handle experimentally. Our computations indicate that the binding of our truncated
nitrile ligands to FAP should have greater Eon and smaller Eoff likely rendering it less active,
while the aldehyde and boronic acid remain reactive enough to inhibit the enzyme. This
prediction is consistent with our literature survey, which revealed that although many POP
inhibitors feature a nitrile, most potent FAP inhibitors feature a boronic acid or an activated
nitrile (Figure 1). The computations were also in agreement with our experimental data on
FAP inhibitory activity of compound 5a (20% inhibition at 100 µM) and compound 10a (80%
ACS Paragon Plus Environment
22

Page 23 of 39
Journal of Medicinal Chemistry
1
2
3
4
5
6
inhibition at 100 µM), although the large bicyclic core of our inhibitor may also hinder
binding to FAP.
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
Conclusion
In conclusion, this study aimed to streamline the investigation of two covalently druggable
targets using computational methods to provide a better understanding of experimentally-
obtained thermodynamic and kinetic factors involved in their covalent inhibition and to
further study the factors controlling general covalent drug potency. With POP, an enzyme
which can easily be expressed and purified, experimental and computational data were in
excellent agreement and revealed that the commonly held belief that covalent bond
formation is rate limiting for covalent inhibitors does not necessarily stand when highly
reactive enzyme residues and/or inhibitors are involved. Our validated computational
protocol then rationalized the wide use of boronic acids in FAP inhibition versus the more
commonly used nitrile in POP inhibition. The results presented here are a first step towards
using computational methods to complete a larger study of covalent binding kinetics, a
concept which is entirely unaccounted for in current computational prediction tools such as
molecular docking. The ability to integrate kinetic data into prediction tools will improve the
ability to rank ligands shown to be active.
Our collaborative approach to this model system aims to facilitate future covalent drug
discovery endeavors. By applying our computational methods to predict the relative
reactivity of a newly-discovered target’s catalytic residue, biologists and chemists can
determine whether the target is covalently druggable and therefore more efficiently design
ACS Paragon Plus Environment
23

Journal of Medicinal Chemistry
Page 24 of 39
1
2
3
4
5
6
the most promising drug candidates. These predictions will, in the long run, save valuable
time and resources in the very costly drug discovery and development processes.
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
Experimental Section.
Chemistry
General information. All commercially available reagents were used without further
purification unless otherwise stated. The 4 Å molecular sieves were dried at 100°C prior to
use. FTIR spectra were recorded using a Perkin-Elmer Spectrum One FT-IR. 1H and 13C NMR
spectra were recorded on Varian Mercury 400 MHz, Varian 300 MHz, Unity 500 MHz, Bruker
400 MHz, or Bruker 500 MHz spectrometers. Chemical shifts are reported in ppm using the
residual of deuterated solvents as internal standard. Thin layer chromatography
visualization was performed by UV or by development using KMnO4, H2SO4/MeOH, Mo/Ce,
CAM solutions. Chromatography was performed on silica gel 60 (230-40 mesh) or using
Biotage Isolera One purification system with ZIP cartridges. Low resolution mass
spectrometry was performed by ESI using a Thermoquest Finnigan LCQ Duo. High resolution
mass spectrometry was performed by EI peak matching (70 eV) on a Kratos MS25 RFA
double focusing mass spectrometer or by ESI on a Ion Spec 7.0 T FTMS at McGill University.
Prior to biological testing, reversed phase HPLC (water and MeCN or MeOH gradient) was
used to verify the purity of compounds on an Agilent 1100 series instrument equipped with
VWD-detector, C18 reverse column (Agilent, Zorbax Eclipse XDB-C18 150 mm 4.6 mm, 5
µm), UV detection at 254 nm. All biologically tested compounds exhibited ≥95% purity.
Measured purities for all tested compounds are listed in Table S1 in the Supporting
Information.
ACS Paragon Plus Environment
24

Page 25 of 39
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
General Procedure for peptidic coupling. The carboxylic acid 18 (1 eq) was suspended in
anhydrous DCM (0.1 M) under argon atmosphere, and Et3N (5 eq) was added. The resultant
solution was cooled to 0°C, and pivaloyl chloride (1.1 eq) was added. After 1h of stirring at
0°C, the amine (1.1 eq) was added, and the reaction stirred at room temperature overnight.
Water was added, and the product was extracted with EtOAc or with DCM (depending on the
amount of original DCM solvent). The combined organic layers were washed with 1M HCl,
saturated NaHCO3, and brine, dried over Na2SO4, filtered, and concentrated in vacuo. The
crude residue was purified by flash chromatography on a silica gel column to give the
product. Residues were triturated in hexanes and/or Et2O and filtered under vacuum to give
solids.
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-benzyl-7-(pyrrolidine-1-carbonyl)isoindolin-1-one (4a) Compound 4a was synthesized
following the general procedure for peptidic coupling, using pyrrolidine as the
1
corresponding amine (40%). H NMR (500 MHz, Chloroform-d) δ 7.52 (t, J = 7.6 Hz, 1H),
7.37 (dd, J = 7.5, 1.7 Hz, 2H), 7.34 – 7.25 (m, 5H), 4.97 – 4.80 (m, 1H), 4.73 – 4.50 (m, 1H),
4.24 (s, 2H), 3.92 – 3.75 (m, 1H), 3.75 – 3.59 (m, 1H), 3.42 – 3.20 (m, 1H), 3.17 – 2.99 (m,
1H), 2.06 – 1.73 (m, 4H); 13C NMR (126 MHz, CDCl3) δ 24.61, 26.01, 45.70, 46.48, 48.25,
49.33, 77.16, 123.34, 126.22, 127.85, 128.28 (2C), 128.39 (2C), 128.91, 131.80, 135.51,
136.94, 141.76, 166.88, 167.02. Spectral and experimental data previously published by our
group.13
1-(2-benzyl-3-oxoisoindoline-4-carbonyl)pyrrolidine-2-carbonitrile (5a) Compound 5a
was synthesized following the general procedure for peptidic coupling, using (S)-
1
pyrrolidine-2-carbonitrile pTsOH salt as the corresponding amine (92%). H NMR (400
MHz, Acetone-d6) δ 7.70 – 7.48 (m, 2H), 7.47 – 7.38 (m, 1H), 7.37 – 7.13 (m, 5H), 5.03 – 4.54
ACS Paragon Plus Environment
25

Journal of Medicinal Chemistry
Page 26 of 39
1
2
3
4
5
6
7
8
9
(m, 3H), 4.41 (s, 2H), 3.79 – 3.66 (m, 0.5H), 3.38 – 3.17 (m, 1.5H), 2.48 – 2.34 (m, 1H), 2.34 –
2.21 (m, 1H), 2.21 – 2.07 (m, 1H), 2.04 – 1.96 (m, 1H); Carbon peaks reported for major
rotamer. 13C NMR (101 MHz, Acetone-d6) δ 25.75, 29.84, 31.25, 46.56, 47.04, 48.58, 50.05,
119.70, 125.15, 126.83, 128.31, 128.84 (2C), 129.11, 129.55 (2C), 132.54, 134.47, 138.48,
143.32, 167.22, 167.88. Spectral and experimental data previously published by our group.13
(S)-1-(2-benzyl-3-oxoisoindoline-4-carbonyl)pyrrolidine-2-carbaldehyde (10a). Oxalyl
chloride (108 mg, 1.2 eq) was dissolved in DCM (3 mL), and the solution was cooled to –
78°C. DMSO (139 mg, 2.5 eq, in DCM, 2 mL) was added dropwise, and the solution stirred for
2 minutes. 20 (250 mg, 1 eq, in 2 mL DCM) was added dropwise, and the solution stirred for
15 minutes. Et3N (361 mg, 5 eq) was added, and the solution stirred 15 mins. Water was
added, and the mixture was warmed to room temperature. The product was extracted with
EtOAc, and the combined organic layers were washed with brine, dried over Na2SO4, filtered,
and concentrated in vacuo to give an orange oil, which was purified by flash chromatography
on a silica gel column (eluent 100% EtOAc) to give the product as a white solid (100 mg,
40%). R.f. 0.30 (100% EtOAc); IR (film) νmax (cm-1) 3387, 3006, 1675, 1626, 1601, 1434; 1H
NMR (500 MHz, Acetone-d6) δ 10.13 – 9.17 (m, 1H), 7.76 – 7.20 (m, 8H), 4.94 – 4.58 (m, 2H),
4.56 – 4.18 (m, 3H), 3.91 – 3.11 (m, 2H), 2.29 – 2.09 (m, 1H), 2.02 – 1.76 (m, 3H); 13C peaks
reported for the major rotamer. 13C NMR (126 MHz, Acetone) δ 14.50, 20.83, 23.27, 26.05,
27.59, 29.84, 32.28, 46.56, 49.28, 50.01, 60.53, 65.74, 124.86, 126.43, 128.31, 128.83 (2C),
129.54 (2C), 129.56, 132.49, 135.22, 138.54, 143.36, 167.38, 167.98, 170.88, 202.63; HRMS
(ESI+): calculated for [C21H20N2O3 + H]+, 349.15467; found, 349.15418.
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-benzyl-7-((R)-2-((3aR,4R,6R,7aS)-3a,5,5-trimethylhexahydro-4,6-
methanobenzo[d][1,3,2] dioxaborol-2-yl)pyrrolidine-1-carbonyl)isoindolin-1-one (12a).
ACS Paragon Plus Environment
26

Page 27 of 39
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Compound 12 was synthesized following the general procedure for peptidic coupling, using
19 as the corresponding amine. The crude residue was purified using (1:1 EtOAc/hexanes)
as the eluent system (43%, white foam); IR (film) νmax (cm-1) 3225, 2922, 1686, 1606, 1451;
Peaks and coupling constants reported for major rotamer. Full integrated proton spectrum
provided. 1H NMR (500 MHz, Chloroform-d) δ 7.52 (t, J = 7.6 Hz, 1H), 7.42 (d, J = 7.5 Hz, 1H),
7.37 (d, J = 7.6 Hz, 1H), 7.35 – 7.26 (m, 5H), 4.84 (d, J = 14.8 Hz, 1H), 4.68 (d, J = 14.8 Hz,
1H), 4.39 (dd, J = 8.9, 2.2 Hz, 1H), 4.25 – 4.19 (m, 2H), 3.47 (dt, J = 9.2, 6.3 Hz, 1H), 3.42 –
3.31 (m, 1H), 3.22 (dd, J = 17.3, 7.9 Hz, 1H), 2.41 – 2.12 (m, 3H), 2.09 (t, J = 5.5 Hz, 1H), 2.01
– 1.85 (m, 5H), 1.61 (d, J = 10.9 Hz, 1H), 1.45 (s, 3H), 1.29 (s, 3H), 0.86 (s, 3H); 13C NMR (126
MHz, CDCl3) δ 24.28, 26.43, 27.29, 27.30, 27.33, 28.92, 35.82, 38.39, 39.79, 44.60, 46.53,
48.15, 49.31, 51.57, 77.16, 78.08, 86.04, 123.60, 127.03, 127.84, 128.42 (2C), 128.67, 128.92
(2C), 128.95, 131.65, 137.08, 141.75, 166.58, 166.69; 11B NMR (161 MHz, CDCl3) δ 32.45;
HRMS (ESI+): calculated for [C30H35B1N2O4 + H]+, 499.27736; found, 499.27634.
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
(S)-1-(2-benzyl-3-oxoisoindoline-4-carbonyl)-4,4-difluoropyrrolidine-2-carbonitrile
(13a) Compound 13a was synthesized following the general procedure for peptidic coupling,
using 21 as the corresponding amine. The crude residue was purified using 3:1
EtOAc/hexanes as the eluent system (30%, white solid); (Rf = 0.53; 3:1 EtOAc/hexanes); IR
(film) νmax (cm-1) 2918, 2850, 1666, 1604, 1411, 1108; 1H NMR (500 MHz, CDCl3) δ 7.75 –
7.57, (m, 2H), 7.57 – 7.45 (m, 2H), 7.43 – 7.28 (m, 4H), 5.34 (br s, 0.5 H), 4.96 (d, J = 14.9 Hz,
0.5 H), 4.89 – 4.80 (m, 1H), 4.71 (br s, 0.5H), 4.62 (d, J = 14.9 Hz, 0.5 H), 4.41 – 4.24 (m, 2.5
H), 4.17 (t, J = 15.1 Hz, 0.5 H), 3.82 (br s, 0.5 H), 3.63 (br s, 0.5 H), 3.08 – 2.89 (m, 1H), 2.83
(t, J = 13.2 Hz, 0.5 H), 2.70 (t, J = 13.2 Hz, 0.5H); Peaks include rotamers. 13C NMR (101 MHz,
CDCl3) δ 46.69, 49.62, 49.79, 52.76, 53.92 (t, J = 31.9 Hz), 77.16, 116.78, 124.96, 125.13,
ACS Paragon Plus Environment
27

Journal of Medicinal Chemistry
Page 28 of 39
1
2
3
4
5
6
7
8
9
126.89, 127.97, 128.14, 128.23, 128.37 (2C), 129.11 (2C), 129.15, 131.68, 132.36, 132.79,
136.30, 136.51, 141.93, 167.06, 167.52; HRMS (ESI+): calculated for [C21H17F2N3O2 + H]+,
382.13616; found, 382.13599.
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
(2S,4S)-1-(2-benzyl-3-oxoisoindoline-4-carbonyl)-4-fluoropyrrolidine-2-carbonitrile
(14a) Compound 14a was synthesized following the general procedure for peptidic coupling,
using 22 as the corresponding amine. The crude residue was purified using 100% EtOAc as
the eluent system (30%, white solid); Rf: 0.18 (100% EtOAc); IR (film) νmax (cm-1) 3012,
1
2920, 1682, 1653, 1605, 1409, 1212; H NMR (500 MHz, Chloroform-d) δ 7.72 – 7.41 (m,
3H), 7.40 – 7.20 (m, 5H), 5.58 – 5.18 (m, 1.6H), 4.87 (dd, J = 18.8, 14.9 Hz, 1H), 4.70 – 4.52
(m, 1.4H), 4.40 – 4.20 (m, 2H), 4.20 – 4.02 (m, 1H), 3.79 – 3.45 (m, 1H), 2.78 – 2.49 (m, 2H);
Carbon peaks reported for both rotamers. 13C NMR (126 MHz, CDCl3) δ 36.51, 36.67, 38.24,
38.40, 44.97, 46.60, 46.67, 47.12, 49.63, 49.77, 53.18, 53.37, 54.14, 54.33, 77.16, 90.34, 91.32,
91.77, 92.76, 117.93 (2C), 124.53, 124.74, 126.89 (2C), 127.53 (2C), 128.10, 128.15, 128.27
(2C), 128.33 (2C), 129.07 (2C), 129.10 (2C), 132.38, 132.49, 132.66, 132.69, 136.38, 136.47,
141.77 (2C), 166.89, 166.91, 167.64, 167.66; HRMS (ESI+): calculated for [C21H18FN3O2 +
H]+, 364.14558; found, 364.14500.
(2S,4R)-1-(2-benzyl-3-oxoisoindoline-4-carbonyl)-4-fluoropyrrolidine-2-carbonitrile
(15) Compound 15a was synthesized following the general procedure for peptidic coupling,
using 23 as the corresponding amine. The crude residue was purified using 1:1
EtOAc/hexanes as the eluent system (30%, white solid); (Rf = 0.1; 1:1 EtOAc/hexanes); IR
1
(film) νmax (cm-1) 3008, 2920, 1686, 1648, 1603, 1413, 1205; H NMR (400 MHz, Acetone-
d6) δ 8.38 – 7.55 (m, 2H), 7.52 – 7.24 (m, 6H), 5.36 (dt, J = 52.2, 3.7 Hz, 1H), 5.15 – 4.65 (m,
3H), 4.55 – 4.17 (m, 2H), 3.96 – 3.43 (m, 2H), 2.99 – 2.83 (m, 1H), 2.75 – 2.52 (m, 1H); 13C
ACS Paragon Plus Environment
28

Page 29 of 39
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
NMR (101 MHz, Acetone) δ 29.84, 37.61 (d, J = 22.3 Hz), 45.37, 46.59, 49.99, 54.93 (d, J =
22.1 Hz), 88.77 (d, J = 177.5 Hz) 118.79, 125.61, 127.79, 128.31, 128.87, 129.56 (4C), 132.52,
133.90, 138.53, 143.46, 166.90, 167.78; HRMS (ESI+): calculated for [C21H18F1N3O2 + H]+,
364.14558. *13C—F peak suppressed, HSQC provided as 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
(S)-2-benzyl-7-(3-fluoropyrrolidine-1-carbonyl)isoindolin-1-one (16a) / (R)-2-benzyl-7-
(3-fluoropyrrolidine-1-carbonyl)isoindolin-1-one (17a) Compounds 16a/17a were
synthesized following the general procedure for peptidic coupling, using 24/25 as the
corresponding amines. The crude residues were purified using 100% EtOAc as the eluent
system (90%/90%, white solids); As they are enantiomers, they gave identical spectral
properties. Rf: 0.18 (100% EtOAc); IR (film) νmax (cm-1) 3005, 1682, 1627, 1607, 1433,
1
1206; H NMR (500 MHz, Chloroform-d) δ 7.54 (td, J = 7.5, 2.6 Hz, 1H), 7.40 (t, J = 7.5 Hz,
2H), 7.34 – 7.22 (m, 5H), 5.26 (ddt, J = 82.8, 52.9, 3.9 Hz, 1H), 4.87 – 4.56 (m, 2H), 4.32 – 4.19
(m, 2H), 4.15 – 3.75 (m, 2H), 3.63 – 3.32 (m, 2H), 2.46 – 2.03 (m, 2H); Carbon peaks reported
for both rotamers. 13C NMR (126 MHz, CDCl3) δ 31.20, 31.37, 32.74, 32.91, 43.60, 45.90,
46.47, 49.37, 52.41, 52.60, 54.22, 54.40, 77.16, 92.06 (d, J = 175.3 Hz), 92.80 (d, J = 177.7
Hz), 123.68, 123.75, 126.28, 126.55, 127.86, 127.89, 128.20 (2C), 128.31 (2C), 128.34 (2C),
128.90 (2C), 128.92 (2C), 131.87, 131.99, 134.57, 134.59, 136.75, 136.81, 141.76, 141.82,
166.79, 166.83, 167.27, 167.31; HRMS (ESI+): calculated for [C20H19FN2O2 + Na]+,
361.1323; found, 361.132.
(S)-2-benzyl-7-(2-(hydroxymethyl)pyrrolidine-1-carbonyl)isoindolin-1-one
(20).
Compound 20 was synthesized following the general procedure for peptidic coupling, using
L-prolinol as the corresponding amine. The crude residue was purified using 100% EtOAc as
the eluent system (87%, white solid); (Rf = 0.15; 100% EtOAc); IR (film) νmax (cm-1) 3434,
ACS Paragon Plus Environment
29