Cell-Active Small Molecule Inhibitors of the DNA-Damage Repair Enzyme Poly(ADP-ribose) Glycohydrolase (PARG): Discovery and Optimization of Orally Bioavailable Quinazolinedione Sulfonamides

Article abstract of DOI:10.1021/acs.jmedchem.8b01407

DNA damage repair enzymes are promising targets in the development of new therapeutic agents for a wide range of cancers and potentially other diseases. The enzyme poly(ADP-ribose) glycohydrolase (PARG) plays a pivotal role in the regulation of DNA repair mechanisms; however, the lack of potent drug-like inhibitors for use in cellular and in vivo models has limited the investigation of its potential as a novel therapeutic target. Using the crystal structure of human PARG in complex with the weakly active and cytotoxic anthraquinone 8a, novel quinazolinedione sulfonamides PARG inhibitors have been identified by means of structure-based virtual screening and library design. 1-Oxetan-3-ylmethyl derivatives 33d and 35d were selected for preliminary investigations in vivo. X-ray crystal structures help rationalize the observed structure-activity relationships of these novel inhibitors.

Full text of DOI:10.1021/acs.jmedchem.8b01407

Subscriber access provided by University of Sunderland
Article
Cell-active Small Molecule Inhibitors of the DNA-damage Repair
Enzyme Poly(ADP-ribose) Glycohydrolase (PARG): Discovery and
Optimization of Orally Bioavailable Quinazolinedione Sulfonamides
Bohdan Waszkowycz, Kate M. Smith, Alison E McGonagle, Allan M. Jordan, Ben
Acton, Emma Fairweather, Louise A. Griffiths, Niall Hamilton, Nicola Hamilton, James
Hitchin, Colin P. Hutton, Dominic James, Clifford D. Jones, Stuart Jones, Daniel P.
Mould, Helen Small, Alexandra Stowell, Julie Tucker, Ian Waddell, and Donald Ogilvie
J. Med. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.jmedchem.8b01407 • Publication Date (Web): 07 Nov 2018
Downloaded from http://pubs.acs.org on November 8, 2018
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 88
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Cell-active Small Molecule Inhibitors of the DNA-damage Repair Enzyme
Poly(ADP-ribose) Glycohydrolase (PARG): Discovery and Optimization of
Orally Bioavailable Quinazolinedione Sulfonamides
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Bohdan Waszkowycz , Kate M. Smith , Alison E. McGonagle , Allan M. Jordan†,#,*, Ben Acton†,
†,#
†,#
†,#
†
†
†
†
Emma E. Fairweather , Louise A. Griffiths , Niall M. Hamilton , Nicola S. Hamilton , James R.
†
†
†
§
†
†
Hitchin , Colin P. Hutton , Dominic I. James , Clifford D. Jones , Stuart Jones , Daniel P. Mould ,
†
†
,∞
†
†
Helen F. Small , Alexandra I. J. Stowell , Julie A. Tucker , Ian D. Waddell and Donald J. Ogilvie .
†Cancer Research UK Manchester Institute, The University of Manchester, Alderley Park,
Maccelsfield, SK10 4TG, UK
§Oncology Innovative Medicines Unit, AstraZeneca, Alderley Park, Macclesfield, Cheshire, SK10
4
TG, UK
Structure and Biophysics, Discovery Sciences, AstraZeneca, Alderley Park, Macclesfield, Cheshire,
SK10 4TG, UK
∞Current address: Centre for Immunology and Infection, Department of Biology and HYMS,
University of York, Heslington, York, YO10 5DD, UK
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 2 of 88
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
ABSTRACT
DNA damage repair enzymes are promising targets in the development of new therapeutic agents for
a wide range of cancers and potentially other diseases. The enzyme poly(ADP-ribose) glycohydrolase
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
(PARG) plays a pivotal role in the regulation of DNA repair mechanisms; however the lack of potent
drug-like inhibitors for use in cellular and in vivo models has limited the investigation of its potential
as a novel therapeutic target. Using the crystal structure of human PARG in complex with the weakly
active and cytotoxic anthraquinone 8a, novel quinazolinedione sulfonamides PARG inhibitors have
been identified by means of structure-based virtual screening and library design. 1-Oxetan-3-
ylmethyl derivatives 33d and 35d were selected for preliminary investigations in vivo. X-ray crystal
structures help rationalize the observed structure-activity relationships of these novel inhibitors.
INTRODUCTION
Exploiting the DNA damage response for cancer therapy relies on the fact that rapidly proliferating
cancer cells have higher incidences of DNA damage, defective DNA repair pathways, genomic
instability, or a combination of these facets.1,2,3 Poly(ADP-ribosylation) is a post-translational
modification that plays an important role in the repair of damaged sections of DNA.4,5,6 Poly(ADP-
ribose) polymerases (PARPs), in particular PARP1, signal the presence of DNA damage and facilitate
DNA repair (Figure 1). Upon DNA damage, PARP1 binds to single-strand break (SSB) sites and
autoribosylates using NAD+ as a substrate, to form poly(ADP-ribose) (PAR) chains. These PAR
chains serve to recruit DNA repair proteins, such as XRCC1, to the site of DNA damage. Poly(ADP-
ribose) glycohydrolase (PARG) systematically degrades the ADP-ribose polymers on the PARP
enzyme, which is essential for DNA repair to occur; however, the precise order of events in this
catalytic cycle is still unknown. Figure 1 shows how inhibiting PARG leads to the persistence of PAR
chains, with predicted consequences on NAD recycling, PARP recycling, and SSB repair. Although
PARP inhibitors have received much attention as cancer therapies,7,8 with olaparib approved by the
9
FDA, the dearth of selective, cell-permeable small molecule PARG inhibitors has hampered
ACS Paragon Plus Environment

Page 3 of 88
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
assessment of the therapeutic potential of targeting PARG in human cancer.10,11 Additionally, whilst
there are 17 known PARP family members, of which only a subset are inhibited by the PARP
4
inhibitors, no close homologs of PARG exist. This provides an attractive target for drug discovery,
given the cell’s apparent reliance on this single enzyme in nuclear DNA damage repair.12
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 1. Consequences of PARG inhibition in DNA SSB repair
Crystal structures of the bacterial, protozoal, and mammalian PARG catalytic domains have served to
elucidate the structural basis of PAR substrate recognition and catalytic mechanism.13,14,15,16,17,18,19 The
C-terminal catalytic domain of PARG comprises a macrodomain, an evolutionarily conserved subunit
that binds ADP-ribose and is involved in the regulation of PAR turnover and signalling,20 with a
PARG-specific catalytic loop. PARG demonstrates a preference for hydrolyzing the long, unbranched
chains of PAR, which interact with an extended substrate binding site that is mostly solvent-exposed
and polar in character.
To date, identification of potent and selective PARG inhibitors has been hampered by the lack of a
suitable HTS-compatible assay system, prompting our recent development of a robust Homogeneous
Time-Resolved Fluorescence (HTRF) based assay format.21 Inhibitors reported in the literature
include: ADP-ribose mimetics (e.g., ADP-HPD, 1),22 tannins (e.g., mono-galloyl glucose, 2),23
salicylanilides (3),24 rhodanine-based compounds (4),25 and phenolic hydrazide hydrazones (5)26
(Figure 2). These compounds are typically weakly potent, non-specific inhibitors with poor cell
1
1
permeability, only demonstrating activity in cell lysates. For example, compounds 3 and 4 did not
demonstrate significant PARG inhibitory activity in our biochemical assay (activities > 70 µM in our
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 4 of 88
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
27
hands). Although compound 4 contains the known pan-assay interference (PAINs) motif rhodanine,
it has been successfully crystallized in complex with protozoal PARG, demonstrating binding to the
ADP-ribose site.24
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
NH2
O
O
OH
OH
OH
HO
HO
N
N
N
HO
HO
N
O
O
HO
OH OH
N
O
O
O
P
P
NH
OH
HN
O
HO
O
O
HO
OH
HO
1
OH
2
Cl
O
NH
N
O
O
O
N
Cl
S
S
Cl
N
OH
OH
Cl
H
N
HO
5
O
3
Cl
OH
Cl
O
4
HO
Figure 2. Reported PARG inhibitors
Given the lack of drug-like, cell permeable inhibitors reported to date, there remains a need for
selective tool compounds with improved physicochemical properties to validate the therapeutic
potential of PARG inhibition.11 To address this need, we recently reported the first cell-active
chemical probes against this enzyme.12,28,29 Herein, we describe the early discovery of several novel
lead-like chemotypes arising from a programme of virtual screening and structure-guided library
design, inspired by the crystal structure of a weakly active HTS hit bound to human PARG. We
further report our efforts to optimize one of these chemotypes in order to afford more potent in vitro
tool compounds, and derivatives with physicochemical properties suitable for preliminary in vivo
pharmacokinetic (PK) assessment.
CHEMISTRY
For preliminary hit expansion, compounds were generally prepared using traditional chlorosulfonation
techniques to prepare the sulfonyl chloride, which was then used for sulfonamide formation.
ACS Paragon Plus Environment

Page 5 of 88
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
Anthraquinones 8ao were prepared in this manner as detailed in Scheme 1, as was compounds 16
described in Scheme 3, with purification by preparative HPLC leading to variable recoveries of
material. Symmetrically substituted quinazolinediones 12ac, 13a,b were prepared by alkylation of
the quinazolinedione, followed by chlorosulfonation and sulfonamide formation (Scheme 2). For the
unsymmetrically substituted quinazolinediones 22ah, a route was developed which installed the
acetyl protected alkyl sulfonamide and allowed diversification in the final step, as detailed in Scheme
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
4
.
a
Scheme 1: Synthesis of analogues 8ao
O
O
O
O
O
O O
S
O O
S
NaO S
R
3
a)
b)
Cl
N
H
O
6
7
8a-o
a
Reagents: a) SOCl , DMF, reflux, 53%; b) RNH , Et N, DCM, 15%quant.
2
2
3
a
Scheme 2: Synthesis of 12ac, 13a,b
O
O
O
O
R
O
a)
b)
ClO S
2
R
O
R'HNO2S
R
O
N
c)
NH
N
N
N
R
N
O
N
R
N
R
H
9
10a,b
11a,b
12a-c, 13a,b
a
Reagents: a) RI, K CO , DMF, 3041%; b) ClSO H, 60 C, 6467%; c) R’NH , Et N, DCM,
2
3
3
2
3
1
876%.
a
Scheme 3: Synthesis of 16
O O
S
O O
S
b), c) or d)
a)
Ar
Cl
Ar
RHN
16
Ar
14
15
a
Reagents: a) ClSO H, 60 C, 70%; b) RNH , i-Pr NEt, THF, DCM, 43%.
3
2
2
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 6 of 88
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
a
Scheme 4: Synthesis of Analogues 22at
O
O
O
ClO S
a)
b)
2
N
N
N
H
NH2
N
O
N
H
O
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
H
17
18
19
O
O
O
O
O
e)
O
d)
S
c)
20
S
R = H
N
N
N
N
R
H
f), g) or h)
N
H
O
2
1 R = Ac
N
R
O
22a-h
a
Reagents: a) CDI, DMF, 140 C, 90%; b) ClSO H, 60 C, 99%; c) 1-methylcyclopropanamine
3
hydrochloride, Et N, DCM, 89%; d) Ac O, pyridine, DMAP, 83%; e) BrCH CN, Cs CO , DMF,
3
2
2
2
3
7
5%; f) R-X (where X = I or Br), K CO , DMF, then NH , THF, 28-88%; g) R-X (where X = Cl, Br
2 3 3
or I), K CO , DMF, then K CO , MeOH 2032%; h) R-OH, DIAD, PS-PPh , DMF, then K CO ,
2
3
2
3
3
2
3
MeOH, 93%.
Scheme 5 describes the synthesis of 27a–au, with elaboration at the N3 vector. The quinazolinedione
core was formed from N-methylanthranilic acid and sodium cyanate. Subsequent chlorosulfonylation
and sulfonamide formation afforded key intermediate 26, which was alkylated to give 1-methyl-3-
substituted quinazolinediones 27a-au. Scheme 6 shows the synthesis of quinazolinediones with
elaboration at both N1 and N3 vectors. Chlorosulfonylation of isatoic anhydride afforded 30. Careful
temperature control in the subsequent sulfonamide formation was vital to minimize anhydride ring-
opening with amine 28. Once the sulfonamide formation was complete at a low temperature, the
appropriate primary amine was added to the reaction mixture and warmed to room temperature (RT)
to induce anhydride ring-opening and decarboxylation to give amide intermediates 31a–d. Ring
closure to afford quinazolinediones 32a–d was carried out using triphosgene, and subsequent
alkylation at N1 afforded compounds 33–36. In some cases, Mitsunobu reaction using
diisopropylazodicarboxylate (DIAD) resulted in an inseparable mixture of product and reduced DIAD
by-product. Fortunately, this problem was surmounted by using di-(4-chlorobenzyl)azodicarboxylate
3
0
(
DCAD), as the hydrazine by-product in this case is easily removed by trituration. The approach to
ACS Paragon Plus Environment

Page 7 of 88
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
3
-(5-methyl-1,3,4-thiadiazol-2-yl)methyl quinazolinediones 41a and 41e was slightly different
(Scheme 7). The 3-substituent bearing the thiadiazole was introduced prior to the chlorosulfonylation
reaction as there are no free sites on the five-membered heteroaryl ring that can undergo
chlorosulfonylation in this case.
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Scheme 5: Synthesis of 27a–ata
O
O
O
N
a
b
ClO S
2
OH
NH
NH
N
H
N
O
O
23
24
25
O
O
O
O
N
O
O
c
d, e or f
S
S
R
O
N
NH
N
N
H
H
N
O
26
27a-au
NH HCl
2.
28
a
Reagents and conditions: (a) NaOCN, NaOH, AcOH, H O, 50 °C, 1 h, 79%; (b) ClSO H, 50 °C, 16
2
3
h, quant.; (c) amine 28, Et N, DCM, RT, 3 h, 93%; (d) (i) NaH, DMF, RT, 5 min, (ii) R-X (X= Br, Cl,
3
I, OTs), RT, 6–59%; (e) R-OH, DIAD (or DCAD), PS-PPh , DMF, RT, 10–21%; (f) R-B(OH) ,
3
2
Cu(OAc) , Et N, 4Å molecular sieves, DCM, RT, 18%.
2
3
Scheme 6: Synthesis of 33–36a
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 8 of 88
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
O
O
O
O
O
a
ClO S
2
b
S
R2
O
O
N
N
H
H
N
H
O
N
H
O
NH2
2
9
30
3
1a R2=
S
N
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
31 R2=
N
N
3
1c R2=
1d R2=
O
N
N
3
O
O
O
O
O
O
S
R2
O
S
R2
N
N
c
d or e
N
H
N
H
N
N
H
O
R1
3
3a-f, 34b,c,e,f, 35a-e, 36b,c,e
32a-d
a
Reagents and conditions: (a) ClSO H, 50 °C, 4 h, 79%; (b) (i) amine 28, Et N, DCM, −10 °C, 30
3
3
min, (ii) R -NH , Et N, −10 °C–RT, 18 h, 20–67%; (c) triphosgene, THF, 0 °C–RT, 18 h 78–97%; (d)
2
2
3
(
i) NaH, DMF, RT, 5 min, (ii) R -X (X= Br, I, OMs, OTs), RT; (e) R -OH, DIAD (or DCAD), PS-
1
1
PPh , DMF, RT, 16–75%.
3
Scheme 7: Synthesis of 41a and 41ea
ACS Paragon Plus Environment

Page 9 of 88
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
O
O
O
a
S
b
O
N
S
N
H
N
N
N
N
N
H
O
NH2
N
H
O
2
9
37
38
O
O
O
O
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
O
O
c
d
S
S
S
S
N
N
Cl
N
H
N
N
N
N
N
O
N
H
O
H
39
4
0
O
O
O
S
S
e
N
N
S
H
H N
2
N
N
N
O
N
N
N
42
R1
4
1a R = Et
1
41e R1=
N
a
Reagents and conditions: (a) amine 42, Et N, DMF, 60 °C for 5 h, then RT for 16 h, 65%; (b)
3
triphosgene, THF, 0 °C–RT, 18 h, 58%; (c) ClSO H, 50 °C, 2 h, 69%; (d) amine 28, Et N, DCM, 0
3
3
°C–RT, 18 h, 32%; (e) EtI, K CO , DMF, µW, 80 °C, 20 min, 20% for 41a; and 5-(chloromethyl)-
2
3
1,3-dimethyl-1H-pyrazole, K CO , DMF, µW, 80 °C, 20 min, 44% for 41e.
2
3
RESULTS AND DISCUSSION
Crystallographic binding mode and initial optimization of anthraquinone 8a
The low druggability of PARG was highlighted by a high-throughput screen at AstraZeneca,
comprising 1.4 million compounds, which resulted in only three confirmed hits of related chemical
structure, exemplified by the anthraquinone derivative 8a (Figure 3). Despite this compound’s modest
inhibitory potency (EC50 = 81 µM) in an HTRF biochemical assay, we were encouraged by its
approximately equipotent activity in an acute (1 h) cell-based proof-of-mechanism PAR chain
persistence assay,21,31 where the compound was shown to block the PARG-mediated degradation of
PAR chains in a dose-dependent manner. However, the limited solubility of the compound precluded
an accurate estimation of EC , and further progression of this anthraquinone derivative was
5
0
additionally hampered by the observation of non-specific 72 h cytotoxicity at cell-active doses.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 10 of 88
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
O
O
O O
S
N
H
8a
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 3. Anthraquinone HTS hit 8a.
Determination of the co-crystal structure of 8a with the catalytic domain of human PARG at 1.9 Å
resolution confirmed the direct interaction of the inhibitor with the PAR binding site, and revealed
some unexpected binding contacts that prompted our efforts to pursue structure-guided optimization
of this series. As was previously reported, the substrate binding site in human PARG comprises an
extended, relatively solvent-exposed channel.15, 16 The anthraquinone 8a is observed to bind in the
region occupied by the adenine moiety of the truncated PAR analogue, ADP-ribose (Figure 4).
However, the plane of the anthraquinone ring system is oriented approximately orthogonal to that of
the adenine moiety of ADP-ribose, with the result that 8a is highly solvent-exposed compared to the
relatively well-buried adenine. The significant difference in the shape of the binding site between the
two crystal structures can be attributed to Phe902 adopting a different rotamer when bound to 8a
compared with ADP-ribose (Figure 4C). As a consequence, the anthraquinone core of 8a is able to
form extensive face-to-face aromatic stacking interactions with both Phe902 and Tyr795. Although
this rotamer of Phe902 is incompatible with substrate binding, in the published apo structure of
human PARG (PDB 4A0D) both rotamers of Phe902 are observed, highlighting the flexibility of this
side chain.16
ACS Paragon Plus Environment

Page 11 of 88
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 4. Comparison of the X-ray crystal structures of the catalytic domain of (A) human PARG
bound to 8a (PDB accession code 6HMM) and (B) ADP-ribose (PDB 4B1H), shown in the same
coordinate frame. Protein solvent-accessible surfaces are coloured by electrostatic potential. (C)
Detail of relative binding orientations of 8a (cyan carbon atoms) and ADP-ribose (green) highlights
movement of Phe902. (D) 8a and ADP Ribose bound to entire PARG domain, colored by atom type.
The binding mode of 8a is anchored by a network of three hydrogen bonds from the sulfonamide
moiety to the sidechain of Glu727 and the backbones of Ile726 and Gln754, with the tert-butyl group
deeply buried in a mainly hydrophobic cavity (Figure 5). We note that in the crystal structure of
PARG in complex with ADP-ribose, the adenine moiety forms hydrogen bonds to the Glu727
sidechain and Ile726 backbone NH (as well as, several water-mediated contacts); hence, this is a rare
example of a sulfonamide moiety mimicking the hydrogen-bond network of adenine.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 12 of 88
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 5. (A) Detail of the hydrogen-bonding environment around 8a (PDB accession code 6HMM).
(B) Protein solvent-accessible surface highlights the exposure to solvent of the anthraquinone core
and the buried pocket around the tert-butyl moiety. Hydrogen atoms modelled using Maestro
(Schrödinger LLC). (C) Protein-ligand interaction diagram with hydrogen bonds indicated by pink
arrows and aromatic interactions by green arrows.
Optimization of 8a focused on a series of hydrophobic and polar replacements of the tert-butyl moiety
to explore the limits of the constrained binding pocket observed around the N-alkyl sulfonamide head-
group (Scheme 1, Table 1). A preference was observed for compact branched or cyclic alkyl groups,
notably for the simple cyclopropyl 8f, and particularly the methylcyclopropyl and cyanocyclopropyl
derivatives (8h and 8i), which yielded single-digit micromolar biochemical activites. A series of
unbranched polar-substituted groups (e.g., 8jn) were designed to probe a narrow solvent-filled
channel extending from the tert-butyl pocket, but these all proved ineffective. N-Methylation of the
sulfonamide, or replacement with alternative linkers such as amides and the reversed sulfonamide
abolished activity (data not shown), emphasizing a strict requirement for the specific network of
hydrogen-bonding contacts observed in the crystal structure. In summary, the steep SAR highlighted
the limited opportunities for optimization of the N-alkyl sulfonamide moiety. Although an
encouraging improvement in activity had been achieved for 8i and 8h compared with 8a, cell toxicity
remained an issue with this series and was presumed to be related to the anthraquinone ring system, a
motif reminiscent of known DNA intercalators.32
ACS Paragon Plus Environment

Page 13 of 88
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
a
Table 1. Initial optimization of the sulfonamide N-alkyl moiety of 8a
O
O O
S
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
R
N
H
O
PARG
Biochemical EC50
µM
PARG
Cell EC50
/µM
~30
Compound
R
/
8
8
8
8
a
b
c
tert-butyl
81 (21)
>30b
Methyl
nd
Ethyl
110 (7.1)
49 (27)
24 (5.2)
13 (5.3)
>50b
nd
d
1,1-dimethylpropyl
Cyclopropylmethyl
Cyclopropyl
nd
8e
8f
nd
22 (6.8)
nd
8
8
8
8
8
g
h
i
Cyclobutyl
1-methylcyclopropyl
1-cyanocyclopropyl
Carbamoylmethyl
2-carbamoylethyl
2-hydroxyethyl
2-methoxyethyl
6.0 (2.3)
3.3 (2.3)
>150
6.6 (3.5)
19 (5.2)
nd
j
k
>150
nd
8l
>150
nd
8m
>150
nd
8
8
n
o
2-methoxy-1,1-dimethyl-ethyl >150
2-cyanoethyl 88 (18)
nd
nd
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 14 of 88
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
aBiological data are stated as the geometric mean of at least three independent determinations unless
otherwise stated, with standard deviations quoted in parentheses. nd = not determined.
bMaximum concentration limited by interference and/or solubility issues.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Scaffold-hopping by virtual screening and library design
In parallel with elaboration of the anthraquinone sulfonamide moiety, a structure-based virtual
screening campaign was initiated to identify alternatives to the anthraquinone ring system that might
obviate the observed cell toxicity and prove more amenable to optimization. As the hydrogen-bonding
network observed around the sulfonamide appeared to be a key pharmacophore, we decided to retain
this moiety while exploring alternative bi- and tricyclic ring systems that would maintain the observed
stacking interactions with Tyr795 and Phe902, and introduce more drug-like physical properties.
To this end, commercially available screening libraries were initially triaged to identify and select all
aryl sulphonamides bearing a variety of aliphatic N-substituents. These selections were then filtered
for suitable drug-like properties and any PAINS or undesirable structural motifs removed. These
approaches yielded around thirty one thousand putative hits which were then docked into the induced
cavity observed in the X-ray structure of 8a bound to human PARG (Glide SP docking, Schrodinger,
LLC, New York). Manual inspection of the top 5 docking poses per ligand ensured maintenance of
the key sulphonamide binding motif and credible binding modes and from this manual analysis, a set
of 68 compounds was selected for purchase and assessment, based on an assessment of binding mode,
site complementarity and docking score.
The biochemical affinity of these initial compounds proved to be very weak, with the notable
exception of the quinazolinedione 12e (Table 2). Replacing the sulfonamide N-cyclopropyl
substituent of 12e with N-ethyl (12d, 13c) resulted in a large drop in activity, consistent with the SAR
observed for the anthraquinone scaffold (Table 1). The only other N-cyclopropyl example in Table 2,
the tricyclic scaffold 43, displayed modest activity but with significantly reduced ligand efficiency
ACS Paragon Plus Environment

Page 15 of 88
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
compared with 12e. While some examples with the N-tert-butyl moiety, such as 44 and 45, were of
similar activity to the anthraquinone 8a, other bicyclic systems proved to be inactive (16). Table 2.a
Example hits from the initial virtual screen of commercial compounds.
PARG
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Compound Structure
Biochemical
EC /µM
5
0
O
O O
S
N
N
12d
>150
H
N
O
O
O
O O
S
N
N
N
1
1
2e
3c
11 (4.0)
H
N
O
O O
S
N
H
140 (30)
N
O
O O
S
16
43
N
>150
H
O O
S
N
50 (9.8)
H
N
H
O
O
O O
S
N
4
4
110 (26)
H
N
H
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 16 of 88
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
O O
S
N
4
5
57 (25)
H
N
O
aBiological data are stated as the geometric mean of at least three independent determinations, with
standard deviations quoted in parentheses.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
To determine if the weak level of activity for these compounds was a consequence of sub-optimal N-
alkyl substitution on the sulfonamide, analogues of selected cores were synthesized with the preferred
N-alkyl substituents identified from the optimization of 8a, e.g., methylcyclopropyl and
cyanocyclopropyl (Scheme 2 and Scheme 3, Table 3). This modification resulted in a significant gain
in biochemical activity for the quinazolinediones 12b, 12c, 13a and 13b, which demonstrated a
greater than 20-fold improvement compared with the tert-butyl analogue 12a, and were accompanied
33
by an increase in biochemical lipophilic ligand efficiency (LLE) from a value of 2.1 for 12a to 3.9
and 4.7 for 12b and 12c, respectively. Notably, the quinazolinediones also demonstrated a similar
level of potency in the cell assay, with the best examples achieving low micromolar potency.
However, N-methylcyclopropyl derivatives of other ring systems such as the naphthalene 16 did not
achieve levels of activity to justify further consideration. From docking studies, the quinazolinedione
scaffold hop was predicted to achieve very similar binding modes to the anthraquinone 8a, while
offering different vectors for enhancing binding affinity through additional protein interactions
(
Figure 6).
a
Table 3. Optimization of sulfonamide N-alkyl groups for hits from the first round of virtual
screening.
PARG
PARG Cell
Compound Structure
Biochemical EC /µM
5
0
ACS Paragon Plus Environment

Page 17 of 88
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
EC /µM
5
0
O
N
O O
S
N
N
1
1
2a
2b
110 (30)
3.9 (1.7)
5.0 (1.6)
nd
H
O
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
O
O O
S
N
N
1.8 (1.1)
14 (3.8)
H
N
O
O
O O
S
N
N
N
12c
H
N
N
O
O
O
O O
S
N
1
3a
H
1.5 (0.56)
1.1 (0.53)
N
O
N
O O
S
13b
N
N
H
1.8 (0.50)
>150
5.6 (3.0)
N
O
O O
S
1
6
N
nd
H
aBiological data are stated as the geometric mean of at least three independent determinations, with
standard deviations quoted in parentheses. nd = not determined.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 18 of 88
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
O
O O
S
N
N3
H
N1
O
Figure 6. Predicted binding modes of quinazolinedione 12b (docked into PDB accession code
6
HMM) The N1 position presents a vector towards solvent and the shelf adjacent to Phe902, while the
N3 position presents a vector toward the pocket occupied by the ribose moiety of ADP.
The requirement for specific N-alkyl sulfonamide substituents, such as tert-butyl and cyclopropyl, in
order to achieve significant levels of potency suggested that potentially useful scaffold hops may have
been missed in the original virtual screening campaign because of the limited commercial availability
of appropriate sulfonamides. Therefore, a second round of virtual screening was initiated based on
docking virtual libraries generated from the in silico enumeration of a diverse set of commercially
available aryl sulfonyl chlorides with several of our preferred alkyl amines. A further set of scaffold
hops was designed on the basis of 2D similarity to the active cores identified previously. Synthesis of
selected examples from these approaches resulted in a chemically diverse set of additional scaffolds,
several of which displayed promising levels of activity, though none proved to be superior in potency
or lipophilic ligand efficiency to the quinazolinediones (Supporting Information Table S2). Note that
for synthetic expediency, no attempt was made to optimize fully the substituents around each distinct
ring system; rather, the aim of the exercise was to perform a rapid exploration of synthetically
ACS Paragon Plus Environment

Page 19 of 88
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
tractable SAR to understand whether any particular ring systems were significantly favored or
disfavored. In general, we observed that fully aromatic/delocalized bicyclic cores were preferred over
partially saturated ring systems, consistent with the modelled binding modes in which the aromatic
core is tightly stacked between Tyr795 and Phe902. Among aromatic systems, there appears to be a
preference for one or more hydrogen bond acceptor groups such as a carbonyl oxygen (as in the
quinazolinedione 12b) or a nitrogen lone pair (as in the quinoline S12, Supporting Information Table
S2). In contrast, incorporation of a nitrogen atom in the benzo ring bearing the sulfonamide resulted
in between 3 and 5-fold loss of activity (e.g., compound S1, Supporting Information Table S2). We
note that, in the X-ray crystal structure of PARG in complex with 8a, a number of polar protein atoms
are located in close proximity to the anthraquinone ring system, including the potential hydrogen bond
donor Tyr795 OH, and it may be that a combination of hydrogen bonding (perhaps through water) and
enhanced π-stacking interactions account for the preferences seen for certain ring systems over others.
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Overall, by utilizing this approach, we succeeded in moving from an unattractive HTS hit to several
lead-like scaffolds displaying improved ligand efficiency and physicochemical properties.
Comparison of the initial hit compound 8a with 12b reveals a reduction in molecular weight and
LogP, with associated increases in LE and LLE (Table 4). In addition, the scaffold hop is devoid of
the acute non-specific cytotoxicity of the parent anthraquinone. On the basis of its promising potency,
synthetic tractability and favorable physicochemical profile, the quinazolinedione series was
prioritized as an attractive starting-point for further optimization.
Table 4. Comparison of measured and predicted physicochemical properties of representative
compounds.
Measured
LogD7.4
Calculated
Turbidimetric
Compound
MW
LEb
LLEc
Cytotoxicity
µM)
LogPa
Solubility (µM)
(
8
1
a
343.4 Nd
3.18
1.50
20
0.23 0.91
0.34 3.9
16 (2)
>30
2b
323.4 1.73
>100
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 20 of 88
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
a
b
XlogP calculated using Dotmatics Elemental package (Dotmatics, Bishops Stortford, UK). Ligand
Efficiency calculated from biochemical EC (i.e., LE = 1.4(pEC )/HAC, where HAC is the number
5
0
50
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
c
of heavy atoms). Ref 33. nd = not determined.
Optimization of quinazolinedione N1 position
Initial exploration of the SAR around the quinazolinedione ring system focused on substitution at the
N1 position (Scheme 4), as this vector presented an opportunity for interaction with a solvent-exposed
shelf adjacent to Phe902 (Figure 6). Methylene-linked saturated and aromatic ring systems were of
interest as a strategy to improve binding affinity, as modelling suggested these would readily access
the Phe902 shelf; while more conformationally flexible groups were deemed likely to extend into
solvent and thus, would be more useful to modulate physicochemical properties.
In general, substitution at N1 yielded relatively modest gains in biochemical and cellular EC (Table
50,
5) with the ethyl analogue 22a similar in potency to the parent methyl 12b. Incorporation of more
polar groups, such as in 22b and 22c, resulted in low micromolar compounds with improved LLE. We
were pleased to observe that simple lipophilic groups such as the prop-2-ynyl 22d and
cyclopropylmethyl 22f yielded our first sub-micromolar compounds in both enzyme and cell assays.
This level of cellular potency represents a significant advance in the identification of cell-active
PARG tool compounds when compared with our original HTS hit and also other inhibitors described
in the literature.
In comparison with 22f, larger carbocyclic substituents (e.g., 22g and 22h) did not result in any
further gain in potency. The predicted binding mode of the benzylic derivative 22h was confirmed by
soaking the compound into pre-formed crystals of human PARG, resulting in a 2.9 Å resolution X-ray
crystal structure (Figure 7). The compound was observed to bind similarly to 8a, with the
quinazolinedione ring system well aligned with the anthraquinone of 8a and forming aromatic
stacking interactions with Tyr795 and Phe902. The sulfonamide moiety forms the same network of
three hydrogen bonds as observed for 8a, with the N-methylcyclopropyl group occupying a similar
ACS Paragon Plus Environment

Page 21 of 88
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
location to the tert-butyl of 8a. As these groups are very similar in size, the significant improvement
in activity observed for the methylcyclopropyl may be due to the greater π-character of the
cyclopropyl ring:34 we note close non-bonded contacts between the cyclopropyl ring and the
.
..
sidechains of Gln754 and Glu727, with a potential CH O hydrogen bond to the backbone carbonyl of
Gln754. The N1 benzyl group partly occupies the Phe902 shelf, stacking against the Phe902–Gly903
peptide bond. The interaction between the benzyl group and the Phe902 shelf appears sub-optimal in
terms of shape complementarity, which likely accounts for the modest potency of 22h and suggested
opportunities for further optimization. Pleasingly, the determined crystallographic structure of this
derivative closely matched the binding mode predicted by modelling, facilitating the prioritization of
latter derivatives (such as the improved analogue 22f) for synthesis and precluding the requirement
for extensive iterative crystallographic determination.
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 7. X-ray crystal structure of 22h bound to the catalytic domain of human PARG at 2.9 Å
resolution (PDB accession code 6HMN) (A) Interaction of the N1-benzyl moiety with the Phe902
shelf. (B) Detail of the hydrogen-bonding environment around the sulfonamide, with a putative
…
CH O interaction highlighted by the cyan dashed line. Hydrogen atoms modelled using Maestro
(Schrödinger LLC). (C) Ligand interaction diagram coloured as in Figure 5.
a
Table 5. Evaluation of quinazolinedione N1 substituents.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 22 of 88
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
O
O O
S
N
N3
H
N1
R
O
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Biochemical
EC /µM
Compound
R
Cell EC /µM
50
5
0
2
1
2
2
0
H
4.0 (2.0)
3.9 (1.7)
1.3 (1.9)
1.1 (0.13)
8.1 (9.2)
2b
2a
2b
methyl
ethyl
1.8 (1.1)
3.6 (1.3)
6.1 (3.1)
6.4 (5.3)
2-methoxyethyl
22c
22d
2-(dimethylamino)ethyl 2.0 (0.77)
prop-2-ynyl
0.71 (0.17)
2.8 (1.6)
0.25 (0.088)
8.9 (4.5)
2
2
2
2
2e
2f
cyanomethyl
cyclopropylmethyl
cyclohexylmethyl
benzyl
0.44 (0.18)
0.56 (0.28)
1.3 (0.62)
0.62 (0.62)
2.7 (1.6)
2g
2h
2.3 (0.97)
aBiological data are stated as the geometric mean of at least three independent determinations, with
standard deviations quoted in parentheses. nd = not determined.
We determined to further investigate structure-activity relationships (SAR) at the N1 position of the
quinazolinedione, as this vector presented a good opportunity for interaction with a solvent-exposed
shelf adjacent to Phe902. It can be seen from the crystal structure of 22h in complex with PARG that
the N1 vector projects away from the PARG active site towards solvent (Figure 7). A library of more
elaborate N1-quinazolinediones (with N3 fixed as methyl) was therefore synthesized to explore
whether the interactions with this shelf could be optimized in order to boost affinity. Further N1
variations are also shown in Table 6.
a
Table 6. Evaluation of quinazolinedione N1 substituents.
ACS Paragon Plus Environment

Page 23 of 88
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
O
O O
S
N
N3
H
N1
R
O
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Biochemical
EC /µM
Compound
R
5
0
2
2
2
2i
2-F-Bn
1.4 (1.0)
2j
2k
3-F-Bn
1.6 (0.80)
0.53 (0.40)
3.2 (0.17)
0.86 (0.77)
0.45 (0.13)
5.8 (1.8)
4-F-Bn
22l
2-OMe-Bn
3-OMe-Bn
4-OMe-Bn
2-Pyridylmethyl
3-Pyridylmethyl
4-Pyridylmethyl
22m
2
2
2
2
2n
2o
2p
2q
1.3 (0.26)
2.4 (0.85)
2
2
2r
2s
1.2 (0.37)
N
N
0.38 (0.19)
0.18 (0.058)
N
N
S
N
22t
aBiological data are stated as the geometric mean of at least three independent determinations, unless
otherwise stated, with standard deviations quoted in parentheses.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 24 of 88
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
The results in Table 6 show that substitution at the 4-position of the phenyl ring of the benzyl
derivatives was usually favored over the 2- or 3-substituted analogues. Although the SAR was
generally quite flat, only modest improvements in potency were observed compared to the parent N1-
benzyl analogue 22h. No improvement was observed with pyridylmethyl derivatives 22o-q. However,
significant improvements were obtained with methylene-linked 5-membered heteroaryl substituents.
Pyrazole derivative 22r showed low micromolar activity, and this was enhanced further by
methylation at the 3-positon (22s). Bis-methylated thiazole derivative 22t gave the most potent
quinazolinedione in the series. Modelling suggested that these substituted 5-membered heterocycles
achieved improved lipophilic contacts with the Phe902 shelf compared with benzyl derivative 22h.
Although some polar protein sites are accessible in this region, it was not apparent from the SAR that
any productive hydrogen-bonding interactions were being achieved.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Exploring the N3 vector
We then turned our attention to exploring the SAR at the N3 vector of the quinazolinedione. Although
this vector also points primarily towards solvent, its proximity to the ribose binding pocket suggested
that additional productive binding interactions may be achievable. A library with a range of
substituents at N3 (with N1 fixed as methyl) was modelled into the crystal structure of PARG (PDB
accession code 6HMN), and a selection of analogues was prepared in order to investigate what is
tolerated at this position. A representative set of N3 variations is shown in Table 7.
Table 7. Elaboration at the N3 vector
O
O
O
S
R
O
N
N3
H
N1
PARG Biochemical
Compound
R
H
a
EC (μM)
5
0
2
6
5.2 (1.2)
ACS Paragon Plus Environment

Page 25 of 88
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
1
2
2
2
2b
7a
7b
7c
Me
Et
3.9 (1.7)
4.0 (1.3)
Cyclopropylmethyl 4.6 (0.27)
Cyclohexylmethyl
Ph
>150
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
27d
27e
23 (4.5)
26 (2.8)
16 (3.4)
2.5 (1.0)
8.0 (0.64)
Bn
2
2
2
7f
2-Pyridylmethyl
3-Pyridylmethyl
4-Pyridylmethyl
7g
7h
Ph
O
2
2
7i
7j
1.6 (0.12)
1.7 (0.18)
N
O
27k
7l
Propargyl
Cyanomethyl
3.0 (2.0)
2
1.1 (0.81)
27m
7.0 (4.6)
6.0 (1.5)
20 (0)
O
2
2
2
7n
7o
7p
O
S
N
S
10 (0.78)
21 (1.2)
13 (4.0)
N
O
2
7q
S
N
27r
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 26 of 88
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
S
2
2
7s
7t
1.2 (0.60)
N
S
0.48 (0.36)
N
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
aAll EC50 data are means of at least n = 4 independent measurements, with standard deviations quoted in
parentheses.
We initially investigated lipophilic alkyl and aryl groups, and found that these gave either no
improvement over the 3-methyl analogue 12b, or were detrimental to activity, consistent with the
generally polar nature of the protein surface in this region. However, introducing polarity to the
substituent at the N3-position yielded promising compounds with activity in the low micromolar
range (e.g., 27g, 27i, 27j and 27l). Modelling studies suggested that the hydrogen bond acceptor atom
common to these groups may form a hydrogen bond with Asn869. The difference in activity observed
across the pyridyl isomers suggested that the position of this hydrogen bond acceptor was important,
and this was further demonstrated using a series of methylene-linked heterocyclic substituents at N3.
Initial exploration with this motif resulted in moderate potencies. However, moving from thiazol-2-
ylmethyl derivative 27r (13 μM) to thiazol-5-yl isomer 27s resulted in a 10-fold improvement in
potency to 1.2 μM. Methylation of the thiazole (27t) gave a further boost to potency to yield a sub-
micromolar inhibitor.
Determination of the X-ray crystal structure of 27t bound to the human PARG catalytic domain
revealed that the quinazolinedione sulfonamide exhibited a similar binding mode to that observed for
the N1-benzyl derivative 22h (Figure 8). The N3 substituent of 27t accesses a narrow solvent-exposed
ledge above Phe902, potentially making a favorable edge-to-face stacking interaction with the Phe902
sidechain. The crystal structure confirmed the hypothesis that the N atom in the thiazole ring can
make a hydrogen bond interaction with the Asn869 sidechain, as had been suggested by modelling.
This explains the noticeable gain in potency for this specific thiazole isomer compared to other
heterocycles lacking an appropriately positioned hydrogen bond acceptor (e.g., 27q). The reduced
affinity noted for furan 27n and oxazole 27q is in line with the weaker hydrogen bonding strength
ACS Paragon Plus Environment

Page 27 of 88
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
expected from an aromatic oxygen atom compared with a nitrogen lone pair. The 2-methyl group on
the thiazole ring of 27t makes an additional hydrophobic contact with the β-carbon of the Asp904
sidechain. Overall, this data suggests that the complementarity in shape and properties of the 2-
methylthiazole substituent enables productive binding to an apparently unpromising, narrow shelf on
the protein surface. The relatively limited conformational flexibility around the N3 methylene, due to
the steric restrictions imposed by the adjacent carbonyl oxygens of the quinazolinedione core, may
also help to stabilize the observed binding mode.
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
A
Figure 8. (A) Crystal structure of 27t bound to human PARG at 2.06 Å resolution (PDB accession
code 6HMK). Hydrogen atoms modelled using Maestro (Schrödinger LLC); (B) Protein-ligand
interaction diagram with hydrogen bonds indicated by pink arrows and aromatic interactions by green
arrows.
The data in Table 7 indicated that the SAR at the N3 vector of the quinazolinedione was steeper than
at the N1 vector. Changing the position of the nitrogen atom in the heteroaryl ring evidently had a
significant effect on the PARG activity, and we were keen to explore the interaction of the nitrogen
with Asn869 further. A focused set of quinazolinediones substituted at the N3 position with
methylene-linked 5-membered N-containing heterocycles was therefore synthesized (Table 8).
Table 8. Exploring the interaction with Asn869: focused N3 library
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 28 of 88
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
O
O
O
S
R
O
N
N3
H
N1
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
PARG Biochemical
Compound
27t
R
a
EC (μM)
50
S
0.48 (0.36)
0.48 (0.17)
0.91 (0.49)
N
S
27u
NH2
N
S
27v
N
S
2
7w
12 (7.0)
N
Ph
2
2
7x
7y
2.3 (1.2)
S
N
N
N
S
2.7 (0.65)
27z
0.88 (0.27)
0.47 (0.12)
4.0 (0.43)
N
S
N
N
S
2
2
7aa
N
H
N
7ab
7ac
N
N
N
2
3.3 (0.85)
ACS Paragon Plus Environment

Page 29 of 88
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
N
2
2
7ad
7ae
3.1 (1.2)
N
N
N
3.0 (0.75)
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
N
N
HN
N
N
N
N
2
2
7af
3.2 (0.97)
4.1 (1.1)
7ag
N
2
2
7ah
7ai
N
12 (0.29)
N
0.39 (0.28)
0.38 (0.20)
0.45 (0.19)
N
NH
27aj
N
N
N
2
2
7ak
7al
N
0.37 (0.027)
N
N
2
2
2
7am
7an
7ao
1.2 (0.77)
0.47 (0.17)
5.2 (0.91)
O
O
N
N
N
O
2
2
7ap
7aq
0.77 (0.24)
0.62 (0.30)
O
N
O
N
ACS Paragon Plus Environment