Synthesis of 2,3-dihydrobenzo[b][1,4]dioxine-5-carboxamide and 3-oxo-3,4-dihydrobenzo[b][1,4]oxazine-8-carboxamide derivatives as PARP1 inhibitors

Article abstract of DOI:10.1016/j.bioorg.2020.104075

Poly(ADP-ribose) polymerase 1 (PARP1), a widely explored anticancer drug target, plays an important role in single-strand DNA break repair processes. High-throughput virtual screening (HTVS) of a Maybridge small molecule library using the PARP1-benzimidazole-4-carboxamide co-crystal structure and pharmacophore model led to the identification of eleven compounds. These compounds were evaluated using recombinant PARP1 enzyme assay that resulted in the acquisition of three PARP1 inhibitors: 3 (IC₅₀ = 12 μM), 4 (IC₅₀ = 5.8 μM), and 10 (IC₅₀ = 0.88 μM). Compound 4 (2,3-dihydro-1,4-benzodioxine-5-carboxamide) was selected as a lead and was subjected to further chemical modifications, involving analogue synthesis and scaffold hopping. These efforts led to the identification of (Z)-2-(4-hydroxybenzylidene)-3-oxo-3,4-dihydro-2H-benzo[b][1,4]oxazine-8-carboxamide (49, IC₅₀ = 0.082 μM) as the most potent inhibitor of PARP1 from the series.

Full text of DOI:10.1016/j.bioorg.2020.104075

Journal Pre-proofs
Synthesis of 2,3-dihydrobenzo[b][1,4]dioxine-5-carboxamide and 3-oxo-3,4-
dihydrobenzo[b][1,4]oxazine-8-carboxamide derivatives as PARP1 inhibitors
Xuwei Shao, Steven Pak, Uday Kiran Velagapudi, Shruthi Gobbooru, Sai
Shilpa Kommaraju, Woon-Kai Low, Gopal Subramaniam, Sanjai Kumar
Pathak, Tanaji T. Talele
PII:
S0045-2068(20)31372-9
DOI:
https://doi.org/10.1016/j.bioorg.2020.104075
YBIOO 104075
Reference:
To appear in:
Bioorganic Chemistry
Received Date:
Revised Date:
Accepted Date:
17 January 2020
29 June 2020
3 July 2020
Please cite this article as: X. Shao, S. Pak, U. Kiran Velagapudi, S. Gobbooru, S. Shilpa Kommaraju, W-K. Low,
G. Subramaniam, S. Kumar Pathak, T.T. Talele, Synthesis of 2,3-dihydrobenzo[b][1,4]dioxine-5-carboxamide
and 3-oxo-3,4-dihydrobenzo[b][1,4]oxazine-8-carboxamide derivatives as PARP1 inhibitors, Bioorganic
Chemistry (2020), doi: https://doi.org/10.1016/j.bioorg.2020.104075
This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover
page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version
will undergo additional copyediting, typesetting and review before it is published in its final form, but we are
providing this version to give early visibility of the article. Please note that, during the production process, errors
may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
©
2020 Published by Elsevier Inc.

Synthesis of 2,3-dihydrobenzo[b][1,4]dioxine-5-carboxamide
and 3-oxo-3,4-dihydrobenzo[b][1,4]oxazine-8-carboxamide
derivatives as PARP1 inhibitors
a,1
a,1
a,1,2
a
Xuwei Shao , Steven Pak , Uday Kiran Velagapudi , Shruthi Gobbooru , Sai Shilpa
a
a
b
b,c,*
Kommaraju , Woon-Kai Low , Gopal Subramaniam , Sanjai Kumar Pathak , Tanaji T.
a,*
Talele
aDepartment of Pharmaceutical Sciences, College of Pharmacy and Health Sciences, St. John's
University, 8000 Utopia Parkway, Queens, NY 11439, USA
bChemistry and Biochemistry Department, Queens College of the City University of New York,
6
5-30 Kissena Blvd., Flushing, NY 11367, USA;
cChemistry Doctoral Program, Biochemistry Doctoral Program, The Graduate Center of the
City University of New York, 365 5th Ave, New York, NY 10016, USA
*
7
Corresponding authors: E-mail address: talelet@stjohns.edu, Tel.: +1 718 990 5405; Fax: +1
18 990 1877; Sanjai.Kumar@qc.cuny.edu, Tel.: +1 718 997 4120, Fax: +1 718 997 5531
1Authors contributed equally to the work.
2Present Address: PACE Analytical Life Sciences, LLC, Suite 102, 19 Presidential Way,
Woburn, MA 01801 (U.K.V.).
Abstract: Poly(ADP-ribose)polymerase 1 (PARP1), a widely explored anticancer drug target,
plays an important role in single-strand DNA break repair processes. High-throughput virtual
screening (HTVS) of the Maybridge small molecule library using the PARP1-benzimidazole-4-
carboxamide crystal structure and pharmacophore model led to the identification of eleven
compounds. These compounds were evaluated using recombinant PARP1 enzyme assay that
resulted in the acquisition of three PARP1 inhibitors: 3 (IC = 12 µM), 4 (IC = 5.8 µM), and
5
0
50
1
0 (IC = 0.88 µM). Compound 4 (2,3-dihydro-1,4-benzodioxine-5-carboxamide) was selected
50
as a lead and it was subjected to further chemical modifications, involving analogue synthesis
and scaffold hopping. These efforts led to the identification of (Z)-2-(4-hydroxybenzylidene)-3-
1

oxo-3,4-dihydro-2H-benzo[b][1,4]oxazine-8-carboxamide (49, IC50 = 0.082 µM) as the most
potent inhibitor of PARP1 from the series.
Keywords: virtual screening, 1,4-benzodioxine, 1,4-benzoxazin-3-one, PARP1, Knoevenagel
condensation
Introduction
6
Poly(ADP-ribose) polymerase 1 (PARP1) is the most abundant (~10 molecules/cell)
member of the PARP family of proteins. It is a chromatin-bound nuclear protein known to
facilitate DNA repair processes by catalyzing the poly(ADP)ribosylation (PARylation) of itself
(
automodification) or on an array of DNA damage repair-associated target proteins
heteromodification) such as topoisomerase I/II, p53, histones, DNA polymerases and DNA
(
ligases [1]. Activation of PARP1 plays a critical role in mediating DNA single-strand break
repair in cancer cells, thereby providing them survival advantages. PARP1 inhibitors are
therefore being considered clinically as potentiators of chemo- and radio-therapy through
suppression of the base excision repair (BER) pathway [2,3]. In addition, some of the FDA
approved PARP1 inhibitors (e.g., olaparib, niraparib, rucaparib and talazoparib) exhibited single-
agent cytotoxic effect on tumors with germline mutations in BRCA1- or BRCA2-dependent DNA
double-stranded repair mechanisms (homologous recombination, HR) [4,5]. However, the rapid
emergence of restored HR pathway function has led to the limited use of existing PARP1
inhibitors [6-8]. Thus, development of new classes of PARP1 inhibitors is highly desirable, and
is considered a promising strategy for anticancer drug development.
In our continuing effort to identify novel PARP1 inhibitors [9,10], we conducted a
structure- and pharmacophore-based virtual screening of a Maybridge database comprised of a
2

refined 15,621 small molecule compound library; this initial refinement of the library was
generated by applying six pre-defined filtering parameters (Figure 1) to the original 54,836
compound library. Among the 868 virtual hits identified this way, 11 were selected as potential
inhibitors of PARP1 enzyme by consideration of novelty, diversity, and lead optimization
potential. Further in vitro evaluation of these compounds in PARP1 enzyme assay led to the
identification of three lead compounds 3 (IC = 12 µM), 4 (IC = 5.8 µM), and 10 (IC = 0.88
5
0
50
50
µM). Since compound 3 had the weakest PARP1 inhibition and compound 10 belonged to the
phthalazinone scaffold, already present in clinically used PARP-inhibitors (e.g., olaparib and
talazoparib), we selected 2,3-dihydrobenzo[b][1,4]dioxine-5-carboxamide (compound 4, Figure
1
) as a tractable lead for further SAR optimization. Using analogue synthesis and scaffold
hopping strategies on compound 4, we obtained a modestly potent compound 49, (Z)-2-(4-
hydroxybenzylidene)-3-oxo-3,4-dihydro-2H-benzo[b][1,4]oxazine-8-carboxamide (IC = 0.082
5
0
µM, Figure 1) that exhibited a high potential as a lead compound in future anticancer drug
development studies targeting PARP1 enzyme.
Results and Discussion
Structure-based Virtual Screening
Initially, 54,836 compounds from Maybridge library were subjected to a ligand
preparation protocol using six pre-defined parameters such as inclusion of molecular weight
filter, controlling ionization, exclusion of tautomers, restricting stereoisomers to one per ligand,
number of aromatic rings (1-3) and that all ligands must possess a primary or secondary
carboxamide group as depicted in Figure 1. This generated a refined library of 15,621
compounds. These compounds were subsequently subjected to a high-throughput virtual
screening (HTVS) protocol using the complex of the catalytic domain of human PARP1 enzyme
3

and the benzimidazole-4-carboxamide ligand (RCSB Protein Data Bank (PDB) ID: 2RCW). The
HTVS with Glide v5.5 docking (Schrodinger LLC, New York, NY) involved constraints such as
(
1) applying a -7.0 Glide score cutoff to eliminate ligands which have a higher Glide score and
2) employing a hydrogen bond constraint to eliminate the ligands which do not interact with the
(
backbone of Gly863. These efforts led to narrowing of the potential PARP1 inhibitory leads to
00 candidates for pharmacophore-based screening.
9
Pharmacophore-based Virtual Screening
PHASE algorithm [11] was employed to quantitatively correlate the chemical structures
of selected known PARP1 inhibitors to their biological activity for generating a valid
pharmacophore model. A training set of 23 known PARP1 inhibitors [12] was chosen for this
activity, with their experimental pIC values ranging from 0.568 to 2.853 and seven compounds
5
0
were employed as the test set to validate the accuracy of the pharmacophore models (Table S1,
supporting information). Compounds with pIC values greater than 1.5 were deemed active for
5
0
the development of pharmacophore models. A pharmacophore site point generation protocol
containing the default variant list: acceptor (A), donor (D), hydrophobic (H), negative (N),
positive (P), and aromatic ring (R) along with a diversity of conformers for the training set
molecules was selected to generate pharmacophore hypothesis models. Four top-scored
pharmacophore hypotheses containing a combination of ADR variants were thus generated
(Table S2, supporting information), exhibiting comparable scores for survival, site, vector,
volume, and number of active sites. Visual inspection led to elimination of pharmacophore
models AADRR.1 and ADRRR.5 due to the incorrect orientation of a vector, thereby preventing
the models to form a key hydrogen bonding interaction between carboxamide group of mapped
ligands and carbonyl oxygen of Gly863 (Figure S1, supporting information). The pharmacophore
4

models AADRR.2 and ADRRR.7 were then subjected to atom-based 3D QSAR analysis which
provided comparable results for the two models (Table S3, supporting information). Further
results of overlapping pharmacophore models on an active PARP1 inhibitor revealed that
ADRRR.7 could not identify the key intramolecular hydrogen bonding interaction between
carboxamide group and the heteroatom located on the adjacent ring of the ligand (Figure S1,
supporting information). Therefore, AADRR.2 was selected for further analysis, and the
previously obtained 900 ligands were screened against the AADRR.2 pharmacophore model.
During this process, all ligands were required to contain at least three (A2, D3 and R9) of the
five features of the AADRR.2 model. This step resulted in 868 ligands, from which 11
compounds were selected for empirical testing in biochemical PARP1 enzyme assay based on
their ability to form hydrogen bonding interaction between the carboxamide and Gly863/Ser904,
novelty, diversity, and the potential for further optimization (Figure 1)[13].
5

Figure 1. A schematic representation of the high-throughput virtual screening protocol and
subsequent hit optimization that led to the identification of the most active compound 49.
PARP1 Inhibition Data
Ten of the eleven selected compounds (1–10) were obtained from Maybridge Chemicals
Ltd (Tintagel, United Kingdom). Compound 11 was purchased from Matrix Scientific
(Columbia, SC). These compounds were then tested for in vitro inhibition of PARP1 enzyme
activity as described in our earlier reports [9,14]. A previously characterized PARP1 inhibitor, 8-
amino-2-methylquinazolinone, developed in our laboratory and 3-aminobenzamide,
recommended in the assay kit, were included as internal positive controls [14]. The preliminary
screening was performed at a single concentration of 10 µM that resulted in three inhibitory
classes of compounds: with high, moderate, and low PARP1 inhibitory activity (Table 1). Based
on these results, the three best compounds (3, 4 and 10) were selected for further analysis using
dose-response relationship. The IC values of these inhibitors thus obtained were 12 μM, 5.8 μM
5
0
and 0.88 μM for 3, 4 and 10 respectively (Table 1). Based on the potency and novelty aspects
described earlier, compound 4 was selected for further investigation. To that end, the binding
model of compound 4 at the active site of PARP1 (Figure S2, supporting information) was
generated to facilitate further lead optimization via extension of substituents at the C2- or C3-
positions of the dioxine ring or by employing a scaffold hopping strategy.
Table 1. Structures and PARP1 inhibitory activities of virtual screening hits 1-11
PARP1
Compound
Vendor Number
Structure
a, b
IC (µM)
5
0
6

O
H
N
1
2
S14246
>10
N
O
O
NH
AC40100
O
>
10
NH2
O
H
N
NRB02674
3
4
12 ± 1.6
(R/S)
O
O
NH2
O
O
MO00789
5
.8 ± 0.10
O
O
NH2
OH
JFD03569
O
5
N
~10
O
(R/S)
N
N
N
O
NH
N
6
7
GK03180
GK02796
~10
>10
S
O
NH
N
S
7

O
NH
8
9
GK00357
S
>
>
10
10
N
O
NH
N
N
O
BTB08875
O
H
N
N
1
0
BTB02780
0
.88 ± 0.090
N
H N
2
O
N
NH2
N
N
1
1
003736
>10
CF3
a3-Aminobenzamide sample included in the assay kit was used as an internal standard, and
b
showed a consistent result as reported (60 – 80% inhibition at 50 µM). All compounds were
assayed at a single concentration of 10 µM in triplicates. Compounds with <40% inhibition at 10
µM were denoted as PARP1 IC >10 µM. Compounds showing 40 – 60% inhibition at 10 µM
5
0
were considered to have PARP1 IC ~10 µM. Compounds showing >60% inhibition at 10 µM
5
0
were further evaluated to obtain PARP1 IC .
5
0s
Chemistry
To conduct SAR around compound 4, it was first synthesized according to Scheme 1.
The 2,3-dihydroxybenzoic acid 12 was esterified using concentrated sulfuric acid in refluxing
methanol to obtain 13. Methyl ester 13 was alkylated with 1,2-dibromoethane in the presence of
8

potassium carbonate to afford cyclized ester 14. The ester group in 14 was hydrolyzed using
lithium hydroxide. Resulting acid was obtained through basic and acidic work up and converted
to carboxamide 4 using the mixed-anhydride method. Target compounds with C7-nitro (15) and
C8-nitro (16) were obtained via a nitration reaction involving a mixture of nitric acid and
trifluoroacetic acid. Corresponding amino derivatives 17 and 18 were obtained using catalytic
hydrogenation reaction conditions [15].
COOH
OH
COOCH3
OH
COOCH3
O
CONH2
O
CONH2
O
CONH2
O
a
b
c
d
e
OH
OH
O
O
O
O
R
R
12
13
14
4
15: R = 7-NO2 17: R = 7-NH2
6: R = 8-NO2 18: R = 8-NH2
1
Scheme 1. (a) MeOH, concentrated H SO , reflux, 12 h, 85%; (b) 1,2-dibromoethane, K CO ,
2
4
2
3
DMF, reflux, 12 h, 60%; (c) (i) LiOH, H O, THF, rt, 8-12 h; (ii) isobutyl chloroformate, N-
2
methylmorpholine, THF, 0 °C, 30 min; (iii) NH in MeOH, rt, 60 min, 55%; (d) HNO ,
3
3
CF COOH, 0 °C, 6 h, 22% (15:16 = 40:60); (e) Pd/C, H , MeOH, rt, 8-10 h, 60-70%.
3
2
To avoid the complexity of introducing racemates during installation of substituents at
prochiral centers C2 or C3 of the dioxine ring, we reacted intermediate 13 with (2S)-glycidyl
tosylate and (2R)-glycidyl tosylate, respectively, to obtain esters 19a/19b (40:60 ratio) (Scheme
2
) and 24a/24b (40:60 ratio) (Scheme 3) [16]. Without further purification, these methyl esters
were converted to their corresponding amides 20a/20b and 25a/25b via hydrolysis followed by
amide formation using mixed-anhydride method. The primary alcohol group of these amides was
subjected to mesylation by treating with methanesulfonyl chloride in the presence of
triethylamine to obtain 21a/21b and 26a/26b. Displacement of the mesylate group in 21a/21b
9

and 26a/26b by N-methylpiperazine or pyrimidinyl-piperazine yielded 22a/22b, 23a/23b,
7a/27b, and 28a/28b as shown in Schemes 2 and 3.
2
COOCH3
O
CONH2
O
CONH2
O
CONH2
O
CH3
O S O
O
b
c
c
d
OH
OH
OH
OH
R
R
O
O
O
O
COOCH3
OH
19a
20a
21a
22a and 23a
a
OH
COOCH3
O
CONH2
O
CONH2
O
CONH2
CH3
d
O
S
O
13
b
O
O
O
O
O
O
19b
20b
21b
22b and 23b
2
2a and 22b: R =
3a and 23b: R =
N
N
N
N
2
N
N
Scheme 2. (a) (2S)-glycidyl tosylate, K CO , DMF, 40 °C, 30 h, 20-30%; (b) (i) LiOH, H O,
2
3
2
THF, rt, 10 h; (ii) isobutyl chloroformate, N-methylmorpholine, THF, 0 °C, 15-25 min; (iii) NH3
in MeOH, rt, 1 h, 60-70%; (c) methanesulfonyl chloride, triethylamine, DCM, rt, 10-12 h, 50-
5
5%, (d) N-methylpiperazine (for 22a and 22b), 1-(2-pyrimidinyl)piperazine (for 23a and 23b),
K CO , DCM, 0 °C to rt, 6-8 h, 20-33%.
2
3
1
0

COOCH3
O
CONH2
O
CONH2
O
CONH2
O
CH3
O S O
O
c
b
d
OH
OH
R
O
O
O
O
24a
25a
26a
27a and 28a
COOCH3
OH
a
COOCH3
O
CONH2
O
CONH2
O
O
CONH2
O
CH3
OH
S
OH
b
OH c
O
d
R
13
O
O
O
O
O
24b
25b
26b
7a and 27b: R =
8a and 28b: R =
27b and 28b
2
N
N
N
N
N
N
2
Scheme 3. (a) (2R)-glycidyl tosylate, K CO , DMF, 40 °C, 30 h, 20-30%; (b) (i) LiOH, H O,
2
3
2
THF, rt, 10 h; (ii) isobutyl chloroformate, N-methylmorpholine, THF, 0 °C, 15-25 min; (iii) NH3
in MeOH, rt, 1 h, 60-70%; (c) methanesulfonyl chloride, triethylamine, DCM, rt, 10-12 h, 50-
5
5%, (d) N-methylpiperazine (for 27a and 27b), 1-(2-pyrimidinyl)piperazine (for 28a and 28b),
K CO , DCM, 0 °C to rt, 6-8 h, 18-31%.
2
3
With the expectation that scaffold hopping could lead to more potent PARP1 inhibitors, a
series of analogues with different core structures were also prepared. Compounds 31 and 32 were
first prepared to evaluate the influence of varying dioxine ring size on PARP1 inhibition
(Scheme 4). Thus, intermediate 13 was reacted with the dibromomethane or 1,3-
dibromopropane, respectively, to obtain 29 and 30 [17]. Esters 29 and 30 were then converted to
target carboxamides 31 and 32 following the similar method used for preparing the carboxamide
4
[15].
1
1

COOCH3
OH
COOCH3
CONH2
O
O
R
O
2
9 and 31: R = CH2
a
b
R
O
OH
30 and 32: R = (CH2)3
13
29 and 30
31 and 32
Scheme 4. (a) dibromomethane (for 29), 1,3-dibromopropane (for 30), K CO , DMF, reflux, 8-
2
3
1
2 h, 55-70%; (b) (i) LiOH, H O, THF, rt, 8-12 h; (ii) corresponding acid, isobutyl
2
chloroformate, N-methylmorpholine, THF, 0 °C, 15-30 min; (iii) NH in MeOH, rt, 50-60 min,
3
6
0-70%.
To obtain new scaffold hopping analogues, target compounds 44-50 were designed and
synthesized by following a synthetic route outlined in Scheme 5. Commercially available acid 33
was initially converted to a methyl ester 34, and the phenolic hydroxyl group in 34 was subjected
to O-alkylation with appropriate ethyl -bromoacetate that afforded diesters 35-38 [18].
Reductive cyclization of 35-38 with iron and acetic acid afforded intermediates 39-42. The cyclic
amide group in ester 39 was reduced to 43 using NaBH in the presence of boron trifluoride-
4
diethyl etherate complex [19]. Target compounds 44-48 were synthesized according to the
procedure used for the synthesis of 4. Finally, target compounds 49 and 50 were prepared via
Knoevenagel condensation reaction conditions using 45 and corresponding benzaldehydes. It is
noted that the Knoevenagel condensation reaction produced 49 and 50 in <10% yields. This is
attributed to the weak acidic nature of the C2-methylene protons. To increase the amount of
product formation, we conducted the reaction using following modifications: (a) conventional
reflux of starting reagents in pyridine containing catalytic amounts of the piperidine, (b)
microwave irradiation at 120 °C in a 1:1 ratio of acetic anhydride and triethylamine, and (c)
microwave irradiation at 120 °C in pyridine containing catalytic amounts of the piperidine.
However, the modified conditions did not result in improved product yields.
1
2

COOH
OH
COOCH3
OH
COOCH3
O
COOCH3
O
COOC H
R
O
2
5
a
b
c
R
NO2
NO2
NO
N
2
H
33
34
35 - 38
39 - 42
d
e
3
5, 39, 45: R = H
CONH2
O
R
COOCH3
O
3
3
3
6, 40, 46: R =
N
H
O
7, 41, 47: R =
8, 42, 48: R =
OMe
Br
N
H
45 and
R'
O
4
3
46 - 48
f
e
CONH2
O
CONH2
O
N
H
N
H
4
4
4
9: R' = OH
50: R' = COOH
Scheme 5. (a) MeOH, conc. H SO , reflux, 12 h, 88%; (b) appropriate ethyl α-bromoacetate,
2
4
K CO , DMF, rt, 8 h, 69-84%; (c) Fe, acetic acid, reflux, 4 h, 69-78%; (d) BF -Et O, NaBH ,
2
3
3
2
4
THF, 5 °C, 2 h, 23%; (e) (i) LiOH, H O, THF, rt, 10 h; (ii) isobutyl chloroformate, N-
2
methylmorpholine, THF, 0 °C, 20 min; (iii) NH in MeOH, rt, 30 min, 30-60%; (f) 45, 4-
3
hydroxybenzaldehyde (for 49), 4-carboxybenzaldehyde (for 50), Et N/Ac O (1/1), reflux, 12 h,
3
2
8
-10%.
Structural Elucidation of Regioisomers 21a and 21b by HMBC Experiments
Regioisomerism assignments were performed based on the NMR analyses of compounds
1
13
2
1b and 21a, as illustrated in Figures 2 and 3 respectively. While 1-bond H- C heteronuclear
single quantum coherence (HSQC) spectrum showed similar patterns for the two regioisomers
1
3

1
13
(
Figures S4 and S5, supporting information), the 3-bond long range H- C coupling experiment
(HMBC = Heteronuclear Multiple Bond Correlation) not only allowed for assignment of
quaternary carbons, but also showed the differences between the two regioisomers. In the 1D
proton spectrum H6 and H4 appear as doublets whereas H5 appears as a triplet. Assignment of
H6 was confirmed by the 3-bond coupling to the carbonyl carbon in the HMBC spectra (peak C
in Figures 2 and 3). Identifying H6 allowed for the identification of quaternary carbon, C2 (peak
B in Figures 2 and 3) through its 3-bond coupling, while quaternary carbon, C3 was identified by
its 3-bond coupling to H5 (peak E in Figures 2 and 3). Unequivocal assignment of C2 and C3
was utilized to find if C8 or C9 were substituted with a functional group. In the expanded HMBC
spectra of 21b shown in Figure 2, C3 showed 3-bond coupling to a pair of methylene protons
(peaks F and G). This was possible only if there was no substituent on C9. On the other hand, in
the expanded HMBC spectra of 21a shown in Figure 3, C2 showed 3-bond coupling to a pair of
methylene protons (peaks F and G) clearly indicating that C8 was unsubstituted. The
regiochemistry of the hydroxymethyl, N-methylpiperazine and pyrimidinyl-piperazine
substituted compounds was also assigned based on the fact that they shared the same
intermediates such as 19a/19b and 24a/24b.
1
4

1
13
Figure 2. H- C HMBC spectra of 21b regioisomer: Crosspeaks identified as A, B, and C are 3-
bond couplings from H6 to C4, C2, and C7 respectively. Crosspeaks identified as D and E are 3-
bond couplings from H5 to C1 and C3. Crosspeaks identified as F and G are 3-bond couplings
from the C9 methylene protons to C3 confirming that the C9 carbon is unsubstituted.
1
5

1
13
Figure 3. H- C HMBC spectra of 21a regioisomer: Crosspeaks identified as A, B, and C are 3-
bond couplings from H6 to C4, C2, and C7 respectively. Crosspeaks identified as D and E are 3-
bond couplings from H5 to C1 and C3. Crosspeaks identified as F and G are 3-bond couplings
from the C8 methylene protons to C2 confirming that the C8 carbon is unsubstituted.
Structure-Activity Relationship
Initial structure- and pharmacophore-based virtual screening studies led to the
identification of 11 compounds that were categorized into three classes: with high, moderate, or
low PARP1 inhibition (Table 1). Based on novelty of structure and ease of synthesis, we selected
compound 4 (IC = 5.8 µM) as our lead for further optimization. PARP-inhibitory positive
5
0
1
6

controls (olaparib and veliparib) were used for comparison. Table 2 summarizes the installation
of small to large substitutions at various positions on the bicyclic ring system of 4 and their
resulting effects on activities. To obtain a direct comparison of activities between optimized
analogues and lead compound 4, percentage inhibition against PARP1 by target compounds at 5
µM was determined. The compounds showing >60% inhibition were evaluated for their IC50
values. Introduction of nitro (15 and 16) or amino (17 and 18) groups on the phenyl portion of
the benzodioxine ring led to a dramatic loss of activity. Further we wanted to explore the effect
of various polar substituents on the dioxine portion of lead compound 4. Synthesis of compounds
with such substitutions will generate a stereogenic center for each modification. To avoid the
complexity involved in the purification and identification of active enantiomers, a synthetic route
that would simplify both chemistry and SAR interpretation was necessary. As a result, we
optimized the synthesis using two homochiral glycidyl tosylates as the starting materials. This
assisted us in achieving the synthesis of the pure enantiomers with a desirable purity and
additionally, the synthesis gave rise to two regioisomers for each enantiomer. The compound
library thus generated, comprising 20a-23a, 20b-23b, 25a-28a, and 25b-28b, was screened in
vitro against PARP1 enzyme (see Table 2).
While analogues with hydroxymethyl extension, 25a (with C2-substitution, IC = 6.9
5
0
µM) and 25b (with C3-substitution, IC = 7.8 µM) displayed comparable activities as that of
5
0
lead compound 4, the modifications leading to compounds 20a and 20b proved detrimental.
Enantiomerically pure hydroxymethyl regioisomers were then converted to respective
methanesulfonates 21a, 21b, 26a and 26b with the hypothesis that they would capture additional
hydrogen bonding interactions in the adenosine-binding pocket (ABP) of PARP1 active site.
Interestingly, C3-substituted enantiomers 21b (IC50 = 0.69 µM) and 26b (IC50 = 1.2 µM)
1
7

exhibited ~8-fold and ~5-fold improvement in activity, compared to the unsubstituted lead 4
IC = 5.8 µM), while analogues 21a and 26a with C2-mesylate demonstrated a loss of PARP1
(
5
0
inhibition. Compound 21b also showed ~10-fold improved activity over hydroxymethyl
precursor 25a. From these results, it was evident that methanesulfonyl extensions at the C3-
position were more favored regardless of its stereochemistry, compared to those at the C2-
position. In conclusion, compound 21b with (S)-configuration was the most potent amongst these
four analogues.
To further assess the importance of methanesulfonyl group, we developed a compound
library where this group was displaced with N-methylpiperazinyl extension resulting in
compounds, 22a, 22b, 27a, and 27b (see Table 2). This modification was anticipated to capture
ionic and/or hydrogen bonding interactions within the ABP of PARP1 active site; however, these
analogues showed a significant decrease (<20% inhibition at 5 µM) in the activity profile
compared to the lead compound 4 (IC = 5.8 µM). Furthermore, replacement of the methyl
5
0
group in N-methylpiperazine with a 2-pyrimidinyl extension (analogues 23a, 23b, 28a and 28b)
also exhibited a significant loss of activity (<20% inhibition at 5 µM). These results together
suggested that bulky extensions at the C2- or C3-positions, were perhaps pointing toward a
sterically hindered region in the active site of PARP1.
Table 2. PARP1 inhibitory activities of target compounds 15-18, 20-23, and 25-28
PARP1
PARP1
IC50 (µM)
Compd
Structure
IC50
Compd
Structure
CONH2
a,b
a,b
(µM)
CONH2
O
CH3
O S O
O
5.8 ±
.10
O
4
26a
>5
0
O
O
1
8

CONH2
CONH2
O
O
CH3
O
S
1
1
1
1
5
6
7
8
>5
>5
>5
>5
>5
>5
26b
22a
22b
27a
27b
23a
O
1.2 ± 0.21
O
O N
O
O
2
CONH2
CONH2
O
O
CH3
N
>5
>5
>5
>5
>5
O
N
O
NO2
CONH2
O
CONH2
O
N
N
N
H N
2
O
O
CH3
CH3
CONH2
CONH2
O
O
O
N
N
O
NH2
CONH2
O
CONH2
O
2
0a
OH
OH
N
O
O
CH3
CONH2
O
N
CONH2
O
N
N
2
0b
N
O
O
CONH2
O
CONH2
O
N
6.9 ±
N
N
>5
>5
2
5a
23b
O
0
.95
OH
OH
O
N
N
CONH2
O
CONH2
O
2
5b
7.8 ± 2.7 28a
N
N
O
N
O
1
9

CONH2
O
CONH2
O
CH3
O S O
O
N
2
1a
>5
28b
N
N
>5
O
N
O
CONH2
O
CH3
O
S
O
0.69 ±
O
2
1b
0
.27
O
0
0
.0012 ±
.00020
0.0015 ±
0.00020
Olac
Velid
-
-
a3-Aminobenzamide sample included in the assay kit was used as an internal standard, and
b
showed a consistent result as reported (60 – 80% inhibition at 50 µM). All compounds were
assayed at a single concentration of 5 µM in triplicates. Compounds with <40% inhibition at 5
µM were denoted as PARP1 IC >5 µM. Compounds showing >60% inhibition at 5 µM were
5
0
c
d
further evaluated to obtain PARP1 IC . olaparib; veliparib
50s
Finally, to further improve the inhibitory potency of benzodioxine lead compound 4, the
scaffold hopping strategy was implemented. We therefore explored the effect of modifying the
size of the fused ring system and studied its effect on inhibitory potency (Table 3). With the
intention of securing compounds with enhanced potency, target compounds that exhibited >60%
inhibition at 2.5 µM concentration were prompted for PARP1 IC50 determination. The
synthesized benzodioxole compound 31 exhibited an IC value of ≥2.5 µM (literature IC = 5.3
5
0
50
µM [20] while benzodioxepine 32 showed significantly decreased activity (15% inhibition at 2.5
µM). This data suggested that the six-membered ring was perhaps the optimum size ring for
PARP1 inhibitory activity. To evaluate the contribution of the oxygen atom in the dioxine ring,
we replaced one of the oxygen atoms (H-bond acceptor) in compound 4 with bioisosteric –NH–
(H-bond donor) group. Our hypothesis, based on structural evidence, was that this modification
2
0

could significantly improve PARP1 inhibition through favorable water-mediated hydrogen
bonding interaction with the catalytically crucial Glu988 residue in the nicotinamide binding
pocket. To our surprise, 1,4-benzoxazine derivative 44 (IC = 11 µM) was ~2-fold less potent
5
0
than 4. However, converting a moderately basic ring amine in 44 to a neutral cyclic carboxamide
led to 45 that exhibited 15-fold enhanced activity (IC = 0.75 µM) in comparison to 44. This
5
0
improvement in potency may be attributed to a water-mediated hydrogen bonding interaction of
5 with Glu988 residue. Analogue 45 was considered as a refined lead compound and
4
subsequent substitutions were performed at its C2-position to enhance hydrophobic/hydrogen
bonding interactions. A phenyl extension at C2-position was thus introduced as a racemate to
capture hydrophobic interactions, which resulted in decreased activity (46, IC50 ~ 2.5 µM).
Electron-donating (methoxy) or electron-withdrawing (bromo) substitutions at the para-position
of the C2-phenyl led to racemic mixtures 47 (methoxy) and 48 (bromo). Compound 47 (IC =
5
0
0
.80 µM) displayed a comparable activity whereas 48 (IC > 2.5 µM) showed decreased activity
50
with respect to 45. Since benzylidene substitutions in our previously published work proved
successful in generating potent analogues, [9,15] we decided to introduce a rigid substitution
such as para-substituted benzylidene moiety at the C2-position. This led to the synthesis of two
compounds with further gain in potency, e.g., compounds 49 (IC = 0.082 µM) and 50 (IC =
5
0
50
0
.42 µM). The pattern in which various structural modifications of the benzodioxine and
benzoxazinone scaffolds favor or disfavor PARP1 inhibition is depicted in Figure 4.
Table 3. PARP1 inhibitory activities of Scaffold Hopping analogues
PARP1
Compound
Structure
a,b
IC (µM)
5
0
2
1

CONH2
O
4
5.8 ± 0.10
>2.5 (5.3)c
>2.5
O
CONH2
O
3
1
2
O
CONH2
O
3
4
O
CONH2
O
4
5
6
7
11 ± 1.9
0.75 ± 0.090
~2.5
N
H
O
NH2
O
4
4
4
N
H
O
O
NH2
O
N
O
H
O
NH2
O
O
0.80 ± 0.15
N
H
O
2
2

O
NH2
O
Br
4
4
8
9
>2.5
N
H
O
OH
O
NH2
0.082 ± 0.045d
O
N
H
O
COOH
CONH2
0.42 ± 0.046d
5
0
O
N
H
O
Olae
Velif
0.0012 ± 0.00020
0.0015 ± 0.00020
-
-
a3-Aminobenzamide sample included in the assay kit was used as an internal standard, and
b
showed a consistent result as reported (60 – 80% inhibition at 50 µM). All compounds were
assayed at a single concentration of 2.5 µM in triplicates. Compounds with <40% inhibition at
2
.5 µM were denoted as PARP1 IC >2.5 µM. Compounds showing 40 - 60% inhibition at 2.5
50
µM were considered to have PARP1 IC ~2.5 µM. Compounds showing >60% inhibition at 2.5
5
0
c
d
µM were further evaluated to obtain PARP1 IC . Literature IC value; Upon observing >70%
5
0
50
inhibition at 0.5 µM in PARP1 enzyme assay, IC values of compounds 49, 50, olaparib, and
5
0
2
3

veliparib were determined by chemiluminescence assay (BPS Bioscience, San Diego, CA);
e
f
olaparib; veliparib
(
(
S or R)-methanesulfonylmethyl favorable;
S)-hydroxymethyl retains activity;
CONH2
(R)-hydroxymethyl, (S or R)-4-methylpiperazinylmethyl,
and (S or R)-pyrimidinylpiperazinylmethyl detrimental
O
R
R
R
O
R
(R)-hydroxymethyl retains activity;
(
S)-hydroxymethyl, (S or R)-methanesulfonylmethyl,
nitro or amino substitutions
lead to loss of activity
(S or R)-4-methylpiperazinylmethyl, and
(S or R)-pyrimidinylpiperazinylmethyl detrimental
dioxole as active as 1,4-dioxine;
phenyl/4-bromophenyl
unfavorable;
4-methoxyphenyl favorable;
4-hydroxy- and 4-
carboxybenzylidene
increases potency
CONH2
CONH2
O
1,4-dioxepine and 1,4-oxazine
R
O
ring unfavorable;
3-oxy-1,4-oxazine leads to
N
H
improved activity
Figure 4. SAR summary of the 1,4-benzodioxine and 3-oxo-2,3-dihydrobenzoxazine scaffolds.
Because PARP-isoform selectivity is a key to obtain clinically useful new compounds to
treat cancers, we conducted screening of the most potent compound 49 along with positive
controls against seven clinically investigated PARP-isoforms (e.g., PARP2, PARP3, TNKS1
(PARP5A), TNKS2 (PARP5B), PARP8, PARP10, and PARP14 (Figure 5 and Table S5).
Compound 49 (PARP1 IC = 0.082 µM) showed complete inhibition of PARP2 at 0.5 µM
5
0
concentration, which is not surprising due to a highly conserved catalytic domain (>69%
homology) being shared by both PARP1 and PARP2.[21] Interestingly, 49 also substantially
inhibited PARP3, TNKS1, and TNKS2, which are involved in cancer progression.[22,23]
Conversely, 49 did not interfere with other PARP-isoforms such as PARP8, PARP10 or
PARP14. Further investigation on PARP-isoform screening with low testing concentrations will
be conducted based on the progress of future structure optimization.
2
4

Figure 5. Percentage inhibition data of 49 against seven PARP-isoforms. The reported data is
based on the two independent experiments, and each experiment was performed in duplicate.
Olaparib for PARP2, PARP3, PARP8, and PARP10, XAV- 939 for TNKS1 and TNKS2 and
rucaparib for PARP14 were used as the positive controls (%inhibition data of the positive
controls are reported in supporting information, Table S5).
To elucidate the plausible binding interactions of the most potent compound 49 at the
active site of PARP1, we docked compound 49 using the Glide docking tool and co-crystal
structure of full length PARP1-dihydrobenzofuran-7-carboxamide (Figure 6) [15]. In the binding
model, the pharmacophore carboxamide group of 49 captures three key hydrogen bonding
interactions with the side chain of Ser904 and the backbone atoms of Gly863, similar to what we
observed for lead 4 (Figure S2, supporting information). Furthermore, a productive
2
5

intramolecular hydrogen bond between the amide proton and the oxazinone ring oxygen atom
was also evident. The phenyl ring of the benzoxazinone moiety formed a π-π stacking interaction
with the side chain of Tyr907. The 4-hydroxybenzylidene moiety extended toward the ABP of
the PARP1 active site while forming a hydrogen bond with the side chain of Ser864.
Additionally, the phenolic hydroxyl group of 49 could form an electrostatic contact with the side
chain of a key residue Asn868 due to the proximity between the two groups [9]. The
benzoxazinone ring –NH- is found located near Glu988, which suggests a plausible water-
mediated hydrogen bonding. Future studies will explore the effect of a wide range of substituents
on the benzylidene ring to extend the molecule further into the ABP of PARP1 enzyme. These
efforts will be facilitated by optimizing the reaction conditions for the Knoevenagel
condensation, which is the low-yielding final step to produce target compounds.
Figure 6. A binding model of 49 in the active site of PARP1. Key amino acids are depicted in
stick model with the atoms colored as carbon – green, hydrogen – white, nitrogen – blue, and
oxygen – red. The ligand is shown in ball and stick model with the same color scheme as above
2
6

except carbon atoms are represented in orange. The broken red, pink and cyan lines indicate
hydrogen bonding interaction, potential electrostatic contacts and π-π interaction, respectively.
Conclusion
In this study, PARP1 inhibitory hit compound 4 was first identified from a structure- and
pharmacophore-based virtual screening protocol. Polar substitutions installed at the C2- or C3-
positions of the benzodioxine ring of 4 led to analogues 21b and 26b with improved enzymatic
inhibitory activities. Finally, utilization of the scaffold hopping strategy identified a refined lead
compound 45 that was further optimized to generate a potent PARP1 inhibitor 49. Binding
models of key compounds in the PARP1 active site allowed rationalization of the observed SAR
trends. It is anticipated that compound 49 will serve as a promising lead compound for
generation of selective PARP1 inhibitory agents with low nanomolar potency.
Experimental
High-throughput virtual screening
A detailed information regarding high-throughput virtual screening (ligand preparation,
protein refinement, Glide docking procedure, and 3D pharmacophore modeling) can be found in
supporting information.
Materials and methods
The selected 11 compounds from HTVS were obtained from Maybridge Chemicals Ltd.
(
Tintagel, United Kingdom) and Matrix Scientific (Columbia, SC). Starting materials for
synthesis were received from the following vendors and used as received: Aldrich Chemical Co.
Milwaukee, WI), AK scientific (Union City, CA), Combi-Blocks Inc. (San Diego, CA), TCI
America (Portland, OR), Gold Biotechnology (St. Louis, MO), Sigma-Aldrich (St. Louis, MO)
(
2
7

and Alfa Aesar (Ward Hill, MA). The uniformity of all synthesized compounds was confirmed
by thin layer chromatography (Agela Technologies). Thin layer chromatography (TLC) was
employed to monitor the progress of reactions at 254 nm ultraviolet light. Bruker 400
UltrashieldTM spectrometer ( H at 400 MHz and 13C at 100 MHz) equipped with a z-axis
gradient probe was used to record NMR experiments. Tetramethylsilane (TMS) was used as an
1
1
internal standard, and chemical shift (δ) of H NMR was reported downfield of the TMS signal
1
in parts per million (ppm). The H NMR data are presented as follows: chemical shift
[multiplicity s (singlet), br (broad singlet), d (doublet), t (triplet), q (quartet), dd (doublet of
doublets), and m (multiplet), number of protons, coupling constant]. 2D NMR experiments were
1
13
carried out on Avance III NMR spectrometer with a H frequency of 500 MHz and C frequency
of 125 MHz and equipped with an inverse probe. Samples were dissolved in DMSO-d or
6
1
13
chloroform-d and experiments were carried out at room temperature. H and C chemical shift
1
1
1
13
1
13
assignments were carried out with the aid of H- H COSY, H- C HSQC and H- C HMBC
experiments. Purification by flash chromatography was performed on Reveleris X2 flash
chromatography system (BUCHI Corporation, New Castle, DE). High resolution mass spectra
(HRMS) were collected for all unknown target compounds using a Waters Xevo G2-XS QToF
mass spectrometer equipped with H-Class UPLC inlet and a LockSpray ESI source. Chiral
HPLC analysis was performed using Agilent 1100 HPLC system and column (250 mm × 4.00
mm, serial no. 111156). Compounds were dissolved in acetonitrile and injected (10 µL) into
Chirex column (phenomenex, Torrance, CA) with stationary phase as (S)-valine and (R)-1-(α-
naphthyl)ethylamine. Optimum resolution of the enantiomers was achieved using isocratic
mobile phases eluting at a flow rate of 1 mL/min. The elution was monitored at UV 254 nm in
the form of major and minor peaks representing each enantiomer. Enantiomeric excess (%ee)
2
8

was calculated based on the following equation, ee = ((R-S)/(R+S)) x 100%, where (R) and (S)
denote individual optical isomers in the mixture and (R) + (S) = 1.
Synthetic procedures
General procedure to convert ester-containing intermediate to respective amides using
mixed-anhydride method (A) [15]. Mixture of ester containing intermediate (2.0 mmol),
lithium hydroxide (0.096 g, 4.0 mmol), and water (2 mL) in 10 mL of tetrahydrofuran was stirred
at room temperature for 10 h. After removal of solvent under vacuum, the reaction mixture was
diluted with ethyl acetate and extracted with water. The aqueous extract was then acidified using
2
N aqueous hydrochloric acid solution followed by extraction with ethyl acetate (3 X 20 mL).
The combined organic portion was dried over anhydrous magnesium sulfate, and concentrated
under reduced pressure to obtain the corresponding acid, which was subjected to the next step
without further purification.
The acid was dissolved in 10 mL tetrahydrofuran, and the temperature was reduced to 0
°
C under nitrogen atmosphere. After addition of isobutyl chloroformate (0.3 mL, 2.2 mmol) and
N-methylmorpholine (0.24 mL, 2.2 mmol), the same condition was maintained until acid was
completely converted to the desired mixed anhydride intermediate as monitored by TLC. Then a
2
2
.0 mL solution of 7N ammonia in methanol was added, and the reaction mixture was stirred for
0 min at room temperature. After extraction with ethyl acetate and water, crude product was
purified using flash chromatography with appropriate combination of ethyl acetate and hexanes.
Methyl 2,3-dihydroxybenzoate (13) [24]. To a solution of 2,3-dihydroxybenzoic acid (0.3 g,
1
.9 mmol) in 20 mL of methanol, concentrated sulfuric acid (0.3 mL) was added at 5 °C. The
2
9