2
A. Ward et al. / Bioorganic & Medicinal Chemistry Letters xxx (2017) xxx–xxx
Ligands were synthesized, as shown in Scheme 1, to allow for
independent optimisation of the three motifs (Scheme 1). Two gen-
eral strategies were implemented, both involving the incorporation
of motif 1 (R1) onto initial motif 3 (D to E, or H to I). The difference
between early (A to B) or late (I to F) stages allowed incorporation
of motif 2 (R2). The synthetic pathways were chosen so as to main-
tain the common intermediate as late as possible for structural
diversification. To introduce motif 1 (R1 in F), the common inter-
mediate acid D (e.g. R2 = 3-pyridyl) was assembled by coupling
ethyl anthranilate A with the corresponding acyl chloride, followed
by ester hydrolysis. Depending on the availability of the building
blocks, the cinnamide analogues (B) were also constructed by Heck
coupling of the corresponding aromatic bromide with acrylamide
C. After introducing motif 1 as an amine through amide coupling
(D to E), the intermediates were cyclized under mild dehydration
conditions with iodine and hexamethyldisilazine18 to give the
desired quinazolinone products (F). In this way, one series of com-
pounds incorporated alkyl and cycloalkyl substituents (1–8), and
another series contained substituted aromatics with a variable
spacer –(CH2)n– (n = 0–2, 9–27). The latter series was designed
to optimally target residues F195 and Y191 through pi-interac-
tions. Various substituents, such as halogen, hydroxy, amino and
its precursor nitro, carboxylate and acetamide were incorporated
to improve properties or polar interactions.
Fig. 1. Structurally different RAD51 inhibitors, including B02. Components of B02 to
Motif 2 (R2 in F, R1 = 4-chlorobenzyl) was assembled using
either a similar linear process (A to E to F), or more efficiently from
common intermediate 2-methylquinazolinone I through one-step
divergent enamine-aldehyde coupling (I to J to F). A one-pot syn-
thesis from anthranilic acid G to 2-methylquinazolinone I, through
the mixed anhydride 2-methyloxazinone H, was used to prepare
variations in Motif 3 (R3). To probe the shallow hydrophobic cleft
where threonine of Phe-His-Thr-Ala bound (Fig. 2), one amino
group was introduced at position 6 of the quinazolinone core
(44), which was further derivatized by either acylation (45–51)
or guanidinylation (52).
The potency of ligands was assessed using immunofluorescent
assay for their inhibition of DNA damage induced RAD51 foci for-
mation (Fig. 3), a critical property of RAD51 in HR. An initial mod-
ification at motif 1 (1–17) resulted in promising compounds, with
both saturated cyclohexylmethyl (6) and 4-chlorobenzyl (17) ana-
logues displaying improved inhibition of RAD51 foci formation.
One methylene spacer shorter (5 vs 6) or longer (10 vs B02)
reduced the potency of RAD51 functional inhibition. Restricting
rotation (indane 11) or introducing potential charged isosteres,
such as morpholine (7–8) or pyridine to replace benzene (12–
14), all reduced efficacy. Smaller alkanes (1–4) also displayed
reduced activity. Varying spacer length in compound 17, with
one methylene unit shorter (15) or longer (16), did not improve
activity and indicated optimal positioning of the aromatic ring in
17. Of the substituted benzyl series, fluoro (18–20), nitro (21–22)
and p-hydroxy (23) analogues showed comparable activity. A polar
substrate, such as p-acetamidomethyl (24) and carboxylic acid
(25–26), were detrimental, while 3,4-dichloro (27) conferred a
slight improvement. This demonstrated that a hydrophobic inter-
action was important at this site.
be varied in this study are separately colored.
approaches. Here we describe some structure-activity relation-
ships for analogues of B02, leading to the discovery of an inhibitor
selective for several human breast cancer cell lines including those
expressing high RAD51.
Several years ago, we developed a homology model of human
RAD51, based on a full-length homologue from Pyrococcus furiosus
(PDB code: 1PZN16), to aid compound design. B02 was docked into
the model in several different putative binding sites, including the
ATPase domain known to bind small fragments like tetrapeptide
and bicyclic aromatics.17 One preferred conformation of B02
showed motif 2 (3-pyridyl) occupying the same cleft that accom-
modated aromatic groups, like the phenylalanine side chain of
Phe-His-Thr-Ala (Fig. 2). Motifs 1 and 3 instead spread-eagled
across the shallow hydrophobic entrance to the cleft with the
charged residue D187 nearby. The cleft was surrounded by
hydrophobic residues (L104, M158, I160, A190, A192, L203, A207,
L219). A second shallow indentation close to motif 3, accommodat-
ing the threonine side chain of the tetrapeptide in the crystal struc-
ture with a truncated RAD51,17 is formed by hydrophobic residues
(F166, P168, L171, V185, L186, V189). These features were used to
design our compound library in this report.
Keeping motif 1 as 4-chlorobenzyl, any modifications at motif 2
apart from 3-pyridyl were detrimental, including its regioisomer 4-
pyridyl (36), mono-amino substituted 3-pyridyl (37–39), and a ser-
ies of mono-substituted (hydroxy, nitro or amino) phenyl (28–35).
This suggested that there were limits to both the substituent size
and the polar interaction, with only the 3-pyridyl moiety being
effective at this site.
In motif 3, incorporating an extended pi-system (40) or a
nitrogen isostere (41–43) led to similar or increased cytotoxic-
ity. In particular, the 6-aza quinazolinone analogue (43) pro-
duced the greatest cytotoxicity. Surface plasmon resonance
Fig. 2. Docking of B02 in the ATPase domain of a homology model of human RAD51.
Ward, Ambber
