362
K. Murphy-Benenato et al. / Bioorg. Med. Chem. Lett. 24 (2014) 360–366
Table 2
CN
CN
Cl
Crystallographic data collection and refinement statistics
S
S
a
b
Br
Compound
1
5c
Br
1
N
N
PDB codea
Space group
Cell constants a, b, c (Å) A, B,
4LH6
C222(1)
4LH7
C222(1)
HN
3
4
PMB
c (°) 49.0, 104.4, 136.7 48.8, 105.0, 136.1
90.0, 90.0, 90.0
52.19–1.65
(1.71–1.65)
98.9 (94.0)
42.077
4.6 (3.3)
11.9 (2.2)
0.058 (0.436)
22.4
25.4
2659
14
22
90.0, 90.0, 90.0
48.99–1.90
(1.97–1.90)
95.8 (91.7)
26.859
4.4 (4.5)
9.9 (4.0)
0.102 (0.249)
22.5
25.9
2621
16
22
Scheme 1. Reagents and conditions: (a) PMBNH2, K2CO3, NMP; (b) H2SO4.
Resolution range (Å)
(Highest resolution shell)
Completeness overall (%)
Reflections, unique
Multiplicity
to 3, followed by hydrolysis with sulfuric acid afforded the desired
product 1. With thienopyridine 1 in hand, experiments were de-
signed to address key questions: (1) does this kinase scaffold bind
in the AMP-binding site of bacterial ligase, and if so (2) what would
be the effect of the resistance mutation on binding affinity?
Thienopyridine 1 was tested for activity against a panel of LigA
enzymes (Table 1). The compound was found to have IC50 values in
I/
r
b
Rmerge
Rvalue
Rvalue
overall
overall
(%)c
(%)
free
Non hydrogen protein atoms
Non hydrogen ligand atoms (INH)
Non hydrogen ligand atoms (NMN)
Solvent molecules
the moderate to high
lM range against H. influenzae, S. aureus,
253
159
Enterococcus faecalis (E. faecalis), and S. pneumoniae. Due to its
moderate biochemical potency, compound 1 did not demonstrate
any cellular activity against any bacterial species tested. These en-
zyme potencies are approximately 100-fold higher than an adeno-
sine-based inhibitor 2 that incorporates a hydrophobic sidechain
that occupies the hydrophobic pocket in which the resistant muta-
tion Leu75Phe occurs9 (Table 1); these compounds also show po-
tency losses of >100-fold when tested biochemically against this
resistance mutation. In contrast, when the thienopyridine com-
pound 1 was tested for activity against the S. pneumoniae enzyme
containing the Leu75Phe mutation, the inhibitor exhibited only a
ꢁ2-fold IC50 difference between the wild type and mutant proteins
(Table 1). These results validate the methodologies used in the vir-
tual screen, as the identified thienopyridine compound contained
all the properties designed into the constraints used in the in silico
screening including activity against both Gram positive and Gram
negative isoforms of LigA (see Table 2).
A high resolution crystal structure of thienopyridine 1 bound to
the adenylation domain of E. faecalis LigA is shown in Figure 2. Thi-
enopyridine 1 bound in the adenosine binding site of the ligase
protein, making the multiple hydrogen bonds (Fig. 2A) predicted
by modeling. The amide makes a donor/acceptor interaction with
Glu118 and Lys291, the thienopyridine nitrogen engages with
the backbone NH of Ile121 and the amine interacts with the back-
bone carbonyl of Ile121. In addition to these hydrogen bonds, there
R.m.s. deviations from ideal values
Bond lengths (Å)
Bond angles (°)
Average B values (Å2)
Protein main chain atoms
Protein all atoms
Ligand (Compound)
Ligand (NMN)
0.006
0.960
0.006
0.992
26.3
27.0
34.0
28.6
36.3
24.7
25.5
25.0
31.9
34.2
Solvent
U, W
angle distribution for residuesd
In most favoured regions (%)
In additional allowed regions (%)
In generously regions (%)
95.1
4.9
0.0
96.1
3.9
0.0
In disallowed regions (%)
0.0
0.0
Rfree is the cross-validation R factor computed for the test set of 5% of unique
reflections.
a
Coordinates have been deposited in the Protein Data Bank.
P
P
P
b
c
Rmerge
Rvalue
¼
½ð jIi ꢂ hIijÞ= Iiꢃ.
i
P
P
¼
hkjljFobsij ꢂ jFcalcjj= hkljFobsj.
hkl
d
Ramachandran statistics as defined by PROCHECK.
starting from 1, 3, or 4 (see Scheme 2). The bromine atom allowed
for a diversity of chemistry: Zn(0) insertion (5a), Pd-catalyzed
Negishi reaction (5b), nitrile displacement with CuCN (5c), Pd-cat-
alyzed carbonylation (5d, 5e), and Pd-catalyzed Suzuki reaction
(5f). Beginning with Br intermediate 13 (Scheme 3),28 analogous
chemistry allowed access to R2 variation (see 16a, 16c and 16d).
En route to 16c, nitrile hydrolysis was observed when para-
methoxybenzylamine was introduced, presumably due to advanta-
geous water present in the reaction. Methyl substituted thieno-
pyridine 16b was synthesized from chloropyridine 17.33 Utilizing
intermediate 3, a variety of amines were introduced, offering R3
diversity (23a and 23b, Scheme 4). Finally disubstituted thieno-
pyridine 24 was synthesized by bromination of intermediate 19
(Scheme 5).
With access to a diversity of analogs, we examined their ability
to inhibit a panel of Gram positive NAD+-dependent DNA ligase en-
zymes. The data is summarized in Table 3. Similar to compound 1,
the compounds tended to be more active against the Gram positive
isozymes and none of the compounds demonstrated comparable
activity against H. influenzae. The active sites of the H. influenzae
and E. faecalis LigA proteins are extremely similar in overall struc-
is a
p-stacking interaction between the thienopyridine ring and
Tyr227. When overlaid with the E. faecalis LigA structure bound
to NAD+ (Fig. 2B),30 it is clear that the thienopyridine scaffold occu-
pies the same space in the binding pocket as the adenine base, and
avoids the hydrophobic pocket where the Leu75Phe resistance
mutation occurs in S. pneumoniae (equivalent to Leu89 in E. faecal-
is). The crystal structure of thienopyridine 1 was also solved in the
H. influenzae isozyme and was found to bind identically in both
proteins (data not shown). This result was not unexpected due to
the previous observations regarding the similarities of the LigA
binding site and allowed the higher resolution E. faecalis crystal
system to be used for SAR development.
The thienopyridine scaffold was an attractive starting point for
optimization.31 The lead compound had low molecular weight
(MW = 272), good physical properties (calculated logP = 1.57)
excellent ligand efficiency (LE = 0.455),32 and the core offered mul-
tiple points for diversification (Fig. 3). Guided by the experimen-
tally determined structure and modeling, two main areas of the
molecule were targeted for improvement of the biochemical po-
tency: (1) building a new interaction with the Tyr87 loop via R1
(Figs. 2 and 3), and (2) creating new interactions in the ribose pock-
et through R2 and R3 substitutions.
ture (C
a
RMSD of adenosine binding region ꢁ0.78 Å2) and in amino
acid composition. Yet differences in potencies were seen between
the Gram positive enzymes and H. influenzae enzyme tested due
to the subtle differences in the active site structures, highlighting
the potential challenge of finding a potent inhibitor for both Gram
positive and Gram negative LigA enzymes.
Compounds 5a–5f were designed to understand the effect of
the Br and to target new interactions with Tyr87. Removal of the
The chemistry to access these compounds is highlighted in
Schemes 2–5. Analogs with R1 variation (Fig. 3) could be accessed