Development of Single-Stranded DNA Bisintercalating Inhibitors of Primase DnaG as Antibiotics

Article abstract of DOI:10.1002/cmdc.202100001

Many essential enzymes in bacteria remain promising potential targets of antibacterial agents. In this study, we discovered that dequalinium, a topical antibacterial agent, is an inhibitor of Staphylococcus aureus primase DnaG (SaDnaG) with low-micromolar minimum inhibitory concentrations against several S. aureus strains, including methicillin-resistant bacteria. Mechanistic studies of dequalinium and a series of nine of its synthesized analogues revealed that these compounds are single-stranded DNA bisintercalators that penetrate a bacterium by compromising its membrane. The best compound of this series likely interacts with DnaG directly, inhibits both staphylococcal cell growth and biofilm formation, and displays no significant hemolytic activity or toxicity to mammalian cells. This compound is an excellent lead for further development of a novel anti-staphylococcal therapeutic.

Full text of DOI:10.1002/cmdc.202100001

Full Papers
doi.org/10.1002/cmdc.202100001
ChemMedChem
1
2
3
4
5
6
7
8
9
Development of Single-Stranded DNA Bisintercalating
Inhibitors of Primase DnaG as Antibiotics
Keith D. Green,[a] Ankita Punetha,[a] Nishad Thamban Chandrika,[a] Caixia Hou,[a]
Sylvie Garneau-Tsodikova,*[a] and Oleg V. Tsodikov*[a]
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
Many essential enzymes in bacteria remain promising potential
targets of antibacterial agents. In this study, we discovered that
dequalinium, a topical antibacterial agent, is an inhibitor of
Staphylococcus aureus primase DnaG (SaDnaG) with low-micro-
molar minimum inhibitory concentrations against several S.
aureus strains, including methicillin-resistant bacteria. Mechanis-
tic studies of dequalinium and a series of nine of its synthesized
analogues revealed that these compounds are single-stranded
DNA bisintercalators that penetrate a bacterium by compromis-
ing its membrane. The best compound of this series likely
interacts with DnaG directly, inhibits both staphylococcal cell
growth and biofilm formation, and displays no significant
hemolytic activity or toxicity to mammalian cells. This com-
pound is an excellent lead for further development of a novel
anti-staphylococcal therapeutic.
Introduction
inhibit bacteria, while sparing human cells.[3] DnaG has been
genetically validated as a potential target in different bacteria,
owing to the classical identification of temperature-sensitive
mutants of the dnaG gene[4] and, more recently, by demonstrat-
ing cell growth inhibition in response to controlled down-
regulation of dnaG.[5] The search for inhibitors of DnaG, which
has started in industry over 10 years ago, has since been
continued in academia. These initial studies found low-micro-
molar inhibitors of DnaG that did not have antibacterial
activity,[6] which was likely the reason for the lack of their
further development. Another recent study identified very weak
(~100 μM) DnaG inhibitors.[7] In the high-micromolar/low-milli-
molar concentration range, where such small molecules display
the activity of interest, they are expected to bind multiple low-
affinity targets and require a 100-fold or greater improvement
in potency by synthetic methods, which is neither likely nor
practical. Our group has developed a robust high-throughput
coupled primase-pyrophosphatase colorimetric assay, which
yielded several low-micromolar DnaG inhibitors from screening
a small (~2500-compound) library of clinically used drugs.[8]
One of the low-micromolar inhibitors that we discovered,
tilorone,[8a] that itself displayed no significant antibacterial
activity has led to analogues with modest antibacterial activity
(MIC=16 μM) that were synthesized by a different group.[9] We
also showed that the inhibitory activity against Mycobacterium
tuberculosis DnaG (MtbDnaG) of one of the assay hits, doxor-
ubicin, and similar molecules with some propensity for DNA
intercalation, correlated with their anti-mycobacterial activity,
strongly suggesting that DnaG was a target of chemical
inhibition in the bacterial cell.[10] Doxorubicin is used in clinic as
an anticancer agent. Although doxorubicin is not considered for
development as an antibiotic, it has served as a useful tool to
validate bacterial DnaG as a target. Doxorubicin inhibits human
topoisomerase II, which shares the same TOPRIM (topoisomer-
ase-primase) fold both with primase DnaG and with the
bacterial gyrase. Doxorubicin is a very weak (~100 μM) inhibitor
of gyrase,[11] which indicates that, as with the human top-
Drug-resistant bacterial infections are a significant medical
concern, as their emergence and spread are outpacing anti-
biotic discovery and development. For example, so-called
ESKAPE (Enterococcus faecium, Staphylococcus aureus, Klebsiella
pneumoniae, Acinetobacter baumannii, Pseudomonas aeruginosa,
and Enterobacter) pathogens are targets of vigorous research
efforts in academia, due to the frequent resistance of these
bacteria to multiple common antibiotics and a consequent high
mortality rate. ESKAPE bacteria account for a large fraction of
the 100000 yearly deaths from hospital-acquired infections in
the USA.[1] Target-based drug discovery, once considered
unpopular, is gaining momentum with development of sensi-
tive assays, affordable compound libraries and liquid dispensing
technologies, enabling screening of a large number of mole-
cules to find potent inhibitors. This initial discovery process is
followed by validation and compound optimization by synthetic
methods of medicinal chemistry, which are often guided by X-
ray crystallography. Prominent classes of targets for which
target-based drug design has been especially fruitful are kinases
and G protein-coupled receptors.[2]
The bacterial primase DnaG is an enzyme that is essential
for the replication of the bacterial chromosome. DnaG is the
only enzyme that can generate RNA primers as a part of the
replication machinery in bacteria. DnaG is conserved in all
bacteria and is dissimilar to the human primase, making DnaG
an attractive potential target for agents that could selectively
[a] Dr. K. D. Green, Dr. A. Punetha, Dr. N. T. Chandrika, Dr. C. Hou,
Prof. S. Garneau-Tsodikova, Prof. O. V. Tsodikov
Department of Pharmaceutical Sciences
University of Kentucky
Lexington, KY 40536–0596 (USA)
E-mail: oleg.tsodikov@uky.edu
sylviegtsodikova@uky.edu
Supporting information for this article is available on the WWW under
https://doi.org/10.1002/cmdc.202100001
ChemMedChem 2021, 16, 1–11
1
© 2021 Wiley-VCH GmbH
��
These are not the final page numbers!

Full Papers
doi.org/10.1002/cmdc.202100001
ChemMedChem
oisomerase II, the mechanism of action of doxorubicin in
bacteria involves interactions with DnaG, in addition to
interactions with DNA. Conversely, widely clinically used
fluoroquinolones, which are weak DNA intercalators (Kd ~
70 μM),[12] are potent inhibitors of bacterial gyrase, owing to
gyrase-specific interactions, but they are inactive against either
human topoisomerase II or bacterial DnaG.[8b]
In this study, we continued our search for DnaG inhibitors
that could be used as a starting point for development of novel
antibacterial agents. The discovery, initial medicinal chemical
optimization, and mechanistic characterization of a selective
anti-staphylococcal agent with a sub-micromolar minimum
inhibitory concentration (MIC) is presented herein.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
Results and Discussion
Discovery of compounds 3 and 8 as SaDnaG inhibitors
As a target for inhibitor discovery, we chose primase DnaG from
S. aureus (SaDnaG), a pathogen that causes over 100000 new
infections and ~20000 deaths yearly in the US alone. Especially
dangerous are the infections caused by methicillin-resistant S.
aureus, or MRSA, for which the antibiotic armamentarium is
very limited. SaDnaG was cloned, expressed in Escherichia coli
BL21(DE3) and purified. A total of 2395 structurally diverse
drug-like compounds were tested for SaDnaG inhibition using
our previously established assay.[8] The assay was carried out
using multichannel pipets in 96-well plates. The Z’ scores[13] of
this medium-throughput assay varied in the range 0.26–0.77 for
different plates, indicating its usability. The tested compounds
were relatively rigid molecules that have a few rotatable bonds,
few or no chiral centers and do not contain weakly bonded
heteroatoms, organometallics, or polycyclic aromatic hydro-
carbons. The dose-response dependence of the compounds
that displayed inhibition with Z scores of 2.5 or greater in the
initial assay was measured to validate these assay hits. We
identified and validated two hit compounds, 3 and 8 (Figure 1)
with Z scores of 2.6 and 8.8, respectively. These compounds are
bola-amphiphilic, containing two polar aminoquinoline groups
(head groups) linked to each other by their ring nitrogens (as in
compound 3) or the secondary amino groups (as in compound
8) via an alkyl linker. Compound 3 contains a positively charged
head group while compound 8 is neutral. Compound 3, also
known as dequalinium, is a bacteriostatic agent used as a
topical antiseptic. It has been used against bacterial wound
infections, particularly oral and vaginal infections. Despite its
use for over 60 years, the dominant mode of action by which
this compound inhibits bacterial growth has not been estab-
lished, although two potential mechanisms have been identi-
fied. Dequalinium was shown to intercalate into double-
stranded DNA, but this interaction was shown to be relatively
weak.[14] Another potential target of compounds 3 and 8, given
their amphiphilic nature, could be the cytoplasmic membrane.
Dequalinium (3) was previously found to be a much weaker
inner membrane disruptor than classical membrane perturbing
agents cetrimide and chlorhexidine, leading the authors to
Figure 1. Synthetic schemes for the preparation of compounds A) 1–5 and
B) 6–10.
conclude that the primary mode of action of this molecule was
likely through nucleic acid binding.[15] Compounds 3 and 8 may
function as SaDnaG inhibitors solely or partially through bind-
ing DNA, but this mechanism merits deeper investigation, as
DnaG initiates its primer synthesis on single-stranded DNA. To
investigate the potential of the identified SaDnaG inhibitors to
be developed as antibiotics, we synthesized compounds 3, 8,
and their analogues (1, 2, 4–7, 9, and 10), tested their activities
against bacterial and mammalian cells, and interrogated their
mechanism of action.
Synthesis of compounds 1–10
In designing analogues of compounds 3 and 8 for studying
their mechanism of action and potential optimization as
antibacterial agents, we varied the length of the linker between
the two heads, we removed one head, and varied the length of
the tail attached to one head. We also considered an impact of
the positive charge. We synthesized, characterized (Figures S1–
ChemMedChem 2021, 16, 1–11
www.chemmedchem.org
2
© 2021 Wiley-VCH GmbH
��
These are not the final page numbers!

Full Papers
doi.org/10.1002/cmdc.202100001
ChemMedChem
S30 in the Supporting Information; note: all compounds were at
least 95% pure), and scaled up the synthesis of initial hits 3 and
8 as well as generated eight other analogues (1, 2, 4–7, 9, and
10) for structure-activity relationship (SAR) studies. The ten
compounds used in this study, which are either a combination
of two head groups joined by a linker (C4–C12) or a single head
with a tail (C8–C10) were synthesized as depicted in Figure 1.[16]
Compounds 1–5, which are quaternary ammonium cations,
were synthesized by a substitution reaction between 4-amino-
quinaldine and diiodoalkanes (for compounds 1–3) or iodoal-
kanes (for compounds 4 and 5; Figure 1A). More precisely, the
reaction of 4-aminoquinaldine with corresponding diiodoal-
kanes 1,6-diiodohexane, 1,8-diiodooctane, and 1,10-diioodode-
chloroquinaldine and 1-aminodecane resulted in the formation
of compound 10 in 84% yield.
1
2
3
4
5
6
7
8
9
Inhibition of bacterial growth
MIC values of compounds 1–10 were first measured for a broad
panel of bacteria derived from clinical isolates, including high-
priority ESKAPE[17] pathogens (strains A, D, G, N, P--T, and Y;
Tables 1 and 2) and M. tuberculosis. We chose to expand the
panel for three types of bacteria: i) Escherichia coli (strains A–C;
Table 1), because some compounds displayed excellent activity
against E. coli, ii) staphylococci (strains E–O; Table 1), and iii) P.
aeruginosa (strains U–X; Table 2), since the P. aeruginosa strain
ATCC 27853 (strain T; Table 2) tested initially was highly
resistant to the control antibiotic kanamycin (KAN). All com-
pounds, except for compounds 1 and 6–8 displayed low- or
sub-micromolar MIC values against the majority of strains (i)
and (ii) above and against M. tuberculosis (Table 1). These strains
included KAN-resistant strains. All compounds were significantly
less active against the other bacterial strains tested (with MIC of
at least 64 μM; Table 2), with a few exceptions. Compounds 2,
3, 5, 9, and 10 had a low-micromolar activity against Bacillus
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
°
cane in butanone at 83 C for 72 h resulted in the formation of
compounds 1–3 in 9–40% yields. The reactions of 1-iodooctane
and 1-iododecane with 4-aminoquinaldine under similar reac-
tion conditions resulted in compounds 4 and 5 in 41 and 21%,
respectively. Compounds 6–9 (27–45% yields) were prepared
by nucleophilic aromatic substitution reactions of 4-chloroqui-
naldine with the respective diamines 1,4-diaminobutane, 1,9-
diaminononane, 1,10-diaminodecane, and 1,12-diaminodode-
°
cane at 120 C for 96 h (Figure 1B), where the chlorine atom was
displaced by an amino group. Finally, the reaction between 4-
Table 1. MIC values [μM] for compounds 1–10 and the control KAN against a series of E. coli, M. tuberculosis, S. aureus, and Staphylococcus epidermidis
bacterial strains.
Cpd#
Strain[a]
A
B
C
D
E
F
G
H
I
J
K
L
M
N
O
1
2
3
4
5
6
7
8
4
>64
16
8
32
16
>64
64
>64
32
16
0.5
>64
16
2
64
16
>64
64
64
32
16
32
2
4
16
1
>64
16
16
2
2
8
16
0.25
16
0.5
0.5
>64
32
32
4
1
0.5
64
4
32
1
1
16
2
>64
32
16
2
8
1
1
2
64
4
0.5
8
1
>64
64
32
4
64
4
1
64
>64
16
1
32
2
>64
64
64
16
8
8
4
0.25
0.25
8
1
64
2
4
0.5
2
0.5 (2)
0.25 (1)
4
4
>64
16
>64
1 (8)
2
0.5
0.5
4
0.5
0.5
8
0.5
>64
16
64
2
0.25 (4)
0.25 (1)
4
0.5
>64
16
16
2 (8)
2
1
>64
8
>64
64
64
32
4
1
1
16
>64
32
32
2
4
2
64
>64
>64
2
>64
>64
64
>64
4
9
10
KAN
8
0.5
2
4
4
2
1
>64
0.5 (16)
32
>64
>64
0.25
64
>64
0.5
>64 (16)
[a] Strains: A=E. coli MC1061 (outer-membrane hyperpermeablility); B=E. coli AG006 wt; C=E. coli ATCC BAA2469 (high resistance); D=M. tuberculosis
mc26230; E=S. aureus ATCC 6538; F=S. aureus ATCC 29213; G=S. aureus ATCC 25923; H=S. aureus C1; I=S. aureus G1; J=S. aureus G14; K=S. aureus NRS77;
L=S. aureus S14; M=S. aureus USA600/NRS22; N=S. epidermidis ATCC 12228; O=S. epidermidis ATCC 35984. Values in parenthesis for strains E and O are MIC
in tryptic soy broth (TSB)).
Table 2. MIC values [μM] for compounds 1–10 and the control KAN against a broad panel of bacterial strains where the compounds were mostly inactive.
Cpd#
Strain[a]
P
Q
R
S
T
U
V
W
X
Y
1
2
3
4
5
6
7
8
32
4
2
32
4
>64
>64
64
4
>64
64
32
64
16
>64
64
64
4
>64
>64
64
>64
64
32
>64
32
>64
>64
64
>64
16
>64
>64
>64
>64
NT
>64
>64
>64
NT
>64
>64
64
NT
NT
>64
>64
>64
NT
>64
64
64
64
64
>64
>64
NT
64
>64
>64
64
64
64
64
2
NT
NT
NT
NT
>64
>64
>64
>64
>64
>64
NT
>64
NT
>64
NT
>64
NT
>64
>64
64
NT
NT
NT
NT
64
64
64
4
9
10
KAN
>64
>64
64
>64
>64
16
>64
>64
32
>64
>64
32
4
0.25
16
2
1
[a] Strains: P=B. anthracis Sterne F32; Q=A. baumannii ATCC 19606; R=E. cloacae ATCC 13047; S=Klebsiella pneumoniae ATCC 27736; T=P. aeruginosa
ATCC 27853; U=P. aeruginosa 4; V=P. aeruginosa 6; W=P. aeruginosa 13; X=P. aeruginosa 94; Y=S. enterica ATCC 14028. NT=not tested.
ChemMedChem 2021, 16, 1–11
www.chemmedchem.org
3
© 2021 Wiley-VCH GmbH
��
These are not the final page numbers!

Full Papers
doi.org/10.1002/cmdc.202100001
ChemMedChem
anthracis (strain P), and compound 9 additionally had a low-
micromolar activity against A. baumannii (strain Q). We
compared the effects of different structural features of com-
pounds 1–10 on their antibacterial activity against bacteria in
Table 1.
structural types (e.g., single-headed, double-headed, charged,
and neutral).
1
2
3
4
5
Effect of compounds 2, 3, and 9 on biofilms formation and
disruption
Double-headed cationic compounds 1, 2, and 3 generally
followed a trend of inhibitory potencies 1!2<3 against most
of the E. coli and staphylococci tested. The potency increased
dramatically (~16-fold) with the linker length increasing from
six to eight carbon atoms in compounds 1 and 2, respectively,
and it increased further approximately fourfold, when the linker
was lengthened to ten carbon atoms in compound 3. As a
consequence, compounds 2 and 3 displayed sub-micromolar
MIC values against some bacteria. We concluded that the 6-
carbon linker of compound 1 (MIC values 2 to >64) was too
short for this compound to function by the same mechanism as
compounds 2 and 3. Next we examined the effects of double-
headed neutral compounds 6–9. These molecules displayed an
analogous trend of increasing inhibitory potency with increas-
ing linker length to that observed for the cationic compounds:
6<7=8<9. Compound 6 with a four-carbon linker bridging
exocyclic amino groups had little or no activity. Notably, despite
effectively longer linkers in these compounds than in the
cationic compounds 1–3, compounds 6–9 were not as active. A
significant activity improvement was observed between the
ten-carbon linker in compound 8 and a 12-carbon linker in
compound 9 (achieving low-micromolar MIC values), but even
compound 9 was not as active as compound 2, which
contained a much shorter, eight-carbon linker. The similar trend
of increasing potency with increasing tail length was observed
in the single-headed cationic compounds 4 and 5, but these
single-headed compounds were not as potent as their double-
headed counterparts 2 and 3. We conclude that the cationic
compounds that are double-headed and contain long linkers
are most potent. A caveat of this conclusion is the inevitable
structural difference in the linkers and in the disposition of the
methyl group relative to the linker between the cationic and
neutral compounds. Therefore, these structural differences, and
not the charge, or in addition to the charge, may contribute to
or be responsible for the differences in potencies between the
two sets of compounds.
6
7
8
9
To extend our understanding of the effect of these compounds
on bacteria, we investigated whether they disrupted preformed
biofilms and whether they inhibited biofilm formation. For
these studies, we chose to examine the effects of compounds 2,
3, and 9 on the biofilm formed by two strains, S. aureus ATCC
6538 (strain E) and S. epidermidis ATCC 35984 (strain O), against
which these compounds displayed low- or sub-micromolar
potencies in the MIC assays (Table 1). The medium used in the
biofilm assays was much richer than that used in the MIC assay;
therefore, we repeated MIC measurements in this new medium.
The MIC values of compounds 2, 3 and 9 for strains E and O
grown at these new conditions were four to eight times higher
than in the previous medium (Table 1; values in parentheses).
Neither compounds 2, 3, 9, nor KAN had any significant effect
on reducing the biomass, as measured by crystal violet, of
preformed biofilm for the two staphylococci tested (Figure 2A,
B). The inactivity of the compounds against the preformed
biofilm could be explained by inadequate biofilm penetration,
by bacterial adaptation to stresses in the biofilm, or by the
presence of persister cells.[18] The same compounds were tested
to examine if the formation of biofilm biomass was blocked. All
compounds had a significant inhibitory effect on the biomass
formed by S. aureus ATCC 6538 (strain E; Figure 2C). Compound
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
9, whose MIC was 8 μM against strain E, exerted a strong
1
activity against biofilm formation, inhibiting it tenfold at = ×
4
MIC (2 μM) and completely at 1= ×MIC (4 μM). Similarly,
2
compound 2 inhibited biofilms formed by strain E by 25% at
1
1
= ×MIC (0.5 μM) and completely at = ×MIC (1 μM). Compound
4
2
3
also displayed
a
significant activity, inhibiting biofilm
1
formation by this strain fivefold at = ×MIC (0.25 μM) and 20-
fold at = ×MIC (0.5 μM). The control antibiotic, KAN, which was
4
1
2
relatively weakly active against strains E and O (MIC=16 μM),
also had a significant biofilm inhibition behavior relative to its
1
1
MIC: the inhibition was 2-fold at = ×MIC and complete at = ×
4
2
For single-headed compounds 5 and 10 with a ten-carbon
tail, the differences in charge and structure did not have as
large of an effect on antibacterial potency as for their double-
headed counterparts 3 and 7. The MIC values for compounds 5
(MIC=0.5–8 μM) and 10 (MIC=1–16 μM) were often within
twofold dilution. The preference for the double-headed over
the single-headed structure for the cationic compounds was
reversed for neutral compounds, as observed when comparing
neutral compounds 8 and 10, each with a ten-carbon linker and
tail, respectively. The neutral single-headed compound 10
(MIC=1–16 μM) was more potent than its double-headed
counterpart 8 (MIC=4 to >64 μM).
MIC. Formation of the biofilm by S. epidermidis ATCC 35984
(strain O) was highly resistant to inhibition by compounds 3
and 9, whereas compound 2 exhibited substantial inhibition
activity (Figure 2D). Compound 2 completely inhibited biofilm
1
formation at = ×MIC (1 μM). KAN weakly inhibited biofilm
4
1
formation, displaying eightfold inhibition at = ×MIC, but no
2
1
measurable inhibition at = ×MIC. To determine if this is due to
4
an effect on the biomass production of the biofilm rather than
a significantly reduced bacterial load caused by sub-MIC value
concentration of the drugs, the colony forming units (CFUs)
were counted. It has been published that the biomass produced
does not always coincide with the number of viable cells.[19]
Control CFU measurements showed that the observed inhib-
itory effect of the compounds on the biofilm could not be
simply attributed to cell growth inhibition, as the scale of the
CFU value changes was different from that of the biofilm
Taking all this information into account, we proceeded to
other cell-based biological studies with compounds 2, 3, 5, 9,
and 10 that were the most potent compounds of different
ChemMedChem 2021, 16, 1–11
www.chemmedchem.org
4
© 2021 Wiley-VCH GmbH
��
These are not the final page numbers!

Full Papers
doi.org/10.1002/cmdc.202100001
ChemMedChem
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
Figure 2. Graphs showing the effects of treatment with select compounds on preformed biofilms of A) S. aureus ATCC 6538 (strain E) and B) S. epidermidis
ATCC 35984 (strain O), and prevention of biofilm formation in C) S. aureus ATCC 6538 (strain E) and D) S. epidermidis ATCC 35984 (strain O). Colony forming
units (CFUs) are noted over the bars for compound 3 and KAN in (C), where a reduction in biofilm was seen at a concentration lower than the MIC for that
strain. Dashed vertical lines indicate the MIC values of the compounds with the bacteria in the TSB medium. The MIC values are indicated by the vertical
dashed lines color-coded analogously to the bars.
formation. For example, for S. aureus ATCC 6538 (strain E), 7.0×
evaluated by dose-response measurements using the initial
coupled primase-inorganic pyrophosphatase assay (Table 3 and
Figures S31–S34). The potencies of the compounds against the
two primases were correlated. For example, compound 2, a
two-headed cationic compound with an eight-carbon linker,
was the most potent compound against both SaDnaG and
MtbDnaG, displaying IC50 values of 2.5�0.2 and 15�2 μM,
respectively. The IC50 values were in the low-micromolar range
for the most potent compounds when tested against SaDnaG,
but they were two to six times higher against MtbDnaG; this is
consistent with SaDnaG being the molecular target used in the
initial screening assay. This result also indicated that the
inhibition activity was species selective. Because of the much
higher potency of the compounds against SaDnaG, we will
focus the description of the SAR for this target.
1010, 1.3×1010, 7.5×109 CFU (one-, five-, and tenfold reduction)
were measured for the concentrations of KAN of 0, 1= ×MIC and
4
1
= ×MIC, respectively (Table S1). The corresponding one-, two-,
2
and >100-fold reductions in the biofilm biomass at the
respective KAN concentrations (Figure 2C, Table S1) do not
scale the same way as the CFU values. However, one can argue
that for biofilm formation the cells need to grow to a
sufficiently high CFU value. In the example given above, the
1
10% cell at growth at the KAN concentration of = ×MIC may
2
not be sufficient for any biofilm formation. On the other hand,
the observation that different compounds affect biofilm for-
mation differently with respect to the MIC values, suggests that
the effect on the biofilm formation is not a trivial consequence
of cell growth inhibition.
The potency of the compounds as SaDnaG inhibitors
generally displayed a very well-defined SAR. Two-headed
cationic compounds 1–3 all had low-micromolar potency
against this primase regardless of linker length or charge. One-
headed compounds 4, 5, and 10 lacked potency independent
of linker length or charge. The most complex SAR was observed
DnaG inhibition and DNA intercalation
Inhibitory potencies of all of the analogues (1–10) against two
homologous primases, SaDnaG as well as MtbDnaG, were
ChemMedChem 2021, 16, 1–11
www.chemmedchem.org
5
© 2021 Wiley-VCH GmbH
��
These are not the final page numbers!

Full Papers
doi.org/10.1002/cmdc.202100001
ChemMedChem
Table 3. DnaG inhibition and DNA intercalation potencies of compounds 1–10.
1
2
3
Cpd#
DnaG inhibition
SaDnaG
DNA intercalation
ssDNA
MtbDnaG
IC50 (μM)
dsDNA
IC50 (μM)
IC50 (μM)
Hill coefficient (h)
Hill coefficient (h)
IC50 (μM)
Hill coefficient (h)
Hill coefficient (h)
4
1
2
3
4
5
6
7
8
9
5.1�0.7
2.5�0.2
4.3�0.2
>100
>100
>100
1
1
1
1
1
1
1
1
2
1
36�3
15�2
19�1
>100
>100
>100
26�1
18�1
25�2
>100
1
1
2
1
1
1
2
2
2
1
16.4�0.5
40�3
2
2
2
1
1
1
1
1
2
2
>100
>100
75�3
>100
>100
>100
76�3
96�11
48�5
>100
1
1
2
1
1
1
2
2
2
1
5
6
7
8
11.0�0.9
>100
86�14
>100
9
14�1
23�3
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
5.2�0.4
17.0�0.1
94�20
13.9�0.7
10.6�0.7
66�5
10
for two-headed neutral compounds 6–9. Compound 6 with the
shortest, four-carbon, linker lacked potency, compounds 7 and
9 with a nine-carbon and a 12-carbon linker, respectively, were
modestly potent (IC50 ~15 μM), whereas compound 8 with a
ten-carbon linker was potent, with IC50 =5.2�0.4 μM. Therefore,
in the absence of the positive charge, the inhibitory potency
was sensitive to linker length, with the maximum potency
observed for linker length of about ten carbons.
Next, we asked whether the observed DnaG inhibition could
be a result of DNA intercalation, or whether it involved
inhibitor-DnaG interactions. Because DnaG catalyzes RNA syn-
thesis on single-stranded DNA as a template, we have tested
intercalation of compounds 1–10 into both single- and double-
stranded (ss and ds) 25-bp oligomer DNA with an arbitrary
sequence (Table 3 and Figures S35–S38). We investigated how
the compounds competed with a known DNA intercalator,
ethidium bromide (EtBr), at a 2:1 molar ratio of EtBr:DNA (on a
molecule basis for DNA), by monitoring changes in EtBr
fluorescence upon titrating the compounds. All compounds
very weakly intercalated into dsDNA (IC50 >48 μM), in agree-
performed with EtBr, as a control (Figure 3C, D), in the
concentration range where DNA was being saturated by EtBr.
Neither 2 nor EtBr displayed a significant CD signal in the
absence of DNA, as expected, based on their lack of chirality.
The spectrum for compound 2-dsDNA complex (Figure 3A)
exhibited significant induced CD signal at 255 and 350 nm and
a blue shift of the negative peak from ~250 to 240 nm. The
negative peak became narrower. These spectral changes were
not observed for EtBr-dsDNA complex (Figure 3C), for which an
induced signal at 310 nm characteristic of EtBr intercalation into
dsDNA was observed.[20] Both compound 2 and EtBr resulted in
the similar increase of the positive peak at 280 nm; although at
the highest concentration of 120 μM of compound 2, the height
of this peak was similar to that of unbound DNA. Both
compounds increased the magnitude of the negative peak to a
similar extent. These data indicate that even though compound
2 likely intercalates into dsDNA, it probably has a more complex
mode of binding, potentially involving groove binding, as
judged by its pronounced induced CD.[21] For ssDNA, we
observed that the CD across the entire examined wavelength
range became substantially weakened upon titrating com-
pound 2 (Figure 3B), indicating the loss of DNA helicity. A
similar induced CD was observed at 255 nm, but not at 350 nm,
disappearing at higher concentration of 2, in agreement with
the loss of helicity. Since the helicity of ssDNA is expected to be
critically dependent on base stacking, and it would be disrupted
by intercalation, these data are consistent with intercalation as
a dominant mode of compound 2, DNA interactions. EtBr also
caused some weakening of the positive signal of ssDNA,
indicative of intercalation, but, in contrast with compound 2,
EtBr strengthened and blue-shifted the negative signal (Fig-
ure 3D). Additionally, a strong induced CD peak appeared at
310 nm, whereas an induced peak was much weaker with
dsDNA. These data indicate that EtBr may have an alternative or
additional mode of interaction with ssDNA, which is likely also
present with dsDNA, but to a lesser extent. Taken together,
these observations are consistent with the greater propensity of
compound 2 to displace EtBr from ssDNA than from dsDNA.
The preference for intercalation into ssDNA could be a
structural consequence of the flexible linker connecting two
DNA base-sized ring systems. Such structure can adapt to a
relatively greater flexibility of ssDNA over dsDNA. The two
ment with
a
previous study of dequalinium-dsDNA
interactions.[14] On the other hand, we observed significant
ssDNA intercalation with IC50 values as low as 10 μM, which
strongly correlated with DnaG inhibition activity. Two-headed
compounds were strong ssDNA intercalators regardless of linker
length or charge, whereas single-headed compounds were
intercalating into ssDNA similarly weakly to dsDNA. These
results strongly suggest that both head groups of the two-
headed compounds are engaged with ssDNA, that is, these
compounds are bisintercalators. The intrinsic ssDNA intercala-
tion Kd values for these compounds are likely to be ten to
20 times lower than the observed IC50 values in the intercalation
assay for the following reason. Because binding to most DNA
binding sites in the assay was not competitive with EtBr (which
could occupy at most two sites per oligomer), such binding
could not be observed. In contrast, most DNA binding events
likely inhibited primer synthesis at initiation or elongation
phases, resulting in full or partial inhibition of DnaG activity.
In order to probe DNA interactions with the compounds
without a competitor, we examined the effect of binding of
compound 2 to the same ds- and ssDNA by circular dichroism
(CD; Figure 3A, B, respectively). Similar CD measurements were
ChemMedChem 2021, 16, 1–11
www.chemmedchem.org
6
© 2021 Wiley-VCH GmbH
��
These are not the final page numbers!

Full Papers
doi.org/10.1002/cmdc.202100001
ChemMedChem
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
Figure 3. CD spectra of A) dsDNA and B) ssDNA with compound 2. CD spectra of C) dsDNA and D) ssDNA with EtBr. The concentrations of DNA and the
compounds are indicated in the insets.
aminoquinoline head groups may sandwich a base or stack into
non-adjacent gaps between the bases, thereby disrupting the
intrastrand base stacking of the ssDNA. Another known
example of an intercalator with a preference for ssDNA is a
fluorescent dye thiazole orange homodimer, which is also a
symmetrical structure containing two cationic chromophores
connected by a flexible linker.[22] On the other hand, ds
intercalating small molecules, such as ethidium bromide or
anthracyclines, are often large rigid ring systems that span a
base pair.
Apparently, ssDNA bisintercalation contributes significantly
to SaDnaG inhibition activity of most two-headed compounds.
However, because these two inhibition activities are not
perfectly correlated, at least some compounds must also
directly interact with SaDnaG. For example, compounds 2 and
8, which stood out as the most potent cationic and neutral two-
headed SaDnaG inhibitors, respectively, were not the strongest
ssDNA intercalators. The differences in inhibitory potencies of
the compounds against SaDnaG and MtbDnaG are also
consistent with enzyme-specific interactions.
Mechanistic insight about the mode(s) of action of com-
pounds 1–10 in the S. aureus cell can be gained from
comparing the MIC values in Table 1 for S. aureus strains E–M
with SaDnaG inhibition and DNA intercalation potencies in
Table 3. Cationic compounds 2 and 3 were most active in both
inhibiting cell growth and inhibiting DnaG, however some S.
aureus strains (E and F) were much more sensitive to inhibition
by these compounds than SaDnaG was. Similarly, the MICs of
compounds 2 and 3 for M. tuberculosis were six- to seven times
smaller than the IC50 values for inhibition of MtbDnaG. The
correlation of cell growth and SaDnaG inhibition by compounds
2 and 3 indicates that, for those S. aureus strains whose cell
growth and DnaG are similarly sensitive to these compounds,
DnaG inhibition contributes to cell growth inhibition. This is
apparently not the case for neutral double-headed compounds,
among which the most potent SaDnaG inhibitor 8 has a very
weak antibacterial activity. Another mechanism must also play a
role in the inhibition of bacterial growth.
ChemMedChem 2021, 16, 1–11
www.chemmedchem.org
7
© 2021 Wiley-VCH GmbH
��
These are not the final page numbers!

Full Papers
doi.org/10.1002/cmdc.202100001
ChemMedChem
The effect of compounds 3, 5, 9, and 10 on cell permeability
perturb the cell membrane of bacteria and fungi, and DMSO, as
a negative control.[23] After the treatment with the compounds,
the cells were washed and incubated in the presence of
propidium iodide (PI). When the cell membrane is compro-
mised, PI passes into the cytosol of the bacterium and
intercalates into DNA, where it can be detected by its character-
istic red fluorescence. As expected for non-disruptive com-
pounds, treatment with DMSO or KAN, at concentrations of 1×
MIC and 4×MIC, resulted in little to no observed fluorescence
(Figure 4). Compounds 3, 5, and 9 at a concentration of 1×MIC
resulted in increased PI fluorescence. The percentage of
fluorescent cells is comparable to that observed with the
positive control, C14-TOB (71% cell staining). At concentrations
of 1×MIC compounds 3, 5 and 9 resulted in 91, 71, and 87%
cell staining, respectively. Relatively low levels of cell staining
observed when these compounds were used at 4×MIC (21, 8,
and 12%, respectively) was likely due to the death of the
compromised cells. In contrast, compound 10 did not cause cell
staining at 1×MIC, but it resulted in a significant increase in the
1
2
3
4
5
6
7
8
9
Cell membrane disruption has been observed for dequalinium
(3)[15] and other compounds with long alkyl chains.[23] Cationic
amphiphilic compounds could bind the membrane phospholi-
pids by the alkyl linker or tail, while establishing a salt bridge
between the positively charged amino group and the neg-
atively charged phosphate group of the lipid. Because DnaG
inhibition and DNA intercalation could not fully account for the
activity of the compounds against the bacterial cell, we asked if
the antibacterial activity was due to the ability of the
compounds to affect the bacterial cell membrane. As we
explained in above sections, compounds 3, 5, 9, and 10 were
selected to represent the double-headed charged (3), single-
headed charged (5), double-headed neutral (9), and single-
headed neutral (10) types. S. epidermidis ATCC 35984 (strain O)
was treated with these four compounds as well as with three
additional compounds: KAN, a previously published C14 thio-
ether tobramycin derivative (C14-TOB) that was shown to
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
Figure 4. A) Pictures showing the bright field, fluorescent, and overlaid images of S. epidermidis ATCC 35984 (strain O) cells treated sequentially with noted
compounds and PI. B) Graph representing the quantification of the fluorescent cells relative to the total cells.
ChemMedChem 2021, 16, 1–11
www.chemmedchem.org
8
© 2021 Wiley-VCH GmbH
��
These are not the final page numbers!

Full Papers
doi.org/10.1002/cmdc.202100001
ChemMedChem
number of fluorescent cells when used at 4×MIC (57% cell
staining). This data indicates that the mechanism of action of
these compounds involves membrane disruption.
Hemolytic activity
1
2
3
Given the ssDNA intercalation and bacterial membrane activ-
ities of the compounds, we sought to probe their toxicity
against mammalian cells. We investigated the hemolytic activity
of compounds 2, 3, 5, 9 and 10 using murine red blood cells
(mRBCs; Figure 5). All five compounds at 1–2 μM displayed low
levels of hemolysis of less than 10%. At higher concentrations
compounds 3, 9, and 10 caused significantly more hemolysis
than compounds 2 and 5 (Figure 4). The concentrations at
which significant hemolysis was observed for compounds 3, 9
and 10 were eight to 16 times higher than the MIC values of
these compounds against sensitive bacteria. Compound 2,
whose low hemolytic activity was similar to that of KAN,
emerged as the most promising compound for further develop-
ment as a novel antibacterial. Notably, the clinically used
dequalinium (3) was much more hemolytic than compound 2.
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
Cytotoxicity in mammalian cells
To consider potential systemic use of these compounds, we
tested cytotoxicity of compounds 2, 3, 5, 9, and 10 to four
mammalian cell lines: human lung carcinoma (A549, Figure 6A),
Figure 5. Graph showing the hemolytic activity of selected compounds using
murine RBCs. TritonTM X-100 (TX) was used as a positive control.
Figure 6. Graphs showing the cytotoxicity of selected compounds against four mammalian cell lines normalized at 100%. A) A549, human lung carcinoma,
B) HEK-293, human embryonic kidney, C) BEAS-2B, normal human bronchial epithelial cells, and D) J774A.1, mouse macrophages. Note: The corresponding
non-normalized graphs are presented in Figure S39.
ChemMedChem 2021, 16, 1–11
www.chemmedchem.org
9
© 2021 Wiley-VCH GmbH
��
These are not the final page numbers!

Full Papers
doi.org/10.1002/cmdc.202100001
ChemMedChem
[1] R. M. Klevens, J. R. Edwards, C. L. Richards, Jr., T. C. Horan, R. P. Gaynes,
D. A. Pollock, D. M. Cardo, Public Health Rep. 2007, 122, 160–166.
[2] G. E. Croston, Expert Opin. Drug Discovery 2017, 12, 427–429.
[3] a) A. Robinson, R. J. Causer, N. E. Dixon, Curr. Drug Targets 2012, 13,
352–372; b) G. Sanyal, P. Doig, Expert Opin. Drug Discovery 2012, 7, 327–
339.
[4] a) D. Karamata, J. D. Gross, Mol. Gen. Genet. 1970, 108, 277–287; b) A.
van der Ende, T. A. Baker, T. Ogawa, A. Kornberg, Proc. Natl. Acad. Sci.
USA 1985, 82, 3954–3958.
[5] A. Kuron, M. Korycka-Machala, A. Brzostek, M. Nowosielski, A. Doherty,
B. Dziadek, J. Dziadek, Antimicrob. Agents Chemother. 2014, 58, 1699–
1706.
[6] A. Agarwal, S. Louise-May, J. A. Thanassi, S. D. Podos, J. Cheng, C.
Thoma, C. Liu, J. A. Wiles, D. M. Nelson, A. S. Phadke, B. J. Bradbury, M. S.
Deshpande, M. J. Pucci, Bioorg. Med. Chem. Lett. 2007, 17, 2807–2810.
[7] M. Singh, S. Ilic, B. Tam, Y. Ben-Ishay, D. Sherf, D. Pappo, B. Akabayov,
Chem. Eur. J. 2020, 26, 10849–10860.
[8] a) T. Biswas, K. D. Green, S. Garneau-Tsodikova, O. V. Tsodikov, Biochem-
istry 2013, 52, 6905–6910; b) T. Biswas, E. Resto-Roldan, S. K. Sawyer, I.
Artsimovitch, O. V. Tsodikov, Nucleic Acids Res. 2013, 41, e56.
[9] S. R. Choi, M. A. Larson, S. H. Hinrichs, P. Narayanasamy, Bioorg. Med.
Chem. Lett. 2016, 26, 1997–1999.
[10] C. Gajadeera, M. J. Willby, K. D. Green, P. Shaul, M. Fridman, S. Garneau-
Tsodikova, J. E. Posey, O. V. Tsodikov, J. Antibiot. (Tokyo) 2015, 68, 153–
157.
human embryonic kidney cells (HEK-293, Figure 6B), human
bronchial epithelial cells (BEAS-2B, Figure 6C), and mouse
macrophages (J774A.1, Figure 6D). Compound 2 was the only
molecule that was not significantly toxic to all four cell lines up
to 8 μM. In fact, compound 2 showed significant toxicity only
against the J774A.1 cell line, at higher concentrations. These
data also support further development of compound 2.
1
2
3
4
5
6
7
8
9
Conclusion
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
In this study, we have screened known bioactive molecules for
their inhibitory potency against SaDnaG. Dequalinium (3), an
antiseptic agent was identified as a low-micromolar SaDnaG
inhibitor. By investigating a series of dequalinium and its
analogues, we established that compounds in this series are
single-stranded DNA bisintercalators, which also compromise
the bacterial membrane. Some of these compounds likely
interact with DnaG directly, as DNA intercalation alone cannot
account for the observed potent DnaG inhibition. Even though
dequalinium itself was, unsurprisingly, substantially hemolytic
and toxic to mammalian cells, one of its analogues, compound
2, was not only highly potent, but it exhibited low mammalian
cytotoxicity and hemolytic activity. It is possible that the
relatively weak DNA intercalation propensity of this compound
may be the reason for its low toxicity against mammalian cells,
whereas the mechanism of the membrane disruption by this
and other compounds may be somewhat specific to the
bacterial membrane. Further medicinal chemistry optimization
starting with compound 2 and preclinical studies may yield a
novel antibacterial agent for clinical use.
[11] B. T. Glaser, J. P. Malerich, S. J. Duellman, J. Fong, C. Hutson, R. M. Fine,
B. Keblansky, M. J. Tang, P. B. Madrid, J. Biomol. Screening 2011, 16, 230–
238.
[12] N. Shahabadi, A. A. Asadian, M. Mahdavi, Nucleosides Nucleotides Nucleic
Acids 2017, 36, 676–689.
[13] J. H. Zhang, T. D. Chung, K. R. Oldenburg, J. Biomol. Screening 1999, 4,
67–73.
[14] R. G. Wright, L. P. Wakelin, A. Fieldes, R. M. Acheson, M. J. Waring,
Biochemistry 1980, 19, 5825–5836.
[15] W. B. Hugo, M. Frier, Appl. Microbiol. 1969, 17, 118–127.
[16] a) C. Abeywickrama, S. A. Rotenberg, A. D. Baker, Bioorg. Med. Chem.
2006, 14, 7796–7803; b) D. Galanakis, C. A. Davis, C. R. Ganellin, P. M.
Dunn, J. Med. Chem. 1996, 39, 359–370.
[17] a) J. N. Pendleton, S. P. Gorman, B. F. Gilmore, Expert Rev. Anti-Infect.
Ther. 2013, 11, 297–308; b) D. M. P. De Oliveira, B. M. Forde, T. J. Kidd,
P. N. A. Harris, M. A. Schembri, S. A. Beatson, D. L. Paterson, M. J. Walker,
Clin. Microbiol. Rev. 2020, 33, e00181–00119; c) World Health Organ-
ization, 2017.
[18] P. S. Stewart, Int. J. Med. Microbiol. 2002, 292, 107–113.
[19] C. D. Cruz, S. Shah, P. Tammela, BMC Microbiol. 2018, 18, 173.
[20] C. H. Huang, R. Baserga, Nucleic Acids Res. 1976, 3, 1857–1873.
[21] N. C. Garbett, P. A. Ragazzon, J. B. Chaires, Nat. Protoc. 2007, 2, 3166–
3172.
[22] H. S. Rye, A. N. Glazer, Nucleic Acids Res. 1995, 23, 1215–1222.
[23] a) I. M. Herzog, K. D. Green, Y. Berkov-Zrihen, M. Feldman, R. R. Vidavski,
A. Eldar-Boock, R. Satchi-Fainaro, A. Eldar, S. Garneau-Tsodikova, M.
Fridman, Angew. Chem. Int. Ed. 2012, 51, 5652–5656; Angew. Chem.
2012, 124, 5750–5754; b) S. K. Shrestha, M. Y. Fosso, K. D. Green, S.
Garneau-Tsodikova, Antimicrob. Agents Chemother. 2015, 59, 4861–4869;
c) M. Y. Fosso, S. K. Shrestha, N. Thamban Chandrika, E. K. Dennis, K. D.
Green, S. Garneau-Tsodikova, Molecules 2018, 23, E899.
Acknowledgements
We thank Mari Johnson for assistance with initial DNA intercala-
tion experiments and Dr. Trevor Creamer for sharing the CD
spectrometer.
Conflict of Interest
The authors declare no conflict of interest.
Keywords: antibacterial agents · DNA intercalators · library
screening · medicinal chemistry · Staphylococcus aureus
Manuscript received: January 1, 2021
Revised manuscript received: March 2, 2021
Accepted manuscript online: March 12, 2021
Version of record online: ■■■, ■■■■
ChemMedChem 2021, 16, 1–11
www.chemmedchem.org
10
© 2021 Wiley-VCH GmbH
��
These are not the final page numbers!

FULL PAPERS
1
2
3
4
5
6
7
8
9
The topical antibacterial agent de-
qualinium is a potent inhibitor of S.
aureus primase DnaG even in methi-
cillin-resistant bacteria. Mechanistic
studies of dequalinium analogues
show them to be ssDNA bisintercala-
tors that enter the bacterium by
compromising its membrane. The
best analogue likely interacts with
DnaG directly, inhibits bacterial
growth and biofilm formation, and
has no significant toxicity to
Dr. K. D. Green, Dr. A. Punetha,
Dr. N. T. Chandrika, Dr. C. Hou, Prof. S.
Garneau-Tsodikova*, Prof. O. V.
Tsodikov*
1 – 11
Development of Single-Stranded
DNA Bisintercalating Inhibitors of
Primase DnaG as Antibiotics
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
mammalian cells.