Small molecule fluoride toxicity agonists

Article abstract of DOI:10.1016/j.chembiol.2015.03.016

Fluoride is a ubiquitous anion that inhibits a wide variety of metabolic processes. Here, we report the identification of a series of compounds that enhance fluoride toxicity in Escherichia coli and Streptococcus mutans. These molecules were isolated by u

Full text of DOI:10.1016/j.chembiol.2015.03.016

Article
Small Molecule Fluoride Toxicity Agonists
Highlights
Authors
d
d
d
A riboswitch was used to report increased fluoride levels
inside bacteria
James W. Nelson, Mark S. Plummer, ...,
Tyler D. Ames, Ronald R. Breaker
Molecules that trigger riboswitch reporter activity were
isolated via HTS
Correspondence
ronald.breaker@yale.edu
Improved hit compounds enhance fluoride toxicity in bacteria
In Brief
Fluoride toxicity mitigation systems are
widespread in nature. Nelson et al. have
developed a high-throughput small
molecule screen to identify compounds
that enhance the innate toxicity of fluoride
in bacteria. Compounds derived from the
initial hits exhibit improved function,
which suggests a route to creating novel
fluoride-mediated antibacterial agents.
Nelson et al., 2015, Chemistry & Biology 22, 527–534
April 23, 2015 ª2015 Elsevier Ltd All rights reserved
http://dx.doi.org/10.1016/j.chembiol.2015.03.016

Chemistry & Biology
Article
Small Molecule Fluoride Toxicity Agonists
James W. Nelson,1 Mark S. Plummer,2 Kenneth F. Blount,2 Tyler D. Ames,2,5 and Ronald R. Breaker2,3,4,
*
1Department of Chemistry, Yale University, New Haven, CT 06520, USA
2Department of Molecular, Cellular and Developmental Biology, Yale University, New Haven, CT 06520, USA
3Howard Hughes Medical Institute, Yale University, New Haven, CT 06520, USA
4Department of Molecular Biophysics and Biochemistry, Yale University, New Haven, CT 06520, USA
5Present address: Phosplatin Therapeutics, New York, NY, USA
*Correspondence: ronald.breaker@yale.edu
http://dx.doi.org/10.1016/j.chembiol.2015.03.016
SUMMARY
termed crcB (also called fex or fluc), codes for a fluoride-specific
channel protein (Baker et al., 2012; Li et al., 2013; Stockbridge
Fluoride is a ubiquitous anion that inhibits a wide et al., 2013). Another commonly controlled gene, eriCF, encodes
a fluoride-selective antiporter (Baker et al., 2012; Stockbridge
et al., 2012). Numerous other genes presumably involved in miti-
variety of metabolic processes. Here, we report the
identification of a series of compounds that enhance
fluoride toxicity in Escherichia coli and Strepto-
coccus mutans. These molecules were isolated by
using a high-throughput screen (HTS) for com-
pounds that increase intracellular fluoride levels as
determined via a fluoride riboswitch reporter fusion
construct. A series of derivatives were synthesized
gating fluoride toxicity are also associated with fluoride ribo-
switches (Baker et al., 2012).
An Escherichia coli strain in which the fluc gene was deleted
(Dfluc) is ꢀ200-fold more sensitive to fluoride (Baker et al.,
2012). Similarly, some eukaryotic species use homologous
channel proteins to improve resistance to fluoride toxicity (Li
et al., 2013), which underscores the importance of fluc to fluoride
detoxification. Moreover, the deleterious phenotype resulting
from fluc deletion in E. coli can be compensated by expressing
to examine structure-activity relationships, leading
to the identification of compounds with improved ac-
tivity. Thus, we demonstrate that small molecule an eriCF gene derived from another bacterium (Baker et al.,
2012), indicating that the proteins expressed from eriCF appear
to be biologically equivalent to those encoded by fluc. These
fluoride toxicity agonists can be identified by HTS
from existing chemical libraries by exploiting a natu-
ral fluoride riboswitch. In addition, our findings sug-
gest that some molecules might be further optimized
findings demonstrate that a major strategy for overcoming
toxicity is for cells to eject fluoride into the environment.
We speculated that small molecules that specifically inhibit the
to function as binary antibacterial agents when com-
protein products Fluc or EriCF, increase the membrane perme-
bined with fluoride.
ability of fluoride, or increase fluoride retention through some
other mechanism could enhance fluoride toxicity. If such mole-
cules could be found, then perhaps they could be optimized
and made to function as novel antibacterial chemotherapeutics,
INTRODUCTION
Although fluoride is commonly added to oral hygiene products to particularly when used in combination with high fluoride concen-
increase the strength of tooth enamel, it also has substantial anti- trations. This speculation is supported by the fact that some pre-
bacterial effects (Barbier et al., 2010; Li, 2003). For example, viously characterized pore-forming antibacterial and antifungal
when in complex with a divalent metal ion and ADP, fluoride compounds have been demonstrated to enhance the cytotoxic
forms a nonfunctional mimic of ATP that can inhibit enolase activity of fluoride (Li and Breaker, 2012; Nelson et al., 2014;
(Curran et al., 1994; Qin et al., 2006), and similar complexes Zasloff and Steinberg, 1993). However, these existing molecules
can inhibit a large diversity of other metabolic enzymes including only modestly increase fluoride toxicity, and no non-peptidic
phosphatases (Barbier et al., 2010; Nakai and Thomas, 1974). molecules have been identified that increase fluoride toxicity
Due to these general mechanisms for enzyme inhibition, fluoride toward bacteria.
is toxic to organisms from all three domains of life.
To identify novel fluoride toxicity agonists for bacteria, we
Until recently, very little was known about how cells sense and developed an HTS strategy based on an E. coli strain in which
respond to fluoride toxicity. New insights into the genes and a b-galactosidase reporter gene controlled by a fluoride ribo-
mechanisms used by bacterial cells to overcome fluoride toxicity switch provides a readout of intracellular fluoride concentration
were revealed by the discovery of fluoride-responsive ribo- (Baker et al., 2012; Nelson et al., 2014). A number of studies
switches that control the expression of a number of genes have been conducted to identify compounds that directly target
conferring resistance to fluoride (Baker et al., 2012). Ribo- riboswitches and function as antibiotics (Blount and Breaker,
switches are structured RNAs that are typically found in the 50- 2006; Deigan and Ferre-D’Amare, 2011; Kim et al., 2009; Lee
UTRs of bacterial mRNAs where they regulate the expression et al., 2009; Lu¨ nse et al., 2014a, 2014b; Mulhbacher et al.,
of genes in response to binding a small molecule or ion (Breaker, 2010; Ster et al., 2013). By contrast, our study is designed to
2011; Peselis and Serganov, 2014; Serganov and Nudler, 2013). exploit a riboswitch as a tool to identify compounds that target
The most common gene associated with fluoride riboswitches, other aspects of bacterial physiology. Hit compounds from the
´
´
Chemistry & Biology 22, 527–534, April 23, 2015 ª2015 Elsevier Ltd All rights reserved 527

(Baker et al., 2012; Nelson et al., 2014). Whereas most ribo-
switches control gene expression by regulating the formation
of an intrinsic transcription terminator stem or by regulating ac-
cess to the ribosome binding site of the mRNA, the precise
mechanism by which this specific riboswitch representative con-
trols gene expression has not been established. The relative
amount of fluoride present in cells can be established by quanti-
tatively measuring reporter gene expression. This is achieved by
the addition of 4-methyl-umbelliferyl-b-D-galactopyranoside
(4-MUG), which is a proto-fluorescent b-galactosidase sub-
strate. A Dfluc E. coli strain containing the riboswitch reporter
was utilized as a positive control for cells that uptake or retain
an unusually high amount of fluoride. Reporter gene expression
in the Dfluc E. coli strain was 15- to 20-fold higher compared with
the WT strain.
S
O
O
S
1
2
3
4
5
N
O
F3C
Cl
N
H
N
H
N
H
N
H
O
HO
CF3
Br
O
O
O
N
H
CF3
F3C
HTS assays were performed using 1 mM fluoride, which is
considerably higher than the concentration of fluoride present
in most natural sources of water. However, this fluoride concen-
tration is far below that tolerated by WT E. coli cultures, which
can still grow at concentrations as high as 100 mM (Baker
et al., 2012). Although fluoride concentrations in the millimolar
range might seem high, amounts are known to vary widely in
the environment, and local concentrations will increase greatly
as relatively modest amounts of fluoride are concentrated in soils
or surface water pools upon evaporation. Moreover, fluoride in
the form of HF will dominate under acidic conditions (HF pKa is
3.14), and this acidic form is far more bioavailable. Therefore,
bacteria that have genetic and biochemical responses to even
rare exposures to high fluoride concentrations or to low concen-
trations of fluoride under slightly acidic conditions have a large
O
Cl
N
H
N
H
O
HO
N
CF3
F3C
O
O
N
H
N
Figure 1. Structures of Hits from a High-Throughput Screen for
Compounds that Enhance Fluoride Uptake or Retention
HTS were validated via a variety of biochemical assays and were selective advantage. Our chemical screening approach is target-
shown to sensitize bacteria to fluoride. A brief survey of analogs ing this considerable capacity of bacteria to overcome the toxic
of the original hits identified more potent compounds with effects of high concentrations of fluoride.
enhanced fluoride sensitization activity. Overall, this study illus-
After optimizing screening conditions, approximately 75,000
trates the potential for optimization of fluoride toxicity agonists compounds from Chembridge, Maybridge, ChemDiv, and pro-
as antibacterial compounds.
prietary Yale small molecule libraries were tested. In all cases,
Z0 values (Zhang et al., 1999) did not vary beyond 0.5 and 1,
and typically ranged between 0.6 and 0.8. The ratio of the fluo-
rescence signal of the positive control well (fluoride-treated Dfluc
cells) to background (average fluorescence of all wells on a plate)
was higher than 15:1. The action of each compound, when
RESULTS AND DISCUSSION
Design and Validation of a High-Throughput Screen for
Fluoride Toxicity Agonists
To efficiently screen large compound collections for molecules added to WT cells with fluoride, was established by determining
that enhance fluoride uptake and/or retention, a method for the a score between 1 and 100, where 1 represents fluorescence of
rapid, facile, and inexpensive detection of intracellular fluoride WT cells treated with fluoride alone and 100 represents fluores-
was needed. A number of methods exist for the detection of fluo- cence of fluoride-treated Dfluc cells. Hit compounds that yielded
ride, including small molecule sensors (Cametti and Rissanen, a signal greater than 3-fold above background (scores typically
2009, 2013), direct monitoring of 18F accumulation (Drescher above 4) were chosen for further analysis. Hit compounds that
and Suttie, 1972; Li et al., 2013; Quissell and Suttie, 1972), and are inherently fluorescent were discarded, after which the hit
YFP variants that fluoresce in the presence of fluoride (Jayara- rate was approximately 1%.
man et al., 2000). However, none of these methods are
Hit validation consisted of repeating the screening assay and
sufficiently selective or sensitive to rapidly and easily detect confirming that the signal increase was only observed in the
intracellular fluoride concentrations in bacteria. Therefore, we presence of both cells and fluoride, resulting in five molecules
designed a screen that utilized the most selective biosensor for that were chosen for future analysis (Figure 1 and Table S1). Of
fluoride currently known: a recently discovered fluoride-sensing these five hits, compound 1 was the most active, yielding an
riboswitch (Baker et al., 2012).
approximately 8-fold increase in reporter signal at 32 mM when
Specifically, we employed a previously described wild-type 10 mM sodium fluoride was present relative to the addition of
(WT) E. coli strain in which a fluoride-sensing riboswitch from fluoride alone (Figure 2A). These compounds do not increase
Pseudomonas syringae controls the expression of a lacZ re- the activity of the b-galactosidase enzyme in vitro, and there
porter gene in a fluoride concentration-dependent manner was no increase in fluorescence upon treatment of cells
528 Chemistry & Biology 22, 527–534, April 23, 2015 ª2015 Elsevier Ltd All rights reserved

Figure 2. Small Molecules Enhance Fluo-
ride Uptake and Toxicity in E. coli
A
B
5
4
3
2
1
0
0.8
0.6
0.4
0.2
0.0
(A) Reporter activity in E. coli treated with 1 and
NaF. Data points are the average of three repli-
cates and error bars represent SDs. Decreases in
reporter expression at high sodium fluoride and 1
are due to cell growth inhibition. See Figure S1 for
additional control experiments.
NaF (mM)
100
30
10
0
(B) Growth of bacteria cultured in the absence or
presence of 1 and increasing concentrations of
sodium fluoride following 16 hr of incubation at
37ꢁC. DMSO (ꢀ1% [v/v] in the growth medium)
was used to solubilize and deliver test com-
pounds. Data points are the average of three
replicates and error bars represent SDs.
DMSO
DMSO + 1
-6.0
-5.5
-5.0
Logc, 1 (M)
-4.5
-4.0
-3.0
-2.5
-2.0
-1.5
-1.0
-0.5
Logc, NaF(M)
containing a reporter construct carrying a fluoride riboswitch and the presence of hydrogen-bond donor groups in the urea
with a disruptive mutation (Figure S1). These results are consis- or thiourea linker.
tent with our hypothesis that 1 increases intracellular fluoride
concentrations.
Interestingly, the SAR data derived from substitutions within
the urea-linked series was not the same as with thiourea-linked
We next assessed whether 1 enhances the antibacterial activ- (e.g., see 12 and 13). This suggests that the determinants of ac-
ity of fluoride. Although 1 did not significantly change the minimal tivity for these two series might be different. It is possible that the
inhibitory concentration (MIC) of fluoride toward E. coli, it did compounds from the two series might enhance fluoride toxicity
reproducibly enhance the ability of 1 to slow E. coli growth (Fig- via different mechanisms. Alternatively, the preferred conforma-
ure 2B). Addition of 1 alone at 3.2 mM causes no decrease in the tions of the urea-linked molecules may be distinct from the thio-
viability of E. coli but enhances the natural toxicity of fluoride. urea-linked examples (Lepore et al., 1973; Brooks et al., 2006;
These results demonstrate that high-throughput screening of a Bryantsev and Hay, 2006), which could potentially generate the
small molecule library with cells carrying a fluoride-dependent differences in the activities we observe.
reporter gene can reveal compounds that exhibit antibacterial
activity in the presence of fluoride. However, given the modest Hit Derivatives with Enhanced Fluoride-Dependent
enhancement in fluoride toxicity generated by 1, we sought to Toxicity
identify analogs with improved activity.
We expected that compounds with improved activity in the fluo-
ride riboswitch reporter assay might also enhance the antibacte-
rial activity of fluoride. Specifically, we sought to identify analogs
Structure-Activity Relationships
All five selected hits isolated from the HTS contained an elec- that stop bacterial growth rather than simply slow replication as
tron-deficient aryl group linked by either an amide or urea to is observed for compound 1, our most active initial hit (Figure 2).
another ring system. To better understand the possible shared Accordingly, we chose to examine the ability of our best urea- or
pharmacophore and to identify more potent fluoride toxicity thiourea-linked compounds (based on the fluorescence assay)
agonists, a small library of related compounds was synthesized to enhance the toxicity of fluoride to certain bacteria. Treatment
with high yield and purity. This was achieved via one-step with a subinhibitory concentration (0.53 MIC) of 14 resulted in a
coupling of acyl chlorides, isocyanates, or isothiocyanates to a greater than 8-fold improvement in the MIC of fluoride against
variety of commercially available aniline derivatives (Figure 3). E. coli and a ꢀ2-fold improvement against Streptococcus mu-
The fluoride-sensing riboswitch reporter strain was again used tans, a causative agent of gingivitis (Figure 6). A subinhibitory
to assess the extent to which each compound increased intra- concentration of 17 was less effective at increasing fluoride
cellular fluoride concentration.
toxicity compared with 14. However, 17 achieves its more
The relative activities of various analogs (Figure 4 and Table modest effects at a lower concentration. In addition, we
S2) confirmed that an electron-deficient aromatic ring, such as observed significant bacterial toxicity for certain compounds in
those in compounds 6 and 7 (Figures 4B and 4C), were indeed the absence of any added fluoride (Figures 4 and 6). No
necessary for enhancing intracellular fluoride concentrations. enhancement of fluoride toxicity was observed upon treatment
The locations and identities of the electron-withdrawing groups, with ciprofloxacin (Figure S2).
however, do not appear to be of critical importance (e.g., see 6,
Our findings are consistent with the hypothesis that these mol-
8, and 9). Notably, carboxamide-linked compounds were gener- ecules might enhance fluoride toxicity by targeting a cellular pro-
ally less active than the corresponding urea- or thiourea-linked cess that is critical for cell growth even in the absence of fluoride.
compounds (e.g., see 9–11). The most potent compounds (14– However, their antibacterial effects are enhanced by the pres-
17) all include a trifluoro-substituted aryl ring, suggesting that ence of fluoride. Importantly, the concentrations of fluoride
this moiety is particularly beneficial for activity. A graphical sum- used in our study are substantially higher than bacteria typically
mary of the structure-activity relationships (SAR) (Figure 5) de- experience in a natural environment, except for those exposed to
notes the importance of these electron-withdrawing moieties, fluoride-containing oral hygiene products. Thus, any use of such
Chemistry & Biology 22, 527–534, April 23, 2015 ª2015 Elsevier Ltd All rights reserved 529

O
O
Figure 3. Synthesis of Analogs for Struc-
ture-Activity Relationship Analyses
X
R1R2NH
R1
R1R2NH
Cl
N
R2
R1
F3C
N
F3C
N
H
N
R2
THF
DCM
C
60-90%
F3C
F3C
50-90%
X
X=O,S
compounds as potential antibacterial agents would likely require cells, resulting in an increase in fluoride retention. However,
the concomitant application of additional fluoride. given that so little is known about fluoride toxicity mitigation
There are several possible mechanisms by which these mole- mechanisms in bacteria, it is not yet possible to rule out a mech-
cules may be acting. The simplest might be that the compounds anism involving the inhibition of some other process that is crit-
disrupt bacterial membranes in a manner that allows fluoride to ical for fluoride exclusion or expulsion from cells.
more rapidly enter cells. There is precedence for compounds
Intriguingly, some of our hits are also structurally similar
that disrupt membrane integrity to enhance fluoride toxicity (Li to known small molecule fluoride receptors (Boiocchi et al.,
and Breaker, 2012; Nelson et al., 2014; Zasloff and Steinberg, 2004). Receptors for fluoride and other anions often utilize diaryl
1993). Compounds that create nonspecific ion pores in cell ureas or thioureas as hydrogen-bond donors to bind anions
membranes should permit the rapid equilibrium between fluoride (Davis et al., 2010), and the removal of one such potential
outside and inside cells.
hydrogen-bond donor results in a dramatic decrease in com-
Curiously, evidence presented to date convincingly demon- pound activity (Figure 4, 18). Furthermore, incorporation of fluo-
strates that Fluc-type ion channels permit the selective and pas- rine and trifluoromethyl substituents onto aryl rings of small
sive diffusion of fluoride through membranes (Stockbridge et al., molecule chloride binders is known to improve their ability to
2013; Ji et al., 2014), and so members of this protein class theo- transport chloride across phospholipid membranes (Busschaert
retically allow fluoride transport both into and out of cells. How- et al., 2011, 2012).
ever, the large negative membrane potential of bacterial cells
Since we observe that the incorporation of similar substituents
should greatly favor fluoride efflux through the channel (Stock- onto the aryl rings of our molecules enhances fluoride uptake
bridge et al., 2013). Also, under slightly acidic conditions, fluoride and/or retention of E. coli (Figure 4 and Table S2), it is tempting
is expected to accumulate in cells relative to the environment to speculate that the compounds identified in the current study
(Stockbridge et al., 2013; Li et al., 2013), and therefore fluoride function as membrane-permeant fluoride binders that directly
outflow should again be favored.
facilitate the transport of fluoride into bacteria. To evaluate this
By contrast, amphotericin B increases fluoride toxicity to hypothesis, we examined 19 (Figure 4), which is a compound
fungal cells presumably by allowing fluoride to pass either way known to promote the transport of chloride out of lipid vesicles
across cell membranes, resulting in enhanced fluoride toxicity (Busschaert et al., 2012). This compound exhibits the greatest
(Zasloff and Steinberg, 1993; Li and Breaker, 2012). One key dif- ability to increase fluoride toxicity when tested with Strepto-
ference between Fluc proteins and compounds such as ampho- coccus mutans (Figure 6). This finding suggests that the com-
tericin B is that the protein channel is highly specific for fluoride. pound series identified in our HTS might indeed function by
Formation of nonselective ion pores or general membrane directly facilitating the transport of fluoride into bacteria. How-
disruption will also disrupt the membrane potential of the cell, ever, additional studies are needed to convincingly establish
which is critical for the productive use of Fluc channel proteins the true mechanism of action.
to expel fluoride from inside the cell. In this regard, it is interesting
to note that various sodium-proton antiporters are believed to Concluding Remarks
be controlled by fluoride riboswitches, and this might be critical Our findings indicate that small nonpeptidic molecules can be
for maintaining a negative membrane potential to facilitate fluo- used to increase fluoride toxicity against bacteria and that
ride movement out of cells. The disruption of membranes in a such molecules might be frequently identified as hit compounds
nonspecific fashion via various externally added compounds via high-throughput screening. Compounds that could find utility
will also disrupt membrane potential. Thus, these compounds as fluoride toxicity agonists should have a number of character-
might either directly facilitate the transport of fluoride into cells istics. First, for the development of broad-spectrum agents, a
or, by disrupting membrane potential, inhibit its transport out good hit compound should have a provable mechanism that dis-
(Nelson et al., 2014).
rupts a key component of the fluoride defense system used by
Some similarities exist between our molecules and known in- many bacteria or fungi. Second, a hit compound should be a
hibitors of bacterial cell wall biosynthesis (Li et al., 2003; Fran- member of a molecular series that has the potential to be
cisco et al., 2004) and lipid biosynthesis (Russell, 2004). It is improved through the synthesis and testing of analogs. Third,
possible that the disruption of these processes could potentially the optimized agent to be used in a useful formulation should
enhance fluoride transport into cells. Furthermore, the specific induce strong fluoride toxicity against the pathogen but remain
nature of the hits obtained via our HTS would suggest these nontoxic to the host when used alone or in combination with
compounds might function by interacting with a specific target fluoride.
rather than by nonspecifically disrupting membranes. General
membrane disruptors should be far more tolerant of changes sufficiently robust that it can be used to reveal hit compounds
to chemical substructures than our SAR data indicate. by HTS that only weakly increase the intracellular concentration
The fluoride riboswitch reporter system used in this study is
Alternatively, the compounds could inhibit Fluc or some other of fluoride in bacteria. Our chemical library was modest in com-
transporter that is necessary to eject fluoride as it slowly enters pound number and diversity, and therefore it is not surprising
530 Chemistry & Biology 22, 527–534, April 23, 2015 ª2015 Elsevier Ltd All rights reserved

CF3
CF3
OCH3
A
6
S
O
O
O
7
9
additional
ring system
possible
F3C
F3C
N
H
N
H
F3C
NC
N
H
N
H
O
meta/para
EWG
CF3
CF3
O
O
N
H
N
H
8
N
H
N
H
N
H
N
H
CF3
S
O
O
H-bond donor
10
12
11
13
N
H
F3C
F3C
N
H
N
H
F3C
Figure 5. Summary of Structure-Activity Relationships
Incorporation of electron-withdrawing groups (EWG) at the meta or para po-
sition of either aryl ring results in enhanced activity. Utilization of a thiourea
typically has little effect on activity compared with a urea, and likewise the
asymmetric incorporation of an additional ring system has little effect. Both
amide protons are necessary for activity and might be important for hydrogen
bonding (dashed lines).
S
N
H
N
H
F3C
NC
N
H
N
H
F
F
F
F
F
F
F
O
14
15
F3C
F
N
H
N
H
N
H
N
H
F
F
F
that nearly all identified hits were based on the same chemical
scaffold. Although this collection of hits likely operate by a similar
mechanism, it is likely that many other compounds that increase
fluoride toxicity by different mechanisms could be identified by
our HTS process. Although additional chemically distinct mole-
cules were rarely observed to increase fluoride reporter expres-
sion, fluoride toxicity agonists with different chemical scaffolds
and mechanisms likely remain to be discovered. Novel com-
pounds with diverse mechanisms might be more easily identified
in larger and more diverse libraries.
F
F
F
F
F
O
O
S
16
18
17
19
F
N
H
N
H
F
N
H
N
H
CF3
CF3
F3C
Cl
O
N
H
N
CH3
Cl
F3C
N
H
N
H
CF3
B 5
C 5
NaF (mM)
NaF (mM)
100
30
10
0
100
30
10
0
4
3
2
1
0
4
3
2
1
0
Regardless, we have succeeded in identifying novel small
molecules able to render both Gram-positive and Gram-negative
bacteria susceptible to fluoride. Optimization of the ability of
these molecules to enhance fluoride toxicity via additional
rounds of SAR studies could yield antimicrobial agents that
when combined with fluoride offer a novel mechanism of action.
The fluoride riboswitch reporter system described in this study
could be used throughout this optimization process to determine
whether new compounds in the lead series can increase fluoride
concentration in cells. There are numerous challenges ahead for
those who seek to create compounds that could conceivably be
used in topical formulations in conjunction with fluoride in a novel
antibacterial approach, but, given the spread of antibiotic-resis-
tant bacteria, such novel approaches are needed.
-7
-6
-5
-4
-7
-6
Log c, 6 (M)
-5
-4
Log c, 7 (M)
20
15
10
5
D
10 µM
100 µM
SIGNIFICANCE
Herein we describe a high-throughput screen that exploits a
fluoride-sensing riboswitch to regulate reporter gene
expression in E. coli. This screen was successfully used
to isolate compounds that increase intracellular fluoride
concentrations. Hit compounds identified with this screen
reproducibly increase reporter gene expression and
enhance the ability of fluoride to slow bacterial growth.
Based on SAR data, additional and more potent analogs
were identified by the synthesis of analogs. The most
potent derivatives substantially increase the antibacterial
potency of fluoride toward representative Gram-negative
0
1
6
7
8
9
10 11 12 13 14 15 16 17 18
Compound
Figure 4. Increase in Reporter Expression for Selected Hit Analogs
(A) Structures of selected compounds synthesized or purchased for determi-
nation of SAR. Analytical data are presented in the Supplemental Information.
(B) Reporter gene expression in E. coli treated with 6 and various concentra-
tions of fluoride. When 6 was tested at or above 10 mM, bacterial growth was
inhibited in 100 mM fluoride. Data points are the average of three replicates.
Error bars are the SD of the measurement.
(C) Reporter gene expression in E. coli treated with 7 and various concentra-
tions of fluoride. Details are as described for (B).
(D) Fold increase in reporter gene expression in E. coli treated with compound
and 10 mM NaF. Values are normalized to that measured with cells treated with
fluoride in the absence of any compound (value set to 1, dashed line). White
bars indicate the addition of 10 mM compound, while black bars indicate the
addition of 100 mM compound, with the exceptions of 1, 14, and 15 (32 mM) and
16 and 17 (3.2 mM). The absence of a bar indicates no substantial bacterial
growth was observed for that condition. See Table S2 for data on additional
compounds tested.
Chemistry & Biology 22, 527–534, April 23, 2015 ª2015 Elsevier Ltd All rights reserved 531

Figure 6. Structural Derivatives of Hit Com-
pounds Enhance Fluoride Toxicity
A
B
C
(A) MIC of hit compounds for E. coli and S. mutans
in the absence of fluoride. Arrows indicate that the
values for the analyses denoted by the symbols
were not measurable at the highest concentrations
tested.
(B) Decrease in the MIC of sodium or potassium
fluoride for E. coli in the presence of varying con-
centrations of hit compounds. For compounds
with a MIC in the absence of fluoride higher than
32 mg mlꢂ1, half and quarter MIC values were
calculated as if the MIC were 64 mg mlꢂ1. Values
are the average of three replicates.
(C) Decrease in the MIC of sodium or potassium
fluoride for S. mutans in the presence of varying
concentrations of hit compounds. Values are the
average of three replicates.
(0.05 mmol) was placed in the vial and dissolved in dichloromethane
(1.5 ml), to which the isocyanate (0.05 mmol) was added. The vials were sealed
and stirred overnight at room temperature. In most cases, the product urea
precipitated or crystalized from the reaction. The solid was isolated from the
reaction by drawing the solution away from the solid with a fine tipped glass
pipette. The solid was washed with dichloromethane and the liquid was again
withdrawn. In the few cases where solid did not precipitate from the reaction,
the reaction was evaporated under a stream of nitrogen and the resulting solid
was washed with 1:1 ether/hexanes to yield the purified product. Solids were
air dried in the vial and then placed under high vacuum. Proton and fluorine nu-
clear magnetic resonance (NMR) and liquid chromatography-mass spectrom-
etry (LC-MS) were obtained for each compound to verify identity and purity,
which was greater than 90% for all compounds tested. Isolated yields were
60%–90%.
and Gram-positive bacteria. There are numerous potential
mechanisms of action for these compounds and for other
such agents that might be identified in the future, including
the disruption of cell barriers or the direct transport of fluo-
ride across bacterial membranes. Additional molecules that
enhance fluoride toxicity could conceivably be optimized for
use as antibacterial or antifungal agents when used in
conjunction with this anion.
EXPERIMENTAL PROCEDURES
Riboswitch Reporter Expression Analyses
Reporter gene expression analyses were performed as described previously
with some modifications (Nelson et al., 2014). E. coli transformed with fluoride
reporter constructs in pRS414 (Baker et al., 2012) were grown with aeration
overnight in lysogeny broth (LB) supplemented with 75 mg mlꢂ1 carbenicillin
at 37ꢁC. 8 ml of either the overnight culture or the culture diluted 1:10 with
fresh LB were then added to 96-well plates containing 36 ml of 23 modified
LB (20 g lꢂ1 tryptone and 10 g lꢂ1 yeast extract) and 36 ml of various concen-
trations of aqueous NaF. Cultures not supplemented with NaF were instead
supplemented with 1 mM NaCl (final concentration). 0.8 ml of compound dis-
solved in DMSO was added and the plates were sealed with parafilm and
grown, with shaking, at 37ꢁC for either 4 hr (1:10 diluted; data presented in
Tables S1 and S2) or overnight (1:100; all other data). Cultures grown without
compound contained 0.8 ml of DMSO alone.
Diarylthioureas were synthesized exactly as described above using isothio-
cyanates in place of isocyanates, except reactions were stirred for 3 days to
ensure completion. Many of the thiourea products were soluble in dichlorome-
thane necessitating the above-described evaporation and trituration proce-
dure. Solids were air dried in the vial and then placed under high vacuum.
Proton and fluorine NMR and LC-MS were obtained on each compound to
verify identity and purity, which was greater than 90% for all compounds
tested. Isolated yields were 50%–90%.
Arylbenzamides were synthesized by adding the aniline (0.29 mmol) to
1-dram vials. Silicycle morpholine resin (1.3 mmol) was added followed by
the addition of dry tetrahydrofuran (1.5 ml) and, subsequently, 4-trifluoro-
methyl benzoyl chloride (49 ml, 0.29 mmol) was added. The reaction was
stirred overnight before Silicycle amine resin (0.30 mmol) was added to
quench any excess acyl chloride. Additional dry tetrahydrofuran was
added as needed to aid stirring. After 24 hr, the reactions were filtered
Following incubation, bacterial growth was determined by measuring the
optical density at 600 nm. Reporter gene expression was established by first
adding 80 ml of modified Z buffer (60 mM Na2HPO4, 40 mM NaH2PO4,
10 mM KCl, 1 mM MgSO4 [pH 7.0] at 25ꢁC), followed by 40 ml of 4-methylum-
belliferyl-b-D-galactopyranoside (1 mg mlꢂ1 in 50:50 v/v deionized H2O/
DMSO). Samples were mildly agitated and allowed to stand at room temper-
ature for 15 min. The reaction was halted by the addition of Na2CO3 and
gene expression was measured via fluorescence (excitation 360 nm, emission
460 nm).
through
a
pipette with
a
cotton plug with the aid of tetrahydrofuran
syringe and 0.45-mm
and the filtrates were further filtered using
a
a
polytetrafluoroethylene syringe tip filter. The solutions were evaporated
under a stream of nitrogen and dried under high vacuum. Proton and
fluorine NMR and LC-MS were obtained for each compound to verify
identity and purity, which was greater than 90% for all compounds
tested. Isolated yields were 60%–90%. Compounds that were not synthe-
sized were purchased from Chembridge, Maybridge, or Sigma-Aldrich
(compound 19).
The solubility of selected compounds in LB and 0.1 M Tris (pH 7.0 at 25ꢁC)
was determined via nephelometry. NMR spectroscopy data for key com-
pounds (those in Figure 4, with the exception of 19, which was purchased
from Sigma-Aldrich) were obtained using an Agilent 400 mHz spectrometer
equipped with an auto tunable probe. LC-MS spectra were obtained using
an Agilent 1260 infinity liquid chromatograph attached to an Agilent 6120
quadrupole mass spectrometer. The compounds were eluted using an
increasing gradient of acetonitrile (0.1% formic acid) and water. The gradient
varied from 10% acetonitrile to 100% acetonitrile over 2 min with a total run
time of 5 min. Compounds were separated using an Agilent Eclipse plus
C18 5.1 3 50 mm column and monitored at 254 nm.
High-throughput screening was performed largely as described above, with
some modifications. LB supplemented with 1 mM sodium fluoride was added
in 10 ml volumes to each well of 384-well plates. 20 nl of a 10 mM solution of
each compound to be tested was added via pronging, followed by the addition
of 10 ml of an overnight culture of E. coli containing the fluoride reporter. WT
and Dfluc E. coli cells, as well as media alone, were included on each plate
as controls. Bacteria were grown by shaking in a humidified incubator at
37ꢁC for 24 hr, at which time reporter gene expression was assayed as
described above.
Chemical Synthesis
Diaryl ureas were synthesized by the one-step coupling of various anilines with
various isocyanates in preweighed 1-dram (3.7-ml) vials. Each aniline
532 Chemistry & Biology 22, 527–534, April 23, 2015 ª2015 Elsevier Ltd All rights reserved

Bacterial Growth Curves and MIC Determinations
Brooks, S.J., Edwards, P.R., Gale, P.A., and Light, M.E. (2006). Carboxylate
complexation by a family of easy-to-make ortho-phenylenediamine based
bis-ureas: studies in solution and the solid state. New J. Chem. 30, 65–70.
Bacteria were grown as described above, overnight, and diluted 1:10 in 23 LB.
8 ml of the resulting solution was then added to 42 ml of 23 modified LB and
42 ml of aqueous NaF, to which 0.8 ml of either DMSO or 1 dissolved in
DMSO was added. Cultures were then incubated in a 100-well Honeycomb
plate at 37ꢁC in a Bioscreen C system (Growth Curves USA) as described pre-
viously (Nelson et al., 2014).
Blount, K.F., and Breaker, R.R. (2006). Riboswitches as antibacterial drug tar-
gets. Nat. Biotechnol. 24, 1558–1564.
Bryantsev, V.S., and Hay, B.P. (2006). Conformational preferences and internal
rotation in alkyl- and phenyl-substituted thiourea derivatives. J. Phys. Chem. A
110, 4678–4688.
MIC assays were conducted in a final volume of 100 ml according to
methods established by the Clinical Laboratory Standards Institute (CLSI,
2012). Briefly, test or control compound dissolved in DMSO or water as
appropriate was serially diluted (1:2) into successive tubes of the same sol-
vent. A 5-ml aliquot of each dilution was transferred to the appropriate wells
of 96-well clear round-bottom plates. To each dilution series, 95 ml of a
freshly prepared bacterial suspension was added with mixing to provide a
final bacterial inoculum of approximately 105–106 colony-forming units per
well and a final DMSO concentration of 5%. DMSO was added regardless
of whether the compound was dissolved in water or DMSO. Cation-adjusted
Mueller-Hinton broth (Fluka) was used for all MIC assays, and lysed horse
blood (Cleveland Scientific) was included at a final concentration of 5% for
assays with S. mutans. Assay cultures were incubated for 18–24 hr at
37ꢁC and in an atmosphere enriched to 5% CO2 for S. mutans. The MIC is
defined as the lowest concentration of antimicrobial agent that completely in-
hibits the growth of the organism as detected by the unaided eye. Ciproflox-
acin was used for comparison with a known antibiotic and as a benchmark
for each screening assay.
Busschaert, N., Wenzel, M., Light, M.E., Iglesias-Hernandez, P., Perez-Tomas,
R., and Gale, P.A. (2011). Structure-activity relationships in tripodal transmem-
brane anion transporters: the effect of fluorination. J. Am. Chem. Soc. 133,
14136–14148.
Busschaert, N., Kirby, I.L., Young, S., Coles, S.J., Horton, P.N., Light, M.E.,
and Gale, P.A. (2012). Squaramides as potent transmembrane anion trans-
porters. Angew. Chem. Int. Ed. Engl. 51, 4426–4430.
Cametti, M., and Rissanen, K. (2009). Recognition and sensing of fluoride
anion. Chem. Commun. 20, 2809–2829.
Cametti, M., and Rissanen, K. (2013). Highlights on contemporary recognition
and sensing of fluoride anion in solution and in the solid state. Chem. Soc. Rev.
42, 2016–2038.
CLSI. (2012). Methods for Dilution Antimicrobial Susceptibility Tests for
Bacteria that Grow Eerobically; Approved Standard, Ninth Edition. (Clinical
and Laboratory Standards Institute).
Curran, T.M., Buckley, D.H., and Marquis, R.E. (1994). Quasi-irreversible inhi-
bition of enolase of Streptococcus mutans by fluoride. FEMS Microbiol. Lett.
119, 283–288.
SUPPLEMENTAL INFORMATION
Davis, J.T., Okunola, O., and Quesada, R. (2010). Recent advances in the
Supplemental Information includes two figures, two tables, and characteriza-
tion of newly synthesized compounds and can be found with this article online
at http://dx.doi.org/10.1016/j.chembiol.2015.03.016.
transmembrane transport of anions. Chem. Soc. Rev. 39, 3843–3862.
Deigan, K.E., and Ferre´ -D’Amare´ , A.R. (2011). Riboswitches: discovery of
drugs that target bacterial gene-regulatory RNAs. Acc. Chem. Res. 44,
1329–1338.
AUTHOR CONTRIBUTIONS
Drescher, M., and Suttie, J.W. (1972). Intracellular fluoride in cultured mamma-
T.D.A. and R.R.B. designed the HTS. T.D.A. performed it with J.W.N. M.S.P.
synthesized and verified all other molecules. J.W.N. conducted the reporter
assay and growth curve experiments. K.F.B. determined the MIC of selected
compounds with and without fluoride. J.W.N., K.F.B., and R.R.B analyzed
the data and wrote the paper.
lian cells. Exp. Biol. Med. 139, 228–230.
Francisco, G.D., Li, Z., Albright, J.D., Eudy, N.H., Katz, A.H., Petersen, P.J.,
Labthavikul, P., Singh, G., Yang, Y., Rasmussen, B.A., et al. (2004). Phenyl
thiazolyl urea and carbamate derivatives as new inhibitors of bacterial cell-
wall biosynthesis. Bioorg. Med. Chem. Lett. 14, 235–238.
Jayaraman, S., Haggie, P., Wachter, R.M., Remington, S.J., and Verkman,
A.S. (2000). Mechanism and cellular applications of a green fluorescent pro-
tein-based halide sensor. J. Biol. Chem. 275, 6047–6050.
ACKNOWLEDGMENTS
We thank Mariya Kolesnikova, Janie Merkel, and the entire Yale Center for
Molecular Discovery for their assistance in performing high-throughput
screening, and Edward Bickham, Daniel Moonan, and Meredith Redick
for assistance in acquiring the SAR data. We also thank Timothy Newhouse
at Yale University for bringing compound 19 to our attention. This work was
supported by the NIH (5R01DE022340) and the Howard Hughes Medical
Institute.
Ji, C., Stockbridge, R.B., and Miller, C. (2014). Bacterial fluoride resistance,
Fluc channels, and the weak acid effect. J. Gen. Phys. 144, 257–261.
Kim, J.N., Blount, K.F., Puskarz, I., Lim, J., Link, K.H., and Breaker, R.R. (2009).
Design and antimicrobial action of purine analogues that bind guanine ribo-
switches. ACS Chem. Biol. 4, 915–927.
Lee, E.R., Blount, K.F., and Breaker, R.R. (2009). Roseoflavin is a natural anti-
bacterial compound that binds to FMN riboswitches and regulates gene
expression. RNA Biol. 6, 187–194.
Received: September 30, 2014
Revised: March 9, 2015
Lepore, G., Migdal, S., Blagdon, D.E., and Goodman, M. (1973).
Conformations of substituted aryl ureas in solution. J. Org. Chem. 38, 2590–
2594.
Accepted: March 16, 2015
Published: April 23, 2015
Li, L. (2003). The biochemistry and physiology of metallic fluoride: action,
REFERENCES
mechanism, and implications. Crit. Rev. Oral Biol. Med. 14, 100–114.
Baker, J.L., Sudarsan, N., Weinberg, Z., Roth, A., Stockbridge, R.B., and
Breaker, R.R. (2012). Widespread genetic switches and toxicity resistance
proteins for fluoride. Science 335, 233–235.
Li, S., and Breaker, R.R. (2012). Fluoride enhances the activity of fungicides
that destabilize cell membranes. Bioorg. Med. Chem. Lett. 22, 3317–3322.
Li, Z., Francisco, G.D., Hu, W., Labthavikul, P., Petersen, P.J., Severin, A.,
Singh, G., Yang, Y., Rasmussen, B.A., Lin, Y.I., et al. (2003). 2-Phenyl-5,6-di-
hydro-2H-thieno[3,2-c]pyrazol-3-ol derivatives as new inhibitors of bacterial
cell wall biosynthesis. Bioorg. Med. Chem. Lett. 13, 2591–2594.
Barbier, O., Arreola-Mendoza, L., and Del Razo, L.M. (2010). Molecular mech-
anisms of fluoride toxicity. Chem. Biol. Interact 188, 319–333.
Boiocchi, M., Del Boca, L., Gomez, D.E., Fabbrizzi, L., Licchelli, M., and
Monzani, E. (2004). Nature of urea-fluoride interaction: incipient and definitive
proton transfer. J. Am. Chem. Soc. 126, 16507–16514.
Li, S., Smith, K.D., Davis, J.H., Gordon, P.B., Breaker, R.R., and Strobel, S.A.
(2013). Eukaryotic resistance to fluoride toxicity mediated by a widespread
family of fluoride export proteins. Proc. Natl. Acad. Sci. USA 110, 19018–
19023.
Breaker, R.R. (2011). Prospects for riboswitch discovery and analysis. Mol.
Cell 43, 867–879.
Chemistry & Biology 22, 527–534, April 23, 2015 ª2015 Elsevier Ltd All rights reserved 533

Lu¨ nse, C.E., Scott, F.J., Suckling, C.J., and Mayer, G. (2014a). Novel TPP-
riboswitch activators bypass metabolic enzyme dependency. Front. Chem.
2, 53.
Quissell, D.O., and Suttie, J.W. (1972). Development of a fluoride-resistant
strain of L cells: membrane and metabolic characteristics. Am. J. Physiol.
223, 596–603.
Russell, A.D. (2004). Whither triclosan? J. Antimicrob. Chemother. 53,
Lu¨ nse, C.E., Schu¨ ller, A., and Mayer, G. (2014b). The promise of ribo-
switches as potential antibacterial drug targets. Int. J. Med. Microbiol.
304, 79–92.
693–695.
Serganov, A., and Nudler, E. (2013). A decade of riboswitches. Cell 152, 17–24.
Ster, C., Allard, M., Boulanger, S., Lamontagne Boulet, M., Mulhbacher, J.,
Lafontaine, D.A., Marsault, E., Lacasse, P., and Malouin, F. (2013).
Experimental treatment of Staphylococcus aureus bovine intramammary
infection using a guanine riboswitch ligand analog. J. Dairy Sci. 96, 1000–1008.
Mulhbacher, J., Brouillette, E., Allard, M., Fortier, L., Malouin, F., and
Lafontaine, D.A. (2010). Novel riboswitch ligand analogs as selective inhibitors
of guanine-related metabolic pathways. PLoS Pathog. 6, e1000865.
Nakai, C., and Thomas, J.A. (1974). Properties of a phosphoprotein phospha-
tase from bovine heart with activity on glycogen synthase, phosphorylase, and
histone. J. Biol. Chem. 249, 6459–6467.
Stockbridge, R.B., Lim, H.H., Otten, R., Williams, C., Shane, T., Weinberg, Z.,
and Miller, C. (2012). Fluoride resistance and transport by riboswitch-
controlled CLC antiporters. Proc. Natl. Acad. Sci. USA 109, 15289–15294.
Nelson, J.W., Zhou, Z., and Breaker, R.R. (2014). Gramicidin D enhances the
Stockbridge, R.B., Robertson, J.L., Kolmakova-Partensky, L., and Miller, C.
(2013). A family of fluoride-specific ion channels with dual-topology architec-
ture. Elife 2, e01084.
activity of fluoride. Bioorg. Med. Chem. Lett. 24, 2969–2971.
Peselis, A., and Serganov, A. (2014). Themes and variations in riboswitch
structure and function. Biochim. Biophys. Acta 1839, 908–918. http://dx.doi.
org/10.1016/j.bbagrm.2014.02.012.
Zasloff, M., and Steinberg, W.H. (1993). Composition and treatment with bio-
logically active peptides and certain anions, U.S. Patent 5,217,956.
Qin, J., Chai, G., Brewer, J.M., Lovelace, L.L., and Lebioda, L. (2006). Fluoride
inhibition of enolase: crystal structure and thermodynamics. Biochemistry 45,
793–800.
Zhang, J.H., Chung, T.D.Y., and Oldenburg, K.R. (1999). A simple statistical
parameter for use in evaluation and validation of high throughput screening
assays. J. Biomol. Screen. 4, 67–73.
534 Chemistry & Biology 22, 527–534, April 23, 2015 ª2015 Elsevier Ltd All rights reserved