Identification of Small Molecules That Antagonize Diguanylate Cyclase Enzymes To Inhibit Biofilm Formation

Abstract

Bacterial biofilm formation is responsible for numerous chronic infections, causing a severe health burden. Many of these infections cannot be resolved, as bacteria in biofilms are resistant to the host's immune defenses and antibiotic therapy. New strategies to treat biofilm-based infections are critically needed. Cyclic di-GMP (c-di-GMP) is a widely conserved second-messenger signal essential for biofilm formation. As this signaling system is found only in bacteria, it is an attractive target for the development of new antibiofilm interventions. Here, we describe the results of a high-throughput screen to identify small-molecule inhibitors of diguanylate cyclase (DGC) enzymes that synthesize c-di-GMP. We report seven small molecules that antagonize these enzymes and inhibit biofilm formation by Vibrio cholerae. Moreover, two of these compounds significantly reduce the total concentration of c-di-GMP in V. cholerae, one of which also inhibits biofilm formation by Pseudomonas aeruginosa in a continuous-flow system. These molecules represent the first compounds described that are able to inhibit DGC activity to prevent biofilm formation.

INTRODUCTION

Biofilms are multicellular bacterial communities encased in an extracellular matrix. Biofilms have been estimated by the National Institutes of Health to be associated with 80% of all bacterial infections (13). It was recently estimated that biofilm-based disease is responsible for 19 million infections annually in the United States, resulting in hundreds of thousands of fatalities and billions of dollars in medical expenses (50). Biofilm formation promotes increased antibiotic tolerance to levels 1,000 times greater than those observed in planktonic bacteria (14, 23, 24). Furthermore, biofilms resist host immune defense strategies, such as mechanical clearance, complement-mediated killing, antibody recognition, and phagocytosis (14). Chronic infections, such as lung pneumonia of cystic fibrosis patients, otitis media, chronic nonhealing wounds, and contamination of artificial medical implants, are also associated with biofilm formation (14). Often, due to ineffective antibiotic therapy, these infections cannot be effectively treated (5).

The second messenger cyclic di-GMP (c-di-GMP) has recently emerged as a novel signal that controls biofilm formation and represses motility (6, 18, 34, 36, 47). Synthesis of c-di-GMP occurs via diguanylate cyclase (DGC) enzymes encoding GGDEF domains, while degradation of c-di-GMP occurs via phosphodiesterase (PDE) enzymes encoding either an EAL or a HD-GYP domain (8, 35, 37, 41). Analysis of bacterial genome sequences revealed that enzymes predicted to synthesize or degrade c-di-GMP are found in 85% of all bacteria, including many prominent human pathogens (9). Deletion of active DGCs completely abolishes biofilm formation, suggesting c-di-GMP is essential for this process in bacteria that utilize the signal (30, 45). Importantly, the enzymatic mechanism of DGCs and PDEs is conserved between species. For example, the unrelated DGCs hmsT from Yersinia pestis and adrA from Salmonella enterica were able to cross-complement mutations in one another, even though they share no homology outside the DGC domain (42). Moreover, there is no evidence that DGCs synthesize other signals besides c-di-GMP.

Because of the widespread conservation of c-di-GMP signaling systems in bacteria and the critical role of c-di-GMP in promoting biofilm formation, inhibition of c-di-GMP signaling systems offers an attractive approach to interfere with biofilm formation (44). Importantly, enzymes associated with c-di-GMP are not encoded in eukaryotic organisms. Thus, small molecules inhibiting this system would be predicted to have less toxicity to the infected host. c-di-GMP is not essential for growth, and small molecules that reduce the intracellular concentration of c-di-GMP would not directly select for resistant organisms. To our knowledge, a glycosylated triterpenoid saponin (GTS) isolated from Pisum sativum is the only know inhibitor of DGC enzymes (31, 32). However, these are complex molecules that were not able to inhibit DGC activity in whole cells, likely due to an inability to cross the outer membrane. Moreover, GTS has not been demonstrated to have antibiofilm properties.

Here, we performed a high-throughput screen leading to the identification of seven small molecules that inhibit multiple DGC enzymes. These compounds also reduce Vibrio cholerae biofilm formation. Two of these molecules are able to significantly reduce the intracellular concentration of c-di-GMP in V. cholerae; however, the remaining five compounds inhibit biofilm formation without significantly altering total c-di-GMP levels. One compound that reduces the concentration of c-di-GMP in V. cholerae significantly inhibits biofilm formation by Pseudomonas aeruginosa in a continuous-flow system. The seven DGC inhibitors we have identified can serve as a foundation to develop improved inhibitors of DGC enzymes to prevent bacterial biofilm formation.

MATERIALS AND METHODS

Bacterial strains and media.

The bacterial strains and plasmids used in this study are listed in Table 1. V. cholerae C6706str2 and P. aeruginosa PAO1cells were grown at 37°C with constant aeration in Luria-Bertani broth (LB). For expression studies, isopropyl β-d-1-thiogalactopyranoside (IPTG) was used at concentrations of 100 μM. When necessary, antibiotics were used at concentrations of 100 μM.

Table 1.

Strains, plasmids, and primers used in the study

Strain or plasmid	Description	Reference
Strains
V. cholerae C6706str2		47a
V. cholerae ΔVC1086		This study
P. aeruginosa PAO1		46a
E. coli BL21(DE3)	F− ompT hsdSB(rB− mB−) gal dcm (DE3)	Invitrogen
Plasmids
pVC1216	Overexpression of VC1216 GGDEF	This study
pVC1673-lux	C-di-GMP inducible transcriptional fusion	46
pETVC142	VC2370(142–579) purification plasmid	This study
pVC484E	VC2370(142)-D484E purification plasmid	This study
pET21bW	WspR purification plasmid	This study
pWR242A	WspR-R242A purification plasmid	This study
Primers
VC2370-142-15b	GGAATTCCATATGGAGTGTCCCAGCCATGACATTC	This study
VC2370-rev-1	GGAATTCCATATGCTGATTTTTATTAAAAGTACCGAAGGC	This study
VC_D484E_top	CCAATCGCCATTCTGATTGTGTTGCCCGCTACG	This study
VC_D484E_bottom	CGTAGCGGGCAACACAATCAGAATGGCGATTGG	This study
wspR_F	GAAGGAGATATACATATGCACAACCCTCATG	This study
wspR_R	GTGGTGGTGGTGCTCGAGGCCCGCCGGGGCCGGC	This study
wspR_R242A	CGCGAGGGCTGCAGTGCCTCCTCGGACCTGGC	This study

High-throughput screen to identify DGC inhibitors.

The high-throughput screen to identify compounds that interfere with c-di-GMP signaling was previously described (38). Briefly, a V. cholerae reporter strain containing two plasmids was utilized. The first plasmid encoded the DGC VC1216 under the control of the Ptac promoter, which allowed induction of this enzyme with IPTG, leading to increased c-di-GMP levels. The second plasmid encoded a transcriptional fusion of a c-di-GMP-inducible promoter located near the gene VC1673 to the luciferase operon (lux) (46). Therefore, luciferase functioned as a reporter for c-di-GMP levels. The culture was incubated overnight at 30°C, and growth was monitored as the optical density at 600 nm (OD₆₀₀). Luminescence was determined using a Pherastar plate reader (BMB Labtech, Cary, NC). Approximately 66,000 compounds and natural-product extracts were screened once, and 1,039 small molecules and 357 natural-product extracts exhibiting differences greater than 3 standard deviations from the negative control were rescreened in triplicate. The top 331 compounds from this rescreen were selected, and the concentration at 50% inhibition (IC₅₀) was determined in duplicate.

Assessment of biofilm formation.

Biofilm formation was measured under both static and flow conditions. To measure biofilm formation under static conditions, we used a quantitative crystal violet assay on minimum-biofilm-eliminating-concentration (MBEC) plates (Biosurface Technologies, Bozeman MT), as described previously (15, 39). The MBEC technology consists of a microtiter plate cover containing 96 polystyrene pegs that sit in the 96 wells of a conventional plate. Briefly, cultures grown overnight were standardized to an OD₅₉₅ of 0.05, 150 μl was transferred to the wells of a 96-well polystyrene microtiter plate, and the MBEC lid was placed on top of the wells. Biofilms were grown on the pegs of the lid under shaking conditions for 8 h. The lid was removed, and the pegs were gently washed three times with 160 μl of phosphate-buffered saline (PBS) to remove nonadherent cells. Adherent biofilms on the pegs were fixed with 160 μl of 100% ethanol prior to staining for 2 min with 160 μl of 0.41% (wt/vol) crystal violet in 12% ethanol (Protocol crystal violet; Biochemical Sciences, Swedesboro, NJ). The pegs were washed several times with 200 μl PBS, pH 7.5, to remove excess stain. Quantitative assessment of biofilm formation was obtained by immersing the pegs in a sterile polystyrene microtiter plate containing 200 μl of 100% ethanol and incubating the plate at room temperature for 10 min to dissolve the crystal violet (33). The absorbance at 595 nm was determined using a SpectraMax M5 microplate spectrophotometer system (Molecular Devices Sunnyvale, CA). At least three independent experiments were performed for each of the assays.

Determination of biofilm formation under flow conditions utilized disposable flow cells (Stovall Life Science, Greensboro, NC), as previously described (39). In brief, the inlet side of the flow cell was connected to a sterile reservoir filled with the appropriate growth medium. The outlet side was connected to a waste reservoir to create a “once-through” flow cell system. Tubing upstream of each individual cell was injected with 0.5 ml of an overnight culture of the test strain adjusted to an OD₆₀₀ of 0.05, and the chamber was incubated in the upside down position at 37°C for 20 min. Flow was then resumed at a rate of 0.3 ml/min. The nonadherent bacteria were eventually flushed by the flow of the medium, thereby replacing the volume of the flow cell once every minute. Biofilm formation on the flow cell was imaged both macroscopically and microscopically at the indicated times.

For biofilm dispersal studies, biofilms were allowed to grow on the pegs of MBEC plates for 24 to 48 h before transfer to a fresh microtiter plate containing 100 μM inhibitors. After 30 min of incubation, the biofilms were removed and the dispersed bacteria were measured either by plating for CFU or by growing the plate with shaking at 37°C with constant monitoring of the OD₆₀₀ for 5 h.

Microscopy.

Confocal laser scanning microscopy (CLSM) analysis of biofilms was performed by stopping the medium flow and then injecting the fluorescent dye Syto-9 (Molecular Probes, Eugene, OR) into the flow cell chamber. The chamber was incubated for 20 min in the dark. Confocal microscopic images were acquired using a Carl Zeiss Pascal laser scanning microscope (Carl Zeiss, Jena, Germany) equipped with a 40×/1.4 numerical-aperture Plan-Apochromat objective. The Syto-9 and propidium iodide fluorophores were excited with an argon laser at 488 nm, and the emission band-pass filters used for Syto-9 and propidium iodide were 505 to 530 nm and low-pass 560 nm, respectively. CLSM z-stack image analysis of 112-μm² images and processing were performed using Carl Zeiss LSM 5 Pascal software (version 3.5). Image stacks of biofilms were acquired from nine distinct regions on the flow cell. The thickness of the biofilm was measured starting from the z section at the flow cell/biofilm interface to the z section at the top of the biofilm surface containing <5% of the total biomass. Image analysis of biofilms was performed with Comstat version 2.1 (16).

Protein production.

V. cholerae DGC VC2370 (residues 142 to 579) was cloned into the NdeI and XhoI sites of pET15b (Novagen) by PCR amplification using primers VC2370-142-15b and VC2370-rev-1 (Table 1) to give pETVC142. Quickchange (Stratagene) site-directed mutagenesis was performed on pETVC142 using primers VC_D484E_top and VC_D484E_bottom (Table 1) to obtain the VC2370(142)-D484E expression vector (pVC484E). N-terminally His-tagged VC2370(142)-D484E was overexpressed in E. coli strain BL21(DE3) by first growing the cells in LB medium supplemented with 100 μg/ml ampicillin to an OD₆₀₀ of 0.5, followed by induction of expression with 0.5 mM IPTG for 18 h at 16°C. The cells were then pelleted and stored at −80°C. All subsequent protein purification steps were carried out at 4°C. The cells were lysed in buffer A (20 mM Tris [pH 7.5], 150 mM NaCl, and 20 mM imidazole) supplemented with 1 μM pepstatin, 20 μg/ml DNase, and 1 mM phenylmethanesulfonyl fluoride (PMSF). The crude lysate was centrifuged for 1 h at 13,000 rpm at 4°C. The cell-free supernatant was applied to Ni-nitrilotriacetic acid (NTA) agarose (Novagen) equilibrated with buffer A. The column was then washed, and the resin was resuspended in buffer A. To remove the His affinity tag, thrombin was added at 0.3-mg/ml Ni-NTA bed volume, and the resin was gently rotated at 4°C for 16 h. VC2370(142)-D484E was then eluted, diluted 3-fold with 20 mM Tris (pH 7.5), and loaded onto a Source 15Q column (GE Healthcare) equilibrated in buffer C (20 mM Tris [pH 7.5] and 50 mM NaCl). VC2370(142)-D484E was eluted in a 0.05 to 1.0 M NaCl gradient of buffer C. Fractions containing VC2370(142)-D484E were concentrated and further purified using a Superdex 200 (GE Healthcare) column equilibrated with buffer D (20 mM Tris [pH 7.5] and 150 mM NaCl). Fractions containing VC2370(142)-D484E were concentrated to 540 μM and stored at −80°C.

wspR was amplified from P. aeruginosa PAO1 genomic DNA with the primer pair wspR_F and wspR_R (Table 1) using Advantage HD Polymerase (Clontech). The PCR product was cloned into the NdeI and XhoI sites of pET21b using the In-Fusion method (Clontech) to give pET21bW. The WspR-R242A expression vector (pWR242A) was generated by site-directed mutagenesis of pET21bW using the ChangeIT Mutagenesis Kit (USB) and the oligonucleotide wspR_R242A (Table 1). C-terminally His-tagged WspR-R242A was overexpressed and purified as previously described (7) with the following modifications. Protein production was induced at 16°C for 18 h; the lysis buffer contained 1 μM pepstatin, 1 μM leupeptin, 1 mM PMSF, and 20 μg/ml DNase; and the protein was subjected to gel filtration using a Sephacryl 100 16/60 column (GE Healthcare) equilibrated in buffer containing 30 mM Tris (pH 7.6), 100 mM NaCl, and 1 mM dithiothreitol (DTT). The fractions containing WspR-R242A dimers were pooled and concentrated. WspR-R242A (2.87 mM) was then stored at −80°C in 15 mM Tris (pH 7.6), 50 mM NaCl, 0.5 mM DTT, and 50% glycerol.

Measurement of diguanylate cyclase activity in vitro.

The ability of compounds to inhibit DGC activity was determined using the EnzChek Pyrophosphate Assay (Invitrogen), as previously described (7), in a 100-μl volume to allow high-throughput measurements. In brief, the assay allows spectrophotometric detection of inorganic pyrophosphate (PP_i) released during the reaction. The enzyme inorganic pyrophosphatase catalyzes conversion of PP_i produced during c-di-GMP synthesis into two equivalents of inorganic phosphate (P_i). In the presence of P_i, the substrate 2-amino-6-mercapto-7-methylpurine ribonucleoside (MESG) is enzymatically converted by purine nucleoside phosphorylase (PNP) to ribose 1-phosphate and 2-amino-6-mercapto-7-methylpurine. Enzymatic conversion of MESG results in a shift in the absorbance maximum from 330 nm for the substrate to 360 nm for the product. Two DGC enzymes, VC2370(142)-D484E from V. cholerae and WspR-R242A from P. aeruginosa, were used as the target enzymes. VC2370(142)-D484E contains the cytoplasmic portion of DGC VC2370 with a mutation in the RXXD inhibition site. This mutation was engineered, as it prevents copurificaiton of c-di-GMP with the protein and blocks feedback inhibition during kinetic assays. WspR-R242A contains a mutation that locks the enzyme in a constitutively active state. The inhibitors (2 μl resuspended in dimethyl sulfoxide [DMSO]) were added to 100-μl reaction mixtures that contained the components of the EnzCheck Pyrophosphatase Assay, adjusted for volume as directed, plus 24 mM Tris, pH 7.5, 5 mM MgCl₂, 45 mM NaCl, and 5 μM DGCs. The inhibitors were incubated with the enzyme for 30 min at room temperature before starting the reaction with the addition of 62.5 μM GTP. Reaction mixtures with no enzyme added or no GTP added were examined simultaneously to verify that increased absorbance was due to c-di-GMP synthesis. These controls displayed no DGC activity. Absorbance was continuously monitored at 360 nm using a SpectraMax M5 microplate spectrophotometer system for 5 min. The rate of the reaction in the absence of compound was normalized to 100%, and the OD₃₆₀ increased linearly for all analyses under these conditions. To determine if the compounds specifically inhibited DGC enzymes, a control reaction was performed to determine the ability of the molecules to inhibit calf intestinal phosphatase (CIP) (NEB), using GTP as a substrate with all other reaction conditions identical.

Measurement of the intracellular c-di-GMP concentration in vivo.

Lead compounds identified from the chemical screen were evaluated for the ability to inhibit c-di-GMP production in vivo using ultraperformance liquid chromatography-tandem mass spectrometry (UPLC-MS-MS), as previously described (3). Two milliliters of bacterial culture containing different lead compounds were grown from an overnight inoculum to an optical density of 1.0 at 560 nm. The cells were then centrifuged at 12,000 rpm for 30 s and extracted with 300 μl methyl 40% alcohol-40% acetonitrile-0.1 N formic acid buffer. The cells were then placed at −20°C for 30 min, and cell debris was removed by centrifugation at 15,000 rpm for 5 min. All compounds were independently analyzed four to six times.

Eukaryotic-cell toxicity.

The cytotoxicity assay was carried out in triplicate by growing THP-1 human cells suspended in RPMI medium with 4.5 g/liter glucose plus 10% fetal bovine serum (FBS) for 8 h with direct inhibitor 3. The cells were then stained with 0.2% trypan blue, which labels dead cells. As a positive control, the cells were killed by addition of 0.025% glutaraldehyde. The cells were counted at ×100 total magnification under bright-field microscopy to determine viability.

Statistical analysis.

Statistical significance was determined using a paired one-tailed Student's t test based on our hypothesis that the lead compounds would lower the activity of DGC enzymes, biofilm formation, and the in vivo concentration of c-di-GMP.

RESULTS

Small-molecule screen.

We initiated a chemical screen to identify novel compounds and natural-product extracts that inhibit the activity of DGC enzymes responsible for c-di-GMP synthesis. The details of the screen have been previously described (38). To identify DGC inhibitors, we grew V. cholerae cells containing a transcriptional reporter that is induced by c-di-GMP in the presence of 66,000 compounds/natural-product extracts at the Center for Chemical Genomics at the University of Michigan. V. cholerae is an excellent model organism to identify compounds targeting c-di-GMP signaling because c-di-GMP and biofilm formation are well studied in the organism (20, 48, 49) and V. cholerae has relatively high intracellular concentrations of c-di-GMP (43). The reporter plasmid encodes a transcriptional fusion of a c-di-GMP-inducible promoter located in the gene VC1673 to luciferase (VC1673-lux) (46). A second plasmid drove expression of an active DGC enzyme to increase intracellular c-di-GMP levels. As high intracellular c-di-GMP promotes the expression of VC1673-lux, we identified compounds and natural-product extracts that reduced VC1673-lux expression without negatively impacting growth. One mechanism by which compounds reduce reporter gene expression is to inhibit the activity of the expressed DGC. Importantly, the screen utilized intact bacteria, indicating that active compounds were likely able to enter the cytoplasm. Alternatively, the small molecules could signal via a receptor on the cell surface.

We identified 358 compounds that exhibited greater than 50% inhibition of lux and 466 compounds that exhibited between 50% and 30% lux inhibition. In addition, 274 natural-product isolates, which will be described elsewhere, that showed greater than 30% lux inhibition were identified. None of these compounds or natural products affected bacterial growth. Twenty-seven of the 358 most active compounds that had known toxicity to eukaryotic cells were removed from further analysis. The IC₅₀s for the remaining 331 small molecules showing greater than 50% inhibition were determined by measuring in duplicate the lux expression of the reporter gene at eight different concentrations. A summary of the IC₅₀s for these 331 compounds is shown in Fig. 1. Importantly, 184 compounds had IC₅₀s of less than 10 μM.

Fig 1.

Summary of lead hits from chemical screens. The numbers of hits with the given IC₅₀s isolated from the small-molecule screen are shown.

Identification of DGC inhibitors.

To determine which lead compounds reduce luciferase expression through inhibition of c-di-GMP signaling, using an in vitro enzyme assay, we tested 166 of the top 184 lead compounds to determine if they inhibited the activity of two purified DGCs. Briefly, conversion of GTP to c-di-GMP by DGCs produces pyrophosphate, which was monitored using the EnzCheck Pyrophosphate Assay (Invitrogen). We modified the assay to allow screening in a high-throughput microtiter format. The first DGC that we examined was the cytoplasmic fragment of the DGC enzyme VC2370 from V. cholerae, as we had previously found, using high-performance liquid chromatography (HPLC)-MS-MS analysis, that the protein actively synthesizes c-di-GMP in vitro (unpublished results). We mutated the RXXD allosteric inhibition site of the protein (VC2370(142)-D484E) because we observed, using HPLC-MS-MS, that c-di-GMP copurifies with native VC2370, complicating further analysis. Mutation of this RXXD site prevented c-di-GMP copurification. Also, mutation of the site ensures that c-di-GMP produced during the in vitro reaction is not able to inhibit enzyme activity. Importantly, since an RXXD mutant was used in the assay, it is unlikely that any of the identified DGC inhibitors function through interaction with the motif.

Thirteen of the 166 test compounds significantly reduced VC2370(142)-D484E activity, exhibiting IC₅₀s below 50 μM (Table 2). However, bacteria typically encode multiple proteins with DGC domains. For example, the strain of V. cholerae used in this study encodes 40 distinct DGC domains (9). Therefore, the most desirable lead compounds exhibit general inhibition of multiple DGC enzymes. To determine if the compounds that inhibited VC2370(142)-D484E were general inhibitors of DGC enzymes, we purified a constitutively active form of the DGC WspR (WspR-R242A) from P. aeruginosa. This allele of WspR does not require phosphorylation of its N-terminal receiver domain to exhibit DGC activity (7). Purified WspR-R242A possessed DGC activity, as confirmed by HPLC-MS-MS (data not shown). Analysis of the 13 VC2370 inhibitors revealed that 9 of the molecules also significantly inhibited WspR activity and exhibited IC₅₀s below 50 μM (Table 2).

Table 2.

Properties of 13 DGC inhibitors

Compounda	IC50 (μM)
	VC2370	WspR	CIP
3	1.0	17.83	None
4	2.9	8.17	428.6
8	8.0	18.9	7559.0
10	0.9	12.2	None
12	1.4	69.7	334.4
13	2.9	39.94	93.9
14	1.1	137	None
15	9.8	30.0	201.9
18	6.6	13.5	283.8
19	6.9	17.5	280.3
22	34.6	2200	None
146	37.5	None	None
159	2.6	15.92	None

^aDIs are shaded.

We considered the possibility that the identified inhibitors might function nonspecifically by precipitating proteins, binding substrate GTP molecules, or interfering with the pyrophosphate detection assay. To test these possibilities, we developed an enzyme counterscreen that measured removal of phosphate from GTP by CIP using the EnzCheck Pyrophosphate assay, which also detects phosphate. Importantly, this counterscreen was identical to the DGC assays described above, except that the DGC was replaced with CIP. Of the nine general DGC inhibitors, only compound 13 inhibited CIP at an IC₅₀ of less than 200 μM, showing that eight compounds were specific antagonists of DGC activity (Table 2). To illustrate typical results for these experiments, concentration response curves for compounds 18 and 19 against VC2370(142)-D484E, WspR-R242A, and CIP are shown in Fig. 2. Upon further analysis, we observed that compound 159 significantly inhibited bacterial growth at a concentration of 100 μM, and the compound was not analyzed further. Thus, 7 compounds, which we named DGC inhibitors (DIs) (shaded in Table 2), were identified as general DGC inhibitors that do not significantly impair bacterial growth. The chemical structures and names of the DGC inhibitors are indicated in Fig. 3.

Fig 2.

Representative enzyme inhibition assays. Inhibition of the DGCs VC2370(142)-D484E (circles) and WspR-R242A (squares) and of CIP (triangles) at various inhibitor concentrations is shown for DI-18 and DI-19.

Fig 3.

Chemical structures of the DGC inhibitors.

The seven DGC inhibitors prevent biofilm formation of V. cholerae.

We next determined if the seven general DGC inhibitors possessed antibiofilm activity against V. cholerae using the MBEC biofilm assay. This system consists of a 96-well microtiter plate with 96 corresponding pegs attached to the plate lid. These pegs are immersed in the culture and provide a surface for biofilm formation. For these experiments, the V. cholerae ΔVC1086 mutant strain was utilized. VC1086 encodes a protein with an EAL domain that actively degrades c-di-GMP in V. cholerae (49), and the ΔVC1086 mutant exhibits slightly elevated levels of c-di-GMP compared to wild-type (WT) V. cholerae (see Fig. 6). Because WT V. cholerae has relatively low levels of c-di-GMP in the high-cell-density quorum-sensing state (49), analysis of the inhibition of in vivo DGC activity is more tractable in the ΔVC1086 mutant. Biofilm formation by a V. cholerae ΔvpsL mutant was simultaneously examined as a negative control, because disruption of the gene inhibits synthesis of extracellular polysaccharide production and biofilm formation (51). Indeed, all seven of the DGC inhibitors significantly inhibited biofilm formation (P < 0.0012) by V. cholerae when added at 100 μM (Fig. 4A).

Fig 6.

DI-3 and DI-10 significantly reduce the intracellular concentration of c-di-GMP. The intracellular concentrations of c-di-GMP in the WT, the ΔVC1086 mutant, and the ΔVC1086 mutant grown with 100 μM each inhibitor were determined by UPLC-MS-MS. The data are normalized to the untreated ΔVC1086 strain. The error bars indicate the standard deviations. *, P < 0.02 compared with the DMSO-treated control.

Fig 4.

Abilities of the seven DGC inhibitors to reduce biofilm formation in V. cholerae. (A) Biofilm formation of the V. cholerae ΔVC1086 mutant analyzed using an MBEC assay with and without 100 μM direct inhibitors. The ΔvpsL mutant of V. cholerae (vpsL), which cannot form biofilms, was a negative control. All strains/conditions were statistically significantly different from the ΔVC1086 mutant treated with DMSO (n = 6; P < 0.012). (B) Representative false-color flow cell image depicting the biofilm depth of V. cholerae, untreated or grown in 100 μM DI-3 and DI-10. (C) The biofilm biomass was determined by averaging nine separate images for each flow cell. The experiment was repeated 3 to 5 times for each treatment, and the graph displays the average biofilm biomasses with the associated standard deviations. *, P < 0.05; **, P < 0.001. 3, DI-3; 10, DI-10.

To more thoroughly examine the activities of select lead compounds, we synthesized DI-3, DI-10, DI-18, and DI-19 to test their abilities to inhibit biofilm formation in a continuous flow cell system. In this assay, bacteria formed biofilms on a glass surface under a constant flow of fresh medium with or without the test compound. These conditions more closely mimic natural biofilms that might form in environmental reservoirs or during infection of a human host. The total biofilm biomass formed by the ΔVC1086 strain in the absence and presence of 100 μM DI-3, DI-10, DI-18, and DI-19 was determined in nine random images for each individual flow cell using the software Comstat. Representative false-color images depicting the depth of the biofilm are shown in Fig. 4B. The experiment was repeated three to five times, and the average biofilm biomasses with the associated standard deviations are indicated in Fig. 4C. DI-18 and DI-19 showed no significant reduction of biofilm formation in the flow cell system (data not shown). However, treatment with DI-3 produced biofilms that showed 32% of the biofilm biomass of the untreated control (P < 0.001), whereas DI-10 treatment reduced biofilm formation to 60% of that of the untreated control (P < 0.05) (Fig. 4B and C).

DI-3 reduces biofilm formation by P. aeruginosa under flow.

We next examined if the DGC inhibitors we had identified were able to reduce biofilm formation by P. aeruginosa. We chose to examine P. aeruginosa, as the pathogen is known to evolve different hyper-biofilm-forming morphotypes during colonization of the lungs of cystic fibrosis patients (2). As a negative control, we utilized a pel fliA biofilm-deficient mutant. None of the seven DGC inhibitors were able to significantly reduce biofilm formation by P. aeruginosa in the MBEC biofilm formation assay (Fig. 5A). Because DI-3 and DI-10 inhibited biofilm formation by V. cholerae under flow conditions, we examined if these compounds showed similar activity against P. aeruginosa (Fig. 5B and C). Under flow, DI-3 significantly reduced the biofilm biomass of P. aeruginosa biofilms to 22% of that of the untreated control (P < 0.03). DI-10 did show a reduction in biofilm biomass to 74% compared with the untreated control, but this reduction was not statistically significant. Therefore, we conclude that DI-3 is also able to inhibit biofilm formation by P. aeruginosa under flow.

Fig 5.

Abilities of the seven DGC inhibitors to reduce biofilm formation in P. aeruginosa. (A) Biofilm formation by P. aeruginosa strain PAO1 with and without 100 μM direct inhibitors. The pel fliA mutant of P. aeruginosa, which cannot form biofilms, was the negative control. No treatments were statistically significantly different from the DMSO-treated control. (B) Representative false-color flow cell image depicting the depths of P. aeruginosa biofilms, untreated or grown in 100 μM DI-3 and DI-10. (C) The biofilm biomass was determined by averaging nine separate images for each flow cell. The experiment was repeated 3 to 5 times for each treatment, and the graph displays the average biofilm biomasses with the associated standard deviations. * = P < 0.03). 3, DI-3; 10, DI-10.

DGC inhibitors do not induce dispersal.

In the experiments thus far, the inhibitors were added concurrently with inoculation of the bacteria. To determine if the compounds could disperse preformed biofilms, V. cholerae biofilms were developed on MBEC pegs and then exposed to the seven lead compounds at 100 μM in fresh medium for short time intervals. After removal of the pegs, the amount of dispersed bacteria in the remaining media was quantified by determination of CFU or by performing comparative growth curves of the resultant suspension. In each case, there was no evidence of increased biofilm dispersal when biofilms were treated with DGC inhibitors compared with the DMSO controls. Therefore, at least under the conditions examined here, the DGC inhibitors we have identified do not disperse preformed V. cholerae biofilms.

DI-3 and DI-10 reduce the intracellular concentration of c-di-GMP in V. cholerae.

We hypothesized that the seven DGC inhibitors inhibit biofilm formation in V. cholerae by reducing the intracellular concentration of c-di-GMP. To test this hypothesis, the V. cholerae ΔVC1086 mutant was grown in the presence of 100 μM each compound or an appropriate DMSO control. The WT strain of V. cholerae was similarly examined. c-di-GMP was extracted and quantified by UPLC-MS-MS, as previously described (3), and the results are expressed as a percentage of the c-di-GMP observed in the DMSO-treated ΔVC1086 mutant. As expected, WT V. cholerae exhibited 50% of the levels of c-di-GMP in the ΔVC1086 mutant, as VC1086 is an active PDE in V. cholerae (Fig. 6). The concentration of c-di-GMP in the ΔVC1086 mutant treated with DI-3 and DI-10 was significantly reduced (P < 0.02) compared to the DMSO-treated control (Fig. 6). Alternatively, treatment with DI-4, DI-8, DI-15, DI-18, and DI-19 did not significantly alter the levels of c-di-GMP compared with the ΔVC1086 mutant (Fig. 6). Based on these results, we conclude that DI-3 and DI-10 inhibit biofilm formation in V. cholerae by reducing the total concentration of c-di-GMP in the cell, whereas DI-4, DI-8, DI-15, DI-18, and DI-19 function via different mechanisms (see Discussion).

Determination of the IC₅₀ of DI-3 for inhibition of biofilm formation.

Based on the above-described experiments, DI-3 is the most promising lead candidate, as it specifically inhibits two distinct DGCs in vitro, reduces biofilm formation in both V. cholerae and P. aeruginosa, and decreases the in vivo c-di-GMP concentration in V. cholerae. We determined the IC₅₀ at which DI-3 reduces biofilm formation by V. cholerae strain ΔVC1086 by triplicate analysis of a concentration series using the MBEC assay. These data indicated that the IC₅₀ of DI-3 was 26.2 μM with a 95% confidence interval of 15.1 to 45.6 μM (Fig. 7).

Fig 7.

Concentration response curve for DI-3. The IC₅₀ for the inhibition of V. cholerae biofilm formation in an MBEC assay by DI-3 was determined to be 26.2 μM, with a 95% confidence interval of 15.1 to 45.6 μM. The concentration response curve was generated in triplicate, and each point represents the mean value and standard deviation. The line is the best-fit curve as generated by the software Prism.

DI-3 exhibits druggable properties.

DI-3 possesses chemical properties that fall within the values of potential druggable molecules, as described by Lipinski and others (11, 22). The molecular mass of DI-3 is 288.35 g/mol, less than the 500-g/mol upper limit predicted to be optimal for small-molecule drugs. The predicted polar surface area of DI-3 is 41.125, and the predicted partition coefficient is 4.82, both of which fall within the optimal range. Finally, addition of up to 200 μM DI-3 to the THP-1 macrophage mammalian cell line showed no significant decrease in viable cells as measured by trypan blue dye exclusion, showing that the therapeutic index of DI-3 is at least 7.6 (Fig. 8).

Fig 8.

DI-3 is not toxic to mammalian cells. THP-1 macrophages were treated as indicated, and viability was measured at 8 h by trypan blue staining. The cells were killed by addition of 0.025% glutaraldehyde. The error bars indicate standard deviations.

DISCUSSION

Here, we report the identification of seven novel small molecules that inhibit DGC enzymes. Targeting c-di-GMP is an attractive strategy to develop antibiofilm compounds, as c-di-GMP is widely conserved among bacteria, and we expect that inhibitors of DGC enzymes will possess broad-spectrum antibiofilm activity. We are currently determining if the compounds described here can inhibit biofilm formation and reduce the levels of c-di-GMP in other bacterial pathogens that possess c-di-GMP signaling systems.

Our results suggest DI-3 is the most promising lead candidate, as it possesses broad-spectrum activity, reducing biofilm formation by both V. cholerae and P. aeruginosa. Interestingly, although DI-3 was able to significantly reduce biofilm formation in flow cell assays, we did not observe a significant reduction of biofilm formation in a static assay. A number of differences could account for this discrepancy. First, in the static assay, the medium is not replenished and the culture grows to stationary phase, ultimately using up all of the available nutrients. This is quite distinct from the flow cell assay, where nutrients are not limited. Therefore, induction of stationary-phase stress response genes could alter the response to DI-3. Moreover, if DI-3 weakens the structural integrity of the biofilm, the shear force generated during the flow cell experiment might be required to disrupt biofilm formation. Regardless of the reason, flow cell biofilm assays are generally thought to more closely mimic physiologically relevant conditions than microtiter-based biofilm assays.

The seven lead compounds show certain structural similarities. Their lengths range from 13 to 15 longest countable atomic linkages end to end. They are generally linear in shape, sharing similar steric demands. They each have one or two hydrogen bond-accepting oxo moieties and one or two hydrogen bond-donating moieties, as well. All lead compounds have two aryl moieties at either end of the molecules, suggesting the possibility of folding the molecule in half to achieve pi-pi stacking. c-di-GMP has been shown to undergo similar pi-pi stacking to form higher-order multimers, and these multimers are required for binding to RXXD allosteric sites (4, 10). Moreover, DI-15, DI-18, and DI-19 are structurally quite similar, with different substituents on one benzene ring. We are performing structural studies of these molecules with purified DGC enzymes to determine the molecular interactions driving the interaction of these molecules with DGCs. These results, combined with structure-activity analysis, will be used to develop more potent DGC antagonists that may lay the foundation for the development of a new class of antibiofilm compounds.

Addition of DI-3 and DI-10 was able to significantly reduce the intracellular concentration of c-di-GMP. The strain of V. cholerae used in this study, C6706str2, encodes 40 distinct DGC enzymes. Therefore, we hypothesize that these compounds must be able to inhibit multiple DGC enzymes in the bacterium. Although the remaining five compounds inhibited both VC2370(142)-D484E and WspR-R242A in vitro, addition of these molecules did not alter the in vivo global c-di-GMP levels. However, these compounds exhibited antibiofilm properties, raising the question of their mechanism of inhibition. One possibility is that these 5 compounds inhibit the activities of specific DGC enzymes that induce biofilm formation without affecting the global concentration of c-di-GMP. A number of lines of evidence suggest that the total intracellular concentration of c-di-GMP cannot be directly correlated with biofilm formation (17, 19, 30), and we have shown that the DGCs in V. cholerae do not induce biofilm formation via modulation of a global low-specificity signaling pool (25). These results suggest that c-di-GMP signaling in V. cholerae is a highly specific process consisting of multiple parallel signaling pathways. Therefore, these compounds might inhibit the activities of specific DGCs that are important in biofilm formation but that do not contribute to the global c-di-GMP concentration.

In addition to inducing biofilm formation through synthesis of c-di-GMP, enzymatically inactive DGCs and PDEs also function as c-di-GMP effector proteins that control biofilm formation in response to changes in c-di-GMP. For example, two enzymatically inactive DGCs, namely, VC0900 from V. cholerae, named CdgG (1), and PelD, encoded by P. aeruginosa (21), are both predicted to bind c-di-GMP via RXXD allosteric binding site motifs to control biofilm formation. In addition, the DGC FimX (28) and PDE LapD (29) encode c-di-GMP signaling proteins with degenerate active sites that bind to c-di-GMP to control biofilm formation posttranscriptionally. If the DGC inhibitors identified here mimic the structural properties of c-di-GMP, they could compete with c-di-GMP binding to degenerate DGC or PDE domains or other c-di-GMP effector proteins, such as transcription factors (20, 46). However, as these molecules are able to inhibit transcription of a c-di-GMP-induced gene, as evidenced by their identification in the original small-molecule screen, these effectors must function at the level of transcription.

Evidence is accumulating that decreases in c-di-GMP trigger dispersion from a biofilm. Exposure of P. aeruginosa to starvation conditions leads to biofilm dispersal and is associated with decreased levels of c-di-GMP (12, 40). Furthermore, mutation of the gene bdlA in P. aeruginosa prevents biofilm dispersal, possibly through elevated levels of c-di-GMP (27). In the related bacterium Pseudomonas fluorescens, c-di-GMP induces dispersal from biofilms through an inside-out signaling pathway mediated by the c-di-GMP effector protein LapD (26, 29). Thus, the available evidence suggests that, at least in the pseudomonads, decreased levels of c-di-GMP may be a biofilm dispersal signal. Under the conditions studied here, none of the DGC inhibitors promoted dispersal of V. cholerae biofilms, including DI-3 and DI-10, which decreased the levels of c-di-GMP. Whether these compounds induce dispersal under other conditions or in other bacteria is currently being investigated. Optimization of the interaction of these molecules with DGCs might generate novel compounds that are more effective at triggering dispersion from biofilms.

To our knowledge, the only other DGC inhibitor that has been reported was isolated from the garden pea (P. sativum). This molecule, a GTS, inhibits DGC activity of a DGC enzyme purified from Gluconacetobacter xylinus (formerly Acetobacter xylinum) via a noncompetitive reaction mechanism (31, 32). However, the molecule did not affect c-di-GMP cellulose synthesis of intact bacteria, suggesting that it could not cross bacterial membranes. In addition, the impact of GTS on intracellular c-di-GMP levels has not been examined. Therefore, this work describes the first molecules shown to reduce biofilm formation and to decrease the intracellular levels of c-di-GMP by direct inhibition of DGC enzymes. These molecules will form the foundation for the development of new compounds that prevent biofilm formation through targeting of c-di-GMP signaling.