Small-Molecule Inhibitor of Vibrio cholerae Virulence and Intestinal Colonization

See allHide authors and affiliations

Science  28 Oct 2005:
Vol. 310, Issue 5748, pp. 670-674
DOI: 10.1126/science.1116739


Increasing antibiotic resistance requires the development of new approaches to combating infection. Virulence gene expression in vivo represents a target for antibiotic discovery that has not yet been explored. A high-throughput, phenotypic screen was used to identify a small molecule 4-[N-(1,8-naphthalimide)]-n-butyric acid, virstatin, that inhibits virulence regulation in Vibrio cholerae. By inhibiting the transcriptional regulator ToxT, virstatin prevents expression of two critical V. cholerae virulence factors, cholera toxin and the toxin coregulated pilus. Orogastric administration of virstatin protects infant mice from intestinal colonization by V. cholerae.

We are entering a challenging era where microbial resistance to antibiotics will complicate the treatment of nearly all common bacterial infections. The development of antimicrobials has lagged behind the development of antibiotic resistance for many life-threatening bacterial species. Current antimicrobials, for the most part, target a relatively small number of essential gene functions, such as inhibition of cell wall synthesis, DNA replication, RNA transcription, protein synthesis, and folate synthesis. What development there has been has been largely limited to improving existing antibiotics through chemical modification and developing synergistic drugs that augment the efficacy of existing antibiotics. Nevertheless, a recent report identifying a new antituberculosis drug suggests it is still possible to identify entirely new classes of antibiotics (1).

Targeting the regulation of virulence factors of specific pathogens represents one approach to identifying antibiotics. Here we report the identification of a small molecule inhibitor of Vibrio cholerae virulence regulation and demonstrate its ability to inhibit bacterial colonization in an animal model of cholera.

Cholera is caused by the Gram-negative bacterium V. cholerae, which elaborates two major virulence factors, cholera toxin (CT) and the toxin coregulated pilus (TCP) (2). CT is an adenosine diphosphate–ribosylating toxin encoded in the genome of the filamentous, lysogenic CTXΦ phage. Secretion of CT into the intestinal lumen results in elevated cyclic adenosine monophosphate levels in intestinal epithelial cells and the subsequent secretory diarrhea that is the hallmark of the disease. TCP is thought to play a role in early attachment to intestinal epithelium and is required for intestinal colonization in an infant mouse model of cholera (2). The regulation of CT and TCP expression has been studied extensively (2), but the mechanisms by which environmental signals stimulate virulence expression in vivo are not clear.

We performed a high-throughput screen of a 50,000-compound small molecule library from Chembridge Research Laboratories (Microformats) to identify inhibitors of V. cholerae virulence factor expression. We constructed a screening strain of the classical biotype strain O395 that carries a chromosomally integrated tetracycline resistance gene (tetA) controlled by the cholera toxin (ctx) promoter. In a 384-well format, the strain was treated with 10 μg/mL samples from the library to identify compounds that conferred tetracycline sensitivity when the strain was grown under in vitro conditions that would normally induce CT expression. We identified 109 compounds as inhibitors of ctx promoter activation by their ability to prevent tetA expression in the reporter strain. Fifteen of these compounds were notable for their inhibitory effects with minimal bacterial toxicity.

One of these 15 compounds, 4-[N-(1,8-naphthalimide)]-n-butyric acid (virstatin) (Fig. 1A) was selected for further study and was subsequently synthesized in gram quantities (3). We generated growth curves for two different V. cholerae biotype strains, O395 and C6706, and showed that virstatin did not inhibit growth (50 μM) (Fig. 1B). Minimal bacteriocidal concentrations (MBC) of 600 and 1200 μM were determined for the two strains, respectively. We confirmed that under in vitro virulence-inducing conditions for these strains, CT production was undetectable in the presence of virstatin (50 μM), as assayed by CT-enzyme-linked immunosorbent assay (CT-ELISA) (Fig. 1C). The minimal inhibitory concentrations for CT expression were 3 μM and 40 μM for O395 and C6706, respectively.

Fig. 1.

Virstatin inhibits virulence expression in V. cholerae. (A) Chemical structure of virstatin. (B) Growth curves in LB at 37°C for O395 (dotted line) and C6706 (solid line). Dimethyl sulfoxide (DMSO) control, blue; virstatin, pink; OD600, optical density measured at 600 nm. (C) CT and TcpA expression in the presence or absence of virstatin. Top: CT expression was measured by ELISA from O395 and C6706 in the presence of DMSO control or virstatin (50 μM) after overnight growth under standard virulence-inducing conditions. Error bars represent the standard deviations from samples performed in triplicate. n.d., not detected. Bottom: TcpA expression was detected by Western blot with an α-TcpA antibody. (D) CTX-KmΦ transduction of O395 in the absence and presence of virstatin. Functional TCP was assayed by growing O395 to mid-log phase under virulence-inducing conditions (LB at pH 6.5 and 30°C) in the presence of DMSO control or virstatin (50 μM). After 30 min incubation with phage, cultures were plated on LB-Kanamycin plates to enumerate transduction events. Total cells and kanamycin-resistant (KanR) cells were counted to reflect transduction efficiency. In the presence of virstatin, no transduction events were measured, which is a 6-log reduction compared to DMSO control.

Because TCP is coregulated with CT, we examined virstatin's ability to inhibit the expression and function of TCP. Western blot analysis showed that the major subunit of TCP (TcpA) was not expressed under virulence-inducing conditions when either O395 or C6706 was treated with virstatin (Fig. 1C). Virstatin prevented the assembly of a functional pilus (TCP) in O395, as determined by its ability to prevent transduction of a bacteriophage (CTX-KmΦ) that uses TCP as its receptor (Fig. 1D) (4). CTX-KmΦ encodes kanamycin resistance and thus confers resistance when transduced into V. cholerae. Although 2.6 × 106 kanamycin-resistant colonies were obtained from phage transduction of O395 grown under virulence-inducing conditions, no kanamycin-resistant colonies were obtained when O395 was grown in the presence of virstatin, suggesting the absence of TCP on these cells.

We next examined whether virstatin affects known regulators of CT and TCP expression. In the canonical model for V. cholerae virulence regulation, environmental signals stimulate virulence by activating a cascading series of transcriptional regulators that ultimately induce transcription of the ctx and tcp genes (Fig. 2A) (5). The major role of regulators ToxRS and TcpPH is to activate transcription of toxT, which encodes an AraC-like transcriptional activator (6). ToxT activates multiple virulence genes (including ctxAB, acfA, and the tcp genes) and is autoregulated, presumably by self-inducing read-through transcription from the upstream tcpA promoter (7).

Fig. 2.

Virstatin inhibits ToxT post-transcriptionally. (A) Cascade model for CT and TCP regulation. Environmental signals are transduced by AphAB, TcpPH, and ToxRS, of which the latter two transcriptionally activate toxT. ToxT then activates ctx and tcp transcription. (B) Virstatin inhibits ctx but not toxR, tcpP, or toxT transcription. Transcriptional reporter strains (O395) were grown overnight under virulence-inducing conditions (pH 6.5 and 30°C) in the presence of DMSO control or virstatin (50 μM). Transcriptional fusions of toxR, toxT, and ctxA with lacZ were assayed for β-galactosidase activity. A transcriptional fusion of tcpP with uidA was assayed for β-glucuronidase activity. Data are presented as the percentage of reporter activity in the presence of virstatin compared to the DMSO control. Error bars represent the standard deviation for samples performed in triplicate.

We investigated the effect of virstatin on these known virulence regulators by examining the transcription of toxRS, tcpPH, and toxT using transcriptional fusion reporters (Fig. 2B). In the presence of virstatin, the transcript levels of toxRS, tcpPH, and the most downstream gene, toxT, were all relatively unaffected, as determined by measuring β-galactosidase or β-glucuronidase activity. As expected, ctx transcription was notably decreased. Thus, virstatin blocks ctx transcription downstream of toxT transcription.

We also examined the effect of 50 μM virstatin on V. cholerae global transcription patterns by using a genomic microarray (8). Eleven of 15 genes in O395 and 21 out of 22 genes in C6706 significantly repressed by virstatin (to levels less than one-third of control levels) are within the tcp or ctx loci (3). These findings are consistent with the results obtained with transcriptional reporters, with ToxT regulating most genes that are down-regulated by virstatin. Moreover, in accordance with the established model for CT and TCP regulation, virstatin appears to inhibit ToxT post-transcriptionally.

Virstatin inhibited ToxT activity when ToxT was expressed under the control of a heterologous pBAD promoter, induced by arabinose, in V. cholerae strain O395ΔtoxT (Fig. 3A). To confirm that virstatin has no effect on ToxT expression, we constructed a toxT variant containing a C-terminal Myc5 tag that both is active (albeit slightly less active than wild-type) and can be inhibited by virstatin (Fig. 3A) (3). Western blot analysis showed comparable levels of ToxT-Myc5 expression in the presence and absence of virstatin, confirming that the inhibition is post-translational.

Fig. 3.

Virstatin inhibits ToxT. (A) Top: Virstatin inhibits CT expression in O395 when ToxT is expressed under pBAD promoter control. O395ΔtoxT, O395ΔtoxT with pBAD24, pBAD24-toxT, pBAD24-toxT-myc5, or pBAD24-toxTL113P were induced (with 0.001% arabinose) in the presence of DMSO or virstatin (50 μM). Virstatin inhibited CT expression in cells expressing wild-type and Myc-tagged ToxT but not ToxTL113P. DMSO, black; virstatin, white. Bottom: Western blot with α-Myc demonstrates that virstatin does not alter ToxT-Myc5 expression. (B) Top: Virstatin inhibits ctx transcription in E. coli reporter strain DTH3060 carrying ctx-lacZ, as measured by β-galactosidase activity. Constitutive ToxT (pEP99.1) and ToxT-Myc5 (pEP99.2) expression under tet promoter control resulted in 6 to 10 times more than in induction over control strain DTH3060 without or with control plasmid (pJB658). Virstatin repressed induction to control levels. DMSO, black; virstatin, white. Bottom: Western blot with α-Myc demonstrates that virstatin does not alter ToxT-Myc5 expression. (C) Virstatin, at increasing concentrations, inhibited β-galactosidase activity in DTH3060 when wild-type ToxT but not ToxTL113P was expressed (under pBAD promoter control; induced with 0.1% arabinose). Activity is presented as percentage of reporter activity in the presence of varying concentrations of virstatin compared to no virstatin. ToxT, solid line; ToxTL113P, dotted line.

Virstatin also inhibited ToxT induction of ctx-lacZ in Escherichia coli reporter strain DTH3060 free of all other V. cholerae factors that might otherwise affect ctx induction (Fig. 3B) (9, 10). When ToxT was expressed constitutively from plasmid pEP99.1 (3), β-galactosidase activity was induced 6 to 10 times more than in strain DTH3060 without and with control plasmid pJB658 (11). With virstatin present, this induction was suppressed to baseline levels. We obtained similar results using ToxT-Myc5 (pEP99.2) and performed Western blot analysis to confirm that ToxT expression was not altered by virstatin.

A virstatin-resistant mutant of ToxT was isolated from a library of ToxT mutants by propagating pBADtoxT in the E. coli mutator strain XL1-Red (Stratagene), transforming the resulting plasmid library into the reporter strain DTH3060, and screening the resulting colonies on LB agar containing virstatin, arabinose, Xgal, ampicillin, tetracycline, and kanamycin. One intensely blue colony was isolated from ∼20,000 colonies screened, indicating a clone that expressed a ToxT mutant capable of inducing ctx-lacZ in the presence of virstatin. Sequencing of the toxT gene revealed a single point mutation, L113P, that occurs in the N-terminal, putative dimerization domain based on its sequence homology to other AraC-like proteins.

ToxTL113P was resistant to virstatin when expressed in O395ΔtoxT with equivalent amounts of CT produced in the absence or presence of virstatin (50 μM) (Fig. 3A). A similar phenomenon was observed in the heterologous E. coli strain DTH3060, with no inhibition of ToxTL113P by increasing concentrations of virstatin (until 60 μM), in stark contrast to the inhibition observed of wild-type ToxT (Fig. 3C). These data demonstrate that ToxT carrying a mutation in the N-terminal domain displays relative resistance to virstatin. Because the N-terminal domain of ToxT is its putative dimerization domain, virstatin may alter the dimerization state of ToxT, thus inactivating it.

In order to determine if the inhibition of virulence observed in vitro could affect in vivo infection, we tested the ability of virstatin to inhibit V. cholerae infection in an animal model. It has previously been shown that deletion of toxT attenuates TCP-dependent colonization of the small intestine of infant mice by V. cholerae (12).

We examined the effect of virstatin on the colonization of infant mice by V. cholerae strains that colonize in a TCP-dependent versus -independent manner. V. cholerae El Tor biotype strain C6706 was used as the TCP-dependent strain, because of its similar growth kinetics in the presence and absence of virstatin to TCP-independent V. cholerae strain S533. S533 is a non-O1 non-O139 clinical isolate that lacks the toxT, tcp, and ctxAB genes but is nevertheless able to colonize infant mice (13).

To validate comparison of the two strains, we examined their competitive capacity in vitro, in the presence and absence of virstatin, by growing S533 and C6706 together. The competitive index (CI) (14) of S533 versus C6706 was close to 1 in the presence and absence of virstatin (3).

Inoculation with single strains into the infant mouse in the presence of virstatin demonstrated a marked reduction in colonization of C6706 but not S533. Infant mice (5 to 6 days old) were orogastrically inoculated as previously described with V. cholerae strain C6706 or S533 (15) in the presence or absence of virstatin and killed at 18 to 24 hours (3). Small intestine homogenate from each mouse was plated on LB-agar containing streptomycin for enumeration of live bacterial counts. Under optimized conditions, S533 colonization was not affected by the presence of virstatin. However, C6706 colonization dropped by four logarithms in the presence of virstatin (Fig. 4A).

Fig. 4.

Virstatin inhibits ToxT-dependent colonization of infant mice. (A) When inoculated alone, C6706 wildtype (C6706wt) colonization was reduced 4 logs in the presence of virstatin under conditions that do not affect S533 or C6706toxTL113P (C6706mut) colonization. Bacteria were recovered from mice 18 to 24 hours post-orogastric inoculation and plated for enumeration. Each data point represents the output from a single animal and the bar represents the log of the geometric mean of data obtained from individual mice. Control buffer inoculum, no boost, blue; control buffer inoculum, control buffer boost, green; virstatin in both inoculum and boost, pink. CFU, colony-forming units. (B) When strains were co-inoculated, virstatin increased the CI of S533 versus C6706wt by 4.5 logs, and the CI of C6706mut versus When C6706wt and C6706DtcpA C6706wt by 1.5 logs. No virstatin, blue; virstatin, pink. (C) were co-inoculated, virstatin decreased recovery of C6706wt by 3 logs, down to the levels of C6706DtcpA. Recovery of the two strains in the presence of virstatin at 24 hours was nearly 1:1, whereas no C6706DtcpA was recovered from a competition experiment in the absence of virstatin. C6706wt, solid; C6706DtcpA, dotted; buffer, blue; virstatin, pink. (D) When C6706wt infection was allowed to establish for 12 hours and mice then were treated with virstatin, colonization was reduced 3 logs in comparison to control buffer–treated mice. Control boost, blue; virstatin, pink.

Competition experiments comparing the relative ratios of C6706 and S533 recovered after co-inoculation into the infant mouse showed the same selective effect of virstatin on C6706 but not S533 as the single-inoculation studies. A mixture of S533 and C6706 in the presence or absence of virstatin was orogastrically inoculated into infant mice. In the absence of virstatin, each mouse was colonized by both strains (CI 0.3 to 35). In contrast, in the presence of virstatin, very few C6706 colonies could be recovered. The CI of S533/C6706 increased over four logarithms (Fig. 4B). Thus, virstatin is able to significantly attenuate the TCP-dependent infection of C6706 relative to S533, a strain whose colonization is independent of ToxT and insensitive to the activity of virstatin.

Because virstatin's inhibition of ToxT and subsequent TCP expression is the likely cause of attenuation in C6706 colonization, we examined the ability of a C6706ΔtcpA strain to compete with the wild type in the presence of virstatin (Fig. 4C). Mice were orogastrically inoculated with or without virstatin and boosted again at 3 and 6.5 hours post-inoculation. Mice were killed at 2, 4.5, 7.5, and 24 hours post-inoculation. In the absence of virstatin, we observed the previously described, severely attenuated phenotype of the ΔtcpA strain (16) and were unable to recover any C6706ΔtcpA bacteria at 24 hours, whereas C6706 wild-type colonized efficiently (Fig. 4C). However, in the presence of virstatin, recovery of wild-type C6706 was significantly diminished, and nearly equivalent numbers of ΔtcpA and wild-type bacteria were recovered at all time points (Fig. 4D). Thus, virstatin is able to eliminate wild-type C6706's competitive advantage over C6706ΔtcpA during in vivo infection.

In vivo studies with C6706toxTL113P, a mutant of wild-type C6706 carrying the mutation in the chromosome at the native toxT locus, confirmed that the differences in colonization are due to the effect of virstatin on ToxT. When inoculated alone, C6706toxTL113P colonization was unaffected by the presence of virstatin (Fig. 4A). When inoculated together in the presence of virstatin, C6706toxTL113P was able to colonize better than wild-type C6706 with a CI of 50 (Fig. 4B) [in vitro CI ∼1, (3)]. Together, these data demonstrate that virstatin inhibits intestinal colonization specifically by blocking the activity of ToxT in vivo.

Finally, we examined whether virstatin could affect long-term infection if administered after colonization has already been established in the infant mouse model. This effect would be analogous to treatment of cholera patients with antibacterials such as tetracycline after the onset of diarrhea, which can reduce the duration of symptoms (17). We found that delayed administration of virstatin 12 hours after inoculation with C6706 still reduced the recovery of C6706 by over three logarithms relative to C6706 recovered from untreated infant mice (Fig. 4D). These data complement prior observations on the requirement of early in vivo virulence expression (18) and suggest that ongoing, late expression of ToxT-dependent genes is also necessary for optimal colonization in this animal model. This result also demonstrates that drugs such as virstatin, like conventional antibacterials, could have utility even after disease has been diagnosed.

Other small molecule inhibitors of virulence regulation have been reported, including inhibitors of a two-component regulator of alginate synthesis in Pseudomonas aeruginosa (19) and inhibitors of quorum sensing in Staphylococcus aureus (20) and P. aeruginosa (21). Inhibitors of virulence factors have also been explored, including compounds that block type III secretion in Yersinia (22) or anthrax toxin protease activity (23). Here we show that even in the absence of any chemical or target structural information, a high-throughput phenotypic screen can be used to identify small molecule virulence inhibitors that exhibit in vivo efficacy against bacterial infection after simple orogastric administration. Thus, identification of inhibitors of virulence represents a path to anti-infective discovery that is quite different from conventional approaches that target only bacterial processes that are essential both in vivo and in vitro. We further predict that drugs such as virstatin may act synergistically with conventional antibiotics, because they act through independent mechanisms to block in vivo bacterial replication or survival.

Supporting Online Material

Materials and Methods

SOM Text

Tables S1 to S3

References and Notes

References and Notes

View Abstract

Stay Connected to Science

Navigate This Article