Reprofiled anthelmintics abate hypervirulent stationary-phase Clostridium difficile

Prolonged use of broad-spectrum antibiotics disrupts the indigenous gut microbiota, which consequently enables toxigenic Clostridium difficile species to proliferate and cause infection. The burden of C. difficile infections was exacerbated with the outbreak of hypervirulent strains that produce copious amounts of enterotoxins and spores. In recent past, membrane-active agents have generated a surge of interest due to their bactericidal property with a low propensity for resistance. In this study, we capitalized on the antimicrobial property and low oral bioavailability of salicylanilide anthelmintics (closantel, rafoxanide, niclosamide, oxyclozanide) to target the gut pathogen. By broth microdilution techniques, we determined the MIC values of the anthelmintics against 16 C. difficile isolates of defined PCR-ribotype. The anthelmintics broadly inhibited C. difficile growth in vitro via a membrane depolarization mechanism. Interestingly, the salicylanilides were bactericidal against logarithmic- and stationary-phase cultures of the BI/NAP1/027 strain 4118. The salicylanilides were poorly active against select gut commensals (Bacteroides, Bifidobacterium and Lactobacillus species), and were non-hemolytic and non-toxic to mammalian cell lines HepG2 and HEK 293T/17 within the range of their in vitro MICs and MBCs. The salicylanilide anthelmintics exhibit desirable properties for repositioning as anti-C. difficile agents.

membrane-active agents have demonstrated potential in eliminating stationary-phase C. difficile cells, which subsequently led to a substantial decrease in toxin production and sporulation 16,18,19 .
Salicylanilides have been reported to exhibit antimicrobial properties 20,21 albeit they are chiefly exploited as antiparasitic agents. Closantel (1), rafoxanide (2), niclosamide (3) and oxyclozanide (4) represent four of the widely used salicylanilide anthelmintics (Fig. 1). Niclosamide is an FDA-approved drug for the treatment of tapeworm infections, while the other three are marketed as veterinary drugs for liver fluke/roundworm infections in ruminants 22 . The exact antibacterial mode of action of salicylanilides is not well defined but is thought to involve dissipation of the (trans)membrane potential or the proton motive force (pmf) 23 . The pmf modulates the spatial organization of morphogenetic proteins 24 as well as ATP homeostasis that is vital for bacterial survival 25 . These functions of the pmf offer an explanation for the effects observed with certain membrane-active compounds, albeit inhibition of which does not always result to cell death in many bacterial pathogens 26 . The potential use of salicylanilides as antimicrobials has drawn considerable interest as exemplified by recent studies demonstrating the anti-staphylococcal properties of closantel, niclosamide and oxyclozanide 27,28 . A limiting aspect is the low oral bioavailability of salicylanilides, which may render them ineffective in treating systemic infections. For instance, niclosamide was found to be only partially absorbed from the GI tract (with a maximal serum concentration ranging from 0.25 to 6 μ g/mL after oral administration to human volunteers) and was also poorly distributed to tissues 29 . Closantel, rafoxanide and oxyclozanide exhibited similar pharmacokinetic (PK) attributes and were minimally metabolized and mostly excreted unchanged (up to ~90% for closantel) in the feces in ruminants 22 .
Despite the seemingly "unfavorable" PK profile, the salicylanilides may find niche therapeutic utility where poor GI absorption is desired. Herein, we provide evidence that the gut pathogen C. difficile makes for a viable target of such compounds. We show that the salicylanilide derivatives efficiently inhibited the growth of C. difficile via membrane depolarization, and more importantly, killed both logarithmic-and stationary-phase cells in a concentration-dependent manner. The bactericidal property against stationary-phase C. difficile could in principle lower the production of toxins and spores, which may in turn lead to improved response and mitigate CDI recurrence.

Results and Discussion
In vitro susceptibilities of C. difficile isolates. Over the last decade, C. difficile ribotype 027 was the predominant ribotype found in CDI fecal samples across North America and Europe 30 , however in recent years, other ribotypes are becoming increasingly prevalent 31 . We evaluated sixteen C. difficile clinical isolates of defined PCR ribotype (including the more prevalent ribotypes 027, 002, 014, 020, 078 and 106) 31 for their in vitro susceptibilities to anthelmintics closantel, rafoxanide, niclosamide and oxyclozanide (Supplementary Table S1). The salicylanilides displayed broad activity against C. difficile, with MIC as low as 0.06-0.13 μ g/mL for rafoxanide (Table 1). Based on the MIC against 16 isolates of C difficile, we determined the MIC 90 values to be 0.25, 0.13, 2 and 1 μ g/mL for closantel, rafoxanide, niclosamide and oxyclozanide, respectively (Table 1 and Supplementary  Table S1). Metronidazole and vancomycin were also included as controls, which displayed MIC ranges of 0.13-0.25 μ g/mL and 0.5-2 μ g/mL, respectively, against the C. difficile isolates (Table 1 and Supplementary Table S1).

Salicylanilides inhibit C. difficile growth via membrane depolarization.
In order to ascertain that the observed activity of the salicylanilides occurs through dissipation of the bacterial membrane potential, we prepared analogues 5 and 6 ( Fig. 2) as previously described 32 , and evaluated their growth inhibitory activity against C. difficile strains 630 (CD630, ATCC BAA-1382-FZ, ribotype 012) and 4118 (CD4118, ATCC BAA-1870, ribotype 027). As a side note, CD630 is a virulent, multidrug-resistant strain whose genome has been completely sequenced 33 , while CD4118 is a BI/NAP1/027 hypervirulent pathogen. We have earlier delineated the structural features that are necessary for protonophoric activity of salicylanilides, requiring both a dissociable phenolic OH group and an amide proton that forms an intramolecular hydrogen bond to maintain hydrophobicity and stabilize the anionic form of the molecule (see Supplementary Fig. S1) 32 . The MIC values that were determined for 5 and 6 are consistent with a membrane depolarization mechanism as the compounds devoid of protonophoric activity [i.e. analogues that lack either the weakly acidic OH (5b, 5c, 5h, 5i, 6b and 6c) or the amide proton (5d)] were inactive, whereas protonophores 5a, 5e, 5f, 5g, and 6a exhibited in vitro activity against CD630 and CD4118 ( Table 2). Encouraged by these results, we explored several other derivatives, which harbor the diidosalicylate moiety coupled to varying substituents including biphenyl (7a), halogenated mono-aryl rings (7b-d), a fused-ring fluorenyl core (7e) and the more flexible ethylbenzenes (7f-i). Compounds 7a-i demonstrated low MIC values (≤ 2 μ g/mL), except for the ortho-chloro analogue 7c, which showed reduced in vitro activity against CD630 and CD4118 (MIC = 8 μ g/mL). Replacement of the diiodosalicylate with its dichloro congener 8 resulted in a 4-fold decrease in MIC relative to 5g (Table 2). Compound 8 was also active against 14 other C. difficile isolates, with MIC values ranging from 0.03 to 0.25 μ g/mL (Supplementary Table S1). The foregoing observations led us to probe other ionophores such as tropolones 9a,b 34 and β -carbolines 10a,b 35 , as well as other structurally related compounds lacking the salicylanilide moiety (compounds 11a-f and 12, Supplementary Fig. S2); however, none of these were found to be active against CD630 and CD4118 (MIC > 32 μ g/mL, Supplementary Table S2).
An attractive feature of membrane-active compounds is their low propensity for resistance 17 . In order to assess the potential for resistance development in vitro, we conducted serial passaging of CD4118 at sub-inhibitory concentrations of the salicylanilides. A total of 21 serial passages were performed and the results are shown in Supplementary Fig. S3. The changes in MICs after serial passages were minimal; final MICs of the salicylanilides were ≤ 4-fold higher than the initial (pre-passage) MICs, suggesting that emergence of resistance is likely to be low. The same held true for both metronidazole and vancomycin, with no substantial increase in MIC (~2-fold) even after 21 passages of CD4118 ( Supplementary Fig. S3).  Salicylanilides are bactericidal against logarithmic-and stationary-phase cultures. Although ionophores are known to dissipate the pmf that is crucial for bacterial energy metabolism, they do not always display bactericidal activity 26,28 . We were interested whether the salicylanilides possess cidal properties against stationary-phase C. difficile, because these cells are the primary producer of toxins and spores that contribute to C. difficile pathogenesis 4 . We selected the more active compounds (closantel, rafoxanide and 8) and assayed them for minimum bactericidal concentration (MBC, defined as the lowest concentration of the antibacterial agent required to kill ≥ 99.9% of the initial inoculum) against growing and non-growing cells of the BI/NAP1/027 pathogen CD4118. As shown in Table 3, all three compounds displayed bactericidal activities against both logarithmic-and stationary-phase cells of CD4118 at concentrations close to their MIC values. The MBC log values of the protonophores were determined to be 0.25-2 μ g/mL (~4 to 8-fold greater than their respective MIC values). Importantly, the salicylanilides retained bactericidal activities against stationary-phase C. difficile cells, in stark contrast to metronidazole and vancomycin, which did not result in ≥ 3-log reduction of CD4118 cells at 32 μ g/mL (Table 3). Next, we determined the time-kill kinetics of closantel, rafoxanide and 8 (at 1× , 4× , and 16× their respective MICs) against stationary-phase cultures of CD4118. As depicted in Fig. 3, all three salicylanilides showed a concentration-dependent mode of killing of the quiescent cells. At 16× the MIC of each protonophore, rafoxanide (at 2 μ g/mL) eradicated > 99.9% of viable cells in 6 h (Fig. 3b), while closantel (at 4 μ g/mL) and compound 8 (at 1 μ g/mL) achieved a similar activity in 24 h (Fig. 3a,c). At four-fold lower concentrations (i.e. 4 × MIC), rafoxanide caused a 2.7-log decrease in the number of CFUs in 24 h, comparable to those of closantel and 8, which reduced bacterial cell viability by 2.2-and 2.4-log, respectively. In comparison, neither metronidazole (at 4 μ g/mL) nor vancomycin (at 32 μ g/mL) reached ≥ 3-log killing of CD4118 stationary-phase cells, even at 16× their   respective MIC values (Fig. 3d). The rapid bactericidal property demonstrated by closantel, rafoxanide and 8 is an interesting finding because quiescent C. difficile cells are notoriously recalcitrant to antibiotic-mediated killing 16 . We surmise that the cidal effect of such protonophores on stationary-phase C. difficile cells would ameliorate the effect of toxin production and spore formation, similar to what was observed with other membrane-active compounds 16 .
Antimicrobial spectrum of salicylanilides. In an effort to assess the antibacterial spectrum of the salicylanilides, we evaluated representative compounds (closantel, rafoxanide, 6a, 7b, 8 Table S3). Moreover, the compounds were ineffective against aerobic Gram-negative bacteria (MIC ≥ 32 μ g/mL against A. baumannii M2 and P. aeruginosa PAO1) and had modest MIC values of ≥ 4 μ g/mL against gut commensals B. fragilis CL05T00C42, B. thetaiotaomicron ATCC 29148, P. distasonis ATCC 8503 and P. nigrescens ATCC 33563. These results are consistent with those of niclosamide and oxyclozanide, which were shown to primarily target Gram-positive bacteria 28 . Compound 5i, which does not possess protonophoric activity 32 , lacked antibacterial activity whereas metronidazole and vancomycin mainly targeted anaerobic bacteria and Gram-positive organisms, respectively (Supplementary Table  S3). The complex multilayered cell envelopes of Gram-negative organisms impose a permeability barrier to antimicrobial agents and most likely account for the diminished activity observed for the salicylanilide molecules. Of note, rafoxanide had MIC values of ≤ 0.13 μ g/mL for C. difficile (Supplementary Table S1), which rendered ≥ 32-fold selectivity over Bifidobacterium/Lactobacillus strains and Gram-negative gut commensals (Bacteroides species) that were tested (Supplementary Table S3). These results warrant further investigations in vivo to fully decipher the effect of the salicylanilides on the gut microbiome.
In vitro cytotoxicity and hemolytic activity of salicylanilides. Although the salicylanilides have been used extensively in veterinary medicine, there is little information available concerning their biological effects on humans, except for niclosamide, which is FDA-approved for treatment of intestinal cestode infections. In order to gauge potential cytotoxicity of the salicylanilides, hemolysis using sheep erythrocytes and MTS 36 assay using two human cell lines (liver carcinoma HepG2 and embryonic kidney HEK 293T/17) were performed. An important finding was that the salicylanilides (closantel, rafoxanide, niclosamide, oxyclozanide and compound 8) did not cause rupture of red blood cells when tested at 32 μ g/mL (Supplementary Fig. S4). However, treatment of human cell lines with niclosamide led to a marked decrease in viability even at a low concentration of 0.125 μ g/mL ( Supplementary Fig. S5). Despite its high in vitro cytotoxicity, niclosamide is considered a "safe drug" because of its minimal absorption from the GI tract and high plasma protein binding 29 , thus sparing the host cells from its uncoupling property. An intriguing observation was the comparably lower in vitro toxicities of compound 8 and the veterinary drugs (closantel, rafoxanide, oxyclozanide) toward HepG2 and HEK 293T/17 ( Supplementary Fig. S5). For example, rafoxanide had no apparent effect on mammalian cell viability even at a concentration of 8 μ g/mL, which is ≥ 64-fold higher than its MIC values against C. difficile (Supplementary Table S1). We note that these results do not guarantee drug safety (relative to niclosamide) but nevertheless indicate the potential for repositioning of the veterinary anthelmintics as human drugs. A common cause of antibiotic failure is the inadequate penetration of the target infection site. In the case of CDI, it is imperative that the active drug achieves therapeutic levels in the colon to repress or eliminate the outgrowth of toxigenic C. difficile. This places the salicylanilide anthelmintics at a definite advantage; their low oral bioavailability and high fecal excretion (as observed in ruminants and humans) 22,29 would in theory result in adequate gut concentrations necessary to disarm the target pathogen. A substantial feature of the salicylanilides (as we have shown for closantel, rafoxanide and 8) is their bactericidal activity against stationary-phase cultures of hypervirulent C. difficile-a property that is not exhibited by many antibiotics including metronidazole and vancomycin 16 . Killing of hypervirulent stationary-phase C. difficile could likely suppress toxin production and sporulation, which in principle may lead to an improved sustained response and reduced recurrence rate. The clinical potential of membrane-active agents is demonstrated by daptomycin and telavancin, which function through permeabilization/depolarization of bacterial membranes and are FDA-approved to treat complicated skin and skin structure infections 37,38 . Our results exemplify notable attributes of membrane-active salicylanilide anthelmintics and demonstrate their potential for repurposing as anti-Clostridium difficile agents.

Determination of minimum inhibitory concentration (MIC). MIC values were determined in
cation-adjusted Mueller-Hinton broth (CAMHB) for bacteria that grow aerobically (B. subtilis, S. aureus, S. epidermidis, A. baumannii, P. aeruginosa) or in CAMHB supplemented with 5% laked horse blood for the Lactobacillus species, in accordance to CLSI guidelines 39,40 .
Bacteria that belong to the B. fragilis group (B. fragilis, B. thetaiotaomicron, P. distasonis), including the closely related P. nigrescens, were grown in Brucella broth supplemented with hemin (5 μ g/mL), vitamin K 1 (1 μ g/mL) and 5% lysed horse blood, according to CLSI guidelines 41 . For growth of Bifidobacterium species, the broth used was lactic acid bacteria susceptibility medium (LSM, consisting of 90% v/v Iso-Sensitest broth and 10% v/v MRS broth) containing 0.03% L-cysteine, consistent with ISO guidelines 42 . MIC assays for all Clostridium species were performed in brain-heart infusion broth supplemented with 0.5% yeast extract (BHIS) containing 0.03% L-cysteine, as previously described 18,43 . Anaerobic bacterial cultures were performed in an anaerobic cabinet (Coy Lab Products Inc., Grass Lake, MI, USA) at 37 °C in a reducing anaerobic atmosphere (8% H 2 , 8% CO 2 , 84% N 2 ). All broths and 96-well microtiter plates were pre-reduced (incubated anaerobically overnight) prior to use for anaerobic culture.
All MIC determinations were performed in 96-well microtiter plates using the broth microdilution method. A summary of bacterial growth medium and conditions is tabulated in the Supplementary Procedures. Briefly, two-fold serial dilutions of test compounds were inoculated with ~5 × 10 5 cfu/mL bacteria. MIC was recorded as the lowest concentration of the test compound that inhibited visible bacterial growth after 20 Serial passage experiments. C. difficile strain 4118 serial passages were performed in BHIS containing 0.03% L-cysteine, using the broth microdilution method. The MIC for each compound was first determined according to the procedure detailed above. After incubation for 20-24 h at 37 °C in an anaerobic chamber, the treated culture growing at 0.5 × MIC was used to inoculate the subsequent passage and this process was repeated a total of 21 times. Serial passage experiments were performed in duplicate.

Determination of minimum bactericidal concentration (MBC).
The procedure for MBC determination was adapted from that of Alam et al. 44 . Briefly, a single overnight colony of C. difficile strain 4118 (streaked onto blood agar plate) was suspended in BHIS and grown to OD 600 ~ 0.4-0.5 (logarithmic phase) or for 24 h (stationary phase). In 96-well microtiter plates, cultures were added to two-fold serial dilutions of test compounds in BHIS and incubated for 20-24 h at 37 °C in an anaerobic chamber. Viable counts were then enumerated on BHIS agar plates and the MBC was determined as the lowest concentration of the test compound that resulted in ≥ 3-log reduction of the initial cell inoculum, after incubation for 20-24 h at 37 °C. MBC measurements were performed in duplicate.
Time-kill kinetics assay. The procedure for the killing rate assays was the same as described for MBC determination except that treated cultures were enumerated on BHIS agar plates at specific time points (t = 0, 3, 6 and 24 h). Briefly, stationary-phase cultures of C. difficile strain 4118 (prepared as described above) were treated with closantel, rafoxanide, compound 8 at 1× , 4× , 16× MIC or with metronidazole and vancomycin at 16× MIC. At t = 0, 3, 6 and 24 h, sample aliquots were withdrawn from the treated cultures and bacterial viability was scored by plating serial dilutions on BHIS agar plates. The agar plates were incubated in the anaerobic chamber for 20-24 h at 37 °C. Kinetic experiments were performed at least twice.
HEPG2 or HEK 293T/17 cells were cultured in 96-well plates, and incubated at 37 °C in a 5% CO 2 humidifying chamber for 24 h. Cells were then treated with test compounds at concentrations ranging from 0.125-32 μ g/mL, and an MTS assay was performed at 16-h post-incubation at 37 °C in a 5% CO 2 humidifying chamber, using the CellTiter 96 aqueous non-radioactive cell proliferation assay kit (Promega, Madison, WI, USA) per manufacturer's instructions. MTS assays were performed in duplicate.
Hemolysis assay. Sheep red blood cells (Innovative Research, Novi, MI, USA) were washed three times with PBS pH 7.4. A 3% cell suspension in PBS (100 μ L) was added to test compounds in PBS (100 μ L, to a final concentration of 32 μ g/mL), and then incubated at 37 °C for 1 h. The plate was centrifuged at 500 × g for 10 min, and supernatants (100 μ L) were transferred to a clean 96-well plate. Hemolysis was determined by measuring absorbance at 540 nm, with 1% Triton X-100 as the positive control and 0.5% DMSO in PBS as the negative control. Hemolysis assays were performed in triplicate.