The small molecule nitazoxanide selectively disrupts BAM-mediated folding of the outer membrane usher protein

Bacterial pathogens assemble adhesive surface structures termed pili or fimbriae to initiate and sustain infection of host tissues. Uropathogenic Escherichia coli, the primary causative agent of urinary tract infections, expresses type 1 and P pili required for colonization of the bladder and kidney, respectively. These pili are assembled by the conserved chaperone–usher (CU) pathway, in which a periplasmic chaperone works together with an outer membrane (OM) usher protein to build and secrete the pilus fiber. Previously, we found that the small molecule and antiparasitic drug nitazoxanide (NTZ) inhibits CU pathway–mediated pilus biogenesis in E. coli by specifically interfering with proper maturation of the usher protein in the OM. The usher is folded and inserted into the OM by the β-barrel assembly machine (BAM) complex, which in E. coli comprises five proteins, BamA–E. Here, we show that sensitivity of the usher to NTZ is modulated by BAM expression levels and requires the BamB and BamE lipoproteins. Furthermore, a genetic screen for NTZ-resistant bacterial mutants isolated a mutation in the essential BamD lipoprotein. These findings suggest that NTZ selectively interferes with an usher-specific arm of the BAM complex, revealing new details of the usher folding pathway and BAM complex function. Evaluation of a set of NTZ derivatives identified compounds with increased potency and disclosed that NTZ's nitrothiazole ring is critical for usher inhibition. In summary, our findings indicate highly specific effects of NTZ on the usher folding pathway and have uncovered NTZ analogs that specifically decrease usher levels in the OM.

Pili are critical for initiating and sustaining infection and thus represent attractive targets for the development of anti-virulence therapeutics. Anti-virulence therapeutics offer an alternative to traditional antibiotics and a promising approach to combat the ever-increasing rate of antibiotic resistance among pathogenic bacteria (16,17). By specifically targeting systems required to cause disease within the host, such therapeutics should limit detrimental side effects on the beneficial commensal bacteria and decrease the selective pressures that lead to antibiotic resistance (18,19). A number of approaches are under investigation for developing anti-virulence therapeutics that target pili, including vaccines against pilus subunits, competitive inhibitors of pilus-mediated adhesion, and small molecules that disrupt pilus biogenesis, termed pilicides (20,21).
In previous studies, the small molecule nitazoxanide (NTZ) was found to inhibit assembly of aggregative adherence fimbriae (AAF) by the CU pathway on the surface of enteroaggregative E. coli (22). NTZ is a synthetic nitrothiazolyl-salicylamide compound (see Fig. 7) used clinically to treat parasitic diseases such as giardiasis and cryptosporidiosis (23,24). We subsequently showed that, in addition to AAF, NTZ also inhibits biogenesis of the E. coli type 1 and P pili (25). This suggests that NTZ has broad inhibitory activity against CU pili. Treatment of E. coli with NTZ was found to interfere with proper folding of the usher protein in the bacterial OM, specifically its transmembrane ␤-barrel domain, leading to degradation of the usher by the periplasmic protease DegP (25). Given that the usher is essential for assembly and secretion of the pilus fiber, loss of functional usher proteins explains the loss of pili upon NTZ treatment. Notably, NTZ appears to selectively interfere with folding of the usher, because levels of other outer membrane proteins (OMPs) remain unaffected by drug treatment (25). This mechanism of action is unique compared with other existing pilicides, which target different aspects of the CU pathway, such as chaperone-subunit interactions or chaperonesubunit binding to the usher (21,(26)(27)(28).
Folding of bacterial OMPs occurs via the conserved ␤-barrel assembly machine (BAM) complex (29 -31). In E. coli, the BAM complex comprises the integral OM ␤-barrel protein BamA and four accessory lipoproteins, BamB-E (Fig. S1). Following translocation into the periplasm by the Sec translocon, nascent unfolded ␤-barrel OMPs bind chaperones such as SurA and Skp, which transport these substrates to the BAM machinery (29). The BAM complex recruits the nascent OMP-chaperone complexes and catalyzes efficient OMP insertion and assembly in the OM (32). Should a ␤-barrel substrate fail to be efficiently inserted into the OM, it can be targeted for degradation by DegP (33,34). Many questions remain regarding the mechanistic contributions of the individual BAM components (BamA-E) to OMP folding. BamA and BamD constitute the core components of the complex and are both essential for bacterial viability (35,36). BamD is implicated in recruitment of OMP substrates, whereas BamA plays a central role in OMP folding and insertion into the OM. The nonessential components BamB, BamC, and BamE perform accessory roles in enhancing the kinetics of OMP biogenesis (37)(38)(39)(40). OMPs vary in size and complexity and are differentially affected by deletion of individual accessory BAM components (30,41,42). This sug-gests that distinct, specialized arms of the BAM complex exist for the folding and membrane insertion of different OMP substrates. Analysis of the type 1 pilus usher FimD showed a strong dependence on SurA and BamB for folding in the OM, whereas deletion of BamC or BamE only minimally affected usher folding (43).
We hypothesized that NTZ inhibits pilus biogenesis by interfering with a specific aspect of BAM complex function required for folding of the usher, but not other OMPs. Consistent with this hypothesis, we show here that sensitivity of the PapC usher to NTZ is modulated by expression levels of the BAM complex and requires the BamB and BamE lipoproteins. Furthermore, using an unbiased genetic screen, we identify a point mutation in the essential BamD lipoprotein that confers reduced sensitivity to NTZ. These results define new aspects of the usher folding pathway and new functionality for the BamD and BamE lipoproteins, supporting a model in which NTZ targets a specialized arm of the BAM complex required for usher folding. We also report the identification of NTZ analogs with increased potency in specifically inhibiting the usher and determine that the nitrothiazole ring of NTZ is critical for the drug's inhibitory activity against the usher.

Sensitivity to NTZ correlates with BAM expression levels
We hypothesized that NTZ interferes with folding of the usher in the OM by altering activity of the BAM complex. If the BAM machinery is the direct target of NTZ, we should observe altered NTZ activity in bacterial strains with increased or decreased levels of the BAM complex; increased BAM expression should confer resistance to NTZ (more folded usher), whereas decreased BAM expression should potentiate NTZ activity (less folded usher). To test this, we quantitated levels of the P pilus usher PapC in OM fractions isolated from E. coli strains that over-or underexpressed the BAM complex, compared with the WT parental strain. For overexpression of the BAM complex, E. coli strain BW25113 expressing His-tagged PapC (pMJ3) was compared with the same strain transformed with a plasmid containing each of the bam genes under control of an inducible promoter (pBamABCDE). OM fractions were isolated from the strains grown in the presence of 0, 10, or 20 g/ml NTZ; the fractions were blotted with anti-His tag antibody; and PapC levels were quantitated by densitometry. In the WT strain background, treatment with 10 g/ml NTZ caused a 50% decrease in usher levels, and treatment with 20 g/ml NTZ caused a 70% decrease in usher levels, compared with vehicletreated control (0 g/ml NTZ) (Fig. 1, A and C). In contrast, overexpression of the BAM complex from pBamABCDE resulted in reduced sensitivity to NTZ, with usher levels decreasing by only 27% at 10 g/ml NTZ and 57% at 20 g/ml NTZ (Fig. 1, B and C). Examination of the OM fractions with an anti-BamA antibody confirmed overexpression of the BAM complex in the strain containing pBamABCDE, with BamA levels four times higher compared with the parental WT strain (Fig. 1B). The changes in BamA, as well as PapC, expression levels were also apparent in Coomassie Blue-stained gels of the OM fractions (Fig. 1, A and B). Moreover, inspection of the

Nitazoxanide selectively disrupts BAM-mediated usher folding
Coomassie-stained gels confirmed our previous findings (25) that treatment of bacteria with NTZ selectively affects levels of the usher, with other OMPs (such as the major porins OmpA, OmpF, and OmpC) remaining unchanged. We also did not observe discernable changes in the overall OMP profile of bacteria overexpressing the BAM complex compared with the parental WT strain.
We next examined the effect of BAM complex underexpression on sensitivity to NTZ. Because BAM complex activity is essential for bacterial viability, we used the MC4100bamA101 mutant strain (44) for this purpose. This strain contains a transposon insertion in the upstream promoter region of bamA, resulting in an ϳ10-fold decrease in BamA expression (45). BamA is the central component of the BAM complex (36); therefore, reducing BamA levels results in a corresponding decrease in overall levels of functional BAM complex. Parental strain MC4100 and the MC4100bamA101 mutant strain, both transformed with pMJ3 (expressing His-tagged PapC), were grown in the presence of 0, 10, or 20 g/ml NTZ, and OM fractions were isolated. Quantitation of anti-His tag immunoblots showed that strain MC4100/pMJ3 exhibited a similar dose-dependent decrease in OM usher levels in response to NTZ as observed for BW25113/pMJ3 (Fig. 2, A and D). In comparison, the bamA101 strain exhibited increased sensitivity to NTZ, with usher levels decreasing by 68% at 10 g/ml NTZ and 96% at 20 g/ml NTZ (Fig. 2, B and D). Complementation of the bamA101 mutant with a BamA expression plasmid (pBamA) restored WT sensitivity of the user to NTZ (Fig. 2, C and D), confirming the specificity of the mutant phenotype. Examination of the OM fractions with an anti-BamA antibody confirmed decreased expression of BamA in the bamA101 mutant compared with the parental WT strain, as well as restoration of BamA expression in the complemented strain (Fig. 2, B and C). The changes in PapC and BamA levels in the different strain backgrounds and NTZ treatment conditions were also apparent in Coomassie Blue-stained gels of the OM fractions. Taken together, these experiments show an inverse correlation between BAM expression levels and the activity of NTZ; BAM overexpression resulted in increased resistance of the usher to NTZ, and BAM underexpression resulted in increased sensitivity of the usher to NTZ. These findings are consistent with the BAM complex as a relevant target for the mechanism of action of NTZ.

The effect of NTZ on usher levels depends on BamB and BamE
If the BAM complex is the target of NTZ, NTZ binding must alter assembly or function of the complex in a manner that specifically affects folding of the usher protein, but not other OMPs. The BAM proteins form a stable complex in the OM that can be isolated from bacteria by pulldown assays using any individual component (46,47). To determine whether NTZ affects assembly of the Bam complex, E. coli strain BW25113 expressing the BAM complex with a His-tagged BamE (pBam-ABCDE-His 6 ) was cultured in the presence of 0 or 20 g/ml NTZ, OM fractions were isolated, and the detergent-extracted complex was captured by nickel affinity chromatography. Proteins eluted from the column were then analyzed by staining with Coomassie Blue or immunoblotting with anti-His tag and anti-BamA antibodies. We did not observe any changes in BAM protein stoichiometry or loss of co-purifying proteins in response to NTZ treatment (Fig. S2), indicating that NTZ does not interfere with the formation or stability of the BAM complex.
The folding of OMPs is differentially impacted by the loss of individual nonessential BAM lipoproteins (BamB, BamC, and BamE) (30,41,42), suggesting that different OMPs require distinct subcomplexes of the BAM machinery. To determine whether NTZ alters the function of the BAM complex in a manner specific to the usher, we first investigated contributions of the nonessential BAM components to folding of PapC and sensitivity to NTZ. The E. coli BW25113 WT strain and ⌬bamB, ⌬bamC, and ⌬bamE mutant derivatives, each express-

Nitazoxanide selectively disrupts BAM-mediated usher folding
ing His-tagged PapC (plasmid pMJ3), were grown in the presence of 0, 10, or 20 g/ml NTZ; OM fractions were isolated; and usher levels were analyzed by Coomassie staining and immunoblotting with anti-His tag antibody. For the ⌬bamC strain, growth in the presence of NTZ resulted in a similar dose-dependent decrease in PapC levels as observed for the WT strain ( Fig. 3, A and D), suggesting that BamC does not play a role in the sensitivity of the usher to NTZ. In addition, in the absence of NTZ, the ⌬bamC strain expressed similar levels of PapC compared with the WT strain, and the overall OMP profile was also unchanged (Fig. 3A). This is consistent with previous studies in which deletion of BamC caused only minor defects in general OMP folding (41,48). In contrast to the ⌬bamC mutant, the ⌬bamB and ⌬bamE strains appeared resistant to NTZ, with growth in the presence of the drug no longer causing a decrease in PapC levels (Fig. 3). This suggests that BamB and BamE are required for NTZ activity against the usher. Interestingly, PapC levels increased in the ⌬bamB strain compared with the WT strain, whereas deletion of bamE did not alter PapC levels (Fig. 3, B and C). The behavior of the usher was in contrast to the overall OMP levels, which were reduced in the ⌬bamB and ⌬bamE strains (Fig. 3, B and C).
BamB was previously shown to be required for folding of the FimD usher in the OM (43), whereas a role for BamE in the usher folding pathway has not been described. To probe the basis for the NTZ phenotypes of the ⌬bamB and ⌬bamE mutants, we investigated effects of the mutations on folding of the usher in the OM and function of the usher in pilus assembly. To discern effects on usher folding, we exploited the intrinsic resistance of ␤-barrel proteins such as the usher to denaturation by SDS, which results in heat-modifiable mobility on SDS-PAGE, where a fraction of the protein remains folded and migrates more rapidly than its predicted molecular mass in the absence of heat (49,50). To determine the function of the usher in pilus biogenesis, we measured assembly of adhesive P pili (from plasmid pFJ29, encoding an inducible pap operon) on the bacterial surface using a hemagglutination (HA) assay (51). The HA titer was measured as the maximum -fold dilution of bacteria able to agglutinate human red blood cells.
For the ⌬bamB mutant, analysis of OM samples incubated in SDS sample buffer at 25°C revealed that mutant was unable to produce any folded PapC usher (Fig. 4A). Complementation of the ⌬bamB mutant with a BamB expression plasmid (pBamB) restored usher folding back to WT levels (Fig. 4A). Consistent with a lack of folded usher, the ⌬bamB strain was unable to assemble adhesive P pili, with an HA titer of 0 (Fig. 4C). The lack of folded usher in the ⌬bamB mutant provides an explanation for its apparent resistance to NTZ, as NTZ specifically interferes with usher folding. Thus, the absence of an NTZ-dependent decrease in usher levels in the ⌬bamB mutant represents an accumulation of unfolded usher that is not degraded, rather than a decreased sensitivity to NTZ. In contrast to the ⌬bamB mutant, we found that the ⌬bamE mutant was still able to produce folded usher, although the amount of SDS-stable, faster-

Nitazoxanide selectively disrupts BAM-mediated usher folding
migrating species was lower compared with the WT strain (Fig.  4B). Complementation of the ⌬bamE mutant with a BamE expression plasmid (pBamE) restored usher folding back to WT levels (Fig. 4B). Intriguingly, the ⌬bamE mutant was fully functional for assembly of adhesive pili, with an HA titer 4 times greater than for the WT strain (Fig. 4C). The increased HA titer is not explained by a change in usher levels, which were comparable between the ⌬bamE and WT strains (Fig. 3C). Together, these results implicate BamE in the mechanism of action of NTZ and reveal a previously unknown role for BamE in the usher folding pathway.

A BamD P100S mutation reduces sensitivity to NTZ
As an unbiased approach to identify the target and mechanism of action of NTZ, we performed a genetic screen to isolate mutants resistant to NTZ (i.e. strains that retain high OM usher levels when grown in the presence of drug). We used nitrosoguanidine mutagenesis (52) to generate a library in E. coli BW25113 expressing a PapC construct (plasmid pJP1) with a 3X-FLAG tag insertion in a surface-exposed loop (Fig.  S3A). The FLAG tag allowed detection of the usher in intact bacteria using a phycoerythrin (PE)-conjugated anti-FLAG antibody, as confirmed by fluorescence microscopy (Fig. S3B). Using flow cytometry of intact bacteria, we confirmed that growth in the presence of 20 g/ml NTZ caused a large decrease in fluorescence of the bacteria, indicating loss of Pap-C FLAG in the OM (Fig. 5). We then set up a gate that contained almost no highly fluorescent bacteria in the WT population treated with 20 g/ml NTZ. Analysis of the nitrosoguanidine mutagenized bacteria showed a 7-fold increase in the percentage of bacteria within this gate after NTZ treatment, indicating the emergence of resistant clones (Fig. 5). Using this gate for fluorescence-activated cell sorting (FACS), we isolated ϳ200 mutagenized E. coli cells that remained highly fluorescent following NTZ treatment. Viable clones (ϳ50) were then individually assessed for their sensitivity to NTZ by growth in the presence or absence of 20 g/ml NTZ. OM fractions were isolated and analyzed by Coomassie Blue staining and immunoblotting using an anti-FLAG antibody to detect PapC levels. Through this analysis, five mutant strains were verified to maintain higher usher levels in the presence of NTZ compared with the WT strain. These five mutant strains, as well as our WT strain, were then subjected to whole-genome sequencing to identify the mutation(s) that conferred resistance to NTZ. Each of the resistant strains had 100 -200 nonsynonymous mutations (Data File S1).
A mutation of particular interest isolated from our screen was a proline-to-serine mutation in residue 100 of the BamD lipoprotein. To confirm that this mutation confers resistance to NTZ, we constructed a bamD P100S mutation in the chromosome of WT E. coli BW25113. The WT and bamD P100S strains, each expressing His-tagged PapC (pMJ3), were grown in the presence of 0 or 20 g/ml NTZ, OM fractions were isolated,

Nitazoxanide selectively disrupts BAM-mediated usher folding
and relative usher levels were analyzed by immunoblotting with anti-His tag antibody and quantitation by densitometry. Growth of the bamD P100S mutant in the presence of 20 g/ml NTZ caused a 54% decrease in OM usher levels, compared with a 79% decrease for WT BW25113 (Fig. 6, A and B). Thus, the P100S mutation in BamD, by itself, reduces sensitivity to NTZ. In the absence of NTZ, expression levels of the PapC usher in the OM were comparable between WT BW25113 and the bamD P100S mutant (Fig. 6A). Furthermore, inspection of Coomassie Blue-stained gels showed that the general OMP profile was not altered in the bamD P100S mutant compared with the WT strain in the presence or absence of NTZ (Fig. 6A). This indicates that the BamD P100S mutation specifically affects sensitivity of the usher to NTZ. Analysis of heat-modifiable mobility showed that, in the absence of NTZ, the bamD P100S mutant produced folded PapC at similar levels as the WT strain (Fig.   6A). The usher also remained fully functional in the bamD P100S strain, with an HA titer similar to WT in the absence of NTZ (Fig. 6C). However, when grown in the presence of increasing concentrations of NTZ, the HA titer decreased by only 4-fold for the bamD P100S mutant compared with 16-fold for the WT strain (Fig. 6C). Together, these results demonstrate that the BamD P100S mutation enables bacteria to assemble more functional usher in the presence of NTZ.

The inhibitory effect of NTZ on usher folding depends on its nitrothiazolyl ring
NTZ is a broadly active compound with a wide range of reported activities (53)(54)(55). At the concentrations used for our experiments, NTZ does not exhibit toxicity toward E. coli. Instead, NTZ specifically affects pilus biogenesis by the CU pathway by interfering with folding of the usher protein. To generate information on the structure-activity relationship of NTZ and its specificity in disrupting usher folding, we screened analogs derived from the NTZ scaffold (Fig. 7A) for effects on levels of the PapC usher in the OM. E. coli strain BW25113 expressing His-tagged PapC (pMJ3) was grown in the presence of vehicle only (DMSO), 32 M (10 g/ml) NTZ, or 32 M NTZ derivative. OM fractions isolated from bacteria were immunoblotted with anti-His tag antibody, and PapC levels were quantified by densitometry (Fig. 7, B and C). Treatment of bacteria with 32 M NTZ caused a 47% decrease in OM usher levels compared with control. Nitrothiazole analogs of NTZ (161183, 16a1011, 161282, and 16a1039) exhibited greater potency compared with NTZ, whereas dinitrothiophene analogs (16b2025 and 16b2031) had no appreciable effect on OM usher levels (Fig. 7, B and C). Substitution of a chloro group at position 4 on the nitrothiazole ring (162088) abolished the drug's effect as well. This identifies the nitrothiazole ring of NTZ as critical for the inhibitory effect of the drug on the usher and, by extension, CU pilus assembly. The most potent NTZ derivative was 161282, a thiophene analog of 2-amino-5-nitrothiazole (Fig. 7). Treatment of bacteria with 32 M compound 161282 decreased OM usher levels by 91% compared with vehicle-treated control (Fig. 7B). Analysis of the OM fractions by Coomassie staining revealed no effect of NTZ or the NTZ analogs on levels of other OMPs. This indicates that the increased potency of the nitrothiazole analogs is due to increased efficacy against the usher, rather than a general targeting of OMP biogenesis or other nonspecific effect, consistent with the mechanism of action of NTZ.

Discussion
Pili are critical virulence factors for a wide range of pathogenic bacteria. Their roles in initiating infection, promoting colonization, and dictating tropism for different sites within the host make pili attractive targets for therapeutic intervention (2)(3)(4). We have determined that the small molecule NTZ acts as a pilicide to disrupt pilus assembly by the conserved CU pathway (25). In previous work using the type 1 and P pilus systems expressed by UPEC, we found that treatment of bacteria with NTZ causes misfolding and subsequent degradation of the usher protein (25). The usher is an essential and rate-limiting factor for pilus biogenesis, explaining the loss of pili upon NTZ  A and B). The OM fractions were incubated in SDS sample buffer at 25°C, subjected to SDS-PAGE, and probed with anti-His tag antibody to visualize PapC. Positions of the folded (F) and unfolded (U) PapC species are indicated on the right of each gel image. WT BW25113 and the ⌬bamB, ⌬bamC, and ⌬bamE deletion mutants, each expressing P pili (pJF29), were tested for HA activity (C). Serial dilutions of normalized whole bacteria were incubated with human red blood cells, and the HA titer was recorded as the maximum -fold dilution of bacteria able to agglutinate the blood cells. Bars, the log 2 transformed titers, which were calculated from three independent experiments of three replicates each; all values for each of the experiments and replicates were identical.

Nitazoxanide selectively disrupts BAM-mediated usher folding
treatment. Here, we investigated the mechanism by which NTZ affects usher folding. We present evidence that NTZ acts on an usher-specific aspect of the BAM complex and assign new roles for BamD and BamE in the usher folding pathway.
Prior analysis of the FimD usher found that proper folding and insertion into the OM is mediated by the BAM complex and requires the SurA periplasmic chaperone and BamB lipoprotein (34, 43, 56). The translocation and assembly module   , middle and bottom panels). Positions of the folded (F) and unfolded (U) PapC species are indicated on the right for samples incubated at 25°C. PapC levels were measured by densitometry of the anti-His tag blot (95°C) and percentage PapC levels were calculated relative to 0 g/ml NTZ (B). The graph shows individual data points and means Ϯ S.D. (error bars) from four independent experiments. *, p Ͻ 0.05 for comparison of OM PapC levels at 20 g/ml NTZ. WT and bamD P100S BW25113, each expressing P pili (pJF29), were tested for HA activity (C). Serial dilutions of normalized whole bacteria were incubated with human red blood cells, and the HA titer was recorded as the maximum -fold dilution of bacteria able to agglutinate the blood cells. Bars, log 2 transformed titers, which were calculated from three independent experiments of three replicates each; all values for each of the experiments and replicates were identical.

Nitazoxanide selectively disrupts BAM-mediated usher folding
(TAM) proteins TamA and TamB have also been shown to work together with the BAM complex to increase the efficiency of usher folding in the OM (57). The BAM complex mediates the folding of a broad range of OMPs in addition to the usher; however, NTZ selectively disrupts usher folding, leaving other OMPs unaffected (25). A goal of the current study was to define the usher-specific folding pathway that is targeted by NTZ. We found that the sensitivity of ⌬tamA or ⌬tamB deletion mutant strains to NTZ was similar to the WT strain, suggesting that NTZ is not targeting the function of the TAM complex. 6 SurA assists in the delivery of many OMPs to the OM, including the major porins (58). Treatment of bacteria with NTZ does not alter porin levels or the general OMP profile, and therefore NTZ is unlikely to target SurA activity. In contrast, evidence has accumulated for the existence of distinct arms of the BAM complex for folding particular substrates (30,41,42), leading us to hypothesize the BAM system as an NTZ target.
Consistent with the BAM system as a target of NTZ, we show here that overexpression of the BAM complex leads to increased resistance of the PapC usher to NTZ. BAM overexpression by itself did not alter PapC levels in the OM; nor did it alter the overall OMP profile. Importantly, this argues that the observed resistance of PapC to NTZ upon Bam overexpression was not simply due to increased production of the usher overcoming the effects of the drug. Instead, this result supports a model in which NTZ disrupts an usher-specific aspect of BAM function, with BAM overexpression providing a larger pool of functional complexes to compensate for complexes rendered incompetent for usher assembly by NTZ. Also consistent with the BAM system as a target of NTZ, we observed increased sensitivity of the usher to NTZ in the bamA101 mutant strain, which has reduced BamA expression and therefore reduced levels of functional BAM complexes (45). Thus, in this case, fewer BAM complexes competent for usher folding should remain following NTZ treatment. The bamA101 mutation did not by itself alter steady-state PapC levels in the OM, although the levels of other OMPs, such as the porins, decreased compared with the WT strain. Therefore, the increased sensitivity of PapC to NTZ in the bamA101 strain cannot be explained by a lower starting level of usher molecules. The different response of PapC compared with other OMPs such as the porins to the bamA101 mutation also supports the existence of a distinct folding pathway on the BAM complex for the usher.
We took advantage of the sensitivity of the usher to NTZ to use the drug as a tool to probe the usher-specific folding pathway on the BAM complex. In agreement with prior analysis of the FimD usher (43), we found that production of properly folded and functional PapC in the OM required BamB but not

Nitazoxanide selectively disrupts BAM-mediated usher folding
BamC. An interesting difference between our results and the prior FimD study is that we observed an accumulation of unfolded PapC in the OM in the ⌬bamB mutant, whereas FimD levels decreased (43). Our results suggest that, in the absence of BamB, the unfolded PapC is trapped at an intermediate step along its folding pathway, presumably on the BAM complex, where it is inaccessible to degradation by the DegP protease. Alternatively, the absence of BamB prevents handoff of PapC to the BAM complex, and PapC is protected by the chaperone SurA. PapC levels no longer decreased in response to NTZ in the ⌬bamB strain. This apparent resistance to NTZ is attributable to the accumulation of unfolded usher in the ⌬bamB mutant. Thus, NTZ would have no apparent effect in the ⌬bamB strain, as the mutant BAM complex is incompetent for usher folding, and the unfolded usher is trapped at a stage where it is not subject to degradation. Further, this suggests that NTZ targets a step in the usher folding pathway subsequent to BamB function. In contrast to the ⌬bamB mutant, PapC remained sensitive to NTZ in the ⌬bamC strain, consistent with the lack of a role for BamC in the usher folding pathway.
In addition to BamB, our analysis revealed a previously unrecognized role for BamE in the usher folding pathway and a requirement for BamE in the mechanism of action of NTZ. The ⌬bamE mutant was resistant to NTZ, with PapC levels no longer decreasing in response to the drug. Total levels of PapC in the OM were similar between the ⌬bamE mutant and WT strain, and thus this resistance was not due to increased usher expression. In contrast to the ⌬bamB mutant, PapC remained fully functional for pilus assembly in the ⌬bamE strain, with an HA titer even greater than for the WT strain. Given that the usher must be folded to assemble pili, this shows that the NTZ resistance of the ⌬bamE strain was not due to an accumulation of unfolded usher. However, loss of BamE did alter usher stability, with less folded species appearing in the heat-modifiable mobility assay. The basis for the NTZ resistance and increased HA titer of the ⌬bamE mutant remains to be determined, but our results suggest a role for BamE, either directly or indirectly, in achieving optimal final stability of the usher. To our knowledge, such a phenotype has not been described previously. The function of BamE is poorly understood, and loss of BamE causes only minor defects in general OMP assembly (30,47). BamE has been shown to modulate the conformation of BamA through its interactions with BamD (46,(59)(60)(61). If NTZ targets the BAM complex directly, an altered conformation of the BAM complex in the ⌬bamE mutant might affect the drug-binding site. It was recently shown that BamB and BamE have specialized, nonoverlapping roles in coordinating BamA-BamD interactions during the assembly of OMPs in complex with the stress-sensing lipoprotein RcsF (39,40,62). Our results suggest that BamB and BamE also play distinct roles in the usher folding pathway. Whereas deletion of BamB completely abolished usher folding, deletion of BamE resulted in a less stable yet functional state of the usher. Therefore, the folding pathway of the usher in the ⌬bamE mutant might be changed in such a way that it is no longer susceptible to interference by NTZ.
A further piece of evidence for the BAM complex as a target of NTZ came from our isolation of a BamD P100S mutation in a screen for mutations conferring decreased sensitivity of the usher to the drug. The nitrosoguanidine mutagenesis protocol used for our screen generated multiple nonsynonymous mutations in each of the resistant strains (Data File S1). We focused here on the role of the BAM complex in the mechanism of NTZ and constructed the bamD P100S mutation in a clean genetic background. Other mutations identified in our screen will be the subject of a separate study. In the absence of NTZ, the BamD P100S mutation did not cause changes to usher expression levels, usher folding, or usher function in pilus assembly. The mutation also did not alter the general OMP profile in the presence or absence of drug. Thus, the BamD P100S mutation presumably affects an usher-specific aspect of the BAM complex that is targeted by NTZ, possibly decreasing affinity of NTZ for its binding site or altering the usher folding pathway in a way that bypasses the drug's effect. Studies are under way to determine whether NTZ binds directly to the BAM complex and the impact of the BamD P100S mutation on this binding. An alternate possibility is that NTZ binds to an interface between the unfolded usher and the BAM complex in a way that affects productive folding by the BAM complex. Proline 100 of BamD is located on a flexible loop that is part of an inner periplasmic ring formed by BamD and POTRA domains 1, 2, and 5 of BamA (Fig. S4) (31,63,64). This region is located under the BamA ␤-barrel pore and is thought to serve as the first contact point for interaction of substrates with the BAM complex (31,63,65,66). The resistant phenotypes of the ⌬bamE and bamD P100S mutants suggest that BamE and BamD cooperate in the usher folding pathway and that NTZ disrupts a specialized BamE-BamD-BamA arm of the BAM complex.
This study analyzed steady-state protein levels, and further work is needed to understand the contributions of BamE and BamD to the kinetics of usher folding in the OM and whether the specificity of NTZ for the usher might relate to differences in folding kinetics or turnover rates for the usher compared with other OMPs. Whereas the BAM complex is generally conserved in Gram-negative bacteria, there are differences in BAM complex structure and domain composition among species (67)(68)(69). The individual contributions of BAM complex components to OMP assembly may thus vary in distantly related organisms. However, the BamE and BamD lipoproteins and the proline 100 residue of BamD are conserved in many different organisms, including Klebsiella pneumoniae, Proteus mirabilis, Enterobacter cloacae, Yersinia pestis, Salmonella enterica and Pseudomonas aeruginosa. The mechanism of NTZ is thus likely to be conserved across multiple Gram-negative pathogens, and NTZ may affect a wide variety of CU pathways.
NTZ is used clinically as an anti-parasitic agent (23,24). NTZ also exhibits broad-spectrum activity against anaerobic bacteria and has a wide range of other reported activities (53)(54)(55). In contrast to this broad activity, we have identified a highly specific effect of NTZ on the usher folding pathway. With the goal of obtaining more potent and narrowly active NTZ analogs that could help define the target and mechanism of action of the drug against the usher, we screened a set of analogs derived from the NTZ scaffold. Our studies identified the nitrothiazole ring of NTZ as critical for inhibition of PapC levels in the OM. Among the most potent compounds were 16a1011 and 161282,

Nitazoxanide selectively disrupts BAM-mediated usher folding
which incorporate halide-substituted thiophenes coupled to 2-amino-5-nitrothiazole (Fig. 7A). Consistent with the proposed mechanism of action for NTZ, the NTZ analogs specifically decreased usher levels in the OM, with no effects on the general OMP profile. These results hold promise for the development of novel NTZ analogs that could be used as anti-virulence therapeutics to disrupt pilus biogenesis by the CU pathway in diverse bacterial pathogens.

Strains, plasmids, and growth conditions
The E. coli strains and plasmids used in this study are listed in Table S1. Unless otherwise stated, E. coli overnight cultures were diluted 1:100 into fresh LB medium containing appropriate antibiotics (100 g/ml ampicillin, 15 g/ml tetracycline) and 0 -20 g/ml NTZ as indicated. For 0 g/ml NTZ, a final concentration of 0.2% DMSO was added as a vehicle-only control, and 0.2% DMSO final concentration was maintained for all NTZ concentrations. Cultures were grown at 37°C with aeration. When the cultures reached OD 600 of 0.6, the expression of plasmid-encoded genes was induced for 1 h by the addition of the appropriate inducing agent: 50 M isopropyl-␤-D-thiogalactopyranoside (IPTG) or 0.1% arabinose.
Plasmid pBamB was generated by digesting pBAD18::bamB (36) with EcoRI and HindIII and ligating the bamB fragment into pTRYC (73). Plasmid pBamE was generated by amplifying the bamE-His8 fragment from pBamE-His (47) using a forward primer containing an EcoRI site and a reverse primer with a HindIII site. All primers used for cloning are listed in Table S2. The fragment was then ligated into pTRYC. Plasmids pBamAB-CDE and pBamA were generated by subcloning the desired bam genes from plasmid pJH114 (37) into pTRYC. The bamE gene from pJH114 also contains a C-terminal His tag. For cloning of these genes into pTRYC, a NdeI site was introduced into the multiple cloning site of pTRYC using the QuikChange XL mutagenesis protocol (Stratagene) to generate pTRYC-NdeI. The desired bam genes were then amplified from pJH114 using a forward primer containing a NdeI site and a reverse primer with an XmaI site and then ligated into pTRYC-NdeI. Plasmid pJP1 encoding the PapC usher with a 3X-FLAG tag insertion at residue Asn-197 was generated by site-directed, ligase-independent mutagenesis (SLIM) as described (74,75) from plasmid pFJ20 (76).

NTZ and NTZ analog preparation
NTZ (Cayman Chemical) and NTZ analogs were obtained in powder form and dissolved in DMSO to give a stock concentration of 10 mg/ml. Synthesis of NTZ analogs has been described previously (77). All structures were confirmed by NMR and were Ͼ95% pure by LC-MS.

HA assay
HA assays were conducted as described (51) by serial dilution in microtiter plates. E. coli strain BW25113/pFJ29 was used to test P pilus functionality. Plasmid pACYC184 served as the vector control for pFJ29. Following growth in 0 -20 g/ml NTZ and induction with 50 M IPTG, bacteria were harvested and washed with PBS before being resuspended and normalized to an OD 540 of 1.0. Following serial dilution of the bacteria in microtiter plates, human erythrocytes were added, and HA titers were determined visually as the highest dilution of bacteria able to maintain agglutination. For each assay, three independent experiments were performed, and each experiment contained three replicates.

Isolation and analysis of the OM
OM fractions were isolated from the indicated E. coli strains grown and induced as described. PapC usher was expressed from plasmid pMJ3 or pJP1, with plasmid pMON6235⌬cat serving as the vector control for the PapC plasmids. The entire BAM complex was expressed form plasmid pBamABCDE, with plasmid pTRYC serving as the vector control. BamA, BamB, and BamE were expressed from pBamA, pBamB, and pBamE, respectively. Bacteria were harvested, washed, resuspended and normalized to an OD 600 of 1.0 in 20 mM Tris-HCl (pH 8.0) plus 1ϫ complete protease inhibitor mixture. The OM fractions were isolated as described (49). The OM pellets were resuspended in 20 mM Tris-HCl and 0.3 M NaCl, followed by the addition of SDS sample buffer. The OM samples were incubated at 25°C or heated to 95°C for 10 min. Samples were analyzed by SDS-PAGE and immunoblotted with a mouse monoclonal anti-His 6 antibody (1:1,000; Covance) for strains expressing pMJ3 or with a mouse monoclonal anti-FLAG antibody (1:1,000; Sigma-Aldrich) for strains expressing pJP1, followed by a secondary IRDye 680 goat anti-mouse conjugated IR antibody (1:40,000; LI-COR Biosciences). Samples were immunoblotted with anti-BamA antibody (1:30,000) (47), followed by a secondary IRDye 800 goat anti-rabbit-conjugated IR antibody (1:15,000; LI-COR Biosciences). Blots were visualized, and band densitometry was performed using a LI-COR Odyssey CLx imager. The specificity of the anti-FLAG antibody is demonstrated by a lack of signal in bacteria expressing non-FLAGtagged PapC (Fig. 5). The specificity of the anti-His 6 antibody is demonstrated by a lack of signal in vector controls (Figs. 1 and  2). The specificity of the anti-BamA antibody is demonstrated by expected changes in signal strength for the BamA WT, mutant, and complemented strains (Figs. 1 and 2).

Purification of the BAM complex
The BAM complex was purified as described previously (37). E. coli strain BW25113/pBamABCDE-His 6 was cultured in the presence of either 0 or 20 g/ml NTZ and lysed by passage through a French press, and a detergent-solubilized clarified lysate was passed over a nickel affinity column to capture the His-tagged BamE, together with stably associated proteins. Protein fractions eluted from the column with imidazole were then concentrated, normalized, and analyzed by Coomassie staining and immunoblotting with the anti-His tag and anti-BamA antibodies.

Nitazoxanide selectively disrupts BAM-mediated usher folding Nitrosoguanidine mutagenesis
A mutant library of BW25113 expressing the FLAG-tagged PapC usher (pJP1) was generated using N-methyl-NЈ-nitro-Nnitrosoguanidine (MNNG) (Bio-Rad), as described (52). BW25113/pJP1 was grown to an OD 600 of 1.0. A 2-ml aliquot of cells was incubated with 50 l of MNNG (2.5 mg/ml in 95% ethanol) for 10 min at 37°C. Cells were harvested by centrifugation (10 min, 1,500 ϫ g) and washed twice with 2 ml of M63 medium. Cells were then resuspended in 5 ml of LB medium and grown overnight at 37°C with aeration. The mutagenized BW25113/pJP1 bacteria were then subcultured in the presence of 0 -20 g/ml NTZ and subjected to FACS.

FACS
WT and nitrosoguanidine-mutagenized BW25113/pJP1 were grown in the presence of 0 -20 g/ml NTZ and induced with 0.1% L-arabinose for expression of the FLAG-tagged PapC usher. Bacteria were harvested, washed, resuspended, and normalized to an OD 600 of 1.0 in PBS. A 500-l aliquot was resuspended in PBS containing 1% BSA and rocked at 25°C for 30 min. Bacteria were then resuspended in 500 l of 1:1000 PEconjugated anti-FLAG antibody (BioLegend) and rocked at 25°C for 1 h. Bacteria were washed three times with PBS and subjected to FACS using a BD FACS ARIA IIIU high-speed cell sorter. Gating on side and forward scatter was used to focus on bacterial cells, and doublets were excluded. Mutagenized bacteria that maintained high levels of fluorescence (high levels of PapC FLAG ) in the presence of NTZ were sorted as single cells into 96-well plates containing LB medium. Plates were then incubated at 37°C, 300 rpm for 24 h and stored at Ϫ80°C after the addition of 5% glycerol.

Whole-genome sequencing
Five mutant FLAGϩ strains as well as the parent strain BW25113 were submitted as frozen cell pellets to GENEWIZ for whole-genome sequencing. Genomic DNA was isolated as specified for Gram-negative bacteria and sequenced by MiSeq (Illumina) to generate 150-bp paired-end data. Reads from each isolate were mapped onto the reference genome (E. coli BW25113, accession number NZ_CP009273), and variants were identified.

Statistical analysis
Densitometry analysis of protein bands was performed using a LI-COR Odyssey CLx imager on samples from three independent experiments. Statistical significance was calculated by one-way analysis of variance and by Bonferroni's multiple-comparison post-test using Prism version 8 (GraphPad Software). p values of Ͻ0.05 were considered significant.