Campylobacter jejuni 81–176 was the primary model strain used for the selection of the potential Tat-dependent inhibitors by high-throughput screening. C. jejuni 81–176 is resistant to 0.5 and 1 mM CuSO₄ (Supplementary Figure S1). C. jejuni 81–176 tatC knockout mutant (C. jejuni ΔtatC) was used to confirm the Tat-dependent inhibitory effect of the selected SM. C. jejuni ΔtatC is susceptible to 0.5 mM and higher of CuSO₄ (Rajashekara et al., 2009). The specificity of the Tat-dependent inhibitors was also tested on 11 additional Campylobacter strains (C. coli ATCC33559 and 10 other C. jejuni strains isolated from chickens) and 7 commensal/beneficial bacteria (Supplementary Table S1). The 1 C. jejuni strains (Au-13, Au-20, Au-38, Au-39, Au-47, Au-50, Au-24, Au-32, Au-44, and Au-45) were selected based on their single nucleotide polymorphism (SNP)-type and percent of prevalence in chickens (Merchant-Patel et al., 2008; Kumar et al., 2016). Additional details about the bacterial strains are shown in Supplementary Table S1.

Colonic epithelial cells (Caco-2) were used to evaluate the cytotoxicity and the ability of the selected eight SMs to clear intracellular C. jejuni. The Caco-2 cells were obtained from the American Type Culture Collection (ATCC HTB-37; Rockville, MD, United States). For in-vivo testing, 3 and 5-week-old White Cronish broiler chickens, obtained from Meyer’s hatchery (Polk, OH, USA), were used to validate the anti-C. jejuni efficacy (log reduction in cecal content) of the selected four SMs (T1, T2, T7, and T8) in poultry and their impact on the cecal microbiota. Broiler chickens were grown in accordance with The Ohio State University Animal Care and Use Program (accredited by the Association for Assessment and Accreditation of Laboratory Animal Care International) and performed following the Institutional Animal Care and Use Committee. Chickens were fed a standard broiler diet (i.e., broiler starter phase from day 0 to 10; broiler growth phase from day 11 to 25; and broiler finisher phase from day 26 to 42; PNW extension #658) which was obtained from local feed mill at the Ohio Agricultural Research and Development Center (OARDC). Chickens and associated rooms were observed at least twice daily to assure that no shortage in feed or side effects occur due to the bacterial inoculation or treatments provided during the experiments.

A total of 50,917 SMs obtained from 11 libraries were screened in this study. SMs were suspended in 100% DMSO at concentration ranging from 2.5 μg/mL to 12.5 μg/mL between libraries and stored at −80°C (Supplementary Table S2). These libraries originated from The National Screening Laboratory for the Regional Center of Excellence in Biodefense and Emerging Infectious Disease (NSRB, United States)—The New England Regional Center of Excellence for Biodefense and Emerging Infectious Diseases (NERC, USA) collection. The NSRB and NERC libraries included FDA-approved bio-active SM, SM used by NIH in recent clinical trials, bio-active screens, and diversified SM synthesized for favorable physico-chemical properties (e.g., solubility, low/no toxicity and increased stability). Details about the origin and concentration of the libraries are shown in Supplementary Table S2.

Our in silico study highlighted that the unique assembly and disassembly feature of Tat system is required for the translocation of several relatively large folded proteins, such as formate dehydrogenase (Fdh) and multi-copper oxidase (CueO) (Rajashekara et al., 2009). Since CueO function requires the transport of the Tat system, increased susceptibility to copper was used as an indicator during our in vitro screen for Tat-specific SM inhibitors. The identification of SM increasing sensitivity of C. jejuni 81–176 to copper sulfate (1 mM CuSO₄, non-lethal dose) was performed using a high-throughput screening in 384 plate formats at NSRB facilities (Harvard Medical School, Cambridge, MA, USA). Columns 1 to 24 of assay plates were filled with 40 μL of MH broth +1 mM CuSO₄ using a Matrix WellMate (Thermo Fisher, United States) automatic plate filler (Drozd, 2012). In total, 100 nL of SM were transferred to each well using The Institute of Chemistry and Cell Biology (ICCB, USA) Longwood Screen Facility Seiko D-TRAN XM3106-31 PN 4-axis cartesian robot (V&P Scientific), which was controlled by SRC-310A Controller/SPEL. Columns 1–23 of the assay plates were inoculated with 40 μL of C. jejuni 81–176 normalized at optical density (OD₆₀₀) of 0.16 in fresh MH broth supplemented with 1 mM CuSO₄. Wells in column 24 were filled with 40 μL of C. jejuni ΔtatC (susceptible to 1 mM CuSO₄) normalized at 0.16 OD₆₀₀ in the same conditions. The ΔtatC mutant is susceptible to 0.5 mM CuSO₄; however, for screening purpose, we used the stringent 1 mM CuSO₄. Plates were incubated at 42°C for 36 h under microaerophilic conditions (85% N₂, 10% CO₂, and 5% O₂). Turbidimetric measurements (OD₆₀₀) were recorded before and after incubation using a Biotek Synergy HT spectrophotometer. The positive and negative controls on each assay plate were used to calculate a Z’ value for that plate. If the Z’ value was >0, the threshold for defining a compound as positive was set at three standard deviations above the average positive control value. Z’ is calculated as described in the study mentioned in the reference (Zhang et al., 1999): Z’ = 1- (3σneg + 3σpos)/(μneg-μpos), where μ is the mean and σ is the standard deviation. The ΔtatC mutant is the positive control mimic, and wild type without drug is the negative control.

Furthermore, validation of the primary assay was performed in a 96-well plate format for the selected 177 SMs with high drug-like properties that significantly increased susceptibility of C. jejuni to CuSO₄. In total, 100 μL of C. jejuni suspension normalized at 0.08 OD₆₀₀ in fresh MH broth supplemented with 1 mM CuSO₄ were transferred to each well of a 96-well plate and treated with 6.25 μg of SM. A parallel 96-well plate was prepared as described above but without CuSO₄, to identify SM that did not inhibit the growth of C. jejuni in a Tat-specific manner. Plates were incubated at 42°C under microaerophilic conditions for 24 h. Turbidimetric measurements were recorded before and after incubation, to remove a potential increase in OD₆₀₀ caused by the SM. Only SM inhibiting C. jejuni WT growth in the presence of CuSO₄ without inhibiting C. jejuni WT growth in the absence of CuSO₄ was selected for further analyses.

Based on previous publicly available bioassay data, the ICCB-longwood/NSRB Screensaver database version v2010.10.29 and v2012.01.26 was used to eliminate SMs that were less likely to target the Tat pathway (Tolopko et al., 2010). Commercial and pharmaceutical based libraries with known antibacterial applications were cross-referenced with the selected SM. In addition, SMs that had positive effects on eukaryote-based screens were deprioritized because of the absence of eukaryotic homolog Tat system.

A series of filters were established to select SM with the optimized drug-like properties. The criteria were based on physicochemical descriptors, potential liabilities (i.e., predicted toxicity), chemical structural diversity, and novelty. ChemDraw suite (CambridgeSoft, PerkinElmer, United States) was used to calculate molecular weight from the simplified molecular-input line-entry system (SMILES) notation. Molecular weights less than 200 Daltons (Da) and more than 550 Da were deprioritized due to the golden triangle measurements for drug-like characteristics (Johnson et al., 2009). ChemBioFinder (CambridgeSoft, PerkinElmer, United States) was used to identify structural redundancy between the SM and remove highly similar candidates. SciFinder® (Chemical Abstracts Service) was used to deprioritize SM, showing more than 90% similarity with previously investigated drugs.

Formate is an essential source of energy for Campylobacter and plays a role in optimizing the adaptation of C. jejuni to the oxygen-limited gastrointestinal tract of the host (Kassem et al., 2017). Fdh is translocated by the Tat system in the periplasm, and thus, the inactivation of Tat system results in a reduction of the Fdh activity. Hence, this activity was used as an indicator during our in vitro screening procedures, to validate Tat-dependent SM inhibitors of C. jejuni. The Fdh inhibition activity assay was performed with the 177 SM with drug-like properties that increased the susceptibility of C. jejuni to CuSO₄ (1 mM). C. jejuni 81–176 suspension normalized to 0.08 OD₆₀₀ in fresh MH broth was incubated for 24 h in microaerophilic condition with 6.25 μg of SM. Subsequently, the treated cultures were suspended in an oxygen-restricted solution containing 25 mM sodium phosphate buffer (pH 7) with 1 mM benzyl-viologen and 10 mM sodium formate. The increase of OD₅₇₈ was measured using a SpectraMax Plus 384 absorbance plate reader (Molecular Devices, USA), to monitor the reduction in benzyl viologen, an indicator of Fdh activity. N = three replicates per SM. C. jejuni 81–176 treated with 1% DMSO and C. jejuni ∆tatCmutant were used as controls. SMs that inhibited at least 30% Fdh activity (which is equivalent to the inhibitory effect of 1 mM of azide on FDH activity (Davies, 2017)) compared with the DMSO control were further down selected for analysis.

In total, 19 SMs were selected using CuSO₄ sensitivity and Fdh inhibition assay and further tested on 11 Campylobacter strains grown in MH broth supplemented with 0.5 mM CuSO4 and treated with 6.25 μg/mL of SM, as described above. A similar experiment was performed with seven commensal/beneficial gut bacteria to determine species specificity of these SMs. In brief, an overnight suspension was normalized at 0.05 OD₆₀₀ using the appropriate medium and challenged with 6.25 μg/mL of SM. Details about the strains and their growing conditions are shown in Supplementary Table S1. Medium alone and 1% DMSO were used as controls (N = 3 replicates per SM).

A CuSO₄ sensitivity dose–response assay was performed to determine the minimal concentration of SM that completely inhibited (MIC) or killed (MBC) C. jejuni 81–176 in the presence of CuSO₄ (0.5 mM). In total, 19 SMs were two-fold serially diluted to obtain a final SM concentration ranging from 6.25 to 0.012 μg/mL. C. jejuni 81–176 was then treated with a determined concentration of SM, as described in the copper sensitivity assay. The lowest SM concentration that completely inhibited the growth without killing C. jejuni in the presence of 0.5 mM CuSO₄ was considered the MIC (no increase in OD₆₀₀ over time, but viable cells were recovered on agar plate after challenge). The lowest bactericidal SM concentration was considered the MBC (no increase in OD₆₀₀ over time and no viable cells were recovered on agar plate after challenge) (N = 3 replicates per SM). A similar copper sulfate sensitivity dose–response assay was performed with the eight most potent SMs (T1-T8) against other Campylobacter strains (n = 11; Supplementary Table S1) using methodology described above. However, these SMs were two-fold serially diluted to obtain a final SM concentration ranging from 5 μg to 0.0625 μg for testing (N = two independent experiments with four technical replicate for each SM).

Cytotoxicity of the eight SMs (T1-T8) was tested on Caco-2 cells at 5, 25, and 50 μg/mL, as previously described (Kumar et al., 2016; Deblais et al., 2019). In brief, a 96-well plate seeded with approximately 1.4 × 10⁵ Caco-2 cells/well in MEM medium was challenged with a final SM concentration of 5, 25, and 50 μg/mL. After 24 h of incubation at 37°C in a humidified 5% CO₂ incubator, cytotoxic effects were determined using the PierceTM Lactate Dehydrogenase Cytotoxicity Assay Kit (Thermo Fisher Scientific). Cell death was measured based on the production of formazan (chromogenic dye), which can be read at OD₅₇₀. Equal concentrations (5, 25, and 50 μg/mL) of kanamycin or chloramphenicol, 1% DMSO, and 10X lysis buffer were used as control. The cytotoxicity level was calculated according to the manufacturer’s instructions (N = two independent experiments with three technical replicate for each SM).

The intracellular reduction in C. jejuni in the presence of a given SM (T1-T8) and 0.5 mM CuSO₄ were evaluated using Caco-2 cells, as previously described (Kumar et al., 2016; Deblais et al., 2019). In brief, cells were infected for 2 h using a multiplicity of infection (MOI) of 100. Infected cells were treated with SM at a final concentration ranging between 5 and 0.315 μg/mL and incubated at 37°C for 24 h in a humidified 5% CO₂ incubator. Following incubation, cells were washed once with 1X PBS, lysed with 0.1% Triton-100X, serially 10-fold diluted in 1X PBS, and plated on MH agar plate. Plates were incubated in microaerophilic conditions at 42°C for 24 h, to determine the intracellular survival. Cells not infected and not treated and cells infected and treated with 1% DMSO were used as controls (N = two independent experiments with four technical replicate for each SM).

Four SMs (T1, T2, T7, and T8) were selected from the initial list of eight SMs based on their low MBC in Caco2 cells (<5 μg SM/ml), low cytotoxicity indexes (<20% toxicity at 5 μg SM /ml), and their minimal effect on commensal bacterial species (less than two species inhibited by the SM out of the seven strains tested; Table 1). The antimicrobial efficacy of our SM was tested on 5-week-old chickens, to assess the clearance of C. jejuni load right before slaughter age. Overall, 5-week-old Campylobacter-free, i.e., specific pathogen-free [SPF] chickens were inoculated orally with a mixture of five C. jejuni strains (10⁵ CFU/chicken, Supplementary Table S3). Rectal swabs were collected 1 day post-inoculation (DPI) to confirm C. jejuni colonization (CFU/g of feces) in birds. From 2 DPI to 7 DPI, groups of 4–5 chickens were treated orally twice a day with one of the four SMs (T1, T2, T7, or T8; approximately 0.225 mg of SM per kg body weight). Details of the treatment groups and the inoculum are shown in Supplementary Table S3.

The antimicrobial efficacy of T1, T7, and T8 was also tested on 3-week-old chickens to assess the clearance of C. jejuni load in chicken right after inoculation. Overall, 3-week-old Campylobacter-free chickens were inoculated orally with a mixture of five C. jejuni strains (10⁵ CFU/chicken, Supplementary Table S3). Rectal swabs were collected 1 DPI to confirm the intestinal colonization of the birds by C. jejuni (CFU/g of feces). From 2 DPI, chickens were treated orally twice a day for 5 days (from 2 DPI to 7 DPI) with one of the SMs (T1, T7, or T8; approximately 0.127 mg of SM per kg body weight; n = 5–6 chicken per group). Details of the treatment groups and the inoculum are shown in Supplementary Table S3.

For both experiments, C. jejuni colonized-chickens treated with 0.0001% DMSO and colonized chickens without additional carrier treatment (non-treated) were used as controls (n = 3–6 chicken per group). Details of the treatment groups and the inoculum are shown in Supplementary Table S3. Both 3 and 5-week-old chickens were euthanized at 7 DPI using carbon dioxide gas and ceca (cecal content and the pouch) and were aseptically collected. One of the ceca pairs was immediately stored at −80°C for microbiota studies. The other ceca were suspended in 1X PBS, homogenized, serially diluted, plated on MH media supplemented with Campylobacter Selective Supplement (CSS) agar plate, and incubated for 48 h at 42°C in microaerophilic conditions to determine C. jejuni load in the ceca (CFU/g of ceca).

All the samples collected during the chicken experiments (n = 58) were selected for microbiota analysis. Genomic DNA was extracted from 150 to 200 mg of cecal contents using the PureLink Microbiome DNA Purification Kit (Life Technologies, Invitrogen Corp.) and combined with RNAse treatment (10 units/h). After quality control with nanodrop, the 16S rRNA V4-V5 variable region was amplified, purified, and sequenced. Amplicon libraries were prepared by using Phusion® High-Fidelity PCR Kit (New England Biolabs Inc., Ipswich, MA, United States), as previously described (Kumar, 2015; Deblais et al., 2018, 2019; Kumar et al., 2018; Srivastava et al., 2020). PCR products were cleaned using AMPure XP PCR (Beckman Coulter Inc., Beverly MA, USA) and were sequenced using Illumina MiSeq 300-base paired-end kit at the Molecular and Cellular Imaging Center. ¹ Sequencing raw data files are publicly available at NCBI Bioproject #PRJNA1023035.

Quality control of the raw reads was performed using FastQC (Babraham Bioinformatics, Cambridge, United States). Trimmomatic was used for trimming and removal of NexteraPE-PE adapter sequences (Bolger et al., 2014). Trimmed reads were processed using QIIME2 v. 2020.11 (Bolyen et al., 2019). The DADA2 plugin was used to process and check the quality of the reads (Callahan et al., 2016). A sequencing depth of 6,600 reads was used for the rarefaction. Taxonomic assignment was performed using QIIME2 and the latest SILVA reference database (version 138.1; 99% homology cut off) (Quast et al., 2013). The obtained reads were further filtered for eukaryotic, mitochondrial, and chloroplastic genetic signatures.

Autodock 4.0 (BIOVIA discovery studio visualizer) (Morris et al., 2009) was used for docking with a homology model, which was generated using an online platform Phyere 2 (Kelley et al., 2015). Due to the unavailability of the 3D crystal structure of C. jejuni TatC, we turned to Aquifex aeolicus VF5, which possessed a TatC homolog. The sequence alignment and secondary structure of A. aeolicus VF5 closely resemble those of C. jejuni TatC (Ramasamy et al., 2013). Of particular significance, the 3D crystal structure of A. aeolicus VF5 is readily accessible in the Protein Data Bank (PDB ID: 4HTS). Consequently, we opted to utilize this structure for the generation of a homology model and Discovery Studio Visualizer along with Chimera to visualize protein–ligand (SM) interaction. Graphical User Interface: AutoDock Tool (ADT) was used for the preparation of pdbqt files for protein and ligand and grid box creation. AutoGrid was employed for the preparation of the grid map, and the grid size was set to 60X60X60 xyz points with grid spacing at 0.375 A. During docking, both protein and ligand were considered as rigid, and the outcomes of docking with 1.0 A in positional root-mean-square deviation (RMSD) were clustered together. The docking pose with most favorable parameters (i.e., lowest energy or binding affinity) was aligned with protein structure and was further analyzed using BIOVIA discovery studio visualizer.

A one-way ANOVA combined with a Tukey’s test was used to analyze the difference in the abundance of C. jejuni in ceca between treatments. A p-value ≤0.05 was considered statistically significant. Alpha diversity was analyzed using Shannon (richness and evenness) and Faith’s PD (phylodiversity) indices. The Kruskal–Wallis test was used to identify difference in alpha diversity. Based on Bray–Curtis distance matrices, permutational multivariate analysis of variances (PERMANOVA) was used to identify difference in beta diversity (unweighted and weighted uniFrac). An analysis of composition of microbes (ANCOM) was used to identify differences in the relative abundance between phylum and species levels (Mandal et al., 2015). A p-value ≤0.01 was considered statistically significant. A multivariate analysis was performed to identify potential correlations between the spectrum of activity and antimicrobial efficacy (MIC/MBC) of the selected SM and the microbiota data and the C. jejuni load in the chicken ceca. Table 2 shows summary of the cutoff selection criteria used in this study, to select the best Tat-dependent SM inhibitors (Ramasamy et al., 2013).

A total of 50,917 SMs divided into 11 libraries were screened against C. jejuni 81–176, to identify potential inhibitors of the Tat system (Supplementary Table S2). The hits were identified by screening in the presence of 1 mM CuSO_4. Since CueO contributed to CuSO₄ resistance and its function requires Tat system transport, increased susceptibility to copper was used as an indicator during high-throughput sequencing for Tat-specific SM inhibitors. The growth profile obtained for each SM tested was compared with the growth of the ΔtatC mutant (susceptible to 1 mM CuSO₄; no OD₆₀₀ increase over time) and C. jejuni 81–176 (resistant to 1 mM CuSO₄; Supplementary Figure S1) in the presence of 1 mM CuSO₄. Out of them, 679 SMs completely inhibited the growth of C. jejuni in the presence of 1 mM copper sulfate after comparison of the turbidimetric values obtained with the C. jejuni ΔtatC mutant. The number of SM identified per library was not proportional to the concentration of the libraries (ranging between 2.5 and 12.5 μg/mL). However, known bioactive collection (Biomol 4, MSDiscovery 1, Microsource 1, and MIH Clinical Collection 1 and 2) displayed higher hit rate (approximately 3.54%) than commercial libraries (Asinex, Chembridge 3, Chemdiv 4, Enamine 2, Lifechemicals 1, and Maybridge 5; approximately 1.20%), suggesting that each library might be composed of SM with distinct chemical structures (Supplementary Table S2). The average hit rate across all libraries was 1.33%, which is significantly higher than the optimal hit rate proposed by the NSRB screening guidelines (0.3% or approximately 1 hit/plate) (Drozd, 2012). Out of the 679 SMs identified, 177 SMs had no predicted bioactivity in eukaryotic cells (i.e., selected SM had no predicted interactions with known eukaryotic targets based on in silico counter screens) and followed the Lipinski rule of five based on in silica analyses (Lipinski et al., 2001; Lipinski, 2004). Among the 177 SMs, 66 SMs had a thiourea group, 46 SM had a benzimidazole group, 38 SM had an acylhydrazone group, and 11 SM had an oxadiazole group. These 177 SMs (107 SM from Asinex, 28 SM from Chembridge, 17 SM from Life chemicals, and 25 SM from Maybridge) were selected for the secondary screen upon resynthesis.

Of the selected 177 SMs tested, only 33.3% of the SM (n = 59/177) reduced at least 30% growth of C. jejuni in the presence of 0.5 mM CuSO₄ compared with the DMSO control (Figure 1A). Twenty-one SM completely inhibited C. jejuni growth, 19 SM reduced the growth of C. jejuni by 75 to 99%, and 19 SM reduced the growth of C. jejuni between 30 and 75% in the presence of 0.5 mM CuSO₄ compared to the DMSO control. The Fdh inhibition activity assay showed that 56.5% of the SMs (n = 100/177) reduced C. jejuni Fdh activity by 30 to 100% compared with the DMSO control (Figure 1B). Overall, 21 SMs that sensitized C. jejuni to copper sulfate also displayed a significant reduction in FDH activity (>30%), suggesting that these SMs affect the growth of C. jejuni in a Tat-dependent manner. These 21 SMs were selected for further testing.

The inhibitory activity of the 19/21 SM (2 SM could not be resynthesized) was also tested on other Campylobacter strains (n = 11; Supplementary Table S1) and commensal/beneficial gut bacteria (n = 7) using a copper sensitivity assay at 0.5 mM CuSO₄ with 6.25 μg of SM_, as mentioned above. Overall, 19 SMs completely inhibited the growth of 12 Campylobacter strains in the presence of copper sulfate while having minimal impact on the growth of two commensal/beneficial gut bacteria in a similar growth condition (growth inhibition in the presence of copper sulfate lower than 50% compared with the DMSO control). The selected 19 SMs displayed a very similar spectrum of activity profiles. They completely inhibited the growth of all Campylobacter strains tested at 6.25 μg. Furthermore, the 19 SMs had no growth effect on Escherichia coli Nissle 1917, Streptococcus bovis, Bifidobacterium adolescentis, Bifidobacterium longum, and Bacteroides thetaiotaomicron at 6.25 μg in the presence of 0.5 mM CuSO₄; however, they affected the growth of Lacticaseibacillus rhamnosus GG (n = 13 SM), Enterococcus faecalis (n = 16 SM), and Levilactobcillus brevis (n = 2 SM; Table 1).

A copper sulfate sensitivity dose–response assay was performed in vitro on C. jejuni 81–176 using the selected 19 Tat-dependent inhibitors (Figure 2). One SM had an MIC of 3.13 μg/mL; two SMs had MIC at 1.56 μg/mL; three SMs had MIC at 0.78 μg/mL; one SM had MIC at 0.39 μg/mL; two SMs had MIC at 0.19 μg/mL; two SMs had MIC at 0.098 μg/mL; two SMs had MIC at 0.049 μg/mL; three SMs had MIC at 0.024 μg/mL; and one SM had MIC at 0.012 μg/mL in the presence of 0.5 mM CuSO₄. Interestingly, the number of strains affected by the compounds (i.e., spectrum of activity) was not correlated with their antimicrobial efficacy (e.g., MIC and MBC).

Based on the antimicrobial activity (i.e., efficacy and spectrum of activity) and in silica data obtained, we selected eight SMs with high drug-like properties, little or no growth effect on commensal/beneficial gut bacteria, and high efficacy against several Campylobacter strains (Table 1; Supplementary Table S1). Overall, the antimicrobial efficacy of the selected eight SMs was similar across the 12 Campylobacter strains tested with MIC values of 0.01 μg/mL and higher, and MBC values of 0.04 μg/mL and higher (Supplementary Tables S4, S5, respectively). In addition, the three-dimensional analysis of the chemical structure of the eight selected SMs showed that T8 and T2, T6 and T1, and T7 and T3 displayed high structural similarities, while T4 and T5 had unique chemical structures (Figure 3). Furthermore, most of the SMs have sulfur and/or nitrogen-based functional groups (thiourea, imidazole uracil, sulfonamide, thiomorpholine dioxide, phenylurea, pyridine, piperidine, oxadiazole, and quinoline). However, no association was detected between the chemical structure of the SM and their antimicrobial properties.

All SMs completely cleared internalized C. jejuni 81–176 in Caco-2 cells after 24 h of treatments with a concentration of SM ranging from 0.63 μg/mL to 10 μg/mL (Table 1). Most of the SMs displayed low toxicity (at most 10%) to Caco-2 cells when treated with 50 μg/mL for 24 h (Figure 4). The toxicity values were comparable to the kanamycin and chloramphenicol-treated cells. Only T1 and T8 displayed toxicity level (36 and 33%, respectively) when treated with 25 μg/mL or 50 μg/mL of SM for 24 h; however, T1 and T8 cleared intracellular C. jejuni 81–176 at 0.63 μg/mL and 5 μg/mL, respectively, which represent concentrations up to 80-fold lower compared with ones used for the toxicity assay. Overall, T1, T2, T7, and T8 displayed the most promising antimicrobial properties in vitro were selected for further analyses.

As a proof of concept, the selected SMs (T1, T2, T7, and T8) were tested on 5-week-old chickens inoculated with a mixture of C. jejuni strains (Supplementary Table S1), to assess the clearance of C. jejuni in chicken right before the slaughter. Following 5 days of SM treatment (twice a day, approximately 0.225 mg of SM per kg body weight per treatment), colonized chickens treated with T2 displayed a significant reduction (1.2-log) in C. jejuni population per gram of cecum compared with the DMSO control group, which harbored approximately 5×10⁷ CFU per gram of cecum (p < 0.01; Figure 5A). Colonized chickens treated with the T7 and T8 groups displayed a 0.5-log reduction in C. jejuni population in ceca compared with the DMSO control group (p > 0.05), while the T1 treatment did not reduce the abundance of C. jejuni in ceca compared with the DMSO group. No significant difference in body weight was recorded between treatment groups before or after 5 days of treatment (p > 0.05; Supplementary Figure S2A).

The antimicrobial activity of T1, T7, and T8 was also tested on 3-week-old chickens inoculated with a cocktail of C. jejuni strain (Supplementary Table S1), to assess the clearance of C. jejuni in young chicken right after inoculation. T2 could not be resynthesized in sufficient quantities and, thus, was not tested on this 3-week-old experiment. Following 5 days of SM treatment (approximately 0.127 mg of SM per kg body weight per treatment), 3-week-old chickens treated with T7 displayed a significant reduction (0.9-log) in C. jejuni population per gram of cecal contents compared with the DMSO-treated group, which harbored approximately 4 × 10⁸ CFU per gram of cecal contents (p < 0.01; Figure 5B). Chickens treated with T1 and T8 harbored similar C. jejuni abundance in ceca compared with the DMSO control group. No significant difference in body weight was recorded between treatment groups before or after 5 days of treatment (p > 0.05; Supplementary Figure S2B).

The impact of the SM treatments on the cecal microbiota of the 3 and 5-week-old chickens was studied using 16S sequencing. After processing and taxonomic assignment with the SILVA reference database, 682,777 sequences were obtained from the 58 samples studied. Sequencing depth varied between 6,648 and 15,749 reads per sample (mean = 9,753 reads per sample). Cecal samples were normalized to 6,600 sequences per sample for the data presented below.

For the 3-week-old chicken experiment, the analysis of the cecal alpha diversity, using Faith’s PD and Shannon diversity index, indicated no significant differences between the treatment groups. On the other hand, the DMSO treatment (DMSO, T1, T2, T7, and T8 groups) significantly increased the Shannon index value (approx. 6.3) compared with the colonized, non-treated groups (PC; approx. 5.5; p < 0.04). The beta diversity analysis demonstrated that the SM treatments (T1, T2, T7, and T8 groups) had minimal impact on the global microbiome composition compared with the DMSO group (Supplementary Figure S3A). Most of the variations were detected between the colonized chickens treated with DMSO (DMSO, T1 and T7 groups) and the chickens that were not treated with DMSO (NC and PC groups; p < 0.01). Additional details concerning the impact of the DMSO and C. jejuni colonization on the cecal microbiome at the genus and species levels are shown in Supplementary Figures S3A–C. Overall, the cecal microbiome was composed of Firmicutes (90–99%), followed by Proteobacteria (0.1–6.7%), Tenericutes (0.6–2.1%), Actinobacteria (0.06–1.4%), and Epsilonbacteraeota (0.02–0.1%, Figure 5C). No significant differences were detected at the phylum level between the SM-treated groups (T1, T2, T7, and T8) and the associated control group (DMSO). At the species level, T8 treatment significantly increased GCA-900066225 (11.6-fold), and T7 treatment significantly increased Ruminococcaceae UCG-014 (2.2-fold) compared with the DMSO group (p < 0.01).

For the 3-week-old chicken experiment, the analysis of the cecal alpha diversity, using Faith’s PD and Shannon diversity index, indicated no significant differences between the treatment groups of (Supplementary Figures S3D,E). Similarly, beta diversity analysis, using the weighted uniFrac, confirmed that the SM treatments (T1 and T7 groups; T8 was not included for microbiota study since it did not have positive impact on C. jejuni load in the cecum) had minimal impact on the global microbiome composition compared with the DMSO group (Supplementary Figure S3F). Most of the variations were detected between the colonized chickens treated with DMSO (DMSO, T1 and T7 groups) and the chickens not treated with DMSO (NC and PC groups; p < 0.01). Additional details concerning the impact of DMSO and inoculation of C. jejuni on the cecal microbiome at the genus and species levels are shown in Supplementary Figures S3D–F. Overall, the cecal microbiome was composed of Firmicutes (62–82%), followed by Proteobacteria (15–35%) and Epsilonbacteraeota (0–8%, Figure 5D). No significant differences were detected at the phylum level between the SM-treated groups (T1 and T7) and the associated control group (DMSO). At the species level, Caproiciproducens (6.5-fold and 4.4-fold, respectively) and Flavonifractor (1.7-fold) were significantly higher in T1 and T7 groups compared with the DMSO group. Eubacterium coprostanoligenes group (8-fold) and Neglecta timonensis (only detected in the T7 group at 0.8%) were significantly higher in the T7 group compared with the DMSO group (p < 0.01).

TatC subunit is the largest and most important part of Tat system, while Tat A and Tat B are much smaller units. The docking studies were conducted with all Tat subunits; however, TatA and B protein folding could not be predicted with high certainty due to smaller protein size. Therefore, we focused on TatC, which was modeled with a high degree of predictability and well-defined binding pockets, giving repeatable docking results. The in silico docking study demonstrated that both active compounds T2 and T7 bind in the same hydrophobic binding pocket associated with the key active residue Glu 165 (Figure 6). The calculated binding energy for T2 is −6.26Kcal/mol, while T7 has a higher binding energy of −8.0 Kcal/mol. The T2 thiourea nitrogen forms a hydrogen bond with Ile80, while the phenyl ring forms Pi-Pi stacking interaction with Phe87. Trp85 and Phe87 form Pi-sigma interaction with the T2 pyridine ring and tert-butyl group, respectively (Figures 6A,C). In addition, there are several hydrophobic interactions within carbon chains of T2 and Ile80, Ser 77, Ile162, Phe111, Ser107, Ile168, Met166, Gln83, and Ser17. The T7 functional groups (i.e., oxadiazole and phenyl groups) bind to several amino acids that are located inside the hydrophobic core of TatC (Figure 6B). More precisely, all aromatic rings of T7 have π-anion interactions with the key Glu165 (Figures 6B,D); the phenylacetamide group has π-alkyl interactions with Val169, π-π interactions with Phe84 and Trp85, and van der Waals interactions with Ile168 and Ser107; the methylthio-phenyl group has π-alkyl interactions with Ile80, Ile162, and Leu81, and van der Waals interactions with Phe111, Ser77, and Phe73. π- Σ interactions are also detected on the amide and methanethiol groups. Thus, anti-C. jejuni activity of both T2 and T7 may be attributed to their binding to hydrophobic pocket in TatC and the interaction with key residues Glu165 and Trp85 (Figure 6).