Bacterial strains used in this study were: E. coli ATCC 9723H, Pseudomonas aeruginosa ATCC 10148, Staphylococcus aureus ATCC 6538P (penicillin resistant), Streptococcus pyogenes (hospital isolate, strain unknown), Mycobacterium smegmatis MC²155, N. gonorrhoeae ATCC 700825 (streptomycin resistant), and Salmonella enterica subsp. enterica serovar Typhimurium SL1344 (streptomycin resistant). E. coli, P. aeruginosa, S. aureus, M. smegmatis, and S. enterica were cultured aerobically at 37°C in Mueller Hinton broth (MHB) (Sigma-Aldrich, MO, United States) and on nonselective Mueller Hinton agar (MHA) plates. S. pyogenes was cultured in 5% carbon dioxide at 37°C in MHB supplemented with 5% lysed horse blood (Quad Five, MT, United States) (MHB + 5% HB) and on nonselective MHA plates supplemented with 5% defibrinated sheep blood (Quad Five, MT, United States) (MHA + 5% SB). N. gonorrhoeae was cultured in 5% carbon dioxide at 37°C in gonococcal (GC) chocolate broth medium and on nonselective GC agar plates [GC medium base (BD Difco, MD, United States) supplemented with 1% BBL™ hemoglobin (BD Biosciences, MD, United States) and 1% IsoVitaleX (BD Biosciences, MD, United States)]. Prior to antimicrobial susceptibility assays (as described below), N. gonorrhoeae cultures were subcultured on nonselective GC agar plates (as described above), to ensure bacterial viability. All bacterial stocks were stored in 20% glycerol at –80°C.

Treponema pallidum subsp. pallidum (Nichols strain) was propagated in, and extracted from, New Zealand White rabbits as described elsewhere (Lukehart and Marra, 2007), and stored in liquid nitrogen. Frozen treponemal stocks were then used for in vitro culture and sub-culture of T. pallidum in the presence of Sf1Ep (NBL-11) cottontail rabbit epithelial cells (ATCC CCL-68) [American Type Culture Collection (ATCC), Rockville, MD, United States]. Continuous axenic culture of T. pallidum in the absence of mammalian cells has not been achieved, and it is believed that the direct adherence of T. pallidum to Sf1Ep cells is required for the long term replication of T. pallidum in vitro (Edmondson et al., 2018). Dissociation of T. pallidum from Sf1Ep cells was accomplished using trypsin-free dissociation buffer [2 mL: 64% cell culture grade water (Sigma Aldrich), 10% modified EBSS (Earle’s Balanced salt solution, 10×), 1% non-essential amino acids (Thermo Fisher Scientific), 0.15% sodium bicarbonate (Sigma Aldrich), 0.728% 100 mM sodium pyruvate (Sigma Aldrich), 0.136% 0.5M EDTA (Thermo Fisher Scientific), 0.16 mg dithiothreitol (DTT) (Sigma Aldrich)] followed by a low speed centrifugation step (220 × g) to separate T. pallidum from the rabbit cells, as previously described (Edmondson et al., 2018; Edmondson and Norris, 2021).

The flow diagram shown in Figure 1 outlines all major steps that comprised our bioinformatics pipeline for the identification of potential AMPs and AMPCCRs in the T. pallidum proteome. As the first step in this approach, the whole proteome of T. pallidum (Nichols strain NC_021490) was obtained from the National Center for Biotechnology Information (NCBI) Genome database¹ and manually searched in order to identify all functionally-unannotated miniproteins containing 150 amino acids or less (Figure 1A). All in silico proteomic analyses performed on the T. pallidum strain reported in the current study were based on the NCBI whole proteome annotation from July 2021.

The full-length amino acid sequences of all functionally-unannotated miniproteins identified in the above analyses were submitted to three AMP prediction servers: (i) AMP Scanner Version 2 (Deep Neural Learning Method for predicting antibacterial activity only)² (Veltri et al., 2018), (ii) iAMPpred [Support Vector Machine (SVM) algorithms for predicting antibacterial, antiviral, and antifungal activities; machine learning method based on amino acid composition, physicochemical, and structural features]³ (Meher et al., 2017), and (iii) CAMP [SVM, Random Forest (RF), Artificial Neural Network (ANN), and Discriminant Analysis (DA) algorithms for predicting antimicrobial activity; machine learning method based on different physicochemical properties of proteins]⁴ (Waghu et al., 2016; Figure 1B). All T. pallidum miniproteins of unknown function were ranked from most likely AMP to least likely AMP based on the number of different server algorithms (out of eight total) that produced positive AMP predictions. Corresponding mean probability scores (1 [most likely prediction score]-0 [least likely prediction score]) were then used to rank all miniproteins within each positive AMP prediction class (Figure 1B).

To determine if any of the miniproteins of unknown function from T. pallidum contain potential Sec-dependent leader peptides (Sec/SP1 peptides), we used the signal peptide prediction servers, SignalP 5.0⁵ (Almagro Armenteros et al., 2019) and LipoP 1.0⁶ (Juncker et al., 2003). Manual searches of the T. pallidum miniproteins for the conserved double-glycine/glycine-alanine leader peptide motif (M[R/K]ELX₃E[I/L]X₂[I/V]XG[G/A]) that has been observed in AMPs from Gram-negative bacteria (Michiels et al., 2001; Dirix et al., 2004) were performed to identify proteins that contain Glycine-Glycine and/or Glycine–Alanine pairs within the first 31 residues of the N-terminus (Figure 1B). A multiple sequence alignment of the N-termini of each of the Glycine-Glycine and/or Glycine-Alanine-containing proteins was generated using Clustal Omega⁷ (Madeira et al., 2019). WebLogo⁸ (Crooks et al., 2004) was then used to generate a sequence logo for the identification of sites within the N-terminal residues of these proteins that exhibit homology to the conserved double-glycine/glycine-alanine leader peptide motif found in AMPs from Gram-negative bacteria (Michiels et al., 2001; Dirix et al., 2004). Physicochemical properties, including hydrophobicity, net charge, and presence/absence of cysteine residues, of the T. pallidum miniproteins of unknown function were analyzed using the APD3 database Antimicrobial Peptide Calculator and Predictor⁹ (Wang et al., 2016; Figure 1B). Tertiary structure modeling of functionally-unannotated proteins of interest (>150 amino acids) found to be located in close proximity to T. pallidum miniproteins identified above, was performed using the protein structure modeling server, Phyre2¹⁰ (Kelley et al., 2015).

Two identified AMP candidates [Tp0451a (accession number WP_014342798) and Tp0749 (accession number WP_010882194)] were further analyzed using our in-house bioinformatics pipeline (Figure 1C). This analysis was performed to help confirm the initial AMP predictions generated by the three prediction servers and to identify potential antimicrobial peptide critical core regions (AMPCCRs) within Tp0451a and Tp0749. This pipeline was comprised of the following four stages:

Full-length amino acid sequences of Tp0451a and Tp0749 were submitted to four servers: AMPA¹¹ (Torrent et al., 2012), CAMP antimicrobial region prediction server (CAMP-ARP)¹² (Waghu et al., 2016), AntiBP¹³ (Lata et al., 2007), and AntiBP2¹⁴ (Lata et al., 2010). Potential AMPCCRs were identified in each protein based on high probability scoring regions (15–23 amino acid stretches) that were predicted by three or more servers.

Secondary structure analyses of potential AMPCCRs were performed using the structure prediction servers Jpred 4¹⁵ (Drozdetskiy et al., 2015) and PSIPRED 4.0¹⁶ (Jones, 1999) and the alpha helix screening and physicochemical characterization server, HeliQuest¹⁷ (Gautier et al., 2008). Structure modeling of potential AMPCCRs based on PSIPRED secondary structure predictions were performed using the de novo peptide structure prediction server, PEP-FOLD 3¹⁸ (Thevenet et al., 2012; Shen et al., 2014) and Swiss Model¹⁹ (Bienert et al., 2017; Waterhouse et al., 2018). All models were then vetted through the comparative protein structure modelling server, Modeller²⁰ (Webb and Sali, 2016) and the homology modeling program, ICM (Molsoft L.L.C., CA, United States)²¹ (Cardozo et al., 1995), for their lowest normalized discrete optimized protein energy value (zDOPE) and GA341 score closest to 1. Comparative homology modeling using structure-based alignment was performed using PROMALS3D²² (Pei et al., 2008). Together, these structure prediction analyses were used to support the multi-server AMPCCR predictions through the identification of structural folds known to be important for AMP function and for facilitating subsequent peptide design via the prediction of intact secondary structure elements.

Amino acid homology searches using the APD3 database (see text footnote 9) (Wang et al., 2016) were employed to determine similarity and identity of predicted T. pallidum AMPCCRs with known and experimentally-validated AMPs and for the identification of short orthologous AMP sequences that would be otherwise missed using the NCBI BLAST tools²³ (Altschul et al., 1990);

AMPCCR cell penetrating abilities, a key functional feature of AMPs, were predicted using the CellPPD Protein Scanning Tool²⁴ (Gautam et al., 2013) (peptide fragment length = 10; prediction method = SVM based with scores ranging from –1.0 to +1.0). To increase prediction stringency, the SVM threshold for positive cell penetrating peptide predictions was increased from the default threshold (0) to ≥0.1.

RNA was isolated and purified from in vivo-harvested T. pallidum subsp. pallidum (Nichols strain) using Invitrogen TRIzol™ reagent (Thermo Fisher Scientific, MA United States) and the RNeasy mini kit (Qiagen, ON, Canada), according to the manufacturer’s instructions. RT-PCR was performed (after genomic DNA digestion/removal) using the orientation-specific RT-PCR sense (5’-aatgtcggctaccatcgctc) and antisense (5’-acgtgctctgccaattactgc) primers for tp0451a and the Invitrogen SuperScript™ IV First-Strand Synthesis System (Thermo Fisher Scientific), according to the manufacturer’s instructions. The negative control RT-PCR reaction did not include reverse transcriptase. PCR products were electrophoresed on agarose gels and visualized with ethidium bromide staining.

For the experimental validation of antimicrobial and immunomodulatory activity, four putative AMPCCR peptides (Table 1) from two T. pallidum miniproteins that were identified using our AMP bioinformatics prediction pipeline (Tp0451a_N, Tp0451a_C, Tp0749_N, and Tp0749_C), a cysteine-to-serine substituted version of Tp0749_C (Tp0749_C_C61S), and a cysteine-to-serine substituted version of Tp0451a_C (Tp0451a_C_C85S), were synthesized without chemical modifications using the PepPower™ solid state peptide synthesis (SSPS) platform at GenScript (NJ, United States). The known AMP, human cathelicidin LL-37 (Turner et al., 1998), the known bullfrog (Rana [Lithobates] catesbeiana) AMP, RaCa-2 (Li et al., 2022), a scrambled version of LL-37 (sLL-37), and a peptide (Tp0751_p5) from the T. pallidum adhesin Tp0751 (Cameron et al., 2005) (Table 1) were also synthesized via the same SSPS platform at GenScript, and used as positive (LL-37 and RaCa-2) and negative (sLL-37 and Tp0751_p5) controls in antimicrobial susceptibility and immunomodulation assays, as described below.

The broth microdilution technique (Wiegand et al., 2008) was used to determine if the T. pallidum peptides are capable of exhibiting bacteriostatic [minimal inhibitory concentration (MIC) measurements] and/or bactericidal [minimal bactericidal concentration (MBC) measurements] activity. Bacterial suspensions were prepared by transferring bacterial colonies into MHB and resuspending using a vortex mixer to ensure complete suspension of any bacterial aggregates. Turbidity of the colony suspensions was adjusted spectrophotometrically to the required optical densities to achieve a turbidity equivalent to that of a 0.5 McFarland standard (1–2 × 10⁸ CFU/mL) followed by dilution in MHB to achieve the standardized microbial inoculum of approximately 5 × 10⁵ CFU/mL. Total viable counts (TVC) were routinely performed on all inoculum suspensions to ensure correct bacterial cell densities. The standardized bacterial suspensions were then incubated with two-fold serial dilutions of each peptide (dissolved in 11 μL of ultrapure sterile water; final peptide concentration range of 256 μg/mL–0.5 μg/mL) in Greiner polypropylene round bottom 96-well microtiter plates (Sigma-Aldrich, MO, United States). Each peptide was tested once per experiment, with a range of 3–9 independent experiments performed per peptide. Negative growth/sterility control wells contained bacterial growth media (100 μL) and the peptide solvent (11 μL of ultrapure sterile water). Positive growth control wells contained the standardized number of bacterial cells (100 μL of ∼ 5.0 × 10⁵ CFU/mL) and peptide solvent (11 μL of ultrapure sterile water). Plates were incubated at 37°C for 16–24 h and MICs were determined using the unaided eye to identify the lowest concentration of AMP that inhibited visible growth of the tested bacterial species. If the sterility control well was turbid, the test was not considered valid. MBCs were determined by plating the entire content of the wells containing the peptide/bacteria mixture representing the MIC and the entire contents of the preceding wells containing two-fold and four-fold more concentrated AMP dilutions onto nonselective agar plates. Plates were incubated for 24 h at 37°C and MBCs were calculated as the percentage of bacteria killed at the different AMP concentrations tested (decrease in TVCs from the MBC plates compared to the initial bacterial suspension of ∼5 × 10⁵ CFU/mL).

The potential antimicrobial activity of the T. pallidum peptides against N. gonorrhoeae was determined using a modified agar dilution method. In this assay, N. gonorrhoeae colonies from GC chocolate agar plates were resuspended in MHB and the turbidity of the suspension was adjusted, as described above, to achieve the standardized microbial inoculum of approximately 5 × 10⁵ CFU/mL. A two-fold serial dilution of the peptides (11 μL in sterile ultrapure water) was then prepared in wells 1–10 of a 96-well sterile polypropylene plate to obtain a dilution series corresponding to 10 times the required testing concentrations (2,560, 1,280, 640, 320, 160, 80, 40, 20, 10, 5 μg/mL). The bacterial suspension (100 μL) was dispensed into the wells containing the peptides. A negative growth/sterility control (well 12) contained bacterial growth media (100 μL) and the peptide solvent (11 μL of ultrapure sterile water). A positive growth control (well 11) contained the standardized number of bacterial cells (100 μL of ∼ 5.0 × 10⁵ CFU/mL) and peptide solvent (11 μL of ultrapure sterile water). The 96-well plate was then incubated at 37°C in an atmosphere of 5% CO₂ for 3 h to allow for peptide binding and antimicrobial activity. After the incubation period, an aliquot (20 μL) from wells 1–12 was removed and spotted onto the surface of a GC chocolate agar plate. N. gonorrhoeae spotted plates and TVC plates were incubated at 37°C in an atmosphere of 5% CO₂ for 18–24 h. Following the incubation period, MICs were determined by identifying the lowest concentration of peptide that completely inhibited visible growth on the agar plate. To determine the bactericidal activity of peptides, total viable counts (TVCs) were also prepared on GC chocolate agar plates for the 3 h-incubated peptide/bacteria mixtures. These counts were compared with TVCs from the corresponding positive growth wells to give the percentage of bacteria killed by each of the peptides following the 3-h incubation.

An antimicrobial susceptibility assay was developed to assess the activity of the four treponemal peptides (Tp0451a_N, Tp0451a_C, Tp0749_N, and Tp0749_C), an equimolar mix of Tp0451a_N and Tp0451a_C, and the negative (Tp0751_p5) and positive (LL-37) control peptides, against T. pallidum. In vitro-grown T. pallidum (100μL; 1.0–1.2 × 10⁶ Tp/mL), prepared as described above, were incubated with each peptide at three concentrations (4, 16, 64 μg/mL) or the Tp0451a_N/Tp0451a_C mix (21.6 μM; ∼85 μg/mL Tp0451a_N and ∼64 μg/mL Tp0451a_C) at 34°C in an atmosphere of 93.5% nitrogen, 5% carbon dioxide, and 1.5% oxygen. Darkfield microscopy was used to monitor T. pallidum viability by counting motile treponemes at 0, 1, 2, and 4 h post co-incubation. For each viability measurement, at least 50 treponemes were observed for each sample at each time point.

Human THP-1 (ATCC TIB-202) monocytes (American Type Culture Collection, VA, United States) were cultured and maintained in 5% CO₂ at 37°C in RPMI-1640 medium (Gibco, Life Technologies, ON, Canada) supplemented with 10% (v/v) fetal bovine serum (FBS) (Fisher Scientific, ON, Canada), 0.05 mM 2-mercaptoethanol (BME) (Sigma-Aldrich, ON, Canada), penicillin (100 units), and streptomycin (0.1 mg/mL) (Sigma-Aldrich, ON, Canada) (hereafter referred to as “complete RPMI-1640 medium”). Cells were passaged at a density of 8 × 10⁵ cells/mL to a maximum of 15 passages and re-seeded at 3 × 10⁵ cells/mL for maintenance. For differentiation into plastic-adherent macrophage-like cells, THP-1 monocytes were seeded at a density of 3 × 10⁵ cells/mL in T75 tissue culture flasks (Fisher Scientific, Ottawa, ON, Canada) and stimulated with a low concentration (25 ng/mL) of phorbol-12-myristate-13-acetate (PMA) (Sigma-Aldrich, ON, Canada) in complete RPMI-1640 medium for 48 h in 5% CO₂ at 37°C. The PMA-mediated differentiation method results in the generation of cells with phenotypic characteristics that are similar to human peripheral blood mononuclear cell (PBMC) monocyte-derived macrophages; they are adherent, larger, more phagocytic, are less proliferative, and exhibit cell surface markers that are characteristic of macrophages (Chanput et al., 2014).

Following incubation with PMA, light microscopy was used to ensure the differentiated cells were adherent, and exhibited morphological changes consistent with PBMC monocyte-derived macrophages. Non-adherent cells were then removed by washing with sterile, calcium- and magnesium-free phosphate-buffered saline (PBS) (ThermoFisher Scientific, MA United States). Plastic-adherent macrophage-like cells were detached by a three- to five-minute treatment with trypsin-EDTA (0.05%) (ThermoFisher Scientific, MA, United States) and physical agitation. The macrophage-like cells were then centrifuged at 1,500 rpm using a Sorvall Model STR04 centrifuge (ThermoFisher Scientific, MA United States) for 5 min at room temperature and seeded in PMA-free complete RPMI-1640 media, as described below.

THP-1 monocytes and THP-1 macrophage-like cells were seeded into the wells of 12-well plates (Thermo Fisher Scientific, ON, Canada) in complete RPMI-1640 medium (1 mL) at a density of 0.5 × 10⁶ cells/mL. Following seeding and prior to stimulation, macrophage-like cells were rested overnight. For AMP alone stimulation, monocytes or macrophages were exposed for 24 h at 37°C in 5% CO₂ to either (i) no stimulation (negative control for baseline cytokine production), (ii) lipopolysaccharide [LPS; from S. enterica serovar Typhimurium (Sigma-Aldrich, ON, Canada)] (positive control for cytokine production; final concentration 1.0 μg/mL), or (iii) the test peptides (control peptides and potential T. pallidum AMPs listed in Table 1; final concentration 25 μg/mL). For AMP/IL-32γ co-stimulation, 20 ng/mL IL-32γ (R&D Systems, MN, United States) in fresh complete RPMI-1640 medium was added to the rested macrophages. IL-32γ stimulation was immediately followed by co-stimulation by the addition of the test peptides at a final concentration of 25 μg/mL. Macrophage cells left unstimulated or stimulated with IL-32γ alone were used as negative and positive controls, respectively. Cells were stimulated for 24 h at 37°C in 5% CO₂. Following stimulation, monocytes and macrophage cells were centrifuged at 1,500 rpm using a Sorvall Model STR04 centrifuge (ThermoFisher Scientific, MA, United States) for 5 min at room temperature and the cell-free supernatant was stored at -80°C prior to quantification of cytokine levels, as described below.

The BD™ Cytometric Bead Array (CBA) system (BD Biosciences, CA, United States) was used to quantify the expression of tumor necrosis factor (TNF), MCP-1, IL-6, IL-8, IL-10, and IL-1β according to manufacturer’s instructions. For statistical analyses, data were analyzed for normality using a D’Agostino-Pearson omnibus normality test and a Shapiro–Wilk normality test. An ordinary one-way ANOVA followed by Dunnett’s multiple comparisons test was used to assess differences between three or more groups of normally distributed data. A Kruskal–Wallis test followed by Dunn’s multiple comparisons test was used to assess differences between three or more groups of data that were not normally distributed.

Although the size of AMPs can vary greatly, ranging from approximately five amino acids to several hundred amino acids, a search of the AMP database, APD3²⁵ (Wang et al., 2016), indicated that 97% of the 3324 AMPs listed at the time of analysis fall within the 10–150 amino acid size range. With this knowledge, we sought to identify potential T. pallidum AMPs by manually searching the whole proteome of T. pallidum and filtering for proteins containing 150 amino acids or less. This resulted in the identification of 151 miniproteins (≤150 amino acids) (Supplementary Table 1), representing ∼16% of the T. pallidum proteome. We then filtered for miniproteins with no assigned function or weak/incomplete annotated functions and for miniproteins with a potential AMP-related function, resulting in the identification of 68 proteins (Supplementary Table 2), representing 7% of the T. pallidum proteome. Genes corresponding to four of these 68 proteins (Tp0039, Tp0130, Tp0451a, and Tp0867) were not included in the latest annotation of the T. pallidum proteome (Nichols strain NC_021490), however, all four genes have been shown to be expressed at the transcript level (Smajs et al., 2005 and current study) justifying their inclusion in this study (Supplementary Table 2). Sixty-seven of the 68 miniproteins of unknown function that were identified in T. pallidum were annotated in the published proteome from July 2021 as either “hypothetical proteins,” “DUF (Domains of Unknown Function) domain-containing proteins,” or as proteins with motifs/domains that do not provide enough insight to confidently indicate potential protein functions (e.g., helix-turn-helix domain-containing proteins, DNA- or RNA-binding proteins, zinc ribbon domain-containing protein) (Supplementary Table 2). One of the 68 miniproteins was annotated as a putative CPBP family intramembrane metalloprotease (TPANIC_RS05485), some of which may be involved in AMP processing (Pei et al., 2011). However, this 91-amino acid treponemal protein is at least two-four-fold smaller than other bacterial CPBP proteins, is only predicted to contain one transmembrane segment (unlike the four or more present in known CPBP proteins), and does not contain the four conserved sequence motifs required for proteolytic activity that are found in other CPBP proteins (Pei et al., 2011). In light of these findings, Tp_RS05485 was included in the list of 68 miniproteins for further bioinformatics analyses (Supplementary Table 2).

Full-length amino acid sequences of all 68 miniproteins identified in the analyses described above were submitted to three AMP prediction servers. Prediction data were then used to rank the 68 treponemal proteins from most likely AMP (ranking 1/68) to least likely AMP (68/68) depending on (i) how many of the eight server algorithms produced positive AMP predictions and (ii) the mean probability scores from each of the AMP predictions for each protein. In summary, 45 high-priority T. pallidum AMP candidates predicted by at least four of the eight algorithms were assigned mean probability scores of at least 0.505 (50.5% probability) (Figure 2 and Supplementary Table 3). AMP prediction results for all 68 T. pallidum miniproteins and corresponding probability scores from each of the AMP prediction servers are listed in Supplementary Table 3.

Many bacterial AMPs are synthesized as inactive preproteins with N-terminal leader/signal peptides whose presence and cleavage are required for export and activation, respectively. These include signal peptides recognized by the general secretory (Sec) pathway [Sec-dependent signal peptides (Sec/SP1 peptides)] (Leer et al., 1995; Chiorean et al., 2018) and Sec-independent double-glycine/glycine-alanine (GG/GA) leader peptides that have been documented in AMPs from both Gram-positive and Gram-negative bacteria (Havarstein et al., 1994; Oman and van der Donk, 2010). The signal peptide prediction servers, SignalP and LipoP, were used to search for the presence of Sec-dependent SP1 signal peptides, which predicted the presence of Sec/SP1 signal peptides in only four of the 68 miniproteins (Supplementary Table 4). Manual searches of the 68 T. pallidum AMP candidates for the conserved double-glycine/glycine-alanine leader peptide motif identified 24 proteins that contain Glycine-Glycine and/or Glycine-Alanine pairs within the first 31 residues of the N-terminus (Supplementary Figure 1). WebLogo analysis of the 24 proteins identified an N-terminal region with similarity to the double-glycine/glycine-alanine leader peptide motif from AMPs from Gram-negative bacteria, suggesting the presence of a similar secretion/activation recognition signal in T. pallidum candidate AMPs (Supplementary Figure 2).

Physicochemical properties known to be important for AMP function were calculated using the APD3 online AMP calculator and predictor tool. Consistent with the high content of arginine and lysine residues in AMPs, the top 22-ranking potential T. pallidum AMPs were found to have mean net charges at pH 7.0 of 5.56 (8 proteins with 8/8 positive AMP predictions) and 7.91 (14 proteins with 7/8 positive AMP predictions) (Figure 3A and Supplementary Table 4). Although no trend was observed between AMP likelihood rankings and the percentage of hydrophobic amino acids found within this group of proteins, the mean hydrophobic residue content of all 68 miniproteins was high (43.1%), with 66/68 proteins comprised of more than 30% hydrophobic residues (Figure 3B and Supplementary Table 4). Also observed was a trend in AMP likelihood rankings and the number / percentage of cysteine residues per protein. In general, cysteines were found to be more common in higher-ranking predicted AMPs (Figure 3C and Supplementary Table 4). In comparison to an average cysteine content of ∼1.9% found in all T. pallidum proteins [calculated from the Nichols strain (NC_021490) proteome], the top eight-ranking predicted AMPs (8/8 positive AMP predictions) contained on average almost three-fold more cysteines (mean cysteine residue content = 5.31%). Interestingly, the cysteine-rich nature of these T. pallidum miniproteins is shared with distinct classes of eukaryotic (Simmaco et al., 1994; Fahrner et al., 1996; Shafee et al., 2016) and prokaryotic (Baindara et al., 2017; Sugrue et al., 2020) AMPs, and thus may represent an important physicochemical property for protein structure and/or function.

Most of the 68 T. pallidum miniproteins from the current study are annotated in the published T. pallidum proteome as “hypothetical” proteins. However, DNA microarray-based analysis of the T. pallidum transcriptome following experimental rabbit infection (Smajs et al., 2005) demonstrated that 56/68 genes encoding these functionally uncharacterized proteins are expressed at the transcriptional level (Supplementary Table 5). Transcripts from most of the other 12 genes were not searched for in the study (Smajs et al., 2005) as they were not annotated in the T. pallidum genome at the time the study was performed. In addition, peptides from 15/68 miniproteins were detected in mass spectrometry-based proteomics studies of rabbit infections (McGill et al., 2010; Osbak et al., 2016), including the protein with the highest level of expression in the Osbak and colleagues study, Tp0214 (Supplementary Table 5). The use of trypsin for T. pallidum protein digestion in the two mass spectrometry studies may have contributed to the low number of miniproteins detected in these experiments. Given that miniproteins are small and contain high numbers of lysine and arginine residues, trypsin treatment, which results in cleavage after lysine and arginine residues, would be expected to cleave the miniproteins into many small peptides. Many of these peptides would be below the size detection limit, a major limiting factor for protein identification in mass spectrometry studies.

Bacterial genomic analyses show genes with related functions tend to form gene clusters (Tamames et al., 1997). To determine the spatial arrangement of all 68 miniproteins (≤150 amino acids) of unknown function within the T. pallidum proteome, each protein was arranged from the lowest (Tp0004) to the highest (Tp1032) locus tag number and clusters comprised of at least two miniproteins separated by five or less intervening proteins were identified. Forty-three of the 68 miniproteins (63%) were found to be located within one of 17 clusters, with 23 of the 43 proteins located in clusters comprised of at least three miniproteins of unknown function (Supplementary Table 6 and Figure 4A). Twenty of the top 30-ranking predicted AMPs (67%) were found to be located within 13 miniprotein clusters, with 11 of the 20 proteins located in clusters containing at least three miniproteins of unknown function (Supplementary Table 6 and Figure 4B). The top-ranking predicted AMP (Tp_RS02215) was found in a three-miniprotein cluster including the 8^th-ranking predicted AMP (Tp0451a). Interestingly, analysis of the surrounding proteome identified two annotated proteins of note including the outer-membrane inner-leaflet-associated lipoprotein, Tp0453, which contains multiple outer membrane-inserting amphipathic alpha helices that result in membrane bilayer destabilization and enhanced permeability (Hazlett et al., 2005; Luthra et al., 2011). Also included in this region is Tp0454; structure modeling of Tp0454 using Phyre2 predicted tertiary structure similarity to several response regulators (Supplementary Table 6), including the DNA-binding response regulator, PhoP, from the PhoP-PhoQ two-component system that is a central regulator for AMP resistance in Gram-negative bacteria (Brodsky and Gunn, 2005). Proteome functional annotation analyses and Phyre2 structure modeling of open reading frames located close to other putative AMPs identified several additional potential homologs and structural orthologs with potential functions that are consistent with AMP secretion, activation, transport, and self-immunity (Supplementary Table 6), including the ORFs Tp0405 and Tp0688 that have been previously annotated as self-immunity proteins (Fraser et al., 1998). The close spatial arrangement of the miniproteins, in particular the high-ranking predicted AMPs, together with the observed proximity of potential AMP accessory proteins in the proteome of T. pallidum is consistent with the concept that functionally-related genes have a tendency to form clusters within bacterial genomes (Tamames et al., 1997).

Two AMP candidates, Tp0451a (accession number WP_014342798) and Tp0749 (accession number WP_010882194), were selected for the identification of potential AMPCCRs, the important minimalistic functional regions of AMPs. Tp0451a was selected as it (i) is one of the top-ranking predicted AMPs (8/8 positive AMP predictions, mean probability score of 92.6%) (Supplementary Table 3), (ii) possesses classical AMP properties (high content of positively-charged and hydrophobic amino acid residues) (Supplementary Table 4), (iii) is clustered in the proteome with several other potential AMPs/related proteins, as described above, and (iv) tp0451a is expressed at the transcript level, as described below. Although Tp0749 is a lower-ranking predicted AMP (ranked 24/68, 6/8 positive AMP predictions, mean probability score of 78.4%) (Supplementary Table 3), it was selected for further analyses as it (i) is highly positively-charged and contains high hydrophobic content, consistent with pore-forming AMPs (Supplementary Table 4), (ii) was identified in preliminary bioinformatics analyses as having clearly defined potential critical core regions, indicative of future success in AMPCCR design and synthesis, (iii) is known to be expressed at the transcript level (Smajs et al., 2005), unlike several of the higher-ranking predicted AMPs (Supplementary Table 5), (iv) is the second highest expressed ORF at the transcript level in the top-30 ranking T. pallidum miniproteins (Smajs et al., 2005) (Supplementary Table 5), and (v) of particular importance, it is one of only six minproteins within the top-30 ranking predicted AMPs whose expression has been detected at the protein level in experimental rabbit infections (Osbak et al., 2016) (Supplementary Table 5). To date, protein expression of all eight miniproteins from the top-eight ranking predicted AMPs (8/8 positive AMP predictions, mean probability score range 98.9–92.6%) has not been demonstrated in rabbit models of infection (McGill et al., 2010; Osbak et al., 2016), and only three of the top-eight ranking predicted AMPs have been shown to be expressed at the transcript level (Smajs et al., 2005). The strong experimental evidence confirming expression of Tp0749 at both the RNA and protein levels in rabbit infections increased the prioritization of this predicted treponemal AMP over higher-ranking predicted AMPs for further bioinformatics and functional characterization studies.

To confirm expression of tp0451a, we analyzed RNA isolated from T. pallidum by reverse transcription PCR (RT-PCR) using sense and antisense primers. When reverse transcriptase was present (RT+), the primer pair amplified a 198 base pair product, matching a similarly sized amplicon generated from T. pallidum genomic DNA (Supplementary Figure 3 lanes 2 and 4, respectively). In comparison, only a very faint amplicon was detected when reverse transcriptase was omitted from the RT-PCR reaction (RT-) (Supplementary Figure 3 lane 3), indicating that the 198 base pair product from the RT+ reaction was amplified from RNA and not contaminating DNA. Together with the previous finding that showed expression of Tp0749 at the protein level (Osbak et al., 2016), this result allowed us to proceed with investigations into potential AMP activity in T. pallidum by focusing on two miniproteins, Tp0451a and Tp0749, that are known to be expressed at either the transcript or protein level.

The first step for mapping AMPCCRs in Tp0451a and Tp0749 involved using four prediction servers [AMPA (one algorithm), CAMP (three algorithms), AntiBP (three algorithms), and AntiBP2 (one algorithm)] to identify the amino acid boundaries of potential antimicrobial active regions (critical core regions, CCRs) within the two T. pallidum proteins based on clusters of high probability scoring regions predicted by at least three of the four servers. For both Tp0451a and Tp0749, two potential active regions were identified in the N-and C-termini of each protein (Figures 5A,B).

In the second step of our pipeline, secondary structure analyses and structure modeling of the two proteins and their identified potential active regions were performed. For Tp0451a, Jpred 4, and PSIPRED analyses predicted a predominantly coiled/beta strand structure that corresponded to the N-terminal predicted active region and an alpha helical structure that corresponded to the C-terminal predicted active region (Figure 5A). HeliQuest analysis showed that the predicted C-terminal alpha helix exhibited amphipathic properties (Figure 5A), a common structural characteristic in AMPs that is important for membrane integration and pore formation. Structure modeling using a combination of PEP-FOLD 2, Swiss Model, Modeller and Molsoft ICM was unable to generate high confidence models for either the N- or C-terminal regions that correspond to the predicted antimicrobial active regions. However, a robust model was generated for the intervening central region (residues E36-I71) (Figure 5A and Supplementary Figure 4). In agreement with the secondary structure predictions, this central region was modeled as two hydrophobic alpha helices. The structure prediction analyses were consistent with the multi-server AMPCCR mapping predictions by defining potential structural elements, one of which is important for AMP function, that corresponded to high-scoring predicted active regions. Together, these findings allowed for the identification of two potential critical core regions within the N- (Tp0451a_N) and C-terminus (Tp0451a_C) of Tp0451a (Figure 5A).

For Tp0749, both Jpred 4 and PSIPRED analyses predicted alpha helices that corresponded to the N- and C-terminal predicted active regions (Figure 5B). Consistently, a high confidence (92%) alpha helix was modeled for the N-terminal region that was predicted to exhibit antimicrobial activity (Figures 5B, 6A). Importantly for potential AMP function, this modeled N-terminal alpha helix was also shown to be amphipathic with one face of the helix rich in positively-charged residues and the opposing face rich in hydrophobic residues (Figures 6A,B). In addition, a structure-based alignment using PROMALS3D of the Tp0749 N-terminal alpha helix model and a solved structure from the known AMP, human cathelicidin LL-37 (PDB:5NMN), predicted structural similarity between the two peptides (RMSD—0.31 Å over 19 Cα atoms) and conservation of 3/6 positively charged residues involved in binding target cell membrane lipids (Sancho-Vaello et al., 2017) (Figure 6C). In agreement with secondary structure predictions, a lower confidence (69%) partially amphipathic alpha helix model was also modeled for the C-terminal predicted antimicrobial active region (Figures 5B, 6D,E). The combined approach of multi-server AMPCCR mapping, secondary structure prediction, and modeling allowed for the identification of two potential critical core regions located in the N- (Tp0749_N) and C- (Tp0749_C) terminus of Tp0749 (Figure 5B).

To help further resolve the potential active regions identified above, the four candidate AMPCCRs were then analyzed for similarity with known, experimentally-validated AMPs from the APD3 database and for their predicted cell penetrating capabilities using CellPPD. Amino acid sequence-based homology searches identified similarities of each of the four AMPCCRs with established AMPs with homologies ranging between 37 and 44% (Table 2). CellPPD analysis also predicted that three of the four potential AMPCCRs with predicted amphipathic alpha helices (Tp0451a_C, Tp0749_N, and Tp0749_C) (Figures 5A, B, 6) contain peptide stretches that may have the ability to penetrate cell membranes (Table 3), a key functional feature of AMPs. The positive control peptide (LL-37) was also predicted to contain cell penetrating peptide regions, unlike the negative control peptide, Tp0751_p5 (Table 3). Together, these results bolstered the initial AMP predictions generated by the four prediction servers and identified two potential AMPCCRs within each of Tp0451a and Tp0749. These putative functionally-active core regions were prioritized for peptide synthesis and AMP functional characterization studies.

To evaluate the potential antimicrobial activity of the four predicted T. pallidum AMPCCRs identified via our bioinformatics pipeline, synthetic peptides were produced and antimicrobial susceptibility assays were performed to test for bacteriostatic and bactericidal activities against a panel of biologically and clinically relevant Gram-negative and Gram-positive bacteria. In broth microdilution assays, all four candidate AMPCCRs were active against M. smegmatis (Table 4). Consistent with our bioinformatics pipeline analyses, the three T. pallidum peptides with predicted amphipathic alpha helices (Tp0451a_C, Tp0749_N, and Tp0749_C) (Figures 5A,B, 6) and highest AMPCCR mapping scores (Figures 5A,B) all exhibited robust anti-mycobacterial activity, unlike the lower-scoring Tp0451a_N. The lack of activity in the latter peptide aligns with the observation that this was the only peptide that lacked predicted (amphipathic) helical structure (Figure 5A). With the exception of Tp0451a_N, the treponemal peptides exhibited anti-mycobacterial activity that was similar to the positive control AMPs, LL-37 and RaCa-2. Tp0451a_N showed no antimicrobial activity against any of the other tested bacteria, whereas Tp0451a_C showed considerable bacteriostatic and bactericidal potency toward the Gram-positive bacterium, S. pyogenes, to a level that exceeded that observed with the positive control AMPs, LL-37 and RaCa-2. None of the other three treponemal peptides exhibited anti-streptococcal activity and all four treponemal peptides were inactive against S. enterica and S. aureus. Together, these findings demonstrate that each of the four treponemal peptides identified in our AMP discovery bioinformatics pipeline are capable of exhibiting both bacteriostatic and bactericidal activity.

Based on our six-member panel of bacteria, we found Tp0749_C to be the most broad-spectrum AMP of the four treponemal peptides. In addition to its anti-mycobacterial properties, it also showed moderate bacteriostatic activity towards E. coli, moderate-low bacteriostatic activity against P. aeruginosa, and low bactericidal activity against both E. coli and P. aeruginosa (Table 4). The positive control peptide, LL-37, was more active against E. coli and P. aeruginosa than Tp0749_C. The positive control peptide, RaCa-2, was more active against E. coli than Tp0749_C but exhibited lower anti-pseudomonal activity. The other three treponemal peptides showed no activity against these two Gram-negative bacteria. The negative control peptide, Tp0751_p5, a 24-mer peptide from the T. pallidum adhesin Tp0751 with similar physicochemical properties to AMPs, was inactive against all six bacteria in all experiments. Surprisingly, the negative control peptide, sLL-37, a scrambled version of LL-37, exhibited a low degree of antimicrobial activity against M. smegmatis and minimal bacteriostatic activity against P. aeruginosa (Table 4). Since sLL-37 retains many of the same physicochemical properties, including overall charge and amino acid composition, as LL-37, but is predicted to lack the helical content found in native LL-37, this may explain the low-level activity observed with the scrambled version of this peptide. Indeed, a similar low level of antimicrobial activity for sLL-37 has been demonstrated previously in an independent study (Gordon et al., 2005). Together, the Tp0451a and Tp0749 results suggest that T. pallidum is capable of producing AMPs that target Gram-negative bacteria, Gram-positive bacteria, and mycobacteria, and established proof-of-concept for our AMP discovery bioinformatics pipeline.

To evaluate if cysteine residues are important for T. pallidum AMP function, cysteine-to-serine substituted versions of Tp0451a_C (Tp0451a_C_C85S) and Tp0749_C (Tp0749_C_C61S) were synthesized and tested for antimicrobial activity using our panel of six bacteria and broth microdilution assays. These peptides were chosen as they have broad-spectrum AMP activity, and only contain one cysteine residue each. As shown in Table 4, compared to the unmodified, cysteine-containing AMPCCR, Tp0451a_C, the antimicrobial activity of Tp0451a_C85S against M. smegmatis was reduced two- to eight-fold, whereas the cysteine substitution had no effect on anti-streptococcal activity. Interestingly, the cysteine substituted version of Tp0451a_C exhibited moderate antimicrobial activity against E. coli, unlike the unmodified version. Compared to the unmodified, cysteine-containing AMPCCR, Tp0749_C, the bacteriostatic activity of Tp0749_C_C61S against E. coli and P. aeruginosa was reduced two- to four-fold, and reduced eight-fold against M. smegmatis (Table 4). Furthermore, the low bactericidal activity of Tp0749_C against E. coli and P. aeruginosa was abolished in Tp0749_C_C61S and the strong bactericidal activity against M. smegmatis was reduced 16-fold. Consistent with the high abundance of cysteines in the top-ranking predicted AMPs, these findings show that the cysteine residues are important for the AMP activity.

To investigate the potential antimicrobial activity of the T. pallidum AMPCCRs against the frequently co-infecting sexually transmitted pathogen, N. gonorrhoeae, an antimicrobial susceptibility assay based on agar dilution was developed and performed to test for bacteriostatic and bactericidal activities. Although growth was visibly inhibited by Tp0451a_C, none of the four treponemal AMPCCRs completely inhibited Neisseria growth on the agar plates at any of the peptide concentrations. The visible inhibition of growth by Tp0451a_C prompted us to test whether any of the treponemal peptides were bactericidal against N. gonorrhoeae by comparing TVCs of the 3 h-incubated peptide/bacteria mixtures with TVCs from the corresponding positive control growth wells to give the percentage of bacteria killed by the peptides. In agreement with our modified agar dilution results, we found Tp0451a_C to be the only treponemal peptide capable of exhibiting bactericidal activity against N. gonorrhoeae, with Neisseria-killing activity consistently observed at peptide concentrations of 64 μg/mL and higher (Table 5). The loss of anti-Neisseria activity in the cysteine substituted version of Tp0451a_C (Tp0451a_C_C85S) further suggested the importance of cysteine residues in treponemal AMP function. These findings suggest that T. pallidum may express proteins that are capable of killing Neisseria during co-infections involving these two sexually transmitted pathogens.

Treponema pallidum exhibits vigorous motility which can be used as an indication of bacterial viability. To assess the activity of the treponemal peptides against T. pallidum, in vitro-cultured T. pallidum was incubated with each peptide at three different concentrations. Treponeme viability was then determined by counting motile treponemes using darkfield microscopy. As shown in Supplementary Figure 5, only Tp451a_C showed antimicrobial activity against T. pallidum at 64, 16, and 4 μg/ml peptide concentrations, when compared to the negative control peptide, Tp0751_p5. The positive control peptide, LL-37, and Tp0749_N also showed a low level of inhibitory activity against T. pallidum at 64 μg/ml. These findings demonstrate that, similar to other bacteria (Beis and Rebuffat, 2019; Smits et al., 2020), T. pallidum is susceptible to the antimicrobial activity of some of the AMPs it produces, at least when added exogenously.

Based upon the well-established propensity for AMPs to suppress or induce cytokine production and the persistent nature of T. pallidum infection, we assessed potential immunomodulatory functions of the T. pallidum AMPCCRs through examining their capacity to influence cytokine production (IL-1β, IL-6, IL-8, IL-10, MCP-1, and TNF) from a human monocyte/macrophage cell line. Initially, we quantified cytokine production from monocytes (THP-1 cells) stimulated with the treponemal peptides under non-inflammatory conditions. Stimulation with treponemal peptides resulted in the production of several cytokines from monocytes, but the levels of cytokine were low (Supplementary Figure 6). No statistically significant differences in cytokine production between monocytes stimulated with Tp0751_p5 (peptide with no antimicrobial activity in the present study), and the treponemal AMPCCR, Tp0451a_C, was detected. Exposure of each of the peptides to macrophages (differentiated THP-1 cells) failed to induce a significant difference in cytokine expression when compared to the unstimulated control (Supplementary Figure 7). In summary, no treponemal AMPCCR-specific effect on cytokine secretion by undifferentiated monocytes or macrophages was observed in these non-inflammatory immunomodulation assays.

Previously, it has been shown that LL-37 modulates macrophage cytokine production under conditions where macrophages have been highly activated by exposure to the pro-inflammatory cytokine IL-32γ (Choi et al., 2014). IL-32γ also plays key roles in pathogen defense and persistence of chronic infection (Bai et al., 2010; Dos Santos et al., 2017; Li et al., 2018). To assess whether the T. pallidum AMPCCRs identified in the current study also modulate cytokine expression levels under cytokine-induced pro-inflammatory conditions, macrophages were activated with IL-32γ and immediately additionally stimulated with one of our control peptides (LL-37, sLL-37, and Tp0751_p5), or one of the four treponemal AMPCCRs (Tp0451a_N, Tp0451a_C, Tp0749_N, and Tp0749_C), or not exposed to a peptide (IL-32γ only) or IL-32γ (no stimulation). IL-32γ elicited robust production of the chemokine MCP-1 from macrophages compared to non-IL-32γ stimulated cells, demonstrating that the cells were successfully activated, and notably, both LL-37 and the treponemal AMPCCR Tp0451a_C were found to be the only peptides that significantly downregulated (p < 0.0001 for both peptides) IL-32γ-induced expression of MCP-1, in three independent experiments (Figure 7A). Additionally, LL-37 and Tp0451a_C were able to modulate macrophage-production of the chemokine IL-8 following IL-32γ stimulation: IL-32γ stimulation resulted in lower levels of IL-8 release compared to unstimulated cells, and co-stimulation of macrophages with IL-32γ and LL-37 or Tp0451a_C caused significantly higher levels of IL-8 release compared to from macrophages stimulated with IL-32γ alone (p < 0.0001 for both peptides in three independent experiments (Figure 7B). Co-stimulation of macrophages with each of the peptides failed to significantly affect the expression levels of IL-1β, IL-6, TNF, or IL-10 (Supplementary Figure 8). In addition to the antimicrobial activities described above, these findings demonstrate that the T. pallidum AMPCCR Tp0451a_C is also capable of immunomodulatory activities in certain inflammatory contexts.