Thuricin CD, a posttranslationally modified bacteriocin with a narrow spectrum of activity against Clostridium difficile

Initial Screening for Bacteriocin Producers.

Our goal was to isolate a narrow-spectrum bacteriocin effective against C. difficile that would not impact on beneficial Gram-positive flora in the intestine. We hypothesized that spore-forming anaerobic bacteria from the gut may be a likely source of bacteriocins active against a related bacterium like C. difficile. Spores were selected by including an ethanol wash step of human fecal samples to kill vegetative cells before plating. Using this approach, over 30,000 isolates were screened from fecal samples acquired from both healthy and diseased adults. Only one colony that was isolated produced a large zone in a C. difficile overlay culture (Fig. 1A). The colony was purified and grown in brain–heart infusion (BHI) broth, and it was confirmed that a potent antimicrobial compound was produced and secreted into the growth medium. The activity was lost after digestion with Proteinase K, indicating that the antimicrobial substance is proteinaceous (Fig. 1 B and C). The producing strain was identified as a Bacillus based on biochemical tests in the API 50CHB and 20E kits. The strain was further characterized to the species level as B. thuringiensis using species-specific primers based on the gyrB gene that encodes the subunit B protein of DNA gyrase. No product was amplified with the B. cereus- or B. anthracis-specific primers. The possibility that the strain could belong to the anthracis species was excluded by the demonstration of inherent hemolytic activity, for which B. anthracis is generally deficient, and by the absence of the two large plasmids corresponding to the native plasmids pXO1 (181 kb) and pXO2 (96 kb) that harbor the machinery necessary to produce and regulate the anthrax virulence factors (20). The strain, designated B. thuringiensis Dairy Products Centre Culture Collection (DPC) 6431 [Alimentary Pharmabiotic Centre Culture Collection (APC) 20], is unusual in that it produces a nontypical oval-shaped parasporal inclusion, and it was proven to not produce crystals or harbor a wide variety of cry genes. These inclusions would seem indicative of a newly described class of B. thuringiensis strains in which the inclusion is encoded by the slp gene and quickly degrades on cell lysis (21). B. thuringiensis DPC 6431 (APC 20) is deposited with National Collection of Industrial and Marine Bacteria Patent Deposit (accession no. 41490).

Physical Characterization of the Bacteriocin from DPC 6431.

The highest concentration of bacteriocin, designated thuricin CD, was detected during the late log phase and stationary phase of growth. Activity was stable during an extended stationary phase. The pH decreased during the exponential growth phase to ~5.8 from an initial pH of ~7.5; during the stationary phase, the pH rose again to ~6.4 (Fig. S1). Microscopic examination throughout growth did not indicate a correlation between thuricin production and spore formation. Cell-free supernatants of thuricin were active throughout the pH range of 2–9 and heat stable up to 85 °C. There was, however, a reduction in activity at 90 °C and a complete loss of activity at 100 °C after 15 min. Experimental details of the physical characterization of thuricin CD are described in SI Text.

Inhibition Spectrum of B. thuringiensis 6431.

The cell-free supernatant (CFS) of B. thuringiensis 6431 was tested against a range of Gram-positive and Gram-negative bacteria using the well diffusion assay (WDA). Thuricin CD displayed a narrow spectrum of inhibition that was restricted mainly to spore-forming clostridia and bacilli. Within the Clostridium sp. tested, all C. difficile isolates, including the ribotypes commonly associated with C. difficile infection and most importantly, ribotype 027 (NAP1), were sensitive to the culture supernatants of DPC 6431 and exhibited large zones of inhibition. C. tyrobutyricum, C. lithuseburense, C. indolis, and C. perfringens were also inhibited. Among the other Gram-positive pathogens tested, only Listeria monocytogenes was sensitive to thuricin CD. A range of LAB was tested, but only L. fermentum was strongly inhibited by thuricin CD. None of the Bifidobacterium sp. tested were sensitive, including species commonly associated with the gastrointestinal tract. Thuricin CD showed no activity against any Gram-negative organisms tested, including Escherichia coli, Pseudomonas sp., Salmonella sp., and Bacteroides fragilis. The complete inhibition spectrum of B. thuringiensis DPC 6431 is outlined in Table S1.

Characterization and Purification of Thuricin CD.

Thuricin was purified from the CFS using both hydrophobic XAD-16 beads and extraction of the cell pellet with 70% propan-2-ol (pH 2.0) and subsequent purification of the combined fractions using C18 solid-phase extraction columns. The eluted material was further purified by RP-HPLC, resulting in two well-separated peaks at 34 and 41 min that eluted at ~59% and ~64% acetonitrile, respectively. MALDI-TOF MS analysis of the two distinct HPLC peaks revealed molecular masses of 2,763 and 2,861, respectively. These hydrophobic peptides were designated thuricin CD Trn-α (molecular mass = 2,763) and Trn-β (molecular mass = 2,861) (Fig. 1E). Activity was present in both the CFS and propan-2-ol extract of the cell pellet, indicating that thuricin also adheres to the cell surface. Trn-β has activity at higher concentrations but is significantly enhanced by the presence of Trn-α; at low concentrations, both peptides are required for activity (Fig. 1D).

Thuricin CD Activity Against C. difficile, L. monocytogenes, and Probiotic Cultures in Broth Cultures.

Initial experiments determined the concentration of thuricin required to kill C. difficile in liquid culture; 100 arbitrary units (AU) thuricin/mL reduced the viable cells of C. difficile PCR ribotype 027 from ~10⁶/mL to 0 within 3 h. The same concentration of thuricin reduced the cell numbers of L. monocytogenes by 1.5 log cycles, but cell numbers did not decrease further with time. The same concentration of thuricin had no effect on the viability of the commensal strains Lactobacillus casei 338 and B. lactis Bb12 in the same time period (Fig. S2). The addition of thuricin CD to logarithmically growing C. difficile caused a gradual reduction of OD600 nm; this decrease in OD was paralleled with a concomitant release of the intracellular enzyme acetate kinase into the growth medium, indicating a low level of cell lysis. In contrast, there was no increase in the concentration of acetate kinase in the control sample without thuricin (Fig. S3).

The specific activity was determined by calculating the minimum inhibitory concentration (MIC)₅₀ using the purified Trn-α and Trn-β peptides where each of the peptides are titrated against each other, resulting in an isobologram plot. This shows (Fig. 2) that the MIC₅₀ of thuricin Trn-β (0.5 μM) is 10-fold less than Trn-α (5 μM) when present as individual peptides. However, the MIC₅₀ of Trn-β is reduced to 0.05 μM when combined with 0.025 μM Trn-α, indicating that the inhibitory activity of Trn-β is increased ~100-fold when low concentrations of Trn-α are added. These results show that, under these conditions, the two peptides act synergistically in an optimal ratio of 2:1 (Trn-β:Trn-α) (Fig. S4). Procedures for determining the specific activity of thuricin CD are described in SI Text.

Efficacy of Thuricin CD and Metronidazole in a Distal Colon Model.

Treatment of C. difficile 001 with thuricin CD or metronidazole at equimolar concentrations showed that, in an ex vivo distal colon model, thuricin CD compared favorably with metronidazole, a common therapeutic agent used for treatment of CDAD. The effect of thuricin and metronidazole on the viability of C. difficile was significant relative to the control (P < 0.001 for thuricin CD and P < 0.05 for metronidazole) when data were analyzed for repeated measures using the SAS program (Fig. 2).

Genetic Sequencing of the Putative Thuricin CD Operon.

Edman degradation yielded incomplete peptide sequences, and therefore, the identification of the genetic determinants of thuricin CD was resolved by inverse PCR. Approximately 17 kb of contiguous sequence containing 25 putative ORFs were generated from seven overlapping PCR products. The structural genes trnB and trnA were identified as two complete ORFs, each consisting of 150 bp and 144 bp (Fig. 3A). Although a promoter sequence could not be identified immediately upstream of the structural genes, both genes are preceded by canonical ribosomal binding sequences (RBS) embedded within a stretch of 21 identical nucleotides (AAAAATAAGGAGGAATTAATC). A short inverted repeat sequence occurs downstream of trnA, and a hairpin loop structure with typical features of a Rho-independent transcriptional terminator was predicted in this region (ΔG = −4.20 kcal/mol). Such a terminator structure is not an unusual occurrence in bacteriocin gene clusters and may allow for low levels of a longer transcript encompassing downstream genes in a putative thuricin operon (22).

The deduced amino acid sequences for both Trn-β (49 aa) and Trn-α (47 aa) share 45.3% sequence similarity (39.6% identity). Both peptides possess a conserved double-glycine (GG) motif, which is a common feature among bacteriocins that are synthesized as biologically inactive precursor peptides (prepeptides); however, unusually in this case, cleavage occurs between the GG residues rather than after the GG motif. The thuricin CD leaders are highly conserved (sequence similarity is 57.9% with 47.4% identity) and also share similar features with a number of other double glycine leaders in Class I [type-A(II)] and class II bacteriocins (22). The thuricin CD leaders are both followed by 30 amino acid propeptide moieties that share 38.2% sequence similarity (35.3% identity). Interestingly, cysteine residues occur at positions +5, +9, and +13 in both peptides; Cys +5 is flanked by small hydrophobic amino acids Ala and Val, and those at positions +9 and +13 are coupled with variant small hydrophobic residues.

Upstream of the structural genes, two ORFs of 633 and 858 nucleotides were identified and designated trnF and trnG, respectively. A putative δ^A promoter sequence was identified 88 nucleotides upstream from the RBS of trnF with a −10 region of TATAAT and a −35 region of TTGAAA. The putative TrnF (285 aa) and TrnG (210 aa) proteins contain regions with homology to ATP binding cassette (ABC) transport proteins (48% and 28% sequence identity, respectively). The TrnF protein is predicted to function as the C-terminal ATP-binding domain because of the presence of putative Walker motifs A and B. A perfect Walker A consensus motif beginning at Gly₃₃ was identified (GX₂GXGKS/T), and a Walker B site (hhhD; h is any hydrophobic residue) was located in the region of amino acid 152 (23) The TrnG protein is predicted to function as the N-terminal hydrophobic integral membrane domain, including six transmembrane α-helices. Because two copies of each domain are required for functionality, it is expected that both proteins dimerize at the cytoplasmic membrane to form an active translocation complex (24).

Two complete and overlapping ORFs designated trnC (1,521 bp) and trnD (1,245 bp) were identified 10 nucleotides downstream of the structural genes. Homology searches revealed that TrnC and TrnD belong to the radical S-adenosylmethionine (SAM or AdoMet) superfamily of proteins. These enzymes generally generate catalytic radicals through the reductive cleavage of AdoMet by combining SAM and an unusual iron sulfur cluster (4Fe-4S) (25). A conserved single hallmark signature motif (CX₃-C-X₂-C) that coordinates the (4Fe-4S) cluster was identified within the N termini of both putative enzymes. TrnC exhibited sequence similarity to conserved domains of the Fe-S oxidoreductase arylsulfatase regulatory proteins (AslB), and TrnD had sequences similar to the MoaA family of molybdenum cofactor enzymes. Radical SAM enzymes rarely occur in bacteriocin gene clusters, but they are included in those of subtilosin A and propiocin F (26, 27). Sequence alignments indicated that AlbA of subtilosin A exhibits only 19% and 17% identity to TrnC and TrnD, respectively, although AlbA exhibits some features of each individual thuricin CD protein.

Another ORF consisting of 1,194 nucleotides was identified immediately downstream and designated trnE. Homology searches revealed that TrnE belongs to the S41-type superfamily of C-terminal processing peptidases (CtpA) and has a putative intracellular function given the absence of a signal sequence. A perfect 18 nucleotide inverted repeat sequence was identified upstream of the TcdE protein, and a sequence corresponding to a Rho-independent transcriptional terminating hairpin loop was predicted in this region (ΔG = −19.90 kcal/mol).

MS/MS Sequencing and Exact Mass Determination of Trn-α and Trn-β.

The complete sequence of each thuricin CD peptide was determined using infusion MS/MS. The sequences matched with the genetic sequences of the peptides at all positions other than residues 21, 25, and 28 of both peptides (Fig. 3B). At these positions, the MS/MS spectra indicated that the residues located at these positions in the mature peptides were two mass units lighter than expected. To confirm our findings, we determined the exact mass of Trn-α and Trn-β using MALDI Fourier transform ion cyclotron resonance MS. The exact mass of Trn-α was found to be [M + H⁺] = 2,763.3218 (−2.4 ppm error), enabling us to predict a molecular formula of C₁₁₈H₁₉₂N₃₁O₃₉S₃ for the full-length peptide. Similarly, the exact mass of Trn-β was found to be [M + H⁺] = 2,861.2698 (1.2 ppm error), giving rise to a predicted molecular formula of C₁₂₈H₁₈₆N₃₁O₃₈S₃ (Figs. S5 and S6). Notably, both masses are consistent with the loss of 6 mass units from each peptide. Analysis of in-source fragments from MALDI FT-ICR MS confirmed the sequences determined by infusion MS/MS. In particular, analysis and comparison of the b₂₀, b₂₄, and b₂₇ ions of Trn-α and Trn-β (Table S2) confirmed that residues 21, 25, and 28 of each peptide have lost two hydrogens from the predicted parent amino acid residue.

NMR Analysis of Trn-α and Trn-β.

Carbon, nitrogen, and proton assignments for both peptides were determined using a suite of multidimensional NMR experiments. The chemical shifts of the backbone α-carbons and β-carbons were found using the HNCACB and CBCACONH experiments. It was immediately noted that the α-carbons of residues 21, 25, and 28 of both peptides have chemical shifts (Table S3) that are 15 ppm downfield of average values expected for unmodified residues (28). These findings are consistent with an interpretation that an electronegative heteroatom such as sulfur is bonded to the α-carbon. Analysis of the ¹³C HCCH-TOCSY and ¹³C HSQC experiments confirmed that these α-carbons are fully substituted, having no attached α-protons. At the same time, the β-carbons of residues 21, 25, and 28 were found to have a normal substitution pattern with β-proton shifts (Table S3) close to expected values for unmodified residues (28). Long-range ¹H-¹H NOEs observed in the ¹⁵N HSQC-NOESY and ¹³C HSQC-NOESY experiments are also consistent with a structure in which the sulfur atoms of cysteine side chains form linkages to the α-carbons of these modified residues. For Trn-α, it was found that the α- and β-protons of Cys13 show NOEs to the β-protons of Ser21. Similarly, the β-protons of Cys9 show NOEs to the γ-protons of Thr25. In addition, the amide proton of Thr28 shows NOEs to the β-protons of Cys5. This evidence shows the presence of sulfur to α-carbon linkage between Ser21 and Cys13, Thr25 and Cys9, and Thr28 and Cys5 in Trn-α (Fig. 4). For Trn-β, the β-proton of Thr21 shows an NOE to the β-protons of Cys13, whereas the α-proton of Cys13 shows an NOE to the amide proton of Thr21. NOEs are also observed between the β-protons of Ala25 and the β-protons of Cys9. Likewise, NOEs appear between the β-protons of Cys5 and the β-protons of Tyr28. Taken together, the NOE data reflect the presence of sulfur to α-carbon thioether linkages between Thr21 and Cys13, Ala25 and Cys9, and Tyr28 and Cys5 in Trn-β (Fig. 4).

Bacterial Strains Used.

C. difficile ATCC 43593 was used to isolate antimicrobial agent producing bacteria from mixed populations in fecal samples. C. difficile R1/NAP1/027, Lactobacillus casei 338, Bifidobacterium lactis Bb12, and Listeria monocytogenes NCTC 5348 were used for bacteriocin sensitivities in time-kill studies. C. difficile ATCC 43593 and Bacillus firmus LMG 7125^t were used for determination of activity of thuricin by well diffusion assay. A full list of the target organisms and their sources that were used for determination of the spectrum of inhibition of the bacteriocin-producing cultures is outlined, together with the media and growth conditions, in Table S1. B. cereus NCIMB 700577 and B. thuringiensis NCIMB 701157 were used as positive controls for the PCR using gyrB primers.

Isolation of Bacteriocin-Producing Cultures.

Fecal samples from both diseased and healthy individuals were received in the laboratory and frozen at −80 °C. On the day of analysis, samples were thawed at room temperature, mixed with equal volumes of absolute ethanol, and allowed to stand at room temperature for ~30 min. Samples were subsequently serially diluted in anaerobic diluent, 100 μL spread on the surface of Wilkens Chargrin anaerobic agar (WCAA), and grown for 5 days at 37 °C in an anaerobic chamber. Colonies that developed were overlaid with ~10 mL of reinforced Clostridium agar (RCA) inoculated at 1.25% with a log-phase culture of C. difficile. The plates were incubated for another 18 h and inspected for zones of inhibition of the overlaid culture. Colonies showing a clear zone of inhibition were subcultured onto fresh WCAA having first removed the agar overlay using a sterile scalpel blade. Approximately 30,000 colonies were screened, and one colony showing potent antimicrobial activity against the overlaid C. difficile strain was purified and stocked at −80 °C on Microbank Beads and designated as DPC6431 (APC 20); the inhibitory substance produced was designated thuricin CD.

Bacterial Speciation.

Full experimental details of bacterial speciation and analysis of parasporal body/crystal information are outlined in SI Text.

Production of Thuricin CD from CFS.

Experimental procedures for preparation of CFS and determination of the spectrum of inhibition are outlined in SI Text.

Purification and Molecular Mass Determination of Thuricin CD.

Thuricin CD was prepared as described for lacticin 3147 with minor modifications (18). Full experimental details are given in SI Text.

Effect of Thuricin CD and Metronidazole on the Viability of C. difficile 001 in the Distal Colon Model.

Full experimental detail is given in SI Text.

Determination of N-Terminal Amino Acid Sequence of Biologically Active Peptides.

N-terminal amino acid determination of biologically active fractions was carried out by Edman degradation at Aberdeen Proteome Facility, University of Aberdeen, Aberdeen, Scotland.

Genetic Sequencing of Thuricin CD Operon.

Full experimental procedures for the characterization of the putative thuricin operon by chromosome walking and the sequence analysis and DNA hybridization methods are given in SI Text.

MS/MS Sequencing and Exact Mass Determination of Trn-α and Trn-β.

Full experimental details of MS/MS sequencing and exact mass determination of peptides are outlined in SI Text.

Production and Purification of [¹³C, ¹⁵N]Trn-α and [¹³C, ¹⁵N]Trn-β.

Full experimental details of production and purification of [¹³C, ¹⁵N]Trn-α and [¹³C, ¹⁵N]Trn-β are outlined in SI Text.

NMR Spectroscopy of [¹³C, ¹⁵N]Trn-α and [¹³C, ¹⁵N]Trn-β.

Full experimental details of NMR spectroscopy of [¹³C, ¹⁵N]Trn-α and [¹³C, ¹⁵N]Trn-β are outlined in SI Text.