Strains used in this study are listed in Supplementary Table S1. E. coli and Campylobacter strains were grown as previously described (Kienesberger et al., 2007). Antibiotics were added to final concentrations of 100 μg ml^-1 ampicillin, 75 μg ml^-1 nalidixic acid, and either 12.5 or 25 μg ml^-1 chloramphenicol, 40 or 25 μg ml^-1 kanamycin for E. coli or Campylobacter cultivation, respectively.

Plasmids and oligonucleotides used in this study are listed in Supplementary Tables S1, S2. For expression in E. coli, genes of interest were amplified with PCR and the fragments were ligated to pBAD24 vector derivatives with distinct antibiotic resistance genes (see Supplementary Tables S1, S2).

For 3D structure prediction amino acid sequences (CDF65254.1, CDF65253.1, CDF65920.1, and CDF65967.1) were analyzed using the Phyre2 web portal (Rollins and Colwell, 1986). The output files were rendered with PyMOL (Schrodinger, 2010). Templates for Fic protein fold recognition via Phyre2 are listed in Supplementary Table S3.

Escherichia coli DH5α harboring pBAD plasmids with fic genes were grown with shaking overnight at 37°C in LB-broth supplemented with appropriate antibiotics and 0.2% glucose to repress expression, or 0.05% arabinose to induce expression via the P_BAD promoter. Bacterial growth was either monitored by survival plating of bacteria grown in 100 ml LB-broth or in 24-well plates with a culture volume of 1 ml per well and starting optical density measured at 600 nm (OD₆₀₀) of 0.05. Plates were incubated at 37°C under shaking at 180 rpm. OD₆₀₀ was measured hourly in triplicate. Determination of colony forming units (CFUs) normally corresponded with OD₆₀₀ measurements except at late time points where the CFU count of filamentous cells remained low.

Cultures of E. coli DH5alpha harboring pBAD derivatives with a starting OD₆₀₀ of 0.05 were grown in LB-broth with 0.05% arabinose for 2 h. Cells were harvested, suspended in 1x phosphate buffered saline (PBS), pH 7.4 and incubated with Nile red for up to 60 min at room temperature in the dark. For immediate microscopy, 1 μl of the pellet was applied to an agar slide (1% agar solution poured on microscopy slide) to immobilize the cells. For later imaging, cells were fixed with 0.4% formaldehyde, before collected by centrifugation, resuspended in 1x PBS, pH 7.4 and stained as described before. Confocal microscopy was performed on a LEICA AOBS SP2 MP microscope (380 nm extinction, 510 nm emission).

Pairs of FLAG-tagged and hemagglutinin (HA)-tagged proteins were co-expressed in E. coli C41(DE3) from respective plasmids (Supplementary Table S1). 100 ml LB broth supplemented with 0.2% glucose was inoculated with overnight cultures to OD₆₀₀ of 0.1. When cultures reached OD₆₀₀ = 0.5–0.8 protein expression was induced with 0.05% arabinose for 2 h. Thirty OD of cells were pelleted and washed with 50 ml buffer A (50 mM Tris-HCl pH 6.8, 100 mM NaCl). Cell lysis was performed as previously described (Gruber et al., 2016) except that the formaldehyde crosslinking step was omitted for this study. All further steps were as in Gruber et al. (2016). For protein detection, OD₆₀₀ 0.015–0.05 equivalents of lysate and pull-down fractions were mixed with sample buffer containing DTT (0.09%) and SDS (0.1%), heated at 95°C for 10 min and resolved on SDS-PAGE (12.5%, Hoefer) or NUPAGE (12%, MES buffer, Invitrogen) gels. Proteins were transferred for 1.5 h onto PVDF membranes. Blocking was done overnight at 4°C in TST: (0.5 M Tris-HCl pH 7.5, 1.5 M NaCl, 1% Tween-20) supplemented with 3% milk powder. FLAG-tagged proteins were detected with HRP-conjugated α-FLAG antibody (A8592, Sigma) and HA-tagged proteins with HRP-conjugated α-HA antibody (12013819001, Roche). After washing (3 min × 10 min) with 1x TST blots were developed with ECL (Bio-Rad) according to the manufacturer’s instructions (1:5 dilution in ddH₂O of substrate for FLAG detection).

Escherichia coli DH5α with plasmids were grown in 100 ml LB-broth to an OD₆₀₀ of 0.4 to 0.6, then shifted to medium containing 0.05% arabinose for 1 h to induce expression. 30 s prior to cell harvest, chloramphenicol was added to a final concentration of 200 or 300 μg ml^-1. Cells were harvested by centrifugation for 10 min at 14,300 × g and 4°C. The cell pellet was resuspended in 500 μl cell lysis buffer (10 mM Tris-HCl pH 7.5, 10 mM MgCl₂, 30 mM NH₄Cl, 100 or 150 μg ml^-1 chloramphenicol) and either mixed with an equal volume of glass beads (300 μm in diameter) and vortexed for 5 min at 4°C or immediately frozen in liquid nitrogen (Bronowski et al., 2014). Suspensions with glass beads were centrifuged for 10 min at 6,400 × g at 4°C. The supernatant was collected, centrifuged for 3 min at 17,649 × g at 4°C and immediately applied to sucrose density centrifugation or stored at -70°C. Frozen suspensions were thawed in an ice bath, frozen again in liquid nitrogen and stored at -70°C.

The protocol for the sucrose density centrifugation was adapted from Jiang et al. (2007). An A₂₆₀ of 8 of the cleared cell lysate was loaded onto a gradient of 5–45% sucrose in buffer (10 mM Tris-HCl pH 7.5, 10 mM MgCl₂, 100 mM NH₄Cl). Ultracentrifugation was carried out for 4 h at 4°C and 253,483 × g in a Beckman SW-41Ti rotor. Fractions of the gradient were collected using an UA-6 system (Teledyne ISCO) with continuous monitoring at A₂₅₄.

Ribosome profiles were scanned, traced in CorelDraw to increase contrast and xy-coordinates were extracted using DataThief III (Pascoe et al., 2015). To increase the reliability of calculations, in addition to peak values we calculated the Xgrad-values using a code (XSpan) written in VisualBasic. The code is available to interested readers upon request. The program ‘XSpan’ places the largest rectangular surface with a predefined width (= X) under individual peaks of a given curve and then calculates the Xgrad-values, which are height (H) and area (A) of the surfaces. XSpan can also extrapolate clipped curves (e.g., when a maximum value exceeds the measurement range) by fitting a cubic function such that it tangents the two flanks of a clipped peak. This option was utilized in this study to obtain the 70S heights (H). The Xgrad-value H is the highest amplitude (absorbance) measured for the surface of predefined width describing fractions of the analyzed sedimentation gradient. X, the width of the rectangular surfaces, was selected such that it covers approximately 1.2% of the gradient fractions analyzed. Statistical analysis was performed using the H values only. For each profile, the H values of the 30S and 50S subunit peaks were normalized to the H values of the first polysome peak and statistical significance was calculated using the paired Student’s t-test. Statistical significance was assumed with p-values below 0.05.

Prevalence of TA genes was surveyed among C. fetus isolates via PCR using chromosomal DNA as template. We applied primer pairs 1/2 for fic1, 26/27 for fic2, 5/6 for fic3, 7/8 for fic4, and 9/10 for fti3 (Supplementary Table S2). Sequencing of fic2 amplicons from C. fetus subsp. venerealis strains V9, V20, V32, V60, V62, and V69 was performed with primers 28/29.

The conserved Fido motif sequence HPFXXGNXR and full length Fic1-4 of C. fetus subsp. venerealis 84-112 were used in BlastP analysis to identify Fido proteins in whole genomes of Campylobacter species (if possible, finished whole genomes, if not available, genomes with low scaffold numbers were used). BlastP analysis was also performed with full-length Fics and the Fic2 specific motif sequence [HPFREGNTRTIA] under exclusion of epsilon-proteobacteria to screen for hits outside this class. Selected Campylobacter proteins, selected Fic reference proteins, as well as other bacteria from the urogenital tract identified by the BlastP were then used to generate the phylogeny tree. Retrieved proteins were aligned with MEGA6.06 using the BLOSUM matrix. The Neighbor joining tree was constructed with MEGA6.06 (Tamura et al., 2013). The tree was rooted to the translated ORF of housekeeping gene glnA of C. fetus subsp. venerealis 84-112. Protein accession numbers are listed in Supplementary Table S4.

fic1 and fic2 genes (Figure 1A) are chromosomally encoded and form part of a pathogenicity island (PAI) that harbors additionally a functional T4SS (Gorkiewicz et al., 2010). C. fetus subsp. venerealis 84-112 also carries extra-chromosomal DNA with features of an integrative conjugative element (ICE_84-112) (Kienesberger et al., 2014). Two additional fic gene homologs, fic3 and fic4, were identified on the ICE (Figure 1B). Residues of the Fido superfamily core motif that enable FIC-containing enzymes to act as AMP transferases have been defined as HxFx(D/E)GNGRxxR (Kinch et al., 2009; Worby et al., 2009; Xiao et al., 2010; Engel et al., 2012). Fic1, Fic3, and Fic4 contain the complete signature of invariant residues [protein accession numbers CDF65254.1 (Fic1); CDF65920.1 (Fic3); CDF65967.1 (Fic4)]. In contrast, in Fic2 (CDF65253.1), the second conserved glycine at position 191 is replaced with threonine and the final arginine of the signature motif (R195A) is absent, suggesting that Fic2 does not have adenylylation activity. Fic1 and Fic4 also contain a conserved inhibitory motif (S/T)xxxE(G/N), which was shown to suppress adenylylation in well-studied systems (Harms et al., 2016b) (Figure 1C). Fic proteins containing this inhibitory helix (inh) are classified depending on whether the inh is part of the FIC fold as an N-terminal helix (class II) or a C-terminal helix (class III) (Engel et al., 2012). Fic1 thus belongs to class II, and Fic4 to class III. Class I Fic proteins do not contain the inhibitory motif themselves, but have an interaction partner that provides the inh in trans. Fic2 and Fic3 lack a motif with this overall consensus, thus they belong to class I. Fic1 may act as antitoxin for the degenerated toxin Fic2. Alternatively, mutations of the core motif in Fic2 may have altered enzyme activity and thus have bypassed the need for an inh motif. We also note that the 78 amino acid ORF (protein accession number CDF65919.1) upstream and partially overlapping fic3 includes residues GHAIEN, which might provide the invariable glutamate (Engel et al., 2012; Goepfert et al., 2013) of a poorly conserved inhibitory motif (Figure 1B). Presence of a 48% identical homolog, fti4 (protein accession number CDF65966.1), upstream and partially overlapping fic4 strengthens the hypothesis that each ORF encodes a small interacting protein to control the cognate FIC enzyme.

Alignment of the predicted proteins shows strong conservation of the Fic core motif but low general similarity (not shown). A structure prediction was performed with the Phyre2 server using templates listed in Supplementary Table S4. All C. fetus subsp. venerealis homologs are predicted to share a similar FIC domain fold (Figures 2A–D). The set of common α-helices are colored from blue (N-terminal) to red (C-terminal), according to the core FIC domain secondary structure topology (Kinch et al., 2009). The predicted active site loops with the conserved core motif including the catalytic histidine are highlighted in black. Fic proteins typically carry a β hairpin close to the active site (gray). This structure, also called “the flap,” constitutes the major target-protein docking site (Kinch et al., 2009; Xiao et al., 2010; Palanivelu et al., 2011; Garcia-Pino et al., 2014). The inhibitory motifs of Fic1 and Fic4, expected to prevent the adenylylation reaction by active site obstruction, are shown in pink. The remaining protein structure outside of each FIC core domain is shown in white. One additional shared feature we noted is the conserved KEKE motif (asterisks) at the C-termini of Fic1 and Fic2 that is reiterated in Fic4, once internally, and again at the C-terminus. Bacterial effector proteins secreted via a given T4SS typically display a short C-terminal stretch of conserved residues that mediates their specific recognition by the transfer machinery (Zechner et al., 2012; Christie et al., 2014). The conserved KEKE motif may represent such a dedicated translocation signal, but this has not been validated experimentally.

To gain insights to the function of the C. fetus proteins we expressed these in a heterologous bacterial host and asked whether the fic1-fic2 module acts as a TA system. In that case, the inhibitory domain of Fic1 would be required to act both intra- and intermolecularly to regulate the enzymatic activity of Fic1 and Fic2. The fic genes of C. fetus were placed under transcriptional control of the P_BAD promoter and their effects on growth of E. coli were investigated. Shifting E. coli cells from LB broth with glucose to medium containing arabinose induced synthesis of the Fic proteins and culture density was monitored over time. Induction of fic2 expression delayed growth of E. coli severely compared to the vector control, demonstrating that Fic2 is toxic despite its degenerate core motif (Figure 3A). Exchange of the catalytic histidine in variant Fic2_H184A eliminated toxicity and allowed the host to grow comparably to the vector control strain. Loss of phenotype could occur either because the histidine is indeed important to the activity of the enzyme, as predicted, or because the mutant variant is unstable. To exclude the latter possibility we purified the mutated protein and verified its stability during overexpression in E. coli and in isolated form (not shown). E. coli expressing fic1 displayed logarithmic growth, but culture densities obtained after 8 h were lower than cells carrying the empty vector. To examine the role of the conserved inh of Fic1, key residues Ser31 and Glu35 were exchanged for alanine. The substitution apparently disrupted the protective function of this motif, since expression of Fic1_S31A/E35A was incompatible with cell growth. Cells were rescued from Fic2-induced growth arrest by co-expression of wild type Fic1, suggesting that Fic1 can act as an antitoxin for Fic2. The importance of the inh module in toxin neutralization was again shown when co-expression of Fic1_S31A/E35A and Fic2 arrested growth fully. The data imply that Fic1 catalyzes an activity detrimental to bacterial growth, but which is normally blocked intramolecularly by the protein’s inh helix. Moreover, the bacterial cytotoxicity of Fic2 depends on the enzyme core motif and is neutralized by antitoxin Fic1.

To characterize the proposed toxin–antitoxin activities, we next compared the impact of Fic protein production on cellular morphology. Fic2 alone caused an extreme filamentous phenotype (Figure 3B) via a mechanism requiring the catalytic histidine since; by comparison, cells expressing Fic2H184A were similar to wild type. Cells expressing Fic1 appeared normal but formed filaments when the Fic1 inh motif was mutated. Coexpression of wild type Fic1 reversed the filamentous phenotype caused by Fic2 consistent with the neutralization observed during growth (Figure 3A).

Antitoxins similar to the PhD-Doc paradigm frequently inactivate the toxin by forming a stable complex. We asked whether inactivation of Fic2 toxicity by Fic1 might involve binding of the two proteins. Codons for a FLAG epitope were added to fic1 and the hemagglutinin (HA) tag was added to fic2. Lysates of E. coli cells expressing both fusion proteins were incubated with FLAG-affinity beads. After elution of bound proteins, lysates and eluates were analyzed by western immunoblotting (Figure 3C). Anti-FLAG antibodies confirmed the presence of FLAG-tagged Fic1 in cell lysates and the absence of signal in control samples expressing native Fic1. Antibody to HA detected Fic2-HA fusion protein in the same cell lysates. HA signal in the pull down fraction indicated retention of Fic2 by Fic1. The specificity of this interaction was confirmed by the absence of signal when partner protein Fic1 lacked the FLAG epitope. These properties indicate that Fic1 and Fic2 of C. fetus subsp. venerealis 84-112 form a functional toxin–antitoxin system. Given that the enzymatic activity of antitoxin Fic1 is autoregulated via inh, this protein exhibits a mode of concomitant intra- and intermolecular- toxin neutralization novel for bacterial Fic proteins.

Fic3, like Fic2, carries the enzyme core motif but lacks an inh motif (Figure 1). Similar to the result of fic2 expression, E. coli carrying fic3 failed to grow under inducing conditions (Figure 4A). However, dual expression of fic3 and its neighboring gene encoding the putative inhibitor protein restored E. coli growth completely. E. coli cells expressing the inhibitor alone grew indistinguishably from cells carrying the vector control. Since, this protein acts as an antitoxin for Fic3 we named the gene fti3 (Fic toxin inhibitor 3). Microscopy of the toxin/antitoxin expressing E. coli revealed that inhibitor protein alone had no impact on cell morphology (Figure 4B). By contrast we observed extreme filamentation due to Fic3 that could be reversed by either mutation of the catalytic histidine in variant Fic3H147A, or co-expression of wild type Fic3 with Fti3. To assess the viability of cells expressing Fic3, samples of a culture before and after 4 or 8 h of arabinose-induced expression were plated on LB agar without arabinose. Viability of the culture dropped by several orders of magnitude after Fic3 induction (Figure 4C). By contrast cells producing the non-toxic variant Fic3H147A exhibited similar viability as the vector control strain. Direct interaction between Fti3 and toxin Fic3 was tested following coexpression of FLAG-tagged Fti3 and HA-tagged Fic3. FLAG-tagged Fti3 from the cell lysate bound the affinity matrix and specifically retained Fic3-HA in the pull down reaction (Figure 4D). No retention of Fic3 was detected when Fti3 lacked the FLAG epitope. We conclude that Fti3-Fic3 form another TA module on the extrachromosomal ICE in addition to the chromosomal system fic1-fic2.

Expression of fic4 for up to 8 h had no effect on cell growth (Figures 5A,C) and affected cell morphology only mildly (Figure 5B). To test whether the protein’s inhibitory motif was suppressing the predicted enzyme activity, residues Thr209 and Glu213 were exchanged for alanine. Expression of the mutant variant was compatible with normal growth comparable to cells expressing wild type Fic4 or the vector control (Figure 5A) but a filamentous phenotype was observed upon Fic4_T209A/E213A-HA expression (Figure 5B). Since cells expressing wild type Fic4 are phenotypically normal under these conditions we used the mutant variant to test for a possible antitoxin activity for the adjacent ORF, Fti4. Coexpression of Fti4 and mutant Fic4 lessened filamentation substantially (Figure 5B) but had no impact on growth or survival (Figures 5A,C).

To test whether Fti4 and Fic4 physically interact, fusion proteins with epitope tags were created and simultaneously produced in E. coli as described above. FLAG-Fti4 (∼10 kDa) was not directly detectable in the cell lysates but was visible after enrichment of the protein on the affinity matrix (Figure 5D). Fic4-HA was detected in lysates of both test and control strains. Fic4-HA was also retained on the FLAG affinity beads in a manner dependent on FLAG-Fti4. Since the functional tests described above showed phenotypes for Fti4 only when combined with the mutant derivative of Fic4, we also assayed for protein binding using the mutant allele. Similar to wild-type Fic4, co-retention of Fic4_T209A/E213A by FLAG-Fti4 was observed (Figure 5D, right panel).

In summary, some of the observed characteristics of Fic4 are consistent with the function of a toxin, yet the toxicity of the mutant form was quite mild compared to the inh-deficient Fic1 derivative and the wild type class I proteins Fic2 and Fic3. It is possible that evolution has introduced mutations outside of the Fic4 active site that impair enzyme activity. It is further possible that the surrogate host E. coli simply lacks the specific protein targeted by Fic4. Another hypothesis that we could test was to ask whether Fic4 might actually function as an antitoxin for a distinct locus (below).

To explore potential in trans interactions involving components of the distinct systems each toxin was expressed pairwise with every putative antitoxin. We found that the ICE_84-112 encoded antitoxin Fti3 reversed the growth defect caused by toxin Fic2 (Figure 6A). In contrast co-expression of Fic4 with Fic2 had no effect. Neutralization of Fic2 toxicity by Fti3 was confirmed by plating samples of the induced cultures (Figure 6B). Cells survived 4 and 8 h of Fic2 expression when co-expressing Fti3, but Fic4 was not able to counteract the toxicity of Fic2. Functional interaction between Fic2 and Fti3 was also supported by the normalized morphology of cells following co-expression compared to the filamentous phenotype caused by Fic2 alone (Figure 6C). A partial reversal of the Fic2-induced filamentation was observed with Fic4.

To test for direct binding interactions between the protein pairs, pull-down assays were performed with cells expressing Fic2-HA with either FLAG-tagged Fti3 or FLAG-tagged Fic4. Consistent with the functional results shown above, Fti3 was able to retain Fic2 (Figure 6D). Remarkably, although Fic4 showed little antitoxin activity for Fic2, a complex of these proteins was detected nonetheless (Figure 6E).

To complete the analyses for toxin Fic2, the same tests were performed for the last putative antitoxin Fti4. No neutralizing activity by Fti4 was observed during cellular growth or by monitoring cell morphology. The pull down assay combining FLAG-tagged Fti4 with Fic2-HA also failed to detect interaction between these proteins (all data not shown). Given that Fti4 is 48% identical to Fti3, the lack of activity observed for Fti4 shows that the ability of antitoxin Fti3 to inactivate Fic2 is specific. As a final specificity check we also tested whether the toxic form of Fic1 (Fic1_S31A/E35A) was affected by co-production of either Fti3 or Fti4. No reversion of the poor growth, reduced survival or filamentous phenotypes caused by Fic1_S31A/E35A was observed (data not shown).

The sum of these data demonstrate that the bacterial growth phenotype caused by Fic2 is counteracted by the cis encoded antitoxin, Fic1, and independently by Fti3 in trans. Fic4 partially reversed the toxic effect of Fic2 in E. coli. Both the cis acting antitoxin Fic1 and the ICE-encoded proteins Fti3 and Fic4 were shown to bind toxin Fic2. These findings support a model of functional crosstalk occurring between chromosomally and ICE_84-112 encoded Fic proteins that act to control the toxin Fic2.

To test for potential regulatory crosstalk occurring between Fic3 and antitoxins of the distinct systems, we again performed phenotypic tests following dual expression of each protein pair. Both Fic1 and Fic4 were unable to neutralize the extreme growth phenotype caused by Fic3 (Figure 7A). Dual expression of Fic3 with Fic1 did not revert the filamentation induced by Fic3, but partial recovery was apparent upon co-expression of Fic3 with Fic4, suggesting some neutralizing interactions (Figure 7B). Consistent with these results the pull-down assay was clearly negative for binding between Fic1 and Fic3 (Figure 7C), but a small yield of co-purified Fic3 was detected using FLAG-tagged Fic4 (Figure 7D). We performed the same analyses with cells coexpressing Fti4 with Fic3. Again despite its similarity to antitoxin Fti3, Fti4 had no affect on Fic3 toxicity and the proteins failed to bind under these conditions (not shown). In summary, we conclude that Fic3 is effectively neutralized by the cis encoded Fti3. Moreover, modest levels of complex formation with trans-acting factor Fic4 may contribute to regulation of this enzyme.

The identity of specific protein targets modified by Fic enzymes in bacteria is difficult to predict. It is known, however, that the activities of many TA toxins interfere with the translation process either directly, e.g., by cleavage of mRNA or tRNA, or as a downstream effect (Rajashekara et al., 2009; Park et al., 2013). To measure translation in E. coli cells expressing C. fetus Fic proteins, we performed sucrose gradient centrifugation of cell lysates and recorded polysome profiles (Figures 8A,B). In these analyses, the height of the polysome peaks is directly proportional to the translation levels, therefore translation defects can be faithfully detected by the reduction of polysome peaks. Moreover, because free ribosomal subunits, 70S monosomes and polysomes can be resolved, changes in the ratios between these different ribosomal (sub-) complexes can give additional information on the type of defect causing reduced translation. We recorded profiles from fic-expressing E. coli cells and compared them to those of the vector control strain. The signal corresponding to ribosomal subunits and translating ribosomes is indicated for each gradient. To make fic-dependent shifts in the relative abundance of these populations more apparent, profiles from different expressing strains were overlaid in the figure. Expression of Fic3 inhibited translation severely, as obvious from the massive reduction of polysome levels (Figure 8A, red trace; Table 1) compared to the vector control strain (black trace) or cells expressing just antitoxin Fti3 (green trace). Concomitantly, a strong increase of the 70S peak was detected, suggesting that ribosomal subunits are competent for joining into 70S ribosomes, but fail to enter into translation. We conclude that Fic3 blocks a step after subunit joining but before translation elongation. Dual expression of Fic3 and antitoxin Fti3 (blue trace) largely restored translation to normal levels. Moreover, the abnormally high 70S peak observed upon Fic3 overexpression was partially reduced upon Fti3 co-expression.

Fic2 expression also had a mild inhibitory effect on translation, as obvious from an accumulation of free 30S and 50S ribosomal subunits relative to the amount of 70S ribosomes and polysomes (Figure 8B). We compared the free subunit accumulation relative to the polysome abundance in multiple independent experiments (n = 7) and determined a quantitatively significant increase (Table 1). Notably, in contrast to Fic3 expression, Fic2 expression did not result in increased 70S levels, and even reduced 70S amounts compared to the vector control. Reduced 70S and increased free subunit levels are indicative of inefficient subunit joining. E. coli cells expressing antitoxin Fic1 (green trace) showed no significant variation in the ratio of free subunits versus polysomes compared to profiles from the vector control strain (Figure 8B and Table 1). The relative abundance of 70S species was mildly reduced in fic1-expressing vs. vector control cells however (Table 1), consistent with the observation that Fic1 expression slightly inhibits cell growth (Figure 3). In line with the neutralizing activity observed in our previous functional tests, simultaneous expression of antitoxin Fic1 with Fic2 (blue trace) resulted in a profile similar to the empty vector control.

The sum of our findings suggests that the C. fetus fic genes act as TA systems. In that case the loci should be well-conserved within the species. We used PCR to survey the prevalence of the fic genes and fti3 in 102 C. fetus isolates from geographically and ecologically diverse sources (summary in Table 2; detailed information in Supplementary Table S5). All of the C. fetus subsp. venerealis strains (n = 62) were positive for fic1 and 59 out of 62 (95%) were positive for fic2. This finding is consistent with genetic linkage of the Fic2 toxin to the Fic1 antitoxin. Sequence analysis of full-length fic2 amplicons randomly selected from our strain collection (n = 6) showed complete conservation for this subspecies (data not shown). In contrast, only 5 out of 40 (12.5%) C. fetus subsp. fetus strains harbor fic1 and only two carry the fic2 gene, whereby strain C. fetus subsp. fetus 98/v445 (F37) lacks the corresponding fic1 antitoxin gene. Sequence analysis of this solitary fic2 allele revealed 36 nucleotide changes, corresponding to 13 amino acid substitutions. Expression of the F37 fic2 gene in E. coli confirmed that the mutated toxin is functionally impaired (data not shown). The fic3 and fic4 genes were detected exclusively in C. fetus subsp. venerealis, fic3 in 11.3% (7/62) and fic4 in 4.8% (3/62) of the isolates. Gene fti3 shows higher abundance: 59.7% (37 of 62) C. fetus subsp. venerealis isolates and four C. fetus subsp. fetus isolates (4/40) carry the gene. Consistent with the predicted selective pressure for co-existence, all strains positive for fic3 additionally encode the corresponding antitoxin Fti3. Moreover the high prevalence of fic2 in C. fetus subsp. venerealis may select for stable maintenance of fti3 even in the absence of the cognate toxin fic3.

In summary, we conclude that the presence of a fic toxin gene in C. fetus is typically linked to carriage of the paired antitoxin gene. The chromosomal TA system is highly conserved in C. fetus subsp. venerealis. The ICE-associated loci are also unique for C. fetus subsp. venerealis but are comparatively rare in the strains surveyed.

The significant association of these fic genes with C. fetus subsp. venerealis and their relative absence in C. fetus subsp. fetus led us to next ask whether they are present in other Campylobacter species and/or whether they are conserved in bacteria which inhabit the urogenital tract. Using the HPFXXGNXR motif in a BlastP analysis revealed that genomes of several Campylobacter species encode from one to four Fido proteins. BlastP analyses were then performed with the full-length C. fetus proteins Fic1-4 to identify related Fido proteins from epsilon-proteobacteria or distantly related bacteria. Interestingly, BlastP analysis using full-length Fic2 or the degenerated motif of Fic2 consequently retrieved proteins of bacterial species linked to human fertility complications (Moreno et al., 2016; Pelzer et al., 2017). We used these proteins, related Campylobacter proteins and selected reference Fic-proteins (Harms et al., 2016b) to generate the Neighbor joining tree shown in Figure 9. The tree architecture placed the proteins in two main branches. Cluster A includes Fic3 and Fic4 of C. fetus subsp. venerealis 84-112. Reference FIC-domain proteins of Bartonella, Yersinia enterocolitica, and E. coli (marked with asterisks) were also placed in this cluster.

Fic1 and Fic2 are both grouped in the B branch and resolve in separate subclusters. Interestingly, Fusobacterium spp. also harbor FIC proteins closely related to Fic1. Fusobacteria inhabit mucous membranes of humans and animals and both Fusobacterium nucleatum and Fusobacterium necrophorum cause abortion in cattle (Kirkbride et al., 1989; Otter, 1996). Moreover F. nucleatum causes intra-amniotic inflection and premature delivery in humans and mice (Han et al., 2004; Gauthier et al., 2011). We also note that C. ureolyticus ACS-301-V-Sch3b, isolated from the human vaginal tract, encodes a protein related to Fic1 and another related to Fic3 and Fic4 of cluster A.

Fic2 clusters with a hypothetical protein of Campylobacter upsaliensis JV21 and proteins of C. helveticus and the more recently described novel species C. cuniculorum and C. corgagiensis. C. upsaliensis is a human enteropathogen that typically causes diarrhea, bacteremia and sepsis, but human infection with C. upsaliensis has also been associated with spontaneous abortion (Gurgan and Diker, 1994). Moreover, proteins of bacterial species suspected or confirmed to play a role in human fertility and pregnancy outcome cluster in the same branch (indicated with a plus sign).

In conclusion, our findings show conserved FIC protein sequences in a variety of bacteria that either inhabit the urogenital tract of humans and animals or are able to establish pathology in this niche.