The bacterial strains and plasmid used in this study are listed in Supplementary Table S1. Bacteria were cultured in Luria-Bertani (LB) broth at 37°C in shaking (150 rpm) conditions. To prepare material from mid-exponential cultures, the overnight culture was diluted 1:100 in fresh LB medium and collected at optical density (absorbance at 600 nm) of ∼0.2–0.3. The remaining culture was incubated for additional 18 h to obtain stationary phase cultures. When required, the medium was supplemented with ampicillin (50 μg/ml), kanamycin (30 μg/ml), tetracycline (10 μg/ml), or chloramphenicol (10 μg/ml).

The mutagenesis was carried out with the Quick-change^TM site-directed mutagenesis kit from Stratagene, following manufacture recommendations. The oligonucleotides used for these procedures, including those degenerated introducing the desired point mutations, are listed in Supplementary Table S2. A copy of the S. Typhimurium wild-type igaA gene was cloned in the pBAD18 vector (plasmid pNG1062, Supplementary Table S1) and used as template for the mutagenesis kit to obtain pNG1062-derivate plasmids containing the mutant alleles (C404S, C425S, C498, and C504S), which were cloned in E. coli DH5α (Supplementary Table S1). The desired mutations were confirmed by sequencing. To generate the C404S–498S and C404S-C504S mutant alleles, a BlpI/BlpI fragment from the pLR1435 [pBAD18::igaA(C404S)] was used to replace the same region in pLR1438 [pBAD18::igaA(C498S)] and pLR1481 [pBAD18::igaA(C504S)] plasmids. The series of pNG1062-derivate plasmid (pBAD18 backbone) was transferred to the S. Typhimurium MD0835 strain [igaA2::KXX Δ(apbE′-rcsC′)], a mutant not producing IgaA and with an additional mutation in rcsC (Mariscotti and Garcia-Del Portillo, 2008). The production of the distinct IgaA variants with mutated cysteine residues was confirmed in the MD0835-derivate strains grown in LB-0.2% L-arabinose and subsequent analysis of total protein extracts by Western assay using anti-IgaA antibody (Cano et al., 2002).

To transfer the igaA mutant alleles to the chromosome, they were first moved from the pBAD cloning vector to the pCVD442 suicide vector (Donnenberg and Kaper, 1991), which has the counter selectable marker sacB, using E. coli DH5α as host strain (Supplementary Table S1). Since we did not initially know whether the igaA mutant alleles could support viability (igaA is an essential gene in a rcsCDB⁺ background), the igaA mutant alleles were first moved to the chromosome of S. Typhimurium strain MD1446 (igaA2::KXX zhf-6311::Tn10dTet rcsC::MudQ). This was done by conjugation using as donor E. coli SM10aaapir carrying the respective series of pCVD442 derivate plasmids. Loss of kanamycin resistance in the S. Typhimurium recipient strain was indicative of replacement of the igaA null allele (igaA2::KXX). The cysteine-defective igaA alleles were further transferred by P22 phage transduction to a clean wild-type background selecting by tetracycline resistance (flanking marker zhf-6311::Tn10dTet). The igaA gene was PCR-amplified from all Tet^R transductants to confirm presence in the chromosome of the mutation in the codon of the corresponding cysteine residue. Whereas all single igaA mutants (C404S, C425S, C498S, and C504S) proved to support viability, no transductants were obtained attempting to transduce to a wild-type background the double mutant alleles C404S–C498S and C404S–C504S. This result was consistent with the viability test performed with the different igaA variants expressed from pBAD using different L-arabinose concentrations (see below, Table 1).

Preparation of protein extracts, electrophoresis and Western assay conditions using polyclonal rabbit anti-IgaA antibody were as described (Dominguez-Bernal et al., 2004).

These assays were performed using the 4′-acetamido-4′-maleimidylstilbene-2,2′-disulfonic acid (AMS) reagent, as described (Jurado et al., 2006).

Levels of β-galactosidase derived from the gmm::lacZ transcriptional fusion were assayed as described by Miller, following the chloroform/SDS permeabilization procedure (Miller, 1972). For these assays, bacteria were grown in LB medium to mid-exponential phase (OD₆₀₀ ∼ 0.2–0.3).

Motility of the different strains used was monitored by motility assays in soft agar plates, as described (Rosu et al., 2006).

Data were analyzed by one-way ANOVA using Prism version 5.0 (Graph-Pad Software). Differences in values with P < 0.05 were considered significant.

Our previous studies in S. Typhimurium showed that IgaA is an inner membrane protein produced at relatively constant levels in all growing conditions (Dominguez-Bernal et al., 2004). Programs that predict transmembrane helix regions and protein topology (THMM, SOSUI, TMPred, PredictProtein) indicate that the 710 amino acid protein IgaA has five transmembrane segments, resulting in two cytosolic domains (residues 22–202, 247–336), one small periplasmic loop (221–225) and one major periplasmic domain (residues 358–652) (Figure 1A). Mutations in specific residues of the two cytosolic loops (R188H, T191P, G262R) as well as in the periplasmic domain (L514P, L643P) impact negatively the capacity of IgaA to repress the RcsCDB phosphorelay (Dominguez-Bernal et al., 2004). We further noted that the periplasmic domain of S. Typhimurium IgaA has four cysteines, (C404, C425, C498, C504) highly conserved in IgaA orthologs found in distinct genera of enteric bacteria (Figure 1B). Based on this observation, we assessed whether these periplasmic cysteines could form disulfide bonds and play an important role in function.

To determine the role played by the C404, C425, C498, and C504 residues, we generated isogenic S. Typhimurium mutants lacking each of these cysteines. To this aim, we first introduced in the chromosome the corresponding igaA point mutant allele using as recipient an igaA::KXX rcsC strain, to subsequently pass the allele to a wild-type (rcsCDB⁺) genetic background by P22 phage transduction (see section “Material and Methods”). Importantly, none of the C404S, C425C, C498S, and C504S mutations affected protein stability in actively growing bacteria. Thus, in bacteria grown to exponential phase these IgaA variants were detected with similar levels than those of wild-type IgaA or the previously characterized R188H variant (Figure 2A) (Dominguez-Bernal et al., 2004). Nonetheless, all these four cysteine variants (C404S, C425S, C498S, and C504S) were unstable in stationary phase (Figure 2A). This phenomenon was reminiscent of that observed for other partially inactive variants such as R188H and L514P (Dominguez-Bernal et al., 2004). Therefore, we concluded that the elimination of any of the four conserved residues (C404, C425, C498, and C504) may lead to structural changes in IgaA that affect its stability when bacteria reach stationary phase. Considering the essentiality of IgaA linked to necessary repression of the RcsCDB system, all these cysteine variants were, however, expected to retain some partial function in growing bacteria.

To determine the capacity of the IgaA cysteine variants to repress the RcsCDB phosphorelay, we monitored several phenotypic traits associated to the activity of this regulatory system. We included production of colanic acid capsule, which is positively controlled by RcsCDB; and, flagella production, negatively regulated by the system. In a first series of experiments, we measured in actively growing (exponential) and resting (stationary phase) bacteria the expression levels of gmm (wcaH), a gene encoding GDP-mannose mannosyl hydrolase, an enzyme involved in colanic acid capsule synthesis. The results showed that the C404S mutation caused partial de-repression of the RcsCDB system, even in actively growing bacteria when the protein remained stable (Figures 2A,B). Among the other mutants, a gradual variation in the level of RcsCDB activity was noted following the order C404S > C425S ∼ R188H > C504S > C498S > wild-type (Figure 2B). Interestingly, the levels of the RcsCDB system in stationary phase inferred from the gmm::lacZ reporter fusion increased only slightly despite the partial degradation observed for some of the IgaA mutant proteins such as C404S, C425S, C498S, or C504S (Figures 2A,B). These data are consistent with a major role of IgaA in repressing the RcsCDB that is critical only during active growth. These data also indicated that C404 and C425 are cysteines more important for function in comparison to C498 and C504.

Additional phenotypic traits that were examined included the formation of mucoid colonies on plates (signal of capsule formation) and motility assays in soft agar plates. In agreement with the data obtained with the gmm(wcaH)::lacZ reporter fusion, the mutant producing IgaA-C404S was highly mucoid and non-motile (Figures 3A,B). For the rest of mutants, a gradual variation in RcsCDB activity was noted in the mucoidity and motility tests (Figures 3A,B), which in some cases were not completely matching the gmm::lacZ assays performed in liquid culture (Figure 2B). Thus, although the IgaA-C504S variant exhibited lower gmm::lacZ expression than R188H (Figure 2B), bacteria producing this C504S variant were slightly more mucoid on plates (Figure 3A). R188H, C425S, and C504S variants also displayed an intermediate phenotype in motility despite their variations in the gmm::lacZ assays or mucoidy on plates (Figures 2B, 3A,B). Such discrepancies in the different tests may be influenced by the different growth conditions used -liquid culture vs. solid agar media plates-. Interestingly, mutations in defined RcsB residues alter the phosphorylation status of this regulator with consequences in either mucoidy or motility, but not in both phenotypic traits (Casino et al., 2017). Some of the mutations described here in the periplasmic cysteines of IgaA could result in distinct RcsB∼P/RcsB ratios, a hypothesis to be tested in future studies. Despite the minor phenotypic differences noted in IgaA for the R188H, C425S, and C504S variants; taken together, our data support that among the four periplasmic cysteines analyzed, C404 and C425 are residues with a more critical role for repression of the RcsCDB phosphorelay.

To elucidate whether the contribution of the periplasmic cysteines to IgaA function relies in the formation of disulfide bonds, we carried out alkylation experiments using the 4′-acetamido-4′-maleimidylstilbene-2,2′-disulfonic acid (AMS) reagent (Jurado et al., 2006; Denoncin et al., 2013). These experiments were performed in exponential phase (OD₆₀₀ ∼0.2–0.3), in which all IgaA variants with mutated cysteines are stable (see Figure 2A). AMS is a maleimide compound that binds to free thiol groups. Thus, it is possible to differentiate the presence of a disulfide bond if changes in electrophoretic mobility are detected in samples incubated with AMS and previously treated or not with a reducing agent such as di-thio-threitol (DTT). When S. Typhimurium wild type cells were incubated in solutions with or without DTT and further treated or not with AMS, we detected four IgaA forms with distinct electrophoretic mobility (Figure 4A). This result proved the presence of at least one disulfide bond in IgaA.

Besides the four periplasmic cysteines, IgaA of S. Typhimurium has four additional cysteines that map in the first and second transmembrane regions (C14, C219, respectively), the second cytosolic domain (C259) and, the short cytosolic domain encompassing the C-terminal end of the protein (C697). To discard the contribution of these cysteines to the mobility shift displayed by wild-type IgaA in the alkylation assays, we generated a variant lacking the periplasmic domain (ΔI358-W652). This variant, unlike the full-length protein, did not exhibit electrophoretic shift in cells exposed to DTT and subsequently to AMS (Figure 4A). Therefore, none of the non-periplasmic cysteines of IgaA contribute to the formation of disulfide bonds. Interestingly, the different electrophoretic forms of IgaA observed in the AMS alkylation assays in wild-type bacteria were also detected in a ΔdsbA mutant (Figure 4A). This result indicated that the disulfide bond present in IgaA can be formed in the absence of the major disulfide oxidase DsbA.

To define the configuration of the periplasmic disulfide bond(s) inferred in wild-type IgaA, we next performed AMS alkylation assays in isogenic strains lacking each of the four conserved periplasmic cysteines. These assays revealed an important contribution of C404 in the formation of a disulfide bond. Thus, unlike the C425S, C498S, and C504S variants, the C404S variant did not migrate differently in the presence/absence of DTT (Figure 4B). Interestingly, the four variants in the conserved periplasmic cysteines (C404, C425, C498, and C504) showed differences in mobility when comparing the samples further exposed to AMS (Figure 4B). This result led us to consider the presence of one “native” C404–C425 disulfide bond and contrasting outcomes when one or the other cysteine residue is mutated. In the absence of C404, we only see possible a non-native C498–C504 bond with minor consequences in electrophoretic migration as evidenced by the similar behavior of -/+ DTT samples (Figure 4B). In contrast, the lack of C425 might favor the formation of non-native C404–C498 or C404–C504 disulfide bonds, which could explain the clear electrophoretic shift observed in the -/+ DTT samples of the C425 mutant (Figure 4B). The more important role assigned to C404 and C425 in comparison to C498 and C504 agrees with the much higher RcsCDB activity registered when any of these two cysteines (C404 or C425) is lacking (Figures 2B, 3A,B).

Since the IgaA-R188H variant is only partially efficient in repressing the RcsCDB system (Cano et al., 2002; Dominguez-Bernal et al., 2004), we sought to determine whether this mutation could have effect on the C404–C425 disulfide bond. Unexpectedly, we detected molecules with distinct redox state that co-exist in this IgaA-R188H variant (Figure 4A). Thus, two forms with distinct electrophoretic mobility were clearly distinguishable in samples non-treated with DTT or the alkylating agent AMS (Figure 4A). Interestingly, DTT moved these two forms to a single one that was coincident to the fully reduced form of wild type IgaA (see Figure 4A). This result demonstrated that a mutation such as R188H mapping in a cytosolic loop can affect proper formation of a disulfide bond in the major periplasmic loop.

Taken together, these data are consistent with a model in which the function of IgaA as repressor of the RcsCDB system depends on the redox state of its periplasmic domain, which in its functional conformation may involve the formation of a C404–C425 disulfide bond (Figure 4C).

To further support the essential role that the C404–C425 disulfide bond plays in repression of the RcsCDB phosphorelay, we generated IgaA double mutants. Our aim was to affect the native bridge (C404–C425) and to simultaneously impair alternative non-native disulfide bonds predicted by the AMS alkylation assays and involving either C498 or C504 (Figure 4B). Therefore, the new IgaA variants were C404S–C498S and C404S-C504S. The AMS alkylation assays revealed no difference in migration between the non-alkylated and alkylated samples not exposed to DTT for any of these two double mutants (Figure 4D). This result, much more evident in the case of the C404S–C498S variant, supported the absence of free periplasmic cysteines and, therefore, the presence a “non-native” disulfide bond (C425–C504). In the C404S-C504S variant, the alkylation assays revealed two bands in the case of the samples non-treated with DTT and exposed to AMS (Figure 4D), implying the co-existence of molecules with and without a non-native C425–C498 disulfide bond. This phenomenon resembled at some extent the co-existence of molecules with different redox state observed for the IgaA-R188H variant (Figure 4A).

We next assessed whether the non-native C425–C498 or C425–C504 bonds could provide functionality to the protein. To this aim, we analyzed the capacity of the different IgaA variants to suppress lethality associated to the presence of an igaA::Km null allele in an RcsCDB⁺ genetic background (Cano et al., 2002; Mariscotti and Garcia-Del Portillo, 2008). The number of igaA::Km transductants obtained in strains bearing inducible pBAD18 vectors expressing the different IgaA variants was determined in the absence/presence of the inducer, L-arabinose. As a negative control, we used a strain with the pBAD18 empty vector, for which no Km^R transductants carrying the igaA:.km null allele were obtained regardless the absence/presence of inducer.

These assays showed that the C404S–C498S and C404S–C504S variants were not capable of suppressing the lethality associated to the igaA::km mutant when induced at 0.02% L-arabinose, a concentration sufficient to prevent lethality by any of the single mutants lacking each of the conserved cysteine residues (Table 1). Furthermore, we observed that when suppressing lethality at higher arabinose concentrations, all transductans were mucoid (Table 1). This result was indicative of the extremely limited function of these double C404S–C498S and C404S–C504S variants, even when produced at high levels. Therefore, the non-native C425-C498 and C425–C504 disulfide bonds are not optimal to provide function to IgaA.