Strains used in this study are listed in Table 1. P. putida was routinely grown in liquid LB (Sambrook et al., 1989) at 30°C on a rotary shaker operated at 200 rpm. For IutY induction experiments, P. putida cells were grown in liquid CAS medium (Llamas et al., 2006) containing 200 μM of 2,2′-bipyridyl (iron-restricted conditions) and 1 mM isopropyl β-D-1-thiogalactopyranoside (IPTG) supplemented with aerobactin containing supernatant derived from iron-restricted cultures of E. coli C600 (ColV-K30) in 1:1 proportion (Bastiaansen et al., 2014). For non-inducing conditions, the supernatant of iron-restricted cultures of E. coli C600 was used. Induction experiments with ferrioxamine and ferrichrome were performed following P. putida or P. aeruginosa growth in CAS medium supplemented with 400 μM of 2,2′-bipyridyl and in the absence or presence of 1 μM iron-free ferrioxamine B or 40 μM iron-free ferrichrome (both purchased from Sigma-Aldrich). Construction of null mutants was performed by allelic exchange using the suicide vector pKNG101 as previously described (Bastiaansen et al., 2014). Mutants were verified by PCR. When necessary, antibiotics were used at the following final concentrations (μg ml⁻¹): ampicillin (Ap), 100; piperacillin (Pip), 25; streptomycin (Sm), 50 for E. coli and 100 for Pseudomonas; tetracycline (Tc), 12.5 for E. coli and 20 for Pseudomonas.

Plasmids used are listed in Table 1 and primer sequences in Table S1. Phusion® Hot Start High-Fidelity DNA Polymerase (New England BioLabs) was used for PCR amplifications. Point mutations were introduced by nested PCR. All constructs were verified by DNA sequencing and transferred to P. putida or P. aeruginosa by electroporation (Choi et al., 2006).

Random mutagenesis of the entire N-terminally HA-tagged iutY gene was performed by error prone PCR on the pMMBK1-HA plasmid (Bastiaansen et al., 2014) using the low fidelity polymerase Pfu-Pol (exo⁻) D473G^* (Biles and Connolly, 2004) with primers NHA-PP2192-E and PP2192R-X (Table S1). PCR products were subsequently cloned in the EcoRI-XbaI restriction sites of the pMMB67EH plasmid (Fürste et al., 1986). After transformation to E. coli TOP10F' all clones were pooled prior to the isolation of plasmid DNA and transferred to P. putida KT2440 bearing the pMPK4 plasmid by electroporation (Choi et al., 2006). Clones with high β-galactosidase activity were selected on LB plates containing 50 μg/ml 5-bromo-4-chloro-3-indolyl-beta-D-galacto-pyranoside (X-gal) and 1 mM IPTG and sequenced following plasmid DNA isolation. For the second round of the screen, mutations in the cytosolic sigma factor domain of IutY were excluded by using the presence of a unique SacII restriction site at the position of the transmembrane domain of iutY. The pool of randomly mutagenized PCR products were cloned as SacII-XbaI fragments, thus containing only the periplasmic region of the protein, in the corresponding sites of the pMMBK1-HA plasmid, thereby keeping the HA-tagged cytosolic domain of iutY contained in this plasmid intact.

β-Galactosidase measurements were performed in soluble cell extracts using the o-nitrophenyl-β-D-galactopyranoside (ONPG) substrate (Sigma-Aldrich) as previously described (Llamas et al., 2006). Activity is expressed in Miller units. Each assay was run in duplicate at least three times and the data given are the average with error bars in all graphs indicating standard deviation (SD).

P. putida was grown to late log-phase in CAS medium containing 200 μM of 2,2′-bipyridyl and 1 mM IPTG, in the presence of either aerobactin deficient or aerobactin containing supernatant. Subsequently, cells were pelleted by centrifugation and heated for 10 min at 95°C following solubilization in SDS-PAGE sample buffer. Sample normalization was done according to the OD₆₆₀ of the bacterial culture. Proteins were separated by SDS-PAGE containing 15% (w/v) acrylamide and electrotransferred to nitrocellulose membranes (Millipore). As a loading control staining with a Ponceau S dye (Serva) was performed. Subsequently, HA-tagged proteins were detected using a monoclonal antibody directed against the influenza hemagglutinin epitope (HA.11, Covance) as previously described (Bastiaansen et al., 2014).

P. putida and P. aeruginosa genome sequences were obtained at www.pseudomonas.com (Winsor et al., 2011). Sequence alignments were performed with ClustalW (Goujon et al., 2010). Tertiary structure prediction was performed with the Phyre2 program at www.sbg.bio.ic.ac.uk/~phyre2 (Kelley and Sternberg, 2009). The quality of this prediction was assessed using DALI (Holm and Rosenstrom, 2010) and TM-align (Zhang and Skolnick, 2005), which calculate the root-mean-square deviation (RMSD) of the model by comparing it with its structure template.

To identify residues in the IutY protein important for activation of the cytosolic σ^IutY domain, we performed a random mutagenesis of the entire iutY gene and selected for gain of function mutations. An N-terminally HA-tagged version of iutY, which encodes a protein with similar activity as the non-tagged variant (Bastiaansen et al., 2014), was mutagenized by error prone PCR. The resulting products were cloned in the pMMB67EH vector behind an IPTG-inducible Ptac promoter and the pool of plasmids was transferred to the P. putida KT2440 wild-type strain bearing the σ^IutY-dependent iutA::lacZ transcriptional fusion (Bastiaansen et al., 2014). Subsequently, a blue-white screening was performed on solid LB medium supplemented with X-gal. Activity of σ^IutY induces expression from the iutA promoter and therefore increases lacZ production, resulting in blue colonies. A total of 46 bright blue colonies were selected. Sequence analyses showed that only seven of the blue clones encoded a full-length IutY protein, while the majority (39) acquired a premature stop codon in the iutY gene and therefore produced truncated IutY variants. This indicates that the most efficient method to create constitutive activity of IutY is to introduce a premature stop codon and these mutations will be discussed first. The shortest truncated variant (until amino acid 168) contained only the cytosolic σ^IutY domain, while the longest truncated variant obtained (until amino acid 365) only lacked the last nine amino acids of the protein (selected mutants are shown in Figure 2A and Figure S1). Other premature stop codons were randomly distributed over the periplasmic domain of the protein, suggesting that IutY can be truncated at any position in the anti-sigma domain to produce a constitutively active σ^IutY. Some of the truncated IutY variants also contained point mutations in the N-terminal IutY sigma factor domain. Although it was unlikely that these mutations were the cause of the increased σ^IutY activity, we exchanged the N-terminus of several truncated IutY variants for the wild-type sequence to exclude any effect of alterations in the sigma domain.

To validate the results of the screen and quantify σ^IutY activity, we measured β-galactosidase activity of the P. putida wild-type strain bearing the iutA::lacZ fusion upon expression of five selected IutY truncated variants (HA-IutY-168, -201, -260, -318, and -365). We also included in this analysis four additional IutY constructs, encoding HA-IutY-225, -236, -293, or -338 (Table 1). This resulted in a total of nine truncated variants of different length, covering the complete periplasmic anti-sigma domain of IutY (Figure 2A and Figure S1). Activity of these proteins was compared with that of the full-length IutY protein upon growth of P. putida in LB medium. All strains expressing a truncated IutY protein displayed very high lacZ activity (i.e., high iutA promoter activity; Figure 2B). This suggests that removing (part of) the periplasmic anti-sigma domain of IutY strongly triggers σ^IutY activity in a signal-independent manner. To confirm this, the HA-IutY-168, -201, -236, and -365 truncated variants (Figure 2A) were expressed in a ΔiutY deletion mutant (Bastiaansen et al., 2014) and σ^IutY activity was measured in the absence and presence of aerobactin. As expected, addition of the siderophore to the medium significantly increased expression from the iutA promoter in the strain expressing the full-length HA-IutY protein (~5-fold; Figure 2C, WT protein). All four tested truncated IutY variants exhibited very high promoter activity in the absence of aerobactin, which did not increase upon addition of the siderophore (Figure 2C). The level of lacZ activity produced by the mutant proteins was similar to that obtained with the wild-type protein upon induction with aerobactin. This indicates that removing part of the C-terminal region of IutY is by itself sufficient to produce maximal σ^IutY activity.

IutY is known to be processed by both the Prc and RseP proteases in response to aerobactin, which allows the release and activation of the cytosolic σ^IutY domain (~23 kDa; Figure 1; Bastiaansen et al., 2014). To assay whether the activity of the truncated IutY variants still depends on these proteases, σ^IutY activity was tested by measuring iutA::lacZ expression of the truncated iutY variants in Δprc and ΔrseP mutant strains (Bastiaansen et al., 2014) upon growth in LB (Figure 2B). The presence of an N-terminal HA-tag in the truncated variants allowed us to also examine protein processing by Western-blot (Figure 2D). These analyses showed that the activity of all truncated IutY variants, except that of the IutY-168 protein, was completely dependent on the RseP protease (Figure 2B). RseP is known to generate the active σ^IutY domain by cleaving within or near the transmembrane domain (Bastiaansen et al., 2014), a domain that is lacking in IutY-168 (Figure 2A). In accordance, two protein bands corresponding to the predicted size of the full-length truncated variant and the σ^IutY domain, respectively, were visible for all constructs except IutY-168 (Figure 2D). For IutY-168, only the band that corresponds to the predicted full-length size of the protein was detected (Figure 2D). However, although IutY-168 is not processed by RseP, its activity was significantly reduced in the ΔrseP mutant when compared with the wild-type strain (Figure 2B). This may be related with a reduced stability of IutY-168 in the absence of RseP, since the amount of this protein in this mutant was significantly lower (Figure 2D). Surprisingly, the other truncated proteins were not detected in the ΔrseP mutant (Figure 2D), which could suggest a role for RseP in the assembly and/or stability of these proteins in the membrane.

The β-galactosidase measurements in the Δprc mutant showed that the activity of the four shortest truncated variants (IutY-168, -201, -225, and -236) was not affected by the deletion of Prc, while the activity of the longer truncated variants was either completely abolished (IutY-293) or significantly reduced (IutY-260, -318, -338, and -365; Figure 2B). Western-blot analyses revealed that the stability and processing of the four truncated variants that were independent of Prc were not affected, whereas clear differences were observed for the other truncated variants between the wild-type background and the Δprc mutant (Figure 2D). Importantly, stability of the IutY proteins in the Δprc mutant correlated with activity. The protein levels of the truncated variants with impaired activity in this protease mutant (namely IutY-260, -293, -318, -338, and -365) were significantly lower than that of the fully active proteins (IutY-168, -201, -225, and -236; Figures 2B,D). Altogether, our results indicate that IutY proteins shorter than 236 amino acids are probably not processed by Prc and likely become direct substrates of the RseP protease.

In addition to the truncated IutY variants, we obtained several clones encoding full-length IutY protein variants in the random mutagenesis screen. Most of these mutants contained multiple mutations (between 2 and 4), both in the periplasmic anti-sigma domain and the cytosolic σ^IutY domain (data not shown). In order to avoid amino acid substitutions in the cytosolic region of the protein, we repeated the screen but with a different cloning strategy to exclude mutations in σ^IutY (described in Section Material and Methods). Subsequently, the plasmids were transferred to P. putida KT2440 and constitutive σ^IutY activity was detected on LB plates supplemented with X-gal. Bright blue clones were first analyzed by Western-blot to select for strains expressing a complete IutY protein and then sequenced. This approach identified in total 20 full-length active IutY mutants (Figure S2). Two of these clones contained only a single changed residue in the anti-sigma domain (R231Q and W301G), while the others contained several mutations. We selected 11 positions that were affected in several IutY mutant proteins (Figure S2) and introduced these single amino acid substitutions in the wild-type N-terminally HA-tagged iutY gene. Moreover, we generated two additional mutants in conserved residues (Figure S2). The functionality of these 13 IutY mutant proteins was tested by determining β-galactosidase activity of the iutA::lacZ fusion upon expression in the ΔiutY mutant in the absence or presence of aerobactin. Activity of six IutY variants was similar to that of the wild-type IutY protein (data not shown), indicating that these single mutations did not confer constitutive activity to the protein. In contrast, the activity of the IutY-D210A, -R231Q, -A240E, -F251S, -V253D, -W301G, and -LtoP (in which three leucines that form a conserved motif implicated in the interaction with the CSS receptor (Enz et al., 2003) were mutated to proline residues) proteins (Figure 3A) was significantly increased in the absence of aerobactin (Figure 3B). Activity of these mutant proteins did not further increase upon addition of the siderophore (Figure 3B), which indicates that these IutY protein variants completely lost their capacity to respond to the inducing signal. The production of the σ^IutY domain was analyzed by Western-blot (Figure 3C). It should be noted that the wild-type (WT) protein was detected in this analysis in which the strains were grown in low iron conditions (Figure 3C), whereas the same protein could not be detected after growth in LB medium (Figure 2D). This is likely due to a higher stability of IutY in iron limitation, which is not surprising given that iron limitation is required in vivo for expression and activation of CSS σ^ECF/anti-sigma factor pairs that regulate iron uptake (Llamas et al., 2014). While in the strain expressing the wild-type IutY protein the σ^IutY domain (~23 kDa) was detected in high amounts only upon growth with aerobactin, it was highly present in all constitutively active mutants even in absence of the siderophore (Figure 3C) and correlated well with the measured activity of the proteins (Figure 3B). Together, these results show that the constitutively active IutY mutants are, as the wild-type protein, activated by the proteolytic release of the σ^IutY domain.

To examine whether RseP and Prc are required for the proteolytic activation of the IutY full-length mutant proteins (Figure 3A), we measured iutA::lacZ activity in ΔiutY Δprc and ΔiutY ΔrseP double mutants upon bacterial growth in aerobactin. Similar to the wild-type protein, activity of all mutant proteins was completely abolished in the ΔiutY ΔrseP mutant (Figure 4A), indicating that the RseP protease is essential for the activation of these proteins. Prc deletion is known to completely abolish activation of IutY when this protein is produced from the bacterial chromosome (Bastiaansen et al., 2014). Interestingly, activity of wild-type IutY expressed from plasmid in the absence of Prc (ΔiutY Δprc mutant) was considerably reduced but not completely abrogated, since the protein retained ~7% of the lacZ expression measured in the ΔiutY single mutant (Figure 4A). Therefore, this result suggests that overexpression of iutY from plasmid produces protein levels that can bypass the absolute requirement for Prc. Similar to the wild-type protein, the IutY-W301G mutant also retained ~10% of its activity in the ΔiutY Δprc strain, whereas the IutY-R231Q and -LtoP proteins retained ~30% of their activity (Figure 4A). This implies that these proteins are mostly dependent on Prc for activation. In contrast, activation of the IutY-F251S and -V253D proteins was largely independent of Prc, since ~70% of their activity was retained in the protease mutant (Figure 4A). Western-blot analyses showed that all IutY proteins were notably less stable in conditions in which they were not or only partially active (Figure 4B). In accordance, all proteins were hardly detectable in the ΔiutY ΔrseP mutant (Figure 4B), in which activity is completely abolished (Figure 4A). In the ΔiutY Δprc mutant, both the full-length IutY form and the σ^IutY domain were clearly detectable for those IutY variants that were active in this mutant (IutY-F251S and -V253D), only low amounts were detectable for those in which activity was compromised (IutY-R231Q and -LtoP), and they were hardly detectable for proteins with impaired activity (wild-type IutY and IutY-W301G; Figure 4B).

Importantly, our results show that single amino acid substitutions in the periplasmic anti-sigma domain of IutY can lead to constitutively active proteins that require RseP and/or Prc for activation, but not the presence of aerobactin. Since the signal in CSS pathways is sensed by the CSS outer membrane receptor (Llamas et al., 2014), this suggests that the IutY mutant proteins do not depend on IutA. To test this, a ΔiutY ΔiutA double mutant was constructed and the activity of the mutant proteins in response to aerobactin was assayed. As expected, deletion of iutA resulted in only residual activity of the wild-type IutY protein, with only ~5% of the lacZ production (Figure 4A; Bastiaansen et al., 2014). In contrast, all tested IutY mutants (R231Q, F251S, V253D, W301G, and LtoP) were active in the ΔiutY ΔiutA double mutant (Figure 4A). In agreement with these results, Western-blot analyses showed that only the amount of the wild-type IutY protein was significantly affected in the absence of IutA (Figure 4B). This confirms that the activity of the IutY mutant proteins does no longer depend on the IutA receptor.

Alignment of the periplasmic regions of all P. putida and P. aeruginosa CSS anti-sigma factors (Llamas et al., 2008, 2014) showed that several of the constitutive mutations described above were altering conserved residues (i.e., Asp-210, Arg-231, Phe-251, Val-253, Trp-301, and Leu-313/317/320; Figure 5, boxed residues). To examine whether similar amino acid substitutions in other CSS anti-sigma factors could also induce signal-independent activity of their cognate σ^ECF, we introduced point mutations in the FoxR and FiuR proteins of P. aeruginosa (Figure 5, boxed residues). These P. aeruginosa anti-sigma factors are part of classical CSS pathways in which the sigma and the anti-sigma factor proteins are produced as two separate proteins and control the activity of σ^FoxI and σ^FiuI in response to the siderophores ferrioxamine and ferrichrome, respectively (Llamas et al., 2006). Single point mutations equivalent to the D210A, R231Q, and V253D substitutions in P. putida IutY (Figure 3) were introduced in N-terminally HA-tagged versions of both P. aeruginosa FoxR (yielding the HA-FoxR-D132A, -R153Q, and -V180D protein variants) and FiuR (yielding the HA-FiuR-D133A, -R153Q, and -V177D protein variants). Activity of these proteins was assayed by β-galactosidase measurements in P. aeruginosa ΔfoxR and ΔfiuR mutant strains (Mettrick and Lamont, 2009) bearing the σ^FoxI- or σ^FoxI-dependent transcriptional fusion foxA::lacZ or fiuA::lacZ, respectively (Llamas et al., 2006). Expression of the wild-type FoxR or FiuR proteins only allowed σ^ECF-mediated lacZ expression in response to the inducing signal (ferrioxamine or ferrichrome, respectively; Figure 6). Production of all mutant proteins resulted in significantly higher σ^ECF activity in the absence of the inducing signal (Figure 6, white bars). One of the mutants, FoxR-R153Q, still responded to ferrioxamine (Figure 6), which indicates that this protein is, in contrast to the other mutant proteins, still able to sense and respond to the presence of the siderophore. Several other mutants, i.e., FoxR-D132A, FoxR-V180D, and FiuR-R154Q, did show constitutive activity, albeit at a lower level as compared to the wild-type protein. These mutants seem to have lost the ability to respond to the inducing signal. Finally, we also identified two mutants, FiuR-D133A and -V117D that induced full σ^FiuI activity independent of ferrichrome (Figure 6). This indicates that some of the mutations identified in the IutY system have a crucial and conserved function in signal transduction in CSS anti-sigma factors.

In order to provide more insight into the mutations that cause constitutive activity of σ^IutY, we generated a structural model of the periplasmic region of P. putida IutY (IutYperi) using the Protein Homology/analogy Recognition Engine (Phyre2) server (Kelley and Sternberg, 2009; Figure S3). IutYperi is predicted to fold in two distinct regions (i.e., the N- and C-terminus), which are connected by a stretch of 12 amino acids (Figure S3A). We mapped the point mutations that resulted in constitutive activity of σ^IutY (Figure S3A, in blue), which showed that Trp-301 is located in the connecting region between the N- and C-terminus of IutYperi. Asp-210, Arg-231, Ala-240, Phe-251, and Val-253 appear to be clustered in the N-terminal periplasmic domain. Interestingly, most of the mutations, with the exception of Ala-240, are located on the same side of the N-terminal domain of IutY. This part of the N-terminal domain is expected to be facing or in close proximity of the cytoplasmic membrane, since the attachment point of the transmembrane segment is also located on this side of the protein. The Leu-313, -317, and -320 residues are present in a α-helix in the C-terminus and substitution of these three amino acids by prolines in the IutY-LtoP mutant is likely to disrupt the secondary structure of this helix. Subsequently, we created similar models for the periplasmic regions of P. aeruginosa FiuR (FiuRperi) and FoxR (FoxRperi; Figures S3B,C). This showed that the overall fold of these proteins is predicted to be similar, with the mutations tested in Figure 3 (IutY) and Figure 6 (FoxR and FiuR) in identical positions.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.