The giant non-fimbrial adhesin SiiE is essential to establish intimate contact between Salmonella enterica and the apical surface of polarized epithelial cells. SiiE is secreted by a type I secretion system (T1SS) encoded by Salmonella Pathogenicity Island 4 (SPI4). We identified SiiA and SiiB as two regulatory proteins encoded by SPI4. Mutant strains in siiA or siiB still secrete SiiE, but are highly reduced in adhesion to, and invasion of polarized cells. SiiA and SiiB are inner membrane proteins with one and three transmembrane (TM) helices respectively. TM2 and TM3 of SiiB are similar to members of the ExbB/TolQ family, while the TM of SiiA is similar to MotB and a conserved aspartate residue in this TM is essential for SPI4-encoded T1SS function. Co-immunoprecipitation, bacterial two-hybrid and FRET demonstrate homo- and heterotypic protein interactions for SiiA and SiiB. SiiB, but not SiiA also interacts with the SPI4-T1SS ATPase SiiF. The integrity of the Walker A box in SiiF was required for SiiB–SiiF interactionand SiiF dimer formation. Based on these data, we describe SiiA and SiiB as new, exclusively virulence-associated members of the Mot/Exb/Tol family of membrane proteins. Both proteins are involved in a novel mechanism of controlling SPI4-T1SS-dependent adhesion, most likely by formation of a proton-conducting channel.
The ability to establish intimate contact to eukaryotic cell surfaces is a key virulence trait of most pathogenic microbes. This adhesion allows colonization of a host surface, protection against innate defence mechanisms but may also lead to subsequent pathogenic interferences such as host cell invasion.
Salmonella enterica is a versatile pathogen with the ability to infect various host organisms and to colonize various niches in infected hosts (Haraga et al., 2008). A key virulence trait of S. enterica is the invasion of non-phagocytic cells (Schlumberger and Hardt, 2006). Invasion is mediated by the translocation of a cocktail of effector proteins by the Salmonella Pathogenicity Island 1 (SPI1)-encoded type III secretion system (T3SS). These effector proteins subvert the regulation of the host cell actin cytoskeleton, resulting in macropinocytosis and internalization of the bacteria.
We recently demonstrated that the function of the SPI1-T3SS is not sufficient to mediate invasion of polarized cells by Salmonella (Gerlach et al., 2008). Here, the function of genes in Salmonella Pathogenicity Island 4 (SPI4) is required. SPI4 encodes a type I secretion system (T1SS) and the giant non-fimbrial adhesin SiiE which is the only known substrate of the SPI4-T1SS (Morgan et al., 2004; Gerlach et al., 2007a; Main-Hester et al., 2008). SiiE mediates adhesion to the apical side of polarized epithelial cells and this function is essential for subsequent SPI1-T3SS mediated invasion (Gerlach et al., 2008). We postulated that cooperation of SPI1-T3SS and SPI4-T1SS functions is necessary to tightly bind bacteria to microvilli on the apical membrane and to position the SPI1-T3SS for efficient translocation.
T1SS commonly secrete proteins into the culture supernatant and a well-characterized paradigm is the haemolysin of Escherichia coli (Delepelaire, 2004). Recently, T1SS have also been shown to secrete large repetitive proteins that contribute to biofilm formation. Such substrate proteins are LapA of Pseudomonas fluorescens (Hinsa et al., 2003), LapF of P. putida (Martinez-Gil et al., 2010) and BapA of S. enterica (Latasa et al., 2005). SiiE does not contribute to biofilm formation but mediates adhesion to host cells (Latasa et al., 2005; Gerlach et al., 2007a). Biofilm formation, as well as adhesion to host cells both requires that the adhesin is at least temporarily retained to the bacterial surface. A novel mechanism of retention and release by c-di-GMP regulated proteolytic modification has been reported for LapA of P. fluorescens (Newell et al., 2009; 2011). We also observed that SiiE is secreted into the culture supernatant but retained on the bacterial surface (Wagner et al., 2011). This retention is maximal in the phase of highest invasiveness.
We have recently characterized the SPI4-encoded T1SS for the secretion of the giant adhesin SiiE. The T1SS is composed of three canonical subunits, namely, SiiF as the inner membrane ATP-binding cassette (ABC) protein, SiiD as periplasmic adaptor protein (PAP) and SiiC as outer membrane protein (OMP). The function of these three subunits was essential for the secretion of the giant non-fimbrial adhesin SiiE as well as for virulence functions mediated by SPI4 such as adhesion to polarized epithelial and host cell invasion. The SPI4 locus contains two additional genes, siiA and siiB, that are co-regulated with siiCDEF and we proposed that siiABCDEF form an operon (Gerlach et al., 2007b). This was supported by a recent study identifying the SPI4 transcriptional start site with dRNA-seq (Ramachandran et al., 2012).
Since the mechanism of retention of SiiE to the bacterial envelope is essential for the understanding of SPI4-mediated adhesion, we performed a functional characterization of the SPI4-T1SS and investigated possible contributions of SiiA and SiiB. The data reported here indicate that S. enterica deploys SiiA and SiiB as a novel molecular mechanism to modulate the function of the adhesin SiiE.
The SPI4 locus encodes SiiA and SiiB with essential functions for polarized cell invasion
Previous work showed that SPI4 encodes SiiC, SiiD and SiiF as the canonical subunits of a T1SS, and SiiE as the only substrate protein of the T1SS known so far. SiiA and SiiB were shown neither to be important for SiiE secretion, nor to represent further substrate proteins (Gerlach et al., 2007a). siiA and siiB are part of the sii operon and expression is co-regulated with other sii genes (Gerlach et al., 2007b) (Fig. S1).
Based on the genetic data, we speculated that SiiA and SiiB may also contribute to SPI4-T1SS-mediated virulence functions of Salmonella. To test this possible contribution, we generated mutant strains deleted for siiA or siiB and determined invasion of non-polarized and polarized epithelial cells and adhesion to polarized epithelial cells in comparison to WT and SPI1- or SPI4-deficient strains. Invasion of non-polarized cells was highly reduced for the SPI1-T3SS-deficient invC mutant strain (Fig. 1A). However, neither a deletion of the entire SPI4 locus, nor defects in siiA, siiB or siiAB affected invasion of HeLa cells. In contrast to this, using polarized MDCK cells invasion rates of invC or SPI4 mutant strains were more than 1000-fold reduced. Similar invasion defects were observed for the siiA or siiB strains, although the invasion was slightly higher than for the SPI4 strain (Fig. 1B). Adhesion to MDCK cells was highly reduced for the SPI4-deficient strain, while the invC strain showed adherence similar to the WT strain. Mutant strains deficient in either siiA or siiB showed about 100-fold reduced adhesion, similar to the SPI4 strain (Fig. 1C). The invasion defects of siiA and siiB strains were restored by episomal expression of siiA::HA or siiAB::HA, respectively, under control of PsiiA (Fig. 1D).
Taken together, these data demonstrate a role for SiiA and SiiB in the SPI4-dependent adhesion to, and invasion of polarized epithelial cells. Since SiiA and SiiB were neither secreted nor belong to the canonical subunits of T1SS, we speculated that SiiA and SiiB have functions in the control of the T1SS or the substrate SiiE.
SiiA and SiiB are integral inner membrane proteins
We have previously shown that significant amounts of both proteins could be enriched in Salmonella inner membrane fractions (Gerlach et al., 2007a). Since these analyses were performed with epitope-tagged variants of siiA and siiB expressed from low-copy-number vectors, artefacts in subcellular localization due to overexpression have to be considered. Accordingly, we generated antisera against recombinant SiiA and SiiB and investigated the subcellular localization of native subunits. Subcellular fractionation of WT, SPI4, siiA and siiB strains was performed to separate membrane and cytosolic fractions (Fig. 2A). The previous analyses of plasmid-encoded, epitope-tagged SiiA indicated the presence of the protein in membrane as well as in cytosol fractions. In contrast, the present work demonstrates that chromosomally encoded SiiA is exclusively present in the membrane fraction. The same subcellular localization was obtained for native SiiB. We also investigated if the membrane localization of SiiA is affected by presence or absence of SiiB and vice versa. Cell fractionation was performed for siiA and siiB strains. The membrane association of SiiA or SiiB was not affected by the absence of SiiB or SiiA respectively. These data show that SiiA and SiiB each are membrane proteins that associate with the membrane independent of the other subunit. To investigate if SiiA and SiiB are integral or membrane-associated proteins, we performed selective extraction of protein from purified membrane fragments (Fig. 2B). Incubation with 2 M urea did not result in release of SiiA and SiiB in the soluble fraction. In contrast, extraction with 1% Triton X-100 released a majority of SiiA from the membrane into the soluble fraction. Treatment of membranes with 1% Triton X-100 resulted in partial solubilization of SiiB and increased concentrations of 2% or 10% resulted in increased amounts of solubilized SiiB. These characteristics are similar to control CpxA, a sensor protein of a two-component regulatory system anchored in the inner membrane (Fleischer et al., 2007). In contrast, none of the extractions resulted in significant solubilization of OmpA, which is in line with the stable insertion of OMPs in the outer membrane (Reusch, 2012). DnaK was not detected in any membrane or solubilized fraction, indicating the absence of cytosolic contamination. The requirement of detergent extraction for release of SiiA and SiiB from membranes indicates the membrane-integral nature of these proteins.
To further characterize SiiA and SiiB and their contributions to the functions of the SPI4-T1SS, their subcellular localizations were analysed. Plasmids for the expression of siiA::mCherry or siiB::mCherry were generated and these constructs complemented the adhesion and invasion defects of siiA or siiB strains respectively (Fig. S2). Strains expressing siiA::mCherry or siiB::mCherry fusions were used to infect MDCK and LPS staining was performed without prior solubilization of host cells in order to distinguish adherent and internalized bacteria. High magnification micrographs obtained after deconvolution of CLSM images indicated a presence of foci of SiiA and SiiB in association with the cell envelope (Fig. 2C). There was no detectable difference in the distribution of SiiA and SiiB foci in adherent (positive for LPS staining) or invaded (negative for LPS staining) bacteria. SiiA–mCherry and SiiB–mCherry foci were located at various positions of the cell envelope, but some cells showed more intense signals at the poles.
Sequence similarity of SiiA and SiiB to membrane proteins forming proton channels
SiiA and SiiB do not show significant sequence similarity to known bacterial proteins. However, blast searches using the N-terminal portions of SiiA and SiiB revealed sequence similarity to a family of proteins involved in formation of proton channels in the inner membrane (Fig. 3). The MotAB, TolQR and ExbBD proton channels are able to transduce energy stored in the proton motive force (PMF) to allow for flagellar rotation, outer membrane (OM) biogenesis and transport processes across the OM respectively (Lloubès et al., 2001; Minamino et al., 2008; Noinaj et al., 2010). In all these complexes the proton channel consists of three alpha helical transmembrane (TM) domains where two α-helices are donated from the larger subunit (MotA, TolQ, ExbB) and the third α-helix stems from the corresponding smaller subunit (MotB, TolR, ExbD) (Zhai et al., 2003; Braun et al., 2004; Zhang et al., 2011). Although the functions of these complexes are quite diverse, they show remarkable sequence homologies within these TM domains (Cascales et al., 2001; Goemaere et al., 2007). We generated sequence alignments of the TM domains of SiiA and SiiB. For SiiA amino acids (aa) 8–30, we found similarities to the TM region of MotB and, to a lesser extent, to the TM regions of ExbD and TolR (Fig. 3A). Strikingly, we observed a highly conserved aspartate (D13, bold red in Fig. 3A) in SiiA which has been shown to be essential for proton conduction in MotB, TolR and ExbD (Zhou et al., 1998; Cascales et al., 2001; Ollis et al., 2009). In case of two hydrophobic regions of SiiB located between aa 140 and 199, there were significant homologies to the two TM α-helices of ExbB and TolQ (Fig. 3B). A hallmark of these TM domains is the presence of a knobs-into-holes helix packing motif characterized by small aa (G/A/S) (green in Fig. 3B) separated by three larger aa (P/W/L/I/M). We also found this motif to be present in SiiB and in ZP_07375189, a putative Arhensia spp. protein of unknown function identified in blast searches using SiiB. Besides their putative TM domains, the C-terminal parts of SiiA and SiiB are quite different in aa sequence and length compared to known proton channel components (see Fig. S3 for complete alignments). To investigate whether SiiA and SiiB indeed form a protein conducting channel which is essential for SPI4-mediated adhesion we further characterized both proteins with respect to membrane topology and protein–protein interactions.
Topology of SiiA and SiiB in the cytoplasmic membrane
For SiiB, a Sec-dependent signal peptide comprising its first 25 aa was predicted using SignalP v4.1 (Petersen et al., 2011) (Fig. S3B). This peptide might guide SiiB for Sec/YidC-dependent insertion into the inner membrane (Luirink et al., 2012). Secondary structure predictions indicated the presence of one and three putative membrane-spanning regions in SiiA and mature SiiB (mSiiB) respectively (Fig. S3A). For a more detailed understanding of the topology of SiiA and SiiB, we generated reporter fusions with alkaline phosphatase and β-galactosidase at various positions of the proteins (Fig. 4A). These reporters have been frequently used to probe the membrane topology of bacterial proteins (reviewed in Traxler et al., 1993). The β-galactosidase (β-Gal) portion of a fusion protein is only enzymatically active when located in the cytoplasm, while the alkaline phosphatase (PhoA) portion of fusion protein requires at least transient contact to the periplasmic environment for activity (Fig. 4B). A novel approach for the rapid generation of reporter fusions to target genes within the chromosome of Salmonella was used (Gerlach et al., 2007c). SiiA fusions were generated after codons 60 and 210, resulting in reporter fusion after a hydrophobic, putative TM region and at the C-terminus respectively. For both positions, low β-Gal and high PhoA activities were determined (Fig. 4C). For SiiB (mSiiB), reporter fusions were generated after codons 101 (76), 166 (141), 241 (216) and 466 (441). Fusions after codons 76, 141 and 216 of mSiiB would result in reporter portions located after the first, second and third putative TM domain respectively. The fusion at codon 441 should position the reporter portions at the C-terminus of SiiB. Fusions after codon 76, 216 and 441 resulted in high β-Gal activity and very low PhoA activities. Strains expressing fusions after codon 141 showed high PhoA, but very low β-Gal activities (Fig. 4C).
Taken together, the reporter fusion analyses of SiiA and SiiB largely support the membrane topology prediction shown in Fig. 4A. In contrast to the prediction by TMHMM v2.0, the orientation of SiiA within the inner membrane seems to be reversed having a cytoplasmic N-terminus, followed by a TM domain, and the rest of the protein located in the periplasm. Besides the sequence homologies in the TM region (Fig. 3A), SiiA also shares its membrane topology with MotB (Chun and Parkinson, 1988), ExbD (Kampfenkel and Braun, 1992) and TolR (Kampfenkel and Braun, 1993a). For SiiB, our model proposes a periplasmic N-terminus, followed by TM domain 1, a short linker located in the cytoplasm, TM domain 2, a short linker located in the periplasm, TM domain 3, followed by the C-terminal half of SiiB located in the cytoplasm. These experimental data are in accordance with the predicted secondary structure of mSiiB (Fig. 4A) and those determined for ExbB (Kampfenkel and Braun, 1993b) and TolQ (Kampfenkel and Braun, 1993a). A topology model for SiiA and SiiB is shown in Fig. 9A.
The aspartate residue at position 13 of SiiA is critical for SPI4 function
Sequence alignments of the TM domain of SiiA with TM domains of MotB from different organisms and E. coli ExbD and TolR indicate a single highly conserved aspartate residue D13 in SiiA (Fig. 3A). For E. coli MotB, D32 is essential for torque generation and function of the flagellar motor (Zhou et al., 1998). Non-conservative substitution of D25 of ExbD or D23 of TolR also led to loss of function (Cascales et al., 2001; Ollis et al., 2009). Based on these previous results we speculated that D13 is essential for SiiA function and introduced D13N and D13H mutations. The mutations were introduced by a QuikChange protocol in the low-copy plasmid p3187 encoding siiA::HA (Gerlach et al., 2007a). The modified plasmids were introduced in the ΔsiiA strain and we determined the ability of the resulting strains to invade polarized MDCK cells. Whereas ΔsiiA [siiA::HA] was able to invade the MDCK cells at 70% of the WT level, both SiiAD13N and SiiAD13H resulted in highly reduced invasion rates comparable to those of a siiF mutant or the empty vector control (Fig. 5A).
In TolR, ExbD and MotB, the conserved aspartate residue is thought to be localized at the cytoplasmic side of the TM α-helix. Furthermore, there are indications that this negatively charged aa might get directly protonated in the proton conduction process (Braun and Blair, 2001; Zhai et al., 2003; Goemaere et al., 2007). Asp to Glu exchange experiments performed for TolR D23 and MotB D32 supported this model. The conservative aa exchange kept a negatively charged residue at these particular positions and retained protein function to full extent (TolR) or at least partially (MotB) (Zhou et al., 1998; Cascales et al., 2001). We generated and analysed a SiiAD13E variant and observed an intermediate phenotype with 4.6-fold increased invasion rates compared to SiiAD13N or SiiAD13H (Fig. 5A). In contrast, the invasion by ΔsiiA with SiiAD13E was still 15-fold reduced compared to ΔsiiA [siiA::HA] (Fig. 5A).
The single aa exchanges introduced within SiiA could lead to misfolding of the protein, resulting in decreased protein stability and/or altered subcellular localization. We performed a subcellular fractionation with all strains and compared the amount of SiiA in different fractions. Introducing the D13N or D13H mutations led to reduction of detectable SiiA in cell extracts, as well as inner membrane fractions (Fig. 5B). For SiiAD13E this effect was less pronounced and proper subcellular localization of SiiA to the inner membrane was observed for any of the mutations introduced (Fig. 5B).
In summary, these results indicate a critical role of aspartate 13 for the SPI4-encoded T1SS, presumably in proton conductance through an inner membrane channel formed by SiiAB.
SiiE surface expression is affected by proton motive force
If SiiA and SiiB are acting as a proton conducting channel, we speculate that this function would be affected by the actual proton motive force (PMF) at the inner membrane. To test this hypothesis, we experimentally destroyed the PMF using CCCP, an uncoupler that allows H+ movement over the membrane. We tested various amounts of CCCP and found concentration in the range of 50–100 μM to be effective in affecting bacterial growth and motility (Fig. S4A). This is within the concentration range used in previous studies on the role of PMF in secretion of flagellar subunits (Paul et al., 2008). Addition of CCCP or the solvent DMSO had no effect on the invasion of non-polarized cells by Salmonella WT (Fig. S4B), indicating that the general physiology of Salmonella and ability to deploy the SPI1-T3SS were not affected. Also, if CCCP was added to MDCK or HeLa cells prior to bacterial infection, the invasion was not affected, demonstrating that the uncoupler does not alter the Salmonella-induced actin reorganization required for internalization (Fig. S4C and D). If bacteria were pre-treated with CCCP 30 min before host cell infection, we observed a highly reduced invasion (Fig. 6A). Cells treated with CCCP or solvent DMSO were analysed for SiiE surface expression and in contrast to the DMSO sample we detected a strong decrease in SiiE on bacterial cells after treatment with CCCP (Fig. 6B and C).
To further test the potential role of PMF on SiiE surface expression, we recovered bacteria from subculture in rich media and subjected cell to washes in PBS adjusted to neutral to slightly acidic pH. This treatment should lead to an artificial increase of the PMF due to acidification of periplasmic space (Padan et al., 1976; Felle et al., 1980). We observed that lower buffer pH resulted in higher amounts of SiiE surface expression (Fig. 6D). The correlation was largely linear and highest surface expression was observed at pH 5.0. External pH did not affect SiiE surface expression of siiA or siiB mutant strains that showed low levels of surface-localized SiiE at the various pH values tested.
These data are in line with a role of the PMF on the control of surface retention of SiiE and suggest that SiiA and SiiB are required for sensing or transducing of PMF.
SiiA and SiiB show homotypic and heterotypic interactions
To function as a putative proton channel, SiiA and SiiB have to form a complex within the inner membrane. In order to detect this putative interaction we did co-immunoprecipitation (IP) assays using antisera against SiiA and SiiB. These experiments were done with bacterial lysates which had been treated with detergent in order to solubilize the integral membrane proteins SiiA and SiiB. In IP with anti-SiiB antiserum, we used ΔsiiA or ΔsiiB strains expressing siiA::HA (Fig. S5A). In IP with anti-SiiA antiserum, ΔsiiA or ΔsiiB strains harbouring p3394 for expression of siiB::HA were used (Fig. S5B). Both plasmids can functionally complement the corresponding mutant strains (Fig. 1D and data not shown). With anti-SiiB antiserum we were able to co-IP SiiA-HA in the complemented ΔsiiA strain. When we used the ΔsiiB strain in co-IP, no SiiA-HA could be detected (Fig. S5A). Vice versa, using the anti-SiiA antiserum we could co-IP SiiB-HA in the complemented ΔsiiB strain. Again, no SiiB-HA was detected in Western blots when the ΔsiiA strain was used (Fig. S5B). Depending on the presence of the bait protein we were able to co-IP SiiA with anti-SiiB antiserum and SiiB with anti-SiiA antiserum.
To investigate the interaction between SiiA and SiiB by an independent approach, we performed bacterial two-hybrid (B2H) assays. B2H allows analysis of protein–protein interactions within an intact bacterial cell and is based on the functional complementation of the T18 and T25 fragments of Bordetella pertussis adenylate cyclase CyaA. Putative interaction partners were either N- or C-terminally fused to T18 and T25 respectively. Upon interaction, CyaA enzyme activity is restored and cAMP production is visualized with the help of a cAMP/CAP-dependent reporter gene (Karimova et al., 1998). To assess interactions between SiiA and SiiB, we tested production of β-Gal by the cyaA-deficient E. coli reporter strain BTH101 (Karimova et al., 2000) using the chromogenic X-Gal substrate. Based on the membrane topology of SiiA and SiiB determined here (Fig. 4), we fused the CyaA fragments to the N-terminus of SiiA and C-terminus of SiiB to allow the cytosolic localization of the enzyme moieties. Coexpression of T18-SiiA and SiiB-T25, or T25-SiiA and SiiB-T18, yielded light blue colonies indicating interaction of the fusion proteins (Fig. 7). Using B2H we also tested for homotypic interactions of SiiA and SiiB. Coexpression of T18-SiiA and T25-SiiA led to a strong induction of lacZ exceeding the activity of the positive control (Fig. 7). When we expressed SiiB-T18 together with SiiB-T25 in E. coli BTH101, we also observed a high β-Gal activity comparable to that of the positive control (Fig. 7). Our results indicate that SiiA and SiiB form dimers or homooligomers as described for other components of proton channels (Higgs et al., 1998; Journet et al., 1999; Braun and Blair, 2001).
The enzymatically active ABC protein SiiF interacts with SiiB
We have previously shown that SPI4-mediated adhesion critically depends on the adhesin SiiE and its cognate T1SS SiiCDF (Gerlach et al., 2007a; 2008; Wagner et al., 2011). How can SiiAB control the function of SiiE and/or the T1SS? Due to their similar subcellular localization within the inner membrane we speculated that there might be an interaction between SiiAB and the ABC protein SiiF. We tested interactions between SiiA and SiiF in B2H without indications of β-Gal activity for any of the combinations tested (Fig. 7). However, we observed blue colonies with the reporter strains where SiiB was combined with SiiF (Fig. 7). Coexpression of SiiB-T18 with SiiF-T25 led to a slightly higher signal compared to the SiiB-T25/SiiF-T18 pair (Fig. 7). These results indicate that the SiiAB complex interacts with SiiF exclusively via SiiB.
Dimerization upon ATP binding is a feature of ABC proteins (Smith et al., 2002). A SiiFG500E variant mutated within the Walker A motif of the nucleotide-binding domain (NBD) cannot bind ATP and this mutant does not secrete SiiE (Gerlach et al., 2009). We wondered if this effect is due to reduced protein stability and/or altered subcellular localization of SiiFG500E. Subcellular fractionation of strains harbouring chromosomally 3xFLAG tagged WT siiF (WRG115) and siiFG500E (WRG118) indicated that both proteins were exclusively in the inner membrane (Fig. S6). Compared to WT SiiF, the SiiFG500E mutant only showed slightly reduced protein levels (Fig. S6).
In B2H assays, constructs with SiiFG500E fused to T18 and T25 fragments, or WT SiiF combined with the SiiFG500E mutant resulted in white colonies indicating absence of protein interactions (Fig. 7). We observed white colonies regardless of the combination of fusion partners used in the assay. We next tested if the SiiFG500E mutation affects interaction with SiiB. Interestingly, no interaction between the enzymatically inactive SiiFG500E and SiiB or other fusion proteins were observed, independent of the combination of CyaA fragments used (Fig. 7).
In summary, our data propose that there is a direct protein–protein interaction between SiiB and the ABC protein SiiF. This interaction was not observed with a SiiF mutant harbouring a non-functional NBD.
FRET analyses of interactions between the SPI4 components SiiA, SiiB and SiiF
To investigate if the identified protein–protein interactions are also present in Salmonella in vivo, we constructed expression vectors for the coexpression of CFP and YFP fusion proteins, which would allow detection of interactions by Förster resonance energy transfer (FRET). The combination of tetracycline-inducible coexpression and a low-copy vector backbone allowed us to adjust protein levels to nearly endogenous amounts (T. Wille and R.G. Gerlach, unpubl. data). Upon induction with AHT, all the constructs were able to complement at least partially the respective Salmonella deletion mutants (Fig. S7). We measured FRET efficiencies using the acceptor photobleaching approach (Fig. 8A) in both S. Typhimurium MvP103 and E. coli MACH1™ T1R. We considered an apparent FRET efficiency (EFRET) equal to or greater than 1% (red line in Fig. 8B) indicative of positive interactions between the fusion proteins, as the EFRET values for negative controls were below 1%. High FRET efficiencies were observed for the SiiB–YFP/SiiB–CFP pair regardless of the expression host. An EFRET of more than 6% was detected when both proteins were expressed in E. coli, which was significantly higher than in S. Typhimurium (Fig. 8A and B). For the interaction of individual SiiF fusion proteins we observed opposite results depending on the host used for expression. In contrast to the B2H assays done in E. coli BTH101 (Fig. 7), we failed to detect homotypic SiiF interactions in E. coli using FRET (Fig. 8B). Surprisingly, when expressed in S. Typhimurium, we detected not only an interaction for the WT SiiF FRET pair but also for the SiiF–YFP/SiiFG500E–CFP pair (Fig. 8B). If both SiiF fusion proteins had the Walker A mutation, no interaction could be detected, regardless of the expression host (Fig. 8A and B). We also investigated the heterogeneous interactions between SPI4 components using FRET. The FRET assays in E. coli as well as in S. Typhimurium indicated SiiA–SiiB and SiiB–SiiF interactions. The FRET efficiencies of both pairs only slightly exceeded 1% in either host (Fig. 8B).
In summary, in vivo FRET experiments support our results from B2H that the interaction between SiiB and SiiF depends on a functional Walker A motif in SiiF (Fig. 8A). In accordance with B2H data, no interaction between CFP–SiiA and SiiF–YFP or SiiFG500E–YFP could be detected (Fig. 8B).
In this study we provide experimental evidence that the SPI4-encoded proteins SiiA and SiiB form a multimeric proton-conducting channel in the inner membrane of Salmonella. We previously demonstrated that SiiA and SiiB were not secreted substrates of the SPI4-T1SS and that siiA or siiB mutant strains were still able to secrete SiiE (Gerlach et al., 2007a). However, this work shows that siiA or siiB strains are as attenuated in adhesion to, and invasion of polarized epithelial cells as previously observed for strains deficient in the adhesin SiiE or the T1SS subunit SiiF. These findings, together with the lack of secretion of SiiA and SiiB and the co-transcription of siiA and siiB with siiCDEF, strongly suggest that SiiA and SiiB are novel regulatory subunits of the SPI4-encoded T1SS.
SiiA and SiiB are both integral proteins of the cytoplasmic membrane (Fig. 2) (Gerlach et al., 2007a). Their predicted and experimentally determined membrane topologies are very similar to those observed for MotAB, ExbBD and TolQR (Kampfenkel and Braun, 1992; 1993a,b). On the sequence level we found that SiiA and SiiB share conserved aa motifs in the putative TM domains with these proteins (Fig. 3). Most notably, we demonstrated that one of these residues within SiiA, a conserved Asp, is absolutely required for SPI4-T1SS function (Fig. 5). Co-IP, B2H and FRET jointly demonstrate that SiiA and SiiB interact in order to form a complex. However, the exact stoichiometry of this complex is unclear. Both, SiiA and SiiB form dimers or higher-order oligomers due to homotypic interactions. This was also demonstrated for MotB (Braun and Blair, 2001), ExbBD (Higgs et al., 1998) and TolQ (Journet et al., 1999). Most detailed data on stoichiometry of these complexes were obtained for the flagellar stator complexes with a MotA4:MotB2 stoichiometry (Braun and Blair, 2001; Kojima and Blair, 2004). ExbB and TolQ are reported to form complexes composed of four to six monomers, whereas ExbD and TolR might form dimers (Journet et al., 1999; Cascales et al., 2001; Ollis et al., 2009; Pramanik et al., 2010). The analyses of local sequence similarity, protein topology, protein interactions and effects of mutations in the TM helix of SiiA reported here are in full agreement with a model in which a SiiA2:SiiB4 complex forms two proton channels. A model showing the possible organization of the different TM helices within such a complex, seen from the periplasmic side of the inner membrane, is depicted in Fig. 9B.
In the assembled T1SS the OMP and the PAP each form trimers whereas the ABC protein is a dimer (Delepelaire, 2004). Dimer formation of ABC proteins is directly linked to bound ATP which is sandwiched between the Walker A motif of the cis monomer and the ABC-signature motif of the trans monomer (Smith et al., 2002; Zaitseva et al., 2006). In accordance with these data we observed interactions between individual SiiF fusion proteins depending on a functional Walker A motif using B2H (Fig. 7) and FRET (Fig. 8). One exception is the lack of SiiF–SiiF interaction in E. coli using FRET (Fig. 8B) which was possible using B2H (Fig. 7). Here, the different E. coli strains used in the experiments might have influenced the homotypic interactions of SiiF or differences in the sensitivity of both assays led to the observed results. Supporting the first possibility, the same construct yielded high FRET efficiencies when expressed in the Salmonella background (Fig. 8B).
Surprisingly, similar high EFRET was observed after coexpression of SiiF::YFP and SiiFG500E::CFP fusion proteins in Salmonella but not in E. coli (Fig. 8B). It has been speculated that binding of a single ATP might be sufficient to stabilize a ‘closed dimer’ conformation of certain ABC proteins (Linton and Higgins, 2007). Our FRET data provide some evidence that this might be the case for SiiF, at least in Salmonella. In Salmonella the substrate SiiE, the PAP SiiD and the OMP SiiC were present which might support and/or stabilize SiiF–SiiFG500E dimerization in a cooperative manner. It is possible that SiiD forms a complex with SiiF in the IM, similar to HlyB–HlyD (Thanabalu et al., 1998), thus stabilizing the dimer with a single ATP bound.
Both B2H and FRET demonstrate that SiiB binds to ABC protein SiiF in E. coli and S. Typhimurium. This interaction was abolished when SiiFG500E was used, independent of the host or method (Figs 7 and 8B). All steps of the T1SS transport cycle go hand in hand with significant conformational changes within the ABC protein including dimerization upon ATP binding (Zaitseva et al., 2006). Therefore, SiiF might only be recognizable as binding partner for SiiB in its ATP-bound, dimeric form. Formation of a continuous channel bridging the periplasm requires substrate binding by the ABC protein and of course the presence of a PAP and OMP (Thanabalu et al., 1998). Interactions between SiiB and SiiF were also observed in E. coli where neither the substrate SiiE nor SiiCD were present (Fig. 8A). Our results point towards a specific recognition of the ATP-bound SiiF dimer by SiiB. Apparently, interaction between SiiB and SiiF does not require formation of the active T1SS channel itself. A model for SiiA, SiiB and SiiF interaction based on our data is depicted in Fig. 9A.
Given the interaction between SiiF and SiiB, this is the first study that shows the involvement of proton-conducting inner membrane complexes in protein secretion by a T1SS. Therefore, the SiiAB complex represents a novel accessory component of a T1SS. There are several examples of accessory components which can specifically modulate T1SS substrate activity. HlyC of pathogenic E. coli is involved in post-translational acylation of the α-haemolysin HlyA (reviewed in Stanley et al., 1998). In P. fluorescens, the c-di-GMP-binding protein LapD regulates the activity of the periplasmic protease LapG which controls surface localization of the adhesin LapA (Newell et al., 2011). In contrast to this, multimeric SiiAB complexes seem to be in close proximity to the SPI4-encoded T1SS itself presumably influencing its activity. Reduced amounts of secreted SiiE in the absence of SiiA or SiiB have been reported previously (Kiss et al., 2007). This rather mild impact is in contrast to the strong in vivo (Kiss et al., 2007) and in vitro phenotypes (Fig. 1) observed for both mutants. It seems that SiiAB are absolutely required for SPI4 function in a way that goes beyond control of SiiE secretion.
Why is regulation by SiiA and SiiB required for the SPI4-T1SS? SiiE is a large non-fimbrial adhesin required for the interaction of Salmonella with the apical surface of epithelial cells. The function requires the tight binding of the adhesin to the bacterial surface. While substrate proteins of T1SS are usually secreted into the medium, the specific function of SiiE requires an additional mechanism for the surface expression of the adhesin. SiiE secreted into the medium was not capable to functionally complement a siiE-deficient strain (Gerlach et al., 2007a). We observed a cooperativity of the SPI4-T1SS with the SPI1-T3SS (Gerlach et al., 2008). This interaction is restricted to a rather short phase of invasiveness, and we recently reported that surface expression of SiiE is drastically reduced if invasiveness decreases (Wagner et al., 2011). Surface expression of SiiE might interfere with the intracellular lifestyle Salmonella initiates after successful host cell invasion. Alternatively or in addition, Salmonella might temporarily restrict the surface expression of SiiE as a highly antigenic structure. In a mutagenesis approach we identified regions in the N-terminal portion of SiiE that contribute to its surface expression and the effectiveness of the subsequent invasion process (Wagner et al., 2011).
It is conceivable that SiiAB controls surface expression of SiiE, but so far there are no comparable systems described. We speculate that SiiAB form a proton conductive channel in the inner membrane. This channel may use the PMF to energize so far unknown reactions in the T1SS that are required for the SiiE surface retention or release. From similar proton channels it is known that the energy harvested from the PMF is transduced to specific binding partners by inducing energy-rich protein conformations. Whereas MotAB drives flagellar rotation interacting with FliG, the ExbBD channel transduces energy to TonB and TolQR to TolA/B-Pal (Cascales et al., 2001). The energy stored in conformations of TonB or TolA is further transduced to OM proteins by direct interactions thus energizing biogenesis of OM proteins and active transport processes. In case of SiiAB we identified the ABC protein SiiF as the only interaction partner so far. If we propose a similar mechanism, energy from proton flux might be transduced to SiiF in form of a conformational change. This energy-rich conformation of SiiF might, directly or indirectly, influence the activity of the adhesin SiiE. Besides this, energy could be also transduced via SiiA to a so far unknown interaction partner. SiiA possesses a long periplasmic C-terminal domain which might interact with an OM protein. Such a mechanism would support surface expression and therefore function of SiiE as an adhesin depending on energy stored in the PMF and converted by the SiiAB proton channel.
Alternatively, SiiAB may form a complex that senses the PMF and transmits this information to other components of the T1SS. In this model, the PMF would not provide energy for a coupled function, but rather provide information about the physiological state of the cell. The effect of external pH on SiiE surface expression (Fig. 6D) and our prior observation that SiiE surface expression is highly reduced if bacteria enter the stationary growth phase (Wagner et al., 2011) would be in line with this model. Additional experiments will be conducted to address whether one of these hypotheses hold true.
Bacterial strains and growth conditions
Salmonella enterica serovar Typhimurium NCTC 12023 was used as wild-type strain in this study and as background strain for all mutant strains generated. Strains are listed in Table S1. E. coli strains were used for cloning and for synthesis of recombinant SiiE proteins.
Bacterial strains were grown in appropriate volumes of LB broth at 37°C with aeration on a rotating or shaking platform. If required, antibiotics carbenicillin (Cb) and/or kanamycin (Km) were added to media at a concentration of 50 μg ml−1. For most experiments, bacteria that were grown over-night were diluted 1:31 (or 1:50 for synthesis of recombinant proteins) in fresh medium and subcultured for indicated periods of time at 37°C. Before cultivation, bacterial strains were streaked freshly from stock cultures, which were stored in 7% DMSO at −70°C.
Construction of recombinant DNA construct for complementation of mutations, protein synthesis or expression of mCherry fusion proteins was performed according to standard methods (Sambrook et al., 1989). Reporter fusions in siiA and siiB were constructed using a modification of the ‘one-step inactivation of chromosomal genes’ method (Datsenko and Wanner, 2000) to generate translational reporter fusions in the chromosome (Gerlach et al., 2007c). Further details are given in the supplementary materials and bacterial strains, plasmids and oligonucleotides used in this study are listed in Table S1, Table S2 and Table S3 respectively.
Culture of mammalian cells
MDCK cells were cultured in MEM medium (Earle's salts, glutamine, PAA) supplemented with 10% FCS, 2 mM Glutamax (Invitrogen), non-essential amino acids (PAA), 100 U ml−1 penicillin and 100 μg ml−1 streptomycin (PAA). For invasion, adhesion and binding assays, cells were seeded at a density of 1 × 105 per well in 24-well plates (Cellstar bio-one, Greiner). Growth medium was replaced by fresh medium every second day and cells were allowed to differentiate for 5–6 days. HeLa cells were cultured in DMEM medium (high glucose, glutamate, sodium-pyruvate, PAA) supplemented with 10% FCS and 2 mM Glutamax (Invitrogen). Cells were seeded at a density of 5 × 104 per well in 24–well plates (Cellstar bio-one, Greiner) 24 h before infection assays.
Adhesion and invasion assays
The medium of MDCK cells was changed at least 4 h before the infection with fresh complete cell culture medium without antibiotics after one washing step with sterile PBS. Bacteria were subcultured from an overnight culture (1:31) in fresh LB medium and grown for 3.5 h at 37°C. Bacteria were adjusted to an OD600 of 0.2 in sterile PBS. Bacteria were then diluted in complete cell culture medium without antibiotics to get an moi of 5 and cells were infected with 350 μl of bacterial suspension. Infection was allowed for 25 min at 37°C. To determine adhesion, bacteria were allowed to adhere for 25 min at 37°C. Afterwards, non-adherent bacteria were removed by five washing steps with pre-warmed sterile PBS. To quantify internalization, non-adherent bacteria were removed by three washing steps with pre-warmed PBS and cells were further incubated with complete cell culture medium containing 100 μg ml−1 gentamicin for an additional hour to kill non-invaded extracellular bacteria. Cells were washed once with PBS and lysed by the addition of 500 μl pre-warmed 0.5% sodium deoxycholic acid or 0.1% Triton X-100 in PBS for 5 min at 37°C. Serial dilutions were made in PBS containing 0.05% Tween 80 and plated on Mueller-Hinton agar plates. Adherent or invaded bacteria were calculated as percentage of bacteria used for infection or they were normalized according to the level of invasion of the wild-type strain.
For immunostaining of bacteria, rabbit α Salmonella O-antigen 4, 5 (Difco, BD) was used and a goat α rabbit Alexa488 conjugate (Life Technologies) was used for detection.
For invasion assays with the protonophore carbonyl cyanide 3-chlorophenylhydrazone (CCCP) bacteria were pre-treated with CCCP for the last 30 min of a 3.5 h subculture at 37°C and directly diluted in complete cell culture medium without antibiotics, but containing corresponding CCCP concentrations. Additionally the bacteria were centrifuged onto the cells (500 g, 5 min, RT) to rule out motility effects. To test effects of CCCP on host cells during invasion experiments, HeLa or MDCK cells were incubated with CCCP for 10 min prior to infection. After treatment with CCCP the culture medium was removed and the cells were infected with untreated bacteria without any further washing steps.
As for invasion assays bacteria were subcultured from an overnight culture (1:31) in fresh LB medium and grown for 3.5 h at 37°C. Initial samples were taken at time point 3 h and adjusted to an OD600 of 0.2 in complete cell culture medium before the addition of CCCP (final concentration 100 μM) or DMSO (final concentration 0.1%) as solvent control. After 30 min of CCCP/DMSO treatment 3.5 h samples were adjusted to an OD600 of 0.2 in complete cell culture medium containing CCCP (100 μM) or DMSO (0.1%). In parallel samples were prepared without any incubation with CCCP or DMSO. Diluted bacteria were fixed with 37% formaldehyde solution for 15 min at RT, pelleted (10 000 g, 5 min, RT) and resuspended in blocking solution. After blocking was performed for 20 min at RT the cells were washed once with PBS (10 000 g, 5 min, RT) and then resuspended in 500 μl blocking solution for the following staining. One hundred microlitres of bacterial solution were mixed with 1 μl of SiiE-DyLight488 antibody to stain SiiE-positive cells, incubated for 1 h under rotation in the dark and washed three times with PBS (10 000 g, 5 min, RT). SiiE-positive bacteria were determined under the microscope (40× objective). For this total number of bacteria and SiiE-positive cells of five fields of vision were counted for each strain and condition.
Assays for reporter enzymes, subcellular and membrane fractionation
Activities of alkaline phosphatase and β-galactosidase reporter fusions were assayed according to Maloy et al. (1996). The subcellular fractionation of Salmonella cells was basically performed as described before (Gerlach et al., 2007a). See supplementary materials for detailed descriptions.
Overnight cultures were reinoculated 1:31 into fresh media and grown with aeration for 3.5 h. Cells of 1 ml culture were pelleted (21 000 g, 10 min, 4°C), and resuspended in 50 μl of 0.2 M Tris pH 8.0. To this suspension, 100 μl of 0.2 M Tris pH 8.0, 1 M sucrose; 5 μl protease inhibitor cocktail (Sigma); 10 μl of 10 mM EDTA; 10 μl of 10 mg ml−1 lysozyme (Sigma) in H2O; 315 μl H2O and 5 μl TURBO DNase (Ambion) were added in this order. After incubation for 10 min at room temperature, 500 μl of 3% Elugent (Merck), 20 mM MgCl2 were added to solubilize membrane proteins. Extraction of proteins was allowed to continue for 30 min at 4°C and insoluble material was subsequently removed by centrifugation (21 000 g, 30 min, 4°C). For immunoprecipitation (IP) of SiiA and SiiB, 1 μl of the respective antiserum was added to the soluble fraction and tubes were incubated in a head-over-head rotator for 1 h at room temperature. Protein A agarose (GE) equilibrated in PBS was added to the samples and subsequently incubated in a head-over-head rotator overnight at 4°C. The unbound fraction was separated from the Protein A agarose by centrifugation (2000 g, 3 min, 4°C) and after that the agarose was washed five times with 1 ml of high stringency wash buffer (0.1 mM EDTA, 1% Elugent, 1 M NaCl, 30 mM Tris pH 8.0). Bound proteins were eluted from the Protein A agarose beads by adding 40 μl of 1× SDS-PAGE sample buffer and heating to 95°C for 5 min. Ten microlitres of each sample was analysed by SDS-PAGE electrophoresis and subsequent Western blot.
Bacterial two-hybrid assays (B2H)
B2H assays were essentially carried out as described before (Wille et al., 2012). Briefly, chemical competent cells of the cyaA−E. coli reporter strain BTH101 (Karimova et al., 1998) were freshly co-transformed with 10 ng (each) of two plasmids coding for the T18 or T25 fragments derived from CyaA with or without fusion partners as indicated. After transformation, 3 μl bacterial suspensions were spotted onto LB plates containing 25 μg ml−1 Km, 100 μg ml−1 Cb for selection of both plasmids, 100 μM IPTG (Thermo Scientific, St. Leon-Rot, Germany) to induce protein expression as well as 40 μg ml−1 X-Gal (5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside; Thermo Scientific) as an indicator for β-Gal activity. Plates were incubated at room temperature for 48–72 h, and colonies were documented using a digital camera.
Constructs for parallel tetracycline-inducible expression of CFP and YFP fusion proteins from a single low-copy vector were generated using MultiSite Gateway® recombination (Life Technologies) with custom entry and destination vectors (T. Wille and R.G. Gerlach, unpublished). As donor/acceptor pair for FRET, the CFP derivative SCFP3A was combined with the YFP derivative SYFP2 (Kremers et al., 2006). All constructs were transferred in electro-competent E. coli MACH1™ T1R as well as S. Typhimurium MvP103 (BSL1 strain) cells. Quantification of FRET efficiency was carried out as follows. Bacterial overnight cultures were subcultured into fresh LB broth containing 100 μg ml−1 Cb and incubated to an OD600 of 0.45–0.5 at 37°C. Expression of CFP and YFP fusion proteins was induced by adding 25 ng ml−1 anhydrotetracycline (AHT, Sigma-Aldrich) and growth was continued for another 2 h. The cell suspension was centrifuged (5000 g, 5 min, 4°C), washed twice with phosphate buffer (10 mM potassium phosphate, 67 mM NaCl, 10 mM lactic acid, pH 7.0), incubated for 1 h at 4°C in this phosphate buffer and finally 10-fold concentrated by centrifugation (5000 g, 5 min, 4°C). The concentrated bacterial suspension was used to generate a dense monolayer of cells on 1% agarose pads. Acceptor photobleaching FRET measurements were carried out on a fluorescence microscope derived from the one described before (Kentner and Sourjik, 2009). The set-up is based on an Olympus IX81 inverted fluorescence microscope equipped with a 60× UPlanFLN 0.9 NA objective. Acceptor photobleaching was obtained by illuminating the sample for 25 s using a 100 mW 515 nm laser (Cobolt, Sweden). FRET efficiency (EFRET) was calculated from donor intensity before (FCFP-pre) and after acceptor photobleaching (FCFP-post) using the following formula (Kentner and Sourjik, 2009): EFRET = 100 × (FCFP-post – FCFP-pre)/FCFP-post. FRET efficiency equal or greater than 1% was considered indicative of a positive interaction between the fusion proteins, as all negative controls showed values below 1%.
This work was supported in part by DFG grants HE1964/13-1 and project P4 of the collaborative research centre SFB 944 ‘Physiology and dynamics of cellular microcompartments’ at the University of Osnabrück (to M.H.) and GE2533/1-1 (to R.G.G.), an ERC grant 294761 (to V.S.), and intramural research grant of the RKI to R.G.G. We like to thank Sabine Hunke for antisera against CpxA and Christiane Schmidt for excellent technical support.