HxcQ Liposecretin Is Self-piloted to the Outer Membrane by Its N-terminal Lipid Anchor*

Secretins are an unusual and important class of bacterial outer membrane (OM) proteins. They are involved in the transport of single proteins or macromolecular structures such as pili, needle complexes, and bacteriophages across the OM. Secretins are multimeric ring-shaped structures that form large pores in the OM. The targeting of such macromolecular structures to the OM often requires special assistance, conferred by specific pilotins or pilot proteins. Here, we investigated HxcQ, the OM component of the second Pseudomonas aeruginosa type II secretion system. We found that HxcQ forms high molecular mass structures resistant to heat and SDS, revealing its secretin nature. Interestingly, we showed that HxcQ is a lipoprotein. Construction of a recombinant nonlipidated HxcQ (HxcQnl) revealed that lipidation is essential for HxcQ function. Further phenotypic analysis indicated that HxcQnl accumulates as multimers in the inner membrane of P. aeruginosa, a typical phenotype observed for secretins in the absence of their cognate pilotin. Our observations led us to the conclusion that the lipid anchor of HxcQ plays a pilotin role. The self-piloting of HxcQ to the OM was further confirmed by its correct multimeric OM localization when expressed in the heterologous host Escherichia coli. Altogether, our results reveal an original and unprecedented pathway for secretin transport to the OM.

The presence of an outer membrane (OM) 4 in Gram-negative bacteria constitutes a second barrier for the secretion of exoproteins into the extracellular medium. At least six different secretion pathways have evolved in these bacteria for the secretion of a very diverse pool of extracellular proteins (1)(2). Among them, the type II secretion pathway is a two-step process in which exoproteins with an N-terminal signal peptide (SP) are first exported through the cytoplasmic membrane by either the Sec or Tat translocons. Following removal of the SP, they are released into the periplasm (3)(4). The periplasmic intermediates are specifically recognized by the type II secretion system (T2SS), also called secreton, for their transport across the OM. This pathway, therefore, promotes the specific transport of exoproteins requiring intracellular folding, like periplasmic disulfide bridge formation, and, in some cases, assembly into multimeric complexes prior to their secretion. Such a requirement implies that the secretion process uses a large and tightly controlled secretion channel in the OM. The T2SS is a highly complex nanomachine embedded in the bacterial envelope consisting of 12-16 different proteins, depending on the organism (1,5). Interestingly, there is only one integral OM protein in this system, which therefore constitutes the only candidate for the OM translocation channel. This OM component belongs to a family of proteins generically designated as secretins (6). This family also includes members that are involved in type III protein secretion (T3SS), type IV pilus assembly, type IV bundle-forming pili, toxin co-regulated pili, and assembly and export of filamentous phage (7)(8)(9)(10)(11)(12). Therefore, secretins constitute an important group of transporters specialized in the translocation of bulky macromolecules or macromolecular complexes across the OM.
Several secretins have been purified and analyzed by electron microscopy, revealing that 12-14 identical secretin monomers form ring-like complexes with a central channel large enough to accommodate their substrates (7,(13)(14). The homology between the members of the secretin family is contained within the C-terminal half of the protein (see Fig. 2) (15). Therefore, this domain has been proposed to form the secretion channel, whereas the much less conserved N-terminal domain that largely protrudes into the periplasm probably undertakes more specific functions, such as substrate recognition and/or interaction with the other components of the corresponding machineries (13,16).
Among the large diversity of identified secretins, most of them depend on a small pilot protein for their correct final insertion into the outer membrane. In most cases, pilot proteins are outer membrane-linked lipoproteins called pilotins. To date, characterized secretin/pilotin couples are: PulD/PulS of Klebsiella (17)(18), OutD/OutS of Erwinia (19) for T2SS; YscC/ YscW of Yersinia (8), InvG/InvH of Salmonella (12), MxiD/ MxiM of Shigella (20) for T3SS, and PilQ/Tgl of Myxococcus (21) for Type IV pilus systems. For T2SS secretin/pilotin couples, the specific pilotin binding domain is localized at the extreme C terminus of the secretin (19,22). The majority of the genes encoding pilotins are found in the same cluster as the genes encoding the corresponding secretion systems. However, in several secretin-containing systems, a pilotin gene has yet to be identified, suggesting the existence of possible alternatives to the pilotin biogenesis pathway. Recently, a soluble nonlipidated periplasmic protein has been shown to be important for the OM localization of XcpQ secretin in P. aeruginosa (23). Interestingly, three secretins are themselves lipoproteins, but no function has so far been attributed to their atypical N-terminal lipid anchor. One, XpsD of Xanthomonas campestris pv. campestris, belongs to a T2SS (24), and two others, BfpB of enteropathogenic E. coli (25) and TcpC of Vibrio cholerae (11) are members of type IV pili systems.
In Gram-negative bacteria, most lipoproteins are periplasmic proteins anchored to the inner or outer membrane through a lipid moiety attached to their invariable N-terminal cysteine residue (3). Lipidation and maturation of lipoproteins take place after their translocation through the inner membrane via Sec machinery (3). Lipoprotein-specific signal peptides (SPs) are characterized by a specific consensus motif (V/L)XXC called the Lipobox (26). The Lipobox is both the lipidation site and the maturation site recognized by the lipoprotein signal peptidase II, which cleaves the SP upstream of the cysteine (27)(28).
P. aeruginosa strain PAO1 possesses two complete and nonredundant T2SS, referred to as the Xcp and Hxc systems (1). While more than a dozen exoproteins utilize the Xcp T2SS for their secretion, the Hxc T2SS, which is induced under phosphate starvation, is dedicated to the secretion of one single low molecular mass protein, the alkaline phosphatase LapA (29).
In the present work, we reveal that the atypical HxcQ secretin of the Hxc T2SS of P. aeruginosa is a lipoprotein. Moreover, we demonstrate that the HxcQ liposecretin is self-piloted to the OM via its N-terminal lipid anchor, therefore revealing a new pathway for secretin biogenesis.

EXPERIMENTAL PROCEDURES
Bacterial Strains, Plasmids, and Growth Conditions-The bacterial strains, vectors, and plasmids used in this study are listed in Table 1. Recombinant DNA methods were performed essentially as described previously (30). Oligonucleotides used for PCR are listed Table 2. P. aeruginosa and E. coli strains were grown at 37°C in Luria-Bertani medium. To induce LapA production and secretion via the P. aeruginosa Hxc T2SS, cells were grown at 30°C under phosphate-limiting conditions using proteose peptone medium (Difco Laboratories) containing 0.4% glucose, with horizontal shaking (29). When required, media were supplemented with the following antibiotics used at the indicated concentrations: 50 g⅐ml Ϫ1 kanamycin, 20 g⅐ml Ϫ1 gentamicin; and 50 g⅐ml Ϫ1 ampicillin; 50 g⅐ml Ϫ1 for E. coli and 250 g⅐ml Ϫ1 carbenicillin; 50 g⅐ml Ϫ1 gentamycin; and 2,000 g⅐ml Ϫ1 streptomycin for P. aeruginosa. Bacterial growth was measured by optical density at 600 nm (A 600 ). 1 A 600 unit corresponds to 10 9 cells/ml. The E. coli CC118pir strain was used to propagate pKNG101 and derivative plasmids, while TG1 and the DH5␣ strains were used for other plasmid manipulations. Plasmids were transferred to P. aeruginosa using the conjugative properties of the helper plasmid pRK2013 in triparental matings (31). Transconjugants were selected on Pseudomonas isolation agar (Difco) containing 2.5% glycerol (v/v) supplemented with corresponding antibiotic(s). For classical arabinose induction, bacterial cultures were induced with 0.2% filtered L-arabinose (Sigma) at A 600 0.8 for 2.5 h.
Preparation of Culture Supernatants-P. aeruginosa strains were grown under phosphate-limiting conditions to an A 600 of 1.5. Cells and extracellular medium were separated by centrifugation; proteins contained in the supernatants were precipitated by adding trichloroacetic acid (10% (w/v) final concentration) and incubated overnight at 4°C. Samples were subsequently centrifuged (30 min at 15,000 ϫ g), the pellets were washed with 90% (v/v) acetone, resuspended in SDS-PAGE sample buffer, and analyzed under denaturing conditions.
Inhibition of Lipoprotein Signal Peptidase with Globomycin-Bacteria were grown in Luria-Bertani medium to an A 600 of 0.8, and arabinose at 0.2% final and globomycin at 100 g⅐ml Ϫ1 final were added to the culture, and incubation was continued for 30 min at 37°C. Bacteria were harvested by centrifugation and resuspended in SDS sample buffer, and solubilized proteins were examined by SDS-PAGE and immunoblotting.
[ 3 H]-Palmitic Acid Labeling-Bacteria were grown at 37°C in a liquid minimal medium, proteose peptone 1ϫ (supplemented with 0.4% glucose as carbon source) to an A 600 of 0.3.
Gene expression was induced with 0.2% L-arabinose. At the same time, 50 Ci of [ 3 H]-palmitic acid was added to 1 ml of the culture. Cells were grown for 150 min and collected, and total proteins were analyzed by SDS-PAGE. Gels were then either dried on filter paper and subjected to autoradiography for 5 months at Ϫ80°C and revealed or blotted for HxcQ detection.
Isolation and Separation of P. aeruginosa Membranes by Density Sucrose Gradient Centrifugation-250 A 600 units equivalent of bacterial cells were harvested by centrifugation at 2,000 ϫ g. The pellet was resuspended in 1.5 ml buffer A (10 mM Tris, pH 7.4, 1 mM p-toluenesulfonyl fluoride (Sigma-Aldrich); 10 g⅐ml Ϫ1 DNase and RNase (Roche); and sucrose 20% (w/w)). The cells were passed twice through a French press cell disrupter (Thermo) at 15,000 pressure units using a 3/8-inch diameter piston (20K French pressure cell, AMINCO). Unbroken cells were removed by centrifugation at 4°C for 15 min at 1,600 ϫ g. The supernatant was centrifuged at 4°C for 30 min at 125,000 ϫ g. The crude membrane pellet was resuspended in 0.5 ml buffer M (10 mM Tris, pH 7.4, "Complete EDTA-free" proteases inhibitor mixture (Roche) and 5 mM EDTA) containing 20% (w/w) sucrose and then loaded on top of a discontinuous sucrose gradient consisting of 1.5 ml layers of buffer M solution containing 60 (bottom), 55, 50, 45, 40, 35, and 30% (w/w) sucrose. The membrane separation was performed by centrifugation at 4°C for 65 h at 39,000 rpm in a Beckman SW41 rotor. The gradients were visually checked, and predicted inner membrane (IM) (upper disc) and OM (lower disc) fractions were collected for experiments as presented (Fig. 5A). Fractions were electrophoresed in (i) 11% denaturing SDSpolyacrylamide gel followed by Coomassie Blue staining and visual identification of the OM protein OprF; (ii) 11% denaturing SDS-polyacrylamide gel followed by Western blotting for XcpY detection; (iii) 3.5-9% gradient seminative polyacrylamide gel followed by Western blotting with antibody against HxcQ peptide 28 for detection of multimers of secretins; and (iv) 9% denaturing SDS-polyacrylamide gel followed by Western blotting with antibody against HxcQ peptide 29 for detection of monomers of secretins.
Isolation and Separation of E. coli Membranes by Density Sucrose Gradient Centrifugation-500 A 600 units equivalent of bacterial cells were harvested by centrifugation at 2,000 ϫ g. The pellet was resuspended in 5 ml of buffer B (10 mM Tris, pH 7.4; 1 mM p-toluenesulfonyl fluoride; 10 g⅐ml Ϫ1 DNase and RNase (Roche); sucrose 20% (w/w); and 400 g⅐ml Ϫ1 lysozyme (Euromedex)). The cells were passed twice through a French press cell disrupter (Thermo) at 15,000 pressure units using a 3/8-inch diameter piston (20K French pressure cell, AMINCO). Unbroken cells were removed by centrifugation at 4°C for 15 min at 1,600 ϫ g. The supernatant was centrifuged at 4°C for 30 min at 125,000 ϫ g. The crude membrane pellet was resuspended in 0.5 ml buffer M (10 mM Tris, pH 7.4; "Complete EDTA-free" proteases inhibitor mixture (Roche), and 5 mM EDTA), containing 20% (w/w) sucrose and then loaded on top of a discontinuous sucrose gradient consisting of 1.5 ml layers of buffer M solution containing 60 (bottom), 55, 50, 45, 40, 35, and 30% (w/w) sucrose. The membrane separation was performed by centrifugation at 4°C for 18 h at 39,000 rpm in a Beckman SW41 rotor. The gradient was further collected in 16    (10% (w/v) final) and incubated overnight at 4°C. Samples were subsequently centrifuged (30 min at 15,000 ϫ g), and the pellets washed with 90% acetone. Pellets containing insoluble or precipitated soluble membrane proteins were resuspended in denaturing or seminative SDS-PAGE sample buffer for SDS-PAGE and immunoblotting.
Transmembrane Potential Measurements-P. aeruginosa strains were grown exponentially for 2 h in the presence of 0.2% L-arabinose and harvested by centrifugation at room temperature. The transmembrane potential (⌬) was measured essentially as described previously (34). Briefly, the cell pellet from two A 600 units was resuspended in 100 l of 100 mM Tris-HCl, pH 7.8, and 1 mM EDTA for outer membrane permeabilization and incubated for 3 min at 37°C. The cell suspension was then diluted 20-fold in the transport buffer (100 mM phosphate buffer, pH 7.8, 1 mM KCl, and 0.4% glycerol). A tritiated triphenylphosphonium bromide (Br-TPP) solution ([ 3 H]Br-TPP; Amersham Biosciences; diluted 40-fold in 2 mM cold Br-TPP) was added at the final concentration of 10 M in 200 l of the cell suspension and further incubated at 37°C for 10 min. Cells were recovered by filtration (Whatman) and washed twice with transport buffer and once with transport buffer without glycerol. As a control for nonspecific TPP binding, cell ali-quots were first incubated with 10 M of carbonyl cyanide m-chlorophenylhydrazone for 15 min at room temperature before addition of the Br-TPP solution, incubation, and filtration.

HxcQ Is a Member of the Secretin
Family-Among the different T2SS Hxc proteins encoded by the hxc cluster, HxcQ is predicted to be the secretin component of the system (29). As presented in Fig. 1, HxcQ forms SDS-resistant high molecular mass (HMM) complexes, which has been shown to be a general characteristic of secretins. We observed such HMM complexes for HxcQ secretin when total cell fractions of P. aeruginosa ⌬hxcQ producing a C-terminal V5-hexahistidine tagged HxcQ (HxcQ V5 ) were loaded on a standard SDS-polyacrylamide gel (Fig. 1). However, secretin complexes can show different behaviors in response to heat treatment. For example, HMM complexes formed by PulD or pIV secretins are fully resistant to heat (18,35), whereas HMM complexes formed by XcpQ, BfpB, TcpC, or OutD secretins are totally dissociated after boiling (7, 10 -11, 36). We found that HxcQ multimers are partially heat-resistant even when samples are incubated at up to 95°C for 10 min (Fig. 1, lane 4).
HxcQ secretin encodes an 803-amino acid protein that is 30% identical and 49% similar to XcpQ, a well characterized T2SS secretin of P. aeruginosa. HxcQ primary sequence analysis revealed the typical two subdomains found in XcpQ, the highly conserved C-terminal domain (residues 424 -803) involved in pore formation, and the dissimilar N-terminal domain (residues 81-362), predicted to be periplasmic (Fig.  2). Primary sequence comparison between HxcQ and XcpQ also revealed the presence of two supplemental linker regions on both sides of HxcQ N-terminal domain that are absent in XcpQ (Fig. 2). The region located between the signal peptide and the N-terminal domain is called L1. L1 is 71-amino acids long and is mostly composed of small amino acids such as alanine, serine, and glycine. A comparable linker region is also present in the Xanthomonas campestris XpsD T2SS secretin (Fig. 2). The second linker region, L2, located between the N-and the C-terminal domains, is 62amino acids long and has a composition of 58% serine and glycine. A similar polyserine/glycine region has already been described for OutD and BfpB (16,37) and is also present in XpsD and TcpC secretins (Fig. 2).

FIGURE 3. HxcQ V5 globomycin sensitivity and [ 3 H]-palmitic acid labeling.
Globomycin inhibition of HxcQ V5 maturation. Immunoblotting of total cell proteins from strain PAO1⌬hxcQ/pJN105hxcQ V5 probed with either V5 antibody (top panel) for HxcQ V5 detection or XcpQ antibody (bottom panel) for XcpQ detection. In the absence of globomycin (Ϫ), mature HxcQ V5 monomers as well as HxcQ V5 multimers are observed, whereas in the presence of globomycin (ϩ), the maturation of HxcQ V5 is inhibited leading to the loss of multimers and accumulation of the precursor form of HxcQ V5 monomers. In contrast, mature XcpQ is detected with or without globomycin treatment. A, PAO1⌬hxcQ cells producing HxcQ V5 or HxcQnl V5 from plasmids were labeled with [ 3 H]-palmitic acid. Cell samples were electrophoresed on an 8% stacking/9% running SDS-polyacrylamide gel and radiolabeled lipoproteins were detected by autoradiography. Similar amounts of low molecular mass radiolabeled proteins are detected at the migration front in both samples. Full-length HxcQ V5 protein and molecular mass markers (in kDa) are indicated on the left. The asterisk indicates a specific radiolabeled protein that might correspond to an HxcQ V5 degradation product (B). DECEMBER 4, 2009 • VOLUME 284 • NUMBER 49

Self-targeted HxcQ Liposecretin
HxcQ Secretin Is a Lipoprotein-The comparison of Xcp and Hxc SPs revealed that, in contrast to XcpQ, which has a classical type I SP, HxcQ presents a characteristic type II or lipoprotein SP ending with a typical lipobox (supplemental Fig. S1). This observation was also supported by the lipoprotein prediction program DOLOP (38). To experimentally demonstrate the lipoprotein nature of HxcQ, we treated P. aeruginosa cells with globomycin, a specific lipoprotein signal peptidase II inhibitor (39). As shown in Fig. 3A, the maturation of HxcQ V5 was significantly affected by the globomycin treatment, leading to the accumulation of the precursor form and loss of HxcQ multimers. As a control, we found that XcpQ remained unaffected in agreement with the resistance of signal peptidase I to globomycin. The lipidation of HxcQ was furthermore confirmed by the recovery of radiolabeled HxcQ V5 when the bacteria were grown in the presence of [ 3 H]palmitic acid (Fig. 3B). As a negative control, no radiolabeling was observed for a nonlipidated form of HxcQ V5 called HxcQnl V5 (see below, and for description, see supplemental Fig. S1). A control experiment where proteins from palmitic acid-treated cells were blotted following SDS-PAGE and probed with antibody against the V5 epitope indicated that both HxcQ V5 and HxcQnl V5 were equally produced (data not shown) and that HxcQ V5 did migrate at the position corresponding to the band designated as HxcQ V5 in Fig. 3B. In conclusion, both globomycin treatment and [ 3 H]palmitic acidlabeling assays clearly demonstrated that, in contrast to XcpQ, HxcQ is a lipoprotein. From now on, we will refer to this variant of secretin as liposecretin.
Lipidation of HxcQ Is Essential for Its Function-Given that most secretins, including the P. aeruginosa type II secretin XcpQ, are not lipoproteins (6), we wanted to determine if the N-terminal lipid anchor of HxcQ is required for its function. For this purpose, we constructed a nonlipidated HxcQ V5 variant (HxcQnl V5 ). This construction was made by substituting the type II SP of the HxcQ wild type for the type I SP of XcpQ. To maintain a compatible environment for type I signal peptidase recognition, we also included the four amino acids downstream of the XcpQ SP cleavage site (supplemental Fig. S1).
As both the wild type and non-lipidated HxcQ V5 possess a C-terminal V5-hexahistidine tag, we first tested the influence of the tag on HxcQ V5 function in the quadruple PAO1⌬hxcQ⌬xcpQ⌬pA⌬qA mutant that is deficient in Hxc, Xcp, and hybrid Xcp T2SSs (40, 5) (Fig. 4, lane 4 versus lane 2). The expression of hxcQ V5 from pJNhxcQ V5 in this mutant specifically restored secretion of the unique Hxc substrate LapA in the extracellular medium (Fig. 4, lane 6), indicating a functional complementation and no influence of the V5 tag on HxcQ function. We then tested the functionality of the nonlipidated recombinant HxcQ. Although the amount of HxcQnl V5 produced by P. aeruginosa was similar to that of the lipidated form (data not shown), HxcQnl V5 was unable to restore secretion of LapA (Fig. 4, lane 8). Instead, LapA accumulated within the cells (Fig. 4, lower panel, lane 7), which indicates that the HxcQ N-terminal lipid anchor fulfills an essential secretion function. We constructed a tag-free HxcQnl to definitely exclude a possible effect of the V5 tag in HxcQnl nonfunctionality. We did not observe any phenotypic differences between tagged and untagged HxcQnl variants (data not shown).

The HxcQ Lipid Anchor Has a Pilotin
Function-In order to understand why HxcQnl V5 was not functional, we examined its cellular localization. These studies were carried out in P. aeruginosa PAO1⌬hxcQ producing wild type or nonlipidated HxcQ from plasmids and under arabinose-inducing conditions. Bacterial cells were disrupted and both HxcQ secretins localized in the total membrane fraction (data not shown). In order to investigate the presence of the secretin multimers in the inner membrane (IM) or the OM, total membrane fractions were then separated by centrifugation on a sucrose density gradient. Regions corresponding to the IM and OM (Fig. 5A) were directly sampled from the tube and analyzed. The quality of the fractionation procedure was verified by the presence in the corresponding fractions of the integral IM protein XcpY and the major P. aeruginosa outer membrane protein OprF (Fig. 5B). Interestingly, we clearly detected both wild type and nonlipidated HxcQ multimers. However, whereas HxcQ V5 multimers were correctly localized in the OM (Fig. 5B, lane 2), HxcQnl multimers were mislocalized and accumulated in the IM fraction (Fig. 5B, lane 3). In contrast to wild type HxcQ multimers, multimers of HxcQnl could only be detected under seminative conditions (see experimental procedure). Multimers of HxcQnl indeed appeared to be more sensitive to heat than wild type HxcQ multimers since they could not be detected under classical denaturing conditions (supplemental Fig. S2). We therefore used semi-native conditions for all HxcQnl multimers detection described in this study.
To determine whether IM-recovered HxcQnl multimers were integrated or peripherically associated with the IM, total membrane fractions containing HxcQ V5 or HxcQnl V5 were treated with various solubilizing agents. As shown in figure 6, HxcQnl V5 behaves as an integral IM protein since it remained insoluble upon treatment of the membranes with 100 mM sodium carbonate or 4 M urea (Fig. 6, lane 2 and 6), both known to solubilize only peripheral membrane proteins, such as XcpR.
In contrast, treatment with the nonionic detergent Triton X-100, which typically solubilizes proteins inserted into the IM (XcpY), specifically affected the non-lipidated secretin (Fig. 6, lane 10), indicating its IM insertion. Since no HxcQnl V5 was recovered in the soluble fraction, its solubilisation probably led to its degradation or at least the degradation of the V5 epitope used for HxcQ detection. As a control, the wild type HxcQ V5 was not found to be solubilized by Triton X-100 (Fig.  6, lane 9), which is congruent with its OM localization.
The IM localization of secretin multimers in the absence of a functional pilotin has already been reported for the PulD and YscC secretins, respectively involved in type II and type III secretion (41,8). As shown for PulD, the absence of the pilotin led to partial dissipation of the proton-motive force (pmf) indicative for IM perturbation. This increase in IM permeability was attributed to IM insertion of the mislocalized secretin multimers. Interestingly, we found similar and significant pmf dissipation when Hxc-Qnl V5 was produced in P. aeruginosa ⌬hxcQ (Table 3) and we attribute this effect to the integral IM insertion of HxcQnl V5 multimers. Together, our results show that lipidation of HxcQ V5 is essential for correct localization of the protein in the OM. Moreover, the recovery of HxcQnl V5 multimers inserted in the IM suggests that the N-terminal lipid anchor of HxcQ plays a pilotin role, since such behavior was earlier reported for secretins produced in the absence of their cognate pilotin.
HxcQ Correctly Localized in the OM of the Heterologous Host E. coli-The biogenesis of secretins is impaired in the absence of their cognate pilotins or pilot proteins, and they either remain monomeric or mislocalize to the IM. We tested whether HxcQ liposecretin also requires the assistance of another P. aeruginosa protein for its biogenesis i.e. correct insertion of multimers into the OM. To this end, HxcQ V5 was expressed in the heterologous host E. coli. Membrane samples were fractionated by sucrose density gradient and analyzed by SDS-PAGE. Data presented in Fig. 7 clearly indicates that in such a heterologous environment, HxcQ remains correctly localized in the OM as a multimer. This result contrasts with those observed for other secretins such as PulD which requires co-expression of a pilotin for proper localization in E. coli. This important finding demonstrates for the first time the pilotin-independent biogenesis of a member of the secretin family.

DISCUSSION
Secretins are an unusual and important class of bacterial OM protein involved in various membrane transport pathways such as T2SS and T3SS, type IV pili assembly, and export and assembly of filamentous phage. They form, in the OM, about 1 MDa  . Differential solubilization of membrane-associated HxcQ V5 and HxcQnl V5 . Membrane fractions containing HxcQ V5 or HxcQnl V5 were treated with 2% (v/v) Triton X-100, 100 mM sodium carbonate, pH 11, or 4 M urea at pH 6.5 for differential solubilization. Soluble and insoluble fractions were analyzed under semi-native conditions for HxcQ V5 and HxcQnl V5 multimer detection and under denaturing conditions on a 12% (w/v) SDS-polyacrylamide gel for XcpR and XcpY. HxcQ V5 and HxcQnl V5 were probed with anti-V5 antibody. Molecular masses (in kDa) are indicated on the left. multimeric pore-forming structures that display relatively low ␤-strand content (13) and high resistance to dissociation in SDS (17). Such specialized and complex OM proteins require custom-made biogenesis pathways involving additional partners. Depending on the secretin, different routes and partners have been described (42), but so far no secretin has been shown to be self-transported to its final destination.
In the present work we report on HxcQ liposecretin, the first example of a self-piloted secretin. Interestingly, we showed that the N-terminal lipid anchor of HxcQ which plays a critical role in its biogenesis might compensate for the lack of specific partner and directly participate in the proper targeting of HxcQ to the OM. Altogether our data reveal a new pathway for secretin transport. As proposed earlier, the biogenesis of secretins sometimes requires special assistance conferred by pilotin lipoproteins (17)(18)19). In the case of the T2SS PulD/PulS secretin/pilotin pair, the pilotin binds to the secretin emerging from the IM translocon and either keeps it in a competent state, or prevents its non-productive aggregation, before its insertion into the OM. The pilotin may first maintain the secretin in its monomeric form and, second, assist its transport through the periplasm (41). Here, we demonstrate that the HxcQ N-terminal lipid moiety functions as a pilotin since a nonlipidated version of HxcQ behaves like a secretin in the absence of its cognate pilotin i.e. multimers accumulation in the IM. Given that HxcQ does not need any additional partner for its biogenesis, we propose that HxcQ liposecretin carries an intramolecular pilotin.
In type II secretion, the fatty-acylated pilotin binds the C-terminal domain of the secretin (22) whereas in HxcQ liposecretin, the secretin is fatty-acylated at its extreme N terminus. The C terminus of a secretin is embedded in the OM and is therefore well situated for interacting with a pilotin which is also anchored in the OM. The situation seems more conflicting for the N-terminal domain which needs some flexibility to interact with other periplasmic or inner membrane components (43). The extra glycine/ alanine/serine rich domain between the lipid anchor and the N-terminal domain that we identified in HxcQ (Fig. 2) could give to the N-terminal extremity the flexibility necessary for its function. It is interesting to note that this domain, absent in nonlipidated PulD and XcpQ T2SS secretins, is also present in XpsD (Fig. 2), another T2SS-lipidated secretin.
HxcQ is the fourth secretin described to be a lipoprotein. Previously, BfpB, TcpC and XpsD were experimentally demonstrated to be lipoproteins (10 -11, 24). The involvement of lipidation in secretin biogenesis was only tested for XpsD where this post translational fatty acylation turned out to be dispensable for secretin function (24). For BfpB and TcpC two small nonlipidated periplasmic proteins have been shown to be required for their stabilization and multimerization respectively (25,11). N-terminal lipidation plays a key role for HxcQ transport and no additional specific partner is required. We therefore suggest that among the liposecretins, HxcQ defines a distinctive subclass whose biogenesis is guided by a new and unprecedented transport pathway.
Although the presence of a lipoprotein is often associated with secretin transport, the involvement of the Lol lipoprotein sorting pathway (44) in this process is still an open question. The discovery here that HxcQ is itself a lipoprotein might suggest that the Lol pathway is directly involved; however the Loldependent transport of HxcQ remains to be demonstrated. On the other hand and based on the broad diversity of secretin transport pathways it is also possible that certain secretins might follow an alternative Lol-independent pathway. This is particularly true when looking at XcpQ, another P. aeruginosa secretin. XcpQ is not a lipoprotein and so far, no cognate lipidated pilotin has yet been identified. The situation is also puzzling regarding the implication of the Bam general OM protein assembly machinery in secretin transport (32). Bam dependence was demonstrated for PilQ secretin (32) but invalidated for PulD (45). It would therefore be interesting to experimen- tally test the Bam-and Lol-dependence of HxcQ in order to reveal the contribution of these systems to the biogenesis of this liposecretin. Lol dependence should also be tested for other secretins, although it will be difficult to discriminate between the requirement of secretin and pilotin for Lol.