Type 3 Secretion Translocators Spontaneously Assemble a Hexadecameric Transmembrane Complex*

A type 3 secretion system is used by many bacterial pathogens to inject proteins into eukaryotic cells. Pathogens insert a translocon complex into the target eukaryotic membrane by secreting two proteins known as translocators. How these translocators form a translocon in the lipid bilayer and why both proteins are required remains elusive. Pseudomonas aeruginosa translocators PopB and PopD insert pores into membranes forming homo- or hetero-complexes of undetermined stoichiometry. Single-molecule fluorescence photobleaching experiments revealed that PopD formed mostly hexameric structures in membranes, whereas PopB displayed a bi-modal distribution with 6 and 12 subunits peaks. However, individually the proteins are not functional for effector translocation. We have found that when added together, the translocators formed distinct hetero-complexes containing 8 PopB and 8 PopD molecules. Thus, the interaction between PopB and PopD guide the assembly of a unique hetero-oligomer in membranes.

The transport of proteins across membranes is essential at many stages of pathogen infection and colonization of human cells. This process typically involves the discharge of proteins from the pathogen (secretion) and the introduction of these secreted toxins/effectors into the cytosol of the target cell (translocation). Many pathogens, including the Shigella, Salmonella, Yersinia, and Pseudomonas species, exploit a sophisticated and efficient mechanism of protein secretion and translocation known as type III secretion (T3S) 3 system (1, 2). The T3S system is a syringe-like macromolecular machine formed by more than 20 different proteins organized in three major structures to span: (i) the inner bacterial membrane, the periplasmic space, and the outer bacterial membrane (the secreton); (ii) the extracellular space (the needle); and (iii) the host cellular membrane (the translocon) (3)(4)(5)(6).
A phylogenetic analysis of bacterial T3S systems based on conservation of their basal body ATPase indicates the presence of at least 7 families of T3S machines. The Pseudomonas aeruginosa genome encodes a single T3S system grouped within the Ysc family, named after the Yersinia spp. T3S system (the archetypical T3S system in this family) (7). The Ysc family includes pathogens like Yersinia pestis, Y. pseudotuberculosis, Y. enterocolitica, Bordetella pertussis, Vibrio parahemeolyticus, and P. aeruginosa, among others. The Ysc family shares structural similarity with the Inv-Mxi-Spa family of T3S systems, which includes the secretion systems used by Salmonella enterica and Shigella spp. (8).
Great progress has been made in the structural characterization of the secreton and the needle for different T3S system families (9). However, little is known about how T3S-secreted proteins are translocated across the plasma membrane of the target cell to alter the normal function of the host (4). Two T3S-secreted proteins, known as the T3S translocators, insert into the target membrane to facilitate effector translocation. P. aeruginosa translocators PopB/PopD and the tip forming protein PcrV are functionally conserved with the Yersinia homologues YopB/YopD and LcrV, suggesting a common translocation mechanism. Placing PopB and PopD in a YopB/YopDdeficient Y. pseudotuberculosis strain rescues full translocation of T3S effectors into HeLa cells as well as a T3S-dependent pore-forming phenotype in erythrocytes, if PcrV is also provided (10). Translocators from the Inv-Mxi-Spa are not able to rescue translocation in Yersinia. The sequence identity between the translocators from the Ysc family and Inv-Mxi-Spa family is poor (less than 20%), and the Shigella and Salmonella translocators are considerably larger than the Yersinia or P. aeruginosa counterparts. These differences suggest that the proteins have evolved to adapt to a different mechanism of invasion and/or to interact with different host cell membranes. Therefore, one needs to be cautious when trying to extrapolate the properties observed for translocators from one family to the other.
Current models for the T3S translocon complex of the Ysc family are quite rudimentary, and they are mostly based on the following observations: (i) the translocators are found associated with cell membranes after incubating the pathogen in close contact with red blood cells (11); (ii) both translocators co-immunoprecipitate after Triton X-100 solubilization of membrane-associated proteins (11); and (iii) ring-like struc-tures are observed using EM when the translocators are incubated with model membranes (12). The proteins are presumed to form a pore in which each PopB/YopB is speculated to have two transmembrane (TM) segments and each PopD/YopD only one TM segment based on sequence analysis (Fig. 1A). Both PopB and PopD can insert individually in model membranes and form discrete and stable pores, yet they interact with each other on the membrane when incubated together (13,14). Other elements like coiled-coil segments and amphipathic ␣-helices can be predicted using bioinformatics analysis on PopB and PopD (Fig. 1A), however, how these proteins interact with other proteins or with the membrane is still uncertain.
We have established the experimental conditions necessary to maximize association and insertion of the P. aeruginosa translocators PopB and PopD into model membranes (14). Both proteins form pores in model membranes and allow the passage of small proteins and other molecules (13,14). We sought here to provide specific insights about the stoichiometric arrangement and mechanism of assembly of the PopB and PopD translocators. Steady-state fluorescence spectroscopy of membrane-bound complexes indicated that when added together, PopB and PopD assemble hetero-complexes. Using single-molecule fluorescence photobleaching, we quantified the stoichiometry of membrane-assembled complexes. In the absence of PopD, PopB oligomers presented a bi-modal size distribution with peaks at 6 and 12 subunits. In the absence of PopB, PopD assembled mostly hexameric structures. Strikingly, when PopB and PopD were both present, they assembled into complexes containing 8 PopB and 8 PopD molecules. These findings provide the basis to explain the requirement of both translocators to assemble a functional TM hetero-complex with unique architecture.

Experimental Procedures
Protein Expression, Purification, and Characterization-The purification, structural characterization, and pore forming properties of isolated PopB and PopD were described before in detail (14). The expression and purification of hisPcrH-PopD, hisPcrH-PopB, their derivatives, and the isolation of the translocators from hisPcrH chaperone were done as previously described (14). The protein concentration was estimated using molar absorptivity of 18,910 M Ϫ1 cm Ϫ1 for hisPcrH, 13,980 M Ϫ1 cm Ϫ1 for PopD, 6,990 M Ϫ1 cm Ϫ1 and for PopB as described (14). Single Cys protein derivatives were obtained by site-directed mutagenesis. The functionality of the protein derivatives employed in these studies was evaluated by their ability to form pores in model membranes (14) and their ability to rescue translocation of P. aeruginosa PAK strain effectors into HeLa cells (15). PopB S164C or PopD F223C derivatives were able to rescue effector translocation in P. aeruginosa PAK strains with deletions for PopB (PAK⌬popB) or PopD (PAK⌬popD), respectively (Fig. 1). PAK⌬popB and PAK⌬popD deletion strains were generously provided by Dr. Stephen Lory. A synthetic 1494-bp DNA fragment (named GHD) encoding the P. aeruginosa PAK (Gene Bank AY232997.1) pcrG promoter (400 -475), pcrH (1663-2179), and popD (3331-4231) was acquired from Biomatik. The construct was provided in the pUC57-Amp vector with an EcoRI site and a HindIII site at its 5Ј and 3Ј, respectively. The GHD fragment was cut by double digestion with EcoRI and HindIII and cloned into the pUCP18 vector (generously provided by Dr. Dara Frank 16,17) to generate the pUCPHD vector. The pUCPHB vector was generated by the Gibson assembly method (18) as instructed by the manufacturer (New England Biolabs) ( Table 1). The infection assay was adapted from Kaufman et al. (15). Briefly, HeLa cells cultured in DMEM (Caisson) containing 10% FBS were maintained in a 37°C incubator supplemented with 5% CO 2 . Before infection, cells were washed once with pre-warmed PBS and incubated in fresh DMEM. Bacteria were grown overnight at 37°C in Miller lysogeny broth, the next day the cells were diluted to an absorbance at 600 nm of 0.1 in fresh broth and grown until the absorbance reached a value of 1. To establish infection a bacterium/HeLa cell ratio of 30 was used. Bacteria were spun down at 16,000 ϫ g for 2 min and suspended in 100 l of DMEM, then incubated with HeLa cells grown to confluence in 6-well plates for 4 h at 37°C (1.2 ϫ 10 6 cells/well). The pore-forming activity of labeled derivatives was indistinguishable from the one observed for WT translocators (14).
Fluorescent Protein Labeling-PopB S164C and PopD F223C were labeled using the iodoacetamide derivative of BodipyFL (BpyFL C 1 -IA, Invitrogen) as previously described (14). PopD F223C was labeled using Bodipy TMR C 5 -Maleimide (BpyTMR, Invitrogen) using the same described procedure (14). Labeling efficiencies were calculated using the molar absorptivities at 280 nm in urea 6 M for PopB and PopD (see above), at 502 nm for BpyFL (55,000 M Ϫ1 cm Ϫ1 ), and at 540 nm for BpyTMR (51,120 M Ϫ1 cm Ϫ1 ). Protein concentration was calculated after correcting the absorbance by the contribution of the fluorescent dyes at 280 nm. The absorbance of BpyFL at 280 nm was 4% of its absorbance at 502 nm. The absorbance of BpyTMR at 280 nm was 20% of its absorbance at 540 nm. The percentage of labeling of selected preparations used for single molecule experiments were 96% for PopD F223C-BpyFL , 100% for PopB S164C-BpyFL , and 95% for PopD F223C-BpyTMR (hereafter PopD BpyFL , PopB BpyFL , and PopD BpyTMR , respectively). For single-molecule experiments the labeled proteins were kept in the dark and all incubations were done in the absence of light to avoid photobleaching. To avoid bleaching of protein complexes during focusing of the sample prior to a photobleaching experiment, the focal plane containing single TM complexes was

PCR primers used to generate the pUCPHD and pUCPHB vectors
The pUCPHB vector was generated using Gibson assembly combining the Cys-less P. aeruginosa PA01 popB gene amplified from plasmid pETDuet1-hispcrHpopB (14) using primers 1 and 2; a fragment from GHD containing the pcrG promoter and pcrH gene amplified using primers 3 and 4; and the pUCP18 vector linearized by double digestion with EcoRI and HindIII. Vectors containing the genes codifying for PopD F223C or PopB S164C were generated similarly using Gibson assembly with the linearized pUCP18 vector and fragments obtained from pUCPHD or pUCPHB with primers 3 and 7, 6 and 5, or 3 and 9, 8 and 2, respectively.

Primer sequences
5Ј-CAGACCGCACTTCTGGGATTCTTTCGC found manually and a Perfect Focus System (PFS, Nikon) was activated. This system monitors the position of the focal plane using a dedicated near infrared 870-nm light-emitting diode and line charge-coupled device sensor that adjusts focus automatically in milliseconds given a drift event in the Z-axis. After focusing and perfect focus system activation, laser excitation was turned off, the objective lens was moved to a new region never exposed to light, and the single-molecule photobleaching experiment was initiated acquiring the fluorescence emission from time 0 without requiring prefocusing of the sample. Liposome Preparation-Liposomes were prepared using identical composition and procedures previously described (14). NBD-PE (N-(7-nitrobenz-2-oxa-1,3-diazol-4-yl)-1,2-dihexadecanoyl-sn-glycero-3-phosphoethanolamine) was purchased from Life Technologies.
Liposome Flotation-Membrane Binding Assay-Association of PopB and PopD to liposomes at the indicated pH was assessed by flotation of proteoliposomes through a sucrose gradient, SDS-PAGE, and quantification of the bound proteins by gel densitometry as described previously (14). When added together, PopB and PopD were pre-mixed in 6 M urea at the indicated ratio.
Supported Lipid Bilayers (SLB)-Clean glass coverslips were prepared by boiling coverslips in a 10% (v/v) solution of 7X detergent (MP Biomedicals) for 20 min. Coverslips were then extensively washed with tap water, followed by extensive wash with pure water, followed by 10 min boiling in pure water. Dried coverslips were heated for 4 h at 400°C in a kiln oven and stored until use. SLB were formed as follows: a sample containing 400 l of liposome or proteoliposome (50 mM sodium acetate buffer, pH 4.0, 0.22 mM final lipid concentration, and 5 mM CaCl 2 ) was placed on a Petri dish between two 1-mm thick aluminum spacers. A clean glass coverslip was placed on top of the sample and incubated for 30 min to allow SLB formation. After this, a glass slide was placed next to the coverslip and the Petri dish was filled with buffer. While submerged in buffer, the coverslip was mounted on the glass slide spaced by two strips of double-sided adhesive tape to form a chamber.
Assembly of Protein Complexes for Imaging-For homo-oligomer reaction mixtures (100 l) containing liposomes (0.1 mM total lipids) and BpyFL-labeled PopB or PopD (30 nM total protein) were incubated in buffer, 50 mM sodium acetate, pH 4.0, at 20 -23°C for 15 min (protein to lipid ratio 1/3,333). For dual labeled hetero-oligomers reaction mixtures were done similarly but total lipid concentrations were 0.3 or 0.6 mM and labeled proteins were 30 nM each (total protein to lipid ratio 1/5,000 or 1/10,000). Before assembly into SLB, proteoliposomes were diluted by addition of 1390 l of buffer, 50 mM sodium acetate, pH 4.0, plus 10 l of a solution of liposomes (30 mM total lipids stock concentration). For single-color labeled hetero-oligomers reaction were done similarly as for dual-color labeled ones, maintaining a total protein to lipid ratio of 1/4,550.
Imaging-Single molecule images were acquired using an electron multiplier charge-coupled device Cascade II camera (Photometrics, Tucson, AZ). Total internal reflection fluores-cence (TIRF) microscopy used a home-built laser system around a Nikon Eclipse Ti microscope and a high numerical aperture objective (ϫ60, NA 1.49) (Nikon, Melville, NY). TIRF excitation was achieved using a 488-nm argon-ion air laser (Melles-Griot) for PopD BpyFL and PopB BpyFL , and a 532-nm solid state (Crystalaser) laser for PopD BpyTMR . Neutral density filters were employed to attenuate laser light to the desired level and increase the average photobleaching time. The standard exposure and acquisition interval time was 70 ms for BpyFL and 100 ms for BpyTMR probes. Dual-color photobleaching was performed first by constant illumination from a 532-nm laser line to photobleach the red emitting dye BpyTMR, followed by constant illumination with a 488-nm laser line to photobleach the green emitting BpyFL. The consecutive photobleaching of BpyTMR followed by BpyFL was done to minimize any problems associated with potential FRET between the dyes and any direct excitation or photobleaching of BpyTMR when using the 488-nm wavelength. Time point image stacks were independently acquired for PopD BpyTMR and PopB BpyFL , and emission intensity as a function of the illumination time was analyzed for individual protein complexes.
Single-molecule Photobleaching-Photobleaching of labeled translocators was performed on SLB assembled using liposomes with membrane-inserted proteins. Photobleaching data were inspected and analyzed manually. Diffraction-limited spots in image stacks were identified and quantified over time using ImageJ. Time-resolved fluorescence intensity plots of single protein complexes were generated importing the measured intensities into Origin software. The number of fluorescent molecules in each complex was obtained by dividing the background-subtracted initial fluorescence intensity by the intensity contribution of a single dye revealed by single stepwise photobleaching events, most frequently discernible within the last few photobleaching events within each trace (19,20). Single monomers were not considered in our analysis given the difficulty of distinguish them from background noise. Because ϳ100% of the proteins were labeled, and we were able to temporally separate the bleaching of a single fluorophore, the overall photobleaching corresponded to the presence of all the subunits of the complex. This is the ideal situation for photobleaching enabling each subunit to be directly counted without the need to implement discrete statistical distributions to deduce the number of molecules in the complex.
Steady-state Fluorescence Spectroscopy-Steady-state fluorescence measurements and spectra for fluorophore characterization and quantification were collected using a Fluorolog 3-21 spectrofluorometer as described earlier (14). Anisotropy and emission intensity measurements were taken using a Chronos fluorescence lifetime spectrometer (ISS, Champaign, IL) equipped with a 470-nm laser diode and Glan-Thompson prism polarizers (10 ϫ 10 mm aperture for excitation and 14 ϫ 14 mm aperture for emission) (14). Emitted light was collected through a Melles Griot cutoff glass filter GG495 to eliminate scattered light. All measurements were blank subtracted with an equivalent sample lacking the fluorophore. For emission intensity measurements the amount of PopD BpyFL was kept constant (final concentration 10 nM) and the amount of WT PopD (PopD WT ) or PopB WT adjusted to render the indicated percentage of labeled protein. The protein:lipid ratio was kept constant at 1/3,000 or 1/10,000 by adjusting the amount of lipids added. For anisotropy measurements total protein concentration was kept constant at 100 nM and total lipid concentration at 1 mM. The percentage of labeling was adjusted by premixing PopD BpyFL with PopD WT (or PopB WT ) in 6 M urea. BpyFL was excited with polarized light, and both vertically polarized (ʈ) and horizontally polarized (Ќ) emission intensities were recorded. Anisotropy was calculated using r ϭ (Iʈ Ϫ I Ќ )/ (Iʈ ϩ 2I Ќ ), where I is the background-subtracted emission intensity of vertically (or horizontally) polarized light. While working with membrane bilayers it is important to select experimental conditions where the anisotropy (homoFRET) signal originates from the association of proteins, and not as a result of the proximity of randomly distributed proteins. Based on determinations of liposome size by electron microscopy (14), and the protein:lipid ratio used in these studies, we estimated that randomly distributed proteins will be on average more than 200 Å apart in our system.
Time-resolved Fluorescence Spectroscopy-Lifetime measurements were done using a Chronos fluorescence lifetime spectrometer (ISS, Champaign, IL) as described previously (14).

PopB Assists the Association of PopD with Membranes-Both
PopB and PopD can efficiently form stable pores at acidic pH on liposomes containing negatively charge phospholipids (12). In addition to their individual ability to insert into membranes, we have shown that the translocators can also interact with each other and form hetero-oligomers that span the bilayer (14). To examine the molecular mechanism of PopB-PopD interactions, we investigated the association of the translocators when simultaneously binding to membranes over a broad pH range, expanding from the optimal acidic pH to neutral conditions. A single Cys residue was introduced into each translocator at a location predicted to be solvent-exposed (Fig. 1A), and specifically labeled with the fluorescent probe BpyFL. Labeled PopB and PopD were incubated with membranes at the indicated pH, and the association of proteins with lipid bilayers was determined using a liposome flotation assay. The membranecontaining fraction was analyzed by SDS-PAGE and membrane-bound proteins were quantified using gel densitometry ( Fig. 2A). In this assay only liposomes and proteoliposomes float to the top of the gradient, whereas free proteins or protein aggregates remain in the bottom fraction. To detect any coop-  Location of the single Cys modifications introduced for probe labeling is indicated on PopD and PopB. Predicted hydrophobic segments, amphipathic helices, and coiled coils are shown. The molecular mass for each translocator is shown on the right. The scale at the bottom indicates amino acid number. The activity of the PopB (or PopD) derivative was indistinguishable from the wild-type protein when evaluated for their ability to complement a popB deletion strain (or a popD deletion strain). The activity of the translocators was assessed by the characteristic rounding up of the infected target cells due to actin cytoskeleton disruption caused by effector translocation (15). Strains used in this assay were P. aeruginosa PAK WT (B), PAK⌬popD complemented with pUCP18 (C), PAK⌬popB complemented with pUCPHB S164C (D), and PAK⌬popD complemented with pUCPHD F223C (E). The PAK⌬popB control showed no effect on cell morphology as shown for the PAK⌬popD control in panel C (not shown).
erative interaction between PopB and PopD, their membrane association was first quantified for individual translocators and then when the two translocators were added together. Any difference in the binding properties of PopB or PopD when incubated alone or in combination suggest a potential interaction between the translocators.
Membrane association for PopD was higher than 90% at pH 4.0, 50% around pH 5.3, and minimal at neutral pH (Fig. 2B). Interestingly, when PopD and PopB were mixed in equimolar amounts prior to incubation with membranes, the binding of PopD was significantly enhanced in the 5.2-5.8 pH range. PopB membrane association was higher than 90% at pH 4.8, 50% around pH 5.8, and less than 30% at neutral pH. PopB membrane binding was not significantly affected by the presence of PopD (Fig. 2C).
These results suggested that the simultaneous addition of PopB and PopD influences how the proteins interact with membranes, and we hypothesized that this affects their final membrane-inserted arrangement. We therefore focused our analysis on the ability of each translocator to form oligomers on membranes when added individually or combined. We characterize the assembly of membrane complexes at pH 4.5, where binding to membranes reached saturation.
PopD and PopB Form Multimers on Lipid Membranes-FRET or homoFRET (FRET between like molecules), as a function of fractional fluorescent labeling, has been used to determine the oligomerization state of small protein complexes (21). When homoFRET occurs, energy can migrate from dye to dye before a photon is emitted, thereby depolarizing the light and reducing the anisotropy of the sample. How much depolarization occurs depend, among other parameters, on the distance among dyes. Therefore, the steady-state fluorescence anisotropy signal reports on the efficiency of homoFRET (i.e. the distance) between protein molecules labeled with the same fluorophore. In addition to homoFRET, the fluorescence anisotropy can also be influenced by changes in the rotational diffusion of the labeled protein. However, this source of ambiguity is overcome by analyzing anisotropy as a function of the molar fraction for the labeled protein when diluted with unlabeled protein. Under these conditions, changes in fluorescence anisotropy are due to differential separation between the fluorescent probes and reflect on the size of the protein oligomer (21). Both homo-and hetero-oligomerization can be assessed using this approach (22).
In addition to the relatively long range distance information provided by homoFRET measurements (R 0 ϳ50 Å), BpyFL dyes can form non-fluorescent dimers when positioned a few angstroms apart, which is easily detected by a decrease in the fluorescence intensity, and reports on molecular interactions within a shorter distance range compared with homoFRET (23). BpyFL self-quenching is indicative of adjacent subunits because orbital contact between fluorophores is required (24), and this effect has been exploited in various analytical methods, like biotin-avidin binding detection (25) and single molecule folding studies (26). Moreover, BpyFL is especially useful to study protein-membrane interactions because its fluorescence properties are not sensitive to the pH or polarity of the environment (27,28).
The ability of PopD to interact with other PopD molecules or with PopB prompted us to investigate the nature of the oligomers formed by these proteins. For PopD-PopD interactions, both the anisotropy (homoFRET) and fluorescence intensity (self-quenching) were measured as a function of the fraction of labeled PopD BpyFL present, which was controlled by addition of PopD WT . Although keeping the total amount of fluorophore constant, an increase in the emission (de-quenching) was observed as the fractional amount of PopD BpyFL was reduced, showing the amount of purified PopD BpyFL and/or PopB BpyFL isolated in the membrane containing fraction after bound and unbound proteins were separated using a liposome flotation membrane-binding assay as described in the text. Total lipids were 2 mM and total protein was 400 nM. Individual proteins or pre-mixed proteins were incubated with membranes at 20 -23°C for 1 h before ultracentrifugation using a sucrose gradient. Proteins were visualized using BpyFL fluorescence. B, PopD BpyFL binding at the indicated pH when the translocator was incubated with membranes alone or when premixed and incubated with an equimolar amount of PopB WT . C, PopB BpyFL binding at the indicated pH when the translocator was incubated with membranes alone or when premixed and incubated with an equimolar amount of Pop-D WT . BpyFL fluorescence was quantified by gel densitometry and each data point represents the average of at least two independent assays and error bars indicate the data range. Average lines to individual translocators (dashed) and both translocators (solid) are shown only as a guide for the reader.
clearly indicating that PopD formed oligomers on membranes ( Fig. 3A and Table 2). At the same time, the homoFRET between subunits decreased (i.e. increase in fluorescence anisotropy) when the fractional amount of PopD BpyFL was reduced.
The interaction of PopD with PopB was explored using a similar approach. While keeping a constant amount of PopD BpyFL , PopB WT was added in increasing amounts. Emission measurements revealed that PopD BpyFL associated with PopB as efficiently as with PopD, because the same final recovery of the fluorescent intensity was observed when the molar fraction of PopD BpyFL was decreased ( Fig. 3B and Table 2). In contrast, very little decrease in homo-FRET (i.e. increase in fluorescence anisotropy) took place when the molar fraction of PopD BpyFL was reduced using PopB WT . These results suggested that PopB was able to intercalate between PopD subunits (as shown by the decrease in the BpyFL self-quenching), but the longer range homoFRET still occurred among subunits in a complex. Given that PopD BpyFL efficiently interacts with PopB, the lack of increase in anisotropy was indicative of the presence of at least two or more PopD BpyFL subunits per PopB/PopD hetero-complex. Taken together, these results showed that PopD assembles homo-oligomeric structures in membranes, but when added together with PopB it forms hetero-oligomeric structures of yet undetermined size and stoichiometry.
Individual Translocators Adopt Mostly Hexameric Arrangements When Assembled into Lipid Bilayers-To effectively determine the stoichiometry of membrane-inserted oligomers and the heterogeneity of the samples, we applied single-molecule fluorescence photobleaching. This technique allows for quantification of the number of protein molecules within a protein complex (19,29). In practice, fluorescence detection with single-molecule efficiency allows visualization of single photobleaching events as a stepwise decrease in the fluorescence signal. The high sensitivity of this technique combined with the spatial resolution of optical microscopy imaging of single protein complexes, allowed us to quantify the number of photobleaching events occurring in a single membrane inserted oligomer (Fig. 4). The number of photobleaching events is equivalent to the number of fluorophores attached to a protein complex, or in the case of single-labeled polypeptides, to the number of proteins forming part of the complex. We adapted the technique to quantify proteins in their native lipid bilayer environment (i.e. in the absence of detergents) by using SLB (30).
PopB BpyFL or PopD BpyFL were reconstituted into liposomes using previously optimized conditions for membrane association and pore formation (14). The resulting proteoliposomes were diluted with excess liposomes to achieve the optimal low protein density required for single-particle imaging. Diluted samples were used to assemble SLB on glass coverslips by the vesicle fusion method (31). Bilayers were imaged using singlemolecule TIRF microscopy as detailed under "Experimental Procedures." Assembly of a continuous and fluid lipid bilayer was confirmed in a parallel experiment by including a fluorescently labeled phospholipid in the membranes, NBD-PE, and observing fluorescence recovery after photobleaching of a membrane patch. Fluorescence recovery after photobleaching under our experimental conditions occurred in a time scale consistent with phospholipid diffusion in a fluid lipid bilayer (calculated 10 Ϫ12 -10 Ϫ11 m 2 /s lateral diffusion coefficient for NBD-PE, data not shown) (32). In contrast to the freely diffusing lipid molecules, protein complexes remained immobile during the time scale of visualization before complete photobleaching of the attached probes (no significant drifting of the fluorescence complex was observed on membranes). This behavior has been observed for integral membrane proteins inserted into SLB, presumably due to interaction or absorption of the protein to the glass surface caused by protein segments facing the glass side of the membrane (33).
A quantitative analysis of photobleaching steps from 100 individual PopD oligomers and more than 150 individual PopB oligomers was employed to determine the stoichiometry of complexes formed by PopD (or PopB) when assembled into membranes. More than 70% of the total PopD translocators were found in complexes formed by 6 Ϯ 1 subunits, indicating that PopD BpyFL assembled mostly hexameric structures when incubated alone with membranes (Fig. 4C). For PopB, a broader size distribution was observed dominated by hexameric and dodecameric oligomers (Fig. 4F).
PopB and PopD Assemble Hexadecameric Hetero-complexes in Membranes-To effectively determine the stoichiometry of hetero-oligomers we applied a technique that allows for quantification of the number of different protein molecules within a single protein complex (19). Dual-color fluorescence photobleaching was used to simultaneously determine the number of

lifetimes of BpyFL in homo-and hetero-oligomers at varying percent of the fluorescent label
Proteoliposome preparations containing the same protein and lipid concentration, while varying the molar fraction of PopD BpyFL and either PopD or PopB (see Fig. 3), were analyzed using frequency-domain fluorescence spectroscopy as described previously (14). Phase-Delay and Modulation-Ratio data points for each individual sample were fit to a double-exponential decay model. Addition of a third exponential decay component did not significantly improved the fit. The table summarizes the Intensity weighted (͗͘ I ) and Amplitude weighted (͗͘ A ) average fluorescence lifetimes derived from each individual best fit (in nanosecond time units), as well as the reduced chi-squared ( 2 ) goodness of fit test parameter. BpyFL in sample  100  90  80  70  60  50  40  30  20   PopB and PopD molecules present in membrane-assembled oligomers. Each translocator was quantified based on the unique properties of the different fluorescence dyes covalently attached to them. PopD F223C was labeled at ϳ100% efficiency with the red-emitting fluorescent probe BpyTMR, whereas PopB S164C was labeled with same efficiency using the green emitting fluorescent probe BpyFL. Control experiments showed that no labeling occurred on WT translocators (which lack Cys residues) when they were incubated with labeling reagents (14). PopB BpyFl and PopD BpyTMR were mixed and reconstituted into liposomes, and SLB were prepared as described above. This technique permitted the analysis of hetero-complexes regardless of the presence of co-existing homo-oligomers of PopB and PopD, as only hetero-complexes display co-localization of green (PopB) and red (PopD) fluorescence (Fig. 5A, yellow  spots).

% PopD
A quantitative analysis of the time-dependent photobleaching for more than 220 individual protein complexes showing yellow spots was used to determine the stoichiometry of oligomers formed by both PopD and PopB (Fig. 5B). Remarkably, we found that the distribution of PopB⅐PopD complexes centered at 8 photobleaching counts for PopD BpyTMR , as well as 8 photobleaching counts for PopB BpyFL (Fig. 5C), constituting an hexadecameric complex. This stoichiometry was independent of the total amount of protein distributed on the vesicles because identical results were obtained when the experiment was done using a 2-fold higher protein:lipid ratio (1/10,000, data not shown).
Individual PopD and PopB homo-oligomers were still observed under these experimental conditions (when the PopB: PopD ratio was 1/1), but it was clear that the proteins arranged into a different oligomeric structure when interacting with each other. The ability of the proteins to form both homo-and hetero-oligomers prompted us to analyze the reversibility of formed complexes. We therefore analyzed if the formation of PopD homo-oligomers was reversible after addition of PopB taking advantage of the PopD Bpy self quenching properties. As showed in Fig. 3B, the emission of PopD Bpy was low when forming homo-oligomers (Bpy self quenching), but increased when PopD Bpy was pre-mixed with an excess of PopB before their addition to the membranes. However, if excess PopB was added after the PopD Bpy homo-oligomers were formed, no change in the emission of Bpy was observed (Fig. 6A). This suggested that after assembling a homo-oligomer, the PopD subunits cannot dissociate and form a hetero-oligomer.
Because the assembly of PopD homo-oligomers is not reversible, how much of each complex would be formed in this model system is dictated by the relative rates of homoversus heterooligomerization. By increasing the PopB/PopD ratio from 1:1 to 10:1 in the reconstitution assay, we were able to optimize the incorporation of PopD into hetero-complexes (Figs. 3B and 6B). Taken together, these data showed that although PopD is able to assemble stable hexamers on membranes, an early interaction event between PopB and PopD along the assembly path-way is required to secure the formation of a hexadecameric hetero-complex.

Discussion
Our spectroscopic analysis of reconstituted membrane protein complexes has provided important insights into the structural organization and assembly of the complex formed by the T3S translocators PopB and PopD. First, PopB facilitated the association of PopD with membranes. Second, in the absence of PopB, PopD formed mostly hexameric membrane complexes. Third, the interaction of PopB with PopD drove the formation of membrane-bound hetero-oligomer with a stoichiometry that differed from the one adopted by isolated proteins in homo-oligomers. Fourth, single molecule dual color photobleaching determinations showed that hetero-oligomers contained 8 PopB and 8 PopD subunits. Therefore, the interaction of PopB with PopD during translocon assembly was critical to chaperone the formation of membrane-inserted hetero-complexes.
Negatively charged lipids and acidic pH have been often used to optimize the interaction of pore-forming toxins and other proteins with membranes in vitro (34 -37). In some cases, acidic pH can induce a molten globular state in proteins due to changes on charged amino acids. This relaxed form may facilitate membrane insertion (38). The exposure of hydrophobic surfaces by molten globular states at neutral pH favor protein aggregation in solution (as in the case of PopB and PopD (12,13,39)). However, the increase of the net positive charge of the protein at low pH would increase the electrostatic repulsion between proteins monomers and favor the electrostatic interaction with negatively charged membranes. It is therefore not surprising that the association of purified PopB and PopD (i.e. in the absence of the needle complex) to model membranes requires acidic pH and is facilitated by the presence of anionic lipids (13,14).
The hydrophobic character of PopB and PopD and their propensity to aggregate in aqueous solutions demand a robust, reproducible, and controllable experimental system to study the structural arrangement of the T3S translocon. Unfoldingdependent membrane insertion is a reasonable approach because PopB and PopD are expected to be secreted unfolded through the narrow conduit of the T3S needle.
We therefore optimized the reconstitution of PopB and PopD membrane complexes by spontaneous refolding of ureasolubilized proteins in the presence of membranes (14). This folding-based reconstitution system allowed the study of individual WT translocators and to detect any difference in their properties that may result from the interaction between them. As shown in Fig. 2, when individually incubated with membranes, optimal binding for both PopB and PopD required pH lower than 5, however, PopB assisted the binding of PopD in a narrow pH range (i.e. pH [5][6]. The mechanistic details that govern the interaction between PopD and PopB are far from being understood. Mostly because no high resolution information is available on these proteins. However, the interaction between PopB and PopD was independently confirmed in this work using two spectroscopic properties of the fluorescent dye BpyFL, emission self-quench-   (PopD only). Center bar, PopD BpyFL was mixed in 6 M urea buffer with a 10-fold excess of PopB WT and the mixture was added to membranes. Excess PopB favored the incorporation of PopD into hetero-oligomers, therefore it de-quenches the emission of the BpyFl probe (see also Fig. 3B). Right bar, PopD BpyFL was incubated with membranes to allow homo-oligomer formation and a 10-fold excess of PopB was subsequently added. No emission de-quenching was observed in this case, indicating that the simultaneous interaction with membranes of both translocators is necessary to form hetero-oligomers. A constant amount of PopD BpyFL (83% labeled with BpyFL) was present in every sample at a final concentration of 10 nM. The protein to lipid ratio was 1/6000 in all samples. Error bars indicate the range between two measurements. B, PopD size distribution shifts from 6 to 8 subunits per complex when membrane incubation is done in the presence of PopB. PopD BpyFL was incubated with membranes in the presence of 10-fold excess PopB whereas maintaining a constant protein:lipid ratio (black bars). Size distributions were obtained using single-molecule photobleaching and compared with the distribution obtained when PopD BpyFL was incubated alone with membranes (Fig. 4C, reproduced in gray bars). Lines indicate Gaussian fits to the histogram data. Oligomers are represented as coded in Fig. 3. ing, and energy migration. Using these complementary approaches, we found that PopB formed hetero-oligomers with PopD BpyFL (Fig. 3B), and these complexes contained several PopD BpyFL subunits even when a high PopB:PopD BpyFL ratio was used. These observations suggest that the formation of a hetero-dimer may re-direct the assembly and architecture of the TM complexes.
We therefore employed TIRF microscopy and single molecule photobleaching (30,40,41) to determine the stoichiometry of homo-and hetero-oligomers formed by PopB and PopD inserted in membranes. These experiments clearly showed that a unique hetero-oligomer is formed when PopB and PopD are added together. These data are in good agreement with results showing that both translocators are required for effector translocation, and that both translocators have to be secreted from the same bacterium for translocation to take place (11). When secreted separately, the encounter between PopB and PopD may take too long and homo-oligomerization dominates under these conditions. Hetero-oligomers formed by PopB and PopD contained mostly 8 PopB and 8 PopD molecules (Fig. 5C), indicating that the presence of both translocators in a particle was not simply a random distribution of translocators or a co-localization of individual homo-oligomers (which are mostly hexameric, Fig.  4). It is worth noting that a translocon complex containing 8 PopD molecules and 8 PopB molecules will have a molecular mass of 570 kDa (or 601 kDa for a YopD 8 YopB 8 , see Fig. 1). Therefore, the molecular mass obtained using this model system is in good agreement with the molecular mass roughly estimated for Yersinia translocons isolated from erythrocyte membranes using blue native gel electrophoresis (600 Ϯ 100 kDa) (42). Taken together, single molecule fluorescence and ensemble experiments suggest a model where the interaction of PopD with PopB, presumably a dimer formation, leads to the assembly of unique hetero-oligomeric structures with defined stoichiometry on lipid bilayers. We can speculate that the sequential secretion of PopB and PopD in the proximity of the target membrane would favor the formation of hetero-oligomers in the proximity of the needle tip. The tip may facilitate the PopB-PopD association and chaperone their insertion into the membrane (4,43,44). Unfortunately, the hierarchical mechanism of secretion for the translocators is far from being understood.
Some advances have been made on the assembly of the Shigella counterparts IpaD, IpaB, and IpaC (equivalents to PcrV, PopB, and PopD in P. aeruginosa, respectively (45)(46)(47)). However, as mentioned above given the poor sequence identity among the translocators from the Inv-Mxi-Spa and the Ysc families, it is risky to extrapolate results obtained from one family to the other (48). In particular, when these bacteria have evolved to invade different organs and to interact with different cell types. For example, in the Shigella system, binding of bile salts trigger IpaB association with the tip and secretion of IpaC. After secretion, the interaction of IpaC with membranes is required to complete the association of the membrane-associated translocator with IpaD/IpaB at the needle tip (49). In contrast, neither YopB nor PopB are found associated with the tip of the needle (50). Moreover, P. aeruginosa strains where PopD is absent cannot translocate effectors, but PopB is still found associated with the target membrane (11). It is therefore clear that the translocators evolved to adapt to different cell targets, and their functions may differ among different families.
Although homogeneous hetero-complexes were observed on membranes when PopB and PopD are added in equimolar amounts, a great number of homo-complexes were also observed (Fig. 5A). Are there any physiological role(s) for homooligomers? It has been claimed that in addition to the essential role in effector translocation, homo-oligomerization of IpaC (or SipC in Salmonella) is essential for events that occurs inside the cytoplasm of the target cell. For example, SipC initiates actin nucleation (51) and IpaC has the ability to induce membrane extensions on macrophages (52,53). Moreover, IpaB induces Golgi fragmentation and reorganization of the recycling compartment (54) and oligomerization of IpaB form channels that permit potassium influx within endosomal compartments (55). Thus, it is possible that homo-oligomerization of PopD and PopB play other roles in the cytosol of the infected cell, but such abilities remain to be identified for the Ysc family members.
Finally, we would like to emphasize the unique properties of the T3S translocators in the context of protein membrane insertion and folding. Proteins that insert into membranes can be broadly classified in two major groups: (i) those that insert as unfolded polypeptides and require a complex proteinaceous machinery to properly fold and assemble into the lipid bilayer (56,57), and (ii) those that spontaneous insert but are secreted as well folded proteins, as best exemplified by bacterial poreforming toxins and complement system proteins (58 -60). The T3S translocators do not belong to any of these groups. They are secreted through the needle in an unfolded conformation (61), they do not adopt a well defined folded structure in aqueous solution (39,62), and no proteinaceous machinery has been identified for insertion and assembly into the target membrane. Therefore, the spontaneous assembly of unfolded proteins into large hetero-complexes with a defined stoichiometry constitutes a novel paradigm among membrane inserted proteins. As shown here, early interactions between PopB and PopD that occur during refolding, membrane binding, and/or insertion direct the spontaneous assembly of a hetero-complex of defined stoichiometry in a membrane bilayer.