Structure–function analysis reveals that the Pseudomonas aeruginosa Tps4 two-partner secretion system is involved in CupB5 translocation

Pseudomonas aeruginosa is a Gram-negative opportunistic bacterium, synonymous with cystic fibrosis patients, which can cause chronic infection of the lungs. This pathogen is a model organism to study biofilms: a bacterial population embedded in an extracellular matrix that provide protection from environmental pressures and lead to persistence. A number of Chaperone-Usher Pathways, namely CupA-CupE, play key roles in these processes by assembling adhesive pili on the bacterial surface. One of these, encoded by the cupB operon, is unique as it contains a nonchaperone-usher gene product, CupB5. Two-partner secretion (TPS) systems are comprised of a C-terminal integral membrane β-barrel pore with tandem N-terminal POTRA (POlypeptide TRansport Associated) domains located in the periplasm (TpsB) and a secreted substrate (TpsA). Using NMR we show that TpsB4 (LepB) interacts with CupB5 and its predicted cognate partner TpsA4 (LepA), an extracellular protease. Moreover, using cellular studies we confirm that TpsB4 can translocate CupB5 across the P. aeruginosa outer membrane, which contrasts a previous observation that suggested the CupB3 P-usher secretes CupB5. In support of our findings we also demonstrate that tps4/cupB operons are coregulated by the RocS1 sensor suggesting P. aeruginosa has developed synergy between these systems. Furthermore, we have determined the solution-structure of the TpsB4-POTRA1 domain and together with restraints from NMR chemical shift mapping and in vivo mutational analysis we have calculated models for the entire TpsB4 periplasmic region in complex with both TpsA4 and CupB5 secretion motifs. The data highlight specific residues for TpsA4/CupB5 recognition by TpsB4 in the periplasm and suggest distinct roles for each POTRA domain.


Introduction
Pseudomonas aeruginosa is a gram-negative opportunistic pathogen of animals, including humans, which can infect the eyes, lungs, kidneys, and urinary tract and lead to fatal consequences. P. aeruginosa is considered to be a model organism for studying bacterial biofilms 1,2 and uses a large number of extracellular appendages and adhesins. Several Chaperone-Usher Pathways [3][4][5][6] have been identified in P. aeruginosa (CupA to CupE) 7-12 that sustain this extracellular lifestyle, although for many of them their specific targets still remain unknown. However, it has been shown that CupA pili promote adherence to solid surfaces, 9 CupB/ CupC pili act in synergy to support microcolony formation, 12 CupD pili function to increase biofilm formation and decrease motility 11,13 and CupE pili aid in colony formation during early biofilms while also shaping the three-dimensional architecture of maturing biofilms. 10 Furthermore, these structures are differentially regulated, 14 which supports the notion that they play distinct roles during biofilm formation and/or allow the colonization of specific environmental niches. 7,8,[10][11][12] The CupC pili form hair-like structures on the bacterial surface 12 and their production is tightly controlled by the Roc systems, 15 specifically the RocS1 sensor. 16 CupB pili are also controlled by the Roc system, but are assembled by an atypical chaperone-usher pathway [ Fig. 1(A)]. This system contains two chaperones (CupB2 and CupB4) 17 and a 95 kDa gene product, CupB5, that is not found in other chaperone-usher pathways. 18 In P. aeruginosa there are also five known two-partner secretion (TPS) systems: Tps1-Tps5, 19 which secrete specific high molecular weight substrates that play distinct roles in virulence. 20 The Tps3 system transports a large adhesive structure (TpsA3 or CdrA) onto the surface of P. aeruginosa that promotes interbacterial aggregation through interactions via Psl polysaccharides. 21 On the other hand, the Tps4 system [ Fig.  1(A)] encodes an extracellular protease (TpsA4 or LepA) that modulates host responses to infection and liberates nutrients for bacterial growth. 22,23 As the name suggests, TPS systems are usually composed of just two components: a dedicated outer membrane transporter belonging to the Omp85 superfamily 24 and a single passenger exoprotein generically termed TpsB and TpsA, respectively, which are usually positioned in tandem within the genome. 25,26 The translocation unit is comprised of a membrane-embedded b-barrel pore with two N-terminal periplasmic POTRA (POlypeptide TRansport Associated) domains and an N-terminal a-helix that is located within the barrel. 27 POTRA domains have well-defined domain orientations 28 and in TpsB transporters they recognize extended conformers of the secretion domain of their TpsA passenger. 29,30 In the resting state, the N-terminal helix of the transporter has been observed in the periplasm, although this compartment is believed to be sparsely populated. 31 However, in the presence of cargo the equilibrium shifts toward this state, which allows the threading of the substrate through the pore and its eventual folding at the cell surface. 27,31,32 The formation of b-helical architectures when outside of the cell is believed to generate the force needed for translocation. 27 Although there is no direct structural evidence, periplasmic POTRA domains are thought to function by augmentation of their three-stranded bsheet with an extended cargo conformation. [33][34][35][36] Although CupB5 is predicted to form a TpsAlike structure, it was previously understood that CupB5 secretion was not dependent on a TPS system, but instead exploited the usher CupB3 from the operon in which it is embedded. 18 We now show using nuclear magnetic resonance (NMR) spectroscopy that TpsB4 recognizes both CupB5 and its predicted cognate substrate TpsA4. Additional genetic manipulation of the cupB and tps4 operons coupled with cellular assays also clearly demonstrates that TpsB4 is responsible for the translocation of CupB5, rather than the P-usher CupB3. Furthermore, we provide new molecular insight into POTRA domain interactions and show that each domain plays a unique role in substrate recognition and coordinating translocation. As many other bacterial pathogens also use TPS systems to cause disease, this work highlights a general secretion mechanism that could be exploited in the fight against emerging bacterial resistance to antibiotics.

Results
CupB3-independent secretion of the TpsA-like protein CupB5 Secretion of the TpsA-like protein CupB5 was previously thought to be dependent on a P-usher, 18 CupB3, suggesting a subtle integration of chaperoneusher pathways and TPS systems. In the present Immunoblot analysis of sheared fractions, harvested from bacteria grown on M63 agar plates, obtained from PAO1DD (DD: DpilA DfliC mutant) and PAO1DDDcupB3*R (DB3*R; cupB3 mutant from Ruer et al. 18 ) conjugated with pMMBrocS1, using antibodies derived against CupB1, CupB5 and CupB6. (C) Immunoblot analysis of cell and sheared fraction, harvested from bacteria grown static in liquid M63 medium, obtained from PAO1DD (DD) and PAO1DDDcupB3 (DB3; cupB3 mutant from this study), conjugated with pMMBrocS1, using antibodies derived against CupB1, CupB5 and DsbA. (D) Fold induction of cupB1, cupA1, tpsA1 (PA2462), tpsA2 (PA0041), tpsA3 (PA4625), tpsA4 (PA4541) and tpsA5 (PA0690) mRNA transcription in PAO1DD in the presence (procS1) or absence (pEV) of RocS1. The fold induction was normalized to the house keeping gene rpoD, which encodes the sigma factor 70. (E) Primary sequence alignment between the N-terminus of mature P. aeruginosa TpsA4 and CupB5 using ClustalW2. 66 Identical residues are red (*), highly similar residues are blue (:), weakly similar residues are green (.) and none homologous residues are orange. The conserved NPNG box and Asn residue of TpsA4 are boxed. Numbering is based on mature protein minus the N-terminal signal sequences. (F) Immunoblot of cell and sheared fraction obtained from PAO1DD (DD), PAO1DDDcupB5, PAO1DDDtpsB4 (DtpsB4) and PAO1DDDtpsB4::tpsB4 (DtpsB4::tpsB4) using antibodies derived against CupB5 and DsbA. The cupB operon was induced with RocS1. study and using similar growth conditions on agar plates set up in our laboratory, we observed that CupB5 could be readily recovered in the sheared fraction of this cupB3 mutant 18 (PAO1DDDcupB3*R: DcupB3 mutant within the DpilA DfliC mutant PAO1 strain PAO1DD) and thus was still able to reach the surface [ Fig. 1(A,B)]. At the same time the CupB1 and CupB6 pilin subunits were recovered in the sheared fraction of the parental strain, but not in the cupB3 mutant, confirming the expected defect in secretion for the CupB system. In addition we explored this observation using alternative conditions, in which the strains are grown in static culture instead of on agar plates. Again, using Western blot and CupB5-specific antibodies, it was found that CupB5 is systematically released in the sheared fraction. CupB5 secretion was also confirmed in a new cupB3 mutant lacking the P-usher (PAO1DDDcupB3), which was created independently for this study [Fig. 1(C)]. The periplasmic protein DsbA was used as a control and its absence in the sheared fractions indicates a lack of cell leakage. Taken together this suggests that in different laboratory environments alternative transporters to CupB3 play a role in CupB5 secretion.

TpsB4 is a genuine transporter of CupB5
TpsA proteins are usually transported by cognate TpsB transporters, therefore it is plausible that CupB5, by virtue of its similarity with TpsA-like proteins could use one of the five TpsB transporters that have been identified in P. aeruginosa. 37 qRT-PCR analysis was performed to investigate whether any of the five tpsA/tpsB gene couples could be upregulated upon overexpression of rocS1 and thus be coregulated with the cupB operon, including cupB5. As a negative control we monitored the induction of cupA1, which unlike the expression of the cupC operon, is not synchronized with the cupB gene cluster. 7,12 Strikingly, whereas cupB1 gene expression was readily induced upon introduction of the rocS1 gene, the tpsA4 gene also displayed an over 80-fold increase in expression [ Fig. 1(D)]. The product of the tpsA4 gene has been identified as a protease (LepA) 22,23 and its cognate transporter is believed to be the product of the adjacent tpsB4 (lepB) gene. None of the other tps genes, or the cupA1 gene, responded to the RocS1 control, suggesting that TpsB4 is a likely candidate for CupB5 transport.
TPS-dependent transport requires the presence of a 250-residue secretion or TPS domain at the N-terminus of the TpsA protein. 29,30,37 This domain contains residues involved in the recognition of TpsB POTRA domains and provides the minimum sequence required to promote folding of TpsA b-helical structures on the extracellular surface and export from the cell. We hypothesized that TpsB4 is the transporter for CupB5 and should possess a similar TPS domain pattern to TpsA4. Amino acid sequence alignment between mature CupB5 (minus the N-terminal signal peptide) and mature TpsA1 (PA2462), TpsA2 (PA0041), TpsA3 (PA4625), TpsA4 (PA4541), and TpsA5 (PA0690) gave sequence identities of 14.4%, 14.4%, 34.8%, 57.2%, and 16.4%, respectively. Furthermore, in addition to CupB5 and TpsA4 sharing significant sequence identity within this domain, 40-residues between S68-D107 (TpsA4) and S69-D108 (CupB5) are either identical or highly conserved [ Fig. 1(E)]. Overall, our data led us to reevaluate previous observations that the CupB3 usher is the dedicated transporter of CupB5 18 and test whether TpsB4 plays a role in this process.
The tpsB4 gene was deleted from the PAO1DD chromosome and upon overexpression of RocS1 secretion of CupB5 was assessed in liquid media as described above. Our data showed that lack of tpsB4 readily abrogates CupB5 secretion, which could then be recovered by reintroducing the tpsB4 gene at the att site on the chromosome of the PAO1DDDtpsB4 strain [PAO1DDDtpsB4::tpsB4; Fig. 1(A,F)]. The lack of CupB5 in the cell fraction when the transporter is not present is expected for TPS systems, as substrates are rapidly degraded in the periplasm if not secreted efficiently. [38][39][40][41] Our data indicate that CupB5 secretion is dependent on the TpsB4 transporter.

Solution structure of the TpsB4 POTRA1 domain
To address the molecular mechanism by which TpsB4 potentially recognizes two substrates, that is, both CupB5 and TpsB4, we used the multiple-threading alignment combined with fragment assembly simulation protocol of I-Tasser to model TpsB4 and facilitate the design of constructs for interactional analysis by NMR 42 [Fig. 2(A)]. I-Tasser produced a TpsB4 model with a C-score of 0.58, sharing tertiary homology with the structure of the Bordetella pertussis TpsB protein, FhaC. 27 The model contains a C-terminal 350-residue b-barrel pore and an N-terminal 200residue periplasmic region (TpsB4-NT) that can be further subdivided into a 20-residue N-terminal plug helix (involved in gating the pore), followed by a 20-residue linker and two adjacent POTRA domains (POTRA1 and POTRA2, respectively). Based on these predictions we produced recombinant His 6 -tagged TpsB4 proteins containing the entire periplasmic region (TpsB4-NT; Residues 1-194), POTRA1-2 domains (TpsB4-P12; , POTRA2 domain (TpsB4-P2; Residues 122-194), POTRA1 domain (TpsB4-P1; Residues 42-121), and the N-terminal plug helix, linker, and POTRA1 domain (TpsB4-a1P1; Residues 1-121; Fig. 2(B)). Residue numbering is based on mature TpsB4, minus the N-terminal signal sequence. These were expressed in Escherichia coli K12 strain and purified by nickel aeruginosa TpsB4 generated using the I-Tasser server. 42 (B) Schematic representation of the constructs used in this study and domains are colored as in (A). (C) Backbone superimposition of the ten best NMR structures of TpsB4-P1 shown in two orientations. An interactive view is available in the electronic version of the article (D) Secondary structure of one of the final TpsB4-P1 structures shown as cartoon. Numbering is based on mature TpsB4 minus the N-terminal signal sequence. An interactive view is available in the electronic version of the article affinity and gel filtration chromatography. 1D 1 H and 2D 1 H-15 N HSQC NMR spectra were recorded on purified proteins and showed only TpsB4-P1 and TpsB4-a1P1 to be correctly folded under these conditions. For example, NMR analyses of samples containing the POTRA2 domain showed significantly large methyl peaks in 1D spectra relative to TpsB4-a1P1 and TpsB4-P1, and although chemical shifts for the POTRA1 domain were present in 2D spectra, no dispersion was observed for the second domain (data not shown). The mis-folding of recombinant material for the second POTRA domain is likely due to lack of stabilizing interactions with the periplasmic face of the outer membrane b-barrel.
To elucidate the solution structure of the POTRA1 domain we used heteronuclear NMR methods. From both manual and ARIA (Ambiguous Restraints for Iterative Assignment) NMR assignment methods, 43 a total of 1818 nuclear Overhauser effects (NOEs) were assigned in TpsB4-P1 15 N/ 13 Cedited NOESY spectra at pH 7.0. The structure determination was also supplemented with //u dihedral angles and hydrogen bond restraints. The average pair-wise root-mean squared deviation (RMSD) for the water-refined final structures is 0.17 6 0.04 Å for the backbone atoms and 0.49 6 0.06 Å for the heavy atoms of residues within secondary structure. Structural statistics are shown in Table I.
All areas of secondary structure are well defined, however, there is less precision in the b4-b5 loop and at the N-terminus preceding the b1-strand Another interesting feature is that we observe NOE correlations between the N-terminal residues of the POTRA domain and the C-terminal pole of the a3-helix. This region, which is directly adjacent to the N-terminal plug helix linker in the intact TpsB4 structure, adopts a helical conformation that creates a pseudo-continuous helix with a3, mediated by electrostatic interactions from the a3-helix dipole, the a3-helix residue S79 and the carbonyl groups of G42 and P44.
In the crystal structure of FhaC, 27 electron density for the linker that connects the N-terminal plug helix and POTRA1 was not observed suggesting that it is flexible. Also, a recent biophysical study demonstrated that this linker is most likely constrained in the FhaC closed state, potentially through interactions with the POTRA domains. 31 We next used NMR to probe this hypothesis in TpsB4 by comparing 1 H-15 N HSQC spectra of TpsB4-P1, and for a construct extended with the plug helix linker (TpsB4-a1P1). Significant chemical shift differences are observed [ Fig. 3(A)]. The largest of these are mapped to the N-terminus, a3-helix and b4-b5-loop, indicating that the linker to the plug helix may fold back and interact transiently within this region [ Fig. 3

TpsB4-P1 recognizes a conserved sequence in both TpsA4 and CupB5
We next examined how the TpsB4 POTRA1 domain recognizes its cargo. Rather than initiating these studies with CupB5, however, we focussed on TpsA4, which is believed to be its cognate substrate. 23 It has been shown in other systems that these events can only be established when the secretion domain is unfolded, 30 therefore a His 6 tagged TpsA4 secretion domain (TpsA4 1-242 ; Residues 1-242 with numbering based on mature TpsA4 minus the N-terminal signal sequence) was purified under denaturing conditions prior to dissection. This was diluted into nondenaturing buffer in the presence of TpsB4-P1, treated with the protease trypsin and  with the corresponding peptide (CupB5 80-96 ; Q 80 VFLVNPNGVLFGRGAQ 96 with numbering based on mature CupB5 minus the N-terminal signal sequence) show that it also interacts with TpsB4-P1 with a virtually identical pattern of CSPs [ Fig. 4(B)], suggesting that these regions are functionally analogous. Although it was not possible to derive an accurate dissociation constant for CupB5 80-96 due to its poor solubility, it is consistent with the estimated average for TpsA4 79-95 ; K d 5 1.49 mM [60.18 mM; Fig. 4(C)]. The affected residues are localized to a distinct, contiguous region in POTRA1 [ Fig. 4(D)]: the b4-b5-loop, the adjoining b5-strand and the C-terminal region of the adjacent a3-helix and N-terminus/b1-strand. Interestingly, some of the largest CSPs lie at the extreme N-terminus of the POTRA domain (G42-T45) and into the neighboring N-terminal His 6 -tag. Furthermore, some of these CSPs coincide with those observed between TpsB4-a1P1 and TpsB4-P1, implying that interactions between TpsA4 or CupB5 and the POTRA domains of TpsB4 may displace its N-terminal linker connected to the plug helix.
TpsB4 POTRA domains recognize tandem TpsA4/CupB5 TPS motifs Three motifs have been identified in TPS domains that are important for the secretion of TpsA proteins: an NPNL box, an NPNG box and a single conserved asparagine position. 37,44 While the NPNL sequence is not present in TpsA4 or CupB5, the NPNG box is present in the TpsA4 79-95 and CupB5 80-96 peptides that recognize the TpsB4 POTRA1 domain [Figs. 1(E) and 4(B)]. The conserved Asp site (N97/N98 in TpsA4/CupB5) is 10-residues downstream of this NPNG box and as the entire sequence between S68-D107 in TpsA4 is almost identical in CupB5 [ Fig. 1(E)] we postulated that amino acid residues C-terminal to TpsA4 79-95 (e.g., within residues V96-D107) may be recognized by the second TpsB4 POTRA domain (TpsB4-P2). We therefore created a series of mutants that would both validate our NMR derived observations of POTRA/TPS interactions and also examine the role of the second TpsB4 POTRA domain in the secretion of CupB5.
We first generated mutations in the conserved TPS motifs of CupB5. We introduced the double point mutation N85A/P86A, part of the conserved NPNG box, which was designed to disrupt interactions with the POTRA1 domain. We also created the double point mutation V97W/V99W, which flanks the conserved N98 residue and was intended to disrupt interactions with the POTRA2 domain [ Fig. 1(E)]. In addition we created an S69A mutant outside of the expected TPS motifs as a negative control. The point mutations were generated in the CupB5-encoding plasmid, pBBR-MCS4-cupB5, introduced in the cupB5 mutant strain overexpressing RocS1 (PAO1DDDcupB5/pMMBrocS1), and secretion of CupB5 or mutant derivatives was monitored. Western blot analysis using anti-CupB5 antibodies clearly show that the N85A/P86A and V97W/V99W substitutions abrogate secretion, while S69A can still be recovered from the sheared fraction, thus supporting the role of these CupB5 TPS motifs in the interaction with TpsB4 POTRA domains [ Fig. 5(A)]. However, it should be noted that levels of recovery of the S69A control is lower than PAO1DD, which may reflect its close proximity to the TPS motifs. Consistent with the effect observed for our tpsB4 mutant [ Fig. 1(F)], and other TPS transporter mutants, abrogation of secretion results in CupB5 accumulation and subsequent degradation in the periplasm.
We next engineered mutations in the POTRA domains of TpsB4 at positions predicted to interact with the TPS motifs of CupB5, with control mutations that should not affect binding. To confirm the importance of the POTRA1 domain in the secretion of CupB5, we created a DPOTRA1 mutant and a double Q107A/E108A point mutation, localized in the dynamic b4-b5 loop and targeted based on significant CSPs observed in this region during titrations against TpsA4 79-95 and CupB5 80-96 . An additional double mutant was also produced (H82W/ P105A), positioned at the periphery of the interac-tion interface based on NMR CSPs, which we expected would have negligible effect on the secretion of CupB5. Within POTRA2 we created two double point mutations, G154W/E158W in the a4-helix and K173A/T175P in the b7-strand, designed to examine whether a groove created between these secondary structure elements were important for substrate recognition. Substitutions of G154 and E158 were expected to block CupB5 binding within this groove, while mutation of K173 and T175 would prevent b-sheet augmentation between the CupB5 substrate and the TpsB4 b7-strand. Finally as a negative control, residues D61 (POTRA1) and D149 (POTRA2), which are located on opposite side of the CupB5/TpsB4 interaction surface, were each substituted by alanine. The mutations were generated in the plasmid-encoded tpsB4 gene, and the chimeric genes were recloned into the integration-proficient vector mini-CTX1. For complementation, the mini-CTX1-tpsB4 or its derivatives were conjugated into the tpsB4 mutant overexpressing rocS1, PAO1DDDtpsB4/-pMMBrocS1, thus placing the tpsB4 allele at the att site on the chromosome. Western blot analysis to monitor CupB5 secretion showed that DPOTRA1, and K173A/T175P mutants completely abolished, while Q107A/E108A and G154W/E158W mutants significantly impaired CupB5 transport to the cell surface [ Fig. 5(B)]. At the same time and as hypothesized, the control and the H82W/P105A double mutant had no influence on CupB5 secretion [ Fig. 5(B)].

/CupB5 80-101 complexes
Interactions between TpsB4 POTRA1 and CupB5 80-96 or TpsA4 79-95 are in fast exchange on the NMR timescale, precluding the elucidation of NOE-based, high-resolution structures for these complexes. We therefore generated structural models that would describe a secretion complex of TpsB4 with the translocation signals from CupB5/TpsA4 by incorporating our experimental NMR and mutagenesis data. Using the HADDOCK approach, 45 Table II.
These complexes share identical overall features and can be considered as an interaction between tandem POTRA domains (POTRA1 and POTRA2) and the adjacent TPS motifs [TPS1 and TPS2; Fig.  5(C,D)]. TPS1 contains the conserved NPNG box while TPS2 contains the completely conserved N97/ 98 residue (TpsA4/CupB5). POTRA1 is recognized by the CupB5 TPS1, kinking under the N-terminal base of this POTRA domain at position P86 (P85 in TpsA4). Here the first substantial interaction is observed between TPS1 F82 (CupB5; F81: TpsA4) and POTRA1 in a pocket created by the b4-b5-loop and b1-strand [ Fig. 5(D)]. In the unbound POTRA1 domain structure, this loop is "pinned" back with the side chains of Q110 in the loop and R115 in the b5-strand (Supporting Information Fig. S2). Our models for these complexes suggest that when the peptide binds, F81/82 (TpsA4/CupB5) is accommodated by a rearrangement of this loop region.

Tps4 is likely a multisubstrate TPS system
In this study, we show that under various growth conditions the TpsB4 transporter is able to translo-cate CupB5 across the P. aeruginosa outer membrane which contradicts earlier observations in which the P-usher CupB3 was shown to secrete CupB5. 18 Our new data described in this manuscript clearly identifies TpsB4 as the major CupB5 transporter and is supported by extensive mutagenesis and structural insight. While this may point to an artefact in the earlier work which did not probe the role of TpsB4 extensively, we also cannot rule out that the laboratory environment and reagents in which strains have been grown may affect the transport process and any interplay between these two systems. The cupB gene cluster is under the tight control of a complex regulatory network, which involves several sensors, including RocS1, and a yet unknown response regulator, 15 so subtle environmental effects may be highly relevant in accounting for phenotypic differences observed. There may also be interplay, or at least coordination, between CupB3-and TpsB4-dependent secretion of CupB pili, CupB5 and TpsA4. For example, it was shown in the case of the Klebsiella oxytoca pullulanase (PulA) secretion that interaction with either the type II secretion machine (T2SS) or the lipoprotein sorting pathway (Lol) might be cooperating or competing events which impact on the final destination of PulA. 47 Importantly, Ruer et al. 18 also observed evidence for a CupB3-independent transport of CupB5, in that a deletion of the CupB3 N-terminus, a predicted POTRA-like domain, did not prevent CupB5 from reaching the surface. In light of this unexpected observation, we have further scrutinized the sequences of P. aeruginosa CupB and TPS systems and revealed here that a TPS motif is present in CupB5 that is in essence identical to the TpsA4 recognition signal, the putative cognate substrate of the TpsB4 transporter. Expression of the cupB genes is under the control of the RocS1 network and we show here that the tpsA4 gene is coregulated with the cupB genes in a RocS1-dependent manner. systems have illustrated that TPS substrate recognition is defined by the recognition of specific TpsA TPS motifs by both POTRA domains of a TpsB transporter. 48 We have confirmed using biophysical and genetic approaches that both TpsA4 and CupB5 TPS motifs are recognized by POTRA domains of the TpsB4 transporter, one of five TpsBs identified in P. aeruginosa. 19,37 Therefore, together with the tandem organization of tpsA4/tpsB4 genes, our observations that the TpsB4 POTRA1 recognizes both TpsA4 and CupB5, and CupB5 secretion is dependent on TpsB4, this strongly supports the notion that Tps4 is a multisubstrate TPS system. The existence of multisubstrate TPS systems, however, is not unique. For example, Bordetella bronchiseptica encodes two TpsA proteins (FhaB and FhaS) that are both secreted by FhaC. 40 Although they contain 95% identity over the majority of their sequences, they display different functionalities. The bacterium N. meningitidis has three TPS systems that are scattered across its genome but are nonetheless coexpressed. Two of these, Tps1 and Tps2, each encodes a single transporter and two substrates, while Tps3 contains a single tpsA3 gene. Whereas TpsB1 secretes both of its cognate TpsA1a and TpsA1b cargo, the TpsB2 transporter exports TpsA2a, TpsA2b, and TpsA3. 49,50 Usually, however, when a tpsA gene is clustered with chaperone-usher genes, it is also found with a cognate tpsB gene. Both genes can be separated by other genes in the chaperone-usher cluster as in B. pertussis 51 or organized in tandem and embedded in the chaperone-usher cluster as in Pseudomonas fluorescens. 52 Our observation that the product of the orphan P. aeruginosa tpsA gene, cupB5, can be transported by a distally encoded TpsB, reflects that an associated tpsB gene may have been lost due to the presence of the highly similar, alternative TpsB. In fact in P. aeruginosa, the TpsB1 and TpsB2 transporters share 99% sequence identity, as do the 250residue TPS-domains of their predicted TpsA1 and TpsA2 substrates (71% identity over the entire TpsA1/A2 sequences) and this may represent another multisubstrate TPS system in this organism. The Tps4/CupB system we describe here reflects a coevolution between chaperone-usher pathways and two-partner secretion systems, which can be genetically linked. Although sequence differences between CupB3 and classical ushers could also suggest an alternative evolution, we have now collected clear evidence that the interaction between CupB5 and TpsB4 is effective and functional, involving key residues in the TPS translocation signal of CupB5 and the POTRA domain of TpsB4 as described here.

TpsB4 POTRA domains have distinct functions
The solution structure of TpsB4-P1 contains the usual core b-a-a-b-b structural motif, but also possesses an additional b1-b3 sheet pairing at the N-terminal base of the domain. This was not observed in FhaC, 27,53 the only other TpsB structure, although it is present in one other POTRA domain containing structure: POTRA1 from a Cyanobacterial Omp85 54 (pdb: 3mc8). Interestingly, significant conformational dynamics exist within the b4-b5 loop and at the N-terminus preceding the b1-strand [ Fig. 2(C) and Supporting Information Fig. S1], which is where TPS1 binds and are likely important for selection of the cargo (Supporting Information Fig. S2).
A conserved proline in the NPNG box of TPS1 allows TpsA/CupB5 to be kinked around POTRA1, which would set the initial register of the TPS motifs on the POTRA domains. The main interactions between TPS1 and POTRA1 are localized to this N-terminal pole of TpsB4 and do not extend substantially along the face of the domain [ Fig.  4(D)]. We also show that the linker adjoining the TpsB4 N-terminal a1 helix (involved in gating the b-barrel pore) specifically interacts with POTRA1 and is likely displaced during recognition of TpsA4/CupB5 TPS1. These observations are supported by other recent studies 31 and indicate that recognition of TPS motifs by POTRA domains is coordinated with release of the N-terminal plug helix, shifting the equilibrium toward an open TpsB conformation.
Using the entire periplasmic region of FhaC and an extended 30 kDa FHA TpsA-secretion domain, low micromolar dissociation constants have been reported. 44 As shown in this study, the interaction between a single POTRA domain of the TpsB4 transporter and its corresponding TPS region is significantly weaker. The presence of two POTRA domains increases the affinity between TpsB transporters and their cargo substantially, while enabling greater specificity. It has been suggested that POTRA domains recognize their substrates through b-sheet augmentation, [33][34][35][36] although there is no direct structural evidence for this. Our NMR structure of POTRA1 clearly shows that b-sheet augmentation would not be possible here, as chemical shift data demonstrates that TPS1 primarily interacts with the b4-b5 loop of the domain. However, our data does support the potential for augmentation between TPS2 with the b7-strand in POTRA2. When the CupB5 80-101 conformation is transposed onto the POTRA domains from the FhaC crystal structure 27 (pdb: 2qdz), the FHA TPS motifs follows a similar trajectory through grooves with complementary electrostatic surface potential (Supporting Information Fig. S3) and it is likely that our observations can be generalized for other TPS systems.
The G100/101 residue in TpsA4/CupB5 is positioned at the C-terminal pole of POTRA2, which in the context of intact TpsB4 would lie at the opening of the barrel. The 10 residues that follow are either identical or highly conserved between TpsA4 and CupB5 [ Fig. 1(E)]. This corresponds to approximately 35 Å of extended polypeptide, which is sufficient to cross the outer membrane and to interact within the barrel of TpsB4 to facilitate transport. These observations suggest that once the TPS motifs of the cargo binds the POTRA domains (TpsA N-terminus facing into the periplasm) a hairpin could form within the TpsB4 pore and polymerization at the tip of this hairpin result in translocation of the remaining Cterminus across the outer membrane. In addition to our work here, others have shown that interactions between POTRA domains and TPS motifs are quite dynamic, which may be essential for substrate release upon folding. 44 While the specific cargo release mechanism is unknown, it is interesting to note that extracellular TpsA4 is processed at its Nterminus and begins at His22, just prior to the TPS1 motif. Therefore, proteolysis of TpsA4, albeit slightly upstream of the TPS1 motif, may destabilize POTRA-TPS interactions and contribute to its release.

Tps4, CupB and CupC systems may act in synergy under specific conditions to promote virulence
TpsA4 (LepA) functions as an extracellular protease, 22,23 that can activate the NF-jB pathway through cleavage of protease-activated receptors (PARs) allowing the modulation of host responses against bacterial infection. 23 Furthermore, in cooperation with a P. aeruginosa hemolysin (PlcH), TpsA4 can degrade hemoglobin to liberate peptide fragments and iron as a source of nutrients for bacterial growth. In this study and in previous investigations 12 we have shown that tps4, cupB, and cupC operons are induced by overexpression of RocS1. While CupC fimbriae promote the development of microcolonies during early biofilm formation, the specific role for the CupB pilus and CupB5 have yet to be determined. However, a recent study has shown that peptides released via degradation of CupB5 at the outer membrane can stimulate the overproduction of alginate, which is a major component of P. aeruginosa biofilms. 55 CupB pili display a putative tip adhesin (CupB6) and it is plausible that this plays a role in adhesion to biotic surfaces, complementing the interbacterial agglomeration induced by CupC fibres. Analysis of CupB5 using the I-Tasser server 42 identifies a number of b-helical structures with excellent alignment scores, including the E. coli Hbp protease (sequence coverage: 0.9, normalized Z-score: 4.0) 56 and the Haemophilus influenzae Hap adhesin (sequence coverage: 0.9, normalized Z-score: 2.0) 57 (Supporting Information Fig. S4). It has previously been shown by transmission electron microscopy and immunogold labelling that CupB5 is localized on the surface of P. aeruginosa alongside CupB pili. 18 In enterotoxogenic E. coli, the TpsA-like molecule EtpA is secreted into the extracellular space where it associates with the flagella and mediates adhesion to host cells. 58 It is tempting to speculate that CupB5 plays a similar role here by bridging or modulating adherence between the CupB pilus and specific host receptors and we are currently developing cellular adhesion assays to test these theories. While we still do not fully appreciate the implications of synergy between the cupB, cupC, and tps4 operons, it is clear that a coordination of bacterial agglutination, mechanisms of nutrient acquisition and the ability to modulate host responses to infection will provide great benefits to an organism during colonization of niche environments.

P. aeruginosa strains and growth conditions
The strains and plasmids used in this study are listed in Supporting Information Table S1. Overnight cultures of P. aeruginosa strains were grown at 37 C in LB medium containing the appropriate antibiotic concentration: 80 mg/mL gentamicin (Gm), 300 mg/mL carbenicillin (Carb), 50 mg/mL tetracycline (Tet), or 2 mg/mL streptomycin (Sm). For appendage preparation, P. aeruginosa strains conjugated with pMMBrocS1 were grown at 30 C either static for 24 h in liquid M63 medium or for 4 days on M63 agar plates. Liquid medium and plates were supplemented with 0.4% L-arginine, 1 mM MgSO4 and 80 mg/mL Gm. 18 Bacterial expression was induced with 0.1 mM IPTG.

Construction of PAO1DD mutants
The oligonucleotides used in this study are listed in Supporting Information Table S2. Polymerase chain reaction (PCR) was used to amplify 500-bp DNA fragments upstream of the cupB3, and tpsB4 genes from the P. aeruginosa genome using the primer pairs 950/ 951, and 976/977, respectively, whereas 500-bp downstream DNA fragments were generated by using the primer pairs 952/953and 978/979, respectively. The generated DNA fragments were used as a matrix in an overlapping PCR using primer pairs 950/953, and 976/979, to obtain the cupB3, and tpsB4 mutator DNA, respectively. The mutator DNA fragments were ligated into the cloning vector pCR2.1 and recloned into the suicide vector pKNG101 by restriction digest. Plasmids were conjugated in PAO1DD by three-partner conjugation. 59

Construction of complementation vectors
PCR was used to amplify the tpsB4 gene with its promoter and ribosome-binding site (RBS) and the cupB5 gene and its RBS from the chromosome of PAO1DD by using primer pairs 1315/1316 and 812/ 813, respectively (Supporting Information Table S2). The cupB5 DNA fragment was ligated into pCR-Blunt and the tpsB4 gene into pCR2.1. tpsB4 was recloned into mini-CTX1 using the restriction enzymes SpeI/EcoRI thereby creating miniCTX-tpsB4. To obtain pBBR-MCS4-cupB5, cupB5 was transferred into pBBR1MCS-4 by using PstI/SpeI/ SmaI, whereas SmaI was used to cut pCR-Blunt.

Quantitative real-time PCR
Total RNA was prepared from triplicate cell samples of PAO1DD and PAO1DD conjugated with pMMBrocS1, stored in RNAlater, using the RNeasy extraction kit (Qiagen) in combination with on column DNaseI treatment (Applied Biosystems). The DNase treatment was carried out twice to remove all traces of chromosomal DNA. mRNA concentration was measured with the NanoDrop 1000 spectrophotometer (NanoDrop Technologies) and normalized to 20 ng/mL. cDNA was synthesized using SuperScript III Rnase H-Reverse Transcriptase (Invitrogen), 10 mM dNTPs, RNAguard (Roche), random hexamer oligonucleotides (Amersham), 0.1 M DTT, and 1X first strand buffer (RT Superscript Kit). Negative controls without reverse transcriptase were run with all reactions. The resulting cDNA was 1:5 diluted and stored at 220 C. For quantitative real-time PCR, SYBR green PCR master mix (ABI), an upstream and downstream primer pair (Supporting Information Table S2) and water were used. Samples were analyzed by the StepOne Real-Time PCR System (Applied Biosystems). The relative quantity gene expression was determined by defining the expression of rpoD mRNA as 1.

Shearing experiments
Shearing experiments from bacteria grown on agar plates were performed as described by Ruer et al. 18 Briefly, bacteria were scrapped from agar plates and an equivalent of 5 OD units was resuspended in LB. Appendages were sheared off by gentle agitation at 4 C overnight. Cells were harvested by centrifugation at 4 C: first, 4000g for 15 min, second, 4000g for 20 min and third, 13,000g for 15 min. Cell pellets, obtained after the first centrifugation step, were washed in 10 mM Tris, pH 8, sedimented by centrifugation (5 min, 13,000g, 4 C) and resuspended in 1 3 SDS-loading buffer. The sheared fraction was then subjected to an ammonium sulfate precipitation at a concentration of 50% for 5 min at room temperature and sheared fraction pellets were harvested by centrifugation (75 min, 13,000g, 4 C). Precipitated proteins were harvested by ultracentrifugation (70,400g, 45 min 4 C) and pellets resuspended in 10 mM Tris, pH 8 and mixed with 5 3 SDS-loading buffer. Cell and sheared fractions were heated for 10 min at 95 C. Bacteria grown in liquid medium were diluted to an equivalent of 10 units of OD600 and appendages were sheared off by gentle agitation for 3 h at 4 C. Cell and sheared fractions were separated by centrifugation (as described above) and sheared fractions subjected to ammonium sulfate precipitation. The pellet was resuspended in 10 mM Tris, pH 8 and 5 3 SDSloading buffer was added. Appendage and cell pellets were resuspended in SDS loading buffer and heated for 10 min at 95 C.

SDS-PAGE and Western blot analysis
Proteins were separated by electrophoresis on a 12% polyacrylamide gel at 160 V. The proteins were either stained with Coomassie blue or transferred to a nitrocellulose membrane (Whatman) using a semidry transfer system (Biorad). The membrane was blocked by incubation overnight at 4 C in 5% (w/v) milk containing 0.01% (v/v) Tween in PBS (PBST). The membrane was incubated for 2 h at RT with the primary antibody, diluted in 5% (w/v) milk in PBST.
Membranes were washed three times for 5 min in PBST and then incubated for 45 min at RT in horseradish peroxidase (HRP)-conjugated goat-anti-rabbit/ rabbit-anti-mouse secondary antibody at a dilution of 1:5000 in 5% (w/v) milk in PBST. Membranes were again washed three times for 5 min in PBST. Proteins were detected using a chemiluminescence revelation kit (Thermo Fisher Scientific).
NMR solution structure determination NMR measurements were performed at 310 K on a 15 N 13 C-labelled TpsB4-P1 sample in 50 mM NaPO 4 pH 7.0, 150 mM NaCl, 10% D 2 O. NMR experiments for backbone and side-chain assignment were performed on three different Bruker spectrometers, a DRX500, DRX600, and DRX800, equipped with TCI and TXI cryoprobes. Assignments were completed using standard triple-resonance assignment methodology 60 and data were analyzed using in-house algorithms with the program NMRView. 61,62 A total of 96% of the potential backbone (e.g., disregarding the N-terminal methionine and proline residues) and 76% of the potential side-chain resonances were assigned; this corresponds to 98% and 84%, respectively, when the 15-residue N-terminal histidine tag is ignored. Heteronuclear NOE cross-relaxation rates on a 15 N-labelled TpsB4-P1 sample were recorded at 600 MHz. NOE data for structure calculations of TpsB4-P1 were obtained from an 800-MHz edited 1 H-15 N/ 13 C nuclear Overhauser effect spectroscopy (NOESY)-heteronuclear single/multiple quantum coherence (HSQC/HMQC) experiments.
However, due to spectral overlap only 63 out of a potential 91 peaks could be analyzed. The ARIA protocol for automated NOESY assignment interfaced with the CNS program was used for structure calculation. 43 Secondary structure in the TpsB4-P1 POTRA domain was first identified using the chemical shift-based dihedral angle prediction software TALOS. 63 For residues located in secondary structure, experimentally derived hydrogen bonds and //u backbone dihedral angles from TALOS, were introduced as restraints in the ARIA structure calculation. The frequency window tolerances for assigning NOEs were 60.03 ppm for direct proton dimensions, 60.06 ppm for indirect proton dimensions, and 60.65 ppm for nitrogen and carbon dimensions. One hundred structures were calculated in the final iteration and these had no NOE violations >0.3 Å and no dihedral angle violations >5 . The ten lowest energy structures have been deposited in the Protein Data Bank with PDB ID code 2mhj and a summary of NMR-derived restraints and statistics is reported in Table I.

Peptide synthesis
To increase solubility all peptides were produced with free N-and C-termini. Initial peptides used in this study (including TpsA4 80-111 ) were synthesized by ThermoScientific to >95% purity, while all other peptides (TpsA4 79-95 and CupB5 80-96 ) were synthesized using standard Fmoc solid phase methods, using an automated ResPepSL synthesiser (Intavis) running Multipep software, adapted from Ref. equivalents of CupB5 80-96 peptide, were used to measure 1 H-15 N HSQC NMR spectra. Experiments were performed at 310 K on a Bruker DRX600. Due to peptide solubility issues saturation could not be reached. Kinetic analyses of TpsB4-P1/TpsA4 79-95 interactions were estimated in NMRView 62 using 1 H-15 N CSPs from 24 separate spin systems.

Experimental driven docking
Ambiguous distance restraints from chemical shift perturbations and in vivo mutational analysis were used to drive a structural calculation of TpsB4  / TpsA 79-100 and TpsB4  /CupB5 80-101 complexes using the "conformational selection and induced fit protein-peptide HADDOCK docking" approach. 46 For the calculations, a composite NMR/I-Tasser 42 model of TpsB4  and an extended conformation of either a 22 residue TpsA 79-100 or CupB5 80-101 peptide was used with a 1:1 stoichiometry. Superposing our NMR structure of TpsB4 42-121 onto Residues 42-121 from the I-Tasser 42 model of TpsB4 and then combining this with Residues 122-194 produced the model of TpsB4  . PyMOL 65 was used, inputting 2139 for phi and 2135 for psi, to generate the extended peptide conformations. Based on chemical shift perturbation, filtered for a relative solvent accessibility greater than 50%, 17 residues in TpsB4-P1 (42,43,44,45,46,78,79,82,83, 86, 104, 105, 106, 107, 109, 110, and 111) were identified as active residues. In this study we have shown that mutants in CupB5 (Residues 85, 86, 97, and 98; equivalent Positions 84, 85, 96, and 98 in TpsA4), Residues 107-108 in TpsB4 POTRA1 and Positions 154, 158, 173, and 175 in POTRA2 abrogate CupB5 secretion and these were therefore also assigned as active residues. Residues juxtaposed to these in the monomer structures that have a relative solvent accessibility greater than 50% were termed passive residues. TPS peptides were allowed to be completely flexible and the interfacial residues in TpsB4  were allowed to move during the simulated annealing and water refinement. 3000 initial complex structures were generated by rigid body energy minimization and the best 1000 by total energy were selected for refinement and analysis with 2500 followed by 5000 molecular dynamic cycles. The best 200 by total energy were selected for torsion angle dynamics and subsequent Cartesian dynamics in an explicit water solvent. Default scaling for energy terms was used as described previously. 45 Details of the structural statistic are shown in Table II.