Direct interactions between the secreted effector and the T2SS components GspL and GspM reveal a new effector-sensing step during type 2 secretion

In many Gram-negative bacteria, the type 2 secretion system (T2SS) plays an important role in virulence because of its capacity to deliver a large amount of fully folded protein effectors to the extracellular milieu. Despite our knowledge of most T2SS components, the mechanisms underlying effector recruitment and secretion by the T2SS remain enigmatic. Using complementary biophysical and biochemical approaches, we identified here two direct interactions between the secreted effector CbpD and two components, XcpYL and XcpZM, of the T2SS assembly platform (AP) in the opportunistic pathogen Pseudomonas aeruginosa. Competition experiments indicated that CbpD binding to XcpYL is XcpZM-dependent, suggesting sequential recruitment of the effector by the periplasmic domains of these AP components. Using a bacterial two-hybrid system, we then tested the influence of the effector on the AP protein–protein interaction network. Our findings revealed that the presence of the effector modifies the AP interactome and, in particular, induces XcpZM homodimerization and increases the affinity between XcpYL and XcpZM. The observed direct relationship between effector binding and T2SS dynamics suggests an additional synchronizing step during the type 2 secretion process, where the activation of the AP of the T2SS nanomachine is triggered by effector binding.

well as with the periplasmic domain of GspC (14,17). In contrast to GspM and GspC, GspL harbors an additional N-terminal cytoplasmic domain presenting structural homology with actin-like ATPases (20). This domain is, together with the integral inner membrane protein GspF, involved in the recruitment of the ATPase GspE at the secretion site (21)(22)(23)(24)(25)(26). Further activation of GspE requires interactions with phospholipids (27). It has been proposed that upon sensing of a signal, the AP interaction network is displaced to ensure proper functioning of the system possibly through intrinsically disordered domains (4). This includes signal transduction across the IM between the periplasmic and cytoplasmic sides of the secreton (4,28). Such transmembrane dynamic signaling has also been reported in the archetypal Tff member, the type IV pilus (29).
Type 2 secretion is a two-step process during which effectors are first exported across the IM by the Sec or Tat systems (30,31). Then, the folded periplasmic effectors are recognized and transported to the extracellular milieu by the secreton (32). How T2SS effectors are specifically recognized by the secreton in the periplasmic soup remains an open question. In contrast to other secretion systems and despite intense research, no common secretion signal has been identified in T2SS effectors. However, several direct interactions have been described between secreted effectors and the secretin GspD, the AP component GspC, and the pseudopilus tip (24,(33)(34)(35)(36). It is accepted that effector recognition and recruitment are performed by GspC (17,24,34,37) followed by its transfer into the secretin vestibule before being extruded out of the cell upon contact with the pseudopilus tip (32). A direct interaction between effectors and GspC HR and PDZ subdomains (24,37) suggests that effector recruitment involves multiple contacts with GspC. In addition, it has been shown that the TMHR domain of GspC is indirectly involved in effector recognition specificity in Pseudomonas aeruginosa (17). Various GspD N-domains are also involved in effector binding, depending on the organism (5,24,33,34,36). Finally, a direct interaction has also been reported between the effector and the pseudopilus tip constituted by the periplasmic domains of GspHIJK (34).
Here, we report that in addition to interacting with GspC, the secretin, and the pseudopilus tip, the T2SS secreted effector also interacts specifically with the periplasmic domains of GspM and GspL inner membrane components of the AP. We further show that these periplasmic interactions trigger conformational changes in the AP that may lead to ATPase activation and pseudopilus assembly. We thus propose that these newly discovered interactions constitute an additional step of the T2SS secretion process, synchronizing effector loading and pseudopilus assembly.

Direct and specific interaction between the Xcp effector CbpD and XcpY L periplasmic domain (XcpY L p)
Our T2SS working model is the Xcp system of P. aeruginosa where 11 different components are named P-Z with the species-specific prefix Xcp. Because protein letters are specific to the Pseudomonas genus, we will systematically refer to the gen-eral Gsp nomenclature using a subscript; i.e. the GspL homolog in P. aeruginosa is named XcpY L .
We previously showed by surface plasmon resonance that the purified periplasmic domains of the secretin XcpQ D , the AP component XcpP C , and the pseudopilus tip quaternary complex (XcpU H V I W J X K ) directly bind secreted effectors, thus allowing us to propose an integrated model of effector recognition and transport by the T2SS (34). To have a better view of the Xcp/effector interactome in the periplasm, we tested the interaction between the secreted effector CbpD and the periplasmic domains of the bitopic AP component XcpY L (XcpY L p). We used biolayer interferometry (BLI), an in vitro protein-protein interaction technique similar to surface plasmon resonance. XcpY L p and CbpD proteins were produced and purified by consecutive affinity and size-exclusion chromatography steps (Fig. 1A), following the procedure used previously (34). CbpD was then biotinylated and immobilized on the sensor tip to be used as bait, and interaction experiments were performed in triplicate with purified XcpY L p used as prey, following the protocol described under "Experiment procedures." The graph, presented in Fig. 1A and reporting the response (nm) as a function of the XcpY L p concentration (M), was used to measure the dissociation constants (K D ). BLI data reveal that XcpY L p directly interacts with the secreted effector CbpD with a relatively low K D of 5.0 M.
To validate this in vitro interaction between CbpD and XcpY L in a more biological context, we set up and performed an in vivo copurification experiment. To reconstitute the natural periplasmic context of the interaction in the absence of the other Xcp T2SS components and secreted effectors, the two partners were produced in the periplasm of the heterologous host Escherichia coli. When produced in E. coli, CbpD naturally accumulates in the periplasm because of its Sec signal peptide (Sp). The second partner, XcpY L p, was artificially targeted to the periplasm by the addition of LasB Sp to its N terminus (Sp-XcpY L p). We first tested protein production under inducing conditions and verified the proper periplasmic localization of CbpD and XcpY L p (Fig. S1). The soluble cell fraction of the E. coli/pCbpD H /pSp-XcpY L p strain grown under inducing conditions was extracted and analyzed by immobilized metal affinity chromatography (IMAC). The copurification experiments shown in Fig. 1B indicate that XcpY L p is coeluted with the histidine-tagged CbpD H , used here as bait, thus supporting the direct interaction found by BLI between the two proteins. The observation (Fig. 1B, bottom) that XcpY L p is not recovered in the elution fractions in the absence of CbpD H excludes a nonspecific affinity of XcpY L p for the IMAC resin.
We then took advantage of the presence of two independent T2SSs in P. aeruginosa, the Xcp and Hxc systems, each secreting their own effectors (38), to test the effector specificity of this newly characterized interaction. We performed in vivo crosscopurification experiments using E. coli strains coproducing the Hxc effector LapA H together with XcpY L p (E. coli/pLapA H / pSp-XcpY L p). Soluble protein lysates obtained under inducible conditions were analyzed by IMAC (Fig. 1C), and we noticed that XcpY L p did not copurify with the heterologous Hxc effector. These data confirms that, as is also the case for XcpQ D ,

Effector recognition in type 2 secretion
XcpP C , and the pseudopilus tip (34), the interaction of XcpY L with its cognate effector is system-specific.

Direct interaction between the Xcp effector CbpD and the periplasmic domain of XcpZ M
To further characterize the Xcp/effector periplasmic interactome, we next tested whether CbpD interacts with the periplasmic domain of the inner membrane component XcpZ M (XcpZ M p). XcpZ M corresponds to the only periplasmic globular domain of the secreton not yet tested for interaction with the secreted effector. As above for XcpY L , we combined complementary in vitro and in vivo protein-protein interaction experiments to investigate the interaction between XcpZ M p and secreted effectors. The BLI experiment using purified CbpD as bait and XcpZ M p as prey was done in triplicate and showed a direct interaction between the two proteins, with a K D of 3.4 M ( Fig. 2A). This in vitro interaction was confirmed by a crosslinking experiment using the short cross-linking agent bis(sulfosuccinimidyl)glutarate (BS2G). Analysis of the cross-linking products by SDS-PAGE followed by Coomassie Blue staining showed a protein complex specifically recovered in the presence of the two partners and the cross-linker (Fig. S2) with a molecular weight corresponding to a heterodimer composed of XcpZ M p and CbpD. This was confirmed by immunoblotting, which showed that XcpZ M p and CbpD are both present in the corresponding complex.
Furthermore, the interaction between XcpZ M p and CbpD was tested and validated by in vivo copurification experiments. To do this, we constructed an E. coli strain producing in its periplasm CbpD H and XcpZ M p (E. coli/pCbpD H /pSp-XcpZ M p) and proceeded with IMAC copurification experiments following the procedure used for XcpY L p.
The analysis of the eluted fractions by SDS-PAGE followed by immunoblotting with anti-CbpD and anti-XcpZ M antibodies shows the specific coelution of XcpY L p by CbpD H (Fig. 2B). As for XcpY L p, the XcpZ M p/effector interaction is Xcp T2SSspecific because the Hxc effector LapA H does not copurify the Xcp GspM component XcpZ M p (Fig. 2C). Altogether, the presently discovered direct interactions between the secreted effector and the globular periplasmic domains of XcpZ M and XcpY L reveal for the first time a direct and specific interaction between secreted effector and components of the T2SS assembly platform.

Competition between XcpY L p and XcpZ M p for binding to CbpD
Protein-protein interaction data revealed that the AP components XcpY L and XcpZ M interact directly with the effector through their periplasmic domains, which raises to five the number of Xcp periplasmic domains or complexes that directly and specifically interact with secreted effectors. In all cases, interaction affinities between T2SS components and the secreted effector are in the M range, in agreement with their transiency during the secretion process. We attempted to better understand the CbpD/XcpY L p/XcpZ M p interactome by evaluating possible competition effects thanks to the multiple coexpression capacity of our in vivo periplasmic reconstitution effector/Xcp interactome assay. We therefore quantitatively compared the copurification levels of XcpY L p and XcpZ M p with CbpD H in the presence or absence of the other CbpD interactant. Hence, the soluble cell lysates of the E. coli strains coproducing CbpD H with XcpY L p, XcpZ M p, or XcpY L p together with XcpZ M p were generated in quadruplicates and analyzed by IMAC for XcpZ M p and XcpY L p copurification (Fig.  3). Although the proportion of XcpZ M p copurified with CbpD

Effector recognition in type 2 secretion
is unchanged with or without coproduction of XcpY L p (Fig. 3, gray bars and corresponding immunoblots), a statistically significant reduction of XcpY L p binding to CbpD was observed in the presence of XcpZ M p (Fig. 3, black bars and corresponding immunoblots). This competition experiment suggests that, during the secretion process, the secreted effector interacts sequentially with XcpY L and XcpZ M .

CbpD effector triggers XcpZ M dimerization and increases XcpZ M /Y L interaction
The above data indicate that both XcpY L and XcpZ M interact with the secreted effector through their periplasmic domains. To understand the possible consequences of such interactions on the global AP interactome within the secreton, we used the bacterial adenylate cyclase two-hybrid (BACTH) method developed by Karimova et al. (39). In this technique, proteins of interest are coexpressed in an E. coli cya mutant (BTH101) as fusions with one of the two fragments (T18 and T25) from the catalytic domain of Bordetella pertussis adenylate cyclase. Interaction of two-hybrid proteins results in a functional complementation between T18 and T25, leading to cAMP synthesis and transcriptional activation of the lactose operon that can be easily detected by ␤-gal activity measurement. We chose this technique because it is particularly well adapted to quantify interactions between membrane proteins (40). Full-length XcpY L , -Z M , and -S F proteins were therefore fused to the T18 and/or T25 domains via their N termini (see "Experimental procedures").
We first evaluated the heterodimerization capacities of different AP components by measuring ␤-gal activity of the three combinations, T18-Y L /T25-S F , T18-Y L /T25-Z M , and T18-Z M / T25-S F , and comparing it with positive and negative controls (Fig. 4, light gray bars). Results indicate that full-length XcpY L directly interacts with XcpZ M . A similar observation, using the same approach, has already been reported in different T2SSs (10,11,14). Interestingly, we also found that XcpS F , the polytopic IM component of the AP, directly interacts with XcpY L and XcpZ M , thus confirming the physical interconnection between components of the AP. In addition and as previously observed by Lallemand et al. (14), full-length XcpZ M does not self-dimerize (or self-dimerization is very weak) when produced alone in the E. coli membrane (Fig. 4). This negative result is not due to nonfunctional fusion proteins because both T18/25 XcpZ M fusions give positive signals when combined with XcpY L and XcpS F partners (Fig. 4). This result contrasts with the homodimerization property of the periplasmic domain of XcpZ M revealed by size-exclusion chromatography (Fig. S3) and suggests that XcpZ M periplasmic homodimerization might be prevented by its transmembrane domain.
We then decided to challenge this AP interactome in the presence of the T2SS effector CbpD (Fig. 4, dark gray bars). CbpD was therefore coproduced in the periplasm of the BTH101 strains producing the various Xcp T18/T25 pairs. No difference was seen for the pairs involving XcpS F , indicating that CbpD binding to XcpY L has no significant effect on the XcpY L /S F interaction. In contrast, a statistically significant increase in ␤-gal activity was measured in the presence of CbpD for the T18-Z M /T25-Z M pair, showing that the presence of CbpD triggers XcpZ M homodimerization, possibly through

Effector recognition in type 2 secretion
their periplasmic domains (Fig. S3). Similarly, CbpD coproduction was also performed with the T18-Y L /T25-Z M pair. In this case, the presence of the effector significantly increased the ␤-gal levels, revealing a significant strengthening of the interaction between XcpY L and XcpZ M in the presence of the secreted effector. Those findings indicate that the presence of the effector triggers structural rearrangements of the AP, thus suggesting a possible synchronization between effector arrival and activation of the system.

Discussion
Effector recognition by the T2SS remains enigmatic because no common secretion signal has been identified in the numerous effectors reported so far. All attempts to identify the secretion signal of the T2SS have converged to the existence of a still unknown conformational signal, in agreement with the folded state of the effectors prior to recognition by the secreton (41). Here, we focused on the Xcp T2SS of P. aeruginosa, which secretes at least 19 different exoproteins, to study effector recognition and transport. Based on the identification of a set of direct periplasmic interactions between secreted effectors and components of the three subcomplexes of the secreton, we previously proposed a model of effector recognition and transport by the T2SS (34). In this model, the substrate is recognized by the secreton peripheral component XcpP C and then transferred into the secretin (XcpQ D ) vestibule to be expelled out of the secretin pore upon contact with the pseudopilus tip (XcpU H V I W J X K ). In the present study, we completed the Xcp/ effector periplasmic interactome by testing XcpY L and XcpZ M , the last two Xcp components harboring a periplasmic domain not included in previous studies. We applied two complementary protein-protein interaction approaches and found that the two AP components directly and specifically interact with the secreted effector CbpD through their periplasmic domains. This brings to five the number of Xcp partners physically and specifically encountered by the effector during the secretion process, thus indicating a system-specific route for substrate recruitment and transport all along the Xcp T2SS secreton.
IMAC experiments revealed a competition between XcpY L and XcpZ M for the effector, suggesting interaction sequentiality. However, it is difficult at this stage to establish an order for these two interactions and propose a hierarchical positioning in the previous T2SS secretion model. Our data nevertheless show for the first time a direct involvement of both GspL and GspM in substrate recognition.
We further evaluated the possible consequences of these new interactions on the T2SS by addressing their impact on AP dynamics. Using BACTH as a quantitative protein-protein interaction technique, we first established the interactome network among the three AP components, XcpY L , XcpZ M , and XcpS F . We then found that XcpZ M oligomerization and XcpY L /XcpZ M heterodimerization are, respectively, triggered and strengthened upon effector binding. Those important observations constitute the first experimental evidence of effector-mediated conformational changes of the T2SS AP components, suggesting a possible synchronization between effector arrival and activation of the system. One possible type of activation mediated by substrate binding was described in a recent report by Lallemand et al. (14). In this work, using an elegant combination of in vivo protein-protein interaction and crosslinking experiments, the authors showed the molecular details of the dynamic interplay between full-length GspL and GspM. In their model, upon sensing an unknown signal, the interactions between GspL and GspM periplasmic domains shift from homo-to heterodimeric, mediating coordinated shifts or rotations of their cognate TM domains. We propose that the dynamic interplay leading to signal transduction from the periplasm to the cytoplasm is triggered by effector binding on the GspM and/or GspL periplasmic domain. It would be interesting to localize the effector-binding domains in XcpY L and XcpZ M and their possible overlap with the ferredoxin-like domains of GspL and GspM directly involved in the dynamic interplay.
The next obvious question brought by our observations is what is the target of the signal transduced from the periplasm to the cytoplasm? A possible scenario, elaborated from the numerous interactions reported between the cytoplasmic domain of GspL and its partner, the ATPase GspE, was recently proposed by Gu et al. (4). Taking into account that activation of the ATPase necessitates an interaction between membrane lipids and the GspL's segment adjacent to the TM domain (27), the authors propose that activation of the ATPase could be medi-

Effector recognition in type 2 secretion
ated by a displacement of GspL in the inner membrane, itself induced by the transmembrane signal generated by effector binding in the periplasm. Therefore, considering that activation of the ATPase mediates pseudopilus assembly (42,43), we propose that effector binding on the periplasmic domains of GspL and GspM triggers pseudopilus assembly, thus synchronizing this late step with effector arrival in the machinery. Therefore, an additional effector-sensing step, coupled with pseudopilus assembly, can be added to our model of effector recognition and transport by the T2SS (Fig. 5). Whether this effector sensingstep mediated by GspL and GspM is linked with their direct involvement in pseudopilus assembly (10,43) remains to be determined.
Further investigations are now required to understand, at the molecular level, the chronology of the effector/XcpY L and effector/XcpZ M interactions and the effectors' subsequent transition into the secretin vestibule as well as the structural basis of the transmembrane signal transduction. In this respect, interestingly, the XcpY L periplasmic domain presents structural homology with proteins harboring the Per-Arnt-Sim (PAS) domain. These PAS domains are present in all three domains of life and are involved in signaling and transmembrane signal transduction, such as in the bacterial two-component systems (44). This structural homology supports the sensing properties of GspL, which may play an important role in T2SS function.

Bacterial strains and plasmids
Bacterial strains and plasmids used in this study are listed in Table 1.

DNA manipulation
Plasmid preparation, DNA purification, gel extraction, and PCR product purification were performed using appropriate Macherey Nagel kits. Restriction enzymes, DNA polymerase, and other molecular biology reagents were purchased from New England Biolabs or Promega. The high-fidelity polymerase Q5 (Biolabs) was used for PCR amplification. The list of oligonucleotides (synthesized by Integrated DNA Technologies, Inc.) used for cloning is provided in Table 2. To construct the BACTH plasmids, the xcp genes were PCR-amplified using corresponding primers and cloned into pKT25 and pUT18C vectors using the SLIC method between BamHI and EcoRI sites. To construct the expression plasmids for heterologous reconstitution, the xcpY L and xcpZ M genes were PCR-amplified using corresponding primers and cloned into the pCDFDuet vector using the SLIC method or digestion ligation methods between NcoI and SalI (MCS1) sites for XcpZ M p or NdeI and EcoRV (MCS2) sites for XcpY L p. The CbpD H gene was subcloned from pT7.5 vector to pETDuet vector using the EcoRI site. All plasmids were sequenced by GATC Co.

Protein production and purification
The DNA sequences encoding the periplasmic domains of XcpY L (XcpY L p; from residue 255 to residue 381) and XcpZ M (XcpZ M p; from residue 53 to residue 173) were cloned into pLIC07 vector using the SLIC method between BsaI sites. These constructs allow the production of XcpY L p and XcpZ M p proteins fused to thioredoxin (Trx) at their N terminus. The Trx is cleaved off after purification, and the resulting proteins are soluble, stable, and produced in sufficient amount for biochemical and biophysical characterization. Competent cells of

Effector recognition in type 2 secretion
strain BL21 (DE3) were transformed with pLIC-XcpZ M p or pLIC-XcpY L p. The bacteria were grown until reaching an A 600 nm of 0.5 at 37°C on TB medium with kanamycin at 50 g/ml. Induction was performed with 0.1 mM IPTG for 12 h at 25°C. Bacteria were collected by centrifugation and broken by sonication (4 ϫ 1 min) in cold buffer (50 mM Tris-HCl, pH 8, 300 mM NaCl, 1 mM EDTA, 20 mM MgCl 2 , 1 mM phenylmethylsulfonyl fluoride, 0.5 mg/ml lysozyme, 20 g/ml DNase, 10 mM imidazole). The lysate was cleared by ultracentrifugation (20,000 ϫ g) to remove unbroken debris and membranes. The cleared lysates containing Trx-XcpY Lp and Trx-XcpZ Mp were loaded onto a 5-ml nickel column (HisTrap TM FF) using an ÄKTA Prime apparatus (GE Healthcare), and the immobilized proteins were eluted in buffer B (50 mM Tris-HCl, pH 8.0, 300 mM NaCl, 500 mM imidazole). XcpY L p and XcpZ M p were obtained after cleavage of the Trx fusion using 2 mg of tobacco etch virus protease for 18 h at 4°C and dialysis in a dialysis bag to remove imidazole. Untagged soluble proteins were then collected in the flow-through of a 5-ml nickel column; the histidine-tagged tobacco etch virus and Trx proteins remained bound to the column. The proteins were concentrated using Centricon technology (Millipore; 10-kDa cutoff) and subjected to size-exclusion chromatography purification using a HiLoad Superdex200 16/600 column pre-equilibrated with 50 mM Tris-HCl, pH 8, 150 mM NaCl.
For CbpD, the extraction and purification protocols of the periplasmic material were described previously (5). Purity and quality of the purified proteins were checked by analyz- Figure 5. Schematic representation of the effector-sensing step mediated by XcpY L /Z M during secretion process by T2SS. ➊, effector (blue sphere) binding to the periplasmic domains of GspL Y (L) and GspM Z (M) induces their homo-and/or hetero-oligomerization (purple arrows). ➋, the effector-mediated gathering of GspM Z and GspL L generates a transmembrane signal (red flash) triggering GspE R (E) activation (ATP to ADP) and pseudopilus (GspG T H U I V J W K X (GHIJK)) assembly and elongation (dashed red arrows). ➌, the growing pseudopilus interacts with the effector, which first transfers it inside the vestibule of the secretin and then leads to its translocation in the extracellular milieu. Also represented on this cartoon are the secretin connector GspC P (C) and the polytopic AP component GspF S (F), which is involved in pseudopilus assembly upon cycles of GspE R -mediated rotation (circular red arrows).

Effector recognition in type 2 secretion
ing samples by SDS-PAGE followed by Coomassie Blue staining.

Coproduction of XcpY p , XcpZ p , LapA H , and CbpD H in E. coli and affinity chromatography
Competent cells of strain BL21 (DE3) were cotransformed with pCDFDuet, pETDuet, and derivatives. The bacteria were grown until reaching an A 600 nm of 0.5 at 37°C on LB or TB medium with appropriate antibiotics (30 g/ml streptomycin, 50 g/ml ampicillin). Induction was performed with 0.1 mM IPTG for 12 h at 17 or 25°C. Bacteria were collected by centrifugation and broken using a French press (10,000 p.s.i.) in cold buffer (50 mM Tris-HCl, pH 8, 150 mM NaCl, 1 mM EDTA, 20 mM MgCl 2 , 1 mM phenylmethylsulfonyl fluoride, 0.5 mg/ml lysozyme, 20 g/ml DNase). The lysate was cleared by ultracentrifugation to remove unbroken debris and membranes. The cleared lysate was loaded onto a 1-ml nickel column (HisTrap FF) using an ÄKTA Prime apparatus. The immobilized proteins were eluted in buffer B (50 mM Tris-HCl, pH 8, 150 mM NaCl, 500 mM imidazole). The loaded, flow-through, and elution fractions were analyzed by SDS-PAGE and immunodetection. For competition experiments, the complete CbpD copurification IMAC procedure for XcpY L , XcpZ M , and XcpY L together with XcpZ M was performed in quadruplicate. The total amount of XcpY L or XcpZ M proteins in the eluate fractions (in percent; compared with the total amount in the loaded fraction) was measured and quantified from immunoblots of each replicate using ImageJ software. Microsoft Excel software was used for data processing and presentation. Statistics were determined using the Student's t test function of Excel using a bilateral model and assuming equal variance.
To study the binding of CbpD to XcpY L p or XcpZ M p, increasing concentrations of XcpY L p (5-160 M) and XcpZ M p (6.25-100 M) were used, and the association and dissociation phases were monitored for 1000 and 3000 s, respectively. Xcp proteins were dialyzed against IB before titration experiments. To avoid nonspecific binding of XcpY L p or XcpZ M p to the streptavidin biosensors, the biosensors were incubated with 10 ug/ml biocytin (Sigma) for 200 s. In all experiments, the response of the nonbiotinylated proteins on the free sensors was subtracted during experiment processing.
The K D values were calculated using GraphPad Prism 5.0 software on the basis of the steady-state levels of the responses in nanometers, directly related to the concentration of the Xcp protein. The K D was calculated from a triplicate experiment by plotting on the x axis the different concentrations of the Xcp protein and the different responses of the Xcp protein at saturation (990 s after the start of the association step) on the y axis. Nonlinear regression fitting for xy analysis and a one-site binding (specific binding) model, corresponding to the equation y ϭ B max ϫ x/(K D ϩ x), were used to calculate K D values.

Effector recognition in type 2 secretion Bacterial two-hybrid system and statistical analysis
To investigate the interaction between XcpZ M and XcpY L periplasmic domains, competent cells of strain BTH101 were cotransformed with pUT18C and pKT25 derivatives, and bacteria were grown for 48 h at 30°C on LB plates containing Ap100, Kan50, and Sm100. Colonies were picked at random; inoculated into 600-l cultures in LB containing Amp100, Kan50, Sm100, and 0.5 mM IPTG; and grown overnight at 30°C. ␤-Gal activity was measured as described (45). At least two independent experiments were performed with three randomly picked transformants. Mean values are presented as bar graphs, and error bars indicate S.D. Microsoft Excel software was used for data processing and presentation. Statistics were determined using the Student's t test function of Excel using a bilateral model and assuming equal variance.
To study the interaction network among the full-length proteins, competent cells of BTH101 containing pJN105 or pJN105-CbpD vectors were cotransformed with pUT18C and pKT25 derivates. Bacteria were grown for 48 h at 30°C on LB plates containing Ap100, Kan50, Sm100, and Gm15. Colonies were picked at random and inoculated into 600-l cultures in LB containing Ap100, Kan50, Sm100, Gm15, 0.5 mM IPTG, and 0.5% arabinose to allow CbpD production. Cells were grown overnight at 30°C, and ␤-gal activity was measured as described (45).