The Molecular Complex between Staphylococcal Adhesin SpsD and Fibronectin Sustains Mechanical Forces in the Nanonewton Range

Binding of Staphylococcus pseudintermedius surface proteins SpsD and SpsL to fibronectin (Fn) plays a critical role in the invasion of canine epithelial cells. Here, we discover that both adhesins have different mechanisms for binding to Fn. The force required to separate SpsD from Fn is extremely strong, consistent with the unusual β-sheet organization of a high-affinity tandem β-zipper. By contrast, unbinding of the SpsL-Fn complex involves the sequential rupture of single weak bonds. Our findings may be of biological relevance as SpsD and SpsL are likely to play complementary roles during invasion. While the SpsD β-zipper supports strong bacterial adhesion and triggers invasion, the weak SpsL interaction would favor fast detachment, enabling the pathogen to colonize new sites.

immunodeficiencies can lead to skin infections such as pyoderma caused by this organism (1). In addition, over the last 2 decades, methicillin-resistant S. pseudintermedius has emerged as a major problem in veterinary clinics worldwide (2,3). Several episodes of life-threatening human infections by S. pseudintermedius have also been reported, mainly after contacts with dogs (4,5).
In staphylococci, a family of cell wall-anchored surface proteins termed microbial surface components recognizing adhesive matrix molecules (MSCRAMMs) mediate bacterial adherence to extracellular matrix proteins of the host (6). Compared to Staphylococcus aureus, the interaction of S. pseudintermedius with host proteins is less characterized, but several strains have been shown to bind to fibronectin (Fn), fibrinogen, cytokeratin 10, elastin, collagen type I, vitronectin, and laminin (7,8). A genomewide screen revealed 18 genes encoding putative cell wall-anchored S. pseudintermedius surface proteins (9,10). Of these, Fn-binding proteins SpsD and SpsL are believed to be important in host tissue colonization and infection (9). The primary translation product of the spsD gene contains 1,031 residues, has an N-terminal secretory signal sequence and a C-terminal cell wall-anchoring domain comprising an LPDTG motif, a hydrophobic transmembrane domain, and a short sequence rich in positively charged residues. The N-terminal end of SpsD consists of an A domain 40% identical to the fibrinogen-binding domain of FnBPB from S. aureus and is involved in binding to fibrinogen, cytokeratin-10, and elastin (11). This domain is followed by a connecting region, region C, which interacts with Fn, and the repeat region R (12) (Fig. 1A). SpsL is a protein of 930 residues that includes a signal sequence at the N terminus followed by a fibrinogen-binding A domain with three IgG-like folds (N1 to N3) (13), an R domain containing seven tandem repeats that confer Fn-binding capacity (12), and a C-terminal sorting signal (Fig. 1A).
In S. aureus, cellular invasion is triggered by the interaction between Fn-binding proteins FnBPA and FnBPB and the ␣ 5 ␤ 1 integrin in the host cell membrane (12). The key to this process is the formation of a Fn bridge between FnBPs and integrin. Soluble Fn is made of multiples modules (Ͻ100 amino acids [aa]) called type I, II, III repeats. FnBPA features eleven nonidentical, unfolded Fn-binding repeats (FnBRs) that bind with four sequential modules of the N-terminal FI domain via a tandem ␤-zipper featuring an unusual ␤-sheet organization (14,15).
While Fn binding by SpsL and SpsD supports invasion of canine epithelial cells by S. pseudintermedius (12), the molecular interactions involved are not known. Two important yet unsolved questions are the following. (i) What are the binding strengths of SpsD and SpsL? (ii) As the two adhesins fulfil similar functions, do they share the same binding mechanism? Using single-molecule atomic force microscopy (AFM) experiments (16,17), we show that SpsD and SpsL are engaged in very different interactions with Fn. While SpsL and Fn form weak bonds, SpsD binds Fn via extremely strong forces, reflecting the ␤-sheet organization of a tandem ␤-zipper. These results may contribute to the development of antiadhesion approaches to treat infections caused by S. pseudintermedius and other bacterial pathogens engaged in tandem ␤-zipper interactions.

RESULTS
SpsD and SpsL favor bacterial adhesion to immobilized Fn. We studied Fn binding using S. pseudintermedius ED99 mutant spsL and spsD strains expressing either SpsD or SpsL. As controls, we used cells expressing SpsD and SpsL (wild-type [WT] cells) and mutant cells lacking both adhesins (spsD spsL mutant cells). As reported previously (10), deletion of one type of adhesin does not affect expression of the other.
Using optical microscopy, we found that spsL mutant cells adhered strongly to Fn-coated substrates, while spsD mutant and WT cells showed lower levels of adhesion (Fig. 1B). Poor adhesion was observed with the double mutant, implying that SpsD and SpsL are the only Fn-binding proteins expressed at the cell surface. The same adhesion pattern was observed in an assay in which the bacteria were allowed to attach to microtiter wells coated with Fn (Fig. 1C). The higher level of adhesion of the spsL mutant compared to that of the WT strain could be due to a better exposure of SpsD proteins to the ligand on the bacterial surface. It is also possible that, unlike in the WT where SpsD and SpsL proteins compete with each other for Fn binding, no such competition occurs in the spsL mutant and Fn is only captured by SpsD.
Moreover, adhesion of cells from the WT and single-mutant strains was inhibited in the presence of the N-terminal fragment of Fn (N29), indicating that this region is involved in Fn binding by SpsD and SpsL as for FnBPA (Fig. 1D). To further evaluate the specificity of Fn binding, the active regions (based on the sequence alignment with S. aureus FnBPA) SpsD 520Ϫ846 (C domain, connecting region) and SpsL 538Ϫ823 (R domain, repetitive region) were immobilized onto microtiter wells and tested for their ability to bind various extracellular matrix and plasma proteins, i.e., Fn, fibrinogen, collagen type I, factor H, and factor I. Substantial binding was only observed for Fn, demonstrating that the latter interacts specifically with SpsD 520Ϫ846 and SpsL 538Ϫ823 (Fig. 1E).
SpsD and SpsL bind to Fn with different affinities. We analyzed the interaction of Fn with SpsD 520Ϫ846 and SpsL 538Ϫ823 using surface plasmon resonance (SPR) by immobilizing the bacterial domains on a chip and injecting Fn in the mobile phase ( Fig. 2A and B). The best fit of the data points was obtained with the Langmuir isotherm equation describing a one-site binding model. From this analysis, we obtained dissociation constant (K D ) values of 1.2 Ϯ 0.4 nM and 50 Ϯ 3 nM for the SpsD-Fn and SpsL-Fn complexes, respectively. This shows that both adhesins strongly bind to Fn but that SpsD clearly exhibits a higher affinity than SpsL.
SpsD and SpsL share conformational epitopes with FnBPA. We also wondered if ligand-induced binding site (LIBS) monoclonal antibodies (MAbs) could recognize epitopes involved in Fn (or N29) binding in the SpsD 520Ϫ846 and SpsL 538Ϫ823 domains. We previously showed that a family of LIBS MAbs against S. aureus FnBPA recognizes specific repeats of the adhesin in the presence of Fn; specifically, MAbs 6B7 and 7D4 recognized conformational neoepitopes in the FnBPA-5 and FnBPA-9 repeats, respectively (18). We found that MAbs 6B7 and 7D4 strongly reacted with the C region of SpsD (Fig. 2C), while only 6B7 showed reactivity for the SpsL fragment (Fig. 2D). Although the connecting C region of SpsD and the repetitive region of SpsL bind to Fn, the conformational epitopes present in SpsD are different from those present in the R region of SpsL. Hence, 7D4 recognizes a specific epitope that is only formed when SpsD, but not SpsL, binds to Fn. We hypothesize that these regions may have a flexible structure that can shift from a disordered to an ordered structure in the presence of Fn, as in the FnBPA-Fn interaction. To support this, the secondary structures of the connecting region of SpsD (SpsD 520Ϫ846 ) and the repetitive domain of SpsL (SpsL 538Ϫ823 ) were analyzed by circular dichroism (CD) (see Fig. S1 in the supplemental material). Both proteins featured quite similar CD spectra and harbored significant unordered regions (43% random coil). The reactivity of LIBS antibodies was correlated with sequence alignments of the FnBPA-5 and FnBPA-9 repeats versus the Fn-binding regions of SpsD 520Ϫ846 and SpsL 538Ϫ823 ( Fig. 2C and D, right). A high degree of identity/similarity was observed for SpsD as follows: 64%/92% for FnBPA-5 644Ϫ672 versus the SpsD 607Ϫ634 sequence and 56%/72% for FnBPA-9 783Ϫ801 versus the SpsD 682Ϫ700 sequence. There was also identity and similarity between FnBPA-5 639Ϫ662 and SpsL 562Ϫ585 (43%/65%) yet to a lower extent than that in SpsD, while FnBPA-9 showed no identity/similarity to SpsL 538Ϫ823 . So, there is a higher level of identity/ similarity for SpsD than for SpsL. Together, these data suggest that S. pseudintermedius SpsD and SpsL share with S. aureus FnBPA disordered epitopes that acquire an ordered structure upon binding to Fn. As a proof of the specificity of the formed neoepitopes recognized by MAbs 6B7 and 7D4, no reactivity for SpsD and SpsL was exhibited by the MAbs 1F9 and 5G3 ( Fig. 2C and D).
S. pseudintermedius engages in two modes of interaction with Fn. The molecular interactions of SpsD and SpsL were first studied by measuring the forces between a single bacterium and Fn-coated surfaces. Figure 3A shows the adhesion forces and rupture lengths obtained for three representative spsL mutant cells (for more cells, see ; n ϭ 111 adhesive curves), 1,754 Ϯ 174 pN (n ϭ 85), and 1,541 Ϯ 154 pN (n ϭ 74) for cells 1, 2, and 3, respectively. These forces were specific to SpsD, as they were abolished in spsD spsL mutant cells (Fig. 3D). The rupture length is the distance from the contact point to the unbinding point. Most bonds ruptured around ϳ300 nm, but some ruptures up to ϳ700 nm were also observed. Assuming that the processed mature adhesin comprises 1,031 residues and that each amino acid contributes 0.36 nm to the contour length of the polypeptide chain, the fully extended protein should be ϳ371 nm long. This The spsD mutant cells featured very different interactions in that strong adhesion was never observed. Instead, weak forces of 168 Ϯ 93 pN (mean Ϯ SD; n ϭ 246 adhesive curves; 3 cells) and 219 Ϯ 111 nm rupture lengths were observed ( Fig. 3B; for more cells, see Fig. S2B). These forces were specific as their frequency was largely reduced in the double mutant (from 66% to 8%; mean from 3 cells) (Fig. 3D). WT cells featured a combination of both weak and strong forces (Fig. 3C), which is not surprising  as they express both adhesins. Strikingly, spsD mutant cells, but not spsL mutant cells, featured adhesive curves (ϳ10% of all adhesive curves) with sawtooth patterns with successive unbinding events (7 to 15) of 141 Ϯ 30 pN magnitude and peak-to-peak distance of 28 Ϯ 5 nm, matching the 28-nm unfolding distance of FnIII repeats (Fig. 3E and F) (19). Therefore, rupture of the SpsL-Fn complex seems to be associated with the unfolding of multiple FnIII repeats. In SpsD, unfolding of FnIII repeats was not observed. Another set of force curves (ϳ5%) showed multiple peaks separated by 10 Ϯ 2 nm, consistent with the unfolding of SpsL repeats (7 ϫ 37 residues) ( Fig. 3E and G). This fits with a bond rupture in which individual SpsL repeats unfold sequentially and detach from the Fn region. Collectively, these results show that SpsD and SpsL mediate bacterial adhesion to Fn through different interactions, i.e., single strong forces versus multiple weak forces, therefore, suggesting that distinct ligand-binding mechanisms occur.
The SpsD-Fn interaction is extremely strong. To study the mechanical strength of individual bonds, single adhesins were picked up and pulled with an AFM tip modified with Fn (Fig. 4). Strong unbinding forces were measured between spsL mutant cells and Fn tips ( , implying that Fn is in a globular form or only partially extended. The ϳ158 pN force is close to that of whole cells (ϳ168 pN) and agrees well with the binding strength between Fn-and single FnBPA-binding repeats (ϳ200 pN [19]). Again, a number of curves showed multiple peaks separated by ϳ10 nm, consistent with the unfolding of individual SpsL repeats. Force peaks were well described by the worm-like chain (WLC) model (Fig. 4B, insets and red lines) using a persistence length of 0.4 nm as follows: where L c and l p are the contour length and persistence length of the molecule, k b is the Boltzmann constant, and T is the absolute temperature.
We believe that the forces reported herein mostly reflect single interactions since (i) force distributions were narrow and did not feature intermediate values usually associated with multiple bonds breaking simultaneously, (ii) single-cell and singlemolecule experiments lead to very similar sharp distributions, and (iii) 1-ethyl-3-(3dimethylaminopropyl)-carbodiimide (EDC)/N-hydroxysuccinimide (NHS) chemistry, used in this study, has been shown to favor single-molecule detection over the years, which is further supported by the occurrence of single well-defined peaks (versus multiple peaks).
Single-molecule mapping revealed that SpsD binding events were localized heterogeneously and formed nanodomains (Fig. 4A, left insets), while SpsL was randomly localized on the bacterial cell surface (Fig. 4B, left insets), suggesting a different binding mechanism. Finally, plots of the binding strengths versus rupture lengths ( Fig. 4E and F) emphasize the strong similarities between single-molecule and whole-cell experiments and demonstrate the major difference between the strong, long-range SpsD interaction and the weak, short-range SpsL interaction.

Dynamics of the SpsD-Fn interaction.
To investigate the dynamics of the SpsD-Fn interaction, force curves were recorded while varying the rate at which force is applied (loading rate [LR]). Typically, the unbinding force of specific bonds increases with the LR, an observation that has been widely described by the Bell-Evans (20) (22,23). These proteins bind their ligand through the multistep dock, lock, and latch (DLL) mechanism involving dynamic conformational changes of the protein. Upon increasing the LR, the binding strength of ClfA and ClfB is dramatically enhanced from 100 to 250 pN to ϳ1,500 pN, similar to a catch bond mechanism. We, therefore, asked whether the SpsD-Fn complex follows one of these two behaviors. The strength of the interaction (F) was measured at different LRs (the effective LR was estimated from the force versus time curves). The probability of forming strong bonds decreased with the LR (Fig. 5A), while the binding strength remained unchanged. We also found that the bond is weak (ϳ0.25 nN) at low tensile force but is dramatically enhanced (ϳ2 nN) by mechanical tension as observed with catch bonds (Fig. 5B and C).  We hypothesize that the strong bond is activated by tensile force and that, when it is loaded quickly, the interaction time between SpsD and Fn might be too short to allow for conformational changes and optimal fitting between the active binding sequences. Supporting this idea, we found that increasing the interaction time from 100 to 600 ms increased the probability of forming strong SpsD-Fn bonds (Fig. 5C), while it had no effect on SpsL-Fn bonds (Fig. 5D).

DISCUSSION
We have shown that the SpsD-Fn interaction is extremely strong, which, together with the recently discovered DLL binding mechanism (24, 25), represents the highest mechanical strength reported to date for a noncovalent biological interaction. As in DLL Binding Strength of Staphylococcal Adhesin SpsD ® complexes, the strong SpsD-Fn interaction is activated by mechanical tension as observed with catch bonds (26).
Owing to single-molecule experiments, we are now starting to appreciate that pathogens have evolved molecular interactions that are extremely strong, enabling them to firmly attach to their host during colonization and infection. The SpsL-Fn interaction is much weaker, with a binding strength similar to that of classical receptorligand complexes (Ͻ0.2 nN). Interaction strengths correlate with dissociation constants, with SpsD featuring a remarkably high affinity (K D of ϳ1 nM versus ϳ50 nM for SpsL). These observations suggest that SpsD and SpsL have different mechanisms for binding to Fn, which is unexpected and surprising as the two adhesins share sequence similarity with S. aureus FnBPs and fulfill the same invasion function.
We speculate that the mechanostability of the SpsD-Fn interaction originates from the ␤-sheet organization of a tandem ␤-zipper (Fig. 6, top) as identified previously for FnBPs, Sfb1, and BBK32 from the pathogens Staphylococcus aureus (14), Streptococcus pyogenes (14), and Borrelia burgdorferi (27), respectively. When SpsD binds to FnI modules, its intrinsically disordered connecting region (region C) would shift into an ordered structure by forming additional ␤-strands along triple peptide ␤-sheets in the Fn molecule. We can also hypothesize that like the allosteric regulation of the Fn-␣ 5 ␤ 1 interaction by FnBPA (28), the globular form of soluble Fn undergoes a conformational change to an extended form so that ligand (integrin)-binding sites on FnIII modules are exposed and become available for interaction. This model is supported by LIBS MAbs  and sequence alignments revealing that SpsD shares with FnBPA disordered epitopes that acquire an ordered structure upon Fn binding. The use of LIBS antibodies has also provided cues to identify minimal Fn-binding units in both SpsD and SpsL, and this can pave the way for designing peptide analogs with inhibitory potential on Fn interactions (18).
Moreover, the SpsD-Fn complex has a high affinity of 1 nM, in line with the work of Meenan et al. (18), who found that 6 of the 11 Fn-binding sites of FnBPA bind with dissociation constants in the nanomolar range. The strong, well-defined SpsD adhesion peaks are consistent with an unbinding model in which multiple bonds of the SpsD-Fn ␤-zipper rupture simultaneously and cooperatively, thus resulting in a mechanically stable complex (Fig. 6, top). Structural analysis of the FnBPA-Fn ␤-zipper have suggested that electrostatic interactions, hydrophobic forces, and hydrogen bonds are likely to be important for binding (14,29). Further molecular dynamics simulations and structural analysis will greatly contribute to understanding the molecular bonds behind the extreme mechanostability of the SpsD-Fn interaction. Milles et al. (24) and Herman-Bausier and Dufrêne (25) applied this approach to the DLL interaction between the staphylococcal adhesin SdrG and fibrinogen. They discovered that the target peptide is confined in a screw-like manner in the binding pocket and that the binding strength of the complex results from numerous hydrogen bonds between the peptide backbone and the adhesin.
The interaction between SpsL and Fn involves weak bonds that rupture sequentially, indicating that they are not engaged in a strong ␤-zipper complex (Fig. 6, bottom). That SpsL requires much lower separation force than SpsD is likely to be due to sequence diversity, resulting in conformational differences between the adhesins. This hypothesis is supported by recent studies showing that amino acid substitutions in the repeat region of FnBPA significantly affect bond strength and influence the conformation of Fn upon binding (19,30). Therefore, intrinsically disordered sequences from both adhesins may bind to FnI ␤-sheets, but SpsL would not form a ␤-zipper because of differences in the spacing, flexibility, and conformation of active sequences. SpsL-Fn rupture forces feature multipeaks that fit with the unfolding of single SpsL FnBRs, while these are intrinsically disordered in their native state. Perhaps FnBRs become more ordered upon Fn binding. Due to the weak SpsL-Fn interaction, Fn remains in a compact conformation as in its soluble state, and upon pulling the complex apart, Fn becomes partially extended explaining the short extensions we observed (ϳ100 nm) ( Fig. 4E and F). Upon pulling the complex apart, Fn becomes partially extended and some FnIII modules become unfolded (26) (Fig. 6, bottom). The weak binding strength correlates with the lower level of identity/similarity of SpsL with FnBPA compared to that of SpsD, as well as with its relatively lower affinity for Fn. We propose that binding strengths (at nonequilibrium) are more appropriate than binding affinities (at equilibrium) to understand the binding mechanism of SpsD-Fn complexes. As most surface-attached bacteria are subjected to physical stress, it seems more appropriate to study their molecular bonds under force as in the present study. Also, it is known that adhesins sharing a common binding mechanism to Fn do not have similar affinity. K D values for FnBPA/ FnBPB are in the range of 1 to 10 nM, while the N terminus of BBK32 is in the order of 100 nM.
The different interaction strengths of SpsD and SpsL could be of biological significance in that they may play complementary roles in invasion and dissemination. There is evidence that invasion by S. pseudintermedius involves an Fn bridge between SpsD or SpsL and the ␣ 5 ␤ 1 integrin in the host cell membrane (12). We, therefore, postulate that the strong SpsD-Fn ␤-zipper may expose high-affinity RGD sites, as in the FnBPA ␤-zipper, leading to a mechanically stable bridge. Formation of SpsD-Fn nanodomains on the bacterial cell surface would favor integrin clustering and invasion. The weak SpsL-Fn interaction and lack of clustering would mediate moderate adhesion to the ␣ 5 ␤ 1 integrin and favor detachment of the pathogens from the host cell surface, thus enabling the colonization of new sites. residue weight of 110. All measurements were performed in 20 mM phosphate buffer, pH 7.4, at 25°C. Ten scans were averaged for each spectrum, and the contribution from the buffer was subtracted in each case. Quantification of secondary structural components was performed using the deconvolution programs CONTIN, CDSSTR, and SELCON3, and the values reported are an average of results obtained.
Single-cell force spectroscopy. Colloidal probes were prepared as described earlier (34). The nominal spring constant of cantilevers was determined by the thermal noise method, giving an average value of ϳ0.08 N/m. Briefly, 50 l of a suspension of ca. 1 ϫ 10 6 cells was transferred into a glass petri dish containing Fn-coated substrates on the other corner, the whole being immersed in PBS. The colloidal probe was brought into contact with a bacterium, which is first caught through electrostatic interactions with polydopamine. The cell probe was then positioned over Fn substrates without dewetting. Cell probes were used to measure interaction forces on Fn surfaces at room temperature by recording multiple force curves (16 by 16) on different spots, a maximum applied force of 250 pN, and approach and retraction speeds of 1,000 nm s Ϫ1 and a contact time of 100 ms. Data were analyzed with the data processing software from JPK Instruments (Berlin, Germany). Adhesion force and rupture distance histograms were obtained by calculating the adhesion force and rupture distance of the last peak for each curve. At least 10 cells of each strain from 3 independent cultures were probed.
Single-molecule force spectroscopy. Cantilevers (k, ϳ0.02 N/m) were prepared as described above, and bacteria were immobilized on polystyrene substrates. Measurements were performed at room temperature in PBS buffer with a NanoWizard 4 atomic force microscope (JPK Instruments). Adhesion maps were obtained by recording 32 by 32 force-distance curves on areas of 500 by 500 nm 2 with an applied force of 250 pN, a constant approach and retraction speed of 1,000 nm · s Ϫ1 , and a contact time of 100 ms. For some experiments, the contact time was increased from 100 ms to 200 and 600 ms. For loading rate experiments, arrays of 32 by 32 force curves were recorded on 500-nm by 500-nm areas at increasing retraction speeds as follows: 0.5, 1, 3, 5, and 10 m · s Ϫ1 . Single-molecule data were processed and analyzed the same as for single-cell experiments. Adhesion force and rupture distance histograms were obtained by calculating the adhesion force and rupture distance of the last peak for each curve. At least 10 cells of each strain from 3 independent cultures were probed.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.