Structural Basis of Ca2+-Dependent Self-Processing Activity of Repeat-in-Toxin Proteins.

The Ca2+-dependent clip-and-link activity of large repeat-in-toxin (RTX) proteins is an exceptional posttranslational process in which an internal domain called a self-processing module (SPM) mediates Ca2+-dependent processing of a highly specific aspartate-proline (Asp-Pro) peptide bond and covalent linkage of the released aspartyl to an adjacent lysine residue through an isopeptide bond. Here, we report the solution structures of the Ca2+-loaded SPM (Ca-SPM) defining the mechanism of the autocatalytic cleavage of the Asp414-Pro415 peptide bond of the Neisseria meningitidis FrpC exoprotein. Moreover, deletion of the SPM domain in the ApxIVA protein, the FrpC homolog of Actinobacillus pleuropneumoniae, resulted in attenuation of virulence of the bacterium in a pig infection model, indicating that the Ca2+-dependent clip-and-link activity plays a role in the virulence of Gram-negative pathogens.

restraints is displayed in Fig. 1C. The Ca-SPM structure has a compact fold composed of eight anti-parallel ␤-strands connected by a single helix-turn-helix motif and several surface-exposed turns and loops (Fig. 1D). These structures appear to be stabilized by calcium ions as the overall structure of Ca-SPM is well defined with a backbone root square mean deviation (RMSD) of 0.70 Ϯ 0.16 Å (Table 1). However, the most abundant calcium isotope, 42 Ca, is not observable by NMR, and the number and exact position of calcium ion(s) could not be identified by this NMR approach.
Calcium stoichiometry for Ca-SPM. In order to determine the number of Ca 2ϩbinding sites in SPM, Ca-SPM was dialyzed against a calcium-free buffer to remove free Ca 2ϩ ions and was next subjected to NMR titration with the metal-chelating agent EDTA. The extensive dialysis of Ca-SPM did not affect the overall fold of the protein, indicating a high stability of the Ca-SPM structure at residual free Ca 2ϩ concentrations. For precise quantification of bound Ca 2ϩ , we took advantage of the fact that complexation of free EDTA with Ca 2ϩ ions is manifested by specific peak shifts in the 1 H NMR spectra that are associated with changes in the chemical environment of methylene and ethylene protons of the acetyl and ethylenediamine moieties, respectively (8) (Fig. 2A). As demonstrated in Fig. S2, titration of Ca-SPM by EDTA resulted in progressive accumulation of singlet (2.47 ppm) and quartet (3.05 to 2.95 ppm) peaks in the 1 H NMR spectra, corresponding to a gradual displacement of Ca 2ϩ ions from Ca-SPM and formation of Ca 2ϩ -chelated EDTA (EDTA.Ca 2ϩ ) complexes. This was further accompanied by a decrease of peak intensities in the aromatic, backbone amide and methyl regions of the 1 H NMR spectra, illustrating an EDTA-induced unfolding of Ca-SPM. No peaks of partially folded SPM intermediates were detectable in the 1 H-15 N HSQC spectra, indicating a cooperative binding of Ca 2ϩ ions (9). Once Ca-SPM was unfolded, additions of EDTA did not further affect the 1 H NMR spectra, except of two singlet peaks at 3.5 and 3.15 ppm, which corresponded to proton signals from free EDTA (Fig. S2).
Changes in relative intensities of proton signals were used to determine Ca 2ϩ stoichiometry for Ca-SPM. As shown in Fig. 2B, ϳ4 molar equivalents of EDTA were required to completely unfold the Ca-SPM molecule. Moreover, the slope of the titration curve was close to Ϫ1/4, indicating a Ca 2ϩ stoichiometry of 4:1 (Ca 2ϩ /SPM). In parallel, four calcium equivalents were required to saturate the signal of methylene protons of the Ca 2ϩ -EDTA complex, confirming that Ca-SPM contains four Ca 2ϩbinding sites.
Structural model of Ca-SPM. The amino acid sequence of SPM has been predicted to contain three putative Ca 2ϩ -binding regions located between residues D 462 and E 474 , D 499 and D 511 , and D 521 and E 532 (2). These regions exhibit sequence homology to a 12-residue-long canonical EF-hand motif (D-X-[D/N]-X-[D/N]-G-X 5 -[D/E]) that binds the Ca 2ϩ ion via side chain carboxyl (Glu or Asp) and carbonyl (Gln or Asn) groups in the conserved positions 1, 3, 5, and 12 (10). Visual inspection of the Ca-SPM structure revealed that the listed regions form very stable loops. In particular, the regions of D 499 to D 511 and of D 521 to E 533 form well-defined helix-loop-helix and helix-loop-strand structural motifs, respectively. Moreover, the SPM sequence contains a stretch of conserved residues (D 419 , D 421 , D 423 , and G 424 ) that strongly resemble the conserved residues (positions 1, 3, 5, and 6) of an incomplete EF-hand-like motif. These residues form a stable loop in the Ca-SPM structure. Along with the adjacent D 456 residue, these can provide two oxygens for calcium coordination and are likely to constitute the fourth Ca 2ϩ -binding site in Ca-SPM. In order to build a reliable model of Ca-SPM, four Ca 2ϩ ions were docked into four putative Ca 2ϩ -binding sites of the Ca-SPM structure and subjected to a series of molecular dynamics (MD) simulations. Analysis of the final Ca-SPM structures after 200 ns of unrestrained MD runs revealed that all four Ca 2ϩ ions bind within the structure without triggering any significant conformational change at the level of the polypeptide backbone and without violating NMR restraints. However, calcium coordination provoked local structural changes at the level of side chains within Ca 2ϩbinding loops, where the surface-exposed Asp or Glu residues were reoriented toward the Ca 2ϩ ions. The reorientation and coordination of Ca 2ϩ by the negatively charged residues was associated with a significant shift of the electrostatic surface potential to less negative values, except for the binding site of D 419 to E 426 that remained negative after Ca 2ϩ binding (Fig. S3). Titration of Ca-SPM by EDTA. Ca-SPM was dialyzed overnight at 4°C against the Ca 2ϩ -free buffer before being titrated by stepwise addition of EDTA. Relative intensities of methyl (circles) and Ca 2ϩ -chelated EDTA (triangles) peaks in the 1 H NMR spectra were plotted against the molar EDTA/SPM ratio, with the maximal intensity of peak set arbitrarily to 1. (C) Structural model of the lowest energy structure of Ca-SPM. (D)Detailed view of the Ca 2ϩ -binding sites in Ca-SPM. The final structural model of Ca-SPM is displayed in Fig. 2C. It is composed of eight anti-parallel ␤-strands, forming two ␤-sheets oriented perpendicularly to each other. The N-terminal ␤-strand is adjacent to the C-terminal ␤-strand (in antiparallel orientation), with the N and C termini of the molecule in close contact. Both ␤-strands form an interface between two ␤-sheets, constituting a hydrophobic core of the protein. This is enclosed by surface-exposed turns, a helix-turn-helix motif, and a long flexible loop at the N terminus of the protein. The structure is stabilized by binding of four Ca 2ϩ ions, each coordinated by six oxygen atoms from side chains or backbone carbonyl groups of the engaged residues (Fig. 2D). Three of the four Ca 2ϩ -binding sites are made up of sequential motifs (D 462 to E 474 , D 499 to D 511 , and D 521 to E 532 ) that closely resemble that of EF hands, while the fourth Ca 2ϩ -binding site consists of a structural motif, where the position 12 of an incomplete EF-hand-like motif (D 419 to G 424 ) is structurally supplemented by D 456 . The calcium-binding site at D 521 to E 532 is adjacent to a specific pair of two tryptophan residues (W 451 and W 519 ), whose aromatic rings are arranged in the T-shaped or edge-to-face orientation (center-to-center distance of 6.1 Ϯ 0.3 Å; angle between the ring planes of 85 Ϯ 5°).
Ca-SPM adopts a novel fold. Structural similarity searches using CATH (11), CO-FACTOR (12), DALI (13), and PDBeFold (14) revealed that the Ca-SPM structure does not resemble any of the structures deposited in the Protein Data Bank (PDB) (Fig. S4). No significant hits were found using DALI and COFACTOR (template-modeling [TM] score of Ͻ0.5 and Z score of Ͻ2). The closest structural match was identified between the Ca-SPM structure and the human split pleckstrin homology domain of phospholipase C-␥ (PDB 2W2W). However, the overall homology was very low, with a Q-score of 0.12. Hierarchic classification of protein domain structures using CATH revealed that the Ca-SPM structure shows signs of similarity with a ␤-barrel of the lipocalin fold (CATH superfamily 2.40.128.50) of D-aminopeptidase from the Gram-negative bacterium Ochrobactrum anthropi and a two-layer sandwich domain (CATH superfamily 3.30.60.30) of a wild turkey serine protease inhibitor. However, the CATH analysis yielded SSAP (sequential structure alignment program) scores of 71.6 and 72.1 (with RMSDs of 5.1 Å and 9.8 Å for 82 and 51 residues, respectively), indicating that the Ca-SPM structure does not contain the same fold and only belongs to the same protein class with common structural motifs. Hence, the structure of Ca-SPM reveals a novel fold.
Backbone dynamics of Ca-SPM. In order to assess the dynamic properties of Ca-SPM on a fast (subnanosecond) timescale, we measured the steady-state 1 H-15 N nuclear Overhauser enhancement and the longitudinal (R 1 ) and transverse (R 2 ) relaxation rates (Fig. S5). Based on a model-free approach, the analysis yielded the overall rotational correlation time ( c ) of 7.54 Ϯ 0.01 ns. The generalized order parameter S 2 , describing the amplitude of fast (picosecond) internal motions, was higher than 0.8, indicating that the Ca-SPM backbone is rigid, similar to that of other well-folded proteins. Lower S 2 values (Ͻ0.8) were observed only for six residues (G 422 , L 463 , D 484 , S 500 , V 551 , and Q 563 ), illustrating high flexibility of these residues within the Ca 2ϩbinding sites. Moreover, the NMR relaxation experiments revealed a conformational exchange on the micro-to millisecond timescale: (i) the R 2 values were elevated (Ͼ15 s Ϫ1 ) for residues V 428 and I 468 to A 493 , and (ii) the ratio between the intensities of the 1 H-15 N cross-peaks at 30°C and 10°C (I 30°C /I 10°C ) was decreased for residues in the regions of I 425 to L 456 , I 468 to D 493 , and Q 520 to R 534 against the expected I 30°C /I 10°C value. Most of these residues are in close proximity to the invisible portion of the structure, documenting that the region of T 430 to G 450 is in a slow conformational exchange.
Solution structure of a catalytic intermediate of SPM (SPM-P415A). To unravel the mechanistic details of the D 414 -P 415 cleavage, we prepared a cleavage-incompetent mutant of SPM, where the proline residue (P 415 ) of the D 414 -P 415 cleavage site was replaced by an alanine residue (P415A). Unlike Ca-SPM, which represents a stabilized low-energy conformation of SPM after the cleavage and in which the N-terminal sequence begins with the P 415 residue ( 415 PLALD 419 ), the SPM-P415A construct possesses the GSDALALD 419 sequence at its N terminus. This is not processed, and it is likely to represent a structure of the SPM precursor before cleavage. As shown in Fig. 3A, the overlay of the HSQC spectra demonstrated that the chemical shift patterns of the Ca 2ϩ -saturated SPM-P415A (Ca-SPM-P415A) and Ca-SPM almost overlap, indicating that the cleavage of the D 414 -P 415 peptide bond is preceded by Ca 2ϩ -induced folding of the SPM polypeptide and that the overall structure of SPM remains unaffected after the cleavage.
Calcium titration followed by far-UV circular dichroism (CD) spectroscopy revealed that the SPM-P415A protein undergoes Ca 2ϩ -induced structural transition from an unfolded to a folded conformation that is undistinguishable from that of SPM (Fig. 3B). Moreover, Ca-SPM-P415A exhibited a strong CD exciton couplet at 230 nm, indicating that the T-shaped orientation of two tryptophan residues was maintained in the Ca-SPM-P415A structure. However, thermal unfolding experiments performed by nanodifferential scanning fluorimetry (nanoDSF) revealed that Ca-SPM-P415A unfolds at a lower temperature than Ca-SPM, with the midpoint of the thermal melting curves (or melting temperature, T m ) being shifted by about 10°C (from 89°C to 79°C) (Fig. 3C). The T m values were dependent on Ca 2ϩ concentrations in the sample buffer, and reversible unfolding of the proteins enabled us to determine thermodynamic parameters of the folding/unfolding process. The values of Gibbs free energy of Ca-SPM and Ca-SPM-P415A unfolding, calculated for the Ca 2ϩ -saturated protein (10 mM CaCl 2 ) at 30°C, were determined to be approximately 80 kJ/mol and 40 kJ/mol, respectively. Thus, the three extra residues at the N terminus negatively affected the thermal stability of Ca-SPM-P415A, indicating structural destabilization of the overall structure prior to cleavage.
The number of Ca 2ϩ -binding sites in Ca-SPM-P415A was determined by NMR titration experiments. Unlike results for Ca-SPM, the dialysis of Ca-SPM-P415A against the Ca 2ϩ -free buffer yielded a partial unfolding of the protein, indicating a lower conformational stability of the Ca-SPM-P415A structure. In order to avoid inaccurate Structure of the Self-Processing Module ® interpretation of the titration data, the dialyzed Ca-SPM-P415A sample was supplemented with an excess of two molar equivalents of Ca 2ϩ (to ensure folding of 100% molecules) and was next titrated with EDTA. As shown in Fig. 3D, the EDTA-induced unfolding of Ca-SPM-P415A was characterized by a nonlinear equilibrium and was completed after addition of ϳ6.5 equivalents of EDTA (including two Ca 2ϩ equivalents added prior to the EDTA titration). The titration data fit well to binding models describing the cooperative binding of both four and five Ca 2ϩ ions, but none of the models could unambiguously assign the exact number of Ca 2ϩ -binding sites in the Ca-SPM-P415 structure. However, the overall calcium stoichiometry for Ca-SPM-P415A was higher than that for Ca-SPM (Fig. 2B), indicating the presence of an additional Ca 2ϩ -binding site in the Ca-SPM-P415A structure.
The solution structure of Ca-SPM-P415A was determined using a procedure identical to that for Ca-SPM. A total of 2,250 distance and 119 dihedral angle restraints were used for the structure calculations (Table 1), and the overlay of final structures is depicted in Fig. 3E. Comparison of the Ca-SPM-P415A structure with that of Ca-SPM revealed that the two structures are highly similar across the models, with the average RMSD of 1.12 Å for the assigned residues. Differences in chemical shifts in most of the residues were very low (Ͻ0.1 ppm), except for residues located at the N and C termini (L 416 and G 578 to L 582 ) and the loop adjacent to the N terminus (G 477 to N 479 and H 490 ). Close examination of the Ca-SPM-P415A structure revealed that the N terminus of the protein adjoins two carboxy side chains of D 478 and D 579 , which were proposed to be involved in the catalysis of the D 414 -P 415 peptide bond cleavage by promoting protonation of P 415 (1). Moreover, D 478 and D 579 along with D 414 form a region with a negative net charge in the vicinity of the cleavage site, which is likely to form the fifth Ca 2ϩ -binding site in the Ca-SPM-P415A structure. Therefore, the fifth Ca 2ϩ ion, coordinated by D 414 , D 478 , and D 579 , was modeled within the Ca-SPM-P415A structure. D 478 and D 579 residues do not catalyze the D 414 -P 415 peptide bond cleavage. To assess whether the carboxylate anions of D 478 and D 579 residues catalyze the D 414 -P 415 peptide bond cleavage, the D 478 and D 579 residues were replaced by alanine or asparagine residues, and the Ca 2ϩ -dependent self-processing activity of the mutants was examined using the respective GST-SPM fusion proteins. As shown in Fig. S6, the replacement of D 478 with alanine (D478A) resulted in a significant reduction of the self-processing activity, while only a moderate reduction in the cleavage rate was observed upon replacement of D 478 by asparagine (D478N). In contrast, the removal of the carboxyl group from D 579 completely abolished the Ca 2ϩ -dependent processing of the GST-SPM-D579A protein, while the D579N substitution had no significant impact on the cleavage of the GST-SPM-D579N protein. The double D478N D579N substitution reduced but did not eliminate the cleavage activity of SPM, indicating that the carboxyl groups of the D 478 and D 579 residues do not catalyze the cleavage of the D 414 -P 415 peptide bond but, rather, play a key structural role.
Structural destabilization of the SPM precursor in simulations. To get insight into structure of the autoproteolytic site prior to cleavage, the A 415 residue in the Ca-SPM-P415A structure was replaced in silico by a proline, and the resulting structural model of an SPM precursor (DP-SPM) was subjected to molecular dynamics (MD) simulations. The purpose of the analysis was not to determine the exact value of the 414 dihedral angle (which is usually planar and fixed to 180°) but to monitor the balance between violation of an ideal peptide bond geometry (strongly imposed by the force field regardless of the actual chemical context) and the overall structure derived from the experimental data and force field requirements. The MD simulations revealed that the noncleavable D 414 -A 415 peptide bond in the Ca-SPM-P415A structure containing four Ca 2ϩ ions is almost completely in trans conformation ( 414 ϭ 180°). This is in line with the experimental data and validates our computational approach (Fig. 4). In contrast, the Ca-SPM-P415A structure containing five Ca 2ϩ -binding sites exhibited a significant distortion of the D 414 -A 415 peptide bond geometry, with a deviation from a near planar geometry by about 25°and 15°(depending on the NMR experimental restraints used in the calculations). In line with that, the D 414 -P 415 peptide bond in the DP-SPM structure containing four Ca 2ϩ ions had a deviation of about 20°and even 40°i n a structural model containing five Ca 2ϩ ions. The cis isomer of 414 occasionally appeared during simulations but with experimental restraints violated (Fig. 4). Taken together, these data suggested that the autocatalytic cleavage of the D 414 -P 415 peptide bond is associated with a ground-state destabilization of the D 414 -P 415 peptide bond through a twisted amide.
The Ca 2؉ -dependent clip-and-link activity of the ApxIVA protein plays an important role in A. pleuropneumoniae infection. Since N. meningitidis is an exclusively human pathogen, the existing animal models do not accurately simulate meningococcal disease. It has thus proved difficult to study the role of individual proteins involved in the pathogenesis of N. meningitidis in vivo. However, the Ca 2ϩ -dependent clip-and-link activity has been observed also for ApxIVA (1), one of the RTX proteins of the animal pathogen A. pleuropneumoniae that causes porcine pleuropneumonia infections (15). Like FrpC, ApxIVA comprises a highly conserved SPM segment (residues 639 to 815 of the full-length ApxIVA) that mediates Ca 2ϩ -dependent cleavage of the D 638 -P 639 peptide bond and a covalent linkage of the processed N-terminal segment to adjacent proteins (1) (Fig. 5A). To assess the biological importance of the Ca 2ϩdependent clip-and-link activity of ApxIVA in vivo, we introduced an in-frame deletion of codons 629 to 827 in the apxIVA gene on A. pleuropneumoniae chromosome (Fig. 5B) and examined the virulence of the mutant strain (ΔSPM) in a pig infection model.
Intranasal administration of high doses (ϳ10 9 CFU per animal) of the wild-type and mutant strains provoked an acute form of pleuropneumonia, causing deaths of most of piglets within 24 h. The surviving animals were in the terminal stage of the disease and were euthanized at day 1 postinfection. Postmortem examination revealed that the lungs of infected animals exhibited the characteristic, dark red-to-black necrohemorrhagic areas of consolidation accompanied by fibrinous pleuritis, with an average lung score of 28 and 32% for the wild-type and the mutant strains, respectively (Fig. 5C). Structure of the Self-Processing Module ® However, at lower bacterial challenge doses (ϳ10 6 CFU per animal), moderate clinical symptoms were elicited, including lethargy, respiratory distress, and tachypnea, with no pleuropneumonia-related deaths during the time frame of the experiment (3 days). The body (rectal) temperature was largely above the physiological range (Ͼ40.5°C) and remained elevated for about 48 h. Regular hematological examinations showed neutrophilic leukocytosis, reaching the maximum at day 1 postinfection in all infected animals (Fig. 5D). However, compared to the challenge with the wild-type bacteria, no monocytosis was observed in pigs challenged with the ΔSPM mutant, and the overall monocyte counts were the same as for uninfected controls. Moreover, the average lung score for the ΔSPM-challenged pigs was lower than for pigs challenged with the wild-type strain (6.8 versus 14.7%), although the wild-type challenge strain was reisolated from lungs of only 3 out of 6 animals (compared to 5 of 6 animals for the ΔSPM strain). This result showed a significant reduction of lung lesions in pigs challenged with the ΔSPM strain, suggesting that the Ca 2ϩ -dependent clip-and-link activity of ApxIVA plays a specific role in the pathogenesis of porcine pleuropneumonia.

DISCUSSION
Ca 2ϩ -dependent protein clip-and-link activity of large RTX proteins represents a unique mechanism of posttranslational processing of proteins that involves a rearrangement of a polypeptide backbone through the highly specific Ca 2ϩ -dependent autocatalytic cleavage of an Asp-Pro peptide bond and a nonspecific covalent linkage of the released carboxyl of the Asp residue to -amino groups of adjacent Lys residues. The cleavage of the Asp-Pro peptide bond is catalyzed by the Ca 2ϩ -dependent structural transition of the SPM domain, which is highly conserved and defines a specific subclass of RTX exoproteins (1, 2). Very low Ca 2ϩ concentrations in the bacterial cytosol (Ͻ100 nM) maintain the RTX proteins in an unfolded state that is required for singlestep translocation across the bacterial envelope via the T1SS apparatus (16). Once exported to the calcium-rich extracellular environment (Ͼ100 M), the proteins fold and acquire their biological activity. Given the fact that the binding affinity of SPM for Ca 2ϩ is about 150 M (2), the Ca 2ϩ -dependent clip-and-link activity of the RTX proteins occurs exclusively outside the bacteria. In this work, we determined the solution structure of SPM of the N. meningitidis FrpC protein and provided the structural insight into the Ca 2ϩ -dependent cleavage of the D 414 -P 415 peptide bond. The Asp-Pro bond belongs to the most labile peptide bond, which is known to be selectively hydrolyzed within several days under acidic conditions at higher temperatures (17)(18)(19)(20). This is due to a higher basicity of the secondary amine group in the proline than that of the primary amine group of other amino acids. At low pH, the free electron pair on the proline nitrogen may polarize the carbonyl C-N bond and make the carbonyl carbon more susceptible to nucleophilic attack. The increased basicity of the nitrogen atom may promote protonation of the nitrogen atom of the intermediate, resulting in the cleavage of the Asp-Pro peptide bond. In contrast, the SPM-mediated cleavage of the Asp-Pro peptide bond is a highly specific catalytic reaction that occurs quickly at physiological pH and with reaction rates of about 2 orders of magnitude higher than the uncatalyzed chemical cleavage. Osicka and coworkers (1) proposed that the protonation of the proline nitrogen in SPM could be promoted by an as yet uncharacterized residue(s), which can interact with the proline residue. Inspection of the Ca-SPM-P415A structure revealed two aspartate residues (D 478 and D 579 ) positioned in close proximity to P 415 . However, site-directed mutagenesis ruled out that any of the aspartates acted as the proton-donating amino acid (see Fig. S6 in the supplemental material). Instead, both aspartates appear to play a structural role in binding of an additional (fifth) calcium ion. The fifth calcium-binding site was deduced to reside only in the SPM-P415A structure, indicating the essential structural role of the D 414 residue prior to cleavage. Molecular docking revealed that the fifth calcium ion could be coordinated by the carbonyl oxygen of the D 414 residue and together with D 478 and D 579 would constitute a transient calcium-binding site that is formed during the Ca 2ϩ -dependent folding of SPM and falls apart upon the cleavage of the D 414 -P 415 peptide bond. The binding of the fifth calcium ion enhances distortion of the C-N bond of the carboxyamide group, generating a twisted amide in a scissile bond, which destabilizes the overall structure of SPM. This is well documented by the experimental data showing that the thermal stability of the Ca-SPM-P415A structure before cleavage is much lower than that of the Ca-SPM structure after the cleavage.
Twisted amides represent one of the approaches for activation of amide bonds, which are usually stable due to formation of a resonating structure provided by the conjugation of the nitrogen lone pair with the carbonyl group (21,22). The distortion of amide bonds causes a loss of double-bond character and thus increases their reactivity compared to planar counterparts. Such an unusual conformation of the scissile amide bond is the driving force for a nucleophilic or electrophilic attack that is proposed to be a central design element of a variety of enzymatic processes, such as cis-trans isomerization (23), amide hydrolysis (24), N-linked glycosylation of proteins (25), and intein-based protein splicing (26). Based on the data above, we propose a mechanism for the SPM-mediated processing of the Asp-Pro peptide bond (Fig. 6). Upon calcium binding, the SPM polypeptide undergoes a highly cooperative structural transition from an intrinsically unstructured conformation to the compact protein fold that is stabilized by binding of four calcium ions in EF-hand-like Ca 2ϩ -binding sites. The scissile D 414 -P 415 peptide bond adopts a highly strained turn that is structurally constrained by the adjacent ␤1-strand and the residues of a narrow port (W 451 , L 475 , G 477 , G 491 , and D 579 to L 582 ) that shields the D 414 -P 415 cleavage site from the solvent (Fig. 6A and B). Coordination of the fifth calcium ion by the oxygen atom of the carbonyl group of D 414 provokes rotation of the D 414 -P 415 peptide bond over the C-terminal ␤14-strand that apparently accentuates rotation of the C-N bond to produce a twisted amide. Such a distortion generates a reactive lone electron pair at the nitrogen that abolishes the conjugation of the nitrogen electrons with the carbonyl group and facilitates a nucleophilic attack of the carboxylic acid side chain of D 414 on its carbonyl carbon (Fig. 6C). This results in formation of a cyclic imide intermediate and D 414 -P 415 peptide bond cleavage, with subsequent formation of a reactive cyclic D 414 anhydride. This can next be attacked by a free amino group of a lysine residue to generate an isopeptide bond.
The SPM represents a highly conserved domain that appears to be present in many RTX proteins of Gram-negative bacteria. However, the SPM-mediated processing of the Asp-Pro peptide bond has as yet been reported only for the N. meningitidis Our in vivo pig experiments showed that the abrogation of the Ca 2ϩ -dependent clip-and-link activity of ApxIVA led to significant reduction of necrotizing lung lesions after A. pleuropneumoniae infection (Fig. 5C). These data are in good agreement with the previous result showing that the deletion of the apxIVA gene attenuates the virulence of A. pleuropneumoniae in a pig infection model (29). A. pleuropneumoniae produces four different RTX proteins, where only ApxI, ApxII, and ApxIII are toxins possessing both hemolytic and cytotoxic activity (30). With respect to the fact that the fourth RTX protein, ApxIVA, does not exert any of these activities, the biological activity of ApxIVA could be ascribed to the Ca 2ϩ -dependent clip-and-link activity. It is tempting to speculate that colonization of pig respiratory airways by A. pleuropneumoniae is associated with the Ca 2ϩ -dependent processing of the D 638 -P 639 peptide bond of the type I-secreted ApxIVA and covalent linkage of the ApxIVA 1-638 fragment to the host respiratory epithelium, which would serve as a high-affinity target for an as yet unidentified protein structure exposed on the bacterial cell surface. This would allow tight adherence of the bacterial cells to the host respiratory epithelia, which may be essential for the full virulence of A. pleuropneumoniae. Hence, covalent attachment of the N-terminal fragments of specific RTX proteins to the host proteins through the Ca 2ϩ -dependent protein clip-and-link activity could represent an unconventional strategy for pathogenic microorganisms to adhere to the target host cell surface.

MATERIALS AND METHODS
Reagents. The suicide vector for allelic exchange on the Actinobacillus pleuropneumoniae chromosome (pEMOC2) and the Escherichia coli ␤2155 (ΔdapA) strain were generously provided by Gerald F. Gerlach (Institute for Innovative Veterinary Diagnostics, Hannover, Germany).
For construction of the pEMOC2-ApxIVAΔ629-827 allelic exchange vector, two DNA fragments corresponding to the 5= and the 3= flanking regions of the in-frame deletion were amplified from genomic DNA purified from the Czech field isolate Actinobacillus pleuropneumoniae KL2-2000 (biotype 1, serotype 9) using PCR, as follows: the 998-bp DNA fragment upstream of the deletion was amplified from DNA by using the forward primer 5=-TTTGCGGCCGCTTGCGGGCAAAGAAGTTACG-3= containing the NotI restriction site and the reverse primer 5=-TTTCTCGAGATTTGGCGCATTCACATCGC-3= containing the XhoI site. Similarly, the 1,004-bp DNA fragment downstream of the deletion was amplified from DNA by using the forward primer 5=-TTTCTCGAGCGCACAATTAATCTAACCGG-3= containing the XhoI site and reverse primer 5=-TTTGGGCCCAATTTTAAGGTGTCAATATCGC-3= containing the ApaI site. The PCR products were cut with appropriate restriction enzymes and collectively ligated with the NotI/ApaI-cleaved pEMOC2 vector. All constructs were confirmed by DNA sequence analysis with an ABI Prism 3130XL analyzer (Applied Biosystems, USA) using a BigDye Terminator cycle sequencing kit.
Protein expression and purification. All GST-SPM fusion constructs were expressed in E. coli strain BL21(DE3) transformed with the appropriate plasmid. Exponential 500-ml cultures were grown in a shaking incubator at 37°C in M9 minimal medium supplemented with trace metals, vitamins, and kanamycin (60 g/ml). For NMR experiments, the cells were grown in M9 medium supplemented with 0.5 g/liter 15 NH 4 Cl (Cambridge Isotope Laboratories, USA) and 2 g/liter [ 13 C]glucose (Cambridge Isotope Structure of the Self-Processing Module ® Laboratories, USA). Expression of proteins was induced by adding 1 mM isopropyl-␤-D-thiogalactopyranoside (IPTG) at an optical density at 600 nm (OD 600 ) of 0.6 to 0.8, and bacteria were grown for an additional 4 h. The cells were harvested by centrifugation (1,500 ϫ g for 15 min), washed in TNE buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 5 mM EDTA), and disintegrated by sonication in TN buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl) at 4°C. The cell extracts were centrifuged at 20,000 ϫ g for 30 min, and the supernatants were used for protein purifications.
The purification of the wild-type Ca-SPM was previously described (31). Briefly, the cleared cell lysate was loaded onto Ni-Sepharose 6 Fast Flow beads (GE Healthcare) and washed with TN buffer, and the dHis-GST-SPM protein was eluted with TN buffer supplemented with 500 mM imidazole. The collected fractions were mixed with dithiothreitol (DTT) to a final concentration of 10 mM before the protein solution was dialyzed overnight at 4°C in TN buffer supplemented with 10 mM DTT and 10 mM CaCl 2 . Addition of calcium ions induced a self-processing activity of SPM, resulting in the cleavage of the GST-SPM fusion proteins. The cleaved GST protein was precipitated by incubation of the protein mixture at 70°C for 15 min, and the resulting suspension was centrifuged at 5,000 ϫ g for 20 min. The supernatant was loaded onto a PLRP-S reverse-phase column (Agilent Technologies) in buffer containing 50 mM triethylamine (pH 8.5) and 5% acetonitrile, and SPM was eluted from the column by the gradient of acetonitrile (5 to 95%). The SPM fractions were concentrated by rotary vacuum evaporator and loaded onto a Superdex HR 200 gel filtration column (GE Healthcare) equilibrated with TN buffer supplemented with 10 mM CaCl 2 . For purification of SPM-P415A, the cell lysate containing GST-SPM-P415A was loaded onto a glutathione agarose column (Life Technologies), extensively washed with TN buffer, and eluted with TN buffer supplemented with 10 mM L-glutathione (reduced). Collected fractions were pooled, mixed with the purified recombinant tobacco etch virus (TEV) protease (1:20, wt/wt), and dialyzed at 4°C overnight against TN buffer. The protein mixture was incubated at 70°C for 15 min, and the resulting suspension was centrifuged at 5,000 ϫ g for 20 min. The supernatant was loaded onto a PLRP-S reverse-phase column (Agilent Technologies) in buffer containing 50 mM triethylamine (pH 8.5) and 5% acetonitrile, and SPM-P415 was eluted from the column by the gradient of acetonitrile (5 to 95%). The collected fractions were concentrated by rotary vacuum evaporator and loaded onto a Superdex HR 200 gel filtration column (GE Healthcare) equilibrated with TN buffer supplemented with 10 mM CaCl 2 . The single (D478A, D478N, D579A, and D579N) and double (D478N D579N) substitution mutants of GST-SPM constructs were purified by Ni-Sepharose 6 Fast Flow column chromatography as described above for the wild-type SPM. The collected fractions were mixed with dithiothreitol (DTT) to a final concentration of 10 mM before overnight dialysis at 4°C in TN buffer. The purity of the proteins was monitored by SDS-polyacrylamide gel electrophoresis (SDS-PAGE), and protein concentrations were determined by Bradford assay (Bio-Rad) using bovine serum albumin as a standard.
Structure calculation of SPM and SPM-P415A. Torsion angle restraints were determined from chemical shifts by TALOS-N (35) using the 1 H N , 1 H ␣ , 15 N, 13C=, 13 C ␣ , and 13 C ␤ resonances. Distance restraints were derived from the automatically assigned NOESY spectra with a manual check of the results. Distance and torsion restraints were used to generate 100 structures in CYANA, version 3.97. Twenty structures with the lowest target functions were used as an input for the molecular dynamics (MD) simulations. In calculations with calcium ions, four Ca 2ϩ ions were introduced in CYANA, version 3.97, by defining three 3-Å distance restraints between Ca 2ϩ and oxygen OD1 of D419, D421, D423, D462, N464, N468, D499, N501, D503, D521, N523, and D525. The fifth Ca 2ϩ ion was introduced into the constructs containing D414, where the distance restraints were to oxygen OD1 of D414, D478, and D579. The Ca 2ϩ ions were not restrained in the final refinement during the MD runs.
MD simulations. Molecular dynamics (MD) simulations were performed in GROMACS, version 5.0.5 (36,37) using the force field parameters AMBER99SB-ILDN (38). On average, the systems were solvated in 8,200 explicit TIP3P (39) water molecules and 12 Cl Ϫ and 18 (minus two times number of Ca 2ϩ ions introduced) Na ϩ ions in a periodic dodecahedron simulation box mimicking the electroneutral system with a salt concentration of 70 mM. The MD step of 1 fs was used in the leapfrog integration scheme. Van der Waals interactions were calculated using a triple-range cutoff scheme, with the updated interaction distance of 1 nm. Electrostatic interactions were calculated by the particle mesh Ewald (PME) method with a Coulomb cutoff of 1 nm and a relative dielectric constant of 1. The temperature and pressure were controlled by a Berendsen thermostat (40) with a coupling constant of 0.1 ps and a Parrinello-Rahman barostat (41) at the constant pressure of 10 5 Pa with the coupling constant of 1 ps. Simulated annealing of the system was initiated with a relaxation phase of 100 ps, equilibration at 300 K, and subsequent heating to 1,000 K in 50 ps. The temperature was kept constant for 3 ns and then slowly decreased to 0 K over an additional 12 ns. The structure was restrained during MD by experimental distances and dihedral angles with force constants of 1,000 kJ mol Ϫ1 nm Ϫ2 and 350 kJ mol Ϫ1 rad Ϫ2 , respectively.
The quality of the structures was analyzed by PROCHECK (42), WHAT IF (43), and CING (44), available as the online service iCING. Ramachandran statistics by program PROCHECK categorized 91.3 and 92.4% residues in the most favored regions, 8.6 and 7.4% in the additional allowed regions, and 0.1 and 0.1% in the generously allowed regions for SPM and SPM-P415A, respectively. No residues were found in the disallowed regions. Secondary structure motifs were identified from the models by the program DSSP (45) with a threshold of 75% of occurrence. The electrostatic potentials were calculated at 303.2 K in APBS (46). The chemical shifts were predicted from calculated structures by SPARTAϩ (47).
Titration of Ca-SPM and Ca-SPM-P415A by EDTA. An unlabeled SPM or SPM-P415A sample was prepared in the same buffer as used for the structure determination and was dialyzed overnight against a CaCl 2 -free buffer, consisting of 50 mM Tris-HCl (pH 7.4) and 50 mM NaCl; 10% D 2 O and 0.1% NaN 3 were added to the sample prior to the measurement. The final volume and protein concentration were 0.54 ml and 0.175 mM or 0.52 ml and 0.707 mM for SPM or SPM-P415A, respectively. The titration was monitored by a 1D proton NMR experiment with water suppression using the WATERGATE W5 pulse sequence with gradients in a double echo at 700 or 850 MHz, and the delays for water suppression were 205 or 169 s. The total number of scans was 192 or 80, and the spectral width was 20 or 30 ppm for SPM or SPM-P415A, respectively. The delay for the interscan was 5 s, and spectra were sampled by 64,000 points. The samples were titrated with the solution of 50 mM EDTA, 50 mM NaCl, and 50 mM Tris-HCl (pH 7.4). The 1D spectra were processed with apodization exponential 1 Hz function and phased.
15 N Relaxation measurements. The NMR 15 N relaxation experiments were carried out at 30°C using 15 N heteronuclear single-quantum coherence (HSQC)-based pulse schemes on a uniformly 15 N-labeled Ca-SPM (0.33 mM). The temperature was calibrated with a 100% methanol sample, where difference between peaks was set up to 1.518 ppm (48). The following relaxation delays were used in the 15  Experiments were acquired with 8 scans per free induction decay (FID) and an interscan delay of 2 s. The relaxation rates were obtained by fitting peak intensities to a mono-exponential decay by using Relax (49). The 1 H-15 N ssNOE measurements were achieved with a uniformly 15 N-labeled Ca-SPM (0.6 mM) under a steady-state condition, achieved by a 5-ms 1 H irradiation with 225 repeats of 200-s 180°pulses separated by 22.22-s delays and with a 15-s interscan delay. The recorded spectra consisted of 320 by 2,048 complex points in the indirect 15 N and direct 1 H dimensions, respectively, corresponding to respective acquisition times of 57.7 ms and 80.3 ms for a reference and steady-state spectra, which were measured in an interleaved manner. Experiments were acquired with 16 scans per free induction decay (50). Fast backbone motions of Ca-SPM were calculated by a model-free approach (51). The program ROTDIF (52) was used to calculate a rotational diffusion tensor in order to separate the influence of the slow exchange from the effect of the rotational diffusion. The axially symmetric model was used for description of the experimental data.
CD spectroscopy. The far-UV circular dichroism (CD) spectra were recorded on a Jasco-815 spectropolarimeter in rectangular quartz Suprasil cells of 1-mm path length (110-QS; Hellma). Protein samples (100 g/ml) were diluted in 5 mM Tris-HCl (pH 7.4) plus 50 mM NaCl in the absence or presence of 10 mM CaCl 2 and measured for wavelengths from 195 to 280 nm at 25°C. Two spectrum accumulations with standard instrument sensitivity and scanning speed of 10 nm/min with response time of 16 ms were acquired. The spectra of the buffers were subtracted from the protein spectra, and mean molar ellipticity () was expressed in degrees per square centimeter per decimole.
Thermal stability assay. Thermal stability of SPM proteins was performed by nano-differential scanning fluorimetry (nanoDSF) using a Prometheus NT.48 instrument (NanoTemper Technologies, Munich, Germany). Ca-SPM and Ca-SPM-P415A were dialyzed overnight at 4°C against TN buffer, and the freshly prepared protein samples were supplemented with increasing concentrations of calcium ions (0 to 10 mM CaCl 2 ) before 10 l was used for filling the Prometheus NT.48 high-sensitivity capillaries (NanoTemper Technologies). The measurements were conducted from 20 to 95°C (with a temperature ramp of 2.5°C/min) under constant monitoring of tryptophan fluorescence at 350 and 330 nm. The melting temperature (T m ) values, corresponding to the inflection points of the unfolding curve, were determined via the first derivative of the curve.
Construction of the apxIVA mutant strain of A. pleuropneumoniae. In-frame deletion in the apxIVA gene (codons 629 to 827) was performed on chromosome of the Czech field isolate Actinobacillus pleuropneumoniae KL2-2000 (biotype 1, serotype 9) by homologous recombination using a pEMOC2 allelic exchange vector as previously described (53). Briefly, A. pleuropneumoniae cells grown in a 5% CO 2 atmosphere at 37°C on Bacto tryptic soy broth agar (BD Biosciences) supplemented with 10 g/ml ␤-NAD (NAD; Sigma) were mated on fresh agar plates supplemented with 10 g/ml NAD, 1 mM diaminopimelic acid (DAPA; Sigma), and 10 mM MgSO 4 with E. coli ␤2155 grown on LB agar supplemented with 1 mM DAPA and 25 g/ml chloramphenicol (Cm) before being transformed with the pEMOC2-ApxIVAΔ629-827 plasmid construct. After 4 h at 37°C, the A. pleuropneumoniae transconjugants were selected on Bacto tryptic soy broth agar plates supplemented with 10 g/ml NAD and 5 g/ml chloramphenicol and incubated in 5% CO 2 for 24 h at 37°C. For sucrose counterselection, the single Cm r colonies were inoculated into salt-free LB broth supplemented with 10% sucrose, 10% horse serum, and 10 g/ml NAD, incubated with shaking at 37°C for 2 h, and plated on salt-free LB broth supplemented with 10% sucrose, 10% horse serum, and 10 g/ml NAD. Sucrose-resistant colonies were screened for the truncated variant of the apxIVA gene by a colony PCR and restriction analysis of the resulting PCR products (XhoI).
Animal experiments. The animal experiments were carried out according to the guidelines of the Animal Care Act (no. 246/1992 Coll.) of the Czech Republic, approved by the Animal Welfare Commission of the Ministry of Agriculture of the Czech Republic, and conducted in accredited barrier-type stables (accreditation certificate no. 5843/2007-10001). Thirty 4-weak-old Large White pigs with body weights of 7 to 10 kg were purchased from a porcine reproductive and respiratory syndrome virus-free and A. pleuropneumoniae infection-free herd (as tested by serological reaction in enzyme-linked immunosorbent assays [ELISAs]) and randomly assigned to five groups, each containing six animals. Pigs were regularly monitored throughout the housing period at least three times per day.
Experimental infection. The virulence of A. pleuropneumoniae strains was assessed in an intranasal infection model as described previously (54). Briefly, pigs were anesthetized by intravenous injection of ketamine (4 mg/kg of body weight) and xylazine (2 mg/kg) and intranasally administered 4 ml (2 ml per nostril) of phosphate-buffered saline (PBS) solution (control group) or infected with two doses of 2 ml of bacterial suspension corresponding to an infection dose of 10 6 and 10 9 CFU per animal. The CFU number was measured spectrophotometrically in the bacterial culture grown at 37°C in brain heart infusion broth (HiMedia, India) supplemented with 10 g/ml NAD, in which an optical density at 550 nm (OD 550 ) of 1.0 corresponded to 5 ϫ 10 9 CFU/ml. Pigs were monitored for 3 days after infection, and the A. pleuropneumoniae-related clinical signs (respiratory rate, dyspnea, coughing, anorexia, and lethargy) were recorded. At Ϫ24, 0, 24, 48, and 72 h postinfection, body temperature was measured in the rectum. Blood samples were collected by jugular venipuncture into heparinized tubes (30 IU/ml) at Ϫ24, 24, and 72 h postinfection and immediately used for cytological examination. Cell counts were determined by a BC-2800Vet Auto Hematology Analyzer (Mindray, China). All pigs were examined postmortem for macroscopic pathological changes and for the presence of bacteria. The lung score was determined as the extent of lung tissue damage (lung lesions) according to the percentages for the following anatomical segments: left apical lobe, 8%; right cranial lobe, 12%; left cardiac lobe, 8%; right medium lobe, 12%; left diaphragmatic (caudal) lobe, 25%; right diaphragmatic (caudal) lobe, 30%; and accessorius lobe, 5%. For bacterial examination, nasal and tracheal swabs were inoculated on Columbia agar supplemented with 5% sheep blood with a streak of Staphylococcus aureus (as a source of NAD) and cultivated aerobically at 36°C for 18 h.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.  This work was supported by the project 19-15175S of the Grant Agency of the Czech Republic and the projects LM2018133, LM2018127 (CIISB research infrastructure), and LQ1601 (CEITEC2020) of the Ministry of Education, Youth and Sports of the Czech Republic. The project RO0518 of the Ministry of Agriculture of the Czech Republic is also acknowledged. Access to computing and storage facilities owned by parties and projects contributing to the National Grid Infrastructure MetaCentrum provided under