Structural and Functional Characterization of an Essential RTX Subdomain of Bordetella pertussis Adenylate Cyclase Toxin*

The adenylate cyclase toxin (CyaA) is one of the major virulence factors of Bordetella pertussis, the causative agent of whooping cough. CyaA is able to invade eukaryotic cells by a unique mechanism that consists in a calcium-dependent, direct translocation of the CyaA catalytic domain across the plasma membrane of the target cells. CyaA possesses a series of a glycine- and aspartate-rich nonapeptide repeats (residues 1006–1613) of the prototype GGXG(N/D)DX(L/I/F)X (where X represents any amino acid) that are characteristic of the RTX (repeat in toxin) family of bacterial cytolysins. These repeats are arranged in a tandem fashion and may fold into a characteristic parallel β-helix or β-roll motif that constitutes a novel type of calcium binding structure, as revealed by the three-dimensional structure of the Pseudomonas aeruginosa alkaline protease. Here we have characterized the structure-function relationships of various fragments from the CyaA RTX subdomain. Our results indicate that the RTX functional unit includes both the tandem repeated nonapeptide motifs and the adjacent polypeptide segments, which are essential for the folding and calcium responsiveness of the RTX module. Upon calcium binding to the RTX repeats, a conformational rearrangement of the adjacent non-RTX sequences may act as a critical molecular switch to trigger the CyaA entry into target cells.

The adenylate cyclase toxin (CyaA) 2 is one of the major virulence factors of Bordetella pertussis, the causative agent of whooping cough (1)(2)(3). The 1706 residue-long CyaA is a bi-functional protein endowed with both catalytic (adenylate cyclase) and hemolytic activities (2,4,5). Synthesized as an inactive precursor, it is converted to the active toxin by a post translational palmitoylation of two internal lysine residues (Lys 860 and Lys 983 ) (6,7). This active CyaA toxin is then able to deliver its catalytic domain directly across the plasma membrane of a variety of eukaryotic cells and disrupts their physiological functions by uncontrolled synthesis of cAMP (5, 8 -11), leading to the cell death by apopto-sis (12)(13)(14). CyaA is constructed in a modular fashion; the calmodulinactivated catalytic domain is located in the 400-amino-proximal residues, whereas the C-terminal moiety (residues 400 -1706) is endowed with hemolytic activity (4,5,15,16), which results from its ability to form cation-selective channels in membranes (17,18). It also mediates the binding and internalization of the toxin into eukaryotic cells (5,11,19). The hemolytic and the RTX domains display structural characteristics that link CyaA to the RTX (repeat in toxin) family of bacterial toxins (20,21). Indeed, it contains a pore-forming domain (from residues 500 -700) with four hydrophobic segments (17,18,22,23), the target site for the post-translational palmitoylation (7,24), 30 -40 copies of a characteristic glycine-and aspartate-rich nonapeptide repeats (residues 1006 -1613) of the prototype GGXG(N/ D)DX(U)X (X represents any amino acid, and U represents any large hydrophobic residue such as Ile, Leu, Val, Phe, Tyr), representing the main calcium-binding sites of the protein (25) (see Fig. 1), and a nonprocessed C-terminal secretion signal (4,26). The crystal structure of the Pseudomonas aeruginosa alkaline protease that has six of these consensus RTX repeats revealed that these sequences constitute a new kind of calcium binding structure, called a parallel ␤-helix or parallel ␤-roll motif (27,28).
The main originality of the CyaA toxin stems from its unique mechanism of penetration into eukaryotic cells, as its catalytic domain appears to be directly translocated through the plasma membrane of the target cells (5, 9 -11, 29). Recently, it has been demonstrated that CyaA binds specifically to target cells through the ␣ M ␤ 2 integrin receptor (CD11b/CD18) (30,31), which is expressed on a restricted subset of leukocytes including neutrophils, macrophages, and dendritic cells. Expression of this receptor most likely accounts for the high sensitivity of these cells to CyaA (13,30).
CyaA is a calcium-binding protein that undergoes conformational changes upon binding of calcium (9,19,25). The entry/translocation of CyaA into target cells is strictly dependent upon the presence of calcium ions in the millimolar range, and the RTX domain is supposed to be directly involved in this process as it harbors the main low affinity Ca 2ϩbinding sites (9,19,25). How binding of calcium ions to the RTX motifs might trigger the translocation of CyaA into cells remains totally unknown. The RTX domain of CyaA is organized in five successive blocks (Ile to Val) of about 8 nonapeptide RTX motifs ( Fig. 1) separated by linkers of variable length (from 23 to 49 residues). Previous results from Iwaki et al. (32) have suggested that the last 217 C-terminal residues of CyaA that encompasses the block V (residues 1490 -1706) might constitute a functional subdomain; a truncated CyaA lacking the last 217 C-terminal residues was unable to intoxicate target cells but could be partially complemented by other CyaA fragments harboring at least these last 217 C-terminal residues (32). Subsequently, Bejerano et al. (33) corroborated these data by showing that a purified polypeptide comprising these 217 C-terminal residues of CyaA could fully restore toxic activity of an inactive truncated CyaA lacking the 76 C-terminal amino acids.
To gain insight into the potential function of the RTX calcium binding motif/domain, we have constructed different fragments derived from the RTX domain and characterized the functional (ability to complement in vivo an inactive truncated CyaA), biophysical (fluorescence, circular dichroism spectroscopy, and channel-forming activity in lipid bilayers), and biochemical (calcium binding) properties in vivo and in vitro. Our results suggest a model for the folding of the RTX modules in which the functional calcium binding structure extends beyond the RTX glycine/aspartate-rich motifs to include adjacent polypeptide sequences that might directly participate in the stabilization of the parallel ␤-helix fold.

EXPERIMENTAL PROCEDURES
Construction of the CyaA-derived Proteins-DNA manipulations were performed according to standard protocols (34) in the Escherichia coli XL1-Blue strain (Stratagene, Amsterdam, The Netherlands) as host cells. The plasmids pT7CACT1, coding for the acylated wild type CyaA, and pACT⌬1-1006, coding for CyaA 1006 -1706 , the RTX domain of CyaA, have already been described (35)(36)(37). The plasmid pTRCyaA⌬1-1006⌬1490 -1706, coding for the truncated CyaA RTX domain (CyaA 1006 -1490 ), is a derivative of pTRCyaA⌬1-1006, in which the sequence coding for the last block of repeated sequences (between the SmaI and BspE1 sites) has been replaced by an appropriate synthetic double-stranded oligonucleotides encoding a termination sequence. The plasmid pTRCyaA⌬1491-1706 encoding the acylated truncated protein (CyaA 1-1490 ) is a derivative of pTRCAG (37), in which the molecule C-terminal DNA sequence (encompassing the RTX last block of repeated sequences and the secretion signal, located between the SmaI and BamH1 sites) was deleted and replaced by an appropriate synthetic double-stranded oligonucleotide encoding an amino acid termination sequence. The plasmid pTRCyaA⌬1-1489⌬1682-1706 encoding the last block of repeated sequences and its N-and C-terminal-flanking regions (CyaA 1490 -1681 ) is a derivative of pACT⌬1-1006 (17), constructed in two steps. First, the DNA sequence coding for the N-terminal part of the CyaA RTX domain (between the NdeI and SmaI sites) was deleted and replaced by an appropriate synthetic double-stranded oligonucleotide encoding the amino acids Met-Leu-Glu-Gly; then the C-terminal DNA sequence (located between the BspE1 and BamH1 sites) was replaced by a synthetic double-stranded oligonucleotide encoding a hexahistidine tag and a termination signal (Pro-Asp-His-His-His-His-His-His-Stop). The plasmid pTRCyaA⌬1-1527⌬1613-1706 encoding the RTX last block of repeated sequences (CyaA 1528 -1612 ) was obtained by amplifying by PCR the CyaA fragment located between amino acids 1528 and 1612, using as primers two single-strand synthetic oligonucleotides (5Ј-GGGAATTCCATATGGGCAGCGCGCGTGATGACGTGCTGA-TCGGC-3Ј and 5Ј-TCCGGTATCCGGATAGCGGATGGTGTCATC-FIGURE 1. Alignment of the RTX repeated sequences of CyaA. The potential RTX nonapeptide repeats localized within the last 700 residues of CyaA are aligned according to the RTX consensus GGXGX(D/N)X(U)X (U is any large hydrophobic residues), where X represents any amino acid. Highlighted in bold are the RTX sequences that match at least four of the five positions of the core consensus (underlined). Non-RTX sequences that interspersed the five separate blocks of successive RTX motifs (numbered I-V) are displayed on the left. The arrowheads indicated the position of the trypsin cleavage sites identified in limited proteolysis experiments (see Fig. 9). Numbers correspond to the amino acid position within the wild type CyaA sequence.
GCCGCCGCCCGATTC-3Ј, including in-frame NdeI and BspE1 sites, respectively). The fragment obtained was then inserted into the NdeI and BspE1 sites of pTRCyaA⌬1-1489⌬1682-1706 (see above), leading to the addition of an in-frame His tag in the C-terminal end of the CyaA 1528 -1612 protein.
Production and Purification of the CyaA-derived Proteins-Protocol for CyaA production has already been described elsewhere (25,38,39).
All proteins were expressed in E. coli BLR strains (Novagen, Merck). Wild type CyaA and CyaA 1-1490 were purified to greater than 95% homogeneity (as judged by SDS-gel analysis, Fig. 2B) from inclusion bodies by a two-step procedure including DEAE-Sepharose and phenyl-Sepharose chromatography as described (38,39). CyaA 1006 -1706 and CyaA 1006 -1490 were purified by Ca 2ϩ -dependent phenyl-Sepharose chromatography as described previously (25). The cells were disrupted by sonication in 50 mM Tris-HCl, pH 8.0, and 2 mM CaCl 2 . After centrifugation (20 min at 25,000 ϫ g at 4°C), the supernatant was loaded on a phenyl-Sepharose column equilibrated with 50 mM Tris-HCl, pH 8.0, and 2 mM CaCl 2 . The column was washed extensively with the same buffer, and the proteins were eluted in 50 mM Tris-HCl, pH 8.0, and 5 mM EDTA. CyaA 1006 -1706 and CyaA 1006 -1490 were then dialyzed against 20 mM ammonium bicarbonate, pH 7.5. CyaA 1490 -1681 and CyaA 1528 -1612 were directly purified from bacterial lysate (20 mM Hepes-Na, 50 mM NaCl, pH 7.5) on an immobilized nickel column (nickel nitrilotriacetic acid-agarose) according to the instructions of the manufacturer (Qiagen). The proteins eluted in the presence of 100 mM imidazole, pH 7.2, were then directly loaded (after dilution with 2 volumes of 20 mM Hepes-Na, pH 7.5) on a DEAE-Sepharose resin, equilibrated with 20 mM Hepes-Na, pH 7.5. After extensive washing with the 20 mM Hepes-Na, 100 mM NaCl, pH 7.5, the proteins were eluted in 20 mM Hepes-Na, 250 mM NaCl, pH 7.5, and dialyzed against 20 mM ammonium bicarbonate, pH 7.5. Protein concentrations were determined spectrophotometrically from the adsorption at 278 nm using a molecular extinction coefficient of 141 mM Ϫ1 cm Ϫ1 for the full length CyaA toxin, 132 mM Ϫ1 cm Ϫ1 for CyaA   Assay of Hemolytic and Invasive Activities-The hemolytic and cellinvasive activities of the CyaA molecules and/or mixed CyaA fragments were characterized on sheep erythrocytes as described previously (25,38). Purified proteins in 8 M urea, 20 mM Hepes-Na were directly diluted (at least 100-fold) into suspensions of sheep erythrocytes (5 ϫ 10 8 cells/ml in 20 mM Hepes-Na, pH 7.5, 150 mM NaCl, 2 mM CaCl 2 ). Adenylate cyclase activity was measured as previously described (25,38); one unit of enzymatic activity corresponds to 1 mol of cAMP formed/min at 30°C and pH 8.0.
Membrane Experiments-Black lipid bilayer membranes were formed as described previously (17,23,40). The instrumentation consisted of a Teflon chamber with two aqueous compartments connected by a small circular hole. The hole had a surface area of about 0.4 mm 2 . Membranes were formed across the hole by painting onto a 1% solution of asolectin (lecithin type III from soybeans from Sigma) in n-decane. The 1 M KCl solutions (Merck) were buffered with 10 mM Hepes-KOH to a pH around 7. The temperature was kept at 20°C throughout. The membrane current was measured with a pair of silver/silver chloride electrodes with salt bridges switched in series with a voltage source and an electrometer (Keithley 617). In the case of the channel recordings the electrometer was replaced by a homemade current amplifier. The amplified signal was monitored with a storage oscilloscope and recorded with a tape or a strip chart recorder.
Fluorescence Spectroscopy-The fluorescence spectra were recorded at 20°C on a 2-ml sample using an LS-5B spectrofluorimeter (PerkinElmer Life Sciences). The excitation wavelength was set at 292 nm with an excitation bandwidth of 5 nm, and the fluorescence emission spectra were recorded from 300 to 420 nm (emission bandwidth of 5 nm). CyaA 1006 -1706 and CyaA 1490 -1681 protein concentrations were 0.2 and 1.0 M, respectively, in a 20 mM ammonium bicarbonate buffer, pH 8. The fluorescence emission spectra of the proteins were recorded at different calcium concentrations ranging from 0 to 2 mM. After each addition of CaCl 2 , a 5-min equilibrium period was allowed before recording the spectrum. Finally, EGTA was added to a final concentration of 5 mM to record the fluorescence spectra of calcium-free proteins. The background fluorescence was recorded similarly in the absence of added proteins and subtracted from all spectra. Circular Dichroism Spectroscopy-Circular dichroism spectra were recorded on a Jobin-Yvon CD6 dicrograph at 20°C with a 0.02-cm path length cylindrical suprasil quartz cell (Hellma). Protein samples were equilibrated in 20 mM ammonium bicarbonate, pH 8.0, by overnight dialysis at 4°C, and spectra were determined at a protein concentration of 0.5 mg/ml in the presence or absence of 2 mM CaCl 2 . Far UV region spectra represent the average of 5 successive scans between 180 and 260 nm with 0.5-nm steps using an integration time of 5 s between 180 and 200 nm and of 1 s between 200 to 260 nm. Base lines acquired on the 20 mM ammonium bicarbonate, pH 8, buffer (with or without the addition of 2 mM CaCl 2 ) under the same conditions were subtracted, and spectra were deconvoluted using the CDSSTR program from W. C. Johnson (41) included in the CDPro software package.
Analytical Ultracentrifugation-CyaA RTX fragments were subjected to sedimentation equilibrium on a Beckman XL-A analytical ultracentrifuge equipped with standard double sector cells with 1.2-mm thick aluminum centerpieces. Protein samples were dialyzed extensively at 4°C against 50 mM Hepes-Na, pH 7.5, 100 mM NaCl, 2 mM CaCl 2 . Insoluble materials were removed by centrifugation at 14,000 ϫ g for 15 min, and aliquots of CyaA 1006 -1706 (0.5 mg/ml; 6.95 M) and CyaA 1490 -1681 (0.71 mg/ml; 31.9 M) were then centrifuged at 20°C at 20,000 and 30,000 rpm respectively, until perfect superposition of consecutive scans (about 20 h). The final equilibrium distributions were determined from absorption measurements at 280 nm. Theoretical partial specific volumes, calculated from the amino acid composition of CyaA 1006 -1706 and CyaA 1490 -1681 were 0.717 and 0.719 ml⅐mg Ϫ1 , respectively, and the solvent density of the Hepes/NaCl buffer was calculated to be 1.003 g ml Ϫ1 at 20°C. To calculate molecular weights, data were fitted using the Origin-based Optima XL-A data analysis software (Beckman). Different fitting models (single ideal component, two ideal components, and association of identical ideal components) for single data sets were systematically tested, and the best fit was retained on the basis of both the 2 value and the lack of systematic deviation of the residuals.
Ca 2ϩ Binding Assay-Binding of calcium to CyaA 1006 -1706 , CyaA 1490 -1681 , and CyaA 1528 -1612 was determined by ultrafiltration, as described previously using radioactive 45 Ca as a tracer (42). Purified proteins were extensively dialyzed against 20 mM Hepes-Na, pH 7.5, 100 mM NaCl, and insoluble materials were removed by centrifugation at 14,000 ϫ g for 15 min (42). For the binding assays, 0.2-ml samples of a 12-90 M solution of dialyzed proteins were placed in the top compartment of Amicon Ultrafree-MC centrifugal filter devices (Millipore Corporation, Bedford, MA; molecular mass cut-off of 5000 Da). Ten l of a 5 mM CaCl 2 solution containing about 500 cpm of 45 Ca/nmol of calcium were added to the protein and thoroughly mixed by vortexing. Ten l of this solution were taken for counting of total radioactivity in a scintillation counter. The mixtures were then centrifuged at room temperature in a tabletop centrifuge for about 30 s until 10 l were filtrated through. The ultra-filtrate was added back to the top compartment (i.e. protein solution) and again mixed and centrifuged as above (this second centrifugation was done to alleviate any dilution due to the dead-volume of the filtration membrane). Radioactivity of a 10-l sample of the filtrate, which contained the free calcium, was determined by scintillation counting. Then 10 l of the radioactive CaCl 2 solution as well as 10 l of the protein solution were added to the top compartment, and the whole centrifugation procedure described above was repeated. At each step the ratio radioactivity in 10 l of filtrate/radioactivity in 10 l of top compartment is equal to free calcium/total added calcium. This allows calculating the number of calcium ions bound per protein molecule as a function of free calcium.
Limited Proteolysis-CyaA 1006 -1706 or CyaA 1490 -1681 was diluted to 1.0 mg/ml in 20 mM Hepes-Na, pH 7.5, in the presence of either 10 mM EDTA or 10 mM CaCl 2 . The solutions were equilibrated at 37°C, and then trypsin was added to a final concentration of 1.0 g/ml. After 5, 10, and 30 min of incubation at 37°C, aliquots of 20 l (20 g of protein) were taken out from the reaction mixture, and proteolysis was stopped by the addition of 500 M 4-(2-aminoethyl) benzene sulfonyl fluoride and SDS-gel loading buffer. The samples were stored at Ϫ20°C until analysis by gel electrophoresis. N-terminal sequencing of the proteolytic fragments (separated by SDS-PAGE, electrotransferred to polyvinylidene difluoride membrane, and stained with Coomassie Blue) was performed on an Applied Biosystems ABI 494 sequencer at the Plate-forme de microséquencage of Institut Pasteur. Kinetic analyzes of the proteolytic cleavage of the CyaA 1006 -1706 or CyaA 1490 -1681 were carried out by recording the decay of the protein intrinsic fluorescence. CyaA 1006 -1706 or CyaA 1490 -1681 was diluted to 0.5 M and 2 M, respectively, in 20 mM Hepes-Na, pH 7.5, containing either 10 mM EDTA or 10 mM CaCl 2 . The fluorescence emission intensities were recorded at 320 nm (the excitation wavelength was 292 nm with a 5-nm band pass) on a Jasco FP-750 fluorimeter thermostatted at 37°C under constant agitation. Proteolysis was initiated by the addition of 3 nM (final concentration) TPCK-treated trypsin, and the kinetics were recorded for 2000 s.

RESULTS
Functional Complementation of the Truncated CyaA 1-1490 by Different RTX Fragments-Several fragments derived from the CyaA RTX domain ( Fig. 2A) were expressed in E. coli and purified to near homogeneity as judged from SDS-PAGE analysis (Fig. 2B). CyaA 1006 -1706 and CyaA 1006 -1490 were purified by Ca 2ϩ -dependent phenyl-Sepharose chromatography. A hexahistidine tag was genetically fused to the C terminus of the two shorter variants to permit their purification on a nickel column. These polypeptides were first tested for their ability to functionally complement an inactive CyaA variant, CyaA 1-1490 , deleted from its last 217 C-terminal residues. This truncated toxin had neither hemolytic nor cytotoxic activities (Fig. 3) as previously reported (32,33). Similarly, the RTX-derived polypeptides were not hemolytic (data not shown) and obviously not cytotoxic as they have no adenylate cyclase catalytic domain. As shown in Fig. 3, the full-length RTX domain (CyaA 1006 -1706 ) was able to partially restore the hemolytic and cytotoxic activities of the truncated CyaA 1-1490 , whereas the truncated RTX The incubation was carried out in presence of 2 mM CaCl 2 overnight at 37°C for the hemolytic assays and for 20 min at 30°C for the catalytic activity assay. No activity was observed in absence of CaCl 2 (data not shown). Results are presented as the means for triplicate assays, which differed by less than 10%.
domain, lacking the last block (block V) of repeated sequences, CyaA 1006 -1490 , had no complementing activity. The last 217 C-terminal residues of CyaA, which contain the block V of RTX motifs with N-and C-terminal-flanking regions (CyaA 1490 -1681 ), was also able to restore a significant hemolytic activity and a low but detectable cytotoxic activity of CyaA  . The partial complementation of CyaA 1-1490 by CyaA 1490 -1681 suggests that this latter fragment had a lower affinity for CyaA 11490 than the full-length RTX domain. In contrast, the polypeptide fragment, CyaA 1528 -1612 , which corresponds precisely to the nine RTX nonapeptide motifs of block V (that is, without the N-and C-terminal-flanking regions), was unable to restore the hemolytic and cytotoxic activities of CyaA 1-1490 . This indicates that the polypeptide sequences flanking the RTX block V have an essential functional role in restoring the biological activities to CyaA 1-1490 (Fig. 3), in agreement with previous report of Bejerano et al. (33).
CyaA is able to form small cation-selective channels in lipid bilayers (17,18,23). Recent studies have shown that the addition of calcium above 0.6 -0.8 mM causes a large increase of bilayer conductance without noticeably changing the unit conductance of the channels formed by CyaA (23). We examined the channel-forming ability of the truncated CyaA variant, CyaA 1-1490 , in the absence or in the presence of the two RTX-derived polypeptides, CyaA 1006 -1706 or CyaA 1490 -1681 . The truncated form of CyaA formed channels that were indistinguishable from those formed by wild type CyaA, which means that they had a The channel-forming activity of CyaA derivatives was assessed on membranes formed from asolectin/n-decane. The aqueous phase contained 1 M KCl, pH 7. The temperature was 20°C, and the applied voltage was 50 mV at the cis side. The panels show the original strip chart recordings of membrane current/conductance after the indicated addition to the cis side of the membrane, proteins, CaCl 2 , or EDTA as a calcium chelator. A, 440 ng/ml CyaA 1-1490 was added to the cis side of the membrane (left arrow), which led to a small increase of membrane conductance. About 7 min later 1.8 mM CaCl 2 was added to the cis side (middle arrow), which did not influence the membrane conductance. After about 4 min, 1 g/ml CyaA 1006 -1706 was added to the cis side of the membrane, resulting in a steep increase of membrane conductance. B, 440 ng/ml CyaA 1-1490 was added to the cis side of the membrane about 6 min before the start of the recording, which led to a small increase of membrane conductance. About 11 min afterward (5 min after the start of the recording) 2 mM CaCl 2 was added to the cis side (left arrow) followed by the addition of 1.6 g/ml CyaA 1490 -1681 3 min later (middle arrow). Note that the CyaA 1-1490 -mediated conductance only increased when the calcium concentration was increased to 4 mM (right arrow). C, 160 ng/ml wild-type CyaA was added to the cis side of the membrane about 10 min before the start of recording. About 14 min afterward (4 min after the start of recording), 1 mM CaCl 2 was added to the cis side (left arrow), resulting in a steep conductance increase. The addition of 4.5 mM EDTA to the cis side (right arrow) led to a transient decrease of the membrane conductance. The amplification of the signal was decreased 10 times toward the end of recording. Note that the addition of EDTA stopped any further CyaA-mediated conductance increase.
single channel conductance of about 50 picosiemens in 1 M KCl, 10 mM Hepes-KOH, pH 7 (data not shown). As shown in Fig. 4A, irrespective of the presence or the absence of calcium ions in the medium, CyaA 1-1490 caused a small increase in conductance across the bilayer similar to that observed previously with the full-length wild-type CyaA in the absence of calcium or in the absence of the RTX domain (23). The full-length RTX domain, CyaA 1006 -1706 or CyaA 1490 -1681 did not affect the CyaA 1-1490 -mediated conductance when added in the absence of calcium or low calcium concentration (see Fig. 4B). However, when sufficient calcium was present in the medium, the addition of either CyaA 1006 -1706 (or of CyaA 1490 -1681 ) strongly stimulated the overall conductance of CyaA 1-1490 (see Fig. 4, A and B). This result suggested that the RTX-derived polypeptides were able to mimic the effect of the missing RTX domain in terms of channelforming activity in asolectin membranes.
Interestingly, the addition of the calcium-chelator EDTA only transiently blocked cation conductance of either the wild type CyaA or of the complemented pair CyaA 1-1490 /CyaA 1006 -1681 or CyaA 1-1490 / CyaA 1490 -1681 . Fig. 4C shows an experiment of this type for wild-type CyaA. 160 ng/ml wild-type CyaA was added to the cis side of a black asolectin lipid bilayer. After about 10 min, 1 mM CaCl 2 was also added to the cis side of the membrane to enhance channel formation (left arrow in Fig. 4C). After about 4 min, 4.5 mM EDTA was added to the cis side to reduce the calcium concentration below the critical concentration (right arrow in Fig. 4C). Subsequently, the membrane conductance showed some transient decrease presumably caused by the addition of EDTA to the cis side. Afterward, the CyaA-mediated conductance was constant and did not increase further (see Fig. 4C). This was also observed when the complemented pair CyaA 1-1490 /CyaA 1006 -1681 or CyaA 1-1490 /CyaA 1490 -1681 was used in the experiments. These results suggested that calcium binding to the RTX domain of CyaA is required mainly for the insertion of the toxin into the membrane but not for the maintenance of the CyaA-mediated conductance when the channels were already formed.
Altogether these data confirm that CyaA 1490 -1681 constitutes an autonomous domain capable of restoring partial hemolytic and cytotoxic activities to the inactive truncated CyaA lacking the corresponding polypeptide segment. Nevertheless, some differences concerning the calcium-induced activation of the CyaA-mediated conductance were observed. Whereas 0.6 -0.8 mM calcium was sufficient to induce conductance increase of wild-type CyaA, the concentration for the complemented pair CyaA 1-1490 /CyaA 1006 -1681 was about 1.5 mM calcium, whereas about 4 mM calcium were needed to increase conductance in the case of CyaA 1-1490 /CyaA 1490 -1681 (see Fig. 4B).
Fluorescence Spectroscopy of the CyaA RTX Fragments-CyaA 1006 -1706 and CyaA 1490 -1681 contain seven and 2 tryptophan residues, respectively. The tryptophan emission spectra of both proteins were recorded in the presence and absence of Ca 2ϩ . The binding of Ca 2ϩ induced an ϳ2-fold increase in the fluorescence intensity of both proteins with no significant shift in the emission maximum wavelength (Fig. 5A). The dependence of the fluorescence emission intensity at 336 nm upon the added calcium concentrations is shown in Fig. 5B. In both cases the maximal change in fluorescence signal was reached at a calcium concentration of about 1 mM. Fast kinetic analysis of the fluorescence changes upon the addition (or removal) of calcium indicated that the binding (or dissociation) of calcium was a very fast process (data not shown); the apparent k on and k off constants were in fact too high to be accurately determined because a major fraction of the overall fluorescence changes occurred during the mixing time (4 ms) of stopped-flow apparatus. Altogether, these data show that the tryptophans of CyaA 1006 -1706 and CyaA 1490 -1681 are sensitive to conformational changes induced by Ca 2ϩ binding. Interestingly, the two tryptophan residues of CyaA 1490 -1681 (residues at position 1621 and 1645 of wild type CyaA) are located outside the RTX nonapeptide motifs, in the C-terminal extension (see Fig. 1). Hence, the calcium-induced enhancement of the tryptophan fluorescence of CyaA 1490 -1681 strongly suggests that the C-terminal peptide downstream of the block V undergoes a structural rearrangement upon calcium binding to the adjacent RTX motifs.
The effect of various divalent cations on the fluorescence of CyaA 1006 -1706 and CyaA 1490 -1681 were also tested. As shown in Fig. 5C, the fluorescence intensities at 336 nm ( max ) of CyaA 1006 -1706 and CyaA 1490 -1681 were unaffected by the addition of 2 mM MgCl 2 , MnCl 2 , NiCl 2 , RbCl 2 , or CoCl 2 . Only SrCl 2 induced a significant enhancement of the fluorescence of these proteins. These data are in good agreement with electrophysiological studies that revealed that only SrCl 2 can trigger a very small increase of CyaA-mediated membrane conductance in lipid bilayers starting with about 3 mM, whereas Mg 2ϩ , Sr 2ϩ , and Ba 2ϩ cations could not (23). Yet Rhodes et al. (43) detected Mn 2ϩ binding to full-length CyaA by EPR spectroscopy and reported that Mn 2ϩ at high concentrations (above 10 mM) was able to confer some cytotoxic activity to CyaA. At variance with their results, in our conditions no hemolytic activity and also no channel formation in lipid bilayers could be detected in the presence of Mn 2ϩ .
Because CyaA 1528 -1612 contained no tryptophan but 3 tyrosine residues (Fig. 1), we attempted to examine the effect of calcium on the tyrosine fluorescence properties of this polypeptide. The tyrosine fluorescence emission spectrum of CyaA 1528 -1612 was unaffected by the presence or the absence of Ca 2ϩ (data not shown); therefore, no conclusion regarding calcium binding to CyaA 1528 -1612 could be drawn.
Circular Dichroism of the CyaA RTX Fragments in the Far UV Region-To further characterize the structural changes induced by calcium binding, the secondary structure of the different fragments were studied by circular dichroism. The far UV spectra of the proteins, equilibrated in ammonium bicarbonate, were recorded in the presence or the absence of 2 mM CaCl 2 . As shown in Fig. 6, calcium binding triggered a large change in the CD spectra of CyaA 1006 -1706 and CyaA 1490 -1681 , characterized by an increase of the positive band at 192 nm and a decrease of the negative band at 222 nm. In contrast, the CD spectrum of CyaA 1528 -1612 was not affected by the addition of calcium (Fig. 6C). When equilibrated in 6 M guanidinium hydrochloride, the spectrum of CyaA 1490 -1681 had the characteristics of an unfolded polypeptide. The addition of calcium had no effect on the CD spectrum recorded in these conditions. Therefore, the changes of the CD spectra observed upon the addition of calcium to CyaA 1490 -1681 and CyaA 1006 -1706 equilibrated in ammonium bicarbonate buffer represent authentic calcium-induced conformational modifications of the secondary structures of the native polypeptide(s). The CD spectra of the different polypeptides were deconvoluted according to the CDSSTR procedure of Johnson (41), and the deduced compositions of secondary structures are shown in Table 1. This analysis shows that Ca 2ϩ binding to CyaA 1006 -1706 is accompanied by an increase in the ␣-helical and ␤-strand contents of the protein, respectively, whereas calcium binding to CyaA 1490 -1681 mainly induces an increase in its ␤-structure content. As expected, the secondary structure content of CyaA 1528 -1612 was unchanged upon the addition of calcium. Very similar deconvolution results were obtained with the CON-TINLL program included in the CDPro package.
From the above experiments we conclude that the full-length CyaA RTX domain as well as the C-terminal fragment, CyaA 1490 -1681 , are both sensitive to calcium binding. Surprisingly, the short fragment,   Fig. 7, A and B. The curves through the data are global least square fits of the expression for a single ideal species. The small, symmetrically distributed residuals (top panels) demonstrate the compatibility of the single-species model with the data. The molecular weights returned by these analyses were 75,230 Ϯ 1,400 for CyaA 1006 -1706 (Fig. 7A) and 21,260 Ϯ 1,100 for CyaA 1490 -1681 (Fig. 7B), in good agreement with the molecular weights derived from sequence (M r of CyaA 1006 -1706 ϭ 72,400; M r of CyaA 1490 -1681 ϭ 22,800). These data demonstrate that both CyaA 1006 -1706 and CyaA 1490 -1681 proteins are in a monomeric state in the presence of calcium, thus excluding that the calcium-induced changes in the secondary structure were due to protein oligomerization and/or aggregation.
Ca 2ϩ Binding Properties of the CyaA-RTX Fragments-Direct Ca 2ϩ binding to the CyaA-RTX fragments was carried out by ultrafiltration using 45 Ca 2ϩ as tracer. As shown in Fig. 8, the full-length RTX domain of CyaA, CyaA 1006 -1706 , bound about 30 Ca 2ϩ ions with an affinity in the millimolar range. This stoichiometry is in good agreement with the number of canonical RTX sequences (as defined in Fig. 1) (15) found in the RTX domain of CyaA, suggesting that each of these motifs can bind one calcium ion. The additional less conserved RTX motifs might also bind calcium in certain conditions as was found in previous studies (25,43). The CyaA 1490 -1681 fragment bound ϳ6 -7 calcium ions per molecule at the highest free calcium concentration tested. This value is in good agreement with the presence of 7 canonical RTX repeat motifs in this polypeptide and confirms that each of these consensus nonapeptide repeated sequences could bind a calcium ion at saturation. Surprisingly,

in the absence or in the presence of calcium
Deconvolution of spectra shown in Fig. 4 was performed using the CDSSTR software as described under "Experimental Procedures." Deconvolution of spectra was also carried out with the CONTINLL software with very similar results (not shown). the shorter polypeptide CyaA 1528 -1612 was unable to bind calcium in our experimental conditions, although it possesses the same RTX repeated sequences as CyaA 1490 -1681 . These data indicate that binding of calcium to these RTX motifs is critically dependent upon the presence of the adjacent polypeptide sequences that might directly participate in the stabilization of the parallel ␤-helix fold. Limited Proteolysis of the CyaA-RTX Fragments-Limited proteolysis experiments were carried out to probe the conformational changes induced by calcium binding to the CyaA-RTX fragments. CyaA 1006 -1706 and CyaA 1490 -1681 were incubated in the presence of either calcium or EDTA, with trypsin (at a 1/1000 ratio) for various times, and analyzed by SDS-PAGE. As shown in Fig. 9, A and B, the two proteins in their calcium-free form were rapidly proteolyzed as compared with the calcium-bound polypeptides. Kinetic analysis of the proteolytic degradation, as monitored by the decay of intrinsic fluorescence of tryptophan residues (Fig. 9C), confirmed that the calcium-free forms of the CyaA fragments were much more susceptible to trypsin proteolysis than calcium-bound forms. Fig. 9D shows the N-terminal  (fragments a, b, c1, c2, d) or CyaA 1490 -1681 (fragment e) were determined by microsequencing of the polypeptides separated by SDS-PAGE and electrotransferred to a polyvinylidene difluoride membrane. The position of the N-terminal residues in the CyaA sequence is indicated in the 3rd column. Note that the two fragments, c1 and c2, exhibited the same electrophoretic mobility.

Secondary structure composition
sequences of the main trypsin-resistant fragments derived from the calcium-bound CyaA 1006 -1706 (fragments a-d) or CyaA 1490 -1681 protein (fragment e). These data revealed that the main cleavage sites are located within the polypeptide segments that link the different RTX blocks (Fig. 1). The kinetics of appearance of these proteolytic fragments (Fig. 9A) suggest a hierarchical process for the proteolysis of calcium-bound CyaA 1006 -1706 in which the protein was first cleaved at Arg-1367 into fragments a and b (Fig. 9). Fragment a was further cleaved at Arg-1220 into fragments c1 and d, whereas fragment b was cleaved at Arg-1491 to yield fragment c2. When the proteins were digested in the presence of EDTA, additional cleavage sites were found within the RTX blocks (data not shown). This suggests that, in the absence of calcium, these proteolytic cleavage sites were much more exposed to the protease than they are in the calcium-bound conformation. Altogether, these data indicate that, upon calcium-binding, both CyaA 1006 -1706 and CyaA 1490 -1681 adopt compact conformations as compared with the calcium-free proteins, thus shielding potential cleavage sites from the action of trypsin.

DISCUSSION
The RTX motif is a structural motif found in a number of cytolysins produced by Gram-negative bacteria as well as in several other potential bacterial pathogenic proteins (20,21). The name RTX stems from a Glyand Asp-rich nonapeptide repeat of canonical sequence GGXGX-DXUX, present in variable number, generally in a tandem fashion. This represents a specific type of Ca 2ϩ binding domain that is essential for the function of the proteins. Indeed most of the cytolysins are calciumdependent, and it has been shown that purified E. coli ␣-hemolysin and B. pertussis CyaA can bind calcium in solution (25, 44 -46). Threedimensional structure analysis of the P. aeruginosa alkaline protease, which possesses six of these consensus RTX motifs, revealed that these repeated sequences constitute a new kind of calcium binding structure called a parallel ␤-helix or parallel ␤-roll (47). In this structure the first six residues of each motif form a turn that binds calcium, and the remaining three residues build a short ␤-strand. The arrangement of consecutive turns and ␤-strands builds up a right-handed helix of paralleled ␤-strands, one turn of this helix consisting in two consecutive nine residues. Calcium is bound between two consecutive turns by the conserved aspartic acids. We previously showed that CyaA, which possesses about 30 -40 of such RTX sequences, binds a large number of calcium ions in solutions with an affinity in the millimolar range, suggesting that each RTX motif can bind a calcium ion (25). A noticeable difference is that calcium ions appeared to be bound to the alkaline protease with a much higher affinity, as they could not be removed from the protein by calcium-chelators such as EDTA. Although the role of the RTX motifs in calcium binding is clearly demonstrated, the structural and functional consequences of calcium binding to these domains remain largely unknown.
In this work we have characterized an essential subdomain from the RTX-repeated structure of CyaA, located at the C-terminal moiety of the toxin. This CyaA fragment encompasses the last block of RTX repeat motifs (block V) as well as flanking non-RTX sequences at both the N and C termini. We showed that this fragment, CyaA 1490 -1681 , constitutes an autonomous domain that is essential for the cytotoxic and hemolytic activities of CyaA, in agreement with previous results from Iwaki et al. (32) and Bejerano et al. (33). The CyaA 1490 -1681 fragment is able to partially restore hemolytic and cytotoxic activities to an inactive, truncated CyaA lacking the corresponding segment (CyaA 1-1490 ). Electrophysiological characterization of the pore-forming capacities of the toxin on artificial lipid bilayer supports the view that the CyaA 1490 -1681 subdomain is required for a proper insertion of the molecule into the plasma membrane of target cells but is not by itself necessary for the maintaining of an open pore conformation. The ability of the CyaA 1490 -1681 fragment to partly complement the inactive, truncated CyaA 1-1490 implies that these two polypeptides could assemble to form an active complex able to bind to and insert into the membrane of target cells. However, no evidence of direct interaction between the two complementary CyaA polypeptides could be obtained in experiments of co-retention/purification of the CyaA 1-1490 protein onto the His-tagged CyaA 1490 -1681 fragment immobilized on a metal affinity column. 3 Most likely, the association of CyaA 1490 -1681 and CyaA 1-1490 might be undetectable in vitro in the absence of target cells and/or artificial membranes. Interestingly, Bejerano et al. (33) previously showed that the polypeptide composed of the last 217 C-terminal residues of CyaA could associate in a calcium-dependent manner with another inactive truncated CyaAs that lacked the 76 C-terminal amino acids. A stable association between CyaA complementary fragments might require an extended overlap between them (in the Bejerano et al. (33) study, the complementing fragments shared 141 overlapping residues including all the RTX motifs) and/or the very C terminus of CyaA (residues 1682-1706).
The CyaA 1490 -1681 functional module is made of the last block of RTX repeat motifs (block V) flanked by non-RTX sequences at both N and C termini. Strikingly, we found that these flanking sequences were essential for calcium binding to the RTX repeat motifs. The CyaA 1528 -1612 polypeptide that contains only the 9 RTX tandem repeats of block V was unable to bind calcium, and its secondary structure was not affected by the presence or absence of calcium. However, when the 9 RTX tandem repeats of block V were expressed with the N-and C-terminal-flanking sequences, the CyaA 1490 -1681 protein was able to bind up to 6 -7 calcium ions per polypeptide with an affinity in the millimolar range. Because the first two of the nine repeats of block V only poorly match the RTX consensus sequence (see Fig. 1), this result suggests that only the seven truly canonical RTX nonapeptide repeats could bind calcium in vitro. This is in good agreement also with the stoichiometry of about 30 calcium ions bound to the whole RTX domain (CyaA 1006 -1706 ) that contains 31 canonical RTX sequences as defined in Fig. 1. Whether the degenerated RTX motifs might also bind calcium and/or contribute to structural stabilization of the calciumbound conformation of the CyaA repeat domain, in particular when the toxin is bound to the receptor and target cell membrane, remains to be clarified. Welch (48) has indeed proposed a model in which calcium binding to the non-canonical RTX repeats of HlyA (and other RTX cytolysins) could contribute to the transition from the water-to the lipid-soluble state of these proteins.
Fluorescence, circular dichroism spectroscopy, and limited proteolysis studies revealed that, upon binding of calcium, the CyaA 1490 -1681 protein undergoes significant conformational changes, as is the case for the whole RTX domain of CyaA. The increase in the ␤-sheet content of CyaA 1490 -1681 upon calcium binding may indeed indicate a stabilization of the ␤-roll structure. Furthermore, the 2-fold enhancement of the tryptophan fluorescence emission intensity of the CyaA 1490 -1681 protein upon binding of calcium is particularly noteworthy, as the two tryptophan residues of the protein are located outside the RTX motifs within the C-terminal-flanking polypeptide segment (see Fig. 1). This indicates that these two residues (Trp-1621 and Trp-1645) experienced a different environment in the calcium-bound versus calcium-free forms of the CyaA 1490 -1681 protein. Hence, calcium binding to the RTX motifs of CyaA 1490 -1681 is able to induce conformational changes in the adjacent C-terminal non-RTX polypeptide segment.
Altogether, these results suggest a model for the folding of this essential RTX motif in which the functional calcium binding structure extends beyond the RTX glycine/aspartate-rich motifs to include adjacent polypeptide sequences that might directly participate in the stabilization of the parallel ␤-helix fold. Upon calcium binding to the RTX repeat motifs of CyaA 1490 -1681 , the N-and C-terminal polypeptide extensions could interact with the RTX ␤-roll structure to stabilize its calcium-bound conformation. Such a model is in line with the previous study of Lilie et al. (28), who showed that a synthetic 75-mer polypeptide containing exactly 8 consensus RTX motifs of the sequence GGS-GNDNLS did not possess any regular secondary structure even in the presence of high concentration of Ca 2ϩ (100 mM). ␤-Structure could only be detected upon the addition of Ca 2ϩ and polyethylene glycol, an unspecific secondary-structure stabilizing agent. Further studies showed that the acquisition of secondary structure was tightly associated with the polymerization of this model RTX polypeptide, suggesting that intermolecular interactions were essential to stabilize the ␤-roll fold of this synthetic protein. The properties of this synthetic minimal RTX peptide are somewhat similar to what was observed with the CyaA 1528 -1612 fragment, containing only the block V RTX repeats, in that they are both unable to respond to calcium. It is noteworthy that in the crystal structure of the P. aeruginosa alkaline protease, the RTX ␤-roll is tightly packed against the protease domain (47). This may indicate the RTX ␤-roll is generally unstable unless stabilizing contacts can be established with additional polypeptide segments.
The observation reported here on the CyaA subdomains might be pertinent to other RTX cytolysins. We propose that the RTX functional unit consists of the tandem repeated nonapeptide motifs flanked by adjacent polypeptide segments that are essential for the calcium-induced folding of the protein. Calcium binding to the RTX nonapeptide repeats induces a conformational rearrangement of the adjacent sequences, acting as a molecular switch to trigger the biological activity of the proteins. Indeed, prior studies on different RTX cytolysins have highlighted the critical role of polypeptide segments adjacent to the RTX-repeated sequences. In the case of CyaA, Bejerano et al. (33) have shown that a stretch of 15 amino acids called "block A," located C-terminal to the last RTX repeats, is essential for the toxic activity of the protein since it is required for the toxin binding and insertion into the membrane. Hence, a conformational change in the CyaA block A segment as a result of calcium binding to the adjacent RTX motifs might be required for the toxic activity of the protein. It is noteworthy than one of the 2 tryptophan residues from the CyaA 1490 -1681 , Trp-1621, shown to be sensitive to calcium binding, belongs to a highly conserved triamino acid stretch (N/D)W(F/Y) in the core of the block A segment. Deletion and/or mutations of the homologous A block of the E. coli ␣-hemolysin, were similarly shown to abolish the hemolytic activity of HlyA (49). Besides, Cortajarena et al. (50) showed that the same region of HlyA is also a major determinant for the specific binding of HlyA to the red blood cell surface through interaction with glycophorin. Other studies on related RTX toxins have suggested that the polypeptide segments at the C terminus of the RTX motifs might be involved in the binding of these toxins to their ␤ 2 -integrin receptor on target cells (51). We also showed recently that a region essential for CyaA binding to its cellular CD11b/CD18 (␣ M␤2 integrin) receptor encompasses a non-RTX segment between the RTX block II and block III motifs. Calcium-induced conformational rearrangements of this segment might explain the strict calcium dependence of the CyaA/CD11b association (31).
One of the key questions that remains to be clarified is how calcium binding to the repeat region located at the C terminus of CyaA can modulate insertion of the molecule into the target cell membrane and formation of cation channels. Both events involve the hydrophobic segments located in the central region of the protein. This implies that the Ca 2ϩ -induced conformational changes of the RTX domain, as evidenced by CD and fluorescence spectroscopy, must be transmitted somehow to the hydrophobic region of CyaA. In the case of E. coli ␣-hemolysin, Schindel et al. (52) have shown that Ca 2ϩ binding to HlyA triggers fluorescence changes of a fluorescent probe introduced in the hydrophobic region of the protein. One attractive hypothesis, in line with the model proposed earlier by Welch (48), could be that binding of Ca 2ϩ to the RTX nonapeptide repeats could trigger their transient association with distinct segment(s) of the protein, thus promoting the transition from the water-soluble to the membrane-associated state of the cytolysins. Validation of this model will await further characterization of such putative intramolecular interactions.