Allosteric Activation of Bordetella pertussis Adenylyl Cyclase by Calmodulin

Background: Adenylyl cyclase (AC) from Bordetella pertussis is activated when it interacts with calmodulin (CaM). Results: A triple mutant of AC, which was predicted by molecular modeling, exhibited a highly reduced affinity for CaM. Conclusion: This study suggests that a long range connection between CaM and the AC catalytic loop is crucial for AC activation. Significance: Molecular modeling identified critical molecular determinants for the allosteric activation of AC. Adenylyl cyclase (AC) toxin is an essential toxin that allows Bordetella pertussis to invade eukaryotic cells, where it is activated after binding to calmodulin (CaM). Based on the crystal structure of the AC catalytic domain in complex with the C-terminal half of CaM (C-CaM), our previous molecular dynamics simulations (Selwa, E., Laine, E., and Malliavin, T. (2012) Differential role of calmodulin and calcium ions in the stabilization of the catalytic domain of adenyl cyclase CyaA from Bordetella pertussis. Proteins 80, 1028–1040) suggested that three residues (i.e. Arg338, Asn347, and Asp360) might be important for stabilizing the AC/CaM interaction. These residues belong to a loop-helix-loop motif at the C-terminal end of AC, which is located at the interface between CaM and the AC catalytic loop. In the present study, we conducted the in silico and in vitro characterization of three AC variants, where one (Asn347; ACm1A), two (Arg338 and Asp360; ACm2A), or three residues (Arg338, Asn347, and Asp360; ACm3A) were substituted with Ala. Biochemical studies showed that the affinities of ACm1A and ACm2A for CaM were not affected significantly, whereas that of ACm3A was reduced dramatically. To understand the effects of these modifications, molecular dynamics simulations were performed based on the modified proteins. The molecular dynamics trajectories recorded for the ACm3A·C-CaM complex showed that the calcium-binding loops of C-CaM exhibited large fluctuations, which could be related to the weakened interaction between ACm3A and its activator. Overall, our results suggest that the loop-helix-loop motif at the C-terminal end of AC is crucial during CaM binding for stabilizing the AC catalytic loop in an active configuration.

The adenylyl cyclase toxin (CyaA) from Bordetella pertussis, the causative agent of whooping cough, plays an essential role in host invasion (1)(2)(3). CyaA is able to invade eukaryotic target cells, where it is activated by interacting with calmodulin (CaM), 5 thereby overproducing cAMP, which disorganizes cellular signaling processes and triggers host cell death (4 -12). Several crystallographic structures have been reported of the CyaA catalytic domain (adenylyl cyclase; AC) in complex with the C-terminal moiety of CaM (C-CaM) (13) (Fig. 1). These structures are physiologically relevant because C-CaM has been shown (14) to be as potent in activating the enzymatic activity of AC as the full-length CaM. In the AC⅐C-CaM complex (Fig. 1a), the AC domain includes three main regions: CA (residues 7-61, 187-197, 261-299, and 313-345; colored in green) in the middle of the structure and CB (residues 62-186; colored in orange) and SA (residues 192-254; colored in purple) at the two extremities. The C-terminal tail (residues 346 -364; colored in cyan) and the C-loop of the catalytic site (residues 300 -312; colored in yellow) are found in the CA region. At the C-terminal end of AC, a loop-helix-loop (LHL) motif that includes residues from Arg 338 to Asp 360 (Fig. 1c) is present at the interface between C-CaM and a major AC catalytic loop that spans residues 300 -312. In C-CaM, the calcium ions are bound to EF-hands 3 and 4, where EF-hand 3 is close to the interaction interface with AC.
In a previous study (15), we analyzed the interaction between the catalytic domain of AC from B. pertussis and C-CaM based on the molecular dynamics (MD) trajectories using maps of the energetic influences (16,17). Three residues in the LHL motif (i.e. Arg 338 , Asn 347 , and Asp 360 ) (Fig. 1b), were predicted to be important for the stability of the AC/CaM interaction, where Arg 338 and Asp 360 make contact with C-CaM, whereas Asn 347 is in direct contact with the AC catalytic loop. In the present study, to further characterize the potential roles of these residues in AC/CaM interactions, we performed in silico MD simulations and in vitro biochemical and biophysical studies of the three modified AC proteins, where one (Asn 347 ; ACm1A), two (Arg 338 and Asp 360 ; ACm2A), or three residues (Arg 338 , Asn 347 , and Asp 360 ; ACm3A) were substituted with Ala. We found that the ACm3A variant exhibited a strongly reduced affinity for CaM, whereas the affinities of the ACm1A and ACm2A variants were similar to that of the wild-type enzyme. This demonstrated the strong synergistic effects of the three mutations. The MD simulations showed that the differences in the behavior of the modified and wild-type complexes were related to differences in the internal mobility of the C-CaM calcium loops. These differences in mobility may induce direct or indirect weakening of the C-CaM/AC interaction observed in the ACm3A modified protein. These results as well as an analysis of the geometrical strain suggest that the LHL motif at the C-terminal end of AC is important for establishing a long range connection between CaM binding and the stabilization of the AC catalytic loop in an active configuration; thus, it is a critical module during the allosteric activation of AC by CaM.

EXPERIMENTAL PROCEDURES
MD Simulations-The wild-type AC⅐C-CaM complexes in the presence (ACwt_2Ca) and absence (ACwt_0Ca) of calcium were prepared as described previously (15), and the three modified systems (i.e. ACm1A, ACm2A, and ACm3A) were generated in the following manner. The last snapshot of the simulation WT2Ca_T1 (see Table 1 for the definition of this trajectory) was extracted, and the side chains of the modified residues were replaced by Ala side chains using the LEaP module from AMBER 10 (18) and the force field FF99SB (19). Sodium counterions were added to neutralize the system.
The five systems (i.e. ACwt_2Ca, ACwt_0Ca, ACm1A, ACm2A, and ACm3A) ( Table 1) were then hydrated in a box of TIP3P (20) water molecules using cut-off values of 10 or 12.5 Å and periodic boundary conditions. The calcium Lennard-Jones parameters of the Ca 2ϩ ions comprised a van der Waals radius (R) of 1.7131 Å and a well depth (⑀) of 0.459789 kcal/mol (21).
The MD trajectories were initiated and ran as described previously (15). For each system, two simulations were recorded, which were labeled using the system name and the strings "T1" and "T2" ( Table 1).
The formation of hydrogen bonds between protein acceptor and donor groups was monitored by selecting donor and acceptor pairs with a proximity of less than 3 Å every 10 recorded frames. This selection was implemented using the Biskit python library (22). In the following, the AC and C-CaM residues are labeled based on their numbers in the crystallographic struc-ture 1YRT (13), and the C-CaM residue numbers are followed by the letter a.
Along the MD trajectories, the geometric strain (23) of a residue i was calculated using the equation, where (d ij Ϫ ͗d ij ͘) 2 is the instantaneous deviation from the average of the distance between the ␣ carbons of residues i and j. It was shown previously (23) that hinge residues are characterized by large p(i) values.
Plasmids-Plasmid pTRAC384GK was used to express the wild-type AC (corresponding to the first 384 codons of CyaA, followed by the two residues Gly and Lys), as described previously (24). Plasmids pKTRACm2A and pKTRACm3A, which were used to express ACm2A and ACm3A, respectively, were constructed from plasmid pKTRAC, an expression vector for the CyaC and CyaA proteins of B. pertussis. This plasmid was obtained from plasmid pTRACG (25) by replacing the ampicillin resistance gene with a kanamycin-resistant gene from plasmid pKT25 (26). In pKTRAC, both the cyaC and cyaA genes are expressed under the control of the phage Pr promoter. The plasmid has a ColE1 replication origin, and it also expresses a thermosensitive repressor cI857, which strongly represses gene transcription at the Pr promoter at temperatures below 32°C. Thus, expression of the CyaC and CyaA proteins can be triggered by shifting the cells to 37-42°C.
Plasmid pKTRACm3A was constructed by subcloning between the unique AgeI and BamHI sites of pKTRAC a synthetic gene (synthesized by GeneArt, Invitrogen) that encodes CyaA residues 320 -384, followed by the two residues Gly and Lys, and where the three codons, Arg 338 , Asn 347 , and Asp 360 , were changed to Ala codons (the sequence of the ACm3A gene is available upon request).
Plasmid pKTRACm2A was constructed by subcloning between the unique SnaBI and BsiWI sites of plasmid pKTRACm3A using a double-stranded synthetic oligonucleotide produced via the hybridization of the following two oligonucleotides: 5Ј-GTTTTCTACGAAAACCGCGC-3Ј and 5Ј-GTACGCGCGGTTTTCGTAGAAAAC-3Ј (Eurofins MWG GmbH, Ebersberg, Germany), thereby restoring the wildtype Asn 347 codon (the SnaBI site was abolished after oligonucleotide insertion; data available upon request). The DNA sequences of the modified cyaAЈ genes in pKTRACm2A and pKTRACm3A were confirmed by DNA sequencing (Eurofins MWG GmbH, Ebersberg, Germany).
Plasmid pKT7AC contained a synthetic AC gene (synthesized by GeneArt, Invitrogen) that encoded the first 384 codons of CyaA, which was optimized for high expression in Escherichia coli, followed by the two residues Gly and Lys (the full DNA sequence is available upon request). This synthetic AC gene was cloned under the control of a T7 promoter and a synthetic RBS sequence in the vector pMK-RQ (GeneArt, Invitrogen), a plasmid with a ColE1 replication origin and a kanamycin resistance gene. Plasmid pKT7ACm1A was used to express ACm1A, and it was constructed by subcloning between the unique SnaBI and BsiWI sites of plasmid pKT7AC, a double-stranded synthetic oligonucleotide, which was obtained by hybridizing two oligonucleotides: 5Ј-GTGTTCTACGAAGCTCGCGC-3Ј and 5Ј-GTAC-GCGCGAGCTTCGTAGAACAC-3Ј (Eurofins MWG GmbH, Ebersberg, Germany), thereby replacing the wild-type Asn 347 codon with an Ala codon (sequence available upon request). The DNA sequence of the modified cyaAЈ gene in pKT7ACm1A was confirmed by DNA sequencing (Eurofins MWG GmbH, Ebersberg, Germany).
Purification of AC Proteins-The wild-type AC, ACm2A, and ACm3A were expressed in the E. coli BLR strain (Novagen, Darmstadt, Germany) by transformation with the plasmids pTRAC384GK, pKTRACm2A, and pKTRACm3A, respectively. The transformants were grown at 30°C in LB medium that contained either 100 g/ml ampicillin (for plasmid pTRAC384GK) or 50 g/ml kanamycin (for plasmids pKTRACm2A or pKTRACm3A). When the culture reached an optical density of 0.6 -0.8 at 600 nm, expression of the proteins was triggered by shifting the growth temperature to 42°C. After 150 min of additional growth at 42°C, the cells were collected by centrifugation (20 min, 10,000 ϫ g, 4°C), and the cell pellets were frozen at Ϫ20°C.
ACm1A was expressed in the E. coli KRX strain (Promega, Madison, WI), which was transformed with the plasmid pKT7ACm1A. The transformants were grown at 37°C in LB medium containing 50 g/ml kanamycin to an optical density of 0.6 -0.8. Next, expression of the proteins was triggered by the addition of 0.1% rhamnose (to induce the expression of the chromosomally encoded T7 RNA polymerase). After 180 min of additional growth at 37°C, the cells were collected by centrifugation, as described above.
The cell pellets were resuspended in 20 mM HEPES-Na, pH 7.5, and disrupted by sonication at 4°C. The sonicated suspension was centrifuged for 20 min at 13,000 ϫ g and 4°C. The supernatant was discarded, and the pellet was resuspended in 8 M urea with 20 mM HEPES-Na (pH 7.5) and agitated overnight at 4°C. After 20 min of centrifugation at 13,000 ϫ g at 4°C, the supernatant ("urea extract") that contained the solubilized AC proteins was collected.
The AC proteins were purified according to a previously described protocol using two sequential chromatographic treatments with DEAE-Sepharose (24). Briefly, the urea extract was first loaded onto a DEAE-Sepharose column (20 ml of packed resin) equilibrated in 8 M urea with 20 mM HEPES-Na (pH 7.5). In these conditions, the AC protein did not bind to the resin, and it was recovered in the flow-through fractions. The collected flow-through fractions were then diluted 5 times with 20 mM HEPES-Na (pH 7.5) and applied to a second DEAE-Sepharose column (20 ml of packed resin), which had been equilibrated in 20 mM HEPES-Na (pH 7.5). In these conditions, the AC proteins were retained on the resin and, after extensive washing with 20 mM HEPES-Na (pH 7.5), the proteins were eluted in a soluble form using 20 mM HEPES-Na (pH 7.5) containing 100 -200 mM NaCl. All of the AC protein preparations exceeded a purity of 95% according to SDS-PAGE analysis (Fig.  6A). The AC protein content was determined based on the absorption spectra using an extinction coefficient at 278 nm of 28,000 M Ϫ1 /cm Ϫ1 .
AC Enzymatic Activity Assays-The activity of AC was measured using a sensitive colorimetric assay, which was reported recently (27). AC converts ATP into cAMP and pyrophosphate (PP i ), and the latter can be hydrolyzed further by an exogenously added pyrophosphatase into two phosphate molecules (P i ), which can be quantified using a standard colorimetric assay based on the change in the absorbance of malachite green dye in the presence of phosphomolybdate complexes. The amount of P i produced was determined using the P i Color-Lock TM ALS colorimetric assay from Innova Biosciences (Cambridge, UK).
The AC enzymatic assays were performed using a 96-well microtiter plate at 30°C in a final volume of 50 l, which contained 50 mM Tris-HCl (pH 8.0), 10 mM MgCl 2 , 0.1 mM CaCl 2 , 0.5 mg/ml bovine serum albumin, 2 units/ml E. coli inorganic pyrophosphatase (Sigma-Aldrich; the amount of pyrophosphatase added to the reaction medium was found to be sufficient to ensure the immediate and complete conversion of the released PP i into P i ), 0 -1 M CaM, and 0.1-1 nM (depending on the specific activity) AC proteins (diluted from stock solutions in 10 mM Tris-HCl with 0.2% Tween 20, pH 8.0). The mixtures were preincubated for 10 min at 30°C. Then the enzymatic reactions were initiated by adding 2 mM ATP (final concentration), and the microplate was incubated at 30°C with agitation. At various incubation times (typically 2-10 min), 10-l samples were taken from each well and transferred into a second 96-well microtiter plate, where each well contained 100 l of a P i -ALS mixture made of 2 volumes of H 2 O plus 8 volumes of P i Color-Lock ALS reagent (provided in the P i -ALS kit from Innova Biosciences). The enzymatic reaction was stopped immediately by the acidic conditions of the P i -ALS mixture. After 10 -13 min of incubation at room temperature, 10 l of stabilizer solution (from the P i -ALS kit) was added to prevent further nonenzymatic breakdown of the phosphorylated substrate in acidic conditions (according to the kit instructions). After further incubation for 30 -60 min at room temperature, the optical density at 595 nm (A 595 ) was recorded using a microplate reader (Tecan, Lyon, France). A standard curve was obtained in parallel by adding known concentrations of P i to the P i -ALS mixture, which was used to convert the A 595 values into moles of PP i produced. The enzymatic activity was calculated based on the initial velocity of PP i synthesis (in the conditions described above, the accumulation of PP i was linear with time).
Biophysical Characterization of AC Proteins-Size exclusion chromatography (SEC) was performed using a Superdex 200 column (GE Healthcare), which was controlled by a GPCmax module that was connected online to a triple detector array (TDA), model 302 (Viscotek Ltd., Houston, UK), as described previously (28). The running buffer was buffer A (20 mM Hepes, 100 mM NaCl, pH 7.4), and the protein concentrations were 10 -20 M. Synchrotron radiation CD experiments were conducted at the SOLEIL synchrotron facility (DISCO beamline, Gif-sur-Yvette, France), as described previously (29). Briefly, the synchrotron radiation CD spectra were recorded at 25°C using an integration time of 1.2 s and a bandwidth of 1 nm, with a 1-nm resolution step. Each far-UV spectrum represented the average of at least three individual scans. Optical cells with a 26-m path length and CaF 2 windows (Hellma) were used for recording CD signals in the far-UV region (180 -260 nm). The protein concentrations were 25-50 M for AC proteins and 250 M for CaM in buffer A. Equimolar mixtures of enzyme and activator were used to obtain the AC-CaM and ACm3A-CaM spectra. The thermodynamic stability of the AC proteins was investigated by following their urea-induced denaturation (at 25°C, in buffer A), which was monitored by tryptophan fluorescence spectroscopy as described previously (30), using an FP-6200 spectrofluorimeter (Jasco, Japan) in a Peltier-thermostated cell holder, with a 1-cm path length quartz cell (101.QS, Hellma). The thermodynamic parameters (urea concentration required to unfold half the population of native proteins ([urea]1 ⁄ 2 ), the cooperativity (m), and free energy (⌬G)) were deduced from the fluorescence data, as described previously (30).

C-CaM/AC Interface in the Crystallographic Structures-To
explore the potential role of AC residues Arg 338 , Asn 347 , and Asp 360 in CaM-induced activation, as suggested previously (15), we first checked the conservation of the atomic contacts established by these residues in the crystallographic structures of the AC⅐C-CaM complex (13). The analysis of the hydrogen bonds and water-mediated contacts established between C-CaM and AC residues showed that the residues Arg 338 and Asp 360 were often involved in hydrogen bonds or in water-mediated contacts with C-CaM (data not shown; available on request). Furthermore, most of the contacts and hydrogen bonds between the LHL motif (residues 338 -360) (Fig. 1c) and other AC residues were present in more than three crystallographic structures (data not shown; available on request), thereby suggesting that the LHL motif plays an important role in the stabilization of the AC structure.
C-CaM/AC Interaction along MD Trajectories-Six independent MD simulations were recorded using the AC⅐C-CaM complex, where AC was modified in three different ways. Two simulations (i.e. "T1" and "T2") were performed for each AC variant, ACm1A (N347A), ACm2A (R338A/D360A), and ACm3A (R338A/N347A/D360A) (Fig. 1b). The aim of these simulations was to analyze the relative stability of the different ACm⅐C-CaM complexes and to relate these to the observations described previously based on the x-ray crystallographic structures.
The conformational drift of the AC⅐C-CaM complex was monitored during trajectories ACm1A_T1, ACm2A_T1, ACm3A_T1, ACm1A_T2, ACm2A_T2, and ACm3A_T2 (Fig.  2). The global RMSD calculated for the different AC⅐C-CaM complexes (Fig. 2, a and e) were similar for all variants, and they rapidly reached a plateau around 3 Å. The RMSD plateau was similar to that obtained based on the MD trajectories of the wild-type AC⅐C-CaM complex (Fig. 2a in Ref. 15). Thus, these modifications did not destabilize the overall AC structure at least during the time scales considered by the MD trajectories.
The conformational drifts (RMSD) of the different protein regions in the two sets of simulations (Figs. 2, b-d and f-h) were very similar for CB (orange curves); the C loop (yellow curves); the F, G, H, and HЈ ␣-helices; and the loop at the extremity of SA, which all had RMSD plateaus that were similar to or less than 2 Å. The CA region (green curves), SA region (violet curves), and C-terminal tail (cyan curves) had more significant drifts, where the RMSD values were larger than 4 Å. The CA and C-terminal drifts are not surprising because the modified residues were located in these regions. Interestingly, the increased internal mobility of the C-terminal tail is analogous to that observed in a wild-type AC⅐C-CaM complex (Fig. 3  (a and b) of Ref. 15) when the Ca 2ϩ ions were removed (Fig. 3 (a  and b) of Ref. 15). The latter had a lower affinity in vitro; thus, the modifications of residues Arg 338 , Asn 347 , and Asp 360 may have decreased the affinity of C-CaM for AC.
The water-mediated connections between AC and C-CaM were also analyzed along the MD trajectories (data not shown). The connections that involved the C-terminal tail of AC were present mainly in one or two trajectories, which shows that the interaction at this interface underwent reorganizations.
C-CaM and AC Fluctuations in the MD Trajectories-The behavior of C-CaM was monitored along the two series of mutant trajectories and compared with previous observations (15) obtained using the wild type AC⅐C-CaM complex. In the simulation of the wild-type AC⅐C-CaM complex in the absence of calcium ions (WT0Ca_T1) (Fig. 3a, solid red curve), the calcium loop of EF-hand 3 (i.e. between ␣-helices V and VI) and the C-terminal part of ␣-helix V exhibited highly increased fluctuations (RMSF). A similar increase was observed in ACm3A_T1 (Fig. 3b, solid red curve), but in ACm2A_T2 and ACm3A_T2 (Fig. 3b, dashed green and red curves), increased fluctuations were found in the calcium loop of EF-hand 4 (i.e. between ␣-helices VII and VIII). Interestingly, the increased fluctuations in the C-terminal part of ␣-helix V in ACm3A_T1 were associated with the small number of hydrogen bonds that connect C-CaM and the C-terminal tail (i.e. only a single hydrogen bond is present for over 80% of the trajectory length between Arg 90a and Glu 346 ).
Analysis of the internal fluctuations of AC along the various trajectories (data not shown) detected no major variations between the different modified proteins, which agreed with the similar global RMSD that was observed for all trajectories (Fig.  2). The fluctuations measured in the wild-type systems were similar to those observed in the modified proteins, where the largest differences were observed for the 226 -232 loop and the C-terminal tail of the protein. It should be noted that loop 226 -232 is not visible in the x-ray crystallographic structures, indicating that it is probably a highly flexible region.
The coordination of calcium ions by the C-CaM residues was stable in the EF-hand 3 along the trajectories of the modified proteins (data available on request), but there were some variations in the EF-hand 4. Indeed, the calcium coordination by Asp 133 in the EF-hand 4 was perturbed in the two series of simulations. This perturbation agrees with the increased fluctuation observed in the calcium loop of EF-hand 4 for ACm2A_T2 and ACm3A_T2 (Fig. 3b) and with the increase in the EF-hand 4 angle up to about 100°.
The C-CaM conformations were extracted from ACm2A_T2 and ACm3A_T2 (Fig. 4) at times when different distances were observed between Asp 133 -O␦1/O␦2 and the calcium ion. In both simulations, the side chain of Asp 133 moved apart, and at the end of the trajectory, the calcium ion was coordinated by the backbone carbonyl instead of the side chain. This conformational tendency of Asp 133 was increased in ACm3A_T2 compared with ACm2A_T2. Interestingly, in a steered MD study of the calcium dissociation in CaM (31), it was found that Asp 133 was among the latest residues to lose calcium coordination. The destabilization of the calcium/Asp 133 interaction observed in the present study corresponds to the destabilization of one of the strongest interactions that defines the calcium coordination in EF-hand 4. Thus, this destabilization indicates that the AC mutations decreased the calcium binding to CaM.
Overall, the perturbations of the EF-hand angles and calcium coordination are precursory indicators of the decrease in the C-CaM affinity for calcium ions (32,33). On longer time scales, this decrease could lead to the dissociation of calcium ions from C-CaM and to a decrease in the affinity of AC for C-CaM, as shown previously (14).
Network of Interactions Involving LHL along the MD Trajectories-The LHL motif, which contains the three modified residues, lies at the interface between C-CaM and the AC catalytic loop. The hydrogen bonds established by the residues in the LHL motif were analyzed in the different AC⅐C-CaM complexes (data available on request). Only long range hydro-gen bonds (that connected residues separated by more than 10 residues in the protein sequence) were considered in order to exclude interactions related to local secondary structures. Most of these hydrogen bonds were conserved along the MD trajectories, except for one hydrogen bond between Asn 35 and Tyr 342 , which was present only in the modified complexes, and two hydrogen bonds between Glu 346 and Leu 357 and between Tyr 350 and Glu 308 , which were observed in less than two simulations. A lesser number of stable hydrogen bonds that involved modified residues were observed in the ACm3A_T1 and ACm3A_T2 simulations. Indeed, Asn 347 established hydrogen bonds with the residues Glu 301 , Gln 302 , and Asn 304 , which are located in the catalytic loop. These hydrogen bonds were disrupted by the change from Asn 347 to Ala except for the hydrogen bonds that involved the Asn 347 backbone hydrogen. Similarly, the hydrogen bonds between Asp 360 and Arg 338 disappeared when Asp 360 or Arg 338 was modified to Ala. These observations show that the overall structure of the LHL motif is conserved in all systems and that the LHL connection to C-CaM on one side as well as to the C-loop on the other side were significantly weakened in the presence of the triple modification.
The hydrogen bonds that involved LHL residues as well as the protein fluctuations and the correlations in the fluctuation indicated that there was limited variation among the protein variants (data not shown). Thus, the possibility of long range communication through LHL was investigated using a parameter that was potentially more sensitive to small variations in geometry. The geometric strain, as defined by Chiappori et al. (23) (see "Experimental Procedures"), exhibits sufficient sensitivity because it depends on the geometry of the environment of the analyzed residue within a sphere of 7 Å, whereas it also follows the distance variations with respect to the average distance values.
The variations in the geometric strain along the MD trajectories exhibited correlated transitions between the given residues, which were due to simultaneous displacements of the environments of these residues. To summarize these variations, the time correlation matrices of the geometric strain were analyzed. The strain correlation peaks (in pink/yellow) correspond to residue pairs that are undergoing correlated displacements in their environment; thus, they are mechanically correlated. In order to estimate the reliability of the correlation predictions, the calculation was repeated five times by selecting at each time one data point over five along the analyzed time interval. The average values were calculated among LHL residues (Fig. 5) and among the LHL, C-loop, and calcium loop residues (data not shown; available on request). The S.D. values obtained from the repeated calculation of the correlations allowed us to estimate the error on the correlation, which was 4 -6.5% for correlations greater than 0.15. In all trajectories, the peaks were observed repeatedly between residues in the ␣-helix at 346 -355. A general attenuation of the correlations, with pixel colors changing from pink to yellow, was observed for the trajectories recorded in the protein variants compared with those recorded in the wild type. This suggests that there was a decrease in the long range correlation along the LHL for the modified proteins.

In Vitro Characterization of AC Variants That Harbored
Mutations at Position 338, 347, or 360-To further delineate the effect of the mutations on the activity of AC and its affinity for CaM, the three variants ACm1A, ACm2A, and ACm3A were characterized in vitro (see "Experimental Procedures"). The three modified proteins and the wild-type catalytic domain (AC) were overexpressed in E. coli and purified to homogeneity, as described previously (24) (see "Experimental Procedures"). The results of the SDS-PAGE analysis of the purified preparations are shown in Fig. 6A.
The enzymatic activities of each AC variant were determined at different CaM concentrations, which ranged from 0.03 nM to 1 M. Fig. 6B shows the CaM dependence of the enzymatic activity for each variant as a percentage of the maximal activity (100%) determined at the saturating CaM concentration of 1 M. The affinities of the AC enzymes for CaM, which were defined as the CaM concentration at half-maximal activation (K1 ⁄ 2 ), were determined (Fig. 6C) by curve fitting with a single binding isotherm, according to A ϭ A Max (C/C ϩ K1 ⁄ 2 ), where C is the concentration of CaM, A is the activity at the concentration C of CaM, and A Max is the maximal activity measured in the presence of 1 M CaM.
As shown in Fig. 6, B and C, wild-type AC (labeled AC WT in Fig. 6C) exhibited a k cat of 4600 s Ϫ1 (at saturating CaM and 2 mM ATP) and a half-maximal CaM activation (K1 ⁄ 2 ) of 0.11 nM, which agreed well with previous studies (13,24,34). The ACm1A variant, where the Asn 347 residue that interacted with the catalytic loop was modified to Ala, exhibited an enzymatic activity that was reduced by about half (k cat ϳ2250 s Ϫ1 ) compared with the wild type and a ϳ5-fold lower affinity for CaM (K1 ⁄ 2 CaM of 0.61 nM). The dual mutations R338A and D360A in the ACm2A variant did not affect the catalytic efficiency (k cat ϳ6000 s Ϫ1 was even slightly higher than that of the wild type), but the affinity for CaM was reduced by about 6-fold (K1 ⁄ 2 CaM of 0.66 nM), as might be expected with modifications at the CaM interface. Finally, the triple mutant ACm3A was the enzyme that was affected most significantly, where its catalytic turnover (k cat ϳ650 s Ϫ1 ) was decreased significantly compared with the wild-type AC (ϳ15% of the wild type turnover), and its affinity for CaM (K1 ⁄ 2 CaM of 26 nM) was reduced by more than 200 times. These data indicate that the separate single (N347A) and dual (R338A and D360A) modifications had limited effects on the catalytic and CaM-binding properties of AC, whereas their combination in a single enzyme (ACm3A) had strong synergistic effects on both the enzymatic activity and the affinity for CaM. This fully supports our prediction that the LHL region is critical for the allosteric coupling of CaM binding to the stabilization of the catalytic loop in an enzymatically active configuration. The 2-5-fold higher K m ATP levels of all of the modified AC enzymes compared with the wild-type AC also demonstrate that the LHL motif is important for maintaining the optimal conformation of the active site.
To verify that the triple modification of R338A, N347A, and D360A in ACm3A did not affect the structural integrity of the protein, we performed the following biophysical analyses. First, the proteins were analyzed by SEC followed by triple detector array (SEC-TDA) to determine whether the mutations affected their oligomeric status. SEC-TDA provides the molecular FIGURE 5. Correlation between geometric strains among the LHL residues, displayed for each recorded trajectory. The time correlation matrices were calculated between the variations of the geometric strains of LHL residues, and the first 10 ns were discarded from the analysis. In order to estimate the reliability of the prediction of correlation, the calculation was repeated five times by selecting, at each time, one data point over five along the analyzed time interval. The average values calculated for the LHL residues are shown in the figure. The intensity scale varies from 1.0 (pink) to Ϫ0.2 (dark green).
masses of the eluted species by combining static right angle light scattering with UV absorbance. The results showed that both the wild-type and ACm3A proteins were monomeric species in solution (Fig. 7A). The secondary structure contents of the CaM and AC proteins, either isolated or in complexes, were analyzed by synchrotron radiation CD in the far-UV range. The shapes of the far-UV CD spectra of wild-type AC and ACm3A were similar (Fig. 7B). The addition of CaM induced similar changes in the overall secondary structure content of both AC proteins. These results clearly indicate that the secondary structure content of ACm3A was not affected by the mutations. Finally, the thermodynamic stability of both AC proteins was characterized. The urea-induced denaturation of the AC proteins was monitored by tryptophan fluorescence. The denaturation profiles (Fig. 7C). and the thermodynamic parameters (Fig. 7D) obtained for both proteins were very similar. Overall, these experiments indicate that the three modifications in ACm3A did not affect the structural properties or the overall stability of the protein.

DISCUSSION
Maps of the energetic influences (16,17) that were calculated (15) previously using the AC⅐C-CaM complex allowed us to identify three residues (i.e. Arg 338 , Asn 347 , and Asp 360 ) that might affect the interaction between AC and C-CaM. In the present study, we characterized three modified proteins (i.e. ACm1A (N347A), ACm2A (R338A, D360A), and ACm3A (R338A, N347A, and D360A)), using in silico MD simulations and based on in vitro biochemical and biophysical studies.
Our in vitro studies showed that the affinity of the triple mutant ACm3A for CaM was greatly reduced, whereas that of the single (ACm1A) or dual mutant (ACm2A) was not affected significantly compared with that of the wild-type enzyme.
These results indicate that these three mutations had a strong synergistic effect on the CaM-binding affinity. A similar synergistic effect was noted for the catalytic efficiency of the enzyme because the enzymatic activity of ACm3A was also reduced significantly (about 15% of the wild-type enzyme activity), whereas that of ACm1A was about 50% of the wild-type activity. The lower activity of the ACm1A and ACm3A variants might be expected because the Asn 347 residue interacts directly with the catalytic loop. Thus, any perturbation of the interaction between Asn 347 and the catalytic loop might directly affect the catalytic efficiency of AC.
The dynamic behaviors of the three modified proteins in complex with C-CaM were also analyzed in silico based on two series of MD simulations. High variability in the establishment of hydrogen bonds in the complex was observed in the different modified proteins as well as among repeated trajectories. However, this variability could not be assigned to simulation artifacts because similar variability was observed among the four crystallographic structures of the complex AC⅐C-CaM (13). The analysis of the trajectories recorded using modified and wild-type AC⅐C-CaM complexes, which were conducted in parallel with the experimental characterization of the AC/CaM interaction, identified two different routes that allowed the affinity of AC for C-CaM to decrease. In the first simulation of the triple modified protein (ACm3A_T1), a destabilization of the calcium loop of EF-hand 3 was observed as well as a large decrease in the interactions between C-CaM and the CA region. In the second trajectory recorded using ACm3A (ACm3A_T2), a destabilization of the calcium coordination in the EF-hand 4 of C-CaM led to the disruption of the hydrogen bond between the Asp 133 side chain and calcium. This could be the first step in the possible dissociation of calcium from C-CaM. It is known that calcium-free CaM has a lower affinity for AC (14); thus, the destabilization of calcium coordination would result in a reduction of the CaM affinity for AC.
Guo et al. (13) described the crystallographic structure of the complex AC⅐C-CaM and characterized one AC modification: the double modification E346A/R348A in the C-terminal tail. The modified protein exhibited a reduced enzymatic activity as well as a reduced affinity for CaM. The two amino acids changed in the E346A/R348A mutant flanked the Asn 347 residue that was modified in the present study. The hydrogen bonds that involve Glu 346 and Arg 348 residues varied in the wild-type AC when calcium ions or C-CaM were removed from the system as well as in the trajectories of the modified AC protein, but no systematic variations were detected. However, FIGURE 7. A, SEC-TDA analysis of the oligomerization state of ACwt and ACm3A. Red curve, right angle light scattering; blue curve, UV absorbance; green curve, molecular masses of the eluted species. See "Experimental Procedures" for details. B, synchrotron radiation circular dichroism spectra of CaM and AC proteins in isolation or in a 1:1 complex. MRE, mean residual ellipticity. See "Experimental Procedures" for details. C, thermodynamic stability of AC proteins following urea-induced denaturation. The ratio of the fluorescence intensity at 360 and 320 nm (excitation wavelength ϭ 295 nm) for the proteins as a function of the urea concentration was measured as described under "Experimental Procedures." D, thermodynamics parameters of the urea-induced denaturation of AC proteins.
the observations of Guo et al. (13) based on the E346A/R348A AC variant agree well with our present study as well as supporting the hypothesis that the LHL motif is critical for linking CaM binding with the stabilization of the AC catalytic loop in a configuration that is favorable for catalysis. Indeed, the LHL motif is in direct contact with C-CaM on one side via the Arg 338 and Asp 360 residues as well as the catalytic loop on the other side via the Asn 347 residue. Thus, it is ideally positioned to relay information related to CaM-binding on the distant catalytic site of the AC enzyme via a dense network of atomic contacts, which primarily involve the Arg 338 , Asn 347 , and Asp 360 residues, as demonstrated in our MD simulations. Long range communications within the LHL motif were further suggested by our analysis of the geometric strain along the MD trajectories of the different ACs. Therefore, the LHL motif appears to be a critical module for the allosteric activation of AC by CaM. The analysis performed in the present study was more qualitative than the method proposed recently by Weinkam et al. (35,36), but we should note that the availability of only one structure of AC in complex with Ca 2ϩ -loaded C-CaM as well as the conformational equilibrium that is probably adopted by AC in the unbound state prevent a more straightforward application of this method.
From a more general perspective, the detection of the residue networks that are responsible for long range communication inside a given protein has been attracting much interest during the last decade (37)(38)(39)(40)(41). These networks are thought to be involved in biomolecular interactions (42) with fundamental roles in many biological processes (43). Homologous protein sequence alignment (44 -46) or analysis of the covariance of the NMR chemical shifts (47) has predicted long range protein communication networks that agree with experimental observations. Some of these analyses have facilitated protein engineering (48). In the present study, we combined MD and mutagenesis to identify a long range communication network related to the activation of a key virulence factor derived from a bacterial pathogen. This investigation may facilitate new methods for interfering with the AC/CaM interaction, thereby allowing a novel drug discovery approach (49).