An Extended Conformation of Calmodulin Induces Interactions between the Structural Domains of Adenylyl Cyclase from Bacillus anthracis to Promote Catalysis*

The edema factor exotoxin produced byBacillus anthracis is an adenylyl cyclase that is activated by calmodulin (CaM) at resting state calcium concentrations in infected cells. A C-terminal 60-kDa fragment corresponding to the catalytic domain of edema factor (EF3) was cloned, overexpressed inEscherichia coli, and purified. The N-terminal 43-kDa domain (EF3-N) of EF3, the sole domain of edema factor homologous to adenylyl cyclases from Bordetella pertussis andPseudomonas aeruginosa, is highly resistant to protease digestion. The C-terminal 160-amino acid domain (EF3-C) of EF3 is sensitive to proteolysis in the absence of CaM. The addition of CaM protects EF3-C from being digested by proteases. EF3-N and EF3-C were expressed separately, and both fragments were required to reconstitute full CaM-sensitive enzyme activity. Fluorescence resonance energy transfer experiments using a double-labeled CaM molecule were performed and indicated that CaM adopts an extended conformation upon binding to EF3. This contrasts sharply with the compact conformation adopted by CaM upon binding myosin light chain kinase and CaM-dependent protein kinase type II. Mutations in each of the four calcium binding sites of CaM were examined for their effect on EF3 activation. Sites 3 and 4 were found critical for the activation, and neither the N- nor the C-terminal domain of CaM alone was capable of activating EF3. A genetic screen probing loss-of-function mutations of EF3 and site-directed mutations based on the homology of the edema factor family revealed a conserved pair of aspartate residues and an arginine that are important for catalysis. Similar residues are essential for di-metal-mediated catalysis in mammalian adenylyl cyclases and a family of DNA polymerases and nucleotidyltransferases. This suggests that edema factor may utilize a similar catalytic mechanism.

cAMP is a key second messenger that modulates a particularly diverse set of physiological responses including sugar and lipid metabolism, cell differentiation, apoptosis, neuronal activity, and ion homeostasis. Certain pathogenic bacteria have evolved exotoxins that severely alter the internal cAMP levels of infected cells. These exotoxins work through two common mechanisms. The first is to ADP-ribosylate the ␣-subunits of heterotrimeric G-proteins. 1 Cholera and pertussis toxin ADPribosylate G s and G i , respectively. In doing so the toxins increase the activity of endogenous adenylyl cyclases to pathogenic levels (1). The second strategy is to secrete an adenylyl cyclase directly into the host. This strategy is employed by Bordetella pertussis, Bacillus anthracis, and Pseudomonas aeruginosa. To overcome the problem of unregulated cAMP production within the pathogen, these cyclases have evolved to become active only upon binding a factor endogenous to the host (2,3). Adenylyl cyclases from B. pertussis and B. anthracis are activated by calmodulin (CaM). The exotoxin from P. aeruginosa is activated by an unknown, soluble cellular factor (4). The adenylyl cyclase from B. anthracis, also known as edema factor, is secreted as an inactive enzyme by bacterial cells and associates with protective antigen, another B. anthracis exotoxin. Protective antigen is a pH-dependent protein transporter that upon proteolytic activation, endocytosis, and pH change, translocates edema factor into the cytoplasm of the target cells (5). After gaining access to the cytoplasm, edema factor binds CaM at the resting calcium concentration and is thus activated to produce unregulated levels of cAMP within the cytosol (6,7). The rate of ATP conversion to cAMP by edema factor is two to three orders of magnitude greater than membrane-bound eukaryotic adenylyl cyclases (8,9). This catalytic activity is thought to be involved in both cutaneous and systemic anthrax infection (10). In the course of anthrax pathogenesis, edema factor works in concert with another enzyme exotoxin, lethal factor. Lethal factor has been shown to proteolytically inactivate ERK kinase (MAP-kinase kinase) (11).
The adenylyl cyclase from B. anthracis is a 92.5-kDa soluble protein and can be divided into N-and C-terminal functional * This research was supported in part by National Institutes of Health Grants GM53459, DA05778 (to C. D.), and AR41637 (to Z. G.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 1 The abbreviations used are: G-protein, guanine nucleotide-binding regulatory protein; CaM, calmodulin; EF, edema factor; EF3, C-terminal 60-kDa edema factor; EF3-N, N-terminal 43-kDa of EF3 (aa 291-640); EF3-C, C-terminal 19-kDa EF3 (aa 640 -800); MLCK, myosin light chain kinase; M13, synthetic peptide corresponding to the CaMbinding region of smooth muscle MLCK; G s␣ , the ␣-subunit of G-protein that stimulates adenylyl cyclase; AC, adenylyl cyclase; PAGE, polyacrylamide gel electrophoresis; 1,5-IAEDANS, N-iodoacetyl-NЈ-(5-sulfo-1-naphthyl)ethylenediamine (AEDANS when attached to a protein); domains (10). The N-terminal 30-kDa domain bears strong homology to the N-terminal domain of lethal factor and is centrally involved in edema factor association with protective antigen and entrance into host cells. The C-terminal 60-kDa domain consists of CaM-sensitive adenylyl cyclase activity. The N-terminal portion of this 60-kDa region is similar to the adenylyl cyclase exotoxins from B. pertussis and P. aeruginosa, with 34 and 29% sequence identity, respectively. However, this region has no significant similarity to the adenylyl cyclases from enterobacteria and eukaryotes. In contrast to mammalian adenylyl cyclases that contain a catalytic core composed of two homologous domains, edema factor and its related exotoxins have no internal sequence repeat and are active as monomers (12).
CaM, a 16.5-kDa soluble protein, is a four-site member of the EF-hand family of Ca 2ϩ -binding proteins (13). CaM is present in all eukaryotic cells, and it binds to and modulates, in a Ca 2ϩ -dependent manner, a number of enzymes including myosin light chain kinase, cyclic nucleotide phosphodiesterase, calcineurin, phosphorylase kinase, and adenylyl cyclase (14). The molecular structures of Ca 2ϩ -bound and Ca 2ϩ -free CaM have been determined by x-ray crystallography and NMR, respectively (15)(16)(17)(18)(19). There are two globular domains in CaM, each containing two Ca 2ϩ -binding EF-hand motifs. The two globular domains are connected by a continuous single ␣-helix that gives CaM its characteristic dumbbell shape. Binding of calcium ions has been shown to induce the movement of a pair of helices in a two EF-hand domain. This leads to an opening of the domain and an exposure of a hydrophobic pocket, which becomes a target interaction site. Blocking this opening by introducing a disulfide bond in either domain reversibly inactivates CaM (20).
CaM binding sites have been identified in a number of enzymes. With a few exceptions they are amphipathic helices capable of interacting with both domains of CaM in a 1:1 complex. The structures of the Ca 2ϩ /CaM complexes with the CaM-binding domain of MLCK and CaM-dependent kinase type II have been solved (21)(22)(23). In both structures, the two globular domains of CaM (residues 6 -73 and 83-146) remain essentially unchanged upon complex formation. However, the long central helix (residues 65-93) of CaM, which connects the two domains in the crystal structure of calcium-bound CaM, is disrupted. This enables the C-and N-terminal domains to clamp the helical peptide, bringing the two domains of CaM close to each other.
In this study, we use biochemical, genetic, and spectroscopic methods to analyze the regulation of edema factor by CaM and gain insight into its catalytic mechanism. Our results show that whereas the mode of activation for edema factor is distinctly different from that seen in mammalian adenylyl cyclases, these two classes of enzymes are likely to share a common catalytic mechanism.
Expression of Edema Factor and CaM-To construct the plasmid for the expression of EF-cya62 (aa 262-800) and EF3 (aa 291-800) (pPro-ExH6-EF-cya62 and pProExH6-EF3, respectively), restriction site NotI was first introduced to the appropriate codons by site-directed mutagenesis (QuikChange kit, Stratagene). The resulting plasmid was then digested with NotI and KpnI, and the fragment that encoded EF-cya62 or EF3 was isolated and ligated to pProEx-H6, which was digested with the same enzymes to express a recombinant edema factor, hexahistidine-tagged at its N terminus. A similar strategy was employed to generate pProEx-CH6-EF3 that had a C-terminal hexahistidine tag. The plasmid for the expression of EF3-N (aa 291-640) was generated by introducing a termination codon at residue 641 using pProExH6-cya62 as a template. The plasmid for the expression of EF3-C (aa 641-800) was made by first introducing a NotI site at the codons for aa 638 -640 of pProExCH6-EF3 and subsequent removal of the fragment for the coding region (aa 291 to 640).
To express the truncated forms of edema factor, the appropriate plasmids were transformed into Escherichia coli BL21(DE3) cells that harbored pUBS520, a plasmid that encodes tRNA for the AGA and AGG codons. The resulting cells were grown in a modified T7 medium with 50 g/ml ampicillin and 25 g/ml kanamycin at 25-30°C to A 600 ϭ 0.4, induced by adding isopropyl-1-thiogalactopyranoside to a final concentration of 100 M and harvested 12-19-h postinduction. All the subsequent procedures were performed in a 4°C cold room. E. coli cells were lysed in T 20 ␤ 5 P 0.1 buffer (20 mM Tris-HCl, pH 8.0, at 4°C; 5 mM ␤-mercaptoethanol; 0.1 mM phenylmethylsulfonyl fluoride) with added lysozyme to 0.1 mg/ml final concentration and a 4-min sonication with a 1-s on and 3-s off cycle. The supernatant of E. coli lysate was applied directly onto a Ni 2ϩ -NTA column that was equilibrated with T 20 ␤ 5 P 0.1 N 100 (T 20 ␤ 5 P 0.1 containing 100 mM NaCl). The Ni 2ϩ -NTA column was washed with T 20 ␤ 5 P 0.1 N 500 and then T 20 ␤ 5 P 0.1 N 100 I 20 (T 20 ␤ 5 P 0.1 N 100 containing 20 mM imidazole, pH 7.0. The column was then eluted with the high imidazole buffer, T 20 ␤ 5 P 0.1 N 100 I 150 . The eluate was diluted 5-fold with M 20 E 1 D 1 P 0.1 (30 mM Mes, pH 6.5, 1 mM EDTA, 1 mM dithiothreitol, 0.1 mM phenylmethylsulfonyl fluoride) and applied to an Amersham Pharmacia Biotech FPLC SP-Sepharose column that had been equilibrated with M 20 E 1 D 1 P 0.1 , pH 6.5. The column was washed with one-column volume of M 20 E 1 D 1 P 0.1 , pH 6.5, and absorbed protein was eluted at 2 ml/min with a linear NaCl gradient (500 -550 mM, 350 -400 mM, and 450 -500 mM NaCl for purification of EF3, EF3-N, and EF3-C, respectively) in M 20 E 1 D 1 P 0.1 . The protein peaks of edema factor fragments were analysed for molecular size on SDS-PAGE and for EF3 by adenylyl cyclase activity. Purified EF3, EF3-N, and EF3-C were concentrated in a Centricon 10 microconcentrator (Amicon) and stored at Ϫ80°C (protein concentration Ͼ5 mg/ml).
An alternative scheme was also used to purify EF3 based on the fact that little binding of EF3 to a DEAE column at pH 7.0 was observed. In this case, E. coli cells were lysed in M 30 E 4 N 25 P 0.1 , pH 7.0, as described above. Lysate was passed first through a DEAE column and then directly onto an Amersham Pharmacia Biotech FPLC SP-Sepharose column, and columns were washed with three volumes of M 30 E 4 P 0.1 , pH 7.0. EF3 protein was then eluted off with a NaCl gradient in M 30 P 0.1 as the major protein peak and identified by SDS-PAGE. The fractions containing EF3 were pooled and loaded onto a Ni 2ϩ -NTA column. The Ni 2ϩ -NTA column was washed with T 20 ␤ 5 P 0.1 N 100 I 20 and eluted with T 20 ␤ 5 P 0.1 N 100 I 150 . Purified EF3 protein was concentrated as described above. This procedure modestly increased purification of the protein.
Proteolytic Analysis of Edema Factor-Purified EF3 and CaM were exchanged into digest solution (10% glycerol, 50 mM Tris-HCl, pH 7.5, 1 mM MgCl 2 , 1 mM CaCl 2 ) by dialysis and two protein stock solutions were prepared for digestion. One contained only EF3, the other a 1:2 molar mixture of EF3 and CaM. The concentration of EF3 (0.5 mg/ml) was identical in both samples. Nine different protease stock solutions were prepared in the same digest solution. Digests with thermolysin, endoproteinase Lys-C, endoproteinase Glu-C (Roche Molecular Biochemicals), elastase, chymotrypsin, trypsin, protease type X, endoproteinase Arg-C (Sigma) and thrombin (heme Tech) were performed at two different protease concentrations, one 10-fold more dilute that the other. The protease concentrations were 1:100 and 1:1000 (w/w) for all proteases except for endoproteinase Lys-C, which was used at 0.03 and 0.003 units of protease/mg of protein. All digestions were held at 4°C. Samples were taken after 10 min and 1, 4, and 14 h. All samples were transferred directly from the digestion mixture into hot SDS loading buffer and subjected to SDS-PAGE.
Adenylyl Cyclase Assay-Activity was assayed for 5-20 min at 30°C in the presence of 10 mM MgCl 2 (9). Full-length human and fruit fly CaM was expressed in bacteria and purified using a previously published protocol and yielded ϳ5-10 mg of purified protein per liter of culture (25,26). When applicable, recombinant EF3 was premixed with CaM on ice for 5 min before the assay, and assays were done in the presence of 50 M CaCl 2 .
Genetic Screen for Loss-of-function EF3 Mutants-This screen is based on the observation that overexpression of EF3 and CaM significantly reduces the viability of E. coli. The coding region encoding fruit fly CaM was generated by polymerase chain reaction using Vent DNA polymerase and cloned into pBB131, resulting in pBB131-dCaM. A library of plasmids that had mutations in the coding region of EF3 were constructed by propagating plasmid pProExH6-EF3 in E. coli mutator strain, XL-1 red (mutD5, mutS, mutT). Plasmid DNAs were isolated and transformed to E. coli ⌬cya strain, TP2000 that harbored pBB131-dCaM[TP2000(dCaM)]. The transformants were plated directly onto Luria broth plate containing 100 M isopropyl-1-thiogalactopyranoside, 50 g/ml ampicillin, and 25 g/ml kanamycin). Plasmid DNAs from the resulting colonies were isolated and retransformed into TP2000(dCaM) to confirm the viability. The coding region of EF3 was then sequenced.
Measurements of Fluorescence Resonance Energy Transfer-Steadystate fluorescence emission spectra were recorded on a SPEX Fluorolog 2/2/2 photon-counting spectrofluorometer (Edison, NJ). Fluorescence lifetime measurements were carried out by the method of time-correlated single photon-counting on a modified ORTEC 9200 nanosecond fluorometer (27)(28)(29)(30)(31). The fluorescence donor was 1, 5-IAEDANS and the acceptor was DAB. The CaM mutant having cysteine at amino acid residues 34 and 110 was constructed and labeled randomly with donor and acceptor or with donor alone as described. 2

Expression and Characterization of the Catalytic Domain of
Edema Factor-We have developed a novel purification scheme to facilitate large-scale production of the catalytic domain of edema factor. Our approach began with modifications of the edema factor gene that replaced the binding region of protective antigen with a hexahistidine tag at the N terminus. The expression of the truncated protein was monitored both by adenylyl cyclase assay and by immunoblot with anti-H5 antisera. We first constructed and analyzed the truncation mutant, EF-cya62 (aa 262-800) as previously reported (8). As expected, it had high adenylyl cyclase activity when stimulated by Ca 2ϩ / CaM. A progressive deletion of residues from the N and C termini of EF-cya62 was performed to search for an optimal construct for the expression of the functional, CaM-sensitive adenylyl cyclase. Deletion of 290 amino acids from the N terminus significantly improved the expression of a functional protein, and the product could be clearly visualized after the whole-cell lysate was run on a gel under the optimum induction conditions (100 M isopropyl-1-thiogalactopyranoside over 12 h in enriched media at 25-30°C). This construct is designated "EF3" (aa 291-800). Further deletion from either the N-or C-terminal end progressively reduced both protein expression and catalytic activity.
EF3 can be effectively purified using Ni 2ϩ -NTA and SP-Sepharose, resulting in greater than 95% pure protein as judged by SDS-PAGE. This preparation yields approximately 42 mg/liter E. coli culture, roughly 10-fold better than that from cells expressing EF-cya62. The enzyme activity of EF3 is similar to that of EF-cya62. Both exhibit relatively low activity without CaM (0.5-1 mol/mg/min) and can be stimulated by CaM by about 1000-fold. Because of its relatively basic pI (8.9), EF3 does not bind to the DEAE column at pH 7.0. Injecting the cell lysate through the DEAE column and then purifying it on SP-Sepharose and Ni 2ϩ -NTA columns, produced a slightly purer EF3 (ϳ97% pure) preparation. A C-terminal-tagged EF3 was also constructed and purified. Its yield was lower, 12 mg/liter culture, though the proteolytic contamination was mildly reduced. We analyzed the molecular weight of native EF3 by Superdex 200 gel permeation column and found that EF3 migrated as a 60-kDa protein. This is consistent with the notion that EF3 is a monomer.
Structural Insights from Proteolytic Digestion Experiments-We have probed the structural stability of EF3 with digestion by nine different proteases (thermolysin, endoproteinase Lys-C, elastase, chymotrypsin, trypsin, protease type X, endoproteinase Glu-C, endoproteinase Arg-C, and thrombin) in the presence or absence of Ca 2ϩ /CaM. The first six of these proteases showed a very similar digestion pattern (Fig. 1). In the presence of CaM (Fig. 1, bottom), little digestion of EF3 was observed. In the absence of CaM (Fig. 1, top), all six enzymes digested EF3 producing a 43-kDa fragment. Using N-terminal sequencing and mass spectrometry we found that the most stable proteolytic fragment (that generated by endoproteinase Lys-C) is a product of C-terminal truncation of 160 amino acids. These results suggest that the N-terminal 43-kDa fragment of EF3 constitutes an independent structural domain and that the 160-residue C-terminal of EF3 is structurally disordered in the absence of CaM.
To address the question of whether the catalytic domain of edema factor can indeed be further divided into two subdomains, we constructed and purified EF3-N (aa 291-640) and EF3-C (aa 641-800) ( Fig. 2A). The 43-kDa EF3-N was expressed and purified with a yield of 2 mg/liter. It had only one-tenth of the basal activity of the EF3 fragment and was only modestly activated by CaM (about 10 -30-fold compared with more than 1000-fold stimulation for the EF3 fragment, Fig. 2B). The 19-kDa EF3-C fragment was also expressed and purified with a yield of only 0.5 mg/liter and a minor contamination of a 25-kDa protein ( Fig. 2A). By itself, EF3-C had no detectable catalytic activity (data not shown); however, with addition to EF3-N, EF3-C fully restored the CaM activation of EF3-N (Fig. 2B). Taken together with the proteolysis data, these results indicate that the 43-kDa EF3-N fragment indeed comprises the catalytic core of the edema factor. This domain appears to be structurally independent and capable of being regulated by Ca 2ϩ /CaM. It appears that the C-terminal 160-residue long EF3-C fragment has little structure on its own, but greatly enhances the Ca 2ϩ /CaM-dependent activation of EF3-N. In the presence of Ca 2ϩ /CaM, both segments of EF3 form a structurally stable catalytic-active entity. We further infer from these data that both EF3-N and EF3-C contain CaM binding sites.
Global Conformation of CaM in Complex with EF3-To determine the conformation of CaM in the complex with EF3, we employed a fluorescence energy transfer technique. With CaM34/110, a two-cysteine CaM mutant in which Cys residues had been substituted for Thr-34 and Thr-110 in the N-and C-terminal domain, respectively. 2 The C␣ atoms of Thr-34 and Thr-110 are 54 Å apart in the crystal structure of Ca 2ϩ -bound 2 P. O'Hara, Y. Mabuchi, and Z. Grabarek, submitted for publication. CaM and ϳ16 Å apart in CaM complexes with the synthetic peptides corresponding to the CaM-binding sites of myosin light chain kinase (M13) and CaM-kinase II. Thus, when labeled with a resonance energy transfer pair of probes, CaM34/ 110 is an extremely sensitive indicator of the global conformation of CaM. 2 CaM34/110 was randomly labeled with 1,5-IAEDANS as a donor and DAB as an acceptor. For the control experiments, the protein was labeled with the donor alone. The fluorescence signal and the lifetime of IAEDANS alone attached to CaM34/110 are virtually insensitive to Ca 2ϩ and to the addition of EF3 or M13 (Figs. 3 and 4), indicating that the immediate environment of the probes attached to CaM34/110 is not affected. Addition of Ca 2ϩ to CaM34/110-IAEDANS/DAB causes a small decrease in fluorescence, which apparently reflects an increased dynamic of CaM and consequently a decrease in the average distance between the two domains of CaM in solution. An addition of M13 causes a large decrease in fluorescence intensity and a decrease in the lifetime to 3 ns, consistent with the compact conformation of the CaM⅐M13 complex (Fig. 3). 2 In contrast to the M13 peptide, EF3 did not cause any decrease in fluorescence of CaM34/110-AEDANS, DAB. On the contrary, an addition of EF3 causes a reversal of the Ca 2ϩ effect, and the fluorescence intensity of the complex exceeds that of CaM34/110-AEDANS, DAB alone in the absence of Ca 2ϩ (Figs. 3 and 4). Because the CaM34/110 labeled only with the donor showed little change in fluorescence upon EF3 binding, we conclude that the change observed for the donor/acceptor-labeled CaM34/110 results from an increase in the distance between the N-and C-terminal domains of CaM upon interaction with EF3. The fluorescence decay of CaM34/ 110-AEDANS, DAB in the presence of CaCl 2 could be fitted with two lifetimes ( 1 ϭ 10.8 ns and 2 ϭ 6.9 ns) with relative contributions of 0.63 and 0.34, respectively, consistent with the dynamic nature of CaM in solution. 2 In contrast, the fluorescence decay of EF3 complexes with singly and doubly labeled CaM34/110 were strictly monoexponential with lifetimes of 12.9 ns and 10.3 ns, respectively (Fig. 4). The distance between the probes calculated from these values assuming the orientation factor 2 ϭ 2/3 was 49.6 Å. We conclude that CaM adopts an extended conformation in its complex with EF3, approximately similar to the conformation in the crystal structure of the Ca 2ϩ -bound free CaM.
Equilibrium Binding Constant for the CaM34/110-AE-DANS, DAB Interaction with EF3-We have utilized the change in fluorescence signal of CaM34/110, AEDANS, DAB upon binding EF3 to estimate the apparent equilibrium binding constant for the complex formation. Fig. 5 shows a fluorescence titration of CaM34/110, AEDANS, DAB with a concentrated solution of EF3 in the presence of Ca 2ϩ . Fitting of the data gave K d ϭ 2.9 ϫ 10 Ϫ8 , which is in excellent agreement with the K d ϭ 2.3 ϫ 10 Ϫ8 value obtained by Labruyẽre et al. (8) from the EC 50 of the EF-cya62 fragment activation by bovine brain calmodulin. Thus, neither the mutations introduced into CaM34/ 110, nor the attached spectroscopic probes affect in any measurable way the interaction between CaM and edema factor.
Activation of EF3 by Mutant Forms of CaM-Two sets of CaM mutants (Q and K series) that have single point mutations at each of four Ca 2ϩ -binding sites have been used to probe the individual contributions of the four calcium binding sites in modulating the activity of various enzymes regulated by CaM (32). The conserved glutamate at position 12 of the Ca 2ϩbinding site was mutated to either glutamine (B1Q to B4Q) or lysine (B1K to B4K). In isolated CaM, each mutation effectively eliminates Ca 2ϩ -binding of the affected site and the associated Ca 2ϩ -induced conformational changes (25,33,34). We found that mutations at either site 3 or 4 drastically reduced the ability of CaM to activate the EF3 fragment of edema factor whereas mutations at either site 1 or 2 had little effect (Fig. 6). Thus we hypothesize that calcium binding at the CaM C-terminal domain plays a particularly important role in edema factor activation. This is a very different pattern than that seen in activity assays using the same mutants to stimulate smooth and skeletal muscle MLCKs and type I mammalian adenylyl cyclase (32). The order of potency for a mutation within a given calciumbinding site to decrease the ability of CaM to activate the two MLCKs or AC1, followed the series 4,2,3,1 in decreasing potency.
In view of the above results we tested the possibility that the C-terminal domain of CaM is not only necessary but also sufficient for the activation of EF3. We have expressed the N-and C-terminal domains of CaM that contained only Ca 2ϩ binding sites 1/2 and sites 3/4, respectively, and tested their ability to activate EF3. We observed no activation of EF3 by the individual CaM domains (data not shown). This suggests that a contribution from both N-and C-terminal domains of CaM are required for the activation of EF3.
Mutational Analysis to Probe the Residues That are Crucial for Catalysis of EF3-Mutational analysis of edema factor and related proteins has indicated several amino acid residues that are important in catalysis. The results are summarized in Fig.  7. Single mutations at lysine 346 and 353 of edema factor and one of the homologous residue (Lys-65) of B. pertussis adenylyl cyclase significantly reduced catalytic activity in both enzymes (35)(36)(37). These two residues are hypothesized to bind nucleotide based on their fit to the consensus sequence GXXX(G/A)KS (fit in both forward and reverse directions) found generally in nucleotide triphosphate-binding proteins (37). Histidine 63 of B. pertussis AC (351 of edema factor) is proposed to serve as an essential acid/base catalyst (38). However, in the homologous adenylyl cyclase from P. aeruginosa, this residue is lysine, casting doubt on this hypothesis. Single point mutations at the two conserved aspartates of both B. pertussis and P. aeruginosa AC (residues homologous to 491 and 493 of edema factor) reduced catalytic activity over 100-fold and therefore are also possibly involved in nucleotide binding (4,39).
In an attempt to coexpress the edema factor and CaM simultaneously in E. coli, we uncovered findings that allowed us to develop a genetic screen for residues critical for catalysis by the edema factor. The expression of edema factor alone did not complement the defect of cAMP in E. coli. Interestingly, we found that it was extremely difficult to transform the plasmid for the expression of edema factor into the cells that already harbor the compatible plasmid for CaM expression. We hypothesized that the expression of edema factor and CaM might be toxic because of cAMP overproduction. Similar toxicity was observed when adenylyl cyclase from E. coli was overexpressed (40). To test this hypothesis, we passed the plasmid that harbors the gene for the expression of edema factor through the mutator strain, XL-1 red (mutD5, mutS, mutT) to introduce random mutations. After the pool of the mutated plasmids was transformed into E. coli ⌬cya, a small number of colonies could be recovered. Two of such colonies were characterized further. Although they expressed normally, based on immunoblot analysis, the adenylyl cyclase activities were significantly reduced. Sequence analysis showed that both carried a single mutation resulting in the change of the highly conserved amino acid residue 491 from aspartate to asparagine (D491N) (Fig. 7).
The conserved aspartate pair is involved in "two-metal" mediated catalysis of nucleotidyltransferases, including DNA polymerases and mammalian adenylyl cyclases (41,42). Edema factor family has three conserved negatively charged residues, Glu-443, Asp-491, and Asp-493, and we made two additional mutants, E443Q and D493N to analyze the role of these three negatively charged residues (Fig. 7). An arginine in mammalian adenylyl cyclase is implicated in stabilizing the pentaphosphate intermediate (42,43). We thus made and analyzed a mutant that had a mutation at one of the two conserved arginine residues, R329M (Fig. 7). All four mutants, R329M, E443Q, D491N, and D493N were purified to homogeneity with the protein yields similar to wild-type EF3. All Where F o is the initial fluorescence, ⌬F is the maximal change in fluorescence, and is the fraction of the CaM34/110-AEDANS, DAB bound to EF3. is a root of the quadratic equation: nA 2 Ϫ (K d ϩ nA ϩ B) ϩ B ϭ 0; where A and B are the total concentrations of CaM34/110-AEDANS, DAB and EF3, respectively; n is the apparent stoichiometry and K d is the apparent equilibrium dissociation constant. The parameters of the fit of the above titration data were that K d ϭ 2.9 ϫ 10 Ϫ8 and n ϭ 0.85. Note that non-integer values for the fitted stoichiometry (n) result mainly from imprecise estimation of protein concentration. mutants had relatively normal EC 50 values for CaM activation (Fig. 8). All four mutants also had relatively normal K m for ATP (0.5 Ϯ 0.1, 0.2 Ϯ 0.02, 0.3 Ϯ 0.1, 0.22 Ϯ 0.07, and 0.27 Ϯ 0.03 mM for wild type, R329M, E443Q, D491N, and D493N, respectively). However, D491N, D493N, and R329M had significantly reduced catalytic activity (10,000-, 5,000-, and 300-fold in V max values of CaM activation, respectively) whereas E443Q only had about a 10-fold reduction (Fig. 8). These results suggest that Asp-491, Asp-493, and Arg-329 residues are important for catalysis. DISCUSSION We have combined biochemical, spectroscopic, and mutational approaches to analyze the mechanism of edema factor activation by CaM. Based on the above results, we have developed a hypothetical "low resolution" model. Our model is derived from the following observed constraints. (i) The 60-kDa monomeric EF3 fragment of edema factor consists of a highly structured core domain and of a loosely structured, approximately 160-residue long, C-terminal domain. (ii) The C-terminal region acquires a well defined structure (becomes resistant to proteolysis) upon CaM binding to EF3. (iii) The C-terminal region is essential for full CaM activation of edema factor. (iv) CaM has an extended conformation while bound to EF3. (v) Whereas both domains of CaM are required for the activation of edema factor, calcium binding in the C-terminal domain plays a privileged role in activation of edema factor when compared with calcium binding in the N-terminal domain. We infer from these empirical constraints that the CaM-dependent activation of edema factor involves a change in relative position of a part of or the entire C-terminal domain with respect to the Nterminal core domain. This type of reorientation could result in the stabilization of a catalytically active state of the N-terminal region or even directly contribute residues to aid in catalysis.
Because the functional unit of EF3 is a monomer and be- cause of the expected large conformational change in the Cterminal domain of EF3 induced by CaM, we predict that the mechanism of activation of EF3 by CaM is very different from that seen in mammalian adenylyl cyclases. Mammalian membrane-bound adenylyl cyclases (AC1-AC9) have two homologous cyclase domains (C1 and C2) punctuated by two transmembrane domains (M1 and M2) (12,44). They are diversely regulated by intracellular proteins, i.e. G-proteins, protein kinases, and CaM. Molecular structures of a catalytically inactive C2a dimer and catalytically active C1a⅐C2a⅐G s␣ complex have recently been determined (44,45). The catalytic site occurs at the interface of the two homologous cyclase domains and includes residues from both regions. The structural mode of activation of adenylyl cyclase does not involve a major conformational change in either C1a or C2a domains upon binding of the stimulatory G-protein. Instead, catalysis is activated by an induced juxtaposition of these two domains to properly form the catalytic cleft at their interface. With the lack of any internal homologous repeat within EF3 and a different potency order for the importance of calcium binding sites in CaM, our model predicts that the structural events leading to EF3 activation by CaM will have little relation to those responsible for mammalian adenylyl cyclase activation by CaM.
The homologous internally repeated cyclase domains of mammalian adenylyl cyclase contain substructures that resemble the palm domains of the polymerase I family of prokaryotic DNA polymerases, including E. coli DNA polymerase and Thermus aquaticus (Taq) polymerase. More strikingly, DNA polymerase and adenylyl cyclase also share the same catalytic mechanism, using two metal ions to polarize the 3Јhydroxyl of the nucleotide and thus stabilize a pentavalent intermediate (42,47). Two metal ions are coordinated by a pair of aspartates that are spatially close, commonly seen as a DXD sequence in DNA polymerases. Interestingly, edema factor has a single DXD motif and we have shown that mutations at these two aspartates can drastically reduce catalysis (41). In addition, there is an arginine residue in mammalian adenylyl cyclase that is involved in the stabilization of a pentaphosphate intermediate (42,43). We have found that arginine 329 of edema factor is crucial for catalysis and may play a similar role in stabilizing the catalytic intermediate. With no other structural data available on the catalytic site of edema factor, the toxin's structural solution for the performance of a two-metalmediated catalysis remains unclear.
CaM adopts both extended and compact structures. It is well established that the compact conformation of CaM can bind an amphipathic peptide that leads to the activation of numerous CaM-regulated enzymes. There is significant evidence for other entirely different types of CaM target complexes and therefore possibly different mechanisms of target activation. For example, low angle x-ray scattering data shows that whereas one of the two CaM-binding peptides from phosphorylase kinase (PhK5) induces a compact conformation in CaM, the other (PhK13) induces an extended conformation (46). Described above, we have shown that CaM adopts an extended conformation to activate edema factor.
Labruyẽre et al. (8), have identified an amphipathic helix derived from residues 499 -532 of edema factor that binds CaM, thus possibly localizing a site for CaM interaction. In addition, Munier et al. (24), have performed a cross-linking study that localized CaM binding to residues 613-767. Whereas distances in primary sequence are poor predictors of spatial distances within folded proteins, the differing results for the location of a single CaM-binding site are commensurate with two binding sites interacting with an extended CaM. Additional spectroscopic and mutational analyses in combina-tion with molecular structure determination of edema factor alone and edema factor/CaM complex are currently in progress, and the solutions should yield valuable insight into how CaM activates the catalytic activity of edema factor.