Adenylyl cyclase Rv1264 from Mycobacterium tuberculosis has an autoinhibitory N-terminal domain.

Mycobacterium tuberculosis contains 15 class III adenylyl cyclase genes. The gene Rv1264 is predicted to be composed of two distinct protein modules. The C terminus seems to code for a catalytic domain belonging to a subfamily of adenylyl cyclase isozymes mostly found in Gram-positive bacteria. The expressed protein was shown to function as a homodimeric adenylyl cyclase (1 micromol of cAMP x mg(-1) x min(-1)). In analogy to the structure of the mammalian adenylyl cyclase catalyst, six amino acids were targeted by point mutations and found to be essential for catalysis. The N-terminal region represents a novel protein domain, the occurrence of which is restricted to several adenylyl cyclases present in Gram-positive bacteria. The purified full-length enzyme was 300-fold less active than the catalytic domain alone. Thus, the N-terminal domain appeared to be autoinhibitory. The N-terminal domain contains three prominent polar amino acid residues (Asp(107), Arg(132), and Arg(191)) that are invariant in all seven sequences of this domain currently available. Mutation of Asp(107) to Ala relaxed the inhibition and resulted in a 6-fold increase in activity of the Rv1264 holoenzyme, thus supporting the role of this domain as a potential novel regulator of adenylyl cyclase activity.

cAMP serves as a second messenger in virtually all organisms; yet at least three independent classes of adenylyl cyclases (AC) 1 exist. Class I ACs are present in bacteria such as Escherichia coli or Yersinia. Class II ACs are toxins, e.g. from Bacillus anthracis or Bordetella pertussis. Class III ACs are present in all phyla (1). The catalytic domain of the huge number of class III ACs has been termed cyclase homology domain (CHD). Based on distinct amino acid motifs, class III ACs have been subclassified (2), and in lower organisms, particularly in bacteria, class III CHDs seem to be linked with different protein domains that most likely impart peculiar regulatory features. However, so far only a few studies addressed bacterial class III ACs.
In the completed genome of Mycobacterium tuberculosis, 15 open reading frames were identified that contain a CHD (2). The availability of this information enables us to study each mycobacterial AC isoform individually in vitro with the per-spective to determine its contribution to the cAMP regulatory system during tuberculosis disease development in vivo. Two of the 15 AC open reading frames (Rv1625c and Rv2435c) belong to the mammalian-type ACs, and recent work concentrated on the membrane-bound mammalian-type AC present in Mycobacterium, Rv1625c (3,4). Four predicted mycobacterial ACs (Rv1318c, Rv1319c, Rv1320c, and Rv3645) contain CHDs that are part of a subclass consisting of ACs from, among others, Anabaena, Stigmatella, Rhizobium, and Treponema (2). The remaining nine mycobacterial CHDs (Rv1264, Rv1647, Rv2212, Rv0386, Rv1358, Rv1359, Rv2488c, Rv0891c, and LipJ) are most similar to CHDs detected in other Gram-positive bacteria.
Here we investigated the gene product of Rv1264. Earlier, AC genes of this subtype were identified by complementation cloning in Brevibacterium liquefaciens and Streptomyces (5,6). Those genes code for modular proteins with the CHD located C-terminally. The dismal expression of these ACs in E. coli precluded detailed biochemical studies (5,6). Thus, Rv1264 was used in an attempt to characterize this AC subtype. In addition, the biochemical characterization of the Rv1264 catalyst might constitute a starting point for future studies on the remaining eight related AC genes present in M. tuberculosis.
We were able to express reasonable amounts of the Rv1264 AC catalytic domain in E. coli with high AC activity. The catalytic site is the result of homodimerization, and catalysis depends on the same amino acids previously identified as crucial in mammalian ACs. The holoenzyme was much less active than the catalytic domain alone, suggesting an autoinhibitory function of this unique N-terminal domain, which contains no similarity to any other known protein module and so far is identified only in altogether seven ACs, i.e. four in Mycobacteria, two in Streptomyces, and one in Brevibacterium.   was cloned into the BamHI and XmaCI sites of pQE30, thus adding an MRGSH 6 GS tag N-terminally. Similarly, Rv1264  plus 26 bp at the 3Ј end were amplified, fitted with BamHI sites at both ends, and cloned into the BamHI site of pQE30. Point mutations were introduced by PCR using the expression cassettes as templates. Nearby unique restriction sites were used for assembly of the fragments. The following additional restriction sites were introduced by silent mutations: a StuI site at base position 411 for construc-* This work was supported by the Deutsche Forschungsgemeinschaft and the Fonds der Chemischen Industrie. 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 U.S.C. Section 1734 solely to indicate this fact.
The expression plasmid for the N-terminal domain was generated from pQE30 containing Rv1264  by digestion with NaeI and circularization of the largest fragment. This created an expression cassette of Rv1264  with an N-terminal His 6 tag and a C-terminal SSP tripeptide extension. Primer sequences will be provided on request.
Expression and Purification of Proteins-Expression plasmids were transformed into E. coli BL21(DE3)[pRep4]. The cells were induced with 60 M isopropyl-1-thio-␤-D-galactopyranoside and harvested after expression for 6 h at room temperature. The bacteria were washed with buffer (50 mM Tris-HCl, 1 mM EDTA, pH 8), frozen in liquid nitrogen, and stored at Ϫ80°C. For purification the cells were suspended in 20 ml of lysis buffer (50 mM Tris-HCl, pH 8, 2 mM 3-thioglycerol), sonicated for 30 s, and treated with 0.2 mg/ml lysozyme for 30 min on ice. The incubation was continued for a further 30 min at 0°C after addition of 5 mM MgCl 2 and 10 g/ml of DNase I. After centrifugation (31,000 ϫ g, 30 min) 250 mM NaCl and 15 mM imidazole (final concentrations) were added. 250 l of nickel-nitrilotriacetic acid-agarose equilibrated in 5 ml of buffer A (lysis buffer containing 250 mM NaCl, 15 mM imidazole, 5 mM MgCl 2 ) were then added, and the protein was allowed to bind for 60 min. The resin was transferred into a column, washed with 10 ml buffer A, and subsequently washed with 5 ml of buffer B (lysis buffer containing 15 mM imidazole, 5 mM MgCl 2 ). The protein was eluted with 0.4 ml of buffer C (37.5 mM Tris-HCl, pH 8, 250 mM imidazole, 2 mM MgCl 2 , 1.5 mM 3-thioglycerol). Purified proteins could be stored without loss of activity in the presence of 50% glycerol at Ϫ20°C.
Adenylyl Cyclase Assay-AC activity was measured for 10 min at 37°C in a volume of 0.1 ml. The standard reaction contained 22% glycerol, 50 mM Tris-HCl, pH 8.0, 3 mM MnCl 2 , 0.5 mM [␣-32 P]ATP (25 kBq), and 2 mM [2, H]cAMP (150 Bq) to determine yield during separation of ATP from cAMP (7). For kinetic analysis variable amounts of MnATP were used in the presence of 3 mM free Mn 2ϩ . Mixtures of mutants were incubated for 10 min on ice prior to the start of the reaction. All data are the means of two to six points and are denoted with their standard deviations.
Cross-linking by Glutaraldehyde-Rv1264 (211-397) was dialyzed against cross-linking buffer (50 mM sodium phosphate, pH 7, 20% glycerol, 5 mM MgCl 2 ). After removal of precipitates, 310 nM protein were incubated with 7 mM glutaraldehyde (final concentration) for 60 min at room temperature. The reaction was quenched by the addition of 0.25 volumes of SDS-PAGE sample buffer, and 15 l were separated on a 15% SDS-PAGE gel. The proteins were blotted onto a polyvinylidene difluoride membrane and visualized with an anti-RGS-His 4 antibody (3).

Sequence Features of the M. tuberculosis AC Rv1264 -AC
Rv1264 was analyzed by BLAST (8) and Smith-Waterman (9) searches. The N terminus was similar to the respective Nterminal AC domains detected in another six firmicute-type ACs, one from B. liquefaciens, two from Streptomyces, Rv2212c from M. tuberculosis, and two DNA sequences of Mycobacterium smegmatis, which probably are homologs of the M. tuberculosis Rv1264 and Rv2212c. An alignment of the seven Nterminal domains demonstrated similarities scattered over several patches (Fig. 1A). The overall identities between different genera ranged from 14.3% (M. smegmatis versus Streptomyces griseus) to 33.3% (B. liqufaciens versus S. griseus). Among six amino acids that appear invariable in all sequences, an aspartate and two arginines were particularly noteworthy (marked in Fig. 1A).
The C terminus of Rv1264 (amino acids 211-397) contains a canonical class III CHD. Alignment with other bacterial class III CHDs of this subfamily and with the mammalian-type AC Rv1625c from M. tuberculosis demonstrates the presence of all six amino acids in register in a single polypeptide chain that have been identified by x-ray data to be critical for catalysis in the mammalian pseudoheterodimeric ACs. There these six amino acids are provided by two separate CHDs, termed C1 and C2, forming a pseudoheterodimeric superstructure ( Fig.  1B and Refs. 10 -12). In the bacterial ACs, these catalytic amino acids are imbedded in clusters of sequence identities, whereas otherwise the sequences diverge considerably in amino acid composition and number, i.e. polypeptide chain length. The first, most N-terminal Asp, which is required for metal cation binding, is connected via a sequence of 37 or 38 amino acids to a conserved stretch that includes a substrate binding Lys invariably separated by three rather conserved amino acids from the second metal-binding Asp residue. What follows is a stretch of about 50 amino acids, which again is highly variable in length, in essentially all CHDs irrespective of their origin. Usually, the next block, which shows a high extent of length and amino acid conservation, starts with an Arg (marked in Fig. 1B), which binds to the ␥-phosphate of the substrate ATP (10). In an alignment of 15 CHDs from Grampositive bacteria (nine from M. tuberculosis including Rv1264, two from M. smegmatis, one from Mycobacterium leprae, two from Streptomyces, and one from B. liquefaciens), a highly significant gap of three to seven amino acids was observed in this length-conserved region (indicated by a bar in Fig. 1B). This gap is immediately ahead of a conserved region comprising 12 amino acids (DX 6 NX 3 R) that includes a second substrate-binding Asp and an Asn and an Arg that stabilize the transition state. The crystal structure of mammalian ACs indicates that this sequence of invariable length constitutes a prominent dimerization region (10,13). Therefore, we examined whether the catalytic domain of Rv1264 functions as a homodimer in the canonical sense that two domains associate in a head-to-tail fashion using the pinpointed six catalytic amino acid residues. This would indicate significant differences to the known AC structure and suggest a considerable flexibility with respect to the protein scaffold that provides the backbone for the catalytic AC fold.
Biochemical Characterization of the Catalytic Domain of Rv1264 -The CHD Rv1264 (211-397) was expressed in E. coli mostly in inclusion bodies. The minor soluble fraction of the expression product (Ͻ10%) was purified to homogeneity (Fig.  2). Rv1264 (211-397) displayed high AC activity with Mn 2ϩ as a cofactor (1.2 Ϯ 0.1 mol of cAMP⅐mg Ϫ1 ⅐min Ϫ1 ), whereas Mg 2ϩmediated catalysis was rather poor (6 Ϯ 2 nmol of cAMP⅐mg Ϫ1 ⅐min Ϫ1 ). ATP substrate specificity was stringent, GTP was not accepted irrespective of the metal cation employed. The temperature optimum was at 37°C, the activation energy of 53 kJ/mol was within the usual range, and the pH optimum was around 7.3 (tested range, 5-8.5). The V max of 1.3 mol of cAMP⅐mg Ϫ1 ⅐min Ϫ1 was derived from a Michaelis-Menten plot. The specific activity is thus comparable with that of the catalytic domain of AC Rv1625c (3) ( Table I). The Hill coefficient of 1.6 indicated a pronounced cooperativity for ATP, consistent with the presence of two catalytic sites. This was supported by the nonlinear protein dependence of the cyclase reaction (Fig. 3A). The specific activity tripled with increasing protein concentration as expected for an equilibrium between monomers and oligomers. An apparent association constant of 50 Ϯ 15 nM was calculated from a reciprocal plot of the dependence of the specific activity on the enzyme concentration, indicating a high affinity of the Rv1264 (211-397) monomers for each other. The formation of dimers was substantiated by glutaraldehyde cross-linking experiments. At 310 nM Rv1264 (211-397) , which was used in standard assays, a band of a dimer-sized protein was detected by a Western blot (Fig. 3B). The limited extent of cross-linking is rationalized by the fact that Rv1264 (211-397) has only two Lys residues, which are the preferential targets for cross-linking by glutaraldehyde.
Site-directed Mutagenesis of Rv1264 (211-397) -Because of the peculiar gap in a conserved stretch of protein mentioned above (Fig. 1B), it was necessary to ascertain whether the same residues crucial for catalysis in mammalian-type CHDs were used (10 -12). Six critical amino acids were individually mutated to alanine in Rv1264 (211-397) . The proteins were expressed in E. coli and purified to homogeneity via their Nterminal His 6 tag (Fig. 2). In all mutants, activity was below 10% of the wild-type level. Rv1264 (211-397) N319A and Rv1264 (211-397) R323A, i.e. mutants in the amino acids stabilizing the transition state (10), showed 3.1 and 8.7% residual activity, respectively (Fig. 2). The mutants K261A and D312A, targeting the purine-binding residues (10,14,15), showed very little activities. D312A had a detectable activity of 0.2 nmol of cAMP⅐mg Ϫ1 ⅐min Ϫ1 only at a protein concentration of 1.3 M, whereas the mutants D222A and D265A, which cannot anymore coordinate the chain of phosphate residues via metal-ion binding (12), were inactive irrespective of the protein concentration. To determine whether the mutants were misfolded or retained a functional conformation, we tested whether they could complement each other to a functional heterodimer provided that the mutations affected amino acids contributed to the catalytic fold by different protein loops as reported previously for the mammalian-type mycobacterial AC Rv1625c (3). Accordingly mutants Rv1264 (211-397) D222A and Rv1264 (211-397)D265A, which are predicted to be functional equivalents of a mammalian C2 domain, should positively interact with mutants Rv1264 (211-397) N319A and Rv1264 (211-397) R323A, which correspond to mammalian C1 equivalents.
In general, reconstitution was possible but poor when compared with wild-type activity (Fig. 4 for a model and three  experimental examples). Upon the addition of an excess of inactive Rv1264 (211-397) D222A to Rv1264 (211-397) N319A, a 3.5fold increase in cAMP production was observed (Fig. 4B). Sim-ilarly, a molar excess of inactive Rv1264 (211-397) D222A or Rv1264 (211-397) D265A activated Rv1264 (211-397) R323A 3-fold to about 200 nmol⅐mg Ϫ1 ⅐min Ϫ1 . This was further evidence of a dimeric catalyst and for correctly folded mutant proteins. The lower than wild-type levels of cyclase activity could indicate that the mutations caused a slight and local conformational deterioration or were to some extend involved in dimerization (see below).
Characterization of Rv1264 Holoenzyme-Next the Rv1264 holoenzyme was expressed in E. coli and purified (Fig. 5A). The specific activity was 2.5 Ϯ 0.2 nmol of cAMP⅐mg Ϫ1 ⅐min Ϫ1 , i.e. more than 300-fold less than that of Rv1264 (211-397) . Activity of the holoenzyme was only detectable with Mn 2ϩ as a metal cofactor; Mg 2ϩ did not support activity. A kinetic characterization showed a calculated V max of 3% compared with Rv1264 (211-397), and the apparent affinity for ATP was reduced about 6-fold (Table I). This suggested that the N-terminal domain is an intrinsic inhibitor of AC Rv1264. The Hill coefficient of 1.9 indicated a strong cooperativity, i.e. the dimerization was also a prominent feature of the holoenzyme.
The similarity of AC Rv1264 to the cloned B. liquefaciens AC was taken as a hint to look for small metabolites as effectors because recombinant B. liquefaciens AC was reported to be stimulated by pyruvate (6). This was in agreement with an unspecific activation of a purified AC from B. liquefaciens by compounds such as ␣-ketocarbonic acids, glycine, alanine, and lactate (16). The Rv1264 holoenzyme was not activated by 10 mM D,L-lactate, 1 mM pyruvate, ␣-ketobutyrate, D-alanine, L-   . 3. Dimerization of Rv1264 (211-397) . A, protein dependence of specific AC activity of Rv1264 (211-397) . B, glutaraldehyde (GA) crosslinking of Rv1264 (211-397) analyzed by Western blot. The apparent molecular masses (in kDa) of the bands (left) were derived from standards on the blotted membrane. alanine, phosphoenolpyruvate, and 19 other compounds from the glycolysis, citric acid cycle, and amino acid-related biochemical pathways.
Inhibition by the N-terminal Domain-The N terminus might inhibit the AC activity of Rv1264 autonomously or only if fused to the catalytic domain. Therefore, we expressed and purified the N terminus (amino acids 1-207) (Fig. 5A). Even at 15 M the N-terminal domain reduced the activity of Rv1264 (211-397) by only 30% (data not shown). Thus, the inhibition of the catalytic domain appeared to require a linked configuration.
We hypothesized that the invariably conserved Asp and Arg residues in the N terminus might be important for inhibition and that mutation of these residues to Ala might attenuate inhibition. Therefore, we generated Rv1264D107A and Rv1264R132A mutant holoenzymes. The recombinant proteins were purified from E. coli (Fig. 5B). The Rv1264R132A mutant had the same low activity as the wild-type protein (Fig. 5C).
However, the removal of a carboxyl group in the Rv1264D107A mutant resulted in a 7-fold more active AC (Fig. 5C). A kinetic analysis of Rv1264D107A revealed a 3-fold increase in V max and a slightly enhanced affinity for ATP (Table I). This established this unique N-terminal domain of Rv1264 as a likely autoinhibitory module.

DISCUSSION
The occurrence of 15 genes coding for class III ACs in M. tuberculosis (2) implies that cAMP-mediated signal transduction may be a central and versatile tool that this pathogen can employ to process multiple environmental challenges. Nine of these genes code for CHDs that are typical for ACs in Grampositive bacteria. Thus, the understanding of the structurefunction relationships in this type of CHDs appears important for an understanding of cAMP signaling in M. tuberculosis and other bacteria. AC Rv1264 was chosen for this study because it has the same modular composition as the single AC identified in Streptomyces and an AC present in B. liquefaciens (5,6).
We addressed three major questions: Is AC Rv1264 operating as a homodimer like mammalian-type AC Rv1625c despite a major gap in a conserved region implicated in dimerization? Are the same amino acids that are known to be critical for catalysis in mammalian ACs also involved in catalysis in Rv1264? What is the potential function of the novel N-terminal domain?
The former two aspects were investigated by use of the recombinant CHD, i.e. Rv1264 (211-397) . A sequence comparison of Rv1264 with mammalian-type ACs demonstrated that all critical amino acids can be properly aligned. This argues for a canonical homodimerization. However, a prominent seven-amino acid gap will result in the almost complete elimination of a distinct loop between an antiparallel ␤-sheet that is formed by the ␤ 4 and ␤ 5 strands in the mammalian AC heterodimer (10,13). This loop has been shown to be important for dimerization, and its absence in Rv1264 raised doubts about its catalytic mechanism. Four experimental approaches were used to demonstrate a functional homodimerization of Rv1264. First, the ATP dependence of Rv1264 (211-397) was positively cooperative, consistent with more than one catalytic site in a dimer. Second, the specific activity increased with the protein concentration as expected for an assembly of monomers in catalyst formation. Third, glutaraldehyde cross-linking demonstrated the physical presence of Rv1264 (211-397) dimers. Fourth, two almost inactive mutants of Rv1264 (211-397) , N319A and R323A, which functionally correspond to mammalian AC C1 domains, were partially complemented by addition of inactive mutant proteins such as Rv1264 (211-397) D222A and D265A, which were the functional equivalents of mammalian AC C2 domains. In the latter experiments the reconstitution of a single wild-type catalytic center was predicted (see Fig. 4A for a pictogram). However, the observed specific AC activity was much less than Rv1264  . The lack of a full complementation of these Rv1264 C1/C2 equivalents may be the consequence of a profound difference of dimerization epitopes compared with mammalian ACs, as suggested by the gap in the dimerization arm (marked by a bar in Fig. 1B). In fact, it cannot be excluded that amino acids that were targeted by the point mutations also affect the orientation of the monomers and thus play an additional role in catalyst formation that obviously could not be substituted in a mixture of two otherwise complementary mutants.
The individual mutation of all six amino acids that are implicated in catalysis (10 -12, 14) was detrimental. The protein folding of these mutants appeared to be largely intact because there was either measurable residual activity in mutants targeting the stabilization of the transition state (N319A and R323A) and the adenine binding site (K261A and D312A) or a complementation with inactive mutants in the metal-binding sites (D222A and D265A). These features are highly similar to results obtained by alanine mutagenesis of the mammaliantype AC Rv1625c of M. tuberculosis (3). Finally, of the nine mycobacterial ACs that belong to this subtype, only three (Rv1264, Rv1647, and Rv2212) have the unaltered set of six amino acids assumed to be critical for catalysis. It will be interesting to investigate whether the amino acid changes in the other six ACs that comprise two to four of the six catalytic amino acids in each isozyme will be compatible with AC activity. Preliminary data indicate that the conservation of the catalytic center is considerably relaxed in those ACs. Obviously, this raises novel questions concerning the structural details of the catalytic cleft in different AC subtypes.
The AC activity of the Rv1264 holoenzyme was almost completely masked by its N-terminal domain. This was in agreement with and explained the tiny activity of Rv1264 (1-397) that has been reported recently (4). Because the recombinant and purified N-terminal domain alone (Rv1264  ) did not inhibit the CHD Rv1264 (211-397) when added separately, the physical linkage of both protein modules seems to be required and sufficient for the autoinhibitory effect. One prerequisite for a domain to be a regulator is the flow of information from the regulator to the effector, and the key property of an inhibitor is that a perturbation of its function leads to the relief of inhibition, i.e. an activation. Both conditions are met by the Nterminal domain of Rv1264 because the mutation of a conserved Asp 107 to Ala led to a highly significant increase of cyclase activity. A comparison of kinetic parameters of Rv1264  , Rv1264 (211-397) , and Rv1264 (1-397) D107A indicated that the inhibitory module decreased the catalytic rate as well as the apparent affinity for ATP. Thus, it may put a conformational constraint on the catalytic domain via the linker region.
A similar modularized AC from B. liquefaciens was reported to be activated by pyruvate, other ␣-ketocarbonic acids, lactate, and some amino acids (5,16). Similar tests with the mycobacterial Rv1264 AC failed to relieve the inhibition. Therefore, the physiological signal that may alleviate the inhibition of the catalytic region remains enigmatic at present.
The inhibitory domain of AC Rv1264 is novel and was not identified previously in other proteins. Therefore, the attempt of a mutational analysis had to rely fully on the analysis of similarity patterns present in six other ACs of this subtype. Because Rv1264 holoenzyme can be purified in large quantities, an elucidation of its structure will be possible and probably allow a more detailed insight into the relationship between the autoinhibitory and catalytic domains.