Structural Requirements for N -Trimethylation of Lysine 115 of Calmodulin*

Calmodulin is trimethylated at lysine 115 by a highly specific methyltransferase that utilizes S -adenosylme-thionine as a co-substrate. Lysine 115 is found within a highly conserved six-amino acid loop (LGEKLT) that forms a 90° turn between EF-hand III and EF-hand IV in the carboxyl-terminal lobe. In the present work a mutagenesis approach was used to investigate the structural features of the carboxyl-terminal lobe that lead to the specificity of calmodulin methylation. Three structural regions within the carboxyl-terminal lobe appear to be involved in methyltransferase recognition: the highly conserved six-amino acid loop-turn region that contains lysine 115 as well as the adjacent a -helices (helix 6 and helix 7) from EF-hands III and IV. Site-directed mutagenesis of residues in the loop show that three residues, glycine 113, glutamate 114, and leucine 116 are essential for methylation. In addition, subdomain (individual helix or Ca 2 1 binding loop) exchange mutants show that the substitutions of either helix 6 (EF-hand III) with helix 2 (EF-hand I) or helix 7 (EF-hand IV) with helix 3 (EF-hand II) compromises methylation. Charge-to-alanine mutations in helix 7 show that substitution of conserved charged residues at positions 118, 120, 122, 126, and 127 reduced lysine 115 methylation rates, suggesting possible electrostatic interactions between this helix and the methyltransferase. Single substitutions in helix 6 did not affect

Calmodulin is a highly conserved calcium sensor protein that modulates the activities of multiple enzymes. Calmodulin is a monomer consisting of two structurally similar globular calcium binding lobes (1) connected by a flexible linker region (2,3). Each lobe consists of two helix-loop-helix EF-hand calcium binding sites, with EF-hand domains I and II constituting the amino-terminal lobe and EF-hands III and IV constituting the carboxyl-terminal lobe. Many naturally occurring calmodulins are posttranslationally trimethylated on a single lysine residue at position 115 (reviewed in Ref. 4). Lysine 115 is a solventexposed residue that is found on a highly conserved six-amino acid loop-turn region (LGEKLT) located between helix 6 of EF-hand III and helix 7 of EF-hand IV (Fig. 1). Trimethylation of calmodulin at lysine 115 is catalyzed by an N-methyltransferase that utilizes S-adenosylmethionine as a co-substrate (Refs. [5][6][7][8][9]reviewed in Ref. 4). This enzyme appears to have the dedicated function of trimethylating lysine 115 in calmodulin from a wide variety of species. From a functional perspective, calmodulin methylation selectively affects the regulation of certain enzymes such as NAD kinase (10 -12) and might also influence posttranslational ubiquitination of the protein (13).
Previous work shows that the site on calmodulin recognized by the calmodulin N-methyltransferase resides solely on the COOH-terminal lobe (residues 78 -148) (7). Mutations or chemical modifications that affect the hydrophobic core and conformation of calmodulin disrupt methylation (7,14), and it seems as if the enzyme requires more than a linear sequence of amino acids or the simple surface exposure of lysine 115 for recognition and methylation.
An examination of the calmodulin structure shows that the amino-terminal and carboxyl-terminal lobes share remarkable structural similarity and symmetry (1). In previous work we performed a series of domain duplication and exchange mutagenesis experiments in which EF-hands III and IV were substituted with the symmetry-related EF-hands I and II (15). These experiments showed that structural features unique to both EF-hands are required for methyltransferase binding and methylation. To define more precisely the regions responsible for calmodulin methyltransferase specificity, we exploited this domain exchange approach further and performed site-directed mutagenesis of various residues in the carboxyl-terminal lobe surrounding the methylation site. The results implicate specific regions in the methylation loop/turn region between EFhands III and IV as well as residues within the adjacent ␣-helices in the binding and recognition of the enzyme.

EXPERIMENTAL PROCEDURES
Molecular Cloning Techniques-All mutagenesis and expression experiments were done with the calmodulin expression plasmid pVUCH (16), which contains the cloned synthetic VU-1 calmodulin gene (17). Synthetic oligonucleotides were obtained either from Oligos Etc. or Life Technologies, Inc. All site-directed substitutions of calmodulin were generated by oligonucleotide-directed mutagenesis using the QuikChange TM site-directed mutagenesis kit (Stratagene). Substitution of entire subdomains of calmodulin was done by cassette mutagenesis, as described previously (18). Briefly, targeted regions in the coding region for the carboxyl-terminal lobe of calmodulin were removed by digestion with restriction enzyme pairs that flank the regions of interest. Synthetic oligonucleotide cassettes containing the homologous sequences from the amino-terminal lobe were engineered with complementary ends of the corresponding restriction enzyme. These were inserted by ligation into digested VU-1 gene as described previously (18). To enable greater latitude in the restriction sites available for cassette mutagenesis, two restriction sites were removed from pVUCH-1 by site-directed mutagenesis: an AatII site within the vector * This work was supported by United States Department of Agriculture National Research Initiative Competitive Grants Program Award 9703548 and National Science Foundation Grant MCB-9904978. 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. and a BamHI site flanking the ptac promoter. Neither mutation affected vector performance or calmodulin expression. Subdomain mutants and corresponding cassettes and enzymes are as follows.
All expression plasmids were transformed into Escherichia coli JM101 as described previously (18). All mutants were confirmed by automated DNA sequencing on a Perkin-Elmer Applied Biosystems 373 DNA sequencer at the University of Tennessee Molecular Biology Research Facility. Sequencing reactions were done with a Prism dye terminator cycle sequencing kit (Perkin-Elmer).
Protein Purification-Calmodulin was isolated from E. coli expression clones by phenyl-Sepharose chromatography as described previously (15,17). Calmodulin N-methyltransferase was purified from rat testes by a modification of a previous procedure (7). The enzyme was extracted and purified through the differential ammonium sulfate step as in Han et al. (7). At this step, the enzyme was applied to a Sephadex G-10 (6 ϫ 5 cm) column equilibrated in 10 mM Hepes-NaOH, pH 7.4, 4 mM ␤-mercaptoethanol, 0.01% (w/v) Triton X-100, and 0.1 mM phenylmethylsulfonyl fluoride. The absorbance of the effluent was continuously monitored at 280 nm. The fractions containing protein were combined and were applied to a 2.5 ϫ 16-cm column of DEAE-cellulose (Whatman DE-53) equilibrated in the same buffer. The non-binding protein fraction, which contains the calmodulin N-methyltransferase, was collected. Calcium was added to 2 mM and calmodulin-Sepharose chromatography was done as described previously (7). The fractions with calmodulin methyltransferase activity were pooled and concentrated by ultrafiltration on a Centricon-30 unit (Amicon). Calmodulindependent cyclic nucleotide phosphodiesterase (sheep brain) and NAD kinase (pea seedlings) were prepared as described previously (14).
Other Analytical Methods-Calmodulin methyltransferase was assayed as described previously (7) in a standard assay buffer of 0.1 M sodium glycylglycine (pH 8.0), 0.15 M KCl, 2 mM MgCl 2 , 5 mM dithiothreitol, 0.01% (w/v) Triton X-100, and either 1 mM CaCl 2 or 1 mM EGTA. The determination of kinetic parameters for the methyltransferase were derived under pseudo-first order conditions with a constant concentration of 12 M S-[methyl-3 H]adenosylmethionine (1.25 Ci/ nmol) and various concentrations of calmodulin (14). Apparent K m and V max parameters were determined by fitting the data to the Michaelis-Menten equation, and k cat was determined by using 38,000 as the molecular weight of the methyltransferase as described previously (7). Pea NAD kinase activity was assayed as described previously (17). Cyclic nucleotide phosphodiesterase assays were done as described previously (19). The pea NAD kinase and calmodulin-dependent cyclic nucleotide phosphodiesterase activation curves were generated by best fits to the data using the Hill equation, where v is the initial enzyme rate, and K 0.5 is the concentration of calmodulin for half-maximal activation. V max is the enzyme rate at maximal activation of the enzyme, [CaM] 1 is the concentration of calmodulin, and n represents the Hill coefficient. Protein concentrations were determined by the bicinchoninic acid assay (Pierce) by using bovine serum albumin as a standard.

Mutations in the Methylation
Loop-Lysine 115 is located on a highly conserved, six-amino acid (LGEKLT, residues 112-117) loop-turn region between EF-hands III and IV (Fig. 1). To test whether this conserved six-amino acid motif is required for calmodulin methylation, a series of mutations were generated ( Fig. 1). Three mutations severely affected lysine 115 methylation: G113S, E114A, and L116T. In all three cases, the substitutions resulted in a complete loss of the ability of calmodulin to serve as a methylation substrate ( Fig. 2; Table I). Examination of the other two conserved residues (Leu-112, Thr-117) in the methylation loop showed that they are less important in activity (Table I). Of the two, substitution of a threonine at Leu-112 showed the most significant effect, with a 4.5-fold reduction in the catalytic efficiency of methylation when assayed in the absence of calcium (Fig. 2B, Table I). However, unlike calmodulin with substitutions at residues 113, 114, and 116, the methylation properties of L112T were essentially identical to wild type VU-1 calmodulin in the presence of saturating calcium ( Fig. 2A), suggesting that this mutation selectively affects the substrate properties of apo-calmodulin.
In contrast to their influence on methylation, all methylation loop mutants showed essentially indistinguishable NAD kinase and calmodulin-dependent cyclic nucleotide phosphodiesterase activation profiles (not shown), suggesting that the substitutions of these highly conserved residues in the methylation loop do not significantly alter calmodulin activator functions.
Production and Analysis of Sub-domain Exchange Mutations of Calmodulin-Although the conserved residues in the methylation loop are essential for trimethylation of lysine 115, it is clear that additional flanking regions within EF-hands III and 1 CaM, calmodulin; VU-1 calmodulin, calmodulin derived from a cloned synthetic gene (17). IV are also essential for methyltransferase recognition (15). To identify the critical regions within EF-hands III and IV necessary for methyltransferase recognition, subdomain exchange mutants were generated. In these mutants, the various helices and calcium binding loops of EF-hands III and IV were replaced with the homologous regions of EF-hands I and II (Fig.  1A).
The conservative nature of these substitutions is underscored by the observation that all subdomain mutant exchange mutants activated calmodulin-dependent cyclic nucleotide phosphodiesterase and NAD kinase similar to VU-1 calmodu-lin, with the exception of CaM H6 , which showed a modest reduction in maximal activation of NAD kinase (Table II). In contrast to the activator properties, the methylation properties of calmodulin were substantially affected by the exchange of certain subdomain elements (Fig. 3). The two subdomain mutant calmodulins with substitutions adjacent to the methylation loop, CaM H6 and CaM H7 , showed the most substantial effects on the rate of calmodulin methylation (Fig. 3). CaM H6 was a poor methylation substrate in both its apo and calciumsaturated state (Fig. 3), showing a catalytic efficiency (k cat /K m ) that was 25-fold (apo) or 13-fold (Ca 2ϩ -bound) lower than that  (17) are aligned with symmetry-related sequences of EF-hands I and II. The boxed areas represent the sequences in the carboxylterminal lobe that were replaced with the homologous sequence from the amino-terminal lobe for the generation of subdomain mutants: CaM CL3 , CaM H6 , CaM H7 , CaM CL4 , and CaM H8 . B, site-directed substitutions within the methylation loop and adjacent ␣-helices 6 and 7. Number signs (#) indicate residues on the hydrophobic faces of the helices 6 and 7. The arrows below the sequence alignment indicate residues that have been mutated, and the star indicates the site of methylation, Lys-115.  of VU-1 calmodulin (Table III). In contrast, the ability of CaM H7 to serve as a methylation substrate was strictly dependent on calcium. In the absence of calcium, CaM H7 was incapable of being methylated (Fig. 3B). However, in the presence of saturating calcium, the methylation kinetics of CaM H7 were nearly restored to the level of VU-1 CaM (Fig. 3A).
The other subdomain mutants exhibited a smaller effect on methyltransferase kinetics (Table III). The mutations within the two calcium binding loops (CaM CL4 and CaM CL3 ) showed catalytic efficiencies with the methyltransferase that were comparable with VU-1 (Table III). The CaM H8 mutant showed slightly defective methylation in its apo form, with a slightly lower k cat (0.0172 versus 0.0124 s Ϫ1 ) and 4-fold higher K m . However, relatively normal kinetic behavior with CaM H8 was observed under conditions of saturating calcium (Table III).
Scanning Mutagenesis of Helix 6 -A comparison of the se-quence in helix 6 of wild type calmodulin (LRHVMTN (105-111)) with the substituted sequence in mutant CaM H6 (LGTVMRS) reveals no change in the hydrophobic residues that are involved in the packing of helix 6 with the other helices in the carboxyl-terminal lobe (Fig. 1B). The possibility that surface residues in helix 6 are important for calmodulin interaction with the methyltransferase was therefore addressed. The CaM H6 mutant has three principal changes compared with wild type: the removal of two positive charged residues (Arg-106 and His-107 replaced with Gly and Thr) and the introduction of a bulky, charged arginine for Thr-110. The remaining substitution, an Asn-111 to Ser change has previously been shown not to affect methylation (5,20). To test whether these residues are important in methylation, charge-to-alanine mutants were generated for Arg-106 and His-107, and Thr-110 was changed to an arginine. In contrast to CaM H6 , these point mutants show little effect on calmodulin methylation (Fig. 4A).
Overall, the data suggest that the removal or alteration of individual surface charge residues on helix 6 does not appear to affect the methylation of either apoCaM or Ca 2ϩ -CaM but rather that the entire substitution of helix 6 with helix 2 results in a structural change that disrupts the site of methylation. Scanning Mutagenesis of Helix 7-An examination of helix 7 (DEEVDEMIREA (118 -128)) shows that this region possesses an extremely high number (7 out of 11) of conserved charged (mainly acidic) residues. These residues provide a surface-exposed patch of negative charge on the carboxyl-terminal lobe adjacent to lysine 115 (Fig. 5). This observation and the previous finding that calmodulin methylation is sensitive to ionic strength (7) raises the possibility that charge-charge interactions could potentially be involved in methyltransferase binding. To test this, a series of mutations of charged residues in   helix 7 were generated and assayed for their effects on methylation (Fig. 4B). The substitution of residues within the charged cluster (residues 118 -120) immediately adjacent to the methylation loop showed the greatest effects on methylation. The calmodulin mutant DEE118 -20KKK showed the greatest defect, with the methylation rate decreased 10-(in the presence of calcium) to 30-(in absence of calcium) fold relative to VU-1 calmodulin. This calmodulin possesses three lysine substitutions within this conserved cluster, resulting in a neutralization of nearly half of the negative electrostatic potential surface on the carboxyl-terminal lobe (21). Each negative charge was then removed by alanine scanning mutagenesis. E120A and D118A showed the greatest effects with a 6-fold and 3-fold reduction in methylation rate, respectively, in the absence of calcium (Fig.  4B). E120A was methylated normally in the presence of calcium, whereas D118A showed a 2-fold reduced methylation rate. In contrast, E119A was methylated normally (Fig. 4B), suggesting this residue is less important in calmodulin methyltransferase recognition.
Other charged residues on helix 7 were also analyzed (Fig.  4B) and showed more modest effects. A decreased activity was observed with E127A to (46% of VU-1) as well as D122A and R126A (60% of VU-1) in the absence of calcium. The effect of these mutations was less apparent in the presence of calcium, suggesting that calcium binding enhances substrate activity and at least partially overcomes the effects of the mutations. DISCUSSION Calmodulin is trimethylated at lysine 115 with a high degree of specificity by a dedicated calmodulin lysine N-methyltransferase. In the present study, domain exchange and scanning mutagenesis were done to attempt to identify regions of the protein that contribute to this specificity. The results suggest that three structural regions within the carboxyl-terminal lobe appear to be involved: the highly conserved six-amino acid loop-turn region that contains lysine 115 as well as the adjacent ␣-helices (helix 6 and 7) from EF-hands III and IV.
The six-amino acid methylation loop (LGEKLT) is highly conserved among phylogenetically diverse calmodulins, and it is reasonable to suggest that its structure provides features necessary for calmodulin methyltransferase recognition. Structural studies suggest that the loop shows greater flexibility and dynamics compared with the calcium binding loops and helices of the EF-hands (22)(23)(24). The loop provides a 90°hairpin turn between EF-hands III and IV, which is facilitated by Gly-113 (/ ϭ 93°/10°(22)), and its conformation is stabilized by three hydrogen bonds between the backbone amide nitrogens of Gly-113 and Glu-114 and the backbone carbonyl oxygens of Met-109 and Thr-110 in helix 6 of EF-hand III (22). In addition, Leu-116 is imbedded in the core of the carboxyl-terminal lobe, forming hydrophobic interactions with residues on the hydrophobic faces of the helices from EF-hands III and IV (1). Glu-114 and Lys-115 are solvent-exposed charged residues with no apparent contacts with other parts of the calmodulin structure (Fig. 5).
Calmodulin methylation is exquisitely sensitive to the substitutions of G113S, E114A, and L116T, which essentially abolish the methylation of lysine 115. These defects were observed regardless of calcium concentration, and thus, both the recognition of calcium-bound as well as apo-calmodulin was affected. Based on the structural features of these residues discussed above, some potential roles in methyltransferase recognition can be suggested. The substitution of glutamate 114 with an alanine removes a surface negative charge adjacent to the site of methylation and, as discussed further below, could provide an electrostatic contact for the enzyme. The substitution of a serine for the highly conserved glycine 113 likely alters the conformational flexibility of the loop-turn structure and might prohibit the residues within the loop from adopting an orientation suitable for methyltransferase binding and catalysis. The substitution of L116T, which is one of 14 residues composing the hydrophobic core of the carboxyl-terminal lobe, could alter the packing of the hydrophobic side chains and the interaction of the methylation loop with the hydrophobic core. Interestingly, none of these mutations significantly affects activation of two calmodulin-dependent enzymes, suggesting that their structural effects are subtle, selectively affecting methyltransferase recognition but not other calmodulin functions.
The conserved residues of the methylation loop are not in themselves adequate to confer methylation. For example, previous work (15) showed that the introduction of the methylation loop at a symmetrical position within the amino-terminal lobe did not result in lysine methylation. Furthermore, the replacement of either EF-hand III or IV with the homologous EF-hand I or II also results in the loss of lysine methylation (15). In the present study, we find that the critical regions are the ␣-helices adjacent to the methylation site, helix 6 of EFhand III and helix 7 of EF-hand IV.
The substitution of helix 2 (LGTVMRS) for helix 6 (LRH-VMTN) resulted in a substantial reduction in the rate of lysine methylation in both the presence and absence of calcium. A comparison of the structure of these related regions shows that they have remarkably similar backbone structures and form nearly identical packing interactions in their respective helical bundles within the amino or carboxyl termini (1,23,25). Thus, the loss of methylation was thought to be the result of alterations of surface residues that presumably interact with the methyltransferase. However, individual charge-to-neutral substitutions at these positions showed essentially normal methylation. Thus, the conservation of these surface residues is not apparently required for methyltransferase activity; however, the packing interactions of helix 6 with others in the carboxylterminal lobe may be important for stabilizing the conformation of the residues that are recognized and bound by the methyltransferase. The substitution of helix 2 apparently perturbs these interactions, a result that was not anticipated based on the similarity of the two structures. The reason for this defect in CaM H6 is not yet clear.
In contrast, helix 7 shows a much different influence on the methylation of lysine 115. This helix introduces a high density of electrostatic charge on the carboxyl-terminal lobe (21) adjacent to the site of methylation (Fig. 5). Mutagenesis of these various charged groups show that the removal of charges at positions 118 and 120 and to a lesser degree from positions 122, 126, and 127 results in a reduction in the rate of methylation. These findings along with the E114A results discussed above suggest that electrostatic interactions may play a role in the binding of the methyltransferase.
Interestingly, many of the defects associated with the substitutions within helix 7 apparently are more severe in apoCaM compared with Ca 2ϩ -CaM. Additionally, other mutations, such as the L112T substitution within the methylation loop, are only defective in apoCaM. This difference in the recognition of apoCaM and Ca 2ϩ -CaM by the calmodulin methyltransferase is supported by several previous findings. For example, the methylation of apoCaM shows different kinetics and considerably greater sensitivity to conditions of increasing ionic strength than Ca 2ϩ -CaM (7). Conversely, the methylation of Ca 2ϩ -CaM, but not apoCaM, is sensitive to peptides and ligands that bind to the hydrophobic cleft (7-9). The inability of the cam2 mutant of Paramecium to be methylated normally in vivo (26) was found to be due to an inability to selectively recognize the apo form of calmodulin (14). Thus, although both Ca 2ϩ -and apoCaM are trimethylated by the calmodulin methyltransferase, they interact with the enzyme in a distinct fashion.
Based on these previous studies and the present mutagenesis work, we propose a model for the interaction of the methyltransferase with the two forms of calmodulin (Fig. 5). ApoCaM exists predominantly in a closed conformation consisting of the four ␣-helices of the EF-hand pair packed in an antiparallel (128°-137°) orientation relative to one another (23,25). This results in fewer exposed hydrophobic residues and a high density of surface charge residues (Fig. 5). Based on the charge substitutions, electrostatic interactions between the methyltransferase and the charged residues of helix 7 and the methylation loop of apoCaM may help contribute to binding/ orientation of the calmodulin substrate. This supports previous findings that the interaction of apoCaM with the calmodulin methyltransferase (7) is sensitive to ion concentrations. Interestingly, the interaction of apoCaM with other target proteins shows a similar sensitivity (27)(28)(29)(30).
The binding of calcium induces a conformational change in the lobe, resulting in a shift of the EF-hand interhelical angles to an almost perpendicular state (86°-101°). Although the relative conformation of helix 6-methylation loop helix 7 (residues 106 -126) undergoes a small change between apoCaM and Ca 2ϩ -CaM (root mean square deviation is 2.6 Å), the major change between the two structures is the surface exposure of a pronounced hydrophobic pocket adjacent to the site of methylation (Fig. 5). This surface might provide additional interactions with the methyltransferase. This could explain why reagents such as drugs and peptides, which interact selectively with the hydrophobic cleft, block the binding of the methyltransferase to Ca 2ϩ -CaM but not to apoCaM (7). Furthermore, this could also help explain why several of the charge-to-alanine mutations have less severe effects on calmodulin methylation in the presence of calcium.