Modification of Calmodulin on Lys-75 by Carbamoylating Nitrosoureas”

This paper describes characterization of the reaction of calmodulin with a series of nitrosoureas which are capable of releasing amine-reactive isocyanates of varying hydrophobic character. The site of calcium- dependent carbamoylation on calmodulin by the antineoplastic agent l-(2-chloroethyl)-3-(4-methylcyclo- hexyl)-l-nitrosourea (methyl CCNU) was determined to be Lys-75 as demonstrated using [ring-14C]methyl CCNU and sequence analysis of the sole labeled peptide obtained from tryptic digestion of reversed-phase high pressure liquid chromatography (HPLC)-purified radiolabeled calmodulin. CCNU, the 4-desmethylcyclo- hexyl derivative of methyl CCNU, and its reactive hydrolysis product, cyclohexyl isocyanate, were also determined to modify calmodulin in a similar manner and at the same site, as demonstrated by specific block- ade of modification by the calmodulin antagonist calmidazolium. Nitrosoureas which release the less hydrophobic 4-hydroxy- and 4-carboxycyclohexyl iso- cyanates are unable to modify calmodulin at %-fold higher concentrations than those required for modifi- cation with methyl CCNU, CCNU, or cyclohexyl isocyanate. With this monomodified Lys-76 derivative, purified to homogeneity by HPLC, differential effects of modi- fication on the activation of bovine amino acid analysis (26). The resulting protein was desalted by gel filtration and concentrated by lyophilization prior to assay. The ATPase was assayed for stimulation by calmodulin and its derivative following reconstitution into phospholipid vesicles using a coupled enzyme determination of released inorganic phosphate according to the procedure of Niggli and co-workers (23). Standard curves for control levels of activation were obtained using HPLC-purified native calmodulin. Studies with Calmidazolium-The ability of calmidazolium, a cal-cium-dependent calmodulin antagonist (30), to block the modifica- tions with methyl CCNU, cyclohexyl isocyanate, and CCNU was assayed by adding calmidazolium dissolved in a minimum volume of absolute ethanol (<5% of total volume) to the standard reaction mixtures containing 1 pM calmodulin and 20 pM carbamoylating agent. Extent of modification was determined by automatic integration of native and modified calmodulin peaks eluted from the C-3 column using a Hewlett-Packard 1090 liquid chromatograph and integrator with effluents monitored at 230 nm. Parallel assays of dose-dependent calmidazolium inhibition of phosphodiesterase activation by calmodulin were performed by adding the compound dis- solved in absolute ethanol (<5% of total volume) to the standard phosphodiesterase assay and using amounts of calmodulin sufficient to provide submaximal levels of activation.

of these enzymes are mediated. In the presence of micromolar calcium concentrations, calmodulin undergoes conformational changes associated with calcium binding that result in increased a-helicity and exposure of several hydrophobic sites (3). The recent determination of the three-dimensional structure of the calcium-replete protein as a highly asymmetric molecule with two calcium-binding structures at either end separated by a long central helix brings new insight into possible mechanisms of target enzyme activation (4).
The activation of calmodulin-regulated enzymes is inhibited by a wide variety of compounds including phenothiazines (5), naphthalene sulfonamides (6), antineoplastic agents (7), and hydrophobic peptides (8). The nitrosoureas, a class of chemotherapeutic agents that release alkylating and aminedirected carbamoylating reagents (for review, see Ref. 9), inhibit the ability of calmodulin to stimulate bovine brain 3',5'-cyclic nucleotide phosphodiesterase (phosphodiesterase) in a manner related to the carbamoylating activity of the nitrosourea (10).
Differential trace labeling of calmodulin with acetic anhydride has demonstrated that Lys-75 undergoes a 25-fold increase in reactivity as the protein binds calcium (11). The reactive nature of Lys-75 and the propensity for carbamoylating nitrosoureas to act almost exclusively on lysyl residues in proteins (12) led us to evaluate and characterize the usefulness of these antineoplastic agents as affinity-directed calmodulin probes.
Other affinity-directed calmodulin probes have been synthesized by generating reactive phenothiazine derivatives. Norchlorpromazine isothiocyanate will modify calmodulin with a 1:1 stoichiometry in a calcium-dependent manner (13). The site of modification is believed to be Lys-75, as determined by amino acid analysis of tryptic peptides (14). The monoadduct, while incapable of activating phosphodiesterase, can act as a competitive inhibitor of native calmodulin for activation of the enzyme, much in the same manner as the amino-terminal half of calmodulin (15). The modification is blocked by trifluoperazine. Calmodulin has also been modified in a 1:1 stoichiometry with POS-TP, a reactive trifluoperazine derivative (16). The site of modification has been proposed to be Lys-148, as demonstrated by amino acid analysis of a labeled tryptic peptide (17). The calmodulin-mediated activation of phosphodiesterase and the ATPase are unaffected by modification at this site, whereas plant NAD kinase is not activated by this derivative (17).
Recently, the role of lysyl residues in the binding of calmodulin to two target enzymes has been studied by acetylation. Differential trace labeling showed substantial protection of Lys-75 from acetylation in the calmodulin-MLCK complex (18). Manalan and Klee (19) reported similar protection of this residue in the presence of calcineurin. However, Winkler and co-workers (20) have presented data suggesting that Lys-77, as well as Lys-21 and -148, in calmodulin show substantial protection from acetylation in the presence of calcineurin phosphatase. Thus, the results of trace labeling protection studies with calcineurin are controversial. Affinity selection studies with preacetylated calmodulins indicate that acetylation of any one lysyl residue reduces the affinity of calmodulin for calcineurin, although maximal activation is still possible (19).
Clearly, for covalent modification studies to provide unambiguous information about the chemistry of lysine modification on calmodulin, the site(s) of modification must be determined unequivocally. We report herein the specific formation and characterization of a Lys-75 monoadduct on calmodulin which has been obtained through reaction of the protein with methyl CCNU.' This determination was based on sequence analysis of the sole labeled peptide obtained from a tryptic digest of calmodulin modified with [rir~g-'~C]rnethyl CCNU. Through use of a series of carbamoylating nitrosoureas which release reactive isocyanates of differing hydrophobic character, we have further characterized the role of simple hydrophobic character of the modifying agent in directing calcium-dependent modification of calmodulin.

Materials
Calmodulin was isolated from bovine testes according to the procedure of Jamieson and Vanaman (21). Calmodulin-deficient 3',5'cyclic nucleotide phosphodiesterase was partially purified from bovine brain according to the procedure of Klee and Krinks (22). Human erythrocyte CaZf,Mg2+-ATPase was isolated from outdated human blood according to the procedure of Niggli and co-workers (23). Nitrosoureas were obtained from the Drug Synthesis and Chemistry Branch of the National Cancer Institute. [ring-*4C]Methyl CCNU (specific activity, 28 mCi/mmol) was obtained from Dr. Robert R. Engle of the Chemical Resources section of the National Cancer Institute. This compound was dissolved in absolute ethanol and checked for purity by thin-layer chromatography. Cyclohexyl isocyanate was purchased from Aldrich and dissolved in acetone prior to use. CCNU, 4-hydroxy-CCNU, and 4-carboxy-CCNU were obtained from the National Cancer Institute. Calmidazolium was purchased from Boehringer Mannheim, and t-l-chloro-3-(4-tosylamido)-4-phenyl-2-butanone-trypsin and soybean trypsin inhibitor were purchased from Worthington. HPLC-grade acetonitrile was from Burdick and Jackson. Aqueous buffers, prepared with Milli-& water, were filtered through 0.2-pm Millipore filters prior to use. HPLC columns (DEAE-TSK Spherogel2SW-4 X 250 mm and Ultrapore RPSC C-3 4.6 X 75 mm) were purchased from Altex. Except as noted, all other chemicals and reagents were purchased from Sigma.

Methods
Reaction of Nitrosoureas with Calmodulin-The hydrophobic nitrosoureas CCNU and methyl CCNU (see Fig. 1 for structures and chemical names) were dissolved in absolute ethanol immediately prior to use and added to reaction mixtures containing 30 mM sodium borate (pH 8.01, 2 mM calcium chloride, and 1-10 p~ bovine testes calmodulin. Incubation was performed for specified times at 37 "C. The less hydrophobic nitrosoureas 4-hydroxy CCNU and 4-carboxy CCNU ( Fig. 1) were dissolved in 50% ethanol (in deionized water) before addition to the reaction mixtures. Cyclohexyl isocyanate was diluted with acetone and added to similar reaction mixtures. To assess the calcium dependence of the reactions, 2 mM EGTA was used to replace the 2 mM calcium chloride in each reaction mixture.
The time and methyl CCNU concentration dependence of the reaction of methyl CCNU with calmodulin were assessed by precipitating the calmodulin from a standard reaction mixture containing 10 p~ calmodulin with 5% (v/v) trichloroacetic acid at various times  In the presence of hydroxyl ion, the present nitrosourea breaks down to yield chloroethyl diazonium hydroxide ( B ) and a reactive R-substituted organic isocyanate (C). B spontaneously degenerates to release (D), a chloroethyl carbonium ion, which is thought to be responsible for the antitumor activity of the nitrosoureas. The reactive isocyanate (C) can react with primary amines such as the e-amino group of lysyl residues, to give rise to F, a substituted urea, or can react with water in a hydrolysis reaction ( E ) (36).
following initiation of the reaction with 10, 20, and 50 p~ methyl CCNU. The precipitated protein was collected and dissolved in a minimum volume of 200 mM Tris-HCl (pH 8.0) and analyzed as described below. For reactions allowed to run to completion (2 h), extents of modification were quantified by direct analysis of the reaction mixture by HPLC following quenching of excess reagent by reaction with a 10-fold molar excess of lysine. The breakdown of CCNU and its 4-substituted derivatives was followed spectrophotometrically by the method of Forist (24), which measures the disappearance of the nitroso moiety. HPLC Analysis and Purification of Modified Calmodulins-HPLCbased procedures were used not only to purify modified calmodulin but also as an analytical tool to follow the time course and extent of modification. For example, the modification of calmodulin with methyl CCNU was followed by C-3 reversed-phase chromatography. Following reaction of methyl CCNU and calmodulin in the pH 8.0 borate solution described above, native and modified calmodulins were separated on a C-3 reversed-phase column using a Varian 5000 liquid chromatograph. Separation of native and modified calmodulins was achieved using a l%/min gradient increasing from 20 to 30% B with 10 mM dibasic sodium phosphate (pH 6.0), 2 mM EGTA, and 5% acetonitrile as buffer A, and 100% acetonitrile as the mobile phase.
Reversed-phase chromatography was unreliable for detecting and separating adducts formed with other less hydrophobic compounds (25). However, these derivatives were easily separated by anionexchange HPLC as modification of lysine c-amino groups leads to loss of a positive charge in the calmodulin molecule. Therefore, all nitrosoureas and cyclohexyl isocyanate were assayed for the ability to modify calmodulin using both reversed phase on C-3 and DEAE anion-exchange chromatography. For anion-exchange chromatography, gradient elution from 40 to 60% B over 20 min was utilized to separate modified and unmodified calmodulins. Buffer A was 10 mM Tris-HC1 (pH 7.0) and buffer B the same, with 0.8 M NaCl added. Column effluents from both systems were monitored for UV-absorbing material using a Waters 441 UV detector at the wavelengths indicated in each trace. Peaks of UV-absorbing material were collected manually and analyzed as described in the following sections. Aliquots of peak fractions were counted by liquid scintillation spec-trometry in the case of samples modified with radiolabeled compound agent. Relative amounts of native and modified calmodulin were determined by manual integration of peaks in the UV elution profiles or by automated peak integration using an HP 3392A integrator.
Determination of Sites of Modification-The site of modification was determined for the primary derivative resulting from the reaction of methyl CCNU with calmodulin. In a final concentration of 30 p~ methyl CCNU, 1 pCi of [ring-"Clmethyl CCNU was added to a standard reaction mixture containing 10 p~ calmodulin. Following incubation, native and modified calmodulins were separated using reversed-phase HPLC as described above; the purified proteins were desalted into 10 mM ammonium bicarbonate on Sephadex G-50 (2 X 50 cm), concentrated by vacuum centrifugation, and digested with trypsin. Peptides were separated on a Waters Phenyl pBondapak column as described by Klevit and Vanaman (26). The sole radiolabeled peptide obtained from the "C-labeled protein was desalted into 10 mM ammonium bicarbonate on a 2 X 50-cm Sephadex G-25 column, lyophilized, and the amount of peptide recovered quantified by amino acid analysis (26). The labeled peptide was then sequenced using an Applied Biosystems vapor-phase sequenator as described by Hunkapillar and co-workers (27). Aliquots from each cycle were assayed for radioactivity by liquid scintillation spectrometry. Pthderivatives recovered from the sequenator were identified and quantified using a Du Pont Zorbax reversed-phase C-18 column by HPLC and automated peak integration (28).
Functional Characterization of Modified Calmodulin-Native and modified HPLC-purified calmodulins were assayed for the ability to activate phosphodiesterase by the procedure of Wallace et al. (29). In order to remove contaminating unmodified calmodulin, derivatives were repurified on a new C-3 reversed-phase column as described above. For use in these assays, protein calmodulin concentrations were determined by amino acid analysis (26). The resulting protein was desalted by gel filtration and concentrated by lyophilization prior to assay. The ATPase was assayed for stimulation by calmodulin and its derivative following reconstitution into phospholipid vesicles using a coupled enzyme determination of released inorganic phosphate according to the procedure of Niggli and co-workers (23). Standard curves for control levels of activation were obtained using HPLCpurified native calmodulin.
Studies with Calmidazolium-The ability of calmidazolium, a calcium-dependent calmodulin antagonist (30), to block the modifications with methyl CCNU, cyclohexyl isocyanate, and CCNU was assayed by adding calmidazolium dissolved in a minimum volume of absolute ethanol (<5% of total volume) to the standard reaction mixtures containing 1 p M calmodulin and 20 p M carbamoylating agent. Extent of modification was determined by automatic integration of native and modified calmodulin peaks eluted from the C-3 column using a Hewlett-Packard 1090 liquid chromatograph and integrator with effluents monitored at 230 nm. Parallel assays of dose-dependent calmidazolium inhibition of phosphodiesterase activation by calmodulin were performed by adding the compound dissolved in absolute ethanol (<5% of total volume) to the standard phosphodiesterase assay and using amounts of calmodulin sufficient to provide submaximal levels of activation.

RESULTS
Calmodulin Modifications-The reaction of calmodulin with methyl CCNU has been characterized in detail. Fig. 2 shows reversed-phase HPLC analysis of reaction mixtures containing 10 pM calmodulin and 20 pM methyl CCNU incubated for 2 h at pH 8.0 in the presence or absence of Ca". A single major new peak (Peak 2), eluting from C-3 at 27% acetonitrile, was formed in the presence of 2 mM caz+ (lower chromatogram). Only material eluting at 25% acetonitrile, the position of unmodified calmodulin (Peak l), was observed following incubation in 2 mM EGTA (upper chromatogram). Studies performed with 14C-labeled methyl CCNU described below confirmed that Peak 1 was unmodified calmodulin and demonstrated that Peak 2 contained only a single specific calmodulin derivative which will be referred to as "monoadduct" hereafter. The elution of this monoadduct from C-3 at higher acetonitrile concentration than the unmodified protein is indicative of the incorporation of a hydrophobic moiety into the protein.

TIME (min)
FIG. 2. The differential modification of 10 p~ calmodulin by 20 p~ methyl CCNU at pH 8.0 in the presence (lower) and absence (upper) of 2 mM calcium chloride. A calcium-dependent derivative peak (2) is formed in the presence of calcium. Peak 1 represents unmodified calmodulin. HPLC separations were done using reversed-phase C-3 as described under "Methods." The rate of modification is dependent on the methyl CCNU concentration employed in the reaction mixture (Fig. 3). Nearly total modification of 10 p~ calmodulin was obtained with 50 p M methyl CCNU at the 120-min time point with formation of the monoadduct almost exclusively. The modification is also strongly time-dependent (Fig. 3). Measurement of exact rates of modification is difficult with these reagents because of the fact that the amine-reactive isocyanates must first be generated by hydrolysis of the parent compound and they are subsequently destroyed by reaction with hydroxyl ion (Fig. 1).
The calcium-dependent modification of calmodulin by CCNU, cyclohexyl isocyanate, 4-hydroxy CCNU, and 4-carboxy CCNU was assayed by HPLC analysis of reaction mixtures. Cyclohexyl isocyanate and CCNU modify calmodulin in a completely calcium-dependent manner, indistinguishable from that observed for methyl CCNU modification. The less hydrophobic nitrosoureas were unable to generate any adduct formation at any concentration or pH tested. These results are summarized in Table I. These compounds were tested for modification by both reversed-phase C-3 and DEAE anionexchange HPLC. Breakdown of all nitrosoureas was followed spectrophotometrically, and good agreement with published values for CCNU decomposition (24) was obtained for all nitrosoureas.
Determination of Site of Modification-In order to fully characterize the modification of calmodulin by methyl CCNU, 50 nmol of the bovine testes protein was reacted with 30 pM methyl CCNU containing 1 pCi of the [ring-"C]compound (final specific activity, 11,828 cpm/nmol) in a 5.0-ml reaction mixture, as described under "Methods." Separation of reaction products by reversed-phase HPLC on Altex C-3 (Fig. 44) yielded four peaks of UV-absorbing material, three of which also contained radiolabel. The first peak (peak 1) contained no radiolabel and eluted at the position of unmodified calmodulin (25% acetonitrile).
As before, the major peak of UV-absorbing material (Peak 2) eluted at 27% acetonitrile which was also coincident with  gives 50% modification of 10 WM calmodulin.
NM refers to no modification detected.
a large peak of I4C radiolabel. The specific radioactivity of the modified calmodulin present in this peak was determined by amino acid composition analysis and scintillation counting to be 11,400 cpm/nmol, within experimental error of that expected for incorporation of 1 mol of reagent/mol of protein.
Two smaller peaks of radioactive material (peaks 3 and 4), eluting at higher acetonitrile concentrations in the profile shown in Fig. 4A, contained considerably smaller amounts of calmodulin as judged both by the corresponding peaks of UVabsorbing material and by amino acid analysis. Based on their specific radioactivities estimated from the latter (-19,000 4. A, the modification of 10 p~ calmodulin by 30 p~ methyl CCNU with 1 WCi of [ring-14C]methyl CCNU. Peak 1, eluting at 25% acetonitrile, represents unmodified calmodulin. Peak 2 represents the primary derivative (monoadduct), and peaks 3 and 4 are multiply modified calmodulins based on comparisons of radioactivity incorporated with the monoadduct peak (see text). This chromatogram was obtained using a reversed-phase C-3 column. B, a tryptic peptide map of HPLC-purified monoadduct (Fig. 2, peak 2). One labeled peptide eluting at 33 min was recovered. The broken line shows the amount of radioactivity present in each 1-min fraction. cpm/nmol), the calmodulin in these peaks was modified by 2 or more mol of reagent/mol of protein. A total of 60% of the calmodulin applied to the separation shown in Fig. 4A was recovered in peaks 1-4, similar to recoveries routinely obtained for the unmodified protein from Altex C-3 columns.
The major ['4C]methyl CCNU-labeled calmodulin derivative, isolated as peak 2 from multiple separations identical to that in Fig. 4A, was further characterized to demonstrate that it was specifically modified. Fig. 4B shows the elution profile obtained for a trypsin digest of this radiolabeled peak 2 material separated by reversed-phase HPLC. A single major radioactive peptide was obtained eluting from C-3 at 33 min into gradient elution performed under conditions where the intact protein eluted at 46 min. Greater than 90% of the applied radioactivity was recovered in the fraction containing this peak. The amino acid composition of this material was in excellent agreement, except for the presence of glycine, with that expected for the peptide spanning residues Lys-75 through Arg-86 of bovine calmodulin, assuming that the detection of only 1 mol of lysine/mol of peptide resulted from complete modification of one of the two lysyl residues in this sequence (not shown).
In order to unequivocally identify the site of modification, a sample of this peptide was subjected to automated Edman degradation on an Applied Biosystems Vapor-Phase Sequencer. The 14C radiolabel and Pth-derivative content obtained at each cycle were determined and are shown in Table 11. Approximately 80% of the radiolabel recovered was obtained in the fraction from the first cycle where a Pth-lysine having a different elution time than unmodified Pth-lysine was recovered. All the radioactivity recovered in this cycle was accounted for in this altered Pth-lysine peak. Cycle 2 contained the remainder of the radiolabel as well as Pth-methionine which was shown to be devoid of 14C by scintillation counting of the Pth-methionine peak collected from this separation. The remainder of the cycles contained no radiolabel and gave those Pth-derivatives on HPLC analysis expected for the sequence Lys-75 through Arg-86.
It should be noted that the modified Pth-lysine present in the first cycle was not identified with the Du Pont Zorbax C-18 column used initially to resolve Pth-derivatives. An identical labeled peptide was isolated from a subsequent trypsin digest of ['4C]methyl CCNU-labeled calmodulin and the Pthderivatives obtained from vapor-phase sequencing resolved isocratically on a Du Pont Bioseries Pth column (31). A peak  of UV-absorbing material, eluting earlier than Pth-lysine and containing >90% of the radioactivity present, was found in the first cycle (data not shown).

Lys-Met-L y s -Asp-T h r -Asp-Thr-Glu-G l u -Glu-Ile-
Effect of Calmidazolium on Modifications-The calciumdependent modification of calmodulin by 20 pM methyl CCNU was almost completely inhibited by addition of 1 pM calmidazolium to the reaction mixture. The extent of modification produced following 2 h of incubation when various concentrations of calmidazolium were added is shown in Fig.  5 . A similar curve was observed for the calcium-dependent modification of 1 ~L M calmodulin by 20 p~ cyclohexyl isocyanate (closed triangles), with the same concentration required for 50% inhibition in both curves. Calcium-dependent CCNU modification was inhibited over an identical range of calmidazolium concentrations (not shown). The effect of a range of calmidazolium concentrations on the ability of 36 ng of calmodulin to activate phosphodiesterase is shown by the closed squares. Again, 1 pM calmidazolium was sufficient to inhibit the majority of the activation.
Effects of Modification on Function-To determine the effect of the Lys-75 modification on the functional integrity of calmodulin, calmodulin derivatives were repurified on a new C-3 column for use in phosphodiesterase and ATPase activation assays. Derivatives prepared in this way contained no unmodified calmodulin as determined by HPLC. The effect of Lys-75 modification on the activation of phosphodiesterase is shown in Fig. 6A. The closed circles represent the activation by native HPLC-purified calmodulin and the closed triangles the stimulation by Lys-75 monoadduct. Although maximal activation is still reached, 7-fold higher amounts are required. In Fig. 6B, the activation of the ATPase by native (closed circles) and monoadduct (open squares) calmodulin is shown. There was no discernible effect of Lys-75 modification on the ability of calmodulin to activate the ATPase.

DISCUSSION
The formation of specific calmodulin derivatives wherein the site(s) of modification have been unambiguously determined continues to be important in relating calmodulin structures to both functional and chemical properties. Through use of [ring-14C]methyl CCNU, the calcium-dependent modifica-tion of calmodulin has been shown to occur on Lys-75. We have been able to trace the modification from incorporation of the label into the intact protein to a single labeled peptide and ultimately to a modified Pth-lysine derivative which behaves chromatographically different than unmodified Pthlysine. Hydrolysates of other proteins (albumin, histones) treated with CCNU have yielded cyclohexylcarbamoyllysine as a modification product (32). It therefore is likely that 4methylcyclohexylcarbamoyllysine is the structure of the Lys-75 monoadduct on calmodulin obtained through reaction with methyl CCNU. The altered retention time of the resulting Pth-4-methylcyclohexylcarbamoyllysine (25) is also anticipated due to the prevention of incorporation of an additional Pth moiety on the t-amino group.
Both CCNU and cyclohexyl isocyanate, the carbamoylating moiety of CCNU, are capable of specific calcium-dependent modification of calmodulin. Modification by the above compounds and methyl CCNU is inhibited to the same extent by identical concentrations of the calmodulin antagonist calmidazolium, indicating that modification for all of these compounds most likely occurs at Lys-75. Additionally, these calmodulin derivatives all activate phosphodiesterase equally.
That the modification occur at Lys-75 is not particularly surprising, since this residue has been shown to undergo a 25fold increase in reactivity toward acetic anhydride when calmodulin binds calcium (11). Hydrophobic reactive phenothiazine derivatives such as norchlorpromazine isothiocyanate (13) and a fluorenyl-based spin labeling reagent (33) are also likely to modify calmodulin on Lys-75, although both also modify a second lysyl residue (probably Lys-148) at a substantial rate. It is somewhat surprising that another reactive phenothiazine derivative, POS-TP, produces calcium-dependent modification of calmodulin preferentially on Lys-148 (17). The selectivity of this reagent for Lys-148 may reflect the orientation of the reactive moiety of the POS-TP in that the reagent may still bind to the hydrophobic pocket near Lys-75, but the reactive moiety is inaccessible to the t-amino group of this residue.
Interestingly, Lys-77, located in an environment expected to be similar to that occupied by Lys-75, is one of the least reactive residues on calmodulin (11). We have as yet been unable to demonstrate any modification of this residue. Its unreactivity may be due to the formation of a salt bridge with Asp-80. These structures are anticipated to participate in the stabilization of the central helix (34). The hydrophobicity of the modifying agent may be important in directing modification at Lys-75 in that the side chain of this residue appears to be oriented toward the proposed hydrophobic pocket formed by a-helices I1 and I1 in the three-dimensional structure of the calmodulin molecule (4). This may explain the failure of the less hydrophobic nitrosoureas, such as the 4hydroxy and 4-carboxy CCNU derivatives, to modify calmodulin at Lys-75. The failure of these compounds to produce any modification is not understood at this time. That calmidazolium concentrations equal to the calmodulin concentration employed (1 p M ) were able to almost completely inhibit modification at Lys-75 may indicate that the hydrophobic modifying agents such as methyl CCNU must first be able to bind to calmodulin prior to reaction.
Since calmidazolium blocks the specific modification of calmodulin at Lys-75, as well as calmodulin-stimulated phosphodiesterase activation over similar concentration ranges, it seems likely that this region of the molecule plays important roles in mediating both the binding of standard hydrophobic calmodulin inhibitors to the hydrophobic site located in the NHp-terminal half of the molecule and the activation of phosphodiesterase. Modification at Lys-148 by POS-TP is also inhibited by trifluoperazine (17). Previous studies have shown that each half of the calmodulin molecule contains 1 phenothiazine binding site (35). Studies with N-hydroxysuccinimidobiotin have shown that calcium-dependent modification at Lys-94 is not blocked by calmidazolium. ' The modification of Lys-75 by methyl CCNU results in a derivative that is still able to stimulate phosphodiesterase maximally, but much greater amounts are required. Since the charge on this t-amino group is destroyed by this reaction, it is clearly not essential for maximal stimulation of phosphodiesterase. Due to the effect observed, however, Lys-75 and/ or the structures adjacent to it are involved in this interaction. Larger aromatic calmodulin derivatives with substituents on Lys-75 (13), such as norchlorpromazine isothiocyanate, are unable to activate phosphodiesterase at all. Based on the three-dimensional structure of calmodulin, Lys-75 is anticipated to reside at the beginning of helix IV. The side chain of this residue has been observed to point directly into the hydrophobic "pocket" formed by helices I1 and I11 (4). It is attractive to speculate that the charge on this residue is important in orienting the formation of the phosphodiesterase-calmodulin complex. The final interaction, which dictates the "tightness" of the binding, would be expected to involve both the central helix and the hydrophobic pocket. The central helix (IV) may function by conferring a-helical structure on the calmodulin-binding region of phosphodiesterase, as has been observed for a synthetic peptide representing the calmodulin-binding domain of myosin light chain kinase (37). The hydrophobic pocket may serve as an anchor by binding aromatic residues such as tryptophan which have been observed in calmodulin binding domains of several dependent enzymes (e.g. MLCK (38) and muscle phosphofructokinase (39)) as well as small peptide antagonists (e.g. melittin (40) and mastoparan (41,42)).
The size and possibly the aromaticity of the adduct on Lys-75 may be crucial determinants in the ability of these derivatives to stimulate phosphodiesterase. These effects may be related to the ability of these derivatives to substitute for binding of target enzymes to the hydrophobic pocket. By reaction of calmodulin with a phenothiazine-based spin-labeling reagent Jackson and Puett (33) have obtained a Lys-75 and -148 dimodified calmodulin. That treatment of this calmodulin with sodium hydroxide, which removes the phenothiazine portion of this adduct, restores some of the ability of this calmodulin to activate phosphodiesterase supports the above contention. We also have observed this phenomenon when an HPLC-purified calmodulin adduct, believed to be modified at Lys-75 by reaction of calmodulin with a cleavable heterobifunctional arylazido compound, is treated with hydroxylamine to remove the arylazide portion of the derivative (25). Therefore, the resulting smaller adducts, while still lacking a positive charge on Lys-75, can have a free hydrophobic pocket, which enables maximal activation.
Acetylation by acetic anhydride to a stoichiometry of 6.6 mol of ['4C]acetyllysine/mo1 of calmodulin inhibits the ability of calmodulin to stimulate phosphodiesterase 7-fold (18). This extent of inhibition is similar to that observed with methyl CCNU-modified calmodulin. Both derivatives were able to stimulate the enzyme maximally if sufficient amounts were employed. The fact that studies of acetylated derivatives were performed with unpurified mixtures of products presumably acetylated to varying extents precludes precise interpretation. Therefore, while modification of Lys-75 with reagents that destroy the charge on the e-amino group may be expected to D. M. Mann and T. C. Vanaman, manuscript in preparation.
reduce the affinity of calmodulin for phosphodiesterase, they do not prevent calmodulin from activating the enzyme maximally.
A model based on a computer-predicted structure of calmodulin has postulated electrostatic interactions and sequestering of hydrophobic residues to be the primary driving forces of calmodulin-basic amphiphilic peptide interactions (43). As discussed above, the data discussed herein are not inconsistent with this model. The modification of charge-containing lysyl residues, while reducing the apparent initial affinity for interaction, does not prevent maximal stimulation from occurring if large amounts of calmodulin are employed, thus indicating that the electrostatic binding mediated by these residues, while not of primary import in the two types of calmodulintarget enzyme interactions described, still plays some role. As the derivative is fully able to activate the ATPase, it is unlikely that this modification perturbs the overall structure of calmodulin.
The calmodulin-ATPase interaction is clearly mediated by different structures on calmodulin than the phosphodiesterase-calmodulin interaction. The modification of Lys-75 did not effect the activation of the ATPase at all, thus indicating that this region is not involved in the activation of this enzyme. On the basis of the limited ability of trypsin-generated fragments of calmodulin encompassing residues 78-148 and 1-106 to stimulate the ATPase (44), calcium-binding domain I11 would seem to be of primary importance, although the necessary structures within this domain remain unclear.
The ATPase clearly binds to different structures within calmodulin than enzymes such as phosphodiesterase, whose activation by calmodulin is strongly inhibited by synthetic MLCK calmodulin-binding peptides (37). In this regard, we have recently identified what appears to be an inhibitory domain of the ATPase which bears a striking homology to domain I11 of calmodulin (45).
Certainly specific lysine modification on calmodulin, when used to produce specifically modified well characterized derivatives can be a valuable tool for understanding structures within calmodulin necessary for target enzyme binding and activation. As different structures within calmodulin are utilized in these interactions with different target enzymes, the generation of other specifically modified derivatives will be invaluable in determining the nature of these pleiotropic interactions.